Membrane topology of the cyanobacterial bicarbonate transporter, BicA, a member of the SulP (SLC26A) family

Abstract

We have completed the first comprehensive transmembrane topology determination for a member of the ubiquitous and important SulP/SLC26 family of coupled anion transporters found in eukaryotes and prokaryotes. The prokaryotic member that we have mapped, namely BicA from Synechococcus PCC7002, is an important Na⁺-dependent bicarbonate transporter that is likely to play a major role in global primary productivity via the CO₂ concentrating mechanism in cyanobacteria. We experimentally determined the topology based on phoA-lacZ topology mapping combined with reference to a range of predictive models based on hydropathy analysis and positive charge distribution. The 12-TMH structure for BicA is characterized by tight turns between several pairs of TMH and it features a prominent cytoplasmically-located STAS domain that is characteristic of the SulP family. A key difference from previous predicted models is that we identify a cytoplasmic loop between helices 8 and 9 where previous models suggested a TMH. This region includes a highly conserved motif that defines the SulP family. The identification of this region as cytoplasmic, rather than transmembrane, has implications for the function and perhaps regulation of SulP family members. This finding is used to reinterpret mutagenesis data relating to highly conserved residues in this region from both plant and human SulP transporters.

Keywords::

• BicA
• bicarbonate transporter
• SulP
• SLC26
• topology
• membrane protein

Introduction

BicA is a low/medium affinity, Na⁺-dependent bicarbonate transporter that is ubiquitous in a wide range of cyanobacterial families and some proteobacteria [1–3]. It was first discovered and characterized in the marine coastal cyanobacterium, Synechococcus PCC7002; in addition, two other cyanobacterial orthologs were identified and characterized by gain-of-function analysis and many more homologs have been identified by bioinformatics [1]. In cyanobacteria, HCO₃^- transporters form an integral part of the CO₂ concentrating mechanism (CCM) where, in conjunction with active CO₂ uptake systems, HCO₃^- is actively accumulated inside the cell, and then used within carboxysome micro-compartments to support optimal rates of photosynthetic CO₂ fixation via the primary carboxylase, ribulose bisphosphate carboxylase-oxygenase or Rubisco [3]. A functional CCM is an obligate growth requirement for all species of cyanobacteria so far examined and it is arguable that the cyanobacterial CCM contributes to primary productivity, especially in oceanic regions where cyanobacteria are estimated to contribute as much as 25% to annual global primary productivity [4,5]. The archetypal BicA transporter from Synechococcus PCC7002 is predicted to be 566 amino acids in length (59.6 kD) and topology predictions indicate it could have up to 10–12 transmembrane domains. The BicA transporter, like all the active CO₂ and HCO₃^- uptake systems in cyanobacteria, is believed to be inactive in the dark, a process that potentially involves protein phosphorylation events [3]. As a prelude to understanding the post-translational or allosteric activation of BicA we have sought to topology-map the structural domains of BicA so that putative cytoplasmic loops involved in regulation can be identified.

Interestingly, BicA is member of the large SulP family that includes the SLC26 family in mammals. SulP members include a wide range of anion transporters or exchangers with diverse substrates and physiological functions in prokaryotic and eukaryotic species. These include sulphate transport into and within plants [6], chloride/HCO₃^- exchange in mammalian epithelial cells, iodide transport into the thyroid and sulphate uptake into fibroblasts [7]. Structural and topological information for this family is extremely limited, with no structure determined for any member and topological predictions suggesting between 10 and 14 transmembrane helices (TMH). Although the sequence identity across the family is low, the distribution of hydrophobic regions is conserved, suggesting a similar structure for all family members. For several of the mammalian members, it has been established that the N- and C-termini are intracellular, indicating that there is an even number of transmembrane helices [8–12]. A characteristic of the SulP family is a conserved C-terminal domain, the STAS domain (Sulfate Transporter Anti-Sigma factor domain) which has been suggested to be involved in interactions with other proteins and to have a role in regulation of transport activity. This is consistent with an intracellular location and for some transporters, interacting partners have been identified [8,9].

