c Indian Academy of Sciences Wheat kernel dimensions: how do they contribute to kernel weight at an individual QTL level?

Abstract

Kernel dimensions (KD) contribute greatly to thousand-kernel weight (TKW) in wheat. In the present study, quantitative trait loci (QTL) for TKW, kernel length (KL), kernel width (KW) and kernel diameter ratio (KDR) were detected by both con-ditional and unconditional QTL mapping methods. Two related F 8:9 recombinant inbred line (RIL) populations, comprising 485 and 229 lines, respectively, were used in this study, and the trait phenotypes were evaluated in four environments. Uncon-ditional QTL mapping analysis detected 77 additive QTL for four traits in two populations. Of these, 24 QTL were verified in at least three trials, and five of them were major QTL, thus being of great value for marker assisted selection in breeding programmes. Conditional QTL mapping analysis, compared with unconditional QTL mapping analysis, resulted in reduction in the number of QTL for TKW due to the elimination of TKW variations caused by its conditional traits; based on which we first dissected genetic control system involved in the synthetic process between TKW and KD at an individual QTL level. Results indicated that, at the QTL level, KW had the strongest influence on TKW, followed by KL, and KDR had the lowest level contribution to TKW. In addition, the present study proved that it is not all-inclusive to determine genetic relationships of a pairwise QTL for two related/causal traits based on whether they were co-located. Thus, conditional QTL mapping method should be used to evaluate possible genetic relationships of two related/causal traits.

Full text

Page 1

c ? Indian Academy of Sciences
RESEARCH ARTICLE
Wheat kernel dimensions: how do they contribute to kernel weight
at an individual QTL level?
FA CUI1,2†, ANMING DING1†, JUN LI1,3†, CHUNHUA ZHAO1†, XINGFENG LI1, DESHUN FENG1,
XIUQIN WANG4, LIN WANG1,5, JURONG GAO1and HONGGANG WANG1∗
1State Key Laboratory of Crop Biology, Shandong Key Laboratory of Crop Biology, Taian Subcenter of National Wheat
Improvement Center, College of Agronomy, Shandong Agricultural University, Taian 271018, People’s Republic of China
2Center for Agricultural Resources Research, Institute of Genetics and Developmental Biology, Chinese Academy
of Sciences, Shijiazhuang, 050021, Hebei, People’s Republic of China
3Tianxing Biotechnology, Handian Industrial Zone, 256200 Zouping, Shandong, People’s Republic of China
4Municipal Academy of Agricultural Sciences, Zao’zhuang 277100, Shandong, People’s Republic of China
5Municipal Academy of Agricultural Sciences, Ji’ning 272031, Shandong, People’s Republic of China
Abstract
Kernel dimensions (KD) contribute greatly to thousand-kernel weight (TKW) in wheat. In the present study, quantitative trait
loci (QTL) for TKW, kernel length (KL), kernel width (KW) and kernel diameter ratio (KDR) were detected by both con-
ditional and unconditional QTL mapping methods. Two related F8:9recombinant inbred line (RIL) populations, comprising
485 and 229 lines, respectively, were used in this study, and the trait phenotypes were evaluated in four environments. Uncon-
ditional QTL mapping analysis detected 77 additive QTL for four traits in two populations. Of these, 24 QTL were verified
in at least three trials, and five of them were major QTL, thus being of great value for marker assisted selection in breeding
programmes. Conditional QTL mapping analysis, compared with unconditional QTL mapping analysis, resulted in reduction
in the number of QTL for TKW due to the elimination of TKW variations caused by its conditional traits; based on which
we first dissected genetic control system involved in the synthetic process between TKW and KD at an individual QTL level.
Results indicated that, at the QTL level, KW had the strongest influence on TKW, followed by KL, and KDR had the lowest
level contribution to TKW. In addition, the present study proved that it is not all-inclusive to determine genetic relationships of
a pairwise QTL for two related/causal traits based on whether they were co-located. Thus, conditional QTL mapping method
should be used to evaluate possible genetic relationships of two related/causal traits.
[Cui F., Ding A., Li J., Zhao C., Li X., Feng D., Wang X., Wang L., Gao J. and Wang H. 2011 Wheat kernel dimensions: how do they
contribute to kernel weight at an individual QTL level? J. Genet. 90, 409–425]
Introduction
Wheat (Triticum aestivum L.) is a major food crop world-
wide, and high yield is a predominant objective in breed-
ing programmes. Kernel weight, one of the three major
yield components, is greatly influenced by kernel dimensions
(KD), such as kernel length (KL), kernel width (KW), etc.
Therefore, it is of utmost interest to obtain more information
about the underlying genetic control of KD traits.
With the rapid development of molecular marker tech-
nology in wheat, increasing numbers of QTL studies have
been conducted in an attempt to dissect the genetic basis of
∗For correspondence. E-mail: hgwang@sdau.edu.cn.
†These authors contributed equally to this work.
thousand-kernel weight (TKW), and all the 21 wheat chro-
mosomes have now been proven to harbour factors affect-
ing it (Halloran 1976; Giura and Saulescu 1996; Araki et al.
1999; Shah et al. 1999; Kato et al. 2000; Varshney et al.
2000; Ammiraju et al. 2001; Zanetti et al. 2001; Böner
et al. 2002; Campbell et al. 2003; Groos et al. 2003; Huang
et al. 2003, 2004, 2006; McCartney et al. 2005; Verma et al.
2005; Kirigwi et al. 2007; Li et al. 2007; Hai et al. 2008;
Röder et al. 2008; Golabadi et al. 2010; McIntyre et al.
2010; Su et al. 2010; Zheng et al. 2010). To obtain infor-
mation about genetic relationships between TKW and KD at
the QTL level, some researchers performed QTL detection
for both TKW and KD simultaneously in the same biparental
mapping populations (Campbell et al. 1999; Dholakia et al.
2003; Breseghello and Sorrells 2007; Sun et al. 2009; Ramya
Keywords. wheat; kernel dimensions; thousand-kernel weight; conditional QTL mapping; genetic relationship.
Journal of Genetics, Vol. 90, No. 3, December 2011
409

