Substrate-specific binding and conformational changes involving Ser³¹³ and transmembrane domain 8 of the human reduced folate carrier, as determined by site-directed mutagenesis and protein cross-linking

Folates are members of the B class of vitamins that are cofactors for the synthesis of nucleotide precursors, serine and methionine in one-carbon transfer reactions [1]. Mammalian cells, unlike bacteria, cannot synthesize folates de novo. Hence, folate requirements must be met entirely from dietary sources [2,3]. Because of their hydrophilic charged character, there is minimal passive diffusion of anionic folates across cell membranes. Accordingly, specific transporters have evolved to mediate intestinal absorption of dietary folates, renal tubular secretion and reabsorption of folates, and transport of circulating reduced folates into systemic tissues [3].

The ubiquitously expressed RFC (reduced folate carrier) is considered to be the major transport system for folate cofactors in mammalian cells and tissues [3,4]. RFC serves a generalized role in folate transport and provides specialized tissue functions [5–9] such that loss of RFC expression or function may have potentially profound physiological and developmental consequences associated with folate deficiency [10]. RFC is also a major transporter of antifolate drugs used for cancer chemotherapy such as MTX (methotrexate), pemetrexed and raltitrexed [4]. Furthermore, the effectiveness of chemotherapy with these agents is closely linked to levels and activity of RFC in tumours [4,11].

Transport protein structural information is a prerequisite for understanding the mechanism of membrane transport. RFC is a mammalian prototype of the MFS (major facilitator superfamily) of transporters [4] that includes a large group of carriers that mediate uptake of diverse substrates including amino acids, neurotransmitters, sugars, vitamins, nucleosides and organic phosphate [12]. MFS proteins typically contain 400–600 amino acids and a structural motif composed of two halves, each with six transmembrane-spanning α-helices connected by a large hydrophilic loop, and cytoplasmic N- and C-termini. X-ray crystallographic structures of the bacterial MFS proteins, lactose/proton symporter (LacY) [13] and inorganic phosphate/glycerol-3-phosphate antiporter (GlpT) [14], were reported in 2003 at resolutions of 3.5 Å (1 Å=0.1 nm) and 3.3 Å respectively. In both the LacY and GlpT structures, hydrophilic cavities form substrate-binding sites from helices-I, -II, -IV and -V of the N-terminal domain, and helices-VII, -VIII, -X and -XI of the C-terminal domain. Helices-III, -VI, -IX and -XII are embedded in the lipid bilayer and are not directly involved in substrate binding.

By contrast, for mammalian MFS transporters such as RFC, structural data are limited due to difficulties in isolating sufficient quantities of purified proteins and in crystallizing proteins for X-ray diffraction studies. With hRFC (human RFC), we used scanning cysteine mutagenesis to generate 282 mutants with cysteine residues individually inserted into TMDs (transmembrane domains) 1–12 [15,16]. For the active 272 mutants, aqueous accessibilities were confirmed by monitoring transport and protective effects of substrate [leucovorin or (6R,S)-5-formyl tetrahydrofolate] upon treatment with membrane-impermeable MTSES (2-sulfonatoethyl methanethiosulfonate). By homology modelling from the solved structures for bacterial MFS proteins and results of biochemical studies, a three-dimensional structural model for the hRFC monomer was generated that includes TMDs 1, 2, 4, 5, 7, 8, 10 and 11 as components of an aqueous membrane-spanning translocation pathway flanked by TMDs 3, 6, 9 and 12. A two-dimensional model of helix packing for monomeric hRFC that incorporates these features is presented in Figure 1. Most recently, hRFC monomers were found to form homo-oligomers [17].

While powerful, scanning-cysteine accessibility methods provide only modest detailed information on the roles of individual residues in substrate binding and/or membrane translocation, let alone dynamic structural changes in the carrier that accompany substrate binding. In our cysteine-scanning studies for hRFC, the ten inactivating cysteine substitutions included a stretch of residues in TMD4 (Arg¹³³, Ile¹³⁴, Ala¹³⁵, Tyr¹³⁶ and Ser¹³⁸), Tyr²⁸¹ in TMD7, Ser³¹³ in TMD8 and Arg³⁷³ in TMD10, suggesting their functional or structural importance [16]. Of particular interest is Ser³¹³, flanking MTSES-reactive positions 311 and 314 in the proximal (extracellular) end of TMD8 which lines the aqueous transmembrane pathway for hRFC. In murine RFC, replacement of the homologous Ser³⁰⁹ with phenylalanine resulted in loss of MTX transport and MTX resistance, although the extent of the transport defect varied with different transport substrates [18].

