The expression of STAT3 is positively related to METTL3 in HCC

First, we assessed the relationship between METTL3 expression and the survival of HCC patients by online analysis tools GEPIA [24] and Kaplan–Meier Plotter [25]. The results revealed that higher expression of METTL3 correlated with shorter survival times (Fig. 1A, B), suggesting the potential role of METTL3 on HCC progression. STAT3 is a classic oncogene that plays a crucial role in promoting the proliferation and metastasis of cancer cells [19, 26, 27]. In a previous study, we revealed the pro-metastatic effect of STAT3 on HCC cells [19]. Thus, we wondered if there was a potential relationship between METTL3 and STAT3 in the promotion of HCC metastasis. To test this conjecture, we quantified the expression level of STAT3 and METTL3 by Western blotting assay. Compared with MHCC97L and HepG2 cell lines, METTL3 and STAT3 were more strongly expressed in the high-metastatic MHCC97H and HCCLM3 cell lines (Fig. 1C). A microarray involving 122 clinical HCC tissues was used for immunohistochemistry (IHC) staining of METTL3 and STAT3. The results indicated that STAT3 was positively correlated to METTL3 in the examined HCC specimens (Fig. 1D, E; spearman correlation analysis, r = 0.656, p < 0.001). Notably, strong staining of METTL3 and STAT3 was observed in metastatic tissues compared with non-metastatic tissues (Fig. 1F, G), suggesting the potentially vital role of METTL3 and STAT3 in cell metastasis. Moreover, Western blotting analysis indicated that METTL3 and STAT3 were highly expressed in HCC tissues in comparison with paired non-tumor tissues (n = 10) (Fig. 1H, I). These findings reveal that the expression of STAT3 is positively related to METTL3 in HCC.

METTL3 mediates m6A modification of STAT3 mRNA in HCC cells

Given the close association between METTL3 and STAT3, we sought to explore the regulatory mechanism between them in HCC cell metastasis. First, we evaluated the m6A level of total RNA in paired HCC tissues and corresponding non-tumor tissues (n = 10). The m6A levels in HCC tissues were upregulated compared to non-tumor tissues (Fig. 2A). Moreover, the m6A levels in high-metastatic HCC cell lines MHCC97H and HCCLM3 were higher than those in low-metastatic MHCC97L and HepG2 and normal liver L-O2 cells (Fig. 2B). These results indicated the association between high m6A levels and cell metastasis. As shown in Fig. 1C, higher STAT3 expression levels were observed in high-metastatic MHCC97H and HCCLM3 cell lines. Thus, we hypothesized whether m6A modification was involved in regulating STAT3 expression. Methylated RNA immunoprecipitation (MeRIP) analysis showed that STAT3 mRNA was indeed subjected to m6A modifications. Notably, the m6A levels of STAT3 mRNA in MHCC97H and HCCLM3 were higher than those in MHCC97L and L-O2 cells (Fig. 2C). To localize the m6A site of STAT3 at a transcriptome-wide level, we performed m6A sequencing (m6A-seq) using mRNA isolated from MHCC97H cells. In line with previous studies, the m6A peaks were mostly enriched in the CDS and 3’-untranslated region (UTR) (Fig. 2D, E). The m6A consensus sequence GAAC (RRACH) motif was identified to be highly enriched within m6A sites in the immunopurified RNA (Fig. 2F). M6A-seq data showed that STAT3 mRNA had two m6A peaks, both of which were in the 3'UTR of STAT3 mRNA, and the positions of these two peaks almost overlapped (peak1:40,466,894–40,467,104; peak2:40,466,931–40,467,111, the details were presented in Supplementary table S5). Then, we visualized the sequencing results and found the m6A peak according to the location information (Fig. 2G). Subsequently, we performed MeRIP and qRT-PCR to validate the m6A-seq data in STAT3 mRNA (Fig. 2H, I). Different primers were used to amplify the m6A peak region or a control (non-peak) region of STAT3 mRNA. Meanwhile, EGFR (a methylated mRNA) and RCN2 (a non-methylated mRNA) were separately amplified as positive control and negative control. The results confirm the presence of the m6A peak in STAT3 mRNA (Fig. 2H, I).