Page 2

Fa Cui et al.
et al. 2010; Tsilo et al. 2010). In these studies, a common
problem associated with the analyses of the data reported so
far is that the analysis of QTL for TKW and its causal traits
was conducted based on the phenotypic values separately.
Generally, they determined genetic associations of two QTL
for pairwise traits based on whether they were co-located.
Due to the interference of other traits, it should be difficult
to clarify the actual relationships among these traits precisely
and comprehensively at the QTL level (Ye et al. 2009).
More recently, a method for multivariate conditional anal-
ysis was proposed for analysing the contributions of compo-
nent traits to a complex trait and for investigating the genetic
relationship between two traits at the QTL level (Wen and
Zhu2005).Basedonthismethodology,itispossibletoreveal
that the gene expression for a complex trait may be con-
tributed by its different causal genes expression at different
levels, thus helping us in understanding the nature of compo-
nent traits in determining the phenotypic value of a complex
trait. To our knowledge, few reports of QTL analysis based
on this methodology have been reported (Guo et al. 2005;
Zhao et al. 2006; Mei et al. 2007; Li et al. 2008; Liu et al.
2008a; Ye et al. 2009; Cui et al. 2011). However, none of
them considered TKW and its components in wheat. In view
of the strong positive correlations between TKW and KD, it
will be of great value to probe their genetic control system
and to evaluate their genetic relationships on individual loci
by conditional QTL mapping method.
Population size has a great effect on the estimation of QTL
number and genetic effects (Beavis 1998; Mackay 2001;
Bernardo 2004; Schön et al. 2004; Vales et al. 2005; Zou
et al. 2005; Buckler et al. 2009). The precision and efficiency
of QTL detection will be enhanced by combining at least two
related populations (Breseghello and Sorrells 2007; Kumar
et al. 2007; Ma et al. 2007; Buckler et al. 2009; Uga et al.
2010; Gegas et al. 2010). In the present study, we performed
QTL detection for TKW, KL, KW and kernel diameter ratio
(KDR) based on the combination of two related populations,
of which one was a large population with up to 485 lines and
the other was smaller comprising 229 lines. Both uncondi-
tional mapping methods and conditional mapping methods
for multivariate conditional analysis were used. The objec-
tives of this study were to: (i) accurately identify the genetic
factors affecting TKW, KL, KW and KDR; (ii) specify the
genetic relationships between TKW and KD at individual
QTL level; and (iii) discuss the effect of combining two
related populations of different size on the efficiency and
precision of QTL detection.
Materials and methods
Experimental populations and their evaluation
Two related wheat F8:9recombinant inbred line (RIL) map-
ping populations were used in this study, which will be
referred here as populations ‘WJ’ and ‘WY’. WJ was derived
from the cross between Weimai 8 and Jimai 20, compris-
ing 485 lines. WY consisted of 229 lines, derived from the
cross between Weimai 8 and Yannong 19. The common par-
ent Weimai 8 is a large-spike type of the ideotype model
and was released by Weifang Municipal Academy of Agri-
cultural Sciences, Shandong, China, in 2003; Jimai 20 and
Yannong 19, two superior quality wheat varieties, are multi-
panicle types,and they werereleased by Crop Research Insti-
tute, Shandong Academy of Agricultural Sciences, China, in
2003, and by Yantai Municipal Academy of Agricultural Sci-
ences, Shandong, China, in 2001, respectively. Among these
three parental varieties, Weimai 8 has the highest TKW and
the largest kernels (table 1). In addition, the common par-
ent Weimai 8 is a 1BL/1RS translocation line whereas the
other two parents have a common 1B chromosome. The par-
ents together with the RILs, were planted in four environ-
ments in Shandong Province, China; Tai’an in 2008–2009
(E1), and Tai’an in 2009–2010 (E2), Zao’zhuang in 2009–
2010 (E3) and Ji’ning in 2009–2010 (E4). This study sites
at Tai’an (36◦09?N, 117◦09?E, altitude 128 m), Zao’zhuang
(36◦50?N, 117◦33?E, altitude 65 m) and Ji’ning (35◦27?N,
116◦35?E, altitude 37 m) represent three different wheat-
growing agroclimatic regions. The temperature in winter was
lower in Tai’an in 2008 than 2009; and the wind in June was
stronger in Tai’an in 2009 than 2010. A two-row plot with
rows 2-m long and 30 cm apart was used, and 50 seeds were
planted in each row. Since large variation in plant height
existed among RIL populations in both WJ and WY (Cui
et al. 2011), the RIL populations were planted in adjacent
plots according to plant heights to avoid severe shading by
adjacent plants which could affect a shorter plant surrounded
by tall plants. The RILs together with parents were sown
on 7th October 2008 for E1, 8th October 2009 for E2, 11th
October 2009 for E3 and 17th October for E4, respectively.
Concerning anthesis data, small variation existed among RIL
populations in both WJ and WY, the range being about five
days. All recommended agronomic practices were followed
in all the four experiments, and all the experimental fields
had loamy soil. However, gibberellic disease was very seri-
ous in Zao’zhuang in 2009–2010 in the watery stage and
milk ripe stage. Kernel traits were evaluated after harvest.
Seed was thoroughly cleaned and all nonwheat materials
and broken kernels were removed before trait evaluation.
TKW was evaluated in grams by weighing two samples of
1000 kernels from each plot. Two samples of 20 kernels
from each plot were lined up length-wise along a ruler with
a precision of 0.1 mm, to measure KL, and then the ker-
nels were arranged breadth-wise to measure KW. All lengths
were reported in centimetres. The KDR was calculated as
KDR = KL/KW.
Analysis of molecular and biochemical markers
Molecular markers of G-SSR, EST-SSR, ISSR, STS, SRAP
and RAPD were used to genotype the three parents
and their derived lines. Of these, relevant information
410
Journal of Genetics, Vol. 90, No. 3, December 2011

Page 3

QTL for wheat kernel weight
Table 1. Phenotypic values for thousand-kernel weight and kernel dimensions of three parents and two RIL populations in four growing environments in wheat.
Parent
WJc
WYc
Traita
En.b
Weimai 8
Jimai 20
Yannong 19
Mean
SD
Min–max
Skewness
Kurtosis
Mean
SD
Min–max
Skewness
Kurtosis
TKW (g)
E1
49.11
28.53
44.68
40.95
5.504
19.59–55.33
−0.203
0.6442
41.68
5.039
23.41–51.90
−0.731
0.582
E2
50.46
32.43
42.88
39.45
4.843
23.89–53.96
0.133
0.041
40.56
3.842
30.95–53.12
0.031
0.172
E3
35.41
28.70
28.71
30.37
4.26
16.36–42.34
−0.166
−0.244
30.67
4.049
19.22–42.65
0.088
0.022
E4
49.00
30.12
41.28
43.09
19.16
25.00–54.57
−0.244
0.559
42.60
4.665
30.58–56.89
−0.112
−0.241
KL (cm)
E1
0.707
0.667
0.680
0.666
0.032
0.588–0.753
0.109
−0.248
0.690
0.033
0.620–0.780
0.369
−0.130
E2
0.710
0.637
0.680
0.642
0.033
0.530–0.752
0.277
0.311
0.664
0.037
0.562–0.805
0.281
0.740
E3
0.645
0.575
0.631
0.603
0.037
0.450–0.700
−0.637
1.658
0.629
0.031
0.555–0.751
0.692
1.145
E4
0.680
0.641
0.676
0.645
0.034
0.465–0.765
−0.329
1.723
0.667
0.036
0.570–0.765
0.053
−0.241
KW (cm)
E1
0.360
0.313
0.327
0.332
0.020
0.247–0.380
−0.554
1.182
0.330
0.019
0.240–0.373
−0.625
1.447
E2
0.360
0.299
0.320
0.324
0.023
0.251–0.400
−0.040
0.402
0.322
0.020
0.265–0.396
0.032
0.504
E3
0.333
0.281
0.295
0.314
0.027
0.250–0.490
2.20
11.67
0.307
0.018
0.258–0.366
−0.019
0.213
E4
0.363
0.330
0.318
0.342
0.018
0.275–0.395
−0.258
0.487
0.336
0.019
0.266–0.395
−0.226
0.654
KDR
E1
1.963
2.128
2.082
2.011
0.143
1.691–2.756
0.998
2.980
2.098
0.141
1.788–2.610
0.469
0.561
E2
1.972
2.130
2.125
1.991
0.142
1.560–2.498
0.308
0.410
2.071
0.161
1.744–2.548
0.385
−0.197
E3
1.940
2.048
2.139
1.941
0.169
1.010–2.470
−1.259
7.809
2.058
0.149
1.764–2.612
0.777
0.493
E4
1.874
1.944
2.126
1.889
0.120
1.606–2.301
0.476
0.234
1.991
0.156
1.603–2.496
0.399
−0.037
SD, standard deviation.aTKW, thousand-kernel weight; KL, kernel length; KW, kernel width; KDR, kernel diameter ratio.
bE1, E2, E3 and E4 represent the environments of 2008–2009 in Taian, 2009–2010 in Taian, 2009–2010 in Zaozhuang and 2009–2010 in Jining, respectively.
cWJ and WY represent the populations derived from the cross between Weimai 8 and Jimai 20 and between Weimai 8 and Yannong 19, respectively.
regarding G-SSR markers, including BARC, CFA, CFD,
CFT, GWM, GDM, GPW, WMC and PSP codes, as well
as PCR-based STS markers of the MAG code, were taken
from the GrainGenes web site (http://wheat.pw.usda.gov).
Relevant information about EST-SSR markers prefixed
CFE, KSUM and CNL are publicly available (http://wheat.
pw.usda.gov/ITMI/EST-SSR/). EST-SSR markers of SWES
and WW codes were developed and kindly provided by Pro-
fessor Sishen Li, College of Agronomy, Shandong Agricul-
tural University, China. EST-SSR markers with the prefixes
CWEM, EDM and CWM were published in reference arti-
cles by Peng and Lapitan (2005), Mullan et al. (2005) and
Gao et al. (2004), respectively. ISSR markers were devel-
oped by the University of British Columbia Biotechnol-
ogy Laboratory (UBCBL) (Nagaoka and Ogihara 1997).
Relevant information about chromosome 1RS-specific mark-
ers of rye were detailed by Zhao et al. (2009), and func-
tional markers, were detailed by Liu et al. (2008b) and
Liang et al. (2010). The differences of high molecular weight
glutenin subunits (HMW-GS) at Glu-a1, Glu-b1 and Glu-d1
between the parents were detected and used as biochemical
markers.
Each PCR reaction for G-SSR, EST-SSR and PCR-based
STS markers was conducted in a total volume of 25 μL in
a TakaRa PCR thermal cycler (Takara, Shiga, Japan) or in
a Bio-Rad 9600 thermal cycler (Bio-Rad, California, USA).
PCR reaction mixture was compounded according to the
formula described by Röder et al. (1998). Amplifications
were performed using a touchdown PCR protocol detailed
by Hao et al. (2008). PCR reaction mixture, as well as
PCR protocol for SRAP and ISSR markers followed the for-
mula and the procedure detailed by Li et al. (2007), and for
RAPD markers, by Suenaga et al. (2005). The PCR products
were separated in 6% nondenaturing polyacrylamide gels.
Gels were then silver stained and photographed. Types of
HMW-GS were detected by using sodium dodecyl sulphate
polyacrylamide gel electrophoresis (SDS-PAGE) (Singh and
Sheperd 1991). Markers of BARC, CFA, CFD, GWM, GDM
and WMC codes were also screened on the nullisomic and
tetrasomic stocks of Chinese Spring (CSNT) to assign them
to chromosomes, where possible.
Construction of the genetic linkage map
Linkage groups were constructed by MAPMAKER 3.0
(Lander et al. 1987). First, the ‘ANCHOR’ command
was used to locate marker loci on chromosomes based
on the CSNT identification and the public genetic maps
in GrainGenes 2.0 (http://wheat.pw.usda.gov/GG2/index.
shtml). Then, the assignment of the remaining loci to chro-
mosomes was made using the ‘ASSIGN’ command at a
LOD score of 3.0. Based on the linkage group defined
above, JoinMap, version 3.0 (Biometris, Wageningen, The
Netherlands; http://www.joinmap.nl), was used to construct
the linkage map, and centiMorgan units were calculated
using the Kosambi mapping function (Kosambi 1944).
Journal of Genetics, Vol. 90, No. 3, December 2011
411