In the present study, we use systematic site-directed mutagenesis for Ser³¹³, and cysteine-insertion mutagenesis and homobifunctional cross-linking between TMD8 and juxtaposed TMD5 to explore the functional significance of TMD8 and Ser³¹³ in membrane transport by hRFC. Our results indicate substantial differences between various (anti)folate substrates in their binding to hRFCs with mutated Ser³¹³, and in inducing conformational changes involving the proximal end of TMD8 by protein cross-linking, in direct support of a essential role for this region and Ser³¹³ in binding and/or membrane translocation of (anti)folate substrates.

[3′,5′,7-³H]MTX (20 Ci/mmol) was purchased from Moravek Biochemicals. The sources of the classical antifolate drugs were as follows: MTX and aminopterin (Drug Development Branch, National Cancer Institute, Bethesda, MD, U.S.A.); edatrexate (10-ethyl-10-deazaaminopterin; CIBA-GEIGY Corporation); PT523 [N^α-(4-amino-4-deoxypteroyl)-N^δ-hemiphthaloyl-L-ornithine from Dr Andre Rosowsky (Dana Farber Cancer Institute, Boston, MA, U.S.A.)]; raltitrexed (N-{5-[N-(3,4-dihydro-2methyl-4-oxyquinazolin-6-ylmethyl)-N-methyl-amino]-2-thienoyl}-L-glutamic acid) and ZD9331 [(2S)-2-{O-fluoro-p-[N-(2,7dimethyl-4-oxo-3,4-dihydro-quinazolin-6-ylmethyl)-N-(prop-2ynyl)amino]benzamido}-4-(tetrazol-5-yl)-butyric acid] were from AstraZeneca Pharmaceuticals; lometrexol [(6R)-5,10dideaza-5,6,7,8-tetrahydrofolate] and pemetrexed [N-{4-[2(2-amino-3,4-dihydro-4-oxo-7H-pyrrolo[2,3-d]pyrimidin-5-yl)ethyl]benzoyl}-L-glutamic acid] (Alimta) were from Eli Lilly; and GW1843U89 {(S)-2-[5-({[1,2-dihydro-3-methyl-1-oxo-benzo(f)quinazolin-9-yl] methyl} amino)1-oxo-2-isoindolinyl] glutaric acid} was from GlaxoSmithKline. Leucovorin and folic acid were purchased from Sigma Chemical Company. Both labelled and unlabelled MTX were purified by HPLC prior to use [19]. Synthetic oligonucleotides were obtained from Invitrogen. Tissue culture reagents and supplies were purchased from assorted suppliers with the exception of fetal bovine and iron-supplemented calf sera, which were purchased from Hyclone Technologies. p-PDM (p-phenylenedimaleimide) was purchased from Sigma Chemical Company and BMH [1,6-bis(maleimido)hexane] was obtained from Pierce Chemical Company.

The previously described HA (haemagglutinin)-tagged TMD1–6 (N₆) and Myc-tagged TMD7–12 (C₆) half-molecule constructs in pcDNA3.1 and pcDNA3 respectively [20], were used as templates to construct cysteine-less (cl) N₆/C₆ hRFC (cl-N₆/C₆) devoid of cysteine residues. cl N₆ (cl-N₆) was prepared by replacing three cysteine residues (at positions 30, 33 and 220) of N₆ hRFC with serine by site-directed mutagenesis using the QuikChange® kit (Stratagene). cl C₆ (cl-C₆) hRFC was constructed by a combination of restriction digestions and site-directed mutagenesis. First, the SfiI–NotI fragment of C₆ hRFC in pcDNA3 was replaced with the corresponding DNA fragment from full-length cl-hRFC [21], resulting in replacement of three cysteine residues (at position 365, 396 and 458) with serine. The fourth cysteine (at postion 246) was changed to serine by site-directed mutagenesis. Using cl-N₆ hRFC as a template, an analogous approach was used to insert single cysteine residues into TMD5 (positions 160, 161, 163, 164, 167, 168, 171, 172, 174 and 175). Similarly, single cysteine residues were inserted into TMD8 (311, 314, 315, 317, 318, 321, 322, 325 and 326) of cl-C₆ hRFC. Ser³¹³ mutants of hRFC were constructed by site-directed mutagenesis from HA-tagged full-length wild-type (wt) hRFC (hRFC^HA) [22] using the QuikChange® mutagenesis kit. Mutagenesis primers are shown in Supplementary Table S1 (at http://www.BiochemJ.org/bj/430/430/bj4300265add.htm). All mutations were confirmed by DNA sequencing at the Wayne State University DNA Sequencing Facility.

