Impaired Membrane Transport in Methotrexate-resistant CCRF-CEM Cells Involves Early Translation Termination and Increased Turnover of a Mutant Reduced Folate Carrier*

The basis for impaired reduced folate carrier (RFC) activity in methotrexate-resistant CCRF-CEM (CEM/Mtx-1) cells was examined. Parental and CEM/Mtx-1 cells expressed identical levels of the 3.1-kilobase RFC transcript. A ∼85-kDa RFC protein was detected in parental cells by photoaffinity labeling and on Western blots with RFC-specific antiserum. In CEM/Mtx-1 cells, RFC protein was undetectable. By reverse transcriptase-polymerase chain reaction and sequence analysis, G to A point mutations were identified in CEM/Mtx-1 transcripts at positions 130 (P1; changes glycine 44 → arginine) and 380 (P2; changes serine 127 → asparagine). A 4-base pair (CATG) insertion detected at position 191 (in 19–30% of cDNA clones) resulted in a frameshift and early translation termination. Wild-type RFC was also detected (0–9% of clones). Wild-type RFC and double-mutated RFC (RFCP1+P2) cDNAs were transfected into transport-impaired K562 and Chinese hamster ovary cells. Although RFC transcripts paralleled wild-type protein, for the RFCP1+P2 transfectants, disproportionately low RFCP1+P2 protein was detected. This reflected an increased turnover of RFCP1+P2 over wild-type RFC. RFCP1+P2 did not restore methotrexate transport; however, uptake was partially restored by constructs with single mutations at the P1 or P2 loci. Cumulatively, our results show that loss of transport function in CEM/Mtx-1 cells results from complete loss of RFC protein due to early translation termination and increased turnover of a mutant RFC protein.

Despite the availability of newer antifolates, methotrexate (Mtx) 1 continues to play an important role as an antineoplastic agent. To reach its intracellular target, dihydrofolate reductase, the preferred route of Mtx entry involves the reduced folate carrier (RFC; 1, 2). RFC transport of Mtx is critical to drug action because of its role in generating sufficient unbound intracellular antifolate to sustain maximal enzyme inhibition (1). Furthermore, high levels of Mtx are also necessary for the synthesis of Mtx polyglutamates (1).
Defective membrane transport of Mtx by RFC has been identified as a major mechanism of Mtx resistance (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11). Transport alterations can manifest as reduced rates of carrier translocation (reduced V max ), decreased affinities for transport substrates (increased K t ), or both, and may involve decreased levels of normal RFC (6) or the expression of structurally altered RFC proteins (7)(8)(9)(10)(11). For instance, in Mtx-resistant K562 (K500E) cells, impaired Mtx transport is accompanied by decreased RFC transcripts and protein (6). A G to A transition at position 890 of the murine RFC cDNA resulted in a substitution of serine 297 by asparagine and a selective decrease in Mtx binding affinity (ϳ4-fold) without effects on other antifolate analogs (aminopterin, 10-ethyl-10-deazaaminopterin; Ref. 9). Likewise, replacement of serine 46 by asparagine (10) or glutamate 45 by lysine (11) in murine RFC resulted in greater impairment of uptake for Mtx than (6S)-5-formyl tetrahydrofolate. In severely transport defective L1210 cells (Mtx r A), loss of transport activity appeared to reflect a single (G to C) point mutation at nucleotide 429 of the murine RFC cDNA sequence which resulted in the substitution of proline 130 by alanine (7). However, these cells also contained a wild-type RFC allele that was not transcribed. A silent wild-type RFC allele was described for Mtx-resistant MOLT-3 cells (MOLT-3/Mtx 10,000 ; Ref. 8). Moreover, two mutations in the RFC coding region were detected which resulted in the creation of new stop codons and synthesis of truncated nonfunctional RFCs (8).
