Skip to main content

Full text of "Heterosis : a record of researches directed toward explaining and utilizing the vigor of hybrids"

See other formats


Heterosis 






HETEROSIS 



A record of researches directed toward explaining 
and utilizing the vigor of hybrids 



Edited by 

JOHN W. GOWEN 

Professor of Genetics 
Iowa Stafe College 




I tc l» 



IOWA STATE COLLEGE PRESS 

AMES • IOWA 



Copyrlghi 1952 by The Iowa Sfate College Press 

All rights reserved. Composed and printed in fhe 

United States of America 




Preface 



Heterosis grew out of a desire on the part of Iowa State College to gather to- 
gether research workers from marginal fields of science, each with something 
to contribute to a discussion of a central problem of major national interest. 
The problem of heterosis, as synonymous in large part with that of hybrid 
vigor, formed a natural theme for discussion. As the reader will note, many 
fields of science have contributed or stand to make significant contributions 
to the subject. Major steps in the advance have led to divergent views which 
may be rectified only through joint discussions followed by further research. 
The conference of students of this problem was held June 15 to July 20, 1950. 

In furnishing the opportunity for these discussions by active research 
workers in the field, Iowa State College hoped: to facilitate summarization 
and clarification of the accumulated data on the subject, to encourage formu- 
lation and interpretation of the observations in the light of present day bio- 
logical information, to stimulate further advances in the controlled success- 
ful utilization and understanding of the biological processes behind the phe- 
nomenon of heterosis, and to increase the service rendered by this discovery 
in expanding world food supply. 

Iowa has a direct, vested interest in heterosis. Today the agricultural 
economy of the state is based upon hybrid corn. The scene portraying a hy- 
bridization block of corn, shown here, is familiar to all who travel within the 
state as well as to those in surrounding regions, for this method of corn 
breeding has been shown to be surprisingly adaptable and useful in producing 
more food per acre over wide areas of the world's agricultural lands. 

Iowa's indebtedness to heterosis, generated through crossing selected and 
repeatedly tested inbred strains, is well known. Few outside the workers in the 
field realize the full magnitude of this debt. 

With the progressive introduction of hybrid corn in 1936 there came a 
steady increase in corn yields over both the former yields and over the 
yields of other agricultural crops, as that of tame hay, which were not sub- 
ject to this genetic method of yield improvement. It seems likely that in no 
other period of like years has there been such an increase in food produced 
over so many acres of land. The return from hybrid corn has been phenome- 
nal, but it is now evidently approaching an asymptotic value. It behooves us 
to find out as much as possible about the techniques and methods which 



VI 



PREFACE 



made these advances possible. Even more we should determine what is going 
on within the breeding and physiological systems through which heterosis 
finds expression, if further increases in yields are to be obtained or better 
systems of breeding are to be developed. 

Toward this end the conference topics were arranged under four major 




j^'u'X ''■*Je.Ai 







mi 



'.»*.. 




Controlled heterosis in the making through pollinations and fertilizations of selectively 
purified genetic strains of corn (maize). (From G. F. Sprague.) 



lA 

z a: 



E z 
o 






O V> -J 

Z -10- 



1 1, 



4 O <t 



K - Z 

u >- e 



> * . 



100 






















/ 


'^' 


^ 
















90- 






















/ 






































/ 






PERCENTAGE OF CORN ACREAGE 
PLANTED WITH HYBRID SEED 




80 


/ 
























/ 






















70 


















/ 


' 






































1 
























60 


















1 






































i 


1 






/ 


\ 


-. 


s. 


/ 


\ 


/ 


\ 




50 




















f 

-7] 


/ 




ACRE YIELD 
OF CORN 




r\ 


/ 


\ 


/ 


\ 


^ 








40 
30 
20 




/' 




\ 






1 














\ 


/ 






w 


/ 


N. 


\ 


1 


\ 


/ 


I 


\ 






/ 
/ 
/ 




--' 


■ -- 


^-. 


_.\ 


/ 


/ 


.-' 


\ 

\ 
\ 

> 


/ 

/ 
/ 


\ 




1 
1 
1 


\ 








— "" " 


"'^ ^. 


1 




ACRE YIELD OF 
TAME HAY / 


/^- 


y' 














\ 


1 




























10 












/ 






































/ 


/ 






































_^ — 


— -^ 



































1930 



1935 



1940 
YEARS 



1945 



1950 



Trends in acre production of maize before and after heterosis was in use and tame hay 
over which there has been no such breeding control 1930-1950. (From G. F. Sprague.) 



PREFACE vii 

headings. The early history and development of the heterosis concepts and 
the cy to logical aspects of the problem occupied the first week. The contribu- 
tions of physiology, evolution, and specific gene or cytoplasmic effects to the 
vigor observed in hybrids were dealt with the second week. The third week's 
meetings covered postulated gene interactions, as dominance, recombination, 
and other possible gene effects. During the fourth week breeding systems and 
methods of utilizing and evaluating heterosis effects were considered. In the 
final week the students considered the problems that lie ahead and recent 
methods of meeting them. 

At each daily conference the speaker of the day presented a formal morn- 
ing lecture covering his subject. In the afternoon, he led a conference session 
on the subject of the morning lecture. At this time, all present had an oppor- 
tunity to participate. 

Accompanying, and as a supplement to the Heterosis Conference, a 
Methods Workshop was held from July 3 to July 13. The Workshop was de- 
voted to recent techniques for evaluating the kinds of data which occur fre- 
quently in animal breeding experiments. Workshop meetings were organized 
by Professor R. E. Comstock of North Carolina State College and Professor 
Jay L. Lush of Iowa State College. 

The meetings were led by men from several institutions besides Iowa 
State College. Professors Oscar Kempthorne, Jay L. Lush, C. R. Henderson, 
G. E. Dickerson, L. N. Hazel, F. H. Hull, A. E. Bell, A. M. Button, J. Bruce 
Griffing, C. C. Cockerham, F. H. W. Morley, R. M. Koch, and A. L. Rae 
contributed much to this phase of the program. It is with regret that it is 
impossible to present the meat of the methods presented and developed in 
the Workshop and the afternoon discussions. To many, this material con- 
tributed much to the merit of the conference and the use to which the results 
were put later. 

In the field of worth-while living, as well as to see heterosis in operation, 
conferees were guests, on various weekends, of three nearby companies putting 
heterosis to the practical test of commercial seed stock production in crops 
and live stock — the Ames Incross Company, the Farmers Hybrid Corn 
Company, and the Pioneer Hi-Bred Corn Company. 

Finally, the organization of the conference was the product of the joint 
effort of the genetic group of Iowa State College. This group transcends all 
departmental lines having as the common interest what goes on in inher- 
itance. They were Jay L. Lush, G. F. Sprague, Oscar Kempthorne, S. S. 
Chase, Janice Stadler, L. N. Hazel, A. W. Nordskog, Iver Johnson, W. A. 
Craft, J. Bruce Griffing, and John W. Gowen. 

In last analysis it was the interest of the audience and their participations 
in the discussions that made the Conference worth while. The papers cover- 
ing material presented by the leaders of these discussions follow. 



Table of Contents 



1. Early Ideas on Inbreeding and Crossbreeding .... conway zirkle 1 

2. Beginnings of the Heterosis Concept . . . George Harrison shull 14 

3. Development of the Heterosis Concept H. k. hayes 49 

4. Preferential Segregation in Maize M. M. rhoades 66 

5. Inbreeding and Crossbreeding in Seed Development . . . r. a. brink 81 

6. Physiology of Gene Action in Hybrids w. Gordon whaley 98 

7. Hybrid Nutritional Requirements william j. robbins 114 

8. Origin of Corn Belt Maize and Its Genetic Significance 

EDGAR ANDERSON and WILLIAM L. BROWN 124 

9. Heterosis in Population Genetics . Adrian o a. buzzati-tra verso 149 

10. Fi.xing Transgressive Vigor in Nicotiana Rustica . . harold h. smith 161 

11. Hybridization in the Evolution of Maize . . . paul c. mangelsdorf 175 

12. Biochemical Models of Heterosis in Neurospora . . sterling emerson 199 

13. Nature and Origin of Heterosis th. dobzhansky 218 

14. Plasmagenes and Chromogenes in Heterosis . . . donald f. jones 224 

15. Specificity of Gene Effects m. r. irwin 236 

16. Genetics and Cytology of Saccharomyces .... carl c. lindegren 256 

17. Genetic Implications of Mutations in S. Typhiniurium . . h. h. plough 267 

18. Dominance and Overdominance james f. crow 282 

19. Gene Recombination and Heterosis leroy powers 298 

20. Gene Interaction in Heterosis A. j. mangelsdorf 320 

21. Inbred Lines for Heterosis Tests? Gordon e. dickerson 3vS0 

22. Specific and General Combining Ability .... c. R. Henderson 352 

23. Rotational Crossbreeding and Heterosis L. m. winters 371 

24. Gamete Selection for Specific Combining Ability 

e. l. pinnell, e. h. rinke, and h. k. hayes 378 

25. Monoploids in Maize sherret s. chase 389 

26. Early Testing and Recurrent Selection g. F. sprague 400 

27. Heterosis in a New Population E. j. wellhausen 418 

28. Recurrent Selection and Overdominance fred H. hull 451 

29. Hybrid Vigor in Drosophila John w. gowen 474 

30. Estimation of Average Dominance of Genes 

R. e. comstock and h. f. robinson 494 

Bibliography 517 

Index 537 



IX 



67867 



CONWAY ZIRKLE 

University of Pennsylvania 



OGIC^^ 




Chapter 1 

Early Ideas on Inbreeding 
and Crossbreeding 



In tracing the historical background of a great scientific advance or dis- 
covery, the historian nearly always has the opportunity of showing that the 
scientists who receive the credit for the work are really late-comers to the 
field, and that all the basic principles and facts were known much earlier. 
Finding these earlier records is always something of a pleasure; comparable, 
perhaps, to the pleasure a systematist experiences in extending the range 
of some well known species. 

The historian may be tempted, in consequence, to emphasize these earlier 
contributions a little too strongly and to re-assign the credits for the scientific 
advances which have been made. In the present state of the history of sci- 
ence, it requires only a little searching of the records to discover contributions 
which have been overlooked and which are very pertinent to the advance 
in question. This wealth of data, which accumulates almost automatically, 
seems to deserve emphasis. But great steps forward generally are made 
not by the discovery of new facts, important as they are, or by new ideas, 
brilliant as they may be, but by the organization of existing data in such 
a way that hitherto unperceived relationships are revealed, and by incor- 
porating the pertinent data into the general body of knowledge so that new, 
basic principles emerge. 

For example, even so monumental a work as Darwin's Origin of Species 
contains few facts, observations or even ideas which had not been known 
for a long time. The work of many pre-Darwinians now appears important, 
especially after Darwin's synthesis had shown its significance. Of course, 
this does not belittle Darwin in the slightest. It only illustrates the way 
science grows. 

The emergence of the scientific basis of heterosis or hybrid vigor is no 

1 



2 CONWAY ZIRKLE 

exception. Practically all of its factual background was reported before 
Mendel's great contribution was discovered. Even workable methods for 
utilizing hybrid vigor in crop production were known, but it was not until 
the classic post-Mendelian investigations of Shull, East, and Jones were 
completed, that heterosis took its proper place in genetics. The following 
discussion of the importance of heterosis will be confined to its pre-Men- 
delian background. 

Heterosis can be described as a special instance of the general principles 
involved in inbreeding and outbreeding. To fit it into its proper niche, we will 
trace first the evolution of our ideas on the effects of these two contrasting 
types of mating. Since our earliest breeding records seem limited to those 
of human beings and primitive deities, we will start with the breeding 
records of these two forms. 

Hybrid vigor has been recognized in a great many plants during the 
last two hundred years. We will therefore describe briefly what was known 
of its influence on these plants. Because heterosis has reached its greatest 
development in Zea mays, we will trace briefly the pre-Mendelian genetics 
of this plant, and show how the facts were discovered which have been of 
such great scientific and economic importance. 

The ill effects of too-close inbreeding have been known for a long time. 
Indeed, Charles Darwin (1868) believed that natural selection had pro- 
duced in us an instinct against incest, and was effective in developing this 
instinct because of the greater survival value of the more vigorous offspring 
of exogamous matings. One of his contemporaries, Tylor (1865), noted that 
many savage tribes had tabooed the marriage of near relatives, and he 
assumed that they had done so because they had noticed the ill effects of 
inbreeding. The Greeks looked upon certain marriages between near rela- 
tives as crimes. This has been known almost universally ever since Freud 
popularized the tragedy of King Oedipus. At present, we outlaw close in- 
breeding in man, and our custom is scientifically sound. 

We are apt to be mistaken, however, if we read into the standards of our 
distant preceptors the factual knowledge which we have today. The in- 
tellectual ancestors of European civilization approved of inbreeding and 
actually practiced it on supposedly eugenic grounds. The fact that their 
genetics was unsound and their eugenic notions impractical is irrelevant. 
They had their ideals, they were conscientious and they did their duties. 
The Pharaohs married their own sisters when possible so that their god- 
like blood would not be diluted. Marriage between half brother and sister 
was common in other royal families of the period. Actually, as we shall see, 
the two great pillars of European thought, Hebrew morality and Greek 
philosophy, endorsed inbreeding as a matter-of-course. 

The Hebrews, who derived mankind from a single pair, were compelled 
to assume that the first men born had to marry their sisters — as there were 



EARLY IDEAS ON INBREEDING AND CROSSBREEDING 3 

then no other women on the earth. Indeed Adam and Eve themselves were not 
entirely unrelated. The marrying of a sister was obviously respectable, and 
it seems to have occurred routinely among the Hebrews and their ancestors 
for several thousand years. Abraham's wife, Sarah, was also his sister. At 
times even closer inbreeding took place. Abraham's nephew, Lot, impreg- 
nated his own two daughters. The latter instances occurred, however, under 
exceptional circumstances — and Lot was drunk. But as late as the time of 
King David, brother-sister marriages took place. The imbroglio between 
David's children, Tamar, Ammon, and Absalom, shows that a legal mar- 
riage between half-brother and sister would then have been a routine oc- 
currence. 

The Greeks also could hardly have had scruples against inbreeding, as 
evinced by the pedigrees they invented for their gods. Their theogony shows 
many instances of the closest inbreeding possible for either animals or gods 
in which the sexes are separate. Zeus, the great father of the gods, married 
his sister, Hera. Their parents, Kronos and Rhea, also were brother and 
sister, and were in turn descended from Ouranos and Gaea, again brother 
and sister. Thus the legitimate offspring of Zeus — Hebe, Ares, and Hephaes- 
tus — were the products of three generations of brother-sister mating. 
Moreover, the pedigrees of the Greek heroes show an amount of inbreeding 
comparable to that in our modern stud books for race horses. They were 
all related in one way or another and related to the gods in many ways. A 
single example will be cited. Zeus was the father of Herakles and also his 
great-great-grandfather on his mother's side. Herakles' great-great-grand- 
mother, Danae, who had found such favor in the eyes of Zeus, was herself 
descended from Zeus through two different lines. With immortals, back- 
crossing offered no real problems. 

East and Jones (1919) have pointed out that close inbreeding was com- 
mon among the Athenians even at the height of their civilization. These 
scientists were of the opinion that most of the freemen in Attica were 
rather closely related to each other. Marriage between half brother and 
sister was permitted, and marriage between uncle and niece fairly common. 
A Grecian heiress was nearly always taken as a wife by one of her kinsmen 
so that her property would not be lost to the family. Common as inbreeding 
was during the flowering of Greek culture, it was as nothing compared with 
the inbreeding which occurred in the period after the Trojan War and before 
the true historical period. In this intervening time, Greece was divided into 
innumerable independent political units, many of them minute. One island 
six miles long and two miles wide contained three separate kingdoms. 
Political boundaries as well as bays, mountains, and seas were functional, 
isolating mechanisms; and the Greeks were separated into many small 
breeding units for fifteen to twenty generations. Isolation was never com- 
plete, however, and there were enough wandering heroes to supply some 



4 CONWAY ZIRKLE 

genie migration. There were also some mass migrations and amalgamations 
of different tribes. The general situation was startlingly close to the condi- 
tions which Sewall Wright (1931) describes as the optimum for rapid 
evolution. 

We may be tempted to explain as cause and effect what may be only an 
accidental relationship in time; and, while recognizing that it is far fetched, 
to ascribe the sudden appearance of what Galton called the ablest race in 
history to the ideal conditions for evolution which their ancestors had. We 
would also like to consider, as the necessary preliminary to the hybrid vigor, 
that period of inbreeding which preceded the flowering of Grecian culture. 
This hybrid vigor we would like to recognize as an important factor in the 
production of the great geniuses who flourished in the later, larger city 
states of Greece. 

So much for the classical attitude toward endogamy. It slowly changed, 
and exogamy which had always existed became the exclusive custom. At 
the time of Sophocles, all forms of inbreeding were not considered ethical 
and pleasing to the gods. The sin of Oedipus lay in his having made a for- 
bidden backcross rather than in mere inbreeding which was lawful. We do 
not find any records of degeneracy appearing in his children — indeed his 
daughter Antigone was a model of feminine virtue. It seems that close 
human inbreeding came to an end without its ill effects ever having been 
recognized. 

The Nordics also were unaware of any degeneracy inherent in inbreeding. 
Their great god Wotan included a bit of inbreeding in his plan for creating 
a fearless hero who could save even the gods themselves from their im- 
pending fate. Wotan started the chain reaction by begetting Siegmund and 
Sieglinde, twin brother and sister. The twins were separated in infancy. 
They met again as adults and, recognizing their relationship, had an il- 
legitimate affair — begetting the hero Siegfried. Although Siegfried was not 
exactly an intellectual type, he was certainly not a degenerate — represent- 
ing rather the ideal male of a somewhat primitive culture. 

As the centuries passed, incest was extended to cover brother-sister 
mating, even when the parties involved were unaware of their relationship. 
There is no need to cite here the many examples of the later tragedies based 
upon this plot. It soon became an almost universally accepted standard in 
literature, from epics to novels. The luckless Finnish hero, Kullervo {The 
Kalevala, Rune XXXV), thus brought disaster to his family by seducing his 
sister unknowingly. Defoe's long suffering heroine Moll Flanders (1722) 
had to abandon an apparently successful marriage when she discovered that 
her husband was her brother. On the other hand, as late as 1819, Lord 
Byron defended brother-sister marriage passionately in his drama Cain — 
but this was a scandalous exceptioii to the rule. The marriage of kin nearer 



EARLY IDEAS ON INBREEDING AND CROSSBREEDING 5 

than first cousins had become legally and morally taboo. Perhaps we may 
follow Westermarck in assuming that endogamy became passe, not because 
its biological ill effects were recognized, but because men knew their kins- 
women too well to marry one of them if they could possibly get a wife 
elsewhere. 

It is possible that we have thus far paid too much attention to inbreeding 
and outbreeding in man. Our excuse is that there are almost no other records 
of inbreeding from classical times. There are no plant records, of course, for 
sex in plants was not understood in spite of the general practices of caprifica- 
tion and hand pollination of the date palm. Records of inbreeding and out- 
crossing in domestic animals are almost completely lacking even in the 
copious agricultural literature of the Romans. Aristotle's History of Animals. 
576al5 (Thompson 1910) does state that horses will cover both their mothers 
and their daughters ". . . and, indeed, a troup of horses is only considered 
perfect when such promiscuity of intercourse occurs" — but he seems to 
be almost alone in referring to the subject. Later on in the same book 
(630b30) he cited a happening which we quote. 

The male camel declines intercourse with its mother; if his keeper tries compulsion, he 
evinces disinclination. On one occasion, when intercourse was being declined by the 
young male, the keeper covered over the mother and put the young male to her; but, when 
after the intercourse the wrapping had been removed, though the operation was completed 
and could not be revoked, still by and by he bit his keeper to death. A story goes that the 
king of Scythia had a highly-bred mare, and that all her foals were splendid; that wishing 
to mate the best of the young males with the mother, he had him brought to the stall for 
the purpose; that the young horse declined; that, after the mother's head had been con- 
cealed in a wrapper he, in ignorance, had intercourse; and that, when immediately after- 
wards the wrapper was removed and the head of the mare was rendered visible, the young 
horse ran away and hurled himself down a precipice. 

This behavior of the stallion was considered so remarkable that it was 
described by Aelian, Antigonus, Heirocles, Oppian, Pliny, and Varro. 
Varro confused the tradition and made the horse bite his keeper to death. 

It is fairly safe for us to assume that in both classical and medieval times 
the flocks and herds were greatly inbred. Transportation difficulties would 
have insured inbreeding unless its evil effects were realized, and we have at 
least negative evidence that they were not. Varro, who gave many detailed 
directions for the breeding of all domestic animals, does not even mention 
the question of kinship between sire and dam. We do have an interesting 
literary allusion by Ovid, however, to the routine inbreeding of domestic 
animals in his account of the incest of Myrrha in the tenth book of the Meta- 
morphoses. The affair between Myrrha and her father Cinyras was like that 
of Oedipus and his mother Jocasta. The fates had decreed that Myrrha 
should become the mistress of her father. Torn by her unholy desires she 
debates the matter with her conscience. Her better nature argues (From 
the metrical translation of Brookes More, 1922): 



6 CONWAY ZIRKLE 

But what more could be asked for, by the most 
Depraved? Think of the many sacred ties 
And loved names, you are dragging to the mire; 
The rival of your mother, will you be 
The mistress of your father, and be named 
The sister of your son, and make yourself 
The mother of your brother? 

In Stating the other side of the case Myrrha describes the "natural" in- 
breeding of animals. 

A crime so great — If it indeed is crime. 

I am not sure it is — I have not heard 

That any God or written law condemns 

The union of a parent and his child. 

All animals will mate as they desire — 

A heifer may endure her sire, and who 

Condemns it? And the happy stud is not 

Refused by his mare-daughters: the he-goat 

Consorts unthought-of with the flock of which 

He is the father; and the birds conceive 

Of those from whom they were themselves begot. 

Happy are they who have such privilege! 

Malignant men have given spiteful laws; 

And what is right to Nature is decreed 

Unnatural, by jealous laws of men. 

But it is said there are some tribes today. 
In which the mother marries her own son; 
The daughter takes her father; and by this, 
The love kind nature gives them is increased 
Into a double bond. — Ah wretched me! 

The debate ends as we would expect, and in due course Myrrha is de- 
livered of an infant boy who certainly showed none of the ill effects of the in- 
breeding which produced him. He grew up to be quite an Adonis. In fact 
he was Adonis. 

We can profitably skip to the late eighteenth century before we pursue 
further the matter of inbreeding. This was the period when Bakewell was 
emphasizing the importance of breeding in improving farm animals, when 
the various purebreds were beginning to emerge, and when the efficacy of 
artificial selection was beginning to be understood. 

By the beginning of the nineteenth century, practical attempts to im- 
prove the different breeds of cattle led to intensive inbreeding. A prize bull 
would be bred to his own daughters and granddaughters. At first, the breed- 
ers seemed to believe that a selection of the very best individuals followed 
by intensive inbreeding was the quickest method for improving the stock. 
On theoretical grounds this seemed to be the case, and great advances 
were actually made by this method — but sooner or later something always 



EARLY IDEAS ON INBREEDING AND CROSSBREEDING 7 

happened. The inbred stock seemed to grow sterile, but vigor could be re- 
established by outcrossing. The actual cause of degeneracy in the inbreds 
was not understood until Mendelian inheritance was discovered, but the 
remedial procedures of the practical breeders could hardly have been im- 
proved on. We owe to them the basis of our finest stocks. They inbred to 
add up and concentrate desirable qualities and then crossbred to prevent 
degeneration, then inbred again and crossed again, all the time selecting 
their breeding stocks most carefully. Charles Darwin (1868) described this 
process most accurately and listed the pertinent publications. 

There was a striking divergence in this work between theory and prac- 
tice, which is just as well, as the only theories available at the time were in- 
adequate. Those breeders who held that inbreeding was the suninmm bonum 
did not hesitate to crossbreed when the occasion demanded, and those who 
emphasized the virtues of hybridization inbred whenever inbreeding gave 
them the opportunity of adding up desirable qualities. Darwin, himself, 
stated, "Although free crossing is a danger on the one side which everyone 
can see, too close inbreeding is a hidden danger on the other." We await 
the twentieth century for a real improvement in breeding methods. 

The first plant hybrid was described as such in 1716, and during the next 
forty-five years many descriptions of hybrid plants were published. Some 
attempts were even made to produce new varieties, but in retrospect the 
work seems somewhat dilettante. 

From 1761 to 1766, Josef Gottlieb Koelreuter (1766) published the several 
parts of his well-known classic, and plant hybridization was put upon a 
different and more scientific basis. His investigation of hybridization was 
intensive, systematic, and scientific. He described, among other things, 
hybrid vigor in interspecific crosses in Nicotiana, Dianthus, Verbascum, 
Mirabilis, Datura, and other genera (East and Jones, 1919). He also observed 
floral mechanisms which insured cross pollination and assumed in conse- 
quence that nature had designed plants to benefit from crossbreeding. It is 
worth emphasizing that hybrid vigor in plants was first described by the 
person who first investigated plant hybrids in detail. Koelreuter continued 
to publish papers on plant hybrids until the early nineteenth century. 

Meanwhile other contributions had been made to our knowledge of the 
effects of outcrossing and the mechanism for securing it. In 1793, Sprengel 
depicted the structure of flowers in great and accurate detail, and showed 
how self pollination was generally avoided. In 1799, Thomas Andrew Knight 
described hybrid vigor as a normal consequence of crossing varieties and 
developed from this his principle of anti-inbreeding. Other hybridizers 
noted the exceptional vigor of many of their creations. Indeed, hybrid 
vigor in plants was becoming a commonplace. Among the botanists who 
recorded this vigor were: Mauz (1825), Sageret (1826), BerthoUet (1827), 
Wiegmann (1828), Herbert (1837), and Lecoq (1845). Gartner (1849) was 



8 CONWAY ZIRKLE 

especially struck by the vegetative luxuriance, root development, height, 
number of flowers and hardiness of many of his hybrids. 

Naudin (1865) found hybrid vigor in twenty-four species crosses out of the 
thirty-five which he made within eleven genera. In Datura his results were 
spectacular. In reciprocal crosses between D. Stramonium and D. Tatula 
the offspring were twice the height of the parents. Knowledge of plant 
hybridization was increasing more rapidly at this time than the biologists 
knew, for this was the year in which Mendel's (1865) paper Versuche iiber 
Pflanzen-Hyhriden appeared. Mendel discovered hybrid vigor in his pea 
hybrids and described it as follows: 

The longer of the two parental stems is usually exceeded by the hybrid, a fact which is 
possibly only attributable to the greater luxuriance which appears in all parts of the 
plants when stems of very different lengths are crossed. Thus, for instance, in repeated 
experiments, stems of 1 ft. and 6 ft. in length yielded without exception hybrids which 
varied in length between 6 ft. and 1\ ft. 

We shall cite but one more scientist who wrote on the general subject of 
hybrid vigor in plants. This is Charles Darwin, whose Cross and Self Fertiliza- 
tion in the Vegetable Kingdom appeared in 1876. This was a book of great 
importance and influence, but no attempt will be made here to summarize 
this work of nearly five hundred pages. At the beginning of his concluding 
chapter, Darwin stated: 

The first and most important conclusion which may be drawn from the observations 
given in this volume, is that cross-fertilization is generally beneficial and self-fertihzation 
injurious. 

There is a special reason why this book of Darwin's is of such great 
importance for any historical background to heterosis. Darwin worked 
carefully and quantitatively with many genera, including Zea mays. He 
measured accurately the amount of hybrid vigor he could induce, and he pub- 
lished his data in full. His work stands in the direct ancestral line to the 
twentieth century research on the subject, and the great advances made 
from 1908 to 1919 are based solidly on this work. There are no great gaps 
in the steady progress and no gaps in the literature. 

Zea mays was brought to Europe in 1493 by Columbus on his home- 
ward voyage. This was sometime before the great herbals were written, 
so our first descriptions of the new grain are to be found in the books of the 
travelers and explorers. Later, Indian corn appeared under various names 
in the early herbals, and it was described in detail in the famous Krautehuch 
of Tabernaemontanus, first published in 1588. The author obviously yielded 
to his enthusiasm in devoting five and a half folio pages to corn and includ- 
ing thirteen illustrations in his treatment. He was the first to describe the 
results of xenia — the occurrence of difi'erent colored grains on the same ear — 
but his explanation of the phenomenon has nothing to do with cross pollina- 
tion. He ascribed it directly to God Almighty. 



EARLY IDEAS ON INBREEDING AND CROSSBREEDING 9 

And one sees an especially great and wonderful mystery in these spikes, Gott der Ilerr, 
through the medium of nature which must serve everyone, disports himself and performs 
wonders in his works and so notably in the case of this plant that we must rightly be 
amazed and should learn to know the One True Eternal God even from his creatures alone. 
For some of the spikes of this plant, together with their fruit, are quite white, brown and 
blue intermixed. Thus, some rows are half white, a second series brown and the third blue; 
and some grains, accordingly are mixed with each other and transposed. Again, sometimes 
one, two, or three rows are white, the next rows blue, then again white and after that 
chestnut-brown; that is, they are interchanged on one row and run straight through on 
another. Some spikes and their grains are entirely yellow, others entirely brown, some are 
white, brown, and blue, others violet, white, black, and brown: of these the white and 
blue are prettily sprinkled with small dots, as if they had been artistically colored in this 
way by a painter. Some are red, black, and brown, with sometimes one color next to the 
other, while at other times two, three, even four colors, more or less, are found one next 
to another in this way. 

During the next century and a half, many other descriptions of the 
occurrence of different colored grains on a single ear were published. I have 
found about forty of them and there are doubtless many more. The earliest 
correct interpretation of this phenomenon had to await the eighteenth cen- 
tury and is contained in a letter written by Cotton Mather in 1716. Here 
the different colored grains occurring together on an ear are ascribed to a 
wind-born intermixture of varieties. This letter is the first record we have of 
plant hybridization, and antedates Fairchild's description of a Dianthus 
hybrid by one year. In 1724, Paul Dudley also described hybridization in 
maize, and he was able to eliminate one of the hypotheses which had been 
used to explain the mixture. As a broad ditch of water lay between the mix- 
ing varieties, he could show that the mixed colors were not due to the root- 
lets of different strains fusing underground, a view held at the time by 
many New Englanders, both white and red. 

Hybridization in maize was described again in 1745 by Benjamin Cooke, 
in 1750 by the great Swedish traveler and naturalist, Pehr Kalm, and in 
1751 by William Douglass. By the early nineteenth century, knowledge of 
plant hybrids was widespread. Plant hybridization was becoming a routine 
practice, and there is little doubt that different varieties of maize were 
crossed many times by American farmers who did not record their breeding 
experiments in writing. 

Brown and Anderson (1947, 1948) have recently shown that the modern 
races now grown in the corn belt are derived from both the northern flint 
and the southern dent varieties. Hybridization in corn was easy to perform 
and the results were easy to recognize. The intermixtures of colors were so 
spectacular that they were frequently described, by Gallesio (1806), Burger 
(1808), Sageret (1826), Gartner (1827), and others. 

We detour briefly here into some of the technical aspects of xenia. Double 
fertilization and the mixed nature of the endosperm were discovered by Na- 
waschin in 1899. In 1881, Focke introduced the term xenia but he used it 
to include what we now call melaxenia. Focke collected from the literature 



10 CONWAY ZIRKLE 

many supposed instances where the pollen influenced directly the color 
and form of the flowers, the flavor and shape of the fruits, and the color 
and content of the seeds. How many of these cases were really due to Men- 
delian segregation we will probably never know, since the investigators did 
not know enough to take proper precautions. 

We can, however, divide the history of true xenia into three periods: 
first, when its visible effect was considered a lusus naturae (1588); second, 
when it was known to be caused by foreign pollen (1716); and third, when 
the embryo and endosperm were recognized as two different structures and 
when the influence of the pollen upon the latter was recorded specifically. 
In the paragraph on Zea in the section on xenia, Focke cites the work of 
Vilmorin (1867), Hildebrand (1868), and Kornicke (1876), who described 
the effect of pollen on the endosperm. 

We should note a brief comment on the subject which has been overlooked 
and is earlier than the papers cited by Focke. In 1858, Asa Gray described 
xenia in maize. He reported starchy grains in ears of sweet corn and many 
different kinds and colors of grains on the same ear. He had two explana- 
tions for this occurrence: (1) cross pollination of the previous year and (2) 
direct action of the pollen on the ovules of the present year. It is obvious 
that by ovules he did not mean embryos. This may be the earliest authentic 
recognition of the real problem of xenia. 

In reviewing the nineteenth century records of hybrid vigor in Zea mays, 
we start with those of Charles Darwin (1876). Darwin planned his experi- 
ments most carefully. He crossed and selfed plants from the same stock, and 
raised fifteen plants from each of the two types of seed he had obtained. 
He planted the seed from both the selfed and crossed plants in the same 
pots, from six to ten plants per pot. When the plants were between one and 
two feet in height, he measured them and found that the average height of 
the plants from the selfed seed was 17.57 inches, while that from the crossed 
seed was 20.19 inches or a ratio of 81 to 100. When mature, the two lots 
averaged 61.59 inches and 66.51 inches, respectively, a ratio of 93 to 100. 
In another experiment when the corn was planted in the ground, the ratio 
of the selfed to the crossed was 80 to 100. Darwin called in his cousin, Francis 
Galton, to check his results and Galton judged them to be very good after 
he had studied the curves that he drew. 

The direct connection between Darwin's work and our present hybrid 
corn is shown by Darwin's influence on W^ J. Beal who was the real leader 
in the American research designed to improve maize. Beal reviewed Dar- 
win's book in 1878, and even wrote an article which was little more than a 
paraphrase of what Darwin had published. Beal's own contributions ap- 
peared a little later. 

In 1880, Beal described how he had increased the yield of corn on a large 
scale. Two stocks of the same type of corn which had been grown a hundred 



EARLY IDEAS ON INBREEDING AND CROSSBREEDING 11 

miles apart for a number of years were planted together in alternate rows. 
All of one stock grown in this field was detasseled and thus it could not be 
self fertilized but could produce only hybrid seed. The tasseled stalks of the 
other lot would still be pure bred as there was no foreign pollen to contami- 
nate their ears and they could again serve as a parent to a hybrid. A small 
amount of the first parental stock which furnished the detasseled stalks was 
grown apart for future hybridization. The hybrid seed was planted, and 
produced the main crop. Beal increased his yield by this method by as 
much as 151 exceeds 100. This method and these results, it should be 
emphasized, were published in 1880. 

E. Lewis Sturtevant, the first director of the New York Agricultural 
Experiment Station, made a number of studies of corn hybrids starting in 
1882. His findings are interesting and important but not directly applicable 
to heterosis. Singleton (1935) has called attention to this work and to the 
excellent genetic research which the western corn breeders were carrying on 
at this time — such geneticists as W. A. Kellerman, W. T. Swingle, and 
Willet M. Hays. They anticipated many of Mendel's findings and described 
dominance, the reappearance of recessives (atavisms), and even Mendelian 
ratios such as 1 to 1 and 3 to 1. They were all concerned with practical 
results. Hays (1889), in particular, tried to synthesize superior breeds of 
corn by hybridizing controlled varieties. 

Sanborn (1890) confirmed Beal's results and reported that his own 
hybrid corn yielded in the ratio of 131 to 100 for his inbred. He also fol- 
lowed Real's method of planting his parental stocks in alternate rows and of 
detasseling one of them. He made an additional observation which we know 
now is important: 

It is this outcrossed seed which will give the great crops for the next year. It will be 
noted that I gained twelve bushels i)er acre by using crossed seed. The operation is simple 
and almost costless and will pay one hundred fold for the cost involved. The cross must be 
made every year using nen' seed, the product of the outcross of two pure seed. (Italics C. Z.) 

If our farmers had known of this discovery reported in 1890 they might 
not have tried to use their own hybrid corn as seed. 

Singleton (1941) also called attention to a pre-Mendelian interpretation 
of hybrid vigor by Johnson (1891) which, in the light of our present knowl- 
edge, deserves more than passing notice. We can state it in Johnson's own 

words : 

That crossing commonly gives better offspring than in-and-in breeding is due to the 
fact that in the latter both parents are likely to possess by inheritance the same imperfec- 
tions which are thus intensified in the progeny, while in cross breeding the parents more 
usually have different imperfections, which often, more or less, compensate each other in 
the immediate descendants. 

We come next to a j)ublication of G. W. McClure (1892). This paper is 
deservedly famous, and its many contributions are incorporated into our 
modern genetics literature. Here we shall cite only the observations which 
pertain to heterosis. McClure noted (1) that sterility and deformity often 



12 CONWAY ZIRKLE 

follow selfing, (2) that crossing imparts vigor, (3) that it is impossible 
to tell in advance what varieties will produce corn of increased size when 
crossed, (4) that what appears to be the best ear does not always produce 
the largest crops, and (5) nearly all of the hybrid corn grown a second year is 
smaller than that grown the first year, though most of it is yet larger than 
the average size of the parent varieties. 

McClure also called attention to the fact that our fine varieties of fruits 
have to be propagated vegetatively, and hinted that the deteriorations of 
the seedlings from fruit trees was not unrelated to a like deterioration which 
occurred in the seedlings grown from hybrid corn. 

The year following McClure's publication. Morrow and Gardiner (1893) 
recorded some very pertinent facts they had discovered as a result of their 
field experiments with corn. They reported that, "In every instance the 
yield from the cross is greater than the average from the parent varieties: 
the average increase per acre from the five crosses [they had made] being 
nine and a half bushels." They noted further in a paper published later the 
same year that, "It seems that cross bred corn gives larger yields at least 
for the first and second years after crossing than an average of the parent 
varieties, but how long this greater fruitfulness will last is undetermined." 
Gardiner continued the work and in 1895 published the data he obtained 
by repeating the experiments. He found that in four of six cases the yield 
was greater in the cross, the average being twelve bushels per acre. 

We now come to the great corn breeding research project which was 
undertaken at the University of Illinois in 1895 by Eugene Davenport 
and P. G. Holden. Both of these scientists had been students of Beal and 
were interested in his work on inbreeding and cross breeding maize. We 
are indebted to Professor Holden for an account of this work which he printed 
privately in 1948. This account gives us valuable historic data not to be 
found elsewhere, as most of the University of Illinois records were destroyed 
by fire. 

An intensive series of inbreeding experiments was undertaken by Holden, 
and later on the inbred lines were crossed. Hybrid vigor was noted, and it 
was found in addition that the crosses between different inbred lines differed 
widely in their yield and in their general desirability. The main purpose of 
the experiments was to find out how to use controlled crossing early and 
effectively. After Holden left Illinois in 1900, the project was taken over by 
C. G. Hopkins, a chemist, who was interested in increasing the protein con- 
tent of maize. He hired as his assistant in 1900 a young chemist named 
Edward Murray East, whom we shall hear about later. 

Our account of the background of heterosis is coming to an end as the 
beginning of the twentieth century makes a logical stopping point. We should 
mention, however, the great hybrid vigor discovered by Webber (1900) 
when he crossed a Peruvian corn, Cuzco, with a native variety, Hickory 



EARLY IDEAS ON INBREEDING AND CROSSBREEDING 13 

King. The average height of the parental stocks was 8 feet 3 inches while 
the cross averaged 12 feet 4 inches, an increase of 4 feet 1 inch. 

The next year Webber (1901) called attention to the marked loss of vigor 
in corn from inbreeding. From 100 stalks of selfed corn he obtained 46 
ears weighing 9.33 pounds, while from 100 stalks obtained from crossing 
different seedlings he obtained 82 ears weighing 27.5 pounds. When he 
attempted to "fix" his Cuzco-Hickory King hybrid by selling he got a great 
loss of vigor and almost complete sterility, but when he crossed the different 
seedlings there was little loss of vigor. He concluded that to fix hybrids 
one should not self the plants. 

In 1900, the discovery of Mendel's long-forgotten paper was announced. 
Both Hugo de Vries and C. Correns, two of the three discoverers of Mendel, 
published papers on Zea mays and all future work on Indian Corn was on a 
somewhat different level. 

SUMMARY OF KNOWLEDGE OF HYBRID VIGOR AT 
BEGINNING OF 20th CENTURY 

1. Inbreeding reduces vigor and produces many defective and sterile indi- 
viduals which automatically discard themselves. 

2. Cross breeding greatly increases vigor both in interspecific and inter- < 
varietal hybrids. Crossing two inbred stocks restores the lost vigor and 
frequently produces more vigor than the stocks had originally. 

3. All inbred stocks do not produce the same amount of vigor when crossed. ^ 
Certain crosses are far more effective than others. 

4. The simplest method of hybridizing Zea on a large scale is to plant two 
stocks in alternate rows and to detassel one stock. The hybrid corn grown 
from the detasseled stock produces the great yields. 

5. Hybridization must be secured each generation if the yield is to be kept 
up, although a second generation of open pollinated corn may still be 
better than the original parental stocks. 

6. In inbreeding, both parents are apt to have the same defects which are V 
intensified in the offspring. The cause of hybrid vigor is that in crosses 
the parents usually have different defects which tend to compensate for 
each other in the immediate progeny. 

7. The fact that hybrid vigor in Zea is not permanent but decreases if the 
hybrids are open-pollinated, seems to be related to the fact that fruit 
trees, whose desirable qualities are preserved by vegetative propagation, 
produce seedlings which are inferior. 



GEORGE HARRISON SHULL 

Princefon University 



Chapter 2 

Beginnings of 
the Heterosis Concept 



The heterosis concept was first definitely recognized in the work with hybrid 
corn. Before attempting to define this concept, however, we will take a brief 
look at some of the observations of early workers which indicated the prob- 
able presence of heterosis, and where recognition of heterosis as an important 
biological principle might have been expected. 

The first hybridizer of plants, Dr. J. G. Koelreuter, noted some impres- 
sive examples of excessive luxuriance in his Nicotiana hybrids. These were 
isolated observations which suggested no theory as to why these hybrids 
should exceed their parents in size and general vigor. Koelreuter cannot be 
said to have had a heterosis concept. Probably every conscious producer of 
hybrids since Koelreuter's time has made similar observations of the exces- 
sive vigor of some hybrids over their parents, so that such hybrid vigor has 
ceased to cause surprise. But the general acceptance of hybrid vigor as a nor- 
mal phenomenon did not estabUsh a heterosis concept. It was merely the 
summational effect of oft-repeated experience. 

Thomas Andrew Knight noted the deterioration of some of the old stand- 
ard horticultural varieties, and concluded that such varieties have a natural 
life-span and gradually decline as the result of advancing senility. He saw 
that such decline makes it necessary to develop new varieties which will start 
off with the vigor of youth. Although Knight himself produced many such 
new varieties, some of which were produced by hybridization, it is not ap- 
parent that he thought of hybridization as an agency for the production of 
such new vigor. Although he advanced a theory concerning physiological 
vigor and its decline, he did not recognize the heterosis concept. 

Luther Burbank also produced numerous varieties, often following inten- 

14 



BEGINNINGS OF THE HETEROSIS CONCEPT 15 

tional hybridizations, and it is easy to recognize heterosis as a potent factor 
in the remarkable values displayed by many of these new varieties. But 
while Burbank made great use of hybridizations in his plant breeding work, 
he did not recognize hybridization, as such, as the source of the large size 
and remarkable vigor of his new varieties. For him the role of hybridization, 
aside from the bringing together of desirable qualities possessed separately 
by the two chosen parents, was merely the "breaking of the types." In this 
way the variability in subsequent generations was greatly increased, thus 
enlarging the range of forms from among which to select the most desirable 
for recognition as New Creations. 

There are many other important observations and philosophical considera- 
tions that bear a close relationship to our current understanding of heterosis, 
and which antedated the recognition of heterosis. It would take us too far 
afield, however, to discuss these related observations at length. We can 
make only this passing reference to the highly significant work of Charles 
Darwin in demonstrating that cross-fertilization results, in many cases, in 
increased size, vigor, and productiveness as compared with self-fertilization 
or with other close inbreeding within the same species. 

Darwin did not recognize this increased vigor as identical with hybrid 
vigor, nor specifically attribute it to the differences between the uniting 
gametes. To him it only demonstrated a method which would inevitably 
preserve by natural selection any variation that might occur — whether me- 
chanical or physiological — which would make cross-fertilization more likely 
or even an obligate method of reproduction. With heterosis established as a 
recognized pattern of behavior, or type of explanation, we can now interpret 
Darwin's demonstrated superiority of crossbreds as examples of the occur- 
rence of heterosis. We may go even further and include the whole field of 
sexual reproduction in showing the advantages of heterosis. These result 
from the union of two cells — the egg and the sperm — extremely difTerentiated 
physiologically, and in all dioecious organisms also dififerentiated genetically. 

Let us briefly consider several investigations which foreshadowed the 
procedures now used in growing hybrid corn — for somewhere in the course 
of this work with corn the heterosis principle was first definitely recognized. 

Two techniques are characteristically associated with the work of the 
"hybrid-corn makers." Uncritical commentators have mistakenly considered 
these techniques synonymous with the development of the hybrid-corn pro- 
gram itself. These are (a) cross-pollination by interplanting two different 
lines or varieties, and the detasseling of one of these lines which then sup- 
plies the seed to be planted; and {b) controlled self-pollination. 

In deciding what part these two methods played in the develoi)mcnt of the 
heterosis concept, we must first consider why these methods were used by 
various workers and how their use affected the experimental conclusions. 

Dr. William J. Beal, of Michigan Agricultural College, apparently was 



16 GEORGE HARRISON SHULL 

the first to make extensive use of controlled cross-pollination in the breed- 
ing of corn. Beal was a student of Asa Gray from 1862 to 1865, when the 
latter was in active correspondence with Charles Darwin. Darwin was be- 
ginning the studies on cross- and self-fertilization, which were reported in 
1877 in an important book on the subject. It has been thought that Darwin's 
views on the significance of crossbreeding may have been instrumental in 
inciting and guiding Beal's experiments in the crossing of corn. There seems 
to be no supporting evidence, however, for such a surmise. 

Beal's lectures before various farmers' institutes stressed the importance 
of being able to control the source of the pollen, so that the choice of good 
ears in the breeding program would not be nullified by pollen from barren 
stalks and other plants of inferior yielding capacity. On this point Professor 
Perry Greeley Holden, for several years assistant to Dr. Beal, has stated that 
controlled parentage, not heterosis, was the aim of the corn breeding pro- 
gram at Michigan and at Illinois before 1900. 

In 1895 Holden was invited by Eugene Davenport to become professor of 

agricultural physics at the University of Illinois. Davenport also had served 

for several years as assistant to Dr. Beal at Michigan. Like Holden, he was 

very enthusiastic about the importance of Beal's program, so it was natural 

that Davenport and Holden should agree that corn improvement be a major 

undertaking of Holden's new department at the University of Illinois. On 

initiating this work at the University of Illinois, they learned that Morrow 

and Gardner already had tested Beal's variety crossing at Illinois before they 

got there, and with confirmatory results. Concerning the motivation of all 

this early work, both at Michigan and at Illinois, Holden says: 

1. Hybrid corn [as we know it today] was unknown, not even dreamed of, previous to 
1900. 2. Controlled parentage was the dominant purpose or object of this early corn improve- 
ment work. 

Holden thus makes it clear that while heterosis was at play in all of this early 
work, it was not the result of, nor did it result in, a heterosis concept. 

I refer next to the matter of inbreeding, which some writers have confused 
with the crossing that has brought the benefits of heterosis. Enough selfing 
had been done with corn prior to 1900 to convince all of those who had had 
experience with it that it resulted in notable deterioration. The results of these 
early observations are aptly summed up by Holden in the statement that 
"Inbreeding proved to be disastrous — the enemy of vigor and yield." No- 
where, so far as I have been able to determine, did any of the early inbreed- 
ers discover or conceive of the establishment of permanently viable pure lines 
as even a secondary effect of inbreeding. 

In 1898 A. D. Shamel, then a Junior in the University of Illinois, offered 
himself to Holden as a volunteer assistant without pay. He did so well that 
when Holden severed his connection with the University in 1900, Shamel 
was appointed his successor, and continued in this capacity until 1902. He 



BEGINNINGS OF THE HETEROSIS CONCEPT 17 

then transferred to the United States Department of Agriculture and did no 
further work with corn. In Shamel's final report of his own corn experiments 
(1905), he laid no stress on the positive gains which resulted from cross- 
breeding, but only on the injurious effects of inbreeding. His "frame of ref- 
erence" was the normally vigorous crossbred (open-pollinated) corn, and the 
relation between self-fertilized and cross-fertilized corn was that of something 
subtracted from the crossbred level, not something added to the inbred level. 
The prime objective in a breeding program, he said, "is the prevention of the 
injurious effects of cross-fertilization between nearly related plants or in- 
breeding." In summing up the whole matter he said: 

In general, ... it would seem that the improvement of our crops can be most rapidly 
effected with permanent beneficial results by following the practice of inbreeding, or cross- 
ing, to the degree in which these methods of fertilization are found to exist naturally in the 
kind of plant under consideration. 

This means, for corn, practically no self-fertilization at all, and makes it 
obvious that, at least for Shamel, the heterosis concept had not yet arrived. 

Edward Murray East was associated with the corn work at the University 
of Illinois, off and on, from 1900 to 1905. He worked mainly in the role of ana- 
lytical chemist in connection with the breeding program of C G. Hopkins 
and L. H. Smith. He must have been familiar with the inbreeding work of 
Shamel, if not with that of Holden. It is generally understood that he did 
no self-fertilizing of corn himself, until after he transferred to the Connecti- 
cut Agricultural Experiment Station in 1905. Some of his inbred lines at 
Connecticut may have had the inbreeding work at Illinois back of them, as 
he secured samples of seeds of the Illinois inbreds sent to him by Dr. H. H. 
Love, who assisted him for one year and succeeded him at Illinois. But ac- 
cording to his subsequently published records these older inbred lines did not 
enter to any important extent into his studies in Connecticut. 

As reported in Inbreeding and Outbreeding (East and Jones, pp. 123, 124), 
"The original experiment began with four individual plants obtained from 
seed of a commercial variety grown in Illinois known as Leaming Dent." 
Table III (p. 124) presents the data for these four lines for the successive 
years from 1905 to 1917, and clearly indicates that the selfing was first made 
in 1905. East's work is so adequately presented in this excellent book that it 
seems unnecessary to comment on it further here except to recall that, as 
shown by his own specific statements, my paper on "The composition of a 
field of maize" gave him the viewpoint that made just the difference between 
repeated observations of heterosis and the heterosis concept. In proof of this 
we have not only his letter to me, dated February 12, 1908, in which he says: 
"Since studying your paper, I agree entirely with your conclusion, and won- 
der why I have been so stupid as not to see the fact myself"; but we also 
have the published statements of his views just before and just after the 
publication of my paper. Thus, we read in his Conn. Agr. Exp. Sta. Bull. 158, 



18 GEORGE HARRISON SHULL 

"The relation of certain biological principles to plant breeding," which was 
published in 1907, only a few months before I read my paper in his presence 
in Washington, D.C., what seems like an echo of the final conclusion of 
Shamel, above cited. In this bulletin East urged that "corn breeders should 
discard the idea of forcing improvement along paths where nothing has been 
provided by nature," specifically rejecting a program of isolation of uniform 
types because of a "fear of the dangers of inbreeding," adding that he was 
"not able to give a reason for this belief beyond the common credence of the 
detrimental effects of inbreeding." He returned to this problem of the de- 
terioration due to inbreeding in his Annual Report to the Conn. Agr. Exp. 
Sta. for 1907-8, prepared in 19C8, with my paper before him. In this report 
he says: 

I thought that this deterioration was generally due to the establishment and enhance- 
ment of poor qualities common to the strain. ... A recent paper by Dr. George H. Shull 
("The composition of a field of maize") has given, I believe, the correct interpretation of 
this vexed question. His idea, although clearly and reasonably developed, was supported 
by few data; but as my own experience and experiments of many others are most logically 
interpreted in accordance with his conclusions, I wish here to discuss some corroboratory 
evidence. 

We have thus far failed to recognize the existence of a general heterosis 
concept among plant breeders, prior to the reading of my paper on "The 
composition of a field of maize" in January, 1908, even when they were using 
the methods of inbreeding and controlled crossing in which such a concept 
could have developed. 1 must mention, however, a near approach to such a 
concept from the side of the animal breeders. Before the American Breeders' 
Association, meeting in Columbus, Ohio, 1907, Quintus I. Simpson, an ani- 
mal breeder from Bear Creek Farm, Palmer, Illinois, read a paper which 
definitely recognized hybridization as a potent source of major economic 
gains beyond what could be secured from the pure breeds. The title of his 
paper, "Rejuvenation by hybridization," is more suggestive of the views of 
Thomas Andrew Knight than of the current students of heterosis, but the 
distinction seems to me to be very tenuous indeed. 

Although I listened with great interest to Simpson's paper, I do not think 
that I recognized any direct applications of his views to my results with 
maize. I was working within the material of a single strain of a single species, 
and not with the hybridizations between different well established breeds to 
the superiority of whose hybrids Simpson called attention. 

Students may make varying estimates as to how closely the work of men 
to whom I have referred approached the heterosis concept as we understand 
it today. But there can be no doubt that there was a beginning of this concept 
in the course of my own experiments with corn. At the beginning of 1907 I 
had not the slightest inkling of such a concept. By the end of 1907 I had 
written the paper that brought such concept clearly into recognition. At that 
time I knew nothing of the work of Beal, Holden, Morrow and Gardner, 



BEGINNINGS OF THE HETEROSIS CONCEPT 19 

McCluer, Shamel or East, in the selling and crossing of the maize plant. 
This will become obvious as I explain the motivation and plan of procedure 
of my corn experiments. 

Upon arriving at the Station for Experimental Evolution at Cold Spring 
Harbor on May 2, 1904, 1 found the laboratory building unfinished. It was in 
fact not ready for occupation until the following November. The potentially 
arable portion of the grounds was in part a swampy area in need of effective 
provision for drainage. The rest had been at one time used as a garden. But 
it had lain fallow for an unknown number of years, and was covered with a 
heavy sod that would need a considerable period of disintegration before it 
could be used satisfactorily as an experimental garden. The total area avail- 
able was about an acre. 

In the middle of this small garden plot was a group of lusty young spruce 
trees. These had to be removed in order to use the area for experimental 
planting the following spring. The ground was plowed, disked, and planted 
as soon as possible to potatoes, corn, sorghum, buckwheat, sugar beets, tur- 
nip beets, and many kinds of ordinary garden vegetables. None of them 
were designed as the beginning of a genetical experiment, but only as an ex- 
cuse for keeping the ground properly tilled so it would be in best possible 
condition for use as an experimental garden later. Due to this fact, no ade- 
quate record was made of the origin of the several lots of seeds which were 
planted. This is unfortunate in the several cases in which some of these cul- 
tures did provide material for later experimental use. 

There were two cultures of corn, one a white dent, the other a Corry 
sweet corn. These two varieties were planted at the special request of Dr. 
Davenport, who wished to have available for display to visitors the striking 
illustrations of Mendelian segregation of starchy and sugary grains on the 
single ears of the crossbred plants. I planted the white dent corn with my 
own hands on May 14, 1904, and must have known at the time that the grains 
came from a single ear. Although I have found no contemporary record to 
that effect, I am now convinced from a well-remembered conversation with 
Mrs. Davenport, that this ear of white dent corn came from the farm of her 
father, Mr. Crotty, who lived near Topeka, Kansas. 

When I was last in Ames, after almost forty years of devotion to other 
lines of genetical experimentation, my memory played me false when Profes- 
sor J. C. Cunningham asked me about the source of the foundation stock for 
my experimental work with corn, and I told him that my studies on corn 
began with some corn I had purchased in the local market as horse feed. I re- 
peated the same unfortunate misstatement to several other highly reputable 
historians of science. I deeply regret this error because these men were tr}-ing 
so hard to get the record straight. My recollection was restored by iinding 
the statement at the very beginning of the record of my formal corn studies 



20 GEORGE HARRISON SHULL 

under date Nov. 7, 1904: "Counted the rows on the ears of White dent corn 
raised in Carnegie garden this year." In fact, as I think of it now, I doubt 
that I could have bought white dent corn in the feed market of Long Island 
at that time. 

I planted the Corry sweet corn on May 17. On July 18 I bagged the corn 
preparatory to making crosses between the two varieties. This crossing was 
carried out on the Corry sweet on July 25, and the crosses for the reciprocal 
combination were made on July 27 and 28. These were the first controlled 
pollinations I ever made in corn, and they were not part of a scientific ex- 
periment. 

My interest in investigating the effects of cross- and self-fertilization in 
maize arose incidentally in connection with a projected experiment with 
evening primroses (Oenothera) to determine the effect, if any, of these two 
types of breeding on the kinds and the frequencies of occurrence of mutations. 
A critic of De Vries's mutation theory had urged that the mutations dis- 
covered by De Vries in Oenothera lamarckiana were artifacts produced by 
selfing a species which, in a natural state, had been always cross-fertilized. I 
developed a program to put this question to a crucial test . Then, it occurred to 
me that it would be interesting to run a parallel experiment to test the effects 
of crossing and selfing on the expressions of a purely fluctuating character. 
Since I had available this culture of white dent maize, I chose the grain-row 
numbers on the ears of corn as appropriate material for such a study. The 
Oenothera problems thus begun, continued to be a major interest throughout 
my genetical career, but it is not expedient to pursue them further here. It is 
important, however, to keep them in mind as a key to my motivation in 
launching my studies with maize. 

In this double-barreled exploration of the genetical effects of cross-fertili- 
zation versus self-fertilization, I had no preconception as to what the out- 
come of these studies would be in either the mutational or the fluctuational 
field. Certainly they involved no plan for the demonstration of distinctive 
new biotypes, nor any thought of the possible economic advantages of either 
method of breeding. I was a faithful advocate of the early biometricians' slo- 
gan: Ignoramus, in hoc signo lahoremus. Until the middle of summer of 1907, 
certainly, I had no premonition of the possible existence of a heterosis prin- 
ciple which would have important significance either scientifically or eco- 
nomically. I was forced to recognize this principle by direct observations of 
manifestations in my cultures which had not been anticipated, and there- 
fore could not have been planned for. 

Let us proceed then to a description of my experiments with corn which 
forced the recognition of this important phenomenon. The culture of white 
dent corn which we had growing, almost incidentally, on the Station grounds 
that first year, showed no variations that seemed to indicate the presence of 
any segregating characteristics. It appeared to be ideal material for the study 



BEGINNINGS OF THE HETEROSIS CONCEPT 



21 



of fluctuations of so definite and easily observed a quantitative character as 
the number of the rows of grains on the ears. The crop was carefully har- 
vested and placed in a crib. On November 7, 1904, I counted the rows of 
grains on every ear, with the result shown in figure 2.1. The 524 ears ranged 
over the seven classes from 10-rowed to 22-rowed. The most populous classes 





10 ... 


.. 3 




12 ... 


. . 93 




14 ... 


. . 201 




16 ... 


. . 153 




18 ... 


. . 58 




20 ... 


. . 12 




22 ... 


.. 4 




Total 


. . 524 


Mean, 


14.827 


± .061 


c>, 


2.082 


± .043 


c. v., 


14.02 


±: .29 




8 



10 



12 14 

NUMBER 



16 18 20 

OF GRAIN ROWS 



Fig. 2.1 — Frequency curve of grain-rows of 524 ears of white dent corn. The total progeny 

of presumably a single ear of corn received from the Crotty farm near Topeka, Kansas, and 

grown at the Station for Experimental Evokition in 1904. 

were the 14-rowed with a frequency of 201, and 16-rowed with 153 individual 
ears. The mean was 14.85 ± .06. 

No photograph nor verbal description was made of the parent ear, since 
there was no intention at the time of its planting to use it in a breeding ex- 
periment. But its characteristics must have been accurately duplicated in all 
of the crossbred families subsequently grown, as well as in most of the Fi hy- 



22 



GEORGE HARRISON SHULL 



brids between the several selfed lines. From each of the grain-row classes, 
several good ears were saved for planting in the spring of 1905, and the rest 
was used as horse feed. 

The plantings from this material were made on May 25, 26, 27, 1905, again 
with my own hands, in the form of an ear-row planting. Two ears from each 
grain-row class of the 1904 crop were used. The seeds were taken from the 
mid-region of each seed ear. An additional row was planted from grains of 
each of the two parent ears with 16 grain-rows. Only modified basal grains 
and modified distal grains for the two halves of the same row in the field 
were used. In Table 2.1 these cultures from modified grains are indicated by 

TABLE 2.1 

GRAIN-ROW COUNTS OF PROGENIES GROWN IN 1905 FROM PARENT 

EARS SELECTED FOR DIFFERENT NUMBERS OF 

GRAIN-ROWS IN NOVEMBER, 1904 



Culture 


Parental 
Grain- 
Rows 


Frequencies of Progeny Grain-Row Numbers 


Numbers 


10 


12 


14 


16 


18 


20 


22 


24 


26 


Totals 


Al 


10.4* 
10 B* 
UA 
UB 
14.4 
UB 
16 A 
\6B 
16.4bt 
16.44 
1656 
16 Bp 
18.4 
18 B 
20 A 
20 B 
22 A 
22 B 
22 CJ 


8 
11 
12 

3 

7 

1 

3 

4 

3 

2 

3 

3 

'r 


55 
50 
36 
30 
32 
11 
62 
31 
7 

19 

3 

5 

12 

20 

3 

10 

2 

3 

2 


47 
57 
45 
43 
58 
47 
81 
79 
19 
18 
5 

18 
36 
33 
28 
14 
9 
9 
12 


16 
15 
10 
28 
13 
26 
44 
66 

7 
16 

8 
12 
39 
29 
38 
28 
21 
20 
32 


3 

1 

1 

4 

5 

13 

10 

14 

2 

4 

4 

11 

17 

7 

14 
14 
27 
28 
24 










129 


A2 


1 








135 


A3 








104 


\4 










108 


AS 










115 


A6 

A7and8 

A9and 10. .. 
All, 


1 








100 






200 


1 








195 








38 


All., 










59 


A12i 

A 12.. 










23 










49 


A13 


3 


1 






108 


A14 






89 


A15 


2 

10 
19 
18 
16 


1 
2 
7 

"3" 






86 


A16 






79 


A17 






85 


A18 


2 

1 


1 
1 


81 


A19 


91 






Totals... 




61 


393 


658 


468 


203 


71 


15 


3 


2 


1,874 













* The significance of the A and B in this column involved the plan to use the A rows for selfing and the B 
rows to be crossed with mixed pollen of plants in the corresponding A rows. 

t The subscript b signifies the use for planting of only the modified basal grains of the given ear; and the sub- 
script p refers to the planting only of modified grains at the "point" or distal end of the ear. 

t C represents an added row grown to increase the probability of finding ears with still higher numbers of 
grain-rows. 

Ab and Bb for the basal grains, and Ap and Bp for the modified "point" 
grains. A second row was planted from each of the two chosen ears having 16 
grain-rows, and these additional rows (A8 and AlO) were detasseled, begin- 
ning July 24, 1905, and received pollen from the intact plants in the corre- 
sponding rows (A7 and A9) beside them. 

In harvesting these two pairs of rows, one detasseled, the other intact, the 



BEGINNINGS OF THE HETEROSIS CONCEPT 23 

two rows from the same parent ear, through an oversight, were not kept 
separate. No further detasseling was done. Since the self-fertilized plants 
could not be detasseled and still utilized for selling, the method of controlling 
cross-fertilization by detasseling would prove a distorting factor in comparing 
the effects of selling and crossing. 

Consequently, no detasseling was practiced in any of my subsequent ex- 
perimental work with corn, but every pollination was controlled by bagging 
with glassine bags and manipulation by hand. The bags were tied in place 
by ordinary white wrapping-cord passed once around and tied with a loop 
for easy detachment. Each plant was labeled at the time of crossing with a 
wired tree-label attached to the stalk at the height of the operator's eyes, 
and marked with the exact identification of the plant to which it was attached 
and the source of the pollen which had been applied. On harvesting these 
hand-pollinated ears, the label was removed from the plant and attached 
securely to the ear, thus assuring that the ear and its label would remain 
permanently associated. A third row (A 19) from an ear having 22 grain-rows 
was added to improve the chances of finding ears with still higher numbers 
of grain-rows. 

In November, 1905, these 19 pedigree cultures were carefully harvested 
by my own hands and the grain-rows counted, with the results tabulated in 
Table 2.1. 

The only observation noted on these 1905 cultures was that there was no 
clear indication of mutations or segregations of any kind, but the aspect of 
the field was that of any ordinarily uniform field of corn. Row counts did 
show the expected indication of Galtonian regression, in that the parents 
with low numbers of grain-rows produced progenies having lower numbers of 
grain-rows than did the ears having higher than average numbers of grain- 
rows. Thus, the two ears with 10 rows of grains each had the average of 13.2 
rows of grains on their progeny ears. The two 20-rowed ears showed an aver- 
age of 15.5 rows of grains on their progeny ears. The three 22-rowed parent 
ears produced progenies with an average of 17.5 rows of grains. 

The same general plan was followed in 1906, except that the pollen for 
the crossbred cultures was no longer taken from the plants set aside for 
selfing. The reason for this change, as specifically stated in my notes written 
at the end of the 1906 season, being "to avoid the deleterious effects of self- 
fertilization in the cross-fertilized series." This indicated that at the end of 
19C6 I had only the concept held by Holden, Shamel, East, and all other 
corn breeders who had had experience with the selfing of maize — that selfing 
has deleterious effects, not that crossing has advantageous effects other than 
the simple avoidance of the deleterious effects of selfing. 

The new method of handling the crossbred cultures was to divide each 
such culture by a marker set at the midpoint of the row. All the plants in 
these rows were bagged. Mixed {)ollen from the plants in the first half of the 



24 GEORGE HARRISON SHULL 

row was collected and applied at the appropriate time to the silks of all the 
plants in the second half of the row. Then the mixed pollen from the plants 
in the second half of the row was applied in turn to the silks of all the plants 
in the first half of the row. It was realized that this still involved a con- 
siderable degree of inbreeding, but it seemed about the only way of carrymg 
on a continuing program of crossing while still keeping the breedmg com- 
pletely under the operator's control. 

Two major observations made on the 1906 crop were: (1) that every one 
of the seven families from selfed parents could be readily detected by their 
less height, more slender stalks, and greater susceptibility to theattack ot 
Ustila2o maydis. When the ears were harvested each lot was weighed and 
it was found that cross-fertilized rows produced on an average about three 
times as much grain as the self-fertilized. (2) The family A3, from a self- 
fertilized ear having 12 grain-rows, was practically all flint corn, showing that 
to be probably recessive. This occurrence of a rather obvious segregation in 
the 1906 crop remained at the end of the season only an isolated observation 
which led to no generalization. From the fall of 1905 until his retirement 
in 1943 Charles Leo Macy assisted me in many of the technical details of my 
experimental cultures. While I handled the planning and breeding operations 
as well as the actual poUinations, Macy prepared the plants for selfing and 
crossing, and counted the grain-rows and weighed the ear corn. The results 
of these counts for the 1906 crop are given in Table 2.2. _ 

The following quotation from my notebook seems justified here, since it 
includes the first formulation of the considerations and conclusions which 
appeared in my report to the American Breeders' Association in 1908, on 
"The composition of a field of maize": 

^- A c;,, 1007 p'; in 1906) namely each self -fertilized row was 

The same plan was continued, ^i" l^^^;^^ ^^^^^^^ row was divided in half, 

the offspring of a smgle self-fertilized far, and each cross termzea ^^^^ 

each half coming from a single cross-fertilized e^^^' f^^^.^f" ^"J^'e ^Vhlr ear coming from the 
the first half of the corresponding row of the precedmg > ear, the other ear c n g 

second half. ... . mnA tv,o coif fprtili/pd rows being invariably 

The obvious results were the same as in ^^O^, the self -fert^ed o^^ s g ..^ 

smaller and weaker than the corresponding """f "^^ ^iliz'-^j^^f l^^^ upon me 

evidence on the self-fertilized A v^ry df erent explana ion of^ h^ P^.^.^^ 

by the fact that the several self- ertlllzedro^^s differ rom each ^^^^^^^^ elementary 

morphological characteristics, thusindicatng that the, be^^^^ ^^^^^ ^^.^.^^ ^^^, 

strains. The same point appeared last >ear in the c^se oi tne ^ ^^^^ 

almost a uniform flint corn, but the ^^^'^^^""^^f.^^^J^Z^^SSr^^^ that my corn- 
It now appears that self-fertilization simply ^^'^f. ^ P^"^ > j^^' but between pure strains 
parisons are not properly between "^^J^^f .^/^"^Jj^'^j^^^^^ 

and their hybrids; and that a well '•^g^^^^^f ,^^^ ^ °4e^^ ^ of 'corn must have as 

its ?b.-t?^rn!?inraTc^ ^^s^ h^d ^^^^^^^^^^^^^ to\e most vigorous and 

productive and give all desirable qualities of ear and grain. 

The ideas in this quotation represent a discovery in complete disagree- 
ment with my preconception that my white dent foundation stock, which 
had been the progeny of a single ear, was essentially a ^^^^'^^^^y f'^l?^'^^' 
I had before me seven distinct biotypes, clearly distinguishable m their sev- 



BEGINNINGS OF THE HETEROSIS CONCEPT 



25 



eral morphological characteristics. They had been derived from seven sepa- 
rate self-pollinations of sibs in a family which I had reason to think was 
genetically homogeneous. This could not fail to make a great imj)ression. 
Had these several pure-bred self-fertilized strains come from different 
breeders and from more or less disconnected experiments, as did the selfed 

TABLE 2.2 

GRAIN-ROW COUNTS AND YIELDS OF EAR CORN IN CULTURES OF 

WHITE DENT MAIZE GROWN AT THE STATION FOR 

EXPERIMENTAL EVOLUTION IN 1906 



Culture 
Numbers 


Parental 
Grain-Rows 


Frequencies of Progeny 
Grain-Row Numbers 


To- 
tals 


Weights 
Lbs.Av. 


Yield 
Bu./A. 


8 


10 


12 


14 


16 


18 


20 


22 


24 


26 


28 


30 


32 


Al.l 


10 selfed 
10 crossed 
10 crossed 
12 selfed 
12 crossed 
12 crossed 
14 selfed 
14 crossed 
14 crossed 
16 selfed 
16 crossed 
16 crossed 
16(22)X10 
16^ crossed 
16j, crossed 

16p crossed 

16 J, crossed 

18 selfed 
18 open-pol. 
20 selfed 
20 crossed 
20 crossed 
22 selfed 
24 crossed 
24 open-pol. 
26 open-pol. 
26 open-pol. 
18 crossed 
18(22)X10 
14(22) XIO 


4 
'5' 


36 
3 
2 

13 

1 
"1 


62 

32 

26 

40 

13 

16 

12 

6 

6 

8 

14 

5 

8 

9 

20 

7 

10 

2 


14 
25 
29 
19 
26 
34 
41 
28 
17 
17 
16 
16 
23 
28 
23 

39 

22 

8 
16 
11 

2 
3 
4 
4 
12 
1 


1 

7 
11 

4 
12 

9 
34 
18 
19 
28 
15 
28 
22 
20 
15 

18 

18 

6 
29 
23 

8 
17 
10 
11 
14 

8 

5 

17 
20 
22 


















117 
68 
69 
82 

581 
61/ 
107 
59\ 
58J 
74 
471 
61/ 
71 
60 
63 

74 

57 

26 
94 

72 

,t?l 

55 
91 
66 
60 
68 
78 
72 
92 






A2.2i 


1 

1 

1 

6 

1 
15 

7 
12 
17 

1 

11 
11 

3 

9 

5 

5 

18 
18 
21 
20 
17 
25 
17 
11 

9 
25 

6 
26 




















A2.22 




















A3.3 
















21.6 
65.8 
33.6 
61.3 
29.6 
58.3 
22.1 


37.7 


A4.4i 


1 
















A4.42 














78.9 


AS.5 


4 














44 9 


A6.6i 














A6.62 


4 
4 

1 
1 
5 














74.5 


A7.7 














59.1 


A9.8i 
















.'^9.82 














77.1 


A19.9 














44.5 


A12i.lOi 












AI22.IO2 




















A12i/"' 

A122i"' 

A13.12 


1 

2 

3 
19 
10 

13 
13 
13 
24 
11 
17 
14 
21 
































2 
9 
4 
5 
4 
7 

18 
6 
6 

19 
7 












9.6 
58 3 
23 6 

56.3 

24.1 
57.3 
32.6 
34.6 
40 6 


52 9 


A14.13 


1 


1 








88.5 


A15.14 






6 


46 9 


A16.15i 


1 












A16.152 
















75. 1 


A17.16 






1 


3 
3 
2 

10 

13 

2 










62 4 


A18.17 


4 








89.9 


A19.18 




1 


3 


70.6 


A19.19 


7 

5 








82 4 


A18.20 














85.4 


A16 21 




"i 


1 
16 
11 


5 
29 
31 




A19. 22 . . 














A19. 23 














46 1 


71.6 


















Totals 




9 


58 


334 


543 


469 


323 


183 


89 


36 


17 


3 


2 


1 


2,067 















lines available to Dr. East, the observation that they showed themselves to 
be genetically distinguishable biotypes would have given no cause for the 
special conclusions I drew from them. It would have been strange, indeed, 
if strains thus derived from heterogeneous sources had not been genetically 
different, one from another. 

Comparison of the results for 1907, presented in Table 2.3, with those for 
1906 in Table 2.2, shows a heavy accentuation of grain-row classes 8 and 10 
and a marked decrease in classes 18 to 20, inclusive. There was also a sig- 
nificant increase in all higher classes, with further extension of the range from 
a maximum of 32 to about 40. The increase in the frequencies of the low 



26 



GEORGE HARRISON SHULL 



grain-row classes was attributed in part to the fact that the 1907 season had 
seemed less favorable in general than 1906. 

It was also noted, as a possible contributory condition, that this was the 
third season in which this corn was grown on the same area north of the 
laboratory building, and that ''the yield may have been lessened by the 
gradual accumulation of injurious substances in the soil." The fact that the 




Fig. 2.2 — Young corn cultures growing in East Garden of the Station for Experimental 
Evolution in 1911, illustrating that no two were alike despite their descent from a single ear 
of 1904 by meticulously controlled pollinations that precluded the introduction of pollen 

from any other strain of corn. 



average grain-row numbers were not significantly different in the two years — 
15.8 in 1906, 16.0 in 1907 — in fact a trifle higher in what was thought to have 
been the poorer year, does not seem to support these suggested explanations 
of the observed differences of distribution in the two years. 

My contemporaneous notes proposed an additional explanation, namely, 
that "each successive generation of close inbreeding still further reduces the 
strains to their simple constituent biotypes, and as these are weaker than 
hybrid combinations, this too would tend to lessen the vigor, and this 
lessened vigor might readily be evidenced by a decrease in the average num- 
ber of [grain-] rows and the total number of ears in the crop." 

If we accept this latter suggestion as valid, it is clear that the occurrence 



BEGINNINGS OF THE HETEROSIS CONCEPT 



27 



of essentially the same average numbers of grain-rows in the two years gives 
only a specious indication of the relative climatic and soil effectiveness in 
these two seasons. It must mean simjjly that the diminution of grain-row 
numbers produced by increasing homozygosity happened to be balanced by 
the increased frequencies in the higher classes, produced by the gradual ac- 
cumulation by selection of more potent hybrid combinations. 

TABLE 2.3 

GRAIN-ROW COUNTS AND HEIGHTS OF PLANTS IN 
THE CULTURES OF 1907 



Pedigree 


Grain-Rows 
of p.\rents 




Frequencies 


DF Progeny Grain- Row Numbers 


To- 
tals 


Av. 
Ht 


Numbers 


8 


10 


12 


14 


16 


18 


20 


22 


24 


26 


28 


30 


32 


34 


36 


38 


40 


IN 

Ft. 


Bl.l 


10 selfed 
10 crossed 
10 crossed 
8 selfed 
12 selfed 
12 crossed 
12 crossed 
14 selfed 
14 crossed 
14 crossed 
16 selfed 
16 crossed 
16 crossed 

16 crossed 
I61 crossed 
18 selfed 
18 open-pol. 
20 selfed 
20 crossed 
20 crossed 
22 selfed 
22 crossed 
20 crossed 
24 crossed 
32 open-pol. 
Branched ear 
30 open-pol. 
Branched ear 
16 crossed 
24 selfed 
26 selfed 


20 

2 

6 

23 

10 

"1 


25 
22 
28 
48 
21 
1 

' '4 


20 
21 
18 
17 
18 
16 

7 
23 

5 

5 
19 

2 


2 

5 
5 


1 


























68 

501 

57/ 

88 

54 

52 

44 

73 

48! 

43/ 

55 

371 

36/ 

311 

37/ 

15 

71 

52 

631 

52/ 

45 

63 

62 

50 

41 

58 

58 

49 

64 

36 

12 


7.25 


B2i.2i 




























B22.22 




























9.00 


B1.3 




























7.63 


B3.4 


4 
19 
14 
29 
15 
18 
26 
8 
9 

13 
15 


1 
13 
14 
15 
19 
19 

9 
14 
15 

10 

9 

5 

22 

21 

16 

10 

6 

9 

7 

1 

3 

9 

5 

14 

14 

5 


























6.25 


B4i.Si 


3 

7 

1 
7 
7 
























8 00 


B42.S2 


1 
1 
1 
1 
























B5.6 






















8.50 


B6i 7i . . 


1 
1 






















B62.72 




















9 67 


B7.8 




















8 25 


B8i.9i 


7 
9 

1 

1 

7 

15 

11 

18 

13 

9 

17 

22 

7 

6 

13 

12 

17 

15 

5 


3 
3 

1 

1 

1 

3 

6 

13 

16 

17 

17 

17 

16 

8 

8 

17 

12 

17 

7 

3 


3 






















B82 92 ... 




















8 75 


BIO2 lOi. .. 






6 
11 
























BlOi IO2 . 






















8.67 


B12.il 


2 

1 

1 

10 

8 

8 

11 

11 

14 

8 

1 

11 

3 

6 

7 

2 




















7.00 


B13 12 






9 


21 
8 
3 
4 




















8 33 


B14.13... . 




















7.25 


B15i 14i... 


2 

1 
5 
4 
4 
5 
7 
1 
9 




















B152.142... 


















8 83 


B16.15 
























7.00 


B17.16... . 








1 
1 


3 


1 














8 67 


B15.17 




















8 83 


B19.18 








4 
4 


3 

2 














9 50 


B20.19... . 










2 












9 50 


B? 20. . . 








6 

2 
3 

5 








8 33 


B17 21... . 


1 
















8 33 


B?.22 


















BIS. 23 








3 
6 
2 


4 
6 


















B20 24 






















8.00 


B20.2S... . 












1 


2 




1 




1 


7.83 


















Totals . . 




62 


150 


204 


236 


282 


228 


189 


108 


49 


22 


6 


3 


3 


1 


1 




1 


1,545 











A truer measure of the relative favorableness of the two seasons for growth 
and productiveness of these cultures can be derived from a study of the 
middle classes with 12, 14, 16, and 18 grain-rows. These grain-row classes 
making up 80 per cent of the 1906 crop and 61.5 per cent of the 1907 crop, 
must be relatively free from most of the distortion assumed to be produced 
either by increasing homozygosity or by the accumulation of the more po- 
tent hybrid combinations. If we average these four grain-row classes by them- 
selves for the two years, we find that in 1906 their average was 15.5 grain- 
rows, and for 1907 only 15.0, thus agreeing with my general impression 
that 1907 was the less favorable year. 

With the fundamental change in my understanding of the nature of my 
corn population came a reorientation of the experiment. I found myself at 



28 GEORGE HARRISON SHULL 

the end of 1907 only ready to make a beginning on the problems of the rela- 
tionship between pure lines and their hybrids, which I now saw was the cru- 
cial field that needed exploration. 

As a first step in that direction, but without as yet a full comprehension 
of its importance, I made in July, 1907, pollinations between plants of C4, 
which I later designated "Strain A," and a plant of C6, which later became 
my "Strain B." I also made two sib crosses within these two strains. The 
cross of Strain A X Strain B, which gave rise in 1908 to Fi family, D9, in- 
volved an 8-rowed ear of the former strain (from an original selection for 
12 grain-rows) and a 12-rowed ear of Strain B which had originated in a selec- 
tion for 14 grain-rows. The near-reciprocal cross (Fi family, D13) resulted 
from the application of pollen from a 12-rowed plant of Strain A to silks of 
the same plant of Strain B, which supplied the pollen for the near- reciprocal 
cross. 

At the time when these two near-reciprocal crosses were made between 
Strains A and B, the truth had not yet dawned upon me that I should do the 
same with all of my other selfed families. Aside from these two sets of crosses, 
the handling of the cultures was the same as in previous years. The results 
of the grain- row counts are given in Table 2.4. Unfortunately, there was con- 
siderable damage from crows, and failures to germinate for unknown reasons. 
The missing hills were replanted on June 8, 1908, and all of the new plantings 
made on this date seem to have reached maturity. To overcome the suggested 
deteriorating effect of soil depletion, the cultures were grown this year on the 
area east of the laboratory building (occasionally referred to in subsequent 
notes as "Fast Garden"). 

In summarizing the results for the year 1908, it may be noted first that 
the tendency to concentrate the frequencies of the grain-rows in the extremes 
of the range, at the expense of those in the middle, has continued strongly. 
As before, the most noteworthy concentration is at the lower extreme. All 
classes below 16 are considerably stronger in 1908 than in 1907 and the 
maximum frequency is now on 12 instead of 16. This is in part due to the fact 
that several of the lower-class families were grown in duplicate. Between 
classes 14 and 26 the relative strength of the classes was lessened in 1908. 
Above class 24 the frequencies were increased, there being 84 ears above 
class 24 in 1908 and only the equivalent of about 50 in the same region in 
1907, when raised to the same total number. The highest number of grain- 
rows noted was 42. 

The important new features brought in by the near-reciprocal crosses be- 
tween Strain A and Strain B and a sib cross in Strain A are presented in my 
report to the American Breeders' Association at Columbia, Mo., in January, 
1909, on "A pure line method in corn breeding." I find a discrepancy in that 
the 78 ears produced by the sib cross weighed only 16.25 pounds instead of 
16.5, as stated in my 1909 paper. Whether by an oversight or intentionally, 



BEGINNINGS OF THE HETEROSIS CONCEPT 29 

I cannot now determine, the corresponding sib crosses in Strain B were not 
included in my 1909 report. The results were essentially the same as were re- 
ported for the sib cross in Strain A. Selfed Strain B (see Table 2.4, family 
C6.ll) showed average heights of plants 2.3 meters, and yielded 66 ears 
weighing 13.0 pounds. The two sib crosses produced plants 2.5 meters tall 
and yielded 89 ears weighing 28.5 pounds. Distribution of the grain-row 
frequencies was closely similar in selfed and in sib-crossed Strain B, but sig- 
nificantly higher in the latter: 

_, . Totals Averages 

Grain-rows 10 12 14 16 18 

Selfed 2 20 26 17 1 66 1.3 8 

Sib-crossed 3 15 45 18 8 89 14.2 

There was abundant evidence that the sib crosses showed a greatly re- 
stricted advantage over self-fertilization. It was also clearly indicated that 

TABLE 2.4 

GRAIN-ROW COUNTS, HEIGHTS, AND YIELDS OF 
WHITE DENT MAIZE GROWN IN 1908 



Pedigree 


Gr.^in-Rows 

of p.4rents 


Frequencies of Progeny 
Grain-Row Numbers 


To- 
tals 


Av. 

Hts. 

IN 

Dm. 


Wts. 

IN 

Lbs. 


Yield 

Bu./ 

A. 




8 


10 


12 


14 


16 

17 
11 

6 
11 

3 
28 
25 
13 
34 
16 
14 
39 

4 

6 
17 

9 
22 

3 
tvs t 

1 

14 
20 

323 


18 

"1 

1 

5 

3 

1 

2 

2 

1 

11 

4 

21 

12 

4 

23 

15 

17 

19 

20 

26 

4 

30 d 

2 

19 
31 

244 


1 
20 

"i 

6 

1 

6 
19 
10 
18 
24 
25 
21 
fBci 
5 

16 
22 

172 


22 

1 

"7 
11 
3 
13 
19 
16 

It t( 
10 

14 

7 

91 


24 

8 

16 
21 
24 
) coi 
10 

4 
3 

60 


26 

"2 

"4 

"7 

int; 

16 

2 
31 


28 

"i 

3 

silks 
18 

2 
24 


30 

i 

sh 
5 

6 


32 

3 

ort 
6 

9 


34 

i 

er 
9 


3d 


38 


40 


42 


C3.1 

C1.2 

C2,.3, 

C22.32 

C1.4 

C2i.5i 

C22.52 

C4.6 

C4.7 

C4.8 

C4.9 

C5i.ini.... 

C52.102.... 

C6.ll 

C6.12I..., 

C6.122 

C6.13 

C7,.14i.... 
C72.142.... 

C8.15 

C9i.l6i.... 
C92.162.... 
C13.17.... 
CI22.I81... 
C122.I82... 
C13.19.... 
C14,.20i... 


10 selfed 
8 selfed 
8 crossed 
8 crossed 
10 selfed 
10 crossed 
10 crossed 
12 selfed 
lOXsib 
8 selfed 
8X12 
12 crossed 
14 crossed 
14 selfed 
16Xsib 
12Xsib 
12X12 
14 crossed 
14 crossed 
16 selfed 
16 crossed 
16 crossed 
18 selfed 
18 crossed 
18 crossed 
20 selfed 
20 crossed 
20 crossed 
20 selfed 
22 crossed 
22 selfed 
28 crossed 
36? selfed 
28(?)X26(?) 
Branched ear 

open-pol. 
20 open-pol. t 


52 
51 

6 

6 
28 

9 
12 
11 

8 
65 

"2 
1 


39 

41 

29 

22 

48 

32 

18 

41 

50 

6 

19 

9 

9 

2 

2 

ii 


13 
2 

14 

12 

12 

9 

3 

32 

19 

2 

64 

31 

17 

20 

4 

11 

56 

18 

9 

31 

4 

6 

6 

6 

8 

2 


"2 

1 
2 

"5 
1 

'9 
15 
14 
26 
25 
20 
31 
28 
18 
32 
14 
18 
10 
20 
19 
15 
3 
2 

13 
3 
9 


tha 
3 


n'h 


usl 


cs. 

1 


104 
94 
51 
42 
90 
50 
33 
89 
78 
73 
92 
59 
42 
66 
47 
42 
100 
59 
43 
94 
55 
41 
77 
56 
46 
85 
58 
46 
70 
84 
92 
83 

86 

83 
93 


19.5 
19.7 

23.4 

is'o 

21.5 

'i7'0 
16.5 
16.5 
24.0 
24.5 
22.5 
23.0 
25.0 
25.0 
26.0 
25.0 
27.0 
24.4 
26.8 
25.2 
19.3 
23.5 
25.5 
21.6 
. . . .* 


31.5 
22.0 
25.01 
21.0/ 

22 
20.81 
14.0/ 
28.0 
16.3 
12.0 
48.0 
34.8 

23 3 
13 
16.8 
11.8 
55 . 
30.0 

19 3 
31.5 
31.5 

20 
16.5 
31.3 
28.3 
23 
31.0 
24.5 
20.5 
48.3 
33.8 
43.3 

50.5 

50 
51.8 


43.3 
33.4 

70.7 

30.5 

59.8 

44.9 
29.8 
23.5 
74.5 
84.1 
79.1 
28.1 
50.9 
40.0 
78.6 
72.6 
64.0 
47.9 
80.8 
69.7 
30.6 
79.7 
87.7 
38.7 


CI42.2O2.. . 








76.4 


C15.21... . 








79.2 


CI61.22. . . 








41.8 


C24.23.... 
CIS. 24.... 






2 


82.1 

52.4 


C25.2S.... 
C19.26.... 






Grai 


n-ro 


74.4 


C22.27.... 
C22.28.... 






1 
1 


11 
9 


83.9 

86.2 
79.5 


Totals . 




^S^ 


387 


415 


375 


10 


3 






1 
















2,403 











* The remaining nine rows were not measured and described, "for lack of time.' 
lanlil^'' '''''"' ""'^"^ ^°"' ^^'' "^'^^ '■*' ''^'^^' ^""^ ^° '"""^ °^ ^''''"'' °^ ""'^'""^ °"'^' "'"^ twenty-rowed ear was used for 




Fig. 2.3— Vegetative habits of Strain A (righl) and Strain B, drawn by J. Marion Shull 
from a photograph taken in the summer of 1908. At upper right typical ears of these two 
strains {Strain A at right) and between them their reciprocal Fi hybrids, each hybrid stand- 
ing nearest to its mother type. 



BEGINNINGS OF THE HETEROSIS CONCEPT 31 

if the advantage consisted solely of the effects of heterozygosity, both Strain 
A and Strain B were still a good way from being homozygous, Strain B being 
as yet more effectively heterozygous than Strain A. 

In the reciprocal crosses between these nearly homozygous strains A 
and B, we have our first opportunity to arrive at an approximation to the 
actual amount of heterosis. The most important new discoveries these 
crosses made possible were: (1) As a result of such a cross it is possible to 
completely cancel in a single year the accumulated deterioration which 
had gradually accrued, although with lessening annual increments, over a 
period of several years; and (2) the approximate identity of the results of the 
reciprocal crosses gave assurance that the amount of heterosis resulting from 
a given hybridization is a specific function of the particular genetical combi- 
nation involved in the cross. 

Several new cultures of yellow- and red-grained corn were added to my 
experimental field in 1908, but these will not be followed here. They are 
mentioned only because they were included in my numbered pedigrees, and 
their omission in the following tables leaves a break in the series of numbered 
families which might lead to some question as to the reason for the apparent 
vacancies. The data from the 1909 cultures of white dent corn are presented 
in Table 2.5. 

The families grown in 1909, as tabulated in Table 2.5, fall into three major 
classes: (1) Twelve families involve continuations of the original self- fer- 
tilized lines, whose average yields range from 18.8 to 41.2 bushels per acre, 
with the average for all twelve at 32.8 bushels per acre; (2) Twelve are con- 
tinuations of crossbred families in which strictly controlled cross-fertiliza- 
tions were made with mixtures of pollen taken from the other plants in the 
same crossbred strain. These yielded from 58.1 to 83.3 bushels per acre with 
the average of all at 73.3 bushels per acre; and (3) there were fourteen Fi hy- 
brid families from crosses between pairs of individuals representing two dif- 
ferent selfed lines. The yields of these range from 60.3 to 87.5 bushels per 
acre, the average for all fourteen being 78.6 bushels per acre. As stated in my 
1910 paper, the three highest yields of any of these cultures were from the 
families produced by crossing representatives of different selfed strains (see 
D8.13, D8.16, andD11.21). 

Besides these, there were two cousin crosses involving matings between 
different families of the same selfed line. These produced, respectively, 27.1 
and 44.6 bushels per acre. One cross between two sibs in Strain A gave 26.0 
bushels per acre. The other cross was two F2 families, each from crosses with 
mixed pollen within one of the Fi families of my 1908 cultures. These Fo 
families yielded 54.2 bushels per acre from the (A X B)F2, and 70.6 from 
the (B X A)F2. These yields should be compared with those of the corre- 
sponding Fi families grown in the same season, in which (A X B)Fi yielded 
74.9 and 83.5 bushels in two different families, and (B X A)Fi produced 
82.6 bushels per acre. 



32 GEORGE HARRISON SHULL 

In 1910 I was absent from the Station for Experimental Evolution during 
the entire summer and my experiments with corn, evening primroses, 
Lychnis, etc., were continued by an assistant, R. Catlin Rose, assisted by 
Mr. Macy, who carried out the operations meticulously described by myself 
in more than one thousand typewritten lines of detailed instructions. 

The data on the white dent corn grown in 1910 are presented here in 

TABLE 2.5 

GRAIN-ROW COUNTS, HEIGHTS OF ST.ALKS, AND YIELDS OF 
EARS OF WHITE DENT CORN IN 1909 



Pedigree 
Numbers 


Grain-Rows 
OF Parents 


Frequencies of Progeny 
Grain-Row Numbers 


To- 
tals 


Hts. 

IN 

Dms. 


Wts. 

IN 

Lbs. 


Yield 
Bu./A. 


8 


10 


12 


14 


16 


18 


20 


22 


24 


26 


28 


30 


32 


34 


36 


38 


40 


Dt.l 

D2.2 

D3i.3i. . . 
D32.32. . . . 

D4.4 

D4 5 


8 selfed 
8 selfed 
8 crossed 
8 crossed 
10X12 
10X14 
lOXsib 
10 selfed 
10 crossed 
10 crossed 
12 selfed 
12Xcousins 
12Xcousins 
A selfed 
AX20 
AX22 
AXB 
AX20 
AX 16 
AXB 

(AXB)Fisibs 
12 crossed 
12 crossed 
B selfed 
BXA 
BX20 

(BXA)Fisibs 
14 crossed 
14 crossed 
16 selfed 
16 crossed 
16 crossed 
18 selfed 
20X16 
20XA 
16Xcousin 
1 8 crossed 
18 crossed 
20 selfed 
20X16 
20 crossed 
20 crossed 
22 selfed 
22 crossed 
22 crossed 
24 selfed 
24 crossed 
24 crossed 
30 selfed 
28 crossed 
28 crossed 
22 crossed 
24 crossed 


21 
29 
18 
8 
30 

io 

7 
3 
4 
5 
1 
3 
66 


51 
70 
25 
39 
55 

8 
53 
32 
23 
22 
50 
31 
29 

5 


30 
6 

12 

3 

21 

44 

32 

55 

17 

15 

35 

18 

20 

3 

4 

44 

18 

21 

74 

71 

57 

28 

25 

10 

58 

6 

26 

13 

14 

25 

2 

4 

2 

1 

16 

3 

5 


ii 

1 
4 
1 

4 

1 

'46 
50 

9 
33 
32 

5 
11 
16 
17 
18 

9 
20 
40 
23 
18 
51 
11 
19 
14 

4 
46 

9 
18 
18 
14 
12 
























'3 


4 


102 

105 
55 
50 

106 
63 
96 
98 
44 
41 
94 
50 
53 
74 
96 

102 
31 
60 

115 
86 

108 
51 
51 
40 
86 
80 
84 
48 
41 
81 
35 
48 
73 
96 
85 
36 
47 
51 
87 

113 
52 
67 
91 
32 
44 
97 
34 
21 
68 
39 
44 

100 

104 


18 
20 
21 
22 
20 
24 
17 
19 
24 
24 
18 
19 
19 
17 
24 
26 
24 
26 
28 
27 
25 
25 
23 
26 
28 
28 
27 
28 
29 
24 
25 
26 
20 
27 
24 
23 
28 
28 
24 
28 
30 
29 
26 
25 
27 
23 
27 
27 
25 
29 
28 
29 
29 


24.0 

24.8 

21.0'i 

22.5/ 

44-8 

35.3 

17.5 

25.0 

17. 3\ 

18. 3J 

23,5 

9.5 
10,3 

9.8 
54 
60.0 
16,3 

29 8 
61,3 
50.3 
41 
29,5 

30 
7.3 

49.8 

49 

41,5 

23,81 

20 8/ 

21.0 

22.51 

24.0/ 

17.3 

53.8 

46 

11.3 

26.01 

28.5/ 

20.3 

63 3 

29, 0\ 

34 8/ 

25,3 

17,51 

26.8/ 

28 

14 81 

12.8/ 

11 5 

14,31 

19,5/ 

37,3 

53 , 5 


53 
33.7 

59.2 

60,3 
80 


D4.6 

D4.7 

DS1.81 

D52.82. . . . 

D6.9 

D7.10i.... 
D7.102. .. . 

D8.11 

D8 1' 


45 
7 
2 

5 
7 

■2 

"3 
12 

38 

15 

10 

8 

4 

9 

19 

42 

27 

22 

19 

17 

23 

36 

54 

2 

5 

4 

5 

6 

5 


9 

3 

15 

"8 

6 

15 

43 

' '4 

5 

4 

27 

36 

12 

19 

30 

12 

5 

22 

2 


' '4 

18 

"1 

1 

1 

8 

9 

23 

21 

41 

11 

8 

36 

10 

2 

2 

5 

4 

27 

28 


"3 

"i 

8 

14 

12 

4 

10 

16 

7 

5 

1 

5 

4 

11 

21 


"5 
7 
4 

"s 

12 
11 
3 
4 
3 
6 
2 
7 


' '2 
1 

'2 
5 
3 
7 

12 
5 
7 

'4 


' 1 
1 
1 
2 

14 
8 
8 
1 


'2 

"2 
9 

5 
5 

"1 


li 
1 
3 


4 


■ ■ 

4 
'2 


26.0 
56.4 

59.7 

35.7 
27.1 
27.6 
18 8 
80 4 


D8.13 

D8.14i.... 
D8 142 


::: 


1 
2 


84.0 
74.9 
70.8 


D8.15 

D8.16 

D9.17 

DIO1.I81. . 
DIO2.I82. . 
D 1 1 19 


1 

2 
3 
2 


1 
8 

32 

5 
5 


76.2 
83.5 
54,2 

83^3' 
25 9 


1)11.20.... 
Dll M 




19 


82 6 

87,5 


D13.22.... 

D14i.23i. 

D142.232 




1 
2 


70,6 
71.4 


D15.24.... 
D16i.25i 




1 


37.0 
80.0 


D16'.252 






D17.26.... 






33 8 


D17 27 






80 


D17.28.... 
D17.29. . 






77.3 
44.6 


DI81 3O1 






79.4 


DI81.3O2 






D19.31 






33 3 


D19.32.... 






80.0 


D20i.33i 






76.5 


D2O2.332 










D21.34.. 










39.6 


D22 35i 










83.2 


D22.352 








1 


D23.36.... 








41.3 


D24 37i 










71.4 


D24.372 












D'5 39 














24.2 


D26.40i 










3 

2 

23 

18 


4 

2 

31 

22 


58.1 


1)26.402 










1)27 41 








5 
3 


81.8 


D28 42 








73.5 












Totals 




214 


570 


846 


588 


497 


341 


261 


123 


73 


48 


36 


24 


16 


5 


6 


3 


4 


S6S'; 





















Without- 

Se)jfftrti)«* 

tjon. 

live Greneyatiew'X 




1 1 k I i 



f7 libtr-u. 







I 


■' 


I 


1 


^L 


■t.. 


wm^ 


Bi. 






Fig. 2.4 — An exhibit set up in the Genetics Department of Cornell University in 1910, dis- 
playing materials grown at the Station for Experimental Evolution in 1909. 




Fig. 2.5— The best eleven ears of the highest-yielding selfed line (F 29.70 in Table 2.7) 
grown in 1911 {top row) ; the best eleven ears of the best Fi hybrid grown in the same year 
(F 32.75 in Table 2.7) ; and the best eleven ears of a crossbred strain (F 55.84 in Table 2.7) 
in which selling was completely prevented during five years. This shows the relative vari- 
ability which is characteristic of these three types of families, the Fi being no more variable 
than the inbred, while the crossbred is quite noticeably more variable. 



34 



GEORGE HARRISON SHULL 



summary form. Some 73 ears were selected for planting, and 5,343 ears were 
harvested. The complete grain-row distribution was as follows: 

Grain-rows 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 Total 

Frequencies ... . 401 812 1271 921 716476275 141 118 74 53 41 24 8 6 4 1 1 5343 
Percentages. ... 7.5 15.2 23.8 17.2 13.4 89 5.2 2.6 2.2 14 10 0.8 0.5 0.2 0.1 0.1 00 0.0 100.0 



To save space and still indicate as completely as possible the significant 
results of these studies in 1910, the data from the several kinds of families 
of white dent corn grown at the Station for Experimental Evolution that 
year are presented in the form of averages in Table 2.6. The several quanti- 
tative indicators of physiological vigor, namely, the average number of 
grain-rows, heights of stalks, and bushels of ear-corn per acre, can be readily 
compared as follows: 



Types of Families 



Inbreds selfed .... 
InbredsXsibs. . . . 

Crossbreds 

Fi between inbreds 
F2 from Fi selfed. . 
F2 from Fi X sibs . . 



No. of 


Av. No. of 


Families 


Grain-Rows 


10 


12.6 


8 


13.7 


11 


16.9 


6 


15.2 


11 


13.3 


11 


13.5 



Av. Heights 
in Dms. 



19.3 
19.8 

23.5 
25.7 
23.3 
23.1 



Av. Yields 
in Bu./A. 



25.0 
28.7 
63.5 
71.4 
42.6 
47.9 



Si.x interesting comparisons can be made among these summaries: (1) 
comparisons between inbreds selfed and inbreds crossed with pollen from 
one or more of their sibs; (2) comparisons between inbreds and crossbreds 
in which selfing has been completely prevented, but which still represent a 
(fairly low) degree of inbreeding; (3) comparisons between inbreds and their 
Fi hybrids; (4) comparisons between the crossbreds in which selfing has been 
prevented through six generations and the Fi hybrids in which five successive 
generations of selfing have been succeeded by a single cross; (5) comparisons 
between the Fi and the F2 hybrids of the inbreds; and (6) comparisons be- 
tween F2 hybrid families produced by selfing the Fi and those F2 families 
produced by sibcrosses in the Fi. 

On making these comparisons we see that the evidence for residual hetero- 
zygosity in the inbreds is indicated by excesses in the sibcrossed families of 
the inbreds over the selfed inbreds of 8.7 per cent in grain-row number, 2.8 
per cent in heights of stalks, and 14.7 per cent in yield of ear-corn. In the Fo 
families (sections E and F, of Table 2.6) those produced from sibcrosses in 
the Fi surpass those families produced from selfings in the Fi by 0.9 per cent 
in grain- row number and 12.5 per cent in yield. 

The average heights of stalks reverse the expectation by showing an in- 
significantly less height from the sibcrossed matings than from the selfings, 
the difference being 0.9 per cent. The contrast between the results of six 
successive selfings and the continued prevention of selfing for the same six 



TABLE 2.6 

AVERAGE V.ALUES IN THE FAMILIES OF WHITE DENT MAIZE 

GROWN IN 1910, GROUPED ACCORDING TO THE 

TYPES OF MATING OF THE PARENTS 



Pedigree 
Numbers 


Parental 
Grain-Rows 


Number 
of Stalks 


Av. No. 
of Grain- 
Rows 


Heights 
in Dms. 


Wts. in 
Lbs. Av. 


Yields 
Bu./A. 




(A) Families from Inbreds Selfed 


El. 16 . . 


8 selfed 
8 selfed 
10 selfed 
12 selfed 
A(8) selfed 
B(14) selfed 
14 selfed 
18 selfed 
22 selfed 

26. 28 selfed 

■ 


57 
83 
79 
80 
75 
53 
66 
82 
62 
72 


10.0 

9.0 
11.1 
12.3 

8.8 
12.9 
13.8 
15.2 
17.9 
15.2 


17 

18 

20 

17 

16.5 

24 

23 

19 

19 

19 


9.8 
22.0 
18.3 
11.4 
9.1 
7.3 
16.3 
15.3 
11.0 
17.5 


24 4 


E2.19 


^9 6 


E7.29 


33 9 


E9.32 


20 9 


E11.34 

E19.47 

E24.54 

E26.56 

E34.67 

E36.71 


18.1 
11.0 
25.8 
22.9 
19.2 
34.2 




Unweighted averages 


71 


12.6 


19.3 


10.7 


25.0 




(B) Families from Inbreds Pollinated by Sibs; Selfing Prevented 


El. 17 


lOXsibs 
10 X sib 
12 X sib 
A(8)Xsib 
B(12)Xsib 
18 X sib 
20 X sib 
?(fasc.)Xsib 


61 

75 
85 
55 
54 
89 
65 
73 


10.2 
9.9 
11.0 
9.5 
12.7 
15.8 
17.9 
22.5 


19 
18 
22 
16 
24 
20 
20 
20 


13.8 
21.0 
18.3 
7.5 
5.3 
24.5 
15.3 
18.3 


29 8 


E2.20 


39 5 


E7.30 


37 ^ 


E11.33 

E19.48 

E26.57 

E34.68 

E36.72 


16.0 
7.8 
37.8 
25.6 
35.2 




Unweighted averages 


61 


13.7 


19.8 


15.5 


28.7 




(C) Families from Parents Given Mi.^ed Pollen in Each Generation; 
Selfing Prevented 


E3.23 ... 


8, 10 crossbred 
10 crossed 
12 crossed 
14 crossed 
16 crossed 
18 crossed 
20 crossed 
20, 22 crossed 
24, 20 crossed 
32 crossed 
32 crossed 


88 
65 
91 
94 
95 
202 
100 
45 
69 
56 
99 


9.5 
10.3 
13.2 
13.7 
14.9 
16.0 
18.5 
20.0 
24.2 
19.2 
26.2 


22 

22 

24 

27 

28 

22.5 

23 

21 

22 

24 

23 


30.8 
31.0 
51.0 
49.0 
48.8 
76.8 
35.8 
26.3 
24.5 
22.5 
39.0 


49 9 


E8.31 


68 1 


E18.46 

E23.53 

E25.55 

E30.63 

E33.66 

E35.70 

E37.73 

E40.75 

E40.76 


80.1 
74.5 
73.3 
54.3 
51.1 
83.3 
50.7 
57.4 
56.3 




Unweighted averages 


91.3 


16.9 


23.5 


39.6 

• 


63.5 



TABLE 2.6 — Continued 



Pedigree 

Numbers 


Parental 
Grain -Rows 


Number 
of Stalks 


Av. No. 
of Grain- 
Rows 


Heights 
in Dms. 


Wts. in 
Lbs. Av. 


Yields 
Bu./A. 




(D) Fi Hybrids between Different Inbred Lines 


E2.21 


A(10)X16 

A(10)XB 

A(8)X10 

A(8)XB 

18X14 

18X26 + (fasc.) 


95 
94 
95 
84 
109 
92 


13.8 
12.8 
11.0 
12.3 
17.8 
23.3 


24 
28 

25 
25 
27 
25 


50.3 
50.0 
33.5 
28.5 
60.8 
62.5 


75.6 


E2.22 


76.0 


E11.36 

E11.37 

E26.58 

E34.69 


51.5 
48.5 
79.6 
97.1 




Unweighted averages 


93 


15.2 


25.7 


47.6 


71.4 




(E) F2 Families from FiXSelf 


E4.24 


(lOXA)Fi selfed 
(10Xl4)Fiselfed 
(A X20)Fi selfed 
(AX 22)Fi selfed 
(AX 16)Fi selfed 
(A XB)Fi selfed 
(BXA)Fi selfed 
(BX20)Fi selfed 
(20Xl6)Fi selfed 
(20XA)Fi selfed 
(20 Xl6)Fi selfed 


86 
86 
76 
83 
94 
96 
95 
92 
97 
95 
93 


10.6 
12.1 
13.9 
12.8 
12.8 
12,0 
11.7 
15.1 
16.6 
13.0 
15.9 


21 

22 
19.5 

24 
25 
25 
24 

25 
25 
22 
24 


30.8 
29.8 
20.5 
18.8 
33.5 
24.0 
25.3 
28.0 
35.3 
22.0 
29.5 


51.1 


E5.26 


49.4 


E12.38 

E13.40 

E15.42 

E16.44 

E20.49 

E21.51 

E27.59 

E28.61 

E32.64 


38.5 
31.4 
50.9 
35.7 
38.0 
43.5 
51.9 
33.1 
45.3 




Unweighted averages 


90.3 


13.3 


23.3 


27.0 


42.6 




(F) F2 Families from FiXSibs 


E4.25 


(10Xl2)F,Xsibs 

(10Xl4)FiXsibs 

(AX20)FiXsibs 

(AX22)FiXsibs 

(AXl6)FiXsibs 

(AXB)FiXsibs 

(BXA)FiXsibs 

(BX20)FiXsibs 

(20Xl6)FiXsibs 

(20XA)FiXsibs 

(20Xl6)FiXsibs 


85 
83 
80 
96 
95 
93 
80 
93 
89 
92 
97 


10.7 
12.2 
14.2 
13.4 
12.3 
11.8 
11.6 
15.5 
17.2 
13.7 
15.4 


21 
22 
21 
25 
23 
24 
24 
25 
25 
23 
21 


31.3 
35.0 
28.8 
27.0 
37.3 
21.0 
23.5 
31.8 
37.3 
30.0 
25.3 


52.5 


E5.27 


60.2 


E12.39 

E13.41 

E15.43 

E16.45 

E20.50 

E21.52 

E27.60 

E28.62 

E32.65 


51.3 
40.2 
56.0 
32.3 
42.0 
48.8 
59.8 
46.6 
37.6 




Unweighted averages 


89.4 


13.5 


23.1 


29.8 


47.9 



BEGINNINGS OF THE HETEROSIS CONCEPT 37 

years (sections A and C, Table 2.6) shows the latter in excess of the former 
by 34.0 per cent in grain-row number, 22.1 per cent in height of stalks, and 
154.2 per cent in per acre yields of ears. The superiority of the Fi hybrids 
between different inbreds and the families in which selfing had been pre- 
vented during six generations of controlled breeding (sections D and C, 
Table 2.6), is indicated by an excess in heights of stalks of the Fi families 
over the crossbreds, of 9.4 per cent, and in yields of ear-corn per acre of 12.3 
per cent. But here there is a notable reversal in grain-row numbers. Not- 
withstanding these proofs of the superior vigor of the F/s over the cross- 
breds, the latter exceed the former in grain-row number by 10.8 per cent. 

The reason for this reversal is easily recognized when we consider that 
parents were selected in these studies for their grain-row numbers, with no 
noticeable selection for heights and yields. In section D of Table 2.6, we note 
that only one parent of any of the Fi families had a grain-row number in 
excess of 18. The crossbred families ranged in parental grain-row numbers 
from 8 to 32. Five of the families came from parents having more than 18 
rows of grains. 

To make a fair comparison between the two types of breeding in their re- 
lation to grain-row number, it is necessary to use only the crossbred families 
having parents with no more than 18 grain-rows. When we make such a limi- 
tation, we find the average grain-row number for the remaining six crossbred 
families is only 12.9. The grain-row average for the six Fi families, namely, 
15.2, exceeds the crossbreds by 17.1 per cent. Limiting the other indicators 
of physiological vigor to the same six crossbred families, we find that the F/s 
exceed the corresponding crossbreds on the average by 6.3 per cent in height 
of stalks and 7.0 per cent in yield of ear-corn. 

