A Review of Small-Molecule Inhibitors of One-Carbon Enzymes: SHMT2 and MTHFD2 in the Spotlight

Abstract

Metabolic reprogramming is a key hallmark of cancer and shifts cellular metabolism to meet the demands of biomass production necessary for abnormal cell reproduction. One-carbon metabolism (1CM) contributes to many biosynthetic pathways that fuel growth and is comprised of a complex network of enzymes. Methotrexate and 5-fluorouracil were pioneering drugs in this field and are still widely used today as anticancer agents as well as for other diseases such as arthritis. Besides dihydrofolate reductase and thymidylate synthase, two other enzymes of the folate cycle arm of 1CM have not been targeted clinically: serine hydroxymethyltransferase (SHMT) and methylenetetrahydrofolate dehydrogenase (MTHFD). An increasing body of literature suggests that the mitochondrial isoforms of these enzymes (SHMT2 and MTHFD2) are clinically relevant in the context of cancer. In this review, we focused on the 1CM pathway as a target for cancer therapy and, in particular, SHMT2 and MTHFD2. The function, regulation, and clinical relevance of SHMT2 and MTHFD2 are all discussed. We expand on previous clinical studies and evaluate the prognostic significance of these critical enzymes by performing a pan-cancer analysis of patient data from the The Cancer Genome Atlas and a transcriptional coexpression network enrichment analysis. We also provide an overview of preclinical and clinical inhibitors targeting the folate pathway, the methionine cycle, and folate-dependent purine biosynthesis enzymes.

1. Introduction

One-carbon metabolism (1CM) is commonly dysregulated in cancer due to its contribution to core cellular building blocks (e.g., purine and pyrimidine nucleotides), epigenetics, post-translational modifications (PTMs), and redox homeostasis, and it is comprised of a complex network of enzymes.^1,2 Targeting of 1CM was one of the original chemotherapeutic strategies with the discovery of aminopterin by Sidney Farber in 1948 and has led to some of the most successful approaches to halt cell growth.³⁻⁵ Methotrexate (MTX) and 5-fluorouracil (5FU) were pioneering drugs in this field and are still widely used today. Dihydrofolate reductase (DHFR) and thymidylate synthase (TYMS) are the targets of these drugs, respectively, and have been successfully targeted for other proliferative diseases such as arthritis.⁶ Two other enzymes complete the folate cycle arm of 1CM but have not been targeted clinically at this point: serine hydroxymethyltransferase (SHMT) and methylenetetrahydrofolate dehydrogenase (MTHFD). Literature continues to suggest that the mitochondrial isoforms of these enzymes are clinically relevant in the context of cancer.⁷⁻⁹

2. The 1CM Pathway

2.1. Preliminary Metabolism before Folate Cycle Entry

1CM begins with the folate cycle and provides the 1C unit to the methionine cycle. The one-carbon metabolism pathway is illustrated in Figure 1. The 1C unit is largely supplied by serine, either from the extracellular environment or de novo synthesis, but can be donated by other amino acids as well.⁴ The extracellular serine is transported by solute carrier family 1 member 4/5 (SLC1A4/SLC1A5).¹⁰ De novo serine biosynthesis is carried out by a series of three enzymes from the glycolytic intermediate 3-phosphoglycerate (3PG): phosphoglycerate dehydrogenase (PHGDH), phosphoserine aminotransferase 1 (PSAT1), and phosphoserine phosphatase (PSPH) which extends off of glycolysis.¹¹ Folate is not able to diffuse through the plasma membrane due to its high polarity (cLog P = −2.37); therefore, it must be transported. Folate and its derivatives enter cells through three mechanisms: internalization of the folate receptors (FRα/FOLR1 and FRβ/FOLR2), passive transport by the reduced folate carrier (RFC, encoded by SLC19A1), and active transport by the proton-coupled folate transporter (PCFT, encoded by SLC46A1).¹²⁻¹⁴ In the context of cancer, the α isoform of the folate receptor has been shown to be more relevant.¹² Polyglutamylation of folate and its derivatives are regulated by the extracellular folate hydrolase 1 (FOLH1, also known as PSMA and GCPII), intracellular folylpolyglutamate synthase (FPGS), and lysosomal gamma glutamyl hydrolase (GGH).^15,16 Only monoglutamate forms of folate are able to be transported. Notably, polyglutamylated folate derivatives are usually better substrates for folate-dependent enzymes, and therefore regulation by FOLH1, FPGS, and GGH play a key role in sensitivity to chemotherapies.^15,17 While nearly all dietary folate is in a reduced, bioactive form,¹⁸ folic acid (vitamin B9) must be reduced to dihydrofolate (DHF) and further to tetrahydrofolate (THF) by DHFR to yield the physiological form.

Figure 1.

Key enzymes involved in one-carbon metabolism. Cytosolic folate metabolism enzymes: FPGS; DHFR; TYMS; SHMT1; MTHFR; MTHFD1; MTHFS, methenyltetrahydrofolate synthetase, ALDH1L1. Serine biosynthesis enzymes: PHGDH, phosphoglycerate dehydrogenase; PSAT1, phosphoserine aminotransferase 1; PSPH, phosphoserine phosphatase. Mitochondrial folate metabolism: SHMT2, serine hydroxymethyltransferase 2; MTHFD2/2L, methylenetetrahydrofolate dehydrogenase 2/2-like; MTFMT, methyltetrahydrofolate methyltransferase; MTHFD1L, methylenetetrahydrofolate dehydrogenase 1-like; ALDH1L2, aldehyde dehydrogenase 1 family member L2; GCS, glycine cleavage system. Plasma membrane transporters: RFC, reduced folate carrier; PCFT, proton-coupled folate transporter; FOLR1/2, folate receptor α or β; FOLH1, folate hydrolase; SLC1A4/SLC1A5, solute carrier family 1 member 4/5. Mitochondrial transporters: SFXN1, sideroflexin 1; MFT, mitochondrial folate transporter.

