Dietary methionine promotes HCC by proteome remodeling

Dietary methionine has a remarkable effect on cancer progression [31, 32], but its effect on the HCC progression remains further explored. We first considered patient-derived xenograft (PDX) models of liver cancer for 2 weeks, then mice were randomized into two groups when the tumour was palpable, with one group fed a standard diet (0.86% methionine, w/w) and the other fed a methionine-restricted diet (0.12% methionine, w/w) (Fig. 1A). Methionine restriction (MR) significantly inhibited tumour growth and prolonged mice lives (Fig. 1B, C), which was also observed in another disease model induced by N-nitrosodiethylamine (DEN) and carbon tetrachloride (CCl4) (Additional file 1: Fig. S1A–C). Note that the inhibitory effect was not because of caloric restriction as similar amounts of food intake were observed for both groups.

Fig. 1

The effect of dietary methionine restriction on mTOR signaling and cancer progression. A Schematic of experimental design using HCC PDX model. Treatment, n = 5 mice per group. B Tumour growth curves and images of tumours at the end point (Day 28). C Survival analysis of mice under Ctrl or MR conditions. D Heatmap displaying global transcriptional changes of liver tissues between Ctrl and MR described in A. Red color represents the genes’ expression upregulated, while blue means the genes’ expression decreased. E KEGG pathway enriched for up-regulated and down regulated genes of liver tissues by comparing Ctrl to MR conditions. F Migration of SNU449 cells (left panel) with quantification data on the right panel. G Sphere formation efficiency of SNU449 cells (left panel) with quantification (right panel). H, I Tumour volumes and survival analysis of mouse allografts. n = 5 mice per group. Data are means ± SEM. Group differences were analyzed by two-tailed Student’s t test (B, C), two-way ANOVA followed by Tukey’s multiple comparison test (F, G, H) or log-rank test (I) (**p < 0.01, ****p < 0.0001). n.s., not significant. Ctrl, Control and MR, Methionine restriction

This intriguing phenomenon promoted us to study what might cause this striking difference during methionine restriction compared to the control set. Therefore, we determined the consequences of MR on gene expression using RNA sequencing (RNA-seq) in the liver. As shown in Fig. 1D, MR greatly remodeled the gene expression in the liver. In the control set, 327 genes out of 30,058 detected genes were downregulated. These genes were involved in steroid hormone biosynthesis, retinol metabolism, and other metabolic processes, suggesting that MR could stimulate specific metabolic pathways (Fig. 1E). Interestingly, 1277 genes were significantly upregulated in the control diet (p-value < 0.05, fold change > 2) (Fig. 1D and Additional file 2: Table S1A). These genes are mainly involved in the cytokine-cytokine receptor interaction, inflammatory bowel disease and several signaling pathways, including mTOR signaling pathway. The elevated expression of canonical mTOR target genes, including c-Myc, fzd9, prkcb and wnt9b, was confirmed by reverse transcription-quantitative PCR (qRT-PCR) (Additional file 1: Fig. S1D). The upregulation of protein c-Myc was further confirmed by western blotting (Additional file 1: Fig. S1E). We also found that SIRT4 downregulation, TRIM32 upregulation and unchanged level of MAT2A (Additional file 1: Fig. S1E) in the Ctrl liver samples compared to that in MR. These results consistent with the prevailing idea that mTORC1 is a central hub of nutrient signaling to coordinate cellular metabolism and growth in response to nutrient availability [13, 33] (Fig. 1E). We hypothesized that mTORC1 might play a role in this striking difference between dietary restriction and the control set. Indeed, immunohistochemical staining revealed that the mTORC1 signaling was activated in the control set, as indicated by higher phosphorylation level of p70 S6 Kinase, a putative mTORC1 substrate (Additional file 1: Fig. S1F). To further confirm our hypothesis, we sought to investigate the effect of rapamycin, an inhibitor of mTORC1, on the HCC development. Rapamycin intensively diminished the cellular growth rate, migration, regeneration and tumorigenesis compared to the control set, and the effect was further augmented by methionine restriction (Fig. 1F–I and Additional file 1: Fig. S1G, H). These results indicated the essential role of mTORC1 in initiating HCC tumorigenesis in response to the availability of methionine.

