METTL3/YTHDF2 m6A axis promotes tumorigenesis by degrading SETD7 and KLF4 mRNAs in bladder cancer

Abstract N6‐Methyladenosine (m6A) modification, the most prevalent modification of eukaryotic messenger RNA (mRNA), is involved in the progression of various tumours. However, the specific role of m6A in bladder cancer (BCa) is still poorly understood. In this study, we demonstrated the tumour‐promoting function and specific regulatory mechanism of m6A axis, consisting of the core ‘writer’ protein METTL3 and the major reader protein YTHDF2. Depletion of METTL3 impaired cancer proliferation and cancer metastasis in vitro and in vivo. Through transcriptome sequencing, m6A methylated RNA immunoprecipitation (MeRIP) and RIP, we determined that the METTL3/YTHDF2 m6A axis directly degraded the mRNAs of the tumour suppressors SETD7 and KLF4, contributing to the progression of BCa. In addition, overexpression of SETD7 and KLF4 revealed a phenotype consistent with that induced by depletion of the m6A axis. Thus, our findings on the METTL3/YTHDF2/SETD7/KLF4 m6A axis provide the insight into the underlying mechanism of carcinogenesis and highlight potential therapeutic targets for BCa.


| INTRODUC TI ON
BCa in the United States; thus, BCa ranks the forth in incidence and eighth in mortality in men. 2 The increasing trend in these numbers constantly urges researchers to better understand the mechanisms underlying the pathogenesis of BCa to identify potential therapies against BCa. m 6 A, a modification first identified in mRNA-enriched RNA fractions in 1974, 3 refers to methylation of the N6 position of adenosine bases, which are widely distributed in the mammalian mRNA. 4,5 With the application of available methods for detecting m 6 A, insights into the regulatory mechanism have been revealed in recent years. m 6 A RNA modification is a dynamic and reversible posttranscriptional modification process maintained by a multicomponent methyltransferase 'writer' complex (METTL3, METTL14 and their cofactors such as WTAP, etc) [6][7][8] and removed by demethylase 'eraser' (FTO and ALKBH5). 9,10 The function of m 6 A in mRNA metabolism primarily depends on reader proteins, which include YT521-B homology (YTH) domain family (YTHDF1-3, YTHDC1-2) and IGF2 mRNA binding protein family. 11 Recognition by YTHDF2 could induce mRNA degradation 12 and the bindings of YTHDF1 and YTHDF3 could initiate the translation of m 6 A-containing transcripts 13,14 while identification by IGF2BP family could promote the stability of target mRNA. 15 Studies documenting the importance of m 6 A modification in physiological bioprocesses and the pathogenesis of diseases have emerged in large numbers, mainly involving the regulation of cell fate decisions, embryonic development, and different kinds of cancers. 11,16 For example, knockout of m 6 A catalysing enzyme METTL3 is embryonic lethal during mouse development due to the failed termination of murine naïve pluripotency. 17 Conditional knocking out the m 6 A reader protein YTHDF2 in mice, delayed degradation of neuron differentiation-related m 6 A-containing mRNAs impaired the self-renewal of neural stem/progenitor cells and the spatiotemporal generation of neurons leading to neural development deficiency. 18 Moreover, the overexpression of FTO in acute myeloid leukaemia could prevent the cell differentiation but promote oncogene-mediated hematopoietic cell transformation and enhances leukemogenesis. 19 However, the mechanism of tumorigenesis influenced by m 6 A modification dysregulation in BCa remains elusive.
In this study, we explored the biological role of m 6 A modification in BCa and principally focused on the regulatory mechanism of the m 6 A axis mediated by METTL3 and YTHDF2. We demonstrated that the m 6 A/SETD7/KLF4 axis was involved in the progression of BCa.
Consequently, we hope that this study will offer an insight into the underlying molecular mechanism of tumorigenesis and provide potential therapeutic targets for BCa.

| Cell culture and transfection
The human BCa cell lines T24, UM-UC-3 and the normal human
To avoid off-target effects, different siRNAs of genes were merged together as the siRNA pool to cotransfect BCa cell lines in all interference experiments.

