METTL14 suppresses proliferation and metastasis of colorectal cancer by down-regulating oncogenic long non-coding RNA XIST

Tumor tissues and adjacent normal tissues (> 5 cm away from tumor) were collected from 37 CRC patients undergoing radical surgery in our institute. Clinical parameters including age, sex, stage, pathological diagnosis, TNM status and recurrence were also collected. No patient had received local or systemic treatment before operation. Fresh tumor and paired normal tissues were frozen in liquid nitrogen and stored at − 80 °C. The rest part of specimens was fixed in the formalin and embedded in paraffin for pathological and immunohistochemical (IHC) analysis. Our study protocol has been approved by the Ethics Committee of Ruijin Hospital, Shanghai Jiao Tong University, School of Medicine. Informed consent was obtained from each enrolled patient.

Tissue sections were deparaffinized, rehydrated, and microwaved-heated in sodium citrate buffer (10 mmol/L, pH 6.0) for antigen retrieval. Then, the slides were incubated with primary antibody. Then the expression levels of target proteins in tissue were examined by two independent pathologists blinded to the clinical characteristics of the patients according to proportion of cell staining (0 = 0%, 1 = ≤ 25%, 2 = 26 to 50%, 3 = 51 to 75%, 4 = > 75% positive cells) and the staining intensity (0 = no staining, 1 = weak, 2 = moderate, 3 = strong). A final score was calculated by multiplying the above two scores. Protein expression was considered high if the final score was greater than 6 points and low if the final score was 6 points or less.

Human colon epithelial cell line NCM460 was purchased from the Cell Bank of Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). Five CRC cell lines (SW480, SW620, HCT116, LoVo, and HT29) were previously purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA) and preserved in the research institute of general surgery. SW480 and NCM460 cells were cultured in RPMI-1640 medium. HT29 and HCT116 were cultured in McCoy’s 5A medium, SW620 cells were cultured in Leibovitz’s L-15 medium. LoVo was cultured in Ham’s F-12 K (Kaighn’s) Medium. All kinds of medium were purchased from Invitrogen (Carlsbad, CA, USA) and supplemented with 10% fetal bovine serum (FBS, HyClone, Logan, USA), 100 U/ml penicillin and 100 μg/ml streptomycin. Cells were maintained in a 37 °C incubator with 5% CO₂. The full length METTL14 cDNA was synthesized by RT-PCR from normal colon epithelial cells (NCM460) and then sub-cloned into the pcDNA3.1 vector to construct pcDNA-METTL14 overexpression (METTL14-OE) plasmid. The shRNA sequences targeting METTL14 were synthesized with a well-established annealing method and then cloned into pLKO.1 plasmid (Sigma Aldrich, St. Louis, USA). SiRNAs (siMETTL14, siXIST, siWTAP) and negative control were purchased from HanYin Bio-Tech Co.,Ltd. (Shanghai, China). For transient transfection, cells were transfected with plasmids encoding target sequences or siRNAs using Lipofectamine 3000 reagent (Invitrogen) according to the manufacturer’s instructions. Stable clones were selected by puromycin (1 μg/ml). All transfects were tested regularly by western blot to ensure the efficiency of over-expression or knockdown. Details of sequences or hairpin sequences used in this study were listed in Additional file 2: Table S2.

Cell viability was measured with cell count assay and colony formation assay. For cell count assay, control and transfected cells were cultured in a 96-well plate (3000 cells/well). Triplicate wells were measured in each group. Cell viability was determined every 24 h. The plate was incubated at 37 °C for 2 h after each well was added with 10 μl CCK-8 solution. Then the spectrophotometric absorbance was measured at 450 nm for each sample. For the colony formation assay, a certain number of control or transfected cells were planted into a six-well plate and maintained in culture media containing 10% FBS for 2 weeks. Colonies were fixed with methanol and stained with crystal violet (Sigma-Aldrich). The colony number was determined by counting stained colonies using ImageJ software. Transwell invasion assay were performed as we described elsewhere. Briefly, 8 × 10⁴ cells in 200 μL serum-free media were seeded in the top chamber (8.0 μm pore size, Corning, USA) with membrane coated by Matrigel (BD Bioscience, USA). 600 μL of 10% FBS-containing medium was placed into the bottom chamber as an attractant. After incubation for 24 h, the cells that did not invade to the lower side of the chamber were removed from the top side. The chambers were then stained with crystal violet and photographed.

