Bacterial Anion Exchange: Reductionist and Integrative Approaches to Membrane Biology

Carriers related to the protein known as UhpT (for the uptake of hexose phosphates) catalyze an obligatory anion exchange. Such porters are termed ‘P_i-linked’, because in each case P_i (inorganic phosphate) is accepted with relatively low affinity while some selected organic phosphate substrate is accepted with high affinity (Maloney et al. 1990). Genetic and biochemical criteria identify three such P_i-linked porters in Escherichia coli and Salmonella typhimurium: UhpT is involved in the transport of glucose 6-phosphate (G6P); GlpT accepts glycerol 3-phosphate as its high-affinity substrate; and PgtP favors phosphoenolpyruvate and triose phosphates. Phenotypic tests point to other important examples in organisms such as Streptococcus lactis and Staphylococcus aureus (Maloney et al. 1990).

The study of UhpT and similar sugar phosphate antiporters has been under way for some time. Indeed, in 1953 a description of the P_i–self exchange reaction in S. aureus established the field of membrane biology in microorganisms (Mitchell, 1953). Somewhat later, work with E. coli focused on the unusual regulation of UhpT, whose expression is controlled by extracellular, but not intracellular, G6P (Winkler, 1966; Island and Kadner, 1993). It is now clear that this regulatory asymmetry reflects a signal transduction. Information transfer is initiated by occupancy of an external site on a membrane receptor (the UhpC protein); the signal then moves via phosphorylations catalyzed by the UhpB protein to a transcriptional regulator, UhpA, whose activity recruits synthesis of UhpT itself (Island and Kadner, 1993; Kadner et al. 1994).

That UhpT emerges once again as a model for transport (rather than regulation) reflects the considerable progress made in the past decade. In particular, biochemical studies have proved that UhpT functions as an anion exchange mechanism (Sonna et al. 1988). These efforts have also provided the tools needed for analysis of UhpT in solubilized and reconstituted systems (Ambudkar and Maloney, 1986; Sonna et al. 1988); that work has established that the monomer is the minimal functional unit (Ambudkar et al. 1990).

A biochemical model of UhpT function has derived from considerations of substrate selectivity and reaction stoichiometry, as noted in S. lactis (Ambudkar et al. 1986), but the model is fully consistent with what is known of other examples of UhpT (discussed in Maloney et al. 1990). Those experiments suggest that UhpT favors monovalent over divalent P_i, yet makes no substantial discrimination between monovalent and divalent forms of G6P. These and other conclusions, along with the observation that antiport of phosphate against sugar phosphate is an electroneutral event, can be easily understood if UhpT possesses a bifunctional active site that accepts two negative charges. More importantly, these charges would be taken either as a pair of monovalent anions or as a single divalent species (Ambudkar et al. 1986; Maloney et al. 1990). This reaction mechanism would therefore resemble that of the ionophore A23187, whose paired carboxyl groups accept either two protons or a single divalent cation. Note that for UhpT this mechanism preserves an electroneutral reaction at the expense of a pH-dependent variable stoichiometry, for at a molecular level substrates will exchange at either a 2:1 or a 2:2 ratio, depending on the nature of the substrate(s) and the pH at either membrane surface (Ambudkar et al. 1986). Perhaps most striking, in the presence of a pH gradient (alkaline inside) one may even imagine that E. coli takes up two sugar phosphate monoanions (2HG6P⁻) from the relatively acidic medium in exchange for a single divalent sugar phosphate (G6P^2-) from the relatively alkaline interior (Fig. 1). Despite its unusual character, this reaction, a masquerade of a proton-linked symport, fits well with the known cell biology of UhpT in E. coli (see Maloney et al. 1990).

