Topological and Functional Analysis of the Human Reduced Folate Carrier by Hemagglutinin Epitope Insertion*

The membrane topology of the human reduced folate carrier protein (591 amino acids) was assessed by single insertions of the hemagglutinin epitope into nine sites of the protein. Reduced folate carrier-deficient Chinese hamster ovary cells expressing each of these constructs were probed with anti-hemagglutinin epitope monoclonal antibodies to assess whether the insertion was exposed to the external environment or to the cytoplasm. The results are consistent with the 12-transmembrane topology predicted for this protein. The hemagglutinin epitope insertion mutants were also tested for their effects on the function of the reduced folate carrier. For these studies, each of the constructs had a carboxyl-terminal fusion of the enhanced green fluorescent protein to monitor and quantitate expression. Insertions into the external loop between transmembrane regions 7 and 8 (Pro-297), the cytoplasmic loop between transmembrane regions 6 and 7 (Ser-225), and near the cytoplasmic amino and carboxyl termini (Pro-20 and Gly-492, respectively) had minor effects on methotrexate binding and uptake. The insertion into the cytoplasmic loop between transmembrane regions 10 and 11 (Gln-385) greatly reduced both binding and uptake of methotrexate, whereas the insertion into the external loop between transmembrane regions 11 and 12 (Pro-427) selectively interfered with uptake but not binding.

The membrane topology of the human reduced folate carrier protein (591 amino acids) was assessed by single insertions of the hemagglutinin epitope into nine sites of the protein. Reduced folate carrier-deficient Chinese hamster ovary cells expressing each of these constructs were probed with anti-hemagglutinin epitope monoclonal antibodies to assess whether the insertion was exposed to the external environment or to the cytoplasm. The results are consistent with the 12-transmembrane topology predicted for this protein. The hemagglutinin epitope insertion mutants were also tested for their effects on the function of the reduced folate carrier. For these studies, each of the constructs had a carboxyl-terminal fusion of the enhanced green fluorescent protein to monitor and quantitate expression. Insertions into the external loop between transmembrane regions 7 and 8 (Pro-297), the cytoplasmic loop between transmembrane regions 6 and 7 (Ser-225), and near the cytoplasmic amino and carboxyl termini (Pro-20 and Gly-492, respectively) had minor effects on methotrexate binding and uptake. The insertion into the cytoplasmic loop between transmembrane regions 10 and 11 (Gln-385) greatly reduced both binding and uptake of methotrexate, whereas the insertion into the external loop between transmembrane regions 11 and 12 (Pro-427) selectively interfered with uptake but not binding.
Mammalian cells require reduced folates for several biochemical pathways involving purine, pyrimidine, and amino acid metabolism. Since mammals cannot synthesize folates, these compounds must be obtained in the diet and imported by cells via either the glycosylphosphatidylinositol-linked folate receptor (1)(2)(3)(4)(5) or the reduced folate carrier (RFC), 1 an integral membrane protein (6 -9). A third, low pH-dependent component for folate transport has been reported (10,11), but its involvement in folate delivery is not clearly understood. The RFC is the main route for transport of 5-methyltetrahydrofolate, the major circulating folate in the bloodstream, 5-formyltetrahydrofolate (folinic acid), and the folate analog, methotrexate (Mtx) (9,(12)(13)(14).
The rfc gene has been cloned from hamster (15), human (16 -19), murine (20), and rat 2 sources. The identity of the rfc gene was verified by cDNA transfections that restored Mtx sensitivity to RFC-deficient cells (15)(16)(17)(18)(19)(20). The human RFC protein is 591 amino acids as deduced from its cDNA sequence, and the hydropathy profile predicts 12 transmembrane-spanning (TM) domains with the amino and carboxyl termini located in the cytoplasm. Recently, the cytoplasmic location of the carboxyl-terminal tail has been confirmed for the human RFC (22). The mouse and hamster homologues (518 amino acids each) have significantly shorter carboxyl-terminal tails than the human RFC (15)(16)(17)(18)(19)(20), although all three RFCs share about 68% identical or similar amino acid residues. The human RFC is glycosylated at an asparagine residue in the first extracellular loop (9,22,23), and while the hamster RFC contains the conserved sequence in this loop (15), its glycosylation status has not been confirmed. The murine RFC lacks the conserved glycosylation site and is not glycosylated (20).
The deduced 12-TM topology can place the RFC in a large family of major transport facilitators that includes the bacterial Lac permeases and the mammalian glucose transporters (24,25). The lack of an ATP-binding domain in the RFC suggests that it does not belong to the 12-TM family of ABC transporters that includes the multi-drug resistance protein (P-glycoprotein), and the cystic fibrosis transmembrane conductance regulator (26). The three-dimensional structure is known for only a small number of membrane proteins, although for many others the structures have been deduced by statistical and biophysical models (27,28). The modeled structure or topology for each protein must be verified before it is possible to understand how the extracellular, intracellular, and TM domains contribute to the function of the protein.
