Molecular Genetics and Genomics

https://doi.org/10.1007/s00438-020-01646-0

ORIGINAL ARTICLE

Genome‑wide association study of soybean seed germination

under drought stress
Zhangxiong Liu1 · Huihui Li1 · Zuowang Gou2 · Yanjun Zhang2 · Xingrong Wang2 · Honglei Ren3 · Zixiang Wen4 ·
Beom‑Kyu Kang5 · Yinghui Li1 · Lili Yu1 · Huawei Gao1 · Dechun Wang4 · Xusheng Qi2 · Lijuan Qiu1

Received: 22 April 2019 / Accepted: 9 January 2020

© Springer-Verlag GmbH Germany, part of Springer Nature 2020

Abstract
Drought stress, which is increasing with climate change, is a serious threat to agricultural sustainability worldwide. Seed
germination is an essential growth phase that ensures the successful establishment and productivity of soybean, which can
lose substantial productivity in soils with water deficits. However, only limited genetic information is available about how
germinating soybean seeds may exert drought tolerance. In this study, we examined the germinating seed drought-tolerance
phenotypes and genotypes of a panel of 259 released Chinese soybean cultivars panel. Based on 4616 Single-Nucleotide
Polymorphisms (SNPs), we conducted a mixed-linear model GWAS that identified a total of 15 SNPs associated with at least
one drought-tolerance index. Notably, three of these SNPs were commonly associated with two drought-tolerance indices.
Two of these SNPs are positioned upstream of genes, and 11 of them are located in or near regions where QTLs have been
previously mapped by linkage analysis, five of which are drought-related. The SNPs detected in this study can both drive
hypothesis-driven research to deepen our understanding of genetic basis of soybean drought tolerance at the germination stage
and provide useful genetic resources that can facilitate the selection of drought stress traits via genomic-assisted selection.

Keywords  Soybean · Drought tolerance · Germination stage · Genome-wide association analysis (GWAS) · Single-
nucleotide polymorphisms (SNP)

Introduction

Drought is by many measures the most significant environ-

mental stress adversely affecting world agricultural produc-
tion (Boyer 1982; Cattivelli et al. 2008; Tuberosa and Salvi
2006), and predictions suggest that water deficit will con-
Communicated by Stefan Hohmann.
tinue to be a hugely influential abiotic factor affecting global
Zhangxiong Liu, Huihui Li, and Zuowang Gou have contributed crop yields (Sharma and Lavanya 2002). Moreover, in light
equally to this work. of current trends in climate change, it is also appreciated that
drought could become more frequent and severe (Mittler
Electronic supplementary material  The online version of this
and Blumwald 2010). To address the challenges of global
article (https​://doi.org/10.1007/s0043​8-020-01646​-0) contains
supplementary material, which is available to authorized users.

2
* Lijuan Qiu Institute of Crop Sciences, Gansu Academy of Agricultural
qiulijuan@caas.cn Sciences, Lanzhou 730070, China
3
Xusheng Qi Maize Research Institute, Heilongjiang Academy
qixusheng6608@sina.com of Agricultural Sciences, Harbin 150086, China
4
1 Department of Plant, Soil and Microbial Sciences, Michigan
National Key Facility for Gene Resources and Genetic
State University, East Lansing 48824, USA
Improvement/Key Laboratory of Crop Germplasm
5
Utilization, Ministry of Agriculture/Institute of Crop Upland Crop Breeding Research Division, Department
Sciences, Chinese Academy of Agricultural Sciences, of Southern Area Crop Science, National Institute of Crop
Beijing 100081, China Science, Miryang 52402, Korea

13
Vol.:(0123456789)
 Molecular Genetics and Genomics

food security, it is thus critical to increase our understand- biparental populations. Few studies were on QTLs underly-
ing of plant responses to abiotic stress; insights in this area ing drought-tolerance traits at the germination stage. Due
can support efforts to breed crops that can maintain higher to the unknown molecular mechanism of drought tolerance
improved yield under stress conditions (Condon et al. 2004; at germination stage in soybean, it is hard for breeders to
Morison et al. 2008). exploit genetic information on genomic-assisted breeding
Soybean is the most widely grown seed legume in the on drought tolerance.
world; and it is an inexpensive source of both protein and The present study used association analysis based on
vegetable oil. Drought stress is one of the most common genotypic and phenotypic information for a large germ-
adverse environmental conditions affecting soybean produc- plasm diversity panel of 259 released Chinese soybean cul-
tion (Dogan et al. 2007; Hufstetler et al. 2007; Mohammadi tivars and sought to identify genetic mechanisms underlying
et al. 2012), and it is now understood that drought affects drought tolerance at the seed germination stage. We used a
soybean seed yields (to at least some degree) during all popular method for phenotyping the drought tolerance of
growth stages (Boyer 1982). Typical yield losses associated germinating seeds for each cultivar in our association panel,
with water deficit are about 40% (Specht et al. 1999); how- and used a total of 4616 SNPs for our mixed-linear model
ever, depending upon the intensity and stage at which such GWAS. In total, we identified 15 associations between given
stress occurs, yield losses can be 80% or higher (Dias et al. SNPs and drought-tolerance indices, and three of these SNPs
2012; Oya et al. 2004). Due to the high costs of irrigation were significantly associated with more than one of the
and/or lack of water availability, irrigation is not a feasible drought-tolerance phenotypic indices. Five of the identified
option for most of the soybean growing regions. Therefore, associations have been previously associated with drought-
the development of soybean genotypes with drought toler- tolerance traits (canopy architecture and seed hardness),
ance, particularly those with the ability to germinate and supporting both the utility of our GWAS analysis and sug-
become established in water-deficit conditions, is viewed gesting that physiological mechanisms controlling drought
as one of the most economical ways to increase the stabil- tolerance in the germination stage may be similar to those
ity and sustainability of soybean production. Thus, drought known to help plants tolerate drought stress at later stages
tolerance has been identified as a major breeding target for of the plant lifecycle. Our study thus lays the foundation for
crop improvement (Rönnberg-Wästljung et al. 2005; Pennisi the molecular breeding of soybeans with increased drought
2008; Xiong et al. 2006). tolerance and generates multiple functional hypotheses relat-
A large number of studies have highlighted that tremen- ing to the basic biology of how germinating seeds manage
dous genetic variability underlies observed differences in water deficits.
soybean drought-tolerance traits (Frederick et  al. 2001;
Mederski and Jeffers 1973; Specht et al. 1986). Tradition-
ally, drought-tolerant genotypes have been selected based Materials and methods
on observations of phenotypes for plants grown under
controlled conditions. However, conventional breeding for Plant materials
drought-tolerant crop varieties is time-consuming and labor
intensive, due in large part to the complex quantitative nature An association mapping panel composed of 259 cultivars
of drought-tolerance traits and the attendant difficulties in released in the Northeast of China and the Yellow River
making selection for drought tolerance (Charlson et al. 2009; Valleys was constructed (Liu et al. 2017). Of the 259 Chi-
Ribaut et al. 1997). Thus, if quantitative genetics or other nese cultivars (which were from 11 provinces), 217 were
analytical methods can identify genomic regions associated derived from the Northern spring (Nsp) ecotype, while
with drought tolerance, breeders would have more useful 42 were derived from the Huang-huai-hai summer (Hsu)
guidance to facilitate the rapid development of improved ecotype. Notably, these cultivars are known to exhibit both
cultivars with increased drought tolerance, for example using clear geographic differentiation and genetic diversity (Liu
marker-assisted selection. et al. 2017). Detailed information for these 259 accessions
QTL mapping on drought-related traits in soybean has is presented in Table S1.
been conducted, focusing for example on yields under
drought stress conditions (Du et al. 2009a, b), fibrous roots Phenotyping
(Abdel-Haleem et al. 2011), water use efficiency (Mian et al.
1996, 1998; Specht et al. 2001), canopy wilting (Abdel-Hal- Fundamentally, our strategy for assessing the drought tolerance
eem et al. 2012; Charlson et al. 2009; Hwang et al. 2015, of the seeds of the various cultivars was based the PEG6000
2016), leaf pubescence density, leaf wilting coefficient, and concentration for soybean seeds drought treatment reported
rate of excised leaf drying (Du et al. 2009c). However, all by Shu et al. (2015) (the PEG6000 concentration which we
of these studies have been based on linkage mapping using used was 22%) and the data analysis was based on Thabet

