Elimination of human folypolyglutamate synthetase alters programming and plasticity of somatic cells

Abstract

Folates are vital cofactors for the regeneration of S-adenosyl methionine, which is the methyl source for DNA methylation, protein methylation, and other aspects of one-carbon (C1) metabolism. Thus, folates are critical for establishing and preserving epigenetic programming. Folypolyglutamate synthetase (FPGS) is known to play a crucial role in the maintenance of intracellular folate levels. Therefore, any modulation in FPGS is expected to alter DNA methylation and numerous other metabolic pathways. To explore the role of polyglutamylation of folate, we eliminated both isoforms of FPGS in human cells (293T), producing FPGS knockout (FPGS^ko) cells. The elimination of FPGS significantly decreased cell proliferation, with a major effect on oxidative phosphorylation and a lesser effect on glycolysis. We found a substantial reduction in global DNA methylation and noteworthy changes in gene expression related to C1 metabolism, cell division, DNA methylation, pluripotency, Glu metabolism, neurogenesis, and cardiogenesis. The expression levels of NANOG, octamer-binding transcription factor 4, and sex-determining region Y-box 2 levels were increased in the mutant, consistent with the transition to a stem cell–like state. Gene expression and metabolite data also indicate a major change in Glu and GABA metabolism. In the appropriate medium, FPGS^ko cells can differentiate to produce mainly cells with characteristics of either neural stem cells or cardiomyocytes.—Srivastava, A. C., Thompson, Y. G., Singhal, J., Stellern, J., Srivastava, A., Du, J., O’Connor, T. R., Riggs, A. D. Elimination of human folypolyglutamate synthetase alters programming and plasticity of somatic cells.

Tetrahydrofolate (THF) and its derivatives, collectively referred to as folates, constitute a group of cofactors critical for several major metabolic pathways in the cell. These cofactors participate in the addition and removal of one-carbon (C1) units in a set of reactions commonly referred to as C1 metabolism (1). The products of these C1 transfer reactions include purines, thymidylate, methionine, S-adenosyl methionine (SAM), and pantothenate (vitamin B5), all of which are crucial for normal cell function (2, 3). The partitioning of carbon units into various cellular outputs involves the following 4 major pathways: the folate cycle, the methionine cycle, the transsulfuration pathway, and the transmethylation metabolic pathways (4–6) (Fig. 1; BioRender, Toronto, ON, Canada). The transmethylation metabolic pathways closely interconnect choline, methionine, and methyl-THF (7) and play a crucial role in the production of SAM (8).

Figure 1.

Schematic representation of THF production and C1 metabolism and their distribution in different compartments of the mammalian cell [adapted from Tibbetts and Appling (1) and Ducker and Rabinowitz (94)]. Activated C1 units, monoglutamylated THFs, are transported from cytoplasm to mitochondria where they are polyglutamylated by FPGS, and polyglutamated folates are utilized in C1 metabolism by SHMT. MTHFD1 is trifunctional in the cytoplasm. MTHFD1L is monofunctional, and MTHFD2 or MTHFD2L are bifunctional in mitochondria. 10-formyl-THF dehydrogenase is functional in both compartments in mammals. All abbreviations are standard gene names. Certain descriptions utilize the common protein name for clarity. ALDH1L1, cytosolic 10-formyl-THF dehydrogenase; ALDH1L2, mitochondrial 10-formyl-THF dehydrogenase; ATIC, 5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase/IMP cyclohydrolase; DHFR, dihydrofolate reductase; dTMP, thymidine monophosphate; GART, phosphoribosylglycinamide formyltransferase; Hcy, homocysteine; MTFMT, mitochondrial methionyl-tRNA formyltransferase; MTHFD1, MTHFD, cyclohydrolase, and formyl-THF synthetase 1; MTHFD1L, monofunctional THF synthase (mitochondrial); MTHFD2L, MTHFD2-like; MTHFR, methylene THF reductase; MTR, methionine synthase; SHMT1, cytosolic SHMT; SHMT2, mitochondrial SHMT; THF-Glu₁, tetrahydrofolate monoglutamate; THF-Glu_n, tetrahydrofolate polyglutamate; TYMS, thymidylate synthetase.

Epigenetic marking by DNA and histone methylation depends on SAM (9). These epigenetic marks are established during mammalian development (10, 11), and are essential for the maintenance of cell identity (12, 13). A large body of evidence suggests that either deficiency or excess of folate can modulate DNA methylation in a cell-, gene-, and site-specific manner, and imbalanced methylation can reinforce a plethora of health conditions ranging from cardiovascular disease to depression (14–19). Loss-of-function mutations in enzymes that are involved in the folate cycle, methionine cycle, transsulfuration pathway, and the transmethylation metabolic pathways can lead to growth defects both in animals and in humans, underscoring the role of C1 metabolism in modulating cell growth (20, 21). Various studies indicate that a steady pool of folate cofactors is essential for actively dividing cells and is required for normal growth and development (22–24). Although much progress has been made toward understanding the biochemistry of enzymes involved in folate metabolism (25), genetic evidence for the biologic roles of these enzymes is still limited.

One of the key enzymes in folate metabolism is folypolyglutamate synthetase (FPGS), which catalyzes ATP-dependent sequential conjugation of Glu residues to folate, forming folypolyglutamates (Fig. 1). Polyglutamylation is essential for the retention of folates within cellular compartments because nonglutamylated or monoglutamylated folates can transport across the mitochondrial membrane in either direction (25). The polyglutamate chain lengths of the folates differ from 1 cell type to another and within different organelles of a given cell, but in most eukaryotic cells, the penta- and hexaglutamate forms predominate (1). In mammals, there is only 1 gene for FPGS, but there are 2 isoforms that are independently required in the cytosol and in the mitochondrial matrix (25, 26). Mitochondria receive folates from the cytoplasm only in a reduced, monoglutamylate form (Fig. 1), which is then polyglutamylated and charged with C1 units in situ (1). Folypolyglutamates cannot traverse mitochondrial membranes in either direction, so both mitochondrial and cytosolic isoforms of FPGS are required to maintain subcellular folate compartmentalization and function (25). Serine is oxidized in the mitochondria and is transferred to THF by serine hydroxylmethyltransferase (SHMT), resulting in glycine and 5,10-methylene-THF. A series of C1 reactions in mitochondria eventually produce formate, which flows to the cytoplasmic THF pool through the activity of mitochondrial methylene THF dehydrogenase (MTHFD) (27). The reductive incorporation of the formate into the cytosolic folate pool results in thymidine production (Fig. 1). Essentially, the cytosolic form of FPGS is required to synthesize purines and thymidine (28), and mitochondrial FPGS is required to produce glycine (29) in a mammalian cell.