Transport-defective MTX-resistant HeLa cells, designated R5 [23], were a gift from Dr I. David Goldman (Albert Einstein College of Medicine, Bronx, NY, U.S.A.). R5 cells were maintained in RPMI 1640 medium, supplemented with 10% fetal bovine serum, 2 mM L-glutamine, penicillin (100 units/ml), and streptomycin (100 μg/ml) in a humidified atmosphere at 37 °C in the presence of 5% CO₂. Transient transfections of wt and mutant hRFC constructs (see below) were performed with Lipofectamine™ Plus reagent (Invitrogen), as described previously [15,16]. Cultures were split 24 h after transfection and assayed for transport and expression on Western blots after an additional 24 h.

The MTX transport-deficient K562 subline, designated K500E, was selected from wt K562 cells (American Type Culture Collection) and maintained in complete RPMI 1640 medium containing 10% iron-supplemented calf serum, 2 mM L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin, and 0.5 μM MTX [24]. Wt and mutant hRFC constructs (see below) were transfected into K500E cells by electroporation (155 V, 1000 μF capacitance). After 24 h, cells were treated with G418 (1 mg/ml) and stable clones were selected by cloning in soft agar in the presence of G418 [24]. Both wt and transfected K500E cultures were cultured in complete RPMI 1640 with 10% supplemented calf serum and antibiotics in a humidified atmosphere at 37 °C in the presence of 5% CO₂. For transfected cells, the medium was supplemented with G418 (1 mg/ml).

Uptake of [³H]MTX (0.5 μM) in transiently transfected R5 HeLa cells was measured over 2 min at 37 °C in 60 mm dishes in HSM buffer (Hepes/sucrose/Mg²⁺ ‘anion-free’ buffer; 20 mM Hepes and 235 mM sucrose, pH adjusted to 7.14 with MgO). Uptake of [³H]MTX was quenched with ice-cold Dulbecco's PBS. Cells were washed with ice-cold PBS (3 times) and proteins were solubilized with 0.5 M NaOH. [³H]MTX uptake into stably transfected K500E cells was measured over 180 s (wt and Ser³¹³ mutants) in physiological HBSS (Hank's balanced salts solution) in a shaking water bath at 37 °C, as described previously [24,25]. For both cell line models, levels of intracellular radioactivity were expressed as pmol/mg of protein, calculated from direct measurements of [³H]MTX and protein content of cell homogenates. Protein assays were based on the method of Lowry et al. [26]. For the stable transfected K500E cells, kinetic constants (K_t, V_max) were calculated from Lineweaver–Burk plots for [³H]MTX, and K_i values for assorted transport substrates were determined from Dixon plots with [³H]MTX (1 μM).

Transfected cells were harvested, flash-frozen and stored at −80 °C. Plasma membranes were prepared from the frozen cell pellets as described previously [25]. Membrane preparations used for cross-linking were suspended in 20 mM Tris/HCl (pH 7.5) containing 100 mM sucrose with a protease inhibitor cocktail (Roche) at a protein concentration of 1–2 mg/ml. Otherwise, membrane preparations were suspended in the aforementioned buffer without sucrose. Membrane preparations were stored at −80 °C as aliquots for further use.

Both in vitro (plasma membranes) and in vivo (transfected cells) cross-linking reactions were performed at 4 °C or 25 °C with thiol-specific homobifunctional cross-linkers [p-PDM (10 Å; rigid) or BMH (16 Å; flexible)]. Final concentrations of cross-linkers were 1 mM (in vitro cross-linking) or 0.25 mM (cross-linking with intact cells) and treatments were for 30 min. Reactions were terminated by adding 10 mM dithiothreitol. For the cell treatments, following cross-linking, membranes were prepared. In either case, membranes were diluted into SDS/PAGE sample buffer [62.5 mM Tris/HCl (pH 6.8), 10% glycerol, 0.7% SDS and 0.7 M 2-mercaptoethanol] and analysed by SDS/PAGE (10% gels) [15]. Proteins were transferred on to PVDF membranes (Pierce) [15] and immunoblotted with Myc- or HA-specific monoclonal antibodies (Covance) against epitope-tagged C₆ and N₆ hRFC proteins respectively, and anti-mouse secondary IRDye™ 800-conjugated antibody (Rockland). Detection and densitometry used the Odyssey® IR imaging system and software (LI-COR Biosciences).

For studies of substrate-induced conformational changes, cells or membranes were incubated with transport substrates (1 mM final concentration) at room temperature (25 °C for 10 min and 1 h respectively), followed by cross-linking and immunoblotting, as described above.

