Autophagy of the m6A mRNA demethylase FTO is impaired by low-level arsenic exposure to promote tumor

Here we show that FTO as an N6-methyladenosine (m6A) RNA demethylase is degraded by selective autophagy, which is impaired by low-level arsenic exposure to promote tumorigenesis. We found that in arsenic-associated human skin lesions, FTO is upregulated, while m6A RNA methylation is downregulated. In keratinocytes, chronic relevant low-level arsenic exposure upregulated FTO, downregulated m6A RNA methylation, and induced malignant transformation and tumorigenesis. FTO deletion inhibited arsenic-induced tumorigenesis. Moreover, in mice, epidermis-specific FTO deletion prevented skin tumorigenesis induced by arsenic and UVB irradiation. Targeting FTO genetically or pharmacologically inhibits the tumorigenicity of arsenic-transformed tumor cells. We identified NEDD4L as the m6A-modified gene target of FTO. Finally, arsenic stabilizes FTO protein through inhibiting p62-mediated selective autophagy. FTO upregulation can in turn inhibit autophagy, leading to a positive feedback loop to maintain FTO accumulation. Our study reveals FTO-mediated dysregulation of mRNA m6A methylation as an epitranscriptomic mechanism to promote arsenic tumorigenicity.

Analogous to epigenetic chemical modifications of DNA and histones, RNA modifications regulate RNA fate and gene expression. The most abundant RNA modification is N6-methyladenosine (m6A)1,2,3,4. m6A methylation regulates several aspects of RNA metabolism such as mRNA decay, nuclear processing, translation, transcription, and RNA-protein interactions1,2,3,4,5. m6A modification plays an important role in development, stem cell homeostasis, and diseases such as cancer3,6,7, as well stress response to UV radiation8 or heat shock9. At the molecular level, RNA m6A methylation is installed by writers (methyltransferases), removed by erasers (demethylases), and recognized by readers, a piece of well-coordinated machinery to regulate m6A’s dynamics1,2,3,4. Among these m6A effectors, FTO (fat mass and obesity-associated protein) was the first RNA m6A demethylase discovered10 and has been shown to play oncogenic roles in leukemia11, glioblastoma12, and melanoma13. However, the role of RNA methylation in arsenic-induced skin tumorigenesis remains unclear.

Exposure to inorganic arsenic in contaminated drinking water poses an environmental public health threat for hundreds of millions of people in the US and in the world14,15,16. Arsenic is a natural component of the earth’s crust and is widely distributed throughout the environment. In nature, sources of arsenic exposure include food, air, and water, with the main one being arsenic-contaminated drinking water17. Arsenic is a human carcinogen. Chronic exposure to arsenic can cause skin cancer and several internal cancers such as lung, bladder, and kidney cancer17. A major target organ of arsenic is the skin. Arsenic-induced skin lesions are an early manifestation of arsenic exposure and toxicity18 and are a risk factor for subsequent cancers19. Arsenic seems to act in concert with sunlight exposure, smoking, and occupational exposures in increasing the risk of skin lesions20. At the molecular level, arsenic has been shown to cause oxidative stress, DNA damage, chromosomal aberration, and epigenetic modifications, including DNA methylation and histone modification17,21,22. However, the mechanism by which arsenic causes tumorigenesis is still poorly understood.

In this study, we discovered that chronic low-level arsenic exposure upregulates FTO and downregulates m6A RNA methylation in keratinocytes, and induces malignant transformation and tumorigenesis through FTO and m6A mRNA methylation. Our study reveals that arsenic inhibits autophagy and autophagic degradation of FTO through suppressing p62 transcription, leading to FTO stabilization. Furthermore, FTO forms a positive feedback loop with autophagy inhibition to maintain FTO accumulation in arsenic tumorigenesis. Taken together, these results demonstrate that m6A RNA methylation acts as an epitranscriptomic mechanism for arsenic damage response and tumorigenesis.

FTO is upregulated, while m6A is downregulated, by arsenic in human keratinocytes and in epidermal keratinocytes in human skin

To determine whether m6A RNA methylation and its regulators play a role in arsenic-induced tumorigenicity, we first treated HaCaT cells, a human keratinocyte cell line, continuously with a relevant low level (100 nM) of inorganic arsenite for 28 weeks to generate HaCaT cells with chronic arsenic damage (As cells)23,24 (Supplementary Fig. 1a). Arsenic substantially upregulated FTO and downregulated m6A, while it had moderate or no effect on other m6A regulators, including the m6A methyltransferase proteins METTL14 and METTL3 and the m6A demethylase ALKBH5 (Supplementary Fig. 1b). We found that treatment with a low level of arsenic for up to 72 h upregulated FTO protein and downregulated m6A in HaCaT cells (Fig. 1a and Supplementary Fig. 1c). The effect of arsenic on FTO seemed to be dose-dependent, since only low-level arsenic upregulated FTO, whereas high-level arsenic downregulated FTO (Supplementary Fig. 1d). Similarly, FTO upregulation was also detected in NHEK cells (normal human epidermal keratinocytes) (Fig. 1b).

