~-,

u c
-
o A
m...“
.4“.

»v

    
   

I ,
._.- t I I u
:5 ~51. I; c .‘ 3:2 , Inn!
1, .I ~ » 3g 44...}... .1» L I
.. I' ”II“ -. :sz 33“.:7il;:':‘l;‘..2, ..; x‘.':l.";II"lI.;.". J2" ' ~I - ..
‘.‘ . "I '7 I I'LTI" . -‘ ' ,‘ ' v. _’
'II\(-. I r..t‘1'IIII'I':II~”II ., I.‘- . ' LI . '
. . ‘. 5"} : (IIIEI; . H ‘. ., .(. ‘.I"‘:I:"' l_ "I=,-?I “ . H: . . I III.-

     

     

UH'I -. ". ‘WIV 1"}. ”'1': -. . I .-. , ..' --
PHI?“ 'I‘le - . ’ II I I - In .‘ .. 3.1" ’1. I I- r . :hzan “MI“.

    

_I ‘ ' " :I:I'II‘:"I" ‘ 1' i I" ' ‘1. I1)" :1.” 5n "4":
"if. I .,....IIIII, '. I; III I“. W"- "I'. ”.1 I’ISI'IIII'S'IaniL'I-é" .I
u!

I.
I
II‘ '. " III
E'IIIII III" '.. IIII. .IIIIIIIII ...’5III. ‘IIIIIIIAII IIJII'IIIIMI‘I-I’I-H'5'1'.II" P—fim'WL‘Jm-II

 

 

A. Ill“lllllll'llllllll‘ll lllllllll

LIBRARY 1293 10200 908
Michigan State

Umvernty

ll

This is to certify that the

thesis entitled

An Analysis Of Selected Breeding Approaches
For Oats And Barley

presented by

James Laurence Nelson

has been accepted towards fulfillment
of the requirements for

Doctor of Philosophy degmmin Plant Breeding

 

 

 

Major pro 501'

Date-Wfi/gll /Cig/

0-7639

 

ov'snoug nuts:
25¢ per day per item

RETURNIfi LI§RARY MATEIQALSL
Mace in book return to mauve
charge from circulation records

 
        
 

  

"All! "‘-\\\\ s '

V‘ : x- .un‘illll ‘

 

 

AN ANALYSIS OF SELECTED
BREEDING APPROACHES FOR
OATS AND BARLEY

By

James Laurence Nelson

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Crop and Soil Sciences

1981

ABSTRACT

AN ANALYSIS OF SELECTED BREEDING
APPROACHES FOR OATS AND BARLEY

by

James Laurence Nelson

Oats (Avena sativa) and barley (Hordeum vulgare) are studied

 

 

by different means with respect to the value of the approaches to
plant breeding. Chapter one includes an analysis of twenty—one

oat genotypes grown in common in four locations for the years 1977
through 1979. Yield and the primary components of yield, as well

as other agronomic data, were recorded and used to construct a new
index of stability and superiority. Additionally, the oat population
was studied by means of path coefficient analysis and three potential
population parameters were noted. Chapter two contains an analysis of
the oat variety Heritage analyzed by means of regressing the primary
components of yield for Heritage against the mean components for the
population of adapted cultivars. Seed number per panicle, Y, is the
component which establishes Heritage as a superior variety. Chapter
three tests the hypothesis that mean seed number per unit area, XY,
for pure lines with respect to each other, is an effective predictor
of the relative frequencies of each genotype after one generation of
growth in bulk mixtures. Three barley genotypes and two oat genotypes
constituted the pure lines and the bulk mixtures. The hypothesis was

rejected and the explanation was differential seedling vigor in the

early stages of growth for several of the genotypes. Chapter four
presents a method for parental selection such that two populations
with similar yields can be seperated by regressing one yield component
against another. Outliers are thus indentified as lying off the
primary regression line and these genotypes may then be used in
crossing combinations which produce unselected progeny whose grain
yield exceeds that of each of the highest yielding parents. Chapter
five details the work to date on the wide hybrid between H. vulgare
and H. Qubatum with respect to the introgression of wild germplasm
from H. jubatum into the genome of H. vulgare. Additionally, the
results of a series of experiments, intended to increase the frequency
of haploid sectors of H. vulgare for the hybrid, are reported. A
chemical called griseofulvin was demonstrated as inducing a higher
grequency of haploidization -- both for H. vulgare and for H. Jubatum --
from the hybrid. Griseofulvin was shown not to induce the same pheno-
menon in six diploid cultivars of barley. Therefore, an hypothesis

was proposed that a degree of genomic instability is necessary for
griseofulvin to induce sectoring and that this sectoring is a function

of whole genomes and not of individual chromosome elimination.

This dissertation is dedicated to

L. S. Nelson, M.D., F.A.C.S., F.I.C.S. and
L. S. Nelson, Jr., M.D.

11'

ACKNOWLEDGEMENTS

The author is deeply grateful for the assistance provided to him
by Drs. Peter Carlson, William Tai, M. N. Adams and especially by
Dr. Dale Harpstead in the drafting of this dissertation and in the
execution of the research herein contained. Additionally, the author
wishes to thank Mr. Dimon Wolfe for his considerable contributions as
field boss. Mrs. Brenda Floyd and Dr. Robert Griesbach are contribu-
tors to this document through their friendship and suggestions.
Lastly, I wish to acknowledge the guiding force behind this disserta-
tion and the disparate, yet substantial, research reported within:

the late Dr. John E. Grafius.

PREFACE

This dissertation is appropriately divided into five chapters.
Each is sufficiently different in terms of the research it reports to
merit such a division. Chapters 1 through 4 are related insofar as
the common thread running current throughout them is the use of the
primary components of yield to study various phenomena.

This is appropriate in view of the unique set of yield component
data which, but for the author, might well have remained unused. As
the last graduate student of the late Dr. John E. Grafius and, to a
considerable extent, his spiritual heir, it seemed the responsibility
of the author to resurrect some of these data and apply them to con-
temporary problems in plant breeding. There is much, however, which
remains to be done.

Chapter 1 introduces yield components and their possible use in a
new index of stability and superiority in oat populations. This index
is complemented by a path coefficient analysis of the same oat popula-
tion with yield components. Chapter 2 introduces a slight modification
of simple linear regression with yield components to assess the stabil-
ity and superiority of performance of Heritage, a new oat variety.
Chapter 3 attempts to relate the expression of yield components in pure
stands of oats and barley to the performance of the varieties when
grown as bulk populations. Chapter 4 details the use of yield compon-

ents in parental selection for higher yield. These chapters will,

iv

possibly, suggest the many uses of yield components, ranging from
studies of the genotype—environment interaction (Chapter 1) through
their effective manipulation in improving yield.

Chapter 5 is an anomaly only in the context of this dissertation.
Of all the chapters, it represents the greatest effort on the part of
the author and it represents his original research topic. The methods
used in attempting to transfer genetic variability from a wild species

to Hordeum vulgare and the results of these efforts are reported.

TABLE OF CONTENTS

LIST OF TABLES .........................
LIST OF FIGURES ........................
CHAPTER 1: AN ALTERNATIVE APPROACH TO ESTIMATION OF

STABILITY AND SUPERIORITY OF A COOL SEASON
OAT POPULATION ...................

Introduction .......................
Literature Review .....................
Materials and Methods ...................
I. Stability Index .................
II. Path Analysis ..................
Results and Discussion ..................
I. Stability Index .................
II. Path Analysis ..................

APPENDIX 1. Notation for the Eleven Year-Sites ........
APPENDIX 2. Notation for the Twenty-One Genotypes .......
CHAPTER 1--BIBLIOGRAPHY ....................

CHAPTER 2: ANALYSIS OF A NEW MICHIGAN OAT VARIETY BY

YIELD COMPONENT REGRESSION .............
Introduction .......................
Literature Review .....................
Materials and Methods ...................
Results and Discussion ..................

APPENDIX 1. Correlational Matrix for Site and Heritage and

Their Associated t-Statistics (r=O) ........

CHAPTER 2--BIBLIOGRAPHY ....................
CHAPTER 3: OAT AND BARLEY COMPOSITE EXPERIMENTS ........

Introduction .......................
Materials and Methods ...................
Results and Discussion ..................

CHAPTER 3--BIBLIOGRAPHY ....................

vi

Page

viii

m-bthi—a

10
10
23
33
34
35

36
36
37
38
4O

50
51

52
52
56
58
7O

Page

CHAPTER 4: THE USE OF OUTLIERS IN BREEDING FOR YIELD

IN BARLEY ..................... 71
Introduction ....................... 71
Materials and Methods .................. 72
Results and Discussion .................. 75

APPENDIX 1. Notation for Genotypes of Possible Parents,
Parents and Bulk Progeny ............ 88

APPENDIX 2a.-
2j. Analysis of Variance Table for Randomised
Block Design .................. 89-98

CHAPTER 4f-BIBLIOGRAPHY .................... 99

CHAPTER 5: CHEMICAL INDUCTION OF GENOME ELIMINATION IN
SOMATIC TISSUE OF A WIDE HYBRID WITH BARLEY . . . . 101

Introduction ....................... 101
Literature Review .................... 102
Materials and Methods .................. 104
Chemical Induction of Chromosome Elimination ..... 104
I. Chloramphenical (CAP) .............. 104
II. Para-fluorophenylalanine (PFP) ......... 105
III. Griseofulvin .................. 105
IV. 2 Percent DMSO (Control) ............ 106
V. Solubility of Griseofulvin in a 2 Percent
DMF Solution .................. 106
VI. Griseofulvin Treatment of Six Diploid

Cultivars of Barley ............... 107

VII. DMF Toxicity Study in Barley and a Wide
Hybrid with Barley ............... 108
Results and Discussion .................. 108
CHAPTER 5--BIBLIOGRAPHY .................... 115

vii

LIST OF TABLES

Page
CHAPTER 1
Table 1. Example of Data Transformation for Heritage Oat . . . . 7

Table 2. Comparison between Genotypes and Between Year-Sites
for 01k, 6i and OOi ................ 13

Table 3. Comparisons of Intravarietal Component Correlation
Coefficients between Non-Transformed and
Transformed Data. Intravarietal Correlation

Coefficients: r ................. 16
Table 4. Comparison between Genotypes and Between Year-Sites

for Vector Lengths, Wik and W1 ........... 17
Table 5. Intervarietal Paired Observations for Stability

and Superiority .................. 22

Table 6. Primary Paths and Correlation Coefficients from
Transformed Data .................. 24

Table 7. Mean Yield (Ni) and the Three Genotypic Components
of the GxE Interaction for the Twenty-One
Genotypes ..................... 26

Table 8. Three Environmental Components of the GxE Inter-
action for the Eleven Environments, the Coeffi-
cients of Determination for the Yield Equations
and the Mean Fitted Yields in Standard Deviation
Units ....................... 28

CHAPTER 3

Table 1. Component and Yield Data for Oat and Barley Pure
Lines and Bulks, Where W = Yield in Grams per
Plot, 2 = Seed Weight and XY = Seeds per Plot . . . 59

Table 2. Mean Square of XY (Seeds/Plot) Estimates of Orbit
and Hulless Terra Grown as Pure Lines and of
W (Yield) ..................... 61

Table 3. Mean Squares of XY (Seeds/Plot) Estimates of Coho,
Red Lemma Titan and Hulless Vantage Grown as
Pure Lines and of W (Yield) ............ 62

Table 4. Estimates of Spike Weights for Coho, Red Lemma
Tital and Hulless Vantage as Determined by
Sampling Pure Line Plots .............. 65

viii

Table 5. Chi Square Test of Significant Differences between
XY Estimated and Observed Values in Oats and

Barley ....................... 67
CHAPTER 4
Table 1a. Table of Unadjusted Means ............... 76
Table 1b. Table of Unadjusted Means ............... 77
Table 1c. Table of Unadjusted Means ............... 78
CHAPTER 5

Table 1. Frequency of Haploid Sectors Arising from VJJ' Clones
Treated with CAP, PFP, Griseofulvin and DMSO . . . . 109

Table 2. Contingency Analysis of Griseofulvin versus Controls
for Haploid Sector Induction in the Hybrid VJJ' . . 110

ix

LIST OF FIGURES

CHAPTER 1

Figure 1. Example of Minimum, Maximum and Optimum Traits
Regressed Against the Site Mean Yield (W)
for a Hypothetical Oat Variety ........... 2

Figure 2. Methods Used for the Calculation of the Transforma-
tions and Associated Statistics Used in the
Stability Index .................. 6

Figure 3. Non-Linear Relationship between r and C05’ r ..... 9

Figure 4. Diagram of Path Coefficients from the Primary
Components of Yield (X,Y,Z) to Yield (W) and
from Environmental Resources (R R R ) to the

Components. (Tai, 1975) . . . i’.2: ? ....... 11
Figure 5. Yield Model Based Upon Path Analysis (Tai, 1975)
and the Modification Proposed by Nelson ...... 12

Figure 6. Intravarietal Distribution of Vectors for Heritage
for Eleven Year-Sites (Length = Yield in

Standard Deviation Units) ............. 20
Figure 7. Predicted Mean Yields (Standard Deviations)

Plotted Against Actual Mean Yields ......... 29
Figure 8. Residuals Plotted Against the Twenty-One Genotypes . . 30
CHAPTER 2
Figure 1. Regression of Heritage X onto Site X ......... 41
Figure 2. Regression of Heritage Y onto Site Y ......... 43
Figure 3. Regression of Heritage XY onto Site XY ........ 44
Figure 4. Regression of Heritage 2 onto Site Z ......... 45
Figure 5. Regression of Heritage W onto Site W ......... 47
Figure 6. Regression of Heritage TW onto Site TW ........ 48
CHAPTER 3

Figure 1. Change in Composition of a Four Component Barley
Bulk Population (Data from Suneson, 1949) ..... 53

CHAPTER 4

Figure 1a.
Figure 1b.

Figure 2a.
Figure 2b.

Figure 2c.
Figure 2d.

Figure 2e.

Figure 3.

Regression of Z onto Y for Possible Parents .....

Regression of 2 onto XY for Possible Parents
and Their Respective Yields ............

Parental and Bulk Progeny Performance for Yield . . .

Parental and Bulk Progeny Performance for Tiller
or Spike Number ................

Parental and Bulk Progeny Performance for Seed
Number per Spike ................

Parental and Bulk Progeny Performance for Seed
Weight ......................

Parental and Bulk Progeny Performance for the
Combined Trait, YZ ................

Regression of 2 onto XY for Parents and Progeny
Grown in a Common Experiment ...........

xi

.8319.

73

74
79

81
82
83
84

86

CHAPTER 1

AN ALTERNATIVE APPROACH TO ESTIMATION OF STABILITY
AND SUPERIORITY OF A COOL SEASON OAT POPULATION

Introduction

 

It is the author's view that all agronomic traits may be classified
as either maxima, minima or optima, for the purpose of plant breeding.
Yield components and test weight in cereals are examples of maximum
traits, disease and lodging are minima and the various quality factors
(malt and milling quality) are optima, for which some range of values
exist, and within which the optimum traits should lie. These classifi-
cations are not rhetorical since they clarify breeding goals. In this
light a plant breeder does not want absolute yield stability in a
variety, for the consequence of this would be a variety buffered against
all environmental variables, including those which would maximize yield.
Simultaneous with breeding for maximum traits, the breeder seeks to im-
pose through genetics a minimum expression of negative traits, of which
disease is a ubiquitous example. Yet, over the range of environments,
the breeder seeks to stabilize the expression of the many quality fac-
tors (optima) which are necessary constituents of a good variety.

Figure 1 expresses the three categories as a function of an independent
variable, site W (yield), for a hypothetical oat variety.

It had been suggested to the author by Dr. M. W. Adams that stabil-

ity might also be broken into components, most logically those of yield.

FIGURE 1. Example Of Minimum. Maximum and Optimum Traits
Regressed Against The Site Mean Yield (H) For
A Hypothetical Oat Variety.

___ Yield (Maximum)
m.-- Milling Quality (Optimum) /'

-._.. Lodging (Minimum) I

   

ba 1 regression line

Varietal

Trait

  

Site Mean Yield

A normalizing transformation would standardize the traits and the corre-
lation between the sets of yield components grown in different year-
sites might offer some statistical measure of stability. It then
occurred to the author that a combination of the two ideas, trait classi-
fication and the concept of stability as a function of multiple compon-
ents, might be effected and tested on an appropriate data set.
Additionally, the author desired to test another measure of stability,
that resulting from path anlysis, and the yield model proposed by its

author, Dr. George C. C. Tai.

Literature Review

 

It is not the purpose of this paper to present an exhaustive review
of the many methods utilized by plant breeders to adjudge stability and
superiority. In general, however, two watershed papers (1,2) presented
methods which allowed plant breeders to compare many varieties over a
range of locations and to establish some measure of statistical confi-
dence in their judgements. Aside from the regression papers of Finlay
gt 31. (1) and Eberhart gt 31. (2), other parameters have been suggested
(3,4) which relate to stability of yield performance.

A possible limitation to all the preceeding methods is their reli-
ance upon yield alone, which, as Grafius has shown, is an artifact
resulting from the multiplicative interactions of its primary components
(5). Just as yield has been broken into its components, perhaps stabil-
ity, properly expressed as a function of components, might better

explain varietal behavior in differing environments.

