METTL14-mediated N6-methyladenosine modification of SOX4 mRNA inhibits tumor metastasis in colorectal cancer

The raw gene expression data in CRC were downloaded from The Cancer Genome Atlas (TCGA) (http://cancergenome.nih.gov) and GEO database. The independent data sets from (GSE9348 [23], GSE44076 [24‐28], GSE41657) were analyzed in this study.

Human normal colonial epithelial cell lines (NCM460) and colorectal cancer cell lines (HCT116, HCT8, SW620, SW480, HT29 and DLD-1) were all obtained from American Type Culture Collection (ATCC). NCM460, HCT116 and HCT8 cells were cultured in RPMI-1640 supplemented with 10% fetal bovine serum (FBS, Hyclone, USA), and SW620, SW480, HT29 and DLD-1 were maintained in Dulbecco’s modified Eagle’s medium (DMEM) with 10% FBS. All these cells were cultured at 37 °C with 5% CO2.

A total of 136 CRC and corresponding adjacent normal tissues (ANTs) were collected from Nanjing First Hospital, Nanjing Medical Hospital. Our study was approved by the Institutional Review Board of Nanjing First Hospital, and written informed consent were obtained from all patients prior to our study. The patients who have achieved system treatment were not permitted in this study. The clinical characteristics in 136 CRC patients was presented in Table 1.

For transwell migration and invasion assays, CRC cells were seeded into the upper chamber without (transwell migration assay) or with (transwell invasion assay) matrigel (BD Biosciences, USA). After 24 h of incubation, non-migrated or invaded CRC cells were scraped off using a cotton swab, and CRC cells on the bottom of chamber were fixed with methanol for 10 min, and stained using 0.5% crystal violet. Then 5 fields(× 200 magnification) were selected and photographed randomly using an inverted microscope (Nicon, Japan). The experiments were performed in triple.

TRIzol Reagent (Invitrogen, USA) was employed to extract total RNA from CRC tissues and cells following manufacturer’s instructions. The mRNA levels was assessed using PrimeScript RT reagent Kit and SYBR Premix Ex Taq (Takara, Dalian, China). All results were normalized to GAPDH. The relative expression of mRNAs was quantified using the 2^–∆∆Ct method. The primers used are listed in Additional file 1: Table S1.

The full-length complementary cDNAs of human METTL14 and SOX4 were synthesized and cloned into the pcDNA3.1(Invitrogen, China). The small hairpin RNA (shRNA) targeting KDM5C, METTL14, SOX4, YTHDF1, YTHDF2 and YTHDF3 were designed and synthesized by GenePharma (Shanghai, China). The shRNA of SOX4, METTL14 and their negative control were synthesized and cloned into the pGLVH1/GFP/Puro vector (GenePharma, China). The plasmids were transfected into CRC cells using lipofectamine 3000(Invitrogen, USA) in accordance with the protocol. The sequences of shRNAs were supplemented in Additional file 1: Table S1. To achieved the METTL14 and SOX4 stable knockdown cell line, HCT116 cells were infected with LV-shMETTL14–1, LV-shSOX4 and LV-NC, and selected using 10 μg/ml puromycin.

To measure RNA stability in METTL14 stable knockdown or control HCT116 cells, actinomycin D (MCE, USA) at 5 μg/ml was added to cells, and the cells were collected after incubation at the indicated times (0, 1, 2, 4, 8 h), and RNA was isolated from these cells for qRT-PCR.

The chromatin immunoprecipitation (ChIP) assay kit (Beyotime, China) was employed to fulfill the ChIP assay following the manufacturer’s instruction. In brief, CRC cells were collected and soniacated to generate DNA fragments ranging from 200 to 500 bp. Then the lysate was immunoprecipitated with anti-KDM5C, anti-H3K4me3 or IgG antibodies (negative control) overnight. Immunoprecipitated DNAs were extracted and analyzed by qPCR. The 2000 bp upstream and 500 bp downstream of the METTL14 promoter were divided into eight parts(C1, C2, C3, C4, C5, C6, C7, C8), and the ChIP primer sequences were listed in Additional file 1: Table S3.