To investigate the potential effect of METTL3 on the m6A modification of STAT3, we performed RIP assays in MHCC97H and HCCLM3 cells. The data revealed the interaction between METTL3 and STAT3 mRNA (Fig. 2J). Moreover, the m6A level of STAT3 was upregulated in METTL3-overexpressed MHCC97L cells relative to control cells (Fig. 2K). Decreased m6A level of STAT3 was also observed in METTL3-knockdown MHCC97H cells compared to control cells (Fig. 2L). Next, we performed dual-luciferase reporter and mutagenesis assays in HEK293T cells (Fig. 2M). The data revealed that knockdown of METTL3 obviously depressed the luciferase activity of reporter carrying wild-type 3'UTR fragment of STAT3 (Fig. 2N). However, this phenomenon was disappeared when the RRACH motif was deleted or mutated in the abovementioned m6A peak (Fig. 2N), suggesting that METTL3 induces m6A modification of STAT3 mRNA at this site. Overall, the above results support that METTL3 induces the m6A modification of STAT3 in HCC cells.

METTL3 upregulates STAT3 by promoting the translation of STAT3 mRNA

Next, we explored whether METTL3-induced m6A modification affected STAT3 expression. qRT-PCR and Western blotting analyses were performed in different cell lines. Overexpression of METTL3 in MHCC97L cells upregulated the protein levels of STAT3 while mRNA levels were not altered (Fig. 3A, B). Meanwhile, silencing of METTL3 decreased the protein levels of STAT3, but no changes were observed at the mRNA level (Fig. 3C-E, Fig. S1A). Next, we treated cells with STM2457, a METTL3-specific inhibitor, and detected the changes in STAT3 m6A level. We found that the m6A level of STAT3 mRNA was reduced upon STM2457 treatment (Fig. 3F). It was worth noting that treatment of STM2457 also caused the decrease of STAT3 protein level, however, the mRNA level had no significant alteration (Fig. 3G, Fig. S1B, C). Moreover, METTL3 overexpression did not affect the stability of STAT3 mRNA (Fig. 3H, Fig. S1D). These results indicated that METTL3 might regulate the translation or protein stability of STAT3. To assess whether METTL3 modulated STAT3 protein stability, we treated cells with cycloheximide (CHX, a protein translation inhibitor) to block translation and measured STAT3 levels. The results showed that overexpression of METTL3 yielded no significant stabilizing effect on STAT3 protein levels (Fig. 3I). Therefore, we hypothesized that METTL3 could regulate the translation of STAT3 mRNA. Rodrigues et al. have reported that the ribosome-engaged RNA fraction is a reflection of the translation activity of gene [28]. They used GFP-tagged RPL10A to track the translating mRNA [28]. Thus, we performed RIP assay and found that METTL3 overexpression enhanced the abundance of RPL10A-enriched STAT3 mRNA (Fig. 3J); meanwhile, knockdown of METTL3 inhibited the enrichment of RPL10A on STAT3 mRNA (Fig. 3K).

Fig. 3

METTL3 upregulates STAT3 by promoting the translation of STAT3 mRNA. A, B qRT-PCR (A) and Western blotting analysis (B) of METTL3 and STAT3 in MHCC97L cells transfected with the indicated plasmids. The right panel in (B) shows the quantification of the intensity relative to tubulin. C qRT-PCR analysis of METTL3 and STAT3 in MHCC97H cells transfected with the indicated siRNAs. D, E Western blotting analysis of METTL3 and STAT3 in MHCC97H cells transfected with si-Control or si-METTL3. (E) shows the quantification of the intensity relative to tubulin. F MeRIP-qRT-PCR analysis of m6A levels of STAT3 in HCCLM3 and MHCC97H cells treated with the indicated concentration of STM2457 for 48 h. G Western blotting analysis of STAT3 in HCCLM3 and MHCC97H cells treated with the indicated concentration of STM2457 for 48 h. H qRT-PCR analysis of STAT3 mRNA stability in MHCC97L cells upon treatment with transcription inhibitor Actinomycin-D (ActD) for the indicated timepoints. The cells were transfected with empty vector or METTL3. I Western blotting analysis of STAT3 in HEK293T cells time-dependently treated with 100 μg/ml cycloheximide (CHX) after being transiently transfected with the indicated plasmids. The upper panel is the quantification of the intensity of STAT3 relative to tubulin. J qRT-PCR analysis of STAT3 mRNA enrichment in anti-RPL10A immunoprecipitated RNA in MHCC97L cells transfected with the indicated plasmids. K qRT-PCR analysis of STAT3 mRNA enrichment in anti-RPL10A immunoprecipitated RNA in MHCC97H cells transfected with the indicated siRNAs. All experiments were repeated at least three times. Error bars represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001 by 2-tailed Student’s t-test