It is important to determine topology experimentally as a first step in understanding structure and function of a membrane protein. Identification of intracellular regions paves the way for further studies on potential regulatory regions, which may not be restricted to the STAS domain. One advantage of working with bacterial transporters is the availability of reliable fusion reporter systems for the determination of topology. One such system is the phoA/lacZ reporter system. This dual construct is fused to successive C-terminal truncations of the transporter gene and the activity of both enzymes determined [13,14]. Alkaline phosphatase is active only in the periplasm while β-galactosidase is active only in the cytoplasm so the relative activity indicates the topological orientation of the construct. An advantage of the dual reporter system is that the ratio of the two activities can be calculated and this corrects for any differences in the expression level of different fusion constructs [13]. Fusion protein topology mapping has been used for several transporters for which the structure was determined subsequently and in these cases the topology was confirmed by the structure [14,15]. We therefore aimed to use the phoA/lacZ system to determine the topology of BicA and then fitted our data to a range of predictive models for a best fit. Prior to the present report, no member of the SulP family had been topology-mapped and so the present report includes basic structural information that may be applicable to other members of the SulP (SLC26) family.

Materials and methods

Bacterial strains and growth conditions

Escherichia coli JM109 was used as plasmid host in this study and strains were grown aerobically at 37°C in Luria Broth (LB). Dual indicator plates [13] contained 1% Bacto-tryptone, 0.5% yeast extract, 0.5% NaCl, 1.5% Bacto-agar, 80 mM K₂HPO₄ (pH 7.0), 1 mM IPTG, 80 μg/ml X-phos (5-bromo-4-chloro-3-indolyl phosphate, disodium salt, Research Organics), 100 μg ml^-1 Red-Gal (6-chloro-3-indolyl-β-D-galactoside, Research Organics) and 30 μg ml^-1 of chloramphenicol.

Cloning of BicA into the dual reporter construct

Plasmid pNV1216 [16] harbouring the phoA-lacZ fusion was digested with Bam HI and Xba I to remove the gtrII gene and replaced with the 1.7 kb bicA gene from Synechococcus PCC7002 that was PCR amplified from genomic DNA with forward primer BicA(7002)BamH1(F), and reverse primer BicA(7002)XbaI(R); tcatcacggatccgaaaaatgc and cgcagatcatctagataacccatctc, respectively. The phoA-lacZ reporter fusion was constructed according to Alexeyev and Winkler [13].

Construction of BicA-phoA/lacZ fusions using Exonuclease-III

BicA(7002)/pNV1216 was linearized with the unique restriction sites, NsiI and XbaI. The Promega Erase-a-base kit was used to create a series of BicA-phoA/lacZ fusions according to the manufacturer's recommendations. After ligation, plasmid DNA was transformed into JM109 cells and plated onto dual indicator plates. Plasmid DNA was purified from blue, red and purple colonies, digested with Bam HI and Sac I to confirm deletion size, and sequenced with the PhoSeq primer (cgctaagagaatcacgcaga). Big-dye Sequencing was carried out following the manufacturer's protocol (Applied Biosystems).

Construction of BicA-phoA/lacZ fusions using PCR

PCR was used to generate a number of BicA-phoA/lacZ fusions at desired locations to create BicA-phoA/lacZ C-terminal fusions G⁵⁹, T¹²¹, L³⁴⁰, A³⁶⁹ and I³⁹³ (amino acid position of last residue before reporter). Primers were designed with an XbaI restriction site 3′ of the fusion site to allow in-frame directional cloning into pNV1216. BicA (7002) was excised from pNV1216 using double digests with BamHI and XbaI. The following primer combinations were used for PCR reactions: BicA(7002)BamH1(F) and one of BicA(G59)XbaI(R), BicA(T121)XbaI(R), BicA(L340)XbaI(R), BicA(A369)XbaI(R) or BicA(I393)XbaI(R); tcctctagacccaaagagggcggcaaagaa, cattctagaggtgacgtatttgccgagtt, ccgtctagataaactagccgccccaag, cgttctagaggcccgttttaggaaactc, or gactctagaaatcaagtcaactaagaccgtca, respectively. PCR products were ligated into BamHI/XbaI digested pNV1216 and transformed into JM109 cells, sequenced and then colonies were analysed on dual indicator plates.

