GNMT Expression Increases Hepatic Folate Contents and Folate-Dependent Methionine Synthase-Mediated Homocysteine Remethylation

Yi-Cheng Wang,1 Yi-Ming Chen,2* Yan-Jun Lin,1 Shih-Ping Liu,2 and En-Pei Isabel Chiang1

1Department of Food Science and Biotechnology, National Chung Hsing University, Taichung, Taiwan, R.O.C; 2Institute of Microbiology and Immunology, National Yang-Ming University, Taipei, Taiwan, R.O.C.

Glycine N-methyltransferase (GNMT) is a major hepatic enzyme that converts S-adenosylmethionine to S-adenosylhomocys- teine while generating sarcosine from glycine, hence it can regulate mediating methyl group availability in mammalian cells. GNMT is also a major hepatic folate binding protein that binds to, and, subsequently, may be inhibited by 5-methyltetrafolate. GNMT is commonly diminished in human hepatoma; yet its role in cellular folate metabolism, in tumorigenesis and antifolate ther- apies, is not understood completely. In the present study, we investigated the impacts of GNMT expression on cell growth, folate status, methylfolate-dependent reactions and antifolate cytotoxicity. GNMT–diminished hepatoma cell lines transfected with GNMT were cultured under folate abundance or restriction. Folate-dependent homocysteine remethylation fluxes were investi- gated using stable isotopic tracers and gas chromatography/mass spectrometry. Folate status was compared between wild-type (WT), GNMT transgenic (GNMTtg ) and GNMT knockout (GNMTko ) mice. In the cell model, GNMT expression increased folate con- centration, induced folate-dependent homocysteine remethylation, and reduced antifolate methotrexate cytotoxicity. In the mouse models, GNMTtg had increased hepatic folate significantly, whereas GNMTko had reduced folate. Liver folate levels corre- lated well with GNMT expressions (r = 0.53, P = 0.002); and methionine synthase expression was reduced significantly in GNMTko, demonstrating impaired methylfolate-dependent metabolism by GNMT deletion. In conclusion, we demonstrated novel findings that restoring GNMT assists methylfolate-dependent reactions and ameliorates the consequences of folate depletion. GNMT ex- pression in vivo improves folate retention and bioavailability in the liver. Studies on how GNMT expression impacts the distribution of different folate cofactors and the regulation of specific folate dependent reactions are underway. © 2011 The Feinstein Institute for Medical Research, www.feinsteininstitute.org Online address: http://www.molmed.org doi: 10.2119/molmed.2010.00243

INTRODUCTION ating methionine from homocysteine hepatic folate-binding protein (2,3) that Different forms of folate serve as carri- remethylation (1). binds to, and, subsequently, may be in- ers of one-carbon units in DNA synthesis Glycine N-methyltransferase (GNMT, hibited by 5-methyl-THF (2). The expres- and biological methylation in mammals. EC2.1.1.20) is an abundant liver protein sion and function of GNMT have been in- The 10-formyl tetrahydrofolate is re- that converts S-adenosylmethionine to vestigated in human diseases (4–7) and quired for purine synthesis whereas the S-adenosylhomocysteine while generat- mouse models (8–10). Downregulation of 5,10-methylenetetrahydrofolate is essen- ing sarcosine from glycine. The GNMT GNMT has also been reported in human tial for pyrimidine synthesis. Among all reaction serves as an alternative pathway hepatocellular carcinoma. Loss of het- forms of folates, the 5-methyl tetrahydro- to regulate the S-adenosylmethionine to erozygosity within the GNMT gene in the folate (5-methyl-THF) is the most abun- S-adenosylhomocysteine balance and liver tissues of hepatocellular carcinoma dant that transfers the methyl group to availability of methyl group in mam- patients has been reported, and GNMT the enzyme methionine synthase, gener- malian cells (2). GNMT is also a major alteration appears to be an early event in human hepatocellular carcinoma (11). GNMT is commonly diminished in *Y-MC is a co–first author. human hepatoma and hepatoma cell lines Address correspondence and reprint requests to En-Pei Chiang, National Chung Hsing (4–10); hence it is believed to be a suscep- University, Taiwan, 250 Kuo-Kuang Road, Taichung, Taiwan 402, R.O.C. Phone: +886-4- tibility gene and a potential tumor sup- 22840385 ext 2190; Fax: +886-4-22876211; E-mail: [email protected]. pressor for human hepatoma (11). Indi- Submitted November 30, 2010; Accepted for publication December 30, 2010; Epub viduals with mutant GNMT and GNMT (www.molmed.org) ahead of print January 3, 2011. knockout mice showed that inactivation

