Topological Analysis of DcuA, an Anaerobic C₄-Dicarboxylate Transporter of Escherichia coli

Abstract

Escherichia coli possesses three independent anaerobic C₄-dicarboxylate transport systems encoded by the dcuA, dcuB, and dcuC genes. The dcuA and dcuB genes encode related integral inner-membrane proteins, DcuA and DcuB (433 and 446 amino acid residues), which have 36% amino acid sequence identity. A previous amino acid sequence-based analysis predicted that DcuA and DcuB contain either 12 or 14 transmembrane helices, with the N and C termini located in the cytoplasm or periplasm (S. Six, S. C. Andrews, G. Unden, and J. R. Guest, J. Bacteriol. 176:6470–6478, 1994). These predictions were tested by constructing and analyzing 66 DcuA-BlaM fusions in which C terminally truncated forms of DcuA are fused to a β-lactamase protein lacking the N-terminal signal peptide. The resulting topological model differs from those previously predicted. It has just 10 transmembrane helices and a central, 80-residue cytoplasmic loop between helices 5 and 6. The N and C termini are located in the periplasm and the predicted orientation is consistent with the “positive-inside rule.” Two highly hydrophobic segments are not membrane spanning: one is in the cytoplasmic loop; the other is in the C-terminal periplasmic region. The topological model obtained for DcuA can be applied to DcuA homologues in other bacteria as well as to DcuB. Overproduction of DcuA to 15% of inner-membrane protein was obtained with the lacUV5-promoter-based plasmid, pYZ4.

Escherichia coli can utilize C₄-dicarboxylates as sole carbon and energy sources under both aerobic and anaerobic conditions (7). Aerobically, uptake of C₄-dicarboxylates (fumarate, malate, and succinate) and l-aspartate is mediated by a secondary transporter, designated DctA (14, 18). The corresponding gene, dctA, has been sequenced, and the role of its product in the utilization of C₄-dicarboxylates (and the cyclic monocarboxylate orotate) has been established by complementation studies in Salmonella typhimurium dctA or outA mutants (2, 28).

Uptake, exchange, and efflux of C₄-dicarboxylic acids under anaerobic conditions is mediated by the Dcu systems (K_m for fumarate uptake = 51 μM), which are genetically distinct from the aerobic Dct system (7, 8, 34). Measurements of both C₄-dicarboxylate uptake and exchange have suggested that the Dcu systems are exclusively expressed under anaerobic conditions, activated by the anaerobic activator protein FNR, and repressed in the presence of nitrate. Three independent Dcu systems have been identified, DcuA, DcuB, and DcuC (27, 34). DcuA and DcuB are homologous proteins (36% identical), whereas DcuC is only 22 to 24% identical to DcuA and DcuB. Growth tests and transport studies with dcuA, dcuB, and dcuC single, double, and triple mutants showed that DcuA, DcuB, and DcuC each mediate exchange as well as uptake (27, 34). The triple mutants were almost completely devoid of Dcu activity. The single mutants exhibited no phenotype, but the dcuA-dcuB double mutant displayed marked deficiencies in C₄-dicarboxylate transport and growth by fumarate respiration, suggesting that DcuA and DcuB have analogous and mutually complementary transport functions in the anaerobic uptake of C₄-dicarboxylates (27, 34). The affinities of DcuA and DcuB for C₄-dicarboxylates are similar, except for the lower affinity of DcuA for malate (27).

DcuA and DcuB have 433 and 446 amino acid residues, respectively, and their sequences suggest that they are highly hydrophobic and lack N-terminal signal sequences, which together indicate that they are polytopic inner-membrane proteins (27). A combination of SOAP, Helixmem, hydropathy plot, and “von Heijne positive-inside rule” analyses were used to predict that DcuA and DcuB have either 12 or 14 transmembrane spanning helices and that their N and C termini are located in the cytoplasm or periplasm (27).

