3.2 | Host selection footprints
After combining the results of all the genome scan analyses, a total of 32 genomic regions with sizes ranging from 0.59 to 57 kb were detected (Figure 2, Table 2). Among the 32 regions, 25 were only identified with BayPass, two only with PoolFreqDiff, and three with both methods (Table 2). Fifteen genomic regions were detected using the six populations sampled in 2011. Twelve of the regions had already been identified in our previous study (Carlier et al., 2021b), including the five major and convergent regions: S1R2-Cu, S2R1-Cu, S2R5-Cu, S12R1-Cu and S12R2-Cu. The detection of three new regions in the present study was certainly because a larger number of SNPs were found using the latest more efficient version of the software used for the mapping step. We ran the same analysis on the eight additional populations sampled in 2013 and a total of 19 genomic regions were detected including two (S1R2-Cu, S12R1-Cu) already detected in the 2011 pool and 17 newly detected.
The genetic differentiation estimated between populations in the 32 candidate genomic regions (from 4 974 SNPs) was higher than that of the whole genome (Figure S1, Table S1) with FST values ranging from 0.003 to 0.226. Although none of the regions was found in all seven population pairs analyzed, 29/32 were significantly differentiated in at least two pairs, including three regions (S1R2-Cu, S6R5-Cu, S12R1-Cu) in five pairs. A complex population structure was observed with no apparent effect of the location, the sampling year, or the variety of origin (Figure S1). Tajima’s D computed from 1-kb non-overlapping window was significantly greater in the candidate genomic regions than in the whole genome in all the study populations and in either pool or individual sequencing in location 1 (Figure 3).