Regulation of de Novo Purine Biosynthesis by Methenyltetrahydrofolate Synthetase in Neuroblastoma*

5-Formyltetrahydrofolate (5-formylTHF) is the only folate derivative that does not serve as a cofactor in folate-dependent one-carbon metabolism. Two metabolic roles have been ascribed to this folate derivative. It has been proposed to 1) serve as a storage form of folate because it is chemically stable and accumulates in seeds and spores and 2) regulate folate-dependent one-carbon metabolism by inhibiting folate-dependent enzymes, specifically targeting folate-dependent de novo purine biosynthesis. Methenyltetrahydrofolate synthetase (MTHFS) is the only enzyme that metabolizes 5-formylTHF and catalyzes its ATP-dependent conversion to 5,10-methenylTHF. This reaction determines intracellular 5-formylTHF concentrations and converts 5-formylTHF into an enzyme cofactor. The regulation and metabolic role of MTHFS in one-carbon metabolism was investigated in vitro and in human neuroblastoma cells. Steady-state kinetic studies revealed that 10-formylTHF, which exists in chemical equilibrium with 5,10-methenylTHF, acts as a tight binding inhibitor of mouse MTHFS. [6R]-10-formylTHF inhibited MTHFS with a Ki of 150 nm, and [6R,S]-10-formylTHF triglutamate inhibited MTHFS with a Ki of 30 nm. MTHFS is the first identified 10-formylTHF tight-binding protein. Isotope tracer studies in neuroblastoma demonstrate that MTHFS enhances de novo purine biosynthesis, indicating that MTHFS-bound 10-formylTHF facilitates de novo purine biosynthesis. Feedback metabolic regulation of MTHFS by 10-formylTHF indicates that 5-formylTHF can only accumulate in the presence of 10-formylTHF, providing the first evidence that 5-formylTHF is a storage form of excess formylated folates in mammalian cells. The sequestration of 10-formylTHF by MTHFS may explain why de novo purine biosynthesis is protected from common disruptions in the folate-dependent one-carbon network.

in genes that encode folate-dependent enzymes, from low intracellular folate concentrations, and/or from other environmental factors (5,6). De novo thymidylate synthesis and homocysteine remethylation are highly sensitive to network disruptions, which result in increased incorporation of dUTP into DNA, elevated cellular homocysteine, and impaired S-adenosylmethionine-dependent methylation reactions. De novo purine biosynthesis is less sensitive to network disruptions, although the mechanisms for this protection are unknown. Impaired folate metabolism increases risk for pathologies and developmental anomalies, including epithelial cancers, cardiovascular disease, and neural tube defects (7). Because of its role in nucleotide biosynthesis and cellular methylation reactions, this network continues to be an attractive target for the development of antiproliferative drugs (8 -10).
In the cell, folates differ by the reduction state of the pteridine ring, one-carbon substitution at the N5 and/or N10 positions, and the length of the glutamate polypeptide, which can range from 1 to 9 glutamate residues linked through ␥-peptide linkages ( Fig. 1) (1). Folates are transported across cell membranes as monoglutamate derivatives that are converted to folate polyglutamates by the enzyme folylpolyglutamate synthetase (1,11). The polyglutamate chain serves both to retain folates within the cell and to increase the affinity of folate derivatives for folate binding enzymes (1). Typically, folate polyglutamates bind to proteins two to three orders of magnitude tighter than corresponding monoglutamate forms (12).
There are five naturally occurring one-carbon-substituted forms of THF (Fig. 2). 5-MethylTHF carries the one-carbon unit at the oxidation level of methanol for the remethylation of homocysteine to methionine. 5,10-MethyleneTHF carries the one-carbon unit at the oxidation level of formaldehyde for thymidylate biosynthesis. Three THF derivatives, 10-formylTHF, 5-formylTHF, and 5,10-methenylTHF, carry the onecarbon unit at the oxidation level of formate. Of these forms, only 10-formylTHF is an enzyme cofactor and supplies the number 2 and number 8 carbons of the purine ring. Folate-dependent biosynthetic enzymes show absolute substrate specificity for a single folate cofactor. However, other one-carbon-substituted forms often function as potent enzyme inhibitors that regulate flux through the network (1,13).
