Identification of an Extracellular Gate for the Proton-coupled Folate Transporter (PCFT-SLC46A1) by Cysteine Cross-linking*

The proton-coupled folate transporter (PCFT, SLC46A1) is required for intestinal folate absorption and folate homeostasis in humans. A homology model of PCFT, based upon the Escherichia coli glycerol 3-phosphate transporter structure, predicted that PCFT transmembrane domains (TMDs) 1, 2, 7, and 11 form an extracellular gate in the inward-open conformation. To assess this model, five residues (Gln45-TMD1, Asn90-TMD2, Leu290-TMD7, Ser407-TMD11 and Asn411-TMD11) in the predicted gate were substituted with Cys to generate single and nine double mutants. Transport function of the mutants was assayed in transient transfectants by measurement of [3H]substrate influx as was accessibility of the Cys residues to biotinylation. Pairs of Cys residues were assessed for spontaneous formation of a disulfide bond, induction of a disulfide bond by oxidization with dichloro(1,10-phenanthroline)copper (II) (CuPh), or the formation of a Cd2+ complex. The data were consistent with the formation of a spontaneous disulfide bond between the N90C/S407C pair and a CuPh- and Cd2+-induced disulfide bond and complex, respectively, for the Q45C/L290C and L290C/N411C pairs. The decrease in activity induced by cross-linkage of the Cys residue pairs was due to a decrease in the influx Vmax consistent with restriction in the mobility of the transporter. The presence of folate substrate decreased the CuPh-induced inhibition of transport. Hence, the data support the glycerol 3-phosphate transporter-based homology model of PCFT and the presence of an extracellular gate formed by TMDs 1, 2, 7, and 11.

Mammals meet their biosynthetic needs for one-carbon donors by absorption of dietary folates mediated by the protoncoupled folate transporter (PCFT) 2 (SLC46A1) (1,2). Mutations in the PCFT gene that result in loss of the protein or its function are the basis for hereditary folate malabsorption (OMIM 229050), a rare autosomal recessive disorder (1,3).
PCFT transports protons along with folates as reflected in the current and acidification accompanying folate transport in oocytes that express this transporter (1, 4 -6). Structure-function studies, based on site-directed and random mutagenesis, have begun to define residues and domains that play a key role in PCFT function. Elucidation of the secondary structure of PCFT, along with the residues that contribute to the aqueous translocation pathway, have been documented by the substituted cysteine accessibility method (7). The protein consists of 12 transmembrane helices, divided in half by a large intracellular loop, with C and N termini oriented into cytoplasm (8 -10). Residues critical to folate/antifolate binding, proton coupling, and proton binding have been identified (5,(11)(12)(13). The roles of residues in the 2nd, 4th transmembrane domains (TMDs), and the loop connecting the 2nd and 3rd TMDs have been studied (14,15). Analyses of the functional defects that occur in subjects with hereditary folate malabsorption have provided insights into residues important to PCFT protein folding, stability, and/or trafficking to the cell membrane, substrate binding, and oscillation of the carrier between its conformational states (13, 16 -20).
PCFT expression is limited to eukaryotes; there are no closely related proteins present in archaea or bacteria, so that a highresolution x-ray crystal structure with a high level of sequence identity is not available that would facilitate a structure/function analysis of the human or rodent proteins. Rather, homology modeling has identified the structure of the bacterial homolog, the glycerol-3-transporter, GlpT, as having a protein structure most like PCFT despite a low (14%) sequence identity (5,10,21). In this model, the transporter is in a conformation that is closed to the outside and open to the inside. According to the alternative access model, PCFT oscillates between this conformation and a conformation that is closed to the inside and open to the outside. This computational model has been used for correlation with functional data and for structural predictions as in the current study (5, 12-14, 21, 22).
In this inward facing homology model of PCFT, TMD-1, -2, -7, and -11 are predicted to cluster together in proximity to the extracellular interface to form an extracellular gate. To confirm the presence of such a gate, Cys pairs predicted to be in proximity were introduced into these TMDs at locations near the extracellular surface. The possibility that the pairs are sufficiently close to be cross-linked and/or form a complex was evaluated by assessment of function, which should be decreased when the mobility of the gate is restricted and the rate of oscil-lation of the carrier is suppressed, under three conditions: (i) spontaneous formation of a disulfide bond accompanied by diminished accessibility of Cys residues to a biotinylation reagent, (ii) disulfide bond formation induced by oxidation with CuPh, and (iii) complex formation between a Cd 2ϩ ion and two Cys-substituted residues. In this way, evidence was obtained to indicate that pairs of Cys-substituted residues were crosslinked or complexed involving all of the predicted 1, 2, 7, and 11 TMDs. The data are consistent with the proximity of these residues and an external gate comprised involving these TMDs that controls access to the aqueous translocation pathway of this transporter.

