Dynamic m6A mRNA methylation reveals the role of METTL3-m6A-CDCP1 signaling axis in chemical carcinogenesis

N6-methyladenosine (m6A) is the most abundant internal modification in mammalian mRNAs. Despite its functional importance in various physiological events, the role of m6A in chemical carcinogenesis remains largely unknown. Here we profiled the dynamic m6A mRNA modification during cellular transformation induced by chemical carcinogens and identified a subset of cell transformation-related, concordantly modulated m6A sites. Notably, the increased m6A in 3′-UTR mRNA of oncogene CDCP1 was found in malignant transformed cells. Mechanistically, the m6A methyltransferase METTL3 and demethylases ALKBH5 mediate the m6A modification in 3′-UTR of CDCP1 mRNA. METTL3 and m6A reader YTHDF1 preferentially recognize m6A residues on CPCP1 3′-UTR and promote CDCP1 translation. We further showed that METTL3 and CDCP1 are upregulated in the bladder cancer patient samples and the expression of METTL3 and CDCP1 is correlated with the progression status of the bladder cancers. Inhibition of the METTL3-m6A-CDCP1 axis resulted in decreased growth and progression of chemical-transformed cells and bladder cancer cells. Most importantly, METTL3-m6A-CDCP1 axis has synergistic effect with chemical carcinogens in promoting malignant transformation of uroepithelial cells and bladder cancer tumorigenesis in vitro and in vivo. Taken together, our results identify dynamic m6A modification in chemical-induced malignant transformation and provide insight into critical roles of the METTL3-m6A-CDCP1 axis in chemical carcinogenesis.


Introduction
The dynamic modifications on mRNAs are essential for posttranscriptional regulation of gene expression [1,2]. N6-methyladenosine (m 6 A), the most abundant internal mRNA modification, has important roles in regulating mRNA splicing, export, stability, and translation [1,2]. m 6 A is catalyzed by an RNA methyltransferase complex consisting of the catalytic subunit METTL3 and its assisting partners METTL14 and WTAP [3,4]. On the other hand, two demethylases, ALKBH5 and FTO, are responsible for the removal of m 6 A from mRNAs [5,6]. Several different m 6 A reader proteins, including the YTHDF and YTHDC family members, can specifically recognize m 6 A modification and regulate the processing, degradation, and translation of m 6 A-modified mRNAs [7][8][9][10][11]. The dynamic interplay between m 6 A pathway components cooperatively regulates fine-turning mRNA metabolism and translation.
Chemical carcinogenesis is a multistage process that leads to the malignant tumor formation in animals and humans [26]. Chemical carcinogens can bind to DNA and induce genetic mutations in cancer-susceptibility genes, resulting in the development of carcinogenesis [27]. Importantly, exposure to chemical carcinogens also leads to epigenetic changes such as aberrant DNA methylation or histone modifications, which coordinate with the genetic mutations to initiate oncogenesis [27,28]. The m 6 A modification pathway is frequently mis-regulated in cancers; however, the role of m 6 A modification in chemical-induced cellular transformation remains unknown.
Here we studied the functions and the underlying molecular mechanisms of m 6 A mRNA modification in chemical carcinogenesis. We found that chemical carcinogen treatment induces dynamic changes of m 6 A modification and identified oncogene CDCP1 as a critical m 6 A target in chemical-induced transformation. We further showed that the METTL3-m 6 A-CDCP1 axis interplays with chemical carcinogens to promote the malignant transformation. Our data uncovered novel epitranscriptomic mechanisms in chemical carcinogenesis.

Topology of m 6 A mRNA methylome during cell malignant transformation
To study the role of m 6 A pathway during cell malignant transformation, we first performed the transcriptome-wide profiling of m 6 A modification using the chemical carcinogen-induced transformation model. The human uroepithelial SV-HUC-1 cells, human prostate epithelial RWPE-1 cells, and human bronchial epithelial 16HBE cells were treated with chemical carcinogens (Cadmium (Cd), 3methylcholanthrene, and Nickel) to induce malignant transformation [29,30] (Figure S1). Then total RNA samples from the normal and transformed cells were isolated for m 6 A profiling using m 6 A methylated RNA immunoprecipitation sequencing (MeRIP-Seq). Metagene and distribution analysis of the m 6 A modification revealed that m 6 A peaks are enriched mostly in CDS and 3′-intranslated region (UTR) region (Fig. 1b). The density of m 6 A peaks increases steadily along transcript in CDS. Conversely, in the 3′-UTR, the peaks decrease in abundance along the length of the 3′-UTR (Fig. 1a). Comparison of m 6 A distribution between normal and malignant cells revealed that chemical treatment has little effect on the m 6 A distribution in mRNAs (Fig. 1a). Motif searching identified the consensus "GGAC" motif within the m 6 A sites (Fig. 1c). Overall, our m 6 A profiling results are consistent with the published m 6 A features, suggesting that we have successfully identified the specific m 6 A sites in the control and malignant cells.

