Identification of a novel seed size associated locus SW9-1 in soybean

Jiji Li, Jinghui Zho,Yinghui Li, Yli Go, Sunn Hu,Muhmm Neem,Genlou Sun,c, Wenming Zhng, Jinfeng Hou, Xioo Wng,*, Lijun Qiu,*

aSchool of Agronomy, Anhui Agricultural University,Hefei 230036,Anhui,China

bThe National Key Facility for Crop Gene Resources and Genetic Improvement(NFCRI)/Key Laboratory of Crop Gene Resource and Germplasm Enhancement(MOA),Institute of Crop Sciences,Chinese Academy of Agricultural Sciences,Beijing 100081,China

cBiology Department, Saint Mary's University,Halifax,NS B3H 3C3,Canada

dCollege of Horticulture,Anhui Agricultural University,Hefei 230036,Anhui,China

Keywords:Soybean [Glycine max (L.)Merr.]Seed size/weight SW9-1 Molecular breeding

A B S T R A C T Seed size is one of the vital traits determining seed appearance,quality,and yield.Untangling the genetic mechanisms regulating soybean 100-seed weight(100-SW),seed length and seed width across environments may provide a theoretical basis for improving seed yield.However,there are few reports related to QTL mapping of 100-SW across multiple ecological regions. In this study,21 loci associated with seed size traits were identified using a genome-wide association of 5361 single nucleotide polymorphisms (SNPs) across three ecoregions in China, which could explain 8.12%-14.25%of the phenotypic variance respectively.A new locus,named as SW9-1 on chromosome 9 that explained 10.05%-10.93%of the seed weight variance was found significantly related to seed size traits,and was not previously reported.The selection effect analysis showed that SW9-1 locus has a relatively high phenotypic effect (13.67) on 100-SW, with a greater contribution by the accessions with bigger seeds (3.69) than the accessions with small seeds(1.66).Increases in seed weight were accompanied by increases in the frequency of SW9-1T allele,with ＞90%of the bred varieties with a 100-SW ＞30 g carrying SW9-1T.Analysis of SW9-1 allelic variation in additional soybean accessions showed that SW9-1T allele accounting for 13.83%of the wild accessions, while in 46.55% and 51.57% of the landraces and bred accessions,respectively,this results indicating that the SW9-1 locus has been subjected to artificial selection during the early stages of soybean breeding,especially the utilization of SW9-1T in edamame for big seed.These results suggest that SW9-1 is a novel and reliable locus associated with seed size traits,and might have an important implication for increasing soybean seed weight in molecular design breeding. Cloning this locus in future may provide new insights into the genetic mechanisms underlying soybean seed size traits.

1.Introduction

The main objectives of soybean breeding are to increase seed yield and improve the quality of seed products to satisfy the requirements for human and animal consumption. Seed size is one of the major traits targeted by soybean breeders because it is vital for determining seed appearance, quality,and yield[1,2].Soybean seed weight is often expressed as 100-seed weight (100-SW), and is affected by seed size, which is measured based on seed length,width,and thickness as well as their corresponding ratios [1]. Seed size is an important soybean yield component[3-6].Thus,selecting soybean lines with increasing 100-SW is one of the main goals for soybean breeding programs [7,8]. However, soybean seed yield and 100-SW are influenced by multiple quantitative trait loci(QTL)or genes as well as by various environmental conditions[2,9,10]. Thus, it is difficult to improve soybean yield via traditional breeding methods[11].

The application of molecular markers represents a fast and accurate approach for identifying new sources of genetic variation and for investigating the genetic factors controlling quantitative traits [12,13]. The use of molecular markers to indirectly select important agronomic traits (e.g., seed size and weight) can improve the efficiency of traditional plant breeding programs [14]. Along with the development of soybean consensus genetic maps [15-23], molecular markers have facilitated the identification of QTL controlling important quantitative traits in many soybean populations [24].Additionally, the construction of genetic maps and analyses of QTL have enabled researchers to estimate the number of QTL regulating genetic variations in soybean [25]. This strategy has also been widely applied for detecting and identifying QTL for soybean seed size (weight) traits. Consequently, many QTL, single nucleotide polymorphism (SNP)markers, and genes associated with these traits have been identified over the past two decades [1,2,8,11,12,25-39].However, most of the identified QTL and SNP markers have rarely been applied in soybean breeding programs.

