p53 deficiency induces MTHFD2 transcription to promote cell proliferation and restrain DNA damage

Significance

Cancer cells acquire distinct metabolic adaptations for survival. However, how metabolic remodeling enables cancer cells to survive genomic instability stress remains poorly understood. This report shows that p53 loss or mutation reprograms folate metabolism through MTHFD2. Moreover, MTHFD2 not only influences reactive oxygen species but also unexpectedly participates in nonhomologous end joining (NHEJ) in response to DNA damage. Down-regulation of MTHFD2 preferably induces apoptosis and cell proliferative arrest in p53-deficient cells but protects p53 wild-type cells by inducing the AICAR-mediated AMPK-p53-p21 activation. Collectively, it reveals both metabolism-dependent and -independent functions of MTHFD2 in controlling p53-dependent cell survival and uncovers a new role for MTHFD2 in NHEJ. MTHFD2 depletion may be a promising therapeutic strategy for the treatment of p53-deficient tumors.

Abstract

Cancer cells acquire metabolic reprogramming to satisfy their high biogenetic demands, but little is known about how metabolic remodeling enables cancer cells to survive stress associated with genomic instability. Here, we show that the mitochondrial methylenetetrahydrofolate dehydrogenase (MTHFD2) is transcriptionally suppressed by p53, and its up-regulation by p53 inactivation leads to increased folate metabolism, de novo purine synthesis, and tumor growth in vivo and in vitro. Moreover, MTHFD2 unexpectedly promotes nonhomologous end joining in response to DNA damage by forming a complex with PARP3 to enhance its ribosylation, and the introduction of a PARP3-binding but enzymatically inactive MTHFD2 mutant (e.g., D155A) sufficiently prevents DNA damage. Notably, MTHFD2 depletion strongly restrains p53-deficient cell proliferation and sensitizes cells to chemotherapeutic agents, indicating a potential role for MTHFD2 depletion in the treatment of p53-deficient tumors.

The tumor suppressor p53 regulates expression of a range of genes involved in various biological events, among which metabolic reprogramming appears to be central for the tumor suppression function of p53 (1). The activation of p53 by diverse stresses controls many cellular metabolic processes such as central carbon metabolism (2–4), NADPH homeostasis (4, 5), and the mevalonate pathway that affects biosynthesis of cholesterol and nonsterol isoprenoids (6). Recently, we discovered a role of p53 in regulating ammonia metabolism by suppressing the urea cycle to control polyamine biosynthesis (7). Intriguingly, mild oxidative stress enhances p53 antioxidant capability to maintain cell survival (8, 9). Moreover, transient activation p53 induces metabolic remodeling and p21-dependent cell cycle arrest to enable cell survival during serine starvation (10). Nevertheless, what metabolic pathways are controlled by p53 and the key to p53’s tumor suppression function still require further investigation.

Folate metabolism is a pivotal metabolic process that provides folate intermediates to fuel the one-carbon metabolism. The alteration of folate metabolism or up-regulation expression of one-carbon metabolic enzyme(s) are thought to be associated with high cancer risk, and serine administration or supplementation has potential to foster cancer growth in vivo and in vitro (11, 12). Although folate metabolism is comprised of cytosolic and mitochondrial pathways, most of the cytosolic one-carbon units are derived from mitochondrially catabolized serine, and the role of the cytosolic enzymes in tumorigenesis is unclear (12) (Fig. 1A). By contrast, the mitochondrial one-carbon metabolism is of more interest in cancer, and increased expression of the mitochondrial folate cycle enzymes was found to be significantly correlated with cancer development. Notably, mitochondrial one-carbon metabolism was shown to be critical for supporting quinoid dihydropteridine reductase–mediated reversal of cytosolic tetrahydrofolate (THF) oxidation in cancer cells (13). Induction of MTHFD2 expression by mTOR complex 1 leads to elevated de novo purine synthesis flux in response to growth signals in both normal and tumor cells (14). Interestingly, MTHFD2 may have a role in DNA synthesis, yet the underlying mechanism is unknown.

Fig. 1.

MTHFD2 is highly expressed in human tumors and negatively correlated with wild-type p53 expression. (A) Cytoplasmic and mitochondrial folate metabolic pathways. Metabolic enzymes are indicated in bold. (B) Fold changes of mRNA levels in different tumors versus their normal counterparts. Every dot represents one type of tumor, and the mRNA level for each dot was determined using at least seven patient samples for each type of tumor. The total number of tumor types is 76. Reference SI Appendix, Table S1 for details. The data were mined from ONCOMINE (http://www.oncomine.org/). Data are mean ± SEM, two-tailed Student’s t test. **P < 0.01; ***P < 0.001; ns, not significant. (C) Relative survival tendency of patients bearing high expression of MTHFD2 relative to its parallel isoenzymes, MTHFD1 and MTHFD2L. Survival tendency was calculated as the difference between the 5-y survival rate of patients with high MTHFD2 expression and that of patients with relatively low expression of MTHFD2. Source data were provided (SI Appendix, Table S2). (D) Comparison of up-regulation frequency (%) of MTHFD2 in different types of human tumors harboring wild-type p53 or p53 mutation or deletion as indicated (32–34) (data collected from http://tcga-data.nci.nih.gov/tcga). Detailed statistics are also provided in SI Appendix, Table S3. (E) Immunohistochemical images of MTHFD2 staining in breast and lung tumors with high or normal p53 expression versus those bearing low or no p53 expression. (https://www.proteinatlas.org/.) (F) mRNA levels of MTHFD2, SHMT2, p53, and p21 in HCT116 cells treated with control siRNA (siCtrl) or p53 siRNA (sip53) were analyzed by qRT-PCR. Data are mean ± SEM, n = 3 independent wells per group, two-tailed Student’s t test. **P < 0.01; ***P < 0.001; ns, not significant. (G) Protein expression of MTHFD2 and SHMT2 in p53^+/+ and p53^−/− HCT116 cells treated with increasing amounts of nutlin-3α for 48 h. Relative MTHFD2/Actin ratios are given. (H) Western blot analysis of MTHFD2 expression in pancreas, spleen, lymph node, and bone marrow tissues from p53^+/+ and p53^−/− C57BL/6J mice. n = 5 mice for each group. MTHFD2 expression levels were normalized to actin. (I) ChIP assay for the binding of p53 to the potential p53 response element (MTHFD2-RE, reference SI Appendix, Fig. S2A for detailed sequence) in p53^+/+ and p53^−/− HCT116 cells using anti-p53 DO-1 antibody (Top). 293T cells transfected with control vector or Flag-p53 were analyzed by ChIP assay using anti-Flag antibody (Bottom). Bound DNA was amplified by PCR and quantified. The results are representative of three independent experiments. (J) Protein levels of MTHFD2 in p53 wild-type and p53-null HCT116 cells or p53^−/− HCT116 stably expressing different types of p53 mutants (R175H, R273H, R282W, G245S, and R249S) individually as indicated. R, arginine; H, histidine; W, tryptophan; G, glycine; S, serine. Data are representative of three independent experiments.

Here, we report both metabolism-dependent and -independent functions of MTHFD2 in controlling p53-mediated purine synthesis and cell survival. Briefly, MTHFD2 acts as a transcriptional target for p53, links p53 to folate metabolism, dampens cellular reactive oxygen species (ROS), and contributes to tumor cell proliferation triggered by p53 depletion or mutation. Interestingly, MTHFD2 restrains p53 activity by reducing 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR) to prevent 5′AMP-activated protein kinase (AMPK) activation. Moreover, MTHFD2 unexpectedly participates in NHEJ in a metabolism-independent manner by physically interacting with PARP3. Thus, the reciprocal regulation of MTHFD2 and p53 renders p53-deficient or mutant tumor cells highly susceptible to serine starvation or MTHFD2 deletion.

Results

MTHFD2 Is Frequently Overexpressed in Human Tumors and Negatively Correlated with Wild-Type p53 Expression.

Folate metabolism is critical for tumor cell physiology (15). By systematically analyzing transcriptomics data from various cancer databases, we found that, among the folate-dependent one-carbon metabolic enzymes, MTHFD2 was highly expressed in multiple cancers (Fig. 1B and SI Appendix, Table S1). Furthermore, cancer patients with higher MTHFD2 expression usually displayed relatively poor survival tendency (Fig. 1C and SI Appendix, Table S2). Strikingly, the messenger RNA (mRNA) level of MTHFD2 was frequently higher in tumors harboring p53 mutation(s) than that in wild-type p53 tumors (Fig. 1D and SI Appendix, Table S3). Similarly, immunohistochemical analysis revealed that human tumors with low or absent p53 expression had increased expression of MTHFD2 (Fig. 1E). Thus, these data suggest that mitochondrial MTHFD2 may be crucial for tumor development and/or survival and is positively associated with p53 abnormalities.

