Shmt1 Heterozygosity Impairs Folate-Dependent Thymidylate Synthesis Capacity and Modifies Risk of Apc^min-Mediated Intestinal Cancer Risk

The interactions among nutrients and genetic factors play an important role in the development of numerous cancers including colorectal cancer (CRC). A strong, inverse association of folate status and CRC has been demonstrated; individuals with lowest dietary folate intake show a 40% to 60% increase in CRC risk when compared with individuals with highest folate intake (1–3). Genetic variation that alters folate metabolism and utilization also influences cancer risk (2). The mechanism by which folate metabolism alters CRC risk is not known, which has led to concerns regarding the potential impact of elevated dietary folate intake and folate fortification initiatives on CRC incidence (4, 5).

Folate-mediated one-carbon (1C) metabolism is a metabolic network of interdependent biosynthetic pathways required for the de novo biosynthesis of purines, dTMP, and the remethylation of homocysteine to methionine (Fig. 1; ref. 6). Methionine can be converted to S-adenosylmethionine (AdoMet), the major 1C donor for cellular methylation reactions including the methylation of DNA, RNA, phospholipids, proteins, and small molecules (6, 7). Impairments in 1C metabolism due to nutrient deficiencies and/or single-nucleotide polymorphisms diminish dTMP synthesis, leading to elevated dUTP pools, increased rates of dUTP misincorporation into DNA, and consequently futile cycles of DNA excision repair and chromosomal strand breaks (8–10). Altered folate metabolism also influences chromatin methylation patterns, including genome-wide CpG hypomethylation and site-specific hypermethylation and altered histone methylation, which modify gene expression patterns (11–13). Therefore, the loss of DNA integrity due to increased genome instability and/or changes in gene expression due to altered genome methylation are candidate causal pathways for folate-mediated CRC.

Cytoplasmic serine hydroxymethyltransferase (SHMT1) acts as a metabolic switch that regulates the partitioning of 1C units between the dTMP and methionine biosynthetic pathways (Fig. 1; ref. 14). Folate-activated 1Cs can be derived in the cytoplasm from formate through the Mthfd1 gene product or from serine through SHMT1 to support methionine and thymidylate biosynthesis (Fig. 1). Thymidylate synthesis can occur in both the cytoplasm and nucleus (7, 15). When expressed, SHMT1 preferentially partitions methylenetetrahydrofolate (methyleneTHF) into the dTMP synthesis pathway through the small ubiquitin-like modifier (SUMO)-mediated compartmentalization of the dTMP biosynthetic pathway components in the nucleus during S phase (14, 16). In the cytoplasm, SHMT1 tightly binds and sequesters 5-methylTHF making it unavailable for the methionine cycle, inhibiting AdoMet synthesis, and reducing the cellular methylation potential (14).

Aberrant Wnt signaling resulting from altered Apc gene expression is observed in approximately 85% of cases of sporadic CRC (reviewed in ref. 17). In rats, folate deficiency has previously been associated with strand breaks in the mutation cluster region within the Apc gene, which were inversely correlated with Apc gene expression (18). Folate-mediated strand breaks in the Apc gene could be due to hypomethylation of the gene or genomic instability induced by misincorporation of uracil into genomic DNA. In the present study, we have modified the expression of SHMT1 in mice to determine the role(s) played by de novo purine, thymidylate, and methionine synthesis in the development of Apc^min-mediated intestinal cancer.

SHMT1 null (Shmt1^−/+) mice were generated as previously described (19) and backcrossed a minimum of 10 generations onto the C57BL/6J strain. Mice were genotyped using the forward primer 5′-GACACTGTTCACATCCCTC-3′ and the reverse primer 5′-CAAAACATTCGGGAGCCTC-3′. The forward primer corresponds to an intron 6 sequence located 5′ to a loxP site and exon 7 and the reverse primer corresponds to an intron 7 sequence located downstream of a 3′ loxP site (19). C57BL/6J-Apc^Min/J (Apc^min/+) mice were obtained from The Jackson Laboratory. Genotyping of Apc^min/+ mice was performed using the following primers recommended by The Jackson Laboratory protocol: wild-type forward primer 5′-GCCATCCCTTCACGTTAG-3′, Min forward primer 5′-TTCTGAGAAAGACAGAAGTTA-3′, and a common reverse primer 5′-TTCCACTTTGGCATAAGGC-3′. Shmt1^−/+ mice were mated to Apc^min/+ mice. Double heterozygous offspring were intercrossed to achieve Apc^min/+Shmt1^+/+, Apc^min/+Shmt1^−/+or Apc^min/+Shmt1^−/− mice.

