Acetylation of MAT IIα represses tumour cell growth and is decreased in human hepatocellular cancer

Metabolic alteration is a hallmark of cancer. Dysregulation of methionine metabolism is implicated in human liver cancer. Methionine adenosyltransferase IIα (MAT IIα) is a key enzyme in the methionine cycle, catalysing the production of S-adenosylmethionine (SAM), a key methyl donor in cellular processes, and is associated with uncontrolled cell proliferation in cancer. Here we show that P300 acetylates MAT IIα at lysine residue 81 and destabilizes MAT IIα by promoting its ubiquitylation and subsequent proteasomal degradation. Conversely, histone deacetylase-3 deacetylates and stabilizes MAT IIα by preventing its proteasomal degradation. Folate deprivation upregulates K81 acetylation and destabilizes MAT IIα to moderate cell proliferation, whereas a single mutation at K81 reverses the proliferative disadvantage of cancer cells upon folate deprivation. Moreover, MAT IIα K81 acetylation is decreased in human hepatocellular cancer. Collectively, our study reveals a novel mechanism of MAT IIα regulation by acetylation and ubiquitylation, and a direct functional link of this regulation to cancer development.

F olate is essential for rapidly proliferating cells and plays a pivotal role in one-carbon metabolism. Its metabolic derivative tetrahydrofolate (FH4), known as the carrier of one-carbon units, is involved in de novo synthesis of thymidylate and purine, amino-acid inter-conversion and so on 1 . Coupling with methionine cycle, which produces the major methyl donor S-adenosylmethionine (SAM), folate metabolism also presents methyl-groups to global methylation of macromolecules.
Methionine adenosyltransferase (MAT) is the key enzyme in methionine cycle, catalysing biosynthesis of SAM. In mammals, two distinct genes, MAT1A and MAT2A, encode two homologous catalytic subunits-a1 and a2, respectively 2 . MAT1A mainly expresses in healthy hepatocytes. Its encoding product, the a1 subunits, assemble into two MAT isozymes, MAT III (dimer) and MAT I (tetramer) 2 . In contrast, MAT2A and its coding product, MAT IIa (dimer formed from a2 subunits), is widely expressed in all exhepatic tissues at relatively low level under normal conditions 3 , whereas overexpressed in various human epithelial tumours, including gastric, colon and liver cancer [4][5][6][7] .
Studies indicated that enhanced MAT2A expression boosts cancer cell proliferation and potentially promotes tumour development and progression 2,8,9 . In addition, MAT2A, but not MAT1A, is found expressed in fetal liver, whereas healthy adult liver expresses mainly MAT1A 3 . MAT expression is progressively switched from MAT2A to MAT1A during liver development, whereas is reversed to MAT2A again during liver malignant transformation. In addition, knockdown MAT2A dramatically suppresses tumour cell proliferation and induces cell cycle arrest and apoptosis 7,10,11 . Therefore, the transcriptional switch from MAT1A to MAT2A is believed to play a role in facilitating cancer cell survival and proliferation 11,12 .
MAT2A expression is regulated at both transcriptional and post-transcriptional levels. Sp1 (transcription factor: specificity protein 1), c-Myb (transcription activator: avian myeloblastosis viral oncogene), NF-kB (nuclear factor of kappa, enhancer in B cells) and AP-1 (adaptor-related protein complex 1) were identified as the trans-activating factors involved in MAT2A transcriptional upregulation 13,14 . Tumour-necrosis factor-a upregulates MAT2A via NF-kB and AP-1 (ref. 14). Moreover, histone acetylation, promoter methylation and mRNA stabilization were also reported to regulate MAT2A expression 15 . Although regulation of MAT2A expression was heavily studied at transcriptional and post-transcriptional levels, its post-translational regulation remains largely unknown.
Covalent lysine acetylation has been identified as an evolutionarily conserved modification in metabolic enzymes, and plays critical roles in regulation of multiple enzymes [16][17][18] . In this study, we demonstrate that MAT IIa protein is acetylated at lysine residue 81 in response to folate deprivation. Acetylation promotes MAT IIa ubiquitylation and its subsequent proteasomal degradation, inhibiting tumour cell proliferation. Our observations reveal a novel mechanism of MAT IIa upregulation in human cancers.

