Functional Role of Arginine 373 in Substrate Translocation by the Reduced Folate Carrier*

The reduced folate carrier (RFC) plays a critical role in the cellular uptake of folates. However, little is known regarding the mechanism used to transport substrates or the tertiary structure of the protein. Through the analysis of a Chinese hamster ovary cell line deficient in folate uptake, we have identified a single residue in TM10 (Arg-373) of RFC that appears to play a critical role in the translocation of substrate. Replacement of this position with various amino acids (KHQNA) diminished the rate of translocation by 16–50-fold, although substrate binding, protein stability, and localization were unaffected. Furthermore, the translocation capabilities of an R373C mutant in a cysteine-less form of the reduced folate carrier were enhanced 2.5-fold by the positively charged methanethiosulfonate reagent, confirming the essential role of a positive charge at this position. When considering the membrane-impermeable nature of this reagent, the data further suggest that the Arg-373 residue is located within the substrate translocation pathway of the RFC protein. Moreover, cross-linking analysis of the Arg-373 residue demonstrates that it is within 6 Å of residue Glu-394 (TM11), providing the first definitive tertiary structural information for this protein.

The reduced folate carrier (RFC) plays a critical role in the cellular uptake of folates. However, little is known regarding the mechanism used to transport substrates or the tertiary structure of the protein. Through the analysis of a Chinese hamster ovary cell line deficient in folate uptake, we have identified a single residue in TM10 (Arg-373) of RFC that appears to play a critical role in the translocation of substrate. Replacement of this position with various amino acids (KHQNA) diminished the rate of translocation by 16 -50-fold, although substrate binding, protein stability, and localization were unaffected. Furthermore, the translocation capabilities of an R373C mutant in a cysteine-less form of the reduced folate carrier were enhanced 2.5-fold by the positively charged methanethiosulfonate reagent, confirming the essential role of a positive charge at this position. When considering the membrane-impermeable nature of this reagent, the data further suggest that the Arg-373 residue is located within the substrate translocation pathway of the RFC protein. Moreover, crosslinking analysis of the Arg-373 residue demonstrates that it is within 6 Å of residue Glu-394 (TM11), providing the first definitive tertiary structural information for this protein.
Folates are essential compounds required by mammalian organisms for numerous biosynthetic pathways, including the synthesis of pyrimidines, purines and several essential amino acids (1). These nutrients are transported into the cell primarily by the reduced folate carrier (RFC) 1 system. This transporter has been implicated in clinical resistance to the chemotherapeutic drug, methotrexate (Mtx) (2)(3)(4), further underlining the importance of characterizing the RFC protein and its mechanism of transport.
The RFC protein is consistent with a predicted 12-transmembrane (TM) topology and cytoplasmically located N and C termini, as determined by epitope mapping (5) and cysteine scanning. 2 Recent work has also demonstrated that these cytoplasmic termini, as well as the loop between TM6 and TM7, do not appear to play a direct role in protein function although they are essential for protein stability and trafficking (6,7). The characterization of various mutations in the RFC protein using human, mouse, and hamster systems has demonstrated that single amino acid changes can lead to drastic alterations in substrate affinity (8 -15), substrate translocation (16,17), protein stability, and trafficking (18,19). Preliminary analysis has suggested an interaction between amino acid residues in TM2 and TM4 (20) providing some insight regarding the tertiary structure. Overall, however, there is limited information available on the folding of the RFC protein or the mechanism of transporting folates.
In this report, we examined the RFC protein encoded by a folate transport-deficient Chinese hamster ovary (CHO) line. The protein has a single point mutation in the predicted TM10, resulting in the substitution of arginine for histidine (R373H). Functional analysis of modified proteins with amino acid replacements for Arg-373 indicates that this residue plays a critical role in substrate translocation and may form part of the translocation pathway. Furthermore, the Arg-373 (TM10) residue and another (Glu-394; TM11) are shown to be within close proximity of each other, presenting the first definitive tertiary structural information for the RFC protein.
Cell Lines-The maintenance of clonal cell lines of CHO wild-type Pro Ϫ 4, and mutant Mtx-resistant Pro Ϫ 3 MtxRII 5-3 (MtxRII 5-3) have been described previously (22,23). The Pro Ϫ 4 MtxRII 4-5 (MtxRII 4-5) cell line was generated using single-step selection after ethyl methane sulfonate mutagenesis in a manner similar to that used for the MtxRII 5-3 line, except it was derived from a different parental line (Pro Ϫ 4). Neither the MtxRII 5-3 nor the MtxRII 4-5 lines transport Mtx and, although the former contains no detectable rfc message by Northern analysis (24), the latter has levels similar to those in wild-type cells (25).