The biochemical basis of exchange via UhpT is reasonably well-established, and further insight requires the application of molecular biology. Early progress in this area came from the laboratories of Kadner, Boos and Hong, who provided the amino acid sequences of UhpT, GlpT and PgtP (Friedrich and Kadner, 1987; Eiglmeier et al. 1987; Goldrick et al. 1988), and this information has now been supplemented with the results of gene fusions in UhpT, GlpT and UhpC (Gott and Boos, 1988; Island et al. 1992). The combined data show that these proteins resemble many other secondary carriers in having a hydrophobic core consisting of 12 transmembrane segments (Fig. 2), with cytosolic N and C termini. A more recent analysis also shows that UhpT and its relatives (GlpT, PgtP, UhpC) constitute one of the five clusters of the major facilitator superfamily (MFS), a collection of 67 related carriers that includes examples from both prokaryotes (e.g. LacY, AraE) and eukaryotes (e.g. the GLUT proteins) (Marger and Saier, 1993). As is typical of carriers within this superfamily (Griffith et al. 1992; Marger and Saier, 1993), the 12 transmembrane segments of UhpT are organized as two groups of six segments each, connected by a central cytoplasmic loop, suggesting that the overall structure may have pseudo-twofold symmetry. Although there is no internal evidence of this asymmetry in UhpT, other carriers in the MFS show clear homology between the N-terminal and C-terminal groups of transmembrane segments (Maiden et al. 1987; Griffith et al. 1992), supporting both the idea of symmetry and the suggestion that such carriers may operate as covalent heterodimers (Maloney, 1990, 1994).

These early experiments have answered questions concerning the two-dimensional organization of UhpT (Fig. 2), but to address more complex issues – for example, the mechanisms underlying substrate selectivity and reaction stoichiometry – new approaches are needed. For this reason, we have undertaken a program of cysteine scanning mutagenesis, a strategy in which cysteine is systematically targeted to strategic locations within the protein. This general methodology was used in early studies of bacteriorhodopsin (Altenbach et al. 1990) and is currently being used to probe functional domains in the acetylcholine receptor ion channel (Akabas et al. 1992) and a possible H⁺-coupling region in LacY (Sahin-Toth and Kaback, 1993). In work with UhpT, our immediate objective is to examine new cysteine variants with SH-reactive probes that resemble the natural substrate in size and charge. In this way, we should be able to identify the translocation pathway and locate the residues that G6P encounters while passing through UhpT. With this information in hand, it is almost certain that we will be better able to understand the origins of substrate selectivity and stoichiometry.

As the first step in applying cysteine scanning mutagenesis to UhpT, we asked whether the cysteines normally present in UhpT are essential to function (see Fig. 2). Site-directed mutagenesis was used to replace, one at a time, each of the six UhpT cysteines with an isosteric serine residue (Yan and Maloney, 1993). These variants were then used, along with in vitro restriction enzymology, to obtain a complementary set of derivatives, each containing a single cysteine at one of the normal positions. Assays of transport by both sets of mutants showed that UhpT function was retained at normal or near normal levels in all variants, even in the cysteine-less derivative (Yan and Maloney, 1993).

This analysis of UhpT cysteines shows that this amino acid plays no essential role in antiport, and such findings have considerably simplified the approach to our long-term goal, since manipulation of old cysteines or the addition of new ones can now take place against a neutral background. It may be that bacterial systems are especially favorable for such work, since the same conclusion has been reached for LacY (van Iwaarden et al. 1991) and GalP (cited in Henderson, 1993). By contrast, cysteines may not be so readily eliminated in other settings. There is a disulfide linkage in the membrane sector of the plasma membrane proton pump of Neurospora crassa (Rao and Scarborough, 1990), and the nature of several mitochondrial carriers is profoundly altered by cysteine-directed agents (Dierks et al. 1990a,b).

UhpT-type proteins are inhibited by low concentrations of mercury and mercurials (Mitchell, 1953; Maloney et al. 1990). Using the UhpT variants containing single cysteines, we sought to identify the residues responsible for this phenotype. Such an effort seemed worthwhile, if only because this information, along with use of permeant and impermeant mercurials, might be relevant to confirming the suspected locations of cysteines in UhpT (see Fig. 2).

Studies of inhibition by a lipid soluble agent, p-chloromercuribenzoic acid (PCMB), showed unambiguously that mercurial sensitivity was attributable to reactions at both C143 and C265 (Fig. 2). Note that PCMB is membrane-permeant in its protonated form and, therefore, that its attack on UhpT might occur from either the extracellular or intracellular surface. To distinguish these alternatives, we next used p-chloromercuribenzosulfonate (PCMBS), a hydrophilic membrane-impermeant derivative of PCMB. We anticipated that PCMBS would not affect G6P transport by intact cells. We also expected that studies of everted membrane vesicles would be informative, since a latent sensitivity to PCMBS might become evident.

