We have traced the evolution patterns of 2480 transmembrane transporters from five complete genome sequences spanning the entire Hemiascomycete phylum: Saccharomyces cerevisiae, Candida glabrata, Kluyveromyces lactis, Debaryomyces hansenii, and Yarrowia lipolytica. The use of nonambiguous functional and phylogenetic criteria derived from the TCDB classification system has allowed the identification within the Hemiascomycete phylum of 97 small phylogenetic transporter subfamilies comprising a total of 355 transporters submitted to four distinct evolution patterns named “ubiquitous,” “species specific,” “phylum gains and losses,” or “homoplasic.” This analysis identifies the transporters that contribute to the emergence of species during the evolution of the Hemiascomycete phylum and may aid in establishing novel phylogenetic criteria for species classification.

THE Hemiascomycete yeasts are believed to have diverged from a common ancestral fungus at least 400 million years ago. This phylum comprises >1200 known species (Kurtzman and Fell 1998; Boekhout 2005). The first Hemiascomycete genome sequence, that of Saccharomyces cerevisiae, has unraveled the existence of >30% of duplicated genes. This observation led to the hypothesis of a recent whole-genome duplication of the Saccharomyces genome (Philippsen et al. 1997; Wolfe and Shields 1997). A massive exploration of partially sequenced genomes from 13 Hemiascomycete species (Souciet et al. 2000) as well as the sequence comparison of a selected set of 40 duplicated gene sequences from six different species further investigated the origin of duplicated genes within the Hemiascomycete phylum (Langkjaer et al. 2003). Analysis of several nearly complete genome sequences from Hemiascomycete species closely related to S. cerevisiae has dated the whole-genome duplication after the emergence of Kluyveromyces waltii (Kellis et al. 2004), Ashbya gossypii (Dietrich et al. 2004), and K. lactis (Dujon et al. 2004) but before that of all sensus stricto and sensus lato Saccharomyces species (Cliften et al. 2003; Kellis et al. 2003) and Candida glabrata (Dujon et al. 2004). These data established the high frequency of gene loss among the duplicated genes. The existence in fungi of numerous gene duplication events followed by differential evolutionary drift of one of the two gene copies is believed to be a major force for adaptation to novel ecological niches and further speciation (Ohno 1970; Kellis et al. 2003; Langkjaer et al. 2003; Dujon et al. 2004).

The complete genome sequence of four species widely spread over the Hemiascomycete phylum became recently available: C. glabrata, the second most prominent causative agent of human fungal infection; K. lactis, a milk-loving yeast, believed to have diverged from the Saccharomyces clade at least 150 millions years ago; Debaromyces hansenii, a halotolerant species contaminating many dairy products; and Yarrowia lipolytica, a distantly related yeast that shares a number of properties with filamentous fungi (Dujon et al. 2004). In the context of this Genolevures project, a uniform nomenclature was designed to facilitate comparisons between species (Durrens and Sherman 2005). These data established that in addition to a recent whole-genome duplication, variable levels of segmental duplication, tandem repeats, and other duplication events of still unknown mechanisms have occurred during evolution of the Hemiascomycete phylum. The analyses of more ancient genomes such as those from the major human Hemiascomycete pathogen C. albicans (Jones et al. 2004) and from the Euascomycete filamentous fungus Neurospora crassa (Galagan et al. 2003) are fully consistent with this view.

Most of the gene products and gene families analyzed so far concerned nonmembrane proteins or RNAs. We wish to focus our analysis here on the evolutionary fate of transmembrane transporter proteins that correspond to ∼10% of the coding genes in the Hemiascomycete phylum. In fact, the transport in and out of the cell or organelles of precursors and end products is expected to be an efficient checkpoint for emergence of complex metabolic pathways. An example of coevolution of a transporter and a related metabolic pathway is the loss or inactivation of the galactose transporter Gal2p gene and six other genes of the galactose catabolism that has been shown in several Hemiascomycete species (Hittinger et al. 2004).