The reasons for having three independent Dcu systems in E. coli and their specific roles in anaerobic C₄-dicarboxylate transport are unknown. In particular, the presence of the homologous and apparently mutually redundant DcuA and DcuB systems remains to be explained. Homologues of DcuA and DcuB are present in Serratia marcescens, Haemophilus influenzae (two homologues), Wolinella succinogenes, Helicobacter pylori, and Salmonella typhimurium, suggesting that the DcuA-like transporters are widespread among the proteobacteria. However, members of the DcuA family have no significant sequence similarity to members of other transporter families, indicating that the DcuA-like proteins are a distinct group (1). In order to better understand the roles and properties of DcuA-like transporters and to provide a robust topological model that would facilitate more accurate comparisons between this group and other transporter families, the topological organization of DcuA within the inner membrane has been analyzed with in-frame translational fusions between a series of progressively truncated forms of the dcuA gene and a downstream “reporter gene” (blaM, encoding β-lactamase). The resulting topological model differs from the predicted models and indicates that the DcuA-like transport proteins represent a unique subgroup of the “duo-decimal transporters” (DDTs).

MATERIALS AND METHODS

Media, growth conditions, strains, vectors, and general methods.

Standard genetic procedures were performed (25) with E. coli DH5α (genotype) as the transformation host. E. coli was grown at 37°C in Luria broth (LB; 250 rpm) or Luria agar containing antibiotics where appropriate (ampicillin [Ap], 100 μg ml⁻¹; kanamycin [Kn], 50 μg ml⁻¹). Amino acid sequence similarity searches were performed by using the BLAST program with the nonredundant databases at the Swiss-BLAST interface of the National Center for Biotechnology Information (Bethesda, Md.). Amino acid sequences were analyzed at the Seqnet Computing Facility (Daresbury, Warrington, United Kingdom) with the GCG and EGCG programs (5) (Peter Rice, Sanger Centre, Cambridge, United Kingdom).

The plasmid pGS745 was constructed by cloning the 6-kb SphI-SalI dcuA-containing fragment of plasmid pGS73 (11) into the corresponding sites of pBR322.

Subcloning the dcuA gene into pYZ4.

A 1.3-kb DNA fragment containing the dcuA coding region was PCR amplified by using the dcuA-containing plasmid pGS73 as a template, Pfu DNA polymerase (Stratagene), and two primers—DCUA2, 5′-ccgaattcc²¹²⁹ATGgTAGTTGTAGAACTC²¹⁴⁶-3′; and DCUA4, 5′-ccggatcc³⁴³⁶TGATCATTACAGCATGAAG³⁴¹⁸-3′ (start codon is underlined, mismatches are in small capitals, EcoRI and BamHI sites are in boldface, the NcoI site is in italics, and the coordinates are from Six et al. [27])—designed to introduce flanking EcoRI-NcoI and BamHI sites. The EcoRI- and BamHI-digested PCR product was subcloned initially into plasmid pUC118 (31), generating pGS748. Plasmid pGS767 was then constructed by subcloning the 1.3-kb NcoI-BamHI dcuA fragment of pGS748 into the corresponding sites of the Kn^r plasmid pYZ4 (33) such that the dcuA gene is appropriately positioned downstream of the IPTG (isopropyl-β-d-thiogalactopyranoside)-inducible lacUV5 promoter.

Construction of dcuA-blaM fusions.

Nested 3′ deletions of dcuA were generated by digesting BamHI- and KpnI-treated pGS767 with exonuclease III using the Double-Stranded Nested Deletion Kit (Pharmacia) according to the manufacturer’s instructions. The products were digested with SacI and ligated with the 0.85-kb SmaI-SacI blaM cassette from the plasmid pLH21 (supplied by J. K. Broome-Smith). The ligation products were used to transform competent E. coli TG1, and Kn^r transformants were selected. Kn^r transformants were tested for growth when inoculated at low or high density on Luria agar containing Ap (35 μg ml⁻¹) plus Kn. The MICs of Ap for transformants carrying dcuA-blaM fusions were determined as previously described (33). Plasmids were isolated from Ap^r Kn^r transformants and analyzed by restriction enzyme mapping to determine the approximate positions of dcuA-blaM fusion sites.