Fluorometric assays of alkaline phosphatase and β-galactosidase activities

E. coli JM109 cells containing BicA-phoA/lacZ fusion constructs were grown overnight, diluted 1:10 in fresh LB containing chloramphenicol and induced with 1 mM IPTG. Cells were grown for a further 2–3 h until A₆₀₀ ∼ 0.5, then harvested by centrifugation and permeabilized as described previously [17]. For the β-galactosidase assay, cells were harvested as above, except prior to permeabilization, cells were resuspended in the following buffer (25 mM Tris-HCl pH 7.5, 125 mM NaCl, 2 mM MgCl₂ and 12 mM 2-mercaptoethanol). Stock solutions of 4-methylumbelliferyl β-D-galactopyranoside (4-MUG) and 4-methylumbelliferyl phosphate (4-MUP) were prepared by dissolving in sterile water and DMSO, respectively. The substrates were diluted in reaction buffer immediately prior to experiments. The reaction buffer for alkaline phosphatase assay was: 1 M Tris- HCl pH 8.0, 0.1 mM ZnCl₂ and 100 μM 4-MUP, and for the β-galactosidase assay: 25 mM Tris-HCl pH 7.5, 125 mM NaCl, 2 mM MgCl₂,12 mM 2-mercaptoethanol and 100 μM 4-MUG. Assays were performed using the TKO 100 Mini Fluorometer (Hoefer, Inc., Holliston, MA, USA) following the manufacturers' protocol. Reactions were stopped with 5% trichloroacetic acid (TCA) and the solution clarified by centrifugation. Glycine carbonate reagent (133 mM glycine, 83 mM Na₂CO₃, pH 10.7) was added and fluorescence was measured. Excitation occurs at 380 nm, and 4-MU emits at a wavelength of 460 nm. At least three independent cultures for each construct were assayed. A standard curve was generated with 4-methylumbelliferone (4-MU) at concentrations between 0 and 500 nM allowing the activity of samples to be calculated.

Topology prediction analysis

A number of computer programs available on the Internet, were used to examine the BicA protein sequence for predicted membrane topology. The TOPCONS website (http://topcons.net/index.php) was used to give a consensus prediction of membrane protein topology using several different algorithms: SCAMPI (single sequence mode) [18], SCAMPI (multiple sequence mode) [18], OCTOPUS and TMHMM [19]. A number of other programs were used including MEMSAT3 [20] and TopPred 2 [21].

Results

PhoA-LacZ mapping of BicA from Synechococcus PCC7002

A total of 47 random in-frame fusions were initially identified on dual reporter plates and scored as red (putative cytoplasmic locations), blue (putative periplasmic locations) or purple colonies (putative membrane spanning domains; Table I). Several different topology prediction programs were used to predict topologies for BicA (Table II) and the locations indicated for the initial fusions were compared to the models generated. The results of different programs suggested 10–12 transmembrane helices (TMH). Some predictions (for example those with a periplasmic location for the STAS domain) could be immediately excluded based on the initial data. To distinguish between the remaining predicted topologies, a further five targeted C-terminal fusions of BicA were generated using PCR (see Methods). DNA sequencing was used to verify the location of all fusions generated by both methods.

Table I.  Enzyme activity of BicA dual reporter fusions.