Briefly, cells were washed with PBS and lysed with RIPA buffer (Thermo Scientific, MA, USA). The lysates were then sonicated on ice and centrifuged at 14,000 g for 5 min at 4 °C to remove cell debris. Next, the supernatants were resolved in 12.5% SDS-PAGE (30 μg/lane) and transferred onto 0.22-lm polyvinylidene fluoride (PVDF) membranes (Millipore, MA, USA). The membranes were probed with proper antibodies overnight at 4 °C. After washing for three times, the membrane was incubated in HRP-labeled secondary antibody for 2 h at room temperature. Protein bands were revealed by enhanced chemiluminescence (ECL) method. The following antibodies used in this study were purchased from Cell Signaling Technology (Beverly, MA, USA): rabbit anti-METTL14 (D8K8W), rabbit anti-WTAP (56501), rabbit anti-GAPDH (D6H11).

Total RNA was extracted using TRIzol reagent (Ambion) and cDNA synthesis was performed using a reverse transcription kit (Promega, Madison, USA) according to the manufacturer’s instructions. PCR was conducted using the SYBR Green Master Mix (Applied Biosystems, MA, USA) and the Applied Biosystems 7900HT sequence detection system. After the reactions were complete, relative gene expression level was calculated using the 2^−ΔΔCt method. GAPDH was used as an endogenous control. Primers used were as follows: METTL14: sense: 5′-GAACACAGAGCTTAAATCCCCA-3′; antisense: 5′-TGTCAGCTAAACCTACATCCCTG-3′. XIST: sense: 5′-GCATAACTCGGCTTAGGGCT-3′, antisense: 5′-TCCTCTGCCTGACCTGCTAT-3′.

Cells cultured in 10 cm plate was washed twice with ice-cold PBS and scraped off in 1 mL PBS. Then the cell was centrifuged and re-suspended in an equal pellet volume of complete RIP lysis buffer (Merck Millipore). 5 μg antibody was pre-bound to Protein A/G magnetic beads in immunoprecipitation buffer (20 mM Tris-HCl pH 7.5, 140 mM NaCl, 0.05% TritonX-100) for 2 h and then incubated with 100 μl cell lysates over night at 4 °C with rotation. Then RNA was eluted from the beads by incubating with 400 μl elution buffer for 2 h. The eluted RNA was precipitated with ethanol and dissolved with RNase-free water. Enrichment of certain fragments was determined by real-time PCR. Primer used for XIST quantification was designed as follows: sense: 5′-CTGCTGCAGCCATATTTCTTAC-3′, anti-sense: 5′-TACGCCATA AAGGGTGTTGG-3′. IgG was used for negative control. For m6A-RNA immunoprecipitation (Me-RIP), XIST extracted from equal amount cell lysates was used as input to measure the m6A-methylated rate of XIST. Antibodies used in this experiment were as follows: anti-m6A (ab190886, Abcam), anti-YTHDF1 (ab99080, Abcam), anti-YTHDF2 (ab170118, Abcam), anti-YTHDF3 (ab103328, Abcam), anti-YTHDC1 (ab122340, Abcam), anti-YTHDC2 (ab176846, Abcam). Anti-IgG (Cell Signaling Technology, #2729).

For RNA sequencing, purified RNA from shMETTL14 and control cells was used for library construction with Illumina TruSeq RNA Sample Prep Kit (FC-122-1001) and then sequenced with Illumina HiSeq 2000. Raw reads were aligned to the human genome GRCh37/hg19 by Bowtie2. Differentially expressed genes (DEGs) between treatment and control samples were identified with limma-voom method. A heatmap clustered by k-means was used to show DEGs or transcripts.

Establishment and analysis of nude mice (male BALB/c nu/nu nude mice, 4-week-old) subcutaneous xenograft model was performed as we described before [21]. For liver metastasis model, mice were anaesthetized and an incision was made through the skin and peritoneum to expose the spleen. 1 × 10⁶ HCT116 cells were injected into the spleen (n = 4 each group). All mice were kept until death due to the neoplastic progression or until the end of the experiment (6 weeks). The mice livers were removed and carefully dissected to evaluate the metastatic lesions. Animal experiments were all carried out according to the Guide for the Care and Use Laboratory Animals of Ruijin Hospital, Shanghai Jiao Tong University School of Medicine.