Identification of chemical transformation-induced m 6 A targets
We next investigated the differentially modified targets in the normal and transformed cells with m 6 A MeRIP-Seq data from four different sets of comparable cells: SV-HUC-1 control and Cd-transformed cells (SV-HUC-1 vs. Cd-SV-HUC-1), SV-HUC-1 control and 3-methylcholanthrenetransformed cells (SV-HUC-1 vs. MC-SV-HUC-1), RWPE-1 control and Cd-transformed cells (RWPE-1 vs. Cd-RWPE-1), and 16HBE control and Nickel-transformed cells (16HBE vs. NSTC2). As shown in Fig. 1d, e, chemical transformation induced dynamic changes of m 6 A modifications in hundreds of mRNAs in each set of control and transformed cells. Overlapping of the upregulated (Fig. 1d) and downregulated (Fig. 1e) sites revealed the dynamic and diverse m 6 A changes in different chemical-mediated transformation models. Gene ontology analysis revealed that the common chemical transformation induced m 6 A gene function in regulation of RNA processing and metabolic process, protein process, cell cycle, and cell death. On the other hand, the common genes with decreased m 6 A modifications in the transformed cells function in RNA processing regulation (Fig. 1f). Comparison of m 6 A profiles in control and transformed cells identified many cancer-promoting genes as m 6 A targets during chemical carcinogenesis. For example, the m 6 A modification on the mRNA of oncogene CDCP1 was upregulated in all four transformation models (Fig. 1g), suggesting the potential role of chemical-induced m 6 A modification of CDCP1 in regulation of oncogenic transformation. CDCP1 mRNA translation, but not degradation, was mediated by the m 6

A modification
The dysregulated expression of oncogene CDCP1 promotes cancer transformation and progression in vitro and in vivo [31]. Therefore, we further studied the molecular mechanisms and functions of CDCP1 m 6 A modification in chemical transformation. m 6 A MeRIP-qRT-PCR (quantitative reverse-transcriptase PCR) confirmed that the CDCP1 m 6 A (II) analysis using 90 genes with downregulated m 6 A peaks in both Cd and MCA-induced transformed cells. Each corresponding set of targets was subjected to DAVID GOBP analysis and an enrichment map was built by Cytoscape with Enrichment Map Apps. Each node denotes one enriched GOBP pathway (p < 0.005, FDR q < 0.1, overlap cutoff > 0.5), with its color reflecting the p-value. Node size is proportional to the total number of genes in each pathway. GOBP pathways with similar function were sorted into one cluster, marked with circles and labels. G representative differentially regulated m 6 A of CDCP1 mRNA in chemical transformation; IP immunoprecipitation modification is significantly increased upon chemical transformation (Fig. 2a), whereas the CDCP1 mRNA stability remained unchanged (Fig. 2b). We further compared the CDCP1 mRNA and protein levels in the chemicaltransformed cells and other bladder cancer cells. Interestingly, CDCP1 protein expression, but not its mRNA, is upregulated in transformed cells and bladder cancer cells (Fig. 2c, d). Overall, these data suggested the potential posttranscriptional regulation of CDCP1 expression through m 6 A modifications during chemical carcinogenesis.

The m 6 A methyltransferase METTL3 promotes CDCP1 mRNA modification and translation
The m 6 A methyltransferase METTL3 catalyzes m 6 A modifications to facilitate target mRNA degradation or translation. Given that CDCP1 mRNA stability and level remained unchanged but CDCP1 protein increased in the chemicaltransformed cells, we tested the possibility of METTL3mediated m 6 A modification in regulation of CDCP1 mRNA translation. Overexpression of METTL3 in SV-HUC-1 increased the CDCP1 protein level (Fig. 2e, f) but had little effect on CDCP1 mRNA (Fig. 2g). METTL3 overexpression significantly promoted the m 6 A modification in CDCP1 mRNA and the mRNA of known m 6 A target SON, but not the negative control HPRT1 mRNA (Fig. 2h), suggesting that METTL3-mediated m 6 A modification promotes the CDCP1 protein expression. On the other hand, depletion of METTL3 in the transformed MC-SV-HUC-1 cells and bladder cancer T24 cells resulted in decreased level of CDCP1 protein but not the mRNA (Fig. 2i, j, S2A, S2B). Moreover, similar to chemical transformation, METTL3 overexpression has little effect in the stability of CDCP1 mRNA (Fig. 2k). As controls, the stability of m 6 A containing SON mRNA is significantly downregulated in the METTL3-overexpressing cells ( Figure S3A), whereas the stability of the m 6 A-negative mRNA HPRT1 remain unchanged ( Figure S3B). To rule out potential effects of METTL3 on CDCP1 protein stability, we treated cells with cycloheximide (CHX) to block translation and measured CDCP1 degradation. Degradation rates of CDCP1 protein were significantly unchanged between control and METTL3-overexpressing cells (Fig. 2l). Furthermore, polysome-bound (translationally active) CDCP1 mRNA levels significantly increased in METTL3-overexpressing cells compared with control cells (Fig. 2m). In contrast, depletion of METTL3 in T24 cells significantly reduced the proportion of CDCP1 transcripts in polysomal fractions ( Figure S2C), suggesting that METTL3 enhances CDCP1 mRNA translation. Overall, our data indicated that METTL3 catalyzes the m 6 A modification on CDCP1 mRNA and promotes its translation upon chemical transformation.