In this report, the molecular mechanisms responsible for the transport-impaired phenotype (ϳ3% of wild-type) of Mtx-resistant (ϳ243-fold) CCRF-CEM (CEM/Mtx-1;12) cells were examined. We show that although the levels of RFC transcripts are essentially unchanged from wild-type cells, there is a complete loss of RFC protein due to early translation termination and increased turnover of a double mutant RFC protein. The residual transport activity previously described in this transport-impaired line (12) presumably reflects extremely low levels of wild-type RFC and/or, possibly, non-RFC modes of Mtx uptake (13)(14)(15). 32 (20 Ci/mmol) and [4, H]leucine (120 Ci/mmol) were purchased from Moravek Biochemicals (Brea, CA). Unlabeled Mtx was provided by the Drug Development Branch, NCI, National Institutes of Health, Bethesda, MD. Both labeled and unlabeled Mtx were purified by high performance liquid chromatography prior to use (16). GW1843U89 (17) was obtained from Glaxo-Wellcome Pharmaceuticals (Research Trian-* This work was supported by National Institutes of Health Grant CA 53535 and a grant from the United Way of Michigan (Detroit, MI). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Cell Culture-Wild-type CCRF/CEM and transport-impaired CEM/ Mtx (18) lymphoblastic leukemia lines were gifts of Dr. Andre Rosowsky (Boston, MA). Cells were cloned in soft agar (6,19) and clonal lines (designated CEM-4 and CEM/Mtx-1 for the parental and Mtx-resistant cells, respectively) were used for all experiments. Transport-deficient K500E cells were selected from wild-type K562 cells by cloning in soft agar with 500 nM Mtx (6). K500E cells were transfected with wild-type RFC (KS43) to generate the K43-1 and K43-6 sublines, as described previously (6). The cell lines were maintained in RPMI 1640 medium as described previously (6,12).
Transport-defective Mtx-resistant Chinese hamster ovary (CHO) cells, MtxRIIOua R 2-4 (20), were a gift of Dr. Wayne Flintoff (London, Ontario, Canada). Cells were grown in ␣-minimal essential medium with 10% iron-supplemented bovine calf serum, penicillin (100 units/ ml), and streptomycin (100 g/ml). The pC43/10 CHO line was derived from MtxRIIOua R 2-4 cells by transfection with the full-length human RFC cDNA (KS43; Ref. 21). CHO cells were grown as monolayers for transfection and general maintenance; for transport experiments, cells were grown in suspension in spinner flasks.
Southern and Northern Analysis-Genomic DNAs were isolated from cultured cells using the Puregene TM DNA isolation kit from Gentra Biosystem, Inc. (Minneapolis, MN). Aliquots (10 g) were digested with restriction enzymes (either BamHI or HindIII), fractionated on a 0.6% agarose gel, and blotted onto a nylon membrane (Genescreen Plus, DuPont), following standard protocols (22).
Total RNA was isolated from log phase cells using the TRIzol Reagent (Life Technologies, Inc.). RNA samples were analyzed on a formaldehyde-agarose gel, exactly as described previously (21). Equal loading was established by probing with 32 P-labeled ␤-actin cDNA or by staining with ethidium bromide. All membranes were hybridized with 32 P-labeled full-length RFC cDNA and processed as described previously (21).
Analysis of RFC Coding Region and Genomic Sequence-RFC cDNAs from parental and CEM/Mtx-1 cells were synthesized from total RNA with random hexamers using a RT-PCR kit from Perkin-Elmer. Four sets of PCR primers were used to generate overlapping partial cDNAs spanning the entire RFC coding region. The PCR primers for RFC cDNA amplification are shown in Table I. PCR conditions were 94°C for 30 s, 63°C for 45 s, and 72°C for 1 min (35 cycles), and ending with 72°C for 7 min (1 cycle). PCR products were subcloned into the pCR 2.1 plasmid using the T-A cloning kit (Invitrogen) and the nucleotide sequences were determined by dideoxynucleotide sequencing (23). RT-PCR reactions were repeated 2-3 times for regions containing the P 1 and P 2 mutations and for each primer set, multiple cDNA clones were sequenced.