Although 5-formylTHF is the most stable natural folate, little is known about its physiological role and regulation in mammalian systems. 5-FormylTHF is synthesized from 5,10-methenylTHF in a reaction catalyzed by serine hydroxymethyltransferase (14 -18). 5,10-Methenyltetrahydrofolate synthetase (MTHFS, EC 6.3.3.2) is the only enzymethatmetabolizes5-formylTHFandirreversiblycatalyzesitsATPdependent cyclization to 5,10-methenylTHF. In prokaryotes, there is evidence that 5-formylTHF is a storage form of folate and that its accumulation in dormant cells, including seeds and spores, is mediated by alterations in MTHFS expression (19,20). However, 5-formylTHF is not known to account for more than 10 -15% of total folate in mammalian cells (19). In mammalian systems, 5-formylTHF has been shown to inhibit several folate-dependent enzymes. 5-FormylTHF polygluta-mates are tight binding inhibitors of serine hydroxymethyltransferase and phosphoribosylaminoimidazole carboxamide formyltransferase (21,22). In addition to catalyzing 5-formylTHF synthesis, serine hydroxymethyltransferase also catalyzes the interconversion of serine and THF to glycine and methyleneTHF, a reaction that generates onecarbon units for purine, thymidine, and methionine biosynthesis (21). Phosphoribosylaminoimidazole carboxamide formyltransferase catalyzes the incorporation of formate into the C2 position of the purine ring. Inhibition of MTHFS in human MCF-7 cells by exposure to the MTHFS antifolate inhibitor, 5-formyltetrahydrohomofolate (5-formyl-THHF), results in an accumulation of cellular folate as 5-formylTHF. This creates a purine auxotrophy because elevated levels of 5-formylTHF inhibit phosphoribosylaminoimidazole carboxamide formyltransferase (22). Therefore, it has been proposed that alterations in cellular 5-formylTHF concentrations, mediated through changes in MTHFS activity, may regulate purine biosynthesis (19). In this study, we elucidated a mechanism for feedback metabolic regulation of MTHFS by 10-formylTHF and provide evidence that this regulatory mechanism supports a role for 5-formylTHF as storage form of folate in mammalian cells. Furthermore, we discovered a novel role for MTHFS-bound 10-formylTHF in the regulation of de novo purine biosynthesis.   All other materials were of high quality and obtained from various commercial vendors.

Materials
Synthesis of Folate and Antifolate Derivatives-[6R]-10-formylTHF (the natural isomer) was synthesized from [6S]-5-formylTHF as described previously (23). 5-Formyl, 6-methyltetrahydropterin and 5-formyltetrahydrohomofolic acid were synthesized from 6-methylpterin and homofolic acid, respectively, as described by others (11). Both homofolic acid and 6-methylpterin were reduced to tetrahydrohomofolic acid and 6-methyltetrahydropterin, respectively, by reduction over palladium(II) oxide hydrate in water or formic acid, respectively; formylation at N5 was achieved by the immediate addition of formic acid and N-(3-dimethylaminopropyl)-NЈ-ethylcarbodiimide at pH 4.0 as described previously (24). All synthesized compounds were purified by anion exchange chromatography and desalted by gravity filtration on a 0.5 ϫ 64-inch G-10 Sephadex column using water as the mobile phase as described elsewhere (23). Samples were lyophilized and their identity verified by UV spectroscopy and 1 H NMR.
Homology Model Construction-A model of the mouse MTHFS protein was constructed using the crystal structure of Mycoplasma pneumoniae MTHFS with ADP, phosphate, and 5-formylTHF bound (PDB code 1U3G; Fig. 3) (25,26). A BLAST search of the Swiss-Prot data base (27), using the SIB BLAST Network Service (28), revealed a number of sequences related to mouse MTHFS. A set of eight sequences comprising three mammalian and five bacterial proteins, including the murine and M. pneumoniae MTHFS enzymes, was aligned using ClustalW (29). A small adjustment was made to the location of insertions in the murine relative to the M. pneumoniae sequence to minimize interruption of secondary structure elements by the insertions. The SWISS-MODEL homology modeling server (30) was used to generate a model of mouse MTHFS. The M. pneumoniae and mouse sequences are 28% identical and 70% similar, and the alignment of the portions of the sequence that are involved in ATP and THF binding is unambiguous. The conformations and locations of inserted loops are uncertain, but none of these is close enough to the active site to be directly involved in catalysis.