486 | WANG ET AL. | MOL MED 17(5-6)486-494, MAY-JUNE 2011 RESEARCH ARTICLE

of GNMT had significant impacts on enzyme that catalyzes the irreversible con- restoring GNMT protein on folate metab- methyl donor S-adenosylmethionine sup- version of 5,10-methylene-tetrahydrofolate olism. Establishing stable clones with and plies as well as the S-adenosylmethionine to 5-methyl-THF, is inhibited by without GNMT was done in the human to S-adenosylhomocysteine balance S-adenosylmethionine (22–23). In mam- hepatoblastoma cell line HepG2 by trans- (8–10). Deletion of GNMT increased the mals, the only known reaction that requires fection and hygromycin (300 μg/mL) se- susceptibility of liver cancer in mice (10). 5-methyl-THF is the synthesis of methion- lection (25). A stable cell line, cotrans- The presence of GNMT in the liver and ine from homocysteine (16). On the other fected with pGNMT and pTK-Hyg kidney implies that this protein could hand, 5-methyl-THF is tightly bound to (Clontech, Palo Alto, CA, USA) plasmid participate in gluconeogenesis (12). The GNMT, and such binding inhibits GNMT DNAs, was used to represent cells with function of GNMT in extra hepatic tis- (24). Both we and others have shown that normal GNMT function (GNMT+). The sues is less clear. GNMT also is present in GNMT deletion led to abnormally high other stable cell line that cotransfected pancreas, prostate, intestinal mucosa, elevations in S-adenosylmethionine and with pFLAG-CMV-5 and pTK-Hyg plas- plasma and semen (12–17). The localiza- altered methylation status in vivo (8–10). mids was used to represent cells with di- tion of GNMT in the exocrine cells in Via its regulation in intracellular minished GNMT (GNMT–). GNMT ex- the above tissues suggests a potential role S-adenosyl methionine homeostasis pressions were confirmed by Western in secretion (12). GNMT serves as an al- and/or its tight binding to folate enzymes, blot analyses in both cell lines. ternative mechanism for utilizing it is plausible that GNMT can alter the To investigate the impacts of GNMT S-adenosylmethionine that does not nec- availability of cellular folate cofactors and expression on folate status, intracellular essarily involve the methylation of physi- further affect folate-dependent reactions. folate contents and folate-dependent ologically important acceptors (12). The In the present study, we hypothesized homocysteine remethylation fluxes were crystal structure of rat GNMT demon- that deletion of GNMT will result in low compared between GNMT+ and GNMT– strated two 5-methyl-THF binding sites intracellular folate levels owing to de- cells under folate-adequate (regular located in the intersubunit areas of the creased folate retention in the liver; and αMEM media containing 2.2 μmol/L tetramer (18). Each folate binding site the loss of GNMT can cause folate defi- folic acid) and folate-restricted consists of two 1–7 N-terminal regions of ciency specifically in those tissues origi- (10 nmol/L folinate) conditions. Cells one pair of subunits and two 205–218 re- nally expressing this folate binding pro- were grown in α-MEM containing 10% gions of the other pair of subunits, thus, tein. Alternatively, the tight binding (v/v) FCS, 0.12% NaHCO3, penicillin each GNMT tetramer binds two folate between GNMT and 5-methyl-THF might (100,000 units/L), streptomycin molecules. The N-terminal fragments of reduce the availability or bioactivity of (100 mg/L), amphotericin (0.25 mg/mL)

GNMT need significant conformational the enzyme-bound 5-methyl-THF, result- and 5% CO2 in an incubator at 37°C. The freedom to provide access to the active ing in a decrease in the 5-methyl-THF- medium was replaced every 72 h. The sites for folate binding as well as for the dependent homocysteine remethylation treatments, including adequate folate, inhibition by 5-methyl-THF (18). Although and methionine synthase expression. Im- and low folate (10 nmol/L folinate), were the binding between GNMT and 5-methyl- pacts of GNMT expression on folate sta- treated for 144 h. Cellular folate, THF has been well characterized, the sig- tus were investigated both in vitro and in S-adenocylmethionine and S-adenosyl - nificance of maintaining optical folate sta- vivo using cell lines with and without homocysteine concentrations were deter- tus in human pathological conditions is GNMT transfection and by using genetic mined as described previously (21). less clear, and the specific impact of mouse models with GNMT transgene or GNMT expression on hepatic methyl- GNMT disruption. The 5-methyl-THF de- Stable Isotope Tracer Studies folate-dependent reactions has not been pendent methionine synthesis metabolic GNMT-expressing HepG2 cells elucidated. In recent years, stable isotope fluxes were investigated in cell models (GNMT+) and the negative control tracers have been widely utilized to eluci- with and without GNMT expression. The (GNMT–) cells were cultured under con- date how specific folate enzymes mediate isotopic tracer studies enabled us to eluci- ditions of folate abundance (100 nmol/L and regulate the fluxes of one-carbon date the impacts of GNMT overexpres- folinate) or mild folate restriction units among folate- dependent reactions sion and deletion on methyl-folate- (10 nmol/L folinate) for 144 h. These lev- (19–20). We demonstrated that trans- dependent metabolic fluxes under els were based on our previous experi- formed human lymphoblasts with re- various folate conditions. ments and were used to generate mildly duced methylenetetrahydrofolate reduc- low intracellular folate without affecting tase (MTHFR) have advantages in de MATERIALS AND METHODS protein turnover significantly, and the novo purine synthesis when folate is ade- tracer can still be taken (21). To investi- quate, but they are more susceptible to Cell Lines and Culture Conditions gate the impact of GNMT expression on S-adenosylmethionine depletion when The diminished GNMT activity in the folate- dependent homocysteine folate is restricted (21). MTHFR, the HepG2 enabled us to study the impacts of remethylation fluxes, cells were plated in

