Evidence for direct inhibition of de novo purine synthesis in human MCF-7 breast cells as a principal mode of metabolic inhibition by methotrexate.

We have investigated the role of dihydrofolate (H2PteGlu) accumulation in the inhibition of de novo purine synthesis by methotrexate (MTX) in human MCF-7 breast cancer cells. Previous studies have shown that cytotoxic concentrations of MTX that inhibit dihydrofolate reductase produce only minimal depletion of the reduced folate cofactor, 10-formyltetrahydrofolate, required for purine synthesis. At the same time, de novo purine synthesis is totally inhibited. In these studies, we show that 10 microM MTX causes inhibition of purine synthesis at the step of phosphoribosylaminoimidazolecarboxamide (AICAR) transformylase, as reflected in a 2-3-fold expansion of the intracellular AICAR pool. The inhibition of purine synthesis coincides with the rapid intracellular accumulation of H2PteGlu, a known inhibitor of AICAR transformylase. When the generation of H2PteGlu is blocked by pretreatment with 50 microM 5-fluorodeoxyuridine (FdUrd), an inhibitor of thymidylate synthase, MTX no longer causes inhibition of purine synthesis. Intermediate levels of H2PteGlu produced in the presence of lower (0.1-10 microM) concentrations of FdUrd led to proportional inhibition of purine biosynthesis, and the exogenous addition of H2PteGlu to breast cells in culture re-established the block in purine synthesis in the presence of FdUrd and MTX. The early phases of inhibition of purine biosynthesis could be ascribed only to H2PteGlu accumulation. MTX polyglutamates, also known to inhibit AICAR transformylase, were present in breast cells only after 6 h of incubation with the parent compounds and were not formed in cells preincubated with FdUrd. The lipid-soluble antifolate trimetrexate, which does not form polyglutamates, produced modest 10-formyltetrahydrofolate depletion, but caused marked H2PteGlu accumulation and a parallel inhibition of purine biosynthesis. This evidence leads to the conclusion that MTX and the lipid-soluble analog trimetrexate cause inhibition of purine biosynthesis through the accumulation of H2PteGlu behind the blocked dihydrofolate reductase reaction.

It produces cytotoxic effects through the inhibition of the de nouo synthesis of purines, thymidylate, and certain amino acids (1). Many investigations have demonstrated that inhibition of purine synthesis is an important component in MTX-induced cytotoxicity. Whereas either purine (hypoxanthine) or thymidine can partially rescue murine sarcoma 180 cells, both are required to effect complete rescue from the toxic effects of MTX (2, 3), a conclusion corroborated with work in in vitro and in uiuo systems using normal and malignant murine and human cell lines (4)(5)(6)(7)(8)(9)(10). Finally, more recent studies have demonstrated the ability of inosine alone to obviate completely the toxicity of mice treated with high-dose MTX (10 Fg/h for 72 h), whereas thymidine alone was ineffective (11). Each of these studies suggests an important role for the inhibition of de noua purine synthesis in the cytotoxic action of MTX.
The previous concept of MTX action has been that the drug inhibits the metabolic pathways by depleting reduced folates as a consequence of its primary action, the inhibition of dihydrofolate reductase (EC 1.5.1.3, 5,6,7,8-tetrahydrofolate:NADP+ oxidoreductase). In the face of dihydrofolate reductase inhibition, the intracellular pools of reduced folate are converted to H,PteGlu by the thymidylate synthase reaction and cannot be reduced to the H,PteGlu state required for cofactor activity. This hypothesis provides a mechanism whereby reduced folates would ultimately become trapped as H2PteGlu and metabolic reactions requiring reduced folate cosubstrates would cease. This indirect mechanism of metabolic inhibition fails to account for the competitive nature of Leucovorin rescue (1, 4, 12,13) and is inconsistent with evidence that certain reduced folate cosubstrate pools are relatively preserved during exposure to MTX (14)(15)(16). In particular, we found that 5-methyl-H4PteGlu and H,PteGlu were the only reduced folates depleted by greater than 50% in the presence of MTX; 10-formyl-H4PteGlu, the required folate cosubstrate for de nouo purine synthesis, was preserved at 80% of control levels for up to 21 h during exposure of human MCF-7 breast cells to 1 FM MTX, a concentration that markedly curtailed de nouo purine synthetic activity. In addition, we found that the rapid accumulation of HpPteGlu in the MCF-7 breast cells following MTX exposure correlated temporally with the inhibition of de novo purine synthetic activity and that the polyglutamates of this oxidized folate as well as MTX polyglutamates potently inhibit AICAR transformylase (17). These studies suggest that inhibition of the de novo purine pathway may occur through direct enzyme inhibition by the polyglutamates of H,PteGlu and/or MTX rather than by depletion of 10-formyl-H,PteGlu. This study provides further evidence for a direct inhibition of the folaterequiring purine synthetic enzymes in that it demonstrates a close temporal and quantitative relationship between the ac-Inhibition of de Novo Purine Synthesis by Methotrexate 13521 cumulation of H,PteGlu and the inhibition of de novo purine synthesis under a variety of conditions. Furthermore, this report lends additional support for the concept of direct inhibition of AICAR transformylase by dihydrofolate polyglutamates as the principal mechanism of de novo purine inhibition by MTX.

