Introduction

Flowering time is an important agronomic trait in plants, as it refers to the resources accumulated in vegetative storage tissues that will be reallocated to the seeds. Optimizing reproduction requires that the flowering date matches environmental conditions so that seeds can mature and disperse when appropriate conditions appear (Brachi et al. 2010). In soybean, flowering time is closely related to yield, quality and tolerances to environmental stresses (Cheng et al. 2011). Genetic modulation of the flowering time has been suggested as an effective means for improving yield, since it has a positive correlation with many yield-determining factors (Hanson 1985; Curtis et al. 2000; Kantolic and Slafer 2007). Natural variation in flowering time is considerable, but is a complex trait involving multiple genes. Recent developments in molecular biology and statistical methodologies for quantitative trait loci (QTLs) mapping have made it possible to identify genetic factors affecting flowering time. Such developments have the potential to significantly increase the rate of genetic improvement of plant species, through marker-assisted selection (MAS) of specific loci, genome-wide selection, gene introgression and positional cloning (Andersson 2001). In the past decades, most studies aiming to unravel the genetics of flowering time variation were performed by conventional linkage mapping in soybean. Since the first QTL experiment on growth stage traits was reported two decades ago (Keim et al. 1990), more than forty loci for flowering time and more loci for maturity have been detected on soybean A2, B1, C1, C2, F1, F2, E, G, H, I, J, K, L, M and N2 linkage groups (LG) in different populations, explaining 3.80–69.7 % of the phenotypic variation (Mansur et al. 1993, 1996; Orf et al. 1999; Wang 2001; Tasma et al. 2001; Zhang et al. 2004; Yamanaka et al. 2005). These reports, however, usually using intercross-derived populations, such as F2 and recombinant inbred lines (RILs), have some limitations in genome resolution because relatively few new recombination events are generated in the single generation separating parents. Therefore, additional strategies are required to locate QTL more precisely. The use of near isogenic lines (NILs) that differ at a single QTL is an effective approach and some of the flowering-related genes were cloned (Molnar et al. 2003; Liu et al. 2008; Watanabe et al. 2009). However, the development of NILs through repeated backcrossing is time-consuming and laborious (Tuinstra et al. 1997).

At present, fortunately, these limitations can be overcome by association mapping, which enables increased mapping resolution from a QTL interval to the candidate gene level. Recent years, many QTL studies were reported using association mapping in plants (Stracke et al. 2007; Jun et al. 2008; Hu et al. 2011; Li et al. 2011). Use of linkage-disequilibrium information increases the precision of QTL mapping because it exploits the entire number of recombinations accumulated since the original mutation generating the new QTL allele occurred (Meuwissen et al. 2002). Combined association and linkage mapping make it possible to exploit recombinations occurring both within and outside the pedigree and genotyped population. It also gives a clearer signal for QTL positions compared with association and linkage mapping alone. Additionally, the approach (combined association and linkage mapping) reduces the risk of false-positive QTL identification caused by accidental marker-phenotype associations when association and linkage mapping are used separately, and also increases the power and resolution of QTL mapping by combining all available information (Goddard 2005). This joint strategy for dissecting complex quantitative traits has been successfully employed, using linkage analysis to narrow suspected genes to centimorgan-scale regions, followed by an association analysis to fine map the genetic variation in regions showing linkage (Nemri et al. 2010; Lu et al. 2010).

Although several studies have reported the presence of one or more QTL affecting flowering time on chromosome 6, the results differ among studies with respect to the mapping procedure; the number of QTL detected, their positions, and the extent are affected by these QTL. The present study aimed at confirming and refining the previously identified flowering time-related QTLs, and to define the candidate region and to identify the polymorphisms underlying natural variation of flowering time. To achieve these, we first identified QTLs controlling flowering time using a segregating population (RIL) derived from a cross between ‘Bogao’ and ‘Nannong 94-156’. To further narrow down the QTL region, natural accessions were phenotyped under field conditions. Fine mapping is achieved by combining genome wide association statistical methods that control for population structure with primary mapping using RILs. Together, these methods enabled us to fine mapping the QTL and detecting the candidate genes affecting flowering time in soybean.

