Molecular cloning, characterization, and regulation of the human mitochondrial serine hydroxymethyltransferase gene.

The human mitochondrial serine hydroxymethyltransferase (mSHMT) gene was isolated, sequenced, and characterized. The 4.5-kilobase gene contains 10 introns and 11 exons, with all splice junctions conforming to the GT/AG rule. The 5′ promoter region contains consensus motifs for several regulatory proteins including PEA-3, Sp-1, AP-2, and a CCCTCCC motif common to many genes expressed in liver. Consensus TATA or CAAT sequence motifs are not present, and primer extension and 5′-rapid amplification of cDNA ends studies suggest that transcription initiation occurs at multiple sites. The mitochondrial leader sequence region of the deduced mRNA contains two potential ATG start sites, which are encoded by separate exons. The intervening 891-base pair intron contains consensus promoter elements suggesting that mSHMT may be transcribed from alternate promoters. 5′-Rapid amplification of cDNA ends analysis demonstrated that the first ATG is transcribed in human MCF-7 cells. However, transfection of Chinese hamster ovary cells deficient in mSHMT activity with the human mSHMT gene lacking exon 1 overcame the cell's glycine auxotrophy and restored intracellular glycine concentrations to that observed in wild-type cells, showing that exon 1 is not essential for mSHMT localization or activity and that translation initiation from the second ATG is sufficient for mSHMT import into the mitochondria. Mitochondrial SHMT mRNA levels in MCF-7 cells did not vary during the cell cycle and were not affected by the absence of glycine, serine, folate, thymidylate, or purines from the media.

natively, one-carbon units can also be generated from glycine in cells that contain a glycine cleavage activity. SHMT is a pyridoxal phosphate-dependent enzyme that catalyzes the reversible interconversion of serine and H 4 PteGlu to glycine and 5,10-CH 2 -H 4 PteGlu. SHMT is present in both the mitochondria (mSHMT) and the cytoplasm (cSHMT) in mammalian cells. The human SHMT cDNAs encoding the two isozymes have been isolated and the genes localized to chromosomes 12q13 and 17p11.2, respectively (1). Currently, the metabolic role of the individual SHMT isozymes is not clearly understood. Chinese hamster ovary cells lacking mSHMT activity are auxotrophic for glycine, suggesting that the mitochondria are the primary site of glycine synthesis, whereas the enzymes responsible for thymidylate, purine, and methionine synthesis are present in the cytoplasm (2). The central role of SHMT isozymes in producing one-carbon-substituted folate cofactors has suggested that the regulation of these enzymes may influence cell growth and proliferation and that they may be targets for the development of antineoplastic agents.
Total SHMT activity and the concentration and metabolism of serine and glycine varies among tissues, reflecting the different roles of these amino acids in different organs (3). This effect is most pronounced in the brain, where it has been demonstrated that there is a direct correlation between SHMT activity and glycine concentration in different regions of rat brain (4). SHMT activity may also be developmentally regulated as SHMT-specific activity is 2-fold higher in the the optical lobe of the rhesus monkey neonate and adult compared with the fetus (5). There is evidence that SHMT is hormonally regulated as SHMT activity is elevated in the uterus after injection of 17␤-estradiol to ovariectomized rats, with a 6-fold acceleration of [3-14 C]serine incorporation into purines, whereas testosterone increases the specific activity of SHMT in the prostate (6). SHMT may also be controlled by nutrient availability as SHMT activity is increased 50% in folate-deficient versus folate-supplemented chickens (7). Total SHMT enzyme activity has also been demonstrated to be increased in tumor cells (8).
Several studies have suggested that mSHMT is primarily responsible for glycine synthesis in the cell (2,9). Revertants of mutant Chinese hamster ovary (CHO) cells that lack mSHMT activity showed a correlation between SHMT levels, intracellular glycine concentrations, and protein synthesis rates (2). While SHMT enzyme activity and one-carbon flux display both developmental, nutritional, and tissue-specific regulation, the metabolic significance of these changes has been difficult to interpret as most of these studies have been performed with crude homogenates that did not distinguish between mitochon-drial and cytoplasmic SHMT activities. In order to better understand the differential metabolic roles of the SHMT isozymes, we have cloned and structurally characterized the human mSHMT gene and have commenced studies to determine the factors responsible for regulating its endogenous expression.