mTOR signaling is activated by dietary methionine for the HCC progression

Next, we would like to investigate how the cancer cells alter gene expression to maintain fast proliferation rate after tumour initiation by mTORC1 signaling. Since transcription factors are proteins involved in the process of converting DNA to RNA, we found that 61 transcription factors were upregulated while only 18 were downregulated (Additional file 2: Table S1B). c-Myc, Jun, Spi1, Ets1, Atf3 and Esr1 plays central hubs in the transcription networks with the first 4 proteins being recognized as oncogenes in the Uniprot database (Fig. 2A). Interestingly, only c-Myc acts downstream of mTORC1 signaling [15], therefore, we explored whether the mTORC1 signaling is responsible for the robust c-Myc expression in a methionine dependent manner. Methionine withdraw resulted in a strong reduction in c-Myc expression and the phosphorylation level of p70 S6 Kinase in HEK293 cells, a signal indicative of mTORC1 inhibition (Fig. 2B and Additional file 1: Fig. S2A). However, rapamycin treatment remained the c-Myc expression at low level despite of increased methionine. We also observed similar results in the SNU449 cells (Fig. 2B and Additional file 1: Fig. S2A). These results together demonstrate that c-Myc expression is dependent on mTORC1 signaling in response to methionine availability. The activation of oncogenic gene c-Myc by mTORC1 may facilitate c-Myc mediated transcription, thereby providing metabolic demanding for the cancer progression.

Fig. 2

mTORC1 activates c-Myc to promote HCC tumorigenesis. A String interaction of differentially expressed transcription factors upon methionine restriction. B Western blot analysis of indicated proteins in the presence or absence of rapamycin (20 nM) when HEK293 or SNU449 cells were grown in Medium with or without methionine. C, D Migration abilities of primary liver cells derived from c-Myc+/+ and c-Myc−/− mice under control or methionine restriction conditions. Quantification was shown in the right panel. D Sphere formation efficiency of primary liver cells derived from c-Myc+/+ and c-Myc−/− mice under control or methionine restriction conditions. Quantification was displayed in the right panel. E Immunoblot showing c-Myc protein level in the liver lysates of c-Myc+/+ and c-Myc−/− mice. F, G Tumour volumes and survival analysis of c-Myc+/+ and c-Myc−/− mice. n = 6 mice per group. Data are means ± SEM. Group differences were analyzed by two-way ANOVA followed by Tukey’s multiple comparison test (C, D, F) or log-rank test (G) (**p < 0.01, ****p < 0.0001). n.s., not significant

To explore the role of c-Myc in the HCC tumorigenicity during methionine restriction, we then expressed doxycycline (DOX) inducible c-Myc shRNA in primary liver cells (Additional file 1: Fig. S2B, C). The inducible knockdown reduced the proliferation rate of these cells in the standard diet to a similar extent as that of methionine restriction (Additional file 1: Fig. S2D). We also inoculated these cells into nude mice. When tumours reached ~ 65 mm3 in volume, experimental mice groups were subjected to DOX with or without MR in the diet. The induction of c-Myc silencing by DOX substantially attenuated the growth of the tumours, almost similar to the MR set (Additional file 1: Fig. S2E). Indeed, c-Myc ablation can significantly reduce migration and sphere formation abilities of HCC cells (Fig. 3D, E), and reverse DEN-induced HCC tumorigenesis and prolonged the mouse lives (Fig. 3F, G). These results together indicated that mTORC1 signaling promotes HCC tumorigenesis through c-Myc activation.

