High NAT10 expression in gastric cancer is correlated with increased mRNA acetylation and poor patient prognosis
To explore the functional roles of mRNA modifications in GC, we first examined 9 nucleoside modifications in mRNA from 20 paired GC and adjacent normal gastric mucosa tissues via HPLC–MS/MS assay. We found a significant increase in ac4C on mRNA in gastric tumor tissues compared with that in adjacent normal tissues (Fig. 1A and Additional file 1: Fig. S1A). Based on the fact that the ac4C modification is catalyzed by NAT10 [16, 17], we sought to determine whether the elevated ac4C mRNA levels in GC are caused by altered expression of NAT10. Upon analysis of the TCGA data and GEO datasets, we observed that NAT10 was significantly upregulated in gastric and other tumor tissues relative to the corresponding normal tissues (Fig. 1B and Additional file 1: Fig. S1B). We then performed quantitative RT–PCR (qRT–PCR) analyses for 20 matched pairs of GCs and adjacent normal tissues and observed upregulation of NAT10 in GCs compared to the corresponding normal tissues (Fig. 1C). Moreover, increased NAT10 expression correlated with elevated ac4C levels in GC tissues (Fig. 1D). Subsequent Western blot analysis validated the upregulation of NAT10 in GC tissues and cell lines at the protein level (Fig. 1E and Additional file 1: Fig. S2A). Immunohistochemistry (IHC) of GC tissue microarrays further validated the elevated NAT10 expression in GC samples (Fig. 1F and Additional file 1: Fig. S1C). More importantly, high NAT10 expression was significantly associated with tumor grade, invasion depth, clinical stage and metastasis as well as with inferior overall survival (OS) of GC patients (Fig. 1G and Additional file 1: Table S2). Bioinformatics analysis of the GEO data using the Kaplan–Meier Plotter further demonstrated that high expression of NAT10 was correlated with shorter overall survival, first progression (FP) and post progression survival (PPS) in gastric cancer patients (Fig. 1H). In addition, NAT10 was found to be an independent risk factor for overall survival by multivariate Cox analysis (Fig. 1I). These data suggest that enhanced mRNA acetylation and NAT10 overexpression may promote the development and progression of GC.
NAT10 promotes gastric cancer cell proliferation and growth in an ac4C-dependent manner
To investigate the roles of NAT10-mediated ac4C modification on gastric carcinogenesis, we used the CRISPR/Cas9 technique to establish NAT10-knockout AGS cell lines (Fig. 2B and Additional file 1: Fig. S2B). As expected, NAT10 knockout dramatically reduced global ac4C modification in both total RNA and mRNA (Fig. 2C, D and Additional file 1: Fig. S2C-E). Importantly, NAT10 knockout decreased cell proliferation, colonic growth and invasion and induced substantial G2/M arrest and moderate apoptosis (Fig. 2E-G and Additional file 1: Fig. S3D, E, F, H). Similar phenomena were observed when NAT10 was knocked down by lentiviral-mediated shRNA in BGC823 and MKN45 cells (Fig. 2B-F and Additional file 1: Fig. S3). To test whether NAT10-promoted malignant phenotypes of cancer cells were dependent on its ac4C activity, NAT10-knockout cells were used to express wild-type NAT10 or mutant NAT10 lacking a functional acetyltransferase domain (G641E) or RNA helicase domain (K290A) (Fig. 2A, B) due to point mutations (K290A or G641E) previously shown to disrupt the RNA acetyltransferase function of NAT10 [19,20,21]. As shown in Fig. 2C-G and Additional file 1: Fig. S2C, S3E-F, re-expression of wild-type NAT10 in NAT10-knockout cells effectively rescued ac4C acetylation and corresponding changes to cancer cell physiology, whereas neither of the two NAT10 mutants exhibited these effects, suggesting that the RNA ac4C modification function of NAT10 is indispensable for its role in promoting gastric carcinogenesis. To further confirm this, Flag-tagged NAT10 Δhelicase lacking the RNA helicase domain was transiently transfected into NAT10-knockout cells. The results showed that the expression of NAT10 lacking the helicase domain failed to effectively reconstitute either RNA ac4C or cell proliferation (Additional file 1: Fig. S2D-E, S3D).