| Western blot assay
Total proteins were extracted by lysing cells in 6-well plate with 60 μL RIPA buffer supplemented with protease inhibitors (#G2006, Wuhan Goodbio Technology) and phosphatase inhibitors (#G2007, Wuhan Goodbio Technology) and quantified by BCA protein assay kit (#P0011, Beyotime). One microgram of proteins was separated by 10% SDS-PAGE and transferred to PVDF membrane (#IPVH00010, Merck Millipore). The membrane was blocked with 5% Non-Fat Milk (#A600669, Sango Biotech) for 1 hour at room temperature and incubated with specific antibody at 4°C overnight followed by HRP-conjugated secondary antibody incubation for 1 hour at room temperature. Images of membrane with proteins were taken with imager (Bio-rad ChemiDoc MP, Bio-rad). The antibodies are recorded: anti-SETD7 (#ab14820, Abcam).
Reverse transcription was proceeded with PrimeScript RT reagent Kit (#RR036A, Takara), and quantitative PCR (qPCR) was performed by SYBR Premix Ex Taq (#RR820A, Takara) in thermocycler (CFX96 Touch Real-Time PCR, Bio-rad). GAPDH was chosen as the endogenous reference for calculating the relative expression of target genes with 2 −ΔΔC t (delta-delta-C t algorithm) method. All

| Cell viability assay
When the confluence was up to 30%-50% per well in 96-wells plates, cells were transfected with merged RNA duplexes at different concentration as 0, 25, 50, 75 and 100 nmol/L. After 48 and 72 hours treatment, cell viability was detected by cell counting kit 8 test (#CK04, Dojindo) according to the protocol.

| Colony formation assay
Five hundred treated cells (transfected with siRNAs or plasmids for 48 hours) per well were counted and seeded in 6-well plates.
After cultured for 7-14 days, cells were fixed with methanol and then stained with 0.3% crystal violet.

| Trans-well assay
Nearly 5 × 10 4 transfected cells were seeded in the trans-well chamber (#MCEP24H48, Merck Millipore) with 0.2 mL serum-free medium and the compartment between chamber and 24-plate well was filled with 0.6 mL medium with 10% FBS. After culturing for 24 hours, the migrated cells were fixed on the lower surface of the membrane with methanol and stained with 0.3% crystal violet.
Images were taken with phase-contrast microscopy (IX71, Olympus) with a 20 × objective lens.

| Wound healing assay
When treated cells grew to 100% confluence, the pipette tip (200 μL) was used to scratch the cell layer on plate to create a break.
When the cellular proliferation obstructed by serum-free medium, the cellular migration process was monitored by phase-contrast microscope (IX71, Olympus) with 5 × objective lens.

| RNA m 6 A dot blot assay
RNA was isolated from cells or tissues using RNAiso plus (#9109, Takara), and the concentration was adjusted to 50 ng/μL in 36 μL RNase-free water. Secondary structure was removed using RNA incubation buffer (mixture of MOPS, formamide and formaldehyde).

| m 6 A-RNA immunoprecipitation assay
Total RNA was extracted from two 15-cm plates of cell lines infected by lentiviruses with the shRNA using RNAiso plus (#9109, Takara), and the process was completed according to the instructions of Magna methylated RNA immune-precipitation (MeRIP) m 6 A Kit (#17-10499, Merck Millipore). In brief, RNA was diluted to 1 μg/μL and sheared into ~100 nt in length. After precipitation at −80°C overnight, nearly 300 μg (except 10% of input) of RNA in 300 μL of RNase-free water was incubated with magnetic beads loaded with 10 μg of antibody of m 6 A (#ABE572, Merck Millipore) or immunoglobulin G (IgG) for 2 hours at 4°C. Then, the fragmented RNA containing m 6 A was eluted and purified with RNA purification kit (#74204, QIAGEN). Enriched fragments were analysed by RT-qPCR. The primers were designed from the 100 nt around the predicted m 6 A peak (http://www.cuilab. cn/sramp) and listed: SETD7 F 5′-CTGGCTTTGGGGTTCAGAGA-3′;