All animal experiments were approved by the animal care Committee of Nanjing First Hospital, Nanjing Medial University (acceptance No. SYXK 20160006). 2 × 10⁶ transfected HCT116 cells in 0.2 ml PBS were injected into the tail vein of nude mice which were randomly divided into nine groups (eight mice per group). After 2 months of injection, mice were sacrificed, and their lungs were removed and stained by Hematoxylin and Eosin (HE) Staining.

A complete description of the methods, including Western blot, Immunohistochemistry (IHC) analysis, RNA m6A dot blot, RNA immunoprecipitation (RIP), RNA-Seq and MeRIP-Seq and Luciferase Reporter Assays are available in Additional file 2: supplementary materials and methods.

In order to detect the expression of METTL14 in CRC, we first sought to determine the METTL14 expression levels from TCGA database, the results showed that the expression of METTL14 was significantly downregulated in CRC tissues (Fig. 1a). We also complied the gene expression from GEO dataset (GSE9348, GSE44076, GSE41657), and confirmed that the METTL14 expression levels were decreased in CRC tissues (Fig. 1b). Moreover, METTL14 was examined by qRT-PCR and IHC on samples from 136 CRC patients compared with corresponding ANTs, and elevated METTL14 expression both on mRNA and protein levels were observed in CRC tissues (Fig. 1c, d). Furthermore, downregulated expression of METTL14 was significantly associated with Lymph node metastasis(P = 0.018), distant metastasis(P = 0.002) and TNM stage(P = 0.005) (Table 1). Survival analysis using Kaplan-Meier method indicated that CRC patients with low METTL14 expression exhibited a worse overall survival (OS) (Fig. 1e). Unvariate and multivariate analyses showed that METTL14 expression could be an independent prognostic for OS (Fig. 1f). Collectively, these results indicated that METTL14 was significantly downregulated in CRC and might be associated with CRC progression.

To uncover the mechanism of low METTL14 expression in CRC, we analyze the ChIP-Seq data of H3K4me3 in the Encyclopedia of DNA Elements (ENCODE) database, and found that the enrichment of H3K4me3 in the promoter region was markedly lower in CRC cells than that in normal colorectal tissues (Fig. 2a). KDM5C belongs to KDMs family, and could catalyze H3K4me2/3 demethylation and suppress gene transcription via decreasing H3K4 methylation [29, 30]. Using UALCAN database (http://ualcan.path.uab.edu/analysis.html), we found that KDM5C mRNA expression was markedly upregulated in CRC tissues (Additional file 3: Fig. S1a), in agreement with the results from TCGA database, KDM5C mRNA expression levels was elevated in our CRC cohort (Additional file 3: Fig. S1b), and the expression of METTL14 was negatively correlated with the KDM5C expression levels (Additional file 3: Fig. S1c). We then treated HCT116 and HCT8 cells with KDM5C inhibitor (KDM5A-IN-1), and found that both RNA and protein levels of METTL14 were significantly upregulated after the treatment (Fig. 2b, c). Moreover, we knocked down KDM5C, and found that KDM5C-deleption significantly elevated the expression levels of METTL14 (Fig. 2d, e). Furthermore, the ChIP results showed that the promoter of METTL14 was enriched in KDM5C binding and H3K4me3 signals, and knockdown of KDM5C could obviously increase the enrichment of H3K4me3 signals in the promoter of METTL14 (Fig. 2f, g). Our data indicated that KDM5C-mediated histone H3K4me3 loss in the promoter region might account for the decrease of METTL14.

