Nuclear Localization of de Novo Thymidylate Biosynthesis Pathway Is Required to Prevent Uracil Accumulation in DNA*

Background: S phase nuclei contain the thymidylate synthesis pathway. Results: Mice overexpressing a Shmt1 transgene exhibit elevated expression of SHMT1 and TYMS, impaired nuclear localization of the thymidylate biosynthesis pathway, and elevated uracil in DNA. Conclusion: SHMT1 and TYMS localization to the nucleus is essential to prevent uracil accumulation in DNA. Significance: SHMT1-mediated nuclear de novo thymidylate synthesis is critical for maintaining DNA integrity. Uracil accumulates in DNA as a result of impaired folate-dependent de novo thymidylate biosynthesis, a pathway composed of the enzymes serine hydroxymethyltransferase (SHMT), thymidylate synthase (TYMS), and dihydrofolate reductase. In G1, this pathway is present in the cytoplasm and at S phase undergoes small ubiquitin-like modifier-dependent translocation to the nucleus. It is not known whether this pathway functions in the cytoplasm, nucleus, or both in vivo. SHMT1 generates 5,10-methylenetetrahydrofolate for de novo thymidylate biosynthesis, a limiting step in the pathway, but also tightly binds 5-methyltetrahydrofolate in the cytoplasm, a required cofactor for homocysteine remethylation. Overexpression of SHMT1 in cell cultures inhibits folate-dependent homocysteine remethylation and enhances thymidylate biosynthesis. In this study, the impact of increased Shmt1 expression on folate-mediated one-carbon metabolism was determined in mice that overexpress the Shmt1 cDNA (Shmt1tg+ mice). Compared with wild type mice, Shmt1tg+ mice exhibited elevated SHMT1 and TYMS protein levels in tissues and evidence for impaired homocysteine remethylation but surprisingly exhibited depressed levels of nuclear SHMT1 and TYMS, lower rates of nuclear de novo thymidylate biosynthesis, and a nearly 10-fold increase in uracil content in hepatic nuclear DNA when fed a folate- and choline-deficient diet. These results demonstrate that SHMT1 and TYMS localization to the nucleus is essential to prevent uracil accumulation in nuclear DNA and indicate that SHMT1-mediated nuclear de novo thymidylate synthesis is critical for maintaining DNA integrity.

Uracil accumulates in DNA as a result of impaired folate-dependent de novo thymidylate biosynthesis, a pathway composed of the enzymes serine hydroxymethyltransferase (SHMT), thymidylate synthase (TYMS), and dihydrofolate reductase. In G 1 , this pathway is present in the cytoplasm and at S phase undergoes small ubiquitin-like modifier-dependent translocation to the nucleus. It is not known whether this pathway functions in the cytoplasm, nucleus, or both in vivo. SHMT1 generates 5,10methylenetetrahydrofolate for de novo thymidylate biosynthesis, a limiting step in the pathway, but also tightly binds 5-methyltetrahydrofolate in the cytoplasm, a required cofactor for homocysteine remethylation. Overexpression of SHMT1 in cell cultures inhibits folate-dependent homocysteine remethylation and enhances thymidylate biosynthesis. In this study, the impact of increased Shmt1 expression on folate-mediated one-carbon metabolism was determined in mice that overexpress the Shmt1 cDNA (Shmt1 tg؉ mice). Compared with wild type mice, Shmt1 tg؉ mice exhibited elevated SHMT1 and TYMS protein levels in tissues and evidence for impaired homocysteine remethylation but surprisingly exhibited depressed levels of nuclear SHMT1 and TYMS, lower rates of nuclear de novo thymidylate biosynthesis, and a nearly 10-fold increase in uracil content in hepatic nuclear DNA when fed a folate-and choline-deficient diet. These results demonstrate that SHMT1 and TYMS localization to the nucleus is essential to prevent uracil accumulation in nuclear DNA and indicate that SHMT1-mediated nuclear de novo thymidylate synthesis is critical for maintaining DNA integrity.