Mice were maintained under specific pathogen-free conditions at the Cornell University Transgenic Mouse Core Facility in accordance with standard use protocols and animal welfare regulations. 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.

Mice were randomly weaned onto either a control (AIN-93G; Dyets, Inc.) diet or a modified AIN-93G diet lacking folic acid and choline (Dyets, Inc.) at 3 weeks of age. The control diet contained 2 mg/kg folic acid and 2.5 g/kg choline bitartrate, and the folate/choline-deficient (FCD) diet contained 0 mg/kg folic acid and 0 g/kg choline bitartrate. Mice were maintained on the diet for 5 weeks (Apc^+/+) or 11 weeks (Apc^min/+).

The colons from wild-type FVB/NJ (Friend virus B NIH Jackson) mice were fixed as Swiss rolls in 10% neutral-buffered formalin. Samples were embedded in paraffin and sectioned (5 μm). Polyclonal sheep anti-SHMT1 antibody was used to probe the sections for cellular localization of SHMT1 (20). Nonimmune sheep serum was used as a negative control. Biotinylated rabbit anti-sheep IgG was used as the secondary antibody. SHMT1 was visualized with AEC (3-amino-9-ethylcarbazole) developer (Dako). Sections were counterstained using Gill's hematoxylin 2 (Sigma). Coverslips were mounted using Fluoromount-G (Sigma).

The small intestine and colon were removed, flushed with cold PBS, opened longitudinally, and laid flat lumen side up for examination using a dissecting microscope as previously described (21). Tumors were counted according to intestinal location (small intestine or colon) and their diameter measured by a pathologist-trained investigator who was blinded to the genotype of the intestinal specimen. Tumor load is a function of tumor number and area and was calculated as the total tumor area per mouse.

Folate concentration of plasma and tissues was quantified using the Lactobacillus casei microbiological assay (14). Protein concentration was quantified (22).

Uracil content in hepatic nuclear DNA was determined by gas chromatography/mass spectrometry as previously described (19). Uracil content in liver nuclear DNA has been shown to correlate with uracil content in colonic nuclear DNA in Shmt1^−/+ mice (A.J. MacFarlane and P.J. Stover, in preparation).

The animal feeding cycle was synchronized prior to tissue harvest to ensure AdoMet levels reflected homocysteine remethylation capacity with minimal contributions from dietary methionine. Food was removed 24 hours before killing the animals. After 12 hours, each animal was given one food pellet and the animals were killed 12 hours later by cervical dislocation. Tissues were harvested and immediately flash frozen and stored at −80°C until analysis. Frozen tissues were sonicated in 500 μL of 0.1 mol/L NaAcO buffer (pH 6), and protein was precipitated by adding 312 μL of 10% perchloric acid to each sample. After vortexing, samples were centrifuged at 2,000 × g for 10 minutes at 4°C. AdoMet and AdoHcy (S-adenosylhomocysteine) were determined by high-performance liquid chromatography as described previously (14). AdoMet and AdoHcy values were normalized to cellular protein content (22).