Fig. 3

c-Myc promotes TRIM32 mediated proteasomal degradation of SIRT4. A Western blot analysis of SIRT4 and c-Myc expression in the whole cellular lysates treated with c-Myc shRNA or MG132 for 48 h. B Western blot analysis of TRIM32 and c-Myc expression in whole cellular lysates treated with c-Myc shRNA. C SIRT4 interacts with TRIM32 in vivo. SNU449/HEK293 cell lysates were immunoprecipitated (IP) with control IgG, anti-SIRT4 or anti-TRIM32 antibodies, and then the precipitated proteins were detected by anti-TRIM32 or anti-SIRT4 antibodies, respectively. D SIRT4 degradation curves under various conditions. TRIM32 was knockdown with shRNA or forced overexpressed in SNU449 cells with or without being treated by proteasome inhibitors MG132 or lactacystin. After being treated with protein synthesis inhibitor cycloheximide (CHX; 50 μg/ml), the above cells were collected at the indicated hours. SIRT4 protein levels were detected by immunoblotting in the SNU449 total cell lysates. E SNU449 cells expressing FLAG-SIRT4 and co-expressing with or without hemagglutinin-tagged ubiquitin (HA-Ub) were co-transfected c-Myc and TRIM32 shRNA (left) or co-transfected c-Myc shRNA and TRIM32 (right) for 48 h. After being treated with 5 µM MG132 for 8 h, cells were collected and IP was performed by using FLAG antibody. Polyubiquitination of FLAG-SIRT4 were detected by anti HA-Ub antibody. F The indicated lysine sites of SIRT4 were mutated into arginine. G Lysates from SNU449 cells expressing the above SITR4 site mutants and co-expressing with or without hemagglutinin-tagged ubiquitin (HA-Ub) were pulled down with anti-FLAG, and then immunoblotted with anti HA-Ub antibody. Group differences were analyzed by two-tailed Student’s t test (D) (*p < 0.05, **p < 0.01)

c-Myc promotes TRIM32-mediated proteasomal degradation of SIRT4

Our previous finding revealed that SIRT4 ablation activated mTOR activity and increased its downstream gene c-Myc while SIRT4 expression altered mTOR phosphorylation and c-Myc gene expression in an opposite way [23]. However, the vice verse effect has not been investigated yet. Compared to that of the MR set, the lower level of SIRT4 in the control set (Additional file 1: Fig. S3A, B) also indicated a possible link between mTORC1-c-Myc pathway and SIRT4 in the HCC tumorigenesis.

We found that SIRT4 at the transcriptome remained almost unchanged (Additional file 2: Table S1A). Therefore, we deliberated whether c-Myc might play a role in repressing SIRT4 expression at the protein level. Interestingly, knockdown of c-Myc significantly increased SIRT4 expression, and the suppressive effect of c-Myc on SIRT4 expression was attenuated by treating the cells with proteome inhibitor MG132 (Fig. 3A and Additional file 1: Fig. S4A). These results implied that c-Myc promoted proteasome-mediated SIRT4 degradation, most probably through regulating E3 ligase gene expression. To identify the potential E3 ligase that might be responsible for SIRT4 turnover, we analyzed a known SIRT4 interactome [22]. We found that TRIM32 as a putative candidate (Additional file 1: Fig. S4B), as c-Myc knockdown in different cell lines reduced TRIM32 expression level (Fig. 3B and Additional file 1: Fig. S4C). Western-blot analysis confirmed the interaction between SIRT4 and TRIM32 (Fig. 3C).

