Lack of specialization of P. oryzae populations to their hosts.
Cross-infection compatibility was tested for 33 rice plants and their paired P. oryzae isolates from YYT (1,089 possible combinations; Supplementary information SI3, Fig. SI3.1). Qualitative results showed a lack of phenotypic specificity for the vast majority of P. oryzaeisolates to their native rice accession or to plants belonging to the same landrace as their native plant. Indeed, among the 1,082 combinations yielding analysable results (data were missing for 7 combinations), 1,025 interactions (94.7%) were fully compatible, only 3 (0.3%) were incompatible, and 54 (5%) were scored as undetermined. YYT indica landraces were previously shown to include numerous R genes: not less than 16 known R genes were detected in two fully sequenced genomes of the accessions Acuce and Xiaogu (Liao et al. 2016), and they probably also include more unknown R genes. Our result suggest that all major resistance genes present in this set of indica landraces genotypes are overcome by P. oryzae .
Analysis of quantitative interactions in the matrix, measured as the average diseased leaf area, revealed a lack of adaptation of P. oryzae to their native host or landrace. ANOVA of the average diseased leaf area, which can be interpreted as the performance of a givenP. oryzae isolate on a given accession, showed that the effect of the isolate*accession combination (F = 1.8, P < 10-16, df = 1088) remains highly significant after removing the significant effect of the experimental replicate (F = 2092.6, P < 10-16, df = 1). We used this ANOVA model to estimate the adjusted performance of each isolate on each accession. Heatmaps of the adjusted performance (Fig. 4) showed differential quantitative responses on the different rice accessions for all P. oryzae isolates, with only one isolate being very weakly aggressive (green color on Fig. 4) on all rice accessions (CH1877) and no isolate being highly aggressive (red color on Fig. 4) on all rice accessions. Except for four isolates (CH1897, CH1900, CH1901, CH1905), the adjusted performance of each P. oryzae isolate was not significantly better on its native rice accessions than on all other accessions (Fig. 4A, Fig. SI3.3). Also, the average performance of allP. oryzae isolates originating from plants of a given landrace was not significantly better on all plants from this landrace than on plants of other landraces (Fig. 5). Local adaptation of P. oryzaeto different rice accessions should also imply that different genes are involved in the interaction with different rice accessions. To test this hypothesis, we performed GWAs analyses on the fungal side, using the 33 adjusted performances on the different YYT accessions as phenotypic traits to analyse (Supplementary Information SI4). Among the 27 markers statistically correlated with at least one phenotypic trait, the majority (22/27) was involved in the interaction with at least two accessions.
Hierarchical clustering by columns and lines (Fig. 4B) showed a lack of structure in the matrix, neither according to rice landraces, or to genetic lineages of P. oryzae isolates themselves. We further analysed nestedness and modularity within the matrix following Moury et al. (2021). The WINE estimate of nestedness was 0.55 and was significant (P=0 and 0.01 after 100 random with null models R1 and R2, respectively). According to Moury et al., this shows that the variance of the quantitative trait value is better explained by a statistical model that does not include a rice accession*P. oryzae isolate interaction term, in other words, that there is no specificity between accessions (or groups of accessions) and P. oryzae isolates (or groups of isolates). The modularity estimated with the spinglass algorithm was low (0.06), albeit significant (P=0 after 100 random simulations with null models R1 and R2).
Altogether, these results strongly suggest that P. oryzaepopulation in YYT did not adapt specifically to their native rice accessions or to any indica landrace.