Evolution and functional diversification of yeast sugar transporters

Sugar transport is often the parameter that limits yeast growth and fermentation [1–4]. Considerable research effort has been made in the past decades to study various aspects of sugar metabolism in yeast, showing that the uptake of sugar in the yeast cell plays a crucial role in the control of the metabolic flux, having influence on catabolic repression and the equilibrium between respiratory and fermentative metabolism. Biochemical characterization of sugar uptake in yeast cells started around six decades ago [5,6], followed by the first isolation 30 years ago of a hexose transporter gene [7], encoding the Saccharomyces cerevisiae galactose permease Gal2. After that, a plethora of monosaccharide transporters in the so-called Hxt class was identified and characterized in this yeast [8]. Nowadays, due to development of bioinformatic tools and sequencing methods, a multitude of putative sugar transporters and sugar sensors has been identified in yeast genomes, allowing deeper study of their evolutionary relationships and biochemistry [1,9–15].

Sugar transporters mainly belong to the major facilitator superfamily (MFS), which includes proteins from Bacteria, Archaea and Eukarya. These proteins are integral plasma-membrane polypeptides that exhibit high structural conservation, although they may share little sequence similarity. MFS sugar transporters consist of two sets of six hydrophobic transmembrane-spanning (TMS) α-helices connected by hydrophilic loop [16] (Figure 1). Most of these sugar transporters operate via facilitated diffusion, a mechanism that permits the movement of a solute along the concentration gradient across the membrane with no need of external energy [16]. Some, however, are able to transport the sugar against the concentration gradient by coupling the transport with the simultaneous movement of protons in the same direction, hence they are known as proton-symporters and are energy-consuming systems. Finally, even though they share a similar structure with other members of MSF, some members of the family (e.g. Snf3 or Rgt2 in S. cerevisiae) are unable to transport sugars, instead acting as sensors, signalling the presence of sugars at the membrane, influencing gene expression [17].

Before going in detail on each class of transporter, it is important to understand that sugar transporters are characterized by high plasticity and promiscuity. One sugar transporter can have a broad uptake spectrum of sugars, while some others evolved to be very specific for sugars present in higher concentrations in that yeast’s environment. For example, some glucose transporters belonging to the Hxt and the Hgt families can also mediate the import of xylose, arabinose and/or galactose with various affinities [19,20] or lactose and cellobiose can share a sugar transporter [11,21], or a glycerol transporter was found to also be a xylose transporter [9].

Although the evolution of MFS has been discussed previously, there has not been a lot of consideration of links between an organisms’ adaptation to specific natural environments and the evolution of transporters. Multiple studies showed that sugar transporters and sensors are present in different copy numbers amongst yeasts [9–11,22]. The gain, loss and diversification of transporters over the course of evolution is a strong indication of niche adaptation as it represents specialization to use specific sugars. Here, we took advantage of the availability of high-quality genomes to survey the diversity and evolution of sugar transporters in the budding yeast sub-phylum (Saccharomycotina). For this analysis, we selected nine species spanning the Saccharomycotina, as well as Schizosaccharomyces pombe (Sp), which represents the Taphrinomycotina, the basal sub-phylum in the Ascomycetes [18,23] (Figure 2).

Complete genome sequences were retrieved and explored leading to the identification of 139 candidate MFS-type putative sugar transporters. Multiple sequence alignments and phylogenetic analysis were performed to establish the detailed relationships between the MFS proteins within and between species [24–26] (Figure 3). This also enabled us group transporters based on sub-class or predicted substrate (Table 1). The extent to which these proteins have been previously studied varies substantially. In some cases, especially for the S. cerevisiae hexose transporters (HXT family), very detailed knowledge including kinetics of sugar transport is known, whereas for others, function has been inferred based on sequence similarity. In some cases, this was based on limited information and thus multiple transporters now have gene names that do not reflect their substrate range. Other putative transporters are identified here for the first time and do not have gene names, though now based on our more complete dataset, it is possible to infer potential substrates for some of them. For the purposes of our analysis, we labelled clades and sub-clades based on the known or presumed functions, notwithstanding the limitations already mentioned. From examination of the tree (and Table 1), it is clear that the pattern of evolution varies dramatically among the different yeasts. We explored this further, considering also the inferences that could be drawn of these evolutionary trajectories.