Genome-wide association study (GWAS) has become a powerful approach for elucidating complex agronomic traits and identifying causal variants with modest effects on target traits in crops[6,40-44].An earlier investigation indicated that SNP genotyping by genome-wide identification effectively promotes the applicability of association analyses for QTL mapping in crops[45].Specifically,a GWAS with SNP markers has been applied to analyze various important traits in many plant species, resulting in the identification of causal genes for a broad range of complex traits in diverse plant species,including Arabidopsis thaliana [41], rapeseed (Brassica napus)[43], rice (Oryza sativa L.) [46], maize (Zea mays) [47,48],sorghum [49], and soybean (Glycine max) [2,6,11,34,39,50-55].Therefore, identifying SNPs associated with seed size-related traits may enable soybean breeders to combine the causal genes during the breeding of new lines.

In our previous study, we observed that the GmCYP78A10 gene associated with seed size may have undergone artificial selection during soybean breeding [2]. In A. thaliana, the P450/CYP78A gene family, which includes AtCYP78A5, AtCYP78A6,and AtCYP78A9,has very important effects on seed size[56-58].Moreover,Du et al. [59]recently confirmed that overexpressed GmCYP78A5(i.e.,a close homolog of GmCYP78A10)in transgenic soybean plants significantly increased seed size and weight.Our previous results also revealed two highly conserved GmCYP78A10 alleles, GmCYP78A10a allele mainly distributed in wild soybean (G. soja), while GmCYP78A10b predominantly presented in cultivated soybean(G.max)[2].We speculated that soybean breeders initially selected big seeds (cultivated soybean) carrying GmCYP78A10b, however, this artificial selection did not significantly affect soybean seed yield. To further elucidate the evolution of loci or genes related to soybean seed size,we analyzed genome regions associated with soybean seed size traits including 100-SW, seed length (SL), and seed width(SW) using GWAS and SNP genotyping of 146 soybean accessions under multiple environments from three major ecoregions in China,and found a novel locus SW9-1 associated with soybean seed size. SW9-1 allelic variation in wild and cultivated soybean accessions was characterized. Artificial selection effect on this locus and its application in molecular breeding to develop new soybean lines with increased seed weight were discussed.

2. Materials and methods

2.1. Plant materials and phenotypic data collection

A total of 146 soybean accessions mainly from three major ecoregions in China were selected for association mapping including 49 accessions from Huanghuai (HH), 49 from Northeast South (NS), and 48 from Northeast North (NN)(Table S1).As soybean is much more sensitive to photoperiod than other field crops like wheat, rice and corn, some of soybean varieties from northern part of China could not reach maturity when planting in Huanghuai geographic region[31].To ensure the normal maturation, and obtain soybean seeds,thus, 49, 49, and 48 different soybean accessions were grown at 6, 6, and 2 experiment locations(environments)in HH,NS,and NN in China, respectively, with single-row of 3.00 m length and 0.50 m spacing between rows in a randomized complete-block design with three replicates (Table S1). One hundred and eighty-four bred accessions, 94 wild accessions, and 116 landrace from the Chinese mini-core soybean collection were used to validate the association of a derived cleaved amplified polymorphic sequence(dCAPS) marker for SW9-1 locus. All these accessions were grown in single-row with 3.00 m length and 0.50 m spacing between rows in a randomized complete-block design with three replicates at Dayangdian experiment station of Anhui Agricultural University (31°56′6.65″N, 117°11′49.39″E) and the experimental farm of Suzhou Academy of Agricultural Sciences (33°38′30.61″N, 117°04′36.32″E) in 2016 and 2017,respectively.