Results
MAT IIa is acetylated at lysine 81. Previous mass spectrometry (MS) analyses indicated that MAT IIa was potentially an acetylated protein ( Supplementary Fig. 1a). To confirm the acetylation modification, pan acetyl-lysine antibody was used to detect the acetylation level of ectopically expressed MAT IIa. Result showed that MAT IIa was indeed acetylated in HEK293T and Chang's cells (Fig. 1a). Furthermore, trichostatin A (TSA), an inhibitor of histone deacetylase HDAC family I, II and IV, increased the acetylation level of flag-MAT IIa approximately two-to threefold (Fig. 1a). Given that lysine (K) 81 is the only putative acetylation site identified in MAT IIa by MS ( Fig. 1b and Supplementary  Fig. 1a), and is evolutionarily conserved from D. melanogaster to mammals, we mutated K81 to arginine (R) and glutamine (Q) and found that both mutations resulted in a significant reduction in acetylation (Fig. 1c). Notably, TSA treatment dramatically increased the acetylation level of wild-type MAT IIa, but not K81R or K81Q mutant, indicating K81 might be the primary acetylation site of MAT IIa (Fig. 1c). To further confirm acetylation of K81, we generated a K81 site-specific antibody specifically targeting to acetylated K81 residue in MAT IIa (designated as 'K81Ac antibody' henceforth). Dot blotting assay was performed to characterize the specificity of this antibody and found that K81Ac antibody preferentially detected K81 acetylated, but not unmodified peptide (Fig. 1d). In addition, strong and specific signal of K81 acetylation was observed in ectopically expressed wild-type MAT IIa but not K81R mutant ( Supplementary  Fig. 1b). Further results showed that both MAT2A knockdown (Fig. 1e) and peptide blocking by K81 acetylated peptide (Fig. 1f) significantly reduced K81Ac signal, indicating a high specificity of this antibody to MAT IIa K81 acetylation. More importantly, with the help of this site-specific antibody, we detected enhanced K81 acetylation signal of endogenous MAT IIa in different cell lines upon TSA treatment (Fig. 1g). These results indicate that K81 is the major acetylation site of MAT IIa under the tested condition.
Folate decreases K81 acetylation and stabilizes MAT IIa. Folate's metabolic derivative tetrahydrofolate plays an essential role in transferring one-carbon units. Restoration of methionine from homocysteine in methionine cycle greatly depends on methyl-groups (one type of one-carbon units) presented from folate metabolism. Therefore, methionine cycle is tightly related to folate availability and one-carbon unit metabolism. As serine hydroxymethyl transferation and glycine cleavage are two major sources of one-carbon units, methionine cycle is also related to serine and glycine metabolisms. To determine whether K81 acetylation of MAT IIa is dynamically regulated by serine, glycine or folate in vivo, we cultured HEK293T cells in medium containing these metabolites with different concentrations and found that K81 acetylation level of MAT IIa increased as concentrations of folate decreased (Fig. 2a, upper panel). In contrast to this, serine and glycine showed no effect on MAT IIa K81 acetylation (data not shown). More interestingly, steady-state level of MAT IIa protein decreased as concentrations of folate decreased (Fig. 2a, lower panel). Similar results were obtained in human hepatocellular carcinoma cell line-Huh7 ( Supplementary  Fig. 2a). These results indicate a potential inverse correlation between MAT IIa acetylation and protein levels in response to folate concentration.