Glucose uptake occurs via the Hxt (hexose transporters) and Hgt (high-affinity-glucose transporters) families of transporters (Figure 3). Although named from original studies, as will be discussed later, these proteins have varying affinity and specificity for diverse sugars. Hxts are undoubtedly the best-studied family and many of these proteins have been extensively characterised in S. cerevisiae. It is a large family that has been subject to expansion in most species. This is particular evident in S. cerevisiae, which has 17 genes encoding Hxts. Reconstructing the evolutionary trajectory of HXTs within and between species is very instructive for understanding this class of transporters more generally. There are multiple examples of orthologous genes that arose via speciation, as well as paralogous genes that arose by gene duplication in particular lineages. Analysis is complicated by the whole genome duplication (WGD), a hybridisation between parents at the base of the ZT (Torulaspora) and KLE (Kluyveromyces) lineages that took place about 100 million years ago, giving rise to the lineage that includes S. cerevisiae, C. glabrata and other genera [27,28].

From a phylogenetic tree featuring all the Hxt proteins, it is clear that the HXT family originated from a common ancestral gene and, since the furthest related species in the Saccharomycotina, Y. lipolytica (YAL), has just one copy, this may represent the ancestral state (Figure 4). There are multiple HXTs in S. pombe (SCHPO), but these cluster together indicating expansion within that species. Moving from Y. lipolytica in the Saccharomycotina, the next two diverging species, S. stipitis (genes annotated as PICS) and D. hansenii (DH), show independent intraspecies expansion of the ancestral HXT (four and three copies, respectively). The situation becomes more complex when we move to the Saccharomycetaceae family, which includes E. gossypii (annotated as AG), S. cerevisiae (Sc), Kluyveromyces (Kl and KMAR), T. delbrueckii (TDEL) and C. glabrata (CAGL). E. gossypii shows expansion of a single HXT, resembling earlier diverging yeasts, but all other species carry expansions of genes that arose at various stages during speciation. Thus, there are multiple clades carrying orthologous sets of genes. From examination of the phylogenetic tree, and also taking the conserved gene synteny into account, we propose a particular pattern and sequence by which the HXT family evolved in the Saccharomycetaceae (Figure 5). The major underlying event was a duplication of the ancestral HXT to create a locus with two paralogues. Although there have been various expansions and losses, and duplication of some of the genes to other regions of the genome, the original HXT locus has largely remained intact and syntenic throughout the Saccharomycetaceae and thus is the key to understand the protein family. K. lactis and K. marxianus are the most straightforward species to analyse first (green labelling in Figure 4). K. lactis retains the two original genes, which are now named KHT1 and KHT2, and that have been studied as glucose transporters in that species. In the best-studied laboratory strains of K. lactis, there was a recombination between KHT1 and KHT2, giving rise to a chimeric gene RAG1, which is actually the best studied Kht variant [29]. In K. marxianus, both KHT1 and KHT2 have been subject to gene duplications such that there are now two copies of KHT1 and four copies of KHT2. Similar to the RAG1 example in K. lactis, recombination within the locus was reported, with some K. marxianus strains showing variable numbers of KHTs [11]. Although possibly a laboratory phenomenon in the strains described, these recombinations also illustrates one of the ways by which variation can arise at the locus. Moving to the Torulaspora (ZT) lineage, the current pattern of HXT genes in T. delbrueckii allows reconstruction of evolutionary events in this clade. Early in the ZT lineage, there was a duplication of the KHT1-like gene, such that there were now three HXTs in total. In T. delbrueckii, the two KHT1-like genes are now named TDELOE2280 and TDELOE2290. Later, during T. delbrueckii evolution, the KHT2-like gene also duplicated, with these genes named LGT1 and IGT1 [10]. Finally, the WGD lineage, that includes S. cerevisiae, C. glabrata and other ‘post-WGD’ species, is explored. As the WGD was a hybridization between strains from the KLE and the ZT lineages, the founder of this clade carried five HXT-family genes – two from the KLE parent and three from the ZT parent. In S. cerevisiae, one of the KHT1-like genes duplicated giving rise ultimately to HXT1 and HXT3. Similarly, a duplication of one of the KHT2-like genes gave rise to HXT6/7 and HXT4. The descriptor HXT6/7 is used to denote this ancestral gene as is clear that HXT6 and HXT7 arose from a much recent tandem duplication within S. cerevisiae. It is implicit in this phylogenetic reconstruction that ancestrally in the S. cerevisiae lineage there were four copies of a KHT1-like gene, though now, S. cerevisiae possesses three genes in this family. In addition, although HXT1 and HXT3 arise from a tandem duplication, they are now on different chromosomes. Both of these can be explained by recombination between homologous regions of the parental chromosomes, giving rise to loss of one allele and switching of HXT5 and HXT3. The consequence of all these events are the two extant S. cerevisiae loci: HXT7-HXT6-HXT3 and HXT4-HXT1-HXT5.

