The tripartite ATP-independent periplasmic (TRAP) transporters of bacteria and archaea

Abstract

Until recently, extracytoplasmic solute receptor (ESR)-dependent uptake systems were invariably found to possess a conserved ATP-binding protein (the ATP-binding cassette protein or ABC protein), which couples ATP hydrolysis to the translocation of the solute across the cytoplasmic membrane. While it is clear that this class of ABC transporter is ubiquitous in prokaryotes, it is now firmly established that other, unrelated types of membrane transport systems exist which also have ESR components. These systems have been designated meriplasmic (TRAP) transporters, and they form a distinct class of ESR-dependent secondary transporters where the driving force for solute accumulation is an electrochemical ion gradient and not ATP hydrolysis. Currently, the most well characterised TRAP transporter at the functional and molecular level is the high-affinity C4-dicarboxylate transport (Dct) system from Rhodobacter capsulatus. This consists of three proteins; an ESR (DctP) and small (DctQ) and large (DctM) integral membrane proteins. The characteristics of this system are discussed in detail. Homologues of the R. capsulatus DctPQM proteins are present in a diverse range of prokaryotes, both bacteria and archaea, but not in eukaryotes. The deduced structures and possible functions of these homologous systems are described. In addition to the DctP family, other types of ESRs can be associated with TRAP transporters. A conserved family of immunogenic extracytoplasmic proteins is shown to be invariably associated with TRAP systems that contain a large DctQM fusion protein. All of the currently known archaeal systems are of this type. It is concluded that TRAP transporters are a widespread and ancient type of solute uptake system that transport a potentially diverse range of solutes and most likely evolved by the addition of auxiliary proteins to a single secondary transporter.

Tripartite ATP-independent periplasmic transporter, Periplasm, Extracytoplasmic solute receptor, Binding protein, C4-dicarboxylate, Symport, DctP, DctQ, DctM, Rhodobacter capsulatus

1 Introduction

Bacterial active transport systems have traditionally been divided into several general classes, defined by their energetic coupling mechanism, primary sequence and subunit composition [1–3]. The majority of prokaryotic transport families are either secondary transporters, that utilise an electrochemical ion gradient to facilitate solute transport, or primary transporters, in which movement of the solute is directly coupled to ATP hydrolysis [1–3]. Additionally there is a third group of transporters that phosphorylate their substrates during transport by a group translocation process [1–3].

Within these three groupings of active transport proteins there are many different families of transporters that have different structures and protein composition. An analysis of the predicted transport capabilities of a wide range of prokaryotes demonstrated the predominance of two particular families of solute transport systems encoded in their genomes [4]. The first of these are the major facilitator superfamily (MFS) of proteins that are a widespread grouping of secondary transporters containing a single subunit with 12 membrane-spanning helices, for example lactose permease from Escherichia coli (LacY). The second major family are the ATP-binding cassette (ABC) protein transporters, a large family of primary transporters that contain a cytoplasmic domain (the ABC protein) that binds and hydrolyses ATP to energise transport [5, 6]. It has been known for many years that ABC transporters involved in solute uptake contain an extracytoplasmic solute receptor (ESR)¹ component that is required for transport. The association of ESRs with transport systems was thought to be strictly limited to these ABC systems [2, 3, 6].

However, there is now evidence that a novel class of transport system exists which, although using an ESR in the transport mechanism, does not possess an ABC protein [7–9]. Biochemical experiments indicate that solute accumulation by such systems is not driven by ATP hydrolysis but by an electrochemical ion gradient. Thus, these systems are clearly secondary transporters which operate with an associated ESR. It is astonishing how such a distinct type of transporter remained undiscovered for so long, as it is now clear from genome sequence analysis that they are found in an incredibly diverse range of both bacteria and archaea, yet there is functional information for only one or two systems to date. The purpose of this review is to describe the characteristics and diversity of these transporters, discuss the roles that ESRs might play in transporters other than those of the ABC type, and to consider the implications for understanding energy coupling and evolution of microbial solute transport systems.

2 What is a TRAP transporter?

TRAP transporters [8] are ESR-secondary transporters composed of three protein components or domains; an extracellular solute receptor and two distinct integral membrane proteins of unequal size. In some cases the integral membrane components are fused into a single large membrane protein but with two distinct domains. The ‘TRAP’ designation has been recognised and incorporated in the transport classification (TC) system proposed by Saier [1, 10]; the TRAP family has been included in the electrochemical potential-driven category, and designated family 2.A.56 [1, 10]. The relationship between the subunit organisation of TRAP transporters and other types of solute transport system is shown in Fig. 1. The key property of TRAP transporters is the possession of an ESR in the absence of an ABC protein. Many questions are raised by discovery of the association of ESRs with secondary transporters, particularly concerning the transport mechanism itself in these systems as compared with conventional secondary transporters, and it re-focuses attention on the precise role of ESRs in transport. The functions of ESRs in ABC transport systems and what this can tell us about TRAP systems is discussed in more detail in Section 3.

Figure 1

Subunit composition of TRAP transporters compared with a typical ABC solute uptake system and a typical conventional secondary transporter. In an ABC uptake system, an extracytoplasmic solute binding protein (red circle with the solute depicted as a small black-filled circle) interacts with two membrane proteins or domains (green rectangles) and solute translocation is coupled to the hydrolysis of ATP by the ABC protein subunits (brown circles). The only common feature between TRAP transporters and ABC systems is the possession of the extracytoplasmic solute binding protein (red circle) which in the case of TRAP systems may be a DctP homologue but can also be an unrelated binding protein. Mechanistically, TRAP transporters are secondary transporters in their mode of energy coupling to a transmembrane electrochemical gradient (depicted here by the nH⁺ arrow, where n is the stoichiometry of proton translocation, although other coupling cations are also possible), and also possess a 12-TM helix protein (DctM homologue; cyan) which is distantly related to some types of secondary transporters in the ‘ion-transporter’ superfamily [9]. The DctQ homologue (blue rectangle) is a protein of uncertain function unique to TRAP transporters. The members of the largest family of secondary transporters (the MFS) consist of just a single solute-translocating protein, often with 12-TM helices (depicted here as a cyan rectangle to emphasise the functional relationship to the DctM homologue in TRAP transporters). Solute is depicted by black circles in each case. Q, DctQ homologue; M, DctM homologue.

The high-affinity binding protein-dependent C4-dicarboxylate transporter, encoded by the dctPQM genes in Rhodobacter capsulatus, was the first member of this new class of systems to be identified by a combination of molecular and biochemical methods [7, 8]. This system is described in some detail in Section 4, as a ‘model’ TRAP transporter along with two other TRAP systems that have been partially characterised either genetically or biochemically. The availability of microbial genome sequence information has allowed the identification of homologous systems in a wide range of bacteria and archaea [8, 9], defining the three constituent protein families of DctP, DctQ and DctM homologues. However, there are significant and interesting variations in different organisms in the gene arrangements, domain structure and function of these systems, and these aspects are discussed in Section 5. Finally, in Section 6, we briefly consider the evolutionary implications of the discovery of TRAP transporters.

3 The roles of ESRs in solute transport

There are several comprehensive reviews of the structure, mechanism and phylogeny of ESR-dependent transport systems in the ABC superfamily [1, 6, 10]. Here, attention will be drawn only to certain features of these systems that are highly conserved and which are useful in comparisons with the TRAP transporters. The structure and protein composition of ESR-dependent transport systems has usually been deduced from primary sequence information. All of the solute uptake systems known to date contain an ESR, either free in the periplasm of Gram-negative cells or else anchored in the inner membrane in Gram-positive cells and archaea [11]. ESRs generally share little primary sequence similarity, except those proteins that bind chemically similar ligands (e.g. [12]), yet their tertiary structures, where they have been determined, are all very similar [13]. This indicates that considerable sequence divergence has taken place during the evolution of ESRs, but that similar functional requirements have constrained large variations in structure. An extensive phylogenetic analysis of ESRs has revealed the existence of at least eight families, each of which bind chemically similar ligands [14]. Detailed ligand binding kinetics have been determined for several ESRs, and those studied appear to operate by a similar mechanism involving the closure of a binding cleft around the ligand (e.g. [15, 16]). The open-unliganded, open-liganded and closed-liganded forms of several ESRs have been observed by X-ray crystallography [13, 17]. Interestingly, it is now clear that several eukaryotic receptors possess extracellular domains that are homologous to ESRs, implicating a role for the ‘venus-fly trap’ mechanism in signal transduction in a range of cell-types (e.g. [18]).

A very large number of prokaryotic solute transporters have now been characterised and classified [1, 10], from which two generalisations can be made that are especially relevant. First, it is well known that none of the well-studied symport systems have ESRs [2, 3]. Secondly, ESRs are always present in solute uptake systems in the ATP-driven, ABC transporter family, whereas efflux systems in the same family, such as those for drugs and antibiotics [19] do not contain such proteins. Other, unrelated, types of efflux pump that are driven by ATP hydrolysis, such as that for arsenite [20], also do not have an associated ESR. These generalisations appear to indicate an important role for ESRs in controlling the directionality of a particular transport process. ESRs have very low K_d values for their ligands and are thus potentially well suited for scavenging low concentrations of substrates; they impart a high overall affinity and specificity to a given transport system and this is often cited as their primary role. However, it is important to emphasise that although these proteins are produced in large molar excess over the membrane components of the transport system, the ability of ESRs from Gram-negative bacteria to diffuse freely in the gel-like periplasm is restricted [21]. Quiocho [13] has stressed the fact that ESRs are able to remove water molecules from around the ligand during binding and that this dehydration may be important in the subsequent steps in the translocation mechanism.

