One-carbon (1C) metabolism has a key role in metabolic programming with both mitochondrial (m1C) and cytoplasmic (c1C) components. Here we show that activating transcription factor 4 (ATF4) exclusively activates gene expression involved in m1C, but not the c1C cycle in prostate cancer cells. This includes activation of methylenetetrahydrofolate dehydrogenase 2 (MTHFD2) expression, the central player in the m1C cycle. Consistent with the key role of m1C cycle in prostate cancer, MTHFD2 knockdown inhibited prostate cancer cell growth, prostatosphere formation, and growth of patient-derived xenograft organoids. In addition, therapeutic silencing of MTHFD2 by systemically administered nanoliposomal siRNA profoundly inhibited tumor growth in preclinical prostate cancer mouse models. Consistently, MTHFD2 expression is significantly increased in human prostate cancer, and a gene expression signature based on the m1C cycle has significant prognostic value. Furthermore, MTHFD2 expression is coordinately regulated by ATF4 and the oncoprotein c-MYC, which has been implicated in prostate cancer. These data suggest that the m1C cycle is essential for prostate cancer progression and may serve as a novel biomarker and therapeutic target.

Significance:

These findings demonstrate that the mitochondrial, but not cytoplasmic, one-carbon cycle has a key role in prostate cancer cell growth and survival and may serve as a biomarker and/or therapeutic target.

Cell proliferation requires energy, the availability of building blocks for new cellular components, and the ability to maintain cellular redox homeostasis (1). For building block generation and redox homeostasis, amino acid metabolism involving serine and glycine, and the carbon units that they provide, are essential (2). The one-carbon (1C) cycle mediates the folate-mediated transfer of 1C units from donor molecules, mainly serine, to acceptor molecules, such as purines, methionine and thymidylate; this is necessary for essential cellular processes including DNA synthesis, DNA repair, and the maintenance of cellular redox status.

Eukaryotic cells have complementary pathways for 1C metabolism in the cytosol and mitochondria comprising distinct serine hydroxymethyltransferases (SHMT) and methylenetetrahydrofolate dehydrogenases (MTHFD). Although the cytoplasmic 1C pathway (c1C) prevails in nonproliferating somatic tissues, the mitochondrial pathway (m1C) is predominantly active in proliferating cells, as well as in cancer cells (3). In fact, the central player of the m1C cycle, methylenetetrahydrofolate dehydrogenase 2 (MTHFD2), is overexpressed in many different tumor types (4). MTHFD2 is also critical during embryonic development (5), but is typically not expressed in normal adult tissues, except in highly proliferative cells, such as during T-cell lymphocyte activation (4, 6).

Although MTHFD2 is implicated in various cancers, little is known about its potential role in prostate cancer, which is the most frequently diagnosed noncutaneous cancer and the second most common cause of cancer death in men (7). The androgen receptor (AR) plays a key role in normal prostate growth, as well as in prostate carcinogenesis and progression. We previously found that AR signaling, a central driver of prostate cancer, increased expression of activating transcription factor 4 (ATF4; ref. 8). We have recently found that ATF4 has essential prosurvival functions in prostate cancer cells in vitro and in vivo through direct activation of a broad range of genes including key metabolic pathways (9).

Here, we show that ATF4 directly and specifically regulates expression of genes encoding m1C cycle enzymes in prostate cancer cells. Among these, MTHFD2 is critical for prostate cancer growth in vitro and in vivo and may serve as a novel therapeutic target. In addition, the oncoprotein c-MYC interplays with ATF4 in regulating MTHFD2 expression establishing a new mode of action for c-MYC in prostate cancer.

Cell culture

293T, RWPE1, LNCaP, DU145, and 22Rv1 cell lines were purchased from the ATCC. The VCaP, C4-2B, and LNCaP-c-MYC cell lines were kind gifts from Dr. Frank Smit (Radboud University Nijmegen Medical Centre, Nijmegen, the Netherlands), Dr. Lelund Chung (Cedars-Sinai Medical Center, Los Angeles, CA), and Dr. Ian G. Mills (Norwegian Center for Molecular Medicine, University of Oslo, Oslo, Norway), respectively. Cells were routinely maintained in a humidified 5% CO2 and 95% air incubator at 37°C. Prostate cancer cells were cultured in RPMI-1640, and 293T cells in DMEM, containing 10% fetal calf serum, 50 U/mL penicillin, 50 μg/mL streptomycin, and 4 mmol/L L-glutamine (all purchased from BioWhittaker-Cambrex). Where indicated, cells were treated with 30 nmol/L thapsigargin (Tg; Sigma-Aldrich) for 5 hours, unless stated otherwise. All cell lines were used within 15 passages after reviving from the frozen stocks and routinely tested and were free of Mycoplasma contamination.

Ectopic expression of ATF4

ATF4 ORF entry clone was obtained from the Arizona State University plasmid repository (HsCD00073682) and cloned into doxycycline-inducible pLIX_403 destination vector (Addgene #41395), a gift from Dr. David Root, through standard gateway cloning procedure. Viruses were produced by transfecting HEK293T cells with packaging (psPAX2), envelope (pMD2.G), and pLIX403-ATF4 plasmids, using Lipofectamine 3000 reagent. LNCaP cells were then transduced with the harvested lentivirus.

Cell proliferation and viability assays

Briefly, cells were reverse transfected using Lipofectamine RNAiMAX transfection reagent (Thermo Fisher) and plated into 96-well or 6-well plates. Cells in 6-well plates were cultured for the indicated times, trypsinized, stained with trypan blue, and counted using a hemocytometer. The data shown are representative of at least three independent experiments performed in triplicate. Cells plated into the 96-well plates were cultured for 48 hours and cell viability was measured using the CCK-8 kit (Bimake).

Colony formation and prostatosphere assays

Cells were trypsinized, seeded at a density of 5,000 cells per well into 6-well plates, and cultured for 2–3 weeks. The cells were then fixed with methanol and stained with 0.4% crystal violet. Colonies were quantified by extracting crystal violet in 10% acetic acid and measurement of absorbance at 590 nm. Prostatosphere assays were performed as described previously (10). The data shown are representative of at least two independent experiments performed in triplicate.

Quantitative PCR

RNA extraction, reverse transcription, and quantitative polymerase chain reaction (qPCR) were performed as described previously (8). The values were normalized to the relative amount of the internal standard GAPDH, TBP, or ACTB. Results normalized to GAPDH are presented unless indicated otherwise. PCR primer sequences are available upon request. The data shown are representative of at least two independent experiments performed in triplicate.

Western blot analysis

Whole-cell extracts and Western blot analyses were performed by standard methods as described previously (8). The antibodies used were: ATF4 (11815, Cell Signaling Technology), ATF4 (A5514; Bimake), ASNS (146811AP; Proteintech); MTHFD2 (sc-390708), GAPDH (sc-47274), β-actin (sc-47778; Santa Cruz Biotechnology). All antibodies were used at a dilution of 1:1,000, except for MTHFD2 (1:100), GAPDH (1:5,000), and β-actin (1:2,000). The data shown are representative of at least two independent experiments.

Other methods

Descriptions of the cell culture, RNA interference, chromatin immunoprecipitation (ChIP), ChIP-seq, patient-derived xenograft organoids, mitochondrial membrane potential assay, nanoliposomal siRNA targeting in prostate cancer xenografts, IHC, and bioinformatics analysis are available as Supplementary Materials.

Statistical analysis

Mean and standard deviation values were calculated using Microsoft Excel software. The potential effects were evaluated using Student two-sided t test unless indicated otherwise. Values of P < 0.05 were considered as significant. Statistically significant differences are denoted by *, **, and *** indicating P < 0.05, P < 0.01 and P < 0.001, respectively. Error bars indicate SEM.