Materials and methods

Plant materials

To map QTLs for flowering time, a soybean segregating population, consisting of 152 F8:10 RILs was used. These RILs were derived from the cross between ‘Bogao’ and ‘Nannong 94-156’. For association mapping, the 192 soybean accessions, including landraces, cultivars, and breeding lines collected in China from latitude 53 to 24°N and longitude 134 to 97°E was used. The population of accessions was selected not only to represent all six ecological regions of soybean in China, but also soybeans with diverse reproductive related traits. Seeds of all accessions were obtained from the Germplasm Storage facility in the National Center for Soybean Improvement (Nanjing, China).

Field experiments and measure of flowering time

For linkage mapping, two-year trials using 152 RILs to map QTLs for flowering time were carried out at Jiangpu Station (32.0°N, 118.8°E), Nanjing, China in 2006 (sowing on 25 June) and 2007 (sowing on 20 June). There were two replicates with six plants in each replicate. Genotypes were sown in a 10 L pot with 4 kg dry soil, with two plants per pot. For association mapping, three field experiments using 192 accessions were carried out at Jiangpu Station, Nanjing (sowing on 15 June), and Maozhuang Station (34.4°N, 113.4°E), Zhengzhou (sowing on 25 June), Henan, China in 2008, and a repeat experiment was carried out in Zhengzhou, Henan in 2010 (sowing on 17 June). These experiments were in a completely randomized design with three replications, using 2 m in row length, 0.4 m in row space and 0.1 m in plants spacing. The flowering time was calculated as the number of days from the germination to the beginning bloom (R1, 50 % of the plants in a plot had an open flower at one of the top nodes with a fully expanded leaf) (Fehr et al. 1971).

Population genotypic data analysis

The 192 soybean accessions were genotyped with 80 unlinked SSRs, providing an even coverage of the soybean genome. The employed SSR markers are available at http://bldg6.arsusda.gov/pooley/soy/cregan/soymap1.html. The number of alleles and the polymorphism information content (PIC) per locus were calculated using the POWERMAKER 3.25 software (Liu and Muse 2005). The population structure was inferred from the SSR data using the STRUCTURE software version 2.2 (Falush et al. 2007). Six independent runs were carried out using with the following parameters: number of populations (K) from 1 to 10, burn-in time and Markov-chain Monte Carlo replication number both set to 500,000, model of admixture and correlated allele frequencies (See the Structure 2.2 documentation at http://pritch.bsd.uchicago.edu/software). The likely number of sub-populations present was estimated following Evanno et al. (2005), in which the number of sub-groups (Δk) is maximized. STRUCTURE produces a Q matrix that lists the estimated membership coefficients for each individual in each cluster and this was used in the subsequent association analysis.

QTL mapping

For linkage mapping, the composite interval mapping (CIM) program of WinQTLCart version 2.5 (Wang et al. 2005) was used to detect QTLs for flowering time, using the 152 soybean RILs. Empirical thresholds were computed using the permutation test (1,000 permutations, overall error level 5 %) for CIM (Churchill and Doerge 1994). Confidence intervals were set as the map interval corresponding to a 1-LOD decline on either side of the LOD peak. A genetic map, which has been constructed in our laboratory, comprising 306 markers was used for linkage mapping (Zhang et al. 2009). For association mapping, the 192 natural accessions, eight newly developed SSR markers and 10 SNP markers based on the physical map (http://soybase.org/gbrowse/cgi-bin/gbrowse/gmax1.01/#search) were used to further narrow the major QTL region (qFT6). LD between pairs of markers in qFT6 was estimated by the TASSEL software version 2.1 (Bradbury et al. 2007). Squared allele frequency correlations (D′ and r 2) were chosen for LD calculations. The significance of LD between markers was determined by a Fisher’s exact test. The general linear model (GLM, not considering population structure) analysis and GLM + Q (considering population structure) in TASSEL was employed to identify associations. Those polymorphisms with P < 0.05 were considered significantly associated to the trait.

Gene prediction in the target genomic region was conducted based on the soybean genome information and bioinformatics. The flowering related candidate genes were identified after submitting the predicted genes to a BLASTP query of the UniRef database (http://www.ebi.ac.uk/uniref/) and synteny comparisons between soybean and other dicotyledonous plants.