2.2. Cytosolic Folate Metabolism

After DHFR reduces folate to THF, SHMT1 transfers a 1C unit from serine to produce 5,10-methylene-THF (5,10-CH₂-THF). This can be recycled back to DHF by TYMS, which transfers the 1C unit to dUMP to form dTMP.² This is the first contribution to nucleotide biosynthesis (pyrimidines). Alternatively, 5,10-CH₂-THF can go through a series of oxidation reactions catalyzed by MTHFD1. MTHFD1 has three functions: dehydrogenase, cyclohydrolase, and synthetase. The first intermediate, 5,10-methenyl-THF (5,10-CH⁺-THF), is short-lived due to its instability and is subsequently reduced to 10-formyl-THF (10-CHO-THF).¹⁹ 10-CHO-THF can go on to participate in purine biosynthesis (GAR transformylase, GART, and AICAR formyltransferase, AICARFT)²⁰ or be converted back to THF by MTHFD1 or aldehyde dehydrogenase 1 family member L1 (ALDH1L1) releasing formate or CO₂, respectively.²¹ Methenyltetrahydrofolate synthetase can contribute to this pathway by converting 5-CHO-THF into 5,10-CH⁺-THF.²² 5,10-CH₂-THF can also be further reduced to 5-methyl-THF (5-CH₃-THF) by methylenetetrahydrofolate reductase (MTHFR), acting as the methyl donor to the methionine cycle.²³

2.3. Mitochondrial Folate Metabolism

In order for mitochondrial folate metabolism to function, serine and THF must first be transported into the organelle by sideroflexin 1 (SFXN1) and the mitochondrial folate transporter (MFT), respectively.^24,25 SHMT2 is the mitochondrial isoform of SHMT1 and performs the same enzymatic reaction to donate a 1C unit from serine to THF.² In addition to serine, glycine is another amino acid that can donate a 1C unit, and this is facilitated by the glycine cleavage system (GCS).²⁶ Both SHMT2 and the GCS generate 5,10-CH₂-THF, which is a substrate for MTHFD2/2L. MTHFD2 and MTHFD2L catalyze the oxidation of 5,10-CH₂-THF to 10-CHO-THF.^27,28 MTHFD1L completes what the trifunctional MTHFD1 is able to catalyze in the cytosol: regeneration of THF and production of formate.²⁹ 10-CHO-THF is also a substrate for ALDH1L2 and MTFMT. ALDH1L2 has the same function as the cytosolic ALDH1L1; production of THF and carbon dioxide.²¹ MTFMT, or mitochondrial methionyl-tRNA formyltransferase, transfers the formyl group from 10-CHO-THF to methionine-tRNAs to produce formyl-methionine-tRNA which is required for the initiation of translation of mitochondrial proteins.³⁰

2.4. Select Inhibitors of 1CM

Inhibitors have been discovered for most enzymes and transporters involved in 1CM (Figure 2), also recently succinctly summarized by Robinson et al.³¹ Those that remain to be targeted are PSAT1, MFT (SLC25A32), PCFT (SLC46A1), SFXN1, ALDH1L1/2, MTFMT, MTRR, and DMGDH. A list of inhibitors of 1CM proteins and their structures, activities, and status can be found in Tables 1–6. Inhibitors of subsections of 1CM are discussed briefly below. Biochemical activity data is only included for nonpolyglutamyl forms of folate derivatives, but it is important to note that polyglutamylated forms typically bind with much higher affinities.¹⁷ It should be noted that only experimentally validated compounds with activity levels ≤50 μM are included, and structures likely to be pan-assay inhibitors (PAINS) are excluded.

Figure 2.

Summary of select preclinical and clinically used compounds targeting 1CM. Coloring: black indicates FDA-approved, green indicates clinical trials, red indicates preclinical. BSP, bromosulfophthalein; GPC, glycerylphosphorylcholine.

Table 1. Serine Biosynthesis Inhibitors^32,34−47.

Table 6. Inhibitors of Folate-Dependent Enzymes in Purine Biosynthesis^{66,107,110−114,116−120}.

2.4.1. Serine Biosynthesis Inhibitors

Two out of the three enzymes that make up the serine biosynthesis pathway have inhibitors discovered for them (Figure 2, Table 1). A good portion of the reported serine biosynthesis inhibitors have antibacterial activity and have been recently reviewed.³² PHGDH is an emerging anticancer target, and there has been extensive effort to develop compounds that inhibit this enzyme.³³ BI-4924 is the most potent among them with an IC₅₀ of 2 nM.³⁴

2.4.2. Folate Glutamylation-Modifying Enzyme Inhibitors

Glutamylation states of folates or derivatives are important for activity against their targets and also determine cell permeability/uptake by transporters.¹⁵ Uptake by the folate transporters is only possible with the monoglutamate forms. Inhibitors have been discovered for both FOLH1 and FPGS although none are FDA-approved (Figure 2, Table 2). FOLH1 inhibitors have suffered from poor oral bioavailability (due to their acidic nature in mimicking the glutamate substrate) or toxicity which prevented progress in clinical trials.¹⁶ However, there is still interest in pursuing pro-drug strategies for the inhibitors in which there was not dose-limiting toxicity because of promising preclinical data.¹⁶ Folate analogs are reported to inhibit FPGS, albeit only modestly in the midmicromolar range.⁴⁸

Table 2. Folate Glutamylation-Modifying Enzyme Inhibitors^16,48,49.

^aConsult ref (16) for details on the clinical use of FOLH1 inhibitors.