The topology of a number of proteins with transmembrane domains has been elucidated by glycosylation scanning mutagenesis (29,30), truncation and fusion strategies (31)(32)(33), and by epitope insertion into putative loops (34 -40). In this study, we have inserted the hemagglutinin epitope (HA) (41,42) of the human influenza virus into nine sites of the human RFC. These HA insertion constructs were transfected into a Chinese hamster ovary (CHO) cell line defective for RFC expression. The accessibility of the epitope to anti-HA monoclonal antibodies was evaluated in permeabilized and non-permeabilized cells to determine its orientation. The results support the predicted 12-TM structure of the RFC. In addition, clones expressing these insertion mutants have been evaluated for Mtx binding and transport.

EXPERIMENTAL PROCEDURES
Chemicals-Restriction endonucleases and modifying enzymes were obtained from Amersham Pharmacia Biotech, except for CpoI (MBI Fermentas). Unlabeled Mtx was obtained from Sigma; [(3Ј,5Ј,7-3 H]Mtx was purchased from Moravek Biochemicals (Brea, CA); N 5 -formyltetrahydrofolate (folinic acid) was purchased from ICN Biomedicals, and Polybrene was from Aldrich. Geneticin® (G418-Sulfate), LipofectAMI-NE TM and Opti-MEM® I were purchased from Life Technologies, Inc. DNA Sequencing-Correct insertion of the HA epitopes was initially confirmed by DNA sequencing using the T7 polymerase kit from Amersham Pharmacia Biotech and using [␣-32 P]dCTP (3000 Ci/mmol) obtained from ICN Biomedicals. The entire sequence of each construct was verified by the ABI 377 sequencing facility in the Robarts Research Institute (London, Ontario, Canada).
Cells and Cell Culture-Wild-type Pro Ϫ3 and Pro Ϫ3 MtxRII 5-3 CHO cells (MtxRII 5-3) were maintained as monolayer cultures in ␣-medium supplemented with 10% fetal bovine serum as described previously (43). MtxRII 5-3 is an Mtx-resistant cell line that is defective in reduced folate transport because it does not express the rfc mRNA (15).
For transient transfections, MtxRII 5-3 cells were grown on sterile round glass coverslips placed in 24-well tissue culture plates and incubated with 0.2 g of plasmid DNA complexed with LipofectAMINE TM . After addition of ␣-medium containing 20% fetal bovine serum to each well, the cells were incubated a further 18 h prior to immunofluorescence labeling and microscopy. Stable transfectants of MtxRII 5-3 cells were obtained using the Polybrene procedure described previously (44). Stable clones of the HA insertion constructs in either the 5Ј⌬RFC or 5Ј⌬RFC-EGFP background (described below) were selected and maintained in folic acid-free medium containing 10% dialyzed fetal bovine serum and 2 nM folinic acid (45). Stable clones of MtxRII 5-3 cells transfected with the 5Ј⌬RFC-EGFP fusion products were also selected with G418 (1.2 mg/ml) in ␣-medium supplemented with 10% fetal bovine serum. Individual colonies resistant to G418 were screened by fluorescence microscopy to ensure expression of the fusion protein.
Three to five clones of each transfectant from the low folinic acid and the G418 selection regimes were isolated by limiting dilution from independently generated transfectants.
The phenotype of transfectants selected for growth in low folinic acid medium were tested for Mtx resistance by dose-response curves, as described previously (43). The G418-selected transfectant clones were used for Mtx uptake and binding assays (46,47). In all experiments, at least two independently selected clones were used for the analyses.
Construction of HA Insertion Mutants-The human influenza virus HA epitope (YPYDVPDYA) was inserted into selected sites of the rfc cDNA by two strategies. In the first approach, based on that described by Canfield and Levenson (34), the cDNA was digested at unique restriction endonuclease sites (ApaI, 254; NotI, 883; CpoI, 1374; StuI, 1570; numbering based on the human rfc cDNA reported in Ref. 16) predicted to be in intra-or extracellular loops. These were blunt-ended using either nuclease S1 or the Klenow fragment of DNA polymerase I, as appropriate to maintain reading frame, and treated with calf intestinal alkaline phosphatase. HA oligonucleotides of the sense strand were synthesized with either one or two additional G deoxyribonucleotides at the 5Ј end (see oligos 1 and 2 of Table I). Each was annealed with an equimolar amount of the antisense HA oligonucleotide (oligo 3) which contains two additional G residues at the 5Ј end. The doublestranded products were blunt-ended either with nuclease S1 or the Klenow fragment of DNA polymerase I, as appropriate to maintain reading frame, treated with T4 polynucleotide kinase, and ligated using molar ratios of insert:linearized plasmid of 5:1 and 1:1. The HA insertion constructs were used to transform XL1-Blue Escherichia coli (Stratagene). Positive clones were selected by hybridization with [␥-32 P]dCTP end-labeled oligo 3 and verified by DNA sequencing. Because of the insertion strategy at the CpoI site (HA-P427), an extra valine residue was inserted at the carboxyl end of the HA epitope.
