a L .. .
4.. (3:5
5:.

1..

 

 

 

 

 

491:):
2.. 1' . I...
:\:.l. 1t!!!
arty.)

‘4 ‘4

(ill! 9
t~ ‘5}...
i491} .2
1 “I

L...’ .534.
a:

v
a

5:!
3:}?!
x.
r

lite 3:) 1.2.5.55...
El. aaru..r4 .

.33

. x .
A35 ,5.

$11.23..

\x. :32
[£15 I
00"... I

.1
or .I. t I.
6a.: x

t

. \..1.r;»t, 1

 

 

Lawn; :2
if tthlyNo‘Z
~17 \ v! :2:

 

31::
.52.. Pl...v

1.: 1.1:.
.3». s.» I.
1

is;
3 t :9)!
0.32. 1......

)2
4.3.17. .

I}: '7.

 

 

My
5‘

7.:
1‘ :. .§

1:44;
i‘ 53‘

 

. . 5.1.x,
.r...x...‘vt . v.5

 

THEStS

                                                                                     

llllflllllllllllll

 

L 1293 01402 7266
O 3. 1:“)
LIBRARY
Michigan State
University

 

 

 

This is to certify that the
thesis entitled

APPLICATION OF MARKER ASSISTED SELECTION
TO IMPROVING A QUANTITATIVE TRAIT
IN COMMON BEAN (PHASEOLUS VULGARIS L.)

 

presented by

Kristin Ann Schneider

has been accepted towards fulﬁllment
of the requirements for

Crop & Soil Sciences —
Plant Breeding &
Genetics

Wma/

Major professor

Masters degree in

 

 

 

Date K'sg‘?~5/

 

0-7639 MS U is an Afﬁrmative Action/Equal Opportunity Institution

 

PLACE Ii RETURN BOXtoromovothbchodtomtmm yourrocord.
TO AVOID FINES return on or More data duo.

DATE DUE DATE DUE DATE DUE

 

[$8 9 431938

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU lsAn Nﬂmativo Action/Equal Opponmity Instituion
W

 

APPLICATION OF MARKER ASSISTED SELECTION TO IMPROVING A
QUANTITATIVE TRAIT IN COMMON BEAN (PHASE 0L US VULGARIS L.)

By

Kristin Ann Schneider

A THESIS

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

MASTERS OF SCIENCE

Plant Breeding and Genetics Program - Crop and Soil Sciences

1995

ABSTRACT

APPLICATION OF MARKER ASSISTED SELECTION TO INIPROVING A
QUANTITATIVE TRAIT IN COMMON BEAN (PHASEOLUS VULGARIS L.)

By

Kristin Schneider

Much eﬁ‘ort has been dedicated to improving drought tolerance in common bean.
Two bean populations were grown from 1990 to 1994 at eight locations in Michigan and
Mexico under stress and non-stress conditions. Yield under drought, yield potential,
drought susceptibility index, harvest index, and geometric mean were examined as
potential indicators of drought tolerant genotypes. Two genotypes from each population
were identiﬁed as having high yield potential under stress and non-stress conditions.
Heritability estimates ranged ﬁom 0.08 to 0.47 for yield, ﬁ'om 0.05 to 0.56 for biomass
and 0.58 to 0.77 for seed weight. Marker assisted selection was examined as a method to
indirectly select for drought tolerance in common bean. Random Ampliﬁed Polymorphic
DNA (RAPD) markers were analyzed against both populations. Using one-way AN OVAs
and multiple regression, RAPD markers were identiﬁed that were signiﬁcantly associated
with yield under stress, non-stress, and geometric mean. These markers were used to
select genotypes from either extreme. To test the eﬁ‘ectiveness of these selections, yield
data from three locations were examined. Selection based on marker genotype was found

to be more eﬁ‘ective than conventional selection gaining 10% more than expected.

ACKNOWLEDGMENTS

I would like to express my appreciation for Dr. James D. Kelly for serving as my
major adviser. His patience, wisdom and wealth of knowledge has contributed more to my
education than any academic curriculum. I would like to thank him for his time and
unwavering support throughout this research.

I would also like to thank Dr. Eunice Foster for her support and encouragement to
seek additional avenues of learning outside of Plant Breeding and Genetics. She was a
very valuable member of my guidance committee. My thanks are also extended to Dr.
Amy Iezzoni, the third member of my guidance committee, for her suggestions and patient
deliberation on this research.

For all of those on the bean crew, Jerry Taylor, Norman Blakely, and Lucia
Afanador, I could not have completed this research without you and I sincerely appreciate
your time and cooperation. A special thanks to Mary Brothers for her contribution to the
continuation of this research, her generosity and training.

I would like to thank Jorge Acosta and all those involved with the Bean/Cowpea
Collaborative Research Support Program. These were the people most closely involved
with the logistics and experimentation of this project. Their help and ﬁnancial support

have made a generous contribution to the completion of this project.

iii

Lastly, I would like to thank my friends and family for their patience and support
for the continuation of my education. They welcomed me home and made my stay here

unforgettable.

iv

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................... viii
LIST OF FIGURES ...................................................................................................... xiii
LIST OF ABBREVIATIONS ....................................................................................... xiv
INTRODUCTION ........................................................................................................... 1
Drought ....................................................................................................................... 1
Marker Assisted Selection ............................................................................................ 7
Quantitative Marker Assisted Selection ........................................................................ 8
Current research and approaches ................................................................................ 10
Marker Assisted Selection for Quantitative Traits ....................................................... 11

Is a genomic map necessary? .................................................................................. 13
Identifying and Evaluating QTLs over Multiple Locations ...................................... 16
Statistical Methods for Identifying QTLs ................................................................ 17
Population Development ........................................................................................ 19
Population Size ...................................................................................................... 20
Markers ..................................................................................................................... 21
MATERIALS AND METHODS ................................................................................... 24
Population Development ............................................................................................ 24

Parents ................................................................................................................... 24

Populations ............................................................................................................ 24
Field Study ................................................................................................................ 25
Marker Protocol ........................................................................................................ 29

DNA Extraction ..................................................................................................... 29

Polymerase Chain Reaction (PCR) Protocol ........................................................... 30

Electrophoresis ...................................................................................................... 3O

Marker Scoring and Nomenclature ......................................................................... 31
Statistical Analysis ..................................................................................................... 31

Field Studies .......................................................................................................... 31

Markers ................................................................................................................. 33

RESULTS ..................................................................................................................... 36
Field Study ................................................................................................................ 36

Statistical Analysis ................................................................................................. 40

Heritability Estimates and Gain ﬁ'om Selection ....................................................... 41

Drought Intensities ................................................................................................. 48

Yield Correlations .................................................................................................. 48
Marker Analysis ......................................................................................................... 54
Marker Assisted Selection .......................................................................................... 62
Marker Assisted Selection Compared to Conventional Selection ................................ 75

DISCUSSION ............................................................................................................... 78
Field Study ................................................................................................................ 79

vi

Marker Analysis ......................................................................................................... 89

Marker Assisted Selection .......................................................................................... 92
Marker Assisted Selection Compared to Conventional Selection ................................ 96
APPENDIX ................................................................................................................. 103
LIST OF REFERENCES ............................................................................................. 132

LIST OF TABLES

Table 1. Analysis of variance combined over ﬁve locations ﬁ'om 1990-1993 in
Michigan and Mexico for 80 bean genotypes. Stress and non-stress
treatments were analyzed individually for the Sierra/A0028 population. ........... 37

Table 2. Analysis of variance combined over ﬁve locations ﬁ'om 1990-1993 in
Michigan and Mexico for 97 bean genotypes. Stress and non-stress
treatments were analyzed individually for the Sierra/Lef—ZRB population. ........... 39

Table 3. Heritability estimates for yield, biomass and 100 seed weight for two bean
populations grown under stress and non-stress treatments over six
locations and ﬁve years (1990-1994) in Michigan and Mexico. ........................... 46

Table 4. Gain in yield per cycle of selection at varying selection intensities (k)
based on heritability estimates (Table 3) for Sierra/AC1028 population
grown under stress and non-stress environments over ﬁve years (1990-
1994) and six locations in Michigan and Mexico. ................................................ 47

Table 5. Gain in yield per cycle of selection at varying selection intensities (k)
based on heritability estimates (Table 3) for Sierra/Lef-ZRB population
grown under stress and non-stress environments over ﬁve years (1990-
1994) and six locations in Michigan and Mexico. ................................................ 47

Table 6. Drought intensities index (DII) for two bean populations grown at six
locations over four years. ................................................................................... 49

Table 7. Correlations between yield, biomass, and 100 seed weight (100 SW) in
stress (S) and non-stress (NS) environments for two bean pepulations
grown over four years (1990-1993) and ﬁve locations in Michigan and
Mexico ............................................................................................................... 51

Table 8. Correlations between yield, biomass, 100 seed weight (100 SW), and
number of pods per plant (PPP) measured under stress (S) and non-stress
(NS) treatment and drought susceptibility index (D81) and geometric mean
(GM) for the Sierra/AC1028 experiment conducted at Kalamazoo County,
MI in 1994. ........................................................................................................ 52

Table 9. Correlations between yield under stress (S), yield under non-stress (NS),
tolerance and mean productivity for two bean populations grown over four
years (1990-1993) in ﬁve environments in Michigan and Mexico. ....................... 53

viii

Table 10. Signiﬁcant F-tests (F) from multiple regression analyses performed
between linkage groups (as deﬁned in Fig. 5) yield under drought (Yd),
yield under non-stress (Yp), and the geometric mean (GM) for the
Sierra/AC 1028 population grown under two moisture treatments. Analyses
were performed using data ﬁ'om each of ﬁve individual environments
(1990-1993) and for one combined over all environments. .................................. 58

Table 11. Signiﬁcant F-tests (F) from multiple regression analyses performed
between linkage groups (as deﬁned in Fig. 6) yield under drought (Yd),
yield under non-stress (Yp), and the geometric mean (GM) for the
Sierra/Lef—ZRB population grown under two moisture treatments. Analyses
were performed using data ﬁom each of individual ﬁve environments
(1990-1993) and for one combined over all environments. .................................. 60

Table 12. A comparison between the number of signiﬁcant F-tests from multiple
regression analyses between all the RAPD markers in linkage group 9 for
the Sierra/AC1028 population and yield under drought (Yd), yield under
non-stress (Y p) and geometric mean (GM) versus the one-way analyses of
variance performed on each individual unlinked marker found in linkage
group 9. Data is shown for each of the ﬁve environments (1990-1993),
separately, and the combined analysis over locations ........................................... 64

Table 13. Grand mean and means for each genotypic group selected based on
marker genotype, LSD and CV for yield, biomass, 100 seed weight, and
number of pods per plant resulting from the analysis of variance for stress
and non-stress treatments, individually, in Kalamazoo County . Eleven
genotypes ﬁ'om the Sierra/AC1028 population were selected as potentially
above average (PAA) and ten as potentially below average (PBA)
performance based on marker genotypes. ........................................................... 65

Table 14. Grand mean and means for each genotypic group selected based on
marker genotype, LSD and CV for yield and biomass resulting from the
analysis of variance for stress and non-stress treatments in 1994 conducted
at Madero, Durango, Calera, Zacatecas and combined over both locations.
Eleven genotypes from the Sierra/AC1028 population were selected as
potentially above average (PAA) and ten as potentially below average
(PBA) performance based on marker genotypes. ................................................ 67

Table 15. A comparison between the number of signiﬁcant F-tests from multiple
regression analyses between all RAPD markers in linkage group 1 for 8/1.
and yield under drought (Yd), yield under non-stress (Yp) and geometric
mean (GM) versus the one-way analyses of variance performed on each
individually unlinked marker. Data is shown for each of the ﬁve
environments (1990-1993), separately, and the combined analysis over
locations. ........................................................................................................... 70

ix

Table 16. Grand mean and means for each genotypic group selected based on

Table

marker genotype, LSD and CV for yield biomass, and 100 seed weight
resulting from the combined analysis of variance over locations for stress
and non-stress treatments, individually, conducted in 1994 at Madero,
Durango, Calera, Zacatecas and combined over both locations. Thirty-six
genotypes from the Sierra/Lef-ZRB population were selected as potentially
above average (PAA) and 28 as potentially below average (PBA)
performance based on marker genotypes. ........................................................... 72

17. Comparisons between marker assisted selection and conventional
selection. Using data over 4 years (1990-1993) and ﬁve locations in
Michigan and Mexico, 3 genotypic groups of 35 RILs were selected from
the Sicrra/Lef-ZRB population based on above average performance (AA)
for yield under drought (Yd), yield potential (Y p), and geometric mean
(mean). Similar selection was used to identify 3 groups selected for below
average performance (BA). The mean yields ﬁ'om these groups calculated
from data collected ﬁ'om 1994 Madero Dgo. and Calera Zac. experiments
were compared to groups identiﬁed by marker genotype. ................................... 76

Table A. 1. Analyses of variance for yield, biomass, and 100 seed weight of 81

bean genotypes from Sierra/AC1028 population and checks grown over
seven locations (1990-1994) under stress conditions in Michigan and
Mexico ............................................................................................................. 103

Table A 2. Analyses of variance for yield, biomass, and 100 seed weight of 81

bean genotypes from Sierra/AC1028 population and checks grown over
seven locations (1990-1994) under non-stress conditions in Michigan and
Mexico ............................................................................................................. 106

Table A. 3. Analyses of variance for yield, biomass, and 100 seed weight of 97

bean genotypes from Sierra/Lef-ZRB population and checks grown over
seven locations (1990-1994) under stress conditions in Michigan and
Mexico ............................................................................................................. 109

Table A. 4. Analyses of variance for yield, biomass, and 100 seed weight of 97

bean genotypes ﬁ'om Sierra/Lef-ZRB population and checks grown over
seven locations (1990-1994) under non-stress conditions in Michigan and
Mexico ............................................................................................................. 112

Table A 5. Split-plot analysis of variance for yield, biomass and 100 seed weight

of 80 bean genotypes ﬁ'om the Sierra/AC1028 population combined over
ﬁve locations (1990-1993) in Michigan and Mexico .......................................... 115

Table A 6. Split-plot analysis of variance for yield, biomass and 100 seed weight

of 98 bean genotypes from the Sierra/Lef-ZRB population combined over
ﬁve locations (1990-1993) in Michigan and Mexico .......................................... 116

Table A. 7. Yield, biomass and 100 seed weight (100 SW) for both stress and non-
stress treatments of the top and bottom ﬁve genotypes ranked by yield
under stress in the Sierra/AC1028 population grown over ﬁve locations
(1990-1993) in Michigan and Mexico. .............................................................. 117

Table A 8. Yield, biomass and 100 seed weight (100 SW) for both stress and non-
stress treatments of the top and bottom ﬁve genotypes ranked by yield
under non-stress in the Sierra/AC1028 population grown over ﬁve
locations (1990-1993) in Michigan and Mexico. Values for both parents
are included ...................................................................................................... 117

Table A. 9. Yield, biomass and 100 seed weight (100 SW) for both stress and non-
stress treatments of the top and bottom ﬁve genotypes ranked by yield
under stress in the Sierra/Lef-ZRB population grown over ﬁve locations
(1990-1993) in Michigan and Mexico. .............................................................. 118

Table A 10. Yield, biomass and 100 seed weight (100 SW) values for both stress
and non-stress treatments of the t0p and bottom ﬁve genotypes ranked by
yield under non-stress in the Sierra/Lef-ZRB population grown over ﬁve
locations (1990-1993) in Michigan and Mexico. ............................................... 118

Table A 11. Signiﬁcant F-tests from one-way analyses of variance between yield
under stress and 70 RAPD markers. One-way analyses of variance were
performed between RAPD marker genotypes and yield under stress from
ﬁve locations (1990-1993) in Michigan and Mexico and the combined
analysis over all environments in the Sierra/A0028 population ........................ 119

Table A 12. Signiﬁth F-tests ﬁ'om one-way analyses of variance between yield
under non-stress and 70 RAPD markers. One-way analyses of variance
were performed between RAPD marker genotypes yield under non-stress
from ﬁve locations (1990-1993) in Michigan and Mexico and the combined
analysis over all environments in the Sierra/AC1028 population ........................ 120

Table A. 13. Signiﬁcant F-tests from one-way analyses of variance between
geometric and 70 RAPD markers. One-way analyses of variance were
performed between RAPD marker genotypes and the geometric mean of
yield under stress and non-stress from ﬁve locations (1990-1993) in
Michigan and Mexico and the combined analysis over all environments in
the Sierra/AC1028 population. ......................................................................... 122

Table A 14. Signiﬁcant F—tests from one-way analyses of variance between yield
under stress and 70 RAPD markers. One-way analyses of variance were
performed between RAPD marker genotypes and yield under stress from
ﬁve locations (1990-1993) in Michigan and Mexico and the combined
analysis over all environments in the Sierra/AC1028 population ........................ 123

Table A 15. Signiﬁcant F-tests from one-way analyses of variance between yield
under non-stress and 70 RAPD markers. One-way analyses of variance

xi

were performed between RAPD marker genotypes and yield under non-
stress from ﬁve locations (1990-1993) in Michigan and Mexico and the
combined analysis over all environments in the Sierra/Lef—ZRB population. ...... 125

Table A. 16. Signiﬁcant F-tests ﬁom one-way analyses of variance between

geometric and 70 RAPD markers. One-way analyses of variance were
performed between RAPD marker genotypes and the geometric mean of
yield under stress and non-stress from ﬁve locations (1990-1993) in
Michigan and Mexico and the combined analysis over all environments in
the Sierra/AC1028 population. ......................................................................... 127

Table A. 17. Split-plot analyses of variance for yield, biomass, 100 seed weight,

number of pods/plant, and days to maturity using data ﬁ'om the 1994
Kalamazoo County experiment. Twenty-one genotypes ﬁ'om the
Sierra/AC1028 population selected based on their marker genotype were
grown under stress and non-stress conditions. Included are orthogonal
contrasts between the marker-based, potentially above-average group
(PAA) and the potentially below average group (PBA) ..................................... 129

Table A. 18. Grand mean and means for each genotypic group selected based on

marker genotype, LSD, CV and analysis of variance for yield resulting
from analysis of the split-plot design combined over locations conducted at
Madero, Durango and Calera, Zacatecas in 1994. Eleven genotypes from
the Sierra/AC1028 population were selected as potentially above average
(PAA) and ten as potentially below average (PBA) based on their marker

genotypes ......................................................................................................... 130

Table A. 19. Grand mean and means for each genotypic group selected based on

marker genotype, LSD, CV and analysis of variance for yield resulting
from analysis of the split-plot design combined over locations conducted at
Madero, Durango. and Calera Zacatecas in 1994. Thirty-ﬁve genotypes
ﬁ'om Sierra/Lef-ZRB were selected as potentially above average (PAA) and
28 as potentially below average (PBA) based on their marker genotypes. ......... 131

xii

LIST OF FIGURES

Figure 1. Frequency distributions for yield under stress (Yd) and non-stress (Y p)
using means for each of 80 genotypes combined over four years (1990-
1993) and ﬁve locations in Michigan and Mexico for the Sierra/AC1028
populations. Sierra is marked with an “O” and AC1028 by an “X” ...................... 42

Figure 2. Frequency distributions for biomass under stress and non-stress using
means for each of 80 genotypes combined over four years (1990-1993) and
ﬁve locations in Michigan and Mexico for the Sierra/AC1028 populations.
Sierra is marked with an “O” and AC1028 by an “X”. ........................................ 43

Figure 3. Frequency distributions for yield under stress (Yd) and non-stress (Y p)
using means for each of 98 genotypes combined over four years (1990-
1993) and ﬁve locations in Michigan and Mexico for the Sierra/Lef-ZRB
populations. Sierra is marked with an “O” and Lef-ZRB by an “X”. .................... 44

Figure 4. Frequency distributions for biomass under stress and non-stress using
means for each of 98 genotypes combined over four years (1990-1993) and
ﬁve locations in Michigan and Mexico for the Sierra/Lef-ZRB populations.
Sierra is marked with an “O” and Lef-2RB by an “X”. ........................................ 45

Figure 5. Linkage groups generated by MAPMAKER using seventy RAPD
markers for the Sierra/AC1028 population. No marker is more than 10 cM
away from its adjacent marker based on a LCD 2 4.0 ......................................... 56
Figure 6. Linkage groups generated by MAPMAKER using seventy RAPD
markers for the Sierra/Lef-ZRB population. No marker is more than 10 cM
away ﬁom the adjacent marker based on a LCD 2 4.0. ....................................... 57

xiii

A

LIST OF ABBREVIATIONS

Carbon isotope discrimination

100 sw 100 Seed weight

1df
AA
BA
cM
CV
Dgo.
DII
DNA
DSI
GM
h2
HI

1:
LCD
LSD
MAS
MI
NS
PAA
PBA
PCR
PPP
QTL

RAPD

S/A
S/L

Xd

1 degree of freedom

Above average genotypic group identiﬁed by conventional selection
Below average genotypic group identiﬁed by conventional selection
Centimorgans

Coeﬁcient of variation

Durango

Drought intensity index (l-Xd/Xp)

Deoxyribonucleic acid

Drought susceptibility index ((1-Yd/Yp)/DII)

Geometric mean of yield under stress and non-stress

Heritability estimate

Harvest index

Percent selection intensity constant

Likelihood of the odds ratio

Least signiﬁcant difference

Marker assisted selection

Michigan

Non-stress treatment

Potentially above average genotypic group identiﬁed by marker assisted selection
Potentially below average genotypic group identiﬁed by marker assisted selection
Polymerase chain reaction

Number of pods per plant

Quantitative trait loci

Coefficient of determination

Random ampliﬁed polymorphic DNA

Restriction ﬁagment length polymorphism

Recombinant inbred line

Stress treatment

Sierra/AC1028 population

Sierra/Lef-ZRB population

Water use eﬁiciency

Mean yield of the stress treatment

XIV

Mean yield of the non-stress treatment
Yield under stress

Yield under non-stress

Zacatecas

INTRODUCTION

Drought

Sixty percent of common bean production worldwide is grown under water stress,
making drought the second largest contributor to yield reduction after disease (Singh,
1995). One of the largest production areas in the world is the Mexican Highlands where
over one million (hectares of common bean are planted annually. Ninety-eight percent of
this region is subjected to intermittent rainfall (Acosta and White, 1995) and natural
conditions in much of this area do not provide suﬁicient moisture (annual precipitation
200-400 mm) (Singh, 1995). Common bean is well adapted to this region because of its
short growing season (100 days ﬁom planting to maturity) which signiﬁcantly reduces its
water requirement below that of other species generally considered more drought-adapted
(White, 1993). Although management practices can contribute to increased yields in
moisture-stressed environments, it is commonly acknowledged that major progress must
come from genetic improvement (White et al., 1994b).

Mechanisms exploited by crop species to survive periods of moisture stress include
avoidance and tolerance. Avoidance can involve escaping drought periods through
developmental timing or by preventing dehydration of tissue. Tolerance refers to the

ability of certain genotypes to tolerate low water potential (Acosta, 1988). By combining

2
aspects ﬁ'om both of these mechanisms a breeder could expect to improve performance
under drought.

A major consideration in breeding for drought tolerance is the type of drought
conditions prevalent in the environment of interest. Two types of moisture stress
conditions are generally recognized: intermittent and terminal (Ludlow and Muchow,
1990). Intermittent stress can occur at any time during the growing season and can vary in
intensity and duration. Intermittent stress is generally observed in the semi-arid tropics
where plants are grown during the rainy season for which rainfall is often sporadic. In
terminal drought situations, crops are planted during the dry season and rely solely on
stored water reserves in the soil. Thus as the season progresses, the crop gradually
depletes its water resources resulting in moisture stress at the end of the growing season
(Ludlow and Muchow, 1990). The highland region of Mexico (1800-2000 masl) is
characteristic of intermittent drought conditions although intermittent stress in this region
is not aggravated by associated high temperatures inherent to the tropics (Acosta, 1988).

In common bean, differences in morphology, phenology, partitioning and yield
response contribute to diﬁ‘ering abilities of genotypes to survive periods of moisture stress.
An important morphological response to drought in common bean is loss of leaf area. This
can be the result of reduced number of leaves, reduced size of younger leaves, inhibited
expansion of developing foliage, or leaf loss brought on by senescence (Acosta, 1988).
Root morphology can also have a major affect on performance under stress. Differences in
growth habit are mirrored by differences in root morphology. Type II growth habit is

characterized by an indeterminate, upright plant structure with reduced branching angle

3

whereas type III habit is typical of an indeterminate prostrate sprawling plant structure
(Brothers and Kelly, 1993). Type 11 plants develop a thick tap root which can exploit
deeper soil levels where water is often stored and type III plants, which are better suited
for terminal stress conditions, exhibit a shallow expansive root morphology (Lynch and
van Beern, 1993). Gennplasm adapted to the Mexican highlands exhibits the type III
grth habit. The reason for this is twofold. An indeterminate prostrate growth pattern
can establish an expansive canopy early in development to prevent water loss through
evaporation from the soil surface. In addition to the prostrate growth habit of type III
plants, a sprawling root system is advantageous since soils in the Mexican highlands are
shallow with little water holding capacity. A sprawling root system that is in contact with
surface water is more adapted to this region where intermittent drought is prevalent
(Acosta, 1988). Roots have been implicated as the primary sensor for drought stress
although this is a much debated subject (Passioura, 1988; Kramer, 1988). White and
Castillo (1989; 1992), using grafting studies have demonstrated that shoot growth is of
lesser importance when compared to root genotype. Clearly, root morphology plays
critical roles in determining performance under drought.

An equally important trait attributed to type III growth habit, especially in terms of
drought avoidance, is phenotypic plasticity. In certain Mexican gerrnplasm the duration
between emergence and maturity varies ﬁom season to season. In the Mexican highlands,
planting depends on the onset of summer rains. This can vary anywhere between June to
early August. ‘ In seasons where early rains permitted early planting, genotypes

demonstrating phenotypic plasticity showed a longer grth cycle whereas these same

4

genotypes dramatically shortened there growing cycle at later planting dates (Acosta and
White, 1995). In this way, plants can avoid drought conditions or low temperatures later
in the season by reducing the time nwded to mature. This trait allows farmers in the
Mexican Highlands the opportunity to avoid drought conditions by timing their planting
dates around the onset of summer rains.

Leaf angle or movement isanother trait which may contribute to drought tolerance
especially where high temperatures are a problem. If a leaf is capable of orienting the
majority of its surface area away from or parallel to incident sunlight, damage due to
stomatal closure can be avoided. This type of movement has been demonstrated in
common bean under drought conditions. Genetic variation has also been observed, with
genotypes of Andean origin exhibiting the strongest effect (Kao et al., 1994).

Another contributing factor to drought tolerance is the ability of a genotype to
partition assimilates to the reproductive organs. This is especially true for indeterminate
plants where the reproductive and vegetative growth stages overlap (Acosta, 1988). An
important aspect of this is the ability to remobilize stored reserves ﬁom other tissue into
the developing seed. A common measure of this capability is harvest index (HI) which is
the ratio between seed yield and biological yield. However it has been reported that
common bean already possesses inherently high HI values suggesting that it would be
impossible to exploit this trait any further (Acosta, 1988). However, utilizing genotypes
which are more capable of remobilizing carbohydrates, stored during non-stress periods,

to the reproductive sink could improve performance under drought.