Depletion of de novo dTMP synthesis caused by folate deficiency, antifolate inhibitors, or genetic disruption of the pathway results in deoxyuridine misincorporation into nuclear DNA leading to genome instability (1)(2)(3). De novo thymidylate biosynthesis is distinct from the other nucleotide synthesis pathways in that it is compartmentalized at the sites of DNA replication, namely the mitochondria (4) and nucleus (5,6). The de novo thymidylate biosynthesis pathway requires three enzymatic activities: serine hydroxymethyltransferase (SHMT), 4 which catalyzes the conversion of serine and tetrahydrofolate (THF) to form glycine and methyleneTHF; thymidylate synthase (TYMS), which catalyzes the conversion of methyleneTHF and dUMP to dTMP; and dihydrofolate (DHF). The cycle is completed by the conversion of DHF to THF in a NADPH-requiring reaction catalyzed by dihydrofolate reductase (DHFR) (Fig. 1). In mitochondria, the pathway is encoded by SHMT2, TYMS, and DHFR1 (4). Disruption of the pathway in Chinese hamster ovary cell mitochondria results in a glycine auxotrophy and elevated levels of uracil in mitochondrial DNA (4). In the nucleus, the de novo thymidylate biosynthesis pathway is encoded by SHMT1 and, to a lesser extent SHMT2, which are functionally redundant in supplying methyleneTHF, TYMS, and DHFR. TYMS is the only gene that is common to and essential for mitochondrial and nuclear thymidylate biosynthesis and is essential for early embryonic development (7).
Nuclear thymidylate biosynthesis is cell cycle-regulated through the small ubiquitin-like modifier (SUMO)-dependent nuclear translocation of SHMT1, TYMS, and DHFR during S phase (5,6,8). Isolated intact nuclei exhibit de novo thymidylate synthesis activity, but this activity is lost in sonicated nuclei, suggesting that this pathway functions through a multienzyme complex and requires compartmentalization in an intact nucleus (8). It is not known whether thymidylate biosynthesis occurs in the cytoplasm. Shmt1 Ϫ/Ϫ mice are viable and fertile (3) and retain about 25% de novo thymidylate biosynthesis capacity in isolated hepatic nuclei because of redundant function of the enzyme SHMT2␣, encoded by Shmt2 through alternative promoter usage (8). Despite this functional redundancy between SHMT1 and SHMT2␣ in the cytoplasm and nucleus, SHMT1 is the major provider of methyleneTHF; mice are sensitive to reductions in SHMT1 expression, as indicated by elevated levels of uracil in nuclear DNA in Shmt1 ϩ/Ϫ mice (3,9), and their increased susceptibility to neural tube defects (10) and Apc min/ϩ -mediated intestinal tumors (9). Interestingly, a common SHMT1 human variant, L474F, is not an effective substrate for UBC-9-catalyzed sumoylation in vitro and exhibits impaired nuclear translocation at S phase (8), suggesting that its association with cardiovascular disease (11,12) and lung cancer (13) risk may be due to impaired nuclear de novo thymidylate synthesis. SHMT1 activity can be limiting for nuclear de novo thymidylate biosynthesis and is also involved in homocysteine remethylation in the cytoplasm. Expression of the SHMT1 cDNA in cell cultures increases rates of de novo thymidylate biosynthesis, while impairing homocysteine remethylation (14). SHMT1 impairs folate-dependent homocysteine remethylation in the cytoplasm by binding and sequestering 5-methylTHF, the substrate for methionine synthase, thereby making it unavailable for the homocysteine remethylation cycle ( Fig. 1) (3,14). SHMT1 expression is regulated over a wide dynamic range in cell culture models by zinc chelators (15), ferritin (16), UV radiation (17), and vitamin A (18), but the effects of increased SHMT1 expression on one-carbon metabolism have not been examined in an animal model. Here, we have generated a transgenic mouse model that overexpresses SHMT1, Shmt1 tgϩ mice, to determine the impact of elevated SHMT1 expression on homocysteine remethylation and de novo thymidylate synthesis in vivo. In this mouse model, increased SHMT1 expression impaired nuclear localization of the de novo dTMP synthesis pathway and markedly increased uracil accumulation in nuclear DNA, indicating an essential role for nuclear thymidylate biosynthesis in the maintenance of DNA integrity and suggests that de novo thymidylate biosynthesis does not occur in the cytoplasm at rates sufficient to prevent uracil misincorporation into DNA.

