A Structurally Altered Human Reduced Folate Carrier with Increased Folic Acid Transport Mediates a Novel Mechanism of Antifolate Resistance*

CEM/MTX is a subline of human CCRF-CEM leukemia cells which displays >200-fold resistance to methotrexate (MTX) due to defective transport via the reduced folate carrier (RFC). CEM/MTX-low folate (LF) cells, derived by a gradual deprivation of folic acid from 2.3 μm to 2 nm (LF) in the cell culture medium of CEM/MTX cells, resulted in a >20-fold overexpression of a structurally altered RFC featuring; 1) a wild typeK m value for MTX transport but a 31-fold and 9-fold lower K m values for folic acid and leucovorin, respectively, relative to wild type RFC; 2) a 10-fold RFC1 gene amplification along with a >20-fold increased expression of the main 3.1-kilobase RFC1 mRNA; 3) a marked stimulation of MTX transport by anions (i.e. chloride); and 4) a G → A mutation at nucleotide 227 of the RFC cDNA in both CEM/MTX-LF and CEM/MTX, resulting in a lysine for glutamate substitution at amino acid residue 45 predicted to reside within the first transmembrane domain of the human RFC. Upon transfer of CEM/MTX-LF cells to folate-replete medium (2.3 μm folic acid), the more efficient folic acid uptake in CEM/MTX-LF cells resulted in a 7- and 24-fold elevated total folate pool compared with CEM and CEM/MTX cells, respectively (500 versus 69 and 21 pmol/mg of protein, respectively). This markedly elevated intracellular folate pool conferred a novel mechanism of resistance to polyglutamatable (e.g. ZD1694, DDATHF, and AG2034) and lipophilic antifolates (e.g. trimetrexate and pyrimethamine) by abolishing their polyglutamylation and circumventing target enzyme inhibition.

The reduced folate carrier (RFC) 1 has long been recognized for its essential role in mediating the cellular uptake of reduced folate cofactors required for several biosynthetic processes (1)(2)(3)(4). Furthermore, RFC-mediated transport is also known to be essential for the cytotoxic activity of folate-based chemotherapeutic drugs such as methotrexate (MTX) (5)(6)(7). One of the characteristic transport kinetic properties of the RFC is its poor affinity for folic acid (K m 200 -400 M) as compared with reduced folate cofactors and MTX (K m 1-10 M) (3, 4, 8 -10). Human RFC is a transmembrane glycoprotein that undergoes a prominent glycosylation resulting in an apparent molecular mass of 80 -120 kDa (11)(12)(13).
Insight into the molecular aspects of RFC was obtained from recent studies reporting the isolation of rodent and human cDNA clones encoding for a protein, RFC1, that is involved in the uptake of folates and antifolates (13, 14 -17). Correlates between RFC1 and RFC function have been implicated from observations that transfection with RFC1 cDNA could restore MTX uptake and drug sensitivity in RFC transport-defective murine, hamster, and human cells (13, 14 -18). Moreover, RFC1 gene amplification and consequent mRNA overexpression has been demonstrated in human CEM-7A leukemia and K562.4CF cells, which overexpressed RFC protein following gradual adaptation to low extracellular concentrations of leucovorin (11, 13, 19 -21).
Defective transport via the RFC as a result of qualitative and/or quantitative defects has been recognized as a common and frequent mechanism of antifolate resistance (4,6,7,(22)(23)(24). The biochemical and molecular basis of RFC transport defects is starting to be recently unraveled (11, 12, 16, 24 -28) and includes structural alterations of the RFC protein (11,13), additional membrane protein changes along with a translocation defect of the RFC (29,30). In the rodent RFC gene a number of different mutations have been reported which resulted in single amino acid substitutions in either the first (26,67), fourth (27), or the loop between the 7th and 8th predicted transmembrane domains of the RFC protein (28). These mutations resulted in a marked defect in MTX transport (26,27,67). Interestingly, however, a serine to asparagine substitution at amino acid 46 (26) and a glutamate to lysine substitution at amino acid 45 of the mouse RFC gene (67) selectively enhanced transport of natural folates (leucovorin, folic acid) over folate analogues such as MTX. In contrast and thus far in the human RFC gene, only N-terminal premature translation termination mutations, which disrupt RFC protein expression, were reported (25).
This study is a first report on the identification of a mutation in the human RFC gene that results in marked changes in the kinetic properties of RFC-mediated transport of (anti)folates. Upon growth in low folate (LF) medium, CEM/MTX cells expressed high levels of a structurally altered RFC that is functionally characterized by a 9-and 31-fold increased affinity (decreased K m ) for transport of reduced folate cofactors and folic acid, respectively. In folate-replete medium, CEM/ MTX-LF cells can accumulate intracellular folates to such a high level that it confers resistance to lipophilic antifolate inhibitors of dihydrofolate reductase and to some polyglutamatable antifolates due to an abolishment of polyglutamylation.  (39) (40,41). All labeled antifolates were purified prior to use by thin layer chromatography as described previously (42). [