Next, we investigated the impacts of translation-associated m6A reader proteins, including YTHDF1/3, YTHDC2, and IGF2BP1/2/3 on the expression of STAT3. We observed that both of them had no obvious effect on STAT3 expression level (Fig. S1E-L). Interestingly, Gregory et al. reported that METTL3 could recruit eIF3 to the translation initiation complex and facilitate the translation of m6A-contained mRNA independent of methyltransferase activity or downstream m6A reader proteins [29]. eIF3b is the core subunit of the translation initiation factor eIF3 [29]. We performed a RIP assay and found that overexpression of wild-type of METTL3 and catalytic mutant METTL3 both effectively enhanced eIF3b-immunoprecipitated STAT3 mRNA (Fig. S1M, N). However, the catalytic mutant METTL3 could not upregulate the protein level of STAT3 while the wild-type of METTL3 did (Fig. S1O). This data was consistent with the previous reports, which showed that catalytic mutant METTL3 could interact with the translation initiation machinery but not elevate the protein level of target genes [30]. The above data indicated that the methyltransferase activity of METTL3 was indispensable for the upregulation of STAT3.

STAT3 regulates the nuclear localization of METTL3 via WTAP

Given the significant association between METTL3 and STAT3 in HCC, we explored whether STAT3 played a potential role in modulating the expression and/or function of METTL3. qRT-PCR and Western blotting analysis showed that silencing STAT3 did not affect the expression of METTL3 (Fig. S2A, B). Previous studies corroborate that METTL3 is primarily localized in nuclear speckle, convenient for the methyltransferase activity of METTL3 in regulating the splicing and export of mRNA [31, 32]. Thus, we investigated the impact of STAT3 on the cellular distribution of METLL3. The nuclear and cytosolic fractions were analyzed in the stable-knockdown-STAT3 cells and control cells. STAT3 depletion decreased the nuclear level of METTL3 and increased its cytosolic level to a certain extent, while the overall protein level of METTL3 was not altered (Fig. 4A). Immunofluorescence staining showed that compared with the non-deleted cells, the nuclear localization of METTL3 was decreased in the STAT3-deleted cells (Fig. 4B, Fig. S2C), suggesting that STAT3 might modulate the cellular distribution of METTL3 by interacting with METTL3. However, the Co-IP assay displayed that STAT3 could not interact with METTL3 (Fig. S2D, E). WTAP and ZC3H13 have been reported to modulate the cellular distribution of METTL3 [31, 33]. Yang et al. found that WTAP could interact with METTL3 and was responsible for nuclear speckle localization, essential for methyltransferase activity of METTL3 [31]. Diao et al. showed that ZC3H13, a zinc-finger protein, played a crucial role in anchoring WTAP in the nucleus to modulate RNA m6A modification [33]. We examined the effect of STAT3 on WTAP and ZC3H13 expression levels. STAT3 overexpression and knockdown significantly upregulated and downregulated WTAP expression, but the expression levels of ZC3H13 were unaltered (Fig. 4C-F, Fig. S2F, G). Thus, we speculated that WTAP might be involved in STAT3-modulated nuclear localization of METTL3. Notably, subcellular fractionation and immunofluorescence staining analyses both demonstrated that overexpression of WTAP could prevent nuclear export of METTL3 induced by STAT3 depletion (Fig. 4G-I). Collectively, STAT3 regulates the nuclear localization of METTL3 via WTAP.