Page 4

Fa Cui et al.
Data analysis and QTL mapping
To estimate the variance components, all the four traits
were first analysed with the MINQUE method proposed
by Zhu (1992). Phenotypic correlation coefficients between
TKW and TKW components (TKWC) were calculated sep-
arately for each environment. Basic statistical analysis was
implemented by the software SPSS13.0 (SPSS, Chicago,
USA). Conditional genetic analysis was conducted based
on the phenotypic values of TKW conditioned on each
of its component traits, that were obtained by the mixed-
model approach (Zhu 1995; Wen and Zhu 2005). Conditional
phenotypic values y(TKW|TKWC)indicate the value of TKW
without the influences of its component traits.
Both the observed phenotypic values (y(TKW)) and the con-
ditional phenotypic values (y(TKW|TKWC)) obtained from each
environment of E1, E2, E3 and E4, and the pooled data
collected from the average of the four environments above
(P) were used for QTL mapping analysis. In addition, phe-
notypic values of KL, KW and KDR obtained from each
environment and the pooled data were also used for QTL
scan, with a view to analyse pleiotropic effects between
TKW and its related traits at the QTL level. QTL screen
was conducted using inclusive composite interval mapping
by IciMapping 3.0 based on step-wise regression of simulta-
neous consideration of all marker information (http://www.
isbreeding.net/). The walking speed chosen for all QTL was
1.0 cM. The threshold LOD scores were calculated using
1000 per mutations. However, here we ignored the QTL with
a LOD value of <2.5 to make the QTL reported authentic
and reliable.
Rules for naming QTL
The assignment of a QTL name is named according to the
following rules: italic upper case ‘Q’ denotes ‘QTL’; let-
ters following it are the abbreviation of the corresponding
trait; the next upper case letters sandwiched the two dashes
‘-’ indicates the population in which the corresponding QTL
was detected; next, a numeral plus an upper case letter, ‘A’,
‘B’ or ‘D’, indicates the wheat chromosome on which the
corresponding QTL was detected; if a break occurred on a
chromosome, a dash ‘-’ plus a numeral are placed as suffixes
to distinguish different segments of the corresponding chro-
mosome; the last numeral after a period denotes the num-
ber of environments in which the corresponding QTL was
detected; and if the name of two different QTL for the same
trait look the same, a lower case letter is used to distinguish
them.
Results
Phenotypic variation of traits and correlations with TKW
The parental performance and variation among the two RIL
lines for TKW, KL, KW and KDR in four environments
are shown in table 1. Over all in four environments, the
significant differences of all the four traits existed at the 0.05
level between Weimai 8 and Jimai 20, and between Weimai 8
and Yannon 19. Weimai 8 was characterized by higher TKW
and larger kernels, compared to both Jimai 20 and Yannong
19. In addition, both RIL lines and parents had the lowest
phenotypic values for both TKW and KD in E3 among all
the four trials in both WJ and WY. This might be due to the
outbreak of gibberellic disease in Zao’zhuang in 2009–2010
in the watery stage and milk ripe stage. The phenotypic vari-
ations of all the four traits among the RIL lines were obvious
in both populations and segregated continuously. Both abso-
lute values of skewness and kurtosis for TKW were less than
1.0 in all the four trials in both WJ and WY, indicating a nor-
mal distribution. The absolute values of skewness for KL,
KW and KDR were less than 1.0, with the exception of KW
in E3 and KDR in E3. The absolute values of skewness for
KL were less than 1.0 in E1 and E2, as that of KW in E2 and
E4 and of KDR in E2 and E4. The results indicate that all
the four traits were typically quantitative traits controlled by
a few minor or major genes and that the data are suitable for
QTL analysis. In addition, strong transgressive segregations
were observed in all the four traits in all environments, indi-
cating that alleles with positive effects are distributed among
the parents. The evaluation of the phenotypic correlations
between TKW and KL, KW and KDR are listed in table 2.
Positive and significant correlations were observed between
TKW and KL, and between TKW and KW, in both WJ and
WY in all environments, whereas negative significant corre-
lations were observed between TKW and KDR. The highest
correlation coefficients in absolute were observed between
TKW and KW, and the lowest were between TKW and KDR.
Overall, the results were consistent in four environments in
both WJ and WY. These findings indicate a strong stable
genetic association between TKW and KW. In both WJ and
WY, conditioning TKW on KW led to the strongest reduction
of phenotypic variance, while TKW conditioned on KDR
showed variances nearly as high as the unconditioned TKW
(table 2). These findings indicate that KW contributes to
the highest level of TKW phenotypic variation, next to KL,
consistent with the results of phenotypic correlation analysis.
Construction of genetic linkage maps
The genetic map constructed based on the WJ population
included 344 loci on the wheat chromosomes and spanned
2855.5 cM, with an average density of one marker per
8.30 cM. There were six linkage gaps with linkage dis-
tances >50 cM (figure 1). Marker distribution ranged from
45 on chromosome 4A to 3 on chromosomes 4D and 7D.
The WY population was used to establish a genetic map
consisting of 358 loci distributed in 27 linkage groups with
six linkage gaps, and it covered 3010.70 cM of the whole
genome with an average distance of 8.41 cM between the
adjacent loci (figure 1). The number of markers per chro-
mosome ranged from 40 on chromosome 1B to 3 on chro-
mosome 3D. The two linkage maps contained 69 common
412
Journal of Genetics, Vol. 90, No. 3, December 2011

