A Functional-Phylogenetic Classification System for Transmembrane Solute Transporters

TRANSPORT NOMENCLATURE

Communication of concepts relevant to transmembrane transport phenomena generally depends upon the use of a uniform, well-defined and accepted, universally understood set of terms that can be used by the international community of scientists regardless of national origin or discipline of training. In this section I therefore present the terms currently in use in the field and mention which of these terms have been recommended for adoption by the TC panel of the IUBMB. It is anticipated that the acceptance of these terms will greatly facilitate the interchange of information by scientists and students of transport internationally.

Almost all transmembrane transport processes are mediated by integral membrane proteins, sometimes functioning in conjunction with extracytoplasmic receptors or receptor domains as well as with cytoplasmic energy-coupling and regulatory proteins or protein domains (51, 112, 130, 139). Each such complex of these proteins and/or protein domains is referred to as a transport system, transporter, porter, permease system, or permease. These are all equivalent terms that are in general use by members of the transport community. A permease (porter) is a protein or protein complex that catalyzes a vectorial reaction, irrespective of whether or not it also catalyzes a chemical or electron transfer reaction that drives the vectorial process. Thus, many transport systems can be thought of as catalytic proteins or protein complexes analogous to enzymes or enzyme complexes. By definition, transporters facilitate vectorial rather than, or in addition to, chemical reactions. The preferred terms for these transport systems are transporters or porter.

Permease-mediated transport can occur by any one of three distinct but related processes. First and simplest is facilitated, equilibrative, or protein-mediated diffusion, a process that is not coupled to metabolic energy and therefore cannot give rise to concentration gradients of the transported substrate across the membrane. Two primary modes of facilitated transport have been recognized in biological systems: channel type and carrier type (Fig. 1). In channel-type facilitated diffusion, the solute passes in a diffusion-limiting process from one side of the membrane to the other via a channel or pore that is lined by appropriately hydrophilic (for hydrophilic substrates), hydrophobic (for hydrophobic substrates), or amphipathic (for amphipathic substrates) amino acyl residue moieties of the constituent protein(s). The structures of several such channel proteins have now been examined and elucidated by X-ray crystallographic techniques (see below). In carrier-type facilitated diffusion, some part of the transporter is classically presumed to pass through the membrane together with the substrate (143, 151). Whether or not this presumption is correct is not known, as no classical carrier has yet yielded to the analytical tools of the X-ray crystallographer.

• Open in new tab
• Download powerpoint

Fig. 1.

Scheme illustrating the currently recognized primary types of transporters found in nature. These proteins are initially divided into channels and carriers. Channels are subdivided into α-helical protein channels, β-barrel protein porins (mostly in the outer membranes of gram-negative bacteria and eukaryotic organelles), toxin channels, and peptide channels. Carriers are subdivided into primary active carriers, secondary active carriers (including uniporters), and group translocators that modify their substrates during transport. Primary sources of chemical energy that can be coupled to transport include pyrophosphate bond (i.e., ATP) hydrolysis, decarboxylation, and methyl transfer. Oxidation-reduction reactions, light absorption, and mechanical devices can also be coupled to transport (see text). Secondary active transport is driven by ion and other solute (electro)chemical gradients created by primary active transport systems. The only well-established group-translocating system found in nature is the bacterial phosphoenolpyruvate:sugar PTS, which phosphorylates its sugar substrates during transport.

Carriers usually exhibit rates of transport that are several orders of magnitude lower than those of channels. Moreover, in contrast to most channels, they exhibit stereospecific substrate specificities. Although both channels and carriers may exhibit the phenomenon of saturation kinetics, this is a more common characteristic of carriers. Very few carriers have been shown to be capable of functioning by a channel-type mechanism, and the few that exhibit this capacity generally do so only after the protein has been modified, either by covalent or noncovalent ligand binding or by imposition of a large membrane potential. Moreover, while most channels are oligomeric complexes, many carriers can function as monomeric proteins. These observations led to the suggestion that channels and carriers are fundamentally, not superficially, different.

If energy expenditure is coupled to transmembrane solute translocation, then a system catalyzing facilitated diffusion can become an active transporter. Such a system is considered to be a primary active transporter if a primary source of energy (i.e., a chemical reaction, light absorption, or electron flow) is coupled to the process. It is considered to be a secondary active transporter if a secondary source of energy (i.e., an ion electrochemical gradient, termed the proton motive force [PMF] in the case of protons or the sodium motive force [SMF] in the case of sodium ions), generated at the expense of a primary energy source, is coupled to the process. The transport panel considered all of these terms to be acceptable.

Active transporters (or porters) can function by uniport, symport, or antiport. Uniporters (the preferred term), also called single-species transporters or facilitated diffusion carriers (the less-preferred terms), catalyze the transport of a single molecular species, and transport therefore occurs independently of the movement of other molecular species. Symporters (the preferred term), also classically called cotransporters, catalyze the transport of two or more molecular species in the same direction. The fact that a single point mutation in a symporter can convert a carrier into a uniporter (41, 62, 66, 75, 147) emphasizes the superficial distinction between these two types of carriers. Antiporters (the preferred term), also called countertransporters, exchange transporters, and exchangers, catalyze the exchange of one or more molecular species for another. Antiport processes can be subdivided into two categories: antiport of like molecules (i.e., solute-solute antiport) and antiport of unlike molecules (i.e., solute-cation antiport). Many uniporters and symporters also catalyze solute-solute antiport, sometimes at rates that are substantially greater than those of uniport or symport. Some carriers catalyze solute-solute antiport at rates that exceed those of uniport or symport by 10³- to 10⁵-fold, and uniport via these carriers is of little or no physiological consequence (110). Such systems are said to be obligatory antiporters or exchangers.

Accelerative solute-solute antiport or countertransport has long been considered to be a diagnostic characteristic of carriers. Early transport kineticists concluded that its demonstration eliminated the possibility that a transporter functions by a channel-type mechanism and suggested that clear boundaries exist between carriers and channels (79, 135). Subsequent observations that certain “carriers” could apparently be converted into “channels” by chemical treatment (16, 17, 28, 29, 56), by imposition of large membrane potentials (131, 132, 149), or by ligand binding (13) led many students of transport to consider these boundaries indistinct. Our in silico phylogenetic and protein structural analyses suggest that these examples may be special cases and tend to reemphasize the importance of the channel-versus-carrier distinction (123, 127).

A few carriers modify their substrates during transport. The best-characterized such system is the bacterial phosphotransferase system (PTS), which phosphorylates its sugar substrates using phosphoenolpyruvate as the phosphoryl donor. Sugars taken up from the external milieu via the PTS are thus released into the cytoplasm as sugar-phosphates. Any process in which the substrate is modified during transport is termed group translocation. Although originally proposed in different form by Peter Mitchell as a general mechanism, its occurrence appears to be highly restricted in nature.