In plants and cancer cells, mutation of FPGS changes the glutamylation status of the folates, and this alteration in polyglutamylated folates and associated compounds affects DNA methylation and releases chromatin silencing on a genome-wide scale (30–32). Polyglutamylated folates are better substrates for methylene THF reductase and methionine synthase, and both of these enzymes are involved in the generation of SAM (33, 34). In cancer cells, FPGS down-regulation by small interfering RNA reduces global DNA methylation and DNA methyltransferase (DNMT) activities (32, 35).

Although FPGS plays a central role in C1 metabolism, folate metabolism, and transmethylation pathways, it was unclear how an imbalance in these pathways caused by FPGS mutation would affect mammalian cell growth and differentiation. Therefore, we decided to characterize a null mutant of FPGS in a mammalian cell, eliminating cytoplasmic and mitochondrial isoforms, and 4 splicing variants (36, 37). We successfully created homozygous deletions of FPGS in the human embryonic kidney (HEK) 293T cell line using clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9). The FPGS knockout (FPGS^ko) cell lines are viable, displaying stem-cell markers in cell culture, but proliferate extremely slowly with a tendency toward cardiogenesis and neurogenesis.

MATERIALS AND METHODS

Cell lines and production of FPGS mutants

HEK cell line 293T was cultured in DMEM (Corning, Corning, NY, USA) supplemented with 10% fetal bovine serum (FBS; Thermo Fisher Scientific, Waltham, MA, USA) and 1% GlutaMax (Thermo Fisher Scientific). The cell lines were cultured at 37°C in a humidified 5% CO₂ incubator. Stable clonal cell lines were created by transfecting 293T cells with GeneArt CRISPR Nuclease Vector with orange fluorescent protein (OFP) reporter gene (Thermo Fisher Scientific) (38) (Supplemental Fig. S1). For transfection, cells were seeded into a 6-well plate and transfected at 70% confluence using XFect (Takara, Kyoto, Japan) according to the manufacturer’s protocol. Transfections were performed with 2 µg of a plasmid coexpressing Cas9, a chimeric single guide RNA (sgRNA), and OFP. At 24–36 h post-transfection, cells were refreshed with 2 ml of growth medium and collected at 72 h after transfection. Transfected positive clones were selected using single-cell sorting [BD FACSAria Cell Special Order Research Product Sorter (BD Biosciences, San Jose, CA, USA)], and cells were collected in a 96-well plate for single-cell growth. Single-cell colonies were expanded in DMEM supplemented with 10% FBS (stem-cell quality; U.S. origin; Thermo Fisher Scientific), Minimum Essential Medium Nonessential Amino Acid (NEAA) solution (Thermo Fisher Scientific), and 1% Glutamax (Thermo Fisher Scientific), and evaluated by sequencing of genomic DNA.

Plasmid and sgRNA design

The sgRNAs targeting the human FPGS gene were designed using Integrated DNA Technologies (Coralville, IA, USA) guide RNA (gRNA) design tools to minimize off-target, and the potency of these sgRNAs was also tested using the Basic Local Alignment Search Tool (BLAST; U.S. National Center for Biotechnology Information, Bethesda, MD, USA) analysis. The gRNAs were designed to target the conserved region of the FPGS and knockout function of both isoforms. The target sequences and plasmid construct map are shown in Fig. 2 and Supplemental Fig. S1. DNA oligos of sgRNAs were cloned using the GeneArt Seamless Cloning and Assembly Kit (Thermo Fisher Scientific) as per the manufacturer’s protocol.

Figure 2.

Generation of FPGS^ko 293T cell lines using CRISPR/Cas9. A) Schematic representation of the FPGS exons (E1–E15), with exon 4 indicated. PAM, protospacer-adjacent motif. B) Sequences of the targeted regions of WT and FPGS^ko-1 and FPGS^ko-2 cell lines. C) Western blots showing loss of FPGS in FPGS^ko lines (C1). Probing the membrane for β-actin showed that sample loading for all 3 samples was similar (C2). The arrows indicate the FPGS or actin positions, as well as that of a nonspecific band that appears.

DNA and RNA isolation and analysis

Genomic DNA from putative clones was extracted using DNAzol Reagent (Thermo Fisher Scientific) and a modified protocol of the previously published method in Ausubel et al. (39). For genotyping, PCR reactions were performed in duplicate with genomic DNA using High-Fidelity Taq DNA polymerase (Thermo Fisher Scientific) according to the manufacturer’s protocol. The target-specific primer sets used for PCR are listed in Supplemental Table S1A, B.

The total RNA from putative clones was extracted using the miRNeasy Mini Kit (Qiagen, Germantown, MD, USA). cDNA synthesis from 293T RNA was performed with 1 μg of total RNA using SuperScript III Reverse Transcriptase and the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific) according to the manufacturer’s protocol.

Cell energy phenotype analysis using the Seahorse XFe96 extracellular flux analyzer

Basal mitochondrial function and metabolic potential of FPGS^ko-1 and wild-type (WT) cells were measured using the Seahorse Bioscience XFe96 Cell Energy Phenotype Test (Agilent Technologies, Santa Clara, CA, USA). This assay simultaneously measures the 2 major energy-producing pathways in live cells (mitochondrial respiration and glycolysis), allowing a rapid determination of energy phenotypes of cells and investigating metabolic potential of the cell. The experiments were performed according to the manufacturer’s protocol. Briefly, cells were seeded in DMEM supplemented with 10% FBS in 96-well tissue culture plates at a density of 20,000 cells/well and allowed to adhere for 24 h. Prior to the assay, the medium was changed to DMEM containing 10 mM glucose, 1 mM pyruvate, and 2 mM Gln (pH 7.4), and the cells were equilibrated for 30 min at 37°C. The oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured under basal conditions. All treatment conditions were analyzed with 6–8 wells/treatment and repeated at least twice. OCR and ECAR values were normalized to cell numbers.

Clariom S Human Array and gene expression analysis

RNA was isolated using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. Total RNA was assessed for the RNA quality verification and microarray hybridization. The Agilent 2100 Bioanalyzer (Agilent Technologies), a microfluidics-based platform, was used for sizing, quantification, and quality of RNA. The RNA integrity number score was generated on the Agilent software. For the microarray analysis, the RNA quality for all of the samples had an RNA integrity number score >7.

For microarray analysis, 3 biologic replicates were included for both control and FPGS mutant. For each array experiment, 500 ng of total RNA was used for labeling using the Clariom S Human Array (Thermo Fisher Scientific). Probe labeling, chip hybridization, and scanning were performed according to the manufacturer’s instructions. A Probe Set (gene-exon) was considered expressed if ≥50% samples had detection above background (DABG) values below the DABG threshold (DABG < 0.05).

