Pharmacodynamic determinants of mitochondrial one-carbon flux and serine hydroxymethyltransferase inhibition in human tumors

Abstract

One-carbon (C1) metabolism includes cytosolic and mitochondrial pathways connected by interchange between serine, glycine, and formate. Mitochondrial C1 metabolism through serine hydroxymethyltransferase (SHMT) 2 generates glycine and C1 units for de novo nucleotide biosynthesis in the cytosol, whereas cytosolic SHMT1 consumes C1 units and glycine. Folates and classical antifolates are transported into tumors by facilitative folate transporters (reduced folate carrier [RFC] and proton-coupled folate transporter [PCFT]) and are metabolized to polyglutamates by folylpolyglutamate synthetase (FPGS). Folate transporter–null HeLa cells were engineered to express RFC under the control of a tetracycline-inducible promoter. Constitutive expression of PCFT and/or FPGS increased cytosolic and mitochondrial folates over that of RFC alone. By targeted metabolomics, the C1 flux in mitochondria through SHMT2 paralleled RFC transport and folate accumulation in mitochondria and cytosol, whereas the SHMT1 flux was constant. Expression of PCFT resulted in further increased C1 flux through SHMT2, in excess of SHMT1. In vitro inhibition of cell proliferation by targeting SHMT1/2 with pyrrolo[3,2-d]pyrimidine antifolates (eg, AGF347) decreased with increasing RFC and with PCFT. Inhibition by AGF347 (not SHIN1/2) was stimulated with ectopic FPGS, accompanying increased AGF347 polyglutamates; decreased sensitivities were seen for nonclassical SHMT1/2 inhibitors (SHIN1/2), which are neither substrates for facilitative transport nor polyglutamylation. Our results document the complex interrelationships among (anti)folate membrane transport, polyglutamylation, and C1 fluxes through SHMT1 and SHMT2. They also demonstrate the profound impact of physiologic folates on antitumor activities and the extraordinary promise of multitargeted pyrrolo[3,2-d]pyrimidine antifolates for cancer therapy.

Significance Statement

Novel pyrrolo[3,2-d]pyrimidine antifolates typified by AGF347 target serine hydroxymethyltransferase (SHMT) 2 in the mitochondria and SHMT1 and de novo purine biosynthesis in the cytosol. This manuscript documents the complex interrelationships among (anti)folate membrane transport, polyglutamylation, and one-carbon fluxes through SHMT1 and SHMT2 in the context of physiologic folate levels. The results document the therapeutic promise of classical multitargeted pyrrolo[3,2-d]pyrimidine antifolates typified by AGF347. These novel compounds offer an exciting new platform for one-carbon–targeted drug development for cancer.

1. Introduction

One-carbon (C1) metabolism is essential for cancer cell survival, proliferation, and maintenance of redox homeostasis.¹ Reflecting this, drugs targeting C1 metabolism have been used to treat cancer for decades.² Although C1 metabolism relies on folate cofactors, reflecting their hydrophilic nature, extracellular folates do not diffuse across cellular membranes and therefore require facilitative transport.³ Major facilitative folate transporters in mammalian cells include the reduced folate carrier (SLC19A1; RFC)⁴ and the proton-coupled folate transporter (SLC46A1; PCFT).^3,5 RFC is the major tissue folate transporter and is ubiquitously expressed in normal tissues and tumors.⁴ Whereas PCFT is expressed in tissues such as the choroid plexus and the upper gastrointestinal tract,³ PCFT has limited expression in most normal tissues.⁶ Notably, PCFT is highly expressed in many solid tumors,⁷ making it an attractive target for tumor-selective delivery of cancer therapeutics.^6,8

C1 metabolism encompasses distinct cytosolic and mitochondrial pathways connected by an interchange between serine, glycine, and formate^1,9,10 (Fig. 1).¹¹ Serine is synthesized from glucose in the cytosol. Following its transport into the mitochondria,¹¹ serine is metabolized by serine hydroxymethyltransferase 2 (SHMT2), 5,10-methylene tetrahydrofolate dehydrogenase (MTHFD) 2, and MTHFD1L, generating glycine and formate. In the cytosol, mitochondrial-derived formate associates with tetrahydrofolate (THF) and is used for de novo purine nucleotide (10-formyl THF) and thymidylate (5,10-methylene THF) biosynthesis, and for the synthesis of S-adenosyl methionine (5-methyl THF) for biological methylation reactions.

Fig. 1.

One-carbon metabolism. This schematic illustrates the transport of folates by RFC and PCFT and folate-dependent C1 metabolic pathways in the cytosol and mitochondria. Following internalization, folates are transported into the mitochondria by SLC25A32. Following its transport into the mitochondria by SFXN1 and 3,¹¹ serine is metabolized by SHMT2 to glycine, accompanying the conversion of THF to 5,10-methylene tetrahydrofolate (5,10-me THF). 5,10-me THF is oxidized in a NAD(P)⁺-dependent step by MTHFD2 (or MTHFD2L) to 10-formyl tetrahydrofolate (10-CHO-THF). Finally, 10-CHO-THF is hydrolyzed by MTHFD1L to THF and formate, the latter of which is transported into the cytosol.¹ In the cytosol, mitochondrial-derived formate associates with THF forms and is used for de novo purine nucleotide (10-CHO-THF) and thymidylate (5,10-me THF) biosynthesis. DHF, dihydrofolate; FAICAR, formyl 5-aminoimidazole-4-carboxamide ribonucleotide; fGAR, formyl glycinamide ribonucleotide; GR, glutathione reductase; GS, glutathione synthetase; GSH, glutathione; MTHFD1, 5,10-methylene tetrahydrofolate dehydrogenase 1; MTHFD2(L), 5,10-methylene tetrahydrofolate dehydrogenase 2(-like); MTHFR, 5,10-methylene tetrahydrofolate reductase; PGs, polyglutamates; PHGDH, phosphoglycerate dehydrogenase; PRPP, phosphoribosyl pyrophosphate; PSAT1, phosphoserine aminotransferase 1; PSPH, phosphoserine phosphatase; TS, thymidylate synthase. Created in BioRender, https://BioRender.com/t66v200.

Mitochondrial C1 metabolism from serine is the principal source of glycine and C1 units for a wide range of biosynthetic and regulatory processes in the cytosol.^1,9 Further, mitochondrial C1 metabolism is frequently reprogrammed in cancer cells, and the mitochondrial C1 enzymes SHMT2 and MTHFD2 are among the top differentially expressed metabolic enzymes from a substantial number of cancers and corresponding normal tissues.^12,13

We discovered novel pyrrolo[3,2-d]pyrimidine antifolates (eg, AGF347, AGF359) which target SHMT2 and deplete tumor cells of glycine and C1 units.14, 15, 16 These novel compounds also inhibit key secondary cytosolic targets, including serine hydroxymethyltransferase 1 (SHMT1) and the C1-dependent purine nucleotide biosynthetic enzymes glycinamide ribonucleotide (GAR) formyltransferase (GARFTase) and 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR) formyltransferase (AICARFTase)¹⁵ (Fig. 1). Multitargeted inhibition by AGF347 resulted in a broad spectrum of antitumor activity in vitro with promising in vivo efficacy toward pancreatic and ovarian cancer xenografts.^14,16

Cellular accumulation of AGF347 and related pyrrolo[3,2-d]pyrimidine compounds is facilitated by RFC and PCFT.^16,17 Following internalization, these compounds (like their folate counterparts) are metabolized to polyglutamate conjugates by folylpolyglutamate synthetase (FPGS) in the cytosol and mitochondria,^17,18 resulting in drug retention and increased inhibition of C1 enzyme targets.¹⁹ Thus, FPGS levels are important determinants of the antitumor efficacies of AGF347 and related pyrrolo[3,2-d]pyrimidine antifolates.¹⁸ Whereas others reported pyrazolopyran inhibitors of SHMT1 and SHMT2 (SHIN1, SHIN2),^20,21 these nonclassical analogs are unlikely to rely on facilitative transport, nor are they metabolized to polyglutamates.