Total protein was extracted and quantified from frozen tissue (22). Immunoblotting was performed as previously described (19). The membranes were visualized using the SuperSignal West Pico chemiluminescent substrate system (Pierce). For actin detection, polyclonal rabbit antiactin antibody conjugated to horseradish peroxidase (HRP; Abcam) was diluted 1:40,000. For SHMT1 detection, sheep anti-mouse SHMT1 antibody (20) was diluted 1:10,000, and rabbit anti-sheep IgG secondary antibody conjugated to HRP (Pierce) was diluted 1:20,000. For thymidylate synthase (TYMS) detection, monoclonal mouse anti-TYMS antibody (Invitrogen) was diluted 1:3,000 and goat anti-mouse IgG secondary antibody conjugated to HRP (Pierce) was diluted 1:5,000. For cytoplasmic thymidine kinase (TK1) detection, monoclonal mouse anti-TK1 antibody (Abcam) was diluted 1:1,000, and horse anti-mouse IgG secondary antibody conjugated to HRP (Cell Signaling Technology) was diluted 1:1,000. Densitometric analysis of Western blots was performed using Scion Image for Windows (Scion Corporation). Three animals per group were included in the analysis. Values were normalized to glyceraldehyde 3 phosphate dehydrogenase. Densitometric data are presented relative to the mean of wild-type animals fed the control diet.

Mouse embryonic fibroblasts (MEF) were freshly isolated as described in the work of MacFarlane and colleagues (19). Cells were cultured in α-minimal essential medium (Hyclone Laboratories) supplemented with 10% fetal calf serum (FCS; Hyclone Laboratories), 0.1 mmol/L nonessential amino acids (Invitrogen), 1 mmol/L sodium pyruvate (Invitrogen), and penicillin/streptomycin (Invitrogen) and incubated at 37°C in a 5% CO₂ atmosphere.

[¹⁴C]Formate and [³H]hypoxanthine are precursors for purine nucleotide biosynthesis through the de novo and salvage pathways, respectively (23). For tracer experiments, FCS was dialyzed against PBS for over 24 hours with 6 buffer changes to remove folate and other small molecules and then charcoal treated to remove any residual folate. The tracer medium was Defined Minimal Essential Medium (Hyclone Laboratories) that lacked glycine, serine, methionine, hypoxanthine, and folate but was supplemented with 10% dialyzed and charcoal-treated FCS, 1 mg/L pyridoxyl-l-phosphate, 200 μmol/L methionine, 20 nmol/L leucovorin, 2 nmol/L [³H]hypoxanthine (Moravek), and 20 μmol/L [¹⁴C]formate (Moravek). MEFs were plated 2 × 10⁴ cells/well in 6-well plates and grown at 37°C, 5% CO₂ in tracer medium until confluent and harvested. The cell pellets were stored at −80°C. Nuclear DNA was isolated using a DNA blood kit from Qiagen, RNase A treated and isotope levels quantified on a Beckman LS6500 scintillation counter in dual disintegrations per minute (dpm) mode.

[¹⁴C]Deoxyuridine and [³H]thymidine are precursors for thymidylate nucleotide biosynthesis through the de novo and salvage pathways, respectively. The tracer medium was Defined Minimal Essential Medium (Hyclone Laboratories) that lacked glycine, serine, methionine, hypoxanthine, and folate but was supplemented with 10% dialyzed and charcoal-treated FCS, 1 mg/L pyridoxyl-l-phosphate, 200 μmol/L methionine, 20 nmol/L leucovorin, 500 nmol/L [³H]thymidine (Moravek), and 10 μmol/L [¹⁴C]deoxyuridine (Moravek). MEFs were plated 2 × 10⁴ cells/well in 6-well plates and grown at 37°C, 5% CO₂ in tracer medium until confluent and harvested. The cell pellets were stored at −80°C. Nuclear DNA was isolated using a DNA blood kit from Qiagen with RNase A treatment and isotope levels quantified on a Beckman LS6500 scintillation counter in dual dpm mode. Data were normalized to the wild-type MEF line in each independent experiment.