| RNA-binding protein immunoprecipitation assay
For RIP assay, we primarily used Magna RIP Kit (#17-700, Merck Millipore) according to the instructions. Cells cultured in two 15-cm plates were lysed, precipitate removed and incubated with magnetic beads loaded with 7.5 μg of antibody of YTHDF2 (#24744-1-AP, Proteintech) or IgG overnight at 4°C. And RNA was isolated from binding protein by proteinase K and purified followed by analysis by RT-qPCR.

| RNA sequencing
Total RNA was extracted from cells using RNAiso plus (#9109, Takara). The purity of RNA was assessed using the ND-1000 NanoDrop, and the integrity was evaluated using the Agilent 2200 TapeStation (Agilent Technologies). In brief, rRNAs were removed by Epicentre Ribo-Zero rRNA Removal Kit (Illumina). The remains were fragmented to approximately 200 bp, purified and subjected to first strand and second strand cDNA synthesis following by adaptor ligation and enrichment with a low-cycle according to instructions of NEBNext ® Ultra™ RNA Library Prep Kit for Illumina (NEB). The purified I library products were evaluated using the Agilent 2200 TapeStation and Qubit ® 2.0 (Life Technologies) and then diluted to 10 pmol/L for cluster generation in situ on the pair-end flow cell followed by sequencing (2 × 150 bp) HiSeq3000. After removal of reads containing adapter, ploy-N and at low quality from raw data, the clean reads were obtained. The software HISAT2 was used to align the clean reads with default parameters, and HTSeq was employed to convert aligned short reads into read counts for each gene model. Differential expression was analysed by DEseq. The Benjamini-Hochberg multiple test correction method was enabled and the chosen criteria of fold change >2 and adjusted P-value < .05 was obeyed. The differentially expressed genes were ranked by P-value, and top 200 were used for heat map in website: (http:// www.ehbio.com/Image GP/index.php/Home/Index /index.html). For KEGG enrichment analysis, a P-value < .05 was used to determine significant enrichment of the gene sets.

| Animal experiments
Male BALB/c nude mice (4 weeks old) were used in this study.

| Statistical analysis
Statistics in this study are described as the mean ± SD in GraphPad prism. Statistical distinction between two groups was estimated by a two-tailed Student's t test. Statistical significance was defined as P-value of <.05.

| METTL3 and YTHDF2 are highly expressed in BCa
To investigate the expression pattern of METTL3 and YTHDF2, data sets in the TCGA database and Oncomine online database 21 (https://www.oncom ine.org) were analysed. We observed that the transcripts of METTL3 and YTHDF2 were promoted in primary tumours compared to normal subjects ( Figure 1A,B). In addition, at different cancer stages, the expression differed and displayed an increasing-tendency as disease malignancy increased ( Figure 1C). We reported the crucial role of the m 6 A axis conducted by 'writer' and reader, 22,23 we performed the correlation analysis between 'writer' proteins and reader proteins ( Figure S2B,C). Further correlation analysis also indicated a positive correlation between the expression of METTL3 and YTHDF2 ( Figure 1E). However, the depletion of METTL3 did not directly affect the expression level of YTHDF2 ( Figure S2A). Thus, we speculated that METTL3 cooperated with YTHDF2 and the METTL3/YTHDF2 m 6 A axis was involved in BCa.

| Depletion of METTL3 impairs the proliferation and migration of BCa cells
To exclude off-target effects, we depleted METTL3 by using a homolog-specific siRNA pool ( Figure 2A) and rescue experiments with siRNA-resistant exogenous METTL3 were conducted ( Figure S3B). We In addition, down-regulation of METTL3 markedly weakened the migratory capabilities of BCa cell lines, which was confirmed by a wound healing assay ( Figure 3A) and trans-well assay ( Figure 3B).

