Population structure
The analysis of population structure was conducted using two different clustering approaches, model-based maximum likelihood hierarchical clustering via ADMIXTURE (Alexander and Lange 2011), individual based principal component analysis (PCA) on genetic distances. Cross validation as implemented in ADMIXTURE across 15 replicates identified K = 4 as the optimal number of genetic clusters in the data (Figure S1). Visualization of the coefficients of ancestry for each individual are arrayed by geographic location from southwest to northeast in Figure 4A. Figure 4B shows the admixture among populations as arranged by hosts. As with genetic distance estimates, ADMIXTURE reveals evidence of structuring among populations and hosts. In this case, patterns of structuring among hosts is more evident than strictly along a geographic gradient. For example, faba bean hosts have more similar patterns of shared ancestry than that shared between faba bean and chickpea or tomato, despite covering a broad range of geographical space. In contrast, populations AT and BM are spatially and genetically isolated. These data are consistent with the AMOVA results supporting a role for an interaction between host and location partitioning genetic variance among populations (Table 4).
The first 4 components from a PCA of genetic distance explained 66.16% of the total variation in the data. To further examine the relationship between samples within this PCA we conducted a Discriminant Analysis of Principle Components (DAPC). DAPC identified 4 genetic clusters in the data across 3 axes (Figure 5). Cluster assignments supported both the model-based and AMOVA identified genetic differentiation by host/population (Figure 6). Posterior assignment supports admixture within chickpea, tomato, and faba bean hosts but not within carrot or pea. Chickpea and tomato samples predominate cluster 1. Pea samples are all placed in cluster 4 while carrot samples are all placed in cluster 2. Faba bean samples have significant overlap between cluster 3 (N=35) and cluster 2 (N = 22) with two samples having shared membership.