Microarray analysis was performed on colon tissue from Shmt1^+/+, Shmt1^−/+, and Shmt1^−/− mice fed the AIN-93G minus folate and choline diet. Total RNA was isolated from colons and 50 ng was amplified using the Nugen Ovation V2 kit (Nugen). Amplified RNA from 9 mice (3 independently processed samples from each genotype) was hybridized to Affymetrix 430A mouse gene expression chips (Affymetrix) and scanned using an Affymetrix GeneChip 3000 scanner. The raw array data were processed using GCOS software to obtain signal values scaled to a default target of 500. Further statistical data analysis was performed using R (24). The signal values were log₂-transformed after offset by 8, and gene filtering was applied to include only the 19,871 probe sets having at least one present call in the data set. A 2-sample t test was applied on the normalized log ratio of each probe set. At a P value cutoff point of 0.01, 134 differential genes between wild-type and null mice were selected whereas 229 differential genes were selected between wild-type and heterozygous mice (Supplementary Table S1). Because only 3 replicates were performed for each sample, multiple testing correction was not performed. The microarray data have been deposited in the National Center for Biotechnology Information's Gene Expression Omnibus (GEO; ref. 25) and are accessible through GEO Series accession number GSE14645.

Gene expression levels of candidate genes identified in the microarray screen were tested by quantitative reverse transcription PCR (qRT-PCR) using the Taqman-based Universal Probe Library (Roche). Colon RNA used in the microarray screen was primed with oligo(dT) and reverse transcribed with Superscript III (Invitrogen) to generate cDNA template. Quantitative RT-PCR primers were designed using the Universal Probe Library Assay Design Center, (Roche; ref. 26) and validated on an Applied Biosystems 7500 machine by comparison against a standard curve generated with rodent GAPDH control primers (Applied Biosystems). Relative quantitation analysis was then performed in triplicate for all 9 samples. Genes tested, Affymetrix ID, primer sequences, and universal probe library numbers used are in Supplementary Table S2.

Differences between 2 groups were determined by Student's t-test analysis. Differences among more than 2 groups were analyzed by 2-way ANOVA and Tukey's HSD post hoc test. Nonnormally distributed data were normalized by log transformation for analysis. Groups were considered significantly different when the P value was ≤ 0.05. All statistics were performed using JMP IN software, release 5.1.2 (Copyright 1989–2004, SAS Institute Inc.).

SHMT1 exhibits tissue-specific expression and is present in all tissues associated with folate-mediated pathologies including the colon (19, 27, 28). Here, we show that SHMT1 is specifically expressed in the cytoplasm of mature colonic enterocytes (Fig. 2A), in mitotically active epithelial cells in the proliferative zone of colonic crypts (Fig. 2A and C), and in the nucleus of cells residing at the base of colonic crypts (Fig. 2A and E).

To determine the effect of Shmt1 on intestinal tumor development, Shmt1^−/+ mice (19), were mated to Apc^min/+ mice, a model of spontaneous intestinal neoplasia (29). Double heterozygous offspring were intercrossed to achieve Apc^min/+Shmt1^+/+, Apc^min/+Shmt1^−/+, or Apc^min/+Shmt1^−/− mice. Mice were weaned at 21 days and randomly assigned to either a control (C) or an FCD diet. Plasma and colon folate concentrations were determined in Apc^+/+ littermates after 5 weeks on diet to estimate folate status at the midpoint of the experiment. Folate in colon was significantly reduced by the FCD diet after 5 weeks (Table 1). This confirms that the colon is sensitive to dietary folate depletion (30). Liver folate concentrations were significantly reduced in mice only after 11 weeks on diet (Table 1).

The small and large intestines from Apc^min/+ mice were examined for neoplasia at 14 weeks of age, after 11 weeks on diet. In mice fed the control diet, decreased SHMT1 expression had no effect on tumor number or load (Fig. 3A and B). The FCD diet was not associated with increased tumor number or load in Apc^min/+Shmt1^+/+, or Shmt1^−/− mice. However, the FCD diet was associated with more than a 50% increase in total tumor number and load in Apc^min/+Shmt1^−/+ mice (Fig. 3A and B). The effect in Apc^min/+Shmt1^−/+ mice on tumor number was driven by an approximately 50% increase in small intestinal tumors and a doubling of colon tumors, although the latter was not statistically significant

To determine the mechanism by which Shmt1 hemizygosity modifies CRC susceptibility in Apc^min/+ mice, the effects of SHMT1 depletion on de novo purine and thymidylate biosynthesis and cellular methylation potential were determined. MEFs were used to determine the impact of Shmt1 hemizygosity on de novo nucleotide synthesis relative to synthesis by the salvage pathway. We found that de novo purine synthesis relative to synthesis by the salvage pathway was not significantly affected by Shmt1 genotype in MEFs (Fig. 4A).