CONSIDERATIONS FOR THE SYSTEMATIC CLASSIFICATION OF TRANSMEMBRANE SOLUTE TRANSPORTERS

The introduction of Linnaeus of a universal classification system for living organisms allowed the rationalization of the tremendous complexity of biological relationships into an evolutionary framework. Similarly, the introduction by the international Enzyme Commision of a universal enzyme classification system greatly increased our conception of the functional relationships of these proteins. Although protein-domain classification systems have been suggested, no comparable classification system has yet been proposed for proteins that catalyze vectorial reactions rather than (or in addition to) chemical reactions. In this section I describe the proposal for a universal system of classification for transporters based on both function and phylogeny.

As noted above, enzymes have long been classified in accordance with the directives and recommendations of the Enzyme Commission (31). The commission developed its directives decades ago, long before protein sequence data became available. Their system of classification was based solely on function. It was tacitly assumed that proteins of similar catalytic function would be closely related and that they therefore should be grouped together. We now know, however, that two different enzymes catalyzing exactly the same reaction sometimes exhibit completely different amino acid sequences and three-dimensional structures, function by entirely different mechanisms, and apparently evolved independently of each other, converging only with respect to the final reactions catalyzed. The enzyme classification system is thus limited in that it reflects only the reactions catalyzed by and the substrate specificities of the enzymes. It does not recognize the phylogenetic origins of these proteins and therefore does not reflect structural or mechanistic features.

As has been extensively documented, molecular phylogeny provides a reliable guide to protein structure and mechanism of action. It also provides an indication (albeit less definitive) of the specific process catalyzed and the substrate acted upon (127). Since the former characteristics are fundamental traits of a protein while the latter characteristics are more superficial traits, sometimes merely reflective of single amino acyl residue changes in a protein, it would be reasonable to suggest that as more and more sequence and phylogenetic data become available, these data should be used to provide the most reliable basis for protein classification. Since single amino acyl residue substitutions in permeases can give rise to different substrate-binding specificities (15, 23, 44, 94), these characteristics should be used in the final level of classification rather than in a primary level. We conclude that recognition of the evolutionary process provides a reliable guide to structure, mechanism, and function, although a few exceptions may exist (102, 127). If molecular phylogenetic studies can accurately retrace the evolutionary process, they should be used as a basis for any rational system of protein classification.

Some of the enzymes classified within the enzyme classification system are asymmetrically situated within an anisotropic, hydrophobic lipid membrane that separates two aqueous compartments. The resultant asymmetry allows these enzymes to catalyze vectorial as well as chemical modification reactions, as clearly enunciated decades ago by Peter Mitchell (83-85). Some of these integral membrane enzymes do, in fact, catalyze transmembrane transport of ions or other small solutes. However, most currently recognized solute permeases do not catalyze a chemical reaction and consequently are not included within the enzyme classification system. The comprehensive system of permease classification proposed here has the potential to encompass all types of transporters, both those that are currently recognized and those that are yet to be discovered.

THE TC SYSTEM

Early studies revealed that transport proteins could be grouped into families based exclusively on the degrees of similarity observed for their amino acid sequences (118). The significance of family assignment remained questionable until the study of internal gene duplications that had occurred during the evolution of some of these families established that these families had arisen independently of each other, at different times in evolutionary history, following different routes (119). In this section I will evaluate and utilize both function and molecular phylogeny for the purpose of conceptualizing transport protein characterization and classification (see also reference 120).

According to the proposed classification system, now recommended by the transport nomenclature panel of the IUBMB, transporters are grouped on the basis of five criteria, and each of these criteria corresponds to one of the five entries within the TC number for a particular permease. Thus, a permease-specific TC number has five components, V, W, X, Y, and Z. V corresponds to the transporter class, while W corresponds to the subclass (see Table 1). X specifies the permease family (or superfamily), while Y represents the subfamily in a family (or the family in a superfamily) in which a particular permease is found. Finally, Z delineates the substrate or range of substrates transported as well as the polarity of transport (in or out). Any two transport proteins in the same subfamily of a permease family that transport the same substrate(s) using the same mechanism are given the same TC number, regardless of whether they are orthologs (i.e., arose in distinct organisms by speciation) or paralogs (i.e., arose within a single organism by gene duplication). The mode of regulation proves not to correlate with phylogeny and was probably superimposed on permeases late in the evolutionary process. Regulation is therefore not used as a basis for classification.

There are four recognized classes of transporters: channels, porters, primary active transporters, and group translocators (Table 1). Sequenced homologs of unknown function or mechanism and functionally characterized permeases for which sequence data are not available are included in a distinct class, class 9. Deficiencies in our knowledge will presumably be eliminated with time as more sequenced permeases become characterized biochemically and as sequences become available for the functionally but not molecularly characterized permeases. One additional class (class 8) is reserved for auxiliary transport proteins. It should be noted that each subclass of transporters has a two-digit TC number (V.W); each family has a three-digit TC number (V.W.X); each subfamily has a four-digit TC number (V.W.X.Y); and each permease type has a five-digit TC number (V.W.X.Y.Z).

As mentioned above, the primary level of classification in the TC system is based on mode of transport and energy-coupling source. The classes and subclasses of transporters currently recognized are listed below.

FAMILIES OF TRANSPORTERS

The current index of transport system families is presented in Table 2. There are more than 250 entries, each of which usually describes a single family. Some of these families are actually large superfamilies with more than a thousand currently sequenced members (e.g., the voltage-gated ion channel (VIC) family (TC 1.A.1) (88); the major facilitator superfamily (MFS) (TC 2.A.1) (96, 125), and the ATP-binding cassette (ABC) superfamily (TC 3.A.1) (130, 139)). Others are very small families with only one or a few currently sequenced members. Most families, however, are currently of intermediate sizes, with between 5 and 500 sequenced members.

All of the families included in Table 2 will undoubtedly expand with time, and new families will be identified. The availability of new protein sequences will occasionally allow two or more currently recognized families to be placed together under a single TC number. In a few cases, two families are already known for which some evidence is available suggesting that they are related, e.g., the monovalent cation:proton antiporter-1 (CPA1) and CPA2 families (TC 2.A.36 and 2.A.37), the nucleobase-cation symporter-1 (NCS1) and NCS2 families (TC 2.A.39 and 2.A.40), as well as the l-lysine exporter, resistance to homoserine/threonine, and cadmium resistance families (TC 2.A.75, 2.A.76, and 2.A.77, respectively) (124, 148). Such evidence is usually based on limited sequence and/or sequence motif similarities, common function, and/or similar protein size, topology, and structure. When “missing link” sequences or three-dimensional structural data become available so that proteins of two families can be unequivocally grouped together within a single family, the lower TC number will be adopted for all of the family members, and the higher TC number will be abandoned.

The rigorous criteria used to delimit a family have been defined previously (121, 122). Briefly, in order for two proteins to belong to the same family, they must exhibit a region of 60 residues or more, in comparable portions of the two proteins, that have a comparison score in excess of 9 standard deviations (27). At this value, the probability that the degree of sequence similarity observed for these two proteins occurred by chance is less than 10⁻¹⁹ (25). It is considered that this degree of sequence similarity could not have arisen either by chance or by a convergent evolutionary process (32, 118). A minimum of 60 residues was arbitrarily selected because many protein domains in water-soluble proteins are of about this size.