Methods
Cell Culture-A late-passage, mycoplasma-free human MCF-7 breast cancer cell line was used for all experiments requiring intact cells. The cells were grown in T-75 tissue culture flasks (Falcon Labware, Oxnard, CA) in 15 ml of RPMI 1640 (Biofluids, Rockville, MD) medium with 2 mM glutamine and 10% heat-inactivated (57 "C for 30 min), dialyzed fetal bovine serum. Fetal bovine serum was dialyzed against 0.9% NaCl in a 1:40 ratio for five 24-h exchanges. The cells were grown for 72 h after plating (50-70% confluency) prior to use in any experiments. Furthermore, all cells were grown in dialyzed fetal bovine serum for at least two passages prior to their use in experiments.
De Novo Purine Synthetic Activity-De novo purine activity was measured according to published methods (18). MCF-7 breast cells (2 X lo6 cells) were plated onto T-75 tissue culture flasks and allowed to grow for 72 h prior to incubation in various concentrations of drugs, including MTX, FdUrd, trimetrexate, and thymidine, for varthe cells were incubated with 10 pCi of ["Clglycine (final specific ious timed exposures. One h prior to termination of each experiment, activity, 5.17 mCi/mmol) in order to label the purine nucleotides synthesized by the de novo pathway. After labeling, the cells were washed three times with ice-cold phosphate-buffered saline (PBS) (Biofluids) and then harvested from the plates with 8 ml of 0.04% EDTA in PBS. The cells were pelleted by centrifugation at 200 X g for 15 min, and the supernatant was decanted. Perchloric acid (0.5 ml) was added to the cells, which were then vortexed and heated in a boiling water bath for 60 min, followed by sedimentation at 2000 X g for 10 min. The pelleted protein was used as a measure of total cellular protein; and the supernatant, containing acid-soluble purine bases, was neutralized to pH 7.0 with the addition of 1 N KOH. After removal of the resulting precipitate by centrifugation at 2000 X g for 10 min, the supernatant was filtered through a 2.2-pm filter (Acro-15, Gelman Sciences, Inc., Ann Arbor, MI) and loaded onto an HPLC column for separation and quantitation of 14C-labeled adenine and guanine as previously described.
Radioisotope dilutional effects were minimized by relatively long incubation times (1 h) and a high mass of external glycine (final concentration in media, 0.13 mM). To ensure that the observed changes in de novo purine activity were not the result of changes in the intracellular specific activity of [14C]glycine, intracellular glycine pools were measured using an automated amino acid analyzer (Beckman Instruments) in control cells and cells exposed to MTX (1 p~) and thymidine (10 p M ) for 2, 5, and 24 h. The specific radioactivity in MTX-treated cells was found to be indistinguishable from control cells under these experimental conditions. The glycine content in Glu, refers to the total number of glutamyl residues in the compound, e.g. MTX-Glu2 = 4-amino-10-methy1pteroy1g1utamy1-yglutamic acid.