R software (version 3.5.1) was used for statistical analysis and data visualization. Protein and RNA levels were compared using two-tailed Student’s t test. Correlations between METTL14 expression in CRC tissues and clinicopathological features were analyzed with Pearson Chi-square (χ2) test. Overall survival (OS) and recurrence free survival (RFS) was assessed by Kaplan-Meier method, difference between survival curves was determined by log-rank test. Differences with a p value < 0.05 were considered as statistically significant.

To investigate the potential role of m6A in colorectal cancer, we first detected the mRNA levels of major m6A methyltransferase including METTL3 and METTL14 in 37 CRC and paired normal samples. As shown, METTL14 was remarkably down-regulated in cancerous compared to paired normal samples (Fig. 1a, left panel). In contrast, no significant difference was observed in the expression level of METTL3 (Fig. 1a, right panel). Moreover, similar expression pattern of METTL14 and METTL3 (Fig. 1b) in CRC were validated in an expanded cohort containing 387 cases of CRC patients from Gene Express Omnibus (GEO) dataset (GSE14333), which further supported our initial findings. However, analysis of RNA sequencing data from The Cancer Genome Atlas (TCGA) shown decreased METTL14 (Additional file 3: Figure S1A) but elevated METTL3 levels (Additional file 3: Figure S1B) in human CRC tissue relative to normal tissue. Due to the inconsistency of METTL3 analysis results, we chose METTL14 for further research in this study. Tissue microarray containing all 37 pairs of cancerous and matched normal tissue was analyzed by immunohistochemical (IHC) staining (Fig. 1c). Consistently, high METTL14 expression was observed in 67.6% (25/37) of normal tissues. Meanwhile, in paired CRC tissue, only 32.4% (12/37) cases showed high METTL14 signal (p = 0.002). Besides, loss of METTL14 was also observed in multiple human CRC cell lines (Fig. 1d). With above evidence, we concluded that METTL14 was frequently down-regulated in human CRC and might be implicated in pathogenesis and progression of CRC.

Next, we sought to explore the clinical significance of METTL14 by performing correlation analysis of METTL14 levels with clinicopathological features of enrolled CRC patients. As shown, low expression of METTL14 was positively correlated with larger tumor size, lymphatic invasion, remote metastasis and more advanced TNM stage (Table 1), indicating METTL14 might play an important role in modulating proliferation and invasion of CRC. Moreover, Kaplan-Meier analysis revealed that patients with low METTL14 expression exhibited unfavorable recurrence free survival (RFS, Fig. 1e). Moreover, we established an independent validation cohort by extracting RNA-sequencing and clinical data of CRC patients from GSE14333 dataset. Consistent with our results, decreased METTL14 expression also implicated worse RFS (Fig. 1f). In addition, multivariate analysis showed that METTL14 was an independent risk factor for RFS of CRC patients from our cohort (Table.2) and GEO dataset (Fig. 1g). Time-dependent ROC analysis showed that METTL14 had stronger predictive ability than TNM stage, especially in predicting 24-month RFS (Fig. 1h, Additional file 4: Figure S2A). Furthermore, TCGA data also showed that METTL14 was positively correlated with OS (Additional file 4: Figure S2B) and represented as an independent risk factor (Additional file 4: Figure S2C). These results suggested METTL14 is a reliable prognostic marker of CRC patients.