ATF4 specifically activates the mitochondrial 1C cycle in prostate cancer

To decipher novel ATF4 targets that may play essential roles in prostate cancer, we performed a ChIP-seq experiment and identified ATF4 binding sites in LNCaP cells upon Tg treatment. This analysis revealed 7,488 ATF4 binding sites in close proximity (±1,000 bp) to the transcription start sites of 5,597 protein-coding genes (Supplementary Table S1). There was a significant overlap between these binding sites with those that were previously identified in various tissues or cell lines in the Gene Transcription Regulation Database (GTRD), suggesting effective capture of target sequences (Supplementary Fig. S1A; Supplementary Table S2). To assess the potential functionality of the binding sites, we analyzed these data together with our data from global transcriptomic and proteomic analyses upon siRNA-mediated ATF4 knockdown in Tg-treated LNCaP cells (9). Among the genes identified by ChIP-seq, 29 were downregulated in both transcriptomic and proteomic analyses (Fig. 1A; Supplementary Table S3). In addition to well-established ATF4 target genes, such as asparagine synthetase (ASNS) and phosphoserine phosphatase (PSPH), two 1C metabolism genes, MTHFD2 and MTHFD1L were among these 29 genes. Another 1C metabolism gene, SHMT2, also harbored an ATF4 binding site and was among the downregulated genes in the microarray experiment (Fig. 1A). Intriguingly, although mammalian 1C metabolism is comprised of two parallel pathways (cytosolic and mitochondrial) with almost identical core enzymatic capabilities (11), all three identified ATF4-regulated 1C metabolism genes belong to the mitochondrial pathway (Fig. 1B). Notably, the identified ATF4 binding sites in the vicinity of m1C cycle genes overlapped with ATF4 binding sites that have been reported in the GTRD, and were highly conserved among various mammalian genomes (Fig. 1C). In contrast, expression of two c1C metabolism genes (SHMT1 or MTHFD1) was not affected; consistently, they did not harbor any ATF4 binding sites near their TSS (Supplementary Fig. S1B). Consistent with this observation, in the great majority of the 16 distinct cancer types, ATF4 or and its target ASNS mRNA levels were highly correlated with m1C, but not c1C, enzyme gene expression (Supplementary Fig. S1C). Moreover, analyses of the protein and mRNA expression data from the Cancer Cell Line Encyclopedia data set showed significant correlations between ATF4 protein levels and expression of the m1C, but not c1C enzymes (Supplementary Fig. S2). In fact, in this data set, after Sperm flagellar protein 2 (SPEF2; its functional role in prostate cancer, if any, is currently not known), MTHFD2 is the second gene with the highest correlation to protein levels of ATF4. Similarly, there was a high correlation between ASNS and m1C cycle enzyme gene expression. Evaluation of 1C metabolism gene expression in the Oncomine database revealed that m1C pathway genes are more consistently upregulated in diverse cancers compared with c1C genes (Fig. 1D). Additionally, CRISPR-based cancer cell line dependency profiles of the 1C metabolism genes showed a significant correlation between the mitochondrial, but not cytosolic, members of the pathway (Fig. 1E). These data suggested that ATF4 is involved in mediating m1C metabolism gene expression in prostate cancer.

Figure 1.

ATF4 binds to and regulates expression of mitochondrial but not cytoplasmic 1C metabolism genes. A, Venn diagram depicting the number of genes affected by the different analyses (microarray gene expression in yellow, mass spectrometry (MS) in blue, and ChIP-seq in light red) and their intersections, upon ATF4 knockdown. B, A simplified diagram showing c1C and m1C metabolic pathways. C, ChIP-seq analyses of ATF4 target genes reveal strong binding sites in the vicinity of genes that encode m1C cycle enzymes. The GTRD lane shows the number of ATF4 peak calls and their location based on this database. Conservation track shows conservation among vertebrate genomes (phastCons scores). D, Summary of 1C metabolism gene regulation in cancer versus normal tissues in the Oncomine database (top 5% threshold). Darker color in each box indicates larger number of studies and thus more significant association. E, Diagram showing the correlation of cancer cell line dependency profiles between m1C metabolism genes and their functionally relevant neighbors. In contrast to the others, ATF4-regulated m1C cycle genes are more strongly linked to each other (red lines). m1C and c1C cycle genes are represented by blue and red boxes, respectively. The numbers in the ovals on the connectors show Pearson correlation coefficients between the linked genes. The red lines have a Pearson correlation coefficient that is greater than 0.15.

Figure 1.

ATF4 binds to and regulates expression of mitochondrial but not cytoplasmic 1C metabolism genes. A, Venn diagram depicting the number of genes affected by the different analyses (microarray gene expression in yellow, mass spectrometry (MS) in blue, and ChIP-seq in light red) and their intersections, upon ATF4 knockdown. B, A simplified diagram showing c1C and m1C metabolic pathways. C, ChIP-seq analyses of ATF4 target genes reveal strong binding sites in the vicinity of genes that encode m1C cycle enzymes. The GTRD lane shows the number of ATF4 peak calls and their location based on this database. Conservation track shows conservation among vertebrate genomes (phastCons scores). D, Summary of 1C metabolism gene regulation in cancer versus normal tissues in the Oncomine database (top 5% threshold). Darker color in each box indicates larger number of studies and thus more significant association. E, Diagram showing the correlation of cancer cell line dependency profiles between m1C metabolism genes and their functionally relevant neighbors. In contrast to the others, ATF4-regulated m1C cycle genes are more strongly linked to each other (red lines). m1C and c1C cycle genes are represented by blue and red boxes, respectively. The numbers in the ovals on the connectors show Pearson correlation coefficients between the linked genes. The red lines have a Pearson correlation coefficient that is greater than 0.15.

Close modal

We next examined 1C metabolism gene expression upon siRNA-mediated knockdown of ATF4 in three independent prostate cancer cell lines. ATF4-specific siRNAs effectively hindered expression of well-known ATF4 target genes, such as phosphoglycerate dehydrogenase (PHGDH), phosphoserine aminotransferase 1 (PSAT1), and PSPH in the LNCaP, VCaP, and 22Rv1 cell lines (Fig. 2A; Supplementary Fig. S3A and S3B), whereas Tg-mediated stimulation of ATF4 expression increased their expression (Fig. 2B). Consistent with the ChIP-seq experiment, m1C gene expression for SHMT2, MTHFD2, and MTHFD1L was decreased upon ATF4 silencing and increased upon Tg-mediated ATF4 activation (Fig. 2A and B; Supplementary Fig. S3A and S3B). In contrast, neither ATF4 silencing nor ATF4 induction significantly affected expression of c1C cycle genes SHMT1 and MTHFD1 (Fig. 2A and B; Supplementary Fig. S3A and S3B). Moreover, in a previously published RNA-seq data set (12), treatment of LNCaP cells with tunicamycin (Tm—an inducer of ER stress/ATF4 pathway through inhibition of N-linked glycosylation in the endoplasmic reticulum) also resulted in the expression of m1C genes in an ATF4-dependent manner (Fig. 2C). We validated the Tm-induced upregulation of m1C genes in LNCaP and PC3 cell lines (Supplementary Fig. S3C and S3D). In Tg-treated cells, ATF4 knockdown effectively reduces MTHFD2 protein levels (Fig. 2D), and ectopic expression of ATF4 effectively rescues it back to control levels (Fig. 2E). In contrast, in nonstressed LNCaP cells, knockdown of ATF4 does not alter MTHFD2 mRNA or protein levels, as under these conditions, despite significant mRNA levels, ATF4 is not translated into protein (Fig. 2F, bottom). However, ectopic ATF4 expression effectively increased both mRNA and protein levels of MTHFD2. Interestingly, without knockdown of endogenous ATF4 expression, ectopically introduced ATF4 mRNA was not detectable (Fig. 2F, top). This was due to the downregulation of endogenous ATF4 mRNA expression upon induction of ectopic ATF4 and suggested an autoregulatory negative feedback loop by ATF4 on its transcription. Consistent with these findings, ATF4 binding to the vicinity of the MTHFD2 gene was verified by ChIP, and ATF4 silencing abolished this interaction (Fig. 2G). Taken together, these data establish ATF4 as a key regulator of the m1C, but not the c1C, cycle gene expression.

Figure 2.

ATF4 induces expression of mitochondrial but not cytoplasmic 1C cycle enzymes. A, siRNA-mediated ATF4 knockdown decreases expression of serine synthesis and m1C cycle enzyme gene expression, but not that of c1C cycle. LNCaP cells were transfected with either control or two independent ATF4-specific siRNAs, treated with Tg (300 nmol/L for 5 hours), and were analyzed by qPCR. The efficiency of ATF4 knockdown was confirmed by Western blot analysis as shown in the inset. B, Expression of genes encoding serine synthesis and 1C cycle enzymes was analyzed by qPCR upon treatment of LNCaP cells with Tg (30 nmol/L) for the indicated time points. Inset shows ATF4 protein levels upon Tg treatment in the time course experiment. C, Data extracted from a RNA-seq experiment that was published previously (12). LNCaP cells were transfected with nontargeting siRNA (siCTRL) or two distinct ATF4-targeting siRNAs. Cells were then either treated with DMSO as control or tunicamycin (2.5 μg/mL for 18 hours). RNA was isolated and subjected to RNA-sequencing. D, ATF4 knockdown decreases MTHFD2 protein expression in multiple prostate cancer cell lines. Indicated cell lines were transfected with either control siRNA or two independent siRNAs targeting ATF4. Cells were treated with Tg (30 nmol/L, 5 hours), harvested, and used in Western blot analysis. E, Downregulation of m1C gene expression upon ATF4 knockdown is rescued by ATF4 reexpression. Doxycycline-inducible LNCaP-ATF4 cells were transfected either with scrambled siRNA or an ATF4-specific siRNA targeting the 5′-UTR of the gene (siATF4); 4 days later, cells were treated with Tg (30 nmol/L for 5 hours). Doxycycline-mediated induction of ATF4 effectively restored MTHFD2 levels as analyzed by qPCR. F, Under non-stressed conditions, despite its high mRNA expression, ATF4 is not translated to a significant protein level; hence, its knockdown fails to downregulate MTHFD2 levels. Doxycycline-inducible LNCaP-ATF4 cells were transfected with either scrambled siRNA or two independent siRNAs targeting the open reading frame (siATF4 #1) or 5′-UTR (siATF4 #2) of the ATF4 mRNA. At the same time with the transfection, cells were treated with indicated amounts of doxycycline to induce ATF4 expression and processed after 48 hours. Note that ectopic ATF4 expression (upon Dox treatment) effectively upregulates MTHFD2 levels while downregulating endogenous ATF4 levels (detected by UTR-specific primers). ASNS expression, a well-characterized ATF4 target gene, is shown as a reporter of ATF4 activity. #, P < 0.001; ⁁, P < 0.01; *, P < 0.05; ns, nonsignificant. G, Individual ChIP analysis verified ATF4 binding to the intronic region (chr2:74,426,212–74,426,487—hg19) of the MTHFD2 gene identified in the ChIP-seq experiment. LNCaP cells were transfected with either scrambled siRNA or ATF4-specific siRNA, treated with vehicle or Tg, and ChIP assay was performed using an ATF4-specific antibody. ns, nonsignificant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 2.