The second HA insertion strategy was based on the two-step PCR method described by Howard et al. (35), using the primer sets and templates listed in Table II. The PCR reactions contained standard PCR buffer (48), 1 ng of template, and optimized levels of MgCl 2 . In some cases the reactions included 10% Me 2 SO. The PCR cycler parameters were as follows: (i) "hot start": 1 min at 94°C, then 5 min at 80°C, during which the Taq DNA polymerase was added; (ii) 26 cycles of amplification: 1 min at 94°C, 1 min at 52°C, 2.5 min at 72°C; and (iii) final extension for 8 min at 72°C. The products were separated on agarose gels, and the band of the correct size was purified by the Geneclean procedure (Bio 101). The two "first-step" products for each

TABLE II
Oligonucleotides for HA insertions by two-step PCR The forward and reverse primers and template of set 1a were used for PCR amplification of the upstream portion proximal to and including the HA insertion. In a separate reaction, the primers and templates of set 1b were used for PCR amplification of the downstream portion of the RFC distal to and including the HA insertion. The PCR reactions produce double-stranded products that have the HA epitope at their 3Ј and 5Ј termini, respectively. The two products were mixed and used as the template for the second round of PCR, with their only region of homology being the HA epitope. The extension primers from sets 1a and 1b are provided in the second round of PCR to fill in the single-stranded DNA regions flanking the annealed HA region to yield the final PCR product containing the HA-P20 insertion. The other primer sets produced the following insertions: 2 a,b, HA-V152; 3 a,b, HA-S225; 4 a,b, HA-P297; and 5 a,b, HA-Q385. The underlined nucleotides of the HA insertion primers represent the coding region for the HA epitope.

5Ј
- T7 primer p5Ј⌬RFC set were diluted, mixed, and used as the template for the second round of PCR. The extension primers for each reaction were the same as used in the first round of PCR (Table II), and the same PCR conditions were used. The second-step products were purified as above, digested with either HindIII and EcoRI or ApaI and EcoRI, and cloned into the p5Ј⌬RFC and p5Ј⌬RFC-EGFP vectors described below. The HA insertion mutants are named according to the amino acid that precedes the HA epitope (i.e. HA-P20 is the HA epitope inserted after Pro-20).
Construction of p5Ј⌬RFC and p5Ј⌬RFC-EGFP-The expression plasmid pRFC contains nt Ϫ94 to ϩ1937 of the rfc cDNA from pHuMtxT4 (16) cloned into the pcDNA3 expression vector (Invitrogen). This construct lacks most of the rfc 3Ј-UTR including the polyadenylation signal. Plasmid p5Ј⌬RFC contains nt ϩ1 to ϩ1937 of the human RFC. The enhanced green fluorescent protein (EGFP) was fused in-frame to the carboxyl terminus of the rfc gene in p5Ј⌬RFC by replacing the stop codon with a BglII site. Using oligos 4 and 5 (Table I) as PCR primers, the EcoRI-BglII fragment of the rfc (nt 1697-1870) was amplified and cloned into a pcDNA3-based plasmid which contained the BamHI-NotI fragment of pEGFP-N1 (CLONTECH). The resulting plasmid, containing the 3Ј end of the rfc gene fused to the EGFP, was linearized with HindIII and EcoRI, and the HindIII-EcoRI fragment of p5Ј⌬RFC was inserted to produce p5Ј⌬RFC-EGFP.
The HA insertion constructs in the p5Ј⌬RFC and p5Ј⌬RFC-EGFP backgrounds were purified by the Qiagen Plasmid Midi Kit (Qiagen, Chatsworth, CA). The EGFP fusion constructs were used in transfections to determine the cellular localization of the protein and for the Mtx binding and uptake studies; the HA insertion constructs without the fusion were used for anti-HA immunofluorescence microscopy to determine protein topology.