After the soybean seeds were harvested and dried naturally until reaching a stable seed weight,100-SW(g),SL(mm),and SW (mm) of all soybean accessions were centrally evaluated using High Precision Electronic Balance (APTP456 series)and Multifunction Seed Analyzer(SC-G type,Hangzhou Wanshen Detection Technology Co., Ltd., Hangzhou, China,http://www.wseen.com/).

2.2. DNA extraction and SNP genotyping

Genomic DNA was extracted from individual dry seeds as described by Kang et al. [60]. All soybean accessions were genotyped by the Illumina SoySNP6k iSelect BeadChip(Illumina, San Diego, CA, USA), which consists of 5361 SNPs[21]. Analyses of the chromosomal distributions, coding, and quality of the SNPs were described by Wen et al. [51]. The exclusion of SNPs with a missing data rate ＞0.25 and a minor allele frequency ＜0.05 resulted in 4987 (93.67%) SNP markers being retained for the subsequent analysis.

2.3. Population structure analysis

The population structure of the 146 soybean varieties was investigated using 4987 SNPs according to the Bayesian model-based program, STRUCTURE 2.3.4 [61,62]. The length of the burn-in period and the number of Markov Chain Monte Carlo replications after the burn-in were set at 100,000, with an admixture and allele frequencies-correlated model. Five independent runs were completed with the hypothetical number of subpopulations (k) ranging from 1 to 10. The correct k estimation was determined by combining the log probability of data [ln P(D)] from the STRUCTURE output and an ad hoc statistic,Δk,which was based on the rate of change in the log probability of data between successive k values[62].On the basis of the correct k value, each soybean variety was assigned to a subpopulation for which the membership value(Q-value)was ＞0.5[61,63],and the population structure matrix(Q)was generated for further analyses.

2.4. Phenotypic data analysis

Phenotypic data across all environments were analyzed with SAS 9.3 (SAS Institute, Inc., Cary, NC, USA). Additionally,location within in a year was treated as a single environment.The analysis of variance (ANOVA) for each trait was performed to evaluate the effect of genotype and environment on phenotypic variance with R ‘lm' function (R Core Team, 2012,https://www.r-project.org/). The model is yij= μ + τi+ βj+ εij,where i is for each line, j is for each environment. τiis for genotype effect and βjis for environmental effect, μ is for average value and εijis for error. The broad-sense heritability(h2)for each trait was calculated as:

where, δ2gis the genetic variance, δ2eis the residual variance and n is the number of environments.The estimates of δ2gand δ2ewere obtained by the mixed linear model using lme4 package in R (R Core Team, 2012, https://www.r-project.org/),treating genotype and environment as random effects, as follow: model = lmer (trait ～(1|Line) + (1|Env), data = data)[64].

The effects of the population structure on the phenotypic variation were calculated based on the mean values for each trait using PROC general linear model (GLM). The model statement included one of the two components of the Q matrix (k = 3) [65]. Correlation coefficients between soybean yield and yield components were estimated with PROC CORR.

2.5. Genome-wide association study

The SPAGeDi [66] and TASSEL 4.0 [67] programs were used to calculate the pair-wise relatedness coefficients (K, kinship matrix)for estimating the genetic relatedness among individuals, with the negative value of kinship set as 0. A GWAS based on the GLM and mixed linear model (MLM) was conducted for three soybean seed size traits with the TASSEL 4.0 program and 4987 SNPs [68]. For the MLM analysis, a marker-based kinship matrix (K) obtained with TASSEL 4.0 was used along with the Q matrix to correct for family and population structures. By Bonferroni correction, a P-value threshold of 1 × 10-3was used to declare the significance of marker trait associations. The SNP sites with relatively high and stable contribution rates were selected for further analyses.