To evaluate the oncogenic role of NAT10 in gastric cancer in vivo, we applied both a subcutaneous xenograft model and a peritoneal dissemination model. In the subcutaneous xenograft model, NAT10 knockout significantly reduced the proliferation and growth of xenograft tumors, which could be offset by re-expression of wild-type NAT10 but not its mutant forms (Fig. 2H and Additional file 1: Fig. S4). The peritoneal dissemination assay showed that NAT10 knockout significantly reduced the formation of tumor nodules in the peritoneal cavity, and this effect could be reversed by the overexpression of wild-type NAT10 but not its mutant counterparts (Fig. 2I). Collectively, these results demonstrate the ac4C-dependent oncogenic role of NAT10 in GC progression.
NAT10-mediated ac4C modification maintains the stability of the MDM2 transcript
To explore the molecular mechanism by which NAT10 promotes GC progression, we performed RNA sequencing (RNA-seq) and ac4C-RNA immunoprecipitation sequencing (acRIP-seq) assays in stable NAT10 knockout and control AGS cells with independent biological replicates. RNA-seq revealed 2576 differentially expressed genes upon NAT10 knockout (Fig. 3A), which were found to be significantly enriched in gene sets involved in cell mitotic division, DNA replication, cell cycle, cell proliferation and so on by Gene Ontology (GO) enrichment analysis (Additional file 1: Fig. S5A). Notably, KEGG pathway analysis revealed that the cell cycle, p53 signaling pathway and DNA replication were affected by NAT10 knockout (Additional file 1: Fig. S5B). In acRIP-seq analysis, we identified on average 4,342 ac4C peaks in 3,591 transcripts and 2,709 ac4C peaks in 2,348 transcripts in control and NAT10-KO cells, respectively (Additional file 1: Fig. S5C). ac4C peaks were enriched in the 3′UTRs and coding sequences (Additional file 1: Fig. S5D, E). Among these peaks, 318 peaks within 306 mRNAs displayed hypoacetylation, and 430 peaks within 413 mRNAs showed hyperacetylation in NAT10-knockout cells relative to those in control cells (Fig. 3B and Additional file 2: Table S3). Since NAT10 positively mediates ac4C modification, only ac4C peaks with decreased abundance (termed ac4C hypo-peaks) upon NAT10 knockout were theoretically anticipated to include genuine targets of NAT10. We assessed whether these ac4C hypo-peaks were associated with differentially expressed mRNA genes in the RNA-seq analysis. The intersection of 318 hypo-peaks (within 306 mRNAs) with the 2576 differentially expressed genes identified by RNA-seq led to the identification of 36 candidate genes (Fig. 3C). To narrow down the scope of downstream targets, we additionally performed an RNA-seq assay in BGC823 cells with or without NAT10 knockdown and found that 313 coding genes were differentially expressed following NAT10 knockdown (Additional file 1: Fig. S5F), among which only one gene, MDM2, displayed ac4C hypoacetylation upon NAT10 knockout and consistent mRNA downregulation following either NAT10 knockout or knockdown (Fig. 3C).
Our ac4C-seq data revealed that MDM2 mRNA was modified by ac4C in its 3′UTR and that NAT10 knockout caused a significant decrease in ac4C levels (Fig. 3D). To confirm this, we applied acRIP-qPCR to examine the effects of NAT10 expression changes on ac4C levels in MDM2 mRNA. In accordance with the acRIP-seq results, significant ac4C enrichment of MDM2 mRNA was observed, and importantly, ac4C levels of MDM2 mRNA were significantly decreased in NAT10-depleted cells, while this effect was reversed by re-expression of wild-type NAT10 but not its mutants (Fig. 3E and Additional file 1: Fig. S6A). Similar to genomic NAT10 deletion, pharmacologic inhibition of NAT10 with Remodelin, a recently described NAT10 inhibitor [19], reduced global ac4C levels in both total RNA and mRNA and, more importantly, suppressed MDM2 ac4C modification (Additional file 1: Fig. S6B-D). In addition, RNA immunoprecipitation (RIP) followed by qRT–PCR assays showed that NAT10 could be significantly enriched in MDM2 mRNA (Fig. 3F). We subsequently measured the change in the mRNA and protein levels of MDM2 upon NAT10 knockout or knockdown. Consistent with the RNA-seq data, knockout of NAT10 substantially reduced MDM2 expression at both the mRNA and protein levels, whereas overexpression of wild-type NAT10, but not the mutants K290A or G641E, could enhance MDM2 expression (Fig. 3G, I), indicating that NAT10 mainly affects MDM2 expression through its RNA acetylation activity. Decreased MDM2 was also observed when we knocked down NAT10 in BGC823 and MKN45 cells (Fig. 3G, I and Additional file 1: Fig. S6E, F). Not surprisingly, treatment with Remodelin resulted in a significant reduction in MDM2 expression levels in a dose-dependent manner (Fig. 3H, J). Conversely, NAT10 expression was not altered by MDM2 overexpression or knockdown (Fig. 4A, B, G). Similar to the data obtained from human cell lines, the NAT10-knockout tumors showed a reduction in ac4C modification of the MDM2 transcript and decreased MDM2 expression compared to those in the control tumors, whereas the tumors with re-expression of wild-type NAT10 exhibited enhanced levels of MDM2 mRNA acetylation and MDM2 expression (Fig. 3K and Additional file 1: Fig. S6G, H). These data strongly support MDM2 as a bona fide target of NAT10.
