2014 Volume 63 Issue 5 Pages 441-449
Wild soybean, the progenitor of cultivated soybean, is an important gene pool for ongoing soybean breeding efforts. To identify yield-enhancing quantitative trait locus (QTL) or gene from wild soybean, 113 wild soybeans accessions were phenotyped for five yield-related traits and genotyped with 85 simple sequence repeat (SSR) markers to conduct association mapping. A total of 892 alleles were detected for the 85 SSR markers, with an average 10.49 alleles; the corresponding PIC values ranged from 0.07 to 0.92, with an average 0.73. The genetic diversity of each SSR marker ranged from 0.07 to 0.93, with an average 0.75. A total of 18 SSR markers were identified for the five traits. Two SSR markers, sct_010 and satt316, which are associated with the yield per plant were stably expressed over two years at two experimental locations. Our results suggested that association mapping can be an effective approach for identifying QTL from wild soybean.
The wild relatives of crops have been undeniably beneficial to modern agriculture, providing plant breeders with a broad pool of potentially useful genetic resources (Hajjar and Hodgkin 2007). However, these wild relatives are consistently ignored for yield improvement because, in general, they have smaller seed sizes, greater tendency to shatter and other undesirable traits. Nevertheless, there are an increasing number of cases of high-yielding derivatives of hybrids that have been created through the use of wild relatives, including tomato, wheat, rice, oat, barley, sorghum, maize and soybean (Frey et al. 1984, Fu et al. 2010, Kan et al. 2012, Li et al. 2008, Reeves and Bockholdt 1964, Rick 1974, Tanksley and McCouch 1997, Xiao et al. 1996), which indicate that a crop’s wild relatives can be used as a gene resource to improve the yield of cultivated crops through traditional breeding or molecular marker-assisted selection. In China, two introgressions from a wild relative of rice have been associated with a 30% increase in the yields of the world’s highest yielding hybrid rice (Deng et al. 2004). For tomato, yield increases of greater than 50% have resulted from pyramiding three independent, yield-promoting genomic segments from a wild relative (Gur and Zamir 2004). Nevertheless, we have learned little about the chromosomal regions that contribute to yield increases, or the genetic bases of these traits (Swamy et al. 2008). Through the implementation of linkage-based QTL mapping and linkage disequilibrium (LD)-based association mapping in crop genetics, it is possible to locate the genomic regions that contribute to yield-related traits, clone the gene/QTL from a wild relative and used this information to improve cultivated crops.
Cultivated soybean or soybean landraces are consistently selected as germplasm to improve soybean yield and linkage mapping is the main method for investigating the genetic basis of yield. However, the genetic diversity of cultivated soybean was lost through artificial selection. Wild soybean, which possesses high genetic diversity compared to cultivated soybean, can be as a germplasm to improve cultivated soybean and provide plant breeders with a broad pool of potentially useful genetic resources. Tanksley and McCouch (1997) noted to the potential role of genome mapping to efficiently utilize the genetic diversity of wild relatives and suggested that the continued sampling of wild germplasm would result in new gene discoveries and utilization.
The limited research on wild soybean mainly focuses on biotic or abiotic stress. However, only a few studies on yield in wild soybean have suggested that wild soybean can be used as the germplasm to improve soybean yield traits and some favorable alleles have been identified in wild soybeans. Concibido et al. (2003) mapped a QTL from wild soybean PI407305 using BC2, which was derived from a cross between cultivated soybean (HS-1) and wild soybean (PI407305). Wang et al. (2004) mapped eight QTL for yield using BC2F4, which was derived from a cross between cultivated soybean (IA2008) and wild soybean (PI468916) and four favorable alleles were found in wild soybean. Li et al. (2008) mapped a QTL closely linked to a SSR marker (satt511) from wild soybean in three environments using BC2F4, which was derived from a cross between cultivated soybean (7499) and wild soybean (PI245331). It was also found that the additive effect of wild soybean can increase the yield from 191 kg ha−1 to 235 kg ha−1. Wen et al. (2008) conducted association mapping for agronomic and quality traits in wild and cultivated soybean, respectively and they found some association only detected in wild soybean population. Kan et al. (2012) mapped two QTL for pod number per plant and one QTL for yield per plant from wild soybean over two years. All of these results suggest that wild soybean contains yield-favorable alleles and that it is feasible to identify a favorable allele for yield in wild soybean. Wild soybean provides a large variation of naturally occurring alleles for QTL mapping and using in soybean improvement (Iyer-Pascuzzi et al. 2007) and many useful new alleles for yield-related traits can be mined from wild soybean. As additional QTL for yield are identified from different wild accessions, it will become clear whether all of the accessions or only a few wild accessions that are distant from cultivars have yield-enhancing QTL by linkage or association mapping.