We have recently published a phylogenetic inventory of 402 established transporter proteins from S. cerevisiae (De Hertogh et al. 2002). Each of these proteins has been classified according to the transport classification (TC) system (Saier 2000). This classification allocates five digits to each phylogenetic cluster of transporters. The two first digits (“class” and “subclass”) identify the global transport mechanism. The third digit characterizes phylogenetic “families” or “superfamilies.” The fourth digit identifies phylogenetic “subfamilies.” The fifth digit (“clusters”) corresponds to the transported substrate or range of substrates, as presumed by experimental data or stringent sequence identity (Saier 2000). When the TC system did not provide the complete five-digit identification of S. cerevisiae transporters, we introduced X, Y, and Z (with numerical indexes such as X1, X2, X3, …) as the provisional last three digits, using only sequence homology criteria (De Hertogh et al. 2002). Moreover, we compiled a full inventory in S. cerevisiae of the class 9.B.X.Y.Z of “possible (or putative) transporters,” which comprises membrane proteins of unknown function. This system termed yeast transporter information (YETI) provides a five-digit identifier to each S. cerevisiae transporter with more than two predicted transmembrane spans (De Hertogh et al. 2002).

As an example of the usefulness of our S. cerevisiae YETI database, its comparison with the 50,000 random sequence tags provided by the partial sequence of 13 Hemiascomycete species (Tekaia et al. 2000) has unraveled 55 novel transporters, present in Hemiascomycete species but absent in S. cerevisiae (De Hertogh et al. 2003).

We now have extended our YETI database to a total of 2480 transporters encoded by the genomes of five species covering the entire Hemiascomycete phylum, namely S. cerevisiae, C. glabrata, K. lactis, D. hansenii, and Y. lipolytica. The N. crassa species belonging to the Euascomycete phylum served as reference outgroup. We have classified the Hemiascomycete transporters into hierarchized phylogenetic families, subfamilies, and clusters and inferred their evolutionary history by tracing their expansion or contraction in each species.

Comparative analyses of these sequences have allowed us to identify the emergence and loss of “species-specific” transporter subfamilies within the Hemiascomycete phylum and to distinguish them from the “ubiquitous” subfamilies of transporters that are conserved throughout the entire phylum. Moreover the “phylum” transporter subfamilies that are “gained or lost” definitively during evolution have been identified as well as the “homoplasic” subfamilies of transporters that are conserved transiently in the different Hemiascomycete species.


A total of 24,165 proteins have been identified in S. cerevisiae, C. glabrata, K. lactis, D. hansenii, and Y. lipolytica genome sequences (Dujon et al. 2004).

A semiautomatic classification pipeline was developed in Perl programming language, with a MySQL database server for data storage and scripts to submit and retrieve data from remote prediction servers HMMTOP (Tusnady and Simon 2001) and TMHMM (Krogh et al. 2001). The Bioperl Blast wrapper was used to handle blast queries and parse blast reports (http://gemo28.gene.ucl.ac.be/bioflow/). This pipeline, illustrated by Figure 1, comprises three components: selection of putative transporters, exclusion of false positive proteins, and final annotation.

Figure 1.—

Scheme for semiautomatic annotation.

The selection of putative transporters comprises four steps:

  • Step 1: Selection of all proteins with at least two predicted transmembrane spans according to TMHMM or HMMTOP within the 24,165 Hemiascomycete proteins.

  • Step 2: Selection by BLAST of each of the 24,165 Hemiascomycete proteins against the Transporter Classification Database (TCDB) (version 09, March 2004) (Saier 2000; Busch and Saier 2002) that comprises ∼1300 examples of transporter proteins from all species. Proteins with an E-value <10−19 are retained.

  • Step 3: Selection by BLAST of each of the 24,165 Hemiascomycete proteins against a raw version of YETI that comprises the 1068 proteins from S. cerevisiae predicted to have at least two transmembrane spans according to either TMHMM or HMMTOP software. Proteins with an E-value <10−19 are retained.

  • Step 4: Combination of proteins selected by steps 1, 2, and 3 in a nonredundant database of putative transporters.

The exclusion of false positive proteins in the putative transporter database comprises two extra steps:

  • Step 5: Exclusion of soluble proteins that BLAST with an E-value <10−19 against a database that comprises soluble proteins such as some transcription factors and other proteins known to exhibit falsely predicted transmembrane spans in S. cerevisiae.

  • Step 6: Exclusion of membrane proteins that BLAST with an E-value <10−19 against another database comprising membrane proteins from S. cerevisiae known to have nontransport activity.

The annotation of the transporters database comprises five steps:

  • Step 7: Allocation of a third digit (family) to each ORF according to the TC of the best BLAST hit in YETI and TCDB, respectively.