Seventeen codon-specific fusions were made by amplifying the desired dcuA fragment by PCR as before, except that codon-specific primers (26-mers, with the EcoRV site-containing sequence ccgatatc at the 5′ termini, and 18 homologous bases at the 3′ termini) were used in place of DCUA4. The 0.4- to 1.3-kb NcoI- and EcoRV-treated PCR fragments and the 0.85-kb SmaI-SacI blaM cassette of pLH21 were coligated into the corresponding sites of pYZ4, and the desired transformants were selected and treated as described above. It should be noted that the PCR-generated fusions contain an Asp-Gly linker derived from codons at the EcoRV-SmaI hybrid site, whereas the exonuclease III-generated fusions contain a linking Gly residue specified by the SmaI half site.

DNA sequencing.

DNA was sequenced by the chain termination method with the Sequenase kit (U.S. Biochemicals, Ltd.). The dcuA-blaM junctions were sequenced with a primer (BLAM1, 5′-ctcgtgcacccaactga-3′) identical to codons 14 to 18 of blaM.

Immunodetection of DcuA-BlaM fusion proteins.

TG1 transformants expressing dcuA-blaM fusions were grown on 0.4% glucose minimal medium plates plus Kn (50 μg ml⁻¹) for ∼36 h. Samples of bacteria were then taken from the plates, resuspended in 1 ml of ice-cold saline, centrifuged (10,000 × g, 10 min), and resuspended in buffer A (100 mM Tris-HCl, pH 6.8; 5% [vol/vol] β-mercaptoethanol; 1% bromophenol blue; 7% [vol/vol] glycerol) to give an optical density at 650 nm of 5. Samples (10 μl) were incubated at 37°C for 1 h (or boiled for 10 min), centrifuged as described above, and fractionated by electrophoresis in a sodium dodecyl sulfate (SDS)–12.5% polyacrylamide gel. The proteins were then transferred to a nitrocellulose membrane (Schleicher and Schuell-Protan BA83) with a Trans-Blot electrophoretic transfer cell according to the manufacturer’s instructions (Bio-Rad). The β-lactamase fusions were detected by using the Bio-Rad protocol (Immuno-Blot Assay Kit) except that a 1/500 dilution of anti-ampicillinase polyclonal antiserum (5 Prime→3 Prime, Inc.) was used as the primary antibody and a 1/30,000 dilution of alkaline-phosphatase-conjugated anti-rabbit antibody (Sigma) was the secondary antibody.

Preparation of membrane fractions.

TG1 transformants were grown at 250 rpm for 16 h in ca. 300 ml of LB plus Kn at 37°C. The cultures were harvested by centrifugation (3,000 × g, 30 min), and membrane fractions were prepared by sucrose density-gradient centrifugation as previously described (21). The membrane fractions were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE). Protein concentrations were determined by a modification of the Lowry method (22).

RESULTS AND DISCUSSION

Overproduction of DcuA.

A 1.3 kb NcoI-BamHI PCR-generated DNA fragment containing the dcuA coding region was cloned immediately downstream of, and with the same polarity as, the IPTG-inducible lacUV5 promoter of the Kn^r expression vector pYZ4 (33), generating the plasmid pGS767 (see Materials and Methods). Transformants of E. coli TG1 carrying pGS767 were analyzed for DcuA overproduction after overnight growth in LB at 37°C in the presence (data not shown) or absence of the inducer IPTG (Fig. 1). A major component (apparent molecular mass, 34 kDa) was present at 10% of total membrane protein (15% of inner-membrane protein). No equivalent component was observed with the control strain, TG1(pYZ4) (Fig. 1), or in whole-cell extracts (data not shown). The “34-kDa” protein probably corresponds to DcuA, which has a predicted molecular mass of 45.8 kDa. The 11.8-kDa (26%) underestimation is consistent with previous reports that integral membrane proteins have higher electrophoretic mobilities than nonmembrane proteins during SDS-PAGE (12, 23, 29). The apparent presence of overproduced DcuA in the inner-membrane fraction suggests that pGS767 can be used for functional overexpression of dcuA. However, measurements of [¹⁴C]fumarate (Sigma) uptake failed to demonstrate any enhanced Dcu activity for strain TG1(pGS767) when compared to the control strain TG1 (data not shown). This is presumably because DcuA is inactive under aerobic conditions (27).