Cell Lines
Human CCRF-CEM leukemia cells and a variant CEM-7A subline, exhibiting 95-fold increased V max for MTX influx along with a 30-fold overexpression of RFC protein, were cultured as described previously (12,20). CEM/MTX cells were originally isolated by Rosowsky et al. (43) by stepwise selection for MTX resistance and displayed a 200-fold resistance to MTX due to defective transport via the RFC (5,22). CEM/MTX cells are maintained in folate-replete RPMI 1640 medium, supplemented with 10% FCS, 2 mM glutamine, and 100 g/ml penicillin and streptomycin, in the presence of 1 M MTX.

Selection of CEM/MTX-LF Cells
CEM/MTX cells were transferred to folate-free medium supplemented with 10% dialyzed FCS in which the folic acid concentration was gradually deprived from 2.3 M to 2 nM over a 5-month period. In addition, 50 nM MTX (5-fold the IC 50 of wild type CCRF-CEM cells) were also present during the folate deprivation of CEM/MTX cells to prevent the emergence of possible revertants of wild type CCRF-CEM cells. The CEM/MTX cells selected to grow at 2 nM folic acid were cloned and will be further designated CEM/MTX-LF (for "low folate").

Growth Inhibition Experiments
One ml of CCRF-CEM, CEM/MTX, and CEM/MTX-LF cells were plated at an initial density of 1.25 ϫ 10 5 /ml in the individual wells of a 24-well tissue culture plate containing a 2-log range of antifolate drug concentrations. Following 72 h of continuous drug exposure, viability cell counts were determined with a hemocytometer and trypan blue exclusion. IC 50 values are defined as drug concentrations at which cell growth is inhibited by 50% compared with untreated controls (5).

Enzyme Assays
All enzyme assays were carried out on cells collected in their exponential growth phase. Cells were washed three times with a HEPESbuffered saline solution (HBSS) containing 107 mM NaCl, 20 mM HEPES, 26.2 mM NaHCO 3 , 5.3 mM KCl, 1.9 mM CaCl 2 , 1.0 mM MgCl 2 , and 7.0 mM D-glucose, adjusted to pH 7.4 with NaOH. The final cell pellets were either used immediately or stored at Ϫ80°C.
Folylpolyglutamate synthetase activity was determined according to a procedure as described previously (44). Two hundred micrograms of protein extract were assayed per experiment using 250 M MTX as a substrate for the folylpolyglutamate synthetase reaction.
Folylpolyglutamate hydrolase activity was measured essentially as described by O'Connor et al. (45). Twenty five micrograms of cellular protein extract and 100 M MTX-Glu 2 were used as substrate for the folylpolyglutamate hydrolase reaction assay. The assay pH was 6.9. The reaction products were quantitated by HPLC (8,41).
Thymidylate synthase catalytic activity and the number of FdUMP binding sites was determined as described by Van der Wilt et al. (46). Dihydrofolate reductase (DHFR) levels were quantitated in a [ 3 H]MTX binding assay as described by Schuetz et al. (29).