2.4.3. Folate and Serine Transporter Inhibitors

Folate transporters play an important role in the efficacy of antifolate chemotherapies.^12,14,50 The small molecule inhibitors of these cell surface transporters require high concentrations in order to be effective (>100 μM). In pursuit of delivering oncolytic virotherapy to ovarian cancer cells, Hulin-Curtis and colleagues identified a peptide that competes with folic acid for binding to FOLR1 (Table 3).⁵¹ There are many antibody–drug conjugates and folic acid conjugates built on the same premise of delivering chemotherapy selectively to the cancer cells.¹² The unconjugated monoclonal antibody farletuzumab specifically targets FOLR1, but interestingly, inhibition of folic acid uptake is not part of its mechanism of action.¹² Despite encouraging preclinical data, farletuzumab has not successfully completed a Phase III trial.¹² As these transporters are important for the efficacy of antifolate drugs, development of inhibitors for these proteins would need to be carefully considered in the context of cancer.

Table 3. Folate and Serine Transporter Inhibitors^12,51−55.

^aAntifolates are excluded as they are also substrates.

^bCompounds with K_i or IC₅₀ > 50 μM are not included

2.4.4. Cytosolic and Mitochondrial Folate Metabolism Inhibitors

DHFR and TYMS were the first 1CM proteins to be clinically validated as targets for cancer therapy and remain the most successful in this context to date (Figure 2, Table 4). All FDA-approved DHFR and TYMS inhibitors are classical or nonclassical folate derivatives, with the obvious exception of 5FU. There are also several compounds targeting these two enzymes in the clinical trial pipeline: raltitrexed, piritrexim, ZD-9331, GS7904L, and ONX-0801 (Figure 2, green text; Table 4). Detailed reviews are available for these classical inhibitors.^56,57 Considerable interest exists for ONX-0801 because it specifically targets FOLR1-overexpressing tumors.⁵⁸ Pemetrexed and raltitrexed, though originally designated as TYMS inhibitors, also potently inhibit DHFR (Table 4) and to a lesser extent purine biosynthesis enzymes (Table 6).

Table 4. Cytosolic and Mitochondrial Folate Metabolism Inhibitors^{12,56,58−90}.

^aOnly cancer-related studies outside of approved indications and with drug as a single agent are included (recruiting or completed).

The only known inhibitor for MTHFR is S-adenosylmethionine (SAM), which is a product feedback inhibitor.^23,59 There are no reported inhibitors for human MTHFS, but 5-formyltetrahydrohomofolate (5-formylTHHF) inhibits the murine homologue with a K_i of 0.7 μM.⁶⁰ The GCS is a multienzyme complex composed of four proteins which all work to convert glycine and THF to 5,10-CH₂-THF, carbon dioxide, and ammonia. To our knowledge, cysteamine is the only described inhibitor for the GCS with a K_i of 5 μM.⁶¹ Cysteamine is FDA-approved for nephropathic cystinosis and is undergoing clinical trials for multiple indications (Table 4).

2.4.5. Inhibitors of the Methionine Cycle Arm of 1CM

The methionine cycle has great implications because through it the 1C unit is transferred to S-adenosylhomocysteine (SAH) to form SAM, a cofactor used by many methyltransferases.⁹¹ MTRR (methionine synthase reductase) and DMGDH (dimethylglycine dehydrogenase, mitochondrial) are the two enzymes yet to be targeted of the methionine cycle. Most of the established inhibitors are folate, adenosine, or amino acid analogs; however, there are some exceptions (Table 5). Methionine synthase (MS) inhibitors contain interesting electrophilic folate analogs such as compounds 6 and 6c which take advantage of MS’s mechanism of action and possess anticancer activity.^92,93 Inhibitors of glycine N-methyltransferase (GNMT) include one of its products, SAH, and 5-CH₃-THF, which acts as a regulator of pathway flux,⁹⁴ but no inhibitors outside of these downstream molecules have been discovered. A clinical trial with a compound targeting the 2A isoform of methionine adenosyltransferase (MAT2A), AG-270, is currently underway (NCT03435250). Biochemical or preclinical data from Agios Pharmaceuticals is not available at this time. The first nonsubstrate or product based allosteric MAT2A inhibitor with nanomolar activity came from Pfizer in 2017.⁹⁵ The natural product AKBA was recently identified as a potent and selective allosteric inhibitor of MAT2A with a K_d of 129 nM.⁹⁶ There are many potent inhibitors discovered for SAHH (S-adenosyl-l-homocysteine hydrolase) (Table 5). Many are, unsurprisingly, adenine analogs.⁹⁷⁻⁹⁹ However, several compounds with new scaffolds for SAHH were recently identified using a novel high-throughput screening technique.¹⁰⁰ The physiological roles of betaine-homocysteine methyltransferases 1 or 2 (BHMT1/2) are still incompletely understood, especially in the context of cancer. Potent inhibitors of both isoforms have been discovered and could be important tools in elucidating BHMT/2’s function.^101,102

Table 5. Inhibitors of the Methionine Cycle Arm of 1CM⁹²⁻¹⁰⁶.

