m⁶A Modification in Coding and Non-coding RNAs: Roles and Therapeutic Implications in Cancer

Abstract

N⁶-methyladenosine (m⁶A) RNA modification emerges in recent years as a new layer of regulatory mechanism controlling gene expression in eukaryotes. As a reversible epigenetic modification found not only in messenger RNAs (mRNAs), but also in non-coding RNAs (ncRNAs), m⁶A affects the fate of the modified RNA molecules and plays important roles in almost all vital bioprocesses, including cancer development. Here we review the up-to-date knowledge of the pathological roles and underlying molecular mechanism of m⁶A modifications (in both coding and non-coding RNAs) in cancer pathogenesis and drug response/resistance, and discuss the therapeutic potential of targeting m⁶A regulators for cancer therapy.

The regulation and function of RNA m⁶A modification

RNA m⁶A modification in eukaryotic RNA

Nucleotides, composed of nucleosides and phosphate groups, are the molecular building blocks of DNA and RNA. Accumulating number of chemical modifications on adenosine (A), guanosine (G), cytidine (C), and uridine (U) nucleosides on RNA have been identified since 1950s, adding additional complexity into the RNA world by broadly affecting the structure, biogenesis and function of RNA. N⁶-methyladenosine (m⁶A), methylated adenosine at the N⁶ position, is a widespread and abundant modification in messenger RNA (mRNA) and non-coding RNAs (ncRNAs), and represents one of the well-studied RNA modifications thus far. Since the first discovery in 1970s (Desrosiers et al., 1974; Perry and Kelley, 1974), m⁶A has been identified as the most prevalent internal modification in mRNA in most eukaryotic species including mammals (Adams and Cory, 1975; Desrosiers et al., 1974; Perry and Kelley, 1974; Perry et al., 1975), insects (Levis and Penman, 1978), plants (Nichols, 1979), yeast (Clancy et al., 2002), as well as in some viruses (Aloni et al., 1979; Beemon and Keith, 1977). However, this field did not move forward too much in several decades due to the lack of molecular biology, quantification and sequencing methodologies to comprehensively study m⁶A modifications in the transcriptome (Deng et al., 2018a; Deng et al., 2018c; Huang et al., 2020b).

The identification of the fat mass and obesity-associated protein (FTO) as the first m⁶A demethylase in 2011 suggested that m⁶A modification is reversible and dynamic and thus is likely functionally important (Jia et al., 2011). Thereafter, the functional importance of m⁶A modifications has been reported in virtually all major bioprocesses, normal development and diseases (including cancers) (reviewed in (Deng et al., 2018c; Frye et al., 2018; Roundtree et al., 2017a)). The landscape of m⁶A in the transcriptome (so called “epitranscriptome”) was first delineated by next generation sequencing (NGS) in 2012 (Dominissini et al., 2012; Meyer et al., 2012). m⁶A were identified in approximately one third of the mammalian mRNAs, with an average of 3-5 m⁶A modifications in each mRNA, and many m⁶A sites are evolutionally conserved between humans and mice. The deposition of m⁶A modification in the transcriptome is not random. The m⁶A modification sites have a typical consensus sequence DRACH (D=G, A or U; R= G or A; H= A, C or U), and are enriched in the coding sequence (CDS) and 3’ untranslated region (3’ UTR), with a particularly high enrichment around the stop codon area (Dominissini et al., 2012; Meyer et al., 2012). Thus far, several antibody-dependent (such as methylated RNA immunoprecipitation sequencing (MeRIP-seq) and m⁶A individual-nucleotide-resolution cross-linking and immunoprecipitation (miCLIP)) (Dominissini et al., 2012; Linder et al., 2015; Meyer et al., 2012) and -independent (such as MAZTER-seq, m⁶A-sensitive RNA-endoribonuclease–facilitated sequencing (m⁶A-REF-seq), and deamination adjacent to RNA modification targets sequencing (DART-seq)) (Garcia-Campos et al., 2019; Meyer, 2019; Zhang et al., 2019e) NGS methods have been developed for m⁶A sequencing, which make high-resolution detection of m⁶A epitrancriptome in diverse cell contexts a reality (reviewed by Huang et al., 2020a). The advent of such NGS-based methods advanced our understanding of this epigenetic mark. It has now become clear that m⁶A exists on almost every type of RNAs, including mRNAs, ribosomal RNAs (rRNAs), long non-coding RNAs (lncRNAs), microRNAs (miRNAs), small nulear RNAs (snRNAs), circular RNAs (circRNAs), and is dynamically regulated during many physiological and pathological processes, including tumorigenesis.

Dynamic regulation of m⁶A by ‘writers’ and ‘erasers’

Similar to DNA and protein, RNA can be methylated and demethylated by dedicated methyltransferases (also known as ‘writers’) and demethylases (also known as ‘erasers’), respectively. As the first known reversible mRNA modification, m⁶A has become a main focus in the epitranscriptomics field and a set of ‘writers’ and ‘erasers’ proteins of m⁶A have been characterized (Figure 1).

Figure 1.

Reversible m⁶A modification on mRNA. The adenosine (A) bases reside in mRNA could be methylated to form N⁶-methyladenosine (m⁶A) by the large MTC writer complex comprised of the METTL3-METTL14-WTAP core component and other regulatory cofactors or by METTL16 alone. This enzymatic reaction uses S-Adenosyl methionine (SAM) as a methyl donor. m⁶A could be recognized by m⁶A binding proteins (readers) to affect mRNA fate, or could be reversibly removed by m⁶A eraser proteins (i.e., FTO and ALKBH5). The demethylatio process requires α-Ketoglutaric acid (α-KG) and molecular oxygen (O₂) as co-substrates and ferrous iron (Fe²⁺) as a cofactor.

The deposition of m⁶A modifications in mRNAs is executed by a multicomponent m⁶A methyltransferase complex (MTC), which is comprised of a core component, the methyltransferase-like 3 (METTL3)/methyltransferase-like 14 (METTL14) heterodimer, and other regulatory factors including WTAP, KIAA1429, ZC3H13, and RBM15/RBM15B (Knuckles et al., 2018; Liu et al., 2014; Ping et al., 2014; Schwartz et al., 2014; Wen et al., 2018; Yue et al., 2018). Within this complex, METTL3 is the sole catalytic subunit that binds to the methyl donor S-adenosyl methionine (SAM) and catalyzes methyl group transfer (Sledz and Jinek, 2016; Wang et al., 2016a; Wang et al., 2016b), whereas METTL14 is critical for m⁶A deposition by stabilizing METTL3 conformation and recognizing substrate RNAs (Sledz and Jinek, 2016; Wang et al., 2016a; Wang et al., 2016b). Recently, it was reported that METTL14 could recognize histone H3 lysine 36 trimethylation (H3K36me3) modification and mediate the selectivity of m⁶A deposition in mRNA (Huang et al., 2020a; Huang et al., 2019a). Distinct from the regulatory role of H3K36me3 in universal m⁶A modification, some transcription factors, such as ZFP217, SMAD2/3, and CEBPZ, could regulate m⁶A deposition in certain mRNAs in some special cell contexts by recruiting or repelling MTC (Aguilo et al., 2015; Barbieri et al., 2017; Bertero et al., 2018). Although MTC catalyzes most of the m⁶A methylation in poly(A) RNA, methyltransferase like 16 (METTL16), zinc finger CCHC-type containing 4 (ZCCHC4), and methyltransferase like 5 (METTL5) have also been identified as m⁶A methyltransferases, which could function alone and catalyze m⁶A on some structured RNAs, such as U6 snRNA, 28S rRNA, and 18S rRNA, respectively (Brown et al., 2016; Ma et al., 2019; Mendel et al., 2018; Pendleton et al., 2017; Tran et al., 2019; Warda et al., 2017).