These arguments were justified by the behavior of C143. PCMBS failed to inhibit transport by intact cells expressing the single-cysteine UhpT variant with cysteine at position 143; nevertheless, G6P accumulation by everted vesicles was substantially blocked (Table 1). This result shows that C143 is exposed to the cytoplasm, and only to the cytoplasm, and verifies the topological model of UhpT in this region (Fig. 2). By contrast, when we tested the UhpT variant containing only C265, PCMBS blocked transport by both intact cells and everted vesicles. This important finding shows that water-soluble, membrane-impermeant PCMBS has access to C265 from both internal and external phases (Yan and Maloney, 1993) (Table 1).

Clearly, UhpT contains a translocation pathway through which G6P moves. Thus, our finding that PCMBS can approach C265 from either membrane surface can be understood readily if this anionic probe moves through UhpT using a pre-existing substrate translocation pathway. To test this possibility in the simplest way, we asked whether G6P could protect C265 against an attack by PCMBS – one might expect this protection to occur if the two anions used the same pathway. Indeed, substrate protection could be readily documented, whether the attack by PCMBS was from the extracellular medium or from the cytoplasmic surface (not shown, see Yan and Maloney, 1993). Such findings strongly suggest that C265 is one of the residues that lines the translocation pathway of UhpT.

The experiments summarized in the preceding section suggest that C265 lies on the UhpT substrate translocation pathway. Equally important, this work shows how one might use targeted cysteine mutagenesis to identify residues that constitute a functional domain. To validate the approach further, we have now begun to implant cysteines into new locations, and one of these new derivatives merits special comment. Starting with the cysteine-less variant of UhpT, we engineered a Ser→Cys mutation at position 199. Analysis of S199C suggests that this variant provides an essential control that justifies extended use of this overall strategy. The S199C derivative is inhibited by PCMBS from the external surface (intact cells), but not from the cytoplasmic surface (everted vesicles) (Table 1). This finding has great practical value, for it suggests that our tacit assumption regarding the polarity of everted membrane vesicles is verifiable by controls using UhpT itself.

Together, these UhpT variants (Table 1) define the experimental limits that should allow identification of residues in the translocation pathway of this membrane carrier. In one case (S199C), the hydrophilic and impermeant PCMBS inhibits only from the extracellular phase; in another instance (C143), PCMBS attacks only from the cytoplasmic phase; whereas in the remaining example (C265), inhibition is found from either surface. It is this last attribute that marks residues lining the translocation pathway.

Fig. 3 describes the population of UhpT cysteine substitution mutants to be examined while holding these experimental criteria in mind. We envision this collection as supporting two kinds of investigations. Because at least one residue on transmembrane segment VII lies on the transport pathway, we plan an initial ‘vertical’ experiment to examine the full length of segment VII. Among other things, this survey should tell us whether the transport pathway involves small or large numbers of residues; moreover, if several such residues are found, it may be possible to use the patterns of reactivity towards PCMBS as an indirect probe of local structure, as has been argued for positional-reactivity in the acetylcholine receptor (Akabas et al. 1992). Early results from this study are encouraging and suggest that the pathway has considerable size; it is also likely that transmembrane segment seven is structured as an alpha-helix, at least between positions 261 and 273.

We anticipate that a ‘horizontal’ experiment will sample a relatively small number of residues on all transmembrane segments. In this case, we should be able to make rapid preliminary decisions about whether a segment does or does not contribute to the transport pathway. At the moment, we have only generated the necessary reagents for this series of trials (Fig. 3); their analysis has not yet begun. On completion of the vertical and horizontal experiments, we may even be able to generate informed models of the tertiary structure of this membrane carrier (Maloney, 1994) and, since UhpT belongs to the major facilitator superfamily (Marger and Saier, 1993), these inferences may be more broadly applicable.

The topics emphasized above are typical of those in the current literature – that is, much work in this field is directed to an understanding of membrane protein structure. Nevertheless, we should not forget that these transporters operate in the world of biology and that they might assume significant new roles in new settings; these ‘emergent’ functions are as equally worthy of study as are issues dealing with protein structure. In the microbial world, this integrative view is exemplified by OxlT and similar carriers that participate in proton-motive metabolic cycles. In effect, these carriers work as proton pumps.