5

For subsistence agriculture and production areas, like the Mexican Highlands,
where drought conditions vary ﬁom season to season a cultivar which performs
consistently well under stress and non-stress environments is preferred (Rosielle and
Hamblin, 1981). Singh (1995) and Ramirez (1992) have demonstrated a positive
correlation between stress and non-stress yields. This would make it possible to improve
yield under moisture stress by selecting for higher yields in non-stressed conditions.
Additionally, heritability estimates (hz) are slightly higher under non-stress than stress
conditions. Selection under drought conditions is confounded by increased environmental
error due to variation in water deﬁcit, soil fertility, and increased susceptibility of
genotypes to soil borne pathogens and pests (Singh, 1995). Although selection for
increased yield under non-stress could be useful to increase performance under drought
stress, h2 for yield is still considered moderately low in any environment. Selection for
yield is also confounded by additive gene action and a strong environmental inﬂuence.
Consequently, selection cannot begin until later generations when sufﬁcient additive
variance and adequate seed is available. It would be advantageous to develop a method to
indirectly select superior genotypes in earlier generations. Many factors which contribute
to drought tolerance have been explored for this purpose. These traits include phenology,
photoperiodism, developmental plasticity, mobilization of assimilate, root hydraulic
conductance, early vigor, leaf area maintenance, low lethal water status, leaf movements,
leaf reﬂectance, epidermal conductance, transpiration eﬁiciency, high temperature
tolerance, osmotic adjustment and carbon isotope discrimination (Ludlow and Muchow,

1990). However, most of these characteristics are laborious, inefﬁcient and expensive to

6
measure. More often, these correlated factors are ineffective to use for indirect selection
because they are not applicable to large breeding populations. To further compound the
problem, the majority of these traits must be tested under drought conditions which are
inconsistent and difﬁcult to reproduce.

The trait most recently regarded as having the greatest potential for indirect
selection of drought tolerance is carbon isotope discrimination (A). A is the ratio between
ambient l3C concentration and the 13C concentration in the plant. This equation,

A = RJR,-1
(R. and R, are the isotope concentrations in the air and plant, respectively), in turn, is
directly related to the ratio between the internal and ambient C02 concentrations in C3
species. This relates A to water use efﬁciency (WUE) (White et al., 1994a). Water use
eﬁciency is deﬁned as the moles of C02 ﬁxed per unit of water lost ﬁ'om a leaf. Thus, the
more carbon ﬁxed per water lost, the more eﬁicient a genotype is at conserving its
resources making that genotype more drought tolerant. A has been negatively correlated
with WUE in wheat, barley, sunﬂower and peanuts in potted plants (Ludlow and
Muchow, 1990). A is not strongly correlated with WUE, however, in ﬁeld studies of
common bean. White (1993) measured A at two locations in Colombia with similar
climatic conditions but different soil types. At both locations bean gerrnplasm was grown
under normal rainfed or stressed conditions. One location, however, had an additional,
irrigated experiment. Correlations between A and seed yield were inconsistent over trials

and h2 for A was lower than expected. When selecting for reduced A, both rainfed

7
locations demonstrated a low gain ﬁ'om selection whereas under irrigation a 7% gain from
selection in seed yield was observed. Although correlations between A and seed yield were
inconsistent, A has been associated with root length density (White et al., 1990; White,
1993; White et al, 1994) in common bean. It was concluded that A is more an indicator of
root morphology than WUE and that A represents a reduction in photosynthesis.
Therefore, increases in WUE are due to a reduction in the amount of water transpired and
not an increase in ﬁxed moles of C02. Root growth and leaf movements are believed to be
major contributors aﬁ‘ecting the amount of water transpired (White, 1993). Thus, in
common bean, A measurements are more susceptible to environmental factors making
carbon isotope discrimination a weak indicator of stress tolerance (White et al., 1994). As

a result, the need for an eﬁicient indirect-selection method still remains.

Marker Assisted Selection
Genetic maps based on linkage analysis were ﬁrst suggested by Punnet while
researching sweat pea in 1908 (Ellis et al., 1992). Linkage relationships between genetic
markers and many agronomic and physiological traits have been studied and included on
genomic maps. This technology has advanced over the years to include complete genomic
maps of many important crop species. With the development of more abundant molecular
or DNA-based markers, marker technology including the mapping of quantitative or

complex traits has become possible. By tagging a particular trait of interest with a tightly

8
linked marker, the marker genotype can be used to indirectly select for that trait. This
method is commonly referred to as Marker Assisted Selection (MAS).

Marker assisted selection has particular potential for improving major-gene disease
and pest resistance in crop species (Melchinger, 1990). Diseases, for which screening is
laborious and expensive or for which the pathogen is quarantined, lend themselves
particularly well to an indirect method of selection. The maintenance of large populations
through the reproductive stage as well as progeny testing can be eliminated because DNA
screening can be done at the seedling stage and heterozygotes can be distinguished from
the homozygous-dominant individuals with certain co—dorninant markers (Tanksley, 1983).
Furthermore, gene pyramiding or the combining of more than one gene in a single
genotype is facilitated with the use of markers. This is particularly true in the case of genes
which are normally masked and unidentiﬁable at the phenotypic level due to epistatic
eﬁ‘ects. Markers have been identiﬁed in common bean which are associated with single-
gene traits such as the rust resistance genes Up-2 (Miklas, et al., 1993), Ur-3 (Haley et al.,
1994b), and 3-190 (Haley et al., 1993) in addition to bean common mosaic virus
resistance genes, I (Haley, 1994a) and bc-3 (Haley et al., 1994d), and an anthracnose
resistance gene, Are (Young et al., 1994). These markers are currently being utilized in

selection for resistant genotypes in addition to facilitating gene pyramiding (Kelly, 1995).

Quantitative Marker Assisted Selection
Because the facility of marker assisted selection for qualitative traits has proven

successﬁll, this technology can be further applied to quantitative traits. A quantitative trait

9
is deﬁned as one that exhibits continuous variation. This is due to one of two reasons or a
combination of both: 1) the trait is controlled by many genes and 2) the trait is strongly
inﬂuenced by the environment (Paterson et al., 1990). When breeding for a quantitative
trait such as drought tolerance in common bean, selection cannot be performed until later
generations. For a conventional breeding program, improvement would be initiated with
with a cross between two inbred lines that demonstrate superior performance under
drought. Individual F2 plants from this cross are advanced to the F5 generation by single
seed descent. F5 seed ﬁ'om individual F5 plants is then bulked to create Fg-derived
recombinant inbred lines (RILs). At this time, replicated ﬁeld trials are planted with both
stress and non-stress treatments. Selection will begin based on the results from this season.
Superior genotypes are not selected prior to the F6 generation for several reasons.
Common bean is a self-pollinating crop where the ﬁnal product will be an inbred line, so
increasing additive variance is a primary goal. In addition, if heterozygosity is too high, the
breeder may be inadvertently selecting for hybrid vigor making it necessary to inbreed until
suﬁcient additive variation and homozygosity are available. Since drought tolerance is a
yield-related trait and is strongly inﬂuenced by the environment, replicated ﬁeld studies in
a number of locations must be conducted. To accurately select consistently superior
genotypes, extensive testing is required and an adequate supply of seed ﬁom each
genotype is needed. It is not until the F4 or F5 generation that sufﬁcient homozygosity is
present to ensure uniform progeny after which a generation or two is required to increase
available seed. If simply inherited markers were used to indirectly select superior

genotypes, selection could begin in earlier generations such as F; or F4 or in environments

10
where drought conditions are inconsistent or difﬁcult to simulate. This does not, however,
eliminate the need to conduct replicated ﬁeld studies in both stress and non-stress
conditions, but ensures that the genotypes which are advanced to the yield trials have the
potential to be among the highest perfomring ﬁom that population.

Soller and Beckrnann (1990) stated that MAS for quantitative traits with high
heritability would not be as eﬁicient as conventional breeding. The predictive value of
MAS is inversely proportional to the heritability of that trait (Lander and Botstein, 1989).
The heritability for yield performance under stress in common bean is low (Ramirez,
1992), indicating that if markers were identiﬁed that explained a signiﬁcant amount of the
variation in a population, MAS would facilitate selection. For a quantitative trait with high
heritability, MAS could still be effective after the major quantitative trait loci (QTL) are
ﬁxed and the heritability is reduced (Paterson et al., 1991). Although markers associated
with QTLs controlling quantitative traits, have been reported (Dudley, 1993; Edwards et
al, 1987 and 1992; Lander and Botstein, 1989; Paterson et al., 1990 and 1991; Martin et
al., 1989; Freyre and Douches, 1994a and 1994b; Stuber, 1986; Stuber et al, 1987),
limited research has been conducted regarding the effectiveness of MAS for improving

quantitative traits (Stromberg et al., 1994).

Current research and approaches
The recent mapping trend has elucidated the chromosomal position of many
quantitative trait loci in several important agronomic crops. Edwards, et al. (1987)

identiﬁed signiﬁcant associations between isozyme loci and 82 quantitative traits in two F2

11
populations of corn. Stuber et al. (1987), using the same hybrid crosses, further identiﬁed
isozyme loci associated with QTLs inﬂuencing 25 yield-related traits. Edwards et al.
(1992) identiﬁed RFLP markers ﬁ'om a highly-saturated genomic map which were
associated with the same quantitative traits evaluated in the Stuber study (1987). They
concluded that associated RFLP loci corresponded to those locations already identiﬁed by
the isozyme markers. RFLPs, however, are more informative than isozymes because they
are more abundant. Paterson et al. (1988), using an RFLP map of tomato, identiﬁed six
QTLs inﬂuencing ﬁ'uit mass, four for soluble solids concentration and ﬁve for fruit pH in
an interspeciﬁc back-cross. They were the ﬁrst to adapt interval mapping, a method
previously used in human genetic studies which involves maximum likelihood ratios and
LCD (Likelihood of the odds ratio) scores, to crop species. QTLs controlling gray leaf
spot resistance in maize have also been mapped (Bubeck et al., 1993) and Beavis et al.
(1994) have identiﬁed QTLs inﬂuencing 24 other agronomic traits in corn. Freyre and
Douches (1994a; 1994b) have mapped QTLs controlling speciﬁc gravity and tuber
dormancy in potato using RFLPs, isozymes and RAPDs, and Martin et al. (1989) have

identiﬁed three RFLP loci associated with A in tomato.

Marker Assisted Selection for Quantitative Traits
Aﬁer identifying map positions and associated markers, MAS can be practically
applied to breeding for QTLs. Discreet regions of the genome can be identiﬁed that
contribute to a signiﬁcant portion of the variation of a quantitative trait. By selecting for

these regions, the trait can be improved. Stuber and Edwards (1986) used markers

12

associated with grain yield to determine if improvement in yield could be realized. Based
on marker genotype, potentially below average performing and potentially above average
performing F2 individuals were selected for three traits: grain yield, ear height, and car
number. Concurrently, conventional mass selection was conducted on an open-pollinated
F2 population for comparison. Stuber and Edwards (1986) concluded that MAS was as
eﬁ‘ective as phenotypic selection for improving quantitative traits despite the fact that the
markers used to identify QTLs represented only 40% of the genome. This would suggest
that with a more saturated map, MAS could be more effective than conventional
phenotypic selection (Stuber and Edwards, 1986). Although similar responses were
observed between phenotypic and genotypic selection, a three-fold increase in marker
frequency occurred in the populations selected on the basis of marker genotype. This
suggests that there are unmapped regions of the genome for which marker-based selection
would be ineﬁ‘ective (Stuber and Edwards, 1986). If, however, a breeder were to combine
genotypic selection with phenotypic selection, breeding efforts to improve quantitative
traits could be accelerated (Bubeck, 1993).

One important question is whether markers are applicable to populations other
than the ones from which the markers were identiﬁed. In maize, Stuber et al. ( 1992)
demonstrated that map positions of major factors inﬂuencing grain yield in the cross B73 x
M017 corresponded to those identiﬁed in the earlier studies. Using the same cross, Beavis
et al. (1994) identiﬁed QTLs associated with 24 agronomic traits using a small but
representative sample of top-crossed and F 4 progeny. Only a few of the yield-associated

QTLs from this study corresponded, in position, to those identiﬁed in the Stuber study

13
(1992) despite using the same parental hybrids. This suggests that the markers are not only
population-speciﬁc but ineffective for different populations that have been developed ﬁ'om
similar crosses. Confounding factors could be attributed to these results which include,
different genetic backgrounds of the progeny, diﬁ‘ering sources of parental lines, different
envirorunents, and the small number of progeny analyzed by Beavis et al. (1994). Beavis et
al. (1994) also suggested that if a trait is inﬂuenced by a large number of genes then
probably only a few of these loci will be identiﬁed by marker analysis. Consequently, the
probability that two experiments will identify the same fraction of loci is unlikely. This is
most likely the case for diﬁ‘erent yield-related traits. If associated markers prove to be
population-speciﬁc, their usefulness is not precluded. Many of the same elite lines or good
combining parents are used repeatedly to develop new cultivars. By identifying genomic
regions of interest in this recurrent parent, markers can be used to make selections in

populations for the same genomic regions that were valuable in the parent.

Is a genomic map necessary?

Based on the majority of research, identiﬁcation of markers associated with
quantitative traits necessarily begins with a highly-saturated genomic map. In fact, Dudley
(1993) states, ‘For marker-assisted selection to be effective, a highly saturated map is
necessary.” A map is useful in that it provides a pool of markers ﬁ'om which to choose
such that a good representation of the entire genome is obtained. In this way, one can
choose markers every 10 cM (centimorgans) and no region of the genome is overlooked.

Although maps are available for many crop species, some are still under development and

14
others may not be applicable to conventional breeding programs. To provide suﬁ'rcient
polymorphisms, maps are oﬁen generated from crosses between diverse germplasm of
interspeciﬁc origins. In common bean, for example, Nodari et al. (1992 a and b) developed
a map from two diverse crosses: Andean X Mesoamerican and an Andean X wild bean. It
is uncommon, however, for a traditional plant breeding program to develop populations
ﬁ'om such wide crosses. The most useﬁrl crosses for successful cultivar development are
made between related germplasm, and in the case of common bean, germplasm which
belongs to the same market class. Consequently, the markers located on the map
developed by Nodari et al. (1992a and b) will not exhibit polymorphism between
genotypes within the same market class and will not be useful to the plant breeder.
Random Ampliﬁed Polymorphic DNA (RAPD) markers are available which demonstrate
polymorphisms between related germplasm in common bean (Haley et al., 1994c). For
genomic maps to be effective, they would need to be developed for each market type and
possibly speciﬁc crosses. Such limitations could reduce the eﬁcacy of MAS especially for
a self-pollinated crop such as common bean.

To create an eﬁcient tool for plant breeders involved in diverse population
improvement, marker identiﬁcation should be feasible without a genomic map. For plant
breeding purposes, the exact location of QTLs controlling a trait is not necessary.
Traditionally, improvements within crop species have been made without knowledge of
speciﬁc gene locations or even fundamental knowledge of the genetic material. If one

could, however, identify reproducible markers which were consistently associated with a

15
trait of interest, these markers could be used to indirectly select for that trait whether or
not the exact chromosomal position is known.

This does not preclude the useﬁrlness of genomic maps for gene tagging. Where
they are available, maps facilitate selection of markers which uniformly cover the entire
genome. There has been much debate over the number of markers which should be used
to identify accurate associations. Lande and Thompson (1990) suggested that the total
number of markers required depended on a variety of factors which include the total
recombination map length, number of generations since the last hybridization and mode of
reproduction. A rough estimate would be a few hundred markers which could be reduced
for a self-pollinating species. Zeng (1994) suggested that too many markers in a model,
even those which are unlinked to a particular region, could increase the critical value of
the test statistic and decrease the power of the test especially if sample sizes are small.

If a molecular map is not available or mapping the exact position of a QTL is not
an objective, polymorphic markers must be randomly selected. In this case, the number of
markers which are needed to detect useful QTLs cannot be estimated. By choosing
markers at random, the probability that many of these markers will be associated in linkage
groups increases. This would dramatically increase the number of markers required to
cover the entire genome. However, if an efﬁcient marker which is both time and cost
eﬂ‘ective is available, the time and expense used to analyze more markers would more than

compensate for the cost of developing a complete genomic map.

16

Identifying and Evaluating QTLs over Multiple Locations

Conventional breeding for quantitative traits, which are dramatically inﬂuenced by
the environment (as with yield-related traits) involves evaluation over several locations.
We can conclude from this, that to obtain accurate data for marker identiﬁcation, marker
studies should also include evaluation of the quantitative trait in several environments.
Two studies, however, demonstrated no QTL X Environment interaction (Edwards et al.,
1992; Beavis et al., 1994) whereas Bubeck et al. (1993), did observe inconsistencies over
environments for QTLs controlling gray leaf spot resistance in maize. In tomato, Paterson
et al. (1991) identiﬁed only a few QTLs that were consistently associated in all three
environments evaluated. The majority were associated in only one environment (Paterson
et al., 1991). The predictive value of the set of markers associated with speciﬁc gravity in
potato was much weaker when analyzed across environments (F reyre and Douches,
1994a). Conclusions ﬁ'om single environment studies are limited to the environment in
which the trait is measured and can often underestimate the number of QTLs that
contribute to that trait (Edwards et al., 1987). Thus it is imperative that, for a broad
application of markers over environments, populations must ﬁrst be tested over a range of
locations. From combined data, markers can be more accurately associated. If there is a
strong QTL X Environment interaction, environment-speciﬁc markers are still useful. If
associated markers are speciﬁc for this particular region, MAS remains applicable for a

conventional breeder who is working in that region.

17

Statistical Methods for Identifying QTLs

As more research is conducted on mapping quantitative traits, the selection of the
most appropriate statistical analysis may vary with objective. Stuber et al. (1987) and
Edwards et al. (1987) ﬁrst identiﬁed QTLs in maize using single factor analyses of
variance where individuals were grouped by marker genotype; homozygous for parent A,
homozygous for parent B and heterozygous. Signiﬁcant F -tests at P < 0.05 were used to
determine marker associations. The percent variation explained by this marker was
indicated as an R2 (coeﬁcient of determination) value. When the R2 is small one cannot
distinguish between QTLs that are tightly linked but have a small effect ﬁ'om those that are
not as tightly linked but have a large etreet (Edwards et al., 1937). To resolve this problem
interval mapping was adopted for use in mapping crop species which is equivalent to the
regression of the phenotype on the F1 genotype (Doerge and Churchill, 1994). Instead of
looking at individual markers, the interval between pairs of markers is analyzed making it
possible to distinguish closely linked QTLs with small effects from QTLs with a higher
recombination ﬁ'equency (Doerge et al., 1994). The maximum likelihood estimate involved
in this method is a ratio between the likelihood a QTL is present in a particular interval
and the likelihood it is absent. Based on the data, a LOD score is calculated and if it
exceeds a predetermined threshold, one would assume a QTL is present within this
interval (Lander and Botstein, 1989). Interval mapping is favorable because when a QTL
is identiﬁed between two marker loci, both these ﬂanking markers are available for
selection. This will increase the efﬁciency of selection especially if the marker interval is

large (Tanksley, 1983). Although this is the most widely recognized method of analysis, it

18
is subject to problems inherent to quantitative analysis. Speciﬁcally, a generalized
approach for determining threshold values is unavailable.

Doerge and Churchill (1994) outlined three major problem areas for determining
threshold values. First, unless many loci contribute to a particular trait, the true
distribution will be a mixture of distributions. Unfortunately, quantitative trait mapping
often involves the positioning of a few major genes with large eﬁ‘ects. The second problem
arises due to multiple hypothesis testing. Because the same observed data is used to test a
number of markers, these tests can no longer be considered independent. Therefore,
typical threshold values cannot be applied. Lastly, many experiment-dependent factors
such as sample size, genome size, marker density, proportion of missing data and
segregation distortion all inﬂuence the distribution of the test statistic (Doerge and
Churchill, 1994).

A permutation test has been proposed by Doerge and Churchill (1994) to eliminate
the problem of determining effective threshold values. In this test, the observed data is
randomly reassigned to other individuals. If there is truly no association between a marker
and QTL, the shuﬁling will have no effect. Any signiﬁcant marker-QTL association arising
ﬁ'om the permuted “data will be the result of chance. Repeating this a number of times
(recommend 1000 times for a a. = 0.05) results in the distribution of the test statistic from
which an experiment-speciﬁc threshold value is obtained.

Alternative statistical methods have been recently suggested but not widely
utilized. One such approach expands on interval testing. Problems with multiple QTL

eﬁ‘ects arise, using interval mapping, when two or more QTLs are located on the same

19
chromosome. This can result in mapping QTLs to the wrong position (Zeng, 1994). Zeng
(1994) proposed a method whereby the test statistic is not inﬂuenced by the QTLs located
outside the interval being tested. Multiple linear regression is combined with interval
testing to more thoroughly utilize the mapping information. Jansen and Starn (1994) have
also proposed an alternative approach. To eliminate requirements of large population sizes
like those required for conventional interval mapping, this technique uses markers as
cofactors to simulate simultaneous mapping of multiple QTLs and uses F1 and parental
data to estimate environmental error. In this way, environmental and genetic variation can

be more discretely separated thereby reducing error (Jansen and Starn, 1994).

Population Development

When developing populations for use in quantitative trait marker studies, several
factors must be considered. One such factor is the mode of reproduction of the crop to be
studied. If the crop is naturally cross-pollinated, an F2 or backcross population will be the
most appropriate. For a self-pollinated species, recombinant inbred lines are typically used.
Another consideration is the type of marker to be used in the study. If a co-dominant
marker is being employed, an F2 or backcross population will be most informative since
heterozygous marker genotypes can be readily scored. Dominant markers, however,
cannot distinguish the heterozygous from the homozygous-dominant individual.
Therefore, for dominant markers it is advantageous to develop RILs where individuals are
completely homozygous. Because the density of crossovers is doubled in RILs, the LOD

threshold for interval mapping is increased by 0.3, thereby increasing the number of

20
required progeny. This is offset, however, because replicated progeny will decrease the
environmental error which, in turn, reduces progeny number (Lander and Botstein, 1989).
Additionally, any form of replicated progeny is more effective because the genotypes can
be reproduced indeﬁnitely. This is most applicable to self-pollinated crops like common
bean. Experiments can be conducted simultaneously in several environments with the
information being cumulative (Soller and Beckrnan, 1990). For a self-pollinating crop,
RILs are the most effective type of population if estimating additive and dominance eﬁ‘ects

is not an objective (Paterson et al., 1991).

Population Size

Population size is another aspect of quantitative trait mapping which is much
debated. It was originally postulated that from 1000 to 2000 progeny must be evaluated to
determine signiﬁcant associations between QTLs and molecular markers (Beavis et al.,
1994). However, this is dependent on the trait and population type being studied. Lande
and Thompson (1990) suggested an inverse relationship between heritability and
population size. If a trait has low heritability, the study will require a large population size
-- up to 1000 individuals. Edwards et al. (1987) and Stuber et al. (1987) used

approximately 2000 individual F2 plants to identify QTLs associated with 82 quantitative

traits in maize. Paterson et al. (1991) demonstrated that QTLs could be identiﬁed using
350 F2 individuals in addition to multiple location experiments using F2 derived progeny in
corn. For plant breeding purposes, it is not economical to maintain populations of more

than 100 individual genotypes (Beavis et al., 1993). Beavis et al. (1993) proved that QTLs

21
can be identiﬁed ﬁ'om a small typical population of about 100 individuals. He also
emphasized the advantage of replicated progeny. Studies in potato that mapped tuber
dormancy and speciﬁc gravity used 110 genotypes (F reyre and Douches, 1994a and b)
whereas gray leaf spot resistance was mapped in maize using two populations of 139 and
193 Pm RILs (Bubeck et al, 1993). Although markers have been identiﬁed and mapped
which are associated with quantitative traits in potato and maize, the application of these
markers to plant breeding remains undecided. However, it is possible to identify QTLs
controlling quantitative traits with smaller populations but it must be stated that if a trait is
controlled by many genes the number of QTLs inﬂuencing the trait will be underestimated
(Beavis et al, 1994).
Markers

In addition to population type and size, the type of marker must also be
considered. Factors contributing to this decision include, crop species, relatedness of
parents, population type, and resources (economic and labor). Although isozymes
(different electrophoretic mobilities of an enzyme) are inexpensive, easy to generate and
c0odominant, they are not widely used because they are limited in number (Dudley, 1993).
Restriction Fragment Length Polymorhpisrns (RFLPs) are probably the most commonly
used marker because they are co-dominant and abunth in most crop species.
Unfortunately, they are expensive, time-consuming, and involve radioactivity and a large
supply of DNA (Tanksley, 1983). Additionally, polymorphisms are limited within
intraspeciﬁc crosses between related germplasm in some species (Foolad et al., 1993;

Haley et al., 1994c).

22

Randomly Ampliﬁed Polymorphic DNA (RAPDs), a PCR-based marker (\Vrlliams
et al., 1990), do not involve radioactivity or large amounts of DNA They are relatively
inexpensive and less labor intensive than RFLPs. But, more importantly, RAPDs generate
a high degree of polymorphisms between related germplasm. A RAPD map was developed
in' tomato as an alternative to the interspeciﬁc RFLP-based map because there is
insuﬁicient polymorphism within the cultivated species when using RFLPs. Consequently,
the interspeciﬁc tomato map was not useful for breeders who manipulate intraspeciﬁc
variation (Foolad et al., 1993). For the same reason RAPDs are often chosen for MAS
development in common bean. RAPDs will generate sufﬁcient levels of polymorphism
even between bean genotypes of the same market class (Haley et al., 1994c). Although
their potential usefulness is limited because of their dominant nature, RAPDs are suitable
for analyses involving RIL populations where progeny are completely homozygous.

The objective of this study was twofold. The ﬁrst was to identify RAPD markers
that were associated with performance under drought in common bean in the absence of a
genomic map. Because drought is a major constraint in bean production areas, worldwide,
it is an important quantitative trait to target for improvement. By identifying associated
markers, drought tolerant genotypes could then be selected based on their marker
genotype in areas where drought conditions are not prevalent, in earlier generations, or
both. This would greatly accelerate cultivar improvement. The second objective was to
determine if the identiﬁed markers would be effective in selecting superior genotypes and
test the applicability of MAS for a quantitative trait in common bean. Although elaborate

and in-depth processes for identifying and mapping quantitative traits exist, we wanted to

 

 

23
determine if markers could be identiﬁed without a map and then used successfully to select
superior germplasm. In this way, the applicability of MAS as a tool for conventional plant

breeders interested in improving a quantitative trait could be more accurately evaluated.

MATERIALS AND METHODS

Population Development

Parents

Three lines, one an adapted pinto cultivar from Michigan and two Mexican
breeding lines, were selected as parents to create two populations. Sierra pinto which
exhibited a degree of drought tolerance was used as a parent in both crosses because of its
commercial seed type and upright architecture. AC1028, a bayo seed type, and Lef-2RB,
an ojo de cabra seed type, were both drought-tolerant Mexican breeding lines adapted to
the Mexican highlands. Although the main objective for germplasm selection was to
create populations that will facilitate genetic studies and molecular marker identiﬁcation,
breeding considerations such as seed type and architecture, which could contribute to the
release of an adapted, drought-tolerant cultivar, were also included.
Populations

Original crosses were made in 1986 between Sierra/Lef-ZRB and Sierra/AC1028
to create two separate populations. Two F3 plants ﬁ'om each of ﬁfty Frderived families
from both crosses were advanced by single seed descent to the F5 generation ﬁ'om 1987 to
1990 using nurseries in both Michigan and Puerto Rico (winter nursery). Seed ﬁ'om F5
plants was bulked to create 95 F5 derived RILs in the Sierra/Lef-ZRB (S/L) population

and 78 RES in the Sierra/AC1028 (S/A) population.