K81 acetylation promotes MAT IIa degradation. To determine whether downregulation of MAT IIa protein by folate occurs at transcriptional level, quantitative reverse transcription-PCR was performed. Results showed that folate treatment had no effect on MAT IIa mRNA level (Fig. 2b), indicating that folate regulates MAT IIa protein mainly at the post-transcriptional level. To determine whether acetylation induced MAT IIa degradation is mediated by proteasomal pathway, we treated HEK293T cells with proteasome inhibitor MG132 and found both TSA and folate deprivation induced decrease in MAT IIa protein can be readily restored by MG132 ( Fig. 2c and Supplementary Fig. 2b), suggesting K81 acetylation in response to TSA or folate deprivation might target MAT IIa protein to proteasomal degradation. The fact that MAT IIa was accumulated in HEK293T, H1299 and U937 cells after treated with MG132 further supported this speculation ( Supplementary Fig. 2c). In addition, TSA treatment combined with protein synthesis inhibitor cycloheximide (CHX) dramatically decreased the half-life of endogenous MAT IIa ( Supplementary Fig. 2d). These results indicate that TSA-and folate deprivation-induced MAT IIa protein degradation is proteasomal dependent. To further test the possibility that TSA or folate deprivation promotes MAT IIa degradation via K81 acetylation, half-lives of wild-type, K81R and K81Q MAT IIa were determined by CHX chase experiment. Wild-type MAT IIa, but not the mutants, was found to display a shortened half-life under folate-deprived condition (Fig. 2d), indicating proteasomal degradation of MAT IIa might be related to K81 acetylation. In addition, we employed ubiquitylation ladder assay and observed a significant increase in wild-type MAT IIa ubiquitylation in the presence of TSA or upon folate deprivation, whereas ubiquitylation of K81R mutation was mostly blocked and no longer respond to either TSA or folate deprivation (Fig. 2e,f). Consistently, we found MG132 could block TSA-or folate deprivation-induced MAT IIa destabilization ( Fig. 2g and Supplementary Fig. 2e). Collectively, these results demonstrate that K81 acetylation of MAT IIa promotes its degradation via ubiquitylation-mediated proteasomal degradation.
UBR4 targets MAT IIa for degradation. Three types of enzymes are required in the process of ubiquitylation: ubiquitin-activating enzymes (E1s), ubiquitin-conjugating enzymes (E2s) and ubiquitin-ligases (E3s), among which, only E3s define the target specificity of ubiquitylation reaction. To further investigate the mechanism of TSA-or folate deprivation-induced MAT IIa (c) TSA increases acetylation level of wild-type MAT IIa but not mutants. Ectopically expressed wild-type MAT IIa, K81R and K81Q were transfected into HEK293T cells followed by TSA treatment, MAT IIa acetylation was analysed by western blotting. (d) Specificity of K81 site-specific acetylation antibody was determined by dot blotting assay. (e) MAT2A knockdown confirmed the K81Ac antibody detection. HEK293T cells were transfected with MAT2A siRNA or scramble control and cell lysates were subjected to western blotting. Signal detected by K81Ac antibody decreased as MAT IIa was knocked down. (f) K81Ac antibody is specific to MAT IIa K81 acetylation. HEK293T cells were lysed and directly subjected to western blotting. The four repetitively loaded lanes were clipped into four membranes and were exposed to antibodies as indicated, separately. For peptide block, acetylated K81 peptide (the antigen we used for K81Ac antibody generation) or non-acetylated K81 peptide (mimicking MAT IIa protein) were added to the diluted K81Ac antibodies (1:200) in a final concentration of 300 mg ml -1 , cultivated at 37°C for 1 h before applying to membranes. Four clipped membranes were aligned and exposed together when developing film to make sure their relative signal intensities are comparable. (g) TSA promotes MAT IIa acetylation at K81 in different cell lines. HEK293T, Hela and HepG2 cells were treated as indicated. K81 acetylation was detected by K81Ac antibody and normalized against MAT IIa protein.
proteasomal degradation, we established an HEK293-derived cell line stably expressing both streptavidin-binding protein and flagtagged MAT IIa, cultured upon folate deprivation and purified MAT IIa protein using tandem affinity purification to identify its E3 ligase via MS. Among the consistently identified interacting proteins, we identified UBR4 (ubiquitin protein ligase E3 component n-recognin 4, belongs to the zinc finger E3 ligase family) as a putative E3 ligase for MAT IIa. To verify whether UBR4 is an E3 ligase for MAT IIa, we overexpressed UBR4 (D) (expressing a truncated form of UBR4 protein, with substrate-binding and catalytic domains only) in HEK293T cells and found a significant drop in MAT IIa protein, which was totally reversed by adding MG132 to the culture medium (Fig. 3a). Furthermore, knocking down UBR4 expression by three independent short interfering RNAs (siRNAs) could instantly block folate deprivation-induced MAT IIa degradation (Fig. 3b). Consistently, CHX chase experiment showed that UBR4 knockdown could totally block the decreased half-life of MAT IIa upon folate deprivation (Fig. 3c). Moreover, UBR4 knockdown increases MAT IIa protein level and its acetylation level ( Supplementary Fig. 3, left panel). The efficacy of UBR4 knockdown was validated by quantitative PCR (qPCR; Supplementary Fig. 3, right panel). In addition, we performed ubiquitin ladder experiment to visualize ubiquitylation. Consistent with previous findings, polyubiquitylation signal of MAT IIa was detected upon folate deprivation (Fig. 3d), whereas HAtagged ubiquitin could no longer be detected at high molecular weight after knocking down UBR4 expression (for quantification of UBR4 mRNA level, please refer to Fig. 3d). Collectively, these results verify UBR4 as an E3 ligase participating in polyubiquitylation-mediated proteasomal degradation of MAT IIa.