EXPERIMENTAL PROCEDURES
Materials-PteGlu, reduced folates, amino acids, nucleosides, nucleotides, alcohol dehydrogenase, pyruvic acid, and NADH were obtained from Sigma. [ 32 P]dCTP (800 Ci/mmol) and [2-3 H]glycine (43.8 Ci/mmol) were obtained from DuPont NEN. Restriction and modifying enzymes were obtained from Boehringer Mannheim, Promega, or New England BioLabs. Taq polymerase was from Perkin-Elmer. DMEM, a modification of ␣-minimum essential medium lacking serine, glycine, folic acid, thymidylate, and hypoxanthine, was purchased from JHR Biosciences. All other materials were of high quality and obtained from various commercial vendors.
Cell Culture-MCF-7 human breast cancer cells (HTB22) were obtained from the American Type Culture Collection; wild-type CHO cells (WTT2) were obtained from Dr. Sharon Krag, Johns Hopkins University; GlyA, a CHO cell mutant lacking mSHMT activity, was obtained from Dr. Larry Thompson, Lawrence Livermore Labs. Cells were determined to be free of mycoplasm contamination by fluorescent DNA staining using 4,5-diamidino-2-phenylindole (Boehringer Mannheim) according to the manufacturer's instruction. Cells were cultured in 100-mm dishes containing 12 ml of media and incubated at 37°C in a 5% CO 2 atmosphere. Wild-type CHO and MCF-7 cells were maintained in DMEM supplemented with 10% fetal bovine serum (Hyclone Inc.) and 20 nM folinic acid, whereas culture media for GlyA were supplemented with 200 M glycine. For some experiments, the fetal bovine serum was dialyzed (dFBS) against 10 volumes of phosphate-buffered saline for 24 h with buffer changes every 4 h to deplete serum glycine and folate. Media were also supplemented with combinations of glycine (200 M), hypoxanthine (20 M), thymidine (20 M), and serine (200 M) for studies concerning effects of nutrient availability on mSHMT mRNA levels in MCF-7 cells. MCF-7 cells were synchronized using Lovastatin and mRNA isolated from the cells at 4-h intervals following release of the cell cycle block with mevalonic acid as described previously (10).
Library Screening-A human genomic library (Lambda FIX II vector (Stratagene)) generated from the lung fibroblast cell line W138 was screened (3 ϫ 10 6 plaques) with [ 32 P]dCTP-labeled oligonucleotides generated using the Random Primed DNA labeling kit (Boehringer Mannheim) and the human mSHMT cDNA as the template. Following plaque purification, five recombinants were obtained, two of which were identical and contained the mSHMT gene as assessed by restriction enzyme mapping and Southern analysis. The genomic insert (12 kb) was excised from the lambda vector using NotI and subcloned into pBluescript II KS ϩ vector (Stratagene) for sequence analysis. The fragment was digested further to obtain a HindIII-NotI fragment (4 kb) that contained exons 2-11 of the mSHMT gene but lacked 121 nucleotides from the 3Ј-untranslated region reported in the mSHMT cDNA (1); and a HindIII-HindIII fragment (1 kb) that contained exon 1 and 314 nucleotides of 5Ј-untranslated nucleotide sequence.
DNA Sequence Analysis-The mSHMT gene was sequenced from double-stranded template by dideoxynucleotide chain termination methods using 35 S-dATP and the Sequenase II kit (U.S. Biochemical Corp.) or using an Applied Biosciences Model 373A automated DNA sequence analyzer located at the Biotechnology Center, Cornell University. The entire SHMT gene sequence was determined and verified by sequencing both DNA strands.