24

25

Field Study

Field experiments designed to study drought tolerance were planted in seven
environments over ﬁve years. Locations included Montcalm County, MI, 1990; Ingham
County, MI, 1991; Madero, Durango, Mexico, 1992; and at two Mexican locations -
Madero, Durango and Calera, Zacatecas, Mexico, 1993 and 1994. For each experiment,
genotypes were grown under two treatments (stress and non-stress) with two replications
per treatment.

On June 12 and 13, 1990, the F45 lines and parents from both S/A and 8/1. were
planted, by hand, in Montcalm County, MI. Experiments were designed as 9X9 (S/A) and
10X10 (S/L) square lattices with two replications per treatment. Sierra was used twice in
the 10X10 lattice to complete the number of entries in the lattice. This design was used for
both stress and non-stress treatments. The soil type in Montcalm County is McBride
sandy loam (coarse-loamy, mixed, mesic Alﬁc Fragiorthods). The ﬁeld has a 2-6% slope
and drought stress was induced by planting along this slope. Seafarer, an adapted bush
cultivar, was planted between rows to establish uniform competition and to compete for
moisture. Single-row plots 1.6 m in length and 0.50 m wide were used with a seeding rate
of 13 plants per meter. Germination was not uniform so yield and biomass per plot were
adjusted using plant stand as a covariate. Fertilizer treatment consisted of 53 kg ha'1 of N,
P205, and K20 banded at planting. Weeds were controlled using pre-plant incorporated
treatment of trifuralin (0.75 kg ha'1 active ingredient), pre-emergence treatment of

chloramben (3 kg ha‘1 active ingredient) supplemented with hand hoeing as needed

26
throughout the season. A single application of dirnethoate (1.12 kg ha") was also used to
control potato leaf hoppers (Emposca fabae) prior to ﬂowering. Plots were harvested at
dry maturity (0.5 m2) and yield per plot, biomass, and days to maturity were measured. An
individual plant ﬁom each F45 line was selected and the seed bulked to create Fs-derived
recombinant inbred lines.

Experiments in 1991, were conducted under the rain shelter at the Agronomy Farm
on MSU’s campus, Ingham County, MI. Plots were planted in the same experimental
design as in 1990 for both populations. The stress treatment was applied by planting RILs
under a rain shelter while non-stress plots were planted outside the shelter and received
normal rainfall. Stress treatment was initiated prior to ﬂowering by closing“ the shelter
ahead of actual precipitation and at night. F 5,6 lines were planted, by hand, on June 7, 1991
in single row plots on Capac soil (ﬁne-loamy, mixed, mesic Aeric Ochraqualfs). Single-
row plots were 1 m long and 0.35 m wide, with a seeding rate of 13 plants per meter.
Fertilizer and weed control was similar to the 1990 growing season. Plots were harvested
at dry maturity (0.33 m2).

In 1992, ﬁeld experiments were conducted in Madero, Durango, in the Mexican
highlands (1932 masl) where drought conditions prevail. Seventy-eight F27 S/A lines,
AC1028, Sierra and a locally adapted pinto cultivar, Pinto Villa, were planted in a 9X9
lattice with two replications per treatment. Ninety-ﬁve F27 8/1. lines, all three parents and
Pinto Villa were planted in a 10X10 lattice with two replications per treatment. Plots were
planted on June 29, 1992 on a Xerosol Haplico (FAO classiﬁcation) soil which has low

water holding capacity. The moisture-stress treatment was considered normal rainfall

27

conditions (226 mm precipitation) whereas the non-stress treatment had three
supplemental irrigation treatments (50 mm of water each) over the growing season. Plots
consisted of two rows 3 m long and 0.5 m wide with a seeding rate of 13 plants per meter.
Supplemental fertilization at the rate of 56 kg ha"1 N, and 78 kg ha"1 P205 was added to
the soil at planting and weeds were controlled by hand hoeing as needed. Lines were
harvested at dry maturity (3 .0 m2) and yield, biomass and 100 seed weight were measured
for each plot.

Field experiments were planted at two locations in Mexico in 1993: Madero, Dgo.,
and Calera, Zacatecas (2200 masl). At Madero, plot and experimental design was similar
to 1992 ﬁeld experiments except that the non-stress treatment received only one
supplemental irrigation. Total precipitation for this location was 207 mm. The F2; lines
ﬁ'om S/L and S/A were planted at Madero, Dgo. in a 10X10 and 9X9 square lattice,
respectively, on July 1, 1993. Parents and a check cultivar, Pinto Villa were also included.
Plots were planted on June 29, 1993 at Calera, Zac. using the same experimental design as
Madero, Dgo. Plots consisted of single-rows 6 m long and 0.76 m wide with a seeding
rate of 13 plants per meter. Five meters were harvested at dry maturity for a total harvest
area of 3.8 m2. Fertilizer and weed control were similar at both locations to the 1992
growing season. The soil at Calera is Luvic Castanozem (F AO classiﬁcation). Stress
treatments received normal rainfall conditions (260 mm of precipitation) whereas non-
stress treatments had two supplemental irrigation treatments (50 mm each) during the

growing season. Evaporation measured at this location was 780 mm.

28

Experiments at Calera, Zac. in 1994 were planted on June 29 and at Madero, Dgo.
on July 8. Similar planting rates as 1993 were used with plots 3 m long and a between row
spacing of 0.76 m. The experimental design was similar to the 1993 ﬁeld experiments.
Calera, Zac. received 345 mm of rainfall and two supplemental irrigations were applied to
the non-stressed treatments. Rainfall at Madero, Dgo., in 1994 was 270 mm and three
supplemental irrigations (20 mm each) were applied to the 9X9 non-stressed experiment
and two supplemental irrigations were applied to the 10X10 non-stress experiment.
Production practices for the 1994 growing season were similar to those used in 1993 at
both locations.

An additional ﬁeld experiment was conducted in 1994 in the rain shelter at the
Kellogg Biological Station, Kalamazoo County, MI. Nineteen selected genotypes and
parents ﬁ'om S/A were grown in a split-plot design with two treatments and four
replications. Nineteen genotypes were indirectly selected based on molecular marker data
associated with yielding ability under water stress (explained in the following section). Ten
potentially above average yielding (PAA) and nine potentially below average yielding
(PBA) genotypes were identiﬁed ﬁom S/A where Sierra was considered above average
and AC 1028 below average. Stress treatments were considered main plots and were split
for genotype. Plots were planted by hand in a Spinks sand (sandy, mixed, mesic
psarnmentic hapludalfs) in single rows 3 m long and 0.5 m wide, on June 15. Plots were
thinned to approximately 13 plants per meter one week later. Seed was treated with a
combination fungicide and insecticide containing thiophanate methyl, captan, and lorsban.

Fertilizer was applied as a split application at the ﬁnal rate of 43 kg ha'1 N, P205, and K20.

29
Plots were treated with the fungicide benomyl (1 kg ha" a.i.) to control anthracnose
(Colletor‘richum Iindemuthianum) and the insecticide dirnethoate (1 kg ha'1 a.i.) was
applied to control potato leaf hoppers (Emposca fabae). The shelter remained open and
plots received normal rainfall until 50% of the plants had at least one ﬂower. At this time,
the rain shelter closed automatically after sensing 15 mm of continuous rain. Control plots
were irrigated by overhead sprinlders once a week with 13 mm of water for 14 weeks (for
a total of 178 mm). Weeds were hand-hood as needed. During the ﬁrst week of October,
plants were pulled and number of pods per ﬁve plants and days to maturity were recorded.
Biomass was measured, plants threshed and yield per plot and 100 seed weight were

recorded.

Marker Protocol

DNA Extraction

Several young trifoliates ﬁ'om each F45 plant and parent genotypes were harvested,
lyophilized and ground. Ground. tissue was stored at -80 °C. DNA was extracted
following the protocol described by Saghai-Maroof et al. (1984) and modiﬁed by Miklas
et al. (1993). The protocol was altered slightly as follows: chloroformzisoamyl alcohol
(24:1) was used instead of chlorofonnzoctanol, RNase A was added prior to the second
chloroformzisoamyl alcohol extraction, and DNA was resuspended in 0.1 M Tris Acetate
EDTA buffer pH 8.0.

When extra DNA was required, a second mini-extraction was performed following

the procedure outlined by Edwards et al. (1991) and modiﬁed according to Haley et al.

30
(1994c). Irrespective of DNA extraction method, samples were quantiﬁed by DNA
ﬂuorometry (Hoefer TK0100, Hoefer Scientiﬁc, San Francisco, CA). A sub-sample was
then diluted to a standard concentration of 10 ng pl" for ampliﬁcation by the polymerase

chain reaction (PCR).

Polymerase Chain Reaction (PCR) Protocol

PCR procedure followed that of Miklas et al. (1993) with slight modiﬁcations
(Haley, et al. 1994a). Random primers were used ﬁ'om selected Operon kits (Operon
Technology, CA) to amplify random regions of the genome. DNA was ampliﬁed using a
Perkin Elmer Cetus DNA Thermal Cycler 480 using the following cycles: 3 cycles of 1
min. at 94 °C, 1 min. at 35 °C, and 2 min. at 72 °C, 34 cycles ofl min. at 94 °C, 1 min. at
40 °C, and 2 min. at 72 °C (ﬁnal step extended by 1 sec for each of the 34 cycles), and a

ﬁnal extension cycle of 5 min. at 72 °C (Haley et al., 1994a).

Electrophoresis
Approximately 20 ul of ampliﬁed DNA from each sample was run on a 1.4%
agarose gel containing ethidium bromide (0.5 ug ml“) 40 mM Tris-acetate, and 1 mM

EDTA. DNA was viewed under ultraviolet light and photographed for permanent record.

3 1
Marker Scoring and Nomenclature

Six hundred single decamer primers were screened against the parents of each
population to detect polymorphisms. Seventy of these polymorphic primers were selected
to use for RAPD analysis on the RILs ﬁ'om each population. The resulting marker data
was scored as a l for RILs which lacked the RAPD band and a 2 for RILs which had the
band.

Nomenclature for RAPDs followed that of Miklas et al. (1993). The preﬁx, ‘0”
designates the commercial RAPD primer kits ﬁom Operon Technologies (Alameda, CA).
The kits used for this study were from A to AZ with 20 diﬂ‘erent primers per kit. Thus a
RAPD speciﬁed as 0A08950 would signify RAPD primer 8 of kit A from Operon
Technologies. The number in subscript is the size, in base pairs, of the identiﬁed RAPD.
For those markers that were not sized and exhibited more than one polymorphism, lower
case letters were used in alphabetical order starting with the longest RAPD from that
primer ﬁrst.

Statistical Analysis

Field Studies

The following analyses were applied to each treatment (stress and non-stress),
every year using both populations, separately. Yield (g m'z), biomass (g m4) and 100 seed
weight (g) were analyzed annually. Data from individual treatments in 1990 and 1991
were analyzed as a randomized complete block design with number of plants per plot as a

covariate. The 1992 data were analyzed as a randomized complete block design with no

 

32

covariate. Four experiments were conducted in 1993 and 1994 for each population. Each
experiment was analyzed as a 9X9 and 10X10 square lattice design for S/A and S/L,
respectively. Means calculated from the above analyses ﬁ'om ﬁve locations in the years
from 1990 to 1993 were combined and analyzed as a split-plot design with environments
as replications (MSTATC, Michigan State University). Means from each stress treatment
analysis of variance were used with the means ﬁ'om the corresponding non-stress
treatment analysis to calculate a single geometric mean for the two treatments for each
RIL in each year. This was also performed with the means resulting from the combined
analysis which generated a single stressed yield (Yd), non-stressed yield (Y p), and
geometric mean (GM) value for each RIL in ﬁve environments and over all years.

Heritability estimates were calculated for yield, biomass and 100 seed weight on an
entry mean basis using data from 1990 to 1993 under stress and non-stress treatments
according to Allard (1960). Using these heritability estimates, gain per cycle of selection
was calculated at varying selection intensities of 5, 10, 20, and 30% according to Allard
(1960). Drought intensity index (DII = l-Xd/Xp where Xd is the mean over genotypes for
the stress treatment and Xp is the mean over genotypes from the control treatment) was
also calculated for each environment and overall experiments according to Fisher and
Mauer (1987).

Correlations between all combinations of the following traits were calculated
within this population: Yd, Yp, biomass, 100 seed weight, and ID under stress and non-
stress conditions, D81 and GM. D81 is a drought susceptibility index developed by Fisher

and Mauer (1978) where DSI = (1-Yd/Yp)/DII. Yd and Yp were correlated independently

33
with drought tolerance (Yd - Yp) and arithmetic mean ((Yd + Yp)/2) according to
Rosielle and Haman (1981). Correlations between arithmetic mean and tolerance were
also examined.

Yield, biomass, 100 seed weight, pods per plant and days to maturity collected
ﬁ'om the 1994 experiment at Kalamazoo County were analyzed as a split-plot design with
four replications for each of the two treatments (MSTATC, Michigan State University).
Stress and non-stress treatments were also analyzed individually as randomized complete
block designs. Genotypic groups based on marker data were contrasted using standard
orthogonal comparison analysis with signiﬁcance set at P < 0.05 (Little and Hills, 1978).
Contrasts between PAA and PBA were analyzed for yield, biomass, 100 seed weight, and
number of pods per plant for each treatment analysis and the combined split-plot design
analysis. Correlations between all combinations of the following traits were calculated
within this population: Yd, Yp, biomass, 100 seed weight, number of pods per plant, and
HI under stress and non-stress conditions, DSI and GM.

Markers

RAPD data from each RIL based on 70 markers per population were analyzed in a
one-way analysis of variance against Yd, Yp and GM (SAS, proc glrn) from the previously
mentioned analyses of variance for each environment and over all years. Associations were
determined by F-tests with signiﬁcance at P < 0.05.

MAPMAKER (Lander and Botstein, 1989) ‘group” command was used to identify
linkage groups among markers. Default linkage criteria were set at a distance of 10

centimorgans (cM) and a LOD score of 4.0 for a stricter conﬁdence interval. All RAPDs

34
from each resulting linkage group were analyzed in a multiple regression analysis against
Yd, Yp and GM ﬁ'om each environment and over all years (SAS, proc glm). Signiﬁcance
was set at P < 0.05.

Based on this data, four markers were identiﬁed that appeared to be associated
with QTLs controlling drought tolerance in S/A Two RAPDs were members of the same
linkage group and the other two were independently associated. These markers were
subsequently used to select genotypes within S/A which potentially differed in their
performance under water stress. The 19 selected genotypes were planted with Sierra and
AC1028 parents at Kalamazoo County in 1994 to determine if signiﬁcant diﬁ‘erences
existed between genotypic groups with potentially above average (PAA) and potentially
below average (PBA) marker genotypes (described above). Similar contrasts were
performed using the mean performance data resulting ﬁ'om the 1994 lattice analyses of the
stress and non-stress treatments at the two Mexican locations (Madero, Dgo. and Calera,
Zac). Genotypic groups were contrasted for yield and biomass using Madero, Dgo. data
and for yield only using data ﬁ'om Calera, Zac. Mean values from yield analyses were also
used in a split-plot design with stress and non-stress treatments as main plots and
genotypes as sub-plots considering the two locations as replications. Comparisons were
made between the means of the genotypic groups resulting from this analysis. Values ﬁ'om
each treatment over the two locations were combined and analyzed as a one-factor
randomized complete block design over locations. The same contrasts mentioned above

were conducted on the combined data.

3 5

Sixty-one RILs were indirectly selected based on their marker genotype in S/L. All
ﬁve markers from linkage group one were used to select a group of 34 potentially above
average (PAA) yielding RILs plus Sierra parent which were compared to a group of 27
potentially below average (PBA) yielding RILs plus Lef-2RB. Orthogonal contrasts were
used to determine if signiﬁcant diﬁ‘erences existed between the two groups. The groups
were contrasted in stress and non-stress environments for yield, biomass and 100 seed
weight at two Mexican locations (Madero Dgo. and Calera, Zac.) in 1994. Signiﬁcance
was set at P < 0.05. An above average (AA) performing group of thirty ﬁve RILs and a
below average (BA) group of twenty eight RILs were selected based on Yd data
combined over four years (1990-1993) and ﬁve locations in Michigan and Mexico. This
was repeated for Yp and GM. The objective was to compare the eﬁiciency of
conventional selection to MAS. Using the yield data from 1994 combined over two
locations, the groups identiﬁed based on conventional selection were compared in the
same fashion to the groups identiﬁed by MAS. Thus, the three above average groups (AA)
selected based on Yd, Yp, and GM data sets were contrasted against the three below
average groups (BA) in stressed and non-stressed environments, independently, in

addition to the combined split-plot analysis.

RESULTS

Field Study

Mean yields for S/A from 1990-1993 ranged from 76 to 150 g m'2 with a grand
mean of 114 g rn'2 for the stress treatment (Tables 1 and A7). The mean yields for the
non-stress treatment ranged from 99 to 214 g rn'2 with a grand mean of 155 g m'2 (Tables
1 and A8). The overall drought intensity index was moderate (0.26) (Table 6). The
difference (41 g m'z) in yield between means of the two treatments was statistically
signiﬁcant (Table A5). The mean values for biomass ranged from 136 to 276 g rn'2 with a
grand mean of 204 g rn'2 for the stress treatment (Tables 1 and A7) and ﬁ'om 157 to 328
g rn'2 with a grand mean of 247 g in2 for the non-stress treatment (Tables 1 and A8). The
diﬁ‘erence between means (43 g m'z) of the two treatments for biomass was non signiﬁcant
(Table A5). One-hundred seed weight ranged ﬁom 22.4 to 34.6 g for the stress treatment
with a grand mean of28.6 g and ranged from 21.9 to 33.9 g with a grand mean of28.2 g
for the non-stress treatment (Table 1). There was a non signiﬁcant difference (0.4 g per
100 seed) between treatments for seed size (Table AS). Sierra ranked 67th for yield under
drought (Yd) with a mean of 96 g rn’2 and ranked 69th for yield under non-stress
conditions (Yp) with a mean of 133 g m'2 (Table A7 and A8). AClozs ranked 44th for

Yp and 50th for Yd with means of 111 and 147 g m'z, respectively (Table A7 and A8).

36

37

Table 1. Analysis of variance combined over ﬁve locations ﬁ'om 1990—1993 in Michigan
and Mexico for 80 bean genotypes. Stress and non-stress treatments were analyzed
individually for the Sierra/AC1028 population. '

 

 

 

 

 

 

 

 

Stress Non-stress
Source DF MS F Test MS F Test
------Yield (g m")————
Grand Mean 114.3 154.6
LSD (at = 0.05) 14.6 16.6
CV (°/o) 29.7 24.9
Location 4 3561246 2969027
Error 8 5 3859.4 33496.0
Genotype 79 2953.0 2.5"" 4822.3 3.2"”
GXE 316 1652.4 1.4““ 2581.8 1.7"”
Error b 395 1150.5 1487.1
------Biomass (g m")—-—
Grand Mean 203.5 247.1
LSD (at = 0.05) 23.0 24.5
CV (°/o) 26.4 23.4
Location 4 6134516 4233422
Error a 5 3986.6 35303.1
Genotype 79 10900.0 3.8"" 11715.2 3.6""
GXE 316 5606.3 1.9"“ 5222.3 1.6""
Error b 395 2879.3 3242.4
-—--—--100 Seed Weight (g)———
Grand Mean 28.6 28.2
LSD (at = 0.05) 0.9 1.0
CV (7.) 7.3 7.8
Location 3 2308.4 1431.0
Error a 4 4.1 4.7
Genotype 79 47.2 11.0"" 49.8 10.6""
GXE 237 10.6 2.5“" 13.0 2.8““
Error b 316 4.3 4.7

 

I’P < .05; “ P <.01; "" P < .001, "" P < .0001

 

38
Two genotypes, T3008-1 and T3016—1, ranked in the top 6% for both Yd and Yp and
three genotypes, T3020-2, T3048-1 and T3004-1 ranked in the bottom 6% for Yd and Yp
(Tables A7 and A8).

Mean yields for S/L from 1990-1993 ranged from 77 to 133 g m'2 with a grand
mean of 104 g m"2 for the stress treatment (Tables 2 and A9). The mean yields for the
non-stress treatment ranged ﬁ'om 112 to 180 g m’2 with a grand mean of 150 g m'2 (Tables
2 and A. 10). The overall drought intensity index for this population was 0.30 (Table 6).
The difference in yield (46 g m'z) between the two treatments was statistically signiﬁcant.
The mean values for biomass ranged ﬁ'om 142 to 242 g rn’2 with a grand mean of 187
g rn'2 for the stress treatment (Tables 2 and A9). Biomass under non-stress ranged ﬁ'om
237 to 277 g m2 with a grand mean of 237 g m2 (Tables 2 and A. 10). The difference (50
g m'z) between treatments was signiﬁcant (Table A. 6). One-hundred seed weight ranged
ﬁ'om 23.1 to 32.5 g under stress and from 24.0 to 32.7 for non-stress with grand means of
27.8 and 28.0 g, respectively (Table 2). There was a non signiﬁcant difference (0.2 g per
100 seed) between treatments (T able A6). Sierra ranked 88th for Yd and 10th for Yp
with means of 86 and 170 g m'z, respectively. Lef-2RB ranked 42nd for Yd with a mean
of 107 g m“2 and 37th for Yp with a mean of 156 g m'2 (Tables A9 and A.10). Two
genotypes, T3110-2 and T3147-2, ranked in the top 5% for both Yd and Yp. One
genotype, T3153-1, ranked in the bottom 5% for both Yd and Yp. However, T3153-2
ranked 5th for Yd and 19th for Yp. The two genotypes, T3153-1 and T3153-2 are derived

from the same F3 plant.

 

Table 2. Analysis of variance combined over ﬁve locations from 1990-1993 in Michigan
and Mexico for 97 bean genotypes. Stress and non-stress treatments were analyzed

39

individually for the Sierra/Lef-ZRB population.

 

 

 

 

 

 

 

 

 

 

Stress Non-stress
Source DF MS F Test MS F Test
—----—Yie1d (g m”)---—
Grand Mean 103.8 149.6
LSD (at = 0.05) 12.1 14.4
CV (7.) 27.1 22.5
Location 4 328121.5 3387443
Error a 5 2815.1 9631.8
Genotype 96 1488.2 1.9"" 2412.5 2.1“"
GXE 384 1290.5 1.6““ 1945.4 1.7"“
Error b 480 788.8 1131.5
-----—-Biomass (g m'2)-——-—
Grand Mean 187.1 236.6
LSD (at = 0.05) 20.7 21.1
CV (7.) 25.7 20.8
Location 4 1576195.] 7185912
Error a 5 9043.3 37829.0
Genotype 96 4296.8 1.9"" 4537.8 1.9"“
GXE 384 3718.5 1.6“" 3861.3 1.6“"
Error b 480 2318.4 2423.4
---—---100 Seed Weight (g)-———--
Grand Mean 27.8 28.0
LSD (or 8 0.05) 0.8 1.1
CV (7.) 6.5 9.4
Location 3 3124.3 2541.6
Error a 4 18.0 26.5
Genotype 96 36.8 11.1"" 35.6 5.2""
GXE 288 18.8 5.7"" 18.5 2.7““
Error b 384 3.3 6.9

 

I"P < .05; " P <.01; "" P < .001, "" P < .0001

 

 

40
Statistical Analysis

The results ﬁom the analyses of variance for each environment and treatment ﬁom
1990 to 1994 are shown in Tables A.1-A.2 (S/A) and Tables A3-A4 (SA) The error
variance was, in general, less under stress than non-stress. The genotypic effect was not
consistently signiﬁcant over all treatments and environments. Additionally, coemcients of
variation (CV) were unexpectedly high except for the S/A experiments grown in Calera,
Zac. in 1993 and 1994. Least signiﬁcant diﬂ‘erence (LSD) values for yield and biomass
based on two replications per treatment were high and equal to approximately half the
grand mean for stress treatments whereas LSDs for the non-stress treatment were from
one half to a third of the grand mean.

Individual treatment analyses of variance combined over locations for yield,
biomass and 100 seed weight are shown in Table 1 (S/A) and Table 2 (S/L). Genotypic
variation for all traits was highly signiﬁcant as was the location by genotype interaction.
Overall, the error variance for all traits in both populations was less under stress than non-
stress. The genotypic variance was also less under stress than non-stress. LSDs were
lower in the combined analysis than was observed for analyses performed on individual
years. CV’s were high for all traits. The combined split-plot analyses over environments
for yield, biomass and 100 seed weight are shown in Tables AS and A6. For both
populations, signiﬁcant genotypic variation was observed for all traits in both populations.
Signiﬁcant treatment effects were demonstrated for yield in both populations and biomass
in S/L. Biomass in S/A and 100 seed weight for both populations failed to show a

signiﬁcant treatment effect. Data ﬁ'om neither population resulted in a signiﬁcant

 

41
treatment X genotype interaction for yield, biomass, nor 100 seed weight. Frequency
distributions for yield and biomass are depicted in Figures 1 and 2 (S/A) and Figures 3 and

4 (S/L).

Heritability Estimates and Gain from Selection

Heritability estimates (hz) for yield, biomass and 100 seed weight, calculated on an
entry mean basis, for both S/A and S/L were similar between moisture stress and non-
stress treatments (Table 3). The genetic variation, in S/A, was greater under non-stress
than stress, however, experimental error was also larger under non-stress than stress
treatments. Experimental error was also larger under non-stress except for Yd but genetic
variance was less under non-stress than stress in S/L. The h2 values calculated on an entry
mean basis for S/L were lower under non-stress than stress conditions for all three traits
whereas lower h2 were observed under stress than non-stress for S/A except for 100 seed
weight. S/A had much higher h2 and genetic variation for yield and biomass estimates than
S/L. Gains in yield per cycle of selection calculated for various selection intensities using
these heritability estimates are shown in Tables 4 and 5. In S/L, gain from selection was
reduced under non-stress compared to stress treatments but not by a large amount. In S/A
selection under non-stress conditions showed equivalent percent gain in selection as that

under stress for this population.

42

Frequency Distribution for Yd a,“ M o 114
F SierraIAC1 028 an: Error - 8.7
28 '1 X

 

 

n as as 104 11: 122 132 141 m.
Yield (9 m“)

Frequency Distribution for Yp

MN I 1“
W SierraIA61028 m Em'f'm
a q

o-Sierra

2° .. X' A0102.
1‘ .1
1O 1
. a
O 4

 

 

as 113 121 142 m m res 19s are
Yield (9 m“)

Figure 1. Frequency distributions for yield under stress (Yd) and non-stress (Y p) using

means for each of 80 genotypes combined over four years (1990-1993) and ﬁve locations

in Michigan and Mexico for the Sierra/AC1028 populations. Sierra is marked with an ‘0”
and AC1028 by an “X”.