  • Step 8: Allocation of a fourth digit (subfamily) to each ORF according to the TC of the best BLAST hit in YETI and TCDB, respectively.

  • Step 9: Allocation of a fifth digit (cluster) to each ORF according to the TC of the best BLAST hit in YETI and TCDB, respectively.

  • Step 10: Manual allocation of digits X,Y, and Z to the proteins from step 4 that exhibit different BLAST values in YETI and TCDB, using the threshold E-value >10−35 for X, between 1E−35 and 1E−65 for Y, and <1E−65 for Z.

  • Step 11: Exclusion or manual allocation of digits to proteins from step 4 that have not been annotated by previous steps.


Global analysis:

In this study, membrane proteins are defined as proteins comprising at least two α-helical transmembrane spans predicted either by HMMTOP or by TMHMM. Because of their high number of false positive predictions, the proteins predicted to contain only one transmembrane span were not included. Predicted membrane proteins constitute on average 17.9% (data not shown) of all proteins encoded in the five Hemiascomycete genomes. Nearly half of the predicted membrane proteins are “established” or “possible” transporters (Table 1). Their frequency ranges from 7.5% in C. albicans to 9.0% in Y. lipolytica. Among the Hemiascomycete 2480 transporters, the secondary porters are by far the most prevalent, followed by transport ATPases (Table 2). The 2480 transporters were classified into 425 phylogenetic clusters characterized by five digits, which were derived from 204 subfamilies (four digits) and 82 families (three digits). The YETI classification of these 2480 transporters is given in supplemental data (supplemental Table D1 at http://www.genetics.org/supplemental/) and includes the 107 9.B.X proteins that are homologous to the S. cerevisiae membrane proteins of which the function is still unknown and that are considered as possible transporters. However the 1296 membrane proteins of unknown function from the non-S. cerevisiae species that have no homolog in S. cerevisiae were not included. They contain a few interesting novel putative transporters but also numerous nontransport proteins that will be analyzed in further work.

View this table:

Global statistics of membrane proteins in Hemiascomycete species

View this table:

Functional distribution of the “established and putative” transporters in the Hemiascomycete phylum

Transporter superfamilies or families (as determined by three digits) are of low sequence similarity to each other. The gain or loss of a complete superfamily or family is not frequent during evolution of the Hemiascomycete phylum. On the other hand, the fifth digit distinguishes clusters that belong to the same phylogenetic subfamily and are likely to transport substrates of similar nature. Obviously, the evolution process involved in the extension or loss of members of phylogenetic clusters is more limited and of less physiological significance than that required to gain or lose subfamilies. Taking into account also that we are dealing with contemporary species that may have continued to evolve after speciation we focus primarily on the evolution of the subfamilies (as determined by four digits—see supplemental Table D2 at http://www.genetics.org/supplemental/) and mention only a few striking examples of cluster variation that we believe to be of particular physiological significance.

Variation of the size of phylogenetic subfamilies:

To compare the relative variations of the size of the small as well as of the large subfamilies, their statistical variance was calculated. The variance is the sum of the squares of the differences between the number of transporters in each species and the average number of transporters across species. This sum is divided by 5 (the number of species). The variance value increases with the average value and takes into account both the size of the cluster and its variability. Table 3 lists the number of transporters of the subfamilies that have the highest variance across the five studied species. These subfamilies are those that have drifted most obviously during the evolution of the Hemiascomycete phylum.

View this table:

The most variable subfamilies based on the statistical variance of the number of ORF, by subfamily

The anion:cation symporter (ACS-2.A.1.14) subfamily takes up anionic vitamins (allantoate, nicotinate, panthotenate, biotin, …) in symport with protons or other cations. This subfamily is highly amplified in Y. lipolytica (39 transporters including 15 members of the “nicotinate cluster”) and D. hansenii (27 transporters including 9 members of the “allantoate cluster”) compared to the 6–13 transporters in the three more recent Hemiascomycete species.

The sugar porter (SP-2.A.1.1) subfamily that comprises 17 transporters in C. glabrata is considerably amplified in S. cerevisiae (34 transporters) and rather unexpectedly so in D. hansenii (48 transporters). The latter species gained 21 members belonging to the new clusters of undetermined substrates 2.A.1.1.Z1/Z10/Z14 that altogether constitute only one single member in S. cerevisiae. Surprisingly Y. lipolytica has no member of the prominent clusters 2.A.1.1.5/6/30/31, which comprise a total of 18 members (HXT) in S. cerevisiae.

