Population structure and demographic history
To investigate the structure of the dataset, we used a model-based clustering approach as implemented in Admixture (Alexanderet al. 2009). For the Admixture analyses the number of populations was inferred by analyzing different number of populations (K) and the cross-validation (CV) error for each K. The CV error is used to find which K has the best predictive accuracy, but does not try to determine the absolute K. The full data set including all isolates and SNPs were analyzed, in addition to reduced data sets of only European and only Japanese isolates (SNP data set reduced by minor allele count ≥ 1 in VCFtools (Danecek et al. 2011)). All datasets were transformed from vcf format to plink format in VCFtools. Furthermore, the variation and genetic distance between and within populations were visualized by a PCA plot analyzed in Eigensoft (Price et al. 2006). The PCA analyses were run on the full SNP data set, and also on the split data sets (European and the Japanese isolates) as in the Admixture analyses. Three PC axes were produced for each of the three PCA analyses.
Coalescent simulations were used to infer the demographic history ofS. lacrymans in Europe and Japan using the model-based approach implemented in Fastsimcoal2 (Excoffier & Foll 2011; Excoffieret al. 2013). In Fastsimcoal2 the likelihood of predefined evolutionary models can be compared. In addition, demographic parameters, such as the effective population size, population growth rate, as well as timing of evolutionary events, can be estimated for the different evolutionary models. To test the divergence of S. lacrymans , we defined three realistic evolutionary models, supported by what is known from the literature. The first model represents a scenario, where S. lacrymans moved to an indoor environment in Japan, before migrating to the built environment in Europe. The second model represents a scenario where S. lacrymans has moved into the built environment, independently in Japan and Europe, from two natural populations that diverged prior to the colonization into houses. In addition to the divergence between Europe and Japan, the change in population size is important for understanding current patterns of genetic variation. The European population has been shown to be highly reduced in genetic diversity, likely resulting from a founder event when the population was established. To account for that, we included a population growth rate in Europe for both models. In the third model, we also implemented a growth rate for the Japanese population.
The likelihood of each model was inferred from the simulated site frequency spectrum (SFS) fitted to the observed minor allele frequency spectrum with the composite likelihood calculated in Fastsimcoal2. For each model 50 independent Fastsimcoal2 runs of 1000000 coalescence simulations and 40 cycles were analyzed. Confidence intervals for the point estimates were calculated using the parametric bootstrap approach used in Excoffieret al . (2013). We analyzed the data with both 10-7 and 10-8 as the mutation rate per site per year. The number of generations rather than years was calculated, as commonly used for such analyses. The generation time forS. lacrymans is probably highly context-dependent. For instance, under optimal growth conditions, the fungus can colonize, grow and expand extremely quickly and fruit after one year, and probably fruit successively for several years. Alternatively, under sub-optimal conditions the fruiting frequency will vary extensively. It is highly plausible that fruiting in the human-made habitat will lead to a reaction from the home owner and often the death of the fungus. Compared to other taxa, all somatic mutations in a fungal individual have a chance to contribute to the next generation, explaining the different scales of mutation rates across organisms. There are also few available estimates of mutation rates for wood decay basidiomycetes. Recently, a mutation rate of 10-10 was estimated for a single diploid individual of the fungal pathogen and wood decay fungusArmillaria gallica (Anderson et al. 2018). This relatively slow mutation rate is probably not representative of a sexually reproductive and flexible population. Regarding other fungal phyla, higher mutations rates have been estimated, e.g. 7.29 × 10-7 for the chytrid B. dendrobatidis (O’Hanlonet al. 2018), 2.4 × 10−6 to 2.6 × 10−6 for ascomycete yeast Saccharomyces cerevisiae (Gallone et al. 2016) and on average 1.98 × 10-8 in Magnaporthe oryzae (Ascomycota) using tip dating of temporally separated samples (Gladieux et al.2018). Thus, using both mutation rates of 10-7 and 10-8 in our demographic analyses allows us to explore the effect of mutation rates on the analyses.