Results

Phenotypic variation and population structure

The means, standard deviation, skew, broad sense heritability and percentage of phenotypic variation explained by population structure of flowering time are shown in Table 1. The phenotypic values of the parents of showed flowering time ranged from 41 to 46 days and from 37 to 50 days in RILs over two years, respectively. There was obvious transgressive segregation for flowering time and the broad heritability in parents and RIL families averaging 91.8 and 92.2 % across two years (Table 1, Fig. 1). On the other hand, highly significant variation was observed among the 192 natural accessions, with flowering time per accession averaging 55–56 days across different years/sites and ranged from 27 to 77 days (Table 1, Fig. 1). The distribution of flowering time for these natural accessions is roughly normal, and its range exceeds that observed among the RILs at both sides of the distribution (Fig. 1). The ANOVA showed that the genotype effect and the genotype × environmental interaction effects were significant (P < 0.01) for flowering time (Table 1). In addition, The 80 SSR markers examined produced a total of 1,422 alleles among the 192 accessions assayed (Table S1). The average number of alleles per locus was 17.78 (ranging from 5 to 37). The mean PIC value was 0.81. Overall, it was clear that soybean accessions exhibited considerable natural variation in flowering time and showed very high genetic diversity.

Table 1 Descriptive statistics, ANOVA, broad sense heritability and percentage of phenotypic variation explained by population structure for flowering time
Full size table
Fig. 1
figure 1

Distribution of natural variation for flowering time (FT). The x-axis gives the days from the germination date to the mean flowering date. The distribution of flowering time scored for the 152 RILs in 2006 (red) and 2007 (green) and the 192 natural accessions (NA) in 2008 (Nanjing, purple; Henan, blue) and 2010 (Henan, orange)

Full size image

The evaluation of the population structure of the 192 soybean accessions indicated that there was an increase in LnP(D) value with increasing k value. The ad hoc quantity (Δk) showed a much higher likelihood at k = 2 than at k = 3–10 (Fig. 2a), suggesting that the population could be clustered into two major subpopulations (Fig. 2b). This result was consistent with previous work, which population structure was evaluated using 1536 SNPs in the same population (Hao et al. 2012). Statistical description for flowering time in the two subpopulations was summarized in Table S2. Nearly threefold difference (27–77 days) for flowering time was found between accessions. While between two subpopulations, a difference of 5.4 days for flowering time was found because of most accessions is early flowering varieties in subpopulation two (S2). The results of one-way analysis of variance (ANOVA) between subpopulations showed that the variation of flowering time was significantly different between subpopulations (Table S2). In addition, the effects of population structure on soybean flowering time were various in the 192 natural accessions, with R 2 average of 3.2–5.3 % (Table 1).

Fig. 2
figure 2

Population structure of 192 soybean accessions. a Δk value was over six iterations of running with putative k ranging from 1 to 10. b Two major subpopulations for 192 soybean accessions, the numbers on the x-axis show the accession number and numbers on the y-axis show the membership probabilities into the subpopulations. The colors of the bar indicate the two subpopulations identified through the STRUCTURE program

Full size image

Primary QTL mapping for flowering time by linkage analysis

To identify QTLs for flowering time in soybean, a genetic map comprising 306 markers has been constructed (Zhang et al. 2009) and 152 RILs were used to map QTLs for flowering time in 2006 and 2007, respectively. As a result, five QTLs were detected for flowering time and mapped to five chromosomes based on the pot experiments in two years (Table S3). Among them, qFT6, qFT8, and qFT20 were mapped on chromosome 6, 8 and 20 in 2006, while qFT6, qFT11, qFT18, and qFT20 were mapped on chromosome 6, 11, 18 and 20 in 2007. The alleles from Bogao at qFT6, qFT18 and qFT20 loci flowering time shifted earlier by 0.68-1.15 days, whereas those at qFT8 and qFT11 from Nannong 94-156 prolonged flowering time by 0.52–0.93 days. From Two main QTLs were co-located flanked by Sat_402-Satt489 on chromosome 6 and Sat_367-Sat_337 on chromosome 20 across two years, indicating a stable genetic effect for flowering time in different environments (Table S3). The QTL, qFT6, which explained 23.1 and 27.0 % of phenotypic variation in different years, could be a target region for identifying genes associated with flowering time (Table S3, Fig. 3a). In addition, this locus had been previously identified by Yamanaka et al. (2001) as a QTL controlling soybean flowering.

