Molecular Characterization of Vibrio parahaemolyticusvSGLT

The Na+/galactose cotransporter (vSGLT) of Vibrio parahaemolyticus, tagged with C-terminal hexahistidine, has been purified to apparent homogeneity by Ni2+ affinity chromatography and gel filtration. Resequencing the vSGLT gene identified an important correction: the N terminus constitutes an additional 13 functionally essential residues. The mass of His-tagged vSGLT expressed under its native promoter, as determined by electrospray ionization-mass spectrometry (ESI-MS), verifies these 13 residues in wild-type vSGLT. A fusion protein of vSGLT and green fluorescent protein, comprising a mass of over 90 kDa, was also successfully analyzed by ESI-MS. Reconstitution of purified vSGLT yields proteoliposomes active in Na+-dependent galactose uptake, with sugar preferences (galactose > glucose > fucose) reflecting those of wild-type vSGLT in vivo. Substrates are transported with apparent 1:1 stoichiometry and apparent K m values of 129 mm (Na+) and 158 μm(galactose). Freeze-fracture electron microscopy of functional proteoliposomes shows intramembrane particles of a size consistent with vSGLT existing as a monomer. We conclude that vSGLT is a suitable model for the study of sugar cotransporter mechanisms and structure, with potential applicability to the larger SGLT family of important sodium:solute cotransporters. It is further demonstrated that ESI-MS is a powerful tool for the study of proteomics of membrane transporters.

the downhill translocation of Na ϩ with the intracellular accumulation of glucose (14) and is directly involved in the human disorder glucose/galactose malabsorption (15) and in the efficacy of oral rehydration therapy (16). Expression of SGLT1 and its site-directed mutants in Xenopus oocytes has been valuable in characterizing the structure and function of the protein (17,18), but mutants often fail to traffic to the oocyte plasma membrane (19 -21). Recently, a bacterial Na ϩ -dependent galactose/glucose cotransporter of the SGLT family has been cloned and characterized from Vibrio parahaemolyticus (22,23). Bacterial expression circumvents the oocyte trafficking problem by direct translocation into the plasma membrane and permits large scale protein purification. The Vibrio ortholog, which we term vSGLT, is thus more amenable to molecular analysis, which makes it an advantageous model for structure/ function studies.

MATERIALS AND METHODS
Construction of VNH6A and VNGFPH6 Expression Plasmids-The plasmid pYAT271A bears a genomic fragment carrying the vSGLT (SglS) gene from V. parahaemolyticus (23). The correct N terminus of vSGLT was discovered upon resequencing the pYAT271A insert (see "Results"). The full vSGLT coding region was then inserted after the L-arabinose-inducible pBAD promoter in the pBAD18 vector (24) and immediately downstream of a 44-base pair T7 phage phi10 ribosome binding region. A sequence encoding a 10-residue C-terminal spacer and six histidines (VLYKSGGSPGHHHHHH; His tag) was appended by polymerase chain reaction mutagenesis (25).
A fusion of vSGLT and green fluorescent protein (GFP) 1 has been expressed also (VNGFPH6). In this construct, the vSGLT C terminus is fused to a fifteenth membrane span sequence NCDAEITLIIFGVMAG-VIGTILLISYGIRRLIK (human glycophorin A membrane span is underlined), which is fused to a bright GFP variant minus the Met initiator (Maxygen, Palo Alto, CA) (26), terminating in SGGSPGHHHHHH. Sequencing assured the fidelity of constructs.
Cells were pelleted 10 min at 3000 ϫ g and resuspended in 100 mM NaCl, 100 mM Tris⅐Cl, pH 8.1, 5 mM Na 3 EDTA, 20% (w/v) sucrose. Hen lysozyme was added to 1 mg/ml, and the addition of one volume of water induced mild osmotic shock (27,28). After 5 min at 22°C, spheroplasts were cooled in ice to 5-10°C and stabilized with MgCl 2 to 10 mM. Phenylmethylsulfonyl fluoride was added to 0. 2  to 1 mM, galactose was added to 20 mM, and NaCl was added to 0.1 M. Spheroplasts were centrifuged for 10 min at 3000 ϫ g at 2°C.