Transport Studies
(Anti)folate transport studies were carried out with (variant) CEM cells in the mid-log phase of growth. CCRF-CEM, CEM/MTX, and CEM/MTX-LF cells were washed and resuspended in HBSS transport buffer (20) to a final density of 1.5 ϫ 10 7 cells/ml; for CEM-7A cells a density of 3 ϫ 10 6 cells/ml was used. Incubation times varied from 1.5 min (for CEM-7A cells) to 3 min (for other cells) at extracellular concentrations of radiolabeled folate (analogues) ranging from 0.25 to 250 M. Transport of radiolabeled folate (analogues) was stopped by the addition of 10 ml of ice-cold transport buffer, after which the cell suspension was centrifuged (500 ϫ g, 5 min, 4°C) and the cell pellet was washed once more with 10 ml of ice-cold transport buffer. The final cell pellet was suspended in water and processed for radioactivity counting.

Affinity Labeling
Quantitation of folate transporter level, using 100 nM of an N-hydroxysuccinimide ester of [ 3 H]MTX or [ 3 H]folic acid, followed by SDSpolyacrylamide gel electrophoresis analysis of labeled protein, was carried out as described previously (20,42).

Membrane Folate Receptor Expression Analysis by MOv18 Flow Cytometric Analysis
Flow cytometric analysis for the cellular expression of membrane folate receptors by a monoclonal antibody directed to membrane folate receptors (MOv18) was carried out as described previously (47,48).

Determination of RFC1 Gene Copy Number and mRNA Levels
High molecular weight genomic DNA was extracted from CCRF-CEM, CEM-7A, CEM/MTX, and CEM/MTX-LF cells maintained in folate-deplete and folated-replete media. DNA was digested with EcoRI, HindIII, PstI, and BamHI, fractionated by electrophoresis on 0.8% agarose gels, transferred to a Zetaprobe (Bio-Rad) nylon membrane, and UV cross-linked. RFC1 gene copy number was determined by Southern blot analysis using a 32 P-labeled (49) human RFC1 cDNA probe isolated from a cDNA library from CEM-7A cells. Poly(A) ϩ RNA was size-fractionated on 1.5% agarose/formaldehyde gels, blotted onto a Gene Screen ϩ (NEN Life Science Products) nylon membrane and UV cross-linked. Human RFC1, ␤-actin (50), and MDR1 cDNAs (51) were labeled by random hexamer priming (49). Southern and Northern blot hybridizations and post-hybridization washes were carried out under high stringency conditions as described previously (23,52). Southern and Northern blots were quantitated with a BAS 1000 Bio Imaging Analyzer (Fujix).

Genomic and cDNA Screening of Mutations and Sequencing
In order to identify the putative alteration(s) present in the RFC1 gene in CEM/MTX and CEM/MTX-LF cells we have undertaken several experimental approaches.
Construction of a cDNA Library from CEM/MTX Cells and Screening of RFC cDNA Clones-A cDNA library was constructed from CEM/ MTX poly(A) ϩ mRNA (2 g) as described previously (52) and screened with a 32 P-oligo-labeled human RFC cDNA probe. Several clones were isolated of which a full-length cDNA clone was fully sequenced thus revealing only a single G to A substitution at nucleotide position 227 (based on the nucleotide numbering of Prasad et al. (15), GenBank TM accession no. UI15939).
Reverse Transcriptase-Polymerase Chain Reaction (PCR) of the Entire RFC Coding Region-1-g aliquots of total RNA isolated from parental CEM, CEM/MTX, and CEM/MTX-LF cells were reverse transcribed using Super-Script II reverse transcriptase (Life Technologies, Inc.). Subsequently a PCR-based amplification of total cDNA was performed using the high fidelity Taq polymerase according to the manufacturer's protocol (Boehringer Mannheim); this was performed using primers previously described by Gong et al. (25). The reverse transcriptase-PCR cDNA products were cloned into the pGEM-T Easy vector system (Promega), and several clones from each cell line were fully sequenced (see Table III under "Results").
Genomic PCR-Single Strand Conformational Polymorphism and Isolation of Multiple Clones-High molecular weight genomic DNA from parental CEM, CEM/MTX and CEM/MTX-LF cells was amplified with Taq polymerase in 1xPCR buffer supplied by the manufacturer (Promega) in the presence of [␣-32 P]dATP. PCR was performed using the conditions described by Gong et al. (25); the downstream primers for exon 2 and exon 3 were those previously described (25), whereas the upstream primers were various 5Ј-intronic-based oligonucleotides. Aliquots of the PCR products were resolved on nondenaturing polyacrylamide gels using the single strand conformational polymorphism analysis as described previously (53,54). Following thymidylate synthase screening for changes in single strand DNA fragment mobility, the PCR products were cloned into the pGEM-T Easy vector system (Promega) after which a large number of wild type and altered clones were sequenced (see Table III under "Results").
Genomic PCR and Direct Sequencing-We have also undertaken a direct sequencing approach of PCR-amplified genomic DNAs from CEM, CEM/MTX, and CEM/MTX-LF cells. This PCR was performed with nonradioactive deoxynucleotides triphosphate. DNA sequence analysis of multiple genomic PCR products, genomic PCR clones and cDNA clones was performed by the fluorescent dideoxy chain termination method using ABI 373A and 377 sequencers (DNA Sequencing Unit, The Weizmann Institute, Rehovot, Israel).