13
Molecular Genetics and Genomics

et al. (2018). Very briefly, 30 healthy seeds for each culti- The method described above was implemented in tool
var were selected; seeds were put into sterilized Petri dishes “ANOVA” in the QTL IciMapping software (Meng et al.
(15 × 15 × 2.5 cm3) containing two vermiculite layers: the mass 2015).
of the vermiculite in each of the two layers was 35 g. Next,
120 ml of PEG6000 or control (aqueous) solutions (22% or Genotyping and genotypic data analyses
0%) were added to the dishes wherein the seeds were sand-
wiched between the vermiculite layers. Subsequently, the In general, the genotyping and genotypic data analyses fol-
plates were incubated in the dark in a germination chamber at lowed the same pipeline as in Liu et al. (2017). DNA sam-
25 ± 1 °C, and the number of germinate seeds were counted ples were extracted from soybean seedlings leaves follow-
after 7 days of incubation. Note that seeds were considered ing a previously described method (Kisha et al. 1997; Liu
to have germinated when the radicle was as long as the seed et al. 2017). The population was genotyped using an Illu-
and when the radical was half as long as the seed. For each mina SoySNP6k iSelect BeadChip; this contains 5361 SNPs
cultivar, the five germinated seeds with longest radicle were (Akond et al. 2013). Chromosomal distributions and qual-
selected; the lengths of these radicles were measured, and their ity control of these SNPs followed a previous study (Wen
fresh weights were measured and recorded. This seed germina- et al. 2014). Powermarker 3.25 was used to assess minor
tion experiment, which also tracked overall germination rates, allele frequencies (MAFs), polymorphic information con-
was repeated three times independently. Therefore, the evalu- tent (PIC), heterozygosity, and gene diversity (Liu and Muse
ated germination traits which we examined included the fresh 2005). A subset of 4616 SNPs (with missing rates lower than
radicle weight (RW), the fresh radicle length (RL), and the 0.25 and MAF higher than 0.05) was selected from amongst
germination rate [GR% = (number of germinated seeds/total the 5361 SNPs for use in the following analyses.
seed number used in the test) × 100]. We used the data for Multivariate analyses, including principal component
these traits to generate indices that were defined as the ratio of analysis (PCA), model-based population structure analysis,
the germination-related traits (RW, RL, and GR) under drought and cluster analysis with a neighbor-joining algorithm, were
conditions to the same traits under the control drought-free used to cluster the soybean lines of the population into sub-
condition: DT-RW, DT-RL, and DT-GR. groups; the PCA and cluster analysis were implemented in
TASSEL 5.0. The Bayesian model-based program STRU​
Phenotypic data analysis CTU​RE 2.3 (Pritchard et al. 2000) was also used to infer
the population structure; this assigned the 259 genotypes
For the three germination-related traits (i.e., RW, RL, and GR), into subpopulations based on the aforementioned 4616 poly-
phenotypic effect was partitioned into overall mean, genotypic morphic SNP markers. Ten independent analyses were per-
effect (Geno), treatment effect (Treatment), genotype-by- formed, with each including 100,000 burn-ins and 100,000
treatment effect (Geno × Treat), replication within treatment MCMC replications. We considered ten different hypotheti-
effect (Block/Treat), and random error effect in the ANOVA cal numbers of subpopulations (k) (1–10). The number of
model. Since the three drought-tolerance indices (i.e., DT-RW, subpopulations was declared when Δk reached its highest
DT-RL, and DT-GR) were derived from the two respective value (an ad hoc statistic; Evanno et al. 2005).
treatments, phenotypic effect of each DT-RW, DT-RL, and Using Arlequin V3.11, analysis of molecular variance
DT-GR was partitioned into overall mean, genotypic effect (AMOVA) (Excoffier et al. 1992) and F-statistics (FST) anal-
(Geno), replication effect, and random error effect in the ysis were performed (Excoffier et al. 2005) to investigate the
ANOVA model. Once the linear model of ANOVA is defined, population differentiations among the subpopulations that
total degree of freedom and total sum square can be partitioned were identified using STRU​CTU​RE. For assessing LD, r2
into the components defined in the linear model, from which was calculated using TASSEL 5.0 (Bradbury et al. 2007);
mean square (MS) of each source of variation can be calcu- the length of an LD block was determined when the decay
lated, and the significance test can be conducted. To calculate distance of LD reached r2 = 0.1.
the heritability in broad sense( (H2) for
) DT-RW, DT-RL,(and
DT-GR, the genetic variance 𝜎Geno and error variance 𝜎𝜀2 Genome‑wide association analysis
2
)

can be estimated from the theoretical expectation of MS, and

the heritability in broad sense can be estimated from the fol- For marker–trait associations, a mixed-linear regression
lowing equation: model controlling both population structure and kinship
2
matrices (denoted as Q + K model) was adopted in TAS-
H2 =
𝜎Geno SEL 5.0. The kinship coefficient between each accession pair
2
𝜎Geno + 𝜎𝜀2 was estimated in TASSEL 5.0 using the Loiselle algorithm
(Loiselle et al. 1995). Markers were defined as being sig-
nificantly associated with the trait by losing the Bonferroni

