Chang2021 Article ExploringGeneticArchitectureFo

Mol Breeding (2021) 41: 28
https://doi.org/10.1007/s11032-021-01223-2
Exploring genetic architecture for pod-related traits

in soybean using image-based phenotyping
Fangguo Chang & Wenhuan Lv & Peiyun Lv & Yuntao Xiao & Wenliang Yan & Shu Chen &
Lingyi Zheng & Ping Xie & Ling Wang & Benjamin Karikari & Salah Fatouh Abou-Elwafa &
Haiyan Jiang & Tuanjie Zhao
Received: 13 November 2020 / Accepted: 18 March 2021 / Published online: 29 March 2021
# The Author(s), under exclusive licence to Springer Nature B.V. 2021
Abstract Mature pod color (PC) and pod size (PS) parameters, i.e., qPC3 and qPC19, located to chromo-
served as important characteristics are used in the soy- somes 3 and 19, respectively, and 12 stable QTLs for
bean breeding programs. However, manual phenotyp- pod size-related traits across nine chromosomes. Several
ing of such complex traits is time-consuming, laborious, genes residing within the genomic region of stable QTL
and expensive for breeders. Here, we collected pod were identified as potential candidates underlying these
images from two different populations, namely, a soy- pod-related traits based on the gene annotation and
bean association panel (SAP) consisting of 187 acces- expression profiling data. Our results provide the useful
sions and an inter-specific recombinant inbred line information for fine-mapping/map-based cloning of
(RIL) population containing 284 RILs. An image- QTL and marker-assisted selection of elite varieties with
based phenotyping method was developed and used to desirable pod traits.
extract the pod color- and size-related parameters from
images. Genome-wide association study (GWAS) and Keywords Soybean . Image analysis . Pod color . Pod
linkage mapping were performed to decipher the genetic size . GWAS . QTL mapping
control of pod color- and size-related traits across 2
successive years. Both populations exhibited wide phe-
notypic variations and continuous distribution in pod Introduction
color- and size-related traits, indicating quantitative
polygenic inheritance of these traits. GWAS and linkage Soybean (Glycine max L. Merr.) is a crucial crop in
mapping approaches identified the two major quantita- terms of its agronomic and economic importance due
tive trait loci (QTL) underlying the pod color to its nutritional qualities for human consumption,
F. Chang : W. Lv : P. Lv : Y. Xiao : W. Yan : L. Zheng :

P. Xie : L. Wang : T. Zhao (*)
National Center for Soybean Improvement, Key Laboratory of B. Karikari
Biology and Genetics and Breeding for Soybean, Ministry of Department of Crop Science, Faculty of Agriculture, Food and
Agriculture, State Key Laboratory for Crop Genetics and Consumer Sciences, University for Development Studies, P. O.
Germplasm Enhancement, Nanjing Agricultural University, Box TL, 1882 Tamale, Ghana
Nanjing 210095, China
e-mail: tjzhao@njau.edu.cn
S. F. Abou-Elwafa
S. Chen : H. Jiang (*) Agronomy Departments, Faculty of Agriculture, Assiut
College of Artificial Intelligence, Nanjing Agricultural University, University, Asyut 71526, Egypt
Nanjing 210095 Jiangsu, China
28 Page 2 of 21 Mol Breeding (2021) 41: 28
animal feed, and industrial utilization. The soybean pod (Barabaschi et al. 2016). Given that thousands or mil-
is regarded as a key fruit tissue determining the seed lions of molecular markers can now be easily devel-
yield via its photosynthetic ability (Yang et al. 2008). oped, phenotyping has increasingly become the bottle-
Pod-related traits such as mature pod color (PC) and pod neck for large-scale genetic analyses in crops (Zhou
size (PS) are also served as the critical characteristics et al. 2019; Furbank and Tester 2011). Although lots
that are usually used to identify and describe a new of commercial platforms and equipment have been de-
variety in breeding programs (Bernard 1967). It is ex- veloped in the last decade for plant phenotyping, the
tensively believed that modern cultivated soybean was high cost is one of major restricting factors for their
domesticated from wild soybean in China 5000 years widespread utilization (Vasseur et al. 2018). In recent
ago (Wang et al. 2016a). The allelic diversity in wild years, image-based phenotyping as a popular and acces-
soybeans is higher than cultivated soybeans. According sible technique has opened new opportunity to evaluate
to the USDA-Germplasm Resources Information Net- various phenotypes that had previously to be manually
work (GRIN, https://www.ars-grin.gov/), all wild quantified or could not be easily quantified (Gage et al.
soybeans have black and small pods, while most 2018; Diaz-Garcia et al. 2018a). This approach over-
improved cultivars have light-colored (tan or brown) comes the drawbacks of the manual phenotyping ap-
and relatively large pods, and the landraces are between proach with relatively low cost, wide applicability, and
them, with a certain proportion of black pods. Thus, high efficiency (Yang et al. 2020a). So far, a variety of
such pod-related characteristics could also be consid- image-based analysis approaches and software have
ered as domestication-related traits and that evolved been developed to carry on the image-based plant phe-
from wild soybean to modern varieties. notyping in many species (Fahlgren et al. 2015; Guo
Phenotyping of large population is essential for most et al. 2018; Fernandez-Gallego et al. 2018; Ali et al.
plant breeding programs or genetic studies (Ghanem 2020; Neumann et al. 2017; Diaz-Garcia et al. 2018b;
et al. 2015; Zhou et al. 2019). Traditional manual phe- Turner et al. 2018; Baek et al. 2020).
notyping approaches are mostly time-consuming, labo- GWAS and linkage mapping have been widely used
rious, and costly and prone to subjective errors (Chen to dissect the genetic architecture of quantitatively
et al. 2014; Uzal et al. 2018). According to the descrip- inherited traits in multiple crops (Korte and Farlow
tors and data standard for soybean proposed by Qiu and 2013; Tian et al. 2020; Yano and Tuberosa 2009;
Chang (Qiu and Chang 2006), based on visual evalua- Würschum 2012). In general, GWAS takes full advan-
tion for PC, it can be classified broadly into five cate- tages of all recombination events occurred in a natural
gories, i.e., grayish brown, yellowish-brown, brown, population, while linkage mapping is suitable for bi-
dark brown, and black. However, from a practical per- parental population. Combining these two mapping
spective, only three main classes, i.e., tan (Tn), brown strategies in one study can complement each other in
(Br), and black (Bl), could readily be distinguished. As overcoming the limitations associated with either of
for pod size-related parameters such as pod length (PL), them for dissecting the genetic basis of a given trait
pod width (PW), pod area (PA), and pod shape (straight, (Pascual et al. 2016). Several pod-related traits as com-
sickle, or bow-shaped), these attributes still are evaluat- plex quantitative traits have been investigated by previ-
ed by using the caliper and visual evaluation at harvest. ous researches either by GWAS or linkage mapping,
Thus, pod phenotyping based on manual phenotyping such as pod wall thickness, pod wall weight and ratio of
approaches are time-consuming and have greatly ham- pod wall to pod (Guo et al. 2011), pod number and pod
pered the progress of genetic studies and molecular maturity date (Zhang et al. 2010; Yang et al. 2013; Fang
breeding for target traits. et al. 2017), pod shattering/dehiscence (Hu et al. 2019)
The fruit-related traits are usually complex and con- and pod color (Bandillo et al. 2017; He et al. 2015; Song
trolled by large numbers of quantitative trait loci (QTL) et al. 2004). Although numerous QTLs underlying these
with both major and minor effects. Genotyping and traits have been identified, no promising progress has
phenotyping are two major determinants affecting the been achieved in the past decade except for a few studies
effectiveness and efficiency of genome-wide associa- identified the gene responsible for the QTL detected
tion studies (GWAS) and linkage mapping (Ibrahim (Dong et al. 2014). Hence, we employed both GWAS
et al. 2020). Recently, the cost of genotyping has be- and linkage mapping to identify the genomic regions
come relatively lower compared to the past decades underlying these pod-related traits. Then, the available
Mol Breeding (2021) 41: 28 Page 3 of 21 28
access to bioinformatics platforms/tools facilitated in long single-row plots with 0.5 m spacing between rows,
predicting the putative candidate genes for them while NJRINP was planted in 2-m-long row with five
(Abou-Elwafa and Shehzad 2018). hills per row. Field management was performed under
Therefore, this study was aimed to (1) employ a high- normal conditions.
resolution imaging pipeline to collect pod images from a
soybean association panel (SAP) consisting of 187 ac- Imaging-based phenotyping
cessions as well as an inter-specific recombinant inbred
line (RIL) population designated NJRINP with 284 Image acquisition
lines, (2) develop a new phenotyping approach based
on the imaging analysis to rapidly generate phenotypic Soybean pod is similar to that of some legumes, which
data for basic soybean pod color- and size-related pa- usually contains 1~3 seeds and rarely 4 seeds. Given the
rameters for breeding programs, (3) implement the phe- practical operability, the 3-seeded pod was regarded as
notypic data obtained from these two populations to the object in the current research. The phenotypic eval-
conduct GWAS and linkage mapping to identify the uation of soybean pod-related traits was performed
genomic regions associated with target traits, and (4) based on the generated pod images. Five soybean pods
predict the putative candidate genes for relatively stable harvested from each genotype at maturity stage (around
QTLs detected in the present study. These findings will the October 15 for SAP and September 26 for NJRINP)
provide valuable information to unravel the genetic were photographed using an industrial camera (MER-
basis of these traits meanwhile developing marker- CURY, MER-310-12UC, Daheng Group, Inc., China)
assisted selection (MAS) to breed the soybean varieties with a lens (M5018-MP2, 50mm, F1.8-F16C, comput-
with desirable pod-related traits and providing the sup- er, Inc., Japan) on a white background with the control-
port for applying the similar procedures in exploiting the lable lighting source (OPT-LI38037-w/AP1024-2,
genetic basis of fruit-related traits in other crops. OPT, Inc. China), and a 7-cm ruler was used as scale
bar information to calculate the data corresponding to
the actual size. To ensure a consistent height relative to
Materials and methods the sample, the camera was mounted to a fixed camera
stand (DH-FP380/500, Daheng Group, Inc., China). A
Plant materials and experimental design backlight and light-emitting diode (LED) lighting was
used to remove the shadows of soybean pods and hence
The SAP consists of 187 cultivated soybeans, and an reduced image errors caused by shadows during image
inter-specific RIL population was designated as generation. Pod samples were always laid out horizon-
NJRINP. The NJRINP population was generated by a tally along the longest axis of the pod and upward at
cross between the elite cultivar NN86-4 and the wild both ends in the center of the field below the camera.
soybean accession PI 342618B contains of 284 RILs. The software Daheng MER-Series Viewer was
The NN86-4 cultivar is characterized by tan pod color employed to capture images via the attached laptop
and large pod size and 100-seed weight (~17.9g), while computer. To facilitate separation of pods and the back-
the wild accession PI342618B exhibits black pod color ground, the initial values of the red, green, and blue
and small pod size and 100-seed weight (~1.1g). The (RGB) channel were set as 255, 255, and 255, respec-
detailed constructive procedures of the SAP and tively, for the white balance correction. The size of
NJRINP populations were described by Li et al. (Li finally generated image was 2048 × 1536 pixels.
et al. 2016a) and Wang et al. (Wang et al. 2016b),
respectively. All materials were obtained from the Na- Image processing and analysis
tional Center for Soybean Improvement, Nanjing Agri-
cultural University, Nanjing, China. Field trials of the The user-friendly Soybean Pod and Seed Phenotyping
SAP and NJRINP were conducted at the Jiangpu Ex- software SPSP version 1.0 was implemented to process
perimental Station of Nanjing Agricultural University and analyze the images and extract the pod color- and
(31 °02′ N and 118° 04′ E) in the summer of 2017 and size-related parameters from each image. This software
2018 using a completely randomized experimental de- was developed based on the Windows 10 operating
sign with three replications. The SAP was sown in 1-m- system, using Python and PyCharm as the programming
language and platform, respectively, and a series of Genotypic data