Analysis of S. cerevisiae HXTs has always been complicated because of the presence of additional copies at other loci that must have arisen from gene duplications. Some of these are specific to S. cerevisiae but others may be indicators of ancient duplications where the orthologous gene was lost from both the ZT and KLE lineages. An example of this would be the deep-branching the clade with HXT13, HXT15, HXT16 and HXT17. The clade with HXT2 and HXT10 (arising from a recent duplication within S. cerevisiae) is also interesting as it also has a T. delbrueckii orthologue (TDEL0A04250). The branching position, supported by good bootstraps, indicates that the ancestor of this gene even precedes the duplication of the “original” HXT that gave rise the KHT1-like and KHT2-like genes. It is therefore inferred that the orthologous gene was retained in the ZT lineage but lost in the KLE lineage. The distance between the two C. glabrata and the two S. cerevisiae (HXT2 and HXT10) representatives in this clade are consistent with independent patterns of duplication in these two post-WGD species. The additional S. cerevisiae genes within the KHT2-like clade have mixed ancestry. GAL2, encoding a galactose transporter, arose in the ZT lineage as a duplication of one of the KHT2-like ancestral genes and was provided to S. cerevisiae by that parent at the WGD event with HXT8 being a paralogue. HXT9, HXT11 and HXT12 are paralogues within S. cerevisiae and cluster strongly with Torulaspora IGT1 and LGT1. There was a single ancestral gene in the S. cerevisiae parent that, as well as giving rise to HXT6/7 also duplicated to a distal locus giving the HXT9/11/12 ancestor. Note that it is the conserved syntenic arrangement that allows the assignation of particular S. cerevisiae HXTs to the ancestral locus. C. glabrata was briefly mentioned above and follows a similar pattern of expansion via tandem duplication. HXTs are ancestral in the ascomycetes and the paradigm of expansion of a single ancestral copy via sequential tandem duplications is also seen in S. pombe. In this case, it is known that certain members of the Hxts work not only as hexose transporters, but they are involved in the transport of other substrates or in different cellular function. For instance, in S. pombe, Ght4, one of the eight proteins (Ght1–8) homologous to the Hxts, is responsible for the uptake of gluconate instead of glucose [30].