13
 Molecular Genetics and Genomics

correction as 2.6 to balance the rates of false positives and Table 2  Descriptive statistics of three drought-tolerance indices for
false negatives. If many SNPs were in one LD block with the the 259 soybean cultivars
chromosome segment controlling the trait of interest, they all Traits Minimum Maximum Mean SD CV
could be called as QTL. In practice, only the SNP having the
DT-RLa 0 0.2286 0.0683 0.0474 69.34
highest − log10(P) in the chromosome region was declared as
DT-RWb 0 0.0935 0.0348 0.0213 61.27
a QTL. In this study, LD decay distance (r2 = 0.1) was used
DT-GRc 0 1.0000 0.3624 0.2681 73.97
to determine LD blocks; and this was the support interval
used to declare significant SNPs associated with a given SD standard deviation, CV coefficient of variation
target trait. a
 DT-RL was defined as the ratio of radical length under drought
stress conditions and radical length under no-drought stress condi-
tions
Results b
 
DT-RW was defined as the ratio of fresh radical weight under
drought stress conditions and fresh radical weight under no-drought
stress conditions
Phenotypic variation and correlation analysis c
 DT-GR was defined as the ratio of germination rate under drought
stress conditions and germination rate under no-drought stress condi-
Three germination-related traits, namely RL, RW, and GR, tions
were measured in seeds for the 259 soybean accessions
under 0 (C) or 22% PEG6000 (D). The means, ranges,
standard deviations, and coefficients of variation for these distributions of DT-RL, DT-RW, and DT-GR among the cul-
traits are shown in Table S2. The means of RL, RW, and tivars of the population (Fig. 1), the three drought-tolerance
GR values for the seeds treated with 22% PEG6000 were indices which we calculated are apparently quantitative traits
1.5200 cm, 0.0260 g, and 35.6550%, while the means for that are likely controlled by multiple genes.
the controls were of 22.2885 cm, 0.7612 g, and 97.5167%, Pearson’s correlations between the three drought-toler-
respectively. Obviously, all of these traits were substantially ance indices were analyzed based on the means for the 259
higher under normal conditions than under drought-stressed accessions (Table 4). DT-GR was found to be positively cor-
conditions. ANOVA indicated significant (P < 0.001) vari- related with the DT-RL (r of 0.85; P < 0.001) and DT-RW (r
ation among genotypes, treatments, and genotype-by-treat- of 0.79; P < 0.001). DT-RL and DT-RW were also positively
ment interactions for all three measured traits among the correlated (r of 0.91; P < 0.001). Furthermore, the broad
cultivars of the population (Table 1), and genotype-by-treat- sense heritabilities for DT-RL, DT-RW, and DT-GR were
ment interaction had a much smaller variance compared to 85.69%, 84.59%, and 92.33%, respectively.
genotype, indicating that genotypic variation was the major
player in the observed phenotypic variation for the three Genetic diversity of the 259 accessions based
traits. The mean values for the calculated indices (DT-RL, on 4616 SNPs
DT-RW, and DT-GR) were 0.0683, 0.0348, and 0.3624
(Table 2), respectively, with significant (P < 0.001) variation Of the 4,616 SNPs of the selected subset from the chip,
for all three (Table 3). Given the substantial variation for the the mean values for MAF, gene diversity, heterozygosity,

Table 1  Analysis of variance Trait Source DFd Sum of square Mean square F value Pr > F
(ANOVA) of three germination-
related traits under 0% PEG and RL Geno 258 4230.71 16.40 17.26 < 0.0001
22% PEG conditions for seeds
Treatment 1 167,456.56 167,456.56 176,248.47 < 0.0001
of the 259 soybean cultivars
Block/treat 4 16.69 4.17 4.39 0.0016
Geno × treat 258 3280.25 12.71 13.38 < 0.0001
RW Geno 258 6.16 0.02 13.90 < 0.0001
Treatment 1 209.49 209.49 121,920.35 < 0.0001
Block/treat 4 0.50 0.12 72.25 < 0.0001
Geno × treat 258 5.81 0.02 13.11 < 0.0001
GR Geno 258 321,549.25 1246.31 39.65 < 0.0001
Treatment 1 1,483,266.88 1,483,266.88 47,189.53 < 0.0001
Block/treat 4 638.88 159.72 5.08 0.0005
Geno × treat 258 261,366.23 1013.05 32.23  < 0.0001

RL radical length, RW fresh radical weight, GR germination rate, DF degree of freedom

13
Molecular Genetics and Genomics

Table 3  Analysis of variance Traits Source DF Sum of square Mean square F value Pr > F
(ANOVA) of three drought-
tolerance indices in the 259 DT-RL a
Geno 258 1.7360 0.0067 18.94 < 0.0001
soybean cultivars
Replication 2 0.0029 0.0015 4.08 0.0174
DT-RWb Geno 258 0.3494 0.0014 17.38 < 0.0001
Replication 2 0.0011 0.0005 6.74 0.0013
DT-GRc Geno 258 55.5808 0.2154 36.99 < 0.0001
Replication 2 0.0702 0.0351 6.02 0.0026

DF degree of freedom
a
 DT-RL was defined as the ratio of radical length under drought stress conditions and radical length under
no-drought stress conditions
b
 DT-RW was defined as the ratio of fresh radical weight under drought stress conditions and fresh radical
weight under no-drought stress conditions
c
 DT-GR was defined as the ratio of germination rate under drought stress conditions and germination rate
under no-drought stress conditions

and PIC were 0.2773, 0.3649, 0.0751, and 0.2904, respec- with k = 3 was used in the subsequent genome-wide asso-
tively (Table S3, Fig. S1). The mean LD of the whole ciation analysis.
genome was r 2 = 0.2440. When the LD decay distance
was around 2,000 kb, r2 decreased to half of its maximum QTLs for the three drought‑tolerance indices
value; and when the decay distance was about 10,500 kb,
r2 was below 0.1 (Fig. S2); these results suggest relatively Fifteen QTLs, distributed in 13 soybean chromosomes were
slow LD decay in this population. The mean marker den- found to be significantly associated with the three drought-
sity was 204.27 kb for the whole population and, there- tolerance indices, and each of these explained (on aver-
fore, was assumed to have adequate power to detect major age) 5.81% of the phenotypic variation (Table 5, Figs. 3
QTL-related germination drought traits in the association and 4). There were two QTLs on each of chromosomes
panel. 12, and 20. Among the 15 detected QTLs some were asso-
Consideration of Δk values obtained in STRU​CTU​RE ciated with more than one trait: 11 were associated with
analysis showed that k = 3 was the optimal population DT-GR; 3 QTLs were associated with DT-RL; and 4 QTLs
subdivision, indicating that the accessions formed three were associated with DT-RW (Table 5). The QTL located
major clusters (Fig. 2a, b). To validate and gain further around 16,289,406-bp on chromosome 11 (ss247449682)
insight into the genetic diversity of the soybean germplasm was commonly associated with DT-GR and DT-RW. The
panel, we constructed a neighbor-joining tree based on the QTL located around 39,325,657-bp on chromosome 17
frequency of shared alleles among the accessions. Con- (ss249472124) was commonly associated with DT-GR
sistent with the results from STRU​CTU​RE analysis and and DT-RL. The QTL located around 46,546,046-bp on
the geographical origins, the 259 soybean accessions were chromosome 20 (ss250606162) was commonly associated
classified into three major groups (Fig. 2c). One group was with DT-RW and DT-RL. These three QTLs highlight the
composed largely of accessions from the Huang-huai-hai aforementioned strong positive correlations between these
valley region (41 out of 48); one was composed largely of drought-tolerance indices (Table 4). We consider the poten-
accessions from Heilongjiang province (107 out of 122); tial functional information related to these QTLs below, in
and the third group was mainly composed of accessions the discussion. Notably, 11 of the QTLs we identified are
from Jilin province (76 out of 89) (Table S4). located within 500 kb of loci reported in previous quan-
PCA has been proposed as an alternative to population titative genetics analyses in soybean, with 5 of them hav-
structure analysis for studying population stratification ing been previously associated with drought-related traits
from genotypic data (Patterson et al. 2006). A PCA of the (Table 5).
entire set of 259 accessions with the 4616 SNPs (Fig. 2d)
also showed a clear separation of the same three groups
that were identified in the STRU​C TU​R E and neighbor- Discussion
joining tree analyses. The significant pairwise FST values
(Table S5) found between the three subpopulations (0.21, Seed germination is obviously an essential stage in the
P < 0.001) further confirmed the existence of three sub- growth cycle of most seed plants. The amount of water
populations for the 259 accessions. Therefore, a Q matrix needed for the germination stage is much smaller than