To investigate the function role of METTL14 in CRC cells, we first examine METTL14 expression levels in CRC cell lines, consistent with the results from CRC tissues, METTL14 was significantly downregulated in CRC cell lines compared with non-CRC cell line (Fig. 3a). Next, METTL14 was knocked down using two shRNA targeting METTL14(shME-1,shME-2), while overexpressed METTL14 using METTL14 expression vector (pcDNA3.1-METTL14) in HCT116 and HCT8 cells (Fig. 3b). Transwell migration assay revealed that forced expression of METTL14 apparently impeded the migratory ability of HCT116 and HCT8 cells, attenuation of METTL14 expression significantly elevated the migration speed of HCT116 and HCT8 cells (Fig. 3c). Correspondingly, transwell invasion assay showed that the invasive ability of HCT116 and HCT8 cells was significantly suppressed in response to METTL14 upregulation, while it was obviously enhanced by knockdown of METTL14 (Fig. 3d). Furthermore, we assessed the physiological relevance of METTL14 to CRC metastasis in vivo, stable cells with modified METTL14 expression were tail-vein injected into the BABL/c nude mice, after 8 weeks, HCT116 cells with METTL14 deficiency dramatically facilitated CRC cell metastasis (Fig. 3e), as evidenced by the number of lung metastatic lesions, in contrast, overexpression of METTL14 markedly inhibited CRC cell metastasis, as shown by the number of lung metastatic lesions (Fig. 3f).

To identify the molecular mechanism by which METTL14 inhibits CRC metastasis, we used RNA-seq in CRC cells with stable METTL14 downregulation and conducted MeRIP-seq in CRC cells with stable METTL14 downregulation and control cells. RNA-seq data showed that 2074 transcripts were markedly upregulated on METTL14 knockdown (Additional file 4: Table S1). MeRIP-seq presented that m6A peaks of 51 transcripts exhibited decreased abundance (fold change<− 1, P < 0.05) (Additional file 4: Table S2). Intriguingly, 25 transcripts were overlapped in the transcriptome-sequencing and MeRIP-seq data (Fig. 4a). Of these candidate genes, SOX4 was reported be closely with tumor progression, and was selected as a candidate target of METTL14-mediated m6A modification for further study. SOX4 was significantly upregulated in METTL14-knockdown HCT116 and HCT8 cells both on mRNA and protein levels (Fig. 4b, c).

The m6A usually happens in RRACH(R = G or A, H = A, C or U) consensus sequence, our MeRIP-seq results showed that an m6A peak was detected around the stop condon of SOX4 mRNA in nontarget control shRNA HCT116 cells and were all decreased upon METTL14 knockdown (Fig. 4d). Therefore, a preliminary conclusion was drawn that SOX4 was a METTL14 downstream target.

Through analyzing the transcriptome data in TCGA database, we found that SOX4 mRNA was markedly upregulated in CRC tissues in comparison with normal tissues (Fig. 4e), similar result was also found in GSE9348, GSE44076 and GSE41657 (Additional file 3: Fig. S2). Our qRT-PCR and IHC results further confirmed the change (Fig. 4f, g). In addition, we also found that the expression of METTL14 was significantly reversely correlated with SOX4 expression (Fig. 4h).

To further verify that METTL14 targets SOX4 mRNA via m6A modification, we first detected the global level of m6A in control and METTL14 stable knockdown group through m6A dot blot assay (Fig. 5a). As expected, m6A levels were obviously reduced with the deletion of METTL14 in HCT116 and HCT8 cells. Then, MeRIP-qPCR was conducted to measure the enrichment of m6A in SOX4. Our results showed that the m6A abundance in the SOX4 mRNA was dramatically increased on METTL14 overexpression as well as substantially diminished on METTL14 knockdown (Fig. 5b). To further determine the effect of m6A modification on SOX4 expression, luciferase reporters containing either wild-type or mutant SOX4 to address the effect of m6A modification on SOX4 expression. For the mutant form of SOX4, the adenosine bases in m6A consensus sequences (RRACH) were replaced by cytosine, thus m6A modification was abolished (Fig. 5c). Luciferase reporter assay revealed that transcriptional level of wild-type SOX4, but not the mutation, obviously increased in the absence of METTL14 (Fig. 5d), revealing that the regulation of SOX4 level was the under control of METTL14-related m6A modification.