IntraceUular Folate Pool Measurements-Intracellular folate pools were measured according to published methods (14). Briefly, MCF-7 cells (2 X lo6 cells) were plated onto T-75 tissue culture plates with 2.2 p~ [3H]folic acid (final specific activity, 0.9 Ci/mmol); and after 72 h of growth, the cells were exposed to various drugs for specified amounts of time, followed by three washes with ice-cold PBS. The cells were then harvested in 1 ml of PBS with a rubber policeman. A 100-pl aliquot of the cellular suspension was used to measure total protein, and the intracellular folates were extracted from the remainder of the suspension. The extracted folates were treated with partially purified hog kidney conjugase, concentrated using a Sep-Pak C18 cartridge, and loaded onto an HPLC column for separation and quantitation of the 3H-labeled intracellular folate pools. Intracellular MTX Polygbtamates-MTX polyglutamates were measured by plating MCF-7 cells onto T-75 tissue culture plates, and after 72 h of growth, the cells were exposed to various combinations of drugs and [3H]MTX (specific activity, 21 Ci/mmol). The cells were then washed three times with PBS, followed by harvesting in 5 ml of ice-cold PBS with a rubber policeman. The cells were sedimented at 200 x g for 10 min, and the supernatant was decanted. One ml of 10% trichloroacetic acid was added to the cell pellet, and the precipitated protein was used as a measure of total cell protein. The MTX polyglutamates found in the supernatant were concentrated using a Sep-Pak Cla cartridge and loaded onto an HPLC column for separation and quantitation according to published methods (19).
Catalytic Activity of AICAR Transformylose-The catalytic activity of AICAR transformylase was measured using a ~pectrophotometric assay (20). A 1-ml reaction cuvette contained 0.02 unit of 300-fold purified human AICAR transformylase (1 unit = 1 pmol of H4PteGlu formed per min at 37 "C), the folate cosubstrate, 1-10-formyl-H4PteGlu, and various concentrations of inhibitor in 25 mM KC1, 50 mM 2-mercaptoethanol, and 50 mM Tris/HCl, pH 7.4. After a 10-min equilibration period at 37 "C, the reaction was initiated with the addition of 50 nmol of AICAR. The reaction velocity was measured as the change in optical density at 298 nm using an extinction coefficient for the reaction of 19,700 cm" m-'. Intracellular AICAR Pools Measured by [14ClGlycine Labeling-Intracellular nucleotide pools were measured by plating MCF-7 cells as outlined above with the addition of ["C]glycine (final specific activity, 5.17 mCi/mmol). After 72 h of growth, the cells were treated with 10 p~ MTX or trimetrexate for various timed intervals and washed three times with ice-cold PBS, and the nucleotides were extracted with 1 ml of 10% perchloric acid. The nucleotides were then neutralized with 1 N KOH and separated by an HPLC system based on that described by Brown and Parks (21). We used a Waters Associates HPLC model 6000-A solvent delivery system, Model 440 absorbance detector, and Model 720 system controller and data module and a Whatman Partisil-10 SAX anion-exchange column. The low-concentration eluent was 0.007 M KH2P04 in 0.007 M KCl, pH 3.5, and the high-concentration eluent was 0.25 M KH2P04 in 0.5 M KCl, pH 4.5. All eluent solutions were filtered through Millipore membrane filters (HA, Millipore, Bedford, MA) and degassed prior to use. An automatic injector, WISP Model 710A, applied the sample (150 pl) to the column; and a Model 660 solvent programer monitored a No. 8 convex gradient with a flow rate of 1 ml/min. The gradient time was 30 min from 100% low-concentration eluent to 100% highconcentration eluent. After completion of the gradient, the column was eluted for an additional 40 min with 100% high-concentration eluent, and peaks were identified by the use of nonradioactive standards. To quantitate the incorporation of [14C]glycine into AICAR, we collected the effluent in 1.0-min (1 ml) fractions and determined radioactivity in each fraction by liquid scintillation spectrometry. The retention time for AICAR was found to be 40 min.