EXPERIMENTAL PROCEDURES
Shmt1 Expression Vector-The Shmt1 Z/EG vector ( Fig. 2A) was designed to express the ␤geo reporter cassette under the FIGURE 1. Folate-dependent de novo thymidylate biosynthesis. The de novo thymidylate biosynthesis pathway is comprised of SHMT1, SHMT2␣, TYMS, and DHFR. During the S phase, these enzymes are sumoylated by Ubc9, which serves as a signal for nuclear import. FIGURE 2. Generation of SHMT1 overexpressing mice. A, an inducible ␤geo/Shmt1 expression vector was created to allow for Cre/lox activation of the Shmt1 transgene. The ␤geo/Shmt1 expression cassette allows for detection and localization of the transgene promoter activity via the lacZ reporter gene and for conditional overexpression of functional SHMT1. Transgene activation is achieved when Shmt1 tg mice are mated to Cre-expressing mice, in this case, Cre expression was controlled by the CMV promoter. Cre expression results in the deletion of the ␤geo reporter gene while simultaneously repositioning the Shmt1 transgene 3Ј of the chicken ␤-actin promoter; we refer to mice expressing Cre and the transgene as Shmt1 tgϩ . B, Shmt1 tgϩ genotyping in Cre ϩ Shmt1 wild type mice and Cre ϩ Shmt1 tgϩ SHMT1 overexpressing mice. The Shmt1 transgene is detected as a 215-bp PCR product, whereas the wild type (Wt) allele is detected as a 542-bp PCR product. The presence of the Cre transgene was detected as a 119-bp PCR product, and the Il2 internal control gene was detected as a 324-bp PCR product. C, Southern blot of genomic DNA demonstrating a single insertion of the Shmt1 transgene. Total nuclear DNA from tail snips was isolated from PCR verified Shmt1 wt and Shmt1 tg mice. The DNA was digested with BglII (lane 1, does not cut within the Z/EG vector), EcoR5 (lane 2, does not cut the Z/EG vector but cuts at position 418nt of the Shmt1 cDNA), or XhoI (lane 3, the site used for cloning that cuts 5Ј and 3Ј of the Shmt1 cDNA insert within the vector). Shmt1 wtϩ , Cre ϩ Shmt1 wild type mice; Shmt1 tgϩ , Cre ϩ Shmt1 tg mice.
control of the chicken ␤-actin promoter. The Z/EG vector contains the chicken ␤-actin promoter with an upstream cytomegalovirus enhancer that drives the expression of a loxP-flanked ␤geo cassette (lacZ-neomycin resistance fusion) and an internal ribosome entry sequence that enables the translation of an enhanced GFP (19). The murine Shmt1 cDNA (20) was cloned 3Ј of the second loxP site and 5Ј of the internal ribosome entry sequence-enhanced GFP into the XhoI site ( Fig. 2A). Following Cre-mediated excision, the murine Shmt1 cDNA is located directly 3Ј of the ␤-actin promoter.
Generation of Shmt1 tg Mice-Mice were maintained under specific pathogen-free conditions in accordance with standard use protocols and animal welfare regulations. All of the study protocols were approved by the Institutional Animal Care and Use Committee of Cornell University and conform to the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The Z/EG vector containing the Shmt1 cDNA was linearized with SfiI, which cuts 3Ј of the enhanced GFP cDNA. Shmt1 tg mice were generated by pronuclear injection of the linearized construct into FVB/N embryos at the Cornell University Transgenic Mouse Facility. Single site integration of the vector was confirmed in Shmt1 tg founder mice by Southern blot analysis of purified tail DNA (Qiagen) (Fig. 2C). Total nuclear DNA was digested with: BglII, which does not cut within the Shmt1 cDNA or the vector; EcoR5, which cuts at nucleotide 418 of the Shmt1 cDNA but not in the vector; or XhoI, which was the cloning site for the Shmt1 cDNA insertion into the vector. The digests were run on a 1.5% agarose gel, transferred to nitrocellulose, and hybridized with a 32 P-labeled probe generated from nucleotides 443-1233 of the Shmt1 cDNA using protocols described elsewhere (21). DNA isolated from the FVB/N-Tg(ACTB-LacZ/Shmt1)B1Stov (Shmt1 tg ) founder exhibited a high molecular band when digested with BglII or EcoR5. Digestion with XhoI resulted in a prominent band of 1.6 kb and a weaker smaller band, which were not observed in DNA isolated from wild type mice (Fig. 2C). The major 1.6-kb band approximates the size of the Shmt1 cDNA insert, whereas the smaller minor band likely reflects XhoI star activity. Southern blots were repeated over three generations without changes in the Southern banding pattern, indicating a single integration site. A single male founder was used to generate the colony. Thereafter, Shmt1 tg mice were genotyped by PCR using purified tail nuclear DNA (Qiagen) using the forward primer 5Ј-gatcccaagactggcaaagagactt-3Ј and the reverse primer 5Ј-gatgcactcacagagctaggctacaaa-3Ј, which correspond to exon 9 and 10, respectively, of the Shmt1 gene, which amplify 215-and 542-bp products representing the Shmt1 transgene and wild type alleles (Fig. 2B).