Dot Blot Analysis-Genomic DNA was isolated from cell lines as * This work was supported by an operating grant (to W. F. F.) and a studentship (to H. S.) from the Canadian Institute of Health Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed.  (26) and digested overnight with BamHI restriction endonuclease. The DNA was denatured with 3 M NaOH prior to loading onto nitrocellulose membrane in dot blot apparatus. Samples were washed with sodium acetate and baked onto membranes, which were subsequently hybridized using 32 P-labeled rfc-specific probes as described previously (24). After autoradiography, membranes were stripped and hybridized with dihydrofolate reductase (dhfr)-specific probes.
Constructs-Site-directed mutations were incorporated into the hamster rfc cDNA background using either the Stratagene QuikChange kit or a two-step PCR method. In the latter case, complementary primers encoding the desired sequence were designed and used in combination with an upstream forward primer or a downstream reverse primer, respectively, in two separate reactions. After the first round of amplification, the two PCR products were diluted 50-fold, mixed together, and reamplified using the same nonmutagenic upstream and downstream primers. These fragments were cloned into pGEM-T (Promega) and, along with appropriate restriction endonucleases, were used to replace the homologous region of the hamster RFC-EGFP backbone in pCDNA 3 (19). The C termini of all proteins were tagged with enhanced green fluorescent protein (EGFP), which had previously been shown not to affect the properties of RFC (6).
The cysteine-less RFC was generated as described elsewhere. 2 Factor Xa sites were inserted into the double cysteine mutants using one of two methods that did not otherwise interfere with the coding region: 1) a blunt-cutting enzyme site was generated at the desired location by site-directed mutagenesis, and a primer cassette containing two tandem factor Xa recognition sites was inserted (ATC GAG GGA CGC ATC GAG GGT AGG); or 2) primers with 5Ј extensions of tandem factor Xa recognition sites were used as described in the Stratagene QuikChange kit. The factor Xa site was inserted into the D86C,R373C mutant in the large central loop (P231), although it was located within inner loop 5 (Ile-387) of the R373C,E394C mutant. All mutations were sequenced to confirm the presence of alterations.
DNA Transfections-Transfection of the construct plasmids into the recipient MtxRII 5-3 cells was performed using 10 g of purified DNA in polybrene/1 ϫ 10 5 cells as described previously (24). After transfection, the cells were selected for growth in either low levels of folinic acid (4 nM) or in normal medium containing 1.2 mg/ml G418 as described previously (5). Colonies were picked from individual transfection experiments and cloned by limiting dilution; those under G418 selection were isolated based on EGFP fluorescence. In some cases, mutants were selected and cloned under higher levels of folinic acid (10 nM). At least two independently generated isolates from separate transfection experiments with each construct were used for analysis. As both isolates for each construct showed similar characteristics, representative data from only one of each type are shown.
Northern Analysis-Poly(A ϩ ) RNA (ϳ5 g) was separated on a 1.2% agarose gel in formaldehyde buffer and transferred to Hybond XL membrane by the capillary method as described (27). Membranes were subsequently hybridized using 32 P-labeled rfc-specific probes, washed, and exposed to x-ray film as described previously (24). Membranes were then stripped and hybridized with dhfr-or actin-specific probes.
Dose Response-Clonal cell lines stably expressing the appropriate RFC molecule were incubated in varying concentrations of Mtx or folinic acid for 7 days. At this point, cells were stained with methylene blue, and colonies were counted. Dose-response curves were generated and D 10 (concentration allowing 10% survival under Mtx selection), or D 50 (concentration allowing 50% growth with folinic acid) values were determined.
Folate Binding and Uptake-Kinetic analyses for the determination of V max and K t for [ 3 H]Mtx were carried out as described previously (21,28) with the exception that cells were grown in flasks until ϳ90% confluent and then gently trypsinized. After two washes in phosphatebuffered saline (PBS) and a third in uptake buffer (0.15 M Hepes, pH 7.4, 1 mM MgCl 2 ), cells were resuspended to ϳ1 ϫ 10 7 or ϳ3 ϫ 10 6 cells/ml for binding and uptake analyses, respectively. The V max values were normalized relative to surface protein and corrected for cellular expression levels (see "Results"). At least two independent sets of data were analyzed for significant differences using the two-tailed Student's t test.
