INTRODUCTION

Biochemical alterations associated with acquired resistance of tumor cells to classical folate analogues are highly diverse (1-5). This phenotypic diversity most likely reflects the common occurrence of multiple genomic alterations among clonal variants selected for resistance to these agents either in vitro or in vivo. As a consequence, acquired resistance to methotrexate and newer folate analogues under development will remain (6-8) a major limitation to their effective clinical utility. One of the determinants of cytotoxicity shared (9-11) by most folate analogues, whether targeted to dihydrofolate reductase or folate-dependent biosynthetic enzymes, is the process of intracellular polyglutamylation mediated by the enzyme folylpolyglutamate synthetase (FPGS).¹ This process is not only important to the conservation and efficient utilization by proliferating cells of folate coenzymes (12-18) but is responsible, as well, for converting folate analogues (9-18) to longer chain polyglutamate anabolites that are more retentive and more effective inhibitors of folate-dependent enzymatic targets. Subsequently, examples have been reported from our laboratory (4, 19) and elsewhere (20, 21) of tumor cell variants with acquired resistance to methotrexate and other folate analogues that exhibit lower levels of FPGS activity when compared with parental cells and are cross-resistant to most other classical folate analogues. In one of our own studies with the L1210 leukemia, resistant variants (L1210/EDX) were independently selected (4) during therapy of tumor-bearing mice with edatrexate (EDX). Some of these variants exhibited levels of FPGS activity that were substantially lower than in parental cells. In the current studies, we address the underlying basis for these alterations in FPGS activity in these resistant variants at the level of the FPGS protein elaborated by these cells as well as cognate gene expression. Our results appear to define a novel mechanism underlying these modifications in FPGS activity in these variants that may have broad significance in regard to the regulation of FPGS gene expression and for therapy with classical folate analogues. A preliminary report of some of these findings has already been presented (22) in abstract form.

EXPERIMENTAL PROCEDURES

Methods for the isolation of variants of the L1210 cell used in these studies were described in detail earlier (4). These variants were independently isolated from ascite fluid in the peritoneal cavity of mice following treatment with EDX. Cloning of each of these variants was by limiting dilution in mice (4), and the cloned variants were maintained in mice under therapy with EDX. Each variant was transplanted into mice not receiving therapy prior to processing cells from ascitic fluid for the various experiments. Cells were usually collected in the form of a pellet by centrifugation, washed once with phosphate-buffered saline (0.14 M NaCl, 0.01 M potassium phosphate), centrifuged again, and frozen at 80 °C for storage.

After processing of cells (4 and above), the protein concentration of the resulting cell-free extract was determined (23), and the assay for FPGS activity was carried out as described earlier (4) using EDX as substrate. Product formation (edatrexate plus 1 additional glutamate linked by a peptide bond at the -carboxyl group) under these conditions was linear for at least 2 h at 37 °C, and these cell-free preparations under the assay conditions used exhibited (24) no detectable folylpolyglutamate hydrolase activity. In some experiments, FPGS activity in cell-free extract was determined for cells exposed to 50 µg/ml of cycloheximide for varying periods of time.

A murine FPGS cDNA (ZAP-L1210/R83-1) 77.3% homologous to a human FPGS cDNA was obtained by hybridization screening (19) of an L1210 cell cDNA library in  gt11 (Stratagene, La Jolla, CA) using the human cDNA, pTZ18U (25), as a probe. This  gt11 cDNA construct incorporates a 2.247-kilobase pair insert ligated at the XhoI polylinker site including 3- and 5-untranslated region sequences of 460 and 26 base pairs respectively, and an open reading frame of 1761 base pairs that codes for a putative mitochondrial leader peptide as well as the enzyme protein. Sequencing of the murine cDNA was by the method of Sanger et al. (26).

A method (27, 28) utilizing rapid isolation of poly(A)⁺ RNA directly from the cell lysates by means of an oligo(dT)-cellulose column was used. An aliquot (2-10 µg) of the RNA was analyzed (29) by Northern blotting and radioautography using a murine FPGS cDNA, ZAP L1210/R83-1 (see above), as a probe and normalized to mRNA content with a human -actin probe, PCD--actin (30). Labeling of each probe was by random priming (Random Primers DNA labeling kit; Boehringer Mannheim) using [-³²P]dCTP (3000 Ci/mmol and 10-20 ng of insert).