Although SHMT1 catalyzes the conversion of glycine to serine in the cytosol (Fig. 1),¹ loss of SHMT2 activity is accompanied by a compensatory reversal of SHMT1 catalysis and synthesis of C1 units and glycine from serine.²² Accordingly, maximal suppression of tumor cell proliferation by SHMT2-targeted antifolates requires the inhibition of both SHMT1 in the cytosol and SHMT2 in the mitochondria.^14,16,22

However, a recent study²³ inferred a more complex relationship between the C1 fluxes through SHMT1 vis-à-vis SHMT2 and folate transport and suggested that previously unrecognized pharmacodynamic factors, most notably levels of RFC, were important determinants of these respective C1 fluxes.²³ Indeed, based on studies in an assortment of cancer cell lines (including both leukemias and tumors), Lee et al²³ concluded that cancer cells expressing low RFC levels are far more reliant on the cytosolic C1 flux through SHMT1 than on SHMT2 in the mitochondria, and that SHMT1 rather than SHMT2 is likely to be a more relevant drug target under these conditions. However, the impact of PCFT or FPGS was not considered. Further, these experiments were all performed in the presence of folic acid (FA), a nonphysiological folate form, raising uncertainty about the physiologic relevance of the authors’ conclusions.

With this in mind, we systematically assessed critical cellular determinants of C1 homeostasis and antitumor activities of SHMT1/2-targeted inhibitors as a necessary pretext for further developing this promising class of tumor-targeted agents. These include the impacts of media folate forms and concentrations, the expression of the major facilitative folate transporters RFC and PCFT, and the role of FPGS metabolism of (anti)folates to their polyglutamate forms in the cytosol and mitochondria. Our goals are to unequivocally identify relevant cellular pharmacodynamic determinants of SHMT targeting, while also establishing critical cellular determinants predictive of antitumor responses to this emerging new class of antitumor agents.

2. Materials and methods

2.1. Chemicals

Methotrexate (MTX) and leucovorin ((6R, S) 5-formyl-THF) (LCV) were provided by the National Cancer Institute. [2,3,3-²H]-L-Serine (98%) was purchased from Cambridge Isotope Laboratories, Inc. Syntheses of AGF347 [4-(4-(2-amino-4-oxo-3,4-dihydro-5H-pyrrolo[3,2-d]pyrimidin-5-yl)butyl)-2-fluorobenzoyl)-L-glutamic acid] and AGF359 [4-(3-(2-amino-4-oxo-3,4-dihydro-5H-pyrrolo[3,2-d]pyrimidin-5-yl) propyl)-2-fluorobenzoyl-L-glutamic acid] were previously reported.¹⁵ SHIN1²⁰ and SHIN2²¹ were purchased from MedChem Express. [3′,5′,7-³H]MTX (10–30 Ci/mmol), [3′,5′,7,9-³H]FA (32.9 Ci/mmol), and [³H]AGF347 (9 Ci/mmol) were purchased from Moravek Biochemicals. [³H](6S) 5-formyl THF ([³H]LCV) (13.9 Ci/mmol) was purchased from Moravek Biochemicals or was synthesized from [³H]FA.

2.2. Tumor cell line models

R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT HeLa cells expressing C-terminal hemagglutinin (HA)-tagged transporters were generated as previously described.²⁴

The human PCFT gene was knocked out in HCT116 colon cancer cells by CRISPR/Cas9 (Synthego); the guide RNA sequence was GCGGTGCCACAGATACTGCG. Single clones were isolated from a knockout (KO) pool, and genomic DNAs were isolated (QIAmp DNA Minikit; Qiagen). Genomic DNAs from wild-type (WT) and KO clones were amplified by polymerase chain reaction (PCR) using Easy-A high fidelity polymerase (Agilent Technologies) and forward (CCACCATCCAAAATGCACCC) and reverse (CTCCCCGCCGGACATTTAAG) primers. Amplified DNAs were electrophoresed on 1% agarose gel in 1X TAE (Tris-acetate-EDTA) buffer, extracted and purified using the QIAEXII Gel Extraction Kit (Qiagen), and sequenced by Sanger sequencing (Azenta) using the forward primer (above). The “Inference of CRISPR Edits” tool (Synthego) was used to analyze DNAs from WT and KO clones to generate KO scores. Clones were selected that gave a minimum KO score of 99% and were confirmed by direct assay of PCFT transport activity with [³H]MTX (below).

2.3. Generation of FPGS overexpressing cells

Full-length human FPGS DNA, including the N-terminal 42 amino acid leader sequence²⁵ was isolated from the hFPGS-pCMV-SPORT6 plasmid (Horizon Discovery) by introducing an EcoRV restriction site via mutagenesis PCR at the 5′ end of the gene upstream of the start codon. The full-length FPGS DNA fragment was excised from the plasmid using EcoRV and XbaI restriction enzymes and inserted into the pcDNA3.1 Hygro (+) vector (Invitrogen Life Technologies) (under control of the CMV promoter) between the EcoRV and XbaI cloning sites. A FLAG tag (DYKDDDDK) was introduced at the 3′ terminus of the FPGS construct by mutagenesis PCR in the pcDNA3.1 Hygro (+) plasmid, and the construct was confirmed by Sanger sequencing (Azenta).

R1-11/Tet-On-RFC cells were transfected with the pcDNA3.1-Hygro-FPGS^FLAG construct using Lipofectamine 2000 (Thermo Fisher Scientific), followed by selection with hygromycin (300 μg/mL) until a stable mixture was obtained. Single clones were isolated and screened in 96-well plates with an Odyssey M infared imager (Licor) by “in-cell-westerns” per the manufacturer’s protocol using a mouse anti-FLAG antibody (Sigma). Multiple “positive” clones with FPGS overexpression were confirmed by western blotting with anti-FLAG and rabbit anti-FPGS (Invitrogen) antibodies (below) and were selected for subsequent analyses (Supplemental Fig. 6). Tet-On-RFC/PCFT HeLa cells were transfected with the pcDNA3.1-Hygro-FPGS^FLAG construct using the 4D-Nucleofector X-Unit (Lonza) and program CN-114. Stable FPGS-expressing cell clones were identified, isolated, and characterized as described above.

2.4. Cell culture

R1-11/RFC2 and R1-11/PCFT4 HeLa cells with stable expression of RFC and PCFT, respectively, were previously described.^6,26 The R1-11/RFC2, R1-11/PCFT4, R1-11/Tet-On-RFC, R1-11/Tet-On-RFC/PCFT, R1-11/Tet-On-RFC/FPGS, and R1-11/Tet-On-RFC/PCFT/FPGS clonal cell lines were maintained in RPMI 1640 (Sigma-Aldrich) supplemented with 10% FBS (Sigma-Aldrich), 100 units/mL penicillin/100 μg/mL streptomycin, and 2 mM L-glutamine. Five to 7 days prior to experimental set-up (below), the R1-11/Tet-On-RFC, R1-11/Tet-On-RFC/PCFT, and FPGS overexpressing cells were cultured in folate-free RPMI 1640 (FFRPMI) (Thermo Fisher Scientific) supplemented with 10% FBS, 100 units/mL penicillin/100 μg/mL streptomycin, 2 mM L-glutamine, 60 μM adenosine, and 10 μM thymidine. HCT116 WT and HCT116 PCFT KO cells were maintained in FFRPMI supplemented with 10% dialyzed FBS (dFBS), 100 units/mL penicillin/100 μg/mL streptomycin, 2 mM L-glutamine, and 25 nM LCV. Cells were routinely monitored for Mycoplasma (Venor GeM Mycoplasma Detection Kit; Sigma-Aldrich). Cultures were maintained in a humidified incubator at 37 °C with 5% CO₂.

Cell proliferation assays were performed in 96 well plates (4000 cells/well; 200 μL/well) with increasing concentrations of inhibitor (0–5000 nM) in folate-, glycine-, and serine-free RPMI 1640 (Thermo Fisher Scientific), supplemented with 10% dFBS, 100 units/mL penicillin/100 μg/mL streptomycin, 2 mM L-glutamine, and 285 μM serine and 25 nM LCV. Tet-on cells were seeded in 96-well plates with increasing concentrations of doxycycline (Dox) (2.5, 5, 10, 25 ng/mL) to induce RFC along with the SHMT2 inhibitors (eg, AGF347, SHIN1) over a range of concentrations. Cells were incubated at 37 °C (8 days for Tet-on-RFC and Tet-on-RFC/PCFT cells; 4 days for HCT116 cells); cell viabilities were assayed using the Cell Titer Blue fluorescence assay (Promega) and a fluorescence plate reader (Agilent BioTek Synergy H1; excitation 560 nm, emission 590 nm). IC₅₀ values were calculated using Prism 10.0 (GraphPad Software).