The drug:proton antiporter-1 (12 transmembrane spans) (DHA1-2.A.1.2) subfamily pumps out a variety of hydrophobic drugs and is believed to be driven by the proton motive force. The DHA1 subfamily comprises 12 transporters in S. cerevisiae and is highly amplified in Y. lipolytica (33 transporters) and D. hansenii (24 transporters while K. lactis comprises only 8 transporters). The major expansion in Y. lipolytica corresponds to two novel clusters of undetermined substrates: 2.A.1.2.Z1 and more specifically 2.A.1.2.Z6 that comprises 8 members unique to Y. lipolytica. Note that a related subfamily of similar function, the DHA2 subfamily that comprises 14 transmembrane spans, is much more stable during evolution.

The peroxisomal protein importer (PPI-9.A.5.1) subfamily groups members of the peroximal membrane translocon. This subfamily is highly amplified in Y. lipolytica (27 members) compared to <11 members in the other species.

The oligopeptide transporter (OPT-2.A.67.1.) subfamily carries out the uptake of small oligopeptides such as the sexual peptides, glutathione, and others. This subfamily does not exist in C. glabrata and is extremely amplified in Y. lipolytica (17 transporters including 16 members of the 2.A.67.1.4 cluster named oligopeptide transporter type 4). Noteworthy is the rather parallel pattern of amplification within the five Hemiascomycete species of OPT and the proton-dependent oligopeptide transporter (POT-2.A.17.2) subfamily that also takes up oligopeptides.

The siderophore iron transporter subfamily (SIT-2.A.1.16) comprises a total of seven clusters, from which the substrates of only four have been determined in S. cerevisiae. In Y. lipolytica, a cluster of undetermined substrate is highly amplified (14 members) while C. glabrata has no SIT member.

The cluster of fructose uniporters (FTU-2.A.1.1.13) that has no member in S. cerevisiae and is highly amplified in K. lactis (12 members) is present in all other Hemiascomycete species.

The putative long chain fatty acid transporters subfamily (Fat1-9.B.17.1) is composed of fatty acyl coa-synthases that have been proposed to control the uptake of long chain fatty acids. This subfamily is considerably amplified in Y. lipolytica, which contains a large species-specific cluster.

The proton-translocating NADH dehydrogenase (NDH1-3.D.1.2) subfamily that corresponds to site 1 of mitochondrial oxidative phosphorylation was known to be lost in S. cerevisiae (Onishi et al. 1967). Our analysis dates the loss of phosphorylation site 1 between the emergence of D. hansenii (where it is present) and that of K. lactis (where it is absent).

The large amino acid-polyamine-organocation family (APC-2.A.3) contains the YAT-2.A.3.10 subfamily that transports a variety of amino acids in yeasts. The YAT subfamily is expanded in D. hansenii (24 members) compared to 14–18 members in the other species. Noteworthy is the different pattern of expansion of another APC subfamily (ACT-2.A.3.4) that transports choline as well as amino acids and is expanded in Y. lipolytica (9 members) compared to 2 members in C. glabrata.

The human phagocyte NADPH oxidase-associated cytochrome b558 H+-channel family (CytB-1.A.20) comprises electrogenic heme-binding proton channels that are believed to compensate for production of superoxide radicals. It includes the yeast metal transporter FRE-1.A.20.5 subfamily that has ferric and cupric reductase activity. The latter comprises only 1 member in C. glabrata compared to 7 members in S. cerevisiae and expands up to 11 members in Y. lipolytica.