Experimental Procedures
Cells and Culture Conditions-R1-11 cells were derived from HeLa cells and express neither the reduced folate carrier, due to deletion of its gene, nor PCFT due to methylation of its gene promotor (23,24). R1-11 cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, 100 units/ml of penicillin, and 100 g/ml of streptomycin. R1-11 cells were thawed regularly (once every 3 months) from liquid nitrogen stocks to ensure that PCFT expression was absent (25).
Site-directed Mutagenesis-An engineered human PCFT, which harbors two point mutations (C66S and C298S) and is tagged with HA (hemagglutinin) at the C terminus, was used as a template for site-directed mutagenesis in most cases. In some cases, HA-tagged, Cys-less PCFT was also used for mutagenesis. A desired mutation was first introduced into the template using the QuikChange II Site-directed Mutagenesis Kit (Stratagene, La Jolla, CA). After sequence verification, a second mutation was introduced into PCFT in a similar way to generate double mutants. Automated sequencing was performed in the Albert Einstein Cancer Center Genomics Shared Resource.
Transient Transfection-R1-11 cells were used as recipients for all transient transfections. For transport studies, cells were seeded in 20-ml Low Background glass scintillation vials (Research Products International Corporation, Prospect, IL). For biotinylation studies, cells were seeded in 6-well plates. In both cases, transfections were performed with Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol 2 or 3 days after seeding cells. Transport and biotinylation assays were always performed 2 days after transfection. Cell growth medium was refreshed 1 day after transfection.
Membrane Transport-Cells were washed twice with HBS (HEPES-buffered saline: 20 mM HEPES, 5 mM dextrose, 140 mM NaCl, 5 mM KCl, 2 mM MgCl 2 , pH 7.4) and incubated in the same buffer at 37°C for 20 min. The incubation buffer was then removed and transport was initiated by the addition of 0.5 ml of pre-warmed (37°C) MBS (MES-buffered saline: 20 mM MES, 5 mM dextrose, 140 mM NaCl, 5 mM KCl, 2 mM MgCl 2 , pH 5.5) containing 0.5 or 50 M[ 3 H]MTX. For [ 3 H]pemetrexed influx studies, transport was initiated by addition 0.5 ml of prewarmed (37°C) HBS containing 1 M drug and halted after 1 min by the addition of 5 ml of ice-cold HBS. Cells were washed three times with ice-cold HBS and digested in 0.5 ml of 0.2 M NaOH at 65°C for 1 h. Radioactivity in 0.4 ml of lysate was measured on a liquid scintillation spectrometer and normalized to the protein level obtained with the BCA Protein Assay (Pierce). All transport assays were performed at least three times on separate days. In each experiment every data point was conducted in duplicate.
Analysis of PCFT at the Cell Surface and Accessibility of PCFT Cys Residues by Biotinylation-Although biotinylation of Lys residues detects the level of PCFT at the plasma membrane, Cys biotinylation with a membrane impermeant reagent determines the accessibility of a Cys residue to the extracellular milieu (7,9,10,14). For surface labeling, cells were treated with EZ-Link Sulfo-NHS-LC-Biotin at a concentration of 1 mg/ml in HBS. For Cys biotinylation, cells were treated with MTSEA biotin at a concentration of 0.2 mg/ml in HBS. Both treatments were performed at room temperature for 30 min. Following this, processing of both biotinylation reactions was identical. Cells were washed in HBS at room temperature, then overlaid on ice with 0.7 ml of hypotonic buffer (0.5 mM Na 2 HPO 4 , 0.1 mM EDTA, pH 7.0) containing protease inhibitor mixture (Roche Applied Science). The cells were then detached from the plates with a disposable cell lifter and centrifuged at 16,000 ϫ g for 10 min at 4°C. The pellet was resuspended in 0.4 ml of lysis buffer (50 mM Tris base, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, pH 7.4) and a 50-l portion was collected as a sample from crude membranes and stored at Ϫ20°C. The remaining suspension was mixed on a rotisserie for 0.5 to 1 h and centrifuged at 16,000 ϫ g for 15 min, all at 4°C. The supernatant was mixed on a rotisserie overnight at 4°C with 50 l of streptavidin-agarose beads (Fisher Scientific) that had been pre-washed three times with the lysis buffer. The agarose beads were then washed at room temperature twice with the lysis buffer and then an additional two times with the lysis buffer containing 2% SDS, each with a 20-min mixing on a rotator. The precipitated proteins were then released from the beads by heating at 95°C for 5 min in 2ϫ SDS-PAGE sample loading buffer with or without DTT.