To validate microarray results, quantitative 2-step RT-PCR was performed. One microgram of total RNA was reverse transcribed to first-strand cDNA with the Qiagen cDNA Synthesis Kit (Qiagen), and this cDNA was subsequently used as a template for quantitative PCR with gene-specific primers. The ubiquitous β-actin gene served as a control for constitutive gene expression. The quantitative RT-PCR (qRT-PCR) reactions were performed using Power Sybr Green PCR Master Mix (Thermo Fisher Scientific) on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). Relative expression levels (2^−ΔCt) were calculated according to the Livak and Schmittgen method (40). Expression levels of each gene were compared with the expression level of actin. Values are the means of 3 biologic and 3 technical replicates and the oligonucleotides used in the study are presented in Supplemental Table S1.

Global DNA methylation measurement

Global 5-methylcytosine (5-mC) levels were quantified using the MethylFlash Methylated DNA Quantification Kits (Epigentek, Farmingdale, NY, USA). The DNA concentration was determined using the Qubit Assay (Thermo Fisher Scientific) according to the manufacturer’s protocol. Briefly, 100 ng of DNA was used for incubation with both capture and detection antibodies using MethylFlash Methylated DNA Quantification Kit (Colorimetric) from Epigentek. Subsequently, measurements of the absorbance of the sample at 450 nm in a microplate spectrophotometer (BioTek Instruments, Winooski, VT, USA) were performed with the percentage of the whole genome 5-mC calculation according to manufacturer’s instructions. Genomic methylation levels in study samples were expressed as percentage of 5-mC.

Western blotting

Cells were lysed with Mammalian Protein Extraction Reagent lysis buffer (78501; Thermo Fisher Scientific) containing 1 mM PMSF and 1× protease inhibitor cocktail (MilliporeSigma, Burlington, MA, USA). Proteins (40 µg/well) were separated on 4–12% gradient Bis-Tris NuPage gels (Thermo Fisher Scientific) and blotted on methanol-activated PVDF membrane (Thermo Fisher Scientific) using 1× transfer buffer (LC3625; Thermo Fisher Scientific) according to the manufacturer’s instructions. Subsequently, blocking was performed using Li-Cor Biosciences (Lincoln, NE, USA) Odyssey blocking buffer (PBS; 927-40000). Thereafter, blocked membranes were incubated with a specific anti-human FPGS antibody (1:1000; AB184564; Abcam, Cambridge, MA, USA) overnight at 4°C in the same blocking buffer. The membrane was washed 3 times (5 min each wash) with 10 ml PBS with 0.05% Tween and 1 time with 1× PBS on a shaker and incubated with anti-rabbit antibody dye 680RD (25-68071; Li-Cor Biosciences) for 90 min at room temperature. After incubation, the membrane was washed 3 times (5 min each wash) with 10 ml PBS with 0.05% Tween and 1 time with 1× PBS on a shaker. Immunocomplexes were visualized with the Li-Cor Odyssey CLx in 700 channel (red). To visualize β-actin on the membrane, an anti–β-actin monoclonal antibody (012M4821; A1978; MilliporeSigma) and secondary antibody IR-Dye 800 (926-32210; Li-Cor Odyssey; Li-Cor Biosciences) were used.

Immunofluorescence analysis

The relative optimal image-based analysis of 2 key human pluripotent stem cells markers [octamer-binding transcription factor 4 (OCT4) and stage-specific embryonic antigen 4 (SSEA4)] was performed using the Pluripotent Stem Cell Immunocytochemistry Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Briefly, the cells were grown in 10% FBS DMEM supplemented with NEAAs and Glutamax in wells coated with 0.1% gelatin. For immunofluorescence localization, the cells were stained for SSEA4 and Oct4 and counterstained with DAPI using the kit (Thermo Fisher Scientific). The endogenous proteins were labeled using primary antibodies (anti-SSEA4 anti-mouse IgG3 and anti-OCT4 host-rabbit) followed by secondary antibodies conjugated to Alexa Fluor 488 goat anti-mouse IgG3 and Alexa Fluor 594 donkey anti-rabbit.

Chemical complementation assays of FPGS mutants

Cells were maintained as monolayers in DMEM with GlutaMax (Corning) supplemented with 10% FBS at 37°C in a 5% CO₂ atmosphere. To supplement the growth medium with amino acids, we added 1× and 2× doses of essential amino acids (Thermo Fisher Scientific) in the medium and grew the cells as previously described for 5 d. In addition to this, we used Iscove’s modified Dulbecco’s medium (IMDM) and DMEM with 1× NEAAs (Thermo Fisher Scientific) along with FBS and Glutamax as described earlier. For 5-formyl-THF (5-CHO-THF) supplementation experiments, 6S-5-formyl-5,6,7,8-tetrahydrofolic acid (calcium salt; natural calcium folinate) was purchased from Schircks Laboratories (Jona, Switzerland). The stock solution (10 mM) of 5-CHO-THF was made using tissue culture–grade Dulbecco’s PBS buffer (Thermo Fisher Scientific), and 100 µl from the stock solution was applied to the 1 ml IMDM to achieve the desired (1 mM) working concentration. The cells were grown for 5 d in the medium supplemented with 5-CHO-THF, and comparative analysis for cell growth was performed with the cells growing with only solvent control (only Dulbecco’s PBS without 5-CHO-THF) in DMEM. All the experiments were carried out in triplicate, and cell proliferation was measured using a Cellometer Auto T4 Counting Chamber (Nexcelom Bioscience, Lawrence, MA, USA).

Genetic complementation of FPGS mutants

To confirm that FPGS^ko phenotype was caused by the loss of FPGS function, we cloned the WT-FPGS coding region from 293T and performed genetic complementation assays on FPGS^ko-1 and FPGS^ko-2 mutants. Cells were grown (293T) in DMEM supplemented with 10% FBS, total RNAs were isolated, and first-strand cDNA was reverse transcribed as previously described. The FPGS coding region was amplified using the Phusion High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, MA, USA) and gene-specific primers (Supplemental Table S1). The resulting amplified product was cloned into pENTR-D-TOPO according to the manufacturer’s protocol (Thermo Fisher Scientific), and the fragment was subsequently cloned into the pDest47 vector using gateway cloning according to the manufacturer’s instructions (Thermo Fisher Scientific). The resulting expression plasmid (pDest47-FPGS-GFP) containing a functional fusion FPGS [FPGS fused to green fluorescent protein (GFP)] was delivered to FPGS^ko-1 and FPGS^ko-2 mutants using Lipofectamine 2000 Transfection reagent (Thermo Fisher Scientific) according to the manufacturer’s protocol. Transfected positive clones (transiently expressing FPGS-GFP) were selected 3 d post-transfection using single-cell sorting as we did for selection of the FPGS^ko mutants. Expression of the transgenic FPGS was confirmed by qRT-PCR.