Fig. 4

STAT3 regulates the nuclear localization of METTL3 via WTAP. A Western blotting analysis of METTL3 in the whole-cell lysate (WCL), cytoplasmic (Cyto), and nuclear (Nuc) fractions from the HCCLM3 (MOCK), HCCLM3-knockdown-control (Control), and HCCLM3-STAT3-knockdown (KD-STAT3) cells. β-actin and Lamin B1 were used as cytoplasmic and nuclear markers, respectively. The right panel shows the quantification of the intensity relative to β-actin or Lamin B1. B Immunofluorescence analysis of METTL3 (green), STAT3 (red) and DAPI (blue) in HCCLM3-STAT3-knockdown and HCCLM3-control cells. Scale bar, 10 μm. C qRT-PCR analysis of STAT3, WTAP, and ZC3H13 in MHCC97L cells transfected with the indicated plasmids. D qRT-PCR analysis of STAT3, WTAP and ZC3H13 in STAT3 knockdown and control MHCC97H cells, STAT3 knockdown and control HCCLM3 cells. E Western blotting analysis of WTAP and STAT3 in MHCC97L cells transfected with the indicated plasmids. F Western blotting analysis of WTAP, ZC3H13, and STAT3 in STAT3 knockdown and control MHCC97H cells, STAT3 knockdown and control HCCLM3 cells. G Western blotting analysis of METTL3 in the whole-cell lysate (WCL), cytoplasmic (Cyto), and nuclear (Nuc) fractions from the HCCLM3-knockdown-control, HCCLM3-STAT3-knockdown, and HCCLM3-STAT3-knockdown-WTAP-overexpression cells. β-actin and Lamin B1 were used as cytoplasmic and nuclear markers, respectively. The right panel shows the quantification of the intensity relative to β-actin or Lamin B1. H, I Immunofluorescence analysis of METTL3 (green), STAT3 (red) and DAPI (blue) in the indicated cells. Scale bar, 10 μm. (I) shows the quantification of nuclear fluorescence intensity of anti-METTL3 cell. All experiments were repeated at least three times. Error bars represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001 by 2-tailed Student’s t-test

STAT3 upregulates WTAP expression by stimulating its transcription

Given the effect of STAT3 on WTAP expression, we further explored whether STAT3 could act directly upon the promoter of WTAP to stimulate its transcription. Online resources (http://gpminer.mbc.nctu.edu.tw/, http://gene-regulation.com/pub/programs.html) were used to analyze the binding site of STAT3 on WTAP promoters. Five putative binding sites of STAT3 were identified on the promoter of WTAP. The ChIP assay was performed, and the specific primers were used to amplify the five regions containing the corresponding binding sites. The data showed that STAT3 could occupy the promoter of WTAP at region 3 (Fig. 5A, B). Then, we cloned a series of fragments of WTAP promoter (Fig. 5C). The luciferase reporter gene assay revealed that the activity of pGL-P4 was highest among the five fragments of WTAP promoter. Knockdown of STAT3 efficiently suppressed the activities of WTAP promoter except for pGL-P5 (Fig. 5C). This data suggested that the region (-630 ~ -480) of the WTAP promoter might contain the core sequence modulated by STAT3 (Fig. 5C). Notably, region 3 (-679 ~ -528) assessed by the ChIP assay was overlapped with the (-630 ~ -480) region confirmed by the luciferase reporter assay, indicating the significance of site 3 (-602 ~ -595) in activating the WTAP promoter. Next, we constructed the site3-deletion mutant based on pGL-P4 and performed a luciferase reporter gene assay. Site 4 (-454 ~ -447) and site 5 (-438 ~ -430) were separately deleted as negative controls (Fig. 5D). Compared with the wild-type (WT), the activity of Delete-3 significantly decreased. Moreover, the activity of Delete-3 was not altered in response to STAT3 knockdown, while Delete-4 and Delete-5 activities were significantly suppressed upon STAT3 knockdown (Fig. 5E). These findings reveal that STAT3 stimulates WTAP transcription via binding to site 3.