None of these proposed roles for ESRs explain their particular association with ABC transporters. Indeed, many of these functions could be considered useful in other types of membrane transport systems. It is well known that ESR-independent mutants of ABC transporters can be isolated, as in the maltose system [22], in which amino acid substitutions in the integral membrane proteins (selected in a PBP deletion background) revealed the ability of these proteins to bind and transport the substrate directly, albeit with much lower affinity. Thus, under defined circumstances, ESRs can be dispensable in these systems and the existence of a substrate-binding site on the integral membrane components is clearly in accord with the ESR-independent nature of the solute export systems in the ABC superfamily. Experiments made possible by the availability of reconstituted systems [23, 24] explain this phenomenon and indicate a specific role for the ESR in transmitting a signal to the ABC protein(s), through its normal interaction with the transmembrane components, so that ATP is only hydrolysed at high rate when substrate is present [25–27]. For example, reconstitution of the maltose transporter from an ESR-independent mutant led to the continual hydrolysis of ATP, in the absence of maltose or binding protein, whereas in the wild-type complex, ATP hydrolysis was strictly dependent on liganded ESR [25]. In the histidine transport system, ESR-independent mutants mapped in HisP, the ABC protein, rather than the integral membrane components, but they too hydrolysed ATP independently of any signal [26]. There is also evidence that both liganded and unliganded ESR (HisJ) can interact with the membrane complex with equal affinity but whereas the unliganded conformation induces only slow ATP hydrolysis, the addition of ligand to the receptor increases the rate of ATP hydrolysis which leads to substrate translocation through the integral membrane proteins [27]. This general mechanism whereby the ESR goes through cycles of ligand binding and release while attached to the membrane complex is fully in keeping with the restricted ability of binding proteins to diffuse in the gel-like periplasm of Gram-negative bacteria and the potential anchoring of the ESR to the membrane in Gram-positive bacteria and archaea.

In this article we will review data that demonstrate that ESRs can be associated with secondary transporters in addition to ABC systems. What is their role in such systems? In general, symporters have lower substrate affinities than ESR-dependent transporters in the ABC family [2, 3] and so the ability to interact with an ESR may simply serve to increase the affinity of the system. Dehydration of the ligand by the ESR prior to delivery to a membrane symport system may also be advantageous, as mentioned above for the ABC systems. However, bearing in mind the fact that ESRs seem only to be associated with solute uptake systems in the ABC family, the addition of such a protein on the extracytoplasmic side of the membrane might also serve to convert a potentially reversible symport system into one which is unidirectional. Unidirectionality of some types of transport system might be of physiological significance in helping to maintain a large concentration gradient of an anionic solute against an opposing membrane potential (inside negative) [28].

4 Functional characteristics of TRAP transporters

4.1 The high-affinity C4-dicarboxylate transport (Dct) system of R. capsulatus

The Dct system of R. capsulatus was the first TRAP transport system to be characterised and is the only system where there is both biochemical and genetic information, hence we consider this to be the archetypal TRAP system and will describe its properties in detail in the following section.

4.1.1 Function and composition of the system

R. capsulatus is a purple photosynthetic bacterium which grows well on the C4-dicarboxylates malate and succinate. Early studies established the presence of a specific transport system for these substrates in the genus Rhodobacter[29] and a single osmotic shock sensitive, binding protein-dependent system (Dct) transporting malate, succinate and fumarate was identified as the sole uptake route for these substrates in R. capsulatus grown chemoheterotrophically in the dark [30–33]. The dct locus was cloned by complementation of Tn5 mutants defective in aerobic C4-dicarboxylate transport [33]. Sequencing, further mutagenesis and complementation studies [8, 32, 34] revealed the presence of three structural genes and two regulatory genes, organised in two divergently transcribed operons (dctPQM and dctSR respectively; Fig. 2).

Figure 2

Model for the regulation and synthesis of the R. capsulatus Dct system. DctS (S) is a sensor-kinase which detects C4-dicarboxylates (small filled circle) in the periplasm and autophosphorylates (P) on a conserved histidine residue. Phosphotransfer to DctR (curved arrow) results in DctR-P-dependent transcriptional activation from the dctP promoter and synthesis of the DctPQM proteins. DctP (red circle) is a periplasmic-binding protein specific for C4-dicarboxylates (small filled circle within the red circle). The presence of a partial transcriptional terminator between dctP and dctQ (shown as a filled stem–loop) results in greater abundance of dctP transcripts (short thick arrow) compared to those for dctPQM (long thin arrow). This explains the greater abundance of DctP compared to the membrane proteins of the system (DctQ and DctM). The stem–loop structure between dctM and the rypA gene (encoding a protein-tyrosine phosphatase; not part of the Dct system) is thought to terminate the dctPQM transcript. dctS and dctR are translationally coupled and are expressed constitutively at low levels (thin arrow above the dctSR genes denotes the transcript).

The dctS and dctR genes are translationally coupled and encode proteins with sequence similarity to the ‘two-component’ sensor–regulator systems [34]. DctS is predicted to be a sensor-kinase and DctR is predicted to be a response-regulator. It is thus likely that DctR transcriptionally activates expression of the dctPQM operon in response to a signal from DctS. When either dctS or dctR were insertionally inactivated by interposon mutagenesis, the resulting mutants were unable to grow aerobically on C4-dicarboxylates and did not synthesise the periplasmic C4-dicarboxylate-binding protein [34], demonstrating the essential role of these proteins as positive regulatory components. DctS contains two potential membrane-spanning sequences in the N-terminal region, with the intervening hydrophilic sequence likely to be located in the periplasm, forming a potential C4-dicarboxylate sensing domain. DctS also contains a PAS domain [35], which suggests that the protein may be responsive to signals in addition to the presence of C4-dicarboxylates, for example oxygen or redox state.

The dctP gene encodes the ESR of the transport system [32]. DctP is an abundant periplasmic protein which can be readily purified from osmotic shock fluid. In many of its biochemical properties DctP appears to be a typical ESR. Comparison of the deduced sequence with the N-terminal sequence of the purified protein revealed the presence of a typical signal sequence which is cleaved to give a mature 307-residue protein of M_r 33 567 [32]. DctP has been subjected to a detailed kinetic and thermodynamic analysis, by exploiting the intrinsic tryptophan fluorescence changes it exhibits upon interaction with ligands [36, 37]. This is discussed in more detail in Section 4.1.3.1.

The dctQ and dctM genes encode proteins that are very hydrophobic, with 62% and 73% apolar residues respectively, suggesting that they are integral membrane proteins. Heterologous expression of the cognate genes in E. coli under the control of a T7 promoter, and localisation of the products to the cytoplasmic membrane after IPTG induction has confirmed this [8]. Although DctQ migrates in SDS–PAGE at about 26 kDa, consistent with its deduced size of 24.6 kDa, DctM runs anomalously (at about 30 kDa rather than 47 kDa), presumably due to its more hydrophobic nature [8]. The hydropathy profile of DctQ indicates that there are four particularly hydrophobic regions of 18–20 amino acids which are potential membrane-spanning sequences. Several topology prediction programs suggest that the N- and C-termini of DctQ are located in the cytoplasm [38]. The actual topology of the protein has been investigated [38] by the use of the β-lactamase fusion technique, which supports the four-helix model and the cytoplasmic location of the N- and C-termini. This is also consistent with the asymmetric distribution pattern of positively charged residues, according to von Heijne’s rules [39]. The two-dimensional structure of the protein deduced from these studies is shown in Fig. 3.

Figure 3

Model of the topology of R. capsulatus DctQ derived from secondary structure prediction and DctQ’-BlaM fusion analysis. The amino acid sequence of DctQ is shown with the ends of the TM regions predicted by the TMHMM topology prediction program [38]. Positively charged residues (arginine+lysine) are shown in red to illustrate their asymmetric distribution according to von Heijnes rule [39]. The BlaM fusions are labelled alphabetically in yellow rectangular boxes along with the ampicillin MIC value of the fusion bearing E. coli strain. Fusions with high MIC values (>20 μg ml⁻¹ are localised to the periplasm, whereas fusions with a low MIC value (<20 μg ml⁻¹) are cytoplasmic. PL1,2, periplasmic loops; TM1-4, transmembrane helices. Figure taken from [38].

In contrast, the hydropathy profile of DctM implies a total of 12 potential membrane-spanning segments, arranged as two groups of six helices separated by a markedly hydrophilic region [8]. The hydropathy profile of DctM is remarkably similar to that of the twelve transmembrane helix transporters exemplified by many families of membrane transport proteins in bacteria and eukaryotes [1, 10]. This is consistent with a central role in solute translocation across the cytoplasmic membrane. Downstream of DctM there is an 80-nucleotide gap before the next ORF. Within this region, a potential stem–loop structure is present that may act as a transcriptional terminator. The gene 3′ of dctM (rypA) might normally be expected to encode the conserved ABC protein which acts as the energy coupling subunit of periplasmic permeases. However, RypA is a bacterial protein-tyrosine phosphatase which is not required for transport [8]. There is no other gene at the dct locus which could encode an ABC protein.

