Disruption of the mthfd1 gene reveals a monofunctional 10-formyltetrahydrofolate synthetase in mammalian mitochondria.

The Mthfd1 gene encoding the cytoplasmic methylenetetrahydrofolate dehydrogenase-methenyltetrahydrofolate cyclohydrolase-formyltetrahydrofolate synthetase enzyme (DCS) was inactivated in embryonic stem cells. The null embryonic stem cells were used to generate spontaneously immortalized fibroblast cell lines that exhibit the expected purine auxotrophy. Elimination of these cytoplasmic activities allowed for the accurate assessment of similar activities encoded by other genes in these cells. A low level of 10-formyltetrahydrofolate synthetase was detected and was shown to be localized to mitochondria. However, NADP-dependent methylenetetrahydrofolate dehydrogenase activity was not detected. Northern blot analysis suggests that a recently identified mitochondrial DCS (Prasannan, P., Pike, S., Peng, K., Shane, B., and Appling, D. R. (2003) J. Biol. Chem. 278, 43178-43187) is responsible for the synthetase activity. The lack of NADP-dependent dehydrogenase activity suggests that this RNA may encode a monofunctional synthetase. Moreover, examination of the primary structure of this novel protein revealed mutations in key residues required for dehydrogenase and cyclohydrolase activities. This monofunctional synthetase completes the pathway for the production of formate from formyltetrahydrofolate in the mitochondria in our model of mammalian one-carbon folate metabolism in embryonic and transformed cells.

Folate is a B vitamin that plays important roles in cellular growth and division. The one-carbon substituted forms of tetrahydrofolate (THF) 1 are involved in the de novo synthesis of purines and thymidylate and support cellular methylation reactions through the regeneration of methionine from homocysteine. In eukaryotes folate metabolism is separated into two compartments, the cytoplasm and the mitochondria. Some enzymes, such as serine hydroxymethyltransferase, are found in both compartments, usually encoded by separate nuclear genes (1). In yeast, trifunctional methyleneTHF dehydrogenase-methenylTHF cyclohydrolase-formylTHF synthetases (DCS) are found in both the cytoplasmic and mitochondrial compartments (2,3). 13 C tracer studies in yeast have demonstrated that one-carbon units from serine flow through the mitochondria to the cytoplasm through the action of the mitochondrial DCS (4,5). The one-carbon units exit the mitochondria as formate, produced by the reversal of the 10-formylTHF synthetase activity. The formate is then used by the cytoplasmic DCS, which converts it into 10-formylTHF for use in purine synthesis and other biosynthetic reactions.
The cytoplasmic DCS has been identified and cloned in several organisms (2, 6 -8). The genes encode a single polypeptide possessing the three DCS activities. The 100-kDa monomer dimerizes to form the active enzyme. However, until recently the only mitochondrial homologue of Saccharomyces cerevisiae mitochondrial DCS to be identified in mammals was the NADdependent methyleneTHF dehydrogenase-methenylTHF cyclohydrolase (NMDMC) (9). Metabolic studies of transformed fibroblasts derived from murine NMDMC knockouts revealed that they are glycine auxotrophs (10). The glycine auxotrophy is most likely because of the formation of a type of folate trap; the available THF in the mitochondria is trapped as methyl-eneTHF. Moreover, the glycine auxotrophy indicates that NMDMC is the only mitochondrial methyleneTHF dehydrogenase expressed in transformed cells. These studies also supported the concept that mammalian mitochondria generate formate to supply the cytoplasmic DCS with one-carbon units during periods of rapid growth, such as embryogenesis. It was suggested that a yet to be identified 10-formylTHF hydrolase or a monofunctional 10-formylTHF synthetase might provide the missing link between the 10-formylTHF produced by NMDMC and the formate released to the cytoplasm. Barlowe and Appling (11) reported the detection of all three of the DCS activities in purified adult rat liver mitochondria preparations. However, because NMDMC null cell lines appear not to express another mitochondrial methyleneTHF dehydrogenase, it is unlikely that a trifunctional DCS is expressed in the mitochondria of transformed and ES cell lines. The detection of the DCS activities in mitochondrial preparations is complicated by contamination by the high activities of the cytosolic protein. To determine whether either of the dehydrogenase or synthetase activities exist in the mitochondria of immortalized cell lines we undertook to inactivate the cytoplasmic DCS gene.