His-tagged protein fractions were pooled, made 10 mM in EDTA, and concentrated to Ͻ0.25 ml by ultrafiltration in Centricon YM50 devices (Millipore). Retentate was centrifuged to pellet any aggregates and then applied to a 1 ϫ 30 cm Superose R12 gel filtration column (Amersham Pharmacia Biotech) equilibrated in buffer N, 0.2 mM EDTA. The column, pre-calibrated with globular proteins, was developed at 0.2 ml/min. Fractions of purified protein were pooled, made 15% (v/v) in glycerol, and stored at 4°C.
Electrospray Ionization Mass Spectrometry-Purified protein was precipitated with CHCl 3 /MeOH (29), dissolved in 88% formic acid, separated from lipid and detergent by HPLC in organic solvent, and analyzed by ESI-MS (30).
Lipid/detergent/protein mixtures were made at 22°C. 30 l of emulsified lipid, 50 l of 10% C10M, and 250 l of 2ϫ KG buffer (20 mM Tris, 20 mM HEPES, 300 mM KCl, 200 mM choline⅐Cl, and 20% glycerol) were combined and incubated for 5 min. Undissolved lipid was removed by centrifugation. Purified VNH6A (2.5 g in 170 l) was added and incubated for 5 min. 5 l of a mixture of 100 mM each MgCl 2 and CaCl 2 was added and incubated for 5 min. PLs formed upon addition of 270 l of CD2 (60 mM ␤-cyclodextrin, 15% formamide in 1ϫ KG buffer) in 10-l aliquots, mixing after each addition. PLs were incubated 5 min and then desalted on a 7-ml bed of Superdex 30 (Amersham Pharmacia Biotech), 1 cm in diameter, in buffer W2 (10 mM Tris, 10 mM HEPES, 150 mM KCl, 3 mM MgSO 4 , 1 mM CaCl 2 , 0.1 mM EDTA, 1 mM dithiothreitol) at 0.5 ml/min. Desalted PLs were pelleted at 400,000 ϫ g for 1 h at 4°C. Proteoliposome pellets for sugar uptake assays were resuspended in 200 l of buffer W2 with a 28G needle and stored at Ϫ80°C.
Transport Assay of His-tagged vSGLT in Proteoliposomes-PLs were thawed and subjected to three more freeze/thaw cycles in liquid N 2 before assay at 22°C. Uptake was initiated by mixing 10 l of PLs and 10 l of buffer 2ϫ U ([ 14 C]D-galactose, 200 mM choline⅐Cl plus NaCl, 20 mM Tris, 20 mM HEPES, 300 mM KCl, 6 mM MgSO 4 , 2 mM CaCl 2 ). Each reaction was stopped with 1 ml of cold buffer RIN (10 mM Tris, 10 mM HEPES, 150 mM KCl, 50 mM choline⅐Cl, 50 mM NaCl, 3 mM MgSO 4 , 1 mM CaCl 2 ) applied centrally to a 0.22-m nitrocellulose filter type GSWP (Millipore) over vacuum and washed with 4 ml of cold RIN, and the filter was assayed by scintillation counting.
Freeze Fracture and Electron Microscopy of Proteoliposomes-For freeze fracture EM of large vesicles, PL pellets were freeze/thawed twice in liquid N 2 , resuspended in 20 l of buffer W2, and then freeze/ thawed twice more. After 16 h at 4°C, they were prepared for freeze fracture (32). Particle densities and dimensions were determined (33).
Analytical Methods-Protein was quantitated by BCA assay (Pierce) after precipitation with 20 g of polyadenylate, 7% (w/v) trichloroacetic acid. SDS-PAGE (34) and Ag ϩ stains were done as described (35). Western blots were probed with a monoclonal antibody to tetra-His (Qiagen) and visualized by ECL assay (Pierce).

The Native N Terminus Comprises an Additional 13
Residues-The plasmid pYAT271A carries a V. parahaemolyticus genomic fragment bearing the SglS (vSGLT) Na ϩ /galactose cotransporter gene (23). An initial attempt to place the reported 530-residue coding sequence under an arabinose-inducible pBAD promoter resulted in a surprising inability to induce expression of Na ϩ -dependent sugar transport in the glucose transport-impaired E. coli strain JM1100 (36). Resequencing the pYAT271A insert revealed two bases that are missing in the original sequence report (D78137). The CG dinucleotide (boxed) is upstream of the incorrect initiator codon (ATG, above an aligned translation) and within the actual N-terminal coding region.