In 1911 I was again in full personal charge of the corn experiments at 
the Station for Experimental Evolution, and was able to expand the work 
considerably, both quantitatively and in the types of matings studied. 
We planted 84 cultures in the white dent series as well as 25 cultures of 
other types of corn. The total number of white dent ears of which the grain- 
rows were counted was 6,508 which showed the following frequencies: 

Grain-rows 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 Total 

Frequencies 267 767 1725 1298 931 683 363 164 114 95 65 23 7 3 3 6508 

Percentages 4.1 11.8 26.5 19.9 14.3 10.5 5.6 25 1.8 1.5 0.9 0.4 0.1 0.1 0.1 99 9 

In Table 2.7 the 1911 results are presented in condensed form. Families 
are grouped in eleven sections representing fairly homogeneous groups, 
mostly based on the types of matings involved. Sections D and E are both 
made up of the same five families of F2 hybrids produced by selfing the same 
number of different Fi's. For these families each seed ear was used to plant 
two rows. The one row of each such family was grown with the other cultures, 
as usual, in the East Garden. The second row of each of these families was 



TABLE 2.7 

AVERAGE GRAIN-ROW NUMBERS AND YIELDS PER ACRE OF WHITE 

DENT MAIZE GROWN IN 1911 GROUPED ACCORDING TO THE 

TYPES OF MATINGS OF THE PARENTS 



Pedigree 
Numbers 


Parental Strains Involved 


Number 
of Stalks 


Av. Num- 
ber Grain- 
Rows 


Weights 
in Lbs. 


Yields 
Bu./A. 




(A) Families from Inbreds Selfed 


FI6.681 

E2.68.2 

F29.70 

F32.73 

F34.76 

F0.77 

E19.79i 

F47.792 

F0.80 

E24.82 

F56.85 

E36.92 

F74.94 


8 selfed 

8 selfed 

10 selfed 

12 selfed 

Strain A selfed 

A from L. H. Smith 

B selfed 

B selfed 

B from L. H. Smith 

16 selfed 

20 selfed 

26, 28 selfed 

*"Cobs" selfed 


12 
44 
89 
95 
98 
101 
3 
46 
95 
84 
90 
79 
64 


8.7 1.5 17.9 

9.0 6.0 19.5 

10.9 16.5 26.5 

11.8 11.3 16.9 

8.4 8.3 12.1 

8.9 8.8 12.4 

Not counted nor weighed 

Not counted nor weighed 

14.3 4.3 6.8 

14.0 7.5 12.8 

15.3 13.8 21.8 

22.7 11.5 20.8 

Not counted nor weighed 

1 




Unweighted averages (omit- 
ting the three uncounted 
families) 


78.7 


12.4 


8.9 


16.7 




(B) Families from Parents Given Mixed Pollen in Each 
Generation; Selfing Prevented 


F23.69 

F31.72 

F46.78 

F53.81 

F55.84 

F632.86 

F66.87 

F70i.91 

F73.93 

F76.96 


8 crossed 
10 crossed 
12 crossed 
14 crossed 
16 crossed 
18 crossed 
20 crossed 
22 crossed 
24 crossed 
32 crossed 


71 
95 
92 
97 
101 
105 
99 
63 
68 
94 


10.4 
10.7 
12.2 
13.7 
15.2 
18.2 
19.4 
22.3 
23.8 
25.2 


30.3 
30.3 
44.5 
40.8 
33.0 
42.5 
40.0 
20.8 
34.5 
50.5 


60.2 

45.5 
69.1 
60.0 

46.7 
51.8 
57.7 
45.9 
72.5 
60.4 




Unweighted averages 


88.5 


17.0 


36.7 


57.0 




(C) Fi Hybrids between Different Inbreds 


F29.71 

F32.74 

F32.75 

F54.83 


(10Xl2)Fi 
(lOXB)Fi 
(10Xl6)Fi 
(16X20)Fi 


62 
106 
100 
100 


12.2 
12.8 
14.3 
18.4 


24.5 
65.3 
63.0 
58.2 


56.5 
87.9 
90.0 
83.2 




Unweighted averages 


92 


14.4 


52.7 


79.4 



♦ This was a slightly fasciated brevistylis type, with silks about half as long as the husks. Usually it pro- 
duced no grains except when given artificial help. 



38 



TABLE 2.7— Contimie 



Pedigree 

Number 


Parental Strains Involved 


Number 
of Stalks 


Av. Num- 
ber Grain- 
Rows 


Weights 
in Lbs. 


Yields 
Bu./A. 




(D) F2 Families from Fi Selfed, Clrown in Annex No. 1 


F21.24 

F22.28 

F36.31 

F37.36 

F58.54 


(8X20)F, selfed 
(8XB)F, selfed 
(A XlO)Fi selfed 
(A XB)Fi selfed 
(20 XI 6) Fi selfed 


69 
61 
99 
93 
103 


13.8 
13.4 
11.3 
11.8 
16.2 


23.0 
31.3 
33.3 
17.0 

54.3 


47.6 
73.2 
48.0 
29.3 

47.5 




Unweighted averages 


83 


13.3 


31.8 


49.1 




(E) Same Families as in (D), but Grown in East Garden 


F21.24 

F22.28 

F36.31 

F37.36 

F58.54 


(8 X20)Fi selfed 
(8XB)Fi selfed 
(AX 10)F, selfed 
(A XB)F, selfed 
(20 Xl6)Fi selfed 


98 

101 

98 

76 

97 


13.4 
13.4 
11.1 
11.0 
16.8 


36.0 
56.0 
31.3 
15.3 
34.3 


52.5 
79.2 
45.9 
28.7 
50.8 




Unweighted averages 


94 


13.2 


34.6 


51.4 




(F) F2 Families from FiXsibs, All Grown in East Garden 


F21.25 

F22.29 

F36.34 

F37.37 

F58.55 


(8X20)FiXsib 

(8XB)FiXsib 

(AXlO)FiXsibs 

(AXB)F,Xsib 

(20Xl6)FiXsib 


59 
97 
93 
71 
110 


12.9 
12.8 
10.8 
11.3 
16.0 


22.0 
42.8 
26.3 
18.5 
35.0 


53.3 
63.0 
40.3 
37.2 
45.5 




Unweighted averages 


86 


12.8 


28.9 


47.9 




(G) Fs Families from F2 Selfed 


F38.39 

F40.42 

F42.45 

F44.46 

F49.49 

F51.52 

F59.57t 

F59.57 

F61.59 

F64.62 


(A X20)F., selfed 
(AX22)F.2 selfed 
(A Xl6)Fo selfed 
(A XB)F.. selfed 
(16XA)F2 selfed 
(16 X20)Fo selfed 
(20 Xl6)Fo selfed 
(20X16)F2 selfed 
(20 XA)F2 selfed 
(BX16)F.2 selfed 


84 
108 

67 

92 
112 

95 
100 
100 
117 
107 


13.0 
11.6 
10.2 
11.0 
11.4 
15.0 
15.9 
16.4 
12.0 
17.0 


9.8 
19.3 

10.5 

6.0 

24.3 

23.8 

24.5 

25.5 

9.8 

12.5 


16.6 
25.5 
22.4 
9.3 
30.9 
35.7 
35.0 
36.4 
13.6 
16.7 




Unweighted averages 


98.2 


13.3 


16.6 


24.2 



t This family was divided and this section was grown in the North Hill-field. All of the other families were 
grown, as usual, in East Garden. 



39 



TABLE 2.1— Continued 



Pedigree 
Numbers 


Parental Strains Involved 


Number 
of Stalks 


Av. Num- 
ber Grain- 
Rows 


Weights 
in Lbs. 


Yields 
Bu./A. 




(H) Fs Families from FiXSibs 


F38.40 

F40.43 

F44.47 

F49.50 

F59.58 

F61.60 

F64.63 


(AX20)F2Xsib 

(AX22)F2Xsib 

(AXB)F,Xsib 

(16XA)F2Xsib 

(20Xl6)F.2Xsib 

(20XA)F,Xsib 

(BXl6)F2Xsib 


106 
112 

94 
104 

90 
111 
104 


13.5 
11.9 
11.2 
11.8 
16.5 
13.8 
15.1 


26.0 
26.5 
21.8 
29.8 
38.5 
25.0 
27.5 


35.0 
33.8 
33.1 
40.9 
61.1 
32.2 
37.8 




Unweighted averages 


103 


13.4 


27.9 


39.1 




(I) Families from "Three-Way" and Iterative Crosses 


F58.56 

F74.95 

F21.27 

F22.30 

F36.33 

F27.38 

F51.53 


(20Xl6)FiX22 

"Cobs"X(20Xl6)Fi 

(8X20)FiX20 

(8XB)F,XB 

(AXB)FiXA 

(AXB)F,XB 

(16X20)F2X20 


114 
29 
67 

103 
84 
79 

108 


18.9 
20.6 
15.0 
14.3 
10.5 
12.8 
17.1 


61.8 
23.3 
28.5 
37.8 
23.0 
23.5 
42.3 


77.4 
114.5 
60.8 
52.4 
39.1 
29.8 
55.9 




Unweighted averages (three- 
way) 


71.5 


19.7 


42.5 


96.0 




Unweighted averagesj (iter- 
ative) 


83.3 


13.1 


28.2 


45.5 




(K) Families from "Four-Way" Crosses, the So-called "Double-Cross"' 


F21.26 

F36.35 

F69.66 

F36 32§ 


(8X20)F,X(AXlO)Fi 
(AX10)F,X(20X16)F, 
(22X"Cobs")F,X(8XlO)Fi 
(AXlO)FiX(AXB)Fi 


67 
106 

75 
102 


12.7 
12.8 
16.3 
11.2 


28.5 
47.0 
58.5 
45.5 


60.8 

63.3 

111.4 

63.7 




Unweighted averages 


87.5 


14.3 


44.9 


74.8 




(L) F3 Families from Four-Way F2 Crosses, and Imperfect 
Iteratives of Same Form 


F61.61 

F38.41 

F40.44 

F44.48 

F49.51 


(20XA)F2X(BX16)F2 

(AX20)F..X(AX22)F2 

(AX22)F2X(AX16)F2 

(AXB)F2X(16XA)F2 

(16XA)F2X(16X20)F2 


102 
103 
110 

78 
117 


15.3 
12.9 
13.2 
11.4 
13.3 


31.8 
27.0 
43.5 
28.0 
44.3 


44.5 
37.5 
56.5 
51.3 
61.6 




Unweighted averages 


102 


13.2 


34.9 


50.3 



X Does not include F51.53. 

§F36.32 is an imperfect 4-way, being partly iterative, involving only 3 inbreds. 



40 



BEGINNINGS OF THE HETEROSIS CONCEPT 41 

planted in new plots of ground about one-fourth mile north of the original 
Station grounds. 

The purpose of this replication was to determine the degree of consistency 
of results secured in these new locations with those recorded for the cultures 
grown in the different conditions of soil, drainage, exposure, lighting, etc., in 
the East Garden. Summaries of these two sections of Table 2.7 show the cul- 
tures grown in the new plot with average grain-row number 1.29 per cent 
higher than in the same families grown in the East Garden. However, the 
East Garden cultures produced a higher average yield of ear-corn by 4.70 
per cent. 

Comparison between selfing and sibcrossing was made a subject of special 
study in the inbred and Fi families in 1910. This was not continued in 1911 
in the inbreds, but was given a further test in the derivation of the F2 families 
from the Fi, and was carried forward to the derivation of F3 families from the 
F2. These comparisons as they relate to Fi families are given in sections E 
and F of Table 2.7. They show the F2 families derived from selfing their Fi 
parents slightly superior to those Fo families produced from sibcrosses in 
the Fi. This is indicated by an average grain-row number 3.1 per cent higher 
and average yield 7.5 per cent higher in the Fo families from selfed Fi par- 
ents, thus reversing the indications from the 1910 cultures. 

The comparison of selfing versus sibcrossing in the production of the F3 by 
these two methods of breeding in F2 can be derived from section G for selfings 
and section H for the sibcrosses. Summaries of these two sections show a 
superiority from sibcrosses of 0.4 per cent in average grain-row number and 
61.6 per cent in yield. A part of this discrepancy is clearly due to the inclu- 
sion of families in the selfed group which had no direct counterpart in the 
sibcrossed group. If we limit the comparison to the families which are repre- 
sented in both groups, we can avoid this cause of distortion. We then find 
the sibcrossed families superior to the selfed by 1.5 per cent in grain-row 
number, and 48.6 per cent in yields. 

Comparative values between inbreds and crossbreds, as shown in sections 
A and B of Table 2.7, and between crossbreds and Fi hybrids, are essentially 
the same as in 1910. The ratios of inbreds, crossbreds, and Fi hybrids, with 
respect to yields, is 0.29 to 1.00 to 1.22. Again the average grain-row number 
is less in the Fi than in the crossbreds, and for the same reason. This particu- 
lar group of Fi families came from parents with low average grain-row num- 
bers, as compared with the broader parentage of the crossbreds. 

The relationship of F.-s to Fo can now be noted by comparing the results 
in sections G and H of Table 2.7, with sections D, E, and F. There are sev- 
eral ways in which such comparisons can be made. Perhaps as good a way 
as any is simply to combine all of the F2's together, regardless of the con- 
siderations which led these to be tabulated in three separate sections, and 
compare the results with all the F3 families of sections G and H likewise 



42 GEORGE HARRISON SHULL 

averaged in an undivided population. When treated in this way, we find that 
the Fo's have an average grain-row number of 13.1 and average yields of 49.5 
bushels per acre, while the F3 had an average of 13.4 grain-rows and pro- 
duced an average of 30.4 bushels per acre. If we associate the average yield 
of the Fi families, 79.0 with these values for Fo and F3, we see the beginning 
of the characteristic curve in which the loss of yield from one generation to 
the next is about twice as great as the loss for the next following generation . 
It remains to consider the last three sections of Table 2.7, in which are 




Fig. 2.6 — Total yields of ear corn of two selfed strains, Strain 16 and Strain 20, in the fore- 
ground (exaggerated, of course, by foreshortening), and their Fi, F2, and F3 hybrids, left to 
right, successively, in the background. As may be seen in Table 2.7, these yields, calculated 
in terms of bushels per acre, are 12.76 and 21.82 for the two inbreds, and 83.21, 50.81, 

and 36.43 for the three hybrid families. 

included the results of more complicated crossing which had become possible 
through the accumulation of simpler crossing in preceding years. In section I 
are given two "three-way" crosses and four iterative crosses involving Fi 
combinations and one iterative cross involving an F2 combination, each repre- 
senting a cross between a hybrid and an inbred. As might be e.xpected, these 
seven families although similar in form show no special consistency, since 
they involve various combinations of five different inbreds and five different 
hybrids. 

In Table 2.7, section K, are presented what I believe to be the first "four- 
way" or so-called "double crosses" ever made among inbreds. The elements 
of one of these double crosses are shown in Figure 2.7. These double crosses 
were made some five or six years before Dr. D. F. Jones pointed out the 
potentialities of such crosses in producing hybridized seed corn at a price 




X 





X 




10 





(Ax 10) F, 




Fig. 2.7 — One of the first four-zcay or double crosses ever grown from selfed strains of maize. 
The single crosses for this double cross were made in 1909, the cross between the Fi's was 
made in 1910, and the double-cross ear at bottom (G35.62) was grown in 1911 and grains 

from it were used for i)lanting in 1912. 



44 GEORGE HARRISON SHULL 

that could make the pure-line method of corn production practical. No credit 
is sought for the fact that I made these four-way crosses some years prior to 
the similar combinations made by Dr. Jones. They are presented here only 
because they belong in a historical account. 

In the last section of Table 2.7 I have entered five families which have the 
form of four-way crosses, but in which the single crossings used were Fo in- 
stead of Fi. Only the first of these five families actually involved four differ- 
ent inbreds, the others being partially iterative, in that only three inbreds 
contributed to each. A comparison of the double crosses both of Fi and F2, 
with the corresponding single crosses, is instructive. Comparison of the sum- 
mary of section C with that of section K shows the double cross families 
slightly inferior to the single cross families, as indicated by a 1 per cent higher 
grain- row number and 6 per cent higher yield of the single cross families 
over the double cross. Comparing sections L and E, it is to be noted that the 
double cross retains the vigor of the F2, instead of declining to the vigor of 
the F3 families produced by the usual methods, as seen in sections G and H, 
Table 2.7. 

In 1911 1 realized that the effective exposition of the important discoveries 
we were making required photographs of prepared exhibits. A number of such 
exhibits were set up and photographed, and have been presented in lantern 
slides on many occasions. I have included the most instructive of these here. 

Here the detailed account of these studies must end, for although they 
were continued in 1912, 1 have been unable to locate the field and harvesting 
notes including grain-row counts and weighings for the 1912 cultures. These 
1912 cultures were especially designed to explore the evidences of Mendelian 
segregations in the F2 and the F3 families, with respect to grain-row num- 
bers and yields. They included 11 families of the breeding Fi X self, 8 families 
of Fi X sib, 21 F2 X self, 10 Fo X sibs, and five families of F3 X self. There 
was also an interesting pair of approximations to eight-way combinations or 
quadruple crosses produced by reciprocal combinations of the four-way 
crosses included in the 1911 cultures. While these had the form of quadruple 
crosses, they were imperfect in that one of the inbreds was repeated, so that 
only seven different inbreds were represented, instead of eight. This was in- 
evitable since I initiated only seven inbred lines in the beginning of these 
experiments. 

The 1912 crop completed the experimental work with corn at the Station 
for Experimental Evolution, and I spent the next year in Berlin, Germany. 
In a lecture I gave at Gottingen about three weeks before the beginning of 
the first World War the word heterosis was first proposed. I used the occasion 
to discuss the bearing of the results of these studies on the practical work of 
breeders of various classes of organisms, both plant and animal. I stressed 
the point that the breeder should not be content, as had long been the case, 
to seek merely to avoid the deterioration incident to inbreeding, but should 



10 



18 



10 



13 



1904 1905 1906 1507 1908 1909 I'JiO igu 191.5 




SELECTION FOR ROW-NIJHBER 
IN CROSS-FERTILIZED MAIZE 



"■?58=^ 



^;2^"- 



»-^. 



;o>*\ 




SELECTIOtJ FOR RaJ-WIKDER 
IN SELF-FEHTILIZEU MIZE 



1904 



1905 



1906 



1907 



1908 



1909 



1910 



1911 



1912 



Fig. 2.8 — Diagrams of the progressive results of selection for grain-row number under the 
two systems of breeding: selling completely prevented in the upper diagram; selling the 
sole method of breeding in the lower. The numbers on the lines indicate the numbers of 
rows of grains on the parent ears. The circles show by their position on the scale at left the 
average grain-row numbers of the resulting [progenies. 



46 



GEORGE HARRISON SHULL 



recognize in heterosis a potent source of practical gains, to be investigated, 
understood, and utilized as a new tool in deriving from plant and animal 
life their maximum contributions in the service of man. 

Although no further experimental work was done with corn at the Station 
for Experimental Evolution after 1912, I tried to resume the work in my 
first two years at Princeton University, by planting 77 cultures of pedigreed 

8 



V, 7 




u 




■; i 




& 




^3^ 






«» 




V * 


f 


V- 




♦ • < 


3 






*" 


z 






^H*-^' 




V, 



'^ /' %fi> X ^h QA >S^ 




A 



^^^ 



3 /^ X B 




S 3. 



Fig. 2.9 — Ears of my white dent "strain" of corn grown at Princeton University in 1916. 
The ears, each typical of the progeny to which it belonged, are from left to right : SA, Shull's 
Strain A; SA X BA, Fi hybrid between Shull's Strain A and Biakeslee's "branch" of the 
same strain; BA X SA, reciprocal of the last; BA, Shull's Strain A, after two successive 
sellings by Dr. A. F. Blakeslee; B.\ X B, Fi between Biakeslee's branch of Strain A and 
Shull's Strain B; and SB, Shull's Strain B. About as much heterosis is shown by a cross be- 
tween two sub-hnes of Strain A as between one of these sub-lines and Strain B, the impli- 
cation being that something more specific may be involved in this example of heterosis than 
the mere number of genetic difTerences. (Photo by W. Ralph Singleton in 1945.) 

corn in 1916 and 65 in 1917. 1 used some of the materials from these cultures 
for laboratory studies in biometry in my classes in genetics. The interesting 
results shown in Figure 2.9 are from my 1916 crop at Princeton. The plantings 
at Princeton were made late and the young plants were decimated by pigeons 
and crows, so that some valuable connections were lost, and with them some 
of my interest in their continuation. 

As we all know, heterosis is not limited to corn, and my own interest in 
the matter was in no wise restricted to its manifestation in corn. There were 
examples presented in many other of my genetical experiments. I was par- 
ticularly interested in the discovery of such special mechanisms as balanced 
lethal genes in the Oenotheras and self-sterility genes in Capsella grandiflora 



BEGINNINGS OF THE HETEROSIS CONCEPT 



47 



which, along with many types of asexual reproduction including partheno- 
genesis, specifically enable the organisms possessing these special mecha- 
nisms to maintain the full advantages of heterosis. On one occasion, one of 
my new hybrid combinations in Oenothera happened to be planted through 
an area in my experimental field where the soil had become so impoverished 
that none of my other cultures reached their normal growth. Many of the 




Fig. 2.10 — The Fi hybrids between a cultivated form of Helianthus anniius and a wild form 
of the same species received from Kansas. This photograph, taken at the Station for Ex])eri- 
mental Evolution in 1906, shows the author alfixing a glassine hag to a head of one of the 
hybrid plants. The two parents of this hybrid averaged from 5 to 6 feet tall, while 51 of 
these Fi hybrids, measured on August 28, 1906, ranged in height from 6.7 to 14.25 feet, the 
average being 10.46 feet. This may be considered my iirst experience with hybrid vigor. 



48 GEORGE HARRISON SHULL 

plants remained rosettes or formed only weak depauperate stems. But this new 
hybrid became a vigorous upstanding form in this impoverished area as well 
as on better soil elsewhere. I recorded this as a notable example of making 
heterosis take the place of manure or commercial fertilizers. 

Figure 2.10 is a notable hybrid, which represents my first direct personal 
contact with a recognized case of hybrid vigor. This hybrid resulted from a 
cross I made in 1905 between the so-called "Russian" sunflower and the wild 
Helianthus annuus of our western prairies. Both of these forms have been re- 
ferred, botanically, to the same species. Both are of approximately equal 
height, scarcely as tall as the six-foot step-ladder shown in the figure. The 
tallest of these Fi hybrids was 14.25 feet in height. 

Returning now to the question which I sidestepped in the beginning — 
what we mean by the expression the heterosis concept — I suggest that it is the 
interpretation of increased vigor, size, fruitfulness, speed of development, 
resistance to disease and to insect pests, or to climatic rigors of any kind, 
manifested by crossbred organisms as compared with corresponding inbreds, 
as the specific results of unlikeness in the constitutions of the uniting parental 
gametes. 

I think the first clear approach to this concept was involved in a statement 
which I have already quoted, that "a different explanation was forced upon 
me" (in my comparisons of cross-fertilized and self-fertilized strains of 
maize). That is, "that self-fertilization simply serves to purify the strains, 
and that my comparisons are not properly between cross- and self-fertiliza- 
tion, but between pure strains and their hybrids." Since heterosis is recog- 
nized as the result of the interaction of unlike gametes, it is closely related to 
the well known cases of complementary genes. It differs from such comple- 
mentary genes, however, mainly in being a more "diffuse" phenomenon in- 
capable of analysis into the interactions of specific individual genes, even 
though it may conceivably consist in whole or in part of such individual 
gene interactions. 



H. K. HAYES 

University of Minnesofa 



Chapter 3 

Development of 
the Heterosis Concept 



Hybrid vigor in artificial plant hybrids was first studied by Koelreuter in 
1763 (East and Hayes, 1912). The rediscovery of Mendel's Laws in 1900 
focused the attention of the biological world on problems of heredity and led 
to renewed interest in hybrid vigor as one phase of quantitative inheritance. 

Today it is accepted that the characters of plants, animals, and human 
beings are the result of the action, reaction, and interaction of countless 
numbers of genes. What is inherited, however, is not the character but the 
manner of reaction under conditions of environment. At this time, when 
variability is being expressed as genetic plus environmental variance, one 
may say that genetic variance is the expression of variability due to geno- 
typic causes. It is that part of the total variance that remains after eliminat- 
ing environmental variance, as estimated from studying the variances of 
homozygous lines and Fi crosses between them. 

Early in the present century. East, at the Connecticut Agricultural Ex- 
periment Station, and G. H. Shull at Cold Spring Harbor, started their 
studies of the effects of cross- and self-fertilization in maize. The writer has 
first-hand knowledge of East's work in this field as he became East's assist- 
ant in July, 1909, and continued to work with him through 1914. In 1909, 
East stated that studies of the effects of self- and cross-pollination in maize 
were started with the view that this type of hiformation was essential to a 
sound method of maize breeding. In addition to studies of maize, which is 
normally cross-pollinated, East carried out studies in tobacco of crosses be- 
tween varieties and species. This gave an opportunity of studying the effects 
of self- and cross-pollination with a self-pollinated jilant. A 1912 j)ublication 
of East and Hayes made the following statement: 

The decrease in vigor due to inbreeding naturally cross-fertilized species and the increase 
in vigor due to crossing naturally self-fertilized species are manifestations of one phenome- 
non. This phenomenon is heterozygosis. Crossing produces heterozygosis in all characters 
by which the parent plants differ. Inbreeding tends to produce homozygosis automatically. 

49 



50 H. K. HAYES 

Several photographs from this bulletin are of some interest. A picture of 
two inbred lines of maize and their Fi cross was one of the first published field 
views of hybrid vigor from crossing inbred lines of maize. East told me that 
such a demonstration of hybrid vigor would create a sensation if the material 
had been grown in the corn belt. 

Some Fi crosses between species and sub-species in tobacco gave large in- 
creases in vigor. Some species crosses were sterile. Some varietal crosses 
within species showed little or no increase in vigor, other crosses gave an aver- 
age increase of 25 per cent in height over the average of their parents. A few 
wide species crosses were very low in vigor. One such cross beween Nicotiana 
tabacum and Nicotiana alata graiidiflora was sterile and very weak in growth. 
Photographs of the parents and hybrids bring out the fact that a lack of vigor 
in a few cases was known to accompany the heterozygous condition. Natural- 
ly such undesirable combinations had little importance either to the plant 
breeder or as a basis for evolution. 

In 1910, G. H. ShuU summarized the effects of inbreeding and crossbreed- 
ing in maize in a clear, concise, and definite manner. The student of heredity 
in this early period had little conception of the complexity of inheritance. 
Hybrid vigor was in many cases not clearly Mendelian. The term heterosis 
was coined by Shull and first proposed in 1914. He used the term to avoid 
the implication that hybrid vigor was entirely Mendelian in nature and to 
furnish a convenient term to take the place of such phrases as "the stimulus 
of heterozygosis." 

At this time it was usually stated that increased vigor in hybrids was due 
to a more rapid cell division as stimulated by the heterozygous condition of 
the genotype. A. F. Shull in 1912 attributed the vigor "to the effect of a 
changed nucleus and a (relatively) unaltered cytoplasm upon each other." 

The purpose of this chapter is to discuss some phases of the development 
of the heterosis concept since 1910. Three main topics will be presented cover- 
ing utilization, breeding methods, and genetic concepts with particular ref- 
erence to practical applications and to genetic explanations. 

UTILIZATION OF HETEROSIS BY THE PRODUCER 

The presentation of East and Hayes in 1912 emphasized the probable 
practical value of heterozygosis. A review of experiments with maize was 
made. In discussing Shull's (1909) plan for the use of single crosses between 
inbred lines, it was stated that the procedure was desirable in theory but 
difficult of application. At this early time the inbred lines of maize that were 
available seemed so lacking in vigor that the use of Fi crosses between selfed 
lines in maize for the commercial crop seemed impractical. Both Shull and 
East believed that some method of direct utilization of hybrid vigor in maize 
would be found. 

One is inclined to forget that the inbred lines of maize of today are marked- 



DEVELOPMENT OF THE HETEROSIS CONCEPT 51 

ly superior, on the average, to those of 1910. Jones's discovery about 1917 
of the double cross plan of producing hybrid seed in maize, and the subse- 
quent proof by many workers that double crosses can be obtained that closely 
approach the vigor of Fi crosses between selfed lines, furnished the basis for 
the utilization of hybrid vigor in field corn. With sweet corn, however, F] 
crosses between selfed lines are used very widely today for the commercial 
crop. 

East and Hayes emphasized that Fi crosses probably would be of com- 
mercial value in some truck crops where crossing was easy. Eggplants, to- 
matoes, pumpkins, and squashes were considered to offer promise for a prac- 
tical use of such vigor. The writers also mentioned the fact that heterozygosis 
had been used in vegetatively propagated plants, though not purposely, and 
that it seemed feasible to make a practical application in the field of forestry. 

The use of heterosis in practical plant and animal improvement has borne 
out and surpassed these early predictions as shown in Table 3.1. 

TABLE 3.1 

USE OF HETEROSIS IN CROP PLANTS AND LIVESTOCK 

Farm crops: Maize, sugar beets, sorghums, forage crops, and grasses 

Horticultural crops: Tomatoes, squashes, cucumbers, eggplants, onions, 

annual ornamentals 
Silkworms 

Livestock: Swine, poultry, beef and milk cattle 

Vegetatively propa- 
gated plants 

In the corn belt of the United States nearly 100 per cent of all maize is 
hybrid. Hybrid corn is rapidly being developed in other countries of the 
world, and is one of the best illustrations of the practical utilization of mod- 
ern genetics. Considerable evidence leads to the conclusion that heterosis can 
be used extensively in farm crops, including such widely different plants as 
sugar beets, sorghums, tobacco, forage crops, and grasses. 

With horticultural plants, where the individual plant is of rather great 
value, planned heterosis has proven worth while. First generation crosses 
of tomatoes, onions, egg plants, cucumbers, and squashes have proven their 
value and are being grown extensively by home and truck gardeners. Similar 
use is being made of heterosis in some annual ornamentals. 

Heterosis has become an important tool of the animal breeder. Its use in 
silkworm breeding is well known. Practical utilization of hybrid vigor has 
been made in swine and poultry, and applications are being studied with beef 
cattle, dairy cattle, and sheep. A somewhat better understanding of the 
effects of inbreeding and crossing by the breeder has aided in applications 
with livestock. As in plants, inbreeding makes controlled selection possible, 
while controlled crosses may be grown to utilize favorable gene combinations. 



52 H. K. HAYES 

METHODS OF BREEDING FOR HETEROSIS 

In general there is a much closer relation between the characters of par- 
ents and of their Fi crosses in self-pollinated plants than between the char- 
acters of inbred lines of cross-pollinated plants and their Fi crosses. 

Characters of Parents and Fi Crosses in Self-pollinated Plants 
A recent study by Carnahan (1947) in flax, which is normally self-polli- 
nated, may be used for illustrative purposes. Four varieties of flax were se- 
lected to represent desirable parental varieties. Each was crossed with four 
other varieties, of different genetic origin from the first group, to be used as 
testers. Sufficient seed for Fi and F2 progenies was produced so that all 

TABLE 3.2 

PARENT AND Fi CROSSES, YIELD 
IN BUSHELS PER ACRE* 

Parent Tester Varieties 

Varieties 5 6 7 8 

16 14 17 13 



1 


19 


31 


25 


22 


19 


2 


18 


24 


26 


19 


20 


3 


13 


26 


24 


20 


18 


4 


17 


22 


21 


20 


19 



* Parent yields outside rectangle, Fi crosses 
within. 

progenies could be planted in replicated, 8-foot rows at the rate of 200 seeds 
per row. Combining ability was studied in Fi and F2 in comparison with the 
parents for yield of seed, number of seeds per boll, number of bolls per plant, 
weight of 1000 seeds, date of full bloom, and plant height. 

As shown in Table 3.2, each Fi cross yielded more than its highest yielding 
parent, although for one cross the difference was only slightly in favor of 
the Fi. For an average of all crosses, the Fi yielded 40 per cent more than the 
average of the parents, and the Fo, 26 per cent more. The lowest yielding 
cross, 3X8, was produced from a cross of the two lowest yielding parents. 
The highest yielding cross, 1X5, however, could have been selected only 
by actual trial. It was obtained by crossing the highest yielding selected 
variety with the second highest yielding tester variety. 

There was excellent agreement, on the average, for each of the characters 
studied between the average expression of the characters of the parents and 
their Fi crosses. Carnahan concluded that for each character studied there 
appeared to be a good relationship between the performance of the parents 
and the average performance of their Fi crosses. The characters of the par- 
ents in this study were as good or better indication of the combining ability 
of a parental variety as that obtained from a study of average combining 
ability in four crosses. 



DEVELOPMENT OF THE HETEROSIS CONCEPT 



53 



Powers (1945) obtained also relatively good agreement in tomatoes be- 
tween the parental yield of 10 varieties and that of all possible Fi crosses 
between the 10 varieties (see Table 3.3). 

Moore and Currence (1950) in tomatoes made a somewhat comparable 
study to that of Carnahan with flax. They used two three-way crosses as 
testers for a preliminary evaluation of combining ability of 27 varieties. 
Based on this, eight varieties were selected that gave a wide range in aver- 
age combining ability for sev^eral characters including early yield and total 
yield. These varieties were crossed in all combinations, and yield trials of the 

TABLE 3J 

YIELD OF RIPE FRUIT IN GRAMS 
IN TOMATOES (AFTER POWERS) 





Yield of Ripe Fruits (per Plant) 


Variety or Inbred 








Variety or Inbred 


9 Crosses (av.) 




Grams 


Grams 


L. escidsntum 






Bounty 4101 


513+ 39 


1280 + 53 


4102 


607+ 86 


1267 + 46 


4105 


332+ 64 


1081+33 


4106 


828 + 108 


1236 ±45 


Es.XL. pirn 






4103 


1066+159 


1597 + 54 


4104 


808 + 114 


1340 + 44 


4107 


801 + 111 


1181+47 


4108 


857 + 108 


1192+41 


4109 


1364+151 


1968 + 46 


4110 


1868 + 149 


2231 + 52 



varieties and Fi crosses were made. There was relatively good agreement 
between the early test for combining ability and the average yield of Fi 
crosses, but the relationship did not seem superior to the varietal performance 
as a means of predicting combining ability in crosses. In the studies by Carna- 
han, Moore, Currence, and Powers the only means of selecting the most de- 
sirable Fi cross was by actual trial. 

Characters of inbred Lines and Their Fi Crosses in Maize 

Numerous studies have been made with maize of the relation between 
characters of inbred lines and of their Fi crosses. There usually have been 
indications of significant correlations for most characters of inbred lines and 
their Fi crosses. In most cases, however, the relationship was not very large 
or highly important when one studied individual characters, or the more com- 
plex character — yield of grain. The studies have been reviewed by numerous 
workers (see Sprague, 1946b). 



54 H. K. HAYES 

Hayes and Johnson (1939) in Minnesota studied the relation between the 
characters of 110 inbred lines of maize and their performance in top crosses. 
The characters studied in selfed lines in replicated yield trials are given in 
Table 3.4. 

All possible correlations were made between the individual characters of 
the inbreds and of these characters and the yield of grain of top crosses. The 

TABLE 3.4 

CHARACTERS OF 110 INBRED LINES IN 
CORN CORRELATED WITH INBRED- 
VARIETY YIELDING ABILITY 

1. Date silked 7. Stalk diameter 

2. Plant height 8. Total brace roots 

3. Ear height 9. Tassel index 

4. Leaf area 10. Pollen yield 

5. Pulling resistance 11. Grain yield 

6. Root volume 12. Ear length 

TABLE 3.5 

TOT.\L CORRELATIONS BETWEEN CHARACTERS OF 110 INBREDS, 
LABELED 1 TO 12, AND YIELDING ABILITY OF INBRED- 
VARIETY CROSSES DESIGNATED AS 15 



Characters Correlated 




2 3 4 5 


6 


7 


8 


9 


10 


11 


12 


15 


1 


0.51 0.61 0.48 0.65 


0.62 


0.55 


0.38 


0.37 


0.22 


0.07 


-0.06 


0,47 


2 


0.76 0.44 0.48 


0.43 


0.40 


0.26 


0.19 


0,36 


0.25 


0.08 


0.27 


3 


0.43 0.54 


0.50 


0.41 


0.35 


0.33 


0.22 


0.15 


-0.01 


0.41 


4 


0.50 


0.44 


0.48 


0.40 


0.29 


0.18 


0.20 


0.08 


0.29 


5 




0.76 


0.51 


0.60 


0.41 


0.21 


0.15 


0.04 


0.45 


6 






0.55 


0.74 


0.39 


0.29 


0.19 


0.03 


0.54 


7 








0.54 


0.24 


0.27 


0.21 


0.15 


0.41 


8 


Multiple value of R 








0.26 


0.22 


0.20 


0.07 


0.45 


9 


for inbred-variety yield 










0.20 


-0.00 


0.03 


0.19 


10 


and twelve characters of 












0.35 


0.32 


0.26 


11 


inbred = 0.67 














0.64 


0.25 


12 


















0.28 



Significant value of r for P of .05 = 0.19. 
Significant value of ;■ for P of .01 = 0.25. 

characters, in general, were those that were considered to evaluate the in- 
breds in developmental vigor. 

The total correlations between characters are summarized in Table 3.5. 
Most correlations were significant at the 5 per cent or 1 per cent point ex- 
cept the relation between ear length and other characters of the inbreds. All 
relationships between the characters of the inbreds, including grain yield, and 
the yield of top crosses were significant at the 1 per cent point except for 
tassel index of the inbreds, and that was significant at the 5 per cent point. 
The multiple correlation coefficient of 0.67 indicated that under the condi- 
tions of the experiment about 45 per cent of the variability of inbred-variety 



DEVELOPMENT OF THE HETEROSIS CONCEPT 



55 



yield was directly related to characters of the inbreds. These relationships 
between the parents and their Fi crosses were somewhat larger than those 
obtained by others with maize. Nevertheless, relationships were much 
smaller than has been obtained in similar studies with self-pollinated plants. 

Richey (1945b) compared the yield of inbred parents in the S3 and S4 gener- 
ations of selling with the mean yield of their single crosses from data taken 
by Jenkins and Brunson. Similar comparisons were made between the yield 
in top crosses and the mean yield in single crosses (see Table 3.6). 

Although for various reasons the r values are not strictly comparable, the 
yield of inbreds was as strongly correlated with the mean yield of their 
single crosses as the yield in top crosses was correlated with the mean yield 
of single crosses. 

TABLE 3.6 

CORRELATION COEFFICIENTS FOR YIELDS OF 

INBRED PARENTS OR TOP CROSSES WITH 

MEAN YIELDS OF SINGLE CROSSES* 



Hybrids 
Correlated 


Previous Generations 
OF Inbreeding 


WITH 


S3t 


84 


Inbred parents 
Top crosses 


.25, .64, .67 
.53 


.41, .45 
.53 



* After Richey, after Jenkins and Brunson. 
t S3 = three years selfed, etc. 



Comparison of Methods with Self- and Cross-pollinated Plants 

In self-pollinated plants it seems probable that the first natural step in 
the utilization of heterosis normally may consist of the selection of available 
parental varieties that in themselves produce the best combination of char- 
acters. It seems important to continue breeding for the best combination of 
genes that can be obtained in relatively homozygous varieties. Where hybrid 
seed can be produced cheaply enough, or new methods can be found to 
make crosses more easily, heterosis can be used to obtain from the hybrid an 
advance in productivity over the homozygous condition. 

In cross-pollinated plants two general methods of breeding for heterosis 
are now being widely utilized. One consists, as in maize, of the selection with- 
in and between selfed lines and the use of single, three-way, or double crosses 
for the commercial crop. The second general method consists of selecting 
or breeding desirable clones of perennial crops. These are evaluated for com- 
bining ability by polycross, or other similar methods, and the desirable clones 
used to produce Fi crosses, double crosses, or synthetic varieties. 



56 H. K. HAYES 

There seems to be some difference of opinion regarding the selection proc- 
ess in its application to maize improvement. One school of thought practices 
a somewhat similar method of breeding selfed lines as is used in self-pollinated 
plants, with the viewpoint that controlled selection makes it possible to iso- 
late in the inbred lines the genes for characters needed in the hybrids. Ap- 
parently the relationship between the characters of inbreds and their Fi 
crosses will become greater as inbred lines themselves improve. The other 
extreme of viewpoint (Hull, 1945a) is that the greater part of hybrid vigor is 
due to interallelic interaction of genes to such an extent that selection based 
on appearance may be harmful. In a recurrent selection program Hull, 
therefore, does not recommend selection for vigor of growth, although he 
states that plants showing pest or weather damage should be avoided. 

It is probable that differences between these two so-called schools may 
have been overstated. Both believe that the actual test for combining ability 
in hybrid combination is necessary. The stage in the breeding program when 
such test should be made will depend on the material worked with and the 
nature of the breeding program. In both cross- and self-pollinated plants an 
actual trial will be needed to determine the combination that excels in 
heterosis. 

Where clonal lines can be propagated vegetatively, a method of selecting 
for heterosis in alfalfa was suggested by Tysdal, Kiesselbach, and Westover 
(1942), by means of polycross trials. The method is being used extensively 
today with perennial forage crops that normally are cross-pollinated. The 
writer is studying the method with early generation selfed lines of rye. With 
perennial crop plants, selection for combining ability is made for heterozy- 
gous parent clones. Where disease and insect resistance or winter hardiness 
are important, it may be essential to insure that the clones used in the poly- 
cross trials excel for these characters. Polycross seed is produced on selected 
clones under open-pollinated conditions where the clones are planted together 
at random under isolation. 

In one study of progenies of eight clones by Tysdal and Crandall (1948) 
yields were determined from polycross seed in comparison with top cross seed 
when each of the clones was planted in isolation with Arizona common alfalfa 
(see Table 3.7). The agreement for combining ability was relatively good in 
the two trials. 

An early suggestion of utilization of heterosis in alfalfa was by double 
crosses, from single crosses between vegetatively propagated clones, without 
entire control of cross-pollination. Synthetic varieties also have been sug- 
gested as a means of the partial utilization of heterosis. In one comparison 
the progeny of a synthetic combination of four clones of high combining 
ability yielded 11 per cent more forage than a similar combination of four 
clones of low yielding ability. A recent comparison of eight synthetics led 
Tysdal and Crandall to conclude that the first synthetic and second syn- 



DEVELOPMENT OF THE HETEROSIS CONCEPT 



57 



thetic seed progenies gave about the same forage yield. In this comparison, 
heterosis continued through the second seed increase of the high yielding 
synthetic. 

Other Studies with Maize 

Combining ability, that is ability to yield in hybrid combination, has been 
shown by various workers to be an inherited character (Hayes and Johnson, 
1939), (Cowan, 1943), (Green, 1948). It seems feasible to breed for high com- 
bining ability as for other quantitative characters. In the breeding program 

TABLE 3.7 

FORAGE YIELDS OF POLY- 
CROSSES COMPARED TO 
TOP CROSSES OF THE 
SAME CLONES* 





Yield Relative to 




Grimm 


AS 100 


Clone No. 




Arizona 




Polycross 


Top Cross 


1 


121 


130 


2 


111 


122 


3 


101 


117 


4 


'J9 


103 


5 


97 


105 


6 


96 


101 


7 


89 


101 


8 


76 


101 







* After Tvsdal and Crandall. 



for the production of imj^roved inbred lines, it is often possible to select as 
parents of crosses, select lines having high combining ability as parents of 
crosses, in addition to selection for other characters that are desired. In 
breeding for heterosis, however, it seems evident that genetic diversity of 
parentage is equally as important as combining ability (see Hayes, and 
Immer, 1942; Sprague, 1946b). 

All relatively homozygous, inbred lines in maize are much less vigorous 
than the better Fi crosses. It is apparent that heterosis is of great impor- 
tance in crosses with inbred lines of maize. 

Inbred lines that have undesirable characters may be easily imjjroved by 
the application of any one of several methods of breeding. The breeder may 
select for each problem the method or methods that seem to him most ap- 
plicable. In breeding selfed lines the selection of parents that have comple- 
mentary characters that together include the characters desired in the im- 
proved inbred is a natural first step. Subsequent methods of breeding may 



58 



H. K. HAYES 



be used according to the viewpoint of the breeder and the particular prob- 
lem to be solved. 

While combining ability is an inherited character, it seems of special in- 
terest that single crosses of high X high combiners have not been greatly su- 
perior in yield, on the average, to crosses of high X low. Both, however, were 
clearly higher in yielding ability than low X low crosses (Johnson and 
Hayes, 1940), (Cowan, 1943), (Green, 1948). An illustration from Johnson 
and Hayes (Table 3.8) shows the type of results obtained. The crosses were 
classified for yielding ability in comparison with recommended double 
crosses of similar maturity. 

Two recent studies in Minnesota may be used to illustrate other breeding 
problems. A further study was made by Johnson (1950) of the combining 
ability of F4 lines that were studied in earlier generations by Payne and 
Hayes (1949). Yield relations in the double cross Min. 608 (A344 X A340) 
(A357 X A392) are illustrated in Table 3.9. 



TABLE 3.8 

FREQUENCY DISTRIBUTION FOR YIELD OF SINGLE CROSSES 

OF SIMILAR MATURITY IN COMPARISON WITH 

RECOMMENDED DOUBLE CROSSES AS 



Type of 


Class Centers of —1 to —2, +1 to +2, etc. Times 
THE S.E. OF A Difference 


Cross 


-7 
-8 


-5 
-6 


-3 
-4 


-1 

-2 





+ 1 
+2 


+ 3 
+4 


+ 5 
+ 6 


+ 7 
+ 8 


Total 


Mean 


Low X low . 




1 
3 
1 


1 

.... 
"5 


2 

11 

12 


4 4 








12 
52 
83 


-0 5 + 7 


Low X high 

HighXhigh 


1 


6 

8 


16 
35 


9 
20 


5 
4 


1 


+ 1.1 + 0.4 
+ 1 1+0 2 









TABLE 3.9 

YIELD RELATIONS IN MIN. 608 

( A334 X A340) (A357 X A392) 





% M. 


Yield 
(Bu.) 


A334XA357 and A392 

A340XA357 and A392 


19.6 

18.5 


66.8 
62.4 


Average 


19.0 


64.6 


A357XA334and A340 

A392XA334and A340 


19.5 
18.6 


66.0 
63.2 


Average 


19.0 


64.6 


Min. 608 


19.0 


64.0 



DEVELOPMENT OF THE HETEROSIS CONCEPT 



59 



In these studies the usual method of predicting combining ability of a 
double cross gave excellent agreement between both predictions and the 
actual double cross yield. 

The studies of the performance in early and later tests of F2 to F4 lines 
from L317 X A116 when crossed with (A334 X A340) in comparison 
with A357(A334 X A340) were carried out by Payne and Johnson. The 
methods of comparing combining ability in different generations were 
adapted by the writer, who alone is responsible for the conclusions drawn. 
The lines were first placed in +1, — 1, etc. X L.S.D. at the 5 per cent point 
with the performance of A357(A334 X A340) as 0. Classes for performance 
of individual lines were made by adding the yield class of a line to its moisture 
class with the sign of the latter changed. 

The F2 and F3 crosses were both grown the same year, the F3 and F4 were 
grown in different years, and the F4and the top crosses were grown the same 
year (see Tables 3.10, 3.11, 3.12). 

In these studies no new lines seemed markedly superior to A357 in com- 



TABLE 3.10 

COMBINING ABILITY RELATION OF F2 AND F3 LINES 

OF (L317XA116) IN CROSSES WITH (A334XA340) 

GROWN IN SAME TRIAL IN 1947 



o 
•-t 

o 



+2 




1 

















2 




2 


1 


1 




-1 




1 


2 


1 


1 






-2 








1 


2 


3 




-3 


1 


1 




1 


3 


1 




-4 








? 


2 






-5 














2 


-7 














2 



+2 +1 -1 -2 -3 -4 -5 
F2 crosses, performance classes 



Total 
1 
6 

5 
6 
7 
4 

2 
2 

3i 



TABLE 3.11 

COMBINING ABILITY RELATION OF F3 AND F4 LINES 

OF (L317XA116) IN CROSSES WITH (A334XA340) 

F3 GROWN IN 1947, F4 IN 1949 

Total 



o 

o 

en 



+2 




1 















1 


2 




2 


1 






-1 








5 








-2 




1 


1 


3 


1 






-3 




1 1 




2 


3 






-4 




1 




1 


2 






-5 










1 




1 


-7 












1 


I 



+2 +1 -1 -2 -3 -4 
F4 crosses, performance classes 



1 
6 

5 
6 
7 
4 
2 
2 

33 



60 



H. K. HAYES 



bining ability with (A334 X A340). As A357 is rather outstanding in com- 
bining ability the result may not be so surprising. There was much greater 
relation between the combining ability of F3 and F4 lines and of F4 with top 
crosses than between F2 and F3. 

In an unpublished study of gamete selection, with a different but highly 
desirable double cross, there was an indication that a lower yielding inbred 
could be improved by an application of gamete selection (Stadler, 1944). 
The study is from one phase of a breeding program to improve Min. 406. 
The yield relations of inbreds in an average of single crosses are given in 
Table 3.13. 

Approximately 60 Fi plants of A25 X Golden King were selfed and top 
crossed with A73 X A375. Thirty-two of the more desirable plants were se- 
lected to study in yield trials. In this study both yield and moisture classes 
of plus 1, plus 2, etc. X L.S.D. at 5 per cent were used around the mean of 

TABLE 3.12 

COMBINING ABILITY RELATION OF F4 LINES OF 

(L317XA116) IN CROSSES WITH (A334XA340) AND 

WITH GOLDEN KING. GROWN IN 1949 

Total 



In 

A334 
X 

A340 
Crosses 



+ 2 


1 












+ 1 


1 















1 3 


1 


1 








-1 








1 






-2 


9 




1 


3 






-3 






3 


4 


1 




-4 












1 


-5 








1 


1 





+ 1 -1 -2 -3 -4 -5 
In Golden King Crosses 



1 
1 
6 
1 
13 
8 
1 
2 

33 



TABLE 3.13 

GAMETE SELECTION IN THE IMPROVE- 
MENT OF MINHYBRID 406 

(A25XA334)(A73XA375) 





Av. OF Crosses 




%M. 


Bu. 


A25XA73, A375 

A334XA73, A375 

A73XA25, A334 

A375XA25, A334 


24.6 
24.7 

24.6 
24.7 


76.2 
79.4 

74.8 
80.8 



Proposal for improvement of A25 and A73: 
A25XG. King gametes 
A73XMurdock gametes 



DEVELOPMENT OF THE HETEROSIS CONCEPT 



61 



A25 X tester as 0. The results (see Table 3.14) indicate that gametes from 
Golden King are a desirable source of improvement of A25 in crosses with 
A73 X A375. 

From this first trial three high and three low yielding lines were selected, 
and selfed progeny grown in Si. Plants in each of the three Si high and three 
low combining lines were selected, selfed, and again top crossed on A73 X 
A3 75. The agreement for So and Si lines was very good (see Table 3.15). 
It appears that gamete selection is an excellent breeding method for the 
early selection of material to improve the specific combining ability of a 
known inbred. 

SOME GENETIC CONCEPTS OF HETEROSIS 
It seems very evident to the writer that heterosis, the increased vigor of F] 
over the mean of the parents or over the better parent, whichever definition 
is "used, is not due to any single genetic cause. A brief summary of various 

TABLE 3.14 

DISTRIBUTION OF % MOISTURE AND YIELD OF 32 So 
PLANTS OF A25XG. KING CROSSED TO A73XA375. 
CLASSES OF L.S.D. 5% AROUND MEAN OF A25X 
TESTER 



%ear 
mois. 



+2 
+ 1 


1 

3 


2 


-1 

-2 


2 
3 


8 

5 


5 
3 




— 2 


-1 


+ 1 +2 



(mean of A25X 
tester) 



Yield 
(mean of A25X tester) 



TABLE 3.15 

PERFORMANCE INDICES OF So AND Si LINES 
FROM A25XG. KING WHEN CROSSED TO 
A73XA375 TESTER AND COMPARED WITH 
A25XTESTER 





So 


Si 




Gamete 








No. OF 


Number 








Si's 




1947 


1949 


1949 




19 H 


+ 11 


+ 19 


+25 


5 


20 H 


+ 14 


+ 9 


+ 14 


7 


36 H 


+ 9 


+ 16 


+ 11 


7 


5L 


-11 


- 3 


+ 5 


7 


29 L 


-11 


- 1 


- 


1 


46 L 


- 5 


+ 1 


+ 2 


7 



62 H. K. HAYES 

theories advanced to explain heterosis seems desirable to set the stage for 
later discussions. Bruce (1910) explained heterosis on the combined action 
of favorable dominant or partially dominant factors, based as Richey (1945a) 
has emphasized on mathematical expectations. 

Keeble and Pellew (1910) used a similar hypothesis on a di-hybrid basis 
to explain hybrid vigor in peas. East and G. H. Shull (1910-1914) believed 
vigor was dependent on heterozygosis on the basis that the stimulus of hy- 
bridity was not entirely Mendelian. A. F. Shull (1912) preferred the explana- 
tion that heterosis was due to a stimulus resulting from a changed nucleus 
on a relatively unaltered cytoplasm. Jones (1917) restated Bruce's concept 
and added the concept of linkage. 

Collins (1921) and Richey (1945) have pointed out that where large num- 
bers of factor pairs are involved it would be very difficult to recover all fac- 
tors in a favorable condition in F2, or in later segregating generations. With 
multiple factors involved, however, linkage must of necessity make the re- 
combination of factors more difficult. East (1936) presented a Mendelian 
concept of the interaction of alleles at the same locus to explain heterosis, 
where two alleles of a particular gene pair had each developed a divergent 
physiological function. The writer believes he continued also to accept the 
previous explanation that heterosis was dependent on the cumulative effect 
of dominant or partially dominant linked genes. 

Gustafsson (1947), Hull (1945a), Jones (1945), Castle (1946), and others 
have emphasized the importance of interallelic action in relation to heterosis. 
Castle has suggested also that the effect of interallelic action of a single pair 
of genes "is similar to that of the killer mutation of Sonneborn, except that 
the action induced in the dominant gene by its sensitized recessive, instead 
of being harmful, in this case is beneficial." 

In certain cases a homozygous recessive pair of genes may completely 
modify the normal expression of either a homozygous or heterozygous or- 
ganism. Homozygous dwarfs in maize condition such a result. A cross be- 
tween two different dwarfs, however, releases the inhibition of each dwarf 
and results in marked heterosis. Both dominant factors, where two dwarfs 
are crossed, appear to be necessary to condition normal development. In this 
case the dominant conditions of both factor pairs act as complementary fac- 
tors for normal growth. 

It is evident that genes are greatly affected in their expression by differ- 
ences in both external and internal environment. Cytoplasmic inheritance of 
male sterility may be used for illustrative purposes. Several cases of male 
sterility in sugar beets and onions, for example, are known that are due to 
maternal cytoplasmic inheritance which may be modified in expression by 
the dominant or recessive condition of one or more factor pairs. 

Recently Hsu (1950) at Minnesota has studied the effect of two pairs of 
dwarf factors of maize in their homozygous dominant and recessive condi- 



DEVELOPMENT OF THE HETEROSIS CONCEPT 



63 



tions, and also when heterozygous in near isogenic, homozygous, and highly 
heterozygous backgrounds. 

The factor pair for Didi was studied in the near isogenic background of 
inbred A 188, that of D^dx in the near isogenic background of A95-344, and 
both factor pairs were studied in crosses between A 188 X A95. Particular 
attention was given to total dry matter produced at various periods of growth 
under field conditions and to the growth in length of the coleoptile and meso- 
cotyl under controlled laboratory conditions. 

One comparison of the growth of the mesocotyl during a 12-day period 
for DiDi and Didi on three different near isogenic backgrounds will be con- 
sidered: the near isogenic background, A188, and the highly heterozygous 
backgrounds of A 188 X A95 in the presence of D^D^ and D^d^, respec- 
tively. While Di conditioned greater growth of mesocotyl in length than d\, 
Dx conditioned less development of the mesocotyl in length than dx. 

The mesocotyl length of six strains consisting of comparisons of DiDi 
with Didi on three dififerent backgrounds was taken as 100. The comparisons 
are summarized in Table 3.16 and in Figure 3.1. 

It is apparent that the superiority of DiDi over Didi in mesocotyl length 
becomes less in the highly heterozygous background than in the homozygous 
background of A188. This may be more evident from the diagram in Fig- 
ure 3.1. 

TABLE 3.16 

COMPARATIVE LENGTH OF MESOCOT- 
\T. FOR SIX STRAINS OF CORN 



Background 


Percentage 
Difference in 

Mesocotyl 

Length, DiDi 

minus Didi 


Percentage 
Expression of 
Background 


A188 


19 

16 

4 


89 


A188XA95Z?xZ?x... 
^.\2,%X^9SDxdx.... 


101 

no 



It seems of some interest that the differences between DiDi and Didi were 
smaller in the highly heterozygous background than in the homozygous 
background, and that in the presence of D^x that the differences were 
further reduced over those in the presence of DxDx. It may be well to recall 
that dx conditioned greater length of mesocotyl than Dx- 

Reference may be made to an explanation by Torssell (1948) of tlie decline 
in green weight or length of stem in alfalfa in different generations of in- 
breeding. It was not greatest in the first inbred generation. He suggests there 
was a surplus of vigor genes in a heterozygous condition in the early genera- 
tions of selfing, and that great loss of vigor was not observed until about I3 



C/) 



< -o 



(/) 



(A 



105 



(/) 


Q 


(/) 




Ixl 


O 


tr 


1- 


Q. 




X 


o 




z 


UJ 


h- 


o 


< 


< 


_J 


1- 


UJ 


z 


rr 


UJ 




o 




or 




UJ 




CL 





95 



85 



75 



A D,D, 




/ 



/ 



/ 



/ 



/ 



/ 



A' 



A 



B 

i± 



C 



80 90 100 110 

PERCENTAGE EXPRESSION OF 
BACKGROUND 



Fig. 3.1 — Relative expression of A A vs. A^i regarding final length of mesocotyl on vari- 
ous backgrounds: (A, A188; B, A188 X A95-344 carrying AA; C, A188 X A95-344 D,d^). 



DEVELOPMENT OF THE HETEROSIS CONCEPT 65 

when selfing reduced the necessary genes below a stage needed by the or- 
ganism. The following quotation from Thorssell emphasizes the viewpoint 
that the relative importance of genes controlling heterosis is greatly in- 
fluenced by other factors of the organism: 

The cumulative effect of heritable characters, however, brings it about that develop- 
ment, that is to say green weight, does not stand in arithmetical proportion to the number 
of pairs of the dominant genes in question. From this it follows also that the said number can 
be reduced within a certain limit without perceptible or any great influence upon green 
weight. If this limit is exceeded, a considerable degeneration sets in. 

The speaker has chosen to consider heterosis as the normal expression of 
a complex character when the genes concerned are in a highly heterozygous 
condition. x'Vs most normal characters are the end result of the action, reac- 
tion, and interaction of countless numbers of genes, and as gene mutation 
constantly occurs although relatively infrequently, it may be impossible to 
obtain all essential genes in the most favorable homozygous state. After 
selecting the best homozygous combinations, further vigor will be obtained 
due to heterozygous combinations of factors. Dominance or partial domi- 
nance seems of great importance as an explanation of hybrid vigor. In some 
cases there may be extra vigor correlated with the heterozygous condition of 
pairs of alleles. The types of response of inter and intra allelic factor interac- 
tions are without doubt dependent upon both external and internal environ- 
ment. 



M. M. RHOADES 

University of Illinois 



Chapter 4 

Preferential 
Segregation in Maize 



The outstanding example of the utilization of heterosis in plant improve- 
ment is that of hybrid corn. Extensive studies on maize genetics have clearly 
demonstrated that chromosome and gene segregation are in accordance with 
Mendel's laws of segregation and recombination. It would appear, therefore, 
that any unusual mechanism operating in maize to produce deviations from 
normal Mendelian behavior should be worthy of our consideration, even 
though the principles involved have no bearing on the nature or manifesta- 
tion of heterosis. The purpose of this section is to present data on preferential 
segregation in maize and to offer a tentative interpretation of this phe- 
nomenon. 

Two kinds of chromosome 10, the shortest member of the haploid set of 
ten, are found in populations of maize. The common or normal type gives 
typical Mendelian ratios when the two homologues are heterozygous for 
mutant loci. The second kind of chromosome 10, which has been found in a 
number of races from Latin America and the southwestern United States, 
also gives normal Mendelian ratios for chromosome 10 loci in plants homozy- 
gous for this chromosome. This second or abnormal kind of chromosome 10 
differs from the normal chromosome 10 by a large, chiefly heterochromatic 
segment of chromatin attached to the end of the long arm and also in the 
chromomeric structure of the distal one-sixth of the long arm (see Fig. 4.1 
and Fig. 1 of Plate I). As is illustrated in Figure 4.1 the chromomeres in this 
region are larger and more deeply staining than are the correspondingly 
situated chromomeres of the normal homologue. 

Although normal Mendelian ratios are obtained for segregating loci in 
chromosome 10 in both kinds of homozygotes, we were able to show in an 
earlier paper (Rhoades, 1941) that preferential segregation occurs at mega- 

66 




n 






Fig. 4.1 — Camera lucida sketch at pachynema of bivalent consisting of one normal and one 
abnormal chromosome 10. Note the dissimilarity in chromomere pattern in the distal one- 
sixth of the long arm. The identical chromomere pattern found in the remainder of the 

chromosomes is not figured here. 

Fig. 4.2 — Anaphase I of cell illustrated in Figure 4 of Plate I. Some of the disjoining dyads 
are normal appearing while others have active neo-centric regions. 

Fig. 4.3 — Metaphase I with eleven dyads. Five of the dyads have precocious neo-centro- 
meres at sub-terminal portions of their long arms. 

Fig. 4.4 — Anaphase II of cell illustrated in Figure 7 of Plate II. In some of the inverted 
V-shaped monads the true centric regions are attracted toward the opposite pole. 



68 M. M. RHOADES 

sporogenesis in plants heterozygous for a normal and an abnormal type of 
chromosome 10. Approximately 70 per cent of the functioning mega spores 
possessed the abnormal 10 instead of the usual 50 per cent. The excess of 
female gametes with the abnormal 10 was not due to lethal factors or to 
megaspore competition. The disjunction of the two dyads comprising the 
heteromorphic bivalent at anaphase I, and of the two monads of each dyad 
at anaphase II, was such that an abnormal 10 chromosome tended to pass 
with a high frequency to the basal spore of the linear set of four. 

The factor or factors responsible for this preferential segregation reside 
in the chromatin segments which differentiate the two kinds of chromosome 
10. Whether the distal one-sixth of the long arm or the large heterochromatic 
piece of extra chromatin carries the causative genes for preferential segrega- 
tion has not yet been determined — since these two regions of the abnormal 
chromosome 10 have never been separated by crossing over. The locus of 
the gene R is in the long arm of chromosome 10. There is approximately 1 per 
cent recombination between R and the end of the long arm in plants hetero- 
zygous for the two kinds of chromosome 10; but every crossover distal to R 
occurred to the left of the dissimilar chromomeres in the distal one-sixth of 
the long arm. Apparently little or no crossing over takes place here, although 
pairing at pachytene is intimate. 

Strictly terminal chiasmata in the long arm have not been observed at 
diakinesis in heterozygous plants. The close linkage of the R locus with the 
extra segment of abnormal 10 is due to a suppression of crossing over in the 
end regions of the long arm. E. G. Anderson (unpublished) has studied a re- 
ciprocal translocation involving normal 10 with the break distal to R, and 
found 5 per cent recombination between R and the translocation point. 
There is an undetermined amount of crossing over between the translocation 
point and the end of the chromosome. It should be possible to locate the re- 
gion or regions in abnormal 10 responsible for preferential segregation by ob- 
taining successively larger terminal deficiencies, but this has not been at- 
tempted. 

The dissimilarity in chromomere pattern in the distal portion of the long 
arms of the abnormal and normal chromosomes 10, together with the lack of 
crossing over in this region, suggest the possibility that the gene content may 
not be identical in the two kinds of chromosome 10. Inasmuch as plants 
homozygous for the abnormal chromosome 10 are not noticeably different in 
growth habit and general appearance from sibs carrying only the normal 10, 
it would appear that some kind of structural modification was responsible for 
the suppression of crossing over. To assume that this distal region consists 
of non-homologous loci in the two types of chromosome would mean that 
plants with two abnormal 10 chromosomes would be homozygous deficient 
for certain loci found in the comparable region of normal 10. This appears 
unlikely. 



PREFERENTIAL SEGREGATION IN MAIZE 69 

That a structural difference, aside from the extra chromatin of abnormal 
10, exists between the two kinds of chromosome 10 also is indicated by the 
pairing relationships in plants trisomic for chromosome 10. in plants with 
two normal and one abnormal chromosome 10, trivalent associations were 
observed in 251 (60.2 per cent) among a total of 417 microsporocytes. When 
a chain of 3 was found at diakinesis, the abnormal 10 occupied a terminal 
position in 90 per cent of the cells. It was united with a normal chromosome 10 
by a chiasma in the short arm. A univalent chromosome 10 was found at 
diakinesis in 39.8 per cent of the pollen mother cells. 

If pairing, as reflected by chiasmata formation, were random among the 
three chromosomes, the ratio of normal -.abnormal chromosomes 10 in the 
univalent class should be 2 : 1. Actually the unpaired chromosome was a nor- 
mal 10 in 28 cells among a total of 166, while in the remaining 138 cells the 
univalent was an abnormal 10. In individuals again trisomic for chromo- 
some 10, but possessing one normal and two abnormal chromosomes, the 
percentage of trivalent associations at diakinesis was 57.9 in a total of 513 
cells. In the chains of 3, the two abnormal homologues were adjacent mem- 
bers, joined by a chiasma between their long arms, in 70 per cent of the 
cases. An unpaired chromosome 10 was found in 42.1 per cent of the micro- 
sporocytes. 

If pairing were random, two times as many abnormal lO's as normal lO's 
should be found as univalents; but in a total of 216 cells an abnormal 10 
was the univalent in 69, while a normal chromosome 10 was the univalent 
in 147. Chiasma formation among the three chromosomes 10 of trisomic 
plants clearly is not at random. There is a marked preference for exchanges 
in the long arm between the two structurally identical homologues. If synap- 
sis usually begins at the ends and progresses proximally, the non-random as- 
sociations found in trisomic plants become understandable. Normal recom- 
bination values for the li-gi and gi-R regions which lie proximal to R (see 
Table 4.1 for gi-R data) indicate that any suppression of crossing over is 
confined to the region beyond the R locus in disomic plants heterozygous for 
the two kinds of chromosome 10. It is no doubt significant that differences 
in chromomeric structure are not found in regions proximal to the R locus. 

Inasmuch as the R locus is closely linked with the extra chromatin of ab- 
normal 10, the ratio oi R:r gametes from heterozygous plants gives a good 
approximation of the frequency with which the abnormal chromosome passes 
to the basal megaspore. The genetic length of the long arm of chromosome 10 
is such that at least one chiasma is found in the arm. If one chiasma invari- 
ably occurs in the long arm of heteromorphic bivalents, each of the two dis- 
joining dyads of anaphase I will possess one normal chromatid and one ab- 
normal chromatid. Preferential segregation would be restricted to the sec- 
ond meiotic division, and occur only if the orientation of the dyad on the 
spindle of metaphase II were such that the abnormal chromatid passed to 



70 M. M. RHOADES 

the lower pole of the spindle. Normal segregation would occur in those mega- 
sporocytes which had homomorphic dyads. 

If the terminal segment of abnormal 10 determines preferential segrega- 
tion, it follows that loci near the end of the long arm will be preferentially 
segregated more frequently than loci further removed from the end of the 
chromosome. From the data in Tables 4.1 and 4.2 it is evident that the dis- 
tortion from a 1 : 1 ratio is greater for the R locus than for the more proximal- 
ly situated gi locus. The li locus which is proximal to gi was less affected 
than ^1. 

Longley (1945) reported non-random segregation at megasporogenesis for 
chromosome pairs other than chromosome 10 when one of the two homologues 
had a prominent knob and the other was knobless. Segregation was random 
for these heteromorphic bivalents in plants homozygous for the normal chro- 
mosome 10, and non-random if abnormal 10 was heterozygous. He studied 
preferential segregation of chromosomes 9 and 6. The data for chromosome 9 
are the most instructive. Some strains of maize have a chromosome 9 with a 
knob at the end of the short arm, others have a knobless chromosome 9. The 
C, Sh, and Wx loci lie in the short arm of this chromosome, with Wx nearer 
to the centromere. C and Sh are in the distal one-third of the short arm. Ap- 
proximately 44 per cent recombination occurs between Wx and the terminal 
knob — they approach independence — while C and Sh are 23 and 26 recombi- 
nation units distant from the knob. * 

When plants of knob-C/knobless-c constitution, which were also heterozy- 
gous for abnormal 10, were pollinated by recessive c, 64 per cent of the func- 
tioning megaspores possessed the C allele. The Sh locus, close to C, showed a 
similar degree of preferential segregation in comparable tests, but the Wx 
locus was little affected. Such a progressive decrease in effect is expected if 
the terminal knob on the short arm is instrumental in producing preferential 
segregation. The part played by the knob of chromosome 9 was wholly un- 
expected. Obviously this heterochromatic structure can no longer be con- 
sidered as genetically inert. The data on various loci in chromosomes 9 and 
10 prove that the degree of preferential segregation of a locus is a function 
of its linkage with heterochromatic regions which, in some way, are con- 
cerned with non-random segregation. 

The data presented above show that alternative alleles are not present in 
equal numbers among the female gametes when abnormal 10 is heterozygous. 
We have here an exception to Mendel's first law. Are deviations from Men- 
del's second law, the independent assortment of factor pairs on non-homolo- 
gous chromosomes, also occurring? This question is answered by Longley's 
data where the C and R loci are both segregating preferentially. In separate 
experiments he found the C locus was included in 64 per cent and the R locus 
in 69 per cent of the functioning megaspores. Assuming that these percent- 
ages hold in plants where both are simultaneously segregating, the observed 




•fppl 






2 



* ^< 



^ i' 



•» ' .' 






4 



Plate I: Fig. 1 — Pachytene showing homozygous abnormal 10. Carmine smear. The proximal jjortion of 
the extra chromatin is euchromatic as is a smaller distal piece. A large and conspicuous knob lies between 
the two euchromatic portions. Fig. 2 — Metaphase I in microsporocyte homozygous for abnormal 10. 
Carmine smear. The ten bivalents each have their true centric regions co-oriented on the spindle. The 
onset of neo-centric activity is manifest in the second, sixth, and seventh bivalents from the right. The 
third and fourth bivalents from the right are somewhat superimposed. Figs. 3 and 4 — .\naphase I in mi- 
crosporocyte homozygous for abnormal 10. Carmine smear. Some of the dyads are undergoing a normal 
anaphase separation while in others the neo-centric regions are pulling the ends poleward. Note that 
the normal ap[)earing dyads are slower in their poleward migration. F'^ig. 4.2 is a drawing of Fig. 4 above. 



Plate II: Figs. 1 and 2 — Metaphase II in plant homozygous for abnormal 10. Carmine 
smear. Precocious poleward movement of neo-centric regions is clearly evident. One dyad 
has a single neo-centric region (Fig. 4.5, dyad No. 8) while the left-most dyad has a neo- 
centric region in both long arms (Fig. 4.5, dyad No. 7). This cell was figured in Rhoades 
and Vilkomerson 1942. Figs. 3 and 4 — .Anaphase II in jilant homoz\gous for abnormal 10. 
Carmine smear. Note that the rod-shaped monads with precocious neo-centromeres are 
the first to reach the poles. Fig. 5 — Metaphase II in plant homozygous for abnormal 10. 
Carmine smear. The only chromosome of the haploid complement which can be recognized 
at metaphase II is chromosome 6 which has a satellite at the end of the short arm. In this 
cell the chromosome 6 d\'ad is the second from the left. That the terminal chromosome of 
the satellite is actuall\- a small knob is indicated by the formation of neo-centric regions al 
the end of the short arm. Fig. 6 — Early anaphase II in plant heterozygous for abnormal 10. 
Carmine smear. That the poleward movement of neo-centric regions is less rapid in hetero- 
zygous than in homozygous abnormal 10 plants is indicated here by the relatively slight 
attenuation of the rod-shaped monads. Fig. — 7 Late anaphase II in plant homozygous for 
abnormal 10. Carmine smear. The previously greatlx- stretched rod monads with precocious 
neo-centromeres have contracted. Note the inverted V-shaped chromatids. This is the same 
cell shown in Figure 4.4. Fig. 8 — Side view of metaphase I in a normal plant showing the 
fibrillar nature of the chromosomal fibers. Fi.xed in Benda, stained with haemotoxylin. 
Paratiine section. The only chromosomal fibers j^resent are those formed by the true cen- 
tromeres. Ordinarily chromosomal fibers are not evident in carmine smears since they are 
destroyed by acetic-alcohol fi.xation and it is nccessar\' to use special techniques to demon- 
strate them. Similar fibrillar chromosomal fibers are found at neo-centric regions when 
proper fixation and staining methods are employed. Fig. 9 (top) — Polar view of meta- 
phase I in normal plant. Fixed in Benda, stained with haemotoxylin. Parafhne section. 
Note the arrangement of the ten bivalents on the ecjuatorial plate. This microsporocyte 
was cut slightl}' above the metaphase plate. The next section, which includes the remaining 
portion of this cell, is a cross section through the ten sets of chromosomal fibers. 



.// 



J 



> 







/ 



j> 



1 



/ 



' 2 

'\ h I- 
't J.^f 

i . S 



■*l!7 



^.-■IH.- ;^^'4«H*' 



# 



8 



// 






T^ • > .' 



A V V, 



^ A. 



' M» /• 



••» 




1 



^lMjJ 






TABLE 4.1 

LINKAGE DATA FROM THE CROSS OF G r ABNORMMVg R 
NORM.\L X gr d'd' 



Linkage 
Phase 


Constitution of Chromosomes 




Repulsion 


(0) 

G 
r 


(o) 

g 
R 


(X) 

G 
R 


(x) 
S 

r 


Total 


Ratio of 
i?:r ON Ear 




243 
102 
150 
396 
154 
169 
215 
231 


138 
86 

114 
50 
81 
90 
61 
79 


29 
9 

18 
7 

11 
21 
24 
35 


49 
13 
20 
59 
29 
30 
77 
81 


459 
210 
302 
512 

275 
310 
377 
426 


186/?:326r 
136/?:319r 
145/?:288r 
169/?:588r 
120/?:277r 
127/?:223r 
102/?:338r 
133/?:358r 




1660 


699 


154 


358 


2871 


1118/?:2717r 



%Rm total = 29.7 % g in total = 36.8 

% i? in non-crossover classes = 29.6 
% /J in crossover classes = 30.1 

G — R recombination = 17.8% 



29.2% R 



TABLE 4.2 

LINKAGE DATA FROM THE CROSS OF G r NORMAL/g R 

ABNORMAL X gr 



Linkage 
Phase 


Constitution of Chromosomes 




Repulsion 


(o) 
G 
r 


(o) 

g 
R 


(x) 
G 
R 


(x) 

g 

r 


Total 


Ratio of 
R\r (m Ear 




12 
38 
35 
39 


87 

96 

86 

107 


13 
29 

21 


1 

6 

7 
9 


113 
169 
161 
176 


182/?: 42r 
188/?: 59r 
230/?: 74r 
241/?: 77r 




124 


376 


96 


23 


619 


841/?:252r 



% r seeds in total = 23.8 

% r seeds in non-crossover classes = 24.8 

% T seeds in crossover classes = 19.3 

G — R recombination = 19.2% 



72 M. M. RHOADES 

frequencies of F2 phenotypes can be compared with those calculated on the 
assumption of independent assortment. The two values agreed very closely, 
indicating little or no deviation from the law of independent assortment. 
His data, from plants where loci in chromosomes 9 and 6 are both segregat- 
ing preferentially, likewise permit such a conclusion to be drawn. 

In my 1942 paper on preferential segregation the statement was made 
that the chromosomes in plants with the abnormal chromosome 10 formed 
extra chromosomal (half spindle) iibers at regions other than the true centro- 
mere region. Rhoades and Vilkomerson (1942) found these supernumerary 
chromosomal fibers were produced only in plants homozygous or heterozy- 
gous for the abnormal 10, and that sister plants homozygous for the normal 
10 had chromosomal fibers originating solely from the localized centric re- 
gion in an orthodox manner (see Fig. 8 of Plate II). Although the abnormal 
chromosome 10 was clearly responsible for the formation of these neo-centric 
regions, they were not restricted to this chromosome since many of the non- 
homologous chromosomes had supernumerary chromosomal fibers. The ab- 
normal chromosome 10 is thus responsible for the formation of neo-centric 
regions, as well as for preferential segregation. Since 1942, a considerable 
body of data has been obtained bearing on the behavior of abnormal 10. 
Some of the more pertinent observations have suggested a cytological mecha- 
nism for the phenomenon of preferential segregation. 

The unorthodox formation of supernumerary chromosomal fibers from neo- 
centric regions is limited to the two meiotic divisions. (For a description of 
normal meiosis in maize see Rhoades, 1950.) The first meiotic division is in 
no way exceptional until metaphase I is reached. Normal appearing bivalents 
are co-oriented on the spindle figure in a regular manner with the half spindle 
fibers, arising from the true centric regions, extending poleward. Normally 
these fibers effect the anaphase movement of the disjoining dyads with the 
localized centromere region leading the journey to the spindle pole. How- 
ever, in plants with the abnormal 10, chromosomal fibers arise from distal 
regions of the chromosome while the bivalents are still co-oriented on the 
spindle at metaphase I. The neo-centric regions are drawn poleward more 
rapidly than the true centric regions. Consequently the distal ends, instead 
of being directed toward the spindle plate during anaphase I, lead the way 
to the pole. 

The appearance of many disjoining dyads at anaphase I suggests that 
their poleward migration is due largely, even exclusively, to the fibers origi- 
nating from the neo-centric regions. The primary centric region appears to 
play no active role even though it possessed chromosomal fibers at meta- 
phase I when the tetrad (bivalent) was co-oriented. At mid-anaphase there is 
no indication of the presence of these fibers in many of the dyads with the 
precocious neo-centric regions. 

Figure 4.5 and Figures 3 and 4 of Plate I illustrate some of the observed 



4 





ANAPHASE I DYADS 



A 

11 



7 8 

METAPHASE H DYADS 



J ii I ^ 

12 13 »U ^15 



10 



i 



12 



ANAPHASE IE MONADS 



t 



16 



Fig. 4.5 — All figures are from carmine smears of homozygous abnormal 10 plants. Figures 
1-5 represent various configurations found at anaphase I. Figure 1 is a normal dyad with 
chromosomal fibers formed only at the true centric region. In Figure 2, two arms have 
formed neo-centric regions. The true centric regions appear to be inactive. Figure 3 shows a 
dyad with two neo-centric regions and an active true centric region whose chromosomal 
fibers are directed away from the nearest pole. Figure 4 is a dyad with a single neo-centric 
region. In Figure 5 the two neo-centric regions are directed to opposite poles. Figures 6-7 
illustrate various metaphase II dyads. The location of the equatorial plate is represented 
by horizontal lines. Figure 6 is essentially normal with no formation of neo-centromeres. 
Figure 7 is a dyad with two neo-centric regions directed toward opposite poles. There is a 
single neo-centric region in Figure 8. Figure 9 is a dyad which is disjjlaced from the equa- 
torial plate. The true centric region has divided to form two independent monads. Each 
monad has formed two neo-centric regions which are oriented toward opposite poles. In 
Figure 10 one of the monads has its two neo-centromeres directed to opposite poles. Fig- 
ures 11-16 are illustrations of monads found at anaphase II. Figure 11 is a normally dis- 
joining monad. In Figure 12 a single neo-centromere is evident. Figure 13 shows two neo- 
centric regions. Figure 14 has a single neo-centromere which was active at metaphase II. 
In Figure 15, chromosomal fibers have arisen from two neo-centric regions and also from 
the true centric region. The true centric region and the neo-centromeres are acting in op- 
posite directions. Figure 16 shows a monad with two neo-centric regions which are directed 
toward opposite poles. This type of monad is derived from those shown in Figure 9. 



74 M. M. RHOADES 

anaphase I configurations. Chromosomal fibers may arise from one or both 
of the long arms of each dyad at late metaphase or early anaphase I. Al- 
though it was not always possible to differentiate between long and short 
arms, the neo-centric regions in general appear to be confined to the long 
arm. When both long arms of the two chromatids of a dyad possessed a neo- 
centric region, the chromosomal fibers arising from these centric regions were 
usually directed toward the same pole. Occasionally they were oriented to 
opposite poles thus causing a great attenuation. In such cases, however, 
those chromosomal fibers nearest to one pole were powerful enough to over- 
come the oppositely directed force of the second neo-centromere. Despite the 
great complexity of configurations at anaphase I resulting from interacting 
and conflicting half-spindle fibers arising from both the true and neo-centric 
regions, the end of anaphase I usually finds ten dyads at each pole. Some- 
times, however, greatly stretched chromosomes undergo breakage. This 
breakage doubtless accounts for the higher pollen abortion (about 10 per 
cent) found in homozygous abnormal 10 plants as contrasted to the lower 
(0-5 per cent) pollen abortion of normal sibs. 

Even though one or two arms of some dyads are markedly stretched at 
anaphase I, the ensuing telophase is normal. All four arms of each dyad con- 
tract to form a spherical mass of chromatin which is loosely enveloped by 
the lightly-staining matrical substance. The chromonemata uncoil during 
interphase and early prophase II finds each daughter cell with ten, long X- 
shaped dyads of typical appearance. The two chromatids comprising each 
dyad are conjoined by the undivided primary centric region. There is no indi- 
cation of neo-centric regions, although some of the long arms possessed chro- 
mosomal fibers at the preceding anaphase. 

The onset of metaphase II sometimes occurs before the dyads have under- 
gone their usual contraction. Occasionally chromosomal fibers arising from 
neo-centric regions in the long arms are found at late prophase II. These 
precociously acting fibers produce an extension of the long arms before any 
spindle is visible. This observation is of singular importance. Some authori- 
ties believe that the centromere region is attracted (whatever this term may 
signify) to the spindle pole. Here we have a movement produced by the 
chromosomal fibers of neo-centric regions in the absence of an organized 
spindle. The way in which these neo-centric fibers act can only be conjec- 
tured, but no interaction between centric regions and spindle pole is essential. 
It is, indeed, probable that the only role of a bipolar spindle is to provide a 
structural frame which channels the chromosomes to the spindle poles. 
Clark's (1940) studies on divergent spindles are pertinent in this respect. 

The objection may be raised that the chromosomal fibers of neo-centric 
regions are not comparable to those arising from the true centric region. I 
do not believe this is a valid criticism. Not only are both kinds of fibers con- 
cerned with chromosome movement, but, as will be shown in a later section, 



PREFERENTIAL SEGREGATION IN MAIZE 75 

the fiber-producing activity of the neo-centric regions is a ])r()<luct of the true 
centric region. 

The appearance of neo-centric fibers in prophase II is not the rule. Usually 
the dyads come to lie with the true centric region on the spindle plate at 
metaphase II before any pronounced activity of neo-centric regions is ap- 
parent. Before the primary centric region divides, thus permitting a normal 
anaphase, chromosomal fibers again arise near the distal ends of the long 
arms of some dyads. These newly formed fibers move the long arms poleward 
while the dyad is still held on the metaphase plate by the undivided true 
centric region. This poleward movement is so rapid that the ends of the 
chromosomes may reach the spindle poles before the true anaphase occurs. 
Eventually the true centric region becomes functionally split, and the two 
monads fall apart and pass poleward. It is evident from Figures 4.4 and 4.5 
and Figure 7 of Plate II that the configurations of the disjoining monads 
(chromatids) at anaphase II are greatly different from normal. 

Neo-centric activity, as shown by formation of additional chromosomal 
fibers, occurs in plants both homozygous and heterozygous for the abnormal 
10, but it is much more striking in homozygous plants. Plants trisomic for 
abnormal 10 were not greatly different from homozygous disomic sibs. 

Precocious chromosomal fiber formation by neo-centromeres at metaphase 
II appears in general to be confined to the long arms of the dyads, although 
it is often difficult to differentiate between two unequal arms when one is 
stretched poleward. Some chromosomes have arm ratios so extreme that 
the distinction between long and short arms is clear, and in these chromo- 
somes the precocious fibers at metaphase II arise from the long arms. It is 
perhaps significant that, with the exception of the terminal knob on the short 
arm of chromosome 9, all remaining knobs in our material were situated in 
the long arms. (Chromosome 6 had two small knobs in its long arm but a 
maximum of one knob was present in the other chromosomes.) Corre- 
spondingly, only one of the two arms of any chromatid had neo-centric 
activity at metaphase 11.^ The number of dyads with precocious spindle 
fibers, as judged by the number of arms pulled poleward at metaphase II, 
varied in different strains. The maximum number in some plants was seven, 
in others five, etc. Plants with seven knobbed chromosomes had a maximum 
of seven dyads with arms stretched poleward at metaphase II. Those with 
four knobs had four such dyads. That is, a strong correlation exists between 
knob number and the number of dyads with neo-centric activity at meta- 
phase II. 

A further observation of some interest was that in plants homozygous for 
all knobs both homologous arms of a dyad usually were pulled poleward at 
metaphase II; while in plants heterozygous for some knobs many of the 
dyads had only one arm with neo-centric activity (see Figure 4.5 and Figures 

1. With the possible exception of chromosome 6. See Figure 5 of Plate II. 



76 M. M. RHOADES 

1 and 2 of Plate II). It is not unreasonable to assume that dyads with both 
homologous arms exhibiting neo-centromeres at metaphase II carried a knob 
in each chromatid, while dyads with one neo-centromere consisted of one 
knobbed and one knobless chromatid. Such heteromorphic dyads would arise 
from heteromorphic bivalents by a crossover between the true centromere 
and the knob. We believe that only knobbed chromatids have active neo- 
centromeres at metaphase II, and that knobless ones are normal at this stage. 
Unfortunately, knobs cannot be recognized at metaphase II, and the validity 
of the above assumptions rests upon indirect but convincing evidence. 

Two types of disjoining monads are found at anaphase II, those which 
are rod-shaped and those which are V-shaped. Monads which had one arm 
extending poleward at metaphase II are rod-shaped. They are the first to 
reach the pole. Indeed distal portions of such chromatids already had arrived 
there during metaphase II owing to the early action of their neo-centromeres. 
The V-shaped monads of anaphase II are derived from those chromatids 
devoid of neo-centromeres at metaphase II. The poleward migration of some 
monads is first begun by the chromosomal fibers emanating from the true 
centric region, but shortly after anaphase is initiated chromosomal fibers 
may arise from the ends of both arms. These terminally placed fibers, which 
are directed to the same pole, propel their ends poleward with such rapidity 
that the ends first overtake and then pass the centric region in the course of 
anaphase migration. Consequently these monads reach the poles as inverted 
V-shaped chromosomes (see Fig. 4.4). The spindle fibers from the true centric 
region now are directed toward the spindle plate rather than to the pole — they 
have reversed their orientation. This would be impossible if chromosomal 
fibers were of a thread-like structure. It is more likely that these fibers repre- 
sent nothing more than lines of force emanating from the centromere. In- 
verted V-shaped chromatids are not invariably found at anaphase II. 

Some monads have chromosomal fibers only at the true centric region and 
move poleward in a normal fashion. Either neo-centric regions are not pres- 
ent, or else arise too late to be effective. It should be emphasized that a funda- 
mental distinction exists between the rod and inverted V chromatids found 
at anaphase II. The rod-shaped monads come from dyads with neo-centric 
activity at metaphase II. Their supernumerary chromosomal fibers arise 
from one arm. Their sub-terminal location suggests they may arise adjacent 
to the knob, but this is merely a conjecture. The later-formed extra chromo- 
somal fibers of the inverted V chromatids, which are knobless, are terminal 
and arise from both arms. 

If a dyad is oriented on the spindle plate at metaphase II before the onset 
of precocious neo-centromere activity, the supernumerary chromosomal 
fibers arising from the knobbed arm of the chromatid situated slightly above 
the spindle plate are directed toward the upper (nearest) pole, and those 
from the bottom chromatid go to the lower pole — they are co-oriented (see 



PREFERENTIAL SEGREGATION IN MAIZE 11 

Fig. 4.3). No such regularity is found in those infrequently occurring dyads 
which are longitudinally displaced from the spindle {)late at metaphase II. 
Their true centric regions divide prematurely. Consequently, the two 
chromatids of these displaced dyads no longer remain conjoined, but fall 
apart to become independent monads which lie side-by-side, parallel with 
the longitudinal axis of the spindle. 

The neo-centric activity which these monads now manifest is similar to 
that found at anaphase II for those monads derived from normally oriented 
dyads lacking precocious neo-centromeres at metaphase II, in that neo- 
centromeres may arise from the ends of both arms. When this occurs, the 
orientation of the two neo-centromeres of each monad is usually to opposite 
poles, but sometimes both ends of a monad are directed toward the same 
pole. Although the monads from displaced dyads have neo-centromeres at 
the end of each arm, one end being attracted to the nearest pole and the other 
to the more distant pole, normal disjunction usually occurs. This requires 
one monad to move away from the nearest pole toward which one of its ends 
is attracted, and to pass to the more distant pole, while the other monad goes 
to the nearest pole. It is difficult to interpret this phenomenon in terms of 
strength of attraction as a function of distance from centromere to pole. 

The formation of neo-centric regions requires the presence of the abnormal 
chromosome 10. In its absence, no such regions are found. It appears highly 
probable that heterochromatic knobs located on other chromosomes also are 
concerned in the formation of precocious centric regions at both meiotic 
metaphases, since the cytological observations show a correlation between 
number of knobs and number of precocious centric regions. Knobless arms 
later form neo-centric regions, but not until anaphase movement has already 
been initiated by the true centric region. 

It is possible that maize chromosomes possess latent centric regions which 
are activated by the abnormal 10. It has been demonstrated, however, that 
the true centric region is involved in the formation of neo-centromeres. 
Plants homozygous for abnormal 10 and heterozygous for the long para- 
centric inversion in chromosome 4, studied by McClintock (1938) and Mor- 
gan (1950), were obtained. Both the normal and inverted chromosome 4 
carried a large knob in the long arm which is included in the inverted seg- 
ment. Single crossovers within the inversion give rise to two non-crossover 
monocentric chromatids, one dicentric chromatid which forms a bridge at 
anaphase I, and an acentric fragment. The knobbed acentric fragment lies 
passively on the spindle with no indication of spindle fiber activity. Neo-cen- 
tromeres arise from the same chromatin segments comprising the acentric 
fragment when they constitute a portion of a whole chromosome 4. It fol- 
lows that the true or primary centromere plays an essential role in the pro- 
duction of neo-centromeres. 

The localized centromeres of maize chromosomes are concerned with the 



78 M. M. RHOADES 

elaboration of fiber-producing material. Normally this unique substance is 
confined to the true centric region, hence chromosomal fibers arise solely 
from this part of the chromosome. 

It is our belief: (1) that these centric regions produce an over-abundance 
of fiber-forming material if abnormal 10 is present in the nucleus; (2) that a 
portion of this substance escapes from the confines of the centric regions and 
moves distally along the chromosome to produce supernumerary chromo- 
somal fibers; and (3) that the knobs either stimulate centric activity or else 
cause the excess fiber-forming substance to move preferentially along knob- 
bearing arms so that neo-centric activity is first manifested by these arms. 

The failure of the acentric fragment to form chromosomal fibers suggests 
that the postulated movement of the material from the true centric region 
occurs after crossing over has taken place. If it happened prior to pachytene, 
the regions which later constitute the acentric fragments would receive some 
of this fiber-producing substance which subsequently could form spindle 
fibers. In support of the above interpretation is the observation that small 
aggregations of a substance similar in appearance to that located in the true 
centric region are sometimes found near the distal regions of some chromo- 
somes at metaphase I and metaphase II. This observation is subject to vari- 
ous interpretations. But in conjunction with the behavior of acentric frag- 
ments, it strengthens the hypothesis that the production of neo-centromeres 
is intimately related to the presence or activity of the primary centric region. 
It is obvious that the presumed movement of the products of the centromere 
along the arms of the chromosome has a bearing on the kinetic theory of Posi- 
tion Effect. 

Evidence has been presented that the abnormal chromosome 10 produces 
the phenomenon of preferential segregation, and that it also causes the for- 
mation of neo-centromeres. Are these two phenomena related — does prefer- 
ential segregation occur as a consequence of neo-centric activity? While no 
definite answer can be given at this time a tentative hypothesis has been de- 
veloped. Sturtevant and Beadle (1936), seeking to account for the absence of 
egg and larvae mortality following single crossovers in paracentric inversions 
in Drosophila, postulated that the crossover chromatids were selectively 
eliminated from the egg nucleus. The two spindles of the second meiotic divi- 
sion in Drosophila eggs are arranged in tandem. Following a crossover within 
the inverted segment, the tetrad at metaphase I consists of two non-crossover 
chromatids, a dicentric and an acentric chromatid. 

They assumed that the chromatin bridge arising from the dicentric chro- 
matid, when the homologous centromeres pass to opposite poles at anaphase 
I, ties its two centromeres together. The spatial arrangement thus produced 
is such that the two monocentric chromatids lie nearer the two poles than 
does the dicentric chromatid. 

The persistence of this relationship into the second division results in a 



PREFERENTIAL SEGREGATION IN MAIZE 79 

non-random orientation on the metaphase II spindles. The monocentric, non- 
crossover chromatids are free to pass to the two terminal poles, while the two 
centromeres from the dicentric chromatid are directed to the two inner poles. 
Consequently, at anaphase II the terminal poles each receive a non-crossover 
chromatid. Since the egg nucleus arises from the innermost terminal pole it 
would contain a non-crossover chromatid with a full set of genes. The cor- 
rectness of this ingenious hypothesis was established by Darlington and La 
Cour (1941) in Lilium and Tulipa and by Carson (1946) in Sciara. 

It is possible that a somewhat similar mechanism is operating in Zea to 
produce preferential segregation. In maize, as in Drosophila, the two 
spindles of the second meiotic division of megasporogenesis are arranged in 
a linear order. The basal megaspore of the linear set of four develops into 
the female gametophyte, the remaining three aborting. We know that in 
plants heterozygous for knobbed and knobless chromosomes, one arm of 
some of the disjoining dyads at anaphase I possess precociously-acting 
chromosomal fibers not present in the homologous arm. There is reason to 
believe that the knobbed arms form precocious neo-centromeres while knob- 
less arms do not. Owing to the rapidity with which neo-centric regions pass 
poleward at anaphase I, those chromatids with neo-centromeres reach the 
pole in advance of knobless arms lacking neo-centromeres. In a dyad con- 
sisting of one knobbed and one knobless chromatid, the knobbed chromatid 
would come to lie closer to the pole, while the knobless one would face the 
spindle plate. 

In order to account for preferential segregation, it is necessary to assume 
that this orientation persists until the second metaphase, and that it results 
in the knobbed chromatids facing the two terminal poles while the two knob- 
less ones would be oriented toward the two inner poles. On such a mechanism, 
preferential segregation would occur only when a crossover takes place be- 
tween the knob and the true centromere in a heterozygous bivalent. The 
extent of preferential segregation would be a direct function of the amount of 
crossing over in the knob-centromere region. 

Such an explanation can only be considered as a working hypothesis. It 
can be critically tested, however, and such experiments are being conducted 
by Jean Werner Morgan, who also participated in the studies reported here. 
They include varying the crossover distance between knob and centromere 
by translocation and inversion, testing for preferential segregation of hetero- 
morphic chromosomes other than chromosome 10 in plants homozygous for 
abnormal 10, determining neo-centric activity in chromatids with knobs in 
both the long and short arm, etc. I prefer not to mention her incomplete 
findings at this time, since to do so would detract from continued interest in 
her work. 

The phenomenon of preferential segregation is by no means confined to 
maize. Sturtevant (1936) found a non-random segregation of three chromo- 



80 M. M. RHOADES 

somes IV in Drosophila. Bridges, in Morgan, Bridges, and Sturtevant (1925), 
established that the distribution of the chromosomes in triploid Drosophila 
was not according to chance. Beadle (1935) reported that crossing over in 
triploid Drosophila near the centromere region between one member of at- 
tached -X's and a free X chromosome was correlated with autosomal dis- 
junction. Lower crossover values were found in 1X2,4 and XX L4 combina- 
tions than in IX 1.1 and XX 2.4 gametes. This non-random distribution 
indicates a correlated orientation of non-homologous chromosomes on the 
equatorial plate. 

In Sciara the paternal set of chromosomes moves away from the pole of 
the monocentric spindle of the primary spermatocyte. The two sister .Y 
chromosomes pass to the same pole at the second spermatocyte division 
(Metz, 1938). Schrader (1931) observed a non-random orientation in Pro- 
tortonia which led to selective distribution in secondary spermatocytes. 
Catcheside (1944), in an analysis of Zickler's data on spore arrangement in 
the Ascomycete Bombardia lunata, found that certain genes were prefer- 
entially segregated. Not all of the above examples are strictly comparable to 
the situations found in maize, Sciara, and Bombardia. In the latter cases a 
specific spindle pole receives a certain chromosome or set of chromosomes, 
while in the Drosophila cases particular chromosomes pass preferentially to- 
gether, but presumably at random, to either pole. 

The neo-centromeres arising from chromosome ends, reported in rye by 
Prakken and Muntzing (1942) and Ostergren and Prakken (1946), closely 
resemble those found in maize. In both maize and rye the neo-centric 
regions are found on arms possessing knobs (heterochromatin), and the pole- 
ward movement of neo-centromeres is precocious in both plants. Unfortu- 
nately, nothing is known about preferential segregation in rye, but it should 
occur if our hypothesis is correct. 



R. A. BRINK 

University of Wisconsin 



Chapter 5 

Inbreeding and Crossbreeding 
in Seed Development 



It is now generally recognized that the effects on growth of inbreeding and 
crossbreeding are intimately interwoven in the whole complex fabric of 
development and reproduction. Not only are the effects widespread and 
often of major consequence in the economy of the organism, but sometimes 
they are manifested in devious ways. Such is the case in the seed of flowering 
plants. 

The success or failure of seed development turns primarily, not on the 
embryo which embodies the line of descent, but upon an accessory organ of 
reproduction, the endosperm. The novel origin and sensitivity of this latter 
tissue to changes in genetic composition render early seed development one 
of the critical stages in the life cycle of flowering plants. My colleague, D. C. 
Cooper, and I have been exploring these relations during the past decade. An 
attempt will be made here to review some of the evidence upon which our 
point of view rests, and to call attention to some of the broader implications 
of the main facts. 

As a means of bringing the important aspects of the problem in flowering 
plants into focus, seed development in the angiosperms and gymnosperms 
will be compared. Essential features of the general hypothesis by which we 
have been guided will then be set fo"rth. The central role of the endosperm in 
formation of the angiosperm seed and the responsiveness of this tissue to 
variations in genetic composition will be illustrated by a consideration of the 
immediate effects of self- and cross-fertilization in alfalfa. It will then be 
shown that the means by which the embryo in the common dandelion, an 
autonomous apomict, is nourished is of a type which would be expected 
according to the hypothesis proposed. 

* Paper from the Deparlment of Genetics, College of Agriculture, University of Wiscon- 
sin, No. 432. 

81 



82 R. A. BRINK 

An illustration will next be given of endosperm failure as an isolating 
mechanism. Finally, the significance of the present results for the problem 
of artificially rearing embryos whose development in the seed is blocked by 
endosperm disfunction will be pointed out. 

Complete literature citations are not given. These may be found in the 
summary paper (Brink and Cooper, 1947) in which much additional evidence 
bearing on the present thesis also is presented. 

The endosperm is a special structure intercalated between the female 
parent and the embryo, serving to mediate the relations between the two. 
The tissue originates from the central cell of the female gametophyte, follow- 
ing a fertilization distinct from that giving rise to the embryo. The secondary 
fertilization is unusual in that two identical haploid nuclei of maternal origin 
are united with one contributed by the pollen. The endosperm thus becomes 
3x in chromosome number in contrast with the 2x condition of the embryo 
and the mother plant, respectively. Endosperm and embryo carry the same 
kinds of genes, but the genie balance may be unlike in the two tissues by 
virtue of the double contribution to the endosperm from the maternal 
parent. A further element of genetic heterogeneity in the seed arises from the 
fact that nucellus and integuments, which are maternal structures, may 
differ in genotype from the endosperm and embryo which they enclose, 
since they belong to the previous generation. 

These facts, of course, have long been known. Certain of their implica- 
tions, however, are only now becoming apparent. Particularly is this true of 
the secondary fertilization on which our attention will be focussed. 

A word should be said at this point concerning the manner in which the 
endosperm should be visualized. Many are familiar with the tissue only in 
the mature seeds of species in which the endosperm persists as a storage 
organ. This condition, well known in the cereals, for example, is exceptional 
among flowering plants, and represents a secondary adaptation of signifi- 
cance mainly for the future seedling. In most species the endosperm either 
does not persist in the fully developed seed or occurs therein as a residue 
only. On the other hand, the endosperm is regularly a prominent organ in 
the juvenile seed. It is especially active directly following fertilization, during 
what may be termed the lag phase of embryo growth. This period is seldom 
longer than a few days, and varies according to the species. In spite of its 
typically ephemeral character, the endosperm plays a critical role in (1) 
transforming the mature ovule into a young seed and (2) nourishing the 
embryo during its initial period of growth. We are here concerned with the 
endosperm in these two relationships only. 

THE SEED IN GYMNOSPERMS AND ANGIOSPERMS 

It is helpful in understanding the significance of the secondary fertilization 
to compare the circumstances of seed development in the angiosperms with 



INBREEDING AND CROSSBREEDING IN SEED DEVELOPMENT 83 

those in the other great class of seed forming plants, the gymnosi)erms. A 
secondary fertilization does not occur in the gymnosperms. The endosi)crm 
is a haploid tissue derived from the megas])ore by continuous cell division. 
The tissue is a part of the gametophyte rather than an integral structure 
distinct from both gametophyte and sporophyte, as in the angiosperms. 

On the other hand, the endosperms in the two classes of seed plants have 
an important common function, namely, nourishment of their respective 
associated embryos. The genetic equipment with which the two kinds of 
endosperms are furnished differs in a fundamental respect. That of the 
gymnosperm is a sample half of the mother plant's inheritance, whereas the 
angiosperm endosperm, being of biparental derivation, has two chances in- 
stead of only one of receiving a physiologically effective genie complement. 
Insofar as the two tissues are autonomous in their functional properties, the 
angiosperm endosperm, therefore, is equipped to meet much more exacting 
requirements than its counterpart in the gymnosperms. A summary review 
of the differences in the gymnosperm and angiosperm ovules and seeds at 
fertilization, and during the immediately subsequent period, shows the im- 
portance of (or necessity for) a secondary fertilization in the flowering 
plants in order to maintain continuity of the life cycle at this stage. 

The differences between the mature ovules of gymnosperms and angio- 
sperms which appear to have a direct bearing on the present problem may be 
summarized as follows: 

1. The seed coat in the gymnosperms approaches its mature size at the 
fertilization stage. The angiosperm seed coat undergoes extensive growth sub- 
sequent to fertilization. These facts are of interest in relation to the total 
food requirements of the two respective classes of growing seeds and the 
post-fertilization distribution of nutrients between the seed coat and the 
enclosed tissues. 

2. The female gametophyte in the gymnosperms is an extensively de- 
veloped multicellular (multinucleate, in some higher forms) structure. Its 
counterpart in the angiosperms typically consists of only seven cells. The 
potential disadvantage of the extreme reduction of the female gametophyte 
in the flowering plants will be considered below. 

3. Generally speaking, the gymnosperm ovule is rich in food reserves, 
whereas the angiosperm ovule is sparsely supplied. This means that in the 
latter, the large volume of nutrients required for growth of the endosperm, 
embryo, and seed coat must be moved in from other parts of the plant. In 
the gymnosperms an extensive supply is directly at hand. 

4. So far as may be inferred from the published accounts, fertilization in 
the gymnosperms initiates a new cycle of growth in the embryo only. Other 
parts of the ovule do not appear to be stimulated. Double fertilization in the 
angiosperms, in contrast, not only marks the inception of endosperm and 
embryo formation, but also incites pronounced mitotic activity and en- 



84 R. A. BRINK 

largement of the cells in the integuments. Thus, with the exception of the 
nucellus which is broken down and absorbed by the rapidly expanding endo- 
sperm, all the elements of the young seed which were previously quiescent, 
suddenly spring into active growth following syngamy. 

Consideration of these differences between the seeds of gymnosperms and 
angiosperms led us some ten years ago to explore the hypothesis that the 
secondary fertilization in angiosperms is essentially a means of enhancing 
the competitive power of the endosperm relative to the maternal portions of 
the seed — by conferring upon the endosperm the advantages of hybridity. 
The nutritive requirements of the young seed suddenly are raised from a low 
to a high level since fertilization starts a new cycle of growth in the massive 
integuments. The nutrient supply, on the other hand, quickly falls to the 
plane which can be maintained by movement of foods into the seed from 
other parts of the plant as a result of exhaustion of the limited ovule reserves. 

It seemed reasonable to assume that, within the seed, the incoming nutri- 
ents would tend to be partitioned between the different tissues according to 
the respective amounts of growth occurring in them. On this basis, the ex- 
tensively developed integuments would consume the major portion. The 
diminutive endosperm and embryo would receive but a small fraction of the 
total. Under these conditions, failure of the young seed through starvation of 
the embryo could arise, unless the endosperm — as the nutritive agent of the 
embryo — were endowed with special properties which offset its initially small 
size. It seemed essential that the endosperm, by one means or another, be 
enabled to quickly acquire a position of physiological dominance in the 
juvenile seed in order to insure continued development. 

Two genetic characteristics of the endosperm suggest themselves as being 
important in this connection. The first is the triploid condition of the nuclei. 
Little is known of the physiological effects of ploidy in general, and virtually 
nothing of its meaning in special situations of this kind. One suspects, how- 
ever, that the endosperm gains some advantage from its extra chromosome 
garniture, as such, in mediating the relations between the diploid maternal 
parent and the young diploid embryo. It is also probably significant that, 
whereas the embryo inherits equally from the two parents, two-thirds of the 
endosperm's genie complement is derived from the plant upon which it is 
nutritionally dependent and one-third of the complement from the male 
parent. 

Heterozygosis is the second characteristic of the endosperm which might 
enhance the inherent physiological efficiency of this tissue. The possibility of 
heterozygosity arises, of course, from the biparental origin of the endosperm 
mother nucleus. The condition is realized in matings between genetically 
different plants. Haploidy of the endosperm, as occurs in the gymnosperms, 
appears to be genetically insufficient for seed development in the flowering 
plants. Early post-fertilization circumstances, particularly the dependence 



INBREEDING AND CROSSBREEDING IN SEED DEVELOPMENT 85 

upon and competition for an outside nutrient supply in the latter, require 
that the tissue shall share in the advantages of sexuality. The advantage 
gained is not that of amphimixis in general, as in the embryo, but solely the 
extra vigor of growth associated with the union of unlike nuclei in the mother 
cell. Thus hybrid vigor in the endosperm has some claim to uniqueness. The 
sole object gained by entry of a sperm into the nuclear makeup of this sterile 
tissue is the added vigor of growth thus acquired. Some of the evidence by 
which the validity of this point of view may be tested will now be considered. 

INBREEDING AND CROSSBREEDING EFFECT ON 
SEED COLLAPSE IN MEDICAGO SATIVA 

Two classes of matings on seven alfalfa plants were carried out under 
favorable growtli conditions in a greenhouse. After removal of the anthers 
from the flowers used, a part of the flowers were pollinated with pollen from 
the same respective plants. This constitutes the self-fertilized series. Other 
flowers on the same plants were cross-pollinated, the pollen being derived in 
each case from an unrelated plant within the group. These matings comprise 
the cross-fertilized series. 

Since alfalfa is regularly cross-fertilized, the second series of matings is 
designed to maintain the level of heterozygosity normal to the endosperm 
and embryo in this species. The enforced self-fertilization, on the other 
hand, would be expected to reduce heterozygosity in the endosperm mother 
nucleus and the zygote by 50 per cent. It is proposed to review the conse- 
quences for seed development of this sharp reduction in heterozygosis. 

Following the above two series of matings, the pistils were collected at 30, 
48, 72, 96, 120, and 144 hours and imbedded in paraffin. After sectioning and 
staining, data were taken on fertility of the ovules, frequency of fertile ovules 
collapsing, number of cells in the embryo, and number of nuclei in the endo- 
sperm. Detailed observations were made subsequently on growth of the 
integuments. 

Alfalfa was known previously to be partially self-incompatible. It was 
not unexpected, therefore, to find that only 15 per cent of the ovules became 
fertile after selfing in contrast to 66 per cent after cross-pollination. The new 
fact which emerged was the much higher incidence of collapse of ovules sub- 
sequent to fertilization in the selfed than in the crossed group. The data are 
summarized in Table 5.1. Fertilization occurred within about 30 hours after 
pollination under the prevailing conditions. It was somewhat delayed after 
selfing. Little evidence of breakdown of the seeds was found at 48 hours. In 
the 72 hour and subsequent collections, however, the phenomenon was com- 
mon. The results presented in the table cover the period from 72 hours to 
144 hours, inclusive, and are based upon 433 seeds and 1682 seeds in the 
selfed and crossed series, respectively. 

Growth of the young seed at this stage appears to be quite independent 



86 



R. A. BRINK 



of that of its neighbors in the same ovary. Furthermore, the quickly succeed- 
ing secondary effects of fertilization, such as enlargement of the surrounding 
fruit, are at a minimum. Studies on the reproductive physiology of the flower- 
ing plants are rendered difficult by the multiplicity of changes which are 
eventually set in motion in the tissues of the seed, the fruit, and the maternal 
plant following fertilization. The sequence and interrelations of the events 
immediately subsequent to syngamy are simpler to analyze than those which 
occur later, in view of the fact that each very young seed may be considered 
to behave independently of the others. 