Samples of a 0-30% ammonium sulfate fraction (4) of cytosol were solubilized in SDS sample buffer, electrophoresed (31) on a 7.5% polyacrylamide slab gel, and transferred to nitrocellulose using a Bio-Rad Trans-Blot cell (30). Western blotting was performed (32, 33) using anti-human FPGS peptide polyclonal antibody (0.5 µg/ml) prepared in rabbits and purified on a peptide-Sepharose column (33, 34). The human FPGS peptide utilized was deduced from the cDNA nucleotide sequence (23) and linked to keyhole limpet hemocyanin (35). The amino acid sequence of this peptide and its use as an antigen for antibody production has already been described (19). During blotting, the anti-rabbit horseradish peroxidase-IgG conjugate (Sigma) at a 1:3000 dilution was used as the secondary antibody. The blots were used to expose hyperfilm after incubation with enhanced chemiluminescence (ECL) reagents (Amersham Corp.) and quantitated by scanning densitometry (Stratagene).

The isolation of poly(A)⁺ RNA and the preparation of FPGS cDNA from the various cell types has already been described above. The murine FPGS cDNA (variant I in Ref. 37) was used as a template to amplify a 1-kilobase pair biotinylated PCR product. Twenty pmol of antisense primer (5-biotin-TCCTGTAACCAGTACATGGATGGC-3) was biotinylated during synthesis by Operon, Inc. The sense primer (5-GGAGAAAATAGCATGGCAGAAACG-3) used was not biotinylated. A 40-cycle PCR reaction in 50 µl was subsequently carried out. The initial denaturation step was 5 min at 95 °C and then 40 cycles of PCR (step 1, 95 °C for 1 min; step 2, 60 °C for 1 min; step 3, 72 °C for 1 min). After completion, the final extension of the reaction was 7 min at 72 °C. The product was electrophoresed on an agarose gel in 1 × TAE buffer (40 mM Tris acetate, 2 mM Na₂EDTA, pH 8.5), excised from the gel, and purified with the aid of BIO-101 GeneClean (BIO-101, Inc.).

DYNABEADS M-280 (DYNAL) streptavidin (100 µl) was used to immobilize the individual PCR products. After washing once in 400 µl of 6 × SSC (1 × SSC: 150 mM NaCl, 15 mM sodium citrate, pH 7.0), the beads were resuspended in 100 µl of the same buffer. The biotinylated PCR product (50 µg) in 6 × SSC was added to the suspension and gently shaken for 5 min. The suspension of beads was placed in a DYNAL MPC to facilitate removal of the supernatant. The beads were resuspended in 400 µl of 0.125 N NaOH plus 0.1 M NaCl and incubated for 5 min at room temperature with occasional shaking. After placement in the MPC, the supernatant was removed. This step was repeated four times. The beads in the MPC were washed five times with 400 µl of 50 mM Tris-HCl, 2 M NaCl, and 1 mM EDTA (pH 7.5), and the pH was obtained to ensure that removal of NaOH was completed. The beads are resuspended in 200 µl of the Tris buffer (see above). Poly(A)⁺ RNA in water (200 µl), which was first denatured at 65 °C for 5 min, was added to the suspension of DYNABEADS and mixed. After 5 min, the beads were washed with 600 µl of Tris buffer (25 mM) Tris-HCL, 1 M NaCl, and 0.5 mM EDTA, pH 7.5) five times. The FPGS mRNA remaining on the beads was removed by boiling the beads in the 100 µl of H₂O for 2 min and rapidly chilling in an ice bath. The beads were again immobilized with the MPC in the ice bath, and the supernatant was removed. This step was repeated once. The sample was treated with RNase-free DNase I (1 unit for 10 min), washed with phenolchloroform, precipitated with sodium acetate (0.3 M, pH 5.2), washed twice with 70% ethanol, and dissolved in H₂O. The integrity of the RNA from several pooled samples was determined by electrophoresis and ethidium bromide staining. After quantification by UV absorbance spectrophotometry, the RNA was used in an in vitro translation reaction. The yield of FPGS mRNA was 30-40 ng/mg of total poly(A)⁺ RNA. The combined product of five to seven individual hybridization selections were pooled, and 200 ng was used in each in vitro translation reaction (see below).