MOL MED 17(5-6)486-494, MAY-JUNE 2011 | WANG ET AL. | 487 GNMT EXPRESSION INCREASES FOLATE STATUS IN VIVO

a 60 mm dish at 30% to 50% confluence the liver compared with the WT mice version of the Clifford/Kuory Folate De- in the treatment medium supplemented during the early stage. The GNMT ficient Amino Acid Rodent Diet #517777 2 ko with L-[2, 3, 3- H3] serine (50% of total knockout mouse model (GNMT ) was with RDA folate level. See http://www. 2 serine) combined with L-[5, 5, 5- H3]- established by gene targeting disruption: dyets.com/510000.html). Protein source leucine (50% of total leucine) for 72 h. the vector was constructed and trans- for rodent diets could have variable The folate-dependent homocysteine ferred into embryonic stem cells, and amino acid composition and uncertain remethylation flux in the HepG2 cells then microinjected into blastocytes of quantities of protein bound vitamins. To was calculated as the relative enrichment C57BL/6 mice described previously (8). minimize this effect and carefully control in methionine + 1 from 13C-serine (26). In comparison with the WT littermates the vitamin content throughout the ex- (GNMTwt[+/+]), the GNMT protein ex- perimental period, investigators, includ- Effects of GNMT Expression on pression decreased by approximately ing us, often choose amino acid–defined Methotrexate-Induced Cytotoxicity 50% in the liver of heterozygous mice diets in folate metabolism and vitamin The impacts of GNMT expression on (GNMThet[+/–]), and was undetectable restriction experiments (28–29). In study the antifolate drug cytotoxicity were in- in the nullizygous mice (GNMTnul[–/–]). III, parental mice and pups were fed an vestigated by treating these cells with var- To investigate the impacts of GNMT ex- amino acid–defined diet containing 2 mg ious doses of methotrexate. GNMT+ and pression on folate status, these genetic folate/kg diet (RDA for rodents) during GNMT– cells were plated in a 24-well mouse models were bred and raised mating, lactation, weaning and through- flat-bottomed culture plate in RPMI (3.5 × under specific pathogen-free conditions out the study periods. Folate status in 104 cells/mL) and treated with methotrex- with temperatures between 20°–25°C at the plasma, liver and spleen was com- ate (50–1000 nmol/L) for 72 h. Cell sur- 50% humidity with a 12-h light–dark pared between GNMTwt, GNMThet and vival was determined by the 3-(4,5- cycle throughout the study periods. Tail GNMTko. The offspring mice were eutha- cimethylthiazol-2-yl)-2,5-diphenyl DNA was obtained for genotyping the nized after an overnight fasting at 7 wk tetrazolium bromide (MTT) assay. The ab- offspring. of age. Plasma and tissues were har- sorbance was measured with a test wave- vested to investigate the effects of the length at 570 nmol/L by sunrise enzyme- Conditions and Diets GNMT defect on folate status and folate- linked immunosorbent assay (ELISA) All animal protocols were approved by dependent methionine synthase protein reader (Tecan, Salzburg, Austria). The IC50 the Institutional Animal Care and Use expression in the liver. values were calculated and compared be- Committee of National Chung Hsing tween GNMT+ and GNMT– cells. University. The animal experiments were Determination of Plasma and Tissue Methotrexate-induced cell apoptosis conducted as follows: Study I was con- Folate was then determined quantitatively by ducted to investigate the effects of Blood samples were collected in flow cytometry using the Annexin GNMT overexpression on folate status. potassium-EDTA tubes and a plasma V-FITC Apoptosis Detection Kit (Becton- Liver folate concentrations were com- sample was isolated by centrifugation at Dickinson, Franklin Lakes, NJ, USA). pared between GNMTtg and WT at wk 2, 977g for 10 min at 4°C and stored at GNMT+ and GNMT– cells were treated 5, 7, and 52 (1 year old). Study II was –80°C before analysis. Tissue folates were with low-dose (50 nmol/L) methotrexate performed to examine the impacts of extracted in freshly prepared folate ex- for 72 h, harvested, washed with cold GNMT deletion on folate status using traction buffer (30) containing 0.1% ascor- PBS and resuspended in Annexin bind- the GNMTko model. Male GNMThet were bic acid using a polytron homogenizer. ing buffer supplemented with Annexin mated with female GNMThet to obtain The supernatant fraction was collected V-FITC and PI. The mixture was incu- GNMTwt, GNMThet and male GNMTnul and then treated with mouse serum con- bated at 4°C in the dark for 10 min littermates. Liver folate concentrations jugase to convert the folylpolyglutamates and immediately analyzed using the were compared among GNMTwt, to their corresponding monoglutamate FACSCalibur system. GNMThet, and GNMTko 1 month after (31) at 37°C. The reaction was terminated weaning at 7 wks of age. at 121°C for 5 min. Folate concentrations Genetic Mouse Models GNMTtg and GNMTko mice received were measured by a standard microbio- The GNMT transgene mouse model chow ad libitum throughout the periods logical microtiter plate assay using Lacto- (GNMTtg) was established by microinject- in study I and II. To avoid lot-to-lot vari- bacillus casei (31). ing the Friend leukemia virus B (FVB)- ability in the protein-bound folates from fertilized eggs that had been transfected the chow diets, we further investigated Determination of Methionine Synthase with the pEGFP-GNMT vector containing folate status among tissues in GNMTko Expression full-length human GNMT cDNA as de- mice that were fed the modified Clifford The expression of folate dependent scribed previously (27). GNMTtg had a amino acid–defined rodent diet (Dyets, protein methionine synthase (MTR) was significant increase in GNMT protein in Bethlehem, PA, USA. We modified Dyets determined by western blot. Approxi-