To assess the effects of MTS reagents on [ 3 H]Mtx uptake, cells expressing cysteine-less RFC-EGFP or cysteine-less RFC-EGFP with an R373C substitution were used. Adherent cells (ϳ7 ϫ 10 5 /sample) were washed twice in uptake buffer, and the appropriate MTS reagent was added to a final concentration of 10 mM. After gentle shaking for 20 min at 37°C, the buffer was removed and At the appropriate time points, the [ 3 H]Mtx was removed by aspiration and the cells lysed in 100 mM NaOH at 4°C. Samples were assayed for radioactivity, and the radiolabeled drug (pmol) taken up was normalized to protein content.
Fluorescence Detection of GFP-For confocal microscopy, cells were grown on glass coverslips and stained with the tracking reagent (brefeldin A BODIPY 558/568 conjugate) as described previously (19). Detection was performed on an LSM410 inverted Zeiss laser scanning microscope with LSM410 software using a krypton/argon laser and a 63ϫ oil immersion lens under standard conditions. Images were taken at various integration values to give the clearest representation of the fusion protein distribution because there were variations in the intensities of the signals from various cell lines.
Analysis of Turn-over-For examination of protein turn-over, 0.2 mg/ml cycloheximide was added to cells stably expressing RFC-EGFP mutant proteins in order to inhibit further protein synthesis. At various time points, cells were lysed on ice (1% Nonidet P-40, 0.1% sodium dodecyl sulfate, 0.5% sodium deoxycholate, 50 mM Tris, pH 8, 150 mM NaCl, 0.2 mM orthovanadate, 0.5 mM phenylmethylsulfonyl fluoride, 5ϫ protease inhibitor mixture (Roche Molecular Biochemicals)) as described previously (19). After separation of lysates by SDS-PAGE, Western analysis was used to detect the EGFP-tagged proteins as described previously (19). Images were scanned, bands quantitated using the BioRad Multi-Analyst program, and t1 ⁄2 determined by linear regression of band intensities. Only images in the linear range of the film were quantitated.
Cell Surface Biotinylation-Equal numbers of cells (ϳ7 ϫ 10 5 ) expressing the various RFC mutations were washed twice with PBSCM (137 mM NaCl, 2.7 mM KCl, 8 mM Na 2 HPO 4 , 1.5 mM KH 2 PO 4 , 1 mM MgCl 2 , 0.1 mM CaCl 2 ) and biotinylated with 1.5 mg/ml biotin succinimide for 1 h at room temperature. Cells were washed twice in 100 mM glycine/PBSCM and once in ice-cold PBS before lysis on ice as described above. The protein concentration of lysates was determined using the Bradford assay, and equal amounts of total protein were incubated with an excess of monoclonal antibody to GFP (Molecular Probes; 3E6). Protein G-Sepharose beads (Amersham Biosciences) were added and, after a 2-h incubation at 4°C, were washed twice (0.1% Nonidet P-40, 50 mM Tris, pH 8, 50 mM NaCl, 1ϫ protease inhibitor mixture). Beads were resuspended in SDS loading buffer, heated at 55°C for 20 min, and separated by SDS-PAGE before electrophoretic transfer to nitrocellulose. Membranes were blocked in 5% bovine serum albumin/TBS-T (20 mM Tris-HCl, 137 mM NaCl, 0.2% Tween 20, pH 7.6) prior to detection with streptavidin-HRP (diluted 1:2000 in 0.5% bovine serum albumin/TBS-T) and subsequently visualized using ECL TM reagent. Blots were then stripped and probed for fusion protein expression with EGFP-specific antibody as described previously (19). Images were scanned and bands quantitated as described above.
Cross-linking of Membranes-Membranes were isolated as described (19) from ϳ1 ϫ 10 7 cells stably expressing double-cysteine factor Xa constructs and quantitated using the Bio-Rad Bradford assay. Samples to be cross-linked (20 g) were resuspended in TNE buffer (10 mM Tris-HCl, pH 8.0, 10 mM NaCl, 10 mM EDTA) at 0.2 mg/ml. Membranes were incubated with 1 mM N,NЈ-o-phenylenedimaleimide (o-PDM), diluted from a 50 mM stock in N,N-dimethylformamide. For oxidative treatment of membranes, solutions of copper sulfate (200 mM) and 1,10 phenantholine (600 mM) were prepared in water just prior to use and then mixed in an equal ratio for a 200 mM solution of Cu(phenantholine) 3. Membranes were rotated with 2 mM Cu(phenantholine) 3 for 30 min at room temperature.