4.1.2 Regulation of synthesis and activity of the Dct system

Transport activity via the Dct system in R. capsulatus is inducible by C4-dicarboxylates and is repressed by other organic acids (e.g. lactate) and by glucose. Under inducing growth conditions, DctP becomes one of the most abundant periplasmic proteins in the cell and can easily be visualised by SDS–PAGE of crude periplasmic extracts [32]. However, DctQ and DctM are much less abundant. Although dctP insertion mutants are polar on dctQM, and cannot be complemented unless intact dctQ and dctM genes are also present [32], a GC-rich region of dyad symmetry which has the potential to form a stable stem–loop structure is located immediately 3′ of dctP, and is followed by a repeated CTTT motif [32]. This type of structure has been found in other R. capsulatus operons, and could act as a partial transcriptional terminator, permitting some readthrough into dctQ and dctM. This suggestion is supported by Northern blots using either dctP or dctQ probes hybridised to mRNA derived from fructose or malate grown cells [40]. The dctP probe detected an abundant 1kb transcript (from dctP) in addition to a less abundant 3-kb transcript (derived from dctP,Q and M) in malate, but not fructose-grown cells. The dctQ probe only detected the 3-kb transcript with mRNA from malate grown cells [40]. The promoter of the dctP operon has been localised in the dctP-dctS intergenic region by primer extension analysis, and DctR has been overexpressed, purified and shown to bind to this region by gel-shift assays [40]. All these data support the model for the regulation of synthesis of the Dct system shown in Fig. 2.

Many bacterial transport systems can be modulated in their activity by variations in the intracellular pH value [41]. Although, in vivo, cytoplasmic buffering and pH homeostasis normally keeps the internal pH value within narrow limits [42], this form of regulation may be important under some physiological conditions. The activity of ESR-dependent systems in particular has been shown to vary significantly with internal pH [43]. A study of the effect of external and internal pH on the activity of the R. capsulatus Dct system in intact cells [31] showed that a marked activation of transport occurred above an internal pH value of 7 and that below neutrality the rate of transport fell to zero. The data were obtained using the ionophore nigericin (which catalyses an electroneutral exchange of protons and potassium ions) over a range of external pH values. The molecular mechanism of such regulation is unknown, but conceivably may involve changes in the protonation state of key amino acid residues in the transport proteins. The effect may be physiologically significant, since utilisation of C4-dicarboxylates as sole carbon sources in batch cultures of R. capsulatus leads to a rise in the external pH, which to a smaller extent will increase the intracellular pH, thus increasing the activity of the transport system, particularly as the culture approaches stationary-phase. Control of the activity of the transport system by the value of the intracellular pH might also be important in other situations. For example, despite the fact that a significant membrane potential exists under anaerobic-dark conditions in R. capsulatus cells (thought to be generated by the electrogenic efflux of K⁺, [44]), malate transport via the Dct system was not detectable [33]. This is probably due to a lowered internal pH resulting in the formation of a reversed ΔpH across the cytoplasmic membrane under these conditions [45].

4.1.3 The mechanism of transport

4.1.3.1 The interaction of ligands with DctP

Exploitation of the intrinsic tryptophan fluorescence properties of ESRs has proven a useful way of studying their interaction with ligands [15, 16, 46, 47]. For example, the rates of ligand binding to the galactose, maltose, arabinose and histidine ESRs have been determined by stopped-flow fluorescence spectroscopy and this has revealed a simple association of protein and ligand, with a linear concentration dependence of the association rate constant [15, 16]. The mechanism of ligand binding to DctP has been studied in some detail by using both steady-state and stopped-flow fluorescence spectroscopy [36, 37]. The fluorescence properties of DctP are characterised by excitation and emission peaks centred around 280 and 319 nm respectively. The addition of fumarate to the protein causes a 17% quench in fluorescence without shifting the emission maximum, while succinate causes a 4.5% enhancement in fluorescence. d- or l-malate also cause a small (2%) enhancement [36]. Titration of the fluorescence indicated a binding stoichiometry of 1:1 and ligand dissociation constants (equilibrium K_d values) of 0.05 μM for l-malate, 0.17 μM for succinate, 0.25 μM for fumarate and 6.3 μM for d-malate were determined [36]. Stopped-flow fluorescence spectroscopy revealed that DctP displays kinetics which are inconsistent with a mechanism in which there is a simple association of ligand and protein [36]. The most important observation on which this conclusion was based was that the rate of ligand-induced fluorescence change was found to decrease in a hyperbolic fashion as the concentration of ligand was increased. The same kinetic behaviour was observed with d-malate, l-malate, succinate or fumarate as ligands and with protein that had been reversibly denatured with urea, thus removing any endogenously bound ligand which may have been responsible. These data are consistent with a slow isomerisation of the protein, with the ligand binding to only one form of an equilibrium mixture of two pre-existing protein conformations (Eq. 1):

1

BP1 represents the non-binding (closed) conformation, BP2 represents the binding (open) form and L denotes the ligand. It was assumed that the rate of ligand binding was much faster than the rate of the conformational change, so that the rate constant for the fluorescence change would be a single exponential function which would decrease with ligand concentration in the hyperbolic manner observed.

Fumarate binding was studied in the greatest detail, since this ligand gave the largest change in fluorescence. Fitting the data to Eq. 1 gave values for the rate constants k₁ and k₋₁ of about 6 s⁻¹ and 174 s⁻¹, respectively, and an equilibrium constant for fumarate binding (K₂) of 0.12 μM [36]. However, because of the distribution of the protein molecules between the two conformational forms, it was clear that the overall equilibrium (dissociation) constant (i.e. the K_d) would be higher than K₂, according to Eq. 2:

2

where K₁ is the equilibrium constant for the transition between closed (BP1) and open (BP2) forms. From the values obtained for the rate constants k₁ and k₋₁, it was calculated that K₁ was 0.035 and thus that the overall k_d value should be about 3.5 μM. A comparison of this value with that obtained by titration of the fluorescence of the protein under steady-state conditions (0.25 μM) suggested that there must be a further isomerisation of the protein following fumarate binding, but one which is optically silent (Eq. 3):

3

In this expanded scheme, BP2-L would represent the open-liganded conformation and BP3-L would represent the closed-liganded conformation, the latter being the form in which ESRs deliver their ligand to the membrane component(s) of the transport system. Using the overall affinity constants (1/K_d) for each ligand and the values for K₁ and K₂, the equilibrium constant for the third step (K₃) could be calculated. Assuming that ligand binding is a rapid equilibrium process, k₋₃ was taken to be the measured ligand dissociation rate constant in ligand-displacement experiments. Thus, values for k₃ could also be calculated. The results obtained [36] showed that the differences in ligand-binding affinity for the C4-dicarboxylates studied (l-malate>succinate>fumarate>d-malate) are due to differences in their respective association rate constants, rather than in their dissociation rate constants, which were very similar in each case; K₂[L] varied from 1.8×10⁸ M⁻¹ s⁻¹ for l-malate, to 3.1×10⁵ M⁻¹ s⁻¹ for d-malate, yet for each ligand the values of k₋₃ were between 7.9 and 11.7 s⁻¹[36]. This kinetic behaviour is in complete contrast to results obtained with a number of sugar and amino acid ESRs [15, 16], where the association rate constants were all of the same order but where observed differences in affinity were attributable to variations in the ligand dissociation rate constants.

The DctP protein is thus kinetically rather distinct from other ESRs which have been characterised to date, and this derives largely from the remarkable stability of the closed unliganded (BP1) conformation. It can be calculated from the values obtained for k₁ and k₋₁[36] that about 96% of the protein molecules adopt the BP1 conformation in the absence of ligand. The molecular basis for the position of the equilibrium between the open and closed forms of DctP in both the presence and absence of ligands was investigated in a separate study [37]. The rate of the conformational change from the BP1 to the BP2 form (i.e. k₁) was found to be pH dependent and the data could be described in terms of a two pKa function, which indicated the involvement of two distinct groups on the protein. The pK values obtained from an analysis of the pH dependence of k₁ were 5.4 and 10.3. Similar values were obtained from an analysis of the pH dependence of the amplitude of the fluorescence change. Investigation of fumarate binding at pH 6, 8 and 10.5 indicated that k₋₁ also followed a two-pKa function and that perturbation of the BP1–BP2 equilibrium towards the open conformation (above and below pH 8) concomitantly resulted in a decrease in the ligand binding affinity of the latter. By pre-equilibrating DctP with 10 μM fumarate and subsequently displacing this ligand with 2 mM succinate buffered to the desired pH, it was possible to determine that the fumarate dissociation rate constant (k₋₃) had a very similar pH profile to that of k₁[37]. In addition to these pH effects, an increase in ionic strength at pH 8 affected both k₁ and k₋₃ in a similar way, giving a hyperbolic increase in these rate constants to maximal values that were similar to those obtained at high pH. The data were interpreted as most likely reflecting the presence of a salt-bridge between an acidic residue (pK 5.4; possibly glutamate) and basic residue (pK 10.3; possibly lysine) in the protein which is responsible for stabilising the closed conformation. Thus, an increase in ionic strength or a change in pH away from neutrality appears to specifically destabilise this salt-bridge, resulting in a shift towards the open conformation. It is significant that crystallographic studies of both the sulfate [48] and phosphate- [49] binding proteins have demonstrated the importance of salt-bridges in ligand binding. Site-directed mutagenesis of the sulfate-binding protein, which has two salt-bridges, to convert Glu15 and Asp68 to the corresponding amide side-chains [50] resulted in a large increase in the dissociation rate constant with little change in the affinity for sulfate. This is consistent with a role of these salt linkages in stabilising the closed conformation.