In some experiments, proteins were digested with N-glycosidase F (New England Biolabs) prior to SDS/PAGE. For this, membranes were collected by ultracentrifugation (48000 rev./min; TLA 100.2 rotor and Beckman TL100 ultracentrifuge) after cross-linker treatment and resuspended in 10 mM Tris/HCl (pH 7.5) containing the protease inhibitor cocktail. Samples were denatured for 10 min with 0.5% SDS and 40 mM dithiothreitol. An equal volume of 50 mM sodium phosphate (pH 7.5) and 1% Nonidet P40 were added, along with N-glycosidase F (1000 units), followed by incubation at 37 °C for 14 h. Control samples were incubated in parallel in buffer without N-glycosidase F. Samples were diluted with 3× SDS/PAGE sample buffer, fractionated on 10% polyacrylamide gels, and analysed by Western blotting.

Ser³¹³ is located in the proximal end of TMD8. In murine RFC, replacement of Ser³⁰⁹ (homologous with Ser³¹³ in hRFC) with phenylalanine resulted in loss of transport and MTX resistance, although the transport phenotype showed substantial substrate dependence [18]. Replacement of Ser³¹³ in hRFC with cysteine was inactivating [16].

To further explore the potential role of Ser³¹³ in binding and translocation of hRFC substrates, we performed systematic site-directed mutagenesis of this residue using HA-tagged wt hRFC as a PCR template. Both conservative and non-conservative replacements of Ser³¹³ were tested, including alanine, cysteine and threonine, and results were compared with those for the phenylalanine hRFC mutant, based on the published report on murine RFC [18]. All mutants were expressed in hRFC-null R5 HeLa cells at high levels on Western blots probed with an HA-specific antibody (Figure 2, inset). However, when assayed for transport with [³H]MTX, all replacements were poorly tolerated. When transport results were normalized to levels of hRFC protein on Western blots, S313A and S313T preserved significant albeit low level activity over vector control-transfected cells with activities ~17% and ~8% respectively, of that for wt hRFC levels (Figure 2).

To characterize the role of Ser³¹³ in the specificity of substrate binding to hRFC, we generated stable transfectants of the S313A and S313T hRFC mutants in hRFC-null K562 (K500E) cells. The levels of hRFC proteins for vector control K500E, and for K500E cells transiently transfected with wt, S313A and S313T hRFCs were measured on Western blots and are shown in Supplementary Figure S1 (at http://www.BiochemJ.org/bj/430/bj4300265add.htm). Kinetic constants (K_t, V_max and V_max/K_t, including normalized values) for MTX with S313A and S313T hRFC, along with results for wt hRFC, were calculated from Lineweaver–Burke plots and are summarized in Table 1. The results show that K_t values for MTX were minimally changed from the wt values, whereas the V_max values were substantially decreased (Table 1).

To extend our kinetic analyses to additional transport substrates with disparate structures, we used ten (anti)folate substrates (raltitrexed, pemetrexed, leucovorin, folic acid, lometrexol, GW1843U89, PT523, ZD9331, aminopterin and edatrexate) as competitive inhibitors of [³H]MTX (1 μM) uptake over a range of concentrations with wt, S313A and S313T hRFC-expressing cells. K_i values were calculated from Dixon plots (Table 2). Whereas K_i values for wt and mutant hRFCs were essentially identical for lometrexol, PT523, ZD9331 and edatrexate, statistically significant differences in binding were seen with other substrates. For instance, both S313A and S313T showed significantly decreased K_i values for pemetrexed, leucovorin and aminopterin (~2–3-fold) compared with those for wt hRFC. S313T hRFC showed selectively increased binding for raltitrexed and GW1843U89 compared with wt hRFC (~2–3-fold decreased K_i values), whereas S313A showed a ~2-fold increased K_i for folic acid compared with wt carrier. These results support the notion that Ser³¹³ located in the proximal end of TMD8 of hRFC directly participates in (anti)folate binding.

TMDs 8 and 5 are juxtaposed in our hRFC monomer models (Figure 1). We previously reported that cysteine substitutions at multiple positions in TMD5 (Val¹⁶⁰, Leu¹⁶¹, Val¹⁶⁴, Ser¹⁶⁷, Ser¹⁶⁸, Gly¹⁷¹, Gln¹⁷², Val¹⁷⁵) and TMD8 (Ala³¹¹, Thr³¹⁴) were reactive with MTSES [16]. Although substrate protection from MTSES was variable among these positions, the protection afforded T314C was particularly notable (2.1-fold) [16]. These results unambiguously established aqueous accessibilities for positions spanning the entire length of TMD5 and in the proximal end of TMD8 in hRFC, and they strongly implied that Thr³¹⁴ in TMD8 faces the substrate-binding pocket (Figure 1).