2.5. Synthesis of [³H](6S) 5-formyl THF

[³H]LCV was synthesized from [³H]FA using dihydrofolate reductase (DHFR) (AbCam) based on the protocol by Moran and Colman.²⁷ To enzymatically convert [³H]FA to [³H](6S) 5-formyl THF, 10 μCi of [³H]FA (∼325 pmol) was dried down under nitrogen to remove ethanol, then solubilized in 100 μL sodium phosphate buffer (50 mM, pH 7.0) containing 2-mercaptoethanol (50 mM) and incubated with excess DHFR (4.5 μg) and NADPH (45 nmol) at 37 °C in the dark under a nitrogen atmosphere. After a 2-hour incubation, 900 μL of 100 mM formic acid in 50 mM sodium phosphate buffer containing 1% 2-mercaptoethanol was added. The pH was adjusted to 4.5 (7.5 μL 10N NaOH), 5 mg 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (Sigma) was added, and the reaction was sealed under a nitrogen atmosphere, and then incubated in the dark at room temperature. After 1 hour, the pH was adjusted to pH 6.4 (5 μL 10 N NaOH) and the [³H](6S) 5-formyl THF was purified by high-performance liquid chromatography (HPLC) on an Infinity II 1260 HPLC system (Agilent Technologies Inc) using a 5 μm C18 column (Kromasil). Elution used a linear gradient from 100% mobile phase A (100 mM sodium acetate, pH 5.5) to 80% mobile phase A/20% mobile phase B (100% acetonitrile) over 40 minutes. Fractions containing the [³H](6S) 5-formyl THF were collected, combined, and dried down with a SpeedVac vacuum concentrator (Eppendorf), and then reconstituted in ethanol/water (50%/50%). The radiospecific activity was calculated from the measured counts per minute, the counting efficiency, and the UV absorbance at 280 nM (extinction coefficient 37,200 M^–1 cm^–1).

2.6. Accumulation of [³H]LCV, [³H]FA, and [³H]AGF347

To measure the accumulations of different ³H-labeled (anti) folate compounds ([³H]LCV, [³H]FA, [³H]AGF347) in culture, folate-depleted R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells were seeded in 60 mm dishes (1.3 × 10⁶ cells/dish and 1.2 × 10⁶ cells/dish, respectively) in folate-, glycine-, and serine-free RPMI 1640 supplemented with dFBS, 2 mM L-glutamine, 100 units/mL penicillin/100 μg/mL streptomycin, 130 μM glycine, 285 μM serine, 60 μM adenosine, 10 μM thymidine, and 25 nM LCV (or 200 nM FA or 2 μM FA). After 48 hours, RFC was induced in the R1-11/Tet-On-RFC cells at low (1 ng/mL) or moderate (10 ng/mL) Dox concentrations (0 ng/mL Dox was not used as the cells are not viable in the absence of RFC without adenosine and thymidine supplementation). The R1-11/Tet-On-RFC/PCFT cells were uninduced (0 ng/mL Dox) or had moderate induction of RFC (10 ng/mL Dox). Tritiated (anti)folates, [³H]LCV (25 nM), [³H]FA (200 nM, 2 μM), or [³H]AGF347 (0.2 μM), were added at the same time as Dox and incubated for 48 hours. Uptake was quenched using ice-cold PBS. The cells were washed with PBS 3 times to remove residual [³H](anti)folates; proteins were solubilized with 0.5 N NaOH. Radioactivity was measured using a Tri-Carb 2910 TR scintillation counter (Perkin Elmer) and normalized to proteins using the Folin phenol reagent.²⁸ Accumulations of [³H](anti)folates were expressed as pmol/mg protein.

To measure mitochondrial and cytosolic [³H](anti)folates, the cells were seeded in three 100 mm dishes per condition (R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/FPGS, 3.4 × 10⁶ cells/dish; R1-11/Tet-On-RFC/PCFT and R1-11/Tet-On-RFC/PCFT/FPGS, 3.0 × 10⁶ cells/dish), as described above. After 48 hours, RFC was induced with Dox, and [³H]LCV (25 nM) or [³H]AGF347 (0.2 μM) was also added. In cells treated with AGF347, 130 μM glycine and 60 μM adenosine were added. Following a 48-hour incubation with tritiated substrates, the cells were washed 3 times with ice-cold PBS, and mitochondrial and cytosolic fractions were isolated²⁹ (Mitochondrial Isolation Kit for Cultured Cells, Thermo Scientific). All buffers were supplemented with EDTA-free cOmplete Mini protease inhibitor (Roche Diagnostics). The mitochondrial pellets were washed an additional 2 times with reagent C. The mitochondrial pellets were solubilized with 0.5 N NaOH; radioactivity from the cytosolic and mitochondrial fractions was measured as above and normalized to protein concentrations of the individual compartments.

2.7. Measurement of [³H]AGF347 polyglutamates

To quantify [³H]AGF347 polyglutamates, mitochondrial and cytosolic fractions were isolated (above); 400 μL of 50 mM sodium phosphate buffer (pH 6.0) was added to the mitochondrial pellets, followed by sonication (2 × 10 seconds, 20% amplitude on Qsonica sonicator). Sodium phosphate buffer was added to the cytosolic fraction to achieve a final concentration of 50 mM. To determine the total radioactivity in the mitochondrial and cytosolic fractions, a portion (40 μL and 160 μL, respectively) was counted with a scintillation counter, and the amounts (in pmol) were normalized to the protein concentrations. A 220 μL aliquot from each fraction (including 12.5 nmol nonradioactive AGF347 standard) was boiled for 10 minutes and centrifuged at 14,000 rpm at 4 °C for 15 minutes. Supernatants were fractionated on an Agilent Infinity II 1260 HPLC system using a Waters 4 μm Nova-Pak C18 column equipped with a Nova-Pak 4-mm C18 guard column (Waters Corporation). Mobile phase A was 100 mM sodium acetate (pH 5.5), and mobile phase B was 100% acetonitrile. A gradient from 100% mobile phase A to 80% mobile phase A/20% mobile phase B over 30 minutes was used to separate the polyglutamate forms. Absorbance at 280 nm was used to monitor the elution of the nonradioactive AGF347 standards. Fractions were collected every 10 seconds, and radioactivity was measured with a scintillation counter. The fractional amounts of the AGF347 parent drug and AGF347 polyglutamates were calculated (as pmol/mg) from the total pmol/mg and the relative percentages of the individual mitochondrial and cytosolic AGF347 monoglutamate and polyglutamate forms. Relative retentions of the AGF347 polyglutamyl and monoglutamyl forms were compared to AGF347 triglutamate (AGF347-PG₃) and pentaglutamate (AGF347-PG₅) standards.¹⁹ To further verify that the radioactive peaks were polyglutamate forms of AGF347, an extract from R1-11/Tet-On-RFC/PCFT cells was treated for 16 hours at 32 °C with chicken pancreas γ-glutamyl hydrolase¹⁸ in 0.5 mL sodium borate (0.1 M, pH 7.8) containing 10 mM 2-mercaptoethanol.

2.8. Membrane transport assays

Transport assays for RFC and/or PCFT were previously described.²⁴ Cells were seeded in 60 mm dishes as above. After 48 hours, increasing concentrations of Dox were added (0, 1, 2.5, 5, 10, 25 ng/mL) in FFRPMI supplemented with dFBS, 2 mM L-glutamine, 100 units/mL penicillin/100 μg/mL streptomycin, 130 μM glycine, and 285 μM serine, and different folate forms and concentrations (25 nM LCV, 200 nM FA, or 2 μM FA). After 48 hours, transport assays were performed. To assay RFC transport in the HeLa cell lines, incubations were at 37 °C in Hepes-buffered saline (20 mM Hepes, 140 mM NaCl, 5 mM KCl, 2 mM MgCl₂, and 5 mM glucose; for pH 7.2). PCFT transport was assayed at 37 °C in MES-buffered saline (20 mM MES, 140 mM NaCl, 5 mM KCl, 2 mM MgCl₂, 5 mM glucose, pH 5.5). Following 2-minute incubation with [³H]MTX (0.5 μM), the buffer was removed, and uptake was quenched with ice-cold PBS. The dishes were washed 3 times with ice-cold PBS, and the cells were solubilized using 0.5 N NaOH. Radioactivity and proteins were quantified as described above; uptake was expressed in units of pmol/mg.

Transport kinetics experiments were performed using R1-11/RFC2 and R1-11/PCFT4 HeLa cells.^7,26 Cells were cultured as above and seeded at a density of 1.2 × 10⁶ cells/dish in 60 mm dishes in FFRPMI supplemented with dFBS, 2 mM L-glutamine, 100 units/mL penicillin/100 μg/mL streptomycin, and 25 nM LCV. For RFC kinetics assays, the cells were washed twice with room temperature PBS and once with 37 °C Hanks’ balanced salts solution (HBSS). The cells were incubated at 37° C for 10 minutes with HBSS, and then incubated with HBSS containing [³H]MTX (0.2–5 μM) or [³H]AGF347 (0.05–2 μM) for 2 minutes. Dishes were then processed as described above. For PCFT transport kinetics, cells were washed with room temperature PBS and once with 37 °C PBS; cells were incubated at 37 °C for 10 minutes with 37 °C PBS, then incubated with pH 5.5 MES-buffered saline containing [³H]MTX (0.2–5 μM) or [³H]AGF347 (0.05–2 μM) for 2 minutes and processed as above. Data were analyzed with Lineweaver-Burk plots for calculating K_m and V_max values.