Determination of Intracellular Folate Pools
Measurement of reduced folates was made by a series of assays that are based on enzymatic cycling of reduced folates to N 5 ,N 10 -methylenetetrahydrofolate (CH 2 THF) followed by entrapment into a stable ternary complex with excess thymidylate synthase and [ 3 H]FdUMP (20 Ci/mmol) (55,56). Frozen cell pellets were thawed (25 ϫ 10 6 cells/ml in an extraction buffer containing 50 mM Tris-HCl, 50 mM sodium ascorbate, and 1 mM EDTA, pH 7.4. Cells were lysed by boiling for 3 min and centrifuged at 15,000 ϫ g for 10 min at 4°C. Aliquots (20 -60 l) of the supernatant were used to determine CH 2 THF, THF, DHF, leucovorin, and 5-CH 3 THF. The 10-CHO-THF pool was assayed in the presence of 0.4 milliunit of 10-CHO-THF dehydrogenase and 10 M NADP instead of the previously reported 10-CHO-THF deacyclase system. Protein concentration was determined by the method of Bradford (57).

RESULTS
Folate Dependence of CEM/MTX Cells-Defective transport via the RFC has been associated with Ͼ200-fold resistance to MTX in human CEM/MTX leukemia cells (5,22,43). However, previous studies from our laboratory (58) and others (59 -61) have shown that CEM/MTX cells surprisingly exhibited only a minor (3-10-fold) cross-resistance to 5,10-dideazatetrahydrofolate (DDATHF), a folate-based inhibitor of glycinamide ribonucleotide formyltransferase. Moreover, since the presumed defective RFC did not compromise growth of CEM/MTX cells in folate-replete medium (containing 2.3 M folic acid), we inves-tigated whether alternative routes for folic acid and DDATHF may be present in CEM/MTX cells. In order to obtain further insight into the mechanistic basis of folate-dependent growth of RFC-defective CEM/MTX cells, we rationalized that a gradual deprivation of folic acid could impose a selective pressure to increase the expression of the folate transporter that is responsible for this phenotype.
We noted that folic acid rather than purines/pyrimidines in undialyzed fetal calf serum was essential for cell growth of CEM/MTX cells (58). This was illustrated by the fact that growth of CEM/MTX cells was unaffected in medium containing 2.3 M folic acid, 10% dialyzed FCS, whereas cessation of cell growth was observed when CEM/MTX cells were transferred to medium containing 2 nM folic acid and 10% undialyzed FCS (data not shown). Another line of evidence for folic acid transport capacity in CEM/MTX cells was obtained from the marked protection by folic acid against growth inhibition by DDATHF; 11-and 330-fold protection in medium supplemented with 20 and 200 M folic acid, respectively (data not shown). Finally, Fig. 1 shows that while the accumulation of cells but for CEM/MTX-LF cells it was 17-fold higher than in wild type CEM cells (Table I). Affinity labeling of RFC in CEM-7A cells (which display a 30-fold overexpression of RFC protein over wild type CEM cells) (20) was only 2-fold higher than for CEM/MTX-LF cells. The increased expression of the folate transporter was stably retained when CEM/MTX-LF cells were grown for a period of 1 month in folate-replete medium (2.3 M folic acid). Although wild type RFC has a 2 orders of magnitude lower affinity for folic acid than for MTX (2,4,10), we used N-hydroxysuccinimide-[ 3 H]folic acid as a labeling reagent for CEM/MTX-LF cells. As expected, specific affinity labeling with this reagent was below the limit of detection for CEM cells and CEM/MTX cells, whereas CEM/MTX-LF cells surprisingly contained 2.1 pmol of [ 3 H]folic acid/mg of protein, a value which is even higher (1.5-fold) than that observed for the highly RFC overexpressing CEM-7A cells.