Cycloheximide (CHX) chase assays further revealed the positive role of TRIM32 in SIRT4 stability. Forced overexpression of TRIM32 accelerated the degradation rate of SIRT4, while TRIM32 knockdown strongly stabilized SIRT4, however, the effect of TRIM32 on SIRT4 degradation was abolished by treating the cells with proteasome inhibitor MG132 or lactacystin (Fig. 3D and Additional file 1: Fig. S4D). To assess whether TRIM32 ubiquitinates SIRT4, we purified SIRT4 in SNU449 cells that were co-transfected with FLAG-SIRT4, hemagglutinin-tagged ubiquitin (HA-Ub), c-Myc/TRIM32 shRNA (left) or c-Myc shRNA/TRIM32 (right) (Fig. 3E). Western blot analysis of SIRT4 immunoprecipitation showed that c-Myc overexpression enhanced ubiquitination of SIRT4, which was attenuated by TRIM32 knockdown (Fig. 3E, left panel and Additional file 1: Fig. S4E). In contrast, TRIM32 overexpression blunted the inhibitory effect of c-Myc silence on SIRT4 ubiquitination (Fig. 3E, right panel and Additional file 1: Fig. S4F). Point mutation of SIRT4 (lysine to arginine) showed that ubiquitination sites of SIRT4 are in the lysines 78 and 229, with the former as the main site (Fig. 3F, G and Additional file 1: Fig. S4G). Moreover, we did co-immunoprecipitation of TRIM32 in cells where truncated SIRT4 with different length was expressed. WB analysis showed that TRIM32 was strongly bound to SIRT4 with length 1–314 or 45–314, whereas TRIM32 interacted weakly to SIRT4 with length 100–314 or 150–314 (Additional file 1: Fig. S5A). Indeed, SIRT4 K78R mutant reduced association with TRIM32 while SIRT4 K299R mutant remained bound to TRIM32 (Additional file 1: Fig. S5B). Furthermore, TRIM32 overexpression had less effect on the expression of SIRT4 K78R mutant than that of SIRT4 K299R mutant (Additional file 1: Fig. S5C). As lysine 78 is located at NAD+ binding region, SIRT4 K78R mutant also reduced cell proliferation rate (Additional file 1: Fig. S5D). Taken together, these results provide convincing evidences that c-Myc promotes SIRT4 degradation by mediating the expression of the ubiquitinase TRIM32.

SIRT4 attenuates c-Myc-promoted tumorigenesis by altering methionine metabolism

Given that c-Myc negatively regulates SIRT4 and SIRT4 has been reported as a cancer suppressor [23], we next explored whether there is any interplay between c-Myc and SIRT4 for the cancer progression. Interestingly, SIRT4 overexpression strikingly suppressed c-Myc induced cell proliferation and stemness (Additional file 1: Fig. S6A-C), a strong hint that SIRT4 retarded c-Myc mediated tumorigenesis through an unknown mechanism.

To dissect the possible role of MR in the metabolism, we analyzed the metabolites changes upon methionine restriction in vivo (Fig. 4A). MR profoundly rewired the metabolism in mouse liver. Of 744 metabolites identified by liquid chromatography-tandem mass spectrometry (LC–MS/MS), 51 metabolites were significantly increased and 243 were decreased (p < 0.05) (Additional file 3: Table S2), with pathways enriched in histidine metabolism, thiamine metabolism and cysteine and methionine metabolism (Fig. 4A, B). We checked the metabolites related with methionine metabolism and found that SAM and S-adenosyl-homocysteine (SAH) declined in the MR set (Fig. 4C, D). To explore the possible relationship between SIRT4 and methionine metabolism, we measured the relative levels of metabolites related with methionine metabolism by comparing SIRT4 overexpression to the control cells. To our surprise, SAM, SAH, homocysteine from methionine metabolism was greatly reduced while other one-carbon cycle related metabolites remained almost unchanged in the SIRT4 overexpression cells (Additional file 1: Fig. S7A). Indeed, SIRT4 overexpression strongly reduced histone trimethylation (Additional file 1: Fig. S7B,C), which has important roles in defining chromatin states, most notably at active genes. Histone methylation are regulated by balanced metabolite levels in the one-carbon and methionine metabolism, for example, methionine, serine and folate [34]. These metabolites contribute epigenetic regulation by providing one-carbon units. Interestingly, only SAM supplementation resorted the level of histone methylation (Fig. 4E and Additional file 1: Fig. S7D).