After being washed once with factor Xa digestion buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 mM CaCl 2 ), membranes were incubated in the same buffer with 1.2% digitonin and 2 units of factor Xa overnight at 4°C. Some Cu(phenantholine) 3 -treated samples were incubated with 200 mM dithiothreitol at room temperature for 30 min. Samples (5 g) were analyzed by SDS-PAGE and Western analysis as above.

Identification of Alterations in the rfc Gene-The
MtxRII 4-5 cell line was selected based on its ability to survive in concentrations of Mtx that are normally toxic to wild-type cells. This phenotype was attributed to a reduction of drug permeability (22), which was confirmed in a preliminary analysis of radiolabeled Mtx uptake. Furthermore, although this cell line demonstrates a similar level of Mtx resistance as found in the RFC-deficient line (MtxRII 5-3), it is able to survive at much lower (ϳ5-fold) levels of folinic acid (Table I). As RFC is a primary route of reduced folate and Mtx influx, the rfc gene was examined for the basis of these characteristics.
Using cycle sequencing, a single G3 A mutation at nucleotide 1218 was identified in the MtxRII 4-5 cDNA, which leads to a R373H substitution in the gene product. The entire rfc coding region was subsequently PCR-amplified from MtxRII 4-5 genomic DNA and, although sequencing of multiple products confirmed the alteration at 1218 bp, no further mutations were identified. As there are normally two rfc alleles in CHO cells (29), the presence of a single mutation indicated that either both alleles are similarly altered or there is only a single copy of the gene. This latter possibility was addressed by dot blot analysis of the genomic DNA, where the intensity of rfcspecific signal was compared among the mutant cell line and two wild-type lines (Fig. 1A). Normalization of the signal to the dihydrofolate reductase gene confirmed that the MtxRII 4-5 cell line has a single copy of the rfc gene.
Functional Characterization of the R373H Mutation-The functional effects of the arginine to histidine substitution were examined by transfecting a construct encoding the R373H mutant with a C-terminal green fluorescent protein fusion (referred to as R373H-EGFP) into the RFC-deficient cell line (MtxRII 5-3). Under folinic acid-restricted (4 nM) conditions, fewer colonies were observed as compared with wild type (Arg-373-EGFP) ( Table I). Although this finding indicates that the mutant protein has some ability to transport substrate, it is contrary to the phenotype of the parental MtxRII 4-5 cell line, which demonstrates approximately a 20-fold higher requirement for folinic acid than the Arg-373-EGFP line. Expression levels could account for this discrepancy, and thus Northern analysis was used to examine the rfc message. As described previously (24), the recipient cell line, MtxRII 5-3, has undetectable levels of rfc-specific message, whereas the MtxRII 4-5 line is similar to that of wild-type cells (Fig. 1B). However, the rfc mRNA levels in two independent cloned cell lines stably expressing the R373H construct were much higher (20 -50-fold) as compared with both wild type and the parental MtxRII 4-5 cells (Fig. 1C). Taken together, it appears that when overexpressed, the mutant protein can provide the cell with sufficient levels of folate to survive (see kinetic analysis of Arg-373 substitution). This is supported by the observations that the folinic acid requirement for growth as well as the Mtx sensitivity of cells stably expressing the R373H-EGFP mutant are similar to those found in cells expressing Arg-373-EGFP (Table I).
Site-directed Mutagenesis of Arg-373 and Functional Characterization-To clarify the role of the Arg-373 residue in protein function, site-directed mutagenesis was used to replace the arginine with lysine, glutamine, asparagine, glutamate, and FIG. 1. Comparison of rfc gene and mRNA levels. A, genomic DNA was isolated from cell lines, and 2-fold dilutions from 10 to 0.63 g were dotted onto nitrocellulose membrane (see "Experimental Procedures"). Blots were probed using 32 P-rfc cDNA followed by autoradiography. Blots were stripped and reprobed with radiolabeled dhfr cDNA, and the signal was quantitated. Pro  alanine; this allowed an examination of the size, charge, and polarity requirements at this site. Each mutant was tagged at the C terminus with the EGFP protein and transfected into the RFC-deficient cell line (MtxRII 5-3). Table I summarizes the transfection frequencies under low folate growth conditions as well as the folinic acid requirements and Mtx sensitivity of stable transfectants. The alteration to lysine, similar in size and charge to arginine, resulted in a comparable ability to rescue the RFC-deficient cell line as Arg-373-EGFP. Additionally, the folinic acid requirement and Mtx sensitivity of cells stably expressing R373K-EGFP suggest that this altered protein is efficient at transporting both substrates. In fact, the sensitivity to Mtx was ϳ3-fold greater than in Arg-373-EGFPexpressing cells.