ATF4 induces expression of mitochondrial but not cytoplasmic 1C cycle enzymes. A, siRNA-mediated ATF4 knockdown decreases expression of serine synthesis and m1C cycle enzyme gene expression, but not that of c1C cycle. LNCaP cells were transfected with either control or two independent ATF4-specific siRNAs, treated with Tg (300 nmol/L for 5 hours), and were analyzed by qPCR. The efficiency of ATF4 knockdown was confirmed by Western blot analysis as shown in the inset. B, Expression of genes encoding serine synthesis and 1C cycle enzymes was analyzed by qPCR upon treatment of LNCaP cells with Tg (30 nmol/L) for the indicated time points. Inset shows ATF4 protein levels upon Tg treatment in the time course experiment. C, Data extracted from a RNA-seq experiment that was published previously (12). LNCaP cells were transfected with nontargeting siRNA (siCTRL) or two distinct ATF4-targeting siRNAs. Cells were then either treated with DMSO as control or tunicamycin (2.5 μg/mL for 18 hours). RNA was isolated and subjected to RNA-sequencing. D, ATF4 knockdown decreases MTHFD2 protein expression in multiple prostate cancer cell lines. Indicated cell lines were transfected with either control siRNA or two independent siRNAs targeting ATF4. Cells were treated with Tg (30 nmol/L, 5 hours), harvested, and used in Western blot analysis. E, Downregulation of m1C gene expression upon ATF4 knockdown is rescued by ATF4 reexpression. Doxycycline-inducible LNCaP-ATF4 cells were transfected either with scrambled siRNA or an ATF4-specific siRNA targeting the 5′-UTR of the gene (siATF4); 4 days later, cells were treated with Tg (30 nmol/L for 5 hours). Doxycycline-mediated induction of ATF4 effectively restored MTHFD2 levels as analyzed by qPCR. F, Under non-stressed conditions, despite its high mRNA expression, ATF4 is not translated to a significant protein level; hence, its knockdown fails to downregulate MTHFD2 levels. Doxycycline-inducible LNCaP-ATF4 cells were transfected with either scrambled siRNA or two independent siRNAs targeting the open reading frame (siATF4 #1) or 5′-UTR (siATF4 #2) of the ATF4 mRNA. At the same time with the transfection, cells were treated with indicated amounts of doxycycline to induce ATF4 expression and processed after 48 hours. Note that ectopic ATF4 expression (upon Dox treatment) effectively upregulates MTHFD2 levels while downregulating endogenous ATF4 levels (detected by UTR-specific primers). ASNS expression, a well-characterized ATF4 target gene, is shown as a reporter of ATF4 activity. #, P < 0.001; ⁁, P < 0.01; *, P < 0.05; ns, nonsignificant. G, Individual ChIP analysis verified ATF4 binding to the intronic region (chr2:74,426,212–74,426,487—hg19) of the MTHFD2 gene identified in the ChIP-seq experiment. LNCaP cells were transfected with either scrambled siRNA or ATF4-specific siRNA, treated with vehicle or Tg, and ChIP assay was performed using an ATF4-specific antibody. ns, nonsignificant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

MTHFD2 is critical for prostate cancer cell growth in vitro and in vivo

Because MTHFD2 has a key role in the m1C cycle (13) and was previously implicated in cancer (4), we assessed whether it affects prostate cancer cell growth. siRNA-mediated MTHFD2 silencing effectively reduced its mRNA and protein levels (Fig. 3A). MTHFD2 knockdown was maximal (∼80%) at 72 hours after transfection and remained significantly downregulated for more than one week, but returned to basal levels by day 12 (Supplementary Fig. S4A). Short-term MTHFD2 knockdown significantly reduced the viability (Supplementary Fig. S4B) and long-term knockdown nearly abolished both viability and colony formation ability of LNCaP, DU145, VCaP, and 22Rv1 cells (Fig. 3BD). Furthermore, MTHFD2 knockdown significantly hindered LNCaP and DU145 prostatosphere growth (Fig. 3E). However, viability of a normal prostate epithelial cell line, RWPE1, was not affected by MTHFD2 knockdown (Fig. 3F).

Figure 3.

MTHFD2 knockdown inhibits growth of prostate cancer cells, PDX-derived organoids, and tumor xenografts. A, MTHFD2 knockdown efficiency was determined in LNCaP and DU145 cells that were transfected with either scrambled siRNA or two independent MTHFD2-specific siRNAs for 48 hours by both qPCR and Western blot analyses. B, LNCaP, VCaP, or 22Rv1 cells transfected with control or MTHFD2-specific siRNAs were cultured for the indicated times, and cell numbers were determined by trypan blue staining. C–E, MTHFD2 knockdown hinders colony (C and D) and prostatosphere (E) formation ability of prostate cancer cells. Indicated cells were transfected with control siRNA or MTHFD2-specific siRNAs and cultured for two weeks. Colonies formed were stained and quantified as described in Materials and Methods. Prostatospheres were pictured and counted under a light microscope; representative areas are presented. F, Viability of normal prostate cells is not affected by MTHFD2 knockdown. RWPE1 cells were transfected with either control siRNA or two independent MTHFD2-targeting siRNAs. After 48 hours, relative cell viability and MTHFD2 expression were determined using the CCK8 assay and qPCR, respectively. G, MTHFD2 expression and knockdown efficiency in various LuCaP organoids were assessed by Western blot analysis. H, MTHFD2 knockdown significantly decreased LuCaP organoid formation without affecting cell death in LuCaP 35 and LuCaP 96 that express MTHFD2, but not in LuCaP 136, which does not express MTHFD2. I, Nanoliposomal systemic delivery of MTHFD2-specific siRNA profoundly inhibited the growth of VCaP and 22Rv1 xenograft tumors in vivo. VCaP and 22Rv1 cells were implanted subcutaneously in male nude mice. Once tumors were palpable, mice (n = 5 per group) were given either empty nanoliposomes or MTHFD2-specific siRNA as described in Materials and Methods. Tumor volumes were measured at the indicated time points. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 3.

MTHFD2 knockdown inhibits growth of prostate cancer cells, PDX-derived organoids, and tumor xenografts. A, MTHFD2 knockdown efficiency was determined in LNCaP and DU145 cells that were transfected with either scrambled siRNA or two independent MTHFD2-specific siRNAs for 48 hours by both qPCR and Western blot analyses. B, LNCaP, VCaP, or 22Rv1 cells transfected with control or MTHFD2-specific siRNAs were cultured for the indicated times, and cell numbers were determined by trypan blue staining. C–E, MTHFD2 knockdown hinders colony (C and D) and prostatosphere (E) formation ability of prostate cancer cells. Indicated cells were transfected with control siRNA or MTHFD2-specific siRNAs and cultured for two weeks. Colonies formed were stained and quantified as described in Materials and Methods. Prostatospheres were pictured and counted under a light microscope; representative areas are presented. F, Viability of normal prostate cells is not affected by MTHFD2 knockdown. RWPE1 cells were transfected with either control siRNA or two independent MTHFD2-targeting siRNAs. After 48 hours, relative cell viability and MTHFD2 expression were determined using the CCK8 assay and qPCR, respectively. G, MTHFD2 expression and knockdown efficiency in various LuCaP organoids were assessed by Western blot analysis. H, MTHFD2 knockdown significantly decreased LuCaP organoid formation without affecting cell death in LuCaP 35 and LuCaP 96 that express MTHFD2, but not in LuCaP 136, which does not express MTHFD2. I, Nanoliposomal systemic delivery of MTHFD2-specific siRNA profoundly inhibited the growth of VCaP and 22Rv1 xenograft tumors in vivo. VCaP and 22Rv1 cells were implanted subcutaneously in male nude mice. Once tumors were palpable, mice (n = 5 per group) were given either empty nanoliposomes or MTHFD2-specific siRNA as described in Materials and Methods. Tumor volumes were measured at the indicated time points. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

To further evaluate the potential effects of MTHFD2 on prostate cancer growth, organoids of LuCaP patient-derived xenograft (PDX) models were used (14). Three of the six analyzed LuCaP organoids expressed high levels of MTHFD2 expression (Fig. 3G). Three organoids, two with high expression and one with low expression, were analyzed further. These three PDX models, LuCaP 35, LuCaP 96, and LuCaP 136, express AR, lack PTEN expression, and were developed from lymph node metastasis, localized prostate cancer, and adenocarcinoma cells from ascites, respectively (14). siRNA-mediated MTHFD2 knockdown effectively suppressed formation of LuCaP35 and LuCaP96 organoids that express high levels of MTHFD2 without inducing cell death (Fig. 3G and H; quantification is presented in Supplementary Fig. S4C). On the other hand, LuCaP 136 organoid that expresses very low levels of MTHFD2 was not affected by MTHFD2 knockdown.