Identification of MTHFD2 as a Transcriptional Target for p53.

p53 is a transcriptional factor that controls various cellular activities. To determine whether MTHFD2 is a target for p53, we knocked down p53 in human colon cancer HCT116 cells and human osteosarcoma U2OS cells and examined the expression of enzymes in the folate-mediated one-carbon metabolic pathway. Interestingly, among these enzymes, only the MTHFD2 mRNA level augmented in p53 knockdown cells (SI Appendix, Fig. S1A). Similarly, p53 knockdown increased the mRNA level of MTHFD2 in HCT116 cells as determined by quantitative RT-PCR analysis (Fig. 1F) and elevated MTHFD2 protein abundance in A549 and U2OS cells (SI Appendix, Fig. S1B). Conversely, ectopic expression of p53 led to both a time- and dose-dependent reduction in the MTHFD2 mRNA level in human embryonic kidney (HEK293) cells (SI Appendix, Fig. S1 C and D). Pharmacological activation of p53 by Nutlin-3 decreased the protein level of MTHFD2 (Fig. 1G and SI Appendix, Fig. S1 E and F), while this effect was totally blocked when p53 was absent, suggesting that regulation of MTHFD2 is p53 dependent (Fig. 1G). We also extended our studies into animals. By comparing MTHFD2 expression in p53^+/+ and p53^−/− mice, we found a noticeable increase in MTHFD2 expression in p53-deficient tissues, including pancreas, spleen, lymph node, and bone marrow (Fig. 1H and SI Appendix, Fig. S1H).

In response to DNA damage, p53 becomes activated and induces expression of genes involved in various biological responses. The treatment of cells with DNA damage agent doxorubicin (DOX) resulted in p53 activation as indicated by the increased p21 expression and, correspondingly, decreased MTHFD2 expression at protein and mRNA levels (SI Appendix, Fig. S1 G and I–K). Like Nutlin-3, DOX showed no ability to reduce MTHFD2 expression in p53-deficient cells (SI Appendix, Fig. S1K). Together, these findings indicate that p53 represses the expression of MTHFD2 under both stressed and unstressed conditions.

To determine whether p53 is a transcriptional regulator for the MTHFD2 gene, we analyzed the human MTHFD2 gene sequence for potential p53 response elements (REs), which share the consensus sequence of 5′-RRRCWWGYYY-(0 to 13 base pair [bp] spacer)-RRRCWWG YYY-3′ (where R is a purine, Y a pyrimidine, and W an A or T; ref. 16). As shown in SI Appendix, Fig. S2A, a potential RE was identified. To evaluate whether this RE confers p53-dependent transcriptional repression, we cloned the DNA fragment into the promoter region of a firefly luciferase reporter plasmid. The expression of p53 was able to suppress luciferase expression driven by the genomic fragment containing this RE, and mutant RE substantially impeded this effect (SI Appendix, Fig. S2B). Consistently, chromatin immunoprecipitation (ChIP) assays in HCT116 cells revealed that endogenous p53 bound to the genomic region of RE (Fig. 1 I, Top). Likewise, when expressed in HEK293 cells, Flag-p53 associated with the MTHFD2 genomic DNA (Fig. 1 I, Bottom). These results suggest that p53 binds to the MTHFD2 gene and represses its transcription.

p53 is frequently mutated in human cancers. To examine whether MTHFD2 expression is also influenced by cancer-associated mutant p53, we generated HCT116 cell lines stably expressing different types of human p53 mutants. Like p53-deficient cells, cells with p53 mutation showed higher MTHFD2 expression compared to wild-type cells (Fig. 1J, lane one versus lane two to eight). Coherently, the silencing of mutant p53 had little effect on MTHFD2 expression (SI Appendix, Fig. S2C). Nutlin-3 does not affect mutant p53 activity. Consistent with this, the treatment of MDA-MB-231 cells that harbor a p53 mutation (R280K) with Nutlin-3 did not have a significant change in the expression of MTHFD2 (SI Appendix, Fig. S2D). Likewise, introduction of p53A138V, a mutant p53 that exhibits a mutant conformation at 37 °C and a wild-type–like activity at 32 °C (17), decreased mRNA levels of MTHFD2 at 32 °C compared to that at 37 °C (SI Appendix, Fig. S2E). Collectively, these data indicate that MTHFD2 is a transcriptional target for p53.

p53 Deficiency Reprograms One-Carbon Metabolism and Mitochondrial Respiration through MTHFD2.

One-carbon metabolism catabolizes the nonessential amino acid serine to generate one-carbon formyl units that contribute to purine synthesis. To examine whether p53 loss induces one-carbon metabolism reprogramming through MTHFD2, we used targeted liquid chromatography–tandem mass spectrometry (LC–MS/MS) to measure the relative flux of stable isotope–labeled serine (U-[¹³C]-serine), which can be incorporated into the purine ring (SI Appendix, Fig. S3 A, Top). As serine could be quickly consumed by tumor cells (SI Appendix, Fig. S3B), we cultured p53^+/+ and p53^−/− HCT116 cells with U-[¹³C]-serine for 6 h and observed a substantial proportion of purine nucleotides (IMP, AMP, and GMP) were labeled with ¹³C from U-[¹³C]-serine, whereas pyrimidine (UMP, CMP, and TMP) were not (SI Appendix, Fig. S3C). Interestingly, p53 deficiency increased the abundance of U-[¹³C]-serine–derived AMP and GMP, suggesting that p53 represses de novo purine synthesis from serine (SI Appendix, Fig. S3C).

Serine provides carbons for glycine and the THF cycle, and therefore, de novo synthesis of the purine ring integrates up to four carbons derived from serine: two carbons from glycine and two one-carbon units from the THF cycle (SI Appendix, Fig. S3 A, Bottom). In a U-[¹³C]-serine isotopic tracer experiment, the interconversion of ¹³C-labeled and unlabeled serine and glycine causes fully labeled (m+4) purine nucleotides to represent unadulterated contribution of carbons required for de novo purine synthesis from U-[¹³C]-serine (18). Strikingly, p53 knockout resulted in increased amounts of fully labeled (m+4) (Fig. 2A) as well as overall labeled purine nucleotides (Fig. 2B), and this effect was largely attenuated by MTHFD2 silencing (Fig. 2 A and B). In both p53^+/+ and p53^−/− cells, knockdown of MTHFD2 decreased levels of ¹³C-labled purine nucleotides (Fig. 2 A and B). Serine also supports the methionine cycle and cysteine-associated glutathione (GSH) generation through the THF cycle (19, 20). In line with the purine flux data, p53^−/− cells displayed elevated fluxes of U-[¹³C]-serine to SAH, S-adenosylmethionine (SAM), L-cystathionine, and S-adenosylhomocysteine (GSH), and this phenomenon declined with MTHFD2 ablation (Fig. 2 C–G and SI Appendix, Fig. S3D).

Fig. 2.

p53 regulates one-carbon metabolism through MTHFD2. (A) p53^+/+ and p53^−/− HCT116 cells treated with control siRNA or MTHFD2 siRNA for 48 h (all siRNAs used in this work are at a concentration of 20 nM unless otherwise indicated) and then starved in serine-free medium overnight prior to being cultured in medium containing 0.4 mM U-[¹³C]-serine for another 4 h. Relative ¹³C labeling IMP (m+4), AMP (m+4), and GMP (m+4) levels were analyzed by LC–MS and normalized to cell number. Data are mean ± SEM, n = 3 independent wells per group, two-tailed Student’s t test. **P < 0.01; ***P < 0.001; ns, not significant. (B–G) p53^+/+ and p53^−/− HCT116 cells treated with control siRNA or MTHFD2 siRNA for 48 h and then cultured in medium containing 0.4 mM U-[¹³C]-serine for another 6 h. Percent labeling of purines (B), glycine (C), SAH (D), SAM (E), L-cystathionine (F), and GSH (G) was measured by LC–MS. The isotopic labeling of each metabolite is denoted as m+n, where n is the number of ¹³C atoms. Data are mean ± SEM, n = 3 independent wells per group. (H) p53^+/+ and p53^−/− HCT116 cells stably expressing control plasmid or an RNA-resistant MTHFD2 plasmid were transfected with a control siRNA or MTHFD2 siRNA as indicated. After starvation overnight and then culturing in 0.4 mM U-[¹³C]-serine-supplied medium for 4 h, relative ¹³C labeling IMP (m+4), AMP (m+4), and GMP (m+4) levels were analyzed by LC–MS and normalized to cell number (Left). Protein expression was determined by Western blot analysis (Right). Data are mean ± SEM, n = 3 independent wells per group, two-tailed Student’s t test. *P < 0.05, ***P < 0.001; ns, not significant. (I) p53^+/+ and p53^−/− HCT116 cells were treated as in B through G, and the remaining level of U-[¹³C]-serine in medium was measured. Data are mean ± SEM, n = 3 independent wells per group, two-tailed Student’s t test. *P < 0.05, ***P < 0.001. (J and K) p53^+/+ and p53^−/− HCT116 cells were transfected with control siRNA or MTHFD2 siRNA for 48 h. Levels of cellular formate (J) and AICAR (K) were analyzed, respectively. Data are mean ± SEM, n = 3 independent wells per group, two-tailed Student’s t test. *P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant. (L–O) p53^+/+ and p53^−/− HCT116 cells transfected with control siRNA or MTHFD2 siRNA were cultured in medium containing increasing amounts of formate for 8 h. Relative levels of cellular IMP (inosine 5′-monophosphate) (M), AMP (adenosine 5′-monophosphate) (N), and GMP (guanosine 5′-monophosphate) (O) were determined by LC–MS analysis and normalized to cell number. (L) Protein expression was analyzed by Western blot. Data are mean ± SEM, n = 3 independent wells per group, two-tailed Student’s t test. *P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant.