Translation of FPGS mRNA using a rabbit reticulocyte system (Promega) was carried out in accordance with the manufacturer's instructions. Fifty µl of reaction volume contained 35 µl of nuclease-free rabbit reticulocyte lysate, 1 µl of a 1 mM amino acid mixture minus methionine, 4 µl of [³⁵S]methionine (1175 Ci/mmol), 1 µl of RNasin ribonuclease inhibitor (40 units/µl), 4 µl of RNA substrate (200 ng), and 5 µl of H₂O. The reaction was carried out at 30 °C for varying periods of time. After the indicated time, 4 volumes of SDS sample buffer were added followed by heating for 100 °C for 2 min. Electrophoresis on a 7-12% polyacrylamide gel was carried out using standard conditions (31). After drying, the gel was subjected to autoradiography. In addition to the various samples of FPGS mRNA, 1 µg of luciferase mRNA was utilized as a control.

EDX was provided by the CIBA-GEIGY Corporation (Summit, NJ). This material was assayed by high performance liquid chromotography (10) and found to be >99% pure. Folylpolyglutamate hydrolase activity was measured as described previously (24). All other reagents were analytical grade. Radioactive isotopes were purchased from DuPont NEN.

RESULTS AND DISCUSSION

A total of 14 L1210/EDX-resistant variants were isolated (4) during therapy with EDX of tumor-bearing mice. Eleven of these variants when compared with parental L1210 cells exhibited increases in dihydrofolate reductase activity, nine exhibited decreases in transport inward of folate analogues, and seven showed a decrease in FPGS activity. Among these latter seven variants, FPGS activity determined in crude cell-free extracts varied (Table I) from 2- to 23-fold lower than that seen in parental L1210 cells. In addition to this enzymic alteration, four of these variants (L1210/EDX-4, -5, -13, and -14) exhibited (4) elevated levels of dihydrofolate reductase, and two variants (L1210/EDX-7 and -12) showed lower inward transport of folate analogues. The remaining variant (L1210/EDX-6), exhibiting the greatest decrease in FPGS activity, showed no alteration in either one of these other properties. In every case, where adequate kinetic measurements could be made, the decrease in FPGS activity observed in these variants compared with parental L1210 cells using EDX as substrate was associated (Table I) with a commensurate decrease in value for V_max for catalytic activity and no change in value for apparent K_m. Otherwise, FPGS activity from variant and parental L1210 cells exhibited (4) similar requirements and preferences among various folate compounds as substrate. Preferences were in the order of ,L5CHO-folateH₄  aminopterin > EDX > methotrexate.

Table I. Kinetic properties and level of FPGS activity in EDX-resistant, L1210 cell variants Methodological details are provided under "Experimental Procedures." The values shown were derived with edatrexate as the substrate. Cell line Specific activity Km Vmax Vmax/Km nmol/h/mg protein µM nmol/h/mg protein L1210 1.52  ± 0.3 29.8  ± 4 1.74  ± 2 0.058 L1210/EDX-4 0.39  ± 0.07 34.6  ± 5 0.44  ± 0.7 0.013 L1210/EDX-5 0.23  ± 0.04 28.2  ± 5 0.25  ± 0.4 0.009 L1210/EDX-6 0.066  ± 0.02 -a -a -a L1210/EDX-7 0.48  ± 0.06 31.0  ± 5 0.56  ± 0.8 0.019 L1210/EDX-12 0.78  ± 0.13 26.3  ± 4 0.85  ± 0.1 0.032 L1210/EDX-13 0.74  ± 0.08 32.9  ± 5 0.91  ± 0.1 0.028 L1210/EDX-14 0.56  ± 0.08 28.6  ± 4 0.64  ± 0.07 0.022 a Activity was inadequate for a kinetic analysis.

In order to rule out the possible role of endogenous stimulatory or inhibitory factors other than FPGS, itself, in determining the relative level of FPGS activity among variant and parental cell-free extracts, a series of mixing experiments were carried out. In these experiments, equal amounts of crude cell-free extract from parental and variant (L1210/EDX-4 to -7 and L1210/EDX-12 to -14 cells) were added to the reaction mixture, and FPGS activity was compared with that obtained with the same amount of each extract added alone to each reaction mixture. In the case of all of the mixtures prepared with parental and variant-derived cell-free extract, the results (data not given) showed that the level of FPGS activity obtained with these mixtures was very similar to that activity expected from the calculated sum of the activity found in each extract when assayed alone. Thus, the difference in FPGS activity seen with variant and parental cell-free extract appeared to be related to the amount or properties of FPGS itself. In support of this conclusion, it was also found that the same relative level (variant/parental cell) of FPGS activity was obtained (data not shown) in preparations of partially purified FPGS from variant and parental cells following ammonium sulfate fractionation of cell-free extract.