Genomic fragments containing point mutations identified in the CEM/Mtx-1 cDNAs were PCR amplified and the PCR products subcloned and sequenced as described above. Primers for genomic amplification were based on the human RFC cDNA (21) and gene (24,25) nucleotide sequences. For amplifying the fragment containing the P 1 mutation, a nested PCR approach was used. In the primary PCR reaction, two RFC intron-specific primers, RFC-IP1 (5Ј-ctgcagaccatcttccaaggtgccctga; upstream of the splice acceptor site at Ϫ49) and RFC-IP2 (5Ј-gcagaccatcttccaaggtgccctga; downstream of the splice donor site at 189), were used. For the secondary nested PCR reaction, the primers used were the exon-specific primer P8 (Table I) and another intronspecific primer RFC-IP3 (5Ј-acctactggtgctgctgcccctgc; downstream of the splice donor site at 189). The fragment containing the P 2 mutation was amplified with intron-specific RFC-IP4 (5Ј-gcggcagcattgctaacacctggtg; upstream of the splice acceptor site at 190) and exon-specific P7 (Table I) primers. PCR conditions for amplifying genomic DNAs were 94°C for 10 s, 63°C for 60 s, and 72°C for 60 s (35 cycles), and 1 cycle of 72°C for 7 min.
Preparation of Recombinant RFC Antiserum-The complete coding sequence of the KS43 RFC cDNA (21) was subcloned into a pGEX glutathione S-transferase (GST) fusion vector (Pharmacia Biotech, Piscataway, NJ). Following transformation of Escherichia coli (BL21) cells and induction by isopropyl-␤-D-thiogalactoside (0.5 mM) for 4 h at room temperature, GST-RFC fusion proteins were purified from bacterial lysates by affinity chromatography using glutathione-Sepharose 4B (Pharmacia Biotech), as recommended by the manufacturer. Authenticity and purity of the purified RFC fusion protein were confirmed by Coomassie Blue staining and Western analysis with anti-GST (Pharmacia Biotech) and RFC peptide-specific (RFC/ps; Ref. 26) antibodies. Anti-GST-RFC antiserum was raised in rabbits using purified GST-RFC fusion protein as antigen (Pocono Rabbit Farms and Laboratories, Canadensis, PA). Both immune and preimmune sera were purified on protein A-agarose columns prior to use (27).
Preparation of Plasma Membranes and Western Analysis-Plasma membranes were prepared by differential centrifugation (19,28). Where noted, particulate membrane fractions were additionally purified on discontinuous sucrose gradients (19,28). Plasma membrane purity and endoplasmic reticulum contamination of crude particulate and sucrose density gradient-purified membranes were established by 5Ј-nucleotidase (29) and NADPH-cytochrome c reductase (30) assays, respectively.
Membrane proteins were electrophoresed on 7.5% gels in the presence of SDS (31) and electroblotted onto polyvinylidene difluoride membranes (DuPont) for detection with protein A-purified GST-RFC antibody and enhanced chemiluminescence (Pierce, Rockford, IL). A few experiments employed RFC peptide-specific (RFC/ps) antibody (26). Light emission was recorded on x-ray film with various exposure times, and the signal was analyzed with a computing densitometer and Im-ageQuant software (Molecular Dynamics, Sunnyvale, CA). For some experiments, the heterogeneously glycosylated RFCs were enzymatically deglycosylated with N-glycosidase F (Boehringer Mannheim), as described previously (6,28).

Transport of [ 3 H]Mtx in Transfected Cells-Initial [ 3 H]
Mtx uptake rates were determined over 180 s using 1-2 ϫ 10 7 cells/ml (6,12,19,21) and a Mtx concentration of 0.5 M. The levels of intracellular radioactivity were expressed as picomoles/mg of protein, calculated from direct measurements of radioactivity and protein contents of the cell homogenates. Protein assays were by the method of Lowry et al. (32). Kinetic constants (K t and V max ) were calculated from Lineweaver-Burk plots.