Kinetic studies on DctP have led to some general insights into the ligand-binding mechanism of ESRs. Nevertheless, as emphasised above, although apparently typical in terms of its structural and general biochemical properties, DctP is somewhat distinct with regard to the stability of the closed unliganded conformation. An important question is whether this is restricted to DctP, related entirely to the nature of the ligands involved, or whether the mechanism will prove to be generally applicable to binding proteins which are specific to TRAP transporters. This can only be answered by further comparative kinetic studies of these proteins.

4.1.3.2 Energy coupling

At physiological pH, the C4-dicarboxylates malate, succinate and fumarate are all negatively charged (e.g. for malate pK₁=3.40, pK₂=5.11) and thus the anionic forms cannot be accumulated by diffusion or uniport against an opposing membrane potential, inside negative. If ATP is not the energy source, symport with either protons or another cation is the most likely alternative possibility. The overall driving force for the operation of a proton-symport system for a dicarboxylate anion with two negative charges is given by the following equation:

4

where ΔμD²⁻/F is the transmembrane electrochemical gradient of the dicarboxylate, n is the number of protons co-transported, Δψ is the membrane potential, ΔpH is the pH gradient across the cytoplasmic membrane and Z is a term that converts ΔpH into mV. At external pH values at or above neutrality, ΔpH is small and the proton-motive force is dominated by the membrane potential in neutrophilic bacteria [3], including Rhodobacter. Thus, symport with more than two protons is needed to obtain active transport according to Eq. 4. In intact cells of R. capsulatus treated with increasing concentrations of structurally different uncouplers, an excellent correlation has been found between the decrease in membrane potential (measured by carotenoid electrochromism) and the decrease in the rate of succinate transport via the Dct system, in the absence of any change in intracellular pH [8]. Importantly, no correlation was found with the intracellular ATP concentration, which remained above 1 mM at each uncoupler concentration tested [8]. These observations provide evidence for the membrane potential as the driving force for C4-dicarboxylate transport. It must be emphasised that although the transport of C4-dicarboxylates in both Gram-negative (e.g. [51–55]) and Gram-positive (e.g. [56]) bacteria is well known, such systems are ‘classical’ secondary transporters which are ESR-independent [1, 10]. It has proven possible to complement a dctABD deletion strain of Rhizobium meliloti with a plasmid carrying only the known dct genes from R. capsulatus[8]. As the Dct system in R. meliloti is a typical proton-symport system, with DctA the only structural component [51], this indicates that all of the genetic information for a functional transport system in R. capsulatus is encoded at the dct locus already identified.

The orthovanadate anion is a potent inhibitor of E₁–E₂-type ATPases in eukaryotic cells [57]. In a reconstituted system [58] orthovanadate was found to inhibit transport through the ESR-dependent histidine permease, possibly indicating the involvement of a phosphorylated intermediate in the transport mechanism. Succinate transport via the Dct system in R. capsulatus has been shown to be vanadate insensitive while transport of aminoisobutyrate via an alanine transporter is abolished under the same conditions [8]. These findings are consistent with the absence of an ABC protein in the Dct system and suggest that vanadate sensitivity is a useful tool in the investigation of energy coupling in other ESR-dependent transporters.

The data available thus far therefore strongly suggest that the energy coupling mechanism of the Rhodobacter Dct system is distinct from the ABC family of periplasmic transporters and involves proton-symport. This is fully consistent with what is known about the structure of the system as determined from molecular genetic and protein studies (Section 4.1.1). The most desirable way of confirming the energy coupling mechanism would be in membrane vesicles where energisation of transport by either added ATP or illumination (to generate a membrane potential) and its dependence on the presence of DctP could be determined. Unfortunately, attempts to use either chromatophores (inside-out vesicles) or right-side out vesicles to study succinate transport have been unsuccessful thus far (N.R. Wyborn and D.J. Kelly, unpublished observations).

4.2 Other functionally characterised TRAP transporters

4.2.1 A C4-dicarboxylate TRAP transporter in Wolinella succinogenes

W. succinogenes is an obligately anaerobic Gram-negative rumen bacterium that can grow by respiration with fumarate as the terminal electron acceptor, using formate or hydrogen as the electron donor. C4-dicarboxylate transport is a pre-requisite for fumarate respiration, since the active site of fumarate reductase is located on the cytoplasmic side of the inner membrane. W. succinogenes has been shown to possess several distinct C4-dicarboxylate transport systems [59]. Growth by fumarate respiration depends on the operation of a sodium-dependent electroneutral dicarboxylate antiport (exchanging fumarate for succinate) catalysed by either of two systems encoded by the dcuA and dcuB genes. These are homologous to the systems found in E. coli [52, 54]. A dcuA dcuB double mutant was unable to grow by fumarate respiration, but could grow by anaerobic nitrate respiration with succinate as the carbon source [59]. Three adjacent genes were also identified that encode a TRAP transporter, with the gene products showing extensive sequence similarity to the R. capsulatus DctPQM proteins (DctP, 44% identity to R. capsulatus DctP; DctQ, 22% identity to R. capsulatus DctQ; DctM, 49% identity to R. capsulatus DctM). A dctQM mutant, constructed by allelic replacement, was not defective in anaerobic fumarate respiration and also grew by nitrate respiration with succinate as the carbon source. However, the doubling time of this mutant under the latter condition was about 50% greater than that of the wild-type or dcu mutants [59]. This implies that the DctPQM system in W. succinogenes is the major, but not the sole transporter of C4-dicarboxylates for carbon metabolism. The C4-dicarboxylate accumulation ratio of the dctQM mutant was only 50-fold (thought to be catalysed by an additional unidentified uptake system), compared to about 10³ with the wild-type or dcu mutants, indicating that this system can indeed catalyse active dicarboxylate transport, although the substrate specificity was not investigated in detail. No evidence for sodium dependency was found, and the mechanism of transport is likely to be an electrogenic dicarboxylate symport with more than two protons. This system is thus functionally and structurally extremely similar to the R. capsulatus Dct system, and appears to have a similar physiological role.

4.2.2 Binding protein-dependent secondary transport of glutamate in Rhodobacter sphaeroides

The transport of glutamate and aspartate in R. sphaeroides can be mediated by single ABC transport system which utilises two distinct binding proteins specific for glutamate/glutamine and aspartate/asparagine respectively [60]. Although mutants in this system do not grow on glutamate as sole carbon and nitrogen source, growth can be restored by the addition of millimolar amounts of sodium salts to the medium [61]. Glutamate transport in such mutants is dramatically stimulated by sodium ions, with a half-maximal rate at about 25 mM sodium chloride. This sodium-stimulated transport system was specific for glutamate, and was found to have a high substrate affinity (K_t for glutamate of 0.2 μM). Osmotic shock completely inhibited transport, and the addition of a periplasmic fraction to sphaeroplasts restored glutamate uptake, indicating the involvement of a ESR [61]. This protein was purified and shown to bind glutamate with a low K_d and with high specificity. However, transport was not inhibited by vanadate, an inhibitor of ABC systems, but was inhibited by the ionophores valinomycin and nigericin. These data indicate that the proton-motive force drives uptake. In energised membrane vesicles, glutamate transport was strictly dependent on the presence of both binding protein and sodium ions, and was prevented by the addition of the uncoupler FCCP. The ionophores valinomycin and nigericin significantly reduced the transport rate [61].

All of the characteristics of this system show that a binding protein-dependent secondary transporter is responsible for sodium stimulated glutamate uptake in R. sphaeroides. Although it seems likely that this is a TRAP transporter, no sequence information has been reported for any component of the system, so it is not possible to determine its relationship to currently recognised members of the TRAP transporter family. It is also unclear whether sodium is co-transported with glutamate or whether it stimulates transport allosterically.

5 The TRAP transporter family: DctPQM homologues are widespread in bacteria and archaea

5.1 Diversity in gene organisation and subunit composition of TRAP transporters

As a result of the increasing availability of genome sequence information, it has become clear that homologues of the dct gene products are present in a large number of diverse prokaryotes, including representatives of both bacteria and archaea, but not eukaryotes. Some organisms contain only a single recognisable system (e.g. E. coli), while others have multiple systems encoded within their genomes (e.g. Pseudomonas aeruginosa and Bacillus halodurans). Nevertheless, the tripartite nature of TRAP transporters is apparent in virtually all cases. The exception is where orphan dctP-like genes are found. This is discussed further in Section 5.2.1. An analysis of the phylogenetic relationships of a number of these proteins established the three well-defined families of DctP, DctQ and DctM homologues [9]. An updated and expanded list of proteins composing TRAP-transport systems is presented in Tables 1–3, and their characteristics are discussed below.