The data in Table 5.1 show that, for each of the seven plants tested, the 

TABLE 5.1 

FREQUENCY OF FERTILE OVULES COLL.APSING IN SEVEN .ALFALFA 
PLANTS FOLLOWING SELF- AND CROSS-FERTILIZATION. DATA BASED 
ON COLLECTIONS AT 72, 96, 120, AND 144 HOURS .\FTER POLLINATION 
(AFTER COOPER AND BRINK, 1940) 



Self-fertiliz.'\tion 


Cross-fertiliz.ation 


Plant 


No. of Fertile Ovules 


Percentage 
Collapsing 


Plants 
Crossed 


No. of Fertile Ovules 


Percentage 


Selfed 


Total 


Collapsing 


Total 


Collapsing 


Collapsing 


A 

B 

C 

D 

E 

F 

G 


37 
37 
20 
17 
39 
109 
55 


9 

19 

7 

7 

8 

39 

19 


24.3 
51.4 
35.0 
41.2 
20.5 
35.8 
34.5 


AXB 

BXC 

CXD 

DXE 

EXA 

FXG 

GXF 

Total. . 


187 
110 
171 
171 
146 
228 
198 


13 
5 

13 
16 
9 
14 
16 


7.0 
4.5 
7.6 
9.4 
6.2 
6.1 
8.1 


Total. . 


314 


108 


34.4 


1211 


86 


7.1 



frequency of seeds collapsing is much higher in the selfed than in the crossed 
series. The proportions vary in different individuals from about 3 to 1 to 
over 11 to 1. On the average, approximately five times as many seeds con- 
taining inbred endosperms and embryos collapse within the first six days 
after pollination as in the crossbred group. Since other factors were not 
varied, the decrease in survival in the selfed series must be attributed to 
the inbreeding. 

The evidence, both general and particular, points to the endosperm 
rather than the embryo as the seat of the inbreeding depression effect. The 
endosperm in alfalfa is free nucleate up to about 144 hours after pollination, 
although it develops as a cellular tissue thereafter. Successive waves of 
mitotic divisions traverse the tissue, the number of nuclei being doubled in 
each cycle. Thus growth during this period proceeds at an exponential rate. 



INBREEDING AND CROSSBREEDING IN SEED DEVELOPMENT 87 

The concurrent development of the embryo, on the other hand, is relatively 
slow. The zygote divides to form a two-celled proembryo. Successive divi- 
sions of the apical cell give rise first to a six-celled proembr>'o and then to the 
initials of the definitive embryo. 

The pronounced difference in rate of development of the two tissues is il- 
lustrated by the fact that at 144 hours the modal number of cells in the 
embryo is only 16, whereas the typical number of nuclei in the endosperm 
at this time is 128. Rapid and precocious development of the endosperm as 
seen in alfalfa is characteristic of the angiosperms in general. The much 
higher level of activity of the endosperm is presumptive evidence tlvit this 
tissue, rather than the embryo, is especially subject to develo{)mental upsets 
in the young seed. Data available in the present instance provide direct con- 
firmation of this interpretation. 

The comparative rates of growth of endosperm and embryo in the selfed 
and crossed alfalfa series up to 144 hours after pollination are illustrated in 
Figure 5.1. Not only are the values for the embryo low, but also there is 
little difference between those for the inbred and crossbred series. The con- 
clusion appears warranted that the direct effect of inbreeding on the embryo 
at this stage, if indeed there is a demonstrable effect, is too small to account 
for the high frequency of seed collapse. In contrast, there is a very sharp 
decline in rate of nuclear division in the endosperm, following enforced self- 
fertilization of this naturally cross-fertilized plant. The lower rate is shown 
from the first division onward. There are about twice as many nuclei present 
at 144 hours in the crossbred as in the inbred endosperms. 

Due to the partial self-incompatibility in alfalfa, fertilization on the 
average, is slightly delayed following selfing. A comparison of the rate of 
growth of the two classes of endosperms independent of time as shown in 
Figure 5.2, however, establishes the reality of the difference in rate of growth 
between the inbred and crossbred endosperms. When the seeds are arrayed 
in terms of cell numbers of the enclosed embryo, it is found that for all nine 
classes occurring in the material the endosperms are more advanced in the 
crossbred than in the inbred series. That is to say, the embryos at a given 
stage of development have associated with them more vigorously growing 
endosperms following cross-fertilization than after selfing. Moreover, the 
decrease in size resulting from the inbreeding is so large that one is led im- 
mediately to suspect that herein lies the primary cause of the frequent seed 
collapse following selfing. 

Why should impairment in rate of endosperm growth lead to arrested seed 
development? The answer in the present case is clear. As was pointed out 
earlier, double fertilization initiates not only endosperm and embryo develop- 
ment, but also a new cycle of growth in the integuments. The latter compete 
directly with the endosperm for the nutrients moving into the young seed. 
If the endosperm is developing subnormally, a disproportionate amount of 



88 



R. A. BRINK 



the incoming nutrients is diverted to the integuments. As a result this 
tissue frequently becomes hyperplastic. The overgrowth in the case of al- 
falfa characterizes the inner integument. As Dr. Cooper observed, it begins 
at a point opposite the distal end of the vascular bundle where the concen- 
tration of nutrients maybe assumed to be the greatest. The inner integument, 
which is normally two cell layers in thickness, becomes multilayered and 
somewhat callus-like in the region of the greatest mitotic activity. This pro- 
nounced overgrowth of the inner integument quickly reacts upon the endo- 
sperm, further impairing its development. In the seeds which fail, a complete 




48 72 % 120 

Time in hours after pollinatior\ 



m 



Fig. 5.1— Increase in number of cells in embryo and in number of nuclei in endosperm 
following self- {broken line) and cross-fertilization {continuous line). After Brink and 

Cooper, 1940. 



INBREEDING AND CROSSBREEDING IN SEED DEVELOPMENT 



89 



collapse of the endosperm then ensues. Significantly, breakdown of the 
endosperm tissue begins in the region opposite the end of the vascular 
bundle where the inner integument is especially hyperactive. Following col- 
lapse of the endosperm, the young seed dies. 

SEED DEVELOPMENT WITHOUT FERTILIZATION 

There are a few species of flowering plants in which both endosperm and 
embryo develop without fertilization. These so-called autonomous apomicts 



1-celled 
proembri/o 

2-celled 
proem bryo 

3-celled 
proembryo 

4-celled 
proembryo 

5-ceiied 
proembryo 

6-celied 
proembryo 



2-celied 
emb.*5usp. 



3-ceiled 
emb.-t-5u5p. 

4-celled 
emb.+sasp. 



1 5elf-fcrtilization 
cross-fertilization 




10 20 50 40 50 60 70 

Number of endosperm nuclei 

Fig. 5.2 — Number of endosperm nuclei associated with proembryos and embryos at various 
stages of development following self- and cross-fertilization. After Brink and Cooper, 1940. 

should provide an independent test of the hypothesis that aggressive develop- 
ment of the endosperm is requisite to seed development, and that the sec- 
ondary fertilization is a device by which aggressiveness of the tissue is en- 
hanced. On the basis of the reasoning applied to sexual species, one would 
e.xpect to find in autonomous apomicts that the embryo is not basically de- 
pendent on an active endosperm for its nourishment. So far as I am aware, 
the evidence bearing directly on this question is limited to a single study 
which Cooper and I carried out on the common dandelion, Taraxacum 
officinale (Cooper and Brink, 1949). 

The common dandelion is triploid (3x = 24). The regularity and abun- 
dance of seed production in the plant is well known. A full complement of seed 



90 R. A. BRINK 

forms in the absence of pollination, as may be demonstrated easily by re- 
moving the corollas and anthers — by cutting off the distal portion of the 
head in the bud stage. Ordinarily the anthers do not open in the intact 
mature flower. 

The female gametophyte is formed without reduction in chromosome 
number of the nuclei. Otherwise it is a typical eight-nucleate, seven-celled 
structure lying in direct contact in the mature ovule with the innermost 
layer of cells of the single thick integument. The polar nuclei fuse to give a 
hexaploid primary endosperm nucleus. The single layer of cells comprising 
the nucellus disintegrates during formation of the embryo sac. 

Sexual forms of the common dandelion are not known to occur. Accord- 
ingly another species, T. kok-saghyz, the Russian dandelion, was examined 
as a control. T. kok-saghyz is diploid (2x = 16) and, since it is self-incom- 
patible, requires cross-pollination for seed formation. A comparative study 
of T. officinale and T. kok-saghyz was made with a view to discovering, if 
possible, the means by which the former is enabled to dispense with the 
secondary fertilization, which is essential to seed formation in the latter. 
Heads were collected at four stages: late bud, just prior to anthesis, open 
flower, and with seeds ranging up to six days of age. After sectioning and 
staining, the number of cells in the endosperm and embryo was determined, 
and observations were made on the amount and distribution of food ma- 
terials. 

Seed formation in T. kok-saghyz follows the course typical of the angio- 
sperms. Endosperm and embryo development are initiated by double 
fertilization. Subsequently, the two tissues grow very rapidly, and in tune 
with each other. Cell number in the endosperm increases exponentially. The 
endosperm, however, is somewhat less precocious than in most flowering 
plants. The seed is mature 9-12 days after fertilization. 

A markedly different set of relations present themselves in the seed of 
the apomictic T. officinale. The seed in this species begins development when 
the flowers are in the late bud stage. By the time the flowers open, there may 
be 100 cells or more in the endosperm, the embryo, or in both tissues in some 
seeds. A further significant fact is the extraordinary amount of variability 
in the size ratios of endosperm and embryo from seed to seed of even age. 
There is a positive relation between cell number in endosperm and embryo 
over the period studied — as would be expected in view of the fact that in 
most seeds both tissues are growing. As measured by the correlation co- 
efficient, this value is low (r = .57) compared with that for T. kok-saghyz 
(r = .76). 

Average cell number in the embryo in relation to endosperm size is de- 
picted for the two species in Figure 5.3. Cell number in the endosperm in- 
creases geometrically, so that size of the tissue may be expressed appropriate- 
ly in terms of division cycles. Embryo cell number, in contrast, increases 



INBREEDING AND CROSSBREEDING IN SEED DEVELOPMENT 



91 



arithmetically. It will be noted from Figure 5.3 that the mean embryo cell 
number in T. officinale, before the endosperm mother cell divides (0 cycle), 
is about 16. The corresponding value T. kok-saghyz is 1. This is a reflection of 
the fact that the embryo in the apomictic sj)ecies usually starts growth in 
advance of the endosperm. Although they start from different levels, the two 
curves are not greatly dissimilar. The embryo in the common dandelion, on 
the average, is consistently larger in the young seed than that of T. kok- 
saghyz, relative to given stages in endosperm develo])ment. 

100 

T. officinale 

T. kok-soghyz 



o 


90 


>- 




oc 




m 


60 


z 




UJ 




z 


70 


UJ 


60 


m 




z 


50 


z> 




■^ 




_J 


40 


-I 




UJ 

o 


30 


2 


20 


< 




UJ 




Z 


10 



4 



K.SJ 



12 3 4 5 6 7 

NUMBER OF DIVISION CYCLES IN ENDOSPERM 

Fig. 5.3 — Early growth of embryo of T . kok-saghyz and T. officinale in relation to endo- 
sperm size. After Cooper and Brink, 1949. 

More instructive than the mean values on which Figure 5.3 are based, is 
the variability in the frequency distributions concerned. The data are sum- 
marized in Table 5.2. A logarithmic scale was used in expressing embryo 
sizes merely as a convenient way of summarizing the widely dispersed values. 
As mentioned above, growth of the embryo during this period is approxi- 
mately linear. 

Table 2 reveals that the variability is low in embryo cell number at suc- 
cessive stages of endosperm development in T. kok-saghyz. This means that 
embryo and endosperm are closely synchronized in their growth in the sexual 
species. The variability in embryo size in the apomict, on the other hand, is 
enormous. For example, in seeds in which the endosperm is still at the mother 
cell stage (0 cycle), the associated embryos are distributed over all size 
classes from 1 to 128. The standard deviation for embryo cell number is 



92 



R. A. BRINK 



15.6, a value equal to the mean. The range is even greater in the class of 
seeds having 128-cell endosperms, and the standard deviation rises to 51 
cells. 

The extreme variability in embryo size for given stages of endosperm de- 
velopment in T. officinale is a fact of cardinal importance in the present 
analysis. Inspection of Table 5.2 reveals certain details which emphasize 
the significance of the summary data on dispersion. Note, for instance, that 

TABLE 5.2 
DISTRIBUTION OF EMBRYOS BY CELL NUMBER RELA- 
TIVE TO ENDOSPERM DIVISION CYCLE 
(AFTER COOPER AND BRINK, 1949) 



Endo- 
sperm 


Species 


Total 
Seeds 
Ex- 
amined 


Embryo Cell Number — Logarithmic 
Class Values 


Stand- 
ard 


Division 
Cycle 


1 


2 


4 


8 


16 


32 


64 


128 


256 


Devia- 
tion 





T. officinale 
T. kok-saghyz 

T. officinale 
T. kok-saghyz 

T. officinale 
T. kok-saghyz 

T. officinale 
T. kok-saghyz 

T. officinale 
T. kok-saghyz 

T. officinale 
T. kok-saghyz 

T. officinale 
T. kok-saghyz 

T. officinale 
T. kok-saghyz 


111 
All 

253 

77 

145 

32 

108 

25 

111 
34 

115 
68 

99 

55 

60 
19 


9 

All 

23 
31 

18 

1 

12 


16 


ii 


57 


66 


38 


7 


1 




15.6 



1 


11 
46 

11 
31 

6 


37 


70 


55 


50 


6 


1 




13.6 
0.5 


2 


7 


23 


43 


IZ 


9 


1 




17.0 
0.2 


3 


6 

22 

2 
5 

4 


19 

3 

9 

27 

4 
24 


27 


25 


12 


1 




21.1 
0.7 


4 


4 


1 


39 

2 

23 
40 

7 
10 

1 


40 


14 


2 




19.2 
1.9 


5 


6 




50 
4 

31 
41 

8 
3 


23 


5 




24.2 
4.1 


6 


1 


1 




46 

4 

17 
16 


13 




29.9 








9.0 


7 


2 




1 




28 


3 


51.0 
16.7 





















among the seeds still in the endosperm mother cell stage (0 cycle) one con- 
tains an embryo in the 128-cell class and seven have embryos in the 64-cell 
class. Similar, although less extreme, cases occur in the 1 -cycle and 2-cycle 
endosperm distributions. Study of the histological preparations shows that 
the seeds in which the embryos are found are growing vigorously and appear 
capable of completing development. This can mean only that either very 
small endosperms in T. officinale are extraordinarily efhcient structures, or 
embryo growth in this species is not dependent on an endosperm. 

At the opposite corner of the table, on the diagonal, two seeds are entered 
in the 7-cycle endosperm array in which the embryos are still in the one-cell 
stage. These seeds also appeared to be healthy and capable of continued 



INBREEDING AND CROSSBREEDING IN SEED DEVELOPMENT 93 

development. These extreme examples {)oint unmistakably to the conclusion 
that in the apomictic dandelion the endosperm, as the master tissue in the 
young seed, has been disestablished. Embryo growth must be sustained by 
other means. 

The substitute arrangement for nourishing the embryo in T. officinale was 
disclosed by a histological study of the ovules of this species and T. kok- 
saghyz. Basically the structure of the ovule is the same in both. As the female 
gametophyte expands, the nucellus disintegrates so that the gametophyte 
comes to lie in direct contact with the endothelium which comprises the in- 
nermost layer of cells of the massive integument. The endothelium [)ersists 
and appears to function in the transfer of nutrients during the course of seed 
development. In T. kok-saghyz the inner layers of integumentary cells ad- 
jacent to the endothelium lose their contents during formation of the game- 
tophyte, and contain shrunken and misshapen nuclei when the ovule is 
mature. The cells of the integument immediately surrounding this depleted 
region are densely cytoplasmic and possess well-defined nuclei. The outer- 
most parenchymatous cells of the integument are highly vacuolate. The 
single vascular bundle makes an arc about the greatest circumference of the 
ovule in both species. Only limited amounts of stainable reserve food ma- 
terials occur anywhere in the T. kok-saghyz ovule. 

The T. officinale ovule differs conspicuously from that of T. kok-saghyz in 
possessing an abundance of reserve food. The cells of the integument just 
outside the endothelium enlarge as the ovule matures and become gorged 
with a homogeneous material which appears to be proteinaceous in composi- 
tion. This substance also extends between the cells at the outer edge of the 
storage region proper. 

This extensive prestorage of protein-rich food material in the integument 
provides an explanation of the fact that embryo development in the apomict 
may proceed normally in spite of very limited endosperm growth. The con- 
ditions render superfluous an aggressively functioning endosperm. The 
embryo draws directly on a food supply already at hand. From the physio- 
logical point of view, the nutritive mechanism in the apomict is analogous to 
that in the gymnosperms. In both these classes of plants certain of the 
processes essential to seed development, which follow double fertilization in 
sexual species of flowering plants, are pushed back into the ovule. The 
secondary fertilization, which through its efifect on vigor of endosperm growth 
may be looked upon as a means of offsetting the tardy provision of nourish- 
ment for the embryo, thus can be dispensed with. 

SEED DEVELOPMENT GRADE AND EMBRYO 
GROWTH POTENTIALITIES 

The conclusion that growth of the angiosperm seed is basically controlled 
by the endosperm has an interesting corollary. That is, that the grade of seed 
development attained after a given mating is not a definitive index of the 



94 R. A. BRINK 

intrinsic vigor of the embryo. This statement is not intended to imply that 
the two phenomena are unrelated, but rather that they vary independently 
of each other to a significant degree. Many interspecific matings, for example, 
yield poorly developed seeds. Often the embryos in these seeds give rise to 
relatively weak plants. Sometimes, however, the embryos within such seeds 
are capable of forming plants of great vegetative vigor. In other words, the 
fact that development of the seed is impaired, even to a degree that calls for 
special methods of propagation, does not necessarily mean that the embryo is 
intrinsically weak. The hybrid during the seed stage may merely be the 
victim of a faulty endosperm. Only when released from this stricture can the 
inherent potentialities of the new individual be expressed. 

Two examples of such intrinsically vigorous hybrids in which the condi- 
tions of seed development have been explored will be briefly mentioned. They 
differ in the grade of seed development attained. Small but nevertheless 
germinable seeds are formed in the one case, whereas in the other the embryo 
egularly dies unless special precautions are taken to save it. 

Cooper and I found that when the diploid (2» = 24) Red Currant tomato, 
Lycopersicon pimpinelUfolium, is pollinated with a particular strain of L. 
peruvianium., likewise a diploid, fertilization occurs with high frequency but 
all the seeds collapse before the fruit is ripe. Seed development follows a 
familiar pattern. The endosperm grows less vigorously than in normal L. 
pimpinelUfolium seeds, and the endothelium enclosing it tends to become 
hyperplastic. Endosperm cells become highly vacuolate and starved in ap- 
pearance. Densely staining granules of unknown composition accumulate in 
the chalazal region just outside the endosperm, suggesting that the latter 
tissue is incapable of absorbing the available supply of nutrients. All the 
seeds in the ripe fruit are shrivelled and incapable of germination. 

Following the application of pollen from the same diploid strain of L. 
peruvianium to a tetraploid (2w = 48) race of L. pimpinelUfolium, about 
one-half the fertile ovules develop into small but germinable seeds containing 
triploid embryos. The other seeds collapse at various stages of growth. Histo- 
logical examination of the 4n L. pimpinelUfolium X 2n L. peruvianium seeds 
shows retarded embryo development and a less rapid endosperm growth 
than occurs in the normally pollinated tetraploid parent. The endosperm in 
sixteen-day-old hybrid seeds lacks the rather densely packed starch reserves 
characteristic of tomato seeds at this stage. The peripheral layers of endo- 
sperm cells adjacent to the endothelium break down. An unusually large 
cavity is formed in the interior of the tissue as a result of digestion of the 
cells by the slowly differentiating embryo. Endosperm function is markedly 
impaired in this cross, but in many seeds remains somewhat above the 
threshold at which complete failure occurs. 

The triploid plants resulting from germinable 4» L. pimpinelUfolium X 
2w L. peruvianium seeds are extraordinarily vigorous. Although partially 



INBREEDING AND CROSSBREEDING IN SEED DEVELOPMENT 95 

sterile, they considerably exceed both the parents in capacity for vegetative 
growth. The inference is clear that the genie combination resulting from this 
cross yields markedly different results in the endosperm and the sister 
sporophy te. The difference in part may be a consequence of the 2 : 1 balance 
of L. pimpinellifolium and L. peruvianium genes in the embryo as compared 
with the 4: 1 ratio in the endosperm. The important point, however, is that 
the mechanism of seed formation in the flowering plants is such that the two 
products of a given double fertilization may be quite differently endowed in 
terms of the genes necessary to perform their respective functions. 

The second example to be discussed in this connection will enable us to 
visualize the limits which may be reached in endosperm disfunction with 
retention of embryo viability. 

Fertilization freely occurs when squirrel-tail barley, Hordeum jubatum is 
pollinated by cultivated rye, Secale cereale. The resulting seeds all die, how- 
ever, within less than two weeks. Space does not permit me to recount here 
the steps leading to the breakdown. They have been described in detail else- 
where (Cooper & Brink, 1944; Brink & Cooper, 1944). The endosperm early 
becomes completely disorganized. Some of the embryos formed, however, 
reach a stage previous to collapse at which time they may be dissected from 
the seed and successfully reared on an artificial nutrient medium. A single 
plant was grown to maturity from an embryo treated in this way. The 
plant was thrifty, although sterile. Representatives of the parent species 
grown under comparable conditions were not available, so that a valid com- 
parison of relative vigor could not be made. The hybrid, however, appeared 
to be intermediate in stature and number of tillers. 

The extreme character of the endosperm disturbances in the H. jubatum X 
S. cereale seed indicates that this hybrid could not arise under field condi- 
tions. Although the embryo is demonstrably capable of continued develop- 
ment its growth is terminated in the seed due to failure of the associated 
endosperm. Death of the embryo, as an indirect result of endosperm disfunc- 
tion following wide crosses, appears to be commoner than was thought before 
the physiological implications of the secondary fertilization in flowering 
plants were recognized. Realization of this fact has stimulated additional 
interest in circumventing the phenomenon by excising such embryos from 
the seed and rearing them artificially. 

Artificial methods of cultivating embryos removed from abortive seeds 
often have been used to extend the area within which gene transfers may be 
effected. Numerous interspecific hybrids have thus been grown which other- 
wise are not realizable. The nature of the general problem involved may now 
be seen in somewhat broader perspective. Two points of particular interest 
may be noted. 

The first, briefly adverted to above, is that the frequency with which em- 
bryos are formed following matings between distantly related plants is much 



96 R. A. BRINK 

higher than earlier believed. Various investigators have expressed the opinion 
that the mere presence of growing pollen tubes in the style causes enlarge- 
ment of the ovules. This view now appears to be incorrect. 

On the other hand, there is a steadily increasing amount of evidence to 
show that the incipient growth of the ovules, following many interspecific 
matings which do not yield functional seeds, is a response to fertilization. 
That is to say, the block in the reproductive cycle which was assumed to 
intervene prior to fertilization actually occurs following syngamy. Embrycs 
are formed in these cases, but they perish when the young seed fails to de- 
velop. Some rather extreme examples of this phenomenon which have been 
observed in our laboratory include Nicotiana ghitinosa X Petunia violacea, 
N. glntinosa X Lycopersicon esculentum, and Medicago saliva X M. scutellata. 

It is not to be inferred that all hybrid embryos of this general class are 
capable of growing into mature plants. The fact that the seeds containing 
them collapse is not proof, however, of intrinsic inviability. An unknown but 
probably significant proportion of these novel zygotic combinations are po- 
tentially propagable. The problem is to discover the means by which they 
may be reared. This brings us to the second point — the nature of the problem 
to be faced in growing very small excised embryos. 

With few exceptions, the embryos which have been successfully culti- 
vated artificially have been removed from the seed at rather advanced stages 
of development. Unless they are multicellular and differentiation has at least 
begun, the embryos usually do not grow on the media which thus far have 
been devised. There are reasons for thinking that the nutritional require- 
ments of these older embryos are simpler than those in a juvenile condition. 
Histological evidence shows that at the early stages of seed development the 
embryo is enclosed, or nearly enclosed, in the highly active, young endo- 
sperm. The endosperm cells adjacent to the proembryo and the very young 
embryo remain intact. A little later, as the embryo enlarges, these cells 
begin to break down and their contents disappear. Eventually all the endo- 
sperm tissue is consumed in most species. 

One may infer from these facts that the embryo is dependent upon the 
endosperm for certain metabolites which initially the embryo is quite in- 
capable of synthesizing. The endosperm may be pictured as secreting the 
needed materials at the early post-fertilization stage, and yielding them 
later in a more passive fashion as the tissue becomes lysed. Meanwhile the 
embryo becomes progressively less dependent upon the endosperm by acquir- 
ing for itself the synthetic capabilities previously limited to the nurse tissue. 
On this view the very young embryo is an obligate parasite on the endo- 
sperm. Once past the state of obligate parasitism, growth of the embryo may 
be effectively supported by comparatively simple nutrients such as may be 
provided in artificial culture media. 

Visualized in those terms, the problem of cultivating very young, excised 



INBREEDING AND CROSSBREEDING IN SEED DEVELOPMENT 97 

embryos resolves itself into the discovery of means of duplicating the un- 
known but presumably special nutritive functions of the normal endosperm. 
Two possibilities suggest themselves in this connection. One is to determine 
natural sources of the special metabolites produced by the endosperm and 
then add these materials to the nutrient medium. Van Overbeek (1942) ob- 
tained significant improvement in the growth of small Datura stramonhim 
embryos by supplying them with unautoclaved coconut milk. Blakeslee and 
Satina (1944) later reported that the coconut milk could be replaced by un- 
autoclaved malt extract. The other possibility is to cultivate the embr>^os 
artificially in association with actively functioning endosperm tissue. Cur- 
rent findings offer some encouragement that the latter procedure may prove 
efficacious. 

Dr. Nancy Ziebur, working in our laboratory, recently has shown that 
the growth of very young embryos of common barley (0.3-1.1 mm. long) 
may be greatly improved by surrounding them on a nutrient agar medium 
with aseptically excised endosperms. The basic medium employed permits a 
satisfactory growth of older barley embryos but does not yield transplantable 
seedlings from embryos shorter than about 0.6 mm. except in conjunction 
with endosperms. Coconut milk and malt extract are ineffective with barley 
embryos. Water extracts of fresh barley endosperms gave positive, although 
smaller effects than the intact tissue. Further exploration of the living endo- 
sperm as a source of nutrients for very young, excised embryos should prove 
rewarding. The interrelationships of these two tissues in the juvenile seed 
give strong credence to this approach. The success which has so often at- 
tended efforts to grow older embryos artificially on rather simple media may 
have blinded us to the fact that the young embryo, divorced from the endo- 
sperm, may have quite different requirements. 



W. GORDON WHALEY 

The Plant Research Insfifufe of fhe Universify of Texas and 
the Clayton Foundation for Research 



Chapter 6 

Physiology of Gene 
Action in Hybrids 



The physiology of gene action in hybrids is not a subject apart from the 
physiology of gene action in organisms in general. The approach to specific 
problems of gene action is probably better made in non-hybrid organisms 
than in hybrids. Hybrids do, however, represent one type of genetic situation 
which in certain instances is particularly favorable for the study of gene 
action. Most useful in this respect are those hybrids which exhibit the phe- 
nomenon referred to, often rather loosely, as hybrid vigor. The terms hybrid 
vigor and heterosis often are used synonymously. A more precise usage, and 
one in accord with the original definitions, refers to the developed superior- 
ity of hybrids as hybrid vigor, and to the mechanism by which the superior- 
ity is developed as heterosis. By this definition, hybrid vigor is heterosis 
manifest. Because in studies of growth and development it is often desirable 
to distinguish clearly between mechanism and end result, this use of the two 
terms will be followed in this chapter. 

Heterosis has been the subject of many experiments and a great deal of 
speculation on the part of geneticists. The concern has been mostly with the 
genetic bases of heterosis, and relatively little attention has been given to the 
physiological mechanisms involved. As a matter of fact, the literature on 
heterosis mirrors faithfully the changing emphasis in genetics in the last two 
or three decades. Practically all of the early investigations of heterosis had 
to do with the comparison of mature characteristics of inbred lines and their 
vigorous hybrids, and then with attempts to formulate genetic schemes in 
explanation of the differences. Gradually, the focus of investigation has 
turned to a study of developmental differences responsible for the hybrid 
vigor, and more recently to the gene action bases of these developmental 
differences. 

98 



PHYSIOLOGY OF GENE ACTION IN HYBRIDS 9^ 

It is a fair hope that from detailed studies of the nature and development 
of heterosis, much will in time be revealed about specific gene action. Un- 
fortunately, most of the studies up to the present time have been directed 
to general rather than to specific considerations. It has been necessary to deal 
in terms of size differences, yield differences, and growth rate differences, un- 
til enough of the pattern should appear to indicate what specific physio- 
logical considerations are likely to be involved in heterosis. Because we have 
come only to this point and have proceeded but a little way in an analysis of 
these specific physiological considerations, this chapter will have to deal 
more with suggestions of the likely mechanisms than with data from investi- 
gations of them. 

It is neither possible nor desirable to separate wholly the consideration of 
the physiological mechanisms of heterosis from the genetic bases. Our main 
concern will ultimately be with the genes involved and the nature of their 
action. 

The word hybrid has no good, definitive genetic meaning. It can be used 
with equal propriety to refer to organisms which approach complete hetero- 
zygosity or to organisms which are heterozygous for only a small number 
of genes. 

There is at least a rough relationship between the amount of heterosis in a 
hybrid and the extent of the genetic differences between the parents. Physio- 
logical and morphological diversity are dependent both upon the number of 
allelic differences between organisms and upon the nature of the action of 
the particular genes among which these allelic differences exist. It is quite 
possible that organisms differing by only a few genes may be more widely 
separated in certain characteristics than are organisms differing by many 
more genes — the actions of which are of less fundamental significance for 
the control of the developmental pattern. 

In our approach to questions of hybrid vigor, we may be concerned with 
different degrees of hybridity. Consideration of this factor must involve not 
only the number of genes but also the nature of the action of the particular 
genes. Nor is this all, for the action of any specific allele is conditioned by the 
genetic background in which it occurs in a particular individual. Hence, the 
relations among genes may often be of critical importance. 

Of tremendous import, too, are the interactions between the activities of 
the genes and the environment. In speaking of hybrid vigor, we are general- 
ly concerned with such characteristics as size and yield, but these are merely 
end products of the metabolic processes. Patterns of these metabolic proc- 
esses are set by the genes, but the processes themselves may be either ac- 
celerated, inhibited, or otherwise modified by the effects of environmental 
factors. Hybrids which are particularly vigorous under certain conditions 
may show relatively little vigor under other environmental conditions. It is 
true that the enhanced vigor of hybrids frequently gives to them a wide 



TOO W. GORDON WHALEY 

range of environmental adaptability. It is equally true that certain hybrids 
exhibit vigor within only relatively narrow environmental limits. For lack of 
evidence it must be assumed that the distinction lies in the differences be- 
tween the patterns of hybridity and in the action of the genes responsible for 
the hybrid advantages. 

Any attempt to explain the genetic basis of heterosis must make initial 
recognition of one fact. The phenomenon can involve only the recombination 
of alleles already existing in the population or populations from which the 
hybrid organisms have been developed; unless, by rare chance, mutation 
should take place just prior to or just after the actual crossing. We are thus 
concerned with an interpretation limited to different types of recombina- 
tions, and to different kinds of gene action resulting from these recombina- 
tions. 

GENETIC MECHANISM OF HETEROSIS 

Consideration of the characteristics of dominance and heterozygosity has 
been of primary importance to investigators concerned with interpretation 
of the genetic mechanism of heterosis. Jones's dominance of linked factors 
hypothesis (1917) probably is still the most popular explanation of the 
genetic basis of heterosis. 

Dobzhansky (1941) and his co-workers, and many others, have recorded 
that in most species there has been, in the course of evolution, accumulation 
of deleterious recessive characters, which when homozygous reduce the 
efficiency of the organism — but which in the heterozygous condition are 
without efficiency-reducing effects. This revelation calls for a reshaping of no- 
tions regarding the nature of the favorable effects of the dominant alleles, but 
does not otherwise modify the structure of the explanation. The favorable- 
ness of the action of many of the dominant alleles probably is not the result 
either of directional mutation producing more favorable dominants or of 
selection tending to eliminate the unfavorable dominants. Instead, it may 
be due to the accumulation in populations of deleterious recessive mutations. 
These, if their effects are not too deleterious, often can be piled up in sig- 
nificant numbers. 

The piling-up of such deleterious recessives is probably one of the reasons 
why heterosis is a much more important phenomenon in such a plant as corn 
than it is, for example, in the tomato. Corn has been handled for hundreds 
or even thousands of years in a manner that has made possible the accumula- 
tion in populations of relatively large numbers of deleterious recessive modi- 
fiers. The tomato is more than 90 per cent self-pollinated, and any great 
accumulation of deleterious modifiers is unlikely. Corn populations char- 
acteristically contain thousands of individuals, and wind pollination makes 
for maintenance of heterozygosity. In tomato, the effective breeding popula- 
tion size approaches one, and deleterious mutations would tend to become 



PHYSIOLOGY OF GENE ACTION IN HYBRIDS 101 

homozygous with sufficient frequency to bring about the elimination of 
many of them. 

As a matter of observation, it would seem that a comparison of the occur- 
rence and degree of heterosis in different species, along with a consideration 
of the reproductive mechanisms in the various species, supports the proposal 
that heterosis in many cases is the result of the covering uj) in the hybrids of 
deleterious recessive alleles with a consequent return to vigor. The often 
stated argument that hybrids of corn, for instance, frequently are more 
vigorous than the original open-pollinated populations from which the in- 
breds used in their production were derived, has no validity with respect to 
this situation. In the production of the inbreds there is invariably a reassort- 
ing of the alleles of the open-pollinated populations. 

It is highly improbable, however, that dominant alleles operating either 
because of certain inherent favorable characteristics of their own, or simply 
to prevent the deleterious activity of recessives, present the only genetic 
basis of heterosis. Dominance is by no means the clear-cut feature described 
in Gregor Mendel's original paper. The dominance of a particular allele may 
be conditioned by the environment, or it may depend upon the genetic 
background in which the allele exists. A completely dominant effect of one 
allele over another, in the classic sense of our utilization of the word domi- 
nance, is by no means universal. 

Rather unfortunately the so-called heterozygosity concept of heterosis has 
usually been introduced as being in opposition to the dominance explanation. 
Because the concepts of the features of dominance and recessiveness early 
put them into rigid categories, it has been difficult to postulate how a hetero- 
zygous condition with respect to one or more genes could render an organism 
more vigorous than the homozygous condition, usually of the dominant 
alleles. 

Evidence of significance for the interpretation of the importance of hetero- 
zygosity in heterosis has been accumulated slowly. There is now, however, a 
fairly long list of instances in many different species in which the heterozy- 
gous condition for certain alleles is known to be superior to either the homo- 
zygous recessive or the homozygous dominant condition (Stubbe and 
Pirshcle, 1940; Singleton, 1943; Karper, 1930; Robertson, 1932; Robertson 
and Austin, 1935; Gustafsson, 1938, 1946; Nabours and Kingsley, 1934; 
Masing, 1938, 1939a, 1939b; Rasmusson, 1927; and Timofeef-Ressovsky, 
1940. 

The accumulation of data on these cases followed a long period during 
which all the investigations reported seemed to indicate no marked differ- 
ences between organisms heterozygous for certain alleles and those with the 
dominant homozygous condition for these same alleles. At least, in no in- 
stance, was there any marked superiority referable to the heterozygous 
condition. Most of the genes involved in the more recent findings have been 



102 W. GORDON WHALEY 

catalogued as having at least moderately deleterious effects in the mutated 
state. The characteristics controlled by them include: chlorophyll deficien- 
cies, modifications of leaf form and pigmentation, stalk abnormalities, flower- 
ing pattern, and time of flowering. 

The extent to which the actual nature of the genetic situation has been 
analyzed varies, but in several of the cases it seems clear that the mutation 
of a single gene is involved and that the Fi hybrids are heterozygous only 
with respect to the alleles at this particular locus. The amount of heterosis 
manifest also varies greatly. Because of experimental differences, no accurate 
comparisons can be made, but in some instances the amount of hybrid vigor 
appears to be nearly comparable to that which occurs in crosses involving 
large numbers of allelic differences. The situation appears to be one in which 
a mutation takes place, and the mutated allele is definitely deleterious when 
homozygous. In individuals heterozygous for the particular gene, there ap- 
pear none of the deleterious effects. Instead, a definite heterotic effect ap- 
pears. Dominance is of no apparent importance, and the distinction between 
the vigorous hybrids and the less vigorous non-hybrids rests upon hetero- 
zygosity. 

Jones (1944, 1945) has reported several cases of what he has called heter- 
osis resulting from degenerative changes. He first suggested that these cases 
represented instances of heterosis with a genetic basis in the heterozygosity 
of certain of the mutated genes. More recently (private communication) 
Jones has concluded that these cases involve more than single gene differ- 
ences, and that the results may be explained on the basis of an accumulation 
of favorable dominant effects. 

The case of a single locus heterosis reported by Quinby and Karper (1946) 
involves alleles which do not produce any detectable deleteriousness, but in 
certain heterozygous combinations produce hybrid vigor comparable in 
amount to that in commercial hybrid corn. Quinby and Karper have referred 
the hybrid advantage in this case to a stimulation of meristematic growth in 
the heterozygous plants. 

All of these instances involve specific allelic interactions and not superior- 
ity resulting from heterozygosity per se — as was postulated by some of the 
earlier workers concerned with the genetic interpretation of heterosis. These 
examples contribute to the increasing realization that the phenomenon of 
dominance is perhaps of less importance with respect to heterosis than has 
been supposed. There is no a priori reason why the interaction of a so-called 
recessive allele and a so-called dominant allele should not give results differ- 
ent from and metabolically superior to those which are conditioned by either 
two recessives or two dominants. 

This situation bears closely upon the interpretation of heterosis set forth 
by East in 1936. East postulated that at the loci concerned with the 
mechanism of heterosis there might be a series of multiple alleles — with the 
combinations of different alleles giving results metabolically superior to 



PHYSIOLOGY OF GENE ACTION IN HYBRIDS 103 

those determined by the combinations of like alleles, and with no considera- 
tions of dominance being involved. In the light of existing evidence it seems 
a safe assumption that a considerable portion of hybrid vigor is the result 
of allelic interaction between different alleles at the same locus. Although the 
evidence as yet is scanty, it is certainly pertinent to suggest that some 
heterosis may result from the interaction of alleles at different loci, when 
such alleles are brought into new combinations in the hybrids. 

Most of the recent studies of the relation of heterozygosity to heterosis 
have been concerned with the results of the action of single genes. Such 
studies have emphasized that heterosis need not have its basis in the action 
of large numbers of genes but can be, and apparently frequently is, a result 
of the combining of different alleles of a single gene. Any considerable amount 
of hybrid vigor resulting from the action of single genes would seem to indi- 
cate the involvement either of multiple effects of single genes or of genie 
action in the control of relatively fundamental metabolic processes. Both are 
likely probabilities. 

The metabolic system of any organism which grows and functions in a 
satisfactory manner is an exceedingly complicated mechanism with a great 
number of carefully balanced, interrelated processes. The mutation of any 
gene which has control over any of the key processes or functions will almost 
certainly be reflected in a number of processes and activities. For example, if 
a change in the character of some fundamental enzyme system is involved, 
either the addition or subtraction of a functional step, or of a substance 
produced at a particular developmental stage, would be likely to enhance or 
inhibit a number of important processes in the general metabolism of the 
organism. 

The equilibrium factor in genie action is obviously a consideration of 
great importance. If a mutation disturbs this equilibrium after it has become 
fairly well established through selection and elimination processes, the con- 
sequences may reduce the organism's vigor. If, in a hybrid, the mutation is 
then brought together with the original wild type or normal allele, the sum 
total of the actions of the mutated allele and the original allele may well be 
such as to exceed that of two copies of the original allele in the production 
of vigor in the organism. 

When we give attention to physiology of gene action in hybrids which are 
heterotic, we must concern ourselves with all of these considerations in- 
cluding the fact that a single gene, the mutation of which affects some 
processes in a sufficiently fundamental stage of the organism's formation, 
may well have a greater end effect than a number of genes whose functions 
are concerned with more superficial developmental processes. 

SEED AND EMBRYO DEVELOPMENT 

The literature on heterosis contains a number of discussions concerning 
the relation between seed and embryo size and heterosis (Kiesselbach, 1926: 



104 W. GORDON WHALEY 

Ashby, 1930, 1932, 1937; East, 1936; Sprague, 1936; Copeland, 1940; Mur- 
doch, 1940; Kempton and McLane, 1942; Whaley, 1944, 1950). 

Most of the investigations have dealt with mature seed and embryo size. 
The evidence shows that in many instances hybrid vigor is associated with a 
high embryo weight. In some cases the initially high-weight embryo is found 
in a relatively large seed. There is, however, by no means a consistent correla- 
tion between either high embryo weight or large seed size and heterosis. 

The results of studies on corn inbreds and hybrids in our own laboratory 
(Whaley, 1950) seem representative of the general findings. Among some ten 
inbred lines there occurred a great deal of variation from one line to another 
as to both embryo weight and seed weight. There was somewhat more varia- 
tion with respect to embryo weight. Among the Fi hybrids, all of which 
exhibited considerable vigor under central Texas conditions, there were a 
few with embryo weights which exceeded those of the larger-embryo parent. 
For the most part, the embryo weights were intermediate, and in one or two 
cases they were as low as that of the smaller-embryo parent. The weight of 
the seed tissues other than the embryo tended to follow that of the pistillate 
parent, but was generally somewhat higher. Double crosses which had vigor- 
ous Fi hybrids as pistillate parents characteristically had large seeds with 
what were classified as medium-weight embryos. 

The few reports, such as Copeland's (1940), concerning the development 
of embryos in inbred and hybrid corn, suggest that at the earlier stages of 
development some hybrid vigor is apparent in the hybrid embryos. The 
observations of hybrid vigor during early development of embryos and the 
absence of any size advantage at the time of seed maturity are not necessari- 
ly conflicting. In most plants, embryo and seed maturation represent fairly 
definite stages at which a certain degree of physiological maturity and of 
structural development has been attained. It is probably to be anticipated 
that even though certain heterotic hybrids show early embryo development 
advantages, these advantages may be ironed out by the time the embryo 
and the seed mature. The size of both the embryo and the other seed tissues 
is conditioned not only by the genotype of these tissues themselves, but also 
by the nutritional background furnished them by the plant on which they 
grow. 

It is quite possible that this genotype-to-background relationship is an 
important consideration in the determination of whether or not hybrid vigor 
is exhibited in the development of the embryo and seed. The background 
provided by the pistillate parent might be such as to preclude the develop- 
ment of embryo vigor, even though the embryo genotype were of a definitely 
heterotic constitution. The fact that hybrid vigor is apparent during certain 
stages of embryo and seed development may or may not be related to an 
embryo or seed size advantage at maturity. Because of this, it seems doubt- 
ful that embryo or seed size is a reliable measure of hybrid vigor; and that 



PHYSIOLOGY OF GENE ACTION IN HYBRIDS 105 

the rate of development during the embryo and seed maturation i)eriod is 
of any critical importance with respect to the development of hybrid vigor 
during post-embryonic growth. 

EARLY SEEDLING GROWTH AND HETEROSIS 

There have been few studies of early postgermination growth in plants in 
relation to heterosis. It would seem that the usual failure to find higher 
growth rates during the grand period of growth, or longer continued growth 
periods in heterotic hybrids, would suggest that the answer to the develop- 
ment of hybrid vigor lies for the most part in the early postgermination 
growth stages. The work of Ashby and his co-workers (Ashby, 1930, 1932, 
1936; Hatcher, 1939, 1940; Luckwill, 1937, 1939) emphasized that the hybrid 
advantage in their materials was either present in the resting embryo or be- 
came manifest in early postgermination growth. Its development was defi- 
nitely not a characteristic of the later growth phases. This observation has 
now been made for many cases of hybrid vigor (Whaley, 1950). There are 
some instances in which hybrid vigor seems to be the result of longer-con- 
tinued growth on the part of the hybrid. These probably have a dilTerent 
explanation from the majority of cases. 

We have been concerned lately in our own laboratory with an analysis 
of the early postgermination growth of corn inbreds and single and double 
cross hybrids (Whaley, 1950) . Studies of growth during the first ten to twelve 
days after germination have revealed that the hybrid advantage is largely 
the result of the heterotic hybrid plants reaching a high growth rate earlier 
than do the inbreds. Almost without exception, the development of the hy- 
brid advantage takes place very rapidly in the early stages of germination 
and growth. Rarely have we seen evidence of the hybrids having higher 
growth rates during any later part of the developmental cycle. Neither are 
the hybrid growth periods extended appreciably beyond those of the in- 
breds. In most instances the hybrids mature somewhat more rapidly than 
the inbreds — a fact of common observation among plant breeders. 

Since the attainment of the maximum growth rate takes place more 
quickly during the early stages of development, the hybrids do have a longer 
maximum growth rate period. During this period the early advantage is 
compounded, to give a considerably greater maturity advantage. When 
both the inbred lines and the hybrids used in our studies are considered, it is 
apparent that the rapid attainment of high early growth rates is correlated 
with relatively low embryo weights. This apparent higher efficiency of small 
embryos and its importance in relation to hybrid vigor requires further study. 

On the basis of the data at hand one can suggest that the hybrid advantage 
lies in the more rapid unfolding of certain metabolic processes, a suggestion 
which receives support from the recorded studies of later growth. 



106 W. GORDON WHALEY 

LATER GROWTH AND HETEROSIS 

It is unfortunate that most studies of the physiology of heterosis have been 
confined to the later growth period, and consequently do not include that 
part of the growth cycle during which the important differences seem to be 
developed. Nonetheless, we can learn much from these studies of later 
growth as to the nature of the physiological differences which may furnish 
bases for the development of hybrid vigor. 

The early experiments on physiological differences between inbreds and 
hybrids were concerned mostly with the responses of the inbreds and the 
hybrids to different soil conditions. A few examples will serve to indicate the 
type of investigation and the character of the results. Hoffer (1926) deter- 
mined the amounts of the constituents of the ash of heterotic hybrid corn to 
be generally intermediate between those of the parental types. He noted that 
iron and aluminum were present in the ash of the hybrids in smaller amounts 
than in the inbreds. His studies showed that although there were marked 
differences in the absorption of iron and aluminum in different soil types the 
vigorous hybrids tended to absorb less of both these elements than the less 
vigorous inbred lines. 

In the same year Kiesselbach (1926) reported distinct differences in water 
requirements between selfed lines of corn and their heterotic F; hybrids. The 
low productivity inbreds had much higher water requirements than the 
vigorous Fi hybrids, when water requirements were calculated on the basis of 
either water absorbed per gram of ear corn or water absorbed per gram of 
total dry matter. Barley inbreds and heterotic barley hybrids were shown 
by Gregory and Crowther (1928, 1931) to make distinctly different responses 
to various levels of available minerals. These investigators postulated that 
heterosis in barley might be directly related to differences in the ability of 
the hybrids and the inbreds to use certain nutrients. This suggestion has had 
a fairly adequate test, particularly with reference to nitrogen and phos- 
phorus nutrition. 

The work of DeTurk et al. (1933), Smith (1934), Lyness (1936), Harvey 
(1939), Burkholder and McVeigh (1940), and Rabideau et al. (1950), has 
provided a fairly adequate picture of the relation of phosphorus and nitro- 
gen nutrition to the development of hybrid vigor. Smith demonstrated dis- 
tinct differences among inbred corn lines with respect to phosphorus nutri- 
tion, noting that these differences were most apparent when the phosphorus 
supply was limited. He postulated that the higher phosphate utilization effi- 
ciency of the hybrids might be referred, at least in part, to the dominant in- 
heritance in them of a much branched root system. Later studies have shown 
that the root growth pattern is certainly important in relation to heterosis. 

Smith noted particularly that when inbred lines were inefficient in the 
utilization of phosphorus or nitrogen, crossing them failed to produce hybrids 
showing any evidence of physiological stimulation resulting in the more 



PHYSIOLOGY OF GENE ACTION IN HYBRIDS 107 

efTective use of these elements. Lyness (1936) studied heterotic Fi hybrids 
resulting from crosses between a low phosphorus-absorbing capacity inbred 
and a high phosphorus-absorbing capacity inbred. He found the heterotic 
Fi plants to have high phosphorus-absorbing capacity. These results sug- 
gested that phosphorus-absorbing capacity in corn, in some instances at 
least, acts genetically as a dominant factor. Lyness also noted the relation- 
ship between high phosphorus absorption and the extent of root develop- 
ment. He supposed that the extent of root development might be responsible 
for varietal differences in phosphorus absorption, a supposition which is sup- 
ported by later studies. The work of DeTurk et at. (1933) suggested that more 
than simply phosphorus-absorbing capacity is involved. This work revealed 
that the actual phosphorus content patterns of two Fi hybrids of corn were 
quite different. By estimating the amount of phosphorus in various chemical 
fractions, De Turk and his coworkers were able to demonstrate marked phos- 
])horus pattern difTerences and to associate these pattern differences with 
various phosphate fertilizer treatments. 

In our laboratory we have made a study of the phosphorus-absorbing ef- 
ficiency of corn inbreds and hybrids, and have attempted to correlate the 
findings of this study with developmental changes in the vascular system 
and with general growth (Whaley et al., 1950; Heimsch et al., 1950; Rabideau 
et al., 1950). The data indicate that heterotic hybrids definitely absorb more 
radioactive phosphorus than their inbred parents. This advantage in ab- 
sorption on the part of the hybrid is associated with more rapid early de- 
velopment, with earlier attainment of maturity, and with certain features of 
vascular organization. The greater absorption can be referred at least in 
part to better early development of the root system in the hybrids, and to a 
generally higher level of metabolic activity which presumably creates a 
greater phosphorus demand. The greater absorption of phosphorus by the 
hybrids is certainly one of the factors which compounds the heterotic effects, 
but it seems doubtful that it is a primary factor in the development of hybrid 
vigor. 

Harvey's (1939) studies of nitrogen metabolism among inbreds and hy- 
brids of both corn and tomato showed differences from one line to another 
with respect to the ability to use nitrate and ammonium nitrogen. The ex- 
periments were of such a nature as to make it clear that such differences in 
nutritional responses were results of differences in genetic constitution. The 
behavior of hybrids produced from the inbreds reflected a combination of the 
characteristics of the inbreds. Significantly, Harvey's study revealed that not 
only did differences exist among his inbreds and hybrids with respect to the 
ability to use different types of nitrogen, but that there were distinct genetic 
differences in the responses of the plants to various levels of nitrogen avail- 
ability. 

Somewhat similar differential responses to potassium availability were 



108 W. GORDON WHALEY 

revealed by Harvey's studies on tomato inbreds and hybrids. Burkholder and 
McVeigh (1940) have also noted differences in responses of corn inbreds and 
hybrids to various levels of available nitrogen. These investigators corre- 
lated apical meristematic development, and the differentiation of the vascu- 
lar system with the level of nitrogen nutrition, and the efficiency of different 
lines and hybrids in utilizing the available nitrogen. Their results indicate 
that hybrid vigor, involving superiority in the production of dry matter 
and the differentiation of organs, was not correlated with greater growth and 
development of the vascular system. 

There definitely are vascular organization differences between the heterot- 
ic hybrids and the inbreds in the material we have studied. These vascular 
organization differences seem not to be the result of differences in mineral 
absorption and distribution, but rather to be one of the factors responsible 
for the differences in absorption and distribution. All the evidence seems to 
indicate that the greater absorption of minerals by heterotic hybrids can be 
referred to better developed root systems in the hybrids, probably also to the 
presence of more efficient transport systems, and to a generally higher level 
of metabolic activity. 

Recently we have undertaken a rather extensive analysis of both the 
morphological and physiological characteristics of a tomato cross in which 
there is marked heterosis. We have found no significant differences between 
the inbreds and the hybrids as to total phosphorus content of the leaves, 
stems, or roots. There is some suggestion that the phosphorus content of the 
organs of the hybrids reaches a higher level earlier in growth than it does in 
the inbreds. Neither do the hybrid plants have any consistent advantage 
with respect to nitrogen content. 

Analyses of the starch content of the leaves and stems suggest that the 
hybrid plants may have a slightly higher starch content than the inbreds 
during the early growth stages. In terms of average figures over the whole 
growth period, however, there are no marked differences between the in- 
breds and the hybrids. The same appears to be true of the sugar content. 
The hybrids have a somewhat higher sugar content, at least in the leaves, 
early in development. During the greater part of the growth cycle the hy- 
brids do not have significantly more sugar than the inbreds. The only clear 
difference found between the inbreds and the hybrids is in the catalase ac- 
tivity of the shoot tips, the hybrids having an appreciably greater index of 
catalase activity than either of the inbred parents. The catalase activity 
differences are associated with much more active meristematic growth in the 
hybrid plants. 

THE ROLE OF SPECIFIC SUBSTANCES IN HETEROSIS 

Evidence for another sort of physiological differences possibly involved in 
heterosis is furnished by the work of Robbins (1940, 1941a) in assaying the 



PHYSIOLOGY OF GENE ACTION IN HYBRIDS 109 

growth-promoting activities of extracts from inbred and hybrid corn grains. 
Robbins' evidence indicates that a substance or substances, which he has 
designated as factor Z, may be synthesized in greater amounts by the hy- 
brids than by the inbreds. He has stated that factor Z can be fractionated into 
Zi, which is hypoxanthine; and Z2, a still unidentified fraction. Robbins' 
work suggests that among the advantages possessed by heterotic hybrids 
may be the ability to synthesize certain growth substances which the in- 
breds either cannot synthesize or cannot synthesize as well. 

Further evidence of a slightly different nature is provided by the root 
culture work of Robbins (1941b) and of Whaley and Long (1944). Robbins 
used cultures of a strain of Lycopersicou pimpiiieUijolium Mill., a strain of 
L. esculeutum Mill., and their Fi hybrid, in solutions supplemented by thia- 
min, thiamin and pyridoxine, or thiamin, pyridoxine, and nicotinamide. 
Robbins found that the Fi roots grew much more rapidly and produced 
more dry matter than those of either parental line. He was able to show 
further that one parental line made a greater response to the presence 
of pyridoxine than did the other, while the roots of the second parental line 
made a greater response to nicotinamide than those of the first. This suggests 
the combination of complementary factors from the parents in the hybrid. 
Whaley and Long (1944) obtained essentially the same results with a cross 
involving two inbred lines of L. esculeutum. 

In the University of Texas tissue and organ culture laboratory, we have 
been exploring certain aspects of this problem. While the results are not suf- 
ficiently complete for publication, some facts are already clear. Among the 
roots of many inbred lines of tomatoes which we have been culturing, there 
are marked differences in growth responses associated with the availability 
or non-availability of thiamin, pyridoxine, niacin, and certain other sub- 
stances. These differences appear definitely to be inherited and they can be 
studied in either the inbred lines or hybrids. 

It is still too early to say what the inheritance pattern is, but consideration 
can be given to some aspects of the growth response patterns. One of the 
most significant revelations is that the responses of most of the roots to a 
specific substance are conditioned not only by the availability of that sub- 
stance, but by the availability of the other substances and by the gen- 
eral composition of the culture medium. Heterosis in tomato root cultures 
is, like heterosis in whole plants, definitely relative, and conditioned, not 
only by the environment, but, with respect to any specific gene action, by 
the background of other gene actions taking place in the developing or- 
ganism. 

Heterosis in tomato root cultures is definitely related to the inheritance of 
the capacity to synthesize or utilize such substances as thiamin, ])yridoxine, 
and niacin. This is not to suggest that heterosis in whole plants of tomato 
may have its basis in the genetic recombination of factors concerned in the 



no W. GORDON WHALEY 

control of thiamin, pyridoxine, or niacin metabolism. In intact plants, it is 
likely that the green parts supply these substances to their own tissues and to 
the roots, in amounts satisfactory for growth and development. The root 
tissue responses, however, are definitely heterotic in certain instances, and 
these mechanisms merit examination. 

It seems pertinent to explore the role of these B vitamins in growth and 
development. Thiamin appears to be a metabolic requirement for all types of 
cells. Its metabolic activity apparently revolves around a role in enzyme 
systems. Thiamin pyrophosphate is the co-enzyme of the enzyme pyruvate 
carboxylase (Lohmann and Schuster, 1937). The enzyme carboxylase occurs 
in many plant tissues. The possible biochemical basis of thiamin action in 
plants has been set forth in some detail by Bonner and Wildman (1946), 
Vennesland and Felsher (1946), and Bonner and Bonner (1948). It is assumed 
that thiamin represents a step in the development of co-carboxylase which is 
active in one or more of the decarboxylating enzyme systems of the respira- 
tory mechanism. 

Pyridoxine also has an enzymatic role, apparently being important for its 
conversion to pyridoxal phosphate, which is a co-enzyme of one or more of 
the reactions in the nitrogen metabolism of the plant (Bonner and Bonner, 
1948). As a co-enzyme active in nitrogen metabolism reactions, pyridoxine 
may be of extreme importance in amino acid-protein building, and hence 
active in conditioning fundamental growth activities. 

Similarly, niacin activity is enzymatic in character. Niacin appears to be 
involved as a constituent of the nucleotide cozymase, and possibly of tri- 
phosphopyridine nucleotide. Cozymase is a co-enzyme for a whole series of 
dehydrogenase enzymes, including alcohol dehydrogenase, malic dehydrog- 
enase, and glutamic dehydrogenase (Bonner and Bonner, 1948). 

The genetic background of thiamin, pyridoxine, and niacin metabolism is 
thus a genetic background concerned with basic components of the plant's 
enzyme systems. Heterosis, which rests upon recombinations concerned with 
thiamin, pyridoxine, or niacin metabolism, quite obviously rests upon recom- 
binations which are concerned with the acceleration, inhibition, or blocking 
of specific stages or developed substances in the basic enzyme system. 

A considerable amount of supporting evidence for the involvement of such 
fundamental enzyme and other growth substance activities in the develop- 
ment of heterosis has been coming for some time from the work on Neuro- 
spora. In many heterocaryons of Neurospora, increased growth responses 
directly suggestive of heterosis have been observed. In a number of instances 
(Beadle and Coonradt, 1944), the growth responses depend upon the two 
types of nuclei in the heterocaryon — each carrying wild type alleles of de- 
leterious mutant genes carried by the other nucleus. Such instances represent 
essentially the same situation as the recombination of favorable dominant 
alleles in normally diploid organisms. 



PHYSIOLOGY OF GENE ACTION IN HYBRIDS 111 

In one case reported by Emerson (1948) a different situation obtains. A 
mutant strain of Neurospora which requires sulfonamides for growth at cer- 
tain temperatures will grow satisfactorily in the absence of sulfonamides, 
provided that the concentration of available p-aminobenzoic acid is held at a 
particular level. Either higher or lower concentrations of p-aminobenzoic acid 
result in growth inhibitions. Emerson has made heterocaryons between a 
mutant strain carrying the sulfonamide-requiring gene {sfo) and a gene which 
prevents the synthesis of p-aminobenzoic acid (pab), and a strain carrying 
sfo and the wild type allele (+) of pab. The resultant heterocaryons grow 
vigorously on the minimal medium (without sulfonamides), whereas strains 
carrying sfo and pab, or sfo and +, make no appreciable growth on the 
minimal medium. Emerson's explanation of the growth of the hetero- 
caryons is that it results from a balance between the production of p-amino- 
benzoic acid by one of the types of nuclei and the absence of production of 
p-aminobenzoic acid by the other type of nucleus; so that the total produc- 
tion of p-aminobenzoic acid is sufBcient for growth but still within the range 
tolerated by strains carrying sfo. Heterosis-like effects of this sort are sugges- 
tive of the instances of heterosis related to the heterozygosity of particular 
genes in diploid organisms. 

We thus have in Neurospora, heterosis-like effects assignable both to a 
recombination of dominant alleles basis and to a heterozygosity basis. More 
important for this discussion is the fact that these instances are all concerned 
with facilitation in the hybrid of the production or utilization of substances 
which are components of the basic enzyme or other growth substance pat- 
tern of the organisms. 

Various investigations of heterosis in Drosophila, while for the most part 
not concerned with specific growth substances, have nonetheless assigned 
manifestation of heterosis to a background in the fundamental biochemical 
activities of the organisms. Inasmuch as these investigations are discussed in 
detail in another chapter, they will not be treated here. 

THE PHYSIOLOGICAL BASIS OF HETEROSIS 

From consideration of the pertinent data, a definite pattern emerges. 
This associates the development of heterosis with the ability of the hybrid 
to synthesize or to utilize one or several specific substances involved in the 
fundamental growth processes of the organisms. Nutritional factors, water 
absorption factors, and the other more gross considerations with which in- 
vestigators have been particularly concerned seem to be secondary factors^ — 
perhaps responsible for compounding the heterotic effects but probably not 
responsible for their initial development. Much of the evidence agrees with 
the assumption that the primary heterotic effect is concerned with growth 
substances whose predominant activity is registered in the early part of 
the developmental cycle; in plants, especially in early postgermination 



112 W. GORDON WHALEY 

growth. Into this category fall the enzymes, the auxins, and the other "phys- 
iological key" substances. 

Many heterotic hybrid plants seem to gain their advantage within the first 
few hours after germination. This advantage may not be shown as statistical- 
ly significant until it has been further heightened by subsequent growth. 
The primary growth activities during this period are those involved in the 
unfolding of the enzymatic pattern; the mobilization, transformation, and 
utilization of stored materials, and the building up of active protoplasmic 
synthesis. It seems definitely to be here that the hybrid advantage lies. By 
the time growth is well under way, the hybrid advantage is already well 
developed. 

Structural diflferences between inbreds and heterotic hybrids shown by the 
studies of Burkholder and McVeigh (1940), Weaver (1946), and the members 
of our laboratory (Whaley et al., 1950; Heimsch et al., 1950; Rabideau ef al., 
1950) are apparently to be regarded as results of heterosis rather than as 
causal factors. The evidence suggests that heterosis is concerned primarily 
with growth processes and that differentiation activities are most likely in- 
volved secondarily rather than primarily. What seems to be indicated is the 
assignment of the physiological basis of heterosis to the activity of one or 
more of the so-called physiologically active substances involved in early 
growth. 

Much of the apparent hybrid vigor is assignable to these activities only in 
a secondary fashion. Once the advantage of a larger number of growing 
centers or of heightened meristematic activity is established, the greater 
availability of nutrients, the greater amount of protoplasm involved in 
further protoplasm building, and other general advantages tend to increase 
the initial differences. To the general evidence in favor of this supposition 
can be added the specific evidence of the few cases in which the physiological 
action of particular alleles is known. Where these alleles in combination are 
responsible for heterosis, they have — when studied in sufficient detail — 
invariably been shown to be alleles whose action involves basic enzyme or 
other growth substance activity. 

If we are to make significant headway in understanding the physiological 
mechanism of heterosis, we shall have to concentrate on a detailed study of 
the developmental physiology of early growth. Much of the general knowl- 
edge we already have can contribute toward this understanding if we trans- 
late it into terms signifying that when we speak of quantitative differ- 
ences — size, yield, or of rate difTerences — we are really concerned with differ- 
ences in the level of metabolism. We must recognize that these differences in 
the level of metabolism are bound to vary against different environmental 
backgrounds, and where the particular genes involved are associated with 
different genetic backgrounds. 

Our approach to the heterosis problem has been complicated by common 



PHYSIOLOGY OF GENE ACTION IN HYBRIDS 113 

insistence upon attempts to find a single genetic mechanism. It has suffered, 
too, from faikire to recognize that between the gene and the final mature 
organism there lies a system of developmental processes of great complexity. 
The complexity of this system is formidable but it surely can be analyzed, 
at least with respect to its most significant features, if it is taken part 
by part. 

SUMMARY 

The evidence relating to heterosis suggests that the phenomenon is to be 
explained genetically in terms of various recombination effects. In some cases, 
dominance is the important consideration, while in other cases, hetero- 
zygosity must be considered. In any event, it is the resulting specific gene 
action which lies at the basis of the physiological advantage or advantages 
which give rise to hybrid vigor. One or many genes may be involved. Con- 
siderations of genetic balance and genotype-environment balance are im- 
portant. Probably most cases of heterosis are to be explained physiologically 
in terms of differences in the more fundamental aspects of the metabolic pat- 
tern, particularly those concerned with enzyme, auxin, and other growth 
substance activity in plants and with enzyme and hormonal activities in 
animals. 

To clarify the mechanism further, studies must be concerned primarily 
with the genetics and physiology of early development. We have been con- 
cerned with mature characteristics of size and yield, with the inheritance of 
so-called quantitative genes, and with analyses by the classic methods of 
genetics. These studies have brought us close enough to an understanding 
of the phenomenon of heterosis to indicate that its further analysis by 
techniques now at hand will uncover facts of tremendous importance for 
genetics, physiology, and other studies of development, some of them con- 
siderably afield from heterosis itself. 



WILLIAM J. ROBBINS 

Columbia University and New York Bofanical Garden 



Chapter 7 

Hybrid Nutrifionol 
Requirements 



Hybrid vigor has been recognized for more than a century. It has been con- 
sidered from a genetic, morphological, developmental, physiological, and 
commercial standpoint. Although a great deal of information has been ac- 
cumulated about the phenomenon, we are still unable to define exactly why 
a hybrid grows better than the parents from which it comes. 

It is obvious that the cause is physiological — the hybrid functions more 
effectively or for a longer period of time, and accumulates a greater mass of 
cell substance. Its metabolic efficiency is greater (East, 1936). It would be 
illuminating if we could locate specifically the physiological processes which 
are responsible for the greater vigor of the hybrid — recognizing that they may 
be numerous and complex rather than single and simple, and that they may 
not be the same for all examples of hybrid vigor. 

For many years I have been interested in the factors which determine why 
one plant species, variety, or strain grows slowly in a given environment 
where another flourishes. I have dealt mainly with microorganisms, especial- 
ly the filamentous fungi, because the external env ironment can be more easily 
controlled and photosynthesis is not a complicating factor. From my ex- 
perience, as well as from the work of others, it is clear that in many instances 
growth — the accumulation of cell substance — is limited by the efficiency of 
the organism's metabolic machinery, especially the activity of one or more 
enzyme systems. Whether this concept can be applied also to the phenome- 
non of hybrid vigor is still to be determined. However, it is a hypothesis 
which deserves exploration. 

Let us begin with a simple example of growth-limitation. Aspergillus niger 
grows well in a liquid medium of sugar, mineral salts, and asparagine. In the 
same medium Phycomyces Blakesleeanus will not grow at all. 

114 



HYBRID NUTRITIONAL REQUIREMENTS 115 

Does Phycomyces fail to grow in the basal solution because of the absence 
of something essential which it needs for growth, or because of the presence 
of something detrimental? Does Aspergillus niger grow in the basal solution 
because it does not need to be furnished with the "essential" substance, or 
because it is more resistant to the supposed injurious ingredient? 

For the example cited, we have a definite and well demonstrated explana- 
tion. Phycomyces fails to grow in the basal medium because it requires the 
vitamin, thiamine — which it is unable to make from sugar, mineral salts, and 
asparagine. Aspergillus niger also needs thiamine, but it constructs the vita- 
min from the elementary materials present in the basal solution. In this in- 
stance, therefore, the failure to grow is due to the lack of something es- 
sential for growth; namely, thiamine, the precursor of co-carboxylase. 

This is not an isolated example. Many species of fungi grow slowly, or not 
at all, in a basal medium because of their inability to make one or more of the 
essential metabolites. These metabolites may include various vitamins, 
purine and pyrimidine bases, amino acids, fatty acids, or substances as yet 
unidentified. 

ESSENTIAL METABOLITES-RELATION TO GROWTH 

It may be assumed that the complex chemical compounds which make 
up the cell substance of a living organism are constructed by the organism 
from simpler compounds. A series of intermediate chemical compounds are 
formed between the original simple foods and nutrients and the final product, 
cell substance. This step-wise progression from simple to complex is made 
possible by a series of enzymes, also made by the organism, which operate on 
each stage as that stage is completed. Although synthesis is likely to be 
emphasized in considering growth, there are other subsidiary processes — 
necessary concomitants for the building up of new cell substance. The cata- 
bolic processes of digestion and respiration also occur in steps, and are made 
possible by the action of a series of enzyme systems. 

Any substance playing a necessary part directly or indirectly in the chain 
of reactions which end in the synthesis of new cell substance is an essential 
metabolite. Unless each essential metabolite, each chemical substance in the 
step-wise process of growth, each enzyme which facilitates the chemical re- 
actions concerned, is made within the organism or supplied from without, the 
series is interrupted. New cell substance is not made, and growth does not 
occur. If not enough of an essential metabolite is made, growth will be 
slowed. 

Of course, this is an oversimplified statement of a very complicated 
process. The reactions concerned in growth probably do not occur in a 
straight line. Some steps may be bypassed and side reactions may occur, all 
of which may affect the speed and character of the growth which results. 

It would be difficult to estimate the number of essential metabolites in- 



116 



WILLIAM J. ROBBINS 



volved in the growth of even the simplest organism, or to put a limit on the 
number for which some organism may not eventually be found to exhibit a 
deficiency. 

Some species or strains exhibit a complete deficiency for one or more 
essential metabolites. They are unable to synthesize any of the substances in 
question and do not grow unless the substances are supplied in the medium 
in which they are cultivated (Robbins and Ma, 1942). Others suffer from 
partial deficiencies, that is, they grow slowly in the absence of a particular 





Fig. 7.1 — Growth affected by complete and partial deficiencies for essential metabolites. 
Fungi grown on mineral-dextrose medium containing asparagine and purified agar and 
supplemented as follows: (1) no addition; (2) thiamine; (3) pyridoxine; (4) biotin; (5) thia- 
mine and pyridoxine; (6) thiamine and biotin; (7) pyridoxine and biotin; (8) all three vita- 
mins. .\bove, Ceratosiomella midtiannulata, complete deficiency for pyridoxine, partial for 
thiamine; below, C. microspora, complete deficiency for thiamine, biotin, and pyridoxine. 



essential metabolite but more rapidly if it is added to the medium (Fig. 7.1). 

For example, the clone of excised tomato roots, with which we have 
worked for many years, suffers from a complete deficiency of thiamine and a 
partial deficiency of pyridoxine. It will not grow unless the medium contains 
thiamine or its equivalent. When pyridoxine is added to a medium contain- 
ing thiamine, the growth of the excised roots is markedly increased. 

In a sugar, mineral-salt solution, the growth of our clone of excised tomato 
ri)ots is limited by its ability to synthesize thiamine. In a thiamine solution, 
growth is limited by the ability of the roots to synthesize pyridoxine (Robbins, 
1946). We have not been able to define what limits the growth of the root 
in a solution which contains both thiamine and pyridoxine. Other examples 



HYBRID NUTRITIONAL REQUIREMENTS 117 

of partial deficiencies could be cited. Their effect is to decrease the rate of 
growth but not to inhibit it entirely. 

As a result of investigations which have extended over the past decade or 
two, we know of many examples in which poor growth or failure to grow in a 
specific environment is due to the inability of the organism to synthesize 
adequate quantities of one or more essential metabolites. The metabolic 
machinery lacks a part, or some part works slowly, with the result that the 
organism does not make sufficient quantities of one or more growth essen- 
tials, and unless supplied with the missing materials from without, grows 
slowly, or not at all. 

Not all instances of failure to grow or of poor growth in a given environ- 
ment are explainable on the basis of deficiencies of essential metabolites. In 
some instances growth may be limited by autogenic growth inhibitors. 

AUTOGENIC INHIBITORS 

Zalokar (1948), Emerson (1947, 1948), and others have described a mutant 
strain of Neurospora which grows poorly at high temperatures. Growth oc- 
curs if sulfonamide is added to the medium. One might conclude that 
sulfonamide acts for this organism as an essential metabolite. It appears, 
however, that this mutant produces growth inhibitors which are antagonized 
in some way by the sulfonamide. This seems to be an example of poor growth 
caused by the accumulation of autogenic growth inhibitors, and not because 
of the lack of an essential metabolite. 

Information on the role of autogenic inhibitors in limiting growth is less 
specific and more difficult to obtain than evidence for the limitation of growth 
due to a deficiency of an essential metabolite. How commonly do internally 
produced inhibitors reduce growth? What is the nature of these substances? 

From the investigation of antibiotic substances we know that many organ- 
isms form metabolic products, highly inhibitory for organisms other than 
themselves. Do they also produce substances which limit their own growth? 
The role of autogenic inhibitors in limiting growth deserves much more 
attention than it has received. 

It is well known that minute amounts of specific chemical compounds 
materially modify the amount and nature of growth in plants. Zimmerman 
and Hitchcock (1949) treated Kalanchoe plants with small amounts of the 
ortho, para, and meta forms of chlorophenoxyacetic acid. The para form 
caused the apical meristem to develop into a spathe-like organ which could 
be cut off and rooted. It had little resemblance to Kalanchoe. The ortho and 
meta forms of this compound did not have this effect. This modification was 
not a mutation. The effect wore off as the chemical in the plant disappeared, 
and the Kalanchoe eventually returned to its normal growth pattern. If the 
change had been permanent, we would have been inclined to call it a muta- 
tion and look for a genie explanation ; i.e., look for a gene which controlled the 



118 WILLIAM J. ROBBINS 

production of para-chlorophenoxyacetic acid. We might say that this com- 
pound and the Kalanchoe plant acted temporarily as linked genes. 

Many other kinds of abnormal growth in plants are probably the result of 
the effect of minute amounts of specific chemical compounds. Insect galls 
are characterized by an abnormal but specific growth pattern superimposed 
on normal tissue by the presence of a foreign living organism. It seems very 
likely from the observations of Boysen Jensen that the abnormal growth of 
insect galls is caused by specific chemical compounds produced by the larvae 
which inhabit the galls. 

It must be emphasized that growth is an extremely complex process, not 
just a series of chemical reactions. To consider it as such is admittedly an 
oversimplification giving no thought to the organization in which these re- 
actions occur, or to the structural elements, physical processes, and chemical 
reactions which must play a role. 

The concept of growth as a series of catalyzed reactions is useful and 
stimulating, however, in considering the role of essential metabolites — 
especially enzymes — and the action of inhibitors and minute amounts of 
specific chemical compounds. 

HYBRID VIGOR 

Some years ago I attempted to determine whether hybrid corn contains a 
greater quantity of substances which stimulate the early growth of Phyco- 
myces Blakesleeanus than the inbred parents. The effect of extracts of air 
dry grains and of partially germinated grains of the hybrid corn and its in- 
bred parents was determined on the growth of Phycomyces in the presence of 
thiamine (Robbins, 1940, 1941a). 

When compared on the basis of extract per grain, I found that the extracts 
of the grains of the hybrid corn gave a greater dry weight of mycelium of 
Phycomyces than those of either of the inbred parents (Fig. 7.2). The stimu- 
lating material seemed to be present in both the embryo and the endosperm. 
Since the solution in which the beneficial effects of the extracts were exhibited 
contained sugar, asparagine, mineral salts, and thiamine, it appeared that 
the effect was produced by unidentified growth substances. These were 
termed for convenience, factor Z. 

After estimating the amount of factor Z present — from the effects of the 
extracts of the corn grains on the early growth of Phycomyces in the presence 
of thiamine — the following generalities seemed permissible. The amount 
of factor Z increased with the time of the germination of the corn grains, at 
least up to seventy-two hours' germination. The quantity of Z was greater 
per endosperm than per embryo, and was greater in the grains of the hybrid 
than in those of either parent. The amount of thiamine and its intermediates 
in the embryo and endosperm of the grains of the hybrid and its parents 
was not correlated with the amount of factor Z, nor did the amount of biotin 
in the extracts appear to be correlated with the amount of factor Z. 



HYBRID NUTRITIONAL REQUIREMENTS 



119 



These results suggest that there is present in the grains of corn, material 
which stimulates the early growth of Phycomyces in the presence of thiamine, 
and that there is more of this material per grain in heterotic hybrids than in 
those of the inbred parents. 

Interpretation of these results depends in part on the identity of factor Z. 

100 



90 



80 



O 

D 
_1 

u 

O 

>- 

H 

>- 
tr 
a 



70 



60 



50 



40 



30 



20 



10 




025 



0.5 
ML. CORN GRAIN EXTRACT 



Fig. 7.2 — Increase in dry weight of Phycomyces produced by extracts of air dry grains of 

maize. Extracts added to medium of sugar, minerals, asparagine, and thiamine. A = line 

4-8; B = line 187; C = 985, 4-8 X 187; D = 995, 187 X 4-8. 1 ml. extract = 1 grain. 



Unfortunately, we do not know what factor Z is. We succeeded in dividing it. 
We demonstrated that factor Z is multiple, and separated it into a fraction 
adsorbed on charcoal, factor Zi, and a filtrate fraction, factor Z2. Factor Zi 
was identified as hypoxanthine. Factor Z2 may be a mixture of amino acids. 
Although this problem is left in an uncertain and unsatisfactory condi- 
tion, it suggests a line of attack. This would be an investigation of heterosis 
by studying the efifect of extracts of parents and of heterotic hybrids on the 
growth of other organisms. This may serve as a means of bioassay for favor- 
able or unfavorable growth factors. 



120 



WILLIAM J. ROBBINS 



Vigor in Heterocaryons 

Observations of Dodge (1942) on heterocaryosis in Neurospora are of 
interest to the general problem of heterosis. Dodge inoculated three petri 
dishes, one with his Dwarf 16 strain of Xeurospora telrasperma, one with race 
C-8, and the third with mixed mycelium or conidia of both the dwarf and the 
C-8 races. He observed that the mycelium of the mixed culture grew much 
more rapidly and produced more abundant conidia than the mycelium of 
either the dwarf or the C-8 races (Fig. 7.3). 




Fig. 7.3 — Heterocaryotic vigor in Neurospora telrasperma. Growth in 34 hours at room 

temperature in petri dishes. The myceUum of the two heterocaryotic races {16 -\- C 4 and 

16 + C8) has nearly covered the medium in the dishes; C4 and C8 have not grown halfway 

across the medium and Dwarf 16 has made no visible growth. 



When two races of Xeurospora telrasperma are grown together, there is a 
migration of nuclei through the openings at the points of hyphal anas- 
tomoses. The races need not be of opposite sex. After nuclear migration, the 
cells of the resulting mycelium are heterocaryotic. They contain two kinds of 
haploid nuclei. The greater vigor of the mixed culture referred to above ap- 
pears to be the result of the presence in a common cytoplasm of two kinds of 
nuclei. 

Heterocaryotic vigor does not always accompany heterocaryosis. Dodge 
(1942) observed heterocaryotic vigor when the two races, Dwarf 16 and C-4, 
were grown together. But heterocaryosis for races C-4 and C-8 did not result 
in increased vigor in the mixed culture. Not all dwarf races act as race 16 
does. Some of them evidence heterocaryotic vigor with both C-4 and C-8, 



HYBRID NUTRITIONAL REQUIREMENTS 121 

others with C-4 but not with C-8, and still others develop none with either 
C-4 or C-8. 

Dodge has suggested that the heterocaryotic hybrid may synthesize a full 
quantity of growth substances or essential metabolites. Whereas the growth 
of each of the parents is limited by their inability to synthesize adequate 
quantities of one or more essential metabolites. 

Dwarf 16, for example, may be able to make adequate quantities of essen- 
tial metabolites 1, 2, 3, and 4, but unable to construct enough of 5, 6, 7, 
and 8. On the other hand, race C-4 may be unable to synthesize enough of 
1, 2, 3, and 4, but be capable of producing an adequate supply of 5, 6, 7, 
and 8. When nuclei of the two races are brought together in a common 
cytoplasm, the essential metabolites synthesized by one of the nuclear com- 
ponents supplement those synthesized by the other component. The hetero- 
caryotic mycelium is then supplied with adequate quantities of all the 
essential metabolites necessary for rapid growth. 

We have tried to test this hypothesis by supplementing with various 
substances the medium on which race 16 and other dwarf races were grown. 
If it were possible to increase materially the growth rate of the dwarf race by 
supplements in the medium, without introducing the heterocaryotic condi- 
tion, the limiting factors for dwarfness could be identified and the stimulus 
involved in the heterocaryotic condition identified. 

A basal agar medium containing mineral salts, dextrose, asparagine, neo- 
peptone, and thiamine was supplemented by a mixture of purine and pyrim- 
idine bases; by a vitamin mixture containing PAB, calcium pantothenate, 
inositol, nicotinic acid, pyridoxine, riboflavin, thiamine, guanine, hypoxan- 
thine, and 2-methyl-l, 4-naphthohydroquinone diacetate; by malt extract, 
casein hydrolysate, cow's milk, dried yeast, choline, a-tocopherol, hemin, 
oleic acid, ascorbic acid (filtered sterile), coconut milk, Taka-diastase 
(filtered sterile), water extracts of the mycelium of Neurospora, liver ex- 
tracts (both filtered sterile and heated), adrenal cortical extract (unheated), 
estrogenic substance, progesterone, anterior pituitary extract, posterior 
pituitary extract, whey, or potato extract. 

None of the substances or combinations of them as used increased the 
growth rates of any of the dwarf races to an extent adequate to explain 
heterocaryotic vigor. Some beneficial effects, usually noted only in older cul- 
tures, were obtained from cow's milk and from liver extract. These efifects 
were not sufficiently marked to suggest that either supplement supplied the 
missing factors. 

We were unsuccessful, therefore, in defining the factors limiting the 
growth of the dwarf races and conversely those effective in inducing more 
rapid growth in the heterocaryotic mycelium. 

Our failure may be explained in various ways. We may not have included 
in our various supplements the missing essential metabolites. These metabo- 



HYBRID NUTRITIONAL REQUIREMENTS 123 

lites may be non-diffusible or very labile substances such as enzyme pro- 
teins, which could only be introduced into the cell through inserting a nucleus 
and its genes. The original hypothesis may be in error. We may not be 
dealing with limiting quantities of essential metabolites but with inhibitors. 
We might assume that the growth of one or both of the parents is limited by 
autogenic inhibitors, and the presence of both kinds of nuclei in a common 
cytoplasm results in the neutralization in some fashion of the inhibitors. 

Emerson (1948) has succeeded in producing heterocaryons in which one 
kind of haploid nucleus neutralizes the effect of the other. The augmented 
growth of the heterocaryon, as compared to that of strains which are 
homozygous, reminds one, says Emerson, of instances of single gene heterosis 
in maize reported by Jones. 

The importance of internal factors in heterosis is suggested by the results 
I obtained on the growth of the excised roots of a heterotic tomato hybrid 
and its inbred parents (Robbins, 1941b). The hybrid roots and the roots of 
the two inbred parents were grown in liquid culture which contained mineral 
salts and cane sugar. This basal medium was supplemented with thiamine, 
with thiamine and pyridoxine, and with thiamine, pyridoxine, and nicotina- 
mide. 

Growth of the roots of the hybrid exceeded that of either of the inbred 
parents in all three types of media (Fig. 7.4). Growth of one parent was im- 
proved by the addition of pyridoxine to the thiamine solution, but a further 
supplement of the medium with nicotinamide had little effect. Growth of the 
second inbred parent was little affected by the addition of pyridoxine to the 
thiamine medium, but was improved by the further addition of nicotinamide 
to the thiamine and pyridoxine solution. 

These results suggest that the greater vigor of growth of the heterotic 
hybrid is determined in part by its greater ability to synthesize pyridoxine 
and nicotinamide. That is evidently not the whole story, because its growth 
exceeded that of the inbred parents in media containing all three vitamins. 

Although heterosis may be considered and should be considered from the 
genetical standpoint, it should also be studied from the physiological stand- 
point. I have suggested that it may be important to devote attention to the 
question of v/hat the internal factors are which limit growth, what they are 
in inbreds, and how they are removed in heterotic hybrids. We should con- 
sider in such investigations the role of essential metabolites, of growth in- 
hibitors, and of other specific chemical compounds which materially modify 
growth. Microorganisms might be utilized as tools for the detection of growth 
stimulators or growth inhibitors. 



EDGAR ANDERSON 

Missouri Bofanical Garden 

and 

WILLIAM L. BROWN 

Pioneer Hybrid Corn Company 



Chapter 8 

Origin of Corn Belt Moize and Its 

Genetic Significance 



Several ends were in view when a general survey of the races and varieties of 
Zea mays was initiated somewhat over a decade ago (Anderson and Cutler, 
1942). Maize, along with Drosophila, had been one of the chief tools of mod- 
ern genetics. If one were to use the results of maize genetics most efficiently 
in building up general evolutionary theories, he needed to understand what 
was general and what was peculiar in the make-up of Zea mays. Secondly, 
since maize is one of the world's oldest and most important crops, it seemed 
that a detailed understanding of Zea mays throughout its entire range might 
be useful in interpreting the histories of the peoples who have and are using 
it. Finally, since maize is one of our greatest national resources, a survey of its 
kinds might well produce results of economic importance, either directly or 
indirectly. 

Early in the survey it became apparent that one of the most significant 
sub-problems was the origin and relationships of the common yellow dent 
corns of the United States Corn Belt. Nothing exactly like them was known 
elsewhere in the world. Their history, though embracing scarcely more than 
a century, was imperfectly recorded and exasperatingly scattered. For some 
time it seemed as if we might be able to treat the problem only inferentially, 
from data derived from the inbred descendants of these same golden dent 
corns. Finally, however, we have been able to put together an encouragingly 
complete history of this important group of maize varieties, and to confirm 
our historical research with genetical and cytological evidence. 

An even approximate survey of Zea way.s-as-a-whole remains a goal for 

124 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 125 

the distant future, but our understanding of Corn Belt dent corns is already 
more complete than we had originally hoped. Since our evidence is detailed 
and of various kinds, it may make the presentation somewhat easier to follow 
if we give a brief description of the pre-hybrid commercial yellow dents of 
the United States Corn Belt, review their history in broad outline, and then 
proceed to an examination of the various kinds of evidence on which these 
generalizations have been built. 

Corn Belt dents, the commercial varieties which dominated the chief 
centers of corn production in the United States for over half a century pre- 
ceding the advent of hybrid corn, were variable open-pollinated varieties. 
They varied from plant to plant, from field to field of the same variety, and 
from variety to variety. Figure 8.1, based upon an examination of a field of 
Golden Queen, one of the lesser known of these varieties, will indicate the 
kind of variation which characterized the fields of that day. 

In spite of this variation, or one might almost say, impressed on top of it, 
was a remarkably persistent combination of generally prevalent characters. 
Considered from plant to plant or from field to field, as individuals, these 
varieties seemed ephemeral and unimportant. Seen as populations, as col- 
lections of inter-breeding individuals, the Corn Belt dents as a whole were a 
well-marked and definite entity, particularly when contrasted with maize 
in other parts of the world. They tended to have one well-developed ear, fre- 
quently accompanied by a small ear at the node below this primary one. 
The ears had large, nearly cylindrical cobs with red or reddish glumes. The 
usually golden yellow kernels, pronouncedly dented at the tip, had a peri- 
carp frequently roughened by tiny wrinkles. They were set in from 14 to 22 
straight rows with little external indication of the fact that the rows were in 
pairs. The mathematical perfection of the ear was frequently lessened by a 
slight tendency for the whole ear to taper toward the apex, and for the row- 
ing of the kernels and the diameter of the cob to be somewhat differentiated 
in its lowermost quarter. 

Characteristically, the plant on which this ear was borne had a single, up- 
right stem, leaves with tight sheaths and strong, arching blades, and a 
heavy, many -branched tassel. Kernel color was remarkably standardized, 
a faint flush of coppery red in the pericarp and a yellow endosperm, combin- 
ing to give varying shades of deep, golden color. Epidermal color was ap- 
parent on the culm and leaves at the base of the plant, but seldom or never 
were there to be found the brilliant reds, dark purples, and other foliage 
colors which are so characteristic of maize in various parts of Latin America. 
While there was some variation in anther color and silk color, pinks and dull 
reds were commonest though greens and bright reds were not unknown. 

As we have shown elsewhere (Anderson and Brown, 1950) there cannot 
be the slightest doubt that these widespread and standardized Corn Belt 
varieties were the creation of the nineteenth century. They came in large part 



20 



18 



OS. 
Ul 



C9 



16 



% 






^ 






^•V i. *- 



14 



12 








8-9 


10- 


■11 12- 


13 


14-15 






KERNEL WIDTH 






TASSEL BRANCH 
NUMBER 


GLUME LENGTH 
IN MM. 


PITH WIDTH 
IN MM. 


EAR LENGTH 
IN CM. 


• 11-20 


• 8-9 




•4-9 




•26-30 


• 21-23 


4io 




4io-ii 




• 22-25 


#^OVER 24 


• 11-13 




• 12-15 




• 14-21 



Fig. 8.1 — Pictorialized diagram showing relationship between numbers of rows of kernels, 

kernel width, tassel branch number, glume length, pith diameter, and ear length in an 

open-pollinated sample of Golden Queen dent corn. 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 127 

from crosses between White Southern Dents, mostly of Mexican origin, and 
the long, slender Northern Flints which had dominated the eastern United 
States for at least some hundreds of years preceding the discovery of 
America. While these two complexes were of primary importance in the crea- 
tion of Corn Belt corn, it should be pointed out that germ plasm of other 
types of maize has undoubtedly filtered into Corn Belt mixtures. Compared 
to Southern Dents and Northern Flints, these certainly are of minor im- 
portance. There are, nevertheless, to be found among dent inbreds of the 
Corn Belt certain strains which exhibit Caribbean influence and others 
which seem to contain germ plasm of southwestern United States or western 
Mexican varieties. 

Although the following discussion does not go into detail regarding the 
influence of these secondary sources of germ plasm on Corn Belt corn, the 
effects of such influences are important and we have already made small 
beginnings at studying them. The Northern Flints are in some ways strik- 
ingly similar to the common yellow flints of the Guatemalan highlands, strik- 
ingly unlike most Mexican maize. They are one of several cultural traits 
which apparently spread from the Mayan area to the eastern United States 
without leaving any clear record of the route by which they came. In their 
general appearance, as well as in technical botanical details, the Northern 
Flints were very different from the Southern Dents. The hybrid vigor which 
resulted from mixing these diverse types was soon noted by alert agricultur- 
ists. While some of the blending of flints and dents may have been haphazard 
and accidental, much of it was directed and purposeful. The benefits to be 
gained were listed in public, and the exact effects of continued mixing and of 
backcrossing were discussed in detail as early as 1825 (Lorain, 1825). This 
intelligent, controlled hybridizing proceeded for at least a half century until 
the new yellow dents were so ubiquitous and everyday that their very origin 
was forgotten. 

For theoretical reasons this neglect of historical tradition was unfortunate. 
Maize breeders have not understood that the heterosis they now capitalize 
is largely the dispersed heterosis of the open-pollinated flint-dent mongrels. 
Maize geneticists are for the most part unaware that the germ plasm they 
use for fundamental generalizations is grossly atypical of germ plasms in 
general. We shall return to a detailed discussion of these two points after 
referring briefly to the evidence concerning the origin of Corn Belt maize. 

Though there is abundant evidence that our Corn Belt dents came from 
mixtures of Northern Flints and Southern White Dents, the evidence con- 
cerning these two regional types is very one-sided. The Northern Flints 
(Brown and Anderson, 1947) were remarkably uniform from place to place 
and from century to century. The archaeological record is rich going back to 
early pre-Columbian times and there are numerous naive but accurate de- 
scriptions of these varieties in colonial accounts. 



128 EDGAR ANDERSON AND WILLIAM L. BROWN 

The Southern Dents (Brown and Anderson, 1948) are much more vari- 
able. For over a century their variability has been stressed by all those who 
have discussed them. The samples which we obtained from the South differed 
from field to field, and from variety to variety. For an accurate understanding 
of them and their history, we would like many more archaeological specimens 
than we have for the flints, and many more colonial descriptions. Instead, we 
have as yet no archaeological record, merely two accounts in early colonial 
times — one from Louisiana and the other from Virginia. There is one passing 
mention in a pre-revolutionary diary, and then a truly remarkable discussion 
by Lorain in 1825. Finally, the United States Patent Office report for 1850 
gives us, for region after region, a detailed picture of the extent to which this 
purposeful mixing had proceeded by that time. 

To summarize the historical evidence, the Northern Flints were once the 
prevailing type of maize throughout the eastern United States (Brown and 
Anderson, 1947) with an archaeological record going back at least to a.d. 
1000. There is as yet no archaeological evidence for their having been pre- 
ceded in most of that area by any other type of maize, or of Mexican-like 
dents having been used there in pre-Columbian times. The Northern Flints 
belong to a type of maize rare or unknown over most of Mexico, but common 
in the highlands of Guatemala. The Southern Dents, on the contrary, obvi- 
ously are largely derived from Mexican sources, and by 1700 were being 
grown as far north as Louisiana and Virginia (Brown and Anderson, 1948). 
As to how and when they spread north from Mexico, we have no evidence 
other than the negative fact that they are not known archaeologically from 
the eastern United States, and are not represented in the collections of early 
Indian varieties from that region. 

As early as 1800, the benefits of crossbreeding these two different types of 
maize were appreciated by at least a few experts. By 1850 the process was 
actively under way from Pennsylvania to Iowa, and south to the Gulf states. 
By the '70's and '80's, a new type of corn had emerged from this blending, 
although crossing and re-crossing of various strains continued up to the ad- 
vent of hybrid corn. During the latter half of the process, the origin of Corn 
Belt dents from 50 to 100 generations of selective breeding of crosses of 
Northern Flints and Southern Dents was almost completely forgotten. Hav- 
ing at length resurrected the evidence (Anderson and Brown, 1950) for this 
mingling of two fundamentally different types of maize, we shall now turn 
to the genetical and cytological evidence which first called the phenomenon 
to our attention and led us to search for historical proof. 

CYTOLOGY 

The most important cytological contribution on the origin of Corn Belt 
maize is found in a comparison of the numbers and distribution of chromo- 
some knobs in the Northeastern Flints, open-pollinated varieties of Southern 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 129 

Dents, and inbred strains of Corn Belt dents. As has been shown previously 
(Longley, 1938) and (Reeves, 1944), chromosome knobs may be an im- 
portant tool in studying relationshii)S in maize. Our work with North Ameri- 
can corn not only supports this contention, but suggests that knob data may 
be even more important than has previously been supposed. 

The 8-10 rowed flint and flour varieties of New York, Pennsylvania, and 
New England are nearly knobless. In the material we have examined, they 
have to 2 knobs. These observations are in agreement with Longley's 
earlier conclusions that maize varieties of the northern Indians were char- 
acterized by having few knobs. Longley's material, however, included no 
strains from northeastern United States — the area in which the flint an- 
cestors of Corn Belt corn were highly concentrated. It is interesting, more- 
over, to note that varieties from this segment of North America have even 
fewer knobs than do the strains from most Northern Plains Indian tribes. 

In contrast, many more knobs were to be found in the open pollinated 
varieties of Southern Dent corn. In these strains we have found numbers 
ranging from 5 to 12, for those varieties representing the least contaminated 
segment of present-day Southern Dent corn. These cytological data are in 
complete agreement with the known facts regarding the history of Northern 
Flints and Southern Dents. 

There seems little doubt that the Gourdseed-like Dents' of the southeast- 
ern United States have stemmed directly from Mexico wiiere morphological- 
ly and cytologically similar corns can be found even today. Likewise, we 
have found in highland Guatemala varieties of maize with ear character- 
istics strikingly similar to Northern Flints and with as few as three knobs. 
Insofar as cytology is concerned, therefore, it is not at all difficult to visualize 
a Guatemalan origin for Northeastern Flint corn. The Corn Belt inbreds 
with which we have worked (Brown, 1949) have knob numbers of 1 to 8. 
The distribution of numbers in these strains is almost exactly intermediate 
between that of Northern Flints and Southern Dents (Fig. 8.2). This evi- 
dence, based on a character which certainly has not been intentionally 
altered by selection, strongly fortifies the archaeological and historical facts 
pointing to a hybrid origin of Corn Belt dent corns. 

GENETIC EVIDENCE 

The genetical evidence for the origin of Corn Belt maize from mixtures 
of Northern Flints and Southern Dents is of various kinds. In its totality, it 
is so strong that, had we not been able to find the actual historical evidence, 
we could have determined what had happened from genetic data alone. In 
the first place we have demonstrated, by repeating the cross, that it is pos- 
sible to synthesize Corn Belt dents from hybrids between Southern Dents 

I The name "Gourdseed" has been used since colonial times to describe the extremely- 
long seeded, white Southern Dents, whose kernels are indeed not so diflferent in appear- 
ance from the seeds of gourds of the genus Lagenaria. 



130 



EDGAR ANDERSON AND WILLIAM L. BROWN 



and Northern Flints. Our experiments in crossing a typical white gourdseed 
from Texas and a typical yellow flint from New York State are now only in 
the third generation and are being continued. However, it is already evident 
that some of the segregates from this cross are within the range of varia- 
tion of Corn Belt dents (Fig. 8.3). 

In spite of the 50 to 100 generations of mixing which has taken place, the 
characters of Northern Flints and Southern Dents still tend to be associated 
in Corn Belt dents. Anderson (1939) has shown that in crosses between species 



10 



OLD SOUTHERN OENTS 



10 

> 
u 

Z 
lit 

3 

a 

^20 
li. 



10 



DERIVED SOUTHERN OENTS 



CORN BELT INBREDS 



10 







1 


























NOR 


1 

THERN ruiNTS 







1 ' 2 


' 3 




4 ' 


5 ' 


6 


7 




8 •■ 


9 ' 


10 ' 


II 


' 12 


"I 



NUMBER OF CHROMOSOME KNOBS 

Fig. 8.2 — Frequency distribution of chromosome knobs in Northern Flints, Southern 

Dents, and Corn Belt inbreds. 



or between races, all the multiple factor characters which characterize each 
are partially linked with one another and tend to remain associated, even 
after generations of controlled breeding. More recently he has used this 
principle in the development of the method of extrapolated correlates (Ander- 
son, 1949) by which the original characteristics can be deduced from the mix- 
tures even when previously unknown. 

Using this method in a relatively crude form, we were able (in advance 
of our historical evidence) to demonstrate (Brown, 1949) in Corn Belt in- 
breds, the association of low knob numbers, flag leaves, cylindrical ears, few 
tassel branches, and flinty kernels — all characteristics which typify the 
Northern Flints. Similarly, it was possible to show the association among 
these 98 Corn Belt inbreds of high knob numbers, no flag leaves, tapering 
ears, dented kernels, and many tassel branches — a combination of char- 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 131 

acters which is typical of the Southern Dents. As a matter of fact, by this 
technique Brown predicted the knob numbers of the Northern Flints, even 
when that fact was unknown to us. 

The association of characters in actual open-pollinated fields of Corn Belt 
dents is so complex that one might suppose any study of it would be hopeless. 
However, from a study of character association in an open-pollinated field 




Fig. 8.3 — Corn Belt Denl-like segregates from an Fo generation of cross of Longfellow 

Flint X Gourdseed Dent. 

of Golden Queen Dent (Fig. 8.1) we were able to demonstrate the association 
of: (1) wide kernels, (2) low row numbers, (3) short glumes, (4) few tassel 
branches, (5) long ears, and (6) narrow central pith in the ear — all of these 
characterizing Northern Flints. The opposing combination: (1) narrow 
kernels, (2) high row numbers, (3) long glumes, (4) many tassel branches, 
(5) short ears, and (6) wide central pith also tended to be associated and is 
characteristic of Southern Dents. In other words, some of the characters 
which went in together from flints and dents were still in this open-pollinated 
variety tending to stay together on the average. The existence of such char- 
acter complexes has been appreciated by experienced corn breeders, though 
apparently it has never been commented on in print. Of course, corn breed- 
ers and corn geneticists differ in their endowments for apprehending such 



132 EDGAR ANDERSON AND WILLIAM L. BROWN 

phenomena in advance of the published facts, and the existence of these 
strong linkages has been more apparent to some than to others. 

WIDTH OF CROSS 

The demonstration that Corn Belt dents largely are derived from hy- 
bridization between Southern Dents and Northern Flints is of particular im- 
portance because this is such a wide cross. Our evidence for this assertion is 
largely morphological, though there is supporting evidence from cytology 
and genetics. 

In nearly all species of cultivated plants there are conspicuous differences 
in color and shape. These differences give the various cultivated varieties of 
a species a false aspect of difference from one another, and from their wild 
progenitors. False, because these differences are usually due to a few genes, if 
not being actually monofactorial. The striking differences between such 
varieties are therefore no true indication of the distinctness of their germ 
plasms. 

On the other hand, there are subtle differences in form, proportion, and 
indument which, though difficult for a novice to apprehend, are more like 
the differences which distinguish distinct species of the same genus. These 
taxonomically important differences have proven valid criteria for indicating 
the diversity of germ plasms. So it has been proven that the subtle taxonomic 
differences between the Old World and New World cottons are much more 
representative of the genetic diversity and relationships of these two groups 
of varieties than are the conspicuous differences in color and leaf-shape which 
are found within each group. In the Cucurbits the striking differences in 
color and form of fruit, which differentiate the varieties of Cucurbita Pepo 
and of C. moschata, are superficial compared to the taxonomically significant 
features which separate these two groups. The latter, moreover, have been 
proved to be a significant index of genetic diversity, either between these 
two groups of Cucurbits or in assaying the variation within C. Pepo itself 
(Shifriss, 1947) (Whitaker and Bohn, 1950). 

The difficulty in relying upon such taxonomic criteria is that the method 
is highly subjective. Taxonomy is of necessity still more of an art than a 
science. This means that one must personally examine the evidence if his 
opinion is to be worth anything. It also means that the worker's opinion is 
worth no more than his understanding of the taxonomic entities included in 
his judgment. However, until more objective criteria are evolved for this 
field, we shall have to use fairly traditional taxonomic methods for want of 
anything better. Accordingly, the senior author has for two years spent one 
day a week in a technical, agrostological, herbarium survey of all the grasses 
conceivably related to Zea mays — all the genera in the tribes Andropogoneae 
and Maydeae. With that background, his judgments may well be mistaken 
but they are certainly informed. 

From this point of view, the variation within Zea mays is without parallel, 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 133 

not only in the cultivated cereals but in any other domesticated plant or 
animal. There are such superficial characters as aleurone color, pericarp 
color, plant color, carbohydrate composition, and such amazing single factor 
differences as tunicate and teopod. In addition, there are a whole battery of 
characters which are difficult to work with genetically, but which are the 
kinds of differences that agrostologists find significant in the deployment of 
species and genera: spikelet shape and venation, spikelet arrangement, 
rachis morphology, pubescence, leaf-shape, internode proportions, etc. Using 
such criteria, the hybridization of the Southern Dents and the Northern 
Flints represents the mingling of two basically different germ plasms. 

For evidences of relationship, the male inflorescence of maize (the tassel) 
is of particular importance. Inflorescence differences generally have proved 
to be of primary taxonomic importance in the Gramineae. Variation in the 
male inflorescence of Zea would likely be less obscured by domestication than 
the female inflorescence (the ear) which has been deliberately selected for 
various peculiarities. The entire male inflorescence of the Southern Dents 
has been extensively modified by condensation (Anderson, 1944), a sort of 
fasciation which telescopes adjacent nodes, and in the ear produces increases 
in row number. It is an abnormality conditioned by at least two pairs of 
recessive genes and its expression is certainly modified by still other genes. 

Tassels of the Northern Flints are without any condensation. Though 
condensation modifies the general aspect of the tassel, it is relatively super- 
ficial. The presence of so much condensation renders difficult the demonstra- 
tion of a much more fundamental difference. The central spike of the North- 
ern Flints is decussately arranged. That is, the pairs of spikelets are in alter- 
nate whorls of two; whereas the spike of the Southern Dents (allowing for 
the modifications produced by extreme condensation) is fundamentally in 
whorls of 3, or mixtures of whorls of 3 and whorls of 2. The rachis of the 
Northern Flints is slender with long internodes, that of the Southern Dents 
is short and flattened (Fig. 8.5). Pedicels of the upper spikelets always are 
long in the Northern Flints. In the Southern Dents they may be so short that 
one cannot distinguish the normally pedicellate spikelet from its sessile 
partner. 

Correlated differences are seen in the ear. That of the Northern Flints has 
a narrow central pith and is long and slender, characteristically with 8-10 
rows. The ear of the Southern Dents is short and thick with a wide central 
pith, and with from 16 to 30 or more rows. Pairing of the rows is markedly 
evident in the Northern Flints, even when they are pushed closer together 
in those occasional ears with 10 or 12 rows (Fig. 8.4). There is little or no row 
pairing in the Southern Dents. The kernel of the Southern Dents is long, flat, 
and narrow. Its largest diameter is near the base. By contrast, the kernel of 
the Northern Flints is wider than it is high, and is considerably thicker 
at the apex than it is at the base. 

The ear of Zea mays is terminal on a secondary branch, which is hidden by 



.^ r^ 



-n 



-.», -iii*; iafe. 







S-2 



M. 







S-l 




N-3 




N-2 




N-4 



t- '■ 



S-4- 



*•-. 




S-3 



Fig. 8.4 — Typical ears (/), shanks (2), and seeds {3 and 4) of Northern Flint (iV), and 

Southern Dent (5). 




Fig. 8.5 — Typical plants, tassels, and staminate spikelets of Northern Flint and Southern 

Dent. 



136 EDGAR ANDERSON AND WILLIAM L BROWN 

its specialized leaves or husks. When dissected out, these ear shoots (or 
shanks) are diagnostically different in Northern Flints and Southern Dents 
(Fig. 8.4). In the former they are long, with elongated internodes which are 
widest between the nodes, and which have a smooth surface upon drying. In 
the latter they are very short, frequently wider at the nodes than between 
them, and have a characteristically ribbed surface upon drying. 

The leaves of the Northern Flints are long and slender and frequently a 
light green. Those of the Southern Dents are proportionately wider and 
shorter and are often dark green. They are set upon culms whose internodes 
are proportionately longer and more slender in the Northern Flints, and less 
prone to become greatly shortened at the internodes immediately above the 
ear. 

If we ignore such abnormalities as differences in carbohydrate composition 
and condensation, these two races of Zea mays still are widely different from 
one another — as compared to differences between their wild relatives in the 
Andropogoneae or the Maydeae. The differences in internode pattern and 
proportion and in leaf shape are similar to those frequently found between 
species of the same genus. The differences in pedicellation of the upper spike- 
let would be more characteristic of genera and sub-genera. On the other 
hand, in the whorling of the central spike (whorls of 2 versus whorls of 3) 
is the kind of difference which would ordinarily separate genera or even 
groups of genera. On a par with this difference are those in the cupule (the 
bony cup in which the kernels are attached in pairs). They are so difficult to 
observe that we cannot discuss these until the general morphology of this 
organ has been described. If we sum up the morphological evidence, it is clear 
that the fundamental differences between the Northern Flints and the 
Southern Dents are similar to those which differentiate distantly related 
species (or even genera) among related wild grasses. There is every morpho- 
logical indication, therefore, that we are dealing with two fundamentally 
different germ plasms. 

The cytological facts reported above lend further weight to the conclusion 
that the Northern Flints are basically different from the Southern Dents. 
The former have chromosomes which are essentially knobless at pachytene. 
The latter average nearly one knob per chromosome (Fig. 8.2). Heterochro- 
matic knobs are known in other grasses besides Zea mays. In these other 
genera, their presence or absence, from such evidence as is available, seems 
to be characteristic of whole species or groups of species. Such a difference 
between the Flints and Dents indicates that we are dealing with two funda- 
mentally different germ plasms. It has been shown in Guatemala (Mangels- 
dorf and Cameron, 1942) and in Mexico (Anderson, 1946) that the varieties 
with many knobs are morphologically and ecologically different from those 
with low numbers of knobs. 

A further indication that these two germ plasms are physiologically dif- 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 137 

ferent is given by their pachytene behavior. The pachytene chromosomes of 
the Northern Flints are easy to smear and give sharp fixation images. South- 
ern Dents are more ditlficult to smear. The chromosomes do not spread out 
well and do not stain sharply. This is not a result of differences in knob num- 
ber, since some of the Mexican Dents with few knobs are equally difilicult to 
smear. Whatever the physiological significance of this reaction, it is direct 
evidence for a difference in the chemistry of the germ cells. Again such dif- 
ferences in stainability are more often met with, between genera, than they 
are in different strains of the same species. 

There is genetic evidence for the difference between Southern Dents and 
Northern Flints, in the behavior of crosses between them. The F/s are fully 

TABLE 8.1 

PERCENTAGE OF STERILE OR BARREN PLANTS IN 

GOURDSEED, LONGFELLOW, AND Fa GENERATION 

OF CROSS GOURDSEED X LONGFELLOW 



Total 
Number 
of Plants 



Gourdseed 

Longfellow 

Fo Gourdseed X Longfellow 



Sterile 






Normal 


or 






Ear 


Barren 




37 


63 


2 


98 


52 


48 



fertile and exhibit extreme hybrid vigor. The Fo's show a high percentage of 
completely barren plants — plants which formed ears but set little or no seeds, 
either because of sterility or because they were too weak to mature success- 
fully — and plants which managed to set seeds, though their growth habit 
indicates fundamental disharmonies of development. 

Table 8.1 shows the percentages of good ears and plants which were either 
without ears or on which the ears had failed to set any seed, for Gourdseed- 
Dent, Longfellow Flint, and their F2, when grown in Iowa. Like Southern 
Dents generally, the Gourdseed is less adapted to central Iowa than is Long- 
fellow Flint. An F2 between these two varieties, however, has a much greater 
percentage than either parent of plants which are so ill-adapted that they 
either produce no visible ear, or set no seed if an ear is produced. Similar 
results were obtained in other crosses between Northern Flints and Southern 
Dents, both in Missouri and in Iowa. From this we conclude that they are 
so genetically different from one another that a high percentage of their F2 
recombinations are not able to produce seed, even when the plants are care- 
fully grown and given individual attention. 

SUMMARY 
The common dent corns of the United States Corn Belt were created 
de novo by American farmers and plant breeders during the nineteenth cen- 



138 EDGAR ANDERSON AND WILLIAM L BROWN 

tury. They resulted in a large measure from deliberate crossing and re- 
crossing of two races of maize (the Northern Flints and the Southern Dents) 
so different that, were they wild grasses, they would be considered as totally 
different species and might well be placed in different genera. The origin of 
two so-different races within cultivated maize is an even larger problem and 
one outside the scope of this discussion. It may be pointed out parentheti- 
cally that the Tripsacum hypothesis (Mangelsdorf and Reeves, 1945) would 
not only account for variation of this magnitude, it would even explain the 
actual direction of the difference between these two races of maize. However, 
the relation between maize and Tripsacum on any hypothesis is certainly a 
most complicated one (Anderson, 1949). It would be more effective to post- 
pone detailed discussions of this relationship until the comparative morphol- 
ogy of the inflorescences of maize and of Tripsacum is far better understood 
than it is at present. 

