NAT10 expression and the level of ac4C modification are significantly up-regulated in CRC
To reflect the importance of NAT10 in CRC, we first measured NAT10 mRNA levels in 80 CRC tissue samples and their paired adjacent normal tissues via qRT-PCR (Fig. 1A). Thus, we observed that the tumor tissues showed higher NAT10 expression levels, consistent with TCGA and GEO datasets (GSE41258) (Fig. S1A). Next, the tissue microarray (TMA) was prepared from the 80 patient samples to further explore the NAT10 protein level in CRC (Fig. S1C, D). Immunohistochemistry (IHC) in this regard showed up-regulated NAT10 expression in CRC tissues (Fig. 1B), as reflected by the higher degrees of staining and the H-scores obtained (Fig. 1C), consistent with Clinical Proteomic Tumor Analysis Consortium (CPTAC) data (Fig. S1B). Further, an exploration of the relationship between NAT10 protein level and the clinicopathological characteristics of patients with CRC showed that tumor stage, lymph node metastasis, vascular invasion, and distant metastasis were statistically significantly associated with NAT10 expression (Table 1). Specifically, clinical subgroup analysis demonstrated that the N1 + N2 group with lymph node metastasis, M1 group with distant metastasis, and III + IV tumor stage group showed higher NAT10 expression levels, implying that NAT10 can regulate CRC proliferation and metastasis (Fig. S1E). Furthermore, we detected ac4C levels by performing IHC staining on our TMA (Fig. S1F). Notably, consistent with NAT10 expression, the ac4C level was also dramatically upregulated in CRC tissues (Fig. 1E, F). To better reflect NAT10 protein level in this regard, we chose eight samples from the patients’ cohort to perform western blotting (WB), which showed that NAT10 was highly-expressed in CRC samples (Fig. 1G). The eight samples were also used for dot blotting, which also showed high ac4C levels in the tumor tissues (Fig. 1H). In addition to the detection of NAT10 and ac4C levels in tissues, our results also indicated that the levels of NAT10 and ac4C were likewise overexpressed in CRC cell lines (Fig. 1I-K). The analysis of all the CRC tissues and cell lines via immunofluorescence (IF) staining indicated that NAT10 was localized in the nucleus and cytoplasm of CRC cells, but predominantly in the nucleus (Fig. S1G, H). Taken together, these results revealed that the ac4C writer, NAT10 is up-regulated in CRC and is significantly associated with the clinicopathological characteristics of patients with CRC.
Nat10 Enhances The Proliferation, Migration, And Invasion Of Crc Cells In Vitro
For verification using CRC cell lines, we specifically chose SW480 and DLD-1 cells, with high NAT10 expression, and HT-29 cells, with relatively low NAT10 expression. After the shNC, shNAT10-1, shNAT10-2, oeVector, and oeNAT10 plasmids were established, they were cloned into the lentivirus infection system to further construct stable cell lines. Transfection efficiency in the three cell lines mentioned above was then verified via qRT-PCR, WB, and dot blot (Fig. 2A, B, and Fig. S2A). Then, to evaluate the ability of NAT10 to adjust CRC cell proliferation, CCK-8, colony formation, and EdU assays were performed. Consistent with the assay results, NAT10 knockdown dramatically inhibited the proliferation of SW480 and DLD-1 cells, while its overexpression in HT-29 cells showed opposite effects (Fig. 2C-H and Fig. S2B-D). Additionally, it was evident that NAT10 knockdown arrested the growth of SW480 and DLD-1 cells in the G2/M phase and increased the ratio of apoptotic cells based on flow cytometric assays of cell cycle and apoptosis. Conversely, the number of CRC cells in the G2/M phase and the apoptotic rates were significantly decreased when NAT10 was overexpressed in HT-29 cells (Fig. 2I-L and Fig. S2E, F). Further, transwell and wound healing assays, performed to detect changes in cell migration and invasion abilities, showed that NAT10 depletion significantly impaired the migration and invasion abilities of SW480 and DLD-1 cells, while its overexpression led to contrary phenomena in HT-29 cells (Fig. 2M, N and Fig. S2G). Taken together, NAT10 could promote the proliferation of CRC cells and also regulate their migration and invasion.