A possible basis for differences seen between FPGS activity derived from variant and parental cells may relate to the stability of the enzyme in each case. Consequently, the relative stability with time of FPGS was determined by the standard assay after incubation of cell-free extract at 50 °C in the reaction buffer without ATP, glutamate, MgCl₂, 2-mercaptoethanol, or folate. The results obtained with the parentally derived cell-free extract showed that loss of FPGS activity during heating occurred with a t1/2 of 3.1 ± 0.4 (S.E.) min. Incubation of each variant derived cell-free extract under the same conditions gave similar results. In this case, values for t1/2 for inactivation varied only from 2.7 ± 0.3 (S.E.) to 3.3 ± 0.3 (S.E.) min.

The data in Fig. 1 show the results of Western blotting with ECL detection of partially purified FPGS from cell-free extract derived from parental and variant L1210 cells following SDS-PAGE. Blotting with rabbit anti-human FPGS peptide antibody (Fig. 1) but not with preimmune serum (data not shown), revealed a 60-61-kDa protein band in material derived from parental cells. In comparison, the companion blot of material derived from the variant cells showed that the intensity of the same 60-61-kDa protein band decreased progressively in the order L1210/EDX-12, -4, and -5. No 60-61-kDa protein band was detected in the same material from L1210/EDX-6 cells. Quantitation of the intensity of this 60-61-kDa band (data not given) among the different samples by densitometry showed that the relative decrease in intensity of this band correlated with the value for V_max for FPGS activity obtained with the original cell-free extract derived from each cell type.

Variant and parental L1210 cells were harvested from mice and resuspended at 1 × 10⁷ cells in RPMI medium in the presence of 50 µg/ml of cycloheximide. Following various periods of incubation at 37 °C, aliquots of cell suspension were removed, and the cells were processed for protein determination and FPGS activity. The results are given in Fig. 2, where the data show that loss of FPGS activity with time occurred with a half-time of 259 ± 30 min. Measurements of FPGS activity in cycloheximide-treated variant (L1210/EDX-4, -5, -7, -12, and -13) cells yielded similar values (t1/2 = 236-268 min) for half-time for the loss of activity with time.

Northern blotting of poly(A)⁺ mRNA from variant and parental L1210 cells with a murine FPGS cDNA probe (19) revealed (Fig. 3) no differences in FPGS mRNA level among these cell types. In each case, a single 2.3-kilobase pair mRNA was detected at the same relative intensity when compared with a control blot of -actin mRNA. These blots were repeated with different amounts of mRNA from each cell type and different hybridization conditions with the same result (data not shown).

Following hybridization selection of FPGS mRNA, the relative ability of this mRNA derived from both variant and parental L1210 cells to mediate translation in an in vitro assay system was determined. In the initial experiments, the time course for accumulation of [³⁵S]methionine-labeled product of the translation reaction was determined by SDS-PAGE for parental cell mRNA. The results given in Fig. 4 show that a product of the translation reaction occurred only when FPGS mRNA or another mRNA, in this case encoding luciferase, was added to the reaction mixture. The results in Fig. 4 also show that the product of the FPGS mRNA-mediated reaction occurred as a doublet (~61 kDa). These products could be blotted (data not shown) by anti-FPGS polyclonal antibody. Data in Fig. 5 show that the rate of product formation obtained with parental L1210 FPGS mRNA, which occurred in the form of a doublet, was constant with time for a period of at least 50 min. By comparison, product formation utilizing FPGS mRNA from L1210/EDX-4 cells also occurred as a doublet but was appreciably less after 50 min. In subsequent experiments comparing the rate of translation mediated by FPGS mRNA from variant and parental L1210 cells, a reaction interval of 50 min was selected to ensure that the comparison of translation efficiency was made during that portion of the time course that was constant with time. The data shown in Fig. 6, A and B, are the results of translation obtained with FPGS mRNA derived by hybridization selection from four individual L1210/EDX variants (L1210/EDX-4, -5, -6, and -12) and parental L1210 cells. This blot made following SDS-PAGE of the translation products shows that the amount of [³⁵S]methionine-labeled product obtained with the mRNA derived from the variants was decreased progressively in the order of L1210/EDX-12, -4, -5, and -6 when compared with that obtained with parental cell FPGS mRNA. In every case the product was represented on the blot as a protein doublet. The radioactivity associated with the protein product derived with each FPGS mRNA, following excision from the SDS-PAGE gel, was also determined by scintillation counting and shown (Fig. 7) to correlate with the relative FPGS activity found in cell-free extract from these variants and parental L1210 cells. However, the results shown do not fully account for the total differential in FPGS activity observed in cell-free extract derived from these different cell types.