toolkits including PyQt toolkit, Qt Designer, and
Model-View-Controller (MVC) were adopted to design The SAP and NJRINP populations were genotyped by a
and create GUI, which took the soybean pod and seed as restriction-site associated DNA sequencing (RAD-Seq)
the measured object, owning the main functions in pod technique. The detailed information on the sequencing
color recognition and pod size-related parameter deter- of SAP and NJRINP were same as the earlier studies
minations including pod length (PL), pod width (PW), reported by Li et al. (Li et al. 2016a) and Wang et al.
and pod area (PA), as well as seed counting. Image (Wang et al. 2016b), respectively. Briefly, a total of 61,
segmentation as a crucial step in image processing was 541 single nucleotide polymorphisms (SNPs) with a
applied to separate the pod from the background. Brief- minor allele frequency (MAF)≥0.05 for 187 genotypes
ly, SPSP V1.0 software estimated PL (cm) and PW (cm) of SAP and a high-density genetic map composed of
as the length and width of the minimum enclosing 5728 bin markers for 284 RILs of NJRINP were used in
rectangle (MER) of the pod in the image, respectively, this study.
and PA (cm2) as the area of pod body. The RGB channel
values of each pixel for a given image were recorded to Statistical analysis
measure the pod color using the SPSP V1.0. To accu-
rately measure the pod color, the color space was then Descriptive statistics, analysis of variance (ANOVA),
converted from RGB to hue (H), saturation(S), and and correlation analysis were performed for the pheno-
value (V), and the HSV values were also recorded for typic data of the SAP as well as NJRINP using the
the same pixel. Scanning all pixels and counting the PROC MEANS, GLM, and CORR of SAS Version
saturation of each pixel for the given image were per- 9.4, respectively. Descriptions of all traits were deter-
formed, and the color of the pixel with the highest mined by the mean value of each trait over five repli-
saturation was recorded and regarded as the dominant cates. The broad-sense heritability (h2) for combined
color of the image. Because of the white background environments was calculated according to the method
color of the image (RGB values of 255, 255, and 255), proposed by Nyquist and Baker (Nyquist and Baker
image showing RGB pixel values greater than 200 was 1991) and shown as the following:
eliminated from further processing and analysis. The h i
RGB and HSV values of the corresponding dominant h2 ¼ σ2g = σ2g þ σ2ge =n þ σ2e =ðnrÞ
color of each image for pod were generated. The ex-
tracted value of the pod color- and size-related traits for Meanwhile the h2 for the individual environment was
each genotype was exported and saved as a .txt file. calculated as the following:
Details of these traits and their definitions were listed in
Table S1. The workflow for the image-based phenotyp- h2 ¼ σ2g = σ2g þ σ2e
ing was simplified and shown in Fig. S1. The data from
a few images was manually refined or corrected where where σ2g is the genotypic variance, σ2ge is the genotype
faulty. by environment interaction variance, σ2e is the error
variance, n is the number of environments, and r is the
Manual measurements number of replications. The best linear unbiased predic-
tors (BLUPs) for each genotype for all traits were cal-
Although image-based phenotyping method was mainly culated using the R package lme4 (Bates et al. 2015)
used to measure the pod color and size for SAP and taking the effect of environment, replicate within envi-
NJRINP, the manual measurements were also needed ronment, genotype, interaction between genotype and
for the test samples for verification purpose. Visual environment, and error as random effects.
assessment for the PC was performed in the NJRINP
population in 2017. Pod size-related parameters includ- Genome-wide association study
ing PL and PW were measured manually by using a
ruler for the subsets of both SAP and NJRINP in 2017 The mixed linear model (MLM) algorithm was imple-
and 2018. Subsequently, the ratio of PL to PW (PRLW) mented with GAPIT package (Version: 2.0) (Tang et al.
was calculated. 2016) in R to perform GWAS with high-throughput
SNP markers. After filtering the SNPs with a minor population. The loci identified by GWAS were aligned
allele frequency (MAF) < 0.05, a total of 61541 SNPs to the reported QTL using the flanked markers informa-
was used for GWAS in the SAP population. The signif- tion. The physical interval of a QTL was determined
icant cutoff of P = 8.12×10−7 [-log (0.05/61541)] based based on the physical position of the flanking markers
on the Bonferroni correction method and a suggestive located in the confidence interval of that QTL. Genes
threshold of P-value of 1×10−4 were also adopted for annotated in G. max Williams 82 reference gene model
SNP-trait association. Linkage disequilibrium (LD) was 1.0 were the source of candidate genes. Gene annotation
measured by the squared correlation coefficient (r2) of and previously published QTLs/genes for relevant traits
the pairwise SNPs across the whole genome using the in soybean and other crop species as well as the
RTM-GWASV1.1 software (He et al. 2017). LD decay orthologs in Arabidopsis with known functions of reg-
rate was estimated by the previously proposed method ulating the fruit color, size, and shape were retrieved
(Huang et al. 2010). The LD block was identified using from the SoyBase (http://www.soybase.org) and
the Haploview V4.2 software with default algorithm and available literature. The expression profiling data of
plotted by R package LD heatmap (Barrett et al. 2005; “ZH13,” a Chinese soybean reference genome, was
Shin et al. 2006). The association results were displayed also used to further investigate the expression level of
using the Manhattan plots with an R script and package candidate genes at different stages of pod development
CMplot (Yin et al. 2020). The peak SNPs were chosen (Shen et al. 2019). In addition, the known information
to represent the loci for the final GWAS result. on the potential biosynthetic pathway responsible for
target traits was used as a reference.
QTL mapping
Linkage mapping was performed by composite interval Results