Path coefficient analysis does not provide direct measures of
varietal superiority. In the agronomic literature, Dewey and Lu (6)
first applied path analysis to yield components, albeit somewhat con-
fusingly. Subsequently, Duarte and Adams (7) used path analysis to
analyze both primary and secondary yield components in Phaseolus.
Eventually, additional literature appeared which tied together the pro-
mise of path analysis with a more thorough understanding of yield com-
ponents. Based upon the proposal of the sequential development of yield
components in cereals (8), Tai (9,10) developed a method for resolving
the individuals paths. Additionally, he proposed a model for yield,
expressed in standard deviation units. Finally, Hamid and Grafius (11)
proposed a developmental allometry for barley which was later modified
by Grafius (12). These papers represent the entirety of literature
familiar to the author which bears upon path analysis and crop yield

expressed through its components.

Materials and Methods

 

I. Stability Index

 

The stability index which the author proposes utilizes only maximum

traits and, for this data set, it includes the following traits:

X = panicles/ft2
Y = seeds/panicle
Z = weight/seed
W = weight/seed

TW = pounds/bushel

It is of interest to note that the set of maximum traits could be

augmented with minimum traits, were the latter expressed as a differential
between some upper limit (i.e., 100%) and their actual value (the author
appreciates this suggestion from Dr. M. W. Adams).

The use of maximum traits affords both a measure of stability and of
superiority. The measure of stability is the standard deviation of the
vectors and the superiority lies in their length. The transformation
can be found in Figure 2. In using this transformation, which is actually
the Z transformation for normality, the different variables are converted
to common units (standard deviations) with unit variances and means of
zero. An example of non-transformed and transformed data are to be
found in Table 1 for Heritage, entry number 3, of the twenty-one geno-
types.

The data were derived from rod row oat experiments grown in four
locations over three years, 1977-1979. Twenty-one genotypes were grown
in these common experiments and one location was lost--East Lansing in
1979. The experiments were planted and analyzed as 5 x 5 square lat-
tices with four additional varieties which varied from year to year.

The entries were planted in four rows, eight feet in length, with eleven
inch row spacings. Each entry was replicated four times. The following
data were recorded: X (panicles/ftz), Y (seeds/panicle), Z (seed
weight), W (yield in bu/a), test weight (lbs/bu), height, lodging and
heading date. These data were tabulated and means over locations each
trait were calculated. Figure 2 details the transformation and the
various statistics associated with the data.

The assumption behind this stability index is that varietal stabil-
ity might be expressed through the correlation of a number of traits

with the mean of each trait over year-sites. Referring to Table 1 and

FIGURE 2. Methods Used For The Calculation Of The Transformations
And Associated Statistics Used In The Stability Index.

Transformation: Eijk - Eij

 

Where Eijk= jth trait of the ith genotype in the kth
year-site,
Eij = mean of the jth trait of the ith genotype
over year-sites,
Oij = standard deviation of the jth trait of the
ith genotype over year-sites,
i = genotype (1-21),
j = trait (1-5 for X,Y,Z,W and TW, respectively),
k = year-site (1—11) (see Appendix 1).

Statistics: Uij = trait mean of the ith genotype over year-sites,

r(ijk)(U].J.)= correlation coefficient for traits 1,2,3 and 5
or rik = (n=4) between individual year-sites and trait means
over year-sites for the ith genotype (W is
excluded since XYZ=W),

=-1=..
Oik cos rik direction of vector

2:
ll

ik yield of the ith genotype in kth year-site,

length of vector in standard deviation units,

standard deviation of the ith genotypic vector.

Q
II

01

TABLE

7

1. Example Of Data Transformation For Heritage Oat.

Non-transformed

 

 

X= panicles/ftz, Y= seeds/panicle, Z= weight/seed, TW=

Year-Site

Trait 1 2 3 4 5 6 7 8 9 10 11 Uij

X 19.5 18.0 18.3 9.1 17.8 14.9 16.6 16.9 16.1 10.4 9.6 15.2

Y 73.8 84.6 87.3 70.1 88.4 84.6 97.7 71.3 79.2 105 90.6 84.8

2 (mg) 32.6 29.4 29.7 29.5 29.4 32.6 32.3 32.5 34.4 32.8 34.9 31.8

W 141 133 138 57 148 134 171 127 130 108 92 125

TH 35.1 33.0 32.6 30.6 33.6 36.0 34.8 31.8 34.8 33.3 37.3 33.9

HT (in) 35.0 35.8 40.2 29.7 43.9 36.4 44.5 34.0 37.1 42.2 36.4 37.7

LD (%) 8 25 2 --- 20 --- O --- --- 25 --- 13

HD 18 -- --- --- 28 --- -- —-- --- --- 24 23

Transformed
Year-Site

Trait 1 2 3 4 5 6 7 8 9 10 11 Uij
X .26 .60 .13 -.94 -.04 -1.1 - 38 89 1.38 -.88 -.81 -.12
Y 1.0 .82 .90 .38 1.05 1.72 1.41 .13 -.21 .53 1.00 .89
Z .10 .04 .14 O -.69 .44 .04 .32 .57 .19 .46 .21
W 1.35 .52 .26 -.62 .75 1.37 1.59 1.35 1.04 -.10 .50 .47
TN .22 .06 .12 .19 -.74 .57 85 .08 1.24 -.50 1.04 .19
HT -.73 .37 .49 -.49 -.83 -.94 13 O -.26 -.7O -.59 .23
L0 .71 .22 .45 --- .22 --- -1.04 --- --- .45 --- 1.08
HD 1.12 -- --- --- .94 --- --- --- --- --- 1.25 .59

test weight (lbs/bu),

HT= height in inches, LD= lodging and HD= heading date (days after 6/1).

Figure 2, this would involve calculating a pair-wise correlation
coefficient, rik’ for each year-site with the mean. This would generate
eleven correlation coefficients whose distribution is a reflection of

stability. This distribution can be represented graphically if one
1 r
ik‘

Figure 3 represents 0 as a function of rik' Any variety in this study,

converts the correlation coefficients to angles, such that O = cos-

then, has eleven vectors with assigned directions (9). Each vector can
be accorded a length by assigning the respective transformed yields,

wik' These two data comprise the intravarietal stability and superior-
ity measures since the more closely the vectors are clustered, the more
stable their phenotypic expression. Their superiority is a function of
their length and this allows comparisons between years and between loca-
tions. Having calculated these vectors it is possible to contrast
varieties by comparing tabular values for theta (01k), the respective
lengths of the vectors (wik)’ and the standard deviation for the geno-
typic Qik (Oei)' A set of correlation coefficients between the genotyph:
means and the mean over genotypes cannot be calculated since this compar-
ison, or grand, mean is necessarily zero as a consequence of the normal-

izing transformation.

II. Path Analysis

 

The two underlying assumptions behind the path analysis used in
this study are, first, that yield (W) is the product of the multiplica-
tive interaction of the three primary components of yield, X, Y and Z
and, secondly, the allometric plant development hypothesis of Grafius
(8,12) is correct. These assumptions first allowed Tai (9) to develop

a path coefficient system in which the paths are resolved by means of

-l
C<a$ 7

(degrees)

 

I50

IZCJ

SC)

SC)

30

-L0

FIGURE 3. Non-linear Relationship Between r And Cos-

-.8 -.6 -.4 -.2 O .2 .4

1

r.

LO

10

simultaneous equations. Tai proposes that for any year and site there
are essentially three different environmental resource groups, Rijk
(j = 1-3), which are exploited with varying degrees of efficiency by

each genotype, through that genotype's yield components over time. Thus,
paths from each resource may be calculated algebraically from the path
coefficients connecting the yield components to one another and to

yield. Figure 4 demonstrates the relationships.

Tai (9) also proposed a yield equation, expressed in standard devi-
ation units, which is composed of some value for mean yield of a genotype,
three genotypic components derived from the respective paths, three
environmental components associated with each genotypic component and
an error deviate. Figure 5 details this system and the author's modifi-
cations. The author proposes that the grand mean and grand standard
deviation (over genotypes and over year-sites) be substituted for Tai's
genotype-specific values for the same statistics. The oat data set is
more powerful than Tai's and, therefore, allows some estimates of popu-
lation parameters associated with cool season oats. The use of the
grand standard deviation relates each genotype to a population mean and
therefore, affords comparisons between genotypes with respect to their
genotypic strategies for exploiting the three groups of environmental
resources. The solution to the yield equation is by least squares.

The same data set used in the stability index was used in the path

analysis.

Results and Discussion

 

1. Stability Index

 

Table 2 contains the Oik (eleven each for each of the twenty-one

11

FIGURE 4. Diagram Of Path Coefficients From The Primary Components
Of Yield (X,Y,Z) To Yield (H) And From Environmental
1,R2,R3) To The Components. (Tai, 1975)

Resources (R

 

l’xv = 0n

rxz = 02:00:03

rvz = Orland:

raw = 04¢0i05 *Oade‘tal 0300

run 8 0590I04*asalfialatat
r31. == <lci"ll<'d>{Pcm3<15'.CII‘13‘34I‘F<II‘12‘3!5
u = t I I

l 2 ’2
ll: ==1£ ( I " (ll) )3:
u: =1 (i- aux-z- can")

FIGURE 5.

Tai:

Nelson:

Where

12

Yield Model Based Upon Path Analysis (Tai, 1975) AHd The

Modification Proposed By Nelson.

 

Li.
A
H
I
(A)
v
I

”11‘ Uwi I Vlirlj I V2ir2j I V3ir3j I Eij

Owi
”i ' Uw I Vlirlj I V2ir2j I V3ir3j I Eij

0w

”i I ”W I Ow (Vlirlj I v2ir2j I V31I3j) I Eij,

wij = yield of the ith genotype in jth environment,
Uwi = mean yield of the ith genotype over locations,
°wi = standard deviation of yield of the ith genotype over

locations,

<
II

11 genotypic component for yield component, X,

u1(a4 + a a + a a + a a3a6)

1 5 2 6 1
v21 = genotypic component for yield component, Y,
= u2(a3a6 + a5)
v31 = genotypic component for yield component, Z,= u a

3 6
' environmental components associated with the geno-

typic components,

W, = yield of the ith genotype over all locations,
Uw = mean yield of all genotypes over all locations,
Ow = standard deviation of yield of all genotypes over I.
i = genotype (1-21),
j = location (year-site) (1-11).

13

 

m.“ m.o~ ¢.em ~.mfi m.m~ o.hm m.a~ m.e~ m.~m m.- “.mfi m.NN o.Hm one;
m.m ¢.o~ o.HN w.m~ H.~N m.m~ m.» m.H~ m.oH w.cm o.m «.mfi o.o~ comcmm
H.“ c.- e.m~ m.NN ~.om m.~ m.mm 4.8N m.mm m.m~ ~.m~ o.om m.o~ zmc_xumz
e.o~ ~.~¢ o.om m.- H.mm e.w~ ~.ee e.m~ m.mm o.om o.me m.¢~ “.mo peacoa
e.w~ 5.8e o.m~ o.~m m.oc 0.8m ~.¢_ ~.mm 4.4m H.o~ N.mm m.~¢ ~.om mace:
o.m o.¢~ N.- “.8” ¢.m~ o.m~ H.mH m.m~ m.~m ~.om «.mm e.- ~.A~ cmcwcez
«.mm e.mm ~.~m m.mm N.wm w.o~ m.mm o.w~ m.o~ m.m~ o.m¢ m.Hm ~.w~ mPnomaa
m.mH w.om “.mfi o.m~ ¢.~m 8.3H m.mm m.om e.HH ~.oc m.Hm e.mH ¢.He scene
m.m~ m.mm m.mo_ 8.8H m.m~ H.4H w.Nm o.NH a.em m.mH m.~m m.mm m.~e ede---mo
m.A~ e.em “.om “.mm m.m~ ~.Hm N.mm m.m¢ m.HH N.HH “.mm m.om o.Hw coozcox
m.efi e.HN H.om m.H~ o.fl e.HN m.~m m.mm H.m~ «.mm ~.m m.w_ m.- “Coco
m.mm m.m~ m.mm~ «.ma o.wHH H.Amfi o.m- m.HeH H.o~ A.MN ¢.o~ ~.NH m.m~ moe---mo
m.me ~.om m.ee ~.mm o.omH H.mmH A.N~ m.o~ N.me m.~m m.o~ ~.m m.w~ aaauwcmz
~.- c.8e N.¢w m.oH m.Hm o.oo e.He m.oo m.NN m.mm w.om ~.o~ m.m~ mmcwsocmz
m.AH w.- o.~o H.m o.mm m.HH H.eH m.mm H.NN o H.mN H.m 8.H~ mam---mo
li_Qw To #3 OH 0 m A o m e m N H hyped
ap_m-cma>

on uc< we .xwo com mmuwm-cmm> :mmZHmm nz< mquuocmu cmmzpmm comwcmaeou .N ubm<p

l4

 

o.- o.mm H.mm ¢.mN A.m¢ m.mm o.~¢ m.wH m.mm N.m c.8e N.mN «.8m p;u_c3
¢.NN m.me A.H¢ H.m N.mm m.o~ m.me m.m¢ o.mm H.om N.AH H.m~ o.~m em epu
m.mm o.me ¢.HH m.mH 8.5m m.oo~ o.eo o.NH w.ee m.om m.oe m.eH m.e~ .mo
G.HN N.~m m.e~ 9.3m o.o~ N.¢H 8.x“ ~.He 8.5m e.H~ ~.em ¢.- m.om m_noz
N.mH 8.8m o.um o.mN o.~o m.mN m.mN 5.5m o.mH ~.me o.mm ¢.mm m.mm e--w~-mo
«.mfi «.mm e.- H.~m m.oe m.~¢ m.mm H.eN ¢.oH e.em m.m~ m.¢¢ H.~e cospmcaz
_oo _o H“ OH o w a o m e m N H xgpcm
mu. wmime>

.umzcwpcoo .N udm<A

15

genotypes), their means over year-sites (9i) and the standard deviations
(Oei) associated with each set of eleven 91k. The 91k represent the
divergence of the series of maximum traits, properly transformed, in
each year-site, from the mean over year-sites. It is apparent that
there is considerable variation, or plasticity, in a genotype's ability
to form yield since the mean degree of correlation is low (high Oei) in
many of the higher yielding genotypes. There would appear to be no
correlation between stability as determined by this system and the ori-
gin of the genotypes, for the stabilities of Michigan lines range from
7.1 for Mackinaw to 59.3 for genotype 69-27-403.

It is of interest to note what changes are exacted through the use
of the transformation. A sample of genotypes and their non-transformed
values were correlated for the eleven year-sites in a manner identical
to that used on the transformed data set. With only a few exceptions,
the Rik values were exclusively 1.0. The exceptions were few and none
were less than Rik = .99. By imposing the conditions of a mean of zero,
unit variance and common units of measure (standard deviations), it is
obvious that the integrity of the genotype has been severely disrupted.
The transformation, however, does legitimatize comparisons since all
traits in all genotypes are expressed in the same units. This was the
intention when Gauss created the normalizing transformation in the 19th
Century. The effect of the transformation can be seen in Table 3, which
lists the various intra-genotypic correlation coefficients for both
transformed and non-transformed data.

Table 4 lists the vector lengths, or the individual genotypic
yields in standard deviations for the eleven year-sites. This may be

construed as the measure of superiority, since yield is the single trait

Table 3 .

Entry

69-27-389

Menominee

Heritage

69-27-403

Orbit

Korwood

69-27-414

Garry

Ausable

Mariner

Moore

Portal

Mackinaw

Benson

Lang

Marathon

69-28-124

Noble

Dal

CLD 64

Wright

16

Comparisons of intravarietal component correlation coefficients

between non-transformed and transformed data.

