DISCUSSION
In this study we have characterized the genetic and pathotypic population structure of P. oryzae pathogens infecting traditionalindica varieties in the Yuanyang terraces of rice paddies. This traditional agrosystem, maintained over centuries (He et al., 2011), and where rice disease pressure was reported to be low (Sheng, 1990), provides a unique opportunity to decipher the impact of crop diversity on disease epidemics, especially of rice blast.
Our first important observation based on analysis of microsatellite and whole genome genetic variation, is the finding of new lineages of the rice blast pathogen endemic to YYT. Our analysis of population structure based on microsatellite (513 isolates) and whole genome (46 isolates) datasets revealed multiple lineages of P. oryzae coexisting in YYT, with relatively high levels of standing variation compared to previous results from Gladieux et al., [23] and Saleh et al.(Saleh et al., 2014). Three genetic lineages endemic to YYT, coexisted with two of the four worldwide lineages previously described [23]. Although the global linkage disequilibrium inferred from microsatellite data was significantly different from 0, analyses of LD-decay, PHI-tests and reticulations within each of the five lineages provided contrasted information regarding the existence of recombination (Supplementary Information 2, Fig. SI2.2). Both mating types were found in sympatry (same village; data not shown) within the 46 YYT P. oryzae fully sequenced isolates, but they were found within the same lineage only for WL1 (Supplementary Information 2, Table SI2.3), which, together with significant PHI-test and reticulations observed for this lineage, was consistent with the fact that it has been described as recombinant (Gladieux et al., 2018; Latorre et al., 2020; Saleh et al., 2014; Thierry et al., 2021). Conversely, only Mat-2 isolates were found among the 30 fully sequenced isolates assigned to lineage YYT1. Therefore, significant PHI-test and reticulations observed in the minimum spanning network for this lineage could be due to scarce genetic exchanges among lineages (Gladieux et al., 2018), or to footprints of historical recombination. In invasive pathogens, higher genetic diversity and signatures of recombination are expected in older, source populations (Ali et al., 2014; Thach, Ali, de Vallavieille-Pope, Justesen, & Hovmøller, 2016). Our observations therefore suggest that YYT area is very close to, if not included into, the centre of origin of rice-infecting P. oryzaepathogens in continental Southeast Asia hypothesized by previous studies (Gladieux et al., 2018; Saleh et al., 2012; Zhong et al., 2018).
Our second important observation is that P. oryzae populations are not specialized to traditional rice landraces in YYT. For pathogens mating within or onto their hosts, specialization should drastically restrict encounters of potential mates and reduce survival of offspring due to maladaptation of immigrants and hybrid offspring (Giraud, Gladieux, & Gavrilets, 2010; Gladieux et al., 2011), which should align the structure of pathogenic populations on that of the host. Our genotyping-by-sequencing data show that the rice accessions from YYT are structured into landraces with relatively high levels of genetic diversity, both within and among landraces, confirming previous findings based on 24 microsatellite markers (Gao et al., 2012).P. oryzae populations are also structured into different lineages, but our analysis reveals a complete lack of host-pathogen genetic co-structure. The fact that population subdivision in the pathogen does not mirror population subdivision in the host strongly suggests a lack of specialization to the host. However, co-structure between host and pathogen genealogies might be only detectable at those loci in the genomes that are specifically involved in coevolutionary processes (Märkle et al., 2021). The data presented here does not allow to test this hypothesis since we lack full genomic information on the plant side, and further genomic analyses are therefore needed to tackle this issue. GWAs analyses performed on the pathogen side detected loci that were involved in the interaction with at least two rice accessions, which is also consistent with a lack of local adaptation. We also analysed the phenotypic relationships among paired P. oryzae / rice samples by cross inoculating all isolates on their native and non-native plants. We showed that nearly all qualitative interactions were compatible. Liao et al. (2016) sequenced the genomes of two YYT rice accessions (Acuce and Xiaogu) and showed that they content 7 and 8 known R genes, respectively, without counting all the other unkown R genes. When 30 representative YYT isolates from indica accessions were inoculated against the modern rice varieties carrying known resistance genes, and showed that these P. oryzae isolates had lost may avirulence functions (Liao et al., 2016). Our result thus confirm that the great majority of the R genes present in indica YYT accessions were defeated by the P. oryzae population. Quantitative interactions