Transfection of GlyA Cells-A 4.0-kb HindIII-NotI fragment containing the human mSHMT gene lacking exon 1 was subcloned into the mammalian expression vector pCep 4 (Invitrogen) which contains the hygromyocin resistance gene. The construct was linearized with NarI and introduced into GlyA cells (2 ϫ 10 7 log phase cells) by electroporation using a Gene-Pulser I (Bio-Rad) (500 V, 25 microfarads). The electroporation solution contained 20 g of the construct in 0.5 ml of 7 mM sodium phosphate buffer, pH 7.5, containing 272 mM sucrose and 1 mM MgCl 2 in 0.4-cm cuvettes. Following electroporation, the cells were incubated for 10 min on ice and cultured in six 100-mm plates per cuvette. The cells were incubated as described above in DMEM supplemented with 100 M glycine for 40 h. Stable transfectants were selected in DMEM containing 500 g/ml hygromyocin and 20 nM folinate. Resistant colonies were isolated and maintained in DMEM containing 50 g/ml hygromyocin and 20 nM folinate.
Determination of SHMT mRNA Levels by RT-PCR-Total cellular RNA was isolated from MCF-7 and CHO cells using the guanidine hydrochloride method (11) and converted to cDNA using the First Strand Synthesis kit (Clonetech). A competitive reverse transcriptasepolymerase chain reaction (RT-PCR) method, based on the mimic strategy (Clonetech), was used to measure mSHMT mRNA levels. A competitive internal standard with identical primer binding sites used to amplify 300 bp of the mSHMT cDNA was generated by amplifying 500 bp of v-erbB with two composite primers (40-mer) (5Ј-TGTTCCGGGAG-TACTCCCTGCGCAAGTGAAATCTCCTCCG-3Ј) and (5Ј-GGAACT-GTCGAGAAGTTAAGTTGAGTCCATGGGGAGCTTT-3Ј). The first 20 nucleotides of the 40-mer were complementary to mSHMT, whereas the following 20 nucleotides were complementary to v-erb-B. The 540-bp internal standard was synthesized and purified following protocols described by the manufacturer. The primers used to amplify mSHMT cDNA and the internal standard were (5Ј-TGTTCCGGGAGTACTC-CCTG-3Ј) and (5Ј-GGAACTGTCGAGAAGTT-3Ј). The primer binding sites are located in exons 8 and 9 thereby avoiding possible amplification of genomic DNA. The primers used to create the glyceraldehyde-3-phosphate dehydrogenase internal standard were (5Ј-ACCACAGTC-CATGCCATCACCGCAAGTGAAATCTCCTCCG-3Ј) and (5Ј-TCCACC-ACCCTGTTGCTGTATTGAGTCCATGGGGAGCTTT-3Ј), and the primers used to amplify glyceraldehyde-3-phosphate dehydrogenase and its internal standard were (5Ј-TCCACCACCCTGTTGCTGTA-3Ј) and (5Ј-ACCACAGTCCATGCCATCAC-3Ј). RT-PCR experiments were performed by converting total RNA (2 g) to cDNA using the first strand synthesis kit (Clonetech) and protocols suggested by the manufacturer. Mimic internal standard was added to 0.4 g of cDNA at several concentrations until a target/standard ratio of 1.0 was achieved (1 ϫ 10 Ϫ2 to 1 ϫ 10 Ϫ5 amol/reaction). Target cDNA and the internal standard were amplified with Taq polymerase using 32 P-end-labeled primers. The cycling parameters were 94°C for 45 s, 65°C for 45 s, and 72°C for 1 min for a total of 28 cycles. Internal standard and targetamplified DNA fragments were separated on a 1.8% agarose gel, transferred to Zeta Probe membrane (Bio-Rad), and quantified using a Molecular Dynamics PhosphorImaging system. The mSHMT levels were reported relative to mimic concentration required to achieve a mSHMT/ mimic ratio of 1.0. Glyceraldehyde-3-phosphate dehydrogenase mRNA levels were also quantified for each sample in control experiments.
