SHMT1 and SHMT2 Are Functionally Redundant in Nuclear De novo Thymidylate Biosynthesis

The three enzymes that constitute the de novo thymidylate synthesis pathway in mammals, cytoplasmic serine hydroxymethyltransferase (SHMT1), thymidylate synthase (TYMS) and dihydrofolate reductase (DHFR) undergo sumoylation and nuclear import during S-phase. In this study, we demonstrate that purified intact mouse liver nuclei convert dUMP to dTMP in the presence of NADPH and serine. Neither nuclear extracts nor intact nuclei exposed to aminomethylphosphonate, a SHMT inhibitor, exhibit thymidylate synthesis activity. Nuclei isolated from Shmt1−/− mouse livers retained 25% of thymidylate synthesis activity exhibited by nuclei isolated from wild type mice. This residual activity was due to the presence of a cytoplasmic/nuclear isozyme of SHMT encoded by Shmt2. Shmt2 is shown to encode two transcripts, one which encodes a protein that localizes exclusively to the mitochondria (SHMT2), and a second transcript that lacks exon 1 and encodes a protein that localizes to the cytoplasm and nucleus during S-phase (SHMT2α). The ability of Shmt2 to encode a cytoplasmic isozyme of SHMT may account for the viability of Shmt1−/− mice and provide redundancy that permitted the expansion of the human SHMT1 L474F polymorphism that impairs SHMT1 sumoylation and nuclear translocation.


Introduction
Tetrahydrofolate (THF) is a metabolic cofactor that carries and activates single carbons for the synthesis of nucleotides and methionine [1]. Folate-mediated one-carbon metabolism is compartmentalized in the mitochondria and cytoplasm of eukaryotic cells (Figure 1). In the cytoplasm, this metabolic network is required for the biosynthesis of purines, thymidylate, and the remethylation of homocysteine to form methionine. Serine is a major source of one-carbon units for this network through its reversible and tetrahydrofolate-dependent conversion to glycine and methyleneTHF catalyzed by serine hydroxymethyltransferase (SHMT). There are cytoplasmic and mitochondrial SHMT isozymes. SHMT1 encodes the cytoplasmic isozyme (SHMT1) and SHMT2 encodes the mitochondrial isozyme (SHMT2) [2,3,4]. Mitochondrial one-carbon metabolism generates one-carbons from serine through the activity of SHMT2, and the one-carbon is oxidized and exported to the cytoplasm as formate, supporting cytoplasmic one-carbon metabolism [5]. The SHMT1 enzyme generates methyleneTHF for thymidylate and methionine biosynthesis, but isotope tracer studies indicate that SHMT1 preferentially partitions methyleneTHF to thymidylate biosynthesis [6].
The de novo thymidylate biosynthesis pathway requires three enzymes: thymidylate synthase (TYMS), dihydrofolate reductase (DHFR), and SHMT1. MethyleneTHF generated by SHMT is the one-carbon donor for the TYMS catalyzed conversion of dUMP to dTMP generating dihydrofolate (DHF). DHFR catalyzes the NADPH-dependent reduction of DHF to regenerate THF for subsequent cycles of de novo thymidylate synthesis.
Recently, the enzymes that constitute the thymidylate synthesis cycle were shown to undergo post-translational modification by the small ubiquitin-like modifier (SUMO) and nuclear translocation during S and G2/M phases [7,8]. Although the synthesis of thymidylate in the nucleus has never been demonstrated, others have found folate cofactors present in liver nuclei [9], and multienzyme complexes containing ribonucleotide reductase and thymidylate synthase have been isolated from nuclear extracts [10]. In this study, intact nuclei are shown to catalyze the formation of dTMP from dUMP, which accounts for the results of stable isotope studies that indicate SHMT preferentially partitions methyleneTHF to thymidylate biosynthesis. Furthermore, both Shmt1 and Shmt2 are shown to contribute to nuclear de novo thymidylate biosynthesis.