The complete TC system is available on our web site (http://www-biology.ucsd.edu/∼msaier/transport/ ), where the descriptions, primary references, and list of functionally characterized protein members of all families are provided. The whole-genome analysis data upon which this classification system was initially based are found on an included subportion of this web site, which was constructed under the guidance of Ian Paulsen (100, 101). This site will be updated continuously as new information becomes available. Anyone noting errors or incomplete listings is encouraged to contact me to provide the missing information and references.

As noted above, members of a transporter family generally utilize a single mode of transport and energy-coupling mechanism, thus justifying the use of these functional categories as the primary basis for classification. However, a few exceptions have been noted. First, the arsenite efflux permease (ArsAB; TC 3.A.4) of E. coliconsists of two proteins, ArsA and ArsB. ArsB is an integral membrane protein that presumably provides the transport pathway for the extrusion of arsenite and antimonite (134, 153). ArsA is an ATPase that energizes ArsB-mediated transport. However, when ArsB alone is present, as in the case of the arsenical resistance pump ofStaphylococcus aureus, transport is driven by the PMF (14). Expression of the E. coli arsB gene in the absence of the arsA gene similarly gives rise to PMF-driven transport. The presence or absence of the ArsA protein thus determines the mode of energy coupling.

The ArsB protein is a member of a large superfamily of ion transporters, the ion transporter superfamily, in which at least two families exhibit the unusual capacity of being able to incorporate auxiliary constituents that alter the transport characteristics of the carrier (107, 127). Such promiscuous use of energy is exceptionally rare and has been documented in only a very few instances. When such an effect is reported, we shall usually classify the permease in accordance with the more complicated energy-coupling mechanism (in this case, as an ATP-driven primary active transporter [class 3] rather than as a secondary carrier [class 2]). However, in this unique case, the TC nomenclature panel of the IUBMB has recommended that a second family describing the PMF-driven ArsB homologs be included in the TC system (TC 2.A.45), as many ArsB homologs function by ATP-independent, ArsA-independent mechanisms.

Examples of secondary carrier families in which promiscuous transport modes have been reported include the mitochondrial carrier family (TC 2.A.29) and the triose phosphate/nucleotide sugar transporter (TP-NST) family (TC 2.A.50). Proteins of both families are apparently restricted to eukaryotic organelles. Members of these families normally catalyze carrier-mediated substrate-substrate antiport and are therefore classified as secondary carriers. However, treatment of mitochondrial carrier family members with chemical reagents, such asN-ethylmaleimide or Ca²⁺ (16, 17, 28, 29, 56), or imposition of a large membrane potential (ΔΨ) across a membrane into which a TP-NST family member has been incorporated (131, 132, 149), has been reported to convert these antiport-catalyzing carriers into anion-selective channels capable of functioning by uniport. Another secondary carrier that may be capable of exhibiting channel-like properties is the KefC protein of E. coli (13), which is a member of the CPA2 family (TC 2.A.37). “Tunneling” of ions and other solutes through carriers with little or no conformational change has been discussed (42). Again, the more complicated carrier-type mechanism, which appears to be relevant under most physiological conditions, provides the basis for classifying these proteins (i.e., as class 2 carriers rather than class 1 channels).

CHARACTERISTICS OF THE FAMILIES

Table 3 summarizes some of the key characteristics of most of the transporter families that we have identified. Categories 1.D and 2.B (non-ribosomally synthesized channels and carriers, respectively), 8 (auxiliary transport proteins), and 9.B and 9.C (putative but uncharacterized transporters) have been omitted (compare Table 1 with Table 3). Table 3provides the family TC numbers, the abbreviations of the families, and the substrates of transporters included within each family. Substrates that are common to one transporter are separated by commas, while substrates of different transporters within the family are separated by semicolons. Thus, in the major intrinsic protein (MIP) family (TC 1.A.8), aquaporins generally transport water but not organic compounds, while glycerol facilitators generally transport short, straight-chain polyols but not water. A few members of the family may transport both (see reference 97 for a review). A recent report has provided evidence that a member of the MIP family can accommodate anions (154), but this observation is of uncertain physiological significance.

Table 3 also includes the size ranges of the individual protein members of the families and the numbers of (putative) transmembrane α-helical segments (TMSs) included within the permease polypeptide chains. All members of a family usually exhibit similar topological features, although several exceptions have been noted. When a homo- or heterooligomeric structure has been established for an intact permease, this fact is also indicated. Finally, the kingdoms in which members of the family have been identified, the approximate number of members that have been identified in each family, and representative examples of well-characterized members are also provided. The table is largely self-explanatory, but detailed information as well as primary and secondary references are provided on our web site and may be available in book form in the near future (Saier et al., unpublished data).

CROSS-REFERENCING PERMEASES BY ACCESSION NUMBER

Protein accession numbers can generally be used to find protein sequences of any sequenced protein referred to in the TC system. An accession number never changes once entered into a database. It therefore provides a quick and easy means of identifying a specific protein sequence. Moreover, it allows access to the database description of the sequenced protein, including structural, topological, and functional information. SwissProt (SP) database entries provide the most detailed information about the proteins, and SwissProt accession numbers are therefore provided when available. When not available, other accession numbers will be provided.

The accession numbers of all representative transport proteins included in the tables of the current TC system can be found on our web site. Accession numbers usually consist of one or two letters followed by four, five, or six digits. A given letter is used by only one database: GenBank (GB), SP, or Protein Information Resource (PIR). Thus, for example, O, P, and Q are used exclusively by SP; D, J, K, L, M, U, X, Y, and Z are used exclusively by GB; and A, B, C, H, I, and S are used exclusively by PIR. However, when AB, AE, or AF is followed by a six-digit number, this is an alternative GB accession number, and when JC, JH, or JN is followed by a four-digit number, this is a PIR accession number. It should be noted that a single sequenced protein may have multiple accession numbers, but no SP or PIR accession number refers to more than one protein. Because a GB accession number refers to a nucleotide sequence that may encode multiple proteins, a GB accession number may provide the sequences of several proteins.

A table entitled Cross-Referencing Permeases by Accession Number is included in our web site. In this table, accession numbers for most of the proteins included in the TC system as of June 1999 are tabulated in alphabetical and numerical order. These may be of general utility to the student of transport, as their availability allows one to easily search all databases using the various BLAST search tools (3). Knowledge of a TC number allows one to quickly identify (i) the protein referred to, (ii) the transport system of which that protein is a constituent, (iii) the substrate specificity of that system, (iv) the family to which that permease belongs, (v) the mode of transport used, (vi) the energy-coupling mechanism used, and (vii) many of the characteristics of that permease family. Thus, cross-referencing by accession number is useful when trying to identify the family to which a newly sequenced protein belongs.