Diets-Mice used for characterization of protein overexpression and the nuclear dTMP synthesis assays were fed a standard rodent chow (Harlan Teklad LM-485) from weaning. Mice on the controlled diet study were randomly weaned at 3 weeks of age to a control diet (AIN93G containing 2 mg of folic acid/kg and 2.5 g of choline bitartrate/kg; Dyets, Inc., Bethlehem, PA) or a folate/choline-deficient diet (AIN-93G diet lacking folic acid and choline; Dyets, Inc.). The mice were maintained on the controlled diet for 5 or 32 weeks (until 8 or 35 weeks of age), as indicated in the results.
Determination of AdoMet and AdoHcy Concentrations-The animal feeding cycle was synchronized prior to tissue harvest to ensure AdoMet concentrations reflected homocysteine remethylation capacity with minimal contributions from dietary methionine. Food was removed 24 h prior to killing the animals. After 12 h, each animal was given one food pellet, and the animals were killed 12 h later by cervical dislocation. Tissues were harvested and immediately flash-frozen and stored at Ϫ80°C until analysis. Frozen tissues were sonicated in 500 l of 0.1 M NaAcO buffer (pH 6), and protein was precipitated by adding 312 l of 10% perchloric acid to each sample. After vortexing, samples were centrifuged at 2000 ϫ g for 10 min at 4°C. AdoMet and AdoHcy were determined as described previously (14). AdoMet and AdoHcy values were normalized to total protein (22).
Immunoblotting-For tissues, total protein was extracted and quantified by the Lowry-Bensadoun assay (22). Tissue lysis was achieved by sonication in lysis buffer (2% SDS, 100 mM, dithiothreitol, 60 mM Tris, pH 6.8). For analysis of liver nuclei, pelleted nuclei, purified as described below, were disrupted by boiling in SDS-PAGE loading buffer for 10 min, and protein concentrations were quantified.
Proteins (40 g/well for tissues, 20 g/well for nuclei) were separated on a 10% (nuclei) or 12% (tissue) SDS-PAGE gel. Proteins were transferred at 4°C to an Immobilon-P polyvinylidene difluoride membrane (Millipore Corp.) using a Transblot apparatus (Bio-Rad). Following transfer, the membranes were blocked in 5% (w/v) nonfat skim milk in PBS for 1 h followed by overnight incubation in primary antibody at 4°C. The membranes were washed with PBS containing 0.1% Tween 20 and then incubated overnight with the appropriate horseradish peroxidase-conjugated secondary antibody (below). The membranes were visualized using the SuperSignal West Pico chemiluminescent substrate system (Pierce).
Plasma and Tissue Folate Concentration-Folate concentration in plasma and liver was quantified using the Lactobacillus casei microbiological assay as described (14).
For quantification of nuclear folate levels, four livers were isolated from Shmt1 wtϩ and Shmt1 tgϩ mice and placed immediately in cold PBS at 5°C containing 200 mM ␤-mercaptoethanol (Sigma) and 2% (w/v) sodium ascorbate (Sigma). Liver nuclei from four age-matched males for each genotype were combined and prepared as described under "Nuclear de Novo Thymidylate Synthesis." Nuclear folate content was determined in two independent experiments.
Uracil Content in Nuclear DNA-Nuclear DNA was extracted from 25-50 mg of tissue using DNeasy Tissue and Blood Kit (Qiagen), including an incubation with RNase A (Sigma) and RNase T1 (Ambion) for 30 min at 37°C. 10 g of DNA was treated with 1 unit of uracil DNA glycosylase (Epicenter) for 1 h at 37°C. Immediately following incubation, 10 pg of [ 15 N 2 ]uracil (Cambridge Isotopes) was added to each sample as an internal standard, and the sample was dried completely in a speed vacuum. 50 l of acetonitrile, 10 l of triethylamine, and 1 l of 3,5-bis(trifluoromethyl) benzyl bromide were added to each sample and incubated for 25 min at 30°C with shaking at 500 rpm. 50 l of water followed by 100 l of isooctane were added to each sample. The samples were vortexed and centrifuged. Organic extraction of derived uracil was completed by the removal of the aqueous phase and analysis of the organic phase.