mapping (CIM) model in Windows QTL Cartographer
V2.5 software (Wang et al. 2012) with a high-density Phenotypic variation and correlations among trait
bin published genetic map (Wang et al. 2016b). Briefly,
the algorithm with a 10-centimorgan (cM) window and Overall, abundant variations and continuous distribution
a 1-cM walking step were used in this model. The in the three pod color components (H, S, and V) and pod
logarithm of odds (LOD) cutoff was determined with size-related traits (PL, PW, PRLW, and PA) were ob-
1000 times permutation at P = 0.05. The QTL boundary served in the SAP and NJRINP (Fig. S2 and S3), indi-
was determined with a 1-drop of the LOD score from the cating polygenic inheritance of these traits. Moreover,
peak score. For individual environment, the average the significant differences of pod size-related traits were
value of the five biological replicates for 284 RILs was also observed between the two populations (Fig. S4). In
used for QTL mapping, whereas the BLUP values were the SAP population, H, S, and V varied from 29.66 to
used for the combined environment. QTL detected in 52.60, 0.12 to 0.86, and 0.25 to 1.00, with average
different environments and exhibited full or partial over- values of 40.40, 0.67, and 0.81, respectively. PL, PW,
lapping confidence intervals were regarded as the same PRLW, and PA varied from 3.83 to 7.07 cm, 1.10 to
QTL. The QTL identified in at least two environments 2.73 cm, 2.33 to 4.55, and 2.84 to 8.24 cm2, with
was treated as a stable one. Nomenclature of QTL was average values of 5.36 cm, 1.67 cm, 3.27, and 4.65
designated following McCouch’s description with mi- cm, respectively, over 2 years (Table S2). For the
nor modifications (McCouch et al. 1997). NJRINP population, H, S, and V ranged from 22.73 to
82.25, 0.19 to 0.87, and 0.26 to 1.00, with average
Identification of putative candidate genes values of 50.00, 0.50, and 0.60, respectively, over 2
years. PL, PW, PRLW, and PA ranged from 2.61 to
A genomic region flanking the physical position of the 5.52 cm, 0.67 to1.55 cm, 2.45 to 4.65, and 1.14 to 3.69
identified major QTLs in the GWAS panel (500-kb cm2, with average values of 3.48 cm, 1.06 cm, 3.35, and
upstream and downstream of the peak SNP) was used 2.01 cm2, respectively, over 2 years (Table S3). The
to identify candidate genes. The annotated genes within absolute values of skewness and kurtosis were less than
the physical interval of major QTLs were used for 1 for these traits except for the S values in the SAP
identification of candidate genes in the NJRINP population (skewness varied from 2.56 to 3.05 over 2
years) and PL in the NJRINP population (kurtosis=2.67 years (r=0.51–0.62, P<0.05). PW showed moderate to
in 2017), suggesting that these phenotypes conformed to high positive correlation with PA within years (r=0.77–
a normal or skew-normal distribution in the two popu- 0.82, P<0.05) and low to moderate correlation with that
lations. ANOVA indicated that the effects of genotypes, between years (r=0.45–0.51, P<0.05). PW was moder-
environments, and genotype-by-environment interac- ately correlated with both PL and PRLW within years
tion exhibited significant differences on all measured and exhibited low positive and negative correlations
traits in the two populations (Tables S1 and S2). The h2 with PL and PRLW between years, respectively. In
of the three-color components and four-size parameters addition, low negative correlations (lower than 0.5)
ranged from 52.42 to 91.64% in the SAP population. were observed between PA and PRLW both within
While in the NJRINP population, a relatively high h2 and between years except for the correlation between
(>70 %) was estimated for all traits except H component PA in 2018 and PRLW in 2017 (Fig. 1d).
for the PC. Due to the large difference in the H compo- To validate the results of the image-based phenotyp-
nent between 2 years, the h2 was separately estimated ing for pod size-related traits, we randomly selected a
for on a per-year basis as 42.71% in 2017 and 78.45% in sample from each of the populations across 2 years for
2018. The range of h2 suggested that the genetic effects hand measurement of these traits. As expected, high
mainly dominated the performance of these traits in both correlations were observed between size-related traits
populations. generated from image-based analysis and that measured
Pairwise correlations among the three pod color- manually by hand either in the SAP (r=0.81–0.98,
related components and four pod size-related parameters P<0.05) or NJRINP population (r=0.88–0.98, P<0.05)
in the two populations exhibited significant correlations over 2 years (Fig. S5), which demonstrated that the
either within or between years except for the correla- image-based phenotyping approach could be efficiently
tions between H and S as well as PRLW and PA in 2018 employed to measure these traits without the laboriously
(Fig. 1). Specifically, for pod color-related components manual work.
in the SAP population, V showed a positive correlation
with S either within or between years. H exhibited a low Genetic dissection of pod-related traits via GWAS
correlation with both S and V between years. However, and linkage mapping
the correlation between H and S was inconsistent be-
tween years (Fig. 1a). For pod size-related parameter, GWAS for pod-related traits in the SAP
PL was highly correlated with PA within years (r=0.92–
0.93, P<0.05) and showed moderate to high correlation To identify genomic regions governing pod-related
with that between years (r=0.79–0.82, P<0.05). PW traits, GWAS was performed in the SAP for the three
was moderately correlated with both PL and PRLW pod color-related components and four pod size-related
within years and exhibited low to moderate and low parameters using MLM approach with both the mean
correlations with PL and PRLW, respectively, between value of each trait in each year and BLUPs of individual
years. Additionally, low and moderate correlations were performance over 2 years. Because the multiple signif-
observed between PL and PRLW, PW, and PA, respec- icant trait-associated SNPs located in close physical
tively, either within or between years (Fig. 1b). In the proximity, we used the most peak SNP (the strongest
NJRINP population, except for the correlations between trait-associated SNP) to represent the trait-associated
H as well as PL and PRLW either within or between loci. A total of one locus for PC and 38 loci for PS were
years, correlations among all traits either within or be- identified across 14 of 20 soybean chromosomes at the
tween years were significant. In particular, the high and suggestive significant P-value level of 1×10−4 (Table 1;
moderate positive correlations were observed between Figs. 2 and 3). Among these 38 loci, 14 for PL, 6 for
V and S within years (r=0.92–0.96, P<0.05) and be- PW, 8 for PRLW, and 10 for PA were detected
tween years (r=0.57–0.68, P<0.05), respectively. How- (Table 1). Out of these, seven, 2, 2, and 6 loci were
ever, H showed positive and negative correlation with repetitively detected across the two environments (i.e.,
both S and V in 2017 and 2018, respectively (Fig. 1c). 2017 and 2018) for PL, PW, PRLW, and PA, respec-
Among pod size-related parameters, PL was highly tively. Some loci were found to be significantly associ-
correlated with PA within years (r=0.87–0.92, ated with more than one pod size-related parameter. For
P<0.05), and moderately correlated with that between instance, five loci significantly associated with PL (i.e.,
Fig. 1 Pearson correlation among traits in two populations. Pod PL to PW (PRLW), and pod area (PA) in SAP and NJRINP are
color-related traits including hue (H), saturation (S), and value (V) shown in b and d, respectively. Significant at P<0.05; both circle
in SAP and NJRINP are shown in a and c, respectively. Pod size- color and size illustrate the correlation between pairs of traits; non-
related traits including pod length (PL), pod width (PW), ratio of significant correlations are indicated by a blank
PL2-2, PL7, PL14, PL15, and PL19-1) were also sig- components were detected in the SAP population over
nificantly associated with PA (Table 1). The locus at the two years.
40.4 Mb position on chromosome 15 was associated
with three PS parameters (PL, PW, and PA) with similar Linkage mapping of pod-related traits in the NJRINP
effects. Due to the existence of linkage disequilibrium
(LD) (Fig. S6), the physical region of PC3 was finally A total of two major QTLs, qPC3 and qPC19, associ-
identified at the 0.2~1.2 Mb on chromosome 3 for ated with both V component and PC were identified on
further identification of putative candidate genes. There chromosomes 3 and 19, respectively, in the NJRINP
were thirty-six SNPs within this 1Mb region. Among across all environments (Table 2; Fig. 5). The qPC3 was
them, six SNPs were significantly associated with target mapped to the interval of 3.3–7.1 cM on chromosome 3,
trait. In addition, the haplotype block analysis revealed with LOD values ranged from 5.74 to 11.11 and
that seven blocks located in this region, and the high LD accounted for 6.89 to 9.73% phenotypic variation (R2).
level (r2=0.87) was observed between the peak SNP and The qPC19 was detected in the interval of 51.2–52.2 cM
its surrounding SNPs (Fig. 4). Except the PC3 for H on chromosome 19, with LOD scores and R2 values in
component, no loci associated with other pod color the range of 13.21–36.37 and 17.06–40.41 %,
Table 1 List of loci and SNPs significantly associated with pod-related traits and previously reported QTLs for pod color-and seed-related
traits in SAP
Traitsa Loci Gmb Peak SNP Position (bp) MAFc Envd -Log10Pe R2f R2g Effect Known QTLsh
PC_V PC3 3 Gm03_743679 743,679 0.18 2017JP 7.67 0.09 0.26 0.01 L2
2018JP 7.13 0.01 0.18 0.01
BLUP 8.95 0.08 0.28 0.00
PL PL2-1 2 Gm02_14232486 14,232,486 0.07 2018JP 4.59 0.45 0.50 −0.11 SLWR 1-2
PL2-2 2 Gm02_42639481 42,639,481 0.28 2017JP 4.58 0.39 0.46 −0.09 SW 49-8
2018JP 5.04 0.45 0.51 −0.15
BLUP 5.05 0.49 0.55 −0.11
PL3 3 Gm03_6682752 6,682,752 0.14 2018JP 4.26 0.45 0.50 −0.06
PL5 5 Gm05_8270215 8,270,215 0.13 2018JP 4.28 0.45 0.50 −0.04 SLWR 1-3
PL7 7 Gm07_8271611 8,271,611 0.17 2017JP 4.22 0.39 0.45 −0.21 SL 5-3
2018JP 4.37 0.45 0.50 −0.19
BLUP 4.38 0.49 0.54 −0.18
PL8 8 Gm08_5096428 5,096,428 0.10 BLUP 4.59 0.49 0.54 0.05
PL9 9 Gm09_3230814 3,230,814 0.33 2017JP 4.85 0.39 0.46 −0.02
BLUP 4.56 0.49 0.54 0.01
PL11-1 11 Gm11_5671948 5,671,948 0.10 BLUP 4.17 0.49 0.53 0.03 SW 49-1
PL11-2 11 Gm11_34192967 34,192,967 0.11 BLUP 4.26 0.49 0.54 0.09 SW 34-14
PL14 14 Gm14_1830770 1,830,770 0.11 2017JP 4.83 0.39 0.46 −0.02
BLUP 4.61 0.49 0.54 −0.04
PL15 15 Gm15_40405966 40,405,966 0.39 BLUP 4.36 0.49 0.54 −0.06 SWD 1-1
PL16 16 Gm16_35892219 35,892,219 0.45 2018JP 5.65 0.45 0.52 −0.08 SL 4-5; SLWR 1-9
BLUP 4.59 0.49 0.54 −0.07
PL19-1 19 Gm19_45415096 45,415,096 0.23 2018JP 5.67 0.45 0.52 0.15 SL 1-11; SWD 1-7
BLUP 4.70 0.49 0.54 0.12
PL19-2 19 Gm19_48388397 48,388,397 0.06 2017JP 4.73 0.39 0.46 −0.13 SW 4-6
BLUP 4.10 0.49 0.53 −0.12
PW PW1 1 Gm01_5135530 5,135,530 0.44 2017JP 4.59 0.15 0.24 0.05 SW 15-2
2018JP 4.01 0.22 0.29 0.04
BLUP 5.29 0.25 0.34 0.03
PW5 5 Gm05_40583499 40,583,499 0.26 2017JP 4.73 0.15 0.24 −0.02 SH 1-4
2018JP 4.34 0.22 0.30 −0.05
BLUP 4.91 0.25 0.33 −0.03
PW15-1 15 Gm15_17116589 17,116,589 0.07 2018JP 4.60 0.22 0.30 −0.02 SWD 1-1
PW15-2 15 Gm15_40405966 40,405,966 0.39 2018JP 4.29 0.22 0.30 −0.01
PW15-3 15 Gm15_46487361 46,487,361 0.08 2018JP 4.26 0.22 0.30 −0.02
PW19 19 Gm19_2253517 2,253,517 0.07 2018JP 4.80 0.22 0.31 −0.05
PRLW PRLW3 3 Gm03_591006 591,006 0.23 2018JP 4.38 0.02 0.11 −0.03
PRLW7 7 Gm07_37272576 37,272,576 0.10 2018JP 4.61 0.02 0.12 0.08
PRLW8 8 Gm08_3015607 3,015,607 0.13 2018JP 4.53 0.02 0.12 −0.01 SL 5-1
PRLW18-1 18 Gm18_4800828 4,800,828 0.31 2017JP 4.04 0.08 0.16 −0.01 SW/P 6-4
PRLW18-2 18 Gm18_59175484 59,175,484 0.07 2018JP 5.18 0.02 0.13 −0.01
PRLW19 19 Gm19_40492170 40,492,170 0.14 2018JP 4.12 0.02 0.11 0.03 SW 35-7
PRLW20-1 20 Gm20_34083917 34,083,917 0.36 2017JP 4.41 0.08 0.17 0.00 SLWR 3-1
2018JP 4.86 0.02 0.13 0.01
Table 1 (continued)
Traitsa Loci Gmb Peak SNP Position (bp) MAFc Envd -Log10Pe R2f R2g Effect Known QTLsh
BLUP 4.78 0.07 0.17 0.02

PRLW20-2 20 Gm20_40707569 40,707,569 0.21 2018JP 4.40 0.02 0.11 −0.12 SW 4-9
BLUP 4.53 0.07 0.16 −0.12
PA PA1 1 Gm01_32903288 32,903,288 0.05 2017JP 4.20 0.41 0.46 −0.07 SW 35-10
PA2 2 Gm02_42639481 42,639,481 0.28 2017JP 4.02 0.41 0.46 −0.23 SW 49-8
2018JP 4.95 0.47 0.53 −0.26
BLUP 4.83 0.50 0.55 −0.23
PA7 7 Gm07_8271611 8,271,611 0.17 2018JP 4.07 0.47 0.52 −0.28 SL 5-3
BLUP 4.19 0.50 0.54 −0.26
PA8 8 Gm08_14217510 14,217,510 0.11 2018JP 4.06 0.47 0.52 0.01
BLUP 4.04 0.50 0.54 0.01
PA11 11 Gm11_28007526 28,007,526 0.20 BLUP 4.06 0.50 0.54 −0.04
PA14 14 Gm14_1830770 1,830,770 0.11 BLUP 4.09 0.50 0.54 0.02
PA15 15 Gm15_40405966 40,405,966 0.39 2018JP 4.05 0.47 0.52 −0.13 SWD 1-1
BLUP 4.54 0.50 0.55 −0.05
PA16 16 Gm16_36289049 36,289,049 0.20 2017JP 4.43 0.41 0.47 0.02 SL 4-5; SLWR 1-9
2018JP 4.28 0.47 0.52 0.05
BLUP 4.34 0.50 0.54 0.04
PA19 19 Gm19_45415096 45,415,096 0.23 2017JP 4.58 0.41 0.47 0.15 SL 1-11; SWD 1-7
2018JP 4.47 0.47 0.52 0.14
BLUP 5.50 0.50 0.56 0.12
PA20 20 Gm20_35141779 35,141,779 0.11 2017JP 4.30 0.41 0.47 0.13 SLWR 3-1
a
Pod color- and size-related traits. PC_V represents value component of pod color; PL, PW, PRLW, and PA represent pod length, pod
width, pod ratio of length to width, and pod area, respectively
b
Chromosomes of soybean genome
c
Minor allele frequency.
d
2017JP, 2018JP and BLUP represent Jiangpu 2017, Jiangpu 2018, and the combined environment
e
Negative log10-transformed P value
f
The contribution rate of model without SNP
g
The contribution rate of model with SNP
h
Based on the QTL list on SoyBase (www.soybase.org) and previous report. SW seed weight, SL seed length, SWD seed width, SH seed
height, SLWR seed length to width ratio, SW/P seed weight per plant
respectively. To verify the accuracy of the mapped surprisingly, these two QTLs revealed the negative ad-
QTLs for pod color based on the quantitative phenotyp- ditive effect when using the qualitative phenotypic data,
ic data generated from image analysis, we employed the while the positive effect was detected when the quanti-
qualitative phenotypic data obtained from visual inspec- tative phenotypic data were used, implying that the
tion to perform QTL mapping. The results obtained additive alleles were inherited from the opposite parents
from the qualitative data were similar to those from the when using these two different types of phenotypic data.
quantitative data based on image analysis. Interestingly, In total, twenty-three QTLs for pod size-related pa-
the relatively higher LOD values were observed for the rameters were identified in the NJRINP across different
QTLs detected by the quantitative data than those de- environments (Table 2; Fig. S7). Out of these QTLs, 8
tected by the qualitative data in most environments (Fig. for PL, 5 for PW, 3 for PRLW, and 7 for PA were
5), indicating the more powerful of employing the quan- detected (Table 2). Eleven QTLs were repetitively de-
titative phenotypic data for mining QTL. However, tected for pod size-related parameters in at least two
Fig. 2 Manhattan plots of the