2.4.6. Inhibitors of Folate-Dependent Enzymes in Purine Biosynthesis

Purine biosynthesis is carried out by the purinosome, a multienzyme complex, where two of the enzymes (GART and AICARFT) are folate-dependent, both requiring 10-CHO-THF.²⁰ Nearly all discovered inhibitors for these enzymes mimic the folate substrate except for the peptidomimetic identified (compound 14) for AICARFT (Table 6).¹⁰⁷ The FDA-approved inhibitor pemetrexed does possess inhibitory activity against GART and AICARFT but only in the low micromolar range.⁶⁶ Though it has reduced potency compared to other 1CM enzymes, pemetrexed-induced inhibition of AICARFT has important implications as it causes accumulation of the purine intermediate aminoimidazolecarboxamide ribonucleotide (ZMP) thereby affecting AMPK and mTOR signaling.^108,109 Lometrexol is much more potent with a K_i of 60 nM for GART,¹¹⁰ but was discontinued from clinical trials. Another potent inhibitor of GART is AG2034 with a K_i of 28 nM but unfortunately did not advance past Phase I clinical trials due to toxicity.^111,112 Importantly, several series of compounds discovered by a collaboration between the Ganjee and Matherly laboratories potently inhibit GART and are selectively transported by folate receptors and the PCFT over the RFC.^113−115 The discovery of the AICARFT inhibitor LSN3213128 by Eli Lilly and Company is exciting as it is orally bioavailable, efficacious in murine xenografts, and selective against other 1CM enzymes.¹¹⁶ Given the success of other antipurinic antimetabolites (e.g., gemcitabine), the field awaits data on the further development of this potential anticancer therapeutic.

3. SHMT2

3.1. Function and Regulation

SHMT2 simultaneously converts serine and THF to glycine and 5,10-CH₂-THF (Figure 1). SHMT2 seems to promote chemosensitivity or resistance in a context-specific manner.^121,122 Outside of nutrient contributions, SHMT2 knockout caused mitochondrial translation to stall at specific methylated tRNAs, and this was dependent on its enzymatic activity.¹²³ Protein translation initiation requires formylmethionyl tRNAs, and when SHMT2 is knocked out, 10-CHO-THF, a downstream product, is no longer available for conjugation by MTFMT.³⁰ SHMT2 also plays a role in redox balance. Expression of SHMT2 was shown to be associated with changes in the level of expression of mitochondrial respiration complex proteins.¹²⁴ Hypoxia induces SHMT2 expression and was found to be necessary for redox homeostasis and cell survival under hypoxic conditions.^125,126 Additionally, expression of SHMT2 promotes survival in the ischemic tumor microenvironment in glioma.¹²⁷ Recent structural studies have identified a novel function of SHMT2 that contributes to immune regulation through the BRISC complex.^128−130 This function was in turn regulated by a PTM modulated by HDAC11.¹³¹ Other PTMs such as acetylation and succinylation determine enzymatic activity and are modulated by SIRT3 and SIRT5, respectively.^132,133 Besides canonical regulation by known metabolic master regulators (Figure 3, black text),¹ novel epigenetic mechanisms, transcriptional programs, and noncoding RNAs have been reported to modulate SHMT2 expression (Figure 3, red text). Knockdown of the histone methyltransferase G9A resulted in a decrease in expression of SHMT2 as well as reduced H3K9Me1 levels for serine synthesis pathway genes.¹³⁴ Novel transcription factors and signaling pathways that regulate SHMT2 include TGF-β, STAT3, and EWS-FLI1.^135−137 Interestingly, SHMT1 was shown to bind to the 5′-untranslated region of the SHMT2 transcript thereby preventing its expression.¹³⁸ SHMT2 expression has also been reported to be under the control of several micro-RNAs (miRs) and long intergenic noncoding RNAs (lincRNAs).^139−142 A summary of regulation of SHMT2 is shown in Figure 3. Despite encouraging evidence that SHMT2 is a viable anticancer target, only a handful of inhibitors have been published to date (discussed below).^143−146 However, none are specific to SHMT2.

Figure 3.

Summary of the regulation of SHMT2 expression in cancer. Canonical mechanisms are in black text, and noncanonical mechanisms are in red.

3.2. SHMT Inhibitors

The pyrazolopyran scaffold is the predominant series of compounds that make up the SHMT1/2 inhibitor space (Table 7 summarizes select inhibitors for the human isoforms). This class of compounds was originally disclosed as plant SHMT inhibitors in a patent application by BASF (WO 2013182472 A1). This scaffold was then evaluated for activity against Plasmodium SHMT and showed nanomolar enantioselective activity against the enzyme and the parasites.¹⁴⁷ That same year, Marani and colleagues published the first data on the human isoforms. In addition to having selective activity against SHMT1, compound 2.12 (Table 7) also displayed anticancer activity in the mid-micromolar range.¹⁴⁸ Further optimization of this scaffold led to potent inhibitors of both SHMT1 and SHMT2.¹⁴³ However, as was also shown in the antimalarial studies,¹⁴⁷ this scaffold suffered from poor pharmacokinetic properties that prevented in vivo efficacy studies.¹⁴³ A breakthrough came this past year with the disclosure of (+)SHIN2 which showed in vivo target engagement of SHMT2 in addition to anticancer activity in models of T-ALL.¹⁴⁵ Han and colleagues recently disclosed novel pyrazolopyran compounds with midmicromolar activity against SHMT2 (Table 7).¹⁴⁶

Table 7. Inhibitors of Human SHMTs^{143−146,148,149,155,156}.

^aSHMT2 IC₅₀ values were calculated using GraphPad Prism 9 with the Source Data file from ref (156).