488 | WANG ET AL. | MOL MED 17(5-6)486-494, MAY-JUNE 2011 RESEARCH ARTICLE

mately 0.03 g of liver was homogenized Table 1. Effects of GNMT expression on cell growth, protein turnover, and folate contents.a in ten volumes of RIPA buffer containing Adequate Folate 0.1% (v/v) protease inhibitor cocktail Set 1 folateb restrictionb % Change P value (Calbiochem, La Jolla, CA, USA). The tis- c sue lysates were centrifuged at 18,360g Cell growth Experiment Ib for 30 min at 4°C and the supernatants GNMT–b 55.3 ± 0.6 57.0 ± 0.5 +3.0 ± 0.8 0.016 were frozen in –80°C until analyses. The GNMT+b 53.3 ± 1.2 53.6 ± 0.4 0.6 ± 0.7 0.68 protein content was quantified by BCA P value 0.06 <0.001 Protein Assay (Pierce, Rockford, IL, % Difference –3.6 ± 2.3 –5.9 ± 0.6 USA). One-hundred micrograms of pro- Experiment IIb tein from each tissue were denatured in GNMT– 52.5 ± 0.6 73.7 ± 0.3 +40.5 ± 0.5 <0.001 and then separated on 12% SDS-PAGE GNMT+ 51.1 ± 0.8 69.5 ± 0.7 +36.1 ± 1.3 <0.001 gels and transferred onto a PVDF mem- P value 0.07 <0.001 brane. After blocking with TBS contain- % Difference –2.7 ± 1.6 –5.4 ± 0.9 d ing 10% skim milk for 2 h, the mem- Protein turnover branes were incubated with the primary GNMT– 0.16 ± 0.01 0.14 ± 0.02 –7.6 ± 13.7 0.40 GNMT+ 0.15 ± 0.01 0.15 ± 0.01 0.4 ± 1.6 0.93 antibody, anti-MTR (1:1000) (Abcam, P value 0.23 0.72 Cambridge, UK), β-actin (1:5000) (Milli- % Difference –4.9 ± 5.9 3.3 ± 1.6 pore, California, USA) in TBS containing Folate (ng/106 cells) 5% skim milk at 4°C overnight. Mem- GNMT– 7.6 ± 0.4 2.0 ± 0.5 –73.7 ± 6.9 <0.001 branes were washed 3x with TBS contain- GNMT+ 11.4 ± 0.4 4.6 ± 0.5 –54.5 ± 4.7 0.017 ing 0.1% Tween 20 (TBST) and then the P value <0.001 0.003 membranes were covered with HRP- % Difference +49.5 ± 4.9 +130.7 ± 23.8 linked antigoat or antimouse IgG (1:5000) a at room temperature for 2 h. The im- All data are presented as means ± SD (n = 3). Each experiment was repeated at least twice. The P values were calculated by Student t test comparing two cell lines under each munoblots were visualized by enhanced condition. The % change was calculated by comparing the mean with those under chemiluminescence kit (New England adequate folate. The % difference was calculated by comparing the mean with negative Biolabs, Beverly, MA, USA). To ensure control cells under the same culture condition. The bolded data were statistically equal protein loading, each membrane significant (P < 0.05). β was stripped and reprobed with anti- - bCell lines and culture conditions. GNMT–: wild-type HepG2 cells transfected with vector actin antibody. only were used as negative control. GNMT+: HepG2 cells transfected with GNMT. Cells were cultured in a modified α-MEM medium under folate abundance (100 nmol/L STATISTICS folinate) or folate restriction (10 nmol/L folinate) for 144 h (Experiment I) or 240 h All experiments were performed in (Experiment II). c triplicate. The statistical significance of Cell growth is presented as the doubling time (h) of each cell line under different folate difference in the means among groups conditions. d 2 Cells were cultured in α-MEM medium supplemented with L-[5,5,5- H ]-leucine (200 μmol/L, was determined by analysis of variance 3 50% of total leucine) for 72 h. Protein turnover was estimated by the enrichments of leucine (ANOVA). A Student t test was per- in the cellular proteins. Protein turnover was measured in cells cultured in low folate for 144 h. formed to detect any significant differ- ence in the means between the control and treatment groups (SYSTAT, SPSS, deplete conditions (Table 1). Under ade- late for 144 h), doubling time increased Chicago, IL, USA). A P value < 0.05 was quate folate, GNMT expression tended significantly in GNMT– cells (P = 0.016) considered significant. to reduce the doubling time in HepG2 but not GNMT+ cells (P = 0.68). These cells by 1.5 to 2 h (mean doubling time data indicated that GNMT-expressing RESULTS was compared between GNMT+ and HepG2 cells were less sensitive to folate GNMT– by Student t test, P = 0.06 and depletion compared with cells with di- Effects of GNMT and Folate Supplies P = 0.07 in Experiment I and II, respec- minished GNMT expression. Growth re- on Cell Growth, Protein Turnover, and tively). When these cells were cultured tardation occurred in both cell lines Folate Concentrations under folate restriction, the difference in when we prolonged folate depletion to The doubling time, protein turnover, doubling time between GNMT+ and 240 h, yet the doubling time of GNMT+ and intracellular folate concentrations GNMT– became significant in both ex- was significantly shorter than that of the were compared between GNMT– and periments (P < 0.001) (Table 1). In mild GNMT– (P < 0.001) (Table 1). These GNMT+ cells under folate replete and folate restriction (Experiment I, low fo- results suggested that GNMT-expressing

MOL MED 17(5-6)486-494, MAY-JUNE 2011 | WANG ET AL. | 489 GNMT EXPRESSION INCREASES FOLATE STATUS IN VIVO

HepG2 cells are less sensitive to folate depletion. 2 Using L-[5, 5, 5- H3]-leucine as the tracer we found that neither GNMT ex- pression nor mild folate restriction sig- nificantly altered leucine enrichment in the cellular proteins, suggesting that total protein turnover did not change under these conditions (Table 1). S-adenosylmethionine was decreased significantly (GNMT– versus GNMT+ cells = 3,195.8 ± 114.1 versus 2,313.4 ± 134.0 [pmol/mg protein], P < 0.001) and S-adenosylhomocysteine increased (GNMT– versus GNMT+ cells = 18.2 ± 1.2 versus 36.1 ± 0.8 [pmol/mg protein], P = 0.003) by GNMT expression. This obser- vation was in agreement with the induced conversion of S-adenosylmethionine to S-adenosylhomocysteine by GNMT. We discovered that GNMT expression Figure 1. GNMT expression significantly re- specifically improved folate status under duced antifolate methotrexate-induced both folate repletion and mild folate re- cytotoxicity in vitro. striction in HepG2 cells. Folate concentra- tion was 50% higher than that of the GNMT expression can protect HepG2 GNMT– under folate repletion and folate cells from apoptosis induced by low dose concentration was 131% higher than that methotrexate treatments. At 50 nmol/L, of the GNMT– under folate restriction. MTX significantly induced apoptosis in Furthermore, folate concentration reduced GNMT– cells (P = 0.011 versus untreated by 74% in GNMT– whereas folate concen- cells), but not GNMT+ cells, as GNMT+ tration decreased by 55% in GNMT+ cells had a similarly low percentage of (Table 1). These results proved that cellu- apoptotic cells as the untreated cells. lar folate contents in GNMT expressing These results demonstrated that restoring cells are less depleted in response to folate GNMT expression in HepG2 cells can re- restriction, and that GNMT expression can duce methotrexate-induced cytotoxicity directly improve folate status regardless of significantly (Figure 1). medium folate conditions. GNMT Expression Significantly GNMT Expression Significantly Improved Hepatic Folate Status In Vivo Reduced Antifolate Methotrexate- In the mouse models, we discovered Related Cytotoxicity that GNMT expression significantly al- To investigate if the better folate status ters hepatic folate status in vivo (Fig- Figure 2. GNMT expression significantly al- presented in GNMT+ actually protects ures 2–3). Normally, mice express GNMT ters hepatic folate status in (A) GNMT tg these cells from antifolate toxicity, we fur- protein in the liver at around 3 wks of transgenic mice (GNMT ) (n = 6), com- ther examined the effects of GNMT ex- age, around the time of weaning. The en- pared with the wild-type (WT) littermates (n = 6); and in (B) the GNMT knockout pression on the cytotoxicity induced by dogenous murine GNMT protein was ei- mice (GNMTnul homozygous mice n = 5), low dose methotrexate. Methotrexate in- ther undetectable or detected at very low compared with age matched littermates. hibited cell proliferation in both cell lines, levels during the early stage of develop- GNMTwt: wild-type (n = 8); GNMThet: het- but GNMT+ cells were less sensitive to ment before weaning. In contrast, the erozygous (n = 7). (C) Liver folate concen- tg low dose (50–100 nmol/L) methotrexate. GNMT mice express human GNMT at trations were positively correlated with the The IC50 value of methotrexate was birth (27). To investigate the impacts of liver GNMT expression levels when both 550 nmol/L for GNMT+ and 450 nmol/L GNMT expression on folate status, liver models were combined (r = 0.53, P = for GNMT– cells. Furthermore, restoring folate levels were compared between the 0.002, n = 32).