Fig. 4

Effect of SIRT4 overexpression on methionine cycle and HCC tumorigenesis. A Heatmap displaying metabolite changes upon methionine restriction (MR). The red color indicates detected metabolites were increased while the blue ones decreased between MR and Ctrl conditions in six biological replicates. B Pathway enrichment analysis of differentially detected metabolites (fold change > 1.5, p-value < 0.05) upon methionine restriction in MetaboAnalyst 4.0. The circle color is indicative of the level of enrichment significance, with yellow being low and red being high. The circle size is proportional to the pathway impact value. C Schematic of the methionine cycle. D The changes of specific metabolites upon methionine restriction. Blue means increased while red indicates decreased. E Western blot analysis of modified histones in the cell culture supplemented with specific metabolites. Total histone H3 was used as a loading control. F, G Migration and sphere formation efficiency of primary liver cells derived from SIRT4+/+ and SIRT4−/− mice under control or methionine restriction conditions. Quantification was shown in the right panel. H Immunoblot showing SIRT4 protein level in the liver lysates of SIRT4+/+ and SIRT4−/− mice. I, J Tumour volumes and survival analysis of SIRT4+/+ and SIRT4−/− mice. n = 5 mice per group. Data are means ± SEM. Group differences were analyzed by two-way ANOVA followed by Tukey’s multiple comparison test (F, G, I) or log-rank test (J) (**p < 0.01, ***p < 0.001 ****p < 0.0001). n.s., not significant

These results above indicate that SIRT4 plays a key role in the methionine metabolism and inhibits SAM production. Therefore, it was of interest to examine the role of SIRT4 in the HCC in vitro and in vivo. SIRT4 deletion dramatically increased migration, and sphere formation abilities of primary liver cells, and the effect was more significant in the control set (Fig. 4F, G). Consistent with in vitro observations, SIRT4 ablation in mice displayed a more remarkable effect in the promotion of tumour formation in the control set and significantly decreased the life of these mice (Fig. 4I-J). In contrast, SIRT4 overexpression with combination of MR treatment greatly reduced the cell proliferation, migration, sphere formation abilities and tumour size, and prolonged mouse survival (Additional file 1: Fig. S8A-F). Moreover, mice of inducible SIRT4 depletion in the control set showed strikingly increased tumorigenesis and shortened the mice’s lives (Additional file 1: Fig. S8G-I). Collectively, these findings demonstrated that SIRT4 suppress HCC progression by regulating methionine metabolism.

SIRT4 mediates mono-ADP-ribosylation of MAT2A

We further sought to investigate how SIRT4 regulates methionine metabolism. We first checked the transcriptome changes of genes related with methionine metabolism, and found that MAT1A significantly decreased while MAT2A and MAT2B remained almost unchanged (Fig. 5A and Additional file 1: Table S1A). Therefore, we speculated that SIRT4 might modulate the methionine metabolism by regulating post-translational modifications of MAT2A/MAT2B. As reported in published studies, SIRT4 potentially interacted with MAT2A [21, 22]. Western blot analysis confirmed that SIRT4 was physically associated with MAT2A (Additional file 1: Fig. S9A), which is a mono-ADP-ribosylation protein (MARylation) [35]. To characterize the role of SIRT4 as an ADP-ribosyltransferase, we investigated whether altered expressed SIRT4 affect the MARylation level of MAT2A. Specifically, SIRT4 knockdown markedly reduced MARylation level of MAT2A (Fig. 5B and Additional file 1: Fig. S9B), while SIRT4 overexpression strongly increased MAT2A MARylation (Fig. 5C and Additional file 1: Fig. S9C). ADP-ribosyltransferase activity is inhibited by two structurally distinct inhibitors of sirtuins, nicotinamide (NAM) and sirtinol [36]. Inhibition of SIRT4 activity by either NAM or sirtinol significantly deceased MARylation level of MAT2A (Fig. 5D, E and Additional file 1: Fig. S9D, E), further validating that MAT2A MARylation is catalyzed by SIRT4.