Since, in all species studied, all the genes at ancestral locus encode glucose transporters it is therefore reasonably to conclude that glucose transport is the original function of the Hxt proteins. The increase in HXT copy number in S. cerevisiae, was a consequence of selective pressures during its evolution, and must reflect the habitats that this species occupied. The study of HXTs and their evolution serves as paradigm for analogous processes in other yeasts that would have occupied different habitats. In this vein, some authors previously speculated that glucose limitation triggered the HXT6 and HXT7 duplication [31,32]. In its natural environment, S. cerevisiae can experience a broad range of sugar concentrations, which is likely to have led to the development a complex and highly regulated system of hexose transporters with specific features that allows the cell to optimize the uptake and metabolism of the sugar [33]. Seven of the seventeen S. cerevisiae HXTs encode glucose transporters with a range of affinities: Hxt1 and Hxt 3, low affinity; Hxt 2, Hxt 4 and Hxt 5, medium affinity; and Hxt 6 and Hxt 7, high affinity. Their expression is strictly regulated by the concentration of glucose in the media, except for the expression of HXT5, which is regulated by growth rate rather than extracellular glucose [33]. The function and expression of Hxt8 to Hxt17 is not fully-defined, but it is known that Hxt9 and Hxt11 might be related to the subfamily of ATP-binding cassette transporters involved in the resistance of antifungals and xenobiotics, and Hxt 13-15-16-17 are polyol transporters [34]. In K. lactis, a low-affinity glucose uptake system (Hxt) is encoded by two tandemly-arrayed genes, known as KHT1 and KHT2, that in some strains recombined giving origin to a single gene RAG1 [29,35]. In K. marxianus, we find an expansion of KHTs, with a number between 5 and 6 which varies between strains. In this yeast, KHTs encode for glucose transporters, but they are responsible as well for the non-specific uptake of pentoses and galactose [9,11]. In T. delbrueckii, two Hxts, Lgt1 and Igt1, were characterized as glucose, fructose and mannose transporters, and Igt1 displayed biphasic glucose uptake kinetics being able to modulate its affinity (intermediate or high) by glucose concentration in the medium [10,36].

While the only known glucose transporters in S. cerevisiae are encoded by HXTs, some other yeasts have an additional family of transporters called Hgt, which were also first described as glucose transporters. Indeed, the archetype in K. lactis, Hgt1, received its name for high affinity glucose transport [35]. In reality, the specificity of these transporters is broader than originally thought and includes at least glucose, galactose, xylose and arabinose. In K. lactis and T. delbrueckii, respectively, there are one [35] and two Hgt-like transporters, but, in K. marxianus, five tandem copies of HGT1 are located on chromosome 1 and a further single homologue on chromosome 5 (Figure 3). As for the Khts, K. marxianus Hgts are also reported to have several functions and a broad specificity for both hexose and pentose sugars deriving from plant sources, which represent the primary environment of isolation of this yeast [9,11]. In the halotolerant yeast, D. hansenii, we also detected a high number of sugar transporters (four copies of HGT and three copies of KHTs) with a functional overlap between the Hgt and the Kht, that have various affinity both for hexoses and pentoses. In Y. lipolytica, three transporters cluster together with Hgt1 of K. lactis and they are reported to promote basic cell growth through the transport of several different hexoses at varying concentrations [37].

Whereas the Kht/Hxt family transport along a concentration gradient via facilitated diffusion, Hgt transporters are proton symporters [20,38]. In D. hansenii, although both Kht-like and Hgt-like transporters are present, it is likely that the sugar-associated proton uptake exclusively in the presence of salts such as NaCl or KCl that has been reported [39], is performed by symporters of the Hgt family. As described later, transporters in the lactose, glycerol, fructose and maltose transporter clades are also sugar/proton symporter. This wide distribution amongst yeast MFS transporters (Figure 3), is consistent with the idea that ancestral protein in this entire MFS family operated by coupling sugar uptake with proton translocation in the same direction across the membrane. The proton motive force that drives this kind of transporters is maintained by export of protons by the H⁺-ATPases therefore this transport system requires energy for its activity. We postulate that Hxts might have lost the proton symport activity or evolved to not be dependent on the proton gradient. This could have been especially useful to permit aerobic fermentation, which is energetically less efficient than respiration. It can be speculated that the loss of Hgts from S. cerevisiae, might be linked to the Crabtree-positive metabolism of this yeast. Indeed, the same argument was previously made for loss of the Fsy1 fructose symporter (see below) [40].