In this study, the variation of these yield-related traits (i.e., days from sowing to flowering, days from sowing to mature, 100-seed weight, pod number per plant and yield per plant) in wild soybean from China was analyzed. And association mapping was conducted for five yield-related traits to detect yield-favorable QTL in wild soybean. Based on the MLM model (Q+K), a total of 45 marker-trait associations were identified for the five yield-related traits, involving 18 SSR markers.
A total of 113 wild soybean accessions, representing the full geographic range of wild soybean from southern China to northeast China, were selected to construct the association mapping population (Supplemental Table 1). The experiments were conducted at the Jiangpu Agronomic Experimental Station of Nanjing Agricultural University (32°12′N 118°37′48″E), Nanjing, China, in the summers of 2011 and 2012 and at the Nanyang Experimental Station at Henan Agricultural University (38°7′N 110°34′E), Nanyang, China, in the summer of 2012. The accessions were planted in a complete randomized block design, with 100 cm × 100 cm hill plots, 4 plants per plot and 2 replications. Five yield-related traits were evaluated: the days from planting to flowering (DTF), days from planting to maturity (DTM) (without the data from the Nanyang Experimental Station), pod number per plant (PN), 100 seed weight (HSW) and yield per plant (YLD).
SSR genotypingGenomic DNA from all of the materials was extracted from the young leaves of each accession as described by Doyle and Doyle (1990), with slight modifications. A total of 85 SSR markers representing 19 soybean chromosomes were selected from published genetic maps (Hwang et al. 2009, Song et al. 2004) to genotype the 113 wild soybean accessions and the genetic position of the SSR were referenced the genetic maps that was constructed by Song et al. (2004) and the genetic maps constructed by Hwang et al. (2009). The PCR amplification was performed in a 10-μl volume containing 20 ng total DNA, 0.4 μM forward and reverse primers, 200 μM of each dNTP, 19-μl PCR buffer (10 mM Tris-HCl, pH 8.3 and 50 mM KCl), 2 mM MgCl2 and 0.5 U Taq DNA polymerase. The PCR was programmed with an initial denaturing at 94°C for 5 min, followed by 35 cycles of 95°C for 30 s, 54°C for 1 min and 72°C for 1 min, with a final extension at 72°C for 10 min. The PCR reactions were performed using an MJ Research PTC 225 DNA engine thermal cycler (Bio-RAD, USA). The PCR products were separated by 8% non-denaturing polyacrylamide gel electrophoresis with a 29 : 1 ratio of acrylamide : bisacrylamide and then silver-stained, as described by Santos et al. (1993). The stained bands were analyzed based on their migration distance relative to the pBR322 DNA Marker (Fermentas) using Quantity One v.4.4.0 software 4.4 (Bio-Rad, Hercules, CA, USA).
Statistical analysis PhenotypeThe data analysis was performed using the R statistic language (R Development Core Team 2010). Analysis of variance (ANOVA) of all phenotypic data based on the means of traits of each accession three environments was conducted as model: Phen = genotypes + years + locates + e. where phen was the phenotypic observation, genotypes was the genetic effect, years was the effect of the different years, locates was the effect of the different experiment place, and e was the residual. The best linear unbiased predictor (BLUP) values for each line were calculated using the lme4 package (Bates et al. 2011). Heritability was calculated as the genotypic variance divided by the total variance. The spearman rank correlation coefficient between each pair of traits was calculated based on the BLUP using the “cor” function. The effect of population structure on the phenotype was assessed based on the BLUP value for each accession using the GLM procedure in SAS 8.02 (SAS Institute 1999).