Metabolomics by hydrophilic interaction liquid chromatography and GC-MS

The cell lines (FPGS^ko and control) were cultured as described earlier in 100-mm plates. Once they attained confluency, the cells were washed twice with PBS (MilliporeSigma) and detached with 0.25% Trypsin-EDTA at 37°C for 2 min. Subsequently, the cells were collected in 15-ml Falcon tubes, pelleted, and washed 5 times with ice-cold PBS before they were counted in a single-cell suspension using a Cellometer (Nexcelom Bioscience). The experiment was conducted with 5 biologic and 5 technical replicates, and a total of 8 million cells were used in each replicate. Samples were analyzed by the West Coast Metabolomics Center at the University of California–Davis (Davis, CA, USA) for primary metabolites (GC-MS) and biogenic amines [hydrophilic interaction liquid chromatography (HILIC)] using standard operating procedures as described earlier (41, 42).

Glu assay

The relative free Glu concentration in FPGS^ko mutants was assessed using a Glutamate Assay Kit (Abcam) according to the manufacturer’s instructions. Using the kit, we measured free Glu levels in the mutant and compared with the control cells (293T). The amount of Glu was quantified by colorimetric analysis (spectrophotometry at optical density = 450 nm) using a Tecan Microplate Reader (Männedorf, Switzerland).

Cardiac and neural differentiation cell-culture methods

To induce differentiation, FPGS^ko cells were grown in human basal differentiation medium containing 10% FBS (stem-cell quality; Thermo Fisher Scientific), 1% NEAAs (Thermo Fisher Scientific), 1% penicillin-streptomycin (Thermo Fisher Scientific), 0.05 mM 2-ME (MilliporeSigma), and 2 mM l-GlutaMax in IMDM (Thermo Fisher Scientific) in 6-well ultra-low attachment plates until they reach 40–50% confluency. Subsequently, cells with embryoid body (EB)–like morphology were plated onto 0.1% gelatin-coated 12-well plates and grown in Roswell Park Memorial Institute (RPMI) 1640 medium with GlutaMax (Thermo Fisher Scientific) and serum-free B27 supplement (Thermo Fisher Scientific) differentiation factors and incubated from cardiac lineage to functional cardiomyocytes.

For neural differentiation, FPGS^ko cells were initially grown onto 0.1% gelatin-coated 12-well plates in DMEM supplemented with 10% FBS (stem-cell quality; Thermo Fisher Scientific), 1% NEAA (Thermo Fisher Scientific), 1% penicillin-streptomycin (Thermo Fisher Scientific), and 1% GlutaMax (Thermo Fisher Scientific). To induce neural differentiation, the cells were dissociated and passaged onto laminin (20 μg/ml; MilliporeSigma)-coated plates and grown in neurobasal medium (Thermo Fisher Scientific) or RPMI 1640 medium (Thermo Fisher Scientific) for 10–15 d at 37°C and 5% CO₂. The medium was supplemented with 1% serum-free B-27 (Thermo Fisher Scientific), 1% GlutaMax (Thermo Fisher Scientific), 1% NEAA (Thermo Fisher Scientific), and 1% penicillin-streptomycin (Thermo Fisher Scientific).

Cell quantification, imaging, and statistical analysis

Cells were quantified using a Cellometer Auto T4 counting chamber after mixing in a 1:1 ratio with Trypan blue (MilliporeSigma) to exclude dead cells. The average of 2 separate counts was taken to calculate the cell numbers. All live cell imaging was conducted using an EVOS FL Auto microscope (Thermo Fisher Scientific). All statistical analysis was performed using a paired Student’s t test with Bonferroni correction or 1-sample Student’s t test. Values of P < 0.05 were considered significant.

RESULTS

Generation of FPGS depleted (FPGS^ko) human cells

In mammalian cells, the single gene for FPGS undergoes alternative splicing, resulting in 2 different isoforms, with the mitochondrial and cytosolic isoforms differing only in the N-terminal domain. To investigate the role of FPGS, we generated deletions in both isoforms of FPGS in 293T cells (HEK cells) using CRISPR/Cas9. The gRNAs we used targeted conserved exons present in both isoforms. The sequence and targeting strategy is shown in Fig. 2A, B and Supplemental Fig. S1. The targeting plasmid expressed OFP and fluorescence-activated cell sorting was used to isolate single-transfected cells, 15 of which slowly grew into small colonies over a period of ∼60 d. PCR products of the targeted FPGS region from each putative clone were analyzed by sequencing. Two clones, FPGS^ko-1 and FPGS^ko-2, harbored 5- and 7-bp deletions, respectively, in the targeted region (Fig. 2B). The deletions were homozygous because all analyzed sequences showed the same deletion. The deletions in FPGS^ko-1 and FPGS^ko-2 cells both induce frame shifts, which result in premature stop codons. Western blots established that little or no FPGS protein was made in either mutant cell line (Fig. 2C); the mRNA was also rendered unstable because transcript levels were reduced over 10-fold (Supplemental Fig. S1B).

To confirm that the changes seen were not caused by off-target effects, we performed complementation studies. FPGS cDNA was prepared from WT (293T), cloned into an expression plasmid vector, forming pDest47-FPGS-GFP, which was used for transfection and for functional complementation. The hFPGS expression vector was transformed into FPGS^ko-1 and FPGS^ko-2, GFP-positive clones were selected using single-cell sorting, and expression of FPGS was verified using qRT-PCR. All clones complemented with FPGS showed relatively high levels of FPGS expression with cell growth and differentiation similar to that of WT 293T (Fig. 3 and Supplemental Fig. S2B). These findings indicate that the phenotypic alterations described in this paper arose from the disruption of FPGS. This conclusion is confirmed by medium supplementation results.

Figure 3.

FPGS deletion decreases cell proliferation and changes cellular morphology. A) Adherent cell growth of FPGS, FPGS^ko-1, and FPGS^ko-2 was assessed by cell counting at the indicated times. A total of 4350 cells were seeded in 12-well plates, and cells were counted after 8 d. The number of cells for each FPGS^ko cell lines (FPGS^ko-1 and FPGS^ko-2) were compared with the parental 293T cells. B) Genetic complementation of FPGS^ko-1 and FPGS^ko-2 mutants. Transfection of an FPGS expression plasmid into FPGS^ko-1 and FPGS^ko-2 rescues the phenotype (Complemented FPGS^ko Comp-FPGS^ko-1 and Comp-FPGS^ko-2). Error bars indicate means ± se (n = 5). ***P < 0.0001.

FPGS^ko cells show decreased cell proliferation

HEK cells (293T) exhibit rapid cell growth (43) in 10% FBS DMEM. However, of 15 putative clones for which we monitored growth (unpublished results), all clones exhibited extremely slow growth and a distinctive morphology that was different from the control cells (Fig. 3). Repeated passaging of the cells in DMEM supplemented with 10% FBS and NEAAs did not change the growth characteristics or morphologic features.