SIGNIFICANCE TO MAIZE BREEDING 

Derivation of the commercial field corns of the United States by the de- 
liberate mingling of Northern Flints and Southern Dents is a fact. Unfortu- 
nately, it is a fact which had passed out of common knowledge before the 
present generation of maize breeders was educated. From the point of view 
of practical maize breeding, either hybrid or open-pollinated, it is of central 
importance. Briefly, it means that the maize germ plasms now being worked 
with by plant breeders are not varying at random. They are strongly 
centered about two main centers or complexes. Such practical problems as 
the development and maintenance of inbreds, the detection of combining abil- 
ity, and the most effective utilization of hybrid vigor need to be rethought 
from this point of view. Detailed experiments to provide information for such 
practical questions already are well under way. While these experiments are 
not yet far enough along to give definite answers, they have progressed far 
enough to allow us to speak with some authority on these matters. 

HETEROSIS 

The heterosis of American Corn Belt dents acquires a new significance in 
the light of these results, and practical suggestions as to its most efficient 
utilization take on a new direction. We are immediately led to the hypothesis 
that the heterosis we are working with is, in part at least, the heterosis ac- 
quired by mingling the germ plasms of the Northern Flints and the Southern 
Dents. 

Insofar as hybrid vigor is concerned, the hybrid corn program largely has 
served to gather some of the dispersed vigor of the open-pollinated dents. 
Preliminary results indicate that this has not been done efliciently in terms 
of what might be accomplished with somewhat more orientation. 

The early days of the hybrid corn program were dominated by the hy- 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 139 

pothesis that one could inbreed this vigorous crop, identify the inferior 
strains in it, and then set up an elite cross-pollinated germ plasm. This hy- 
pothesis was clearly and definitely stated by East and Jones {Inbreeding and 
Outbreeding, 1919, pp. 216-17). 

Experiments with maize show that undesirable qualities are brought to light by self- 
fertilization which either eliminate themselves or can be rejected by selection. The final re- 
sult is a number of distinct types which are constant and uniform and able to persist in- 
definitely. They have gone through a process of purification such that only those individu- 
als which possess much of the best that was in the variety at the beginning can survive. The 
characters which they (pure lines) have, can now be estimated more nearly at their true 
worth. By crossing, the best qualities which have been distributed to the several inbred 
strains can be gathered together again and a new variety recreated. After the most desirable 
combinations are isolated, their recombination into a new and better variety, which could 
be maintained by seed propagation, would be a comparatively easy undertaking. 

Though other corn breeders and corn geneticists may not have committed 
themselves so definitely in print, such a notion was once almost universal 
among hybrid corn experts. Modified versions of it still influence breeding 
programs and are even incorporated in elementary courses in maize breeding. 

The facts reported above would lead us to believe that heterosis, having 
resulted from the mingling of two widely different germ plasms, will probably 
have many genes associated with characters which in their relatively homo- 
zygous state are far from the Corn Belt ideal of what a corn plant should look 
like. It is highly probable that much of the so-called "junk" revealed by in- 
breeding was extreme segregants from this wide cross, and that it was closely 
associated with the genes which gave open-pollinated dents their dispersed 
vigor. It is significant that some very valuable inbreds (L317 is a typical ex- 
ample) have many undesirable features. For this reason, many such inbreds 
are automatically eliminated even before reaching the testing stage. 

If one accepts the fact that Corn Belt dents resulted from the compara- 
tively recent mingling of two extremely different races of maize, then on the 
simplest and most orthodox genetic hypotheses, the greatest heterosis could 
be expected to result from crosses between inbreds resembling the Southern 
Dents and inbreds resembling the Northern Flints. If heterosis (as its name 
implies) is due to heterozygous genes or segments, then with Corn Belt corn 
on the whole we would expect to find the greatest number of differing genes 
when we reassembled two inbreds — one resembling the Northern Flint, the 
other resembling the original Southern Dent. 

Theory (Anderson, 1939a), experiment (Anderson, 1939b; Brown, 1949), 
and the results of practical breeding show that linkage systems as differenti- 
ated as these break up very slowly. On the whole, the genes which went in 
together with the Northern Flints still tend to stay together as we have 
demonstrated above. This would suggest that in selecting inbreds, far from 
trying to eliminate all of the supposed "junk," we might well attempt to 
breed for inbreds which, though they have good agronomic characters like 
stiffness of the stalk, nevertheless resemble Northern Flints. On the other 
hand, we should breed also for those which resemble Southern Dents as close- 



140 EDGAR ANDERSON AND WILLIAM L. BROWN 

ly as they can and still be relatively easy to grow and to harvest. It would 
seem as if the opposite generally has been done. A deliberate attempt has 
been made to produce inbreds which look as much as possible like good Corn 
Belt maize in spite of being inbreds. 

There are, of course, practical necessities in breeding. In this direction the 
work of corn breeders is a remarkable achievement. Strong attention to lodg- 
ing resistance, to desirable kernel shapes and sizes, and to resistance to 
drought and disease has achieved real progress. The inbred-hybrid method 
has permitted much stronger selection for these necessary characters than 
was possible with open-pollinated maize. Most Corn Belt dents now plant 
well, stand well, and harvest well. 

Perhaps partly because of these practical points there has been a conscious 
and unconscious attempt on the part of many breeders to select for inbreds 
which are like the Corn Belt ideal in all characters, trivial and practical 
alike. The corn shows are now out-moded, but corn show ideals still influence 
corn breeding. For instance, there has been an effort to produce plants with 
greatly arching leaves, whose margins are uniformly ruffled. Such characters 
are certainly of a trivial nature and of secondary importance in practical pro- 
grams. Any potential heterosis closely associated with upright leaves, yellow 
green leaves, tillering, or blades on the husk leaves has seldom had a chance 
to get into inbreds where it could be tested on a basis of achievement. It 
would seem highly probable that, in not basing the selection of inbreds more 
soundly on performance, we have let much potential heterosis slip through 
our sieve of selection. 

Heterosis Reserves 

These considerations lead us to believe that there is probably a good deal 
of useful heterozygosis still ungathered in high yielding open-pollinated 
varieties. There is also a distinct possibility that still more could be added 
by going back to the Northern Flints and Southern Dents with the specific 
object of bringing in maximum heterozygosity. From our experience it is 
more likely that superior heterosis is to be found among the best flints than 
among the best dents. On the whole, the Northern Flints have been farthest 
from the corn breeders' notion of what a good corn plant should look like. 
Flint-like characteristics (tillering, for example) have been most strongly 
selected against, both in the open-pollinated varieties and the inbreds derived 
from them. 

Several of the widely recognized sources of good combining inbreds are 
open-pollinated varieties with a stronger infusion of Northern Flints than 
was general in the Corn Belt. This is particularly true of Lancaster Surecrop, 
the excellence of whose inbreds was early recognized by several breeders in 
the United States Department of Agriculture. In our opinion, it is probable 
that the greater proportion of flint germ plasm in Lancaster Surecrop has 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 141 

made it an outstandin<f source of inbreds of proven highly sj)ecific combining 
ability when used with other Corn Belt inbreds. This is not an isolated ex- 
ample, and even more extreme cases could be cited. We think it is a reason- 
able working hypothesis that Northern Flint varieties of superior productiv- 
ity might be efficient sources of improved heterozygosity for the United 
States Corn Belt. 

Morphological Characters as Related to Heterosis 

To put this hypothesis in different language, morphological characters, if 
carefully chosen, may be used as criteria of specific combining ability in Corn 
Belt inbreds. Before presenting data bearing directly on this hypothesis, two 
points need to be emphasized and discussed: (1) the effective selection of 
morphological criteria, and (2) the relativity of all measures of effective 
combining ability. 

Previous studies (Kiesselbach, 1922; Jenkins, 1929; and others) have indi- 
cated that the only positive correlations between the morphology of inbreds 
and their combining ability are those involving characters of the inbreds 
which are indicative of plant vigor. Reference to these investigations shows 
that the characters chosen were such superficial measurements as date of 
silking and tasseling, plant height, number of nodes, number of ears, ear 
diameter, etc. Unfortunately, the morphology of the maize plant is not a 
simple matter. It is so complex that one needs technical help on morphology 
quite as much as he would in biochemistry were he studying the concentra- 
tions of amino acids in the developing kernel. 

Accordingly, we first familiarized ourselves thoroughly with the technical 
agrostological facts concerning the detailed gross morphology of grasses in 
general and Zea in particular. Just as in the case of a biochemical study of the 
kernel, we found that further original research was necessary if the investiga- 
tion was to be carried on effectively. We have accordingly undertaken de- 
tailed studies of internode patterns and branching of the inflorescence; the 
venation, size, and shape of the male spikelet, the development of the husk 
leaf blades, the external anatomy of the cob, and the morphology of the 
shank. Some of these investigations are still continuing, and must continue 
if inbred morphology and combining ability are to be effectively correlated. 

It is impossible to produce an absolute measure of combining ability. 
When one speaks of combining ability of two inbreds, he always refers to 
their behavior with each other compared to their behavior with certain other 
inbreds or open-pollinated varieties. This is such a relative measure that the 
scoring of a particular Fi cross as very low or very high in combining ability 
might depend solely upon our previous experience with the two inbreds. We 
may illustrate this point with an extreme example. Let us suppose that we 
have inbreds IF and 2F derived directly from Northern Flints, and inbreds 
lOD and IID derived from Southern Dents. Were we to cross IF X 2F and 



142 EDGAR ANDERSON AND WILLIAM L. BROWN 

lOD X IID we would expect relatively little heterosis within either of the 
crosses. Accordingly, when we crossed 2F X IID we would rate this cross 
as having high specific combining ability. On the other hand, had we origi- 
nally crossed 2F X lOD and IID X IF, then there would probably have 
been almost equally great heterosis in each of the crosses. Had these been 
used as a basis for comparing the heterosis of 2F X IID, then our notion as 
to the amount of heterosis in these crosses would have been very different 
than it would have been had comparisons been made with IF X 2F or 
lOD X IID. 

If the germ plasms of the two main races of maize involved in Corn Belt 
dents are still partially intact as a result of linkages, it should be possible to 
classify inbreds on the basis of morphological differences according to their 
flint and dent tendencies. If this can be done, and if genetic diversity is im- 
portant in bringing about a heterotic effect in hybrids, one should be able to 
predict with some accuracy the relative degree of heterosis to be expected 
from crossing any two inbred lines. With this hypothesis as a background, a 
series of experiments was started three years ago to determine whether or not 
hybrid vigor in maize, as expressed in terms of grain yield, could be predicted 
on the basis of morphological differences of inbreds making up the Fi hybrids. 

Fifty-six relatively homozygous inbred lines consisting of eighteen 
U.S.D.A. or experiment station lines, and thirty-eight strains developed by 
the Pioneer Hi-Bred Corn Company were scored for the following character- 
istics: row number, kernel length, denting, development of husk leaf blades, 
number of secondary tassel branches, glume length, and chromosome knob 
number. For each of these characteristics the two extremes in the eastern 
United States are to be found in Southern Dents and Northeastern Flints. 
At least twelve plants of each of the fifty-six inbreds were scored, and these 
scores were then averaged to give a mean value for the line. The resulting 
means were translated into numerical index values, in which a low value 
represents Northern Flint-like tendencies, and a high value Southern Dent- 
like tendencies. For example, the mean row number values for the inbreds 
studied ranged from 11.2 to 19.5. These were arranged in the following index 
classes. 

1 2 3 4 S 6 7 

11.2-11.7 11.8^12.3 12.4-12.9 13.0-13.5 13.6-14.1 14.2-14.7 14.8-15.3 

8 9 10 11 12 13 14 

15.4-15.9 16.0-16.5 16.6-17.1 17.2-17.7 17.8-18.3 18.4-18.9 19.0-19.5 

Index values for the other characteristics were arranged similarly, and 
from the individual characteristic inbred indices (each being given equal 
weight) a total "Inbred Index" was determined as is shown by example in 
Table 8.2. 

After index values had been determined for the inbreds, single cross combi- 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 



143 



nations were made and these tested for yield. In 1948, sixty-six single crosses 
were grown in yield tests in Iowa and in Illinois. Each Fi hybrid was repli- 
cated six times in each test. At the end of the season, yield of grain was de- 
termined on the basis of 15 per cent moisture corn. Actual yields in bushels 
per acre and morphological differences of the inbreds involved in each of the 
crosses were then plotted on a scatter diagram as shown in Figure 8.6. It will 
be noted that although the observations exhibit considerable scatter, there is 
a tendency for grain yields in single crosses to increase as the morphological 
differences between the inbreds making up the crosses become greater. 
Actually the correlation coefficient between yield and index differences in 
this case was r = +.39. 

The experiment was continued in 1949, in which 100 Fi hybrids were 
tested for yield. In this experiment three characters only were used to deter- 

TABLE 8.2 
INBRED INDICES BASED ON SEVEN CHARACTERS 



Inbreds 


Row 

No. 


Kernel 
Length 


Dent- 
ing 


Husk 
Leaves 


Tassel 
Branches 


Spikelet 
Length 


Chromo- 
some 
Knobs 


^ , J Sums of 7 
Inbred ^.„ 
^ , Differences 
Index .,, ^ „. 

without Signs 


Hy 

Oh40b.. 
MYl... 


9 

2 

14 


14 

8 

11 


4 

4 
14 


14 

1 
14 


5 

4 

14 


6 
1 
6 


12 

3 
9 


f± 


IV/ 30 

82 / 



mine the index of relationship between the inbreds used. These were row 
number, kernel length, and degree of development of husk leaf blades. 
Elimination in this experiment of certain morphological characteristics used 
previously was done largely to facilitate ease and speed of scoring. It had 
been determined previously that, of the several characteristics used, those 
having the highest correlation with yield were differences in row number, 
kernel length, and husk leaf blades. There was likewise known to be a rather 
strong association between each of these characteristics and tassel branch 
number, denting, glume length, internode pattern, and chromosome knob 
number. Therefore the scoring of these three characteristics probably covers 
indirectly nearly as large a segment of the germ plasm as would scores based 
on all seven characteristics. 

The 1949 tests in which each entry was replicated six times in each loca- 
tion were again grown both in Iowa and Illinois. Yields from these tests, 
plotted against index differences of the inbreds, are shown in Figure 8.7. 
As in the previous year's data, a pronounced tendency was shown for hybrids 
made up of inbreds of diverse morphology to produce higher grain yields than 
hybrids consisting of morphologically similar inbreds. The correlation co- 



1948 







. 
























; 












# 






• 




115 


" 










• 


• 


• 






u 




\ 






Mk 




• 


W 


• 














V 














DC 


no 


. 






• 




• 








# 


(J 




• 




• 


• 


• 


• 


• 


• 
• 


• 


# 


^ 




• 




• 


• 


• 






• 






D 


105 


• 


• 


• 


• 
• 


• 


• 


• 
• 


• 


• 


• 


OQ 




• • 


• 


• 


• 






• 


• 






1 










• 
• 


• 
• 


• 






# 


• 


Q 


100 


- 


• 




• 


• 




• 








J 




• • 


• 


• 








• 








LiJ 




• # 












• 








> 


95 
90 










• 




• 










85 








• 















5 10 

NDEX DIFFERENCE 



r=+.38 



Fig. 8.6 — Scatter diagram depicting relationship between grain yields of 66 single cross 
hybrids and morphological differences of inbred parents of the hybrids. Explanation in text. 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 



145 



efficient between yield and index differences is r = +.40, a significant value 
statistically. 

In terms of practical corn breeding, the distribution of single crosses in 
Figures 8.6 and cS.7 is of particular significance. If these observations are 
critical (we have produced a repea table result) it means that one could have 
eliminated from the testing program the lower one-third of the crosses on the 
basis of index differences, without losing any of the top 10 per cent of the 
highest yielding hybrids. In the case of the 100 hybrids in Figure 8.7, one could 
have eliminated from testing 35 per cent of the crosses, thereby permitting the 
inclusion of 35 additional hybrids in this particular testing area. If further 

1949 



130 



125 • 



120 ■ 



3 lis 
CD 

I I 10 

a 

LJ 10^ 



LJ 

q: 
u 



> 



00 



95 






• • • • 










• • • • • 



:.• • 



• • 






• • 



5 10 15 20 

INDEX DIFFERENCE 



25 



r= +.40 



^x 



Fig. 8.7 — Scatter diagram depicting relationship between grain yields of 100 single cross 
hybrids and morphological differences of inbred paren ts of the hybrids. Explanation in text . 



146 EDGAR ANDERSON AND WILLIAM L BROWN 

experiments show that the method is reliable, such a procedure should 
expedite most corn breeding programs. 

Our method of scoring does not take into account the variation brought 
about by the infusion of germ plasm other than that from Northern Flints 
and Southern Dents. Perhaps this is one reason why we have not ob- 
tained higher correlations between differences in inbred morphology and 
yield. There are a few inbreds in the Corn Belt which appear to be affiliated 
either with Caribbean flints or the Basketmaker complex. Scoring of such 
inbreds on a scale designed for Northern Flints and Southern Dents un- 
doubtedly leads to conflicting results. It is hoped that experiments now in 
progress will aid in clarifying this situation. 

SIGNIFICANCE OF FLINT-DENT ANCESTRY 
IN CORN BREEDING 

The Flint-Dent ancestry of Corn Belt maize bears upon many other breed- 
ing problems besides those concerned with heterosis. Its widest usefulness is 
in giving a frame of reference for observing and thinking about the manifold 
and confusing variation of Corn Belt maize. When one becomes interested 
in any particular character of the corn plant, he no longer needs to examine 
large numbers of inbreds to understand its range of variation and its general 
over-all direction. He merely needs to examine a few inbreds, and a Northern 
Flint and a Southern Dent. A good part of the variation will then be seen to 
fall into a relatively simple series from an extreme Northern Flint type to the 
opposite Southern Dent extreme, with various intermediates and recombina- 
tions in between. This is quite as true for physiological or biochemical char- 
acters as for glumes, lemmas, or other morphological characters. One is then 
ready to study further inbreds with a framework in his mind for sorting out 
and remembering the variation which he finds. 

The actual breeding plot efficiency of this understanding will be clearer if 
we cite a practical example. Now that corn is picked mechanically, the size, 
shape, texture, and strength of shank are important. When maize was picked 
by hand, the hand had a brain behind it. Variations in ear height, in the 
stance of the ear, and in the strength and shape of the shank were of minor 
significance. Now that machines do the work, it is of the utmost practical im- 
portance to have the shank standardized to a type adapted to machine 
harvesting. When this necessity was brought to our attention a few years 
ago, there were few published facts relating to variation in the shank. Exam- 
ination of a few inbreds showed that though this organ varied somewhat 
within inbreds, it varied more from one line to another than almost any 
other simple feature of the plant. We accordingly harvested typical shanks 
from each of 164 inbreds being grown for observation in the breeding plots 
of the Pioneer Hi-Bred Corn Company. We also examined a number of 
Northern Flints, and had they been available, we would have studied the 



ORIGIN AND SIGNIFICANCE OF CORN BELT MAIZE 



147 



shanks on typical Southern Dents. However, simply b}- using the hyj)othesis 
that one extreme would have to come in from the Northern Flints, the other 
from the Southern Dents, we were able within one working day to tabulate 
measurable features of these shanks and to incorporate all the facts in a 



< 



X 
h- 

o 

z 

LJ 



< 

I- 

o 

I- 











'4 






4 
















V 




)l 






















i 




























V 


4 










265 








i V 
4 4 

4 4 

U 4 


4 

'^4 4 


V 


u 

'i 




4 


V 


t 


210 








4 '* ^ 4 


V / 
^4 


4 




\ 




4 










/ 




U 4 


'i 


u 

w 


4 




)r 








W' 




4^4 ^i 4 ^i 


4 


u 


% 








^w 


135 








i, ^ 

V 


V 


V 

*- 












V 


^ 


¥ 


t )i 


^ 4 




i 


















a 
^ 


4 


• 












60 

























10 



15 



20 



WIDTH OF MID-INTERNODE OF SHANK 



LENGTH OF 


MAXIMUM WIDTH - 


CONDENSED 


LONGEST INTERNODE 


MINIMUM WIDTH 


INTERNODES 


• 0-15 


• 


\o 


¥ 16-35 


i 1-3 


■* 1-2 


436- + 


i4- + 


• 3■-^ 



Fig. 8.8 — Pictorialized diagram showing relationship in 164 Corn Belt inbreds of the fol- 
lowing shank characters: total length, width of mid internode, length of longest internode, 
maximum width minus minimum width, and number of condensed internodes. 



148 EDGAR ANDERSON AND WILLIAM L. BROWN 

pictorialized scatter diagram (Fig. 8.8). Using the method of Extrapolated 
Correlates we were able to reconstruct the probable shank type of the 
Southern Dents. (We later grew and examined them and verified our predic- 
tions.) We arranged most of the facts concerning variation in shank type in 
United States inbreds in a single, easily grasped diagram. All the technical in- 
formation needed as a background for breeding was available after two days' 
work by two people. Without the Northern Flint-Southern Dent frame of 
reference for these miscellaneous facts, we might have worked around the 
problem for several breeding seasons before comprehending this general, over- 
all picture. 

SUMMARY 

1. Archaeological and historical evidence shows that the common dent 
corns of the United States Corn Belt originated mainly from the purposeful 
mixing of the Northern Flints and the Southern Dents. 

2. Cytological and genetic evidence point in the same direction and were 
used in the earlier stages of our investigations before the complete historical 
evidence had been located. 

3. The Northern Flints and Southern Dents belong to races of maize so 
dififerent that, were they wild grasses, they would certainly be assigned to 
different species and perhaps to different genera. Such cytological and 
genetical evidence as is available is in accord with this conclusion. 

4. The significance of these facts to maize breeding problems is outlined. 
In the light of this information, the heterosis of Corn Belt maize would seem 
to be largely the heterosis acquired by mingling the germ plasms of the 
Northern Flints and Southern Dents. It is pointed out that most breeding 
programs have been so oriented as to be inefficient in assembling the dis- 
persed heterosis of the open-pollinated varieties of the Corn Belt. The possi- 
bility of gathering more heterosis from the same sources is discussed and it is 
suggested that more might be obtained, particularly among the Northern 
Flints. 

5. Morphological characters of dents and flints, if carefully chosen, should 
be useful criteria for specific combining ability. The problem of selecting 
such characters is described. Two seasons' results in correlating combining 
ability and flint-dent differences are reported. They are shown to be statisti- 
cally significant and of probable practical importance. 

6. The practical advantages of understanding the flint-dent ancestry of 
Corn Belt maize are discussed and illustrated by example. In brief these facts 
provide a "frame of reference" for detecting, organizing, and understanding 
much of the manifold variability in Corn Belt maize. 



ADRIANO A. BUZZATI-TRAVERSO 

Universiia, Isfituto di Gene/ico, Pavia, Ifaly 



Chapter 9 

Heterosis in 
Population Genetics 



Population genetics is the study of the genetic structure of populations. 
Such a statement may look at first to be a truism, a tautology. The subject 
matter of our research becomes very intricate, however, as soon as we try to 
specify what we mean by the above definition. The terms "genetic structure" 
and "population" may have different meanings according to what we are 
willing to indicate by such words. It therefore seems convenient to start 
with an analysis of the terms we are using. Such discussion will give us a 
chance to see how the problem of heterosis is intimately connected with the 
general theme of population-genetical studies. A few experimental data will 
be used to illustrate such points. 

Let us consider first what we mean by population. If we take a dictionary 
definition, we find in Webster's that population is "all the people or in- 
habitants in a country or section." It means, in this sense, the sum of indi- 
viduals present at a certain moment over a more or less arbitrarily limited 
territory. But this definition does not correspond to the requirements of our 
studies, as I have tried to show elsewhere (Buzzati-Traverso, 1950). Such a 
definition is a static one, while the population, as considered in the field of 
population genetics, is a dynamic concept. We are interested not in the 
number of individuals present at a certain time in a certain place and their 
morphological and physiological characteristics. Instead, we are concerned 
with the underlying mechanisms which bring about such characteristics, and 
the particular size the population reaches at any particular moment. Since 
such mechanisms depend upon the numerical dynamics of the population 
and upon heredity, it follows that our concept of population is typically 
dynamic. On this view, then, a population is an array of interbreeding indi- 
viduals, continuous along the time coordinate. 

149 



150 ADRIANO A. BUZZATI-TRAVERSO 

Consideration of a population as a phenomenon continuously occurring in 
time makes it impossible for the experimental student of population genetics 
to get a direct and complete picture of what is occurring within a population 
at any particular moment. We can attempt to collect data on the population 
under study only by freezing such flowing processes at particular time in- 
tervals. Collecting observations on a population at different times gives us 
a chance to extrapolate the direction and rate of the processes that have 
occurred within the population during the time elapsed between two succes- 
sive sets of observations. If the samples studied are large enough and give an 
unbiased picture of the total population at the time when the sample is being 
drawn, this experimental procedure ma}^ give us a fairly adequate idea of 
what is going on within the array of interbreeding individuals continuous 
along the time coordinate. That sum of individuals at a definite time, which 
one usually means by population, is of interest to the population geneticist 
only as an index of the particular evolutionary stage reached by the array of 
interbreeding individuals. Since there are actual breeding and genetic rela- 
tionships between the individuals of any such array, of any such population, 
the population can be considered as the natural unit of our studies. 

If we consider now what we mean by "genetic structure," our task be- 
comes much more complex. At first we could assume that the genetic struc- 
ture of a population could be properly described in terms of the gene frequen- 
cies present at a certain time within a population. But this is only part of the 
picture. 

For the total description of the genetic structure of a population we have 
to consider not only the frequencies of existing genes, but how these are 
fitted within the chromosomes, how these allow the release of variability by 
means of recombinations, how large is the amount of new variability pro- 
duced by mutations, and several other factors which we cannot analyze now. 
In a few words, the study of population genetics aims at the knowledge of the 
breeding system of populations. This, as we shall see, is a rather difficult task 
because of the complexity of factors responsible for the origin and evolution 
of such systems. 

EVOLUTIONARY FACTORS INVOLVED 

When we take into consideration a species or a natural population at a 
certain stage, we have to assume that such a natural entity is the product of 
a series of evolutionary factors that have been at work in previous times and 
that some, or all of them, are still operating on the population while we are 
studying it. This means that we should try to explain the genetic structure of 
the population in terms of such evolutionary factors. 

Now, if we are willing to examine the nature of the known evolutionary 
agencies, we conclude that these can be classified into two types. On one side 
we find, in sexually reproducing organisms, a limited number of chromo- 



HETEROSIS IN POPULATION GENETICS 151 

somes, linkage between genes, sterility mechanisms, mating discriminations, 
devices favoring inbreeding, and other conservative forces that aim at the 
preservation of certain constellations of genes over a large number of genera- 
tions. On the other side we find mutation pressure, recombination between 
chromosomes, recombination among genes due to crossing over, outbreeding 
devices, migration pressure, and other revolutionary forces that aim at the 
production of genetic novelt3^ 

It seems reasonable to maintain that, at any particular time, a species or 
a natural population can be considered as a sort of compromise between the 
two conflicting forces — a compromise that is brought about through the 
action of natural selection. In other words, the fine adjustment or adaptation 
of a population to its environment is the expression of such compromise. At 
any particular time the terms of the compromise between the conflicting 
forces are always different as compared to other moments, as the compro- 
mise itself is a dynamic process. 

In order to reach the highest possible level of adaptation with respect to a 
certain set of environmental conditions, natural selection is discriminating not 
only for or against a certain individual genetic constitution, but for or against 
a group of individuals, as w-ell. Sometimes selection acts at the level of the in- 
dividual, sometimes it operates at some higher level. If we consider a genotype 
that insures resistance against an infectious disease, present in a certain area of 
distribution of a species, it wall be obvious that an individual carrying it shall 
directly benefit by it. But if we consider a genotype producing fecundity 
higher than the average of the population, this will be selected by the mere 
fact that a larger number of individuals having such genetic constitution will 
be present in the next generation. These, in their turn, shall have a chance of 
being represented in the next generation greater than that of individuals 
having a less fertile genotype. The individual itself, though, obtains no direct 
advantage from such selection. 

The next extreme condition we can consider is the one occurring when the 
advantage of the individual is in conflict with the advantage of the group. 
This is the case, for instance, of a genotype that would extend the span of 
life far beyond the period of sexual activity — or higher fertility linked with 
antisocial attitudes in the case of man. In both cases, natural selection favor- 
ing the preservation of the group will discriminate against the individual. A 
similar mechanism must have played a great role in various critical periods 
of organic evolution. When intergroup selective pressure is in the opposite 
direction from intragroup selection, a sort of compromise has to be reached 
between the two conflicting tendencies. This can be reached in many differ- 
ent ways that are best illustrated by the great variety of life histories and 
mating systems to be found in the living world. 

Those factors which we have classified as conservative tend to j)roduce 
genetic homogeneity, or what is technically known as homozygosis. Factors 



152 ADRIANO A. BUZZATI-TRAVERSO 

that we have named revolutionary tend to produce genetic heterogeneity or 
heterozygosis. Thus we come to the conclusion that the mentioned compro- 
mise brought about by selection consists of the pursuit of an optimum level 
of hybridity with respect to the conditions under which the organism lives. 
Such a hybridity optimum is the product, not only of the mutation rate and 
selective value of single genes, but also depends largely upon the genetic sys- 
tem and the mating system — the breeding system — of the considered species 
or population. 

The genetic structure of natural populations cannot be solved only in 
terms of individual variations observable in the group. Instead, it must be 
integrated into a unitary research on changes in gene frequencies as related 
to the underlying breeding systems. This is why we are justified in consider- 
ing the natural population as a unit, since individual variations must be 
referred to the genetic balance of the whole aggregate of individuals. 

What is that hybridity optimum I was speaking about but heterosis? How 
else could heterosis be defined in population problems other than that type 
and amount of heterozygosity that gives the population or the individual the 
best adaptive value with respect to the conditions in which the organism 
lives? With this view, then, it becomes feasible to analyze experimentally 
what morphological and physiological characteristics of the hybrids produce 
the better adaptation. 

MECHANISMS WHICH PROMOTE HYBRIDITY 

In studying how heterosis mechanisms are brought about in living crea- 
tures, we may attempt a sort of classification of the devices present in plants 
and animals insuring hybridity. Starting from the most complex and proceed- 
ing to the less complex cases, we can distinguish three types of mechanisms: 
(1) mating systems, (2) chromosome mechanisms, and (3) gene effects. 

We will not discuss in detail all the devices insuring hybridity found in 
plants and animals. We will mention a few, in order to show how many differ- 
ent paths have been followed in evolution to reach the same sort of results. 

Under the heading "mating systems" we may mention homo- and hetero- 
thally among fungi; monoecism and dioecism, incompatibility mechanisms, 
and heterostyly among flowering plants. Here, in some cases such as Primula 
scofica, there is close relation between the variability of ecological conditions, 
and, therefore, of selection pressure and the efficiency of the incompatibility 
mechanisms. Other species of this genus present in England are character- 
ized by heterostyly and incompatibility devices to insure the occurrence of 
outcrossing, apparently necessary to meet the requirements of varied eco- 
logical conditions. Primula scofica, living in a very specialized ecological 
niche, shows that such a mechanism has broken down. In fact, it looks as if 
the requirements of a constant environment are met better by populations 
genetically less diversified. 



HETEROSIS IN POPULATION GENETICS 153 

Among animals, the largest ])art of which are not sessile and therefore not 
bound to the ground, the differentiation into two sexes offers the best solution 
to the problem of insuring a wide range of crossing among different geno- 
types. But even here we see that special behavior patterns have been de- 
veloped for this purpose. These may be courtship relationships, sexual selec- 
tion, dominance relationships among a group of animals, or protandry 
mechanisms, where the presence of two sexes in hermaphrodites could reduce 
the amount of outcrossing and therefore endanger the survival of the species. 
Even among parthenogenetic animals, such as Cladoceran Crustacea, the ap- 
pearance of sexual generations after a long succession of asexual ones seems 
to depend upon extreme environmental conditions. For its survival, the 
species must shift over to sexual reproduction in order to obtain a wider 
range of genetic combinations, some of which might be able to survive under 
the new set of conditions. 

At the level of the chromosome mechanisms, several examples of perma- 
nent hybrids are known well enough to be sure that they play an im- 
portant role for the survival of some flowering plants. In animals, too, some 
similar mechanism may be present. In a European species of Drosophila 
which we are studying now^, Drosophila subobscura, one finds that practically 
every individual found in nature is heterozygous for one or more inversions. 
It looks as if the species were a permanent hybrid. 

Rarely, though, one finds individuals giving progeny wath homozygous 
gene arrangement. Such cases have been observed only three times: once 
in Sweden, once in Switzerland, once in Italy; and they are very peculiar 
in one respect. The three homozygous gene arrangements are the same, even 
though the ecological and climatic conditions of the three original popula- 
tions were as different as they could be. It looks as if the species could 
originate only one gene arrangement viable in homozygous condition, and 
that this may occur sporadically throughout its vast distribution range 
(Buzzati-Traverso, unpublished). 

At this level too is the fine example of heterozygous inversions from the 
classical studies of Dobzhansky (1943-1947). They have demonstrated that 
wild populations of Drosophila pseudoobscura show different frequencies of 
inversions at different altitudes or in the same locality at different times of 
the year. Variation in the frequency of inversions could be reproduced ex- 
perimentally in population cages by varying environmental factors such as 
temperature. It is shown in such a case that natural selection controls the in- 
crease or decrease of inversions determining an interesting tyj)e of balanced 
polymorphism. Finally, according to the investigations of Mather (1942- 
1943) on the mechanism of polygenic inheritance, it appears that linkage rela- 
tionships within one chromosome, even in the absence of heterozygous inver- 
sions, tend to maintain a balance of plus and minus loci controlling quantita- 
tive characters. 



154 ADRIANO A. BUZZATI-TRAVERSO 

We come then to the third level, that of gene effects. Here it is well known 
that heterozygotes for a certain locus sometimes show a higher viability or a 
better adaptation to the environment than either homozygote. The most 
extreme examples are those of the widespread occurrence of lethals in wild 
populations of Drosophila, noted in the next section. 

Every population of plants and animals that has been studied from 
the genetic viewpoint has proved to be heterozygous for several loci. We 
have now at our disposal a large series of data showing that the phenomenon 
of genetic polymorphism is frequent in plants, animals, and man. These offer 
to the student of evolutionary mechanisms the best opportunities to test his 
hypotheses concerning the relative importance of selection, mutation pres- 
sure, migration, and genetic drift as factors of evolution. Wherever we find 
a well established example of balanced polymorphism, such as that of blood 
groups and taste sensitivity in man, it seems safe to assume that this is due to 
selection in favor of the heterozygote. How this selection actually may pro- 
duce an increase in the chances of survival of the heterozygote, as compared 
to both homozygotes, is an open question. When the characters favored by 
natural or artificial selection are the result of several genes in heterozygous 
condition, the analysis becomes very difficult indeed, as the experience of 
plant and animal breeders clearly shows. 

EXPERIMENTS WITH HETEROSIS 

The importance of the problem of heterosis forpopulation-genetical 
studies is clearly shown, not only by such general considerations and by the 
few examples mentioned, but also by the everyday experience of people 
interested in such lines of work. I have come across problems involving 
heterosis several times and shall describe some results we have obtained 
which may be of interest for the problem under discussion, especially at the 
level of single gene differences. 

Several Drosophila workers have been able to show the occurrence of 
heterosis in the fruit flies. L'Heritier and Teissier (1933), Kalmus (1945), and 
Teissier (1947a, b) have shown that some visible recessive mutants of Dro- 
sophila melanogaster such as ehony and sepia have a higher selective value in 
heterozygous condition than either of the corresponding homozygotes under 
laboratory conditions. Dobzhansky and collaborators in Drosophila pseudo- 
obscura, Plough, Ives, and Child, as well as other American and Russian 
workers in Drosophila melanogaster, have shown that recessive lethals are 
widely spread in natural populations. It is generally accepted that such genes 
are being maintained in the population because the heterozygotes are being 
selected. Teissier (1942, 1944) has brought similar evidence under labora- 
tory conditions for Drosophila melanogaster. 

It has been shown in several populations of species of the genus Drosophila 
that heterozygous inversions are being selected, under natural and ex- 



HETEROSIS IN POPULATION GENETICS 155 

perimental conditions. It seems, however, that the study of selection in 
favor of the heterozygote for single loci deserves more careful analysis. The 
whole problem of heterosis for several genes affecting quantitative characters 
will be solved, I think, only when the more simple cases of heterosis where 
single gene differences are involved shall be cleared up. I have been lucky 
enough to come across some useful experimental material for the purpose. 

For a number of years I have kept about one hundred different wild 
stocks of Drosophila melanogaster coming from diflferent geographical locali- 
ties. Such stocks were maintained by the usual Drosophila technique of 
transferring about once a month some 30-40 flies from one old vial to a new 
one with fresh food. About twice a year I look at the flies under the micro- 
scope. Since all such stocks were wild type, no change by contamination was 
expected, as these stocks were phenotypically alike. Contamination by mu- 
tants kept in the laboratory could not have produced any appreciable result, 
owing to the well known fact that both under laboratory and natural condi- 
tions mutants are generally less viable than the normal type. To my sur- 
prise, however, I happened to observe at two different times, in two different 
wild stocks, that a fairly large number of the flies showed an eye color much 
lighter than the normal. These two mutants proved to be indistinguishable 
recessive alleles at the same locus in the third chromosome. The presence 
of the homozygotes has been checked at different times over a number of 
years. 

In the summer of 1947 while collecting flies in the wild for other purposes, 
I found in the neighborhood of Suna, a small village on the western shore of 
the Lake Maggiore, in Northern Italy, several individuals of both sexes show- 
ing the same eye color. From these a homozygous stock for such mutant was 
obtained. Crossing tests proved that it was another allele of the same locus as 
the above mentioned. The occurrence of several individuals mutant for an 
autosomal recessive within a free living population was remarkable enough. 
But finding that the same gene was concerned as in the laboratory stocks, I 
suspected that such a mutant might have a positive selective value, both 
under laboratory and natural conditions. 

I began an experiment to check this point. Two populations in numerical 
equilibrium were started, applying the method previously used by Pearl for 
the study of population dynamics of Drosophila, described in detail else- 
where (1947a). Sixteen light-eyed individuals, eight males and eight females, 
were put together in one vial with sixteen wild type flies. The gene frequency 
at the beginning of the experiment was therefore .5. Under the experimental 
conditions the population reached an equilibrium in respect to the amount 
of available food at about 700-900 flies per vial. After about twenty genera- 
tions, assuming that each generation takes 15 days, the frequency of recessive 
homozygotes was about 40 per cent. Assuming random mating within the 
population, taking the square root of .40 one gets a gene frequency for the 



156 ADRIANO A. BUZZATI-TRAVERSO 

light-eyed gene of about .63. Since in both parallel populations the gene fre- 
quency was similar, one could conclude that selection had favored the mutant 
type, shifting its frequency from .5 to .63 in the course of about twenty 
generations. 

Such an experiment did prove that the mutant gene had a positive selec- 
tive value. It was impossible to know whether in the long run it would have 
eventually eliminated its normal allele from the population. At this stage, I 

0.875^ 



Fig. 9.1 — Variation in the frequency of the Hght-eyed gene in selection experiments. In the 

abscissae is the number of generations, in the ordinates the gene frequency. Each line 

represents a single experiment on an artificial population. 

have begun a new experiment along the same lines, but with different gene 
frequencies to start with. Two populations were started with 2 males and 2 
females of the mutant type, plus 14 males and 14 females of the normal type. 
Two populations were started with 16 mutant and 16 wild flies, and two 
populations with 28 mutant and 4 wild type flies. 

I had, therefore, at the beginning of the experiment six populations. Two 
had a gene frequency of the light-eyed mutant approximately equal to .125. 
Two had a gene frequency of .5, and two had a gene frequency of .875. 
Figure 9.1 shows the result of such an experiment after about fifteen genera- 
tions. Crossings of wild type males, taken from the populations, with homo- 
zygous recessive females showed that there was no significant departure from 
random mating within the population. The gene frequencies indicated on the 



HETEROSIS IN POPULATION GENETICS 



157 



orclinates were obtained by taking the square root of the observed frequencies 
of homozygous recessives. 

The following conclusions can be drawn: (1) the three experimental popu- 
lations, each being run in duplicate, have reached the same gene frequency 
at about the .579 point; (2) natural selection has been acting on the three 
populations producing the same end results, irrespective of the initial gene 
frequency; (3) natural selection has been acting in favor of the heterozygous 
flies; and (4) the homozygous mutant seems to be slightly superior in its 
survival value to the homozygous normal allele. 

It was of considerable interest to determine whether the intensity of selec- 
tion operating in the three experiments was the same. Since the three experi- 
mental curves (each being the mean of the two duplicate populations) could 
not be compared directly. Dr. L. L. Cavalli elaborated a mathematical 
analysis of the problem (Cavalli, 1950). The function of gene frequency linear 
with time F, when the heterozygote is at an advantage, is given by: 

Y = Qe^Og p-\-pe\0g q -\0g[pe-p] , 

where p and q are the gene frequencies at the beginning of the experiment in 
a random breeding population, and pe and qe are the equilibrium frequencies. 
By means of this function it is possible to transform the experimental curves 
to linear ones. Results can then be plotted graphically for the three experi- 
ments. Fitting straight lines with the method of maximum likelihood, one 
obtains the following values for the constants of the linear regression equa- 
tion: 



Experi- 
ment 


Initial Gene 
Frequency 
(Observed) 


Slope 


Position 


Initial Gene 

Frequency 

(Theoretical) 


1 

2 

3 


.500 
.125 
.875 


.0879 
.0631 
.0726 


+ 1.21 
- .41 

+ .27 


.425 
.100 
.830 



The position is the transformed value of the initial gene frequency which 
is given in the last column, and is in good agreement with the experimental 
value. If one tests the parallelism of the three regression lines so obtained, one 
gets a chi square of 4.0 with two degrees of freedom. Parallelism therefore 
seems to be satisfactory. This implies that the intensity of selection is inde- 
pendent of initial conditions. 

If we take these results together with the two independent occurrences of 
the same mutant gene in different genotypical milieus, it seems safe to main- 
tain that such a gene has a positive selective value with respect to its normal 
allele, and that selection is acting mainly through a typical heterosis mecha- 
nism. It is worth while to stress that this gene was found both in natural and 



158 ADRIANO A. BUZZATI-TRAVERSO 

experimental conditions. The exceptional occurrence of many mutant indi- 
viduals in a free living population can be accounted for by assuming that 
they have a higher selective value. 

BASIS FOR SUPERIORITY OF THE HETEROZYGOTE 

It would be interesting to try to find out how selection discriminates 
against both normal and mutant homozygotes. I am just beginning to attack 
this problem. 

Dr. E. Caspar! has some interesting results on a similar problem, and I 
wish to thank him for permission to quote them (Caspari, 1950). In free 
living populations of the moth Ephestia kuhniella, this author has observed 
a balanced polymorphism, whereby individuals having brown colored and 
red colored testes occur in various numbers. The character brown behaves as 
a complete dominant with respect to red. The polymorphism seems to be 
determined by a higher selective value of the heterozygote. It has been pos- 
sible to show that the heterozygote is equal or superior to the homozygous 
recessive and the latter is superior to the homozygous dominant with respect 
to viability. It was found that, while the heterozygote is equal or superior to 
the homozygous dominant, the homozygous brown is superior to the homozy- 
gous red with respect to mating activity. The dominance relationships of such 
two physiological characters are therefore reversed. 

There is no decisive evidence for heterosis for any of the characters 
studied. The recessive for the testis color acts as dominant with respect to 
viability, and the dominant testis color acts as dominant with respect to 
mating behavior. The net result is a selective advantage of the heterozygote 
that can account for the observed polymorphism. This seems a good ex- 
ample of how a heterosis mechanism can be determined by the behavior of 
two visible alleles in heterozygous condition. It is hoped that similar analyses 
will be developed for other cases of balanced polymorphism. 

The search for clear-cut examples of heterosis depending on single genes 
seems to me the most promising line of attack on the general problem under 
discussion. If I could find another gene behaving in a way similar to the one 
I have studied in Drosophila melanogaster, and could study the interaction of 
the two, it would be possible to go a step further in the analysis of heterosis 
mechanisms. The evidence derived from such single genes being favored in 
heterozygous condition is likely to be very useful in more complex condi- 
tions where the action of several genes is involved. 

When we come to consider the selective advantage of polygenic charac- 
ters, even in such an easy experimental object as Drosophila, the problem 
becomes very entangled indeed. In recent years I have been studying, for 
example, a number of quantitative characters being favored by natural 
selection in artificial populations in numerical equilibrium, such as the ones 
I have been speaking about. I have set in competition at the beginning of one 



HETEROSIS IN POPULATION GENETICS 159 

experiment two stocks differing for visible mutants. One stock was while- 
and Bar-eyed, the other stock was normal for both characters. The two 
stocks differed, too, in a number of quantitative characters such as fecundity, 
fertility, rate of development, longevity, and size. 

After about thirty generations the two mutant genes had been wiped out. 
This could have been expected on the basis of previous data of L'Heritier and 
Teissier on the elimination of such genes in artificial populations. At that 
time, however, I did not discard the populations, but kept them going for 
some seventy more generations. All the individuals present in the popula- 
tions were phenotypically normal. But testing from time to time the values 
of the above mentioned characters, I could establish that natural selection 
was continuously operating and favoring higher fecundity, higher fertility, 
higher longevity, and quicker developmental rate throughout the four years 
that the experiment lasted. At the end, the flies present in the population 
were superior by a factor of more than six to the mean of the considered 
characters in the two original parental stocks. When I measured such values 
in the Fi hybrids between the two stocks I could observe values higher than 
those obtained after more than one hundred generations of selection. 

The selection experiment could then be interpreted in two different ways. 
Either (a) selection had picked up a new genotype made out of a new com- 
bination of polygenes derived from the two parental stocks, or (b) selection 
had just preserved by means of a heterosis mechanism a certain amount of 
heterozygosity, which was at its highest value at the beginning of the experi- 
ment. The fact that in the course of the experiment the factors had been 
steadily improving seemed to be against hypothesis b, but I could not be sure 
that was the case. 

I then set up a new selection experiment, whereby I put in competition the 
original stock white Bar with the normal type derived from the population 
which had been subjected to natural selection for more than one hundred 
generations. The result was clear. The genes white and Bar were elimi- 
nated in this second experiment at a much higher rate than in the first ex- 
periment. In the first experiment the gene frequency of the gene Bar after 
ten generations had dropped from .50 to .15. In the second experiment, after 
as many generations, the Bar gene frequency had dropped from .50 to .03. 
It seems that the genotype produced by a hundred generations of natural 
selection under constant conditions was so much better adapted to its en- 
vironment that it could get rid of the competing genes with much greater 
ease than the original wild type flies. But could it not be that all or at least 
part of this result could be accounted for by the action of some heterosis 
effect? 

Another example of a similarly puzzling condition is an experiment on 
artificial populations under way now in my laboratory. I would like to find 
out whether it is possible to produce so-called small mutations or polygenic 



160 ADRIANO A. BUZZATI-TRAVERSO 

mutations with X-rays, and whether an increase in the mutation rate may 
speed up the evolutionary rate under selection pressure. 

For this purpose I have set up four artificial populations starting from an 
isogenic stock of Drosophila melanogaster. One of these is being kept as con- 
trol while the other three get, every two weeks, 500, 1000, and 2000 r-units 
respectively. At the start, and at various intervals, I am measuring fecun- 
dity, fertility, and longevity of the flies. The few data so far collected show 
clearly that in the irradiated populations the percentage of eggs that do not 
develop is much higher than in the control. This is due to the effect of 
dominant and recessive lethals. But the startling result is that the fecundity, 
measured by the number of eggs laid per day by single females of the irradi- 
ated populations, is higher than in the control series. Probably X-rays have 
produced a number of mutations for higher fecundity which have been ac- 
cumulated by natural selection in the course of the experiment. But, are spe- 
cific mutations for higher fecundity being produced, or am I dealing with 
heterosis phenomena dependent upon nonspecific mutants? 

These few examples from my own experience with population-genet ical 
studies show, I think, how important the heterosis phenomenon can be in our 
field of work. Both in natural and artificial populations, heterosis seems to 
be at work, making our analysis rather difficult, but stimulating as well. 
Closer contacts between students of selection and heterosis in plant and 
animal breeding and students of evolutionary problems are to be wished. 
Let us hope that a higher level of hybridization between various lines of 
investigation might become permanent, since it surely will make our studies 
more vigorous and better adapted to the requirements of a rapidly growing 
science. 



HAROLD H. SMITH 

Cornell Universify 



Chapter 10 

Fixing Tronsgressive Vigor 
in Nicofiona Rustica 



Hybrid vigor has been observed to varying degrees among certain inter- 
varietal hybrids of the self-pollinated cultivated species Nicotiana rustica L. 
(Bolsunow, 1944; East, 1921). In experiments undertaken to obtain a larger 
;V. rustica plant giving increased yield of nicotine, it was reported (Smith 
and Bacon, 1941) that inbred lines derived as selections from hybrids among 
three varieties exceeded the parents and Fi's in plant height, number of 
leaves, or size of the largest leaf. 

The general experience of breeders of self-pollinated plants has been that 
improved varieties can be developed through hybridization followed by selec- 
tion and inbreeding, to fix desirable transgressive characteristics. Yet it is 
difficult to find data from which quantitative relationships of parents, Fi, and 
transgressive inbred can be adequately evaluated; as from replicated and 
randomized experiments in which the generations have been grown at the 
same time under comparable conditions. In view of the increasing number 
of reports on hybrid vigor in self-pollinated crop plants and its suggested 
utilization (Ash ton, 1946), it was considered opportune to present relevant 
data accumulated on N. rustica. 

Since methods of partitioning phenotypic variance have become generally 
available there was additional interest in making further study of the iV. 
rustica material. Breeding results obtained in advanced selections could be 
related to the heritability estimated from data on early generations. 

MATERIALS AND METHODS 

Four varieties of Nicotiana rustica were used in these experiments. Three 

of them — hrasilia strain 34753, Olson 68, and tall type have been described 

* Published as Paper No. 261 , Department of Plant Breeding, Cornell University, Ithaca 
New York. 

161 



162 HAROLD H. SMITH 

in Smith and Bacon (1941). The fourth was received originally from the 
director of the Tabak-Forschungsinstitut, Baden, Germany, under the name 
of texana, a designation which we have retained. It is a small, early-maturing 
type. The four parental varieties were of highly inbred stocks maintained by 
the Division of Tobacco, Medicinal and Special Crops of the United States 
Department of Agriculture. The earlier part of the breeding program was 
carried out while the writer was associated with this organization. 

The advanced selection, designated Al, used in these experiments has a 
complex genetic history of crossing, backcrossing, and inbreeding. This can 
be briefly summarized by stating that its ultimate composition was, on an 
average, 60 per cent 34753, 22 per cent Olson, 12 per cent tall type, and 6 
per cent texana. About 82 per cent of the Al genotype was, on chance alone, 
contributed by the two most vigorous parents, 34753 and Olson 68. This 
calculation does not take into account any differential effect of selection on 
changing the frequency of genes introduced from diverse parental origins. 
Observation of the Al phenotype led us to believe that selection had further 
increased the proportion of genes from the two most vigorous parents. 

In 1947 the four parents, the six possible F/s, the three double crosses, 
and the F4 generation (preceded by three generations of inbreeding) of line 
Al were grown in a randomized complete block design with fifteen plants in 
each plot and replicated six times. In 1949 the two most vigorous varieties 
(Olson 68 and 34753), the Fi, F2, backcrosses of the Fi to each of its parents, 
and the Fe generation of line Al were grown in a randomized complete block 
design with twenty plants in each plot and replicated eight times. 

Measurements were made on plant height, number of leaves or nodes, and 
length of the largest leaf. In addition, data were taken on the width of the 
largest leaf, number of days from planting to appearance of the first flower, 
and total green weight of individual plants. 

Typical plants of Olson 68, 34753, the Fi between these two varieties, and 
selection Al are illustrated in Figure 10.1. 

EXPERIMENTAL RESULTS 

Data obtained from the 1947 and 1949 plantings are summarized in 
Tables 10.1 and 10.2, respectively. 

Phenotype-Genotype Relations 

Preceding further biometrical analysis of the data, tests for evidence of 
differential environmental effects and genetic interactions were made. For 
the former, the relation between genotype mean and non-heritable variabil- 
ity was determined by comparing means and variances of the parents and Fj 
(1949 data. Table 10.2). For the characters plant height and leaf length, the 
variances were unrelated to the means and the parental variances were not 
significantly different^from each other. For node number, however, the 




Fig. 10.1 — Typical field-grown (1949) ]ilants of Nicotiana riistica. Left to right: Olson 68, 
brasilia strain 34753, Fi Olson 68 X 34753, and selection A^Fe). The scale shown at the 

left is in inches. 



TABLE 10.1 

PLANT CHARACTERISTICS IN PARENTAL VARIETIES, HYBRIDS, 
AND AN INBRED SELECTION OF NICOTIANA RUSTICA* 





Type 


Plant 
Height 


Leaf 
Number 


Leaf 
Length 


Mean 
Leaf 
Width 


Mean 
Days 

TO 

Ma- 
ture 




Gen- 
era- 
tion 


Mean 


Total 

within 
Plot 


Mean 


Total 

within 

Plot 


Mean 


Total 

within 

Plot 


Mean 
Green 

Wgt. 

Per 
Plant 




d.f. 


Var. 


d.f. 


Var. 


d.f. 


Var. 




Pi 


Olson 68 (/4 ) 


in. 

49.9 

29.0 

46.7 

33.6 


73 
84 
83 
83 


17.8 
20.0 
30.3 
11.1 


18 8 
15.8 
16 
12.7 


72 
82 
83 
83 


4.68 
7.96 
1.36 
1.04 


in. 

11.7 
8.7 
6 2 
6 6 


72 
82 
83 
83 


1.07 
1.43 
1.24 
1.01 


in. 

8.7 
7.5 
5.5 
5.4 


70.1 
66.6 
48.0 
40.9 


lbs. 
1 51 




34753 (B) 


89 




Tall type (C) 


SI 




Texana (D) 


48 




Average 

01sonX34753UX5). . 

OlsonXtall(^XC) 

Olson Xtexana {A XD) . 

34753 Xtall(BXC) 

34753Xtexana(SX£>). 
Tall Xtexana (CXC) . . 

Average 

iAXB)X(CXD) 

{AXC)X(BXD} 

{AXD)XiBXC) 

Average 






39.8 




19.8 


15.8 




3.76 


8.3 




1.19 


6.8 


56.4 


0.85 


Fi 


48.5 
42.9 
40.1 
47.1 
40.3 
44.2 


74 
80 
81 
84 
83 
84 


58.3 
25.4 
20 8 
45.2 
28.1 
29.3 


16.8 
13 
11.2 
16 6 

14.4 
15.5 


74 
80 
81 
84 
83 
83 


7.18 
7.74 
4 99 
1.77 
6 34 
1.49 


10 8 

10.1 

10.6 

7.8 

8 7 

7.7 


74 
80 
81 
84 
83 
83 


1.92 
1.82 
1.49 
1.37 
2.30 
1.88 


8.7 
9.6 
9.7 
6.7 
7.4 
7.0 


75.0 
65.1 
70.6 
50.5 
60.0 
51.8 


1.47 
1.16 
1.13 
0.76 
0.93 
0.95 




43.8 




34.5 


14.6 




4.92 


9.3 




1.80 


8.2 


62.2 


1.07 


FiXFi 


41.9 
39 6 

42.5 


75 
82 
81 


25.1 
61.1 
39.9 


14.0 
12.4 
13.9 


74 
79 
80 


10.60 

9.29 

10 80 


8 8 

9 8 
9.7 


74 
79 
80 


2.36 
3.11 
2.50 


7.7 
8.8 
8.7 


61.9 
66.0 
60.4 


0.86 
0.99 
1.06 




41.3 




42.0 


13.4 




10.23 


9.4 




2.66 


8.4 


62.8 


97 




Selection Al 




F4 


54.9 


78 


78.8 


19.9 


77 


S.28 


10.4 


77 


1.51 


8.0 


79.6 


1 83 








Least si 
5%le 
l%le 


?nificant diff. at 

vel 


2.68 
3.56 






1.22 
1.62 






0.89 
1.19 






0.76 
1.00 


4.11 
5.46 


25 


vel 














34 



















* Summary of 1947 data. 



164 



HAROLD H. SMITH 



means and non-heritable variances were linearly related for both 1947 and 
1949 data, and the parental variances were significantly different. 

Tests to reveal the presence or absence of non-allelic interactions were then 
made according to the method proposed by Mather (1949). Results are 
shown in Table 10.3. No significant deviations from zero were found if the 
level of significance was taken as P ^ .01. In each test, however, the P values 
for number of nodes were less than for plant height or leaf length, possibly 
owing to non-additive gene effects. 

It was concluded, on the basis of these tests, that for the two characters 



TABLE 10.2 

PLANT CHARACTERS IN THE TWO MOST VIGOROUS VARIETIES OF 

N. RUSTIC A, THEIR F,, F2, AND FIRST BACKCROSS PROGENY 

AND IN SELECTION AKFe)* 



Gen- 
era- 
tion 



Pi 
P2 
Fi 
F2 
Bi 
B2 
Fe 



Type 



Olson 68 

34753 

Olson 68X34753... 
(Olson X34753) self. 

FiXOlson 68 

FiX34753 

Selection Al 



Plant Height 



Mean 



in. 
47.8 
28.7 
43.2 
40.6 
47.3 
36.2 
55.6 



Total with- 
in Plot 



d.f. 



141 
143 
140 
149 
149 
148 
133 



Var. 



15.46 
22.63 
39.18 
99.19 
40.28 
101.50 
29.77 



No. OF Nodes 



Mean 



24. 
21. 

22. 
23. 
24. 
21. 



31.0 



Total with- 
in Plot 



d.f. 



136 
106 
110 
119 
138 
117 
126 



Var. 



3. 
10. 

8. 



45 
10 
60 



10.52 
10.49 



45 
44 



Leaf Length 



Mean 



in. 
11 
10. 
11 
11 
11 
10.8 
12.0 



Total with- 
in Plot 



d.f. 



142 
127 
131 
142 
144 
135 
141 



Var. 



0.68 
0.81 
0.63 



08 
10 



0.95 
0.69 



Least significant diff. at 

5% level 2.55 

1% level 3.42 



1.37 
1.83 



0.49 
0.66 



* Summary of 1949 data. 

TABLE 10.3 
SCALING TESTS FOR AVERAGE ADDITIVENESS OF GENE EFFECTS* 



Character 



Test A 



Dev. 



Var. 



Plant height. 3.6 2.86 
No. nodes...' 2.8 1.50 
Leaf length.. 0.9 0.47 




Test B 



Dev. 



Var. 




Test C 



Dev. 



Var. 




2. 13. 03-. 04 0.54.13 0. 25'. 80-. 81 -0.515.48 0. 13 .89-. 90 
2. 30. 02-. 03-1. 11. 71 0.84. 40-. 41 3.5 6.18 1.40. 16-. 17 
1.32,. 18-. 19, 0.00.47 0.00 1.00 i 0.5 1.77 0.38 . 70-. 71 



* Based on 1949 means. 



FIXING TRANSGRESSIVE VIGOR IN NICOTIANA RUSTICA 165 

plant height and leaf length, the data, as taken, could be used without 
serious error in partitioning the variance of segregating generations. For 
node number it was indicated that some correction should be made with 
the data before further biometrical analysis was undertaken. 

Mather suggested that difficulties of the sort encountered in these data 
with node number may be overcome by finding a transformation of scale on 
which they would be minimized. The transformations \/X, X'^, X^, and 
\/a -\- bx on the individual measurements were made. In the latter transfor- 
mation b is the linear regression coefficient and a the intercept. Also, for 
^ya + bx — K, \/—K was taken as —\/K. In some cases the transforma- 
tions reduced the departure from the preferred relationship in one test, only 
to make the transformed data less preferable by another test. No transforma- 
tion tried resulted in a consistent improvement over the original scale, and 
consequently none was used. 

It is evident that the significantly different variances in node number of 
the two parental types were due mainly to different interactions between 
genotype and environment. From previous experience we know that under 
ideal conditions of growth, Olson 68 and strain 34753 show approximately the 
same variability. The adverse weather conditions of the 1949 season were ob- 
served to have a more deleterious effect on leaf number in strain 34753. Con- 
sequently it was considered that the greater variability of this variety, com- 
pared to Olson 68, could be attributed to a greater phenotypic interaction 
between genotype and environment. In view of these relationships, the 
analysis of the data on node number was approached in another way, as 
mentioned below under "Partitioning Phenotypic Variance." 

First Generation Hybrids 

Deviations of the Fi means from mid-parent values (arithmetical average 
between parental means) can be used to estimate the preponderance of 
dominant gene effects, acting in one direction, at loci by which the parental 
complements differ. Mid-parent values were calculated from the 1947 da- 
ta on the four original varieties. The results for each line are summarized in 
Table 10.4. The data shown were obtained by first calculating the difference 
between the Fi mean and the mid-parent (Fi — MP) for each cross, then tak- 
ing the average of the differences for each group of three F/s involving the 
parent variety under consideration. The ratio of the deviation of the Fi from 
the mid-parent to half the parental difference, Fi — MP/\{P2 — Pi) , is a meas- 
ure of the relative potence (Mather, 1949; Wigan, 1944) of the gene sets. Po- 
tence ratios, calculated from averages, are shown in Table 10.4. For plant 
height and leaf length the Fi means fall, on an average, about .6 of the dis- 
tance from the mid-parent toward the larger parent. For leaf number the Fi 
means fall, on an average, about .7 of the distance from the mid-parent 
toward the smaller parent. 



166 



HAROLD H. SMITH 



The Fi's were taller and had larger leaves, on an average, than the mid- 
parent. It was concluded, therefore, that a preponderance of dommant+ 
genes was mvolved in determining differences in plant height and leaf length. 
In the development of the parent varieties, selection resulted in the accumu- 
lation of dominant-f modifiers, as is usually the case in naturally cross- 
pollinated plants. 

The result with the character leaf number was different in that the Fi had 
fewer leaves, on an average, than the mid-parent. Evidently, in the evolution 
of the varietal gene sets, there had been accumulated a preponderance of 
recessive+ modifiers (or dominant genes ior fewer leaves) at the loci by which 

TABLE 10.4 

DIFFERENCE BETWEEN THE Fi AND MID-PARENT (Fi - MP) AND THE 
POTENCEt RATIO IN INTERVARIETAL HYBRIDS. BOTH V.ALUES ARE 
EXPRESSED AS THE AVERAGE FOR EACH VARIETY IN CROSSES WITH 
THE OTHER THREE VARIETIES* 



Variety 


PL.4NT Height 


No. Leaves 


Leaf Length 




Fi-MP 


Potencef 


Fi-MP 


Potence 


Fi-MP 


Potence 


Olson 68 

34753 


in. 

+0.7 
+9.1 

+2.6 
+3.8 
+4.0 


+0.10 

+ 1.26 
+0.46 
+0.68 
+0.62 


-3.2 
+0.1 
-0.9 
-1.1 
-1.2 


-1.62 
+0.09 
-0.87 
-0.53 
-0.73 


in. 

+0.9 
+0.3 
+0.8 
+ 1.1 
+0.8 


+0.43 
+0.33 


Tall . . . 


+0.63 


Texana 


+0.97 


Average .... 


+0.59 



* 1947 data. _ _ _ 

t Potence = Fi-MP/kiPr-Pi). 

the parents differed. There can be little doubt that selection for mafty leaves 
was practiced in producing the parent types. This is especially true for Olson 
68 which was developed from hybrid origin by the late Mr. Otto Olson 
(Smith and Bacon, 1941) by selection for plants yielding large amounts of 
nicotine. In crosses with Olson 68, the Fi was consistently below the mid- 
parent. This result, interpretable as due to an accumulation of a preponder- 
ance of recessive genes for the character favored by selection, might be ex- 
pected occasionally in naturally self-pollinated plants. Dominance is of less 
importance here than in cross-pollinated organisms, since selection is largely 
a matter of sorting out superior homozygous combinations. 

The 1949 results (Table 10.2) on Olson 68 X 34753 were consistent with 
those of 1947 discussed above. 

Double Crosses 

The three possible double crosses involving all six Fi hybrids of four varie- 
ties were grown in 1947 in order to obtain evidence on genie interactions by 
comparing experimental results with predicted values. The latter were made 



FIXING TRANSGRESSIVE VIGOR IN NICOTIANA RUSTICA 



167 



in the manner employed in corn breeding, namely Jenkins' method, in which 
the average of the four Fi's not contributing to the double cross was used. 
These comparisons are shown in Table 10.5 for the three plant characters 
studied. The differences between observed means and j)redicted values in the 
nine comparisons made were all within the limits required for odds of 19: 1. 
It was concluded that the double cross means for plant height, number of 
leaves, and leaf length in iV. rustica could be predicted with a high degree of 
precision by Jenkins' method. The results indicated that there were no 

TABLE 10.5 
COMPARISON BETWEEN PREDICTED AND OBSERVED VALUES FOR PLANT 
HEIGHT, NUMBER OF LEAVES, AND LEAF LENGTH IN THREE DOU- 
BLE CROSSES INVOLVING FOUR VARIETIES OF N. RUSTICA 



Double Cross 


Observed 


Predicted 


Difference, 
Obs.-Pred. 


Plant height {in.) : 

{AXB)X{CXD) 

{AXC)X{BXD) 

{AXD)X{BXC) 


41.9 + 2.68 
39.6 + 2.68 
42.5±2.68 


42.6+1.34 
45.0 + 1.34 
44.0±1.34 


-0.7+3.00 

-5.4 + 3.00 
-1.5±3.00 


Average 


41.3 


43.8 


-2.5 


No. leaves: 

{AXB)X{CXD) 

{AXC)X{BXD) 

{AXD)X{BXC) 


14.0+1.22 
12.4±1.22 
13.9+1.22 


13.8 + 0.61 
15.0 + 0.61 

14.9 + 0.61 


+0.2 + 1.36 
-2.6 + 1.36 
-1.0 + 1.36 


Average 


13.4 


14.6 


-1.2 


Leaf length {in.) : 

{AXB)X{CXD) 

{AXC)X{BXD) 

{AXD)X{BXC) 


8.8 + 0.89 
9.8 + 0.89 
9.7 + 0.89 


9.3 + 0.44 

9.2 + 0.44 

9.3 + 0.44 


-0.5 + 0.99 
+0.6 + 0.99 
+0.4±0.99 


Average 


9.4 


9.3 


+0.1 



A , B, C, D represent the parent varieties as shown in Table 10.1. 



marked interactions between the genes or gene sets from the four varieties 
when combined in a variety of associations. 

To illustrate this point, let us assume that each parent is homozygous for 
a different allele at each of two independent loci so that A = XX YY, B = 
X'X'Fip, C = X^XWW\ and D = X^XWW^. The Fi's represent six dif- 
ferent combinations of these alleles. Each double cross contains all four alleles 
of each locus in four particular combinations. For example, the i)opulation 
{AXB)XiCX D) is 1/4A'X2 _^ \/^xX' + \/^X^X- + \/AX'XHoi the X 
locus and 1/4FF2 + 1/477^ + 1/47^72 + \/4:Y^YHor the Y locus. Sixteen 
different combinations of alleles at the two loci are possible in this double 
cross. Accurate prediction of the double cross value on the basis of only four 
of these combinations, namely: Fi's AXC, AXD, BXC, and B X D, 



168 



HAROLD H. SMITH 



indicates that the other 12 possible combinations do not introduce any sig- 
nificant non-additive effects. 

Another indication that epistatic effects were unimportant in the in- 
heritance of plant height, leaf number, and leaf length was afforded by the 
evidence that the means of the double crosses did not differ significantly from 
each other (Table 10.1). 

The average variance of the double crosses was greater than that of the 
parents or Fi's (Table 10.1), as would be expected from segregation. 

Partitioning Phenotypic Variance, Heritability, and 
Number of Effective Factors 

Estimates of the magnitude of the non-heritable variation (0%), in popu- 
lations involving Olson 68 and 34753 (1949 data), were obtained by taking 

TABLE 10.6 

ESTIMATES OF COMPONENTS OF \\\RIABILITY, NUMBER OF EFFECTIVE 
FACTORS (Ai), HERITABILITY, AND GAIN FOR PLANT HEIGHT, LEAF 
LENGTH, AND NUMBER OF NODES IN THE X. RUSTICA CROSS OLSON 
68 X BRASILIA, STRAIN 34753* 



Character 








"l 


A'l 


Herit- 
ability 
Per 
Cent 


Gain 


Plant height 

Leaf length 

Node number. . . 


25.76 + 15.3 
0.71+ 0.45 
7.38+ 4.38 


67.32 + 53.5 
1.04+ 1.05 
8.16 + 13.00 


113.20 + 71.3 
0.22+ 0.69 
2.20+ 8.11 


0.81 
1.38 
0.83 


54.9 
11.2 
12.4 


1.74 
0.91 

2.42 



* 1949 data. 



an average of the total within plot variance of the non-segregating families — 
Pi, P2, and Fi. As shown in Table 10.6, the values obtained were 25.76 for 
plant height, 0.71 for leaf length, and 7.38 for number of nodes. 

The following symbols are used for the components of heritable variance 
(total phenotypic minus environmental) : alb = variance depending on addi- 
tive gene effects, a|) = variance depending on dominance. The heritable 
variance of the F2 was calculated and equated to: 1/2o-g + l/4o-|). The 
pooled heritable variance of the two first backcrosses was equated to 
l/2crG + l/2o-z). Solving for 0%, the values obtained were 67.32 for plant 
height, 1.04 for leaf length, and 8.16 for number of nodes. Values for 0%, 
as calculated by substitution, were 113.20 for plant height, 0.22 for leaf 
length, and 2.20 for number of nodes. 

In view of the influence on node number of a differential interaction of the 
two parental genotypes with environment, an additional way of approach- 
ing an analysis of the data on this character was tried. If a simple relation 



FIXING TRANSGRESSIVE VIGOR IN NICOTIANA RUSTICA 169 

between the environmental variances of the Pi, P2, and Fi is assumed, so 
that a% of the Fi = 1/2((t% of Pi + al of P2), then a% of the Fi = 6.78. 
The environmental variance of Bi may then be equated to 1/2 (variance of 
Pi+ variance of Fi), which is 5.12. By a similar relation, the environmental 
variance of B2 is equal to 8.44. The pooled heritable variance of Bi + B2, i.e., 
l/2ah + 1/2(7?,, may be equated to: (W.49 - 5.12) + (9.45 - 8.44). This 
gave 6.38. The heritable variance of the F2, i.e., l/2ao + l/4o-i), may be 
equated to (10.52 — 6.78). This gave 3.74. Solving: <t% = 10.56 and <to = 
2.20. The former, (xjj, has a somewhat larger value than that obtained by 
the original analysis (8.16, Table 10. 6j; the latter, cc, is the same. 

Heritability of a character was estimated as the ratio, expressed in per 
cent, of the variance component due to additive, fixable gene effects (aa) to 
the sum, <r% -\- a]) -\- a%. Heritability of plant height was calculated to 
be 54.9 per cent, of leaf length 11.5 per cent, and of node number 12.4 per 
cent. 

Estimates of the number of effective factors (Ki) were made on the as- 
sumptions inherent in the equation A'l = (Pi — P^^l^a'a. The values ob- 
tained (Table 10. 6 j were 0.81 for plant height, 1.38 for leaf length, and 0.83 
for number of nodes. These estimates were undoubtedly too low, due in part 
to non-isodirectional distributions of + and — genes in the parents. Ex- 
perimental evidence of non-isodirectional distribution was afforded by the 
fixing of transgressive characteristics in inbred selections following hybridiza- 
tion between varieties. Some -\- genes were contributed by each parent, and 
consequently could not have been concentrated in one. Linkage in coupling 
phase and/or differences in magnitude of effect of the individual genes or 
gene blocks might also have contributed to the low estimates of the number 
of effective factors. 

In the absence of data on Fa's, biparental progenies, and double back- 
crosses (Mather, 1949), the errors of the estimates of 0%, a\,, and 0% for each 
character were computed as follows. From the eight replications, four means 
were calculated by grouping replications 1 and 2, 3 and 4, 5 and 6, and 7 
and 8. The standard error of the four independent means was then obtained 
(Table 10.6). These errors are maximum estimates since there was a pro- 
nounced gradient of environmental effects from replication 1 to replication 8. 

Mather (1949) is in the process of making an extensive biometrical genetic 
analysis of plant height in a Xicotiana rustica cross, and it was of interest to 
compare his published results with corresponding statistics presented in this 
study. From his data so far reported, the average values (mean of 1946 and 
1947) for components of variance for plant height are: 9.30 for 0%, 9.25 for 
(To, and 18.05 for (7%. The heritability calculated from these estimates is 44.1 
per cent. The results reported in this discussion are similar in that heritabil- 
ity is high and <!% has about twice the value of a]}. 



170 HAROLD H. SMITH 

Results of Selection 

The result of selection for tall plants with many, large leaves can be seen 
by comparing the means of Al with those of the parental and hybrid genera- 
tions in Tables 10.1 and 10.2. 

From the 1947 data it is evident that in the F4 generation of selection Al a 
significant increase had been obtained over the parents and Fi's in plant 
height and green weight. This was accompanied by a lengthening in time 
required to reach maturity. With regard to this latter character, it was noted 
that the average time for reaching maturity in five of the six Fi's was later 
than the average of their respective parents. This is contrary to the usual 
result in first generation hybrids of certain other plants, as maize and toma- 
toes; and, where early maturity is an important economic character, would 
generally not be considered a manifestation of hybrid vigor, at least in a 
"beneficial" sense. 

The number of leaves in selection Al was significantly higher (P < .05) 
than in any of the Fi's, and all but the most vigorous parent, Olson 68. Leaves 
of the selection were shorter than the parent with the longest leaves (Olson 
68), not significantly different from the three Fi's that involved this parent, 
and longer than in the other three parents and three Fi's. 

The 1949 data (Table 10.2) corroborated the 1947 results. There was a 
significant increase (P ^ .01) in plant height and in number of nodes over 
the two main parents and their Fi. Number of nodes, rather than of leaves, 
was used since it is a more reliable criterion of the same character. As in 
1947, there was a less marked effect of selection on leaf length, though there 
appeared to be an increase in Al from the F4 to the Fe. For this character 
the selection was superior to 34753 and the Fi, but not significantly different 
from Olson 68, although a close approach to significance at the 5 per cent 
level of probability was reached. 

The total within plot variances of selection AKFe) for plant height, num- 
ber of nodes, and leaf length were in no case significantly higher than for the 
more variable parent. It was deduced, therefore, that the inbred selection 
had reached relative homozygosity. 

The general conclusions were that an inbred selection had been produced 
which had increased plant height, more nodes, heavier green weight, and a 
longer growth period than any parent or Fi. Length of leaf had been main- 
tained at least at the level of the best parent variety. 

It was also noted, though no quantitative data were taken, that selection 
Al had markedly less vigorous sticker growth at topping time than any of the 
other varieties or hybrids. This is an important agronomic character. 

Heritability and Gain 
One of the objectives in conducting these experiments was to attempt to 
determine to what extent the progress realized in actual selection experi- 



FIXING TRANSGRESSIVE VIGOR IN NICOTIANA RUSTICA 171 

ments could be related to the heritability of a character as determined from 
F2 and first backcross data. 

Results on the three main characters studied were similar in that there 
was no indication of complex genie interactions, and that estimates of the 
number of effective factors were low and of the same order of magnitude in 
each. If we wish to assume that the "reach" or selection differential (in terms 
of standard deviations) was the same for each character, and this is approxi- 
mately correct though exact records on this point are lacking, then the gain (in 
terms of standard deviations) due to selection should be roughly pro])ortional 
to the heritability. The gain was calculated as the difference between the 
mean of selection Al and the mid-parent value, divided by the standard 
deviation of the F2 (1949 data. Table 10.2). 

The relationships between heritabilities and gains can be observed by com- 
paring the last two columns in Table 10.6. With regard to plant height and 
leaf length, both heritability and gain are higher in the former character; 
though the gain is less in plant height than would have been anticipated from 
the relative heritabilities. Some possible explanations for this latter result 
could be that the selection differential for plant height was lower than for leaf 
length, that there was a relatively more rapid reduction in heritability, or 
that an approach to a physiological limit for tallness was made. 

The gain in node number is disproportionately high in relation to its 
heritability. Some possible explanations for this result could be that the 
selection differential was higher, that there was a genetic correlation with 
plant height, or that the selected character was determined by a preponder- 
ance of recessive genes (see Fi result), and individual plants selected for high 
node number were largely homozygous for recessive+genes. 

DISCUSSION 

The experimental results have shown that first generation crosses among 
different varieties of Nicotiana riistica exhibit different degrees of character 
expression ranging from the smaller parent value to above the larger parent. 
By selection and inbreeding it was possible to develop an essentially true- 
breeding improved line which exceeded the best Pi or Fi in most character- 
istics measured. 

This same type of result has also been obtained in our experience with the 
commercial species, N. tahacum, and it may be generally characteristic of 
self-fertilized plants, as, e.g., Phaseolus ■vulgaris (Malinowski, 1928), soy- 
beans (Veatch, 1930), and Galeopsis (Muntzing, 1930). 

Crossbreeding 
There have been relatively few fundamental changes in the standard 
domestic varieties of N. tabacum over a long period of years, except for recent 
development of types resistant to destructive diseases (Garner, 1946). 



172 HAROLD H. SMITH 

Houser (1911) originally suggested the use of first generation intervarietal 
tobacco hybrids on a commercial scale to increase yields. He presented breed- 
ing results on cigar filler types, dating back to 1903, in which the hybrids 
outyielded the parent types by as much as 57 per cent. Plant breeders in 
various tobacco-growing areas of the world have observed hybrid vigor 
among first generation hybrids of commercial varieties (Ashton, 1946), and 
have suggested its use in practice to increase production. Currently, con- 
sideration is being given to improving the yield of flue-cured varieties by 
this method (Patel et al., 1949). 

The results of Hayes (1912), Hayes, East, and Beinhart (1913), and East 
and Hayes (1912) showed that by intervarietal hybridization, selection, and 
inbreeding the number of leaves, an important factor in yield of tobacco, 
could be fixed at a level exceeding the parents or Fi. Regarding the use of 
Fi hybrids on a commercial scale, they stated (Hayes, East, and Beinhart, 
1913), 

While it is doubtless true that by this method the yield could be somewhat increased, 
the yield factor, for cigar wrapper types at least, is only of secondary importance com- 
pared with quality. Because of the great importance of quality it seems much more reason- 
able to suppose that further advance can be made by the production of fixed types which in 
themselves contain desirable growth factors, such as size, shape, position, uniformity, vena- 
tion, and number of leaves, together with that complex of conditions which goes to make 
up quaHty, than by any other method. 

The problem of producing higher yielding varieties of N. tabacum with 
acceptable quality characteristics of the cured leaf remains today. Kosmo- 
demjjanskii (1941) bred four families from the cross Dubec 44XTrebizond 
1272, two Russian varieties of .Y. tabacum, which, he reported, were uniform 
for morphological characters and flavor and maintained transgression in 
plant height and number of leaves to the F; generation. 

While first generation hybrids between selected parents may be of use as 
a temporary measure to improve self- fertilizing crop plants, it would appear, 
in so far as can be generalized from the results on Nicotiana, that production 
of fixed types with favorable transgressive characteristics offers a better long- 
time solution. Within any one type of tobacco, such as flue-cured, there are 
currently available a number of high quality inbred varieties which, though 
of similar phenotype, may be expected to differ by genes of a multifactorial 
system affecting size characteristics (Emerson and Smith, 1950). Selections 
from intervarietal crosses may be expected, therefore, to yield fixed types of 
increased size without presenting undue difficulties to the breeder attempt- 
ing to maintain quality. 

In order to discuss the hereditary basis for experimental results on hetero- 
sis and inbreeding, current concepts of the genetic and evolutionary mecha- 
nisms involved are briefly presented. In the evolution of naturally crossbred 
organisms, mutation and selection result in the accumulation of dominant 
favorable genes, hidden deleterious recessives, and alleles or complexes of 



FIXING TRANSGRESSIVE VIGOR IN NICOTIANA RUSTICA 173 

linked polygenes which give heterotic effects as helerozygoles. Heterosis is 
explained genetically as due to the accumulated effect of the favorable domi- 
nants and/or coadapted heterozygous combinations. It is an adaptive evolu- 
tionary phenomenon (Dobzhansky, 1950). 

Selfing 

In naturally selfed populations there are accumulated, for the most part, 
favorable genes that are either dominant, recessive, or lacking in strong 
allelic interactions. Dominance is of little evolutionary significance, and 
hence a preponderance of favorable dominant genes is not to be expected. 
Furthermore, there would ordinarily be no adaptive significance to favorable 
heterozygous combinations. One possible exception is suggested by Brieger's 
(1950) demonstration that "if survival values for both homozygotes should 
be below 0.5 (compared to the heterozygote value of 1.0) in selfed populations, 
a final equilibrium is reached with all three genotypes remaining in the 
population." Such a condition might have adaptive value in maintaining 
variability in selfed organisms. Hybrid vigor in self-pollinated plants, in 
view of the above considerations, is a chance manifestation, an "evolutionary 
accident" causing luxuriant growth (Dobzhansky, 1950), and not an adap- 
tive product of mutation and selection. 

However, from published data on crosses within selfed species of culti- 
vated plants, it appears that hybrid vigor is of frequent rather than chance 
occurrence. Reported results with flax (Carnahan, 1947), wheat (Harrington, 
1944), barley (Immer, 1941), tomatoes (Larson and Currance, 1944), egg- 
plants (Odland and Noll, 1948), and soybeans (Weiss, Weber, and Kalton, 
1947) all demonstrated that hybrid vigor is characteristic of Fi's. If these 
data constitute a representative sample, then, although hybrid vigor is an 
evolutionary accident in naturally selfed species, it is not a genetical accident. 

The result may be interpreted genetically as follows: Selfed species are 
purged of deleterious genes by selection. Different varieties within the 
species have accumulated different alleles all of which control "non-defec- 
tive," slightly different physiological reactions. The combination of divergent 
alleles in heterozygous condition may, more frequently than not, act as East 
has suggested in a complementary manner to produce a more efficient physio- 
logical condition. This is expressed phenotypically by the hybrid manifesting 
more vigorous growth than midway between the homozygotes. Subsequent 
selection and inbreeding, however, would permit an accumulation of the most 
favorable alleles or gene complexes in the homozygous condition. 

As a simplified schematic example, let us assume that two varieties. Pi 
and P2, differ by three alleles or linked polygene complexes: X' is dominant 
and favorable for vigorous growth, F^ is a favorable recessive, and at the Z 
locus the product of the heterozygous condition is above the mean of the 



174 HAROLD H. SMITH 

homozygotes. The composition of parents, Fi, and selected inbred is shown 
below with arbitrary "size" values assigned to each. 

Pi = XKX' (4) + y V (4) +Z'Z' (2) =10 

Po = X'X^ (2) + Y^V (2) -\-Z'-Z-' (6) =10 

Fi = X'X' (4) + I'l ¥' (2) ^Z'Z- (5) =11 

sel. = X'X' (4) + Y'V (4) -^Z-Z'- (6) =14 

Although the difficulty in selecting superior inbreds would become 
greater with increasing numbers of effective segregating units, the following 
advantages of selfed over crossbred systems would enhance the opportunity 
for success: (1) lack of deleterious recessives, (2) less preponderance of 
dominant favorable alleles, (3) homozygous pairs of alleles are superior, as a 
result of an adaptive evolutionary process, to heterozygous combinations. 
Naturally inbred organisms are products of historical evolutionary processes 
in which harmonious systems of homozygous loci have been selected to 
attain optimum adaptation. These considerations favor the expectancy 
and practicability of obtaining maximum advance through selection and 
inbreeding with self-fertilized organisms. 

SUMMARY 

There were two general purposes in conducting these experiments: First, 
to demonstrate that by selection following intervarietal hybridization in a 
self-fertilized organism, inbreds could be produced which transgressed the 
character expression in parents and Fi; secondly, to investigate the relation 
between estimated heritability and the actual results of selection. 

An inbred selection of Nicotiana rustica which transgressed the Pi and Fi 
characteristics in plant height, node number, and leaf length was obtained. 
The heritabilities for these three characters were calculated to be 54.9 per 
cent, 12.4 per cent, and 11.2 per cent, respectively. The gains (in terms of 
standard deviations) due to selection were 1.74, 2.42, and 0.91, respectively. 
Some possible explanations for the lack of direct proportionality between 
heritability and gain were discussed. 

The number of effective segregating factors for each of the three characters 
studied was estimated to be of the same order of magnitude and relatively 
few. Non-isodirectional distribution of -|- and — genes in the parent varieties 
contributed to an underestimation of this number. 

Non-allelic interactions were apparently not an important source of 
variation, as indicated by scaling tests and evidence from double cross means. 

Reasons for expecting greater advances by selection and inbreeding, as 
contrasted to the use of first generation hybrids, in naturally self-fertilizing 
genetic systems were reviewed. 



PAUL C. MANGELSDORF 

Harvard University 



Chapter 1 1 

Hybridization in 
the Evolution of Moize 



All varieties and races of maize so far studied prove upon inbreeding to con- 
tain numerous heterozygous loci, and all respond to inbreeding with a marked 
decline in vigor and productiveness. Since contemporary maize is both 
heterozygous and heterotic, it is probable that the factors which have been 
responsible for bringing about the present conditions are also factors which 
have played an important, if not the principal role, in the evolution of maize. 

All of the steps involved in the evolution of maize are not yet known. 
Archaeological remains have told us something of the early stages of maize 
under domestication, and we can draw additional inferences about its original 
nature from its present-day characteristics. Our knowledge of the nature and 
extent of its present variation, although far from complete, is already sub- 
stantial and is growing rapidly. By extrapolating forward from ancient 
maize, and backward from present-day maize, w^e can make reasonably valid 
guesses about some of the intermediate stages and about some of the evolu- 
tionary steps which have occurred in its history. 

The earliest known archaeological remains of maize, as well as the best 
evidence of an evolutionary sequence in this species, occur in the archaeo- 
logical vegetal remains found in Bat Cave in New Mexico in 1948. This ma- 
terial which covers a period of approximately three thousand years (from 
about 2000 B.C. to a.d. 1000) has been described by Mangelsdorf and Smith 
(1949). It reveals three important things: (1) that primitive maize was both 
a small-eared pop corn and a form of pod corn; (2) that there was an intro- 
gression of teosinte into maize about midway in the sequence; (3) that there 
was an enormous increase in the range of variation during the period of ap- 
proximately three thousand years resulting from teosinte introgression and 
interracial hybridization. 

175 



176 PAUL C. MANGELSDORF 

INTERRACIAL HYBRIDIZATION IN MAIZE 

For additional evidence on interracial hybridization in maize we may 
turn to existing races of maize. Among these the Mexican races are of par- 
ticular interest and significance, not because maize necessarily originated in 
Mexico, since there is considerable evidence that it did not, but because 
Mexico is a country where primitive races, which in other places are to be 
found primarily as archaeological remains, still exist as living entities. It 
is possible in Mexico to find all stages between ancient primitive races and 
modern highly-developed agricultural races. One has only to place these 
racial entities in their proper sequence in order to have at least the outline 
of an evolutionary history. 

Wellhausen et al. (1951) have recently made a comprehensive study of the 
races of maize of Mexico. They recognize twenty-five distinct races as well 
as several additional entities which are still somewhat poorly defined, but 
some of which may later be described as races. They divide the known races 
into four major groups as follows: 

Group No. Races 

1. Ancient Indigenous 4 

2. Pre-Columbian Exotic 4 

3. Prehistoric Mestizos 13 

4. Modern Incipient 4 

Origin of Mexican Races of Maize 

Ancient Indigenous races are those which are believed to have arisen in 
Mexico from the primitive pod-pop corn similar to that whose remains were 
found in Bat Cave in New Mexico. The races in this group are called in- 
digenous not because they necessarily had their primary origin in Mexico, 
but because they are thought to be the product of indigenous differentiation 
from a remote common ancestor. The differentiation is assumed to have re- 
sulted from independent development in different localities and environ- 
ments with hybridization playing little if any part. 

Four races of the Ancient Indigenous group — Palomero Toluqueno, Arro- 
cillo Amarillo, Chapalote, and Nal-tel — are recognized. All of these, like their 
primitive ancestor, are pop corn. Two of the four — Chapalote and Nal-tel — 
are forms of pod corn. All have small ears, and all are relatively early in 
maturity. 

Pre-Columbian Exotic races are those which are believed to have been 
introduced into Mexico from Central or South America before 1492. Four of 
these races — Cacahuazintle, Harinoso de Ocho, Oloton, and Maiz Dulce — 
are recognized. The evidence for their antiquity and exoticism derives prin- 
cipally from two sources: all have South American counterparts; all except 
Maiz Dulce have been parents of hybrid races, some of which are them- 
selves relatively ancient. 



HYBRIDIZATION IN THE EVOLUTION OF MAIZE 177 

Prehistoric Mestizos,^ thirteen in number, are races which are believed to 
have arisen through hybridization between Ancient Indigenous races and 
Pre-Columbian races and hybridization of both with a new entity, teosinte. 
The term })rehistoric rather than pre-Columbian is used for this grouj) be- 
cause, although all are prehistoric in the sense that there is no historical evi- 
dence of their origin, it is not certain that all are pre-Columbian. 

Modern Incipient races are those which have come into existence in the 
post-Columbian period. These races, of which four are recognized, have not 
yet reached a state of genetic equilibrium. They are recognizable entities but 
are still changing. 

The seventeen races comprising the two last groups all appear to be prod- 
ucts of hybridization, either between races in the first two groups, or between 
these races and teosinte. In several cases, secondary and even tertiary hy- 
bridization seems to have occurred. 

That a race is the product of previous hybridization seems highly prob- 
able when the following four kinds of evidence are available. 

1. The race is intermediate between the two putative parents in a large 
number of characteristics. 

2. The putative parents still exist and have geographical distributions 
which make such hybridization possible and plausible. 

3. Inbreeding of the suspected hybrid race yields segregates which ap- 
proach in their characteristics one or the other of the two putative parents — 
in some cases both. 

4. A population quite similar to the race in question can be synthesized 
by hybridizing the two putative parents. 

Wellhausen el al. (1951) have presented all four kinds of evidence for the 
hybrid origin of a number of the present-day Mexican races. They have pre- 
sented similar but less complete evidence for the remainder. 

The variety Conico, for example, which is the most common race in the 
Valley of Mexico, is clearly the product of hybridizing the ancient Palomero 
Toluqueno with the exotic Cacahuazintle. Conico is intermediate between 
these two races in many characteristics. The two putative ancestral races still 
are found in isolated localities in the Valley of Mexico. The race is interme- 
diate in its characteristics between the two suspected parents. Inbreeding 
yields segregates which almost duplicate in their characteristics one of the 
parents — Palomero Toluqueno. Segregates approaching the other suspected 
parent, Cacahuazintle, also result from inbreeding but this parent is never 
exactly duplicated. Obviously the race has become something more com- 
plex than a mixture of equal parts of two earlier races. Nevertheless the 
crossing of Palomero Toluqueno and Cacahuazintle still produces a hybrid 
which in many respects is scarcely distinguishable from the suspected hybrid 
race. The data in Table 11.1 show that Conico is intermediate between Palo- 

1. Mestizo is the Latin-American term for a racial hybrid. 



178 



PAUL C. MANGELSDORF 



mero Toluqueno and Cacahuazintle in a large number of characteristics. They 
also show how closely a recently-made hybrid of these two ancient races re- 
sembles the suspected hybrid race, Conico. Ears of the three races and the 
hybrid are illustrated in Figure 11.1. 

The hybrid race, Conico, has in turn been the ancestor of still more complex 
hybrid races. A Modern Incipient race, Chalqueho, which has originated in his- 
torical times in the vicinity of the village of Chalco in the Valley of Mexico, 

TABLE 11.1 
COMPARISON OF CONICO WITH ITS PUTATIVE PARENTS* 



Characters 



Ears and plants: 

Ear diameter, mm 

No. rows grain 

Width kernels, mm 

Thickness kernels, mm 

Diameter peduncle, mm 

Length ear, cm 

Height plant, cm 

Tillering index 

Pilosity score 

Internal ear characters: 

Ear diameter, mm 

Cob diameter, mm 

Rachis diameter, mm 

Length kernels, mm 

Estimated rachilla length, mm 

Cob/rachis index 

Glume/kernel index 

Rachilla/kernel index 

Pedicel hairs score 

Rachis flap score 



Races 



Palomero 
Toluqueno 



37.1 
21.8 
4.6 
2.8 
8.0 
9.8 
175 
.26 
3 



34.0 
19.5 
10.4 
11.4 
.4 
1.88 



.40 
.04 








Fi 
Hybrid 



Conico 



45.2 
18.6 
6.8 
3.6 
9.2 
11.8 
200 
.35 
4 



45.1 

15.7 

7.4 

3.9 

9.8 

12.6 

193 

.22 
3-4 



42.4 

19.0 

9.6 

14.8 
1.6 
1.98 
.32 
.11 
2-4 



2-3 



Cacahua- 
zintle 



53.2 
16.2 
9.8 
5.3 
10.6 
14.7 
210 
.39 
4 



47.0 
27.7 
11.7 
14.0 

3.6 

2.37 
.57 
.26 

4 

3 



* After Wellhausen el al. 



is undoubtedly the product of hybridizing Conico with Tuxpeiio, a pro- 
ductive lowland race of the Prehistoric Mestizo group. Since Tuxpeno is 
itself a hybrid, the postulated pedigree for Chalqueno, which is shown in 
Figure 11.2, becomes quite complex. 

In the pedigree of Tuxpeno a distinction has been made (by employing 
different styles of type) between the facts which are well-established and 
those which are largely based upon inference. There is little doubt that 
Conico is a hybrid of Palomero Toluqueno and Cacahuazintle, or that Chal- 
queno is a hybrid of Conico and Tuxpeno. There is little doubt that Tuxpeno 
is a hybrid derivative of Tepecintle, but it is not certain that the other par- 
ent is Olotillo, although this is the best guess which can be made with the 




Fig. 11.1 — Ears of the Mexican maize races Palomero Toluqueno, Conico, and Cacahua- 
cintle. Conico is intermediate between the two other races and is thought to be the product 

of their hybridization. 



CHALQUENO 



CONICO 



PALOMERO TOLUQUENO 



"•CACAHUACINTLE 



TUXPENO 



OLOTILLO 



^TEPECINTLE 



HARINOSO FLEXIBLE 



TEOSINTE 



HARINOSO DE GUATEMALA 



^TEOSINTE 



Fig. 11.2— The postulated geneology of the Mexican race Chalqueno. Parts of the geneal- 
ogy not well established by experimental evidence are shown in Italics. 



180 PAUL C. MANGELSDORF 

evidence now at hand. That Olotillo and Tepecintle are both hybrid races 
involving teosinte is even more difficult to prove, although data on chromo- 
some knobs presented by Wellhausen et al. tend to substantiate such a con- 
clusion. 

There is at least no doubt that interracial hybridization has been an im- 
portant factor in the evolution of maize in Mexico. Has this hybridization 
produced heterosis, or has it merely resulted in Mendelian recombination? 
The extent to which the suspected hybrid races remain intermediate be- 
tween the two putative parents suggests that natural selection (operating in 
a man-made environment) has tended to preserve the heterozygote and to 
eliminate the segregates which approach homozygosity. It is at least certain 
that the hybrid races are intermediate between their putative parents in 
their characteristics to a remarkable degree and that they are highly hetero- 
zygous. Even in the absence of natural selection favoring the more heterozy- 
gous individuals, there would seem to be a tendency for repeated interracial 
hybridization to create an ever-increasing degree of heterosis. This is the 
consequence of the fact that maize is a highly cross-pollinated plant, and 
that heterozygosity does not diminish after the Fo in cross-fertilized popula- 
tions in which mating is random. 

Wright (1922) has suggested that the vigor and productiveness of an Fo 
population falls below that of the Fi by an amount equal to l/« of the dif- 
ference between the production of the Fi and the average production of the 
parental stock, where n is the number of inbred strains which enter into the 
ancestry of a hybrid. The formula is also applicable to hybrids in which the 
parental stocks are not inbred lines, but are stable open-pollinated varieties 
in which random mating does not diminish vigor. It is, of course, not ap- 
plicable to hybrids of single crosses which are themselves subject to dimin- 
ished vigor as the result of random mating. 

Hybrid Vigor in Advanced Generations 

The rate at which hybrid vigor diminishes in a population after the F2 gen- 
eration is related to the proportion of outcrossing. This is true whether hybrid 
vigor depends upon heterozygosity or upon the cumulative action of dominant 
genes, and irrespective of the number of genes involved and the degree of 
linkage. With complete selffiig the amount of hybrid vigor retained is halved 
in each succeeding generation. With complete outcrossing the amount of 
hybrid vigor falls to one-half in the F2 and thereafter remains constant. With 
a mixture of selfing and outcrossing an intermediate result is to be expected. 
This can be calculated from the following formula presented by Stephens 

(1950): 

h = l[{\-k)h'-Vk]. 

In this formula h is the proportion of Fi vigor retained in the current gen- 
eration, // is the proportion retained in the preceding generation, and k is 



HYBRIDIZATION IN THE EVOLUTION OF MAIZE 181 

the proportion of outcrossing. The formula is based upon the assumption that 
gene action is, on the average, additive. 

It is obvious (according to this formula) that the percentage of hybrid 
vigor retained in later generations of a cross will approach but never fall be- 
low kjl. Since the value of k in the case of maize lies usually between .9 and 
1.0, it is apparent that the amount of hybrid vigor retained in later genera- 
tions of maize crosses will (with random mating) seldom fall below the 
one-half, which is characteristic of the Fs- 

There are experimental data which tend to show that advanced genera- 
tions of maize crosses behave approximately as would be expected from the 
formulae of Wright and Stephens. 

Kiesselbach (1930) compared the Fi,F2,andF3of 21 single crosses with the 
parental inbred lines. The average yield of the inbreds was 24.0 bushels. The 
average yield of the Fi was 57.0 bushels. The theoretical yield of the Fo is 
40.5 bushels. The actual yield was 38.4 bushels which does not differ signifi- 
cantly from the theoretical. The yield of the F3 was 37.8 bushels which is 
almost identical to the F2 yield. 

Neal (1935) compared the yield in Fi and F2 of 10 single crosses, 4 three- 
way crosses, and 2 double crosses. The theoretical reduction in yield be- 
tween the Fi and F2 in these three groups (based upon Wright's formula) 
should have been 31.1 per cent, 21.0 per cent, and 15.2 per cent respectively. 
The actual reduction was 29.5 per cent, 23.4 per cent, and 15.8 per cent. The 
agreement could scarcely have been closer. 

There is abundant evidence from maize crosses to show that equilibrium 
is reached in F2, and that in the absence of selection there is no further reduc- 
tion in yield in the F3. Data from the experiments of Kiesselbach (1930), 
Neal (1935), and Sprague and Jenkins (1943) are summarized in Table 11.2. 

The data so far presented are concerned with crosses of inbred strains. Do 
hybrids of open-pollinated varieties behave in the same w^ay? Since open- 
pollinated varieties, although not homogeneous, are stable in productiveness 
they should behave in crosses in the same way as inbred strains. Data from 
advanced generations of topcrosses presented by Wellhausen and Roberts 
(1949) indicate that they do. The theoretical yields of the F2 of a topcross 
can be computed from a formula suggested by Mangelsdorf (1939). 

W'ellhausen and Roberts compared the Fi and F2 generations of 31 dif- 
ferent topcrosses each including the open-pollinated variety Urquiza and 
two inbred lines of unrelated varieties. The latter were in all cases first-gener- 
ation selfs. The mean yield of the 31 Fi hybrids (in terms of percentage of 
Urquiza) was 132 per cent. The mean yield of the corresponding 31 F2 hy- 
brids was 126 per cent. Since the yields of the first-generation selfed lines 
entering into the cross is not known, it is impossible to calculate with pre- 
cision the theoretical yield of the F2. However, it is known that good homozy- 
gous inbreds yield approximately half as much as open-pollinated varieties 



182 



PAUL C. MANGELSDORF 



(Jones and Mangelsdorf, 1925; Neal, 1935) which means that inbreds, selfed 
once and having lost half of their heterozygosity, should yield 75 per cent as 
much as the open-pollinated varieties from which they were derived. Assum- 
ing that the single-cross combinations involved are at least equal to the top- 
cross combinations — 132 per cent — we compute the theoretical F2 yield of 
the topcrosses at 117 per cent, which is considerably less than the 126 per 
cent actually obtained in the experiments. From the results it can be con- 
cluded that hybrid combinations including open-pollinated varieties of maize 
retain a considerable proportion of their vigor in advanced generations. 

There is also some evidence to indicate that the amount of heterosis which 
occurs when open-pollinated varieties are used in hybrid combinations may be 

TABLE 11.2 

SUMMARY OF EXPERIMENTS DEMONSTRATING EQUILIBRIUM 
REACHED IN Fo AND NO ADDITIONAL YIELD REDUC- 
TION IN F3 OF MAIZE CROSSES 





Class of Hybrids 


No. 
Hybrids 
Tested 


Yield in Per Cent of Fi 


Investigators 


Fi 


F2 


Fs 


Kiesselbach, 1930 

Neal, 1935 


Single crosses 
Single crosses 
3-way crosses 
Synthetics 


21 

10 

4 

5 


100 
100 
100 
100 


68.0 
70.5 
76.6 
94.3 


66.0 

75.7 


Neal, 1935 

Sprague and Jenkins, 1943. . . . 


75.8 
95.4 


Total and averages 




40 


100 


76.9 


78.2 









considerably higher with Latin-American varieties than with varieties com- 
monly grown in the United States. Wellhausen and Roberts report single 
topcrosses yielding up to 173 per cent of the open-pollinated variety and 
double topcrosses up to 150 per cent. A recent report from the Ministry of 
Agriculture of El Salvador (1949) shows four different hybrids between open- 
pollinated varieties yielding about 50 per cent more than the average of the 
parents. Such increases are not surprising, since the varieties used in the 
experiments are quite diverse, much more so than Corn Belt varieties. 

All of the data which are available on the yields of advanced generations 
of maize crosses, whether the parents be inbred strains or open-pollinated 
varieties, tend to show that a substantial part of the hybrid vigor charac- 
teristic of the Fi is retained in subsequent generations. Thus maize under 
domestication is potentially and no doubt actually a self-improving plant. 
Distinct more-or-less stable varieties or races evolve in the isolation of 
separated regions. Man brings these varieties or races together under condi- 
tions where cross-fertilization is inevitable, and a new hybrid race is born. 
Repeated cycles of this series of events inevitably lead to the development, 



HYBRIDIZATION IN THE EVOLUTION OF MAIZE 183 

without any direct intervention of man, of more productive races. If, in addi- 
tion, natural selection favors the heterozygous combinations as it does in 
Drosophila (Dobzhansky, 1949), then the retention of hybrid vigor in ad- 
vanced generations of maize crosses will be even greater than that indicated 
by the experimental results. 

INTER-SPECIFIC HYBRIDIZATION OF MAIZE AND TEOSINTE 

Superimposed upon these evolutionary mechanisms, at least in Me.xico 
and Central America, is a second kind of hybridization which involves the 
introgression of teosinte into maize. The importance of this evolutionary 
factor would be difficult to overemphasize, for as Wellhausen el al. have 
shown all of the more productive races of maize of Mexico show evidence of 
past teosinte introgression. 

The genetic nature of teosinte need not enter into the present discussion. Dr. 
R. G. Reeves and I concluded some years ago that teosinte is not, as many 
botanists have supposed, the ancestor of maize, but is instead the progeny 
of a cross of maize and Tripsacum. This conclusion has not yet been ex- 
perimentally proven, and although there is much evidence to support it, it is 
by no means universally accepted by other students of corn's ancestry. For 
the purpose of this discussion we need not debate this particular point, since 
we need only to recognize that there is a well-defined entity known as teo- 
sinte which occurs as a weed in the corn fields of central Mexico and as a wild 
plant in Southwestern Mexico, Guatemala, and Honduras. 

Teosinte is far more common than formerly supposed. Twenty -five years 
ago its occurrence was known in only three or four localities in Mexico. Since 
then, numerous additional sites have been described in Mexico and Guate- 
mala, and recently a locality in Honduras has been added (Standley, 1950). 

Teosinte is the closest relative of maize. It has the same chromosome num- 
ber (ten) as maize, and hybridizes easily with it to produce hybrids which are 
completely fertile, or almost so. The chromosomes of corn and teosinte are 
homologous to the extent that they pair almost completely. Crossing over 
between teosinte and corn chromosomes is of the same order as crossing over 
in pure corn (Emerson and Beadle, 1932). 

Present-Day Hybridization 
Since both teosinte and maize are wind-pollinated plants and since they 
hybridize easily, it is almost inevitable that hybridization between the two 
species should occur in any region where both are growing. There is no doubt 
that such hybridization is constantly occurring, and that it has been going 
on for many centuries. Fi hybrids of com and teosinte have been collected 
in both Mexico and Guatemala. They are especially common in Central 
Mexico where teosinte grows as a weed. In 1943, 1 obtained some data on the 
extent to which hybridization occurs near the village of Chalco where teosinte 
is a common weed in and around the corn fields. In a field where teosinte oc- 



184 PAUL C. MANGELSDORF 

curred abundantly as a weed permission was obtained from the owner to tag 
and harvest 500 consecutive plants. Of the 500 plants tagged, 288 proved to 
be maize, 219 were teosinte plants, and 3 were Fi hybrids. Of the 288 ears 
classified as maize, 4 showed definite evidence of contamination with teosinte 
in earlier generations. In addition, one ear was found in an adjacent row (not 
part of the sample of 500 plants) which was identical in its characteristics 
with a first backcross to teosinte. 

The plants in this field therefore furnished unmistakable evidence of hy- 
bridization, both present and during the recent past, between maize and 
teosinte. One plant out of every 167 plants in the field was a vigorous Fi hy- 
brid shedding abundant pollen which became part of the general pollen mix- 
ture in the field. The Fi hybrids themselves, in spite of their vigor, have a low 
survival value. The Mexican farmer makes no distinction between teosinte 
and the Fi hybrids. Both are left standing in the field when the corn is har- 
vested. The pure teosinte disperses its seeds which are enclosed in hard bony 
shells, and a new crop of teosinte plants appears the following spring. But 
the Fi hybrids have no effective means of seed dispersal, and their seeds, only 
partially covered, are quite vulnerable to the ravages of insects and rodents. 

Both maize and teosinte are quite successful in occupying distinct niches 
in Mexican corn fields. The one, a cultivated plant, depends for its survival 
upon its usefulness to man. The other, a weed, depends for survival upon its 
well-protected kernels and its efficient method of dispersal. There is no such 
niche for the Fi hybrid. It is discarded by man as a cultivated plant, and it 
cannot compete with teosinte as a weed. "Finding no friend in either nature 
or man" (to use Weatherwax's apt description) the Fi hybrids would be of 
no evolutionary significance were it not for the fact that they hybridize with 
both parents. Thus there is a constant introgression of teosinte into maize and 
of maize into teosinte. In the vicinity of Chalco, in Mexico, this process has 
gone on so long and the teosinte has become so maize-like in all of its charac- 
ters, that maize and teosinte plants can no longer be distinguished until after 
the pistillate inflorescences have developed. The teosinte of Chalco has "ab- 
sorbed" the genes for hairy leaf sheaths and red color characteristic of the 
maize of the region. Individual plants of teosinte have been found which have 
the yellow endosperm color of corn, although teosinte is normally white- 
seeded. 

The introgression of teosinte into maize in Mexico today is an established 
fact. The question is how long this process has been going on and whether it 
is strictly a local phenomenon or whether it has affected the maize varieties 
of America. 

Practically all students of maize and its relatives recognize that teosinte 
varieties differ in the degree to which they have become maize-like. Longley 
(1941), for example, considers the teosinte of Southern Guatemala to be the 
least maize-like and that of Mexico the most maize-like. 



HYBRIDIZATION IN THE EVOLUTION OF MAIZE 185 

Rogers (1950) has shown that teosinte varieties differ quite markedly in 
their genes governing the characteristics in which maize and teosinte differ, 
especially characters of the j)istillate inflorescence, tillering habit, and re- 
sponse to length of day. He attributes these differences to varieties in the 
type and amount of maize germplasm which has become incorjiorated into 
teosinte. 

If teosinte varieties differ in the amount and kind of maize contamination 
which they now contain, it is difficult to escape the conclusion that maize 
varieties must likewise differ in the amount of teosinte contamination. There 
is little doubt that maize varieties do differ in this respect. 

Ancient Hybridization 

The prehistoric maize from Bat Cave has already been briefly mentioned. 
The earliest Bat Cave corn, dated at approximately 2000 B.C., shows no 
evidence whatever of teosinte introgression. Beginning about midway in the 
series (which would be about 500 b.c. if the sequence were strictly linear but 
which, according to unpublished radio-carbon determinations made by Libby , 
is probably somewhat later) cobs make their appearance which are scarcely 
distinguishable from the cobs which we have produced experimentally by 
crossing corn and teosinte. \\'eatherwax (1950) regards this evidence of teo- 
sinte introgression as far from conclusive, and it is, of course, quite impossible 
to prove that a cob a thousand years or more old is the product of hybridiza- 
tion of maize and teosinte. Nevertheless, it is true that teosinte introgression 
produces certain definite effects upon the cob, as some of us who have studied 
the derivatives of teosinte-maize crosses on an extensive scale are well aware. 

When it is possible to duplicate almost exactly in experimental cultures 
specimens found in nature, the odds are at least somewhat better than even 
that the resemblance between the two specimens is more than coincidence. 
There is little doubt in my mind that the later Bat Cave corn is the product 
of contamination with teosinte. Certainly it differs from the earlier Bat Cave 
corn quite strikingly, and it is exactly the way in which it would be expected 
to differ if it is the product of teosinte introgression. 