RNA m⁶A modification could be removed by alpha-ketoglutarate (αKG)- and Fe(II)-dependent demethylases, specifically, the fat mass and obesity-associated protein (FTO) and alkB homolog 5 (ALKBH5) (Jia et al., 2011; Zheng et al., 2013), thus conferring dynamic regulation of m⁶A methylation under physiological and pathological conditions. As the first reported human demethylase of m⁶A, FTO could mediate demethylation of both internal m⁶A and 5’ cap m⁶A_m (N⁶,2'-O-dimethyladenosine) in mRNA (Jia et al., 2011; Linder et al., 2015). However, internal m⁶A in mRNA is the major substrate of FTO in many cell types studied so far, including acute myeloid leukemia (AML) cells (Su et al., 2018; Wei et al., 2018; Zhang et al., 2019d). FTO plays oncogenic roles by promoting proliferation and inhibiting differentiation in AML cells (Li et al., 2017d), in which FTO is extensively expressed in cytoplasm and dramatically demethylates cytoplasmic m⁶A in mRNA, linking the oncogenic function of FTO to its cytoplasmic m⁶A demethylation activity (Wei et al., 2018). In other cases, functional roles of FTO inside nucleus still need to be addressed by additional studies. Unlike FTO, ALKBH5 seems to be an m⁶A-specific demethylase.

The writers and erasers are important for maintaining proper m⁶A level and gene expression across human tissues and cells. The dynamic balance between the deposition and removal of m⁶A modification is essential for normal bioprocesses and development. Therefore, mutation or dysregulation of writers and erasers is usually associated with diseases, such as cancers, by aberrantly adding or removing of m⁶A in RNA transcripts with critical biological functions.

Multifunction of m⁶A on mRNA fate decision

Gene expression is tightly controlled at different levels, including transcription, post-transcription, translation and post-translation. m⁶A is deposited on native RNA transcripts during transcription and influences gene expression post-transcriptionally through altering RNA structure or specific recognition by m⁶A-binding proteins (also known as ‘readers’) (Figure 2A).

Figure 2.

Functions of m⁶A modifications on coding and non-coding RNAs. m⁶A marks are deposited cotranscriptionally into nascent mRNAs, which undergo alternative splicing with the recruitment of splicing factors to m⁶A sites or flanking sequences. After splicing, m⁶A-containing mRNAs are recognized by YTHDC1 and exported into cytoplasm. Existence of m6A on mature mRNAs affects mRNA stability (stabilized by IGF2BP1/2/3 proteins and destabilized by YTHDF2/3 proteins), translation initiation, and translation elongation. m⁶A marks on lncRNAs play roles in RNA-RNA interaction, RNA-protein interaction, and chromatin remodeling. Anti-sense lncRNAs could modulate m⁶A abundance on the sense mRNA by recruiting m⁶A eraser (i.e., ALKBH5). Presence of m⁶A on pri-miRNA facilitates miRNA processing. In circRNAs, m⁶A could promote protein synthesis or inhibit circRNA immunity.

The m⁶A writers and erasers are located in nuclear speckles, where they are associated with mRNA splicing factors, implying the functional relevance of m⁶A with mRNA splicing (Bartosovic et al., 2017; Jia et al., 2011; Ping et al., 2014; Zheng et al., 2013). Indeed, m⁶A on precursor mRNAs (pre-mRNAs) could recruit splicing factor heterogeneous nuclear ribonucleoproteins A2/B1 (hnRNPA2B1) or increase the accessibility of flanking RNA sequence to splicing factors heterogeneous nuclear ribonucleoproteins C (hnRNPC) and hnRNPG by changing local structure, a mechanism known as “m⁶A switch” (Alarcon et al., 2015a; Liu et al., 2015; Liu et al., 2017; Zhou et al., 2019). On the other hand, m⁶A could also be bound by the reader protein YTH domain-containing protein 1 (YTHDC1), which recruits the splicing factor serine/arginine-rich splicing factor 3 (SRSF3) but repels SRSF10 to promote exon inclusion in targeted mRNAs (Xiao et al., 2016). YTHDC1 also affects the exportation m⁶A-modified mRNA transcripts from nucleus to cytoplasm (Roundtree et al., 2017b).

Cytoplasmic RNA could be loaded on ribosome for active translation or be sorted to messenger ribonucleoprotein (mRNP) foci, such as processing bodies (P-bodies) and stress granules, for degradation or storage. The YT521-B homology (YTH) domain-containing family proteins (YTHDFs) tend to accelerate metabolism of m⁶A-modified mRNAs in the cytoplasm. YTHDF1 selectively recognizes m⁶A and interacts with initiation factor eIF3 to promote translation initiation and protein synthesis (Wang et al., 2015). In contrast, YTHDF2 brings m⁶A-modified translatable mRNAs to mRNA decay sites (e.g., P-bodies), and recruits CCR4-NOT deadenylase complex to trigger deadenylation and degradation of the transcripts (Du et al., 2016; Wang et al., 2014a). YTHDF3 promotes mRNA translation in synergy with YTHDF1 and accelerated decay of m⁶A-containing mRNAs through interaction with YTHDF2 (Li et al., 2017a; Shi et al., 2017). In contrast to the destabilizing function of YTHDF2/3, the insulin-like growth factor-2 mRNA-binding protein (IGF2BP) family, including IGF2BP1, IGF2BP2, and IGF2BP3, protect m⁶A-modified mRNAs in P-bodies and stress granules from degradation through interacting with ELAV like RNA binding protein 1 (ELAVL1, also known as HuR), matrin 3 (MATR3), and poly(A) binding protein cytoplasmic 1 (PABPC1), and facilitate mRNA translation (Huang et al., 2018). Interestingly, METTL3 can also act as a reader in cytoplasm to promote translation of a subset of m⁶A-modified mRNAs independently of its methyltransferase activity (Choe et al., 2018; Lin et al., 2016). Since m⁶A is usually enriched around stop codon and far away from the translation initiation site, the role of m⁶A in promoting translation initiation is achieved by a closed-loop model, where mRNA circularization is mediated through the interaction between the eukaryotic translation initiation factors (eIFs) subunit at the 5’end of the mRNA and METTL3 or YTHDF1 bound to m⁶A sites near stop codon (Choe et al., 2018; Wang et al., 2015). It was also reported that m⁶A resided in 5’UTR could promote cap-independent translation through recruiting eIF3a to nearby translation initiation sites (Meyer et al., 2015). During translation elongation, although m⁶A-modified codon is same with unmodified codon in terms of amino acid coding, m⁶A modification on mRNA acts as a barrier to delay tRNA accommodation and thereby perturbs translation elongation dynamics (Choi et al., 2016).

m⁶A modification in ncRNAs

ncRNAs are a group of endogenous RNA molecules that are not translated into proteins but have specialized functions in the regulation of gene expression, which can be divided into lncRNAs with length exceeding 200 nucleotides (nt) and small ncRNAs less than 200 nt in length. In addition to protein-coding mRNAs, m⁶A modifications were found in ncRNAs including lncRNAs and small ncRNAs (such as miRNAs and snRNAs), and were shown to be important for their expression and functions (Brown et al., 2016; Linder et al., 2015; Liu et al., 2013; Meyer et al., 2012; Pendleton et al., 2017; Warda et al., 2017) (Figure 2B).