Genotypic data analysisThe number of alleles, gene diversity and polymorphic information content (PIC) were calculated using Powermarker version 3.25 (Liu and Muse 2005). Additionally, Nei’s genetic distance (1973) among the individuals was calculated using Powermaker version 3.25 (Liu and Muse 2005) and was then used to construct a neighbor-joining (NJ) phylogenetic tree with 1000 boot-strapping runs using Powermaker version 3.25 (Liu and Muse 2005). The tree was visualized using MEGA version 4.0 (Tamura et al. 2007).
Population structureThe Bayesian model-based program STRUCTURE 2.2 (Pritchard et al. 2000) was used to infer the population structure using 74 SSR markers, which were selected to represent 20 chromosomes. The burn-in period was 100 000 and the number of iterations was 100 000 using a model that allowed for admixture and correlated allele frequencies. The number of subpopulations (K) was set from 1 to 10, with 7 independent runs for each K. The most likely number of subpopulations was then determined using the Delta K method proprsed by Evanno et al. (2005). A kinship matrix was calculated using SPAGeDi software (Hardy and Vekemans 2002). All of the negative kinship values between the individuals were set to zero, according to Yu et al. (2006).
Linkage disequilibrium calculationThe level of LD between pairs of SSRs was calculated using the software TASSEL V2.1 (Bradbury et al. 2007). LD was measured for each pair of loci using D′ and the significance (P-value) for each SSR pair was determined with 1000 permutations.
Association mappingThe association between the phenotypes and markers was evaluated with general linear model (Q) and mixed linear model (Q+K) that was implemented in Tassel V2.1 software (Bradbury et al. 2007, Yu et al. 2006). In this model, we tested the marker association between the phenotype and SSR markers, with Q as a fixed covariate and kinship (K) as a random effect. The markers were identified as significantly associated with traits using a threshold of −Log(P-value) ≥ 2.00.
The phenotypic allele effect of SSR that associated with five traits was estimated through comparison between the average phenotypic value over accessions with the specific allele and the of all accessions based on the BLUP value as: ai = ∑xij/ni − ∑X/n, where ai representing the phenotypic effect of ith allele; xij representing the phenotypic value of the jth material with the ith allele; ni representing the number of materials with the ith allele; ∑X/n representing the mean of the phenotypic value of all materials. If ai > 0, it is supposed to be a positive allele, if ai < 0, it corresponds to be a negative allele.
If not otherwise noted, all of the analyses were performed using the statistical software R (R, Development core Team, Vienna, Austria 2010).
By genotyping the 113 wild soybean accessions with 85 SSR markers, we detected a total of 892 alleles, ranging from 2 to 23 alleles per SSR marker, with an average of 10.49 alleles per locus. The corresponding PIC values ranged from 0.07 to 0.92, with an average of 0.73. The genetic diversity at each SSR marker ranged from 0.07 to 0.93 with an average of 0.75 (Supplemental Table 2).
The genetic relationships among the accessions were investigated using a model-based Bayesian clustering method with 74 SSR markers. Four subpopulations were detected by STRUCTURE, which is based on a Bayesian approach (Fig. 1A). The first, second, third and fourth subpopulations contained 17, 13, 41 and 22 accessions (Fig. 1B), respectively. The information of Unrooted neighbor-joining tree of 113 wild soybean accessions, as based on Nei’s 1973 genetic distance was consistent with the results from STRUCTURE (Fig. 2).
A. calculation of the true K of 113 wild soybean accessions, according to Evanno et al. (2005) and B. the population structure of 113 wild soybean accessions, as based on the 74 SSR loci. Each individual is represented by a single vertical line divided into four colored segments, with lengths proportional to each of the four clusters, and the color proportional in a single vertical line indicated that the proportional of a line belongs to a subpopulation. The number under vertical line represents the accession corresponding to Supplemental Table 1.
