3.3 | Adaptive phenotypic variation in quantitative pathogenicity traits
Genome wide association studies (GWAS) were performed using the 63 samples collected in location 1 in 2011 that were both sequenced and phenotyped. No significantly associated SNP nor genomic regions were detected using two models either with a false discovery rate (FDR) of 5% or the local score procedure. Adaptive phenotypic variation was further investigated considering the 32 genomic regions showing host selection footprints. Phylogenic trees were computed for each region from the genomic sequences of the 63 sequenced isolates (example of the region S1R2-Cu in Figure S3). The two regions with the lowest sizes (S3R0-Cu and S9R0-Cu) were first discarded because no SNP was detected from this data. In most of the other regions, the isolates clustered in two main groups of close DNA sequences, hereafter referred to as haplotype 1 and haplotype 2 (Table S2). However, two genomic regions (S9R1-Cu and S12R3-Cu) were discarded because clusters of haplotypes were difficult to identify due to missing data and possible recombinant haplotypes. In the remaining 28/30 regions, each isolate was assigned to either haplotype 1 or haplotype 2. Between one and six individuals with intermediate haplotypes that may have resulted from recombination between the two main haplotypes in four regions were discarded (Table S2). The distribution of the DLA-R and DLA-S traits was then compared between the two haplotypes using a non-parametric Wilcoxon test. If a significant difference was observed, the haplotype performing the best (which could be either haplotype 1 or haplotype 2) was considered as the advantageous haplotype in the variety concerned. The results suggest association of at least one haplotype with at least one trait in 17/28 regions. A significant difference between haplotypes was observed for both traits (DLA-R and DLA-S) in 3/28 regions, where haplotype 1 appeared advantageous in both cultivars in two regions (S2R1-Cu and S4R4-Cu) while in the latter (S9R3-Cu), haplotype 1 was advantageous in the resistant cultivar and haplotype 2 in the susceptible cultivar (Table 3 and S3). In 9/28 regions only haplotype 1 appeared to be more advantageous in the resistant cultivars. The opposite was observed in 5/28 in which only haplotype 2 appeared to be more advantageous in the susceptible cultivar.
To further investigate correlations with phenotype in the 28 candidate regions in which haplotypes could be defined, a redundancy analysis (RDA) was conducted. Differentiation between the two populations (CU1S2 and CU1R2) was first observed using this analysis according to the x axis (RDA1) but not according to the y axis (RDA2, Figure 4A). The angle between the two vectors corresponding to DLA-S and DLA-R approached 90° suggesting no correlation between the two traits. The sign of the RDA score indicates the direction of the correlation for a given trait, a positive or a negative score indicating that haplotype 1 or haplotype 2 is correlated with the trait, respectively. An unequal contribution to DLA-S and DLA-R of the 28 candidate genomic regions was observed when we plotted the RDA scores corresponding to all candidate regions (Figure 4B). RDA scores ranged from 0.419 to -0.232 for DLA-R and from 0.234 to -0.301 for DLA-S, suggesting heterogeneous contributions of the candidate regions for the two traits (Table S3). A significant correlation (permutation test with 10% threshold) was found using the RDA approach considering both traits in 13 genomic regions for which a significant difference in pathogenicity was previously observed between haplotype 1 and haplotype 2 using the Wilcoxon test (Table 3). The S2R1-Cu region was significantly correlated with both traits, with RDA scores of 0.347 and 0.234 for DLA-R and DLA-S, respectively, supporting the hypothesis that haplotype 1 is advantageous for both traits. Eight out of 13 regions were only correlated with DLA-R with positive RDA scores ranging from 0.419 to 0.285, suggesting that haplotype 1 is the advantageous haplotype for the resistant cultivars. By contrast, the RDA scores of the 4/13 regions showing a significant correlation with DLA-S ranged from -0.301 to -0.264 and this negative correlation supports the hypothesis that haplotype 2 is advantageous for the susceptible cultivar.