Significance of Chromosome Knobs 

Mangelsdorf and Reeves (1939) suggested some years ago that the deeply 
staining heterochromatic knobs, characteristic of the chromosomes of many 
varieties of maize, are the result of the previous hybridization of maize and 
teosinte, or more remotely of maize and Tripsacum. There has been much 
indirect evidence in support of this hypothesis (especially Mangelsdorf and 
Cameron, 1942; Reeves, 1944), and the recent studies of Wellhausen ei al. 
on Mexican races of maize provide additional evidence of this nature. 
Chromosome knob number in Mexican races is closely correlated with the 
characteristics of the races. The four Ancient Indigenous races, assumed to 
be relatively pure corn, have an average chromosome knob number of 4.2. 



186 



PAUL C. MANGELSDORF 



The four Pre-Columbian Exotic races, also believed to be relatively free 
from contamination, have an average chromosome knob number of 4.3. The 
thirteen Prehistoric Mestizos and the four Modern Incipient races (all except 
one of which are assumed to involve teosinte introgression) have chromo- 
some knob numbers of 7.1 and 8.0, respectively. 

It is interesting to note that in races for which hybridization is postulated 
the hybrid race, although usually intermediate in chromosome knob number 
between its two putative parents, resembles most closely the parent with a 
high knob number. For the eleven hybrid races for which chromosome knob 
numbers are available, not only for the hybrid races but for the two suspected 
parent races, the data (Table 11.3) are as follows: the average of the lower- 

TABLE 11.3 

CHROMOSOME KNOB NUMBERS OF MEXICAN HY- 
BRID RACES OF MAIZE AND OF THEIR 
PUTATIVE PARENTS* 



Hybrid Race 



Tabloncillo 

Comiteco 

Jala 

Zapalote Chico. . 
Zapalote Grande 

Tuxpeno 

Vandeiio 

Chalqueno 

Celaya 

Conico Norteno. 
Bolita 

Averages . . . 



Race 



7.6 
5.6 
7.5 
11.7 
7.4 
6.1 
8.1 
6.8 
8.5 
8.0 
8.6 



7.8 



Parents 



Lower 



low 
5.0 
5.6 
5.5 
7.0 
6.3 
6.1 
1.0 
6.1 
1.0 
7.6 



5.1 



Higher 



8.0 



.0 
.6 
.0 

,7 
.0 



7. 
7. 
9. 

11. 
9. 
7.4 
6.1 
7.6 
8.5 

11.7 



8.5 



* Data from Wellhausen et al. 

numbered parent was 5.1 knobs, of the higher-numbered parent, 8.5 knobs, 
of the hybrid, 7.8 knobs. The fact that the average knob number in the 
hybrid races approaches the average knob number of the higher parents 
suggests, perhaps, that natural selection has tended to retain the maximum 
amount of teosinte introgression and hence the maximum number of knobs. 

The Effects of Hybridizing Maize and Teosinte 
There is no doubt that maize and teosinte are hybridizing in Mexico and 
Central America today, and there is at least a strong indication that they 
have done so in the past. What have been the effects of that hybridization? 
One valid way of determining what happens when teosinte introgresses 
into maize is to produce such introgression experimentally. This has been 
done on an extensive scale by crossing an inbred strain of maize, Texas 4R-3, 



HYBRIDIZATION IN THE EVOLUTION OF MAIZE 187 

with four varieties of teosinte, and by repeatedly backcrossing (three times 
in most instances) the hybrids to the inbred strain, retaining various amounts 
of teosinte germplasm through selection. The end result is a series of modified 
inbred strains approximately like the original 4R-3 — all relatively isogenic 
except that parts of one or more chromosome segments from teosinte have 
been substituted for homologous parts from maize. 

That the substitution involves chromosome segments or blocks of genes 
and not single genes is strongly indicated by the fact that the units have 
multiple effects and that there is breakage within the units in some cases, 
although in general they are transmitted intact. Their mode of inheritance 
and their linkage relations can be determined as though they were single 
genes. Yet each of the units affects many if not all of the characters in which 
maize and teosinte differ. The block of genes on chromosome 3, for example, 
although inherited intact as a single hereditary unit, affects number of ears, 
size of ear, number of seeds, size of seeds, number of rows of grain, staminate 
spikelets on the ear, and induration of the rachis. In addition it has a con- 
cealed effect, discussed later, upon such characters as response to length of 
day and the development of single spikelets. The block of genes on chromo- 
some 4 has practically the same effects in somewhat greater degree, but this 
block shows definite evidence of breakage or crossing over which is of the 
order of 30 per cent. 

These blocks of genes are not random samples of teosinte germplasm, but 
represent definite genie entities which are transmitted from teosinte to maize 
in the process of repeated backcrossing. Different varieties of teosinte yield 
comparable if not identical blocks of genes, and the same variety of teosinte 
in different crosses does likewise. Regardless of the amount of introgression 
of maize which teosinte has undergone in its past history, and regardless of 
the differentiation which has occurred between varieties of teosinte, there are 
still regions in all varieties of teosinte, perhaps near the centromeres, which 
have remained "pure" for the original genes. 

Effects in Heterozygous Condition 
When these blocks of genes are introduced into maize they have profound 
effects which differ greatly in the heterozygous and homozygous condition. 
Since maize and teosinte represent completely different morphological and 
physiological systems (especially from the standpoint of their pistillate in- 
florescences and their response to length of day), this substitution, of seg- 
ments of chromatin from one species for homologous segments from the 
other, represents a drastic interchange of parts comparable, perhaps, to in- 
stalling a carburetor or other essential part from one make of car into an- 
other. In the Fi hybrid of corn and teosinte where the blocks of genes are 
heterozygous, there is no particular functional difficulty. Here the two com- 
plete systems are operating simultaneously and the result is a vigorous hy- 



188 



PAUL C. MANGELSDORF 



brid, vegetatively luxuriant, potentially capable of producing great numbers 
of seed. Measured solely by total grain yield, the Fi hybrid does not exhibit 
heterosis since its grain yield is considerably less than that of corn, but meas- 
ured in terms of number of seeds, or number of stalks, or total fodder, the 
hybrid certainly exhibits heterosis. 

In the modified inbred in which a block of genes from teosinte has been 
substituted for a block of genes from maize, the situation is quite different. 
There are no functional aberrations so long as the block of genes from teosinte 
is heterozygous. Under these circumstances it has very little discernible 




Fig. 11.3 — Ears of a teosinte-modified inbred strain 4R-3 which are isogenic except for an 

introduced block of genes from chromosome 3 of Florida teosinte. The ear at the left lacks 

the block of teosinte genes, the center ear is heterozygous for it, the ear at the right is 

homozj'gous for it. Note the high degree of dominance or potence of the maize genes. 



HYBRIDIZATION IN THE EVOLUTION OF MAIZE 



189 



effect. Figures 11.3, 11.4, and 11.5 show ears of corn heterozygous for blocks 
of genes from chromosomes 3 and 4 respectively, compared to "pure" corn 
in the same progeny. The blocks of genes from corn are much more "potent" 
(a term proposed by Wigan, 1944, to describe the integrated dominance 
effects of all genes) than the block of genes from teosinte, at least in the 
striking characteristics which differentiate the two species. This is in itself a 
noteworthy phenomenon since corn is not strongly "dominant" or more po- 
tent than teosinte in the Fi hybrid, where both species contribute more or 
less equally. 




Fig. 11.4 — These ears are the exact counterparts"]<)f those in Figure 11.3 exce])t that the 
block of teosinte genes was derived from chromosome 4 of Florida teosinte. 





-< 



t 







•-^i 



\ r 



m 



Fig. 11.5 — When the inbred 4R-3 is crossed with No. 701 the liybrid ear illustrated above 
(left) is produced. When a modified strain of 4R-3 (right) which has had three blocks of genes 
from Durango teosinte substituted for corresponding maize genes is crossed with No. 701, 
the hybrid (center) is much more maize-like than teosinte-like. The hybrid, being multiple- 
eared, bears a substantially greater number of seeds than either parent and in one experi- 
ment was appreciably more productive. 



HYBRIDIZATION IN THE EVOLUTION OF MAIZE 191 

The reason for the strong jjotence of maize over teosinte in blocks of 
genes introduced from teosinte into maize, is to be found in a phenomenon 
termed "antithetical dominance" which has been postulated by Anderson 
and Erickson (1941) on theoretical grounds. These writers assumed that in 
species hybrids such as that between maize and Tripsacum, the Fi would be 
intermediate but that backcrosses to either parent would strongly resemble 
the recurrent parent. The basis for this assumption is that the possibilities 
for successful recombination of two such different systems is remote. 

The conception of antithetical dominance has some relationship to 
Richey's opinion (1946) that dominance in some cases is no more than a con- 
dition where one allele is capable of doing the entire job, or most of it, while 
the other allele merely stands by. According to this interpretation, genes are 
not favorable because they are dominant, but are dominant because they are 
favorable. They reveal their presence by doing something. 

There is, in any case, little doubt that something of the general nature of 
antithetical dominance or the kind of dominance postulated by Richey is 
involved in the teosinte-maize derivatives. Both teosinte and maize are 
about equally potent in the Fi hybrid, but a small amount of teosinte germ- 
plasm incorporated into maize in the heterozygous condition is definitely 
lacking in potence. 

Effects in Homozygous Condition 

Since a block of teosinte genes introduced into maize is largely recessive 
in its effects when heterozygous, its effects should become much more ap- 
parent in the homozygous condition. This is indeed the case. The ear on the 
right in Figures 11.3, 11.4, and 11.5 illustrates the effects of one or more 
blocks of teosinte genes incorporated in a homozygous condition in the inbred 
strain 4R-3. 

The combination of traits from corn and teosinte which occurs in these 
homozygous teosinte derivatives is characterized by a distinct lack of har- 
mony in the development of the pistillate inflorescence. The husks are too 
short for the ears, the glumes are too small for the kernels and tend to con- 
strict the growing caryopses producing misshapen kernels. The vascular sys- 
tem is inadequate for the number of kernels borne on the ear, and there are 
many shrunken kernels as well as numerous gaps where no kernels have de- 
veloped. Germination of the seeds is often poor, and viability of short dura- 
tion. Homozygous combinations of this kind obviously have a low survival 
value. Indeed it has been difficult to maintain some of them in artificial 
cultures. 

These unfavorable effects of teosinte introgression in the homozygous con- 
dition may be nothing more than the result of substituting parts of one well- 
integrated system for corresponding parts of another. They may, however, 
also involve "cryptic structural differentiation" of the kind suggested by 



192 PAUL C. MANGELSDORF 

Stephens (1950) for species of Gossypium, although the extent of this cannot 
be great, otherwise some combinations would be lethal. But whatever the 
cause, there is little doubt about the reality of the unfavorable effects. 
Therefore, if the repeated hybridization of corn and teosinte which has oc- 
curred in the past has had any permanent effect, one of two things or both 
must have happened: (1) The undesirable effects of teosinte have become 
recessive as the result of natural selection for modifying factors. (2) The 
regions of the chromatin involving teosinte genes have been kept heterozy- 
gous. There is some evidence that both may have occurred. 

There is some evidence, by no means conclusive, that maize varieties of 
today have absorbed teosinte germplasm in the past and are now bufered 
against the effects of teosinte genes. There is at least no doubt that when the 
same variety of teosinte is crossed on a series of maize varieties, considerable 
variation is displayed by the Fi hybrids in the potence of the maize parents. 

In general, varieties which show some evidence of previous contamination 
with teosinte are more likely to produce maize-like Fi hybrids than those 
which do not show evidence of such contamination. Corn Belt inbreds as a 
class produce the most maize-like Fi of any of stocks tested. Figure 11.6 illus- 
trates a case where a South American stock (an inbred strain derived from 
the Guarany corn of Paraguay) is less potent in crosses with two varieties of 
teosinte than is a North American stock (a genetic tester). I also have ob- 
served that blocks of teosinte genes introduced into an inbred strain of 
Guarany by repeated backcrossing have a greater effect than these same 
blocks introduced into Texas 4R-3 or Minn. A158, both of which seem al- 
ready to contain appreciable amounts of teosinte. 

If the increased potency of teosinte-contaminated maize proves to be gen- 
erally true, then the reason for it is that there has been a selection of modify- 
ing factors which have tended to suppress the most unfavorable conspicuous 
effects of the teosinte introgression. Otherwise, varieties of maize containing 
teosinte germplasm should produce hybrids which are more teosinte-like, 
rather than more maize-like, than the average. This is convincingly demon- 
strated experimentally by crossing the original inbred 4R-3 and one of its 
modified derivatives with the same variety of teosinte (Florida type). The 
results are illustrated in Figure 11.7. 

The Fi of 4R-3 X teosinte is a typical Fi hybrid, intermediate between 
its parents. It has both single and double spikelets and, although the fact is 
not revealed by the illustration, it has approximately the same type of re- 
sponse to length of day as does maize. In marked contrast, when a derived 
strain of 4R-3 (in which a block of teosinte genes on chromosome 3 has been 
substituted for a corresponding block of maize genes) is crossed with the 
same teosinte, the Fi hybrid is scarcely distinguishable in its pistillate spike 
from pure teosinte. Furthermore, it has teosinte's response to length of day. 
Plants of this hybrid started in the greenhouse in February did not flower 



HYBRIDIZATION IN THE EVOLUTION OF MAIZE 



193 



until the following October and November. This derivative of a maize-teo- 
sinte hybrid, therefore, carries at least two concealed characteristics of teo- 
sinte: single spikelets and response to length of day. Genes for these two char- 
acters do not express themselves in the derivative itself, but their presence 
becomes immediately apparent when the derivative is crossed with teosinte. 
The situation is comparable to the concealed genes for hair color and texture 



b^ 




i 









, c 



% 



^ 



Fig. 11.6 — A North .\merican stock is more potent in crosses with Nobogame teosinte (A) 
and Durango teosinte (C) than the Guaranj' corn from Paraguay {B and D). This is at- 
tributed to previous introgression of teosinte accompanied by the evolution of modifier 

complexes in North American varieties. 



in persons who are completely bald. The genes are there but have no oppor- 
tunity to express themselves. 

Since varieties of maize which appear to be the product of previous teo- 
sinte contamination, such as those of the Corn Belt, behave quite differently 
in crosses from stocks known to be contaminated, there is at least an indica- 
tion that such contamination has become modified through selection acting 
upon the modifier complex. More data are obviously needed on this problem. 

A second question which arises in considering the effects of the natural 
hybridization of com and teosinte is whether there is any mechanism which 



194 



PAUL C. MANGELSDORF 



tends to maintain the maize-teosinte loci in a perpetual state of heterozy- 
gosity. It already has been shown that homozygous teosinte genes in the 
maize complex are decidedly deleterious. Therefore, if the teosinte genes are 




^*"«*W*^iW- 







D 



Fig. 11.7 — When the inbred 4R-3(^) iscrossed with Florida teosinte (C), thcFi hybrid ears 
(B) are maize-like in having four-ranked ears, some double spikelets, and partially naked 
seeds. When a teosinte-modified strain of 4R-3 (D) is crossed with Florida teosinte (F), 
the Fi hybrid (E) is much more teosinte-like. The spikes are two-ranked, single, and the 
seeds are completely enclosed. The teosinte derivative obviously carries "concealed" genes 

for these teosinte characteristics. 



to survive their deleterious effects, they must be modified through selection 
or the genes must be maintained in a more or less heterozygous state. It may 
be assumed that the latter mechanism would operate only if heterozygosity 
for a group of maize-teosinte genes confers a distinct selective advantage 
making the heterozygous combination superior, not only to the homozygous 
teosinte genes (as it obviously is) but also to the corresponding homozygous 
maize genes. 



HYBRIDIZATION IN THE EVOLUTION OF MAIZE 195 

Data are available both from my experiments and those of R. G. Reeves 
(1950), conducted independently, to indicate that heterozygosity for a block 
of teosinte genes does sometimes confer a selective advantage. In 1944, in my 
experiments, five Corn Belt inbred strains were crossed with the Texas in- 
bred 4R-3, as well as with foar modified strains of 4R-3 in which teosinte 
genes had been substituted for maize genes. The four modified strains may be 
briefly described as follows: 

No. Blocks Teosinte 

Strain Genes Variety 

Modified 4R-3 Strain A 2 Florida 

Modified 4R-3 Strain B 2 Florida 

Modified 4R-3 Strain C 3 Durango 

Modified 4R-3 Strain D 3 Durango 

The Fi hybrids were grown in 1945 in two replications in a modified Latin- 
Square yield test. Several hybrids were omitted for lack of sufficient seed. 
The results are shown in Table 11.4. 

TABLE 11.4 

AVERAGE YIELDS IN BUSHELS PER ACRE OF HYBRIDS OF 

CORN BELT INBREDS WITH TEXAS 4R-3 AND ITS 

TEOSINTE-MODIFIED DERIVATIVES 



4R-3 OR Deriv.^tive 


Corn Belt Inbreds 


K1S5 


38-11 


L317 


701 


CC24 


4R-3 (check) 


108.6 

102.6 

126.6* 

94.2 

93.0 


85.2 


99.0 


100.2 
87.0 


100 2 


Modified Strain A 


88 8 


Modified Strain B 


82.8 
75.6 
57.0 


109.8 
66.0 
71.4 


78 6 


Modified Strain C 

Modified Strain D 


97.8 
146.4* 


92.4 
79.8 



* Difference probably significant. 



Of the 17 hybrids tested, only 3 proved to be better than the correspond- 
ing checks in total yield, and in only 2 of these is the difference significant. 
Although the data are not extensive, there is some indication that the Corn 
Belt inbred strains used in these experiments differ in their ability to "com- 
bine" with the teosinte derivatives. 

Perhaps more important than total yield, from the standpoint of selective 
reproductive advantage, is total number of seeds per plant (Table 11.5). 
Here 6 of the 15 hybrids for which data are available were superior to the 
checks, 4 of these significantly so. 

These results, so far as they go, are in agreement with the recently pub- 
lished results of Reeves (1950). Reeves tested 49 modified 4R-3 lines in hy- 
brids with a common tester. He found none significantly better than the 
check in yield, although several were superior in heat-tolerance. Reeves 



196 



PAUL C. MANGELSDORF 



found, however, that when teosinte germplasm was introduced into another 
inbred strain, 127C, the results obtained in the hybrids were somewhat dif- 
ferent. In 1946, 6 hybrids out of 25 were better than the check, 3 of them 
significantly so. In 1947, 15 hybrids out of 49 were better than the check, 
6 of them significantly so. Reeves suggested that the difference between 4R-3 
and 127C in their response to teosinte introgression lies in the fact that 
4R-3 already contained considerable amounts of teosinte germplasm while 
127C does not. The suggestion is supported by differences in the morphologi- 
cal characteristics of the two lines. 

There was also an indication in Reeves' experiments that the entries with 

TABLE 11.5 

AVERAGE NUMBERS OF SEEDS PER PLANT IN HYBRIDS OF 

CORN BELT INBREDS WITH TEXAS 4R-3 AND ITS 

TEOSINTE-MODIFIED DERIVATIVES 





Corn Belt Inbreds 


4R-3 OR Derivative 


K1S5 


38-11 


L317 


701 


CC24 


4R-3 (check) .... 


849 
756 
937 
1419* 
770 


636 


925 


1132 
1095 


1179 


Modified Strain A 


807 


Modified Strain B 


809' 
573 


1107* 
746 
843 




Modified Strain C 

Modified Strain D 


1696* 
1811* 


885 
864 



* Difference probably significant. 



teosinte genes made their best showing in 1947, a season of severe drought. 

Considering all of the results together it may be concluded that: (c) 
blocks of teosinte genes in the heterozygous condition do in some instances 
improve the total yield of the plants which contain them ; (b) even more fre- 
quently do such blocks of genes increase the total number of seeds produced ; 
(c) there is some evidence that the teosinte derivatives impart resistance to 
heat and drought to their hybrids. 

In those crosses in which the heterozygous combination is superior to 
either of the homozygous combinations, a block of maize genes or a block of 
teosinte genes, natural selection would undoubtedly favor, at least initially, 
the heterozygous combination. If the block of genes were one involving the 
region of the centromere where crossing-over is reduced, it is quite possible 
that the block of genes would be retained more or less intact for a consider- 
able number of generations. The maintenance of heterozygosity through 
natural selection also would be promoted if, as in the case of Drosophila 
studied by Dobzhansky, one set of genes is superior in adapting the organism 
to one kind of environment while the other set contributes to adaptation 



HYBRIDIZATION IN THE EVOLUTION OF MAIZE 197 

to a wholly different environment which the organism also encounters pe- 
riodically. 

It cannot be proved that such a situation exists in the case of maize which 
has become contaminated with teosinte, but it is quite possible that it does. 
For example, human selection when practiced would tend to favor the larger- 
seeded, larger-eared individuals with a minimum of teosinte contamination. 
Natural selection would favor the individuals with the larger number of 
seeds, hence those with an appreciable amount of teosinte contamination. 
These two forces operating simultaneously or alternately would tend to per- 
petuate the heterozygote. Similarly, if maize germplasm were superior in 
seasons of excessive moisture and teosinte germplasm in seasons of drought 
(for which there is some evidence), there would be a tendency for natural 
selection to perpetuate heterozygous combinations. It cannot be demonstrated 
that any of these hypothetical situations actually exist. There is no doubt, 
however, that present-day maize is highly heterozygous, and there is more 
than a suspicion that repeated hybridization with teosinte has been respon- 
sible for part of the heterozygosity. 

DISCUSSION 

The present-day heterozygosity of maize may involve a variety of differ- 
ent factors and forces which have operated during its past history. Two of 
these are now reasonably clear: interracial hybridization, and introgression 
of teosinte into maize. 

When interracial hybridization occurs, hybrid vigor not only manifests 
itself in the first generation, but also persists in part through an indefinite 
number of subsequent generations. Maize under domestication is, therefore, 
potentially a self-improving plant. The evidence from Mexican races of 
maize indicates that repeated interracial hybridization has been an extremely 
important factor in the evolution of maize in Mexico. There is every reason 
to believe that the situation in Mexico, so far as interracial hybridization is 
concerned, is typical of other parts of America. 

The second factor, introgression of teosinte, which is believed to have 
played an important role in the evolution of maize, is not so easily demon- 
strated. There is no doubt, however, that teosinte is hybridizing with maize 
in Guatemala and Mexico today, or that this hybridization has occurred in 
the past. It would be surprising indeed if such hybridization had no effect 
upon the evolution of maize. There is every indication that it has had a pro- 
found effect. All of the most productive modern agricultural races of maize 
in Mexico show evidence of contamination with teosinte, not only in their 
external characters, but also in their internal cytological characteristics. 

It can be shown experimentally that teosinte germplasm, when introduced 
into maize, may sometimes have a beneficial effect when heterozygous, but 
is always deleterious when homozygous. Therefore it follows that after maize 



198 PAUL C. MANGELSDORF 

and teosinte have hybridized, and after there has been an introgression of 
teosinte into maize: (1) the teosinte genes must be eliminated or, (2) their 
effects must be changed through the accumulation of a new modifier com- 
plex, or (3) they must be kept in a heterozygous state. There is evidence, but 
not final proof, that both of the two last-named factors have operated during 
the evolution of maize. Interracial and interspecific hybridization accom- 
panied by sustained heterosis are therefore regarded as two important fac- 
tors in the evolution of maize. 

SUMMARY 

1. Evidence is presented to show that both interracial and interspecific 
hybridization, accompanied by heterosis, have been factors in the evolution 
of maize. 

2. The races of maize of Mexico are cited as an example of interracial hy- 
bridization. Of the 25 Mexican races described by Wellhausen et al., 14 are 
considered to be the products of interracial hybridization. 

3. The hybrid vigor, which occurs when races of maize are crossed, is 
capable of persisting in part in subsequent generations. Maize under domesti- 
cation is therefore potentially a self-improving plant. 

4. Interspecific hybridization of maize and teosinte is occurring in Gua- 
temala and Mexico today, and there is evidence — archaeological, morphologi- 
cal, and cytological — that it has occurred in the past. 

5. Introgression of teosinte into maize in experimental cultures is some- 
times beneficial when the teosinte genes are heterozygous, but is always 
deleterious when they are homozygous. 

6. It, therefore, seems probable that the persistence of teosinte germ- 
plasm in races of maize has been accompanied either by development of 
modifier complexes which have made the teosinte genes recessive in their 
action, or by the maintenance of a continued state of heterozygosity. 

7. The possibility that heterozygosity in maize has been preserved by 
natural selection as it has been in Drosophila is discussed. 



STERLING EMERSON 

California Insfifufe of Technology 



Chapter 12 

Biochemicol Models 
of Heterosis in Neurosporo 



Some of the things that have been learned about gene controlled reactions 
in Neurospora can be used in forming a picture of how individual genes con- 
tribute to heterosis. I wish to consider especially those examples which indi- 
cate that heterozygosity at a single locus may influence the growth of an 
organism to a considerable extent. 

It should be noted at the beginning, however, that one is not justified in 
assuming that the situations found in Neurospora are necessarily similar to 
those occurring in the higher organisms in which heterosis is ordinarily 
studied. It may be unwise to assume that any two organisms are essentially 
similar. There are special reasons for caution in making comparisons between 
Neurospora and higher plants and animals, since the nuclear and chromo- 
somal basis for the expression of heterosis is so dissimilar. On the other hand, 
there is a considerable accumulation of information about the parts played 
in the physiology and biochemistry of Neurospora by individual genes 
(Beadle, 1948; Horowitz, 1950) and, with proper caution, we may assume 
that some of this information may have rather broad application. 

In any haploid organism, such as the ascomycetous fungus Neurospora, 
in which there is a single set of genes in each nucleus, such phenomena as 
dominance, heterozygosis, and heterosis cannot occur. There is, however, 
a condition known as heterocaryosis which permits a loose approximation 
to each. 

CHARACTERISTICS OF HETEROCARYONS 

The plant body of Neurospora can be said to be made up of cells, but they 
are very different from the cells of higher plants. In the first place, the cells 
contain a large and variable number of nuclei in a common cytoplasm. The 
so-called cells themselves are not as discrete as cells are generally supposed 

199 



200 



STERLING EMERSON 



to be. The walls between them have perforations which permit both cyto- 
plasm and nuclei to move from cell to cell. If all nuclei are identical, their 
movement and distribution is probably of minor importance, but if they are 
not identical there may be effects of considerable consequence arising from 
irregularities in nuclear distribution. 

There are two ways in which a mixture of different kinds of nuclei within 
a single cell may come about. In the growth resulting from a sexually pro- 
duced ascospore, or from a uninucleate asexual microconidium, all nuclei 
are directly descended from a single haploid nucleus. Barring mutation, they 
should all have the same genetic constitution. After the growth has become 



Strain X 





Fig. 12.1 — Heterocaryon formation resulting from hyphal fusion (a diagram). 



multinucleate, if a mutation should occur in one nucleus, the descendants of 
that nucleus would then have a different genetic constitution from the re- 
maining nuclei in the common cytoplasm, and a condition of heterocaryosis 
would exist. The second way in which heterocaryons arise is from the direct 
fusion of branches or hyphae of different strains, with the subsequent in- 
termingling of their nuclei. By the latter method, heterocaryons of pre- 
determined genetic constitution can be made at will. 

The controlled production of heterocaryons is shown diagrammatically in 
Figure 12.1. Strain X is represented as having black nuclei to distinguish 
them from the nuclei of strain Y, which are pictured as being white. After 
fusion between hyphae, nuclei of strain Y may migrate into cells of strain X, 
and those of X into Y. It is possible that different hyphal tips, growing from 
this common mass of cells, will have different relative numbers of the two 
sorts of nuclei, as illustrated by the ratios 1:7, 1:1, and 7:1 in three of 



BIOCHEMICAL MODELS OF HETEROSIS IN NEUROSPORA 



201 



the hy})hal tips. To prove that two kinds of nuclei were present in the same 
cells of such heterocaryons, Beadle and Coonradt (1944) cut ofT single hyphal 
tips, transferred them to fresh medium, and then identified two kinds of 
nuclei in the resulting growth by genetic test. 

Where there is freely branching filamentous growth, as in Neurospora, it is 
possible for the two types of nuclei in a heterocaryon to become sorted out 




Fig. 12.2 — Somatic segregation of dissimilar nuclei in the formation of conidia (a diagram). 



purely as a matter of chance, as illustrated in a schematic way in Figure 12.2. 
This diagram actually represents an erect fruiting branch, or conidiophore, 
on which the asexual spores are born. The conidia of Neurospora have 
variable numbers of nuclei, but generally more than one. Dodge (1942) 
proved that two kinds of nuclei were present in the same cell of a heterocar- 
yon by growing cultures from single conidia, and then showing by genetic 
test that some of these cultures had both types of nuclei. In some instances 
he was able to distinguish the heterocaryotic and both homocaryotic types 
in culture derived from single conidia by their mon:)hological characteristics. 
The essential differences between Neurospora and higher organisms with 



202 STERLING EMERSON 

respect to heterosis result from the points just noted. In a diploid which is 
heterozygous for a single gene pair, both alleles are present in the same nu- 
cleus and in equal dosage. Whereas in the corresponding haploid heterocar- 
yon, the two alleles are present in different nuclei, and the relative propor- 
tions of the two alleles vary with the frequencies of the two types of nuclei. 
All cells of a diploid heterozygote have the same genetic constitution, but 
there can be a considerable variation in genetic constitution in different parts 
of a heterocaryotic individual. Interactions between alleles, by which I mean 
such things as the expression of dominance, must result from the ability of 
genes to act at some distance in heterocaryons, in which there is no possibility 
of an intimate association of alleles within a nucleus (Lewis, 1950). It is 
considerations such as these that show that dominance and heterosis-like 
effects in Neurospora are only approximations to the phenomena as known in 
diploid organisms. 

HETEROSIS IN HETEROCARYONS 

An enhancement of growth, closely simulating heterosis, in heterocaryons 
of Neurospora teirasperma was reported by Dodge in 1942. In this paper he 
distinguished between heterocaryotic vigor and the hybrid vigor of diploid 
organisms along much the same lines as I have just done. He suggested that 
the heterocaryotic vigor observed might be the result of complementing 
growth factors whose production was controlled by the two types of nuclei 
(Robbins, 1950). It was later (Dodge, Schmitt, and Appel, 1945) demon- 
strated that genes responsible for enhanced growth segregated and recom- 
bined in a normal fashion. These studies showed that genes residing in differ- 
ent nuclei, but in a common cytoplasm, can cooperate in establishing condi- 
tions favoring rapid growth, and that a condition resembling hybrid vigor 
occurs. 

Meantime, Beadle and Coonradt (1944) had reported on heterocaryons 
between pairs of mutant strains of Neurospora crassa, each of which is unable 
to synthesize a particular vitamin or amino acid. Each mutant strain by itself 
is unable to grow unless supplied with its specific growth requirement, but 
nine heterocaryons involving different combinations of seven mutant strains 
grew at rates approximating that of wild type without the addition of growth 
factors. The authors conclude that the wild type allele is dominant to the 
mutant allele in each of the examples studied. 

Beadle and Coonradt note further that in such heterocaryons, in which 
there is the opportunity for great diversity in the relative numbers of the 
two types of nuclei in different hyphal tips, those tips having the most favor- 
able proportions of nuclei should grow most rapidly. Conversely, rapidly 
growing hyphae should have the two sorts of nuclei in roughly optimal pro- 
portions. In heterocaryons involving pairs of mutant strains, Beadle and 
Coonradt found nuclear ratios varying between approximately 1 : 1 and al- 
most 20: 1. They interpreted these results to mean that the wild type alleles 



BIOCHEMICAL MODELS OF HETEROSIS IN NEUROSPORA 203 

of different mutant genes have different degrees of dominance. A strongly 
dominant wild type allele will need to be present in relatively few nuclei — say 
one in twenty. 

A heterocaryon between two mutant strains could grow at the maximum 
rate over a large range of nuclear proportions, provided the wild type alleles 
concerned were both strongly dominant. A weakly dominant wild type 
allele, on the other hand, must be present in a large j)roportion of the nuclei — 
say nineteen of twenty — to ensure vigorous growth. Heterocaryons in which 
the wild type alleles concerned are both weakly dominant could never result 
in vigorous growth, since the two wild type alleles cannot both be present 
in excess, one being in one type of nucleus and the other in the remaining 
nuclei. 

HETEROSIS DUE TO HETEROZYGOSITY AT ONE LOCUS 

The heterosis effect in heterocaryons studied by Beadle and Coonradt re- 
sults from the mutually complementary nature of the nuclei involved. For 
each deleterious mutant allele in one nucleus there is the corresponding 
favorable and dominant wild type allele in another. In contrast to these 
there are other heterocaryons (briefly reported in Emerson, 1947) in which 
the nuclei differ in only one gene, yet which still show the heterosis effect. 
Heterocaryons in which some nuclei carry the dominant allele and some the 
recessive are superior to homocaryons, all of whose nuclei have the dominant 
allele, or all the recessive. 

Heterocaryotic Suppression of the Sulfonamide-requiring Character 

Most of the heterocaryons of this sort that have been found so far have 
involved the so-called sulfonamide-requiring mutant strain. At 35° on mini- 
mal medium, this strain makes extremely poor growth, but it does keep 
creeping along. After varying lengths of time, it frequently happens that the 
growth will change to a rapid vigorous type. Growth curves of six cultures 
which have reverted to something approaching wild type growth are shown 
in Figure 12.3. When the mycelium had reached the end of the growth tubes, 
inocula from the newest growth were introduced into fresh tubes containing 
minimal medium, resulting in the growth curves shown in the upper part of 
the figure. 

From these curves it can be seen that the reverted type of growth usually 
persists through a conidial transfer. After the mycelium had reached the end 
of the second tube, conidia were removed and used in outcrosses to wild type 
to determine the genetic constitution of their nuclei. These tests showed 
that each of the six cultures represented in Figure 12.3 was a heterocaryon. 
One type of nucleus present in each heterocaryon was identical to those in 
the original sulfonamide-requiring strain. The second type of nucleus in each 
also carried the sulfonamide-requiring gene, sjo (in one instance, that de- 
rived from culture number 1 in Figure 12.3, the sJo gene itself was somewhat 




en 






bo On 



>. ni 

^ S 

^ eg 

O n! 
u 

-2 Pi 

-a o 



.S -^ 

o - 

OS f 

CI- « 

c/i C 

2 ^ 

3 S 

o 3 

c ^ 

3 M 

cr u 

11 o 






en 

cj .5 

c — 

'^ o 



> 



O 



O 

o 

c 



3 

O 

o 






BIOCHEMICAL MODELS OF HETEROSIS IN NEUROSPORA 



205 



modified), and in addition a second mutant gene, S, which was presumably 
responsible for the change in growth (Table 12.1). 

The new mutants appearing in the heterocaryons have been called sup- 
pressors because they overcome the deleterious effect of the sulfonamide- 
requiring gene in heterocaryons. Actually they are not like the usual suj)- 
pressors, because in homocaryotic strains which also carry the sulfonamide- 
requiring gene they do not result in wild type growth. 

Growth characteristics of strains homocaryotic for four of these suppres- 
sors, with and without the sulfonamide-requiring gene, are represented in 



TABLE 12.1 

DISTRIBUTION OF NUCLEI IN 
THE HETEROCARYONS REPRE- 
SENTED IN FIGURE 12.3 





From Culture 
Tube Number 


Nuclei 




sfo. 


+ 


sfo,S 


1. 

2. 
3. 
4. 




3 

6 

15 

8 
8 
1 




5 
5 
2 
1 


5 


1 


6 


14 



Figure 12.4. From these growth curves it can be seen that wild type (+, +) 
is neither inhibited by sulfanilamide in a concentration of 2 X 10~^ M, nor 
stimulated by />-aminobenzoic acid in a concentration of 10"'* M when grown 
at 35°, and is only slightly inhibited by sulfanilamide at 25°. At 35° growth of 
the sulfonamide-requiring strain (sfo, -\-) is stimulated by sulfanilamide and 
inhibited by ^-aminobenzoic acid, though neither substance has an appre- 
ciable effect at 25° in the concentrations used. 

The suppressor from tube 1 (+, ^-1) does not grow at 35°, and grows slow- 
ly on all media at 25°. The suppressor from tube 2 (+, 5-2) differs from wild 
type principally in taking longer to attain its maximum growth rate, though 
there is also some stimulation by sulfanilamide at 35°. When combined as a 
double mutant with the sulfonamide-requiring gene (sfo, 5-2), it almost ap- 
proximates the growth of wild type. The suppressor from tube 4 (+, 5-4) 
differs from wild type in being stimulated by />-aminobenzoic acid and in- 
hibited by sulfanilamide, the inhibition being stronger at 25°. In combination 
with the sulfonamide-requiring gene (sfojSA) it resembles the sulfonamide- 
requiring strain itself except that there is a long lag phase on sulfanilamide 
at 35°, and inhibition at 25°. The suppressor from tube 6, either alone 




HOURS 

Fig. 12.4 — Growth curves of suppressor strains in absence and presence of the sulfonamide-requiring gene (sfo) 
at 35° and 25° on minimal medium (light Hne), 10"* M />-aminobenzoic acid (line of open circles — PABA), and 

2 X 10"" M sulfanilamide (dotted line -SA). 



BIOCHEMICAL MODELS OF HETEROSIS IN NEUROSPORA 207 

(-|_^ 5.6) or in combination with the sulfonamide-requiring gene (sfo, 5-6), 
grows very poorly at 35°. 

Of those illustrated, suppressors numbered 4 and 6 are perhaps the most 
significant to the present discussion. When combined with the sulfonamide- 
requiring gene (sfo, SA and sfo, S-6), neither grows well on minimal medium 
at 35°. Yet heterocaryons between either of these double mutants and the 
sulfonamide-requiring strain are enabled to grow quite well under those 
conditions. In these heterocaryons the sulfonamide-requiring gene is present 
in all nuclei, in some of which it is combined with a suppressor. The suppres- 
sor is not capable of overcoming the ill effects of the sulfonamide-requiring 
gene when present in all nuclei, but is effective when present in only some 

of them. 

Biochemical Basis for the Sulfonamide-requiring Character 

This seeming paradox becomes less important once the nature of the reac- 
tion controlled by the sulfonamide-requiring gene is understood (Zalokar, 
1948, 1950; Emerson, 1950). The diagrams in Figure 12.5 illustrate some of 
the important reactions involved. There are a large number of amino acids, 
vitamins, components of nucleic acid, and so on, that are essential to growth. 
But we shall consider only two amino acids, methionine and threonine, and 
the vitamin />-aminobenzoic acid. Para-aminobenzoic acid is involved in a 
number of reactions essential to growth, one of which is the final step in the 
synthesis of methionine from homocysteine. Wild type carries out all essen- 
tial reactions and produces all essential growth factors, with the exception 
of biotin which must be supplied to all strains. 

The reaction governed by the sulfonamide-requiring gene has not yet been 
identified, but we know quite a little about it. It requires the presence of both 
homocysteine and />-aminobenzoic acid. Presumably homocysteine is used as 
a substrate in this reaction, and />-aminobenzoic acid, or a derivative, is 
needed as a catalyst. The reaction either results in the destruction of threo- 
nine or else interferes with its normal utilization, so that the sulfonamide- 
requiring strain has too little threonine for growth. W^e also know that more 
homocysteine is required for this deleterious reaction than for the syn- 
thesis of methionine, and that in the presence of limiting amounts of homo- 
cysteine, the synthesis of methionine goes on without any interference with 
the utilization of threonine. 

Furthermore, the deleterious reaction requires larger amounts of ^-amino- 
benzoic acid than are needed for all essential reactions combined. Only about 
half as much is needed in the synthesis of methionine, about a quarter as 
much in the production of purines, and very much less still for other essen- 
tial, but still unidentified factors. Both wild type and the sulfonamide-requir- 
ing strain produce about one hundred times as much />-aminobenzoic acid as 
is needed for all essential reactions. 

We know of three ways in which the deleterious reaction leading to threo- 



Wild T ype 



Homocysteine 

CH2-SH 

CH2 
"^ CH-NHz 

COOH 




Methionine 
CH2-S-CH3 

CH2 

CH-NH2 

COOH 



COOH 

^-Aminobenzoic 
acid 




GROWTH 



CH3 
^CHOH 

CH-NH2 
COOH 

Threonine 



Sulfonamide Requirin g, Homoc v steineless 



H-9a 



*- .»-.■■- 














•- c?/f(yr ^ 



Fig. 12.5— 
by genes of 



CH3 

CHOH 

CH-NH2 

COOH 



Certain biochemical reactions essential to growth in Neurospora as influenced 
the sulfonamide-requiring strain (E-15172), the homocysteineless strain (H-98), 
and the aminobenzoicless strain (1633). 



208 



SulfonoTiide Re g uirin g 



CH2-SH , 

CH2 ^'''' 
CH-NHz 
COOH "^2 




CH2-S-CH3 
GHz 

CH-NHg 
COOH 



UNITS 



UNITS^ UNITS 



COOH 





1/2 UNIT 



CH3 

CHOH 

CH-NHz 

COOH 



GROWTH 



Sulfonamide Req uiring, Aminobenzoicless 



CH^-S-CH3 



CH2-SH 
CH2 
CH-NH2 
COOH 




CH3 
^CHOH 

"^'cH-NHz 
COOH 



Fig. 12.5 — Continued 



GROWTH 



209 



210 STERLING EMERSON 

nine deficiency can be prevented by genetic means. The simplest is of course 
by introducing the wild type allele of the sulfonamide-requiring gene, but the 
other two are of more interest. One of these is by introducing a genetic block 
to the synthesis of homocysteine. Mutant strain H-98 blocks the terminal 
step in the synthesis of homocysteine. In the double mutant — sulfonamide- 
requiring, homocysteineless — there is no interference with the availability of 
threonine for growth, since the deleterious reaction does not take place in 
the absence of homocysteine. In the absence of homocysteine, however, there 
can be no synthesis of methionine, so that the double mutant fails to grow 
because of a methionine deficiency. The double mutant will grow if supplied 
with exactly the right amount of methionine — more inhibits growth, because 
methionine is degraded to homocysteine which then supports the deleterious 
reaction (Zalokar, 1950). 

The remaining method is to introduce a genetic block to the synthesis of 
/»-aminobenzoic acid. In the double mutant — sulfonamide-requiring, amino- 
benzoicless — there is again no interference with the utilization of threonine 
since there is no />-aminobenzoic acid to catalyse the deleterious reaction. 
There is again a deficiency for methionine, because />-aminobenzoic acid is 
needed in its synthesis. There is also a deficiency of />-aminobenzoic acid for 
other essential processes. The double mutant will grow if supplied just the 
right amount of /(-aminobenzoic acid to satisfy the essential requirements, 
but not enough to stimulate the deleterious reaction (Zalokar, 1948). 

Model Heterocaryons 

It can be seen that the simple sulfonamide-requiring mutant on the one 
hand, and the two double mutants on the other, have different deficiencies. 
One produces methionine and ^-aminobenzoic acid, but not enough threo- 
nine. The others produce sufficient threonine, but no methionine, and in one 
case, no ^-aminobenzoic acid. In heterocaryons between the simple and 
double mutants, the two types of nuclei should complement each other in 
the production of essential growth substances. If the nuclear ratios can be so 
adjusted that the different substances are produced in appropriate amounts, 
vigorous growth should result. Heterocaryons involving the simple sulfona- 
mide-requiring mutant and the double mutant sulfonamide-requiring, amino- 
benzoicless have resulted in vigorous growth (Emerson, 1948) in every test 
so far made. Growth curves of some of these heterocaryons are illustrated 
in Figure 12.6. 

Growth of these heterocaryons is usually not maintained at a constant 
rate. Growth may stop completely after a time, or it may nearly stop and 
then start again. This is believed to be due to fluctuations in the ratio of the 
two kinds of nuclei in the advancing hyphal tips. Apparently there must be 
many times as many double mutant nuclei as simple sulfonamide-requiring 
nuclei to result in a favorable combination. This is not surprising since the 




'— I 1) C 

oB I 

« <U r- 

[1- X! 00 



sy3i3Piiiim 



212 STERLING EMERSON 

sulfonamide-requiring strain produces something in the order of one hun- 
dred times as much />-aminobenzoic acid as is required for essential reactions, 
or about fifty times as much as is required for the reaction which makes 
threonine unavailable for growth. 

Limited direct tests of nuclear frequencies in such heterocaryons indicate 
that nuclei carrying only the sulfonamide-requiring gene are much less fre- 
quent than those carrying the aminobenzoicless gene as well. In one test of 
about one hundred nuclei, all proved to be double mutants. In another test, 
conidia from heterocaryons were transferred to fresh growth tubes which 
contained a concentration of sulfanilamide sufficient to inhibit growth of the 
double mutant very strongly and still be favorable to the growth of the simple 
sulfonamide-requiring mutant. Only one of five such transfers grew — again 
suggesting that simple sulfonamide-requiring nuclei were infrequent. 

If in order to have rapid growth there must be many double mutant nuclei 
and few simple mutants, it is not surprising that vigorous growth should 
cease rather suddenly. Ryan, Beadle, and Tatum (1943) have shown that 
growth substances can be transported for a distance of about one centimeter 
in the mycelium of Neurospora. One sulfonamide-requiring nucleus at a dis- 
tance of about a centimeter from the tip might supply enough ^-aminoben- 
zoic acid for the growth of that tip. But as the tip grows, that nucleus might 
easily be left behind. A deficiency of />-aminobenzoic acid would then de- 
velop in the tip, and growth would be arrested unless a nucleus of the proper 
constitution happened to migrate into the tip. 

Attempts to obtain rapidly growing heterocaryons involving the sulfona- 
mide-requiring mutant and the sulfonamide-requiring, homocysteineless 
double mutant were unsuccessful. It may be that it is impossible to have a 
nuclear ratio which will produce sufl&cient, but not too much methionine, 
and at the same time sufficient threonine for the requirement of the hetero- 
caryon. 

Interpreting Suppressor Heterocaryosis Based on Model Experiments 

The heterocaryons between the sulfonamide-requiring mutant and its 
double mutants with aminobenzoicless and homocysteineless were set up as 
models which should duplicate the behavior observed in the sulfonamide- 
requiring strain when suppressor mutations occurred, provided the interpre- 
tation placed on them was correct. For this purpose, the results obtained 
were gratifying. We should like to know just where each of the suppressor 
mutations studied fits into the biochemical scheme, but at present it can be 
shown only that they fit in a general way. 

Four suppressors in the first lot of six (those illustrated in Fig. 12.4), 
which are the only ones that have been studied in any detail at all, appar- 
ently represent mutation at four different loci, though almost no direct tests 



BIOCHEMICAL MODELS OF HETEROSIS IN NEUROSPORA 213 

for allelism are available. The inference that they are distinct genes is based 
on the data summarized in Table 12.2. 

The reactions controlled by the suppressor genes have not been identified. 
Suppressor SA is stimulated in growth by additional /)-aminobenzoic acid, 
and is inhibited considerably by sulfanilamide at concentrations twenty 
times less than that required to inhibit wild type. It is possible that a de- 
ficient amount of />-aminobenzoic acid is produced by this mutant, which 
would make it approximate the condition in one of the model heterocaryons. 
Growth of suppressor S-2 is somewhat stimulated by sulfanilamide (Fig. 
12.4) and by threonine, in this respect resembling the sulfonamide-requiring 
mutant which it "suppresses." It is even more stimulated by the purine, 

TABLE 12.2 

EVIDENCE SUGGESTING THAT SUP- 
PRESSORS 5i, S2, Si, AND Se ARE 
DIFFERENT GENES 



Suppressor 


Second 

Division 

Segregation 


Relation 
to 1633 


Genetically 
Independ- 
ent of 


Si 


25% 

50% 

0% 

60% 


none 

allele ? 

none 

none 




52 




Si 

s, 


s. 

Si 



adenine, as shown by the growth curves in Figure 12.7. It was previously 
known that in the presence of methionine, adenine reduces the normal re- 
quirement for />-aminobenzoic acid to about one-tenth its usual value. This 
suggested that the production of adenine also requires /)-aminobenzoic acid. 
The reaction controlled by this suppressor may thus be closely related to 
that controlled by the sulfonamide-requiring gene. No clues have turned up 
to indicate how the reactions governed by the remaining suppressor muta- 
tions may be related to these. 

In the living cell of Neurospora the reactions which are influenced in one 
way or another by the amount of available /»-aminobenzoic acid must be fairly 
numerous. The production of adenine and methionine requires the presence 
of this vitamin as does the reaction in the sulfonamide-requiring mutant 
which makes threonine unavailable. 

Strauss (1950) has studied a mutant strain (44602) which requires pyri- 
do.xine unless grown at high pH with ammonia as nitrogen source. He found 
that under the latter conditions it is inhibited by methionine, and that this 
inhibition is reversed by sulfanilamide, as if />-aminobenzoic acid were re- 
quired for the inhibition. Still another interrelationship has been found by 
Shen (1950) in studies of a mutant strain (84605) which requires sulfur in a 



214 



STERLING EMERSON 



form at least as reduced as thiosulfate. At 35° it has no other requirement, 
but at 25° it needs reduced sulfur, generally supplied as the amino acid 
cysteine, and also tyrosine. When methionine is supplied as the source of sul- 
fur at 25°, growth is strongly inhibited by choline. Under these conditions, 
choline does not inhibit at 35°, but there is an unexpected stimulation in 
growth by ^-aminobenzoic acid at that temperature. 

Mutant strains have been reported on two occasions which require either 
choline or />-aminobenzoic acid — choline may be the source of the methyl 



to 
ce 




HOURS 

Fig. 12.7 — Growth curves of suppressor mutant strain S-2 on minimal medium, on threo- 
nine (5 mg/100 ml), on methionine, and on purines (5 mg/100 ml each adenine sulfate and 

guanine hydrochloride) at 35°. 

group of methionine. Strehler (1950) has reported a strain which requires 
either methionine or /?-aminobenzoic acid. There is also a suggestion that 
/>-aminobenzoic acid may be involved in the metabolism of lysine. In Neuro- 
spora this is suggested only because the double mutants between the sul- 
fonamide-requiring strain and two different mutants which are unable to 
synthesize lysine do not grow on any combination of growth factors we have 
tried. In bacteria a strain has been found which requires either lysine or 
/»-aminobenzoic acid as a growth factor (Koft el al., 1950), strengthening the 
supposition of a similar interrelationship in Neurospora. 

These observations are referred to at this time because they indicate that 
there are a large number of metabolic reactions that are in one way or an- 
other related to the availability of />-aminobenzoic acid. These reactions 
must themselves be interrelated in the sense that an upset in one of them 



BIOCHEMICAL MODELS OF HETEROSIS IN NEUROSPORA 215 

may have a strong effect on one or more of the others, possibly through 
changing the availability of /?-aminobenzoic acid or a derivative. The model 
heterocaryon experiments described earlier show that it is possible for one 
mutation to cause an upset in one reaction and thus be detrimental to growth, 
and for a second mutation to restore conditions favorable to growth by actu- 
ally interfering with a different reaction which is itself essential to growth, 
but which is interrelated with the first reaction. In the reactions related to 
the metabolism of ^-aminobenzoic acid, there is sufficient complexity to ac- 
count for the occurrence of a large number of different suppressors of the sul- 
fonamide-requiring character. 

DISCUSSION 

It has been shown that increased vigor can result from heterocaryosis in 
which the two kinds of nuclei differ by only one pair of alleles. This may be 
true only under very special conditions such as have been present in the 
examples discussed. On the other hand, it is possible that the necessary con- 
ditions may be met with rather frequently in Neurospora, as suggested by 
the following examples. 

In mutant strains which have specific requirements for particular amino 
acids, it is commonly found that their growth is inhibited by the presence of 
other amino acids which do not ordinarily interfere with growth. Some mu- 
tants which require an outside source of threonine are strongly inhibited by 
methionine, (Teas, Horowitz, and Fling, 1948). Mutants specifically requir- 
ing lysine are inhibited by arginine (Doermann, 1944), and so on. In each of 
these instances, the inhibition by a particular amino acid is competitively 
antagonized by the specific amino acid required by the strain in question. 
The growth of these mutants should be favored by a reduction in the amount 
of the inhibiting amino acid, as would occur if some of the nuclei carried a 
genetic block to its synthesis. 

In extreme cases, the specific requirement for an amino acid may not re- 
sult from a failure in its synthesis, but from an oversensitivity to the in- 
hibiting amino acid. Thus, the sulfonamide-requiring strain can be said to 
be oversensitive to homocysteine in a way that leads to a requirement for 
threonine. One of the lysineless mutants (33933) seems to be oversensitive to 
arginine in much the same way. Heterocaryons having the lysineless gene in 
all nuclei, some of which also carry a genetic block to the synthesis of ar- 
ginine (from strain 36703), make considerable growth on minimal medium, 
whereas neither the lysineless nor the double mutant does (Fig. 12.8). 

Mary B. Mitchell (personal communication) recently observed that the 
stock cultures of certain lysineless mutants (4545, 15069, and 33933) had 
become less sensitive to inhibition by arginine. Tests of these showed that 
they were heterocaryons, some of whose nuclei were unchanged. Some car- 
ried mutant genes which lowered the sensitivity to arginine inhibition while 



216 



STERLING EMERSON 



leaving the requirement for lysine. These heterocaryons were more vigorous 
than the original lysineless strain, but no more vigorous than the pure double 
mutant strains extracted from the heterocaryons. 

In studies on reverse mutation in a leucineless strain (33757), Ryan and 
Lederberg (1946) found that heterocaryons, whose nuclei differed only in 





' 








;| 




•150 








/ 




. 








/ 2', 




• 






*> i 


/ f] 




• 






o / 


/''' 




■ 






TJ / 


/ 

'1 t 


ex: 


■100 






*» / 
4. / 


UJ 








■^/ 


•/ / 


h- 










/t/ 


UJ 










'p' 


-J 
_j 








• 
/ 


/ //J • 


S 








/ / 


/^ /^ 




50 




/ 










/ 




// 

y/ 






/y 


> 


*••/ 

^•^x 










,-' 


.y^ X 








, -- 


*• * 


-^l 






^^•-^ 


















^ .'y^ ^^% _ . 






50 






i6o 



HOURS 

Fig. 12.8 — Growth curves of heterocaryons between lysineless (/y, +) and lysineless, 
arginineless (/y, arg) strains of Neurospora at 35° on minimal medium. Curve 1 : heterocar- 
yon in which both nuclear types were of mating type A; curves 2 to 5: heterocaryons made 
up of nuclei of different mating types (/v, +, A and /v, arg, a) — cf. Beadle and Coonradt 

' (1944). 



that some carried the wild type allele and some the mutant allele of the 
leucineless gene, almost invariably reverted to the homocaryotic condition. 
By the time growth had reached the end of a tube containing minimal me- 
dium, nothing but wild type nuclei remained. In tubes containing limiting 
concentrations of leucine, nothing but leucineless nuclei were present after a 
short period of growth. This was under conditions where the growth rate of 
the leucineless strain is considerably less than that of wild type. Under both 
of these conditions, the heterocaryon is at a strong disadvantage compared 
to its components. It is not known whether or not there is a particular con- 
centration of leucine which would favor the heterocaryon. 



BIOCHEMICAL MODELS OF HETEROSIS IN NEUROSPORA 217 

Houlahan and Mitchell (1948) have studied the interactions of mutant 
strains involved in the metabolism of i)yrimidines and lysine. A pyrimidine- 
less mutant (37301) has a specific requirement for pyrimidine. There is a 
suppressor of this mutant which enables it to grow without added pyrimidine, 
unless arginine is also added, whereupon the pyrimidine requirement is re- 
stored. One lysineless strain (33933) can utilize a-amino adipic acid in place 
of lysine. As a double mutant with the pyrimidine suppressor, it can still 
use a-amino adipic acid, but requires four times as much as the simple lysine- 
less strain unless small amounts of arginine, or an arginine precursor, are 
added. The double mutant combining this lysineless with the pyrimidineless 
mutant is unable to use a-amino adipic acid unless a small amount of lysine 
is added — arginine is ineffective in this instance. A second lysineless mutant 
(454.5), which has a specific requirement for lysine and which secretes pyrimi- 
dines into the medium, behaves in a predictable fashion as a double mutant 
with pyrimidineless, or its suppressor, but not as the triple mutant lysineless, 
pyrimidineless, suppressor of pyrimidineless. Instead of requiring only lysine 
for growth, this triple mutant also requires pyrimidines and arginine. This 
example is cited as another in which metabolic interactions may be as com- 
plex as in those discussed earlier which depend in one way or another on 
/>-aminobenzoic acid. 

Applicability to Classical Heterosis 

Observations relating to one-gene heterosis in higher plants are discussed 
in other papers in this series (Crow, Hull, Jones, and Whaley). Studies of 
Neurospora heterocaryons have shown that a very similar phenomenon oc- 
curs under certain special physiological conditions. In a particular genetic 
background, the amount of an essential metabolite normally produced has 
deleterious consequences which are removed by reducing the dosage of a 
gene responsible for the production of that metabolite. This reduction was 
brought about through heterocaryosis in the studies reported, but it should 
also result from heterozygosis under similar physiological conditions. There 
is nothing in the studies of heterocaryosis in Neurospora to suggest that one- 
gene heterosis is of general occurrence and importance, or that other examples 
should have similar biochemical backgrounds. 



TH. DOBZHANSKY 

Columbia University 



Chapter 13 

Nature and 
Origin of Heterosis 



Exploitation of heterosis in cultivated plants and animals is to date by far 
the most important application of the science of genetics in agricultural prac- 
tice. It is therefore unfortunate that few of the studies so far made on 
heterosis go beyond crudely empirical observations and descriptions and 
that little effort is being made to understand the underlying causes of the 
phenomena involved. Such an understanding is needed particularly because 
the advances of general genetics make it evident that several quite distinct, 
and even scarcely related, phenomena are confused under the common label 
of heterosis or hybrid vigor. 

In what follows, an attempt is made to indicate briefly what seem, to the 
writer, promising lines of approach to a classification and study of the various 
kinds of heterosis. The tentative nature of the classification here suggested 
is fully realized. But it is believed that this classification may nevertheless 
serve a useful function if it directs the attention of the students of heterosis 
to factors which are only too often overlooked. 

MUTATIONAL EUHETEROSIS 

Perhaps the simplest kind of true heterosis — euheterosis — is that which 
results from sheltering of deleterious recessive mutants by their adaptively 
superior dominant alleles in populations of sexually reproducing and cross- 
fertilizing organisms. 

Although only a small fraction of the existing species of organisms have 
been investigated genetically, it is reasonable to assume that mutational 
changes arise from time to time in all species, albeit at different rates. Fur- 
thermore, a great majority of the mutations that arise are deleterious, and 
lower the fitness of their carriers to survive or to reproduce in some or in all 

218 



NATURE AND ORIGIN OF HETEROSIS 219 

environments. This deleterious character of most mutations seems surpris- 
ing, especially because in modern biology the process of mutation is regarded 
as the source of the raw materials from which evolutionary changes are con- 
structed. 

A little consideration shows, however, that the ada])tively negative char- 
acter of most mutations is by no means unexpected. Indeed, since every mu- 
tation has a finite probability to occur in any generation, the mutants which 
we observe in our fields and laboratories must have arisen many times in the 
history of the species. The rare mutants which confer adaptive advantages 
on their possessors in the environments in which the species normally lives 
have had the chance to become established in the species populations as 
components of the normal species genotype. In a more or less static environ- 
ment, the genotypes of most species are close to the upper attainable level of 
adaptedness. 

The above argument may seem to prove too much. In the absence of use- 
ful mutants, evolution would come to a standstill. The paradox is resolved 
if we recall that the environment is rarely static for any considerable periods 
of time. Furthermore, most living species occur not in a single but in several 
related environments. Genotypes which are adaptively valuable in a certain 
environment may be ill adapted in other environments, and vice versa. It 
should be possible then to observe the occurrence of useful mutations if we 
place the experimental organisms in environments in which their ancestors 
did not live. 

Progressive improvement of domesticated animals and plants in the hands 
of breeders constitutes evidence that useful mutations do occur. The genetic 
variants which are being made use of by breeders have arisen ultimately 
through mutation. These mutations have been arising from time to time, be- 
fore as well as after the domestication. But while they were deleterious in 
the wild state, some of them happened to be suitable from the standpoint of 
the breeders. They were useful in the man-made environment or they were 
useful to man. Favorable mutations can be observed also in wild species, 
provided that the latter are placed in unusual external or genetic environ- 
ments. This has been demonstrated in experiments of Spassky and the writer 
on Drosophila pseudoobscura. Several laboratory strains of this fly were sub- 
jected to intense selection for fifty consecutive generations, and improve- 
ments of the viability have been observed in most of them. 

Many, perhaps most, deleterious mutants are nearly or completely reces- 
sive. Others are more or less dominant to the "normal," or ancestral, state. 
The fate of the dominant deleterious mutants in jwpulations of sexually re- 
producing and cross-fertilizing species is different from that of the recessives. 
By definition, deleterious mutants in wild species lower the fitness of their 
carriers to survive or to reproduce, and in cultivated species impair the 
qualities considered desirable by the breeders. Natural and artificial selec- 



220 TH. DOBZHANSKY 

tion will consequently tend to lower the frequency, or to eliminate deleteri- 
ous mutants. 

Selection against a dominant deleterious mutant is, however, a far more 
efficient process than that against a recessive mutant. This is because dele- 
terious recessive mutant genes are sheltered from selection by normal domi- 
nant alleles in heterozygotes. Deleterious dominants are eliminated by selec- 
tion within relatively few generations after their origin. Deleterious reces- 
sives accumulate in heterozygotes until their frequencies become so high that 
recessive homozygotes are produced. Dominant alleles are not intrinsically 
beneficial, and recessives are not necessarily deleterious. But at any one time, 
we find in cross-fertilizing populations more deleterious recessives than dele- 
terious dominants, because the former are not eliminated by selection as 
promptly as the latter. 

Analysis of wild populations of several species of Drosophila has revealed 
extensive infestation of the germ plasm by deleterious recessive mutant genes. 
According to the unpublished data of Pavan and collaborators, 41 per cent 
of the second chromosomes in Brazilian populations of Drosophila willistvni 
are lethal or semilethal when homozygous. Among the remainder, 57 per 
cent are sublethal when homozygous. Furthermore, 31 per cent of the second 
chromosomes make the homozygotes completely sterile in at least one sex, 
32 per cent retard the development, and 16 per cent cause various visible 
abnormalities. Comparable figures for the third chromosomes are 32 per 
cent of lethals and semilethals, 49 per cent subvitals, 28 per cent steriles, 
36 per cent retarded, and 16 per cent containing visible mutants. Since 
every fly has two second and two third chromosomes, it is easily seen that a 
great majority of individuals in Brazilian populations carry several deleteri- 
ous variants in heterozygous condition. 

The mass of deleterious recessives carried in normally breeding natural 
populations has no disastrous effects on the average fitness of members of 
such populations. This is because the frequency of recessive homozygotes 
found in a population at equilibrium is equal to the number of the corre- 
sponding recessive mutants that arise in every generation. The loss of fitness 
caused in a normally breeding population by dominant and by recessive mu- 
tants is thus proportional to the frequency of the origin of these mutants by 
mutation. 

The situation changes completely if a normally crossbred population is 
subjected to inbreeding. For inbreeding renders homozygous many reces- 
sives that would remain sheltered in heterozygotes under normal crossbreed- y 
ing. These recessives become suddenly exposed to natural, or to artificial, 
selection. The loss of fitness in inbred lines of normally cross-fertilized species 
is the consequence. Conversely, the heterosis observed in the progeny of 
intercrossed inbred lines is the outcome of restoring the normal reproductive 
biology and the normal population structure of the species. 



NATURE AND ORIGIN OF HETEROSIS 221 



BALANCED EUHETEROSIS 



Balanced heterosis is due to the occurrence of a rather special class of 
mutations and gene combinations, which confer on heterozygotes a higher 
adaptive value, or a higher agricultural usefulness than is found in the cor- 
responding homozygotes. 

The conditions most frequently found in heterozygotes are either domi- 
nance and recessiveness, when the heterozygote is more or less similar to one 
of the homozygotes, or phenotypical intermediacy between the homozygotes. 
A heterozygote may, however, be in some respects phenotypical ly more ex- 
treme than either homozygote. Thus, a heterozygote may be more viable, 
more productive, or otherwise exceed both homozygotes in some positive or 
negative quality. This condition is sometimes spoken of as overdominance 
(Hull). 

Although overdominance is, by and large, an exceptional situation, it is of 
particular interest to a student of population genetics, and especially to a 
student of heterosis. Suppose that a certain gene is represented in a popula- 
tion by a series of alleles, A\ A^,A^ . . . which are deleterious in homozygous 
condition, AKi\ A-A^, A^A^ . . . , but which show a relatively higher fitness 
in heterozygotes A^A'^, A'^A^, A~A^ . . . , etc. Natural or artificial selection 
would preserve in the population all the variants A\ A", A^ . . . , regardless 
of how poorly adapted the homozygotes may be. In fact, one or all homozy- 
gotes may be semilethal or even lethal, and yet selection will establish an 
equilibrium at which every one of the variants will be present with a definite 
frequency. This equilibrium can easily be calculated if the selective dis- 
advantages of the homozygotes, compared to the heterozygotes, are known. 
The resulting situation is referred to as balanced polymorphism. 

Balanced polymorphism may be produced by mutations in single genes, 
provided that the heterozygotes exhibit overdominance in fitness in some 
environments. This has been demonstrated, among others, by Gustafsson 
and Nybom. They observed several mutations in barley that were deleterious 
in homozygotes, but produced heterozygotes superior to the ancestral "nor- 
mal" homozygotes. Ford and others showed that certain color variants in 
butterflies, which are inherited as though caused by a single genetic change, 
are maintained in natural populations by the same mechanism. 

Detailed data are available on balanced polymorphism in several species 
of Drosophila, in which natural populations are very often polymorphic for 
gene arrangements in some chromosomes. These gene arrangements differ in 
inversions of blocks of genes. Thus, in certain populations of Drosophila 
pseiidoobscura from Southern California, at least 70 per cent of the wild indi- 
viduals are inversion heterozygotes. In populations of Drosophila willisloni 
from central Brazil (Goyaz), an average individual is heterozygous for as 
many as nine inversions, and very few individuals are homozygous. 



222 TH. DOBZHANSKY 

Now, it has been shown by observation both on natural and on experi- 
mental populations of some Drosophila species, that the heterozygotes for 
the naturally occurring inversions possess considerable adaptive advantages 
over the homozygotes. For example, taking the adaptive value of the 
heterozygotes for ST and CH inversions in Drosophila pseudoobscura to be 
unity, the adaptive values of the ST/ST and CH/CH homozygotes are 
about 0.8 and 0.4 respectively. Further, it has been shown that the heterosis 
in the ST/CH heterozygotes occurs only if the constituent chromosomes are 
derived from the same population, or from populations of nearby localities. 
Chromosomes with the same gene arrangements, ST and CH, derived from 
remote localities (such as Central and Southern California, or Southern 
California and Mexico) exhibit little or no heterosis. 

This finding is most compatible with the assumption that the over- 
dominance in fitness observed in the heterozygotes is the property not of a 
single gene locus, or of a chromosome structure, but rather of integrated sys- 
tems of polygenes. Such polygenic systems are coadapted by natural selec- 
tion to other polygene complexes present in the same populations. The role 
of the chromosomal inversions in the formation of the heterotic state of bal- 
anced polymorphism is due to the suppression of crossing over caused by 
most inversions, at least in Drosophila. Elimination of crossing over prevents 
the breakup of the adaptively integrated polygene complexes which are 
carried in the chromosomes involved. 

It should be noted that adaptively integrated polygene complexes can be 
maintained in crossbreeding populations with the aid of genetic mechanisms 
other than chromosomal inversions. Any factor which restricts or prevents 
crossing over in chromosomes, or parts of chromosomes, can accomplish the 
same biological function. Localization of chiasmata may be such a factor. 
If, for example, chiasmata are found chiefly or exclusively at some definite 
points in a chromosome, the genes carried in the sections which intervene 
between these points are inherited in blocks. Such gene blocks may act 
exactly as gene complexes bound together by inversions. 

Balanced heterosis differs profoundly from mutational heterosis. The 
latter is due simply to the sheltering of deleterious recessive mutants by 
their dominant alleles. Balanced heterosis is a result of overdominance. Mu- 
tational heterosis is a protective device of a sexual species with a certain 
population structure against the mutation pressure. Balanced heterosis is an 
evolutionary contrivance that permits maintenance in a population of a mul- 
tiplicity of genotypes that may be adaptive in dififerent ecological niches 
which the population occupies. 

LUXURIANCE 

Mutational and balanced heterosis resemble each other in one important 
respect — both are normal adaptive states attained in outbred sexual species 



NATURE AND ORIGIN OF HETEROSIS 223 

as a result of an evolutionary history controlled by natural or by artificial 
selection. The normal heterotic state can be disru])ted by sudden inbreeding, 
which is evidently a disturbance of the reproductive biology to which the 
species is adjusted. The heterotic state can also be restored by intercrossing 
the inbred lines. This is true heterosis, or euheterosis. Euheterosis is a form 
of evolutionary adaptation characteristic of sexually reproducing and cross- 
fertilizing species. 

Numerous instances are known, however, when hybrids between si)ecies, 
neither of which can be regarded as inbred, are larger, faster growing, or 
otherwise exceeding the parental forms in some quality. Similar luxuriance is 
observed in some hybrids between normally self-fertilizing species, races, or 
strains. This kind of luxuriance of hybrids cannot be ascribed to sheltering of 
deleterious recessive mutants, because the latter are sheltered in the parental 
populations. It is also unlikely to arise from overdominance since, at least in 
wild species, natural selection would be expected to have induced such bal- 
anced heterosis in the parental species or strains. 

Luxuriance is, from the evolutionary standpoint, an accidental condition 
brought about by complementary action of genes found in the parental form 
crossed. Two sets of facts are important in this connection. First, in cases of 
luxuriance there is usually no indication whatever that the luxuriant hybrids 
would prove adaptively superior in competition with the parental forms in 
the natural habitats of the latter. Second, luxuriance appears to be more 
frequently encountered in domesticated than in wild species. 

It stands to reason that increase in body size, or in growth rate, is by no 
means always an adaptively superior change. To equate size with vigor, fit- 
ness, or adaptive value would be a height of anthropomorphic naivete. The 
rate of growth and the size attained by an organism in its normal environ- 
ments are evidently controlled by natural selection. Excessive as well as de- 
ficient sizes are adaptively about equally disadvantageous. The checks upon 
excessively rapid growth and excessive size are, however, very often relaxed 
under domestication. In man-controlled environments those qualities often 
become desirable from the standpoint of the breeder if not from that of the 
organism. Luxuriance is, really, pseudoheterosis. 



DONALD F. JONES 

Connecf/'cuf Agricultural Exper'imenf Sfaiion 



Chapter 14 

Plasmagenes and 
Chromogenes in Heterosis 



The word heterosis is essentially a contraction of the phrase stimulus of 
heterozygosis. It was first used by G. H. Shull (1914). The concept of a 
stimulation resulting from the genetic union of unlike elements was de- 
veloped by East (1909). Previous to the Mendelian conception of units of 
heredity, it was generally considered by plant and animal breeders that the 
invigorating effect of crossing unlike varieties of plants and breeds of live- 
stock was due to the correction of imperfections that existed in both parental 
types. This idea is clearly stated by Samuel Johnson in the second edition of 
his book How Crops Grow (1891). 

The early recordings of instances of hybrid vigor and the various means 
of accounting for this phenomenon have been stated and restated so many 
times that there is no need or useful purpose in repeating them here. Excel- 
lent reviews of the literature are readily available (see especially East and 
Hayes, 1912; Jones, 1918; East and Jones, 1919; East, 1936; and Whaley, 
1944). 

THE EXPRESSION OF HETEROSIS 

At the present time, the term heterosis designates the increased growth or 
other augmented action resulting from crossing, however it is produced. As 
generally used, it is essentially synonymous with hybrid vigor. Heterosis has 
two general modes of expression. In one, there is an increase in size or num- 
ber of parts. This is usually the result of a greater number of cells and a faster 
rate of cell division and cell activities. This results in an improvement in gen- 
eral well-being of the organism similar to the result of being placed in a more 
favorable environment. Such luxuriance may be accompanied by partial or 
complete sterility in diverse crosses. 

A somewhat different manifestation of heterosis is an increase in bio- 

224 



PLASMAGENES AND CHROMOGENES IN HETEROSIS 225 

logical efficiency, such as reproductive rate and survival ability. This may 
even be shown with a reduction in productiveness as measured by economic 
characters. Some confusion has arisen by not distinguishing clearly between 
these two different manifestations of heterosis. 

In addition to these two general types of heterotic effects, there may also 
be a reduction in both growth and survival ability; in other words, hybrid 
weakness or a reversed or negative heterosis. This effect is much less com- 
mon and is seldom found in cultivated plants and domesticated animals. 

TYPES OF GENE ACTION 

An understanding of the mode of action of heterosis has now resolved into 
a study of the nature of gene action. The genes usually used to illustrate 
Mendelism are the loss variations that have a major effect such as the inabil- 
ity to produce some essential substance. This results in a block in the normal 
chemical processes, finally resulting in an individual of greatly altered ap- 
pearance, size, or ability to survive. The effect ranges in intensity from a com- 
pletely lethal condition at some stage of development, up to individuals that 
differ only slightly in appearance from normal with no appreciable reduction 
in growth or survival ability. Such genes are illustrated by the long lists of 
Mendelizing characters now tabulated for maize, Drosophila, mice, and many 
other animals, plants, and lower organisms. 

DOMINANT AND RECESSIVE GENES 

In these cases, the normal allele is usually designated by a capital letter, 
with the mutant, deficient allele denoted by the corresponding lower case 
letter. In comparison with the normal allele, the recessive mutants are de- 
ficient in some respect. In their inability to produce certain specific sub- 
stances, as shown in the haploid Neurospora by Beadle and his co-workers, 
they are referred to as .4 -less, 5-less, C-less, etc. In diploid organisms A is 
usually completely dominant over a; that is, one A allele functions as well 
or nearly as well as two. 

There is no question that the accumulation in a hybrid of the normal 
alleles of this type results in heterosis. In the simplest example of a cross of 
^-less by B-less (aaBB XAAbb) the hybrid offspring are all AaBb, and 
essentially normal for whatever effect A and B have. But since the mutant 
recessive alleles of this type are so drastic in their effect, most of these 
deficiencies are removed by natural selection in all species whether self- 
fertilized or cross-fertilized. Therefore they have little part in the heterosis 
that is shown by these organisms when crossed. Furthermore, genes of this 
type are eliminated when naturally cross-fertilized species, such as maize, 
are artificially self-pollinated. Yet such inbred strains show the largest 
amounts of heterosis. 

There is evidence, as will be shown later, that there are many genes of this 



226 DONALD F. JONES 

type having small effects that are not eliminated by natural or artificial 
selection either in the wild or under domestication, and that these deficiencies 
or degenerative mutants do have a large part in bringing about reduced 
growth. Before presenting this evidence, there are other types of gene action 
that should be considered. 

CHROMOSOMAL DELETIONS 
In addition to the recessive mutant alleles that are deficient as compared 
to their normal alleles, there are also chromosomal deletions which result in 
the complete elimination of the normal locus. Large deletions are usually 
lethal and are quickly eliminated. Small deletions that cannot be detected 
cytologically are haplo-viable, and may persist indefinitely if they are closely 
linked with essential loci. Changes of this type have been demonstrated by 
McClintock (1931) and by Stadler (1933). They cannot be readily distin- 
guished from recessive mutants of the .4 -less type. In fact there may be no 
difference. In practically all cases they show varying amounts of germ cell 
abortion, and do not mutate back to normal. Deletions of this type are 
designated Ao. 

DOMINANT UNFAVORABLE GENES 

In many cases of deletion the heterozygote, or the hemizygote, is visibly 
and unfavorably altered from normal, in which event the genes involved are 
listed as dominant, and if partially viable they can bring about negative 
heterosis or hybrid weakness. It is not known whether all dominant unfavor- 
able genes are deletions of this type, but as far as their effect on heterosis is 
concerned it makes little difference whether or not they are. An illustration 
of this type of gene action may be seen in a cross of Ragged and Knotted 
maize plants. Both of these genes result in a marked reduction in growth in 
the heterozygous condition. They are not completely lethal in the homozy- 
gous dominant condition, but seldom produce seed or pollen. When both 
dominant genes are present together in the heterozygous condition, there is 
a marked reduction in size, rate of growth, and reproductive ability as com- 
pared with either parental type. 

Tunicate, teopod, and corn grass are also dominant genes that reduce 
grain yields in both the homozygous and heterozygous condition. They are 
probably reversions to a primitive condition which in suitable genetic com- 
binations maybe favorable to survival in the wild. Dunn and Caspari (1945) 
describe many structural abnormalities in mice that seem to be due to dele- 
tions having a dominant effect in the hemizygote. Some of these counteract 
each other and tend to restore a more normal condition, while others accumu- 
late unfavorable effects. A similar situation has been reported in Drosophila 

by Stern (1948). . 

In addition to recessive deletions with a dominant effect m the hetero- 
zygote, there are also dominant inhibitors that have no indication of being 



PLASMAGENES AND CHROMOGENES IN HETEROSIS 227 

deletions, but do prevent other genes from liaving their usual exi)ression. 
Most of these inhibitors control color characters and are usually not involved 
in heterosis. If they were, there would be more negative heterosis than actual- 
ly is found. 

GENES WITHOUT DOMINANCE 

Unlike the visible Mendelizing genes with their clear-cut dominance and 
unfavorable action of one or the other allele, there are many genes that dif- 
ferentiate size or number of parts, time of flowering and maturing. These are 
the genes usually involved in normal variation. They are the ones the plant 
and animal breeder are mainly concerned with and could expect to have a 
major effect on heterosis. Since neither member of an allelic pair can be con- 
sidered abnormal or deficient, both are designated with a capital letter with 
some prefix to differentiate them, as for example A and A\ 

Genes of this type usually have simple additive effects such as the F endo- 
sperm color gene in maize, in which each allele adds a definite increment in 
total carotene content. Such additive genes without dominance are used to 
interpret the inheritance of quantitative characters which have been shown 
to segregate and recombine in a Mendelian manner. 

No clear distinction can be made between the A a and .4^1' types of genes 
and this has led to much confusion. The first class shows complete or nearly 
complete dominance. The second shows no dominance or very little domi- 
nance, but one type integrates into the other. The principal question at issue 
is whether either type shows over-dominance, or in other words, an interac- 
tion between alleles such that Aa > A A or aa or AA^ > A A or .4 '.4'. Before 
considering the evidence for or against over-dominance, two remaining types 
of genes should be considered. 

CHROMOSOMAL REARRANGEMENTS 

By chromosomal rearrangements such as inversions and translocations, 
genes without alteration are placed in different spatial relations with other 
genes. In their altered position they have different effects. Dobzhansky and 
his associates have studied many geographical races of Drosophila that differ 
by chromosomal rearrangements. Crosses between these chromosomal types 
from the same region exhibit heterosis, whereas the same chromosomal type 
from different regions do not show such a high degree of heterosis. This 
seems not to be a position effect, but is the result of an accumulation of gene 
differences that are protected from random distribution by the prevention of 
crossing over in hybrids of different chromosomal types. 

COMPOUND GENES AND GENES WITH MULTIPLE EFFECTS 
In many organisms, loci are known which have different effects on differ- 
ent parts of the organism. In maize the .1 , P, and R genes have been studied 
in considerable detail by Stadler and his co-workers. These loci each have a 



228 DONALD F. JONES 

series of alleles that produce characteristic color patterns and intensities of 
colors in different parts of the plant such as culm, leaf sheath, leaf blade, 
glumes, anthers, silks, cob and pericarp, and endosperm. They may be con- 
sidered either as genes located so closely together that they never show- 
crossing over, or compound genes with multiple effects. Without going into 
the evidence for or against these two hypotheses, it is obvious that compound 
genes can have an important part in heterosis if they control growth proc- 
esses. More information is needed on the specific effect of compound genes. 

In Godetia a series of multiple alleles has been described by Hiorth (1940) 
that is often cited as an illustration of an interaction between alleles produc- 
ing an effect analogous to heterosis. Actually these are color determiners that 
control pigment production in different parts of the flower quite similar to the 
A, P, and R loci in maize. Each allele has a different manifestation, and all 
tend to accumulate color in the heterozygotes. 

The familiar notation of a chromosome as a linear arrangement of loci, 
each of which is the site of a single gene with one effect function, is probably 
an oversimplification of the actual condition. It is difficult to see how an 
organism could have originated in this way. It is more likely that a chromo- 
some is an association of primitive organisms of varying types and functions. 
These primitive organisms found it to be an advantage in the evolutionary 
process to become associated in some such process as the colonial organisms 
now exhibit. This association has undergone very great modification and 
ramifications, but the compound genes may be vestigial structures of such 
an association, differing greatly in size, arrangement, and function. Many of 
them still retain some independence, and when removed from their normal 
position in the chromosome could function as plasmagene or viroid bodies. 

These compound genes may undergo mutation and possibly recombina- 
tion or reorganization within themselves, but crossing over takes place for 
the most part only between these compound structures. Compound genes 
also arise by unequal crossing over and duplication of loci are shown by the 
Bar eye gene in Drosophila and others of similar type. 

In addition to compound or multiple genes, there are single genes with 
multiple effects. Many of these are important in growth processes and are 
illustrated by chlorophyll production in maize studied by H. L. Everett 
(1949). One major gene is essential for the production of carotene. In the 
recessive condition the seeds are pale yellow in color, in a normal, dark yel- 
low seeded variety. The young seedlings grown from these pale yellow seeds 
are devoid of chlorophyll. The recessive allele is therefore lethal. By using the 
pale yellow endosperm as a convenient marker and crossing with a number of 
standard field corn inbreds, it has been found that these inbreds differ widely 
in their normal chorophyll mechanism. Many of them have genes that can 
restore normal chorophyll production without restoring the production of 
carotene in the seed. Other genes restore chlorophyll production only partial- 



PLASMAGENES AND CHROMOGENES IN HETEROSIS 



229 



]y (see Table 14.1). Hybrid combinations that bring these genes together are 
appreciably more efficient in chl()ro])hyll production than combinations that 
lack some of them. However one of these dominant alleles has a suppressing 
eflfect on chlorophyll development. The combination of all of these chloro- 
phyll genes so far studied is not the most productive. There are many genes 
of this type that block chemical syntheses, that are not lethal in the usual 
genetic assembly, but which combine to give a cumulative efficiency in most 
cases. 

Lethal genes which show complete dominance of the normal allele would 
have no effect on heterosis other than to reduce the number of offspring. Such 

TABLE 14.1 

GENES CONTROLLING CHLOROPH\TL PRO- 
DUCTION IN MAIZE* 



Ch 


Ch 


Ch 


Seed Color 


Chlor. Grade 


Viability 











Pale 


Albino 


Lethal 


— 


+ 


_ 


Pale 


Virescent 


Lethal 


— 




+ 


Pale 


Light green 


Normal 


— 


+ 


+ 


Pale 


Light green 


Normal 


+ 


— 


— 


Yellow 


Light green 


Normal 


+ 


+ 


— 


Yellow 


Med. green 


Normal 


+ 


+ 


+ 


Yellow 


Dark green 


Normal 


+ 




+ 


Yellow 


? 


Normal 



* Data from H. L. Everett. 



genes would be just as effective in the homozygous as the heterozygous con- 
dition. Genes that have any part in the type of heterosis that is manifested 
in increased growth must be viable and have some degree of dominance. In 
other words, Aa must be greater than | AA. Aa may even be greater in 
its effect than A A or aa in which case theoretically there is over-dominance, 
but very little specific evidence is available to show that such a situation 
actually exists. 

I can see no way in which it is possible to separate over-dominance from 
a stimulus of heterozygosis. They seem to be different ways of saying the 
same thing. The essential point at issue at the present time is whether or not 
over-dominance actually occurs, and if so, how important this is in the 
total amount of heterosis in addition to the known accumulation of favorable 
dominant effects. 

INTERACTION BETWEEN ALLELES 

Evidence has been presented from many sources bearing on the problem 
of over-dominance and interaction between different alleles. Much of the 
argument is based on mathematical treatment of data that require many as- 
sumptions. What is needed is more specific evidence where the effect of 



230 DONALD F. JONES 

multiple genes can be ruled out. Very few specific examples of single gene 
action are available. 

In one case studied by the writer there is clear evidence for an interaction 
between alleles (Jones, 1921). A mutation in a variety of normally self-ferti- 
lized tobacco changed a determinate plant into an indeterminate, non- 
flowering variation. It was a change in the normal response to the summer 
day length period. The mutant plants failed to flower in the normal growing 
season and continued in a vegetative condition. Reciprocal crosses between 
the mutant and normal types both grew at the same rate as the normal 
plants showing complete dominance of the normal growth rate. The hetero- 
zygous plants continued their vegetative growth longer and produced taller 
plants with more leaves and flowers than the normal homozygous plants. 
This result I consider not to be heterosis, since there was no increase in 
growth rate. It is merely an interaction between alleles to produce a result 
that is different from either parent. There are undoubtedly many allelic 
interactions of this type. Whether or not they can be considered to contribute 
to heterosis is largely a matter of opinion. 

Other cases in corn where heterosis resulted from degenerative changes 
(Jones, 1945) were at first assumed to be single allelic differences, since they 
originated as mutations in inbred and highly homozygous families. The de- 
generate alterations were expressed as narrow leaves, dwarf plants, crooked 
stalks, reduced chlorophyll, and late flowering. All of these mutant variations 
gave larger amount of growth in a shorter period of time and clearly showed 
heterosis. 

The further study of this material has not been completed, but the results 
to date indicate that the differences involved are not single genes. Both the 
extracted homozygous recessives and the extracted homozygous dominants 
from these crosses are larger than the corresponding plants that originally 
went into the crosses. 

This indicates quite clearly that the visible changes were accompanied or 
preceded by other changes with no noticeable effects, but which are expressed 
in growth rates. A more complete summary of these results will have to 
wait until all of the evidence is at hand. It is a simple matter to extract 
the homozygous recessives from these crosses, but it is difficult to extract 
the homozygous dominants. Many of the self-fertilized plants proved to be 
heterozygous. 

GENES CONTROLLING GROWTH 

Additional evidence that there are a large number of genes having small 
effects on growth without visible morphological changes is becoming clearly 
apparent from a backcrossing experiment now in progress. Several long 
inbred lines of corn, one of which is now in the forty-first generation of con- 
tinuous self-fertilization, were outcrossed to unrelated inbred lines having 
dominant gene markers which could be easily selected. The markers — red 



PLASMAGENES AND CHROMOGENES IN HETEROSIS 231 

cob, yellow endosperm, and non-glossy seedlings— were chosen because Ihey 
had little or no effect on growth of the plant. 

The first generation outcrosses showed the usual large increases in size of 
plant, time of flowering, and yield of grain that is expected in crosses of un- 
related inbred strains of corn. The hybrid plants were backcrossed as seed 
parents with pollen from the inbred with the recessive gene marker. In every 
generation, plants with the dominant gene marker were selected for back- 
crossing. These plants have now been backcrossed six successive times. Many 
progenies have been grown. They are all heterozygous for the gene marker 
plus whatever neighboring regions on the same original chromosome from 
the non-recurrent parent that have not been lost by crossing over. 

The plan is to continue the backcrossing until no measurable differences 
remain between the backcrossed plants and the recurrent parent, or be- 
tween the two classes of backcrossed individuals in the same backcrossed 
progeny, those with the dominant marker and those with the recessive 
marker. When the point is reached where no differences can be detected, the 
plan is to compare successive earlier generations from remnant seed to pick 
up whatever single gene differences there might be that could be measured 
and detected by their segregation. 

So far both classes of backcrossed plants in nearly all progenies are taller 
and flower earlier, showing that they have not been completely converged to 
the parental type (see Table 14.2). The differences are small and not statisti- 
cally significant in the tests so far made, but are nearly all in the direction of a 
heterotic effect. As yet there are not sufiicient data to base final conclusions. 
It is hoped that the comparison of the two classes of backcrossed progeny 
with the original recessive parent will permit a distinction between the favor- 
able action of dominant genes and an interaction between heterozygous 
alleles. Also that it may be possible to show whether or not there is any 
residual cytoplasmic effect, since some of the outcrossed plants have the same 
cytoplasm as the dominant gene marker and some do not. 

Important facts do stand out clearly from this experiment. Since heterosis 
still remains after these many generations of backcrossing, it shows clearly 
that these three chromosome regions selected as samples have an appreciable 
effect on growth. Since the gene markers themselves have no effect on 
growth, as far as this can be determined in other material, these three regions 
are random selections for growth effects. This indicates quite clearly that 
there are genes in all parts of the chromosomes that contribute to normal 
growth and development. While the evidence so far available does not per- 
mit a clear separation between the effects of an accumulation of favorable 
genes as contrasted to an interaction between alleles, or between genes and 
cytoplasm, the results show that there are many loci involved in the heterotic 
effect in addition to the dominant gene markers. 

This follows from the evidence at hand. If the heterosis now remaining 



232 DONALD F. JONES 

were due solely to the interaction between the dominant and recessive mark- 
ers, there would have been a rapid approach to the level of vigor now re- 
maining. If it were due to a larger number of genes distributed rather evenly 
along the chromosome, the reduction in heterosis would be gradual, as it 
has proved to be. Small amounts of heterosis may persist for a long time un- 
til all of the genes contributing to it are removed by crossing over. 

A recent experiment by Stringfield (1950) shows a difference in produc- 
tiveness between an F2 selfed generation and a backcross having the same 
parentage. The amount of heterozygosis as measured by the number of 
allelic pairs is the same in both lots. In the backcross there are more indi- 
viduals in the intermediate classes with respect to the number of dominant 

TABLE 14.2 

INCREASE IN HEIGHT OF PLANT IN SUCCES- 
SIVE BACKCROSSED GENERATIONS HET- 
EROZYGOUS FOR A DOMINANT GENE 
MARKER 



Number of 


Per Cent Increase in Height 


Generations 
Backcrossed— > 


4 


5 


6 


20yX243Y 

20yXP8 Y 

20pX243P 

243glX20Gl 


6.7 
1.9 
6.6 


2.2 
2.3 
3.0 


1.5 

1.2 

1.1 

-1.3 











genes. This indicates a complementary action of favorable dominant genes. 
Gowen et al. (1946) compared the differences in egg yield in Drosophila 
between random matings, 47 generations of sib mating, and homozygous 
matings by outcrossing with marker genes. The differences are significant, 
and indicate a large number of genes having dominant effects on the repro- 
ductive rate. 

INTERACTION OF GENES AND CYTOPLASM 

The suggestion has been made many times that heterosis may result from 
an interaction between genes and cytoplasm. Within the species, differences 
in reciprocal crosses are rare. In commercial corn hybrids, reciprocal differ- 
ences are so small that they can usually be ignored. Evidence is accumulating 
that there are transmissible differences associated with the cytoplasm, and 
that these must be considered in a study of heterosis. Small maternal effects 
are difficult to distinguish from nutritional and other influences determined 
by the genotype of the mother and carried over to the next generation. 

The cross of the two different flowering types of tobacco previously cited 
shows a maternal effect. The first generation cross of the indeterminate or 



PLASMAGENES AND CHROMOGENES IN HETEROSIS 233 

non-flowering type as seed parent grows taller than the reciprocal com- 
bination, and flowers later. These difi"erences are statistically significant. 

Reciprocal crosses between inbred California Rice pop, having the small- 
est seeds known in corn, with inbred Indiana Wf9 having large embryos 
and endospersms, show differences in early seedling growth and in tillering. 
Inbred Wf9 produces no tillers. California Rice, also inbred, produces an 
average of 4.1 tillers per plant. The first generation cross of Rice popXWfQ 
averages 1.0 tillers, while the reciprocal combination under the same condi- 
tions produced 2.2 tillers per stalk. In this case the non-tillering variety, 
when used as the seed parent, produces more than twice as many tillers. 
This seems to be a carry-over effect of the large seed. Tillering is largely de- 
termined by early seedling vigor. Anything that induces rapid development 
in the early stages of growth tends to promote tillering. 

PLASMAGENES AND CHROMOGENES 

In addition to these transitory effects there are many cases of cytoplasmic 
inheritance. Caspari (1948) has reviewed the evidence from fungi, mosses, 
the higher plants, and from Paramecium, insects, and mammals to show 
that many differences do occur in reciprocal crosses and that they persist into 
later backcrossed generations. Reciprocal differences in the amount of 
heterosis have been demonstrated in Epilobium (Michaelis, 1939) and in 
mice (Marshak, 1936). 

Cytoplasmic pollen sterility has been found in Oenothera, Streptocarpus, 
Epilobium, flax, maize, onions, sugar beets, and carrots. In every case that 
has been adequately studied, the basic sterility remains the same in repeated 
generations of backcrossing, but the amount of pollen produced varies in 
different genotypes. There is an interrelation between plasmagenes and 
chromogenes determining the final result (Jones, 1950). 

In maize the amount of pollen produced ranges from to 100 per cent. 
Only by suitable tests can these cases of full fertility be recognized as having 
any cytoplasmic basis. Interest in this problem now centers on the effect of 
these cytoplasmic differences on heterosis. 

A series of standard inbreds have been converted by crossing these onto 
suitable sterilizer stocks, and backcrossing a sufficient number of generations 
to re-establish completely the inbred, and maintaining the inbred in a sterile 
condition by continuous backcrossing. It has been found necessary to select 
both the cytoplasmic sterile seed parent individuals and the individual fertile 
pollen parents for their ability to maintain complete sterility both in inbreds 
and in crosses. In some lines it has proved to be impossible to establish com- 
plete sterility, but the majority are easily sterilized and maintained in that 
condition. 

A comparison of fertile and sterile progenies in inbreds, in single crosses of 
two inbreds, and multiple crosses of three and four inbreds, shows that this 



234 



DONALD F. JONES 



cytoplasmic difference has no appreciable effect on size of plant as measured 
by height at the end of the season, in days to silking, or in yield of grain. 
The results are given in Table 14.3. With respect to pollen sterility-fertility, 
the cytoplasm has no effect on heterosis. 

In the conversion of standard inbreds to the cytoplasmic sterile pollen 
condition, it has been found that many of these long inbred strains, presum- 
ably highly homozygous, are segregating for chromogenes that have the abil- 
ity to restore pollen fertility. In normally fertile plants these genes have no 
way of expressing themselves. They are not selected for or against unless 
they contribute in some way to normal pollen production. It is one more 

T.\BLE 14.3 
A COMPARISON OF FERTILE AND STERILE MAIZE PLANTS 





Fertile 


Sterile 


5 Inbreds 

7 Crosses of two inbreds. . 
7 Crosses of two inbreds . 
3 Crosses of three inbreds. 

1 Cross of three inbreds. . . 
3 Crosses of four inbreds 

5 Crosses of four inbreds. . 

2 Crosses of four inbreds . 


72.3 
102.6 

58.5 
111.7 

99.1 
123.9 

61.1 
115.8 


70.1 Height of Stalk 

97.7 Height of Stalk 

58.3 Days to first silk 
108.9 Yield, bushels per acre 
103.3 Yield, bushels per acre 
119.0 Yield, bushels per acre 

64.5 Yield, bushels per acre 
117.3 Yield, bushels per acre 


14 Crosses, average yield. . . . 


102.8 


102 . 6 Yield, bushels per acre 



source of evidence to show that there is a considerable amount of enforced 
heterozygosity in maize. Even highly inbred families remain heterozygous. 
This has been shown to be true for other species of plants and animals. 

SUMMARY 

Specific evidence from a study of chlorophyll production in maize and 
from similar studies in Neurospora, Drosophila, and other plant and animal 
species proves conclusively that there are numerous mutant genes that re- 
duce the ability of the organism to grow and to survive. Such genes exist 
in naturally self-fertilized and cross-fertilized organisms and in arti- 
ficially inbred families such as maize. The normal alleles of these mutant 
genes show either complete or partial dominance, and any crossbred indi- 
vidual contains a larger number of these dominant, favorable alleles than 
any inbred individual. 

Evidence from Nicotiana shows that there is an interaction between di- 
vergent alleles at the same locus such that the heterozygote produces a larger 
amount of growth and a higher reproductive rate than either homozygote. 
There is no increase in growth rate and this instance is considered not to be 
heterosis. The assumption of an increased growth rate, or true heterosis, in 
such allelic interactions is not supported by specific evidence that cannot be 



PLASMAGENES AND CHROMOGENES IN HETEROSIS 235 

interpreted in other ways. The experimental evidence to date does not dis- 
tinguish clearly between a general physiological interaction and a specific 
contribution from favorable dominant effects. More evidence on this point is 
needed. 

Backcrossing experiments in maize, where dominant gene markers are 
maintained in a heterozygous condition, show heterosis continuing to the 
sixth generation. The approach to the level of growth activity of the recur- 
rent inbred parent is so slow as to indicate that every region of the chromo- 
somes, divisible by crossing over, has an effect on growth. 

The growth rate in these backcrossed generations is maintained at a level 
appreciably above the proportional number of heterozygous allelic pairs. 
This effect can be interpreted in a number of ways other than a general 
physiological interaction, such as enforced heterozygosity, and the comple- 
mentary action of dominant genes at different loci. 

There is no way known at the present time to distinguish clearly between 
the accumulation of favorable dominant effects of compound or multiple 
genes at the same loci and a general physiological interaction or over- 
dominance. 

Reciprocal crosses differ in many species, resulting in appreciable diver- 
gence in the amount of growth, and these differences have a cytoplasmic basis. 
The evidence from maize, however, shows clearly that cytoplasmic pollen 
sterility has no effect on size of plant, time of flowering, or productiveness. 



M. R. IRWIN 

Universify of Wisconsin 



Chapter 15 

Specificity 
of Gene Effects 



If an attempt were made to survey all the possible ramifications suggested 
by the title of this paper, it should include much of the published work in 
genetics. It is of course a truism to all students of genetics to state that some 
sort of differential specificity towards the end product must exist between 
allelic genes or their effects could not be studied. It would be very interesting 
as a part of this discussion to attempt to trace the change in concepts held by 
various workers during these past fifty years concerning the nature and paths 
of action of the gene. However, beyond a few remarks, such considerations 
are hardly within the scope of this chapter. 

Since the effects of genes can be recognized only if there are differences in 
the end product, it is quite natural that the differences in the experimental 
material first subjected to genetic analyses should have been those which 
were visible, as differences in form, color, etc. Although the pendulum has 
swung somewhat away from intensive investigations of such hereditary char- 
acteristics, it should be emphasized that by their use the underlying mecha- 
nisms of heredity have been elucidated. 

Major attention was given by most of the investigators during the first 
quarter of this century to the effects of respective genes upon individual 
hereditary characters. In some quarters there was an oversimplification in 
the interpretation of the relation of the gene to the character affected by it. 
Gradually, however, the concept has become clearer that the majority of 
hereditary characters — even many of those which had previously appeared 
to be most simply inherited — are affected by many genes. 

An early observation of gene specificity, too long neglected by all but a few 
geneticists, was that made by Garrod in 1909 (see 1923 edition of Inborn 

* Paper No. 433 from the Department of Genetics, University of Wisconsin. 

236 



SPECIFICITY OF GENE EFFECTS 237 

Errors of Metabolism) on the inability of some humans to break down homo- 
gentisic acid (2,5-dihydroxy])henylacetic acid), resulting in the disease known 
as alcaptonuria. Observations reported by Gross (1914) indicated that this 
affliction was due to the lack of a ferment (enzyme) in the serum of alcapto- 
nurics, whereas the enzyme capable of catalyzing the breakdown of homo- 
gentisic acid was demonstrable in the serum of normal individuals. As 
Beadle (1945) has stated, no clearer example exists today that "a single gene 
substitution results in the absence or inactivity of a specific enzyme and that 
this in turn leads to the failure of a particular biochemical reaction." (The 
writer distinctly remembers that, while he was a student in a class in physio- 
logical chemistry, the instructor paid considerable attention to the chemical 
explanation of alcaptonuria, but none at all to its hereditary nature.) 

Another example of gene specificity and also of gene dosage is that of yel- 
low endosperm in corn and the content of vitamin A reported by Mangelsdorf 
and Fraps (1931). Their study showed that the amount of vitamin A in the 
endosperm of white corn was almost negligible, but that the presence in the 
endosperm of one, two, or three genes for yellow pigmentation was accom- 
panied by corresponding increases in the amounts of the vitamin. 

GENE EFFECTS IN A SERIES OF REACTIONS 

There are numerous examples which have shown that many genes con- 
tribute to the development of a heritable character. Thus, in corn there are 
many genes which affect the development of chlorophyll. Each recessive 
allele, when homozygous, allows the formation of only partial pigmentation, 
or in extreme cases no pigmentation at all, and the seedlings are albino. It is 
generally believed that the majority, if not all, of these different genes for 
albinism affect different steps in the process of chlorophyll development. A 
breakdown of the process at any one of these steps results in albinism of the 
seedling. Haldane (1942) has likened the complexity of such a synthetic proc- 
ess to the activity of an equal number of students as there are genes, "en- 
gaged on different stages of a complicated synthesis under the direction of a 
professor, except that attempts to locate the professor have so far failed. Or 
we may compare them to modern workers on a conveyor belt, rather than 
skilled craftsmen each of whom produces a finished article." 

One of the earliest examples of the physiological bases of the specificities 
which are the final gene products is that of the chemical analyses of genetic 
variations in flower color. These studies were carried out in England by sev- 
eral workers. See reviews by Beadle (1945), Beale (1941), Haldane (1942), 
Lawrence and Price (1940) for the general results and references to specific 
papers. 

Mention will be made here of only one of the many investigations which 
have defined in chemical terms the hereditary differences in pigmentation. 
Anthocyanin is one of the five types of pigments concerned in flower color, 



238 M. R. IRWIN 

and its presence or absence in several species is genetically determined. One 
way in which anthocyanin may be modified is by the degree of oxidation of 
the prime ring. According to Beale (1941) in the two genera Lathyrus and 
Streptocarpus, the hydroxyl group is at position 4' in the pelargonidin type, 
at positions 3' and 4' in the cyanidin types, and at 3', 4', and 5' in the delphin- 
idin types. The more oxidized pigments are usually dominant to the less 
oxidized types. Thus flowers with genes AB and -46 will be of the delphinidin 
type of pigment, those with aB of the cyanidin type, and those with ab of the 
pelargonidin type. 

These and other extensive chemical studies on the anthocyanin pigments 
genetically modified in various ways are dramatic examples of the specifici- 
ties of gene effects. The analogy drawn above between the various genes and 
students working on a complicated synthesis becomes a little more clear in 
relation to flower pigments, since considerable information is available as to 
what some of the genes accomplish. 

A further example of the effect of many genes upon a character is that of 
eye color in Drosophila melanogaster. Between twenty-five and thirty genes 
are known to modify the brownish-red color of the wild-type eye. There ap- 
pear to be two independent pigments, brown and red, concerned in the devel- 
opment of the wild-type eye, each of these being affected by specific genes. 
Certain components of the brown pigment are diffusible from one part of the 
body to another, and hence are more readily subjected than others to chem- 
ical analyses. 

The details of these analyses are presented in other review articles (Beadle, 
1945; Ephrussi, 1942a, 1942b). Briefly, dietary tryptophan is converted to 
alpha-oxytryptophan by a reaction controlled by the wild-type allele of the 
vermilion gene (?). This substance is oxidized further to kynurenine (the so- 
called i'+ substance). By virtue of the activity of the normal allele of the cin- 
nabar gene, kynurenine is further oxidized to the or substance, which may 
be the chromogen of the brown pigment (Kikkawa, 1941). The production of 
either brown or red eye pigment can be blocked by genes at the white eye 
locus, thus indicating that such genes act on a common precursor of the red 
and brown pigments. 

Mention should be made of the relation between the original designation 
of certain of the genes for eye color and their presently known effects. Thus, 
the eyes of flies with the mutant alleles bw bw are brown. But it is now known 
that this pair of alleles, instead of being concerned with the production of 
brown pigment, restricts the development of red pigment and thus we see 
only the brown color. Similarly, the four gene pairs whose mutants modify 
the red coloration do so by virtue of their effect on the brown pigment, not 
upon the red. 

Wheldale (1910) proposed four decades ago that genetic characters were 
the resultant of a series of reactions, and that if a break in the chain occurred, 



SPECIFICITY OF GENE EFFECTS 239 

the series of steps would have proceeded only to that j)oint. Following the 
initial work in Neurospora by Beadle and Tatum (1941) on mutants which 
blocked certain metabolic processes, this type of approach has expanded 
enormously and profitably. Attention can be called here to but one very sig- 
nificant example of this kind of experimental study in microorganisms. A 
report by Srb and Horowitz (1944) shows clearly how many genes act in the 
synthesis of arginine. Of fifteen mutant strains studied, there were seven dif- 
ferent steps represented in the synthesis of arginine. One of the forms grew 
only if arginine was supplied. Two others required either arginine or citrul- 
line, and these two strains were genetically different. Four other strains, 
genetically different from the first three strains and from each other, would 
grow if arginine, citrulline, or ornithine were provided. For a diagrammatic 
representation of these steps, see Beadle (1945). 

DIRECT EFFECTS OF GENES 

The preceding examples are but a few^ of the many which could be cited to 
illustrate the gene specificities in the development of a genetic character 
which involves the successive activities of many genes. Are there any genetic 
characters which may be the immediate products of the causative genes? An 
example almost unique in higher plants is that of the waxy gene in corn 
(Collins, 19C9) in its effects upon the starch of the pollen grain and the endo- 
sperm reserves. As is well known, the starch granules in the pollen grains 
bearing the waxy gene are stained reddish-brown with iodine, as are the 
endosperm reserves of waxy seeds, in contrast to the typical blue reaction of 
the starch granules of non-waxy pollen and of the endosperm reserves of 
non-waxy seeds. Following studies of the physiological effects of the waxy 
gene, Brink (1929) proposed that this gene has its effect on the enzyme amy- 
lase which functions directly in the synthesis of starch. 

Another class of hereditary characters which in some respects appears to 
satisfy some of the criteria for a direct effect of the causative genes is that of 
the antigenic characters of the red blood cells of animals. With only rare 
exceptions, to be considered later in more detail, each of the known antigenic 
substances has appeared in the cells of an individual only if one or both 
parents also possessed it. If there is but a single pair of contrasting charac- 
ters, each is expressed in the heterozygote. Further, the cells which give rise 
to the hematopoietic tissue from which the red blood corpuscles are derived 
are laid down shortly after the first division of the fertilized egg. The possibil- 
ity cannot be excluded, of course, that there is a chain of reactions within 
each cell leading to the formation of the antigen, but no block in such a chain 
of reactions has yet been observed. There are two statements concerning the 
cellular antigens which are of interest: (1) the antigenic substance must be 
located at or near the surface of the cell in order to be detectable, and (2) 
there is no known effect of the environment upon them. 



240 M. R. IRWIN 

We should avoid misunderstanding about the meaning of the terms com- 
monly used in immunological literature. For example, the word antigen was 
originally defined as any substance which, when introduced parenterally into 
an animal, would invoke the production of antibodies. This definition would 
now be extended to include any substance which will react visibly with an 
antibody. And an antibody would be defined as a constituent of the serum 
which reacts with an antigen in any of several ways. The circle of reasoning 
here is obvious. However, insofar as chemical studies of various antigens have 
contributed to an understanding of their specificities, the specificities have 
always been associated with structural differences of the antigenic sub- 
stances. On the other hand, the reasons underlying the specificities in re- 
activity of the antibodies are almost completely unknown, although it is 
known that the antibodies are intimately associated with the globulins of the 
serum, and in fact may constitute the gamma globulins of the serum. 

CELLULAR ANTIGENS IN HUMANS 

As our first example of these antigenic substances, let us consider the well 
known and extensively studied O, A, B, and AB antigenic characters, or 
blood groups, of human cells. Following their discovery by Landsteiner 
(1900, 1901), it soon became clear that these substances were gene controlled. 
At the present time, the theory of three allelic genes, as postulated by Bern- 
stein (1924) on statistical grounds, is generally accepted. The two other the- 
ories proposed for their inheritance — independent and linked genes, respec- 
tively — are fully discussed by Wiener (1943). Landsteiner noted that the 
serum of certain individuals would agglutinate (clump) the cells of other 
individuals, and from this observation the reciprocal relationship between the 
presence and absence of each antigen and its specific antibody has been 
elucidated. 

A or B Antibody 

Antigen on of the 

Group the Cells Serum 

O None Anti-A, Anti-B 

A A Anti-B 

B B Anti-A 

AB AB None 

It may readily be seen that the presence of an antigen, as A, on the cells 
is accompanied by the presence of the antibody (anti-B) for the contrasting 
antigen, as B, in the serum, and vice versa. If both antigenic characters are 
found on the cells, as in AB individuals, the serum contains no antibodies. 
While if neither A nor B is present on the cells, the serum contains both anti- 
A and anti-B. 

These phenomena pose the question whether the genes producing the cel- 
lular substances also have an effect on the antibodies of the serum. That is, 
does the gene which is responsible for the O antigen (which is definitely an 
entity but is less reactive than A and B) also effect both anti-A and anti-B 



SPECIFICITY OF GENE EFFECTS 241 

in the serum — while in individuals with substance A, only anti-B is found; 
in those with B, only anti-A is present; and in AB individuals the effects of 
the respective genes on the antibodies are somehow neutralized? 

Before attempting to answer this question, it will be advisable to review 
the present knowledge of the chemistry of the A and B substances of human 
cells. See Rabat (1949). 

These antigens (blood groups, cellular characters, antigenic factors, etc.) 
are found in nearly all the fluids and tissues of the human body. They also 
are widely distributed throughout the animal kingdom. The A substance or 
an A-like substance has been found, for example, in hog gastric mucosa, in 
the fourth stomach (abomasum) of the cow, and in swine pepsin, while both 
A and B substances have been noted in the saliva and stomachs of horses. 
Following chemical fractionations, principally of horse saliva and hog gastric 
mucosa, various investigators have obtained preparations with activity re- 
lated to the A substance. These preparations have been largely polysac- 
charide in nature. In addition to the polysaccharides, even in the purest 
preparations, some workers have noted traces of amino acids. 

At present, while it appears that both the A and B substances of human 
cells may be classed as nitrogenous polysaccharides, no information is avail- 
able as to the structural differences between them. Our knowledge of such 
specificities rests entirely upon the technics of immunology, that is, by the 
interaction of either naturally occurring antibodies (as anti-A and anti-B), 
or immune antibodies, with the respective substances A and B. 