Page 5

QTL for wheat kernel weight
Table 2. Phenotypic correlations between thousand-kernel weight and kernel dimensions and the phenotypic variances of thousand-kernel
weight and thousand-kernel weight conditioned on kernel dimensions.
CorrelationDirected and conditioned variances
E1E2TraitE1E2 E3 E4E3E4
TKW
KL
KW
KDR −0.378∗∗/−0.288∗∗−0.273∗∗/−0.189∗∗−0.275∗∗/−0.237∗∗−0.111∗/−0.141∗25.96/23.28 21.80/14.16 16.45/15.15 18.86/21.62
For each entry, the first figure refers to WJ, and the second to WY. For abbreviations, see table 1.
∗Correlation is significant at when P < 0.01 level;∗∗correlation is significant at when P < 0.01 level; –/–, data not available.
–/––/––/– –/– 30.29/25.39 23.46/14.76 18.15/16.39 19.16/21.76
0.458∗∗/0.331∗∗24.81/21.73 17.80/13.47 15.07/14.20 15.09/19.65
0.577∗∗/0.510∗∗14.41/15.73 14.09/11.32 11.84/10.82 12.74/16.32
0.425∗∗/0.380∗∗
0.724∗∗/0.617∗∗
0.495∗∗/0.289∗∗
0.618∗∗/0.479∗∗
0.368∗∗/0.340∗∗
0.555∗∗/0.571∗∗
loci. The chromosomal locations and the orders of the mark-
ers in the two maps were generally in agreement with pub-
lished reports in GrainGenes 2.0 (http://wheat.pw.usda.gov/
GG2/index.shtml). Positions of the loci common to the
two maps were approximately in accordance. In addition, a
1BL/1RS translocation event was confirmed by the linkage
maps of chromosome 1B in both WJ and WY. Functional
markers and biochemical markers were accurately mapped to
their corresponding chromosomes. The recommended map
distance for genomewide QTL scanning is an interval length
less than 10 cM (Doerge 2002). Thus, the maps were suitable
for genomewide QTL scanning in this study.
Unconditional QTL mapping in WJ and WY populations
Thousand-kernel weight: In total, 14 and nine putative addi-
tive QTL for TKW were detected in WJ and WY, respec-
tively (see tables 1 and 2 in electronic supplementary
material at http://www.ias.ac.in/jgenet/; figure 1). They
together covered all the 21 wheat chromosomes except
1B, 1D, 3A, 4B, 4D and 5A. Of these, QTkw-WJ-4A.5 and
QTkw-WJ-6A.5 were verified in all the five trials, and they
individually accounted for 2.36–5.29% and 2.32–5.94% of
the phenotypic variance, respectively. QTkw-WJ-5B.4 was
identified reproducibly in four of the five trials, exhibiting
3.35–4.75% of the phenotypic variance. Of the five QTL
that showed significance in three trials, QTkw-WY-2A.3 and
QTkw-WY-2B-1.3 individually accounted for 8.01–12.60%
and 8.36–14.79% of the phenotypic variance, respectively,
both being major QTL. The remaining 15 QTL showed sig-
nificance in only two or one of the five trials. Alleles of QTL
with increased effect were identified from both parents in
both WJ and WY.
Kernel length: QTL mapping detected 12 and seven chromo-
somal regions governing KL totally in the five trials in WJ
and WY, respectively (see tables 1 and 2 in electronic sup-
plementary material; figure 1). These QTL were located on
1A, 1B, 1D, 2A, 2D (2 QTL), 3B, 5A, 5B, 6A and 6D in WJ,
and on 3A, 6A (2 QTL), 6B (2 QTL) and 6D (2 QTL) in WY.
Of these, only QKl-WJ-3B.5 was stable across all the five
trials, explaining 3.94–8.18% of the phenotypic variance.
Two QTL, QKl-WJ-5B.4 and QKl-WY-6B.4, were stable
across four of the five trials, individually exhibiting 3.51–
6.36% and 7.78–11.64% of the phenotypic variance, respec-
tively. In addition, there were three QTL for KL that
showed significance in three of the five trials, all individ-
ually accounting for less than 10% of the phenotypic vari-
ance. The remaining 13 QTL were significant in only two
trials or one, of which, only QKl-WY-6B.2 was major QTL,
accounting for 6.80–12.95% of the phenotypic variance. The
additive effects for seven QTL were positive with Weimai 8
increasing the QTL effects in WJ, but Jimai 20 contributed
the favourable alleles for all the seven QTL in WY.
Kernel width: For KW, 13 putative additive QTL in WJ and
nine in WY were detected, respectively (see tables 1 and 2
in electronic supplementary material; figure 1). They were
together assigned to 14 wheat chromosomes, all the 21 wheat
chromosomes with the exception of 1B, 4A, 4D, 5B, 6D, 7A
and 7D. Of these, only two QTL, QKw-WJ-5A-3.3 and QKw-
WJ-6A.4, were reproducibly identified in at least three trials,
each exhibiting 5.17–7.83% and 3.45–6.56% of the pheno-
typic variance, respectively. Of the remaining 20 QTL that
were significant in only two trials or one, there were four
QTL individually accounting for more than 10% of the phe-
notypic variance. Favourable alleles for KW were dispersed
among the parents in both WJ and WY.
Kernel diameter ratio: Concerning KDR, up to 25 putative
additive QTL were detected in two populations, distributed
across 13 of the 21 wheat chromosomes (see tables 1 and
2 in electronic supplementary material; figure 1). Of these,
QKdr-WY-6B.5, explaining 5.24–13.97% of the phenotypic
variance, was reproducibly identified in all the five trials.
Two QTL, QKdr-WJ-2A.4 and QKdr-WJ-5A-1.4, were sta-
ble across four of the five trials, individually exhibiting
2.96–4.19% and 13.49–15.08% of the phenotypic variance,
respectively. In addition, there were four QTL that showed
significance in three trials, all individually explaining less
than 10% of the phenotypic variance. In total, 10 of the 25
Journal of Genetics, Vol. 90, No. 3, December 2011
413

Page 6

Fa Cui et al.
Figure 1. Genetic linkage map and location of putative QTL for thousand-kernel weight and kernel dimensions based on 485 RILs derived
from Weimai 8 × Jimai 20 and 229 RILs from Weimai 8 × Yannong 19, with the prefixes WJ and WY, respectively. The positions of
marker loci on chromosomes are listed on the left of the corresponding chromosomes. Map distances are shown in centiMorgans and were
calculated using the Kosambi (1944) mapping function. A putative QTL with LOD > 2.5 is placed on its corresponding flanking markers.
QTL symbols are described at the bottom right of figure 1, and an uppercase letter E plus a numeral, 1, 2, 3 or 4, or the uppercase letter P
under the corresponding QTL symbol indicates the trial in which QTL was detected. Here, we only showed the conditional QTL detected
in the trial where it was undetectable by unconditional QTL mapping method (figure 1 continues on following pages).
QTL alleles increasing KDR were donated by the common
parent Weimai 8 in the two populations.
Conditional QTL mapping in the WJ and WY populations
Of the 14 unconditional additive QTL for TKW in WJ, eight,
five and four of them were undetectable when TKW was
conditioned on KW, KL and KDR, respectively (table 3).
However, one and four conditional QTL showed additive
effects similar to that of the corresponding unconditional
QTL when TKW was conditioned on KW and KDR, respec-
tively. In the WY population, of the nine putative uncondi-
tional additive QTL for TKW, six, two and four failed to
show significance, respectively, when the influence of KW,
414
Journal of Genetics, Vol. 90, No. 3, December 2011

Page 7

QTL for wheat kernel weight
. Figure 1 (contd)
KLandKDRonTKWwasexcluded;butzero,twoandthree,
respectively, showed additive effects similar to that of the
corresponding unconditional QTL (table 4). In both WJ and
WY, changes in additive effects of the remaining conditional
QTL,comparedtoitscorrespondingunconditionalQTL,was
inconsistent over trials, being unchanged in one trial, but
greatly/moderately changed in another one, or being greatly
changed or even undetectable in one trial, but moderately
changed in another one.
In addition, conditional QTL mapping analysis revealed
numerous additional additive QTL that could not be iden-
tified by unconditional QTL mapping method (tables 5 and
6). Ten, seven and 10 additional conditional QTL in WJ, and
five, eight and six in WY were detected when TKW was con-
ditioned on KL, KW and KDR, respectively. Of these, two,
three and four in WJ, and two, one and three in WY, respec-
tively, have been detected in unconditional QTL mapping
analysis in other trial/trials.
Discussion
Kernel dimensions: how do they contribute to the kernel weight
at an individual QTL level?
A comparison of conditional and unconditional QTL map-
ping analysis provides information about the genetic con-
trol system involved in the synthetic process between TKW
and its related traits at the level of an individual QTL.
For example, when performing conditional QTL analysis of
TKW conditioned on KL (TKW|KL), there are four possi-
ble results: (i) a QTL detected by the unconditional method
can be identified with a similar or equal effect, indicating that
this QTL for TKW expresses independently for the given
trait KL; (ii) a QTL detected by the unconditional method
can be identified with either a greatly reduced or a greatly
enhanced effect, suggesting that this QTL for TKW is par-
tially, but not completely, associated with KL; (iii) a QTL
detected by the unconditional method cannot be identified,
Journal of Genetics, Vol. 90, No. 3, December 2011
415

Page 8

Fa Cui et al.
. Figure 1 (contd)
416
Journal of Genetics, Vol. 90, No. 3, December 2011