490 | WANG ET AL. | MOL MED 17(5-6)486-494, MAY-JUNE 2011 RESEARCH ARTICLE

AB C

Figure 3. Tissue specificity of GNMT disruption on folate status. GNMT knockout significantly reduced folate concentrations (A) in plasma; (B) in GNMT expressing tissue liver; (C) but not in non-GNMT–expressing tissue spleen. GNMT wild-type (GNMTwt), heterozygous (GNMThet), homozygous (GNMTnul) mice were bred from crossing heterozygous (GNMThet) parents that received amino acid defined diet with 2 mg/kg folic acid and 0.1 % sulfathiazole. The offspring mice were fed the same diet after weaning.

WT littermates and GNMTtg mice before cant difference in body weight or food minished GNMT. Taken together, we wt het and after weaning in Study I. intake among the GNMT , GNMT and proved that GNMT expression increases ko As we postulated, liver folate status is GNMT mice (data not shown). There- cellular methyl-folate retention, and the closely related and is comparable to the fore we suggest that the reduced folate retained folates are used effectively for ko hepatic GNMT expression pattern. Com- levels observed in GNMT mice was in- the only biochemical reaction in which pared with WT, hepatic GNMT levels dependent of lower folate intake. 5-methyl-THF participates. were only significantly elevated during wks 1 to 3 in GNMTtg. Consistently, liver GNMT Expression Induced Folate- GNMT Deletion Reduced Folate in the folate levels significantly increased in Dependent Homocysteine Liver GNMTtg during this time period (mean ± Remethylation Next, we investigated whether GNMT SD in WT versus GNMTtg = 8.5 ± 4.1 ver- To investigate whether the increased plays an essential role in folate homeo- sus 17.5 ± 8.0 μg/g tissue, P = 0.033, n = folate by GNMT overexpression can be stasis by examining folate status in 6 in each group) (Figure 2A). Further- bioavailable and utilizable for 5-methyl- GNMT knockout mice (study II). Com- wt more, the hepatic GNMT expressions THF-dependent methionine synthesis, pared to GNMT (+/+) (n = 8), mean correlated well with liver folate concen- the homocysteine remethylation fluxes liver folate concentrations were reduced het trations before weaning (r = 0.615, P = were investigated further in stable iso- by approximately 39% in GNMT (+/–) 0.033, n = 12, data not shown). After topic tracer experiments under conditions (P = 0.047, n = 7) and by 50% (P = 0.017, nul weaning, GNMT expressions did not dif- of folate abundance and folate restriction. n = 5) in GNMT (–/–) mice (Figure fer between GNMT tg and WT. Consistent Under our experimental conditions, nei- 2B). These data demonstrated that inac- with this pattern, folate concentrations in ther GNMT expression nor mild folate re- tivation of GNMT directly led to loss of the liver did not differ between WT and striction significantly altered leucine en- hepatic folate in a dose-dependent man- tg 2 GNMT at wk 5 (6.3 ± 2.4 versus 4.0 ± richment from L-[5,5,5- H3]-leucine tracer. ner. Furthermore, when all data were 2.2, P = 0.2), wk 8 (6.4 ± 1.1 versus 6.8 ± On the other hand, GNMT expression sig- combined from these mice expressing wt 3.0, P = 0.8), or up to wk 52 (6.3 ± 2.4 nificantly increased the relative methion- different levels of GNMT (GNMT , 13 tg ko versus 4.0 ± 2.2, P = 0.7) (n = 5–6 per ine +1 enrichment from the C-serine GNMT and GNMT ), liver folate con- group at each time point, data not tracer, both in conditions of adequate fo- centrations correlated significantly with shown). Results from Study I supported late and low folate (Table 2). These results hepatic GNMT expression levels in our hypothesis that GNMT expression supported our postulation that the these animals (R = 0.53, P = 0.002, n = 32) can improve folate status, presumably methyl-folate- dependent homocysteine (Figure 2C). Mice in Study I and Study via increased hepatic folate retention in remethylation fluxes can be promoted by II were fed the chow diet throughout vivo. We have not observed any signifi- restoring GNMT function in cells with di- the study periods.

MOL MED 17(5-6)486-494, MAY-JUNE 2011 | WANG ET AL. | 491 GNMT EXPRESSION INCREASES FOLATE STATUS IN VIVO

Table 2. Intracellular folate concentrations and folate dependent homocysteine remethylation fluxes in HepG2 cells with and without GNMT expression.a