2.9. Metabolomics analysis

R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells were seeded in 60 mm dishes (above) in folate-, glycine-, and serine-free RPMI 1640 supplemented with dialyzed FBS, 2 mM L-glutamine, 100 units/mL penicillin/100 μg/mL streptomycin, 130 μM glycine, 285 μM serine, 60 μM adenosine, 10 μM thymidine, and 25 nM LCV. After 48 hours, the cells were induced with increasing concentrations of Dox (0, 1, 2.5, 5, 10, 25 ng/mL); adenosine and thymidine were not supplemented after the cells were induced. After 24 hours, the medium was replaced with fresh folate-, glycine-, and serine-free RPMI 1640 supplemented with dFBS, 2 mM L-glutamine, 100 units/mL penicillin/100 μg/mL streptomycin, 130 μM glycine, 25 nM LCV, and 250 μM of [2,3,3-²H]serine, along with the Dox additions. After 24-hour incubation at 37 °C, the cells were washed rapidly with ice-cold PBS 3 times, and allowed to rock on dry ice in 80% methanol for 10 minutes. Then, the cells were scraped with a cell scraper, transferred to 1.5 mL tubes, and vortexed for 30 seconds. The samples were centrifuged at 14,000 rpm at 4 °C for 15 minutes to precipitate the proteins. The supernatants containing the metabolites were collected and stored at –80 °C; the protein pellets were solubilized in 0.5 N NaOH, and the protein concentrations were determined by the Lowry assay²⁸ for normalization of the metabolite concentrations to cellular proteins.

The cell extract supernatants were aliquoted and dried in a CentriVap Refrigerated Centrifugal Concentrator at 6 °C, and stored at –80 °C until analysis. The dried cell extracts were reconstituted in 50 μL water, followed by vortex-mixing and centrifugation, and the reconstituted supernatants were subjected to liquid chromatography–mass spectrometry (LC-MS)/MS analysis using an AB Sciex QTRAP 6500 LC-MS/MS system in the Karmanos Cancer Institute Pharmacology and Metabolomics Core, as described previously.^14,16 Briefly, [2,3,3-²H]serine isotopomers were eluted on a Waters ACQUITY UPLC BEH Amide column (1.7 μm, 2.1 × 50 mm) and analyzed under the positive ionization mode using multiple reaction monitoring. ²H-dTTP isotopomers were eluted on a Waters Atlantis T3 column (3 μm, 2.1 × 100 mm) and analyzed under the negative ionization mode using multiple reaction monitoring. The isotopomer concentrations were corrected for natural abundance and normalized to cell protein concentrations.

2.10. Western blot analysis

Cells were plated in 60 mm dishes for western blot analysis. After 48 hours of induction, the cells were washed with ice-cold PBS, collected by scraping, and sonicated in 10 mM Tris-HCl containing protease inhibitor (cOmplete; Roche Diagnostics). Crude membranes were isolated by ultracentrifugation at 37,000 rpm for 30 minutes.²⁴ The supernatants were removed, and the pellets were suspended in 10 mM Tris-HCl (pH 7.0) containing 2% SDS and protease inhibitor. To resolve RFC and PCFT in the Tet-on-RFC/PCFT cells, the membrane fractions were deglycosylated by peptide-N-glycosidase F (New England Biolabs) prior to western blotting. To detect HA-tagged RFC and PCFT proteins, 20 μg of crude membrane was electrophoresed on 7.5% SDS-PAGE at 200 V and transferred to polyvinylidene difluoride membranes (Thermo Fisher Scientific) at 250 mA. Mouse anti–HA-monoclonal antibody (Biolegend) was used to detect HA-tagged RFC and PCFT. Na⁺/K⁺ ATPase was probed as a loading control using a mouse anti-Na⁺/K⁺ ATPase monoclonal antibody (Novus Biologicals). Blots were developed using IRDye 800CW-conjugated goat anti-mouse IgG (LI-COR Biosciences) secondary antibodies. The membranes were scanned using an Odyssey infrared imaging system (LI-COR Biosciences).

To assay FLAG-tagged FPGS on western blots, 60 μg of whole cell extracts were prepared from R1-11/Tet-On-RFC, R1-11/Tet-On-RFC/PCFT, R1-11/Tet-On-RFC/FPGS, and R1-11/Tet-On-RFC/PCFT/FPGS transfectants and electrophoresed on 7.5% SDS-PAGE at 200 V. Proteins were transferred to polyvinylidene fluoride membranes, as described above. The blots were probed using a rabbit anti-FPGS polyclonal antibody (Thermo Fisher Scientific) and a mouse anti-FLAG monoclonal antibody (Sigma-Aldrich). The blots were developed using IRDye 800CW-conjugated goat anti-mouse IgG (LI-COR Biosciences) and IRDye 680CW-conjugated goat anti-rabbit IgG (LI-COR Biosciences) and scanned as described above.

2.11. Statistical analysis

Data were summarized as mean values with standard deviations (SDs). The distributions of continuous outcomes were assessed, and nonparametric methods were applied when necessary. Comparisons were performed using unpaired t tests or the nonparametric Mann-Whitney U tests, followed by Holm’s method to correct for multiple comparisons.

3. Results

3.1. Impact of facilitative folate transporter levels and extracellular folate forms on transport activity and intracellular accumulation of folates

To systematically evaluate the impact of RFC and PCFT on C1 metabolism via SHMT2/SHMT1, we used folate transporter-null R1-11 HeLa cells³⁰ for which RFC is expressed under the control of a tetracycline-inducible promoter in the presence of Dox without PCFT (R1-11/Tet-On-RFC), or with constitutively expressed PCFT (R1-11/Tet-On-RFC/PCFT).²⁴ Both RFC and PCFT levels in the engineered R1-11 cell lines approximate those in cultured tumor cells (including WT HeLa).⁷

The engineered cells were cultured in 25 nM (6R,S) 5-formyl THF (LCV) such that the “active” 6S isomer (12.5 nM) is a “surrogate” of the major circulating folate (eg, 5-methyl THF) at a similar concentration to that in human serum (26.9 nM³¹). For comparison, additional treatments included moderate (200 nM) or excess (2 μM) FA, as described.²³ Both cell lines were treated with Dox (1–25 ng/mL) to induce RFC to various levels without or with constitutive PCFT.^7,24 Transcript levels for C1-related genes in R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells cultured in 25 nM LCV (including folate transporters and C1 metabolism genes) are summarized in Supplemental Fig. 1.

We initially assayed RFC and PCFT protein levels on western blots and measured transport activity with [³H]MTX (0.5 μM) over 2 minutes at pH 7.2 and pH 5.5, respectively (approximating their respective pH optima).^4,5 When R1-11/Tet-On-RFC cells were cultured in 25 nM LCV, both RFC protein (Fig. 2A) and MTX transport at pH 7.2 (Fig. 2B) increased in direct proportion to the Dox concentration up to 25 ng/mL. Identical results were seen for RFC with the R1-11/Tet-On-RFC/PCFT cells (Fig. 2, C and D, left panel), whereas PCFT protein was constitutively expressed. The contribution of PCFT to net [³H]MTX transport over 2 minutes at pH 7.2 was nominal (Fig. 2D, left panel). PCFT transport was elevated in comparison to RFC transport at their respective pH optima (Fig. 2D).

Fig. 2.

Characterization of R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT transporter expression and transport activity. (A) R1-11/Tet-On-RFC cells were treated with increasing concentrations of Dox for 48 hours. Crude membranes were isolated by differential centrifugation, and HA-tagged RFC was detected by SDS-PAGE/western blots on 7.5% gels with HA-monoclonal antibody; Na⁺/K⁺ ATPase was used as a loading control. (B) RFC transport activity was measured in R1-11/Tet-On-RFC cells with 0.5 μM [³H]MTX for 2 minutes at pH 7.2 and 37 °C. (C) Crude membranes were isolated from R1-11/Tet-On-RFC/PCFT cells by differential centrifugation and deglycosylated (dg) using peptide-N-glycosidase F (PNGase F) prior to fractionation by SDS-PAGE on a 7.5% gel and western blotting as for panel (A) (to facilitate the resolution of the heavily glycosylated RFC and PCFT proteins). (D) (Left) RFC transport was measured by assaying 0.5 μM [³H]MTX uptake for 2 minutes at pH 7.2 at 37 °C; (right) PCFT transport was measured with 0.5 μM [³H]MTX uptake over 2 minutes at pH 5.5 and 37 °C. Data are represented as mean values ± SD from 3 biological replicates.