The transfection frequency of R373Q-EGFP is lower than that of R373H-EGFP, whereas the R373A-EGFP and R373N-EGFP constructs required a higher level of folinic acid (10 nM) to demonstrate a significant ability to rescue the RFC-deficient cell line. However, although the cells stably expressing the R373N-EGFP require a ϳ20-fold higher level of folinic acid than Arg-373-EGFP, the R373H-EGFP, R373Q-EGFP, and R373A-EGFP mutants are similar to Arg-373-EGFP (Table I).
Although the R373H-EGFP and the R373Q-EGFP cells show an Mtx sensitivity similar to that of Arg-373-EGFP, the R373A-EGFP and the R373N-EGFP are more resistant with the latter demonstrating a phenotype similar to the RFC-deficient cell lines. The R373E-EGFP and R373D-EGFP mutant constructs were unable to rescue the RFC-deficient cell line (MtxRII 5-3) under either of the folate-selective conditions (4 or 10 nM), and both the folinic acid requirement and Mtx sensitivity of R373E-EGFP-expressing cells (selected in G418) resembled that of the recipient cells (Table I).
Kinetic Analysis of Arg-373 Substitution-The kinetic parameters of Mtx uptake were determined in order to elucidate the basis for the differences in transfection frequency of the replacement amino acids as compared with Arg-373-EGFP. The substitution to lysine seems to increase the affinity for Mtx by ϳ4-fold; however, this appears to be due to unique properties of this residue, as the K t for the R373H-EGFP, R373Q-EGFP, and R373A-EGFP mutants are very similar to Arg-373-EGFP. Furthermore, although the R373E-EGFP mutant is transport-defective, it demonstrates a binding affinity for Mtx (K d ϭ 3.1 M) that is similar to Arg-373-expressing cells (K d ϭ 2.7 M). Thus, it appears that the 373 residue makes little contribution to the substrate-binding site of the RFC protein.
Interestingly, the V max values for Mtx uptake of all the 373-aa mutants are significantly reduced (2-7-fold) when compared with Arg-373-EGFP. To determine whether this was due to variations in cell line expression levels or the amount of protein at the cell surface, the values were normalized as follows. Cells stably expressing each of the mutant proteins were labeled with a membrane-impermeant amide-reactive biotinylation reagent, and the fusion proteins were immunoprecipitated from equal amounts of cellular lysate using an excess of monoclonal antibody. Subsequent Western analysis (see "Experimental Procedures") allowed a relative quantification of the amounts of both biotinylated (cell surface) and fusion protein (Fig. 2, Table II). After normalizing for differences in fusion protein levels, it appears there is Ͻ2-fold variation in the proportion of protein at the cell surface for the mutant cell lines as compared with Arg-373-EGFP; the lowest values belong to those proteins most impaired in folate transport (Table  II). However, there are significant differences in cellular EGFP-protein expression levels relative to Arg-373-EGFP that, when taken into account, decrease the V max values of the mutants further (Table I). Thus, it appears that the alterations to the 373 residue lead to a 16 -50-fold reduction in the efficiency of substrate transport.
Role of Arg-373 in Protein Biogenesis-The effect of the substitutions on protein localization was evaluated using confocal microscopy to visualize cells expressing the various fusion proteins (Fig. 3). The majority of the Arg-373-EGFP protein appears to be at the plasma membrane, with a lesser amount co-localizing with an endoplasmic reticulum-Golgi-specific stain (brefeldin A BODIPY). The localization of each of the 373 mutant proteins is very similar, including the transport-defective R373E-EGFP mutant with a reversal of the wild-type charge. It should be noted that the EGFP signal intensities do not reflect relative expression levels between the cell lines (see "Experimental Procedures").
The stability of the 373 mutant EGFP fusion proteins was evaluated over time and was determined not to be significantly different from Arg-373-EGFP (Fig. 4). These Western analyses further confirm that the majority of each of the mutant proteins are able to obtain a complex glycosylation pattern, with a small amount (ϳ2-5%) that is core-or unglycosylated (19). Based on these analyses, it is apparent that the 373 residue does not play a fundamental role in RFC protein stability or localization.
