Species delimitation analysis
We used the SNAPP software (Bryant et al., 2012) to test if species could be delimited at the molecular level using the Bayes Factors criteria implemented in BEAST v2.6 (Bouckaert et al., 2019). We evaluated four competing species models with the BFD* method (Leache et al., 2014), using as an outgroup the CBS2950 reference strain (group A, Table S2), and as the interest group we used the strains from B: Huilo-Huilo, C: Central Plateau, D: Coyhaique and E: Altos de Lircay. SNAPP analyses were only performed with a subset of individuals per genetic group for computational feasibility; the total number of SNPs in the dataset was 218,266. To change the VCF haploid to diploid we used the java code VcfFilterJdk (Lindenbaum and Redon, 2018). We used the Marginal likelihood estimation with 12 steps for the dataset to steppingstone sampling. We used default settings (e.g., Yule prior for species tree and branch length estimations; gamma distributions for ancestral theta), with five million burn-in iterations. To ensure convergence, we confirmed that our runs reached effective sample size values >500 after burn-in using the program Tracer v1.4 (Rambaut et al., 2018). Maximum credibility trees were generated using TreeAnnotator in BEAST v2.6, and the posterior distributions of species trees were visualized using DensiTree (Bouckaert, 2010).