Leu +3 Ser +1 Met +1 Met +1 from enrichmentb enrichmentb enrichmentb13C-serineb

Adequate folatec GNMT– 0.161 ± 0.002 0.131 ± 0.004 0.007 ± 0.001 0.055 ± 0.001 GNMT+ 0.153 ± 0.01 0.181 ± 0.014 0.028 ± 0.001 0.156 ± 0.016 P value 0.231 0.004 <0.001 <0.001 % Difference –4.9 ± 5.9 +38.6 ± 10.5 +288.9 ± 13.6 +10.1 ± 1.7 Low folatec GNMT– 0.149 ± 0.022 0.138 ± 0.008 0.008 ± 0.001 0.056 ± 0.008 GNMT+ 0.154 ± 0.002 0.173 ± 0.003 0.021 ± 0.001 0.118 ± 0.01 P value 0.725 0.002 <0.001 0.001 % Difference 3.2 ± 1.6 +25.2 ± 2.4 +163.2 ± 18.7 +6.2 ± 1.04 aAll data are presented as means ± SD (n = 3). The P value, calculated by t test, compared two cell lines under each condition. The % difference was calculated by comparing the mean with negative control cells under the same culture condition. Cell lines. GNMT–: WT HepG2 cells transfected with vector only were used as negative control. GNMT+: HepG2 cells transfected with GNMT. b Folate dependent homocysteine remethylation fluxes were calculated as the relative Figure 4. GNMT disruption significantly re- 13 α enrichments in methionine +1 from C-serine. Cells were cultured in -MEM medium duces hepatic MTR protein expression in 13 μ wt supplemented with L-[ C]-serine (237.8 mol/L, 100% of total serine) combined with mice. GNMT wild-type (GNMT ) and 2 L μ nul -[5,5,5- H3]-leucine (200 mol/L, 50% of total leucine) for 72 h. knockout (GNMT ). cCulture conditions. Cells were cultured in a modified α-MEM medium under folate abundance (100 nmol/L folinate) or mild folate restriction (10 nmol/L folinate) for 144 h. methyltransferase (MTR) protein product in Study III was conducted in GNMTko sion not only improves folate status but the liver. A recent study demonstrated that model–offspring mice on amino acid- also plays a crucial role in folate retention low dietary folate led to higher betaine de- based diet containing RDA folate contents in the liver. Finally, the hepatic methyl- mand and reduced MTR expression in throughout the life period. GNMT nul(–/–) folate- dependent enzyme MTR was re- mice (32). Our unpublished work also in- had significantly lower plasma folate con- duced in GNMT nul(–/–) mice (Figure 4). dicated that folate restriction causes low centrations compared with GNMTwt(+/+) hepatic folate concentrations, decreased (P = 0.009) and GNMThet(+/–) (P = 0.049) DISCUSSION MTR protein and reduced folate depen- (Figure 3A). Consistently, GNMTnul(–/–) In the present study, we demonstrated dent homocysteine remethylation fluxes in has reduced folate contents compared numerous findings. (a) Restoring GNMT mice livers. These observations support with GNMT wt(+/+) (P = 0.002) and in cells with diminished GNMT can im- the postulation that reduction in hepatic GNMT het(+/–) (P = 0.077) (Figure 3B). prove intracellular folate status. (b) MTR protein seen in GNMTnul(–/–) could Plasma folate levels reflected folate status GNMT expression can increase 5-methyl- result from hepatic folate depletion. in the liver (r = 0.42, P = 0.066). In con- THF-dependent metabolic fluxes in Folate status can determine the pheno- trast, folate concentrations did not differ GNMT–deficient cells. (c) GNMT amelio- typic expressions of folate metabolic en- among different genotypes in rates the growth retardation induced by zymes and functions. One good example non-GNMT–expressing spleen tissue (Fig- folate depletion. (d) Expression of is found in the interactions between folate ure 3C). Hepatic folate concentrations GNMT can protect HepG2 cells from status and MTHFR C677T poly morphism were significantly associated with GNMT low-dose methotrexate induced apopto- in regulating different folate-dependent expression in the liver. These results sis. (e) In vivo GNMT expression im- biochemical reactions. Previously, we demonstrated that GNMT deletion had proves folate status presumably owing to demonstrated that human lymphoblasts minimal impact on folate concentrations increased retention and bioavailability in with weaker MTHFR (homozygous of in the tissues without endogenous GNMT the liver. (f) Destruction of GNMT in vivo MTHFR C677T) have advantages in de expression; and the impact of GNMT specifically reduces hepatic folate and novo purine synthesis when folate is ade- deletion on folate status was closely re- decreases methylfolate-dependent me- quate, but they are more susceptible to lated to the tissue-specific expression pat- thionine synthase expression in the liver. S-adenosyl methionine depletion when terns of this folate-binding protein. These Our data demonstrated that GNMT dis- folate is restricted (21). In human colon results once again supported our postula- ruption in mice resulted in reduction of and breast cancer cell models, MTHFR tion that normal hepatic GNMT expres- the 5-methyltetrahydrofolatehomocysteine C677T mutation can induce cell-specific