To further investigate whether the reduced abundance of U-[¹³C]-serine derived of IMP, AMP, and GMP in p53^+/+ cells related to p53^−/− cells was mediated by MTHFD2 and also to rule out off-target small interfering RNA (siRNA) effects, we stably expressed an RNA interference–resistant MTHFD2 in p53^+/+ HCT116 cells expressing MTHFD2 siRNA to a similar level to that found in p53^−/− HCT116 cells (Fig. 2 H, Left). In keeping with this, restoration of MTHFD2 expression increased the abundance of U-[¹³C]-serine–derived purine intermediates to a level comparable to that in p53^−/− cells (Fig. 2 H, Right). Thus, these data indicate that p53 inhibits metabolic flux of serine into purine nucleotides and other metabolites via the mitochondrial one-carbon metabolism pathway.

We next examined serine uptake by comparing the level of serine remaining in medium cultured from p53^+/+ and p53^−/− HCT116 cells expressing MTHFD2 siRNA or control siRNA. Correlating with the findings that p53^−/− cells displayed increased fluxes of serine through the THF cycle (Fig. 2 A–H and SI Appendix, Fig. S3C), a noticeable lower level of serine in medium was observed from p53^−/− HCT116 cells compared to their wild-type counterparts, and MTHFD2 knockdown blocked this effect (Fig. 2I). Thus, it is likely that p53 loss leads to more serine uptake by cells. Mitochondrial one-carbon metabolism determines formate production and AICAR utilization (19, 20). Consistent with this, p53 depletion resulted in increased formate and correspondingly decreased AICAR levels (Fig. 2 J and K), and silencing of MTHFD2 sufficiently reversed these alterations (Fig. 2 J and K). Moreover, MTHFD2 ablation reduced the level of formate and increased AICAR production strongly, regardless of p53 status (Fig. 2 J and K). When exogenous serine is scarce, formate supplementation can replenish the one-carbon pool (21). In keeping with this, providing cells with formate restored the levels of purine nucleotides in MTHFD2-depleted cells (Fig. 2 L–O).

We next wanted to know if diminishing mitochondrial one-carbon metabolism by MTHFD2 inhibition would affect mitochondrial function, particularly which might contribute to p53-mediated mitochondrial respiration. Consistent with previous findings (3), we found an enhanced mitochondrial respiration related to glycolytic metabolism (oxygen consumption rate [OCR]/extracellular acidification rate [ECR] ratio) in p53^+/+ HCT116 cells. Interestingly, in line with p53 expression, ablation of MTHFD2 resulted in a block to mitochondrial spare respiration and an increased ratio of OCR/ECR in both p53^+/+ and p53^−/− cells, and the supplementation of exogenous formate was sufficient to reverse these effects in both cell lines (SI Appendix, Fig. S3 E–G), suggesting that repression of MTHFD2 expression may be involved in p53-mediated mitochondrial respiration. Collectively, it is likely that p53 inactivation induces one-carbon metabolism reprogramming through MTHFD2 up-regulation.

Tumor Cells Harboring p53 Mutation or Depletion Require MTHFD2 for Proliferation and Survival.

Next, we determined if MTHFD2 is required for cell proliferation by calculating the number of cells with BrdU incorporation. Surprisingly, knockdown of MTHFD2 nearly totally inhibited BrdU incorporation in p53^−/− HCT116 cells, while p53^+/+ cells were slightly impacted (Fig. 3A). A similar finding was also observed in other cancer cell lines (A549 and U2OS) (SI Appendix, Fig. S4 A and B). Likewise, MTHFD2 silencing impeded cell proliferation and augmented cell death considerably in p53^−/− HCT116 cells as ascertained by counting cell numbers at different culture time points (Fig. 3B and SI Appendix, Fig. S4C). To generalize these findings, we knocked down MTHFD2 in a range of tumor cell lines and found that, compared to that in p53 wild-type cells, MTHFD2 down-regulation resulted in a strong decline in proliferation of cells harboring p53 mutation or deletion (Fig. 3C). MTHFD2 silencing also promoted cell death in these p53 mutant or deficient cells (SI Appendix, Fig. S4D), and, consistent with this, increased apoptosis was observed in p53^−/− cells with deleted MTHFD2 (SI Appendix, Fig. S4 E and F). Moreover, an evaluation of carboxyfluorescein succinimidyl amino ester (CFSE) dilution revealed that the p53-null cells proliferated dramatically slower than p53 wild-type cells upon MTHFD2 ablation (Fig. 3D). MTHFD2 contributes to cellular NADPH generation and ROS detoxification (22). Interestingly, a higher level of ROS was observed in p53^−/− cells when MTHFD2 was absent related to that in p53^+/+ cells (Fig. 3E). In agreement with these findings, depletion of MTHFD2 attenuated xenograft tumor growth in nude mice, and compared to p53^+/+ tumors, MTHFD2 depletion had a more profound effect on p53^−/− tumors (Fig. 3F).

Fig. 3.

MTHFD2 is required for proliferation and survival of tumor cells harboring p53 mutation or depletion. (A) p53^+/+ and p53^−/− HCT116 cells were transfected with control siRNA or MTHFD2 siRNA for 48 h. Cell proliferation (%) was determined by BrdU incorporation assay. Representative images of BrdU staining are shown (Left). Data are mean ± SEM, n = 3, two-tailed Student’s t test. ***P < 0.001; ns, not significant. (B and D) Proliferation of p53^+/+ and p53^−/− HCT116 cells transfected with control siRNA or MTHFD2 siRNA was determined by counting cells at indicated time points (B) or by CFSE staining followed by flow cytometry analysis at day 6 (D). (C) Multiple cell lines harboring wild-type p53 or p53 mutation or deletion as indicated were transfected by control siRNA or MTHFD2 siRNA for 48 h. Cells were then plated (20,000 cells/well) and cultured for another 6 d. The cell number was determined by counting the Trypan Blue negative cells. Data are mean ± SEM, n = 3. (E) ROS levels in p53^+/+ and p53^−/− HCT116 cells transfected with control siRNA or MTHFD2 siRNA for 48 h were determined by 2′,7′-dichlorodihydrofluorescein diacetate (DCF) staining and subsequent fluorescence-activated cell sorting analysis. (F) Average weights of xenograft tumors derived from p53^+/+ and p53^−/− HCT116 cells treated with control siRNA or MTHFD2 siRNA as indicated. n = 5 tumors for each group. Data are mean ± SEM, two-tailed Student’s t test. *P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant. (G) p53^+/+ and p53^−/− HCT116 cells were transfected with control siRNA or MTHFD2 siRNA for indicated times. Protein expression was determined by Western blot analysis. Data are representative of three independent experiments. (H) p53^+/+ and p53^−/− HCT116 cells transfected with control siRNA or MTHFD2 siRNA for 48 h were cultured in medium containing 0 or 1 mM hydrogen peroxide (H₂O₂) for another 24 h. Cell were then washed and pictured. Representative phase-contrast images of the cells (magnification, 10×) are given. (I) p53^+/+ and p53^−/− HCT116 cells transfected with control siRNA or MTHFD2 siRNA for 48 h were cultured in medium with or without 0.5 mM GSH as indicated for another 24 h. DNA fragmentation was detected by DNA ladder analysis. Data are representative of three independent experiments. (J) Colony formation assay of p53^+/+ and p53^−/− HCT116 cells transfected with control siRNA or MTHFD2 siRNA in the absence or presence of dNTPs (0.01 g/L), formate (2 mM), and/or NAC (1 mM) or GSH (2 mM) as indicated. Numbers of colonies with a diameter greater than 20 μm were quantified (means ± SEM, n = 3) after 2 wk of culturing. Data are mean ± SEM, n = 3, two-tailed Student’s t test. *P < 0.05, **P < 0.01; ns, not significant.