As-treated cells developed tumors in all four mice, while control cells did not (Fig. 1c and Supplementary Fig. 1e). We isolated the tumorigenic cells (As-T, Arsenic-Tumorigenic cells) from one tumor in mice (#1) as well as other tumors formed by the As-treated cells (As cells) in mice (Fig. 1c and Supplementary Fig. 1a). In addition to As cells, FTO was also upregulated in the four As-T cells, while the m6A level was downregulated in As and four As-T cells as compared with control cells (Fig. 1d). In addition, as compared with control cells, As and As-T cells showed increased proliferation (Fig. 1e) and survival (Fig. 1f, and Supplementary Fig. 1f). However, in contrast to arsenic, cadmium, another carcinogenic heavy metal present in the air, water, soil, sediment25, had no effect on FTO in HaCaT cells (Supplementary Fig. 1g), suggesting FTO as a target unique for arsenic exposure. Future investigation is needed to determine whether cadmium and other heavy metals affect FTO or other m6A regulators in keratinocytes and other cell types.

To determine the role of FTO and m6A in arsenic-induced tumorigenesis, we next assessed whether the levels of m6A and FTO are altered in arsenic-induced premalignant human skin lesions, arsenical keratoses, as compared with normal human skin. In humans, it is known that chronic arsenic exposure causes arsenical keratoses, which can persist indefinitely and may develop into invasive skin cancers such as squamous cell carcinoma26. Normal human skin samples were obtained from the archives in the tissue bank of the Section of Dermatology (Department of Medicine, University of Chicago). The arsenical keratoses were obtained from the clinical follow-up for arsenic-exposed cohorts that evaluate the health effects of exposure, including skin cancer, in Bangladesh. To remove the signal from DNA, we treated all samples with DNase I8 (Fig. 1g). As compared with normal human skin, we found that FTO is upregulated and the m6A level is downregulated in arsenical keratoses (Fig. 1h–l). In addition, we found that the m6A level is negatively correlated with the FTO level (Fig. 1m). These findings suggest that FTO acts as a protumorigenic factor in arsenic-induced skin lesions.

FTO is required for arsenic-induced tumorigenicity

To determine the functional significance of FTO upregulation in arsenic-induced tumorigenicity, we assessed the consequence of FTO knockout (KO) in As-T cells using CRISPR (Fig. 2a). FTO deletion decreased cell growth (Fig. 2b), anchorage-independent growth (Fig. 2c), clonogenic growth (Supplementary Fig. 2a), 3D cell growth (Supplementary Fig. 2b), and survival when cells were placed in suspension (Supplementary Fig. 2c). Notably, FTO deletion partially reversed the effect of arsenic on cell survival (Supplementary Fig. 2c). It is possible that FTO and other pathways affected by arsenic also promote cell survival. Nevertheless, FTO deletion drastically inhibited tumor growth of As-T cells in mice (Fig. 2d and Supplementary Fig. 2d). Furthermore, knockdown of FTO using siRNA also decreased cell proliferation in As cells (Supplementary Fig. 2e). While overexpression of FTO slightly increased cell proliferation in HaCaT cells (Supplementary Fig. 2f), it did not induce tumor formation in NSG mice (Supplementary Fig. 2g), indicating that FTO overexpression alone is not sufficient to induce tumorigenesis. It is likely that FTO upregulation is critical for arsenic tumorigenesis, while it seems that both FTO upregulation and other molecular alterations induced by arsenic are required for malignant transformation and tumorigenesis.

Next, we used the arsenic-UVB skin co-carcinogenesis mouse model to determine the role of FTO in arsenic-induced skin tumorigenesis. We first created mice with epidermis-specific FTO deletion (K14Cre;Fto flox/flox). Then we treated the mice with arsenic, UVB irradiation, or both, as reported previously27. We found that arsenic accelerates skin tumorigenesis, consistent with the previous report27. Epidermis-specific FTO deletion inhibited skin tumorigenesis induced by UVB irradiation alone or the combination of UVB irradiation and arsenic, with a greater effect on the latter (Fig. 2e, f). In addition, skin-specific FTO deletion reduced epidermal hyperplasia induced by UVB irradiation alone or the combination of UVB irradiation and arsenic (Fig. 2g, h). Histological analysis showed that the tumors developed in these groups are either papillomas or squamous cell carcinomas (SCC) (representative images shown in Supplementary Fig. 2h).