BicA phoA-lacZ fusion point^a	Colony colour^b	AP activity (% of max rate)^c	BG activity (% of max rate)^c	AP:BG ratio	Topology location score^d
G19	Red	0.4	50.6	0.01	cytoplasmic
A32	Blue	46.7	14.8	3.16	periplasmic
G34	Blue	nd	nd	–	periplasmic
A41	Blue	98.3	9.1	10.82	periplasmic
G43	Blue	100.0	34.7	2.89	periplasmic
W46	Blue	65.3	20.5	3.19	periplasmic
F53	Purple	38.0	77.8	0.49	TMH
G59	Red	0.9	24.3	0.04	cytoplasmic
V74	Purple	nd	nd	–	TMH
T84	Purple	4.8	24.5	0.20	TMH
G93	Blue	40.8	68.2	0.60	periplasmic
A97	Purple	29.7	75.6	0.39	TMH
V100	Blue	3.8	2.3	1.65	periplasmic
F107	Purple	0.4	33.5	0.01	cytoplasmic
T121	Red	0.5	38.6	0.01	cytoplasmic
V141	Purple	2.5	50.6	0.05	TMH
Q150	Blue	5.7	5.7	1.00	periplasmic
G154	Blue	31.2	34.7	0.90	periplasmic
N170	Blue	4.4	1.7	2.60	periplasmic
A177	Blue	nd	nd	–	periplasmic
P190	Purple	0.6	36.3	0.02	cytoplasmic
K192	Red	0.6	31.8	0.02	cytoplasmic
T208	Purple	9.4	69.3	0.14	TMH
I210	Purple	24.0	13.3	1.80	periplasmic
A211	Blue	48.0	13.6	3.52	periplasmic
D221	Blue	50.9	85.8	0.59	periplasmic
I225	Blue	14.8	33.3	0.44	periplasmic
V236	Blue	32.1	61.9	0.52	periplasmic
I249	Blue	22.3	83.5	0.27	TMH
V268	Red	0.2	63.2	0.00	cytoplasmic
E277	Red	0.6	100.0	0.01	cytoplasmic
K282	Red	nd	nd	–	cytoplasmic
G288	Red	0.2	13.4	0.01	cytoplasmic
G298	Red	nd	nd	–	cytoplasmic
V311	Red	1.6	17.2	0.09	cytoplasmic
N312	Red	0.2	16.4	0.01	cytoplasmic
G336	Purple	7.2	17.0	0.42	TMH
L340	Blue	44.7	4.1	10.90	periplasmic
L346	Purple	21.7	21.0	1.03	periplasmic
A347	Blue	nd	nd	–	periplasmic
I360	Red	19.3	44.8	0.43	TMH
A369	Red	0.9	68.5	0.01	cytoplasmic
A377	Purple	nd	nd	–	TMH
L385	Blue	nd	nd	–	periplasmic
V388	Blue	87.3	36.4	2.40	periplasmic
I393	Blue	72.0	44.8	1.61	periplasmic
Q415	Red-Purple	1.4	20.5	0.07	cytoplasmic
D426	Red-Purple	nd	nd	–	cytoplasmic
D440	Red-Purple	1.8	40.9	0.05	cytoplasmic
G458	Red-Purple	nd	nd	–	cytoplasmic
E466	Red-Purple	nd	nd	–	cytoplasmic
P561	Red-Purple	0.5	23.6	0.02	cytoplasmic

^aAmino acid position of the residue immediately before the phoA/lacZ reporter. ^bColour of colony as determined on dual indicator plates. ^cThe percentage of AP and BG activity relative to the maximum observed rates of AP or BG; activities were the mean of three independent cultures and SEs were less than ± 14% of the mean (n = 4 or greater). AP max rate = 81.2 nmoles MU/min/OD600 (G43 fusion); BG max rate = 7.76 nmoles MU/min/OD600 (E277 fusion); MU = 4-methylumbelliferone. ^dFusion locations were scored under the following criteria: Blue colonies with AP/BG ratios > 0.45 scores as periplasmic if AP activity was also > 5%; Purple colonies with AP/BG ratios between 0.1 and 2 were scored as TMH located; Red colonies with AP/BG ratios < 0.1 scored as cytoplasmic if AP activity was also < 2.5% of max.

Table II.  Predicted topologies for BicA using available prediction programs.