Nat10 Facilitates The Tumorigenesis And Metastasis Of Crc Cells In Vivo
To explore the effect of NAT10 in vivo, xenograft tumor models and metastasis models were established. In the xenograft tumor model, SW480 and DLD-1 cells, stably transfected with shNC and shNAT10-2 and HT-29 cells stably transfected with oeVector and oeNAT10 were subcutaneously injected into nude mice. Thereafter, the analysis of changes in tumor weight and the trends of tumor volume revealed that NAT10 knockdown inhibited tumor growth in vivo, while its overexpression had an opposite effect (Fig. 3A, B, and Fig. S3A). IHC showed that, compared with the control group, the shNAT10-2 group showed significantly downregulated Ki67 expression, while the oeNAT10 group showed significantly up-regulated Ki67 expression (Fig. 3C, D and Fig. S3B). Further, the liver and lungs of metastasis models were injected with stably transfected CRC cells via the distal tip of the spleen and tail veins of the nude mice, respectively. Thereafter, higher fluorescence intensity and larger numbers of metastasis nodules in the liver and lungs indicated that NAT10 facilitated metastasis, which could be restrained by its knockdown (Fig. 3E-H and Fig. S3C, D). Thus, the animal models revealed that CRC cell tumorigenesis and metastasis in vivo could be promoted by upregulated NAT10 expression, consistent with the findings obtained in vitro.
Identification of the profile of ac4C-modified genes regulated by NAT10 in CRC cells
To determine how NAT10 regulated CRC progression at the transcriptional level, we first subjected NAT10-knockdown and control SW480 and DLD-1 cells to RNA-seq. Thereafter, differentially expressed genes (fold change >1.20 or < 0.83, p <0 .05) between the two groups of SW480 and DLD-1 cells were shown using volcano plots (Fig. S4A), while the down-regulated genes in CRC cells upon the knockdown of NAT10 were shown using a heatmap (Fig. 4A). To explore the biological processes NAT10 might be involved in, 126 genes significantly downregulated and 58232 genes altered by NAT10 in both SW480 and DLD-1 cells were respectively included for gene ontology (GO) analysis and gene set enrichment analysis (GSEA) (Fig. S4B, C). GO analysis showed that the enriched pathways included TOR signaling, protein ubiquitination, the Wnt signaling pathway, and mRNA export from the nucleus, while GSEA indicated that NAT10 might be associated with the CTNNB1 oncogenic signature as well as epithelial-mesenchymal transition (Fig. 4B and Fig. S4D).
Moreover, given that NAT10 mainly participates in epigenetic regulation by binding and affecting ac4C acetylated transcripts and to clarify its role as an ac4C ‘writer’ protein, we further performed RIP-seq and acRIP-seq using SW480 and DLD-1 cells (the acRIP-seq process is shown in Figure S4E). Notably, RIP-seq revealed that NAT10 binds to 58302 transcripts, among which 34.04% were mRNAs. Additionally, acRIP-seq showed 15622 ac4C peaks corresponding to 7608 transcripts, 94.07% of which were observed in mRNA. This observation represented the main type of transcripts ac4C modification happened in and the potential combination of NAT10 with mRNA (Fig. 4C). After analyzing the distribution on mRNA, we noticed the appearance of NAT10-binding regions and ac4C peaks in coding sequences (CDS) and 3’-untranslated regions (3’-UTR), consistent with previously reported data[8, 12] (Fig. 4D). Motif analysis further indicated that the typical ac4C motif, ‘CXXCXXCXX’, was significantly enriched both in the NAT10-binding and ac4C-modified sequences, suggesting that NAT10 possibly modulated the ac4C modification of mRNA by binding to it (Fig. 4E). Based on GO analysis, we also observed that NAT10-binding (FPKM>0 both in IP and IgG groups) and ac4C-modified genes were involved in rRNA processing, translational initiation, mRNA stability regulation, cell cycle, and the positive regulation of the canonical Wnt signaling pathway (Fig. 4F).