Our immediate goal was to express functional cl-hRFC as TMD1–6 and TMD7–12 half-molecules (designated N₆ and C₆ respectively) in hRFC-null R5 cells, cl-N₆ with an HA insertion after Glu²²⁶ and cl-C₆ with a Myc-His₁₀ inserted after Leu⁵³⁷ (Figure 3A). For cross-linking, cl-N₆ and cl-C₆ hRFC constructs were mutated to include cysteine insertions at defined positions spanning the lengths of helices 5 and 8. This approach was based on our previous report that co-expression of wt hRFC half molecules (wt-N₆/C₆) in hRFC-null cultured human cells led to functional complementation and restoration of transport activity, whereas transfections with the N₆ or C₆ hRFC constructs individually were ineffective [20]. The expressed HA-tagged hRFC N₆ fragment was glycosylated at Asn⁵⁸, resulting in a broadly banding (27–58 kDa) pattern on SDS/PAGE that quantitatively reverted to a 27 kDa species upon enzymic deglycosylation, whereas the Myc-tagged hRFC C₆ half-molecule migrated as a sharp 40 kDa band on SDS/PAGE [20]. Confocal analysis showed that when co-expressed, both the N₆ and C₆ half-molecules were targeted to the plasma membrane with no obvious intracellular staining [20].

We initially needed to prepare the cl-N₆/C₆ hRFC construct, for which we used wt-N₆/C₆ hRFCs as a template for cysteine replacements. cl half-molecule constructs were generated by replacing cysteine residues at positions 30, 33 and 220 in N₆, and at positions 246, 365, 396 and 458 in C₆ with serine (Figure 3A). cl-N₆/C₆ was co-transfected into hRFC-null R5 Hela cells to test for restoration of transport function and hRFC protein expression. wt-N₆/C₆ was transfected in parallel as a positive control. At 48 h post-transfection, cells were harvested for assays of [³H]MTX transport and levels of individual wt- and cl- N₆ and C₆ proteins on Western blots with antibodies against HA and Myc epitopes respectively.

As shown in Figure 3(B) (lower panel), a major band migrating at 30 kDa was detected with an anti-HA antibody (detects N₆ hRFC) for both wt-N₆/C₆ and cl-N₆/C₆ hRFCs, along with an assortment of higher molecular mass N-glycosylated N₆ forms. With the anti-Myc antibody (detects C₆ hRFC; Figure 3B, upper panel), a major 40 kDa band was detected for both wt-N₆/C₆ and cl-N₆/C₆. An unidentified low abundant (~38 kDa) species was detected with anti-Myc antibody in some analyses. Although there were nominal differences in expression of N₆ and C₆ between wt- and cl-hRFC proteins, MTX transport was somewhat decreased (~40%) for cl-N₆/C₆ (Figure 3C). Nonetheless, the MTX uptake for cl-N₆/C₆ still exceeded (~3-fold) the residual low level in R5 cells.

The use of functional cl-N₆/C₆ with each half-molecule including a unique epitope tag and strategically placed cysteine insertions provided an ideal approach for identifying cross-linked domains, simply by following changes in migrations of HA- and Myc-tagged proteins on Western blots. Although there may be slight differences in folding between reconstituted half-molecule transporters and full-length wt hRFC, wt-N₆/C₆ hRFC was shown to accurately recapitulate a number of functional characteristics of full-length wt hRFC [20].

Based on their relative proximities and proposed orientations toward the hRFC hydrophilic cavity in two- (Figure 1) and three-dimensional [16] models, cysteine residues were inserted into cl-N₆/C₆ along the juxtaposed faces of TMD5 in N₆ hRFC (positions 175, 174, 172, 171, 168, 167, 164, 163, 161 and 160) and of TMD8 in C₆ hRFC (positions 311, 314, 315, 317, 318, 321, 322, 325 and 326) (Figure 4A). Altogether, ten cysteine N₆/C₆ pairs (175/311, 174/314, 172/315, 171/317, 168/318, 167/321, 164/322, 163/325, 161/326 and 160/326 in TMDs 5/8) were selected for transfections of hRFC-null R5 cells. All of the N₆/C₆ double cysteine mutants were expressed in R5 cells and the six cysteine mutant pairs (175/311, 174/314, 172/315, 168/318, 164/322 and 160/326) with transport activities in excess (≥2-fold) of the basal low level in untransfected cells (Supplementary Figure S2 at http://www.BiochemJ.org/bj/430/bj4300265add.htm) were used for cross-linking experiments (Figure 4A also shows a schematic of the cysteine pairs used for cross-linking).