Results
Shmt1 and Shmt2 contribute to nuclear dTMP biosynthesis. The ability of purified nuclei to catalyze the formation of tritiated dTMP from unlabeled dUMP, NADPH and [2,3-3 H]-L-serine in vitro was investigated ( Figure 2). Intact nuclei isolated from the livers of wild type mice were capable of generating tritiated dTMP, demonstrating that folate-dependent nuclear dTMP synthesis occurs in liver. The addition of the SHMT inhibitor and amino acid analog, aminomethylphosphonate, to the reaction mixture inhibited dTMP synthesis by greater than 95%, demonstrating the essentiality of the SHMT reaction in generating folate-activated one-carbons from serine for dTMP synthesis in nuclei. 5formyltetrahydrofolate pentaglutamate, a natural inhibitor of SHMT [11], did not inhibit nuclear dTMP biosynthesis but may not have been able to traverse the nuclear membrane. Disruption of nuclei by sonication eliminated all dTMP synthesis activity, indicating that cytoplasmic contamination was not responsible for the observed dTMP synthesis activity in nuclei, and suggesting that maintenance of nuclear architecture is essential for nuclear dTMP synthesis.
Surprisingly, intact nuclei isolated from the liver of Shmt1 2/2 mice were capable of generating tritiated dTMP at approximately 25% of the level observed from nuclei isolated from Shmt1 +/+ mice ( Figure 2). Aminomethylphosphonate inhibited dTMP synthesis in nuclei generated from Shmt1 2/2 mice, indicating that a second SHMT activity is present in nuclei which is not derived from Shmt1. PCR was used to confirm the genotype of the purified nuclei and immunoblotting was performed to verify that the nuclei lacked Shmt1 ( Figure 3A). Furthermore, the nuclei were shown to be free of cytosolic and mitochondrial contamination ( Figure 3B). DHFR and TYMS protein were observed in purified nuclei isolated from the livers of wild type and Shmt1 2/2 mice, and surprisingly, SHMT2 protein was also observed in nuclei isolated from the livers of wild type and Shmt1 2/2 mice ( Figure 3A).
Previous evidence suggests that Shmt2 contains two translation initiation sites. Expression of a SHMT2 gene fragment lacking exon 1 in glyA CHO cells, which lack SHMT2 activity, rescued the Figure 1. Compartmentation of folate-mediated one-carbon metabolism in the cytoplasm, mitochondria and nucleus. One-carbon metabolism in the cytoplasm is required for the de novo synthesis of purines and thymidylate, and for the remethylation of homocysteine to methionine. One-carbon metabolism in mitochondria generates one-carbon units for cytoplasmic one-carbon metabolism by generating formate from serine and glycine. One-carbon metabolism in the nucleus synthesizes dTMP from dUMP and serine. SHMT2, mitochondrial serine hydroxymethyltransferase; SHMT1, cytoplasmic serine hydroxymethyltransferase; TYMS, thymidylate synthase; DHFR, dihydrofolate reductase; THF, tetrahydrofolate. doi:10.1371/journal.pone.0005839.g001 glycine auxotrophy [4]. Exon 1 encodes the first translation initiation start site and most of the peptide sequence required for efficient import into mitochondria. Shmt2 contains a potential second translation initiation codon within exon 2, and translation initiation from this site generated a protein capable of import into mitochondria, albeit at lower efficiency ( Figure 4) [4]. To determine if two transcripts were generated from Shmt2, one containing exon 1 and one lacking exon 1, the mouse EST database was probed for Shmt2 cDNA sequences that contained nucleotide sequence from the Shmt2 intron 1/exon 2 boundary. An EST (AA793217) that lacked exon 1 but contained 166 nucleotides from the 39 end of intron 1 at its 59 end was identified. Similarly, the human EST database was probed and two SHMT2 transcripts containing 131 nucleotides from the 39 end of intron 1 were identified (DB184899 and DA597551). These data indicate that SHMT2 encodes two transcripts and an alternative promoter within intron 1.