m 6 A demethylase ALKBH5 regulates CDCP1 mRNA modification and translation
To study the functions and mechanisms of m 6 A modification in regulation of CDCP1 expression during chemical carcinogenesis, we determined the roles of m 6 A demethylases, ALKBH5 and FTO, in regulation of CDCP1 expression. Overexpression of ALKBH5 in bladder cancer J82 cells and UM-UC-3 cells resulted in decreased CDCP1 protein level without changing its mRNA (Fig. 3a-c). On the other hand, depletion of ALKBH5 increased level of CDCP1 proteins but not the mRNA (Fig. 3d, e). Interestingly, overexpression or depletion of FTO did not change CDCP1 expression ( Figure S4A-S4D), suggesting that ALKBH5 and FTO might have different molecular targets. Consistent with ALKBH5's function as an m 6 A demethylase, ALKBH5 depletion resulted in increased m 6 A modification on CDCP1 mRNA (Fig. 3f). Moreover, manipulation of ALKBH5 expression did not significantly change the stability of CDCP1 mRNA and protein (Fig. 3g, h). Overall, ALKBH5 demethylates CDCP1 mRNA and negatively regulates CDCP1 protein expression, further supporting the role of m 6 A pathway in regulation of CDCP1 mRNA modification and translation.  chemical transformation, we first analyzed the CDCP1 mRNA sequence and found three m 6 A motifs within the m 6 A peak region located in the 3′-UTR of CDCP1. We then cloned the CDCP1 3′-UTR region into a luciferase reporter to determine the function of these m 6 A motifs in regulation of gene expression. As shown in Fig. 4a, b, overexpression of METTL3 promoted the translation of luciferase reporters with longer region (F1 WT) or shorter region (F2 WT) of CDCP1 3′-UTR. Mutation of the first m 6 A motif (F1 MUT1, F2 MUT1) slightly decreased METTL3's function in promoting translation (Fig. 4b), whereas mutation of the three motifs (F2 MUT3) completely abolished the translation enhancement driven by METTL3 ( Figure. 4c, S5), suggesting that the three m 6 A motifs are essential for METTL3 to regulate CDCP1 expression. Interestingly, the catalytic mutant METTL3 (METTL3 MUT) can also promote the translation of CDCP1 reporters, although with weaker activity compared with the METTL3 wild type (WT) (Fig. 4b, c, S5), which is consistent with the previous report that the N-terminal METTL3 is sufficient to promote translation when tethered to target mRNAs [18]. In vitro translation using the control and m 6 A-modified reporter All bar plot data are means ± SEM of three independent experiments except f, where error bars denote SD of technical triplicates. *p < 0.05, **p < 0.01, ***p < 0.001 mRNAs further confirmed the role of m 6 A modification in promoting CDCP1 translation (Fig. 4d). Most importantly, transfection of the CDCP1 reporters into the control and chemical-transformed cells revealed that chemical-induced transformation significantly increased the WT CDCP1 reporter expression but not the m 6 A motif-mutated reporter (Fig. 4e). Overall, the above results supported that the chemical transformation induces the m 6 A modification of CDCP1 3′-UTR and promotes the translation of CDCP1 mRNA.