Direct Measurement of Intracellular AICAR-For these experiments, MCF-7 cells, 3 days after replating, were treated with 10 p~ MTX or trimetrexate for various times prior to harvesting in 1 ml of ice-cold PBS. The cells were disrupted by sonication with three 2-s bursts from a Branson sonifier 350 equipped with a microtip. Following a 10-min centrifugation at 2000 X g, a 100-pl aliquot was reserved for protein analysis, and the remaining cell supernatant fluid was used for nucleotide analysis. The extracted nucleotides were then concentrated using a Sep-Pak CIS cartridge as follows. The Sep-Pak cartridge was washed with 10 ml of methanol, followed by 10 ml of water, and then 2 ml of 5 mM tetrabutylammonium phosphate, pH 5.5, followed by an additional 5 ml of water. The sample was then loaded onto the cartridge and washed with 5 ml of water. Three ml of methanol was then used to elute the nucleotides from the Sep-Pak cartridge, and the eluate was dried under a steady stream of nitrogen and stored at -70 "C until analyzed. The recovery rate for the extraction procedure was found to be 80% for AICAR using an unlabeled standard mixed with MCF-7 cytosol. For analysis, the nucleotides were suspended in 100 pl of 5 mM tetrabutylammonium phosphate filtered through a 2.2-pm filter and loaded onto a reversephase HPLC column for separation and quantitation by comparison to standard amounts of AICAR. A reverse-phase HPLC system using a pBondapak Cs column was used for the separation. The nucleotides were eluted from the HPLC column under isocratic conditions with 80% 2.5 mM tetrabutylammonium phosphate and 20% methanol. The pH of tetrabutylammonium phosphate was chosen to adjust the retention time of AICAR to a point where no interfering peaks were found. A pH range of 3.6-3.75 was found to be optimal, and this resulted in retention times of 28-35 min. An evaluation of AICAR retention time uersw pH between 3.0 and 3.75 revealed a linear relationship with a slope of 1.9 min/O.l pH unit. The peak height of the AICAR in the sample measured at 260 nm was quantitated by comparison to AICAR standards and standardized to the amount of cytosolic protein in each sample. Authentication of the putative AICAR peak was accomplished by three criteria: 1) coelution with standard AICAR, 2) UV ratio (254280; 254269; 269280) that matched standard AICAR, and 3) complete utilization of the putative AICAR peak when it was incubated with 0.04 unit of purified human AICAR transformylase in 25 mM KCl, 50 mM Tris/HCl, pH 7.4, 50 mM 2-mercaptoethanol, and 0.7 pmol of 1-10-formyl-H4PteGlu.
Protein Measurement-Cytosolic protein was measured according to the method of Bradford (22) using bovine serum albumin as the standard.

RESULTS
Since de nouo purine synthesis as measured by [14C]glycine incorporation into nucleotide pools is profoundly and rapidly inhibited following brief exposures to MTX (Fig. 4), we sought to define the enzyme step in this pathway affected by MTX. Measurement of the substrate for the second of two folaterequiring enzymes in the pathway provides an indication of which of the two enzymes was inhibited by the antifolate. Intracellular levels of AICAR were measured by two different techniques after exposure of the MCF-7 cells to 10 p~ MTX or trimetrexate for various timed intervals: 1) direct measurement of the AICAR pool using UV detection of the nucleotides separated by HPLC, and 2) [14C]glycine pool labeling, followed by nucleotide separation on HPLC and quantitation by scintillation counting of the peak identified as AICAR. For all experiments, control (unexposed) cells and experimental cells were processed simultaneously. As illustrated in Fig. 1  ( A and B ) , both techniques demonstrate a 2.3-fold increase in the intracellular level of AICAR when the MCF-7 cells were exposed to either MTX or trimetrexate. Using the direct UV measurement of intracellular AICAR, the mean amount of this nucleotide found in the control MCF-7 cells was 6.1 & 1.2 pmol/mg of cytosolic protein (n = 5). These studies suggest that inhibition of the de nouo purine pathway occurs at the level of AICAR transformylase.
Whereas MTX is capable of forming polyglutamates that are potent inhibitors of AICAR transformylase, trimetrexate cannot form polyglutamates; we next determined whether trimetrexate may also be an inhibitor of this enzyme. Using the assay system described under the "Methods," we found that trimetrexate had no effect on the velocity of the reaction catalyzed by purified human AICAR transformylase at concentrations of drug up to M (data not shown). Thus, the inhibition of AICAR transformylase could not be ascribed to direct effects of the antifolates or their polyglutamates on the enzyme.