To determine the therapeutic potential by targeting FTO in As-T cells, we first tested the effect of inducible FTO knockdown using a doxycycline-inducible shRNA knockdown system. Inducible FTO knockdown (Fig. 2i) decreased cell growth (Supplementary Fig. 2i), tumor growth (Fig. 2j), and tumor weight (Supplementary Fig. 2j). To determine the translational potential of targeting FTO, we tested the effect of the FTO inhibitors CS1 and CS2 identified recently28. CS1 and CS2 increased the m6A level (Supplementary Fig. 2k), while it decreased cell growth (Supplementary Fig. 2l), tumor growth (Fig. 2k), and tumor weight (Fig. 2l), supporting the anti-tumor effects of these FTO inhibitors in As-T cells. Overall, these findings demonstrate that FTO is required for arsenic-induced malignant transformation and tumorigenesis and that targeting FTO genetically or pharmacologically inhibits the tumorigenicity of arsenic-treated cells.

m6A seq and RNA seq analyses identify NEDD4L as the critical target for arsenic-induced FTO upregulation

To determine the potential m6A targets of arsenic exposure, we used m6A-seq coupled with RNA-seq to determine the potential m6A demethylation gene targets across the whole transcriptome for arsenic. As compared with control cells, As-T cells showed decreased m6A enrichment in 5’UTR and 3’UTR (Fig. 3a) and altered total peak distribution (Supplementary Fig. 3a). As-T cells also showed a decreased number of m6A genes compared with control cells (Supplementary Fig. 3b). Sequence analysis of m6A peaks showed the previously identified m6A target sites (GGACU)29,30 (Fig. 3b).

Next, we used the following criteria and analyses and identified NEDD4L (neural precursor cell-expressed developmentally downregulated 4 like) as a potential m6A target of arsenic, m6A modification, and FTO: (1) decreased m6A enrichment, (2) altered mRNA level, (3) potential role in cancer, and (4) altered expression in both human and mouse SCC. We compared the genes with altered expression and m6A enrichment in our m6A seq and RNA seq datasets from control and As-T cells, with genes with altered expression from human and mouse skin cancers (squamous cell carcinoma, SCC) (GSE84292 and GSE84293) generated from the previous studies31. There were 40 genes altered among the four groups, including 10 downregulated genes (GSN, LPIN1, ME1, NCOA7, NEDD4L, PDLIM4, TRPV4, CLDN1, CRIP2, GAS6) and 3 upregulated genes (GINS4, RAD54L, SLC16A1) that shared changes in As-T, mouse and human cuSCCs (Fig. 3c, and Supplementary Fig. 3c). Among all these genes, we elected to initially focus on NEDD4L, as it is found to be downregulated in several cancers32,33,34,35, suggesting a tumor-suppressive role for NEDD4L.

As and four As-T cells showed decreased mRNA and protein levels of NEDD4L as compared with control cells (Fig. 3d, e). m6A-seq analysis showed that As-T cells exhibited decreased m6A levels in the exon 1 (NC_000018.10:58221610-58221760) region of the NEDD4L isoform (id = NM_001144966.3) (Fig. 3f). Consistently, m6A IP qPCR analysis showed that in As-T cells the m6A level for the NEDD4L full-length transcript was reduced as compared with control cells (Fig. 3g), consistent with elevated FTO in these cells (Fig. 1d). To determine the m6A site at the NEDD4L transcript regulated by FTO, we performed the m6A IP qPCR analysis of m6A enrichment in the differentially methylated region of the NEDD4L transcript in control and As-T cells. As-T showed decreased m6A enrichment in the exon 1 (NC_000018.10:58221610-58221760) region of NEDD4L as compared with control cells (Fig. 3h). Furthermore, NEDD4LmRNA stability was decreased in As-T compared with control cells (Fig. 3i), suggesting that arsenic regulates NEDD4L RNA metabolism and decay. FTO deletion increased the m6A level and abundance of the NEDD4L transcript (Fig. 3j). m6A IP qPCR analysis of m6A enrichment showed that FTO deletion increased the m6A level in the exon 1 region of NEDD4L in As-T cells (Fig. 3k). Future investigation is needed to determine the importance of other potential m6A sites in the NEDD4L genes in arsenic tumorigenesis. Moreover, we found that FTOdeletion increased the NEDD4L mRNA and protein levels and mRNA stability (Fig. 3l, m, and Supplementary Fig. 3d), suggesting that increased m6A enrichment on the NEDD4Ltranscript elevates its stability instead of leading to transcript decay36,37.

To determine whether FTO regulates NEDD4L mRNA m6A modification and expression through its demethylase activity, we tested the effect of the FTO demethylase-inactive mutants. In cells with FTO deletion, wild-type (WT) FTO decreased NEDD4L mRNA and protein levels, while FTO demethylase-inactive mutants had no effect (Fig. 3n, o). We also showed that the FTO inhibitors CS1 and CS2 increased the protein and mRNA levels of NEDD4L (Supplementary Fig. 3e, f), while it had no effect on the protein or mRNA levels of FTO, consistent with the previous studies28. These data strongly indicate that FTO upregulation by arsenic decreases the NEDD4L mRNA stability through its RNA demethylase activity.