Protein Expression and Purification-Recombinant murine MTHFS protein was expressed and purified as described elsewhere (31).
Steady-state Kinetics-MTHFS activity was monitored by measuring the increase in 5,10-methenylTHF production with time at 37°C in 1-ml cuvettes using a spectrophotometer; 5,10-methenylTHF has a unique absorbance maximum at , 355 nm (⑀, 25,100). Purine Biosynthesis Assay-SH-SY5Y neuroblastoma cells expressing the human MTHFS cDNA (SH-SY5YMTHFS) have been described elsewhere (19,31). These cells were maintained in Minimal Essential Medium, alpha modification (Hyclone) supplemented with 11% fetal bovine serum. For tracer experiments, fetal bovine serum was dialyzed against phosphate-buffered saline for over 24 h with four buffer changes to remove folate and other small molecules and then charcoal treated to remove any residual folate. The tracer medium was Defined Minimal Essential Medium (Hyclone) that lacked glycine, serine, methionine, hypoxanthine, and folate but was supplemented with 200 M methionine, 20 nM leucovorin, 2 nM [ 3 H]hypoxanthine, and 20 M [ 14 C]formate. Cells were split 1:3 and grown in 6-well plates at 37°C, 5% CO 2 in tracer medium until confluent and harvested. The cell pellets were stored at Ϫ40°C. Nuclear DNA was isolated using a DNA blood kit from Quiagen and isotope levels quantified on a Beckman LS6500 scintillation counter in dual dpm mode. For HPLC separation of nucleosides, cells were grown to confluence and harvested. DNA was digested to nucleosides prior to separation by HPLC, using procedures described elsewhere (32). Peaks corresponding to nucleosides were identified from standards, collected, and isotope levels quantified.

Determinants of MTHFS Substrate Specificity-
The contribution of all folate chemical moieties to substrate binding and catalysis was elucidated for recombinant murine MTHFS (Table 1 and Fig. 1). [6R,S]-5-formylTHFGlu 3 , an endogenous form of cellular folate, is a substrate for murine MTHFS (K m , 0.4 M; k cat /K m , 225). Previous studies have shown that the unnatural [6R]-5-formylTHF is not a substrate or inhibitor of mammalian MTHFS enzymes; therefore, the K m value for [6S]-5-formylTHFGlu 3 is likely 0.2 M (33, 34). The K m for [6S]-5-formylTHF is 10 M (k cat /K m , 6.0), in agreement with values determined for rabbit and human MTHFS (22,33). Therefore, the loss of the diglutamate chain increases the K m by 25-fold and decreases the substrate specificity (k cat /K m ) by 97%. [6R,S]-5-formyltetrahydropteroate, an unnatural derivative that lacks a glutamate moiety, is also a substrate (K m , 33 M; k cat /K m , 1.8), indicating that the Glu 1 moiety of 5-formylTHF makes only minor contributions to the K m and substrate specificity (k cat /K m ).
[6R,S]-5-formyl, 6-methyltetrahydropterin, which lacks the p-aminobenzoic acid and glutamate moieties, at a concentration of 60 M did not inhibit the MTHFS-catalyzed conversion of 10 M [6S]-5-formylTHF to 5, 10-methenylTHF, demonstrating that the p-aminobenzoate moiety of folate is essential for substrate binding. Collectively, these data indicate that pterins lack affinity for MTHFS and that the glutamate polypeptide decreases K m and increases substrate specificity.
Inhibition of MTHFS-Previous studies of human MTHFS protein have demonstrated that folic acid, THF, and 5-methylTHF are weak competitive inhibitors of human MTHFS (34). Inhibition of MTHFS by 10-formylTHF has never been investigated. The inhibition of recombinant murine MTHFS by folic acid and 10-formyl-substituted folate   (Fig. 3, blue for positive and red for negative). The positioning of the substrate, 5-formylTHF (in green), is iden-  . Active site models of the MTHFS ternary complexes. THF and ATP are shown with dark CPK coloring and MTHFS with lighter CPK coloring, except for Tyr-151 (on the left) and Tyr-152 (at the top), whose side chains are colored green. These Tyr residues are conserved in all MTHFS sequences and contribute to the active site. The THF formyl group is shown in magenta. The figures were rendered using Molscript (44) and Raster3D (45). A, 5-formylTHF is positioned to attack the ␥-phosphate of ATP. The thin black lines emphasize that the formyl O is positioned in the middle of the triangle formed by three phosphate oxygens with appropriate orientation and approximation for nucleophilic attack on the ␥-phosphate. B, 10-formylTHF, positioned to place the formyl oxygen as close as possible to ATP ␥-phosphate. The thin black lines show that the formyl oxygen approach is blocked by two phosphate oxygens, an unfavorable direction for nucleophilic attack.