Impaired Mtx Transport in CEM/Mtx-1 Cells Is Independent of Changes in RFC Transcripts or Gene Structure-Northern analysis of total RNAs from parental CCRF-CEM (CEM-4) and
Mtx-resistant CEM/Mtx-1 cells showed that essentially identical levels of a major 3.1-kb RFC mRNA transcript were expressed (Fig. 1) despite a ϳ33-fold difference in relative Mtx transport (12). Although a 1-kb RNA species hybridized with the RFC cDNA in parental cells and a unique 9.5-kb band was detected in CEM/Mtx-1 cells (Fig. 1), the significance of these forms is not clear. These bands were still present even when poly(A) mRNAs were used for Northern analysis (data not shown). Restriction analysis (BamHI or HindIII) of genomic DNAs from CEM-4 and CEM/Mtx-1 cells did not reveal any major alterations in RFC gene organization or copy number between the lines (data not shown).
Analysis of RFC Protein Expression in CEM/Mtx-1 Cells by Western Blotting and Photoaffinity Labeling-Expression of RFC protein in plasma membranes from parental and CEM/ Mtx-1 cells was analyzed by Western blotting using antibody to recombinant RFC fusion protein (GST-RFC) and chemiluminescence detection, and by photoaffinity labeling with APA-[ 125 I]ASA-Lys (6,21). For both methods, a broadly migrating RFC band centered at ϳ85 kDa was identified in parental cells (Fig. 2, left panel, and Fig. 3, respectively). Identical results were obtained on Western blots with peptide-specific (RFC/ps) antiserum (not shown). Slight differences were seen in the relative migrations for RFC, reflecting the different gel systems used for separation (7.5% for the Western versus 4 -10% for the photoprobe experiments). By both approaches, the major bands identified as RFC were converted to a single ϳ65-kDa deglycosylated form by treatment with N-glycosidase F (shown for the immunoblotted RFC in parental CCRF-CEM cells; Fig. 2, right panel). This is the size predicted from the RFC cDNA sequence (21,33,34). By contrast, in CEM/Mtx-1 cells none of the ϳ85-kDa RFC protein was detected either by Western blotting with anti-GST-RFC (Fig. 2) or peptide-specific antiserum (not shown), or by photoaffinity labeling with APA-[ 125 I]ASA-Lys (Fig. 3). However, an unidentified 42-kDa protein was specifically labeled with the photoprobe (Fig. 3). Although there were no changes in the background staining on Western blots following treatment of CEM/Mtx-1 proteins with N-glycosidase F, the 42-kDa photolabeled band was converted to ϳ37 kDa by this treatment (not shown).
Identification of Mutations in the RFC Coding Sequence in CEM/Mtx-1 Cells-The RFC coding sequences from parental CCRF-CEM and CEM/Mtx-1 cells were examined by RT-PCR and dideoxynucleotide sequencing of the PCR products. Four primer sets were used to amplify the entire RFC coding sequence (P1/P3, P4/P7, P8/P7, and P9/P10; Table I). Three alterations were identified in a 572-bp segment (positions Ϫ23 to 549, where 1 is the translational start site) amplified from CEM/Mtx-1 cells by primer set P7/P8 and encoding the RFC amino terminus. These include two G to A point mutations at positions 130 (designated P 1 ; nucleotide position 1 is the translation start) and 380 (P 2 ) in all of the 16 CEM/Mtx-1 clones sequenced, and a 4-bp (CATG) insertion at position 191 in 3 of the clones. By contrast, none of the 9 cDNA clones amplified with P7/P8 from parental CCRF-CEM cells contained any alterations from wild-type RFC sequence (21,33,34).
Analogous results were obtained by amplification of a fragment containing the P 2 locus (positions 141-549) with the P4/P7 primer set (21/23 with a P 2 mutation, including 7 with insertion at position 191). However, 2 of 23 clones derived from CEM/Mtx-1 also contained wild-type sequence at this position. All of the 15 clones amplified from parental CCRF-CEM cells with P4/P7 primers contained wild-type sequence at the P 2  locus; however, for 3 of these wild-type clones, the 4-bp CATG insertion was detected.
