Identification of loci controlling mineral element concentration in soybean seeds

Abstract

Background: Mineral nutrients play a crucial role in the biochemical and physiological functions of biological systems. The enhancement of seed mineral content via genetic improvement is considered as the most promising and cost-effective approach compared alternative means for meeting the dietary needs. The overall objective of this study was to perform a GWAS of mineral content (Ca, K, P and S) in seeds of a core set of 137 soybean lines that are representative of the diversity of early maturing soybeans cultivated in Canada (maturity groups 000-II).

Results: This panel of 137 soybean lines was grown in five environments (in total) and the seed mineral content was measured using a portable x-ray fluorescence (XRF) spectrometer. The association analyses were carried out using three statistical models and a set of 2.2 million SNPs obtained from a combined dataset of genotyping-by-sequencing and whole-genome sequencing. Eight QTLs significantly associated with the Ca, K, P and S content were identified by at least two of the three statistical models used (in two environments) contributing each from 17 to 31% of the phenotypic variation. A strong reproducibility of the effect of seven out these eight QTLs was observed in three other environments. In total, three candidate genes were identified involved in transport and assimilation of these mineral elements.

Conclusions: There have been very few GWAS studies to identify QTLs associated with the mineral element content of soybean seeds. In addition to being new, the QTLs identified in this study and candidate genes will be useful for the genetic improvement of soybean nutritional quality through marker-assisted selection. Moreover, this study also provides details on the range of phenotypic variation encountered within the Canadian soybean germplasm.

Keywords: GWAS; Minerals; QTL; Soybean; XRF.

Fig. 1

Pearson correlation between wet chemistry and ED-XRF for Ca, K, P and S content on a dry-weight basis among 30 soybean seed samples

Fig. 2

Distribution of Ca, K, P and S content in the seed of 137 Canadian soybean lines

Fig. 3

Models-based population structure in a core set of 137 Canadian soybean lines. a: Classification into seven populations using fastSTRUCTURE where each individual (from 1 to 137) is represented by a single vertical line and each color represents one cluster. b: Bootstrap consensus phylogenetic tree (2000 replicates) constructed using MEGA 7; each color represents a subgroup and seven subgroups were found in total and c: PCA eigenvalues computed using GAPIT. The total variance explained by each principal component (PC) decreased from PC1 to PC7 and, beyond PC7, the variance explained by each further PC remained low and stable

Fig. 4

Manhattan plots for mineral elements content in a core set of 137 Canadian soybean accessions. Manhattan plots for (a) calcium (b) potassium, (c) phosphorus and (d) sulfur content. Each dot/symbol indicates the degree of association between a single marker and a trait (y-axis) while the x-axis shows the physical position of each marker. A blue horizontal line indicates the significance threshold (FDR ≤ 0.05). Significantly associated markers are indicated as a red dot for FarmCPU while the blue cross (+) and asterisk (*) indicate SNPs that were declared significantly associated by CMLM or MLMM, respectively. These associations were superimposed on the Manhattan plots produced using FarmCPU

Fig. 5

Venn diagram for the 32 identified QTLs through three analytical approaches

Fig. 6

Stability of the eight QTLs detected by at least two models for Ca, K, P and S content. The core set of 137 early Canadian soybean accessions were grown in three additional environments (in 2017 or 2018, with [I_] or without [N_] supplemental irrigation). The phenotypic mean was calculated for the subsets of lines contrasting for the peak SNP at each of 8 QTLs previously detected by at least two of the three GWAS models. Each colored symbol represents the p-value for the contrast observed in one environment. The y-axis shows the -log10(p-value) of each test while the x-axis shows the reported QTLs associated with each trait. A red horizontal line indicates the Bonferroni significance threshold at ≤ − log10 (0.05/n), where n = number of co-identified QTLs per trait (e.g. 0.05/2 for Ca)

Fig. 7

Identification of a candidate gene underlying QTL S_#10 within the haplotype block on chromosome 20. Top panel: marker-trait associations within a ~ 80-kb interval (39,027–39,106 Kb) of Gm20. Middle panel: position and orientation of four gene models present in the 35-kb region that is defined by the left-most (Gm20: 39,042,071) and right-most (Gm20: 39,076,880) markers that are in perfect LD with the peak SNP (Gm20:39,076,484). The most likely candidate gene (Glyma.20G151500, Sulfate assimilation) is highlighted with a green asterisk. Bottom panel: pairwise LD among markers falling within the defined genomic region of interest. LD is indicated as D’× 100 and the empty squares indicate complete LD (D’ = 1). The position of the peak SNP (blue arrow) and candidate gene (green arrow) are shown