13
 Molecular Genetics and Genomics

Fig. 1  Frequency distributions of three drought-tolerance indices DT-RL (a), DT-RW (b), and DT-GR (c) in the 259 soybean cultivars

amount needed over the whole soybean life cycle, but it revealed substantial diversity in an association panel of 259
is now understood that water deficits in the germination released cultivars (Fig. 1), and a genome-wide association
stage of soybean can cause severe effects on overall growth. analysis based on the 4,616 SNPs identified fifteen QTLs
Drought occurs frequently in the three main soybean pro- associated with drought tolerance (Table 5).
duction areas of China, particularly during the sowing time As noted in the results, our association mapping results
for cultivars based on the Hsu and Nsp ecotypes, thus caus- for the drought-tolerance indices were compared with
ing difficulty for sowing and germination. Currently, little QTLs previously reported within a 500 kb vicinity using
is known regarding the genetics of drought tolerance in soy- tools at Soybase (https​://www.soyba​se.org). Five of the
bean at the germination stage. In the present study, the distri- QTLs detected in present study, including ss246991464 on
bution of values for three drought-related phenotypic indices Chr. 9, ss247449682 on Chr. 11, ss248468442 on Chr. 14,

13
Molecular Genetics and Genomics

Table 4  Phenotypic correlations between drought stress indices based (Abdel-Haleem et al. 2012; Hwang et al. 2016; Kaler et al.
on the means of the traits in the 259 soybean cultivars 2017).
Trait DT-RWb DT-GRc Of particular note, the region which contained our QTL
for the SNP marker ss250269376 has been identified in
DT-RLa 0.9138*** 0.8542***
three previously reported studies (Abdel-Haleem et  al.
DT-RWb 0.7917***
2012; Hwang et al. 2016; Kaler et al. 2017); the SNP that
***Significant level under 0.0001 for Pearson correlation test we detected is 3090 bp away from the SNP (ss715635661)
a
 DT-RL was defined as the ratio of radical length under drought detected by Kaler et al. (2017) that was associated with
stress conditions and radical length under no-drought stress condi- delayed canopy wilting. In terms of our association panel, it
tions was notable that cultivars with allele G for the ss250269376
b
 
DT-RW was defined as the ratio of fresh radical weight under position had significantly higher DT-GR values than those
drought stress conditions and fresh radical weight under no-drought
stress conditions
with allele A across all accessions (P value < 0.01) (Fig. 5a).
c Further analysis indicated that the distribution of allele G
 DT-GR was defined as the ratio of germination rate under drought
stress conditions and germination rate under no-drought stress condi- was also uneven across the three geographical subpopula-
tions tions; the frequency of the favorable allele G was only 5.98%
in SG 2 (mainly from Heilongjiang province), 12.35% in
SG 3 (mainly from Jilin province), but rose to fully 82.93%
ss250269376 on Chr. 19, and ss250606162 on Chr. 20, were in SG 1 (mainly from Hsu region) (Fig. 5b). Seed hard-
located in regions (Table 5) that have been previously asso- ness is known to be closely related with drought tolerance
ciated with the genetic control of delayed canopy wilting (Liu 2009). Thus, it was interesting that the region near

Fig. 2  Genetic structure and relatedness of the 259 soybean cultivars. from groups when k = 3; c Neighbor-joining tree constructed using
a Evolution of the natural logarithm probability of the data against K SNP data; soybean accessions from Heilongjiang are shown in red;
and the magnitude of Δk for each K value; b Clustering for K = 3 for those from Jilin are shown in blue; those from Liaoning and Inner
the entire panel of soybean cultivars. Each individual is represented Mongolia are shown in purple; and those from the Huanghuaihai val-
by a vertical bar, partitioned into colored segments with the length of ley region are shown in black. d Principal component analysis for the
each segment representing the proportion of the individual’s genome entire panel of soybean cultivars

13


13
Table 5  SNPs significantly associated with the three calculated drought-tolerance indices
Marker Chr Position Associated traits (R2%) − log10(P) Alleles and allelic effect Reported QTLs/genes

ss245388356 4 16,902,743 DT-RW (4.99) 2.6990 C (0.0476)