Subcellular localization of SHMT2 gene products
The cellular localization of the two SHMT2 isoforms was determined by expression of SHMT-yellow fluorescent protein (YFP) and red fluorescent protein (RFP) fusion proteins. Cyan fluorescent protein fused to a mitochondrial leader sequence (CFPmito) at its amino terminus and Draq5 DNA binding dye were used as mitochondrial and nuclear markers respectively. Confocal microscopy revealed that the SHMT protein expressed from the SHMT2 transcript containing exon 1 (referred to as SHMT2-YFP) and CFP-mito protein co-localized to mitochondria. However, SHMT protein expressed from the SHMT2 transcript lacking exon 1 (referred to as SHMT2a-RFP) localized predominantly to the cytoplasm and nucleus ( Figure 5A). The nuclear localization of the SHMT2a-RFP fusion protein exhibited similar cell cycle dependence as observed for SHMT1 [8]. The SHMT2a-RFP fusion protein localized exclusively to the cytoplasm in G1 phase whereas in S and G2/M phases it localized to both the cytoplasm and nucleus ( Figure 5B).
Both SHMT2 and SHMT2a rescue the glycine auxotrophy in GlyA Chinese Hamster Ovary cells Previously, transfection of a human SHMT2 gene fragment that lacked exon 1 into CHO GlyA cells, which lack SHMT2 activity in mitochondria and exhibit a glycine auxotrophy, resulted in very low levels of SHMT2 activity in mitochondria which was sufficient to rescue the glycine auxotrophy [4]. In this study, the ability of SHMT2a to rescue the glycine auxotrophy in GlyA cells was determined. Twelve to fifteen stable transfectants were selected for G418 resistance from GlyA cells electroporated with either an empty TagRFP-N plasmid, a plasmid expressing the cDNA encoding SHMT2, or a plasmid expressing the cDNA encoding SHMT2a. All plasmids were driven by the immediate early promoter of cytomegalovirus. Stable transfectants containing the RFP empty vector did not rescue glycine auxotrophy and no revertants were observed in the 12 colonies screened. However, the glycine auxotrophy was rescued in all selected cell lines expressing either the SHMT2-RFP or the SHMT2a-RFP fusion protein ( Figure 6). This data confirms that SHMT2a localizes to mitochondria and can rescue mitochondrial one-carbon metabolism.

Expression of SHMT2 and SHMT2a in mouse tissues
The relative levels of each SHMT2 isoform was determined in mouse liver and kidney by western blot analyses ( Figure 7). A single immunoreactive band that migrated at 53 kD was present in nuclear extracts isolated from mouse liver when probed with an antibody against SHMT2. This molecular mass corresponds to the predicted size of the SHMT2a protein expressed from the Shmt2 transcript lacking exon 1. Purified liver mitochondrial extracts exhibited three immunoreactive bands. The upper band, which migrated with a molecular mass of approximately 56 kDa, is consistent with the predicted mass of the full length SHMT2 preprocessed protein. A lower band, which migrated with a molecular mass of approximately 50 kDa, corresponds to the processed SHMT2 protein lacking its leader sequence as present in mitochondria. These 3 bands were also observed in whole liver extract, and the blot indicates that the predominant SHMT2 isoform is the processed mitochondrial form. However, a substantial amount of the SHMT2a protein was also present in liver. In kidney, only the pre-processed SHMT2 and SHMT2a form of the enzyme were observed.

A single nucleotide polymorphism in SHMT1 impairs nuclear localization
Previously, we demonstrated that a common polymorphism in SHMT1, L474F impairs SUMO modification of the SHMT1 protein [8]. To determine if this polymorphism impairs nuclear localization and to explore the potential for the redundancy between SHMT1 and SHMT2a, MEFS derived from Shmt1 2/2 mice were electroporated with plasmids that express the wild-type human SHMT1-YFP fusion protein, the human L474F SHMT1-YFP fusion protein or a mutated K38R/K39R SHMT1-YFP fusion protein, which lacks a SUMO-modification site [8]. MEFs isolated from Shmt1 2/2 mice were used to eliminate the possibility that endogenous mouse SHMT1 protein could oligomerize with the human SHMT1 fusion proteins thereby allowing nuclear import. The SHMT1-YFP fusion protein localized to both the cytoplasm and nucleus in S-phase, whereas the K38R/K39R SHMT1-YFP fusion protein localized exclusively to the cytoplasm in S-phase ( Figure 8). The L474F SHMT1-YFP fusion protein was found in both the cytoplasm and nucleus at S-phase, but in contrast to the SHMT1-YFP protein, its localization was primarily cytoplasmic. These data demonstrate that the SHMT1 L474 polymorphism impairs nuclear localization. The potential for functional redundancy between SHMT1 and SHMT2a in nuclear folate metabolism may have been the permissive factor that allowed this mutation to expand in human populations.