| YTHDF2 promotes the migration of BCa cancer cells
An increasing number of studies have illustrated the importance of the m 6 A reader protein YTHDF2 in the regulation of physiological processes and diseases. 18,20 In the analysis of the TCGA database and detection in expression levels compared to SV-HUC-1 cells, YTHDF2 was predicted to be as a tumour promoter in BCa. To verify this, YTHDF2 was depleted in T24 and UM-UC-3 cells by using a homolog-specific siRNA pool ( Figure 4A). To exclude off-target effects, rescue experiments with siRNA-resistant exogenous YTHDF2 were conducted ( Figure S3C). The depletion led to marked inhibition of cells migration, as shown by the wound healing ( Figure 4B) and trans-well assays ( Figure 4C). In expression levels, a robust reduction in the migration rate and the expression levels of EMT pathway-related proteins and matrix metalloproteinase family protein was observed after silencing YTHDF2 ( Figure 4D). Conversely and concordantly, overexpression of YTHDF2 by plasmids promoted cell migration ( Figure S1A-D).

The observation of similar effects after manipulating METTL3
and YTHDF2 expression supports the concept that the axis consisting of m 6 A writer and m 6 A reader is likely to exist in BCa, analogous to a previous study in hepatocellular carcinoma. 23

| METTL3/YTHDF2 axis directly degrades SETD7 and KLF4 mRNAs in an m 6 A dependent manner
Recently, substantial studies have demonstrated the main catalytic core of METTL3 in producing m 6 A modification 6 ; we obtained evidence of this catalysis in BCa by using an m 6  As predicted by the SRAMP database (http://www.cuilab. cn/sramp), there were at least three highly probable m 6 A-modification peaks in these two targets ( Figure 5F). To confirm the mutual interaction between METTL3 and the two downstream targets, MeRIP was performed. In contrast to the control groups, a significant decrease of mRNA enrichment by m 6 A-specific antibody was detected in the METTL3 stably interfered UM-UC-3 cell lines ( Figure 5G). These results confirmed that the m 6 A-modification on target mRNAs was transferred by METTL3, and this finding was supported by Western blot and RT-qPCR, in which depleted METTL3 enhanced the expression of SETD7 and KLF4 ( Figure 5H,I). As expected, in rescue experiments, silencing SETD7 or KLF4 abrogated the tumour-suppressive phenotypes by depletion of METTL3 ( Figure S4A,B). The accumulated expression of SETD7 and KLF4 upon METTL3 depletion was also discovered in UM-UC-3 derived xenograft tumours obtained from nude mice ( Figure S1E).
Massive studies have confirmed that METTL3 could cooperate with YTHDF2 not only in regulating cell differentiation 25 but also causing tumour suppressor gene mRNA decay to impact the development and progression of tumour. 23 Notably, as expected, compared with IgG pull-down control, the methylated SETD7 and KLF4 F I G U R E 5 METTL3/YTHDF2 m 6 A axis directly degrades the SETD7 and KLF4 mRNAs. A, Dot blot assay demonstrates that the m 6 A level alters with changes in the expression of METTL3. B, Heat map shows representative differentially expressed genes upon METTL3depletion from RNA-seq. Low level is marked blue (|log 2 FC| > 1 and P-value < .05, arranging from smallest). C, KEGG enrichment analysis of differentially expressed genes upon METTL3 depletion from RNA-seq. D, The overlap of mRNAs significantly up-regulated (log 2 FC > 1 and P-value < .05) in T24 cells upon the depletion of METTL3 and genes negatively correlated with METTL3 in TCGA database analysed by LinkedOmics is shown by Venn diagram. E, Gene correlated with METTL3 in LinkedOmics database and negative correlation of METTL3 and SETD7 (r = −.2519, P < .0001) or KLF4 (r = −.2054, P < .0001) in primary BCa from TCGA database by Pearson correlation. F, Potential m 6 A modification sites along mRNA of SETD7 and KLF4 predicted by SRAMP. The peak noted with red arrow was valid in MeRIP assay. G, MeRIP assay shows the decreased amount of two target mRNAs collected by m 6 A antibody upon the repressed expression of METTL3, suggesting the direct methylation of METTL3. H, Representative Western blots demonstrating up-regulation of SETD7 and KLF4 protein upon METTL3 depletion by siRNA pools in cancer cell lines. I, The quantification of SETD7 and KLF4 mRNA abundance in METTL3 depleted T24 cells by RT-qPCR indicates that the down-regulation of METTL3 induced the up-regulated expression of SETD7 and KLF4. GAPDH was used as the normalization. J, RIP assay showing the binding between YTHDF2 and SETD7 or KLF4. K, Representative Western blots documenting the accumulated expression of SETD7 and KLF4 upon down-regulated YTHDF2 by siRNA pools in cancer cell lines. L, The quantification of SETD7 and KLF4 mRNA abundance upon depleted YTHDF2 in T24 cells by RT-qPCR indicating the negative correlation between YTHDF2 and two targets. GAPDH was used as the normalization. Statistical significance was determined by Student's t test: *P < .05, **P < .01, ***P < .001 mRNAs were abundant by the YTHDF2-specific antibody in the RIP assay ( Figure 5J). The recognition and direct binding of YTHDF2 were also demonstrated at the protein and RNA levels ( Figure 5K,L).
And in rescue experiments, knock-down of SETD7 or KLF4 rescued repressed migration ability by depletion of YTHDF2 ( Figure S4C,D).
In summary, all these findings indicated that METTL3-dependent m 6 A-modification depressed the expression of SETD7 and KLF4 via YTHDF2-mediated mRNA degradation.