Identification of NEDD4L as the functional target of FTO

Next, we investigated whether NEDD4L is responsible for FTO’s function in arsenic tumorigenicity. First, we assessed whether NEDD4L is negatively associated with FTO in arsenical keratoses, as compared with normal human skin. We found that FTO is upregulated, while NEDD4L is downregulated in arsenical keratoses as compared with normal skin (Fig. 4a, b). NEDD4L levels were negatively correlated with FTO levels (Fig. 4c).

Second, we analyzed the alteration of NEDD4L in arsenic-UVB-treated skin and tumors in mice and the effect of FTO deletion. We found FTO is upregulated by UVB alone, arsenic alone, or the combination of both (Fig. 4d and Supplementary Fig. 4a). Moreover, while NEDD4L was upregulated by UVB irradiation, it was downregulated by arsenic (Fig. 4e). Other mechanisms may mediate UVB-induced NEDD4L upregulation, which remains to be investigated in detail. Nevertheless, NEDD4L levels were negatively correlated with FTO levels (Fig. 4f). Furthermore, FTO deletion increased NEDD4L levels in skin and tumors from mice treated with both UVB irradiation and arsenic (Fig. 4g, h, and Supplementary Fig. 4b).

Third, we knocked down NEDD4L and observed that it partially reversed the effect of FTOdeletion on cell proliferation, anchorage-independent growth, and cell survival when placed in suspension (Fig. 4i-l). Consistently, NEDD4L knockdown also partially reversed the effect of FTO deletion on tumor growth in vivo (Fig. 4m, n). These data indicate that NEDD4L is an important target of FTO, and other FTO targets may also play an important role in FTO’s function.

Next, we investigated the potential target of NEDD4L. NEDD4L is shown to act as a ubiquitin ligase controlling several pathways, including the WNT signaling by targeting Dvl238. As WNT signaling is one critical pathway in skin cancer39, we elected to test the role of the Wnt/β-catenin pathway in NEDD4L’s function. Using TOP Flash luciferase assay, we found that WNT signaling is increased in As-T cells as compared with control cells (Supplementary Fig. 4d). We observed that Dvl2, as well as β-catenin, a downstream target of WNT/Dvl240, is upregulated in As-T cells as compared with control cells (Supplementary Fig. 4e). FTOdeletion decreased the levels of Dvl2 and β-catenin and WNT activity, which was prevented by NEDD4L knockdown (Supplementary Fig. 4e-f). Future investigation is needed to elucidate the role of other pathways, including the TGF-β41 and PI3K/AKT42 pathways, in the function of NEDD4L in skin cancer. These results demonstrate that NEDD4L is a critical downstream target of FTO responsible for its function in arsenic-induced malignant transformation.

FTO regulates NEDD4L mRNA stability through m6A and IGF2BPs

To determine whether FTO regulates NEDD4L through m6A, we assessed the effect of manipulating the m6A RNA methyltransferase factors METTL14 and METTL343. METTL14knockdown at least partially reversed the effect of FTO deletion on NEDD4L mRNA stability, mRNA level, and protein level (Fig. 5a, b, and Supplementary Fig. 5a). Furthermore, METTL14knockdown partially reversed the effect of FTO deletion on tumor growth (Fig. 5c and Supplementary Fig. 5b). Similarly, METTL3 knockdown also partially reversed the effect of FTO deletion on cell growth (Fig. 5d and Supplementary Fig. 5c), and anchorage-independent growth (Fig. 5e). Overexpression of wild-type METTL3, but not the catalytically dead mutant METTL3 (D395A), increased the protein and mRNA levels of NEDD4L in As-T cells with FTO deletion and METTL3 knockdown (Fig. 5f, g).

In addition to m6A10, FTO has been shown to demethylate the m6Am RNA methylation44. To determine whether m6Am also has a role in FTO’s function, we assessed the effect of knocking down the m6Am methyltransferase PCIF145,46,47,48. We found that PCIF1knockdown has no effect on the regulation of NEDD4L by FTO (Supplementary Fig. 5d–f), consistent with a previous report49. Next, we investigated potential m6A readers responsible for the regulation of NEDD4L by FTO. Since FTO deletion increased the mRNA stability of NEDD4L, we reckoned that plausible m6A readers could be IGF2BP1-3, a group of m6A readers that stabilize m6A-modified transcripts37. Knockdown of IGF2BP1, IGF2BP2, or IGF2BP3 decreased the NEDD4L mRNA level (Supplementary Fig. 5g–j). Knockdown of all three IGF2BPs (IGF2BP1-3) diminished the effect of FTO deletion on NEDD4L protein and mRNA levels, and mRNA stability (Fig. 5h–j).CLIP-qPCR analysis showed that IGF2BPs bind with the NEDD4L transcript, which was decreased in As-T cells as compared with control cells (Fig. 5k). These findings demonstrate that FTO regulates NEDD4L RNA stability through m6A mRNA methylation and IGF2BPs.