PCR amplification of CEM/Mtx-1 genomic DNA with both intron-and exon-specific primers, and sequencing of the PCR products confirmed both P 1 and P 2 mutations at the genomic level. Again, neither of the mutations was detected in parental cells. Notably, the 4-bp (CATG) insertion could not be found in any of the genomic DNAs amplified with the P8 and RFC-IP3 primers, suggesting that this probably arose from alternative splicing of intron sequence at the splice donor junction at position 189 (24,25). Not surprisingly, a number of CEM/Mtx-1 clones (1 of 8 for P 1 and 3 of 4 for P 2 , amplified with separate primer sets) exhibited wild-type genomic sequence.
The P 1 mutation would result in a change of glycine 44 to arginine and, by computer prediction (Garnier-Robson-Osguthorpe; 35), introduce an altered secondary structure in the region immediately upstream from this locus. The P 2 mutation results in a substitution of serine 127 by asparagine in a conserved putative transmembrane domain (residues 124 to 144), yet secondary structure is seemingly unaffected. The CATG insertion at position 191 of the RFC coding sequence generates a frameshift and early translation termination at position 1176, resulting in a truncated RFC protein (ϳ48 kDa) with only 11% of recognizable primary sequence. However, this would appear to account for no more than 30% of the loss of the full size RFC protein.
Characterization of Mutant RFC P1ϩP2 Protein-The lack of detectable RFC protein in CEM/Mtx-1 cells may reflect its inefficient synthesis, membrane targeting, or decreased stability of the double-mutated RFC. To evaluate these possibilities, wild-type RFC and a mutant construct containing G to A mutations at both positions 130 and 380 (designated RFC P1ϩP2 ) were transfected into transport-defective K562 (K500E) and CHO (MtxRIIOua R 2-4) cells, for comparison to wild-type RFC (6, 21). As described previously for K500E and MtxRIIOua R 2-4 transfectants (designated K43-1 and K43-6, and pC43/10, respectively) expressing wild-type RFC, RFC protein in plasma membranes is reflective of the high levels of RFC transcripts (Refs. 6 and 21; data are shown in Fig. 4 for the human K43-6 (left panel) and hamster pC43/10 (right panel) transfectants).

FIG. 4. Expression and transport of wild-type RFC and RFC P1؉P2 cDNA constructs in K500E (left panels) and CHO (right panels)
transfectants. Panels A and D, equal amounts of total RNAs (15 g) from K500E (K43-6, K43 P1ϩP2 /4, and K43 P1ϩP2 /22) and CHO (pC43/10 and RII P1ϩP2 /13) transfectants, and untransfected cells (K500E and MtxRIIOua R 2-4) were analyzed on Northern blots probed with a 32 P-labeled human RFC cDNA (KS43; Ref. 20). Equal loading was established by staining with ethidium bromide (not shown). Size markers (18 S and 28 S, or molecular standards) are noted. Panels B and E, particulate membrane fractions from the K500E and CHO transfectants were analyzed on Western blots with untransfected cells. The protein amounts (in g) analyzed for each of the sublines are noted. Detection was with anti-GST-RFC antibody and enhanced chemiluminescence. Migrations of wild-type RFC (ϳ85 kDa) and RFC P1ϩP2 (70 kDa) are noted. In the wild-type RFC transfectants, a 70-kDa band comigrating with RFC P1ϩP2 is seen at some exposures, likely reflecting less glycosylated variants of wild-type RFC. Panels C and F, initial uptake rates for [ 3 H]Mtx (0.5 M) were assayed over 180 s as described under "Experimental Procedures." Mean data are shown for duplicate incubations from a single representative experiment. RFC P1ϩP2 transcripts were, likewise, detected on Northern blots for the majority of both K500E (8 of 12) and MtxRII-Oua R 2-4 (11 of 18) transfectants (not shown), invariably as multiple hybridizing bands (Fig. 4, A and D, shows representative data). The smallest band approximated the size (ϳ2 kb) of the RFC P1ϩP2 cDNA so that all transcript forms were of sufficient size to encode the RFC P1ϩP2 protein. For only two K500E (K43 P1ϩP2 /4 and K43 P1ϩP2 /22) clones and one MtxRIIOua R 2-4 (RII P1ϩP2 /13) clone was RFC P1ϩP2 protein detected on Western blots (shown in Fig. 4, panels B and E). For these, the levels of mutant protein were exceedingly low (estimated by densitometry as 4 -8% of the wild-type value relative to levels of total RFC or RFC P1ϩP2 transcripts). Furthermore, RFC P1ϩP2 protein was distinctly smaller (ϳ70 kDa) than the wild-type carrier (Fig. 4). Both wild-type RFC and RFC P1ϩP2 completely reverted to 65-kDa deglycosylated forms upon treatment with N-glycosidase F (Fig. 5), establishing that these differences in carrier size reflected their extents of N-glycosylation.