The m⁶A methylation in lncRNAs was observed when mapping m⁶A on poly(A) RNAs (Dominissini et al., 2012; Meyer et al., 2012). Without coding ability, lncRNAs are involved in gene expression regulation via interaction with RNA binding proteins, crosstalk with other RNA species, or chromatin remodeling. It was reported that m⁶A modification on lncRNA could affect RNA-protein interaction. For instance, metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) is a conserved lncRNA whose mutation or upregulation has been consistently associated with tumorigenesis and metastasis. MALAT1 is often highly methylated with m⁶A, with multiple m⁶A sites detected by MeRIP-seq and site-specific cleavage and radioactive-labeling followed by ligation-assisted extraction and thin-layer chromatography (SCARLET) (Dominissini et al., 2012; Liu et al., 2013; Meyer et al., 2012). Two of these m⁶A residues were reported to prevent formation of RNA local secondary structures and increase the recognition and binding of hnRNPC to a U5-tract in the MALAT1 hairpin through an ‘m⁶A switch’ mechanism (Liu et al., 2015). It was recently reported that m⁶A also plays a role on the RNA-RNA interaction function of lncRNAs, where internal m⁶A modification of the large intergenic coding RNA 1281 (linc1281) is required for sequestering pluripotency-related let-7 family miRNAs to ensure mouse embryonic stem cell identity (Yang et al., 2018). On the other hand, lncRNAs could also interact with m⁶A regulators to facilitate their function. For example, FOXM1-AS, a lncRNA antisense to FOXM1, promotes the interaction of ALKBH5 with FOXM1 nascent transcripts (Zhang et al., 2017c). Depleting FOXM1-AS phenocopied depletion of ALKBH5 in regulating FOXM1 methylation and expression (Zhang et al., 2017c). Similarly, GAS5-AS, a lncRNA antisense to GAS5, could enhanced GAS5 stability by interacting with ALKBH5 and regulating m⁶A modifications of GAS5 (Wang et al., 2019c). It appears that the m⁶A modulating function might be common for antisense lncRNAs.

miRNAs are a class of small and highly abundant non-coding RNAs involved in gene silencing or post-transcriptional gene expression regulation. Transcribed from DNA, the primary transcripts of miRNAs (pri-miRNAs) undergo a series of cleavage to form hair-pin precursor, precursor miRNAs (pre-miRNAs), and mature miRNAs. The canonical m⁶A motif GGAC are enriched in pri-miRNAs, but not in pre-miRNAs and mature miRNAs, and most of them can be methylated by METTL3 in cellulo (Alarcon et al., 2015b). The pri-miRNAs with m⁶A marks could be recognized by hnRNPA2B1, which interacts with DGCR8 and promotes miRNA processing (Alarcon et al., 2015a). Therefore, alternation of m⁶A level by METTL3 knockdown or overexpression could result in significant changes in the mature miRNA pool (Alarcon et al., 2015b).

CircRNAs are a new type of ncRNAs which, unlike linear RNAs, form a covalently closed continuous loop by back splicing. Recently, it was found that m⁶A modifications were also widespread in circRNAs and were written and read by the same machinery as did mRNAs (Zhou et al., 2017). However, the m⁶A enrichment pattern in circRNAs is distinct from that of mRNAs, exhibiting enrichment at the translation start site of their corresponding mRNAs (Zhou et al., 2017). It appeared that recognition of m⁶A-modified circRNAs by YTHDF2 did not promote degradation of circRNAs but instead played a role in regulating mRNA stability and circRNA immunity (Chen et al., 2019c; Zhou et al., 2017). m⁶A could also drive protein synthesis from circRNAs through recruiting YTHDF3 and initiation factor eIF4G2 (Yang et al., 2017a).

Aberrant m⁶A regulation in cancers

Although m⁶A modification does not change base pairing and coding, it broadly affects gene expression in multiple levels through interacting with diverse reader proteins and associated complexes. Therefore, dynamic m⁶A modification is critical for many normal bioprocesses, including self-renewal and differentiation of embryonic stem cells and hematopoietic stem cells, tissue development, circadian rhythm, heat shock or DNA damage response, and sex determination (Alarcon et al., 2015b; Chen et al., 2015; Deng et al., 2018c; Geula et al., 2015; Huang et al., 2020a; Wang et al., 2014b; Weng et al., 2019; Weng et al., 2018; Xiang et al., 2017; Zhang et al., 2017a; Zhao et al., 2017a; Zhao et al., 2017b; Zhao et al., 2014; Zheng et al., 2013; Zhou et al., 2015). Emerging data have suggested that the global abundance of m⁶A and expression levels of its regulators, including writers, erasers, and readers, are often dysregulated in various types of cancers and are critical for cancer initiation, progression, metastasis, as well as drug resistance and cancer relapse (Table 1 and Figure 3) (Huang et al., 2020b). Other intracellular events (such as mutations that cause gain or loss of m⁶A sites) and extracellular stimulations could also affect cellular m⁶A modification and are therefore associated with human cancers (Figure 4).

Table 1.

Roles of m⁶A writers, erasers and readers in human cancers.

m6A regulator	Role in cancer	Cancer type	Target genes	References
Writer
METTL3	Oncogene	Acute myeloid leukemia	MYC, BCL2, PTEN, SP1, SP2	(Barbieri et al., 2017; Vu et al., 2017)
		Glioblastoma	SRSFs, SOX2	(Li et al., 2019a; Visvanathan et al., 2018)
		Hepatocellular carcinoma	SOCS2, SNAIL, LINC00958	(Chen et al., 2018; Lin et al., 2019b; Zuo et al., 2020)
		Hepatoblastoma	CTNNB1	(Liu et al., 2019)
		Bladder cancer	AFF4, IKBKB, RELA, MYC, CDCP1, miR-221/222, ITGA6	(Cheng et al., 2019a; Han et al., 2019b; Jin et al., 2019b; Yang et al., 2019a)
		Breast cancer	HBXIP, BCL2	(Cai et al., 2018; Wang et al., 2020)
		Gastric cancer	ZMYM1, HDGF, SEC62, ARHGAP5-AS1	(He et al., 2019; Lin et al., 2019a; Liu et al., 2020b; Wang et al., 2019b; Yue et al., 2019; Zhu et al., 2019b)
		Pancreatic cancer	pri-miR-25	(Taketo et al., 2018; Xia et al., 2019; Zhang et al., 2019b)
		Colorectal cancer	SOX2, pri-miR-1246	(Li et al., 2019c; Peng et al., 2019)
		Colon cancer	TP53(R273H,G>A)	(Uddin et al., 2019)
		Melanoma		(Dahal et al., 2019)
		Osteosarcoma	LEF1	(Miao et al., 2019)
		Prostate	GLI1	(Cai et al., 2019)
		Non-small cell lung cancer	YAP, MALAT1	(Jin et al., 2019a)
	Tumor suppressor	Glioblastoma	ADAM19	(Cui et al., 2017)
		Endometrial cancer	PHLPP2, MTORC2	(Liu et al., 2018a)
		Renal cell carcinoma		(Li et al., 2017c)
METTL14	Oncogene	Acute myeloid leukemia	MYB, MYC	(Weng et al., 2018)
		Breast cancer	TGF–β signaling pathway genes	(Panneerdoss et al., 2018)
	Tumor suppressor	Endometrial cancer	PHLPP2, MTORC2	(Liu et al., 2018a)
		Glioblastomas	ADAM19	(Cui et al., 2017)
		Hepatocellular carcinoma	pri-miR-126	(Ma et al., 2017)
		Bladder cancer	NOTCH1	(Gu et al., 2019)
		Renal cell carcinoma	P2RX6	(Gong et al., 2019)
		Gastric cancer		(Zhang et al., 2019a)
		Colorectal cancer	pri-miR-375	(Chen et al., 2019a)
WTAP	Oncogene	Hepatocellular carcinoma	ETS1	(Chen et al., 2019b)
VIRMA (KIAA1429)	Oncogene	Hepatocellular carcinoma	ID2, GATA3	(Cheng et al., 2019b; Lan et al., 2019)
Eraser
FTO	Oncogene	Acute myeloid leukemia	ASB2, RARA, MYC, CEBPA	(Li et al., 2017d; Su et al., 2018)
		Glioblastomas		(Cui et al., 2017)
		Non-small cell lung cancer	MZF1, USP7	(Li et al., 2019b; Liu et al., 2018b)
		Pancreatic cancer	MYC	(Tang et al., 2019)
		Cervical squamous cell carcinoma	β-catenin, E2F1, MYC	(Zhou et al., 2018; Zou et al., 2019)
		Breast cancer	BNIP3	(Niu et al., 2019)
		Melanoma	PDCD1, CXCR4, SOX10	(Yang et al., 2019c)
		Gastric cancer		(Zhang et al., 2019a)
ALKBH5	Oncogene	Glioblastomas	FOXM1	(Zhang et al., 2017 c)
		Breast cancer	NANOG	(Fry et al., 2018; Wu et al., 2019; Zhang et al., 2016a; Zhang et al., 2016b)
		Non-small cell lung cancer	UBE2C, FOXM1	(Chao et al., 2020; Guo et al., 2018)
		Gastric cancer	NEAT1	(Zhang et al., 2019c)
		Ovarian carcinoma	BCL2	(Zhu et al., 2019a)
		Oral squamous cell carcinoma	FOXM1, NANOG	(Shriwas et al., 2020)
	Tumor suppressor	Bladder cancer	ITGA6	(Jin et al., 2019b)
		Pancreatic cancer	KCNK15-AS1, WIF1	(He et al., 2018; Tang et al., 2020)
Readers
YTHDF2	Oncogene	Acute myeloid leukemia	TNFRSF2	(Paris et al., 2019)
		Hepatocellular carcinoma		(Yang et al., 2017b)
		Prostate cancer		(Li et al., 2018)
		Pancreatic cancer	YAP	(Chen et al., 2017)
		Gastric cancer		(Zhang et al., 2017b)
		Non-small cell lung cancer	6PGD	(Sheng et al., 2019)
		Cervical cancer	GAS5	(Wang et al., 2019c)
	Tumor suppressor	Hepatocellular carcinoma	EGFR, IL11, SRPINE2	(Hou et al., 2019; Zhong et al., 2019)
YTHDF1	Oncogene	Colorectal cancer		(Bai et al., 2019; Nishizawa et al., 2018)
		Hepatocellular carcinoma		(Lin et al., 2019b; Zhao et al., 2018)
		Non-small cell lung cancer	KEAP1	(Shi et al., 2019)
		Markel cell carcinoma		(Orouji et al., 2020)
YTHDF3	Oncogene	Colorectal cancer	GAS5 lncRNA	(Ni et al., 2019)
YTHDC2	Oncogene	Colon cancer	HIF1A	(Tanabe et al., 2016)
IGF2BP1/2/3	Oncogene	Hepatocellular carcinoma and cervical cancer	MYC, FSCN1, TK1, MARCRSL1	(Huang et al., 2018)
IGF2BP1	Oncogene	Hepatocellular carcinoma and ovarian cancer	SRF	(Muller et al., 2019)
IGF2BP2	Oncogene	Colorectal cancer	SOX2	(Li et al., 2019c)
		Pancreatic cancer	DANCR	(Hu et al., 2019)
IGF2BP3	Oncogene	Gastric cancer	HDGF	(Wang et al., 2019b)
METTL3	Oncogene	Non-small cell lung cancer	EGFR, TAZ, BRD4	(Choe et al., 2018; Lin et al., 2016)
		Ovarian carcinoma	AXL	(Hua et al., 2018)