Discussion
The results from this study demonstrate directly the existence of nuclear thymidylate biosynthesis. Pardee and co-workers [10,12,13] first proposed the concept of nuclear nucleotide biosynthesis and put forward the concept of a nuclear multienzyme complex termed the replitase, which synthesized nucleotides at the replication fork during DNA synthesis. Both DHFR and TYMS activities were found in these purified complexes isolated from nuclei, but no direct evidence for nuclear nucleotide biosynthesis was reported. The concept of nuclear folate metabolism is also supported by early studies that demonstrated 10% of the total cellular hepatic folate was present within the . Identification of SHMT2 in purified nuclei. Panel A) PCR was used to confirm the genotype of nuclei isolated from SHMT1 +/+ and SHMT1 2/2 mice. SHMT1 2/2 mice were generated through deletion of exon 7 which encodes the PLP binding site. SHMT1 +/+ mice exhibit a 740-bp PCR product whereas the SHMT1 2/2 mice exhibit a 460-bp PCR product. Western blotting confirmed the presence of TYMS and DHFR in nuclei of both genotypes. SHMT1 was present in purified liver nuclei from SHMT1 +/+ mice, but absent in liver nuclei from SHMT1 2/2 mice. SHMT2 was present in nuclei from both genotypes. Panel B) Western blots confirm the purity of the isolated liver nuclei. The control lane represents purified cytosol from NIH/3T3 cells for both the Lamin A (nuclear marker) and GAPDH (cytoplasmic marker). The control for the COX IV immunoblot represents a purified mitochondrial fraction from NIH/3T3 cells. The absence of GAPDH and COX IV in nuclear extracts indicated that no cytosolic or mitochondrial contamination was present in the nuclear thymidylate biosynthetic assays. doi:10.1371/journal.pone.0005839.g003 nucleus [9]. Previous studies have shown that TYMS localizes to the nucleus at S-phase [8,14], and recently SHMT1, DHFR, and TYMS proteins have all been shown to be SUMOylated, providing a mechanism by which all the enzymes required for de novo thymidylate biosynthesis can be localized to the nucleus [7,8]. Recent evidence indicates that nuclear folate-dependent nucleotide biosynthesis is limited to thymidylate biosynthesis, as the multi-enzyme complex responsible for de novo purine biosynthesis, the purineosome, localizes to the cytoplasm [15].
In this study, we demonstrated that one-carbon units generated from serine are used for nuclear de novo thymidylate biosynthesis, and that the three enzymes that participate in de novo thymidylate pathway are present in nuclei. Compartmentalization of these enzymes to the nucleus accounts for previous isotope tracer studies that demonstrate SHMT1 preferentially shuttles one-carbon units towards thymidylate biosynthesis [6].
The data also demonstrate functional redundancy between Shmt1 and Shmt2 in nuclear folate metabolism. The observation that Shmt2 (and human SHMT2) encodes a nuclear/cytoplasmic isozyme, SHMT2a, may account for the unexpected viability of Shmt1 2/2 mice, and potentially the emergence in human populations of the L474F SHMT1 polymorphism, which impairs nuclear SHMT1 import. However, SHMT2a cannot fully replace SHMT1 function as Shmt1 2/2 mice exhibit elevated uracil content in DNA [16]. The S-phase and G2/M phase dependence of SHMT2a nuclear localization suggests that it is available for de novo thymidylate biosynthesis during DNA synthesis and repair as observed for SHMT1 [8]. The mechanism by which SHMT2a localizes to the nucleus is currently under investigation.
To our knowledge, SHMT2 is the only gene identified that encodes proteins that can localize to the cytoplasm, nucleus and mitochondria. Previously, zebrafish SHMT2 protein was shown to localize exclusively to mitochondria [17]. The zebrafish Shmt2 gene lacks the second translation initiation site in exon 2 and thus can only encode a mitochondrial isozyme. Mammalian SHMT2 genes, with the exception of the lagomorph gene, encode a . The mouse Shmt2 encodes two transcripts and contains two translation initiation sites. The SHMT2 transcript is denoted by the color red. The first nucleotide of the translation initiation site contained in exon 1 is numbered as +1. The first translation initiation site in exon 1, which encodes for the mitochondrial leader sequence, is boxed in blue. The mitochondrial leader sequence is cleaved co-translationally in the mitochondria denoted by the scissors [2]. The second translation initiation site, present in exon 2, is boxed in blue. A transcription start site in intron 1 deduced from a mouse liver EST (AA793217) is denoted by the arrow and starts at 2166 from the second ORF. doi:10.1371/journal.pone.0005839.g004 translation initiation codon in exon 2 and therefore likely encode the cytoplasmic/nuclear SHMT2a isozyme. These data also provide evidence for tissue-specific expression of the SHMT2 isoforms. The SHMT2/SHMT2a ratio differs between liver and kidney, with SHMT2 being the predominant form in liver, whereas only the SHMT2a and the SHMT2 unprocessed protein precursor were observed in kidney. Mitochondria are a primary source of onecarbon units for cytoplasmic one-carbon metabolism through the activity of SHMT2, and the SHMT2a protein lacks most of the mitochondrial targeting sequence and may not be imported into mitochondria efficiently [4]. Therefore, the lack of processed SHMT2 protein in kidney mitochondria may account for previous findings that glycine catabolism through the mitochondrial glycine cleavage system is a major source of one-carbon units in kidney [18]. The molecular mechanisms underlying the differential promoter usage in SHMT2 that results in the synthesis of the SHMT2 isoforms is currently under investigation.
Compartmentation of SHMT2a and the other de novo thymidylate biosynthetic pathway enzymes in the nucleus may allow for folate dependent dTTP biosynthesis directly at the replication fork. It is likely that nuclear thymidylate synthesis requires the formation of an enzyme complex, as sonicated nuclei were not capable of dTMP biosynthesis. It has been shown that the processivity factor PCNA interacts with SHMT1 in both C. elegans and HeLa cDNA yeast two hybrids [8,19]. Additional studies are required to determine why thymidylate biosynthesis, unlike purine nucleotide biosynthesis [15], occurs in the nucleus and if a thymidylate-specific replitase-like complex exists that enables the de novo dTMP synthesis pathway to function at the replication fork and prevent uracil misincorporation into DNA.