The blots were developed with Western Lightning Plus-ECL (PerkinElmer Life Sciences, Waltham, MA).
Reduction, Oxidation, or Complex Formation of PCFT Expressed in Cells-Reduction with DTT, oxidation with CuPh, and complex formation with CdCl 2 were all performed in transient transfectants expressing PCFT mutants immediately before transport or biotinylation assays. For reduction with DTT, cells were washed once with HBS at room temperature and then incubated for 10 min at room temperature with freshly made 12 mM DTT in HBS. After cells were washed twice with HBS at room temperature, transport or Cys accessibility was assessed. The control for DTT reduction was cells treated with HBS that did not contain DTT. For oxidation with CuPh, DTT-treated cells were washed a third time with ice-cold HBS then incubated for 5 min on ice with 10 M CuPh, which was prepared before experiments by a 1 to 1000 dilution of a frozen stock (a mixture of 10 mM dichloro(1,10-phenanthroline)copper(II) and 3 mM CuSO 4 in water) in ice-cold HBS. The cells were washed four times with ice-cold HBS prior to assessment of transport. The control for CuPh treatment was cells exposed to CuPh-free HBS. For complex formation with Cd 2ϩ , cells were treated with DTT then washed twice with HBS as indicated above, then exposed to CdCl 2 (8 M) in HBS for 20 min at 37°C. The buffer was then aspirated and transport was assessed in HBS containing 8 M CdCl 2 . For a control, no cadmium was present in the preincubation or transport buffer.
Oxidative Cross-linkage in the Presence of Folic Acid-Transient transfectants were washed once with HBS and incubated with 12 mM DTT in HBS for 10 min at room temperature. The cells were washed three times with HBS, incubated with 1.7 mM folic acid in HBS (1 ml) for 1 min, all at room temperature, then cooled in an ice water bath for 5 min. A CuPh solution in HBS (2.5 M, 1 ml) at 0°C was added to the cells to achieve a final CuPh concentration of 1.25 M and a final folic acid concentration of 0.85 mM, and the incubation continued for 5 min at this temperature. The cells were then washed three times with icecold HBS and placed in a 37°C water bath for 1 min following which MTX influx was initiated with the addition of 0.5 ml pre-warmed MBS (pH 5.5) containing 0.5 M [ 3 H]MTX.
Statistical Analysis-Statistical analyses were conducted with analysis of variance, GraphPad Software (La Jolla, CA).
Homology Model of Human PCFT-A comparative protein structure model for PCFT has been built for the interpretation of experimental observations within a three-dimensional context. A variety of homology models that differed in their targetto-template input alignments were built and the Prosa energy score evaluated (26). The most stable model was obtained using input from the HHpred method (27), a hidden Markov modelbased fold-recognition and alignment method. The best-scoring template found was the crystal structure of GlpT from Escherichia coli (Protein Data Bank code 1pw54), as in other studies (5, 10, 12-14, 21, 22). The optimal sequence alignment between PCFT and 1pw54 served as input for the comparative protein structure modeling program MMM (28 -31). The resulting model underwent additional quality verification by cross-correlating the location of predicted transmembrane helical segments using HMMTOP (32) with the ones observed in the model. The corresponding models resulted in a largely conserved core of transmembrane segments, but there were substantial variations in the modeling of the extra-and intracellular loop regions.