| SETD7 and KLF4 act as tumour suppressors in BCa
As observed in the analysis of the TCGA database, the expression of SETD7 and KLF4 was repressed in BCa compared to normal samples ( Figure 6A,B). The decreased expression was also observed in distinct cancerous cell lines ( Figure 6C). Our previous studies have demonstrated the explicit anti-cancer role of KLF4 in BCa, 24 which has also been acknowledged in other urologic cancers, such as prostate cancer 26 and renal cancer. 27 Therefore, we focused on exploring the mechanism of SETD7. Monitored by wound healing assay ( Figure 6E) and trans-well assay ( Figure 6F), deletion of SETD7 using the homolog-specific siRNA pool resulted in decreased migration activity, while overexpression of SETD7 via plasmids presented the opposing results ( Figure 6I,J).
Moreover, the tumour metastasis related genes whose levels were substantially affected by SETD7 were consistent with the roles of METTL3 and YTHDF2 in EMT progression ( Figure 6G,K). The phenomenon further highlighted the interaction between SETD7 and METTL3/YTHDF2.

| M 6 A/SETD7/KLF4 axis is established in regulating BCa progression
Collectively, these data emphasized the importance of SETD7 and In these studies, m 6 A modification is implicated in tumorigenesis, which is the core issue of concerned. Currently, m 6 A modification has been found to be associated with leukaemia, 29 liver cancer, 30 brain tumour, 31 breast cancer 32 and lung cancer. 16,33 For a better interpretation of molecular mechanisms, a complete modification process mediated by the cooperation of m 6 A writers, erasers and readers, rather than a single isolated gene, is needed. For instance, METTL3 was documented to degrade SOCS2 mRNA to promote the progression of liver cancer in a YTHDF2-mediated m 6 A-dependent manner. 23 Methyltransferase METTL14, demethylase ALKBH5 and reader protein YTHDF3 were discovered to alter m 6 A levels in breast cancer, influencing the cell cycle, epithelial-mesenchymal transition (EMT) and angiogenesis of cancer cells. 22 Moreover, m 6 A-increased translation of SNAIL mediated by YTHDF1 was confirmed as a part of EMT altered by disorder of METTL3 in cancer cell. 34 However, few studies connect m 6 A modification with BCa, especially involving mutual collaboration between these m 6 A regulators.
Here, we demonstrated that disturbances in the balance of m 6 A modification mediated by the METTL3/YTDHF2 axis played a crucial role in tumour cell proliferation and migration thus influencing the tumorigenesis of BCa. When the expression of METTL3 was repressed, cells were primarily arrested in G1 phase and hardly proliferated or migrated in vitro; these cells had decreased m 6 A levels in total RNA. In vivo, concordantly, the depleted expression of METTL3 triggered the inhibition of tumorigenesis and metastasis.
Considering the identified modification process 11 and a previous study on liver cancer, 23 we showed that cells with down-regulated YTHDF2 were deficient in mRNA clearance capabilities and had impeded migration, which was in accordance with some of the functional roles of METTL3. Consequently, we explained that YTHDF2 recognized the m 6 A modification catalysed by METTL3 and completed the regulation of downstream targets.