Our previous data suggest that Shmt1^−/− mice have increased hepatic AdoMet concentrations compared with wild-type and hemizygous mice under conditions of folate depletion (19). However, AdoMet and AdoHcy levels, as well as the AdoMet/AdoHcy ratio, were not significantly affected by Shmt1 genotype or diet in this study (Table 2).

In addition, we performed a microarray analysis on colons from mice fed the FCD diet to determine the effect of Shmt1 on gene expression and its relationship to tumor susceptibility. Using the NuGen Ovation amplification approach, a Present Call of between 78% and 81% was achieved, indicating high sensitivity in transcript detection. However, at a P value of 0.01, less than 10 genes in Shmt1 hemizygous mice and less than 15 genes in null mice of the possible 19,000 present genes had a 2-fold or greater change in gene expression. These minimal differences strongly suggest that gene expression changes in Shmt1 mutants would be difficult to detect (Supplementary Table S1). The data were confirmed by qRT-PCR for 9 genes of interest (Supplementary Table S2). Only one gene, uridine phosphorylase (Upp1), demonstrated consistent changes in expression.

Previously, we have shown that uracil content in hepatic nuclear DNA was significantly affected by Shmt1 genotype (19). Here, we have found that uracil content in hepatic nuclear DNA, which correlates with colon uracil content in Shmt1 knockout mice (A.J. MacFarlane and P.J. Stover, in preparation), was significantly increased in Apc^min/+Shmt1^−/+ mice in comparison with Apc^min/+Shmt1^+/+ mice, and uracil content doubled when the Apc^min/+Shmt1^−/+ mice were fed the FCD diet (Table 2). Although uracil content also doubled in Apc^min/+Shmt1^+/+ fed the FCD diet, the absolute concentration was consistently one third of that observed in Apc^min/+Shmt1^−/+ regardless of diet. In addition, MEFs derived from Shmt1^−/+ mice demonstrated a tendency for decreased de novo thymidylate synthesis relative to the salvage pathway in comparison with Shmt1^+/+ MEFs as determined by a modified deoxyuridine suppression assay (ref. 31; Fig. 4B). Furthermore, the protein expression of TYMS, which functions in de novo thymidylate synthesis, was decreased in the colon of Shmt1^−/+ mice in comparison with wild-type mice fed the control diet (Fig. 4C; 0.69 ± 0.3 vs. 1.00 ± 0.04). TK1 protein expression was not different between hemizygous and wild-type mice. The data indicate that Shmt1^−/+ mice experience a reduction in de novo thymidylate synthesis capacity relative to their wild-type counterparts, which likely accounts for the increased incorporation of uracil into nuclear DNA observed in these mice.

Surprisingly, Shmt1^−/− mice had lower uracil levels in nuclear DNA compared with Shmt1^−/+ mice (Table 2) and MEFs from Shmt1^−/− mice exhibited increased de novo thymidylate biosynthesis relative to synthesis from salvage compared with those derived from Shmt1^−/+ mice (Fig. 4B). Shmt1^−/− mice demonstrated markedly elevated colonic TYMS (2.86 ± 1.3 vs. 1.00 ± 0.04) and TK1 (2.19 ± 1.0 vs. 0.97 ± 0.06) protein expression in comparison to Shmt1 wild-type mice. Although the results were unexpected, the changes in TYMS and TK1 protein expression observed in Shmt1^−/− mice could explain their increased capacity for thymidylate synthesis and decreased uracil concentration in genomic DNA in comparison with Shmt1^−/+ mice.