Table 1

ESRs for TRAP transporters

Organism	Name of ORF	Size (residues)	Database and accession number
(a) Members of the DctP family.
	Actinobacillus actinomycetemcomitans	ORF539	328	Contig0268b
	Actinobacillus actinomycetemcomitans	ORF942	330	Contig0423
	Bacillus halodurans	BH0701	341	trBAB04420
Bacillus halodurans	BH2673	348	trBAB06392
Bacillus halodurans	BH2750	348	trBAB06469
Bacillus halodurans	BH3390	353	trBAB07109
Bacillus subtilis	YdbE (o)a	350	spP96600
Escherichia coli	YiaO	328	spP37676
Haemophilus influenzae	HI1028	328	spP44992
Haemophilus influenzae	HI0146	329	spP44542
Haemophilus influenzae	HI0052	328	spP71336
Klebsiella pneumoniae	DctP (frag)	216	trQ9RMM7
Pseudomonas aeruginosa	PA0884	331	trAAG04273
Pseudomonas aeruginosa	PA3779	346	trAAG07166
Pseudomonas aeruginosa	PA4616 (o)	332	trAAG08004
Pseudomonas aeruginosa	PA5167	331	trAAG08552
Ralstonia sp. strain TFD41	DctP	209	gbAF045553
Rhizobium sp. strain NGR234	Y4MM	541 (PQ)	spP55572
Rhodobacter capsulatus	DctP	333	spP37735
Rhodobacter sphaeroides	DctP	337	gbAF005842
Rhodopseudomonas palustris	ORF2661	339	Contigs0058
Rhodopseudomonas palustris	ORF2590	334	Contigs0058
Synechocystis sp. strain PCC6803	SLL1314 (o)	369	spP73589
Thermotoga maritima	TM0322	336	trQ9WYF8
Treponema pallidum	TP0957	342	trO83923
Vibrio cholerae	VC1779	321	trAAF94928
Vibrio cholerae	VC1929	332	trAAF95077
Wolinella succinogenes	DctP	329	spQ9Z4J3
(b) Members of the SmoM subfamily
	Anabaena sp.	ORF3386	364	C328c328
	Bordetella bronchiseptica	ORF8706	364	Contig2488
	Bordetella pertussis	ORF3287	364	Contig842
Pseudomonas sp. DJ-12	FcbT1	326	trAAF16407
Rhodobacter sphaeroides	SmoM	365	gbAF018073
Rhodopseudomonas palustris	ORF2587	363	Contig0058
Synechocystis sp. strain PCC6803	SLR1952 (o)	370	spP74100
Vibrio cholerae	VC0488 (o)	359	trAAF93661
Vibrio cholerae	VC1273	401	trAAF94432
(c) Members of the TAXI family.
	Aeropyrum pernix	APE2136	326	trQ9YA03
	Aeropyrum pernix	APE2545	369	trQ9Y8T8
	Archaeglobus fulgidus	AF0467	355	trO29783
Archaeglobus fulgidus	AF0635	330	trO29620
Archaeglobus fulgidus	AF0988	347	trO29274
Bacillus halodurans	BH0601	346	trBAB04320
Bacillus halodurans	BH2937	346	trBAB06656
Bacillus halodurans	BH2947	335	trBAB06666
Deinococcus radiodurans	DR1649	319	trQ9RTV5
Halobacterium sp. NRC-1	Imp	318	trAAG19063
Pseudomonas aeruginosa	PA5545	319	trAAG08930
Rhodobacter capsulatus	ORF2210	471	Contig P
Rhodopseudomonas palustris	ORF0570	315	Contigs0053
Rhodopseudomonas palustris	ORF0352	386	Contigs0051
Rickettsia prowazekii	RP197 (o)	313	trG71730
Synechocystis sp. strain PCC6803	SLL1037	491	trS74613
Vibrio cholerae	VCA0144	351	trAAF96057
Vibrio cholerae	VCO430	328	trAAF93603
(d) Other proteins
	Synechocystis sp. strain PCC6803	SLL1104	296	spP74223
	Abbreviations: sp SWISS-PROT, tr TREMBL, gb GenBank.

^a(o) Indicates a TRAP-transporter gene that is not present in an operon with other genes encoding a complete TRAP transporter. (frag) Indicates that only the incomplete sequence is in the databases. (PQ and QM) Indicate fusion of homologues of DctP/DctQ and DctQ/DctM, respectively.

^bSequences from uncompleted genomes were identified in IGwit (http://wit.integratedgenomics.com/igwit/) and confirmed by using them in PSI-BLAST searches to identify homology to known TRAP-transporter components.

Table 2

The DctQ family

Organism	Name of ORF	Size (residues)	Database and accession number
Actinobacillus actinomycetemcomitans	ORF2353	166	Contig0423
Anabaena sp.	ORF3384	184	C328c328
Bacillus halodurans	BH2672	183	trBAB06391
Bacillus halodurans	BH3392	175	trBAB07111
Bacillus halodurans	BH0702	177	trBAB04421
Bordatella pertussis	ORF3288	209	Contig842
Bordatella bronchiseptica	ORF8707	209	Contig2488
Deinococcus radiodurans	DR1650	239	trQ9RTV4
Escherichia coli	YiaM	157	spP37674
Haemophilus influenzae	HI0147	633 (QM)	spP44543
Haemophilus influenzae	HI0051	165	spP44484
Haemophilus influenzae	YiaM	161	spP44994
Klebsiella pneumoniae	DctQ	136	trQ9RN75
Pseudomonas aeruginosa	PA0885	212	trAAG04274
Pseudomonas aeruginosa	PA3780	164	trAAG07167
Pseudomonas aeruginosa	PA5168	210	trAAG08553
Ralstonia sp. strain TFD41	O85149	185	gbAF045553
Rhizobium sp. strain NGR234	Y4MM	541 (PQ)	spP55572
Rhodobacter capsulatus	DctQ	227	trO07837
Rhodobacter sphaeroides	DctQ	228	trAF005842
Rhodopseudomonas palustris	ORF3081	194	Contigs0058
Rhodopseudomonas palustris	ORF3107	198	Contigs0058
Synechocystis sp. strain PCC6803	SLL1104	179	spP74225
Vibrio cholerae	VC1274	177	trAAF94433
Vibrio cholerae	VC1778	173	trAAF94927
Vibrio cholerae	VC1928	232	trAAF95076
Wolinella succinogenes	DctQ	170	spQ9ZEJ3
For legends see Table 1.

Table 3

The DctM family

Organism	Name of ORF	Size (residues)	Database and accession number
Actinobacillus actinomycetemcomitans	ORF540	616 (QM)	Contig0268
Actinobacillus actinomycetemcomitans	ORF1598	433	Contig0423
Aeropyrum pernix	APE2542	560 (QM)	trQ9Y8U1
Aeropyrum pernix	APE2133	731 (QM)	trQ9YA06
Anabaena sp.	ORF3385	445	C328c328
Archaeglobus fulgidus	AF0989	643 (QM)	trO29273
Archaeglobus fulgidus	AF0636	679 (QM)	trO29619
Archaeglobus fulgidus	AF0466	684 (QM)	trO29784
Bacillus halodurans	BH2671	426	trBAB06390
Bacillus halodurans	BH0703	434	trBAB04422
Bacillus halodurans	BH3391	422	trBAB07110
Bacillus halodurans	BH2945	656 (QM)	trBAB06664
Bacillus halodurans	BH2938	654 (QM)	trBAB06657
Bacillus halodurans	BH0602	652 (QM)	trBAB04321
Bordatella bronchiseptica	ORF8708	552	Contig2488
Bordetella pertussis	Q45366	389	trQ45366
Bordetella pertussis	ORF3289/90	450	Contig842
Deinococcus radiodurans	DR1653	535	trQ9RTV3
Escherichia coli	YiaN	424	spP37675
Haemophilus influenzae	YiaN (HI1029)	425	spP44993
Haemophilus influenzae	HI0050	401	spP44483
Haemophilus influenzae	HI0147	633 (QM)	spP44543
Halobacterium sp. NRC-1	VNG0537C	904 (QM)	trAAG19061
Klebsiella pneumoniae	DctM	236 (frag)	trQ9RN74
Pseudomonas aeruginosa	PA5169	427	trAAG08554
Pseudomonas aeruginosa	PA0886	427	trAAG04275
Pseudomonas aeruginosa	PA3781	426	trAAG07168
Pseudomonas aeruginosa	PA5544	674 (QM)	trAAG08929
Rhizobium sp. species NGR234	Y4ML	419	spP55571
Rhodobacter capsulatus	DctM	440	spO07838
Rhodobacter capsulatus	ORF2212	686 (QM)	Contig P
Rhodobacter sphaeroides	DctM	443	gbAF005842
Rhodopseudomonas palustris	ORF0333	688 (QM)	Contigs0051
Rhodopseudomonas palustris	ORF0518	706 (QM)	Contigs0053
Rhodopseudomonas palustris	ORF2406	631	Contigs0058
Rhodopseudomonas palustris	ORF2442	627 (QM)	Contigs0058
Rhodopseudomonas palustris	ORF2542	441	Contigs0058
Salmonella typhimurium	YgiK	435	spP40800
Synechocystis sp. strain PCC6803	SLL1103	445	spP74224
Thermotoga maritima	TM0324	427	trQ9WYF9
Treponema pallidum	TP0958	632 (QM)	trO83924
Vibrio cholerae	VC0429	854 (QM)	trAAF93602
Vibrio cholerae	VC1275	427	trAAF94434
Vibrio cholerae	VC1777	427	trAAF94926
Vibrio cholerae	VC1927	453	trAAF95075
Vibrio cholerae	VCA0146	721 (QM)	trAAF96059
Wolinella succinogenes	DctM	415	trQ9ZEJ2
For legends see Table 1.