Prediction program	Number of TMH predicted	Location of N-terminus	Location of loop 8/9 region	Location of STAS domain
TMHMM	12	cytoplasmic	TMH	cytoplasmic
TopPred 2	12	cytoplasmic	TMH	cytoplasmic
OCTOPUS	10	external	TMH	external
SCAMPI-seq	11	cytoplasmic	TMH	external
SCAMPI-msa	12	cytoplasmic	cytoplasmic	cytoplasmic
MEMSAT3	12	cytoplasmic	cytoplasmic	cytoplasmic

Prediction server sites used are mentioned in the materials and methods section.

A collection of 41 fusion constructs was quantitatively assessed for alkaline phosphatase and β-galactosidase activity using a fluorescence enzyme assay (Table I). Assay data were normalized and expressed as ratios to allow for any differences in the expression of fusion constructs (Table I). In theory, a ratio of alkaline phosphatase to β-galactosidase activity greater than 1 should indicate a periplasmic location, while a ratio less than 1 would be consistent with a cytoplasmic location. In practice there are two complications. Firstly, some fusions may undergo cytosolic proteolysis which results in abnormally high β-galactosidase activity when a normally periplasmic C-terminus is released in the cytoplasm [14]. However, the converse does not occur so a high alkaline phosphatase activity is a better indicator of a periplasmic location than β-galactosidase is of a cytoplasmic location. Secondly, TMH-located fusions may be targeted to both the periplast and cytoplast, and this can depend on the protein sequence and on the nature of the fusion protein [14]. These two factors were taken into account to give the scoring parameters described in Table I footnotes, which include colony colour and alkaline phosphatase activity as well as the ratio of alkaline phosphatase to β-galactosidase activities.

Data obtained from the activity ratios of the C-terminal fusions suggested a topology model of 12 transmembrane helices with the N- and C-termini in the cytoplasm (Figure 1). Most predictive models used also take into account the distribution of positive charges, and as shown in Figure 2 our topology model conforms well to the positive-inside-rule [22]. Five cytoplasmic loops could be clearly identified, with each containing at least one fusion where the activity ratio was less than 0.1 and the colony was red (Table I; Figures 1 and 2). Six periplasmic loops were also identified, where at least one fusion resulted from a blue colony with an activity ratio of greater than 1.0 (Table I). Five of these fusions had significant alkaline phosphatase activity (> 30% of the maximum activity). In the remaining periplasmic loop, between putative helices 3 and 4, the only fusion that had an activity ratio greater than 1.0 (V¹⁰⁰) appeared to be poorly expressed, as indicated by the low activity of both enzymes. However, the two preceding fusions had high alkaline phosphatase activity (G⁹³ and A⁹⁷) which is inconsistent with this region forming a cytoplasmic loop and therefore supports the existence of helices 3 and 4 in this location.

Figure 1.  The topology map for the Synechococcus PCC7002 Na⁺-dependent HCO₃^- transporter, BicA. Experimentally determined phoA/lacZ mapping positions (Table I) are shown as black (cytoplasmic), mid-grey (TMH located) or light grey (periplasmic); in the online version these are shown as blue, purple and red, respectively. The inset shows a WebLogo (weblogo.berkeley.edu) representation of the consensus of the ungapped alignment of the 50 residue loop between TMH 8 and 9; all the SulP/SLC26A members used to generate the phylogenetic tree shown in Figure 3 were used and BicA numbering is shown. This region corresponds to the second signature sequence identified by Saier et al. [23] for the SulP family; the most highly conserved subset, NSNKELIGQGLGN (279-291), is highlighted. This Figure is reproduced in colour in the online version of Molecular Membrane Biology.

Figure 2.  A simplified version of the topology map for BicA (Synechococcus PCC7002) showing the positions of positive charged residues and three highly conserved residues that are discussed in the text, E²⁸³ and N²⁹¹, that are implicated in human disease and T⁴⁸⁹, which is a putative phosphorylation site.