2.6. Development of molecular markers

The Primer Premier 5.0 software (Premier Bio-soft International, Palo Alto, CA, USA) was used to develop dCAPS markers for three SNP loci related to non-synonymous mutations associated with soybean seed size traits identified by the GWAS.Enzyme digestion information for the SNP loci was confirmed using the dCAPS Finder 2.0(http://helix.wustl.edu/dapaps/dcaps.html)online program,and Eam1105I,Hinf I, and Xba I were selected as candidate enzymes. Details regarding the developed dCAPS markers were provided in Table S2. A PCR was completed in a 20 μL reaction volume consisting of 2.0 μL template DNA (100 ng μL-1), 2.0 μL 10×buffer, 2.0 μL dNTPs, 0.5 μL each primer (10 μmol L-1), 0.2 μL 1.0 U Taq DNA polymerase, and 12.8 μL ddH2O. The PCR was completed using the program: 95 °C for 5 min; 35 cycles of 95 °C for 30 s,55 °C for 30 s with a decrease of 0.2 °C per cycle,and 72 °C for 30 s; 72 °C for 5 min. A quantitative real-time PCR was completed using an ABI PCR System (Applied Biosystems, Foster City, CA, USA). The PCR products were digested with restriction endonucleases for 3 h following the manufacturer's recommended protocols (Takara, Dalian,China).

2.7. Statistical analysis

Data were analyzed using SAS, SPSS, and Microsoft Excel.Additionally, Student's t-test was completed using the independent-samples t-test (version 19.0; http://www.spss.com/). The phenotypic effect values for the allelic variations were calculated with the following equation:

where, acis the phenotypic effect value; xcdis the phenotypic data for the d-trait of the variety carrying the c-allelic variation type; ncis the number of varieties containing the callelic variation type; Neis the phenotypic data of the eindividual carrying the non-allelic variation type; neis the number of varieties containing the non-allelic variation type.

An acvalue ＞0 indicated that allelic variation type had additive effects on soybean seed size, while a value ＜0 indicated the allelic variation had negative effects.

3.Results

3.1. Phenotypic variation and population structure analysis

Each trait varied widely among the different ecoregions(Table 1).The 100-SWs in HH, NS, and NN ecoregions were 12.71-23.15,11.60-24.07, and 13.19-20.71 g, with average values of 18.27,20.16,and 17.29 g,respectively(Table 1).Compared to NS and NN ecoregions, large range of variation for 100-SW (3.06), SL (0.41),and SW (0.33) in the 6 environments of HH ecoregion was observed (Table 1). The h2value of the different soybean accessions across three major ecoregions was 70.13%-93.87%(Table 1). The highest h2value (93.87%) was observed in HH,indicating that the majority of the 100-SW variation was due to genetic effects. In contrast, the 100-SW in the other two ecoregions might be affected primarily by environmental factors.

To avoid false-positive associations because of population stratification, STRUCTURE and a neighbor-joining tree-based method were used to estimate the relatedness among 146 soybean accessions using 4987 SNPs(Fig.1).The distribution of the ln P(D)value for each given k value did not show a clear peak trend(Fig.1-A).The ad hoc quantity(Δk)exhibited a much higher likelihood at k = 3 than at k = 2 and 4-10 (Fig. 1-B), suggesting that the 146 soybean accessions could be divided into three major subpopulations (Fig. 1-C). The neighbor-joining tree results (Fig. 1-D) were consistent with the STRUCTURE results(Fig.1-C).Of the 86 accessions in subpopulation P3,49 were from HH,and exhibited relatively late maturity.The subpopulation P1(26 accessions)and P2(34 accessions)were mainly from NS and NN, and exhibited relatively early maturity. An earlier study suggested that diversity in the photoperiod response among groups that differed regarding maturity might be the primary factor mediating the differentiation of cultivated soybean [69].The co-existence or overlapping of genetic and species differentiation were likely due to natural and artificial selection based on fitness [70]. The Q matrices obtained from the STRUCTURE program for the three subpopulations were used to control the variation of the population structure in the following genetic analysis.

Table 1-Analysis of soybean 100-seed weight as well as seed length and width across three ecoregions.