We extended our studies to include measurements of RFC and PCFT proteins and [³H]MTX uptake at their respective pH optima in R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells cultured in 25 nM LCV compared to cells cultured in 200 nM or 2 μM FA (Fig. 3, A and D). At 10 ng/mL Dox, R1-11/Tet-On-RFC cells grown in 2 μM FA expressed 61% decreased (±5%, P = .0023) RFC protein (Fig. 3A), accompanied by a comparable decrease in [³H]MTX transport (Fig. 3B) relative to cells grown in 25 nM LCV or 200 nM FA. Decreased RFC protein was accompanied by a similar decrease in levels of RFC transcripts (Supplemental Fig. 1), suggesting a post-transcriptional regulation of RFC at excessive concentrations of folate. In contrast, for R1-11/Tet-On-RFC/PCFT cells, neither the levels of RFC or PCFT proteins (Fig. 3D) nor transport activities (Fig. 3E) were appreciably impacted by the folate form or concentration used for cell culture.

Fig. 3.

Effect of physiologic and nonphysiological folates on transport activity and folate accumulation in R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT. R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells were treated with Dox for 48 hours to induce low (1 ng/mL) or physiological (10 ng/mL) levels of RFC. (A) R1-11/Tet-On-RFC cells were cultured in 25 nM LCV, 200 nM FA, or 2 μM FA and treated with Dox. RFC was detected on western blots of crude membranes isolated by differential centrifugation. HA-tagged RFC was detected by SDS-PAGE/western blots on 7.5% gels with HA-specific monoclonal antibody; Na⁺/K⁺ ATPase was used as a loading control. Densitometry was performed with the Odyssey imaging software, and results were expressed relative to the 10 ng/mL Dox condition at 25 nM LCV from 3 biological replicates. At 10 ng/mL Dox, R1-11/Tet-On-RFC cells grown in 2 μM FA expressed 61% decreased (± 5%, P = .0023) RFC protein. (B) RFC transport was measured with 0.5 μM [³H]MTX for 2 minutes at pH 7.2 and 37 °C in R1-11/Tet-On-RFC cells cultured in 25 nM LCV, 200 nM FA, or 2 μM FA. (C) Folate accumulation was measured over 48 hours in R1-11/Tet-On-RFC cells cultured with [³H]LCV (25 nM) or [³H]FA (200 nM or 2 μM). (D) R1-11/Tet-On-RFC/PCFT cells were cultured in 25 nM LCV, 200 nM FA, or 2 μM FA and treated with 0 or 10 ng/mL Dox. Crude membranes were isolated by differential centrifugation. Proteins were deglycosylated (designated “dg”) using peptide-N-glycosidase F (PNGase F) prior to fractionation by SDS-PAGE on a 7.5% gel and western blotting with HA-monoclonal antibody and Na⁺/K⁺ ATPase. (E) (Left) RFC transport was measured with 0.5 μM [³H]MTX for 2 minutes at pH 7.2 and 37 °C; (right) PCFT transport was measured with 0.5 μM [³H]MTX over 2 minutes at pH 5.5 and 37 °C. (F) Folate accumulation was measured over 48 hours in R1-11/Tet-On-RFC/PCFT cells cultured with [³H]LCV (25 nM) or [³H]FA (200 nM or 2 μM). For MTX transport and folate accumulation experiments (panels B, C, E, and F), data are represented as mean values ± SD from 3 biological replicates. An unpaired t test was used to compare 25 nM LCV and 200 nM FA, and 25 nM LCV and 2 μM FA; Holm’s method was used to correct for multiple comparisons. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001.

We measured the accumulation of tritiated folates in R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells during culture over 48 hours in [³H]LCV (25 nM) or [³H]FA (200 nM and 2 μM). For R1-11/Tet-On-RFC cells at a low level of RFC induction (1 ng/mL Dox), there was a ∼2.7-fold increased folate accumulation when cells were cultured with 200 nM [³H]FA compared to 25 nM [³H]LCV (Fig. 3C); however, at 10 ng/mL Dox, tritiated folate accumulations at 200 nM [³H]FA and 25 nM [³H]LCV were similar. In spite of decreased RFC in R1-11/Tet-On-RFC cells grown in 2 μM FA, accumulations of tritiated folates increased ∼16-fold and ∼3.8-fold at 1 and 10 ng/mL Dox, respectively, over those measured at 25 nM LCV (P = .00004 and .00015, respectively). Thus, intracellular folate accumulations are profoundly impacted by not only the level of RFC but also by the concentration and the form of extracellular folate. For the R1-11/Tet-On-RFC/PCFT cells, constitutive expression of PCFT increased the accumulation of radiolabeled folates from [³H]LCV and [³H]FA (2.5-fold to 25-fold) over those in R1-11/Tet-On-RFC cells (even at neutral pH; compare Fig. 3, C and F); differences in folate accumulations were variably preserved for the various treatments (2 μM FA > 200 nM FA ≈ 25 nM LCV) (Fig. 3F).

As folates accumulated in both the cytosol and mitochondria, we measured cytosolic and mitochondrial folate levels in R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells cultured in [³H]LCV (25 nM) over 48 hours at 0, 2.5, and 25 ng/mL Dox. Modest albeit increasing levels of radiolabeled folates accumulated from [³H]LCV in R1-11/Tet-On-RFC cells in both the cytosol (∼4.6-fold) and mitochondria (∼1.8-fold) from 0 to 25 ng/mL Dox (Fig. 4A; Table 1). In the R1-11/Tet-On-RFC/PCFT cells, mitochondrial and cytosolic folates increased (∼2.2-fold to 5.9-fold and ∼2.3-fold to 16-fold, respectively) compared to R1-11/Tet-On-RFC cells.

Fig. 4.

Evaluation of mitochondrial and cytosolic C1 metabolic flux as a function of folate transporter expression and folate supplementation. (A) R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells were treated with Dox (0–25 ng/mL) and [³H]LCV (25 nM) for 48 hours. Mitochondrial and cytosolic fractions were isolated; ³H-folates are expressed as pmol/mg in the respective compartments. (B) Schematic of serine isotope “scrambling”²² is shown. Heavy (²H) atoms in [2,3,3-²H]serine (blue circles) are metabolized in the mitochondria via SHMT2, MTHFD2 and MTHFD1L to [²H]formate which passes to the cytosol where it is metabolized to [²H]10-formyl THF and [²H]5,10-methylene THF (5,10-me THF), resulting in [²H]dTMP and [²H]dTTP (as M+1). In the cytosol, ²H atoms in [2,3,3-²H]serine (red circles) are metabolized via a reversal of SHMT1 catalysis to convert [2,3,3-²H]serine to [²H]glycine and [²H]5,10-methylene THF (M+2), resulting in [²H]dTMP and [²H]dTTP (as M+2). The serine isotopomer distribution in R1-11/Tet-On-RFC (C) and in R1-11/Tet-On-RFC/PCFT (D) cells is shown following incubation with [2,3,3-²H]serine over a range of Dox concentrations for 24 hours. Data are shown for unmetabolized [2,3,3-²H]serine (M+3 isotopomer), along with the other serine isotopomers (M+0, M+1, M+2). The dTTP isotopomer distributions in R1-11/Tet-On-RFC (E) and R1-11/Tet-On-RFC/PCFT (F) cells are shown following incubation with [2,3,3-²H]serine over a range of Dox concentrations for 24 hours. Data are shown for [²H]dTTP (M+0, M+1, M+2) for which the M+1 isotopomer reflects the mitochondrial C1 flux through SHMT2 and the M+2 isotopomer reflects the cytosolic C1 flux through SHMT1.^14,16,22 Data are presented as mean values ± SD from 5 biological replicates. P values were determined by unpaired t tests, followed by Holm’s correction for multiple comparisons. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001.

Table 1.

Mitochondrial and cytosolic accumulation of [³H]LCV and [³H]AGF347

Accumulation of [³H]LCV and [³H]AGF347 in mitochondrial and cytosolic fractions of cells treated with 0, 2.5, or 25 ng/mL Dox for 48 hours. Data are expressed as pmol/mg protein in the individual fractions. Data are represented as mean values (±SD) from 3 biological replicates.