Implications of the BicA topology for the SulP family

Bacterial transporters have frequently been found to be good structural models for eukaryotic homologs so we examined other members of the SulP family in the light of the BicA topology. The SulP family is defined by conserved motifs in helices 1 and 2 and in the cytoplasmic region between helices 8 and 9 [23]. Conserved amino acid residues in helices 1 and 2 have been identified as functionally important in the plant SulP transporter, SHST1 [24] but the other conserved region has not been well studied. One feature of BicA that makes correct prediction of its topology difficult is the extremely tight turns between helices. While eukaryotic members of the family are generally 100–200 amino acids longer than bacterial members, at least some of the tight turns, as indicated by spacing on topology predictions, appear to be conserved. We therefore examined different topology predictions (see Methods) for BicA to determine which ones predicted a topology that agrees with the phoA/lacZ fusion data and compared these to predictions for other members of the family. As was found for BicA, a range of topologies were predicted for several eukaryotic members of the SulP family, usually showing 10–12 TMH, depending on the algorithm used.

The only programs that accurately predicted the topology of BicA were SCAMPI-msa and the latest version of MEMSAT3 (Table II). Other programs were generally unable to distinguish some of the most closely connected helices (helices 1–3 and 9–12) and many also predicted a TMH for the region between helices 8 and 9 that we have shown is cytoplasmic. SCAMPI-msa and MEMSAT3 differ from other programs in that they both use a multiple sequence alignment as a basis for prediction. The entered query sequence is first used by each program to generate a multiple alignment, which is then used for topology prediction. This is thought to improve the accuracy of prediction as it takes into account sequence conservation as well as hydrophobicity [20].

A phylogenetic tree was generated for a wide range of prokaryotic and eukaryotic SulP transporters (Figure 3).

Figure 3.  Phylogenetic tree showing the relationship of the cyanobacterial BicA family (30 members) to the Arabidopsis SulP sulphate transporter family (13 members), the human SLC26A family (9 members) and a selection of bacterial BicA-like proteins (13 members). Protein sequences were aligned using ClustalW and nearest neighbour-joining routines in MEGA4 software (www.megasoftware.net). The scale marker represents 0.2 substitutions per residue.

This shows that the cyanobacterial BicA family, the BicA-like bacterial transporters, the plant sulphate transporters and the mammalian SLC26 members form distinct subfamilies. While the tree emphasises the diversity of this family, there are higher levels of sequence conservation in some parts of the protein sequences and this is demonstrated in Figure 4 which shows an alignment between a single representative from each branch of the tree. The same two representatives from the eukaryotic subgroups, SHST1 and SLC26A2, were used as the query sequences for SCAMPI-msa and this resulted in virtually the same topology being predicted as that found for BicA (Figure 4). However, the sequence alignments used by the program to generate the prediction in each case were different, with each being largely confined to members of the same subgroup (results not shown). The TMH predicted by SCAMP-msa are shown in Figure 4, and demonstrate the similarity of the predictions as well as the higher level of sequence conservation for TMH compared to loops, with the exception of the loop between helices 8 and 9 (Figure 1 inset). MEMSAT3 predicted a BicA-like topology for SLC26A2 but found only nine TMH in SHST1. Overall, these results support our suggestion that all members of the family have a similar topology to BicA.

Figure 4.  BicA (Synechococcus PCC7002) aligned to an extensively investigated member of the plant sulphate transporter family (SHST1, Stylosanthes hamata) and a member of the human SLC26A family (DTDST, SLC26A2) that is linked to the disease, diastrophic dysplasia. SCAMPI-msa predicted TMH regions are shaded grey based on separate queries for each protein. The two disease-causing residues, E²⁸³ and N²⁹¹, and the putative phosphorylation site T⁴⁸⁹, are shown.