3.2. Identification of SNP loci related to soybean seed size

A total of 21 SNPs including 3 SNPs associated with 100-SW,16 SNPs associated with SL,and 2 SNPs associated with SW were identified by GWAS using 4987 SNP markers(Table S3).Eighteen of the 21 SNPs,each explained ＞10%of the phenotypic variance,while the other three SNPs each explained ＞8%of the phenotypic variance.On average,11.34%of the phenotypic variance could be explained by each SNP (Table S3). All of these SNPs were significantly associated with seed size traits in all 14 environments in three soybean ecoregions of China. To better identify significant SNP loci associated with soybean seed size traits,two environmental locations with higher relevance were selected for each of three ecoregions. An association analysis of eight combinations representing six locations was then conducted.Accordingly, 3, 17, and 1 SNPs were identified significantly associated(P ＜0.01)with 100-SW,SL,and SW,respectively,in at least four combinations for three ecoregions(Table S4).One SNP(ss248666643)on chromosome 15 was related to 100-SW and SL in at least five combinations (Table S4). Moreover, two SNPs(ss246792949 and ss244709037) were significantly associated with 100-SW, SL, and SW in at least four combinations, and explained an average of 10.01% of the phenotypic variance of three traits(Table S4).

3.3. Identification of loci with significant effects on seed size and development of molecular marker

To compare the allelic variation of three SNP loci(ss246792949T/C, ss244709037A/G, and ss250164298G/A) significantly associated with seed size (100-SW), the dCAPS markers corresponding to three loci were developed. Sequence analysis results revealed that the PCR products of the three loci were all target fragments, with mismatched bases at the corresponding SNP positions, which were consistent with the expected results. A total of 50 bred soybean varieties with extreme 100-SW (25 100-SW ≤13.60 g;25 100-SW ≥28.90 g)were further used to verify the three loci with the corresponding dCAPS markers. The sizes of the digested DNA fragments were consistent with the predicted sizes(Table S2,Fig.2).These results confirmed the suitability of the dCAPS markers developed for three loci.

3.4. Identification and validation of the SW9-1(ss246792949T/C) locus

Fig. 1 - Analysis of the population structure of 146 soybean accessions. (A) Distribution of the ln P(D) values corresponding to each hypothetical k-value. (B) Calculated k-value over five runs with the putative k ranging from 1 to 10. The highest Δk value occurred when k = 3, indicating that all the tested soybean accessions could be divided into three major subpopulations. (C) Model-based population structure for all 146 soybean accessions. (D) Neighbor-joining tree based on Nei's genetic distance matrix

Based on the average value of 100-SW, we divided soybean accessions into big seeds(BSs)and small seeds(SSs)group.The genotyping of the dCAPS marker revealed that the frequencies of the ss244709037A and ss244709037G alleles were 40% and 60%, respectively. No obvious differences in the selection efficiencies of ss244709037A and ss244709037G were observed(Table 2). However, the frequency of the ss250164298A allele was 28%, and the selection efficiency in SSs was about twice higher than that in BSs. The frequency of the ss250164298G allele was 72%, which was mainly expressed in BSs (55.56%).Additionally,the allele variation at the ss250164298(A/G)locus did not significantly affect soybean seed weight (Table 2). The frequency of the ss246792949T allele(named SW9-1T)was 70%in BSs, and the selection efficiency in BSs was approximately twice higher than that in SSs.In contrast,the frequency of the ss246792949C allele (named SW9-1C) was 30%, which was mainly expressed in SSs (86.67%) (Table 2). The 100-SW of soybean varieties with SW9-1T allele was significantly heavier than those with SW9-1C allele(P ＜0.01)(Table 2).These results indicated that the allelic variation at the SW9-1 locus significantly affected soybean seed weight. Furthermore, the SW9-1 locus had relatively high phenotypic effect (13.67) on 100-SW,with greater contribution by the accessions with BSs(3.69)than those with SSs(1.66).The SW9-1T/C locus had more significant effect on 100-SW than the ss244709037 (A/G) and ss250164298(A/G)loci(Table 2).Meanwhile,the quantile-plots indicated that the allelic variation at the SW9-1 locus had significant effect(Δm = 2.36 g;P = 5.748E-05)on 100-SW(Fig.S1).