SDS-polyacrylamide gel electrophoresis analysis of N-hydroxysuccinimide-[ 3 H]MTX labeled CEM-7A and CEM/ MTX-LF cells, as well as N-hydroxysuccinimide-[ 3 H]folic acidlabeled CEM/MTX-LF cells showed a broadly migrating band which peaked at a molecular mass of 75-80 kDa for both CEM-7A cells and CEM/MTX-LF cells (results not shown). CEM/MTX-LF cells did not express membrane folate receptors (8,62,63) as judged from the lack of a high affinity binding of [ 3 H]folic acid and the absence of immunoreactivity with the membrane folate receptor-specific monoclonal antibody MOv18 (47,48,63). Flow cytometric analysis with MOv18 was found negative for CEM/MTX-LF cells as compared with folate receptor overexpressing nasopharyngeal KB-3-1 cells (results not shown). Furthermore, no evidence was found for the presence of a low pH folate transport route (64,65) (Fig. 3, panel B) it was demonstrated that CEM/MTX-LF cells did not express human membrane folate receptor, which is highly expressed in KB-3-1 nasopharyngeal carcinoma cells (8,62).
To investigate the molecular basis for this RFC1 mRNA overexpression, RFC1 gene copy numbers were examined by Southern blot analysis. Digestion of genomic DNA from CCRF-    (Fig. 4). Reprobing this Southern blot with a multidrug resistance (MDR1) cDNA probe confirmed that similar amounts of genomic DNA were being analyzed (not shown).
Characterization of (Anti)folate Transport Kinetic Parameters in CEM/MTX-LF Versus CEM Cells-The transport kinetic parameters (K m and V max ) for influx of [

H]MTX influx in CEM/MTX-LF cells was similar to that in CEM cells.
Anion Dependence of MTX Transport in CEM/MTX-LF Cells-A characteristic feature of (anti)folate transport via wild type RFC is a differential inhibitory effect by various (in)organic anions (3,4,10,20). We measured [ 3 H]MTX influx in CEM-7A cells (wild type RFC) and CEM/MTX-LF cells as a function of increasing concentrations of chloride in an isotonic extracellular buffer (Fig. 5).  -p), fractionated on a 0.8% agarose gel, transferred to Zetabind GT filter membrane (Bio-Rad), and probed with a 32 P-oligo-labeled RFC1 cDNA clone, hRFC25. The blot was hybridized and washed under high stringency conditions according to the manufacturer's instructions (Bio-Rad). Shown on the right are size markers given in kilobase pairs.  CEM/MTX-LF cells on the biological activity (growth inhibitory effects) of a selected group of (non)polyglutamatable and lipophilic folate analogues (5,66) are presented in Table IV. Except for the lipophilic DHFR inhibitors, all of these folate analogues were reported to be efficiently transported via the wild type RFC (5). Table IV shows the growth inhibitory effects of the panel of folate analogues, following a 72-h continuous drug exposure, against CCRF-CEM cells, CEM/MTX cells, and CEM/MTX-LF cells in folate-deplete and folate-replete medium. CEM/MTX cells exhibited a high degree of cross-resistance to EDX (410-fold), ZD1694 (127-fold), and GW1843 (220fold). An intermediate level of cross-resistance was also noted for the nonpolyglutamatable folate analogues PT523 (30-fold) and ZD9331 (18-fold), whereas only a modest (7-and 4-fold) cross-resistance was found for the two glycinamide ribonucleotide transformylase inhibitors DDATHF and AG2034, respectively. In contrast, CEM/MTX cells displayed collateral sensitivity to the lipophilic DHFR inhibitor TMQ. In folate-deplete medium, CEM/MTX-LF cells displayed a complete wild type CEM cell sensitivity to ZD1694 and GW1843 and even greater drug sensitivity than CEM cells for ZD9331, DDATHF, and AG2034. CEM/MTX-LF cells in folate-deplete medium also exhibited an increased sensitivity for MTX, EDX, and PT523, but the drug sensitivity of wild type CEM cells was not reached. Remarkably, a dramatic drug resistance was observed for ZD-1694 (513-fold), DDATHF (822-fold), and AG2034 (811-fold) when CEM/MTX-LF cells were transferred to folate-replete medium. Likewise, growth of CEM/MTX-LF cells in folatereplete medium provoked a resistance to the lipophilic antifolates TMQ (21-fold) and pyrimethamine (154-fold; IC 50 values, 64 M versus 0.4 M in folate-replete versus folate-deplete medium, respectively), as well as EDX (23-fold) and MTX (6-fold). It is noteworthy that in folate-replete medium CEM/MTX-LF cells retained appreciable sensitivity to PT523, ZD-9331, and GW1843 as compared with CEM/MTX-LF cells in folate-deplete medium.