Identification of Candidate Residues Predicted to Form an Extracellular Gate for the Substrate Translocation Pathway in
Human PCFT-A homolog model has been built for human PCFT based upon the crystal structure of the E. coli GlpT transporter. As indicated in Fig. 1, TMD1, -2, -7, and -11 are predicted to be in a position to form a gate at the entry pathway for substrate in an outward-closed, inward-opened, conformation. Gln 45 in TMD1, Asn 90 in TMD2, Leu 290 in TMD7, Ser 407 , and Asn 411 in TMD11 are located in the termini of these TMDs and pairs are predicted to be in sufficient proximity to form disulfide bonds and/or a complex with Cd 2ϩ when substituted with Cys residues.
Activity and Expression of Single and Double Cys PCFT Mutants-To investigate the dynamic interactions between and among these transmembrane helices, five single mutations (Q45C, N90C, L290C, S407C, and N411C) and nine double mutations (Q45C/N90C, Q45C/L290C, Q45C/S407C, Q45C/ N411C, N90C/L290C, N90C/S407C, N90C/N411C, L290C/ S407C, and L290C/N411C) were generated in a modified PCFT template in which both Cys 66 and Cys 298 residues were substituted with serine. The double mutant containing S407C and N411C was not studied because both mutations are located in the same helix. Previous studies demonstrated that Cys 66 and Cys 298 are the only residues that form a disulfide bond in the wild-type PCFT and this bond is not required for protein stability or function (9). The removal of these Cys residues eliminated the native disulfide bond in the molecule that would confound interpretation of the results but minimally affects function of the transporter. This template is identified as PCFT-DSL and these mutants are identified with a "-DSL" suffix (disulfide bond-less), indicating a platform that lacks this post-translational modification. The activity of PCFT-DSL was 72 Ϯ 5% of that of wild-type PCFT assessed by [ 3 H]MTX influx (0.5 M, pH 5.5) in transient transfectants based upon three independent experiments. A Cys-less PCFT (PCFT-CL), in which all Cys residues were converted to Ser was not routinely used to generate these mutants because introduction of additional mutations in the Cys-less scaffold often results in destabilization of the protein, whereas the same mutations generated in wild-type or DSL PCFT are expressed and functional, a phenomenon also observed for other transporters (33,34).
The activity of the PCFT mutants was assessed in transient transfectants by measuring influx of [ 3 H]MTX at pH 5.5. As indicated in Fig. 2A, activities of the single mutants, N90C-DSL, S407C-DSL, and N411C-DSL were comparable with that of PCFT-DSL (91.8, 100, and 98.1%, respectively). Activity of Q45C-DSL was half, whereas activity of the L290C-DSL mutant was one-fourth that of PCFT-DSL. As discussed below, neither Q45C nor L290C forms a disulfide bond with another Cys residue in the molecule, but both were pre-modified with endogenous thiols. All the double mutants exhibited an activity Յ25% that of PCFT-DSL except the N90C/N411C mutant, which retained 64% of PCFT-DSL activity. The activity of most double mutants was much lower than what might be expected from their two single constituting mutants (e.g. 13% activity for the Q45C/N411C-DSL as compared with 50 and 98% for the Q45C-DSL and N90C-DSL mutants, respectively).
PCFT expression at the plasma membrane was measured by cell surface biotinylation with a reagent targeting free amine groups. The expression of all single mutants in the plasma membrane was comparable with that of PCFT-DSL (Fig. 2B).
Although expression of the Q45C/N90C-DSL and N90C/ L290C-DSL double mutants in the plasma membrane was markedly decreased, expression of the other double mutants was comparable with that of PCFT-DSL. Expression of all PCFT mutants in the crude membranes tended to mirror the cell surface expression suggesting that trafficking of the mutants to the cell membrane was largely intact. Hence, the single mutant, L290C-DSL and the double mutants, Q45C/ L290C-DSL, Q45C/S407C-DSL, Q45C/N411C-DSL, N90C/ S407C-DSL, L290C/S407C-DSL, and L290C/N411C-DSL, had markedly decreased activity despite their robust expression at the plasma membrane. The basis for the loss of activity of the Q45C/N90C-DSL and the N90C/L290C-DSL pairs was less clear because this could be related to their low levels of expression.