To further verify whether P300 acetylates MAT IIa at K81, we designed experiment shown in Supplementary Fig. 4, and proved that P300 overexpression only increases acetylation of wild-type MAT IIa, but not the K81R/Q mutant. Conversely, knocking down P300 by RNA interference significantly reduced K81 acetylation of endogenous MAT IIa (Fig. 4c). P300 decreased MAT IIa protein level in a dosedependent manner (Fig. 4d), whereas P300 knockdown rescued folate deprivation-induced MAT IIa destabilization (Fig. 4e). Moreover, we found folate-deprivation increased the interaction between endogenous MAT IIa and P300 (Fig. 4f), and this interaction was further verified by glutathione S-transferase (GST) pull-down (Fig. 5f). To further examine the hypothesis that P300 is the acetyl-transferase of MAT IIa, in vitro acetylation assay was conducted. As shown in Fig. 4g, recombinant P300 instantly acetylated MAT IIa at K81. Collectively, our results demonstrate that P300 is the acetyl-transferase responsible for MAT IIa acetylation at K81.
HDAC3 deacetylates MAT IIa at K81. According to the previous results that TSA, an inhibitor of type I, II and IV HDACs, increased K81 acetylation of MAT IIa, we postulated that HDACs might be involved in MAT IIa deacetylation. We co-expressed HDAC1-7 individually with MAT IIa, and found that overexpression of HDAC3, but not other tested HDACs, decreased the acetylation level of ectopically expressed MAT IIa ( Fig. 5a and Supplementary Fig. 5). Thereafter, we overexpressed HDAC3 in HEK293T cells, and found a decrease in K81 acetylation of endogenous MAT IIa (Fig. 5b). Notably, HDAC3 knockdown decreased the steady protein level of MAT IIa (Fig. 5c). Furthermore, we found HDAC3 knockdown led to the destabilization of MAT IIa (Fig. 5d), further supported that deacetylation stabilizes MAT IIa. Previous study by Katoh et al. identified HDAC3 as a putative MAT IIa-interacting protein by MS analysis 19 . To verify this interaction between HDAC3 and MAT IIa, we performed co-immunoprecipitation (co-IP) experiment and found that interaction indeed exists between these two proteins, and the interaction could be further enhanced in the presence of folate (Fig. 5e). Similarly, GST pull-down was performed to further verify this interaction (Fig. 5f). Taken together, HDAC3 is the deacetylase for MAT IIa K81 acetylation.

K81 mutants promote tumour growth in vitro and in vivo.
Given MAT IIa overexpression was previously implicated to promote cancer cell proliferation and cancer progression 7,15 , we next examined the effect of MAT IIa acetylation on cell proliferation and tumour growth. To this end, we generated stable cell lines in HepG2 cells in which endogenous MAT IIa was knocked down by short hairpin RNA (shRNA; targeting 3 0untranslated region of MAT2A gene) and the shRNA-resistant MAT IIa wild-type, K81Q or K81R (each contains only coding sequence of wild-type or mutated MAT2A gene) was stably expressed. Western blotting analysis demonstrated that the endogenous MAT IIa was effectively knocked down and that wild-type, K81Q or K81R mutants were expressed at levels similar to that of the endogenous MAT IIa (Fig. 6a). Direct cell count and MTT assays were both applied to monitor cell proliferation velocities of wild-type or K81 mutant stable cells under normal or folate-deprived conditions. MAT2A knockdown instantly caused growth arrest in HepG2 stable cell line ( Supplementary Fig. 6a). HepG2 cells expressing K81R and K81Q mutant stable cells were found proliferating significantly faster than cells expressing wildtype MAT IIa upon folate deprivation, whereas under normal condition, no significant difference in proliferating velocity was detected ( Fig. 6b and Supplementary Fig. 6b). As MAT IIa catalyses the production of main methyl donor SAM, and SAM is converted to SAH (S-adenosyl-homocysteine) after the methyltransferation reaction, we then investigated SAM/SAH  Fig. 6c). As a result, global methylation in wild-type-expressing cells cultured upon folate deprivation was found decreased, compared with that of the mutant-expressing cells grew under the same condition ( Supplementary Fig. 6d).