FPGS^ko mutants show reduced metabolic potential

To help understand the extremely slow growth of the mutants, we examined cellular metabolism by monitoring the OCR and ECAR of mutant and WT cells using the Seahorse XFe96. Both the mitochondrial respiration (OCR) and glycolytic activity (ECAR) of the FPGS^ko-1 cells were significantly decreased in comparison with control cells (Fig. 4), with a significant decrease of basal respiration, ATP production, and glycolytic capacity. These results indicate that the FPGS^ko-1 cells are in a quiescent-like state (Fig. 4), with oxidative phosphorylation being more affected than glycolysis.

Figure 4.

Quiescent energy phenotype of the FPGS^ko-1. An Agilent Seahorse XF was used to determine OCR and ECAR of FPGS^ko-1 and 293T. *P < 0.05, **P < 0.001 (Student’s t test)

Metabolomic profiling of the FPGS^ko indicates a decreased methylation capacity and altered amino acid and nucleic acid profiles

HILIC- and GC-MS–based metabolomic approaches were carried out to understand the differential pattern of C1 and other primary metabolites in FPGS^ko cells as compared with WT cells. A total of 1488 (HILIC) and 503 (GC-MS) compounds were detected in FPGS^ko and 293T cells, of which 131 biogenic amines (by HILIC) and 165 primary metabolites (by GC-MS) could be assigned chemical structures and quantified based on spectral matching to authentic compounds. Considering the main focus of the study, we restricted our analysis to some key C1 compounds and primary metabolites. Compared with the parental cell, we found that the metabolites belonging to the C1 pathway were changed significantly in FPGS^ko cells. The ratio of SAM to S-adenosylhomocysteine (SAH) provides an indication of the cellular capacity to catalyze transmethylation reactions. We found that FPGS^ko cells had a significant accumulation of SAH along with a reduction in SAM. The resulting low SAM/SAH ratio in FPGS^ko suggests a marked difference in the methylation capacity of the mutant (Fig. 5A).

Figure 5.

Quantitative estimation of SAM, SAH, Gln, Glu, and GABA by HILIC and metabolomic profiling by GC-MS. Quantitative estimation of SAM and SAH (A) was determined using HILIC. Metabolic profiling of amino acids (B) depicting fold change (logarithmic values) in the FPGS^ko cell line compared with the parental 293T cells, which was confirmed by the quantitative estimation of Gln, Glu, and GABA by HILIC (C). Metabolic profiling of nucleic acids (D) depicting fold change (logarithmic values) in the FPGS^ko cell line compared with the parental 293T cells. AMOT, angiomotin; CDR1, cerebellar degeneration related protein 1; CHAC1, γ-glutamylcyclotransferase 1; CNPY1, canopy FGF signaling regulator 1; CSMD3, CUB and Sushi multiple domains 3; DDR2, discoidin domain-containing receptor 2; DPYD, dihydropyrimidine dehydrogenase; DUSP6, dual specificity phosphatase 6; ETV5, ETS variant 5; GABRA3, GABA type A receptor α3 subunit; GABRB2, GABA type A receptor β2 subunit; GABRB3, GABA type A receptor β3 subunit; GDPD3, glycerophosphodiester phosphodiesterase domain containing 3; HSPB8, heat shock protein family B member 8; IRS4, insulin receptor substrate 4; KRTAP21-2, keratin-associated protein 21-2; LCP1, lymphocyte cytosolic protein 1; MAP3K12, MAPKK kinase 12; NEFM, neurofilament medium; PSAT1, phosphoserine aminotransferase 1; RHEBL1, Ras homolog enriched in brain like 1; SERPINF1, serpin family F member 1; SLC6A9, solute carrier family 6 member 9; TXNIP, thoredoxin-interacting protein. Error bars represent the se for 5 independent experiments and 5 technical replicates. *P < 0.05, **P < 0.01, ***P < 0.001 (Student’s t test).

Our analysis also showed that some nucleotides and amino acids were depleted in FPGS^ko (Fig. 5). Amino acids that showed a statistically significant reduction in FPGS^ko compared with the WT included IIe, Gly, Ser, Hse, Met, Pro, and Asn (Fig. 5B). However, Gln, Lys, His, Val, and Glu were significantly increased in FPGS^ko compared with 293T cells. (Fig. 5B). Quantitative analysis of Gln, Glu, and GABA further confirm a significant accumulation of these metabolites in FPGS^ko cells (Fig. 5C). In addition to this, we noticed statistically significant reductions in the levels of AMP, uridine, adenosine, guanine, thymine, adenine, and methylthioadenosine in the mutant (Fig. 5D).

Microarray analysis identified 2315 differential genes in FPGS^ko cells

We next conducted a transcriptional analysis of FPGS^ko using Clariom S Human Arrays. The cells were grown in DMEM (Corning) supplemented with 10% FBS (stem-cell quality; U.S. origin; Thermo Fisher Scientific) and 1% GlutaMax. These data revealed that 2315 genes were at least 2-fold differentially expressed between the FPGS^ko mutants and control cells. Among the differentially expressed genes, 1163 had higher expression in the FPGS^ko mutant, whereas 1153 had a lower expression (Fig. 6 and Supplemental Table S2). Major changes in expression were noticed for C1 metabolism, DNA methylation, cell cycle, cellular assembly and organization, Glu metabolism, developmental disorders, hereditary and neurologic disorders, DNA replication, and DNA repair genes.

Figure 6.

Differentially expressed genes of FPGS^ko cell line. A). Scatter plot transcription signals determined by microarray analysis of FPGS^ko-1 and 293T. B). Summary of the number of differentially expressed genes. C). Heat map of the most differentially expressed genes. See Supplemental Table S2 for a complete list. Avg, average.

Close examination of C1 metabolism-related genes showed that around 14 genes were significantly down-regulated in the mutant, including thioredoxin-interacting protein (NM_006472), IGF-binding protein 2 (NM_000597), and cystathionine-β-synthase (NM_001178008). Among the genes directly involved in the folate biosynthesis pathway, expression of aldehyde dehydrogenase 1 family, member L2 (ALDH1L2; NM_001034173), MTHFD (NADP + dependent) 2 (NM_006636), and MTHFD (NADP + dependent) 1-like (NM_001242767) were significantly down-regulated in the mutant. Microarray data further validated that the FPGS (NM_001018078) was down-regulated in the FPGS^ko. In addition to this, expression of 36 genes associated with the methylation process and 26 genes related to DNA repair were affected (Supplemental Table S3A–G).