Arg-373 Appears to be Part of the Translocation Pathway-Although the RFC topology predictions place Arg-373 within the 10th transmembrane domain, functional analyses of the various amino acid substitutions indicate that the 373 position requires a large, positively charged residue for optimum transport capabilities. The presence of a charged residue within a hydrophobic lipid bilayer is energetically unfavorable and can be tolerated only if neutralized by a polar interaction with another amino acid or by exposure to the hydrophilic environment, perhaps through a channel or pore. The latter possibility was examined by attempting to determine whether the 373 residue was accessible to MTS reagents. These small membrane impermeable compounds react specifically with cysteines, and can impart a positive (MTSET) or negative (MTSES) charge on the residue. A cysteine-less form of the RFC protein (Cys-RFC-EGFP) that retains function was generated recently. 2 An R373C substitution was made in this backbone and subsequently transfected into the RFC-deficient cells. A single colony was ex- panded from the small number obtained with increased levels of folinic acid (20 nM) for growth and evaluated for the ability to take up [ 3 H]Mtx (Fig. 5). As compared with cells expressing the Cys-RFC-EGFP, the amount of drug transported by the R373C-EGFP protein was decreased by Ͼ90%. However, when these cells were treated with the MTSET reagent, there was a 3-fold increase in the accumulation of the Mtx over time. This effect was unique to the MTSET (ϩ) compound, as the MTSES (Ϫ) treatment had little effect. These results further confirm that for substrate translocation a positive charge at position 373 is required.
Identification of Charged Residues Potentially Interacting with Arg-373-The possibility that another polar residue may be interacting with aa 373 was also evaluated. As the topology of the RFC protein is not defined precisely, it is difficult to a Numbers were generated from images in which the EGFP signal was within the linear range of the film. Transfected protein levels were quantitated using EGFP signal intensities and expressed relative to Arg-373-EGFP.
b Values indicate the normalized amount of surface protein when cellular EGFP fusion expression levels are equivalent to Arg-373-EGFP (ratio of biotinylated protein to biotinylated Arg-373-EGFP/column 1).
c Values indicate the relative amount of surface protein expression corrected for cellular expression levels (column 1 ϫ column 2).

FIG. 3. Confocal microscopy of 373 mutant RFC cell lines.
The left image in each panel is EGFP fluorescence, and the center image is specific staining of the endoplasmic reticulum and Golgi complex with brefeldin A BODIPY as described under "Experimental Procedures." The right panel is an overlay of the two preceding images.

FIG. 4. Turn-over analysis of aa 373 mutant EGFP-tagged molecules.
Clonal cell lines expressing the various aa 373 mutant RFC molecules were incubated with cycloheximide to inhibit protein synthesis. Samples were harvested over a 24-h period, as indicated by the numbers at the top, and subjected to Western analysis as described under "Experimental Procedures." The numbers on the right indicate an approximate half-life (t1 ⁄2 ). *, indicates core-or unglycoslyated protein. Upon longer exposure of R373, these forms were also evident. identify those residues on a similar plane within the lipid bilayer as the 373 position. Thus, three conserved and negatively charged residues predicted to be within the membrane were chosen for further analysis: Asp-86, Glu-394, and Asp-453 (Fig. 6).
Each of the candidate interacting positions was mutated to arginine, tagged with the EGFP protein, and examined for its ability to rescue the RFC-deficient cell line (MtxRII 5-3) (Table  III). If the residue is involved in a critical interaction, then reversing the charge should affect functionality. Additionally, because the protein may regain transport capabilities if an interaction has been maintained within the molecule, each of the arginine substitutions was paired with the nonfunctional R373D mutation and examined. Each of the single alterations demonstrated some ability to rescue the RFC-deficient cell line (Table III), although the D86R mutation led to a significant reduction in transfection frequency. None of the paired constructs was able to rescue the nonfunctional R373D mutation.
Functional Characterization of the Asp-86 and Glu-394 Residues-As there was nearly a 90% decrease in the number of functional transfectants with the alteration to D86R, it appears this residue is important for substrate transport. This is supported by the significant reduction in transfection frequency when this position was changed to a structurally similar (D86E) or uncharged (D86A) amino acid (Table IV). Furthermore, although the reversal of both the 86-and 373-aa charges in the same molecule did not restore function, an interaction between these two residues cannot be ruled out; an opposite charge may have significant secondary effects in a different local environment. Thus, both the 86 and 373 residues were mutated to alanine in the same molecule, as removal of both charges may be less disruptive. For comparison, the E394 residue was also altered to alanine and paired with R373A, as the E394R substitution only depressed the transfection frequency by 55%. Surprisingly, both combinations showed a level of functionality greater than with R373A alone (Table IV). Kinetic analysis of Mtx uptake indicated that the D86A replacement in the same molecule as the R373A mutation (normalized V max ϭ 0.10 pmol/min/mg) restored some of the substrate translocation efficiency of the latter alteration alone (normalized V max ϭ 0.02 pmol/min/mg). In contrast, the combined alanine mutations in the R373A,E394A,EGFP construct had little effect (normalized V max ϭ 0.03 pmol/min/mg).