FIG. 1.

SDS-PAGE analysis of membrane proteins produced by TG1(pGS767). Samples (50 μg of protein) containing total-membrane and inner-membrane fractions (see Materials and Methods) were fractionated by SDS-PAGE (12.5%) and stained with Coomassie brilliant blue R250. Prior to electrophoresis, samples were dissolved in SDS-containing loading buffer (see Materials and Methods) and incubated for 1 h at 37°C. Lanes 1 and 2 show total-membrane and inner membrane proteins, respectively, of TG1(pYZ4); lanes 3 and 4 show total-membrane and inner-membrane proteins, respectively, of TG1(pGS767). The positions of molecular mass markers and DcuA (arrow) are indicated.

Construction and analysis of dcuA′-blaM fusions.

To test the predicted topology of DcuA (27), fusions (dcuA′-blaM) were created between a nested series of exonuclease III-truncated dcuA derivatives and a blaM gene encoding a leaderless β-lactamase (see Materials and Methods). The MIC of Ap was determined for all Kn^r Ap^r transformants obtained (Fig. 2). Transformants for which the Ap MIC was 5 to 10 μg ml⁻¹ were presumed to encode fusion proteins in which the BlaM region remains in the cytoplasm or is located in the membrane, whereas those for which the Ap MIC was ≥10 μg ml⁻¹ were considered to encode fusion proteins in which the BlaM segment is exported to the periplasm (2). The junctions of the dcuA′-blaM fusions were defined by sequence analysis (see Materials and Methods), and a total of 49 unique in-frame dcuA′-blaM fusions were characterized. Although the fusion sites were fairly evenly distributed within the dcuA coding region, attempts to isolate fusions in some regions, in particular between codons 140 and 200, were not very successful. The occurrence of such “cold spots” has been noted previously (30). To overcome this problem, 17 directed fusions (Fig. 3) were generated by cloning specific PCR-amplified dcuA fragments (together with blaM) into pYZ4, and the MICs of the resulting constructs were determined as described above.

FIG. 2.

Hydropathy plot of the DcuA protein and properties of DcuA′-BlaM fusion proteins. (A) Hydropathy was calculated by using the algorithm of Kyte and Doolittle (16) with a window of 11 residues (27). Regions of hydrophobicity (labeled 1 through 12) and hydrophilicity (labeled A through K) are indicated. The 12 shaded solid bars show the putative membrane-spanning helices of the 12-spanner model which were all strongly predicted by the SOAP and Helixmem programs (6, 15). The shaded broken bars denote the two additional helices of the 14-spanner model that are weakly predicted by the hydropathy plot. The 10 solid bars show the helices predicted from the DcuA′-BlaM fusion data. (B) Ap resistance of strains expressing DcuA′-BlaM fusion proteins. Vertical lines represent the Ap MICs conferred by 65 unique DcuA′-BlaM fusion proteins. The BlaM fusion point for each of the DcuA′-BlaM proteins is shown on the horizontal axis. Five regions predicted to be in the periplasm are indicated by the numerals I through IV. (C) Linear representations of the topologies of the computer-derived 12- and 14-spanner models for DcuA, together with the 10-spanner model generated by using the fusion data in Fig. 2B. Solid bars show the predicted helices, “C” indicates the putative cytosolic loops, and “P” shows the potential periplasmic loops. The numbers of Arg+Lys residues, the lengths of the extramembranous segments, and the charge biases (Δ) are indicated.

