Identifying QTL–allele system of seed protein content in Chinese soybean landraces for population differentiation studies and optimal cross predictions

Soybean originated in ancient China has been quickly extended globally as a major protein and oil crop. The QTL–allele constitution of seed protein content (SPC) in the Chinese soybean landrace population (CSLRP) was studied using a representative sample composed of 365 accessions tested under multiple environments and analysed under the novel restricted two-stage multi-locus genome-wide association study (RTM-GWAS) procedure based on 29,121 SNPLDB (single nucleotide polymorphism linkage disequilibrium blocks) markers. The SPC varied from 37.51 to 50.46% among accessions, for which 89 QTLs, each with 2–9 alleles in a total of 255 alleles were identified, accounting for 83.16% of the phenotypic variation covering most of the genetic variation (h2 = 84.31%). The QTL–alleles of the 365 landraces were organized into a 255 × 365 QTL–allele matrix as the compact form of SPC genetic constitution in CSLRP. Of the 89 QTLs, 53 showed significantly differentiated allele frequency distribution patterns among geographic eco-regions (sub-populations). There were 32.09% alleles not common among sub-populations but found only in some sub-populations; new allele(s) emerged on some loci in some respective sub-populations, with Eco-region III showing less but Eco-region VI more emergence. The QTL–allele matrix was also used for prediction of optimal crosses for breeding purpose to reach a 99th percentile potential of up to 54.81%, more than the highest accession (50.46%). From the 89 QTLs, 59 SPC candidate genes involving biological processes, cellular components and molecular functions were annotated. Among them, Glyma18g13574 and Glyma20g21370 were inferred as two of the major SPC genes in the whole genome.