Unrooted neighbor-joining tree of 113 wild soybean accessions, as based on Nei’s 1973 genetic distance. The color of the lines show the subpopulation they belong to based on the Fig. 1B. The number of each line represents the accessions corresponding to Supplemental Table 1.
The relative kinship estimates based on the 74 SSR data indicated that 80.34% of the pairwise kinship estimates were within the range of 0 to 0.05, the remaining estimates ranged from 0.05 to 0.71, with a continuously decreasing number of pairs filling in the higher estimate categories (Fig. 3). The kinship analysis revealed that the majority of the accessions had a null or weak relationship with the other accessions in this population.
Distribution of pair-wise kinship coefficients for 113 wild soybean accessions. The values are from SPAGeDi estimates using 74 SSR markers.
ANOVA revealed that there were significant differences among the accessions (P < 0.01) for five yield-related traits, indicating a large amount of genetic variation in the population (Supplemental Fig. 1). The effect of the years and location on the five traits was significant (Table 1). The heritability of the five traits ranged from 39.57% for PN to 97.84% for DTF. The population structure had a strong influence on DTF (32.28%) and DTM (28.58%), with P-values < 0.0001. A significant effect was detected for PN (13.25%), with a P-value < 0.0038. No significant effects were detected for HSW and YLD with P-value 0.53 and 0.29, respectively (Table 1).
| Traits | E | Mean + SD | Max | Min | Year | Loc | Gen | h2 | R2q |
|---|---|---|---|---|---|---|---|---|---|
| DTF | 2011NJ | 55.61 ± 12.92 | 84.00 | 27.00 | ** | ** | ** | 97.84 | 32.28 |
| 2012NJ | 57.14 ± 12.58 | 84.50 | 26.00 | ||||||
| 2012NY | 60.55 ± 12.98 | 92.00 | 35.50 | ||||||
| DTM | 2011NJ | 105.70 ± 12.76 | 138.33 | 81.28 | ** | Na | ** | 96.30 | 28.58 |
| 2012NJ | 102.61 ± 11.39 | 131.00 | 78.50 | ||||||
| PN | 2011NJ | 268.43 ± 111.42 | 268.43 | 93.40 | ** | ** | ** | 39.57 | 13.25 |
| 2012NJ | 220.19 ± 80.65 | 419.58 | 50.33 | ||||||
| 2012NY | 736.21 ± 362.08 | 1621.5 | 124.13 | ||||||
| HSW | 2011NJ | 2.58 ± 1.55 | 10.41 | 0.90 | ** | ** | ** | 90.95 | 2.85 |
| 2012NJ | 2.39 ± 1.44 | 9.74 | 0.90 | ||||||
| 2012NY | 2.19 ± 1.50 | 8.48 | 0.66 | ||||||
| YLD | 2011NJ | 15.18 ± 7.34 | 44.65 | 3.62 | ** | ** | ** | 90.31 | 4.36 |
| 2012NJ | 16.37 ± 9.53 | 54.35 | 3.19 | ||||||
| 2012NY | 17.07 ± 8.28 | 43.38 | 1.66 |
E, Environments; SD, Standard deviation; Na: No information; h2: Heritabity; R2q: Phenotypic variance explained by the population structure;
Pearson correlation coefficients between traits based on the BLUP value were calculated and there was a significant negative correlation between HSW and PN (r = −0.44, P < 0.01). Conversely, there were significant positive correlations between HSW and YLD (r = 0.53, P < 0.01), DTF and PN (r = 0.51, P < 0.01), DTM and YLD (r = 0.26, P < 0.01), DTM and PN (r = 0.34, P < 0.01) and DTF and DTM (r = 0.91, P < 0.01) (Table 2).
The LD pattern was assessed based on the 2279 pairwise combinations of the 85 SSR loci. Based on the D′ estimates, 15.05% had a significant LD at P ≤ 0.05 and D′ ranged from 0.0038 to 1 with an average of 0.38.