To assess the therapeutic potential of MTHFD2 inhibition in vivo, we performed xenograft experiments as previously described (15, 16). VCaP or 22Rv1 cells were subcutaneously injected into male nude mice. Upon formation of palpable tumors, empty nanoliposomes or those that carry MTHFD2-specific siRNA were administered by intraperitoneal injection and tumor growth was monitored over time. Whereas tumors continued to grow rapidly in mice injected with the empty nanoliposomes, injection of nanoliposomes containing MTHFD2-specific siRNA dramatically inhibited tumor growth in both models (Fig. 3I). Nanoliposomal siMTHFD2 delivery was well tolerated and did not result in any weight loss (Supplementary Fig. S4D). Together with the findings from above, these data suggest that MTHFD2 is critical for prostate cancer growth and may serve as a novel therapeutic target.

MTHFD2 expression is upregulated in human prostate cancer specimens

Both mRNA and protein expression of MTHFD2 were robust in the normal prostate cell line RWPE1 and in all of the prostate cancer cell lines tested, with some variability in the level of expression (Fig. 4A). We next evaluated MTHFD2 expression in 24 human prostate cancer specimens and their corresponding benign tissues from the same patients using IHC. MTHFD2 expression was significantly increased in prostate cancer compared with benign specimens (Fig. 4B). This observation was verified by an independent tissue microarray cohort consisting of 860 prostate cancer and 223 benign prostate specimens (Fig. 4C). In this large cohort, MTHFD2 expression also correlated with the Gleason score, indicating that it may have prognostic value. Importantly, ATF4 and MTHFD2 protein expression was correlated in a sample subset of this tissue microarray (Fig. 4D). These data show that MTHFD2 expression is significantly increased in prostate cancer compared with normal tissue.

Figure 4.

MTHFD2 expression is increased in prostate cancer. A, Basal levels of MTHFD2 mRNA and protein were determined in various prostate cancer cell lines and the normal prostate cell line by qPCR and Western blot analysis, respectively. B, MTHFD2 expression was analyzed by IHC in matched benign prostate and prostate cancer specimens from 24 patients. Representative images and quantification of staining are shown. C, Tissue microarrays with normal prostate (n = 223) and primary prostate tumors (n = 860) were analyzed by IHC. Middle panel shows increased MTHFD2 expression with increasing Gleason grade of the samples. Representative images and quantification of staining intensity are shown. D, In a subset of the samples used for IHC analysis shown in Fig. 5C, the correlation of ATF4 and MTHFD2 staining scores is depicted. r denotes Pearson correlation between the two stainings. n.s., nonsignificant; **, P < 0.01; ***, P < 0.001.

Figure 4.

MTHFD2 expression is increased in prostate cancer. A, Basal levels of MTHFD2 mRNA and protein were determined in various prostate cancer cell lines and the normal prostate cell line by qPCR and Western blot analysis, respectively. B, MTHFD2 expression was analyzed by IHC in matched benign prostate and prostate cancer specimens from 24 patients. Representative images and quantification of staining are shown. C, Tissue microarrays with normal prostate (n = 223) and primary prostate tumors (n = 860) were analyzed by IHC. Middle panel shows increased MTHFD2 expression with increasing Gleason grade of the samples. Representative images and quantification of staining intensity are shown. D, In a subset of the samples used for IHC analysis shown in Fig. 5C, the correlation of ATF4 and MTHFD2 staining scores is depicted. r denotes Pearson correlation between the two stainings. n.s., nonsignificant; **, P < 0.01; ***, P < 0.001.

Close modal

m1C gene expression signature is strongly associated with prostate cancer prognosis

To assess whether ATF4-regulated m1C cycle gene expression could serve as a potential prognostic biomarker for prostate cancer, we analyzed MTHFD2, SHMT2, MTHFD1L, and MTHFD2L expression in five independent prostate cancer cohorts in the Oncomine database (17–21). MTHFD2 expression was consistently and significantly upregulated in primary and metastatic prostate cancer compared with benign samples (Fig. 5A). SHMT2 was upregulated in three of the five cohorts (Fig. 5B). Only three cohorts had expression data for MTHFD1L and two for MTHFD2L. MTHFD1L was significantly upregulated in the primary and metastatic tumors whereas MTHFD2 L was upregulated in one of the two cohorts (Supplementary Fig. S5). In addition, in The Cancer Genome Atlas (TCGA) data set, m1C metabolism gene expression, but not the cytosolic counterparts, were more prominently upregulated in primary tumor samples (Fig. 5C). These data led us to evaluate whether m1C metabolism gene expression may have prognostic value. Indeed, a gene expression signature consisting of the three m1C cycle enzymes (MTHFD2, MTHFD1L, and SHMT2) was significantly associated with recurrence-free survival in prostate cancer patients (Fig. 5D). Taken together, these observations suggest that activation of ATF4-regulated m1C metabolism could serve as a prognostic biomarker in prostate cancer.

Figure 5.

Mitochondrial 1C gene expression is deregulated in prostate cancer and a gene expression signature derived from it is strongly associated with prostate cancer prognosis. A and B, Expression of MTHFD2 and SHMT2 is upregulated in primary and/or metastatic prostate cancer compared with benign samples. Expression data were retrieved from the Oncomine database. 1, Singh et al. (21); 2, Taylor et al. (17); 3, Vanaja et al. (18); 4, Lapointe et al. (19); 5, La Tulippe et al. (20); N, normal; P, primary; M, metastatic. C, m1C but not c1C cycle genes are upregulated in primary prostate cancer samples in the TCGA dataset. D, A gene expression signature consisting of the three ATF4-regulated m1C cycle genes (MTHFD2 + MTHFD1L + SHMT2) is significantly associated with recurrence-free survival in the TCGA dataset. Prostate cancer samples were ordered based on the expression of the signature genes, and a Kaplan–Meier graph was drawn based on the survival data for top and bottom ∼36% of the samples. NS, nonsignificant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 5.

Mitochondrial 1C gene expression is deregulated in prostate cancer and a gene expression signature derived from it is strongly associated with prostate cancer prognosis. A and B, Expression of MTHFD2 and SHMT2 is upregulated in primary and/or metastatic prostate cancer compared with benign samples. Expression data were retrieved from the Oncomine database. 1, Singh et al. (21); 2, Taylor et al. (17); 3, Vanaja et al. (18); 4, Lapointe et al. (19); 5, La Tulippe et al. (20); N, normal; P, primary; M, metastatic. C, m1C but not c1C cycle genes are upregulated in primary prostate cancer samples in the TCGA dataset. D, A gene expression signature consisting of the three ATF4-regulated m1C cycle genes (MTHFD2 + MTHFD1L + SHMT2) is significantly associated with recurrence-free survival in the TCGA dataset. Prostate cancer samples were ordered based on the expression of the signature genes, and a Kaplan–Meier graph was drawn based on the survival data for top and bottom ∼36% of the samples. NS, nonsignificant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

mTOR/ATF4 signaling regulates m1C expression in normal prostate but not in prostate cancer

Previous work has shown that mTORC1 and PERK/eIF2A signaling regulates cell metabolism by controlling ATF4 levels (22, 23). To assess the impact of mTORC1 signaling on m1C metabolism in prostate cancer, we investigated whether there is a correlation between the expression of three ATF4-regulated m1C metabolism genes and those that are specifically regulated by the mTORC1 signaling cascade. Genome-wide transcriptional alterations were previously determined upon treatment of wild-type or ATF4 knockout human embryonic kidney (HEK) cells with Torin 1, a potent ATP-competitive inhibitor of mTOR (22). We used these data to identify genes that are exclusively regulated by mTORC1 signaling without input from the PERK/eIF2α/ATF4 cascade. Sixty genes were significantly deregulated by more than ±1.7-fold upon Torin 1 treatment in both cell lines. Because some of these genes could still be regulated and influenced by ATF4, whose expression is in fact also low in normal cells, we further narrowed this list by filtering out known or potential ATF4 target genes that were compiled from various public databases, resulting in 38 genes as ATF4 independent targets of mTOR signaling (Fig. 6A).

Figure 6.

Mitochondrial 1C cycle gene expression is differentially regulated in benign and prostate cancer tissues. A, m1C cycle gene expression correlates with ATF4-independent target genes of mTORC1 signaling in normal prostate, but not in prostate cancer samples. Pearson correlation coefficients between indicated genes were calculated in the GTEx data set containing 106 normal prostate tissue samples and TCGA database containing 426 primary prostate cancer samples. Genes in green and red were upregulated and downregulated, respectively, upon Torin 1 treatment. B, Distinct sets of genes correlate with MTHFD2 in normal and prostate cancer samples. Venn diagram shows the number of shared genes that correlate with MTHFD2 in the SEEK, TCGA, or GTEx data sets. In each data set, the top 500 genes that correlate with MTHFD2 expression were included in the analyses. C, ER stress–mediated induction of ATF4 overrides mTORC1/ATF4-mediated MTHFD2 expression. LNCaP cells were treated with 30 nmol/L Tg (for 8 hours) and/or 100 nmol/L rapamycin for 24 hours and expression of m1C cycle genes was determined by qPCR. Bottom panel shows Western blot analysis verifying the activity of the compounds. D, Enrichment analysis of top 500 genes that correlate with MTHFD2 expression in the TCGA data set. E, Prevalence of various mutations among the top and bottom 180 (∼36%) samples in the TCGA data set based on the expression of the m1C cycle gene signature. F, Venn diagram showing proteins that correlate with the three m1C cycle enzyme expression and two mitochondrial DNA encoded proteins (MT-CO1 and MT-ATP8; r > 0.5). G, Gene set enrichment analysis of 136 proteins that were shared between the MTHFD2 and SHMT2 groups identified in F.