ss245627013 5 8,209,039 DT-GR (7.02) 3.3790 T (0.1404)
ss245905576 6 16,437,297 DT-GR (4.97) 2.6383 C (0.5007) Seed yield (Du et al. 2009a), seed protein (Mao et al. 2013), Stem
strength (Chen et al. 2011)
ss246144068 7 746,249 DT-RW (5.34) 2.8268 C (0.0085) Seed sucrose (Maughan et al. 2000), plant height (Yao et al. 2015)
ss246991464 9 38,525,177 DT-RL (5.30) 2.7670 G (0.0263) Canopy wilt (Kaler et al. 2017)
ss247449682 11 16,289,406 DT-GR (5.48), DT-RW (4.80) 2.9469 T (0.5326); T (0.0450) Canopy wilt (Kaler et al. 2017)
ss247772178 12 31,934,231 DT-GR (4.91) 2.6326 A (0.5698)
ss247780010 12 33,115,103 DT-GR (5.57) 2.9208 G (0.0752)
ss248097753 13 31,586,157 DT-GR (6.25) 3.3443 G (0.2081) Seed weight (Hyten et al. 2004; Eskandari et al. 2013a; Teng et al.
2009), seed protein (Hyten et al. 2004; Eskandari et al. 2013a), seed
oil (Eskandari et al. 2013b)
ss248468442 14 44,520,959 DT-GR (6.84) 3.6765 C (− 0.1277) Canopy wilt (Abdel-Haleem et al. 2012), seed weight (Liu et al.
2011), seed oil (Eskandari et al. 2013b)
ss249472124 17 39,325,657 DT-GR (6.72), DT-RL (5.18) 3.5559 C (0.2742); C (0.0390) Seed yield (Kabelka et al. 2004), seed protein (Tajuddin et al. 2003),
seed oil (Liang et al. 2010; Mao et al. 2013; Han et al. 2015)
ss249791040 18 47,660,490 DT-GR (5.23) 2.7144 T (− 0.1576) Pod number (Zhang et al. 2007), seed protein (Lu et al. 2012; Jun
et al. 2008)
ss250269376 19 47,214,600 DT-GR (6.63) 3.4927 G (0.2933) mqCanopy wilt (Abdel-Haleem et al. 2012; Hwang et al. 2016),
canopy wilt (Kaler et al. 2017), seed weight (Csanadi et al. 2001;
Funatsuki et al. 2005), seed number (Funatsuki et al. 2005), pod
number (Yang et al. 2013), seed oil (Hyten et al. 2004), SCN (Guo
et al. 2005), seed hardness (Zhang et al. 2008)
ss250603455 20 46,154,637 DT-GR (5.11) 2.7305 C (0.2772) Seed yield (Yuan et al. 2002; Kabelka et al. 2004), Seed weight (Han
et al. 2012)
ss250606162 20 46,546,046 DT-RL (6.35), DT-RW (7.88) 3.3085 A (0.0309); A (0.0154) Canopy wilt (Kaler et al. 2017)
Molecular Genetics and Genomics
Molecular Genetics and Genomics

signaling may influence protein and oil concentrations of

seeds (Dornbos and Mullen 1992) or, alternatively, that
the genes controlling these traits may simply be located in
similar chromosomal regions. Identification of more trade-
associated markers in such regions could help to delineate
the specific functional contributions of particular loci. It is
even possible that marker-assisted selection could be used to
simultaneously improve both drought tolerance and protein
or oil content traits. The other four drought-associated QTLs
that we identified have not been reported in the previous
studies.
The soybase database (https​://www.soyba​se.org/) was
used to identify candidate genes hit directly by the SNPs
of our detected QTLs or in nearby genes; a total of 22 can-
didate genes were thusly identified (Table S6). Of note,
none of the SNPs was direct hits in the exons of genes;
all 15 of the SNPs were located in intergenic regions or
in regulatory regions. Functional annotations for the 22
genes were obtained from Soybase, which revealed that our
candidate genes potentially encode transcription factors,
protein kinases, auxin response factors and transporters,
and proteins relating to the E3 ubiquitin system. Notably,
three genes (Glyma.04G126000, Glyma.12G174000, and
Glyma.11G167100) are apparently homologs of genes
from other plants that have serine/threonine-protein kinase
or phosphatase functions related to tolerance stress in potato
(Bai et al. 2017), Arabidopsis (Huang et al. 2018), wheat
(Ghorbel et al. 2017), rice (Liao et al. 2016), and Vigna
aconitifolia (Tiwari et al. 2018). Such genes are known to
Fig. 3  Quantile–quantile plots from GWAS across the three drought- be regulated by the well-characterized drought-related plant
tolerance indices: DT-GR (a), DT-RL (b), and DT-RW (c)
hormone ABA (Hou et al. 2016). Work in other plant spe-
cies has also shown that ABA impacts drought-related plant
ss250269376 detected in the present study was previously regulatory networks that includes the type of transcription
found to be associated with seed hardness (Zhang et al. factors putatively encoded by our hits Glyma.06g188400
2008; Table 5). and Glyma.17g239700 (MYB), and Glyma.20g231800
It is interesting to note that the QTLs which we detected (bHLH99) (Chander et al. 2018; Chen et al. 2015; Gao et al.
for the drought indices are genetically linked to markers 2016; Park et al. 2018; Takahashi et al. 2016; Wang et al.
associated with seed weight and yield-related traits on Chrs. 2017; Xu et al. 2018; Zhao et al. 2018). Finally, we were
6, 13, 14, 17, 18, 19, and 20 (Table 5) (Csanadi et al. 2001; also intrigued by a putative E3 ubiquitin-protein ligase gene
Du et al. 2009a; Eskandari et al. 2013a; Funatsuki et al. among our hits, Glyma.20g231900. Many previous studies
2005; Han et al. 2012; Hyten et al. 2004; Kabelka et al. 2004; have reported drought-tolerance-related functions for E3
Liu et al. 2011; Teng et al. 2009; Yang et al. 2013; Yuan ligase proteins (Ding et al. 2015; Serrano et al. 2018).
et al. 2002; Zhang et al. 2007). Furthermore, the genomic In conclusion, we here used GWAS to examine a germ-
regions for six of our QTLs have been previously associated plasm association panel comprising 259 soybean released
with protein content and oil content, including ss245905576 Chinese cultivars for drought-associated traits based on
on Chr. 6, ss248097753 on Chr. 13, ss248468442 on Chr. phenotypic data from drought-treated germinating soybean
14, ss249472124 on Chr. 17, ss249791040 on Chr. 18, and seeds. The analysis was based on a total of 4,616 SNPs,
ss250269376 on 19 (Eskandari et al. 2013a, b; Han et al. and 15 SNP-trait associations were identified by GWAS,
2015; Hyten et al. 2004; Jun et al. 2008; Liang et al. 2010; among which three SNPs were significantly associated
Lu et al. 2012; Mao et al. 2013; Tajuddin et al. 2003). One with two of the drought-tolerance indices. Practically, sub-
possible implication of these findings is that drought stress sequent applied research based on our study research will

13
 Molecular Genetics and Genomics

Fig. 4  Manhattan plot from the Q + K model across the three drought-tolerance indices: DT-GR (a), DT-RL (b), and DT-RW (c)

seek to confirm the functional contributions of the identi- the germination stage of plant growth. Collectively, our
fied QTLs in different genetic backgrounds and in different work will facilitate future molecular breeding of soybean
environments. Scientifically, the loci detected in the present for improved drought tolerance, which will help to ensure
study lay the foundation for deepening our understanding stable high yields in increasingly variable climatological
of the genetic basis of soybean drought tolerance during conditions.