To investigate the mechanism(s) underlying the above observations, we found that MTHFD2 depletion–induced DNA damage, and, correlating with proliferation data (Fig. 3 A–D), more DNA fragments were observed in p53-deficient cells (SI Appendix, Fig. S5A). γ-H2AX is a marker of DNA damage. When DNA damage causes double-stranded breaks (DSBs), γ-H2AX is phosphorylated and localizes to sites of DNA strand breaks. Apparently, p53^−/− cells with MTHFD2 depletion displayed a time-dependent increase in γ-H2AX expression, whereas p53^+/+ cells did not (Fig. 3G). Collectively, these data suggest that MTHFD2 may have a role in maintaining genomic stability and protect p53 mutant or deficient tumor cells from DNA damage.

We further confirmed these findings in p53^−/− HCT116 cells. The silencing of MTHFD2 time dependently increased DNA fragmentation and γ-H2AX expression (SI Appendix, Fig. S5B). Conversely, enforced expression of MTHFD2 reduced the expression level of γ-H2AX (SI Appendix, Fig. S5C).

The above findings indicate that MTHFD2 inhibition may confer p53^−/− cells more sensitive to the induction of DNA damage. To test this, we treated cells with H₂O₂ to induce DNA damage through oxidative stress. When MTHFD2 was knocked down, the supplementation of H₂O₂ markedly impaired p53^−/− cell proliferation, whereas it had a moderate effect on p53^+/+ cells (Fig. 3H). Consistent with this, H₂O₂ treatment dramatically enhanced the level of MTHFD2-depletion–induced DNA fragmentation, particularly in p53^−/− cells (Fig. 3I). Furthermore, DOX strongly promoted γ-H2AX expression (SI Appendix, Fig. S5D) and restrained cell proliferation and colonial formation efficiency in p53-null H1299 cells and p53^−/− HCT116 cells depleted of MTHFD2 (SI Appendix, Fig. S5 E–J). Notably, MTHFD2 reintroduction reversed these effects (SI Appendix, Fig. S5 H–J). Collectively, these findings suggest that p53 mutation or deletion indeed sensitizes tumor cells to MTHFD2 inhibition. We next investigated the mechanism(s) underlining these observations. The inhibition of mitochondrial one-carbon metabolism led to ROS accumulation (Fig. 3E) and restrained purine synthesis (Fig. 2 A, B, and H). The supplementation of deoxyribonucleotides (dNTPs), or formate that could functionally replenish the one-carbon pool for synthesis of purine nucleotides (Fig. 2 M–O), feebly increased proliferation of p53^−/− HCT116 cells with MTHFD2 knockdown (Fig. 3J). Moreover, supplying cells with ROS scavenger N-acetyl-L-cysteine (NAC) or GSH, or together with dNTPs or formate, only partially restored cell proliferation (Fig. 3J). Thus, these findings indicate the existence of other unknown mechanism(s) that may contribute to MTHFD2-mediated cell proliferation and DNA damage prevention.

MTHFD2 Binds to PARP3 and Promotes NHEJ in Response to DNA Damage.

To elucidate potential mechanism(s) of MTHFD2 preventing DNA damage in p53 mutant tumor cells, we fractionated cells and examined the localization of MTHFD2. Besides cytoplasmic localization, a portion of MTHFD2 was also found in the nucleus and present in chromatin fraction (Fig. 4A). These findings promoted us to examine whether MTHFD2 has a previous unrecognized role in regulating DNA damage response to avert DSBs accumulation in the nucleus. Indeed, by using a NHEJ reporter system with different types of DSBs (SI Appendix, Fig. S6A), we found that MTHFD2 depletion significantly reduced NHEJ repair efficiency in p53-deficient or -null cells but not in p53 wild-type cells (Fig. 4B and SI Appendix, S6 B and C). Moreover, depletion of MTHFD2 resulted in a block to chromatin recruitment of XRCC4, a core component of NHEJ (23) (SI Appendix, Fig. S6D). These findings together support the hypothesis that MTHFD2 may positively regulate NHEJ.

Fig. 4.

MTHFD2 binds to PARP3 and promotes NHEJ in response to DNA damage. (A) p53^−/− HCT116 cells were fractionated into chromatin and cytosol and analyzed for the distribution of MTHFD2 by Western blotting. (B) MTHFD2 promotes NHEJ repair efficiency determined by EYFP protein expression. p53^−/− HCT116 cells treated with control siRNA or MTHFD2 siRNA were subsequently transfected with YFP expression vector with complete (not damaged), linearized blunt, compatible, or incompatible ends (see details in SI Appendix, Fig. S6A) for 24 h. YFP protein expression was analyzed by Western blot. Data are representative of three independent experiments. (C) In vitro pull-down interactome screening identifies PARP3 as a binding partner for MTHFD2. Data are representative of three independent experiments. (D) HEK293T cells were transfected EGFP-PARP3 or EGFP vector, and representative immunofluorescent images of GFP and endogenous MTHFD2 staining are shown. The colocalization efficiency of MTHFD2 and GFP was analyzed. Data are representative of three independent experiments. (E) Flag-PARP and Histone H1.2 proteins purified from HEK293T cells were incubated with increasing amounts of purified His-MTHFD2 protein in the presence of 2 µM NAD⁺. PARP ribosylation was measured by Western blot analysis using anti-ADP ribose antibody. Histone H1.2 ribosylation was determined. Data are representative of three independent experiments. (F and G) p53^−/− HCT116 cells stably expressing PARP3 or vector control were transfected with control siRNA or MTHFD2 siRNA for 48 h and exposed to 0 or 200 µM H₂O₂ for another 24 h for Western blot analysis (F) or 72 h for imaging (G). Data are representative of three independent experiments. (H) Enzymatic activities of purified proteins. Proteins were visualized by Coomassie blue staining. Data are mean ± SEM, n = 4. (I) Lysates from HEK293T cells expressing HA-PARP3 alone or together with Flag-MTHFD2 or Flag-MTHFD2 mutant as indicated were subjected to IP using anti-Flag M2 beads followed by Western blot analysis. Data are representative of three independent experiments. (J) H1299 cells transfected with Flag-MTHFD2, Flag-MTHFD2 mutant, or vector control as indicated were treated with 1 µg/mL DOX for 24 h. Protein expression was analyzed by Western blot and relative γ-H2AX band intensity was quantified and normalized to Actin. Data are representative of three independent experiments. (K) p53^−/− HCT116 cells stably expressing PCDH-PARP3 or vector control were transfected with control siRNA or MTHFD2 siRNA and cultured in soft agar (2,000 cells per each condition) in the absence or presence of formate (2 mM), NAC (1 mM), or GSH (2 mM) as indicated for 14 d. The number of colonies with a diameter greater than 20 μm were quantified (means ± SEM, n = 3). Two-tailed Student’s t test. *P < 0.05, **P < 0.01; ns, not significant.

NHEJ is a major pathway for the repair of DSBs in human cells (24). Mechanistically, NHEJ is mediated by a complex containing LIG4, XRCC4, and XLF (23, 25) and accelerated by the PARP3/APLF axis for efficient ligation of damaged DNA ends (26) (SI Appendix, Fig. S6E). In order to pinpoint the primary target(s) for MTHFD2, we ascertained whether MTHFD2 could bind to these key NHEJ repair factors. To this end, we observed that only poly(ADP)ribose polymerase 3 (PARP3) could form a complex with MTHFD2 in pull-down assays (Fig. 4C). Quantitative image analysis revealed substantial colocalization of MTHFD2 and GFP-PARP3, not GFP, in the nucleus (Fig. 4D). To exclude potential artifacts caused by tags, we carried out immunoprecipitation (IP) and pull-down assays using different tags, and similar results were obtained (SI Appendix, Fig. S6 F and G). In addition, it appears that the N-terminal WGR domain of PARP3 mediated its binding to MTHFD2, as PARP3 depleted of this domain lost its capability to interact with MTHFD2 (SI Appendix, Fig. S6H).