The first in vivo active SHMT inhibitor, AGF347, is a folate mimetic.¹⁴⁴ Despite being much less potent than the pyrazolopyran compounds against recombinant protein, AGF347 showed in vivo activity in pancreatic adenocarcinoma models;¹⁴⁴ notably, the least-sensitive cell line in vitro was used for the xenograft model. Importantly, AGF347 was shown to be metabolized to polyglutamate forms.¹⁴⁹ Polyglutamated endogenous folates bind SHMT with increased affinity compared to their monoglutamated counterparts.^150,151 Therefore, it is possible that the polyglutamylation of AGF347 increases the affinity to have similar activity as the pyrazolopyrans, but this remains to be determined experimentally. FDA-approved antifolates do inhibit SHMTs, but higher concentrations are required (≥100 μM).^152−154

Using a novel fluorescent assay measuring SHMT activity directly, Nonaka et al. conducted a high-throughput screen of >200k compounds.¹⁵⁶ They identified two hits with sub-micromolar activity (Table 7). “Hit 1” is ∼20-fold selective for SHMT1, while “hit 2” is the first-reported SHMT2 selective inhibitor (∼5-fold). Prior to Nonaka and co-worker’s study, generally, only marginal selectivity between the two SHMT isoforms has been achieved, with the inhibitors being selective for SHMT1 by 2- to 3-fold.^143,155 The cysteine-reactive inhibitor 3-bromopyruvate is selective for SHMT1 due to the absence of a cysteine residue in SHMT2’s active site.¹⁵⁷ Crystal structures are available for both isoforms (summarized in Table 8), but there are currently no cocrystal structures of SHMT1 with any of the published inhibitors. If selectivity is desired, then it will be critical to experimentally determine differences in the binding pockets. However, one could pose the following question: Is SHMT2 selectivity ideal? In vivo studies of the dual SHMT1/2 inhibitors show that coinhibition is efficacious with minimal systemic toxicities,^144,145 suggesting that selectivity may not be required. Additionally, when the mitochondrial folate pathway is inhibited, the cytosolic pathway reverses flux to compensate; therefore, inhibition of both isoforms is essential for antitumor activity.¹⁵⁸

Table 8. Crystal Structures of the Human SHMTs.

protein	variant	ligands	macromolecule	method	resolution	PDB ID	refs
SHMT1	WT	PLP	NAa	X-ray	2.65 Å	1BJ4	(159)
	H135N, R137A, E168N	PLP	NA	X-ray	3.60 Å	6FL5	(160)
SHMT2	WT	NA	NA	X-ray	2.04 Å	6DK3	TBPb
	A264T	NA	BRISC-complex	cryo-EM	3.90 Å	6H3C	(128)
	WT	NA	BRISC-complex	cryo-EM	3.80 Å	6R8F	(129)
	WT	PLP, pemetrexed	NA	X-ray	2.28 Å	6QVL	(152)
	WT	PLP, lometrexol	NA	X-ray	2.32 Å	6QVG
	WT	PLP	NA	X-ray	2.60 Å	4PVF	(161)
	WT	PLP, compound 2	NA	X-ray	2.47 Å	5V7I	(143)
	K74R, A264T	NA	NA	X-ray	2.85 Å	5X3V	TBP

^aNA, not applicable.

^bTBP, to be published.

3.3. Association with Cancer Patient Survival and Transcriptional Programs

The mitochondrial isoform of SHMT is selectively overexpressed compared to its cytosolic counterpart and is associated with poor patient prognosis, staging, and metastasis in several cancers including kidney, gastric, liver, breast, lung, colorectal, esophageal and intrahepatic cholangiocarcinoma, oral squamous carcinoma, and glioma.^162−172 Across the NCI-60 panel of cell lines, SHMT2 expression held a stronger correlation with proliferation rate than that of SHMT1.⁸ The Cancer Genome Atlas (TCGA) disease patient samples were evaluated for reduced survival by comparing survival outcomes for patients with high SHMT1 and SHMT2 expression to those with low SHMT1/2 expression. Patients were stratified based on median gene expression, and a log-rank statistic was used to quantify differences in survival with p-values FDR-adjusted across 33 cancer types, and FDR-adjusted p-values < 0.05 were considered significant (Figure 4). Gene expression thresholding and Kaplan–Meier (KM) plots were generated using patient sample RNA-Seq RSEM-normalized gene expression values and survival metadata sourced from the TCGA GDAC Firehose.¹⁷³ Survival test statistics and KM plots were generated using the R statistical programming language.¹⁷⁴ We observed higher expression of SHMT2 in adrenocortical carcinoma (ACC), which was significantly associated with reduced survival; SHMT1 was also marginally associated with reduced survival (p-value not significant after multiple testing). This result is novel based on our search of the literature. Not much is known about the metabolic landscape of ACC, but gemcitabine and capecitabine are part of second-line therapy regimens.¹⁷⁵ Additionally, some mTOR inhibitors are currently undergoing clinical trials.¹⁷⁵ We identified three TCGA diseases for which higher levels of SHMT1 was associated with improved survival and would not recommend targeting SHMT1 in these diseases. Interestingly, in brain lower grade glioma (LGG), patients with a higher expression of SHMT1 were significantly associated with reduced survival, and patients with a higher expression of SHMT2 were moderately associated with improved survival (p-value not significant after FDR correction).

Figure 4.