METTL3/YTHDF1 preferentially recognizes m 6 A residues on CDCP1 3′-UTR
We next sought to understand how the m 6 A machinery promotes CDCP1 translation upon chemical transformation. Therefore, we first tested whether YTHDFs bind to CDCP1 mRNA. RNA immunoprecipitation assay revealed that YTHDF1, but not YTHDF2 or YTHDF3, selectively binds to CDCP1 mRNA ( Fig. 5a-c, S6A). Moreover, forced expression of METTL3 significantly increased the interaction between YTHDF1 and CDCP1 mRNA, suggesting that METTL3 mediates the m 6 A modification and facilitates YTHDF1 binding to CDCP1 mRNA ( Fig. 5a-c, S6B). WT METTL3 protein binds to CDCP1 mRNA and, interestingly, the catalytic mutant METTL3 also interacts with the endogenous CDCP1 mRNA, although with less binding capacity compared with WT ( Fig. 5d, S6V, S6D). Using the CDCP1 reporter, we found that only the WT METTL3 can promote the m 6 A modification of CDCP1 reporter mRNA, whereas the METTL3 MUT has no catalytic activity (Fig.  5e). Similar to the findings in the endogenous CDCP1 mRNA, both METTL3 WT and MUT proteins interact with CDCP1 reporter mRNA and overexpression of METTL3 promotes the binding of YTHDF1 to CDCP1 reporter mRNA (Fig. 5f, g). Consistent with the functional assay (Fig. 4c), METTL3 and YTHDF1 only interact with the WT CDCP1 reporter but not the m 6 A motif mutated reporter (Fig. 5h, i, S6E, S6F). In addition, inhibition of METTL3 or YTHDF1, but not YTHDF2/3 expression, decreased CDCP1 expression (Fig. 5j, k) and knocking down both METTL3 and YTHDF1 had synergistic effect in reducing CDCP1 expression. Taken together, our data revealed that METTL3 and YTHDF1 bind to the m 6 A motifs in the CDCP1 3′-UTR and promotes CDCP1 mRNA modification and translation.
METTL3 and CDCP1 are upregulated in bladder cancers and correlated with bladder cancer progression Chemical carcinogens are important risk factors of bladder cancer. Our data identified METTL3-m 6 A-CDCP1 axis as a downstream target during chemical-induced transformation of uroepithelial cells; therefore, we evaluated the clinical relevance of METTL3-m 6 A-CDCP1 oncogenic signaling axis in bladder cancer with patient samples. Immunohistochemistry (IHC) staining of bladder cancer patient tumor microarray revealed that both METTL3 and CDCP1 are moderately or highly expressed in most of the bladder cancer samples, whereas their expression are weak or not detectable in the majority of the paratumor controls (Fig. 6a). We further investigated the association between METTL3-CDCP1 expression and bladder cancer progression. Determination of METTL3 and CDCP1 expression in human cystitis and bladder cancer tissues revealed that METTL3 and CDCP1 expressions are weak or not detectable in majority of the cystitis samples, although the METTL3 and CDCP1 are upregulated in the non-muscle-invasive bladder cancer samples. Moreover, the expression levels of METTL3 and CDCP1 are further elevated in the muscleinvasive bladder cancer samples (Fig. 6b). Taken together, METTL3 and CDCP1 expression are upregulated in human bladder cancer samples and the expression of METTL3 and CDCP1 are associated with bladder cancer progression.

METTL3-m 6 A-CDCP1 axis is essential for the growth and progression of bladder cancer
The upregulated METTL3 and CDCP1 in bladder cancers and the correlation between high METTL3/CDCP1 expressions with advanced bladder cancer stages suggested the important function of METTL3-m 6 A-CDCP1 axis in regulation of bladder cancers. Therefore, we further studied the function of METTL3-m 6 A-CDCP1 axis in the growth and progression of chemical-transformed uroepithelial cells and bladder cancer cells. Our results revealed that depletion of METTL3 or CDCP1 modestly decreased proliferation (Fig. 6c) and significantly decreased the migration (Fig. 6d) and invasion (Fig. 6e) of the chemical-transformed Cd-SV-HUC-1 cells. Data from the bladder cancer T24 cells also supported the essential role of the METTL3-m 6 A-CDCP1 axis in regulation of proliferation, migration, and invasion of bladder cancers ( Figure S8A-S8C). In addition, forced expression of CDCP1 partially rescued the effects of METTL3 depletion on migration (Fig. 6f) and invasion (Fig. 6g) in chemical-transformed Cd-SV-HUC-1 cells, further confirming CDCP1 as an essential downstream target of METTL3 in chemical carcinogenesis. Next, we studied the role of METTL3-m 6 A-CDCP1 axis in regulation of in vivo tumorigenesis using xenograft tumor models. Our data revealed that METTL3 depletion in T24 cells modestly reduced the bladder cancer growth in vivo (Fig. 6h-j). IHC analysis showed that the METTL3-depleted tumors had reduced CDCP1 expression (Fig. 6k). Overall, our in vitro and in vivo loss-of-function studies confirmed that the METTL3-m 6 A-CDCP1 axis is essential for the growth and progression of bladder cancer.