PARP3 is a DNA damage–activated ADP ribosyltransferase that catalyzes the NAD⁺-dependent ADP ribosylation. Notably, we found that expression of MTHFD2 promoted the formation of tetramer forms of PARP3 (SI Appendix, Fig. S7A) and enhanced the ADP ribosylation of PARP3 as well as its target histone H1.2 (Fig. 4E and SI Appendix, Fig. S7B), indicating that MTHFD2 may interact with PARP3 and boost PARP3 activity to promote NHEJ repair. In support of this notion, depletion of PARP3 resulted in elevated abundance of γ-H2AX and reduced chromatin recruitment of XRCC4 in p53^−/− HCT116 cells (SI Appendix, Fig. S7 C–E). Consistently, depletion of PARP3 led to impaired cell proliferation, and this effect was augmented remarkably upon H₂O₂ treatment (SI Appendix, Fig. S7F). By contrast, ectopic expression of PARP3 reduced the elevated level of γ-H2AX induced by H₂O₂ treatment (SI Appendix, Fig. S7G). More importantly, γ-H2AX elevation, as well as cell proliferation inhibition induced by MTHFD2 knockdown, could be largely reversed by ectopic expression of PARP3 (Fig. 4F and SI Appendix, Fig. S7 H and I). In echoing with the findings that MTHFD2 silencing sensitized p53-deficient cells to DNA damage, overexpression of PARP3 significantly protected these cells from death triggered by the addition of DNA damage agents H₂O₂ and DOX (Fig. 4G and SI Appendix, Fig. S7J). Notably, upon DNA damage induced by H₂O₂ treatment, enforced expression of PARP3 led to a significant restoration of cell survival of p53^−/− HCT116 cells depleted of MTHFD2 (Fig. 4G). While enforced expression of wild-type MTHFD2 reduced γ-H2AX expression, introduction of a PARP3- and NAD⁺-binding defect MTHFD2 point mutant (R201A) did not reduce, yet to some extent increased, expression of γ-H2AX (SI Appendix, Fig. S7K). To elucidate whether MTHFD2 repression of DNA repair requires its enzymatic activity, we generated two enzymatic activity-dead MTHFD2 mutants, D155A and QLP-ALA (N132A and P134A), which displayed remarkably binding ability to PARP3 (Fig. 4 H and I). Like wild-type MTHFD2, expression of either of these mutants resulted in conspicuously reduced γ-H2AX level in response to DNA damage (Fig. 4J). In comparison, S85A or K88A mutation dampened the binding of MTHFD2 to PARP3 without affecting MTHFD2 enzymatic activity and failed to reduce γ-H2AX expression (Fig. 4 H–J). These findings together suggest that MTHFD2 may prevent DNA damage through PARP3-dependent activation of NHEJ, which is independent of its enzymatic activity. In addition, PARP3 overexpression in combination of ROS scavenger (GSH or NAC) resulted in a further reduction of DNA damage and significant enhancement of anchorage-independent cell growth (Fig. 4K and SI Appendix, Fig. S7L).

p53/p21 Signaling Activation Compromises MTHFD2 Depletion–Induced DNA Damage.

While p53 wild-type cells are more resistant to MTHFD2 deletion, tumor cells harboring p53 mutation or depletion are dependent on MTHFD2 for proliferation and survival, indicating that p53 wild-type cells display other potential mechanism(s) compromising this effect. To examine this, we found that protein levels of phosphorylated p53 (p-Ser15) and p21 highly increased in HCT116 cells upon MTHFD2 silencing (SI Appendix, Fig. S8A). Similar results were observed in A549 cells and U2OS cells (SI Appendix, Fig. S8B). However, inhibition of MTHFD2 expression by siRNA did not affect the p53 transcript (SI Appendix, Fig. S8C). Notably, depletion of MTHFD2 led to nuclear accumulation of p21, suggesting that p21 might directly participate in maintaining genomic stability (Fig. 5A). In support of this idea, depletion of MTHFD2 resulted in elevated abundance of p21 and PARP3 in chromatin fraction (Fig. 5B). Importantly, p21 silencing augmented γ-H2AX levels induced by H₂O₂ treatment (Fig. 5C) and reduced the live cell population upon DOX treatment (SI Appendix, Fig. S8D). Intriguingly, depletion of MTHFD2 did not further increase γ-H2AX amounts in sip21 cells at the basal level (Fig. 5D), indicating that p21 might act downstream of MTHFD2 in DNA damage elimination. Nevertheless, knockdown of p21 and/or MTHFD2 increased DNA fragmentation, and this effect could be enhanced by the treatment of H₂O₂ (Fig. 5E). A similar effect was found on the amount of cellular γ-H2AX in HCT116 cells treated with increasing amounts of DOX (SI Appendix, Fig. S8E). These results were further verified by the observation that reduction of p21 expression by the knocking down of p53 also led to an enhancement of γ-H2AX expression when cells were treated with DOX (SI Appendix, Fig. S8F).

Fig. 5.

A protective role for AICAR-AMPK-p53-p21 axis in MTHFD2 depletion–induced DNA damage. (A) Representative immunofluorescence staining images of p21 expression in p53^+/+ and p53^−/− HCT116 cells transfected with control siRNA or MTHFD2 siRNA for 48 h. Data are representative of three independent experiments. (B) Subcellular localization and protein expression levels of MTHD2, PARP3, and p21 in p53^+/+ and p53^−/− HCT116 cells treated with control siRNA or MTHFD2 siRNA. Data are representative of three independent experiments. (C) p53^+/+ HCT116 cells were transfected with control or p21 siRNA for 48 h in the presence of increasing amounts of H₂O₂. γ-H2AX and p53 activation were analyzed by Western blot. Data are representative of three independent experiments. (D) HCT116 cells were transfected with control (−), p21, and/or MTHFD2 siRNAs for 48 h. Protein expression was analyzed by Western blot. Data are representative of three independent experiments. (E) p53^+/+ HCT116 cells transfected with control (−), p21, and/or MTHFD2 siRNAs for 48 h were treated with increasing amounts of H₂O₂ for another 24 h. DNA fragmentation was detected by DNA ladder analysis. Data are representative of three independent experiments. (F) p53^+/+ HCT116 cells stably expressing PARP3 or vector control (PCDH-V) were transfected with control siRNA or p21 siRNA for 48 h. An equal number of cells was cultured in medium containing 0, 50, 200 µM H₂O₂ for another 3 d. Representative images of the cells are given (Left), and protein expression was analyzed by Western blot (Right). Data are representative of three independent experiments. (G) Formate levels in HCT116 cells (Left) and HEK293 cells (Right) transfected with control siRNA or MTHFD2 siRNA for 48 h were determined by LC–MS. Data are mean ± SEM, n = 3, two-tailed Student’s t test. *P < 0.05. (H) HCT116 cells (Left) and HEK293 cells (Right) were transfected with control or MTHFD2 siRNA in the presence or absence of formate as indicated for 48 h. The cellular AICAR level was measured by LC–MS analysis. Data are mean ± SEM, n = 3, two-tailed Student’s t test. **P < 0.01; ***P < 0.001. (I) Protein expression in p53^+/+ and p53^−/− HCT116 cells transfected with control (−) or MTHFD2 siRNA was analyzed by Western blot using indicated antibodies. Data are representative of three independent experiments. (J) Protein expression in p53^+/+ HCT116 cells treated with control siRNA (−) or increasing amounts of MTHFD2 siRNA for 48 h was analyzed by Western blot using the indicated antibodies. Data are representative of three independent experiments. (K) p53^+/+ HCT116 cells treated with control (−) or MTHFD2 siRNA were grown in medium with or without formate as indicated. AMPK and p53 activation were determined by Western blot analysis using indicated antibodies. Data are representative of three independent experiments. (L) HCT116 and U2OS cells were transfected with control (−), MTHFD2, and/or AMPK siRNAs as indicated for 48 h. Protein expression was analyzed by Western blot. Data are representative of three independent experiments.

p21 expression induces cell cycle arrest and inhibits cell proliferation. Consistent with this, while MTHFD2 silencing suppressed cell proliferation (indicated by BrdU incorporation), knockdown of p21 increased cell proliferation (SI Appendix, Fig. S8G). Double knockdown of MTHFD2 and p21 resulted in a great reduction in cell proliferation (SI Appendix, Fig. S8G). In keeping with the findings in p53^−/− cells (SI Appendix, Fig. S7 E and G), depletion of PARP3 in p53^+/+ HCT116 cells also led to increased γ-H2AX and sensitized cells to DNA damage agents H₂O₂ and DOX (SI Appendix, Fig. S8H), suggesting that PARP3 may be functionally involved in DNA damage regardless of p53 status. Notably, in these wild-type p53 cells, PARP3 reintroduction reduced γ-H2AX levels and partially restored cell proliferation of both sip21 cells and control siRNA-treated cells upon H₂O₂ treatment (Fig. 5F). Together, these results reveal that p53/p21 activation may confer cellular resistance to MTHFD2 depletion by conquering MTHFD2 depletion–induced PARP3 inactivation and, consequently, NHEJ suppression.

MTHFD2 Depletion Stimulates p21 Expression through AICAR-AMPK-p53 Axis Activation.