As noted above, one needs only to conduct a BLAST search, and all sufficiently similar homologs will be displayed. When the accession number of any one of these retrieved homologs is shown to correspond to one of the established members of a family, the family to which the newly sequenced protein belongs is immediately known. Furthermore, by identifying the proteins with the highest BLAST scores (smallestP values), one can immediately recognize the closest homologs. This information provides an indication of the most likely substrate specificity, energy-coupling mechanism, and physiological function of the newly sequenced permease protein. Cross-referencing of accession numbers and TC numbers therefore provides a simple and rapid approach to the initial characterization of a newly sequenced porter. I and my colleagues, working with Andrei Lupas (SmithKline-Beecham), are currently developing a search tool based on the TC system that will allow anyone to search the complete TC system using sequence, sequence motif, accession number, gene name, protein name, family name, etc. (A. Lupas et al., unpublished data).

GROUPING TRANSPORT SUBSTRATES BASED ON BIOLOGICAL SIGNIFICANCE

In 1993 and again in 1996, Monica Riley presented an extensive tabulation of E. coli gene products (114, 115). To facilitate this endeavor, enzymes were classified based on the nature of the molecule(s) (substrates) acted on. In order to cross-reference transport systems based on substrate specificities, a basis for classifying potential substrates had to be devised. We have done so, creating a system that includes virtually all currently recognized transport substrates. This system of cross-referencing transporters is described here.

All known transport substrates have been classified into eight categories (Table 4). These categories are I, inorganic molecules; II, carbon compounds, III, amino acids and derivatives; IV, bases and derivatives; V, vitamins, cofactors, signaling molecules, and their precursors; VI, drugs, dyes, sterols, and toxic substances; VII, macromolecules; and VIII, miscellaneous compounds.

Most inorganic molecules (category I) are cationic or anionic. However, some channel proteins are nonselective, and aquaporins of the MIP family (TC 1.1) transport water selectively. The four defined subcategories for category I therefore include A, nonselective; B, water; C, cations; and D, anions. Inorganic compounds not falling into one of these subcategories are classified as others (subcategory E), and this subcategory can be subdivided in the future if desired.

Carbon compounds (category II) have similarly been grouped into four defined subcategories: A, sugars, polyols, and their derivatives; B, monocarboxylates; C, di- and tricarboxylates; and D, noncarboxylic organic anions (organophosphates, phosphonates, sulfonates, and sulfates). Subcategory E (others) encompasses all other carbon compounds.

Amino acids and their derivatives (category III) have been subdivided into A, amino acids and conjugates; B, amines, amides, and polyamines; C, peptides; D, other related organocations; and E, others. Bases and their derivatives (category IV) have been subcategorized into A, nucleobases; B, nucleosides; C, nucleotides; D, other related derivatives; and E, others. Vitamins, cofactors, and cofactor precursors (category V) have been subcategorized into A, vitamins and vitamin or cofactor precursors; B, enzyme and redox cofactors; C, siderophores and siderophore-iron complexes; D, signaling molecules; and E, others. Drugs, dyes, sterols, and toxics (category VI) have been subcategorized into A, multiple drugs and dyes; B, specific drugs; C, bile salts and conjugates; D, sterols and conjugates; and E, others. Category VII is devoted to macromolecules: A, complex carbohydrates; B, proteins; C, nucleic acids; D, lipids; and E, others. Finally, category VIII (miscellaneous) encompasses any transport substrate that does not fall into categories I to VII. So far no transport substrate has been relegated to category VIII, and very few of those in categories I to VII have fallen into the “other” category.

A few compounds belong to more than one category. For example, bile acids fall into both category II.B and category VI.C. Theoretically, oligosaccharides (e.g., lactose, raffinose, and maltooligosaccharides) could be classified either in II.A or in VII.A. We have elected to put oligosaccharides into category II.A and reserve category VII.A for larger molecules such as polysaccharides, teichoic acids, and lipooligosaccharides. Thus, category II.A generally refers to smaller carbohydrates normally taken up by cells for purposes of carbon catabolism, while category VII.A refers to structural carbohydrates that are synthesized by cells and exported intact.

Some permease systems transport a range of compounds that fall into more than one category. For example, a single ABC export system may catalyze efflux of multiple drugs (VI.A) and peptides (III.C), and it may also facilitate phospholipid flipping between the two bilayers of a membrane (VII.D). Such permeases are rare, but when they do occur, they will be included in all applicable categories.

DISTRIBUTION OF TRANSPORTER TYPES BASED ON SUBSTRATE SPECIFICITY

As described above, Table 4 groups potential transport substrates according to structure and biological significance. This system of substrate classification has been used to cross-reference transport systems according to the types of substrates and the specific substrate(s) transported.

Table 5 presents the distribution of transporter types based on substrate specificity. In this table, permeases are categorized into four groups: α-type channels, β-type porins, primary carriers (regardless of the primary source of energy utilized and including PTS-type group translocators), and secondary carriers (including uniporters, symporters, and antiporters). Transporter types of unknown mode of transport or energy-coupling mechanism (categories 9.A and 9.B) were not included in Table 5.

α-Type channel proteins (TC 1.A) generally either are nonselective (13 types) or function in the transport of inorganic ions (I.C and I.D; 18 families) or proteins (VII.B; 5 families). One type (aquaporins in the MIP family; TC 1.A.8) transports water, while another type of the same family (glycerol facilitators of the MIP family) transports straight-chain polyols and small organic molecules such as urea. Some MIP family proteins may transport both water and small, neutral organic molecules, but with the possible exception of a single MIP family member (154), none of the MIP family channel proteins have been shown to be selective for ions or larger molecules. Besides MIP family members, only two other recognized channel families include members that are specific for organic compounds. These families are the urea transporter family (TC 1.A.44) and the phospholemman (PLM) family (TC 1.A.27). The PLM family includes members that transport organic anions. Channel-forming toxins (TC 1.C) are generally nonspecific, or they exhibit weak charge selectivity (i.e., anion selective or cation selective).

Porins (TC 1.B) are pore-forming proteins that exhibit β-barrel structures or variations on the β-barrel structural theme. They are localized to the outer membranes of gram-negative bacteria, mitochondria, and chloroplasts. They exhibit a wider range of substrate selectivities than do the α-type channel proteins cited above (Table5). However, like channel protein types, most porins either are nonselective, exhibit some degree of anionic or cationic selectivity, or function in the export of macromolecules across the outer membranes of gram-negative bacteria. Many porins allow passage of any molecule smaller than a certain cutoff size (usually about 500 to 1,000 Da). Of the macromolecular export porin types, more than half export proteins, but several transport complex carbohydrates, and at least one functions in DNA transport.