Page 9

QTL for wheat kernel weight
. Figure 1 (contd)
meaning that this QTL for TKW is entirely contributed by
KL; (iv) an additional QTL can be detected by the condi-
tional mapping method, which means that the expression of
the QTL for TKW is completely suppressed by KL, and the
effects could only be identified by eliminating the influence
of KL. This suggests that the additional QTL has an opposite
additive effect on KL and the other causal trait of TKW.
The present study, together in the two populations, indi-
cated that only two QTL, QTkw-WJ-1A.1 and QTkw-WJ-
5D-2.2, were both entirely due to variation in all the three
TKW-related traits referred here, i.e., KL, KW and KDR
(tables 3 and 4). Three QTL, QTkw-WJ-2D-1.1, QTkw-WJ-
7A.2b and QTkw-WY-7D.1, were completely contributed by
both KL and KW; however, KDR contributed to them at
different levels. Five QTL, QTkw-WJ-1A.2, QTkw-WJ-2B.2,
QTkw-WY-6B.1, QTkw-WY-6D.2 and QTkw-WY-7B.1a, were
detectedentirelyduetovariationinbothKWandKDR;how-
ever, KL contributed to them at different levels, and some-
times they showed inconsistency across environments. Four
QTL, QTkw-WJ-3B.3, QTkw-WJ-7A.3, QTkw-WY-6B.2 and
QTkw-WY-7B.1b, were all entirely contributed by KW con-
sistently over trials in which they were detected; KL made no
Journal of Genetics, Vol. 90, No. 3, December 2011
417

Page 10

Fa Cui et al.
. Figure 1 (contd)
contribution except for one trial where they were detected,
but made entire contribution to them in the remaining
trials, showing inconsistency across environments. Both
QTkw-WJ-7B-2.3 and QTkw-WY-2B-1.3 expressed entirely
dependent of variation in KL, but independent of variation in
KDR, in all trials where they showed significance; however,
KW entirely contributed to them in one trial each and par-
tially contributed to them in the remaining two trials each.
For three QTL, QTkw-WJ-4A.5, QTkw-WJ-6A.5 and QTkw-
WJ-7A.2a, KDR did not have any impact on their expression
in all trials where they showed significance; KL and KW had
inconsistent influence on these QTL over trials. QTkw-WJ-
5B.4 was entirely/partially due to variation in KL, but KW
had no influence on it; due to strong association between
KDR and KL, this QTL was partially contributed by KDR
consistently over trials. QTkw-WJ-3D.2 was partially due to
variation in KW, KL; but KDR had inconsistent influences
on it over trials. QTkw-WY-2A.1 was completely contributed
by KDR, partially contributed by KL, but KW made no
contribution to it. For QTkw-WY-2A.3, all the three TKW-
related traits referred here had inconsistent influences
on it.
In addition, many QTL for TKW were undetectable in
unconditional QTL mapping analysis due to the interference
of related traits (tables 5 and 6; figure 1). Of the extra con-
ditional QTL for TKW, some were suppressed simultane-
ously by two or three related traits referred here (figure 1).
Of these, five each were detected when TKW were condi-
tioned on KW and KDR, respectively, and they were pair-
wise co-located on 3B, 4A, 5B and 7A in WJ, and 4B in WY,
respectively (tables 5 and 6; figure 1). Conditioning TKW on
KL and KW, respectively, one each extra conditional QTL
for TKW was detected on chromosome 7B in WJ, and they
shared a common interval, as did one pair of conditional
QTL on 2A in WY, which was identified when the influence
of KL and KDR on TKW were excluded, respectively. Two
conditional QTL for TKW in WY were both suppressed by
all the three TKW-related traits, distributing on chromosome
6A and 7B, respectively. Notably, several unconditional QTL
were detected in conditional QTL mapping analysis in the
trial in which they failed to show significance in uncondi-
tional QTL mapping analysis. Though we considered them
as extra conditional QTL for TKW, they shared common
intervals with that of their corresponding unconditional QTL.
The results described above demonstrate that some uncon-
ditional QTL for TKW were contributed by more than one
component trait, based on whether the QTL could be iden-
tified by conditional analysis (tables 3 and 4). In QTL map-
ping, the likelihood of detecting a QTL is dependent on
the ratio between the variance caused by the QTL’s effect
and the total variance of the trait, as well as the size of
the mapping population (Lander and Botstein 1989). In
conditional QTL analysis, effects on QTL contributed by a
conditional trait are reduced and the QTL with effects below
a certain threshold become virtually undetectable. Thus, it
is reasonable to obtain the results described, which indi-
cated that the unconditional QTL for TKW was strongly
influenced by the conditional traits, indicating a pleiotropic
QTL. Notably, there was a difference in the conditional
mapping results in different environments. The environment
plays an important role in controlling gene expression, espe-
cially for quantitative traits, and could account for the above
differences.
Conditional QTL mapping analysis above indicated} that,
at the QTL level, KW had the strongest influence on} TKW,
next to KL, and KDR had the least level of contribution}
to TKW (tables 3, 4, 5 and 6). This finding confirmed the}
results of the correlation analysis and variance analysis in
table} 2, and is also consistent with previous researches
(Dholakia et al. 2003; Sun et al. 2009; Gegas et al. 2010).
Therefore, to increase TKW, we should enhance KW, in
accordance with practical wheat breeding.
418
Journal of Genetics, Vol. 90, No. 3, December 2011

Page 11

QTL for wheat kernel weight
Table 3. Conditional QTL for thousand-kernel weight with respect to kernel dimensions in the WJ population.
Condition add (En/PVE%)b
TKW|KWUncona
TKW|KL
0.83 (E3/3.51) −
0.80 (E2/4.06) =
1.12 (E1/4.88) =
−0.94 (E3/5.25) +
−0.67 (P/2.99) =
0.87 (E1/3.03) −
TKW|KDR
QTkw-WJ-1A.2
QTkw-WJ-1A.1
QTkw-WJ-2B.2
QTkw-WJ-2D-1.1
QTkw-WJ-3B.3
−0.77 (E2/2.21) −
0.95 (E1/3.46) −
0.83 (E2/3.11) +
−0.71 (E3/2.77) =
1.23 (E1/6.04) =
0.81 (E2/3.00) =
0.75 (E3/5.48) =
0.87 (E4/3.97) =
0.86 (P/4.39) =
−1.25 (E1/4.40) +
−1.30 (E2/6.71) +
−1.26 (E3/7.19) +
−1.23 (P/4.92) +
−1.22 (E1/4.57) =
−1.04 (E2/4.90) =
−0.69 (E3/2.83) =
−1.09 (E4/6.12) =
−0.92 (P/4.92) =
−0.96 (E2/4.16) =
−1.02 (E4/6.12) =
−0.76 (P/3.30) −
0.91 (E4/4.16) +
0.58 (P/1.92) −
1.88 (E1/4.89) =
1.41 (E2/3.66) =
1.15 (P/2.92) =
QTkw-WJ-3D.2
−0.79 (E3/4.72) +
−0.47 (P/2.42) −
0.68 (E1/3.20) −
1.03 (E2/7.11) +
QTkw-WJ-4A.5
QTkw-WJ-5B.4
−0.72 (E3/2.96) −
−0.64 (P/2.78) −
−0.99 (E1/4.49) =
−1.15 (E2/8.16) =
−1.11 (E3/8.70) =
−0.96 (P/9.82) =
−0.83 (E2/4.82) −
QTkw-WJ-5D-2.2
QTkw-WJ-6A.5
−1.13 (E1/5.14) −
−1.16 (E2/7.54) =
−0.81 (E4/4.33) −
−0.79 (P/4.70) −
−1.06 (E4/6.83) +
QTkw-WJ-7A.2a
−0.86 (E4/5.41) =
QTkw-WJ-7A.2b
QTkw-WJ-7A.3
0.78 (E1/2.34) =
QTkw-WJ-7B-2.3
1.54 (E1/5.89) −
0.84 (P/3.59) −
Here, we only compared the conditional QTL and unconditional QTL in the trials where the unconditional QTL
showed significance.aUnconditional QTL.bConditional QTL. TKW|KL, TKW without the influence of KL;
TKW|KW, TKW without the influence of KW; TKW|KDR, TKW without the influence of KDR. Numerals
before parentheses are estimates of the additive effect of the conditional QTL. A letter E plus a numeral, or the
letter P, in parentheses, indicate the environment in which the conditional QTL was detected. The numeral after
the slash in parentheses is the percentage of phenotypic variance explained by the additive effects of the mapped
QTL. A minus, ‘−’, or a plus, ‘+’, following the parentheses, denotes the additive effect of a conditional
QTL, in absolute, that increases or decreases more than 10% of that of the corresponding unconditional QTL,
respectively. An equal sign, ‘=’, was placed after the parentheses if a conditional QTL with equal additive
effects to that of the unconditional QTL. An unconditional QTL that still showed significance in conditional
analysis in a trial in which it did not show significance in unconditional analysis is given in bold font. For the
remaining abbreviations and descriptions see table 1.
Determining genetic relationship of a pairwise QTL for two
related/causal traits based on their co-location: are the results
consistent with that of conditional QTL mapping analysis?
Coincidence of QTL may indicate either single QTL with
pleiotropic effects or that the genomic regions associated
with these QTL harbour a cluster of linked genes associated
with those traits. Generally, QTL for pairwise traits that have
strong genetic association are prone to be co-located. How-
ever,unconditionalQTLforacomplex traitareusuallyunder
the interference of its more than one causal traits, thus we can
hardly clarify the actual relationships between the complex
trait and its certain causal trait at an individual QTL level
through unconditional analysis (Wen and Zhu 2005; Ye et al.
2009).
Indeed, the present study confirmed this theory, as the co-
location of QTL for pairwise traits was not always consistent
with that of conditional QTL mapping analysis (figure 1).
Based on the conditional QTL mapping analysis, we knew
that all the unconditional QTL for TKW detected in the two
populations were either completely or partially contributed
by at least one of the three TKW-related traits referred here.
This implied that all the unconditional QTL for TKW should
be co-located with at least one QTL for KW, KL, or KDR;
however, six unconditional QTL for TKW in WJ and eight in
WY did not show pleiotropic effects; in other words, no QTL
Journal of Genetics, Vol. 90, No. 3, December 2011
419