FIG. 3.

Proposed model of the membrane topology of the DcuA protein. Ten transmembrane helices (labeled 1 through 10) are connected by four periplasmic loops (labeled I through IV) and five cytoplasmic domains. The N and C termini are located in the periplasm. The nonspanning, highly hydrophobic regions (5a, 10b, and part of 10a) or weakly hydrophobic regions (4a and part of 9a) are enclosed in shaded solid or shaded broken boxes, respectively. Residue numbers are boxed. DcuA′-BlaM fusion positions are indicated by black circles with numbers denoting the MICs of Ap (μg ml⁻¹). Fusions generated by PCR amplification are indicated by a single asterisk, and the fusion containing the Asp-Gly-(Gly)₄ linker is indicated by two asterisks. Encircled plus signs denote positively charged residues.

Analysis of DcuA′-BlaM fusions.

The DcuA hydropathy profile (Fig. 2A) contains 12 highly hydrophobic regions (regions 1 to 9, 5a, 10a, and 10b) separated by 11 hydrophilic regions (regions A to K). The shaded-solid and shaded-broken bars superimposed upon the hydropathy plot denote the 12 well-predicted membrane spanning α-helices (spanners) and the two (spanners 4a and 9a) tentatively predicted spanners, respectively (27). These observations suggest that DcuA consists of either 12 strongly predicted spanners with the N and C termini in the cytoplasm or 14 spanners with the N and C termini located in the periplasm (Fig. 2C) (27). The distribution of positively charged residues in both models shows good agreement with the positive-inside rule (32), although the charge bias is twice as high for the 14-spanner model (Fig. 2C).

The MICs of strains expressing in-frame dcuA′-blaM fusions identified five periplasmic regions for DcuA (labeled I to V in Fig. 2B), three of which (I, III, and IV) correspond to hydrophilic regions B, G, and I, respectively, in the hydropathy profile (Fig. 2A). Periplasmic loop II includes the hydrophilic region D and the weakly hydrophobic region 4a (Fig. 2A). This implies that region 4a is not a spanner but instead forms a hydrophobic domain within periplasmic loop II. This is not surprising since 4a is not recognized as a potential spanner by the Helixmem and SOAP programs. Periplasmic segment V extends from residue 384 to residue 433 (the last residue of DcuA) and includes hydrophilic region K, the hydrophobic region 10b, and part of the hydrophobic region 10a (Fig. 2C). This leads to several conclusions: (i) the C terminus is oriented towards the periplasm; (ii) region 10b does not contain a spanner; and (iii) there is only one (not two) spanner (designated spanner 10) in the 9a-to-10a region. The finding that hydrophobic region 10b is nonspanning was unexpected and indicates that this region could form a hydrophobic domain within the C-terminal periplasmic segment that may interact with the membrane or another periplasmic portion of DcuA. Alternatively, it is possible that region 10b is indeed membrane spanning in the “natural” DcuA protein and that the C-terminally linked BlaM domain interferes with the folding or membrane insertion of region 10b, resulting in its false location in the periplasm. To test this possibility, a DcuA-Asp-Gly-(Gly)₄-BlaM fusion was generated comprising residues 1 to 433 of DcuA, the BlaM domain, and an Asp-Gly-(Gly)₄ flexible linker. The four additional Gly residues (note that all PCR-derived fusions contain the Asp-Gly dipeptide at the fusion site; see Materials and Methods) were included to provide a spacer region that would tether the BlaM domain and region 10b less tightly, thereby minimizing the potential for an interfering interaction. The MIC for the resulting fusion was high (100 μg ml⁻¹ with the Gly₄ linker, cf. 50 μg ml⁻¹ without the linker; see Fig. 3), indicating that the C terminus is periplasmically located in the presence or absence of the flexible linker. This supports the notion that the hydrophobic 10b region is indeed periplasmic in the natural DcuA protein.