Fig. 5

STAT3 upregulates WTAP expression by stimulating its transcription. A The diagram shows the five regions which contain the binding sites of STAT3 in the WTAP promoter. B qRT-PCR analysis of the enrichments of five regions in anti-STAT3, Histone3, and IgG immunoprecipitated DNA in MHCC97H (left panel) and HCCLM3 (right panel) cells. C Activities of corresponding fragments of WTAP promoter were examined by luciferase reporter gene assay in HEK293T cells. The cells were transiently transfected with control siRNA or si-STAT3#1 along with indicated fragments of WTAP promoter. D The diagram of pGL3-basic luciferase reporter constructs containing the fragment of human WTAP promoter (pGL-P4). The constructs of wild-type (WT), Delete-3 (the binding site3 (-602 ~ -595) was deleted), Delete-4 (the binding site4 (-454 ~ -447) was deleted), and Delete-5 (the binding site5 (-438 ~ -430) was deleted) were separately shown in the diagram. E Luciferase reporter gene assay of WTAP promoter activities in HEK293T cells. The cells were transiently transfected with control siRNA or si-STAT3#1 and the WTAP promoter (WT) or constructs with deleted binding sites of STAT3. All experiments were repeated at least three times. Error bars represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001 by 2-tailed Student’s t-test

Furthermore, tissue microarray analysis uncovered a positive relationship between the expression levels of STAT3 and WTAP in cancer tissues and adjacent non-cancerous tissues (Fig. S3A, B). Similarly, Western blotting analysis in 4 paired HCC tissue specimens also revealed higher levels of STAT3 and WTAP in HCC tissues compared with non-cancerous tissues (Fig. S3C). TCGA database analysis by UALCAN (http://ualcan.path.uab.edu/) showed that WTAP mRNA levels in HCC tissues were significantly elevated than in non-tumor tissues (Fig. S3D). In addition, WTAP expression was closely related to the tumor stage of HCC (Fig. S3E). Moreover, Kaplan–Meier analysis [25] uncovered that the patients with higher expression levels of WTAP conversely had a shorter survival time (Fig. S3F, G). Taken together, these results indicate that WTAP is overexpressed and correlated with STAT3 in HCC tissues.

STAT3 modulates the methyltransferase function of METTL3 via WTAP

To further verify the relationship among STAT3, WTAP, and METTL3, we performed Co-IP assays in MHCC97H and HCCLM3 cells. Consistent with the previous study [31], we found that METTL3 could interact with WTAP (Fig. 6A, B). However, STAT3 depletion obviously diminished the interaction of WTAP with METTL3 by depressing WTAP expression (Fig. 6A, B). Next, we measured m6A levels in MHCC97H and HCCLM3 cells following STAT3 knockdown. The m6A level was downregulated upon STAT3 depletion (Fig. 6C, D). Interestingly, we observed that overexpression of WTAP efficiently alleviated STAT3 knockdown-induced decrease of m6A level, while METTL3 overexpression only alleviated this decrease to a certain extent (Fig. 6C, D). This might be because METTL3 overexpression could not compensate for the failure of nuclear speckle localization caused by WTAP deficiency. Overall, these data indicate that STAT3 modulates the methyltransferase function of METTL3 through WTAP.

Fig. 6

STAT3 modulates the methyltransferase function of METTL3 via WTAP. A, B Co-IP analysis of the interaction between METTL3 and WTAP in STAT3 knockdown and control MHCC97H cells (A), STAT3 knockdown and control HCCLM3 cells (B). The lower panels show the quantification of intensity relative to input. C, D The m6A level of total RNA in MHCC97H cells (C) and HCCLM3 cells (D) were determined by the m6A RNA methylation assay kit. The cells were transfected with the indicated siRNAs and/or plasmids. All experiments were repeated at least three times. Error bars represent mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001 by 2-tailed Student’s t-test

METTL3-STAT3 positive feedback loop promotes cell metastasis in vitro and in vivo