Examination of the specificities of porins for organic substrates reveals a wide variety of specificities. The maltoporin of E. coli (TC 1.B.3.1.1) and the raffinose porin of E. coli(TC 1.B.15.1.1) are both inducible by their sugar substrates, but while maltoporin is quite specific for maltooligosaccharides, the raffinose porin transports a variety of oligosaccharides. Other porins have been reported to preferentially transport organophosphates (TC 1.B.1.1.2), fatty acids (TC 1.B.9.1.1), nucleosides (TC 1.B.10.1.1), or organic solvents such as toluene (TC 1.B.9.2.1). Still another type apparently exhibits specificity for short-chain amides and urea (TC 1.B.16.1.1). Members of the outer membrane receptor family (TC 1.B.14) import vitamin B₁₂ and a variety of iron-siderophore complexes in a process that is coupled to the PMF via the TonB-dependent energy-coupling system (TC 2.C.1). To what degree these proteins are biochemically selective is not always clear, although they are often encoded within operons that exhibit specific induction properties, and these systems are constituents of transenvelope transport complexes.

Primary carriers are in general highly specific for one or a few related substrates, and like channels, they are almost always selective for inorganic ions or macromolecules. Those specific for organic molecules of small or intermediate sizes belong to either of two superfamilies, the ABC superfamily (TC 3.A.1) or the PTS functional superfamily (TC 4.A.1-6). The PTS is actually a group translocating system, since it phosphorylates its substrates using phosphoenolpyruvate as the phosphoryl donor. For the purpose of tabulating substrate specificities as presented in Table 5, we have grouped this functional superfamily together with the active transporters. The energy-coupling mechanisms used for the transport of ions are diverse, involving pyrophosphate bond hydrolysis, decarboxylation, methyltransfer, oxidoreduction (both hydride shift and electron flow), and light absorption (Fig. 1). By contrast, all macromolecular primary active transporters use ATP or GTP hydrolysis to drive export, although a few macromolecular secondary active exporters use the PMF as the energy source for transport.

Secondary carrier types (TC class 2, subclass 2.A) exhibit a very different spectrum of substrate specificities. None is nonselective or water selective, but many are selective for specific inorganic cations or anions, and a few appear to function in the export of lipids, proteins, or complex carbohydrates, as is characteristic of primary active transporters. Others function in the transport of the many different types of small organic molecules found in biological systems. Thus, every class of molecules included in Table 5 is transported by one or more currently identified secondary carrier(s). For example, members of four transporter families are known to transport sugars and polyols; 13 types transport monocarboxylates, and 12 types transport di- and tricarboxylates. Eighteen types function in the transport of amino acids and their conjugates. It is clear that secondary carriers are primarily responsible for the transport of small organic molecules in virtually all living organisms.

The last column in Table 5 reveals the total numbers of families involved in the transport of the various types of biologically important compounds. About equal numbers of families are concerned with transport of inorganic and organic compounds, with most of the 105 families for inorganic molecules transporting ionic species. Thirty-three families are concerned with carbon source uptake, while 34 are concerned with the uptake of nitrogen-containing compounds (amino acids, bases, and their derivatives). Only 14 families include members that take up compounds in category V (i.e., vitamins, cofactors, signaling molecules, and related compounds), while only 8 families are concerned with transport of hydrophobic substances (category VI). Macromolecules are exported via the transporters of 26 families. While inorganic molecules and macromolecules are transported by all four types of systems, small organic substances are transported almost exclusively by secondary carriers (Table 5).

SUBSTRATE SELECTIVITIES

CELLULAR MACROMOLECULAR EXPORT SYSTEMS

Table 15 tabulates the transport systems that catalyze the export of macromolecules. The majority of these systems utilize ATP hydrolysis to drive transport, but several also appear to exhibit a dependency on the PMF. PMF-dependent exporters for complex carbohydrates may include those of the polysaccharide transporter family, while those for proteins include members of the twin-arginine-targeting family. Bacterial holins and certain channel-forming toxins are probably energy-independent protein exporters and importers, respectively. Bacterial MscL channels and mammalian Bcl-2 channels probably also function by energy-independent mechanisms. While three recognized families participate in polysaccharide export, 13 tabulated families participate in protein transport. The mitochondrial and chloroplast envelope protein transport systems can be thought of either as matrix uptake systems or as cytoplasmic export systems. It should be noted that the protein-specific holins and ABC exporters as well as the diphtheria and the botulinum and tetanus toxin importers are relatively simple in structure. ABC export systems may function with trans-envelope protein complexes (157, 158). The more general systems, which transport many proteins, however, consist of large complexes of multiple protein constituents. Only a single type of export system tabulated, the type IV secretory pathway family (TC 3.A.7), mediates export of nucleoprotein complexes. However, another type of system, the bacterial competence-related DNA transformation transporter family (TC 3.A.11), mediates uptake of naked single-stranded DNA in bacteria competent for natural transformation, and a poorly characterized family of systems, the septal DNA translocator family (TC 9.A.16), may function in the transmembrane transport of double-stranded DNA. Two types of active transport systems (ABC and P-type ATPases) are believed to mediate phospholipid flipping from the inner leaflet to the outer leaflet of a biomembrane, although the anion exchanger and RND families of secondary carriers include members that have been reported to do the same (4, 46, 137) (Table 15). These last-mentioned transport systems represent the only examples in which macromolecular export systems are ubiquitous, being found in eukaryotes as well as prokaryotes.

CLASSIFICATION OF TRANSPORTERS OF UNKNOWN MECHANISM

Several families of proteins are known in which one or more members have been shown to function as transporters, but either the mode of transport (channel versus carrier) or the energy source driving solute accumulation or expulsion has not been determined. Consequently, it is not possible to assign the transporter family to a defined category (1-4). Such families fall into TC category 9.A. Additionally, families of proteins in which no member of the family has been shown to be a transporter are known, although some indirect experimental evidence, or inferences based on topological analyses and/or operon gene product analyses, supports such a possibility. Such families fall into TC category 9.B. Finally, functionally characterized transporters lacking an identified sequence fall into TC category 9.C. The families listed in these categories will either be transferred to one of the established categories when their transport mechanism becomes defined or be eliminated from the TC system if it is shown that these proteins are not actual transporters. In this section, the families that constitute TC class 9.A will be discussed. Those of classes 9.B and 9.C will not be considered further here.

Table 16 tabulates families of known transporters for which no member has yet been clearly defined in terms of either its mode of transport (channel or carrier) or its energy-coupling mechanism. Many of these permeases belong to families that include members which are specific for inorganic ions. Eleven families are inorganic ion specific, and 10 of these are cation specific. Most of these families include members that are specific for a single ion or a few closely related ions. However, one family (low-affinity cation transporter, TC 9.A.20) transports a variety of cations, exhibiting unexpectedly broad specificity.

Some of the category 9.A permeases (belonging to six distinct families) exhibit specificity for small organic compounds. These compounds vary from amides and amines, including urea and uric acid, to peptides and vitamin precursors. Thus, a variety of organocations, organoanions, and neutral molecules are transported. One family (polysaccharide transporter) transports complex polysaccharides, probably by a PMF-dependent mechanism, but the energy-coupling mechanism is still poorly defined. Considerations to be discussed in the next section allow prediction of the modes of action of several of these systems. Putative transporters (category 9.B) are not discussed here but can be evaluated by consideration of the information provided in our web site.

