INTRODUCTION
A major challenge for modern agriculture is to implement sustainable solutions ensuring food security by promoting crop health while decreasing our reliance on agrochemicals (Tilman, 2011). Globalization and agricultural intensification have disrupted the coevolutionary battle in which plants and pathogens engage in natural ecosystems, generally favoring larger pathogen population sizes (i.e., more widespread and intense epidemics) and thus rapid evolution of pathogen aggressiveness and infectivity (Burdon & Thrall, 2008; Gladieux et al., 2015; Parker & Gilbert, 2004). Monocultures of varieties bred for high yield and disease resistance are also vulnerable to disease outbreaks, because they impose strong directional selection on pathogens, and because mutants that can overcome resistance in one individual plant can infect all plants in a field and hence quickly spread (Hill, 2001; Stukenbrock & McDonald, 2008; Zhan, Thrall, Papaïx, Xie, & Burdon, 2015). The literature in plant pathology provides many examples of so-called boom-and-bust disease dynamics, in which newly deployed resistant varieties are rapidly colonized by pathogen variants able to overcome new resistance genes (Brown, 1994; de Vallavieille-Pope et al., 2012; Guérin, Gladieux, & Le Cam, 2007). In contrast, in unmanaged, natural, ecosystems, pathogen prevalence is generally lower, and disease epidemics more limited in time and space (Burdon and Thrall 2014). Long-term empirical studies and modeling work suggest that ecological and environmental heterogeneity, with highly patchy and variably diverse host plant populations, varying abiotic conditions and the co-occurrence of closely related but distinct or phylogenetically-distant plants can contribute to limit the burden of disease in the wild (Burdon & Thrall, 2008; Zhan et al., 2015). Metapopulation dynamics and frequency dependent selection in heterogeneous environment create a mosaic of local coevolutionary scenarios ranging from local adaptation to maladaptation (Laine, 2007), depending on the biology of the system.
Traditional agrosystems are promising models for deriving new disease management rules for modern agrosystems (Chentoufi et al., 2014; Sahri et al., 2014). Transfering knowledge gained from studies of the mechanisms underlying the stability of plant-pathogen associations in the wild (Burdon & Thrall, 2008) is hindered by divergence in the structure and complexity of unmanaged ecosystems and modern agrosystems caused by marked differences in the impact of humans on the spatio-temporal distribution of host diversity between the two types of systems. Unlike modern agrosystems and modern crops, which have been engineered and intensely selected to improve yield and quality under relatively low-stress conditions, landraces and their agrosystems have been selected and developed for their capacity to provide stable yields in specific environmental conditions and under low-input agriculture. The value of landraces as sources of genetic variation, or the value of traditional agrosystems as models for re-engineering modern agrosystems, are generally accepted (Feuillet, Langridge, & Waugh, 2008). Some studies have also been done at the field based experimental level (Zhan & McDonald, 2013). However, there has been remarkably little effort to investigate causal links between the structure of genetic and phenotypic diversity in crops and pathogens on the one hand and disease dynamics on the other.
The traditional, centuries-old agrosystem of the Yuanyang terraces (YYT) of flooded rice paddies (Yunnan, China) represents an outstanding model system to investigate the factors that render plant agrosystems less conducive to disease (Liao et al., 2016). More than 180 landraces, mostly indica rice, have been grown for centuries in the Yuanyang terraces (Gao, Mao, & Zhu, 2012; Jiao et al., 2012; Yang et al., 2017). The Yuanyang landraces are famous for being little affected by diseases (Sheng, 1990), such as rice blast caused by Pyricularia oryzae (syn., Magnaporthe oryzae ), which is an important rice disease worldwide (R. Dean et al., 2012).
Rice blast is widely spread on all ecotypes of rice and in different ecological zones, where it has a massive socio-economic impact on human populations (R. Dean et al., 2012; Tharreau et al., 2009). Rice blast is caused by one out of several host-specific lineages of P. oryzae(Gladieux et al., 2018). The rice-specific lineage is subdivided in three clonal (Gladieux et al., 2018) and one recombining and genetically more diverse lineage mainly distributed in Southeast Asia (Gladieux et al., 2018; Saleh, Milazzo, Adreit, Fournier, & Tharreay, 2014). Cross inoculation experiments with globally distributed isolates pathogenic on rice have revealed host specialization of P. oryzae to the main groups of modern rice varieties (Gallet et al., 2016; Gladieux et al., 2018). In the traditional YYT agrosystem, local adaptation to indica and japonica host ecotypes was also observed, and was associated with major differences in basal and effector-triggered immunity in the host (Liao et al., 2016). However, the coevolutionary interactions underlying the overall lower disease burden observed in YYT, remains unknown.
In this study we addressed whether the lower disease pressure observed on indica landraces, which represent 90 % of acreage in YYT, could result from the elevated landrace diversity extant in YYT, which hinders the emergence of P. oryzae populations specialized toindica landraces. We first analysed the population structure of YYT rice landraces on the one hand and P. oryzae populations on the other hand. We then used paired samples of P. oryzaepathogens and their plants of origin to address whether host and pathogen populations were genetically co-structured in order to establish if P. oryzae genotypes were specialized to their native host genotypes.