Fig. 4 illustrates the various arrangements of genes encoding components of TRAP-transport systems. The two most common organisations are those that are found in the archetypal dctPQM operon of R. capsulatus and the yiaMNO operon in E. coli (Fig. 4). In these operons the three components of the TRAP transporters are encoded by separate genes, the arrangements differing in the position of the gene encoding the ESR component. There are also examples of TRAP-transport systems where two of the three proteins are fused. In Rhizobium sp. strain NGR234, the genes encoding a TRAP-transport system consist of a fusion of the dctPQ homologues into a single ORF, with a separate dctM homologue [62]. This type of fusion appears to be rare. In contrast, in one of the three TRAP systems of Haemophilus influenzae, the two integral membrane components are encoded by a single gene (dctQM), while the putative ESR is encoded by a separate gene. Fusions of the Q and M homologues appear to be relatively common, but we could not identify any examples of TRAP systems where all components of the tripartite transporter are encoded by a single gene or examples where homologues of dctP and dctM are fused. As noted by Rabus et al. [9], the single TRAP system in Treponema pallidum might have two ESRs; upstream of a dctP homologue is a second gene that is not homologous to dctP, but encodes a similar sized hydrophilic protein with a potential signal sequence [9] (Fig. 4). We found a single case of divergent transcription of the genes encoding a TRAP transporter. In this example, from one of a number of loci encoding TRAP transporters in Actinobacillus actinomycetemcomitans, the dctP homologue is divergently transcribed from the dctQ and dctM homologues. The intergenic region spans ∼200 bp, so has the potential to contain elements for the co-regulation, presumably co-activation, of these two genes (Fig. 4). Interestingly, in some cases the binding protein gene is convergently transcribed with the genes encoding DctQ and DctM homologues. This arrangement is often found in those systems with a SmoM (mannitol-binding protein) homologue. As discussed in Section 5.2.2, there is a distinct family of ESRs that are invariably associated with genes encoding fused DctQM membrane subunits. Again, the position of the gene encoding these potential binding proteins can vary in TRAP-transporter encoding operons (Fig. 4). There is also an example of a TRAP-transporter locus from Synechocystis sp. PC6803 that encodes a homologue of a well-defined glutamate-binding protein (GlnH) downstream of dctQ and dctM homologues [63] (Fig. 4).

Figure 4

Organisation of genes encoding TRAP transporters in bacteria and archaea. The genes are centred on the dctQ and dctM homologues found in all the TRAP systems (coloured blue and cyan, respectively, and often fused into a single gene). Genes encoding the binding proteins flank these genes and those homologous to dctP are shown in red. The gene shown in red dots in the T. pallidum operon may be a second ESR for this TRAP system. In a number of systems, the binding protein is not a DctP family member, but is related to an immunogenic protein first identified in B. abortus[70]. These proteins constitute a new family of ‘TAXI’ (TRAP-associated extracytoplasmic immunogenic) proteins (encoded by genes coloured gold). In other cases, the binding protein is clearly homologous to a functionally characterised protein e.g. GlnH (gene coloured green). In a few systems, an additional gene (blue spots) is present, the function of which is unknown (see text).

Finally, we have found that an extra gene is sometimes associated with TRAP-transporter encoding operons in some organisms (Fig. 4). However, it is not known if this small hydrophilic protein forms an extra subunit of the transporter or has an unrelated function. Thus far, five examples have been found and in the cases of B. halodurans and Halobacterium sp. NRC-1, the gene is interposed between two of the genes encoding other TRAP subunits (Fig. 4). The predicted proteins share some, but not extensive, sequence similarity and are not similar to any other known proteins (G.H. Thomas and D.J. Kelly, in preparation).

5.2 ESRs associated with TRAP transporters: phylogenetic relationships

5.2.1 The DctP family

Many TRAP transporters (but not all; see Section 5.2.2) contain DctP homologues (Table 1) which are presumed to function as ESRs for the cognate membrane transport components of the system. These are hydrophilic proteins of similar size to R. capsulatus DctP (about 280–350 amino acid residues), with the exception of the larger DctP-DctQ fusion (Y4mM) found in Rhizobium sp. Strain NGR234 [62]. Many of the homologues identified in the databases are from Gram-negative bacteria, and they generally possess an easily recognisable signal sequence at their N-terminus, consistent with a periplasmic location of the mature protein.

From phylogenetic analyses of known binding proteins, it has been proposed that homologous proteins can be grouped into families which bind chemically similar ligands [14]. While C4-dicarboxylates may be the ligands for many members of the DctP family, particularly those with high sequence similarity with DctP itself, there is evidence that some homologues bind chemically unrelated substrates. The SmoM protein is a DctP paralogue from R. sphaeroides and has been proposed to bind mannitol [64]. There are now several homologues of SmoM in the databases from different organisms (Table 1b), which appear to form a distinct subfamily [9] and suggests that mannitol-transporting TRAP systems may be common.

There is one potential TRAP-transport system in the model organism E. coli, which is encoded within a large operon that encodes enzymes involved in utilisation of l-lyxose [65, 66]. The ESR from this system, YiaO, has been expressed and purified but does not bind C4-dicarboxylates [40] and its ligand may be a pentose sugar or related molecule as suggested by its genetic location. The H. influenzae binding protein HI1028 is closely related to E. coli YiaO, and the local gene organisation and deduced proteins in the two bacteria are very similar, indicating a similar type of molecule is the transport substrate in each case. Indeed, the HI1028/YiaO proteins and their relatives form a distinct grouping within the DctP family [9]. More work is needed to clarify the ligands for these and indeed the majority of the proteins in the DctP family.

In B. subtilis, an orphan member of the DctP family is encoded by the ydbE gene, in the absence of any DctQ or DctM homologues in the genome. This suggests that YdbE is not part of a TRAP transporter, but has another role. The ydbE gene is divergently transcribed from an operon encoding a sensor–regulator pair (YdbF and YdbG, respectively) and a homologue of the DctA C4-dicarboxylate transport protein from Rhizobium leguminosarum (YdbH; 54% amino acid identity with DctA). It was proposed [9] that YdbE might act as an extracytoplasmic C4-dicarboxylate receptor which interacts with the YdbF sensor-kinase, as the latter protein has two predicted N-terminal transmembrane regions with an intervening periplasmic domain. Activation of YdbF would lead to phosphorylation of the regulator YdbG, which in turn could control transcription of the ydbH gene encoding the putative dicarboxylate transporter. This suggested mechanism has been confirmed experimentally [67]. Individual inactivation of each of the ydbEFG or ydbH genes resulted in a growth deficiency on fumarate or succinate (but not malate). Importantly, inactivation of either the sensor–regulator pair or the solute binding protein caused the complete loss of ydbH expression, as measured with a ydbH transcriptional fusion to β-galactosidase. The utilisation of fumarate and ydbH expression was restored in a ydbE null mutant in which ydbFGH were overproduced [67], confirming that YdbE has a role in signal transduction rather than transport. The role of YdbE in this system is thus similar to that of the periplasmic sugar-binding protein ChvE in Agrobacterium tumefaciens, which has been shown to interact with the periplasmic domain of the VirA kinase to activate the regulator VirG [68]. A similar situation appears to exist for an orphan DctP homologue in B. halodurans. Other organisms which possess orphan DctP homologues include Synechocystis sp PCC6803 and P. aeruginosa. In these cases, the local gene organisation does not immediately suggest a similar role to the B. subtilis homologue. However both of the latter bacteria contain one or more complete TRAP systems encoded elsewhere in their genomes. Hence the orphans could be alternative ESRs for these systems which could endow them with a broader substrate specificity. Clearly, there may also be additional roles for orphan DctP homologues which have not yet been identified.

A number of organisms possess multiple TRAP transporters with paralogues of DctP (Table 1). In most cases these proteins are highly divergent, which may be related to their roles in functionally distinct transport systems and is a reflection of their different substrate specificities. For example, as noted above, the H. influenzae HI1028 protein is much more similar to YiaO of E. coli than it is to the two other binding protein paralogues, all the H. influenzae proteins exhibiting only moderate sequence similarity with each other. An exception to this pattern is seen in P. aeruginosa, where there are four DctP paralogues [69], three of which (PA0884, PA4616 and PA5167) not only show very high sequence similarity with DctP itself, but also share 53–73% amino acid sequence identity with each other. It seems likely that these are all C4-dicarboxylate-binding proteins. The fourth paralogue, PA3779, shares only 23–28% sequence identity with the other paralogues in P. aeruginosa, implying that this protein binds a different type of ligand.