Figure 3.

Deregulation of m⁶A modifiers in human cancers. Modifiers in red indicate an oncogenic role, modifiers in green indicate a tumor suppressive role, while those in yellow have controversial roles reported, in the specific cancer type. AML, acute myeloid leukemia; BRC, breast cancer; NSCLC, non-small cell lung cancer; GC, gastric cancer; BLC, bladder cancer; PRC, prostate cancer; OST, osteosarcoma; GBM, glioblastoma; RCC, renal cell carcinoma; HCC, hepatocellular carcinoma; PAC, pancreatic cancer; CRC, colorectal cancer; GNC, gynecological cancer; CVC, cervical cancer; ENC, endometrial cancer; OVC, ovarian cancer; MLM, melanoma.

Figure 4.

Causes of aberrant m⁶A regulation in cancer. Deregulation of the m⁶A machinery could result in upregulation of oncogenes or downregulation of tumor suppressive genes, leading to cancer development. On the other hand, mutations from other nucleosides to A on mRNA render mRNA to m⁶A-mediated regulation and can also contribute to cancer. Environmental factors could reprogram the epitranscriptome and lead to cell immortalization.

Aberrant global m⁶A abundance in cancer

Aberrant decrease or elevation of global m⁶A abundance has been reported recently in some types of cancers, and the dysregulation could be associated with cancer progression and clinical outcome. For example, it was reported that the global m⁶A abundance in mRNA or total RNA (as detected by m⁶A dot blots or colorimetric ELISA-like assays) was significantly increased in human gastric cancer tissues compared to normal control tissues (Wang et al., 2019b). Another group found that human HCC exhibited a gain of global m⁶A abundance in mRNA and increased mRNA expression due to the reduction of YTHDF2 (Hou et al., 2019). In contrast, Gu et al. reported that the global m⁶A abundance was significantly decreased in human bladder cancer tissues, especially in more advanced bladder cancer, as detected by m⁶A dot blot assays and by bladder cancer tissue array and immunohistochemistry. The authors also showed that decreased m⁶A abundance is associated with poor prognosis in bladder cancer patients (Gu et al., 2019). In addition, it has been indicated that m⁶A is associated with drug response and could be an epigenetic driver of chemo-resistance in leukemia cells, where hypomethylation of m⁶A in response to elevation of FTO during tyrosine kinase inhibitor (TKI) treatment are required for obtaining TKI tolerance and growth advantage (Yan et al., 2018).

Dysregulation of m⁶A writers in cancer

METTL3 was first identified as an m⁶A methyltransferase more than 2 decades ago and found to be widely expressed in human tissues later (Bokar et al., 1994; Bokar et al., 1997; Tuck, 1992). However, it was until 2017 that the role of METTL3, as an m⁶A methyltransferase, was started to be implicated in human cancers. It was reported that METTL3 acts as an essential oncogene in AML cells, where it could be recruited to chromatin by the CAATT-box binding protein CEBPZ to induce m⁶A deposition on the associated mRNA transcripts, such as SP1 and SP2 (Barbieri et al., 2017). Vu et al. also reported that METTL3 is highly expressed in AML and plays a critical role in AML cell survival and leukemia progression through promoting translation of target mRNAs, including MYC, BCL2, and PTEN, in an m⁶A dependent manner (Vu et al., 2017). Since then the overexpression and oncogenic functions of METTL3 were reported in many other types of cancer (Table 1), including hepatocellular carcinoma (HCC) (Chen et al., 2018; Lin et al., 2019b), hepatoblastoma (Liu et al., 2019), gastric cancer (GC) (He et al., 2019; Lin et al., 2019a; Liu et al., 2020b; Wang et al., 2019b; Yue et al., 2019; Zhu et al., 2019b), colorectal cancer (CRC) (Li et al., 2019c; Peng et al., 2019), non-small cell lung cancer (NSCLC) (Choe et al., 2018), and bladder cancer (BLC) (Han et al., 2019b). For instance, METTL3 was shown to promote HCC growth and contribute to HCC progression by repressing SOCS2 expression through an YTHDF2-dependent mechanism (Chen et al., 2018), and was also involved in regulation of epithelial-mesenchymal transition (EMT) of HCC cells through YTHDF1-mediated promotion of Snail mRNA translation (Lin et al., 2019b). Both studies suggested upregulation of METTL3 as an adverse prognostic factor in patients with HCC. In GC, elevated METTL3 level could also serve as an independent predictor of poor prognosis (Wang et al., 2019a; Yue et al., 2019). Higher METTL3 expression was found in CRC metastatic tissues and was associated with a poor prognosis, with SOX2 being identified as a downstream target which could be stabilized by IGF2BP2 (Li et al., 2019c). The correlation of METTL3 expression with CRC metastasis was confirmed by another group, which reported that METTL3 methylated pri-miR-1246 to promote its maturation (Peng et al., 2019). The overexpression of METTL3 and its adverse prognostic impacts have also been reported in BLC, in which it promotes methylation of target mRNAs (e.g., AFF4, MYC, CDCP1, ITGA6) or miRNAs (i.e., pri-miR221/222) (Cheng et al., 2019a; Han et al., 2019b; Jin et al., 2019b; Yang et al., 2019a). Interestingly, the oncogenic function of METTL3 in NSCLC is complicated, where METTL3 could methylate YAP mRNA and recruit YTHDF1/3 and eIF3b to promote translation of YAP mRNA (Jin et al., 2019a), or could interact with eIF3h to promote translation of target mRNAs (e.g., BRD4) independent of its enzymatic activity (Choe et al., 2018; Lin et al., 2016). The enzyme activity-independent oncogenic role of METTL3 was also reported in ovarian cancer (OVC) (Hua et al., 2018).