As increases the FTO stability through impairing autophagy

To determine the mechanism by which arsenic upregulates FTO, we first assessed whether arsenic affects FTO mRNA levels. As compared with control cells, FTO mRNA levels remained unchanged in As and As-T cells (Fig. 6a), indicating that arsenic did not alter FTO mRNA abundance. Next, we assessed the effect of arsenic on FTO protein stability by treating the cells with cycloheximide (CHX) over a time course. As compared with control cells, FTO protein stability was increased in As-T cells (Fig. 6b, c), indicating that arsenic increased FTO stability. Next, we assessed the role of protein degradation through either the proteasome or the autophagic-lysosomal pathways, two critical protein degradation pathways50,51,52. When the proteasome was inhibited by MG132, control cells showed FTO protein accumulation, similar to As-T cells (Supplementary Fig. 6a, b). In contrast, when the lysosome was inhibited by the lysosome inhibitor bafilomycin A1 (BfnA1), control cells showed significantly more FTO protein accumulation than As-T cells did (Fig. 6d, e). In addition, As-T cells showed reduced autophagy, as indicated by decreased accumulation of LC3-II, the surrogate marker for autophagosome accumulation, when lysosome was inhibited (Fig. 6d). To further determine the role of autophagy in FTO stability, we assessed the effect of genetic inhibition of autophagy. Deletion of the essential autophagy genes ATG5 or ATG7 increased FTO protein stability as indicated by the increased half-life (Fig. 6f, g, and Supplementary Fig. 6c). These findings suggest that arsenic mimics the effect of autophagy inhibition in stabilizing FTO.

Next, we used the PLA assay to assess the plausible interaction between FTO and autophagosomes. We first validated the specificity of the PLA assay for FTO and LC3 (Supplementary Fig. 6d-f). Treatment with BfnA1 increased the interaction between FTO and LC3 (Fig. 6h, i). This was further confirmed by the co-localization of FTO and LC3 by immunofluorescence staining (Fig. 6j). It is noteworthy that FTO was localized mainly in the nucleus, consistent with the previous studies10; lysosome inhibition seemed to increase the FTO level in both the nucleus and the cytosol (Fig. 6j). Gene set enrichment analysis (GSEA) detected downregulation of both ‘Lysosome’ and ‘Phagosome’ pathway-related genes in mouse cuSCCs as compared with non-tumor skin (Fig. 6k, l). The downregulated genes were shown in the heatmap (Supplementary Fig. 6g, h). Moreover, we also found that ‘Autophagy-animal’, ‘Lysosome’, and ‘Phagosome’ pathway-related genes are downregulated in As-T cells compared to control cells (Supplementary Fig. 10a, b). Electron microscopy analysis showed that the number of autophagic vacuoles is decreased in As-T cells as compared with control cells (Fig. 6m, n), further supporting the suppression of autophagy by arsenic.

To determine the mechanism by which arsenic inhibits autophagy, we first assessed the role of Metal-activated transcription factor 1 (MTF-1), a multipotent regulator of transcription often involved in the adaptation to stress, including arsenic treatment53,54. While low-dose acute or chronic arsenic treatment did not induce MTF1, acute high-dose arsenic treatment did increase the MTF1 level (Supplementary Fig. 6i-l). However, MTF1 knockdown had no effect on the FTO level (Supplementary Fig. 6m), suggesting that MTF1 is not a major player in chronic low-level arsenic damage. Future studies are required to assess the role of MTF1 in arsenic tumorigenicity in detail. Taken together, these findings demonstrate that arsenic stabilizes FTO protein by inhibiting autophagic degradation.

p62 regulates FTO stability and is suppressed by arsenic

To determine the mechanism by which arsenic dysregulates autophagic FTO degradation, we first mined our RNA-seq data for the effect of arsenic on autophagy-related genes. We found that OPTN, LAPM1, p62, and ATG5 are decreased in As-T cells as compared with control cells (Fig. 7a). qPCR analysis showed that these genes are downregulated in both As and As-T cells as compared to control cells (Fig. 7b). Consistently, the p62 protein level was downregulated by arsenic (Fig. 7c).

Next, we assessed whether p62 regulates FTO protein stability. Indeed, overexpression of p62 reduced the FTO protein level, whereas p62 knockdown increased it (Fig. 7d and Supplementary Fig. 7a). Moreover, p62 inhibition increased FTO protein stability in both HaCaT and MEF cells (Supplementary Fig. 7b–e), whereas overexpression of p62 deceased it in As-T cells that expressed low levels of p62 (Fig. 7e, f). Correspondingly, overexpression of p62 increased NEDD4L mRNA levels in WT cells but not in cells with FTO deletion (Fig. 7g and Supplementary Fig. 7f), indicating that p62 regulates NEDD4L expression through FTO.