In addition to Shmt1-dependent changes to thymidylate synthesis capacity, we also observed an upregulation of members of the de novo and salvage dTMP biosynthetic pathways in response to folate deficiency. SHMT1 protein was increased by 35% (1.2 ± 0.4 vs. 0.9 ± 0.1), and TYMS protein was increased by 80% (1.8 ± 0.6 vs. 1.00 ± 0.04), as determined by densitometry, in the colons of Shmt1 wild-type mice fed the FCD diet in comparison to control fed mice (Fig. 4C), which could account for the lack of an effect of folate deficiency on uracil content in DNA (Table 2) and intestinal tumor number and load in wild-type mice (Fig. 3A and B). Although SHMT1 (0.56 ± 0.2 vs. 0.4 ± 0.06) and TYMS (1.7 ± 0.5 vs. 0.7 ± 0.3) levels were also increased in Shmt1^−/+ mice fed the FCD diet, the absolute levels of these enzymes never exceeded than those observed in wild-type mice (Fig. 4C). TK1 protein levels were also increased when wild-type mice were fed the FCD diet (1.0 ± 0.06 vs. 2.0 ± 0.8), which could support an increased capacity to synthesize thymidylate from the salvage pathway and also contribute to their protection from diet-mediated cancer susceptibility. TK1 also increased in Shmt1^−/+ mice fed the FCD diet (1.5 ± 0.1 vs. 1.0 ± 0.6), but the induction was to a lesser extent than that observed in wild-type mice. Finally, TK1 remained high but unchanged in Shmt1^−/− mice fed the FCD diet (2.0 ± 0.5 vs. 2.2 ± 1.1).

Altered folate metabolism has been shown to be a risk factor for the development of CRC. However, the mechanism(s) involved remain unknown. Because of its metabolic properties and its cellular localization, SHMT1 is poised to regulate 1C metabolism in cells associated with neoplastic transformation and CRC risk (32). Therefore, SHMT1 expression was modified to determine the relative contributions of dTMP, purine, and AdoMet synthesis and consequently indicators of genome stability and methylation, to intestinal tumorigenesis.

Three biosynthetic pathways, purine, thymidylate, and methionine synthesis, compete for 5,10-methyleneTHF, a limited cofactor in the cell. C1THF synthase can convert 5,10-methyleneTHF to 10-formylTHF, the substrate required for de novo synthesis of purines (33). Whereas purine deprivation or inhibition of de novo purine synthesis is usually associated with G₁ arrest, cytostasis, and cytotoxicity (34, 35), reduced purine synthesis capacity has also been associated with abnormal DNA repair and DNA mutagenesis in mammalian cells (36, 37) and could thereby play a role in Apc-mediated intestinal tumorigenesis. However, we observed that de novo purine synthesis relative to synthesis by the salvage pathway was not significantly affected by Shmt1 genotype in MEFs and therefore does not likely contribute to the increased tumor numbers observed in Apc^min/+Shmt1^−/+ mice.

5,10-MethyleneTHF can also be irreversibly reduced by methyleneTHF reductase (MTHFR) to form 5-methyTHF, from which a methyl group is transferred to homocysteine by methionine synthase (MS) to form methionine. Methionine can then be S-adenosylated to form AdoMet, the major methyl donor in cellular methylation reactions. The balance between AdoMet and AdoHcy, the substrate and product, respectively, of transmethylation reactions is an indicator of cellular methylation potential. Cellular methylation potential can impact DNA, RNA, and histone methylation patterns and has been associated with aberrant gene expression patterns in cancer. SHMT1 regulates the partitioning of methyl groups between the thymidylate and methionine synthesis pathways by preferentially shuttling its product, 5,10-methyleneTHF, toward thymidylate synthesis, as well as by tightly binding 5-methylTHF making it unavailable to the methionine cycle (14). Apc^min/+Shmt1^−/− mice tended to have increased hepatic AdoMet under folate-deficient conditions; however, unlike our previous findings, the difference was not statistically significant (19). We determined Shmt1-dependent changes in gene expression patterns in hemizygous mice fed the FCD diet, which were minimal, with only 10 genes identified that had significant changes in expression greater than 2-fold in Shmt1^−/+ mice compared with wild-type mice. One interesting note is that the only gene that demonstrated consistent changes in expression, as confirmed by RT-PCR, was Upp1, which is involved in pyrimidine degradation and uridine homeostasis, indicating another mechanism by which SHMT1 status could impact thymidylate synthesis capacity (38). A limitation to our microarray analysis was that it did not determine diet or gene-by-diet effects as mice fed the control diet were not included in the analysis. However, together the data suggest that cellular methylation potential and any associated changes in gene expression did not significantly contribute to tumorigenesis in this mouse model.