Cell Line (in R1-11)	LCV (pmol/mg)						AGF347 (pmol/mg)
	Mitochondria			Cytosol			Mitochondria			Cytosol
	Doxycycline (ng/mL)
	0	2.5	25	0	2.5	25	0	2.5	25	0	2.5	25
Tet-On-RFC	1.20 (0.14)	2.05 (1.04)	2.21 (2.22)	1.02 (0.52)	4.68 (3.25)	4.69 (3.56)	15.5 (1.98)	19.4 (7.05)	27.5 (4.46)	5.76 (1.47)	9.84 (2.24)	26.5 (20.6)
Tet-On-RFC/PCFT	7.13 (5.50)	9.10 (2.82)	4.96 (2.50)	16.6 (3.90)	18.3 (3.86)	10.7 (4.07)	21.6 (11.3)	17.5 (8.99)	23.3 (15.9)	75.5 (27.7)	63.1 (15.2)	73.3 (17.8)
Tet-On-RFC/FPGS	7.86 (0.89)	10.5 (0.42)	12.9 (1.19)	7.22 (1.28)	39.0 (23.5)	36.3 (1.60)	16.1 (3.08)	57.8 (31.0)	86.9 (18.4)	18.5 (0.76)	179 (76.2)	382 (49.1)
Tet-on- RFC/PCFT/FPGS	18.1 (3.32)	18.4 (4.06)	14.3 (1.54)	27.1 (4.90)	28.9 (5.23)	23.3 (1.07)	262 (17.6)	245 (16.4)	305 (73.5)	557 (94.9)	595 (211)	651 (129)

Collectively, these results document the profound impact of medium folates (both forms and concentrations) and membrane transporter expression on the cellular accumulation of folates. Moreover, these findings also document the importance of using physiologically relevant concentrations of reduced folates for in vitro studies of C1 metabolism.

3.2. Mitochondrial C1 flux is affected by RFC expression levels and predominates in the presence of PCFT

The report by Lee et al²³ challenged a central tenet of C1 metabolism, namely that the C1 flux through SHMT2 predominates over SHMT1 in cancer cells, reflecting the essential role of SHMT2 in providing glycine and C1 units for cellular biosynthesis in the cytosol.²⁰ Rather, Lee et al²³ concluded that cancer cells cultured in 200 nM FA and expressing low levels of RFC depend primarily on the C1 flux through SHMT1 in the cytosol rather than on SHMT2 in the mitochondria, and that under these conditions, therapeutic targeting of SHMT1 over SHMT2 may be warranted.

To directly assess the impact of increasing RFC expression and constitutive PCFT on relative C1 fluxes through the mitochondrial and cytosolic C1 pathways, we induced RFC with Dox in R1-11/Tet-On-RFC (1–25 ng/mL) and R1-11/Tet-on-RFC/PCFT (0–25 ng/mL) cells cultured in 25 nM LCV in the presence of [2,3,3-²H]serine (250 μM).^14,16,22 We measured the metabolism of [2,3,3-²H]serine and incorporation of deuterium from [2,3,3-²H]serine into dTTP (M+1 isotopomer) over 24 hours by LC-MS/MS as a surrogate of the mitochondrial C1 flux, whereas incorporation of deuterium into M+2 dTTP measures the reverse cytosolic flux (serine-to-glycine) via SHMT1^14,16,22 (Fig. 4B).

In R1-11/Tet-On-RFC cells, the impact of RFC levels on serine catabolism was reflected in progressively decreased total (∼2.2-fold) and M+3 (∼2.4-fold) serine with increasing Dox (Fig. 4C). Increasing RFC levels from 1 to 25 ng/mL Dox inversely correlated with total serine and resulted in increasing M+1 dTTP over a ∼3.9-fold range, whereas M+2 dTTP was essentially constant (Fig. 4E). These results establish a direct correlation between RFC levels and the mitochondrial C1 flux through SHMT2. They also show that the flux through SHMT1 is independent of RFC levels and transport. An analogous increase in M+1 dTTP was seen when R1-11/Tet-On-RFC cells were cultured in 200 nM FA and RFC was induced at 1 and 10 ng/mL Dox (Supplemental Fig. 2); however, this was not seen when cells were cultured in 2 μM FA.

In the presence of PCFT (R1-11/Tet-On-RFC/PCFT), the mitochondrial C1 flux (M+1 dTTP) increased compared to that for R1-11/Tet-On-RFC cells and was independent of increasing levels of RFC (Fig. 4F compared to Fig. 4E). Further, [2,3,3-²H]serine (all isotopomers) dramatically decreased in the R1-11/Tet-On-RFC/PCFT cells compared to R1-11/Tet-On-RFC cells not expressing PCFT (Fig. 4D compared to Fig. 4C). This increased mitochondrial C1 flux in R1-11/Tet-On-RFC/PCFT cells was further reflected in substantially deceased levels of steady-state de novo purine biosynthetic intermediates (GAR and AICAR; suggesting increased C1 flux) compared to results for R1-11/Tet-On-RFC cells, although IMP levels were similar (Supplemental Fig. 3).

These results establish that for cells cultured in 25 nM LCV in the absence of PCFT, the mitochondrial C1 flux through SHMT2 closely reflects the level of RFC transport and the accumulation of intracellular (mitochondrial) folates, whereas the reverse flux through SHMT1 in the cytosol is unimpacted. When PCFT is expressed, the mitochondrial C1 flux is exacerbated and is independent of the levels of RFC.

3.3. Facilitated transport of pyrrolo[3,2-d]pyrimidine antifolates by RFC and PCFT

We discovered 5-substituted pyrrolo[3,2-d]pyrimidine inhibitors of SHMT2 (AGF347 and AGF359)^14,15 which are structurally distinct from the “non-classical” pyrazolopyran SHMT2 inhibitors SHIN1 and SHIN2^20,21 (Fig. 5A). Whereas both the pyrrolopyrimidine and pyrazolopyran inhibitors target SHMT2 in the mitochondria and SHMT1 in the cytosol,^14,15,21,22 AGF347 and related compounds also inhibit de novo purine nucleotide biosynthesis (at GARFTase and AICARFTase),^14,15 thus augmenting the impact on SHMT2 inhibition in limiting C1 units for downstream biosynthesis in the cytosol (Fig. 1).

Fig. 5.

Impact of expression of major facilitative folate transporters on antitumor efficacies of SHMT2 inhibitors and effect of FPGS overexpression. (A) Structures of the 5-substituted pyrrolo[3,2-d]pyrimidine antifolates (AGF347, AGF359) and pyrazolopyran compounds (SHIN1, SHIN2) are shown. (B) Total cell accumulations of 0.2 μM [³H]AGF347 over 48 hours in R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells are shown. (C) Results are shown for the distributions of [³H]AGF347 between the cytosolic and mitochondrial fractions in R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells. These results are summarized in Table 1. (D) IC₅₀ values are plotted for SHMT2-targeted antifolates from in vitro proliferation assays with WT and FPGS overexpressing (labeled O/E in legend) R1-11/Tet-On-RFC (upper panels) and R1-11/Tet-On-RFC/PCFT (lower panels) cells treated with increasing Dox. (E) Total cell accumulation of 0.2 μM [³H]AGF347 over 48 hours in R1-11/Tet-On-RFC/FPGS and R1-11/Tet-On-RFC/PCFT/FPGS are shown. (F) Results are shown for the distribution of [³H]AGF347 between the cytosolic and mitochondrial fractions in R1-11/Tet-On-RFC/FPGS and R1-11/Tet-On-RFC/PCFT/FPGS cells. These results are tabulated in Table 1. Data are represented as mean values ± SD from 3 or 4 biological replicates. P values were determined by unpaired t tests, followed by Holm’s correction for multiple comparisons. ∗P <.05; ∗∗P <.01; ∗∗∗P <.001.

We previously confirmed facilitated transport with [³H]AGF347 in engineered R1-11/RFC2 and R1-11/PCFT4 HeLa cells^7,26 with [³H]AGF347 and competition with specific ligands for RFC (PT523) and PCFT (AGF94).¹⁷ As an extension of these studies, we directly measured the transport kinetics of [³H]AGF347 with R1-11/RFC2 and R1-11/PCFT4 cells compared to [³H]MTX. Relative to MTX, K_m values for AGF347 decreased ∼4-fold for RFC (0.20 μM) and ∼10-fold for PCFT (0.06 μM); V_max values decreased ∼1.7-fold and ∼2.5-fold, respectively (Supplemental Table 1). From the V_max/K_m values, AGF347 is a preferred transport substrate over MTX for both RFC (22.5 vs 9.03) and PCFT (47.7 vs 12.8), with PCFT transport ∼2.1-fold greater than by RFC at their respective pH optima.