Analysis of uracil-3,5-bis(trifluoromethyl) benzyl bromide was carried out on a Shimadzu QP2010. 1 l of sample or standard was analyzed in the splitless mode with a purge activation time of 1 min and split vent flow of 50 ml/min with an injection port temperature of 280°C. Ultrapurity helium gas was used as carrier gas with a linear velocity of 55 ml/min. Separation of derived uracil was obtained by using an XTI-5, 30 m, 0.25-mm inner diameter, 0.25-m column (Restek), using the following temperature cycles for the oven: 100°C for 1 min, ramping to 280°C at 25°C/min, holding for 5 min, ramping to 300°C at 5°C/min, and holding for 5 min. The interface temperature was held at 300°C with an ion source temperature of 260°C. Ionization was achieved using the NCI mode using methane as the reagent gas and monitoring for ions 337 m/z for uracil and 339 m/z for [ 15 N 2 ]uracil.
Nuclear de Novo Thymidylate Synthesis-Livers from agematched male Shmt1 wt and Shmt1 tg mice fed a standard rodent chow diet (4 -6 months of age, n ϭ 12 per genotype) were dissected and placed immediately in cold PBS at 5°C. Liver extracts for each genotype group were combined for nuclei purification. The nuclei were prepared using an iodixanol gradient as previously described (8). De novo dTMP reactions were completed as previously described (8). In short, purified nuclei were suspended in 500 l of nuclear assay buffer containing 5 mM NADPH (Sigma), 100 mM ␤-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 l of assay buffer containing equal numbers of suspended nuclei were aliquoted into four 1.5-ml plastic tubes, and 8 Ci of [2,3-3 H]-L-serine (Moravek Biochemicals) was added to each sample. The assay was conducted under three different experimental conditions: lysed nuclei, nuclei were lysed with sonication (Branson Sonifier 150) at 5°C using two 10-s pulses at 10 watts separated by a 10-s resting interval; intact nuclei; or intact nuclei with SHMT1 inhibitor, intact nuclei incubated with aminomethylphosphonate (Sigma) added to a final concentration of 100 mM. The reactions were incubated for 12 h at 37°C with shaking at 300 rpm. The nuclei were pelleted by centrifugation at 2000 rpm for 5 min, and the supernatant was collected and analyzed for radiolabeled thymidylate by HPLC. Sample preparation and HPLC was performed as previously described (23,24). Fractions were collected, and tritium was 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. The data were normalized to the number of nuclei. All of the experiments were performed in duplicate.
Statistical Analyses-Differences in genotype distribution were analyzed by the 2 test. Differences between two groups were determined by Student's t test analysis. Differences among more than two genotypes were analyzed by two-way ANOVA and Dunnett's post hoc test using the wild type genotype as the control group. Diet ϫ genotype effects were analyzed by twoway ANOVA and Tukey's Honestly Significant Difference post hoc test. The groups were considered significantly different when the p value Յ 0.05. The data are presented as the mean Ϯ S.E. All of the statistics were performed using JMP IN software, release 5.1.2.

RESULTS
Shmt1 tgϩ Mice Are Viable-Shmt1 transgene expression was activated by crossing Shmt1 tg mice to Cre-expressing mice, which results in deletion of the ␤geo cassette and relocation of the Shmt1 intronless transgene 3Ј to the ␤-actin promoter, generating Shmt1 tgϩ mice ( Fig. 2A; ϩ denotes Cre positive). PCRbased genotyping results in the generation of a 542-bp PCR product for the wild type Shmt1 allele and a 215-bp PCR product for the transgene (Fig. 2B).