For the initial cross-linking experiments, plasma membranes were prepared and cross-linked in vitro at 25 °C with the membrane-permeable homobifunctional cross-linkers p-PDM and BMH [27]. The proteins were separated by SDS/PAGE for Western analysis with a Myc-specific antibody. In the absence of cross-linkers, a prominent ~40 kDa species was detected (Figure 5A, lane 4 shows this result for the Cys¹⁷⁵/Cys³¹¹ pair). For the Cys¹⁷⁵ N₆ and Cys³¹¹ C₆ mutant pair, treatment with cross-linkers resulted in two major bands [70 and 80 kDa; labelled N₆–C₆ and C₆–C₆ respectively, in Figures 4B (lanes 1 and 2) and 5A (lanes 5 and 6)] not seen in the absence of cross-linkers (Figure 5A, lane 4) or in cl-N₆/C₆ in the presence or absence of cross-linkers (Figure 5A, lanes 1–3). With the other five mutants, identical results were obtained in the absence of cross-linkers (results not shown). Although a very low level of the 70 kDa N₆–C₆ was detected with BMH and Cys¹⁶⁰/Cys³²⁶ with the anti-Myc antibody, for the other mutant pairs and cross-linkers, only the 80 kDa (C₆–C₆) species was detected (Figure 4B). Explanations for the failure to detect the 70 kDa band with these cysteine pairs range from their relative inaccessibilities to the cross-linkers or helix proximities somewhat different from those predicted by the three-dimensional hRFC model [16], to the impact of nearby amino acids on individual cysteine chemical reactivities, all of which may preclude an ability to efficiently cross-link, even though individual cysteine residues may be highly reactive. From their sizes, the 70 kDa and 80 kDa bands probably arose from cross-links between the N₆ and C₆ half-molecules and between two C₆ half-molecules respectively. There was no obvious difference in the extent of cross-linking between p-PDM and BMH.

For the Cys¹⁷⁵/Cys³¹¹ pair without cross-linker, HA-specific antibody identified products derived from N₆, including a major ~30 kDa form and higher mass N-glycosylated forms (Figure 5A, lower panel, lane 4). An identical pattern was seen with cl-N₆/C₆ with or without cross-linkers (Figure 5A, lower panel, lanes 1–3). With p-PDM/BMH treatments of Cys¹⁷⁵/Cys³¹¹, the 70 kDa (but not 80 kDa) band was detected, along with a 60 kDa species and higher molecular mass (glycosylated) forms not seen in the absence of cross-linkers (Figure 5A, lower panel, lanes 5 and 6). N-glycosidase F treatment shifted the major HA immunoreactive bands from 70, 60 and 30 kDa to 67, 54 and 27 kDa respectively (Figure 5B, lanes 3 and 4), further establishing their N-glycosylation and probable identities as N₆–C₆, N₆–N₆ and N₆ respectively. N-glycosidase F treatment also shifted the 70 kDa N₆–C₆ band detected with anti-Myc antibody to 67 kDa but not the 80 kDa C₆–C₆ or the 40 kDa C₆ bands (results not shown). When probed with anti-HA antibody, the five other cysteine pairs (174/314, 172/315, 168/318, 164/322 and 160/326) gave results identical with those for Cys¹⁷⁵/Cys³¹¹ in the absence of cross-linking. Upon cross-linking, the 60 kDa species was detected (results not shown). Our inability to detect higher molecular mass (>70 kDa) glycosylated forms of N₆–C₆ on blots probed with anti-Myc antibody probably reflects differences in sensitivities between anti-Myc antibody and anti-HA antibody (as suggested from the more intense signal for the 70 kDa species with anti-HA over anti-Myc antibodies in Figure 5A).

Several critical controls were performed for our in vitro cross-linking experiments. When the reactions for Cys¹⁷⁵/Cys³¹¹ mutants were performed at 4 °C rather than 25 °C, cross-links were still detected (Figure 5D, upper panel, lanes 1 and 3). Another negative control involved solubilization of the Cys¹⁷⁵/Cys³¹¹ membranes with 0.7% SDS before cross-linker treatment, followed by treatment with cross-linkers for 30 min at 25 °C, and quenching with 10 mM dithiothreitol and SDS/PAGE buffer (labelled II in Figure 5C, lanes 3 and 4). Results were compared with those for the Cys¹⁷⁵/Cys³¹¹ sample cross-linked under established conditions at 25 °C (labelled I in Figure 5C, lanes 1 and 2). For reaction II, neither the 70 kDa nor 80 kDa cross-linked band was detected (with anti-Myc antibody), firmly establishing that the cross-linking detected with the Cys¹⁷⁵/Cys³¹¹ half-molecules does not occur randomly in solution but rather can only occur in situ in the intact plasma membranes.