These results were not unexpected based on the direct connection of FPGS to these pathways and process, but we noticed some unexpected results in the up-regulated and down-regulated genes. Looking at the top 15 up-regulated genes in the transcript profiling of FPGS^ko, we found anosmin-1 (ANOS1) to be 311-fold higher in the mutant, followed by GABA A receptor, β-2 (179-fold), ankyrin repeat domain 1 (ANKRD1; 93-fold), E26 transformation-specific (ETS) variant 5 (53-fold), heat shock 22kDa protein 8 (27-fold), and expression of Dickkopf WNT-signaling pathway inhibitor 1 (DKK1) was 13-fold higher in the mutant as compared with the WT cells (Fig. 6 and Supplemental Table S2). ANOS1 is a glycoprotein expressed in the brain and spinal cord (44). The GABA receptor is a multisubunit chloride channel that mediates the fastest inhibitory synaptic transmission in the CNS (45). Cardiac adriamycin-responsive protein or ANKRD1 is a cardiac ankyrin repeat protein that is highly expressed in cardiac and skeletal muscle (46). Interestingly, all these genes are associated with normal development and are associated with expansion and differentiation of neurons or cardiomyocytes.

Most of the genes down-regulated in FPGS^ko were associated with regulation of cell growth, differentiation, and metabolism. Some genes that manifested notably reduced expression in FPGS^ko as compared with the WT cells included gametocyte-specific factor 1 (GTSF1; 164-fold), solute carrier family 7 (SLC7; 108-fold), serpin peptidase inhibitor (48-fold), discoidin domain-containing receptor 2 (31-fold), insulin receptor substrate 4 (27-fold), and angiomotin (17-fold). Unbiased clustering analysis of the data suggests that FPGS elimination altered the expression of the genes related to cell differentiation, amino acid transport, angiogenesis, C1 metabolism, neurogenesis, and oxidative stress (Fig. 6 and Supplemental Tables S2 and S3A–G).

We validated the microarray results for selected genes by real-time quantitative PCR experiments using FPGS mutant and control cells. The transcript levels of GTSF1, SLC7A11, ALDH1L2, and MTHF were significantly repressed in the FPGS mutant, consistent with the microarray results (Fig. 7). Similarly, and consistent with microarray data, expression of ANOS1, GABA receptor subunit β-2, ANKRD1, and DKK1 was significantly higher (Fig. 7) in the mutant when compared with 293T cells.

Figure 7.

Validation of microarray gene expression data by qRT-PCR. Relative expression levels of GTSF1, SLC7A11, ALDH1L2, MTHFD2, ANOS1, GABA receptor subunit β-2 (GABRB2), ANKRD1, and DKK1 genes in FPGS^ko (mutant) and FPGS (control-293T) cells were checked to validate microarray expression data. Error bars represent the se for 3 independent experiments and 3 technical replicates. *P < 0.05, Student’s t test. *P < 0.05, **P < 0.001 (Student’s t test).

Global DNA methylation is reduced and FPGS^ko cells express pluripotent stem-cell markers

A global decrease in methylated DNA content has previously been observed after treatment with antifolates (15, 18, 19, 47, 48). Therefore, we measured the global level of 5-mC in FPGS^ko cells and found a significant reduction in FPGS^ko cells (Fig. 8A).

Figure 8.

A) FPGS mutants showed a significant reduction in global DNA methylation. 5-mC content in the FPGS^ko cell line was measured and compared with the parental 293T cells. DNA was extracted and equal amounts of genomic DNA (100 ng) were analyzed with 5-mC ELISA. Statistical analysis was performed using Student’s t test. *P < 0.05 indicates significantly lower levels of DNA methylation in the mutant in comparison with controls. Error bars represent the se for at least 3 independent experiments and 3 technical replicates. B, C) Relative expression levels of Oct4 and Sox2 genes (B) and immunofluorescence localization of SSEA4 and Oct4 in FPGS^ko and control (293T) cells to evaluate pluripotency biomarkers (C). The FPGS^ko exhibits different cell morphology compared with WT cells (C). The FPGS^ko illustrates the expression of SSEA4 surface antigens but WT had no signal (C). Immunoreactivity for the OCT4 transcription factor in the mutant was found in the nucleus and cytoplasm; however, no signals were detected in WT (C). DAPI staining in FPGS^ko and WT were used to validate live cells. Scale bar, 200 µm. Error bars represent the se for 3 independent experiments and 3 technical replicates. *P < 0.05, **P < 0.001 (Student’s t test).

Though slow growing, the morphologic features of FPGS^ko cells were similar to those of stem cells. We therefore examined markers for stem cells in FPGS^ko-1 and FPGS^ko-2 cells grown on 0.1% gelatin-coated plates in DMEM supplemented with 10% FBS, Gln, and NEAAs. We found that the expression of key pluripotency marker genes OCT4 and sex-determining region Y-box 2 (SOX2) were significantly higher in FPGS^ko clones when compared with parental lines (Fig. 8B). These findings were consistent with microarray data, and expression of around 25 pluripotency marker genes was significantly higher, including Sox2, Oct4, and Kruppel-like factor 4 (Supplemental Table S3B) in the mutant when compared with 293T cells. To authenticate these findings, we examined 2 key pluripotency markers (OCT4 and SSEA4) using immunochemical staining, and this revealed high levels of OCT4 and SSEA4 in FPGS^ko lines (Fig. 8C) as compared with control cells (Fig. 8C). Interestingly, a distinct staining pattern by OCT4 antibodies was observed in FPGS^ko lines, with staining observed in both cytosol and nucleus (Fig. 8B). Similarly, strong SSEA4 expression was observed in the mutant (Fig. 8B, C). Together, these data indicate that defects in FPGS gene function cause somatic cells to lose cell identity and start expressing pluripotency genes.

FPGS^ko cells can manifest hallmarks of cardiogenesis and neurogenesis

The 93-fold up-regulation of ANKRD1 [a protein that is highly expressed in stressed cardiac muscle (46)], 13-fold up-regulation of DKK1 [important in regulation of heart development, cardiac repair, and heart disease (49)], and 51-fold down-regulation of thioredoxin-interacting protein [controls cardiac hypertrophy through regulation of thioredoxin activity (50)], was suggestive of ailing cardiac progenitor cells (CPCs). To determine if FPGS^ko cells could form cardiomyocyte-like cells, we adapted an earlier reported cardiac differentiation protocol (51). Contractile EBs were noted at d 10 (Supplemental Video S1). The contractile EBs in all groups peaked around d 14 and decreased after 17 d. The expression of cardiac-specific genes was also assessed in the FPGS^ko cells by using RT-PCR. Expression of 3 key cardiomyocyte markers [Nkx2 (early cardiac transcriptional factor indicative of cardiac progenitor phenotype), myosin regulatory light chain 2 (a distinctly expressed protein in cardiac muscle), and cardiac troponin T (a muscle contractility regulatory protein indicative of a mature cardiac phenotype)]. All 3 were significantly up-regulated in the FPGS^ko cells grown in basal 10% FBS and DMEM (Supplemental Fig. S3).

Neurogenesis was seen when FPGS^ko cells were maintained on laminin-coated plates and grown in neurobasal medium or RPMI 1640 basal medium with Glutamax-I and serum-free B27 supplement differentiation factors (Supplemental Fig. S4). In a separate experiment, mutants formed neurons even if the cells were maintained on DMEM and not maintained on a specific differentiation medium. However, a consistent proportion and population of neurons were noted when they were grown in differentiation medium. Close microscopic examinations displayed many bipolar neurons in the neural population (Supplemental Fig. S4).