Next, we assessed whether FTO interacts with p62. First, we observed that the co-localization of FTO and p62 is increased by deletion of ATG7 or by the lysosome inhibitor BfnA1 in MEF cells (Supplementary Fig. 7g). PLA assays supported the co-localization of p62 and FTO in HaCaT cells (Supplementary Fig. 7h, i). Co-immunoprecipitation analysis showed that FTO binds with p62 (Fig. 7h). As an autophagy receptor, p62 can interact with polyubiquitin chains or its ubiquitinated protein substrates through its UBA domain55 and with LC3/GABARAP proteins noncovalently via a short linear sequence known as LIR56 to deliver its substrates to autophagosomes and then lysosomes for degradation. Indeed, we found that FTO is ubiquitinated in As-T cells (Fig. 7i). Moreover, deletion of the UBA domain of p62 abolished the FTO-p62 interaction (Fig. 7j, k), indicating that FTO-p62 interaction requires p62’s UBA domain. Furthermore, overexpression of wild-type (WT) p62, but not the LIR-binding deficient mutant (W338A)57,58, reduced FTO protein levels (Fig. 7l). Immune electron microscopy analysis showed that FTO colocalizes with the p62 cluster in the cytoplasm (Fig. 7m). These p62-positive structures are shown to be liquid droplets formed by liquid-liquid phase separation and degraded by autophagy, rather than monomeric or oligomerized p6259,60,61. It is possible that FTO is translocated into the p62 bodies depending on the context and degraded by autophagy. Lastly, as compared with normal human skin, human arsenical keratoses showed decreased p62 levels (Fig. 7n and Supplementary Fig. 7j). In addition, we found that p62 is negatively correlated with FTO (Fig. 7o).

Next, we assessed the role of p62 in FTO localization in the autophagosomes. Overexpression of p62 increased autophagy in p62-low As-T cells (Supplementary Fig. 8a). In addition, it also increased the co-localization of FTO with LC3 (Supplementary Fig. 8b, c) in As-T cells, whereas p62 deletion decreased it (Supplementary Fig. 8d-h) in MEF cells. Notably, p62 deletion markedly decreased but did not completely abolish, the FTO-LC3 PLA signal, suggesting that while p62 is a crucial autophagy receptor for FTO colocalization with LC3, other autophagy factors may also play a role. Immunofluorescence analysis further confirmed that p62 deletion reduced the co-localization of FTO with LC3 (Supplementary Fig. 8i). Our data indicate that p62 is necessary and sufficient for FTO localization in the autophagosomes.

Furthermore, we showed that FTO also regulates autophagy in response to arsenic as a positive feedback signaling system. We found that FTO deletion in As-T cells increases autophagy, as shown by p62 accumulation and formation of LC3-II upon BfnA1 treatment (Supplementary Fig. 9a). FTO deletion decreased p62 mRNA levels, consistent with decreased p62 protein levels in FTO-depleted cells under homeostatic conditions (Supplementary Fig. 9a, b). In addition, we found that FTO deletion increased the phosphorylation of AMPK, decreased phosphorylation of the mTOR substrate p70S6K, and increased protein levels of ATG5 and ATG7, as well as increasing the LC3-II level, under serum starvation conditions that can induce autophagy62 (Supplementary Fig. 9c).

Next, we assessed the role of TFEB, a transcription factor that controls transcription of CLEAR and autophagy genes and is regulated by phosphorylation63,64,65,66. We found that, as compared with control cells, As-T cells showed reduced levels of TFEB protein and phosphorylation levels (S211), but not the TFEB mRNA level (Supplementary Fig. 9d, e). It is possible that chronic arsenic treatment decreases either translation or protein stability of TFEB and that the decreased TFEB phosphorylation is due to the decreased TFEB abundance, which requires further investigation. FTO deletion decreased phosphorylation of TFEB at S122 and the expression of ATG5 and ATG7 but not p62 mRNA (Supplementary Fig. 9f-h), possibly due to the reduced mTOR signaling (Supplementary Fig. 9c). These findings suggest that TFEB has a complex role in arsenic damage response and FTO function, which requires further investigation. Our results suggest a positive feedback loop between FTO and autophagy dysregulation, leading to self-sustained maintenance of FTO upregulation in arsenic-induced malignant transformation and tumorigenesis. Taken together, these findings demonstrate that p62 binds with FTO to mediates FTO degradation, and that arsenic inhibits FTO degradation by suppressing p62 expression.