One of the most striking differences between the different topology models is the location of the region that we have shown forms a relatively large cytoplasmic loop (50 residues from V²⁶⁹ to R³¹⁸) between helices 8 and 9 of BicA (Figures 1 and 2). All programs other than SCAMPI-msa and MEMSAT3 predicted a TMH in the middle of this region (Table II). The experimentally determined topology for BicA is consistent with a cytoplasmic loop in this region (Table I), although its hydrophobicity suggests that it could form a partially re-entrant loop. Such a loop is unlikely to be detected by a fusion approach because part of the loop will be replaced by the fusion protein. A re-entrant loop might only enter the membrane as a folded structure which would require the presence of the entire loop sequence. This region also includes a number of highly conserved residues, as shown on the WebLogo (weblogo.berkeley.edu) representation (Figure 1, inset) derived from the aligned members shown in Figure 3.

Discussion

We have shown that the cyanobacterial Na⁺-dependent HCO₃^- transporter, BicA, has 12 transmembrane helices identified by phoA/lacZ fusion mapping (Figures 1 and 2). The structure and function of BicA is of interest for several reasons. Firstly, BicA has an important physiological role in the uptake of bicarbonate for photosynthesis in cyanobacteria. As a transporter with a high flux rate that contributes to the cyanobacterial carbon concentrating mechanism and which has homologs in all major clades of cyanobacteria (Figure 3), it is likely to have a significant role in photosynthetic productivity on a global scale. Secondly, BicA is a member of a larger family including eukaryotic as well as prokaryotic members – the SulP or SLC26 family. BicA is the first member of this family for which a comprehensive topological analysis has been done and our bioinformatic analyses suggest that the proposed 12 TMH topology may apply across the family.

The physiological role of BicA in the cyanobacterial CO₂ concentrating mechanism has been well characterized but little is known about its regulation. Two types of allosteric regulation appear to exist, although both may involve protein phosphorylation [3]. The first involves inactivation of HCO₃^- transporters (and the CO₂ uptake systems) on the transition from light to dark, presumably to avoid futile cycling (pump and leak) in the dark, while the second, known as ‘fast induction’ manifests as a rapid activation of transporter affinity on exposure to severe HCO₃^-/CO₂ limitation, and is known to be blocked by Ser/Thr protein kinase inhibitors [25]. To gain a greater understanding of these processes, it is important to identify which parts of the protein are intracellular and may, therefore, contain phosphorylation or other regulatory sites. Topological analysis provides this information and therefore acts as a starting point for the identification of putative regulatory regions. An unexpected finding of this study is the existence of a 50 residue intracellular loop between helices 8 and 9. In this respect, our experimentally determined topology differs from predictions from most programs, which have been used to suggest this region formed a transmembrane helix. Our new topology has implications for the functioning and regulation of BicA and for eukaryotic members of the family, where there exists experimental data relating to the function of this region of the protein (discussed below).

An advantage of studying topology in bacterial transporters is the availability of effective dual reporter methods such as the phoA-lacZ system used here. Dual reporter systems have been shown to provide accurate topological information by comparing results with subsequent structures as they become available [15]. A recent study tested the reliability of reporter fusion topology mapping by using this approach to determine the topology of the E. coli ClcA H⁺/Cl^- exchange transporter, for which the structure is known [15]. These authors used phoA and GFP fusions and found that for normal membrane-spanning helices, the results were highly reliable. It was concluded that the dual reporter accurately predicted ‘normal’ TMH although it did not detect the two shorter, and less hydrophobic, helical hairpins seen in the ClcA structure. Of interest to the present study, is that two TMH linked by a tight turn in ClcA which could not be clearly distinguished on the hydropathy plot were distinguished by the fusion data. In BicA, the first three helices and the pairs, 9 and 10, and 11 and 12, are linked by tight turns. The fusion data is consistent with the existence of these 7 helices, even though they are not predicted by all topology prediction programs (Table II). It is possible that some of these helices may include short sections that are non-helical, as has been found for several transporters [26]. This would serve to lengthen these helices, in addition to introducing a less hydrophobic region which may be important for transport function.