tical to that observed in the active site of the M. pneumoniae crystal structure of the MTHFS⅐ADP⅐P i ⅐5-formylTHF complex. In this model, the MTHFS N terminus is at the lower left, the entrance to the ATP binding site is on the left side, and the view is directly into the 5-formylTHF binding site. All the inserted loops (i.e. the poorly modeled portions of the structure) are in the top half of the molecule. The positively charged region at the bottom of the structure is the putative polyglutamate binding site. In this model, the conformation of 5-formylTHF exposes the ␥-carboxyl of the 5-formylTHF glutamate moiety to the solvent, but this part of the molecule was not well defined in the crystal structure (zero occupancies are given in the Protein Data Bank for all atoms past C9). Adjustment of the 5-formylTHF model using O (35) reveals that it is easily possible for hydrogen-bonding interactions to be made between polar atoms in the substrate glutamate moiety and the side chains of residues Lys-17 and Arg-147. Glu-62 could also be part of a network of H-bonds involving 5-formylTHF and the protein surface. All three of these residues are conserved in known mammalian MTHFS sequences, and Lys-17 of rabbit liver MTHFS can be cross-linked to the ␣-carboxylate of 5-formylTHF (36). The lack of specific contacts for the 5-formylTHF Glu 1 is consistent with its apparent weak contribution to substrate specificity (Table 1).
Residues lining the 5-formylTHF binding site include amino acid sequence regions 58 -64, 97-105, 133-136, and 144 -152 (corresponding to 49 -55, 76 -84, 104 -107, and 115-123, respectively, in the M. pneumoniae protein) (Fig. 4a). The pterin moiety of 5-formylTHF is coordinated by the side chain of Glu-64 and the main chain O of Asp-59 and probably by Gln-58 O ⑀1 and Ser-97 O ␥ . The first two of these interactions are present in the M. pneumoniae structure and in the MTHFS of most species, whereas the latter two are not. Modeling an ATP into the MTHFS active site such that the ␥-phosphate occupies the site of a phosphate in the M. pneumoniae structure illustrates that a simple rotation of the formyl group of 5-formylTHF places the formyl O in good approximation and orientation for nucleophilic attack on the ␥-phosphorous (Fig. 4a). The importance of the highly conserved residues 144, 148, and 150 -152 in defining the active site is apparent: Arg-144 occupies a critical position coordinating both the N5-formyl group of 5-formylTHF and the ␥-phosphate of ATP.
10-FormylTHF can be built into the THF binding site without altering the position of side chain residues (Fig. 4b). Because the active site is not tightly constricted, several folate derivatives can be positioned to form good interactions between the pterin moiety and residues lining the binding site and between the glutamate moiety and surface hydrophilic residues. This is consistent with the kinetic data that demonstrate weak inhibition of MTHFS activity by most folate derivatives (Table 2). However, it is not possible to position the N10 formyl oxygen properly for attack on the ␥-phosphorous of ATP. Approach of the formyl oxygen to the phosphorous is hindered by two of the phosphate oxygens. The angle (formyl O)-P-(opposite P-O bond) is nearly linear with the 5-formyl moiety but is bent by ϳ40°with the 10-formyl moiety. In the case of N10 formyl species, it is possible that an H-bond could be formed between the formyl O and Lys-149 N; whether this is sufficient to explain the tighter binding of N10-formylated inhibitors is unclear ( Table 2).