Arsenic downregulates p62 expression by inhibiting the TNF/NF-κB pathway

To determine how arsenic downregulates p62 expression, we performed pathway analysis to determine which shared signaling pathways were downregulated in As and As-T cells as compared with control cells using Metascape67. Among many pathways downregulated by arsenic, the cytokine production and cytokine signaling pathways, in particular, were downregulated in As and As-T cells as compared with control cells (Fig. 8a). Moreover, KEGG network analysis showed that the ‘Autophagy-animal’,’ Phagosome’, ‘Lysosome’, and ‘NF-κB and TNF signaling pathway’ was downregulated in As-T cells as compared with control cells (Supplementary Fig. 10a, b). GSEA analysis showed that TNF-signaling genes were downregulated in As-T cells as compared with control cells, and in mouse SCC as compared with non-tumor skin (Fig. 8b and Supplementary Fig. 10c-g). qPCR analysis showed that several TNF-related genes (NFKB1, TNF, TNFRSF1A, TNFAIP3) and autophagy-related genes (ATG5, ATG7, and p62) were decreased in arsenic-treated cells as compared with control cells (Fig. 8c).

It is known that p62 expression can be induced by NF-κB activation68. Previous studies have shown that in myometrial cells the known NF-κB-inducing cytokine TNFα induces p62expression69. Therefore, we hypothesized that arsenic downregulates p62 expression through downregulating signaling of cytokines such as TNFα. Indeed, NF-κB activity was decreased in arsenic-treated cells (As and As-T cells) as compared with control cells (Fig. 8d), while it was increased after TNFα treatment (Fig. 8e). TNFα treatment increased the protein and mRNA levels of p62 and NEDD4L, and the protein levels of ATG5 and ATG7, while it decreased FTO protein level (Fig. 8f, g). Knocking down p62 or RELA, also known as p65 as part of the NF-κB heterodimer, prevented the effect of TNFα (Fig. 8h, i). A PLA assay showed that TNFα increased the colocalization of FTO with either p62 or LC3, which was prevented by knocking down either p62 or RELA (Fig. 8j, k, and Supplementary Fig. 10h, i).

Previous studies showed that Nrf2 and TFEB also regulate p62 expression70. However, we found that in HaCaT cells knockdown of either Nrf2 or TFEB had no effect on FTO levels, although it had a small to moderate effect on the p62 levels (Supplementary Fig. 10j–l). Future studies are needed to assess the role of Nrf2 and TEFB in arsenic-regulation of p62 in greater detail. Taken together, our findings demonstrate that arsenic suppresses p62expression by downregulating the NF-κB pathway to upregulate FTO.

RNA m6A modifications are shown to play important roles in cancer development and therapeutic response7. However, the role of m6A in tumorigenicity caused by environmental carcinogens such as arsenic remains unknown. Here we discovered that chronic low-level arsenic exposure upregulates FTO stability and downregulates m6A enrichment in keratinocytes, leading to malignant transformation and tumorigenesis. At the molecular level, arsenic inhibits FTO protein degradation via impairing p62-dependent selective autophagy. FTO interacts with p62 in the autophagosomes, and p62 is required for autophagic degradation of FTO. Furthermore, arsenic inhibits p62 expression by suppressing the TNFα/NF-κB signaling pathway. We identified NEDD4L as a critical downstream target of FTO in arsenic tumorigenicity. NEDD4L RNA stability is regulated by m6A modification, through the m6A readers IGF2BPs and FTO. Furthermore, FTO and autophagy dysfunction form a positive feedback loop to maintain the upregulation of FTO. Taken together, these results establish RNA modification and FTO as an epitranscriptomic mechanism in arsenic-induced tumorigenesis (Supplementary Fig. 11).

We found that FTO is upregulated and m6A enrichment is downregulated by chronic exposure to relevant low levels of arsenic. The downregulation of FTO induced by arsenic seems to depend on the dose of arsenic since only low-level arsenic upregulates FTO (Supplementary Fig. 1d). However, high-level arsenic downregulates FTO, similar to previous reports71,72. Such FTO upregulation is required for arsenic-induced tumorigenesis, as FTO deletion diminished arsenic-induced tumorigenesis in both a xenograft tumor model and an arsenic-UVB co-carcinogenesis model. FTO’s pro-tumorigenic role in arsenic damage response is mediated through m6A RNA methylation, as knockdown of METTL14/METTL3reversed the effect of FTO deletion in vivo and in vitro. Furthermore, the FTO level is higher, while the m6A level is lower in human arsenical keratosis, a premalignant lesion induced by arsenic than in normal human skin. These findings provide an epitranscriptomic mechanism by which arsenic upregulates the RNA demethylase FTO to promote tumorigenesis.

Emerging evidence suggests that FTO and m6A RNA methylation regulate their functions in cancer and other diseases through different downstream gene targets depending on the context and cell lineage6,7. For example, in leukemia, Li and colleagues have shown that FTO promotes tumorigenesis by suppressing the expression of ASB2 and RARA genes11. In melanoma, we have recently shown that FTO promotes the mRNA stability of the melanoma-promoting genes PDCD1, CXCR4, and SOX1013. Using m6A IP seq followed by qPCR validation, we found that FTO reduces the m6A levels of the NEDD4L transcript, leading to its increased decay.