The fusion data were further interpreted by identifying regions where periplasmic loops are separated by a pair of predicted spanners which are themselves separated by a hydrophilic cytoplasmic loop. This revealed three cytoplasmic loops flanked by spanners: 3-C-4, 7-H-8, and 9-J-10 (Fig. 2A). Hydrophobic regions 5 and 6 are likely to be spanners since fusions in these regions gave low MICs suggestive of a cytoplasic location, yet the adjacent hydrophilic regions D and G correspond to periplasmic loops II and III (Fig. 2 and 3). Hydrophobic region 5a must therefore form a hydrophobic domain (that is possibly membrane associated) within the large central cytoplasmic loop that includes hydrophilic regions E and F (Fig. 3). This is surprising given the highly hydrophobic nature of region 5a (Fig. 2A). However, in order to accommodate a spanner in region 5a, a second adjacent spanner would also be required and the fusion data, hydropathy plot, and computer predictions do not support this.

Fusions in the region preceding periplasmic loop I gave MICs of 5 to 10 μg ml⁻¹. These data are difficult to interpret because fusions close to the N terminus are likely to contain insufficient truncated polypeptide to allow meaningful insertion into the membrane (4). However, since the N-terminal region of DcuA contains two strongly predicted and highly hydrophobic spanners, it is likely that hydrophobic regions 1 and 2 are membrane spanning even though spanner 1 is predicted to be of insufficient length to completely span the membrane. Also, since hydrophilic region A is not apparently in the periplasm (and is therefore likely to be cytoplasmic), whereas hydrophilic region B is periplasmic, it is most probable that the N terminus is orientated towards the periplasm as shown in the new model (Fig. 3).

Immunodetection of DcuA′-BlaM fusion proteins.

The β-lactamase approach for investigating membrane protein topology is based on the assumption that levels of Ap resistance exhibited by blaM fusion strains are directly related to the subcellular locations of the BlaM component of corresponding fusion proteins. However, it is possible that Ap resistance reflects the cellular amounts of the fusion protein, rather than the topology. In order to address this possibility, the amounts of DcuA′-BlaM fusion protein present in the 66 dcuA′-blaM strains were compared by anti-β-lactamase Western blot analysis (see Materials and Methods and Fig. 4). Most of the DcuA′-BlaM fusion strains had similar quantities of multiple immunoreactive proteins (data not shown). It was assumed that the proteins with the greatest molecular mass correspond to the full-length fusions, whereas the others represent breakdown products. The instability of topology probe fusions has been documented previously (26, 30). A Western blot analysis of nine representative fusions and two controls is shown in Fig. 4. The observed sizes of the full-length DcuA′-BlaM fusions are similar to the predicted sizes when allowances are made for the abnormal mobility of the DcuA protein. Although the DcuA′-BlaM fusions containing residues 1 to 46, 1 to 110, and 1 to 129 of DcuA conferred high MICs (90 to 115 μg of Ap ml⁻¹), they were present at levels similar to those obtained for the fusions conferring MICs of just 5 to 10 μg of Ap ml⁻¹ (Fig. 4). These examples clearly illustrate that there is no correlation between the level of Ap resistance and the amount of DcuA′-BlaM fusion protein. Furthermore, they strongly suggest that the observed level of resistance is related to the subcellular location of BlaM rather than to the amounts of BlaM in the cell. The Ap^s control strain, TG1(pYZ4), produced a weakly staining immunoreactive unknown protein that appears to be present in all the samples. The Ap^r control strain, TG1(pUC118), produced two closely spaced, strongly staining immunoreactive proteins (molecular masses of ca. 29 kDa) corresponding to the immature and mature forms of β-lactamase encoded by the bla gene of pUC118.

FIG. 4.