Effect of MTHFS Expression on Purine Biosynthesis-The intracellular concentration of folate derivatives is less than the binding capacity of folate-utilizing enzymes, indicating that biosynthetic pathways within the one-carbon network compete for a limiting pool of folate cofactors (6,34). Folate-binding proteins can serve as "sinks" that sequester specific folates and thereby inhibit folate-dependent pathways or can interact with other enzymes to selectively "channel" cofactors and accelerate flux through individual biosynthetic pathways (12). 10-FormylTHF is required by GARFT and phosphoribosylaminoimidazole carboxamide formyltransferase, two folate-dependent enzymes involved in de novo purine biosynthesis (Fig. 2) (1, 11). To determine the metabolic effects of 10-formylTHF sequestration by MTHFS on de novo purine biosynthesis, a "formate suppression" assay was developed. Mammalian cells expressing the MTHFS cDNA were cultured in the presence of [ 3 H]hypoxanthine and [ 14 C]formate. [ 3 H]Hypoxanthine is converted to purine nucleotides via the folate-independent salvage pathway, whereas [ 14 C]formate is incorporated into purine nucleotides via the de novo pathway after condensing with THF to form 10-formylTHF (Fig. 5). The ability of the de novo purine biosynthetic pathway to suppress contributions from the purine salvage pathway to DNA synthesis was investigated in human SHSY-5Y and SHSY-5YMTHFS neuroblastoma cells. SHSY-5YMTHFS cells display 100-fold increased MTHFS activity and protein levels (19,34). The ratio of 14 C to 3 H (dpm) in DNA and purine nucleotides serves as a measure of de novo purine synthesis efficiency. The 14 C/ 3 H ratio in nuclear DNA is 20% higher in SHSY-5YMTHFS cells compared with the parent cell line (Fig. 6). Because 14 C could also  [ 3 H]Hypoxanthine is converted to purines via the folate-independent salvage pathway, whereas [ 14 C]formate is incorporated into purines via the de novo pathway. The 14 C/ 3 H dpm ratio was determined in SHSY-5Y and SHSY-5YMTHFS nuclear DNA. Nuclear DNA was isolated from SHSY-5Y (white bars) and SHSY-5YMTHFS (shaded bars) and digested to nucleosides, which were then separated by HPLC. The 14 C and 3 H content (dpm) in resulting fractions was quantified on a Beckman Coulter LS6500 scintillation counter. Variation is expressed as S.D. of the mean from three measurements for separated nucleosides and six experiments for nuclear DNA. dA, deoxyadenosine; dT, deoxythymidine; dG, deoxyguanosine.
be incorporated into deoxythymidine and methylcytosine (Fig. 2) via equilibration into the folate-activated one-carbon pool, the DNA was digested to nucleosides, which were fractionated by HPLC. The deoxyguanosine and deoxyadenosine 14 C/ 3 H ratio was increased by 43 and 69%, respectively, in SHSY-5YMTHFS cells compared with the parent cell line (Fig. 6). Comparison of 14 C counts derived from purified deoxythymidine normalized to 3 H counts from deoxyadenosine indicates that increased MTHFS expression does not affect the 14 C deoxythymidine/ 3 H deoxyadenosine ratio. The enhancement of de novo purine biosynthesis by MTHFS indicates that MTHFS-bound 10-formylTHF is available for de novo purine biosynthesis.

DISCUSSION
The data presented in this study provide the first evidence that 5-formylTHF concentrations are regulated in mammalian cells. The product of the MTHFS reaction, 5, 10-methenylTHF, exists in chemical equilibrium with 10-formylTHF. Upon the hydrolysis of 5,10-methe-nylTHF to 10-formylTHF, the MTHFS reaction is subject to feedback inhibition by 10-formylTHF (Table 2) (Fig. 7). This discovery that MTHFS is regulated by 10-formylTHF informs the physiological function of 5-formylTHF. 5-FormylTHF can only accumulate when MTHFS is inhibited by 10-formylTHF. Therefore, 10-formylTHF accumulation both suppresses 5-formylTHF metabolism and, through its equilibrium conversion to 5,10-methenylTHF, serves as a substrate for 5-formylTHF synthesis by serine hydroxymethyltransferase. This regulatory mechanism does not support a role for physiological regulation of purine biosynthesis by 5-formylTHF. 5-FormylTHF can only accumulate when the substrate for de novo purine biosynthesis, 10-formylTHF, accumulates. Feedback regulation of MTHFS does indicate that 5-formylTHF represents a stable pool of excess formyl folates that can be mobilized by MTHFS only when 10-formylTHF levels are depleted.