Fig. 3
figure 3

QTL mapping for flowering time on chromosome 6 using 152 RILs and 192 natural accessions at two-environments. a The QTL for flowering time mapped on chromosome 6 using 152 RILs by linkage mapping in 2006 and 2007. b Fine mapping centered on Sat_402 and Satt489 in the qFT6 region on Chromosome 6 by association mapping. GLM denote associations was identified by the general linear model analysis in TASSEL, not considering population structure; GLM + Q denote the estimated Q matrices were used in the association analysis, considering population structure

Full size image

Fine-mapping for flowering time by the partial scan of association analysis

To narrow down the major QTL region qFT6, which stably affected soybean flowering time across years, the 192 soybean accessions, eight previously developed SSR markers and 10 SNP markers inside this region were used to fine map qFT6 (Fig. 4b). Firstly, we assessed the LD structure across the qFT6 locus using eight SSR and 10 SNPs markers in the 192 soybean accessions. Two major LD blocks were found and one is including five loci within 2.4 Mb (positions 17.2-19.6 Mb, Table 2, Fig. S1). The pairwise r 2 of these markers in the LD blocks were higher and showed highly significant LD (P < 0.001) among the 192 soybean accessions. In addition, the scatter plots of the r 2 against distance also showed that there was a high LD in the qFT6 locus (Fig. S2). The pairwise r 2 varied from 0.000 to 0.988 with an average of 0.106. LD decay curve showed that LD extends 0.35 Mb in chromosome 6 for the 192 soybean accessions. The short extent of LD across the qFT6 locus thus provided sufficient genetic resolution for the following association mapping.

Fig. 4
figure 4

Fine mapping and candidate genes predict in the qFT6 region. a A flowering time QTL qFT6 was mapped to the interval between markers Sat_402 and Satt489 on soybean chromosome 6. b This QTL qFT6 was further delimited to ~300 kb region on chromosome 6 using the 192 natural accessions. c The black arrow indicates the site of the predicted gene between the markers BARC-014947-01929 and Satt365

Full size image
Table 2 The markers associated with flowering time (mean value) and their phenotypic variation
Full size table

Then association mapping was performed by general linear model (GLM) analysis (not considering population structure) and GLM + Q (considering population structure) using TASSEL software (Fig. 3b). To reduce both false positive and false negative risks caused by population structure, only the markers that were detected by both analysis methods were taken into account in this study. The results from the partial scan of association analysis showed that 11 markers (Satt286, Satt277, BARC-014947-01929, Satt365, Satt557, BARC-040213-07685, Satt658, Satt489, BARC-031099-06997, Satt289 and Satt100) were significantly associated with the flowering time in the qFT6 genomic region (Table 2, Fig. 4b). From Table 2, BARC-014947-01929 and Satt365, which with the highest significant level were closely associated with flowering time (P = 4.5 × 10−5 and P = 1.2 × 10−7). Then the haplotypes analysis between every two markers and regression analysis of the haplotypes to phenotypic data were performed to determine the joint effect of those loci pairs associated with flowering time. Interestingly, the results showed that the haplotype between BARC-014947-01929 and Satt365 could explain more phenotypic variation (26.5 %) than any other combination of markers. These results suggested that the target gene was located in ~300 kb between BARC-014947-01929 and Satt365 (Fig. 4c).

Candidate genes predicted for flowering time

Comprehensive analysis (combined with soybean genome information, bioinformatics and genome comparison) of this region predicted 26 annotated genes (Table S4). Among them, three were considered as candidate genes (Glyma06g22450, Glyma06g22650 and Glyma06g22680) related to flowering time or flower development after BLASTP queries of the UniRef database (http://www.ebi.ac.uk/uniref/) and synteny analyses between soybean and other dicotyledonous plants. These included the candidate gene Glyma06g22450, encoding a SBP domain, whose homologs play important roles in plant growth and flower development in Arabidopsis, rice and Antirrhinum majus (Klein et al. 1996; Yang et al. 2008). Other studies showed that SBP box gene regulates early flower development because the SBP-box can bind to the promoter of floral meristem identity gene (Yamaguchi et al. 2009). The candidate gene Glyma06g22650, encoding a MADS-Box protein (SRF-type transcription factor), whose homologs play fundamental roles during flower development in plants (Shore and Sharrocks 1995; Pelucchi et al. 2002). The other candidate gene Glyma06g22680, encoding a MYB related gene whose homologs play an essential role in flower development in plants (Preston et al. 2004; Higginson et al. 2003; Li et al. 2006). The candidate genes information identified in this study will be an important resource in marker-assisted selection for flowering time and for cloning the gene.