43

Frequency Distribution for Biomass Under Stress

"mum, SierraIA01028
F Grand Mean I 204
18 - Std Error - 14.8

181 o-Siarra

XII AC1028
12-
104
.1

 

   

O 1 ‘
13. 1“ 171 1” 200 224 141 25. 27‘

Biomass (g m‘)

Frequency Distribution for Biomass Under Non-Stress
PM! SierraIAc1028
Grand Mean - 241
Std Error - 11.:

(“Sierra
x- A0102!

  
    

 

mmmmusauasssorum
mum“)

Figure 2. Frequency distributions for biomass under stress and non-stress using means for
each of 80 genotypes combined over four years (1990-1993) and ﬁve locations in
Michigan and Mexico for the Sierra/AC1028 populations. Sierra is marked with an ‘0”
and AC1028 by an “X”.

44

F SierralLef-ZRB StdError- 11.7

”a
1I<

124
104

 

.N..
aaaa

 

mum‘)

Frequency Distribution for Yp M m a m
sigrralLef-ZRB swam" 1“
Inequeney
1a ~

“4
121
10-

 

 

ON...

112 11. 117 188 142 150 158 1‘5 173 180
Yiﬂd‘gln‘)

Figure 3. Frequency distributions for yield under stress (Yd) and non-stress (Y p) using
means for each of 98 genotypes combined over four years (1990-1993) and ﬁve locations
in Michigan and Mexico for the Sierra/Lef-2RB populations. Sierra is marked with an ‘0”
and Lef-2RB by an “X”.

45

Frequency Distribution for Biomass Under Stress

 

     

    

SierraILef-2RB
Frequency Grand Mean I 187
as . std Error - 15.2
0 I Sierra
2!! , X I Let-2RD
18 a
10 a
‘ a
O -l
178 188 107 208 210 231 242
Biomass (g In")
Frequency Distribution for Biomass Under Non8tress
F SiernILef-ZRB . 1”
m Std Error I 14.8
2‘ 1 X
0 I Sierra
2. , x I Let-2RD
18 .
10 -
‘ 1 .....
a l

 

 

177 122 1” 210 221 282 243 2“ 2“ 277

Figure 4. Frequency distributions for biomass under stress and non-stress using means for
each of 98 genotypes combined over four years (1990-1993) and ﬁve locations in

Michigan and Mexico for the Sierra/Lef-ZRB populations. Sierra is marked with an ‘O”
and Lef-2RB by an “X”.

46

Table 3. Heritability estimates for yield, biomass and 100 seed weight for two bean
populations grown under stress and non-stress treatments over six locations and ﬁve years
(1990-1994) in Michigan and Mexico.

 

 

 

 

 

 

Sierra/AC1028 Sierra/LEF-ZRB
Stress Non-stress Stress Non-stress
Yield 0.45 0.47 0.11 0.08
s32 113.90 185.70 23.21 15.40
Error Variance 1183.90 1617.40 1662.42 1258.60
Biomass 0.50 0.56 0.11 0.05
982 454.98 559.94 28.35 19.78
Error Variance 2998.70 . 3704.70 1826.70 2786.60
100 Seed Weight‘ 0.77 0.74 0.61 0.58
as2 4.60 3.00 2.11 1.99
Error Variance 4.30 4.70 3.40 6.31

 

'Data for 100 seed weight was unavailable for 1990

47

Table 4. Gain in yield per cycle of selection at varying selection intensities (k) based on
heritability estimates (Table 3) for Sierra/AC1028 population grown under stress and non-
stress environments over ﬁve years (1990-1994) and six locations in Michigan and

Mexico.

 

 

 

 

Selection Gain from Selection
Intensity (%) k Stress Non-Stress
.2
5 2.06 16.1 21.0
10 1.76 13.7 17.9
20 1.4 10.9 14.3
30 1.16 9.0 11.8

 

Table 5. Gain in yield per cycle of selection at varying selection intensities (k) based on
heritability estimates (Table 3) for Sierra/Lef-ZRB population grown under stress and non-
stress environments over ﬁve years (1990-1994) and six locations in Michigan and

 

 

 

 

 

Mexico.
Selection Gain ﬁ'om Selection
Intensity (%) k Stress Non-Stress
8 ma

5 2.06 3.6 2.5

10 1.76 3.1 2.1
20 1.4 2.5 1.7
30 1.16 2.0 1.4

 

48
Drought Intensities
Drought intensity index (DII) ranged from 0.19 in 1990 to 0.49 in 1994 Calera,
Zac. experiments (Table 6). The highest value (DII = 0.76) was measured in Kalamazoo
County in 1994. Aside from the Kalamazoo experiment, drought intensities were
moderate. DIIs were equivalent between populations with means of 0.26 and 0.30 in S/A

and S/L, respectively.

Yield Correlations

The correlations calculated between Yp, Yd, and drought tolerance (Yd - Yp),
GM, arithmetic mean, and D81 are open to interpretation since Yd and Yp values were
used in the calculation of these parameters. As a result, trends and not actual correlation
values were considered. In S/A, all combinations of correlations between the measured
factors were signiﬁcant and positive except for those involving DSI, HI and GM (Table
7). DSI was inversely and signiﬁcantly correlated with Yd, and biomass under stress, and
positively and signiﬁcantly correlated to Yp and biomass under non-stress. HI under stress
was inversely related to biomass under stress. HI under stress was positively and
signiﬁcantly correlated with Yd and HI under non-stress. I-II under non-stress was
inversely proportional to biomass under stress and non-stress conditions, but positively
correlated to HI under stress, and Yp.

In S/L, Yd was signiﬁcantly and positively correlated to Yp, biomass under stress
and non-stress conditions, HI under stress and non-stress, 100 seed weight under non-

stress, and GM, but negatively correlated to DSI (Table 7). Biomass measured under

49

Table 6. Drought intensities index (DII)"' for two bean populations grown at six locations

over four years’.

 

 

 

Sierra/AC1028 Sierra/LEF-ZRB
1990 0.19 0.21
1991 0.24 0.34
1992 0.31 0.36
1993 Madero, Dgo. 0.42 0.44
1993 Calera, Zac. 0.25 0.26
1994 Madero, Dgo. 0.30 0.31
1994 Calera, Zac. 0.49 0.42
Combined (1990-1993)t 0.26 0.30
Kalamazoo, MI 0.76 N/A

 

*(D1 = l-Xd/Xp; where Xd = mean yield of stressed plants and XI, = mean yield of non-stress plants)

TCombined analysis was combined over all locations except for the experiments in Kalamazoo County,
Madero, Durango, and Calera, Zacatecas, 1994.

 

 

50

moisture stress was signiﬁcantly and positively correlated with Yp, biomass under non-
stress, and GM but negatively correlated with HI under stress, and DSI. One-hundred seed
weight measured under moisture stress was positively and signiﬁcantly correlated with 100
seed weight under non-stress. HI under stress was signiﬁcantly and positively correlated to
GM and HI under non-stress but negatively correlated with DSI. Yp was signiﬁcantly and
positively correlated with biomass under non-stress, HI under non-stress, DSI and GM.
Biomassunder non-stress was signiﬁcantly and positively correlated to DSI and GM but
was not correlated to HI under non-stress although biomass under non-stress was used in
the calculation of HI. HI under non-stress was positively correlated with GM (Table 7).

Correlations between traits measured at Kalamazoo County are shown in Table 8.
Yd was signiﬁcantly and positively correlated with biomass under stress, 100 seed weight
under stress, number of pods per plant under stress and non-stress treatments, HI under
stress, and GM while negatively correlated with DSI. Biomass under stress was positively
and signiﬁcantly correlated with number of pods per plant under stress, and GM and
negatively correlated with DSI. Signiﬁcant correlations between 100 seed weight under
stress and biomass under non-stress, 100 seed weight under non-stress, number of pods
per plant under non-stress, HI under stress and GM were all positively related. Number of
pods per plant measured under stress was negatively correlated with DSI. HI under stress
was positively and signiﬁcantly correlated to Yd, 100 seed weight under non-stress,
number of pods per plant measured under non-stress and HI under non-stress. Yield under

non-stress demonstrated signiﬁcant and positive correlations with biomass under non-

 

51

:5. V .— see ”SV m as Kc. V as

 

 

 

 

2.0. .3an Ed cached edemwd eeomd :6 2.3th second 20
2 6 n— .c. eeewmd seized on”? 3 .c. :31? ascend. 5D
2... a... :22 :82. Se So. an... 86 E
3... 2 .e 2 .c in»... o. .e .2... 66 3m 2:
:58... 8.? 8.... Same 3.3... GE 8985
2 .o 8.: :53 .12.: 86 so;
2 .c eNNdr seamed Amy E
2... 3... Q 3 8—
eeeohd Amy mama—2m
commence mango
had. ecNd dead .2536 2.8ch h— .c ae— n6 deemed sieved 20
2... So ammo an... 2 .o. 3.? tone. tone. an
S .o .25.. :42. ES... 85 :39 z .o azv E
.42. tom... 2 .c. team... in... an"... 62. 3m 2:
see—ed S .c. ennd 2506.: :5th Ame 35:05
2 .e 1.3 3.8... take 66 29>
3o. .33.... .3... Q E
as... :8... Q 5 8—
.386 Amy mans—ea
deceased «8.335%
in E 3m 2: spasm 29> E 3m 2: neaeE so;
modem 82 .83

 

6&on ecu comm—.22 E Eouaoo. 26 e5 Gee—teem S 28» .58 .26 e395 ocean—neon econ

be be 3898.98 azv 2828 es. E are a $6 8: came; 82 2: es. eases see scene 388on s can...

 

52

:5. v m 1.... ”EV .— ..i ”3. v m...

 

 

 

 

mud. ound 93.2.6 .36 zoned .3666 N66 :6 ...vnd :hnd «scan... 20
...... z... 8... 3... 3.... 8.... 55.... 2.... 5...... ...8... .mo
.3... 2... 8... .......... .....3 2.... ...... X... a... 62. E
...8... .....e... 5...... an... 2... .....3 2... an... 62. ......
.2... .3... .3... z... 52.... 2... 3... $2. 3... ...:
......2. an... 8.... .5... ... ... a... $2. 8.28...
..R... z .... ...... 8.... on... 62. 23>
R... .2... 8.... ....n... .m. E
and $6.: :36 mm.—
..N... ...n... a. 3m ...:
2!..an Am. 85:05
83...... ”8.3.8.3
an E ...... 3m 8. $58... ...»; E ...... 3m ...: 82...... 23>
895m .32 gm

 

.32 5 H2 .3550 conga—«M an 38:33
.85....5 «Nausea... .... .... .29. :8... 0.5.2.8.. ...... :8. .3... 3.53.03... ...»..o... ...... 258...... 62. .89.... ...... .m.
...... 3...... 3.38... ...... ...... .... ...... ... .35.... ...... ..Bm ...... 2...... ..8. ...: ..aeo... ...... =82... 82......8 .. 2......

53

Table 9. Correlations between yield under stress (S), yield under non-stress (NS),
tolerance and mean productivity for two bean populations grown over four years (1990-
1993) in ﬁve environments in Michigan and Mexico.

 

 

 

Yield (S) Yield (NS) Tolerance?

Sierra/AC1028 Population

Yield (NS) 072“"

Tolerance 0.12 ~0.61*“

Mean Productivity’ 0.91m 0.94m -030"
Sierra/Lef-ZRB Population

Yield (NS) 0.72“" .

Tolerance 0.40" -0.64"'“"'

Mean Productivity 082*" 0.88"” -O.19*

 

‘P < .05; " P <.01; *" P < .001, “" P < .0001
f Tolerance = Yd -Yp
1 Mean productivity = (Yd + Yp)/2

54

stress, 100 seed weight under non-stress, number of pods per plant under non-stress, HI
under non-stress and GM. Biomass under non-stress was signiﬁcantly correlated with 100
seed weight under non-stress, number of pods per plant under non-stress and GM with a
positive relationship. One-hundred seed weight grown under non-stress conditions was
signiﬁcantly and positively correlated with number of pods per plant under non-stress and
GM. Number of pods per plant measured under non-stress was positively correlated with
HI under non-stress and GM. HI under non-stress was positively and signiﬁcantly
correlated with D81 and GM.

In both populations, Yd was positively correlated to Yp and the arithmetic mean
(Table 9). Yd was positively correlated with tolerance only in S/L. Yp was negatively
associated with drought tolerance and positively associated with arithmetic mean whereas

drought tolerance was negatively associated with arithmetic mean in both populations.

Marker Analysis

Six hundred markers were screened against the parents of both populations. Fifty
percent of these were polymorphic. F-test results are shown in Tables A. 11 - A. 16. S/L
had 49, 40 and 52 markers signiﬁcant for Yd, Yp and GM traits, respectively. Fewer
RAPDs were signiﬁcantly associated in S/A, with 30, 36 and 32 RAPDs identiﬁed as
being associated with Yd, Yp and GM, respectively. No single RAPD was identiﬁed to be
consistently associated with Yp, Yd nor GM over all ﬁve locations. Only about 3% of the
RAPDs were found to be associated with at least one of these values in three

environments and none in four.

 

55

Mapmaker analysis of the seventy RAPDs in each population revealed nine linkage
groups in S/A and 10 linkage groups in S/L (Figures 5 and 6). Linkage group one in S/A
was the largest linkage group with 11 markers (Fig. 5). Adjacent markers in all linkage
groups were not more than 10 cM apart based on a LOD score of not less than 4.0.

All RAPDs from each linkage group were analyzed together in a multiple
regression analysis against Yp, Yd and GM in all environments separately, and combined.
Signiﬁcant associations are shown in Tables 10 and 1 1. Linkage group nine from S/A was
signiﬁcant for at least one of the parameters, Yd, Yp and GM, in three out of the ﬁve
environments tested and in the combined analysis. R2 values for this group ranged ﬁom
0.08 to 0.14. Overall, from the combined linkage analysis, group nine was associated with
yield potential, explaining 14% of the variation, and geometric mean, explaining 8% of the
variation. The remaining groups were not as consistently associated over locations
(Table 10).

The multiple regression analysis performed on the linkage groups from S/L
generated many more signiﬁcant associations with the parameters Yp, Yd and GM, than
with S/A and explained a greater proportion of the variation for yield and biomass (Table
11). Linkage groups one and two were consistently associated over environments for at
least one of the parameters, Yd, Yp and GM and associated with all three in the combined
analysis. Linkage group two, however, was not signiﬁcantly associated with Yd, Yp, nor
GM in Madero, Dgo. in 1993. Additionally, R2 values were not as high for linkage group
two as they were for group one. Linkage group one had 1?.2 values ranging from 0.17 to

0.62 whereas linkage group two ranged ﬁom 0.14 to 0.27. The highest R2 value (0.62)

 

 

 

 

 

 

 

 

 

 

56
1T 2 4 7
— 0906900: T060640. '— OVOIuo 0001""
"I 1° so
10.0
—- 0201 m
— 0F06
7.0 18 970 02031010
— OACOJs'n 0611500 — 0416850
/ 5.2
2.0 0.4818600 b-
9n '3 _\< 01.12420 01031130 8
2.: 0119340
0A”5u
0103330 r— 0005
‘ g 620
2.0 u 5 1.0
10 —' 01105”
H 0110950 01’0”“ 3., one“. 011101000
t— oan-m 0010550
3.5 3 5.1 9
_ OABI4450
0A3]
— ones ,. 0T18550 6 ‘ "5'
3.5 .
to 0m“.
10 01‘le
—‘ 0118“” 690
—" 011115“ OH!”
5.: rm 6
F— 0w
9.0 —. 0103870 l.”
l 00021010 ‘4
u —/ _/ 0Rl61 180 ’-°
—\ 0.1 ‘_\
| 02045.0 ,5 01102760
7.0 h.
I 0114770 _. 0N09860
0H08490
TUnlinked m are 0A09e, OAlGa, 0A1 83, 0A1 8b, OFOI, OFlO, 0004, 0008a, 0008b, 0009, OHOIa,
0H0“), Ole, 0H1“, OHltb. OLO'I, OMOS, ON03, 0Q06a, OTOl, 01'16, 0003., OU03b, 0010a, 0W20e,
0W20b, 0X01, 0X03.
tRAPD: corresponding to Sim gnotype are in bold 6.1:. 01.18;...) when-I RAPD- oorreepond'mg to AC1028
W m in italics (Le. 0N09..).

 

 

Figure 5. Linkage groups generated by MAPMAKER using seventy RAPD markers for
the Sierra/AC1028 population. No marker is more than 10 cM away from its adjacent
marker based on a LCD 2 4.0

 

57

 

 

 

 

 

1* 2 3 s 8
—0A045601 l-ouosa "oamqs. a ”—0110”... 0604330
9“ 6.9 6.2 ‘ l.l _/0m5440 6.2
7.7
_0C15610
55 0A0)!
mo 4.. __ mucus... 9
35 _0X18980 u h'Olll)9_t,o 6 a [0120“.
01mm 0AG0291. 04309820 02031010
8.5 4
razor... "5
3.4
~0m75“ u —0AC15570 0.4602700 10
2.2 — 01001020 7 u 0mm
—0ABI8600 1.5 470
6.2 6.6 [0101131. 01.1119“
— 0AC02
700 OAEMm.
lUnlinked markers are 0A09, 0A16a, 0A16b, OAEOI, OAElOb, OAGOl, ocro, 01204, OFOSa,
OF06, OFIO, 0602, 0604, 0605a, 0608, 0608b, 0610, 0612, OHOla OHOlb, 0H03a, 01-106,
0H12a, 0H12b, OI-Ilc, 0L18, 0N09, OT02, OWO9a,-OW09b, 0X01, 0Y1 lb, 0201a, OZOlb.
1RAPDs corresponding to Sierra genotype are in bold (i.e. Olemo) whereas RAPDs corresponding to
Lef-ZRB genotype are in italics (i.e. 0A C027oo).

 

 

Figure 6. Linkage groups generated by MAPMAKER using seventy RAPD markers for
the Sierra/Lef-ZRB population No marker is more than 10 cM away from the adjacent
marker based on a LCD 2 4.0.

58

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8... :5 Nu... ..m
.86 egd eegﬁ m 20
x
u
... er
:5 2... 8... 2... «m ......2
eomnd ...anﬁ e35 :vad ..— v> 33
S... ....
9:5. "— 2O
2... ....
......n m 8»
z... 8... ....
:23 .92 ... 3 N8.
2... . c3 ....
:nuﬁ. canﬁ m 20
S... a... ....
Quad eons; ..— n>
...
m 3 .2:
8... ....
32“ m 20
....
m a»
86 5.x
.8.“ .— § 82
a «85 b .85 m .85 o .85 . x85 ... .85 .N .85
35:29.35

=d .83 3:388 28 .8 98 9am _ $3: 3:08:85... 3:232: oz.“ we gone Set 3% maﬁa “co—Eaton 803 8&3 9:253:
0568:. 9.3 825 :38» noun—anon wag U<>E£m 05 com n29 :8:— oEoEoom 05 v5 .3er 9.23-5: 825 203 AFC 23.5.6
8...... 2% a a... ... BE... .3 .283 case: .853 3.5.2. .32.... Ease... cause as. E 3.2... .§s&.m .2 2.5

59

082.853 80:82 2,: an 5.6 2.255 322.50 - 5:80.
5:82 38:8..N .2250 - .uaN.
8:32 omeﬂan .9632 - .32.
.8258... .35.... 8 a... . ...... . 5.85.
8.825.... ... 8.8580 n ....

:5. V m ...: 28V .— .. mnc. V m...

 

 

 

 

 

 

 

8... N. ... ...
obﬂn send "— 20
...... ....
...... m 3
2 ... ....
..3 a E. ......8
.. .o ....
... .n ... 2o
5
N
.. er
... .o a... ... .8...
:85 ..an .— :> 2.6—
. ..85 h ..85 e ..85 . n ..85 .. u.85 n ..85 .N ..85

 

85.8 .... 2......

60

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8... N...
...... . .20
...... . «.... ...... ...
...... .8." . :9... . ...r
8... N... .....2
...... . ...» m8.
...... ...... ...... N...
92... ..va onod .— 20
...
«
. ....
...... n... N...
59§nnd :onn.v .— e> Nag—
..2. ...... a... .2. 2... ...... z... N... ...
Gigi oooaud :33.” 0:36 2..de 3.an can." .38.: m 20
...... ...... ...... 2... .... ...... 8... N...
.36 {:6 36:6 :36 §8N .nvN ...:hvi m a>
...... 2... 2... 2... ...... .... a... N...
.5395 9:3“: 3:32: :32: ..nhd 8:26 9:36 a 6% ga—
...... ...... a...
:86 9!..an m 20
...... ...... N...
:36 35:5.” .— 9?
.... .... 2... z... ...... ....
962M unwd 3.2.6 93; 353...: m a? 8a—
.. ..85 .. ..85 .. ..85 .. .85 e ..85 n .85 . ..85 n ..85 N .85 ...85
3.5.5.0225

.... .26 3:353 28 .2 v.3 @3782. 35:58:60 o>: 12.3%... .... .58 80... 32. 9...: 35.25.. 0.03 .8252 .3858:
0.3.5:. 93. .25.. .56.» 5:23.... mm~¢§t2m 2: .2 A29 :8... 052.com 2: ...... Aer. ....o..?:2. .02... 22.. Ago .53...
3...... 2.... ... ...... ... 8.5.0.. ...... 9.8... 0.5.... 883.8 858...... .355 8.3.052 2&3... 52. 5. 38... 285.5... ... 0......

62
was observed in the combined analysis for linkage group one associated with Yp. Overall,
from the combined linkage analysis, associations with linkage group one explained 54% of

the variation of Yd, 62% of the variation of Yp, and 56% of the variation of GM.

Marker Assisted Selection

RAPDs from linkage group nine (0H19590 and OABlSaso; Fig. 5) combined with
unlinked markers, OFOlm and CHIS-no in S/A were chosen as indirect selection criteria
to improve performance under drought. RAPD OH19590 is associated with genotypes
performing below average whereas RAPDs OABI8650, 01701520, and CHIS-no are all
associated with genotypes performing above average. Therefore, RILs which lacked the
(01119590 band but possessed the 0AB18550, 01701520 and 0H18a bands were chosen as
potentially above average genotypes (PAA). Those RILs with the 0H19690 band present
but lacking the OABISGSO, OFOlm and 0H18a bands were designated as potentially
below average genotypes (PBA). Ten RILs plus Sierra were identiﬁed in the PAA group,
whereas nine RILs plus AC 1028 were classiﬁed as in the PBA group. Both markers from
linkage group nine were used to select genotypes because, when analyzed alone, the F-test
revealed signiﬁcant associations between OH19690 and performance only once in 1992
where it was signiﬁcantly associated with Yd (Table 12). OAB18550 was also signiﬁcantly
associated with performance in one analysis (1991 GM). When analyzed together, in a
multiple regression analysis, signiﬁcant associations occur 8 times out of a possible 18

diﬁ‘erent analyses (Table 12).

63

The 21 RILs identiﬁed by their RAPD genotype were planted in a split-plot design
with two treatments (stress and non-stress). The drought intensity index for this growing
season was 0.76, higher than any previous experiment. From the split-plot analysis of
variance, genotypic variation was signiﬁcant for yield, 100 seed weight, number of pods
per plant, and days to maturity, but not biomass (Table A. 17). Treatments diﬂ‘ered
signiﬁcantly for yield, biomass, 100 seed weight and number of pods per plant but not for
days to maturity. Signiﬁcant genotype X stress interactions occurred for number of pods
per plant, 100 seed weight and yield. Orthogonal comparisons with 1 degree of freedom
were made between the PAA and PBA groups (Table A. 17). Comparisons were
signiﬁcant for yield, 100 seed weight and number of pods per plant. Days to maturity was
not analyzed because no signiﬁcant difference between treatments was observed. For

signiﬁcant contrasts, the difference between the mean of the PAA and PBA groups was 11

g m.2 for yield, 0.5 g for 100 seed weight and 1.2 pods per plant. The PAA group had an
average of 1.2 more pods per plant. Pods per plant was the most severely affected yield
component with a reduction in the grand mean of 34% in the stress treatment.

When treatments were analyzed individually as a randomized complete block
design with four replications, signiﬁcant genotypic variation was observed for all traits
except Yd, biomass under both treatments and number of pods per plant under stress
(Table 13). Orthogonal comparisons between groups were signiﬁcant in all but biomass
and number of pods per plant in the stressed treatment. For yield, the means for PAA
exceeded PBA by 10 and 33 g rn'2 for stress and non—stress treatments, respectively (Table

13). The mean yield of PAA exceeded the grand mean by 3.5 and 15.7 g rn'2 for stress and

Table 12. A comparison between the number of signiﬁcant F-tests from multiple
regression analyses between all the RAPD markers in linkage group 9* for the
Sierra/AC1028 population and yield under drought (Yd), yield under non-stress (Y p) and
geometric mean (GM) versus the one-way analyses of variance performed on each
individual unlinked marker found in linkage group 9. Data is shown for each of the ﬁve
environments (1990-1993), separately, and the combined analysis over locations.

 

1990

1991

1992

1993 Madero, Dgo.
1993 Calera, Zac.
Combined

1990

1991

1992

1993 Madero, Dgo.
1993 Calera, Zac.
Combined

1990

1991

1992

1993 Madero, Dgo.
1993 Calera, Zac.
Combined

Yd Yp GM

Linkage group 9 (01119590 & 0A318550)
3.63"I 3.29“

4.54" 5.23"
5.99" 5.81"

6.39" * 3 .27“
Indivrdual 0H19590

11.51"

Indivrdual GAB 18550

3.92“

I'17' < .05; " P <.01; ‘" P < .001, "" P < .0001

TAsdeﬁnedinﬁgure 5.

65

Table 13. Grand mean and means for each genotypic group selected based on marker
genotype, LSD and CV for yield, biomass, 100 seed weight, and number of pods per plant
resulting ﬁ'om the analysis of variance for stress and non-stress treatments, individually, in
Kalamazoo County . Eleven genotypes ﬁom the Sierra/AC1028 population were selected
as potentially above average (PAA) and ten as potentially below average (PBA)
performance based on marker genotypes.

 

 

 

 

 

 

 

 

 

Stress Non-Sm“
...—weld (g m'z) ......

Grand Mean 47.2 1896""
Mean (PAA) 50.7 205.3
Mean (PBA) 41.1 172.3

Diﬁ'erence 9.6"“ 33.0""

LSD (a < .05) 27.0 ' 51.3

CV (%) 41.5 19.1

----Biomass (g m'z) ----

Grand Mean 104.9 ' 3243""
Mean (PAA) 106.0 337.7
Mean (PBA) 103.7 . 309.6

Difference 2.3" 28.1“
LSD (or < .05) 49.2 80.7
CV (%) 33.2 ' 17.6

----100 Seed Weight (g) ----

Grand Mean 28.8 30.4"
Mean (PAA) 29.3 31.6
Mean (PBA) 27.2 29.1

Difference 2.1"" 2.5""
LSD (at < .05) 2.0 2.0
CV (%) 2.7 4.6

“Number of Pods per Plant-—

Grand Mean 3.6 105*"
Mean (PAA) 3.8 11.4
Mean (PBA) 3.5 9.1

Diﬁ‘erence 0.3" 2.3”
LSD (or < .05) 4.5 3.6
CV (%) 44.2 26.7

 

‘P < .05; " P <.01; "“ P < .001, “"“ P < .0001

66

non-stress treatments, respectively. The difference between mean yields for PAA and PBA
did not generally exceed the LSD (0.05). LSD, however, for the stress treatment was half
as much as the grand mean for yield and biomass. The LSD for number of pods per plant
exceeded the grand mean. CV values for yield, biomass and number of pods per plant
under stress were 41.5%, 33.2%, and 44.2%, respectively, whereas CV values under non-
stress were 19.1%, 17.6%, and 26.7%, respectively. Although CV values under the non-
stress treatment are more typical, the severe drought intensity observed in the stress
treatment signiﬁcantly reduced the grand mean of yield, biomass and pods per plant
thereby inﬂating the CV. This was not evident for 100 seed weight because the difference
between grand means of the stress and non-stress experiments was under 2 g.