OxlT is an anion exchange carrier (antiporter) found in the Gram-negative anaerobe O. formigenes, a cell that derives all of its metabolic energy from the decarboxylation of oxalate (⁻OOC–COO⁻), a process that yields the simple end-products, formate (HCOO⁻) and CO₂ (Allison et al. 1985). A study of oxalate transport by this cell (Anantharam et al. 1989) identified an anion exchange reaction involving the one-for-one electrogenic antiport of divalent oxalate and monovalent formate. The properties of OxlT, the protein responsible for this exchange (Ruan et al. 1992), suggest that it is a typical membrane carrier. The stoichiometry of exchange and the rate with which it occurs further suggest that OxlT is the major pathway by which oxalate and formate move into and out of O. formigenes.

Although the biochemical characteristics of OxlT are of intrinsic interest (e.g. OxlT has a turnover number higher than that of any other organic substrate carrier), OxlT merits special attention for the manner in which its function is integrated within overall cell physiology – OxlT, in combination with the cytosolic decarboxylase, functions as an entirely new kind of proton pump (Anantharam et al. 1989; Ruan et al. 1992). This ‘indirect’ or ‘metabolic’ proton pump is illustrated in Fig. 4. Note that OxlT transports both the precursor (oxalate^2-) and product (formate⁻) of decarboxylation and that decarboxylation itself consumes a single internal proton. Consequently, each completed cycle of this pathway is accompanied by net inward movement of a single negative charge and the disappearance of a single internal proton. This cycle is the thermodynamic equivalent of an outwardly directed proton pump, one whose stoichiometry is 1H⁺/cycle. Three cycles would extrude 3H⁺ and establish a proton-motive gradient sufficient to drive ATP synthesis by ‘decarboxylative phosphorylation.’

With this prototype in mind, it has been possible to search for similar examples, and in recent years a number of proton-motive decarboxylation cycles have been identified (Table 2) (see also Maloney, 1994; and the summary by Poolman and Konings, 1993). In several cases, there is direct evidence in support of the hypothesis, whereas in other instances the required players are present but suitable experiments are not yet available. I believe it is notable that these systems often show acidic pH optima (K. Abe, unpublished results). For this reason, it will be of interest to consider such cycles in connection with the acid-inducible amino acid decarboxylases found in E. coli (Meng and Bennet, 1992). Decarboxylation-linked proton-motive cycles may be especially well adapted to highly acidic conditions, since these new proton pumps are not limited by the traditional thermodynamic restraints associated with ion transport. That is, if CO₂ is freely disseminated, the cycle becomes functionally irreversible and the sustainable proton-motive gradient will be limited by the transmembrane concentration gradient of the precursor. With sufficiently high internal levels of a scalar partner, cytosolic levels of precursor might become extraordinarily small. Accordingly, these metabolic cycles may function as ‘tunable’ proton pumps.

The proton-motive metabolic cycles described in Table 2 resemble the prototype based on OxlT (Fig. 4). This resemblance is not essential, however, since on theoretical grounds these cycles require only a one-for-one stoichiometry between entry of negative charge (exit of positive charge) and consumption of internal protons (or external hydroxyls). There are many ways this final result might be achieved. In fact, consideration of this principle suggests a role for such cycles in the evolution of primitive systems, perhaps as a way of generating a proton-motive force without requiring complex metabolic or physical structures (Maloney and Wilson, 1993). Note also that the presently known cycles are rather simple, involving as they do only a pyridoxal-or CoA-linked decarboxylation system (Table 2). Again, simplicity is not essential, and more complex examples have been suggested. For example, acetate, formate or bicarbonate can serve as substrates for synthesis of methane. If these precursors enter as anions, the collection of internal reactions leading to methane production can be viewed as the scalar portion of a metabolic proton pump – for each anion that moves inward, the path to methanogenesis ensures that a single internal proton will be consumed (Ruan et al. 1992).

Indirect proton pumps appear to be restricted to bacterial systems, but because the possibility of a wider distribution has not been carefully examined, it is probably too soon for strong generalizations. Most importantly, the role played by OxlT argues that membrane carriers can assume unanticipated emergent properties if their biochemical functions are properly articulated in relation to other aspects of cell function. This principle will surely be found in all cell types, wherever membrane carriers are found.

These studies of anion exchange are supported by grants from the United States Public Health Service (GM24195) and the National Science Foundation (MCB9220823). We wish to thank our colleague, Dr Robert Kadner, of the University of Virginia Medical School, for introducing us to the molecular biology of UhpT.