AMPK is known to be able to phosphorylate and activate p53 (27). To further determine how MTHFD2 affects p53 activity, we examined the effect of MTHFD2 on cellular AICAR, which is a natural activator of AMPK (28). In human cells, one-carbon metabolism generates formate, which is further converted into N¹⁰-formyl THF (15). By incorporation of a carbon donated from N¹⁰-formyl THF, AICAR is catalyzed to yield FAICAR by AICAR transformylase (SI Appendix, Fig. S9A). Thus, suppression of formate generation from one-carbon metabolism would reduce N¹⁰-formyl THF and, as a result, may lead to AICAR accumulation and IMP diminution. Indeed, AICAR dose dependently activated AMPK and p53 (SI Appendix, Fig. S9 B and C), and the depletion of MTHFD2 resulted in decreased IMP production (Fig. 2 A, B, and H). More importantly, consistent with previous studies (19, 20), MTHFD2 silencing led to a reduction in formate generation (Figs. 2J and 5G) and indeed an elevated abundance of AICAR (Figs. 2K and 5H). Markedly, providing cells with formate sufficiently reversed this effect and brought the AICAR level back to normal (Fig. 5H).

In echoing with the above findings, ablation of MTHFD2 led to AMPK activation as evidenced by increased phosphorylation of AMPK and its target ACC in both p53^+/+ and p53^−/− HCT116 cells (Fig. 5I). Moreover, with the increasing amount of MTHFD2 siRNA transfected, there was increasing activation of AMPK and p53 (Fig. 5J and SI Appendix, Fig. S9D). Notably, these effects could be abolished by the addition of exogenous formate (19, 20) (Fig. 5K and SI Appendix, Fig. S9 E–G), and the supplementation of formate dose dependently reduced the phosphorylation of AMPK and p53 as well as the expression of p21 (Fig. 5K and SI Appendix, Fig. S9G). Importantly, the knockdown of AMPK in HCT116 cells and U2OS cells nearly completely impaired the activation of p53 and up-regulation of p21 resulting from the depletion of MTHFD2 (Fig. 5L). By contrast, providing cells with ROS scavenger NAC failed to block the activation of the AMPK-p53-p21 axis induced by MTHFD2 inhibition (SI Appendix, Fig. S9H). Collectively, these data indicate that MTFHD2 influences AMPK-mediated p53 activation and p21 expression through regulating AICAR metabolism, not ROS detoxification.

Discussion

Our findings reveal that p53-deficient cells are highly susceptible to MTHFD2 deletion. Up-regulation of MTFHD2 caused by p53 deficiency enhances one-carbon metabolic flux to purines and NHEJ repair ability, which helps cells to proliferate and survival. Speculatively, in p53 wild-type cells, the effect of p53 on MTHFD2 expression could be counteracted to some extent by some oncogenic signaling (14, 29) so that MTHFD2 can be maintained at certain levels necessary for basal cell proliferation. An interesting observation is that supplementation of either formate or nucleotides showed minimal effect on MTHFD2-depleted cell proliferation (Fig. 3J), and the repression of ROS by treatment of GSH or NAC could only partially restore proliferation of these cells. These findings indicate existence of other mechanism(s) for MTHFD2-mediated cell proliferation. Indeed, further investigations have led us to the discovery of an unexpected role for MTHFD2 in the direct modulation of DNA damage response through binding to PARP3 to accelerate PARP3 ADP ribosylation. MTHFD2 depletion–induced DNA damage might be due to the defect in NHEJ repair, as PARP3 overexpression could largely restrain DNA damage in MTHFD2-depleted cells. Interestingly, expression of an PARP3-binding but enzymatically inactive MTHFD2 mutant (e.g., D155A) noticeably impeded DNA damage (Fig. 4 H–J). Thus, it is likely that the enzymatic activity of MTHFD2 is dispensable for its DNA repair function. MTHFD2 has been found to colocalize with DNA replication sites in the nucleus for several years (30), but how it functions in this organelle still remains mysterious. Our findings, together with others reported recently (31), reveal a metabolic-independent role of MTHFD2 in regulating DNA repair. In addition to MTHFD2, other folate metabolic enzymes such as SHMT1 and MTHFD1 also display partial nuclear localization. Thus, it appears that these folate enzymes may have a metabolism-independent function, for instance, by forming a complex with other nuclear proteins (15), yet how exactly they function in nuclear is unclear and needs substantial further investigation.

p53 has been long known as a regulator for DNA synthesis and DNA damage repair. However, the underlying mechanisms remain poorly understood. The current study reveals a potential experimental explanation that controlling one-carbon metabolism would afford p53 capability to limit purine synthesis and suppress PARP3-mediated NHEJ. Moreover, activation of p53 by MTHFD2 depletion leads to p21 up-regulation and, at least in our hand, results in accumulation of p21 in the nucleus. This demonstrates a possibility in space for p21 to be part of the DNA damage repair machinery, which is functionally distinct from the classic role of p21 as a cell cycle regulator. Nevertheless, whether nuclear accumulation of p21 is directly involved in DNA damage repair needs further exploration.

A previous study reveals that p53-deficient cells rely on serine for survival (10). Interestingly, here we found that MTHFD2 expression is strongly elevated by p53 deficiency, suggesting a physiological mechanism for the increased serine utilization in p53-deficient cells. MTHFD2 up-regulation confers p53^−/− cells more advantages for purine synthesis and redox control by consuming serine. Therefore, from a point of view, these cells are susceptible to MTHFD2 depletion and serine starvation as they rely on serine utilization for survival and proliferation.

In the absence of p53, depletion of MTHFD2 causes DNA damage by reducing NHEJ, which in turns leads to more DNA damage accumulation and increased sensitivity to chemotherapeutic agents. Yet, these effects could be compromised by p21 up-regulation as that in p53 wild-type cells. Serine starvation induction of cell cycle arrest by transient activation of p53-p21 directs depleted serine stores to glutathione synthesis, preserving cellular antioxidant capacity (10). Like serine starvation, the depletion of MTHFD2 resulted in decreased glutathione synthesis and more ROS accumulation in p53^−/− cells. Thus, p21 activation–induced cell cycle arrest may also contribute to the survival of p53^+/+ cells from MTHFD2 depletion by reducing ROS accumulation. In conclusion, our study here uncovers that p53-deficient or mutant tumor cells are susceptible to MTHFD2 depletion, which further sensitizes these cells to chemotherapy and oxidative stress. Thus, clinically, depletion of MTHFD2 may have a potential implication in the treatment of p53-deficient tumors.

Methods and Materials

Antibodies and Reagents.

Antibodies and the specific epitopes used in this study were purchased from the indicated companies: anti-p21 (BD, category no. 556431, 1:1,000 for immunoblotting [IB]), anti-p53 (Santa Cruz, category no. sc126-HRP for IB), anti-Phosphorylated-p53 (Ser15) (Cell Signaling Technology, category no. 9284S, 1:1,000 for IB), anti-AMPKα (Cell Signaling Technology, category no. 2603S, 1:1,000 for IB), anti-phosphor-AMPKα (Thr172) (Cell Signaling Technology, category no. 2535S, 1:2,000 for IB), anti-Phospho-Histone H2A.X (Ser139) (20E3) Rabbit mAb (Cell Signaling Technology, category no. 9718, 1:2,000 for IB), mouse anti-mthfd2 (Santa Cruz, category no. sc-390708, 1:1,000 for IB), mouse anti-PARP3 (Santa Cruz, category no. sc-390771, 1:1,000 for IB), Flag-HRP (Sigma, category no. A8592, 1:3,000 for IB), HA-HRP (Roche, category no. 12013819001, 1:3,000 for IB), Anti-GFP pAb from MBL, anti-Actin (Proteintech, category no. 66009–1-LG, 1:5,000 for IB), Goat anti-rabbit IgG-HRP (Santa Cruz, category no. sc-2004, 1:5,000 for IB) and Goat anti-mouse IgG-HRP (Santa Cruz, category no. sc-2302, 1:5,000 for IB), Donkey anti-Mouse IgG (H+L) ReadyProbes Secondary Antibody, Alexa Fluor 594, (Life Technologies, category no. R37115), Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, and Alexa Fluor 488 (Life Technology, category no. A11008).

The following reagents were purchased from the indicated companies: PFT-α (Sigma, P4359), Nutlin-3 (Sigma, SML0580), DOX (Sigma, D1515), Etopside (Sigma, E1383), anti-FLAG M2 affinity gel (Sigma, A2220), anti-HA magnetic beads (Pierce, 88837), Protein A/G agarose (Pierce, 20421), Lipofectamine 2000 (Thermo Fisher Scientific, 12566014), RNAiMAX transfection agent (Thermo Fisher Scientific, 13778075) and crystal violet (Solarbio, C8470-25). The amino acid serine, glycine, and the nucleotides GMP, AMP, IMP, UMP, TMP, and CMP are purchased from Sigma. Formate (category no. 456020–5g) and AICAR (category no. A9978) are purchased from Sigma. Eagle’s Minimum Essential Medium with ribonucleosides and deoxyribonucleosides (category no. M8042) or without ribonucleosides and deoxyribonucleosides (category no. M4526) is from Sigma.

Cell Culture, siRNA Transfection, and CRISP/Cas9-Mediated Deletion of p53.