The N-terminus is thus 13 residues longer (bold type), increasing the total number to 543. A ribosome binding sequence (bold type, underlined) occurs upstream of the correct initiator codon (ATG). Restoration of the sequence encoding these 13 residues to our original, nonfunctional pBAD construct enabled inducible Na ϩ -dependent glucose cotransport in vivo. Uptake of radiolabeled glucose by full-length vSGLT, expressed under the pBAD promoter in JM1100 cells, was 34-fold greater in the presence of sodium (58.6 pmol/min⅐g protein) than in choline, whereas cells bearing the empty vector had no sodium-dependent uptake (1.7 pmol/min⅐g protein in either cation).
Purified vSGLT migrates on SDS-PAGE at an apparent mass of 35 kDa ( Fig. 2A, lane GF) and is recognized by an antibody to tetrahistidine (Fig. 2B). Lipid apparently copurifies with VNH6A: a Ag ϩ -staining band comigrates with a lecithin band from control liposomes ( Fig. 2A, cf. lanes GF and L), and neither band is stained by Coomassie R250 (not shown).
ESI-MS of His-tagged vSGLT Expressed under the pBAD and Native Vibrio Promoters-The genomic fragment in pYAT271A was modified such that vSGLT expressed under the Vibrio promoter contained a C-terminal His tag identical to that of pBAD-expressed VNH6A. Proteins expressed under both pBAD and Vibrio promoters were purified separately for comparison by ESI-MS.
Purified protein was subjected to HPLC gel filtration in organic solvent to remove detergent and lipid. Fig. 3 (A and B) shows the UV absorbance and total ion current of VNH6A (highlighted) eluting at about 500 s. The average ESI-MS spectrum obtained from this eluting peak exhibits VNH6A molecular ions carrying 26 -53 positive charges (Fig. 3C). Twenty four measurements yield a single compound mass of 60,681 Da, which is within 0.015% (change of 9 Da) of the calculated mass of 60,671.7 Da for VNH6A of 559 residues (543 ϩ 16) retaining its formylmethionine. The discrepancy between measured and calculated masses lies just outside the most stringent criterion of 0.008% applied to ESI data collected on a quadrupole instru- The mass determined by ESI-MS is consistent with an extracellular N terminus, because a periplasmic disposition prevents excision of the 150-Da formylmethionyl, or just the 28 Da formyl, group (39,40). vSGLT, bearing the His tag, was also expressed from pYAT271A (i.e. under its native Vibrio promoter) and purified. This protein yielded molecular ions carrying 29 -61 charges, with 16 estimates of compound mass averaging 60,676 Da (Fig.  3E), which is within 0.008% of the calculated mass (60,671.7). Therefore, wild type vSGLT (i.e. without the His tag) must necessarily comprise 543 residues.
A GFP fusion of vSGLT (VNGFPH6, functional in vivo and in PLs) was also purified and analyzed by ESI-MS, yielding 24 molecular ions carrying 47-78 charges (not shown). The compound mass determined averages 90,544 Da, which is within 0.008% of the calculated mass of 90,544.6 Da (before fluorophor formation) or within 0.01% of 90,542.5 (with hydrolyzed fluorophor). The sample preparation procedure for HPLC may lead to hydrolysis of the imidazolone ring or to an oxidation. Additionally, the fraction of analyzed VNGFPH6 protein that had completed the slow pathway of dehydration and oxidation to formation of the green fluorophor is unknown (calculated mass, 90,524.5 Da after fluorophor formation).
Freeze-Fracture Electron Microscopy of Liposome-reconstituted VNH6A-Purified VNH6A was reconstituted into PLs (31). Aliquots (ϳ5 l) of PLs, as prepared for uptake assays, were treated for freeze-fracture EM. In the replicas, PLs appeared as concave or convex surfaces of ϳ200 nm in diameter. The fracture faces contained 1-3 intramembrane particles (Fig.  4, B-D). These particles were absent in control liposomes (Fig.  4A). Moreover, PLs studied by SDS-PAGE showed one band of protein that was absent in control liposomes ( Fig. 2A, lanes PL  and L), suggesting that these particles were induced by the addition of protein.