To determine whether Shmt1 tg mice were viable upon activation of the transgene, the Shmt1 tgϩ genotype distribution was determined from crosses between BALB/c CMV-Cre homozygous male mice and FVB/N Shmt1 tg female mice carrying one copy of the transgenic allele (Fig. 2C). A total of 226 F1 pups from 39 litters were examined ( Table 1). The mean litter size was 5.8 pups, which approximates observed litter sizes for inbred FVB/N and BALB/c mice. The Shmt1 transgene was distributed as expected for Mendelian inheritance with a ratio of Shmt1 wtϩ to Shmt1 tgϩ mice of 97:129, and both sexes were found at the expected frequency. Shmt1 tgϩ mice were not bred, so the effect of SHMT1 overexpression on fertility remains to be determined. SHMT1 Protein Overexpression in Shmt1 tgϩ Mice-Shmt1 tgϩ mice demonstrated variable tissue-dependent SHMT1 overexpression, as detected by immunoblotting (Fig. 3). The kidney, colon, and ileum of Shmt1 tgϩ mice exhibited a 3-4-fold increase, whereas liver demonstrated a 2-fold increase in SHMT1 protein levels in comparison with Shmt1 wt , Shmt wtϩ , and Shmt1 tg (not Cre activated) mice. Endogenous SHMT1 is undetectable in the brain of wild type mice by immunoblotting, but a significant amount of SHMT1 can be detected in the brain of Shmt tgϩ mice. Note that the lower band on the brain immunoblot is nonspecific, as is also observed in Shmt1 null mice.
SHMT1 and Folate Status-SHMT1 overexpression did not impact plasma or liver folate status in mice fed the control or folate/choline-deficient diet ( Table 2). As expected, the folate/ choline-deficient diet resulted in a significant decrease in plasma and liver folate after 32 weeks on the diet (Table 2). SHMT1, de Novo Thymidylate Synthesis and Uracil Misincorporation in Nuclear DNA-SHMT1 preferentially partitions one-carbon units to thymidylate synthesis through the SUMOmediated localization of the thymidylate synthesis pathway to the nucleus during S phase (5,6). Furthermore, TYMS protein levels respond to changes in SHMT1 expression; Shmt1 ϩ/Ϫ mice exhibit decreased SHMT1 and TYMS protein, decreased capacity for de novo dTMP synthesis, and elevated uracil in nuclear DNA (9). Therefore, we hypothesized that overexpression of SHMT1 would be associated with increased TYMS expression and increased nuclear de novo thymidylate synthesis capacity with a consequent decrease in uracil incorporation in nuclear DNA. Unexpectedly, Shmt1 tgϩ mice fed the control diet demonstrated a significant 2-fold increase in uracil content in hepatic nuclear DNA (Table 2) despite an almost 4-fold increase in total liver SHMT1 (3.9 Ϯ 0.3) and TYMS (3.7 Ϯ 0.4) protein content in comparison with Shmt1 wtϩ mice (1.0 Ϯ 0.3 and 1.0 Ϯ 0.3, respectively) (Fig. 4A). When placed on a folateand choline-deficient diet, Shmt1 tgϩ mice exhibited a 10-fold increase in nuclear uracil content compared with a 2-fold increase observed in wild type mice on the same deficient diet ( Table 2). These data indicate that elevated SHMT1 and TYMS expression is not sufficient to increase capacity for de novo thymidylate synthesis.
To explore this unexpected finding, we sought to clarify the specific effect of SHMT1 overexpression on nuclear de novo thymidylate synthesis. Isolated hepatic nuclei from Shmt1 tgϩ mice fed the folate-replete diet contained 75% less SHMT1 and TYMS protein compared with their wild type counterparts (Fig.  4, C and D), whereas SHMT1 overexpression had no effect on DHFR localization to the nucleus (Fig. 4D). The decrease in nuclear SHMT1 and TYMS observed in Shmt1 tgϩ mice was associated with an approximate 50% decrease in nuclear de novo thymidylate biosynthesis capacity in purified hepatic nuclei (Fig. 4B). The decrease in de novo thymidylate synthesis was not due to decreased hepatic nuclear folate content because we did not observe significant differences in nuclear folate concentrations between Shmt1 wtϩ and Shmt1 tgϩ mice fed a standard rodent chow (9.5 Ϯ 0.4 versus 10.7 Ϯ 2.5, respectively).