association analysis for V
component in SAP. Manhattan
plot of V on 20 chromosomes
across different environments.
The dotted red line indicates the
significance threshold of P-value
1×10−4. The significant SNPs are
labeled with red dots. The
outermost ring strip represents the
distribution of SNP markers on 20
chromosomes of soybean. The
color represents the density of the
SNP markers; the inner three
circles from the inside to the
outside represent three
environments, 2017, 2018, and
BLUP, respectively
environments (either 1 year with BLUP or both years linkage mapping. Annotation of the physical interval of
with or without BLUP). Of these, four QTLs (qPL10-1, 491,890–847,720 and 37,463,301–37,938,066 of qPC3
qPL17-1, qPW17-1, qPA17) were stably detected in 2 (PC3) and qPC19 to the SoyBase database identified 35
years and BLUP. Besides, more than half of these QTLs and 48 candidate genes, respectively. Among the model
were only detected in an individual environment. The R2 genes within the physical regions of the qPC3, the
of these detected QTLs ranged from 3.66 (qPL2) to MATE (multidrug and toxic compound extrusion) gene
14.17 % (qPA17). Notably, one QTL was identified cluster that contained four MATE genes
for both PL (qPL17-1) and PA (qPA17) and located in (Glyma03g00770, Glyma03g00780, Glyma03g00790,
the interval of 41.8–45.4 cM on chromosome 17 with and Glyma03g00830) and two MYB (myeloblastosis)
LOD scores and R2 values ranging from 6.13 to 11.03 genes (Glyma03g00890 and Glyma03g00980) were
and 7.17 to 14.17%, respectively. Most identified QTLs identified. Similarly, another four MATE genes
such as qPL17-1 and qPA17 overlapped with previously (Glyma19g29860, Glyma19g29870, Glyma19g29940,
mapped QTLs for soybean seed size-related traits and Glyma19g29970) and two MYB genes
(Table 2), suggesting a potential genetic relationship (Glyma19g29750 and Glyma19g30220) were identified
between pod size and seed size in soybean. These QTLs within the physical region of qPC19 (Table S4). The
exhibited the positive additive effects, indicating that expression profiling of the corresponding gene models
female parent NN86-4 contributed the beneficial alleles in the Chinese soybean (ZH13) genome was utilized to
with exception of qPRLW9 that had its beneficial alleles investigate the pod-related tissues expression level of
from the male parent PI 342618B. the candidate genes. A BLAST search against the ZH13
reference genome revealed that among the identified
Identification of candidate genes MATE genes, two genes, i.e., Glyma03g00830 and
Glyma19g29970, whose corresponding genes in ZH13
Identification of candidate genes for major QTLs genome have higher expression level in pod tissue at
underlying pod color four different developmental stages (2-week-old pods
with seed, 3-, 4-, and 5-week-old pods only). Further-
To further dissect the molecular basis of pod color- and more, another gene within the qPC19 genomic region,
size-related traits, we identified the candidate genes for Glyma19g29880, and the corresponding gene in ZH13
two major QTLs detected either through GWAS and genome, SoyZH13_19G106100, have the highest
Fig. 3 Manhattan plots of the association analysis for pod size- represents the distribution of SNP markers on 20 chromosomes of
related traits in SAP. Manhattan plots of pod size-related traits on soybean. The color represents the density of the SNP markers; the
20 chromosomes across different environments. The dotted red inner three circles from the inside to outside represent three envi-
line indicates the significance threshold of P-value 1×10−4. The ronments, 2017, 2018, and BLUP, respectively. a PL. b PW. c
significant SNPs are labeled with red dots. The outermost ring strip PRLW. d PA
expression level in pod tissue at the same four different distance of the SAP used in this study, the regions
developmental stages (Table S5). within 500 kb on either side of the eight loci were used
for identification of candidate genes. As a result, the
Identification of candidate genes for major QTLs corresponding 228, 213, 99, and 389 annotated genes
associated with pod size-related traits were found in these regions, of which, 34 candidate
genes were further identified within these QTLs based
As to pod size-related traits, among the 38 loci identified on gene ontology (GO) annotation, implicated mainly
in the SAP, eight loci including 2 for PL (PL2-2 and on the GO biological processes such as cell prolifera-
PL7), 2 for PW (PW1 and PW5), 1 for PRLW tion, cell division, elongation, multiplication, cell wall
(PRLW20-1), and 3 for PA (PA2, PA16 and PA19), biosynthesis, and fruit growth and development, as well
marked in bold in Table 1, were simultaneously identi- as response to phytohormone signaling and biosynthesis
fied in all environments, indicating that they were more pathways (Table S6).
stable. These loci were used to retrieve model genes A total of 483 candidate genes were identified within
from the SoyBase database. Considering the LD decay the physical regions of four major pod size-related QTLs
is twice higher than the average expression level of the

genome-wide genes over these three stages of pod de-
velopment (the normalized FPKM value is 7.74).
Furthermore, several promising candidate genes
were mined based on the function of their homologs in
other plant species. For instance, two protein families,
i.e., ovate protein family (OPF) and IQ67-domain (IQD)
protein which coordinate organ development, have been
reported to regulate organ size and shape in multiple
plant species (Snouffer et al. 2020). A candidate gene in
the qPA19, Glyma19g38480, was annotated as an OFP
homolog and homologous to Arabidopsis AT2G36026
(OFP13) (Table S7). Three candidate genes,
Glyma01g05100, Glyma20g31800, and
Glyma17g10660, encode IQD proteins two of which
were associated with PW1 and PR20-1 in the SAP,
respectively, and the other one associated with the
qPW17-1 in the NJRINP are homologs to Arabidopsis
AT1G14380, AT1G74690, and AT5G03040, respective-
ly (Table S7). Moreover, the candidate gene
Glyma17g10660 has high expression level at the three
pod developmental stages mentioned above, with the
normalized FPKM values ranging from 9 to 20
(Table S5).
Fig. 4 Local Manhattan plot and LD heatmap of PC3. The top
panel shows the local Manhattan plot for V component using the
BLUP value, with a 0.5-Mb region on each side of the peak SNP Discussion
(SNP with the lowest P value). The red line indicates the signifi-
cance threshold of P-value (8.12×10−7, 0.05/61541), and the blue
one indicates the suggestive significance threshold of P-value of Image-based phenotyping accelerates the genetic
1×10−4. The two of six significant SNP are highlighted with red dissection for pod-related traits
dots. The bottom panel depicts the extent LD and LD block in this
candidate region. The LD blocks within this region are indicated
with the black border. The blue asterisks indicate the position of Pod-related traits such as PC and PS are essential mor-
the above significant SNPs phological characteristics which are closely related to
the variety description and seed yield in soybean breed-
ing programs (Yang et al. 2008; Bernard 1967). How-
(159 for qPL10-1, 78 for qPL17-1(qPA17), and 246 for ever, because of being time-consuming, labor-intensive,
qPW17-1) detected in the NJRINP. Among these iden- and often influenced by the changing environmental
tified candidate genes, thirty-five genes were further factors and the observer subjectivity, the conventional
identified underlying these QTLs based on the same phenotyping approaches are becoming one of the limit-
terms of biological process (Table S6). On the other ing factors for soybean genetic analysis and molecular
hand, the expression profiles of those candidate genes breeding in the past few decades (Furbank and Tester
in pod tissue were further examined. According to the 2011). With the rapid development of crop phenomics,
SoyBase RNA-seq Atlas, among the selected candidate several phenotyping platforms and approaches have
genes, five genes, i.e., Glyma01g05410, been developed (Vasseur et al. 2018). Image-based phe-
Glyma05g37190, Glyma05g37470, Glyma17g12880, notyping has been widely applied in plant phenotyping
and Glyma17g13470, have high expression levels at across multiple scales from cellular to organ level and
three stages of pod development (1-cm-long pod, pod from single plants to population level. Compared to the
shell 10 days after flowering (DAF), and pod shell 14 conventional phenotyping approach, it provides advan-
(DAF)). Moreover, the expression level of these 5 genes tages in terms of cost, efficiency, and level of
throughput (Zhao et al. 2019). In the current study, potential genomic regions for pod size-related traits as
image-based phenotyping using high-resolution images much as possible. Both populations exhibited wide phe-
of the soybean three-seeded pods was employed to notypic variations and continuous distribution in pod
extract the basic parameters of pod color and size in color- and size-related traits, indicating quantitative
two different types of population. To validate the reli- polygenic inheritance of these traits. The high heritabil-
ability of the generated phenotypic data on pod size- ity (0.80 and 0.79) were observed for the V component
related parameters, we manually measured in a subset in the SAP and NJRINP populations, respectively, indi-
containing random samples from each of the two popu- cating that the V component could be implemented as a
lations. A highly positive correlation between the phe- stable phenotypic indicator facilitated GWAS and QTL
notypic data obtain from either phenotyping approaches mapping for pod color. Meanwhile, pod size-related
was observed, suggesting that the image-based pheno- parameters revealed the high heritability higher than
typing could be efficiently implemented to generate the 0.80 and 0.70 in the SAP and NJRINP populations,
accurate and reliable quantitative and high-throughput respectively. The estimated heritability in both popula-
phenotypic data in a shorter time with less effort and tions indicated that the genetic effects mainly dominated
costs. the performance of the traits.
GWAS and linkage mapping have been extensively Through performing GWAS with 61,541 SNP
used to analyze the genetic basis of quantitative traits. markers, we identified a major pod color locus and 38
GWAS is a powerful tool for identification of the chro- pod size-related parameters loci in the SAP population
mosomal regions associated with target traits by taking (Table 1). Among the pod size-related parameter loci,
full advantage of historical recombination in natural eight loci (PL2-2, PL7, PW1, PW5, PRLW20-1, PA2,
populations (Che et al. 2020). Conventional QTL map- PA16, and PA19) were identified for pod size-related
ping usually relies on linkage analysis in biparental traits across all environments (2017, 2018, and BLUP).
populations. A combination of the two approaches On the other hand, two and twenty-three QTLs were
complemented with a high-throughput phenotyping ap- identified for pod color- and pod size-related traits,
proach might comprehensively elucidate the genetic respectively, in the NJRINP population with 5728 bin
architecture of complex traits. markers and R2 values of 3.66–40.41%. Of these, two
QTLs (qPC3 and qPC19) for PC and four QTLs
Dissecting the genetic basis of pod-related traits (qPL10-1, qPL17-1, qPW17-1, and qPA17) for PS-
related parameters were detected in all environments
Although previous studies have dissected the genetic (Table 2). These QTLs identified by association and
basis of several pod-related traits including pod wall linkage mapping revealed that PC and PS are controlled
thickness, pod number and maturity date, and pod color by both major and minor QTL which partially agree
in soybean (Guo et al. 2011; Zhang et al. 2010; Yang with the report in other crop species where the minor
et al. 2013; Fang et al. 2017; Bandillo et al. 2017; Song QTLs also play important roles in the genetic architec-
et al. 2004), uncovering the genetic control for pod size- ture of fruit traits (Zhou et al. 2019).
related traits is still very limited. Since the PS is the The stable QTLs are requisite for their utilization in
important component of soybean seed yield, the identi- plant breeding, hence, the QTLs including PC3, PL2-2,
fication of major QTL controlling them across different PL7, PW1, PW5, PRLW20-1, PA2, PA16, and PA19
environments is of great importance for fine mapping identified in the SAP and qPC3, qPC19, qPL10-1,
and MAS breeding of new varieties with the desirable qPL17-1, qPW17-1, and qPA17 detected in the NJRINP
pod traits. In the present study, two diverse populations, (Tables 1 and 2) might be targeted for future soybean
including a natural population with 187 breeding lines breeding with the specific PC and PS traits. Besides, the
and an inter-specific bi-parental population derived two identified QTLs (qPC3 (PC3) and qPC19) for PC
from a cross between two genotypes possessing the overlapped with the reported classical loci controlling
distinct pod characteristics with 284 lines, were evalu- pod color in soybean, L1 and L2, which located on the
ated for pod color- and size-related parameters. Two central genomic region of the chromosome 19 and the
QTL mapping approaches were implemented to inves- top of the chromosome 3, respectively (Bernard 1967;
tigate the power and reliability of the data generated by Bandillo et al. 2017; He et al. 2015; Song et al. 2004).
image-based phenotyping approach and to mine the Moreover, these two loci interact with each other in a
Table 2 List of QTLs for pod-related traits and previously reported QTLs for pod color- and seed-related traits in NJRINP
Traitsa QTL Gmb Peak position (cM) LODc CI (cM)d PI (bp)e AEf R2 (%)g Envh Known QTLsi
PC qPC3 3 3.81 7.31 3.3–4.8 491,890–847,720 −0.19 7.30 2017JP L2