Seven of the S. cerevisiae Hxt transporters (Hxt1–7) are reported to be able to mediate fructose uptake via facilitated diffusion but S. cerevisiae in general lacks an active system for fructose uptake [41]. Interestingly, however, the wine strain EC1118 does have a specific fructose-proton symporter named Sc_Fsy1 (marked with an asterisk in Figure 3). Furthermore, S. uvarum, S. eubayanus and S. pastorianus also carry this gene, though it is not widely observed in other species of Saccharomyces. Zygosaccharomyces and Torulaspora also carry FSY1, while Lachancea and Kluyveromyces have an orthologous gene named FRT1. A previous comprehensive analysis of FSY1/FRT1 found that it was distributed in very unusual patterns amongst Ascomycetes and suggested that (1) the gene originated in the Pezizomycotina and (2) was independently lost and reacquired by horizontal transfer multiple times in yeast and fungal evolution [40]. Indeed, when the Saccharomycetaceae are examined, it is evident that this gene has an interesting, if not yet fully resolved, evolutionary history. To try and bring some clarity, the phylogenetic (Figure 6) and syntenic arrangement (Figure 7) of the gene is shown. The phylogenetic tree does not show the expected relationships since it is evident that FRT1 from Kluyveromyces and FSY1, found in Torulaspora and Zygosaccharomyces, share a history consistent with a past transfer of the gene from Kluyveromyces to the ZT lineage. In Kluyveromyces and Lachancea kluyveri, FRT1 is located in the same locus in a non-telomeric region of the chromosome between INO80 and ARO2, though the FRT1 orientation is different (Figure 7B). Inspection of the yeast gene order browser [42] suggests that this is the gene order found in the yeast ancestor, therefore, it seems likely that there was a transfer of FRT1 from Kluyveromyces to an ancestor in the ZT lineage, as suggested by Coehlo et al. (2013). Interestingly, in L. thermotolerans there has been a reciprocal translocation at this location, though FRT1 was retained next to INO80 (not shown). In Torulaspora and Zygosaccharomyces, FSY1 is located on chr3/C adjacent to GDH3 in a sub-telomeric region (Figure 7A). Although the 3’ flanking gene(s) varies, the conserved juxtaposition to GDH3 is consistent with this gene being present here prior to the divergence of these genera. The corresponding region in Saccharomyces does not encode FSY1 and the gene is inly present in a minority of Saccharomyces species and strains. For the strains of Saccharomyces that do carry FSY1, two distinct locations are seen (Figure 7C and D). In the S. uvarum lineage where its carriage seems universal (from data on S. uvarum, S. eubayanus and S. pastorianus), FSY1 is located in a sub-telomeric region on chrIV, whereas in the wine strain S. cerevisiae EC118, this region has sub-telomeric genes and FSY1 is located on a part of chrXV that is termed ‘region C’ and was suggested to have been acquired horizontally in the wine environment from Torulaspora microellipsoides [43–45]. Apart from Kluyveromyces and L. kluyveri, which may represent the ancestral state, FSY1/FRT1 is always located in unstable regions of the genome, which contributes to its propensity to be lost, and which is not untypical for sugar transporter genes. As referred to above, FSY1/FRT1 is found in many species of yeast and fungi, including S. stipitis and D. hansenii (Figure 6). The sporadic distribution suggests that this gene is only retained where it is needed, but the multiple examples of acquisition indicates that, when needed, the gene can be reacquired from an environment gene pool. This is quite interesting in the context on considering the community metagenome as a repository for individual lineages to source genes as required. It should be remembered that since Hxts are able to transport fructose, most if not all yeasts can use this sugar and the key difference is whether is transported by passive diffusion or active transport. It is important to record that fructophilic yeasts such as Z. bailii, possess a fructose transporter, FfZ1, of the DHA rather than the MFS family. FfZ1 is a facilitator, related to drug antiporters, that was originally acquired from bacteria and it is thought to confer some yeasts with this preference for fructose over glucose [46,47]. As with the choice of using a passive or active transport system for fructose, the presence of a transporter that enables preference of fructose over glucose must be a reflection of the habitat and niche of these yeasts.