indicated that P. oryzae isolates did not perform significantly better on their native than on their non-native plants, and thatP. oryzae isolates originating from plants of the same landrace did not perform significantly better on this landrace than on any other landraces, thus leading to rejection of the “home versus away” criteria of local adaptation (Blanquart, Klatz, Nuismer, & Gandon, 2013). Together, the results of our analysis of population genetic and pathotypic structure therefore reveals a complete lack of P. oryzae specialization to rice landraces and thus, a lack of adaptation to specific hosts, at least within the indica host compartment. Some studies have shown adaptation to specific host lines (Goyeau, Halkett, Zapater, Carlier, & Lannou, 2007; Goyeau, Park, Schaeffer, & Lannou, 2006), though this was not the case here. Although our results are consistent with a generalist life style in P. oryzae population, maladaptation cannot be definitely ruled out. Maladaptation in pathogens describes the case where isolates performance is significantly better on non-native hosts than on native hosts (Kniskern, Barrett, & Bergelsong, 2011). Given that we have assessed performance using an overall infection trait (the percentage of diseased leaf surface), we cannot exclude that other traits involved in fungal fitness that are not captured by our index (e.g. number of lesions, proportion of HR lesions among infective lesions, growth speed of lesion, sporulation capacity) would reveal such a maladaptation pattern.
Rice landraces and their P. oryzae pathogens from the Yuanyang terraces therefore represent a model system in which the pathogen is specialized to indica and japonica rice subspecies (Liao et al. 2016), but not specialized to the various landraces ofindica rice. This is consistent with predictions that the nature of the mechanism underlying immunity in the host or avirulence in the pathogen and the magnitude of divergence in immune systems between hosts has an impact on the likelihood of pathogen specialization (Giraud et al., 2010; Schulze-Lefert & Panstruga, 2011). Liao et al. (2016) showed that in the YYT area where japonica and indica rice subspecies are cultivated in sympatry, differences in immune systems between indica andjaponica subspecies would prevent the emergence of populations with a generalist lifestyle on both hosts, because a large effector complement is required to infect japonica rice, whileindica rice has a larger repertoire of immune receptors and therefore greater capacity to detect effectors that will trigger immunity. Differences in repertoires of immune receptors amongindica landraces might be sufficient to lead to the emergence of specialized P. oryzae populations in stable and homogeneous conditions. The maintenance of generalist P. oryzaegenotypes highlighted by our results could be explained by the elevated heterogeneity of the “host landscape” in the YYT area, with spatio-temporal variation in rice genotypes distribution (elevated genetic diversity within and among rice landraces, spatial mosaics of paddy fields sown with different landraces, and temporal turnover of the mosaics) impeding the emergence of specialized pathogens on specificindica plant genotypes or landraces. This could also be possible in case of lack of selection pressure due to R genes in the host, where the host lack many R genes, the absence of R genes in YYT landraces is highly unlikely. Unlike populations of P. oryzae from modern agrosystems, which tend to be largely clonal and infecting relatively stable and homogenous host populations (Gladieux et al., 2018; Zhong et al., 2018), populations infecting YYT landraces displayed higher genotypic and genetic diversities with occurrence of recombination. This last feature may contribute to the maintenance of the generalist lifestyle by re-shuffling virulence alleles amongP. oryzae pathogens. Finally, the Yuanyang Terraces bring to light an interesting observation, in which the populations of the pathogen are diverse and recombinant - characteristics of populations with high adaptive potential - while they generate less yield losses than the clonal populations observed in other parts of the globe. This suggests that the epidemiological models that should help the re-engineering of agrosystems should not ignore the scenarios leading to the emergence of recombinant and diverse pathogenic populations, as observed in our case study. Strategies should be adopted to avoid the emergence of specialist pathogen lineages and encourage the evolution of generalist pathogen which causes lesser disease burden. Varietal mixture at the spatial scale (even if not within the field) and temporal scale along with the encouragement of diverse sources of germplasm could be useful in this regard.
This study could inspire future work to foster the adoption of dynamic diversity (McDonald, 2014) to decrease the disease burden at the landscape level. This could be translated to other areas and other pathosystems to ensure a low disease burden without rapid emergence of virulences and low fungicide application.