FIG. 5. Chloride dependence of [ 3 H]MTX influx in CEM-
Enzyme Levels in CEM, CEM/MTX, and CEM/MTX-LF Cells-The levels of target enzymes (DHFR and thymidylate synthase) and (anti)folate metabolizing enzymes (folylpolyglutamate synthetase and folylpolyglutamate hydrolase) were de-  termined in CEM, CEM/MTX, and CEM/MTX-LF cells. Table V shows that DHFR levels were similar in CCRF-CEM and CEM/ MTX cells, but were 4-fold increased in CEM/MTX-LF cells, most likely due to the use of MTX during the selection procedure of these cells (see "Experimental Procedures"). No major differences in levels of thymidylate synthase catalytic activity and FdUMP binding sites were observed among the three cell lines. Folylpolyglutamate synthetase activity in CEM/MTX cells was 1.7-fold lower than in CEM cells. In CEM/MTX-LF cells, folylpolyglutamate synthetase activity increased to the level of wild type CEM cells. Finally, the activity of folylpolyglutamate hydrolase was lowest in CEM/MTX cells and 1.6and 2.5-fold higher in CEM and CEM/MTX-LF cells, respectively. (Table VI) showed that in folate-replete medium CEM/MTX-LF cells had a total intracellular folate pool (mainly 10-CHO-THF, CH 2 THF, and THF) of 500 pmol/mg of protein, which is almost 500-fold higher than for CEM/MTX-LF cells maintained in folate-deplete medium. In addition, under folate-replete growth conditions, the intracellular folate pools in CEM/MTX-LF cells were markedly higher than in CEM cells (7-fold) and CEM/ MTX cells (24-fold).

Polyglutamylation of MTX and EDX in CCRF-CEM and CEM/MTX-LF Cells-The accumulation and formation of polyglutamates of [ 3 H]MTX and [ 3 H]EDX was analyzed in
CEM and CEM/MTX-LF cells in folate-deplete and folate-replete medium (Fig. 6) showed a similar picture as for MTX, 47% of EDX was unmetabolized, and the Glu 3 , Glu 4 , and Glu 5 metabolites comprised only 8, 6, and 5% of the total accumulated drug, respectively.

