Lack of genetic co-structure between hosts and pathogens
To address the question of specialization to the host in P. oryzae population, we first compared the genetic structure of hosts and pathogens within the 46 paired samples of indica landraces andP. oryzae isolates from YYT. The total evidence tree built from rice GBS data (26,860 SNPs after removal of sites with missing; Fig. 3, left tree) showed that, as expected, YYT rice accessions clustered according to landrace vernacular name. The few exceptions to this clustering pattern could be explained by seed movements due to farmer practices at the time of harvest, resulting in variety mixture within paddies. Assignment analysis based on genomic data (Fig. 3, left barplots) further confirmed the clustering of YYT rice accessions by landrace name, and defined five main genetic clusters corresponding to Xiaogu, Hongyang, Acuce, Hongjiao and Baijao landraces. One accession (BJ_Q_B06) was considered as admixed. These results confirmed that landraces names in YYT correspond to well-defined genetic clusters, corresponding to landraces, i.e., populations of different − though genetically-related − genotypes.
Population subdivision was also clearly evidenced for the pathogen, as shown both by the phylogenetic tree and the clustering analyses (Fig. 3, right panel). While considering five clusters (i.e. at K=5), the grouping of the 46 sequenced P. oryzae isolates was congruent with the subdivision previously inferred from the phylogenetic network (Fig. 2C): one group (yellow) encompassed three isolates assigned to worldwide lineages W-Lineage 1, one group (blue) encompassed five isolates assigned to the worldwide lineage 3, and the remining isolates were assigned to three groups specific to YYT (green, red and pink groups, corresponding respectively to lineages YYT1, 2 and 3 of Fig. 2C). Pairing host and pathogen samples in genome genealogies clearly showed a complete lack of genetic co-structure between host and pathogen populations (black lines in Fig. 3): pathogens did not cluster according to the landrace their host of origin belongs to, thus suggesting a lack of specialization to the host. However, since coevolution is likely driven by a small number of interacting loci in the genomes of hosts and pathogens (Märkle et al. 2021), it is possible that co-structure might be visible only at these coevolving loci, and not at the entire genome scale.
The lack of any subdivision explained by host in P. oryzaepopulations sampled on indica landraces from YYT was also confirmed by the DAPC analysis of microsatellite data, since genetic subdivision inferred from these data did not match the host of origin (Fig. 2B). Genetic diversity of pathogens estimated with microsatellite data was high on most rice landraces in YYT (except for HongYang 3 and an unknow landrace from which a high proportions of P. oryzaeclones were sampled), ranging from 0.292 to 0.626 with a mean value of 0.445 (Supplementary Information 3, Table SI3.1), which compares to the ten most diverse populations described in Saleh et al. (2014). Among the 289 different microsatellite multilocus genotypes (MLGs) recovered from the 557 P. oryzae isolates sampled in YYT, 48 MLGs were detected more than once, 31 of which (64.5%) on multiple indica landraces (Table SI3.2). This confirmed that pathogen genotypes were distributed on all rice landraces.