PREDICTIONS OF TRANSPORT MODE BASED ON PROTEIN TOPOLOGY

Examination of the topologies of families of recognized α-type channels (TC 1.A) and secondary carriers (TC 2.A) reveals that these two functional types of transporters differ fundamentally both in polypeptide structure and in oligomeric composition. This fact suggests that there are fundamental differences between these two functional types of transporters and that channels and carriers truly represent distinct types of proteins. This structural distinction between the two principal functional types of transporters is evaluated in this section.

As illustrated in Fig. 2, most families of cellular integral membrane α-type channel proteins include members that possess three or fewer TMSs per polypeptide chain (Fig. 2A), while almost all families of secondary carriers include members that possess eight or more TMSs (Fig. 2B). When permease families of unknown transport mode are examined (Fig. 2C), some are found to fall into the 1 to 3 TMS range observed for most channel families, while others fall into the 8 to 14 TMS range observed for most carrier families. It can be anticipated that most of the former proteins will prove to be channels, while the latter will mostly prove to be carriers. The disproportionate number of families of unknown mechanism of action with about 6 TMSs leads to the possibility that new types of transporters, not yet characterized, may be found among these families.

• Open in new tab
• Download powerpoint

Fig. 2.

Established or predicted topologies for channel proteins (A), carrier proteins (B), and proteins of unknown transport mode (C). The proteins included in A are the channel proteins of TC category 1.A, while the carriers represented in B are the families of TC category 2.1A. Because most primary carriers of categories 3 consist of heterooligomers, many of very complex structure, these were not included in the analyses depicted.

Interestingly, very few carriers have been shown to be capable of functioning as channels under any experimental set of conditions. Two of those that do exhibit this unusual property prove to consist of polypeptide chains that have 6 TMSs. Families of such transporters include the mitochondrial carrier family (TC 2.A.29) and the TP-NST family (TC 2.A.50) (17, 28, 29, 132). The E. coliKefB and KefC proteins of the CPA2 family (TC 2.A.37) also seem to have the capacity to function either by a carrier-type K⁺:H⁺ antiport mechanism or by a K⁺-specific channel-type mechanism (38, 39). Proteins of this last-mentioned family exhibit 10 to 14 putative TMSs and therefore have the topology of a typical carrier. They exhibit channel-type activities following treatment with certain chemicals. Ambivalent modes of transport for members of a few other secondary carrier families have also been noted (see reference121 for further consideration of this point).

On the basis of all of the observations summarized in this section, we propose that, with only a few exceptions, channel proteins are structurally and functionally different from carriers. Channels are proposed to generally consist of oligomeric structures in which the monomeric protein subunits exhibit ≤3 TMSs. Some exceptions to this rule have resulted from the fusion of non-transport-regulatory domains to the channel-forming constituents of transporters (88). Regardless of topology, however, the channel generally results from the proper association of multiple channel-forming subunits or domains. Carriers, on the other hand, are proposed to generally consist of functional monomers that exhibit 8 to 14 TMSs or, less frequently, of functional dimers that have 4 to 7 TMSs. In these situations, the transport pathway requires the participation of just one or, at most, two polypeptide chains. The numbers of known exceptions to this topological rule are small (Fig. 2).

RECOGNIZED DISTRIBUTION OF TRANSPORTER FAMILIES IN THE THREE DOMAINS OF LIFE

Our studies of the distribution of proteins within the various families of transporters have revealed that most families are restricted to just one of the major domains of life, bacteria, archaea, or eukarya. Other families are ubiquitous, being found in all three domains. If lateral transfer and fixation of genetic material occurred appreciably between these three domains during the past two billion years, one would expect many families to be ubiquitous. Our observations have therefore led to the suggestion that the ubiquitous families are among the oldest families and that they existed before divergence of the three major domains of organisms, some three billion years ago. The domain-specific families are therefore those that arose late, after the “great split.” Alternatively, some of these families may have diverged in sequence from their ancestral system at rates that exceed those observed for the recognized ubiquitous families. Even if this occurred, however, the lack of recognizable homologs in the other domains suggests the absence of appreciable lateral transfer.

Table 17 summarizes the distribution of the identified families of various channel types, secondary carriers, primary carriers, group translocators, and transporters of unknown mechanism in the three domains of life: bacteria, archaea, and eukaryotes. Regardless of transporter category, the distribution is simple. Thus, many families of transport systems are found exclusively in either bacteria or eukaryotes, and four have been identified only in archaea. Many of these families may prove to exist in only one of the three major domains of life, and most such families probably arose within that kingdom after the three domains of living organisms separated from each other. It is also possible that some ancient families that existed prior to the divergence of archaea and eukaryotes from bacteria will prove to be restricted to just one or two domains because a particular transport mode or energy-coupling mechanism is incompatible with (or disadvantageous to) the organisms within a particular domain. It is particularly noteworthy in this regard that although hundreds of genes of the PTS (TC 4.A) have been sequenced from bacteria and many of the genes encoding the cytoplasmic constituents of the PTS function in regulation rather than in transport, not a single such gene has yet been found within an archaeal or eukaryotic genome (J. Reizer and M. H. Saier, Jr., unpublished results). Similarly, although ABC-type efflux pumps are universal, extracytoplasmic receptor-dependent ABC-type uptake permeases (TC 3.A.1) as well as receptor-dependent tripartite ATP-independent periplasmic transporter-type uptake permeases (TC 2.A.56) are found only in prokaryotes (107). These observations have led us to conclude that horizontal transmission and fixation of genetic material across the three domains of life has occurred rarely, at least in the case of genes encoding many types of transporters, during the past two billion years.

Several families are found ubiquitously in all three domains of living organisms or are found in at least two of these domains (Table 17). We predict that many (but not necessarily all) of the latter families will prove to be represented in all three domains. The lower representation of transporter types in the archaeal domain presumably reflects, at least in part, the paucity of both sequence data and functional analyses reported for this domain. It should be noted that the vast majority of ubiquitous families (about two-thirds) are families of secondary carriers. The distribution of transporter types and the identification of the relevant families are presented in Table18 for the various channel types, in Table 19 for the various carrier families, and in Table 20 for the various types of primarily active transporters.

Several interesting conclusions derived from the data in Tables 17 to20 can be tentatively made. First, of the families of channels, only three families (MIP, VIC, and ClC) are ubiquitous. A fourth such family may prove to be the metal ion transporter family (TC 9.A.17). Second, except for these families and two families (MscL and MscS) specific to bacteria, all families of α-type protein channels are found exclusively in eukaryotes. Third, the vast majority of these eukaryotic families are restricted to animals. Finally, the vast majority of protein and peptide toxin families, holin families, and β-strand-type porin families are restricted to bacteria. In the case of the toxin families, this unequal distribution may reflect, at least to some extent, my greater focus (and that of research scientists in general) on bacterial toxins rather than those of eukaryotes.