Here we report the production of DCS null cell lines that we used to detect mitochondrial synthetase activity. We propose that the detected activity is due to a recently identified mitochondrial DCS protein in which the dehydrogenase and cyclohydrolase activities have been inactivated, resulting in a monofunctional synthetase. * 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 1 The abbreviations used are: THF, tetrahydrofolate; DCS, methyl-eneTHF dehydrogenase-methenylTHF cyclohydrolase-formylTHF synthetase; NMDMC, mitochondrial NAD-dependent methyleneTHF dehydrogenase methenylTHF cyclohydrolase; mtDCS, mitochondrial methyleneTHF dehydrogenase-methenylTHF cyclohydrolase-formyl-THF synthetase; ES, embryonic stem; DC, methyleneTHF dehydrogenase-methenylTHF cyclohydrolase.

EXPERIMENTAL PROCEDURES
Knock-out of DCS in Embryonic Stem Cells-The gene and cDNA for DCS were isolated as described previously (8) (GenBank TM accession numbers AF364580 to AF364592) 2 . The targeting vector used for homologous recombination is shown in Fig. 1. The vector is designed to insert the cDNA of DCS into the first exon of the nuclear gene; a rabbit ␤-globin intron 2/SV40 polyadenylation signal cassette obtained from the pSG5 vector (Stratagene) was attached to the cDNA to provide processing signals. The 5Ј arm is a 1.5-kb HindIII NcoI fragment containing exon 1 up to the start codon, and the 3Ј arm is a 2.3-kb SacI fragment containing exon 2. The arms and cDNA were inserted into the pMC1neopA vector (Stratagene) to which a herpes simplex virus thymidine kinase cassette had been added. To knock out the gene encoding DCS with this vector the first 106 nucleotides of the coding sequence were removed, altering the reading frame of the remaining sequence. The cDNA also contained a K386E mutation that abrogates synthetase function (data not shown). Generation of heterozygous and homozygous embryonic stem cells was done as described previously (12). Clones were screened by Southern blot (12); HindIII digested genomic DNA was probed with a 1.65-kb HindIII PstI fragment containing exon 1 and a 1.5-kb PstI HindIII fragment of intron 2 as shown in Fig. 2.
Generation of Null Mutant Fibroblasts-The homozygous null DCS (Ϫ/Ϫ) ES cell lines were injected into blastocysts as described previously (12). The E11.5 chimeric embryos were used to generate primary embryonic fibroblast cell lines as described previously (10). Disrupted embryos were incubated in Dulbecco's modified Eagle's medium containing 15% fetal bovine serum supplemented with 1ϫ non-essential amino acids, 1ϫ glutamine, 1ϫ penicillin/streptomycin (all Invitrogen), 30 M thymidine, and 30 M hypoxanthine (both Sigma) with 0.3 mg/ml G418 (Geneticin, Invitrogen) to select for null mutant cells. All cell lines described in this work were routinely cultured in this medium, unless otherwise noted. The embryonic fibroblasts were spontaneously immortalized by continuous passage in culture.
Purine Auxotrophy of Null Mutant Fibroblasts-Cells lacking the cytoplasmic DCS cannot produce the 10-formylTHF required for the synthesis of purines from either formate or methyleneTHF. To functionally test the knock-out, the growth of the cells with and without supplementation by the purine hypoxanthine was evaluated. SF6 (Ϫ/Ϫ) and IF22 (ϩ/ϩ) cells were cultured in the medium described above with the addition of 1ϫ ITS-S (insulin, transferrin, sodium selenite, and ethanolamine, from Sigma) and the substitution of 15% redialyzed fetal bovine serum (Invitrogen) with or without hypoxanthine as required. The fetal bovine serum was redialyzed, and the cells were adapted to it as described by Patel et al. (10). Cells were plated in 6-well plates at a density of 1 ϫ 10 4 cells/well in medium containing hypoxanthine. After 24 h the medium was changed to the medium required for the experiment. Cells were counted in triplicate at 24-h intervals by the trypan blue exclusion method.
Cell Culture and Enzyme Assays-For assay of enzyme activities in whole cell extracts, the immortalized wild type and null fibroblasts were grown to confluence in 5ϫ T175 flasks, harvested by trypsinization, and washed with phosphate-buffered saline. The cell pellet, about 300 mg of packed cells, was stored at Ϫ85°C. For the isolation of purified mitochondria, SF6 cells (immortalized null fibroblasts) were grown to confluence in 26 ϫ 15-cm tissue culture plates. The cells were harvested and the mitochondria isolated as described in Ruffolo et al. (13). Approximately 0.15 g of mitochondria was obtained per gram of packed cells.