Impact of DTT Reduction on Transport Activities of the Mutants-To assess whether formation of a disulfide bond between two Cys-substituted residues in the double mutants was a basis for the reduction in PCFT function, all transient transfectants were treated with 12 mM DTT at room temperature and neutral pH for 10 min to break disulfide bonds before MTX influx was measured at acidic pH (35). As indicated in Fig.  3A, which illustrates the fold-increase in transport activity induced by DTT, this treatment had no significant effect on the activity of the N90C-DSL, S407C-DSL, and N411C-DSL single DSL mutant, and no effect on the activity of the N90C-DSL, S407C-DSL, and N411C-DSL single mutants. Fig. 3B illustrates activities of the various mutants after DTT treatment as a percentage of function of DTT-treated PCFT-DSL. The activities of the single mutant L290C-DSL and the double mutants, L290C/S407C-DSL and L290C/N411C-DSL, were comparable with the activity of PCFT-DSL under these conditions, suggesting that these mutants are fully functional when the substituted Cys residue(s) are not modified. Although the activity of Q45C-DSL reached 70% of PCFT-DSL, the function of Q45C/L290C-DSL, Q45C/S407C-DSL, and Q45C/ N411C-DSL was increased to ϳ50% of PCFT-DSL with DTT reduction. This degree of impaired function of the double mutants could be attributed to the decreased function of the Q45C single mutant even after DTT. The activities of Q45C/ N90C-DSL and N90C/L290C-DSL were only 20 and 30%, respectively, of PCFT-DSL after DTT treatment despite the much lower level of expression, suggesting that these two mutants are also functional. Therefore, all these mutants exhibited sufficient intrinsic activity for further analysis.
Effect of DTT Reduction on the Accessibility of the Substituted Cys Residues-As previously noted, residues, Gln 45 , Asn 90 , Leu 290 , Ser 407 , and Asn 411 of human PCFT are predicted to be located near the extracellular interface of the aqueous translocation pathway. If this is the case, then the single substituted Cys residues at these positions should be accessible to the extracellular compartment and modified by a membrane-impermeant sulfhydryl-targeting biotinylation reagent such as MTSES biotin. However, if the two Cys residues form a spontaneous disulfide bond in the double mutant, the labeling of each of the two residues will be diminished or eliminated as previously demonstrated for the native Cys 66 and Cys 298 residues of the wild-type human PCFT (9). On the other hand, if the disulfide bond is reduced by DTT, these two Cys residues should become available for labeling (9). Accordingly, accessibility of the substituted Cys residues in all mutants was assessed before and after DTT reduction with MTSES biotin. As indicated in Fig. 3, C and D, the substituted Cys residues in the N90C-DSL, S407C-DSL, and N411C-DSL single mutants were labeled to the same extent prior to, and after, DTT treatment consistent with the lack of effect of DTT on their transport activity (Fig. 3A). However, DTT markedly increased labeling of the Q45C-DSL and L290C-DSL single mutants and increased labeling of all double mutants, to different extents, except for the N90C/N411C-DSL mutant. The increase in accessibility of the N90C/S407C-DSL mutant with DTT, and the lack of increased accessibility for the N90C-DSL and S407C-DSL single mutants (Fig. 3D) was consistent with the presence of a spontaneous disulfide bond as was the loss of intrinsic activity of this double mutant (Fig. 2A). The lack of DTT-induced labeling of the N90C/N411C-DSL mutant suggested the lack of spontaneous formation of a disulfide bond, also consistent with the lack of a DTT effect on the activity of this mutant. Because DTT treatment resulted in labeling of the Q45C-DSL and L290C-DSL single mutants, no conclusion was possible as to whether a disulfide bond formed spontaneously in the double mutants harboring these Cys-substituted residues in which accessibility also increased with DTT. The biotinylation of the substituted Cys residues in the S407C-DSL and N411C-DSL mutants, and the lack of biotinylation of the substituted Cys residue in the Q45C-DSL mutant in the absence of DTT treatment, were also observed for the same mutations introduced into Cys-less PCFT (10).