Correlation Coefficients: r

’43 0-15 Hg

"‘5

xy

-.207
.017

-.405
-.655

-.195
-.624

-.134
-.725

-.612
-.835

-.518
-.790

-.215
-.725

-.524
-.423

-.394
-0649

-.518
-.447

-.507
-.708

-.421
-.364

-.336
-.600

-.464
-.467

-.469
-0426

-.064
-0713

-.718
.682

-.137
.504

. 378
-0240

-0224
-0282

-.684
-.424

X2

-.339
-.220

-.345
.218

-.267
.133

-.068
.659

-.075
.018

-.150
.052

-.234
-.183

-.363
.247

-.267
-.208

-0346
.049

-.264
.059

-.022
.304

.159
.154

.062
. 33‘

.059
.342

-.627
.265

.050
.038

-0361
.039

-.477
-0626

-0179
-0482

-. 279
.024

XV th

.787 .151
.707 .286
.839 .237
.358 .363
.826 -.005
.384 .199
.859 .123
.418 .599
.817 .119
.451 -.058
.759 .405
.056 .293
.807 .490
.620 .535
.745 -.038
.615 .061
.695 .186
.178 -.401
.755 .065
.585 -.082
.706 -.040
.286 -.295
.649 .267
.707 .283
.779 .366
.091 .077
.779 .241
.159 -.050
.808 .477
.723 .323
.702 -.101
.644 .826
.820 .338
.286 .170
.731 .018
.784 .034
.712 .080
.562 -.063
.843 .306
.632 -0257
.744 .003
.898

yz

-.116
-.588

.048
-.286

.146
-.246

-.100
-.848

.197
-.373

.352
.055

.116
-.136

.097
-.513

.022
-0326

.282
-.484

.093
-.163

.212
-.823

-.233
-.496

.119
-.638

.503
-.264

-.027
-.154

.019
.003

.746
-0222

. 352
-.282

.049
-.101

.411

YV

.330
.584

.035
.081

.288
.337

.249
-.048

-.119
-.O49

.096
.379

.283
-.211

.051
.228

.252
.362

.044
-.060

.141
.385

.343
.168

.235
.297

.038
.535

.052
.237

.178
-0914

-.311
.318

.484
-0079

.220
.465

.203
.653

-¢101

.215 -.469 -.125

Intravarietal
ytw zw ztw
.009 -.226 .689
-.332 -.648 .518
.104 -.083 .607
-.037 .428 .546
0352 -0006 .697
-.039 .078 .743
.281 .176 .562
-.332 .379 .417
.432 .122 .738
-.266 .055 .545
.166 .220 .537
-.245 .678 .346
.181 -.028 .398
-.486 -.241 .025
.352 -.170 .529
.182 .271 .508
.133 .015 .576
‘0034 ’0199 .432
.449 .031 .307
.141 .419 .156
.234 -.006 .748
.267 .081 .557
.201 .334 .675
-.796 -.014 .852
.113 .165 .417
-.054 .312 .693
-.120 .451 .865
.086 -.369 .347
.148 .446 .696
-.280 .311 .240
0227 -0196 .186
-0332 -0029 0065
-.067 .271 .551
.136 .536 .756
.434 .262 .705
.257 .302 .374
.310 .076 .355
-.479 -.408 .396
.018 .113 .695
.011 -.379 .760
.194 .126 .725

.436 -.225 -.229

wtw

.298
-.026

.427
.795

.297
.216

.374
.503

.443
-.079

.640
.505

.523
.742

.288
.460

.447
-.650

.347
.510

.323
.081

.429
-.091

.445
.524

.430
-.187

.582
.121

.257
. 378

.450
.787

.446
.139

.412
-0426

.414
-0219

.299
.476

17

o mo.u m¢.- om.u Ho.u No. ~o.- NN. mm.. Hg. N¢.H om. zucpxumz

mm.n ¢H.i mm. mm.- mo.Hu vo.u mm.u cm.u No.n om.~- em.Hu om.Hu pmucom
NH. as. mo.u mo. mm. mw.u mm. mm. mm. om.u Kw. me. mcooz
Ho.u mm. co. ma. mm.- oH. mm.u om.- no. HH. Hm.u 0H. mewgmz
mo. mm. N¢.- ma. mm.. 00.- o¢.u em. mm. mm. mH.u HH. mpnmms<
mH. HN.H o¢.Hu 0m. mm.~ mm.- mm. Hm.u Kw. mm. we. em. xggmw

ow. mm.H Hm. ow. «N.H mm.~ Hm. Hm. mu.u HH.H om. oo.H efivunmnmo
mm. mm. NH.H mm. mm.- HH. No. me. mo.- mH.u mm. mm. voozsox
em. No.n mm.~ mo. HN.H om. mm.H mm. mm.u mo.H mm. ma. uwnco
me. mm.H mN.H mm. we. cm.~ ew.H mm.H mm. No.u om.H mm.H movumwumm
me. om. o~.: eo.H mm.H mm.H um.H mu. No.u mm.H mm.“ mm.H mmmuwcm:
me. mm.H mm.H RN.H mm. mm. mm.H mm. mm. no.H en. am.H mmcwsocmz
Hm. NH.H- mm. NN.H mm. mm. mm. mv.H No. mm. we. “0.- mwmnnmumm

 

.3 HH 0H m N N o m a m N H Hmucm

wHHm-Lma>

.wz use xv: .mgumcmH cowom> com mmumm-cmm> :mmzum cc< mmgxpocwu cmmzpmm comwcmaeau .c m4m<H

18

em.u co.Hu mm.- mo.~- om.H- oo.mn o¢.- mo.~- mm. mo. Ho.Hi o¢.Hu exams:

um.u oo.u ~¢.Hu mo.Hu mm.Hu wo.Hu mo.~u mw.mu N©.Hi m¢.~u mm.u nm.~u we cpu
mm.u mo.Hu ew.u ~H.Hi vo.mu ~m.~- mo.Hu mm.- mm. mH.Hu HH.~- om.u Poo
NN.- mH.- mm. om.u oq. HH. en.Hu mc.u Hm.fiu we. mw.u ~H.Hn wpnoz

0 me.. mm.u mm.u ¢N.- mH. mm.. mm. mm.~ Hm. mm. HH. emfiuwmumo

 

oH.- NN. mo. N..- ao.- om.- mN.- mo.- mN. Hm. «H.N- NN.- cogsmcmz

sH.- NN.- Hm. oe.- Ne. mo.- mN.- as.H Ne.N- mN.H- mm.- mm.- New;

am.- mo.H- Na.H- ON.- om.- NN.- No.- om.H- No. mo.H- ma.- Nc.- comcmm

.IHa HH oH a N N o m a m N H NcHeN
aHHm-c.m>

.emsercou .e NHN<H

19

of greatest importance. Any positive value represents above average
yield with respect to the zero mean of all twenty-one genotypes.
Superior varieties, superior in their yield, are readily apparent in
this table. It is of interest to note that Heritage yield was below

the mean in only two year-sites, those of Kalamazoo County in 1977 and
Lenawee County of 1979. A discussion in greater depth on Heritage may
be found in Chapter 2 of this dissertation. Lest the success of the
Michigan oat program be questioned, observation of the yield performance
of Clintland 64 should dispell any doubts. Clintland 64 was a recom-
mended variety until 1970 when newer varieties, considerably superior to
Clintland 64 in yield, were introduced. For the eleven year-sites in
this study, Clintland 64 had become the poorest variety of the twenty-
one genotypes included. Appendix 1 lists the specific year-sites with
the numbers under which the years and locations may be found.

Based upon Table 2 and 4, it is possible to visually represent each
variety as a cluster of vectors (eleven each) graphed with respect to a
standard (Ux or W1) lying along the x-axis. Figure 6 is an example,
using Heritage as the genotype. The standard, Uw’ is the x-axis and it
equals .91 standard deviation units of yield in length. Each year-site
for Heritage is thus plotted relative to the standard and their lengths
are simply the W for each year-site. Any yield superior to the mean
(=0) of all twenty-one genotypes must lie in the first two quadrants.

This is a function of the cos.l

r1 which produces Cg.of between 0 and 1800
(see Figure 3). This places all angles within the first two quadrants
and any yield which is positive insures a positive direction. Quadrant
3 has two year-sites (4 and 10) expressed. The angles (37.30 and 29.20,

respectively) are positive, but the lengths are negative, since the

20

z . oH

HH

mmmcmmu m.m¢ u goo

HmHHca :oHH.H>mo
ucwucmum :H upmw> u gumcva .mmpwmucmm> cm>mpm Lou

mmmgmcm: com mcopum> we covpsnwcumwo Hmamwcm>mcucH

.m mmszu

21

yield of Heritage in these two year-sites was below the population mean.
The overall stability of this vector cluster is represented in the
value of 48.90, the second highest such value among all entries. It is
interesting to note that if all twenty-one genotypes were similarly
graphed and their respective vectors summed for each year-site, the
entire matrix of vectors would collapse to a single point. Properly
manipulated, there is, indeed, symmetry in nature other than the five
perfect solids!

Table 5 lists the means from Tables 2 and 4. For mathematical
etiquette, this is a better mode of expression since the mean vectors
for each entry are arranged such that each has an assigned directional

stability (0 ) and a corresponding length (Hi). This table facilitates

ei
inter-genotypic comparisons. The author has no reservations concerning
this measure of superiority; however, the measure of relative stability
is more difficult to interpret since high degrees of instability are
correlated (raw= .38) with superior yield. Testing the hypothesis,
Ho:re
interpretation of this phenomenon is that genotypes which are superior

w = O with t = 1.79, the probability is: P< .1. The most logical

in their yield (maximum traits, for this study), are superior due to
their plasticity of response via their yield components, or component
compensation. A stress manifested at one point in their development is
compensated by a concomitant improvement in the remaining components
when constrasted with more poorly performing genotypes, unless the stress
occurs during the development of Z.

There is no final statement to be made concerning this new system
for measuring and presenting cultivar stability and superiority. Further

review and possible additional manipulation of the data are very likely.

22

TABLE 5. Intervarietal Paired Observations For Stability

And Superiority.

Stability (09,)

Superiority (Ni)

 

Entry (in degrees) (in standard deviations)
69-27-389 17.8 .31
Menominee 27.7 .47
Heritage 48.9 .47
69-27-403 59.3 .49
Orbit 14.3 .34
Korwood 27.5 .22
69-27-414 28.5 .40
Garry 15.9 .13
Ausable 23.4 .05
Mariner 8.0 -.01
Moore 18.4 .12
Portal 20.4 -.25
Mackinaw 7.1 0
Benson 8.3 -.39
Lang 7.3 -.14
Marathon 13.4 -.10
69-28-124 15.2 0
Noble 21.6 -.22
Dal 33.5 -.53
Clintland 64 27.4 -.77
Wright 12.6 -.54

23

Since the purpose was to analyze genotypes by a different method for
multiple traits, enormous complexities arise which must be interpreted

with care lest the conclusions be nothing more than specious nonsense.

II. Path Analysis

 

Table 6 lists the six path coefficients and the correlational
matrix from which they were derived. Calculation of ”1’ u2 and u3 pro-
vide the necessary entries for computing the three genotypic components,

v1, v2 and v3 for each genotype. The equations are:

<
ll

1 ”1(a4 I alaS I azas I ala3a6) I ”1wa

v2 = u2(a3a6 + a5)

u3a6,
and the results are detailed in Table 7, accompanied by the genotypic
mean yields. The results are remarkable in view of the highly variable

results of Tai (9,10). First, all v1, v2 and v are positive for each

3
genotype. Secondly, the values are tightly clustered within each Vj and

thirdly, the mean values descend sharply from v1 to v3:
v1 = .770, °v1 = .057
v2 = .504, °v2 = .127
v3 = .217, °v3 = .084

This strongly supports the hypothesis of component compensation and,
further, these results again indicate that the most important yield com-
ponent toward the formation of yield is X, panicles/unit area, followed
in importance by Y and Z. Tai concluded from his positive v3 for

potatoes that Z, or tuber bulking, was the most important component in

24

 

qu. wmo. aux. mHH. moo. ¢o¢.- wmm. mew. mom. me. qu. ¢m¢.- comcmm
moH. mmw. mun. mmm.- mmH. omm.i emH. omm. mmm. mom.- Hmo. omm.- zmcwxumz
emm. mum. one. NHN. Nwo.- HN¢.- new. mam. mew. mew. Hwo. Hmv.- Hence;
000.- HGH. ooN. moo. eom.- Nom.- «mm. ewe. NHH.H mmo.- NmN.- Nom.- mcooz
Hmo. ceo. mmm. mmm. oqm.- me.- com. mmm. omH.H HcH. mm~.- me.- cwcwcmz
mHo. mmm. moo. mmo. NoN.- com.- «um. mam. «No.H moo.i mom.- cmm.- mpamma<
oNH.- Hmo. we“. Nae. mom.- «mm.u emH. Nmo. «8H.H mNH.- Hm¢.- amm.- accmo
mmo.- mwm. Now. 90H. qmm.- mHN.- mmH. mow. cum. moo. mHm.i mHN.- eHv-Nmumo
0mm. moo. mmN. Nmm. omH.- me.- NNH. moo. woo.H oxm. mwo. mHm.- noozcox
NNH. mHH.- NHw. NmH. mNo.- NHo.- omo. wwm. qu.H New. mko. NHm.- awaco
oNH. mew. mmw. ooH.- moo.- cmH.- omm. Nov. Nmm. HHH.- mmo.- cmH.- movium-mo
000.- mwm. omw. 08H. Nom.- mmH.- owH. ome. com. moo. weN.- mmH.- mmmawcm:
mmo.- mmo. mmw. wqo. mqm.- coe.- «mm. mmc. HmH.H moH.- mwm.- ¢o¢.- mm:HEocmz
omm.- omm. Rm“. oHH.- mmm.- ~o~.- ooH. mam. cmm. emH.- m~m.- Nom.- ammunwumo
3N 3» 3x N» Nx xx mm mm cm mm me He Acucm
mucmwuwkemou cowumHmccou mucmmu_w.mou span
.mpmo umEcoemcmcp sock mucwwuwmmmou cowumHmccou uc< mzuma acmEHLQ .o uHm<H

25

89V a co. co.

II
S.

 

Ho.vm .28 mm. uc
mo.vm .6. 3. u...

H.vm .8» so. u.» 6ch 8. :58 Ho: 2. L8 33:; 28.5.5 mumewxocaaq
omH. HoH.- ecu. HHV. mum. woo.- omH. ooo. mom.H oHe. moo. ooo.- “cove:
mHH. mom. moo. moo. mNH. «NN.- mom. oHv. mom. moo. NNH.- «NN.- co ocm_pcvpo
oNo. omN. NH“. mom. NNQ. on.- qu. «we. HoH.H How. Ho¢.- mum.- Foo
mom. mom. HmN. ooh. Hom. an.- omm. mom. omo. oHN. com.- an.- mpnoz
Hum. HHm.- omo. mHo. moo. oHN.- oom. mom. «oN.H mHH. HmH. oHN.- emHuo~-mo
omH.- oNH. mom. “No.- Kmo. voo.- omv. «mm. oom. ooo.- «mo.- ooo.- cospmcmz
ova. moo. ooo. mom. mmo. mo¢.- qu. Hoe. moo.H Hoo. mNm. mo¢.- memo

3N 3x 3x Nm Nx xx on mm mm mm mm Hm xcucu

mucmwuwemmoo cowumHmccoo

chaHoHccmou eHaa

. woo: .Hpcou . o WES.

26

TABLE 7. Mean Yier (Ni) And The Three Genotypic Components Of The

GxE Interaction For The Twenty-one Genotypes.

 

Yield
Entry (bu/A) V1 V2 V3
69-27—389 121.4 .787 .504 .152
Menominee 125.2 .839 .410 .264
Heritage 125.2 .826 .460 .180
69-27-403 125.7 .859 .368 .277
Orbit 122.2 .817 .483 .091
Korwood 119.3 .759 .573 .153
69-27-414 123.6 .807 .467 .134
Garry 117.2 .745 .518 .167
Ausable 115.4 .695 .572 .260
Mariner 114.0 .755 .509 .235
Moore 117.1 .706 .579 .209
Portal 108.4 .649 .679 .200
Mackinaw 114.1 .779 .527 .150
Benson 105.0 .779 .451 .339
Lang 110.8 .808 .488 .136
Marathon 111.9 .702 .224 .421
69-28-124 114.1 .820 .398 .204
NobTe 109.0 .731 .590 .192
0a] 101.7 .712 .528 .373
CTintTand 64 96.0 .843 .402 .263
Wright 101.6 .744 .560 .158

27

potato yield. This, as can be concluded from Table 7, is the exact
opposite of oats, and by inference, of all cereals. Additionally, it
might also be concluded that despite the apparent dissimilarities of
the oat genotypes included in this study for the many traits and with
respect to their different origins, they are, nevertheless, a highly
homogeneous species.

Figure 7 is a plot of the predicted mean yields graphed against
actual mean yields. The expectation here is that the values would be
closelyrcentered around the b = 1 regression line (ascending at 450 from
the origin). The predicted W1 are actually almost randomly distributed.
Based upon the r2 values from Table 8, and as adjudged visually from
Figures 7 and 8, it is apparent that the model of Tai (9,10) might
require modifications to predict yield in oats. The author places con-
siderable confidence in the genotypic components (Table 7) since they
are resolved algebraically from path coefficients. These paths are,
themselves, derived from the correlation coefficients within each geno-
type. Insofar as the correlations are correct, the Vj will be correct.
The clustering of genotypes within the respective vj classes suggests
that the vi are, in fact, population parameters of benefit to biologists.

What is especially disconcerting) to the author, however, is the
fluctuations in correlations which he has noted in using data recorded
by the late Dr. John E. Grafius over the past twenty years. These data
are sound insofar as one can assay the correlational matrix for many
varieties for as many as twenty year-sites (see Chapter 2, Appendix 1).
Comparing the correlations of Heritage in Appendix 1 of Chapter 2 with
the non-transformed set from this chapter (Table 3), the differences are

startling. If these discrepancies are generalized across crops and

28

TABLE 8. Three Environmental Components Of The GxE Interaction For
The Eleven Environments, The Coefficients 0f Determination
For The Yield Equations And The Mean Fitted Yields In

Standard Deviation Units.

2

 

Year-Site* r1 r2 r3 Coef. Det.: r ka
1 1.58 -1.60 -2.16 .144 .26
2 3.74 .94 -.69 .249 -2.94
3 -1.82 -2.43 -3.01 .270 3.72
4 -1.33 -.61 .54 .155 -.92
5 -3.49 -3.09 -5.43 .275 6.28
6 .52 -1.40 -2.22 .219 1.00
7 2.78 -1.49 -2.35 .300 .46
8 -1.27 -2.69 -3.79 .351 3.04
9 1.09 -.94 -2.15 .136 .17
10 -.04 -.67 -1.59 .140 .48
11 -.45 -1.30 -1.18 .074 .17

*See Appendix 1 for explanation of years and locations.

29

FIGURE 7. Predicted Mean Yields (Standard Deviations)

Plotted Against Actual Mean Yields.

-.90

-.99

-1.08 _

Predicted

w.
‘ -1.17 ~-

-1.26 '

-1.35 '

 

-1.80 -1.50 -1.20 -.90 -.60
Actual W1

30

FIGURE 8. Residuals Plotted Against The Twenty-one Genotypes.

 

1 5 10 15 21

Genotype*

*See Appendix 2 for a listing of genotypes.

31

environments for different years, it means that most, if not all, of
the published component correlations are too narrow in sc0pe to draw
conclusions from for population parameters. The implications for plant
breeders is that extreme care must be taken when basing a breeding
strategy upon component correlations drawn from a narrow group of en-
vironments and years. This conclusion would hold true of the author's
work detailed in Chapter 4.