A distinguishing feature of the pyrrolo[3,2-d]pyrimidine antifolates AGF347 and AGF359 vis-à-vis the nonclassical inhibitors SHIN1 and SHIN2 relates to their requirement for facilitated transport by RFC and/or PCFT for internalization. We directly compared binding to RFC and PCFT by competition with [³H]MTX (0.5 μM) by 10 μM of these inhibitors in R1-11/RFC2 and R1-11/PCFT4 cells, respectively. The pyrrolo[3,2-d]pyrimidine compounds potently inhibited transport of [³H]MTX by RFC (95%; Supplemental Fig. 4, left panel,) and PCFT (90%; Supplemental Fig. 4, right panel), essentially equivalent to the established transport inhibitors (PT523³² and AGF94,³³ respectively), whereas the pyrazolopyran compounds SHIN1 and SHIN2 had no impact on [³H]MTX transport (Supplemental Fig 4). This establishes that SHIN1 and SHIN2 do not bind to RFC or PCFT.

We extended studies to include assays of [³H]AGF347 (0.2 μM) accumulation with increasing RFC (0–25 ng/mL Dox) over 48 hours in R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells. In R1-11/Tet-On-RFC cells, total cellular [³H]AGF347 increased ∼3.5-fold with Dox up to 25 ng/mL (Fig. 5B), reflecting ∼4.6-fold increased cytosolic [³H]AGF347 although mitochondrial [³H]AGF347 was modestly impacted (Fig. 5C). In the presence of constitutive PCFT (R1-11/Tet-On-RFC/PCFT cells), total intracellular [³H]AGF347 further increased (∼2.6-fold to 7.9-fold) and was independent of RFC levels (Fig. 5B). This was completely due to the increased cytosolic [³H]AGF347 (∼2.8-fold at 25 ng/mL Dox) as mitochondrial [³H]AGF347 accumulation was unchanged from that for R1-11/Tet-On-RFC cells (Fig. 5C). These results are summarized in Table 1.

3.4. Expression of folate transporters and sensitivity to SHMT2 inhibitors

An important goal of our study was to examine the complex relationships among (anti) folate membrane transport, cytosolic and mitochondrial C1 metabolic fluxes, and the antitumor efficacies of classical (AGF347, AGF359)^14,15 and nonclassical (SHIN1, SHIN2)^20,21 SHMT1/SHMT2 targeted agents (Fig. 5A).

We measured proliferation over 8 days for R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells treated with 2.5 to 25 ng/mL Dox to induce RFC in the presence of AGF347, AGF359, SHIN1, or SHIN2 (Fig. 5D). Paradoxically, increasing RFC levels (2.5 to 25 ng/mL Dox) and cellular accumulation of AGF347 in R1-11/Tet-On-RFC cells were accompanied by ∼3.4-fold decreased in vitro sensitivities to AGF347 (ie, increased IC₅₀ values for growth inhibition) (Fig. 5D, upper panels; Supplemental Table 2). Analogous results were seen with AGF359 and with SHIN1 and SHIN2. The presence of PCFT (R1-11/Tet-On-RFC/PCFT) further decreased the sensitivities to the SHMT2-targeted agents (increased IC₅₀ values), although these were largely independent of RFC levels (Fig. 5D, lower panels; Supplemental Table 2).

To assess whether these patterns of antifolate sensitivity could be generalized, we extended our studies to include HCT116 colorectal carcinoma cells that express endogenous RFC and PCFT.⁷ KO of PCFT in HCT116 cells resulted in ∼83% decreased PCFT transport of [³H]MTX without any effect on RFC transport of [³H]MTX (Supplemental Fig. 5, A and B, respectively). This was accompanied by significantly increased sensitivities to AGF347 and AGF359 (IC₅₀ decreases of 2.7-fold and 2.8-fold, respectively) but not to SHIN1 or SHIN2 (Supplemental Fig. 5C).

These results establish that increasing RFC adversely impacts antitumor sensitivities to SHMT2/SHMT1 targeted agents, and this effect is further exacerbated by the presence of PCFT.

3.5. Increased metabolism of pyrrolo[3,2-d]pyrimidine antifolates to polyglutamate forms increases drug accumulation and antitumor activities

FPGS is expressed as cytosolic and mitochondrial isoforms differing by an N-terminal 42 amino acid leader sequence.²⁵ FPGS metabolizes folates to polyglutamate forms in which additional L-glutamate residues are linked to the γ-carboxyl of the L-glutamate of the folate. Folate polyglutamates accumulate in both the cytosol and mitochondria,¹⁸ do not exchange between compartments²⁵ and are frequently associated with increased binding affinities for C1 enzymes.^34,35 AGF347 is also metabolized to polyglutamates in the mitochondria and cytosol^17,18 and pentaglutamyl AGF347 (AGF347-Glu₅) (Fig. 6A) is a ∼20-fold more potent inhibitor of SHMT2 than the parent AGF347.¹⁹ However, the pyrazolopyran inhibitors SHIN1 and SHIN2 (Fig 5A) cannot be metabolized by FPGS.

Fig. 6.

Analysis of AGF347 to polyglutamate forms in response to folate transporter and FPGS expression. (A) Structure of AGF347-Glu₅. (B) R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT (n = 3), R1-11/Tet-On-RFC/FPGS and R1-11/Tet-On-RFC/PCFT/FPGS (n = 2) cells were treated with Dox and 0.2 μM [³H]AGF347 for 48 hours; mitochondrial (Mito) and cytosolic (Cyto) compartments were isolated²⁹ and AGF347 parent and polyglutamate forms were quantified (by HPLC) and normalized to cell proteins. Results are presented as mean values ± SD. In panels (C–G), AGF347 monoglutamate and polyglutamate forms were resolved by reverse phase HPLC; individual fractions were collected and measured for radioactivity. The chromatograms are representative of 2 to 3 biological replicates. To verify that the early migrating species are AGF347 polyglutamates, a whole cell extract from the R1-11/Tet-On-RFC/PCFT cells was treated with chicken pancreas γ-glutamyl hydrolase to generate diglutamate and monoglutamate [³H]AGF347 forms (E).

We measured the metabolism of [³H]AGF347 (0.2 μM) to its polyglutamate forms in the cytosol and mitochondria of R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT HeLa cells over 48 hours, treated with 2.5 and 25 ng/mL Dox. For both the R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells, essentially all of the [³H]AGF347 forms in both the cytosol and mitochondria were polyglutamyl metabolites (Fig. 6B²⁹; Supplemental Table 3). The majority of the AGF347 polyglutamates had chain lengths exceeding 5 glutamates based on their HPLC elution times relative to AGF347-Glu₅ standard (Fig. 6, C and D). Treatment with γ-glutamyl hydrolase¹⁸ quantitatively converted the [³H]AGF347 metabolites to 2 additional forms, likely AGF347-Glu₂ (20 minutes; 53%) and parent AGF347 (23 minutes; 47%) (Fig. 6E). This confirms the identities of the [³H]AGF347 metabolites as polyglutamate forms of AGF347.^18,36

In the R1-11/Tet-On-RFC/PCFT cells, cytosolic [³H]AGF347 polyglutamates increased 2- to 4-fold over levels in R1-11/Tet-On-RFC cells; however, the levels of mitochondrial [³H]AGF347 polyglutamates were essentially identical in the presence or absence of PCFT (Fig. 6B; Supplemental Table 3). For the PCFT-expressing cells, there was a modest increase in the amount of unmetabolized AGF347 compared to the cells with only RFC (Fig. 6B; Supplemental Table 3).

We considered the possibility that the decreased in vitro efficacies of the pyrrolo[3,2-d]pyrimidine antifolates with increasing RFC and PCFT were the result of elevated cellular folate pools (Fig. 4A) that directly compete with classical antifolates for uptake from the cytosol to mitochondria, and for binding to cytosolic C1 targets (SHMT1, GARFTase, AICARFTase), as well as to SHMT2 in the mitochondria (Fig. 1). We reasoned that this could be circumvented at least in part by augmenting the synthesis of polyglutamate forms of the pyrrolo[3,2-d]pyrimidine antifolates. However, this would not be possible for SHIN1 and SHIN2. To directly test this possibility, we overexpressed full-length FPGS in the R1-11/Tet-On-RFC and R1-11/Tet-On-RFC/PCFT cells. Clones expressing ectopic FPGS (designated R1-11/Tet-On-RFC/FPGS and R1-11/Tet-On-RFC/PCFT/FPGS) were identified on western blots (Supplemental Fig. 6). We measured the impact of ectopic FPGS on the accumulation of cytosolic and mitochondrial [³H]folates from [³H]LCV (25 nM) and total and polyglutamyl [³H]AGF347 at 2.5 ng/mL and 25 ng/mL Dox.