Orthogonal comparisons were made between the same 21 genotypes using data
from 1994 Madero, Dgo. and Calera, Zac. (Table 14). The means of the genotypic groups
(PAA and PBA) did not differ signiﬁcantly for yield nor biomass using the data from
Madero, Dgo. 100 seed weight was not available for analysis. There was no signiﬁcant
genotypic variation in the non-stress treatment for yield nor biomass (Table A. 2). LSDs
were high, approaching half the grand mean for stressed and non-stressed plants.
Coeﬁcients of variation were higher than expected ranging ﬁ'om 21.4 to 23.6% although
they were consistent between treatments. Mean yields between the genotypic groups
(PAA vs. PBA), although non signiﬁcant, differed by 27.2 and 18.7 g rn'2 for stress and
non-stress treatments respectively. PAA exceeded the grand mean for yield under stress by

2.7 g In2 (Table 14).

Table 14. Grand mean and means

stress and non-stress treatments
Zacatecas and combined over both

67

for each genotypic group selected based on marker
genotype, LSD and CV for yield and biomass resulting from the analysis of variance for
in 1994 conducted at Madero, Durango, Calera,
locations. Eleven genotypes ﬁ'om the Sierra/AC1028
population were selected as potentially above average (PAA) and ten as potentially below

average (PBA) performance based on marker genotypes.

 

 

 

 

 

 

 

 

 

Stress Non-stress
Madero, Durango, 1994
———Yreld (g m") ----

Grand Mean 134.7 ' 192.8
Mean (PAA) 137.4 187.0
Mean (PBA) 110.2 168.3
Difference 27.2" 18.7"

LSD (at < .05) 63.5 93.7

CV (%) 23.6 23.3

—-—Biomass (g n?)———

Grand Mean 245.7 - 323.6
Mean (PAA) 249.2 310.2
Mean (PBA) 217.0 292.7
Difference 32.2“ 17.5”

LSD (at < .05) 105.1 149.7

CV (%) 21.4 23.2

Calera, Zacatecas, 1994
...—mew (g m") ...—...

Grand Mean 120.8 240.2
Mean (PAA) 118.4 233.4
Mean (PBA) 109.4 206.6
Difference 9.0“ 16.8"I

LSD (at < .05) 40.7 60.0

CV (%) 16.9 12.5

Combined Analysis over Locations
—---Yie1d (g m") ...——

Grand Mean 127.8 216.5”
Mean (PAA) 127.7 210.1
Mean (PBA) 110.8 187.5
Difference 16.9“ 22.6"

LSD (at < .05) 38.4 54.9

CV (%) 21.6 18.2

 

T Biomass was unavailable at this location

68

The same trend was observed for comparisons between genotypic groups
performed on data from 1994 Calera, Zacatecas (Table 14). There was no signiﬁcant
difference between genotypic groups. However, mean yield for PAA exceeded PBA by
9.0 and 26.8 g m'2 for stress and non-stress treatments, respectively. PAA group did not
exceed the grand mean in either treatment. Data for biomass and 100 seed weight were
unavailable. CV’s were 16.9 and 12.5% for the stress and non-stress treatments,
respectively. This is considerably lower than previous experiments while LSDs were less
than one third the grand mean.

Corresponding treatments ﬁom each location were combined and analyzed as a
one-factor randomized complete block design combined over locations for yield (Table
14). No signiﬁcant difference between genotypic groups was observed and PAA did not
exceed the grand mean in either treatment. Mean yields diﬂ‘ered between the groups,
however, by 16.9 and 22.6 g m'2 for stress and non-stress treatments respectively. CV was
18.3% which is as expected and LSDs were approximately one quarter of the grand mean.

Means ﬁ'om the individual lattice analyses ﬁom both locations and treatments were
combined and analyzed as a split-plot design with locations as replications (Table A 18).
Although there was a non-signiﬁcant treatment effect, there was signiﬁcant genetic
variation. Contrasts between genotypic groups were not signiﬁcant and the mean for PAA
group did not exceed the grand mean. Mean yields between the two groups differed by
19.8 g m'z.

RAPDs from linkage group one in S/L (Fig 6.) were used to select 34 RILs plus

Sierra as potentially above average performing (PAA) and 27 RILs plus Lef-2RB as

69

potentially below average genotypes (PBA). Linkage group one contained ﬁve markers:
0A08m, 0Z08m, OA04550, OX1 1m and 0X18m. RAPD markers OA04560, OX1 1m,
and 0X18,“ were all associated with below average performance whereas RAPD markers
OZOS-m and 0A08m were associated with above average performance. RILs which
lacked the OA04560, OX1 1m, and OX18930 bands but had the 0Z08m and 0A08m bands
were included in the PAA group. RILs with the opposite banding pattern were selected as
the PBA group. All the RAPDs from linkage group one were - used in the same manner
that RAPDs in linkage group nine of S/A were employed. When analyzed independently,
0X18m and OZ08150 were signiﬁcantly associated in only six of the possible 18 different
analyses (Table 15). 0A08m, OA04560, and 0X11m were associated in only seven, four,
and two of the total 18 diﬂ‘erent analyses, respectively. The multiple regression analysis
generated signiﬁcant associations in 15 out of a total 18 analyses performed and explained
up to 62% of the variation.

Orthogonal comparisons were made between groups using 1994 data ﬁom
Madero, Dgo. (Table 16). The means between PAA and PBA did not differ signiﬁcantly
for biomass nor yield. 100 seed weight, however, did show a signiﬁcant difference for
both stress and non-stress environments at P < .0001. Genotypic variation was signiﬁcant
for yield, biomass, and 100 seed weight in both stress and non-stress treatments.
Differences between PAA and PBA for yield and biomass although statistically non-
signiﬁcant were 1.5 and 2.1 g m’2 respectively for the stress treatment and 4.0 and 4.9 g

m'2 for the non-stress treatment. 100 seed weight diﬂ‘ered by 2.0 and 2.4 g for stress and

70

Table 15. A comparison between the number of signiﬁcant F -tests ﬁom multiple
regression analyses between all RAPD markers in linkage group IT for S/L and yield under
drought (Yd), yield under non-stress (Y p) and geometric mean (GM) versus the one-way
analyses of variance performed on each individually unlinked marker. Data is shown for
each of the ﬁve environments (1990-1993), separately, and the combined analysis over
locations.

 

 

 

 

 

 

Yd Yp GM
Linkage group 1 (0A08m, 0Z08750, 0A04560, OX1 15.0, OX189.0)
1990 6.34“" 8.76”" 7.45""
1991 4.19‘“ 4.47"" 6.09‘“
1992 4.32‘“ 2.62‘
1993 Madero, Dgo. 2.85""I
1993 Calera, Zac. 8.67‘" 7.89””I 9.87“”
Combined 1525"“ 20.97"" 16.56‘"
Individual 0A08m
1990
1991 6.39“ 3.54‘ 6.76"
1992 3.48‘
1993 Madero, Dgo.
1993 Calera, Zac. 3.57" 3.42“
Combined 4.02‘I
Individual 0Z08750
1990
1991
1992
1993 Madero, Dgo. ‘
1993 Calera, Zac. 19.16”" 16.13““ 22.83”“
Combined 4.43" 7.26" 7.20“
Individual 0AM,“
1990
1991 10.58“ 6.11‘
1992 4.40'
1993 Madero, Dgo. 4.29"
1993 Calera, Zac.
Combined
Individual OX1 15.0
1990
1991 13.85“" 8.98“
1992

1993 Madero, Dgo.

1993 Calera, Zac.
Combined

Table 15. Continued.

1990

1991

1992

1993 Madero, Dgo.
1993 Calera, Zac.
Combined

 

71

Individual 0X18m

18.07”" 4.68‘ 13.72“”
6.57 4.07
4.87“

‘P < .05; “ P <.Ol; “" P < .001, “W" P < .0001

TAsdeﬁnedinFigurc6.

72

Table 16. Grand mean and means for each genotypic group selected based on marker
genotype, LSD and CV for yield biomass, and 100 seed weight resulting from the
combined analysis of variance over locations for stress and non-stress treatments,
individually, conducted in 1994 at Madero, Durango, Calera, Zacatecas and combined
over both locations. Thirty-six genotypes from the Sierra/Lef-ZRB population were
selected as potentially above average (PAA) and 28 as potentially below average (PBA)
performance based on marker genotypes.

 

 

 

 

 

 

 

 

 

 

 

 

Stress Non-stress
Madero, Durango 1994
Yield (g m") ----

Grand Mean 143.4 194.0
Mean (PAA) 146.3 200.1
Mean (PBA) 144.8 196.1
Difference 1.5“ 3.9"

LSD (or < .05) 66.2 81.1

CV (%) 23.2 21.0

Biomass (g m’z) ---

Grand Mean 246.0 321.1
Mean (PAA) 247.9 327.6
Mean (PBA) 245.8 322.7
Diﬂ'erence 2.1” 4.9III

LSD (at < .05) 95.6 130.9

CV (%) 19.6 20.5

-—--100 Seed Weight (g) ----

Grand Mean 24.8 28.1
Mean (PAA) 25.5 _ 28.9
Mean (PBA) 23.5 26.5
Diﬂ'crencc 2.04M!!! 2.4tttt

LSD (at < .05) 3.6 3.7

CV (%) 7.4 6.7

Calera, Zacatecas, 1994
Yield' (g m’z) ---

Grand Mean 91.2 157.2
Mean (PAA) 111.5 188.1
Mean (PBA) 82.6 144.1
Difference 28.9““ 44.0"“

LSD (a < .05) 61.3 61.1

CV (%) 33.8 19.6

 

 

73

Table 16. Continued

 

 

 

 

Combined over Locations
Yield (g m”)

GrandMean 117.3 175.6“
Mean (PAA) 130.3 190.3
Mean (PBA) 115.1 172.9
Diﬂ‘erence 152* 17.4.

LSD (at < .05) 50.3 51.9

CV 1%) 30.4 22.0

 

‘P < .05; n p <01; m P < .001, mm P < .0001
* Biomass and 100 seed weight were unavailable for this location.

74
non-stress treatments, respectively (Table 16). LSDs were, generally, a little less than half
the grand mean for yield and biomass under stressed and non-stressed treatments. The CV
for yield and biomass averaged around 21% which is moderately high.

The results ﬁ'om 1994 data from Calera, Zac., demonstrated a different trend
(Table 16). Although biomass and 100 seed weight data were unavailable, genotypic
group means differed signiﬁcantly for yield in both stress and non-stress environments.
Means diﬁ‘ered by 29 and 44 g m'2 for stress and non-stress treatments, respectively. PAA
out-yielded the grand mean by 11 g m'2 in the stress environment and by 31 g m’2 in the
non-stress environment. Genotypic variance was signiﬁcant in both treatments. For the
stress treatment, the LSD was high, averaging over two thirds the grand mean. The CV
for the non-stress treatment was 19.6% whereas the stress treatment was 33.8%. This is a
considerable increase in the cv which can be accounted for by the reduction in mean yield
of the stress treatment thereby inﬂating the CV for this treatment.

The results from the combined one-factor randomized complete block design over
locations for data from 1994 were similar to ' the results ﬁ'om 1994 Calera, Zac. data
(Table 16). Genetic variation and orthogonal contrasts (PAA vs. PBA) were signiﬁcant in
both treatments. In the stress treatment, mean yields differed between genotypic groups by
15.2 g m'2 and PAA exceeded the grand mean by 13.0 g m‘z. Results ﬁ'om the non-stress
treatment analysis showed a diﬂ‘erence between genotypic group means of 17.4 g m‘2 and
PAA exceeded the grand mean by 14.7 g m".

Results from the overall split-plot analysis for 1994 conﬁrmed an insigniﬁcant

treatment eﬂ‘ect and treatment X genotype interaction (Table A19). Genotypic variation,

 

 

75
however, was signiﬁcant as well as the orthogonal comparison. Means ﬁ'om the genotypic
groups diﬁ‘ered by 19.5 g m'z. PAA out-yielded the grand mean by 15.6 g m’z. LSD was

one-third the grand mean and coefﬁcient of variation was 26.7%.

Marker Assisted Selection Compared to Conventional Selection

Since signiﬁcant diﬁ‘erences between PAA and PBA were identiﬁed, yield data was
used to compare marker assisted selection to conventional selection. Based on marker
genotype we identiﬁed 35 potentially above average and 28 potentially below average
performance RILs. Using the same selection intensity, above (AA) and below (BA)
average RILs were identiﬁed based on Yd, Yp, and GM combined over ﬁve locations
ﬁom 1990-1993. The 1994 yield data from Calera and Madero for these RILs was
evaluated to test their performance and to compare MAS. The AA group of 35 RILs
selected based on Yd did not yield signiﬁcantly different from the BA group of 28 RILs
(Table 17). The diﬂ‘erences between AA and BA were 9.1, 8.5, 8.9 g rn'2 for yield under
stress, non-stress, and combined over treatments, respectively. AA exceeded the grand
mean in the stress treatment but only by 0.2 g m'z. In the non-stress and combined
treatment analyses, AA yielded 1.6 and 0.8 g m'2 below the experimental mean. AA and
BA-selected lines based on Yp did not diﬂ‘er signiﬁcantly for stress, non-stress, nor
combined treatment analyses. BA out-yielded AA in the non-stress treatment analysis by
0.2 whereas AA out-yielded BA by 2.0 and 0.9 g m'2 for stress and combined treatments
analyses, respectively. AA and BA selected based on GM also did not diﬁ‘er signiﬁcantly

for any treatment analyses. The mean of AA exceeded BA by 6.6, 5.6, and 6.1 g m'z, for

 

 

76

Table 17. Comparisons between marker assisted selection and conventional selection.
Using data over 4 years (1990-1993) and ﬁve locations in Michigan and Mexico, 3
genotypic groups of 35 RILs were selected ﬁom the Sierra/Lef-ZRB population based on
above average performance (AA) for yield under drought (Yd), yield potential (Y p), and
geometric mean (mean). Similar selection was used to identify 3 groups selected for below
average performance (BA). The mean yields from these groups calculated ﬁom data
collected ﬁ'om 1994 Madero Dgo. and Calera Zac. experiments were compared to groups
identiﬁed by marker genotype.

 

 

 

 

 

Selection Based on:
Markers Yd Yp GM
Combined over Locations - Stress Treatment
Experimental Mean 117.3 117.3 117.3 117.3
Mean AA 130.3 117.5 116.9 116.5
Mean BA 115.1 108.4 117.1 109.9
Difference 15.2‘ 9.1 0.2 6.6
Combined over Locations - Non-Stress Treatment
Experimental Mean 175.6 175.6 175.6 175.6
Mean AA 190.3 174.0 178.5 173.3
Mean BA 172.9 . 165.5 176.5 167.7
Difference 17.4“ 8.5 2.0 5.6
Combined over Locations - Combined Treatments
Experimental Mean 146.6 146.6 146.6 146.6
Mean AA 162.1 145.8 147.7 144.9
Mean BA 142.6 136.9 146.8 138.8
Difference 19.5" 8.9 0.9 6.1

 

‘Signiﬁcant differences between AA and BA groups at P < 0.05.

77
stress, non-stress and combined treatment analyses, respectively. In none of the treatment
analyses did AA yield above the experimental mean.

For all cases, the difference between mean yields of AA and BA were negligible
and AA yielded, in general, close to the experimental mean. BA selected based on Yd and
GM, however, yielded, on average 9.6 and 7.7 g m'2 less than the experimental means for
all treatments. The experimental means for PAA and PBA, selected based on marker
genotypes, on the other hand, differed signiﬁcantly for all treatment analyses. The mean of
PAA out-yielded PBA by 15.2, 17.4, and 19.5 g m’2 for the stress, non-stress and
combined treatment analyses respectively. Although the mean of PBA was not more than
4.0 g 111'2 less than the experimental mean, PAA exceeded the experimental mean by 13.0,
14.7 and 15.5 g rn‘2 for the stress, non-stress and combined treatment analyses,
respectively.

The seventy RAPDs used to screen each population for associations were not the
same markers for each population. Although the markers in linkage group one (Fig. 6)
from S/L showed polymorphisms between Sierra and AC1028, they were not screened
against S/A. Therefore, these markers were not identiﬁed as a linkage group associated
with drought performance in S/A. It is highly probable that these ﬁve RAPDs would show
similar gains in yield if used for MAS performed on S/A. The next approach for this

research will be to identify the effectiveness of these markers in S/A and determine

population speciﬁcity.

DISCUSSION

It was our objective to develop a practical breeding method whereby DNA-based
markers could be exploited to improve a quantitative trait associated with performance of
common bean under drought. By using previously developed populations we were able to
identify markers associated with drought tolerance based on ﬁve years of testing. The
application of these markers to other genetic populations is beyond the scope of this
research. However, markers associated with increased yield under drought were identiﬁed.
Since these markers originated from a single parent, Sierra, which is a Michigan pinto
bean cultivar used widely in crosses for variety development, these markers could be used
as a selection criteria in other populations developed from crosses with Sierra as a parent.
The exact location of these markers on the genetic map was not considered. We
postulated that markers could be identiﬁed that were associated with drought tolerance
and then used as a tool to more efﬁciently select genotypes with improved performance.
The process we have outlined is speciﬁc to a diploid, self-pollinating crop.

Tolerance to drought, in agronomic terms, is a yield-related trait and, therefore,
necessarily a quantitative trait. Improved performance under water stress is determined by
a number of genes but is also strongly inﬂuenced by the environment which complicates
breeding for drought tolerance. To improve a quantitative trait like drought tolerance in

common bean, a breeder will initially cross drought tolerant genotypes. Individual F2

78

 

 

79
plants will be advanced by single seed descent until the F4 generation where homozygous
seed is bulked to create recombinant inbred lines. Determination of drought tolerance does
not begin until the F5 or F6 generation when suﬂicient seed is available for replicated
testing. This process is especially expensive since most of the genotypes, which have been
maintained for ﬁve to eight years, will be discarded based on performance. By identifying
markers associated with drought tolerance, genotypes could be evaluated earlier in the F3
or F4 generations using selection based on marker genotype. Because the markers are
DNA-based, a single plant can be genotyped thereby eliminating the need for large
amounts of seed. This will not replace ﬁeld testing but will allow the breeder to make
selections in earlier generations so that inferior material is not maintained. Field testing
restricted to only those potentially superior-performing individuals would then be

conducted.

Field Study

Due to the strong environmental inﬂuence on drought tolerance, yield data was
collected over ﬁve years (1990-1994) and seven locations to ensure that adequate drought
environments were tested. The ﬁrst four years (ﬁve locations) were used to develop
associated markers whereas the ﬁnal year (two locations) was used to determine the
eﬂ‘ectiveness of these markers as a selection criteria for drought tolerance. Since a large
data base was collected over ﬁve years several conventional genetic analyses were
performed. Sierra, the common parent in both populations, was used for population

development because it had demonstrated good performance under drought stress in two

80

previous studies (Acosta, 1988; Ramirez, 1992). AC1028 was an 0118ng of a cross
developed by Acosta (1988) and identiﬁed as having potential for drought tolerance.
However, Ramirez (1992) later classiﬁed it as a poor performer under moisture stress
based on data collected over a two year period from experiments conducted in Michigan.
Acosta (1988) identiﬁed Lef-2RB as a Mexican breeding line having superior
performance under drought which was later supported by Ramirez (1992). Based on
percent reduction in yield, Sierra was a better performer than Lef-2RB whereas Lef-2RB
was better than AC1028. Sierra and Lef-2RB were classiﬁed as good and AC1028 as poor
performers under drought (Ramirez, 1992). In the current study, Sierra was used in both
populations and demonstrated a reduction in yield of 28% in S/A and 50% in S/L
experiments (Tables A7 and A9). AC1028 yielded 24% less and Lef-2RB 31% less under
stress. Lef-ZRB had a higher percent reduction under stress than AC 1028 and yielded less
under non-stress but more under stress conditions than AC 1028. Sierra appears to be
unadapted to the temperature extremes, arid climate, and elevation of the Mexican
highlands which could explain its erratic behavior. This illustrates the limitations in
breeding for stress adaptation across locations and the inability to provide environments
where stress is consistent.

Comparing populations, S/A demonstrated a greater range of values for yield, and
biomass under both treatments than S/L (Tables A7-10; Figures 1-4). S/A had the lowest
and highest yielding genotypes between the two populations and had higher mean values
for yield and biomass under both treatments (Tables 1 and 2). This suggests that S/A had

greater variation for yield traits related to drought tolerance than S/L.

81

This is supported by heritability estimates for yield, biomass and 100 seed weight
for both stress and non-stress treatments (Table 3). Heritability estimates for yield and
biomass under stress and non-stress treatments were over three times greater in S/A than
S/L whereas h2 estimates for one hundred seed weight were similar between populations.
One-hundred seed weight is considered moderately heritable where h2 calculated for two
pinto/navy bean populations were 0.58 and 0.39 (Brothers and Kelly, 1993). White et al.
(1994) reported a heritability estimate for 100 seed weight as 0.72 for bean genotypes
grown in Durango. This is consistent with the estimates observed in this study (Table 3).
Ramirez (1992) reported h2 estimates over two years for yield based on ﬁve genotypes
which include AC1028, Lef-2RB, and Sierra parents grown under stress and non-stress
conditions, of 0.16 and 0.46, respectively. We report h2 estimates of 0.45 and 0.47 for 80
genotypes in S/A for yield under stress and non-stress, respectively (Table 3). Heritability
estimates for 98 genotypes in S/L were 0.11 for stress and 0.08 for non-stress (Table 3).
Scully et al. (1991) reported h2 estimates for yield, in common bean, which ranged from
0.01 to 0.29. However, White et al. (1994) observed 112 for seed yield ranging ﬁ'om 0.09
to 0.75 for rainfed treatments conducted in Mexico and Colombia. The wide range in
values emphasizes the difﬁculty in comparing h2 estimates across experiments and how
dependent h2 is on population type and size, germplasm studied, and independent sources
of variation. Heritability estimates are useful in breeding to estimate expected gains from
selection and determine which breeding approach is most appropriate (Ramirez, 1992).
Our data suggests that h’, for yield, is much higher in S/A than S/L. We would, therefore,

expect a lower gain ﬁ'om selection in S/L than S/A for yield. Calculations for the expected

82

gains per cycle of selection based on the calculated h2 support this point (Tables 4 and 5).
S/A shows a 8.4 fold advantage over S/L for yield under non-stress conditions and a 4.5
fold advantage for yield under stress. Consequently, conventional selection would be more
successﬁll for improving yield performance for both stress and non-stress in S/A than S/L.
In the case of S/L, where h2 for yield under stress and non-stress conditions are extremely
low and gains ﬁom selection are 3 .0 g M2 per cycle, it would be eﬁ‘ective to try alternative
forms of selection such as MAS. Lander and Botstein (1989) demonstrated that the
efﬁcacy of MAS is inversely proportional to the h2 of a given trait.

A reduction in heritability of a given yield-related trait is expected under stress
conditions (Rosielle and Hamblin, 1981). A similar trend was observed in S/A for yield
and biomass but was not observed in S/L (Table 3). Although the differences were small,
genetic variation for yield, biomass, and 100 seed weight were less under non-stress
conditions for S/L. Singh (1995), has reported a similar trend between h2 calculated for
populations grown under stress and non-stress treatments. In one population h2 was
greater under water stress whereas, in the other population, the opposite trend was
reported.

Rosielle and Haman (1981) analyzed a variety of correlations between stress and
non-stress conditions using simulated data. They deﬁned tolerance to stress as the
diﬁ‘erence between stress yield (Yd) and non-stress yield (Y p) and mean production as the
arithmetic mean between the two yields ((Yd + Yp)/2). They postulated that if the genetic
variation under stress is lower than non-stress (which is generally the case) selection for

tolerance will result in a reduction in mean yield in the, non-stress environments. If the

83,
genetic variance is less under non-stress and the correlation between Yd and Yp is high,
selection for tolerance will result in both an increase in Yd and Yp. The correlations
calculated between tolerance and Yp using our experimental data were -0.61 for S/A and
-0.64 for S/L indicating a strong negative relationship (Table 9). In S/A, genetic variation
was less under stress than non-stress conditions. Although the genetic variation in S/L was
less under non-stress conditions, the correlation between Yd and 'Yp was small resulting in
a negative correlation between tolerance and Yp. However, when breeding for
environments like the semi-arid highlands of Mexico where drought conditions are typical
but not consistent and where irrigation is not available, it is preferable to develop a
cultivar that yields consistently over all environments although it yields less than other
genotypes under non-stress conditions. Rosielle and Hamblin (1981) postulated that
selection for tolerance will increase Yd when genetic variance is lower under stress
conditions or when genetic variance is lower under non-stress provided that the
correlation between Yd and Yp is small. This trend was observed in our data since
calculated correlations between tolerance and Yd were positive for both populations
(Table 9). Selection for stress tolerance has been reported to decrease mean productivity
(Rosielle and Hamblin, 1981). Our calculations of the correlation between tolerance and
mean productivity were negative and signiﬁcant for both populations although only
moderately associated. Inferred from these correlations, selection based on mean
productivity will increase both Yd and Yp (Table 9). The drought intensity indexes for all
environments in Mexico were less than 0.50 indicating moderate stress for this region.

Selection for increased Yd or tolerance would be ineffective since severe drought is not

84
consistent and high Yp would be desirable in some years. It would be preferable to select
for mean productivity instead of tolerance for this region. The choice of the geometric
mean between stress and non-stress environments to represent mean productivity is
preferred because GM better accounts for large differences between stress and non-stress
environments than does the arithmetic mean.

Since breeding for increased yield does not concurrently increase drought
tolerance, breeders have been searching for a method to indirectly select for increased Yd
and Yp. Acosta, (1988) suggested increasing biomass to increase Yd indirectly since Yd
and Yp were both highly correlated with biomass (r = 0.80). This relationship is intuitive
since a genotype which cannot produce suﬁicient biomass will not have the photosynthetic
capacity to accumulate high yield. In seasons when drought is not a limitation, Acosta
(1988) postulated that selection for increased biomass would Concurrently increase Yd. In
this study, Yp and Yd were both highly correlated with biomass stress and non-stress,
respectively (Table 7). Correlation coefﬁcients ranged ﬁ'om 0.46 (S/L) to 0.72 (S/A)
between Yd and biomass under non-stress whereas correlation coeﬂicients between Yd
and Yp were 0.52 and 0.77 for S/L and S/A, respectively. Although biomass under non-
stress was signiﬁcantly correlated with Yd, it would be more effective to indirectly select
for Yd based on Yp.