492 | WANG ET AL. | MOL MED 17(5-6)486-494, MAY-JUNE 2011 RESEARCH ARTICLE

alterations in DNA methylation (33) as protective effects of GNMT against liver and utilized for 5-methylfolate- well as uracil incorporations into the cancer tumorigenesis and progression. dependent reactions when needed. DNA (34). Both are potential molecular The interactions and regulations among GNMT is a widely known folate- bases for cell- or site-specific cancer risk S-adenosylmethionine, GNMT, MTHFR binding protein that is sensitive to inhibi- modification. MTHFR, a crucial enzyme and cellular folate are potential molecular tion by 5-methyl-THF polyglutamates that alters the distribution and utiliza- mechanisms for cancer risk modification (40). Here we provide novel in vivo and in tion of different folate cofactors, is inhib- and are under investigation. vitro evidence that, by binding to ited by S-adenosyl methionine. It is As the regulation of the key enzymes 5-methyl-THF, GNMT may serve as a therefore plausible that hepatic GNMT in the folate and methionine cycle is spe- reservoir for intracellular folate that can expression affects the competition of fo- cific to gene, tissue or cell type (33), the be further utilized for folate-dependent late cofactors between folate-dependent impacts of GNMT expression on folate reaction including homocysteine reactions via its regulation of intracellu- status and folate-dependent reactions remethylation. We provide evidence that lar S-adenosylmethionine homeostasis in may well differ among tissues. In our in GNMT expression retains the folate in the liver. vitro experiments, HepG2 was chosen as the liver; however, how GNMT can im- The inhibition of GNMT by 5-methyl- it retains morphological and biological prove plasma folate levels remains to be THF is an effective way to adjust and characteristics of normal human hepato- investigated. We postulate that the re- maintain the methyl group homeostasis cytes (35–37). Also, the endogenous duced plasma folate concentrations in in mammals. High dietary methionine GNMT expression in HepG2 is almost GNMTnul(–/–) mice reflect poor folate intake leads to increased hepatic undetectable, so we can easily distin- status due to chronic folate loss by in- S-adenosylmethionine that inhibits guish the specific impacts derived from creased excretion. The urinary folate ex- MTHFR and lowers 5-methyl-THF levels. GNMT expression in this model. Fur- cretion is currently under investigation. The reduced methylfolates alleviate the thermore, the diminished GNMT activity Our present study demonstrated that in- inhibition of GNMT, leading to more con- in these cells may represent the inactive tracellular folate status could be improved version of S-adenosylmethionine to GNMT protein in human hepatoma cells, effectively when GNMT–diminished cells S-adenosylhomocysteine (12). Normal thus enabling us to study the impacts of were transfected with GNMT, and restor- GNMT function not only can protect ani- restoring GNMT protein on folate metab- ing GNMT could ameliorate the conse- mals against methionine toxicity, but also olism in these cells by transfecting quences of folate restriction. Furthermore, might assist cells to conserve methyl GNMT. To examine the tissue-specific GNMT could further ameliorate the cyto- groups under methionine restriction. We impacts of GNMT expression on folate toxicity of antifolate treatment(s). Results recently discovered in vitro that GNMT status in vivo, folate status in the liver from the present study provided direct expression does not exacerbate methyl and the spleen were determined to repre- evidence that GNMT expression in hepa- group deficiency when the methionine sent tissues with and without normal tocytes could improve folate status and supply is limited; instead it can protect GNMT expressions. Consistent with the that GNMT could play a crucial role in cells from further hypomethylation in- expression pattern of GNMT, low folate folate retention. duced by methionine depletion. Restoring status was found in GNMTko mice, and In conclusion, these mice and cell lines GNMT expression in GNMT diminished folate concentrations correlated with are feasible models for future investiga- cell lines is crucial in maintaining methyl GNMT expression in the liver. Con- tions of the interactions between GNMT group homeostasis via homocysteine versely, no difference was found in folate expression and folate metabolism in dis- transmethylation and transsulfuration ki- concentrations in the spleen between ease occurrence, progression or during netics (41). Adding to the current knowl- GNMTwt and GNMTko mice. antifolate treatments. Based on results edge, results from the present study Studies of mathematical modeling on from the present study, we suggest that further demonstrated evidence that in hepatic one carbon metabolism indicated normal GNMT function is important for GNMT–abundant tissue, that is, the liver, that as total folate decreases, the dissoci- reducing liver cytotoxicity and that nor- normal GNMT function is crucial for ations of numerous folate-enzyme com- mal GNMT function should be consid- optimal folate status and hepatic- folate- plexes increase both the amount of active ered as a factor during antifolate chemo- dependent biochemical reactions in vivo. enzyme and additional free folate therapy or immunosuppressive Our experiments in genetic mouse models (38–39). These results implied that the treatments. As defective GNMT is com- demonstrated that inactivation of GNMT enzyme-bound folates can be released monly found in early hepatoma, the im- directly resulted in loss of hepatic folate in under folate depletion. Results from our pacts of GNMT on folate metabolism a dose-dependent manner, and normal current study suggested that the GNMT may in part account for the protective GNMT expression is critical for folate re- expression can improve the retention of role of GNMT against liver tumorigene- tention in the liver. We suggest that these folate in the liver; and GNMT–bound sis. More studies are needed to deter- mechanisms may in part account for the 5-methyl-THF can be released readily mine whether these mechanisms par-