A signature sequence for the DctP family of proteins (Table 4) has been derived from a conserved gap-free portion of a multiple sequence alignment of 12 homologues [9]. This sequence can be used to identify potential new members of the DctP family.

Table 4

Signature sequences of the three families of TRAP-transporter proteins [9]

Family	Signature sequence
DctP	[SANF]-[RKNSQ]-x(2)-[LIV]-x(2)-[LIVWPF]-[ADES]-[DHE]-[LIVMF]-x-G-[LQM]-[KRD]-[LIVM]-R-[LIVTA]
DctQ	[LIVC]-F-x-[WLY]-[LIVMF]-x(2)-[LIVFG]-G-[AI]-[LIVAS]-[LIVY]-x(6)-H-[LIVM]-x-[LIV]
DctM	[LIVM]-[LIVMT]-A-[LIV]-P-[FL]-[FY]-[LIVMA]-x(2)-[GS]-x(2)-[LM]-x(2)-[GST]-x(2)-[GAST]-x(2)-[LIVM]

5.2.2 Other types of ESR associated with TRAP transporters

Since the study of Rabus et al. [9], a significant number of additional microbial genome sequences have been released and it is now clear that there is an additional family of putative ESRs associated with TRAP transporters. These proteins are hydrophilic, about 300–400 residues in length and contain N-terminal signal sequences; properties expected of ESRs. As yet, no functional information is available and they are currently annotated simply as ‘immunogenic proteins’, as the first example was identified as a 31 kDa cell surface protein from Brucella abortus[70]. Although the gene encoding this protein was cloned from B. abortis, the flanking DNA was not characterised and the function of the protein has not been determined. As we also have no functional information about these proteins we have called them TAXI proteins for ‘mmmunogenic protein’. Three members of this family from Archaeoglobus fulgidus were noted by Rabus et al. [9] who identified some sequence similarity to the E. coli glutamine-binding protein. We have now identified almost 20 similar proteins (Table 1c) and they appear to form a discrete subfamily of ESRs (G.H. Thomas and D.J. Kelly, unpublished data). The vast majority of TAXI proteins are components of TRAP transporters, but, as with the DctP family, there are a few examples of TAXI proteins that are not genetically linked to genes encoding other TRAP-transporter components. Here the TAXI protein may couple to another TRAP-transport system or have a distinct function. It is also clear that TRAP transporters that contain a TAXI protein always have a fusion of the two integral membrane subunits. There is perhaps a single exception of this association as the TAXI-containing TRAP transporter from Deinococcus radiodurans contains separate genes for the dctQ and dctM components, however they are split by two genes for transposases, so this system might not be functional, or this insertion might have forced the separate translation of these membrane components.

Interestingly, all of the archaeal TRAP systems thus far identified are of this TAXI/DctQM-type, but such systems are also found in a diverse range of bacteria, often when the same bacterium also has DctP/Q/M-type transporters as well. This suggests that this is a more ancient form of the TRAP transporter, the DctP/Q/M system having evolved later in bacteria for functions including dicarboxylate transport, however there is as yet no experimental evidence suggesting a function or substrate for these TAXI/DctQM-TRAP systems.

There is one example of a TRAP transporter that is very likely to function in transporting amino acids. This is the system from the cyanobacterium Synechocystis sp. PCC6803 [63]. The Sll1102 and Sll1103 proteins are clear homologues of DctQ and DctM, respectively, while Sll1104 is a homologue of GlnH, a glutamine-binding protein This GlnH/Q/M system very probably transports amino acids such as glutamine and glutamate. This is a distinct protein from the TAXI family of ESRs and is also found in TRAP transporters where the DctQ and DctM homologues are separate proteins.

The occurrence of diverse types of ESRs associated with TRAP transporters not only implies that these systems can transport structurally unrelated solutes, but also suggests such proteins have been specifically recruited to the membrane components during evolution.

5.3 The DctQ family

DctQ homologues are invariably present in TRAP transporters [8, 9], either as discrete subunits or as domains fused to either of the other TRAP proteins, and so far have not been found in any other type of transport system. In the R. capsulatus Dct system, DctQ is a small integral membrane protein (26 kDa) which is essential for C4-dicarboxylate transport (see Sections 4.1.1 and 4.1.2). Compared with several homologues, R. capsulatus and R. sphaeroides DctQ proteins have a larger periplasmic loop between the first and second transmembrane helices and also possess a significant hydrophilic C-terminal extension, which has been shown to be located in the cytoplasm ([38] and Fig. 3). These regions may thus have a role specific to the Dct system in Rhodobacter.

From a multiple sequence alignment of eight members of the DctQ family, a signature sequence was derived [9] that can be used to identify new members of the family (Table 4) and a list of proteins homologous to R. capsulatus DctQ are presented in Table 2. However, the DctQ family is the most divergent of the three types of TRAP-transporter proteins in terms of sequence similarity, and there are some examples of proteins that are encoded in gene clusters along with genes for the other TRAP proteins, which have little or no sequence similarity to DctQ itself or other clearly defined members of the family. Nevertheless, the hydropathy profiles of such proteins are usually predicted to contain four transmembrane helices and are of a similar size to bona fide DctQ homologues (Table 2), suggesting that they are functionally equivalent [9]. Hence, although some of these proteins are not sequence homologues of R. capsulatus DctQ, we have included them into the DctQ family as we consider it likely that they are structurally and probably functionally similar to R. capsulatus DctQ.

The function of DctQ in the transport process is unknown, but some possible roles could include (i) mediation of interactions between the DctP and DctM components, perhaps by acting as an ‘anchor’ for DctP (ii) a chaperonin or assembly factor to stabilise DctM in the membrane, or (iii) participation in energy coupling to solute transport. At present there is little experimental evidence for any of these possibilities. The fact that the DctQ family is so divergent and that only fusions between the P and Q or Q and M homologues in TRAP transporters have been found, (not P–M fusions) has been suggested to be indicative of a role for DctQ homologues in mediating interactions between the P and M homologues [9]. However, attempts to express the R. capsulatus dctM gene in E. coli were not successful unless the dctQ gene was also co-expressed, when both proteins were correctly localised to the cytoplasmic membrane (N.R. Wyborn and D.J. Kelly, unpublished data), suggesting a role in DctM membrane insertion. As DctQ homologues are apparently a unique feature of TRAP transporters, elucidation of their functional role should provide important insights into the mechanism of these systems.

5.4 The DctM family: relationships with secondary transporters

5.4.1 DctM homologues

The translocation of substrate across the cytoplasmic membrane is thought to be accomplished by the DctM homologue in TRAP transporters. These proteins are predicted to contain multiple transmembrane helices and in many cases their hydropathy profiles suggest the presence of 10–13 membrane-spanning regions, with a particularly hydrophilic loop often dividing the protein into two halves [9]. This topology is similar to many typical membrane transport proteins in other families which often have a 12-transmembrane helix organisation [1, 10, 71]. The sizes of the DctM homologues are generally in the region of 390–450 amino acids in length (Table 3). However, some are considerably larger at over 600 residues due to a non-homologous extension of 150–200 residues at the N-terminus. Examples include HI0147 of H. influenzae and TP0958 of T. pallidum amongst eubacteria and all three DctM paralogues in the archaeon A. fulgidus. Indeed, all archaeal members of the DctM family so far identified appear to have this two-domain structure. In the case of HI0147, the non-homologous N-terminal domain has significant sequence similarity with DctQ proteins [9], and it is clear that this represents a QM fusion protein. In the other cases the sequence divergence is considerable (a characteristic of the DctQ family as noted above) but the sizes of these domains and the presence of an average of four predicted transmembrane helices strongly suggests similar types of fusions have occurred during evolution. It is thus clear that although some TRAP transporters appear to contain only one large membrane protein in addition to an ESR, the tripartite nature of these systems is still preserved in their domain structure, contrary to a suggestion in a recent review [28].

When sequence similarities between DctM homologues are analysed, it is apparent that these proteins have diverged least compared with the DctP and DctQ families [9]. Determination of the mean sequence identity between pairs of a selection of TRAP proteins in the three families gave values of 35% for DctM, 27% for DctP and 20% for DctQ comparisons [9]. It is thus much easier to recognise new DctM homologues in genome sequences than members of the other TRAP protein families. From a multiple sequence alignment of 11 homologues, a DctM family specific signature sequence has been derived [9] which can be used to identify new members of the family (Table 3). However, the relationship between the DctM family and other known families of membrane transporters has been more difficult to determine. Sensitive PSI-BLAST comparisons of a large number of known transport protein families has provided some evidence that at least 11 of these, including the DctM family, may have a common ancestry [9]. Many of these families catalyse the transport of organic or inorganic ions and have been proposed to be members of an ‘ion transporter’ (IT) superfamily [9, 10]. Interestingly, the DctM family and the DcuC family share a similar motif (LFPVLVxLGIDP) as identified by MEME and MAST programs [9]. DcuC is a secondary dicarboxylate transport protein first identified in E. coli, and which has been functionally characterised as a succinate efflux protein important in fermentative growth on glucose [72, 73], but the significance of this shared motif is unclear. These in silico analyses suggest that DctM has a specific, if weak, relationship to secondary transporters. It is also significant that one type of transporter in the IT superfamily, the arsenite efflux protein ArsB, can operate either in an ATP-dependent mode in concert with ArsA or as a pmf-driven pump in the absence of the ArsA subunit [74]. Thus the ArsB and DctM members of the IT superfamily can be viewed as able to function in association with additional or auxiliary proteins which modify their transport characteristics in terms of energy coupling or solute reception respectively.