To determine folate's effect on MAT IIa K81 acetylation in tumour growth, we performed xenograft experiments using HepG2 stable cell lines referred previously. Nude mice were subcutaneously injected with HepG2 stable cells and fed with folate-free diet. Tumour growth was continuously monitored thereafter. Measurement of the widths and lengths of tumours demonstrated that HepG2 cells expressing K81R or K81Q  mutants displayed a growth rate significantly faster than those expressing wild-type MAT IIa upon folate-deprived dietary ( Fig. 6c and Supplementary Fig. 6e). At the time of harvesting, mice injected with cells expressing K81R/K81Q mutants developed larger tumours than those injected with cells expressing wild-type MAT IIa (Fig. 6d). In addition, levels of MAT IIa protein in stable cell lines and xenograft tumours derived from stable cell lines were assessed by western blotting, protein levels of K81R and K81Q mutants were found indeed higher than that of the wild-type MAT IIa (Fig. 6e). Collectively, K81 mutants stabilize MAT IIa and promote tumour cell growth upon folatedeprivation both in vitro and in vivo. interaction. HEK293T cells were transfected with flag-tagged MAT IIa and cultured in normal or folate-deprived conditions. The interaction between flag-MAT IIa and endogenous P300 was determined by co-IP followed by western blotting. (g) P300 acetylates MAT IIa at K81 in vitro. Ectopically expressed GST-MAT IIa from BL21 E. coli was purified through GST antibody-conjugated beads and incubated with P300 (immunoprecipitation-purified and eluted). K81 acetylation levels were analysed using K81Ac antibody and normalized against GST-MAT IIa.  Fig. 6f and Supplementary Fig. 6f). To further exam the correlation between K81 acetylation and MAT IIa protein, we determined MAT IIa K81 acetylation in these 40 pairs of samples, and found that 32 pairs of samples showed relatively higher levels of K81 acetylation in the tumour tissues than that of the matched normal tissues (Fig. 6f and Supplementary Fig. 6g). Furthermore, we found that HDAC3 expression was increased in 19 out 32 HCC samples (Supplementary Fig. 6f). The tumour sample analyses demonstrate that MAT IIa protein levels are elevated in hepatocellular cancers, and K81 acetylation inversely correlates with the elevated MAT IIa protein. These data also indicate an application prospect of using MAT IIa and K81 acetylation as potential biomarkers for hepatocellular cancer diagnosis.

Discussion
Dysregulation of cellular metabolism is a hallmark of cancer. Besides elevated glycolysis, increased lipogenesis and so on, dysregulation of amino-acid metabolism is also implicated in cancer. Imbalanced methionine metabolism in human liver cancer has been known for at least 60 years. Liver can be considered as the body's SAM factory, and MAT is the key enzyme catalysing biosynthesis of SAM. Overexpression of MAT2A is common in various types of cancer 6,7,15 , whereas knocking down MAT2A gene expression causes cell cycle arrest and apoptosis in cancer cells 10,11 . Our study reveals a novel mechanism of MAT IIa regulation by acetylation and ubiquitylation in response to folate concentration (Fig. 6g) and provides a new mechanistic insight in MAT IIa upregulation in cancer.
Upregulation of MAT IIa in cancer is in part due to MAT2A overexpression regulated by alerted gene transcription or mRNA stability [13][14][15] . In this study, we report two post-translational modifications of MAT IIa. We propose that MAT IIa acetylation modulated by folate concentration and consequent ubiquitylation play crucial roles in MAT IIa stability control. When folate is sufficient, MAT IIa is stabilized via HDAC3 deacetylation and subsequent inhibition of ubiquitylation. Accumulation of MAT IIa therefore enhances tumour growth. Conversely, when folate is deprived, P300 acetylates MAT IIa at K81. Acetylation therefore promotes UBR4-mediated MAT IIa ubiquitylation, leading to its proteasomal degradation and reduced cell growth.