METTL3-m 6 A-CDCP1 axis promotes uroepithelial cells transformation and tumorigenesis
We next performed gain-of-function assays to study the role of METTL3-m 6 A-CDCP1 axis in uroepithelial transformation and bladder cancer tumorigenesis. As shown in Fig. 7a, overexpression of METTL3 significantly increased the proliferation of SV-HUC-1 uroepithelial cells. In addition, both METTL3 and CDCP1 can promote the migration of SV-HUC-1 cells (Fig. 7b). However, although METTL3-m 6 A-CDCP1 axis can promote the proliferation and migration of SV-HUC-1 uroepithelial cells in vitro, injection of METTL3 or CDCP1-overexpressing SV-HUC-1 cells into immunodeficient mice did not induce tumor formation in vivo (data not shown). Therefore, we further studied whether METTL3-m 6 A-CDCP1 axis and chemical carcinogen treatment have synergistic effect in promoting cells malignant transformation. For this purpose, we treated the METTL3or CDCP1-overexpressing and the control SV-HUC-1 cells with chemical carcinogen Cd for 6 weeks. Our data showed the Cd treatment itself promotes the proliferation (Fig. 7c) and migration (Fig. 7d) of uroepithelial SV-HUC-1 cells. Overexpression of METTL3 together with Cd treatment can further promote the proliferation of the SV-HUC-1 cells (Fig. 7c). In addition, forced expression of METTL3 or CDCP1 functionally interplay with Cd treatment to promote the migration (Fig. 7d) and invasion (Fig. 7e) of SV-HUC-1 uroepithelial cells. Moreover, depletion of CDCP1 partially reverse the effects of METTL3 on migration (Fig. 7f) and invasion (Fig. 7g) in SV-HUC-1 cells treated with Cd for 6 weeks, suggesting CDCP1 is a major contributor to the function of METTL3 in cell transformation. These data further confirmed essential function of the METTL3-m 6 A-CDCP1 axis in promoting the chemical carcinogen-induced malignant transformation.
To study the interplay between METTL3 and chemical carcinogen in tumorigenesis in vivo, the Cd-treated METTL3-overexpressing and control SV-HUC-1 were implanted into immunodeficient mice by subcutaneous injection. Our data revealed that forced expression of METTL3 further promoted the Cd-induced carcinogenesis in the mouse xenograft model (Fig. 7h-j); consistent with previous results, the CDCP1 expression is upregulated in the METTL3-overexpressing tumors (Fig. 7l). Importantly, the m 6 A modification in CDCP1 mRNA also increased in the METTL3-overexpressing tumors (Fig. 7k). These data revealed the synergistic effect of METTL3 overexpression and chemical carcinogen in promoting oncogenic transformation and in vivo carcinogenesis. Overall, our data support the oncogenic function of METTL3-m 6 A-CDCP1 axis in promoting the in vitro and in vivo bladder cancer progression.