Fig.2- Polymorphism analysis involving dCAPS markers and soybean cultivars. (A)Polymorphisms of dCAPS marker ss246792949 in 52 soybean accessions.(B) Polymorphisms of dCAPS marker ss244709037 in 52 soybean accessions. (C)Polymorphisms of dCAPS marker ss250164298 in 52 soybean accessions.M:marker;1:ddH2O;2-3:soybean accessions with a known genotype(A:2 represents the CC genotype and 3 represents the TT genotype;B: 2 represents the AA genotype and 3 represents the GG genotype;C:2 represents the GG genotype and 3 represents the AA genotype);4-53:50 soybean accessions with an unknown genotype,of which 4-28:100-seed weight of 7.40-13.60 g and 29-53: 100-seed weight of 28.90-44.25 g.

Table 2-Allelic variation at three loci related to soybean seed weight corresponding to the type of seed distribution,frequency,selection efficiency,and phenotypic analysis of the allelic variation effects.

Table 3-Verification of the effectiveness of SW9-1T/C based on dCAPS markers as well as bred varieties and wild accessions.

3.5. Distribution of SW9-1 alleles in soybean accessions

To confirm the effect of SW9-1 on soybean seed size(100-SW),we analyzed total of 184 accessions including additional 134 newly bred soybean accessions.The frequency of the SW9-1T allele was 60%, with higher selection efficiency for the accessions with BSs (51.82%) than for the accessions with SSs. The frequency of the SW9-1C allele was 40%, which was mainly detected in SSs (62.16%) (Table 3). Additionally, the SW9-1T/C alleles had a significant effect on 100-SW, with a phenotypic effect value of 3.73(Table 3).The results suggested that SW9-1C allele was primarily in the accessions with SSs,while SW9-1T was mainly in the accessions with BSs(Tables 2 and 3). To confirm the association of these alleles with seed size, 94 wild-type accessions with averaged 100-SW of 2.26 g were used for an additional genotyping with these dCAPS markers. We observed that 81 of 94 wild-type accessions(86.17%) mainly contained the SW9-1C allele, while the remaining 13 accessions mainly comprised the SW9-1T allele(Table 3).

3.6. Genetic and selective effects of SW9-1T/C alleles

We observed that SW9-1T occurred more frequently than SW9-1C in elite cultivars or accessions with BSs, in contrast to the allelic distribution in wild soybean accessions with SSs(Tables 2 and 3).This suggested that SW9-1T might have been artificially selected during the early stages of soybean breeding.To verify this hypothesis, the frequencies of the two alleles in 94 wild accessions,116 landraces,and 318 bred accessions were further determined. The SW9-1T allele was detected in 13.83% of the wild accessions, while in 46.55% landraces and 51.57% bred accessions (Table 4, Fig. 3). The frequency difference between wild and cultivated (bred varieties and especially landraces)soybean populations implied that the SW9-1 locus was under artificial selection during the early stages of soybean domestication. An analysis of the distribution of SW9-1 alleles in bred soybean seeds with different 100-SWs revealed that SW9-1T was in the accessions with heavier seeds,while SW9-1C was in the accessions with lighter seeds(Fig.4).Bred soybean varieties were further divided into four groups according to the 100-SW range. The SW9-1C allele was detected in 77.78% of varieties with a 100-SW ＜10 g, while only 22.22% carried SW9-1T.Increases in 100-SW were accompanied by increases in the frequency of SW9-1T. ＞90.00% bred varieties with a 100-SW＞30 g carrying SW9-1T. Moreover, there was a significantly difference(P ＜0.01)between the 100-SWs associated with SW9-1T and SW9-1C(Table 4,Fig.4).