Folate is needed in large quantities in rapidly proliferating cells. Folate uptake is increased in several types of solid human cancers including ovarian, endometrial, colorectal, kidney, lung and breast carcinomas [20][21][22] . Anti-folate drugs have long been approved for treating cancers 23 . Therefore, one may expect an important role for folate in MAT IIa regulation.
Indeed, folate-induced MAT IIa stabilization results in an increase in SAM production ( Supplementary Fig. 6c). We speculate that SAM accumulation would possibly cause dysregulations of oncogenes and tumour-suppressor genes via affecting global methylation, eventually leading to tumorigenesis and cancer development. In addition, the fact that stabilization of MAT IIa by K81 mutations significantly promotes cancer cell proliferation and in vivo tumour growth indicates a critical role of MAT IIa acetylation in coordinating folate availability and the regulation of cell growth and tumorigenesis. It also suggests a possibility that promoting MAT IIa acetylation may merit exploration as a target for cancer therapy. Notably, the observation that MAT IIa is overexpressed in hepatocellular cancer tissues, whereas K81 acetylation is significantly decreased indicates a pivotal role of MAT IIa acetylation regulation in hepatocellular cancer development.

Methods
Cell culture and treatment. Cells were all purchased from American Type Culture Collection and cultured in DMEM/high-glucose medium (HyClone) supplemented with 10% fetal bovine serum (HyClone), 1% penicillin and streptomycin at 37°C, in a humidified atmosphere containing 5% CO 2 . Standard RPMI-1640 and RPMI-1640 without folate were purchased from Life Technologies, Inc. (GIBCO). For folate-deprivation treatment, cells were cultured in the medium containing different amounts of folate as indicated. For TSA, MG132, CHX treatment, please see descriptions in figure legends, respectively.
Ubiquitin ladder assay. Ubiquitin ladder assay was performed as previously described 24 . Thirty-six hours after transfection, cells were collected and lysed in 1% SDS buffer (50 mM Tris-HCl (pH 7.5), 0.5 mM EDTA, 1 mM dithiothreitol) with inhibitors and boiled for 10 min. Before immunoprecipitation, lysates were diluted ten-fold with 0.3% Nonidet P40 buffer. Ubiquitylation were determined by western blotting.
Knocking down and putting back stable cell lines. Flag-tagged human wild type, K81R and K81Q mutants of MAT2A were cloned into the retroviral vector (pQCXIH) and were co-transfected into HEK293T cells together with vectors expressing gag and vsvg genes (from vesicular stomatitis virus G). Retroviral supernatants were harvested by filtration (0.45 mm filters) 48 h after initial plasmid transfection and mixed with polybrene (8 mg ml -1 as final concentration) to increase the infection efficiency before applied to HepG2 cells. HepG2 cells were infected with the prepared virus for 12 h and screened by hygromycin (350 mg ml -1 as final concentration) for at least 2 weeks. pMKO-shMAT2A and pMKO-shVEC were Knockdown efficiency and reexpression levels of wild-type or K81R/Q mutants were determined by western blotting. (b) K81R and K81Q mutations reverse the proliferative disadvantage of HepG2 cells upon folate deprivation. HepG2 stable cell characterized in a were cultured in normal or folate-deprived medium. Direct cell count was performed every 24 h after seeding. Error bars represent cell numbers ± s.d. for triplicate experiments. The two-tailed Student's t-test was used. **Po0.01; ***Po0.001. (c,d) K81R and K81Q mutants promote xenograft tumour growth. Subcutaneous xenograft experiment was performed in nude mice using HepG2 stable cells. Major and minor diameters of tumours were measured and tumour volumes were calculated. The two-tailed Student's t-test was used. *Po0.05; **Po0.01; ***Po0.001; NS denotes no significance (c). 25 days after injection, tumours were dissected, photographed and weighted. The two-tailed Student's t-test was used. **Po0.01; NS denotes no significance (d). (e) The expression of wild-type, K81R and K81Q MAT IIa in xenografts. Whole-cell lysates were prepared from either original HepG2 stable cell lines or xenograft tumours, followed by western blotting analysis. (f) The hepatocellular cancer clinical samples show an inverse correlation between MAT IIa protein and K81 acetylation. Human hepatocellular cancer samples each paired with cancerous tissue (designated as C) and adjacent normal tissue (designated as N) were lysed and directly subjected to western blotting. Only eight pairs of samples showcasing inverse correlation are shown. For more samples, please refer to supplementary Fig. 6f. (g) Working model. By enhancing MAT IIa association with P300 and UBR4, folate deprivation promotes both MAT IIa K81 acetylation and ubiquitylation, resulting in its proteasomal degradation. Folate stabilizes MAT IIa by dissociating MAT IIa-P300-UBR4 complex, and promoting MAT IIa-HDAC3 interaction. Accumulated MAT II a protein facilitates rapid cell proliferation and, therefore, promotes tumour growth.