To further clarify the target genes and consider the crucial effects of NAT10 on mRNA stability, we overlapped the genes identified based on RNA-seq, RIP-seq, and acRIP-seq. Thus, we observed that 69 genes bound by NAT10 were tagged with ac4C and were down-regulated upon NAT10 knockdown (Fig. 4G). Thereafter, by re-ranking the abovementioned 69 genes via fold-enrichment in RIP-seq and acRIP-seq followed by overlapping the top two-thirds of them considering the two datasets, 33 genes were identified as potential NAT10 direct targets (Fig. S4F). Furthermore, the correlation of these 33 genes’ expression with NAT10 was observed using the public GEPIA dataset based on the TCGA (http://gepia.cancer-pku.cn). Based on the correlation analysis, the expression of 21 genes was notably positively correlated with NAT10, with only seven genes’ (HNRNPH3, MPP6, NAP1L4, POU2F1, UBA1, KIF23, HNRNPA2B1) showing Pearson correlation coefficients > 0.3 (Fig. 4H). Next, using IGV, we visualized the obvious ac4C peaks and NAT10-binding peaks corresponding to these seven genes and via qRT-PCR, observed the mRNA levels of these genes in 24 randomized patient samples. As shown (Fig. 4I and Fig. S4G, H), KIF23 changed markedly in tumor tissues corresponding to adjacent tissues among the 7 candidate genes and the potential ac4C motif in the modified region (chr15:69 448 127-69 448 397) on KIF23 mRNA 3’UTR region might be ‘CUUCUCCAG’. Therefore, we successfully identified the profile of ac4C-modified genes regulated by NAT10 in CRC, with KIF23 seeming to be a direct target of NAT10 in CRC cells.
Nat10 Stimulates Kif23 Expression via ac4C Modification
To explore the correlation between the expression levels of NAT10 and KIF23, we first detected the KIF23 mRNA level of KIF23s in tissue samples from 80 patients with CRC via qRT-PCR. Thus, we observed higher KIF23 mRNA levels in tumor tissues as well as a positive correlation between KIF23 mRNA expression and NAT10, consistent with TCGA or GEO datasets (GSE40967) (Fig. 5A, B and Fig. S5A). Next, the detection of KIF23 protein levels via IHC using the TMA showed similar results implying the existence of a potential regulatory association between NAT10 and KIF23 (Fig. S5B, C). We also noted that KIF23 protein expression was significantly associated with tumor site, tumor stage, lymph node metastasis, nerve invasion, and CEA among the 80 patient samples (Table 2). The verification of the interaction between NAT10 and KIF23 mRNA via RIP-qPCR assays using three CRC cell lines showed a notable enrichment of NAT10 in conjunction with KIF23 mRNA compared with the IgG groups (Fig. 5C and Fig. S5D). Consistent with the results of acRIP-seq, acRIP followed by qPCR confirmed the abundance of ac4C modifications on KIF23 mRNA (Fig. 5D and Fig. S5E). Further, we observed that the interaction between NAT10 and KIF23 mRNA and the abundance of ac4C modification sites on KIF23 mRNA could be changed following NAT10 knockdown and overexpression, which contributed to stronger binding between NAT10 and KIF23 mRNA and up-regulated ac4C modification in KIF23 mRNA. Furthermore, NAT10 disruption led to opposite results (Fig. 5E, F Fig. S5F, G). Considering the correlation between NAT10 and KIF23, we further investigated KIF23 expression via qRT-PCR and WB upon NAT10 knockdown and overexpression. The same expression change tendencies were observed for all three CRC cell lines (Fig. 5G and Fig. S5H). Then, to demonstrate that the regulatory effect of NAT10 on KIF23 could be primarily attributed to its binding to the 3’-UTR region of KIF23 mRNA rather than direct or indirect regulation of the promoter activity, two experimental scenarios were established. In the first scenario, we noticed that the luciferase activity for a reporter containing the KIF23 promoter region remained unchanged upon NAT10 knockdown or overexpression in CRC cells, indicating that NAT10 could not regulate KIF23 expression by directly or indirectly modulating its promoter activity of KIF23 (Fig. S5I). In the second scenario, we investigated whether NAT10 could specifically bind to the ac4C-modified regions in the 3’-UTR region of KIF23 mRNA via luciferase reporter assay with the reporter containing the ac4C-modified regions of the 3’-UTR region. The wild type contained the potential ac4C-modified regions, while the mutation type did not. Thus, we observed that the luciferase activity of the reporter region containing the ac4C-modified regions could be repressed in SW480 and DLD-1 cells with NAT10 knockdown and enhanced in HT-29 cells with NAT10 overexpression. Conversely, the mutation group did not show any significant changes in this regard (Fig. 5H and Fig. S5J). To further confirm the direct binding between NAT10 and KIF23 mRNA, SW480 cells were subjected to RNA electrophoretic mobility shift assay (REMSA) which showed that the complex formed as a result of the reaction between the labeled probes and nuclear proteins could be inhibited by the unlabeled probes, and this inhibition could be attenuated by the mutant probes. The supershifted complex phenomenon was also observed after NAT10 antibody addition, indicating the direct binding between NAT10 with the ac4C motif in KIF23 mRNA (Fig. 5I and Fig. S5K).