3.7. Novelty of SW9-1 locus for soybean weight

On the basis of our results,the SW9-1 locus was localized to chromosome 9 in soybean genome (Table S5). Comparison of SW9-1 position with QTL positions reported in bi-parental mapping studies revealed that 13 QTL/loci on chromosome 9 were associated with soybean seed weight (100-SW) (Table S6).However,the physical location of the SW9-1 locus was notwithin and did not overlap with the physical location interval for known QTL.Thus,SW9-1T/C may be a new allele determining soybean seed size(weight).

Table 4-Distribution and selective effect of SW9-1T/C alleles in soybean accessions.

Fig. 3-Distribution of the SW9-1T/C alleles in different soybean accessions.

4.Discussion

4.1. Analyses of population structure and genetic diversity

Previous studies have indicated that SNPs can be used to obtain accurate sequence information via high-throughput sequencing technology to compensate for bi-allelic shortcomings [70-72]. SNP markers have also widely been used in soybean genotyping [6,11,52-55,73]. These studies suggested that achieving a similar accuracy to that of simple sequence repeats (SSRs) in analyses of population structures and genetic diversity required about 10 times more SNPs than SSRs[74-75].In this study,146 soybean accessions from three different ecoregions in China were classified into three subpopulations, which the highest genetic variability (CV)and h2associated with 100-SW (16.82% and 93.87%, respectively),SL(5.30%and 94.87%,respectively),and SW(5.11%and 95.53%, respectively) were observed in the HH ecoregion (P3).Meanwhile, the lowest values were observed for the NN (P2)ecoregion (Table 1). These results indicated that there were significant genetic differences among the soybean varieties grown in the three ecoregions(subpopulations).

Fig.4- Distribution of the SW9-1T/C alleles in different 100-SW ranges for bred soybean varieties.

4.2. SW9-1 is a reliable locus for soybean 100-SW

Because of functional markers (CAPS/dCAPS markers) can quickly and effectively identify SNP alleles, they have been widely applied to detect mutations causing phenotypic changes in many crops,including wheat(Triticum aestivum L.)[76,77],Brassica oleracea[78],maize[79],rice[80,81],and barley(Hordeum vulgare L.) [82]. Here, the reliability of SW9-1 locus was evaluated by our newly developed dCAPS marker. The results showed that SW9-1 locus has the higher phenotypic effect value(13.67)for 100-SW than the ss244709037(A/G)and ss250164298(A/G) loci, more effect on accessions with BSs(3.69) than on accessions with SSs (1.66) (Table 2). Therefore,we speculated that the SW9-1T/C alleles have more significant effect on soybean seed weight. Further analysis results confirmed this speculation (Table 3, Fig. S1). Important soybean quantitative traits, including seed size, are largely affected by multiple genes and environmental conditions[2,9,10].Consequently,related QTL or loci that can be detected in different environments or genetic backgrounds may be useful because of their apparent stability[83,84].To verify the stability of SW9-1 locus in multiple environments, 184 bred accessions involved in genotyping were divided into different ecoregion according to their sources. The effect of allelic variation at SW9-1 locus on 100-SW showed different distribution in different ecoregion,100-SW of accessions with SW9-1T type was greater than that of varieties with SW9-1C type in different ecoregion, especially in Huanghuai ecoregion, there were significant differences in 100-SW between accessions with SW9-1T and SW9-1C loci. Thus, we concluded that the SW9-1 locus was reliable and persistent across different environments, and significantly influenced soybean 100-SW.The high reliability of this locus across multiple environments might have considerable implication for marker-based breeding of soybean cultivars with enhanced seed size(100-SW).