S. cerevisiae lacks the enzymes to naturally assimilate pentoses, even so, it can still uptake these sugar non-specifically via Hxts and Gal2 [48,49]. The latter can transport D-xylose with low-affinity, but it is competitively inhibited by D-glucose, meaning that xylose starts to be transported and metabolized only when D-glucose is depleted from the media. For biotechnological applications, various approaches have been tested to provide S. cerevisiae with a specific pentose transporter, in particular engineering the binding site of Gal2 to increase the affinity for xylose [8,50]. K. marxianus is a native pentose utilizer and several transporters have been reported to be involved in xylose and arabinose uptake with different affinities. It is concluded that arabinose is exclusively transported into the cell by some members of the Hgt family as a mutant strain lacking the full HGT cluster fails to grown on arabinose [9]. Xylose is taken up with high affinity by the Hgts, KMAR_10529/KmAXT1 and KMAR_10531, which are also capable of glucose and galactose uptake respectively, and with low affinity by other Hgts and some Khts [9,20]. Interestingly, it was found that KMAR_60179, which is a glycerol transporter (see below), can also import xylose with low-affinity [9]. S. stipitis is an excellent xylose-fermenting yeast and is characterized by a plethora of transporters able to import both xylose or arabinose [51,52]. These transporters are found in two clusters, one deep-branching in the HXT clade (annotated as SUT) and a second, a sister clade of the glycerol transporters (annotated as XUT) (Figure 3). D. hansenii also has proteins in both of these clades, with at least one of those in the XUT clade, DEHA2C11374p, shown to be able to uptake both xylose and glucose [53]. In the case of both S. stipitis and D. hansenii, duplications are seen and the general pattern matches those described earlier and provides indications of duplications at different stages of yeast evolution.

Glycerol uptake occurs in S. cerevisiae via the Stl1 glycerol/H+ symporter in response to high osmolarity [54]. T. delbrueckii, A. gossypii, S. stipitis, K. lactis and K. marxianus are also predicted to encode an orthologous glycerol transporter (Figure 3). In D. hansenii, however, there is a remarkable expansion of glycerol/H+ symporters (9 putative members) that is probably related to the reported importance of multiple transporters for the ability of D. hansenii to maintain osmotic balance under salt stress conditions [55,56]. Some of these transporters are symporters of glycerol and sodium which work in a tight equilibrium with sodium-proton antiporters and sodium-potassium pumps. Amplification of the glycerol cluster of transporters in Y. lipolytica is also observed, and this is most likely connected to the capacity of this yeast to use lipids as triglycerides as carbon source [57]. Although only one S. stipitis Stl1 is reported, other two S. stipitis proteins, annotated as xylose-transporters (Xut2 and Xut5), cluster within the ‘glycerol’ clade, perhaps indicating that these are native glycerol transporters as well. Curiously, Km_60179, which is equivalent to Sc_Stl1 is also capable of low-affinity xylose uptake, which is a feature not reported for any other glycerol transporter [9]. Within the larger xylose-glycerol cluster, there is a clade with proteins from several yeasts that do not have predicted substrates. An overall assessment would be that there seems to be some degree of relationship between the capacity to transport glycerol and xylose, but detailed studies are required to explore this more. This will determine the substrate range of each protein, and also what substrates it actually transports when growing in its own environment under physiological conditions.