The 293T cells, HEK 293 cells, human osteosarcoma U2OS cells, human colon cancer HCT116 cells, and human hepatocellular carcinoma Huh-7 cells were purchased from American Type Culture Collection. p53^+/+ HCT116 and p53^−/− HCT116 cells were kindly gifted by Bert Vogelstein, Johns Hopkins University, Baltimore, MD. HEK 293, 293T, HCT116, and Huh-7 cells were maintained in standard culture medium. U2OS was cultured in Maccoy’s 5A with 10% fetal bovine serum (FBS). All cells were cultured without a penicillin-streptomycin solution. All cell lines were subjected to an examination of mycoplasma contamination and were cultured for no more than 2 mo. siRNA transfections were performed with Lipofectamine RNAiMAX (Life Technologies) using siRNA pools targeting human corresponding genes, respectively. All siRNAs were used at a concentration of 20 nM. For cotransfection experiments, the total siRNA concentration was equalized under all conditions by control siRNA (nontargeting). The siRNA sequences for genes are the following: Human TP53, 5′-GAC​TCC​AGT​GGT​AAT​CTA​C-3′; Human AMPKα, 5′-ACC​AUG​AUU​GAU​GAU​GAA​GCC​UUA​A-3′; Human p21, 5′-CUU​CGA​CUU​UGU​CAC​CGA​G-3′; Human ATM, 5′-CGT​GTC​TTA​ATG​AGA​CTA​CAA-3′; Human ATR, 5′-GCU​UAA​GUC​UGA​UUU​GCU​ATT-3′; Human NMNAT1, 5′-CCU​UUG​CUG​UUC​CCA​AUU​UTT-3′; Human MTHFD2, 5′-GCU​GCG​ACU​UCU​CUA​AUG​UTT-3′; Human PARP3, 5′-GAG​AAG​AAA​UUU​CGG​GAA​ATT-3′; and Human NAMPT, 5′-GCA​UCU​UCC​AAU​AGA​AAU​ATT-3′. The siRNA sequence (5′-CGU​ACG​CGG​AAU​ACU​UCG​ATT-3′) targeting luciferase was used as a control throughout this study.

ChIP.

To search for potential p53 REs in the Mthfd2 gene, we used the Genomatix Promoter Inspector software (http://www.genomatix.de/). In general, the consensus sequence for the p53 RE is 5′-RRRCWWGYYY-(0 to 13 bp spacer)-RRRCWWGYYY-3′, in which R is a purine, Y a pyrimidine, and W either A or T. By analyzing the genome sequence, we found one putative p53 RE for each of them, and the sequence is 5′-AGT​AAA​GGT​CAA​GGT​GTT-​GAC​AGG​ATT​GGT​TTC-3′. ChIP assays were performed as previously described. Briefly, cells were cross-linked with 1% formaldehyde solution for 15 min at room temperature (25 °C). The cross-linking reaction was stopped by the addition of 125 mM glycine and lysed in 1 mL sodium dodecyl sulfate (SDS) lysis buffer for 10 min on ice. Lysates were then sonicated to generate DNA fragments with the average size ∼200 to 1,000 bp. Sonicated samples were spun down and subjected to overnight IP with p53 antibodies or control IgG. The next day, bounded DNA fragments were eluted using low salt immune complex buffer, high salt immune complex buffer, and Tris-EDTA buffer sequentially and then amplified by PCR. The primer pairs were MTHFD2-RE-F: GAC​ACT​CGA​GGA​TCA​CGT​GGT​TGA​AGC​AG and MTHFD2-RE-R: GAT​AAA​GCT​TGA​GCC​GAG​ACT​GTG​CCA​CTG​C.

Semiquantitative RT-PCR and qRT-PCR.

Briefly, total RNA was isolated from triplicate wells in each condition using the Total RNA Purification Kit (GeneMark) and 2 µg RNA of each sample was reversed to complementary DNA (cDNA) by the First-Strand cDNA Synthesis System (Thermo Fisher Scientific). Between 0.2 to 0.4 µg cDNA of each sample was used as a template to perform semiquantitative or qPCR. qPCR was performed on a CFX96 Real-Time PCR System (Bio-Rad), and the amplifications were done using the SYBR Green PCR Master Mix (Gene Star). The fold changes in gene expression were obtained after being normalized to actin. The primers used in this study are listed below: p21-F, CCG​GCG​AGG​CCG​GGA​TGA​G; p21-R, CTT​CCT​CTT​GGA​GAA​GAT​C; p53-F, CACGAGCTGCCCCCAGG; p53-R, TCAGTCGACGTCTGAGT; GAPDH-F, GGT​GGT​CTC​CTC​TGA​CTT​CAA​CA; GAPDH-R, GTT​GCT​GTA​GCC​AAA​TTC​GTT​GT; GENOMIC-ACTB-194-F, CTC​CGA​CCA​GTG​TTT​GCC​TT; GENOMIC-ACTB-194-R, CGC​GGC​GAT​ATC​ATC​ATC​CA; SHMT1-F, CCC​TCC​CCA​TTT​GAA​CAC​T; SHMT1-R, GGG​ATC​CAC​ACT​TTT​CAC​TCC; MTHFD2-F, TAG​GAA​CGA​TGA​GCA​CAA​TGC; MTHFD2-R, ACT​GTT​GAT​TCC​CAC​AAC​TGC; MTHFD1-F, TAG​GAA​CGA​TGA​GCA​CAA​TGC; MTHFD1-R, AGA​CAC​TGG​CCA​GAC​TTT​CAA; SHMT2-F, AGT​CTA​TGC​CCT​ATA​AGC​TCA​ACC​C; SHMT2-R, GCC​GGA​AAA​GTC​GAG​CAG​T; MTHFD1L-F, CCC​TTT​GGT​CGG​AAC​GAT​GA; MTHFD1L-R, TGC​CGA​ACA​CCA​TAC​TCC​AC; MTHFD2L-F, CCA​GGA​GGT​GAT​GCA​ACT​GT; and MTHFD2L-R, TCC​TGT​CAC​TGG​ATC​GTG​GA.

Soft Agar Assay and Xenograft Tumor Model.

For soft agar assay, cells were transfected with control siRNA or MTHFD2 siRNA. After 24 h, cells were suspended in 1 mL Dulbecco’s Modified Eagle Medium (DMEM) with 10% FBS containing a 0.3% agarose and plated on a firm 0.6% agarose base in 12-well plates (2,000 cells per well). Cells were then cultured in a 5% CO₂ incubator at 37 °C for 2 wk. Colonies were fixed with 25% formaldehyde and stained with 0.0125% crystal violet till colonies turned into blue. Colonies were then quantified by counting, and images were obtained.

For xenograft experiments, 4 × 10⁶ control or MTHFD2-depleted cells were injected into the flank of 6- to 7-wk-old male athymic Balb-c nu/nu male mice. After 15 d, mice were euthanized, and the tumor growth was evaluated by weight. All the procedures performed in this study were approved by the Tsinghua University Animal Care and Use Committee (TUACUC). In all xenograft experiments, the maximal total tumor burden for a mouse is 1.5 g. All the tumors’ sizes didn’t exceed the TUACUC-approved maximum size (10% of mouse weight, 1.8 to 2.0 g usually). The p53^−/− mice were generated by Beijing Biocytogen Co., Ltd (please reference SI Appendix for details). All animals were kept according to guidelines and regulations approved by the TUACUC.

LC–MS Analysis of Metabolites.

Cells were seeded in 6-cm dishes and cultured overnight. The following day, cells were washed twice in phosphate-buffered saline (PBS) and cultured in medium without serine (Thermo Fisher Scientific, category no. 11091181) or in complete medium (Thermo Fisher Scientific, category no. C11995500BT) overnight. The next day, we removed the medium and washed the cells using PBS twice and added new medium with ¹³C-labeled serine culturing for another 4 h. Cells were collected and washed with PBS and metabolites extracted using complete methanol. Then, extracts were further spun at 16,000 × g for 15 min and collected supernatants. And the metabolites were subjected to vacuum freeze drying. Positive ionization mode data were acquired using an Agilent 1290/6460 (Agilent) triple quadrupole mass spectrometer. Cell extracts (10 μL) were injected onto a 100 × 2.1 mm 1.9 μm Hypersil GOLD aQ (Thermo) that was eluted at a flow rate of 200 μL/min with initial conditions of 100% mobile phase A (10 mM ammonium acetate in water) followed by a 10 min linear gradient to 100% mobile phase B methanol. The [M +H] of analyte was selected as the precursor ion. The quantification mode was multiple reaction monitoring (MRM) mode using the mass transitions (precursor ions/product ions). The nitrogen generator (PEAK Shanghai) was used for solvent removal and atomization, and high purity nitrogen was used as colliding gas. The sheath gas temperature is 350 °C, sheath gas flows at 10 L/min, the gas temperature is 325 °C, and the gas flow is 8 L/min. The capillary voltage is 4,000 V, the nebulizing gas is 45 psi, and the nozzle voltage is 500 V. Acquisition and processing were performed using Quantitative Analysis Version B.04.00 (Agilent). The collision energies for the different MRM pairs were individually optimized with AMP (m/z 348 > 136, CE 20 V), IMP (m/z 349 > 137, CE 10 V), and GMP (m/z 364 > 152, CE 10 V).