Other methods used to determine drought tolerance are drought indexes which can
identify genotypes that exhibit more consistent performance over stress treatments. Two
examples are percent reduction in yield and the drought susceptibility index (DSI). D81 is

based on a reduction in yield but adjusted for the drought intensity of a particular

85

experiment (Fischer and Mauer, 1978). D81 was negatively associated with Yd in both
populations with correlation coefﬁcients of -0.39 and -0. 56 for S/A and S/L, respectively
(Table 7). D81 was positively associated with Yp with values of 0.28 and 0.41 for S/A and
S/L, respectively. Thus selection for low DSI. would improve Yd and reduce Yp, but only
by a moderate amount. DSI was not signiﬁcantly correlated with GM in either population
whereas GM was highly correlated with Yd and Yp. Since the correlation coemcients
between GM and Yd were high in both populations (0.94 and 0.89), selection for high
GM would increase Yd more eﬂ‘ectively than DSI while still increasing Yp.

Harvest index (HI) which is the ratio of seed yield to total biomass, has also been
suggested as a secondary selection criterion to improve yield under drought within the top
ranking genotypes based on GM. High HI could indicate that a genotype has an increased
ability to remobilize biomass to the reproductive organs (Ramirez, 1992). However, this is
a single characteristic which could contribute to increased performance under stress but,
alone, is not suﬂ'lcient to identify superior performing genotypes. Acosta (1988) identiﬁed
genotypes which possessed high HI but had little yield or biomass accumulation capability.
Despite superiority in remobilization and partitioning under stress, these genotypes had
little yield potential under stress or non-stress conditions (Table 7). HI under stress, in our
study, was only moderately correlated with Yd. HI under non-stress was also moderately
correlated with Yp in both populations and not signiﬁcantly correlated with Yd in S/A. In
S/L, the correlation coeﬁ'lcient between HI under non-stress and Yd was signiﬁcant but
small. This indicates that HI under non-stress would not be a useful indirect measure of

Yd nor would HI in either treatment be a good selection index to improve yield. None of

86
these indexes provides an effective indirect measure of performance under drought since
both Yd and Yp must be measured.

Ramirez (1992) stated that Yp and Yd were not good indicators of drought
resistance because the highest yielding genotypes under stress were not the highest under
non-stress and vice versa. This could be explained by differing mechanisms that aﬁ‘ect
yield under the two treatments. The goal of the breeder would, therefore, be to combine
these two mechanisms into a single genotype. A cross between a genotype with high Yd
and a genotype with high Yp would be the initial step. Evaluation of a large number of
progeny would be needed to identify recombinants which have the two mechanisms. The
inability to identify genotypes that yielded equally well under stress as non-stress
(Ramirez, 1992) was probably due to a limited number of progeny tested. In this study, we
identiﬁed two genotypes, which ranked ﬁrst and second for GM from both populations
and yielded in the top 6% under stress and non-stress treatments (Tables A7-A 10).
Within S/A, T3008-1 ranked ﬁrst for GM and Yd (150 g m'z) and third (197 g m'z) for
Yp. T3016-1 ranked second for GM and Yd (150 g m'z) and ﬁrst (214 g m") for Yp
(Tables A7 and A8). T3016cl and T3008-1 ranked 22nd and 47th, respectively for DSI.
Improved drought tolerance is inversely related to DSI such that the number one ranking
is the lowest DSI value for a population. Three genotypes, T3004-1, T3020-2 and T3048-
1 ranked 77th, 78th and 80th for GM. These three genotypes also appeared in the bottom
6% for both Yp and Yd. r3020.2 ranked 9th for DSI whereas T3004-1 ranked 34th and
T3048-1 ranked 66th for DSI. T3110-2 from S/L ranked ﬁrst under stress (133 g m") and

non-stress (180 g m") treatments and T3147-2 ranked second (130 g m'z) under stress and

87

fourth (176 g m’z) under non-stress (Table A9 and A10). These same genotypes ranked
65th and 68th for DSI, respectively. The deﬁciency of the drought susceptibility index in
beans is illustrated by the genotype, T3153-1, that ranked last for GM and in the bottom
5% for both Yd and Yp but ranked 28th for DSI (higher than the top yielding genotypes).

Ramirez (1992) postulated a breeding strategy which involved selecting a
proportion of genotypes from a population based ﬁrst on high GM and then, ﬁ'om within
the selected genotypes retain those with low DSI. In this way, Yp can be increased
without a decline in Yd. When the t0p 25% of genotypes identiﬁed in S/A based on GM
were ranked for DSI, the previously selected genotypes, T3016-1 and T3008-1 ranked
18th and 14th. If genotypes were selected based ﬁrst on GM and then DSI, these two
genotypes would have been eliminated. In S/L, T3147-2 and T3110—2 ranked 6th and 8th
for DSI within the top 20% of genotypes identiﬁed based on GM. Clearly, using DSI as a
secondary selection criterion is inconsistent between populations. When selecting for
increased yield under both stress and non-stress conditions, GM should be used ﬁrst with
high Yd as the second criterion to more consistently identify the genotypes with the
highest-yielding capability under both treatments. Since our objective, in a breeding
program, is to increase performance under both stress and non-stress conditions
simultaneously, selection for increased yielding capacity under both treatments using GM
as the ﬁrst criterion followed by selection within this group for high Yd is the most
appropriate approach.

Correlations between traits measured at the Kalamazoo County experiment in

1994 trended similar to those calculated for S/A and S/L. The stress applied at this

88
location was severe with a D1] of 0.76 (Table 8) and was comparable to that observed
(DII = 0.63 - 0.78) in a previous study (Ramirez, 1992). Under severe stress, drought
tolerance becomes dependent on the ability of a genotype to accumulate biomass. Yd was
most highly correlated with biomass (r = 0.75) under stress and DSI was more highly
correlated with biomass under stress than Yd. This indicates that a genotype which can
accumulate increased biomass under severe stress conditions is less likely to have a large
reduction in yield. Number of pods per plant and Yd were equally correlated with DSI
suggesting that the ability to develop pods is an important determinant in yield reduction
under severe stress (Acosta-G and Kohashi-Shibata, 1989). The correlation coeﬁcient
between D81 and Yd was -0.66 for this experiment which suggests that DSI would be a
better indicator of Yd under severe stress since the correlation coemcient between DSI
and Yd in the combined analysis over Michigan and Mexican locations (DII = 0.26) was -
0.39. It would be effective, therefore, under severe stress conditions to select for GM ﬁrst
and then, as a secondary selection criteria, identify those with low DSI. However, biomass
under stress was even more highly correlated (r = 0.75) to Yd than DSI suggesting that
biomass would be a better indicator of Yd than DSI as a secondary selection criteria.
Although GM was highly correlated to Yd (r = 0.88), biomass. under stress was a close
second which may indicate that either GM or biomass could be used as a primary selection
criteria. However, biomass under stress was not correlated with Yp whereas GM is.
Consequently in areas of severe drought, it would be effective to breed for increased GM

to maintain Yp and then select based on increased biomass under stress conditions or Yd.

89

Drought tolerance per se is an ambiguous term, possessing different deﬁnitions for
diﬂ‘erent disciplines. For agronomic purposes, drought tolerance refers to improved yield
performance under stress conditions. However it is often measured in terms of a percent
reduction in yield ﬁ'om non-stress. Problems arise, when basing selection on this type of
index, since many genotypes that have little yielding capacity in any environment will
exhibit a small reduction in yield across treatments. These genotypes, although they could
be considered drought tolerant, are agronomically undesirable. Alternatively, genotypes
that yield much greater than the population mean under non-stress conditions but
experience a large reduction in yield under stress would be discarded because they are not
drought tolerant per se. However, although these genotypes exhibit a large reduction in
yield under stress compared to non-stress, their yield potential may be so great that they
will still yield more than the population mean under stress conditions. These are not
drought tolerant genotypes but are still regarded as agronomically acceptable. For
breeding purposes, therefore, drought tolerance is a misnomer and improved yield

performance under stress and non-stress conditions is the main objective.

Marker Analysis
GM, percent reduction in yield, and DSI, are all selection criteria that depend on
yield measurements under both stress and non-stress conditions. Because Yd and Yp are
required for these indexes, costly drought experiments under both conditions must be
conducted. Additionally, other selection criteria, such as biomass and HI under non-stress,

have proven to be poor indicators of yield performance under stress. It was our objective

90
to develop a selection method where improved genotypes could be identiﬁed without the
need for testing in replicated ﬁeld trials under stress and non-stress conditions. Since
molecular markers have been associated with QTLs controlling many quantitative traits in
other crop species (Dudley, 1993; Edwards et al, 1987 and 1992; Lander and Botstein,
1989; Paterson et al., 1990 and 1991; Martin et al., 1989; Freyre and Douches, 1994a and
1994b; Stuber, 1986; Stuber et al, 1987), we postulated that linked markers would be an
appropriate tool to evaluate indirect selection of drought tolerance in common bean. Since
our objective was to develop markers without mapping their exact location, one-way
analyses of variance with signiﬁcant F-tests were used to determine associations (Tables
A9-A14). Dudley (1993) demonstrated that one-way analyses of variance were
comparable to interval mapping but do not require the relative location of markers (Lander
and Botstein, 1989). A linkage map of Phaseolrrs vulgar-1's was available, but sufﬁcient
RFLP markers appearing on this map were not polymorphic between the parents used in
the crosses for this study (N odari et al, 19928). The populations used to generate the P.
vulgaris map were developed ﬁ'om wide crosses making the map less useful for breeders
who generally work with related germplasm. Thus, RAPD markers which generated
polymorphisms between the parents were used in this study. The simplest and most direct
method for identifying markers signiﬁcantly associated with a quantitative trait will be
through the use of one-way analyses of variance, particularly when the location of the
QTL is not required. The seventy markers analyzed for each population in this study did
not prove consistent over all locations and years. In fact, no single marker appeared

signiﬁcant in more than three environments. This can be explained by the strong

91

environmental inﬂuence indicative of a quantitative trait such as drought tolerance.
Because yields vary dramatically between environments a marker may explain a signiﬁcant
proportion of the variation for yield using data from one year but not explain a signiﬁcant
proportion the next. Data ﬁ'om a single growing season is not adequate to determine
accurate associations. It follows that a single marker which represents one loci is also not
suﬁcient for identifying accurate associations. In order to account for environmental
variation, 8 region of the genome must be analyzed against the yield data over several
locations. As seen in interval mapping, a curve representing LOD scores will be distributed
over a region of the chromosome with its maximum as the most likely position for the
QTL on that map (Lander and Botstein, 1989).

The markers from each population were analyzed using MAPMAKER to identify
linkage groups within each population (Fig. 5 and 6). The markers within each group were
analyzed as a multiple regression against data from both individual years and the combined
analysis to determine if a region of the genome could better explain the variation (Tables
10 and 11). Multiple regression techniques have been proposed for QTL mapping
however the basis for this technique has been mainly theoretical (Moreno-Gonzalez, 1992
and 1993; Zeng, 1994). In order to factor in environmental variation, we felt that a region
of the genome marked by several markers would more accurately identify associations and
explain a higher degree of variation. Due to the strong environmental inﬂuence on yield,
analyses of variance using single RAPD markers were not resulting in consistent
associations across locations. This could be explained by ﬂuctuations in the yield data

where one RAPD may be observed to be signiﬁcant when using data ﬁ'om one location

92

but another RAPD marker which is adjacent to the ﬁrst may prove signiﬁcant in another
location. Thus, the most likely position for a QTL controlling the trait of interest
ﬂuctuates across a region of the genome due to environmental inﬂuences. A multiple
regression analysis using all RAPD markers Within a linkage group would better explain
the variation for this trait in all locations. Once a linkage group is identiﬁed, QTL(s)
associated with the group can be ﬁne mapped using interval mapping or other statistical
methods. However, for the purposes of improvement, breeders should select on the basis
of all the markers in that region to account for environmental variation present in their test
plots.

Multiple regression analysis detected speciﬁc areas of the genome that were
consistently signiﬁcant over locations. For S/A, linkage group nine was signiﬁcantly
associated with Yp, Yd and GM in both 1990 and 1991 Michigan locations, 1992 Madero,
Dgo., and the overall analysis, and in all cases, could explain a moderate amount of the
genetic variation (112 = 0.08 - o. 14; Table 10). When analyzed independently, however,
each marker in linkage group nine was signiﬁcantly associated in only one environment
(Table 12). The markers in this group were chosen as a selection criteria for drought

tolerance along with two other unlinked markers.

Marker Assisted Selection
To test the effectiveness of the MAS method, RILs within S/A selected based on
markers were identiﬁed and grown in stress and non-stress treatments under a rain shelter

in Kalamazoo County, MI in 1994.~ Orthogonal comparisons performed between

93

genotypes identiﬁed as potentially above average and those which are potentially below
average were signiﬁcant except for biomass and number of pods per plant under stress
conditions (Table 13). Because LSDs from previous experiments were large and
inconsistent, we did not feel that this would be an unbiased measure of differences
between genotypes so orthogonal comparisons were performed between genotypic groups
(PAA and PBA) as demonstrated in previous marker studies (Stuber, 1986; Stromberg,
1994). Marker-based selection improved yield under stress by 3 .5 g m“2 over the grand
mean and 15.7 g 111’2 over the experimental mean for yield under non-stress conditions
whereas MAS selection for low yield was 6.1 g rn'2 and 14.7 g m'2 below the grand mean
for stress and non-stress, respectively. This suggests that improvement would be more
eﬁ‘ective if the PBA group Was eliminated ﬁom the population as opposed to retaining the
PAA group. Under non-stress, the diﬁ‘erences of PAA and PBA from the mean . were
equivalent. A signiﬁcant improvement was observed for PAA over PBA supporting our
hypothesis that selection can be performed using associated markers.

Genotypes T3025-1 and T3053-l were selected within the PAA group and, based
on GM, ranked fourth and ﬁfth for the S/A population. The PAA group, however, did not
include the genotypes T3008-1 and T3016-1 which were previously identiﬁed based on
yield data as making ﬁrst and second for GM. The genotypes selected among PBA
included two genotypes, T3 004-2 and T3024-1 which ranked 70th and 73 rd for GM based
on yield data for 80 genotypes. PBA did not include T3020-2, T3004-1, nor T3048-l,
three genotypes identiﬁed in the bottom 5% for GM. For MAS to be effective, these

genotypes identiﬁed by their performance data, should also have been identiﬁed by their

94

marker genotypes. However, the markers used for selection explain only 14% of the
variation and could only identify 11 RILs for PAA and 10 for PBA. The selection intensity
was therefore low and the markers did not accOunt for a signiﬁcant amount of variation.

All genotypes from S/A were planted in Madero, Dgo. and Calera, Zac. in 1994.
This data was used to test the eﬂ‘ectiveness of MAS for experiments performed in Mexico.
The same genotypes were contrasted using Mexican data in an analysis of each location
separately and a combined analysis over both locations (Tables 14 and A18). In none of
the analyses, were signiﬁcant differences detected between PAA and PBA indicating that
MAS was not effective for drought tolerance in this region. Yield should have improved
by 13.7 g rn‘2 under stress and 17.9 g in2 under non-stress at a selection intensity of 10%
based on the calculated heritability estimates (Table 3) and the gain ﬁ'om selection per
cycle (Table 4). The gain ﬁ'om selection based on MAS was only 0.1 g m‘2 for stress and
6.4 g m'2 for non-stress (Table 14). Clearly, MAS was ineffective in this population. Since
the markers chosen were associated more consistently in Michigan locations we may have
identiﬁed regionally-speciﬁc markers which were ineﬂ‘ective for improvement of
performance under drought in Mexico. Alternatively, these markers may be better
indicators of increased performance under severe stress.

In S/L, linkage group one was signiﬁcant for at least one of the three parameters
(Yd, Yp, GM) for all environments including the overall combined analysis (Table 11).
This linkage group explained up to 62% of the variation for improvement of yield
performance. Because the linkage group consisted of ﬁve RAPD markers, no other

independent markers were chosen for use as selection criteria. The markers within this

95
linkage group spanned an area of 26.8 cM and all ﬁve RAPDs associated with increased
performance under moisture stress were derived from Sierra parent (Fig. 6). Thirty-ﬁve
RILs were selected for the PAA group based on their marker genotype and 28 RILs were
selected for the PBA group as opposed to 10 and 9 genotypes, respectively, from S/A.
Because the ﬁve markers are within the same linkage group, less recombination will occur
between the markers and more individuals within the p0pulation will possess the same
parental genotypes. Only two unlinked markers and one linkage group with two markers
were chosen as selection criteria, for S/A The chance for recombination will increase
because independent assortment will occur among the two unlinked markers and linkage
group nine resulting in the identiﬁcation of fewer individuals with the speciﬁed genotype.
This has serious consequences on the use of MAS as an eﬂ‘ective breeding tool. The basis
for MAS would be to select superior genotypes in earlier generations to eliminate the need
to maintain inferior individuals through subsequent generations. An alternative approach
would be to rogue inferior individuals based on marker genotype. However, a selection
criterion which eliminates or maintains the majority of the population is not desired. A
breeder would require moderate selection pressure of around 30%. In the case of S/A, 10
genotypes would have been retained after selection whereas 9 would have been discarded.
This is not a practical selection I intensity. In either case, too few individuals ﬁ'om a
population of 80. were selected. For S/L, 35 individuals would be maintained (equivalent
to 36% of the population) or, alternatively, 28 individuals (equivalent to 29% of the
population) would be discarded. Too many independently assorting markers will result in

an excessively high selection intensity. For the most effective selection, a linkage group

96 >
which explains a substantial amount of the variation of a population should be identiﬁed
and exploited.

RILs ﬁ'om S/L were grown in Madero, Dgo. and Calera, Zac. in 1994 to test the
eﬂ‘ectiveness of MAS on selection for drought tolerance. The same analyses were
conducted as were performed on S/A data from Mexico (Tables 16 and A. 19). The
orthogonal contrasts identiﬁed highly signiﬁcant differences between PAA and PBA at
Calera but not at Madero (Table 16). However, in the combined analyses, all contrasts

were signiﬁcant (Tables 16 and A. 19). PAA under stress and non-stress conditions yielded

13 and 15 3 m7, respectively, more than the grand mean. This supports our hypothesis
that MAS would be effective in identifying a group of superior performing individuals.
With a more moderate selection intensity of 30%, MAS identiﬁed genotypes
T3110-2, T3147-2, and T3140-2 which ranked ﬁrst, second, and third for GM and within
the top 5% for Yd and Yp. Included in PAA were genotypes T3118-1 and T3107-1 which
ranked second and third for Yp. Fifty-seven percent of the genotypes identiﬁed by MAS to
be included in PAA ranked in the top 30% of the population for GM based on ﬁve years
of data whereas only 28% of the individuals included in PBA ranked in the bottom 28% of
the population for GM. These data suggest that the markers are more suitable to identify

above average individuals than below average individuals.

Marker Assisted Selection Compared to Conventional Selection
There has been limited research conducted to compare the effectiveness of MAS

with conventional selection. Stuber and Edwards (1986), identiﬁed markers associated

97
with 82 quantitative traits in maize. Two populations were constructed from crosses
between parents which exhibited extreme phenotypes for agronomic and morphological
characteristics. Seventeen and 20 markers which covered 40% of the genome were used
to determine associations in the COTX and CMT populations, respectively. Having
identiﬁed markers associated with grain yield, Stuber and Edwards (1986) tested the
eﬁcacy of MAS compared to conventional selection. Selection was performed among F2
plants based on marker genotype. Concurrently, conventional mass selection was
conducted on an open-pollinated F2 population for comparison. Thirty-seven individuals
were selected from each population in divergent directions. Populations consisted of 1,776
and 1,930 F2 plants in COTX‘and CMT, respectively. This is equivalent to a 2% selection
intensity. Using similar contrasts (1 df) to those used in this study, Stuber and Edwards
(1986) reported a 12% increase of PAA over PBA for grain yield (g plant'l) for COTX
and a 40% increase of PAA over PBA in CMT. Markers identiﬁed in CMT accounted for
twice as much variation which would explain the dramatic increase in grain yield.
Comparing conventional selection, AA out-yielded BA by 16% in COTX and by 24% in
CMT (Stuber and Edwards, 1986). Although MAS resulted in a greater diﬁ‘erence
between the opposing groups (PAA vs PBA) than did phenotypic selection (AA vs~ BA),
PAA and AA showed similar mean values. This implies that phenotypic selection and
MAS are equivalent selection techniques for improved performance. However, PBA
yielded much less than BA which did not differ ﬁ'om the experimental mean, indicating
that MAS is more effective in decreasing performance than phenotypic selection (Stuber

and Edwards, 1986).

9.8

In our study, PAA out-yielded PBA in S/L, on average, by 12%, the same value
reported for COTX (Table 17). However, phenotypic selection based on Yd resulted in a
non-signiﬁcant 6.5% increase of AA over BA. Selection based on Yp showed a negligible
difference of 0.5% between opposing groups and selection based on GM was intermediate
with a 4.5% increase. Since GM is the mean of the two treatments, we would expect it to
have a moderate value. Our phenotypic selection was not as dramatic as that reported by
Stuber and Edwards (1986), however our selection intensity was considerably less.
Additionally, our results were based on two locations with only two replications per
location whereas Stuber and Edwards (1986) reported three locations with ten replications
per location. One important difference betWeen studies lies in the genetic structure of
individuals; open pollinated corn populations (Stuber and Edwards, 1986) vs. RIL
populations in common bean. In our study, PAA out-yielded AA and the experimental
mean, indicating MAS is a better selection criteria for improved performance (Table 17).
PBA, on the other hand, did not diﬂ'er much from the experimental mean while BA did.
This also supports our conclusion that the markers used for MAS in this study are more
eﬂ‘ective in identifying above average than below average individuals. Contrary to Stuber
and Edwards (1986), phenotypic selection is a better indicator of low performance than
MAS when selection is based on Yd or GM.

Stuber and Edwards (1986) did not report the amount of phenotypic variation
accounted for by the markers used in the Selection study although the markers used to
identify associations only accounted for 40% of the genome. With 62% of the variation

explained by the markers used in this study, we would expect a greater diﬂ‘erence between

99

divergent groups than was reported (Table 11). The lack of difference can be explained
by the lack of genetic variation for yield in S/L. The parents used for the populations in the
Stuber and Edwards study (1986), were diverse phenotypes. Sierra and Lef-ZRB are both
considered drought tolerant and are representative of the same extreme phenotype for this
trait. Thus, genetic variation is less, h2 is less and expected increase in mean yield is
reduced. The gain in selection based on the calculated h2 (Table 3) at a selection intensity
of 30% was 2.0 and 1.4 g m’2 for stress and non-stress, respectively (Table 5).
Conventional selection did not improve the mean by the expected amount in either stress
or non-stress treatments (Table 17). Although expected gains were not achieved with
conventional selection, similar results have been observed in previous studies with
common bean (White et al., 1994). MAS, however, exceeded the expected gain per cycle
of selection by 10 and 11% for stress and non-stress respectively (Table 17). This is a
substantial improvement which suggests that MAS is more effective than conventional
selection for drought tolerance in this common bean population.

We have demonstrated that MAS was more eﬂ‘ective in S/L than S/A Although
the markers used to select genotypes in S/A accounted for only 14% of the variation for
drought tolerance, it was also more difﬁcult to identify markers consistently associated
across environments in this population compared to S/L. As quantitative theory suggests,
the eﬁ‘ectiveness of MAS is inversely proportional to the heritability of a given trait
(Lander and Botstein, 1989; Lande and Thompson, 1990; Paterson et al., 1991). The
results ﬁ‘om this study are consistent with this statement. The h2 for yield under both stress

and non-stress treatments in S/A was four times greater than the h2 for yield in S/L.

100

Additionally, the genetic variation was much greater in S/A than S/L for yield under both
stress and non-stress conditions. Thus, for populations where improvement of a particular
trait by conventional means is limited due to a lack of adequate genetic variation, MAS
becomes an appealing option. In S/A, where h2 for yield is from 0.45 to 0.47, considerable
improvement can be made by conventional selection. However for S/L, where h2 for yield
is from 0.08 to 0.11, MAS proved more effective than conventional selection. If the
effectiveness of MAS is consistent with quantitative theory, we would expect that
improvement in S/A based on MAS, using the RAPDs from linkage group one of S/L, will
not realize the same gains in selection as was Observed in S/L.

This does not preclude the use of MAS in populations with higher genetic
variation for a particular trait; Situations may arise where the trait of interest cannot be
measured, for example, drought conditions may not exist. Selection can still be performed
using MAS. Additionally, MAS used in conjunction with conventional selection may prove
an effective method to accelerate progress to improve a trait of interest.

It is probable that because the marker genotype for RAPDs in linkage group one
associated with above average performance originated ﬁ'om Sierra, most mechanisms for
drought tolerance in Sierra and Lef-2RB are similar except this region of the genome for
which Sierra aﬂ‘ords an advantage. This would explain the lack of variation observed for
yield and biomass in S/L compared to S/A. Additionally, it would explain why one linkage
group could account for such a large proportion of the variation for a polygenic trait such
as yield performance under moisture stress. Since alleles associated with increased

performance ﬁ'om RAPDs in linkage group one originated from a single parent, Sierra can

101
be used in backcrossing schemes to introgress this region of the genome into deﬁcient
germplasm. Although markers associated with performance under drought were restricted
to a single region of the genome, the polygene theory for yield-related traits is not
precluded.

For markers to be useful in breeding for drought tolerance, early generation
marker-based selection must be tested. Stromberg et al. (1994) attempted this in corn but
MAS performed on F2 genotypes failed to increase performance of the inbred lines above
the population mean. However, conventional selection also proved ineffective. A selection
index which explained only 25% of the phenotypic variation, the inﬂuence of
environmental factors, and the limited number of loci used to identify associations all
contributed to this outcome (Stromberg, 1994). To verify the efficacy of the markers
associated with drought tolerance in this study, new populations must be developed and
MAS performed on F3“ RILs. If markers prove to be population speciﬁc, Sierra could be
used as one parent making MAS useful, presuming that polymorphism for these markers
exists between the chosen parents.

The mechanism that distinguishes drought tolerance in these two genotypes
remains unclear. It has been suggested in previous studies that Sierra has a greater level of
stomatal conductance which could be important to drought tolerance (Ramirez, 1992).
Heliotropic movement, where a genotype will orient leaves parallel to incident sunlight to
avoid direct rays during stressful periods of the day has been observed in this genotype
and implicated as a mechanism for drought tolerance (Ramirez, 1992; Kao, 1994). Due to

Sierra’s more upright plant stnlcture, it also possessed a more distinct tap root than Lef-

102
2RB which can exploit soil moisture to a greater depth (Lynch and van Beem 1993).
However, Sierra did not perform well under stress conditions in either population due to
its lack of adaptation to the Mexican Highlands. It is possible that any of these traits,
combined with the adapted background of Lef-ZRB, could have contributed to the
improved performance of the progeny under drought conditions. Thus, markers would be
useful to introgress this region of the genome from Sierra into more adaptable germplasm
thereby increasing performance under drought. Additionally, the development of near-
isogenic lines that diﬂ‘er only by this speciﬁc region of the genome would facilitate
physiological studies which could lead to a better understanding of drought tolerance

mechanisms in common bean.

APPENDIX

103

 

e4 ﬁuaw ..oov cw 3...“...—

 

 

 

 

s...... 3. 8.. a8: 8.. $8 8 2.889
2 8.5. 3.8 . 8.8.8..
3 ...... a: .8. >0
3 new ...8 .8. v8. :3
can . 8.5. ...8 8o: :85
.8... 8.980 88888.. ”8.8.2 own .28.“: N8.