Immunodetection of DcuA′-BlaM fusion proteins. Extracts of E. coli TG1 transformants were prepared as described in Materials and Methods. Lanes 1 to 11 contained whole-cell extracts of strains carrying the following plasmids: 1, pUC118 (Ap^r control); 2, pYZ4 (control); 3 to 11, plasmids encoding DcuA′-BlaM fusions containing the following DcuA residues, with the apparent and expected molecular masses of the fusion proteins (in kilodaltons): 1 to 46 (32 and 35), 1 to 84 (33 and 38), 1 to 110 (35 and 40), 1 to 129 (39 and 44), 1 to 160 (42 and 48), 1 to 183 (44 and 50), 1 to 227 (47 and 53), 1 to 282 (52 and 59), and 1 to 433 (60 and 75). The positions of molecular mass markers are indicated.

The DcuA topological model.

The DcuA topological model derived from the hydropathic profile, spanner prediction (Fig. 2A), BlaM fusion data (Fig. 2B) and multiple alignment (Fig. 5) contains 10 spanners arranged as two clusters of five helices separated by an 80-residue cytoplasmic loop (Fig. 3). The N terminus is located in the membrane and orientated towards the periplasm, and the C-terminal region consists of a relatively hydrophobic 50-residue periplasmic tail. The 433 residues of DcuA are fairly evenly distributed among the three subcellular compartments: 120 (28%) residues in the cytoplasm, 134 (31%) residues in the periplasm, and 179 (41%) residues in the membrane. This is a typical feature of bacterial secondary transporters (10). In accordance with the positive-inside rule (32), the distribution of positively charged residues shows a strong bias towards the cytoplasm: 8 Arg and Lys residues in the periplasm and 15 in the cytoplasm, giving a charge bias of 7; Arg and Lys residues represent 12.5% of cytosolic residues but only 6% of periplasmic residues (Fig. 2C). This distribution is better than for the 12-spanner model (12% of cytoplasmic and 11% of periplasmic residues are Arg or Lys), but less favorable than for the 14-spanner model, where 17% of cytoplasmic and 5% of periplasmic residues are Arg or Lys residues (Fig. 2C). The failure of the previously published analysis to correctly predict the topology of DcuA was due to the false identification of two (or four) membrane-spanning helices (Fig. 2A, 2C, and 3). It is not clear how the two strongly predicted spanners could have been correctly identified as being extramembranous given their highly hydrophobic nature and high scores in the Helixmem analysis. However, their incorrect identification highlights the potential unreliability of current computer-based predictions, showing that such predictions are not satisfactory replacements for experimental data and should be treated with caution.

FIG. 5.

Multiple alignment and positions of predicted membrane-spanning helices for the DcuA family. The amino acid sequences (and database reference numbers) are as follows: DcuA from W. succinogenes (e1186590); DcuB1 and DcuB2 from H. influenzae (P44855 and gi1573110); DcuA from H. pylori (gi2313848); and DcuA and DcuB from E. coli (P04539 and P14409). Sequences were aligned by using PILEUP and displayed by using Prettyplot (5). The S. typhimurium oxy-d (U84267) and S. marcescens DcuA (P40684) sequences are not included because they are incomplete (36 and 132 residues, respectively) and are very similar (100 and 91% identity) to E. coli DcuA and DcuB, respectively. The codon corresponding to residue X (position 46) in the H. influenzae DcuB2 sequence contains an extra base which was inserted to generate an in-frame translation in the encoding gene. Regions of sequence similarity are boxed and a conserved, repeated motif (GGIGL) is marked by double-headed arrows. The 10 transmembrane helices of E. coli DcuA (1 through 10) are denoted by solid bars, and the highly hydrophobic (5a, 10b, and 10a) and weakly hydrophobic (4a and 9a) regions are indicated by gray solid bars and gray broken bars, respectively.

Many bacterial transporters have 12 membrane-spanning helices, and such proteins are designated “duo-decimal transporters” or DDTs (13). The topological and sequence similarities of DDTs (10, 19), suggest that they possess similar structural and mechanistic properties (19, 24). A common feature of many DDTs (e.g., AraE, UhpT, and XylE, but not DctA, GltP, LacY, and MelB) is the presence of a large, central cytoplasmic loop between spanners 6 and 7 which divides the 12 spanners into two six-helix clusters. The topology of DcuA conforms to this consensus, except that there are only five helices in each half of the protein (i.e., helices 1 and 12 are absent). This suggests that DcuA could have structural and functional properties similar to those of the DDTs.