28 Page 14 of 21
V 3.81 9.71 3.3–7.1 0.06 8.04 2017JP

5.71 5.74 3.7–6.5 0.04 6.89 2018JP
3.81 11.11 3.3–4.4 0.04 9.73 BLUP
PC qPC19 19 51.71 27.44 51.2–52.2 37,463,301–37,938,066 −0.41 33.06 2017JP L1
V 51.71 36.37 51.2–52.2 0.13 40.41 2017JP
52.11 13.21 51.8–57.1 0.07 17.06 2018JP
51.71 33.01 51.2–52.0 0.08 36.35 BLUP
PL qPL2 2 40.91 3.67 39.7–42.5 8,492,250–9,746,989 0.04 3.66 BLUP SWD 4-1
qPL10-1 10 106.31 6.78 105.5–108.0 49,853,666–50,958,736 0.11 8.32 2017JP
109.11 6.53 106.7–109.8 0.09 7.59 2018JP
109.81 8.75 109.5–110.2 0.07 9.63 BLUP
qPL10-2 10 8.51 4.12 7.2–10.4 1,524,835–2,400,798 0.05 4.39 BLUP SW 12-6
qPL12 12 50.41 6.09 50.0–50.9 15,840,212–18,643,308 0.09 7.00 2018JP
qPL15 15 35.51 4.35 34.7–37.7 8,204,888–8,738,986 0.08 5.23 2017JP
35.51 3.93 34.7–36.7 0.05 4.17 BLUP
qPL17-1 17 42.31 9.97 41.8–44.7 9,606,364–10,335,777 0.13 12.70 2017JP SW 13-5
42.31 6.13 41.8–44.5 0.09 7.17 2018JP
43.91 10.07 42.8–45.4 0.08 11.25 BLUP
qPL17-2 17 33.91 6.11 33.2–34.6 8,060,798–8,430,194 0.06 7.04 BLUP SH 1-8
qPL19 19 86.41 4.49 84.5–87.8 43,900,720–44,450,021 0.08 5.10 2018JP SL 1-11; SWD 1-8
PW qPW2 2 30.91 6.28 30.3–32.5 6,748,351–7,107,693 0.04 8.29 2017JP SW 1-9; SH 2-1
31.21 4.78 30.6–32.9 0.02 5.77 BLUP
qPW4 4 61.91 4.38 60.2–62.4 41,414,649–42,555,654 0.02 5.28 BLUP SL 1-15
qPW13 13 123.91 4.59 122.4–124.9 41,263,865–41,899,830 0.04 5.97 2017JP
123.91 3.56 123.1–125.2 0.02 4.29 BLUP
qPW17-1 17 27.71 3.93 26.5–28.9 6,278,808–8,235,617 0.04 5.17 2017JP SW 18-2
30.31 3.85 29.1–31.2 0.04 5.23 2018JP
32.51 6.31 32.4–33.5 0.03 7.69 BLUP
qPW17-2 17 42.91 5.96 41.9–45.0 9,504,436–10,335,777 0.03 7.29 BLUP
PRLW qPRLW5 5 106.41 3.60 104.7–110.4 37,774,179–38,282,413 0.09 4.90 2017JP SW 34-9
106.41 3.97 104.2–110.4 0.06 5.01 BLUP
Table 2 (continued)
Traitsa QTL Gmb Peak position (cM) LODc CI (cM)d PI (bp)e AEf R2 (%)g Envh Known QTLsi
qPRLW9 9 86.51 3.85 85.8–88.6 42,352,126–43,055,603 −0.08 5.25 2018JP SW/P 2-1
89.51 4.09 87.8–90.6 −0.06 5.12 BLUP
qPRLW12 12 36.41 4.80 34.5–37.0 6,959,753–7,626,171 0.06 6.06 BLUP SW 23-2; 44-1
PA qPA2-1 2 31.21 5.84 30.5–32.9 6,662,983–7,553,487 0.10 7.11 2017JP SH 2-1