To estimate the density and molecular dimensions of the reconstituted VNH6A, large PLs (Ն 2m in diameter) with a 12-fold higher protein/lipid ratio were prepared by freeze-thawing a PL pellet, to increase the density of particles and reduce the PL radius of curvature. Consistent with the increase in protein, the particle density also increased, to be 352 Ϯ 72/m 2 membrane (n ϭ 2160 from 6.2 m 2 ) (Fig. 4E). At high magnification, vSGLT particles had an asymmetrical appearance with a long and short axis (Fig. 4E, inset). Measurements indicated a single population of particles with a mean size of 9.1 Ϯ 0.4 by 7.0 Ϯ 0.5 nm (n ϭ 70).
Transport Characteristics of Reconstituted VNH6A-The time course of galactose uptake into PLs exhibited an initial rate 10 -25-fold higher in 100 mM Na ϩ than in choline (Fig. 5A,  E). Maximum uptake at 30 -60 min decayed thereafter with a t1 ⁄2 of ϳ150 min. The peak galactose content of the PLs in NaCl at 30 -60 min is 4-fold higher than that, in Na ϩ , at 300 min, and 19-fold higher than that of PLs in choline at 300 min.
For PLs (and liposomes) in choline, a rapid, small initial sugar "uptake" (t1 ⁄2 Ͻ 1 min) was followed biphasically by slow influx (0.2 fmol/min) to ϳ200 fmol at 300 min (Fig. 5A, q). This Na ϩ -free initial rate in PLs or liposomes increased linearly with 12-400 M galactose (not shown), and thus both the rapid and slow components of the slight Na ϩ -free uptake into PLs are independent of VNH6A and indicate transbilayer equilibration. The initial rate of sugar accumulation into control liposomes was typically about 55% that of PLs in the presence of choline and was not subtracted from the values plotted in Figs. 5 and 6. The initial rate of D-galactose transport into PLs was a saturable function of [Na ϩ ] (Fig. 5B). The K Na was calculated at 129 Ϯ 8 mM, with a Hill coefficient of 0.9, and a V max of 128 Ϯ 4 pmol/g⅐min. Likewise, the initial rate of galactose transport was a saturable function of the external [galactose] (Fig. 5C). In 100 mM Na ϩ , the K galactose was 158 Ϯ 10 M, the Hill coefficient was 1.0, and the V max was 178 Ϯ 5 pmol/g⅐min. Indicative of cotransport, the apparent affinity for galactose decreased as external [NaCl] was reduced, the K galactose increasing by 2.6fold with a 5-fold reduction of external [NaCl] to 20 mM.
Sugar selectivity was assessed by the ability of external sugars (20 mM) to inhibit the initial uptake rate of [ 14 C]Dgalactose into PLs (Fig. 6). Addition of D-galactose reduced Na ϩ -dependent radiotracer uptake to the level observed in choline. D-Glucose inhibited uptake by 92%, whereas D-fucose inhibited by 70%. L-glucose, 2-deoxy-D-glucose, or ␣-methyl-Dglucoside did not inhibit significantly. The presence of 20 mM sugars had no effect on uptake in choline. DISCUSSION vSGLT bearing a C-terminal histidine tag was expressed, purified, and reconstituted as a functional Na ϩ /galactose cotransporter. The native N terminus of vSGLT comprises an additional 13 residues, the omission of which results in a nonfunctional truncation of vSGLT. ESI-MS analysis of protein expressed from the Vibrio genomic fragment establishes unequivocally these 13 residues in wild type vSGLT. This finding and the ability to determine the mass of the VNGFPH6 construct of Ͼ90 kDa and 15 transmembrane spans, affirm the utility of ESI-MS in membrane proteomics (29,30). The N terminus of vSGLT is predicted to reside extracellularly (2), as established for the human SGLT1 Na ϩ /glucose (11), bacterial putP Na ϩ /proline (12), and rat NIS Na ϩ /iodide cotransporters (13). ESI-MS is at least consistent with retention of the Nterminal Met residue and its N-formyl group, which accords with a periplasmic disposition of the N terminus (39,40). Asp 12 of the newly recognized and functionally essential N terminus corresponds to Asp 28 of human SGLT1. Mutations D28N (15) and D28G (41) each cause the disorder glucose/galactose malabsorption, and this aspartate is conserved in all SGLT sugar cotransporters as well as in the Na ϩ /iodide and Na ϩ /myoinositol cotransporters (see Fig. 2 of Ref. 2), suggesting a functional importance.