Primer Extension-Primer extension was performed at 70°C using Tth DNA polymerase in the presence of Mn 2ϩ using protocols described by the manufacturer (Promega) with 5 ng of the primer (5Ј-GCCGC-CCAAAACAAAGAGAAGTACAGCATCGCAACTCGG-3Ј) that corresponds to bases 29 through Ϫ10 of the mSHMT gene. 5Ј-RACE products were generated, and 15 products were cloned and sequenced using a kit from Life Technologies, Inc. following protocols recommended by the manufacturer.
Determination of Mitochondrial SHMT Activity-A sensitive radioassay was used to measure relative changes in mSHMT and cSHMT enzyme activity in cell extracts from as few as 1 ϫ 10 6 cultured cells as described previously (12). The assay is based on the observation that H 4 PteGlu accelerates the SHMT-catalyzed exchange of the pro-2S proton of glycine, and 5-CHO-H 4 PteGlu inhibits this reaction (13). In a typical experiment, 5 ϫ 10 6 CHO cells in 1.0 ml of 10 mM potassium phosphate buffer, pH 7.5, containing 300 mM sucrose were disrupted by 50 strokes of a Wheaton pestle A (tight fitting), and the cytoplasmic and mitochondrial fractions were prepared as described previously (10). Lactate dehydrogenase and glutamate dehydrogenase activities were determined in each fraction to correct for mitochondrial breakage and cytoplasmic fraction contamination (14). The isolated mitochondria were lysed in 200 l of 20 mM sodium phosphate buffer, pH 7.2, 10 mM 2-mercaptoethanol, 0.5% Triton X-100. mSHMT activity was measured by diluting 40 l of the mitochondrial fraction to 500 l with 10 mM potassium phosphate buffer, pH 7.5, 10 mM 2-mercaptoethanol, and glycine such that the final glycine concentration was 1 mM with a specific activity of 2 ϫ 10 6 dpm/mol. The reaction was initiated by the addition of 100 pmol of H 4 PteGlu and incubated at 37°C for 30 -120 min. Control reactions were performed to correct for background exchange by the addition of 100 pmol of 5-CHO-H 4 PteGlu in lieu of H 4 PteGlu. The reaction was terminated by the addition of 3 ml of 50 mM HCl (4°C), and the solution was passed through a column containing 0.8 ml of Dowex 50 AG (Bio-Rad) to remove radiolabeled glycine. The column was washed with an additional 2 ml of 50 mM HCl, and the tritiated water was collected and quantified by scintillation counting. Control reactions containing 5-CHO-H 4 PteGlu exhibited less than 4% proton exchange compared with the H 4 PteGlu-catalyzed exchange reaction. All assays were performed in duplicate, and all experiments were repeated at least twice. Mitochondrial fractions were determined to be free of lactate dehydrogenase activity and were not corrected for cSHMT contamination. Protein concentrations were determined by the method of Lowry as described previously using bovine serum albumin as a standard (15).
Amino Acid Analysis-Intracellular free amino acids were isolated from cultured cells by a modification of the procedure described previously (16). Cells were cultured in 100-mm plates containing 10 ml of ␣-minimum essential medium/10% dFBS lacking glycine for 36 h. The media was removed by aspiration, and the cells were washed five times with 10 ml of phosphate-buffered saline. The cells were harvested using a cell scraper after the addition of 300 l of 5% trichloroacetic acid. Cell extracts were transferred to microcentrifuge tubes, vortexed for 30 s, and centrifuged at 10,000 ϫ g for 10 min. The trichloroacetic acid was removed from the supernatant by extracting three times with an equal volume of water-saturated diethyl ether. The aqueous solution containing the free amino acids was vacuum-dried. The free amino acids were quantified following o-phthaldialdehyde derivatization at the Cornell Biotechnology Analytical/Synthesis Facility. Intracellular serine and glycine concentrations were normalized to intracellular valine or isoleucine concentrations as internal standards.