33.71 3.78 30.9–34.3 0.09 4.73 2018JP
qPA2-2 2 40.91 6.28 40.2–43.7 8,523,421–10,313,659 0.07 7.16 BLUP SWD 4-1
qPA7 7 72.91 3.70 72.3–73.9 37,415,278–37,870,092 0.09 4.62 2018JP SW/P 6-9
qPA9 9 39.61 4.92 38.8–40.2 9,729,212–12,077,804 0.10 6.02 2017JP SW 34-6; 36-6
qPA10 10 9.21 3.77 7.3–9.9 1,524,835–2,517,319 0.08 4.54 2017JP SW 12-6
8.51 3.84 7.2–10.6 0.06 4.05 BLUP
qPA13 13 121.11 4.78 120.4–122.1 40,655,672–41,263,864 0.06 5.40 BLUP
qPA17 17 43.91 11.03 42.7–45.1 9,606,364–10,335,777 0.14 14.17 2017JP SW 13-5
42.31 6.55 41.8–44.9 0.12 8.96 2018JP
43.91 9.77 42.2–45.1 0.09 11.60 BLUP
a
Pod color- and size-related traits. V represents value component of pod color (PC); PL, PW, PRLW, and PA represent pod length, pod width, pod ratio of length to width, and pod area,
respectively
b
Chromosomes of soybean genome
c
The logarithm of odds (LOD) value at the peak position of the QTL
d
1-LOD confidence interval (cM)
e
The corresponding physical interval (bp)
f
Additive effect, the positive value indicates that NN86-4 contributed the increased allele for traits; otherwise, the increased allele for traits was contributed by PI342618B
g
Phenotypic variation explained by the QTL
h
2017JP, 2018JP, and BLUP represent Jiangpu 2017, Jiangpu 2018, and the combined environment
i
Based on the QTL list on SoyBase (www.soybase.org) and previous report. SW seed weight, SL seed length, SWD seed width, SH seed height, SLWR seed length to width ratio, SW/P seed
weight per plant
Page 15 of 21 28
Fig. 5 Graphs of QTL mapping for two major loci controlling pod color in NJRINP. a qPC3. b qPC19.
dominant epistatic manner to produce a F2 phenotypic cultivar NN86-4 and the small-podded wild accession
segregation ratio of 12 (Bl): 3 (Br): 1(Tn) (Bandillo et al. PI 342618B. Additionally, the results further showed
2017). In addition, the qPC19 also overlapped with a that most QTLs identified in our study have been pre-
previously reported QTL hotspot regulating the antiox- viously reported for seed-related traits, suggesting the
idants, phenolics, and flavonoids contents in soybean presence of an inheritance relationship between the size
seeds (Li et al. 2016b). These results give credence to of pod and seed in soybean. Besides, a previously re-
the reliability of the current mapping for the image- ported improvement-selective sweep partially over-
based phenotypes. However, the detected QTLs under l ap p e d w i t h t he g en o m i c r e g i on o f q P C 1 9
the specific environment could be explained by the (37,463,301-37,938,066), and six selective sweep re-
significant influence of environment on these traits as gions overlapped with the three loci identified for PL
revealed by ANOVA (Table S2 and S3). Similarly, the via GWAS, i.e., PL7, PL9, and PL14, and three QTLs
colocalization of genomic regions regulating more than detected for that by linkage mapping, i.e., qPL2, qPL10-
one trait may be the genetic basis for the strong to weak 1, and qPL15 (Table S8). Considering the pod-related
correlation among some of the traits. This phenomenon traits such as PC and PS largely diverged from the wild
may be due to the presence of tightly linked genes that to cultivated soybeans, thus some of these traits might
control the independent traits or possible pleiotropic be targeted by selection in breeding programs.
effects. For instance, PA2, PA7, PA14, PA15, and
PA19 for PA colocalized with the genomic regions of Identifying potential candidate genes within the stable
PL2-2, PL7, PL14, PL15, and PL19-1 for PL on chro- QTLs
mosomes 2, 7, 14, 15, and 19, respectively, in the
GWAS study. Furthermore, the colocalization of Candidate gene analyses are necessary for further gene
qPA17 for PA and qPL17-1 for PL was identified on cloning and functional verifications. Based on the two
chromosome 17 in the conventional QTL analysis. QTLs detected, thirteen candidate genes were identified
In comparison of the results from GWAS and linkage for PC (Table S4). Among them, a gene cluster includ-
mapping for pod size-related traits, no genomic regions ing four genes, Glyma03g00770, Glyma03g00780,
were overlapped between them possibly due to the Glyma03g00790, and Glyma03g00830, was annotated
genetic composition of the bi-parental population which as multidrug and toxic compound extrusion (MATE)
exhibited wide variations in pod size, or the two popu- transporter, and the remaining two genes,
lations used in this study have greatly phenotypic dif- Glyma03g00890 and Glyma03g00980, were annotated
ferences in pod size-related traits. The SAP population as MYB transcription factors (TFs). Interestingly, the
was composed of breeding lines generated by cultivated six genes with the same annotations, including 4 MATE
soybeans, while NJRINP population was derived from transporters of Glyma19g29860, Glyma19g29870,
the inter-specific cross between the large-podded Glyma19g29940, and Glyma19g29970 and 2 MYB
TFs of Glyma19g29750 and Glyma19g30220, were also Fruits such as grape (VvAM1 and VvAM3) and in apple
located in the qPC19 region. In addition, based on the (MdMATE1 and MdMATE2) were involved in the trans-
tissue expression level of the corresponding gene model port and accumulation of anthocyanin and PA (Gomez
in ZH13, we found that the corresponding genes of et al. 2009; Gomez et al. 2011; Frank et al. 2011).
Glyma03g00830 and Glyma19g29970 had higher ex- Therefore, we speculate that one or several of the above
pression level in pod tissue at different developmental genes might be the potential candidate genes involved in
stages than the other MATE genes detected. We also pod coloration in soybean based on the gene annotation
found another gene, Glyma19g29880, whose corre- and previous reports on the known genes involved in the
sponding gene in ZH13 had highest expression level in biosynthesis pathway related to fruit color in other
pod tissue at different developmental stages compared plants (Albert et al. 2014; Takanashi et al. 2014). More
to all other genes, which was annotated as alpha- attention should be paid to the genes highly expressed in
isopropylmalate synthase. pod tissue in the future. Facing such results and consid-
It is known that anthocyanins as one of major sec- ering the nature of soybean genome, we further specu-
ondary metabolites responsible for the coloration of lated that the genes controlling PC might undergo the
many fruits, flowers, leaves, and seeds (Lin-Wang gene duplication and subsequent functional divergence
et al. 2014). Previous studies have identified some key in soybean evolutionary process, which eventually leads
regulators of anthocyanin biosynthesis in various plants to the differentiation of pod color.
(Niu et al. 2010; Li et al. 2012; Espley et al. 2013; Zhang Based on the results of GWAS and linkage mapping
et al. 2014). Of them, a conserved MBW (MYB-Bhlh- in both populations, a total of 12 QTLs for pod size-
WD40) complex, consisting of R2R3-MYB, basic- related traits, including 4, 3, 1, and 4 for PL, PW,
helix-loop-helix (bHLH), and WD-repeat (WDR) pro- PRLW, and PA, respectively, were stably identified in
teins, is widely believed to regulate the common path- all environments. Sixty-nine candidate genes were pre-
way of anthocyanins and proanthocyanidins (PA) bio- dicted based on the biological process of GO with the
synthesis (Jaakola 2013; Albert et al. 2014). As a mem- terms of “cell proliferation, cell division, elongation,
ber of this complex, several MYB genes in Arabidopsis multiplication, cell wall biosynthesis, and fruit growth
and fruits have been reported to influence the organ and development as well as involved in phytohormones
coloration by regulating anthocyanin biosynthetic path- signaling and biosynthesis pathways” (Table S6). Ac-
way. For instance, in lettuce, AtMYB60 as an exogenous cording to the RNA-Seq Atlas in SoyBase, five of these
gene has been reported to participate in the regulation of candidate genes, Glyma01g05410, Glyma05g37190,
leaf pigmentation by inhibiting anthocyanin biosynthe- Glyma05g37470, Glyma17g12880, and
sis (Park et al. 2008). In strawberry, FvMYB10 as a key Glyma17g13470 were highly expressed in pod tissue
activator has been reported to participate in the regula- at three developmental stages, two of which have also
tion of leaves, flowers, and fruits coloration (Lin-Wang been proven to be highly expressed in pod tissue at four
et al. 2014). In pear and apple, PcMYB10 and stages of pod development by the previously published
MdMYB10, respectively, have also been shown to in- data (Shen et al. 2019) (Table S5). Among these five
volve in the regulation of anthocyanin biosynthesis (Li genes mentioned, the ZH13 corresponding gene to
et al. 2012; Espley et al. 2013). On the other hand, the Glyma01g05410 has the highest expression level in
fruit coloration may be related to not only the biosyn- pod tissue across the different developmental stages.
thesis pathway, but also the transport and accumulation In addition, three genes, Glyma01g05100,
process. Previous studies have reported that MATE Glyma17g10660, and Glyma20g31800, were found in
transporters were involved in the transport and accumu- the interval of PW1, qPW17-1, and PRLW20-1, respec-
lation of secondary metabolites such as flavonoid, an- tively, which were annotated as IQ67 Domain (IQD)
thocyanin, and PA (Takanashi et al. 2014). In proteins. One gene, Glyma19g38480, was found in the
Arabidopsis, AtTT12 was first MATE transporter found interval of PA19, which was annotated as ovate family
to transport flavonoids to alter seed coloration (Diener protein (OFP) (Table S7). The OFPs are the protein
et al. 2001). In Medicago, MtMATE1 and MtMATE2 family containing a conserved OVATE domain that
were involved in the transport of PA and flavonoid, regulates organ size and shape in plants. The IQD family
which influence the pigmentation of seeds, flowers, proteins are the calmodulin-binding proteins encoding a
and leaves (Zhao and Dixon 2009; Zhao et al. 2011). plant-specific domain of 67 conserved amino acid
residues (designated as IQ67 domain), which often these stable QTLs, 13 and 73 candidate genes were
serves as the regulator to participate in the complexes predicted for pod color-and size-related traits, respec-
of OFP to form organ shapes by regulating cytoskeleton tively. These QTLs and corresponding candidate genes
activities (Snouffer et al. 2020; Abel et al. 2005). Sev- provide important information for fine-mapping/map-
eral homologous genes of OFP and IQD in Arabidopsis based cloning and MAS for breeding elite varieties with
and other crops have been reported to participate in the desirable pod traits.
regulation of organ size and shape. For example, in
Arabidopsis, AtOFP1 has been shown to suppress cell
elongation (Wang et al. 2007). In rice, OsOFP19 has Supplementary Information The online version contains sup-
been reported to participate in the regulation of grain plementary material available at https://doi.org/10.1007/s11032-
021-01223-2.
width by increasing cell length and cell number (Yang
et al. 2018) and SlOFP20 in tomato (Wu et al. 2018).
For IQD, GSE5, the causal gene of GW5 locus that is Acknowledgements We appreciated Prof. Yufeng Wu and Dr.
widely utilized by rice breeders to control grain size, Wenkai Yan for their technical assistance. The work was support-
which encodes a protein with IQ domain, has been ed in part by the Bioinformatics Center of Nanjing Agricultural
demonstrated to participate in regulation of gain size University.
predominantly by influencing cell proliferation in spike-
Author contribution TJZ and HYJ conceived and designed the
let hulls (Duan et al. 2017; Liu et al. 2017). Another research experiments. FGC, PYLv, LYZ, and PX implemented the
known gene in rice, OsIQD14, has been reported to field experiments. FGC, SC, WHLv, YTX, WLY, and LW per-
control the shape and size of gain seeds (Yang et al. formed phenotyping. FGC and SC analyzed data. TJZ and HYJ
2020b). In tomato, SUN encoding a member of IQ67- designed the overall project. FGC drafted the manuscript. TJZ,
BK, and Abou-Elwafa revised the manuscript. All authors read
domain (IQD) family has been reported to govern the and approved the final manuscript.
elongated fruit shape (Xiao et al. 2008). Thus, these four
genes may also be potential candidate genes regulating Funding This work was supported by grants from the National
pod size and shape in soybean. Although we cannot Natural Science Foundation of China (No. 31872847, 31871646),
accurately determine which genes are directly related the National Key Research and Development Project
(2018YFD0201006), the MOE Program for Changjiang Scholars
to pod color and size based on the data of this study, we and Innovative Research Team in University (PCSIRT_17R55),
provide a list of candidate gene models, which could be the Fundamental Research Funds for the Central Universities
used as the useful information for future works. Further (KYT201801), and the Jiangsu Collaborative Innovation Center
studies should be conducted via fine-mapping of stable for Modern Crop Production (JCIC-MCP) Program.Availability
of data and materialsAll data generated and analyzed during this
QTLs and validating the molecular function of potential study are included in this article.
candidate genes discovered in this study.
Declarations
Conclusion Ethics approval and consent to participate Not applicable.
Consent for publication Not applicable.