Immunofluorescence  (27). No secondary structure is implied for the loops. The longer transmembrane helices in the shaded region are drawn perpendicular to the membrane surface for clarity, although these helices may actually be shorter or inserted into the membrane at an angle. The HA insertion sites are shown by filled arrowheads. The HA insertion mutant nomenclature indicates the amino acid that precedes the HA epitope. 1, HA-P20; 2, HA-G54; 3, HA-V152; 4, HA-S225; 5, HA-R263; 6, HA-P297; 7, HA-Q385; 8, HA-P427; 9, HA-G492. Sites 2, 5, 8, and 9 are at unique restriction endonuclease sites in the RFC cDNA coding region. The HA epitope was inserted into sites 1, 3, 4, 6, and 7 by a two-step PCR process (see "Experimental Procedures"). HA-G54 temperature with freshly prepared 2% paraformaldehyde in 0.5ϫ PBS. The cells were washed twice with PBS containing 1% bovine serum albumin and 0.02% sodium azide (PBS-plus). Cells in one well of each series were permeabilized with 0.2% Triton X-100® in PBS at room temperature and then washed three times with PBS-plus. All cells were incubated at room temperature with 12CA5 mouse monoclonal antibody (Roche Molecular Biochemicals) and then washed twice with PBSplus. Permeabilized and non-permeabilized cells were washed once with PBS containing 0.05% Triton X-100 to remove nonspecifically bound antibody. The cells were washed twice with PBS-plus and stored overnight at 4°C in PBS-plus. The next day, the cells were incubated at room temperature with goat anti-mouse IgG, fluorescein isothiocyanate conjugate (BIOSOURCE International, Camarillo, CA) in PBS-plus containing 10% goat serum. The cells were washed as before. The coverslips were placed on slides with mounting medium containing the DNA binding fluor DAPI, sealed, and viewed with a Zeiss Axioskop fluorescence microscope with 488 nm excitation. Images were captured using a Sony CCD digital camera and analyzed using Northern Eclipse TM (version 2.0) software (Empix Imaging, Inc., Toronto, Ontario, Canada). Mtx Uptake and Binding-Uptake and binding experiments were performed as described previously (46,47) using EGFP-tagged clones of the HA insertion mutants selected for G418 resistance and green fluorescence at the plasma membrane. Total cellular protein was quantified by the Bradford method (49). To normalize for specific expression of the HA-RFC-EGFP fusion proteins, the mean fluorescence values for each clone was determined by flow cytometry (FACScan, Becton-Dickinson) using MtxRII 5-3 cells as the control for cell size and background fluorescence parameters. The clones expressing the RFC-EGFP fusion proteins had mean fluorescence values at least 15 times greater than the control cells.

5Ј⌬RFC-EGFP Fusion
Proteins-Before examining the topology of the RFC in transfected cells using the HA epitope, the EGFP tag was fused in-frame to the carboxyl terminus of various constructs containing the wild-type RFC. This was done to determine whether the EGFP moiety affected RFC function, as it was desirable to have a tag to monitor protein expression and cellular location. MtxRII 5-3 cells were transfected with plasmid DNA from either RFC, RFC-EGFP, 5Ј⌬RFC, or 5Ј⌬RFC-EGFP constructs and selected in low folinic acid medium (45). There was no significant difference in numbers of colonies obtained for these four constructs, indicating that neither the 94 base pairs of the rfc 5Ј-UTR nor the EGFP fusion affect the ability of the transfected RFC to complement functionally the RFC-deficient cell line (Table III). Colonies selected from these transfections exhibited wild-type sensitivity to Mtx, as expected (data not shown). In a second set of experiments (not shown), kinetic analysis of [ 3 H]Mtx uptake was carried out to determine the apparent K t and V max values (46). For these experiments, stable clones that expressed either the human RFC, the RFC-EGFP fusion, or 5Ј⌬RFC were selected for growth in low levels of folinic acid. The cells transfected with the RFC construct had an apparent K t of 2.0 Ϯ 0.6 M and V max of 3.3 Ϯ 1.2 pmol/min/mg total protein for transporting Mtx. The cells expressing the RFC-EGFP fusion protein had similar kinetic parameters with an apparent K t of 1.2 Ϯ 0.7 M and V max of 5.0 Ϯ 1.0 pmol/min/mg total protein. The values for cells expressing 5Ј⌬RFC were K t of 1.7 Ϯ 0.7 M and V max of 5.0 Ϯ 1.2 pmol/min/mg total protein. Based on these results neither the removal of the 5Ј-UTR nor the carboxyl-terminal EGFP fusion affected the functionality of RFC, confirming the suitability of this tag as a passive reporter. The EGFP moiety can then be used to monitor the expression of the RFC, since fluorescence should only occur if the fusion protein were properly translated. In addition, since the fusion protein was shown to be functional, it must be properly localized, and thus the fluorescence could serve as an indicator of the cellular location of the fusion protein (see below).
Epitope Insertion Analysis of RFC Topology-The epitope insertion approach was chosen for examining the membrane topology of RFC because it also allows an assessment of the functional consequences of disrupting the chosen insertion site. The nine insertion sites chosen in the RFC (Fig. 1) were the largest inner or outer loops predicted by the TMpred algorithm (27) as well as the amino-and carboxyl-terminal tails. In addition, these regions contain less than 50% amino acid conservation among the hamster, mouse, and human homologues.