The distribution of secondary carriers is not so polarized. Thus, 31 carrier families (40%) are ubiquitous, compared to 25 (32%) and 14 (19%) that are specific to bacteria and eukaryotes, respectively. Only eight (10%) are found in two of the three kingdoms. If lateral transfer of genetic material coding for transporters has been minimal, as we have proposed (119, 120), then it would appear that a large proportion of the secondary carrier families came into existence early, before the split between the three domains, compared to channel or primary carrier families.

Primary carrier families are found solely in bacteria, ubiquitously, and in bacteria plus eukaryotes, in decreasing numbers in that order. The relatively large percentage of systems in the last category is due to the presence of three families of H⁺-pumping electron or hydride-transferring carriers that are found only in mitochondria and/or chloroplasts of eukaryotes in addition to bacteria. Since both of these eukaryotic organelles are believed to have arisen from bacteria long after the split between bacteria and eukaryotes (93), the actual proportion of primary active transporter families specific to bacteria may be considered substantially greater, while that of carriers shared by bacteria and eukaryotes may be smaller (see lightface values in parentheses in Table 17). Finally, one family, the fungal-archaeal rhodopsin family (TC 3.E.1), is unusual in that although these light-driven ion transporters are restricted to one small group of archaea, homologs that may not function in transport are found in yeasts and other fungi. This may represent one of those rare examples where distant homologs of a transporter family have evolved to serve very different functions (see below). The results summarized in Tables 18 to 20 provide a detailed breakdown of channels, secondary carriers, and primary carriers, respectively, and the TC numbers of the families in each category are provided so that the reader can easily identify the relevant families.

It has been noted that archaeal metabolic enzymes and transporters frequently resemble the homologous bacterial sequences more than those of the corresponding eukaryotic proteins, although archaeal proteins of DNA replication, transcription, and translation are more similar to those of eukaryotes (20, 34). This observation has been interpreted to suggest that archaea are mosaic organisms, with nucleic acid and protein-biosynthetic enzymes derived primarily from an early eukaryotic precursor cell, while transport and metabolic functions are derived primarily from a primordial bacterium. If such a “fusion” event was responsible for the generation of the archaeal lineage, a significant number of transporter families should prove to be restricted to bacteria and archaea but lacking in eukaryotes. The availability of four complete archaeal genome sequences has allowed resolution of this question. Of the 200 families represented in Table17, only 7 (3.5%) are shared by bacteria and archaea but not by eukaryotes. Similarly, very few families are represented in bacteria and eukaryotes but not archaea. Moreover, some of these last-mentioned families are represented only in eukaryotic organelles, suggesting a more recent bacterial origin. Thus, very few families may prove to be restricted to just two of the three domains of living organisms. An alternative view concerning the origin of archaea, such as that proposed recently by Poole et al. (103), may be worth considering.

TRANSPORT PROTEINS FOR WHICH THREE-DIMENSIONAL STRUCTURAL DATA ARE AVAILABLE

An ultimate understanding of transport will depend upon detailed structural data for each of the major classes of transport systems. Until recently, few or no such data were available. The approach of X-ray crystallography has yielded very significant advances in understanding the three-dimensional structures of certain classes of integral membrane proteins. Most of these proteins are of prokaryotic origin, and they do not yet include the major classes represented by secondary carriers, group translocators, and ATP-driven primary pumps. However, channel-type proteins and both light- and electron flow-driven proton pumps are now structurally understood at high resolution.

Table 21 lists the transport proteins for which high-resolution three-dimensional structural data are available. Four types of channel proteins (α-helix-forming channels, β-barrel porins, peptide channels, and protein toxin channels) are represented, as are electron flow-driven and light absorption-driven proton pumps. The structures of these and other membrane proteins have been discussed by Sakai and Tsukihara (128). The fact that no chemically driven primary carriers, no facilitators or secondary carriers, and no group translocators are represented means that we are currently far from a structural understanding of transport. Although the structures of several water-soluble domains (receptors or energy-coupling proteins) of some of these systems have been determined (i.e., ABC-type receptors, the transhydrogenase hydride transfer domains and pumps, and the energy-coupling proteins of the PTS) (105, 106, 126), the structures of the integral membrane constituents of these systems are still unsolved. In fact, high-resolution structures are not available for a single transport system within one of these categories, even though these types of transporters represent the major types found in nature. Much work will be required before molecular transport can be put on a firm structural basis.

TRANSPORTER FAMILIES INCLUDING NONTRANSPORTING HOMOLOGS

Of the currently recognized 250-plus families of established transporters, we have noted that only 7 include transmembrane proteins that have been shown to function in a capacity other than transport. Of these seven families, four include homologs that are believed to serve as receptors (Table 22). In the case of the ammonium transporter family of NH₃ (or NH₄⁺) transporters, a yeast homolog, Mep2p, acts as both a sensor and a transporter. In the MFS and amino acid-polyamine-organocation superfamilies, the putative transcriptional regulatory sensors have not been shown to be incapable of transporting their ligands, although the available evidence is against it (57, 65). In the case of the MFS receptors, protein domains that interfere with transport function may be required to convert a transporter into a signaling receptor (57, 65). This scenario is reminiscent of the sensory rhodopsins, for which interaction with transducer proteins blocks proton transport (159). In the case of the RND superfamily, a homologous integral membrane domain serves as a sterol-binding domain, and this domain is found in several receptors, and even an enzyme, 3-hydroxy-3-methylglutaryl (HMG)-coenzyme A reductase (Table 22). This provides one of the best-documented examples of a family of transporters that has truly diverged in function. As noted above, the established bacteriorhodopsin family includes sensory rhodopsins that mediate phototaxis as well as homologs in S. cerevisiae that probably do not function in transport (52, 67, 159). Indirect evidence suggests that most of these yeast proteins are integral membrane heat shock or organic solvent shock proteins (53, 108). They lack the conserved lysine to which retinal binds in Schiff's base linkage in the archaeal proteins. However, a homologous retinal-containing photoreceptor has recently been identified in Neurospora crassa (9). It is possible that the fungal chaperone proteins contain noncovalent retinal and/or energize protein folding by catalyzing proton transport through themselves.

Finally, water-soluble constituents, and possibly also the integral membrane transporter domains of both the ABC and PTS superfamilies, have been shown to function in various nontransport capacities (50, 126, 136). Thus, the extracytoplasmic receptors of ABC permeases have homologs that are domains within bacterial transcription factors (90) as well as eukaryotic neurotransmitter receptors of the glutamate-gated ion channel family (TC 1.7) (87, 141). Similarly, a few homologs of the ATP-hydrolyzing ABC proteins function in catalysis of bacterial cytoplasmic processes, and some ABC permease homologs function in regulation of other transporters (21). Bacterial PTS II.A proteins and protein domains function in regulatory processes, sometimes in addition to their transport functions and sometimes instead of their transport function (126). A few PTS transporters (II.C constituents) also serve as sensory transducers (24, 71, 78). Nevertheless, it seems surprising that so few transporter homologs function in a nontransport capacity. This fact greatly facilitates the annotation and functional assignment of putative proteins whose sequences are (or will be) revealed by genome sequencing. It also suggests that transporters evolved as a class of proteins independently of other protein types, such as enzymes, structural proteins, and regulatory proteins.