13
Molecular Genetics and Genomics

Fig. 5  Diagrams depicting the

genetic effects of bi-allele vari-
ation at position ss250269376
in relation to a DT-GR across
all accessions and geographi-
cal subpopulations inferred by
STRU​CTU​RE, b allele frequen-
cies in geographical subpopula-
tions inferred by STRU​CTU​
RE. SG 1, 2, and 3 are structure
groups 1, 2, and 3, inferred by
STRU​CTU​RE analysis. P val-
ues of 0.01 are shown as **

Acknowledgements  This work was supported by the National Key R Abdel-Haleem H, Carter TE, Purcell LC, King CA, Ries LL, Chen
& D Program for Crop Breeding (Grant No. 2016YFD0100304), the P, Schapaugh W, Sinclair TR, Boerma HR (2012) Mapping of
Development of Novel Elite Soybean Cultivars and Lines with High quantitative trait loci for canopy-wilting trait in soybean (Glycine
Oil Content (Grant No. Z161100000916005-06), the Crop Germplasm max L. Merr). Theor Appl Genet 125:837–846
Resources Protection (Grant No. 2017NWB036-5), the Platform of Akond M, Liu S, Schoener L, Anderson JA, Kantartzi SK, Meksem
National Crop Germplasm Resources of China (Grant Nos. 2018-004 K, Song Q, Wang D, Wen Z, Lightfoot DA, Kassem MA (2013)
and 2017-004), the Agricultural Science and Technology Innovation SNP-based genetic linkage map of soybean using the SoySNP6K
Program of the Chinese Academy of Agricultural Sciences (CAAS), Illumina Infinium BeadChip genotyping array. J Plant Genome
and the Improvement of Soybean Abiotic Stress Tolerance to Address Sci 1:80–89
the Climate Change (Grant No. PJ0121092018). Bai J, Mao J, Yang H, Khan A, Fan A, Liu S, Zhang J, Wang D, Gao
H, Zhang J (2017) Sucrose non-ferment 1 related protein kinase
Author contributions  ZL, XQ, and LQ conceived and designed the 2 (SnRK 2) genes could mediate the stress responses in potato
experiments. ZL, ZG, YZ, XW, and HR performed the experiments. (Solanum tuberosum L.). BMC Genet 18:41
ZL, HL, ZW, BKW, YL, LY, HG, DW, XQ, and LQ analyzed data and Boyer JS (1982) Plant productivity and environment. Science
wrote the manuscript. All authors read and approved the manuscript. 218:443–448
Bradbury PJ, Zhang Z, Kroon DE, Casstevens TM, Ramdoss Y, Buck-
ler ES (2007) TASSEL: Software for association mapping of
Compliance with ethical standards  complex traits in diverse samples. Bioinformatics 23:2633–2635
Cattivelli L, Rizza F, Badeck FW, Mazzucotelli E, Mastrangelo AM,
Conflict of interest  The authors have declared that no competing in- Francia E, Mare C, Tondelli A, Stanca AM (2008) Drought toler-
terests exist. ance improvement in crop plants: an integrated view from breed-
ing to genomics. Field Crop Res 115:1–14
Ethical approval  This article does not contain any studies with human Chander S, Almeida DM, Serra TS, Jardim-Messeder D, Barros
participants or animals performed by any of authors. PM, Lourenço TF, Figueiredo DD, Margis-Pinheiro M, Costa
JM, Oliveira MM, Saibo NJ (2018) OsICE1 transcription factor
improves photosynthetic performance and reduces grain losses
in rice plants subjected to drought. Environ Exp Bot 150:88–98
References Charlson D, Bhatnagar S, King C, Ray J, Sneller C, Carter T, Purcell L
(2009) Polygenic inheritance of canopy wilting in soybean [Gly-
cine max (L.) Merr.]. Theor Appl Genet 119:587–594
Abdel-Haleem H, Lee GJ, Boerma RH (2011) Identification of
Chen H, Shan Z, Sha A, Wu B, Yang Z, Chen S, Zhou R, Zhou X
QTL for increased fibrous roots in soybean. Theor Appl Genet
(2011) Quantitative trait loci analysis of stem strength and related
122:935–946
traits in soybean. Euphytica 179:485–497