The ubiquitous subfamilies of transporters:

Supplemental Table D1 (http://www.genetics.org/supplemental/) includes 2125 transporters classified in 107 subfamilies that are present in all Hemiascomycete species. Not surprisingly, most of them are large superfamilies (see Table 4) transporting essential substrates such as the mitochondrial carrier (21 subfamilies), several major facilitators (transporting sugars, allantoate, siderophores, drugs, CoA, and phosphate), ABC transporters (7 subfamilies), and P-type ATPases (6 subfamilies). The numerous members of the nuclear pore complex (NPC) (5 subfamilies), the mitochondrial protein translocase (MPT) complexes, the outer mitochondrial pore (MPP), the oxidative phosphorylation complexes (F-ATPase, COB, COX), and the vacuolar-ATPase (V-ATPase) complex are fully conserved in all Hemiascomycete species as well as the calcium (VIC), chloride (ClC), or water (MIP) channels and amino acid (AAA, APC), cation (CDF), metal (MIT), nucleotide-sugar (DMT), or sugar-phosphate (TPT) transporters. The conservation of subfamilies of undetermined substrate such the MFS subfamily 2.1.Y1 or of the YaaH family (which is a controversed ammonium effluxer or acetate influxer) or of the lipid translocating efflux (LTE) family of unknown mechanisms is more unexpected.

View this table:

Mean and standard deviation of the subfamily size according to the different modes of evolution within the Hemiascomycete phylum

The different evolution patterns of transporter subfamilies:

Only 355 transporters are submitted to evolution drift in the Hemiascomycete phylum. They are classified into 97 subfamilies that generally are of smaller size than the ubiquitous subfamilies (Table 4). Three different patterns of evolution were distinguished. They are listed in Table 5 and illustrated in Figure 2. The species-specific subfamilies (in the pie graphs at the top) are present (gain in light shading) or absent (gap in dark shading) in only one species. The phylum-gained-or-lost subfamilies are definitively gained or lost during intervals of evolution spanning the emergence of two distinct species (for instance, between the emergence of D. hansenii and K. lactis) and persisting in the species that emerged thereafter (Figure 2, vertical thick arrows). When subfamilies are gained and lost more than one time independently within the five species, they are considered as being submitted to homoplasic evolution (Table 7). These different evolution patterns were distinguished quite simply by using the matrix of presence or absence of transporters in the different species illustrated in Table 6.

Figure 2.—

A summary of the evolution of the subfamilies of transporters. The numbers in boldface type refer to the numbers of gained or lost subfamilies. The numbers in parentheses indicate the total numbers of ORFs in these subfamilies. Dashed arrows refer to undocumented flows.

View this table:

Species-specific and phylum-evolving transporter subfamilies

View this table:

Identification principles of the different evolution patterns distinguished in Figure 2

View this table:

Homoplasic evolution

The species-specific transporters:

Table 5 identifies the transporter subfamilies that are unique or specifically absent in one of the five yeast species. A total of 17 transporters belonging to 15 subfamilies are unique to one Hemiascomycete species and a total of 21 subfamilies are absent in only one species (see Table 5 and supplemental Table D3 at http://www.genetics.org/supplemental/ for detailed listings and comments).

As shown in Figure 2 and Table 5, the species that exhibit the highest number of unique Hemiascomycete transporters are D. hansenii (nine specific transporters) and Y. lipolytica (five specific transporters) whereas S. cerevisiae, K. lactis, and C. glabrata have each only one unique transporter. As shown by the frequent use of Y as the fourth digit (see Table 5), the nature of the substrates transported by the unique transporters is often undetermined. This may be related to the particular nature of the growth substrates of Y. lipolytica (lipids and hydrocarbons) and of D. hansenii (protein hydrolysates).

The species-specific losses are most abundant in Y. lipolytica and C. glabrata. For the latter a general high loss of genes has been noted (Dujon et al. 2004) and believed to reflect the pathogenic status of the species. Curiously, the specific gains (and losses) of transporters are very limited in the milk-loving K. lactis.

The phylum-gains-and-losses of transporters subfamilies:

Table 5 lists the 15 subfamilies that are gained in D. hansenii and K. lactis and are maintained in the more recent species during the evolution of the Hemiascomyces phylum. In contrast, numerous subfamilies are present in Y. lipolytica (17 subfamilies), D. hansenii (14 subfamilies), and K. lactis (5 subfamilies) but absent in the younger species. They are therefore considered as being definitively lost after the emergence of Y. lipolytica, D. hansenii, or K. lactis, respectively. It has to be noted that most of the transporters lost after emergence of Y. lipolytica were also present in N. crassa. The nature of the transporters gained or lost during phylum evolution is commented upon in supplemental Table D4 (http://www.genetics.org/supplemental/).

The homoplasic transporters:

Table 7 shows that 12 transporter subfamilies comprise members that have emerged two or even three times during the speciation of the five Hemiascomycete species. They belong to a variety of families and the physiological driving forces of these evolution patterns are difficult to comprehend especially as most of the transported substrates are of undetermined nature.