AUXILIARY TRANSPORT PROTEINS

Proteins that in some way facilitate transport across one or more biological membranes but do not themselves participate directly in transport are classified as auxiliary proteins (Table23). These proteins by definition always function in conjunction with one or more transport proteins. They may provide a function connected with energy coupling to transport, play a structural role in complex formation, or serve a regulatory function (see section 8.A in Table 2). Examples include the membrane fusion proteins (TC 8.A.1), which provide a periplasmic bridge between primary, energy-coupled efflux permeases in the cytoplasmic membranes of gram-negative bacteria, and outer membrane factors (TC 1.B.17), which provide porin-type channel functions across the latter structures (63, 99, 138, 157, 158). Membrane fusion protein family proteins allow solute export across both membranes of the gram-negative bacterial cell in a single energy-coupled step (10, 30, 73).

Other proteins that span the cytoplasmic membrane with large domains in the extracytoplasmic space of the gram-negative or gram-positive bacterial cell and sometimes function with additional cytoplasmic domains include members of the cytoplasmic membrane-periplasmic auxiliary 1 (MPA1; TC 8.A.3) and MPA2 (TC 8.A.4) families (33, 98, 152). These proteins are believed to function directly in export and possibly also in the regulation of complex carbohydrate export by virtue of the protein tyrosine kinase activities that are associated with their cytoplasmic domains (147).

A most interesting set of auxiliary transport proteins is the TonB family (TC 2.C.1) of heterotrimeric protein complexes that allow transmission of energy in the form of the PMF across the inner membranes of gram-negative bacteria to energize uptake of iron-siderophore complexes and vitamin B₁₂ across outer membranes via proteins of the outer membrane receptor (TC 1.B.14) family. The latter proteins exhibit structural features superficially resembling those of outer membrane porins (37, 74, 99). However, they differ from typical porins in being monomeric and exhibiting 22 antiparallel β-strands in the β-barrel structure. The heterooligomeric TonB-ExbBD complex may prove to transport protons, explaining their capacity to respond to the PMF. Limited sequence similarity of these proteins to the MotAB proteins (TC 1.A.45) (A. Lupas, personal communication) further suggests this possibility.

Other auxiliary proteins include the energy-coupling proteins of the bacterial phosphoenolpyruvate-dependent sugar-transporting PTS (categories 4.A.1 to 4.A.6). Enzymes I and HPr proteins (TC 8.A.7 and 8.A.8, respectively) serve as phosphoryl transfer proteins, thereby providing both energy-coupling and enzyme-catalytic functions (104, 111). The enzymes I are homologous to phosphoenolpyruvate synthases and pyruvate:phosphate dikinases that normally function in phosphoenolpyruvate synthesis (117). We have suggested that the PTS evolved relatively late and depended on the conversion of preexisting phosphoenolpyruvate synthases into phosphoenolpyruvate-dependent phosphoryl transfer enzymes of the PTS (112).

Finally, many proteins are clearly implicated in transport, but they appear to play indirect and ill-defined roles in the process. These proteins include the rBAT (TC 8.A.9) and MinK (TC 8.A.10) family members. rBAT and MinK are believed to function in conjunction with amino acid carriers and potassium ion channels, respectively (35, 77, 116, 146). They may play roles in stability and subcellular targeting.

Many additional auxiliary proteins are included in the tables describing porters of TC categories 1 to 4. Because of their tight association with particular transport systems, they are described as constituents of these systems rather than as auxiliary proteins of the 8.A class.

CONCLUSIONS AND PERSPECTIVES

In this article I have described a comprehensive classification system for transport proteins that has the theoretical potential to include all transmembrane transport systems found in all living organisms on Earth. We have attempted to design this system so that it can accommodate new information and incorporate new systems as these become available with minimal alteration in structure. We have designated this system the transporter classification (TC) system of the Transport Commission of the IUBMB. This system is based on a combination of functional and phylogenetic characteristics of transporters and their constituents. The incorporation of phylogenetic data is a departure from the classification system devised by the Enzyme Commission years ago for the classification of enzymes, but the use of phylogenetic information provides many advantages, as discussed in the introductory section. Thus, phylogeny provides the most reliable guide to structure, function, and mechanism, and it provides valuable information concerning the evolutionary history of a family. The TC system should be capable of incorporating any novel type of molecular transporter that may be discovered in the future as well as the ever-increasing numbers of novel transporters that fall into existing families. Rules have been presented that allow the systematic consolidation of families as evolutionary links between them become available. Our goal is to eventually automate the incorporation of novel transporters into the system without (or with minimal) human intervention. Since the classification system is based on both function and phylogeny, achievement of this goal will require automation of tree construction as each new sequence becomes available in public databases, as well as the incorporation of biochemical, genetic, and physiological data as these become part of the scientific literature. As additional genomes are sequenced, the achievement of this goal will also require that screening techniques and annotation of novel families and family members be streamlined. In conjunction with Andrei Lupas and the bioinformatics group at SmithKline-Beecham (5), automation is now being implemented. Our web site will soon serve as a search tool that, for the analysis of transport proteins, will hopefully prove to be as useful as the BLAST search tools of the National Center for Biotechnology Information. Continual revamping of in silico methods for achieving these goals represents a major challenge that will require cooperation on the parts of computational scientists, molecular biologists, and cell physiologists. Exactly how these goals should best be achieved cannot easily be anticipated, as they are likely to be tightly coupled to technological advances through the years.

Currently recognized transporters include simple proteins as well as large multisubunit complexes that either facilitate passive diffusion of molecules across membranes or use one or more types of energy to drive transport. A large number of potential energy-yielding reactions have already been shown to be coupled to transport. These include several distinct chemical reactions, such as bond breakage reactions (e.g., decarboxylation and pyrophosphate bond hydrolysis), chemical group transfer reactions (e.g., hydride and methyl transfer), and electron flow. In addition, light absorption and the flow of ions down electrochemical gradients can be used to drive transport. In the reverse direction, ion transport can function to drive flagellar rotation, ATP synthesis, or active transport across the outer membranes of gram-negative bacteria. Variations on the established themes as well as entirely new themes are likely to be revealed by the efforts of future investigators. Perhaps, as three-dimensional structural data become available for the major classes of primary and secondary active carriers as well as group translocators, we will be able to delineate the mechanistic details of these processes. As totally new transport modes, not yet imagined, may be revealed, the transport biologist has exciting new discoveries to look forward to. The classification system proposed here, based on both function and phylogeny, is designed to accommodate any such discoveries and will hopefully aid in delineating the applicability of the structural, mechanistic, and evolutionary principles established with a few model systems to the hundreds of transporter types currently recognized and yet to be discovered.