Pan-cancer SHMT1 and SHMT2 patient survival association survey. (A) SHMT1 and SHMT2 expression are associated with overall survival in several cancers. Patients from each disease were stratified by median expression of SHMT1 and SHMT2. Heat map color indicates directionality and significance of association (red = reduced survival, blue = increased survival) with increased gene expression. Heat map cells with black borders indicate FDR p-value < 0.05. (B) SHMT1 and SHMT2 expression associate with reduced survival in ACC. (C) LGG patients with high expression of SHMT1 had significantly reduced survival, while patients with high levels of SHMT2 marginally associated with improved survival. Log-rank (LR) statistic p-values used to quantify association with survival were FDR adjusted across cancers per gene. ACC SHMT1 and LGG SHMT2 LR p-values not significant after FDR correction. (D) Gene Set Enrichment Analysis (GSEA) performed to identify gene sets enriched for genes correlated with SHMT1 expression in ACC and LGG diseases. 34 common enriched gene sets were identified using Hallmark and KEGG gene sets with FWER < 0.05. (E) GSEA enrichment plots for interferon gamma response gene set. Several immune-related gene sets were significantly enriched for coexpression with SHMT1 expression. GSEA PreRanked analysis performed using Pearson correlation with SHMT1 expression using GSEAv2.2.3, weighted scoring, and 10 000 permutations with MsigDBv7 gene sets. BRCA = breast invasive carcinoma; CESC= cervical and endocervical cancers; CHOL = Cholangiocarcinoma; COAD = colon adenocarcinoma; DLBC = lymphoid neoplasm diffuse large B-cell lymphoma; ESCA = esophageal carcinoma; GBM = glioblastoma multiforme; HNSC = head and neck squamous cell carcinoma; KICH = kidney chromophobe; KIRP = Kidney renal papillary cell carcinoma; LAML = acute myeloid leukemia; LIHC = liver hepatocellular carcinoma; LUAD = lung adenocarcinoma; LUSC = lung squamous cell carcinoma; MESO = mesothelioma; OV = ovarian serous cystadenocarcinoma; PCPG = pheochromocytoma and paraganglioma; PRAD = Prostate adenocarcinoma; READ = Rectum adenocarcinoma; SARC = sarcoma; SKCM = skin cutaneous melanoma; STAD = stomach adenocarcinoma; TGCT = testicular germ cell tumors; THCA = thyroid carcinoma; THYM = thymoma; UCS = uterine carcinosarcoma; UVM = uveal melanoma.

We performed a transcriptional coexpression network enrichment analysis to identify gene sets statistically enriched for genes correlated with SHMT1 in ACC and LGG patient cohorts. Gene Set Enrichment Analysis (GSEA) PreRanked V2.2.3 was used to score Hallmark and KEGG gene sets from the MSigDBv7 database.¹⁷⁶ Interestingly 34 enriched gene sets were commonly enriched in ACC and LGG networks (Figure 4D, GSEA Supporting Information Table) and the statistically significant overlap between diseases indicates common transcriptional programs associated with SHMT1 expression are detected. There were several immune-related gene sets including interferon gamma response, inflammatory response, interferon alpha response, and cytokine signaling identified. Previous studies have identified links between SHMT proteins and interferon signaling, and these findings indicate SHMT expression is associated with immune response transcriptional programs in ACC and LGG patient cohorts.^131,177,178 Tumor necrosis factor alpha (TNF-α) is a key cytokine that regulates immune responses in healthy persons and in diseased conditions. In cancer, TNF-α can elicit myriad effects depending on the context, such as proliferation, apoptosis, necrosis, migration, angiogenesis, or invasion.¹⁷⁹ Little is known about the relationship between cancer cell metabolism reprogramming and TNF-α signaling through NF-κB. To our knowledge, only a couple of studies have been published investigating metabolic changes of cancer cells in response to TNF-α. In noncancerous and malignant breast epithelial cells, TNF-α induces a shift to reliance on aerobic glycolysis, however it is unknown if it was NF-κB dependent.¹⁸⁰ Also, in cooperation with interleukin-17, TNF-α induces glycolysis in colorectal cancer cells.¹⁸¹ Similarly, another study reported metabolic reprogramming of hepatocytes toward the Warburg-effect phenotype in response to viral infection mediated by macrophage-secreted TNF-α.¹⁸² In adipocytes, sustained exposure to TNF-α alters mitochondrial metabolism, notably reducing the NAD/NADH ratio.¹⁸³ Despite the observations toward the Warburg effect, no mechanisms have been elucidated. Given this knowledge gap, it is uncertain as to how folate metabolism is relevant to NF-κB signaling and more specifically SHMTs. Interestingly, a nanoparticle containing NF-κB-targeted siRNA and MTX, a standard drug for the treatment of rheumatoid arthritis, reduced TNF-α secretion and suppressed arthritis progression in a mouse model.¹⁸⁴

4. MTHFD2

4.1. Function and Regulation

MTHFD2 possesses two enzymatic activities, a dehydrogenase and cyclohydrolase (Figure 1). The role of MTHFD2 in cancer was recently reviewed.⁷ Both through genetic and small molecule approaches, inhibition of MTHFD2 reduces proliferation, migration, invasion, tumor growth, stem-cell-like properties, and promotes cell death, differentiation, and chemosensitivity in several cancer types.^7,185−188 In addition to its involvement in nutrient regulation, MTHFD2 plays a role in redox homeostasis through the use of its cofactor NAD⁺, as well as involvement with expression of hypoxia-related proteins such as HIF-2α.^{158,189−192} Other secondary functions of MTHFD2 have also been discovered such as nonenzymatic driven proliferation, potential physical interaction with nuclear-RNA-related proteins, and localization to DNA replication sites.^193,194 The mechanism of how the nonenzymatic function drives proliferation is not well understood. Overexpression of MTHFD2 with an active site residue mutated that is required for catalysis still had a proliferative advantage compared to the vector control.¹⁹³ Similar to SHMT2, regulation of MTHFD2 expression extends beyond canonical metabolism-related transcription factors such as c-MYC, mTOR, ATF4, HIF-1α, AMPK, and NRF2.¹ For example, several microRNAs and transcription factors like the chimeric oncogene EWS-FLI1 have been shown to impact MTHFD2 expression.^7,137 Noncoding RNAs reported to regulate MTHFD2 expression include miR-9 (along with lncRNA TUG1), miR-92a, miR-940, miR-33a-5p, miR-22, and LIN28B.^195−201 MTHFD2 is a predicted target gene of miR-99a-3p, miR-186–5p, and hsa-miR-202.^202−204 Similar to SHMT2, acetylation of an active site residue (Lys88) was reported to be removed by SIRT3 thereby activating MTHFD2’s enzymatic activity.²⁰⁵ Five sites of phosphorylation on MTHFD2 have been identified through phosphoproteomics (S149, T187, T191, T306, and T324).²⁰⁶ However, it is unknown what kinases catalyze these phosphorylations or what purpose they serve. PTMs in general can act as a switch for metabolic enzymes to take on their moonlighting functions.²⁰⁷ For example, phosphorylation of superoxide dismutase 1 (SOD1) causes its translocation to the nucleus where it is able to transcriptionally upregulate antioxidant-related genes.²⁰⁸ As MTHFD2 has been shown to be present at DNA replication sites,¹⁹³ it would be interesting to see if PTMs regulated this function.