Turnover of Wild-type RFC and RFC P1ϩP2 Proteins-The decreased levels of mutant RFC P1ϩP2 compared with wild-type RFC in transfected cells (and by extension, CEM/Mtx-1) may, in part, reflect differential rates of carrier degradation. To explore this possibility, K43-6 and K43 P1ϩP2 /22 transfectants were treated with 0.2 mg/ml cycloheximide (results in Ͼ98% inhibition of protein synthesis, as reflected in trichloroacetic acid-precipitable [ 3 H]leucine). Rates of exponential decline of wild-type RFC and RFC P1ϩP2 were assayed over 24 h on Western blots (Fig. 6). By this analysis, the level of wild-type RFC decreased by approximately 50% over 24 h. In contrast, RFC P1ϩP2 protein exhibited a rapid turnover (Fig. 6). The halflife of RFC P1ϩP2 was calculated as 2.0 Ϯ 0.56 h (mean Ϯ S.E.; n ϭ 3) and no RFC P1ϩP2 protein could be detected after 8 h following addition of cycloheximide.  (Fig. 4, panels C and F); i.e. initial rates of [ 3 H]Mtx uptake over 180 s were identical for the transfected cells and the untransfected lines from which they were derived.
[ 3 H]Mtx transport was partially restored for the single mutant RFC P1 and RFC P2 constructs, expressed in MtxRII-Oua R 2-4 cells (i.e. RII P1 /2, and RII P2 /6A and RII P2 /15A for RFC P1 and RFC P2 , respectively; Fig. 7). For the RII P2 /15A cells, expressing the highest levels of RFC P2 (8-fold greater than wild-type RFC in pC43/10), uptake of [ 3 H]Mtx (0.5 M) was ϳ50% of that for wild-type RFC-expressing cells. Although we were able to identify stably expressed RFC P1 (in RII P1 /2 cells) by screening over 20 G418-resistant colonies, expression was low (Fig. 7, inset). However, both RII P1 /2 and RFC P2 /15A exhibited sufficient transport activity to calculate kinetic constants for Mtx uptake (Table II). With both RFC P1 (RII P1 /2) and RFC P2 (RII P2 /15A), the K t values for Mtx were increased from that for wild-type RFC (ϳ11 and ϳ5-fold, respectively; Table  II). The absolute V max values for Mtx for both RFC mutant constructs were at least 70% of that for wild-type carrier (i.e. pC43/10 cells; Table II). When normalized to levels of immunoreactive RFC protein on Western blots (measured by densitometry; Fig. 7, inset), the relative V max for RFC P1 (RII P1 /2 subline) exceeded that for the wild-type carrier by 3-fold, whereas the relative V max for RFC P2 (RII P2 /15A subline) was only 16% of the wild-type value (Table II). The V max (normalized)/K t values were 27 and 2.7%, respectively, of that for wild-type RFC. In contrast to RFC P1ϩP2 (see above), the substitution of arginine for glycine 44 in the single-mutated RFC P1 construct and asparagine for serine 127 in RFC P2 had no obvious effect on the processing to mature, glycosylated (ϳ85 kDa) RFC proteins (Fig. 7, inset). DISCUSSION CEM/Mtx-1 cells exhibit only 3% of normal levels of Mtx influx associated with a 4-fold increased K t and 6-fold decreased V max (12). This altered substrate binding extends to a range of folate and antifolate transport substrates (10-ethyl-10-deazaaminopterin, aminopterin, ZD1694, GW1843U89, (6R)-5,10-dideaza-5,6,7,8-tetrahydrofolate, folic acid, and