Yeasts have variable and limited ability to assimilate disaccharides, with known disaccharide transporters falling into two clades (Figure 3). Most of the genes in the disaccharide clade have not been functionally characterised and so annotations are based on similarity to those that have been. Maltose permeases are a sister clade of the Hxt transporters and are found in S. cerevisiae, T. delbrueckii, S. stipitis, D. hansenii and K. lactis [58–60]. As mentioned earlier, the observation that these are proton symporters [61] is consistent with the idea that the shared ancestor with the Hxts was also a symporter. MAL1-type genes, encoding the permease, are typically divergently transcribed from a gene encoding an α-glucosidase (MAL2), which cleaves maltose in two molecules of glucose. In S. cerevisiae, where maltose utilisation has been best-studied, a transcriptional regulator (MAL3) is also part of the cluster [62,63]. The MAL cluster is often duplicated one or more times in yeasts, and brewing strains of S. cerevisiae are known to have multiple duplications, generally in sub-telomeric regions [32,64]. As well as multiple more canonical proteins, both S. stipitis and D. hansenii have a divergent protein in this clade and, in the case of S. stipitis, the partner of Ps_Mal5 is a divergent α-glucosidase, probably indicating an alternative substrate for these proteins [65]. As with most of the other transporters studied, the presence of representatives in different branches of the phylogenetic tree is consistent with the gene being ancestral and lost multiple times. Not all the proteins encoded by the MAL cluster have been functionally characterised and it remains to be determined whether the substrate of this ancestral transporter was maltose or a different sugar. The second clade encoding disaccharide transporters shares a common ancestor with the HGT genes (Figure 3). This group comprises three sub-clades, one of which is defined as cellobiose transporters, one that includes various proteins annotated as lactose transporters, and a third with proteins with unknown functions but that is a sister sub-clade of the cellobiose one. Taking the cluster as a whole, the large number of representatives from S. stipitis (10) and D. hansenii (9) is striking. In K. lactis, the cellobiose transporter CEL1 is divergently transcribed from a b-glucosidase (CEL2) that hydrolyses cellobiose in an arrangement that is highly reminiscent of the MAL genes described above. Also similar to MAL1-MAL2, this gene pair has been duplicated at the ends of several (3) chromosomes [66]. A previous study showed that the CEL1-CEL2 gene pair assembled during the evolution of the Kluyveromyces genus, and that CEL1 was lost from K. marxianus [21]. Functional data on the putative lactose transporters also come from studies in K. lactis and K. marxianus. LAC12 encodes the permease and is part of a divergently transcribed gene pair with LAC4, encoding a β-galactosidase [67,68]. Early studies on these genes focused on K. lactis but recently it was shown that K. lactis LAC12-LAC were acquired from K. marxianus via a relatively recent introgression [21]. In fact, from a population-level analysis of K. lactis genomes, it has been suggested that this introgression may have happened several times in anthropogenic environments [69] Furthermore, studies in K. marxianus, where there are four copies of LAC12 indicate that the capacity to transport lactose arose in a domesticated lineage of K. marxianus and the ancestral transporter probably had an alternative substrate [21,70,71]. With this in mind, it should be considered that the S. stipitis and D. hansenii transporters in this sub-clade, while carrying ‘LAC12’ annotations, probably do not have lactose as a primary substrate, as most of the species of these genera are reported to be deficient in growing on lactose [72]. Whether considering the maltose cluster genes or the cellobiose/lactose cluster, it is remarkable to note the shared tendency to put the disaccharide transporters in close proximity to hydrolases (Figure 8). The observation that most assemblages arose within particular lineages, rather than preceding speciation, is indicative of an underlying evolutionary pressure to cluster these functionally related genes. This pressure could have been to allow the transporter and hydrolase co-evolve together for new substrates, to avoid loss of one of a pair of co-operating genes, or later to facilitate co-regulation from a single promoter.

It has been known for many years that the S. cerevisiae proteins Snf3 and Rgt2 are glucose sensors, with Snf3 responding to low glucose concentrations and Rgt2 to high, to regulate expression of various HXTs [73]. Snf3 and Rgt2 belong to the larger Hxt family but have lost the ability to transport glucose. SNF3 and RGT2 arose from the WGD, therefore post-WGD yeasts typically have both sensors, whereas pre-WGD usually have one (Figure 3). The K. lactis sensor Rag4 appears to have the dual function of signalling both high and low concentrations of glucose and oxygen sensing in this yeast [74,75]. The general view would be that the WGD enabled S. cerevisiae separate functions, thereby acquiring greater regulatory control of HXT expression. Perhaps this was also a factor that allowed the loss of the HGT family of genes. The importance of glucose sensing is illustrated by the fact that, with the exception of Y. lipolytica, all the pre- and post-WGD yeasts in this study retain a sensor. It is notable that Y. lipolytica only has a single HXT representative and the logical conclusion is that these two things are linked. Most of the pre- and none of the post-WGD genomes we examined have a gene in a sister clade to the sensors (Figure 3), and in D. hansenii and Y. lipolytica, that gene is duplicated. Although these genes are sometimes annotated as encoding sensors, for example PsRGT2 in S. stipitis, functional data from studies of the Y. lipolytica Yht1 (YALI0C06424) and Yht2 (YALI0C08943) proteins indicate that these are hexose transporters, not sensors [37]. In particular, Yht1 is required for high affinity fructose transport. Thus, this clade are likely to encode sugar transporters as opposed to sensors. The overall pattern is consistent with the belief that sensors evolved from transport proteins. It is important to note, however, that more functional analysis is required to explore this issue. For example, if this hypothesis is correct, proteins may exist that have both sensor and transport activity, or even, the capacity act as a sensor of one sugar and a transporter of another.