Similar to METTL3, METTL14 also contains a MT-A70 domain (i.e., a methyltransferase domain, MTD) and forms a heterodimer complex with METTL3 (Liu et al., 2014). Recent structural studies suggested that METTL14's catalytic center is degenerated and is thus not involved in the catalyzing reaction; instead, METTL14 is required for stabilizing METTL3 conformation and substrate RNA binding (Sledz and Jinek, 2016; Wang et al., 2016a; Wang et al., 2016b). As a critical component of the MTC, METTL14 has also been shown to be overexpressed and play important roles in various types of cancers. METTL14 was first reported to be overexpressed and play a critical oncogenic role in the initiation and progression of AML (Weng et al., 2018). Depletion of METTL14 inhibited self-renewal of leukemia stem/initiating cells (LSCs/LICs) and promoted myeloid differentiation of AML cells, mainly through reducing m⁶A abundance on its target mRNAs MYB and MYC and therefore negatively regulating their mRNA stability and translation (Weng et al., 2018). In normal hematopoiesis, Mettl14 was also found to be highly expressed in murine hematopoietic stem cells and down-regulated during myelopoiesis, especially in mature myeloid cells; knockout of Mettl14 moderately inhibited self-renewal of mouse hematopoietic stem cells (HSCs) and promoted myeloid differentiation (Weng et al., 2018). In breast cancer (BRC) cells, METTL14 and ALKBH5 were shown to control each other’s expression and play pro-tumorigenic function by controlling the m⁶A levels in key EMT and angiogenesis-associated transcripts (Panneerdoss et al., 2018).

WTAP was shown to be upregulated and act as an oncogene in AML before it was identified as a component of the m⁶A methyltransferase complex (Bansal et al., 2014). Recently, an m⁶A related function of WTAP was reported in HCC, where WTAP was upregulated and promoted liver cancer development via the huR-ETS1-p21/p27 axis (Chen et al., 2019b). Another component of the MTC, VIRMA, was also shown to have higher expression in HCC than in normal liver tissues and enhance migration and invasion of HCC, through regulation on the ID2 mRNA or GATA3 pre-mRNA in an m⁶A dependent manner (Cheng et al., 2019b; Lan et al., 2019).

In contrast, a tumor suppressive role of METTL3 was suggested in renal cell carcinoma (RCC) (Li et al., 2017c) and endometrial cancer (ENC) (Liu et al., 2018a), where METTL3 expression was lower in tumor samples compared to adjacent non-tumor samples, leading to the activation of the AKT signaling pathway and the increased proliferation and tumorigenicity of such tumor cells. It should be noted that a controversial role of METTL3 in glioblastoma (GBM) was reported by different groups (Cui et al., 2017, Li et al., 2019a; Visvanathan et al., 2018), which may be caused by the different origin of primary samples used and hence reflect the heterogeneity of GBM. Similarly, in several types of cancer, METTL14 was also shown to be downregulated and exhibit a tumor suppressive role. A hotspot R298P mutation was present in METTL14 in human ENC, resulting in the reduction m⁶A mRNA methylation in the patient endometrial tumor samples and could promote the tumorigenicity of endometrial cancer cells (Liu et al., 2018a). The level of METTL14 as well as m⁶A was reduced in HCC, accounting for high metastatic capacity of HCC (Ma et al., 2017). The down-regulation of METTL14 and decreased global m⁶A abundance were also observed in human BLC, and were associated poor prognosis and more advanced disease stages in patients with BLC (Gu et al., 2019). By post-transcriptional suppression of NOTCH1 expression via an m⁶A-dependent mechanism, METTL14 inhibits bladder tumor-initiating cell (TIC) self-renewal and bladder tumorigenesis (Gu et al., 2019). In addition, the down-regulation and tumor-suppressive role of METTL14 (similar to METTL3) has also been reported in GBM (Cui et al., 2017).

Dysregulation of m⁶A erasers in cancer

FTO is the first identified RNA m⁶A demethylase (Jia et al., 2011), although it had been shown to also exhibit catalytic activity towards other types of modification on RNA and DNA (Jia et al., 2008). The pathological function of FTO in cancer as an m⁶A demethylase was first reported in AML, where FTO was found to be highly expressed (Li et al., 2017b). It was demonstrated that FTO acts as an oncogene to promote leukemogenesis and inhibit all-trans-retinoic acid (ATRA)-mediated leukemia cell differentiation by destabilizing ASB2 and RARA mRNA via reducing m⁶A abundance on these mRNA transcripts (Li et al., 2017b). Further studies revealed FTO as a direct target of R-2-hydroxyglutarate (R-2HG), a metabolite produced to high levels by mutant isocitrate dehydrogenase 1/2 (IDH1/2) enzymes (Su et al., 2018). Activity of FTO could be inhibited by R-2HG, leading to the increase of m⁶A abundance and decreased stability of MYC and CEBPA mRNAs, which subsequently caused growth inhibition, cell cycle arrest and apoptosis of leukemia cells; as a result, R-2HG displays an intrinsic anti-leukemia activity (Su et al., 2018). The oncogenic function of FTO was also reported in solid tumors, such as GBM (Cui et al., 2017), BRC (Niu et al., 2019), and melanoma (MLM) (Yang et al., 2019c). Inhibition of FTO by an FTO inhibitor MA2 phenocopied effects of METTL3 overexpression on suppressing GBM stem cell (GSC) growth and self-renewal, and on suppressing tumor progression in xenograft mice (Cui et al., 2017). In BRC, FTO promotes cancer cell proliferation and metastasis via negative regulation on BNIP3 mRNA, and high level of FTO is associated with poor prognosis (Niu et al., 2019). Recently, it was shown that FTO could be induced by metabolic stress through the autophagy and NF-κB pathway in melanoma; as a result, FTO expression was elevated in human MLM compared to normal skin tissues, and high level of FTO promoted growth of MLM cells in nude mice (Yang et al., 2019c).