To prepare extracts of either whole cells or mitochondria, the appropriate pellets were resuspended in 3 ml of sonication buffer (100 mM potassium phosphate, pH 7.3, 1 mM benzamidine hydrochloride, 1 mM phenylmethylsulfonyl fluoride, and 35 mM ␤-mercaptoethanol, all Bioshop) per gram of pellet and sonicated on ice with 3 ϫ 10-s pulses separated by 1 min of rest on ice. Glycerol (0.24ϫ volume) was added to the sonicate, which was centrifuged for 10 min in a microcentrifuge at 4°C. The supernatants were transferred to fresh tubes and kept at 4°C until assayed.
The dehydrogenase and synthetase activities of the extracts were determined using a multiple time point assay to ensure detection of small amounts of enzyme activity. The assay conditions are as described in Mejia and MacKenzie (9) and Tan et al. (14) with the following changes. All assay mixtures contained 144 M ␤-mercaptoethanol, and the NAD-dependent dehydrogenase assay mixture contained 630 M NAD. NAD-dependent dehydrogenase assays were also done without magnesium to look for dehydrogenase activity not due to NMDMC. 10 mM EDTA was added to assays without magnesium to chelate any trace amounts found in the crude extracts. The protein content of the extracts was determined by Bradford assay using bovine serum albumin as a standard (15).
Northern Blot Analysis of NMDMC and DCS Mutants-Total RNA was isolated from IF22 (ϩ/ϩ), IF74 (NMDMC (Ϫ/Ϫ)), and SF6 (DCS (Ϫ/Ϫ)) cell lines grown to confluence in a T175 flask using the Qiagen RNeasy kit. 20 g of each RNA sample was electrophoresed on a 1% formaldehyde gel in quadruplicate and transferred to a Hybond-N membrane (Amersham Biosciences) by vacuum transfer in 10ϫ SSC for 2 h. The RNA was cross-linked to the membrane under ultraviolet light for 5 min. Each subset of membrane was probed with one of: full-length NMDMC cDNA (2-kb, GenBank TM accession number NM_008638), fulllength DCS cDNA (3-kb, GenBank TM accession number AF364580), a 1.3-kb NcoI fragment of the clone 6311761 (GenBank TM accession number BQ917278), which codes for the 5Ј end of the mtDCS, and a 1.7-kb PstI fragment of the clone 5374413 (GenBank TM accession number BC030437), which codes for the 3Ј end of the mtDCS. Clones 6311761 and 5374413 were obtained from Open Biosystems. Each probe was labeled using the Rediprime II random prime labeling system (Amersham Biosciences) and [ 32 P]dCTP (Amersham Biosciences). Hybridization, washing, and exposure of the blots were performed according to standard protocols.

DCS Null
Fibroblasts-DCS heterozygous and null ES cells were established as described, and their genotypes were verified by Southern blot analysis (Fig. 2). Null ES cells were injected into blastocysts, and DCS null fibroblast cell lines were established from chimeras using high G418 medium to select against wild type cells. Fig. 3 shows that the DCS null cell lines are purine auxotrophs; they have no means of producing the 10-formylTHF required for purine synthesis in the cytoplasm from either serine or formate and are therefore dependent on exogenous sources of purines. This phenotype confirms the knock-out of the cytoplasmic methyleneTHF dehydrogenase and 10-formylTHF synthetase activities.
Enzyme Assays-The dehydrogenase and synthetase activities of wild type and DCS null mutant cell lines are shown in Fig. 4 and Table I. These results show low but significant synthetase activity in cell extracts in the absence of the usually confounding cytoplasmic DCS activities. The synthetase activity of mtDCS is about 6% of the total synthetase activity in the wild type cells.
Although the synthetase activity can be detected in whole cell extracts, assays of mitochondrial extracts were required to confirm the subcellular localization and to allow the assay of more concentrated protein samples to confirm the lack of NADPdependent dehydrogenase activity. The results of these assays are shown in Fig. 5 and Table I. The NAD-dependent dehydrogenase activity serves as a marker for the enrichment of mitochondrial proteins. Although there is significant synthetase activity, confirming the mitochondrial location of mtDCS, there is no detectable NADP-dependent dehydrogenase activity. To confirm that the NAD-dependent dehydrogenase activity observed is NMDMC, the assay was repeated without added magnesium and with EDTA added to the assay mix to chelate magnesium present in the protein sample. In the absence of magnesium ions, which are uniquely required by NMDMC for dehydrogenase activity (17), no NAD-dependent dehydrogenase activity is observed.