Considering the role of NAT10 in regulating MDM2 mRNA levels, we evaluated whether NAT10 may affect MDM2 transcription and mRNA export. A luciferase reporter assay showed that NAT10 ablation did not affect MDM2 promoter activity (Additional file 1: Fig. S6I), suggesting that transcription of MDM2 was not affected by ac4C. Furthermore, no obvious change in the subcellular localization of MDM2 mRNA between the control and NAT10-KO cells was found (Additional file 1: Fig. S6J). Since transcription and mRNA export cannot explain the reduced MDM2 expression in NAT10-depleted cells, we then asked whether NAT10 influences MDM2 mRNA stability. In fact, depletion of NAT10 resulted in a noticeable decrease in the stability of the MDM2 transcript, while overexpression of wild-type NAT10 but not its mutants offset this effect (Fig. 3L). We also observed an obvious decrease in MDM2 mRNA stability in AGS and BGC823 cells with Remodelin treatment (Additional file 1: Fig. S6K). To study the potential roles of ac4C acetylation on the MDM2 3′UTR, we generated luciferase reporters containing a firefly luciferase, followed by the wild-type (Wt) MDM2 3′UTR or its mutant (Mut) counterpart. For the Mut reporters, the region covered by the ac4C peak (181 bases in length, putative ac4C sites) was deleted to eliminate the effect of ac4C acetylation (Fig. 3M). We found that depletion of NAT10 caused a markable reduction in both mRNA expression and activity of firefly luciferase when fused with the wild-type MDM2 3′UTR, and this effect was offset when firefly luciferase was fused with the mutant MDM2 3′UTR (Fig. 3N, O). Furthermore, overexpression of wild-type NAT10 rather than its mutant counterparts substantially increased luciferase mRNA levels and its activity of the reporter construct containing the wild-type MDM2 3′UTR, while the lack of the ac4C peak region in the MDM2 3′UTR abrogated this effect (Fig. 3N, O). In summary, our results indicate the regulation of MDM2 is under the control of NAT10-guided ac4C modifications.
As MDM2, an E3 ubiquitin ligase, targets p53 for proteasomal degradation and functions as a negative regulator of p53 [22], we tested whether NAT10 affects the p53 pathway in the p53 wild-type cell lines AGS and BGC823. As expected, p53 protein levels were significantly induced following genomic deletion or chemical inhibition of NAT10 but were decreased after re-expression of wild-type NAT10 but not its mutant forms (Fig. 3I, J). Similar results were observed in the xenograft tumor tissues (Fig. 3K). Even though p53 protein levels were changed upon regulation of NAT10 expression, the p53 (TP53) mRNA levels remained unchanged (Additional file 1: Fig. S6L), demonstrating that NAT10 reduces the expression of p53 at the protein but not the mRNA. Consistent with these findings, both mRNA and protein levels of the p53 target genes, p21 (CDKN1A) and PUMA, were also increased after depletion or inhibition of NAT10 and reduced after re-expression of NAT10 (Fig. 3I, J and Additional file 1: Fig. S6L, M). NAT10 was reported to acetylate p53 protein at K120 in CRC cells in a previous study [23]. In contrast, we and another group [9] observed that NAT10 failed to affect p53 acetylation (Additional file 1: Fig. S6N), which may be due to the cell type-specific role of NAT10 or bias relating to the experimental conditions. We examined whether NAT10 ablation affects p53 protein degradation and found that NAT10 knockout significantly increased the stability of endogenous p53 protein (Fig. 3P), indicating that NAT10 destabilizes p53 protein.