ALKBH5, belonging to the same AlkB subfamily of the Fe(II)/ αKG dioxygenases as FTO, was the second identified RNA m⁶A demethylase (Zheng et al., 2013). Soon after its identification as an m⁶A eraser, ALKBH5 was found to be involved in the enrichment of breast cancer stem cells (BCSCs) in the hypoxic tumor microenvironment (Zhang et al., 2016a; Zhang et al., 2016b). Expression of ALKBH5 could be induced in a hypoxia-inducible factors-dependent manner in breast cancer cells under hypoxia, leading to the decreased m⁶A abundance in NANOG mRNA and the increased protein level of this pluripotency factor, which allowed for the enrichment of BCSCs (Zhang et al., 2016a; Zhang et al., 2016b). Another study showed that ALKBH5 was highly expressed in GSCs and that its high expression level correlated with poor clinical outcome in GBM patients (Zhang et al., 2017c). The in vitro and in vivo assays using cultured human GSCs and xenograft models suggested that ALKBH5 is required for GSC proliferation and selfrenewal. Interestingly, FOXM1-AS, a chromatin-bound lncRNA transcribed in the antisense orientation of FOXM1, facilitated the recruitment of ALKBH5 to its nascent target transcript FOXM1. Demethylation of FOXM1 pre-mRNA by ALKBH5 promoted HuR binding and resulted in increased stability of the FOXM1 pre-mRNA and enhanced FOXM1 protein expression, which is critical for the maintenance of GSC tumorigenicity (Zhang et al., 2017c). A more recent study investigated the function of m⁶A RNA modification in bladder cancer and found that ALKBH5, in opposite to METTL3, inhibited bladder cancer cell growth and progression repressing ITGA6 expression through m⁶A-dependent post-transcriptional regulation (Jin et al., 2019b). In pancreatic ductal adenocarcinoma (PDAC), ALKBH5 was also shown to exhibit a tumor suppressive role by decreasing m⁶A modification on WIF-1 RNA and mediating Wnt signaling, and overexpression of ALKBH5 sensitized PDAC cells to chemotherapy (Tang et al., 2020).

Dysregulation of m⁶A readers in cancer

The m⁶A reader proteins mediate effects of m⁶A modification by controlling the fate of the modified RNAs. Thus, deregulation of reader proteins may lead to misinterpretation of modified RNAs and the subsequent disorder of RNA metabolism. As the first reported m⁶A reader, YTHDF2 has been extensively studied in various cancer types, and was shown to play an oncogenic role in most cancer types. In prostate cancer (PRC), YTHDF2 was upregulated in cancer tissues compared to the adjacent normal tissues; miR-493-3p was found to be a negative regulator of YTHDF2, and inhibition of miR-493-3p could abrogate the inhibitory effect of YTHDF2 knockdown on PRC cell proliferation and migration (Li et al., 2018). In agreement with the upregulation of METTL3 and METTL14 in AML, YTHDF2 was expressed higher in AML samples compared to healthy controls and in AML cells with LSC activity compared to those without LCS activity (Paris et al., 2019). In addition, YTHDF2 was critical for AML initiation and propagation by shortening the half-life of diverse m⁶A-containing transcripts important for LSC function (Paris et al., 2019). However, YTHDF2 has been reported to play both oncogenic and tumor suppressive roles in HCC (Yang et al., 2017; Zhong et al., 2019). By examining expression of YTHDF2 in clinical HCC tissues using immunohistochemistry staining, Yang et al. showed that YTHDF2 expression is associated with the malignance of HCC, and is negatively regulated by miR-145, a miRNA that is downregulated in HCC patients (Yang et al., 2017). In contrast, Zhong et al. showed that YTHDF2 expression was suppressed in HCC under hypoxic environment, and forced expression of YTHDF2 inhibited HCC cell proliferation and growth in vitro and in vivo by directly binding to the m⁶A modification site of EGFR 3 ’ UTR to promote the degradation of EGFR mRNA (Zhong et al., 2019).

Among all YTH family proteins, YTHDF1 was the only one showing increased expression in CRC samples than in normal portions according to the TCGA database. CRC patients with high YTHDF1 expression had poor overall survival, suggesting YTHDF1 as a prognostic factor (Nishizawa et al., 2018). DNA copy number amplification or potential transcriptional regulation by the oncogenic MYC protein may account for the overexpression of YTHDF1 in CRC (Bai et al., 2019; Nishizawa et al., 2018). Similarly, YTHDF1 was highly expressed in liver cancer and acted as an adverse prognosis factor in liver cancer patients (Lin et al., 2019b; Zhao et al., 2018). A very recent study identified YTHDF1 as one of the genes evolved rapidly in hypoxia adaptation, whose expression decreased in Tibetan domestic mammals compared to lowlanders (Shi et al., 2019). Furthermore, YTHDF1 was shown to be amplified in NSCLC and critical for NSCLC cell proliferation, xenograft tumor formation and lung cancer progression; surprisingly, it was also found that low expression of YTHDF1 rendered cancer cells resistant to cisplatin treatment and resulted in a worse clinical outcome (Shi et al., 2019). This study highlights the important roles of YTHDF1 in both hypoxia adaptation and NSCLC pathogenesis.

Deregulation of other YTH family proteins in cancer was also reported. YTHDC2 expression was found to be positively correlated with the colon tumor stage, including metastasis; high expression of YTHDC2 contributed to colon tumor metastasis by promoting translation initiation of HIF-1α (Tanabe et al., 2016). A GAS5-YAP-YTHDF3 negative feedback loop was recently uncovered in CRC (Ni et al., 2019). The YAP-interacting lncRNA GAS5 could directly bound with YAP to facilitate its phosphorylation and ubiquitin-mediated degradation, leading to attenuated YAP-mediated transcription of YTHDF3 and the subsequent relief of GAS5 from YTHDF3-mediated m⁶A-dependent decay. As YTHDF3 was expressed at significant higher level in CRC tissues than in counterpart normal tissues, this negative feedback loop sustained YTHDF3 expression and established a key role of YTHDF3 in CRC progression (Ni et al., 2019). Although IGF2BP1/2/3 proteins were just recently identified as m⁶A readers (Huang et al., 2018), their elevated expression had been previously observed in various cancers (Bell et al., 2013; Degrauwe et al., 2016). Among them, IGF2BP1 and IGF2BP3 are oncofetal proteins which are produced by tumors as well as by fetal tissues but are downregulated in adult tissues (Lederer et al., 2014). All the IGF2BP proteins were shown to promote cell proliferation, colony formation, migration, and invasion of HCC cells and cervical cancer (CVC) cells through binding to m⁶A sites on target transcripts, including MYC, to stabilize these mRNAs and promote their translation (Huang et al., 2018). Recently, it was shown that IGF2BP1 promoted the expression of SRF in an m⁶A-dependent manner by impairing miRNA-directed decay of the SRF mRNA, leading to enhanced SRF-dependent transcriptional activity and tumor cell growth and invasion (Muller et al., 2019). Significant elevation of IGF2BP3 expression was found in gastric cancer compared to paired normal gastric mucosa, and the binding of IGF2BP3 to the m⁶A site in HDGF mRNA enhanced its stability, leading to the increased expression and secretion of HDGF and the subsequent gastric cancer growth and liver metastasis (Wang et al., 2019b). In addition, IGF2BPs have also been shown to bind to ncRNAs, such as lncRNAs (Hu et al., 2019; Lim et al., 2019). It was reported that IGF2BP2, which is upregulated and associated with poor outcomes in pancreatic cancer patients, stabilizes lncRNA DANCR by binding to m⁶A-modified DANCR (Hu et al., 2019). Whether IGF2BPs recognize other ncRNAs through m⁶A-dependent manners in cancer and what are the functional consequences have yet to be investigated.