The mitochondrial carrier:

The mitochondrial carrier (MC) family is unusual by its large size: 172 Hemiascomycete transporters distributed in 36 subfamilies. The transported substrates of 27 of the 35 S. cerevisiae carriers are known so far. They include monocarboxylic, dicarboxylic, or tricarboxylic acids; sulfate, phosphate, iron, manganese, basic, or acid amino acids; ATP/ADP or GTP/GDP nucleotides; and carnitine, SAM, CoA, FAD, or TPP cofactors. The MC family comprises mitochondrial and peroxisomal members. Even though each carrier is believed to comprise six transmembrane spans, their hydrophobic properties are atypical and their number is underestimated by most software and especially by TMHMM.

Despite a large spread of phylogenetic subfamilies and clusters among the five fungal species listed, each species contains a similar total number of individual transporters (32–39) that can be classified in 36 phylogenetic subfamilies (supplemental Table D1 at http://www.genetics.org/supplemental/). Twenty subfamilies are present in all five species but several species-specific gains and losses are identified. Table 8 lists the 21 MC transporters that have no homologs in S. cerevisiae (as determined by an E-value <10−35. All of them correspond to novel MC subfamilies of undetermined substrates. Compared to S. cerevisiae, a total of 11 novel subfamilies of mitochondrial carriers are identified in the four other Hemiascomycete species. The species that have lost the site 1 of oxidative phosphorylation (K. lactis, C. glabrata, and S. cerevisiae) do not seem to have been submitted to particular gain or loss of carriers. Noteworthy is the presence of 10 mitochondrial carriers in Y. lipolytica; most are also present in N. crassa but are lost in D. hansenii and afterward. All have undetermined substrates. Both Y. lipolytica (2.A.29.Y19) and C. glabrata (2.A.29.Y15) contain a species-specific MC subfamily, also of undetermined substrate.

View this table:

Hemiascomycete mitochondrial carrier subfamilies that are absent in S. cerevisiae


Transmembrane transporters make up to 10% of the proteins encoded by the yeast genomes. They may be considered as metabolic checkpoints for complex anabolic or catabolic pathways. The gain or loss of transporters during speciation is an essential, and possibly initial, step in the differentiation of large intracellular metabolic pathways. Therefore the tracing of transporters during evolution may provide a vital, but rather neglected so far, panorama of the physiological and biochemical speciation processes.

To trace the evolution of transporters from the Hemiascomycete phylum we constructed a database (termed YETI) comprising all transporters within five complete genome sequences spanning ancient and more recent Hemiascomycete species, Y. lipolytica, D. hansenii, K. lactis, C. glabrata, and S. cerevisiae, using N. crassa as a Euascomycete reference outgroup. The YETI database uses the TCDB nomenclature and distinguishes the phylogenetic families, subfamilies, and clusters according to the TCDB five-digit system that we have adapted to cover all established transporters from the five studied species. Due to the large evolution span of the species used, we expect that the YETI database annexed in supplemental Table D1 (http://www.genetics.org/supplemental/) will provide a reference data set for the nonambiguous semiautomatic annotation of the vast majority of transporters from the novel Hemiascomycete genome sequences that are presently accumulating.

Numerous phylogenetic studies have established that the chronological order of emergence of the different species in the Hemiascomycete phylum is the following: Y. lipolytica, D. hansenii, K. lactis, C. glabrata, and S. cerevisiae. To study the speciation process under ideal conditions, one should compare the ancestral genomes of each of these species as they may have continued to evolve after separation from the phylum. These ancestral genomes are not available; we thus consider that the genome sequences of the presently available species provide a valid approximation of speciation events.

A total of 2480 transporters have been identified within the five species and classified into 204 phylogenetic subfamilies, characterized by the fourth digit, which has been adopted as the most significant level of analysis. We found that only 97 subfamilies comprising 355 individual transporters were submitted to the evolution drift. The statistical “variance” value of the size of these phylogenetic subfamilies within each species allowed us to identify those that had shown considerable expansion or contraction during speciation. The most variable subfamilies of transporters are listed in Table 3.