13
 Molecular Genetics and Genomics

Chen T, Li W, Hu X, Guo J, Liu A, Zhang B (2015) A cotton MYB nematode races 2, 3 and 5 in soybean PI 90763. Theor Appl
transcription factor, GbMYB5, is positively involved in plant adap- Genet 103:1167–1173
tive response to drought stress. Plant Cell Physiol 56:917–929 Han Y, Li D, Zhu D, Li H, Li X, Teng W, Li W (2012) QTL analysis
Condon AG, Richards RA, Rebetzke GJ, Farquhar GD (2004) Breeding of soybean seed weight across multi-genetic backgrounds and
for high water-use efficiency. J Exp Bot 55:2447–2460 environments. Theor Appl Genet 125:671–683
Csanadi G, Vollmann J, Stift G, Lelley T (2001) Seed quality QTLs Han Y, Teng W, Wang Y, Zhao X, Wu L, Li D, Li W (2015) Uncon-
identified in a molecular map of early maturing soybean. Theor ditional and conditional QTL underlying the genetic interrela-
Appl Genet 103:912–919 tionships between soybean seed isoflavone, and protein or oil
Dias FG, Borges ACN, Viana AAB, Mesquita RO, Romano E, Grossi contents. Plant Breed 134:300–309
de Sa MF, Nepomuceno AL, Loureiro ME, Ferreira MA (2012) Hou YJ, Zhu Y, Wang P, Zhao Y, Xie S, Batelli G, Wang B, Duan
Expression analysis in response to drought stress in soybean: CG, Wang X, Xing L, Lei M, Yan J, Zhu X, Zhu JK (2016) Type
shedding light on the regulation of metabolic pathway genes. one protein phosphatase 1 and its regulatory protein inhibitor 2
Genet Mol Biol 35:222–232 negatively regulate ABA signaling. PLoS Genet 12:e1005835
Ding S, Zhang B, Qin F (2015) Arabidopsis RZFP34/CHYR1, a Huang X, Hou L, Meng J, You H, Li Z, Gong Z, Yang S, Shi Y
ubiquitin E3 ligase, regulates stomatal movement and drought (2018) The antagonistic action of abscisic acid and cytokinin
tolerance via SnRK2. 6-mediated phosphorylation. Plant Cell signaling mediates drought stress response in Arabidopsis. Mol
27:3228–3244 Plant 11:970–982
Dogan E, Kirnak H, Copur O (2007) Deficit irrigations during soybean Hufstetler EV, Boerma HR, Carter TE, Earl HJ (2007) Genotypic
reproductive stages and CROPGRO-soybean simulations under variation for three physiological traits affecting drought toler-
semi-arid climatic conditions. Field Crops Res 103:154–159 ance in soybean. Crop Sci 47:25–35
Dornbos DL, Mullen RE (1992) Soybean seed protein and oil contents Hwang S, King CA, Ray JD, Cregan PB, Chen P, Carter TE, Li
and fatty acid composition adjustments by drought and tempera- Z, Abdel-Haleem H, Matson KW, Schapaugh W, Purcell LC
ture. J Am Oil Chem Soc 69:228–231 (2015) Confirmation of delayed canopy wilting QTLs from
Du W, Fu S, Yu D (2009a) Genetic analysis for the leaf pubescence multiple soybean mapping populations. Theor Appl Genet
density and water status traits in soybean [Glycine max (L.) 128:2047–2065
Merr.]. Plant Breed 128:259–265 Hwang S, King CA, Chen P, Ray JD, Cregan PB, Carter TE, Li Z, Abdel-
Du W, Wang M, Fu S, Yu D (2009b) Mapping QTL for seed yield and Haleem H, Matson KW, Schapaugh W, Purcell LC (2016) Meta-
drought susceptibility index in soybean (Glycine max L.) across analysis to refine map position and reduce confidence intervals for
different environments. J Genet Genomics 36:721–731 delayed-canopy-wilting QTLs in soybean. Mol Breeding 36:91
Du W, Yu D, Fu S (2009c) Detection of quantitative trait loci for yield Hyten DL, Pantalone VR, Sams CE, Saxton AM, Landau-Ellis D, Ste-
and drought tolerance traits in soybean using a recombinant inbred faniak TR, Schmidt ME (2004) Seed quality QTL in a prominent
line population. J Integr Plant Biol 51:868–878 soybean population. Theor Appl Genet 109:552–561
Eskandari M, Cober E, Rajcan I (2013a) Genetic control of soybean Jun TH, Van K, Kim MY, Lee SH, Walker DR (2008) Association
seed oil: II. QTL and genes that increase oil concentration without analysis using SSR markers to find QTL for seed protein content
decreasing protein or with increased seed yield. Theor Appl Genet in soybean. Euphytica 162:179–191
126:1677–1687 Kabelka EA, Diers BW, Fehr WR, LeRoy AR, Baianu IC, You T,
Eskandari M, Cober E, Rajcan I (2013b) Genetic control of soybean Neece DJ, Nelson RL (2004) Putative alleles for increased yield
seed oil: I. QTL and genes associated with seed oil concentration from soybean plant introductions. Crop Sci 44:784–791
in RIL populations derived from crossing moderately high-oil Kaler AS, Ray JD, Schapaugh WT, King CA, Purcell LC (2017)
parents. Theor Appl Genet 126:403–495 Genome-wide association mapping of canopy wilting in diverse
Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clus- soybean genotypes. Theor Appl Genet 130:2203–2217
ters of individuals using the software STRU​CTU​RE: a simulation Kisha TJ, Sneller CH, Diers BW (1997) Relationship between genetic
study. Mol Ecol 14:2611–2620 distance among parents and genetic variance in populations of
Excoffier L, Smouse PE, Quattro JM (1992) Analysis of molecular soybean. Crop Sci 37:1317–1325
variance inferred from metric distances among DNA haplotypes: Liang H, Yu Y, Wang S, Lian Y, Wang T, Wei Y, Gong P, Liu X,
application to human mitochondrial DNA restriction data. Genet- Fang X, Zhang M (2010) QTL mapping of isoflavone, oil and
ics 131:479–491 protein contents in soybean (Glycine max L. Merr.). Ag Sci China
Excoffier L, Laval G, Schneider S (2005) Arlequin (version 3.0): an 9:1108–1116
integrated software package for population genetics data analysis. Liao YD, Lin KH, Chen CC, Chiang CM (2016) Oryza sativa protein
Evol Bioinform 1:47–50 phosphatase 1a (OsPP1a) involved in salt stress tolerance in trans-
Frederick JR, Camp CR, Bauer PJ (2001) Drought-stress effects on genic rice. Mol Breeding 36:22
branch and mainstem seed yield and yield components of deter- Liu X (2009) Drought. In: Lam HM et al (ed) Research on tolerance
minate soybean. Crop Sci 41:759–763 to stresses in Chinese soyben. China Agriculture Press, Beijing
Funatsuki H, Kawaguchi K, Matsuba S, Sato Y, Ishimoto M (2005) Liu KJ, Muse SV (2005) PowerMarker: an integrated analysis environ-
Mapping of QTL associated with chilling tolerance during repro- ment for genetic marker analysis. Bioinformatics 21:2128–2129
ductive growth in soybean. Theor Appl Genet 111:851–861 Liu W, Kim M, Van K, Lee Y, Li H, Liu X, Lee S (2011) QTL identifi-
Gao F, Yao H, Zhao H, Zhou J, Luo X, Huang Y, Li C, Chen H, Wu cation of yield-related traits and their association with flowering
Q (2016) Tartary buckwheat FtMYB10 encodes an R2R3-MYB and maturity in soybean. J Crop Sci Biotech 14:65–70
transcription factor that acts as a novel negative regulator of salt Liu Z, Li H, Wen Z, Fan X, Li Y, Guan R, Guo Y, Wang S, Wang D,
and drought response in transgenic Arabidopsis. Plant Physiol Qiu LJ (2017) Comparison of genetic diversity between Chinese
Bioch 109:387–396 and American soybean (Glycine max (L.)) accessions revealed by
Ghorbel M, Cotelle V, Ebel C, Zaidi I, Ormancey M, Galaud JP, Hanin high-density SNPs. Front Plant Sci 8:11. https​://doi.org/10.3389/
M (2017) Regulation of the wheat MAP kinase phosphatase 1 by fpls.2017.02014​
14-3-3 proteins. Plant Sci 257:37–47 Loiselle BA, Sork VL, Nason J, Graham C (1995) Spatial genetic struc-
Guo B, Sleper DA, Arelli PR, Shannon JG, Nguyen HT (2005) Iden- ture of a tropical understory shrub, PSYCHOTRIA OFFICINALIS
tification of QTLs associated with resistance to soybean cyst (Rubiaceae). Am J Bot 82:1420–1425