The presence of N and C termini in the periplasm is unusual for a membrane transport protein. Other membrane proteins possessing periplasmic termini include leader peptidase (20), F₁F₀ ATP synthase (17), and the rhamnose sugar transport protein RhaT (30). Interestingly, topological analysis revealed that RhaT, like DcuA, possesses 10 spanners. However, it lacks a C-terminal tail, its central loop contains only 25 amino acid residues, and there is no sequence similarity between DcuA and RhaT. Therefore, DcuA and RhaT appear to be structurally distinct.

DcuA family.

Database searches revealed that there are six dcuA homologues (two completely sequenced and two partially sequenced) in other bacteria. Most of these genes, like the E. coli dcuA and dcuB genes, are adjacent to genes encoding enzymes with roles in C₄ metabolism: dcuA of S. marcesens is next to an aspartase (aspA) gene; and one (dcuB1) of the two dcuB genes of H. influenzae and the dcuA genes of W. succinogenes and H. pylori are next to genes (asnB or asnA) that apparently encode a periplasmic l-asparaginase and cytosolic l-asparaginases, respectively. The other dcuB gene (dcuB2) of H. influenzae is adjacent to genes with no known role in C₄ metabolism, and the location of the dcuB homologue (oxd-5) of S. typhimurium has not been reported, but it is likely to be next to the fumB gene, as in E. coli. Thus, of the eight genes encoding DcuA-like proteins, it appears that seven are linked to genes involved in the interconversion of asparagine, aspartate, fumarate, and malate. This is fully consistent with the deduced roles of the E. coli DcuA and DcuB proteins in the transport of C₄-dicarboxylates and supports the notion that DcuA and DcuB homologues from other bacteria have roles in C₄-dicarboxylate transport. It also indicates that those dcuA genes adjacent to asparaginase genes could encode transporters with specificities for asparagine as well as, or instead of, C₄-dicarboxylates. This would be consistent with many previous examples of close linkage between genes encoding transport systems and enzymes that utilize the same substrates.

The six DcuA and DcuB sequences aligned in Fig. 5 have a high degree of sequence similarity (32 to 57% identity), evenly distributed throughout the membrane-spanning and extramembranous sections. There are only eight regions where padding characters are required to optimize the alignment, and these lie in loop regions: four in the central cytoplasmic loop, two between spanners 9 and 10 in the cytoplasmic loop, and one each between spanners 2 and 3 and spanners 6 and 7 in periplasmic loops I and III. The alignment indicates that all six proteins have a similar topological organization. The hydropathy profiles and predicted membrane-spanning helices are likewise very similar, indicating that the topological model produced for E. coli DcuA can be used to predict the topologies of the other family members (Fig. 5). Interestingly, spanner 2 contains a highly conserved and duplicated motif, GGIGL, of unknown function (Fig. 5).

Sequence searches of available databases using either single amino acid sequences or a “profile” (9) generated from the alignment in Fig. 5 failed to detect any other proteins with significant sequence similarity to those in the DcuA family. This contrasts with a previous report suggesting that the DcuA, DcuC, and DctA proteins (34) are homologous. Furthermore, since the protein fusion-derived topology model of the DctA protein of Rhizobium meliloti contains 12 spanners, with C and N termini in the cytosol, no cytoplasmic central loop, and no periplasmic C-terminal tail, it is unlikely that the DcuA and DctA families are closely related.

Conclusion.

Members of the DcuA family of bacterial polytopic transporters are apparently associated with the metabolism of C₄ compounds and are presumed to mediate the exchange, uptake, and export of C₄-dicarboxylates. Their topological organization suggests that they have structural similarities to the bacterial DDT proteins, but there are sufficient differences to warrant classification as a novel subgroup of DTT-like proteins currently containing eight members from six proteobacterial species.