In our transcriptome sequencing and MeRIP-RT-qPCR confirmation, SETD7 and KLF4 were perceived as the critical downstream targets of METTL3 and were indeed regulated in an m 6 A-dependent manner. The negative correlation between targets and METTL3 reminded us of mRNA degradation mediated by the m 6 A reader YTHDF2. Interestingly, our data showed direct binding between YTHDF2 and the mRNA of targets. Therefore, we concluded that a METTL3/YTHDF2-SETD7/KLF4 m 6 A axis exists in BCa. Herein, we investigated the dysregulation of SETD7 and KLF4 was partially inhibited by m 6 A modification, which was consistent with the KLF4m 6 A regulation in previous studies. 35 Actually, the roles of SETD7 and KLF4 in cancers are controversial. As a member of histone lysine methyltransferase, SETD7 is also an important modifier of non-histone proteins. While SETD7 probably drives the progression of gastric cancer, 36 the modification of SETD7 would favour the tumour suppressor function of E2F1 in breast cancer. 37 The important transcription factor KLF4, acted as a suppressor in metastasis and proliferation of cancer cells in colorectal cancer, 38 while driven expression of KLF4 in tumour-associated macrophages promoted immunosuppression to facilitate tumour growth in glioblastoma. 39 However, by means of this study and our previous research, 24 we demonstrated that SETD7 and KLF4 were two evident tumour suppressor genes in BCa. Consequently, the degradation of tumour suppressor genes by the m 6 A regulatory axis probably plays a vital role in the tumorigenesis of BCa.
In summary, we make a compelling illustration that the overexpressed METTL3 in BCa, the major methyltransferase catalysing m 6 A modification of mRNA, disrupts the dynamic balance of m 6 A F I G U R E 7 M 6 A/STED7/KLF4 axis is established in regulating BCa progression. In BCa, m 6 A modification is assembled on mRNA by METTL3 with other accessory catalytic subunits and eliminated by FTO and ALKBH5. YTHDF2 specifically recognizes m 6 A-mRNA and induces the degradation of target mRNAs, promoting tumorigenesis eventually modification, which is perceived as one of the vital reasons for cancer proliferation and cancer metastasis. In the m 6 A modification mechanism, our studies emphasize that the METTL3/YTHDF2-SETD7/KLF4 axis plays an indispensable part. This m 6 A regulatory axis contains different types of modification, especially mRNA and protein, revealing a sophisticated network in the regulation of gene expression. Prospectively, these findings may supplement the underlying molecular mechanism of gene regulation in the tumorigenesis of BCa and provide a promising strategy for BCa therapy in the future.

| CON CLUS IONS
Altogether, our study demonstrated a complete and important m6A regulatory pathway in BCa, containing the core 'writer' protein METTL3 and the major reader protein YTHDF2. The m 6 A modification catalysed by METTL3 is recognized by YTHDF2 to mediate the mRNA decay of tumour suppressors SETD7 and KLF4, which consequently induces the BCa progression. The METTL3/YTHDF2-SETD7/ KLF4 m 6 A axis is an intricate network of multiple modifications and could be a potential target for the diagnosis and treatment of BCa.

CO N FLI C T O F I NTE R E S T
The authors declare that there are no conflicts of interest in connection with the work submitted.