Post-translational Modifications of the Substituted Cys Residues in Q45C-DSL and L290C-DSL PCFT Mutants-As indicated above, DTT treatment enhanced both the activity and accessibility of the Q45C-DSL and L290C-DSL single mutants. To exclude the possibility that the substituted Cys in Q45C-DSL and L290C-DSL mutants spontaneously form a disulfide bond with one of the five remaining native Cys residues, the Q45C or L290C mutations were introduced in the Cys-less PCFT to generate the Q45C-CL and L290C-CL mutants. As indicated in Fig. 4A, expression of the Q45C-CL and L290C-CL mutants was much lower than that of PCFT-DSL both at the plasma membrane and in the crude membrane fractions. This is consistent with the previous finding that the Cys-less PCFT is  vulnerable to the introduction of additional mutations (15,33). Consistent with the low expression, the activities of the Q45C-CL and L290C-CL mutants were lower than their counterparts in the DSL scaffold. However, the increase in activities of both mutants with DTT was comparable with the increase when generated in the DSL scaffold (Fig. 4B) and, most important, the accessibility of both mutants to biotinylation was increased after treatment with DTT in the Cys-less scaffold (Fig. 4C). Hence, the impact of DTT on the activities or accessibility of the substituted Cys residues was similar for the Q45C or L290C mutations regardless of the template used for the mutants (Cys-less versus disulfide bond-less). These results indicate that post-translational modification of the Cys-substituted residue in either Q45C-DSL or L290C-DSL is not due to the formation of a disulfide bond with one of the native Cys residues. Rather, the data are consistent with modification by endogenous thiols similar to what occurred when a Cys residue was substituted for Thr 338 of the cystic fibrosis transmembrane conductance regulator (36). Most important, the modifications of Q45C and L290C mutants appeared to be reversible with DTT reduction. The Impact of CuPh-DTT reduction not only restored activity of most single mutants and some double mutants as indicated in Fig. 3B, but also freed-up sulfhydryl groups for modification in the mutants. This enabled an analysis of the impact of CuPh, an agent that cross-links adjacent thiol groups. As indicated Fig. 5A, CuPh did not significantly alter the activity of the single mutants suggesting that CuPh neither non-specifically reduces PCFT function nor interferes with the function of the single Cys mutants. A critical factor in this assay was that the L290C-DSL mutant maintained its function after CuPh treatment in contrast to the 80% reduction in activity due to the post-translational modification (Figs. 2 and 3). This indicates that this Cys residue was not modified by CuPh treatment under these conditions. It can be seen that CuPh oxidation resulted in a significant decrease in the activities of all double mutants as compared with that of the single mutant constituents. The greatest reduction in activity induced by CuPh oxidation was observed in the Q45C/L290C-DSL (79%), N90C/ S407C-DSL (85%), and L290C/N411C-DSL (89%) mutants.
The effect of CuPh oxidation on MTX influx kinetics was assessed for the Q45C/L290C-DSL, N90C/S407C-DSL, and L290C/N411C mutants at MTX concentrations of 0.5 and 50 M, the latter Ͼ10 times greater than the influx K t for wild-type PCFT (13), thereby reflecting the influx V max (Table 1). Based upon three experiments, oxidation by CuPh reduced activity of the Q45C/L290C-DSL mutant by 54% at 50 M, comparable with 64% at 0.5 M MTX. Similarly, CuPh oxidation reduced activity of the N90C/S407C-DSL mutant by 87% at 50 M, comparable with 84% at 0.5 M MTX, and decreased activity of the L290C/N411C-DSL mutant by 92% at 50 M, comparable with 88% at 0.5 M MTX. Hence, the major impact of CuPh oxidation appears to be a reduction in the influx V max for both mutants, consistent with an alteration in the mobility of the carrier.