The neurotransmitters GABA and Glu are known to have a major role in survival, proliferation, and integration of newly formed neurons (52–54), and Glu can act as a positive regulator of neurogenesis (55). Gln also regulates CPC metabolism and proliferation in mammalian systems (56). Our gene expression data pertaining to Gln-Glu-GABA metabolism were consistent with this notion (Supplemental Table S3A–G), so we determined free Glu concentration and expression of Glu-ammonia ligase in the mutant. The expression of Glu-ammonia ligase was significantly low in the mutant, and free Glu concentration was significantly high in the FPGS^ko mutant (Supplemental Fig. S5).

Nutrient supplementation rescues the FPGS^ko slow-growth phenotype

We tested if supplementing the growth medium with various small molecules can rescue the FPGS^ko slow proliferation phenotype. First, we spiked the cell-growth medium with additional essential amino acids (1×, 2×). Both FPGS mutants supplied with the additional essential amino acids showed increased cell proliferation (Fig. 9, A2-A3). Higher doses of amino acids (4×) in the medium were toxic (unpublished results). Because IMDM is recommended for embryonic stem-cell growth (57), we grew FPGS^ko cells in IMDM with 1× essential amino acids and NEAAs and noticed an improvement in cell proliferation (Fig. 9-A4). 5-CHO-THF and several amino acids are critical to the function of FPGS (58), so we grew the FPGS^ko mutant in IMDM supplemented with 1 mM 5-CHO-THF and 1× NEAAs and observed substantial increases in cell growth (Fig. 9-A5). Finally, we cultured the FPGS^ko mutant in IMDM supplemented with 1 mM 5-CHO-THF, 1× NEAAs, hypoxanthine (10 mM), and thymidine (1.6 mM; Thermo Fisher Scientific). This modification showed almost full restoration of growth rate (Fig. 9-A6).

Figure 9.

Supplementation with thymidine, 5-CHO-THF, and amino acids complements the phenotype of FPGS^ko. The growth medium was supplemented with 1 and 2× essential amino acids (EAAs), and cells were grown for 7 d (A2-A3). Additionally, to check the growth behavior of the mutant in a different basal medium, regular DMEM and IMDM with 1× NEAAs were tested. To rescue the phenotype of FPGS^ko, 5-CHO-THF (1 mM) with sodium hypoxanthine (10 mM) and thymidine (1.6 mM) mixture (HT; Thermo Fisher Scientific) was exogenously applied to IMDM, respectively (A5, A6), and compared with the FPGS^ko cells (A1) and WT cells (A6) grown with only solvent as a control. All modifications showed improved growth of FPGS^ko-1, ranging from partial (A2–A5) to full (A6) complementation. All the experiments were carried out in triplicate and cell proliferation was measured using a Cellometer. The number of cells for each cell line is compared with the number of colonies for parental 293T cells. Error bars indicate means ± se (n = 4). *P < 0.05, **P < 0.001.

DISCUSSION

FPGS is a critical enzyme not only because it is required for intracellular folate homeostasis (30) but also because it links to the transmethylation pathway (Figs. 1 and 10). The importance of FPGS for C1 metabolism in bacteria, yeast, and plant and mammalian cells (30, 59–62) is well-established, but most studies in mammalian systems have been on cancer cells with the ultimate aim being cancer therapeutics (63–68). For example, Kim et al. (32) used RNA interference to knock down FPGS activity in breast cancer cells (32, 35) and found that FPGS modulation altered global DNA methylation and expression of several genes involved in important biologic pathways. Considering that the human FPGS gene produces 2 proteins by alternative translational initiation of exon 1 [Freemantle et al. (28)], and it has 4 splicing variants (36, 37), suppression of a specific isoform of FPGS is not enough to fully illustrate implications of FPGS disorder in a human cell. There is only 1 report of an FPGS-null mutant in mammalian cells (69); this investigation focused on formaldehyde toxicity and the diversion of endogenous formaldehyde into C1 metabolism, reporting that FPGS-null cells were not able to grow in unsupplemented growth medium. The role of FPGS in energy metabolism and cellular plasticity was not investigated. In this study, we have targeted a conserved region of exon 4 of FPGS, which not only eliminates both isoforms of FPGS but also eliminates all splicing variants of FPGS. Additionally, around 34 single-nucleotide polymorphisms (SNPs) have been verified in FPGS that have altered the FPGS protein sequence (70). Among the 12 SNPs located in exon regions, none are known to be located in exon 4 (70). Here, we report that cells without FPGS are viable, although very slow growing in standard medium supplemented with 10% FBS, and undergo apparent reprogramming to a metabolic and transcriptional state with considerable resemblance to stem cells.

Figure 10.

Schematic model to illustrate role of FPGS and connected pathways in DNA methylation and pluripotency. Cells with functional FPGS (WT) rely on appropriate production and assimilation of Gln-Glu-GABA in a cyclic manner through the tricarboxylic acid (TCA) cycle (left); nonfunctional FPGS (FPGS^ko) perturbs the Gln-Glu-GABA equilibrium and promotes cardio- and neurogenesis utilizing excess GABA in the system (right). Black arrows represent normal enzymatic reactions in the cycle; red arrows indicate possible consequences caused by FPGS deletion. Hcy, homocysteine; Met, methionine.

Somatic-cell reprogramming into stem cells using 4 transcription factors, Oct4, Sox2, Kruppel-like factor 4, and c-Myc or OCT4, SOX2, NANOG, and LIN28 is well-established (71–73), and suppression of the maintenance DNMT1 or treatment with the DNMT1 inhibitor 5-azacytidine can aid this conversion (74, 75). Folate in its various forms is essential for the conversion of homocysteine to SAM, which is the source of methyl groups for both DNA methylation and histone methylation. Folate deficiency and mutations in folate-dependent pathways are well known to affect mammalian development, even sometimes causing transgenerational effects, probably by affecting epigenetic inheritance (76). As an interesting example, a hypomorphic mutation in the mouse 5-methyl-THF-homocysteine methyltransferase reductase gene, which is required for activation of methionine synthase and thus the formation of SAM, results in congenital malformations that can persist through 5 generations (76). It is clear that a homeostatic balance among C1 metabolism, the methionine cycle, and the transmethylation metabolic pathways is required for normal cell function and development. Elimination of FPGS is expected to affect folate retention and function in both mitochondria and cytoplasm, and this is likely to have profound effects on C1 metabolism. Thus, it is not surprising that metabolism of FPGS^ko cells is greatly altered. A second finding is that FPGS^ko cells have features of stem cells.