2 28" .82. 8 88m

e... e... S 5...: ca 8.8: . 33860
at: 2m 2:...” p.88 :1: ~83 ... 2.886
2 38. .....S . 8.8.8...
2.... an" 8.8 .8. >0
a... «.8. «.8 w .8. v8. :3
«.8 3: «...: 8o: ...—co

8... .2. as... .o .882. 2888 .. .8... ".8380 88.8.5. ”..2 8.83 an. .8.

- - ”.3... «.8: 8 88m

- - S... . .N 3 Zone . season
- - 2.... 8.22. .... 2e: 8 8880
- - 4.82 2 «.88 . 8.8.33.
- ...: a..." .8. >0
- as: «.8 e3
- ...E ....n. 8o: :85

8... 8.. 88... .o .882. 2888 .. o.8... 8.8.8 .3888... H..... 588 8.8.82 8...
28> a. m: 28> ... ms. 28> 2. ms. ...a 8.8m
.3 .3 ..oom 8. ...... m. .88... ...a u. so;

 

60.52 E... 535.2 E acouﬁuoo 80...... .095 Ava—tome: 22.30. :38 .96 :38»
88.2.9. 823..an 8.. 8.58.... 82. ... .o Ema; .88 8. 8a .38... ....o... 8. 85.5., .o .885. .3. as:

 

WORN :69: 3 mam

 

104

 

 

 

 

- - 3.8 e.««... e. .8888. 2.3. ......S. 8.8...
- - as. 92.... :8. ....8. 8 2.880
- - 8.38. «.82. . 888:8...
- ....« e.n« .8. >0
- ..m... 33 a... v 8. am.—
- «a: «.8. :82 2.9.0
8.8.. 88$ 3... ”8.82 88:5 882 a...
.... 2... 28 8 deem.
«.« 25 3.2... e. 88%... 2.3. 55.3 8.8...
tees «.... :8.« :«n :82 8.82 8 8880
e. 99.... 4.8. . 8:883.
etc. e.«. ...... .8. >0
...« .....m S... a... v8. :3
«.n« a.e«« «.2. 8o: 89.0
884. 283 a... ”8.082 8.8.88 .880 «8.
m... «.88 98.. 8 88m
n... 3.8 28 e. 8.8.8.. 2.3. 55.3 8.8:.
...en.« e.o« 2:2 .« 3.8. .8. «a: 8 2588
8... I. «.e« . 882.8...
8.... ...: 9.8 .8. >0
5 «..e «.8 .8. v8. 8..
...8 ..8. one and: 8:0
8.8.. 285 a... ”8.82 .8859 canons. «8.
28> ... ms. 28> a. m: 28> ... me. a... 8.8m
3 .883 .88 8. mg 8.8... «new. .8...

 

8:880 . .4 use...

105

88. v m 2!: 48. V m 0: ”EV m ..i an... v m...

 

33> «— m2

...»... 8 Sum

«e.«..« e. .8888. an”. 8...? 8.8...

...:vd @692 cm 0909—00
e.«««e« . 8:883.

...e. .8. >0

...... .8. v8. ......

«.8. 8o: ....20

8.8.. 888 a... “8.82 88283 .880 .8.

m2 83> b m2 ..5 8.50m

 

 

 

.3 .883 .88 8.

...... m. 8.8... .8... m. .8...

88.80 .. .< 28...

106

 

9d @6— S «.33 we acum—

 

 

 

 

 

 

._..o... ...... «.. «.8«« «.. ...«3 8 2:350
S. 3.8 «.R. . 8.38:8...
«... «... ...«« ..\.. >0
.... «...... ...nm .2. vs. 8..
n.n« ....«« ...ﬂ. ...»: ......u
.8... 22.3.... 83.3%.... .83.: ...«o .2022 «a...

n.« ....«2. «a: .... .2...

.... ... .... .....«8« 3.... .33. . 03.3.5
:3... ...: 5.... «.28. 52 ...: .. .. 2588
...: «ﬁg ...“: . 838......
.... 3.: «.«« 3.. >9
«.« 3.8. ......” a... v... :3
...«n 3...: .3. 5oz 23......

.3... .2. 9.3... .3 .352. 03.3.8 .. .8... season. 8.3.8.... ”.2 .3335... .8... a...

- - «.38 3...: 3 8......

- - «.«m n.«.«.....« «.S «.82.. . 2333.50
- - .3. «.82. .n. ......3 8 .388
. - a...“ 3.8... . 838......
- ...n« .....« 3.. >0
- ...... «...... .2. v a. an.
- ...»: 3.9. .82 ...-....

.8... .2. 3.3... ... .8332. 233.80 .. .8... 32.....8 8.3.8.5. H3,. 55.5 6.8.82 8...
83> .. m2 83> .. ms. 83> .. m: ...a 8.8m
.3. ...»..3 ..8m 8. ...... u. 33.8... ...... u. ...»;

 

.8382 ...... cumin: E 80.5.50 80.3.30: .03.... Ava—-83. 9.2.82 30>». ..o>o :39.»
8.33.2. 323335 ...... 8.58... .32 .. ... .....33 ..8. 8. ...... ......o... .2... .... 8.3.3.. .... .33.... .«.< 2.3...

107

 

 

 

 

 

 

 

. . «.88 38a ..« sum
- - «.89. «.82 «_ .3»an as. €53 9.83
- - 2 «.o«_« 2 ._ «.3: 8 03890
- - «.«2 n «.88 . 838:5

- «.3 «.3 3.. >0

- 2: «.2 a... v3 n3

- «.«2 «.5 :82 «are

8:3 283 as ”832 693.5 8082 «a.

S. «.«2: «.3: «_ €833 §§§ 9.85

3.3 «.2 :12 «.2 «v :53 «.8: 8 0388
1: ..S «.22 _ 288:9:

«.« «.2 «.3 3: >0

«.n «.«« «.n« 6... vs :3

..«N «...: «.3 :82 «5.5

«85 20326 .3285 ”82.2 «823 .628 8a

3 «.«8 28 S «Em
an «.3: «.«2 «_ 9.8«3 as. 55.? £85

2:3. ...: :1: «.22 ...«2 ES 8 £588
3 32 «.3 _ 888:3

2 «.2 «.2 3L >9

«... «.«m «.«n 3.73 am.—

«._« «.8. «.5 :8: «5:0

823 £35 33 ”882 ...«gn 8082 «a.
83> m «2 83> m «2 33> ... «2 "E 858
I281? 38 2: Jug 3.85 A236;

 

3580 «.... 23.

108

Sec. v m 2.3... 48. v m .....i ”av A 3. ”ma. v m...

 

 

 

 

 

 

. . - - «.«8 3 33m
- - - - «.23 «_ 9833 33. £55 383.

- - - - :32 «>83 8 33280

- - - - «.382 _ 388:8...

- - «.2 3% >9

- - «.8 5. v3 :3

- - «.33 a8: «5.5

833.. 233 as ”8382 «.8283 .8830 3....
33> .— «2 33> ... «2 33> m «2 "E 83$
3 23333§ $333.83 «33%:

 

8.338 «a. 233.

109

 

ad «den ﬁvun 5 «ohm

 

 

 

 

 

_.« «.««« «.««« «_ «88.33 3323.3» 385
.53. «.2 2.3.. «.32 :5 «.««« «« 25°80
«.« «._««« «.««: ~ 3:38:83
«4. «.5 «...« 3.0 >0
«.... «.«n «.8 «3.13:3
«.3 «.«2 «.2. «8: «:80
«85 «8338 8333.5. ”8382 ...«o .8032 as

«.« «.22 «.«9 «« «Em

«s «a «a «.«««« «.32 _ «33360
...«d «.«« 2:2 «.23 2 «.««« «« 33.80
S n «.38 «.««« _ 838:8...
«.... «.«« «.«« 3b >0
«.... «.«« «.3. «3. vs :3
«.«« «.8. «.3 «82 «:80

co: .3 333 «o .33an 233.50 .. «85 «33:30 «83.3.2 «2 «33.3 33m 82

- - «.«««« «.2 2 «« «Em

- - «.3 «.«8«« «s «.52 _ «3.550
- - 2 «.«««« ...: «.33 «« 258.6
- - «.33.. «.«« _ 8383.3.
- «.«« «...—n 3s >0
- «.«2 «.«« :3
- «.«R «a: .8: «:80

«as 8.— 3.5 «a .356 23380 .. «83. 22980 «33:38. a: .3550 3.8.82 «3.
33> "— «2 33> « «2 33> « «2 «a 8.8m
3 «333 «8« «2 m3. 3 3365 «we «0 «33

 

.3382 3:3 «BE—«£2 E «333:8 «web... «03:: A3263 3 «5:82 :o>o« «o>o ESE
coax—«ac.— mdﬂmodabomm Sou «258% :33 «x: .3 Emma? coo... 2: 3:3 @3803 .203 «cu «Baum—g .«o «8235‘ .m.< 033.

 

«an 9:2 @02— 3 «ohm

 

110

 

 

 

 

 

«.« «.:«: «.««: 2 «383.33 83.223 285
...: «.2 .22 «.«««« .«2 «.««2 «« 82800
«.2 «.2 3 «.««2 _ «838:5
«.« «.«. «.«« 25 >0
«.« «.«« «.«« «2.13.52
«.3 «.«3 «.«3 «32 «5.0
823 28$ «:2 ”8332 .3339 8832 ««2
«.« «.«3 «.3« _« «Em
«.« «.««: :2 .« «.««3 2 2.8230 «.5323 283
:32 .2 «.3 ...«3. «.«««« «.««2 «« £5890
«.« :«3 «.«««« _ «8:835.
«.«. «.2 «.2 3.0 >0
«.« «.«« «.«« 22.23 :3
«.3 «.««2 «.««. «8: «5:0
823 23$ «:8. ”830: «8283 .3830 ««2
«.« «.««. «.««« _« «Em
«.« .52 .« «.««« «.«3 2 9.83.33 83. 25.3 285
«.«« «.««2 :««._ «.23 «« 8.586
:33 «.«« :32 «.33 _ 83835.
I. «.«« «.«« 32 >0
«... «.«« «.«« 5. vs :3
«.«« «.«« «.«« «82 «5:0
823 23$ «:8. ”83...: .8559 8832 ««2
33> « «.2 83> « «2 83> « «2 «a 858
8 «.333 «8« «2 p.340 8355 m3. «335

 

8.3250 .3. 2.33

111

88. v m 21.... .25. V m ...: M5V m .3 Kc. v m...

 

 

 

 

 

- - - - I? a sum
- - - - 2: 2 : €8£3 as. 55;» 9.85
. - - 2:»? v.82 a £388
- . - - 32.2. _ 288:5
- - can soy >0
- - v.3 Ge. v .8 Am.—
- - «.3 :32 ...—9.0
8:3 88$ 2.5. ”832 «.833 .825 as
83> ..— mz 25$ m m: 83> m m: "5 85am
As 2%; Bow 8_ ma 3 38.5 m.& 3 22>

 

82.2.8 .2 23

112

 

6d v.2: _ n6: 3 Sum

 

 

 

 

 

.... .N ...... ...... 3R. .9. “.8... 2. 25890
2... ”.82 ...»..n . 838...»...
a... 3.. ...... 3.. >0
a... «.8 S... G... vs. :3
..ﬂ ...: ...... :82 ......u
.8... 32.58 38.5. ”8.5.2 an .2022 a...

e.“ .22 $8. 8 Sum

n... ... 3. .....SN ...... 28: . Suﬁsm
.53 ...: .....N 22.. ...2 ..82 a £950
...m 32.... .388 . 8.3.3.
a... a... a... A... >0
.... ...? ran. a... v... :3
3.... San n.9,. :82 ......u

.8... .2. 25... ... .852. 2.580 .. “.8... 329:8 89.385. ”.2 $.83 3m .8.

- - .....8. .... .N. .3 sum

. - 93 3:5. 2.. 28:. . 3:380
- - .... .52.. ...... 3.3. 8 25800
- - ..SB: 232. . 5:829.
- .... a..." 3.. >9
- «...: a...» :3
- ......" an: :8... 2.26

.8... .2. as... ... .852. sag .. u.8... 20.9.5 .3583. ”.2 55.8 382°: 8...
82> . m2 33> . ms. 33> . m: .3 8.8m
3 23.»? Bow 8. 0.... m. 35.85 A”... a 23>

 

62on 23 5320.2 5 £822.00 acumen... .09... Avg _ 68 3 m:o_.§_ 55.. .96 .56.»
sown—2.2. 59.95.86 80... non—bosom 52. oo— ..o Emma? 39. 8. E... .3an... .203 .0. 35.53 .3 «on»? .v.< 039—.

113

 

 

 

 

 

n.« «.3. «.8. .« .2...
.... «.««S 33 «. €83.33 ...... 55.3 9.8...
......n 2.. .2... «.«««e ...... «.82 8 8.889
«.«. «.22 «.««. . 838...»...
...» «.«« «.« ..x..>9
«.« «.««. ...« .3. v 5. ..«..
..«« ...«« «...... as: 23.9
8...... 28$ «:8. ”838,. 35.8 8832 a...
«.« ...«8 3.8 .« .8...
«.« ..«.: ««.«« «. .3833... a... 8.5.3 5.8...
....3 ..«. ...... ««.«. ......n ..82 2. 8.389
«... «.32. 3...... . 888......
«.« ...: ..«. 3.. >9
«.« ...... ...... a... v3. ..«..
«.«« e..«« 5.3. :8: «5:9
8.8... 28cm «:8. ”8.5.2 «8285 .8039 «8.
3. 3.8 «.3 .« 8:...
«.« «.8. «.««.“ «. 98...... a... 55.3 5.8...
...«s «.«« ...... ««.«. ..«.. «..«.. a 8.889
«... «.««... «.3: . 388...»...
«... ...... ..«. 3.. >9
«... «.3. ««.« 6... vs. ..«..
«.«« «..«. ...... :32 ......9
8...... 28$ «38. ”838,. «.«an 8832 «a.
83> . «s. 83> . «s. 83> . «2 ..«. 888
.3 «3.03 ..8« 8. ...... u. 38.2.. ...... a so...

 

.3888 .3. 2......

114

88. V .— :o. .86. v .— ...: ”SV .— ..i ”w... v A...

 

 

 

 

- - - - 8.2... .. .8...

. - - - 3.... .. .8853. 2.3. 8.5.3 3.85

. - - . ...... .. 23.. a 8.860

- . - - ..3.8. . 8858.33.

- - e...— .o\.. >0

- - ...» .2. vs. .5.

- - «.5. .82. .52...

853 28$ 2.8. ”8.52 «8285 .835 a...
83> h. ms. 83> .. m: 8...> 5 ms. .... 858
.3 2.5.03 .88 8. ...... m. .8585 ...... m. ...»...

 

82.88 .3. 2...:

115

Table A. S. Split-plot analysis of varianceT for yield, biomass and 100 seed weight of 80
bean genotypes from the Sierra/AC1028 population combined over ﬁve locations (1990-

 

 

 

 

 

 

 

 

1993) in Michigan and Mexico.
Source DF MS F Value
—---Yield (g m")---—-—-

Grand Mean 125.3
LSD (at < .05) 11.9
CV (%) 22.2
Replication (Years) 4 300287 .3
Treatments 1 25654.8 28.7”
Error 4 8918.6
Genotype 79 2380.7 3.1""
Treatment X Genotype 79 565.7 0.7
Error 630 768.0

Biomass (g m") ---—-
Grand Mean 221.9
LSD (a. < .05) 20.1
CV (%) 21.1
Replication (Years) 4 4575587
Treatments 1 2549658 2.92
Error 4 87312.8
Genotype 79 7040.2 3.2""
Treatment X Genotype 79 1518.8 0.7
Error 630 2197.9

”—100 Seed Weight (g)1 ---

Grand Mean 28.3
LSD (a < .05) 1.2
CV (%) 8.7
Replication (Years) 3 1775.8
Treatments 1 25.8 0.23
Error 3 87.9
Genotype 79 43.2 7.1""
Treatment X Genotype 79 3.8 0.6
Error 474 6.1

 

I"P < .05; n P <.01; u" P < .001, tenur- P < .0001
tMain plots were stress treatments split for genotype

1100 Seed weight data was unavailable for 1990, Montcalm, MI

116

Table A. 6. Split-plot analysis of varianceT for yield, biomass and 100 seed weight of 98
bean genotypes ﬁom the Sierra/Lef-ZRB population combined over ﬁve locations (1990-

 

 

 

 

 

 

 

 

 

1993) in Michigan and Mexico.
Source DF MS P Value

Yield (g m")—-—
Grand Mean 121.3
LSD (at < .05) 11.9
CV (%) 22.9
Replication (Years) 4 270107.8
Treatments 1 4506242 1189""
Error 4 3789.2
Genotype 97 1900.7 2.4”"
Treatment X Genotype 97 512.9 0.7
Error 776 771.5

Biomass (g m") ----
Grand Mean 200.6
LSD (at < .05) 15.7
CV (%) 18.2
Replication (Years) 4 5858019
Treatments 1 7994698 206*
Error 4 38789.7
Genotype 97 4658.5 3.5“"
Treatment X Genotype 97 905.7 0.7
Error 776 1329.2
------100 Seed Weight (g)’-—--

Grand Mean 27.9
LSD (a. < .05) 1.1
CV (%) 7.9
Replication (Years) 3 2807.7
Treatments 1 8.9 0.2
Error 3 41.3
Genotype 97 64.4 13.3""
Treatment X Genotype 97 2.9 0.6
Error 584 4.8

 

‘P < .05; “ P <.01; "* P < .001, "" P < .0001
TMainplotswerestresstreatmentssplitforgenotype

‘100 Seed weight data was unavailable for 1990, Montcalm, Ml

 

117

Table A. 7. Yield, biomass and 100 seed weight (100 SW) for both stress and non-stress
treatments of the top and bottom ﬁve genotypes ranked by yield under stress in the
Sierra/AC1028 population grown over ﬁve locations (1990-1993) in Michigan and

Mexico".

 

 

 

 

 

 

Stress Non-stress
11t Genotype Yield Biomass 100 sw Yield Biomass 100 sw
-s m"- -s m”- -s- -s m”-- -s m”-— --s--

1 13008-1 150.2 238.4 (11)' 31.4 196.5 (3) 320.2 (3) 32.0
2 13016-1 149.5 248.2 (9) 30.9 213.5 (1) 328.1 (11) 29.2
3 13025-1 145.5 273.5 (3) 28.8 182.6 (7) 282.5 (13) 27.6
4 13053-1 145.5 263.9 (5) 34.6 182.0 (8) 305.6 (5) 32.4
5 13005-1 145.1 255.6 (8) 31.8 184.8 (6) 286.7 (10) 33.4
76 13044-2 89.6 218.9 (24) 27.9 139.6 (60) 234.8 (54) 28.5
77 13004-1 89.2 171.3 (65) t 30.9 121.9 (77) 202.3 (75) 28.9
78 13033-1 85.4 137.6 (79) 28.5 144.1 (54) 211.4 (73) 29.2
79 T3048-1 84.5 141.3 (78) 24.1 104.5 (79) 165.1 (79) 23.3
80 13020-2 76.1 136.4 (80) 25.7 119.6 (78) 185.3 (77) 26.6
67 Sierra 95.7 159.8 (75) 28.2 132.6 (69) 211.7 (72) 27.5
44 AC1028 111.2 183.9 (56) 27.0 147.2 (50) 237.3 (52) 26.8

LSD ot=o.05 14.6 23.0 0.9 16.6 24.5 1.0
"Values for both parents are included
1 11 = Ranking

' Corresponding rankings are in parentheses

Table A 8. Yield, biomass and 100 seed weight (100 SW) for both stress and non-stress
treatments of the top and bottom ﬁve genotypes ranked by yield under non-stress in the
Sierra/AC1028 population grown over ﬁve locations (1990-1993) in Michigan and

Mexico. Values for both parents are included.

 

 

 

 

 

Stress Non-stress
11* Genotype Yield Biomass 100 sw Yield Biomass 100 sw
-s m’z- -s m"- -s- -8 ma— -8 m”- --s--

1 13016-1 149.5 (2)' 248.2 (9) 30.9 213.5 328.1 (1) 29.2
2 13025-2 127.3 (18) 225.3 (19) 28.4 207.8 308.1 (4) 28.0
3 13008-1 150.2 (1) 238.4 (11) 31.4 196.5 320.2 (2) 32.0
4 13035-2 126.0 (21) 218.4 (25) 26.5 196.1 318.3 (3) 25.7
5 13014-1 111.0 (45) 183.6 (57) 28.5 185.2 279.5 (15) 27.1
76 13043-2 90.9 (74) 212.6 (30) 31.3 124.0 217.6 (63) 30.6
77 13004-1 89.2 (77) 171.3 (65) 30.9 121.9 202.3 (75) 28.9
78 13020-2 76.1 (80) 136.4 (80) 25.7 119.6 185.3 (77) 26.6
79 13048-1 84.5 (79) 141.3 (78) 24.1 104.5 165.1 (79) 23.3
80 13002-1 90.9 (75) 153.2 (77) 28.4 98.8 157.1 (80) 25.4
69 Sierra 95.7 (67) 159.8 (75) 28.2 132.6 211.7 (72) 27.5
50 AC1028 111.2 (44) 183.9 (76) 27.0 147.2 237.3 (52) 26.8

LSD a=0.05 14.6 23.0 0.9 16.6 24.5 1.0
1Values for both parents are included
1 R = Ranking

' Corresponding rankings are in parentheses

118

Table A. 9. Yield, biomass and 100 seed weight (100 SW) for both stress and non-stress
treatments of the top and bottom ﬁve genotypes ranked by yield under stress in the
Sierra/Def-ZRB population grown over ﬁve locations (1990-1993) in Michigan and
Mexico .

 

 

 

Stress Non-stress
11‘ Genotype Yield Biomass 100 sw Yield Biomass 100 sw
“8 “1'2“ “8 m-2__ “8" “8 “1'2" . "8 m-2__ "8"

1 T3110-2 132.5 235.6 (2)' 23.7 180.4 (1) 270.0 (4) 26.1
2 13147-2 130.1 212.5 (11) 29.3 176.4 (4) 256.3 (17) 30.2
3 13140-2 126.6 213.6 (9) 27.8 172.2 (8) 273.2 (2) 29.8
4 13108-1 125.2 227.0 (4) 25.6 168.1 (12) 268.9 (5) 28.7
5 T3153-2 125.0 228.9(3) 29.5 163.7 (19) 246.1 (34) 28.5
93 ”13136-1 81.3 144.4 (86) 28.4 130.1 (86) 199.3 (95) 29.7
94 T3149-2 80.8 160.6 (88) 26.3 152.1 (46) 247.9 (88) 26.9
95 13128-2 80.7 144.2 (96) 29.1 124.7 (88) 199.5 (96) 29.2
96 13153-1 77.4 157.7 (91) 26.0 118.2 (93) 189.9 (91) 26.5
97 13152-2 77.4 141.9 (97) 25.8 153.2 (44) 237.4 (97) 25.9
88 Sierra 85.8 160.8 (87) 29.8 169.8 (10) 255.2 (18) 29.5
42 Lef-2RB 106.8 181.2 (59) 30.6 155.6 (37) 243.1 (41) 28.0

LSD (0.05) 12.1 20.7 0.8 14.4 21.1 1.1
iValues for both parents are included
i R = Ranking

 

' Corresponding rankings are in parentheses

Table A 10. Yield, biomass and 100 seed weight (100 SW) values for both stress and
non-stress treatments of the top and bottom ﬁve genotypes ranked by yield under non-
stress in the Sierra/Lef-ZRB population grown over ﬁve locations (1990-1993) in

 

 

 

 

Michigan and Mexico’.
Stress Non-stress
RI Genotype Yield Biomass 100 SW Yield Biomass 100 SW
—s m”- -s m'z- -s- -8 m”— -s m"- -s-

1 13110-2 132.5 (1). 235.6 (2) 23.7 180.4 270.0 (4) 26.1
2 13107-1 105.5 (47) 175.6 (69) 29.5 177.4 261.6 (13) 31.1
3 13118-1 98.6 (66) 180.6 (61) 28.2 177.1 261.0 (15) 29.1
4 13147-2 130.1 (2) 212.5 (11) 29.3 176.4 256.3 (17) 30.2
5 13118-2 95.2 (75) 170.8 (76) 26.5 174.8 263.2 (9) 26.1
93 13153-1 77.4 (96) 157.7 (91) 26.0 118.2 189.9 (96) 26.5
94 13122-1 99.5 (64) 174.0 (73) 25.3 117.5 206.8 (73) 25.8
95 13126-2 92.5 (79) 162.2 (85) 30.5 116.7 176.5 (85) 29.7
96 13138-1 87.7 (87) 175.4 (70) 29.9 116.4 199.8 (70) 29.1
97 13113-2 88.5 (85) 166.4 (83) 26.9 111.8 199.0 (94) 26.8
10 Sierra 85.8 (59) 160.8 (59) 29.8 169.8 255.2 (18) 29.5
37 ch-ZRB 106.8 (87) 181.2 (87) 30.6 155.6 243.1 (41) 28.0

LSD (0.05) 12.1 20.7 0.8 14.4 21.1 1.1

 

1‘Values for both parents are included

1R=Ranking
' Corresponding rankings are in parentheses

119

Table A. 11. Signiﬁcant F-tests ﬁ'om one-way analyses of variance between yield under
stress and 70 RAPD markers. One-way analyses of variance were performed between
RAPD marker genotypes and yield under stress ﬁ'om ﬁve locations (1990-1993) in
Michigan and Mexico and the combined analysis over all environments in the
Sierra/AC1028 population.

 

 

RAPD 1990 1991 1992 1993Mt 1993ZT Combined
0A161m 5. 19"
OAK.” 6.62”
0Al8a 4.87" 8.81“
0A18b 9.01“

0.409610 4.55“ 10.52"“‘"'l

0ABl8goo 6.00‘ 8.57"“I

0101 5.40‘ 6.50“ 4.49‘ 4.90"
0610 9.73"

0011 10.31“I

0605520 5.38‘

06091000 4.42.
01112 8.90“‘ 4.07‘I

011188 11.90“I

01119 11.51”

0H02 15.16“" 5.26"I
01103 7.70“I

01108 8.04“

0114 11.81"“I 3.98‘l
0118 3.72

0119 8.12"

01038 5.74‘

01031) 11.13""I

01.121050 8.57"

0N03 4.3‘ 8.67""I

“21010 4.53

0P037oo 5.25"

0Q06970 3.85‘

011161)” 13.17"" 3.69‘
0118550 6.03"l

0U10 7.06" 5.95‘I
oz01L 6.91“

 

 

‘P < .05; “ P <.01; "" P < .001, "“ P < .0001
1M = Madero, Dgo. location and Z = Calera, Zac. location

120

Table A 12. Signiﬁcant F-tests from one-way analyses of variance between yield under
non-stress and 70 RAPD markers. One-way analyses of variance were performed between
RAPD marker genotypes yield under non-stress ﬁ'om ﬁve locations (1990-1993) in
Michigan and Mexico and the combined analysis over all environments in the

 

 

Sierra/AC1028 population.