4.3. Domestication and artificial selection for SW9-1 locus

Seed size is an important determinant of the evolutionary fitness of plant species [85]. It is also a key agronomic trait during crop domestication because of its important role in determining the seed yield of crops [85-87]. Previous investigations uncovered many QTL or genes correlated with domestication-related traits, including seed size, seed weight,plant height, flower color, stem determinacy, seed coat color,hilum color, and pubescence form [1,2,32-36,39,88-92]. In this study,we observed that the SW9-1C allele was mainly expressed in soybean populations with small seeds,in contrast,SW9-1T was the predominant allele in soybean varieties with big seeds.Further analysis of additional accessions indicated that SW9-1T was gradually selected for use (13.83% wild accessions, 46.55%landraces and 51.57% bred accessions), while SW9-1C was gradually eliminated. Increases in seed weight were associated with increases in the frequency of SW9-1T,＞90.00%bred soybean accessions (mostly were vegetable soybeans (edamame))with a 100-SW ＞30 g carried this allele. The results indicated that the allelic variation of SW9-1 loci had been subjected to domestication or artificial selection during the early stage of soybean breeding programme, especially SW9-1T in edamame with heavier seeds.Considering that the SW9-1T/C alleles significantly affect soybean seed size(100-SW),the positive effects of SW9-1T on seed size should be applied in soybean modern breeding programs.

4.4.Potential implication of SW9-1 locus for soybean breeding

Currently, ＞400 QTL or markers for soybean seed size traits have been detected on 20 chromosomes, including ＞280 QTL or markers associated with seed weight (100-SW) (SoyBase, http://www.soybase.org/; Table S7). A QTL region controlling plant height,lodging,and seed yield spanning approximately 1.2 Mbp on chromosome 9 was detected in a GWAS [88]. Li et al. [34]subsequently used a pan-genome analysis to identify four genes[Glyma09g06950(Gm9:5,783,764-5,788,819),Glyma09g07090(Gm9:5,932,308-5,935,017), Glyma09g07206 (Gm9: 6,072,782-6,078,905),and Glyma09g07760(Gm9:6,694,656-6,699,276)]in this region with large-effect SNPs and/or indels related to soybean growth and seed yield.Moreover,Zhou et al.[39]also detected a genetic loci associated with seed weight, which was consistent with the previously reported qSW locus on chromosome 17[88].There are four genes and 13 QTL related to soybean seed weight(100-SW)reported on chromosome 9, however, the physical distance gap between SW9-1 and its nearest locus (Sdwt 15-6) is ＞600 kb [24].The physical location of the SW9-1 locus was not overlapped with a known QTL or a gene interval listed in SoyBase (http://www.soybase.org/). Therefore, the SW9-1T/C locus represents a novel locus.

The fact that soybean 100-SW is regulated by multiple genes and is influenced by various environmental conditions makes it difficult to improve soybean yield via traditional breeding methods, however, the development of modern molecular biology technologies (genomics, gene editing technology, molecular design breeding (MDB), etc.) has provided alternative tools to enhance breeding efficiency in soybean breeding programs [2,9-11]. Compared with other breeding techniques,MDB can precisely modify agronomic traits at the genetic level, and avoid the problem associated with traditional breeding methods regarding the transfer of undesirable genes.Thus,MDB can greatly increase breeding efficiency and shorten the breeding cycle [93]. Previous studies confirmed that highly consistent marker loci across environments can be used for MDB to enhance soybean seed weight[12].Therefore,applying specific genes or allele variations for soybean breeding via MDB may be an effective way to genetically improve soybean seed size and yield.This method has been successfully applied to enhance chilling tolerance, nitrogen use efficiency, grain shape, yield and grain quality, and grain size and yield in rice[94-97].In this study,SW9-1(SW9-1T)locus was linked with the reported large-effect QTL to form a“molecular module”according to the MDB theme,which may be relevant for the breeding of new soybean varieties that produce large seeds, as well as for elucidating the theoretical basis for the genetic improvement and prediction of soybean seed weight in different ecoregions.

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2018.12.010.

Conflict of interest

Authors declare that there are no conflicts of interest.

Acknowledgments

This work was supported by the National Key Research and Development Program of China (2016YFD0100201), the National Natural Science Foundation of China (31771819), the China Postdoctoral Science Foundation (2017M621990), the Introduced Leading Talent Research Team for Universities in Anhui Province,and the Natural Science Foundation of Anhui Province,China(1608085QC66).