Introduction
Alcoholic fermentation -the capacity of some yeasts to extract energy from single sugars, generating CO2 and ethanol as metabolic products even in the presence of oxygen- is an important physiological adaptation. The process allowed the utilization of the ecological niche given by modern fruits, an abundant source of food that emerged in the terrestrial environment in the Cretaceous (Dashko et al., 2014, Piskur et al., 2006). Although best known by their capacity to produce and metabolize ethanol (Piskur et al., 2006), the diversity of substrates metabolized by yeasts is enormous, as they exploit the varied habitats provided by the interphase between plants and animals (Paleo-Lopez et al., 2016, Kurtzman et al., 2011). This ecological success is represented by (at least) 1,500 species of known yeasts, which can be found on a broad range of substrates including the skins of fruits, cacti exudates, soils and animals, where they can be either commensal or pathogenic (James et al., 2006, Kurtzman et al., 2011). The fermentation lifestyle, however, has the special advantage of producing a toxic product (alcohol), which displaces other microorganisms and allows yeasts to dominate the environment. For this reason, it represents a key innovation that probably boosted the diversification of fermentative yeasts about 100 millions of years ago (MYA) (Dashko et al., 2014, Piskur et al., 2006). Thus, rapid sugar and nitrogen assimilation and subsequently efficient ethanol production, even in the presence of oxygen at the expense of ATP production, represents a key feature of fermentative yeasts (“Crabtree positive yeasts”, hereafter, (Gutierrez et al., 2016).
The domesticated Baker’s yeast (Saccharomyces cerevisiae ) with its large collection of genetic variants, is normally regarded as the most important yeast for fermentation (Piskur et al., 2006); but several other yeast species, such as wild yeasts from temperate rainforests (S. paradoxus at the Northern hemisphere; S. eubayanus at the South), can produce alcoholic products with considerable efficiency (Williams et al., 2015, Libkind et al., 2011). In fact, comparing ethanol yield (i.e., rate of ethanol production per gram of glucose consumed; a proxy of fermentative performance) among yeasts species does not always gives a clear pattern of superiority in competitive fitness for a given species, as fermentative performance is very variable and depends on a myriad of factors (Williams et al., 2015, Hagman & Piskur, 2015, Hagman et al., 2014, Hagman et al., 2013). Here, mapping trait values measured under homogeneous conditions on a calibrated phylogeny would reveal several interesting patterns of phenotypic variation, for instance, historical events (see below).
It has been proposed that the origin of the fermentative lifestyle in yeasts occurred in a few steps involving some genomic rearrangements that affected the yeast lineage since its origin, about 200 MYA, such as the loss of mitochondrial electron transport (respiratory complex I), the horizontal transfer of URA1 gene, and a whole genomic duplication (Hagman et al., 2013, Paleo-Lopez et al., 2016, Dashko et al., 2014)(see Fig 1a). The relative importance of these rearrangements on fermentative capacity of Crabtree positive yeasts has some debate. Some authors, based on phenotypic comparison of Crabtree positive and negative yeasts concluded that the onset to fermentative capacity in Crabtree positive yeasts was attained in these several steps (Hagman & Piskur, 2015, Hagman et al., 2014, Hagman et al., 2013). However other authors, based on genomic comparisons sustain that it was abrupt and marked only by the whole genomic duplication event that occurred about 100 millions of years ago (Marcet-Houben & Gabaldon, 2015, Wolfe & Shields, 1997).
In order to study the origin of fermentative capacity in a phylogenetic comparative analysis for yeasts, we took advantage of a phenotypic compilation where several proxies of fermentative performance were measured in cultures of several species, including Crabtree positive and negative ones (Hagman et al., 2013). Phylogenetic comparative analyses are useful statistical approaches for the analysis of phenotypic variation, since the phylogeny is used as a template for testing departures from the assumption of common descendance in lineages. Thus, conclusions should be taken exclusively for the phylogeny and the set of traits being measured. In this case, measurements were obtained under strict homogeneous conditions and after several generations. Then, phenotypic differences will only reflect lineage-level differentiation, the hallmark of “common-garden” experiments in ecology and evolution (Kawecki & Ebert, 2004, Linhart & Grant, 1996). We applied a particular comparative procedure to those data (the “lasso-OU” algorithm, see methods), which detects automatically adaptive shifts in phenotypic values, permitting a “blind” identification of evolutionary events that have disproportional influence on phenotypic variation. Specifically, we explored if multiple events or a single event explains the actual fermentative capacity of yeasts, after mapping these traits on the phylogeny. We considered four continuous traits representing performance (i.e., ethanol yield, EthY; Respiratory Quotient, RQ; glycerol production. Gly; respiratory quotient and growth rate). EthY is a measure of general fermentative performance as is quantified as the amount of ethanol produced per unit of glucose consumed, thus being central for characterizing fermentation efficiency (Hagman et al., 2013). RQ, on the other hand is important because fermentation does not need oxygen as the final electron acceptor, and produces just one CO2 in the first decarboxylation step. Then, ethanol-forming yeasts have RQ ratios significantly greater than one, while non-ethanol forming yeasts have an RQ close to, or equal to one (Hagman & Piskur, 2015). The justification of Gly relies on the fact that fermentative yeasts produce this metabolite as a response to hyperosmotic stress, in an alternative pathway of respiration (Aslankoohi et al., 2015) (see Table 1). If these variables are informative enough, then a comprehensive phylogenetic analysis should detect –above the level of reasonable statistical doubt- the positions where major phenotypic shifts occurred. As a null hypothesis we included dry mass growth rate, which represents an undifferentiated measure of growth performance in all lineages. Given that this variable is neutral for clade differentiation, phylogenetic signal should be non-significant and the lasso-OU algorithm should not detect any adaptive shift on it.