The environmental components of the GxE interaction, as resolved
by least squares regression, are extremely heterogeneous (Table 8). It
should be recalled that the matrix of observations which comprised this
aspect of the study were transformed by using the grand mean and grand
standard deviation (Figure 5), intended to relate each genotype and
year-site to a mean of zero and a unit variance. The low coefficients
of determination (Table 8) would indicate that there is insufficient
information used to predict yield by this method. The adjusted r2
values (adjusted for degrees of freedom) lower the r2 values consider-
ably (not listed). Figure 8 is a plot of the residuals as a function
of genotype. Were the r2 values high, one would find the residuals
centered around zero on the ordinate and this tight band moving to the
right as each genotype's residuals are recorded. What is the case in
this study is that the residuals are skewed strongly in the directions
of both higher and lower yielding genotypes. Observation of the mean
fitted yields (the constant terms in multiple regression), ka, in
Table 8 also suggest a bimodal distribution with the smallest terms
clustered around year-sites 1-3 and 9-10. This might indicate an arti-

fact stemming from the nature of the original transformation (Figure 5).

32

In conclusion, the author is generally pleased with the stability
index and the path coefficient study which was so painfully accomplished.
It should be noted that the data set is amenable to further analysis by
modifications of the techniques herein used. Additionally, it would
seem wise to apply other methods of analysis--in particular, factor
analysis--in an attempt to empirically determine the commonality of the

various statistics which result.

33

 

APPENDIX 1. Notation For The Eleven Year-Sites.
Number Year Site
1 1977 E. Lansing
2 1977 Tuscola County
3 1977 Lenawee County
4 1977 Kalamazoo County
5 1978 E. Lansing
6 1978 Tuscola County
7 1978 Lenawee County
8 1978 Kalamazoo County
9 1979 Tuscola County
10 1979 Lenawee County
11 1979 Kalamazoo County

34

APPENDIX 2. Notation For The Twenty-one Genotypes.

 

Number Genotype (Variety)
1 69-27-389 *
2 Menominee *
3 Heritage *
4 69-27—403 *
5 Orbit

6 Korwood *

7 69-27-414 *
8 Garry

9 Ausable *
10 Mariner *
11 Moore

12 Portal

13 Mackinaw *
14 Benson

15 Lang

16 Marathon

17 69-28-124 *
18 Noble

19 Dal

20 Clintland 64
21 Wright

* Michigan lines

10.

11.

12.

35

CHAPTER 1--BIBLIOGRAPHY

Finlay, K.W. and G.N. Wilkinson. 1963. The analysis of adaptation
in a plant breeding programme. Aust. J. Res. 14:742-754.

Eberhart, S.A. and W.A. Russell. 1966. Stability parameters for
comparing varieties. Crop Sci. 6:36-40.

Tai, G.C.C. 1971. Genotypic stability analysis and its application
to potato regional trials. Crop Sci. 11:184-190.

Shukla, G.K. 1972. Some statistical aspects of partitioning
genotype--environment components of variability. Heredity.
28:237-245.

Grafius, J.E. 1964. A geometry for plant breeding. Cr0p Sci.
4:241-246.

Dewey, D.R. and K.H. Lu. 1959. A correlation and path coefficient
analysis of components of crested wheatgrass seed production.
Agron. J. 51:515-518.

Duarte, R.A. and M.W. Adams. 1972. A path coefficient analysis of
some yield component correlations in field beans (Phaseolus
vulgaris L.). Crop Sci. 12:579-582.

Grafius, J.E. and R.L. Thomas. 1971. The case for indirect genetic
control of sequential traits and the strategy of depkyment of
environmental resources by the plant. Heredity. 27:433-442.

Tai, G.C.C. 1975. Analysis of genotype-environment interactions
based on the method of path coefficient analysis. Can. J.
Genet. Cytol. 17:141-149.

1979. Analysis of genotype-environment interaction of

 

potato yield.

Hamid, Z.A. and J.E. Grafius. 1978. Developmental allometry and
its implication to grain yield in barley. Crop Sci. 18:83-86.

Grafius, J.E. 1978. Multiple characters and correlated response.
Crop Sci. 18:931-934.

CHAPTER 2

ANALYSIS OF A NEW MICHIGAN OAT VARIETY BY
YIELD COMPONENT REGRESSION

Introduction

 

An important aspect of a plant breeding program for agronomic
creps is the methodology used in evaluating potential varieties. Aside
from such important traits as disease and lodging resistance and the
various quality factors necessary to a superior cultivar,cwaluation typi-
cally centers around measures of yield. In most cases yield is considenai
as an holistic trait and it is ranked by contrasting a specific variety
with other, presumably adapted, cultivars. There is no question that
such systems are effective in that superior yield is easily identified
in any given experiment. Comparisons between experiments over years and
locations is confounded, however, since the genotype by environment in-
teraction (GxE) is usually prevalent. Because the variability induced
by different environments is so extreme various systems have evolved to
identify individual cultivars which are both stable and superior in their
yield. The primary method used by the Michigan State University oat and
barley breeding programs is that of Pedersen gt al. (1). The system is
a modification of those proposed by Finlay and Wilkinson (2), Eberhart
and Russell (3) and Schmidt (4).

The author proposes a modest change in Pedersen's system which

is simply to regress an individual cultivar's yield components against

36

37

the mean component of the representative population. By this method it
is hoped that the superiority (or inferiority) of a cultivar can be
ascribed to one or more of its components, along with some measure of
its stability. An explanation of the yield components may be found in

the Materials and Methods section.

Literature Review

 

Pedersen, Grafius and Everson (1)1wnposed their system to
simplify analysis of many genotypes grown in different locations over a
number of years. It is very much a hybrid approach, but it is nonethe-
less elegant for this. Beginning with the concept of regressing varietal
yield against the mean of all varieties grown in a common location (first
proposed by Yates and Cochran (2)), Finlay and Wilkinson (3) placed
reliance on the regression coefficient, b, to determine stability. A
value of b = O was interpreted as absolute phenotypic stability, b = 1
as average response over environments and b‘< 1 as inferior performance.
Eberhart and Russell (4) proposed an environmental index constructed by
subtracting the specific environmental mean yield from the mean over
every environment for yield. The mean square deviation from regression
is their stability parameter. Schmidt gt 31. (5) introduced the coeffi-
cient of determination, r2, into consideration for winter wheat evalu-
ation. They showed that newer cultivars performed considerably better
than the checks and that new cultivar response to improving environments
is both superior to the checks and highly predictable. Hence, the
coefficient of determination was suggested as a stability parameter.

The utility of the system is amply demonstrated by its use in

evaluating Michigan oat, barley and wheat cultivars since the regression

38

graphs are comprehensible to a variety of people, including cereal
growers. The method was applied to the traits yield and test weight

for a collection of twenty-four oat varieties routinely grown in Michi-
gan and covering the years 1968 through 1978 (6). The response of each
cultivar was regressed upon the mean response of all twenty-four for the
traits being evaluated. The number of tests ranged from seven (for
older, low acreage varieties) to forty-four (more highly adapted and
newer varieties). For yield the r2 values were clustered as follows:

8 varieties > .90, 13 varieties > .80 and 3 varieties > .70. r2 values
for test weight were: 3 varieties > .90, 8 varieties > .80, 5 varie-
ties > .70, and the other 8 varieties ranging down to r2 = .47. With
respect to yield, high coefficients of determination reflect the high
degree of linearity associated with cultivar response; in other words,
the site mean is an excellent predictor of performance. Test weight had
a higher degree of variability, but the system identifies inherently
poor cultivars and, as one might expect, the single variety with the

lowest r2 was Lang, which is poorly adapted to Michigan growing condi-

tions and is not extensively planted in this state.

Materials and Methods

 

The regression method of Pedersen gt a1. is exceedingly simple and
easily identifies varieties which are superior. Essentially, a variety
is considered a constituent of an adapted population of genotypes devel-
oped in a common region. For Michigan oats this region is roughly
inclusive of all cooperating stations in the Uniform Mid-season Oat

Performance Nursery. More specifically, this population, or gene pool,

39

is of more northerly states within the UMOPN. Thus, the collection of
cultivars developed and released in this region can be viewed as a gene
pool which reflects average oat response. Regressing individual culti-
var response upon the mean response of a collection of adapted genotypes
for such traits as yield and test weight is an excellent measure of
superiority. Plotting 9 values for a given cultivar with respect to the
b = 1 regression line provides a visual method for determining response
in different environments. The statistics which accompany regression
analysis are the "y-intercept," the regression coefficient, b, the
sample size, n, and the coefficient of determination, r2. A superior
genotype should have a positive "y-intercept," a large "n" to insure
adequate sampling and a regression coefficient greater than or equal to
one. A superior genotype which is stable in its superiority (stably
superior) should have a high coefficient of determination, in addition
to the above-mentioned statistics.

This paper proposes a further use of regression analysis, based
not only on yield and test weight, but upon the primary components of
yield as reflected in oats, where X = number of panicles/unit area,

Y = number of seeds/panicle, Z = weight/seed, and the combined com-
ponent, XY, or number of seeds/unit area. The individual cultivar
which will serve as the tester is Heritage, formerly Michigan line
64-152-47. The data are derived from thirty separate experiments grown
over the years 1972 through 1980. The experiments were rod-row nur-
series which include a collection of from 25 to 36 adapted varieties
developed for the aforementioned UMOPN region. For twenty site-years
detailed yield component data for X, Y, and Z were computed. For all

thirty site-years Z, W (yield) and TW (test weight) were recorded.

40

Thus, an estimate 0f the pooled component, XY, was computed since W/Z =
XY. By regressing cultivar yield components upon site mean components
it should be possible to analyze in greater depth the mode of response
of the components (and, hence, of yield) to varying environments.

Since regression analysis is well known and its application is
facilitated by the readily available use of small and inexpensive calcu-
lators, I will not detail the methodology. All experiments were planted
in lattice designs with four replications and analyzed as such. The
data for Heritage were extracted from tables of adjusted means and the
site data were simply the means of all entries included in the rod-row

experiments.

Results and Discussion

 

The utility of the simple linear regression approach, as modified
for the primary components of yield, is demonstrated in Figures 2-5.
Figure 1, the regression for Heritage tiller number on site X, shows a
suppression of panicle production in the poorer sites, since the regres-
sion line lies below the b = 1 regression line. The b = 1 regression
line represents average response to the environment and Heritage responds
more favorably as the site mean improves. Low site X is correlated with
Kalamazoo County, one of the four testing sites for Michigan oats. The
soils are sandy and this frequently results in heat and drought stress
in the crop. For those regions, therefore, in which stress occurs early
in the season, Heritage might not be the variety of choice since the

consequence might be a yield reduction through poor tiller initiation.

41

FIGURE 1. Regression 0f Heritage X Onto Site X.

20

19
-—. b=1

I 8 _____ Heritage

I 6
HERITAGE

l5

y=:- |.61D +- LC)? x *
n== 2 O

2
rc.93
ll

*Equation of the regression line.

 

      
   

 

IO ll l2 I3 I4 15 I6 l7 IO

SITE X

2
tillers / H

42

The expression of seed number strongly influences yield in Heritage

(Figure 2), for the regression for Y lies considerably above the b = 1
regression line throughout the range of site values. The question
arises, however, whether this Y-superiority is sufficient to offset the
yield reductions imposed by low X in the poorer, more stressed, sites.
Figure 3, the regression for the pooled component over thirty year-
sites, shows that X is compensated by the consistently high values for
Y. The regression coefficient is nearly 1.2, which demonstrates the
strong and favorable response of Heritage to improving environments.
At the poorest locations, the compensation is sufficient to insure that
Heritage responds at least as well as the average for most oat cultivars
grown in these locations. The high value for r2 provides strong confir-
mation that the predicted response is reliable.

Figure 4 indicates that Heritage Z is little different from the
site seed weight, except at the extremes of the range of values result-
ing from the different environments over the eight years of testing.
Considering the inferior expression of X at the poorer sites, it is
possible that the slight inferiority of Z, coupled with poor X in
Heritage, would reduce yield. This would be the consequence only if
the locations in which low site 2 was prevalent were highly correlated
with the same locations which manifest a reduced X. An observation of
the data and the correlation coefficients between X and Z. for both the
site and Heritage, would indicate that they are poorly correlated. rxz
(Heritage) = -.13 and rxz (site) = -.11 would imply that seed weight is
randomly distributed. The entire set of correlation coefficients and

their respective t-statistics may be found in Appendix 1.

43

FIGURE 2. Regression Of Heritage Y Onto Site Y.

     
  

 

 

IIO
'00 -—- b=1
H ' I
_. eritage I
.I
.l
4'
I],
90 I
/'
HERITAGE /
Y
80
70 y--so.62.l.:9x*
/'
’I/ n; 20
r: .80

60

1' *Equation of the regression line.

1'
"I'
V
50 60 70 BO 90 IOO

SITE Y

occdsllponlclo

44

FIGURE 3. Regression Of Heritage XY Onto Site XY.

        
  
 
  
   

DSIDCI

-—- b=1

4’
I
1'
4'
HERITAGE
IZIDC)
/ y--Iov.62+i.19x*
IOOO /
/ “2: 30
f I .92
800
*EQUation of the regression line.
610()

600 800 IOOO IZOO I400

ESI'T ES )(‘Y

2:
seeds/ft

45

FIGURE 4. Regression Of Heritage 2 Onto Site 2.

340

    
  
 
   

--- b=1

330 -— Heritage
3 20

310

HERITAGE
300

290 ys-25.B|+l.09x *

III 30

280 f2. .73
270

2 60 *Equation of the regression line.

 

 

260 270 280 290 300 SIG 320 330

SITE 2

m9 /uod

46

Figure 5 is the distillation of the preceeding yield component
regressions. The regression of Heritage yield onto site yield for thirty
year-sites amply demonstrates the superiority of Heritage with a high
degree of determination. This superiority is primarily a function of
the production of more seeds per unit area than competing varieties,
without a severe reduction in seed weight, Z.

It should be remarked that any oat population is, to a considerable
extent, a predictor of the response of any single genotype within the

population. Avena sativa is a population of homozygous genotypes which

 

will, therefore, necessarily predict its own performance. The plant
breeder, however, operates narrowly within the broad parameters expressed
by the species as a whole, since he deals with a sample of the popula-
tion. This sample, or subpopulation, is more adapted to the specific
environments for which the plant breeder engineers his varieties. Thus,
one would expect a high coefficient of determination for the trait of
yield. This high degree of prediction results from the homogeneous
nature of the adapted population as well as from the developmental

nature of yield. This latter point can best be expressed by the term
"component compensation" (7). The pathway to yield, by way of the pri-
mary components, is a fluid one. Therefore, a stress on X (panicles/
unit area) should reflect differential genotypic response with respect
to site X and a low r2 would imply that the environments included in

the sample were highly stressed. However, yield is the product of three
components and of the direct and indirect effects through the components.

A stress in one is frequently buffered by superior expression in another
component, and this will stabilize yield, the holistic trait. It may

be concluded that Heritage represents a stable variety.

47

      

FIGURE 5. Regression Of Heritage W Onto Site W.

100

150 -—- b=1

_. Heritage
I 40

I30

HERITAGE
I20

YIELD
IIO

IOO
r--o,oo+I.I0x

.0 l “.30

z o
oo "' 5
70

*Equation of the regression line.
00

 

GO IO 70 IO .0 100 IIO IIO ISO 140

SITE YIELD
bun/«r0

48

FIGURE 6. Regression Of Heritage TW Onto Site TW.

37

3G

      
    

--- b=1

_ Heritage
3 5

HERITAGE 3‘

TV
33

:-8.‘°+IO'°‘ *
32

08130

:.4
3I r 7

3O

2 9 *Equation of the regression line.

 

 

E. 29 30 3| 32 33 34 33 3G

SITE TW

Ibo/bu

49

There is an elegance to this method of assaying performance
through simple linear regression. It has been of help to the author in
exploring relative genotypic performance and its efficacy is in no way
limited to the six regression graphs herein presented. It seems a
logical choice for contrasting the response of progeny to that of the
parents, for both yield and its components, over a range of environ-
ments, where sufficient data are available. In an unpublished study,
Nelson (1981) generated a series of regressions for such a purpose and

found the information of considerable value.

50

 

 

APPENDIX 1. Correlational Matrix For Site And Heritage And Their
Associated t-Statistics (r=O).
Correlation Site r t Heritage r t
rxy -.02 .08 -.15 .64
rXZ -.11 .47 .13 .56
rxw .80 5.66*** .80 .66***
rxtw .23 1.00 .07 .30
ryz -.12 .51 -.31 .39
ryw .38 1.74* .33 .48
rytw .32 1.43 -.79 .44***
rZw .15 .80 -.01 .05
rztw .60 3.97*** .61 .O7***
rwtw .36 2.04* .38 .17**
* P < .10
** P < .05

*** P < .001

51

CHAPTER 2--BIBLIOGRAPHY

Pedersen, A.R., E.H. Everson and J.E. Grafius. 1978. The gene pool
concept as a basis for cultivar selection and recommendation.
Crop Sci. 18:883-886.

Yates, F. and W.G. Cochran. 1938. The analysis of groups of experi-
ments. J. Agric. Sci. 28:556-580.

Finlay, K.W. and G.N. Wilkinson. 1963. The analysis of adaptation
in a plant breeding programme. Aust. J. Res. 14:742-754.

Eberhart, S.A. and W.A. Russell. 1966. Stability parameters for
comparing varieties. Crop Sci. 6:36-40.