The antigenic substances A and B of human cells are complex polysac- 
charides, while the antibodies are modified globulins, or are found in serum 
protein very closely related to the globulin fraction. If the gene which effects 
antigen A is responsible also for the B antibody, and that for antigen B for the 
A antibody, it would seem that here is a clear-cut case of pleiotropic effects of 
the respective genes. This explanation runs into difficulties in AB individuals 
which, on this proposal, should have both kinds of antibodies but actually 
have none. In contrast, a current explanation of the reciprocal presence of the 
antigenic substance of the cells and the antibody for the contrasting sub- 
stances is that the antibodies for both substances (A and B) are normal con- 
stituents of human serum. Production of the antibodies would then be con- 
trolled by a gene or genes at another locus than that having to do with the 
cellular substances, ij genes were involved in their production. If an indi- 
vidual carries the gene for A, and hence has A substances widely distributed 
throughout his body, the A antibodies are presumed to be absorbed from the 
serum, and of course the B antibodies are left. Also, an individual with the 
B substance would absorb the B antibodies, and the antibodies to A would 
remain, while both anti-A and anti-B would be absorbed in an AB individual. 
Other hypotheses are given by Wiener (1943). Unfortunately, no experimen- 
tal test of the correctness of this or other hypotheses is likely. 



242 M. R. IRWIN 

Landsteiner and Levine (1927) announced the discovery in human cells of 
a new pair of contrasting antigens, called M and N. These were detectable 
only by the use of immune sera produced in rabbits, as was another antigenic 
factor called P. The heritability of the M and N substances is adequately 
explained by the assumption of a single pair of allelic genes, and the sub- 
stance P appears to be dominant to its absence. 

Another antigenic factor in human blood which has aroused wide interest 
is the recently discovered Rh substance, or complex, as it might be termed. 
In 1940, Landsteiner and Wiener (1940) reported that a new antibody, de- 
rived from a rabbit immunized with the erythrocytes of a rhesus monkey, 
was reactive with the cells of about 85 per cent of the white population of 
New York. They gave the name Rh (a contraction of rhesus) to this agglu- 
tinable property of human cells. As Boyd (1945) aptly states: 

The technic of testing for the new factor was difficult, the best available serums were 
weak, and had it not been for a remarkable series of discoveries which followed in the next 
few months, the Rh factor might have aroused no more interest than its practically still- 
born brethren. . . . 

The Rh factor was shown to be involved in previously unexplained com- 
plications following transfusions (Wiener and Peters, 1940), but is most 
widely known for its role as the etiologic agent in the majority of cases 
of hemolytic disease of the newborn. The proposal was first made by 
Levine and Stetson (1939) that an antigen in the fetus, foreign to the mother 
and presumably transmitted by the father, could pass through the placenta 
and immunize the mother. Later studies implicated the Rh factor as the 
foreign antigen, and showed that the antibodies developed in the mother may 
pass back through the placenta and affect the red blood cells of the fetus, 
before or following birth. Although the majority of cases of hemolytic disease 
of the newborn may be justly ascribed to Rh incompatibility between the 
father and mother, there is no satisfactory explanation as to why only about 
one in forty of such potentially dangerous combinations leads to morbidity. 

There exist several subgroups, or subtypes, of the Rh complex, and inves- 
tigations as to their respective specificities occupy the center of interest of 
many workers at the present writing. There are two schools of thought as to 
the mode of inheritance of these subgroups, which also involves the terminol- 
ogy to be used in their identification (see Strandskov, 1948, 1949, for leading 
references). One explanation is that the various subtypes are manifestations 
of a series of multiple allelic genes, the other that they are the result of the 
action of respective genes at three different but closely linked loci. It is not 
within the province of this chapter to discuss the arguments for and against 
these two proposals. However, it should be stated that the genetic results 
under either explanation are essentially the same. 

One of the most pertinent statements which can be made about these 
various antigenic substances of the erythrocytes is that they are detectable 



SPECIFICITY OF GENE EFFECTS 243 

no matter in what gene complex they may occur. That is, other genes than 
the causative ones have no measurable influence upon their expression. A 
possible exception to this statement might be proposed for the A and N char- 
acters, res])ectively, since each is somewhat less readily agglutinated when in 
the heterozygote, AB and MN, than when either occurs singly. 

THE HYBRID SUBSTANCE IN SPECIES HYBRIDS 
Until the early part of this century, most of the workers in immunology 
had reached the conclusion that the specificities obtained in immunological 
reactions were primarily if not entirely concerned with proteins. Therefore, 
the finding by Heidelberger and Avery (1923, 1924) that the immunological 
specificities of the pneumococcal types were dependent upon polysaccharides 
was indeed a forward step in our understanding of the chemical nature of 
biological specificity. It is a pleasure to acknowledge that this work of 
Heidelberger and Avery convinced the writer that immunological technics 
should be a useful tool in studying genetic phenomena. Also, although at that 
time pollen differing in gene content seemed (and still does) to be promising 
experimental material, the species and species hybrids in pigeons and doves 
produced by the late L. J. Cole were tailor-made for further studies. 

Pigeon-Dove Hybrids 

The first step was to determine whether the cells of one species could be 
distinguished from those of the other. In brief, all the comparisons by im- 
munological technics, between any pair of species of pigeons and doves, have 
resulted in the ability to distinguish the cells of any species from those of an- 
other, and to show that each species possessed antigenic substances in com- 
mon with another species, as well as those peculiar to itself — those species 
specific. A dozen or more kinds of species hybrids have been obtained in the 
laboratory, and in general, each kind of hybrid has contained in its cells all 
or nearly all of the cellular substances of both parental species. One such 
species hybrid is that obtained from a mating between males of an Asiatic 
species, the Pearlneck (Slreplopelia chinensis) and the domesticated Ring 
dove females {St. risoria). The corpuscles of these hybrids contained all the 
substances common to each parental species, but did not contain quite all the 
specific substances of either parental species. Further, the cells of these 
hybrids did possess a complex of antigenic substances not found in the cells 
of the parents. These relationships are presented in Table 15.1 and are given 
in diagrammatic form in Figure 15.1. This new antigen has been called the 
"hybrid substance," and it has been ])resent in every hybrid produced be- 
tween these two species. 

Upon repeatedly backcrossing these species hybrids and selected back- 
cross hybrids to Ring dove, ten antigenic substances which differentiate 
Pearlneck from Ring dove have been isolated as probable units. That is, a 



244 



M. R. IRWIN 



backcross bird carrying any one of these unit substances, when mated to a 
Ring dove, has produced approximately equal proportions of progeny with, 
and without, the particular substance in their blood cells. These substances 
peculiar to Pearlneck, as compared with Ring dove, have been called d-1, 
d-2, d-3, d-4, d-5, d-6, d-7, d-8, d-11, and d-12. Each of these is distinct from 
the others (Irwin, 1939) both genetically and immunologically. Thus it ap- 
pears that a gene or genes on each of ten of the thirty-odd pairs of chromo- 

TABLE 15.1 

ANTIGENIC RELATIONSHIPS OF THE BLOOD CELLS 

OF PEARLNECK, RING DOVES, AND 

THEIR HYBRIDS 



Immune Serum 



Pearlneck . 
Pearlneck . 
Pearlneck . 



Ring dove. 
Ring dove. 
Ring dove . 



Fi. 
Fi. 
F,. 
Fi. 



Absorbed by 
Cells of 



Ring dove 
Fi 



Pearlneck 
Fi 



Pearlneck 
Ring dove 
Pearlneck and 
Ring dove 



Agglutination Titers 
WITH Cells of 



Pearlneck 



23040 

11520 

90 

15360 



15360 



7680 





Ring Dove 



23040 



15360 

3840+ 
180 

15360 
3840+ 





23040 

11520 



15360 
3840+ 


15360 
3840+ 
7680 
360+ 



" -o ^'^ o"q a o 



.<>%»o • Pearlneck . <, o.-/ 

On a oQf, a n M /i 1% n ^ O O O O « O 



I 



A a. H.Q a a q q • 



I 



t\\ Pearlneck "aV-sfCs; 



Common 







Common 


Ringdove 


1 


i 



Pearlneck 



Ringdove 



Common- 



Ringdove^ P'N. x R,D.-F, 



Hybrid 



Fig. 15.1 — Diagrammatic representation of the antigenic relationships of the Pearlneck, 

Ring dove, and their hybrids. 



SPECIFICITY OF GENE EFFECTS 245 

somes of Pearlneck produce effects on cellular antigens which differentiate 
Pearlneck from Ring dove. Although the cellular substances particular to 
Ring dove, in contrast to Pearlneck, have not been obtained as units, the 
available evidence indicates strongly that a gene or genes on nine or ten 
chromosomes of Ring dove produce antigenic effects which differentiate that 
species from Pearlneck. 

The question may well be raised as to what this recital of antigenic char- 
acters in man and doves, which in general illustrates gene specificity in the 
production of cellular antigens, has to do with the general topic of heterosis. 
The so-called hybrid substance has one word (hybrid) in common with the 
term hybrid vigor, and suggests a possible relationship of the two terms. 

The hybrid substance seemingly represents a departure from the hypoth- 
esized direct action of a gene on the antigenic substance, in that it appears 
to result from the interaction of two or more genes in the species hybrids to 
produce some antigenic substance different from any detectable in either 
parent. With but one exception proposed by Thomsen (1936) in chickens, and 
for which another explanation will be considered shortly, a hybrid substance 
has thus far been found only in species hybrids. 

Mention should be made of the technics required for the detection of the 
hybrid substance. Briefly, if an antiserum prepared against the cells of an 
individual, whether a species hybrid or not, would be absorbed by the cells of 
both its parents and would then react with the cells of the individual, but not 
with the cells of either parent, there would be evidence of a different anti- 
genic substance in the homologous cells — those used in the immunization. 
(If an antigen were recessive, it would be present in the heterozygote, and 
presumably could absorb its specific antibody.) 

Domestic Fowl Hybrids 

As stated above, Thomsen (1936) reported that within each of two families 
of chickens there was a different antigenic substance present than was found 
in the parents. Attempts in our laboratory by Mrs. Ruth Briles to duplicate 
this finding were without success, but a very interesting and quite unex- 
pected observation was made which may be the explanation of Thomsen 's 
finding. If an antigenic substance were present in an individual different from 
that possessed by either parent, immunization of either parent (as #1) with 
the cells of this individual might engender antibodies against the new sub- 
stance. Absorption of such an antiserum by the cells of the other parent (as 
#2) should remove all antibodies except those formed against the new or 
hybrid substance, and such a reagent should be reactive only with the cells 
containing the new substance. This was the procedure followed by Thomsen, 
except that his tables do not show that the cells of the two parents were used 
as negative controls in the tests made after the various absorptions. 

Immunizations of each of the parents of a family of chickens against the 



246 M. R. IRWIN 

cells of one of the offspring, or the pooled cells of two or more, were made by 
Mrs. Briles. Following the absorption of the antiserum obtained from either 
parent by the cells of the other, it was noted that the absorbed antiserum was 
at least weakly reactive with the cells of the individual from which the anti- 
serum was obtained. That is, such an antiserum would not react (agglutinate) 
with its own cells before absorption with the cells of the mate, but after such 
absorption it definitely would agglutinate the cells of the individual from 
which it was derived. 

To use a concrete example, bird R614 (containing Bi antigen) was im- 
munized with the washed cells of R2C43, to produce B3 antibodies (Briles, 
McGibbon, and Irwin, 1951). After this antiserum from R614 was mixed for 
absorption with the washed cells of R622 (having B3 antigen in its cells and 
having been immunized to produce Bi antibodies), all cells containing the Bi 
antigen were reactive with it, including those of R614 itself. Thus it appears 
that the antibodies to Bi which were circulating in the serum of R622 were 
also attached to the surface of the red blood cells and were transferred to the 
antiserum from R614 during the absorption process. It was possible to dem- 
onstrate that, after washing the cells of R622 in saline, the saline contained 
antibodies, even after nine successive washings. Hence, unless the cells of 
both parents were used as controls in comparable tests for the presence of a 
hybrid substance, agglutination of any cells could be explained as due to a 
transfer of antibody from the blood cells to an antiserum. Unfortunately, 
such controls are not given in Thomsen's paper, and the possibility cannot be 
eliminated that the reactions obtained by him were due simply to segregation 
within the various families of an antigenic character of one of the parents. 
This possibility was mentioned by Thomsen (1936), but was not considered 
applicable to his experiments. 

Hybrid Substances 

Returning to the hybrid substance for which there is definite evidence, it 
should first be stated that such a substance has not been found in all kinds of 
species hybrids, as may be seen from the data given in Table 15.2. It has been 
reported from our laboratory in hybrids between Pearlneck and Ring dove, 
the pigeon {Columba livia) and Ring dove, the Mallard {Anas platyrhynchos) 
and Muscovy duck (Cairina moschata), but not in the hybrids between the 
triangular spotted pigeon (C. guinea) and livia. Irwin (1947) gives the spe- 
cific references to pertinent articles. 

A hybrid substance has been detected but not previously reported in 
hybrids from malings between the Philippine turtle dove (St. dussumieri) 
and Ring dove, the dwarf turtle dove {St. Jmmilis) and Ring dove, the 
Oriental turtle dove {St. orientalis) and Ring dove, and the band tail pigeon 
{C. fasciata) and livia. No such substance has been observed in the hybrids 
between the Senegal dove {St. senegalensis) and Ring dove, an African dove 



SPECIFICITY OF GENE EFFECTS 



247 



(St. semitorquata) and Ring dove, the Senegal dove and the Cape turtle dove 
{St. capicola), the spot wing pigeon (C. maailosa) and livia, and between the 
Grayson dove {Zenaidiira graysoni) and the common mourning dove {Zen. 
macroura). It is possible that a hybrid substance does exist in these latter 
species hybrids, but the same technics by which it was observed in the other 
species hybrids failed to demonstrate its presence in them. 

Three different fractions of the hybrid substance have been demonstrated 
in the hybrids between Pearlneck and Ring dove (Irwin and Cumley, 1945), 
by virtue of a frequent association of each fraction with one or more antigens 

TABLE 15.2 

TESTS FOR HYBRID SUBSTANCES IN THE CELLS OF 

VARIOUS SPECIES HYBRIDS 



Antiserum to 



Fi — Pearlneck X Ring dove 

Fi — C. //zi/aX Ring dove 

Fi — St.dussumieriXRing dove. . . . 

Fi — St. humilisXRing dove 

Fi — St. orientalisXRing dove 

Fi — C. fasciataXlivia 

Fi — Mallard X Muscovy 

Fi — St. senegalensisXRing dove. . . 
Fi — St. semitorquataXRing dove. . 

F: — Senegal X-S/. capicola 

Fi — C. tnaculosaXlivla 

Fi — C. guineaXlivia 

Fi — Zenaidura gray soniX Zen. ma 
crmira 



Absorbed by Cells of 



Parent 1 



Parent 2 



Pearlneck 

livia 

dussumieri 

h II mil is 

orientalis 

C. fasciata 

Mallard 

St. senegalensis 

St. semitorquata 

Senegal 

C. maculosa 

C. guinea 

Zen. graysoni 



Ring dove 
Ring dove 
Ring dove 
Ring dove 
Ring dove 
livia 

Muscovy- 
Ring dove 
Ring dove 
St. capicola 
livia 
livia 

TLen. macroura 



Re.actions of Pa- 
rental AND HYBRID 

Cells with the 

Respective 

Reagents 



Par- 


Par- 


ent 1 


ent 2 

















































































Hy- 
brid 



+ + 

+ + 

+ 

+ 

+ 

+ 

+ + 















peculiar to Pearlneck. Thus one fraction called dx-A was always associated 
in the backcross hybrids with the d-1 1 substance, dx-B seemingly was loosely 
linked with the d-1 character and with certain others as well — thereby pro- 
viding strong evidence that on several chromosomes of Pearlneck there are 
duplicate or repeat genes — and dx-C was always associated with the d-4 
antigen. The pertinent reactions which show these specificities are given in 
Table 15.3 and are represented diagrammatically in Figure 15.2. 

Because of the constant association of the dx-A and dx-C fractions with 
the d-11 and d-4 substances, respectively, one cannot be certain that these 
two fractions, although antigenically distinct from the d-1 and d-11 specific 
characters, are not simply a new specificity conferred upon the specific char- 
acters by some sort of rearrangement of the specific substances following the 



248 



M. R. IRWIN 



interaction of the causative genes. This question cannot be completely an- 
swered until either a genetic separation has been observed, as between the 
dx-A and d-11, or the chemical separation into two distinct substances has 
been done. On the other hand, the dx-B fraction has been separated from 
each of the species specific characters to which it presumably is loosely 
linked, thereby showing that this fraction of the hybrid substance is an 
antigenic entity. 

The reagent which interacts with the hybrid substance (hybrid antiserum 

TABLE 15.3 

TESTS FOR SIMILARITIES AND DIFFERENCES OF THE COM- 
PONENTS OF THE "HYBRID SUBSTANCE" OF THE SPECIES 
HYBRID BETWEEN PEARLNECK AND RING DOVE 





Reactions of Different Cells with Anti-Fi Serxjm 


Cells 


Absorbed 

by Cells 

of Both 

Pearlneck 

and Ring 

Dove 


Absorbed by the Cells of Both Pearlneck and Ring Dove, 
in Combination with Others as Listed 




d-1 
(dx-B) 


d-4 

(dx-C) 


d-11 
(dx-A) 


P.N. 
^iR.D. 


Pgn 
^iR.D. 


Sen. 


Aus. 
cstd. 


Pearlneck 

Ring dove 

Fi— P.N./R.D 

d-1 (dx-B). . . . 

d-4 (dx-C) .... 

d-11 (dx-A). . . 

Fi-Pgn/R.D 

Senegal 

Australian crested. 





++ 

+ 

+ 

++ 

+ 

+ 

± 





++ 



? 

++ 
+ 



+ 





++ 
+ 



++ 
+ 



+ 





-1- 
+ 

-f 
























+ + 
+ 

-1- 

+ 








++ 
+ 

-f- 

++ 
+ 



± 





+ + 

+ 
-f 

+ 
-1- 



Column 


2 


3 


4 


5 


6 


7 


8 


9 



Symbols: -f -f = marked agglutination; -|- = agglutination; ± = definite but weak agglutination; ? 
doubtful reaction; = no agglutination — at the first dilution of the serum cell mixture. 



absorbed by the cells of both Pearlneck and Ring dove) will also agglutinate 
the cells of various species. Thus in the genus Streptopelia, there were five 
species {capicola, dussutnieri, kumilis, orientalis, and senegalensis) other than 
Pearlneck and Ring dove whose cells were reactive, and one {semitorquata) 
with nonreactive cells. Within the genus Columba, the cells of one species 
(rufina) likewise reacted with this reagent, but those of seven other species 
(fasciala,flavirostris, guinea, livia, maculosa, palumbus, and picazura) did not. 
And of twelve species tested in other genera within the Columbidae, only 
three species from Australia (Australian crested dove, or Ocyphaps lopkofes, 
the bronze wing dove or Phaps chalcopiera, and the brush bronze wing dove, 
or Phaps elegans) possessed reactive cells. In Table 15.3, the Senegal cells are 
representative of the parallel reactions of the five species of the Streptopelia, 



SPECIFICITY OF GENE EFFECTS 



249 



as are those of the Australian crested of the equivalent reactivities of the 
three Australian species. 

Although the reagent for the hybrid substance did not agglutinate the 
cells of livia, it invariably clumped those of the hybrids between pigeon 
(livia) and Ring dove. As previously reported (Irwin and Cole, 1936), these 
hybrids also contain a hybrid substance. Because of the cross reactions exist- 
ing between these two hybrid substances, a certain degree of similarity can 
be assumed. That the fraction in the hybrid substance of the Fi-Pearlneck X 

DIAGRAMMATIC REPRESENTATION OF THE'HYBRID SUBSTANCE" 
OF THE HYBRID BETWEEN PEARLNECK AND RINGDOVE 



VPEARLNECK>- 



^^■•^V-VV:- 



COMMON- >•:•.■•. '.<,"•.•••: 



RINGDOVE 



O O O O O 

O O O 

O O O ^ Q 

o O O OO 

O o 



n PEARLNECK 
^i RINGDOVE 



HYBRID SUBSTANCE 



(d-ll) 


Cd-i) (d^) 


M 
















O 0« OO 

O O o o o 
O o 00 
O £7 oO 
O Oft 
o a o OO 




a 00 
a aao 

G O QtJ 
O 00 







dx-A 



d<-B 



dii-C 




MrBRIO SUBSTANCE ASSOCIATED 

WITH SPECIFIC PEARLNECK SUBSTANCES 



RELATED SUBSTANCES 



Fig. 15.2— The separation into constituent parts of the hybrid substance of the species 

hybrid between Pearlneck and Ring dove. 



Ring dove, which is primarily if not entirely responsible for the cross reac- 
tions, is dx-A may be deduced from Table 15.3, in that this fraction (associ- 
ated with d-ll) is the only one which will exhaust the antibodies from the 
reagent for the cells of the pigeon-Ring dove hybrid (column 5). Also, in 
unpublished tests the reagent for the hybrid substance of the pigeon-Ring 
dove hybrid (anti-hybrid serum absorbed by the cells of pigeon and Ring 
dove) did not react with Pearlneck cells, but reacted strongly with d-ll cells, 
presumably by virtue of their content of the dx-A fraction, and not definitely 
with cells carrying dx-B or dx-C. If the dx-A hybrid substance of the species 
hybrids between Pearlneck and Ring dove were partially or largely a rear- 
rangement of an antigenic substance, in this case d-ll, which is species spe- 
cific to Pearlneck — since the Ring dove is a common parent of the two kinds 



250 M. R. IRWIN 

of species hybrids — that specific substance (cl-11) should be detectable in the 
cells of livia. 

To date, reasonably extensive tests (unpublished) have not shown that the 
cells of livia contain more than a trace of an antigenic substance related to the 
d-11 of Pearlneck. Whatever the relationship of the genes in Pearlneck (as- 
sociated with those on a chromosome effecting the d-11 specific substance) 
and livia, respectively, which presumably by interaction with a gene or genes 
from Ring dove in the two species hybrids effect a common fraction of the 
two hybrid substances, they are not associated with genes which produce 
similar antigenic patterns in the two species. On these grounds, it would seem 
unlikely that the hybrid substances in these two kinds of species hybrids are 
merely a different arrangement of a species specific antigen. 

The question is pertinent as to whether such reactivities in the cells of 
these other species, as Senegal and Australian crested, are themselves an in- 
dication of antigenic response to gene interaction within each species, or the 
more direct product of a gene. This cannot be answered directly. But, as 
given in Table 15.3, the fact that absorption of the reagent for the hybrid 
substance by fractions dx-A, dx-B, or dx-C removes the antibodies for the 
cells of Senegal indicates that there is some common constituent of these 
three fractions related to, if not identical with, a reactive substance in Senegal 
cells. However, only the dx-A fraction removes the antibodies for the cells of 
the Australian crested dove. Further, absorption by the cells of the pigeon- 
Ring dove hybrid also removes the antibodies from this reagent for the cells 
of the Australian crested dove. 

The hypothetical explanations could be advanced, (1) that the antigenic 
substances in Senegal and the Australian crested dove, themselves being dis- 
tinct, but both related to the hybrid substance in Pearlneck-Ring dove 
hybrids, are the result of a genie interaction. But there is no evidence for such 
an assumption. Also, (2) the argument could be advanced that the relation- 
ship between these substances in Senegal and Australian crested, and in the 
respective species hybrids, is fortuitous, simulating the occurrence of the 
Forssman antigen in many species of animals and plants, including bacteria 
(Boyd, 1943). That is, the antigenic substances involved (related in some 
manner to the hybrid substance) may be gene controlled in each of the re- 
lated species, since indistinguishable substances to those of Senegal were 
found in four other species of Streptopelia, capicola, dussumierl, humilis, and 
orienlalis, and to those of the Australian crested dove in two species of an- 
other genus, Phaps chalcoptera and Phaps elegans, but the antigenic similarity 
to the hybrid substance is by virtue of some related antigenic component. 
Various ramifications of these and other explanations would be purely 
speculative. 

The hybrid substance, as it has been observed in the cells of various species 
hybrids in birds, simulates for cellular antigens the expression of heterosis in 
plants and animals. That is, it appears as the resultant of an interaction be- 



SPECIFICITY OF GENE EFFECTS 



251 



tween genes. One may well ask if there is any other manifestation of heterosis 
in these species hybrids and backcross hybrids. Extensive measurements of 
eight body characteristics, as over-all length, extent, width of tarsus, width of 
band, length of wing, beak, middle toe, and tail, were made over a 7)eriod of 
years under the supervision of L. J. Cole. The differences in the averages of 
these various characteristics between the parental Pearlneck and Ring dove 
species, as yet unpublished, were statistically significant, and the averages of 
the measurements of these characteristics in the species hybrids showed them 
to be in general intermediate between those of the parental species. Thus 
there was no evidence of heterosis in any external characteristic of the species 
hybrids, and no correlation with the hybrid substance of the blood cells. 

CELLULAR CHARACTERS WITHIN A SPECIES 

The finding that one or more genes on each of nine or ten pairs of chromo- 
somes of Pearlneck had effects on the species specific antigens of the blood 
cells of this species made plausible the belief that many more genes than 
commonly believed would have effects within a species making for individual- 
ity of the cellular patterns. Acting on this assumption, a series of exploratory 
tests w^ere made in experimental animals, principally in cattle and chickens. 
For example, following the transfusion of the blood of a young cow into her 
dam, an antibody was obtained from the serum of the recipient which re- 
acted (produced lysis of the reacting cells upon the addition of complement 
to the serum-cell mixture) with the cells of some individuals, but not with 
those of others. The reactive substance was called A. 

The objective was to be able to detect each antigenic factor separately, ac- 
cording to the following criterion. The reactive cells from any individual 
should remove the antibodies from the reagent specific for those cells, when 
added in excess to the reagent. However, if there were antibodies in the re- 
agent which recognized two or more distinct blood factors, any such ab- 
sorption with cells containing only one such substance would remove only a 
part of the antibodies. Those remaining would still be reactive with all cells 
containing the substance corresponding to the unabsorbed antibody. 

To this criterion was added that of genetics for a single character, using 
the gene-frequency method since controlled matings were not possible. A 
typical example of the analysis is that for substance A, as follows: 



Typk ok 
Mating 


Number of offspring 


With 
Antigen A 


Lacking 
Antigen A 


AXA 

AX- 


217 

76 




23 
51 


-X- 


41 



252 



M. R. IRWIN 



These results illustrate the observation that an individual has any cellular 
character recognized to date only if one or both parents possessed it. Also, 
each behaved as if it were a dominant to its absence. 

From further isoimmunizations in cattle, and from immunizations of rab- 
bits, various antisera have been obtained which detect other antigenic factors 
of cattle cells. Each of these has been subjected to the criteria of both genetics 
and immunology for a single character, as described in reports by Ferguson 
(1941), Ferguson et al. (1942), and Stormont (1950). At present, about forty 
different reagents are regularly used in typing cattle cells. 

Other Antigens in Cattle 

As stated above, the first substance detected in cattle cells was named A. 
The next was called B, the next C, . . . Z. That called A' implies no relation- 
ship to A, nor B' to B, etc. Each of these antigenic factors is therefore recog- 
nized independently, and when subjected to an analysis of gene frequency, 
each has behaved as expected if effected by a single gene in comparison to its 
absence. 

However, some definite associations have been noted among them. For 
example, Ferguson (1941) reported that the C and E factors were not inde- 
pendent, for only C occurred alone, whereas E was present always with C, 
and such cells therefore had CE. It was postulated that there were three al- 
lelic genes involved, one for the components C and E together, one for C 
alone, and a third for the absence of both C and E. 

It was later noted by Stormont that certain additional antigenic factors 
appeared only if one or more other components also were present. For ex- 
ample, the substance B occurs alone, as does that called G. But a third factor 
called K has never been observed unless both B and G were also present. 
(A possible exception to this rule was noted shortly after these factors were 
first demonstrable, and a weak reaction at that test with the reagent for the 
G substance was probably incorrectly recorded.) This association of K with 
B and G has been noted in over eighteen hundred animals of more than six 
thousand tested. Hence the combination of the BGK factors has always oc- 
curred as a unit, and it has also behaved as a unit in the progeny of individ- 
uals possessing it. A compilation of some unpublished data has yielded the 
following information: 





Number of Offspring 


Type of Mating 


With 
BGK 


Without 
BGK 


BGKXBGK 

BGKX- 


151 

185 




44 
137 


— X — 


160 







SPECIFICITY OF GENE EFFECTS 253 

Notwithstanding the fact tliat H, (1, and K are recognized sei)arately by- 
respective reagents, these data, and the observation that K has occurred 
only with both H and G, are strong evidence for tlie conclusion that B, G, and 
K in the cells behave as a unit. 

Further, offsj)ring of some individuals ])ossessing B and G (BG) in their 
cells have given only two classes of offspring, those with B and those with G, 
as would be expected if the causative genes were alleles. But another type of 
BG individual has produced offspring of two quite different types — those 
with both B and G (BG) and those with neither, as if a gene producing B and 

TABLE 15.4 

THE DISTRIBUTION OF THE CONSTITUENTS OF 

THE "B" COMPLEXES IN THE OFFSPRING 

OF SELECTED SIRES 



Sire 


Antigenic 
Complex 


Number 
of Off- 
spring 


Antigenic 
Complex 


Number 
of Off- 
spring 


H-1 

H-4 

H-5 

H-6 

H-7 

H-11 

H-19 

G-19 


BbgioiTza' 

Bb02A'E3 
BbOiY2D' 

BbOi 
Bbgke'. 
Boia' 
Bgy2e; 

BlE'i 


25 
35 
26 
15 
14 
31 
19 
8 


BoiYjA' 
Bo3J'k' 
Bo3j'k' 

BgYjE'i 

BgyoeJ 

B03I'k' 

Bb 
Bb 


23 
31 
24 

23 
15 
23 
13 

7 



G together was allelic to one not effecting either B or G. These combinations 
of antigenic substances, as BG and BGK, have been called antigenic com- 
plexes. 

There are two series of such complexes, called the B and C series, respec- 
tively. In the B series there are twenty-one of the forty-odd antigenic char- 
acters which are associated in various conbinations. At least seven of these 
may appear singly, as was described for B and G. The other fourteen have 
been found only in various antigenic complexes, each of which may be made 
up of from two to eight of the twenty-one characters. The majority of these 
twenty-one characters do not occur at random in a complex with each of the 
others. As was stated above, the character K has always been found with B 
and G, but it has never occurred with I, with which it appears as a contrast- 
ing substance. In contrast, either B or G may be present in a complex with I. 
No separation of the antigenic characters of a complex has ever been ob- 
served in the cells of the offspring of an individual possessing it. A few ex- 
amples are listed in Table 15.4 from more complete data given in a paper by 
Stormont, Owen, and Irwin (1951). All present evidence makes it seem some- 
what more reasonable to assume that each antigenic complex is produced by 
a single gene than by linked genes. The various antigenic complexes in each 



254 M. R. IRWIN 

of the two systems, or series, would then be produced by a series of multiple 
alleles. The possibility of pseudo-alleles cannot be eliminated, but for the 
present may be assumed not to be a complicating factor. 

If the assumption be granted that a single gene controls an antigenic com- 
plex, as BGK, what explanation or explanations can be proposed for the dif- 
ferent antigenic specificities of this and other complexes, and, in turn, what 
can be inferred from such an explanation as to the action of the causative 
gene? 

Antigens of Pneumococci 

By virtue of the ability to attach simple chemical compounds to proteins, 
thereby preparing conjugated antigens with specifically reacting components 
of known constitution, there has emerged from such studies the realization 
that a so-called single antigenic substance may engender a multiplicity of 
antibodies of varying specificities (see Landsteiner, 1945, for a critical review 
and references). A pertinent example of this sort may be found in the anti- 
genic relationship existing between type III and type VIII pneumococci. 
Cross reactions between the respective antisera (produced in horses) and the 
two types of pneumococci have been observed, implying to them some sort of 
antigenic similarity. 

As is well known, the specificities of the pneumococcal types depend upon 
the carbohydrates of the capsules (Heidelberger and Avery, 1923, 1924). 
Thus, the carbohydrate of type III has been found to be a polyaldobionic 
acid (Reeves and Goebel, 1941). The understanding of the structure of the 
polysaccharide of type VIII is not as complete as for type III, but about 60 
per cent of the molecule of the carbohydrate of type VIII appears to be 
aldobionic acid. Cross reactivity may therefore be expected between the 
soluble specific substances of types III (S III) and VIII (S VIII), by virtue 
of the presence in each of multiples of the same aldobionic acid as a structural 
unit. It is probable that the serologically reactive unit in each of these two 
types is a larger portion of the polysaccharide molecule than a single chemical 
structural unit. Type S VIII also contains approximately two glucose mole- 
cules for every aldobionic acid residue, thereby presumably accounting for at 
least a part of the specificity of type VIII in contrast to type III. Thus it may 
be seen that serological cross reactions may be expected when the antigenic 
substances under comparison are closely related chemically. Also to be ex- 
pected is the ability to distinguish between such substances, as was actually 
possible in the case of types III and VIII (Heidelberger, Kabat, and Meyer, 

1942). 

Genetic Significance 

The above example may be combined with other findings in the field of im- 
munochemistry to allow the statement that antigenic substances of related 
but not identical chemical constitution may — but sometimes do not — incite 
the production of cross reacting antibodies. From the serological point of 



SPECIFICITY OF GENE EFFECTS 255 

view, a pertinent question concerning tiiese antigenic complexes in cattle is 
whether the cells which react with the B reagent, or with any other specific 
reagent, do so by virtue of the presence of a specific reacting substance in a 
single antigenic molecule, or otherwise? Does the comj)lex BOX, for exam])le, 
represent (1) three different and separate antigenic substances? Or does it 
represent (2) a single antigenic substance with (a) a possible common base 
and three more or less different reactive groups accounting for B, G, and K, 
respectively, or (6) a single substance capable of inciting many specificities of 
antibodies, of which those for B, G, and K represent only a part of the re- 
activities of the spectrum of antibodies which may be produced? A combina- 
tion of (a) and (b) also may be a possibility. 

At present, very little experimental evidence is available concerning the 
adequacy of any one or combination of the above possibilities to explain the 
antigenic relationships of the components of the antigenic complexes of cattle 
cells. Tests are under way to determine whether the reactive substance called 
B, for example, is the same in all cells in which it appears, whether singly or 
in an antigenic complex. 

In terms of the action of the causative genes, apart from the possibilities of 
linkage and pseudo-allelism, the question seems to resolve itself around two 
main aspects: (1) Do the genes controlling an antigenic complex, as a single 
gene for BGK, have separate specificities for B, G, and K, or (2) does this 
gene produce a single substance with no such separate specificities, and the 
similarities between such a complex as BGK and BGIY, are due primarily if 
not entirely to the general similarities in their chemical structure. The writer 
is inclined to adopt a combination of these two possibilities as a current work- 
ing hypothesis. No matter what may eventually prove to be the correct in- 
terpretation of antigenic structure of the complexes, and the action of the 
controlling genes, it appears that these studies have given some insight into 
the complexities of the gene products and perhaps also of the causative genes. 

The studies of the specificities of the gene products — the antigens of the 
blood cells of cattle — and the resulting inferences of the structure of the genes 
themselves, may not be directly related to the over-all heterosis problem. 
Nevertheless the writer is convinced that somewhat comparable specificities 
might well be obtained in plants, in which attempts are currently in progress 
to measure various aspects of the genetic bases for heterosis. Just how useful 
an additional tool of this sort would he is only a guess. 



CARL C. LINDEGREN 

Southern Illinois Universify 



Chapter 16 

Genetics and Cytology 
of Soccharomyces 



In the middle of the last century, Buchner ground up yeast cells and proved 
that the cell-free filtrate contained a substance capable of fermenting sugar. 
This experiment settled a heated controversy between Liebig and Pasteur 
concerning whether or not living structures were essential to fermentation. 
The substance responsible for the fermentation was called an enzyme, the 
word being derived from the Greek and meaning "in yeast." Since that time, 
yeast has been the organism of choice for experimenting in enzyme chemistry 
because of the abundant supply obtainable from breweries and from factories 
producing bakers' yeast. The biochemistr}^ of fermentation has provided the 
foundation for our present understanding of the biochemistry of respiration 
and of muscular contraction — two of the fundamental problems which have 
intrigued biologists. It has led to an understanding of vitamins and through 
them to an understanding of chemotherapy. 

BIOCHEMICAL DEFECTS AS GENE MARKERS 
The work of Beadle and Tatum has popularized the generally accepted 
view that enzymes are derived somehow or other from genes. Their work 
initiated a new interest in biochemical genetics. They showed that the in- 
activation of a specific gene caused a deficiency which could be met by sup- 
plying a specific chemical. Vitamins, amino acids, purines, and pyrimidines 
were the substances chosen in this analysis. They used the fungus, Neuro- 
spora, because its life cycle had been thoroughly worked out by B. O. Dodge 
and because the Lindegrens had shown by genetical analysis that it contained 
conventional chromosomes on which genes, arranged in linear order, could 
be mapped by the standard procedures used in studying corn and the fruit 

fly- 

256 



GENETICS AND CYTOLOGY OF SACCHAROMYCES 257 

YEAST GENETICS 

Until 1935, yeasts were considered to be devoid of sex and, therefore, un- 
suitable for genetical analysis. At that time, Winge showed that the standard 
yeast cell carried two sets of chromosomes — one contributed from each 
parent — and was, therefore, a typical hybrid. The hybrid yeast cell produces 
four spores, each with a single set of chromosomes. Each of these spores is a 
sex cell. By fusing in pairs they can produce the standard (hybrid) yeast cell 
and complete the life cycle. In this laboratory it was shown that the spores 
are of two mating types, and that each spore can produce a culture each cell 
of which can act as a sex cell, like the original spore. Mass matings between 
two such spore-cultures result in the production of fusion cells, from which 
new hybrids are produced by budding. 

This work made it possible to study the inheritance of biochemical de- 
ficiencies in the organism on which classical enzyme study is based, and to 
attack the problem of the relation of genes to enzymes in this fruitful mate- 
rial. We have related specific genes to several of the most thoroughly studied 
classical enzymes: sucrase, maltase, alpha methyl glucosidase, melibiase, and 
galactase. 

The principal advantages of yeasts for biochemical genetics are: 

(1) Yeast enzymes have been the subject of intensive biochemical study. 

(2) Techniques for studying respiration and fermentation are based prin- 
cipally on work with yeast and thus especially adapted to this organism. 
Yeasts grow as free cells rather than as mycelial matts and, therefore, can be 
subdivided any number of times without injury, thus simplifying weighing 
and dilution of the cells. 

(3) Large quantities of cells are available from industrial sources or can be 
grown cheaply and quickly and are easily stored in living condition. 

(4) A variety of genes concerned with the differential utilization of nu- 
merous monoses as well as di- and poly-saccharides are available. 

(5) A polyploid series of yeast cultures is now available: (a) haploid cells, 
each containing a single set of chromosomes, (b) diploid yeast cells, each con- 
taining the double number of chromosomes, (c) triploid, and (d) tetraploid 
cells (made available by our recent discovery of diploid gametes [Lindegren 
and Lindegren, 1951]). 

(6) With free cells it is possible to study competition between genotypes 
and to observe the advantages or disadvantages in controlled environments. 
The populations involved are enormous and the life cycles short, so it is pos- 
sible to simulate natural selection in the laboratory. Experiments of this type 
have enjoyed an enormous vogue with bacteria, but it has not been possible 
to distinguish gene-controlled variation from differentiation. Eor this reason, 
experiments with bacteria cannot be interpreted in terms of the comparison 
between gene-controlled and other types of inherited characteristics. 



258 



CARL C. LINDEGREN 



CHROMOSOMAL INHERITANCE 

In our selected breeding stocks of Saccharomyces, irregular segregations 
do not occur very frequently. In maize or Drosophila a similar frequency of 
irregularity would not be detectable since tetrad analysis is not possible in 
these forms. Using regularly segregating stocks of Saccharomyces we have 
mapped four and possibly five chromosomes for genes controlling the fer- 
mentation of carbohydrates and the synthesis of various nutrilites (Fig. 
16.1). 



HI 



AN 



PN 



AD I 



IN 



PY 



TH 



24 



23 



22 



10 



30 



24 
AD 2 



26 



ME 

l_ 



40 



45 



I 



22 



PB 



22 



UR 



Fig. 16.1 Chromosome maps of Saccharomyces. 

Chromosome I, PN (pantothenate), centromere, ADl (adenine), IN 
(inositol), PY (pyridoxine), and TH (thiamin). 

Chromosome II, centromere, G (galactose), AD2 (adenine), ME (meli- 
biose). 

Chromosome III, centromere, a (mating type). 

Chromosome IV, centromere, PB (paraminobenzoic acid). 

Chromosome V, centromere, UR (uracil). 

Chromosomes IV and V may or may not be different; UR and PB have 
not been used in the same hybrid. 

HI (histidine) and AN (anthranilic acid) are linked to each other (24 
morgans) but have not yet been located on a chromosome. 

DIRECT TETRAD ANALYSIS 

The focal point in the life cycle is the reduction division, at which the 
chromosomes of a diploid cell are sorted out, and the haploid sex cells (such as 
sperm, eggs, pollen, or yeast spores) are produced. Each diploid parent cell 
divides twice to produce a tetrad of four haploid sexual nuclei. This process 
is substantially the same whether a single yeast cell produces four spores or a 
cell in the testis produces four sperm. In yeast, however, each of the four 
spores of a single tetrad can produce clones which are available for individual 
study, and the reduction division can be analyzed directly instead of by in- 
ference. 



GENETICS AND CYTOLOGY OF SACCHAROMYCES 259 

Many yeast hybrids have been produced by mating sex cells carrying 
chromosomes marked with biochemical mutant genes. The tetrads from these 
hybrids have been analyzed by growing clones from each of the four spores 
of a single ascus and classifying each of the spore-cultures. These exj)eriments 
are direcl tests of the Memieliau theory. They have shown that excei)tions to the 
Mendelian theory occur more frequently than was hitherto supposed. 

CURRENT STATUS OF IRREGULAR MENDELIAN SEGREGATION 

Tetrad analysis of triploid and tetraploid yeasts has revealed that some of 
the irregular (not 2 : 2) segregations in hybrid asci arise from the fact that one 
or both of the parents is diploid (Lindegren and Lindegren, 1951; Roman, 
Hawthorne, and Douglas, 1951). Roman, Hawthorne, and Douglas have 
concluded that all irregular segregations in Saccharomyces arise from the 
segregation of triploid or tetraploid zygotes. We have recently completed the 
analysis of segregation in diploid hybrids heterozygous for both MA /ma and 
MG/mg. This analysis revealed that in many asci in which segregation of 
MA /ma was 2:2 {MA MA ma ma), segregation of MG/mg was 1:3 {MG mg 
mg mg). This finding excludes the possibility that the hybrid was either 
triploid or tetraploid since segregation of both genes would have been equally 
affected. The phenomenon has been explained as conversion of the MG gene 
to mg in the zygote. This conclusion is further supported by evidence indicat- 
ing that both genes are in the same linkage group. 

One hypothesis of the nature of the gene developed during the study of 
irregular segregation seems to have some merit. This is the proposal that the 
gene is a complex of many more or less loosely connected molecules rather 
than a single macromolecule. In this view, the gene is composed of a series of 
identical sites around the periphery of a more or less cylindrical chromosome. 
These sites may be extremely numerous since they are of molecular dimen- 
sions around the periphery of a thread easily visible under the microscope. 
At these sites identical agents responsible for the action of the gene are 
located. 

GENE DIVISION 

The concept of the gene as a bracelet of catalysts arranged on the outside 
of the chromosome simplifies the concept of gene reproduction. When one 
conceived of genes as macromolecules arranged like beads on a string, it was 
difficult to understand how all the genes on a chromosome could divide 
simultaneously. If, however, there are thousands of loci and chromogenes at 
the site of a single gene on the outside of an otherwise inert chromosome which 
is composed principally of skeletal material, any longitudinal splitting of the 
chromosome will partition two qualitatively equivalent parts which may or 
may not be quantitatively equivalent. The restoration of balance by inter- 
dependence of the autonomous organelles may make precise division unneces- 
sary. 



260 



CARL C. LINDEGREN 



EXTRACHROMOSOMAL INHERITANCE 

When a pure haplophase culture of red yeast (adenine dependent) is 
planted on an agar plate, both red and white colonies appear. When the white 
colonies are subcultured, only white colonies appear. The red cells when 
planted on a second plate continue to produce both red and white colonies. 
The white colonies are stable variants derived from red. Bacterial variations 
of this type are ordinarily called gene mutations, but bacteriologists have 
been unable to test their so-called mutations by breeding experiments except 



While Normal 
(Adenine Independent) 



Red Original Culture 
(Adenine Dependent) 




White Normal 
(Adenine Independent) 



Red Subculture 
(Adenine Dependent) 





White Variant 
(Adenine Dependent) 



Red Subculture White Variant White Subculture 

Normal X (Adenine Dependent) (Adenine Dependent) (Adenine Dependent) X Normal 



Hybrid 

(White Culture, Adenine Independent) 

ASCI Analyzed 
White (Adenine Independent) 
White (Adenine Independent) 
Red (Adenine Dependent) 
Red (Adenine Dependent) 



Hybrid 

(White Culture, Adenine Independent) 

ASCI Analyzed 
White (Adenine Independent) 
White (Adenine Independent) 
Red (Adenine Dependent) 
Red (Adenine Dependent) 



Fig. 16.2 — Inheritance of pink versus white colony in Saccharomyces. The white mutants 
derived from pink produce pink offspring and are indistinguishable from the original pink 

genetically. 

in a few cases. The white cultures have lost their color but they are still 
characteristically adenine dependent like their red progenitor. Breeding ex- 
periments (Fig. 16.2) have been carried out with the white yeast cultures 
derived from the red. When the derived white cultures were used as parents, 
they produced precisely the same kind of offspring as the original red culture 
from which they arose. This proves that the change from red to white did 
not affect a gene. The change from red to white may, therefore, be called a 
differentiation since it occurs without gene change. 

The phenomenon of Dauermodifikation which was first described by JoUos 
(1934) has thus been confirmed in yeast genetics. The stable change from 



GENETICS AND CYTOLOGY OF SACCHAROMYCES 261 

red to white resembles the discontinuous variations which occur in the vege- 
tative cycles of bacteria. Hybridization experiments have revealed that the 
origin of white cultures does not involve gene change. This phenomenon in 
yeast, called depletion mutation, is identical with Dauermodifikation in Para- 
mecia. Since neither involves gene change, both are equivalent to differentia- 
tion. 

It is not possible to study Dauermodifikationen using the classical objects 
of genetical research, maize and Drosophila, since each generation of these 
higher organisms is produced sexually — a process during which Dauermodi- 
fikationen revert to normal. The stable variants in vegetative cultures of 
yeast, which revert to normal (produce only normal offspring by sexual re- 
production) have no parallel in maize and Drosophila. This points up a 
striking disadvantage of maize and Drosophila — that they cannot be propa- 
gated vegetatively. One cannot be certain that the characteristic variations 
in flies, which occur when they hatch on wet medium or are subjected to 
shock treatment, would be lost on vegetative cultures unless one were able 
to propagate the bent wings or other peculiarities asexually, possibly in tis- 
sue culture. 

THE AUTONOMOUS ORGANELLES OF THE YEAST CELL 

In addition to the chromosomes (Lindegren, 1949) there are other perma- 
nent structures in the yeast cell which never originate de novo (Lindegren, 
1951). They have the same type of continuity in time as chromosomes but 
are less precisely partitioned than the latter. 

The Cytoplasm 

The cytoplasm is a limpid fluid which is transmitted to each daughter cell. 
It is rich in RNA but varies in basophily and contains the mitochondria, 
usually adhering to the surface of the centrosome or the nuclear vacuole. 

The Mitochondria 
The state of the mitochondria varies from highly refractile lipoidal struc- 
tures, sharply defined from the cytoplasm to less refractile organelles with 
somewhat irregular boundaries. 

The Centrosome 

The centrosome is a solid and rigid structure which stains with acid 
fuchsin but does not stain with basic dyes. This highly basic organelle may 
contain some of the basic proteins which Caspersson and Mirsky have found 
in chromosomes. The centrosome is always attached to the nuclear vacuole 
and is the most rigid structure in the cell as revealed by its behavior following 
shrinkage of the cell. It never originates de novo and plays a leading part in 
budding, copulation, and meiosis. 



262 CARL C. UNDEGREN 

The Centrochromatin 

The centrochromatin is a basophilic, Feulgen-positive substance closely 
attached to the basic centrosome (probably by an acid-base reaction). Some 
portion of it is usually in contact with the nuclear vacuole. It is partitioned 
between the cells following budding by a direct division controlled by two 
tiny centrioles. In the resting cell it may assume a spherical form and cover 
most of the centrosome. In division it is usually present in the form of a long 
strand. The centrosome and centrochromatin have been identified with the 
nucleus by several workers, but this view has been criticized by Lindegren 
(1949), Lindegren and Rafalko (1950), and Rafalko and Lindegren (1951). 
The filament often bends on itself to assume a V- or U-shape. In some 
preparations it appears to be composed of numerous small particles, but this 
is due to poor fixation and is especially prevalent in preparations fixed with 
alkali. The view that the centrochromatin is a single filament external to 
the centrosome is supported by a multitude of observations on well-fixed 
cells. Centrochromatin is probably homologous to the heterochromatin of 
higher forms differing only in being carried on the centrosome rather than 
the chromosome. 

The Nuclear Membrane and the Chromosomes 

The nuclear vacuole contains the chromosomes and the nucleolus. The 
chromosomes are partitioned between mother and bud vacuole in a precise 
orderly manner without recourse to a spindle. The wall of the nuclear vacuole 
does not break down at any time in the life cycle; it is a permanent cellular 
structure. 

The Cell Membrane and the Cell Wall 

The cell membrane is a permanent cell structure. The cell wall appears to 
be formed de novo in the spores, but it may depend on the cell wall sur- 
rounding the ascus for its origin. 

BUDDING 

Figure 16.3-1 shows a cell in which the acidophilic centrosome attached 
to the nuclear vacuole is surrounded by the darkly staining cytoplasm. A 
band of basophilic centrochromatin is securely applied to the side of the 
centrosome and is also in contact with the nuclear vacuole. Greater differen- 
tiation often reveals a small centriole at each end of this band. The nuclear 
contents are unstained. 

Figure 16.3-2 shows the first step in the process of budding. The centro- 
some produces a small conical process which forces its way through the 
cytoplasm and erupts into the new bud shown in Figure 16.3-3. 

Figure 16.3-4. The nuclear vacuole sends out a long, slender process which 
follows the centrosome into the bud. Although the cell wall is not visible in 
these preparations it must be assumed that the cell wall never ruptures but is 





.Mk 





i 








7 



8 







10 



12 . 



Fig. 16.3 — Behavior of the centrosome and centrochromalin during budding. 



264 CARL C. LINDEGREN 

extended to enclose the bud at all times. The vacuolar process follows the 
external surface of the centrosome into the bud, lying between the cell wall 
and the centrosome. 

Figure 16.3-5 shows a cell in which the bud vacuole has received its two- 
stranded chromosome complex. This is an exception to the rule that the 
chromosomes usually are completely destained in the differentiation by 
iron alum. 

In Figure 16.3-6 the bud vacuole is lobed. This is a rather common phe- 
nomenon. The cytoplasm has passed into the bud and completely surrounds 
the centrosome and the bud vacuole. The extension of the centrochromatin 
along the surface of the acidophilic centrosome has begun. 

Figures 16.3-7, 16.3-8, and 16.3-9 show cells in which the separation of 
the centrochromatin has been completed with mother and bud held to- 
gether by the centrosome. 

In Figure 16.3-10 the division of the centrosome is complete, but both 
centrosomes are near the point of budding. In Figure 16.3-11 the bud cen- 
trosome has reached the distal end of the bud while the mother cell centro- 
some still lies in the neighborhood of the point of budding. In Figure 16.3-12 
both centrosomes have reached the distal ends of the cells and are prepared 
for the formation of the next bud. 

CONCLUSIONS CONCERNING EXTRACHROMOSOAAAL INHERITANCE 

Cytological examination of the yeast cell shows that many of its organelles 
may have the same integrity and continuity in time that characterize the 
chromosomes — they cannot arise de novo. In the yeast cell there are seven 
or eight such "continuous" organelles. The cell membrane, the nuclear 
membrane, the centrosome, the centrochromatin, the cytoplasm, the mito- 
chondria, and the chromosomes are permanent cell structures. Because they 
apparently divide in a manner which does not provide for precise trans- 
mission of specific portions to each daughter cell, it appears that the other 
components differ from the chromosomes in a significant manner — they are 
probably homogeneous, or their heterogeneity is simple, possibly a few 
different types of dipolar molecules held together in a specific manner. 

There is no reason to assume that any one of these components is of more 
importance, or directs the "activities" of any one of the other components. 
The cell can function only if all its component parts are present in proper 
structural correlation and in adequate amounts. There is no reason to as- 
sume that any one of these components is unique in the manner in which it 
reproduces itself. The present hypothesis proposes that they all reproduce 
by the simple accretion of molecules like those which they contain, and it is 
their association with each other in an adequate milieu which provides the 
molecules necessary for their increase in size. Each of the different organelles 
is rate limiting in growth. When any one is present in less than the minimal 



GENETICS AND CYTOLOGY OF SACCHAROMYCES 265 

amount, the other organelles cannot obtain the supply of molecules necessary 
for maintenance and increase until the amount of the deficient organelle has 
increased. 

The chromosomes differ from the other permanent organelles in their high 
degree of linear heterogeneity. It is this characteristic which has given them 
the spurious appearance of "controlling" other cellular activities. Mutations 
with which we are familiar in the laboratory constitute defects or deletions 
in the extraordinarily heterogeneous chromosomes. The deficiency in the 
organism caused by the defect — the deletion of the contribution ordinarily 
made by the intact region of the chromosome — becomes apparent only be- 
cause the rest of the chromosome produces sufficient materials to enable the 
defective cell to continue to grow in its absence, although in a manner differ- 
ent from that which was previously characteristic. 

Any transmissible defect in a homogeneous structure like the cell wall, 
the cell membrane, the nuclear membrane, the centrosome, or the centro- 
chromatin would result in total failure of the organism to survive and bring 
all vital activity to a halt. The survival of the defective mutants in their 
altered condition due to the defect in the chromosome (which has been called 
a mutant gene) has led to the view that genes are different from other cellu- 
lar components since they can reproduce variations in themselves. This is an 
incorrect point of view. It is more proper to say that when a defect or dele- 
tion occurs in a small segment of a chromosome, the rest of the organism can 
carry on, albeit in a changed condition due to the absence of the contribution 
previously made by that region, now called the gene. This denies the im- 
portance of the ordinary mutations encountered in the laboratory as factors 
for progressive evolution, and implies that progress in evolution must occur 
in some other way. 

It may be that progressive evolution occurs more frequently as the result 
of changes in the chromosomes than of other organelles. But the present hy- 
pothesis does not exclude the possibility that advances in evolution can 
occur by ''progressive" changes in the composition of any one of the eternal 
organelles such as the nuclear membrane or the centrosome. The condition 
for the perpetuation of any change would be that the mutated organelle 
could be provided with the materials necessary for its continuance by the 
cell as a whole in its surrounding environment at the time of its occurrence. 
On this hypothesis, progressive changes in evolution are not confined to any 
single cellular component, but constitute a potential of every component of 
the cell. Although progressive changes of the different substances compris- 
ing the chromosome may not occur significantly more frequently than 
changes in the substances making up the other organelles, more changes may 
occur in the chromosomes in toto because a change in each individual com- 
ponent of the extraordinarily heterogeneous chromosome registers as a sepa- 
rate change. 



266 CARL C. LINDEGREN 

In many types of organisms the chromosomes are always separated by the 
nuclear membrane from the cytoplasm. The mitochondria (like the chromo- 
somes) are relatively non-homogeneous, but apparently the balance of their 
activities is not so critical since no specific devices appear to be required to 
limit their reproduction or activity. The cytoplasm is probably heterogene- 
ous also, with every separate eternal component having the same continuity 
in time as the chromosomes. However, it comprises substances transmitted to 
the daughter cells in a manner which is apparently subject to control by the 
environment, and this may constitute the basis for differentiation. In the 
germ line, the entire cytoplasmic potential must be maintained. In fact, the 
main function of the germ line under this hypothesis would be to maintain 
an intact cytoplasm. The integrity of the chromosomes is usually provided 
for in either the somatic or the germinal tract. Defects in the extra chromo- 
somal apparati are reconstituted in an outcross, thus differentiating so-called 
cytoplasmic from genie inheritance. 



H. H. PLOUGH 

Amhersf College 



Chapter 17 

Genetic Implications of 
Mutations in S. Typhimurium* 



The contribution that an account of studies in bacterial genetics can make 
to the problem of heterosis must be indirect, since actual sexual or other 
fusion in bacteria has not been observed and the weight of evidence is against 
the view. Even the very interesting genetic evidence of recombination dis- 
covered by Tatum and Lederberg (1947) in the K12 strain of the colon 
bacillus, and now being developed by the capable studies of Lederberg (1947, 
1949) and others, is still susceptible of other interpretations. Diploid strains, 
if they occur at all, are certainly so rare as to be unimportant in the produc- 
tion of hybrid vigor in bacterial populations. 

The apphcations of bacterial genetics to the problem of heterosis must be 
rather in the information they make available concerning the kinds and fre- 
quencies of gene mutations, and the ways in which they interact with each 
other within populations. It has been generally recognized by geneticists 
only recently that the bacteria are excellent material for studies of these 
problems, though bacterial mutation was first mentioned by Massini in 1907, 
and distinctive and precise food requirements for bacterial strains have been 
known since 1913 (Hinselwood, 1946). Studies in the genetics of bacteria 
have, of course, been greatly stimulated by the pioneer work on mutations 
in fungi by Thom and Steinberg (1939), and particularly on Neurospora by 
Dodge, by Lindegren, and by Beadle (1949) and his associates, as well as by 
the important work on yeast as presented in Dr. Lindegren's chapter. 
Long before the currently enlarging wave of interest in bacteria as objects 
of genetic study, Gowen had shown that mutations of the same order of 
frequency as in higher plants or animals were induced by radiation in Phy to- 

* This research was supported bva grant from the Atomic Energy Commission, Division 
of Biology and Medicine #AT (30-1) -930. 

267 



268 H. H. PLOUGH 

monas (1945). He and Zelle had indicated the genetic basis of virulence in 
Salmonella (Zelle, 1942). 

ADVANTAGES OF SALMONELLA FOR GENETIC STUDIES 

I became acquainted with the Enterobacteriaceae and particularly with 
the pathogenic forms in Zinsser's laboratory at the Columbia Medical School. 
My own realization that Salmonella offered excellent material for studies in 
microbial genetics was heightened when, as an Army bacteriologist in the 
Philippines, I had to diagnose enteric infections. I found most of the Salmo- 
nellas which Flexner first described from Manila still present in the islands. 
More than 140 strains or species of Salmonella are recognized which are dis- 
tinguishable by a common pattern of fermentation reactions (dextrose and 
maltose- AG, lactose and sucrose-negative, citrate and H2S positive). Each 
one has been shown by the serological studies of White (1929), Kaufmann 
(1944), or Edwards and Bruner (1942) to have a very precise and readily 
separable antigenic constitution. 

The antigens are determined by agglutination studies using serums from 
different rabbits immunized to one or another of the major strains. They 
fall into two distinct groups: the somatic (0) antigens associated with the 
surface protein layers, and the flagellar (H) antigens determined by proteins 
of the flagella. Each of these groups is known to be compound, with 
some twenty separate O antigens — each strain may carry three or four 
(O) antigens — and eight or ten different specific (H) antigens as well as cer- 
tain alternative and non-specific phases of the latter. Thus each strain can 
be shown to have a distinctive and readily determinable antigenic constitu- 
tion (S. typhimurium is I, IV, V, XII — i, 1, 2, 3). The whole group naturally 
falls into a tree-like pattern very like the evolutionary trees made for fami- 
lies of animals or plants on the basis of structure. 

Tatum's (1946) discovery that mutagenic agents (including radiation and 
nitrogen mustards) could induce mutants of colon bacteria having constant 
growth factor requirements more limited than the parental organism, just 
as with Neurospora, has re-emphasized the one gene-one enzyme hypothesis. 
It has strengthened the idea of bacterial evolution developed by Lwoff (1943) 
that the parasitic forms have been derived from the less exacting hetero- 
trophic organisms by successive losses of synthetic abilities. Thus it gives 
added meaning to the tree-like interrelationships suggested by the antigenic 
analyses. 

Soon after the war our Amherst group entered on an intensive study of 
induced biochemical and antigenic mutations in the food poisoning organ- 
ism, Salmonella typhimurium. It was our hope that this organism would 
prove more favorable for genetic studies than E. coli, not only for the analysis 
of the mode of action of genes, but for evidence on the genetic nature of type 
specificity, virulence, and their bearing on evolutionary relationships. 



GENETIC IMPLICATIONS OF MUTATIONS IN S. TYPHIMURIUM 269 

METHODS OF INDUCING AUXOTROPHIC MUTATIONS 

The strains of Salmonella typhimuriimi which we have used are two: 519 
received from the New York Salmonella center at Beth Israel Hospital, and 
533 (lie) from Gowen. 

Our method for isolating mutations to specific food or growth factor re- 
quirements by penicillin screening is that of Lederberg and Zinder (1948) 
and of Davis (1949) with some additions of our own. S. typhimurium is a 
heterotrophic organism of the least exacting sort. Cultures will grow on a 
basic medium containing ammonium sulphate, sodium chloride, potassium 
phosphate buffers, with traces of other metallic ions, and glucose added as 
an energy source. Better growth is obtained with a supplementary nitrogen 
source, such as asparagin, and a further energy source, citrate, but these are 
not essential. Thus the organism synthesizes all its own food comj)onents, 
coenzymes, and growth factors, as well as the enzymes necessary for food 
and energy tranformations. 

Suspensions are subjected to radiation by X-rays (up to 100,000 roent- 
gens) or ultraviolet light (up to 3,600 ergs per mm.-), and are then transferred 
to an enriched nutrient broth for 24 hours. The broth stimulates active 
division of all organisms. These are centrifuged off, washed, and reinoculated 
for 24 hours into the basic or minimal medium containing 100 units per ml. 
of penicillin. This stops the divisions, and progressively kills the organisms 
which divide actively. 

These organisms which penicillin screens out are called prototrophic 
(Lederberg), and they are, of course, the unchanged originals. Any mutated 
organisms w^hich now require some specific nutrilite will not divide on the 
basic medium, and so they are not affected by penicillin. These are now 
auxotrophic organisms (Davis), and they are isolated by plating on complete 
agar, and identified by paper disc inoculations on successive plates of basic 
medium with single nutrilites added — amino acids, nucleic acid fractions, or 
vitamins, as shown in Figures 17.1 and 17.2. These methods are described 
in more detail by Plough, Young, and Grimm (1950). 

AUXOTROPHIC MUTATIONS FROM RADIATED LINES 
I shall cite only one set of isolations from such a radiation experiment, the 
data for which are given in Table 17.1. Suspensions from an unradiated con- 
trol and from seven successively increased X-radiation dosages were run 
through the penicillin screening, and 500 auxotrophic mutants isolated. Of 
these a total of 459 were recovered and their specific requirements deter- 
mined. Although the control had been derived from successive single colony 
isolations within 3 days of the tests, still 5 per cent of the isolated strains were 
mutants — indicating that spontaneous mutation occurs and accumulates in 
stock strains. 

From the major strain used (#533), 234 strains out of the 459 isolated 



5 ML 



5 ML 



CENTRIFUGE 



24-HOUR 17 ML 

AGAR SLANTS NACL 




CELLS 
IN NACL 



1 ML 



I ML 



COMPLETE 
BROTH 

ISOLATIONS 



If 



3 ML 




10* 100 UNITS 

DILUTION PENICILLIN 
IN MIN. MED. 



COMPLETE 
AGAR 



MINIMAL 
BROTH 



IN 3 ML 
OF NACL 



COMPLETE 
BROTH 



Fig. 17.1 — -Diagram showing methods for the production of radiation-induced auxotrophic 
mutations in Salmonella and for their isolation by screening through minimal medium 

containing penicillin. 



PAPER DISC METHOD FOR TESTING BIOCHEMICAL MUTANTS 



TEST STRAIN 

NEEDLE TRANSFER 




24-HR BROTH DISTILLED Hg 




PAPER DISKS 



SALTS +GLUCOSE + 
AGAR + NUTRILITE 



Fig. 17.2 — Diagram showing method for determining the particular nutrilite required by 
the auxotrophs isolated as in Figure 17.1. .\ series of Petri plates is used, each containing 

a different test substance. 



GENETIC IMPLICATIONS OF MUTATIONS IN S. TYPHIMURIUM 



271 



were auxotrophic mutants, among which 17 different auxotroj^hic mutants 
occur according to tests of the specific nutrilite required. A summary of these 
requirements with the numbers of each is given in Table 17.2. Tlae most 
frequent auxotroph is the one requiring cysteine. The next most frequent is 
the histidine auxotroph, and so on down the list to one which has a double 
requirement of both valine and isoleucine for growth. Only two auxotrophic 
mutants require substances other than amino acids. One must be supplied 

TABLE 17.1 

FREQUENCIES OF AUXOTROPHIC MUTATIONS IN S. TYPHIMURIUM 
AFTER X-RADIATION AND PENICILLIN SCREENING 



1 

X-Ray Dosage 
and Time 


2 

% 
Bacteria 
Surviving 


3 

Total 

No. 

Tests 


4 

Total 

No. 

Mutants 


5 

% 
Mutants 


6 

No. 
Different 
Mutants 


7 

% 
Different 

Mutants 


Strain 533 

I Controls 

II 11,400 R4 min 

III 17,100 R6 min 

IV 22,800 R 8 min 

V 28,500 RIO min 

VI 34,300 R 12 min 

VII 45,600 R 16 min 

VIII 57,000 R 20 min 


100 

40 

25 

14 
6 

1.5 
0.9 


135 
62 
86 
41 
25 
94 
99 
50 


7 

16 
18 
11 
19 
64 
72 
34 


5.1 
25.9 
20.9 
26.8 
76.0 
68.1 
72.6 
68.0 


3 
4 
4 
3 
6 
8 
17 
12 


2.2 

6.4 

4.7 

7.3 

24.0 

8.4 

17.2 

24.0 


Totals 




459 


234 


















IX II + III + IV 




189 
268 


45 
189 


23.8 
70.5 


11 
43 


5.8 


X V+ VI 4- VII 4- VIII. . . 




16.0 








Strain 519 
XI 45,600 R 16 min 


25 


100 


22 


22.0 


9 


9.0 



with adenine, and others (not found in this experiment) must have either 
guanine or thiamin in the medium. 

In our published report of these data (Plough, Young, and Grimm, 1950, 
Table 3) we listed a number of additional strains showing alternative re- 
quirements. Davis (personal communication) retested a number of these 
and found them to be mixtures of single autotrophs. We have just completed 
an extensive recheck of all strains listed originally as alternates, and now 
confirm his results except for the three types of alternates listed in Table 
17.2 (Plough, Miller, and Berry, 1951). 

MUTATION FREQUENCY AND X-RAY DOSAGE 

One of the most interesting results of this experiment is the clear relation 
between the frequency of auxotrophic mutants and the X-ray dosage. This 
is shown in Table 17.1, column 5, lines II- VIII, and I could add to the data 
from other experiments. The numbers of tests vary for the different radiation 



272 



H. H. PLOUGH 



dosages, and some of the values are less significant statistically, but the per- 
centage of mutants is significantly higher at the higher dosages. This is em- 
phasized by lines IX and X where the sums of the first three and the last 
four values are compared. The same conclusion is evident from inspection 
of column 7 in the table, where the numbers of different mutants at the 
successive dosages are shown. Nearly three times as many were isolated from 
the upper group as from the lower. 

TABLE 17.2 
KINDS OF AUXOTROPHIC MUTATIONS IN S. TYPHIMURIUM 



No. 


Strain 533 
Single Amino Acids 


No. 


Strain 519 
Single Amino Acids 


105 

55 

15 

14 

5 

5 

4 

2 

1 


Cysteine 

Histidine 

Leucine 

Proline 

Tyrosine 

Threonine 

Methionine 

Vahne 

Arginine 


8 

3 

3 

3 

1 

1 

1 


Histidine 

Cysteine 

Methionine 

Proline 

Leucine 

Tryptophane 

Phenylalanine 


5 


Nucleic Acid Fraction 
Adenine 






1 

4 


Multiple Amino Acids 
Valine and isoleucine 
Unanalyzed 






21 

1 


Alternative A niino Acids 
Cysteine or methionine 
Tyrosine or tryptophane 


1 


Alternative A mino Acids 
Cysteine or Methionine 


1 


Tyrosine or phenylalanine 







Line XI in the table shows the result of one radiated series made on a 
different initial strain, #519. Comparison of the column 5 and column 7 
totals with line VII above, shows that this strain is much more resistant to 
radiation than is strain #533. It is clear that comparisons of the mutagenic 
effects of radiation dosage must always be made between samples from the 
same strain. 

The data in Table 17.1, column 5, are graphed in Figure 17.3. Compari- 
son of the percentages of mutants at successive dosages shows a positive 
correlation, though rather far from a straight line curve. As the penicillin 
screening method involves a 24 hour growth in complete broth, and another 
24 hours in minimal medium with penicillin, it might be expected that the 
final percentage of mutant strains would not bear the direct relation to dos- 
age shown in tests of mutations produced in germ cells in sexually reproduc- 
ing organisms. Indeed Davis, in his account of the penicillin screening method 



GENETIC IMPLICATIONS OF MUTATIONS IN S. TYPHIMURIUM 



273 



as used in E. coli, stated "... the method as developed so far does not appear 
to yield quantitative survival of mutants." Such a statement assumes that 
the penicillin screening may be expected to be complete, which in fact is not 
true. Rather penicillin acts, as do all antibiotics, in a progressive fashion ac- 
cording to a typical logarithmic killing curve. If two or more mutant cells 

PERCENT 
MUTATIONS 
100 



80 



60 



40 



20 



DATA OF TABLE 1 
A = I - VllI 
O - IX- X 



o 



4 6 8 10 12 14 16 
MINUTES OF RADIATION 



20 



Fig. 17.3 — Graph showing the relation between percentage of mutations isolated and 
X-ray dosage in minutes (2850 R per minute). 



appear in a growing wild type population, they will increase logarithmically 
and form smaller less numerous clones. As the penicillin acts, the far more 
numerous parent clones will be logarithmically reduced in numbers, while 
the mutant clones exposed will have reached a level which may be main- 
tained during the 24 hour period of penicillin action. It is clear that if a 
sample is taken, and plated at any point short of the complete killing oflf of 
the wild type, we may expect frequencies showing the same order as in the 
original population, although the mutant percentages are greatly magnified. 



274 



H. H. PLOUGH 



An actual test of artificially made mixtures of the parent strain and one 
cysteine requiring mutant as screened by the media is shown in Table 17.3. 
The data show that a mixture of 90 per cent wild and 10 per cent mutant 
still gives a greater number of wild survivors after penicillin screening than 
does a mixture having 10 per cent wild and 90 per cent mutant. For the 
actual experiments reported in Table 17.1 the proportion of mutants to un- 
mutated wild type even after 24 hours of growth in complete broth is one 
in many thousands, rather than 10 per cent to 90 per cent. So it seems justi- 
fied to consider the percentage of mutants and wild type as an index of muta- 

TABLE 17.3 

EFFECT OF GROWTH IN COMPLETE MEDIUM FOLLOWED 
BY PENICILLIN SCREENING ON ARTIFICIAL MIX- 
TURES OF CONTROL (533) AND A CYSTEINE AUXO- 
TROPH (533-169) 



Mixture 


Percentages 
Original Mixture 


Percentages 

After 24 Hrs. 

IN Broth 


Percentages 

After Subsequent 

Screening 




533 


533-169 


533 


533-169 


533 


533-169 


A 


90 


10 


70 


30 


8 


92 


B 


50 


50 


33 


67 


2 


98 


C 


10 


90 


5 


95 


2 


98 



tion frequency in comparing X-ray dosages. The trend in Figure 17.3 sug- 
gests a sigmoid curve rather than a straight line as Hollaender (1948) has 
shown for ultraviolet induced visible mutations in fungi. Essentially the 
same interpretation can be drawn from a comparison of the number of differ- 
ent mutations found at the successive X-ray dosages. Much more extensive 
data are now available showing the relation between mutation frequency 
and both X-radiation and ultraviolet dosages and they will appear in another 
publication. In general they all bear out the conclusion that the frequency 
of auxotrophic mutations is directly correlated with radiation dosage as is 
true for gene mutation in other organisms. 

A rather interesting result of comparison of these percentages of mutants 
present after penicillin screening is that the most frequent class changes from 
the lower to the higher dosages. Thus after 11,000 roentgens, a cysteine auxo- 
troph is the most frequent, while after 57,000 r it is a histidine requirer. 
Perhaps we are dealing with a specific effect of dosage or conceivably with 
a differential effect of wave length, but until the complex nature of the cys- 
teine mutants are more fully understood it is unwise to attempt too definite 
an interpretation. 

RECOMBINATION TESTS IN SALMONELLA 

Much interest has been excited among geneticists as well as bacteriologists 
by Lederberg's proof that mixtures of multiple mutant stocks of the K12 



GENETIC IMPLICATIONS OF MUTATIONS IN S. TYPHIMURIUM 275 

strain of E. coli give rise to new strains having the auxotrophic mutants in 
new combinations. These initial observations have been repeated in differ- 
ent combinations and amply confirmed by the observations of many other 
investigators. As Lederberg has suggested, these results are most reasonably 
interpreted as due to bacterial union like a sexual fusion of gametes, fol- 
lowed by an immediate reduction process involving segregation and genetic 
recombination, suggesting linkage in a single chromosome system. More 
recently Lederberg (1949) has found evidence of what appears to be a diploid 
strain which gives highly aberrant segregation ratios. These require assump- 
tions of such an extremely complex and involved type of chromosome 
interchange that it becomes questionable whether some other explanation is 
not after all more probable. 

In S. typ/iinmrium we now have more mutant strains carrying single 
auxotrophic genes or multiple combinations of these than in any other 
bacterial species except E. coli. This makes it especially important to test the 
theory with our strains. Accordingly Miss Marie McCarthy has been mixing 
these in varying combinations, and then plating out in heavy suspensions on 
base medium supplemented so as to show up the transfer of one or more re- 
quirements from one to the other original combination. 

Although more than a hundred such tests have been made and carefully 
checked, the results have been unequivocally negative until very recently. 
This work will be reported in detail in a later publication, but I will describe 
it briefly here. Multiple strain #519-38-94-41 requiring tryptophane, me- 
thionine, and histidine was mixed with #533-486-96-85 requiring leucine, 
threonine, and arginine. On plating in appropriate media it was found that 
in addition to the original parental combinations several colonies each gave 
strains requiring two new sets of requirements. Recombination No. 1 re- 
quired tryptophane, leucine, and threonine. Recombination No. 2 needed all 
six amino acids: tryptophane, methionine, histidine, leucine, threonine, and 
arginine. These new stocks have been retested, and there can be no question 
of the fact that we have here two recombinations of the original stocks used. 
Other recombinations have now appeared but reciprocal classes are never 
found. Thus we have in Salmonella confirmation of the recombination results 
found by Lederberg in the K12 strain of E. coli. In view of the irregularity of 
such results both in E. coli and in Salmonella, it would seem wise to suggest 
that some alternative explanation may yet prove to be more satisfactory than 
recombination or chromosomal crossing-over. 

BIOCHEMICAL STUDIES OF AUXOTROPHIC MUTANTS 

The Neurospora studies of Eeadle and his associates as well as those of 
Lindegren (1949) on yeast have made it evident that in studying the action 
of auxotrophic mutants we are many steps closer to the initial determinative 
activities of the genes themselves than is ordinarily true for characters in the 
higher plants and animals. When a series of auxotrophic genes can be shown 



276 



H. H. PLOUGH 



to block successive steps in the syntheses of particular amino acids or vita- 
mins or more complex products, the one gene-one enzyme hypothesis offers 
the most satisfactory preliminary explanation, even though the presence of 
the particular enzyme as a gene product has not been demonstrated. Each 
set of auxotrophic mutants offers data on the chain of synthetic processes to 
some essential substance, and thus becomes a challenging biochemical prob- 
lem. It is significant that many of those already studied in the fungi have 
also been uncovered in E. coli, but every organism shows individual differ- 
ences. So far in Salmonella we have investigated the biochemical steps in 
only two such series of auxotrophs, but many others await study especially 
as new mutants are added. 

TABLE 17.4 

UTILIZATION OF SULPHUR COMPOUNDS BY VARIOUS 
AUXOTROPHS OF 5. TYPHIMURIUM 



Strain 


NaiSOi 


NajSiOs 


NajS 


Cysteine 


Cysta- 
thionine 


Methio- 
nine 


Block 
in Fig. 17.4 


1. Original 533 

2. 533-575 


+ 




+ 
+ 


+ 
+ 
+ 


I 1 -f + + + 


+ 
+ 
+ 
+ 
+ 


+ 

+ 

+ 
+ 


None 
7-1-2 


3. 533-526 


7 


4. 533-452 


5-h2 


5. 533-P249 

6. 533-535 


4 
2 







The first of these sets of interacting synthetic steps which we have studied 
is the cysteine-methionine auxotroph series. These mutants fall into many 
of the same gradations described by Lampen, Roepke, and Jones (1947) for 
E. coli, by Emerson (1950) for Neurospora, and by Teas (1950) for B. 
subtilis. We have tested all of the apparent cysteine or methionine requirers 
for their ability to reduce inorganic sulfur compounds as well as to utilize 
organic precursors of methionine. The wild type strains can reduce sulphate, 
sulphite, or sulfide, and can grow with no other source of S. It has been 
shown, however, that none of the apparent cysteine requirers can reduce 
sulphate, but some can reduce sulphite and some sulfide. Many, however, 
must have cysteine or cystathionine (kindly supplied by Dr. Cowie) and 
others require methionine as such. 

A summary of representative mutants isolated as cysteine or methionine 
requirers and their abilities to grow on various compounds as the sole source 
of S is given in Table 17.4. This can be visualized as in Figure 17.4 in terms 
of a succession of steps, each catalyzed by an enzyme controlled by a gene 
which is inactivated by the mutation numbered in parentheses. Such a 
straight line series appears to run in the direction of the arrows from sulphate 
to protein. When a mutation occurs, as at (5), it must be assumed that growth 
requirements will be satisfied by any compound succeeding the break in the 



GENETIC IMPLICATIONS OF MUTATIONS IN S. TYPHIMURIUM 277 

synthetic chain, unless a second mutation has occurred. This does not hold 
for methionine which cannot be utilized in mutants #2 and 4 (Table 17.4). 
Such a result suggests that cysteine is enzyme controlled through a gene 
which is inactivated by the mutation numbered in parentheses. Cysteine is 
ordinarily made from methionine (as has been shown for the mammal) and so 
the reverse dotted arrow marked (1) is shown in the figure. It is hardly likely 
that a second mutation is indicated for the mutants cited as showing two 
blocks, but rather that certain mutations cause inhibition of more than one 
enzyme system. A more comprehensive scheme for the cysteine-methionine 
synthesis based on the Neurospora work has been given by Emerson (1950). 
It is certain that more is involved in the series of reactions shown in Figure 
17.4 than the furnishing of essential sulphur for cysteine and methionine. 

^ Protein 

(7) (6) (5) 
SO4 »> SO3 *► S Cysteine 

HOMOSERINE 
(1)/ \ (4) 

y Cystathionine 

? ^ d) \ 

Protein -< Methionine < Homocysteine 



Serine 

Fig. 17.4 — Possible chain of reactions involving sulphur-containing compounds. (Mutant 
blocks indicated by numbers in parenthesis.) 

Sulphate, sulphite, and suliide, as well as cysteine itself, may act as H 
acceptors, cooperating with dehydrogenases involved in the respiratory or 
energy producing activities of the organism. That the organism reduces more 
sulphate than is necessary for the S required in the amino acids is indicated 
by the fact that Salmonella forms a readily testable excess of H2S. We are 
attempting to trace the course of the sulphur by the use of the radioactive 
isotope S^''. Last summer Dr. T. P. Ting and the writer were able to show that 
(NH4)2S*04 is taken up by the wild type 533 organisms and not at all by a 
cysteine requiring mutant, thus confirming our growth tests (see also Cowie, 
Bolton, and Sands, 1950). 

We hope to continue this work using labeled sulphur in sodium sulfide or 
barium sulfide, which should be utilized by wild and mutants number (7), 
(6), (5) (Fig. 17.4). Finally it should be possible to determine by quanti- 
tative tests how much S^* is combined into bacterial protein and how much 
passed out in HoS. Comparisons between different strains in oxygen utiliza- 
tion are being made with the W'arburg respirometer. As already shown in 



278 



H. H. PLOUGH 



Table 17.3, some cysteine requiring strains will overgrow the parent and this 
may be due to differences in energy requirements. 

The second set of steps in synthesis being studied concerns the adenine 
requirer. Here we appear to have rather more definite information than was 
described by Guthrie (1949) for the purine auxotrophs of E. coli. It has been 
shown that the Salmonella auxotroph utilizes adenine and hypoxanthine, but 
not guanine and xanthine. Of the nucleosides and nucleotides only adenosine 
and adenylic acid are used, and much more of the latter is required for com- 
parable growth than of adenine. Thus it appears that in purine metabolism, 
Salmonella and an animal like Tetrahymena (Kidder and Dewey, 1948) 
show almost opposite requirements, for the bacteria do not convert adenine 
to guanine. Preliminary studies by Mrs. Helen Y. Miller demonstrate a spar- 
ing action for adenine utilization by the amino acid histidine. This suggests 



N==C-NH2 



HC— NH 



HC— NH 



H-C 



C-NH 



N 



— C-N 

Adenine 



^ 



CH 



C— N 
CH2 

o=c 



^ 



CH 



-^- 



C — N 
CH2 
H2N-CH 



/- 



:cH 



COOH 

Imidazole Pyruvic Acid 

+ 
Pyridoxamine 



COOH 
Histidine 



Fig. 17.5 — A Possible relation of Adenine to Histidine synthesis (after Broquist and Snell). 



that for this organism as with Lactobacillus (Broquist and Snell, 1949) the 
purine is a precursor of histidine, probably by the utilization of the imid- 
azole ring through pyruvic acid, and the transaminating action of pyridoxa- 
mine (Figure 17.5). While these facts have been revealed by a study of the 
adenine mutant alone, further gene changes and their reactions with the 
histidine auxotrophs already available should help clarify some of the inter- 
actions of purines and amino acids in the bacterial cell. 

ALTERATION OF ANTIGENIC SPECIFICITY 

The auxotrophic mutations reveal a series of biochemical steps or trans- 
formations common to whole groups of organisms. Antigenic analysis, on the 
other hand, has revealed precise specific or strain differences which are as 
distinctive as the form or structural differences of complex animals and 
plants. This has been clearly demonstrated by the blood group analysis 
presented in the studies of Irwin and his colleagues. The specificity is no less 



GENETIC IMPLICATIONS OF MUTATIONS IN S. TYPHIMURIUM 279 

sharp in the antigenic analyses of Salmonella. The tree-like relationship 
which they suggest was our chief stimulus to a study of bacterial genetics in 
this organism. 

Preliminary tests of all of the auxotrophic mutants made by Miss Dorothy 
Farley show that they are unchanged antigenically. Not only the specific 
antigens, but the agglutination titers are the same as the original strains. 
This has been confirmed by reciprocal absorption tests, as well as by precipi- 
tation, and inhibition of agglutination using supernatants from boiled cul- 
tures. Thus it appears that the loss of ability to synthesize a particular amino 
acid in no way alters the antigenic configuration. Apparently if proteins are 
formed at all they take on the antigenic configuration of the cytoplasm al- 
ready there. The auxotrophic mutants and the antigenic patterns fall into 
two quite independent systems so far as present evidence goes. This seems 
to be true also for variations in or loss of virulence. The relation of the auxo- 
trophic mutants to virulence for mice is being studied in detail by Gowen 
and his associates and will be reported separately, but so far at least it ap- 
pears that there is no relation between virulence and the biochemical re- 
quirements of the strain. 

It was originally and is still our hope to be able to induce antigenic 
variants by radiation, but so far such attempts have given negative results. 
We have inoculated radiated suspensions into one end of U tubes of semi- 
solid agar containing low concentrations of O serum from a rabbit im- 
munized against the specific strain, and the organisms grow through the 
medium. When agar containing specific H serum is used, however, the 
organisms grow only at the site of inoculation. If antigenic mutants had 
occurred we would expect that the homologous serum would act as a screen 
to block off the original and let the mutants through, just as the penicillin 
does for the auxotrophs. The result simply means that we have not found any 
antigenic mutants following radiation. Perhaps we should not expect any. 
Antigenic mutants have been induced in several bacteria by other meth- 
ods, especially by McCarty (1946) in the pneumococcus, by Bruner and 
Edwards (1947) in Salmonella, and by Boivin (1947) in E. coU. The pneu- 
mococcus method is not applicable to Salmonella, and the Boivin method in- 
volving exposure of the organism to autolysates of rough variants of other 
strains gives negative results. Tests using similar culture filtrates have 
been unsuccessful in altering the antigenic constitution of our organism. On 
the other hand, Miss Farley has made use of the Edwards technique of grow- 
ing an auxotrophic mutant in a semi-solid medium containing homologous O 
serum previously absorbed with a related organism which lacked one of the 
major antigens, XII (and in another case lacked V but carried an additional 
antigen XXVII). By this method two successful transformations of type 
have been secured out of several tried. Both of these transformations were 
performed on an auxotrophic mutant (519-PlO) requiring histidine. 



280 H. H. PLOUGH 

Preliminary tests showed that these strains were antigenically similar and 
gave the same agglutination titer with homologous serum as the parental 
wild types — (I) IV, V, XII, for the O antigens. The parentheses indicate that 
(I) is very weak or absent. The first case is typical. Specific serum from ani- 
mals immunized by #519 was absorbed with a suspension of organisms of 
#527, an unnamed strain known to have antigens IV, V only. After it was 
passed through semisolid agar containing the absorbed serum now carrying 
XII antibodies only, 519-PlO was retested and shown now to give agglutina- 
tion at a very low titer (1/320 instead of 1/10,000) compared with the origi- 
nal. Further testing has demonstrated that this strain retains the two major 
antigens (IV and V), but has lost XII. Thus it has been transformed to IV, 
V like strain #527. Further tests on differential media prove that the strain 
is unchanged as an auxotrophic mutant, and still cannot grow unless the 
medium contains histidine (519-10). 

In the other case 519 serum was absorbed by S. schleissheim (V, XII, 
XXVII). The mutant after growing through the absorbed serum failed to 
agglutinate in XII serum, and had a higher titer in XXVII than S. schleiss- 
heim. Thus the changed mutant has lost XII and taken on antigen XXVII. 
It still retains its histidine requirement. 

Thus we have two independent cases of the alteration of antigenic speci- 
ficity by the Edwards method of passage through specific serum. Here again 
the evidence indicated no relation between antigenic configuration and the 
biochemical requirements. We are now exposing these antigenically altered 
strains to further radiation with the idea of building up multiple auxotrophic 
stocks combining the two major systems of mutations. These can then be 
used for more conclusive tests of possible fusion and recombination. How- 
ever, this demonstration that antigenic mutants can be induced by specific 
serum adds to the possibility that mutual interaction of genes or gene 
products between organisms in mixtures may give a more acceptable 
explanation of the recorded cases of recombination in bacteria, than does one 
based on genetic analogies with higher forms. 

SUMMARY 

An account has been given of the results of X-radiation of suspensions of 
the two strains of Salmonella lyphimurium, and the isolation of strains with 
specific nutrilite requirements (auxotrophic mutants). These strains are iso- 
lated by the Davis-Lederberg method of growth for twenty-four hours in 
enriched broth, followed by twenty-four hours in minimal broth containing 
100 units per ml. of penicillin. The method screens out the unmutated organ- 
isms according to a logarithmic survival curve, and preserves the mutant 
bacteria. 

Successive tests show a relation between X-ray dosage and the percentage 
of recovered auxotrophic mutants, and also between dosage and the number 
of different mutants. 



GENETIC IMPLICATIONS OF MUTATIONS IN S. TYPHIMURIUM 281 

In all, 249 separate auxotrophic mutants, of which 20 are different, were 
isolated out of 459 tests. Most of these showed requirements for single amino 
acids, but a few required the purine base adenine, and others showed alterna- 
tive, and a small number, multiple requirements. 

A large number of tests involving growth of multiple mutant stocks in 
mixtures followed by re-isolations have been made to test for possible fusion 
and recombination as reported by Lederberg and others in the K 12 strain 
of E. coli. Recombination has been found but it is unlike that in sexually 
reproducing organisms. 

Detailed studies of the different auxotrophs requiring cysteine or me- 
thionine show a step-like series beginning with loss of ability to reduce in- 
organic sulphate, and continuing to the loss of ability to form methionine. 
Many of these mutational steps are explainable as due to the inactivation of a 
specific enzyme, but several require a complex pattern of chemical interac- 
tions. 

Similar studies of the adenine auxotroph suggest that adenine may be a 
source of histidine. 

Tests have been made to determine if antigenic specificity can be altered 
by radiation, with negative results. However, an auxotrophic mutant has 
been antigenically altered in two difTerent cases by the Edwards technique of 
passing through absorbed immune serum. In each case, one of the antigens 
was removed, and in one case another antigen was added. In both cases the 
biochemical requirement of histidine was retained. 

It appears that the auxotrophic and antigenic series represent two quite 
different and unrelated sets of mutations. 



JAMES F. CROW 

Universify of Wisconsin 



Chapter 1 8 

Dominance 
and Overdominance 



Since the first attempts to explain hybrid vigor and the deleterious effects 
of inbreeding in Mendelian terms, there have been two principal hypotheses. 
Both were advanced early, and though each has had its ups and downs in 
popularity, both have persisted to the present time. The first hypothesis is 
based on the observed correlation between dominance and beneficial effect 
(or recessiveness and detrimental effect). Inbreeding uncovers deleterious 
recessives, and typically results in deterioration. 

With hybridization, some of the detrimental recessives brought into the 
hybrid zygote by one parent are rendered ineffective by their dominant 
alleles from the other, and an increase in vigor is the result. If the number 
of factors is large, or if there is linkage, the probability becomes exceedingly 
small of a single inbred line becoming homozygous for only the dominant 
beneficial factors. Consequently, there should be a consistent decrease in 
vigor with inbreeding, and recovery with hybridization. This idea has been 
called the dominance or the dominance of linked genes hypothesis. 

The alternative theory assumes that there is something about hybridity 
per se that contributes to vigor. In Mendelian terms this means that there 
are loci at which the heterozygote is superior to either homozygote, and that 
there is increased vigor in proportion to the amount of heterozygosis. This 
idea has been called stimulation of heterozygosis, super-dominance, over- 
dominance, single gene heterosis, cumulative action of divergent alleles, and 
simply heterosis. 

In accordance with the title of this discussion I shall use the words domi- 
nance and overdominance for the two hypotheses. This leaves the word 
heterosis free for more general use as a synonym for hybrid vigor (Shull, 1948). 

* Paper No. 434 from the Department of Genetics, University of Wisconsin. 

282 



DOMINANCE AND OVERDOMINANCE 283 

In most situations, the hypotheses of dominance and overdominance lead 
to the same expectations. In either case there is a decrease of vigor on in- 
breeding and a gain on outcrossing. Wright (1922c) has shown that with the 
dominance hypothesis the decline in vigor is proportional to the decrease in 
heterozygosis, regardless of the relative number of dominant and recessive 
genes and of the degree of dominance. The same decline in vigor with de- 
creasing heterozygosity is true with overdominance. 

It is usually impossible in a breeding experiment to differentiate between 
true overdominance in a pair of alleles, and pseudo-overdominance due to 
the effects of two pairs of alleles closely linked in the repulsion phase. Only 
in special circumstances, such as when a mutation has recently occurred in 
an isogenic stock, can the experimenter be reasonably certain that the effect 
is due to a single allelic difference. Furthermore, there is the possibility of 
heterosis due to borderline situations, such as might arise in pseudoalleles 
with a position effect, which could not even theoretically be classified as due 
to dominance of linked genes or overdominance. Finally, it should be noted 
that the various hypotheses may not be equally important in all situations. 
For example, it is reasonable to expect that overdominance would be more 
important in determining differences between inbred lines of corn pre- 
viously selected for general combining ability than in lines not so selected. 

If the two hypotheses are not mutually exclusive, neither are they col- 
lectively exhaustive. There is no reason to think that multiple factors are 
any less complex in their interactions than factors concerned with qualitative 
differences. With the number of genes involved in heterosis, and with the 
complexity of interactions known to exist in cases where individual gene 
effects have been isolated and studied, there must surely be all sorts of com- 
plex interactions in heterosis. Therefore no single theory can be expected to 
account for the entire effects of heterosis. Although it is difficult to separate 
by statistical methods the effects of dominance and epistasis, it may be 
possible to construct simple models which are of some utility. 

DOMINANCE 

Davenport (1908) was the first to point out the now well-recognized fact 
that in most cases the dominant character is beneficial to the organism pos- 
sessing it, while the recessive has a weakening effect. He noted that this could 
help explain the degeneration that usually follows inbreeding. Davenport 
was thinking of relatively few factors with individually large effects, whereas 
at present, more emphasis is given to multiple factors. But he was close to 
the ideas now held. 

Keeble and Pellew (1910) found that hybrids between two pure varieties 
of peas were taller than either parent. In this case, two different dominant 
factors were involved — one resulting in longer internodes and the other in- 



284 JAMES F. CROW 

creasing their number. Here only two gene pairs were involved, but it was 
mentioned that similar systems might hold for more complex cases. 

A more general development of the dominance hypothesis was given dur- 
ing the same year by Bruce (1910). He designated the frequencies of domi- 
nant and recessive alleles as p and q in one breed and P and Q in the other. 
The array of individuals in the two groups will then be {p'-DD -\- IpqDR -\- 
q^RRY and {PWD + 2PQDR + Q-RRY, where D and R are the dominant 
and recessive alleles and n is the number of factor pairs involved.' If these 
two populations are crossed, the mean number of homozygous recessive loci 
is nqQ, whereas the average number for the two parent populations is 
n{q- -f Q~)/2. The former is the geometric mean of the two parental recessive 
genotype frequencies while the latter is the arithmetic mean. Since the geo- 
metric mean is always less than the arithmetic, the number of homozygous 
recessive loci will always be less in the hybrid population than the mean 
number in the two parent populations. If either or both the parent popula- 
tions are inbred the decrease will be greater. 

Bruce then said: 

If, now, it be assumed that dominance is positively correlated with vigor, we have the 
final result that the crossing of two pure breeds produces a mean vigor greater than the col- 
lective mean vigor of the parent breeds. ... I am aware that there is no experimental evi- 
dence to justify the assumption that dominance is correlated with a "blending" character 
like vigor; but the hypothesis is not an extravagant one, and may pass until a better takes 
the field. 

The average proportion of recessive homozygotes in the parents, which is 
(9^ + '2")/2, may be rewritten as qQ -{- {q — Q)~/2. This is always larger 
than qQ, the proportion in the hybrid, unless q and Q are equal. Although 
Bruce didn't mention this, after one generation of random mating the propor- 
tion of recessives in the hybrid population becomes {q + QY/'^ = (lQ'\~ 
(q — QY/4-, which shows that half the gain in vigor is lost as soon as ran- 
dom mating begins. 

Bruce concentrated his attention on the decrease of homozygous reces- 
sive loci in the hybrid, and postulated a correlation between recessiveness 
and deleterious effect. He could have used the same algebraic procedures to 
show that crossing produces an increase in heterozygous loci, and thus based 
a theory of hybrid vigor on overdominance. He showed remarkable foresight 
in choosing the former, at a time when he had no evidence of a correlation 
between dominance and beneficial effect. 

1. The notation used by Bruce implies equal frequency of dominant and recessive 
alleles at all loci. This assumption is not at all necessary for the argument, and I think 
that what Bruce really meant was 



Yl {p]DD-^2p,q,DR-^qfRR). 



DOMINANCE AND OVERDOMINANCE 285 

Objections to the dominance hypothesis were made largely on two 
grounds. First, if vigor is not a product of heterozygosity as such, it should 
be possible by selection to obtain individuals which are homozygous for all 
the beneficial dominant factors, and hence have the same vigor as hybrids. 
Secondly, in the Fo of a cross between two inbred strains there should be a 
skew distribution of the trait being measured — since the dominant and re- 
cessive loci would be distributed according to the expansion of (3/4 + 1/4)", 
where n is the number of factors. 

These objections were largely removed when Jones (1917) pointed out 
that, with linkage, the consequences of the dominance hypothesis were 
much closer to those postulating superior heterozygotes. If a detrimental 
recessive were linked with a favorable dominant, the heterozygous chromo- 
some would be superior to both homozygotes, and the linked combination 
might not break up readily. Later, Collins (1921) showed that with a large 
number of factors, regardless of linkage, the skew distribution disappears. 
The probability of getting all the beneficial dominants into one homozygous 
strain becomes vanishingly small, so the objections hold only if a small num- 
ber of factors is assumed. 

Most of the mutations known in Drosophila and elsewhere are recessive, 
and practically all are in some way deleterious. Even if dominant and re- 
cessive mutations were occurring with equal frequency, the deleterious mu- 
tations in a population at any time would be mostly recessive, since the domi- 
nants would be rapidly eliminated. It is to be expected — and it has been often 
observed — that at most unfixed loci the recessive is deleterious in compari- 
son with its dominant allele.^ 

Almost thirty years ago Sewall Wright (1922c) wrote: 

Given the Mendelian mechanism of heredity, and this more or less perfect correlation be- 
tween recessiveness and detrimental effect, and all the long-known effects of inbreeding — - 
the frequent appearance of abnormalities, the usual deterioration in size, fertility, and con- 
stitutional vigor in the early generations, the absence of such decline in any one or all of 
these respects in particular cases, and the fixation of type and prepotency attained in later 
generations — are the consequences to be expected. 

It has been shown many times that populations actually contain a large 
number of detrimental recessives — sufficient to account for a large decline in 
vigor on inbreeding. In Drosophila pseudoobscttra, Dobzhansky et al. (1942) 
found that almost every fly examined had at least one concealed lethal. Fur- 
ther evidence that at least some heterosis is due to dominant favorable genes 
is provided by the experiments of Richey and Sprague (1931) on convergent 
improvement in corn. 

2. I consider the statement that a dominant is beneficial and the statement that a reces- 
sive is deleterious as meaning the same thing. Since a geneticist ordinarily can study gene 
effects only by substituting one allele for the other, he cannot distinguish what each factor 
is doing individually or whether it is harmful or beneficial except relative to its allele. That 
is, he can only tell what the effect of the substitution is. 



286 JAMES F. CROW 

OVERDOMINANCE 

The concept of a stimulating effect of hybridization began independently 
with Shull (1908, 1911b) and East (1908). It was assumed that there was a 
physiological stimulus to development which increased with the diversity 
of the uniting gametes — with increasing heterozygosis. East (1936) elabo- 
rated the idea further by postulating a series of alleles each having positive 
action functions, and with these functions to some extent cumulative. As the 
alleles became more and more divergent in function, the action was postu- 
lated to become more nearly additive in the heterozygote. 

At the time when East and Shull first formulated the hypothesis, there was 
no direct evidence of any locus at which the heterozygote exceeded either 
homozygote. For a number of years, overdominance as an explanation of 
heterosis largely was given up because of the failure to find such loci. 

Stadler (1939) pointed out that in certain of the R alleles in corn a situa- 
tion obtains in which certain heterozygotes have more areas pigmented than 
either homozygote. He suggested that genes acting in this manner could re- 
sult in overdominance for such characters as size and yield. Other such loci 
are known in corn. 

There are now several cases in the literature of single genes with heterotic 
effects. In most of these it is not possible to rule out the possibility of close 
linkages giving pseudo-overdominant effects. In particular, many cases may 
turn out to be pseudoallelism, but the consequences for the animal or plant 
breeder would not be changed. 

Several workers (Teissier, 1942a; Robertson, unpublished) have found per- 
sistent lethals in Drosophila population cage experiments. If these are not 
due to individually heterotic loci, extremely close linkage must be postu- 
lated. Also certain recessive genes, such as ebony, come to an equilibrium 
with their normal alleles in population cages. One of the most convincing 
cases is that of the eye color mutant described by Buzzati-Traverso in this 
volume. This mutant persists in the population, and was found in three in- 
dependent stocks. It is quite improbable that in each of these cases the gene 
happened to be linked in the repulsion phase with another harmful recessive. 

The idea of superior heterozygotes has been upheld by Hull (1945) who 
suggested the word overdominance. Hull's original argument for overdomi- 
nance is a simple one. He noted that in most cases the hybrid between two 
inbred maize lines has a greater yield than the sum of the two inbreds. This 
would not be possible with dominant genes acting in a completely additive 
manner — unless it were assumed that a plant with no favorable dominants 
had a negative yield. 

The validity of this argument depends on the unimportance of epistasis 
in corn yields. Evidence on this point is very incomplete and somewhat con- 
tradictory. Neal (1935) reported that the Fo yields were almost exactly inter- 
mediate between the Fi and the average of the parents. This would suggest tha t 



DOMINANCE AND OVERDOMINANCE 287 

epistatis is not important or else that there is some sort of cancelling out of 
various effects. On the other hand, Stringfield (1950) found that in many 
cases backcrosses showed consistently higher yields than the F2. This sug- 
gests some sort of interaction, as if some of the gene combinations selected for 
during the inbreeding process were active in the backcross, but were broken 
in the F2. None of these data give any evidence as to the importance of 
epistasis in determining the difference between an inbred line and a hypothet- 
ical line with none of the favorable dominants, since the data do not extend 
into this range. It is in this range where non-additivity might be expected 
to be most pronounced. 

Hull's second argument is based on results obtained by the technique of 
constant parent regression. The regression of Fi on one parent, with the other 
parent held constant, has different expectations when there is overdominance 
than when there is dominance. With overdominance the regression may be 
negative when the constant parent is high-yielding, so the regression surface 
is different from that expected with dominance. In this volume Hull gives 
data which conform with this expectation. 

Overdominance is not the only possible explanation of such results, as 
Hull has pointed out. In addition, the constant parent regression technique, 
or any technique making use of yield data on inbred lines, is complicated by 
the difficulty of obtaining consistent results with inbreds. Another possi- 
bility is that the factors responsible for yield in inbreds are largely different 
genes from those determining the yield in the hybrids. This possibility will 
be considered later. 

For these reasons it is still not possible to be sure of the importance of 
overdominance from Hull's methods. They are at least strongly suggestive, 
and recent data from Robinson et al. (1949), obtained by an entirely differ- 
ent procedure, also gave evidence of overdominance. 

MAXIMUM HETEROSIS WITH THE DOMINANCE HYPOTHESIS 

In this discussion several assumptions are made. Most of these have been 
implicit in most discussions of heterosis, but it is best that they be clearly set 
forth at the outset. The assumptions are: 

1. Genes concerned with vigor are dominant, and in each case the domi- 
nant allele is beneficial and the recessive deleterious. This is an assumption 
of convenience which does not alter the essential nature of the hypothesis. 
The conclusions still hold if dominance is not complete. Also there are loci 
in which the recessive is advantageous or in which the heterozygote is inter- 
mediate; but these are of no consequence for heterosis and therefore can be 
omitted from the discussion. 

2. There is complete additivity of effects between loci — no epistasis. 

3. There are no barriers to recombination that prevent each gene from 
reaching its own equilibrium frequency independently of other loci. 



288 JAMES F. CROW 

4. The gene and phenotype frequencies of the parent population are at 
their equilibrium values. 

5. Increased vigor results in, and can be measured in terms of, increased 
selective advantage, though the selection may be natural or artificial. This 
assumption restricts the discussion to those cases in which heterosis results 
in changes in the same direction as selection had previously been acting. 
Such an assumption appears to be valid for yield characters in field crops, 
and for viability and fertility as is measured in Drosophila population 
studies. It is highly questionable for such things as increase in size of hybrids 
between wild varieties or species, where natural selection pressure may well 
have been toward an intermediate size. 

Under this assumption the increase of vigor on hybridization depends di- 
rectly on the number of loci which are homozygous recessives in the parent, 
but which become heterozygous in the hybrid. The individual or population 
of maximum vigor is one in which every allelic pair contains at least one domi- 
nant. The actual attainable heterosis would be less than this in any particu- 
lar case. 

Consider the case of complete dominance. The recessive phenotype is as- 
sumed to have a selective disadvantage of s. That is, the dominant and re- 
cessive phenotypes are surviving and reproducing in the ratio of 1 to 1 — 5. 
The rate of mutation from A to a is u per gene per generation. Reverse muta- 
tion will be ignored as it can be shown to have a negligible effect on the 
equilibrium gene frequency attained. 

Genotype A A A a a a 

Frequency P 2Q R 

Selective value 1 1 1 — 5 

F-\-2Q-\-R= I 

Under these assumptions, the frequency of gene A will be P -j- ^, while 
the frequency of a will he Q -\- R. With mutation from A to a at rate u, the 
frequency of .4 will be reduced in one generation by u{P -f- Q) and the 
frequency of a increased by the same amount. Likewise, due to the effect 
of selection, the frequency of a will be decreased by sR. Therefore the gene 
ratio, {P -j- Q)/(Q + R), will change in one generation due to the effects 
of mutation and selection to 

iP + Q){l -u) 

{P-hQ)u-\-Q-\-R- sR' 

When equilibrium is reached the gene frequency will no longer change from 
generation to generation which, stated algebraically, is 

P^^ ^ {P + Q){l-u) 

Q^R {P-\-Q)u-\-Q-\-R- sR- 



DOMINANCE AND OVERDOMINANCE 289 

This has the sokition, R = u/s. (For a more pedantic demonstration of this, 
see Crow, 1948.) 

The average reduction in selective value of the population due to a detri- 
mental factor will be the product of the selective disadvantage of the factor 
and the proportion of individuals possessing the factor. This amounts to 
(s) (ti/s), or, simply, u, the mutation rate. Hence, the effect of a detrimental 
gene on the selective value of the population is equal to the mutation rate to 
that gene, and is independent of the selective disadvantage which that factor 
causes, as was first pointed out by Haldane (1937). This fact, which at first 
appears paradoxical, is readily understandable when one notes that a mildly 
deleterious mutant persists much longer in the population, and hence affects 
many more individuals than one which has a greater harmful effect. 

The total effect on the population of all the loci capable of mutating to 
deleterious recessives is simply the sum of the individual mutation rates as 
long as the gene effects are additive. If there are n such loci with an average 
mutation rate of u, the net reduction in selective value due to all homozygous 
detrimental recessives at all loci in which they occur is nil. This is also ap- 
proximately correct if the factors are multiplicative, provided the individual 
effects are small. 

The product ml is probably in the vicinity of .05 (Crow, 1948). This means 
that if all the deleterious recessives were replaced by their dominant alleles, 
the selective advantage of an equilibrium population would be increased by 
about this amount. This could be considered as the maximum average im- 
provement in vigor, as measured in terms of selective advantage, that could 
occur due to hybridization. This means that the dominance hypothesis can- 
not, under the conditions postulated, account for average increases of more 
than a few per cent in vigor. 

There are several reasons why the 5 per cent figure given above may be 
too large. One is that many deleterious factors considered to be recessive 
may not be completely recessive. Stern and Novitski (1948) and Muller 
(1950) have shown that the majority of lethals and detrimentals that occur 
in laboratory cultures of Drosophila are not completely recessive. Even if the 
detrimental effect of the heterozygote is much less than that of the homozy- 
gote, the greatest selection effect will still be on heterozygotes because of 
their much greater frequency in the population. Thus, from the population 
standpoint, these factors would be acting more like dominants than reces- 
sives. This means that each locus would have a detrimental effect of 2u in- 
stead of u (since a dominant gene would be responsible for twice as many 
"genetic deaths" as a recessive), but the locus would be unimportant for 
heterosis. Since the n in the formula refers only to the number of loci which 
are capable of mutating to a completely recessive allele, its value may be 
smaller than previously assumed and the product nu proportionately less. 

It has been assumed that the parent populations are at equilibrium be- 



290 JAMES F. CROW 

tween selection and mutation pressures. This assumption probably is not 
strictly correct for any population. Any equilibrium involving occurrences 
as rare as mutations must be slow of attainment. Hence many if not most 
populations must not be at equilibrium. Probably the most common way 
in which a population gets out of equilibrium is by an alteration of the breed- 
ing structure or population number so that the effective amount of inbreeding 
is changed. If the change in population structure is such as to increase the 
amount of homozygosity, a new equilibrium is reached comparatively rapidly 
through the elimination by selection of the recessives which have been made 
homozygous. On the other hand, if the change in population is such as to 
decrease the amount of homozygosity a new equilibrium is attained only 
through the accumulation of new mutations. This is an extremely slow 
process. 

Since the return to equilibrium is much slower when the population 
changes in the direction of less inbreeding, it follows that most populations 
which are out of equilibrium will be out in the direction of having too few 
detrimental recessives. Therefore the effect of fluctuations in population 
size and breeding structure will be on the average such as to increase the 
fitness of the population. For this reason, the average loss of fitness per locus 
is probably less than the mutation rate. Fisher (1949) has pointed out that 
if the yield of a crop is near a "ceiling," the relative effect of each factor con- 
ditioning yield becomes less. There will be a similar tendency for the popula- 
tion to be out of equilibrium because of the slowness of occurrence of the 
mutations required to bring the population to the new equilibrium level. 

Another factor also pointed out by Fisher is that complete lethals and 
highly deleterious factors contribute to the mutation rate but, at least in 
grain crops, have no appreciable effect on yield since they are crowded out 
by other plants. 

All of these factors make the 5 per cent figure an overestimate, so it should 
be regarded as a maximum. The true value may be much less. In this con- 
nection Fisher (1949) said: 

... it would appear that the total elimination of deleterious recessives would make less 
difference to the yield of cross-bred commercial crops than the total mutation rate would 
suggest. Perhaps no more than a 1 per cent improvement could be looked for from this 
cause. Differences of the order of 20 per cent remain to be explained. 

These considerations make it difficult to explain, in terms of the domi- 
nance hypothesis, cases in which two equilibrium populations produce hy- 
brids with considerable heterosis, or in which crosses between inbred lines 
average appreciably more than the randomly mating populations from 
which they were derived. 