Discussion
Chemical carcinogens induce a variety of genetic and epigenetic changes to promote oncogenic transformation in vitro and in vivo [27,32,33]. Aberrant DNA methylation is an important hallmark of cancer [34,35]; chemical carcinogen treatment of normal epithelial cells causes abnormal promoter DNA methylations that are commonly seen in cancers and results in activation of key oncogenic signaling in tumorigenic process [32,36]. Similar to DNA methylation, m 6 A modification of mRNA is catalyzed by methyltransferases to transfer methyl group from S-adenosylmethionine to target sites; moreover, aberrant m 6 A modification is frequently found in cancers and is associated with cancer progression. However, the role of chemical treatment in m 6 A mRNA modification during chemical carcinogenesis is still unknown.
Indeed, more than 20 years ago, it was found that the increase in m 6 A mRNA methyltransferase activity was concomitant with adenovirus-induced rat embryo cells transformation [37]. However, the exact role of increased methyltransferase activity in the regulation of specific gene expression in transformed cells is largely unclear. In this study, we first profiled the m 6 A epitranscriptomic modifications in the chemical carcinogen-induced cellular transformation models. Similar to those of previous studies [38,39], our results showed that the control and transformed cells transcriptomes contain 13,623~16,239 putative m 6 A sites within 7569~9039 coding gene transcripts (Table S3). Moreover, we not only found a subset of carcinogene-dependent or cell type-specific, dynamically altered m 6 A peaks but identified the common differential m 6 A peaks by comparing the methylation profiles of each set of control with the corresponding transformed cells (Fig.  1d, e, Table S4). These results suggested the complexity of m 6 A regulation mechanism of specific cell malignant transformation induced by different carcinogens. More importantly, our data uncovered the commonly dynamic m 6 A modification changes at the mRNAs of oncogenes such as CDCP1 during chemical carcinogenesis (Fig. 1).
CDCP1 is a transmembrame glycoprotein that is highly overexpressed in multiple cancer types and overexpression of CDCP1 is associated with poor prognosis in cancer patients [40][41][42]. CDCP1 interplays with the oncogenic Ras/ERK signaling to promote the cancer cell growth, migration, and invasion [41][42][43], and represents as a promising target for cancer therapeutics. Here we found that CDCP1 is posttranscriptionally regulated by the m 6 A pathway upon chemical carcinogenesis. Mechanistically, the m 6 A methyltransferase METTL3 and demethylase ALKBH5 mediate the dynamic m 6 A modification on CDCP1 mRNA. Furthermore, interestingly, METTL3 and YTHDF1, but not YTHDF2 and YTHDF3, bind to the 3′-UTR of CDCP1 mRNA and promote the CDCP1 translation. Our data support the novel function of the METTL3-YTHDF1 axis in promoting the m 6 A modification and translation of oncogenes such as CDCP1 upon chemical treatment, which could cooperate with other chemical-induced genetic and epigenetic changes to facilitate oncogenic transformation in chemical carcinogenesis (Fig. 7m).
The m 6 A mRNA methyltransferase METTL3 is upregulated in multiple types of cancers and can promote the m 6 A modification and expression of oncogenes including EGFR, TAZ, MYC, and SOX2 [18,20,24]. The increased m 6 A modification and translation of oncogene CDCP1 in the transformed uroepithelial cells indicates the potential role of METTL3/m 6 A/CDCP1 axis in bladder cancer oncogenesis. Analysis of METTL3 and CDCP1 expression in primary patient samples revealed that METTL3 and CDCP1 are strongly upregulated in the bladder cancer patient samples. Most importantly, the expression of METTL3 and CDCP1 is correlated with the progression status of the bladder cancers, further supporting the role of METTL3-mediated m 6 A pathway in bladder cancer oncogenesis.
Our in vitro and in vivo data revealed that inhibition of the METTL3-m 6 A-CDCP1 axis resulted in the decreased proliferation, migration, and invasion of bladder cancer cells and the chemical-transformed uroepithelial cells, suggesting that the METTL3-m 6 A-CDCP1 axis is essential for the growth and maintenance of bladder cancer cells and transformed cells. The fact that METTL3 and CDCP1 can both promote SV-HUC-1 migration and invasion, but only METTL3 can promote proliferation of SV-HUC-1 cells, suggests that METTL3 could have additional targets. Indeed, our m 6 A MeRIP-seq identified a subset of genes undergo m 6 A modification changes during chemical carcinogenesis and our data confirmed that CDCP1 is one of the key downstream target genes of METTL3 mainly related to migration and invasion. We further tested the function of METTL3-m 6 A-CDCP1 axis in bladder cancer initiation and found that forced expression of METTL3 or CDCP1 functionally interplay with Cd to promote the bladder cancer tumorigenesis, supporting the synergistic effect of METTL3-m 6 A-CDCP1 axis and chemical carcinogens in promoting oncogenic transformation.

Conclusions
Overall, our study uncovered that altered m 6 A modifications as a novel epitranscriptomic mechanism in chemical carcinogenesis and suggested that METTL3-m 6 A-CDCP1 axis could be a promising therapeutic target for the treatment of chemical-induced cancers.

Clinical samples
In this study, Formalin-fixed paraffin embedded (FFPE) tissues of 114 bladder cancer and 30 cystitis patients, who underwent radical cystectomy and bladder biopsies between February 2010 and October 2016, were obtained from the

Cell culture and cell transformation
Human uroepithelial cells (SV-HUC-1) and 3methylcholanthrene-transformed SV-HUC-1 cells (MC-SV-HUC T2) were obtained from the American Type Culture Collection (Manassas, VA). Human prostatic epithelial cell line RWPE-1 and human bladder cancer cell lines T24, UM-UC-3 was purchased from the Institute of Cell Biology, Chinese Academy of Sciences (Shanghai, China). Human bronchial epithelial cells (16HBE) and NiStransformed cell lines (NSTC2) were generated as described previously [29]. RWPE-1 cells were cultured in KSF-M (Fisher Scientific) containing 50 mg/ml bovine pituitary extract and 5 ng/ml epidermal growth factor. SV-HUC-1 cells were cultured in F12K, T24 cells were cultured in RPMI 1640, and UM-UC-3, 16HBE, NSTC2 cells were cultured in Minimum Eagle's medium (MEM). All medium were supplemented with 10% fetal bovine serum (FBS; Gibco, USA), in a humidified air atmosphere of 5% CO 2 at 37°C. All the cell lines were tested negative for mycoplasma contamination using a PCR-based universal mycoplasma detection kit. CdCl 2 -induced cells transformation assay were performed as described previously [44,45]. Briefly, SV-HUC-1 and RWPE-1 cells were maintained continuously in medium containing 10 µM CdCl 2 for 8 and 10 weeks, respectively, and then the cells were cultured in Cd-free medium for 2 weeks.