Schmidt, J.W., V.A. Johnson, A.L. Diehl and A.F. Dreier. 1973.
Breeding for adaptation in winter wheats for the Great Plains.
In. Fourth Int. Winter Wheat Symp., E.R. and L.M.S. Sears, ed.
Univ. of Missouri, Columbia, Mo.

Grafius, J.E., R. Leep, D. Wolfe and C. Hoy. 1978. Oat variety
performance in Michigan, 1968-1978. Mich. Coop. Ext. Service,
File No. 22.15.

Adams, M.W. 1967. Basis of yield component compensation in crop
plants with special reference to the field bean, P. vulgaris.
Crop Sci. 7:505-510.

CHAPTER 3

OAT AND BARLEY COMPOSITE EXPERIMENTS

Introduction

 

The simple experiments detailed here originated during the teaching
of CSS 408, Introduction to Plant Breeding. The subject of breeding
systems was the venerable bulk population method. While writing the
class notes the potential for bias in the system was seen: the breed-
ing objective, which is to allow natural selection to operate on the
bulk population, might not be a valid one, couched as it is in vague
terminology. The source of the bias can be found in the primary com-
ponents of yield. A brief review of the bulk population method will
help explain the conflict.

A bulk population usually consists of a large field plot of segre-
gating genotypes or, if homozygosity is well advanced, the bulk is a
heterogeneous mixture of homozygous genotypes. In either case the plot
is harvested in bulk and replanted in bulk, the cycle being repeated as
many times as the breeder deems necessary to achieve his goals. These
goals are several in number. The first is to allow natural selection to
operate on the population to extract those genotypes which are superior
in adaptation and presumably in yield; the secondary goals are the re-
ductions in time, labor and expense which naturally result from the

breeding system. The underlying assumption is that the natural selection

52

53

¢<w»
o¢m.~.¢ fit at ¢¢ n¢ NC .¢ 0v an mm b” mm on V» nnm.

It, I IIIIII o
ecu-.0) C 05.1 II”’
I,
323: 2.6.5.5.!!! IIIIIIII o.
Q'O'O'o'o "t’
'o'o'o'o ,II
'o'o'o'OUUIII’I ON
alu'o'JII,
'O'lg
on
at
:0 72m
uouiou
om .\.
n9..~c.ow._-ua 222:; on
OO.INL . NV..IuD “050:
N m. an. . mo..-oa “2.3;: ..u o
. . . 0
v0 . E . a .no .
N 50¢ a _ <
Aoeofi .comocam soc» annoy
om

co_.o_:aon fan

a. .0 23:09.30 .23
«0:4 _ a 3 oo.

o 30 coEuanou c. oocucu _ mung...

54

in the bulk increases those genotypes that have agronomic value, and
that no artificial selection is practiced by the breeder upon the popu-
lation. There are several potential flaws concerning both the primary
goal and the assumption.

Allard (1) devotes extensive space presenting the method. Central
to his discussion was the published work of Suneson, et a1. (2,3,4,5).
Suneson (3) began a bulk population study in 1933 in which four barley
varieties were planted, each comprising 25 percent of the population.
Each year the bulk plot was harvested, the frequency of each genotype
was recorded and the bulk was replanted. The study was terminated in
1948 after sixteen generations. The results are displayed in Figure 1
which has been constructed using Suneson's data.

Sixteen generations changed the composition dramatically. Atlas
comprised 88 percent, Club Mariout 10.5 percent, Hero 0.7 percent and
Vaughn 0.4 percent. Grown in pure stands, however, the order of yield
was Vaughn > Hero > Atlas > Club Mariout. Suneson was unable to explain
these results. The answer seemed obvious to this author: despite a low
yield in pure stands, Atlas produced more seeds per unit area than the
competing varieties. I exclude from consideration seedling vigor which
may have strongly influenced the outcome. In any case, even considering
vigor, Atlas had to have produced more seeds for any given harvested
area. Expressed as yield components and over years XYAtlas> XHLM.,V.,H.’
where X = # of spikes per unit area and Y = # of seeds per spike, the
combined component representing seed number per unit area. The reason-
ing was that Atlas, as an old variety, had not been selected for high Z,
or individual weed weight. Its smaller seeds would account for high XY

by virture of component compensation, despite its intrinsic low yield.

55

Reviewing the literature more extensively it was found that seed
number was proposed as the force behind changes in bulk composition.
Harlan gt_al, (6) and Laude et al. (7) remarked on the phenomenon, but
Suneson chose to explain the differential composition of his bulk by
ascribing disease as the cause, despite its comparative absence and
despite the fact that the changes in composition were highly linear, as
detailed by the high r2 values in Figure 1. As a consequence of this
an hypothesis was proposed: The percentage composition of a variety
grown in bulk with other pure lines will be a function solely of its XY
component as determined by the XY estimates taken from the pure stands.

Mathematically, the function would be

where PO is the beginning proportion of a variety, n is the number of
varieties grown in the bulk and XY is the seed number per unit area per
variety as determined by the pure stand. Over time the function would
be P = b (At), where b is the regression coefficient and At.is the number
of years the bulk has been harvested.

Although I have no substitute, the term natural selection seems to
me a misnomer. The bulk population method of breeding supposedly oper-
ates independently of man, but as a consequence of planting and harvest-
ing, artificial selection is taking place--the interdiction of man
occurs. My point is more philosophical, or semantical, than that con-
cerning the dynamics of the population. However, natural selection is

a function of great expanses of time and it is not an isolated phenomenon

56

operating upon a crop whose maintenance, as such, is entirely dependent
upon constant human intervention.

The ramifications of artificial selection based upon seed number
were intriguing enough that two experiments were initiated. One experi-
ment would be with barley and the other with oats. The intent was to
test the hypothesis that a genotype's relative composition in a bulk
population could be determined by XY component analysis of that variety

in a pure stand.

Materials and Methods

 

The constituents of the oat and barley bulks were chosen first on
the basis of easily distinguishable differences in their seed to facili-
tate sorting after bulk harvest. Secondly, an attempt was made to maxi-
mize the potential yield component differences in the varieties based
upon the performance summaries of Michigan oat and barley experiments.
Thus, two barleys were chosen from many potential genotypes because one
was a naked barley and the other a red seeded variety. The former was
named Hulless Vantage and the latter Red Lemma Titan. To these two
six-row varieties was added Coho, a well adapted Michigan two-row vari-
ety with large seeds and high tiller number. The oat composite was made
up of two varieties, Orbit and Hulless Terra. As it was impossible to
distinguish red oats from yellow and white when mixed together, a two
component bulk would have to suffice. Orbit was chosen for its large
seed weight and Terra was the other choice since it was a naked oat and
was adapted, if not directly to Michigan, then at least to Indiana where

it was developed. Terra was provided by Dr. David Smith and the two

57

barley varieties, Red Lemma Titan and Vantage, were obtained through
the courtesy of Dr. Eugene Hockett.

Composite bulks were made of the oat and barley varieties such
that each variety was equally represented by seed number. The oat bulks
and pure lines were packaged 35 grams per seed envelope with five repli-
cations for Terra and four each for Orbit and the composite for a total
of thirteen plots. The barley bulks and pure lines were packaged 30
grams per envelope in four replications, totaling sixteen plots. Each
envelope was planted with a four row belt plot planter. Row spacings
were 11 inches and the plots were twelve feet in length. The plots were
subsequently trimmed to eight feet in length. The barley experiment was
planted in Tuscola County April 23, 1980, with 420 lbs. 8-32-16 fertil-
izer. The oat experiment was planted May 5, 1980, in Ingham County
with similar fertilization. The sixteen barley plots were sprayed pro-
phylactically in late May with Benlate to protect against mildew.

All barley plots were mechanically harvested July 25. Only the
center two rows were taken to eliminate border effect. The oat plots
were harvested by hand August 3. Again, only the center two rows were
harvested and the plot harvester was not used due to extreme lodging in
the nursery. The grain was dried for five days, cleaned and final
yields were recorded for both the pure lines and the bulks. Multiple
samples for each pure line from each replication were weighed and the
seeds in each sample were counted to arrive at an estimate of Z, indi-
vidual seed weight. The replicated 2 values were then divided into the
respective plot yields to calculate an estimate of XY, or seed number

per plot.

58

The oat and barley bulks were sampled three times per replication
and the seeds were divided into their respective classes. These geno-
typic seed classes were weighed and the seed number was calculated.

Thus, the proportions of each genotype in the harvested bulk were esti-
mated. The results are tabulated in Table 1. Analysis of variance was
calculated and can be found in Tables 2 and 3. Chi square values adjudg-
ing differences between the expected and the observed values of XY--

based upon pure line component evaluations--are listed in Table 5.

Results and Discussion

 

Analysis of variance for the oat pure lines was not significant at

p

.05 for XY, but was highly significant for W (yield) (Table 2). Dif-
ferences for both XY ans N were highly significant for barley purelines
(Table 3). Chi square estimates for differences between the observed
and estimated values for XY in the bulks, calculated using analyses of
the pure lines, were significant at levels far less than P = .001 (Table
5). Based upon the contingency analysis from one year's data on two
bulk populations I would reject without reservation the hypothesis that
changes in bulk composition can be predicted from cultivar performance
as a pure line.

Table 2 shows that the pure line XY difference between Terra and
Orbit is negligible; their differences in bulk composition are profound,
with deviation from expected values equaling nearly 22 percent (Table 1).
Seed number per plot, or XY, differences for barley based upon estimates
derived from pure lines were significantly different, unlike the oat

estimates (Table 3). The prediction equation which reflected the

59

Table 1. Component and yield data for oat and barley pure lines and
tnflks,where W = yield in grams per plot, Z = seed weight and
XY = seeds per plot.

 

 

 

 

Oats
Entry Replication W Z XY
Terra 1 428 .0223 19,174
2 410 .0219 18,737
3 405 .0207 19,602
4 413 .0206 20,030
5 416 .0196 21,216
means 414 .0210 19,752
SX 8.6 .0011 950
Orbit 1 545 .0293 18,584
2 534 .0305 17,515
3 578 .0287 20,134
4 .521 .0273 19,288
means 546 .0290 18,880
Sx 22.6 .0013 1,109
Composite 1 453
2 508
3 477
4 52.1.
mean 490
Sx 30.7
Barley
Coho 1 633 .0499 12,692
2 647 .0488 13,264
3 635 .0476 13,335
4 .101 .0461 15,212
means 654 .0481 15,212

5 31.9 .0016 1,096

Table 1, continued

60

 

 

 

 

Barley
Entry Replication W Z XY
Red Lemma Titan 1 565 .0400 14,125
2 652 .0399 16,333
3 695 .0394 17,653
4 219. .0402 17,644
means 656 .0399 16,439
Sx 65.1 .003 1,662
Hulless Vantage 1 398 .0399 10,448
2 398 .0381 10,627
3 425 .0373 11,390
4 _498 .0373 10,934
means 407 .0376 10,850
Sx 12.7 .0004 412
Composite 1 456
2 623
3 672
4 139
mean 597
Sx 96.1
.§!EEQ£X
Crop Entry Bulk Compositions
% Expected % Observed A %
Oats Terra 51.13 29.29 -21.84
Orbit 48.87 70.71 21.84
Barley Coho 33.30 58.63 25.33
RLT 40.18 38.56 - 1.62
HV 26.52 2.81 -23.73

 

61

Table 2. Mean square of XY (seeds/plot) estimates of

Orbit and Hulless Terra grown as pure lines
and of W (yield).

Variable XY

 

 

 

 

 

Source of Degrees of Mean
Variation Freedom Sggare F
Blocks 3 1,526,796
Entries 1 822,404 3.73 n.s.
Error 3 220,532

Variable W
Source of Degrees of Mean
Variation Freedom Square F
Blocks 3 219
Entries 1 34,716 39,40 *
Error 3 318

 

* P < 0.005

62

Table 3. Mean squares of XY (seeds/plot) estimates of
Coho, Red Lemma Titan and Hulless Vantage

grown as pure lines and of W (yield).

Variable XY

 

 

 

 

 

Source of Degrees of Mean
Variation Freedom Square F
Blocks 3 2,691,174
Entries 2 31,237,402 36.03*
Error 5 867,067
*P Q 0.005

Variable W
Source of Degrees of Mean
Variation Freedom Square F
Blocks 3 2,988
Entries 2 81,677 55.83*
Error 5 1,463

 

*P ‘ 0.005

63

hypothesis would place Titan as first in the bulk, with Coho and Vantage
following. The observed values were dramatically different and the most
startling difference was the near extinction of Hulless Vantage in a
single generation.

There is an explanation for the deviations from expected values:
vegetative vigor. This is, perhaps, another way of stating that those
varieties best adapted to Michigan predominated in the bulk, regardless
of the expression of their relative XY components. For example, Orbit
constituted 71 percent of the bulk which was 22 percent more than was
predicted. Orbit has long been an approved variety for Michigan, al-
though it is a New York release. Michigan growing conditions are suffi-
ciently good that certified seed is grown in this state for eventual
sale in New York and other eastern states. For the 1980 harvest year
Orbit was the second highest yielding entry in both rod row experiments
grown in Kalamazoo and Tuscola counties. In two additional experiments
containing advanced generation material and various parents, Orbit was
the single highest yielding entry. The year was ideal for a cool
season, large seeded oat of which Orbit has been a premier example for
more than a decade. Terra is an Indiana variety and although it was
tested for several years in the Uniform Cooperative Mid-Season Oat Per-
formance Nursery it never found a market. Like other naked oats it is
a curiosity of no great commercial value considering the current tech-
nological ease by which oats are dehulled. The mean yield of 78 bu/acre
is respectable, especially in the absence of hulls, when contrasted with
a mean yield of 102 bu/acre for Orbit.

Vegetative vigor is again the logical explanation for the disparity

between the observed and the expected values for XY in the barley bulk

64

population. In this case I suspect that early tiller initiation was the
critical component of vigor that determined very early the eventual re-
lative seed number production. Coho predominated due to its adaptation
to Michigan (it is a Michigan-developed variety) and to its two-row
characteristic. Two-row varieties are necessarily high tiller producers
since the lateral florets are sterile and as such contribute nothing to
yield. The competition provided by an abundance of tillers which were
initiated early would retard growth of less vigorous six-row varieties.
Hence, the virtual elimination of Vantage from the composite occurred.
In fact, Vantage had extremely large heads (Table 4) which would be re-
flected in a reduced number of tillers. On the other hand, Red Lemma
Titan, aside from being one of the most phenotypically stunning barleys
this author has ever seen, had relatively small heads and good tiller
production. In short, this variety showed remarkable adaptation and
vigor considering that it was developed in Montana. The relative order
of the varieties in the bulk after one generation was, in fact, the
inverse of the order of the mean weight of individual spikes.

An apologium: The hypothesis which I proposed and which generated
this experiment was necessarily simplistic. It was, however, easily
tested and the results are not without value. First, there is a lesson
to be drawn concerning the composition of bulks or, under a different
appelation, the composition of multilines. With equivalence of yield it
is apparent that other factors are operating in the population which can
quickly and deleteriously eliminate one of the constituent genotypes,
depending upon events determined early in growth as demonstrated by
Wiebe, gt 31. (8). The use of highly adapted varieties with adequate

genetic markers to determine their survival in composites is a possible

65

Table 4. Estimates of spike weights for Coho, Red Lemma
Titan and Hulless Vantage as determined by samp—

ling pure line plots.

 

Entry Replication Sample 25 Spike Mean Spike
Weight (gms)Weight (gms)

 

Coho 1 29.31
30.85
27.65 1.17
29.35
26.58
26.62 1.10
30.55
31.64
31.05 1.24
25.84
28.39
26.37 1.07
32.00
33.15
34.12 1.32
32.84
32.35
33.26 1.31
39.73
37.55
40.66 1.57
34.75
36.31
38.04 1.45

Red 1
Lemma

Titan

uNHwNHwNHuNHwNI—IwNHwNHwNI-i

66

Table 4, continued

 

Entry Replication Sample 25 Spike
Weight (gms)

Mean Spike
Weight (gms)

 

 

 

Hulless 1 1 56.62
Vantage 2 51.60
3 52.79 2.15
2 1 56.60
2 54.69
3 52.14 2.18
3 1 62.68
2 60.98
3 59.04 2.44
4 1 58.91
2 60.71
3 51.29 2.28
Summary
Entry Mean Individual Spike
Weight Over Replications S-
and Over Samples (gms) x
Coho 1.15 0.04
Red Lemma Titan 1.41 0.06
Hulless Vantage 2.26 0.07

67

Table 5. Chi square test of significant differences
between XY estimated and observed values

in oats and barley.

Oats: Chi square 671.6 with 3 degrees of freedom.

P < 0.001

Barley: Chi square = 1290 with 6 degrees of freedom.
P < 0.001

68

solution to the problem of varietal choice. Preferrable to such a
system is the method of Grafius which utilized bulks at the F5 genera-
tion in which neither adaptation nor extreme artifical selection
occurred. Such bulks are composed of F5 head hill selections which
retain a modest amount of heterozygosity which would eventually con-
stitute heterogeneity. Another point is that extreme selection does,
or can, occur in bulk populations, though the selection is neither nat-
ural nor easily predicted. The final point is instructive: the data
are good educational material for any course in plant breeding since
the scope of the experiments is broad and the explanations require no
little thought.

To bring this chapter full circle I will conclude with Suneson's
original experiment (1). I attempted at various times to contact Dr.
Charles Schaller, the current barley breeder at the University of
California, Davis, and Coit Suneson's successor. I was unable to com-
municate with Dr. Schaller until well after my experiments were conclu-
ded. I did speak with Linda Prato, Dr. Schaller's technician in an
attempt to find specific information concerning Suneson's four varieties.
What I desired was data on the four varieties grown in the same year and
location and for which both yields and seed weights were recorded.
Dividing Z into W (yield) would provide an estimate of XY. Despite
Prato's best efforts to find such records, they do not exist.