leucovorin) and initially suggested to us that the synthesis of a structurally altered RFC might be responsible for the drugresistant phenotype (12). Although the demonstration of normal levels of a major 3.1-kb transcript in CEM/Mtx-1 cells lent further credence to this notion, no significant RFC protein could be detected by Western blotting or photoaffinity labeling. In MOLT3/Mtx 10,000 cells, the absence of immunoreactive RFC was previously attributed to the presence of mutations in the RFC coding sequence which resulted in early translation termination and the synthesis of severely truncated RFCs (8).
Since an analogous mechanism could occur in the CEM/Mtx-1 subline, we sequenced partial cDNAs amplified from CEM/ Mtx-1 transcripts. Indeed, a 4-bp (CATG) insertion was identified at position 191 of the RFC coding sequence which resulted in a frameshift and the use of a new stop codon at position 1176. Although this would generate a predicted ϳ48-kDa protein with only 11% of recognizable RFC sequence and unlikely to react with GST-RFC antibody on Western blots, the low frequency at which the 4-bp insertion was detected (19 -30% of cDNA clones) suggested, at most, its minor contribution to the lack of RFC expression in these cells.
Wild-type RFC sequence was also detected in a small number of CEM/Mtx-1 cDNAs. However, its low frequency, combined with the lack of a signal on Western blots (even at high protein loading; data not shown), indicated that an insignifi-cant amount of wild-type RFC protein was actually synthesized in these cells. The variable frequencies at which wild-type cDNA and genomic sequences were detected in these analyses may reflect the localization of RFC to chromosome 21 (21q22. Ref. 33) and presence of a third (and possibly wild-type) RFC allele due to a random trisomy 21 in the CEM/Mtx-1 subline (1 of 8 karyotypes). 2 Rather, the majority of RFC transcripts in the CEM/Mtx-1 subline contained G to A substitutions at both nucleotide positions 130 and 380 (in general, without the 4-bp insertion at position 191) which result in replacements of amino acids 44 and 127. The lack of detectable mutated RFC proteins in these cells could result from impaired translation of mutant RFC transcripts, and/or an accelerated degradation or inefficient plasma membrane targeting of mutant proteins. These possibilities could not be evaluated in CEM/Mtx-1 cells. Consequently, we expressed mutant RFC cDNAs in transport-defective CHO and human cells to better correlate levels of RFC transcripts and immunoreactive protein for comparison with wild-type RFC. As with the wild-type RFC transfectants (6,21), double-mutated RFC P1ϩP2 transcripts were observed on Northern blots at high frequencies for both CHO and K500E transfectants. However, only for the cells transfected with wild-type RFC constructs were significant accumulations of immunoreactive RFC protein detected. For the three clones which expressed sufficient mutant RFC P1ϩP2 protein for immunoblot detection, the carrier migrated as a ϳ70-kDa band, distinguishable from both native wild-type RFC (ϳ85 kDa) and enzymatically deglycosylated RFC (65 kDa).