Page 12

Fa Cui et al.
Table 4. Conditional QTL for thousand-kernel weight with respect to kernel dimensions in the WY population.
Condition add (En/PVE%)b
TKW|KW
−1.39 (E1/12.23) =
0.78 (E3/5.67)−
−1.93 (E4/20.15) +
−1.29 (E3/15.34)−
−0.97 (P/12.74) −
Uncona
TKW|KLTKW|KDR
QTkw-WY-2A.1
QTkw-WY-2A.3
−1.56 (E1/10.94) +
0.98 (E3/6.79) −
−1.92 (E4/16.71) +
1.13 (E3/8.40) =
1.55 (E4/11.18) =
−1.45 (E1/9.03) =
−1.35 (E3/14.79) =
−1.29 (P/11.06) =
−1.50 (E1/9.51) =
QTkw-WY-2B-1.3
QTkw-WY-6B.2
QTkw-WY-6B.1
QTkw-WY-6D.2
QTkw-WY-7B.1a
QTkw-WY-7B.1b
QTkw-WY-7D.1
−1.29 (E1/6.95) =
1.04 (E2/8.49) =
−1.11 (E1/5.71) =
−0.99 (E3/6.89) −
−1.15 (E4/8.01) =−1.17 (E4/5.57) =
−1.07 (E4/4.91) =
See table 3 for title descriptions.
for KL, KW or KDR was co-located with them. Of these,
in WJ, one each was distributed on chromosomes 1A, 4A
and 7B, respectively, and three on 7A; in WY, one each was
located on chromosomes 6D and 7D, two each on 2A, 6Band
7B, respectively (figure 1). Though the remaining nine QTL
co-segregated with one or more QTL for its related traits,
Table 5. Extra conditional QTL for thousand-kernel weight with respect to kernel dimensions in the WJ population.
Traita
Extra conditional QTLb
Intervalc
En.d
LODe
PVE%f
Addg
TKW|KL
QTkw|kl-WJ-2A.1
QTkw|kl-WJ-2B.2
QTkw|kl-WJ-2D-2.1
QTkw|kl-WJ-3A.2
QTkw|kl-WJ-5A-1.1a
QTkw|kl-WJ-5A-1.1b
QTkw|kl-WJ-5D-1.1a
QTkw|kl-WJ-5D-1.1b
QTkw|kl-WJ-7A.1
QTkw|kl-WJ-7D.2
QTkw|kw-WJ-3B.1
QTkw|kw-WJ-3B.2
QTkw|kw-WJ-4A.1
QTkw|kw-WJ-4D.1
QTkw|kw-WJ-5B.1
QTkw|kw-WJ-7A.1a
QTkw|kw-WJ-7A.1b
QTkw|kdr-WJ-1D-2.1
QTkw|kdr-WJ-2A.1
QTkw|kdr-WJ-3B.1
QTkw|kdr-WJ-4A.1
QTkw|kdr-WJ-5A-2.1
QTkw|kdr-WJ-5B.1
QTkw|kdr-WJ-7A.2
QTkw|kdr-WJ-7A.1
QTkw|kdr-WJ-7B-1.1
QTkw|kdr-WJ-7B-2.1
Xgwm382.3–Xgwm558
Xbarc98–Xwmc332.2
PPO29.1–Xmag633
Xswes185–Xmag896.1
Xmag694–Xcfe186
Xbarc165–Xcwm216
Xcfd78–Xcfd189
Xbarc133–Xcfe242.1
Xgwm473–Xedm16.1
Xswe2940.3–Xgwm111
Xbarc164–Xcft53
Xbarc229.2–Xcwm93
Xcfe190.1–Xapr1.2.1
Xcfd23–Xwmc617.2
Xissr854.1–Xwmc73
Xgwm473–Xedm16.1
Xbarc176.2–Xcfe261
Glu-d1–Xbarc346.1
Xdupw210–Xgwm382.1
Xbarc164–Xcft53
Xcfe190.1–Xapr1.2.1
Xbarc40–Xgwm205
Xissr854.1–Xwmc73
Xbarc176.2–Xcfe261
ww160.2–Xbarc49.2
Xcfe233–Xissr844.2
Xbarc65–Xcfe75
P4.10 4.870.85
E4/P
P
E3/P
E3
P
E1
3.61/3.06
2.76
3.88/5.24
4.78
3.32
3.19
3.35
3.34
3.05/3.14
2.9
3.20/2.81
3.73
3.87
2.57
3.09
3.12
5.88
4.05
2.70
4.76
3.39
3.72
3.24/2.63
7.60
2.62
3.01
5.77/4.42
3.19
3.76/4.77
8.39
11.09
2.53
4.62
4.48
3.56/3.45
3.36
5.15/4.01
3.70
3.92
2.21
3.69
2.92
6.84
3.70
2.25
5.80
6.82
3.60
3.68/3.65
6.41
4.02
2.82
1.18/0.97
0.66
−0.76/−0.81
−1.41
−1.29
−0.84
−0.91
−0.86
−0.96/−0.85
0.66
0.86/0.57
−0.70
0.70
−0.55
−0.68
−0.68
1.68
−0.83
0.62
1.00
−1.07
−0.83
−1.01/−0.85
1.05
0.97
1.21
E2 (E3/E4)
E3 (E2/E4)
E1/E2
E4
E1/P
E3
E3
E3 (E1/E2/E4/P)
E3 (E2/E4)
E1 (E3/P)
E1
E3
P
E3
E3
E3 (E1/E2/E4/P)
E1/E4 (E3/P)
E3 (E1/E4/P)
E4
E4 (E1/E2/P)
TKW|KW
TKW|KDR
aSee table 3 for details.bA conditional QTL that still showed significance in unconditional analysis but in other trial/trials is marked by
bold typeface.cFlanking markers of the QTL.dA letter E plus a numeral, or a letter P, before the parentheses, indicates a trial in which
the conditional QTL showed significance in conditional analysis; and that in the parentheses indicates a trial in which the conditional QTL
showed significance in unconditional analysis. In all trials where QTL showed significance in both conditional and unconditional analysis
are given in bold font.eLOD value of the corresponding putative additive QTL.fPhenotypic variance explained by the corresponding
putative additive QTL.gAdditive effect of the corresponding putative additive QTL; positive values indicate Weimai 8 alleles that increase
the value of the corresponding trait, and conversely, negative values indicate Weimai 8 alleles decrease it.
420
Journal of Genetics, Vol. 90, No. 3, December 2011