In early December, 1980, I had an opportunity to talk with Dr.
Schaller. We discussed Suneson's work at some length and Dr. Schaller
concluded that the results were mostly a function of relative varietal
vigor, although no controlled experiments were even conducted to test

the notion. What was most interesting, however, was Dr. Schaller's

69

recitation of farmer preference. Despite a clear superiority of many
varieties over Atlas in yield, growers consistently and for many years
continued to grow Atlas to the exclusion of newer releases. The phen-
omenon was deemed sufficiently important to conduct a survey of barley
growers with respect to their preference for Atlas. The answer was
simple: Atlas always looked better in the field. Thus are aesthetics

and vigor united.

70

CHAPTER 3--BIBLIOGRAPHY

Allard, R.W. 1960. Principles of plant breeding. John Wiley and
Sons, New York, 485 pp.

Suneson, C.A. and G.A. Wiebe. 1942. Survival of barley and wheat
varieties in mixtures. J. Amer. Soc. Agron. 34:1052-1056.

Suneson, C.A. 1949. Survival of four barley varieties in a mixture.
Agron. J. 41:459-461.

Suneson, C.A. and H. Stevens. 1953. Studies with bulked hybrid
populations of barley. USDA Tech. Bull. 1067, 14 pp.

Suneson, C.A. 1956. An evolutionary plant breeding method. Agron.
J. 48:188-190.

Harlan, H.V. and M.L. Martini. 1938. The effect of natural selec-
tion in a mixture of barley varieties. J. Agric. Res. 57:
189-199.

Laude, H.H. and A.F. Swanson. 1943. Natural selection in varietal
mixtures of winter wheat. J. Amer. Soc. Agron. 34:270-274.

Wiebe, G.A., F.C. Petr and H. Stevens. 1963. Interplant competition
between barley genotypes. In: Statistical Genetics and Plant
Breeding. NAS-NRC Publication 982, pp. 546-555.

CHAPTER 4

THE USE OF OUTLIERS IN BREEDING FOR YIELD IN BARLEY

Introduction

 

The first use of the primary components in explaining yield was
first proposed by Balls (1) and their use in raising yields was subse-
quently proposed by Woodworth (2). Frankel (3) cited the strong
genotype-environment interaction as an explanation for why plant breedens
failed to utilize them. The environment notwithstanding, Adams (4)
offered the most plausible explanation for the difficulties which a
plant breeder encounters when working with yield components: the strong
negative correlations between the primary components would confound
attempts to increase yield by the mechanism of component compensation.
In a similar vein Grafius et_al, (5,6,7,8) delineated the effects of
component compensation in barley which, by extension, would apply to
other cereals. The application of this research in conjunction with
regression analysis led to an experiment (9) which highlighted a method
of isolating genotypes which, in proper crossing combinations, would
produce selected homozygous progeny transgressive to both parents for
yield. The selected pwogeny, cited as 1 x 42 (9), was Michigan line
68-105-15 which was eventually released as the variety Bowers, an ex-
tremely high yielding feed barley (10,11,12,13). The confidence engen-

dered by the development of Bowers, coupled with an extended knowledge

71

72

of the use of yield components, provided the inspiration for this

experiment.

Materials and Methods

 

The mathematical identity X Y Z = W, where X = spike number/unit
area, Y = seed number/spike, Z = weight/seed and W = weight/unit area
means, simply, that yield is determined by its primary components.
Hence, any yield differences between genotypes must be reflected in
their determinants, X, Y and Z. The primary problem lies in assessing
differences with respect to the negative correlations between compon-
ents, or component compensation. The method of Grafius (9) to ascertain
the extent of these differences was through the simple expedient of
linear regression.

Data from 1978 barley experiments were plotted for Z (seed weight)
against Y (seed number per spike). Of eleven entries three were "out-
liers" (Figure 1) in the sense that they appeared to lie above and to
the right of the axis of the eight remaining entries. Two of the out-
liers, 68-105-1 (Michigan) and M33 (Minnesota) (see Nos. 5 and 11,
respectively),were each crossed to the four genotypes having the high-
est values for seed number per spike, Y. The constraint on increasing
the number of crosses and utilizing the third outlier was that of
greenhouse space needed to simultaneously advance no more than eight
bulks through two generations of single seed descent (14). The crosses
are detailed in Appendix 1.

The seven surviving F1 lines were increased to F2 and subsequently

grown for two generations under single seed descent, producing F4 seed.

gms/IO3

 

73

 

1 FIGURE 1a Regression of 2 onto Y for Possible Parents
0 11
O 1
45.0.
9 5
Z = 46.8 - .1Y
n . 8
44.0 1
r2 ' .24
r - -.49
Y’2
o 2
43.0
seeds s‘~ '9
-
42.0 \\
-
-
-
-
-
-
“-
.10 .8 \
-
-
-
41.0 . ‘94
o c»5
4 03
40.0
39.0
40 45 50 55 60

Y seeds/spike

74

 

 

 

FIGURE 1b Regression of 2 onto xv for Possible Parents
and Their Respective Yields.

ENTRY YIELD (bu/A) RANK

1 89.5 3

2 91.1 1

' 11

.1 3 91.1 1

“5'0 4 90.0 2

5 87.0 4

. 5 6 85.4 5

7 84.4 6

8 83.5 7

44.0 9 83.3 8

10 82.0 9

Z 11 81.3 10

5mg
103 seedaI
43.0
i .9
42.01
| Z 3 47.1 - .07XY
n = 8
‘ r2 - .26
\“: 10 .rxy.z- -.51
‘s
o 2 “-
41.0 ¥~
‘
7 ‘x‘
.6 ‘\‘ .4
‘\
‘\
40.0.
'3
800 850 900 950 1000 1050

XY seeds/ft2

75

Approximately 1,000 seeds of each bulk were planted in single plots at
East Lansing in the spring of 1979 for seed increase. The plots con-
sisted of four rows on eleven inch centers, eight feet in length. They
were fertilized with 400 pounds/acre 8-32-16 and the center two rows
were harvested using a plot harvester (15).

F5 seed of each bulk, all six parents and seven checks were planted
in Tuscola County April 24, 1980, with 420 pounds/acre 8-32-16. Plots
were four rows, eight feet in length with eleven inch row spacings. The
twenty total entries were planted as a 4 x 5 rectangular lattice with
four replications. The soil type was a Parkhill Clay Loam.

The experiment was harvested July 25 and yield component data were
taken or all entries. Only the center two rows were harvested for
yield. The grain was dried in a greenhouse for four days and subsequent-
ly cleaned before data were recorded. Table 2 summarizes the results
and Table 3 contains the analyses of variance associated with the traits

of interest.

Results and Discussion

 

The expectation upon which these crosses were based was that the
mean yields of all seven bulk populations would exceed all of the
respective parents. Table 2 and Figure 2a display the remarkable re-
sults. There are four points of importance concerning these results
which need to be stated. First, of the seven bulks, five outyielded
all the parents. Secondly, the two bulks which did not outyield all

parents had, nevertheless, very respectable yields. More importantly,

however, these two bulks (879-105 and 879-106) were associated with

TABLE 1a.

8095 PULVS

EXPEPI'FNT

f A V

U N A 0 J U E T E D F

O F

"LC
nRFFPI

T A
12 IN DESCENDING

ISORTED 0N VARIAELE

V A R I A 8 L E

U GHS

OHS/SE

I
--------------- P-------------

PIKE

NU)
F1\
(1)

Y SEED

31
X TILLEFS/SO FT

ENTRY NAHF

OOOO")
0609"?
cocoon
GOOD!)
0 O O O O
annomn
NOOO‘U‘
CCCDF

Odo-96‘
ONU‘OC
COW.”
OOOOO
O G O O O

CIDITICC

76

nO‘CO‘U‘
HOOGO‘
00".”
000°C;
0 o o o o

OQUInc

ohnna
COP)“:
corona
con-no
O O O 0 O
ONCNv-u
nNu-onc-n
hhhhfi

lfianO"
JFCOQ
nfihfilfl
U‘OC‘NO
o o o o 0
nor-cc
fihimno

cannnm
mwmwnc
Ncuwnm
ouncnc:

coo-o
wumnmo
Ho‘flfi"

«0105.11

0

y.

I

C C
O U
n (D!
I “43¢.
Ono-tn
“I'D—l:

PROCEDING 20 MEANS

OF VAPIATION

F
o 5)
o 1)
1 NI

0
0
0

Bolts
0 Parents

TABLE 1b.

"PLVS

Fnos

EXPERIMENT

" E A N S

' T F D

J U

u n o

u

8’

r L r o
DIMF ochre)

I‘

‘FSCE“

1’-

I?

(SORTED ON VARIAPLF

V A R I I R L E

--------------------------------------------------------.---

11

HT

20 H0

30

qU/ACRF

‘3
SEEDS/PLOT

IV

11

THILP/PU)

ENTRY NAME

66°65
0060\0
DOOOO
anoint
O O O O 0
Odhbnh
“.08).

90°05
ODC‘U‘J
Iaocn~m
com-0.4
o o o o o
P184900
GDGOU‘
none-I

OOUOO
Fen-or
meow—-
”NCC—I
O o o o o
DOOR-0‘
“WOO.

CC-F-mfl.

77

”09000
Grant:
Int-math
OGFON
0 o o o O
OTC???
OOO‘OO

Irommo

“whit, U

0!")th
mono-4.4m
bnfhth
”0'1an

000..
NOWU‘C
DODGE;

cacao
IDhHC'ON
O-twnfi
O‘Du‘mr
o o o o o
cacao
ommnann

”F nhn
numwan
Crammer.
ocU‘ND
on .00
CAD-ONO
camera

68-10 0-7

O‘O‘FC
bun—N
0&0”
cmnn
I o o a
«mucus
O‘Ro-

PROCEDING 20 MEANS
OF VARIATION

TABLE 1c.

9095 pULI‘S

EMPERIHFNT

M E t N S

0

E

J

N I

LI

0 F

H L E
ORDER)

T n
12 IN DESCENDING

ISORTED 0N VARIABLE

V a R I A B L F

LODGING

HEIGNT (INCHES)

ENTRY NAME

00
33
00
00
33

CJF'IDOF
ONO Om
DV‘OOQ
an: CO

.00 00
CONS—O
cahtcr4

78

nhcon
”00°F?
”85°C”
manor.
0 o o o o
xc«:om
NNNNO

nnbno
”memo
I'm-mom:
nnwnc
o o o o o
momma-o
thereto

'1

In
C
I—

3

0
III €31

10 4-7
R

"I
FICDZI
”CF on
f mmmm

O
Iwzct
cro<n
WIUJI

|VERAGE OF PROCEDING 20 MEANS
(P2005)

LSD

(P3001)
COEFFICIENT OF VARIITION

79

xcmooco x—om new mucmcmo

0 HH m HH 0 Ha v m

.o_m_> com moonstomcwo Acmooco x~oo one Foucmcmo

HH m

 

mm mmome

om

mm

om

mo

ooH

mod

meum\:o

80

high degrees of lodging (68 percent and 47 percent, respectively). It
is not inconceivable that this lodging contributed to a yield reduction
which, when adjusted for the loss by machine harvesting (15), might have
ranked these bulks above the parents, as well. Thirdly, it must be
borne in mind that in the development of these bulks, that no selection
for yield had been applied. Fourthly, these bulks are collections of
homozygous genotypes which assuredly are normally distributed with re-
spect to the polygenic trait of yield. It is logical, therefore, to
assume that each bulk contains pure lines which are transgressive not
just to their parents, but with respect to the high mean yield of the
bulks themselves. The practical consequences of this are obvious.

With the exception of the late Dr. John E. Grafius, there has been
no corpus of theory which conveniently explains these results since they
are predicated upon the primary components of yield. The difficulty of
explanations using yield components is perhaps best summarized by the
following analogy. Yield components are like eating Jello with fingers:
no sooner does one get a grip in two dimensions, the Jello squirts off
into a third dimension. It was precisely this plasticity or compensa-
tion or collection of negative correlations which this experiment was
designed to circumvent. Yield cannot be explained by one or two com-
ponents, but must be presented and dealt with holistically. Figures
2b through 2e demonstrate this point.

The seven bulks are represented in histograms nested between each
of their parents for the traits X, Y Z and the combined trait YZ. The
individual graphs of X, Y and 2 do not explain the yield performance of
the top five bulks in a consistent fashion. The combined trait, XY

(Figure 2e), however, does. Of the five superior bulks, all were

81

acmooco xpao ucm mucmcmo

m m 0 HH n HH 0 aa e m n HH

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.LmaE:z mxwom no

LQFZF Low mucmELOWLmQ XCmDOLQ x—sm Ucm Fauthma 2N “$.2an—

o.mH

m.mH

.o.oH

-m.oH

.o.mH

m.mH

 

.o.wH

Nsc\mta_2ap

82

zcmooco xpoo new mucmcma

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o fiH N fig o HH e m n HH m
om
mm
.L. F.Il I, ,III. ll . ON.
11L
[II1IL wlL - mm
[1". It
1.- ow
2 mm

.mxwom Lmo twosoz comm

com mucmELowcma xcmooca xpoo vow Papoose; um mmoon

mxwom\muwmm

83

 

 

 

HH

xcmooco xpom vow mucmcmo

o HH

HH

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_r1L

.uzowmz comm com moonstomcma xcmooco xpom ocm Faucmema

 

 

 

 

Um mmome

 

 

0.0m

-o.mm

o.wm

-o.mm

- o.o¢

.o.H¢

. o.~e

-o.m¢

 

- o.¢¢

mamas moa

84

y.

Acmooca xFom ucm mpcmcmo

0 HH m HH m HH ¢ m

Ha

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.N> .saare smeaanu

mop cow mucmscowcmo Acmooco xpoo new Fmpcmcma

 

 

 

 

 

mm mmoomm

o.m

m.m

w.m

o.N

- o.m

H.m

 

m.m

N>

85

superior to each of their parents with the exception of 879-104 which
was superior to one parent and equal to the other. This is not surpris-
ing since the regression upon which the crosses were based was that of
Z onto Y. Net superiority in the progeny with respect to both Y and Z
should necessarily result, provided that the regression related to
heritable traits with a genotype-environment interaction which was
reasonably consistent among all six parents. In other words, the re-
gressions of yield components upon one another should strongly reflect
true genotypic, and not merely phenotypic, performance. The exclusion
of the trait X is not without reason since its expression in the bulks
with respect to the parents is seemingly random (Figure 2b), which shouki
be a consequence of the manner in which the crosses were constructed.
Malting barley requires a high percentage of plump seed to command a
premium in the market. For that reason, Z was a yield component of
major importance in the breeding goals of the Michigan barley project.

A causal analysis of Figures 2a through 2d and Figure 3 would
suggest several points concerning the genetics of these crosses. Taken
separately, the expression of the traits X, Y and Z are randomly dis-
tributed among the bulks with respect to their parents. Some are mid-
parental, some are superior and some are inferior. Thus, additivity
cannot be excluded as a mode of genetic action (16). Neither may
epistasis (non-additivity) be excluded since five of the bulks out-
yielded all parents. Only dominance and its interactions can safely be
excluded from consideration since these bulks were approximately‘96
percent homozygous. The author is reluctant to draw further conclusions
for several reasons. This experiment was not constructed as a genetic

study to extract the various components of variance and the application

86

 

 

FIGURE 3. Regression of 2 onto XY for Parents and Progeny
Grown in a Eamon Experiment.
43.0 i
\
\
42.0 \\ ‘7 x 5
\
\\
Q \
3 x 5 \
\ o
\ 6
\\
41.0 " . \ .
\
\
\
‘\\ I 3 x 11
\\ O
40.0 * 6 x 11
\\\
\
6 x 5 \\
\
\
o 3 \ 9
39.0 \\4 x 11
Z - 56.7 - .14xv “°
31% n = 13 7 x11
10 r2 . .53
seeds xy-z ' "‘73
38.0 .
37.0 .
O
7
36.0 T 1 T
1000 1100 1200 1300

XY seeds/ft?

87

of the quantitative genetical lexicon to this study is largely inappro-
priate. With exactitude, one cannot even describe the yield performance
of the five superior bulks as being transgressive since this term, as
originally conceived, applies only to pure lines (17).

What might be concluded is that this system of identifying yield
component outliers and the genetic variability they impart to crosses is
the first potential technique to aid cereal breeders in parental selec-
tion where yield is the trait of greatest interest. It might also be
proposed that when outliers are found they represent a second population,
distinct from the first, or main, population, despite similarities in
yield. The system might also be extended to other species as well since
it offers a plausible method for predicting specific combining ability
in corn and other hybrid crops. This experiment has been planted
spring, 1981, to improve its resolution and to determine its repeat-

ability.

88

Appendix1.N0tation for Genotypes of Possible Parents, Parents

and Bulk Progeny.