4.2. MTHFD Inhibitors

The mounting evidence of the clinical relevance of MTHFD2 has led to intense interest in developing MTHFD2-targeting therapeutics. The development of MTHFD2 inhibitors was recently reviewed, so it will only be discussed here briefly.⁷ There are fewer published inhibitors of MTHFD2 than for SHMT2 (Table 9), but significant progress in selectivity over the cytosolic MTHFD isoform (MTHFD1) was achieved in a very short amount of time. The first described MTHFD2 inhibitor, LY345899, is a folate analog. While it had activity in the high nanomolar range, it was still ∼7-fold selective for MTHFD1.²⁰⁹ Around the same time, the natural product carolacton was originally discovered as an inhibitor of the E. coli isoform, FolD, but it also possessed potent activity against MTHFD1 and MTHFD2.¹⁸⁵ Interestingly, carolacton was selective for MTHFD2 with the dehydrogenase function, but for MTHFD1 with the cyclohydrolase function.¹⁸⁵ The first truly MTHFD2-selective inhibitor was reported in 2019, DS44960156, although potency was lacking.²¹⁰ Later that year, the same group followed up with an optimized inhibitor, DS18561882, that was not only potent in vitro but also efficacious in vivo.¹⁸⁶ MTHFD2L takes the place of MTHFD2’s function in normal adult tissues, so an important next step would be to evaluate the activity of these molecules against this mitochondrial isoform. Currently there are no crystal structures available for the other two human mitochondrial isoforms (Table 10).

Table 9. Inhibitors of Human MTHFDs^{185,186,189,209,210}.

Table 10. X-ray Crystal Structures of Human MTHFDs.

protein	ligands	resolution (Å)	PDB ID	refs
MTHFD1	NADP	1.50	1A4I	(211)
	NADP, LY249543	2.20	1DIA	(212)
	NADP, LY345899	2.70	1DIB
	NADP, LY374571	2.20	1DIG
	NADP, LY249543	2.20	6ECP	(213)
	NADP, LY345899	2.70	6ECQ
	NADP	2.20	6ECR
MTHFD2	NAD, LY345899	1.89	5TC4	(209)
	DS44960156	2.25	6JIB	(210)
	compound 1	2.50	6JID
	DS18561882	2.25	6KG2	(186)
MTHFD1L	NA	NA	NA	NA
MTHFD2L	NA	NA	NA	NA

^aNA, not applicable

4.3. Association with Cancer Patient Survival and Transcriptional Programs

Of the four MTHFD isoforms, MTHFD2 is the only one that is not expressed in adult human tissues and is primarily expressed during embryonic development or in cancerous cells, making it an attractive anticancer target.⁹ Overexpression of MTHFD2 is associated with poor prognosis and clinicopathological parameters in breast, renal, liver, pancreatic, and colorectal cancers and some brain cancers.⁷ A TCGA pan-disease survival association analysis with MTHFD2 expression was performed using the GDAC Firehose-sourced data as described above (Figure 5). Our analysis corroborated previous findings with pancreatic (pancreatic adenocarcinoma, PAAD), kidney renal clear cell carcinoma (KIRC), and kidney renal papillary cell carcinoma (KIRP) malignancies.^214,215 We identified significant reduction in survival with a higher expression of MTHFD2 in bladder urothelial carcinoma (BLCA) and uterine corpus endometrial carcinoma (UCEC) (Figure 5B). These findings have not been published to our knowledge. Significant reduction in survival was seen with a lower expression of MTHFD2 in glioblastoma multiforme (GBM) and low-grade glioma (LGG) (Figure 5C). The role of MTHFD2 in brain cancer seems to be quite complex as other studies have seen similar and contrasting results.⁷

Figure 5.

Pan-cancer MTHFD2 patient survival association survey. (A) MTHFD2 expression correlates with cancer patient survival in several cancers. Patients from each disease were stratified by median expression of MTHFD2. Heat map color indicates directionality and significance of association (red = reduced survival, blue = increased survival) with increased gene expression. Heat map cells with black borders indicate FDR p-value <0.05. (B) High MTHFD2 expression associates with reduced survival in BLCA, KIRC, KIRP, PAAD, and UCEC. Log-rank (LR) statistic p-values used to quantify association with survival were FDR adjusted across cancers. (C) Chow-Ruskey venn diagram showing results from Gene Set Enrichment Analysis (GSEA) performed to identify gene sets enriched for genes correlated with MTHFD2 expression in BLCA, KIRC, KIRP, PAAD, and UCEC diseases. Nine common enriched gene sets were identified using Hallmark and KEGG gene sets with FWER < 0.05. (D) GSEA enrichment plots for Hallmark MYC Targets V1 gene set. GSEA PreRanked analysis performed using Pearson correlation with MTHFD2 expression using GSEAv2.2.3, weighted scoring, 10,000 permutations with MsigDBv7 gene sets.