Mutations in m⁶A sites in cancers

It has now been realized that dysregulation of m⁶A regulators, especially of writers and erasers, causes abnormal m⁶A modifications in critical transcripts and aberrantly regulates expression of these cancer-associated genes post-transcriptionally. Therefore, it is reasonable to speculate that mutations in the m⁶A sites of such transcripts may disturb m⁶A deposition and thus play a role in cancer. Indeed, a G>A mutation in codon 273 (resulting in the R273H mutation) of TP53 pre-mRNA was reported to confer drug resistance of colon cancer cells in an m⁶A dependent manner (Uddin et al., 2019). The transited adenosine in codon 273 was methylated by METTL3, resulted in promoted preferential pre-mRNA splicing and the production of R273H mutant p53 protein, which led to acquired multidrug resistance in colon cancer cells (Uddin et al., 2019). More recently, a large-scale population study identified a missense variant, the rs8100241 variant located in the exon of ANKLE1 with a G>A change (resulting in Ala>Thr change) in CRC, and showed that it was associated with decreased risk of CRC (Tian et al., 2019). Mechanistically, compared to the rs8100241[G] allele, the rs8100241[A] allele could be methylated by the m⁶A MTC and recognized by YTHDC1, which increased protein expression of ANKLE1, a potential tumor suppressor that inhibited cell proliferation by maintaining genomic stability (Tian et al., 2019). Besides the gain of de novo m⁶A sites due to mutations in cancers, mutations resulting in loss of m⁶A modification may also exist and contribute to cancer development and drug response. Some online tools are available that incorporate m⁶A site information and SNP information to facilitate the investigation of functional m⁶A site mutations in cancers (Jiang et al., 2018; Zheng et al., 2018). It is also of great importance that high-resolution m⁶A sequencing methods are available to obtain more accurate and comprehensive view of the m⁶A methylome in each individual cancer types and corresponding normal tissues.

m⁶A epitranscriptome change upon external stimulations

It was estimated that exposure to a wide variety of environmental factors, including smoke, alcohol, radiation, environmental chemicals, and viruses, accounts for at least two-thirds of all cancer cases in the United States according to a report from the NIH/NCI/NIEHS (https://www.niehs.nih.gov/health/materials/cancer_and_the_environment_508.pdf). In addition to the intracellular events such as dysregulation of m⁶A regulators and mutations in m⁶A sites, external environmental stimulations have also been shown to regulate m⁶A modification and contribute to cancer development.

Chemical carcinogens can cause genetic mutations in cancer-susceptibility genes and/or induce epigenetic changes to promote oncogenic transformation (Clark and Molloy, 2017; Luch, 2005; Vaz et al., 2017). Human bronchial epithelial (HBE) cells chronically treated with arsenite sodium showed significantly elevated m⁶A modification and were transformed with malignant phenotypes, which could be dramatically reversed by silencing of METTL3 (Gu et al., 2018). In another study, different types of human epithelial cells transformed by chemical carcinogens, including cadmium (Cd), 3-methylcholanthrene, and nickel (Ni), showed dynamic changes in m⁶A abundance (Yang et al., 2019b). Mechanistic studies identified CDCP1, an oncogene, as a critical m⁶A target in chemical-induced malignant transformation and suggested that the METTL3-m⁶A-CDCP1 axis could be a potential therapeutic target for the treatment of chemical-induced cancers (Yang et al., 2019b).

The involvement of m⁶A modification in cigarette smoke-induced cancer development was recently reported (Zhang et al., 2019b). Exposure of immortalized human pancreatic duct epithelial cells or pancreatic cancer cells to cigarette smoke condensate caused hypomethylation in the promoter of METTL3 and the recruitment of NFIC transcription factor to induce METTL3 expression. High level of METTL3 significantly catalyzed m⁶A formation in pri-miR-25 and resulted in the increased level of mature miR-25-3p, which suppressed PHLPP2 mRNA expression and evoked the oncogenic AKT-p70S6K signaling to promote the initiation and progression of PDAC (Zhang et al., 2019b).

Epstein–Barr virus (EBV) is the first human oncogenic virus discovered and contributes to approximately 2% of all human cancers. It was recently shown that EBV latent protein EBNA3C could activate transcription of METTL14 and promote METTL14 protein stability by directly interacting with METTL14 in the host cells. Increased expression of METTL14 in EBV latently infected cells reprogrammed the epitranscriptome and upregulated essential EBV latent antigens, including EBNA3C, via an m⁶A-dependent mechanism, and therefore contributed to EBV-mediated tumorigenesis (Lang et al., 2019).

Implications for cancer therapy

Targeting m⁶A regulators as cancer therapies

The identification of FTO as the first m⁶A demethylase revived research in this field, and since then FTO has been the most attractive target for developing inhibitors against m⁶A regulators for cancer treatment (Deng et al., 2018b). MO-I-500 was developed to selectively inhibit the m⁶A demethylase activity of FTO and was shown to inhibit the survival and/or colony formation of a triple-negative inflammatory breast cancer cell line (Singh et al., 2016; Zheng et al., 2014). Meclofenamic acid (MA), a nonsteroidal anti-inflammatory drug, was demonstrated to specifically inhibit FTO over ALKBH5 and further proved to inhibit GBM cell growth and survival (Cui et al., 2017; Huang et al., 2015). R-2HG, a major metabolic product of mutant IDH1/2, was found to be an inhibitor of FTO that binds directly to FTO and inhibits its m⁶A demethylase activity, leading to an inhibition of leukemic cell growth/survival and leukemia progression (Su et al., 2018). Very recently, a more potent FTO inhibitor FB23-2 was developed, which significantly inhibited AML progression in xeno-transplanted mice (Huang et al., 2019b).

Cancer therapeutic resistance occurs as cancer cells develop resistance to treatments including chemotherapy, radiotherapy, and targeted therapies, and is responsible for treatment failure and/or cancer recurrence. Mechanisms underlying therapy resistance vary, including genetic and/or epigenetic changes in cancer cells, the microenvironment that cancer cells resides, and the existence of a small population of self-renewal, drug resistant cancer stem cells (CSCs). The findings that m⁶A regulators are dysregulated and play important roles in various types of cancer suggest that m⁶A modification could be involved in therapy resistance development. Indeed, it was reported that FTO enhanced chemo-radiotherapy resistance in cervical squamous cell carcinoma through reducing m⁶A abundance in β-catenin mRNA (Zhou et al., 2018), while enhanced m⁶A modification in FZD10 mRNA in BRCA-mutated epithelial ovarian cancers owing to downregulation of FTO and ALKBH5 could result in reduced PARPi sensitivity (Fukumoto et al., 2019). In addition, leukemia cells with FTO upregulation and m⁶A hypomethylation showed increased TKI tolerance and growth advantage due to enhanced mRNA stability of proliferation/survival-related transcripts harboring m⁶A (Yan et al., 2018). Such data suggest that m⁶A RNA modification signatures may serve as predictive markers for personalized cancer treatment and shed light on the overcome of therapy resistance in cancer through modulating the m⁶A-related pathways (Figure 5). Encouragingly, proof-of-concept studies have emerged to support such a notion. In AML, depletion of METTL14 or FTO could sensitize leukemia cells to the differentiation-inducing agent ATRA (Hsu et al., 2017; Weng et al., 2018). By suppressing FTO signaling, R-2HG treatment could sensitize AML cells to a variety of first-line therapeutic agents such as ATRA, 5-azacytidine (5-Aza), decitabine (DAC), daunorubicin (DNR), and cytosine arabinoside (Ara-C), and sensitize GBM cells to temozolomide (TMZ) (Su et al., 2018). In addition, silencing of METTL3, which is elevated in GSC and GBM tumors and predicts poor survival in GBMs, led to enhanced sensitivity of GSCs to γ-irradiation (Visvanathan et al., 2018). Similarly, METTL3-depleted pancreatic cancer cells were more sensitive to anticancer reagents or therapy, including gemcitabine, 5-fluorouracil (5-FU), cisplatin, and irradiation (Taketo et al., 2018).

Figure 5.