Tumorigenicity in NOD/SCID mice
To test for malignant transformation, 1 × 10 7 cells were inoculated subcutaneously in the dorsal thoracic midline of ten NOD/SCID mice (Weitong Lihua Experimental Animal Technology Co. Ltd). Tumor formation and growth were assessed every 3 days. Tumor volume = Length × Width 2 × 0.52. Tumor samples were paraffin-embedded, sectioned, stained with hematoxylin and eosin, and analyzed by light microscopy.

m 6 A methylated RNA immunoprecipitation sequencing
MeRIP-seq and data analysis were performed as described previously [46]. Briefly, total RNA was extracted from cells using Trizol (Invitrogen) following the manufacturer's instructions. mRNA was first purified from total RNA using Gen Elute mRNA Miniprep Kit (Sigma-Aldrich). Then 5 μg mRNA was fragmented using a RNA fragmentation kit (Ambion) and immunoprecipitated with the mixture of Protein A beads (Thermo Fisher) and anti-m 6 A antibody (Synaptic Systems); the immunoprecipitated RNA was washed and eluted by competition with m 6 A nucleotide solution (Sigma-Aldrich). The purified RNA fragments from m 6 A MeRIP were used for library construction and sequenced with Illumina Nextseq500. Reads mapping, m 6 A peak calling, motif search, and differentially methylated peaks were analyzed by exomePeak R/Bioconductor package as described [47].

RNA-binding protein immmunoprecipitation
The RNA samples was isolated from HUVEC primary cells and cDNA synthesized by Reverse Transcription PCR using TransScript® All-in-One First-Strand cDNA Synthesis SuperMix for quantitative PCR (qPCR) (Transgene). The YTHDF1, YTHDF2, YTHDF3, and METTL3 CDS regions were PCR amplified and cloned into pcDNA3-FLAG2AB using HindIII and EcoRI restriction sites. The 2AB-FLAG-YTHDF (3.5 μg) and PLEX-CDCP1 (4 μg) plasmids were co-transfected into 10 7 METTL3overexpressing 293T cells by Lipofectamine 3000. The medium was replaced with fresh Dulbecco's modified Eagle's medium (DMEM)/10% FBS 6-8 h after transfection. Forty-eight hours later, cell lysate was using the Magna RIP™RNA-Binding Protein Immunoprecipitation Kit (Millipore). The lysate was then immunoprecipitated with 50 μl protein A/G magnetic beads and 5 μg anti-FLAG antibody(Sigma) for 30 min at room temperature. Normal Rabbit IgG was used as a negative control. The magnetic bead-bound complexes were immobilized with magnet and washed with RIP Wash Buffer. The precipitated RNA samples were extracted with Phenol:chloroform purification, then analyzed by qPCR.

Dual-luciferase reporter assay
Luciferase reporter assay was performed with the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's descriptions. Cells were seeded into 24-well plates (about 3 × 10 5 cells/well) 1 day before the transfection. After 24 h, the cells were co-transfected with the psiCHECK TM -2 Vector (Promega), which is used for internal normalization and various constructs containing the seed sequence or mutant seed sequence of CDCP1 mRNA 3′-UTR. The relative luciferase activities were accessed 48 h post transfection by SYNERGY microplate reader (Bio-Tek). Each group was performed at least three times.

In vitro transcription and translation
The psiCHECK TM -2 vectors with WT or mutant m 6 A motifs of CDCP1 mRNA 3′-UTR were treated with Acc651 and Mlu1 to linearize the DNA templates. The in vitro transcription was performed with mMESSAGE mMA-CHINE Kit (Ambion) according to the manufacturer's instructions. The transcribed mRNA samples were purified with Phenol:chloroform extraction and isopropanol precipitation. The in vitro translation of the purified mRNA samples was performed using Flexi Rabbit Reticulocyte Lysate System (Promega), then the reaction was tested for the synthesis of functional luciferase using the standard luciferase assay.

Immunofluorescence and immunohistochemistry
The cells were fixed with 4% formaldehyde for 15 min and then blocked with 0.1% Triton X-100 in phosphate-buffered saline (PBS) for 5 min and with 1% bovine serum albumin (BSA) for 60 min at room temperature. Immunofluorescence staining was incubated with appropriate primary antibodies and then performed with appropriate Alexa Fluor 488 and Alexa Fluor 568 secondary antibodies (dilution 1:1000) for 60 min at room temperature, respectively. Nuclei were counterstained with 4′,6-diamidino-2phenylindole. Images were taken with a ZEISS Axio Imager Z2 Upright Microscope.
Paraffin sections are baked for 2 h at 55°C. Before immunostaining, paraffin sections were deparaffinized, rehydrated through an alcohol series followed by antigen retrieval with sodium citrate buffer. The sections were then treated in 3% hydrogen peroxide for 30 min and subsequently blocked with 5% normal goat serum. After blocking, the sections were incubated with appropriate primary antibodies at 4°C overnight. IHC staining was performed with horseradish peroxidase (HRP) conjugates using Diaminobenzidine (DAB) detection. Images were taken with a ZEISS Axio Imager.Z2 Microscope.