Cross-linking of the 175/311 cysteine pairs was also confirmed with intact cells co-transfected with Cys¹⁷⁵ N₆ and Cys³¹¹ C₆ half-molecule constructs treated with p-PDM and BMH at 4 °C and 25 °C. In membrane preparations from cross-linked cells, both C₆–C₆ and N₆–C₆ cross-linked species were detected on Western blots (Figure 5D, lower panel). In cells co-transfected with the cl-N₆ and Cys³¹¹ C₆ half-molecule constructs and treated with p-PDM at 4 °C, the 80 kDa band (but not the 70 kDa band) was detected (Figure 5E, lane 1). This unambiguously establishes that the 80 kDa species (but not the 70 kDa form) is the result of cross-links between Cys³¹¹ on separate C₆ molecules and that the 70 kDa cross-linked product is absolutely dependent on co-expression of the Cys¹⁷⁵ N₆ and Cys³¹¹ C₆ hRFC half-molecules.

Our cross-linking results strongly suggest that the TMD5 and TMD8 helices are close together at their proximal (extracellular) ends, as indicated by formation of the N₆–C₆ cross-link between Cys¹⁷⁵ and Cys³¹¹. Detection of C₆–C₆ intermolecular cross-links for multiple cysteine pairs spanning TMDs 5 and 8 suggests that the TMD8 helix in each hRFC half-molecule protomer abuts the corresponding region in another, in support of the notion of higher-order hRFC homo-oligomers, as recently reported [17]. An analogous argument can be made for the apparent N₆–N₆ intermolecular cross-links for the Cys¹⁷⁵/Cys³¹¹ pair and TMD5. While the formation of C₆–C₆ (or N₆–N₆) cross-links was completely unexpected given that the positions selected for cysteine replacement were based on their patterns of MTSES reactivities and apparent aqueous accessibilities, the nature of the homo-oligomeric interface(s) is not yet established. Studies are underway to explore this important question.

If Ser³¹³ directly participates in substrate binding as suggested by our mutant studies, nearby residues (e.g. position 311) might be expected to be conformationally active in the presence of excess substrate. To test this possibility, in situ cross-linking involving position 311 was used as a highly sensitive probe of the conformationally active interfaces between transmembrane helices 8 and 5 upon substrate binding [28–30]. We transfected R5 cells with Cys¹⁷⁵ N₆ and Cys³¹¹ C₆ hRFC, then treated the cells with BMH in the absence and presence of hRFC substrates (aminopterin, leucovorin and raltitrexed). In four independent experiments, N₆–C₆ (but not C₆–C₆) cross-links by BMH were demonstrably increased to 1.6-, 2.3- and 2.4-fold by leucovorin, aminopterin and raltitrexed respectively (Figure 6). Analogous results were observed with BMH-treated plasma membranes from Cys¹⁷⁵/Cys³¹¹-transfected cells, cross-linked in the presence of transport substrates (Supplementary Figure S3 at http://www.BiochemJ.org/bj/430/bj4300265add.htm).

Since cross-link formation is a reflection of dynamic collisions that result in chemical modifications of reactive residues [31], our finding of enhanced cross-linking between positions 175 and 311 in the presence of transport substrates suggests that conformational changes occur involving the proximal end of TMD8 in relation to TMD5 upon substrate binding. This is entirely consistent with the notion that this stretch of TMD8 contributes to the substrate-binding pocket in hRFC, as noted above.