MDM2 mediates NAT10-induced malignant cell phenotypes
We further explored the biological significance of MDM2 in the tumor-promoting function of NAT10. Ectopic expression of MDM2 significantly reduced the induction of p53 and its target p21 elicited by NAT10 depletion, whereas knockdown of MDM2 effectively abolished the inhibitory effect of NAT10 overexpression on the p53 pathway (Fig. 4A, B), indicating that NAT10 can destabilize p53 protein via MDM2. We subsequently examined whether MDM2 affects NAT10-mediated malignant cell phenotypes. Indeed, overexpression of MDM2 significantly rescued the hypoproliferative phenotype of NAT10-deficient GC cells. In contrast, MDM2 silencing largely abrogated the effect of NAT10 overexpression on cell growth/proliferation (Fig. 4C, D). Moreover, MDM2 overexpression significantly abolished G2/M arrest and apoptosis mediated by NAT10 depletion. In contrast, MDM2 silencing resulted in an induction of cell cycle arrest and apoptosis in NAT10-overexpressing cells (Fig. 4E, F). Thereafter, further in vivo rescue experiments were conducted, and our data demonstrated that the impaired potential of in vivo tumor growth triggered by NAT10 knockout could be restored by stable MDM2 overexpression (Fig. 4G, H), showing that MDM2 mediates the oncogenic functions of NAT10.
We next sought to verify whether our findings could be extended to gastric cancer patients. The results showed that GC specimens displayed enhanced levels of both MDM2 ac4C and MDM2 mRNA expression compared to those in paired normal gastric specimens (Fig. 5A, B). Further correlative analysis of the ac4C modification level within MDM2 showed a significant association with MDM2 mRNA levels as well as with NAT10 expression in GC tissues (Fig. 5C), further supporting the notion that NAT10-mediated ac4C modification stabilizes MDM2 mRNA. Moreover, NAT10 mRNA expression positively was associated with MDM2 mRNA expression (Fig. 5D). This correlation was true in gastric cancer sample cohorts based on analysis of the mRNA expression data from the TCGA and GEO datasets (Fig. 5E). IHC staining of MDM2 in the TMA cohort demonstrated high expression of MDM2 in 102 of 202 (50.5%) GC tissues (Fig. 5F and Additional file 1: Fig. S1D). More importantly, MDM2 protein expression was positively associated with NAT10 protein in GC tissues (Fig. 5F). In addition, Kaplan–Meier survival curves showed that high expression of both NAT10 and MDM2 predicted the poorest overall survival (Fig. 5H). These data further support our findings that the NAT10/MDM2 axis promotes GC progression and leads to worsened overall survival.
Hp infection induces NAT10 expression and regulates p53 stability
Hp is a major risk factor for gastric cancer [24,25,26]. Previous studies have shown that Hp infection promotes proteasomal degradation of p53 in gastric epithelial cells, which results in downregulation of p53 protein [27,28,29]. Given that our results suggest that NAT10 maintains the stability of MDM2 mRNA via ac4C modification and consequently destabilizes the p53 protein, we asked whether Hp infection can affect NAT10 expression to regulate p53 stability. To answer this question, we first measured p53 protein in GES1 normal gastric epithelial cells and AGS gastric cancer cells cocultured with two Hp standard strains, SS1 and ATCC43504. Hp-mediated p53 inactivation and p53 destabilization were verified in our study (Fig. 6A-C and Additional file 1: Fig. S7A). We next examined the effect of Hp infection on NAT10 expression. Interestingly, the expression of NAT10 mRNA and protein was significantly induced in the infected cells, and this induction by Hp was consistent across several time points (Fig. 6A, C). We then explored the expression of NAT10 in Hp-infected mice. When mice were challenged with Hp strain SS1, which successfully colonizes the murine stomach [30], they showed a significant induction of NAT10 mRNA and protein expression in the gastric tissues compared with that in control mice (Fig. 6D, E and Additional file 1: Fig. S7B).