For both R1-11/Tet-On-RFC/FPGS and R1-11/Tet-On-RFC/PCFT/FPGS cells, expression of ectopic FPGS increased the accumulation of folates from [³H]LCV (5.1-fold to 8.3-fold and 1.6-fold to 2.9-fold, respectively) and [³H]AGF347 (1.0-fold to 18-fold and 7.4-fold to 14-fold, respectively) in both the cytosol and mitochondria (Fig. 5, E and F; Table 1). Essentially all of the [³H]AGF347 was metabolized to [³H]AGF347 polyglutamates (Fig. 6B; Supplemental Table 3) including long chain length forms, as reflected in a distinct shift in the migration of the radioactive metabolites in the FPGS-transfected cells relative to the AGF347-Glu₅ standard (Fig. 6, F and G).

Ectopic expression of FPGS also sensitized R1-11/Tet-On-RFC and Tet-On-RFC/PCFT cells to AGF347 and AGF359, as reflected in decreased IC₅₀ values (2.3-fold to 12-fold and 31-fold to 72-fold, respectively) (Fig. 5D, upper and lower panels, respectively; Supplemental Table 2). However, for the nonclassical antifolates SHIN1 and SHIN2, the impact of FPGS overexpression on in vitro efficacy generally resulted in decreased drug efficacy (Fig. 5D; Supplemental Table 2).

Collectively, these data demonstrate that FPGS, in addition to the levels and activities of the major facilitative folate transporters, are critical determinants of the antitumor activities of pyrrolo[3,2-d]pyrimidine antifolates targeting mitochondrial C1 metabolism at SHMT2.

4. Discussion

Mitochondrial C1 metabolism from serine is a compelling tumor target, reflecting its role as the principal source of glycine and C1 units (as formate) for a wide array of biosynthetic and regulatory processes.^1,9 In many cancers, mitochondrial C1 metabolism is uniquely dysregulated.^12,13,16,37 This prompted the discovery of 2 classes of mitochondrial C1 inhibitors, each with significant therapeutic potential and both targeting SHMT2. These include the nonclassical pyrazolopyran compounds SHIN1²⁰ and SHIN2,²¹ and the classical pyrrolo[3,2-d]pyrimidine antifolates AGF347 and AGF359.^14,15

Both SHIN1 and AGF347 bind to the same THF binding pocket of SHMT2,^19,20 resulting in enzyme inhibition, with the potency of the SHIN compounds exceeding that for AGF347; for AGF347, its metabolism to polyglutamates increases enzyme inhibition as AGF347-Glu₅ is ∼20-fold more potent in inhibiting SHMT2 than the parent AGF347.¹⁹ AGF347 and AGF359 are multitargeted inhibitors that also target the cytosolic C1 enzymes GARFTase and AICARTase in addition to SHMT2,^14,15 thus compounding the impact of reduced synthesis of mitochondrial formate on the de novo purine biosynthetic pathway.

For both classes of SHMT2 inhibitors, maximal suppression of tumor cell proliferation requires inhibition of cytosolic SHMT1 in addition to SHMT2,^14,20 as reversal of SHMT1 catalysis can compensate for the loss of SHMT2.²² However, this concept was challenged by the recent study by Lee et al²³ which strongly suggested that SHMT1 could be an important therapeutic target independent of SHMT2 for solid tumors and leukemias, as the cytosolic C1 flux through SHMT1 appeared to predominate over the C1 flux in mitochondria accompanying modest levels of folate transport by RFC.

In this report, we systematically tested this notion by evaluating key pharmacodynamic determinants of nonclassical and classical SHMT1/2-targeted inhibitors for cancer in novel engineered tumor models designed to isolate and test these components. These include the contributions of the major facilitative folate transporters RFC (ubiquitously expressed in tumors and normal tissues⁴) and PCFT (expressed in solid tumors with limited expression in normal tissues⁶) at levels approximating those in cultured tumor cells⁷ to the accumulation of folates and SHMT1/2-targeted antifolates, and the impact of FPGS on folate and pyrrolo[3,2-d]pyrimidine antifolate accumulations in the mitochondria and cytosol. We also tested the potential impact of different extracellular folate forms (LCV vs FA) and folate concentrations commonly used to study C1 metabolism in vitro on these pharmacodynamic parameters.

RFC and PCFT mediate the cellular uptake of folates from the extracellular medium and facilitate net accumulation of folates in the cytosol and indirectly in the mitochondria following mitochondrial folate transport,²⁵ although the net impact varies depending on whether a physiologic concentration of a reduced folate (LCV) is used or if nonphysiological FA is provided. FA is a synthetic folate form commonly found in tissue culture media and is a poor substrate for RFC⁴ and must be reduced by DHFR to enter the metabolically active pool of THFs. At a highly elevated concentration of FA (∼2 μM, typically found in culture medium), the accumulation of especially high intracellular concentrations of folates was accompanied by ∼60% decreased RFC protein and transcripts, although this was not reflected in the net levels of folate accumulation. Although PCFT is a comparatively poor transporter of folates at pH 7.2, during sustained exposures, folates accumulated to far higher levels in PCFT-expressing cells than in the absence of PCFT. This may in part reflect a modest decrease in the pH of the culture medium accompanying cell growth, resulting in a slight augmentation of PCFT transport. Most likely, this reflects the extensive metabolism of the modest levels of intracellular “monoglutamyl” folates (transported by both RFC and PCFT) to their polyglutamate forms in both the cytosol and mitochondria (following transport into the mitochondria³⁸).

We used targeted metabolomics with [2,3,3-²H]serine to directly evaluate the impact of increasing RFC transport (at 25 nM LCV) on the cytosolic C1 flux through SHMT1 versus the mitochondrial C1 flux through SHMT2. Our results unambiguously established that the mitochondrial C1 flux directly parallels changes in RFC transport and net folate accumulation in the cytosol and mitochondria, whereas the cytosolic C1 flux through SHMT1 was unchanged. Importantly, when PCFT was expressed, the mitochondrial C1 flux increased substantially and was independent of changes in RFC. An analogous pattern for the R1-11/Tet-On-RFC cells was seen with 200 nM FA but not with 2 μM FA.

We also examined the impact of increasing RFC levels and activity, combined with constitutive PCFT expression and/or elevated FPGS, on the sensitivity or resistance to classical (eg, AGF347) and nonclassical (eg, SHIN1) SHMT2 inhibitors. Paradoxically, for both classes of inhibitors, drug sensitivity was maximal at the lowest levels of RFC. Further, increasing RFC levels and accumulation of AGF347 were accompanied by decreased in vitro efficacies analogous to findings with the nonclassical SHMT2 inhibitors SHIN1 and SHIN2. Further, these effects were exacerbated in the presence of ectopically expressed PCFT. However, for the classical SHMT2 inhibitors AGF347 and AGF359 (but not for SHIN1 and SHIN2), inhibition was dramatically enhanced in the presence of elevated FPGS.

Our results establish several general principles that are important to the future clinical development of SHMT2-targeted antifolates for cancer:

• 1.For in vitro experiments, the type and concentration of extracellular folate profoundly impact the interpretation and in vivo relevance of studies with classical and nonclassical C1 inhibitors. The importance of this result is immense, as for decades the majority of metabolic studies on C1 metabolism and therapy have used tissue culture media with FA at micromolar concentrations largely irrelevant to the circulating folate levels in humans.³¹
• 2.RFC levels are important determinants of mitochondrial C1 flux via SHMT2 but are much less so for SHMT1 in the cytosol.
• 3.RFC and PCFT activities profoundly impact tumor inhibition by SHMT2 inhibitors by directly facilitating drug uptake (eg, AGF347) and by elevating levels of cytosolic and mitochondrial folates, which directly compete for binding to intracellular targets, decreasing target engagement. For classical inhibitors such as AGF347, which are excellent FPGS substrates,¹⁸ the adverse impact of elevated cellular folates can be effectively circumvented by the synthesis of drug polyglutamate forms in both the mitochondria and cytosol; however, this is not possible for nonclassical SHMT2 inhibitors such as SHIN1.

Collectively, our results further document the substantial therapeutic promise of classical multitargeted pyrrolo[3,2-d]pyrimidine antifolates typified by AGF347. These novel compounds offer an exciting new platform for C1-targeted drug development for cancer.14, 15, 16

Conflict of interest

Zhanjun Hou, Larry H. Matherly, Mathew Schneider, Jing Li, Seongho Kim, Charles E. Dann III, and Aleem Gangjee report financial support by National Institutes of Health National Cancer Institute. All other authors declare no conflicts of interest.