Recently, studies have reported that YTHDF1/2/3, a distinct family of m6A readers that could target thousands of mRNA via recognizing m6A motif [31]. Therefore, we explore the effect of YTHDF1/2/3 on SOX4 mRNA stabilisation. Two specific shRNAs targeting YTHDF1/2/3 were employed and western blot was used to confirm the knockdown efficiency (Fig. 5e, f). We found that YTHDF2 knockdown strongly augmented SOX4 expression, while YTHDF1/3 had no effect on SOX4 in HCT116 and HCT8 cells (Fig. 5e, f, Additional file 3: Fig. S3a,b). To further investigate the mechanism regulating SOX4 expression through m6A modification, we firstly detected the expression of precursor (pre) and mature (mat) mRNA of SOX4 in METTL14 stable knockdown and control HCT116 cells, and the results showed that both pre and mat mRNA expression levels were substantially enhanced in METTL14 stable knockdown cells (Fig. 5g). We then treated METTL14 stable knockdown and control HCT116 cells with actinomycin D to block transcription. Our results revealed that half-life of both precursor and mature mRNA in control cells were markedly shorter than in METTL14 stable knockdown cells (Fig. 5h, i). It indicated that m6A modification might trigger the splicing of precursor mRNA and the degradation of mature mRNA of SOX4. As shown in Fig. 5j, when compared to IgG control, YTHDF2-specific antibody substantially reduced the enrichment of SOX4 mRNA in the RIP assay. Above results confirmed that SOX4 was a target of YTHDF2. Moreover, we found that there was no differences in the expression of YTHDF2 between in CRC tissues and non-tumor tissues. (Additional File 3: Fig. S3c). The results showed that METTL14 couldn’t affect YTHDF2 expression levels, but could raise YTHDF2 to m6A modified sites of SOX4 mRNA, and modulated the expression of SOX4. Collectively, METTL14-mediated m6A modification repressed SOX4 expression through YTHDF2-dependent mRNA degradation.

To further investigate the oncogenic function of SOX4 in CRC, the expression of SOX4 in HCT116 and HCT8 cells was strikingly downregulated using shRNA targeting SOX4(shSOX4) (Fig. 6a). Then the loss-of-function experiments were employed, transwell migration assays revealed that SOX4 upregulation lead to increase CRC cell migration (Fig. 6b, Additional file 3: Fig. S4a), and transwell invasion assay strongly showed that SOX4 overexpression elevated the ability of CRC cell invasion (Fig. 6c, Additional file 3: Fig. S4b). To further validate the association between METTL14 and SOX4, we then studied the effects of METTL14 upregulation on cell migration and invasion after SOX4 overexpression, and found that the reduced migratory and invasive capabilities of HCT116 and HCT8 cells caused by METTL14 overexpression could be extensively retarded by SOX4 upregulation (Fig. 6d, e, f, Additional file 3: Fig. S4c, 4d). Mice inoculated with SOX4 stable knockdown HCT116 cells had less metastatic foci in the lungs than mice that received control cells, while there were fewer metastatic foci in the lungs of mice after injecting SOX4-silencing HCT-116 cells in comparison with control groups (Fig. 6g). In addition, SOX4 overexpression partially restored the metastasis capacity inhibited by METTL14 upregulation (Fig. 6h).

A number of studies have verified that SOX4 controlled the TGF-β-induced epithelial-to-mesenchymal (EMT) [14, 32], a process closely associated with increases in invasive and migratory capacity of tumor cells. In the present study, we found that depletion of METTL14 elevated the expression of N-cadherin and Vimentin as well as reduced E-cadherin expression levels, indicating that disruption of METTL14 promotes EMT process, while silencing of SOX4 could reverse this promotion caused by METTL14 knockdown (Fig. 7a). Moreover, It has been reported that SOX4 is an crucial activator of MAPK, PI3K-Akt and Wnt signaling [33]. Here, we found that METTL14 knockdown facilitated the phosphorylation of the PI3K and Akt proteins, indicating that METTL14 knockdown activated the PI3K/Akt signaling, and SOX4 silencing partially reversed the activation of the PI3K/Akt pathway induced by METTL14 knockdown via western blot assays (Fig. 7b).

LY294002, a chemical inhibitor of PI3K, which has been previously reported to block PI3K/Akt signaling [34]. Here, we found that p-Akt expression levels was significantly reduced in LY294002 treated group (Additional file 3: Fig. S5a). Moreover, the abilities of migration and invasion in HCT116 and HCT8 cells treated with LY294002 were significantly decreased in comparison with that in control group (Additional file 3: Fig. S5b, c). Taken together, we concluded that METTL14 exerted its biological function via inactivating SOX4 mediated PI3K/Akt pathway.