Cross-linking Analysis of Asp-86 and Glu-394 -An alternative approach was used to evaluate the potential Arg-373 interactions between Asp-86 or Glu-394 utilizing the cysteineless form of the RFC protein. In the Cys-RFC-EGFP backbone, double cysteine mutants (D86C,R373C-EGFP and   R373C,E394C-EGFP) were generated with a factor Xa recognition cleavage site inserted between each pair of cysteines (see "Experimental Procedures"). Neither mutant protein was able to rescue the RFC-deficient cell line under low folate-selective conditions. However, Western analysis of clones obtained under G418-selective conditions supplemented with high levels of folinic acid (2 M) indicated that a portion of the RFC-EGFP molecules obtained a complex glycosylation modification (Fig.  7). This finding indicates that these molecules have reached the later stages of the secretory pathway and have likely achieved the correct topology, and thus they are valid to use as models for examining amino acid interactions. Membranes isolated from cells stably expressing the mutant proteins were exposed to oxidative conditions in order to induce disulfide bond formation, or they were reacted with a homobifunctional cross-linking agent (o-PDM; 6 Å in length). Following factor Xa digestion and Western analysis (Fig. 7), it was possible to evaluate the proximity of these amino acids to each other.
For the D86C,R373C-EGFP mutant, neither the addition of the o-PDM cross-linker nor the use of oxidative agents significantly prevented the separation of the factor Xa-treated protein when examined by SDS-PAGE analysis (Fig. 7A). In contrast, although only a small portion of the R373C,E394C-EGFP protein is cleaved by the protease, it is evident that treatment with the o-PDM linker reduces the amount of cleaved product. The abundance of the fully glycosylated R373C,E394C-EGFP protein relative to the cleavage product at ϳ40 kDa is 1:1 in the untreated sample and 9:1 when the membranes are treated with o-PDM (Fig. 7B). This indicates that these two residues are within 6 Å of each other in the tertiary structure of the protein. DISCUSSION A number of functionally disruptive mutations have been identified and characterized within the RFC protein, and yet the mechanism of substrate transport has remained undefined. In this study, a charged residue (Arg-373) that appears to have a key role in the efficiency of substrate translocation has been identified. Furthermore, this residue was demonstrated to be in close proximity to one located 21 residues downstream, providing the first delineation of the RFC protein tertiary structure.
The Mtx-resistant phenotype of the MtxRII 4-5 cell line is the combined result of the loss of one rfc allele, possibly during the initial mutagenesis procedure, and the presence of a mutation in the remaining allele. This point mutation resulted in the substitution of histidine for arginine at position 373. As this amino acid is conserved within the RFC protein throughout eukaryotic organisms including the Caenorhabditis elegans homologue, it is likely to play a pivotal role in defining protein structure or function. Biochemical and functional analysis of an array of amino acids substituted into this position indicated that there are strict requirements to allow any degree of substrate transport. First, the amino acid at position 373 appears to require the ability to form hydrogen bonds, as the frequency of functional transfectants diminishes as the polar tendencies of the residues decline (Arg ϳ Lys Ͼ His Ͼ Gln Ͼ Ala). Second, the amino acid needs to be of a certain size, as exemplified by the transfection frequency of the R373Q and R373N mutant proteins. Although the asparagine residue differs by only a single carbon in length, the transfection frequency is significantly reduced from that of R373Q. However, the alanine substitution was also slightly better tolerated than the asparagine, suggesting that the ability to form hydrogen bonds can be detrimental if the residue is too small due to destabilizing polar interactions within the local environment.
The transfection frequencies are drastically reduced for most of the alterations at aa 373 with the exception of lysine, a residue similar in size and charge to arginine. However, most of the other replacements (R373H, R373Q and R373A) yield cell lines similar to the wild-type in folinic acid requirements and Mtx sensitivity; this seems to be the result of increased EGFP fusion protein expression (3-6-fold) over Arg-373-EGFP lines. The low folinic acid growth conditions select for cells with a high level of protein expression, which can compensate for the decreased rate of substrate transport. Thus, in some cases in which clones were not selected for functionality but merely for G418 resistance carried by the plasmid, the protein expression levels are lower (i.e. R373N-EGFP). The low level of R373N-EGFP expression as compared with the other 373 mutant cell lines is reflected in the folinic acid requirements and Mtx sensitivity.