We also established that the emergence and loss of transporter subfamilies during speciation as well as their expansion or contraction in size can follow five different evolution patterns schematized in Table 6. The ubiquitous transporters are conserved in all Hemiascomycete species. The species-specific transporters are gained or lost in only one of the studied species. Other transporters are definitively lost or gained at a given step of the phylum evolution and these phylum gains or losses are definitively maintained thereafter. Homoplasic transporters have emerged and are lost more than one time independently during the evolution span under study.

The molecular mechanisms of emergences or losses of transporter genes during evolution are not understood and are probably diverse. It is, however, reasonable to assume that they generally result from physiological selection of diverging duplicated genes during adaptation to new ecological niches.

When more genome sequences become available, some transporter subfamilies that are considered here as species specific may be found shared by other species or subspecies. Similarly, transporters considered here as being definitively gained or lost only once may become homoplasic. Nevertheless we believe that the general methodology developed here and schematized in Table 6 will remain valid and may lead to the development of a new set of phylogenetic and classification criteria based on the presence or absence of a transporters subfamily. Such phylogenetic criteria may complement the classical systematic physiological parameters such as growth or fermentation on different substrates.

Full physiological analysis of our data is not possible here but some comments on a few striking results can be made. The emergence of the Y. lipolytica and D. hansenii species is accompanied by drastic gains of transporters compared to those identified during emergence of K. lactis, C. glabrata, and S. cerevisiae. For instance, a total of 154 transporters from the three large families, ACS (anion transporters), OPT (peptide transporters), and SP (sugar transporters), are encoded in Y. lipolytica and D. hansenii genomes whereas only 101 transporters from the same families are found in C. glabrata and S. cerevisiae. We may assume that physiological constraints produced by the nature of the available growth substrates may have driven the emergence of transporters. The drastic definitive loss of transporters before and after emergence of D. hansenii is also surprising and may be related to still undefined metabolic and physiological specificities. Another striking observation is the very limited number of species-specific transporters in K. lactis and the specific loss of several transporter subfamilies during emergence of C. glabrata or Y. lipolytica.

Some of our genomic findings could have been expected from known physiological traits such as the loss of the mitochondrial site 1 of oxidative phosphorylation between D. hansenii and K. lactis. Similarly, the proliferation of peroxisomal transporters and fatty acid-related proteins in Y. lipolytica may have been expected as this species is known to grow on fatty acids. The emergence of oligopeptide transporters in Y. lipolytica and D. hansenii may reflect the ability of species to grow in the absence of other nitrogen sources (Ahearn et al. 1968). The physiological drives underlining more unexpected genomic amplifications are puzzling, such as the amplification in Y. lipolytica and D. hansenii of anion:cation symporters (transporting anionic vitamins such as nicotinate and allantoate), of drug:H+ antiporter transporters, and of quinate:H+ symporters. The specific gain in Y. lipolytica of ABC transporters of the ABCA1 subfamily that so far had members only in mammalian cells is unexpected. Also the specific scarcity in S. cerevisiae of monocarboxylate porters of the sialate:H+ symport family (comprising only the Jen1p monocarboxylate transporter in S. cerevisiae compared to six transporters in K. lactis) is surprising but is congruent with some pioneering work (Cassio et al. 1987). The scarcity of SITs in C. glabrata could be explained by the likely absence of siderophores in blood. The unexpected broad sequence divergence of mitochondrial transporters in Y. lipolytica and D. hansenii merits further scrutiny.

In brief, the tracing of the genes encoding transporters within related yeast species genomes has established the involvement of transporter protein in the evolution mechanism of speciation. We hope that the rigorous identification and annotation of transporter proteins through the YETI classification system and further elaboration of the different evolution patterns initiated here for the Hemiascomycete phylum will contribute to such investigations.


We thank our colleagues of the Genolevures consortium and Anne-Catherine Lantin and Jérémie Nsengimana for help in the preparation of the databases. This work was supported by the Interuniversity Attraction Poles Programme-Belgian Science Policy. Development of the annotation pipeline (Bioflow) was supported by funding from the Walloon Region (FIRST EUROPE Objectif 3, no. EPH3310300R0082).


  • 1 These authors contributed equally to this article.

  • 2 Present address: Facultés Universitaires Notre-Dame de la Paix, 61 rue de Bruxelles, 5000 Namur, Belgium.

  • Communicating editor: F. Winston

  • Received June 18, 2005.
  • Accepted August 4, 2005.


View Abstract