Background: Artificial selection played an important role in the origin of modern Glycine max cultivars from the wild soybean Glycine soja. To elucidate the consequences of artificial selection accompanying the domestication and modern improvement of soybean, 25 new and 30 published whole-genome re-sequencing accessions, which represent wild, domesticated landrace, and Chinese elite soybean populations were analyzed. Results: A total of 5,102,244 single nucleotide polymorphisms (SNPs) and 707,969 insertion/deletions were identified. Among the SNPs detected, 25.5% were not described previously. We found that artificial selection during domestication led to more pronounced reduction in the genetic diversity of soybean than the switch from landraces to elite cultivars. Only a small proportion (2.99%) of the whole genomic regions appear to be affected by artificial selection for preferred agricultural traits. The selection regions were not distributed randomly or uniformly throughout the genome. Instead, clusters of selection hotspots in certain genomic regions were observed. Moreover, a set of candidate genes (4.38% of the total annotated genes) significantly affected by selection underlying soybean domestication and genetic improvement were identified. Conclusions: Given the uniqueness of the soybean germplasm sequenced, this study drew a clear picture of human-mediated evolution of the soybean genomes. The genomic resources and information provided by this study would also facilitate the discovery of genes/loci underlying agronomically important traits.

Soybean cyst nematode (SCN, Heterodera glycines Ichinohe) is one of the most fatal pests of soybean (Glycine max (L.) Merr.) worldwide and causes huge loss of soybean yield each year. Multiple sources of resistance are urgently needed for effective management of SCN via the development of resistant cultivars. The aim of the present study was to investigate the genetic architecture of resistance to SCN HG Type 0 (race 3) and HG Type 1.2.3.5.7 (race 4) in landraces and released elite soybean cultivars mostly from China. A total of 440 diverse soybean landraces and elite cultivars were screened for resistance to SCN HG Type 0 and HG Type 1.2.3.5.7. Exactly 131 new sources of SCN resistance were identified. Lines were genotyped by SNP markers detected by the Specific Locus Amplified Fragment Sequencing (SLAF-seq) approach. A total of 36,976 SNPs were identified with minor allele frequencies (MAF) > 4 % that were present in 97 % of all the genotypes. Genome-wide association mapping showed that a total of 19 association signals were significantly related to the resistance for the two HG Types. Of the 19 association signals, eight signals overlapped with reported QTL including Rhg1 and Rhg4 genes. Another eight were located in the linked regions encompassing known QTL. Three QTL were found that were not previously reported. The average value of female index (FI) of soybean accessions with resistant alleles was significantly lower than those with susceptible alleles for each peak SNP. Disease resistance proteins with leucine rich regions, cytochrome P450s, protein kinases, zinc finger domain proteins, RING domain proteins, MYB and WRKY transcription activation families were identified. Such proteins may participate in the resistant reaction to SCN and were frequently found in the tightly linked genomic regions of the peak SNPs. GWAS extended understanding of the genetic architecture of SCN resistance in multiple genetic backgrounds. Nineteen association signals were obtained for the resistance to the two Hg Types of SCN. The multiple beneficial alleles from resistant germplasm sources will be useful for the breeding of cultivars with improved resistance to SCN. Analysis of genes near association signals may facilitate the recognition of the causal gene(s) underlying SCN resistances.

The relative abundance of five dominant fatty acids (FAs) (palmitic, stearic, oleic, linoleic and linolenic acids) is a major factor determining seed quality in soybean. To clarify the currently poorly understood genetic architecture of FAs in soybean, targeted association analysis was conducted in 421 diverse accessions phenotyped in three environments and genotyped using 1536 pre-selected SNPs. The population of 421 soybean accessions displayed significant genetic variation for each FA. Analysis of the molecular data revealed three subpopulations, which reflected a trend depending on latitude of cultivation. A total of 37 significant (p < 0.01) associations with FAs were identified by association mapping analysis. These associations were represented by 33 SNPs (occurring in 32 annotated genes); another four SNPs had a significant association with two different FAs due to pleiotropic interactions. The most significant associations were cross-verified by known genes/QTL or consistency across cultivation year and subpopulations. The detected marker-trait associations represent a first important step towards the implementation of molecular-marker-based selection of FA composition with the potential to substantially improve the seed quality of soybean with benefits for human health and for food processing.

Summary: Landraces often contain genetic diversity that has been lost in modern cultivars, including alleles that confer enhanced local adaptation. To comprehensively identify loci associated with adaptive traits in soya bean landraces, for example flowering time, a population of 1938 diverse landraces and 97 accessions of the wild progenitor of cultivated soya bean, Glycine soja was genotyped using tGBS^®. Based on 99 085 high‐quality SNPs, landraces were classified into three sub‐populations which exhibit geographical genetic differentiation. Clustering was inferred from STRUCTURE, principal component analyses and neighbour‐joining tree analyses. Using phenotypic data collected at two locations separated by 10 degrees of latitude, 17 trait‐associated SNPs (TASs) for flowering time were identified, including a stable locus Chr12:5914898 and previously undetected candidate QTL/genes for flowering time in the vicinity of the previously cloned flowering genes, E1 and E2. Using passport data associated with the collection sites of the landraces, 27 SNPs associated with adaptation to three bioclimatic variables (temperature, daylength, and precipitation) were identified. A series of candidate flowering genes were detected within linkage disequilibrium (LD) blocks surrounding 12 bioclimatic TASs. Nine of these TASs exhibit significant differences in flowering time between alleles within one or more of the three individual sub‐populations. Signals of selection during domestication and/or subsequent landrace diversification and adaptation were detected at 38 of the 44 flowering and bioclimatic TASs. Hence, this study lays the groundwork to begin breeding for novel environments predicted to arise following global climate change.

A genome-wide association study (GWAS) identifies trait-associated loci, but identifying the causal genes can be a bottleneck, due in part to slow decay of linkage disequilibrium (LD). A transcriptome-wide association study (TWAS) addresses this issue by identifying gene expression-phenotype associations or integrating gene expression quantitative trait loci with GWAS results. Here, we used self-pollinated soybean (Glycine max [L.] Merr.) as a model to evaluate the application of TWAS to the genetic dissection of traits in plant species with slow LD decay. We generated RNA sequencing data for a soybean diversity panel and identified the genetic expression regulation of 29 286 soybean genes. Different TWAS solutions were less affected by LD and were robust to the source of expression, identifing known genes related to traits from different tissues and developmental stages. The novel pod-color gene L2 was identified via TWAS and functionally validated by genome editing. By introducing a new exon proportion feature, we significantly improved the detection of expression variations that resulted from structural variations and alternative splicing. As a result, the genes identified through our TWAS approach exhibited a diverse range of causal variations, including SNPs, insertions or deletions, gene fusion, copy number variations, and alternative splicing. Using this approach, we identified genes associated with flowering time, including both previously known genes and novel genes that had not previously been linked to this trait, providing insights complementary to those from GWAS. In summary, this study supports the application of TWAS for candidate gene identification in species with low rates of LD decay.