Additional evidence against direct inhibition of AICAR transformylase by MTX polyglutamates came from the following experiments. 1) For the rate of formation of MTX polyglutamates in MCF-7 cells, we exposed MCF-7 cells to 1 TIME (hrj TIME (hr) FIG. 1. Effect of methotrexate and trimetrexate on the intracellular AICAR pool. The AICAR pool size in antifolate-treated MCF-7 breast cells was measured by two independent methods. A represents the intracellular AICAR levels in breast cells treated with 10 p~ MTX for 8 and 24 h and in untreated control cells. These AICAR levels are quantitated by comparing the UV peak height of the samples to known standards after extraction and separation by HPLC as described under "Methods." The error bars represent the standard error of the mean of six independent experiments. B represents the AICAR level in MCF-7 cells exposed to 10 p~ MTX (stippled bars) or trimetrexate (solid bars) for 5 and 24 h. For these experiments, the purine nucleotide pools in all cells, including controls, were labeled for an identical period of time (48 h) with ["C] glycine (final specific activity, 5.17 mCi/mmol), during which the cells were exposed to the antifolates for the indicated times. The nucleotides from control and experimental cells were extracted simultaneously for quantitation by HPLC as described. Each bar represents the mean of at least two independent experiments. pM MTX for various times, monitored the activity of the de m v o purine pathway using [14C]glycine incorporation into the products of the pathway (adenine/guanine) (Fig. 4), and correlated these findings with the rate of formation of the higher polyglutamates of MTX (Glu3 + GluJ. As depicted in Fig. 4, the activity of the purine pathway is only 27% of untreated control cells after 2 h of exposure and 10% after 5 h. Examination of the rate of higher MTX polyglutamate formation revealed no higher MTX polyglutamates after 2 h of exposure and only 30 f 5.7% of the intracellular MTX in the form of higher polyglutamates after 6 h. After 24 h of drug exposure, 68 f 4.5% of the intracellular MTX is in the higher polyglutamate form. 2) To determine the effect of trimetrexate on de m u 0 purine synthetic activity, since trimetrexate does not undergo polyglutamation and does not inhibit AICAR transformylase directly, we chose to study its effects on purine pathway activity and related these effects to changes in the intracellular folate cofactor pools. Fig. 2 (inset) indicates that purine synthesis is rapidly inhibited by exposure to trimetrexate. Control cells had a base-line rate of de m u 0 purine synthesis of adenine and guanine nucleotides of 3.7 & 0.6 nmol/h/mg ( n = 6). Similar to our previous studies of the folate cofactor pools after MTX exposure, purine pathway inhibition correlates temporally with rapid accumulation of H,PteGlu, whereas there is relative preservation of the 10formyl-H4PteGlu pool, the required cosubstrate for the folaterequiring enzymes of the de nouo purine pathway (Fig. 2). During trimetrexate exposure, other changes in the intracellular folate pools closely mimicked the changes produced by MTX: 5-methyl-H4PteGlu was rapidly depleted, and 10-formyl-H,PteGlu accumulated. The latter compound is presum- ably formed by an enzymatic formylation of H,PteGlu (14,23). In addition, both control and experimental cells had constant levels of folate (14.3 k 2%). Control cells also contained KPteGlu (7.8 k 2%) and 5-formyl-H4PteGlu (5.4 k 1.5%), but these folates were not detectable in cells exposed to trimetrexate. These early effects of trimetrexate on purine synthesis and intracellular folates closely parallel those produced by MTX (14).
Because of the temporal coincidence of intracellular dihydrofolate accumulation and purine pathway inhibition and the lack of correlation of purine inhibition with MTX polyglutamate formation or cosubstrate depletion, we investigated the possibility that the intracellular accumulation of HBPteGlu might be responsible for direct inhibition of the purine pathway. Since thymidylate synthase is the only known biochemical reaction that generates H,PteGlu, we chose to modulate the intracellular level of dihydrofolate by treating MCF-7 cells with variable concentrations of the halogenated pyrimidine FdUrd prior to constant exposures to 1 p~ MTX. 5-Fluorodeoxyuridine monophosphate, a metabolite of FdUrd, is a potent and specific inhibitor of thymidylate synthase; and, as such, the amount of 5,10-methylene-H,PteGlu (the folate cosubstrate for thymidylate synthase) oxidized to HpPteGlu is inversely correlated with the concentration of FdUrd used for pretreatment. To ensure that effects on the de mvo purine pathway were not the result of pyrimi-dine deprivation or FdUrd-related toxicity, all cells treated with FdUrd were simultaneously incubated with 10 pM thymidine, which effectively reversed the toxic effects of thymidylate synthase inhibition. Variable inhibition of thymidylate synthase activity as a function of FdUrd concentration can be demonstrated by the measurement of intracellular folate pools after drug treatment. Fig. 3 illustrates the actual measurement of the intracellular folates 5-methyl-H4PteGlu (a prevalent reduced folate that is rapidly depleted upon antifolate exposure) and HzPteGlu with various permutations of FdUrd/dThd pretreatment