Based on algorithms, epitope mapping (5), and cysteine scanning, 2 the RFC protein is consistent with a 12-TM topology with aa 373 predicted to be in the 10th transmembrane domain. It appears that the presence of this charged amino acid in the hydrophobic membrane environment may be tolerated as a result of accessibility to the extracellular environment, perhaps as part of the translocation pathway. The kinetic parameters of Mtx uptake demonstrate that any alteration to the Arg-373 residue leads to a 16 -50-fold reduction in the efficiency of substrate translocation (V max ), indicating that this residue has a vital role in the structure or function of the protein. As evaluated by protein stability, cellular distribution, and the portion of total molecules at the plasma membrane, the amino acid at position 373 does not significantly affect the folding or localization of RFC. However, the addition of the positively charged MTSET reagent to cells expressing an R373C-EGFP gene product led to a 2.5-fold increase in the amount of Mtx drug accumulation as compared with nontreated cells, confirming the functional importance of the positive charge at 373.
The 373 residue is predicted to be located within a membrane-spanning segment of the RFC protein and thus should not be accessible to the extracellular environment. Preliminary experiments indicate that this is the case, as a cysteine at this position cannot be labeled by the membrane-impermeable biotin maleimide compound (data not shown). However, although the MTSET reagent is also membrane-impermeable, its small size enables the molecule to permeate protein pores. Taken together with the functional role of this amino acid in substrate transport, it appears the Arg-373 residue forms part of the translocation pathway.
It is difficult to determine conclusively whether the Arg-373 residue interacts with another amino acid based on functional analyses of an array of mutant proteins. It is only with the replacement of Asp-86 or Glu-394 with alanines in combination with R373A to limit potential secondary effects that there is a positive influence on protein function. When transfected into the RFC-deficient cell line, each combination demonstrates an increased transfection frequency as compared with R373A-EGFP alone (3-13-fold), although it is still less than for Arg-373-EGFP. Furthermore, the Mtx uptake kinetics of the cell lines expressing the D86A,R373A-EFGP mutant indicate that this protein has a 5-fold better rate of translocation compared with R373A-EGFP alone. This is suggestive of an interaction between these two residues, although secondary structural effects may also play a role. In contrast, it appears that combining the E394A alteration with R373A has little functional effect on the rate of substrate translocation.
The potential interactions between Arg-373 and the Asp-86 or Glu-394 residues were further examined using a technique previously utilized for the elucidation of membrane protein helix packing (30 -32). Cysteines replaced the residues of interest in a cysteine-less RFC backbone, and a protease site inserted between them was used to evaluate the effectiveness of the cross-linking agents. Previous work has indicated that removal of the cysteines has little effect on functionality of the RFC and thus, should not interfere with helix packing. 2 Based on these data, it is clear that E394 is within 6 Å of the Arg-373 residue in the tertiary structure of the RFC protein. However, there was no apparent cross-linking between the D86C and R373C residues after the addition of the o-PDM cross-linker or under oxidative conditions.
The demonstration that Glu-394 (TM11) is within 6 Å of Arg-373 (TM10) is highly significant, as it provides the first evidence for a tertiary structure for this protein. Furthermore, it reflects studies on the bacterial lactose permease protein, where TM10 and TM11 are also located adjacent to each other (33). The two proteins are structurally similar in many respects, in that they are both polytopic 12-TM domain proteins with a large cytoplasmic loop between TM6 and TM7. As the helix packing of very few polytopic proteins has been documented, it is not clear whether there are common tertiary structural characteristics. A recent report of a potential chargepair interaction in the human RFC protein (20) implicated the residues corresponding to Asp-86 and Arg-131 in the hamster protein.
In an attempt to confirm these observations in the hamster system, the same alterations were made (D86V,R131L). However, neither this, nor a double alanine mutant (D86A,R131A) was able to complement the MtxRII 5-3 cell line under restricted folinic acid growth conditions (data not shown). Based on these two opposing results, it appears that there may be differences in tertiary folding of the RFC protein between the human and hamster species. This is somewhat surprising, as there is a high degree of amino acid similarity (ϳ60%) between the two species as well as similar folate transport properties.
The 373 position has strict charge and size requirements such that none of the amino acids tested was able to provide comparable substrate translocation efficiency. The data presented here suggest that the Arg-373 residue is accessible to the extracellular space and may directly interact with the substrate, although it does not appear to be part of the substrate-binding site. Furthermore, the close proximity of this residue to TM11 and the slight stabilizing effect of the D86A alteration suggest that Arg-373 may also have a structural role in defining the translocation pathway. As the structural nature of the arginine residue allows multiple and simultaneous interactions, it is not improbable that the Arg-373 position in RFC may have many different roles.