Genome-Wide Association and Prediction of Traits Related to Salt Tolerance in Autotetraploid Alfalfa (Medicago sativa L.)

Abstract

Soil salinity is a growing problem in world production agriculture. Continued improvement in crop salt tolerance will require the implementation of innovative breeding strategies such as marker-assisted selection (MAS) and genomic selection (GS). Genetic analyses for yield and vigor traits under salt stress in alfalfa breeding populations with three different phenotypic datasets was assessed. Genotype-by-sequencing (GBS) developed markers with allele dosage and phenotypic data were analyzed by genome-wide association studies (GWAS) and GS using different models. GWAS identified 27 single nucleotide polymorphism (SNP) markers associated with salt tolerance. Mapping SNPs markers against the Medicago truncatula reference genome revealed several putative candidate genes based on their roles in response to salt stress. Additionally, eight GS models were used to estimate breeding values of the training population under salt stress. Highest prediction accuracies and root mean square errors were used to determine the best prediction model. The machine learning methods (support vector machine and random forest) performance best with the prediction accuracy of 0.793 for yield. The marker loci and candidate genes identified, along with optimized GS prediction models, were shown to be useful in improvement of alfalfa with enhanced salt tolerance. DNA markers and the outcome of the GS will be made available to the alfalfa breeding community in efforts to accelerate genetic gains, in the development of biotic stress tolerant and more productive modern-day alfalfa cultivars.

Keywords: GBS; abiotic stress; allele dosage; association mapping; genomic selection; polyploid.

Figure 1

Single nucleotide polymorphism variants (SNVs) identified in alfalfa (Medicago sativa) populations developed in Logan, Utah (A) Histogram of filtered variants called by Next Generation Sequencing Experience Platform (NGSEP) showing distribution by minor allele frequency and classified by function after annotation. (B) Distribution of GBS SNP markers across eight Medicago truncatula chromosomes using 1 Mb window. The colored lines represent the marker density as showing on the right color legends.

Figure 2

Frequency of allele dosage in autotetraploid alfalfa (Medicago sativa) for 6862 high-quality biallelic SNVs obtained from NGSEP pipeline in the Logan dataset. A represents dosage of the major allele and B is for the minor allele dosage.

Figure 3

Manhattan plots showing marker–trait association for vigor (V) in alfalfa populations at Othello Washington (WA) and Castle Dale Utah (UT). (A) Markers identified by general model in the UT dataset. (B) Markers identified by diplo-general model in the UT dataset. (C) Markers identified by general model in the WA dataset. (D) Markers identified by diplo-general model in the WA dataset. The threshold of 0.05 was used for significant markers according to the Bonferroni method.

Figure 4

Manhattan plots showing marker–trait associations for yield datasets in alfalfa (Medicago sativa) at Othello, Washington over two years. (A) Markers identified by general model in All 2018. (B) Markers identified by 2-dominant reference model in All 2018. (C) Markers identified by diplo-general model in July 2018 dataset. (D) Markers identified by general model in August 2018. (E) Markers identified by general model in September 2018. (F) Markers identified by diplo-general model in June 2019. (G) Markers identified by diplo-additive model in June 2019. (H) Markers identified by 1-dominant reference model in June 2019. (I) Markers identified by diplo-general model in July 2019. (J) Markers identified by diplo-additive model in July 2019. (K) Markers identified by 1-dominant reference model in July 2019. (L) Markers identified by 2-dominant reference model in September 2019 dataset. Markers threshold was set using Bonferroni > 0.05.

Figure 5

Linkage disequilibrium (LD) among markers associated for yield and vigor under salt stress. Haploview v4.2 [14] and pairwise LD values ($r^2\times 100$) were used for 27 SNPs associated with yield and vigor under salt stress (green color) and their surrounding SNPs in 10 kb (black color). Bright red coloring indicates $D^{\prime }=1,LOD\ge 2$; blue coloring indicates $D^{\prime }=1,LOD<2$; white coloring indicates $D^{\prime }<1,LOD<2$; shades of pink/red coloring indicates $D^{\prime }<1,LOD\ge 2$.