The HA insertion constructs (lacking EGFP) and control plasmids were transiently transfected into MtxRII 5-3 cells. The accessibility of the HA epitope to antibody binding was compared in permeabilized and non-permeabilized cells (Fig.  2). If the RFC is properly localized to the plasma membrane, one would expect to see green fluorescence at the periphery of  (22). This plasma membrane localization was seen in most of the transfectants analyzed in this study. In the cells expressing the constructs HA-P20, HA-V152, HA-S225, HA-Q385, or HA-G492, the HA epitope is clearly facing the cytoplasm since fluorescence at the membrane surface is only detectable in the permeabilized cells. For the HA-Q385 construct, the fluorescence was mainly membrane-localized but punctate in nature. In the case of HA-V152, few cells showed fluorescence. In the cells transfected with the HA-R263 construct, epitope detection required permeabilization. In this case, few cells were detected in which the fluorescence was located at the cell membrane, and unlike most of the other transfectants, this was distributed in a punctate manner. This could be due to inaccessibility of the epitope to the anti-HA antibodies, lack of proper expression of the protein, or to protein degradation (see below). The HA epitope in constructs HA-G54, HA-P297, and HA-P427 is on the external face of the membrane, since it is accessible to anti-HA antibody in both non-permeabilized and permeabilized cells. In the cells expressing HA-P427, fluorescence was detected in the non-permeabilized case but was somewhat diffuse and punctate in nature with a majority of cells showing a membrane localization of the fluorescence. Permeabilization of the cells resulted in a clearer membrane localization of the fluorescence which was still somewhat punctate. The reasons for this are not clear but may represent a reduced accessibility of the antibodies to the epitope. These epitope orientations are consistent with the predicted protein topology presented in Fig. 1.
Transient transfections can result in large variations in expression and can affect protein localization, leading to intracellular inclusions (51,52). To ensure that the determined HA epitope orientations were a true reflection of RFC topology, the experiment was repeated using stable clones of cells selected in low folinic acid medium and expressing either the HA-P20, HA-G54, HA-S225, HA-P297, or HA-G492 constructs. A similar pattern of localization was found for the stable clones, although there was a stronger, more uniform detection at the plasma membrane and fewer intracellular inclusions than their transient counterparts (data not shown).
To ensure that the HA-tagged constructs were properly localized, transient transfections were also carried out with the HA-RFC-EFGP fusions to permit an analysis independent of antibody detection. In general, the EGFP moiety had little or no discernible influence on the cells expressing constructs HA-P20, HA-V152, HA-S225, HA-P297, HA-Q385, HA-P427, or HA-G492 since the fluorescence patterns obtained from the HA detection and the respective EGFP fusions of each construct (without antibody detection) had very similar distributions (Fig. 2). Transfections with either the HA-G54-EGFP or HA-R263-EGFP constructs yielded very few fluorescent cells.
Transfectants bearing these nine HA-RFC-EGFP constructs were selected in G418, thus using a selection criterion independent of RFC functionality. In the initial transfections, 20 -40% of the G418-resistant colonies showed fluorescence at the plasma membrane. This enabled screening for expressing cells that were subsequently cloned. As expected, most clones showed a predominant plasma membrane localization of fluorescence except for those expressing HA-G54-EGFP, HA-R263-EGFP, or HA-V152-EGFP. In the former two, the cells showed diffuse fluorescence throughout the cytoplasm and nucleus, whereas in the latter one, a majority of the cells contained diffuse fluorescent material with a minority fluorescing primarily at the plasma membrane (Fig. 3, panel C). The diffuse fluorescent pattern is similar to the distribution observed with the soluble EGFP control (Fig. 2, panel A).
The large perinuclear inclusions in cells expressing either HA-P297-EGFP (Fig. 3, panel F) or HA-P427-EGFP (Fig. 3, panels H and I) appear to result from trapping in the Golgi as determined by use of a medial Golgi-specific antibody (MG-160, data not shown). In addition, there are also many smaller inclusions throughout the cytoplasm of these cells (Fig. 3,  panel J).
Analysis of RFC Function-To examine the effect of the epitope insertions on RFC function, the transfected cell lines were evaluated for their ability to grow in low folinic acid, to take up [ 3 H]Mtx, and to bind [ 3 H]Mtx.
The insertion of the HA epitope at four sites (HA-P20, HA-S225, HA-P297, and HA-G492) had little discernible effect on the ability of RFC to rescue MtxRII 5-3 cells under low folinic acid conditions (Table III), whereas the efficiency of colony formation with the HA-G54 construct was reduced 5-6-fold. Clones of HA-P20, HA-G54, HA-S225, HA-P297, and HA-G492 selected under low folinic conditions were assayed for sensitivity to Mtx. All clones, except those expressing the HA-G54 construct, had wild-type sensitivity to the drug (results not shown). The HA-G54 stable transfectants showed Mtx sensi-   tivity that was intermediate between the wild-type and the MtxRII 5-3 response (not shown). None of these transfected constructs were tested for binding or uptake of folates since previous work has shown that cells able to grow in low levels of folinic acid can bind and take up folates (15,16,(45)(46)(47).
MtxRII 5-3 cells transfected with either the HA-V152, HA-R263, HA-Q385, or HA-P427 constructs were unable to grow in low folinic acid medium, suggesting that the insertions at these sites affect RFC function. It is of interest to note that in transient transfections only a small number of cells transfected with HA-R263 have the protein located at the cell membrane (see above).