Hence, the presence of the P 1 and P 2 mutations appears to alter processing to the mature N-glycosylated carrier and, likewise, results in markedly decreased levels of membrane RFC P1ϩP2 protein. This, in part, reflects the dramatically accelerated turnover of the mutant carrier, likely due to altered secondary and tertiary structures, however, differences in translation efficiencies between wild-type and RFC P1ϩP2 cannot be discounted as a contributing factor. Increased turnover rates of mutant proteins are well established as a mechanism for maintaining cellular homeostasis (36,37). For other integral membrane proteins, including cystic fibrosis transmembrane regulatory protein (38) and p-glycoprotein (39), increased rates of mutant protein degradation accompany incomplete glycosylation and impaired membrane targeting due to retention in the endoplasmic reticulum. However, for RFC P1ϩP2 , there was no evidence for endoplasmic reticulum retention and degradation since both wild-type RFC and mutant RFC P1ϩP2 co-localized with 5Ј-nucleotidase activity in particulate and sucrose gradient-purified plasma membrane fractions (data not shown). Furthermore, there were no significant differences in the levels of wild-type RFC or RFC P1ϩP2 between sucrose gradient-purified and crude particulate membrane fractions, differing ϳ3-fold in NADPH-cytochrome c reductase (an endoplasmic reticulum marker enzyme) activity. Thus, both carrier  forms are primarily targeted to the cell surface. It was of interest that both point mutations identified in CEM/Mtx-1 RFC resulted in replacement of highly conserved amino acids (glycine 44 and serine 127) and, together, they completely abolished transport activity. When expressed individually in transport-impaired CHO cells, both P 1 and P 2 mutant constructs exhibited low levels of transport activity. Although the presence of arginine 44 in the mutant RFC P1 actually increased the Mtx V max over wild-type RFC in RII P1 /2 cells, this was accompanied by an increased K t so that net transport was appreciably impaired. For RFC P2 , Mtx uptake was low due to effects on both K t and V max and was only detected in the RII P2 /15A transfectant expressing extremely high levels of the mutant carrier. Thus, both glycine 44 and serine 127, or the protein domains including these residues, are likely important for RFC function. Important functional roles were also implied for glutamate 45 (11) and serine 46 (10), and for alanine 132 (alanine 130 in the murine RFC; Ref. 7) from studies in transport-impaired L1210 cells.
Hence, our results with the double-mutated RFC P1ϩP2 construct in transfected CHO and K562 cells approximate those for the mutant RFC proteins and transcripts in CEM/Mtx-1 cells. The complete loss of RFC protein in the CEM/Mtx-1 subline appears to result from early translation termination and rapid turnover of the RFC P1ϩP2 protein, although inefficient translation from mutant RFC transcripts may also be a contributing factor. It is of interest that a very recent report by Jansen et al. (40) described a CEM/Mtx subline of apparently identical origin to the CEM/Mtx-1 clonal line described herein yet exhibiting distinctly different characteristics. These relate to relative levels of folate-dependent enzymes (folylpolyglutamate synthetase and thymidylate synthase; Ref. 12), 5-fold decreased levels of RFC transcripts compared with parental CCRF-CEM cells, and the presence of a point mutation at position 133 of the RFC coding sequence which results in the synthesis of a mutated carrier with a lysine substitution for glutamate 45. These discrepancies with our results cannot be explained simply by differences in experimental methodologies or data interpretation. Rather, the most likely explanation is that the CEM/Mtx cells (18) somehow changed during longterm culture and/or different clonal variants of CEM/Mtx were studied in the different laboratories.
In our studies, the complete absence of RFC protein and the total lack of [ 3 H]Mtx transport activity for RFC P1ϩP2 strongly implies that the residual uptake and anomalous substrate binding properties in CEM/Mtx-1 cells (12) are not due to the RFC P1ϩP2 . While the transport kinetics observed for this resistant subline are clearly incompatible with those of the wildtype carrier, the small amounts of wild-type RFC activity may be modulated by unknown endogenous factors (6,21,41,42). Alternatively, non-RFC modes of uptake (13-15) may also contribute to the CEM/Mtx-1 transport phenotype. Studies are underway to further explore these possibilities.