Based on the MLM model (Q+K), a total of 45 marker-trait associations were identified for the five yield-related traits, involving 18 SSR markers (Table 3 and Fig. 4). Among the 45 significant associations, seven were correlated with DTF, ten with DTM, three with PN, twelve with HSW, thirteen with YLD.
| Trait | Loci | Chr. | Position (cM) | 2011NJ | 2012NJ | 2012NY | BLUP | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| GLM | MLM | GLM | MLM | GLM | MLM | GLM | MLM | ||||
| DTF | satt408 | 1 | 106.69 | 2.07 | 2.01 | ||||||
| satt521 | 3 | 65.46 | 2.92 | 2.34 | 2.55 | 2.51 | |||||
| satt322 | 6 | 82.23 | 2.23 | 2.05 | 2.22 | 2.47 | 2.38 | 2.37 | 2.22 | ||
| satt636 | 7 | 5.00 | 2.06 | ||||||||
| satt150 | 7 | 18.58 | 2.00 | 2.10 | |||||||
| satt417 | 9 | 46.20 | 2.04 | 2.00 | 2.02 | 2.10 | |||||
| satt405 | 16 | 12.41 | 2.08 | ||||||||
| satt285 | 16 | 25.51 | 2.04 | 2.32 | 2.77 | 2.43 | |||||
| sctt010 | 18 | 45.87 | 3.00 | 3.14 | 2.82 | ||||||
| satt564 | 18 | 57.32 | 3.04 | 4.37 | 3.22 | 4.06 | 2.43 | 3.77 | 2.41 | ||
| satt614 | 20 | 31.94 | 2.08 | ||||||||
| DTM | satt408 | 1 | 106.69 | 2.00 | |||||||
| satt641 | 3 | 29.28 | 2.10 | 2.06 | |||||||
| satt521 | 3 | 65.46 | 5.61 | 4.32 | 4.44 | 2.80 | 5.32 | 3.77 | |||
| sat_304 | 3 | 77.10 | 2.22 | 2.15 | 3.72 | 2.59 | 2.96 | 2.80 | |||
| satt718 | 4 | 73.79 | 2.23 | 2.12 | |||||||
| sctt010 | 18 | 45.87 | 3.60 | 2.72 | 3.12 | ||||||
| satt564 | 18 | 57.32 | 2.32 | 2.39 | 2.11 | 2.44 | 2.00 | ||||
| satt614 | 20 | 31.94 | 2.34 | 2.31 | 2.05 | ||||||
| HSW | BE475343 | 2 | 30.74 | 2.26 | |||||||
| satt641 | 3 | 29.28 | 5.60 | 3.84 | 7.08 | 4.58 | 2.21 | 4.25 | 2.47 | ||
| satt521 | 3 | 65.46 | 2.21 | 2.31 | |||||||
| sat_137 | 5 | 0.00 | 2.74 | ||||||||
| satt227 | 6 | 26.65 | 3.31 | 2.31 | 2.14 | 3.64 | |||||
| satt316 | 6 | 127.67 | 2.68 | 2.08 | 2.19 | ||||||
| satt150 | 7 | 18.58 | 4.58 | 3.30 | |||||||
| satt210 | 7 | 112.08 | 2.62 | 2.60 | 2.09 | 2.72 | |||||
| satt445 | 10 | 20.43 | 2.70 | 2.43 | 2.49 | ||||||
| satt123 | 10 | 86.86 | 2.07 | ||||||||
| satt251 | 11 | 36.84 | 2.18 | 2.57 | |||||||
| sat_334 | 12 | 59.01 | 2.59 | 2.82 | 2.16 | 2.31 | 2.82 | 2.12 | |||
| satt516 | 13 | 44.42 | 3.33 | 2.43 | 3.91 | 2.77 | |||||
| sct_188 | 13 | 85.33 | 2.37 | 2.24 | 2.17 | ||||||
| satt706 | 15 | 43.36 | 2.06 | ||||||||
| satt285 | 16 | 25.51 | 2.19 | 3.29 | 2.39 | ||||||
| satt564 | 18 | 57.32 | 2.70 | 3.93 | 2.96 | ||||||
| sct_010 | 19 | 59.52 | 2.24 | 2.68 | 2.12 | ||||||
| GMES2079 | 19 | 90.10 | 2.12 | 2.21 | |||||||
| GMES4376 | 19 | 91.10 | 2.41 | 2.70 | 2.60 | 2.82 | |||||
| PN | satt342 | 1 | 48.14 | 2.34 | 2.38 | ||||||
| satt641 | 3 | 29.28 | 2.30 | 2.19 | |||||||
| sat_137 | 5 | 0.00 | 2.30 | ||||||||
| satt385 | 5 | 64.74 | 2.49 | 2.21 | |||||||
| satt322 | 6 | 82.23 | 2.33 | 2.04 | |||||||
| satt389 | 17 | 79.23 | 2.35 | 2.11 | |||||||
| YLD | satt641 | 3 | 29.28 | 3.45 | 2.96 | 2.36 | 2.92 | ||||
| satt521 | 3 | 65.46 | 2.04 | 3.22 | 4.49 | 4.49 | 3.75 | 2.28 | |||