Fig. 5

SIRT4 regulates methionine metabolism through mediating MAT2A ADP ribosylation. A Volcano plot showing differential gene expression in mice’s livers upon methionine restriction. Red circles indicates genes increased in the Ctrl set while blue ones means genes decreased in the MR set. B Western blot analysis of MAT2A mono-ADP-ribosylation (MARylation) in HEK293 and SNU449 cells in which SIRT4 was silenced (shRNA). C The MARylation of MAT2A was measured by western blot in HEK293 or SNU449 cells stably expressing HA tagged SIRT4. D Western blot analysis of MAT2A MARylation in the FLAG immunoprecipitated samples from HEK293 or SNU449 cells treated with nicotinamide (NAM) for indicated times. E Western blot analysis of MAT2A MARylation in the FLAG immunoprecipitated samples from HEK293 or SNU449 cells treated with Sirtinol for indicated times. F Relative activity of MAT2A was shown in the SNU449 cells in which either c-Myc or SIRT4 was knockdown or overexpressed or c-Myc and SIRT4 overexpressed together. G Abundances of intracellular primary methionine cycle metabolites were compared between the control cells and the SNU449 cells where either c-Myc or SIRT4 was knockdown or stably overexpressed or c-Myc and SIRT4 overexpressed together. H MARylation level of mutant FLAG-MAT2A-E111A ectopically expressed in MAT2A-knockout HEK293 or SNU449 cells was measured by using anti-ADP ribosylation antibody when SIRT4 was silenced. I MARylation level of mutant FLAG-MAT2A-E111A ectopically expressed in MAT2A-knockout HEK293 or SNU449 cells was measured by using anti-ADP ribosylation antibody when SIRT4 was overexpressed. J MAT2A dimer are shown in blue or lightblue (PDB ID: 4NDN). The active site was shown in red circle. K E111 in MAT2A is conserved. Protein sequence alignment surrounding E111 colored in light blue from indicated species. Data are means ± SEM. Group differences were analyzed by two-tailed Student’s multiple comparison test (F, G) (**p < 0.01)

c-Myc negatively regulates SIRT4 expression. Indeed, c-Myc overexpression reduced the MARylation on MAT2A, while SIRT4 overexpression attenuated the inhibitory effect of c-Myc on MAT2A MARylation in different cell lines (Additional file 1: Fig. S9F, G). Furthermore, knockdown of c-Myc and SIRT4 alone or together had the vice versa effect (Additional file 1: Fig. S9F, G). Consistently, forced overexpression of c-Myc increased MAT2A activity, whereas SIRT4 overexpression suppressed the promoting effect of c-Myc on MAT2A activity (Fig. 5F). As a consequence, SIRT4 overexpression reduced SAM production (Fig. 5G). In contrast, c-Myc overexpression promoted elevation of SAM, while c-Myc knockdown reduced SAM level. As expected, SIRT4 overexpression diminished the promoting effect of c-Myc on SAM level (Fig. 5G). Taken together, these data demonstrate that c-Myc relieves MAT2A MARylation by promoting TRIM32 mediated SIRT4 degradation, which consequently enhances the role of MAT2A in the methionine metabolism.

We mutated each of eight putative MARylation sites individually into alanine (A) (Additional file 1: Fig. S10A) [37] and examined MARylation level. Mutation of glutamic acid 111 (E111A), but not other glutamic acid or aspartic acid residues to alanine, lead to a remarkable reduction in MAT2A MARylation (Additional file 1: Fig. S10B). Either SIRT4 knockdown or overexpression did not affect MARylation level of E111A mutant (Fig. 5H, I and Additional file 1: Fig. S10C, D). In addition, NAM or sirtinol treatment had almost no effect on the MARylation level and activity of MAT2A-E111A mutant (Additional file 1: Fig. S10E, F). These results indicate that E111 is a major MARylation site in MAT2A. Importantly, E111 is located in the substrate binding region and is in the vicinity of the gating-loop residue Q113, which directly contacts methionine and is proposed to play a critical role in positioning methionine for catalysis (Fig. 5J) [38, 39]. The modified ADP moiety is larger than methionine, thus, MARylation on E111 may prevent methionine from accessing the gating-loop residue Q113, resulting in low catalytic efficiency. Note that E111 is evolutionarily conserved from Saccharomyces cerevisiae to mammals (Fig. 5K).