RAPD 1990 1991 1992 1993M7 1993z’r Combined
014116.,0 5.989 4.099
0A18b 4.449 4.249

0.407,.o 9.0799

0A09610 5.91‘

0141314.so 7.1099

011318600 6.8399 8.859
OF01 19.45999 4.069
0100...... 5.229
0105..., 3.919 3.929
0010 9.8499

0011 6.389 5.419
06021010 3.90“

0005...o 4.669
0H18a 8.299 4.269

0110111 6.259 5.589

01102 3.849
01103 3.879 5.319 5.869 8.4899
01108 5.159

0110c 7.1199 7.3199
0110.50 6.7799

0110830 5.399

0114 4.139

0118 6.119 5.279
0118 3.66 3.919 3.929
01038 6.7999 4.319
0103b 6.7699 5.339
0103c 5.739

0le"... 6.8399 5.819 5.859
0L07 5.829

019105 9.6999

0Q06970 7.5399

017.11,.o 6.6899

017.16m0 3.949

0101

5.79‘

 

12 1
Table 12. Continued

 

 

RAPD 1990 1991 1992 1993MT 19932T Combined
OT18550 13.12999 3.68
0W201) 5.299

0x03 6.189

0z01.00 6.009 4.599 3.71
0z05 7.2199

 

‘P < .05; n P (.01; ”" P < .001, at!“ P < .0001
1M =9 Madero, Dgo. location and Z = Calera, Zac. location

1.0." "-9; ..ZJVIq

 

 

122

Table A 13. Signiﬁcant F-tests from one-way analyses of variance between geometric and
70 RAPD markers. One-way analyses of variance were performed between RAPD marker
genotypes and the geometric mean of yield under stress and non-stress ﬁ'om ﬁve locations
(1990-1993) in Michigan and Mexico and the combined analysis over all environments in
the Sierra/AC1028 population.

 

RAPD 1990
0A16.,o

0A18a

0A18b

0A07m

0A08m

0A09510

011818550
011318600

OFOIm 7.29"
01:10)”

0105440

0610

061 1

0605.0

01112 6.039
0H18a 15.43 “9
011018

01102

01103

01108

0110950

0114

0118

0119

0103a

0103b

0L12mo

0M05

0N03

0Q06m

0T18550

0U10

0X03

0201M

1991 1992

6.34”I

4.13"

3.92“

17.98"“I

4.59‘

5.94"
4.26“

4.51"I

5.87‘I

4.28‘

4.95“

8.08"l
4.64"
6.49“

4.81‘'
5.05"

*P < .05; “ P <.01; "" P < .001

1M = Madero, Dgo. location and Z = Calera, Zac. location

1993Mf

5.29"I
9.18"

4.54“
8.96“

5.60‘
6.46"

14.50"”
10.55""l
4.95‘I
12.34“”I
5.60“
13.96""

7.68"”I

1339"“
6.20“

1993z’r

3.86“

5.81‘I

7.20"

4.749

8.66""I

6.83""

Combined
5.."18‘I

5.25“
5.41‘
4.34

4.47"
4.30

5.59‘I
7.06"

5.21"
4.57“
5.17“
5.25‘

4.18“

8.22“I

4.13"

123

Table A 14. Signiﬁcant F-tests ﬁom one-way analyses of variance between yield under
stress and 70 RAPD markers. One-way analyses of variance were performed between
RAPD marker genotypes and yield under stress ﬁ'om ﬁve locations (1990-1993) in
Michigan and Mexico and the combined analysis over all environments in the

 

 

Sierra/AC1028 population.

RAPD 1990 1991 1992 1993MT 1993z1 Combined
0.4161220 6.599 5.459 4.059
011116.50 5.229 11.84999 5.40
01912096., 5.779

0110456., 10.5899 4.409

0.407“, 6.059 3.829 8.0399

0A08 6.3999 3.489

0A09 7.1199

OABl44so 15.6444“:

019,318.50 4.559 8.5799

01181860., 4.709 5.679 4.769

011309.20 7.8799 17.309999 5.159
01801557,, 7.0399

01100275., 4.519 3.799
0AC021oo 15.869999

0AC081i4o 4.549

0.4151047, 5.339 9.2699

0C10 7.5999 4.109

0C15610 9.8899 23.979999 8.2099
01304 12.11999 4.499

01705760 10.81999

01905.4(, 18.119999

01706970 20.739999 7.6699
0012 5.069

0002.0", 7.1799 8.4699 4.739

0005610 4.18‘l

0005..., 4.299 8.1599

0608.2", 6.5099 4.539

060872, 10.03999

011121) 7.15"I 6.22"

0110111 4.369 9.7599

OH01c 7.08“

01103,“, 13.33999 4.069
01103.0", 9.4399

01106 9.3999 5.749 9.4199
011095., 4.089

0Ll2m, 6.17*

0L18 4.949

0N09 4.289 -

 

Table 14. Continued

124

 

 

RAPD 1990 1991 1992 1993M 19932 Combined
01116;)” 16.10”"

0120.60 7.36“

0W09a 14.95"“I 9.90”
0X01 22.45”" 6.93"
0X11950 1385""

0X18ggo 18.07“” 6.57

0X20 15.94““

0Y1 1610 5.46‘I

0Z011220 5.66‘ 4.94"

OZOlgmo 5.14" 11.65"" 8.57“
0201.00 20.31"""""I 5.05‘I
0Z031010 1553"“.

0208750 19.16""' 4.43‘

 

‘P < .05; ” P <.01; ”" P < .001, "“ P < .0001
TM = Madero, Dgo. location and Z = Calera, Zac. location

125

Table A; 15. Signiﬁcant F -tests ﬁom one-way analyses of variance between yield under
non-stress and 70 RAPD markers. One-way analyses of variance were performed between
RAPD marker genotypes and yield under non-stress from ﬁve locations (1990-1993) in
Michigan and Mexico and the combined analysis over all environments in the Sierra/Lef-

 

 

2RB population.

RAPD 1990 1991 1992 1993M1 1993ZT Combined
01116.50 10.07IMI 5.06"
0A04550 4.29"

0A0774o 4.19‘I

0A08 3.54"' 3.57‘'

0ABl4450 14.71"" 5.20“
0A309m 10.26"

0AC02750 7.72" 6.58“ 5.43"I
0AC027oo 15.53‘“ 7.39"
0AE01 5.78“ 4.68" 4. 17"
0AE10470 9.25“

0AE10350 5.70“

OC15510 9.16" 4.88"I
01705750 18.62"" 5.58‘I

01705440 6.77"

0F06970 6.79" 1439*" 5.18“
0001b 6.30"

0004330 10.77“"I 4.25“I
00054.0 4.17‘ 6.83"

0608721 4.44‘ 3.78“
011128 4.52‘

0H12b 4.27"

011018 5.23"l 13.04"" 4.32“
0H01b 6.25“"I

01101c 4.45"I

0H06 6.58"

0Ll2mgo 10.80‘"

0N09 7.59“I

011161190 8.16"

0W09a 8.56"" 10.61“I
0W09b 6.60"

0X01 35.74"“ 8.45"
03089.0 4.68“ 4.87‘

0X20 21.11"""""I 4.95"
CY] 1919 4.44‘

 

l 26
Table 15. Continued

 

 

RAPD 1990 1991 1992 1993Mf 19932t Combined
02011220 4.21"I

02011070 6.86" 4.35"
0201.00 11.70"“ 4.21‘I
0208750 16.13“" 7.26""

 

I"P < .05; " P (.01; n" P < .001, at“ P < .0001
1M = Madero, Dgo. location and Z = Calera, Zac. location

127

Table A 16. Signiﬁcant F-tests ﬁom one-way analyses of variance between geometric and
70 RAPD markers. One-way analyses of variance were performed between RAPD marker
genotypes and the geometric mean of yield under stress and non-stress from ﬁve locations
(1990-1993) in Michigan and Mexico and the combined analysis over all environments in
the Sierra/AC1028 population.

 

RAPD
0A161m
OA16350
0A20950
0A04550
0A07740
0A08
0A09
CAB 14450
CAB 18550
CAB 18500
0A309m
OAC 15570
OAC02760
OAC027m
OAEO 1
0AE10470
0C 10

0C 15610
0E04
OFO5760
01:05“)
01:06970
06 12
06021010
06054.0
00031210
011123
011121)
0H01a
0H0 1c
011031010
011031040
01106
0110950
015121050
0N09

1990

8.07”

6.14“

5.08"
4.42‘l

6.04“

7.63“”I

4.71"
4.89‘I
4.63‘I

4.33"

12.10"“I

1991
5.79‘
6.11"
6.76"
5.04“

5.62‘
20.00““

23.38“"
4.02“

16.40.00.

4.86".
5.35“

6.90“
6.57"“I

1992 1993M1
11.1399 5.049
8.2199

9.95“

10.74"
8.70""I
5.71"I
3.99‘I

4.06‘I

9.54"
6.98""I
1126‘I

4.16"
5.21"

6.74""l

6.83“

19932T
5.439
14.12999

3.42"

19.199999

6.39‘I
19.93""

11.62“”

9.23"

22.60“"
4.92"I

3.88“
9.29"

13.81""

10.32"I
7.07"

Combined

6.67‘

4.02"

4.79"

5.62"
6.56"

8.15"

8.20'"I

3.08“I

4.15“
3.68"

6.44“

128
Table A. 16. Continued

 

 

RAPD 1990 1991 1992 19931911I 19932 Combined
011161100 17.16“" 3.76
0W09a 13.24”‘ 12.87""'"'I
0W09b 5.34“

0X01 6.18 37.05“" 9.62“I
0X119so 8.98"

0X18900 1172"” 4.07

0X20 23.61"“ 5.0‘
0Y11610 6.59

02011220 7.49“ ‘
0Z011070 4.19‘| 10.87" 7.99“
0Z01m 20.04”“ 5.92‘
02031010 4.91‘I

0Z08750 22.83""‘ 7.20""I

 

‘P < .05; ” P (.01; “" P < .001, “n P < .0001
1M = Madero, Dgo. location and Z = Calera, Zac. location

129

Table A. 17. Split-plot analyses of variance for yield, biomass, 100 seed weight, number of
pods/plant, and days to maturity using data from the 1994 Kalamazoo County experiment.
Twenty-one genotypes ﬁ'om the Sierra/AC1028 population selected based on their marker
genotype were grown under stress and non-stress conditions. Included are orthogonal
contrasts between the marker-based, potentially above-average group (PAA) and the

potentially below average group (PBA).

 

 

 

 

 

 

 

 

Source DF MS F Test
Yield (g m").—

Blocks 3 14567.3
Stress Treatment 1 8642719 8171"“
Error 8 3 1057.7
Genotype 20 2284.9 2.79”

PAA vs. PBA 1 9497.8 11.2”
Genotype X Stress . 20 1450.02 1.79
Error 1 18 847.8

Biomass (g m") ----

Blocks 3 37342.4
Stress Treatment 1 2021617.4 7745""
Error 8 3 2610.4
Genotype 20 3579.9 1.6

PAA vs. PBA 1 4813.1 2.2
Genotype X Stress 20 3037.4 1.4
Error 120 2233.2

----100 Seed Weight (g) ----

Blocks 3 7.01
Stress Treatment 1 227.3 105.5"
Error 8 3 2.2
Genotype 20 60.0 29.0""

PAA vs. PBA 1 126.7 60.3"“
Genotype X Stress 20 4.3 2.1“I
Error 120 2.1

--—Number of Pods per Plant—~—

Blocks 3 58.5
Stress Treatment 1 1885.4 6243*"
Error 8 3 3.0
Genotype 20 10.0 2.2”

PAA vs. PBA 1 33.2 7.4"
Genotype X Stress 20 8.1 1.8‘
Error 120 4.5

-—---Days to Maturity—~—

Blocks 3 55.9
Stress Treatment 1 5.0 1.0
Error a 3 4.8
Genotype 20 35.0 5.0““
Genotype X Stress 20 4.5 0.6
Error 120 7.0

 

’P < .05; " P <.01; ""‘ P < .001, """ P < .0001

130

Table A 18. Grand mean and means for each genotypic group selected based on marker
genotype, LSD, CV and analysis of variance for yield resulting from analysis of the split-
plot design combined over locations conducted at Madero, Durango and Calera,
Zacatecas in 1994. Eleven genotypes ﬁ'om the Sierra/AC1028 population were selected as
potentially above average (PAA) and ten as potentially below average (PBA) based on

 

 

 

their marker genotypes.
Source DF MS F Test
--—-Yie1d (g m") -—-
Grand Mean 172.1
Mean (PAA) 168.9
Mean (PBA) 149.1
Diﬁ’erence 19.8“
LSD (at < .05) 43.7
CV (%) 18.3
Replication" l 22687.1
Treatment 1 6373006 8.4
Error a l 76059.5
Genotype 80 2331.9 2.4“"
PAA vs. PBA 1 741.4 0.2
Trtrnnt X Genotype 80 246.6 0.8
Error b 160 987.7

 

9P < .05; *9 P <.01; 9“ P < .001, ‘9'" P < .0001
1 Locations were designated as replications

131

Table A. 19. Grand mean and means for each genotypic group selected based on marker
genotype, LSD, CV and analysis of variance for yield resulting ﬁom analysis of the split-
plot design combined over locations conducted at Madero, Durango. and Calera
Zacatecas in 1994. Thirty-ﬁve genotypes from Sierra/Lef-ZRB were selected as potentially
above average (PAA) and 28 as potentially below average (PBA) based on their marker

genotypes.

 

 

 

 

Source DF MS F Test
Yield (g m”)——-—
GrandMean 146.5
Mean (PAA) 162.1
Mean (PBA) 142.6
Diﬁ‘erence 19.59"I
LSD (at < .05) 52.4
CV (%) 26.7
Replication? 1 l97784.8
Treatment 1 3396792 58.3
Error 8 1 5823.2
Genotype 99 3426.0 2.29999
High vs. Low 1 118985.] 7.899
Trtmnt x Genotype 99 833.9 0.5
Error b 198 1524.4

 

9P < .05; 99 P <.01; "'9 P < .001, 9“" P < .0001
t Locations were designated as replications

LIST OF REFERENCES

 

132

LIST OF REFERENCES

Acosta, J .A 1988. Selection of common bean (Phaseolus vulgaris) genotypes with
enhanced drought tolerance and biological nitrogen ﬁxation. PhD. diss. Michigan
State University, East Lansing (Diss. Abstr. 88-24816).

Acosta-G, J.A., and J.W. White. 1995. Phenological plasticity as an adaptation by
common bean to rainfed environments. Crop Sci. 35:199-204.

Acosta-Gallegos, J.A., and J.K.Shibata. 1989. Effect of water stress on growth and yield
of indeterminate common-bean (Phaseon vulgaris) cultivars. Field Crops
Research 20:81-93.

Allard, RW. 1960. Principles of plant breeding. John Wiley and sons, Inc., New York.
pp.92-98.

Beavis, W.D., 0.8. Smith, D.Grant, and R Fincher. 1994. Identiﬁcation of quantitative
trait loci using a small sample of top-crossed and F4 progeny ﬁ'om maize. Crop Sci.
34:882-896.

Brothers, ME, and JD. Kelly. 1993. Interrelationship of plant architecture and yield
components in the pinto bean ideotype. Crop Sci. 33: 1234-1238.

Bubeck, D.M., M.M. Goodman, W.D. Beavis, and D. Grant. 1993. Quantitative trait loci
controlling resistance to gray leaf spot in maize. Crop Sci. 33:83 8-847 .

Doerge, R.W., and GA Churchill. 1994. Issues in genetic mapping of quantitative trait
loci. In: Analysis of Molecular Marker Data. Proc. Joint Plant Breeding Symposia.
American Society for Horticulture Science Annual Meeting, Corvalis, 0R, pp. 15-
25.

Doerge, R.W., Z-B Zeng, and BS. Weir. 1994. Statistical issues in the search for genes
affecting quantitative traits in populations. In: Analysis of Molecular Marker Data.
Proc. Joint Plant Breeding Symposia. American Society for Horticulture Science
Annual Meeting, Corvalis, 0R, pp. 15-25.

133

Dudley, J .W. 1993. Molecular markers in plant improvement: Manipulation of genes
aﬂ‘ecting quantitative traits. Crop Sci. 33:660-668.

Edwards, M.D., C.W. Stuber, J.F. Wendel. 1987. Molecular-marker-facilitated
investigations of quantitative-trait loci in maize. 1. Numbers, genomic distribution
and types of gene action. Genetics 116:113-125.

Edwards, K.C., C. Johnstone, and C. Thompson. 1991. A simple and rapid method for the
preparation of plant genomic DNA for PCR analysis. Nucleic Acids Res. 19: 1349.

Edwards, M.D., T. Helentjaris, S. Wright, and CW. Stuber. 1992. Molecular-marker-
facilitated investigations of quantitative trait loci in maize. Theor. Appl. Genet.
83:765-774.

Ellis, T.H.N., L. Turner, R.P. Hellens, D. Lee, C.L. Harker, C. Enard, C. Domoney and
DR. Davies. 1992. Linkage maps in pea. Genetics 130:649-663.

Fischer, RA and R. Mauer. 1978. Drought resistance in spring wheat cultivars. 1. Grain
yield responses. Aust. J. Agric. Res. 29:897-912.

Foolad, M.R., R.A. Jones, and KL. Rodriguez. 1993. RAPD markers for constructing
intraspeciﬁc tomato genetic maps. Plant Cell Reports 12:293-297.

Freyre, R and BS. Douches. 1994a. Development of a model for marker-assisted
selection of speciﬁc gravity in diploid potato across environments. Crop Sci.
34:1361-1368.

Freyre, R. and BS. Douches. 1994b. Isoenzymatic identiﬁcation of quantitative traits in
crosses between heterozygous parents: Mapping tuber traits in diploid potato.
Theor. Appl. Genet. 87:764-772.

Haley, S.D., P.N. Miklas, J.R. Stavely, J. Byrum, and JD. Kelly. 1993. Identiﬁcation of
RAPD markers linked to a major rust resistance gene block in common bean.
Theor. Appl. Genet. 86:505-512

Haley S.D., L.K. Afanador, and JD. Kelly. 1994a. Identiﬁcation and application of RAPD
marker for the I gene (Potyvirus resistance) in common bean. Phytopathology
84:157-160.

Haley, S.D., L.K. Afanador, and JD. Kelly. 1994b. Selection for monogenic pest
resistance traits with coupling- and repulsion-phase RAPD markers. Crop Sci.
34: 1061-1066.

Haley, S.D., P.N. Miklas, L. Afanador, and JD. Kelly. 1994c. Random ampliﬁed
polymorphic DNA marker variability between and within gene pools of common
bean. J. Amer. Soc. Hort. Sci. 119:122-125.

 

134

Haley, S.D., P.N. Miklas, L.K. Afanador, J.R. Stavely, and JD. Kelly. 1994d.
Heterogeneous inbred populations are useful as sources of near-isogenic lines for
RAPD marker isolation. Theor. Appl. Genet. 88:337-342.

Jansen, RC. and P. Starn. 1994. High resolution of quantitative traits into multiple loci via
interval mapping. Genetics 136:1447-1455.

Kao, W.Y., M.P. Comstock and J .R Ehleringer. 1994. Variation in leaf movements
among common bean cultivars. Crop Sci. 34: 1273-1278.

Kelly, J .D. 1995. Use of random ampliﬁed DNA markers in breeding for major gene
resistance to plant pathogens. Hortscience 30:15-19.

Kramer, R]. 1988. Changing concepts regarding plant water relations. Plant, Cell and
Environment 11:565-568.

Lande, R and R Thompson. 1990. Efﬁciency of marker-assisted selection in the
improvement of quantitative traits. Genetics 124:743-756.

Lander, ES, and D. Botstein. 1989. Mapping Mendelian factors underlying quantitative
traits using RFLP linkage maps. Genetics 121 : 185-199.

Lander, E.S., P. Green, J. Abrahamson, A. Barlow, M.J. Daly, S.E. Lincoln, and L.
Newburg. 1987. MAPMAKER: An interactive computer package for constructing
primary genetic linkage maps of experimental and natural populations. Genomics
1:174-181.

Little, T.M. and F.J. Hills. 1978. Agricultural Experimentation. John “Wiley and Sons, New
York, pp. 65-76.

Ludlow, M.M., and RC. Muchow. 1990. A critical evaluation of traits for improving crop
yields in water-limited environments. Advances in Agronomy 43:107-153.

Lynch, J. and J .J van Beem. 1993. Grth and architecture of seedling roots of common
bean genotypes. Crop Sci. 33:1253-1257.

Martin, B., J. Nienhuis, G. King, and A Schaefer. 1989. Restriction ﬁagment length
polymorphisms associated with water use eﬂiciency in tomato. Science 243 : 1725-
1728. '

Melchinger, AE. 1990. Use of molecular markers in breeding for oligogenic disease
resistance. Plant Breeding Reviews 104:1-19.

Michelrnore, RW., I. Paran, and RV. Kesseli. 1991. Identiﬁcation of markers linked to
disease-resistance genes by bulked segregant analysis: A rapid method to detect

135

markers in speciﬁc genomic regions by using segregating populations. Proc. Natl.
Acad. Sci. USA 88:9829-9832.

Miklas, P.N., J .R Stavely, and J .D. Kelly. 1993. Identiﬁcation and potential use of a
molecular marker for rust resistance in common bean. Theor. Appl. Genet. 85:745-
749:

Moreno-Gonzalez, J. 1992. Genetic models to estimate additive and non-additive effects
of marker-associated QTL using multiple regression techniques. Theor. Appl.
Genet. 85:43 5-444.

Moreno-Gonzalez, J. 1993. Efﬁciency of generations for estimating marker-associated
QTL eﬁ‘ects by multiple regression. Genetics 135:223-231.

Nodari, R.0., E.M.K Koinange, J.D. Kelly, P. Gepts. 19928. Towards an integrated
linkage map of common bean. I. Development of genomic DNA probes and levels
of restriction ﬁagment length polymorphism. Theor. Appl. Genet. 84:186-192.

Nodari, RO, S.M. Tsai, RL. Gilbertson, P. Gepts. 1992b. Towards an integrated linkage
map of common bean. II. Development of an RFLP-based linkage map. Theor.
Appl. Genet. 85:513-520. '

Passioura, J.B. 1988. Response to Dr. P.J. Kramer’s article, “Changing concepts regarding
plant water relations’, volume 11, number 7, pp. 565-568. Plant, Cell and
Environment 11:569-571.

Paterson, AH. S. Damon J.D. Hewitt, D. Zamir, H.D. Rabinwitch, S.E. Lincoln, E.S.
Lander and SD. Tanksley. 1991. Mendelian factors underlying quantitative traits
in tomato: Comparison across species, generations and environments. Genetics
127:181-197.

Paterson, AH., E.S. Lander, J.D. Hewitt, S. Peterson, S.E. Lincoln, and SD. Tanksley.
1988. Resolution of quantitative traits into Mendelian factors by using a complete
linkage map of restriction fragment length polymorphisms. Nature 335:721-726.

Paterson, AH., JW. DeVerna, B. Lanini, S.D. Tanksley. 1990. Fine mapping of
quantitative trait loci using selected overlapping recombinant chromosomes, in an
interspecies cross of tomato. Genetics 1242735-742.

Ramirez, P. 1992. Identiﬁcation and estimation of heritabilities of drought related
resistance traits in common bean (Phaseolus vulgaris). Ph.D. diss. Michigan State
University, East Lansing (Diss. Abstr. 92-26240).

Rosielle, AA and J. Hamblin. 1981. Theoretical Aspects of Selection for Yield in Stress
and Non-Stress Environments. Crop Sci. 21:943-945.

136

Saghai-Maroof, M.A., K.M. Soliman, RA Jorgensen, R.W. Allard. 1984. Ribosomal
DNA spacer-length polymorphisms in barley: Mendelian inheritance, chromosomal
location, and population dynamics. Proc. Natl. Acad. Sci. 81:8014-8018.

Scully, B.T., D.H. Wallace, and DR Viands. 1991. Heritability and correlation of
biomass, grth rates, harvest index and phenology to the yield of common beans.
J. Amer. Soc. Hort. Sci. 116:127-130.

Singh, SP. 1995. Selection for water-stress tolerance in interracial populations of
common bean. Crop Sci. 35:118-124.

Soller, M., and J .S. Beckmann. 1990. Marker-based mapping of quantitative trait loci
using replicated progenies. Theor. Appl. Genet. 80:205-208.

Stromberg, L.D., J.W. Dudley, and GK. Rufener. 1994. Comparing conventional early
generation selection with molecular marker assisted selection in maize. Crop Sci.
34:1221-1225.

Stuber, CW. and MD. Edwards. 1986. Genotypic selection for improvement of
quantitative traits in corn using molecular marker loci. In: Proc. 4lst Annual Corn
and Sorghum Research Conference. American Seed Trade Association,
Washington, DC, pp. 70-83.

Stuber, C.W., M.D. Edwards, and IF. Wendel. 1987. Molecular marker-facilitated
investigations of quantitative trait loci in maize. 11. Factors inﬂuencing yield and its
component traits. Crop Sci. 27:639-648.

Stuber, C.W., S.E. Lincoln, D.W. Wolﬁ‘, T. Helentjaris and ES. Lander. 1992.
Identiﬁcation of genetic factors contributing to heterosis in a hybrid from two elite
inbred lines using molecular markers. Genetics 132:823 -83 9.

Tanksley, SD. 1983. Molecular markers in plant breeding. Plant Molecular Biology
Reporter Vol. 1:1, pp 3-8.

Welsh, J. and M. McClelland. 1990. Fingerprinting genomes using PCR with arbitrary
primers. Nucleic Acids Research 18:7213-7218.

White, J .W. and J .A. Castillo. 1989. Relative effect of root and shoot genotypes on yield
of common bean under drought stress. Crop Sci. 29:360-362.

White, J.W. and J .A Castillo. 1992. Evaluation of diverse shoot genotypes on selected
root genotypes of common bean under soil water deﬁcits. Crop Sci. 32:762-765.

White, J .W. 1993. ‘Implications of carbon isotope discrimination studies for breeding
common bean under water deﬁcits” in Stable Isotopes and Plant Carbon-Water
Relations. Academic Press, Inc. 387-398.

137

White, J .W., J .A Castillo, J .R. Ehleringer. 1990. Associations between productivity, root
growth and carbon isotope discrimination in Phaseon vulgaris under water
deﬁcit. Australian Journal of Plant Pathology 17:189-198.

White, J .W., J.A. Castillo, JR Ehleringer, J .A. Garcia-C and SP. Singh. 1994a. Relations
of carbon isotope discrimination and other physiological traits to yield in common
bean (Phaseolus vulgaris) under rainfed conditions. Journal of Agricultural
Science, Cambridge 1222275-284.

White, J.W., R. oohoa M., F. Ibarra P., and SP. Singh. 1994b. Inheritance of seed yield,
maturity and seed weight of common bean (Phaseolus vulgaris) under semi-arid
rainfed conditions. Journal of Agricultural Science, Cambridge 122:265-273.

Williams, J.G.K., AR. Kubelik, K.J. Livak, J.A. Rafalski and S.V. Tingey. 1990. DNA
polymorphisms ampliﬁed by arbitrary primers are useful as genetic markers.
Nucleic Acids Research 18:6531-6535.

Zehr, a. 1990. Use of RFLP markers in maize as an aid in selection during inbreeding.
Ph.D. diss. University of Illinois. (Diss. Abstr. no. N45-27).

Zeng, Z. 1994. Precision mapping of quantitative trait loci. Genetics 136: 1457-1468.

"11.111111111111411“