Figure 6.

Mitochondrial 1C cycle gene expression is differentially regulated in benign and prostate cancer tissues. A, m1C cycle gene expression correlates with ATF4-independent target genes of mTORC1 signaling in normal prostate, but not in prostate cancer samples. Pearson correlation coefficients between indicated genes were calculated in the GTEx data set containing 106 normal prostate tissue samples and TCGA database containing 426 primary prostate cancer samples. Genes in green and red were upregulated and downregulated, respectively, upon Torin 1 treatment. B, Distinct sets of genes correlate with MTHFD2 in normal and prostate cancer samples. Venn diagram shows the number of shared genes that correlate with MTHFD2 in the SEEK, TCGA, or GTEx data sets. In each data set, the top 500 genes that correlate with MTHFD2 expression were included in the analyses. C, ER stress–mediated induction of ATF4 overrides mTORC1/ATF4-mediated MTHFD2 expression. LNCaP cells were treated with 30 nmol/L Tg (for 8 hours) and/or 100 nmol/L rapamycin for 24 hours and expression of m1C cycle genes was determined by qPCR. Bottom panel shows Western blot analysis verifying the activity of the compounds. D, Enrichment analysis of top 500 genes that correlate with MTHFD2 expression in the TCGA data set. E, Prevalence of various mutations among the top and bottom 180 (∼36%) samples in the TCGA data set based on the expression of the m1C cycle gene signature. F, Venn diagram showing proteins that correlate with the three m1C cycle enzyme expression and two mitochondrial DNA encoded proteins (MT-CO1 and MT-ATP8; r > 0.5). G, Gene set enrichment analysis of 136 proteins that were shared between the MTHFD2 and SHMT2 groups identified in F.

Close modal

In the GTEX database, representing normal prostate samples, expression of majority of the 38 genes were well correlated with m1C gene expression in the expected direction (e.g., the genes inhibited by Torin 1 positively correlated with the investigated genes; Fig. 6A). However, such a correlation was absent in TCGA data set, representing primary prostate cancer. Consistently, there was a significant overlap between the top 500 genes that correlated with MTHFD2 expression in the SEEK database that contains 78 microarray-based prostate cancer gene expression data sets, and the TCGA prostate cancer data set (188 genes, Fig. 6B). However, only 62 genes were shared between GTEX and SEEK, GTEX and TCGA, or in all three data sets. These 62 genes were not enriched for any specific signaling pathway, but contained genes for several translation factors (EEF1B2, EIF2S2, EIF3J, and EIF4EBP1), genes that encode proteins involved in RNA transport (STRAP, THOC7, and RAN), and the gene encoding mitochondrial folate transporter SLC25A32 (Supplementary Table S4). These results indicated that m1C metabolism is distinctly regulated between normal prostate and prostate cancer. In the normal prostate, m1C metabolism appears to be primarily regulated by mTORC1 signaling, but in prostate cancer another signaling pathway(s) may override this regulation. Indeed, in LNCaP cells, Tg-mediated induction of ER stress effectively supersedes mTORC1-mediated MTHFD2 regulation; in contrast to basal conditions, mTORC1 inhibitor rapamycin failed to downregulate ATF4-regulated m1C metabolism gene expression upon Tg treatment (Fig. 6C). These data suggest that m1C metabolism is differentially regulated in prostate cancer compared with normal prostate.

c-MYC is a key mediator of m1C gene expression

To assess which additional pathways could be involved in the regulation of m1C metabolism, we performed gene set enrichment analysis on the top 500 genes that correlate (Pearson r > 0.4) with the m1C gene expression signature in the TCGA data set (Supplementary Table S5). As expected, genes involved in aminoacyl-tRNA biosynthesis, 1C metabolism, cell cycle, and mitotic nuclear division were enriched among the correlated genes (Fig. 6D). In addition, according to both ENCODE and ChEA databases, genes that are associated with c-MYC were exceptionally highly enriched (P = 8.96e−90 in ENCODE, 8.6e−44 in ChEA). Indeed, c-MYC expression itself significantly correlated with the m1C gene signature (r = 0.43). Moreover, the 188 genes that correlated well with MTHFD2 in both SEEK and TCGA databases (Fig. 6B) were also highly enriched for c-MYC-mediated regulation (P = 1.15e−50 in ENCODE, P = 1.18e−17 in ChEA; Supplementary Table S4). These results suggested that c-MYC may be involved in mediating the effects of m1C gene expression in prostate cancer.

We next determined whether m1C gene expression signature may be associated with mutational events in prostate cancer. Comparison of the mutation prevalence between the top and bottom 180 samples based on the expression of the m1C gene expression signature in the TCGA data set revealed several enriched copy-number alterations and point mutations (Fig. 6E). Increased representation of 8p11 and 8p21–22 deletions, and c-MYC amplification among the signature-high group indicated that c-MYC is involved in the regulation of m1C gene expression (Fig. 6E; ref. 24). 8p11 locus harbors the SFRP1 gene that encodes a Wnt signaling inhibitor that is frequently inactivated in a variety of malignancies, including prostate cancer, and has been identified as an essential molecule in c-MYC-dependent transformation (25, 26). In the TCGA data set, SFRP1 expression negatively correlated with those of ATF4 (r = −0.35), ASNS (r = −0.33), and m1C gene signature (r = −0.28), suggesting that loss of SFRP1 could regulate m1C metabolism via ATF4 signaling. Similarly, loss of the 8p21–22 locus is one of the most frequent chromosomal aberrations in prostate cancer and harbors the NKX3.1 gene that encodes a transcription factor, which acts as a tumor suppressor by opposing c-MYC transcriptional activity (27). We further investigated potential association of these mutations on m1C gene expression by assessing expression of the three m1C genes in the mutated versus wild-type samples in the TCGA data set (Supplementary Fig. S6). All three m1C cycle genes were significantly upregulated in all of the eight investigated mutant subgroups. Taken together, these data suggest that mutational events in prostate cancer could activate m1C gene expression through c-MYC.

We next examined m1C gene expression at the protein level using the recently reported proteomics data from 375 cancer cell lines (28). Because the number of mitochondria per cell would vary among different cancer cell lines and can skew the results, we used the expression of mitochondrial encoded proteins MT-CO1 and MT-ATP8 to filter out proteins that correlate with m1C metabolism gene expression simply due to varying numbers of mitochondria in the different cell lines. There was a significant overlap between the expression of proteins that correlated with MTHFD2 and SHMT2 expression, and these did not coincide with mitochondrial proteins (Fig. 6F; Supplementary Table S6). In contrast, although almost one-fourth of the proteins that correlated with MTHFD1L expression were mitochondrial proteins, MTHFD1L did not correlate well with MTHFD2 or SHMT2 protein expression; this suggests that in contrast to MTHFD1L, MTHFD2 and SHMT2 expression is regulated independently from mitochondrial biogenesis. Interestingly, mRNA processing, splicing, transport, and mitochondrial translation-related proteins were highly enriched among the 136 proteins that were shared between the MTHFD2 and SHMT2 groups (Fig. 6F). These proteins were also enriched for harboring a nearby c-MYC binding site in their genes (P = 1.23e−13), further indicating that c-MYC is a prominent player in the regulation of m1C metabolism at the protein level (Fig. 6G).

Under stress conditions, ATF4 counterbalances c-MYC loss to drive m1C gene expression.

A recent study has identified an intricate regulation of protein synthesis through c-MYC–ATF4 cooperation, where c-MYC was involved in the regulation of EIF4EBP1 expression in an ATF4-dependent manner (29). We thus considered the possibility that ATF4 and c-MYC may bind to adjacent or neighboring sites and coordinately regulate m1C cycle gene expression. Intriguingly, data from ENCODE indicated that all three ATF4-regulated m1C genes also harbor c-MYC binding sites that are in close proximity, in fact almost overlapping, with the identified ATF4 binding sites (Supplementary Fig. S7). Moreover, the analysis of a previously published ChIP-seq experiment in LNCaP cells revealed c-MYC binding sites that overlap with those of ATF4 that we have identified in the m1C genes and the EIF4EBP1 gene (Fig. 7A; ref. 30). c-MYC binding to these sites was enriched upon its ectopic expression. Furthermore, analysis of publicly available data sets revealed that m1C gene expression was modulated upon ectopic expression or CRISPRi-mediated inhibition of c-MYC (Fig. 7B and C). In these data sets, modulation of m1C gene expression could not be attributed to an alteration upon cellular stress as in contrast to the report by Tameria et al., indicators of ER stress (such as DDIT3, HERPUD1, ERLEC1, PDIA4, PDIA6) were not deregulated by c-MYC induction or knockdown (Supplementary Fig. 8A–B; ref. 29). We verified the effect of c-MYC on MTHFD2 expression in LNCaP cells, where siRNA-mediated c-MYC knockdown resulted in a significant decrease in MTHFD2 at both mRNA and protein levels (Fig. 7D).