These studies have also identified MTHFS as a 10-formylTHF tightbinding protein. Folate cofactors are bound by enzymes that utilize one-carbon units in transfer reactions and also by folate tight-binding proteins that do not metabolize the cofactor but regulate its availability. For example, 5-methylTHF is a substrate for methionine synthase but is bound tightly to and inhibits glycine N-methyltransferase (37). THF is bound tightly by and inhibits 10-formyltetrahydrofolate dehydrogenase (38), and 5-formylTHF and 5-methylTHF are bound tightly by and inhibit serine hydroxymethyltransferase (39). MTHFS is the first 10-formylTHF tight-binding protein to be identified. Previously, we have shown that increased MTHFS expression in SHSY-5Y cells increased relative concentrations of 10-formylTHF at the expense of 5-methylTHF levels; 10-formylTHF levels constituted as much as 90% of total cellular folate in MTHFS-expressing cells (21). These results are consistent with MTHFS serving as a reservoir for 10-formylTHF that permits its accumulation.
The results from this study indicate that MTHFS may determine the fate of cellular 10-formylTHF (Fig. 7). 10-FormylTHF can be used by three pathways. It is a cofactor for phosphoribosylaminoimidazole carboxamide formyltransferase and GARFT during de novo purine biosynthesis. Alternatively, the 10-formyl group can be reduced and the cofactor converted to other THF one-carbon forms through its conversion to 5,10-methenylTHF by the enzyme 5,10-methenylTHF cyclohydrolase. Finally, 10-formylTHF can be converted to THF and CO 2 by the enzyme 10-formyltetrahydrofolate dehydrogenase, an enzyme that depletes the supply of 10-formylTHF (38). In this study, we have demonstrated that sequestration of 10-formylTHF by MTHFS enhances de novo purine biosynthesis, which supports a role for MTHFS in directing 10-formylTHF to de novo purine synthesis. 10-FormylTHF sequestration by MTHFS may function to protect purine biosynthesis from disruptions in the one-carbon network, including folate deficiency. Although there is no direct evidence that MTHFS associates with phosphoribosylaminoimidazole carboxamide formyltransferase or GARFT and channels 10-formylTHF, the electrostatic surface of MTHFS shows a large negatively charged area near the active site that is a potential protein-protein interaction site (Fig. 3). Previous studies have demonstrated that MTHFS and cytoplasmic serine hydroxymethyltransferase exhibit very different tissue expression patterns (40), indicating that MTHFS may play other roles in one-carbon metabolism distinct from metabolizing 5-formylTHF, including serving as a 10-formylTHF-binding protein.
These results inform the design of inhibitors that target MTHFS. The data in Table 2 indicate that high affinity MTHFS inhibitors should be N10 formyl substituted and be capable of conversion to polyglutamate derivatives in the cell. N5-substituted MTHFS inhibitors, including 5-formylTHHF, are not attractive in vivo inhibitors because they can be phosphorylated and slowly metabolized by some mammalian MTHFS enzymes and because they are not effective substrates for folylpolyglutamate synthetase (1). A number of GARFT inhibitors have been synthesized (4,8,41,42), including 5,10-dideazatetrahydrofolate (DDATHF or Lometrexol). DDATHF inhibited mouse GARFT with a K i of 6 nM and human GARFT with a K i of 60 nM. It also proved to be an effective inhibitor of cell growth with an EC 50 of 10 -30 nM in several different cell lines (42). A similar compound, 10-formyl-5,10-dideaza-acyclicTHF (10-formyl-DDACTHF), was shown to exhibit some selectivity for GARFT. 10-Formyl-DDACTHF is a substrate for folylpolyglutamate synthetase and accumulates in cell cultures over 100-fold (41). The pentaglutamate form effectively inhibits GARFT (K i , 14 nM) and was an effective cytotoxic agent (IC 50 , 60 nM). Given the high affinity of MTHFS for the natural isomer of 10-formylTHFGlu 3 (K i , 15 nM) and its relaxed specificity for the pterin moiety, the results of the present work suggest that N10-substituted analogs of 10-formylTHF may be targeting MTHFS, which may contribute to their cytotoxicity.