and MTX exposure. Control cells had high levels of 5-methyl-H4PteGlu and undetectable levels of H,PteGlu (first bar), and these findings are undisturbed by a 24-h treatment with 50 PM FdUrd, 10 p~ dThd (second bar). Exposure to 1 p~ MTX in the absence of a 4-h pretreatment with FdUrd/dThd leads to a rapid decrease in the 5methyl-H4PteGlu pool with a parallel accumulation of HzPteGlu (third bar). Four-h pretreatment with 10 p~ dThd and 1, 10, and 50 p~ FdUrd results in a progressive preservation of the reduced folate pools such that at the highest FdUrd concentration (50 p~) , the pools are unaffected by MTX (fourth to sixth bars). Fig. 4 depicts the activity of the de novo purine synthetic pathway after various durations of incubation with 1 p~ MTX following an initial 4-h preincubation with 10 p~ thymidine and various concentrations of FdUrd. For each time point, a parallel experiment was performed exposing the cells to the identical concentration of thymidine and FdUrd, and this was used as the control for each MTX time point. When cells were pretreated with 50 p~ FdUrd, a condition that prevented accumulation of H,PteGlu (Fig. 3, sixth bar)  followed by timed exposures to 1 PM MTX (0, 2 h; 0, 5 h; A, 24 h). In each case, the amount of H2PteGlu (as a percent of the total folate pool) accumulation in the cells was quantitated and plotted as a function of de nouo purine synthetic activity. The mean total intracellular folate concentration was found to be 20.4 k 3.1 pmol/mg. The inset represents a linearization of the data plotting the log percent control of purine activity uersus intracellular H,PteGlu. The error burs represent the standard error of three to six independent determinations. The standard error for the H,PteGlu determinations was less than 10%.
FdUrd exposures (0.1, 1, and 10 PM) followed by 24 h of MTX showed a mean pool preservation of 78 f 7%. There was no correlation between the preservation of the 10-formyl-H,PteGlu pool and exposure to FdUrd.
Exposure to 1 PM MTX in the absence of FdUrd resulted in an 80 k 13% 10formyl-H,PteGlu hydrofolate pool.
For each of the experimental points shown in Fig. 4, the amount of intracellular H,PteGlu accumulating under each of the conditions was measured and correlated with the de nouo purine pathway activity as measured by [14C]glycine incorporation. The correlation is shown in Fig. 5 . Fig. 5 (inset) is a replot of the data using the log of the percent of purine activity as a function of H,PteGlu concentration, a plot that yields a log linear relationship with a correlation coefficient of 0.96.
The relation between H,PteGlu and de nouo purine activity shown in Fig. 4 was derived from experiments in which cells were exposed to 1 PM MTX for various times up to 21 h and various FdUrd concentrations. Whereas brief exposures to MTX (2-3 h) result in negligible polyglutamate formation, it would be expected that prolonged exposures would result in a significant generation of MTX polyglutamates. These polyglutamates may have additional inhibitory effects on purine synthesis beyond those expected for H,PteGlu alone; however, Fig. 4 reveals no conclusive evidence for such an effect. However, there was a constant tendency for increased inhibition of purine synthesis activity at a given intracellular HsPteGlu level for 24-h MTX exposures uersm the 2and 5h exposure points (Fig. 4). To investigate further this question, we measured the formation of MTX polyglutamates in cells exposed to 1 pM MTX for 21 h following 4-h treatments with various permutations of thymidine and/or FdUrd. These results are illustrated in Table I and reveal that negligible quantities of MTX polyglutamates are formed under the pretreatment conditions used for these studies. Thymidine (10 FM) alone was sufficient to diminish the conversion of MTX to its higher polyglutamate forms by 40% compared to control experiments; and the addition of FdUrd, even at the lowest concentrations used in these studies (0.1 FM), resulted in a 94% reduction in MTX polyglutamate formation. As additional support for the concept of H,PteGlu acting as a direct inhibitor of purine synthesis, we performed the following experiment. MCF-7 cells were preincubated for 4 h with 10 PM thymidine and 50 FM FdUrd, followed by 1 PM MTX. As previously shown (Fig. 4)

DISCUSSION
These studies indicate that de nouo purine synthesis in MCF-7 breast cells is rapidly inhibited by cell exposure to either methotrexate or the lipid-soluble antifolate trimetrexate and that inhibition occurs at the level of the folaterequiring enzyme AICAR transformylase. Moreover, inhibition of this enzyme cannot be accounted for by reduced folate The MTX polyglutamate profiles for MCF-7 cells treated with 1 p~ for 24 h with either no pretreatment or various doses of FdUrd and dThd pretreatment are tabulated. The MTX polyglutamates were quantitated by exposing log phase MCF-7 cells to [3H]MTX (final specific activity, 1.3 Ci/mmol) for 24 h. The radiolabeled MTX polyglutamates were extracted from the cells and separated by HPLC according to published methods (19) for final quantitation using a liquid scintillation counter. depletion but is best explained by direct enzyme inhibition by H,PteGlu polyglutamates that accumulate intracellularly as the result of inhibition of dihydrofolate reductase by the antifolates.