DISCUSSION
An increased K m and/or decreased V max of RFC-mediated transport has been recognized as a common mechanistic basis of transport-related resistance to antifolates (4,6,7,9,12,(22)(23)(24)(25)(26)(27)(28)(29)(30)43). In contrast, this study describes a first case for human cells in which a decreased K m and increased V max for transport of folic acid and reduced folates provides a phenotype that confers resistance to various antifolates. This novel mechanism of antifolate resistance is associated with the overexpression in human CEM/MTX-LF leukemia cells of a structurally altered RFC that is characterized by a markedly increased affinity (decreased K m ) for folic acid (31-fold) and leucovorin (9-fold) transport as compared with the wild type RFC. In folate-replete medium, the preferential folic acid transport results in a markedly expanded intracellular folate pool. Consequently, polyglutamylation of antifolate compounds is markedly impaired, thereby abolishing their cytotoxic activity. With respect to lipophilic antifolates, the increased intracellular folate pool leads to circumvention of target enzyme inhibition and consequent antifolate resistance.
Previous studies from our laboratory (58) and Matherly et al. (59) had provided preliminary evidence that CEM/MTX cells expressed a folate transporter with altered kinetic properties as compared with the wild type RFC. This evidence was based on observations that CEM/MTX cells displayed a differential drug sensitivity for the tetrahydrofolate-based antifolate DDATHF as compared with 2,4-diaminopteridine or quinazoline-based antifolates such as MTX and ZD-1694 (5,59). However, since the expression level of this putative folate transporter in CEM/MTX cells was very low (12,43), a detailed functional characterization of this transporter was not feasible. In the present study we were able to increase the expression of this folate transporter by adapting CEM/MTX cells to a gradual deprivation of folic acid in the cell culture medium of CEM/ MTX cells. Such a strategy has previously been proven successful in up-regulating the expression of the wild type RFC or membrane folate receptors in human and murine leukemia cells (20,42). Indeed, CEM/MTX-LF cells had undergone 10fold RFC gene amplification and subsequent overexpression of the principal 3.1-kilobase RFC transcript. Consequently, affinity labeling studies (Table I) revealed that the folate-transporter in CEM/MTX-LF cells was overexpressed from below the limit of detection to a level 17-fold higher than that present in wild type CEM cells and comparable to the RFC overexpressing CEM-7A cells (20).
The transport kinetic data with CEM/MTX-LF cells suggested that the marked decrease in the transport K m for folic acid and leucovorin was a result of a structural alteration in the RFC. Various genomic and cDNA analyses followed by sequencing of the RFC gene from CEM/MTX cells consistently revealed both the mutant allele (i.e. G 3 A at nucleotide position 227, see Table III) and the wild type allele. This was in good agreement with the full-length RFC cDNA clones isolated and sequenced in which both the mutant and wild type RFC alleles were present in CEM/MTX cells. These consistent findings suggested that this glutamate to lysine substitution at amino acid 45 in the human RFC in CEM/MTX cells was heterozygous. In contrast, sequencing of the RFC PCR products of the subline CEM/MTX-LF revealed primarily the mutant allele. When combined with the finding of the 10-fold RFC1 gene amplification in CEM/MTX-LF cells (Fig. 4), it appears that the mutant RFC allele but not the wild type allele has undergone gene amplification. According to this scenario of selective mutant allele amplification, the resultant overexpression of the mutant RFC with glutamate 45 to lysine may equip CEM/MTX-LF cells with the ability to preferably transport folic acid thereby enabling their survival under folate deplete con-ditions. Further evidence for a structural alteration in RFC in CEM/MTX-LF cells was also revealed by the striking anion (Cl Ϫ ) dependence for [ 3 H]MTX influx (Fig. 5). Unlike wild type RFC (with the negatively charged Glu-45 residue), functional activity of RFC in CEM/MTX-LF cells is stimulated by chloride, conceivably by Cl Ϫ neutralization of the positively charged Lys-45 residue (67).