In addition, we analyzed the global levels of ac4C modification and MDM2 ac4C levels in Hp-infected cells. Consistently, global ac4C modification and MDM2 ac4C levels were increased in GES1 and AGS cells infected with Hp (Fig. 6F, G), and MDM2 mRNA stability was enhanced accordingly (Fig. 6H), leading to MDM2 overexpression (Fig. 6A, C). Moreover, the upregulation of MDM2 and the downregulation of p53 and p21 were further confirmed in the Hp-infected mice described above (Fig. 6D, E). To verify that Hp-mediated p53 downregulation is dependent on NAT10, NAT10-knockout AGS cells and control cells were cocultured with Hp and then analyzed for MDM2/p53 expression. As expected, the regulatory effect of Hp on MDM2 and p53 expression was reversed by NAT10 knockout (Fig. 6I). Overall, the results demonstrate that Hp infection induces the degradation of p53 at least partly via a NAT10-mediated mechanism.
Targeting NAT10 with Remodelin elicits antitumor activity and improves the sensitivity of GC cells to MDM2 inhibitors
In view of the above results, we subsequently evaluated the anticancer activity of the NAT10 inhibitor Remodelin. We started to evaluate the effect of Remodelin on cell proliferation in a panel of GC cell lines. Remodelin treatment of GC cell lines expressing either wild-type or mutant p53 for 72 h showed a potent inhibitory effect on cell proliferation in a growth IC50 range of 8–16 µM (Fig. 7A, B). Remodelin treatment also showed a dose-dependent inhibition of clonogenic growth and induction of apoptosis in both p53 wild-type GC cell lines and p53 mutant GC cell lines (Fig. 7C, D and Additional file 1: Fig. S8). Interestingly, the normal gastric epithelial cell lines GES1 and NGEC were resistant to Remodelin treatment (Fig. 7A-D).
Considering the potent effects of Remodelin in cell culture, we examined the in vivo antitumor properties of Remodelin in mice using subcutaneous xenografts of BGC823 cells. Tumor-bearing mice were treated with vehicle or different doses of Remodelin via intraperitoneal injection. With 60 mg per kg doses (daily), complete tumor growth inhibition was observed in mice (Fig. 7E), and no significant adverse effects, such as weight loss or treatment-related mortality, were observed during Remodelin treatment (Additional file 1: Fig. S9). We also observed that mice treated with Remodelin exhibited increased survival (Fig. 7F).
Given previous reports that treatment with MDM2 inhibitors (inhibitors of the MDM2-p53 interaction) leads to the accumulation of MDM2 as a p53 target following p53 induction (which forms a feedback loop with p53) [31,32,33], we tested whether Remodelin could potentiate the antitumor effect of MDM2 inhibitors. As expected, treatment with HDM201, a highly selective, orally available active small-molecule MDM2 inhibitor in AGS cells, caused a dose-dependent induction of p53 and p21 expression as well as MDM2 accumulation (Fig. 7G). The combination of HDM201 and Remodelin resulted in higher p53 accumulation and dramatic MDM2 downregulation compared to HDM201 alone, which correlated with stronger activation of the p53 downstream target p21 (Fig. 7G). We subsequently evaluated the antiproliferative effect of combined treatment with HDM201 and Remodelin in GC cell lines and observed remarkable synergistic growth inhibitory activity, i.e., combination index < 1, at most or all concentrations tested in GC cell lines harboring wild-type p53 (Fig. 7H), while p53-mutant cells were resistant to MDM2 inhibitors (Additional file 1: Fig. S10). Similar results were obtained when combining Remodelin and an additional MDM2 inhibitor Nutlin-3 (Fig. 7G, I and Additional file 1: Fig. S10). Using a subcutaneous xenograft model of BGC823 cells, we found that combined administration of HDM201 and Remodelin was significantly more effective in suppressing in vivo tumor growth than either Remodelin or HDM201 alone (Fig. 7J). This correlated with a significant improvement in the survival of mice bearing tumors in the combination treatment group compared with the HDM201-alone treatment group (Fig. 7K). The combination treatment did not induce severe toxicity in mice because there was no significant loss in the total body weight of nude mice bearing tumors (Additional file 1: Fig. S9) or no treatment-related mortality. Together, these data demonstrate that pharmacological inhibition of NAT10 may provide a promising treatment for gastric cancer.