Nuclear thymidylate biosynthesis assay
The generation and characterization of SHMT1 2/2 mice has been described previously [16]. All study protocols were approved by the Institutional Animal Care and Use Committee of Cornell University and conform to the NIH Guide for the Care and Use of Laboratory Animals. Twelve livers were isolated from Shmt1 +/+ or Shmt1 2/2 mice on a 129SvEv background and placed immediately in cold phosphate buffered saline at 5uC. Liver extracts from six age-matched males and females were combined and used for each genotype group. Nuclei were prepared using an iodixanol gradient as previously described [20]. Purified nuclei were suspended in 500 mL of nuclear assay buffer containing 5 mM NADPH (Sigma), 100 mM b-mercaptoethanol, 25 mM HEPES, pH 7.5, 50 mM Sucrose, 5 mM MgCl 2 , 25 mM KCl, and 1 mM dUMP (Sigma) and quantified using a hemocytometer. 125 mL of assay buffer containing suspended nuclei were aliquoted into four 1.5 ml plastic tubes and 8 mCi of [2,3-3 H]-L-Serine (Moravek Biochemicals) was added to each tube. The assay was conducted under 4 different experimental conditions: Experiment 1, nuclei that were lysed with sonication (Branson Sonifier 150) at 5uC using two 10 sec pulses at 10 watts separated by a 10 sec resting interval; experiment 2, intact nuclei; experiment 3, intact nuclei with 5formyltetrahydrofolate pentaglutamate (Schirks Laboratory) added to at a final concentration of 200 mM; experiment 4, intact nuclei with aminomethylphosphonate added to a final concentration of 100 mM (Sigma). Reactions were incubated for 12 h at 37uC with shaking at 300 rpm. Nuclei were pelleted by centrifugation at 2000 rpm for 5 min and the supernatant was collected and analyzed for radiolabeled thymidylate by high performance liquid chromatography (HPLC). Sample preparation and HPLC was performed as previously described [21,22]. Fractions were collected and tritium quantified with a scintillation counter. The retention times of [2,3-3 H]-L-serine (9 min) and 3 H-thymidine (17 min) (Moravek Biochemicals) were verified prior to separation of the reaction mixtures. All experiments were performed in duplicate.