To examine the ability of cells expressing the nine constructs to take up Mtx, G418-selected isolates were used. Stable clones of 5Ј⌬RFC-EGFP, HA-P20-EGFP, HA-S225-EGFP, HA-P297-EGFP, HA-Q385-EGFP, HA-P427-EGFP, and HA-G492-EGFP were obtained and further analyzed. The clones of HA-G54-EGFP, HA-V152-EGFP, and HA-R263-EGFP were not stable and gradually stopped producing the fusion protein. Clones expressing these three constructs were sorted by FACS to enrich for the RFC-EGFP-expressing cells prior to use in the experiments below.
Representative clones expressing either the HA-P20-EGFP, HA-S225-EGFP, HA-P297-EGFP, or HA-G492-EGFP constructs, in which the fusion protein is localized to the membrane, were capable of taking up Mtx (Table IV). This is consistent with the Mtx sensitivity and growth in low folinic acid medium demonstrated for these transfected constructs without the EGFP fusion. The normalized Mtx uptake in cells expressing either HA-S225-EGFP or HA-P297-EGFP is about 70 and 45%, respectively, relative to the control, whereas those expressing HA-P20-EGFP and HA-G492 were 80 and 120%, respectively, of the control. Western blotting with anti-GFP (Fig.  4) showed that each of these transfectants produces a broad band (probably the result of glycosylation) around 105 kDa and comparable to the 5Ј⌬RFC-EGFP control (Fig. 4, lane 4). The non-glycosylated size is predicted to be 91 kDa. The HA-P297-EGFP expressing clones, in which the fusion protein is localized to both the plasma membrane and to intracellular inclusions (Fig. 4, lane 10), also show some degraded product. This may correspond to the material in the intracellular inclusions (not shown, but similar to the inclusions shown for HA-P427-EGFP in Fig. 3, panels I and J), although this was not further explored in the present study. However, this suggests that the reported Mtx uptake value for the clone expressing HA-P297-EGFP may be underestimated as the normalization procedure did not take into account the cellular location of the fusion protein (see "Discussion").
Cells expressing either the constructs HA-G54-EGFP, HA-V152-EGFP, HA-R263-EGFP, HA-Q385-EGFP, or HA-P427-EGFP failed to transport Mtx. This inability to take up Mtx may be due to functional inactivation of the fusion protein that is otherwise properly targeted and inserted in the plasma membrane, degradation of the fusion proteins, improper localization, or improper processing of the protein. Cells expressing either HA-G54-EGFP, HA-V152-EGFP, or HA-R263-EGFP have little, if any, of the fluorescent fusion protein localized to the plasma membrane (Fig. 3). This probably accounts for the inability of these constructs to restore RFC function to the RFC-deficient cells. Western blots probed with anti-GFP antibody show that cells expressing HA-G54-EGFP have a discrete band of about 45 kDa (Fig. 4, lane 6). The fluorescence in these cells is indicative of the complete translation of the fusion protein. However, the degradation product is larger than the soluble EGFP protein (ϳ27 kDa, Fig. 4, lanes 1 and 14), suggesting that a portion of the amino terminus of the RFC has been removed. The HA-V152-EGFP product is not evident on the blot presented (Fig. 4, lane 7), but longer exposures show a faint broad band corresponding to ϳ105 kDa. This is probably the material that is localized to the plasma membrane in a small fraction of the cells expressing this construct. Bands of lower molecular size that might represent the material responsible for the diffuse fluorescence were not detected in longer exposures of the Western blot. Because only a small portion of cells expressing this construct have the product localized to the plasma membrane, it is not yet possible to examine the effect of this HA insertion on RFC function. The HA-R263-EGFP cell extract (Fig. 4, lane 9) has small amounts of two degradation products with apparent masses of ϳ45 and ϳ25 kDa which are visible only in longer exposures of the Western blot (not shown).
The cells expressing either the HA-Q385-EGFP or the HA-P427-EGFP constructs have strong fluorescence at their plasma membranes although the HA-P427-EGFP-expressing cells also have some intracellular material. This indicates, for the most part, that the fusion proteins are correctly localized and maintained under G418 selection (Fig. 3, panels G and H). The Western blot pattern for the non-functional HA-Q385-EGFP resembles that of the functional 5Ј⌬RFC-EGFP control (Fig. 4, lane 4). Cells expressing HA-P427-EGFP also produce protein of the expected size, but about half of the material migrates as smaller products of ϳ45 kDa (Fig. 4, lane 12), which may correspond to the material in the small intracellular inclusions (Fig. 3, panel J), although this was not examined further in this study. Although the HA-P427-EGFP construct has an extra valine at the carboxyl terminus of the HA epitope (see "Experimental Procedures"), this is likely not a contributing factor to disruption of function. This residue is chemically similar to the hydrophobic nature of the HA epitope, and alternative insertion strategies have included single copies with one to four additional amino acids at the carboxyl terminus, or tandem, or triplet copies of the HA epitope to boost antibody detection, with no apparent detrimental effects on protein function (37,39,40). The cells expressing either the HA-Q385-EGFP or HA-P427-EGFP constructs are of particular interest because they produce protein of the expected size that is localized to the plasma membrane, but they do not transport Mtx (Table IV). This inability may be due to disruption of the substrate-binding site(s) or of the substrate translocation process. Binding assays were carried out to distinguish between these two possibilities.