 

Notation Genotypic Designation
1 Michigan 68-104-14

2 " 68-106-9

3 " 68-103-1

4 " 68-104-7

5 " 68-105-1*

6 " 68—105-10

7 " 68-105-15 (Bowers)
8 " 68-104-21

9 " 68-104-14

10 Larker

11 Minnesota M33 *

6 x 11 879-101

7 x 11 B79-102

3 x 11 B79-103

4 x 11 879-104

3 x 5 879-105

6 x 5 B79-106

7 x 5 B79-107

*Outliers

89

«nhmmhh08n

owrmhoc mwmmwwnxomm
nwhmrncco

u~Fm~pchm u munqncw 24m:

bu owxmmuaa—h x

L a m d h

on
«r

«a
N

zoouumu
uc ovcuc

h2uu cum pontoon

20-4~¢<> no hzu~u~uuuou

coupon "0mg «a. «a
meow-N «am; was un

awbcznxocaacu uuuzwauuu~o hz<u~unzunm hm¢ua

nmuhOhmmoth 4(h0h
mmocnomoawnn xuoqm upmaazou cum~t092¢a
KO¢¢E
mwmmawmmomom ouhmafiocz:
munchzu
mmmnmmcmo mzo~h¢u~4uu¢
mma¢30m no tam uuz<—¢«>
no guanow

n» wac¢~¢¢> hzuozunuc

zwnmuo xuoax cum—SCOZ<
E u 2 q n a < ) c

a «on
u m u w > 4 a 2 a

anzm mcoc hzux~¢ukxu

.mm x~ozmaa<

9O

h2mu awn Chum-m

zo~r¢~c¢> kc h2u~u~uuwcp

anouonu new; no. no
«chooOn Mama mo. "a

awhct—xouanma mwuzuauuu—n bz¢u~u~20~m hmcma

cu com—sooo.womh dope»
mocm~m0o88m an ocnnommo.hoou vuoom upmaaxcu oumnxoozac
uoxac
mmrm¢.o «moonwamcmnm an onmcwsogoromw cupuasocza
muunhrt
omouosho.~ o mnccocmu.o wzo~paunanua
Unwwgpchm u muoqnom Zaur sooouam mmccaom to saw uu2¢~mq>
kc ocomp no ucaaow
cx~amxmoEum > m— uoaa~a¢> prooZanL

zc—mcr xuc4¢ Qmo~roozca you
u 4 m ¢ » L u 2 a ~ a 4 > u c m ~ u > a u 2 <

mxmoa «mom p2u:_aunxu

.aN x~ozmoa<

91

hzuu can Foch-w

zo—h<~¢¢> no h2u~u~uuwoo

cwooo now; «a. nu
maooo now; no. no

nuhnx—xozma<~ muuzuauuuun bz¢u~u~zo~m hmcma

mm nhoooo000 ache»
«uncooooo av camcoooOo xuoac whuaaxou cum—IOOZ¢¢
mezzo
on—naonn omnNooooo on nmuncoooo ouwwafio<23
mu—zhzu
amooooooo m mnnoooooo mzo~h<u~anm¢
o—pm—pqpm u wuccoow Zoo: tacomau mucuoow mo tom uuz<~¢<>
go occuc uo wcaoow
ouumxmxm N m~ mox¢~¢<> >2uozuauo

Zonmmo xuodm oum~toozo¢ mom
L a a a p u u 2 4 ~ a c > L o w h u h 4 q 2 q

wxoo: mmom h2ut~1u1xu

.um x~ozmam<

92

onhcocmbonomm

Lagoo8~ upcmcscm.wsnp

Choooo—aooae—
onpmmpqhm u wbcqncw 2am,

Lac 1

m a m a h

k

bzuu mun cwnmoo

zo~h1~¢<> no haunonuuuoo

wmocoonn no“; no. no
cracooon n no no. no

ourcx—x0a0n<o mquKcuuu—o pz-unuuzunw rmoua

mu mmnnnm¢Oocmn¢Gn done»
an «amatocmom0CCcn xucam upmantou CumutOGZQc
couaL
aw «Nooocmoormmnmu DurwaocQZD
mw—KhZE
n mmnnnnmwocuum m20~>¢u~4nuu
aooumwu mmoaaom no 13m UUZ¢~¢q>
uc oucuc uo uucacm
Wm w41d~¢d> bZkDZmQUL
a.“ . . . . .

wxobm ucou rigs—auaxu

.um x~ozwao<

93

«OohOonn

u-m-<~m n

monmmhwno

owwooonnom

watchmcoo
wua<ocw 2<ur

nDn\EJ-3p

m a m c »

ammo.
sumo.»

 

“whctmxomnn¢u m

we unsoooCOoon

or ocmuommnow

ea huncmoonoco

m mnoccmco.
toouuun mmacoow no tom
no odomo

m— n4m¢~¢¢> hauozunno
2c~moc xuoaz cuv~toozda «on
u u z a — a u > n o m n w

bzuu mun Onnoo

zo~»¢~¢¢> no haw—ounnuou

now; no. no
no.3 no. no

uuZuzunnuo p2¢o~n~zo~w pmmuo

ambOh
xucao upmanxou oumutoozoc
zoxcn
oubwDoO¢zo
muncpzw
mzo~h<u~anu¢

uuz<~¢¢>
no uoaoom

r a d 2 a

mxaon meow P2ux~¢unxm

.mm xHozmon<

94

encomoc

onhm~bahm n

mrwnmnuaouwcnanm

nmmwoooo.omochhcn

Nancohumoncwcrc
wucqncm Zdur

hoonxnoncm rx

m A a c h

or
an

an

Al
IOGLna
nc ommw

m— u

2c~nno

n u z c ~

wcuwohuw
nchowmcw

h2mu «on coonoc

zo~h¢~n¢> no bzw~u~nnwou

"o. no
mo. no

.uh‘t—xounn<w mnuzumwnnuo h2¢u~n~zo~m hmoua

ottomanfiowcnwwcmmw o¢hoh
mcncmwnwowwhommho xuoom upmaoaou cumnzoozow
«ozun
accnmnuuounuhohccw ouhmaooazo
mu~c~2u
romnocpcowacnex Macmkcu—Jnuc
c unacoow nc 23m uuz¢~aa>
a no uncocw
oc¢~a¢> FZCCZnnnc
xuooo ouv~sonzaa out
a < > n c w p m > a a 2 a
3,3433 «.2ch hauswuuaxm

.mm x~ozuno<

95

bzuu znn cannon

zo~h¢ucc> no hzu~u~nnunu

huonooa now; «a. un
omomowu no“; no. "a

auhct~xounoto mquuuunn~o b2¢uunnzo~m hw<m4

mm accommomommoc oopo»
oncnowocoom on omummupwohmmm xuomo Upuanzou cum—tooz<c
acacn
nummoom womnmwnoocmu cu mmuohmmo.mmn~ omhwouo<zo
mu—xhzn
«coonovaou— m coomnoomomn mzo~p¢u~4nmz
u-m~pqhm n moa¢ocm Zeb: recuuan wmacoom nc tom uuz<~¢<>
no aucnc no nuaDOm
nau¢xam m» m4m¢~¢¢> h2m02unuc
u 4 : a h n aubuqn—chwn>cmmwsmo2<m we” > 4 < z a

mxaoz mace bzuznzunxu

.mN xHQZMaa<

96

mnonoownoon

mc—chow occoxnowomo

somJHUVho—
ohhm~hchm n wncdaow 2cm:

>3 (1 on

n o m a h

00
In

an

r.

3L0um~u
no 006»;

hZUU awn hxumoo

2o~>¢~¢<> no pronounnuon

osOOon now; do. no
camnow not; no. u

outer—xownhou nnuzuawnnnc b2qU—nnzohm bwauo

cmrnc~m~.whnu dupe»
moomono>.ocm xcomo nhmoaxou cowoxoozaa
«oxen
”cccrama.sncfi ouhwoaoazo
amuuhza
cmnnmc~u.r «zowpcu—onaa
mmaaoom no :30 uuzchaao
no occoom

w» won¢nac> penczuawc

zc—cic yucoa rmu»:
n o z a o 0 c > u

$34.7. pccoo h21.1~1..&>..

.zm x~ozmno<

97

ounowmmmom

nncnoow mnomnonwoua

soocomumom
u-m-¢~m n moadocw Zaur

awozoz—o prouur

u 4 m a h

an
or

o“

N

toouumn
no crow:

bruu cun nunooo

zo~><~¢¢> no h2u~u~nnuou

ow-.c new; no. no
2.3.5 "or... mo. un

«upcxnxocna<o vuuzwuwnnuo hznuunnzonm hmoua

nnnnnnnnocpn Jcbo»
cocoooonoonu xuoom upmantou oumn1002q¢
«oxen
oooooooOonnN oohmofio<zo
munzhzn
nnnnnnnOouu mzo~>¢u~4nm¢
mmucoom no tom uuzcnxc>
no nncDom

m» uam¢~c<> hzu02unnc

zemmoc goose oum_amo2.u soc

LU.¢¢~Q¢>

u o p c r A a 7 a

wnab: moon hzuznnwnxu

.wm xHozmon<

98

Nnonmoawoohm

cmcwmom wommcnonomno

ooooooooomcr
Unhmnpqhm n monqoow anr

ozncnco

. a n a h

L

I
no

on

m

COnuan
ovcuc

hzmu «no Nonuoom

zo~h¢~ac> no >2n~unnnwcn

omnc.rn nova "o. no
soc~.>~ "on; no. nn

anhctnxonnn<o mn02unmnn~n h2¢u~nnzonm hwqua

ooooocoo.m~—nn ache»
ooooooooocumcfi yuoom upmantou ouwntcozqu
comma
ooooooooomnom— curmosoazo
wmhahzu
occoocooocac mzohwcunanma
muncoom no tom uu2<~u<>
nc nca20m

v» noac~a<> ~2u02ncwc

EQ—ULC 19641 Enemicozca nCn

o

z a — a c >

n G c F V > J o 2 a

wsooe Lac: pznx~aunxm

.nm x~ozmno<

10.

11.

12.

13.

14.
15.

99
CHAPTER 4--BIBLIOGRAPHY

Balls, W.L. 1912. The cotton plant in Egypt. MacMillan, London.

Woodworth, C.M. 1931. Breeding for yield in crop plants. J. Amer.
Soc. Agron. 23:388-395.

Frankel, O.H. 1935. Analytical yield investigation on New Zealand
wheat. II. Five year's analytical variety trials. J. Agric.
Sci., Camb. 25:466-509.

Adams. M.W. 1967. Basis of yield component compensation in crop
plants with special reference to the field bean, Phaseolus
vulgaris. Crop Sci. 7:505—510.

Grafius, J.E. and R.L. Thomas. 1971. The case for indirect genetic
control of sequential traits and the strategy of deployment of
environmental resources by the plant. Heredity. 26:433-442.

Thomas, L., J.E. Grafius and S.K. Hahn. 1971a. Genetic analysis
of correlated sequential characters. Heredity. 26:177-188.

1971b. Transformation of

 

sequential quantitative characters. Heredity. 26:189-193.

Thomas, L. and J.E. Grafius. 1976. Prediction of heterosis levels
from parental information. Eucarpia. 7:173-180.

Grafius, J.E., R.L. Thomas and J. Barnard. 1976. Effect of paren-
tal component complementation on yield and components of yield
in barley. Crop Sci. 16:673-677.

Leep, R.H., L.O. Copeland and J.E. Grafius. 1979. Bowers barley.
Michigan State Univ. Extension Bull. E-1314.

Timian, R.G. (Compiler). 1976. Mississippi Valley Barley Nursery
Report. USDA, ARS, NCR, N. Dakota State Univ., 29 pp.

1977. Mississippi Valley Barley Nursery

 

Report. USDA, ARS, NCR, N. Dakota State Univ., 32 pp.
Dreier, A.F., J.W. Schmidt, L.A. Nelson and R.S. Moomaw. 1980.
Nebraska spring small grain variety tests. Nebraska Coop.
Ext. Serv., E.C. 80-102, 27 pp.
Grafius, J.W. 1965. Shortcuts in plant breeding. Crop Sci. 5:377.

Wolfe, D. and J.E. Grafius. 1965. A combine for small grain nur-
series. Crop Sci. 5:376-377.

16.

17.

100

Grafius, J.E. 1965. A geometry for plant breeding. Res. Bull.
7, Michigan State Univ. Agri. Exp. Sta., 57 pp.

Darlington, C.D. and K. Mather.
Allen and Unwin, London.

1949.

The elements of genetics.

CHAPTER 5

CHEMICAL INDUCTION OF GENOME ELIMINATION IN SOMATIC
TISSUE OF A WIDE HYBRID WITH BARLEY

Introduction

 

Barley (Hordeum vulgare) has, as its primary agronomic weakness,

 

low winter survival in cold-stressed environments (1). Furthermore,
the prospects for improving winter hardiness through the exploitation
of the remaining intraspecific variability are exceedingly poor as
stated by Grafius (1) and as judged from other reports in the litera-
ture (2,3,4,5,6). Winter hardiness is by no means the only polygenic
character for which barley might be improved. Other goals include
further improvement for salt tolerance and resistance to other stresses
such as puddled soils. Mendelian traits such as disease resistance are
also of interest.

The barley wide hybridization program at Michigan State University
was initiated by the late Dr. John E. Grafius to find genetical solu-
tions to these problems by incorporating alien genetic variability in
the barley genome. The problem has been a particularly difficult one
in view of the many man-years expended to the purpose at this institu-
tion. The author has been intimately associated with this effort for
four years and he has experienced the many frustrations that have been
the consequence. On the other hand, progress has been made, not so

much in terms of the tangible improvement of barley, but more in terms

101

102

of finding solutions to the many barriers to successful genetic transfer
from wild species.

The research herein reported is narrow in focus and deals exclu-
sively with the wide hybrid between H. vulgare and H. jubatum. What
will be reported are the results of the author's previously reported
efforts to effect somatic recombination by three means (7). Addition-
ally, this research will detail the use of three chemicals intended to
induce genome or chromosome elimination in the wide hybrid with the
intention of recovering haploid H. vulgare (2n = x = 7) which, upon the
doubling of the chromosome complement would restore fertility and allow
the introgression of wild germplasm (if present) into the cultivated
population of barley. The necessity for this system has been previously

reported by the author (7).

Literature Review

 

The value of barley (Hordeum vulgare) as a genetical research organ-

 

ism is not necessarily correlated with its agronomic value. Its diploid
state (2n = 2x = 14) provides for easier genetic studies than those in

wheat (Triticum aestivum) which is an allo-hexaploid (2n - 6x = 42).

 

Wheat has distinct advantages over barley, however, and these advantages
stem from its polyploid nature. There are three distinct genomes in
wheat, each derived from a different species. A deficiency for an agron-
omic trait in one genome may be compensated by genetic expression for
that trait in another genome. Harlan (12) has termed this "buffering"
and it is not found in barley. Buffering has advantages other than

genomic complementation for specific genetical traits: the existence

103

of multiple genomes is what allows the manipulation of one genome, yet
insures adequate physiological response expressed by the undisturbed
genomes.

Polyploidy in wheat and its physiological stability when disturbed
is a valuable condition utilized by cytogeneticists and plant breeders.
The greater economic importance of wheat has also fostered more inten-
sive efforts toward its improvement. The gene transfer systems of
Riley (8), Sharma (9), Knott (10) and Sears (11),‘gt_gl,, are achieve-
ments as yet not accomplished in barley. Barley suffers a paucity of
closely related wild relatives which might serve as a convenient source
for additional variability. Regions of extensive homoeology between
the barley genome and those of other species are rare (12).

Despite these difficulties there have been several reports of suc-
cessful introgressions of wild germplasm into barley. Hamilton gt_gl,
(13) and Schooler gt 21. (14) transferred disease resistance through
wide crosses and Schooler (15) recovered male sterile lines by similar
methods. These few reports represent the extent of successful gene
transfers. No success has been reported for other traits such as insect
resistance, resistance to other diseases, enhanced salt tolerance, or,
most importantly, for improved winter hardiness.

Success in the author's wide hybridization program is contingent
upon the chemical induction of haploid cells in somatic tissue of barley
and this has been variously reported (16,17,18). It has also been

achieved in tissue cultures of a medicago hybrid (20).

104

Materials and Methods

 

Chemical Induction of Chromosome Elimination

 

I. Chloramphenical (CAP)

 

Chloramphenicol is an antibacterial and antirickettsial drug de-

rived from cultures of the soil bacterium Streptomyces benezuelae (16).

 

Aside from its pharmacological properties, CAP is known to be a chloro-
phyl inhibitor and, in at least one case, it has been shown to induce
chromosome elimination in barley root tips (17). Clones from the

perennial hybrid Hordeum vulgare x H. jubatum (2n = 3x = 21), known as

 

VJJ', were treated as follows:

Chemicals and Concentration

 

 

 

Date_ Number of VJJ' Clones CAP DM§Q_ Duration of Treatment
4/9/80 12 0.39/1 2% 2.0 hours

" " " " 3.0 hours

" " " " 4.5 hours

Healthy clones of VJJ' were subdivided into sections of 1.0-1.5 cm.
in crown diameter. Fifty percent of the root tissue was removed and the
clones were then placed in culture tubes into which the solutions were
poured until the level was 2 cm. above the crown. The treatments were
conducted in the greenhouse under supplemental florescent lighting and
the clones were removed according to the time schedule listed above.

The plants were thoroughly washed, trimmed of 50 percent of their leaf
matter and potted in 4 inch pots with fresh soil. All pots were labeled
and the plants were watered daily and fertilized once a week for the

duration of the study.

105

II. Para-fluorophenylalanine (PFP)

 

P-fluorophenylalanine is an antibacterial agent (16) which has been
shown to induce chromosome elimination in root tips of a Ribes hybrid

(18) and in the wide hybrid Festuca pratensis x Lolium multiflorum

 

 

(2n = 4x = 28) (7). Clones of VJJ' were treated as follows:

Chemicals and Concentration
Date Number of VJJ' Clones CAP_ DMSO Duration of Treatment

 

4/9/80 12 0.39/1 2% 2.0 hours
" " " " 3.0 hours

" " " " 4.5 hours

The mode of treatment was identical to that of CAP.