Given the fact that METTL14 is down-regulated in tumor tissue, we speculated that METTL14 may act as a tumor suppressor in CRC. To define the functional roles of METTL14 in CRC, we established METTL14-knockdown cell model in HCT116 and HT29 cells with two independent siRNAs (Fig. 2a, Additional file 4: Figure S2D). The alteration of proliferative ability of CRC cells was then evaluated by cell count assay and colony formation assay, respectively. As expected, depletion of METTL14 markedly enhanced proliferative ability of CRC cells, reflected by expediated growth rate (Fig. 2b) and increase of colony number (Fig. 2c) of siMETTL14 cells. In addition, transwell invasion assay showed that knockdown of METTL14 dramatically promoted invasive ability of CRC cells (Fig. 2d). Moreover, two shRNAs targeting METTL14 (Additional file 4: Figure S2E) were designed and transfected into CRC cell lines to further validate the alterations of phenotypes of CRC cells. Consistently with above results, shMETTL14 cells demonstrated enhanced growing (Additional file 5: Figure S3A, S3B) and invasive (Additional file 5: Figure S3C) ability compared to control cells in cell count assay, colony formation assay and transwell invasion assay. To test our in vitro findings, we established shMETTL14 cell models and injected HCT116-shMETTL14 and control cells subcutaneously in the flank of nude mice to evaluate the effect of METTL14 on CRC tumorigenicity (Fig. 3a). Consistent with our in vitro observations, knockdown of METTL14 resulted in significant increase of the tumor volume (Fig. 3b) and tumor weight (Fig. 3c). Besides, significant higher ki-67 rate was observed in the xenograft tumors formed by shMETTL14 cells by IHC staining (Fig. 3d and e). Furthermore, we generated a liver metastasis model by injecting tumor cells into the spleen of nude mice. As expected, more metastatic nodules were found in shMETTL14 group compared to their control counterparts (Fig. 3f). Collectively, these results suggested knockdown of METTL14 promoted proliferation and invasion of CRC in vitro and in vivo.

We next sought to investigate the mechanism by which METTL14 suppressed malignant phenotype of CRC cells. As a key m6A methyltransferase, METTL14 induces installation of methyl group on m6A residues of target RNAs and activates downstream signaling. Moreover, recent studies demonstrated that pri-miRNAs could be marked by m6A-methylation for subsequent splicing and maturing [22]. Besides, m6A process was also proved to participate in modulating process of lncRNAs [23]. These reports not only highlight the important role of m6A modification in RNA processing, but also remind us that m6A-methylation can affect cancer biology by regulating non-coding RNAs (ncRNAs). We then conducted gene set enrichment analysis (GSEA) to explore the potential downstream pathways of METTL14. As shown, METTL14 was positively correlated with ncRNA metabolism (Additional file 6: Figure S4A), ncRNA processing (Additional file 6: Figure S4B) and RNA degradation (Additional file 6: Figure S4C). In contrast, negative correlation was seen between METTL14 and several signaling pathways namely growth factor binding, collagen trimer and collagen binding (Additional file 6: Figure S4D-4F). Therefore, METTL14 may exert anti-tumor activity via targeting and down-regulating oncogenes or oncogenic lncRNAs in these pathways. We then performed RNA-seq to interrogate the expression changes in HCT116-shMETTL14 cells relative to the control cells. The fold change of gene expression was calculated and genes with log2|FC| > 0.5 were considered as differentially expressed. Among thousands of potential targets, we noticed lncRNA XIST was significantly upregulated upon METTL14 knockdown (Fig. 4a, Additional file 7: Figure S5A). XIST is a potent oncogenic lncRNA expediting growth and metastasis of CRC cells [19]. Importantly, analysis of transcriptome m6A mapping data revealed that at least 78 m6A residues are located across XIST sequence (Additional file 7: Figure S5B and Additional file 1: Table S1) [20]. As typical mRNA only contains 3–5 m6A sites, the high abundance of m6A sites of XIST implied an incredibly important role of m6A in regulating of XIST function. Thus, we assumed that METTL14 inhibited CRC growth and metastasis by down-regulating lncRNA XIST. To verify this hypothesis, expression levels of XIST was first detected in METTL14 knockdown and control cells. As expected, METTL14 depletion led to distinct elevation of XIST in two CRC cell lines (Fig. 4b). Next, we quantified XIST levels in cancer and normal specimens. Consistent with reported results, XIST exhibited higher levels in cancer compared to normal tissues (Fig. 4c), which suggested a tumor-driving effect of XIST in CRC. Moreover, a negative correlation was observed between expression levels of METTL14 and XIST in CRC samples from our cohort (Fig. 4d). Besides, TCGA RNA-seq data of CRC also showed that XIST upregulation often happened in cases with downregulation of METTL14 (Fig. 4e). Additionally, enhanced cell growth (Fig. 4f, g and Additional file 7: Figure S5C) and invasion (Additional file 7: Figure S5D) induced by METTL14 knockdown could be attenuated by XIST inhibition. Notably, these results were also tested in shMETTL14 cells to confirm that knockdown of XIST diminished the growth (Additional file 8: Figure S6A, S6B) and invasion (Additional file 8: Figure S6C) of shMETTL14 cells. Taken together, these results indicated that increased CRC growth and invasion resulted from METTL14 depletion was, at least partly, due to elevation of lncRNA XIST.