This discussion is relevant only when the character is measurable in terms 
of selective value. For yield characters subject to any high degree of artificial 
selection an increase in yield is probably accompanied by a greater proper- 



DOMINANCE AND OVERDOMINANCE 291 

tional increase in selective value. Thus any conclusions about maximum i)ro- 
portional increase in selective value would hold a forliori for yield. Fisher 
(1949) reaches a similar conclusion when he says: "If the chance of survival is 
equated to the yield, as is reasonable with grain crops." 

Another assumption is that the hybrids are compared with equilibrium 
populations. There is room for question, particularly with domestic plants 
and animals, as to whether selection has been occurring long enough and its 
direction has been consistent enough for a gene frequency equilibrium to have 
been attained. Another point that must be remembered in discussions of 
maize is that commercial hybrids are not random combinations of inbred 
lines, but highly select combinations. An average hybrid may have a yield 
very close to that of a randomly mating population. Thus the argument of 
this section may not be relevant for corn. But it can hardly be true that the 
high yield of certain corn hybrids is due to the elimination of deleterious 
recessives during inbreeding. 

The quantitative limit placed on average improvement on hybridization 
with the dominance hypothesis does not hold for overdominant loci. A locus 
at which the homozygote A A has a selective disadvantage of .y with respect 
to the heterozygote, and the homozygote A' A' has a disadvantage of /, will 
come to equilibrium with gene frequency of A equal to t/(s + 0, and the 
frequency of A' equal to s/(s + /) (Wright, 1931b; Crow, 1948). The average 
reduction in selective advantage of the population due to the two homozy- 
gous genotypes comes out to be st/{s + /). The loss in fitness of the popula- 
tion is of the order of magnitude of the selection coefficients, as Haldane 
(1937) has first shown, whereas with a detrimental recessive, the loss is of 
the order of the mutation rate. Hence a single overdominant locus has a 
tremendously greater effect on the population fitness than a single locus with 
dominance or intermediate heterozygote. If such loci are at all frequent they 
must be important. The question is: how frequent are they? 

Even with overdominance it is difficult to understand large average in- 
creases in selective advantage of hybrids between equilibrium populations. 
Such populations should be somewhere near their optimum gene frequencies, 
which means that the hybrids would be about the same as the parents. It 
may be that, on the average, hybrids do not greatly exceed their parents in 
selective advantage, and that the cases of increased size observed in variety 
crosses and occasionally in species crosses are nothing but luxuriance. If so, 
they are much less difficult to explain. 

As Bruce showed in 1910, if the parents differ at all in gene frequencies, 
the hybrids will be more heterozygous. If both parents are at equilibrium 
they should have, for additive genes, approximately the same frequencies. 
But what differences there are — due to chance, for example — will amount to 
much more in an overdominant than in a dominant locus because the former 
has a gene frequency much nearer .5. 



292 JAMES F. CROW 

POPULATION VARIANCE 

The same considerations which show that an overdominant locus has a 
much greater effect on average population fitness than a dominant locus also 
show that an overdominant locus has a much greater effect on the population 
variance. If the selective values of the three genotypes, AA,Aa, and aa are 1, 
1, and 1 — s respectively, the frequency of aa genotypes is u/s and the aver- 
age selective value 1 — u. The variance in fitness will be sii. On the other 
hand, with an overdominant locus where the fitnesses of the three genotypes 
are 1 — ^, 1, and 1 — ^, the mean fitness is 1 — s/2. The variance in fit- 
ness is 5-/4. 

The ratio of these variances is s/4u, which means that an overdominant 
locus causes a population variance s/4:U times as great as that resulting from 
a recessive locus of the same selective disadvantage. If 4u is 10"^, this 
amounts to 100 for ^ = .001, or is 1000 for s = .01. This makes an over- 
dominant locus with these selective values equivalent to 100 or 1000 ordinary 
loci in its effect on the population variance. Haldane (1950) has emphasized 
the importance of loci with adaptively superior heterozygotes in increasing 
the variance of natural populations. 

From this we must conclude that there doesn't have to be a very high 
proportion of overdominant loci for overdominance to be the most important 
factor in the genetic variance of the population. If much of the genetic vari- 
ance of a population is due to overdominance, this would explain the great 
slowness of selection. Characters with high genetic determination but low 
parent-offspring correlation might be due to this cause. 

The facts of hybrid corn also are consistent with this. Ordinary selection 
has not been effective. Yet there is a great deal of variation in an open- 
pollinated variety. It has been relatively easy to find combinations of inbred 
lines that have yields well above the open-pollinated averages. There appears 
to be a relatively high degree of genetic determination of yield, but relatively 
low heritability. These results are not impossible with dominant genes, es- 
pecially with epistasis, but are precisely what would be expected if some of 
the variance were due to overdominant loci. 

A population with many overdominant loci is always well below its maxi- 
mum possible fitness. It is expected that such factors could eventually be 
replaced in long evolutionary periods. This might occur by an appropriate 
mutation, by duplication, or by modifiers. Or a population with too many 
overdominant loci might disappear due to inter-population competition. But 
at any particular time, a population may have a small proportion of such loci, 
and it does not require many for these to be the major source of variation. 

DO THE SAME GENES DETERMINE VARIATION IN 
INBREDS AND HYBRIDS? 

The rarer a recessive phenotype is in a population, the greater will be its 
relative increase in frequency on inbreeding. If the frequency of the recessive 



DOMINANCE AND OVERDOMINANCE 293 

gene is q, the frequency of recessive homozygotes in a randomly mating 
population is (f. With increasing amounts of inbreeding, the frequency 
changes from q- to q. The smaller the value of q, the greater is the ratio of 
q to q~. If a gene is highly deleterious it will be very rare in the population. 
Hence the genotypes which are most deleterious are those which have the 
greatest relative increase in frequency on inbreeding. 

These relationships are brought out in the following figures, based on a 
mutation rate of 10~^. The ratio given is the ratio of homozygous recessives 
in a homozygous population as compared with one which is mating at 
random. 

Selective disadvantage (5) 0001 .001 .01 .1 lethal 

Gene frequency (?) 1 -032 .01 .003 .001 

Ratio (g/5=) 10 32 100 316 1000 

This means that highly deleterious recessives, which ordinarily have an 
effect on the population only of the order of the mutation rate, become much 
more important wdth inbreeding and may become the major factors in deter- 
mining the fitness of an inbred population. This might to some extent be 
offset by selection during the inbreeding process, but such selection would be 
directed against factors which are of no consequence in a more heterozygous 
population. 

The detrimental recessive factors referred to here include the lethals and 
semilethals (such as chlorophyll deficiencies) that show up during inbreed- 
ing. But more important are the larger number of factors, not individually 
detectable, which collectively result in the loss of vigor with inbreeding de- 
spite rigorous selection. 

On the other hand, the major part of the variance of a non-inbred popu- 
lation may well be determined by genes of intermediate frequencies, from 
.1 to .9. The effect of such factors in determining the population variance 
in fitness would change only slightly with inbreeding. 

As an example, consider a hypothetical population mating at random 
whose variance is made up of two components. Ninety per cent of the vari- 
ance is due to relatively common loci with gene frequencies of the order of .5. 
The other 10 per cent is due to loci with recessive gene frequencies of the 
order of .01 or less. Now when this population is inbred without selection, 
the variance due to the common genes will not change greatly but the vari- 
ance due to the recessive loci will increase by a hundred fold or more. Thus 
the factors which originally contributed only 10 per cent to the variance 
may now contribute over 90 per cent of the variance between the various 
inbred lines derived from the population. 

Gene frequencies of the order of .5 might result from several causes. They 
might be genes which are advantageous in one geograj)hical location and 
disadvantageous in another so as to form a cline. Or there might be seasonal 
differences in selective value. They may be due to complex interactions with 
other loci or be of extremely small selective advantage or disadvantage. But 



294 JAMES F. CROW 

another explanation is selective superiority of heterozygotes (Haldane, 1950), 
at least for those factors of importance in heterosis. 

If yield is determined entirely by dominant factors, the correlation be- 
tween inbreds and their hybrids should be positive. If it is due to over- 
dominant loci, the correlation should be generally positive, though there 
would be negative correlations between yield of hybrids and inbreds when 
the other inbred is constant and high yielding. If both factors are involved 
and overdominant loci are relatively important in hybrids while dominants 
are important in inbreds, the correlation would approach zero. The experi- 
ence of corn breeders has been that selection for yield during inbreeding is 
relatively ineffective, and that the correlation of hybrid with inbred yield, 
though positive, is small. 

With overdominant loci the effect of a certain percentage increase in 
heterozygosity is to cause the vigor to increase by a certain amount. De- 
creasing the heterozygosity by the same percentage would cause a decrease 
of approximately the same amount. On the other hand, with dominant loci, 
making the original equilibrium population more heterozygous would cause 
a very slight increase, whereas making the population more homozygous 
would have a decreasing effect of a much greater amount. Therefore it is 
easier to account for inbreeding depression by dominant loci than to account 
for increase in vigor on hybridization above the level of a random mating 
population. 

I should like to suggest the following interpretation of the effects of in- 
breeding and hybridization: The deleterious effects of inbreeding and the re- 
covery on hybridization are mainly due to loci where the dominant is favor- 
able and the recessive allele so rare as to be of negligible importance in a non- 
inbred population. Variance of a non-inbred population, and hybrid vigor 
when measured as an increase over an equilibrium population, are deter- 
mined largely by genes of intermediate frequency, probably mostly over- 
dominants. 

OVERDOMINANCE AND GENE ACTION 

In order to have overdominance it is not necessary that the immediate 
gene products of the heterozygote exceed in quantity or variety those of 
either homozygote. At the level of the immediate gene product, or any inter- 
mediate state, the effect of the heterozygote may be intermediate between 
the two homozygotes and still result in a greater final result. Any kind of 
situation in which something is produced for which an intermediate amount 
is optimum could be such that the heterozygote is nearer this optimum than 
either homozygote. 

A model for such cases is found in the sulfanilamide-requiring strain of 
Neurospora reported by Emerson (1948). When this mutant is present the 
heterokaryotic state of the suppressor gene results in more nearly the opti- 



DOMINANCE AND OVERDOMINANCE 295 

mum amount of para-amino benzoic acid than either homokaryon. Other 
cases, less known biochemically, may be similar. 

I think that it is doubtful whether such a system would persist for long 
evolutionary periods. Alleles of intermediate productivity could arise and 
replace the originals. Also modiiiers altering the expression of the homozy- 
gotes would have considerable selection pressure. Or if the alleles were anti- 
morphic, the situation might be resolved by duplication, as Haldane (1937) 
has suggested. It is significant that the system reported by Emerson is not 
one which is ordinarily of importance, but acts only in the presence of the 
sulfanilamide-requiring mutant. 

A form of gene action that appears more likely to account for instances 
of overdominance is one in which the two alleles differ qualitatively or each 
does something that the other fails to do. Instances of mosaic dominance 
provide excellent examples. This has been demonstrated for the scute series 
of bristle characters in Drosophila and for color pattern in beetles (Tan, 
1946). Other examples are provided by the ^4 and R loci in maize. 

Similar examples of physiological mosaic dominance are found where the 
heterozygote apparently produces something approximating — at least 
qualitatively — the total effect of the two homozygotes. An example is rust 
resistance in flax, where each strain is resistant to a certain rust but the hy- 
brid is resistant to both (Flor, 1947). By the usual tests for allelism, the two 
resistance factors are alleles. Another series of examples is found in the 
blood group antigens in man, cattle, and elsewhere. In almost every instance 
the heterozygote has all the antigenic properties of both homozygotes 
(Irwin, 1947). The presence of both the normal and abnormal types of 
hemoglobin in humans heterozygous for the gene for sicklemia provides 
another example (Pauling, 1950). 

Many instances of overdominance may have a similar explanation. This 
is the kind of action that East (1936) postulated in his discussion of heterosis 
due to cumulative action of divergent alleles. It is not necessary that the 
effects be completely cumulative; only that the net effect on the phenotype 
be greater in the heterozygote than in the homozygote. Any system in which 
the alleles act on different substrates to produce the same or different prod- 
ucts, or convert the same substrate into different products — neomorphs, in 
Muller's terminology — could result in overdominance. 

Any of the examples listed above may turn out to be closely linked genes 
(pseudoalleles) rather than alleles. In most cases it is impossible to distin- 
guish between these alternatives. If the overdominance effect is due to 
linked genes, eventually a crossover should result in a situation where the 
desirable effects could be obtained in a homozygous individual. If there are 
position effects, it may be that no homozygous arrangement is as advanta- 
geous as one which is heterozygous. Unless there are position effects, it does 
not seem likely that heterosis due to pseudoallelism would persist for any 



296 JAMES F. CROW 

great length of time, but in any particular population such factors might be 
important. 

IS INCREASED SIZE ADAPTIVE? 

The foregoing arguments are based on the assumption that heterosis is 
measurable in terms of increased selective advantage. The selection may be 
natural or man-imposed. This assumption would appear to be reasonable for 
such factors as fertility and resistance to disease. It also would apply to in- 
crease in size or yield, if the direction of selection in the past were in this di- 
rection, as in corn. However, it is questionable whether the increase in size 
that is sometimes observed in variety hybrids is really adaptive. 

Mather (1943) and especially Dobzhansky (1950) have emphasized that 
increased size does not necessarily result in increased fitness in natural popu- 
lations. Dobzhansky proposed the words euheterosis and luxuriance, re- 
spectively, for increased selective advantage and for mere non-adaptive in- 
crease in size. In these terms this discussion has dealt entirely with eu- 
heterosis. 

If euheterosis occurs in species or variety crosses, it is very difficult to 
explain. It raises the troublesome question: How can the hybrid between 
two well adapted strains be better adapted than its parents when there has 
been no selection in the past for its adaptation? It may be that euheterosis 
is developed only under some form of selection, as in the inversion heterozy- 
gotes studied by Dobzhansky, or in the series of hybrids between inbred lines 
of corn selected for combining ability. 

If large size is not advantageous, luxuriance may be due to the covering 
of recessive factors which were acting as size bottlenecks and had been 
selected into the population because of this. Each of the parents might have 
its growth limited by or held in check by a series of factors, and if some of 
these were recessive, increased size would be found in the hybrids. 

SUMMARY 

Since the earliest attempts to explain hybrid vigor in Mendelian terms 
there have been two principal hypotheses. The first of these is the domi- 
nance hypothesis. This notes the observed correlation between recessiveness 
and detrimental effect and attributes the increased vigor of heterozygosity 
to the covering of deleterious recessive factors by their dominant alleles. 
The alternative hypothesis, the overdominance hypothesis, assumes that 
heterozygosity per se is important — that there exist loci at which the hetero- 
zygote is superior to either homozygote. 

It is clear that the dominance hypothesis is adequate to explain the de- 
terioration that results from inbreeding and the recovery of vigor on out- 
crossing, but it is difficult to explain how the hybrids could greatly exceed in 
fitness the equilibrium populations from which their parents were derived. 
The overdominance hypothesis demands the assumption of a kind of gene 



DOMINANCE AND OVERDOMINANCE 297 

action known to be rare, but it is pointed out that if only a small proportion 
of the loci are of this type, these may nevertheless be the major factor in the 
population variance. 

The following interpretation is suggested: Inbreeding depression and re- 
covery on crossing are mainly the result of loci at which the favorable allele is 
dominant and the recessives are at low frequency. On the other hand the 
variance of heterozygous populations and the differences between different 
hybrids are due mainly to loci with intermediate gene frequencies. It appears 
likely that such loci are due to selectively superior heterozygotes, but there 
are several other possibilities. 



LEROY POWERS 

USDA, Bureau of Plant Industry 



Chapter 19 

Gene Recombinotion 
and Heterosis 



This article will be confined primarily to the tomato {Lycopersicon) genetic 
work which has a bearing on gene recombination and heterosis. The barley 
(Hordeum) genetic research which is discussed briefly was conducted at the 
University of Minnesota. The tomato genetic research which constitutes the 
bulk of the material discussed was conducted at the United States Horticul- 
tural Field Station, Cheyenne, Wyoming. 

With the present available methods of analysis it is difficult in quantita- 
tive inheritance studies to distinguish between blocks of fairly closely linked 
genes and individual pairs of genes. This has been shown by the work of 
Jones (1917), Warren (1924), Mather (1942, 1949), and Straus and Gowen 
(1943). Consequently, in this article where the two genetic systems are not 
distinguishable the term pairs of genes will be employed. Mather (1949) has 
used the term ejffedive factor to depict such a genetic situation. 

MARKER GENES AND LINKAGE IN BARLEY 

Powers (1936) has shown that in a cross between Bl {Hordeum deficieiis) 
and Brachytic {Hordeum vulgare) the Fi, which is a two-row barley, gave a 
greater yield of seed per plant than either the two-row or six-row parents. 
Then, weight of seed per plant shows heterosis. The data on marker genes 
and linkage in barley presented have some bearing upon whether any of the 
advantages of the Fi hybrid attributable to heterosis can be recovered in 
inbred lines through gene recombinations. 

The deficiens (two-row) character was found to be differentiated from the 
vulgare (six-row) character by one pair of genes designated as Vv, and the 
brachytic character from the normal character by one pair of genes designated 

298 



GENE RECOMBINATION AND HETEROSIS 



299 



as Brbr. Using these symbols, the genotype of the Fi is VvBrbr. The Vv gene 
pair is carried on chromosome 1 and the Brbr gene pair on chromosome 7. 

Table 19.1 gives the comparative effect upon four quantitative characters 
of genes associated in inheritance with Vv and vv and VV and vv, as deter- 

TABLE 19.1 

COMPARATIVE EFFECT UPON FOUR QUANTITATIVE CHARACTERS OF 

GENES ASSOCIATED IN INHERITANCE WITH Vv AND vv, AND 

VV AND vv; F-, GENOT\TES OF A BARLEY HYBRID 





Weight 


3F Seed* 


Spikes per Pl.ant* 


Height of Plant* 


Length 


3F Awn* 


Genotvpe 


Vv-vv 


vv-vv 


Vv-vv 


VV-VV 


Vv-vv 


vv-vv 


Vv-vv 


VV-VV 


BrBr 

Brbr 

brbr 


-2.22 
-2.98 
-1.88 


-3.44 
-3.74 
-2.74 


1.72 
0.94 
0.13 


0.21 

0.39 

-0.94 


1.54 
2.08 
1.03 


0.64 

1.41 

-0.68 


16.58 

16.42 

1.95 


7.. SO 

9.20 

-6.68 



* Weight of seed per plant is expressed in grams, spikes per plant in number, height of plant in inches, and 
length of awn in millimeters. 



TABLE 19.2 

COMPAR.VTIVE EFFECT UPON FOUR QUANTITATIVE CHARACTERS OF 

GENES ASSOCIATED IN INHERITANCE WITH Vv AND VV, AND 

VV AND vv; F. GENOT\TES OF A BARLEY HYBRID 





Weight ( 


3F Seed* 


Spikes per Plant* 


Height of Plant* 


Length 


OF Awn* 


Genotype 


Vv-vv 


VV-VV 


Vv-vv 


vv-vv 


Vv-vv 


vv-vv 


Vv-vv 


VV-VV 


BrBr 

Brbr 

brbr 


1.22 
0.76 
0.86 


-3.44 
-3.74 
-2.74 


1.51 
0.55 
1.07 


0.21 

0.39 

-0.94 


0.90 
0.67 
1.71 


0.64 

1.41 

-0.68 


9.08 
7.22 
8.63 


7.50 

9.20 

-6.68 



* Weight of seed per plant is expressed in grams, spikes per plant in number, height of plant in inches, and 
length of awn in millimeters. 



mined by differences between means of F2 plants. In every case, the differ- 
ences between Vv and vv are greater than the differences between VV and vv 
for spikes per plant, height of plant, and length of awn. With the exception 
of the comparison between VV and vv within the brbr genotype, the differ- 
ences are in favor of the two-row (Vv and VV) segregates as compared with 
the six-row (vv) segregates. Within the brbr genotype, iJt) plants exceed the VV 
plants for all three characters. As regards weight of seed per plant in every 
case the six-row plants outyielded the two-row plants whether heterozygous 
deficiens or homozygous deficiens. However, the differences between vv and 
Vv were less than those between vv and VV. 

The data of Table 19.2 show that for all characters the Vv plants give an 



300 LEROY POWERS 

increase over the VV plants, and with the exception of the Brbr genotype for 
height of plant and length of awn, the differences of Vv-VV are greater than 
the differences for VV-vv. 

These facts concerning the data reveal that Vv is associated with an in- 
crease in all four quantitative characters. For spikes per plant, height of 
plant, and length of awn this increase results in heterosis. 

Hypotheses for Difference in Vigor 

If the increase noted is due solely to an interaction between V and v such 
as is depicted by East's physiological hypothesis, then it would not be pos- 
sible to obtain homozygous lines possessing any of this increase. However, if 
the heterosis noted is due to a combination of favorable and unfavorable 
genes linked with V and v, it should be possible to obtain lines in which 
some of the favorable genes are recombined. These lines should show some 
increase in the four quantitative characters studied. In the event that linkage 
of genes favorable and unfavorable to an increase in the quantitative charac- 
ters was found to furnish the most logical explanation, an intraallelic interac- 
tion such as depicted by East's physiological hypothesis still may be having 
some influence as the two systems are not mutually exclusive. 

Tables 19.1 and 19.2 show that Vv results in an increase of all four charac- 
ters: weight of seed per plant, number of spikes per plant, height of plant, 
and length of awn. This fact is most simply explained by assuming the pro- 
duction of a favorable growth-promoting substance which influences all of 
them. Then such being the case, on the basis of East's (1936) physiological 
hypothesis, V and v supplement each other, resulting in greater development. 
Next consider the development of the lateral florets which determines the 
number of rows of kernels per spike (two-row or six-row spikes). The Vv 
segregates are two-row types, whereas the vv segregates are six-row types. 
Hence, as regards the character number of rows of kernels per spike, the in- 
teraction between V and v is such as to prohibit the development of the 
lateral florets, resulting in a two-row barley spike rather than a six. Summing 
up, on the basis of the physiological hypothesis, in the case of four quantita- 
tive characters the interaction between V and v is such as to stimulate de- 
velopment. In the case of number of rows of kernels per spike the interac- 
tion is such as to prohibit development of the lateral florets. From physio- 
logical genetic considerations such a pleiotropic effect seems rather im- 
probable. 

Explaining the heterosis associated with Vv plants on the basis of linkage, 
a simple interpretation would be that the favorable linked genes and their 
alleles interact according to Jones's (1917) hypothesis to produce a substance 
favorable to growth processes, resulting in the heterosis noted; and that V 
is dominant to v resulting in Vv (Fi) plants having two-row barley spikes. 
This explanation does not require the assumption that V and v stimulate 



GENE RECOMBINATION AND HETEROSIS 



301 



growth in one character and inhibit it in another, and hence is more in ac- 
cord with modern physiological genetic concepts. 

The article by Powers (1936) furnishes additional information j)ertaining 
to gene recombination and heterosis. If genes other than Vv are responsible 
for the heterosis noted, then F2 plants having a genotype identical to the Fi 
generation should give a somewhat lower yield than the Fi. Since the Fi 
plants were not grown in a randomized experiment with the F2 plants, the 
comparison must be made through the Bl parent. As compared through the 
Bl parent an actual reduction of one gram in yield of seed per plant was 
found (Powers, 1936). This reduction could be due to genes carried on chro- 
mosome 1, as are V and v, or to genes carried on other chromosomes. In 
either event, theoretically some of the genes favorable to increased weight 

TABLE 19.3 

COMPARISON BETWEEN PARENTS AND F2 PAREN- 
TAL GENOT\TES FOR WEIGHT OF SEED 
PER PLANT IN A BARLEY HYBRID 



Total 


Weight of Seed per Plant in Grams 


Number of 
Plants 


Fs 


Parent 


Differ- 
ence 


/ 


78 and 266* 

64 and 63 f . 


3.9 
4.5 


1.9 
4.0 


2.0 
0.5 
1.5 


28.189 
761 


Interaction 


5 807 











* VVBrBr, genotype of Bl parent, two-row normal. 

t vvbrbr, genotype of Brachytic parent, six-row Brachytic. 

of seed per plant that resulted in the heterosis noted in the Fi population 
must be capable of recombination. 

Even though some of the genes favorable to increased growth can be re- 
combined, the yield of the lines in which the favorable genes have been com- 
bined depends upon the nature of the interaction of the genes. The weights 
of seed per plant of parents and F2 plants of the parental genotypes are given 
in Table 19.3. From this table it can be seen that the F2 plants of the VVBrBr 
genotype gave an increased yield of 2.0 grams per plant over the Bl parental 
plants having the same genotype. However, the F2 plants of the same geno- 
type as the Brachytic parent gave an increase over this parent of only 0.5 
grams per plant, which is not statistically significant. The interaction of 1.5 
grams (Table 19.3) is statistically significant. This means that a preponder- 
ance of the genes favorable to increased weight of seed per plant must have 
entered the cross from the Brachytic parent. The balance of the unfavorable 
genes that entered the cross from the Bl parent did not cause a correspond- 
ing decrease in weight of seed per plant of the F2 plants possessing the vvbrbr 
genotype. 



302 



LEROY POWERS 



In this same study (Powers, 1936) found that the greater the number of 
genes in the genotype tending to increase a character the greater is the effect 
of any given gene. It is apparent that it is not possible to definitely predict 
the yield of seed per plant resulting from recombining genes favorable to 
growth because of the interactions noted. Either a greater or smaller increase 
than expected may be obtained. Such interactions of genes would affect the 
yield of plants in which the favorable genes were recombined, and hence the 
feasibility of obtaining inbred lines equaling or excelling the Fi hybrid. In 
some cases the probability of getting the desired results would be increased 
and in other cases decreased; depending on the type of interallelic and intra- 
allelic interactions of the genes. 

GENE RECOMBINATIONS DIFFERENTIATING WEIGHT PER LOCULE 
WHICH EXCEED HETEROSIS OF F, POPULATION 

The data for weight per locule of fruit for the Porter X Ponderosa to- 
mato hybrid and parental populations grown at Woodward, Oklahoma, in 

TABLE 19.4 

.ARITHMETIC AND LOGARITHMIC MEANS FOR 
WEIGHT PER LOCUI.E OF PORTER X PONDERO- 
SA TOMATO HYBRID AND PARENTAL POPULA- 
TIONS* 





Mea\ 


P0PUL.\TI0N 


Arithmetic 


Logarithmic 


Porter 


10.2 
11.8 
14.4 
13.5 
13.7 
9.8 


1.018253 + 0.012325 


Bi to Porter .... 


1.070936 + 0.009939 


F, 

Fs 

Bi to Ponderosa 

Ponderosa 


1.168729 + 0.010134 
1.128481+0.011879 
1.124941+0.012651 
0.982054 + 0.011845 



* Grown at Woodward, Oklahoma, in 1041; original data taken in grams 
and transformed to logarithms to obtain the means and standard errors of the 
logarithms. 

1941 (Powers, Locke, and Garrett, 1950) will be analyzed to determine 
whether in Fo and backcross populations gene recombinations are occurring 
which exceed the heterosis of the Fi population. 

The means for weight per locule calculated on both the arithmetic and 
logarithmic scales are given in Table 19.4. Weight per locule is greatest for 
the Fi population, and the means of the Bi to Porter, F2, and Bi to Ponderosa 
populations are larger than the means of the Ponderosa and Porter parents, 
but smaller than the mean of the Fi population. The only means not showing 
significant differences are the means of Porter and Ponderosa, and the means 
of the F2 and Bi to Ponderosa populations. Hence, in these hybrid popula- 
tions weight per locule definitely shows heterosis on either scale. 



GENE RECOMBINATION AND HETEROSIS 



303 



The frequency distributions for weight per locule for the Porter X Pon- 
(lerosa hybrid and parental populations are given in Table 19.5. This table 
shows that the F2 and Bi to Ponderosa populations have plants falling into 
classes of greater value than 1.51 1883, the last class in which Fi or Ponderosa 
plants occur. There are nine such F2 i)lants and three such Ei to Ponderosa 
plants. If no recombination of genes to produce plants with weight per locule 
greater than the Fi plants is possible, these plants with values greater than 
any individual of the Fi population must be chance deviates. Moreover, the 
chance deviates must be those plants in the F2 population having the I'^ 

TABLE 19.5 

OBTAINED FREQUENCY DISTRIBUTIONS FOR WEIGHT PER LOCULE 
OF TOMATO FRUITS FOR PORTER X PONDEROSA HY- 
BRID AND PARENTAL POPULATIONS* 





Upper Limit of Class in Logarithms or Grams 




Popula- 
tion 



d 


d 


00 

d 


On 
rs 

d 


00 

q 




o\ 

so 
On 

q 


00 



00 
•* 






00 


-0 

00 





10 

00 

m 





00 


00 
00 


OS 

00 
IT) 






00 
10 


10 





00 
00 

o> 


Total 
Plants 


Porter... . 


1 




4 


27 


80 


98 


20 


2 




























232 


Bi to Por- 














































ter 


1 


3 


13 


54 


81 


102 


80 


49 


35 


16 


11 


1 




1 




1 












448 


Fi 






1 
13 


6 
31 


22 
68 


35 
8? 


49 

81 


37 
63 


34 
47 


23 
?4 


13 
17 


4 

11 


2 

8 


3 
3 


4 

1 




4 


? 




? 


1 


233 


F2 






453 


Bi to Pon- 














































derosa. . 


1 


6 


17 


29 


45 


71 


72 


62 


52 


26 


19 


18 


1 


4 


4 


4 


1 


1 


1 






434 


Pondero- 














































sa 




10 


21 


28 


25 


16 


18 


9 


4 


6 


2 


3 


1 




1 


1 












145 









* Grown at Woodward, Oklahoma, in 1941; original data taken in grams and transformed to logarithms to 
obtain the means and standard errors of the logarithms. 



genotype or a very similar genotype. The probability of their being chance 
deviates possessing the Fi or similar genotypes can be determined. 

The mean of the logarithms of the Fi population is 1.168729, and the 
standard error of a single determination is 0.123426. Calculations (for meth- 
od see Powers, Locke, and Garrett, 1950) show that only 0.3 per cent of such 
a genotypic population would be expected to have a value greater than 
1.511883. The following tabulation shows the theoretical number of gene 
pairs differentiating the parents, the theoretical percentage of the j)opula- 
tion of the F2 or Bi to Ponderosa populations possessing the same genotyi)e 
as the Fi, the theoretical number of plants of the Fi genotype in a popula- 
tion of 453 F2 plants and in a population of 434 Bi to Ponderosa plants, and 
the theoretical number of plants of the Fi genotype in the F2 population 
and in the Bi to Ponderosa population expected to exceed a value of 1.511883. 

An examination of the data opposite one pair of genes in the tabulation be- 
low shows that only 0.68 Fo plants would be expected to exceed a value of 



304 



LEROY POWERS 



1.511883, whereas 9 plants did so (see Table 19.5). The same comparison 
for the Bi to Ponderosa population is 0.65 expected and 3 obtained. Also, a 
study of the tabulation below reveals that with an increased number of 
gene pairs diflferentiating the parents the odds become even greater against 
those plants which exceed 1.511883 being chance deviates. 

It remains to be seen whether plants of the Fi genotype plus plants of 
genotypes which might have similar effects, but do not possess recombina- 
tion of favorable genes in excess of the total number of favorable genes car- 



Number 

Pairs of 

Genes 


Per Cent 
Population 
OF Fi Geno- 
type 


Number of Plants 
OF Fi Genotype 


Number of Plants 
Expected To Ex- 
ceed A Value of 
1.511883 




F2 


Bi 


F2 


Bi 


1 


50.00 

25.00 

12.50 

6.25 

3.12 


226.50 

113.25 

56.62 

28.31 

14.17 


217.00 

108.50 

54.25 

27.12 

13.56 


0.68 
0.34 
0.17 
0.08 
0.04 


0.65 


2 

3 


0.33 
0.16 


4 


0.08 


5 


0.04 



ried by the Fi, could be responsible for the results noted. The result would be 
to increase the proportion of the F2 and Bi to Ponderosa populations fluctu- 
ating around means very similar in magnitude to that of the Fi population. 
The extreme case (but highly improbable) would be to have all of these two 
populations made up of such plants. On this basis and on the basis that the 
parents are differentiated by one pair of genes, the number of plants of the 
F2 population expected to exceed 1.511883 is 1.36, and for the Bi to Pondero- 
sa population is 1.30. The number of plants obtained (Table 19.5) is 9 for the 
F2 population and 3 for the Bi to Ponderosa population. Furthermore, the 
Bi to Porter population had 1 plant in a class beyond that in which any Fi 
plants occurred. 

The analysis can be carried further. For the F2 population the number of 
plants expected to exceed 1.562293 is 0.3223 and the number obtained is 3. 
Whereas the values for the Bi to Ponderosa population are 0.3087 and 1, 
respectively. Also, the frequency distributions (Table 19.5) in general do 
not support the supposition that over one half of the plants of the F2 and Bi 
to Ponderosa populations are fluctuating around a mean as great as that of 
the Fi generation. Again with an increase in number of gene pairs differentiat- 
ing the parents, the odds against the plants exceeding 1.562293 being chance 
deviates become even greater. It is evident that the data are not in accord 
with the assumption that plants of the Fi genotype have the greatest weight 
per locule. This is true regardless of the number of gene pairs differentiating 
the parents. Therefore, some of the plants falling in classes having values 



GENE RECOMBINATION AND HETEROSIS 305 

greater than 1.511883 must have genotypes composed of more favorable 
genes than the Fi, and therefore recombinations of genes to produce plants 
having a greater weight per locule than the Fi plants have occurred. 

Whether inbred lines retaining this increased weight per locule can be 
established is dependent upon the number of gene pairs differentiating the 
parents and linkage relations (Jones, 1917). Close linkage of genes favorable 
to increase in weight per locule would favor recombination. Whereas close 
linkage of genes favorable to increase in weight per locule with those not 
favorable would hinder recombination and hence reduce the chances of ob- 
taining inbred lines retaining some or all of the advantages attributable to 
heterosis. 

The data furnish evidence concerning the number of gene pairs differen- 
tiating weight per locule. From Table 19.5 it can be seen that the plants of 
the F2 generation falling beyond the value 1.511883 are distributed over four 
different classes, and those of the Bi to Ponderosa population falling beyond 
this same value occur in three different classes. The behavior of these plants 
cannot be explained on the basis of five or more independently inherited 
pairs of genes, as there are too many of these plants falling beyond the 
1.511883 class. In addition, the weights per locule of those falling in these 
classes are greater than can be explained on the basis of chance deviation. 

Further, to account for the plants of the F2 and Bi to Ponderosa popula- 
tions falling in those classes beyond 1.511883, on the basis of five or more 
pairs of independently inherited genes differentiating the parents, it would 
be necessary to assume that 50 per cent or more of the plants were fluctuating 
around a mean greater than that of the Fi generation. Since the means 
(Table 19.4) of the F2 and Bi to Ponderosa populations are less than the 
mean of the Fi, these populations cannot have a greater majority of the 
plants fluctuating around a mean larger in magnitude than that of the Fi 
plants. This deduction is confirmed by the frequency distributions of Table 
19.5, as both of these populations have a greater percentage of their plants 
in lower classes of the frequency distributions than does the Fi population. 
Powers, Locke, and Garrett (1950) have shown that the data give a good fit 
to frequency distributions calculated on the assumption that the parents are 
differentiated by three pairs of genes. 

Here, proof of recombination of genes to produce plants in the F2 and Bi 
to Ponderosa populations with greater weight per locule than Fi plants is 
fairly conclusive. Also, since the number of gene pairs or closely linked blocks 
of genes is few, it should be possible by selection to establish inbred lines re- 
taining this advantage. 

MAIN AND COMPONENT CHARACTERS 

The data from the parental and hybrid populations of tomatoes on the 
main and component characters provide information concerning the rela- 
tions between gene recombination, dominance, and heterosis. 



306 



LEROY POWERS 



Weight of Fruit and Its Component Characters 

The data on weight per locule, number of locules, and weight per fruit for 
the Porter X Ponderosa hybrid and parental populations are given in Tables 
19.4 and 19.6. On the arithmetic scale, smaller numbers of locules show par- 
tial dominance. On the logarithmic scale the means of the Fj and F2 popula- 
tions are not significantly different from the average of the means of the 
Porter and Ponderosa populations. The mean of the Bi to Porter population 
is not significantly different from the average of the means of the Porter 
and Fi populations. The mean of the Bi to Ponderosa population is not sig- 

TABLE 19.6 

THE ARITHMETIC AND LOGARITHMIC MEANS FOR NUMBER OF LOCULES 

AND WEIGHT PER FRUIT OF PORTER X PONDEROSA TOMATO 

HYBRID AND PARENTAL POPULATIONS* 





Number of Locules 


Weight per Fruit 


PoPUL.iTION 


Arith- 
metic 


Logarithmic 


Arith- 
metic 


Logarithmic 


Porter 


2.1 
3.1 
4.5 
4.7 
7.1 
10.0 


0.30707210.002151 
0.468411+0.008158 
0.637265 + 0.007663 
0.628793+0.012522 
0.82940410.007738 
0.983292 + 0.017094 


21.5 
36.6 
65.0 
63.5 
97.3 
97.7 


1.326101+0.012358 


Bi to Porter 

F: 

Fo 

Bi to Ponderosa . . 
Ponderosa 


1.539833 + 0.010394 
1.806845 + 0.009416 
1.762614 + 0.013078 
1.95443010.013269 
1.965097 + 0.008750 



* Grown at Woodward, Oklahoma, in 1941; original data taken in numbers and grams and transformed to 
logarithms to obtain the means and standard errors of the logarithms. 



nificantly different from the average of the means of the Fi and Ponderosa 
populations. Hence, on the logarithmic scale there is no dominance, and the 
data indicate that the genetic variability follows the logarithmic scale. In 
other words, the effects of the genes differentiating weight per locule are 
multiplicative. This is true of both the intraallelic and interallelic inter- 
actions. 

Thus on the logarithmic scale number of locules shows no dominance, 
weight per locule shows heterosis (Table 19.4) and the two combine addi- 
tively to give weight per fruit. For weight per fruit the Fi indicates partial 
dominance of greater weight per fruit, the Bi to Ponderosa complete domi- 
nance, and the Bi to Porter no dominance. On the arithmetic scale the two 
component characters unite multiplicatively, and the Fi indicates partial 
dominance of greater weight per fruit, the Bi to Ponderosa complete domi- 
nance, and the Bi to Porter partial dominance of smaller weight per fruit. 
Then it is clear that regardless of scale, one of the component characters 
shows some degree of dominance, the other heterosis. They combine to pro- 
duce the main character which in turn shows some degree of dominance. 



GENE RECOMBINATION AND HETEROSIS 307 

Powers, Locke, and Clarrett (1950) found the number of major gene j)airs 
differentiating number of locules to be 3. Since weight per locale was found 
to be differentiated by 3 pairs of major genes, a comparatively few (prob- 
ably 6) pairs of major genes differentiate weight per fruit. Hence, the number 
of major gene pairs responsible for heterosis of weight per locule is no greater 
than the number of major gene pairs responsible for no dominance of num- 
ber of locules and partial or complete dominance of weight per fruit on the 
logarithmic scale. Then, in this study the number of pairs of major genes 
differentiating the character has no bearing on whether the hybrid popula- 
tions will show no dominance, partial dominance, complete dominance, or 
heterosis. 

From these results it follows that in this material recombination of genes 
to retain the advantages of heterosis is no different than recombination of 
genes to combine desirable characters. Furthermore, these data furnish 
rather convincing evidence that dominance and heterosis are different de- 
grees of expression of the same physiological genetic phenomena, as was 
postulated by Powers (1941, 1944). 

Main and Component Characters of 45 Hybrids Produced 
by Crossing 10 Inbred Lines of Tomatoes 

Table 19.7 summarizes the dominance relations of the main and com- 
ponent characters of 45 hybrids produced by crossing 10 inbred lines of 
tomatoes. 

The percentage values given in Table 19.7 were calculated from data pre- 
sented in a previous article (Powers, 1945). The reader is referred to this 
article for the experimental design, a description of the material, and meth- 
ods. Here, only the method of compiling the data need be given. All of the 
values of this table with the exception of those listed under heterosis were 
calculated from the formula 100[2Fi/(Pi + A)]. The percentages listed 
under the column headings "heterosis" were calculated from the formula 
lOO(Fi/Pi) and 10G{Fi/F2), respectively. Fi is the mean of the Fi popula- 
tion, Fi the mean of the parent with the smaller value, and A the mean of 
the parent with the larger value. The 11 characters listed in Table 19.7 were 
originally expressed in the following units of measurement: Spread of plant 
in inches, yield per plant in grams, number of fruit that ripened per plant, 
height per plant in inches, weight per locule of the fruit in grams, number of 
days from first fruit set to first fruit ripe, number of days from first bloom 
to first fruit set, weight of fruit in grams, number of days from seeding to 
first bloom, number of locules per fruit, and number of days from seeding 
to first fruit ripe. 

The odds against any value belonging in an adjacent classification (col- 
umn) are greater than 19: 1 with the exception of the two values designated 
with an asterisk. Even for these two values the odds against their deviating 



308 



LEROY POWERS 



more than one class are greater than 19: 1. When interpreting the data it is 
necessary to have in mind that parental percentage values would have fallen 
into the complete dominance columns, the Pi value into the first such col- 
umn, and the Pi into the second such column. Also, it should be kept in mind 
that the values listed in Table 19.7 are for the different Fi hybrids, and with 
the exception of the values listed under the columns headed "heterosis" are 
percentages based on the means of the two respective parents. The percent- 
ages listed under the heterosis columns are based on the mean of the parent 
that fell into the adjacent complete dominance columns. 

TABLE 19.7 

PERCENTAGE RANGE IN EXPRESSION OF DOMINANCE FOR 
DIFFERENT CHARACTERS OF Fi TOMATO HYBRIDS* 





Dominance 


Character 


Heter- 
osis 


Com- 
plete 


Partial 


None 


Partial 


Com- 
plete 


Heter- 
osis 


Spread of plant . .... 












114 
166 
172 
112 
109 


122 


Yield, ripe fruit per plant. . . 








106 
99 

100 
98 

99 


117 
142 
104 

103 


171 


Number, ripe fruit per plant. 






78 

96 
70 

93 


155 


Height of plant 






121 


Weight per locule 






119 


Period, first fruit set to first 
fruit ripe 




90 

80 

'95' 
69 

93 




Period, first bloom to first 
fruit set 


75 
95 
89 
73 

95 


125* 




VVpie^ht ner fruit 


53 
99 
79 

96 


102 

100* 

96 






Period, seeding to first bloom . 

Number of locules per fruit. . 

Period, seeding to first fruit 

ripe 





























* As measured by percentages of averages of values of parents and percentages of parental values. 

If dominance and heterosis are different degrees of expression of the same 
physiological genetic phenomena, then the different genotypes, as represented 
by the different Fi hybrids, might be expected to show ranges in expression 
of a given character from different degrees of dominance to heterosis. 

Every character listed in Table 19.7 except number of days from first fruit 
set to first fruit ripe, in the different hybrids, ranges from some degree of 
dominance to heterosis. Yield in grams of ripe fruit per plant, depending 
upon the genotype (Fi hybrid), varied from no dominance through all classes 
to heterosis for increased yield. Number of ripe fruit per plant and height of 
plant varied through all classes from partial dominance of a decrease in mag- 
nitude of these two characters to heterosis for an increase. Weight of fruit 
in grams, number of days from seeding to first bloom, and number of locules 
per fruit varied from no dominance to heterosis for a decrease of these char- 
acters. Considering all of the characters there is a continuous array of values 



GENE RECOMBINATION AND HETEROSIS 309 

(that is values in all classes) from heterosis for decrease of a character to 
heterosis for increase of a character, depending upon the character and geno- 
type (Fi hybrid). 

The most logical conclusion from these figures is that dominance and het- 
osis to a considerable extent are different degrees of expression of the same 
physiological genetic phenomena. This hypothesis is greatly strengthened by 
findings of Powers (1941) that whether a character shows dominance or het- 
erosis in some cases is dependent upon the environment and in other cases 
upon the genotype. As pointed out previously, gene recombination in rela- 
tion to heterosis is no different from combining any two desirable characters 
by recombination of genes. A study of the component characters of the main 
characters given in Table 19.7 offers further evidence in support of this 
contention. 

Yield of ripe fruit as determined by weight of fruit in grams is dependent 
upon number of fruits that ripen and weight per fruit. The first of these 
component characters, depending upon the Fi hybrid being considered, varies 
from partial dominance of fewer number of ripe fruits to heterosis for an in- 
creased number of ripe fruits. The second component character varies from 
no dominance to heterosis for smaller weight per fruit. They combine mul- 
tiplicatively, and in many cases result in heterosis for yield of fruit (Table 
19.7 and Powers, 1944). Here again, then, is a case involving combination 
of characters to produce heterosis. To retain some of the benefits of heterosis 
in inbred lines would involve recombination of the genes differentiating the 
two component characters. 

In turn the number of fruit that ripens is dependent to a large extent at 
Cheyenne, Wyoming, on earliness of maturity, number of days from seeding 
to first fruit ripe (Powers, 1945). Earliness of maturity varies from partial 
dominance of fewer days from seeding to first fruit ripe to heterosis for the 
shorter period. The component characters of earliness of maturity are period 
from seeding to first bloom, period from first bloom to first fruit set, and pe- 
riod from first fruit set to first fruit ripe. Number of days from seeding to 
first bloom varies from no dominance to heterosis for the shorter period. 
Number of days from first bloom to first fruit set varies from complete domi- 
nance of the longer period to heterosis of the shorter period. Number of days 
from first fruit set to first fruit ripe varies from partial dominance of the longer 
period to complete dominance of the shorter period. 

Weight per fruit is dependent upon weight per locule and number of 
locules per fruit. Weight per locule varies from partial dominance for less 
weight per locule to heterosis for greater weight per locule. Number of 
locules varies from no dominance to heterosis for fewer locules. On the 
arithmetic scale these two component characters combine multiplicatively 
so that weight per fruit varies from no dominance to heterosis for less weight 
per fruit. 



310 LEROY POWERS 

From the above, as was found true for yield per plant, the heterosis noted 
for earliness of maturity results from the combination of component charac- 
ters which in certain Fi hybrids may themselves exhibit heterosis. The same 
is true for weight per fruit. In other words, the study of genetics of heterosis 
has been somewhat simplified by breaking the main characters down into 
their component characters. Also, as before, the study shows that gene re- 
combination to retain some or all of the increase of the Fi hybrid over the 
parents is dependent upon the same physiological genetic phenomena as are 
involved in attempting to combine two or more desirable characters into a 
single inbred line. 

RECOVERING INBRED LINES RETAINING ADVANTAGES 
ATTRIBUTABLE TO HETEROSIS 

The physiological genetic phenomena that hinder or aid, by the recombi- 
nation of genes, the recovery of inbred lines retaining some or all of the 
advantages attributable to heterosis are the same as those emphasized by 
Jones (1917) and East (1936). These are the number of gene pairs differen- 
tiating the parents, linkage relations of the genes, pleiotropy, and the inter- 
action of the genes as determined by the measurement of end products, both 
interallelic and intraallelic. This genetic information can be obtained only 
by rather detailed genetic studies. With the quantitative characters such 
studies are expensive and time consuming. Hence, very few such studies have 
been made with tomato hybrids. Powers, Locke, and Garrett (1950) and 
Powers (1950b) have made a gene analysis for some of the main characters and 
their more obvious components. Even though the gene analysis for number 
of days from seeding to first fruit ripe has been completed for only one of 
the four crosses to be considered, this character and weight per locule will be 
treated as component characters of yield of ripe fruit per plant in the section 
dealing with number of pairs of genes differentiating the parents. 

Number of Gene Pairs Differentiating Parents 

In considering the bearing that number of gene pairs differentiating the 
parents has upon gene recombination and heterosis, just two characters will 
be considered: weight per locule and number of days from seeding to first 
fruit ripe. That both of these characters have an effect upon yield of ripe fruit 
should be kept in mind during the analyses and discussions which follow. 
Also, other component characters listed in Table 19.7 could be studied. How- 
ever, the additional information gained would not justify the time and space 
required, as the fundamental principles involved can be brought out from 
an analysis and discussion of the data for the two characters chosen. The 
number of gene pairs (effective factors; Mather, 1949) differentiating 
weight per locule has been determined for all the hybrid populations listed 
in Table 19.8. For days from seeding to first fruit ripe the number of gene 



GENE RECOMBINATION AND HETEROSIS 



311 



pairs (effective factors) differentiating the parents has been determined for 
the Porter X Ponderosa hybrid populations only. 

In discussing the bearing the number of gene pairs differentiating the two 
parents has upon gene recombination and heterosis, information concerning 
phenotypic dominance of the characters for the hybrid populations is neces- 
sary and will be derived by studying the means of the parental and hybrid 

TABLE 19.8 

MEANS FOR WEIGHT PER LOCULE AND NUMBER OF DAYS FROM SEED- 
ING TO FIRST FRUIT RIPE WITH TYPE AND NUMBER OF GENE PAIRS 
DIFFERENTIATING THE PARENTS FOR WEIGHT PER LOCULE* 





Danmark X 


Danmark X 


JOHANNISFEUER X 


Porter X 




Red Currant 


JOHANNISFEUER 


Red Currant 


Ponderosa 


Population 




No. of 




No. of 




No. of 




No. of 


Weight 


Days 


Weight 


Days 


Weight 


Days 


Weight 


Days 




per 


From 


per 


From 


per 


From 


per 


From 




Locule 


Seeding 


Locule 


Seeding 


Locule 


Seeding 


Locule 


Seeding 




(Gm.J 


to Fruit 
Ripe 


(Gm.) 


to Fruit 
Ripe 


(Gm.) 


to Fruit 
Ripe 


(Gm.) 


to Fruit 
Ripe 


Pit 


0.45 


156.9 


4.61 


164.9 


0.44 


126.0 


10.2 


147.7 


Bi to Pi 


0.97 


155.0 


6.72 


165,0 


1.04 


123.1 


11.8 


152.0 


F, 


2.33 


153.8 


7.96 


165,6 


2,70 


118.9 


14.4 


149.6 


F2 


2.12 


156.6 


8.35 


166.4 


2.12 


125.5 


13.5 


155.0 


Bi toPz 


4.82 


159.7 


8.32 


167.6 


4.48 


124.7 


13.7 


168.8 


P2t 


10.36 


169.8 


9.92 


170.0 


6.20 


136,1 


9.8 


204.8 


Type and num- 


















ber of pairs 


















of genes 


Minor 




Major 





Minor 




Major 


Major 




40 -f 




2 or 

3 




40+ 
Major 
2 or 
3 




3 


8 



* For the hybrid populations of Danmark X Red Currant, Danmark X Johannisfeuer, Johannisfeuer X 
Red Currant, and Porter X Ponderosa. 

t Pi is Red Currant, Johannisfeuer, Red Currant, and Porter, respectively. 
t P2 is Danmark, Danmark, Johannisfeuer, and Ponderosa, respectively. 



populations given in Table 19.8. The means for weight per locule of tomato 
fruits and number of days from seeding to first fruit ripe together with the 
type and number of gene pairs differentiating the parents for weight per 
locule for the hybrid populations of Danmark X Red Currant, Danmark X 
Johannisfeuer, Johannisfeuer X Red Currant, and Porter X Ponderosa are 
given in Table 19.8. 

The first two hybrid populations were grown at Cheyenne, Wyoming, in 
1938, the third hybrid population at the same location in 1939, and the last 
hybrid whose means are listed in the extreme right hand column of Table 
19.8 was grown at Woodward, Oklahoma, in 1941. The means of this table 



312 LEROY POWERS 

were taken from the following publications: Powers and Lyon (1941), 
Powers, Locke, and Garrett (1950), and Powers (1950a). The data will be 
analyzed to obtain information concerning the recombination of the genes 
differentiating weight per locule and number of days from seeding to first 
fruit ripe. Also, the data will be studied to ascertain the probable bearing 
this information has upon the production of inbred lines, by gene recombina- 
tion, that retain some or all of the advantages attributable to heterosis of 
yield of ripe fruit per plant which the hybrid populations would be expected 
to exhibit. 

On the arithmetic scale the Danmark X Red Currant populations show 
partial phenotypic dominance for smaller weight per locule. The parents of 
the Danmark X Red Currant hybrid were found to be differentiated by a 
large number of gene pairs (probably more than 40) which individually had 
minor effects. From these results it is evident that, if somewhere near one- 
half of the genes for smaller weight per locule in the Danmark X Red Cur- 
rant hybrid populations had entered the cross from one parent and the 
balance from the other parent, smaller weight per locule would have shown 
heterosis. Some of the genes must be linked because the parents have a 
haploid chromosome number of 12. In fact, since 40 or more pairs of genes 
are differentiating the parents, it seems highly probable that a system of 
linked polygenes is involved. With 40 pairs of genes differentiating the par- 
ents in the F2, to recover an individual possessing all of the genes for in- 
creased weight per locule (without linkage) would require a population of 
lO^'* individuals. The size of such a population can be appreciated by con- 
sidering the fact that 10" is 100 billion. The bearing this has upon the 
feasibility of recovering from segregating populations inbred lines retaining 
much of the advantage that might be exhibited by Fi hybrids is apparent. 

The Red Currant parent which possesses small weight per locule also 
possesses earliness of maturity. Hence, some of the genes tending to increase 
weight per locule are almost certain to be located on the same chromosomes 
with a non-beneficial gene or genes tending to increase the time required for 
maturity. However, due solely to the large number of gene pairs differentiat- 
ing weight per locule, with no close linkage, pleiotropy, or unfavorable in- 
terallelic and intraallelic interactions of the genes, only a comparatively 
small amount of the increased weight per locule of the Danmark parent could 
be combined with the earliness of maturity of the Red Currant parent by 
selection in the F2 or backcross populations. 

Weight per locule and earliness of maturity have a material influence on 
yield of ripe fruit per plant (Powers, 1945). In some crosses (see Tables 19.7 
and 19.8) greater weight per locule is at least partially dominant. Since the 
shorter period for days from seeding to first fruit ripe for the Danmark X 
Red Currant cross shows heterosis (Table 19.8) the hybrid populations would 
be expected to show heterosis for yield of ripe fruit per plant in crosses hav- 



GENE RECOMBINATION AND HETEROSIS 313 

ing such a polygenic system conditioning weight per locule, provided greater 
weight per locule was at least partially dominant, and provided the genes 
for increased weight per locule and shorter period from seeding to first fruit 
ripe were divided between the two parents. The analyses and discussions in 
the immediately preceding paragraphs show that in such an event it would 
be almost impossible to obtain inbred lines which through gene recombina- 
tion would retain any appreciable amount of the yield of the Fi hybrid. 

On the arithmetic scale the Johannisfeuer X Red Currant populations 
show partial phenotypic dominance of smaller weight per locule with the 
exception of the Bi to P2 which indicates no dominance. The parents of the 
Johannisfeuer X Red Currant hybrid populations were found to be differ- 
entiated by a large number of gene pairs (probably more than 40) each of 
which individually had minor effects and in addition by a few gene pairs 
(probably 2 or 3) having major effects. In these hybrid populations the total 
effect of the minor genes was greater than the total effect of the major genes. 
Again the shorter period from seeding to first fruit ripe showed heterosis. 

With the number and type of gene pairs conditioning weight per locule 
found for the Johannisfeuer X Red Currant hybrid, and provided the genes 
differentiating weight per locule exhibited at least partial dominance, as is 
indicated for the Danmark X Johannisfeuer populations, certain parental 
combinations of the genes would result in the hybrid populations showing 
heterosis for increased yield of fruit per plant. Since comparatively few ma- 
jor gene pairs differentiate weight per locule, it should be possible by re- 
combination of genes through selection in F2 and backcross populations of 
such a cross to combine into inbred lines some of the increased yield at- 
tributable to heterosis. 

The Danmark X Johannisfeuer hybrid populations show partial pheno- 
typic dominance for greater weight per locule, and complete dominance for 
shorter period from seeding to first fruit ripe. Two or three major gene pairs 
were found to be differentiating weight per locule. For weight per locule and 
number of days from seeding to first fruit ripe, dominance is such that had the 
genes tending to increase each of these two characters been divided between 
the two parents, the hybrid populations would have shown heterosis for both 
component characters. Likewise, if the above conditions had been fulfilled, 
yield of ripe fruit per plant would have shown heterosis in the hybrid popu- 
lations. 

The Porter X Ponderosa hybrid populations showed at least partial genie 
dominance for weight per locule (Powers, Locke, and Garrett, 1950). The 
parents were found to be differentiated by three pairs of genes and the genes 
tending to increase weight per locule were distributed between the two par- 
ents. As was to be expected, the hybrid populations showed heterosis for in- 
creased weight per locule. Period from seeding to first fruit ripe showed al- 
most if not complete dominance for the shorter period from seeding to first 



314 LEROY POWERS 

fruit ripe. The number of major gene pairs found to be differentiating the 
parents was eight. Due to the magnitude of the work involved it was not 
possible to measure yield of fruit, but in all probability the hybrid popula- 
tions of this cross would have shown heterosis for yield of ripe fruit per plant. 
In such an event it seems highly probable that some and perhaps a con- 
siderable amount of the increase in yield attributable to heterosis could be 
obtained in inbred lines through recombination of genes. 

Considering the data for all the crosses listed in Table 19.8 the informa- 
tion may be summarized as follows: In the Danmark X Red Currant cross 
a large number of gene pairs differentiates the parents and individually the 
genes have minor effects. The same is true of the Johannisfeuer X Red Cur- 
rant cross with the exception that two or three pairs of genes have major 
effects. In both the Danmark X Johannisfeuer and the Porter X Ponderosa 
crosses weight per locule is differentiated by a comparatively few pairs of 
genes having major effects. It is apparent that in the Porter X Ponderosa 
cross it should be possible by selection in the segregating populations to ob- 
tain by recombination of genes inbred lines equaling if not excelling the Fi 
fruits in weight per locule. 

The discussions treating weight per locule and number of days from seed- 
ing to first fruit ripe as component characters of yield of ripe fruit per plant 
reveal that the recombination of genes to retain some or all of the advantages 
of the Fi hybrid is analogous to recombination of genes for the purpose of 
combining desirable characters. 

Linkage Relations 

Linkage may be an aid or a hindrance to gene recombination. The data 
in Table 19.9 were computed to facilitate a consideration of the manner in 
which different linkage relations may affect recombination of genes. 

Certain assumptions were essential to a calculation of the data. First, it 
was assumed that the coefficient of coincidence is 1. Since in most cases there 
is interference, to assume a coefficient of coincidence of 1 is to err on the 
conservative side. For example, all the values given in the second row head- 
ing (with the exception of the first and last) would increase as the coefficient 
of coincidence became smaller. The reverse is true of the figures in the third 
and fourth columns. The frequencies listed in the second, third, and fourth 
columns of Table 19.9 are the theoretical number of individuals in the F2 pop- 
ulation carrying the 12 plus genes in the homozygous condition. The cross- 
over values expressed as decimal fractions are assumed to be equal for the 
different sections of the chromosomes delimited by any two adjacent genes. 

The conclusions to be drawn from the theoretical data of Table 19.9 are not 
invalidated by these assumptions. They merely serve the purpose of allowing 
the calculation of theoretical values for illustrative purposes. Other assump- 
tions such as different values of crossing over for the various sections of the 



GENE RECOMBINATION AND HETEROSIS 



315 



chromosomes and diflferent numbers of genes, combinations of genes in the 
parents, and number of linkage groups would not alter the conclusions to be 
drawn. In the illustration chosen only two linkage groups are shown and each 
has three pairs of genes. Also, the top row of genes represents the gamete from 
one parent and the lower row of genes the gamete from the other parent. In 
all three assumed cases, 3 plus and 3 minus genes entered the cross from 
each parent. 

It is evident that innumerable plausible cases could be assumed, but the 
fundamental principles derived from a consideration of the theoretical values 
given in the table would not be altered. One further assumption should be 

TABLE 19.9 

THEORETICAL NUMBER OF INDIVIDUALS IN THE F2 POPU- 
LATION THAT CARRY 12 PLUS (+) GENES WHEN THE PAR- 
ENTS ARE DIFFERENTIATED BY 6 PAIRS OF GENES, EACH 
OF 2 CHROMOSOME PAIRS CARRYING 3 PAIRS OF GENES* 



Cross- 


Linxage Relations in Fi (Number per Million) 


over 
Value 


(^^^)(;;;) 


(^i;)(;;^) 


(^;i)(;i;) 


0.000... 
0.075... 
0.225... 
0.375... 
0.450... 
0.500... 


62,500 

33,498 

8,134 

1,455 

523 

244 


0.000 

1.448 

57.787 

188.596 

234.520 

244.141 


0.000000 

0.000063 

0.410526 

24.441630 

105.094534 

244.140625 



* The crossover values for each section of the chromosome being equal and of the magni- 
tude shown. 

mentioned. In every case the plus genes are assumed to give an increase in 
some desirable quantitative character and, comparatively, the minus genes 
a decrease. Finally, in the table two extreme situations are shown, namely 
that in which there is no crossing over and that in which the two sections 
of the chromosome between adjacent genes show 50 per cent of crossing over. 

The data in the second column apply to that situation in which all of the 
plus genes occur in one member of the homologous chromosomes in each of 
the two pairs of chromosomes depicted. In the case of 50 per cent of crossing 
over or independent inheritance, only 244 individuals in a million of the F2 
population possess all twelve plus genes. The number of such individuals 
among a million F2 individuals increases with a decrease in the percentage 
of crossing over until with no crossing over 62,500 individuals in a million 
possess all six pairs of the plus genes in the homozygous condition. 

The data in the third column apply to that situation in which two plus 
genes are linked with one minus gene in one member of a chromosome pair 
and two minus genes with one plus gene in the other member of the same 



316 LEROY POWERS 

chromosome pair. In this column the situation is reversed as compared to 
column two. Again 50 per cent of crossing over gives 244 individuals among 
a million in the F2 possessing all twelve plus genes. This decreases with a de- 
crease in the percentage of crossing over until with no crossing over no indi- 
viduals in the infinite F2 population contain more than eight plus genes. 
However, since two of the plus genes are carried on the same chromosome 
in each of the two linkage groups, an increase in the linkage intensity results 
in an increased number of individuals in the F2 population possessing all 
eight plus genes in the homozygous condition. 

Here, then, is a case in which close linkage facilitates recombination of 
desired genes up to a certain number, and from a practical standpoint further 
advances by selection in that generation are impossible. Also, it would be 
difhcult to make further advances by continued selection in later genera- 
tions. In the F2 population with a crossover value of 0.075 the frequency of 

the ( T T _ ) ( T T _ ) genotype e.xpressed as a decimal fraction is 

0.183024 and of the (t t 7)(i t l) genotype is 0.014840. 

To obtain some F3 families derived from F2 plants of the latter genotype 
would require growing at least 300 selections in the F3 generation. To sepa- 
rate the F3 families derived from the F2 plants of the former genotype from 
those derived from the latter genotype would require an adequately replicated, 
well designed experiment. Anyone who has worked with the quantitative char- 
acters either in genetics or plant breeding realizes the difficulties besetting such 
a task. After such F3 families had been determined, only 25 per cent of the in- 
dividuals would be of the (TTT)(TT_) genotype. These would have 
to be tested in the F4 to separate them from F4 families derived from F3 plants 
of the (i i 7) (i t l) and the (^ ]j] ~) (^ ][] ~) genotypes. Even 

with the small number of genes assumed in the above example, it would not be 
a simple matter to make progress by continued selection in later generations. 
The addition of a few more genes having the plus and minus genes alternating 
on the same chromosome would make further progress by continued selection 
in generations later than the F2 practically impossible. From the above it is 
apparent that any series of plus genes being adjacent without minus genes in- 
tervening would facilitate recombination of desirable genes in the F2 genera- 
tion. It seems that in actual genetic and plant breeding materials many such 
combinations do exist. 

The figures in the fourth column of Table 19.9 are the theoretical fre- 
quency distributions for that situation in which the plus and minus genes 
alternate on the chromosome. Again the number of individuals expected in 
the F2 generations possessing all twelve plus genes decreases rather rapidly 
with a decrease in the percentage of crossing over. Even in the case of 50 per 



GENE RECOMBINATION AND HETEROSIS 317 

cent of crossing over it is doubtful whether it is possible for the plant breeder 
or geneticist to isolate individuals from the F2 population carrying twelve 
plus genes. 

The data in Table 19.9 emphasize that even with the probably over- 
simplified genetic situation depicted it is not possible to recover in a single 
individual all of the genes favorable to the production of a desirable charac- 
ter for which the F2 population is segregating, unless the favorable genes are 
located on the same chromosome and immediately adjacent to each other 
without unfavorable genes intervening. If any of the favorable genes are 
adjacent to each other without unfavorable genes intervening, then decided 
advances can be made by selection in the F2 populations up to a certain point. 
Beyond that point further selection in the F2 will have no effect, and selec- 
tion in advanced generations does not offer much promise. The most difficult 
situation is that in which the linkage relation is such that the favorable and 
unfavorable genes alternate on the chromosome and the number of such 
linkage groups is at a minimum for the number of gene pairs involved. 

For the sake of clarity of illustration only three linkage relations were 
shown. However, it is apparent that undoubtedly in the material available 
to plant breeders and geneticists, the possible different kinds of linkage rela- 
tions are almost innumerable. Some will aid the investigator in obtaining the 
desired recombination of genes and others will be a decided hindrance. In 
the cases of undesirable linkage relations it will be almost impossible for the 
breeder to obtain individuals possessing recombinations of genes making 
that individual equal to or superior to the Fi for the character exhibiting 
heterosis. On the other hand, desirable linkage relations may make it pos- 
sible to obtain the recombination of genes sought even though a large num- 
ber of gene pairs differentiates the parents used in hybridization. 

Pleiotropy, and Interallelic and Intraallelic Interactions 

Powers, Locke, and Garrett (1950) have made a rather detailed genetic 
study of eight quantitative characters in hybrid and parental populations 
involving the Porter and Ponderosa varieties of Lycopersicon esculentum 
Mill. The characters studied and the indicated number of major gene pairs 
differentiating the parents are as listed immediately below. 

Character Gene Symbols 

Percentage of flowers that set fruit FifiFifiFifiFifi 

Period from seeding to first fruit ripe: 

Seeding to first bloom BihiB-ibiBibz 

First bloom to first fruit set SiS\S2Si,SiSi 

First fruit set to first fruit ripe RxfiRiTi 

Weight per fruit: 

Number of locules LcilciLcildLcilci 

Weight per locule WiWiWiW'zWzWz 

With most quantitative characters it is difficult to distinguish between 
pleiotropy and linkage. It seems highly probable that linkage instead of plei- 



318 LEROY POWERS 

otropy produced the relations noted by the above authors between the four 
series of genes F/, Ss, Rr, and Lclc with the exception of the Ff and Ss re- 
lation, because all the associations noted are those expected on the basis of 
linkage. If pleiotropy were involved, such relations would be coincidental, 
which for all these gene series is highly improbable. However, as pointed out 
by Powers, Locke, and Garrett (1950) some of the genes of the Ff and Ss 
series must be identical, as percentage of flowers that set fruit has an effect 
on period from first bloom to first fruit set. The Lclc and Ww series of genes, 
differentiating number of locules and weight per locule, respectively, were 
independent as regards linkage and pleiotropy. In these studies pleiotropy 
was not of major importance. 

Phenotypic and genie dominance furnish some information concerning 
the interallelic and intraallelic interactions of the genes. That genie domi- 
nance is dependent upon the genotypic milieu was pointed out by Fisher 
(1931) and many others (Dobzhansky, 1941). Hence both interallelic and 
intraallelic interactions as measured by end products are second order inter- 
actions, genes X genes X the environment. 

Any of the interactions of genes noted as affecting any of the component 
characters dealt with in the study by Powers, Locke, and Garrett (1950) 
were interactions of genes differentiating yield of ripe fruit per plant. With 
this fact in mind, it is interesting to note the interactions of the genes differ- 
entiating the component characters. The intraallelic and interallelic interac- 
tions of the Ff gene series were such that genie dominance was intermediate. 
The intraallelic and interallelic interactions of the Bb series of genes were 
such that one of the six dominant genes shortened the period from seeding to 
first bloom as much as all six, which shows that both dominance and epistasis 
were complete. For the 5^ series and Rr series of genes, genie dominance was 
complete. Also, the effects of the gene pairs were cumulative. 

Had the dominant genes of the Ss series entered the cross from one parent 
and the dominant genes from the Rr series entered the cross from the other 
parent, the Fi hybrid would have shown heterosis for earliness of maturity. 
Porter would then represent an inbred line which by recombination of 
genes retained the earliness of maturity of the Fi hybrid. Genie dominance 
was partial for genes (LciLc^) tending to produce fewer locules per fruit and 
for the (Lcs) tending to produce more locules per fruit. A series of genes such 
as Lci and Lc2, some entering the cross from one parent and some from the 
other, would produce an Fi hybrid showing heterosis for fewer locules per 
fruit. On the other hand a series of genes such as (Lcs), some entering the 
cross from one parent and some from the other, would produce heterosis for 
more locules per fruit. 

Finally, for the Wiv series of genes, genie dominance was partial for in- 
creased weight per locule and the effects of the gene pairs were cumulative. 
As regards this character, both parents did contribute genes for increased 



GENE RECOMBINATION AND HETEROSIS 319 

weight per locule, and the Fj hybrid did show heterosis for increased weight. 
Also, as has been shown in the F2 and Bi and P2 populations some individuals 
were obtained having greater weight per locule than the Fi plants and this 
greater weight per locule proved to be due to recombination of favorable 
genes. 

Also, the interallelic interactions of the genes as determined by the inter- 
relations of the component characters are of interest because of the informa- 
tion they provide concerning recombination of genes and heterosis. The 
effects of the Bb series of genes, the ^'^ series, and the Rr series, respectively, 
were found to be cumulative. On an average the S genes would be e.xpected 
to shorten the period from first bloom to first fruit set less in the presence 
of the R genes than in the presence of the r genes — if the physiological reac- 
tions affecting these two component characters that were instigated by the 
environment were the same as those instigated by the 6'5 and Rr gene series. 
That such was the case seems probable from the results of Goldschmidt's 
work (1938) with phenocopies. In fact it seems almost axiomatic that this was 
the case, because the second order interaction (Ss gene series X Rr gene 
series X environment) was such that, on an average, when the ^5 series 
responded to a given environment by shortening the period from first bloom 
to first fruit set the Rr series in the same plant tended to produce a longer 
period from first fruit set to first fruit ripe. Then the effects of these two 
series of genes were less than additive as regards the dependent character pe- 
riod from seeding to first fruit ripe 

About the same situation existed in respect to the Lclc series and the Ww 
series of genes in that greater number of locules, on an average, was ac- 
companied by less weight per locule. This type of interallelic interaction 
would tend to decrease the possibility of obtaining inbred lines combining 
desirable characters. This would be particularly true of the interallelic inter- 
action between the Ss and Rr gene series, because a shorter period from first 
bloom to first fruit set tended to be accompanied by a longer period from 
first fruit set to first fruit ripe 

The data do not furnish any evidence concerning that type of intraallelic 
interaction postulated by East's (1936) physiological hypothesis, other than 
to say that no cases of overdominance were found. This would indicate that 
probably overdominance does not play a predominant part in the produc- 
tion of heterosis in the tomato hybrids studied. 



A. J. MANGELSDORF 

Experiment Sfation, Hawaiian Sugar Planters Assodaiion , Honolulu, T.H. 



Chapter 20 



Gene Interaction in Heterosis 



Sugar cane behaves very much like corn in its reaction toward inbreeding 
and outcrossing. Although the sugar cane flower is normally provided with 
both male and female organs, male sterility is not uncommon. Among the 
varieties that produce an abundance of pollen, many are partially or highly 
self-sterile. As a consequence, cross-fertilization by wind-borne pollen is the 
rule in sugar cane, as in corn. When sugar cane is subjected to self-pollina- 
tion, the usual result is a reduction in seed setting and a marked reduction in 
the vigor of the offspring. 

The sugar cane breeder enjoys one great advantage over the corn breeder: 
sugar cane can be propagated asexually. Each node on the stalk is provided 
with a bud and with a number of root primordia. In field practice, stalks of 
the selected variety of sugar cane are sectioned into cuttings of two or more 
internodes each. These cuttings are then placed horizontally in furrows and 
covered lightly with soil. In due course the cutting sends out its roots, the 
buds develop into shoots, and a new plant is established. 

Were it possible to apply this procedure to corn, and thus to perpetuate 
outstanding individuals from whatever source, it is unlikely that the corn 
breeder would have felt obliged to resort to the laborious procedures now 
employed. 

When sugar cane varieties are propagated by cuttings, the traits by which 
we are able to distinguish one variety from another maintain their integrity 
through many cycles of clonal propagation. This is true not only of morpho- 
logical traits, but also of physiological traits. 

Sugar cane has a number of relatives growing in the wild, some of which 
may be ancestral to the original cultivated forms. Wild Saccharums are wide- 
ly distributed in the tropical and sub-tropical regions of the Old World, from 
central Africa through Asia and Malaya, to and including the Indonesian 
and many of the more westerly Pacific islands. This heterogeneous array of 

320 



GENE INTERACTION IN HETEROSIS 321 

wild forms has been somewhat arbitrarily classified into two great groups — - 
the S. spontaneum group and the S. robiistum group. Each of these groups 
comprises a diversity of tyj)es which differ among themselves in morphology 
and in chromosome number. The members of the sponlanenni group have 
slender stalks; they are often strongly stoloniferous. The members of the 
robust U7n group have hard, woody stalks, sometimes of good diameter; sto- 
lons, if present, are not strongly developed. 

The original cultivated varieties likewise may be classified into two great 
groups. The first of these comprises a number of slender varieties which ap- 
pear to be indigenous to India, and which have been lumped together under 
the name S. Barberi. Certain of the Barberi varieties bear a striking resem- 
blance to the wild spontaneums of that region. 

The New Guinea region is the home of a group of large-stalked tropical 
cultivated varieties of the type which Linnaeus named S. officinarum. The 
wild form most closely resembling S. officinarum and possibly ancestral to 
it is S. robustum, which is indigenous to that region. 

In the closely related genus Sorghum, the difference between varieties 
having pithy stalks containing but little sugar, and varieties with sweet 
juicy stalks, has been shown to be determined by a single major gene. In 
Saccharum the change from the dry, pithy, low-sucrose stalks of the wild 
forms to the juicy, high-sucrose stalks of the cultivated varieties appears to 
have been brought about by several, but perhaps by no more than three or 
four major gene changes. 

The cultivated and wild forms also differ in genes for stalk size. In crosses 
between the two, the genes responsible for the slenderness of the wild forms 
show a high degree of dominance. 

A striking feature of this multiform genus is the prevalence of inter- 
fertility among its members. Widely divergent forms can be crossed without 
undue difficulty. The resulting hybrids are rarely completely sterile; they 
are often highly fertile. The explanation is presumably to be sought in the 
polyploidy which is characteristic of both the wild and the cultivated forms. 
They range in chromosome number from 24 to 80 or more pairs. It appears 
that once the minimum chromosomal complement needed to produce a func- 
tional zygote has been supplied, there is considerable latitude in the number 
and in the assortment of chromosomes that can be added without impairing 
the viability, or even the fertility of the hybrids. 

Since the breeder is as yet unable to create superior genes at will, he is 
obliged to content himself with developing new combinations of the genes 
available in whatever breeding material he may be able to assemble. The 
sugar cane breeder is fortunate in having in the wild relatives of sugar cane a 
reservoir of genes for disease-resistance and hardiness. Those are traits that 
had to some degree been lost in the course of domestication. Considerable 



322 A. J. MANGELSDORF 

use has already been made of the wild forms. The important varieties today 
are almost without exception complex hybrids that include in their ancestry 
representatives of both the S. officinarum and the S. Barberi groups of cul- 
tivated varieties, together with representatives of one or both of the wild 
species. 

Thus the sugar cane breeder has been exploiting, to the best of his ability, 
the advantages that heterosis has to offer. He is, however, acutely aware that 
a better understanding of the genetic basis of heterosis is prerequisite to its 
more effective utilization. Since he suffers the disadvantage of isolation from 
the centers of research, he cherishes such rare opportunities as he may have 
to peer over the shoulder of the research worker, to whom he must look for 
new facts that may lead to a better understanding of the mechanism of gene 
action and thus, of heterosis. 

Recently some of us who are engaged in sugar cane breeding in Hawaii 
formulated a number of postulates with the object of providing a basis for 
discussing heterosis and related matters. These postulates have been ex- 
cerpted or inferred from the published literature and from correspondence 
with workers engaged in genetic research, whose helpful suggestions are 
gratefully acknowledged. 

Although the evidence supporting these postulates is sometimes meager, 
and sometimes capable of other interpretations, we have deliberately phrased 
them in a categorical vein in the belief that they might thus better serve 
their primary purpose — that of provoking a free exchange of ideas. 

POSTULATES RELATING TO INCIDENCE OF LESS FAVORABLE ALLELES 

1. Naturally self-fertilized populations tend to keep their chromosomes 
purged of all alleles other than those which in the homozygous condition 
interact to best advantage with the remainder of the genotype and with the 
existing environment^ to promote the result favored by natural selection (or 
by human selection). This does not imply that any single population will con- 
tain all of the best alleles existing in the species. Selection can make a choice 
only between the alleles present in the population. 

2. In addition to their prevailing (normal, plus, or wild type) alleles, cross- 
fertilized organisms such as corn and sugar cane carry in the heterozygous 
condition, at many loci, recessive alleles which in the homozygous condition 
would be inferior in their action to that of their normal or prevailing partners. 

3. These less favorable alleles may be thought of as belonging to one of two 
classes, which, although differing in their past history, may have similar 
physiological consequences: (a) fortuitous, resulting from sporadic mutation, 
and representing the errors in the "trial and error" of the evolutionary proc- 
ess; or {b) relic, representing the residue of what were once the prevailing 

1. The term environment is here used in a broad sense to mean the sum-total of the ex- 
ternal influences acting upon the organism, including its nutrition. 



GENE INTERACTION IN HETEROSIS 323 

alleles but which, in the course of evolution or under a changed environment, 
have been displaced, to a greater or lesser degree, by still better alleles. 

4. The prevailing allele at a given locus has reached its pre-eminent posi- 
tion through the sifting action of natural selection over many generations. 
Given a stable environment, further improvement, through mutation, at that 
locus would long since have materialized if the chances for such improvement 
were high. It is not strange that random mutation should only rarely be able 
to produce a superior new allele. Nevertheless, once the possibilities for im- 
provement through recombination of existing genes have been exhausted, 
further evolutionary progress will be contingent upon just such an event, 
however rare its occurrence may be. 

5. Whether dominant or recessive, and whether in a naturally self-ferti- 
lized or naturally cross-fertilized population, a substantially superior mutant, 
once established in the population, is destined to increase in frequency and to 
become the prevailing allele in the population. 

6. A deleterious dominant is doomed to eventual extinction. In a cross- 
breeding population of sufficient size a deleterious recessive may persist in- 
definitely, its incidence, except for random drift, being determined by the 
balance between its elimination by selection and the rate at which it recurs by 
mutation. 

7. The best allele for one environment may not be best for another envi- 
ronment. The burden of less favorable alleles which cross-fertilized organisms 
carry along generation after generation is not an unmitigated liability. It 
serves as a form of insurance by providing a reservoir of adaptability to 
changing conditions. 

ROLE OF LESS FAVORABLE ALLELES 

Turning now to the role of these less favorable alleles in the heterosis 
phenomenon as manifested in naturally cross-fertilized organisms we may 
formulate a second group of postulates: 

1. At many and perhaps at most loci, .la is as good or nearly as good as 
A A, and both A A and A a are better than aa. 

2. There may be a few loci where aa is better than .LI or .la. This is par- 
ticularly likely to be the case for loci affecting traits which are advantageous 
under domestication, but disadvantageous in the wild under natural selec- 
tion. 

3. There may, for all we know, be occasional loci where .LI' is better than 
A A or A' A' (overdominance). 

4. There may be many regions in the chromosomes which behave as though 
A A' were better than /L4 or A' A'. With deleterious recessive alleles in the 
heterozygous condition at many loci, it seems almost inevitable that some of 
these will be closely linked in the repulsion phase, as for example Ab/aB, 
which in the absence of crossing over would behave as a single locus, the 



324 A. J. MANGELSDORF 

heterozygous condition of which is superior to either homozygote. It is to be 
expected that such a linkage will eventually be broken. However, there may 
be regions in the chromosomes, such as the centromere region, for example, 
where crossing over is reduced, and where a group of genes may act indefi- 
nitely as a single gene. We may for convenience designate the effect of such 
reciprocal apposition of favorable dominants to their less favorable reces- 
sives as a pseudo-overdominance effect. It will be noted that such a balanced 
defective situation conforms with the dominance and linkage hypothesis ad- 
vanced by Jones as an explanation of the heterosis phenomenon. 

5. Even in the absence of linkage, an overdominance type of reaction (but 
resulting from pseudo-overdominance) must assert itself whenever each of 
the two members of a pair of gametes is able to supply the favorable domi- 
nant alleles required to counteract the less favorable recessives carried by 
the other member of the pair. The likelihood of success in retaining, in suc- 
cessive generations of selfing, all of the favorable dominants heterozygous 
in Fi, and eliminating all of the less favorable recessives, diminishes ex- 
ponentially with increasing numbers of loci heterozygous in Fi. It would 
seem that naturally cross-fertilized organisms which carry, at many loci, 
deleterious recessives of low per locus frequency in the population could 
hardly fail to manifest a pseudo-overdominance type of response to inbreed- 
ing and outcrossing. 

6. From an evolutionary standpoint, it may be important to distinguish 
between the consequences of (a) true overdominance (heterozygosis at the 
locus level) and {b) pseudo-overdominance (heterozygosis at the zygote level 
resulting from the reciprocal masking of deleterious recessives by their 
dominant alleles). From the standpoint of the breeder who is of necessity 
working against time, this distinction may have little practical importance 
if many loci are involved in the pseudo-overdominance effect. A breeding 
plan designed to deal efficiently with one of these alternatives should be 
effective also in dealing with the other. 

7. Whether due to true overdominance or to pseudo-overdominance, the 
widespread if not universal occurrence among naturally cross-fertilized or- 
ganisms of an overdominance type of response to inbreeding and outcrossing 
poses a problem which the breeder cannot afford to disregard. 

8. Neither overdominance nor pseudo-overdominance can be called upon 
to explain the differences in vigor between different varieties of wheat, beans, 
sorghums, and other self-fertilized forms. Such differences are determined by 
genes in the homozygous state, as are also the differences between homozy- 
gous inbred lines of corn. 

ROLE OF LIMITING FACTORS 

A consideration of the role of limiting factors in quantitative inheritance 
leads us to a third group of postulates: 



GENE INTERACTION IN HETEROSIS 325 

1. The adequacy of a diet is determined not by those constituents which 
are present in ample amounts, but by those which are deficient to the point of 
acting as limiting factors. Similarly the excellence of a genotype is deter- 
mined not by its strongest but by its weakest links. The term weak link as 
here employed refers to a gene pair at a particular locus which at some mo- 
ment in the life of the organism proves so inadequate in performing the task 
required of that locus as to act as a limiting factor — a bottleneck in an essen- 
tial physiological process. A bottleneck efifect may result from a deficiency of 
an essential gene product or from an excess of a gene product. 

2. At each moment throughout its life the physiological processes of even 
the most vigorous organism are held down to their prevailing rates by bottle- 
necks or limiting factors. We are merely rephrasing a genetic axiom when 
we say that a bottleneck in the physiological reaction system is neither purely 
genetic nor purely environmental. The physiological bottleneck at any given 
moment results from the interaction of a particular locus (which we may for 
convenience refer to as the bottleneck locus) with the remainder of the geno- 
type and with the environment of that moment. When we speak of an en- 
vironmental bottleneck, we are merely focusing attention upon the environ- 
mental component of the genetic-environmental bottleneck. When we speak 
of a bottleneck gene, we are referring to the genetic component of the genetic- 
environmental bottleneck. 

3. The value of an otherwise perfect diet would be seriously impaired by 
the omission of a single essential element. Similarly an otherwise superior 
genotype could be rendered mediocre or worse by a single bottleneck. A po- 
tentially superior genotype is unable to manifest its potentialities so long as it 
is being throttled by a genetic-environmental bottleneck. A breeder looks at 
the bottleneck and sees the need of a better allele at the bottleneck locus. 
An agriculturist looks at the same bottleneck and sees the need for correcting 
its environmental component. Bottlenecks relating to climatic limitations 
usually can be most economically dealt with by breeding.- On the other hand, 
bottlenecks resulting from nutritional deficiencies can often be advantage- 
ously dealt with by correcting the environment. 

4. The substitution, at a bottleneck locus, of a better combination of al- 
leles^ will result in an improvement in yield providing that no other limiting 
factor, genetic or environmental, asserts itself before an appreciable gain has 
been realized. 

5. The substitution of potentially better alleles at loci other than bottle- 
neck loci cannot substantially improve yields any more than the addition of 
calcium to the diet of a plant or an animal can relieve the effect of a phos- 
phorus deficiency in that diet. We take it for granted that each essential 

2. This rule is not without exceptions. For example, a bottleneck resulting from a 
deficiency of rainfall can sometimes be economically eliminated by irrigation. 

3. As already indicated, the best combination of alleles may be AA, Aa, or aa depend- 
ing upon the particular locus. 



326 A. J. MANGELSDORF 

chemical element has its specific role to perform in the physiological reaction 
system. Similarly we accept as well established the thesis that gene action is 
likewise specific — that a particular gene can perform its particular function, 
and that function only. Nevertheless we sometimes engage in speculations 
which ignore these convictions and which appear to assume that genes affect- 
ing quantitative characters such as yield are freely interchangeable, one with 
another, and that one yield gene can serve as well as another, regardless of 
its locus or function. 

6. A bottleneck locus may act as such throughout the life of the individual 
or it may act as a limiting factor only for a short period and under specific 
conditions, such as drought, nitrogen deficiency, or excessively high or low 
temperatures. Under a varying environment the bottleneck of one moment 
may be superseded by a different bottleneck at the next moment. 

7. The physiological bottleneck may be ameliorated or removed by correct- 
ing the particular feature of the environment contributing to the bottleneck. 
In the examples cited above this would entail supplying moisture, or nitrogen, 
or lowering or raising the temperature. Or the bottleneck may be ameliorated 
or removed by substituting a more effective allele at the bottleneck locus, 
providing that such an allele is available. 

8. As already indicated, the amelioration or removal of a bottleneck, either 
by improving the environment or by substituting a better allele at the 
bottleneck locus, will permit a rise in the rate of the essential physiological 
processes. This rise may be small or it may be large, depending upon the 
point at which the next ensuing bottleneck begins to make itself felt. The 
substitution of a more efficient allele at a bottleneck locus in a certain geno- 
type, under a particular environment, may result in a large gain. The substi- 
tution of the same allele in a different genotype or under another environ- 
ment may result in little or no gain. It is not strange that difficulty should 
be encountered in analyzing the inheritance of genes affecting yield and other 
quantitative characters which are subject to the influence of a varied and 
fluctuating array of genetic-environmental bottlenecks. 

9. A diet that is low in calcium may supply calcium at an adequate rate so 
long as growth is being retarded by a lack of phosphorus. But once phos- 
phorus is supplied at an adequate rate, calcium deficiency becomes a bottle- 
neck which limits the rate of growth. Similarly a mediocre gene m at one 
locus may be adequate (not a bottleneck) so long as the rate of physiological 
activity of the organism is being throttled by environmental limitations or 
by a bottleneck gene at some other locus. But once the other genetic-environ- 
mental limiting factors have been removed, the mediocre gene m is unable 
to handle the increased load and becomes the bottleneck in the reaction 
system. 

10. The maximum vigor or yield possible under a given environment will 
be attained when the organism is endowed with the best available allele or 



GENE INTERACTION IN HETEROSIS 327 

combination of alleles at each bottleneck locus. There are presumably many 
loci that never act as bottlenecks in any part of the reaction system affecting 
vigor or yield, no matter which allele or combination of alleles happens to 
occupy such a locus. 

11. The difference between the weakest inbred and the most vigorous hy- 
brid is merely one of degree. Each represents an integration of the many 
genetic-environmental bottleneck effects under which it has labored. The 
weak inbred has been throttled down by one or more bottlenecks to a low 
level. The superior hybrid is able to go much further, even attaining what we 
might concede to be extreme vigor. But both the weak inbred and the vigor- 
ous hybrid have throughout their lives been held down to their respective 
levels by their genetic -environmental bottlenecks.^ 

MISCELLANEA 

The fourth and last group of postulates comprise a heterogeneous popula- 
tion randomly listed as separate topics for discussion. 

1 . If each step in a complex physiological process such as photosynthesis is 
conditioned by the action of a specific gene, and if each successive step in the 
chain of reactions is contingent upon the successful completion of the pre- 
ceding steps, it follows that in attempting a biomathematical analysis of the 
inheritance of quantitative characters such as yield we may not be justified 
in assuming, as a basis for our calculations, that each of the genes concerned 
is independent in its action. 

2. Since our efforts to "improve" the genotype are constantly being 
thwarted by bottleneck genes, we may be tempted to damn all such genes as 
inventions of the Devil. No doubt there are many defective genes that would 
have to be classed as liabilities under any normal environment. But certainly 
there are many bottleneck genes that are indispensable to survival — genes 
that act as governors in regulating physiological reactions and in fitting the 
organism to its particular ecological niche. A mouse or a moss can survive and 
reproduce where larger organisms would perish. And a mouse which, as a 
result of changes in certain of its adaptive bottleneck genes attained the size 
of a rat, might find itself at a disadvantage in a community of normal mice. 

3. If we are correct in assuming that even a single major bottleneck locus 
can act as a limiting factor in the development of an otherwise superior geno- 

* Certain of the foregoing postulates pertaining to the role of bottleneck genes in quanti- 
tative inheritance may be guilty of gross over-simplification. So complex is the physiological 
reaction system of even the simplest organism that we are only now beginning to gain an 
inkling of the extent of its complexity. These postulates may also be guilty of exaggeration. 
Because we believe that the action of limiting factors in quantitative inheritance has not re- 
ceived the attention that it deserves, we have intentionally stressed the importance of the 
bottleneck locus, even at the risk of over-emphasis. Furthermore, we have pictured the 
limiting factor at a given moment as pertaining to a single bottleneck locus. This may or 
may not be the rule. It would not be difficult to imagine a bottleneck which pertains to sev- 
eral loci and which could be relieved or eliminated by substituting a more effective allele at 
any one of these loci. 



328 A. J. MANGELSDORF 

type, it is hardly to be expected that the phenotype of an inbred line will 
afford a wholly reliable indication of its breeding potentialities in hybrid 
combinations. 

4. We need to keep in mind the limitations that pertain to a rating for gen- 
eral combining ability. The best "general" combiner thus far discovered in 
corn is not so general in its combining ability as to be able to combine to ad- 
vantage with itself or with any other genotype that happens to be afflicted 
with the same bottleneck genes. At best, a rating for general combining abil- 
ity can represent nothing more than an average arrived at by lumping a given 
population of specific combinations. An average derived from a different pop- 
ulation of specific combinations could result in quite another rating. 

5. If aseriesof inbreds^,5, etc., be crossed with a tester inbred T, we ob- 
tain the hybrids AT, BT, etc. The yield of AT will be determined by the 
bottleneck genes in the A T genotype. The yield of BT will be determined by 
the bottleneck genes in the 5 T genotype. The test cross can tell us which lines 
combine to best advantage with the tester line, but it cannot reasonably be 
expected to tell us more than that. It cannot, for example, tell us with cer- 
tainty what we may expect from A X B. Both A and B may combine to 
advantage with T, but if A and B each happen to be afflicted with one or 
more of the same bottleneck genes (not present in T) the yield of the cross 
AB will suffer. 

6. The failure of a cross between two convergently improved lines to equal 
the cross between the two original lines from which they were derived cannot 
be taken as critical evidence for the existence of an overdominance mecha- 
nism. The benefits which convergent improvement seeks to achieve can be 
vitiated if a recessive bottleneck gene b, present in only one of the original 
parent lines, should become homozygous in both convergently "improved" 
lines. Selection exercised with the object of preventing such an occurrence 
may be ineffective if b becomes a bottleneck only under the enhanced rate of 
physiological activity of the A{B) X B{A) hybrid. 

7. During recent years several examples of heterosis reported in the litera- 
ture have been attributed to the effect of heterozygosity at a single locus. 
When the amount of heterosis is substantial, it should be possible to verify 
the validity of the hypothesis by breeding tests. If the two parents are really 
isogenic, except for the heterosis locus H, and if HiH2 individuals are more 
vigorous than either homozygote, then by selfing only the most vigorous in- 
dividuals in each generation it should be possible to retain in one-half of 
the population the original vigor of Fi even after many generations of selfing, 
and such a line should continue indefinitely to segregate HiHi, HiH-i, and 
H<iH2 individuals in a 1:2:1 ratio. 

8. East describes the effect of heterosis as "comparable to the effect on a 
plant of the addition of a balanced fertilizer to the soil or to the feeding of a 
more adequate and more chemically complete diet to the animal." The simi- 



GENE INTERACTION IN HETEROSIS 329 

larities noted by East between the beneficial effects of heterosis and those of 
improved nutrition are more than coincidental. The first prerequisite for en- 
hanced well-being is the removal of the bottlenecks that stand in the way- 
This can sometimes be accomplished by improving the nutrition, sometimes 
by substituting more efficient alleles at the bottleneck loci, and sometimes by 
both. 

9. The term heterosis remains ambiguous in spite of the many attempts to 
define it. It continues to have different meanings for different workers. 

10. If heterosis is to be measured by comparing performance of offspring 
with performance of parents, then the higher the standing of the two parents 
in the scale of measurement, the lower the degree of heterosis to be expected 
in their offspring. Conversely the lower the standing of the parents, the great- 
er the heterosis to be expected. (Exceptions to the latter rule will occur when 
both parents owe their enfeeblement to the same bottleneck genes.) 

1 1. Success in crop and livestock production depends largely upon the skill 
of the grower in detecting, diagnosing, and correcting the environmental com- 
ponents of the bottlenecks affecting yield. Success in developing higher yield- 
ing genotypes depends largely upon the ability of the breeder to substitute 
more effective alleles at the bottleneck loci, and to accomplish this without 
establishing new and equally serious bottlenecks at other loci. 



GORDON E. DICKERSON 

UnivarsHy of Missouri 



Chapter 21 



Inbred Lines for Heterosis Tests? 



The justification for considering heterosis tests in breeding work rests on the 
mode of action and interaction of the genes responsible for genetic variability 
in the material available to the geneticist. The nature of this genetic variabil- 
ity may vary widely between species or populations in response to differences 
in the degree of inbreeding and kind of selection, natural or imposed, that 
has characterized the population over an extended period. For any given 
trait or combination thereof, structure of genetic variation will depend upon 
how consistent, intense, and prolonged selection has been. 

It follows that choice of the system of mating and selection appropriate 
for most rapid improvement in economic attributes of any given plant or 
animal population should be guided by as complete knowledge of the kind 
of genetic variation in the population as analysis of all available data affords. 
The discussion which follows is an attempt (1) to interpret the evidence 
presently available concerning the sort of genetic system which underlies 
important economic traits, using swine as the example; and (2) to compare 
expected effectiveness of several alternative breeding methods. 

NATURE OF GENETIC VARIATION IN ECONOMIC TRAITS 

Types of association between the genotype and its phenotypic expression 
have been classified logically as intra-allelic and inter-allelic. The former 
includes all degrees of dominance or levels of expression for the heterozygote 
relative to the corresponding homozygotes. The concept of heterozygote ad- 
vantage or overdominance differs from the usual ideas of dominance in that 
each gene is visualized as exerting certain dominant favorable effects lacking 
in its allele. Inter-allelic gene action or epistasis includes all effects of a gene 
in one set of alleles on the expression of genes in other sets of alleles. Comple- 
mentary, inhibiting, duplicate dominant, and duplicate recessive gene inter- 
actions are extreme examples. 

330 



INBRED LINES FOR HETEROSIS TESTS? 331 

By definition, epistasis is universal in the sense that expression of every 
gene is to some degree dependent on and modified by the effects of genes in 
other sets of alleles. Epistasis would include fixed multij)licative or propor- 
tional effects of each gene on the expression of non-allelic genes. Such 
epistasis, although unlikely to be important, would be of a highly predictable 
sort and would disappear if phenotypes were measured in log scale units. 
A potentially much more important sort of epistasis would be that involved 
whenever a phenotypic maximum is associated with an optimum genetic 
intermediate (Wright, 1935). Here a given gene may have either a positive 
or a negative selective value, depending on whether the individual's average 
genotype is above or below the optimum genetic intermediate. 

Some of the evidence concerning the kind of genetic variability with which 
we must deal in seeking to improve economic characters of swine has been 
considered earlier (Dickerson, 1949, 1951) and may be summarized here as 
follows: 

Inbreeding and Crossbreeding Effects 

Proportion of heterozygous loci has a major influence on total perform- 
ance, affecting most the highly important but lowly heritable characters for 
which selection has been consistently in one direction. Take for example, an 
intra-season comparison of 538 inbred and 325 linecross litters from the same 
lines in four projects of the Regional Swine Breeding Laboratory (Dickerson 
et al., 1947). This showed a decline in performance per 10 per cent increased 
inbreeding of litter amounting to 2.6 and 7.8 per cent, for litter size at birth 
and weaning, respectively; 2.6 per cent for pig weight at 154 days of age; and 
11.4 per cent for total weight of litters at 154 days. Similar estimates per 10 
per cent increased inbreeding of dam, based on sixty-three inbred and fifty 
linecross dams at the Iowa Station, were 2.1 and 5.0 per cent for litter size 
at birth and weaning; 1.6 per cent for pig weights at 21 days; and 5 per cent 
for total weight of litters at 154 days. 

Results from studies of regression of j)erformance on inbreeding of dam 
and litter within line and season (Blunn and Baker, 1949; Stewart, 1945; 
Comstock and Winters, 1944; and Hetzer et al., 1940) agree quite well with 
the figures given. Inbreeding of dam and litter together greatly depresses 
prolificacy, suckling ability, pre- and post-natal viability and growth rate, 
and particularly their product — total litter weight. Inbreeding effects on 
carcass composition, body conformation, and efficiency of food utilization 
were relatively minor (Dickerson et al., 1946). 

The results of the earlier crossbreeding exi)eriments have been summarized 
by Lush (1939) and Winters (1936). When the mean of the two purebred 
stocks crossed is compared with the crossbred litters, the results of many ex- 
periments summarized by Carroll and Roberts (1942) indicate that the 
average performance of crossbred individuals is increased about as much as 
it would be by a 10 per cent reduction of inbreeding (see Table 21.1). More 



332 



GORDON E. DICKERSON 



recent studies of crossbreeding using inbred strains (Hazel el al., 1948; Sierk, 
1948) verify the earlier conclusions. 

Some degree of dominance is the most obvious genetic mechanism by 
which change in heterozygosity from inbreeding or crossbreeding would affect 
the level of performance. Inbreeding decline due to dominance would be a 
function of 2q{l — q)k f, where q is frequency of the dominant allele, / is 
Wright's inbreeding coefficient, and k is the degree of dominance (Hull, 
1945) defined in terms of phenotypic scale as (2 Aa-AA-aa)/{AA-aa). 

TABLE 21.1 

RESULTS OF CROSSBREEDING EXPERIMENTS SUM- 
MARIZED BY CARROLL AND ROBERTS (1942) 



Factors of 
Production 


No. of 
E.xpts. 


Mean of Two 
Pure Breeds 


Mean of 
Crossbreds 


Relative Per- 
formance of 
Crossbreds 
with Purebreds 
= 100 


No. pigs per litter 

Birth weight of pigs (lbs.) . 

Survival ability (%) 

Weaning wt. of pigs (lbs.) . 

Weaning wt. of litters 

(lbs.)* 


12 

6 

15 

15 

13 
9 
6 

32 
32 


9.74 
2.77 

76.3 

32.5 

235.6 

1.381 
374.1 

1.359 
345.4 


9.48 
2.79 
80.2 
33.12 

254.1 

1.436 
368.6 

1.381 
344.3 


97.3 
100.6 
105.1 
101.8 

107.9 


Av. daily gain (lbs.) 

Feed for 100 lbs. gain (lbs.) 
Danish pig-testing sta- 
tions: 

Av. daily gain 

Feed per 100 lbs. gain . . 


104.0 
98.5 

101.5 
99.7 



* From the original publications of these experiments. 

If genetic intermediates in one or more primary functions produce maxi- 
mum performance, the increased total genetic standard deviation (V 1 + /) 
associated with inbreeding would tend to increase the average deviation from 
opti mum genotype and hence depress performance roughly in proportion to 
(vl + / — !)• Inbreeding alone would not alter mean level of performance 
without dominance, if only epistatic factors of the complementary or dupli- 
cate sort were involved. 

Inbreeding depression and crossbreeding advantage indicate some degree 
of dominance or of genetic intermediate optimum, but, alone, they fail to dis- 
tinguish between the two or to indicate the probable degree of dominance. 

EfFectiveness of Selection within Inbred Lines 

Selection within mildly inbred lines has been only slightly effective. De- 
cline in performance with mild inbreeding (2 to 4 per cent per generation) 
has been much the same as would have been expected from inbreeding with- 
out selection. These statements are based largely on a study' of time trends 

1. To be published in more detail, separately. 



INBRED LINES FOR HETEROSIS TESTS? 