Impact of SHMT1 Overexpression on Homocysteine Remethylation-SHMT1 binds 5-methylTHF in the cytoplasm, making it unavailable for the remethylation of homocysteine in cultured MCF-7 cells (14). Therefore increased SHMT1 expression is expected to result in decreased AdoMet levels and elevated homocysteine. Consistent with this expectation, Shmt1 tg mice were crossed with Cre-expressing mice to activate expression of the Shmt1 transgene, and their progeny were genotyped. The expected genotype distribution was calculated based on a Mendelian distribution. Differences between observed and expected genotype distributions were analyzed by chi-squared analysis. p values Յ 0.05 were considered significantly different. Shmt1 wtϩ , Cre ϩ Shmt1 wild type mice; Shmt1 tgϩ , Cre ϩ Shmt1 tg mice. NS, not significant.   1 and 2, respectively), and Cre-activated Shmt1 wild type (wt؉, lanes 3 and 5) and transgenic (tg؉, lanes 4 and 6) mice. Immunoblots of tissue lysates were probed with polyclonal anti-SHMT1 and polyclonal anti-GAPDH antibodies. GAPDH served as a loading control. DECEMBER 23, 2011 • VOLUME 286 • NUMBER 51

Nuclear Thymidylate Biosynthesis
hepatic AdoMet and AdoMet:AdoHcy ratio, which is indicative of the cellular methylation capacity, were significantly lower in Shmt1 tgϩ mice at 32 weeks post-weaning (Table 2). AdoHcy was unaffected by SHMT1 overexpression. We did not observe any genotype ϫ diet effects on AdoMet, AdoHcy, or the Ado-Met:AdoHcy ratio. The impact of SHMT1 overexpression on homocysteineand folate-related metabolites in the plasma of male and female mice was determined at 5 weeks post-weaning. We did not observe a significant effect of SHMT1 overexpression on any of the metabolites queried because of insufficient power to detect significant differences. A significant sex effect on plasma homocysteine, methionine, cysteine, and cystathionine was observed. Female mice exhibited increased plasma homocysteine and cysteine and decreased cystathionine and methionine relative to male mice. Plasma serine tended to be decreased in

Plasma and liver folate, liver S-adenosyl-methionine, S-adenosyl-homocysteine, S-adenosyl-methionine:S-adenosyl-homocysteine ratio, and uracil content in SHMT1-overexpressing mice at 32 weeks post-weaning
Differences between genotypes and diets were analyzed by Student's t test. nuclear DNA genotype ϫ diet effects were analyzed by two-way ANOVA using Tukey's Honestly Significant Difference post-hoc analysis. The data represent the means Ϯ S.E. values. p values Յ 0.05 were considered significantly different. n ϭ 4 -7/group. Folate/cholinedeficient Shmt1 tgϩ mice are significantly different than control and folate/choline-deficient Shmt1 wtϩ and control Shmt1 tgϩ mice, p Ͻ 0.05, as analyzed by two-way ANOVA and Tukey's Honestly Significant Difference post-hoc test Shmt1 wtϩ , Cre ϩ Shmt1 wild type mice; Shmt1 tgϩ , Cre ϩ Shmt1 tg mice. NS, not significant.  (8). De novo thymidylate biosynthesis activity was normalized to that of Shmt1 wtϩ intact nuclei, which was assigned an arbitrary value of 1.0. The reactions were performed in duplicate, and the experiment was repeated twice. We did not observe a run-dependent difference; therefore data from the two replicate experiments were combined. The data are presented as the means Ϯ S.E. C and D, immunoblotting for SHMT1, TYMS, and DHFR was performed on hepatic nuclei isolated from Shmt1 null (C only) (3) and Shmt1 wtϩ and Shmt1 tgϩ mice. Lamin A was used as a nucleus-specific loading control. E, immunoblotting of cytoplasm-restricted GAPDH confirmed the purity of hepatic nuclei.

Diet
Shmt1 tgϩ mice, an effect driven by samples from male mice ( Table 3). The folate/choline-deficient diet was associated with increased plasma homocysteine and cysteine and decreased methylglycine (Table 3). We did not observe any significant genotype ϫ diet effects on any of the queried metabolites.

DISCUSSION
Our study demonstrates that nuclear localization of the de novo dTMP synthesis pathway is essential to prevent uracil accumulation in nuclear DNA and thereby maintain genome stability and confirms previous studies demonstrating that SHMT1 expression is an important determinant of uracil accumulation in DNA. The results demonstrate unequivocally that restricting the de novo dTMP synthesis pathway to the cytoplasm results in elevated uracil accumulation into nuclear DNA.
