Phylogenetic tree
The present analysis was performed on the phylogeny published by Kurtzman and Robnett (2003). Although there has been new and more complete developments for Saccharomycotina (Shen et al., 2018, Salichos & Rokas, 2013), these phylogenies does not contain the number of measured species compiled here. Also, the phylogeny does not change much excepting for some tips involving the Saccharomyces genus (S. paradoxus and S. cerevisiae appear as sister species in Kurtzman & Robnett, 2003). In any event, we re-analyzed the data using Shen et al. (2018) phylogeny, for which we could compile a combination of 24 species and traits, and results were similar to what is reported here. Since the original Kurtzman and Robnett’s phylogeny was not available in digital format, we recompiled it using the descriptions of the original paper (Kurtzman & Robnett, 2003), and the instructions that the first author kindly provided. This phylogeny was obtained using four nuclear genes (large subunit rRNA, small subunit SSU, ITS-5.8S and translation elongation factor-1α) and two mitochondrial genes (mitochondrial SS rRNA and COXII) (Kurtzman & Robnett, 2003). We downloaded the sequences reported by the author to obtain the phylogenetic relatedness among species using the maximum likelihood (ML) function included in the software MEGA v6 (Tamura et al., 2013). ML was used with the defaults provided as well as with the gamma distribution, using similar set parameter Kurtzman and Robnett (2003) phylogeny. The bootstrap was performed with 1000 replicates. We employed same method reported for Kurtzman & Robnett (2003) to obtain same topology using concatenated genes. We time-calibrated the phylogeny using three different historical events: the loss of the respiratory complex I, which is dated to 150 millions of years ago (MYA; (Marcet-Houben et al., 2009), the horizontal transfer of the URA1 gene, which according to (Dujon, 2010) occurred 125 MYA and the WGD, which according to (Wolfe & Shields, 1997) occurred 100 MYA (Fig. 1A). The calibration was performed with the chronopl command in ape (Paradis, 2012).
Traits were compiled from the values published by Hagman et al. (2013) [the complete dataset is provided in Table A1], and pruned to make them suitable for the comparative analysis, respecting branch lengths and species representation in the phylogeny. This is pruning is a common practice in phylogenetic comparative methods whenever trait values are compiled from literature and not necessarily are represented in the phylogenetic tree. In our case, pruning unavoidably reduced the sample size from 50 original strains, to 31 species with trait values. According to Cressler et al. (2015, pp 959), OU methods have relatively good performance in model discrimination even with sample sizes as small as 20 species, as long as the sample is representative of the biological diversity of the group (Cressler et al., 2015). This is our case, as we used a wide variety of Crabtree positive and negative species (see Fig 1). When two trait values were available for a single species, we took trait averages. We averaged strain values for Candida glabrata(two strains), Eremothecium corylii (two strains), E. sinecaudum (two strains), Kazachstania lodderae (two strains),Kluyveromyces lactis (two strains), Kluyveromyces marxcianus (four strains), Lachancea kluyveri (two strains),Saccharomyces cerevisiae (two strains), Saccharomyces pastorianus (two strains), Tetrapisispora iriomotensis (two strains), Torulaspora franciscae (two strains).