The MFS of transporters in budding yeasts are remarkable for the diversity that has been achieved over the course of evolution. Some of these proteins are involved in intracellular transport, for example to the vacuole or mitochondrion, but the majority play a role in importing carbon sources from outside the cell. Without doubt, the habitat and niche that yeasts found themselves in have had a profound influence on evolution of the family. At a global level, this is most evident when looking at the Kht/Hxt family. Cases where this family has been retained and expanded are surely an indication of evolution in a glucose-centric environment. Furthermore, while it would be simplistic to suggest that this is the only factor, it does seem reasonable to propose that the capacity for rapid assimilation of glucose without expending energy for transport is linked to the ability of yeasts to grow by fermentative metabolism, and even to the Crabtree effect of aerobic fermentation. Conversely, it is also clear that this feature is less important for species like Y. lipolytica and D. hansenii, which have instead expanded families of transporters that are relevant for their lifestyle. One of the striking aspects of our survey was the finding that there are many putative transporters that cluster within particular clades, and perhaps have particular annotations, but for which we do not know their true substrates or their intrinsic range. It was also notable that some clades are defined by substrates that we are familiar with from our own activities, for example brewing or dairy fermentation, and these may not be representative of the clade. There are also quite a few clades that have highly conserved proteins where we have no idea what the substrate is.

There is a lot of interest in the biotechnology community in identifying new transporters for particular substrates and, while obvious perhaps, it is evident from this study that the best place to prospect for these transporters is in species that already use that substrate. It is also apparent that yeasts that are highly adapted for using particular sugars, or for operating in an environment where the sugar concentration or type available varies, specialised for this by having a range of functionally similar transporters. Yeasts like S. cerevisiae (glucose) or Y. lipolytica (glycerol) are examples of highly specialized yeasts, whereas a yeast like K. marxianus appears extremely versatile. When engineering yeasts for cell factory applications on sustainable substrates, one of the decisions is whether to introduce transporters into a well-characterised tractable host, or to engineer the novel pathway into a well-adapted host. Looking at the intricate systems that nature uses optimising yeast growth on particular substrates provides a strong argument for very careful consideration of the substrate–yeast combination when designing cell factories for bioprocess development. It is also important to consider that biotechnological substrates often comprise mixtures of sugars and so the interplay between different transporters is critical. Finally, the sensing dimension has received relatively little attention outside of glucose sensors in S. cerevisiae. A deeper understanding of how other sugars are sensed, and how this affects expression of transporter genes, is certainly required to enable the design of well-tuned cell factories.

• The ability to transport simple sugars across the cell membrane is central to the capacity of yeast to grow in natural and biotechnological environments

• Multiple transporter sub-families within the major facilitator superfamily have evolved in yeast by diversification and gene duplication

• The pattern of gene expansion and gene loss, as well as gain by HGT, is a dynamic process that reflects the particular habitat and specialization of different yeasts

• Saccharomyces cerevisiae has lost entire families of active transporters for are responsible for transport of important sugars in other yeast lineages

• The intrinsic plasticity of MFS transporters, as well their diversity and breadth across the budding yeast sub-phylum, signposts future potential for biotechnological research and exploitation.

The authors declare that there are no competing interests associated with the manuscript.

This work was supported by the YEASTDOC project which received funding from the European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skłodowska-Curie [grant number 764927].

We thank Carlos Belloch Molina for reading the manuscript and the useful discussion.

HGT

high-affinity glucose transporters

MFS

major facilitator superfamily

TMS

transmembrane-spanning