These findings stand in perfect agreement with recent results obtained by the group of Goldman (67) which characterized a MTX-resistant murine L1210 cell line with a severe transport impairment that was derived using a chemical mutagen; the same lysine for glutamate 45 (E45K) substitution was observed due to a G 3 A in this mutated murine RFC1 gene. Strikingly, this murine E45K mutant displayed the very same chloride-dependence for MTX influx. At the transport kinetic level, these L1210 mutant cells displayed a 7-fold decreased K m for folic acid transport, whereas the K m for MTX increased by 7-fold. In contrast, here we noted a 31-fold decrease in the K m for transport of folic acid, whereas the K m for MTX, as well as for the antifolates EDX, PT523, GW1843, ZD9331, and ZD1694, 2 was largely unchanged in CEM/ MTX-LF cells. Thus, there seems to be a disparity between the transport kinetic data of the murine E45K mutant (67) and those observed for human CEM/MTX-LF cells. Although the molecular basis for this disparity in kinetic data between the human and murine E45K mutants is yet unclear, several findings suggest that the rodent and human RFC bear some structural and kinetic differences that are likely to translate into some differences in folate transport properties: 1) It should be recognized that kinetic differences in antifolate transport between murine and human RFC are not unusual. The best example is a 15-20-fold higher K m of murine RFC for transport of the antifolate GW1843 as compared with human RFC (5,68). Hence, since such prominent kinetic differences distinguish between the wild type murine RFC from its human counterpart, it is likely that identical mutations occurring in these carriers may show kinetic disparities. 2) While the rodent and human RFC proteins bear a significant amino acid sequence homology especially at predicted transmembrane domains, the C-terminal end of the murine RFC is shorter by 77 amino acids than its human counterpart (15)(16)(17)(18)(19). 3) In transfection experiments in which the human RFC was transfected into rodent cells, hybrid kinetic properties were observed, thus supporting the notion of differences in folate transport function between the two carriers (13). 4) Whereas murine RFC overexpression in cDNA-transfected mouse leukemia L1210 cells shows little MTX concentrative ability (69), transfection of the human RFC cDNA into transport-defective ZR-75-1/MTX carcinoma cells, shows a prominent MTX concentrative ability (70). 5) The human but not the mouse RFC1 gene undergoes a substantial alternative splicing that results in truncation of ϳ100 amino acids (71). 6) Finally, the human RFC undergoes an extensive glycosylation, whereas the rodent RFC does not (11,12).
Collectively, recent studies with rodent cells (26 -28, 67) and the present one with human cells indicate that single amino acid substitutions in the RFC can ultimately influence folate homeostasis by altering the transport efficiency of natural reduced folate cofactors. The impact of this on the growth inhibitory potential of antifolates can vary substantially. Tables IV and VI illustrate that a relatively small increase in the intracellular folate pools (e.g. 7-fold between CEM and CEM/ MTX-LF cells in folate-replete medium) can result in a disproportional level of 1,000-fold resistance to some antifolates. Among the most sensitive for changes in intracellular folate pools, due to competition at the level of folylpolyglutamate synthetase and consequent abolishment of polyglutamylation, were antifolates that are dependent on polyglutamylation (ZD1694, AG2034, and DDATHF) (72)(73)(74). Furthermore, a marked expansion of the intracellular folate pool provoked resistance to lipophilic antifolate inhibitors of DHFR. In this case, the increased reduced folate pool may bypass a decreasing folate pool that arises from the inhibition of DHFR by TMQ/ pyrimethamine. Moreover, TMQ/pyrimethamine inhibition of DHFR brings about a rapid increase in the level of dihydrofolate, which upon very efficient polyglutamylation can readily displace TMQ/pyrimethamine from DHFR (75,76). In this way, DHFR enzyme activity can be restored, at least in part, thereby resulting in resistance to these DHFR inhibitors. Expanded folate pools had a less dramatic effect on the growth inhibitory potential of antifolates for which polyglutamylation is of lower importance (e.g. MTX, EDX, and GW1843) since monoglutamate forms of these compounds are already potent inhibitors of their target enzymes (5,34). Consistently, it is of importance to note that the growth inhibitory effect of the nonpolyglutamatable antifolates PT523 and ZD-9331 was retained even under conditions of high intracellular folate levels.
Taken altogether, these results point to the key role of RFCmediated folate transport in controlling intracellular folate pool size. Along with the recently identified alterations in folate efflux, which can also provoke an increased intracellular folate pool (77,78), this will provide novel mechanistic insights into transport-related resistance to antifolates which may occur in the clinical setting (79).