Isolation and Organization of the mSHMT Gene-A gt11
Fix II library was screened as described under "Experimental Procedures," and five recombinants were obtained, two of which were found to be identical by restriction enzyme mapping and Southern analysis. The identical recombinants were restriction mapped, and the gene was localized to a 4.0-kb HindIII-NotI fragment and a 1.0-kb HindIII-HindIII fragment. Both fragments were sequenced. The 1-kb fragment contained exon 1 and 314 nucleotides of 5Ј-flanking sequence. The 4.0-kb fragment contained the remainder of the coding region of the mSHMT gene but lacked the terminal 121 nucleotides of the 3Ј-untranslated region present in the cDNA. The introns were sequenced, and the mSHMT gene nucleotide sequence has been deposited in the EMBL/Genbank Data Libraries (accession number U23143). Sequence analysis and restriction mapping of the remaining three clones suggested that they were not the mSHMT or cSHMT genes and are currently being investigated. The coding sequence of the gene was in agreement with the previously published cDNA sequence (1) with two exceptions. Codon 281 in the cDNA contains the nucleotide C in the number 1 position, whereas the gene contains the nucleotide T resulting in a Phe to Leu change. Both the human and rabbit cSHMT enzymes contain a Leu in this position, whereas the rabbit mitochondrial protein contains a Phe at this position. The third position of codon 293 coding for Leu in the cDNA contains a T, whereas the gene contains the nucleotide G in this position; however, both codons code for Leu.
The gene contains 10 introns and 11 exons spanning about 4.5 kb (Table I). The entire nucleotide sequence was determined, and all intron/exon splice junctions conform to the gt-ag rule (17). All introns are relatively small ranging from 86 to 891 bp. The 3Ј splice sites do not show any preference for thymidine at the Ϫ4 position as has been found for mitochondrial aspartate aminotransferase and other nuclear-encoded mitochondrial proteins (18). The gene contains a 2-fold higher occurrence of type 2 intron splice junctions than the average for mammalian genes (19), with 50% type 0 introns, 20% type 1 introns, and 30% type 2 introns.
Analysis of the Mitochondrial Leader Sequence-Previous studies have suggested that mSHMT is located in the mitochondrial matrix (20), and the primary sequence of the rabbit liver mSHMT has been determined by amino acid sequencing (21). Amino-terminal sequencing of the rabbit liver mSHMT enzyme did not yield a start methionine suggesting either that the mitochondrial import presequence had been cleaved or the enzyme may have been subjected to proteolysis at its amino terminus during purification (21). Fig. 1 shows the aminoterminal residues of the human mSHMT primary sequence, deduced from the nucleotide sequence obtained by 5Ј-RACE analysis, aligned with the rabbit liver mSHMT protein sequence. Amino-terminal residues 1-8 in the rabbit liver primary sequence align to residues 30 -37 in the human mSHMT ( Fig. 1), suggesting that residues 1-29 in the human mSHMT primary sequence represent a mitochondrial import presequence. The presequence is rich in the amino acids Arg, Leu, and Ser which are favorable for mitochondrial import. Analysis of the mSHMT presequence suggests that it can form an amphipathic ␣-helix with the positively charged Arg residues and hydrophobic residues residing on opposite sides of the helix, typical of classical mitochondrial import presequences (22). The translational initiation site contains a near consensus translation initiation sequence (23), with a G in the Ϫ3 and Ϫ6 positions. However, the mitochondrial leader peptide contains an internal methionine residue at position 22. The two methionine codons are separated by an 891-bp intron that could potentially serve as an alternate promoter as seen in the rat glucokinase gene (24). The second AUG codon is also contained within a near consensus translation initiation sequence, with  To determine if intron 1 serves as an alternate promoter, 5Ј-RACE products were generated from human MCF-7 cell mRNA. Fifteen RACE products were sequenced, and all included the initiation codon present in exon 1. If intron 1 served as an alternate promoter in these cells, primers complementary to the 3Ј-untranslated region of intron 1 and to internal cDNA sequences would be expected to generate PCR products using MCF-7 cell cDNA as a template. However, no such amplification products were detected. Intron 1 does not serve as a promoter region in human MCF-7 cells. However, these studies do not preclude the possibility that this region may serve as a promoter in other cell types.