Several investigators have reported that inhibition of de nouo purine synthesis results from the inhibition of AICAR transformylase by demonstrating an accumulation of AICAR nucleosides in the medium of bacteria treated with MTX or aminopterin and in the urine of rats treated with MTX (24)(25)(26)(27). Our findings are consistent with these studies in that the human breast cell line also demonstrates an elevation of AICAR levels following antifolate exposure.
Previous studies from our laboratory (14) have shown that the reduced folate cosubstrates required for purine synthesis are only partially depleted (20-30%) in cells treated with MTX, and this study shows a similar preservation of the 10formyl-H4PteGlu pool in cells treated with trimetrexate. The other folate pool changes following trimetrexate exposure also mimic those seen with MTX-exposed cells. Others have reported a similar preservation of reduced folate cosubstrates following MTX treatment. Sur et al. (15) found that the 5,lOmethylene-H4PteGlu pool in Krebs ascites tumor cells was preserved at up to 50% of control levels for 24 h following a 10 p~ MTX exposure. Baram et al. (16) have also demonstrated preservation of the 10-formyl-H4PteGlu pool in normal myeloid precursor cells exposed to MTX over a concentration range of lo-' to M. Additional evidence against folate depletion as a mechanism of de novo purine inhibition comes from the fact that this mechanism would be expected to result in pathway inhibition at glycinamide ribonucleotide transformylase since this is the first of two enzymes in the pathway that utilize the same folate (10-formyl-H,PteGlu) cosubstrate. Pentaglutamated 10-formyl-H4PteGlu has a similar affinity ( K , ) for glycinamide ribonucleotide and AICAR transformylase (4.9 and 5.5 p M ) (17,28), and AICAR transformylase has a 1.5-fold higher specific activity in human MCF-7 cytosolic preparation^.^ The accumulation of AICAR following MTX exposure indicates that the purine pathway is blocked at the level of AICAR transformylase. C. J. Allegra and J. Baram, unpublished observation. MTX pools and H,PteGlu polyglutamates potently inhibit AICAR transformylase in cell-free systems (17). Inhibition constants are strongly dependent on the glutamylated state of the folate cosubstrate 10-formyl-H4PteGlu; but even in the presence of 10-formyl-H4PteGlu,, inhibitory intracellular levels of MTX polyglutamates and H2PteGlu polyglutamates are readily attained in the presence of modest concentrations of MTX (1-2 p~) (14,16,19,29,30). Other investigators have also demonstrated that MTX and H,PteGlu polyglutamates are potent inhibitors of other folate-requiring enzymes, including thymidylate synthase (31)(32)(33)(34)(35) and methylene tetrahydrofolate reductase (36,37). Several additional lines of evidence implicate H,PteGlu polyglutamates as the entity responsible for inhibition of the purine pathway. Trimetrexate, which has no direct inhibitory effects on AICAR transformylase and is incapable of forming polyglutamates, nonetheless caused an increase in H2PteGlu and inhibition of purine synthesis. Furthermore, purine synthesis inhibition occurs within 2 h of exposure to MTX, at a time when MTX polyglutamates are not detectable in MCF-7 cells (19,29). Intracellular folate pool measurements in cells treated with either antifolate reveal a rapid accumulation of H,PteGlu that coincides with inhibition of purine synthesis. These findings all strongly suggest that the accumulation of HzPteGlu polyglutamates may be the factor responsible for inhibition of de nouo purine synthesis during antifolate exposures.
