k-mer-based GWAS in a wheat collection reveals novel and diverse sources of powdery mildew resistance

Abstract

Wheat genetic resources hold the diversity required to mitigate agricultural challenges from climate change and reduced inputs. Using DArTseq, we genotype 461 wheat landraces and cultivars and evaluate them for powdery mildew resistance. By developing a k-mer-based GWAS approach with fully assembled genomes of Triticum aestivum and its progenitors, we uncover 25% more resistance-associated k-mers than single-reference methods, outperforming SNP-based GWAS in both loci detection and mapping precision. In total, we detect 34 powdery mildew resistance loci, including 27 potentially novel regions. Our approach underscores the importance of integrating multiple reference genomes to unlock the potential of wheat germplasm.

Fig. 1

Workflow to identify the genetic basis of resistance in the wheat powdery mildew pathosystem. A All accessions from the Swiss wheat collection were phenotyped using 10 Bgt isolates from around the globe. B All accessions from the collection were sequenced using DArTseq. C From the raw reads of the DArTseq, 31-bp k-mers were generated, and a presence/absence matrix was used to run GWAS. D Using the k-mer matrix and the phenotyping data, GWAS was used to find k-mers significantly associated with the phenotype. E All the DArTseq raw data were mapped to ten Triticum aestivum genomes as well as three progenitor genomes and the genome of T. spelta. F Manhattan plots were generated for each genome of reference. The significant peaks were extracted to select candidate genes

Fig. 2

Phenotyping of the Swiss wheat collection with 10 powdery mildew isolates representative of the global genetic diversity of Blumeria graminis f. sp. tritici. A PCA of 400 Bgt isolates with 9 of the 10 isolates used in this study highlighted. B Avirulence (blue) and virulence (red) pattern of the 10 isolates across 37 Pm-tester lines [24]. C Phenotype distribution of the isolate CHE_96224 on the Swiss wheat collection. Pictures represent example phenotypes for fully resistant, partially resistant, and fully susceptible seedling reactions. D Correlation plot of the phenotype of all accessions for each isolate. Background color represents the Pearson correlation value. E Heatmap representing the phenotype of each accession of the Swiss collection for the ten Bgt isolates sorted the same way as B. The three main clusters were split

Fig. 3

k-mer GWAS results. A Summary of the GWAS results based on CS for Pm1, Pm2, and Pm4 for all three matrices for all the isolates, as well as two regions only found using k-mers. The colors blue, brown, and purple respectively represent the SNP-Chip, the SNPs generated from the DArTseq data, and the k-mers. Filled squares represent the presence of an associated region. B UpSet plot for all k-mers significantly associated with resistance using the resistance phenotype of each of the 10 isolates. The biggest set of common significantly associated k-mers (5594) was found between the isolates ARG_4_2 and JPN_CHIKA, GRB_JIW2, CHE_97251, THUN12, and KAZ_1b. The first 14 groups are colored. The entire data set can be explored in the Shiny app at https://benjiapp.shinyapps.io/Manhattan_plot/. C Circular plot representing Manhattan plots comparing three matrices used for the GWAS of Bgt isolate CHE_96224. The inner circle in blue represents the SNP matrix generated using DArTseq, and the middle circle in orange represents the matrix from the SNP-chip. The outer circle represents the k-mer-GWAS where only k-mers significantly associated with resistance are displayed. The y-axis is the same for all three plots. D 3D dot plot of the different Manhattan plots for the region of Pm4 (red arrow on C). The color legends are as in A. The k-mer-GWAS SY Mattis represents the k-mer-GWAS using the SY Mattis genomes as a reference. The line is at the position of the Pm4 gene in the SY Mattis genome. E Proportion of k-mers mapping to all the Triticum aestivum genomes (red), some of the genomes (blue), and none of the genomes (gray). F Proportion of reads that do not map to any of the T. aestivum genomes, but map to genomes of wheat progenitors or relatives. The other 40% do not map to any (light blue) and only 0.02% map to all four genomes

Fig. 4

Identification of known Pm genes and new candidate genes for powdery mildew resistance. Manhattan plot for the Bgt isolate CHE_96224 and the chromosomes 2 A (A) and 5D (B) of the SY Mattis genome. The zoom-in of each of the regions of interest also shows the position of two of the Pm genes known to confer powdery mildew resistance as well as the LD pattern of the region. C Table summarizing the presence of the three copies of the Werner-like candidate genes across 11 assembled wheat genomes (gray part). Phenotypes of the same 11 genotypes for the 9 [10] isolates used in this study. Blue represents resistance, red susceptibility, and white is missing data. D Alignment of the genomic regions (15 Mb) on chromosome 2B containing the Werner-like exonuclease gene candidate for genomes containing the candidate gene (CS, Landmark, Norin 61, and Stanley), as well as other genomes not containing it (Renan, Mace, SY Mattis, and Lancer). E Presence/absence pattern of the associated region at the beginning of Chr3D for 11 genomes and for each of the isolates used in this study. F Alignment of the genomic region around the candidate gene at the beginning of chromosome 3D (4 to 6 Mb). The position of the GWAS peak, as well as the best gene candidates, is marked with a triangle. A longer alignment of the 12 first Mb of chromosome 3D is in Supplemental Fig. 16

Fig. 5

Genetic basis of wheat resistance at adult stage. A Correlation between the 2 years of field phenotyping for spring (black) and winter (green) wheat. r represents the Pearson coefficient. B GWAS using Spring Wheat 2024 for phenotypic values and for comparison of three genotype matrices. The inner circle is DArTseq SNP, the middle SNPs chip, and the outer k-mer GWAS. All SNPs are plotted for the two innermost circles, but only the k-mers above − log10(p value) of 3 are plotted in the outermost circle. Red arrows represent the main significantly associated regions. C Double Manhattan plot comparing k-mer GWAS from the 2 years of Spring wheat phenotyping. 2023 at the top and 2024 at the bottom. Only k-mer with a significance above − log10(p value) of 3 is displayed. The colors are representing the different subgenomes of wheat. Light, dark, and blue represent the A, B, and D genomes, respectively