Genotyping and immunoblotting
After completion of the nuclear thymidylate biosynthesis reactions, pelleted nuclei were genotyped and analyzed by western blots. DNA was isolated using the DNeasy Blood and Tissue Kit (Qiagen) per manufacturer's protocol and genotyping as previously described to confirm the Shmt1 genotype [16]. For western  . SHMT2 and SHMT2a are present in mouse tissues. Immunoblots using a sheep-anti-human SHMT2 antibody revealed the presence of three immunoreactive bands. Protein masses were determined by the migration distance and relative mobility of standards. Purified liver nuclei, whole liver and whole kidney extracts contained an immunoreactive band at ,53 kDa, the predicted mass of SHMT2a. Purified liver mitochondria, whole liver and whole kidney extracts contained a band at ,56 kDa, the predicted mass of the SHMT2 pre-processed protein. Purified liver mitochondria and whole liver, but not whole kidney extracts contained a band at ,50 kDa, the predicted mass of the SHMT2 processed protein. The nuclear fraction was free of cytosolic and mitochondrial contamination as shown by the GAPDH and COX IV immunoblots. doi:10.1371/journal.pone.0005839.g007 blots, nuclei were disrupted by boiling in SDS-PAGE loading buffer for 10 min and protein concentrations quantified by the Lowry-Bensadoun assay [23]. 20 mg of protein were loaded into each lane of a 10% SDS-PAGE gel [23]. Protein transfers, western blotting and DHFR and TYMS detection were performed as previously described [7]. SHMT1 and SHMT2 detection were completed as previously described [24,25]. Equal loading and/or purity of nuclear fractions was confirmed through the detection of GAPDH using a-GAPDH (Novus Biologicals, 1:20000 dilution), Lamin A using a-Lamin A (Santa Cruz Biotechnology, 1:500 dilution), and COX IV using a-COX IV (Abcam, 1:5000 dilution). For Lamin A detection, goat anti-rabbit IgG-horseradish peroxidase-conjugated secondary (Pierce) was used at a1:20000 dilution. For COX IV and GAPDH detection, a 1:10000 dilution of goat anti-mouse IgG-horseradish peroxidase-conjugated secondary (Pierce) was used.
The sizes of the SHMT2a isoforms were determined using the Precision Plus Protein All Blue (BIORAD) and Kaleidoscope Prestained Standard (BIORAD). Processed mitochondrial SHMT2 migrated with the Precision Plus Protein All Blue 50 kDa marker. To determine the size of the bands corresponding to preprocessed SHMT2 and SHMT2a proteins, a standard curve was generated by measuring the migration distances of the protein standards. These distances were plotted as a function of log 10 of the molecular mass standards, and the molecular mass of preprocessed SHMT2 and SHMT2a were determined using the equation MM = 21.4D+137.9; R 2 = 0.98, where MM is molecular mass in kDa and D is migration distance in mm. The migration distance of the largest molecular mass band corresponding to preprocessed SHMT2 migrated 58 mm and SHMT2a migrated 61 mm, corresponding to approximately 56 kDa and 53 kDa respectively.

Identification of Shmt2a expression
The nucleotide sequences of the human SHMT2 and mouse shmt2 intron 1/exon 2 boundaries were BLASTED against both the human and mouse EST databases to identify alternative SHMT2 and Shmt2 transcripts lacking sequence encoded by exon 1 but containing intron 1 sequence within the 59-untranslated region. To identify shmt2 protein products expressed from alternative shmt2 transcripts, liver and kidney were isolated from SHMT1 +/+ mice on a C57Bl/6 background. The tissues were washed with phosphate-buffered saline and proteins solubilized using the Mammalian Protein Extraction Reagent (Pierce) containing 10 mM b-mercaptoethanol and a 1:100 dilution of mammalian protease inhibitor (Sigma). Immunoblotting was performed as described above.