Figure 7.

c-MYC co-occupies ATF4 response elements and regulates m1C cycle gene expression. A, c-MYC ChIP-seq analyses in LNCaP cells revealed colocalization of ATF4 binding sites that were identified in the proximity of m1C cycle genes and EIF4EBP1. The data were obtained from GSE73994, in which doxycycline-inducible LNCaP-c-MYC cell line was used. Orange triangles show the location of ATF4 binding sites identified in our ChIP-seq analysis (Fig. 1D). Please note that for all four genes, in contrast to other nearby sites, the c-MYC binding at the ATF4 target site was relatively more increased upon c-MYC expression. B, CRISPRi-mediated inhibition of MYC expression downregulates m1C cycle enzyme genes and EIF4EBP1 expression in the 22Rv1 prostate cancer cell line. The data were obtained from GSE142808. C, Results of a microarray study showing induction of m1C enzyme gene and EIF4EBP1 expression upon ectopic c-MYC expression in LNCaP cells. The data were obtained from GSE73917. D,MYC knockdown effectively downregulates m1C cycle enzyme genes and EIF4EBP1 expression. LNCaP cells were transfected with siCTRL or sic-MYC (an siRNA pool), and 48 hours later processed for qPCR and Western blot analyses. E, Under stress conditions, ATF4 counterbalances the c-MYC loss to drive m1C gene expression. LNCaP cells were transfected with control, ATF4 and/or c-MYC-specific siRNAs, and 48 hours later treated with DMSO or 30 nmol/L Tg for 5 or 24 hours. ATF4, MYC, and MTHFD2 expressions were determined by qPCR. Under the same conditions, protein levels of c-MYC, ATF4, and MTHFD2 were measured by Western blot analysis. F, Schematic depiction of m1C and c1C cycle gene expression under normal prostate and in prostate cancer. Under normal conditions, ATF4 protein expression is high enough only under transient stress conditions and it specifically induces m1C expression, whereas c-MYC drives both c1C and m1C expression upon transient growth-promoting signals. In tumors, there is chronic stress that activates ATF4 signaling and m1C gene expression. Likewise, various chronic signals, such as proliferative signaling and gene amplification, increase c-MYC expression in prostate cancer tumors that results in higher levels of m1C and c1C gene expression. Having input from both ATF4 and c-MYC, m1C gene expression is markedly higher than that of the c1C cycle in prostate cancer. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 7.

c-MYC co-occupies ATF4 response elements and regulates m1C cycle gene expression. A, c-MYC ChIP-seq analyses in LNCaP cells revealed colocalization of ATF4 binding sites that were identified in the proximity of m1C cycle genes and EIF4EBP1. The data were obtained from GSE73994, in which doxycycline-inducible LNCaP-c-MYC cell line was used. Orange triangles show the location of ATF4 binding sites identified in our ChIP-seq analysis (Fig. 1D). Please note that for all four genes, in contrast to other nearby sites, the c-MYC binding at the ATF4 target site was relatively more increased upon c-MYC expression. B, CRISPRi-mediated inhibition of MYC expression downregulates m1C cycle enzyme genes and EIF4EBP1 expression in the 22Rv1 prostate cancer cell line. The data were obtained from GSE142808. C, Results of a microarray study showing induction of m1C enzyme gene and EIF4EBP1 expression upon ectopic c-MYC expression in LNCaP cells. The data were obtained from GSE73917. D,MYC knockdown effectively downregulates m1C cycle enzyme genes and EIF4EBP1 expression. LNCaP cells were transfected with siCTRL or sic-MYC (an siRNA pool), and 48 hours later processed for qPCR and Western blot analyses. E, Under stress conditions, ATF4 counterbalances the c-MYC loss to drive m1C gene expression. LNCaP cells were transfected with control, ATF4 and/or c-MYC-specific siRNAs, and 48 hours later treated with DMSO or 30 nmol/L Tg for 5 or 24 hours. ATF4, MYC, and MTHFD2 expressions were determined by qPCR. Under the same conditions, protein levels of c-MYC, ATF4, and MTHFD2 were measured by Western blot analysis. F, Schematic depiction of m1C and c1C cycle gene expression under normal prostate and in prostate cancer. Under normal conditions, ATF4 protein expression is high enough only under transient stress conditions and it specifically induces m1C expression, whereas c-MYC drives both c1C and m1C expression upon transient growth-promoting signals. In tumors, there is chronic stress that activates ATF4 signaling and m1C gene expression. Likewise, various chronic signals, such as proliferative signaling and gene amplification, increase c-MYC expression in prostate cancer tumors that results in higher levels of m1C and c1C gene expression. Having input from both ATF4 and c-MYC, m1C gene expression is markedly higher than that of the c1C cycle in prostate cancer. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

We next investigated whether ATF4 and c-MYC could synergistically regulate MTHFD2 expression in prostate cancer cells. To that end, we modulated the levels of the two transcription factors by siRNA-mediated knockdown and assessed MTHFD2 expression, under basal and Tg-treated conditions. Under non-stressed conditions, c-MYC knockdown effectively reduced mRNA and protein levels of MTHFD2, whereas ATF4 knockdown had a slight effect (Fig. 7E). In contrast, under Tg-induced stress conditions, MTHFD2 expression was clearly dependent on ATF4 as Tg induction, in a remarkable fashion, completely inhibited c-MYC expression, and ATF4 knockdown effectively decreased MTHFD2 levels (Fig. 7E). These data suggest that, along with ATF4, c-MYC is a key component of the regulatory network that affects m1C gene expression.

Metabolic reprogramming is critical for cancer cell growth and dissemination (31–33). Some key determinants of this cellular rewiring have been established, but there is an urgent need to identify the molecular mechanisms at play that can be targeted for novel therapeutic approaches. We have recently found that ATF4 is critical for prostate cancer growth in vitro and in vivo (9). We now show that ATF4 makes significant contributions to metabolic reprogramming of prostate cancer cells by significantly increasing m1C cycle gene expression (MTHFD2, MTHFD1L, and SHMT2), without affecting expression of their cytoplasmic counterparts (MTHFD1 and SHMT1). In particular, we found that ATF4-driven deregulation of MTHFD2 expression promotes prostate cancer cell proliferation in vitro and in vivo. Consistently, a gene expression signature based on the m1C cycle has significant prognostic value for prostate cancer progression.

Previous studies have shown that ATF4 regulates metabolic pathways connected to amino acid uptake, tRNA synthesis, and transport (34, 35). In particular, ATF4 is a major regulator of the serine biosynthesis pathway that is upregulated and is associated with poor prognosis in various cancers (2, 36–38). Conversion of serine into glycine and formate is catalyzed by the m1C cycle enzymes in three metabolic steps (Fig. 1B); by concurrently regulating these two cascades, ATF4 enables de novo synthesis of purines, thymidylate, and glutathione, which are essential for rapidly proliferating cells.

It is currently not known what could be the benefit for cancer cells to preferentially use the m1C cycle, rather than the cytosolic counterpart. The products of the m1C cycle, formate and glycine, are transported to the cytosol, where formate is metabolized back to 10-Formyltetrahydrofolate (CHO–THF) to serve as a substrate for purine synthesis (39). Although the c1C cycle could also drive purine synthesis by yielding CHO–THF, the m1C pathway provides the dominant flux for purine synthesis (40). However, neither formate nor glycine treatment was able to recover the viability of prostate cancer cells upon MTHFD2 knockdown (Supplementary Fig. S9), suggesting that m1C metabolism plays other essential roles beyond supplying the building blocks for nucleotide biosynthesis.

One possibility in this regard is the potential contribution of m1C cycle to the energy and redox demands of proliferating cells by generation of ATP and NADH/NADPH (3, 41). In the m1C cycle, the reaction catalyzed by MTHFD2 is a significant source of NADH, which can be used in oxidative phosphorylation to generate 2.5 ATPs, and the reaction catalyzed by MTHFD1L itself generates an ATP molecule giving an overall yield of 3.5 ATPs per cycle (3). Moreover, NADH production by the m1C cycle could contribute to prostate cancer development by enhancing the antioxidant defense of cancer cells (42, 43). Consistently, inhibition of MTHFD2 and SHMT2 expression has been reported to disturb redox homeostasis and impair cell survival under hypoxic conditions in colorectal cancer and glioma, respectively (42, 44). However, N-acetyl cysteine, a potent antioxidant, failed to rescue viability of prostate cancer cells upon MTHFD2 knockdown, suggesting that altered redox homeostasis may not be the primary reason for MTHFD2 knockdown-mediated cell death (Supplementary Fig. S9).

MTHFD2 also participates in the formation of formylmethionyl transfer RNA (fMet) that is required for the initiation of protein synthesis in the mitochondria, thereby regulating mitochondrial protein translation (Fig. 1B), suggesting that its inhibition may impair mitochondria genesis and/or biology. However, in preliminary experiments we did not observe any decrease in the number of mitochondria or mitochondrial membrane potential upon MTHFD2 knockdown (Supplementary Fig. S10A–S10C); nevertheless, it is still possible that the growth advantage that is contributed by MTHFD2 to prostate cancer cells is due to its effects on mitochondrial homeostasis.