We next plan to clarify whether the function of METTL14 down-regulating XIST directly depends on its m6A catalytic activity. METTL14 was ectopically over-expressed in HCT116 and HT29 cells (Fig. 5a, Additional file 9: Figure S7A), the potential alterations of XIST expression level and cell phenotypes were then assessed. As shown, overexpression of METTL14 resulted in remarkable decrease in XIST expression level (Fig. 5b), cell growth (Fig. 5c, d) and invasion (Fig. 5e and Additional file 9: Figure S7B) of CRC cells, which were consistent with the above results and confirmed the suppressive effects of METTL14 on CRC. Recently, METTL3 and METTL14 are identified as interactors with Wilms tumor-associated protein (WTAP). WTAP acts as a scaffold and binds the methyltransferase to form a complex which mediates m6A methylation on RNAs. The loss of WTAP significantly blocks m6A modification process. Thus, we performed WTAP knockdown with siRNA in METTL14-overexpressed (METTL14-OE) cells (Fig. 5a, Additional file 9: Figure S7A) to inhibited the cellular m6A process and analyzed the effect of this act. As shown, the decreased XIST level induced by METTL14 overexpression could be attenuated by WTAP depletion (Fig. 5b). In line with this, WTAP knockdown significantly rescued the impaired growth (Fig. 5c and d) and invasion (Fig. 5e and Additional file 9: Figure S7B) of METTL14-OE cells. These results suggested that the downregulation of XIST induced by METTL14 was dependent on the formation of m6A complex and subsequent m6A methylation process. Moreover, m6A-methylated XIST was immunoprecipitated with m6A antibody in siMETTL14 and siWTAP cell lysates and then quantified with real-time PCR. We found that knockdown of METTL14, or WTAP led to significant reduced levels of methylated XIST (Fig. 5f). These results directly showed that METTL14 promotes m6A-methylation of XIST, resulting in downregulation of this lncRNA and suppression of CRC proliferation and invasion. Intriguingly, we noticed that knockdown of WTAP almost totally abolished the m6A-methylation of XIST while METTL14 depletion significantly but only partly decreased m6A levels (Fig. 5f), which suggested that there might be other mechanisms involved.

m6A methylation is a marking procedure which needs to be recognized by m6A “reader” proteins for further disposition of target RNAs [24]. We next investigated the reader protein which recognized the m6A-methylation of XIST in order to mediate its downregulation. Analysis of previously reported iCLIP-sequencing data identified direct binding between XIST and the m6A readers YT521-B homology (YTH) domain family [20] which comprise three members of the YTHDF proteins (YTHDF1, YTHDF2, and YTHDF3), YTHDC1 and YTHDC2. To determine the binding between XIST and the potential reader proteins in CRC cells, we immunoprecipitated YTH family proteins with proper antibodies from the cell lysates and then measured the amount of bound XIST by quantitative PCR. As shown, YTHDF2 immunoprecipitants contained significantly higher level of XIST than control group (Fig. 6a and Additional file 10: Figure S8A). Consistently, RNA pull-down assay showed that YTHDF2, but not other readers, was significantly enriched by biotin-labeled XIST (Fig. 6b and Additional file 10: Figure S8B). Given that YTHDF2 was reported to facilitate m6A-dependent RNA degradation under normal and stress conditions [25], we hence speculated that YTHDF2-dependent m6A-RNA decay played an important role in the downregulation of XIST. As expected, knockdown of YTHDF2 resulted in remarkable augment of XIST expression in HCT116 and HT29 cells (Fig. 6c), which validated our hypothesis. Moreover, expression levels of YTHDF2 and XIST was negatively correlated with each other in TCGA dataset and our cohort (Fig. 6d). In addition, the decay rate of XIST was significantly slower in shYTHDF2 CRC cells, which directly showed YTHDF2 mediated the degradation of lncRNA XIST (Fig. 6e and Additional file 10: Figure S8C). Taken together, these findings suggested that METTL14-induced m6A process suppressed XIST expression through YTHDF2-dependent RNA degradation.