Analysis of vSGLT in PLs by freeze-fracture EM demonstrates that the protein is integrated into the lipid membrane of unilamellar vesicles. The protein particles are homogeneous and asymmetrical (Fig. 4E). After correction for the replica thickness of 1.2 nm (33), the average particle area measures 24 Ϯ 5 nm 2 , a value comparable with that reported for rabbit SGLT1. The area of freeze fracture images of membrane proteins scales linearly with the number of transmembrane helices (33), and the area of the vSGLT particles is consistent, within experimental error, with a monomeric protein of 14 transmembrane helices.
Galactose transport in PLs was fully dependent on Na ϩ . Rapid vesicular galactose accumulation, followed by slow efflux (Fig. 5A), is explicable only by cotransport. The slowness of galactose efflux in Na ϩ and insignificant galactose influx in choline indicate that these PL bilayers are impermeant. PLs show an apparent K Na of ϳ130 mM in 100 M galactose and an apparent K galactose of ϳ160 mM in 100 mM Na ϩ . Affinities observed here are significantly lower than those measured in vivo in E. coli: K Na Ͻ 10 mM and K galactose ϳ40 M (42). In rightside-out vesicles from E. coli expressing vSGLT, a K Na of 6 -9 mM was observed in 100 M glucose. 2  B, total ion current profile eluting from the column. Both profiles show separation of VNH6A (shaded) from other UV-absorbing and ion-generating compounds. C, ESI mass spectrum from the highlighted VNH6A peak. ESI generated multiply charged ions of VNH6A, bearing between 26 and 53 protons. Spectra from several scans across the peak were compiled to produce the spectrum shown. D, computer reconstruction of the VNH6A molecular weight profile. E, His-tagged vSGLT was expressed under the native Vibrio promoter and purified, and 11 g were analyzed by ESI-MS. The reconstructed mass coincides, within error, with that of the pBAD-expressed VNH6A.
ESI-MS does not suggest significant oxidation or other covalent chemical changes.
Reconstituted vSGLT may reside in the bilayer in two orientations, as suggested by freeze fracture EM images of PLs showing particle retention on concave as well as convex faces. Distinct substrate affinities may be expected for the two orientations, but curve fitting of data of Fig. 5, B and C, to Michaelis-Menten equations for two distinct affinities for Na ϩ or for galactose contraindicates a second K m for either substrate. We infer that the observed K m values reflect the appropriately oriented transporter, because the reverse orientation should exhibit the lowest affinities (43,44), possibly too low to detect in our assays.
Hill coefficients for Na ϩ and galactose of 0.9 and 1.0, respectively, suggest a 1:1 cotransport stoichiometry. Sugar preferences of reconstituted VNH6A (D-galactose Ͼ D-glucose Ͼ Dfucose) and lack of an inhibitory effect by 2-deoxy-D-glucose and L-glucose reflect characteristics of vSGLT in vivo (42). Galactose turnover per VNH6A molecule is estimated from data in  PLs prepared for transport assays at a protein:lipid ratio of 1:470. Intramembrane particles (arrows) appear in both convex (B) and concave faces (C and D) of unilamellar vesicles, indicating a random insertion of the protein in the bilayer. E, larger PLs (Ն 2 m) reconstituted at a 12-fold higher VNH6A protein:lipid ratio than in transport vesicles (B-D) were used to estimate the average particle size, 9.1 Ϯ 0.4 by 7.0 Ϯ 0.5 nm (n ϭ 70). The particle density on the this convex face was 352 Ϯ 72/m 2 (n ϭ 2160 from 6.2 m 2 ). The scale bar (200 nm) refers to all five panels. Inset, High magnification view of one vSGLT particle (scale bar, 20 nm). In summary, we show that a His-tagged fusion of the V. parahaemolyticus Na ϩ /galactose cotransporter can be purified to apparent homogeneity and functionally reconstituted. ESI-MS confirms 13 newly recognized amino acid residues at the N terminus and suggests that the purified protein is unmodified either post-translationally or during purification. Freeze-fracture EM reveals that vSGLT exists in the bilayer as a monomer, as does its homologue, SGLT1. Sodium gradients drive galactose cotransport with apparent 1:1 stoichiometry, and sugar specificities generally correspond to those observed in vivo, although galactose and Na ϩ affinities are unexpectedly low. The ability to purify a functional Na ϩ /sugar cotransporter enables a variety of spectroscopic, labeling, and cross-linking experiments to probe structure and function, with results possibly relevant to other cotransporters in the SGLT family.