333 



in litter size and growth rate in 49 inbred lines from five projects with an 
average of 9 seasons per line (see also Dickerson, 1951). In Figure 21.1 the 
average actual linear time trend (solid line) is negative for both litter size at 
weaning and for pig weight at 154 days of age. An estimate of the eflfective- 
ness of selection was made by adjusting the time trends for the effect of the 
increased inbreeding, using corrections derived from the intra-season com- 
parison of inbreds and linecrosses from the same inbred lines involved in the 
time trends. The adjusted time trend (dashed line) indicates that selection 





.75 










/ 




.50 


■ 










H 












/' 


3 












y 
















a, 












/ 


Q 


.25 


- 








/ 


H 










/ 




Z 








/■ 






< 








/ 






W 








/ 






:* 






/ 








CO 




/ 










O 


00 


-^^- 










0. 




~~~~" 






^^^^ 




.25 










^^^^^-^ 




TIME IN YEARS 

Fig. 21.1 — Linear time trend within mildly inbred strains for pigs weaned per litter and 
154-day weight per pig. Solid line is actual trend, dashed line is trend adjusted for effect 
of inbreeding trend to non-inbred basis, and the top broken line indicates mean superiority 

of selected parents. 



has failed to improve genetic merit for litter size and has allowed growth 
rate to decline, although selection of parents per year has averaged about 
.6 pigs for size of litter weaned by the dam and sixteen pounds for pig 
weight at 154 days (top broken line). 

These results must be accepted with caution, because time trends can be 
influenced by trends in nutrition, parasites, disease, management, or other 
factors. Also, the correction for inbreeding effects may have been underesti- 
mated. It seems clear that improvement has been at best only a small frac- 
tion of what would have been expected from the heritability of these traits 
and the amount of selection practiced for each. Evidence from comparison of 
intra -breed linecrosses with representative purebreds is meager but does not 
suggest any major improvement. Intra-herd comparisons of viability and 
growth rate of progeny from inbred and from representative purebred boars 



334 GORDON E. DICKERSON 

(Hazel et al., 1948) likewise have shown little advantage accruing from 
selection during development of the inbred lines. 

The apparent inability of selection to offset the decline in performance 
from mild inbreeding casts doubt on the assumption that epistasis or ordi- 
nary dominance (between none and complete) can account for the major 
influence of inbreeding on performance in swine. Unless one assumes a pre- 
ponderance of tight repulsion phase linkages, selection should have increased 
the frequency of favorable dominant genes. Similarly, under epistasis in 
which the genetic intermediate is optimum, selection should have prevented 
fixation of the more extreme homozygous combinations, particularly if a 
rather large number of loci determine the genetic range for each primary 
function. 

The type of genetic mechanism that would most surely produce an in- 
breeding decline relatively unresponsive to selection is heterozygote superior- 
ity (^ > 1). Here selection would maintain gene frequency near some inter- 
mediate equilibrium value, rather than move it toward fixation of any one 
allele (qa smaller). Linear regression of genotype on phenotype (heritability) 
would be lower than for lesser degrees of dominance, making selection rela- 
tively ineffective. Inbreeding depression for dominance, which is propor- 
tional to 2 5'a(1 — Qa) kf, would increase with k, particularly since qa would 
be smaller and Qa (1 — q) larger than under partial or complete dominance. 

"Controlled" Selection Experiments 

Results have been published from two ''controlled" experiments on selec- 
tion with minimum inbreeding in swine. In both the Illinois study of growth 
rate (Krider et al., 1946) and the Alabama study of feed efficiency (Dickerson 
and Grimes, 1947), the high and low selection lines separated appreciably 
and significantly. However, it is difficult to judge from the time trends 
whether the difference came partly from improvement in the high line or 
almost entirely from decline in the low line. Taken at face value, the time 
trends indicate that the separation was due to decline in growth rate of the 
low line in the Illinois experiment, but that efficiency increased in both lines 
in the Alabama study. 

In these experiments, the low line involved a reversal in the usual direction 
of selection. This amounted to assigning new selective values to genes affect- 
ing growth and feed utilization, and hence selection might be expected to be 
unusually effective for the first few generations in moving toward some new 
equilibrium. In both experiments, selection was most effective in the first 
generation. 

In Goodale's (1938) and in MacArthur's (1949) selection for size in mice, 
there is no question that a steady increase in size was produced. However, 
these experiments with adult size in mice are not directly analogous to those 



INBRED LINES FOR HETEROSIS TESTS? 335 

with prolificacy, viability, and growth rate in swine, for several reasons. First, 
the history of selection prior to the beginning of the experiment presumably 
had not been consistently positive for adult size in mice, as it was for pro- 
lificacy, viability, and rate of growth in swine. Second, selection for increased 
size of the organism may be quite different from selection for a further in- 
crease in efficiency within the same adult body size. Adult size is generally 
highly heritable but not consistently selected for in either direction in farm 
animals. The steady decline in effectiveness of selection without reduction in 
variability for size in MacArthur's study suggests approach to an equilibrium 
similar to that postulated for total performance in swine. 

Heritability Estimates 

Heritability, the portion of observed variance linearly associated with 
genotype, ranges from about 10 to 50 per cent for individual characters of 
economic importance. But heritability is found to be lower for the highly 
important characters such as prolificacy and viability, for which selection 
has been appreciable and always in one direction, than for traits such as 
carcass composition or external dimensions, for which selection has been mild 
or in opposite directions in different portions of the breed or during different 
periods of time. Heterozygote superiority is more likely to be important for 
genetic variability in the highly important characters, since selection would 
have had greater opportunity to fix those genes whose homozygotes were 
equal or superior to alternative genotypes at the same locus, leaving at 
intermediate frequencies (larger qA[i — Qa]) genes exhibiting heterozygote 
advantage. 

Ineffectiveness of selection for heritable traits suggests that degree of 
dominance may be higher and heritability lower for total performance than 
for its individual components. This has been shown for grain yield and its 
components in corn by Robinson el al. (1949) and by Leng el al. (1949). In 
swine, Cummings el al. (1947) found heritabilities of 22 per cent for size 
of litter at birth, 40 per cent for survival from birth to weaning, but only 7 
per cent for total litter weight at weaning. Heritability of total weaning 
weight jumped from 7 to 59 per cent when effects of size of litter at birth 
and of survival were held constant. These results could have arisen from 
negative genetic-physiological or from high positive environmental correla- 
tions, or both, between numbers per litter and weight per pig at weaning. 

Positive estimates of heritability for economic characters may be obtained, 
even though selection is ineffective due to heterozygote advantage. If ^ > 1 
and rates of reproduction were proportional to phenotypic levels, equilibrium 
frequency for the more favorable allele would be 9.4 = (1 + k)/2k. At this 
point, the linear regression of genotype on phenotype in an unselected popu- 
lation would be zero, and all intra-allelic genetic variability would be due to 



336 



GORDON E. DICKERSON 



dominance deviations (Fig. 21.2). Here both paternal |-sib correlation and 
regression of progeny on parent would yield zero estimates of heritability, if 
only dominance were involved. 

Equilibrium gene frequency actually will be determined by degree of 
dominance expressed in terms of relative selective values or reproductive 
rates {k') rather than in terms of relative performance levels {k) of the sev- 
eral genotypes. Conceivably, k' could be either larger or smaller than k. If 
culling is mild, difference in reproduction rates will be smaller between Aa 
and A A and larger between A A and aa than if proportional to phenotypic 
levels, and effective k' will be smaller and equilibirium qA larger. Conversely 
if phenotypic selection is intense, differences in reproduction rates between 

V^ = 2q (l-q)[(l+k)^ - 2kq (2+k-kq)] d^ 



2 ,2 



V^= 2q(l-q)[uk (l-2q)] ^ d 



AA-aa 




.1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 
FREQUENCY OF MORE FAVORABLE ALLELE (q^) 



Fig. 21.2 — Total variance in phenotype (V„) and portion linearly associated with genotype 
(Vq) in a random breeding population for a single chromosomal unit and heterozygote ad- 
vantage of jfe = 2, at varying frequencies for the more favorable of two alleles. 



INBRED LINES FOR HETEROSIS TESTS? 337 

AasLndAA will be magnified and those between. 1.1 and aa minimized, mak- 
ing k' larger than k and equilibrium (ja smaller than (1 + k)/2k. The larger 
the number of genes controlling genetic variation in the basis of selection, 
the less difference intensity of culling will introduce between k' and k. 

Estimates of heritability from regression of offspring on parent will in- 
crease positively as equilibrium q becomes larger than (1 + k)/2k, and as- 
sume larger negative value as q becomes smaller than (1 + k)/2k. Larger 
positive heritability estimates based on paternal |-sib correlation will be 
obtained as q becomes either larger or smaller than (1 + k)/2k, since this 
method estimates fraction of the phenotypic variance linearly associated 
with genotype regardless of the sign of the regression of ofTs})ring on parent 
(Fig. 21.2). 

It seems clear that positive estimates for heritability of individual char- 
acters do not rule out the possibilities (1) that heterozygote advantage ob- 
tains, especially for net selective advantage or total performance; and (2) 
that effectiveness of selection may be only a small fraction of that indicated 
by the estimates of heritability for individual characters. More attention 
needs to be given estimates of heritability for total performance indices and 
their components. 

Negative Genetic Correlations between Components 
of Total Performance 

Existence of negative genetic correlations would correspond to hetero- 
zygote superiority. This is in the sense that an increase in frequency of 
genes with partially or completely dominant favorable effects on one char- 
acter would amount to a decrease in frequency of their alleles having partial- 
ly or completely dominant favorable effects on one or more other characters. 
This involves the reasonable assumptions that genes have manifold end 
effects and that selection maintains at intermediate frequencies where 
contribution to genetic variability is larger — only those genes having domi- 
nant favorable but recessive unfavorable effects on performance. Mac- 
Arthur's (1949) experiment provided ample evidence that selection for a 
single character (adult size) produces many important changes in other 
characters. 

Direct evidence for negative genetic correlations is not plentiful. Much 
data must be analyzed to estimate genetic correlation with precision, par- 
ticularly when the traits correlated are of low heritability. Also, it is difficult 
to avoid bias from environmental correlations. If leaner hog carcasses are 
considered desirable, the genetic correlations of .3, .7, and —.7 for ratio of 
fat to lean cuts with 180-day weight, daily gain, and feed requirements per 
unit of weight gain found in a study of Iowa Record of Performance data 
(Dickerson, 1947) need to be considered. In the same and in another study 
(Dickerson and Grimes, 1947) evidence for genetic antagonism between 



338 GORDON E. DICKERSON 

good milking ability and rapid, economical fattening in swine is presented. 

Other reasons for expecting negative genetic correlations that might be 
mentioned are (1) lower heritability for total performance than for its com- 
ponents, as outlined previously, and (2) approach to some physiological 
maximum, where increase in one function must necessarily reduce others, as 
in division of nutrient energy available between milk production and fleshing. 

Negative genetic correlation, in some degree, is maintained by the 
process of selection itself and would disappear if selection were relaxed. 
Animals mediocre in any one respect are retained as breeders only if superior 
in several other characters. Thus selection leads to a negative correlation be- 
tween characters among the animals selected as parents. To a much lesser 
degree, these negative relationships would appear among the progeny, where 
fresh selection would magnify them again. Such negative character relation- 
ships may explain in part the discrepancy between rates of improvement 
"expected" and obtained, and could exist quite apart from any real heterozy- 
gote advantage. 

Analogy between Results with Corn and with Swine 

In both corn and swine, (1) inbreeding has been slight during domesticated 
history, until recently at least, (2) degree of heterozygosity exerts a major 
influence on performance, (3) effectiveness of continued phenotypic selection 
is questionable in stocks with a long history of selection for the same complex 
of characters in which further improvement is sought. 

Hull (1945) has postulated overdominance or heterozygote superiority, 
with additive interaction of non-alleles, to explain corn breeding results. He 
does so on the basis that (1) yields of parental, Fi, Fa, and backcross popula- 
tions are linearly related to proportion of loci heterozygous (Neal, 1935), (2) 
yields are usually less than one half as large for homozygous lines as for 
their Fi crosses, (3) regression of Fi yield on parental inbred yield among Fi 
crosses having one parent in common often is zero or negative for the higher 
yield levels of the common parent. Robinson et al. (1949) have obtained esti- 
mates indicating heterozygote advantage {k = 1.64) for grain yield but only 
partial to complete dominance for components of yield. Crow (1948) has 
shown that under complete dominance {k — 1 ) of favorable genes combining 
additively, average superiority of maximum hybrid over the variety at 
equilibrium gene frequency would be the product of mutation rate and 
number of loci, or less than 5 per cent, whereas potential hybrid advantage 
under some degree of heterozygote advantage (^ > 1) at even a small pro- 
portion of loci could be many times greater, in agreement with results al- 
ready obtained 

The impossibility of accounting for the 15 to 25 per cent advantage of 
better corn hybrids over open-pollinated varieties through complete domi- 
nance of favorable genes combining additively can be demonstrated (Dicker- 



\f^ 



PS 

< 



o 

< 

H 

< 
> 

Q 
< 

Q 

K 

e 



u 

O 

a 



Id. 



luT 



o 

Eh 




.9 .8 .7 .6 
FREQUENCY OF MORE FAVORABLE ALLELE (q) 



.5 



Fig. 21.3 — Potential hybrid advantage per locus {Aa — Fi) as a proportion of observable 

inbreeding decline {f^ — p), for varying degrees of dominance and frequencies of the more 

favorable allele in a population of homozygous (/ = 1) lines and their Fi crosses. 



340 GORDON E. DICKERSON 

son, 1949) from the average ratio per locus of maximum potential superiority 
of hybrid over average of all possible Fi's to the observable mean advantage 
of the Fi's over the inbred lines themselves (Fig. 21.3). Using (1 — q) for 
frequency of the less favored allele, /for Wright's inbreeding coefficient, k for 
degree of dominance, as before, and d = {A A — aa)/2: 

H = Maximum heterozygote =C -\- {l -\- k) d 

Fi = Mean of Fi crosses = C-{-2q[\ + k{l-q)]d 

P = Mean of inbred lines = C -\-2q [\ -j- k (1 - q) {I - f) ] d 

(H-Fi) {k-l)[l-2q(l-q)]-h2{l-q)^~ 



Hence , 



Fi-P 2kjq{\-q) 



Underpartialor complete dominance, equilibrium (1 — 9) = {k —\)/2k = 
0, except for reverse mutation pressure. When the parental lines are homozy- 
gous (/ = 1), mean (1 — q) lies between .05 and .1 and mean inbred yield is 
about 40 per cent of that for Fi crosses, the maximum increase of hybrid over 
average Fi would lie between 3 and 7 per cent. There is little reason to sup- 
pose that present better hybrids approach the maximum. The potential maxi- 
mum increase over open-pollinated varieties increases rapidly with degree of 
heterozygote advantage {k), approaching {Fi — P) or about 50 per cent in 
corn yield. 

The evidence thus far obtained in swine suggests that the genetic basis of 
variation in net productivity is fundamentally quite similar to that in corn. 
This indicates challenging possibilities for increasing productivity of swine 
by utilizing potential heterosis. 

Heterozygote Advantage for Single Loci and Chromosome 
Segments in Other Species 

Dobzhansky (1949) has shown experimentally that natural selection 
favors individuals heterozygous for inversion chromosome segments in 
Drosophila. He also has shown that the resulting equilibrium between fre- 
quency of alternative homologous segments fluctuates with locality and sea- 
son of the year, depending on relative selective advantage of alternative 
"homozygous" segments. He postulates natural selection for increased co- 
adaptation between alternative segments in heterozygotes within each inter- 
breeding population. Demonstration of heterozygote advantage at individual 
loci would be difficult, since any one locus usually would account for only a 
small part of the total variability in selective value or in any complex charac- 
ter. However, some cases of presumably single gene mutations exhibiting 
heterozygote advantage have been reported (Jones, 1945; Gustafsson, 1946, 
1947). The "yellow" gene of the agouti series in mice (Danforth, 1927; 
Dickerson and Gowen, 1947) provides a classic example of manifold effects 



INBRED LINES FOR HETEROSIS TESTS? 341 

of genes and of heterozygote su{>eri()rity in food utilization, if not in selective 
value. 

It seems inevitable that manifold effects of genes and equilibrium between 
frequencies of alternative alleles are common{)lace, with relative selective 
values shifting with the characters given emphasis in selection at each stage 
of development from conception through maturity. 

EFFECTIVENESS OF METHODS OF SELECTING 
FOR MAXIMUM HETEROSIS 

The evidence presented provides several related assumptions concerning 
the nature of genetic variability in economic characters of swine as the basis 
for considering how selection for maximum heterosis can be made most effec- 
tive. These are: (1) Heterozygote advantage (^ > 1) is important for total 
performance when its components are characters that have had consistently 
positive selective values, although lesser degrees of dominance may obtain 
for individual characters. (2) Average gene frequency approaches an inter- 
mediate equilibrium near q.i = (1 + k)/2k, whose value and stability depend 
on the intensity, consistency, and duration of selection. (3) Performance 
levels attainable by selection in outbred populations are far below the maxi- 
mum heterozygote, because more than one-half of the individuals are homo- 
zygous at each locus. (4) Inbreeding decline may be considered as due largely 
to the reduced number of genes useful to the species that can be carried by 
the more homozygous individuals, rather than to fixation of unfavorable re- 
cessive genes. 

Under these assumptions, any method of selecting for maximum perform- 
ance will involve (1) selection for maximum proportion of heterozygous 
loci in crosses of complementary strains, and (2) selection based on progeny 
tests of individuals or lines in crosses. These methods are indicated only 
when individual and family selection become relatively ineffective, because 
the intensity of selection per unit of time is much lower for selection based on 
test-cross progeny performance. 

Importance of Recurrent Selection to Achieve Maximum Heterosis 

Hull (1945) has emphasized the great importance of utilizing cumulative 
gains from recurrent selection for heterosis in crosses, rather than relying on a 
single selection among Fi crosses of a group of homozygous lines. This prin- 
ciple may be illustrated by contrasting the observed distribution for number 
of heterozygous loci in a population of Fi crosses among inbred lines with the 
potential range (Fig. 21.4). It can be shown that the standard deviation 
in proportion of heterozygous loci is: 



-4 



\2q{\-q){\-J)[\-2q{\-q ){\-J) ] 



342 GORDON E. DICKERSON 

within linecrosses, and 



<THb 



n 



among linecrosses, where/ is the inbreeding of the population of lines and n 
is the efifective number of segregating chromosomal units. 

Range in degree of heterozygosity among all Fi crosses of a population of 
lines for likely values of n {n = 100 and/= 1 in Fig. 21.4) is small rela- 




.3 .4 .5 .6 .7 .8 .9 1.0 

PROPORTION OF LOCI HETEROZYCX)US 



Fig. 21.4 — Frequency distribution for proportion of 100 loci heterozygous when k = 2 and 
initial q = .75 within lines 50 per cent inbred (.4), between Fi crosses of homozygous lines 
or in a non-inbred population (B), between Fi crosses of hnes inbred only 50 per cent (C), 
and in a cross between complementary strains (^i = .95, go = .15) attainable only through 
recurrent selection for cross performance {D). 

tive to the potential range. Hence recovery of Fi crosses much above the 
average for all Fi's or for non-inbred stock cannot be expected. Inbreeding 
provides a means for steadily reducing the proportion of heterozygous loci. 
What is needed is recurrent selection in complementary strains to make them 
steadily approach opposite extremes in gene frequency at each locus exhibit- 
ing heterozygote advantage. The best Fi of a population of 100 Fi crosses 
would average about 2.6 am above the mean, whereas the best 1 of 10 
would average about 1.54 am above the mean and cumulative selection of 
the best 1 of 10 in each of 10 recurrent cycles of selection would amount to 
choosing Fi's that were 10 (1.54) am = 15.4 am above the original mean. 

Homozygous Tester versus Reciprocal Selection 
Hull (1945) has proposed recurrent cycles of selection in crossbred material 
based on progeny test in crosses with a single homozygous line (alternatively, 
with two related lines or their Fi) as a method of producing highly comple- 



INBRED LINES FOR HETEROSIS TESTS? 343 

mentary lines to be used in production of commercial hybrids. Comstock 
et al. (1949) have compared expected effectiveness of Hull's plan with that 
for reciprocal selection for cross performance between two foundation stocks 
of divergent origin, avoiding inbreeding in both stocks. The}' point out that 
the potential limits of improvement are the same for the two methods, except 
for loci exhibiting only partial dominance {k < 1), where the use of a tester 
homozygous for any of the less favorable alleles would reduce potential hy- 
brid performance. Existence of important epistatic effects also would tend to 
make limits lower for use of a homozygous tester. 

There is no reason to expect initial cross performance to differ between 
reciprocal and homozygous tester selection, other than because of the per- 
formance of the inbred tester line itself (inbreeding effects on maternal en- 
vironment of the litter in swine). If anything, it would be easier to find a 
population differing materially in gene frequency at individual loci from a 
homozygous tester than to find two similarly complementary non-inbred 
populations. 

Relative rates of improvement expected from the two plans depend on (1) 
selection pressure applied, and on (2) size of regression of gene frequency in 
the material under selection on performance of test-cross progeny. Hull's 
homozygous tester plan limits selection to only one of the parental stocks. 
Hence selection applied will be only | as great as in reciprocal selection. 
However, as long as frequencies of the more favorable alleles {q) are any- 
where near their expected equilibrium of (1 + k)/2k, progress toward com- 
plementary gene frequencies (toward maximum proportion of heterozygous 
loci in the cross) per cycle of selection will be far less for reciprocal than for 
homozygous tester selection. 

Comstock (1949) has shown that improvement in performance of random 
crosses between two segregating populations per generation of selection, at a 
given locus, is: 

AP.= [^qA^ + k-2kq.i) -{-Aq2{l + k-2kqi) - Ik- Lq,- Aq^] d (1) 

The change in gene frequency at a given locus within each of the two selected 
populations (A^i and Ag2, respectively) will be determined by (1) the in- 
tensity of selection based on the test-cross progeny means {s = selection dif- 
ferential in a units), (2) the correlation between qi and the mean progeny 
performance (P), and (3) the size of aq^ among the tested individuals, as 
follows: 

Avi = s r,,,.a„ = , Cov,...^ 5^^ q,{\-q,){\ + k -2kq.^ (2) 

0" P 2<T p 

Hence, as Comstock indicates, improvement in cross performance from 
one cycle of reciprocal selection and for any one locus is: 

APr = ^[qi{\-q,)(l + k-2kq,)^-+q2i\-q2)il + k-2kq,)^ ^^^ 

— 2k'Aqi-\q-2 



344 



GORDON E. DICKERSON 



However, equation (3) is not applicable for evaluating Hull's (1945) plan 
of recurrent selection for cross performance with a homozygous tester line. 
Here, cross performance will improve as qi -^ for loci at which the more 
favorable homozygote (A A) is fixed in the tester, and as Qj -^ 1 for loci at 
which the tester is homozygous for the less favorable allele (aa). li qr repre- 

TABLE 21.2 

MEANS, VARIANCES, AND COVARIANCES FOR GENOT\TES OF 

SELECTED POPULATION AND PHENOTYPES OF TEST-CROSS 

PROGENIES FROM HOMOZYGOUS TESTER 



Selected Population 


Mean Phenotypes of Progenies 
FROM Homozygous Tester 


Genotypes (qs) 


AA at gj. of 
Loci (Gi) 


aa at (1— gj-) of 
Loci (Gy) 


AA 


Means 

Dev. 

Freq. 


1.0 
l-qs 


2d 

i\-q,)(l-k)d 

qhr 


(l-f^M 
(l-9,)(l + ^)rf 

q^i^-qr) 


Aa 


Means 

Dev. 

Freq. 


.5 

.5-9. 
29.(1-9.) 


{?>^k)d 

{\-qi){\-k)d 
2q^{\-qi)qT 


{^^k)d 

{\-q,){\^k)d 
29,(1 -9,) (l-9r) 


aa 


Means 

Dev. 

Freq. 




-qs 

(1-9.)^ 


-q,{\-k)d 
{l-q.)'qT 




-q^'^^k)d 
(.^-q^m-qr) 


Means 


9. 


[l + k+qiil-k)]d 


9,(1 + *M 


Variances 


2 


qi{\-qi){\-k)H^ 
2 


9;(1 -9,) (1 + ^)^2 

2 


Covariances (9,, • G) 


qi{\-qi){\-k)d 
2 


9,(l-9y)(l-H^)^ 
2 



sents the proportion of loci segregating in the stock under selection that are 
homozygous A A in the tester, then it can be shown (Table 21.2) that aver- 
age progress in cross performance per locus is: 

AP,= [qT•^qA'^-k) + {\- qr) '^qJ{\ + k)]d (4) 

For loci fixed .4.4 in the tester: 

A^, = s- rpo^' >'G,„,-(rgi = —rr-—'q^ (I- q,)(l- k) (5) 

^ (T p 

Similarly for loci fixed aa in the tester: 

sd 



Aqj 



S ' r p(; • y a ,, ' a 



., ^,, . Vil-qr) 'qj{\-qj)il-\-k) (6) 

Z, (7 p 



INBRED LINES FOR HETEROSIS TESTS? 345 

From equation (4) we can now express average progress per locus from selec- 
tion for cross performance with a homozygous tester as: 

M\ = [ qfq.(i-q.) ( 1 - ^)-+ ( 1 - qr^Qj ( 1 - 9.) d + ^) ^1 ^ (7) 

z a j> 

Rate of improvement in cross performance from reciprocal selection (equa- 
tion 3) approaches zero as gene frequencies approach the equilibrium ex- 
pected if rates of reproduction of individuals were directly proportional to 
their phenotypes (i.e., 9 = [1 + k]/2k). Hence, progress from reciprocal selec- 
tion may be expressed more usefully in terms of the deviation of gene fre- 
quencies from (1 + k)/2k, as follows: 



^''' = £['■('-'■> Gr - "')' + ^= ( ' - '') (Vr - *'T 



4 k-' 



— 2k 'Aqi' Aqod (8) 

Comparisons of expected progress per generation from homozygous tester 
and from reciprocal selection may be made from equations (7) and (8), re- 
spectively. The comparison may be visualized by plotting rate of improve- 
ment against deviations of gene frequencies from an initial equilibrium value 
of (1 + k)/2k, using qi and q-i for the two populations under reciprocal selec- 
tion, and qi and qj for loci that are AA and aa, respectively, in the tester, for 
homozygous tester selection. 

In Figure 21.5, it is assumed that k = 2, and qi is shown approaching 
{k -\- \):ik — 1) times as fast as q, approaches 1. Actually, qt would 
move more slowly than Qj at first because {Aa — A A) = {k — \)d and 
(.4 a — aa) = (1 + k)d. However, A^,- increases as qi falls from .75 toward 
.5 because of the increased variance of qi and consequent increase in genetic 
variance and in covariance with progeny means, and then Aqi declines as 
qi moves from .5 toward 0. There is a steady decline in Aq, as qj rises from 
.75 toward 1.0. 

Under reciprocal selection, if q\ and q^ are near an equilibrium of (1 + k)/2k 
at the outset, initial progress will be slight compared with that from homozy- 
gous tester selection and will not equal APh until qi and q^ differ, in opposite 
directions from (1 -\- k)/2k, by an average of about .50. Only during the late 
generations of selection will reciprocal selection surpass homozygous tester 
selection in effectiveness. 

Another possible disadvantage of reciprocal selection is that gene fre- 
quencies at most loci for which k> \ may be somewhat below (1 -\-k)/2k. 
This will occur if the advantage oi Aa over .4.4 and aa individuals in rate of 
reproduction is made greater by intensive individual or family selection than 
it would be if reproductive rates were directly proportional to phenotypic 
levels of performance. This would amount to increasing the effective degree 



346 



GORDON E. DICKERSON 



of heterozygote advantage from k to k' , and hence making actual equilibrium 
q nearer to .5 (i.e., ^ = [1 + k']/2k'). 

If actual equilibrium frequencies for the more favorable allele are generally 
below (1 + k)/2k in both populations, reciprocal selection will tend to raise 
both qi and 92 toward (1 + k)/2k, but at an ever decreasing rate, until q 
chances to go beyond (1 + k)/2k in one of the populations. However, q\ and 
92 are unlikely to be equal, even when both are smaller than (1 + k)/2k. If 
qi > q-i, then qi will be closer than qi to (1 + k)/2k and will move faster in 
that direction (A^i > d^q-y). Consequently gi will become larger than (1 + k)/ 
2k and direction of Aq2 will be reversed without reducing A^i to zero. Only 




Qj and qj = .75 
q. and q, = .75 

Fig. 21.5 — Mean performance of crosses (upper) and rates of progress per cycle (lower) for 

homozygous tester (solid lines) and for reciprocal (broken lines) selection when k = 2 and 

q = .75, as qi and ^i approach and qj and ^2 approach 1. 



INBRED LINES FOR HETEROSIS TESTS? 347 

then can the slow-starting reciprocal selection begin moving qi and 52 toward 
opposite extremes. 

Use of Partially Inbred Tester Lines 

In large animals, or even in poultry, discussion of selection utilizing homo- 
zygous tester lines is still largely academic. Few very highly inbred and 
usable lines of swine and chickens exist. However, there are many partially 
inbred lines of swine and poultry whose average cross performance has been 
or is now being tested. These partially inbred lines should be extremely use- 
ful in overcoming the initial disadvantage of reciprocal selection, because in- 
breeding will have pushed frequencies of individual genes in these lines much 
further away from equilibrium than in non-inbred stocks. Since effectiveness 
of reciprocal selection (equation 8) increases with 






even a moderately inbred line used as one of two populations under reciprocal 
selection would materially increase initial progress per cycle from Aqi. 

Of course, further selection within the inbred line itself on the basis of 
cross performance would be relatively ineffective (A92 small) until the selec- 
tion on cross performance has had time to shift qi at individual loci in the 
non-inbred population away from (1 + k)/2k in the opposite direction from qi. 
It might be wise to ignore cross performance in selecting replacements 
within the inbred line for a number of cycles to allow time for qi to make this 
shift at loci where initial ^'i and 92 chance to deviate from (1 + k)/2k in the 
same direction. Beyond this point, progress from reciprocal selection between 
the partially inbred and the non-inbred populations should approach and 
finally exceed that from selection for cross performance with a homozygous 
tester. 

In selecting a partially inbred line for use in reciprocal selection, one in- 
stinctively would choose a line known to be superior in its average cross per- 
formance and in its usability as an inbred strain. This seems desirable to 
assure that the line carries at high frequencies any genes whose favorable 
effects on total performance are incompletely dominant. In addition, it 
would be helpful to try a number of different partially inbred lines in crosses 
with a given non-inbred stock, choosing finally for reciprocal selection the 
line showing best initial cross performance. Diversity of origin and previous 
crossing data would of course aid in selecting the lines more likely to be 
initially complementary to a given non-inbred stock. 

Presumably, both initial cross performance and rate of progress from 
reciprocal selection are likely to be greater if the two populations are of dif- 
ferent breeds. However, Dobzhansky's (1949) finding of greater heterozygote 
advantage from alternative homologous chromosome segments within a 
single population than in crosses between non-interbreeding populations of 



348 GORDON E. DICKERSON 

Drosophila suggests the need for further investigation of the importance of 
diversity of origin for attainment of maximum heterosis in crosses. 

Use of an Fi Cross as the Tester 

Hull (1945) also has suggested selection to complement the Fi cross of two 
homozygous lines as a means for developing new lines to replace the poorer 
ones presently used in successful corn hybrids. Here, expected rate of im- 
provement in performance of the 3-line cross (AP/) would be a composite 
of that expected from selection for cross performance with a homozygous 
tester, and with a non-inbred strain in which gene frequency is ^ at each 
locus: 

AP/= [q\'^qd\-k)-\-{\-qT)''^qM + k) + 2qA\-qT)'^qf\d (9) 

where qr is average proportion of loci homozygous for the more favorable 
allele in the lines represented in the Fi tester; qi, qj, and q/ are average fre- 
quencies of the more favorable alleles at loci that are A A, aa, and Aa, 
respectively, in the Fi tester. The Fi tester is A A at qr, aa at (1 — 5^)^ and 
Aa at 2^1.(1 — qr) of the loci. Hence, 

Aqi = ^ qrqii^ - 9,)(1 -k) , A9, = :r— (1 - 9r)9;(l - qj)(l-\-k) 

and 

sd 



A?/ = 7i V2gr(l - 9r) -9/(1 - (//) . 

Substituting in equation (9), we obtain as estimated progress per cycle: 

Ai'/= \q%qS\-q){\-ky-^{\-q^yq^{\-q){\-^ky~ 

sd^- (10) 
+ [2q^{\-q^)]^/'^-qf{\-q;)]j— 

^ (J p 

Apparently one might expect that selection to complement an Fi tester (of 2 
homozygous lines) would be about one-half as effective as selection to fit a 
single homozygous tester. 

In selection for complementary strains in livestock, the Fi tester may be a 
cross of two partially inbred lines, M and X. Selection of a population, L, to 
complement M-N would tend to improve the L{M'N) cross at a rate inter- 
mediate between one-half that for reciprocal selection (APr)/2 and that for 
use of an Fi cross of homozygous lines as the tester (AP/), depending on the 
degree of inbreeding in lines M and N . 

If M and .Y were being selected to complement each other, gene frequency 
in the {M'N) linecross tester would tend to be lowered from equilibrium 
(1 -f k)/2k toward ^ as the limit. Consequently, rate of improvement in the 
L{M-N) cross from selecting L to fit Af •A'^ should approach that expected 
from selecting in population L to complement a non-inbred tester in which 



INBRED LINES FOR HETEROSIS TESTS? 349 

^ = .5 at each segregating locus. Progress per cycle when q^ = .5 should 
approach 

qiii - q\) '^ — . 

lap 

Since qi would be increased above initial equilibrium of (1 + k)/2k, maxi- 
mum progress per cycle should be 

{\-]-k){k-\) sd-' 



W lap' 

and rate of progress would decline as ^i became larger than (1 -(- k)/2k. The 
maximum rate of progress, then, for selecting population L to complement 
the cross of two highly complementary strains M and iV, is expected to be 
little more than one-half that for selection to complement a homozygous 
tester. 

Other Considerations 

Under heterozygote advantage and selection toward complimentary 
strains by either the reciprocal or the homozogous tester method, the strains 
themselves may be expected to decline in performance for characters that are 
depressed by inbreeding. The less favorable allele would tend to become 
fixed at about half of the loci segregating in the foundation stocks. The effec- 
tiveness of this sort of selection in moving gene frequencies toward opposite 
homozygous extremes in the complementary strains would be greater for 
those traits in which heterozygote advantage (^ > 1), and hence inbreeding 
depression, is larger. That portion of the inbreeding depression arising from 
loci at which there is no heterozygote advantage {k < 1) would not be pro- 
duced by selection for cross performance without inbreeding, because selec- 
tion would favor the dominant allele in both strains. Therefore, any serious 
decline in performance of the strains themselves, while under selection for 
cross performance, is indicative of heterozygote advantage and should be 
accompanied by compensatory improvement in performance of the cross. 

In order to develop complementary strains whose own performance would 
make them usable in commercial production of crosses, some compromise 
may be necessary between selection based on test-cross and on individual 
performance. There is much opportunity for selection in choosing young 
breeders, especially males, to be tested in the strain-cross. Individual selec- 
tion for characters little affected by inbreeding would be least apt to impair 
the effectiveness of the complementary selection. Some selection for indi- 
vidual performance characters important for both the strains and their 
cross may become necessary to prevent fixation of rare genes with major 
detrimental effects in the homozygote, but advantageous in the heterozygote. 
Selection for fertility and maternal influences (e.g, hatchability, prolificacy, 
or suckling ability) in test-cross matings should help maintain usable strains. 



350 GORDON E. DICKERSON 

SUMMARY 
Genetic Variability in Economic Characters of Swine 

1. Inbreeding and crossbreeding effects indicate that degree of hetero- 
zygosity exerts a major influence on the important performance characters, 
and that a high degree either of dominance or of epistacy due to deviations 
from an optimum genetic intermediate, or both, characterizes genetic vari- 
ability in performance. 

2. Relative ineffectiveness of selection within mildly inbred strains makes 
ordinary dominance or epistasis doubtful as an explanation of inbreeding de- 
cline, and suggests heterozygote advantage for net desirability in prolificacy, 
suckling ability, viability, and growth rate. 

3. "Controlled" selection experiments with swine show that high and low 
lines for growth rate or feed utilization can be separated, but indicate little 
improvement of high line over foundation stock, particularly for net per- 
formance in all characters. 

4. Lower heritabilities and larger inbreeding declines for characters long 
and intensely selected in one direction, compared with those selected toward 
an intermediate or in varying directions, indicate a higher degree of domi- 
nance for the former. 

5. Some sort of negative relationship between components of total per- 
formance is indicated by lower heritability for total performance than for its 
component characters and by direct estimates of correlation. This would 
correspond to heterozygote superiority, in that increased frequency of genes 
with dominant favorable effects on one character would constitute decreased 
frequency of their alleles having dominant favorable effects on other char- 
acters. 

6. The genetic basis of performance appears to be similar in corn and in 
swine, as indicated by natural degree of inbreeding, extent of inbreeding 
decline in performance, and the effectiveness of phenotypic selection. Ordi- 
nary dominance is inadequate to account for heterosis already achieved in 
corn, and by analogy, is unlikely to be adequate in swine. 

7. Examples of manifold effects and heterozygote advantage for specific 
chromosome segments or loci support their inferred importance for quantita- 
tive economic characters. 

Methods of Selecting for Maximum Heterosis 

1. Intensity of selection per unit of time is lower when based on progeny 
performance in test-crosses than when based on individual and family per- 
formance. Hence, methods of selecting for maximum cross performance be- 
tween complementary strains are indicated only when individual and family 
selection have become relatively ineffective, and when there is evidence for 



INBRED LINES FOR HETEROSIS TESTS? 351 

important heterozygote advantage with attendant intermediate equilibrium 
gene frequencies. 

2. Cumulative gains from recurrent selection pressure are necessary to 
obtain efficiently crosses heterozygous for anywhere near the potential maxi- 
mum proportion of loci, since distribution of Fi crosses within any generation 
is narrow relative to the potential range when numbers of loci are large. 

3. Expected effectiveness of reciprocal recurrent selection between two 
populations and recurrent selection for cross performance with a homozygous 
tester may be compared as follows: 

a. They are alike in potential limits of cross performance for loci exhibiting 
heterozygote advantage, but use of a homozygous tester would be more 
likely to limit ultimate cross performance if partial dominance or special 
epistatic effects were important. 

b. They would be similar in initial cross performance, except that it should 
be easier to deliberately select a stock differing materially from a homo- 
zygous tester in gene frequency at individual loci than to select two 
equally complementary non-inbred stocks. 

c. As long as gene frequencies in the selected populations are anywhere near 
their expected equilibria, improvement in cross performance per cycle 
will be far greater for the homozygous tester than for the reciprocal selec- 
tion plan. The difference between progenies from A and a gametes under 
selection approaches zero as frequency of A in the non-inbreed tester 
approaches an equilibrium of (Aa — aa)/(2 Aa — AA—aa), but discrimi- 
nation between A and a gametes under selection is maximum when the 
tester is homozygous aa or A A. 

d. Rate of progress from reciprocal selection accelerates as the difference in 
frequency of homologous chromosomal units in the two populations be- 
comes larger, and surpasses homozygous tester selection when qi — q^ 
exceeds about .5. 

4. Use of a partially inbred line as one of the two populations in reciprocal 
selection would greatly increase progress in early cycles, since individual 
gene frequencies will be further away from equilibrium in inbred strains than 
in non-inbred stocks. 



C. R. HENDERSON 

Cornell Universify 



Chapter 22 



Specific and General Combining Ability 



By general combining ability we mean the average merit with respect to 
some trait or weighted combination of traits of an indefinitely large number 
of progeny of an individual or line when mated with a random sample from 
some specified population. The merit of the progeny is measured in some 
specified set of environmental circumstances. If maternal effects are present, 
we must specify that the tested individuals are males. If the tested individu- 
als are females, the merit of the progeny is a function of both general com- 
bining ability and maternal ability. 

General combining ability has no meaning unless its value is considered in 
relationship to at least one other individual or line and unless the tester 
population and the environment are specified. For example, suppose two 
dairy bulls used concurrently in an artificial breeding ring each have 500 
tested daughters, and that it can be assumed that the cows to which the two 
bulls were mated were a random sample of cows from herds using artificial 
breeding. Suppose that the mean of the butterfat records of the daughters 
of the first sire is 410 pounds and of the second sire is 400 pounds. Five hun- 
dred tested daughters are sufficient to reduce the sampling variance of the 
progeny mean to a negligible amount. Consequently the general combining 
ability of the first sire is 410 — 400 = 10 pounds better than that of the sec- 
ond in this particular population and in this set of environmental circum- 
stances. The general combining abilities of the two sires might differ by more 
or by less than 10 pounds if they were used in some other region where both 
the genotypes of the cows to which they were mated as well as the environ- 
ment could be quite different from those of the test. 

SPECIFIC COMBINING ABILITY 

We shall define specific combining ability as the deviation of the average 
of an indefinitely large number of progeny of two individuals or lines from 

352 



SPECIFIC AND GENERAL COMBINING ABILITY 353 

the values which would be expected on the basis of the known general com- 
bining abilities of these two lines or individuals and the maternal ability 
of the female parent. As Lush (1948) has pointed out, apparent specific 
effects, or what animal breeders usually call nicking, also can be a conse- 
quence of Mendelian sampling, of inaccurate estimates of the additive genetic 
values of the two parents, and of environments affecting the progeny which 
are different from the average environments in which the general combining 
abilities and the maternal abilities were estimated. 

Genetically, specific combining ability is a consequence of intra-allelic gene 
interaction (dominance) and inter-allelic gene interaction (epistasis). We shall 
assume in this paper that we can estimate only the joint effect of dominance 
and epistasis. As an illustration of specific combining ability let us suppose 
that we know that the general combining ability with respect to weight in 
swine line A is +10 pounds at 154 days, and that the general combining 
ability plus maternal ability of line B is -|-5 pounds at 154 days. Then if an 
indefinitely large number of progeny of the cross A X B has a mean of +7 
pounds, the specific effect for this cross is 7 — 10 — 5 = —8. 

SELECTION FOR GENERAL AND SPECIFIC 
COMBINING ABILITY 

Under some circumstances selection would be largely for general com- 
bining ability, and in other circumstances for a combination of general and 
specific combining ability. For example, those selecting sires for use in a 
large artificial breeding ring are interested primarily in obtaining sires with 
the highest general combining ability with respect to the population of cows 
and environments in which the bulls are to be used. On the other hand, those 
wishing to employ crosses among inbred lines for commercial use select for a 
combination of general, maternal, and specific effects. 

Now let us consider some of the problems involved in selecting for general 
and specific combining ability. There are reasonably good solutions to some 
of these problems, but almost none for others. Some of the questions which 
are involved are : 

1. Given a particular set of records how can one best estimate the general 
combining abilities of individuals, families, or lines, and how can one best 
estimate the value of the progeny of a specific cross between families or in- 
bred lines? 

2. What proportion of the breeder's resources should be put into a testing 
program? For example, if he is dealing with inbred lines, what proportion of 
his resources should be employed in the making of lines and what proportion 
in testing them for general and specific combining ability? 

3. Having decided on the size of the testing program, what kind of tests 
should be made? For example, should lines be tested in topcrosses or in line 
crosses or in some combination of these two procedures? Also what use should 



354 C. R. HENDERSON 

be made of a sequential type of testing in which some lines are discarded on 
the basis of a very preliminary and inaccurate test? 

4. What relative emphasis in selection should be placed on general as 
compared to specific combining ability? 

5. How much inbreeding should be done in the making of lines? How fast 
should the lines be made? 

Obviously a complete discussion of all these problems and their possible 
solutions in the time at our disposal is impossible. Consequently we shall 
discuss primarily the problem of estimating general combining abilities of 
lines and individuals and of estimating the values of specific crosses among 
lines, given a particular set of records. In addition, since estimates of the 
variances play an important role in these selection methods, we shall discuss 
briefly the problem of estimating variance components from the results of 
line-cross tests. 

So far as estimation of general combining abilities of individuals is con- 
cerned, the methods to be presented here are essentially those of the selection 
index. It will be shown that no assumption of normality of distributions is 
required; that joint estimates of general combining abilities and certain 
parameters such as the population means, the yearly effect, the age and in- 
breeding effect, can be obtained; and that certain short-cut computational 
procedures are sometimes distinctly advantageous. An application of the 
principles of the selection index to estimation of general combining abilities 
of lines or families also will be presented. Finally it will be shown that appli- 
cation of the selection index need not be restricted as it has been to selection 
for additive effects, but can be applied equally well to joint selection for 
specific effects and general combining ability. The selection index approach 
to appraising crosses can, under some circumstances, be much more efficient 
than selection based on the mean of the progeny of a particular cross. 

ESTIMATION PROBLEMS IN SELECTION 
Before turning to selection for general and specific combining abilities let 
us consider the type of estimation problem which is involved and some gen- 
eral solutions to it. Later the manner in which the solutions can be applied 
to our present problem will be discussed. Our estimation problem can be 
stated in this way. We have a sample of N observations, yi, y2, • • • , ^a^, 
from which we wish to estimate ^i, 62, . . . , dg. The y's are assumed to have 
a multivariate distribution (precisely what distribution need not be specified 
for the present) with means, biXu-\- ^2-V2i + • • • + bpXpi, and variance- 
co variance matrix, 

Tiie b's are fixed parameters such as the population mean and the regres- 
sion of y on age of the dam, and x is an observable parameter, the first sub- 
script denoting with which b it is associated, the second subscript with 



SPECIFIC AND GENERAL COMBINING ABILITY 355 

which sample observation. As an illustration, .Vi might be associated with bi, 
the population mean. Then Xn would have the value 1 in each observation; 
Xii might denote the inbreeding coefficient of the dam. 

Now comes the really crucial part of the model. 'J'he ^'s are regarded as 
having some multivariate distribution with means zero and variance-co- 
variance matrix, 

Also the d's and ^''s are regarded as having a joint distribution with covari- 
ances aekyi- The way in which this problem differs from the ordinary estima- 
tion problem in statistics is that here we wish to estimate the values of indi- 
vidual ^'s which are regarded as a sample from some specified population. 

Selection for Additive Effects in the Normal Distribution 
What is the ''best" way to estimate the 0's? Suppose that they represent 
additive genetic values of individuals and that any linear function of the y's 
is normally distributed. Lush (1948) has shown that, subject to the normality 
assumptions, improvement in additive genetic merit of a population through 
selection by truncation of the estimates (indexes) of additive genetic values 
is maximized by choosing that index which has maximum correlation with 
additive genetic value. This principle has been used in the index method of 
selection by Fairfield Smith (1936), Hazel (1943), and others. These workers 
have shown that the index can be found in a straightforward manner pro- 
vided certain variances and covariances and all of the 6's, the fixed elements 
of the model, are known. 

The values of Kei which maximize ree where 6 — KeiWi + . . . + KbnWn 
are the solution to the set of simultaneous equations (1). The w's are the 
^''s corrected for the fixed elements of the model such as the population mean 
(not the sample mean). Thus Wi = yi — biXn — ■ ■ . —b,,Xpi. 



ei y^y-i ' B2 y.^ ' ' e^ y.,y\ J/.." 



(1) 



K„,(T + K„„a -|- . . . + K„..(T-^ = o-„ „ 

d\ y^yx ^' y^'Ux *-^ l^X yN'' 

Selection when Form of Distribution is Unspecified 
and b's Are Unknown 
Maximization of ree is a satisfactory solution to the problem of selection 
for additive genetic values under the normality assumption and the as- 
sumption of known b's,. Is a comparable solution available when nothing is 
known of the distribution or of the b's? So far as I am aware there is not. 



356 C. R. HENDERSON 

Consequently let us consider some other criterion of a "best" index. We shall 
use as our criterion of "best" that index from the class of linear functions of 
the sample which is unbiased (coefBcients of all b^s, = in Ed) and for which 
E{d— Oy is a minimum. E denotes expected value. Consequently E{d — dY 
denotes the average in repeated sampling of the squared deviations of the 
index of 6 about the true value of d. When the i's are unknown, the same 
criterion of best is applied to them, that is, minimum E{b — bY for unbiased 
estimates {Eb = b) which are linear functions of the sample. It turns out 
that minimization of E{d — Oy and maximization of ree lead to identical in- 
dexes. Hence the assumption of normality is not essential to construction 
of selection indexes as now used. 

It must be obvious that the selection index method just described is very 
laborious when a number of different 6 need to be estimated, for the solution 
to a set of simultaneous equations is required for each 6. In practice this diffi- 
culty is avoided to a certain extent by choosing arbitrarily only a few sources 
of information to be employed in selection. This is not a wholly satisfactory 
solution, for in most cases if the number of different indexes is not to be en- 
tirely too large, information must be rejected which could add at least a 
little to the accuracy of the index. 

By means of a simple modification it becomes necessary to solve only one 
set of equations no matter how many B are estimated from a particular set 
of data, and precisely the same index as in the conventional method is ob- 
tained. Using the same notation as before, the index for 6 is now 

where the Cs are the solution to a set of equations identical to set (1) except 
that the right members are W\, . . . , Wn rather than a^y^, . . . , agy^.. Con- 
sequently once the C's are computed, any number of 0's can be estimated 
simply by taking the appropriate linear function of the C's. 

More tedious computations result if the b's are not known. One solution 
is of the following general form. In order that each d be unbiased it is neces- 
sary that the K's have these restrictions imposed: 

A'l.Vn + A'2.V]2 + • • • + K^Xix = 

A'i.r2i + A'2.Y22 + . . . + A.vA'l.V = 

(2) 

A'l-Vp, -f A'2-Vp2 + • • • + A.Y.TpTv = 

Subject to these restrictions the values of the A's which minimize E{d — dY 
can then be found. 

If we wish to obtain estimates of the i's which are unbiased and have 
minimum E(J) — bY, we impose the restrictions of equations (2) except that 



SPECIFIC AND GENERAL COMBINING ABILITY 357 

the right member of the equation pertaining to the particular b to be esti- 
mated is 1 rather than 0. 

An easier solution to the problem of unknown 6's often can be obtained 
by regarding the model as, 

where the d are independently distributed with mean zero and variance al^ 
and the z's are observable parameters. For example, ^i might rei)resent the 
general combining ability of inbred line A, di the general combining ability 
of line B, and di a specific effect peculiar to the cross A X B. The observable 
parameters z would have the following values: Zi = 1 when line A is one 
of the parents, = otherwise; Z2 = 1 when line B is one of the parents, = 
otherwise; and Zs = 1 when ji is an observation on the cross A X B ox 
B X .1, = otherwise. Now the joint estimates of 6's and 0's are the joint 
solution to the subsets of equations (3), (4), and (5). 

<^i< +C2V. +• • • +C,^_,^. = y,- b,x^, - ... - 6^x,, 

'. . (3) 

C,a +C,(7 +...+C,.(72 =y -hx -. . .-h X 

1 i/,y_Y ' 2 y„yx A y^ ' -V 1 UV p p.-v 

1 1 j/j9[ A y^v'i 

(4) 



"? 



where 



b,Sx\ + . . . + bpSx.Xj, + ^i^.vi si + • . . + ^,5.vi2, = 5.viy 

'. . (5) 

Si5:*;iXp + . . . + 6p5.Vp + diSXp zi + . . . + ^,>SXp z, = Sxj,y , 

S.v;=2:#. 5.v,.v.= 2:^?^, etc. 

These equations can be solved by the following steps. First solve for the C's 
in equations (3). The results will be in terms of the sample observations and 
the 6's. Second, substitute values of these C's in equations (4) to obtain ^'s 
in terms of the sample and the b's. Third, substitute these values of the ^'s 
in equations (5) and solve for the 6's. Fourth, substitute the computed values 
of the S's in (4) and solve for the ^'s. 



358 C. R. HENDERSON 

An alternative computational procedure which is less laborious when the O's 
are few in number, and in particular when the 5's are uncorrelated, involves 
joint estimation of the i's and 9's by solution of equations (5) to which are 
added equations (6). 



(6) 



b,Sx,z,-\-. . .+ b,SXj,z,-\-di{Sz,z,+ a'')-\-. . .-^d,(Szl+a"')=Sz,y 
where 



,-'MI = IIV,l 



an 



d 



SxiZi = > — -^, etc 



These equations are simply least squares equations (the d's are regarded 
as fixed rather than having a distribution) modified by adding a"-' to certain 
coefficients. 

SELECTION BY MAXIMUM LIKELIHOOD ESTIMATES 

Now let us assume that the ^'s have the multivariate normal distribution 
and that the errors are normally and independently distributed. What are 
the maximum likelihood estimates of the d's and 6's? It just so happens that 
the estimates which are unbiased and which have minimum E(d — 6)- and 
E(b — b)- for the class of linear functions of the sample are also the maxi- 
mum likelihood estimates. Consequently the estimation procedure we have 
described can be seen to have the following desirable properties: unbiased- 
ness, maximum relative efficiency of all linear functions of the sample, maxi- 
mization of genetic progress through selection by truncation when the dis- 
tributions are normal, properties of maximum likelihood estimates when the 
distributions are normal, and equations of estimation which can be set up in 
a routine manner. 

Unknown Variances and Covariances 

An important problem in selection remains unsolved and perhaps there 
is no practical solution to it. What should be done if the variances and co- 
variances are unknown? If our sample is so large that estimates of the vari- 
ances and covariances can be obtained from it with negligible errors, we can 
use these estimates as the true values. Similarly we may be able to utilize 
estimates obtained in previous experiments. But if there are no data available 
other than a small sample, the only reasonable advice would seem to be to 
estimate the variances from the sample, perhaps modifying these estimates 



SPECIFIC AND GENERAL COMBINING ABILITY 359 

somewliat if they appear totally unreasonable. At any rate the estimation 
procedure serves to point out what additional information is needed if an 
intelligent job of selection is to be accomplished. 

SELECTION FOR ADDITIVE GENETIC VALUES IN INDIVIDUALS 
As our first application of the methods described above, consider the esti- 
mation of additive genetic values of individuals with respect to a single trait 
(the single trait might be net merit) from a set of records all made in the 
same herd or flock. It will be assumed for the present that the population 
mean is known and that records can be corrected satisfactorily for all non- 
random environmental factors. For example, the records might represent 
all of the 305 day, mature equivalent butterfat records made in a herd during 
the past ten years. It is desired on the basis of these records to decide which 
cows should be culled, which heifers should be selected for replacements, and 
which bull calves should be grown out for possible use as herd sires. 

In the usual approach to this selection problem by use of the selection 
index, one would decide what particular subset of the records would con- 
tribute most to the estimate of the value of each animal under consideration 
and would then construct separate indexes. The method to be presented here 
employs all available records in estimating the value of each animal. That is, 
no prior decision is made concerning which records to use to construct the 
index for each animal, but instead all available ones are used. 

The first step in the procedure is the computation of what Emik and Ter- 
rill (1949) have called a numerator relationship chart and Lush (1948) has 
called genie variances and covariances for all animals whose records are to be 
used in the index or whose breeding values are to be estimated. In terms of 
Wright's (1922) coefficients of relationship and inbreeding, the genie variance 
of the iih animal is 1 -|- Fi, where Fi is the inbreeding coefficient of the ith 
animal, and the genie co variance between the ith. andjth animal is 



i?oV(l+F,)(l-|-Fy), 

where i?,, is the coefficient of relationship between the two animals. The nu- 
merator relationship or genie eovariance, which we shall denote by a.y, is the 
numerator of the fraction representing relationship. That is 

7? - ^'' 



Vd+FJd-fF,) 



The computation of ||a,_,|| is a routine procedure if it is done systematically as 
described by Emik and Terrill and by Lush. 

Next we need an estimate of heritability of the trait, and if more than one 
record is available on a single animal, as would be true of butterfat produc- 
tion, an estimate of repeatability. Now let yi, yz, ■ ■ • , yp be the mean of the 
■Hi records of each of p animals, these records having been corrected for 



(7) 



360 C. R. HENDERSON 

non-random environment and expressed as deviations about the population 
mean. The next step is to solve the following set of equations for Ci, . . . , Cp. 
In these equations h denotes heritability and r denotes repeatability. 

Cianh-\-C-2 (FoJH j + • • • -hCpG^ph = 5*2 

* • 

Cia^pJi -\-C2a2ph + . . . +Cp [Fp -\ 'A = Vp 

If all available records are to be used in the estimation procedure just de- 
scribed, the number of equations to be solved for the C's is large. It might ap- 
pear, in fact, that the number is too great for the method to have any value. 
However, the equations are ideally suited to an iterative solution. The reason 
for this is that the diagonal elements of the left members of the equations are 
very large compared to the off-diagonal elements thereby making the itera- 
tive solution a particularly rapid one. On the basis of our experience with a 
few herds a solution to sufficient accuracy can be obtained in three or four 
rounds of iteration. 

Once the C's have been computed the estimate of gt, additive genetic value 
of the iXh. animal, is 

gi = h (Cifli, -\-C2a2i + . . . +Cpap,) . 

If the iih animal had one or more records included in the computation of 
the C's the estimate can be computed more easily, for 

-^ ^ . _^ 1 + ("> - 1) >' - n, h 

The estimate of the real producing ability of a tested animal is even more 
simple to express. The estimated real producing ability is 

y'i—Ci . 

It should be pointed out that this estimate differs from the one presented by 
Lush (1945) since his method does not utilize records on relatives. 

Valuable characteristics of the method just described, in addition to its 
ease of computation and its use of all available information, is that the inclu- 
sion of the records of the contemporaries of the ancestors of the animals being 
appraised automatically eliminates the troublesome problem of what effect 
selection has had on the phenotypic and genetic variances of the selected 
group of ancestors. Also changes in additive genetic variances and covari- 



SPECIFIC AND GENERAL COMBINING ABILITY 361 

ances effected by inbreeding are automatically taken into account. If selec- 
tion is intense, the samj^le mean may considerably overestimate the i)o[)ula- 
tion mean appropriate for subtraction from the records. The safest procedure 
is to regard ^t as unknown and to estimate it by the procedure described 
earlier (equations 3, 4, 5) . It is also of interest to note that joint estimation by 
this method of such factors as environmental trends and age effects automati- 
cally eliminates biases in the estimates resulting from use of selected data. 

SELECTION FOR GENERAL COMBINING ABILITY 
IN TOPCROSS TESTS 

When it comes to estimation of the general combining abilities of inbred 
lines or of the values of specific crosses, apparently no application has been 
made of the selection index method. This failure may have been due to diffi- 
culty in obtaining the estimates of the needed variances and covariances, 
failure to see that the method was applicable, or the opinion that since inbred 
lines can be carefully tested more efficient but complex methods of appraisal 
are not worth the extra computational labor. We propose to show here how 
the methods can be applied to such selection problems, to indicate some situ- 
ations in which it may result in considerably more efficiency in selection than 
the use of the straight means of the lines or crosses as the criteria of selection, 
and to present some approximate solutions which are relatively easy to 
compute. 

Let us consider first one of the most simple tests of lines, the topcross test. 
In this test a random sample of individuals from each of several lines is mated 
to a tester population, and measurements are taken on the resulting progeny. 
If only one trait is considered important, the lines are usually rated according 
to the means of their topcross progeny. This method of ranking is as good as 
any, provided either that the same number of progeny is obtained for each 
line or that the sampling errors of the line means are negligible. Seldom, at 
least in large animal tests, would either of these conditions hold. Accidents 
usually preclude attainment of equal numbers, and sampling errors are usual- 
ly large. If sequential testing is done, numbers would always be unequal. By 
sequential testing we mean here that lines are given a preliminary test, and a 
certain fraction of those performing worst are discarded. Then the remaining 
lines, accompanied perhaps by some new lines, are given another test, and so 
on through any number of cycles desired. The lines surviving several such 
tests would obviously have larger numbers of progeny than the new lines, and 
it would be a very inefficient procedure to disregard the results of prior tests 
on the older lines when choosing between them and the newer, less well- 
tested lines. 

The way in which the lines should be ranked on the basis of all information 
is analogous to choosing between individuals with different numbers of rec- 
ords. In the latter case both repeatability of single records and the number of 



362 C. R. HENDERSON 

records need to be considered; in the former case the genetic differences 
among Hnes, the environmental variance, and the number of progeny. Also in 
both cases consideration of the genetic covariances between individuals or 
between lines increases the accuracy of the ranking. 

Assuming that the population mean is known and that it and non-random 
environmental factors have been subtracted from the means of the progeny 
of the various lines, the estimate of gi, the general combining ability of the 
iih line, is 

£.=C,a- --{-... -\-C a - 

°i 1 BiVi P BiVp 

where yi, • ■ ■ , Jp are the corrected means for the p tested lines and the C's 
are the solution to a set of equations with 

as coefificients in the left members and corrected y\, . . . , y? as the right mem- 
bers. Computation of o-^^j, and o-oiSi requires good estimates of 

Ik II 

and of al. Assuming that the corrected mean of a particular topcross is 
yi = gi'\- ^i, and that the errors are independent with common variance af 
we have the following variances and covariances 

a- = a- -]-a-/n. cr- — a~ 

a-- = a (where i 7^ j) a- = a (i ^ j) 

ViVj aivj -' viOj Qiuj ■' 

Frequently good estimates of n and non-random environmental factors are 
not available and consequently must be estimated from the topcross data. 
For example, it is very likely that the environment is not the same from test 
to test and must be taken into account if the data from several tests are to be 
combined into a "best" index. In such cases the method of equations (5) and 
(6) can be employed to distinct advantage unless 

is too difficult to compute. To illustrate this method as applied to topcross 
data we shall assume that y^j, the record of the/th progeny of the ith line, 
can be represented by 

yu = byXuj -\- hoX'iij + gi -\- e,j . 

bi and 62 are examples of fixed parameters, gi is the general combining ability 
of the /th line, and d, is a random error. Assuming that the gi are distributed 
with means zero and known variance-covariance matrix. 



SPECIFIC AND GENERAL COMBINING ABILITY 363 

and that the dj are mde{)endently distributed with means zero and common 
variance a;, the estimates of the b's and g's which are "best" by the criterion 
used in this paper are the solution to the following equations: 



i>i 



(8) 



^i-vi;,. + b.x.^ + ip ("p+ <T?ff^^) + ^ ^pff^cr"' = >' 






Dots in the subscripts denote summation over that subscript, and a'' denotes 
an element of 



cr 



The above procedure for appraising lines on the basis of topcrosses assumes 
either that the lines are homozygous or that only one progeny is obtained 
from each randomly chosen male. If these assumptions are not correct, the 
procedure is moditied to take into account intra-line variances and covari- 
ances and the number of progeny per male. 

What are the consequences of appraising lines on the basis of the arithmet- 
ic average of their respective progeny as compared to the more efficient 
method just described? First, the errors are larger than necessary. Second, 
selection of some small fraction of tested lines will tend to include a dispropor- 
tionately large number of the less well-tested lines. The more efficient meth- 
od discounts the higher averages in accordance with the number of tested 
progeny and the relative magnitudes of a^ and a;. 

What if the number of lines tested is large and certain lines are related? 
This means that a large matrix, 

11%,, II. 

has to be inverted and then a large set of simultaneous equations solved. 
What approximations might be employed in the interest of reducing compu- 
tations? For one thing, we might ignore the covariances between the g's, 
thereby reducing the inverse matrix to l/a^, in the diagonal elements and 
in the off-diagonal elements. Also if we know /i and non-random environmen- 



364 C R. HENDERSON 

tal factors well enough, further simplification is possible. Let Wi be the cor- 
rected mean of the progeny of zth line. Then 



p = 7^ 



Hi e 



This result is a straightforward application of the principles of the selection 
index. 

It must be quite apparent that efficient appraisals of the general combin- 
ing abilities of lines depend on knowledge of the variances and covariances of 
general combining abilities and of the variance of error. It hardly seems like- 
ly that estimates of the line variances and covariances can be obtained with 
accuracy comparable to estimates of additive genetic variances and covari- 
ances with respect to individuals. The latter estimates are based on studies of 
heritability and on the known facts of the hereditary mechanism. In the case 
of inbred lines, however, the sample of different lines tested is usually so 
small as to make the estimates of al less reliable than we should like. A way 
around this difficulty in the case of traits for which heritability is well known 
is to compute the expected variances and covariances based on knowledge of 
(tI in the original population from which the lines were formed, the inbreeding 
of the different lines, and the relationships between pairs of lines. It seems 
likely that such estimates would be more reliable when the number of lines 
is small than would estimates arising from the actual line tests. We cannot 
be any more precise regarding this point until methods are developed for 
placing confidence limits on estimates of variances and covariances arising 
from non-orthogonal data. 

SELECTION FOR GENERAL COMBINING ABILITY, MATERNAL ABILITY, 
AND SPECIFIC ABILITY IN LINE CROSS TESTS 

If we wish to estimate the general combining ability of lines relative to the 
population from which the lines themselves can reasonably be regarded as a 
random sample, line crosses give, for fixed size of testing facilities, more accu- 
rate estimates than do topcrosses. The reason for this is that we obtain from 
each cross estimates of the general combining abilities of two or more lines. 
Also, line crosses enable one to estimate differences in maternal abilities un- 
confounded with differences in general combining abilities and to appraise the 
values of specific crosses. In those species for which hand mating is the cus- 
tomary procedure, little more labor is required for line cross than for topcross 
tests. The estimation of line and line cross characteristics from line cross data 
is no different in principle from what we have already described with respect 
to estimation of additive genetic values of individuals or general combining 
abilities of lines. As before, we wish to obtain unbiased and most efficient esti- 
mates of certain genetic values. For the sake of simplicity of presentation we 



SPECIFIC AND GENERAL COMBINING ABILITY 365 

shall confine ourselves to discussion of the analysis of single crosses. Applica- 
tion of these principles to multij)le cross data involves no new principles. 

Let us consider first what type of model might be reasonable for a single 
cross. It is not too difficult to suppose that the value of a particular observa- 
tion on a single cross is the sum of the general combining ability of the male 
line, the general combining ability of the female line, a maternal effect coming 
from the line used as the female, a specific effect due to dominance and 
epistasis and peculiar to the particular cross, non-random environmental ef- 
fects, and a multitude of random errors such as Mendelian sampling and the 
environment peculiar to the particular progeny on whom the record is taken. 
More complicated models could of course be proposed, but the one which we 
have just described would seem to account for the major sources of variation 
among crosses. Furthermore it is amenable to mathematical treatment. Put- 
ting the above description in a mathematical model we have 

Jiik = ^i-^iu'^- + boXi^jk + g, + gj + W; -f s,j + eijk , 

where yuk is the observation on the ^th progeny of a cross between the ith 
line used as a male parent and thejth line as a female parent, the 6's and a;'s 
are related to the mean and other non-random environmental factors as de- 
scribed in the model for the topcross test, giigj) is the general combining abili- 
ty of the ith(yth) line, nij is an effect in addition to the additive genetic value 
which is common to all progeny of they'th line used as a female parent, 5,, is 
an effect over and above the additive genetic and maternal effects and which 
is common to all progeny of the cross of the ith line by thej'th line or of the 
yth line by the ith line, and Cijk is a random error associated with the particu- 
lar observation. 

In this model the gi are regarded as having some multivariate distribution 
with means zero and variance-covariance matrix, 

11 '.,-., II- 

The wy, Sij, and e^j are all regarded as independently distributed with means 
zero and variances o-^, o",, and al, respectively. It is of course conceivable that 
the variances of the m, and sa and the covariances between them vary with 
the inbreeding and relationships of the lines. Also gi and ntj may be correlat- 
ed. In the absence of any real knowledge concerning such covariances we 
shall ignore them for our present purposes. If, however, something is known 
about these covariances, the estimation procedure can be modified to take 
them into account. The procedure should also be modified if the lines are not 
homozygous and each parent has more than one progeny. 

A single cross test can supply answers to the following questions with re- 
spect to the lines tested: 

1. What are the best estimates of the relative values of the tested lines 



366 C. R. HENDERSON 

when used as the male parent in topcrosses on the population from which 
the lines are regarded as a sample? 

2. What are the best estimates of the relative values of the tested lines as 
female parents in crosses with males from the above population? 

3. What are the best estimates of the relative values of specific single 
crosses among the tested lines? 

Suppose that tin progeny of the cross tth line of male by 7th line of female 
are tested (nn can be zero for some crosses) . Now the easiest way to estimate 
the value of the ith line as a male parent is simply to compute the mean of 
all progeny of the line when used as the male parent. This simple procedure, 
however, fails to take into account the distribution among lines of the mates 
of males of the /th line, the covariances among the general combining abili- 
ties of lines, the consequences of specific effects, the size of the error variance, 
and the number of progeny tested. Furthermore, since the zth line is used also 
as the female parent in certain crosses, something can be gained by employ- 
ing the measurements on these progeny. Estimation by the general procedure 
we have described takes into account all of these factors. Similarly the easiest 
way to estimate the maternal ability of thejth line is to compute the mean 
of all progeny out of females of the /th line, but the most efficient procedure 
takes into account the same factors as are needed in efficient estimation of 
general combining ability. Finally the easy way to appraise the value of a 
particular cross is merely to find the mean of all progeny of the specific cross 
(if that cross has been tested). This latter estimate is subject to large sam- 
pling error since it would seldom be feasible to test many individuals of the 
numerous possible crosses among even a few lines. The error of estimation can 
be materially reduced by utilizing the fact that the true merit of a cross is a 
function of the general combining abilities of two lines, the maternal ability 
of the female line, and the specific effect peculiar to that cross and to its re- 
ciprocal. The method to be described places the proper emphasis on estimates 
of general and maternal abilities and on the progeny averages of the specific 
cross and its reciprocal. The procedure also enables estimates to be made of 
the value of a specific cross even though that particular cross has not been 
tested. 

The major step in these efficient estimation procedures is the setting up 
and solving of a set of simultaneous equations in the 6's, ^'s, ?n's, and s's. 
These equations are as follows: 

^1 2 y^ 2 -VluA- + ^2 23 2 2 •^■l'J>^"2 0-i- + 2^ gi i^l'.. + -^'l. I. ) 

i j k i i k i ( 9 ) 



and similarly for the bo equation. 



i<; i J k 



SPECIFIC AND GENERAL COMBINING ABILITY 367 

biixn.. +^-1.1.) +^2(a-2i.. +:t2.i.) +^i(«i. + ".i + ff^cr*^) 

= 3'!.. +>'.!. 

and similarly for the other gi equations. 

^l-Vl.l. + ^2-V2.1. + gl«.l+2^ gi«.l+ Wi (;/.!+ Cr!/ffl,)+ 2 5i,M,i = J.l. 

and similarly for the other ;«; equations. 

^(■Vn2. +-Vi21.) +6.2(.V212. +-V22I.) + (gl+ g2) ("12 + W21) +/«1«21 

+ W2H12+ ■Jl2(«12+»21 + ae/(rJ = yi2. +3'21. 

and similarly for the other Sij equations. 

These equations are not particularly difficult to solve, for each Sij can be 
expressed as a function of ya., y;,-., 5i, 62, Qi, Qj, and mj. Utilizing this relation- 
ship the equations can be reduced to a set involving none of the ,Sij. Also an 
iterative solution is usually easy because of the relatively large diagonal co- 
efficients. Once the estimates of gi, nij, and s^ are obtained it is a simple mat- 
ter to evaluate the lines and crosses. The estimate of the value of a line as the 
male parent in topcrosses is Qi, and the estimate of its average value as the 
female parent is ^i + ifu. The value of a single cross is estimated simply as 
9 1 + di + '^ij + •^u- It is appropriate to add the estimates in this manner be- 
cause they have the desirable property of invariance. 

If solution of the large set of simultaneous equations required for most ef- 
ficient appraisal of lines is considered too burdensome, certain approximate 
solutions can be employed. An approximation suggested by the common 
practice in construction of selection indexes is the choosing of certain infor- 
mation most pertinent to the particular line or cross to be appraised. For ex- 
ample, the estimate of g, might be based entirely on y^.. and y.,., each cor- 
rected for the b's as best can be done with the information available regarding 
their values. As a further simplification it might be assumed that the gi are 
uncorrelated and have common variance o-|. Similarly w, might be estimated 
entirely from y,. . and y.,. . These approximate solutions are 

g. =C ,a +C><7 
fh ■ =C ,(j -\-C„(j , 
where the C's are the solution to 



368 C. R. HENDERSON 

The variances and covariances needed in this approximate solution can be 
computed easily from a^, a^, a;, and a^. Approximate values of §ij can then 
be obtained by substituting the approximate bi, hi, Qi, and rhj in equations (9) . 

ESTIMATION OF VARIANCES OF GENERAL, MATERNAL, 
AND SPECIFIC EFFECTS 

As mentioned earlier, one might take as the additive genetic variance and 
covariance among the lines the theoretical values based on relationships 
among the lines, degree of inbreeding among the lines, and the genetic vari- 
ance in the original population from which the lines came. It is necessary even 
then to estimate al^, af, and a;. It is well known that methods for estimating 
variance components are in a much less advanced stage than estimation of 
individual fixed effects. It is seldom possible to obtain maximum likelihood 
estimates. Consequently many different methods might be used, and the 
relative efficiencies of alternative procedures are not known. 

We shall consider as desirable criteria of estimation procedures for vari- 
ance components ease of computation and unbiasedness. If the single cross 
experiment is a balanced one, that is if there are the same number of observa- 
tions on each of the possible crosses, it is not difficult to work out the least 
squares sums of squares for various tests of hypotheses, regarding the line 
and cross line characteristics as fixed. Then assuming that there are no co- 
variances between the various effects and interactions, one can obtain the ex- 
pectations of the least squares sum of squares under the assumption that the 
effects and interactions have a distribution (Henderson, 1948). In case the 
experiment is not a balanced one, it is still possible to obtain least squares 
tests of hypotheses and to find expectations of the resulting sums of squares- 
This, however, is ordinarily an extremely laborious procedure (Henderson, 
1950). 

A much easier procedure is available. It probably gives estimates with 
larger sampling variance, although that is not really known, and gives almost 
exactly the same results in the balanced experiments as does the least squares 
procedure. This involves computing various sums of squares ignoring all cri- 
teria of classification except one, taking expectations of these various sums of 
squares, and solving the resulting set of simultaneous equations. The latter 
procedure will now be illustrated for single cross data in which we wish to 
obtain estimates of the variances pertaining to general combining ability, 
maternal ability, specific effects, and error. It will be assumed that the only 
fixed element in the model is /x. Now let us compute certain sums of squares 
and their expectations. These are set out below. 

Total: i^(X;2: Vy?,,)=>/..(M^ + 2a^ + ai + a^4-ap 



SPECIFIC AND GENERAL COMBINING ABILITY 369 

Sires: £ f Y) -'-') = ;/.. (m" + ^D + S — (<t"' + <t2 + a^) +5^2, 

where 5 denotes number of different lines used as the male line. 



n-.. 
'I 



where d is the number of different lines used as the female line. 
Crosses: e( V -^^-^^■'■•^-') = «•• (m^ + 2a^ + a'-') 



C<T- 



where c denotes the number of different crosses (regarding reciprocals as one 
cross) 

Correction Factor: -E (^-— ) = "••M"+ ^ (;/, -f ».,) Vp/M.. 



2 

e 



The above sums of squares and expectations are quite easy to compute and 
once this is done all one needs to do is to subtract the correction factor and 
its expectation from the other sums of squares and expectations and solve the 
resulting set of four equations for a\, cr^,, a,, and af. 

FURTHER RESEARCH NEEDED 
If maximum progress through selection for general and specific combining 
ability is to be attained, much additional research is needed. From a statisti- 
cal standpoint we need to know if an index based on minimization of E{6—dY 
comes close to maximizing progress through selection by truncation when the 
distributions are not the multivariate normal. If such an index does not do 
so, we need to know what practicable index or indexes will. Further, if 
nothing is known of the variances and covariances needed in construction of 
indexes or if there are available only estimates with large sampling errors, we 
need to know if the index based on the assumption that the estimate is the 
true value is best from the standpoint of maximizing genetic progress. Final- 
ly, much more work is needed on the problem of estimating variance and co- 
variance components and placing confidence limits on such estimates. 



370 C. R. HENDERSON 

Although there is a considerable body of literature on heritability esti- 
mates, we need more accurate estimates of the heritabilities of most traits of 
economic importance. Also almost nothing is now known about genetic cor- 
relations between traits, about genetic-environmental interactions, and 
about the magnitude of genetic differences among herds. Estimates of these 
genetic parameters are essential to intelligent selection for additive genetic 
values. In the case of inbred lines, little is known concerning the variances 
of general and specific combining abilities. The work of Sprague and Tatum 
(1942) with corn and Henderson (1949) with swine illustrates the types of 
estimates which are badly needed in selecting for general and specific combin- 
ing abilities from the results of line cross tests. 

Finally, well designed experiments are needed to test how closely predic- 
tions made from indexes or other selection procedures check with actual re- 
sults. 



L M. WINTERS 

University of Minnesota 



Chapter 23 

Rotational Crossbreeding 

and Heterosis 



It is well for all of us, including our most eminent scientists and philosophers, 
to reduce our thinking to relatively simple terms. Genetics is, after all, basi- 
cally rather simple. A fertilized zygote results from the union of two germ 
cells, each of which carries a haploid number of chromosomes, and a haploid 
number of genes which are resident in the chromosomes. By the very nature 
of the procedure, genes are paired which are alike or not alike. As the pairing 
of similar genes is increased, the population approaches increased purifica- 
tion. As the pairing of dissimilar genes increases, the resulting population be- 
comes more heterozygous. Increased heterozygosity has been generally as- 
sociated with increased vigor which is generally spoken of as hybrid vigor. 

PLANNING THE MINNESOTA EXPERIMENTS 

I believe the best way to discuss rotational crossbreeding is to relate briefly 
how the system was developed. When I was asked in 1928 to head the re- 
search in animal breeding at the University of Minnesota, I brought with me 
several proposed projects. One of these was a study of crossbreeding swine. 
A review of the literature of crossbreeding experiments conducted previous 
to 1928 shows that for the most part they were small-scale experiments. 
When the data were all put together, however, the evidence was in favor of 
crossbreeding. Yet, the general sentiment at that time among the stockmen 
was overwhelmingly opposed to the practice of crossbreeding. The statement 
frequently heard was that crossbreeding was quite satisfactory for the pro- 
duction of one crop, but all of the crossbreds must be marketed because it 
was absolutely disastrous to use any of the crossbreds as breeding animals. 
By 1928, it was quite evident that corn breeders were revolutionizing the 
system of breeding corn, and that hybridization was to become the rule rath- 

371 



372 L. M. WINTERS 

er than the exception in the production of commercial corn. Wright (1922) 
had several years previously published what has since turned out to be a 
classic: The Report of the U.S.D.A. Studies of Inbreeding and Crossbreeding 
with Guinea Pigs. Why then should the situation be different in livestock 
than in corn and in guinea pigs? W^as it true that livestock failed to respond 
to crossbreeding, and if so, why? A likely explanation appeared to be that our 
breeds of livestock were not truly comparable to inbred lines of guinea pigs 
and corn because they did not possess sufficient genetic purification. 

I am sorry now that I did not record in advance of this experiment the re- 
sults that I expected to derive. Had I recorded them, they would have been 
something like this: The crossing of the breeds of livestock will result in a 
slight increase in vigor. The increase will be so slight that it is scarcely worth 
while for the commercial producer, in contrast to the more simple procedure 
of grading or the maintenance of a registered herd. Most of my severe critics 
regarding crossbreeding will be quite surprised to read this statement. As 
nearly as I can tell at this time, there were two major reasons for my belief. 
The first was the accumulation of the continued absorption of a large amount 
of teaching toward that end. The second was the general belief on the part 
of geneticists that our breeds of farm animals had not been sufficiently puri- 
fied nor separated genetically to yield hybrid vigor when crossed. 

If I had recorded all of my thoughts, they would have included this reser- 
vation: If the crossing of the breeds does result in increased performance suf- 
ficient to make crossing worth while, then the standard advice that had been 
given through the years regarding the use of crossbred females for breeding 
must be erroneous. This reservation was based on the results that Wright had 
previously obtained in his use of crossbred guinea pigs as dams, and from the 
information already available regarding the production of hybrid corn. At 
this time there was no thought regarding continuous crossbreeding by either 
rotation or crisscross breeding. The objectives were merely to find out if there 
was any advantage in crossing the breeds for the market production of swine. 
If there was an advantage in crossing the breeds for market production of 
swine, was there then any advantage in retaining these crossbreds to become 
future parents? 

In planning the experiment, provision was made whereby as nearly as 
possible the same genetic material was put in the crosses as was produced 
in the purebreds. In planning the use of crossbreds as parents, the original 
plan called for the use of both crossbred females and crossbred males. My 
senior officers informed me that they were willing to go along with me quite 
a way in this crossbreeding study, but that when it came to the use of cross- 
bred males, that was going just a bit too far and I would have to compromise. 
I compromised on this point all too willingly. How I have wished, during the 
last few years, that I had insisted on carrying out my original plan of 
using both crossbred females and crossbred males in the experiment. But little 



ROTATIONAL CROSSBREEDING AND HETEROSIS 373 

did any of us realize at that time that within twenty-two years we would be 
in the midst of a flourishing hybrid boar business. None of us know much 
about their true merits and demerits. Nevertheless, the crossbred boars were 
not included in the experiment. 

Experimental Results 

The experiment did proceed as planned for the production of first cross 
offspring from the mating of purebred females of one breed to purebred males 
of another breed. Crossbred females were then retained as breeding animals 
to be mated in one case to a boar of one of the two parental breeds and in the 
other case to a boar of a third breed. The results of this experiment showed 
that there was a very definite advantage in the production of firstcross pigs. 
There was a slightly greater advantage in the production of backcross pigs 
(that is, where crossbred females were mated back to a boar belonging to one 
of the parental breeds). There was a still greater advantage where the cross- 
bred females were mated to a boar of another breed. 

As I mentioned before, you bring together either genes that are alike or 
genes that are not alike. There appears to be very little likelihood of bring- 
ing about any more heterozygosity as a result of a three-breed or a four-breed 
cross than there is in a two-breed cross. The advantages derived from the 
backcross and from the cross to a third breed appear therefore to have been 
derived from the fact that the female parents were crossbreds or in a more 
hybrid state than their purebred half sisters. Why should this be the case? 
The female produces the eggs, carries the fertilized eggs, and develops them 
to the point where, after a period of about 114 days, they are ready for birth 
and then nurses the little pigs for another 56 days. In general, the advantage 
derived from the crossbred female in contrast to the purebred female is 
about equal to that derived from having the progeny crossbred in contrast 
with having progeny that are purebred. 

The above are the general deductions that I made at the close of our cross- 
breeding experiment. Now I am not so certain that this interpretation is ab- 
solutely correct. The reason for my questioning is that recently I had a long 
visit with one of the largest hybrid seed corn producers in this nation. He is a 
man who has had many years of experience in the field. lie told me that he 
had not yet seen a single cross of hybrid corn that was as useful for com- 
mercial corn production as the double hybrid. He elaborated further to the 
efifect that the single cross hybrid often would yield as heavily as the double 
hybrid, but that under adverse environmental conditions the double hybrid 
fared better. This he attributed to the fact that the double hybrid de- 
veloped from four inbred lines possessed greater genetic diversity toward 
adversity. This appears to be somewhat in contrast to the experimental re- 
sults and interpretations of those results in some of the ])resent-day funda- 
mental studies of Urosophila genetics. Undoubtedly, with time and more 



374 L. M. WINTERS 

experimental results, we will be in a better position to bridge this gap. My 
experiences have convinced me, however, that it is a mistake not to take 
seriously the observations made by competent practical men in the field of 
operations. I am inclined to believe that very often these men see more, al- 
though they measure less accurately, than we in the field of research, with 
our eyes glued carefully to the job of measuring certain details. 

In this experiment we used four standard measurements for appraisal of 
the pigs' worth. These were: number of pigs born alive, number of pigs 
weaned, rate of gain, and feed per unit of gain. Since then we have added a 
fifth measure — appraisal of body form on the basis of judgment. When we 
took the first four factors and compared the performance of the crosses with 
the comparable purebreds, we obtained an advantage of the crosses over the 
purebreds of 6.3 per cent for the first cross, 7.5 per cent for the backcross, 
and 11.7 per cent for the three-breed cross. This was obtained by throwing 
the four factors together as equal in importance. 

By another method of comparison, wherein more factors were thrown into 
the pool, we obtained an advantage for the first cross of 7 per cent, the back- 
cross 6 per cent, and the three-breed cross of 17 per cent. If, however, we 
were to take litter weight at weaning, which in one sense is comparable to 
yield in corn, we would have an advantage of the first cross of approximately 
25 per cent, the backcross 39 per cent, and the three-breed cross 61 per 
cent. If we were to take total litter weight at the close of the experiment, the 
advantages of the crosses would be still greater. In my opinion, total litter 
weight as a sole measure of merit exaggerates the diliference. On the other 
hand, I do not consider the method we have used entirely satisfactory. I do 
not know of an entirely satisfactory measure of performance in livestock. 
We in the livestock field need to do a great deal in the matter of perfecting our 
methods of measurements. The important question to the practical man is 
whether one procedure is better than another, rather than whether this pro- 
cedure gives me exactly 20 per cent or 18 per cent increase. 

ROTATIONAL CROSSBREEDING 

On the basis of these results, we developed and put forward our plan of 
rotational crossbreeding. Even at the time that I started to analyze the data, 
I did not believe that our three-breed cross had given us any worth-while 
advantage over the single cross. I mention this merely to show how strongly 
entrenched the old teaching had become regarding the limitations of cross- 
breeding in livestock production. The results of the experiment were, how- 
ever, very definite. I calculated and recalculated, and the results were always 
essentially the same- — the three-breed cross possessed distinct advantages 
over the first cross and over the backcross. 

Simple calculation shows that, on the average, the first cross will possess 
50 per cent of the chromosomes, or more properly speaking, linkage groups 



ROTATIONAL CROSSBREEDING AND HETEROSIS 375 

of breed 1, and 50 y^er cent of breed 2. The second year, wherein three breeds 
are used, the resulting pigs will, on the average, possess: 25 per cent of 
the chromosomes from breed 1, 25 per cent from breed 2, and 50 per cent 
from breed ^. The third year, the pigs will possess 62.5 per cent of the 
chromosomes from breed 1 , 12.5 per cent of breed 2, and 25 per cent of breed 
3. The fourth year, the pigs will carry, on the average 31.25 per cent of the 
linkage groups from breed 1, 56.25 per cent from breed 2, and 12.5 per cent 
from breed 3. The lifth year, the pigs will possess, on the average: 15.63 per 
cent from breed 1, 28.12 per cent from breed 2, and 56.25 per cent from breed 
3. From that time on, they will remain in a continuous cross, in about that 
general state of equilibrium, but the percentage of relationship to the differ- 
ent breeds will change. 

On the basis of these calculations, we advocated rotational crossbreeding. 
Some of our critics could not understand how we felt Justified in recommend- 
ing rotational crossbreeding when our experiments had been carried only to 
the three-breed cross. Calculations showed so clearly that if the three-breed 
cross was good, then the continuous cross, by rotation, could not help being 
successful, insofar as the system of breeding was concerned. On the basis of 
the theory I have always contended that there was very little advantage in a 
four-breed cross. Now, however, I am not so sure that that is correct, if we 
are to take seriously what my commercial hybrid corn producer told me 
regarding the merits of the double cross of corn in contrast to the single 
cross. There may be merits in the four- or even the five-way cross that are 
not generally revealed in short-time experiments. 

We have recommended rotational crossbreeding for commercial swine 
production, and it seemed, on the basis of theory again, that the rotational 
scheme of crossing had a particular aptitude for swine production, and was 
perhaps questionable with other classes of four-footed farm animals. The 
reason for this is that in swine it is possible for the commercial producer to 
turn a generation every year if he so desires. I have, however, a number of 
friends who are breeding commercial flocks of sheep after this general pattern 
with remarkably good results. If you look at their flocks with the strictly 
commercial viewpoint, they do not have the variance that most critics of the 
plan have contended would result. Further than that, the experiments con- 
ducted by the United States Department of Agriculture with beef cattle and 
dairy cattle have shown that the same basic principles ap})ly to these classes 
of livestock as in swine. Dairymen have perhaps been more reluctant to de- 
part from the purebred philosophy of breeding than any other group of live- 
stock breeders. Yet by a strange coincidence, the experiments of the United 
States Department of Agriculture are showing a greater increased yield as 
the result of crossing dairy cattle than the crossing of any of our other species 
of farm animals. Their data show an increase of 25 per cent in milk and 32 
per cent in butterfat yield. 



376 L M. WINTERS 

ROLE OF INBRED LINES 

The next logical question then is: Where and how do inbred lines enter 
this general picture? I cannot see that it changes the picture appreciably 
unless perhaps it gives an added reason as to why four or five inbred lines 
may (theoretically speaking) prove of advantage over the three-line rotation- 
al cross. We have now carried the continuous rotational cross of three inbred 
lines in two series of crossings to the seventh continuous generation of crossing. 
We have several others in the sixth, and several in the fifth generation. The 
comparative results of the different line crosses have been remarkably similar 
and uniform from generation to generation. 

I have already given the average increased performance of the different 
breed crosses as being 6.3 for the first cross, 7.5 for the backcross, and 11.7 
for the three-breed cross. What then are the increases obtained from crosses 
of inbred lines? By the same method of comparison used in breed crosses, 
except in this case including an estimate on type, we obtained an average of 
approximately 12 per cent increased performance for the crossing of inbred 
lines belonging to the Poland China breed, and an increased performance of 
18 per cent when we crossed the Minnesota No. 2 with our inbred Poland 
China lines, and 20 per cent when we crossed Minnesota No. 1 with Minne- 
sota No. 2 or crossed Minnesota No. 1 with our inbred Poland China lines. 
This is an increased performance over the performance of the inbred lines. 

I am constantly asked what the comparative performance of our crosses 
of inbred lines with the performance of outbred stock is. By the best methods 
with which we have been able to make comparisons to date, the increased 
performance of our crossbred lines in comparison to the performance of the 
old-line breeds is an increase of about 20 to 25 per cent. One of these compari- 
sons was made with outbred stock from our own University of Minnesota 
purebred herds. The other comparison is with the performance of purebred 
herds as given by Lush and Molln (1942). I do not regard either of these 
comparisons as entirely adequate, and again I will frankly state that I do 
not know how to make a comparison that will be entirely adequate. I would 
be much obliged if someone would present me with a plan by which a satis- 
factory comparison can be made. 

I cannot conceive of any sampling method (sampling of the breeds) that 
will constitute an adequate sample of the breeds for comparative purposes, 
unless we go far beyond any funds that I can conceive of being made avail- 
able for this purpose. Field trials such as have been conducted for compari- 
sons with corn have been advocated. Some of the corn breeders inform me 
that they are not at all satisfied that these field trials are adequate. One 
reason is that yield is not a sufficient measure. Many farmers have told 
me that our own estimates of the advantages of crossing both the standard 
breeds and the use of our inbred lines is in error, due to an underestimate 
rather than an overestimate of the benefits. 



ROTATIONAL CROSSBREEDING AND HETEROSIS 377 

Contrary to expectations, our three-line crosses have not given us as much 
increased performance over the two-line crosses as I expected on the basis of 
the results with breed crosses and theoretical expectations. I do not know 
the cause, but I am inclined to believe that it is due to inadequate sampling, 
and that as our samples become larger the advantages of the three-breed 
continuous cross will become more pronounced. 

FARM APPLICATIONS 

How does this work out on farms? The records on one of the large farms 
with which I am working show that their percentage of survival from the 
purified lines (230 litters) is 75 per cent, whereas their survival from the 
crosses of lines (248 litters) is 92 per cent under the same conditions. The 
survival of crossbred pigs out of crossbred sows is 91 per cent, but the cross- 
bred gilts weaned an average of 9.1 pigs, to 8.3 for the first cross pigs and 7.2 
for the purified lines. 

This discussion would not be complete without reference to hybrid boars 
and how they are entering the picture. I have not seen sufl&cient data to al- 
low me to appraise properly the advantages and disadvantages of the so- 
called hybrid boar, but he does seem to be proving popular with a number of 
farmers. If, then, the hybrid boar is here to stay, what is his place in rota- 
tional breeding? In my opinion, it will not change the basic situation ma- 
terially, except that at least six inbred lines will be needed to produce the 
boars for the rotational crossing of the production of commercial stock. In 
this case, we will then use hybrid boar of lines 1 and 2, the following year 
hybrid boar of lines 3 and 4, the next year hybrid boar of lines 5 and 6, and 
then we will go back to 1-2, to 3-4, and to 5-6 in rotation. 

In thinking about rotational crossing, we need to keep in mind that it is 
merely a procedure whereby we are able to maintain our breeding females 
(and perhaps our breeding males), as well as the offspring, in a relatively 
permanent hybrid state. It in no way affects the basic concepts of hybridiza- 
tion. It is just a means of utilizing hybridization, and if at some future date our 
methods of production change, as for instance the general development of 
so-called pig hatcheries, then we may well find some other method of cross- 
breeding better suited to the swine industry. 



E. L PINNELL 

E. H. RINKE 

and 

H. K. HAYES 

Universify of Minnesofa 



Chapter 24 

Gamete Selection for 
Specific Combining Ability" 



Gamete selection as a breeding method was designed for more efficient 
sampling of open-pollinated varieties. It was suggested by Stadler in 1944. 
The method was outlined in detail by Stadler (1945) and preliminary data 
presented. Hayes, Rinke, and Tsiang (1946) proposed that the same technic 
could be used to select gametes from such sources as synthetic varieties, 
single or more complex crosses, and inbred lines. They discussed gamete 
selection in its relation to the improvement of a particular double cross 
combination. 

MATERIALS AND METHODS 

In 1945, three double crosses, Minliybrids 602, 607, and 406, were selected 
for a method study in gamete selection. Single cross performance data shown 
in Tables 24.1 and 24.2 indicate that A344 is low in combining ability in 
Minhybrids 602 and 607, and that the same is true for inbreds A25 and A73 
in Minhybrid 406. 

A344 was crossed to Minnesota #13 (Morris strain) and to 8 inbred lines 
namely, Oh51A, A97, 1234, A315, A348, A367, A396, and 111. 4226 as sources 
of gametes. The inbreds were selected because of their diversity of origin 
and good general combining ability. In addition, A367 had yielded well in 
specific tests with A357, A385, and A392. A315 and A348 had performed well 
in crosses with A392. The remaining five inbreds had not been crossed to 
A357, A385, or A392 in previous years. A25 was crossed with Golden King 
and A73 with Murdock. In 1946, individual Fi plants of these crosses were 

* Paper No. 2591 of the Scientific Journal Series, Minnesota Agricultural Experiment 
Station. 

378 



GAMETE SELECTION FOR SPECIFIC COMBINING ABILITY 379 

selfed and crossed to the opposing single cross parent of the double cross. 
Table 24.3 gives the number and type of test crosses made. 

Approximately 60 inbred X variety Fi plants from each of the three sources 
were selfed and crossed to the testers. Selection of the better Fi plants at 
harvest reduced the number for further study to 35, 32, and 38 as listed in 
Table 24.3. P:xperiments 1 and 2 were tested in separate randomized blocks 

TABLE 24.1 
PERFORMANCE OF A344 AND A334* 



Inbred s 


Av. OF Crosses 


% Moist. 


Bu. 


A344XA357, A392, A385 

A334XA357, A392, A385 


28.3 
30.1 


69.8 
84.3 



* As indicated by average moisture and yield when crossed in 
non-parental single cross combinations of Minhybrid 602 (A344X 
A334) (A357XA392) and Minhybrid 607 (A344XA334) (A357X 

A38.S). 



TABLE 24.2 
PERFORMANCE OF A25, A334, A73, AND A375* 



Inbreds 


Av. OF Crosses 


% Moist. 


Bu. 


A25 XA73, A375 


24.6 
24.7 

24.6 

24.7 


76 2 


A334XA73, A375 

A73 XA25, A334 


79.4 
74 8 


A375XA25, A334 


80.8 



* As indicated by moisture and yield in non-parental single 
crosses of Minhybrid 406 (A25XA334) (A73XA375J. 

using two replicates at each of three locations in central Minnesota. Data 
from the two testers were averaged to give a total of 12 replicates as a test 
for each gamete. Experiments 3 and 4 were also grown in randomized blocks 
with three replicates at each of four locations in southern Minnesota. One 
location of experiment 3 was discarded and one replicate of experiment 4 was 
abandoned prior to harvest. 

On the basis of yield trial results in 1947, gametes were selected from all 
varieties and from two inbreds, for use as parents in the development of 
new lines by straight selfing and by backcrossing to the sampler inbred. 

Study of the performance indices and agronomic characters of the test 
crosses led to the selection of three Golden King gametes and four Murdock 



380 E. L. PINNELL, E. H. RINKE, AND H. K. HAYES 

gametes as higher in general desirability than the sampler inbred parent. In 
addition, gametes of low yield potential but of relatively satisfactory agro- 
nomic characters were selected from both varieties, three from Golden King 
and two from Murdock. 

Fo populations were grown in 1948 from the selfed ears of the twelve Fi 
plants selected in the above manner. Visual selection was practiced for desir- 
able plant and ear characters, and these individual plants were crossed to 
the appropriate single cross tester. Remnant seed of the test crosses of the 
selected Fi plants was used to make a direct yield comparison in 1949 with 
the test crosses of the selected F. plants. Two randomized block experiments 
were made at each of three locations in southern Minnesota with three 

TABLE 24.3 

GAMETE SOURCES, TESTERS USED, AND TEST CROSSES 

MADE IN SELECTING GAMETES FOR IMPROVEMENT 

OF INBREDS ASM, A25, AND A73 



Experi- 
ment 
Number 



1, 

2 

3 

4 



Inbreds 



A344 
A344 

A344* 
A344 

A25 

A73 



Gamete Source 



Tester 



Morris 13 gametes 

Inbred gametes 

A344 

Morris 13 gametes 

Inbred gametes 

A344 

Golden King gametes 

A25 

Murdock 

A73 



Number 

of 
Crosses 



A357XA392 

A357X.\392 

A357XA392 

A357XA385 

A357XA385 

A357XA385 

A73XA375 

A73XA375 

A25XA334 

A25XA334 



35 
8 
1 

35 
8 
1 

32 
1 

38 
1 



* Same plants as in experiment 1. 

replications per location. One experiment consisted of the crosses of 6 Fi 
plants (remnant seed) from A25 X Golden King, and 34 F-, plants by the 
tester compared with the cross of A25 X tester. The other included test crosses 
of 6 Fi and 37 Fo plants from A73 X Murdock gametes in comparison with 
A73 X tester. 

EXPERIMENTAL RESULTS 
The Morris strain of Minnesota 13 has been a very outstanding open- 
pollinated variety in Central Minnesota for many years. Yield trial data 
from plants of A344 X Morris 13 crossed with the testers show that a large 
proportion of Morris 13 gametes have higher yield potential than A344. 
Table 24 4 gives the distribution of yield and moisture data obtained from 
thirty-five test crosses of A344 X Morris 13. Sixteen of the thirty-five gametes 
gave test-cross vields significantly in excess of A344 X tester. Although not 
higher in yield, five other gametes may be considered superior to A344 be- 



GAMETE SELECTION FOR SPECIFIC COMBINING ABILITY 



381 



cause of their significantly lower moisture content at harvest and yields not 
significantly different in test crosses from A344 X tester. 

The eight inbreds tested as possible sources of germ plasm for the improve- 
ment of A344 were A97, A315, A348, A367, A396, OhSlA, Ia234, and 111.4226. 

TABLE 24.4 

DISTRIBUTION OF PER CENT MOISTURE AND 
YIELD OF 35 F, PLANTS OF A344 X MORRIS U 
CROSSED TO S.C. TESTERS A357 X A392 AND 
A357 X A385* 



^(Mean A344 
X tester) 



% Ear +1 




3 


5 1 


Moisture _, 
-2 


2 3 
3 


6 

2 


6 2 
2 




-2 -1 
Yield 


+ 1 


+2 +3 



(Mean A344 X tester) 

* Classes separated by one or more than one LSD (5 per cent) from the 
mean of A344Xtester. 



TABLE 24.5 

DISTRIBUTION OF PER CENT MOISTURE AND YIELD OF 
TEST CROSSES OF 8 INBRED LINES AS SOURCES OF GAM- 
ETES. CROSSES ARE OF THE TYPE (A344 X INBRED) X TEST- 
ERS* 



^(Mean of A344 
X tester) 



%Ear 


+2 
4-1 


2 


1 1 

1 1 


Moisture 


-1 


1 


1 






-2 -1 
Yield 


4-1 4-2 -H3 4-4 4-5 



i 

(Mean of A344 X tester) 
* Classes separated by one or more than one LSD (5 per cent) from the mean of A344 Xtester. 

The distribution for yield and moisture of test crosses of A344 X Inbred is 
given in Table 24.5. Three inbreds, Ia234, A396, and A97, demonstrated 
yield potential significantly higher than A344. 

Golden King and Murdock are old, well-adapted varieties formerly 
grown extensively in southern Minnesota. Test crosses of thirty-two Golden 
King gametes X A25 and thirty-eight Murdock gametes X A73 are shown in 
Tables 24.6 and 24.7, respectively. 

No gamete from Golden King exceeded A25 significantly in test-cross 
yields. However, eight not different in yield were significantly earlier, and 



382 E. L. PINNELL, E. H. RINKE, AND H. K. HAYES 

are considered superior in yield performance on the basis of maturity. Eight 
gametes from Murdock demonstrated yield potential greater than A73, as 
indicated by significantly higher yields in crosses. In addition, fourteen 
gametes not different from A73 in yield were significantly earlier in maturity. 
The proportion of promising gametes extracted from the three varieties 
and the eight inbreds is summarized in Table 24.8. About 25 per cent of the 
total number tested were superior to the sampler inbred in yield potential. 
Another 25 per cent would be considered desirable parents because they had 
a yield potential equal to and a maturity potential which was significantly 
earlier than the sampler inbreds. 

SELECTION OF GAMETES AS PARENTS AND TESTS 
OF F2 PROGENIES 

Years of testing at Minnesota have led to the conclusion that, in general, 
there is a direct relation between yield and moisture content at husking 
among hybrids of equal genetic desirability. On this basis the combining 
ability of inbred X gamete plants was determined by considering both yield 
and moisture percentage at husking. They were effectively placed on a com- 
parable basis by calculation of a performance index using the test cross of 
the sampler inbred as 100 for both ear moisture and yield. For example, if 
the moisture percentage of an A25 X Golden King plant (in test cross) was 
93.5 as compared with A25 X tester, and its yield was 106.5 per cent, its per- 
formance index would be -f 13. Where the comparative moisture percentage 
is higher than the yield percentage the index becomes a negative value. 

The performance indices for the selected gametes for both 1947 and 1949 
trials and similar data for the F2 plant progeny tests appear in Tables 24.9 
and 24.10. The tests of F2 plants from gametes of both high and low yield 
potential were made as explained in "Materials and Methods". This wa scar- 
riedout by selfing selected F2 plants and also crossing them with the appro- 
priate tester, and again comparing the results with the crosses of the appro- 
priate inbred with the same tester. Agreement between the two tests of the 
gametes was very good except for Murdock gametes numbered 12 and 49. 

On the average there was good agreement between Fi and F2 progeny 
performance. Tables 24.11 and 24.12 show that there is evidence of segrega- 
tion for yield factors within almost all of the F2 families tested. 

Mean performance of the F2 progeny from the high testing gametes was 
little different from the Fi for either ear moisture or yield (Table 24.13). 
However, the F2 progeny from the gametes of low yield performance ex- 
ceeded the Fi parent plant in yield performance on the average. This indi- 
cates that visual selection of plants within the F2 populations was more 
effective among the progenies arising from the gametes of low yield per- 
formance than for those F2 plants that were selected from high performing Fi 
crosses (gamete X inbred). 



TABLE 24.6 

DISTRIBUTION OF PER CENT MOIS- 
TURE AND YIELD OF 32 Fi PLANTS 
OFA25 X GOLDEN KING CROSSED 
TO A73 X A375* 



+2 
% Ear +^ 


1 
3 


2 


Moisture _. 
-2 


I 


8 
5 


5 
3 



1 +1 +2 



Yield 



♦Classes separated by one or more LSD (5 per 
cent) from the mean of A25Xtester. 



TABLE 24.7 

DISTRIBUTION OF PER CENT MOIS- 
TURE AND YII:LD of 38 Fi PLANTS 
OF A73 X MURDOCK CROSSED TO 
A25 X A334* 



+3 


1 








+2 










+ 1 




1 


1 




-1 




5 


7 


3 


-2 


1 


2 


9 


3 


-3 




2 


1 


2 



%Ear 
Moisture 



-2 -1 -fl +2 
Yield 

* Classes separated by one or more LSD (5 per 
cent) from the mean of A73Xtester. 



TABLE 24.8 

NUMBER AND SOURCE OF GAMETES SUPERIOR 
IN PERFORMANCE TO SAMPLER INBREDS 



Source of 
Gametes 


Sampler 
Inbred 


Total 

Gametes 

Tested 


Higher in 
Yield and 
as Early or 
Earlier in 
Maturity 


Not Dif- 
ferent in 
Yield but 
Earlier 


Morris 13 

8 inbreds 

Golden King. . . 
Murdock 


A344 
A344 
A25 
A73 


35 

8 

32 

38 


16 

2* 



8 


5 



8 

14 


Total.... 




113 


26 


27 









* An additional gamete higher in yield was also later in maturity. 

383 



TABLE 24.9 

PERFORMANCE INDICES OF TEST CROSSES OF 

SELECTED Fi PLANTS AND F. PROGENY 

FROM A25 X GOLDEN KING 





Perform.«ice Index 




Fi Plant 
Number 


1947 


1949 


Number 

OF Fz 

Plants 




Fi 


Fi 


F2 




19H*. ... 

20 H 

36H 

5L 

29 L 

46L 


+ 11 
+ 14 
+ 9 
-11 
-11 
- 5 


+ 19 
+ 9 
+ 16 
- 3 

- 1 

+ 1 


+25 
+ 14 
+ 11 
+ 5 
- 
+ 2 


5 
7 
7 
7 
1 
7 



H = high- and L = low-performing gametes. 



TABLE 24.10 

PERFORMANCE INDICES OF TEST CROSSES OF 

SELECTED Fi PLANTS AND Fj PROGENY 

FROM A73 X MURDOCK 





Performance Index 




Fi Pl.\nt 
Number 


1947 


1949 


Number 

OF Fa 

Plants 




F, 


F, 


F2 




12H 

14 H 

49 H 

50 H 

6L 

35 L 


+ 18 

+25 
+20 
+29 
-10 
- 4 


- 3 

+33 
+ 1 
+ 19 
-23 
-24 


- 6 

+27 
+ 5 
+ 18 
-10 

- 7 


8 
5 
6 
7 
5 
6 



384 



TABLE 24.11 

FREQUENCY DISTRIBUTION OF PERFORMANCE IN- 
DICES OF 1-2 PROGENY PLANTS FROM A25 X 
GOLDEN KING AROUND MEAN PERFORMANCE 
OF A25 X TESTER 



Fi Plant 


Indices for F2 Plants 


Number 


-15 


-S 


+ 5 


+ 15 


+25 


+35 


19 H 

20 H 

36 H 

5L 

29 L 

46 L 


1 


3 
1 
1 


2 

3 
2 

5 


1 

3 
4 

1 


2 
2 

1 


2 



TABLE 24.12 

FREQUENCY DISTRIBUTION OF PERFORMANCE 
INDICES OF F2 PROGENY PLANTS FROM A73 X 
MURDOCK AROUND MEAN PERFORMANCE OF 
A73 X TESTER 



Fi Plant 


Indices for ¥2 Plants 


Number 


-15 


-5 


+ 5 


+ 15 


+25 


+ 35 


12 H 

14 H 

49 H 

50 H 

6L 

35 L 


5 

2 

2 


1 

2 

2 
3 


1 

2 

2 
1 

1 


1 
2 
2 


1 
2 

2 


2 
1 



TABLE 24.13 

COMPARISON OF Fi PLANTS AND THEIR F2 PROGENY IN 1949 TEST 
CROSS PERFORMANCE FOR EAR MOISTURE, YIELD, 
AND PERFORMANCE INDEX 



Parents 


No. 
Plants 
Tested 


Ear Moisture % 


Yield in Bu. 


Performance 
Index 




F, 


F2 


Fi 


F2 


Diff. 


Fi 


F2 


Diff. 


Fi 


F2 


Diff. 


A25XG. KingH.. .. 

A25XG. KingL 

A73XMurdockH.... 
A73XMurdockL.... 


3 
3 
4 
2 


19 
15 
26 
11 


20.3 

23.3 
21.8 
22.8 


20.2 
23.8 
21.1 
22.4 


-0.1 
+0.5 
-0.7* 
-0.4 


54.0 
52.9 
56.0 
41.7 


54.5 
56.3 
55.0 
48.7 


+0.5 

+3.4** 
-1.0 
+ 7.0** 


+ 14.4 
- 2.1 
+ 12.5 
-23.6 


+ 15.8 
+ 3.3 
+ 9.6 
- 8.6 


+ 1.4 
+ 5.4 
- 2.9 
+ 15.0 



* Exceeds 5% point of significance. 



** Exceeds 1% point of significance. 

385 



386 E. L PINNELL, E. H. RINKE, AND H. K. HAYES 

DISCUSSION 

Almost 50 per cent of the gametes studied showed evidence of having 
combining ability in excess of the sampler lines. The gametes chosen as par- 
ents appear to furnish a desirable source of germ plasm for a selection pro- 
gram designed to improve the yield potential of A344, A25, and A73 in 
combining ability in specific crosses. 

Where a high combining varietal gamete is chosen for an inbred selection 
program, the Fi plant of which it is a parent represents a high X low type of 
cross so far as combining ability is concerned. To the extent that such Fi 
plants are comparable to crosses of inbreds, the breeding results should be 
similar to those from crosses of inbreds differing in combining ability. At 
Minnesota (Hayes and Johnson, 1939), crosses of highXlow combiners have 
given F5 lines ranging from high to low, but with a good proportion of high 

combiners. 

Whether selection of gametes should be followed by test controlled selec- 
tion in the F2 is an important question. In these studies more than 50 per 
cent of the F2 plants from high combining gametes tested at least ten per- 
formance index units higher than the sampler inbred. Thus without further 
test crosses, the chances of choosing high combining F2 plants would still 
have been very good. The number of Fo plants that could be handled in test 
crosses was quite limited. This may make for greater difficulty in recovering 
or improving the agronomic type of the sampler lines. It was very evident 
from field observations that the proportion of agronomically desirable F3 
lines appeared lower than usually found from crosses of highly selected 

inbreds. 

The greater effectiveness of visual selection among the F2 progenies of the 
low testing gametes is at this stage only an interesting development. Only a 
small proportion of the plants arising from the low testing gamete parents 
exceeded the sampler inbred in performance by a significant amount. 

It was not possible to determine by visual examination which F2 popula- 
tions were derived from high gametes and which from low gametes, although 
there were wide differences in plant type between populations. 

Gametes from eight inbred lines compared fairly well with varietal 
gametes from Morris 13, in offering promising sources of germ plasm for the 
improvement of specific combining ability of A344. Where a breeder has 
available large numbers of inbred lines of diverse origin the use of test 
selected inbred parents rather than varietal gamete parents may be the more 
feasible approach. Selection for characters other than yield would presum- 
ably be done more economically. The same advantage can be claimed for the 
use of complex crosses of inbreds. On the other hand, utilization of varietal 
gametes in improvement work does not "use up" inbreds so far as their 
combination in hybrids for commercial use is concerned. Lines recovered 



GAMETE SELECTION FOR SPECIFIC COMBINING ABILITY 387 

from crosses of inbreds may be more restricted than their inbred parents in 
commercial use because of relationship. It seems probable to the writers that 
the method of gamete selection is worthy of considerable use for further 
selection of material from open-pollinated, desirable commercial varieties. 
Studies of lines recovered from selected varietal gametes will have to be 
carried to F5 or later generations to determine if the large amount of out- 
cross testing is justified economically. The writers would like to emphasize 
the importance they attach to method studies of the type presented here. 
New ideas in breeding must be explored constantly if continued progress is 
to be made in corn improvement. 

SUMMARY 

Since 1945, a program has been underway at Minnesota to attempt im- 
provement of Minhybrids 602, 607, and 406 by the method of gamete selec- 
tion. The hybrid pedigrees are respectively: (A344XA334) (A357XA392), 
(A344XA334) (A357XA385), and (A25XA334) (A73XA375). Detailed 
studies of the non-parental single crosses among the inbred parents of each 
hybrid led to the conclusion that A344 in Minhybrids 602 and 607, and A25 
and A73 in Minhybrid 406 were low in combining ability. 

A344 was crossed to the Morris strain of Minnesota 13 and to eight in- 
breds of diverse origin. A25 was crossed to the Golden King variety and A73 
to Murdock. (Inbred X gamete) (tester) crosses were made using the oppos- 
ing single cross parents as testers. These were compared with the appro- 
priate cross of inbred X tester. Yield trial performance was obtained from a 
total of 113 gametes, 35 from Morris 13, 8 inbreds, 32 from Golden King, 
and 38 from Murdock. 

Sixteen gametes from Morris 13, three from the inbreds, and eight from 
Murdock gave significant increases in yield over the test crosses of the checks 
A344 and A73. Five gametes from Morris 13, eight from Golden King, and 
fourteen from Murdock were not significantly different in yield but were 
significantly earlier so that yield performance could be considered better 
than the checks on the basis of ear moisture at harvest. These varieties and 
the three high testing inbreds thus appear to be good sources of gametes for 
improving the relatively low performing inbreds in specific combining ability 
for yield. 

Both high and low testing varietal gametes were selected for use in a study 
of the development of new inbreds. From the crosses, A25X Golden King and 
A73X Murdock, selected F2 plants X the approj)riate tester were compared 
with the progeny of their Fi parental plants when crossed on the same tester. 
While there was excellent agreement, on the average, for combining ability 
in the Fi and F2, there was evidence of segregation for combining ability from 
almost all of the twelve Fo families which were studied. Visual plant selection 
within the F2 generations appeared to be effective in increasing yield per- 



388 E. L. PINNELL, E. H. RINKE, AND H. K. HAYES 

f ormance of the plants from the low testmg gametes, but appeared to have no 
effect in further increasing the yield performance of the F2 plants from the 
high testing gametes. 

The economic feasibility of Fo plant testing in a gamete selection program 
is questioned. 



SHERRET S. CHASE 

Iowa Sfafe College 



Chapter 25 



Monoploids in Maize 



Haploid sporophytes have been reported in jimson weed (Blakeslee et al., 
1922), cotton (Harland, 1920), tobacco (Chipman and Goodspeed, 1927), 
evening primrose (Gates, 1929), maize (Randolph, 1932a, 1932b), wheat 
(Gaines and Aase, 1926), rice (Ramiah et al., 1933) tomato (Lindstrom, 
1929), pepper (Christensen and Bamford, 1943), and in many other genera 
which have been subjects of cytogenetic study. 

A haploid organism, strictly speaking, is one which has only one set of 
chromosomes, that is, one genome per cell. In the common usage of botanists, 
geneticists, and others, a sporophy te originating by reduced parthenogenesis 
or by an equivalent process, and consequently carrying the reduced or gamet- 
ic complement of chromosomes in each cell instead of the normal zygotic 
complement, is referred to as a haploid. 

Thus the term, as applied to a sporophyte, has come to carry the connota- 
tions of both parthenogenetic origin and gametic chromosome number, and 
the actual genomic condition tends to be ignored. Many so-called haploids are 
actually diploids or polyploids. Thus the haploids of common wheat are 
triploids, since the parent species, Triticum vulgare, is a hexaploid. To em- 
phasize the fact that the haploids of maize carry only one set of chromosomes 
per cell, that is, only one chromosome of each type instead of the normal pair, 
the alternate term monoploid is used here to designate these aberrant plants. 

In normal sexual reproduction in maize the pollen tube penetrates the 
eight nucleate embryo sac. One of the two male gametes released fuses with 
the egg nucleus to form the zygote, while the other fuses with the two polar 
nuclei to form the primary endosperm nucleus. In the abnormal type of re- 
production giving rise to monoploid sporophytes the ])rocesses apparently 
are the same except that for some reason the first male gamete fails to fuse 
with the egg nucleus and is lost. The egg nevertheless is activated and devel- 

* Journal Paper No. J-1906 of the Iowa Agricultural Experiment Slalion, Ames, Iowa. 
Project 1201. 

389 



390 SHERRET S. CHASE 

ops into an embryo. Evidence for this is indirect — monoploid embryos are 
found in kernels having normal (3n) endosperm. It is possible that some or all 
monoploids arise from reduced cells of the embryo sac other than the egg, 
from the synergids perhaps. 

As tools for experimental research monoploids offer many possibilities: in 
the cytological field for studies of the meiotic distributions of unpaired chro- 
mosomes, non-homologoiis synaptic relations of the chromosomes and me- 
chanics of chromosome doubling; in the genetic field for direct observation of 
mutational effects, measurement of mutation rates, studies of cytoplasmic 
effects, and biochemical investigations; in the agronomic field for the produc- 
tion of diploid, homozygous stocks directly from the monoploids. The follow- 
ing discussion is concerned primarily with my own investigations into the 
latter possibility. 

A monoploid carries in each of its cells, or nuclei, only one chromosome of 
each type. Thus if the chromosome complement of any cell can be doubled, 
the affected cell and any derivative of it will consequently be both diploid and 
homozygous. If such homozygous diploid sectors include the reproductive tis- 
sues, meiosis should then be normal and the gametes produced functional. 
Thus such plants can produce diploid progeny — homozygous diploid progeny 
if the individual is successfully self pollinated — since every gamete of the 
plant is genetically equivalent to every other gamete. In a monoploid without 
diploid sectors, since the chromosomes lack synaptic mates, meiosis is highly 
irregular. Only rarely are functional gametes carrying the full complement of 
chromosomes produced. If two of these rare functional gametes from a single 
monoploid do fuse in syngamy, the zygote produced will be diploid, and 
homozygous, unless the gametic chromosomes were subject to chromosomal 
aberration during the irregular meiosis. 

Production of homozygous diploid progeny from monoploids results in the 
fixation of a single gametic complex. In any population, desirable gametes are 
more frequent than desirable zygotes. Foi example, if one has on hand an 
individual heterozygous for three pairs of genes and wishes to obtain from it a 
definite homozygous product by selfing, one individual in sixty-four of the im- 
mediate self progeny (Si) will, on the average, carry the desired genotype. 
One gamete in eight, extracted as a monoploid and then converted into a 
homozygous diploid, will furnish the same genotype (see Fig. 25.1). 

Successful production of homozygous diploids in quantity from mono- 
ploids depends upon the adequate solution of two main problems. The first 
of these is the production and recognition in the seedling, or in earlier stages, 
of large numbers of monoploids. This problem has been solved to the extent 
that thousands of monoploids can be produced with relatively small expendi- 
ture of effort. The second problem is that of deriving self, and consequently 
homozygous, diploid progeny from the monoploids isolated. This problem 
also has a practical, though partial, solution. 



MONOPLOIDS IN MAIZE 



391 



MONOPLOIDS IN MAIZE 

It has been known for some time that monoploids occur naturahy in maize 
in measurable frequency. Data of Randolph (Randolph and Fischer, 1939) 
and of Einset (19