The whole-genome dissection of root system architecture provides new insights for the genetic improvement of alfalfa (Medicago sativa L.)

Abstract

Appropriate root system architecture (RSA) can improve alfalfa yield, yet its genetic basis remains largely unexplored. This study evaluated six RSA traits in 171 alfalfa genotypes grown under controlled greenhouse conditions. We also analyzed five yield-related traits in normal and drought stress environments and found a significant correlation (0.50) between root dry weight (RDW) and alfalfa dry weight under normal conditions (N_DW). A genome-wide association study (GWAS) was performed using 1 303 374 single-nucleotide polymorphisms (SNPs) to explore the relationships between RSA traits. Sixty significant SNPs (-log ₁₀ (P) ≥ 5) were identified, with genes within the 50 kb upstream and downstream ranges primarily enriched in GO terms related to root development, hormone synthesis, and signaling, as well as morphological development. Further analysis identified 19 high-confidence candidate genes, including AUXIN RESPONSE FACTORs (ARFs), LATERAL ORGAN BOUNDARIES-DOMAIN (LBD), and WUSCHEL-RELATED HOMEOBOX (WOX). We verified that the forage dry weight under both normal and drought conditions exhibited significant differences among materials with different numbers of favorable haplotypes. Alfalfa containing more favorable haplotypes exhibited higher forage yields, whereas favorable haplotypes were not subjected to human selection during alfalfa breeding. Genomic prediction (GP) utilized SNPs from GWAS and machine learning for each RSA trait, achieving prediction accuracies ranging from 0.70 for secondary root position (SRP) to 0.80 for root length (RL), indicating robust predictive capability across the assessed traits. These findings provide new insights into the genetic underpinnings of root development in alfalfa, potentially informing future breeding strategies aimed at improving yield.

Figure 1

Phenotype analysis of 171 alfalfa samples based on six root system architecture (RSA) traits. (A) Comparison of six root traits among different improved statuses. (B, C) Box plots that display RDW and RL among different improved statuses, respectively. The mean values of the materials in different improved statuses are displayed. (D) Cluster analysis of 171 alfalfa samples based on six RSA traits. (E) Comparison of six root traits among the three clusters (groups 1–3). (F) Proportion of materials from three clusters (groups 1–3) in the landrace and cultivar. RN, root number; TRD, taproot diameter; SRD, secondary root diameter; RDW, root dry weight; RL, root length; SRP, secondary root position. Here, n represents the number of alfalfa materials in different categories.

Figure 2

Correlations between root system architecture (RSA) traits and forage yield-related traits. The size of the pie chart indicates the strength of the correlation. Asterisks denote significance (^*P < 0.05, ^**P < 0.01, ^***P < 0.001). RN, root number; TRD, taproot diameter; SRD, secondary root diameter; RDW, root dry weight; RL, root length; SRP, secondary root position; RPH, regeneration plant height; PH, plant height; BN, branch number; LA, leaf area; DW, dry weight. “N_” and “D_” symbolize normal and drought conditions, respectively. For example, N_DW and D_DW represent the forage dry weight under normal and drought conditions.

Figure 3

GWAS identification of candidate genes for variation in alfalfa root traits. (A) Circular Manhattan plots of the association analysis for the six root system architecture (RSA) traits using Blink. The dotted red line indicates the significance threshold of a P-value of 1 × 10⁻⁵. ①–⑥ represent different traits: ① RN, root number; ② TRD, taproot diameter; ③ SRD, secondary root diameter; ④ RDW, root dry weight; ⑤ RL, root length; and ⑥ SRP, secondary root position. ⑦ distribution of SNP markers on eight chromosomes in the association pool. The color represents the density of the SNP markers. (B) GO enrichment analysis of 308 GWAS candidate genes. These candidate genes were identified in our GWAS results within 50 kb of the significant SNPs. (C and D) GWAS identification of MsMKK6 and MsVTI13 (significant SNP located within the gene) as candidate genes for RSA variation. Each set of plots comprises a partial Manhattan plot (top left), the candidate gene structure, an LD heat map (bottom left), and the phenotype values of different haplotypes (right). (E and F) GWAS identification of MsLBD2 and MsP4H5 (significant SNP located within the LD block) as candidate genes for RSA variation. Each set of plots comprises a partial Manhattan plot (top left), the distribution of genes within the LD block, an LD heat map (bottom left), and the phenotype values of different haplotypes separated using the top significant SNP (right). In each box, asterisks denote significant differences between haplotype groups (t-test, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001). “n” in C–F denotes the number of genotypes in each haplotype group, and each point represents the mean value of six clone plants of each genotype.

Figure 4

Role of favorable haplotypes in alfalfa breeding progress. (A) The proportion of favorable haplotypes in different improved status materials across the six root-related traits. The “n” below each trait indicates the number of associated SNPs (−log₁₀(P) ≥ 4), whose phenotype has significant differences (pairwise.t.test, P < 0.05) between the haplotype group. (B) Distribution of favorable haplotypes related to RDW in alfalfa materials from different improved statuses: cultivar (n = 70), landrace (n = 70), wild (n = 13), and uncertain (n = 18). Further details of the distribution of favorable haplotypes for the six RSA traits are included in Figs S4 and S5 and Table S7. (C) The number of favorable haplotypes corresponding to root dry weight (RDW) in our association panel. Linear regression was used for the analysis, and the coefficient of determination (R²) and P-value of the resulting equation were provided. The x-axis represents the number of favorable haplotypes identified by GWAS. The y-axis represents the weight of the dry root. Additional results for the remaining five RSA traits are shown in Fig. S6. (D, E) Number of favorable haplotypes in alfalfa materials with different forage dry weights (N_DW and D_DW). The x-axis shows the group divided according to the forage dry weight. The y-axis is the total number of favorable haplotypes related to the appropriate root system architecture identified by the GWAS. Five RSA traits (RN, TRD, SRD, RDW, and RL) related to N_DW, and Three RSA traits (RN, TRD, and RL) related to D_DW were selected for this analysis. The lower and upper boundaries represent the 25th and 75th percentiles, respectively. The middle horizontal line represents the median value. Different letters indicate significant differences at P < 0.05 (Tukey's all-pair comparisons). RN, root number; TRD, taproot diameter; SRD, secondary root diameter; RDW, root dry weight; RL, root length; SRP, secondary root position; DW, dry weight. “N_” and “D_” symbolize normal and drought conditions, respectively. N_DW and D_DW represent forage dry weight under normal and drought conditions, respectively.

Figure 5

Genomic prediction accuracy of six machine learning models among seven SNP sets (Set1–7) for each root-related trait. (A–F) Genomic prediction accuracy for RN, TRD, SRD, RDW, RL, and SRP, respectively. Set1 contained 306 411 SNPs filtered by LD. Set2–Set7 were obtained using GWAS. SNPs with -log₁₀(P) ≥ 5 (Set2), SNPs with −log₁₀(P) ≥ 4 (Set3), top 300 association markers (Set4), top 500 (Set5), top 1000 (Set6), and top 5000 (Set7). RN, root number; TRD, taproot diameter; SRD, secondary root diameter; RDW, root dry weight; RL, root length; SRP, secondary root position.