Subcellular localization of SHMT2 and SHMT2a
The human SHMT2 full length open reading frame was purchased from Open Biosystems. The cDNAs for the full length transcript (encoding SHMT2), and the transcript lacking exon 1 but containing 131 nucleotides of intron 1 in the 59UTR (encoding SHMT2a) were amplified by PCR amplified and cloned into PhiYFP-N (Evrogen) and TagRFP-N (Evrogen) vectors respectively. The forward primer used to amplify the SHMT2 transcript was 59-ATATCTCGAGATGCTGTACTTCTCTTTGTT-39. The forward primer to amplify the SHMT2a transcript was 59-ATATCTCGAGGTATGGCCATTCGGGCTCAGCAC-39. The underlined sequence signifies a XhoI (Promega) restriction site. The same reverse primer was used for both amplifications: 59-ATATAAGCTTCTAATGCTCATCAAAACCAG-39. The underlined sequence denotes a HindIII (Promega) restriction site. The PCR conditions for amplification of both the SHMT2 and SHMT2a transcripts were as follows: 95uC for 45 s, 52uC for 45 s, and 72uC for 2 min. A vector containing a mitochondrial marker, pTagCFP-mito (Evrogen) was used for control transfections to identify mitochondria. Plasmids were transfected into HeLa cells using Lipofectamine 2000 (Invitrogen) per manufacturer's protocol. The DNA binding dye, Draq5 (Biostatus Limited) was used to visualize nuclei. For cell cycle analysis, HeLa cells were treated with 1 mM hydroxyurea, 30 mM Lovastatin or 60 ng/ml nocodazole to block at S, G 1 and G2/M phase Figure 8. The L474F polymorphism in SHMT1 impairs nuclear localization. SHMT1 2/2 MEF cells were transfected with cDNAs encoding the YFP-empty vector, YFP-SHMT1 wild-type, YFP-SHMT1-K38R/K39R, and YFP-SHMT1-L474F. S-phase blocked transfectants showed that YFP-SHMT1 is greatly increased in the nucleus in S-phase. Empty vector control and K38R/K39R mutation eliminated nuclear localization, and the L474F polymorphism inhibited nuclear localization. doi:10.1371/journal.pone.0005839.g008 respectively as previously described [8]. Confocal fluorescence microscopy (Leica TCS SP2 system) was used to image all cells at the Cornell Microscope and Imaging Facility.

Rescue of the glycine auxotrophy in GlyA cells
GlyA cells were cultured in dDMEM supplemented with 200 mM glycine and 20 nM leucovorin. TagRFP-N-SHMT2 and TagRFP-N-SHMT2a and TagRFP-N vectors were linearized using NheI (Promega) and electroporated into GlyA cells as previously described [26]. Following electroporation, cells were plated on BD Primaria 100620 mm culture dishes (BD Bioscience) containing 10 ml of dDMEM medium, supplemented with 20 nM leucovorin, 200 mM glycine and 200 mg/ml Geneticin (Gibco). 12-15 colonies from each transfection were isolated and plated in 6-well plates (Corning) containing 3 ml dDMEM medium supplemented with glycine, leucovorin, and Geneticin. Cell growth assays were completed in media with and without glycine as previously reported [27].

Site-directed mutagenesis and subcellular localization of SHMT1
The SHMT1 cDNA [28] was used as a template to generate SHMT1-YFP fusion proteins. The SHMT1 cDNA was amplified for cloning into PhiYFP-N by PCR. The forward and reverse primers used were 59-ATATCTCGAGATGACGATGCCAGT-CAAC-39 and 59-ATATAAGCTTTTAGAAGTCAGGCAGG-CC-39 respectively. The PCR conditions were as follows: 95uC for 45 s, 55uC for 45 s, and 72uC for 2 min. The underlined regions depict XhoI and HindIII restriction sites respectively. Coding mutations K38R/K39R and L474F mutations were made as previously described [8]. PhiYFP-N vectors containing the SHMT1 cDNA, the SHMT1 cDNA containing a K38R/K39R mutation, and a SHMT1 cDNA containing the L474F mutation were transfected into Shmt1 2/2 MEFs using electroporation protocols described above. Cells were blocked at S-phase by exposure to 1 mM hydroxyurea (Sigma) for 24 h. Confocal fluorescence microscopy (Leica TCS SP2 system) was used to image the SHMT1-YFP fusion proteins at the Cornell Microscope and Imaging Facility.