Our transcriptional coexpression network enrichment analysis identified nine gene sets statistically enriched for genes correlated with MTHFD2 in BLCA, KIRC, KIRP, PAAD, and UCEC patient cohorts. (Figure 5C, GSEA Supporting Information Table). Common gene sets were found to be related to the cell cycle, MYC signaling, and unfolded protein response. MTHFD2’s relation to MYC has previously been established.^189,216,217 Recently, SHMT2 and MTHFD2 expression was found to be increased upon induction of the unfolded protein response (UPR).²¹⁸ Additionally, knockout of a key UPR transcription factor XBP1 caused an upregulation of SHMT2 in dendritic cells.²¹⁹

5. Summary

Metabolic reprogramming shifts a cell’s metabolism from an energetically efficient process to one which is more biomass-production-centric. As a result, this allows cancer cells to rapidly proliferate, resist chemotherapies, invade, metastasize, and survive a nutrient-deprived microenvironment. 1CM has its hand in many biosynthetic pathways that fuel growth, regulate redox status, and contribute to post-translational modifications, among others. In this review, we focused on the 1CM pathway as a target for cancer therapy, and in particular, SHMT2 and MTHFD2.

The function, regulation, and clinical relevance of SHMT2 and MTHFD2 were all discussed. These enzymes are critical for their contributions to the synthesis of one carbon donors in different oxidation states (5,10-CH₂-THF and 10-CHO-THF) that are donated to biosynthetic pathways such as pyrimidine and purine biosynthesis, respectively. Several post-translational modifications have been discovered that regulate SHMT2’s activity. Interestingly, both SHMT2’s and MTHFD2’s enzymatic activity is regulated by SIRT3.^133,205 Additionally, the expression of both enzymes are controlled by an increasing number of miRs and other noncoding RNAs. Through our independent analysis of TCGA, we identified several cancers whose survival was associated with the high expression of SHMT2 (ACC) and MTHFD2 (BLCA, ESCA, HNSC, LUAD, MESO, and UCEC).

We also provided a comprehensive overview of preclinical and clinical inhibitors targeting the 1CM pathway. Most clinically approved drugs for this pathway target DHFR and TYMS, and there are still even more new drugs undergoing clinical trials that target these enzymes (raltitrexed, piritrexim, ONX-0801, ZD-9331, and GS7904L). However, there are other 1CM enzymes with inhibitors undergoing clinical trials (MAT2A and GART). There is considerable interest in pursuing development of SHMT2 and MTHFD2 inhibitors due to their selective overexpression and association with myriad of clinicopathological parameters. Significant progress has been made in the discovery of potent inhibitors of these enzymes.^144,145,186 In summary, we believe that targeting 1CM will be a mainstay therapeutic strategy for the treatment of cancer for years to come.

Glossary

Abbreviations Used

1CM

One-carbon metabolism

5FU

5-Fluorouracil

ACC

Adrenocortical carcinoma

ALDH1L1/ALDH1L2

Aldehyde dehydrogenase 1 family member L1/L2

BLCA

Bladder urothelial carcinoma

BRCA

Breast invasive carcinoma

BSP

Bromosulfophthalein

CESC

Cervical and endocervical cancers

CHOL

Cholangiocarcinoma

COAD

Colon adenocarcinoma

DHFR

Dihydrofolate reductase

DLBC

Lymphoid neoplasm diffuse large B-cell lymphoma

ESCA

Esophageal carcinoma

FDA

Food and Drug Administration

FOLR1/2

Folate receptor α or β

FOLH1

Folate hydrolase

FPGS

Folylpolyglutamate synthase

GBM

Glioblastoma multiforme

GPC

Glycerylphosphorylcholine

GSEA

Gene Set Enrichment Analysis

HNSC

Head and neck squamous cell carcinoma

IC₅₀

Half maximal inhibitory concentration

KICH

Kidney chromophobe

KIRC

Kidney renal clear cell carcinoma

KIRP

Kidney renal papillary cell carcinoma

LAML

Acute myeloid leukemia

LGG

Lower grade glioma

LIHC

Liver hepatocellular carcinoma

LUAD

Lung adenocarcinoma

LUSC

Lung squamous cell carcinoma

MESO

Mesothelioma

MFT

Mitochondrial folate transporter

MTHFD1

Methylenetetrahydrofolate dehydrogenase 1, cytosolic (trifunctional)

MTHFD1L

Methylenetetrahydrofolate dehydrogenase 1-like, mitochondrial (monofunctional)

MTHFD2

Methylenetetrahydrofolate dehydrogenase 2, mitochondrial (bifunctional)

MTHFD2L

Methylenetetrahydrofolate dehydrogenase 2-like, mitochondrial (bifunctional)

MTHFR

Methylenetetrahydrofolate reductase

MTHFS

Methenyltetrahydrofolate synthetase

MTX

Methotrexate

OV

Ovarian serous cystadenocarcinoma

PAAD

Pancreatic adenocarcinoma

PAINS

Pan-assay interference compounds

PCFT

Proton-coupled folate transporter

PCPG

Pheochromocytoma and paraganglioma

PHGDH

Phosphoglycerate dehydrogenase

PRAD

Prostate adenocarcinoma

PSPH

Phosphoserine phosphatase

PTM

Post-translational modification

READ

Rectum adenocarcinoma

SARC

Sarcoma

SFXN1

Sideroflexin 1

SHMT1

Serine hydroxymethyltransferase 1, cytosolic

SHMT2

Serine hydroxymethyltransferase 2, mitochondrial

SKCM

Skin cutaneous melanoma

SLC1A4/SLC1A5

Solute carrier family 1 member 4/5

STAD

stomach adenocarcinoma

TGCT

Testicular germ cell tumors

THCA

Thyroid carcinoma

THYM

Thymoma

UCEC

Uterine corpus endometrial carcinoma

UCS

Uterine carcinosarcoma

UVM

Uveal melanoma

TCGA

The Cancer Genome Atlas

THF

Tetrahydrofolate

TYMS

Thymidylate synthase

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsptsci.0c00223.

• Gene Set Enrichment Analysis data set (XLSX)

The authors declare no competing financial interest.