Transfection of GlyA Cells with the mSHMT Gene-To determine if translation initiation could occur from the second ATG in the mSHMT mRNA and result in the synthesis of a functional mSHMT protein, GlyA cells were transfected with the human mSHMT gene lacking exon 1 and under the control of the cytomegalovirus immediate early enhancer/promoter in the mammalian expression vector pCep4 (Invitrogen). Five stable transfectants (GlyA-human-mSHMT) were obtained by selection in media containing hygromycin, and colonies were maintained for over 12 months in the selection media. Fig. 2 shows RT-PCR analysis of total RNA isolated from WTT2, GlyA, and GlyA-human-mSHMT cells using primers designed to amplify 300 bp of the human mSHMT cDNA as described under "Experimental Procedures." A fragment of the expected size was amplified from GlyA-human-mSHMT, WTT2, and GlyA cDNA. This suggests that CHO mSHMT mRNA contains sufficient nucleotide identity to the human mRNA to permit amplification by the human-specific primers and that the hu-man gene is correctly spliced in CHO cells. All five stable GlyA-human-mSHMT transfectants were enriched with mSHMT mRNA compared with WTT2 cells as determined by RT-PCR. The GlyA-human-mSHMT transfectants contained 10 -20 amol of human mSHMT mRNA/mg of total RNA. In comparison, MCF-7 cells contain 1 attomole mSHMT mRNA/mg total RNA.
Expression of the mSHMT gene lacking exon 1 in GlyA cells eliminated the glycine auxotrophy in all transfectants. All CHO cell lines were characterized by measuring SHMT activity and by analyzing intracellular amino acid concentrations (Table II) as described under "Experimental Procedures." Intracellular serine is elevated 9-fold in GlyA cells compared with WTT2 cells, whereas glycine is nearly undetectable in GlyA cells, consistent with the absence of mSHMT activity. GlyAhuman-mSHMT glycine levels were elevated about 10-fold compared with GlyA cells and were similar to WTT2 cells, whereas intracellular serine levels in the transfectants were decreased 50%. These changes in intracellular amino acid concentrations occurred despite only minor increases in mSHMT activity (Table II). WTT2 cells displayed 2-and 50-fold higher total and mitochondrial SHMT activities, respectively, compared with GlyA cells when assayed using the [ 3 H]glycine exchange assay. Expression of the partial human mSHMT gene lacking exon 1 in GlyA cells resulted in less than a 10% increase in total SHMT activity but nearly a 3-fold increase in mSHMT activity. These results suggest that expression of the human mSHMT gene lacking exon 1 in GlyA cells is sufficient for mSHMT expression and mitochondrial import. These results also suggest that only a 2-3-fold increase in mSHMT activity is required to alleviate the glycine auxotrophy of GlyA and that mSHMT activity can be reduced greater than 90% without compromising intracellular glycine concentrations. It is also apparent from whole cell SHMT activity measurements that expression of the partial human mSHMT gene did not result in measurable increased SHMT activity in the cytoplasm. Expression of mSHMT in the cytoplasm would not be expected to overcome the glycine auxotrophy as overexpression of cSHMT in GlyA cells does not overcome the glycine requirement. 2 Analysis of the Promoter Regions-The transcriptional initiation sites of the mSHMT gene were determined by primer extension analysis of total cellular RNA isolated from MCF-7 cells. Multiple transcription initiation sites of equal intensities were observed (Fig. 3) which is consistent with the absence of TATA or CAAT-like sequences (Fig. 4). The 5Ј promoter region of the mSHMT gene contains consensus DNA recognition sequence for Sp-1 (Ϫ290 to Ϫ284), AP-2 (Ϫ249 to Ϫ241), HC3 (Ϫ150 to Ϫ144), and PEA3 (Ϫ239 to Ϫ245). A zeste-white sequence (Ϫ184 to Ϫ190) is present and may account for some of the observed changes in mSHMT expression during development. Additionally, a CCCTCCC motif (Ϫ128) is present that is common to a number of genes that are expressed predominantly in the liver (25).