Previous research has shown that the key regulator of protein synthesis mTORC1 can activate ATF4 through mechanisms distinct from its canonical induction by stress cascades (22, 45). However, according to a recent study, only a very small subset (61 genes) of ATF4 regulated genes are actually induced by both mTOR/ATF4 and PERK/ATF4 cascades (46). All three mitochondrial 1C cycle enzymes (MTHFD2, SHMT2, and MTHFD1L) were among these genes, but not the two cytosolic counterparts (MTHFD1 and SHMT1). This study was performed on normal mouse embryonic fibroblasts, suggesting that ATF4-mediated metabolic regulation is not specific to cancer cells.

Furthermore, our analysis on the GTEx and TCGA data sets suggested that in prostate cancer tumors, cascades other than mTORC1/ATF4 signaling are also involved in regulation of the m1C cycle (Fig. 6A). In particular, c-MYC–associated gene expression was highly correlated with m1C gene expression in prostate cancer. Together with the other data we present here, this suggested that ATF4 and c-MYC may coordinately regulate m1C gene expression. Consistent with this, the ATF4 binding sites in the vicinity of m1C enzymes from our ChIP-Seq analysis coincide with those of c-MYC that were identified earlier (Fig. 7A; ref. 30). Furthermore, both mRNA and protein levels of MTHFD2 were inhibited upon c-MYC knockdown. These observations are intriguing as c-MYC is an established oncoprotein for prostate cancer, and we have recently identified it as a downstream target and mediator of the IRE1-XBP1s arm of the unfolded protein response (UPR; refs. 10, 47). Our data thus establish a new role of c-MYC in modulating the UPR and a potential novel mode of cross-talk between the IRE1–XBP1s and PERK–eIF2α–ATF4 signaling.

Based on our findings herein and recently published studies, we suggest the following model (Fig. 7F): Under normal conditions, ATF4 regulates m1C gene expression, whereas c-MYC is involved in regulating both m1C and c1C gene expression. Under conditions of some types of stress, c-MYC expression is downregulated, and ATF4 can compensate for this to sustain the expression of the m1C cycle gene expression, whereas c1C gene expression remains low. However, in tumors, various mechanisms, such as c-MYC gene amplification and activation of the IRE1/XBP1 cascade, can keep c-MYC expression high, resulting in further elevated levels of m1C gene expression, which will satisfy the metabolic needs of the cancer cell in cooperation with UPR-mediated ATF4 signaling. These data thus establish that UPR activation can induce m1C cycle by promoting both ATF4 and c-MYC expression. There may be other points of interaction of c-MYC with the UPR in prostate cancer to establish autoregulatory mechanisms, such as c-MYC heterodimerization with XBP1s to activate the IRE1–XBP1s pathway, which then activates c-MYC expression, as observed in breast cancer cells (48, 49; for a review, see ref. 50). Thus, there appears to be feedback loops that are likely to be responsive to environmental cues and determine the outcome of the interactions between ATF4 and c-MYC signaling, which converge on activation of m1C expression.

In summary, our findings establish an interplay between ATF4 and c-MYC to drive m1C cycle gene expression as a critical component for prostate cancer growth. As exemplified by the dramatic tumor inhibitory effects of MTHFD2 targeting in vivo and the robust prognostic value of the m1C gene signature, future work should further evaluate the m1C cycle as a potential biomarker and/or therapeutic target in prostate cancer.

No disclosures were reported.

N. Pällmann: Formal analysis, investigation, methodology, writing–original draft. K. Deng: Investigation, methodology, writing–review and editing. M. Livgård: Formal analysis, investigation, methodology. M. Tesikova: Investigation and methodology. Y. Jin: Investigation. N.S. Frengen: Investigation. N. Kahraman: Investigation. H.M. Mokhlis: Investigation. B. Ozpolat: Supervision and methodology. W. Kildal: Investigation. H.E. Danielsen: Resources and data curation. L. Fazli: Formal analysis, investigation, and visualization. P.S. Rennie: Investigation. P.P. Banerjee: Formal analysis, investigation, visualization, writing–original draft. A. Üren: Formal analysis and investigation. Y. Jin: Investigation and methodology. O.F. Kuzu: Conceptualization, formal analysis, supervision, validation, investigation, methodology, writing–original draft, writing–review and editing. F. Saatcioglu: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing.

The authors thank all members of the F. Saatcioglu laboratory for helpful discussions. This work was supported by grants from the Norwegian Research Council (#193337), Norwegian Cancer Society (#419204), and Health South East Norway (#36024) to F. Saatcioglu.

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.