111. Griseofulvin

 

Griseofulvin is an antifungal and antibacterial chemical derived

from cultures of Penicillium griseofulvum (16). It has been shown to

 

induce abnormal chromosome numbers in plant cell cultures (20). Use of
this chemical is confounded by its virtual insolubility in water. On
5/9/80 a solution of 0.3g/1 griseofulvin plus 2 percent DMSO was made.

A white flocculent persisted under stirring, so the solution was placed
in a cold room (griseofulvin is unstable at room temperature) for three
days of continual stirring. On 5/12/80 the solution was filtered to
remove the still-considerable flocculent. The treatment was identical

to that of CAP, except that 43 VJJ' clones were treated and the duration
of the treatment was 24 hours, due to a low concentration of griseofulvin

in solution.

106

IV. 2 Percent DMSO (Control)

 

Twenty-five clones of VJJ' were treated similarly to the CAP
treatments with only 2 percent DMSO. The date of the treatment was
7/20/80 and the duration was 4.5 hours. All other untreated clones

were also regarded as controls.

V. Solubility of Griseofulvin in a 2 Percent DMF Solution

 

A report in the literature by Nutti-Ronchi indicated that as much
as .3 grans of griseofulvin is soluble in one liter H20 (20), This report
ran counter to the author's experience. This experiment was devised to
help quantify the treatments.

Communication with Sigma Chemical Company of St. Louis, Missouri,
informed the author that at least 50 mg. griseofulvin is soluble in
1 ml. dimenthylformamide. The author was informed that a precipitate
is not at all unlikely when attempting to dissolve griseofulvin in H20.
The solubility of griseofulvin in a 2 percent DMF solution was then
undertaken.

Twenty ml. of DMF was placed in a clean, dry 50 ml. beaker.

0.3000 g of griseofulvin was added and magnetically stirred for 15
minutes. All the griseofulvin went into solution. This solution was I
added to 980 ml. of triple-distilled H20 and magnetically stirred for
ten minutes. A precipitate, white in color, formed over the period of
ten minutes.

A piece of Whatman #1 filter paper was weighed and placed in a
Buchner funnel attached to a vaccum apparatus. The solution was fil-
tered through the filter paper; the paper was then removed and placed

in an oven at 600 C until dry. The paper was then reweighed. The

107

procedure was repeated three times and the results were as follows:
Grisefulvin not in solution = .2022 g i .0037 9. Therefore, approxi-
mately 90 + mg. of griseofulvin is soluble in a 2 percent DMF solution.
However, it should be remarked that the filtered solution had a fine,
white precipitate in it the following day after sitting for 24 hours at

room temperature.

VI. Griseofulvin Treatment of Six Diploid Cultivars of Barley

 

Approximately three hundred seeds of each of the varieties--Bowers,
Munn, Coho, Manker, Larker and Morex--were immersed in tap H20, supple-
mented with 5 percent sodium hypochlorite. The seeds were in 250 ml.
Erlenmeyer flasks and were drained and rinsed three times after 24 hours.
The sodium hypochlorite was added to suppress fungal diseases during
seed treatment.

A saturated solution of griseofulvin (approximately .O9g/liter in
2 percent DMF and 2 percent DMSO was filtered and added to the six
flasks, 125 ml. per flask. The treatment time was 24 hours, after
which the solutions were drained and the seeds were rinsed three times
in tap water.

The flasks were removed to the greenhouse and remained there for
another 24 hours to better identify those seeds which had germinated.
The criterion was radicle emergence.

Each variety was planted in 20 pots, split in four groups of five
pots each and these groups were randomized on the greenhouse benches.
Bowers, Munn and Morex were planted three seeds to the pot and the other
varieties were planted more densely, due to a reduced radicle emergence.

The plants were watered daily and fertilized weekly until complete matur-

ity.

108

VII. DMF Toxicity Study in Barley and a Wide Hybrid with Barley

 

Ten clones of VJJ' and twenty-five barley seedlings of approximately
six inches in length were placed in a 2 percent solution of DMF for five
hours in a greenhouse under florescent lighting. The clones and seed-
lings were immersed approximately one inch above their crowns and, when
removed from the DMF solution, they were thoroughly washed, had 50 per-
cent of their top growth removed and were then repotted. They were
watered daily and fertilized once a week. There was no mortality
resulting from the treatments and the twenty-five barley plants were

thrown away at maturity.

Results and Discussion

 

Table 1 summarizes the results of the three treatments and the con-
trols. The haploids arising after the griseofulvin treatments merited
statistical analysis. Therefore a contingency test was conducted test-
ing a null hypothesis that there were no differences between the
pooled controls and the griseofulvin treated material with respect to
the occurrence of haploid sectors. Table 2 shows the results of the
analysis. All haploid lines were confirmedlcytologically through root
t0p squashes for 2n = x = 7.

The extreme significance (P << .001) of the griseofulvin treatment
renders superfluous a discussion of the CAP and PFP treatments.
Griseofulvin has been variously researched in the past decade (20,21,22,
23,24,25,26,27,28) and its mode of action seems to be through inhibi-
tion of microtubule protein synthesis. The site or sites at which it

operates are different from those of colchicine, although griseofulvin

109

Table 1. Frequency of Haploid Sectors Arising from VJJ' Clones Treated

with CAP, PFP, Griseofulvin and DMSO.

 

 

Number of Duration Number of Hap- Date of

Treatments Surviving_Clones of Stugy loid Sectors Sectoring
CAP 35 5.5 mo. 1 9/80
PFP 36 5.5 mo. 0 ----
Griseofulvin 38 4.5 mo. 4 7/80
7 8/80
3 9/80

Total for Griseofulvin: 14

Control I

(2% DMSO) 25 3.0 mo. 1 9/80

Control 11

(All other

untreated

plants) 242 6.0 mo. 1 8/80
5 9/80

Total for both controls: 7

110

Table 2. Contingency Analysis of Griseofulvin versus Controls for

Haploid Sector Induction in the Hybrid VJJ'.

H : The incidence of haploid sectors was independent of griseofulvin

treatment.

HA: The incidence of haploid sectors was associated with griseofulvin

 

 

treatment.
Control Griseofulvin Total
Haploids 7 14 21
No haploids 260 24 284
Total 267 38 305

Chi-square (with Yate's correction for continuity) = 55.54
df = 1

Therefore, reject H0 at the 0.001 level of P.

111

mimics colchicine to a certain extent through an early induction of
polyploidy (20,28). This polyploidy of cultured plant cells is accom-
panied by aneuploidy and, eventually, a range of chromosome numbers,
some of which are representative of haploidy. This latter phenomenon
accounted for the use of griseofulvin in this study.

What is not accounted for in any of the literature which has come
to the author's attention is a plausible, yet exact, explanation for
the extreme delay in the griseofulvin effect. It was fully two months
before it manifested itself in haploid induction and the phenomenon
continued until October, 1980, when the experiment was terminated, some
four and one-half months later. Furthermore, only haploid barley
sectored from the vegetative tissue of the triploid hybrid. No
aneuploidy was observed either cytologically or morphologically in the
sectors. Since there is an innate predisposition of the hybrid to
sector spontaneously (7,29,30)--although at a low frequency--it might
be argued that the aneuploidy and polyploidy reported in the scant
literature concerning the effects of griseofulvin on plant system (20,
28) does, in fact, occur in the hybrid VJJ'. The difference lies in
the possibility that the region containing the crown meristem of the
hybrid exerts a strong selection pressure upon karyotypically abnormal
cells, such that only three conditions have a reasonable probability of
surviving: haploid H. vulgare (2n = x = 7), haploid H. jubatum (2n =
2x = 14) and the normal hybrid tissue, VJJ' (2n = 3x = 21). This might
account for the delayed effects of the drug.

The confidence engendered by the increased frequency of haploidy
resulted in the treatment of many of those hybrids which were themselves

treated earlier with mitomycin-c and gamma irradiation to accomplish

112

genetic transfer from the H. jubatum genomes to that of H. vulgare.
These griseofulvin treatments were initiated in late July, 1980. The
treatments were identical to those of CAP, PFP and the original griseo-
fulvin treatments. In October haploid sectors were seen. For the
first time since this hybrid was made, sectors bearing a pure H.

jubatum morphology resulted. Four such plants aroseg. but contrary to

expectation, they were reconstituted segmental allotetraploids (2n
4x = 28). Therefore, in addition to a strict reduction in chromosome
number (e.g., H. vulgare haploid sectors), the elimination of the H.
vuloare genome was followed by a spontaneous doubling of the chromosome
number in the H. jubatum sectors.

A large and controlled study of six diploid barley cultivars
treated factorially with differing concentrations of griseofulvin with
2 percent DMF (dimethyl formamide) added had completely negative results
with respect to haploid induction. The only noticeable effects were
the partial induction of uniculm in the barley cultivar Bowers and a
generalized low frequency of sterile florets. DMF was determined prior
to the study to have no effect at the 2 percent concentration on any
of the plants with which the author was working. DMF is the only sol-
vent of griseofulvin which is not toxic (Nelson, unpublished). It was
used to increase the concentration of griseofulvin in plant tissue. The
solubility of griseofulvin in a 2 percent DMF solution was still only
0.2022 g/l (Nelson, unpublished). It might therefore be concluded that
the concentrations at which griseofulvin is effective are extremely
small, since DMF was not the solvent used in the original experiment.

The results of this experiment would imply that some degree of genomic

113

instability is, therefore, necessary for griseofulvin to induce chromo-
some elimination in somatic tissue.

What is disconcerting is that of all the haploids produced by the
author, either spontaneously or by induction, in the preceding four
years, none was phenotypically distinguishable from another. The trans-
fer of a polygenic trait was therefore unlikely. For those clones
which were untreated for somatic recombination this is not so surprising
since there is no homoeology between the respective genomes of H. yglggrg
and H. jubatum (31). This would render unlikely the probability of
somatic pairing and consequent genetic exchange. However, the irradi-
ated plants and the haploid plants which they produced should have
evinced some morphological differences. The irradiations were at 500 R
which was near the L050 for clones of this hybrid.

Isozyme analysis is a possible recourse to assay heterogeneity in
the haploid population. Limited material and severe time constraints
have hampered this study, although there is some evidence (32) for dif-
ferences between lines for malic dehydrogenase (MDH) and its associated
five bands. In the absence of gross morphological differences between
haploid lines the author is reluctant to ascribe isozyme heterogeneity
in the haploids to genes found only in H. jubatum and not in H. vulgare.
This is due to a potential source of bias incorporated in the VJJ'
hybrid itself. It has been mistakenly reported (30) that Coho barley
was the female parent of the hybrid. This is not so and the true parent-
age is reported by the creator of the hybrid, Dr. Robert Steidl (29).
The original FIVJJ' seed was produced on two spikes of a male sterile
(nuclear) winter barley. The male sterile character can be maintained

only in the heterozygous condition to produce seed over generations.

114

Consequently, there is considerable potential for heterogeneity through
outcrossing in the H. vulgare genome residing in the hybrid VJJ' and,
possibly, heterogeneity for MDH. Nevertheless, the author believes that
the haploid stocks should be assayed again and with greater specificity
for heterogeneity in isozymes with respect to each other, to the hybrid,
and to the pure H. jubatum recovered through griseofulvin treatment.
Additionally, the hybrid should undergo another cycle of irradiation as
the investment is low and the possible benefits are great. H. jubatum
is exceedingly winterhardy where H. vulgare, winter habit, is not.

H. jubatum also expresses extreme salt tolerance and is vigorous in
puddled soils.

The wide hybridization effort at Michigan State University has
encountered obstacles which were beyond comprehension when the program
was initiated. Years of hard work and a fair amount of ingenuity have
allowed us to develop novel means of transferring genetic variability
from wild to domestic barley. With another four years of concentrated
effort utilizing the recombinant and haploid systems detailed herein,
economically valuable improvements in barley will be recovered. The
VJJ' hybrid is a perennial of immense value and it should not be allowed

to disappear in consideration of its potential in barley improvement.

10.

11.

12.

13.

14.

115
CHAPTER 5--BIBLIOGRAPHY

Grafius, J.E. 1974. Breeding for resistance. In: Winter Hardiness
in Barley. Res. Rep. 247, Michigan State Univ. Agr. Exp. Sta.,
E. Lansing, 20 pp.

Eunus, A.M., L.P. Johnson and R. Akeel. 1962. Inheritance of winter
hardiness in an eighteen-parent diallel cross in barley. Can.
J. Genet. Cytol. 4:356-376.

Reid, D.A. 1965. Inheritance of growth habit in barley (Hordeum
vulgare L. Emed. Lam.). Crop Sci. 5:141-145.

Winter hardiness of progenies from winter x spring bar-

ley (Hordeum vulgare L. Emed. Lam.) crosses. Crop Sci. 5:

 

263-266.

Rhode, C.R. and D.F. Pulham. 1960. Heritability estimates of
winter hardiness in winter barley determined by the standard
unit method of regression analysis. Agron. J. 52:584-586.

Genetic studies of winter hardiness

 

in barley. Nebraska Agr. Exp. Res. Bul. 193.

Nelson, J.L. 1980. Interspecific and intergeneric hybrids with
Hordeum vulgare L. M.S. Thesis, Michigan State University.

 

Riley, R. and G. Kimber. 1966. The transfer of alien genetic
variation to wheat. Rep. Plant Breed. Inst., Cambridge 1964-
1965:6-36.

Sharma, 0. and D.R. Knott. 1966. The transfer of leaf rust resis-
tance from Agropyron to Triticum by irradiation. Can. J.
Genet. Cytol. 8:137-143.

Knott, D.R. 1968. Translocations involving Triticum chromosomes
carrying rust resistance. Can. J. Genet. Cytol. 10:695-696.

Sears, E.R. 1972. Agrogyron--what transfers through induced homo-
eologous paring. Can. J. Genet. Cytol. 14:736.

Harlan, J.R. 1968. On the origin of barley. USDA Agr. Handb.
338:9-31.

Hamilton, D.G., S. Symko and J.W. Morrison. 1955. An anomalous
cross between Hordeum leporinum and H. vulgare. Can J. Agr.
Sci. 35: 287-293.

 

Schooler, A.B. and M.K. Anderson. 1979. Interspecific hybrids
between (Hordeum brachyantherum L. x H. bogdanii Wilensky) x
H. vulgare L. J.THered. ’70:70-72.

 

15.

16.
17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

116

Schooler, A.B. 1967. A form of male sterility in barley. J. Hered.
58:206-211.

Windholz, M. ed. 1976. Merck Index, 9th ed. Merck and Company.

Yoshida, H. and H. Yamaguchi. 1973. Arrangement and association
of somatic chromosomes induced by chloramphenicol in barley.
Chromosoma (Berlin) 43:399-407.

Knight, R.L., A.P. Hamilton and E. Keep. 1963. Somatic reduction
of chromosome number in a Ribes hybrid following treatment with
parafluorophenylalanine. Nature (London) 200:1341-1342.

nitzsche, W. 1973. Mitotic chromosome reduction in higher plants
treated with 3-Fluro-phenylalanine. Die Naturwissenscft. 60:
1-2.

Rorchi, V.N., L. Mela, F.L. Schiava and M. Terzi. 1980. Effect of
grisefulvin on chromosome segregation and its possible use for
the isolation of mutants in plant cell cultures. In. Plant
Cell Cultures: Results and Perspectives, Sala, Parisi, Cella
and Ciferri, eds. Elsevier/North-Holland Biomedical Press,
pp. 139-144.

Malawista, S.E., H. Sato and W.A. Creasey. Dissociation of the
mitotic spindle in oocytes exposed to griseofulvin and vinblas-
tin. Exptl. Cell Res. 99:193-197.

Roobol, A., K. Gull and C.I. Pogson. 1976. Inhibition by griseo-
fulvin of microtubule assembly 19 vitro. FEBS Letters. 67:
248-251.

North, J. 1977. Effects of griseofulvin on diploid strains of
Coprinus lagopus. J. Gen. Microbiol. 98:529-534.

 

Roobol, A., K. Gull and C.I. Pogson. Griseofulvin-induced aggrega-
tion of microtubule protein. Biochem. J. 167:39-43.

Raudaskoski, M. and E. Huttunen. 1977. Effect of griseofulvin on
nuclear distribution in a dikaryon of Schizgphyllum commune.
Nykosen 20:339-348.

 

Valla, G. 1979. Effects of griseofulvin on cytology, growth,
mitosis and branching of Polyporus arcularius. Trans. Br.
Mycol. Soc. 73:135-139.

 

Raudaskoski, M. 1980. Griseofulvin-induced alterations in site of
dividing nuclei and structure of septa in a dikaryon of Schizo-
phyllum commune. 323-331.

 

Achiavo, F.L., V.N. Ronchi and M. Terzi. 1980. Genetic effects
of griseofulvin on plant cell cultures. Theor. Appl. Genet.
58: 43-47.

29.

30.

31.

32.

117

Steidl. R.P. 1976. Hybridization of barley (H. vulgare L. Emend.
Lam.) with its wild relatives. Ph.D. Dissertation, Michigan
State University.

Orton, T.J. 1980. Haploid barley regenerated from callus cultures
of Hordeum vulgare x H. jubatum. J. Hered. 71:280-282.

 

Starks, 0.0. and W. Tai. 1974. Genome analysis of Hordeum jubatum
and H. compressum. Can. J. Genet. Cytol. 16:663-668.

 

Dr. James Hancock, personal communication.

"IIIIIIIIIIIIIIIIIIIIIs