5.4.2 Relationship between the DctM and DedA families

Limited sequence similarity is apparent between members of the DctM family and a large group of proteins of unknown function, the DedA family. DedA is encoded by a gene that was originally identified in the purF-hisT region of E. coli[75] but it is now known that DedA-related proteins are encoded in the genomes of a wide range of both Gram-positive and Gram-negative bacteria. They are all of 140–180 amino acids in length and are predicted to contain four or five potential transmembrane helices by hydropathy analysis. An obvious possibility is that DedA family proteins are membrane transporters that share a common ancestry with the DctM family, but functional information is only available for one DedA family protein [76] which confers 4-hydroxybutyrate utilisation when cloned in E. coli. However, it is not clear whether this protein is involved in 4-hydroxybutyrate degradation or transport. The similarity between the DedA and the DctM families is located in the central portion of DctM and related proteins, covering about 150 residues, i.e. the entire length of the DedA proteins. Clearly, until the physiological role of DedA proteins is elucidated, the physiological significance of the relationship between the two families cannot be evaluated.

5.5 Clues to the function and mechanism of TRAP transporters from genomic data

Examination of the context of genes encoding TRAP transporters can often yield hints to their possible physiological functions. In a few cases genes encoding TRAP transporters are linked to regulatory proteins. This type of organisation is seen in the archetypal dctPQM system from R. capsulatus, where the genes for the TRAP-transporter system are divergently transcribed from genes encoding a two-component regulatory system that senses C4-dicarboxylates; the same substrates that are transported by the TRAP system (Fig. 2) [8]. In B. halodurans one of the TRAP transporters is encoded within a group of genes that encode AtoC and AtoS homologues. In other organisms, these two proteins form a two-component system that senses acetoacetate, suggesting that this might be the substrate for this B. halodurans TRAP transporter.

There are numerous examples of genes encoding TRAP transporters that are in potential operons with genes that encode putative enzymes. This organisation strongly suggests the presence of a metabolic pathway that requires a substrate transported by the TRAP transporter encoded in the same operon. In these potential metabolic operons there is often a transcriptional regulator that is divergently transcribed away from the first gene in these operon. Many of these metabolic pathways are likely to be novel and current annotations are insufficient to be able to predict the substrate for the TRAP transporter without experimental data. However, from an in silico analysis we speculate that mannonate, dihydroxyacetone, xylulose, gluconate and rhamnose might be additional substrates for TRAP transporters (G.H. Thomas and D.J. Kelly, unpublished data). Based on an in silico prediction, an attempt to characterise the substrate specificity of the single TRAP transporter in E. coli has been attempted [40]. The genes are within an operon that potentially form a pathway for l-lyxose metabolism [65, 66]. Although the purified ESR was found to bind certain pentose sugars, the binding kinetics suggested these were probably not the physiological substrates [40].

Aside from our in silico analysis, there is also a number of other genes that have recently been proposed to encode TRAP transporters with potentially new substrates [77, 78, 79]. One study reports the presence of a TRAP system in E. coli consisting of gntT, a known gluconate transporter, and two genes upstream of this, yhgH and yhgI[77]. However, yhgH and yhgI are not homologous to either dctP or dctQ and the intergenic distance between yhgI and gntT does not suggest that these three genes form an operon. Therefore we do not think that these three genes encode a TRAP-transport system.

A second report identified genes homologous to dctP and dctM in a larger cluster of genes from Agrobacterium radiobacter that are involved in protocatechuate and 4-sulfocatechol oxidation [78]. Further sequence analysis suggests the presence of a dctQ homologue between the identified dctP and dctM homologues, implying that a complete TRAP-transport system can be encoded by this region (G.H. Thomas and D.J. Kelly, unpublished data). It has been suggested that this TRAP system might be involved in uptake of protocatechuate and 4-sulfocatechol, the substrates for the protocatechuate 3,4-dioxygenase encoded in this region [78]. However, this needs to be confirmed experimentally.

A third potential TRAP system has been identified in the fcb gene cluster of Pseudomonas sp. DJ-12, which is involved in 4-chlorobenzoate dechlorination [79]. Transfer of the fcb region to E. coli confers the ability to convert 4-chlorobenzoate to 4-hydroxybenzoate and it is tempting to speculate that the TRAP system is responsible for uptake of 4-chlorobenzoate [79]. However, as with the previous example, experimental evidence will be required to confirm this hypothesis, although it is worth noting that in both instances the potentially new substrates are structurally rather different from known substrates for TRAP transporters.

A possible clue to the functional mechanism of the TRAP transporters is revealed by the genome sequence of B. halodurans, which contains the greatest number of TRAP-transport systems in any organism to date [80]. This bacterium encodes six complete TRAP transporters within its genome. In stark contrast, the closely related bacterium Bacillus subtilis has no TRAP transporters and only contains an orphan DctP homologue (see Section 5.2). The main physiological difference between these two organisms is the ability of B. halodurans to grow at alkaline pH. It has an obligatory requirement for Na⁺ ions to grow under these conditions and Na⁺ is essential for effective solute transport [80]. This correlation between the use of sodium in energy conservation and the propensity for TRAP transporters makes it tempting to speculate that a common property of TRAP transporters is the dependence on sodium ions for activity. This would agree with experimental work from the potential TRAP system for glutamate uptake in R. sphaeroides[61]. B. halodurans is also interesting as the six TRAP transporters present are a combination of three DctP/DctQ/DctM-type and three TAXI/DctQM-type systems. The organism also has an orphan DctP protein, homologous to that found in B. subtilis, which is linked to a two-component regulatory system and is probably involved in sensing C4-dicarboxylates in the same way (see Section 5.3).

6 Conclusions: the evolution and function of TRAP transporters

Several conclusions may be drawn from this review. Importantly, the discovery of TRAP transporters has made it clear that ESRs can be found in transport systems that are unrelated to the ABC superfamily and thus not necessarily driven by ATP hydrolysis. However, the notion that a secondary transporter can possess a substrate-binding protein was virtually heretical even a few years ago; it is only now that a large number of genome sequences are available that it is obvious how widely distributed this additional class of transporter is turning out to be. What are the implications for understanding the evolution of multi-component transport systems? Pathways for the evolution of both simple and multi-component solute transport systems have been proposed (e.g. see [28, 81] for recent reviews), based on information from the composition and phylogenetic relationships of many transport families. In one view [81], channels gave rise to secondary transporters which in turn gave rise to primary transporters and group translocation systems, involving both gene duplication and evolution of the integral membrane components as well as the recruitment of additional proteins. TRAP transporters appear to be ancient systems, the components of which have probably evolved together but at different rates. In view of the similarities between DctM and other members of the IT family which operate as conventional secondary carriers, the most likely scenario for the evolution of TRAP transporters is that the QM fusion protein, ubiquitous in the archaeal systems and also found in some bacteria, was the basic ancestral transporter subunit onto which the unrelated ESR was added early during evolution [9, 28, 81], to form a multi-component system endowed with high-substrate affinity. Subsequently, in some bacteria, the QM fusion split into two separate proteins. While ESRs of several different types can be found in TRAP transporters, those in the DctP and TAXI families appear to be most commonly associated with these systems. Some of the functions of the ESR in TRAP transporters may well be different, however, to those played in ABC transporters. One function which may be particularly relevant for the association of a ESR with a secondary transporter is to impose directionality on the system, as discussed in Section 3. However, the more general question of the need for TRAP-transporter systems in overall cellular physiology, as opposed to the use of either a conventional secondary transporter or ABC system for a given substrate, is unresolved. There may be clues in the types of substrates which TRAP transporters have evolved to use; it is interesting that no ABC system (to our knowledge) has been identified which transports malate/succinate/fumarate – only secondary transporters for these C4-dicarboxylates have been identified. Thus, DctPQM-type dicarboxylate transporters may offer the only unidirectional, high-affinity alternative to simple cation:dicarboxylate symporters for these substrates. However, this cannot be argued for substrates like glutamate or glutamine for which ABC, TRAP and symport systems are known.

The true diversity of TRAP transporters has yet to be appreciated, since only two or three systems have been studied in any functional detail. More information on the substrate specificity, driving force for transport and physiological roles of the many homologues in the databases now needs to be established.

Acknowledgements

We would like to thank Mark Hamblin, Jonathan Shaw, Mark Behrendt, Jason Forward, Richard Cross, Mark Gibson, Neil Wyborn, Adrian Walmsley and Simon Andrews for their contributions to the work reviewed in this paper. Work in the corresponding author’s laboratory in this area has been supported by the UK Biotechnology and Biological Sciences Research Council by research grants and studentships, including a grant for G.H.T.

1

In this review we will use the term ESR to encompass all binding protein components in ABC and meriplasmic (TRAP) transporters from bacteria and archaea. The frequently used term periplasmic binding protein (PBP) will be only be included when we wish to refer to the subset of ESRs that are found in Gram-negative bacteria. A named binding protein refers to a PBP in which the substrate specificity is known, e.g. sulfate-binding protein.