Northern Analysis-Two separate laboratories have recently published the discovery of mitochondrial methyleneTHF dehydrogenase-cyclohydrolase-synthetase cDNAs in mouse and human (18,19). The encoded protein conforms to known DCS sequences in that it consists of separate DC and synthetase domains connected by a short linker region. The human protein has about 57% overall homology to the cytoplasmic DCS (19). However, the DC domain has only 33% homology to the cytoplasmic DCS DC domain and has not been demonstrated to possess either dehydrogenase or cyclohydrolase activity (18,19). To determine whether this protein may be responsible for the observed synthetase activity in the DCS null cell lines expression of its message was evaluated by Northern analysis. Fig. 6 shows Northern blots of total RNA isolated from wild type, NMDMC null and DCS null cell lines probed for DCS, NMDMC, and the mitochondrial DCS RNA. Separate probes for the DC and synthetase domains of mtDCS were used to reduce the possibility of confounding results because of cross hybridization. All three messages were detected in all three cell lines, with the exception of the NMDMC and DCS messages in their respective knock-out cell lines. The Northern results also show that the probes for DCS and mtDCS do not crosshybridize. DISCUSSION The model of folate metabolism well elucidated by Appling and co-workers (4, 5, 20) for yeast was also proposed to apply to higher organisms (11) and has been used in several metabolic studies tracking the flow of one-carbon units carried by THF in mammals. This model contains an NADP-dependent DCS in the mitochondria, in analogy to the mitochondrial DCS protein found in yeast mitochondria. Although the inclusion of this hypothetical protein in models of one-carbon folate metabolism does little to change the interpretation of radiotracer studies, it ignores the lack of strong evidence to support a trifunctional DCS in mammalian mitochondria.
The mitochondrial NAD-dependent protein NMDMC activity was first detected in 1960 in Erlich ascites tumor cells (21). This protein remains the only methyleneTHF dehydrogenase in mammalian mitochondria to have been identified and characterized (9). We have shown through the comparison of the genes of NMDMC and DCS in mouse, human, and Drosophila that NMDMC arose from a trifunctional precursor through the loss of the synthetase domain and the change in cofactor specificity from NADP to NAD (8). This cofactor specificity change is often neglected in applying the yeast-based model to mammals. The use of NAD in embryonic and transformed cells shifts the equilibrium of the reaction to favor the production of 10-formylTHF over that of methyleneTHF (22). This shift optimizes the mitochondrial system to produce one-carbon units, presumably as formate, for use in purine synthesis in the cytoplasm as was found in yeast. Patel et al. (10) demonstrated that the growth of NMDMC null fibroblasts is enhanced by the addition of formate to the medium. It was also shown that there is an increase of 14 C-labeled formate into DNA in NMDMC null mutants, as compared with wild type cells. These results confirmed that, as in yeast, the mitochondria serve as sources of one-carbon units in the form of formate for use in purine synthesis in the cytoplasm. There was, however, a gap in our model that could not explain how formate is released from formyl-THF in the mitochondria to enter the cytoplasm. There are several possibilities including a 10-formylTHF hydrolase or a separate 10-formylTHF synthetase. Wild type murine fibroblast extracts have been assayed for 10-formyl hydrolase activity, but no activity was detected. 4 Because of the relatively high activity of the cytoplasmic DCS, attempting to detect a mitochondrial synthetase from whole cell extracts, even if overexpressed, presents a technical challenge. Even when subcellularly fractionated, very small amounts of contamination from the cytoplasmic enzyme could have a significant effect on the results of enzyme assays. Therefore, a knock-out of DCS provides the best opportunity to examine mitochondria for additional dehydrogenase and synthetase activities.
In the DCS null cells 10-formylTHF synthetase activity is observed both in whole cell and mitochondria extracts. Even when the sample is concentrated for mitochondrial proteins, as in Fig. 5, there is no detectable dehydrogenase activity that is not attributable to NMDMC. These assays can detect NADPdependent dehydrogenase activity down to 1 milliunit/ml, which would permit the detection of as little as 1% of the observed NAD-dependent dehydrogenase activity. The mtDCS described by Prasannan et al. (18) and Sugiura et al. (19) may provide the identity of the observed synthetase. Northern blots of total RNA extracted from wild type, NMDMC null, and DCS null transformed cell lines show that the mtDCS is expressed in these cell lines (Fig. 6). Recent work by Di Pietro et al. (12) and Patel et al. (10) has shown that there is only one methyl-eneTHF dehydrogenase expressed in embryonic and transformed cells. This observation is in agreement with the failure of both groups to detect dehydrogenase activity both in yeast 4 N. R. Mejia, unpublished observation. overexpressing the mtDCS cDNA (18) and in purified preparations of the DC domain (19).