Given that KIF23 expression was altered in conjunction with changes in NAT10 expression, we confirmed the bind of NAT10 to KIF23 mRNA via its ac4C motif. Next, we wondered whether KIF23 regulation by NAT10 was ac4C dependent. Reportedly, NAT10 functions as an acetyltransferase, owing to its N-acetyltransferase domain (558–753) and the mutation in G641 can abrogate its acetyl-CoA binding structure[22]. Subsequently, we introduced a point G641E mutation in the N-acetyltransferase domain of NAT10 with FLAG tag (NAT10-mut) or NAT10 wide type (NAT10-wt) and transfected them into SW480 and DLD-1 cells (Fig. S5L). Interestingly, RIP and acRIP followed by qPCR revealed that NAT10-mut could impair ac4C modification on KIF23 mRNA, but not affect the bond between NAT10 and KIF23 mRNA (Fig. 5J). We also observed that NAT10-wt, but not NAT10-mut, could upregulate the expression of KIF23, suggesting that the N-acetyltransferase domain of NAT10 played an important role in mediating ac4C modification (Fig. 5K). In addition, the HA-tagged KIF23 expression vector (KIF23-wt) and its mutant, with mutations ac4C sites (KIF23-mut), were constructed (Fig. S5M), and as expected, we demonstrated that NAT10-wt, but not NAT10-mut, could upregulate KIF23-wt expression and that NAT10-wt could not influence KIF23-mut expression due to the absence of the ac4C motif (Fig. 5L).
Reportedly, NAT10 might regulate the stability of mRNA or its translation efficiency to modulate gene expression[11, 13]. Thus, we treated CRC cells with actinomycin D (5µg/mL) to examine RNA decay following NAT10 knockdown or overexpression. The results obtained indicated that NAT10 could enhance KIF23 mRNA stability, and this effect could subsequently enhance protein translation (Fig. 5M and Fig. S5N). Previous studies have indicated that NAT10 can acetylate histones via its acetyltransferase activity or promote protein degradation via its E3 ligase activity[19]. To exclude the possibility that NAT10 might mediate the stability of KIF23 protein, we treated CRC cells with the protein translation inhibitor cycloheximide (CHX) (100µg/mL) and observed a mild effect on KIF23 protein stability (Fig. 5N and Fig. S5O). Collectively, these results indicated that NAT10 regulated KIF23 mRNA via ac4C modification.
Nat10 Regulates The Wnt/β-catenin Pathway Via The Nat10/kif23/gsk-3β Axis
To clarify the downstream of the KIF23 pathway, we performed GSEA analysis using TCGA and GEO datasets (GSE4097), which demonstrated the importance of NAT10 to cell cycle, G2/M checkpoint, MYC targets, and Wnt/β-catenin signaling (Fig. S6A). Considering previous studies, which showed that KIF23 exerts a regulatory effect on the Wnt/β-catenin pathway and the significant correlation between NAT10 and β-catenin[34–36], we hypothesized that NAT10 possibly regulates the Wnt/β-catenin pathway by mediating KIF23. Next, the detection of β-catenin expression in the CRC TMA showed high β-catenin expression, which was positively correlated with the expression levels of NAT10 and KIF23 according to the Pearson correlation analysis (Fig. 6A and Fig. S6B). β-catenin protein expression level also showed a strong correlation with tumor site, TNM staging system, tumor stage, lymph node metastasis, and distant metastasis (Table S4). Subsequently, to investigate whether NAT10 modulates the Wnt/β-catenin pathway in CRC, we measured phosphorylated GSK-3β, GSK-3β, and β-catenin levels as a function of changes in NAT10 levels, and observed that the Wnt/β-catenin pathway could be notably activated by NAT10 (Fig. 6B and Fig. S6C). Moreover, the subcellular protein fraction assay confirmed that NAT10 knockdown impaired β-catenin expression in the nucleus, while its overexpression significantly enhanced β-catenin expression in the nucleus (Fig. 6C and Fig. S6D). IF assay using SW480 and DLD-1 cells and IHC using xenograft tumor tissues as mentioned above showed similar results (Fig. 6D, E and Fig. S6E, F). Additionally, after the transfection efficiency in CRC cells following KIF23 knockdown or overexpression was verified, a series of rescue experiments were performed in vitro (Fig. S6G). Thus, we