The Impact of Cadmium Chloride-Cd 2ϩ interacts with the sulfhydryl group of two Cys residues to form a complex and the extent of this interaction depends upon the proximity of the residues (37). Cd 2ϩ has been used to probe the proximity of Cys residues in sodium channels and neuronal glutamate transporters (35,37). Fig. 5B indicates the effect of Cd 2ϩ following DTT treatment on the activity of the mutants. This was performed at neutral pH to preserve the capacity of Cd 2ϩ to form a complex with Cys residues, which is much weaker at pH 5.5 (see below). Activity was determined with pemetrexed because this agent retains greater influx activity at neutral pH than MTX (38,39). Cd 2ϩ had no effect on the activity of any of the single mutants nor did it significantly decrease the activity of the Q45C/ L290C-DSL and N90C/L290C double mutants as compared with their respective constituting single mutants. However, Cd 2ϩ significantly reduced the activities of all other double mutants as compared with the respective constituting single mutants. Hence, except for the Q45C/L290C pair, the effect of Cd 2ϩ was consistent with the effect of CuPh oxidation, with the strongest inhibition by Cd 2ϩ observed for the Q45C/S407C-DSL (80%), N90C/S407C-DSL (73%), and L290C/N411C-DSL (85%) mutants. When influx was measured at pH 5.5, only 35 and 20% inhibition was observed for the N90C/S407C-DSL or L290C/N411C-DSL mutants, respectively (data not shown), consistent with much weaker binding of Cd 2ϩ to Cys residues at this low pH. Impact of Folic Acid, a PCFT Substrate, on the CuPh-induced Cross-linkage between Two Cys Residues-The pairs of Cys residues, particularly, the Q45C/L290C-DSL, N90C/S407C-DSL, and L290C/N411C-DSL mutants were found to be in proximity when the mutants were unoccupied with substrates. Experiments were then undertaken to determine whether the presence of transport substrate would alter the extent of cross-linking induced by CuPh, as reflected in inhibition of transport. Folic acid was chosen in these studies as a substrate for PCFT. Preliminary data indicated that folic acid provided the greatest protection when the CuPh concentration was decreased to 1.25 M. Among the nine double mutants studied, sufficient inhibition of activity (Ͼ40%) induced by 1.25 M CuPh could be achieved for the Q45C/L290C-DSL, N90C/S407C-DSL, and L290C/N411C-DSL mutants. Therefore the protective effect of folic acid on the cross-link-induced inhibition was studied in these three mutants. As indicated in Fig. 6, the presence of folic acid alone did not significantly alter the activity of these mutants. CuPh at a concentration of 1.25 M decreased the activity to 50, 55, and 40% of the control in the Q45C/L290C-DSL, N90C/S407C-DSL, and L290C/N411C-DSL mutants, respectively. The presence of folic acid during exposure to CuPh protected the activity of all three mutants. Notably, whereas CuPh inhibition (ϳ60%) was greatest for the L290C/ N411C-DSL mutant, folic acid protection was complete.

Discussion
In the absence of a crystal structure for eukaryotic solute transporters, with few exceptions (40 -44), biochemical methodologies have been employed to characterize tertiary structure based upon homology models of known bacterial transporter. Cross-linking of introduced Cys pairs is an established approach to confirm the proximity of Cys-substituted residues in membrane transporters (45)(46)(47). Recently, this approach was applied to identify residues in transmembrane helices that form an extracellular or intracellular gate at external or internal interfaces of the aqueous translocation pathway based upon the inward-open or outward-open predicted structures, respectively. Hence, an extracellular gate involving two helices of the neuronal glutamate transporter EAAC1 was defined based upon the Glt Ph Pyrococcus horikoshii glutamate transporter structure (35). Three helices of the Leischmania donovani nucleoside LdNT1.1 transporter were shown to form an external gate based upon an ab initio computational model similar to GlpT (48) as was an intracellular gate of the same transporter encompassing four different helices based upon a homology model of the E. coli fucose-proton symporter in the outward facing conformation (34). In the current study, a homology model of the inward-open structure of the GlpT transporter predicted helices that could be components of an extracellular gate for PCFT along with residues within those helices that appeared sufficiently close to allow the formation of a disulfide bond when substituted with a Cys moiety. The experimental data, based upon cross-linking of Cys-substituted residues in the presence and absence of the PCFT substrate along with complexing with Cd 2ϩ , provide strong evidence that these four transmembrane helices are in proximity and form an extracellular gate.
Critical to the experimental design was creating a PCFT scaffold in which Cys-substituted residues only react with each other and not one of the endogenous Cys residues. Ideally, Cyssubstitution studies should utilize the Cys-less transporter to exclude possible interference from native Cys residues. How-  ever, the Cys-less PCFT is vulnerable to the introduction of other mutations posing a serious limitation to the application of the substituted cysteine accessibility method (15,33). But the preservation of only one or two native PCFT Cys residues, in an apparent nonspecific fashion, markedly reduces the vulnerability of PCFT to subsequent mutations (33). Accordingly, Cys substitutions were inserted into a disulfide bond-less PCFT (PCFT-DSL) scaffold in which both Cys 66 and Cys 298 , located in extracellular loops and participants in a sulfhydryl bond, were substituted with Ser. Indeed, this PCFT scaffold tolerated Q45C and L290C mutations, whereas expression of Cys-less PCFT that harbored these mutations was markedly decreased. Vulnerability of the Cys-less background to the introduction of Cys substitutions has been observed for other membrane transporters, such as L. donovani nucleoside transporter 1.1 (34). Hence, the method employed in this study to circumvent this problem may have broad applicability to membrane transporters, in general.