To determine whether there is a structural basis for the lack of detectable dehydrogenase activity a multiple sequence alignment of known mtDCS, NMDMC, and DCS proteins was con-structed. Fig. 7 shows regions of the lineup of particular interest. In the human DCS, Lys 56 is a residue crucial to cyclohydrolase activity (23); mutation to asparagine, as in mt-DCS, will abrogate cyclohydrolase activity. Residues in the positions of human DCS Arg 173 , Ser 174 , Gly 178 , and Ser 197 are important in binding the adenine and diphosphate moieties of NAD(P) (24,25). The amino acid changes in mtDCS observed at these positions introduce oppositely charged, bulky, and hydrophobic side chains into the cofactor binding site. These mutations, both individually and collectively, will compromise NAD(P) binding. In addition, strictly conserved residues identified as interacting with the nicotinamide moiety of NADP (Thr 148 , Val 177 , Ile 238 , Gly 276 , and Thr 279 ) (24) are not conserved in mt-DCS. Residue Asp 208 of the yeast monofunctional NADdependent methyleneTHF dehydrogenase, which is involved in NAD binding (26), is also not found in mtDCS. Based on this comparison, mtDCS is most likely a monofunctional synthetase.  The specific activity for the lower limit of these assays was estimated based on the detection limit of 0.5 nmol min Ϫ1 ml Ϫ1 , and the average protein concentration of the samples. In most cases, monofunctional 10-formylTHF synthetases are found to be tetramers (27). The trifunctional proteins, however, are generally dimers that interact such that the synthetase domains of each monomer would be directed away from each other (23,24), precluding a strong interaction of the synthetase domains. When separated from the DC domain through proteolytic cleavage or protein mutagenesis, the synthetase domain has been found to be unstable (28,29). Although there are some significant deletions of amino acids within the DC domain, all of the structural elements involved in DC dimerization appear to be intact. Therefore, the non-functional DC domain of mtDCS may serve to stabilize the synthetase domain by allowing for dimerization.
It is likely that mtDCS, encoding a monofunctional synthetase, provides the missing metabolic reaction required to link the mitochondria and the cytoplasm in the mammalian model of one-carbon folate metabolism in embryonic and transformed cells (Fig. 8). However, firmly establishing the role of mtDCS in this system will await the development of a knock-out cell line. Together with NMDMC, the mitochondrial synthetase provides an example of divergent evolution to optimize the enzyme activities in the cell. The effects of the cofactor substitution to NAD and the embryonic expression pattern of NMDMC, suggest that its role is to supply one-carbon units to the cytoplasm during periods of rapid growth such as embryogenesis and tumorigenesis. The up-regulation of mtDCS in transformed cells (19) fits with the proposed role of NMDMC. The separation of the two genes allows for a more precise control of the expression of these proteins; NMDMC is expressed only during periods of rapid growth, during which the mitochondrial synthetase is up-regulated. Under other conditions, varying amounts of mitochondrial synthetase are expressed independently of NMDMC (18), perhaps to ensure a pool of formylTHF to supply the formylmethionyl-tRNA used in the initiation of mitochondrial protein synthesis.
It is not clear how this model of one-carbon flux in embryonic tissues and tumor cells can be reconciled with the observations of formate generation by adult liver mitochondria (11). The mRNA for NMDMC has been detected at quite low levels in adult liver (30), although NAD-dependent dehydrogenase activity was not detectable by enzyme assay of liver extracts (9). It is possible that extremely low levels of NMDMC expression, in cooperation with mtDCS, are adequate for the generation of formate by mitochondria in the amounts required by liver cells. However, in adult tissues the system of one-carbon flux could involve an as yet unidentified mitochondrial DCS or additional methylenetetrahydrofolate dehydrogenase activity that is not expressed in transformed cells. Residues important for enzyme activity and cofactor binding are highlighted by the boxes. In the human DCS, Lys 56 is the catalytic residue responsible for cyclohydrolase activity (23). G 172 XS 174 XXXG 178 is the NAD(P) binding consensus sequence of this enzyme family. Arg 173 forms hydrogen bonds with and provides charge balance to the 2Ј-phosphate group of NADP and is essential for NADP binding (24,25). In E. coli this position is occupied by an alanine, and the role of this residue is taken up by Arg 191 (31). Ser 174 interacts with the diphosphate moiety of NADP (24), and a glycine or other small uncharged residue is required at position 178 to maintain the binding site structure (32). Ser 197 hydrogen bonds to NADP to assist in cofactor binding (24,25). In several proteins it is substituted by arginine, where it may have a role similar to that of Arg 173 .