Page 13

QTL for wheat kernel weight
Table 6. Extra conditional QTL for thousand-kernel weight with respect to kernel dimensions in the WY population.
Traita
Extra conditional QTLb
Intervalc
En.d
LODe
PVE%f
Addg
TKW|KL
QTkw|kl-WY-2A.1
QTkw|kl-WY-4D-2.1
QTkw|kl-WY-6A.1
QTkw|kl-WY-7B.1
QTkw|kl-WY-7D.1
QTkw|kw-WY-1A.1a
QTkw|kw-WY-1A.1b
QTkw|kw-WY-2A.2
QTkw|kw-WY-2B-1.1
QTkw|kw-WY-3D.1
QTkw|kw-WY-4B-1.2
QTkw|kw-WY-6A.2
QTkw|kw-WY-7B.2
QTkw|kdr-WY-2A.1
QTkw|kdr-WY-2B-1.1
QTkw|kdr-WY-4B-1.2
QTkw|kdr-WY-6A.2
QTkw|kdr-WY-6B.1
QTkw|kdr-WY-7B.1
Xissr848–Xgwm372
Xcau17.1–Xcfd71
Xcfe179.2–Xswes123.3
Xcau12.4–ww121
Xswm6–Xswes558.2
Glu-a1–Xgdm93
Xmag3489–Xme3em2.8
Xcfe175.2–Xcfe253.2
Xmag1729.1–Xgwm374
Xcfe172–BE470813.1
Xcau8.1–Xgwm495
Xcfe179.2–Xswes123.3
Xcau12.4–ww121
Xissr848–Xgwm372
Xcinau119.2–Xcinau119.1
Xcau8.1–Xgwm495
Xcfe179.2–Xswes123.3
Xswes180.1–Xmag3469
Xcau12.4–ww121
P (E1)
E1
E1
E2 (E3)
E1
E2
E4
E1/P
P
E2
E2/P
E2/P
E2/P (E3)
P (E1)
E2 (E1/E3/P)
E2/P
E2/P
E1
E2 (E3)
5.06
2.70
3.22
2.97
2.74
3.06
2.67
9.36
4.10
6.61
5.73
4.11
9.44
7.46
−1.03
1.00
1.21
−0.88
−1.02
1.13
1.17
TKW|KW
2.52/5.48
2.88
2.53
4.35/4.08
4.67/3.14
3.31/2.60
4.85
2.61
3.39/3.78
5.19/4.95
5.13
2.67
9.93/13.92
8.19
4.38
9.19/7.52
9.62/6.11
7.49/5.26
18.15
8.70
7.35/7.16
11.66/9.40
10.82
4.77
−1.31/−1.06
−0.93
−0.71
1.34/1.00
−1.04/−0.67
−0.92/−0.63
−1.52
−1.11
1.34/1.27
−1.28/−1.11
−1.74
−0.82
TKW|KDR
See table 5 for title descriptions.
they all accounted only for a part of the conditional QTL
mapping analysis. For example, conditional analysis indi-
cated that QTkw-WJ-1A.2 was partially/entirely contributed
by KL and entirely contributed by both KW and KDR;
however, only QKw-WJ-1A.2 co-located with it (table 3;
figure 1).
Based on the above comparisons of results of conditional
QTL mapping analysis and traditional analysis, we con-
clude that it will not precise or efficient enough to determine
genetic relationships of two related/causal traits according to
the co-locations of QTL for the pairwise traits. Conditional
QTL mapping methods for multivariate conditional analysis
is an efficient tool to reveal relationships between a complex
trait and its causal traits.
QTL consistency over environments
If a QTL is independent of the environment, the implica-
tion is that its expression is stable regardless of differences
in environment. We defined a stable QTL that was verified in
at least three of the five trials. Together in WJ and WY pop-
ulations, 24 QTL were stable and were involved in all the
four traits referred here (table 7). Of these, QKdr-WJ-5A-1.4,
QTkw-WY-2A.3, QTkw-WY-2B-1.3, QKl-WY-6B.4 and QKdr-
WY-6B.5 were major QTL that accounted for more than
10% of the phenotypic variance with LOD scores of >3.0
(see tables 1 and 2 in electronic supplementary material).
Generally, a major QTL consistent over environments is of
great value for marker assisted selection (MAS) in breeding
programmes; thus, the five major stable QTL should be of
great value in genetic improvement of wheat kernel-related
traits.
Two-related RIL populations with large/moderate size
If a QTL detection is conducted based on a population with
limited lines, the number of QTL that can be detected are
usually underestimated and their effects are prone to be over-
estimates (Beavis 1998; Bernardo 2004; Schön et al. 2004;
Vales et al. 2005; Buckler et al. 2009). False positive QTL
Table 7. The number of unconditional QTL detected in the WJ and WY populations.
No. of QTL
PVE%
5%–10%
No. of trial
Trait
<5%
>10%12345 Total
TKW
KL
KW
KDR
9/0
7/2
9/1
7/2
5/6
4/3
4/4
4/9
0/3
1/2
0/4
1/2
2/5
6/4
9/8
5/9
6/2
2/1
2/1
1/2
3/2
2/1
1/0
4/1
1/0
1/1
1/0
2/0
2/0
1/0
0/0
0/1
14/9
12/7
13/9
12/13
For each entry, the first numeral refers to WJ, and the second, to WY. See table 1 for abbreviations
and title descriptions.
Journal of Genetics, Vol. 90, No. 3, December 2011
421

Page 14

Fa Cui et al.
Table 8. Congruent QTL resolved in both populations.
WJa
WYb
Allelesc
Common locid
QKw-WJ-1A.2
QKdr-WJ-2A.4
QKl-WJ-6A.1a
QKl-WJ-6D.1
QTkw-WJ-7B-2.3
QKw-WY-1A.1b
QKdr-WY-2A.2
QKl-WY-6A.1a
QKl-WY-6D.1
QTkw-WY-7B.1b
+/−
−/−
+/−
−/−
+/−
Glu-a1/Glu-a1
Xbarc212/Xbarc212
Xwmc580.1/Xwmc580.1
Xswes123.1/Xswes123.1
Xwmv517.1/Xwmc517.1
aQTL detected in the WJ population.bQTL detected in the WY population.cAdditive effect; for
additional details, see table 5.dLoci nearby the corresponding putative additive QTL are common
in the two populations.c,dFor each entry, the first signal refers to WJ, and the second, to WY.
mightbecausedbyparentalsharingwhentheRILpopulation
was not large enough to permit completely random mating
(Zou et al. 2005). Of the 51 unconditional QTL for the four
kernel-related traits reported here, only two QTL, individ-
ually, explained more than 10% of the phenotypic variance
(table 7). In 175 lines randomly sampled from the 485 RIL
lines of WJ and a high-density genetic map enriched with
DArT markers, however, 40 of the 57 unconditional QTL
for the four kernel-related traits, individually, accounted for
more than 10% of the phenotypic variance. Thirteen QTL
showed significance simultaneously in analysis based on 485
and 175 lines of WJ RIL population, and each of them had
much higher additive effects and contribution rates in 175
lines of WJ than 485 lines. For example, QTkw-WJ-5B.4
explained 3.35, 4.22, 4.75 and 4.0% of the phenotypic vari-
ance with additive effect values of −1.09, −1.09, −1.03 and
−0.91 g in E1, E2, E4 and P, respectively, in the 485 RIL
lines of WJ; however, in the 175 lines of WJ RIL popula-
tion, it showed significance in E2, E4 and P with additive
effect values of −1.38, −1.61 and −1.71 g, accounting for
8.42, 12.59 and 15.45% of the phenotypic variance (data
not shown). In the WY population, 11 of 38 unconditional
QTL for the four kernel-related traits accounted for more
than 10% of the phenotypic variance, individually (table 7).
QTL analysis based on 172 lines randomly sampled from
the 229 RIL lines of WY and the high-density genetic map
enriched with DArT markers, 45 unconditional QTL for the
four kernel-related traits were identified, 33 of which indi-
vidually explained more than 10% of the phenotypic vari-
ance. Fifteen QTL were significant in analysis based on both
172 and 229 lines of the WY RIL population, with equal or
slight higher additive effects and contribution rates in the 172
lines of WY compared to that in the 229 lines of WY (data
not shown). The above analysis indicates that: (i) it is diffi-
cult to detect minor QTL using a small population; (ii) QTL
effects are apt to be overestimated with small populations;
and (iii) the coincidence of QTL across environments may be
influenced by the population size to some extent.
In addition, if we conducted QTL detection using a single
mapping population, only a limited number of QTL could
be detected and the result was not conclusive. With the
rapid development of molecular marker technology, addi-
tional research on QTL effects in more than one different
or related genetic backgrounds is warranted (Kumar et al.
2007;Maetal.2007;Breseghelloand Sorrells2007;Buckler
et al. 2009; Uga et al. 2010; Gegas et al. 2010). Breseghello
and Sorrells (2007) have shown that QTL detected on dif-
ferent mapping populations, with identical evaluation meth-
ods, can be very distinct. However, common QTL among
related/associated mapping populations have been proven to
be detectable when the methods of evaluation are identical
(Ma et al. 2007; Buckler et al. 2009). Present study detected
at least five pairwise congruent QTL in the two related RIL
populations, based on common markers in the two genetic
maps(table8;figure1).Ofthese,thecommonparentWeimai
8 alleles of two congruent QTL showed consistent additive
effects, being positive simultaneously in the two populations.
As is well known, a reducing or enhancing additive effect is
not absolute but relative to the effect of two parental alleles,
sotheremainingthreeQTLcanstillberegardedascongruent
QTL. Due to the limited number of common loci in the two
genetic maps, the precise prediction and definition of com-
mon QTL in the two populations were hampered, although
positions of most QTL for the same trait identified in the two
populations were of high congruency. The results showed
that QTL from the common parent in the two related popu-
lations can be detected repeatedly to a certain extent, and the
comparable QTL are authentic.
Comparison of the present study with previous studies
TKW, KL and KW in wheat has been subjected to mono-
somic or QTL analysis in many other reports. In most cases,
they were reported individually but not simultaneously in
one report (Halloran 1976; Giura and Saulescu 1996; Shah
et al. 1999; Araki et al. 1999; Kato et al. 2000; Ammiraju
et al. 2001; Varshney et al. 2000; Zanetti et al. 2001; Böner
et al. 2002; Huang et al. 2003, 2004, 2006; Groos et al. 2003;
Campbell et al. 2003; McCartney et al. 2005; Verma et al.
2005; Li et al. 2007; Kirigwi et al. 2007; Röder et al. 2008;
Hai et al. 2008; Golabadi et al. 2010; McIntyre et al. 2010;
Su et al. 2010; Zheng et al. 2010). Though several other
reports have documented QTL for TKW and KD simulta-
neously by traditional QTL mapping analysis or association
mapping analysis, no conditional QTL mapping method was
implemented to dissect their genetic relationships (Giura and
422
Journal of Genetics, Vol. 90, No. 3, December 2011