1.
Locasale
JW
,
Cantley
LC
. 
Metabolic flux and the regulation of mammalian cell growth
.
Cell Metab
2011
;
14
:
443
51
.
2.
Locasale
JW
. 
Serine, glycine and one-carbon units: cancer metabolism in full circle
.
Nat Rev Cancer
2013
;
13
:
572
83
.
3.
Meiser
J
,
Vazquez
A
. 
Give it or take it: the flux of one-carbon in cancer cells
.
FEBS J
2016
;
283
:
3695
704
.
4.
Nilsson
R
,
Jain
M
,
Madhusudhan
N
,
Sheppard
NG
,
Strittmatter
L
,
Kampf
C
, et al
Metabolic enzyme expression highlights a key role for MTHFD2 and the mitochondrial folate pathway in cancer
.
Nat Commun
2014
;
5
:
3128
.
5.
Di Pietro
E
,
Sirois
J
,
Tremblay
ML
,
MacKenzie
RE
. 
Mitochondrial NAD-dependent methylenetetrahydrofolate dehydrogenase-methenyltetrahydrofolate cyclohydrolase is essential for embryonic development
.
Mol Cell Biol
2002
;
22
:
4158
66
.
6.
Vazquez
A
,
Tedeschi
PM
,
Bertino
JR
. 
Overexpression of the mitochondrial folate and glycine-serine pathway: a new determinant of methotrexate selectivity in tumors
.
Cancer Res
2013
;
73
:
478
82
.
7.
Siegel
RL
,
Miller
KD
,
Jemal
A
. 
Cancer statistics, 2020
.
CA Cancer J Clin
2020
;
70
:
7
30
.
8.
Sheng
X
,
Arnoldussen
YJ
,
Storm
M
,
Tesikova
M
,
Nenseth
HZ
,
Zhao
S
, et al
Divergent androgen regulation of unfolded protein response pathways drives prostate cancer
.
EMBO Mol Med
2015
;
7
:
788
801
.
9.
Pallmann
N
,
Livgard
M
,
Tesikova
M
,
Zeynep Nenseth
H
,
Akkus
E
,
Sikkeland
J
, et al
Regulation of the unfolded protein response through ATF4 and FAM129A in prostate cancer
.
Oncogene
2019
;
38
:
6301
18
.
10.
Sheng
X
,
Nenseth
HZ
,
Qu
S
,
Kuzu
OF
,
Frahnow
T
,
Simon
L
, et al
IRE1alpha-XBP1s pathway promotes prostate cancer by activating c-MYC signaling
.
Nat Commun
2019
;
10
:
323
.
11.
Ducker
GS
,
Rabinowitz
JD
. 
One-carbon metabolism in health and disease
.
Cell Metab
2017
;
25
:
27
42
.
12.
Luhr
M
,
Torgersen
ML
,
Szalai
P
,
Hashim
A
,
Brech
A
,
Staerk
J
, et al
The kinase PERK and the transcription factor ATF4 play distinct and essential roles in autophagy resulting from tunicamycin-induced ER stress
.
J Biol Chem
2019
;
294
:
8197
217
.
13.
Tibbetts
AS
,
Appling
DR
. 
Compartmentalization of Mammalian folate-mediated one-carbon metabolism
.
Annu Rev Nutr
2010
;
30
:
57
81
.
14.
Nguyen
HM
,
Vessella
RL
,
Morrissey
C
,
Brown
LG
,
Coleman
IM
,
Higano
CS
, et al
LuCaP Prostate cancer patient-derived xenografts reflect the molecular heterogeneity of advanced disease and serve as models for evaluating cancer therapeutics
.
Prostate
2017
;
77
:
654
71
.
15.
Jin
Y
,
Wang
L
,
Qu
S
,
Sheng
X
,
Kristian
A
,
Maelandsmo
GM
, et al
STAMP2 increases oxidative stress and is critical for prostate cancer
.
EMBO Mol Med
2015
;
7
:
315
31
.
16.
Jin
Y
,
Qu
S
,
Tesikova
M
,
Wang
L
,
Kristian
A
,
Maelandsmo
GM
, et al
Molecular circuit involving KLK4 integrates androgen and mTOR signaling in prostate cancer
.
PNAS
2013
;
110
:
E2572
81
.
17.
Taylor
BS
,
Schultz
N
,
Hieronymus
H
,
Gopalan
A
,
Xiao
Y
,
Carver
BS
, et al
Integrative genomic profiling of human prostate cancer
.
Cancer Cell
2010
;
18
:
11
22
.
18.
Vanaja
DK
,
Cheville
JC
,
Iturria
SJ
,
Young
CY
. 
Transcriptional silencing of zinc finger protein 185 identified by expression profiling is associated with prostate cancer progression
.
Cancer Res
2003
;
63
:
3877
82
.
19.
Lapointe
J
,
Li
C
,
Higgins
JP
,
van de Rijn
M
,
Bair
E
,
Montgomery
K
, et al
Gene expression profiling identifies clinically relevant subtypes of prostate cancer
.
PNAS
2004
;
101
:
811
6
.
20.
LaTulippe
E
,
Satagopan
J
,
Smith
A
,
Scher
H
,
Scardino
P
,
Reuter
V
, et al
Comprehensive gene expression analysis of prostate cancer reveals distinct transcriptional programs associated with metastatic disease
.
Cancer Res
2002
;
62
:
4499
506
.
21.
Singh
D
,
Febbo
PG
,
Ross
K
,
Jackson
DG
,
Manola
J
,
Ladd
C
, et al
Gene expression correlates of clinical prostate cancer behavior
.
Cancer Cell
2002
;
1
:
203
9
.
22.
Park
Y
,
Reyna-Neyra
A
,
Philippe
L
,
Thoreen
CC
. 
mTORC1 balances cellular amino acid supply with demand for protein synthesis through post-transcriptional control of ATF4
.
Cell Rep
2017
;
19
:
1083
90
.
23.
Harding
HP
,
Zhang
Y
,
Zeng
H
,
Novoa
I
,
Lu
PD
,
Calfon
M
, et al
An integrated stress response regulates amino acid metabolism and resistance to oxidative stress
.
Mol Cell
2003
;
11
:
619
33
.
24.
Xing
C
,
Ci
X
,
Sun
X
,
Fu
X
,
Zhang
Z
,
Dong
EN
, et al
Klf5 deletion promotes Pten deletion-initiated luminal-type mouse prostate tumors through multiple oncogenic signaling pathways
.
Neoplasia
2014
;
16
:
883
99
.
25.
Cowling
VH
,
D'Cruz
CM
,
Chodosh
LA
,
Cole
MD
. 
c-Myc transforms human mammary epithelial cells through repression of the Wnt inhibitors DKK1 and SFRP1
.
Mol Cell Biol
2007
;
27
:
5135
46
.
26.
Zheng
L
,
Sun
D
,
Fan
W
,
Zhang
Z
,
Li
Q
,
Jiang
T
. 
Diagnostic value of SFRP1 as a favorable predictive and prognostic biomarker in patients with prostate cancer
.
PLoS One
2015
;
10
:
e0118276
.
27.
Anderson
PD
,
McKissic
SA
,
Logan
M
,
Roh
M
,
Franco
OE
,
Wang
J
, et al
Nkx3.1 and Myc crossregulate shared target genes in mouse and human prostate tumorigenesis
.
J Clin Invest
2012
;
122
:
1907
19
.
28.
Nusinow
DP
,
Szpyt
J
,
Ghandi
M
,
Rose
CM
,
McDonald
ER
, 3rd,
Kalocsay
M
, et al
Quantitative proteomics of the cancer cell line encyclopedia
.
Cell
2020
;
180
:
387
402
.
29.
Tameire
F
,
Verginadis
II
,
Leli
NM
,
Polte
C
,
Conn
CS
,
Ojha
R
, et al
ATF4 couples MYC-dependent translational activity to bioenergetic demands during tumour progression
.
Nat Cell Biol
2019
;
21
:
889
99
.
30.
Barfeld
SJ
,
Urbanucci
A
,
Itkonen
HM
,
Fazli
L
,
Hicks
JL
,
Thiede
B
, et al
c-Myc antagonises the transcriptional activity of the androgen receptor in prostate cancer affecting key gene networks
.
EBioMedicine
2017
;
18
:
83
93
.
31.
Faubert
B
,
Solmonson
A
,
DeBerardinis
RJ
. 
Metabolic reprogramming and cancer progression
.
Science
2020
;
368
:
eaaw5473
.
32.
Yoshida
GJ
. 
Metabolic reprogramming: the emerging concept and associated therapeutic strategies
.
J Exp Clin Cancer Res
2015
;
34
:
111
.
33.
Sun
H
,
Zhou
Y
,
Skaro
MF
,
Wu
Y
,
Qu
Z
,
Mao
F
, et al
Metabolic reprogramming in cancer is induced to increase proton production
.
Cancer Res
2020
;
80
:
1143
55
.
34.
Wang
Y
,
Ning
Y
,
Alam
GN
,
Jankowski
BM
,
Dong
Z
,
Nor
JE
, et al
Amino acid deprivation promotes tumor angiogenesis through the GCN2/ATF4 pathway
.
Neoplasia
2013
;
15
:
989
97
.
35.
DeNicola
GM
,
Chen
PH
,
Mullarky
E
,
Sudderth
JA
,
Hu
Z
,
Wu
D
, et al
NRF2 regulates serine biosynthesis in non-small cell lung cancer
.
Nat Genet
2015
;
47
:
1475
81
.
36.
Possemato
R
,
Marks
KM
,
Shaul
YD
,
Pacold
ME
,
Kim
D
,
Birsoy
K
, et al
Functional genomics reveal that the serine synthesis pathway is essential in breast cancer
.
Nature
2011
;
476
:
346
50
.
37.
Sun
L
,
Song
L
,
Wan
Q
,
Wu
G
,
Li
X
,
Wang
Y
, et al
cMyc-mediated activation of serine biosynthesis pathway is critical for cancer progression under nutrient deprivation conditions
.
Cell Res
2015
;
25
:
429
44
.
38.
Vie
N
,
Copois
V
,
Bascoul-Mollevi
C
,
Denis
V
,
Bec
N
,
Robert
B
, et al
Overexpression of phosphoserine aminotransferase PSAT1 stimulates cell growth and increases chemoresistance of colon cancer cells
.
Mol Cancer
2008
;
7
:
14
.
39.
Lane
AN
,
Fan
TW
. 
Regulation of mammalian nucleotide metabolism and biosynthesis
.
Nucleic Acids Res
2015
;
43
:
2466
85
.
40.
Ducker
GS
,
Chen
L
,
Morscher
RJ
,
Ghergurovich
JM
,
Esposito
M
,
Teng
X
, et al
Reversal of cytosolic one-carbon flux compensates for loss of the mitochondrial folate pathway
.
Cell Metab
2016
;
23
:
1140
53
.
41.
Shin
M
,
Momb
J
,
Appling
DR
. 
Human mitochondrial MTHFD2 is a dual redox cofactor-specific methylenetetrahydrofolate dehydrogenase/methenyltetrahydrofolate cyclohydrolase
.
Cancer Metab
2017
;
5
:
11
.
42.
Ye
J
,
Fan
J
,
Venneti
S
,
Wan
YW
,
Pawel
BR
,
Zhang
J
, et al
Serine catabolism regulates mitochondrial redox control during hypoxia
.
Cancer Discov
2014
;
4
:
1406
17
.
43.
Martinez-Reyes
I
,
Chandel
NS
. 
Mitochondrial one-carbon metabolism maintains redox balance during hypoxia
.
Cancer Discov
2014
;
4
:
1371
3
.
44.
Ju
HQ
,
Lu
YX
,
Chen
DL
,
Zuo
ZX
,
Liu
ZX
,
Wu
QN
, et al
Modulation of redox homeostasis by inhibition of mthfd2 in colorectal cancer: mechanisms and therapeutic implications
.
J Natl Cancer Inst
2019
;
111
:
584
96
.
45.
Ben-Sahra
I
,
Hoxhaj
G
,
Ricoult
SJH
,
Asara
JM
,
Manning
BD
. 
mTORC1 induces purine synthesis through control of the mitochondrial tetrahydrofolate cycle
.
Science
2016
;
351
:
728
33
.
46.
Torrence
ME
,
MacArthur
MR
,
Hosios
AM
,
Valvezan
AJ
,
Asara
JM
,
Mitchell
JR
, et al
The mTORC1-mediated activation of ATF4 promotes protein and glutathione synthesis downstream of growth signals
.
Elife
2021
;
10
:
e63326
.
47.
Koh
CM
,
Bieberich
CJ
,
Dang
CV
,
Nelson
WG
,
Yegnasubramanian
S
,
De Marzo
AM
. 
MYC and prostate cancer
.
Genes Cancer
2010
;
1
:
617
28
.
48.
Zhao
N
,
Cao
J
,
Xu
L
,
Tang
Q
,
Dobrolecki
LE
,
Lv
X
, et al
Pharmacological targeting of MYC-regulated IRE1/XBP1 pathway suppresses MYC-driven breast cancer
.
J Clin Invest
2018
;
128
:
1283
99
.
49.
Sheng
X
,
Nenseth
HZ
,
Qu
S
,
Kuzu
O
,
Frahnow
T
,
Simon
L
, et al
IRE1α-XBP1s pathway promotes prostate cancer by activating c-MYC signaling
.
Nat Commun
2019
;
10
:
323
34
.
50.
Jin
Y
,
Saatcioglu
F
. 
Targeting the unfolded protein response in hormone-regulated cancers
.
Trends Cancer
2020
;
6
:
160
71
.