GST pull-down assay. BL21 E. coli transformed with pGEX-GST-MAT IIa plasmid was induced (or not induced) by isopropyl-b-D-thiogalactoside (0.1 mM as final concentration) at 20°C for 12 h. Protein was then purified through GST antibody-conjugated beads. Beads were then added to HEK293T cell lysate (lysed by 0.3% Nonidet P40 buffer), and mixed at 4°C overnight. Non-induced group was used as a negative control. Beads were subsequently harvested through centrifugation and washed four times by 0.3% Nonidet P40 buffer before boiled by 1 Â SDS-polyacrylamide gel electrophoresis loading buffer and subjected to western blotting.
In vitro acetylation assay. Ectopically expressed MAT IIa from BL21 E. coli was purified through GST antibody-conjugated beads and incubated with P300 (immunoprecipitation-purified and eluted) in histone acetyl-transferase (HAT) assay buffer (purchased from Millipore), supplemented with acetyl-co-enzyme A (10 mM as final concentration). Reaction was conducted under 30°C for 1 h with continuous shaking. After adding 5 Â SDS-polyacrylamide gel electrophoresis loading buffer and boiled, samples were analysed by western blotting.
Cell proliferation analysis. 5 Â 10 4 HepG2 stable cells were seeded in triplicate in each well of six-well plate, and cultured in normal or folate-deprived medium. Cell numbers were counted every day over a 5-day period via haemacytometer counting method. For MTT (3-(4,5-dimethyl-2-thiazolyl-2,5-diphenyl-2-H-tetrazolium bromide)) assays, 2 Â 10 3 HepG2 stable cells were seeded in quintuplicate in each well of 96-well plate. Standard methods were performed to track the cell proliferation.
Xenograft analysis. The procedures related to animal studies were approved by the Ethics Committee of the Institutes of Biomedical Sciences, Fudan University. Rodent diet without folate was purchased from Research Diets, Inc. Nude mice (nu/nu, six-week-old males) were fed with folate-deprived diet for 1 week, followed by injected subcutaneously with 5 Â 10 6 HepG2 stable cells. The mice were subsequently fed with folate-deprived diet. Major and minor diameters of tumours were measured every 2 days. Mice were killed after 25 days and tumours were collected and weighed.
Hepatocarcinoma cancer samples. Hepatocarcinoma cancer samples were obtained from the 10th People's Hospital, Shanghai (Tongji University Affiliated), with written consents from all investigated patients. The procedures related to human subjects were approved by the Ethics Committee of the Institutes of Biomedical Sciences, Fudan University. Direct immunoblotting was performed as mentioned above.
SAM and SAH quantification. SAM and SAH quantification were performed as previously described 25 . Cells were harvested and weighed. The cell pellets were added with 0.4 M perchloric acid, mixed vigorously and centrifuged. Supernatants were adjusted to pH 5-7 with 2.5 M K 2 HPO 4 and kept on ice for 15 min to precipitate potassium perchlorate. Samples were centrifuged twice and supernatants were analysed by liquid chromatography-tandem mass spectrometry.
Methylated DNA quantification. Genomic DNA was prepared using a DNA isolation kit (Beyotime). A methylated DNA quantification kit (Colorimetric; Abnova) was used to detect methylated DNA.