The data indicate that the substituted Cys residues in the Q45C-DSL and L290C-DSL mutants were post-translationally modified. However, this did not involve any of the native Cys residues. Post-translational modifications of a native Cys residue, primarily by S-glutathionylation, play a role in functional regulation of ion channels under certain physiological and pathological conditions (49). Post-translational modifications of a substituted Cys residue (T338C) were found to reduce conductance in the cystic fibrosis transmembrane conductance regulator and, under some conditions, could be reversed by DTT (36). The modification of the T338C was enhanced by the intracellular glutathione level (50), which prevented the substituted Cys from further reacting with methanethiosulfonates (51). The post-translational modification of the Cys residues introduced at positions Gln 45 and Leu 290 of PCFT has important ramifications for the interpretation of results based upon the substituted cysteine accessibility method in general. Hence, either the absence of biotinylation or a decrease in activity does not necessarily mean that a substituted Cys residue is not accessible to the extracellular compartment, or is not present within the aqueous translocation pathway, or is a poorly functioning transporter. Rather, it is always necessary to exclude the possibility that these properties are due to modification of the substituted Cys residue, and do not reflect its properties in its unmodified state, by evaluating the impact of pretreatment with DTT.
The following experimental results indicate that amino acid residues, Gln 45 , Asn 90 , Leu 290 , Ser 407 , and Asn 411 of human PCFT are in proximity in the inward-open conformation predicted by the homology model. (i) There was spontaneous formation of a cross-link between N90C/S407C based upon the loss of function and lack of accessibility of the pair, both of which were reversed by treatment with DTT. (ii) There was also spontaneous formation of a disulfide bond in all double mutants containing Q45C because the slight increase in function of the Q45C single mutant induced by DTT could not account for the much greater increase in function of double mutants (Q45C/N90C-DSL, Q45C/S407C-DSL, and Q45C/ N411C-DSL) that occurred with the same DTT treatment. (iii) Because the activity of the single L290C mutant could be fully restored, and the modification of the introduced Cys residues could be fully reversed by treatment with DTT, it was possible to study the impact that CuPh and Cd 2ϩ had on the Cys pairs that contained this residue. Hence, the most substantial loss of function was observed for the Leu 290 /Asn 411 Cys pair and a lesser, although still significant loss of function, for the L290C/ S407C pair. (iv) Significant loss of activity was observed for all Cys pairs containing Q45C with CuPh oxidation, and for all but Q45C/L290C with Cd 2ϩ , whereas neither Q45C nor any of the other single mutants lost activity upon these treatments.
Consistent with the role of TMD-1, -2, -7, and -11 in the formation of an extracellular gate, when Cys pairs within these TMDs were cross-linked, the influx V max was decreased (Table 1) indicative of a decrease in the rate of conformational change between the inward-and outward-facing transporter configurations. Furthermore, when folic acid was bound to the carrier the CuPh-induced decrease in transport for three of the Cys pairs (Fig. 6) was diminished or completely prevented. This is consistent with substrate-induced closure of the external gate resulting in decreased accessibility of the CuPh to the Cys-substituted residues as predicted by the alternating access mechanism of transport. Alternatively, as proposed for the external gate defined for the L. donovani nucleoside transporter 1.1, the presence of substrate may increase the distance between Cys residues during a portion of the transport cycle (48).
The formation of a disulfide bond or complex between N411C or S407C and the same residues on other helices is consistent with their positions within the context of the helix turn, the observation that both are accessible to MTSEA biotin, and the flexibility of the carrier as its oscillates between its conformational states. Although, in general, the pattern and magnitude of the effects of Cd 2ϩ were comparable with the effects of CuPh, there were exceptions. The impact of Cd 2ϩ on transport function was considerably less than that of CuPh for the Q45C/ L290C and N90C/L290C mutants. Because Cd 2ϩ should complex Cys residues further apart than occurs with a disulfide bond, this might result in less restriction of the complex and less impact on transport function. The interaction between the Cys residues appeared to be specific because the activity of all single mutants after DTT reduction was not affected by CuPh oxidation nor the presence of Cd 2ϩ . This indicates further that there was no interaction between substituted Cys residues in each single mutant with any of the remaining five native Cys residues in PCFT that are located within TMD4, TMD6, TMD11, the N terminus, and the intracellular loop between TMD-8 and -9.