The Mtx binding capabilities of the G418-selected HA-RFC-EGFP clones (or a FACS-enriched population) were assayed, and representative data for six of the constructs are presented in Fig. 5. Since each clone expresses a different amount of the HA-RFC-EGFP fusion protein, the binding data were first normalized to the total amount of cellular protein and then to the a A value of at least 50% relative to the control (5Ј⌬RFC-EGFP) was considered a ϩ value. The effect of insertions at Gly-54, Val-152, and Arg-263 could not be determined since stable localization of the fusion protein was not observed. b PM, plasma membrane. Inclusions refers to both Golgi-trapped material and the small cytoplasmic inclusions. mean fluorescence determined by flow cytometry. The clones expressing either HA-P20-EGFP or HA-G492-EGFP transport Mtx (Table IV) and have normalized binding values about 80% that of the control cells (Fig. 5). The HA-S225-EGFP and the HA-P297-EGFP expressing clones which have about 70 and 45%, respectively, of the uptake (Table IV) also have about half the Mtx binding of the control cells (Fig. 5). This suggests that the lowered Mtx uptake may be a direct consequence of reduced binding.
The clones expressing either HA-Q385-EGFP or HA-P427-EGFP do not transport Mtx and have different phenotypes with respect to binding. The cells expressing the HA-Q385-EFGP construct present the HA epitope on the cytoplasmic side of the plasma membrane and bind Mtx only 20% as well as the control cells. Conversely, the HA-P427-EGFP-transfected cells, with the HA insertion in the last extracellular loop, bind 40% as much Mtx as the control cells. However, cells expressing HA-P427-EGFP have about half the fluorescent material at the membrane (see Fig. 3, panels H and I) where initial substrate binding would be expected to take place. Thus, the binding capability of this clone may be similar to that of the control transfectants since the normalization procedure did not take into account the cellular localization of the fusion protein. This suggests that the insertion in the last extracellular loop does not affect Mtx binding but rather implicates it in the substrate translocation process.
A summary of the results of the functional analysis of cells expressing the HA-RFC-EGFP constructs is shown in Table V. DISCUSSION In this report, the topology of the human RFC was assessed by an epitope insertion technique that also allows partial analysis of the functional regions of the protein. Although the RFC had not been previously subjected to the same level of systematic topology determination that has been applied to other transmembrane proteins (34 -40), previous reports suggested that the loops between TMs 1 and 2 (9,23) and between TMs 5 and 6 (53) are facing the cell exterior and that the carboxyl terminus is located in the cytoplasm (22). The results presented here confirm these findings and localize other RFC regions. The amino terminus, the loops between TMs 4 and 5, 6 and 7, and 10 and 11 all face the cytoplasm, whereas the loop between TMs 11 and 12 faces the cell exterior.
Although the results of the HA epitope orientation analysis are not exhaustive since every single putative loop has not been targeted, they are consistent with the 12-TM topology predicted for the RFC (Fig. 1). The confirmation of this topology is not an insignificant goal. Placing the RFC within a group of major transport facilitators (24,25) that includes the Lac permease of E. coli, the mammalian glucose transporters, and many other proteins, may help elucidate functional domains by comparison with other family members. All these proteins have a 12-M topology in which the amino and carboxyl termini are intracellular, and the patterns of loop sizes are strongly conserved, although the amino acid sequences are not. For example, each member of the 12-M protein family mentioned above has a very large inner loop between TMs 6 and 7, which for cystic fibrosis transmembrane conductance regulator and Pglycoprotein contains the nucleotide binding domain and, in others, has an undefined function. The carboxyl-terminal tail is the most variable region with respect to length, even among the four species for which the RFC has been characterized, although its functional role is unclear.
In this study, the EGFP tag was used to monitor construct expression in living cells during the clonal selection process in G418 medium. In addition, the EGFP marker allowed us to quantify the level of protein expressed from the constructs and to assess its localization and degradation status.