Finally, 5,10-methyleneTHF can be utilized by TYMS to methylate dUMP for the de novo synthesis of dTMP. We have shown that SHMT1 preferentially partitions 1C units into the de novo thymidylate biosynthetic pathway (14). Consistent with our previous observations, we saw a decrease in de novo thymidylate synthesis in Shmt1^−/+ mice, which was associated with an increase in uracil incorporation into nuclear DNA. Interestingly, Shmt1^−/+ mice also demonstrated decreased TYMS protein expression. TK1 was induced by the FCD diet in hemizygous mice, but to a lesser extent than that observed in wild-type mice. A reduction in TK1 would result in decreased uptake of thymidine from the salvage pathway, in addition to an overall dampening of the capacity to synthesize dUMP, and subsequently dTMP, whereas a reduction in TYMS would specifically decrease the capacity for de novo thymidylate synthesis. We conclude that impairments in de novo thymidylate biosynthesis, resulting from Shmt1 hemizygosity and its consequent effect on TYMS and TK1 protein levels, exacerbate the tumor phenotype of Apc^min/+mice fed the FCD diet. Observed increases in SHMT1, TYMS, and TK1 in wild-type mice fed the FCD diet or dramatic increases in TYMS and TK1 in Shmt1^−/− mice could serve as a protective mechanism to maintain adequate dTMP synthesis thereby minimizing uracil misincorporation into nuclear DNA and consequent genome instability.

In the Apc^min/+ mouse model of CRC, Shmt1 hemizygosity did not significantly affect either de novo purine synthesis or cellular methylation potential, indicating that these pathways were unlikely to contribute to the increased tumorigenesis observed in Shmt1^−/+ mice. However, changes in dTMP synthesis were significantly and inversely correlated with increased risk for intestinal cancer development in a gene-by-diet interaction. Furthermore, de novo thymidylate synthesis capacity appears to be dynamically regulated by diet and Shmt1 expression. TYMS and TK1 protein levels respond to changes in SHMT1 expression, and TYMS, TK1, and SHMT1 levels respond to dietary folate/choline deficiency. As a result of these dynamic changes, thymidylate synthesis capacity, as indicated by uracil content in nuclear DNA and metabolic flux assays in MEFs, was lowest in Shmt1^−/+ mice, which exhibited the highest tumor number and load in Apc^min/+ mice under FCD conditions.

Loss of heterozygosity of the Apc gene is an early and necessary event for intestinal neoplastic transformation in the Apc^min/+ mice (39). It has previously been shown that colonic folate concentrations correlate with strand breaks in the Apc gene in rats fed a severely folate deficient diet for 5 weeks and that the strand breaks were inversely correlated with steady-state transcript levels of Apc (18). In addition, mild folate depletion in combination with other B vitamins, including vitamins B6 and B12, and methionine, was associated with DNA strand breaks within the Apc mutation cluster region, decreased expression of a number of genes involved in the Wnt signaling pathway, and suppression of apoptosis in colonocytes (30). The increased uracil misincorporation into DNA and subsequent genome instability associated with Shmt1 hemizygosity in combination with dietary folate deficiency accelerated the loss of heterozygosity of the Apc gene and resulted in increased cellular transformation and tumor development. Compensatory mechanisms appear to buffer thymidylate synthesis capacity under folate-restricted conditions, as demonstrated by the upregulation of SHMT1, TYMS, and TK1 in the colon of FCD-fed mice, which provide protection from loss of Apc heterozygosity and tumorigenesis. Together, our data provide a plausible mechanism by which folate-dependent thymidylate synthesis and consequential changes to genomic stability contribute to Apc-mediated intestinal cancer.

No potential conflicts of interest were disclosed.

The authors thank Sylvia Allen and Rachel Slater for technical assistance. They also thank the Cornell Microarray Core and Dr. Wei Wang for performing the hybridization and analysis.

This work was supported by Public Health Service grant CA105440 to P.J. Stover.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.