NEDD4L belongs to the NEDD4 family of the E3 HECT domain ubiquitin ligases and has been suggested to regulate several signaling pathways38,41,42, thus acting as a tumor suppressor in several cancers32,33,34,35. We found that knockdown of NEDD4L reversed the effect of FTOknockout on arsenic-induced tumorigenesis, establishing that NEDD4L is a critical mRNA target for FTO. However, in wild-type cells, knockdown of NEDD4L had no effect, suggesting that NEDD4L alone is dispensable for arsenic-induced tumorigenicity when FTO is intact to inhibit NEDD4L. In contrast to melanoma in which m6A decreases the RNA stability of melanoma-promoting gene transcripts13, in arsenic-induced tumor cells, we found that m6A increases the RNA stability of NEDD4L through the readers IGF2BP1-3. Previously IGF2BP1-3 proteins have been shown to be a distinct family of m6A readers that bind to and promote the stability and translation of m6A-modified transcripts37. We found that knockdown of IGF2BPs increases the NEDD4L decay in FTO knockout cells and reverses the effect of FTO. Future investigations are needed to determine whether IGF2BPs also play active roles in arsenic damage response and tumorigenicity and whether the FTO/IGF2BPs axis also regulates NEDD4L translation.

We found that FTO regulates the Dvl2 level through inhibiting NEDD4L, resulting in increased Wnt signaling, suggesting Wnt signaling as a downstream target for the FTO/NEDD4L axis. Taken together, our findings clearly demonstrate that FTO promotes arsenic-induced tumorigenesis through m6A RNA methylation and NEDD4L.

Intriguingly we found that chronic arsenic exposure stabilizes FTO protein by compromising its autophagic degradation. From RNA-seq analysis, arsenic suppressed the expression of several autophagy genes, including the autophagy receptor p62. Arsenic inhibited autophagy by suppressing p62 expression, as adding p62 reversed the effect of arsenic. In addition, we found that FTO degradation by autophagy is mediated by p62. FTO binds with p62’s UBA domain and colocalizes in the LC3-positive autophagosomes in a p62-dependent manner. p62 knockdown or knockout increases FTO protein stability, while p62overexpression decreases it. It appears that arsenic downregulates cytokine signaling pathways, including the TNFα pathway. p62 expression is shown to be regulated by several signaling pathways and transcription factors such as NF-κB68, a major downstream pathway for cytokine signaling. Treatment with TNFα increases NF-κB activity, leading to increased p62 expression and decreased FTO levels. In addition, the roles of other autophagy pathways in FTO degradation will require additional investigation.

Our findings showed that chronic low-level arsenic exposure inhibits p62 expression, which at least partially leads to compromised autophagy and autophagy-mediated FTO degradation, and thus facilitates arsenic tumorigenesis. Other autophagy genes downregulated by arsenic may also play important roles in autophagy inhibition by arsenic, which needs further investigation. Our findings seem to be specific for arsenic, as previous studies have shown that multiple types of tumors exhibit elevated basal autophagy73,74, as well as p62 accumulation70, to promote tumorigenesis and tumor progression70,75,76. In addition, it seems that the effect of toxic metals on autophagy is also dependent on context, including the specific metal, cell type, concentrations of metals, and treatment duration77. We found that in HaCaT cells, cadmium has no effect on FTO levels up to 72 h (Supplementary Fig. 1f). These discrepancies could be due to the specific mechanism of action for arsenic-induced tumorigenesis and suggest a context-dependent role for FTO, as well as p62 and autophagy. The regulatory and functional roles of carcinogenic metals in autophagy warrant further investigation, in particular in the context of clinically relevant low-level chronic exposure.

Moreover, we found that FTO inhibits autophagy. FTO deletion increased the expression of autophagy genes including ATG5 and ATG7, increased phosphorylation of AMPK, and decreased phosphorylation of the mTOR target p70S6K and TFEB, pathways that regulate autophagy63,78,79, and enhanced autophagy. This is consistent with previous reports in which FTO increases activation of the mTORC1 pathway and thus inhibits autophagy80. However, our findings are opposite to those reported in preadipocytes, in which FTOknockdown decreases the expression of ATG5 and ATG7 and reduces autophagy81. These discrepancies could be due to the cell-type specific or context-dependent role of FTO in autophagy, which warrants future investigation. Nevertheless, our findings demonstrate that FTO and autophagy inhibition form a positive feedback loop, leading to self-sustained FTO upregulation and malignant traits in arsenic-induced tumorigenic cells.

In summary, we demonstrate that arsenic promotes FTO protein stability by impairing selective autophagy and thus induces malignant transformation and tumorigenesis. FTO decreases m6A modification and stability of its functional downstream target NEDD4LmRNA. Our results provide an epitranscriptomic molecular axis as the post-transcriptional mechanism in arsenic-induced tumorigenesis, and thus may help the development of new mechanism-based prevention and therapy by targeting FTO and m6A RNA methylation to reduce the skin cancer burden in arsenic-exposed individuals worldwide.