RNA isolation and quantitative real-time PCR
RNA was isolated using TRIzol (Life Technologies) following the manufacturer's protocol. cDNA was generated using the transScript All-in-One First-Strand cDNA Synthesis SuperMix for qPCR (Transgen). Quantitative real-time PCR using Fast SYBR Green PCR Master Mix (Applied Biosystems) was performed on a Step-One Fast Real-time PCR System (Applied Biosystems). For RNA stability assay, the cells were plated in a 6 cm dish and incubated with actinomycin D (Santa Cruz) at 5 mg/ml for indicated time points. Total RNA samples were then isolated for qPCR analysis. The quantitative PCR primers were listed in Table S1.

Sucrose gradient centrifugation and polysomal fractionation
Polysomal fractionation to categorize translationally active transcripts was performed using sucrose gradient centrifugation as previously described [48]. Cell lines for polysomal fractionation were first lysed using Polysome cell extraction buffer (50 mM MOPS, 15 mM MgCl 2 , 150 mM NaCl, 100 μg/ml CHX, 0.5% Triton X-100, 1 mg/ml Heparin, 200 U RNaseout, 2 mM phenylmethylsulfonyl fluoride (PMSF) 1 μM Benzamine). Following cell lysis, nuclei and cell debris were cleared by centrifugation at 13,000 × g for 10 min at 4°C. One milliliter of supernatants was laid on the top of 11 ml 10~50% sucrose gradient tube, then centrifuged at 36,000 r.p.m. for 2 h 30 min at 4°C with max break (Beckman coulter SW 41 Ti rotor) at 4°C. Then the RNA in polysome fraction was extracted and subjected to real-time PCR.
The sequences of sgRNAs were listed in Table S1. The lentiviral vectors were co-transfected with packaging vectors psPAX2 and VSVG (Addgene) into 293T cells for lentivirus production. To establish stable cell lines, the cells were transduced using the above lentiviruses with polybrene (8 mg/ ml, Sigma). Seventy-two hours after transduction, cells were selected with 1 μg/ml puromycin for 5-7 days.

Migration assay
Cell migration assay was performed with the IncuCyte™ 96-well Real-Time Cell Migration System (Essen Bioscience). Before initiating a 96-well assay, cells (1 × 10 5 ) were grown to confluence in a 96-well Essen Ima-geLock™ plate in a standard CO 2 incubator. The 96-pin WoundMaker simultaneously created precise and reproducible wounds in all wells of a 96-well ImageLock plate by gently removing the cells from the confluent monolayer using an array of 96 pins. After washing by cold PBS, the plate was placed inside the IncuCyte. The software was set to scan the experiment every hour for migration assays using "Scratch Wound" as the "Experiment Type". The data were analyzed by the relative wound density and the image collection was created using the representative phase-contrast images. Additional statistics and graphing were completed using GraphPad Prism following data export.

Transwell TM invasion assay
Cell invasion assay was performed in a 24-well Transwell™ (Costar). The upper chamber surface of the filter was coated with Matrigel (Corning) before the experiment. The cells were prepared (1 × 10 5 /200 μl) with serum-free DMEM and loaded into the upper chamber. DMEM medium containing 20% FBS was added to the bottom chamber as the chemoattractant. After 24 h incubation, wet cotton was used to remove the non-invaded cells from the wells. The cells were fixed with carbinol for 15 min at room temperature, stained with 0.1% crystal violet for 20 min, and quantified by counting the total number of cells in four independent areas under the ZEISS Axio Imager.Z2 Microscope.

Quantification and statistical analysis
Data are presented as the mean ± SEMs or SDs. Statistical analyses were performed in GraphPad Prism 6 (GraphPad Software, Inc.) using unpaired two-tailed Student's t-test to compare differences between two groups with significance of P < 0.05. One-way analysis of variance with multiple comparisons tests was used to compare three or more groups with significance of P < 0.05.

Data availability
MeRIP-seq data are deposited at the Gene Expression Omnibus database with the accession Number GSE112970.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Ethics approval and consent to participate This study was approved by the Institutional Ethics Committee for Clinical Research and Animal Trials of the First Affiliated Hospital of Sun Yat-sen University [(2016)067]. Informed consents were obtained from all patients before analysis. All the experimental methods involved in this study comply with the Helsinki Declaration.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.