Different topologies with 10–14 transmembrane helices have been proposed for different members of the SulP family. Our results confirm the intracellular location of the N- and C-termini found for other members of the family [9,12,27]. There have been some studies providing limited topological information for eukaryotic members of the SulP family, including several that examined the functionality of conserved glycosylation sites that occur after the third predicted transmembrane region. Two studies indicated that these are glycosylated [28,29] and therefore outside the cell (consistent with the BicA topology) but a third found no evidence for glycosylation [30]. Epitope tagging was used to suggest a ten helix structure for prestin in which most TMH were inverted compared to other predictions [30]. In the light of these conflicting data, it is hard to draw definitive conclusions. This is partly because these studies were done on different members of the family, using different techniques, in contrast to our comprehensive analysis of BicA.

Although a topology of 12 transmembrane helices has been predicted for many members of the SulP family, our experimentally determined topology differs from most of these in the positions of helices 9–12. Interestingly, the SCAMPI-msa topology prediction, which is based on a multiple sequence alignment, consistently predicts this structure, regardless of which SulP member is used as the query sequence. However, other topology predictions based on individual transporter sequences commonly miss one or more helices and also often predict a TMH in the cytoplasmic loop region between helices 8 and 9. The sequence previously thought to form helix 9 is clearly identified as cytoplasmic in BicA by multiple fusions in this region (Figure 1). The predicted helices 10 and 11 become 9 and 10 in our model and are therefore inverted relative to earlier predictions. The final region of hydrophobicity is identified by our fusion data as a tight helical hairpin, bringing the C-terminal STAS domain into the cytoplasm, consistent with earlier studies. The orientation of these final four helices in BicA is more consistent with the positive-inside-rule than earlier predicted SulP family topologies (Figure 2).

The sequence of the intracellular region between helices 8 and 9 is highly conserved and forms part of the SulP consensus motif [23]. It is also quite hydrophobic, raising the possibility that it may form a partial re-entrant loop, which would not have been detected by our fusion analysis. Mutational analysis of amino acid residues in this region is consistent with it being involved in a protein/protein interaction, rather than forming an unstructured cytoplasmic loop. Mutations in the human SulP members, SLC26A2 and SLC26A4 are responsible for the diseases diastrophic dysplasia and Pendred syndrome, respectively [7]. Some mutations affect conserved amino acid residues in the sequence between helices 8 and 9 [31–33]. Mutations in some of the homologous residues have been introduced into the plant SulP member, SHST1, and primarily affected folding and trafficking of the mutant transporters in a yeast expression system [34]. The two mutations tested either introduced or removed a charge (affecting the highly conserved residues, E²⁸³ and N²⁹¹, BicA numbering, `s 2 and 4) so it was suggested that they affected interaction with other helices and hence folding. These results are more surprising for a cytoplasmic region of the protein but could be accounted for if the loop is involved in protein/protein interactions. This loop may therefore be functionally important and is a good candidate for an intracellular regulatory region.

In the light of these results, we now hypothesize that this loop region interacts with another part of the BicA protein, either within the membrane, or with the C-terminal STAS domain, which is homologous to a protein/protein interaction domain of the sporulation regulator, SpoIIA [35]. Recent work on ammonium transporters from the Amt family has found that a C-terminal domain interacts with cytoplasmic loops to regulate transporter function [36,37]. This interaction is controlled by phosphorylation of the C-terminal domain. BicA, like most SulP transporters, contains a putative phosphorylation site in the STAS domain at T⁴⁸⁹ (Figure 4). Our topology for BicA therefore raises the possibility that a similar mode of regulation to that found in the Amt family exists. Future mutagenesis studies could test this possibility as well as examining the potential functional and regulatory roles of the loop between helices 8 and 9 more generally. Our experimentally determined topology has provided the basis for a revised understanding of existing data for this important transporter family as well as suggesting new directions.

Acknowledgements

This work was supported by an Australian Research Council Discovery grant, number DP0984773, to GDP; we thank Loraine Tucker, Warren Hudson-Taylor, Sara Milward and Soumi Bala for occasional technical support. We thank Dr Fiona Leves for helpful discussions and Naresh Verma's lab for the gift of the pNV1216 plasmid.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.