A high-throughput image-based phenotyping approach
has been successfully employed for dissecting the ge-
Competing interests The authors declare no competing
netic control of pod color- and size-related traits using interests.
GWAS and linkage mapping strategies in two indepen-
dent populations. Two QTLs, qPC3 (PC3) and qPC19,
were identified to be associated with PC. Thirty-eight References
QTLs for pod size-related parameters, including 14 for
PL, 6 for PW, 8 for PRLW, and 10 for PA, were
identified by GWAS in the SAP population, and 23 Abel S, Savchenko T, Levy M (2005) Genome-wide comparative
QTLs containing 8 for PL, 5 for PW, 3 for PRLW, analysis of the IQD gene families in Arabidopsis thaliana and
Oryza sativa. BMC Evol Biol 5:72–72. https://doi.
and 7 for PA were detected by linkage mapping in the org/10.1186/1471-2148-5-72
NJRINP population. Twelve QTLs were stably identi- Abou-Elwafa SF, Shehzad T (2018) Genetic identification and
fied in all environments for the two populations. Around expression profiling of drought responsive genes in sorghum.
Environ Exp Bot 155:12–20. https://doi.org/10.1016/j. Duan P, Xu J, Zeng D, Zhang B, Geng M, Zhang G, Huang K,
envexpbot.2018.06.019 Huang L, Xu R, Ge S, Qian Q, Li Y (2017) Natural variation
Albert NW, Davies KM, Lewis DH, Zhang H, Montefiori M, in the promoter of GSE5 contributes to grain size diversity in
Brendolise C, Boase MR, Ngo H, Jameson PE, Schwinn rice. Mol Plant 10:685–694. https://doi.org/10.1016/j.
KE (2014) A conserved network of transcriptional activators molp.2017.03.009
and repressors regulates anthocyanin pigmentation in Espley RV, Bovy A, Bava C, Jaeger SR, Tomes S, Norling C,
eudicots. Plant Cell 26:962–980. https://doi.org/10.1105 Crawford J, Rowan D, McGhie TK, Brendolise C, Putterill J,
/tpc.113.122069 Schouten HJ, Hellens RP, Allan AC (2013) Analysis of
Ali A, Ullah Z, Alam N, Naqvi SMS, Jamil M, Bux H, Sher H genetically modified red-fleshed apples reveals effects on
(2020) Genetic analysis of wheat grains using digital imaging growth and consumer attributes. Plant Biotechnol J 11:408–
and their relationship to enhance grain weight. Sci Agric 77. 419. https://doi.org/10.1111/pbi.12017
https://doi.org/10.1590/1678-992x-2019-0069 Fahlgren N, Gehan MA, Baxter I (2015) Lights, camera, action:
Baek J, Lee E, Kim N, Kim SL, Choi I, Ji H, Chung YS, Choi MS, high-throughput plant phenotyping is ready for a close-up.
Moon JK, Kim KH (2020) High throughput phenotyping for Curr Opin Plant Biol 24:93–99. https://doi.org/10.1016/j.
various traits on soybean seeds using image analysis. Sensors pbi.2015.02.006
20:248. https://doi.org/10.3390/s20010248 Fang C, Ma Y, Wu S, Liu Z, Wang Z, Yang R, Hu G, Zhou Z, Yu
Bandillo NB, Lorenz AJ, Graef GL, Jarquin D, Hyten DL, Nelson H, Zhang M, Pan Y, Zhou G, Ren H, Du W, Yan H, Wang Y,
RL, Specht JE (2017) Genome-wide association mapping of Han D, Shen Y, Liu S, Liu T, Zhang J, Qin H, Yuan J, Yuan
qualitatively inherited traits in a germplasm collection. Plant X, Kong F, Liu B, Li J, Zhang Z, Wang G, Zhu B, Tian Z
Genome 10:18. https://doi.org/10.3835 (2017) Genome-wide association studies dissect the genetic
/plantgenome2016.06.0054 networks underlying agronomical traits in soybean. Genome
Barabaschi D, Tondelli A, Desiderio F, Volante A, Vaccino P, Biol 18:161. https://doi.org/10.1186/s13059-017-1289-9
Vale G, Cattivelli L (2016) Next generation breeding. Plant Fernandez-Gallego JA, Kefauver SC, Gutierrez NA, Nieto-
Sci 242:3–13. https://doi.org/10.1016/j.plantsci.2015.07.010 Taladriz MT, Araus JL (2018) Wheat ear counting in-field
Barrett JC, Fry B, Maller J, Daly MJ (2005) Haploview: analysis conditions: high throughput and low-cost approach using
and visualization of LD and haplotype maps. Bioinformatics RGB images. Plant Methods 14:22. https://doi.org/10.1186
21:263–265. https://doi.org/10.1093/bioinformatics/bth457 /s13007-018-0289-4
Bates D, Mächler M, Bolker B, Walker S (2015) Fitting linear Frank S, Keck M, Sagasser M, Niehaus K, Weisshaar B, Stracke R
mixed-effects models using lme4. J Stat Softw 67:1–48. (2011) Two differentially expressed MATE factor genes
https://doi.org/10.18637/jss.v067.i01 from apple complement the Arabidopsis transparent testa12
Bernard RL (1967) The inheritance of pod color in soybeans. J mutant. Plant Biol (Stuttg) 13:42–50. https://doi.org/10.1111
Hered 58:165–168. https://doi.org/10.1093/oxfordjournals. /j.1438-8677.2010.00350.x
jhered.a107575 Furbank RT, Tester M (2011) Phenomics-technologies to relieve
Che Z, Yan H, Liu H, Yang H, Du H, Yang Y, Liu B, Yu D (2020) the phenotyping bottleneck. Trends Plant Sci 16:635–644.
Genome-wide association study for soybean mosaic virus https://doi.org/10.1016/j.tplants.2011.09.005
SC3 resistance in soybean. Mol Breed 40. https://doi. Gage JL, de Leon N, Clayton MK (2018) Comparing genome-
org/10.1007/s11032-020-01149-1 wide association study results from different measurements
Chen D, Neumann K, Friedel S, Kilian B, Chen M, Altmann T, of an underlying phenotype. G3 (Bethesda) 8:3715–3722.
Klukas C (2014) Dissecting the phenotypic components of https://doi.org/10.1534/g3.118.200700
crop plant growth and drought responses based on high- Ghanem ME, Marrou H, Sinclair TR (2015) Physiological pheno-
throughput image analysis. Plant Cell 26:4636–4655. typing of plants for crop improvement. Trends Plant Sci 20:
https://doi.org/10.1105/tpc.114.129601 139–144. https://doi.org/10.1016/j.tplants.2014.11.006
Diaz-Garcia L, Covarrubias-Pazaran G, Schlautman B, Grygleski Gomez C, Terrier N, Torregrosa L, Vialet S, Fournier-Level A,
E, Zalapa J (2018a) Image-based phenotyping for identifica- Verries C, Souquet JM, Mazauric JP, Klein M, Cheynier V,
tion of QTL determining fruit shape and size in American Ageorges A (2009) Grapevine MATE-type proteins act as
cranberry (Vaccinium macrocarpon L.). PeerJ 6:e5461. vacuolar H+-dependent acylated anthocyanin transporters.
https://doi.org/10.7717/peerj.5461 Plant Physiol 150:402–415. https://doi.org/10.1104
Diaz-Garcia L, Schlautman B, Covarrubias-Pazaran G, Maule A, /pp.109.135624
Johnson-Cicalese J, Grygleski E, Vorsa N, Zalapa J (2018b) Gomez C, Conejero G, Torregrosa L, Cheynier V, Terrier N,
Massive phenotyping of multiple cranberry populations re- Ageorges A (2011) In vivo grapevine anthocyanin transport
veals novel QTLs for fruit anthocyanin content and other involves vesicle-mediated trafficking and the contribution of
important chemical traits. Mol Gen Genomics 293:1379– anthoMATE transporters and GST. Plant J 67:960–970.
1392. https://doi.org/10.1007/s00438-018-1464-z https://doi.org/10.1111/j.1365-313X.2011.04648.x
Diener AC, Gaxiola RA, Fink GR (2001) Arabidopsis ALF5, a Guo GY, Sun R, Hou M, Guo YX, Xin DW, Jiang HW, Liu CY,
multidrug efflux transporter gene family member, confers Hu GH, Chen QS (2011) Quantitative trait locus (QTL)
resistance to toxins. Plant Cell 13:1625–1638. https://doi. analysis of pod related traits in different environments in
org/10.1105/tpc.010035 soybean. Afr J Biotechnol 10:11848–11854
Dong Y, Yang X, Liu J, Wang BH, Liu BL, Wang YZ (2014) Pod Guo Z, Yang W, Chang Y, Ma X, Tu H, Xiong F, Jiang N, Feng
shattering resistance associated with domestication is medi- H, Huang C, Yang P, Zhao H, Chen G, Liu H, Luo L, Hu H,
ated by a NAC gene in soybean. Nat Commun 5:3352. Liu Q, Xiong L (2018) Genome-wide association studies of
https://doi.org/10.1038/ncomms4352 image traits reveal genetic architecture of drought resistance
in rice. Mol Plant 11:789–805. https://doi.org/10.1016/j. Genet Newsl 14:11–13. https://doi.org/10.1007/s10142-013-