MOL MED 17(5-6)486-494, MAY-JUNE 2011 | WANG ET AL. | 493 GNMT EXPRESSION INCREASES FOLATE STATUS IN VIVO

tially account for the protective role of 11. Tseng TL, et al. (2003) Genotypic and phenotypic 27. Yen CH, et al. (2009) Glycine N-methyltransferase GNMT against liver tumorigenesis. Stud- characterization of a putative tumor susceptibility affects the metabolism of aflatoxin B1 and blocks ies on how GNMT expression impacts gene, GNMT, in liver cancer. Cancer Res. 63:647–54. its carcinogenic effect. Toxicol. Appl. Pharmacol. 12. Yeo EJ, Wagner C. (1994) Tissue distribution of 235:296–304. the distribution of different folate cofac- glycine N-methyltransferase, a major folate- 28. Schwahn BC, et al. (2004) Effects of betaine in a tors and the regulation of specific folate binding protein of liver. Proc. Natl. Acad. Sci. murine model of mild cystathionine-beta- dependent reactions are underway. U. S. A. 91:210–4. synthase deficiency. Metabolism. 53:594–9. 13. Chen YM, et al. (1998) Characterization of 29. Chang HY, Tzen JT, Lin SL, Wu YT, Chiang EP. glycine-N-methyltransferase-gene expression in (2011) Long-term prednisolone treatments in- ACKNOWLEDGMENTS human hepatocellular carcinoma. Int. J. Cancer. crease bioactive vitamin B6 synthesis in vivo. This project was supported in part by 75:787–93. J. Pharmacol. Exp. Ther. 2011;337:102–9. National Science Council (NSC98-2320- 14. Chen YM, et al. (2000) Genomic structure, expres- 30. Chiang EP, et al. (2005) Inflammation causes tis- B005-004MY3, EP Chiang) and by the sion, and chromosomal localization of the human sue-specific depletion of vitamin B6. Arthritis Res. glycine N-methyltransferase gene. Genomics. Ther. 7:R1254–62. Department of Health in Taiwan (DOH 66:43–7. 31. O’Broin S, Kelleher B. (1992) Microbiological 97-TD-D-113-97011, EP Chiang). 15. Kerr SJ. (1972) Competing methyltransferase sys- assay on microtitre plates of folate in serum and tems. J. Biol. Chem. 247:4248–52. red cells. J. Clin. Pathol. 45:344–7. DISCLOSURE 16. Pilch B, Mann M. (2006) Large-scale and high- 32. Christensen KE, et al. (2010) Steatosis in mice is The authors declare that they have no confidence proteomic analysis of human seminal associated with gender, folate intake, and expres- plasma. Genome Biol. 7:R40. sion of genes of one-carbon metabolism. J. Nutr. competing interests as defined by Molec- 17. Schenk S, Schoenhals GJ, de SG, Mann M. (2008) 140:1736–41. ular Medicine, or other interests that A high confidence, manually validated human 33. Stempak JM, Sohn KJ, Chiang EP, Shane B, Kim might be perceived to influence the re- blood plasma protein reference set. BMC Med. YI. (2005) Cell and stage of transformation- sults and discussion reported in this Genomics. 1:41. specific effects of folate deficiency on methionine 18. Luka Z, et al. (2007) 5-methyltetrahydrofolate is cycle intermediates and DNA methylation in an paper. bound in intersubunit areas of rat liver folate- in vitro model. Carcinogenesis. 26:981–90. binding protein glycine N-methyltransferase 34. Sohn KJ, et al. (2009) The methylenetetrahydrofo- REFERENCES J. Biol. Chem. 282:4069–75. late reductase C677T mutation induces cell- 1. Brody T, Shane B. (1984) Folic Acid. In: Handbook 19. Herbig K, et al. (2002) Cytoplasmic serine hy- specific changes in genomic DNA methylation of Vitamins: Nutritional, Biochemical, and Clinical droxymethyltransferase mediates competition and uracil misincorporation: a possible molecu- Aspects. Machlin LJ (ed) Marcel Deller, New between folate-dependent deoxyribonucleotide lar basis for the site-specific cancer risk modifica- York, pp. 459–96. and S-adenosylmethionine biosyntheses. J. Biol. tion. Int. J. Cancer. 124:1999–2005. 2. Cook RJ, Wagner C. (1984) Glycine N-methyl- Chem. 277:38381–9. 35. Fukuda Y, Ishida N, Noguchi T, Kappas A, Sassa 20. Anguera MC, et al. (2006) Regulation of folate- S. (1992) Interleukin-6 downregulates the expres- transferase is a folate binding protein of rat liver mediated one-carbon metabolism by 10- sion of transcripts encoding cytochrome P450 cytosol Proc. Natl. Acad. Sci. U. S. A. 81:3631–4. formyltetrahydrofolate dehydrogenase. J. Biol. IA1, IA2 and IIIA3 in human hepatoma cells. 3. Zamierowski MM, Wagner C. (1977) Identifica- Chem. 281:18335–42. Biochem. Biophys. Res. Commun. 184:960–5. tion of folate binding proteins in rat liver. J. Biol. 21. Chiang EP, Wang YC, Tang FY. (2007) Folate re- 36. Knowles BB, Howe CC, Aden DP. (1980) Human Chem. 252:933–8. striction and methylenetetrahydrofolate reductase hepatocellular carcinoma cell lines secrete the 4. Luka Z, Cerone R, Phillips JA III, Mudd HS, 677T polymorphism decreases adoMet synthesis major plasma proteins and hepatitis B surface Wagner C. (2002) Mutations in human glycine via folate-dependent remethylation in human- antigen. Science. 209:497–9. N-methyltransferase give insights into its role in transformed lymphoblasts. Leukemia. 21:651–8. 37. Aden DP, Fogel A, Plotkin S, Damjanov I, methionine metabolism Hum. Genet. 110:68–74. 22. Jencks DA, Mathews RG. (1987) Allosteric inhibi- Knowles BB. (1979) Controlled synthesis of 5. Avila MA, et al. (2000) Reduced mRNA abun- tion of methylenetetrahydrofolate reductase by HBsAg in a differentiated human liver carcinoma- dance of the main enzymes involved in methion- adenosylmethionine. Effects of adenosylmethion- derived cell line. Nature. 282:615–6. ine metabolism in human liver cirrhosis and he- ine and NADPH on the equilibrium between ac- 38. Wagner C, Briggs WT, Cook RJ. (1985) Inhibition patocellular carcinoma. J. Hepatol. 33:907–14. tive and inactive forms of the enzyme and on the of glycine N-methyltransferase activity by folate 6. Augoustides-Savvopoulou P, et al. (2003) Glycine kinetics of approach to equilibrium. J. Biol. Chem. derivatives: implications for regulation of methyl N-methyltransferase deficiency: a new patient with 262:2485–93. group metabolism. Biochem. Biophys. Res. Com- a novel mutation. J. Inherit. Metab. Dis. 26:745–59. 23. Kutzbach C, Stokstad EL. (1967) Feedback inhibi- mun. 127:746–52. 7. Mudd SH, et al. (2001) Glycine N-methyltransferase tion of methylene-tetrahydrofolate reductase in 39. Nijhout HF, Reed MC, Budu P, Ulrich CM. (2004) deficiency: a novel inborn error causing persis- rat liver by S-adenosylmethionine. Biochim. Bio- A mathematical model of the folate cycle: new tent isolated hypermethioninaemia. J. Inherit. phys. Acta. 139:217–20. insights into folate homeostasis. J. Biol. Chem. Metab. Dis. 24:448–64. 24. Luka Z (2008) Methyltetrahydrofolate in folate- 279:55008–16. 8. Liu SP, et al. (2007) Glycine N-methyltransferase–/– binding protein glycine N-methyltransferase. 40. Reed MC, et al. (2006) A mathematical model mice develop chronic hepatitis and glycogen stor- Vitam. Horm. 79:325–45 gives insights into nutritional and genetic aspects age disease in the liver. Hepatology. 46:1413–25. 25. Chen SY, et al. (2004) Glycine N-methyltransferase of folate-mediated one-carbon metabolism. 9. Luka Z, Capdevila A, Mato JM, Wagner C. (2006) tumor susceptibility gene in the benzo(a)pyrene- J. Nutr. 136:2653–61. A glycine N-methyltransferase knockout mouse detoxification pathway. Cancer Res. 64:3617–23. 41. Wang YC, Tang FY, Chen SY, Chen YM, Chiang model for humans with deficiency of this en- 26. Chiang EP, Wang YC, Chen WW, Tang FY. (2009) EP. (2011) Glycine-N methyltransferase expres- zyme. Transgenic Res. 15:393–7. Effects of insulin and glucose on cellular meta- sion in HepG2 is involved in methyl group ho- 10. Martinez-Chantar ML, et al. (2008) Loss of the bolic fluxes in homocysteine transsulfuration, meostasis by regulating transmethylation kinet- glycine N-methyltransferase gene leads to steato- remethylation, S-adenosylmethionine synthesis, ics and DNA methylation. J. Nutr. 141:777–82. sis and hepatocellular carcinoma in mice. Hepa- and global deoxyribonucleic acid methylation. tology. 47:1191–9. J. Clin. Endocrinol. Metab. 94:1017–25.

494 | WANG ET AL. | MOL MED 17(5-6)486-494, MAY-JUNE 2011