| satt316 | 6 | 127.67 | 3.10 | 3.76 | 2.48 | 2.35 | 2.07 | 2.10 | 3.06 | 3.51 | |
| sat_330 | 7 | 140.69 | 3.22 | 3.24 | 2.00 | ||||||
| sct_010 | 19 | 59.52 | 5.51 | 5.06 | 4.28 | 3.24 | 2.64 | 2.15 | 5.59 | 4.47 | |
| GMES4376 | 19 | 91.10 | 2.37 | ||||||||
NJ, Nanjing; NY, Nanyang.
Soybean simple sequence repeat (SSR) genetic linkage map showing the marker positions and estimated map distances (cM; indicated on the left of the vertical bars) based on the consensus linkage map of Song et al. (2004). Markers associated with any of five yield-related traits are indicated by red characters. Black character SSR markers indicated no association to any of the five yield-related traits in this study, DTF: days to flowering from sow, DTM: days to mature from sow, HSW: hundred seed weight, PN: pod number per plant, YLD: yield per plant.
The 7 SSR-trait associations related to the DTF involved 3 SSR loci; two SSR loci at satt322 on chromosome 6 and satt564 on chromosome 18 were identified under the BLUP value and in two experimental stations in 2012. The ten SSR-trait associations related to DTM involved 5 SSR loci; two SSR loci at sat_304 on chromosome 7 and satt521 on chromosome 6 were identified over two years and under the BLUP value. The 12 SSR-trait associations related to HSW involved seven SSR markers; three SSR loci at satt641 satt285 and satt516 were detected over two years in Nanjing, but only satt641 was detected under the BLUP value. Four SSR-traits associations related to PN, but all of that were detected only in one environment. The 13 SSR-trait associations related to YLD involved 5 SSR markers; two SSR loci at sct_010 on chromosome 19 and satt316 on chromosome 6 were identified under any condition. And, five out of the 18 SSR markers were associated with two or more traits.
Mining of the elite allelesThe phenotypic allele effect of each SSR that significantly associated with five yield-related traits was shown in Supplemental Table 3. Among the alleles associated with DTF, satt521-204 had the most positive phenotypic effect and able to increase DTF by 18.03 days, whereas satt564-191 had the most negative phenotypic effect (−18.31 days). Among the alleles associated with DTM, satt521-204 and sat_304-178 had the most positive phenotypic effect and able to increase DTM by 28.96 days, whereas satt564-191 and sat_304-162 had the most negative phenotypic effect (−21.08 days). Among the alleles associated with HSW, satt516-277 had the most positive phenotypic effect and able to increase HSW by 4.39 g, whereas satt285-275 had the most negative phenotypic effect (−2.04 g). Among the alleles associated with YLD, satt316-277 had the most positive phenotypic effect and able to increase HSW by 22.8 g, whereas sat_330-399 had the most negative phenotypic effect (−6.28 g). Among the alleles associated with PN, satt342-272 had the most positive phenotypic effect and able to increase PN by 95.10, whereas satt342-220 had the most negative phenotypic effect (−64.82).