Energy metabolism of stem cells is predominantly aerobic glycolysis (77). We find that the energy metabolism of FPGS^ko cells is also predominately glycolysis, though at a reduced level (Fig. 4). DNA methylation is reduced (Fig. 8) in FPGS^ko cells, which is consistent with an increased SAH/SAM ratio (Fig. 5), because SAH inhibits transmethylation reactions. This result is consistent with previous reports that perturbing folate and C1 metabolism affects global DNA methylation (15, 17, 32, 48, 76, 78, 79). Perhaps as a result of decreased DNA methylation, several key pluripotency genes such as OCT4, SOX2, and SSEA4 are expressed in FPGS^ko cells (Fig. 8B, C). We also find that FPGS^ko cells will differentiate to either neuron-like cells or cardiomyocyte-like cells, depending on growth medium and conditions. Perhaps this is why our transcriptomic analysis of FPGS^ko cells showed, relative to parental 293T cells, greatly increased transcription of several neuronal and cardiomyocyte-specific genes. For example, we found that expression of ANOS1, GABA A receptor β-2, ANKRD1, and DKK1 was significantly higher in the mutant. ANOS1 and GABA play an important role in the CNS (45). Similarly, cardiac adriamycin-responsive protein or ANKRD1 is a rescue protein for cardiac muscle under stress conditions (46). At least some changes we observed may be generally linked to FPGS reduction and not only to 293T being the parental cells, as a significant change in the expression of ANKRD1, DKK1, and SOX2 caused by FPGS modulation was also noticed by Kim et al. (32), who reduced FPGS levels in HCT116 colon and MDA-MB-435 breast cancer cells by small interfering RNA treatment.

It is not clear why differentiation is preferentially toward neurons or cardiomyocytes, but 1 possibility is a change in Gln and Glu metabolism. In FPGS^ko cells, we see increases in Gln (5-fold), Glu (1.7-fold), and GABA (5-fold). Glu is the key excitatory and GABA is the main inhibitory neurotransmitter in mammals (80, 81). Our transcriptional analysis of FPGS^ko cells showed that the expression of 10 genes pertaining to Glu metabolism was significantly low. This includes glutathione-specific γ-glutamylcyclotransferase 1 (21-fold), asparagine synthetase (10-fold), and several neuronal and Glu transporters. In addition to this, expression of around 8 genes related to GABA receptors were significantly altered (Supplemental Table S2). Because Gln, Glu, and GABA are of special significance for neurons, we checked the expression of neuron-related genes and found that expression of 20 genes connected to the brain and neurons were affected. Ras homolog enriched in brain-like 1, which was 15-fold higher in the FPGS mutant, has been associated with the neuronal development and hippocampal neurogenesis (82). In addition to this, we also noticed 5-fold higher expression of brain acid-soluble protein (83) and 3-fold higher brain-derived neurotrophic factor, which can stimulate neurogenesis in the cell culture (84).

Why do FPGS^ko-derived pluripotent cells preferentially differentiate into cardiomyocytes? There is a possible involvement of Gln, GABA, and C1 metabolism in CPC proliferation (56, 85–87). The GABA A receptor, which is abundant in the heart and brain, plays a significant role in cardiovascular regulation (88). GABA B receptors are also expressed and functional in mammalian cardiomyocytes (85). Interestingly, Salabei et al. (56) have shown that Gln is a primary regulator of CPC growth, differentiation, and survival.

In FPGS^ko cells, we observed a distinct immunostaining of OCT4 in both cytosol and the nucleus (Fig. 8C). Riekstina et al. (89) found that heart mesenchymal stem cells express OCT4, NANOG, SOX2, and SSEA4. Additionally, OCT4 expression is not always localized to the nucleus, but it is a nucleocytoplasmic shuttling protein (90), and expression of OCT4 mediates partial cardiomyocyte reprogramming of mesenchymal stromal cells (91). Altogether, it seems that the altered metabolic state of the FPGS^ko cells predisposes them toward differentiation to cardiomyocytes and neurons. However, in normal, non–stem cell culture conditions, these cells are likely to be stressed, and this may explain the extremely high expression of ANKRD1, which is known to be expressed under stress conditions (46). Of interest, it has recently been reported that reduced cardiac hypertrophy and improved cardiac functions in mice is mediated by activation of serine and C1 metabolism (87).

Although we eliminated both the isoforms and all splicing variants of FPGS in this study, in future work it would be useful to investigate the importance of all SNPs of FPGS, perhaps using CRISPR technology. Genetic variation in folate-regulating enzymes is associated with the risk of various disorders, including cardiovascular disease (92). There are reports that the SNPs are associated with both metabolic disruption and disease risk, including neural tube defects (93). Therefore, a comprehensive study is warranted, which will be useful for individualized therapy of patients with cancer. In summary, we have found that the elimination of both FPGS isoforms in 293T cells triggers epigenetic modifications, influences gene expression, assists cellular plasticity, and reduces cell proliferation. Moreover, the FPGS^ko cells are directed toward cardiac and neuronal lineages.

Glossary

5-CHO-THF

5-formyl-THF

5-mC

5-methylcytosine

ALDH1L2

aldehyde dehydrogenase 1 family, member L2

ANKRD1

ankyrin repeat domain 1

ANOS1

anosmin-1

C1

one-carbon

Cas9

CRISPR-associated protein 9

CPC

cardiac progenitor cell

CRISPR

clustered regularly interspaced short palindromic repeats

DABG

detection above background

DKK1

Dickkopf WNT-signaling pathway inhibitor 1

DNMT

DNA methyltransferase

EB

embryoid body

ECAR

extracellular acidification rate

FBS

fetal bovine serum

FPGS

folypolyglutamate synthetase

FPGS^ko

FPGS knockout

GFP

green fluorescent protein

gRNA

guide RNA

GTSF1

gametocyte-specific factor 1

HEK

human embryonic kidney

HILIC

hydrophilic interaction liquid chromatography

IMDM

Iscove’s modified Dulbecco’s medium

MTHFD

methylene THF dehydrogenase

NEAA

nonessential amino acid

OCR

oxygen consumption rate

OCT4

octamer-binding transcription factor 4

OFP

orange fluorescent protein

RPMI

Roswell Park Memorial Institute

SAH

S-adenosylhomocysteine

SAM

S-adenosyl methionine

sgRNA

single guide RNA

SHMT

serine hydroxylmethyltransferase

SLC7

solute carrier family 7

SNP

single-nucleotide polymorphism

SOX2

sex-determining region Y-box 2

SSEA4

stage-specific embryonic antigen 4

THF

tetrahydrofolate

WT

wild type

AUTHOR CONTRIBUTIONS

A. C. Srivastava designed research, performed research, analyzed data, and wrote the paper; Y. G. Thompson, J. Singhal, J. Stellern, and A. Srivastava performed research; J. Du contributed new analytic tools and analyzed data; and T. R. O’Connor and A. D. Riggs designed research and wrote the manuscript.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.