Shmt1 tgϩ mice demonstrated a 2-4-fold increase in SHMT1 and TYMS protein in tissues that normally express SHMT1, namely the liver, kidney, and gastrointestinal tract (Fig. 3). No significant differences in genotype distribution among Cre-activated F1 wild type or transgenic pups were observed (Table 1), indicating that the achieved level of SHMT1 overexpression had no impact on embryo survival in dams fed a standard rodent chow diet. It will be of interest to determine the effect of a folate deficient diet on Shmt1 tgϩ embryo development, because we have demonstrated folate-responsive neural tube defects in Shmt1 heterozygous null mice, which was attributed to reduced de novo thymidylate synthesis (10).
Shmt1 ϩ/Ϫ mice have been reported to exhibit increased uracil content in hepatic nuclear DNA, reduced levels of nuclear SHMT1 and TYMS, and consequent decreased capacity for de novo thymidylate synthesis (3,9). In this study, contrary to expectations, uracil content in hepatic nuclear DNA was increased in Shmt1 tgϩ mice compared with wild type mice, which was exacerbated markedly by a folate-and choline-deficient diet. Indeed, hepatic nuclear DNA uracil content in Shmt1 tgϩ mice fed the control diet was comparable with that observed in Shmt1 ϩ/Ϫ mice (3,9). Compared with wild type mice, Shmt1 tgϩ mice exhibited a 4-fold increase in total hepatic SHMT1 and TYMS protein expression, whereas nuclear localization of SHMT1 and TYMS was reduced by ϳ75%, resulting in a 50% reduction in dTMP synthesis in isolated nuclei from folate-replete mice (Fig. 4). Although Shmt1 tgϩ have severalfold increased expression of SHMT1 and TYMS compared with Shmt ϩ/Ϫ mice, they both exhibit decreased levels of nuclear SHMT1 and TYMS protein, reduced nuclear thymidylate synthesis, and elevated uracil in DNA compared with wild type mice. Collectively, these results indicate that nuclear localization of the de novo thymidylate biosynthesis pathway is essential for maintaining sufficient thymidine nucleotide pools for DNA replication.
The mechanism by which SHMT1 and TYMS accumulation in the nucleus is inhibited in Shmt1 tgϩ mice is not known. Elevated expression of SHMT1 in the cytoplasm may impair sumoylation of SHMT1 and/or TYMS or prevent their translocation into the nucleus. Interestingly, the accumulation of SHMT1 and TYMS, but not DHFR, in the nucleus is coordinated and dependent on SHMT1 expression.
Similar to our previous studies, in which SHMT1-dependent expression of TYMS was observed in colon and embryonic tissue (9, 10), the present study provides additional evidence for the coregulation of these two proteins. Shmt1 Ϫ/ϩ mice were shown to have ϳ50% SHMT1 protein content, which was con- Plasma metabolic profile of SHMT1 overexpressing mice at 5 weeks post-weaning comitant with a reduction in TYMS (9). Interestingly, SHMT1, TYMS, and thymidine kinase 1 protein expression were also responsive to folate deficiency as demonstrated by their increased expression in mice fed the folate/choline-deficient diet (9). Microarray analysis did not indicate significant transcriptional changes to TYMS expression in Shmt1 Ϫ/ϩ , suggesting that their coregulation occurs post-transcriptionally (9). We also have not queried the effect of SHMT1 expression on sumoylation of itself or TYMS, a mode by which SHMT1 could influence nuclear translocation. The mechanism by which SHMT1 expression influences TYMS levels is actively being investigated.
This study also confirms that increased SHMT1 expression plays a modest role in the regulation of homocysteine methylation in the cytoplasm. SHMT1 is a 5-methylTHF binding protein (27), and therefore overexpression of SHMT1 was anticipated to reduce hepatic methionine and AdoMet synthesis by sequestering cytoplasmic 5-methylTHF. Consistent with this expectation, SHMT1 overexpression decreased hepatic AdoMet, which resulted in a reduced AdoMet:AdoHcy ratio ( Table 2). SHMT1 overexpression did not have a major impact on serum one-carbon metabolites, because our study was underpowered to detect significant differences.
We previously demonstrated that the common human L474F variant impairs the UBC9-SHMT1 interaction and consequently SHMT1 sumoylation and translocation to the nucleus (5). The impact of this SHMT1 variant on TYMS nuclear localization and nuclear de novo thymidylate biosynthesis capacity is unknown. The reduction of nuclear de novo dTMP synthesis capacity in the Shmt1 tgϩ mouse provides a unique experimental model of the human L474F SHMT1 variant and permits mechanistic studies that investigate and validate reported epidemiological associations of this variant with lung cancer and cardiovascular risk (11)(12)(13).