13
Molecular Genetics and Genomics

Lu W, Wen Z, Li H, Yuan D, Li J, Zhang H, Huang Z, Cui S, Du W Specht JE, Hume DJ, Kumudini SV (1999) Soybean yield potential—a
(2012) Identification of the quantitative trait loci (QTL) underly- genetic and physiological perspective. Crop Sci 39:1560–1570
ing water soluble protein content in soybean. Theor Appl Gen Specht JE, Chase K, Macrander M, Graef GL, Chung J, Markwell JP,
126:425–433 Germann M, Orf JH, Lark KG (2001) Soybean response to water:
Mao T, Jiang Z, Han Y, Teng W, Zhao X, Li W (2013) Identification a QTL analysis of drought tolerance. Crop Sci 41:493–509
of quantitative trait loci underlying seed protein and oil contents Tajuddin T, Watanabe S, Yamanaka N, Harada K (2003) Analysis of
of soybean across multi-genetic backgrounds and environments. quantitative trait loci for protein and lipid contents in soybean
Plant Breed 132:630–641 seeds using recombinant inbred lines. Breed Sci 53:133–140
Maughan PJ, Maroof MAS, Buss GR (2000) Identification of quanti- Takahashi Y, Kinoshita T, Matsumoto M, Shimazaki KI (2016) Inhi-
tative trait loci controlling sucrose content in soybean (Glycine bition of the Arabidopsis bHLH transcription factor by mono-
max). Mol Breeding 6:105–111 merization through abscisic acid-induced phosphorylation. Plant
Mederski HJ, Jeffers DL (1973) Yield response of soybean varieties J 87:559–567
grown at two soil moisture levels. Agron J 65:410–412 Teng W, Han Y, Du Y, Sun D, Zhang Z, Qiu L, Sun G, Li W (2009)
Meng L, Li H, Zhang L, Wang J (2015) QTL IciMapping: Integrated QTL analyses of seed weight during the development of soybean
software for genetic linkage map construction and quantitative (Glycine max L Merr). Heredity 102:372380
trait locus mapping in biparental populations. Crop J 3:269–283 Thabet SG, Moursi YS, Karam MA, Graner A, Alqudah AM (2018)
Mian MAR, Bailey MA, Ashley DA, Wells R, Carter TE, Parrott WA, Genetic basis of drought tolerance during seed germination in
Boerma HR (1996) Molecular markers associated with water use barley. PLoS ONE 13:e0206682
efficiency and leaf ash in soybean. Crop Sci 36:1252–1257 Tiwari B, Kalim S, Tyagi N, Kumari R, Bangar P, Barman P, Kumar
Mian MAR, Ashley DA, Boerma HR (1998) An additional QTL for S, Gaikwad A, Bhat KV (2018) Identification of genes associ-
water use efficiency in soybean. Crop Sci 38:390–393 ated with stress tolerance in moth bean [Vigna aconitifolia (Jacq.)
Mittler R, Blumwald E (2010) Genetic engineering for modern Marechal], a stress hardy crop. Physiol Mol Biol Pla 24:551–561
agriculture: challenges and perspectives. Annu Rev Plant Biol Tuberosa R, Salvi S (2006) Genomics-based approaches to improve
61:443–462 drought tolerance of crops. Trends Plant Sci 11:405–412
Mohammadi PP, Moieni A, Hiraga S, Komatsu S (2012) Organ spe- Wang N, Zhang W, Qin M, Li S, Qiao M, Liu Z, Xiang F (2017)
cific proteomic analysis of drought-stressed soybean seedlings. J Drought tolerance conferred in soybean (Glycine max. L) by
Proteomics 75:1906–1923 GmMYB84, a novel R2R3-MYB transcription factor. Plant Cell
Morison JIL, Baker NR, Mullineaux PM, Davies WJ (2008) Improv- Physiol 58:1764–1776
ing water use in crop production. Philos Trans R Soc Biol Sci Wen Z, Tan R, Yuan J, Bales C, Du W, Zhang S, Chilvers MI, Schmidt
363:639–658 C, Song Q, Cregan PB, Wang D (2014) Genome-wide association
Oya T, Nepomuceno AL, Farias JRB, Tobita S, Ito O (2004) Drought mapping of quantitative resistance to sudden death syndrome in
tolerance characteristics of Brazilian soybean cultivars—evalua- soybean. BMC Genomics 15:809
tion and characterization of drought tolerance of various Brazilian Xiong LM, Wang RG, Mao GH, Koczan JM (2006) Identifica-
soybean cultivars in the field. Plant Prod Sci 7:129–137 tion of drought tolerance determinants by genetic analysis of
Park DY, Shim Y, Gi E, Lee BD, An G, Kang K, Paek NC (2018) The root response to drought stress and abscisic acid. Plant Physiol
MYB-related transcription factor RADIALIS-LIKE3 (OsRL3) 142:1065–1074
functions in ABA-induced leaf senescence and salt sensitivity in Xu Z, Ge Y, Zhang W, Zhao Y, Yang G (2018) The walnut JrVHAG1
rice. Environ Exp Bot 156:86–95 gene is involved in cadmium stress response through ABA-signal
Patterson NJ, Price AL, Reich D (2006) Population structure and eige- pathway and MYB transcription regulation. BMC Plant Biol
nanalysis. PLoS Genet 2:e190 18:19
Pennisi E (2008) The blue revolution, drop by drop, gene by gene. Yang Z, Xin D, Liu C, Jiang H, Han X, Sun Y, Qi Z, Hu G, Chen Q
Science 320:171–173 (2013) Identification of QTLs for seed and pod traits in soybean
Pritchard JK, Stephens M, Donnelly P (2000) Inference of population and analysis for additive effects and epistatic effects of QTLs
structure using multilocus genotype data. Genetics 155:945–959 among multiple environments. Mol Genet Genomics 288:651–667
Ribaut JM, Jiang C, Gonzalez-de-Leon D, Edmeades GO, Hoisington Yao D, Liu Z, Zhang J, Liu S, Qu J, Guan S, Pan L, Wang D, Liu J,
DA (1997) Identification of quantitative trait loci under drought Wang P (2015) Analysis of quantitative trait loci for main plant
conditions in tropical maize. 2. Yield components and marker traits in soybean. Genet Mol Res 14:6101–6109
assisted selection strategies. Theor Appl Genet 94:887–896 Yuan J, Njiti VN, Meksem K, Iqbal MJ, Triwitayakorn K, Kassem MA,
Rönnberg-Wästljung AC, Glynn C, Weih M (2005) QTL analyses Davis GT, Schmidt ME, Lightfoot DA (2002) Quantitative trait
of drought tolerance and growth for a Salix dasyclados × Salix loci in two soybean recombinant inbred line populations segregat-
viminalis hybrid in contrasting water regimes. Theor Appl Genet ing for yield and disease resistance. Crop Sci 42:271–277
110:537–549 Zhang D, Cheng H, Wang H, Zhang H, Liu C, Yu D (2007) Identi-
SAS Institute Inc (2007) SAS/STAT software: Version 9.1.3. SAS fication of genomic regions determining flower and pod num-
Institute, Cary, NC bers development in soybean (Glycine max L.). J Genet Genom
Serrano I, Campos L, Rivas S (2018) Roles of E3 ubiquitin-ligases in 37:545–556
nuclear protein homeostasis during plant stress responses. Front Zhang B, Chen P, Chen C, Wang D, Shi A, Hou A, Ishibashi T (2008)
Plant Sci 9:139 Quantitative trait loci mapping of seed hardness in soybean. Crop
Sharma KK, Lavanya M (2002) Recent developments in transgenics Sci 48:1341–1349
for abiotic stress in legumes of the semi-arid tropics. In: JIRCAS Zhao Y, Cheng X, Liu X, Wu H, Bi H, Xu H (2018) The wheat MYB
Working Report. pp 61–73 transcription factor T ­ aMYB 31 is involved in drought stress
Shu Y, Zhou Y, Shi X, Hu N, Shao Q, Du J (2015) Screening of appro- responses in Arabidopsis. Front Plant Sci 9:1426
priate PEG-6000 concentration for the identification of soybean
drought tolerance at germination stage. Soybean Sci 1:56–59 Publisher’s Note Springer Nature remains neutral with regard to
Specht JE, Williams JH, Weidenbenner CJ (1986) Differential jurisdictional claims in published maps and institutional affiliations.
responses of soybean genotypes subjected to a seasonal soil water
gradient. Crop Sci 26:922–934

13