The addition of the EGFP fusion at the carboxyl terminus of the wild-type RFC was shown not to affect the ability of the protein to bind or take up Mtx. The green fluorescent protein has been used in many studies of protein trafficking and localization within the cell since it does not influence the targeting of the protein to which it is fused (52, 54 -58). Cells transfected with the HA insertion constructs HA-P20, HA-S225, HA-P297, and HA-G492 had colony formation efficiencies in low folinic acid medium similar to those transfected with the wild-type RFC, regardless of whether they were fused to the EGFP. However, the EGFP tag did reduce further the colony forming efficiency of cells transfected with the HA-G54 construct which, when transfected by itself under low folinic acid growth conditions, yielded about 15-20% as many colonies as the wild-type RFC. The HA-G54 insertion does not disrupt the glycosylation site but is in a region suspected to be critical for substrate recognition/discrimination as single amino acid substitutions in the corresponding loop of the murine RFC affect substrate binding (21,59). The cells expressing the HA-G54 construct may therefore have a reduced ability to bind or transport reduced folates, although this was not further explored.
The insertion of the HA epitope at four sites resulted in loss of function independent of the EGFP fusion. For two of these sites (Val-152 and Arg-263) the loss of function could be attributed to a lack of stable localization of the protein at the plasma membrane. Thus, we cannot discern the roles of these two sites in Mtx binding and uptake. Clones expressing the HA-V152-EGFP construct showed fluorescence at the plasma membrane which, with extended passage in G418 medium, became diffuse throughout the cytoplasm although no fusion protein degradation products were detected in Western blots. Cells expressing the HA-R263-EGFP construct showed a diffuse pattern of fluorescence in the cytoplasm and nucleus. The nonfunctional status of the HA-R263-EGFP protein may be a result of disrupted secondary structure, improper localization, or impaired function leading to degradation. The boundaries of the TM regions vary when a polytopic membrane protein is analyzed by multiple prediction programs or biophysical models (28). Thus, the HA-R263 insertion, predicted to be close to the membrane, may interrupt the transmembrane region. Alternatively, the hydrophobic HA epitope at this location may result in a novel transmembrane helix. Nonetheless, the other insertion (HA-S225) in this predicted loop indicates that at least a nearby region faces the cytoplasm.
Cells expressing HA-Q385-EGFP have the fusion protein localized to the plasma membrane, but they bind Mtx very poorly. Because this insertion site is in a cytoplasmic loop and should not interact directly with the substrate, the loss of binding is likely due to structural changes in the RFC.
The cells expressing the HA-P427-EGFP construct appear to have about half the Mtx-binding ability of the control, but there is evidence that the insertion at this site interferes with the substrate translocation process and not binding (see Table IV, Fig. 5). The Mtx binding of the cells expressing HA-P427-EGFP is comparable to that of cells expressing either the HA-P297-EGFP or HA-S225-EGFP constructs, both of which transport Mtx. Furthermore, although binding was normalized to the total amount of cellular protein and mean fluorescence, the precise amount of fluorescent material that is at the plasma membrane was not considered. No attempt was made to measure accurately the amount of the fusion protein as techniques for purifying membrane proteins generally aim for product purity rather than quantitative recovery. As well, there is no assurance that all of the protein localized to the membrane is functional. For cells expressing either the HA-P297-EGFP or the HA-P427-EGFP construct, it was evident from microscopy that about half of the fluorescent material was not at the plasma membrane. This was corroborated by Western blots showing that about half of the detected material was not of the size expected for the RFC-EGFP fusion protein. With this data taken into account, the normalized Mtx binding of the HA-P427-EGFP construct would be similar to that of the control. However, unimpaired Mtx binding by insertion at Pro-427 does not rule out the involvement of this loop in substrate binding as it is possible that the insertion would not disrupt a site contributed by several external loops.
Of the four constructs competent for Mtx uptake, HA-P297-EGFP has the potential to interfere directly with the substrate interaction since the insertion is in an external loop. Cells expressing this construct have about half of the fusion protein at the cell membrane and about 50% of the Mtx binding and uptake relative to the control. Thus, at this level of analysis, there is no apparent effect on substrate interaction. A single amino acid change in this loop in the murine RFC-1 was recently shown to result in changes in Mtx uptake via a 4-fold increase in K m values without affecting V max (53).
The insertions in the intracellular regions corresponding to the amino-terminal (HA-P20-EGFP) and the carboxyl-terminal regions (HA-G492-EGFP) had no obvious effects on Mtx uptake or binding. Cells expressing HA-S225-EGFP showed a roughly parallel reduction in Mtx binding and uptake compared with the transfectants expressing the 5Ј⌬RFC-EGFP control. These cells have most of the fusion protein localized to the plasma membrane, and the Western blot indicates that it is of the appropriate size. This suggests that the effects of the epitope inserted at this intracellular site may result from a subtle disruption of RFC structure.
Not every loop of the RFC was targeted for HA insertion. Some are very short and may serve no purpose other than providing a hinge for the next helix to insert in the membrane. Thus, insertion in these loops is more likely to inactivate the protein by disrupting structure rather than selectively destroying a binding site or substrate translocation region.
The work described here is a step toward understanding the structure of the RFC protein. The identification of the exterior and cytoplasmic surfaces of this protein will guide future work characterizing the sites involved in Mtx binding and translocation through the membrane and possible regulatory interactions with other proteins.