The sequence 5Ј to the second translational initiation site (intron 1 region) also does not contain any TATA or CAAT-like sequences. A consensus core sequence for CTF/NF1 is located on the antisense strand (757-762) (20). Consensus sequence motifs were also found for PuF-1, PEA3, Sp-1, UBP1, octB2, Adh1, TRE, GRE, FSE2, HC3, and IE1. The possible role of these sequences in regulating the endogenous expression of the mSHMT is currently under investigation.
Cell Cycle Regulation of mSHMT-The expression of many genes involved in DNA synthesis, including those encoding folate-dependent enzymes such as dihydrofolate reductase and thymidylate synthase, is enhanced upon entry into the S phase of the mammalian cell cycle. Both dihydrofolate reductase and thymidylate synthase contain a cis-acting element that binds the transcription factor E2F, and E2F binding is sufficient for growth-regulated promoter activity at the G 1 /S phase boundary (26). MCF-7 mSHMT mRNA levels were measured by RT-PCR through the cell cycle to determine if the mSHMT gene is co-regulated with other genes involved in DNA synthesis. Cells were synchronized with Lovastatin, an inhibitor of 3-hydroxy-3-methylglutaryl-coenzyme A reductase. Lovastatin blocks the cell cycle reversibly in G 1 , and the block can be released by mevalonic acid, the product of the reductase reaction. Lovastatin is preferable to other cell cycle blocking agents as DNA synthesis is inhibited over 96% and cells remain synchronized for at least three cycles after release of the block. The percentage of cells in S-phase was determined by [ 3 H]thymidine incorporation and by histone H4 expression as described (10). No variations in mitochondrial SHMT or control glyceraldehyde-3-phosphate dehydrogenase mRNA levels were observed (Fig.  5) suggesting that mSHMT mRNA levels are not cell cycle-regulated.
Nutrient Control of SHMT Expression in MCF-7 Cells-A previous study showed that SHMT activity in chicks was increased in folate deficiency (7), suggesting that folate status and perhaps other nutrients may play a role in mSHMT expression. The hydroxymethyl group of serine is incorporated into purines, methionine, and thymidylate, and the availability of these products of one-carbon metabolism may regulate mSHMT expression. MCF-7 cells were cultured in DMEM, 10% dFBS with and without purines, thymidine, methionine, glycine, or serine as described under "Experimental Procedures." The omission of purines, thymidine, methionine, glycine, serine, or combinations thereof did not change SHMT mRNA levels after maintenance in this media for 72 h (data not shown) as determined by competitive RT-PCR. Likewise, mSHMT mRNA levels were unchanged in folate-deficient MCF-7 cells passaged for over 7 weeks in media lacking folic acid. These results suggest that mSHMT gene transcription is not influenced by nutrient status associated with folate-mediated one-carbon metabolism in MCF-7 cells.

DISCUSSION
The differential metabolic role of the two SHMT isozymes in folate-dependent anabolic pathways is not understood, but mSHMT activity has been suggested to be the primary source of intracellular glycine (2,9). In order to understand the differential metabolic roles of the SHMT isozymes, we have cloned the mSHMT gene and initiated studies to determine the factors and mechanisms that cells use to regulate the endogenous expression of the gene. Recent studies have demonstrated that elimination of mitochondrial folate pools in CHO cells results in a glycine auxotrophy (27,28) and that CHO cell glycine auxotrophs lacking mSHMT activity have an elevated intracellular serine concentration but are not able to accumulate intracellular glycine (9). These data suggest that mSHMT is primarily responsible for glycine synthesis and that cSHMT is not effective in catalyzing the formation of glycine from serine even in the presence of elevated serine concentrations (9). However, the possibility that the cell lines had additional mutations that may have been responsible for the observed metabolic disturbances could not be eliminated. Our results confirm that GlyA cells also contain very low levels of intracellular glycine despite the high accumulation of intracellular serine and the presence of cSHMT activity. In addition, we have shown that increasing the mSHMT activity in GlyA cells overcomes the glycine auxotrophy, thereby demonstrating that mSHMT deficiency alone is responsible for the glycine auxotrophy associated with GlyA cells.