observed that CRC cell proliferation abrogated by shNAT10-2 could be rescued by oeKIF23, while that enhanced by oeNAT10 could be decreased by shKIF23-2 (Fig. S6H, I and S7A). Furthermore, oeKIF23 could reverse G2/M arrest and the apoptotic rates caused by NAT10 knockdown, while shKIF23-2 could lead to G2/M arrest and increased apoptotic rates following NAT10 overexpression (Fig. S7B, C). Transwell and wound healing assays yielded similar results (Fig. S7D). Next, considering cell cycle regulation by NAT10, the transcriptional targets that were activated the most by β-catenin, including cyclin D1 and c-Myc, were detected via WB. Survivin and bcl-xl, downstream of c-Myc, acting as typical anti-apoptotic proteins, were found to induce G2/M transition. Our results also indicated that NAT10 knockdown significantly downregulated KIF23 protein levels of KIF23, phosphorylated GSK-3β, β-catenin, cyclin D1, c-Myc, surviving, and bcl-xl, while NAT10 upregulation yielded opposite results. In the co-transfected groups, sh-KIF23-2 could reverse the regulatory effects mediated by oeNAT10, and the effects caused by sh-NAT10-2 were rescued by oeKIF23 (Fig. 6C and Fig. S8A).
Interestedly, we noticed that NAT10 could be modulated by GSK-3β[26]. This observation raised our curiosity regarding the existence of a regulation loop. Following treatment with LiCl (20mmol/L), a GSK-3β inhibitor, for 48 h, IF assays showed increasing NAT10 and KIF23 staining (Fig. 6D and Fig. S8B). RIP or acRIP after the LiCl treatment followed by qPCR also showed an obvious increase in KIF23 mRNA (Fig. 6E and Fig. S8C). Similarly, WB confirmed that GSK-3β inhibition could rescue the effects of shNAT10-2 on the NAT10/KIF23/GSK-3β/β-catenin axis and also enhance the up-regulatory effects of oeNAT10 (Fig. 6F and Fig. S8D). These observations indicated that in CRC cells, NAT10 regulates the Wnt/β-catenin pathway via the NAT10/KIF23/GSK-3β loop.
Targeting Nat10 Using Remodelin Exhibits Potential Therapeutic Effects
To investigate the clinical significance of NAT10, first, Kaplan-Meier analysis using data corresponding to 80 patients with CRC revealed that higher NAT10, KIF23, and β-catenin expression levels were associated with poorer overall survival (OS), and the analysis based on NAT10 was consistent with TCGA and GEO datasets (GSE40967) (Fig. 7A).
Subsequently, remodelin, which was reported to be a chemical inhibitor of NAT10[22, 37, 38], was applied to detect its inhibitory effect on NAT10 in vitro and vivo. By treating SW480 cells with remodelin at different concentrations, the half-maximal inhibitory concentration (IC50) of this NAT10 inhibitor in SW480 cells was 20.89µM (Fig. 7B). Further, after remodelin (20µM) treatment for 48 h, the ac4C modification on KIF23 mRNA and the bond between NAT10 and KIF23 mRNA were notably weakened (Fig. 7C). Furthermore, in SW480 cells the downstream of NAT10, including KIF23 and β-catenin, were found to be significantly downregulated by remodelin (Fig. 7D). In vitro, the proliferation, migration, and invasion abilities of SW480 cells were suppressed by remodelin, and SW480 cells also showed G2/M arrest and an increased rate of apoptosis following remodelin treatment (Fig. 7E-K). In vivo, xenograft tumor models of BALB/c nude mice were administered remodelin via oral gavage (100mg/kg per day) for 15 days, while the metastasis models were administered remodelin via intraperitoneal injection at 5 mg/kg every other day for 4 weeks (Fig. 8A, D). The results thus obtained indicated that remodelin could notably inhibit the growth of xenograft tumors and the IHC analysis of the expression of Ki67, KIF23 and β-catenin presented similar results (Fig. 8B, C). Live imaging and the numbers of metastasis nodules in the liver and lungs of the model mice showed that remodelin could also significantly inhibit SW480 cells metastasis to the lungs or liver (Fig. 8E, F). All these findings demonstrated that NAT10 is a great prognostic indicator and targeting it using remodelin could inhibit CRC cell progression in vitro and in vivo, providing a potential prognosis or therapeutic target for CRC.