The lower levels of FPGS activity and protein that were characteristic of the L1210/EDX variant cell lines, in light of evidence for no alteration in cognate mRNA level, could have had as its basis more rapid turnover of this protein. The results presented here strongly suggest that this was not the case and that the lower levels of FPGS synthesis observed in cell-free extract from these variants were determined at least in part by alterations of their FPGS mRNA. These alterations could possibly affect ribosomal binding to FPGS mRNA, translation initiation or elongation, and/or the interaction of various stimulatory or inhibitory factors (36) with the mRNA. It was of interest to note that the magnitude in the difference between variant and parental cell FPGS mRNA in the ability to mediate product formation in the in vitro translation assay, while appreciable, was less than that expected in view of the differences in FPGS activity and protein among these different cell types. However, this may reflect the fact that the in vitro translation in this assay occurred out of context with respect to the normal environment of the cytosol and that differences seen within the cell were not fully expressed under these conditions. Alternatively, changes involving molecular factors (see above) in addition to the FPGS mRNA, itself, may also play a role in mediating these apparent posttranscriptional modifications of FPGS gene expression in these variants. In addition, it should also be noted that we have not directly excluded the possibility of a separate mutational alteration in some of these variants that affects the catalytic turnover of FPGS. Such an effect could account in part for the lower level of FPGS activity observed in these variants given the limitation on the precision with which FPGS levels could be quantitated by Western blotting. Further work will be required to address these issues and to distinguish between the various possibilities.

It was also of interest to observe from the data given in Figs. 5, 6, 7 that multiple peptide products were generated by in vitro translation of FPGS mRNA from variant and parental cells. In view of their molecular mass relative to each other, these products would appear to represent (25, 37-39) mitochondrial and cytosolic forms of FPGS. If so, the data also suggest that the predominant product of the in vitro translation reaction is the mitochondrial form rather than the cytosolic form of FPGS. This is an unexpected result, since some of our other findings (37) show that the most common splice variant (variant II) of FPGS mRNA found in L1210 cells does not encode a mitochondrial leader peptide. Consequently, it is possible that within the somewhat artificial environment of the in vitro translation assay, their was preferential initiation of FPGS synthesis at the upstream ATG start codon, which is present only in variants I and III, which are less prevalent than variant II. Whether this is actually the case remains to be seen and will require further work. Of greatest interest were our findings showing that formation of both products of the translation reaction were decreased in the case of the variant FPGS mRNA. This would suggest that an alteration of the mRNA modulating translation occurred either in the nucleotide sequence encoded by exon 1b (37) or further downstream.

Including those variants described here, three different classes of antifolate-resistant variants of the L1210 cell have been isolated that exhibit altered FPGS activity. A recent publication (19) from this laboratory reported on the selection of a variant resistant to methotrexate with a decrease in the rate of FPGS mRNA transcript formation. Thus, resistance to classical folate analogues in this murine tumor resulting in lower FPGS activity can occur from both transcriptional and post-transcriptional alterations of FPGS gene expression. In the same report (19), an even more novel group of variants were described that were selected for resistance to the lipophilic inhibitor of dihydrofolate reductase, metoprine. These variants exhibited elevated levels of FPGS activity as a result of the increase in rate of FPGS mRNA transcript formation. The availability of these three classes of variants with altered levels of FPGS synthesis should greatly facilitate studies of the regulation of expression of the corresponding gene at both transcriptional and posttranscriptional levels and perhaps the role of sequence-dependent topology of mRNA and other cytosolic factors in determining translation efficiency.