MAT2A contributes to tumour progression in multiple types of cancers by promoting methionine metabolism [26, 27, 40,41,42]. Interestingly, SIRT4 and MAT2A co-expression leads to marked decrease in the cell proliferation (Additional file 1: Fig. S10G). In contrast, this effect was not observed in the MAT2A-E111A mutant cell line, as it abrogated SIRT4-induced suppression effect on cell proliferation (Additional file 1: Fig. S10G). Consistently, the tumour growth was strongly inhibited by SIRT4 overexpression but increased in the mice injected with MAT2A-E111A mutant (Additional file 1: Fig. S10H, I). Cellular treatment with FIDAS-5, an inhibitor of MAT2A [3], strongly affected MAT2A activity and its related histone methylation (Additional file 1: Fig. S11A-C), as well as cell proliferation and tumour growth (Additional file 1: Fig. S11D, E), similar to that observed in cells stably overexpressing SIRT4 (Additional file 1: Fig. S10G, H). These results demonstrate that SIRT4 exerts its role in the methionine metabolism through MARylation on MAT2A, thereby inhibiting cellular proliferation and tumour growth.

SIRT4 affects transcription through regulating methionine metabolism

Methionine is a key metabolic dependence for the tumour initiation and progression, as methionine metabolism determines levels of histone methylation by modulating SAM and SAH, which influences genomic architecture and subsequent gene expression [3, 31]. Therefore, we took an oncogenic transcription factor c-Myc as an example to investigate how SIRT4 affects gene transcription by modulating methionine metabolism.

SIRT4 overexpression greatly reduced levels of H3K4me3, H3K4me2 and H3K36me3, but not in the MAT2A-E111A mutant cells (Fig. 6A and Additional file 1: Fig. S12A). These results demonstrate that SIRT4 has an indirect effect on the histone methylation by inhibiting methionine metabolism through MAT2A MARylation. Then we checked the epigenetic change at c-Myc gene loci. Indeed, SIRT4 overexpression decreased H3K4me3 and H3K36me3 at the c-Myc promoter, while SIRT4 loss increased H3K4me3 and H3K36me3 at the c-Myc promoter (Additional file 1: Fig. S12B, C). SIRT4 overexpression also altered repressive histone methylation marks, in case of H3K9me2, H3K9me3, and H3K27me3 (Additional file 1: Fig. S12D). To further analyze the consequence of methionine rewiring on gene expression, we specifically checked c-Myc-targeted genes in MAT2A, MAT2A-E111A mutant and SIRT4 expressing SNU449 cells that were supplemented with or without SAM [43]. These genes play keys roles in the cancer metabolism, for example, ODC in the polyamine synthesis and HK2 in the glycolysis. Either MAT2A knockdown or SIRT4/MAT2A co-expression decreased the expression of c-Myc-targeted genes (Fig. 6B). However, SAM supplementation restored these genes’ expression (Fig. 6B), showing simultaneous recovery of their mRNA as well as histone methylation. Interestingly, SIRT4 did not have obvious effect in the MAT2A-E111A mutant cells (Fig. 6B). These results were further confirmed at the protein level (Fig. 6C and Additional file 1: Fig. S12E). These results all together show that SIRT4 has a key regulatory role in methionine metabolism to modulate gene expression.

Fig. 6

SIRT4 modulates gene expression by regulating methionine metabolism. A Western blot analysis of histone methylation level. Histone H3 was used as a loading control. B The mRNA levels of indicated genes were determined by quantitative real-time PCR (qRT-PCR) in