Cell Proliferation and Survival Assays.

For cell proliferation and survival assays, cells were transfected with indicated siRNA or left untreated for 24 h. Then, cells were collected, and the cell number was determined by a TC20 Automated Cell Counter (Bio-Rad). A total of 20,000 cells were distributed for each well of a 6-well plate and duplicated three times for each group. After growth for an indicated number of days, the cells were washed with PBS, typsinized, and counted. For a survival evaluation, cells were stained with 0.4% Trypan Blue for 5 min. Live cells and dead cells were then determined by a TC20 Automated Cell Counter.

BrdU Staining and Imaging.

Cells treated with indicated siRNAs for 48 h were collected and washed twice using PBS. An appropriate number of cells were seeded into a new 6-well plate and cultured overnight. The following day, cells were in a fresh medium containing BrdU at 37 °C for 4 h and were then washed twice with PBS. A total of 1 mL 3.7% formaldehyde in PBS was added into each well and incubated for 15 min at room temperature to fix the cells. Next, the fixing buffer was removed and 1 mL 0.2% Triton X-100 permeabilization buffer was added into each well for another 15 min before further incubation of 1 mL 2N HCl for 10 min at room temperature. BrdU incorporation was then determined and visualized using anti-BrdU antibody followed by imaging on a confocal microscope.

Protein Expression and Purification.

pRK5 plasmids tagged with the Flag or HA epitope coding the full-length human MTHFD2, PARP3, LIG4, XLF, XRCC4, PARP1, or APLF were transfected into HEK-293T cells, respectively. A total of 48 h later, cells were harvested and lysed by sonification in a lysis buffer (50 mM Tris HCl, pH 7.4, with 150 mM NaCl, 1 mM EDTA, and 1% Triton X-100) and then centrifuged at 13,000 × g for 10 min at 4 °C. FLAG M2 or HA beads were used to IP tagged proteins in supernatants according to the manufacturer’s standard procedures. Beads were washed three or five times with wash buffer (20 mM Tris, pH 7.5, 150 mM NaCl, or 1% Triton X-100) at 4 °C on a rotator followed by competitive elution with appropriate synthetic FLAG or HA peptides in tris-buffered saline or PBS solution depending on experimental purposes. Purified proteins were used immediately or stored at −80 °C.

Formate Level Assay.

The formate level was measured using a Formate Colorimetric Assay Kit (category no. K653-100, Biovision). Briefly, cells were collected and homogenized (1 × 10⁶) in 100 µL Assay Buffer. After centrifuging the cells at top speed for 10 min to remove insoluble components, the soluble fraction was assayed directly. A total of 50 µL fraction was transferred into a 96-well plate well, and 50 µL reaction buffer was then added. The reaction was incubated for 60 min at 37 °C, and optical density was measured at 450 nm using a microplate reader.

Western Blot Analysis.

Cells were lysed by using a modified radioimmunoprecipitation assay buffer containing 10 mM Tris HCl at pH 7.5, 150 mM NaCl, 1% Triton X-100, and proteinase inhibitors at 4 °C for 10 to 20 min. Protein samples were quantified using the standard Pierce BCA Protein Assay (Thermo Fisher Scientific) or a BCA Protein Assay Kit (Macgene), boiled in 5× loading buffer and equal amounts of proteins that were resolved by SDS–polyacrylamide gel electrophoresis (10 mL 10% resolving gel including 4 mL H₂O, 3.3 mL 30% acrylamide, 2.5 mL 1.5 M Tris at pH 8.8, 0.1 mL 10% SDS, 0.1 mL 10% ammonium persulfate [APS], and 0.004 mL tetramethylethylenediamine [TEMED]; 10 mL 5% stacking gel including 6.8 mL H₂O, 1.7 mL 30% acrylamide, 1.25 mL 1 M Tris at pH 6.8, 0.1 mL 10% SDS, 0.1 mL 10% APS, and 0.01 mL TEMED), and transferred onto a nitrocellulose membrane. Before immunoblotting with the indicated antibodies, 5% skimmed milk was used to block the membrane. For IP experiments, both input and pull-down samples were loaded together on the same gel, which, to an extent, can be treated as loading controls.

Luciferase Activity Assay.

Briefly, the genomic fragment of MTHFD2 containing either the wild-type or mutant p53-binding region was cloned into a pGL3 basic vector (Promega, catalog no. E1751). The reporter plasmids were then transfected into 293T cells together with a Renilla luciferase plasmid and Flag-p53 or vector control plasmid. A total of 24 h after transfection, luciferase activity was determined using a dual Luciferase Assay System (Promega, catalog no. E1910). Transfection efficiency was normalized on the basis of the Renilla luciferase activity.

Measurements of ROS Levels.

To measure the ROS level, cells were incubated at 37 °C for 30 min in medium containing 10 µM 2′,7′-dichlorodihydrofluorescein diacetate. Cells were then washed twice with PBS, trypsinized, and resuspended in PBS. Fluorescence was immediately measured using a FACScan Flow Cytometer (Becton Dickinson).

CFSE Staining and Cell Proliferation Assay.

Briefly, cells were collected and resuspended in a buffer containing 0.1% bovine serum albumin (BSA) in PBS at a final concentration of 1 to 5 × 10⁶ cells/mL (no more than 1 × 10⁷ cells per milliliter). Then, we added 0.2 µL 5 mM stock CFSE solution per milliliter of cells (final working concentration, 1 µM). Cells were then incubated with dye at 37 °C for 10 min followed by adding five volumes of ice-cold culture media and further incubating for 5 min on ice to stop the staining. Cells were pelleted by centrifugation and washed with fresh medium. A total of 20,000 cells were then loaded into each 6-well plate well. After 5 d of culturing, cells were harvested and analyzed using flow cytometry.

IP Assay.

Briefly, HEK-293T cells were transfected with Flag-tagged wild-type or mutated MTHFD2 plasmid alone or together with HA-tagged PARP3 plasmid for 48 h. Cells were lysed in a lysis buffer (50 mM Tris HCl, pH 7.4, with 150 mM NaCl and 1% Triton X-100) and then centrifuged at 13,000 × g for 10 min at 4 °C. A total of 1 mM NAD⁺ was added into the supernatants prior to IP with FLAG-M2 beads. After overnight incubation at 4 °C, beads were washed three times with a wash buffer (50 mM Tris, pH 7.5, 150 mM NaCl, and 1% Triton X-100, followed by heating at 95 °C for 10 min in 1× SDS loading buffer. Proteins were analyzed by Western blotting. Indirectly, the potential interaction between MTHFD2 and PARP3 is indicated in some databases such as nextprot (https://www.nextprot.org) and the University of California, Santa Cruz Genome Browser (http://genome.ucsc.edu/index.html).

Intracellular Immunofluorescence Staining and Imaging.

Briefly, HEK293 cells transfected with PARP3-YFP or control YFP vector plasmid for 24 h were collected and washed twice with PBS. Cells were then seeded into chamber slides at a density of 2 × 10⁴ cells/well and cultured overnight. The next day, cells were fixed with 4% paraformaldehyde for 15 min and permeabilized with 0.1% Triton X-100 following pretreatment with 2% BSA. Subsequently, cells were stained with anti-MTHFD2 (anti-p21 or anti–γ-H2AX) antibody and DAPI, respectively, followed by the treatment of secondary Ab Alexa Fluor 594–conjugated Donkey Abs specific for mouse IgG. Imaging was performed using a confocal microscope.

OCR.

OCR was analyzed using a XF96 Extracellular Flux Analyzer (Seahorse Bioscience) according to the manufactural instructions. Briefly, cells were cultured in nonbuffered DMEM media with 10 mM glucose, and measurements were obtained after the addition of 2 μM oligomycin and 100 mM 2-DG. The consumption rate was also measured under basal condition.

Statistical Analysis.

No statistical methods were used to predetermine sample size. For animal experiments, enzymatic activity assays, and LC–MS analysis, the authors who did the experiments were blinded to group allocation during data collection and/or analysis. The investigators were not blinded to all other experiments. The results are shown as means ± SEM. All statistical methods used were specified in the figure legends. All statistical analysis was performed and P values were obtained using the GraphPad Prism software 5.0 (GraphPad Software, Inc.). Tests performed with P < 0.05 were considered statistically significant. Statistical significance is shown as *P < 0.05, **P < 0.01, ***P < 0.001.

Data Availability

All data supporting the findings of this study are included in this article and SI Appendix.