Discussion

Combined linkage and association mapping is powerful for fine mapping flowering time QTL in soybean

To date, the combined method (combined linkage and association mapping) for fine mapping have been used widely in humans and animals. These studies showed that fine mapping of an identified chromosomal region using the combined method could greatly reduce the QTL interval and was an important step toward identification of the gene and its causative allele (Olsen et al. 2005; Cohen-Zinder et al. 2005). However, the application of this effective method is very limited in plants. As we know, though it is not difficult to map a QTL to a 5–20 cM interval using primary segregating populations (such as F2, F2:3, RILs, etc.), cloning the underlying gene from a large interval is still a big challenge. In the present study, therefore, we combined linkage and association mapping that takes advantage of current and historical recombination events for QTL fine mapping in soybean based on the publicly available soybean genome information (Schmutz et al. 2010). The results showed that the combined method clearly outperforms each method used in isolation. For example, association mapping improves the positioning accuracy and linkage mapping increases our ability to distinguish true from false associations finely mapped by association mapping. In this study, population structure was not the dominant factor in variation of flowering time, but its effect with a 4 % average effect can be seen in this trait. Therefore, controlling for population structure is necessary for reducing the false associations. This empirical result supports the notion that linkage and association mapping are complementary methods (Liu et al. 2007; Manenti et al. 2009). This may be a solution that is especially powerful for quantitative traits such as flowering time, which variation overlaps with population structure.

Although the significant QTL affecting reproductive traits including flowering and maturity in soybean mapping to chromosome 6 was reported in several previous studies (Yamanaka et al. 2005; Su et al. 2010), in this study, the potential benefit of dual linkage and association mapping would be narrowing QTL intervals in RIL families and greatly increase our power to finely map genomic regions associated with phenotypic variation. At the first stage, primary linkage analysis using a 152 RILs confirmed the existence of the QTL affecting flowering time at a very high significance level (P < 0.001) but suggested the resolution was very low (8.6 cM). To refine the position of the QTL, association analysis of a partial scan of the major QTL region (qFT6) was then performed. A highly associated marker interval (BARC-014947-01929-Satt365), which can explain more phenotypic variation than any other haplotype combination of markers, was selected. These results indicate that combining linkage and association mapping provides an efficient and precise approach to further elucidate the genetic basis of phenotypes and fine mapping of soybean genes.

Favorable alleles for flowering time in research and soybean breeding

Marker assisted selection (MAS) for natural variation has been limited by resolution and germplasm diversity. The association mapping approach used here allows for rapid generation of selectable markers based on the performance of diverse germplasms. In addition, this provides more relevant markers in a broad genetic background, enabling breeders to search for favorable alleles in locally adapted germplasms. In this study, a causal base transition (A/T) of the SNP marker BARC-014947-01929 was highly associated with soybean flowering time, which leads to an 11.1 days variation of flowering time in different alleles. This polymorphism was corresponding to the trait variance and thus can be considered as the candidate sites for functional molecular markers. In addition, using the method of analyzing a ‘null allele’ (Breseghello and Sorrells 2006), we found that 20 accessions with the allele A283 at the Satt365 produce significantly prolonged flowering time (average 58.8 days) than the lines carrying other alleles. There are eight accessions with the allele A358 at the Satt365 shifted earlier for flowering time (average 37.9 days) than the lines carrying other alleles. Although these markers would be useful in MAS, further confirmation is necessary because the marker alleles are correlated with, not entirely predictive of, the gene alleles.

In summary, we identified five QTLs for flowering time; the QTL (qFT6) on soybean chromosome 6 was narrowed down to an interval spanning ~300 kb. The markers tightly linked to the QTL in qFT6 region would help breeders for MAS of appropriate accessions to flowering time. Additionally, the results of this study provide important information for map-based cloning of the causal genes for flowering time.