To test this hypothesis, we designed a series of experiments that would allow modulation of intracellular H,PteGlu levels by FdUrd, an inhibitor of thymidylate synthase. Thus, a correlation could be made between the intracellular accumulation of H,PteGlu and inhibition of de novo purine synthesis. FdUrd (50 p~) prior to MTX completely inhibited thymidylate synthase activity and thereby prevented any HzPteGlu accumulation. Under these conditions, purine synthesis was unaffected by MTX. Lower concentrations of FdUrd pretreatment allowed greater accumulations of H,PteGlu and proportional decreases in purine synthesis. As alluded to under "Results," cells exposed to MTX for 24 h appear to consistently demonstrate a greater inhibition of purine synthesis at a given H,PteGlu concentration when compared to cells exposed for 2 or 5 h (Fig. 5). As there is an insignificant formation of MTX polyglutamates under the experimental conditions, it is conceivable that an additional purine inhibitor is being generated during the prolonged exposures. As we have shown in cells exposed to either trimetrexate or MTX (14,16), a new folate, formyl-HpPteGlu, can be identified. Preliminary observations suggest that this folate may be an inhibitor of human glycinamide ribonucleotide transformylase (28). The addition of exogenous H,PteGlu to cells blocked by pretreatment with 50 p~ FdUrd and 1 p~ MTX (a condition that had no effect on purine synthetic rate) led to inhibition of purine synthesis. The amounts required for this inhibition could not be directly compared to inhibition constants derived from cell-free experiments as the glutamylated state and the extent of uptake of the added H,PteGlu were not defined.
Since inhibition of de nouo pyrimidine synthesis is also an important component of the cytotoxicity of antifolates, the observations need to be followed by a parallel set of experiments designed to define the relative contributions of folate depletion uersu-s direct enzyme inhibition of thymidylate synthase, the critical de nouo pyrimidine synthetic enzyme. This enzyme requires 5,10-methy1ene-H4PteGlu, and measurement of this pool during antifolate exposure would be essential.
The concept that H,PteGlu may act as a controller of metabolic pathways is not novel. Matthews and Baugh (36) and Matthews and Haywood (37) found that the polyglutamates of HZPteGlu were potent inhibitors of methylene tetrahydrofolate reductase, one of several cytosolic enzymes utilizing 5,10-methylene-H4PteGlu. These investigators postulated that intracellular H,PteGlu pools would expand in response to increased thymidylate synthase activity and, by inhibiting methylenetetrahydrofolate reductase, act as a physiologic mechanism for preserving 5,10-methylene-H4PteGlu for additional thymidylate synthesis.
Whereas the reported experiments strongly suggest a role for HzPteGlu as the mediator of the inhibitory effects of antifolates on purine synthesis, they do not exclude the potential role of MTX polyglutamates in these processes. Clearly, the early inhibition of purine synthesis is best explained by the accumulation of HzPteGlu, but the experiments that examined late effects on this pathway employed conditions that inhibited MTX polyglutamate formation. Pretreatment of MCF-7 cells with FdUrd/thymidine or FdUrd alone markedly curtailed the polyglutamation of MTX. This finding is in contrast to that reported by McGuire et al. (38), who found that pretreatment of a human leukemia cell line with concentrations of FdUrd up to 50 PM had no effect on the ability of these cells to polyglutamate subsequently administered MTX. Our results would support the attenuation of MTX polyglutamation by FdUrd pretreatment as a potential explanation for the antagonism reported by others for the sequential use of FdUrd, followed by MTX. Based on clinical observations, one possible explanation for the differences observed in the two cell lines may be a relative insensitivity of the leukemic cells compared to the breast cells to the inhibitory effects of FdUrd. The MTX polyglutamates are capable of inhibition of dihydrofolate reductase, and, by virtue of their long intracellular half-life (39), are critical in maintaining the intracellular HpPteGlu pools. Sensitivity to MTX has been correlated with cellular capacity for polyglutamation by a number of investigators (19,(40)(41)(42)(43). In addition to sustaining inhibition of dihydrofolate reductase, MTX polyglutamates are likely contributors to the direct inhibition of enzymes other than dihydrofolate reductase given their potency of interaction with metabolically important folate-requiring enzymes such as AICAR transformylase (17) and thymidylate synthase (31)(32)(33). Inhibition of these distal enzyme sites by MTX and H,PteGlu polyglutamates and the greater formation of both in malignant cells may explain the selectivity of antifolate action and Leucovorin rescue. Furthermore, the suggestion that the antifolates are exerting their effects through direct inhibition of enzymes distal to dihydrofolate reductase would support the development of inhibitors directed at critical folate-requiring enzymes other than dihydrofolate reductase. Inhibitors such as the glycinamide ribonucleotide transformylase inhibitor 5,10-dideazatetrahydrofolate (44,45) and the thymidylate synthase inhibitor 10propargyl-5,8-dideazafolate (CB3717) (46,47) are currently under active investigation as antineoplastic agents.