The above data revealed the comprehensive interplay between METTL3 and STAT3. They formed a positive feedback loop in HCC. Next, we intended to investigate the role of the feedback loop in HCC metastasis. Based on MHCC97H cells, we constructed different cell lines with luciferase signal, including stably depleted-METTL3, stably depleted-METTL3-overexpressed-STAT3, and corresponding control cells. The efficiencies of knockdown and overexpression were verified, as shown in Fig. 7A. The wound healing and transwell assays showed that METTL3 knockdown effectively inhibited the migration and invasive capacities of HCC cells, while overexpression of STAT3 countered the METTL3 depletion-induced inhibitory effect (Fig. 7B-F). Similarly, the functional assays in HCCLM3 cells after transient transfection showed consistent results as above (Fig. 7G-J). Then, we constructed an orthotopic xenograft model of HCC to evaluate the role of the METTL3-STAT3 feedback loop in HCC metastasis in vivo. The findings showed that the growth and weight of xenograft tumors were suppressed when METTL3 was silenced, and STAT3 overexpression could alleviate the METTL3 depletion induced-suppressive effect (Fig. 8A, B). Subsequently, we evaluated the expressions of proliferation marker Ki67, METTL3 and STAT3 in tumor tissues by IHC staining and Western blotting assay. IHC data uncovered that knockdown of METTL3 depressed the expression of Ki67 and STAT3. Meanwhile, STAT3 overexpression could rescue the reduced Ki67 expression caused by METTL3 knockdown (Fig. 8C, D). More importantly, lung metastasis was significantly reduced in mice with METTL3 knockdown compared to mice with control cells (Fig. 8E, F). Nonetheless, STAT3 overexpression could reverse METTL3 knockdown-induced reduction of lung metastasis (Fig. 8E, F). Overall, our findings suggest that the METTL3-STAT3 positive feedback loop promotes cell metastasis in HCC.

Fig. 7

METTL3-STAT3 positive feedback loop promotes cell metastasis in vitro. A Western blotting analysis of METTL3 and STAT3 in the indicated cells shows the corresponding interfere/overexpress efficiency. The right panel shows the quantification of intensity relative to tubulin. Error bars represent mean ± SD, *P < 0.05, **P < 0.01, ***P < 0.001 by 2-tailed Student’s t-test. B, C Wound healing assay of the migration ability in the indicated cells. D-F Transwell assay of the migration and invasion abilities in the indicated cells. G Wound healing assay of the migration ability in HCCLM3 cells transfected with indicated siRNAs and plasmids. H-J Transwell assay of the migration and invasion abilities in HCCLM3 cells transfected with indicated siRNAs and plasmids. Error bars represent mean ± SD, *P < 0.05, **P < 0.01, ***P < 0.001 by one-way ANOVA

Fig. 8

METTL3-STAT3 positive feedback loop promotes cell metastasis in vivo. A, B Imaging (A) and weights (B) of the xenograft tumors harvested from 5 mice per group. Error bars represent mean ± SD, *P < 0.05, **P < 0.01, ***P < 0.001 by one-way ANOVA. C IHC staining of Ki67, METTL3 and STAT3 in the tissues of xenograft tumors. Scale bar, 20 μm. The right panel shows the quantification of average optical density of Ki67, METTL3 and STAT3 in the tissues of xenograft tumors. *P < 0.05, **P < 0.01, ***P < 0.001 by 2-tailed Student’s t-test. D Western blotting analysis of the protein levels of METTL3, STAT3 and Ki67 in the indicated groups of tumor samples. The right panel shows the quantification of the intensity relative to tubulin. *P < 0.05, **P < 0.01, ***P < 0.001 by 2-tailed Student’s t-test. E The representative bioluminescence images were captured in the indicated time after transplantation. The lower panel shows the bioluminescence intensity (photons/s/cm2/sr) in the lungs was quantitatively analyzed. The significance of differences was assessed by one-way ANOVA, *P < 0.05, **P < 0.01, ***P < 0.001. F Representative images of hematoxylin–eosin (H&E) staining in lung tissue after transplantation