Page 15

QTL for wheat kernel weight
Saulescu 1996; Campbell et al. 1999; Dholakia et al. 2003;
Breseghello and Sorrells 2006, 2007; Sun et al. 2009; Ramya
et al. 2010; Tsilo et al. 2010). The present study first evalu-
atedthegeneticrelationshipbetweenKWandKDatthelevel
of an individual QTL using unconditional and conditional
QTL mapping methods, thus enhancing the understanding of
genetic control system involved in the synthetic process of
TKW and KD. In addition, the combination of two related
populations with a large/moderate population size made the
results authentic and accurate.
Most QTL reported here are consistent with the previ-
ous reports. QTkw-WJ-1A.2 and Qkw-WJ-1A.2, a pairwise
pleiotropic QTL, correspond to a pairwise pleiotropic QTL
for TKW and KW reported by Campbell et al. (1999); in
addition, Varshney et al. (2000) and Ramya et al. (2010)
have detcted QTL for TKW in this interval; this chromo-
somal region may harbour a robust QTL cluster for kernel-
related traits, hence being potentially more useful in breeding
programmes. QKl-WJ-2D-2.3 and QKdr-WJ-2D-2.3, another
pairwise pleiotropic QTL, are consistent with the co-loctaed
QTL for KW and TKW reported by Ramya et al. (2010);
also, Wang et al. (2009) have located a QTL for TKW in
this interval; using 175 lines randomly sampled from the 485
RIL lines of WJ and the high-density genetic map enriched
with DArT markers, these two QTL have been reproducibly
detected, accounting for 9.88–16.16% and 5.58–15.18% of
the the phenotypic variance, respectively (data not shown);
thus, more attention should be paid to this chromosomal
region in MAS in wheat breeding programmes. QTkw-WJ-
4A.5, a environment-independent QTL, is in agreement with
the QTL for KL revealed by Sun et al. (2009). QTkw-
WJ-5B.4, QKl-WJ-5B.4 and QKdr-WJ-5B.1b shared a com-
mon interval on chromosome 5B; in addition, when TKW
was conditioned on KDR and KW, QTkw|kdr-WJ-5B.1 and
QTkw|kw-WJ-5B.1 showed significance in a new environ-
ment in this interval; Ramya et al. (2010) have reported a
pleiotropic QTL for TKW and KW in this interval; as QKl-
WJ-2D-2.3 and QKdr-WJ-2D-2.3, QTkw-WJ-5B.4 showed
significance in the small WJ RIL population, with additive
effect values of −1.38 to −1.71 g, accounting for 8.42–
15.45% of the phenotypic variance (data not shown); hence,
this interval may be potentially of great value in breed-
ing programmes. Both QKl-WJ-5A-1.2 and QKdr-WJ-5A-1.4
were mapped to a position similar to that of Vrn1, corre-
sponding to the QTL for TKW detected by Campbell et al.
(1999) and Wang et al. (2009); in addition, when TKW
wasconditionedonKL,QTkw|kl-WJ-5A-1.1bshowedsignif-
icance in this interval. The positions of QTkw-WJ-7A.3 and
a QTL for TKW reported by Huang et al. (2003) are of high
congruency. QTkw-WJ-7B-2.3 and QTkw-WY-7B.1b, a pair-
wise congruent QTL, were detected in WJ and WY popula-
tions, respectively; Hai et al. (2008) have reported a QTL for
TKW in this interval, indicating a reliable QTL; in addition,
the Vrn-B3 gene was about 3.0 cM distal from Xbarc65, one
flanking marker of QTkw-WJ-7B-2.3, indicating pleiotropic
effects (http://wheat.pw.usda.gov/GG2/index.shtml). QKl-
WJ-3B.5 and QKdr-WJ-3B.3 showed pleiotropic effects;
the two QTL are of high congruency in position to the
QTL for TKW detected Huang et al. (2006), Wang et al.
(2009) and Golabadi et al. (2010). QTkw|kw-WY-4B-1.2
and QTkw|kdr-WY-4B-1.2, two extra conditional QTL for
TKW without the influences of KW and KDR, respec-
tively, shared intervals of QTL for TKW reported by Huang
et al. (2004, 2006), McCartney et al. (2005), Zheng et al.
(2010), and for KL reported by Sun et al. (2009); in
addition, Rht-B1 was about 3.0 cM distal from Xgwm495,
one flanking marker of the two extra conditional QTL
(http://wheat.pw.usda.gov/GG2/index.shtml); this chromo-
somal segment should be a gene-rich region. One extra con-
ditional QTL for TKW excluding the influences of KL,
QTkw|kw-WY-4D-2.1, was mapped to a position similar
to Rht-D1. QKw-WJ-5A-3.3 and QKdr-WJ-5A-3.3, two co-
located QTL, confirmed a QTL for KL and a QTL for TKW
detected by Ramya et al. (2010) and Zheng et al. (2010),
respectively. QKdr-WJ-5A-1.4 and QKl-WJ-5A-1.2 showed
pleiotropic effects and they shared common intervals of QTL
for TKW reported by Wang et al. (2009) and for kernel diam-
eter reported by Tsilo et al. (2010). QTkw-WY-2A.3 corre-
sponds to a QTL for TKW reported by Huang et al. (2004);
a QTL cluster for kernel weight, kernel diameter and kernel
size has been reported in this interval by Tsilo et al. (2010);
thus, this interval may be potentially of great value in breed-
ing programmes. QKdr-WY-6B.5, QKl-WY-6B.2 and QKw-
WY-6B.1a are three co-located QTL; Huang et al. (2006)
have detcted QTL for TKW in this interval. QKl-WY-6D.3
and QKdr-WY-6D.3, two co-located QTL, have not been
reported elsewhere, as has QKl-WY-6B.4. For the remaining
QTL, comparison of the present study with previous studies
was hampered due to lack of common information for their
flanking markers of the corresponding QTL.
In summary, the combination of conditional and uncondi-
tional mapping methods applied to two related populations
can precisely evaluate genetic relationship between KW and
KD at an individual QTL level. In addition, a large popu-
lation size can enhance the authenticity and accuracy of the
QTL detection. Five major QTL that showed consistency in
expression across environments will be of great value for
MAS in breeding programmes.
Acknowledgements
This research was supported by the National Basic Research Pro-
gramme of China (973 Programme, 2006CB101700). The authors
thank Sishen Li, College of Agronomy, Shandong Agricultural
University, Taian, China, for kindly providing EST-SSR markers.
References
Ammiraju J. S. S., Dholakia B. B., Santra D. K., Singh H., Lagu
M. D., Tamhankar S. A. et al. 2001 Identification of inter sim-
ple sequence repeat (ISSR) markers associated with seed size in
wheat. Theor. Appl. Genet. 102, 726–732.
Journal of Genetics, Vol. 90, No. 3, December 2011
423