Increasing the yield of soybean is a major target for soybean breeders. Indeed, there is much concern regarding the reduction of the diversity of the currently cultivated soybeans. Because early farmers used only a limited number of individual progenitors in the domestication process, only the seeds from the best plants were utilized to form the next generation, which led to a loss of genetic diversity (Doebley et al. 2006). After domestication, the genetic variation in soybean has been continually reduced by modern plant breeding. In present study, the average number of alleles per loci was 10.49, which is low compared to the 17.8 that was detected by Wen et al. (2009), but higher than the 5.6 alleles per number that was reported by Wang et al. (2010), perhaps due to differences in the samples, sample size and SSR markers that were selected. In the present study, the 9 EST-SSRs that were selected have very a low number of alleles, ranging from 2 to 6, with an average of 3.5 per locus and the marker that is selected will affect the result of the analysis of the allele number and genetic diversity.
Population structure can lead to the discovery of many false-positive QTL (Zhao et al. 2007) and several models have been developed to resolve the complication, including genomic control, Q+K model, PCA model (Devlin and Roeder 1999, Devlin et al. 2004, Price et al. 2006, Yu et al. 2006). Previous studies have demonstrated that the Q+K method is the most powerful method for performing association mapping (Stich and Melchinger 2009, Yu et al. 2006, Zhao et al. 2007). In the present study, four subpopulations were identified using a STRUCTURE analysis based on the Bayesian model and the population structure has different effect on different trait (Table 1).
Association mapping and potential usages of the results in soybean breedingBased on the GLM model (Q), a total of 118 marker-trait associations were identified for the five yield-related traits, involving 33 SSR markers (Table 3). Based on the MLM model (Q+K), forty-five SSR-trait associations were identified, involving 18 SSR markers. Except BE475343 associated with HSW on Chr2 in 2011NJ and satt285 associated with HSW on Chr16 in 2012NJ, all the SSR-trait associations identified by MLM were detected by GLM, but many of SSR-trait associations identified by GLM were not detected by MLM, which may be resulted from the effect of kinship.
Of these 18 SSR markers identified by MLM, five SSR markers were associated with two or more traits. Sat_334 on chromosome 12, associated with HSW, was close to satt442, which has been reported to be associated with HSW in wild soybean (Wen et al. 2008). These two loci that associated with HSW detected in this study were also identified in an F2 population derived from the crossing of cultivated and wild soybeans (Kan et al. 2012). However, to our knowledge, these QTL have not been identified in cultivated soybean populations, indicating that may be a QTL which involving soybean domestication. Sct_010 on chromosome 19, which is associated with YLD, has been reported by Kan et al. (2012) and the increase allelic from wild soybean. Satt316 on chromosome 6 associated with YLD and which have been reported by Reinprecht et al. (2006). Satt150 on chromosome 7, which is associated with HSW, has been reported to have a close linkage with soybean seed size and volume (Salas et al. 2006).
The majority of the loci that were associated with the five traits could only be identified in a specific environment, either that of the Nanjing Experimental Station or Nanyang Experimental Station, which indicated that wild soybean is very sensitive to the environment. However, some stable associations were identified in our study, such as satt546 and satt322, which were associated with DTF, sat_304 and satt521, which were associated with DTM and sct_010 and satt316, which were associated with YLD. A low threshold, −Log(P-value) ≥ 2.00, was used to detect the marker-trait association due to the limited number of marker used in this study. If high-density DNA polymorphism datasets are used for association mapping, additional markers with high −Log(P-value) may be obtained.
In order to use the results of the association analysis, we assessed the phenotypic allele effect of each SSR that associated with five yield-related traits and a number of elite allele was detected associated with five yield-related traits. These will be useful for molecular marker assist selection and molecular design breeding.
A small sample and limited markers were used in this study and the results need to be confirmed using linkage mapping or a large association population. However, the results are credible because many of the loci that were identified were associated with traits that were common with previous reports of linkage or association mapping.
This work was supported by the National Basic Research Program of China (973 Program) (2010CB125906, 2009CB118400), the National Natural Science Foundation of China (31000718, 31171573, 31201230, 31271749), Jiangsu Provincial Programs (BE2012328, BK2012768, BE2012747) and the Young Scholar Innovation Foundation of the Nanjing Agricultural University (KJ2011004).