molp.2018.03.018 0328-1
He Q, Yang H, Xiang S, Tian D, Wang W, Zhao T, Gai J, Singh R Neumann K, Zhao Y, Chu J, Keilwagen J, Reif JC, Kilian B,
(2015) Fine mapping of the genetic locusL1conferring black Graner A (2017) Genetic architecture and temporal patterns
pods using a chromosome segment substitution line popula- of biomass accumulation in spring barley revealed by image
tion of soybean. Plant Breed 134:437–445. https://doi. analysis. BMC Plant Biol 17:137. https://doi.org/10.1186
org/10.1111/pbr.12272 /s12870-017-1085-4
He J, Meng S, Zhao T, Xing G, Yang S, Li Y, Guan R, Lu J, Wang Niu SS, Xu CJ, Zhang WS, Zhang B, Li X, Lin-Wang K, Ferguson
Y, Xia Q, Yang B, Gai J (2017) An innovative procedure of IB, Allan AC, Chen KS (2010) Coordinated regulation of
genome-wide association analysis fits studies on germplasm anthocyanin biosynthesis in Chinese bayberry (Myrica rubra)
population and plant breeding. Theor Appl Genet 130:2327– fruit by a R2R3 MYB transcription factor. Planta 231:887–
2343. https://doi.org/10.1007/s00122-017-2962-9 899. https://doi.org/10.1007/s00425-009-1095-z
Hu D, Kan G, Hu W, Li Y, Hao D, Li X, Yang H, Yang Z, He X, Nyquist WE, Baker RJ (1991) Estimation of heritability and
Huang F, Yu D (2019) Identification of loci and candidate prediction of selection response in plant populations. Crit
genes responsible for pod dehiscence in soybean via genome- Rev Plant Sci 10:235–322. https://doi.org/10.1080
wide association analysis across multiple environments. /07352689109382313
Front Plant Sci 10:811. https://doi.org/10.3389 Park JS, Kim JB, Cho KJ, Cheon CI, Sung MK, Choung MG, Roh
/fpls.2019.00811 KH (2008) Arabidopsis R2R3-MYB transcription factor
Huang X, Wei X, Sang T, Zhao Q, Feng Q, Zhao Y, Li C, Zhu C, AtMYB60 functions as a transcriptional repressor of anthocya-
Lu T, Zhang Z, Li M, Fan D, Guo Y, Wang A, Wang L, Deng nin biosynthesis in lettuce (Lactuca sativa). Plant Cell Rep 27:
L, Li W, Lu Y, Weng Q, Liu K, Huang T, Zhou T, Jing Y, Li 985–994. https://doi.org/10.1007/s00299-008-0521-1
W, Lin Z, Buckler ES, Qian Q, Zhang QF, Li J, Han B (2010) Pascual L, Albert E, Sauvage C, Duangjit J, Bouchet JP, Bitton F,
Genome-wide association studies of 14 agronomic traits in Desplat N, Brunel D, Le Paslier MC, Ranc N, Bruguier L,
rice landraces. Nat Genet 42:961–967. https://doi. Chauchard B, Verschave P, Causse M (2016) Dissecting
org/10.1038/ng.695 quantitative trait variation in the resequencing era: comple-
Ibrahim AK, Zhang L, Niyitanga S, Afzal MZ, Xu Y, Zhang L, mentarity of bi-parental, multi-parental and association
Zhang L, Qi J (2020) Principles and approaches of associa- panels. Plant Sci 242:120–130. https://doi.org/10.1016/j.
tion mapping in plant breeding. Trop Plant Biol 13:212–224. plantsci.2015.06.017
https://doi.org/10.1007/s12042-020-09261-4
Qiu LJ, Chang RZ (2006) Descriptors and data standard for
Jaakola L (2013) New insights into the regulation of anthocyanin
soybean (Glycine spp.). China Agriculture Press, Beijing
biosynthesis in fruits. Trends Plant Sci 18:477–483.
Shen Y, Du H, Liu Y, Ni L, Wang Z, Liang C, Tian Z (2019)
https://doi.org/10.1016/j.tplants.2013.06.003
Update soybean Zhonghuang 13 genome to a golden refer-
Korte A, Farlow A (2013) The advantages and limitations of trait
ence. Sci China Life Sci 62:1257–1260. https://doi.
analysis with GWAS: a review. Plant Methods 9:29.
org/10.1007/s11427-019-9822-2
https://doi.org/10.1186/1746-4811-9-29
Li L, Ban ZJ, Li XH, Wu MY, Wang AL, Jiang YQ, Jiang YH Shin J-H, Blay S, McNeney B, Graham J (2006) LDheatmap: An
(2012) Differential expression of anthocyanin biosynthetic R function for graphical display of pairwise linkage disequi-
genes and transcription factor PcMYB10 in pears (Pyrus libria between single nucleotide polymorphisms. J Stat Softw
communis L.). PLoS One 7:e46070. https://doi.org/10.1371 16:1–9. https://doi.org/10.18637/jss.v016.c03
/journal.pone.0046070 Snouffer A, Kraus C, van der Knaap E (2020) The shape of things
Li L, Guo N, Niu J, Wang Z, Cui X, Sun J, Zhao T, Xing H to come: ovate family proteins regulate plant organ shape.
(2016a) Loci and candidate gene identification for resistance Curr Opin Plant Biol 53:98–105. https://doi.org/10.1016/j.
to Phytophthora sojae via association analysis in soybean [ pbi.2019.10.005
Glycine max (L.) Merr.]. Mol Gen Genomics 291:1095– Song QJ, Marek LF, Shoemaker RC, Lark KG, Concibido VC,
1103. https://doi.org/10.1007/s00438-015-1164-x Delannay X, Specht JE, Cregan PB (2004) A new integrated
Li MW, Munoz NB, Wong CF, Wong FL, Wong KS, Wong WH, genetic linkage map of the soybean. Theor Appl Genet 109:
Qi X, Li KP, Ng MS, Lam HM (2016b) QTLs regulating the 122–128. https://doi.org/10.1007/s00122-004-1602-3
contents of antioxidants, phenolics, and flavonoids in soy- Takanashi K, Shitan N, Yazaki K (2014) The multidrug and toxic
bean seeds share a common genomic region. Front Plant Sci compound extrusion (MATE) family in plants. Plant
7:854. https://doi.org/10.3389/fpls.2016.00854 Biotechnol 31:417–430. https://doi.org/10.5511
Lin-Wang K, McGhie TK, Wang M, Liu Y, Warren B, Storey R, /plantbiotechnology.14.0904a
Espley RV, Allan AC (2014) Engineering the anthocyanin Tang Y, Liu X, Wang J, Li M, Wang Q, Tian F, Su Z, Pan Y, Liu
regulatory complex of strawberry (Fragaria vesca). Front D, Lipka AE, Buckler ES, Zhang Z (2016) GAPIT Version 2:
Plant Sci 5:651. https://doi.org/10.3389/fpls.2014.00651 an enhanced integrated tool for genomic association and
Liu J, Chen J, Zheng X, Wu F, Lin Q, Heng Y, Tian P, Cheng Z, prediction. Plant Genome 9:1–9. https://doi.org/10.3835
Yu X, Zhou K, Zhang X, Guo X, Wang J, Wang H, Wan J /plantgenome2015.11.0120
(2017) GW5 acts in the brassinosteroid signalling pathway to Tian D, Wang P, Tang B, Teng X, Li C, Liu X, Zou D, Song S,
regulate grain width and weight in rice. Nat Plants 3:17043. Zhang Z (2020) GWAS Atlas: a curated resource of genome-
https://doi.org/10.1038/nplants.2017.43 wide variant-trait associations in plants and animals. Nucleic
McCouch SR, Cho YG, Yano M, Paul E, Blinstrub M, Morishima Acids Res 48:D927–D932. https://doi.org/10.1093
H, Kinoshita T (1997) Report on QTL nomenclature. Rice /nar/gkz828
Turner SD, Ellison SL, Senalik DA, Simon PW, Spalding EP, Yang C, Ma Y, He Y, Tian Z, Li J (2018) OsOFP19 modulates
Miller ND (2018) An automated image analysis pipeline plant architecture by integrating the cell division pattern and
enables genetic studies of shoot and root morphology in brassinosteroid signaling. Plant J 93:489–501. https://doi.
carrot (Daucus carota L.). Front Plant Sci 9:1703. org/10.1111/tpj.13793
https://doi.org/10.3389/fpls.2018.01703 Yang W, Feng H, Zhang X, Zhang J, Doonan JH, Batchelor WD,
Uzal LC, Grinblat GL, Namías R, Larese MG, Bianchi JS, Xiong L, Yan J (2020a) Crop phenomics and high-
Morandi EN, Granitto PM (2018) Seed-per-pod estimation throughput phenotyping: past decades, current challenges,
for plant breeding using deep learning. Comput Electron and future perspectives. Mol Plant 13:187–214. https://doi.
Agric 150:196–204. https://doi.org/10.1016/j. org/10.1016/j.molp.2020.01.008
compag.2018.04.024 Yang B, Wendrich JR, De Rybel B, Weijers D, Xue HW (2020b)
Vasseur F, Bresson J, Wang G, Schwab R, Weigel D (2018) Rice microtubule-associated protein IQ67-DOMAIN14 reg-
Image-based methods for phenotyping growth dynamics ulates grain shape by modulating microtubule cytoskeleton
and fitness components in Arabidopsis thaliana. Plant dynamics. Plant Biotechnol J 18:1141–1152. https://doi.
Methods 14:63. https://doi.org/10.1186/s13007-018-0331-6 org/10.1111/pbi.13279
Wang S, Chang Y, Guo J, Chen JG (2007) Arabidopsis Ovate Yano M, Tuberosa R (2009) Genome studies and molecular
Family Protein 1 is a transcriptional repressor that suppresses genetics-from sequence to crops: genomics comes of age.
cell elongation. Plant J 50:858–872. https://doi.org/10.1111 Curr Opin Plant Biol 12:103–106. https://doi.org/10.1016/j.
/j.1365-313X.2007.03096.x pbi.2009.01.001
Wang S, Basten CJ, Zeng ZB (2012) Windows QTL Cartographer Yin L, Zhang H, Tang Z, Xu J, Yin D, Zhang Z, Yuan X, Zhu M,
2.5. Department of Statistics, North Carolina State Zhao S, Li X, Liu X (2020) rMVP: A memory-efficient,
University, Raleigh visualization-enhanced, and parallel-accelerated tool for
Wang J, Chu S, Zhang H, Zhu Y, Cheng H, Yu D (2016a) genome-wide association study. bioRxiv. https://doi.
Development and application of a novel genome-wide SNP org/10.1101/2020.08.20.258491
array reveals domestication history in soybean. Sci Rep 6: Zhang D, Cheng H, Wang H, Zhang H, Liu C, Yu D (2010)
20728. https://doi.org/10.1038/srep20728 Identification of genomic regions determining flower and
Wang W, Liu M, Wang Y, Li X, Cheng S, Shu L, Yu Z, Kong J, pod numbers development in soybean (Glycine max L.). J
Zhao T, Gai J (2016b) Characterizing two inter-specific bin Genet Genomics 37:545–556. https://doi.org/10.1016
maps for the exploration of the QTLs/genes that confer three /s1673-8527(09)60074-6
soybean evolutionary traits. Front Plant Sci 7:1248. Zhang Y, Butelli E, Martin C (2014) Engineering anthocyanin
https://doi.org/10.3389/fpls.2016.01248 biosynthesis in plants. Curr Opin Plant Biol 19:81–90.
Wu S, Zhang B, Keyhaninejad N, Rodriguez GR, Kim HJ, https://doi.org/10.1016/j.pbi.2014.05.011
Chakrabarti M, Illa-Berenguer E, Taitano NK, Gonzalo MJ, Zhao J, Dixon RA (2009) MATE transporters facilitate vacuolar
Diaz A, Pan Y, Leisner CP, Halterman D, Buell CR, Weng uptake of epicatechin 3'-O-glucoside for proanthocyanidin
Y, Jansky SH, van Eck H, Willemsen J, Monforte AJ, Meulia biosynthesis in Medicago truncatula and Arabidopsis. Plant
T, van der Knaap E (2018) A common genetic mechanism Cell 21:2323–2340. https://doi.org/10.1105/tpc.109.067819
underlies morphological diversity in fruits and other plant Zhao J, Huhman D, Shadle G, He XZ, Sumner LW, Tang Y,
organs. Nat Commun 9:4734. https://doi.org/10.1038 Dixon RA (2011) MATE2 mediates vacuolar sequestration
/s41467-018-07216-8 of flavonoid glycosides and glycoside malonates in
Würschum T (2012) Mapping QTL for agronomic traits in breed- Medicago truncatula. Plant Cell 23:1536–1555. https://doi.
ing populations. Theor Appl Genet 125:201–210. https://doi. org/10.1105/tpc.110.080804
org/10.1007/s00122-012-1887-6 Zhao C, Zhang Y, Du J, Guo X, Wen W, Gu S, Wang J, Fan J
Xiao H, Jiang N, Schaffner E, Stockinger EJ, van der Knaap E (2019) Crop phenomics: current status and perspectives.
(2008) A retrotransposon-mediated gene duplication under- Front Plant Sci 10:714. https://doi.org/10.3389
lies morphological variation of tomato fruit. Science 319: /fpls.2019.00714
1527–1530. https://doi.org/10.1126/science.1153040 Zhou Y, Srinivasan S, Mirnezami SV, Kusmec A, Fu Q, Attigala
Yang Y, Jing C, Wang X, Lin C, Wang X, Zhou ZS (2008) L, Salas Fernandez MG, Ganapathysubramanian B, Schnable
Photosynthetic characteristics of soybean pod and its contri- PS (2019) Semiautomated feature extraction from RGB im-
bution to yield. J Northeast Agric Univ 39:51–56 885 ages for sorghum panicle architecture GWAS. Plant Physiol
Yang Z, Xin D, Liu C, Jiang H, Han X, Sun Y, Qi Z, Hu G, Chen 179:24–37. https://doi.org/10.1104/pp.18.00974
Q (2013) Identification of QTLs for seed and pod traits in
soybean and analysis for additive effects and epistatic effects Publisher’s note Springer Nature remains neutral with regard to
of QTLs among multiple environments. Mol Gen Genomics jurisdictional claims in published maps and institutional
288:651–667. https://doi.org/10.1007/s00438-013-0779-z affiliations.

Chang2021 Article ExploringGeneticArchitectureFo

Uploaded by

Copyright:

Available Formats

Chang2021 Article ExploringGeneticArchitectureFo

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chang2021 Article ExploringGeneticArchitectureFo

Uploaded by

Copyright:

Available Formats

Mol Breeding (2021) 41: 28

Exploring genetic architecture for pod-related traits

F. Chang : W. Lv : P. Lv : Y. Xiao : W. Yan : L. Zheng :

language and platform, respectively, and a series of Genotypic data

Linkage mapping was performed by composite interval Results

BLUP 4.78 0.07 0.17 0.02

Fig. 2 Manhattan plots of the

is twice higher than the average expression level of the

PC qPC3 3 3.81 7.31 3.3–4.8 491,890–847,720 −0.19 7.30 2017JP L2

V 3.81 9.71 3.3–7.1 0.06 8.04 2017JP

PA qPA2-1 2 31.21 5.84 30.5–32.9 6,662,983–7,553,487 0.10 7.11 2017JP SH 2-1

Conclusion Ethics approval and consent to participate Not applicable.

Consent for publication Not applicable.

in rice. Mol Plant 11:789–805. https://doi.org/10.1016/j. Genet Newsl 14:11–13. https://doi.org/10.1007/s10142-013-

You might also like