Expression of the partial human mSHMT gene in GlyA cells resulted in only modest increases in mSHMT activity, representing less than 5% of total WTT2 cell mSHMT activity, despite expression of high levels of human mSHMT mRNA in the GlyA-human-mSHMT cells. We conclude that exon 1 of the mSHMT gene is not essential for mSHMT activity or mitochondrial import. These data also suggest that mSHMT activity can be inhibited greater than 95% in CHO cells without inducing a  3. Transcriptional start sites determined by primer extension analysis. Total RNA was isolated from MCF-7 cells, and the transcriptional start sites of the mSHMT gene were determined by primer extension at 72°C. The 1st lane represents the primer extension products, and the 2nd to 5th lanes are the mSHMT gene sequencing products. All reactions were performed with the same oligonucleotide primer.
glycine auxotrophy or affecting intracellular glycine levels, although serine levels were still elevated.
It has been suggested that the mitochondrial and cytoplasmic SHMT isozymes function cooperatively in shuttling onecarbon units between the cytoplasm and mitochondria analogous to the shuttling of reducing equivalents that occurs in the malate-aspartate shuttle (29). Studies of the regulatory regions of the aspartate aminotransferase genes demonstrated that the mitochondrial and cytoplasmic isozymes do not share common promoter elements (30), and the two genes are regulated independently. Previous studies have demonstrated that the relative ratio of SHMT in the cytoplasm and mitochondria varies among cell and tissues, suggesting that the SHMT genes are also regulated independently. While mSHMT is found in all cell types, it is enriched in certain tissues, including the kidney and liver. Analysis of the human mSHMT 5Ј promoter region suggests that it is a housekeeping gene with multiple transcriptional initiation sites, but it also contains a CCCTCCC element that is present in many genes that are expressed predominantly in the liver (24) including the phenylalanine hydroxylase gene. It is also of interest that the 5Ј promoter region contains a zeste-white element that may be responsible for the observed developmental variations in SHMT activity and serine metabolism (3). We have also demonstrated that mSHMT message levels do not change throughout the cell cycle in MCF-7 cells, consistent with similar observation for the yeast mSHMT message levels (31). Additionally, no evidence for changes in mSHMT mRNA levels in MCF-7 cells were observed during folate, purine, thymidine, or methionine deprivation. Continuing studies will elucidate the contribution of the mSHMT promoter elements to mSHMT expression and their influence on glycine and folic acid metabolism.
Intron 1 appears to contain consensus regulatory elements capable of expressing the mSHMT gene. Analysis of the mitochondrial import peptide suggests that exon 1 contains most of the hydrophobic and basic amino acid residues that may be required for efficient mitochondrial protein import but are not essential. Expression of the human mSHMT lacking exon 1 did not result in measurable increases in SHMT activity in the cytoplasm. The low level of mSHMT activity resulting from the expression of the human mSHMT gene lacking exon 1 in the GlyA transfectants is most probably due to either inefficient translation from the translation initiation site located in exon 2 or rapid turnover of the mSHMT protein in the cytoplasm prior to mitochondrial import due to its shortened import presequence. In light of these observations, we are currently investigating the stability of the mSHMT protein in the cytoplasm and the amino acid residues that are essential for mitochondrial import, and we are determining whether the putative promoter elements present in intron 1 are capable of expressing reporter genes in human cells.