Although hRFC exists as an homo-oligomer [17], each hRFC monomer has its own translocation pathway and appears to function independently [32]. Characterization of the determinants of substrate binding in each hRFC monomer is essential to understanding the molecular mechanism of folate and antifolate membrane transport by this physiologically and pharmacologically important carrier. We previously proposed that Lys⁴¹¹ in TMD11, Arg³⁷³ in TMD10, Tyr²⁸¹ in TMD7 and Ser³¹³ in TMD8 participate in (anti)folate binding and that TMD helices including these residues comprise the hRFC substrate-binding pocket [16]. Thus replacement of Arg³⁷³, Tyr²⁸¹ or Ser³¹³ in hRFC individually with cysteine resulted in nearly complete loss of transport activity [16]. Similarly, aliphatic substitutions of Arg³⁷³ abolished transport, whereas activity was preserved with lysine replacement at this position [33]. Although Lys⁴¹¹ in hRFC can be replaced by any of a number of amino acids of varying bulk and charge with modest effects on transport activity, this residue is nonetheless the primary target for electrophilic attack by N-hydroxysuccinimide-activated MTX ester and can participate in an interaction with (anti)folate substrate, primarily through an ionic association with the γ-carboxy group [33]. However, this interaction with Lys⁴¹¹ is apparently not essential for transport function since the γ-carboxy group is not only expendable, but indeed its replacement by an uncharged hydrogen or methyl group in a series of furo[2,3-d]pyrimidine antifolates actually enhanced high-affinity reversible binding of substrate to the carrier, as long as an ionizable α-carboxy group is intact [33]. Rather, Arg³⁷³ was suggested to forge an ionic association with the α-carboxylate of (anti)folate substrates [33].

The present study sheds new important light on the functional significance of the proximal TMD8 helix in general, and Ser³¹³ in particular. By kinetic analysis with an assortment of structurally diverse transport substrates, there were substrate-selective differences in K_i values between wt and Ala/Thr³¹³ mutant hRFCs, suggesting a possible role for Ser³¹³ in substrate binding to hRFC.

Our results with Cys¹⁷⁵/Cys³¹¹ hRFC half-molecule mutants and cysteine cross-linking established that the proximal (extracellular) ends of the TMD8 and 5 helices are juxtaposed, as predicted by hRFC homology models. Although cross-linking results must be interpreted with caution since cross-link formation is a reflection of dynamic movements and chemical reactivities with individual cysteine residues rather than just their proximities [31], this approach has nonetheless established a close correlation between collision rates and residue proximities [34]. In the presence of transport substrates, the 175/311 interface was conformationally active, as reflected in increased N₆–C₆ cross-links between these positions. Thus while position 311 is not directly involved in substrate binding [16], this result nonetheless supports the notion that residues located in this stretch of TMD8 (i.e. Ser³¹³) participate in substrate binding. Cross-linking between Cys¹⁷⁵ and Cys³¹¹ was substrate-dependent, further implying that hRFC assumes distinct conformations in this region upon binding different transport substrates, consistent with our kinetic analysis of structurally diverse transport substrates.

Based on these collective data, we present a hypothetical model for binding (anti)folate substrates to hRFC involving interactions between the pteridine ring of MTX and Tyr²⁸¹ and Ser³¹³, and between the α-carboxy of MTX and Arg³⁷³ (Figure 7). Although a putative hydrogen bond is depicted between the Ser³¹³ hydroxy group and the 4-amino group of MTX, this must not be obligatory since functionality at position 313 can in part met by alanine. In the model, Tyr²⁸¹ is juxtaposed to the pteridine ring of MTX and may bind with MTX through π–π interactions. Lys⁴¹¹ interacts with the γ-carboxy group of MTX, although this is not essential for binding and transport, as noted above.

Finally, results are presented herein that C₆–C₆ and N₆–N₆ cross-links occur for the TMD5/8 cysteine pairs, providing further independent confirmation of the existence of homo-oligomeric hRFC [17]. As previously suggested, such higher-order hRFC structures should be particularly significant, with profound implications to hRFC mechanism, regulation and antifolate resistance [17]. Further characterization of the structural and regulatory features of homo-oligomeric hRFC will be the topic of future reports.

BMH 1

6-bis(maleimido)hexane

C₆

transmembrane domains 7–12

cl

cysteine-less

HA

haemagglutinin

MFS

major facilitator superfamily

MTSES

2-sulfonatoethyl methanethiosulfonate

MTX

methotrexate

N₆

transmembrane domains 1–6

p-PDM

p-phenylenedimaleimide

RFC

reduced folate carrier

hRFC

human RFC

TMD

transmembrane domain

wt

wild-type

Zhanjun Hou designed and performed the experiments, and wrote the manuscript. Jianmei Wu prepared the Ser³¹³ hRFC mutants, performed stable transfections of K500E cells and did the kinetic analysis. Jun Ye assisted with experimental design and generated the hypothetical molecular model of the hRFC substrate-binding site. Christina Cherian prepared the cysteine-less N₆ and C₆ half-molecule constructs and their cysteine insertion counterparts. Larry Matherly supervised the project and wrote the manuscript.

We would like to thank Dr I. David Goldman (Albert Einstein School of Medicine, Bronx, NY, U.S.A.) for his gift of hRFC-null R5 HeLa cells.

This work was supported by the National Cancer Institute, National Institutes of Health [grant number CA53535].