Therapeutic implications of m⁶A modification. Considering the roles of FTO, ALKBH5, and METTL3 in glioblastoma stem cells (GSCs) and of METTL14 and FTO in leukemia stem cells (LSCs), inhibitors of m⁶A modifiers holds the potential to be combined with chemotherapy or radiotherapy to erase cancer stem cells (CSCs) and achieve complete remission. In addition, m⁶A regulators play a role in the immune system, and thus are promising targets for enhancing anti-tumor immunity when combined with immune therapies.

m⁶A and cancer immunotherapy

The immune system is a host defense system that protects against infections and diseases. In recent years, studies have shown that RNA m⁶A modification is involved in the development of the immune system and the induction of immune response. A dramatic alteration of lymphocyte homeostasis was observed, where T cells were prevented from homeostatic expansion and kept in a naive state, when Mettl3 was depleted (Li et al., 2017b). Mechanistically, Mettl3 deletion resulted in reduced m⁶A level and increased mRNA stability and protein level of Socs1, Socs3 and Cish, which negatively regulated IL-7 signaling in CD4+ T cells (Li et al., 2017b). Another study revealed the role of m⁶A in dendritic cell activation, in which Mettl3-mediated m⁶A modification on CD40, CD80, and TLR4 signaling adaptor Tirap transcripts enhanced their translation in dendritic cells for strengthening TLR4/NF-κB signaling-induced cytokine production and stimulating T cell activation (Wang et al., 2019a). Recently, two groups independently reported the role of m⁶A modification on the control of innate immune response to dsDNA or viruses. Rubio et al. found that METTL14 depletion, in opposite to ALKBH5 depletion, reduced human cytomegalovirus (HCMV) reproduction in host cells and stimulated dsDNA- or HCMV-induced interferon beta 1 (IFNB1) mRNA production and stabilization (Rubio et al., 2018). Similarly, Winkler et al. reported that deletion of METTL3 or YTHDF2 following viral infection stabilized IFNB1 in an m⁶A dependent manner, resulting in the subsequent suppression of viruses propagation in the host cells (Winkler et al., 2019). Interestingly, endogenous human circRNAs were m⁶A-modified and could be recognized by YTHDF2, which sequestered m⁶A-circRNA and was essential for suppression of innate immunity, while foreign circRNAs were unmodified and could induce antigen-specific T cell activation, antibody production, and anti-tumor immunity (Chen et al., 2019c).

Immunotherapy is a new strategy of cancer therapy that stimulates and improves the immune system's natural ability to fight cancer cells. YTHDF1 was recently found to control anti-tumor immunity and improve immunotherapy through regulating expression of lysosomal proteases in an m⁶A dependent manner (Han et al., 2019a). YTHDF1 could recognize m⁶A-marked transcripts encoding lysosomal proteases to increase their translation in dendritic cells, and loss of Ythdf1 enhanced the cross-presentation of tumor antigens and the cross-priming of CD8+ T cells in vivo. Moreover, Ythdf1 deletion enhanced the therapeutic efficacy of PD-L1 checkpoint blockade (Han et al., 2019a). In melanoma, increased level of FTO promoted tumor growth by reducing m⁶A methylation in PD-1 (PDCD1), CXCR4, and SOX10 and preventing their RNA decay mediated by YTHDF2. Knockdown of FTO in melanoma cells sensitized tumor cells to interferon gamma (IFNγ) in vitro and promoted melanoma response to anti-PD-1 antibody in mice (Yang et al., 2019c). These data together implicate that modulation of m⁶A regulators could be combined with anti-PD-1/PD-L1 blockade to improve anticancer immunotherapy (Figure 5).

Conclusions and Perspectives

Within the past few years, extensive efforts have been devoted to study m⁶A modification in cancers, leading to extensive accumulation of experimental data and increased knowledge on the roles of m⁶A modification and the associated machinery in various types of cancers. It is now clear that both global m⁶A levels and expression of m⁶A regulators (writers, erasers, and readers) are dysregulated in various types of cancers, and their dysregulation could be associated with drug resistance and prognosis in cancer patients. Evidence is emerging that m⁶A regulators play important pathological (more often tumor-promoting) roles in various types of cancers by modulating the epitranscriptome in cancers. Notably, in certain types of cancers such as AML, breast cancer, lung cancer and gastric cancer, both m⁶A writer and eraser genes are aberrantly overexpressed and play important oncogenic roles. While the underlying molecular mechanisms have yet to be fully understood, such phenomena suggest that dysregulation of the epitranscriptome by aberrantly expressed either writers or erasers could cause similar phenotypes (Deng et al., 2018c). Since both m⁶A writers and erasers are overexpressed in such cancers, the clinical implications of the global m⁶A abundance may not be very informative. Instead, m⁶A profiles (or signatures) of some particular transcripts or transcript loci could be better biomarkers for cancer diagnosis, classification and prognosis. However, currently available transcriptome-wide m⁶A-seq methods often need a large amount of RNA material (e.g., >20 μg total RNA), thus it is difficult to conduct m⁶A-seq for a large cohort of patients due to the limitation in primary patient samples. Therefore, much improved m⁶A-seq technologies that need much less RNA material and provide base-resolution m⁶A profiles are urgently needed. With such improved m⁶A-seq technologies, we can conduct m⁶A profiling with precious primary patient material or even with limited primary CSCs or TICs; the m⁶A profiles/signatures of certain particular transcripts or transcript loci could be identified as biomarkers for early cancer diagnosis, cancer classification, outcome prediction, and risk stratification. Besides cancer cells, cell-free circulating RNA (cfRNA) collected from plasma of cancer patients could also be used for m⁶A-seq, and particular m⁶A signatures could be used as biomarkers in clinic application. Thus far, more efforts have been focusing on identification of mRNA targets of m⁶A modification in cancers, and m⁶A modification on ncRNAs has been less studied and thus represents one of the future directions. Studying how m⁶A modification affects the production, cellular location, function of ncRNAs and how these processes are linked to cancer will greatly improve our understanding of the cellular function of m⁶A as well as the currently unappreciated functions of ncRNAs. More recently, m⁶A modifications were found on chromosome-associated regulatory RNAs (carRNAs), including promoter-associated RNAs, enhancer RNAs, and repeats RNAs, and were shown to regulate chromatin state and transcription (Liu et al., 2020a). Whether these carRNA species play a role in cancers remains an interesting topic that warrants further studies.

Given the functional importance of m⁶A modification and regulators in various types of cancers, targeting dysregulated m⁶A regulators represents an attractive strategy for cancer therapy. Indeed, some proof-of-concept studies already suggest that targeting dysregulated m⁶A regulators by small-molecule inhibitors holds therapeutic potential for cancer treatment. A few small pharmaceutic or biotech companies (e.g., STORM Therapeutics, Accent Therapeutics, Gotham Therapeutics and Genovel Biotech Corp.) have started to develop highly potent and selective small-molecule inhibitors targeting m⁶A regulators such as METTL3, METTL14, and FTO. Besides small-molecule compounds that target m⁶A regulators directly, PROTAC (proteolysis targeting chimera)-based inhibitors could also be developed to selectively degrade dysregulated m⁶A regulatory proteins for cancer therapy. In addition, because m⁶A modification also plays important roles in mediating responses of cancers to chemotherapy, radiotherapy, and immunotherapy, targeted treatment against m⁶A regulators can also be applied in the clinic together with chemotherapy, radiotherapy, or immunotherapy to achieve much improved cancer therapy in the near future. In addition, similar to genome editing, epitranscriptome editing can also be developed to restore or remove functional essential m⁶A sites that are mutated or dysregulated in cancers, and such editing may also be applicable in the clinic for cancer therapy in the future. Overall, study of m⁶A modification in cancer is a new frontier in cancer research, which not only reveals a new layer of epigenetic regulation in cancer and thereby provide novel insights into the molecular mechanisms underlying tumorigenesis, immune response, and drug resistance, but will also lead to the development of effective novel therapeutics. Targeting dysregulated m⁶A regulators by effective inhibitors (or targeting mutated or dysfunctional m⁶A sites by targeted epitranscriptome editing) alone or in combination with other therapeutics likely holds potent therapeutic potential for the treatment of various types of cancers, especially those resistant to currently available treatments.

Abbreviates:

m⁶A

N⁶-methyladenosine

mRNA

messenger RNA

ncRNA

non-coding RNA

A

adenosine

G

guanosine

C

cytidine

U

uridine

NGS

next generation sequencing

CDS

coding sequence

3’ UTR

3’ untranslated region

MeRIP-seq

methylated RNA immunoprecipitation sequencing

miCLIP

m⁶A individual-nucleotide-resolution cross-linking and immunoprecipitation

m⁶A-REF-seq

m⁶A-sensitive RNA-endoribonuclease–facilitated sequencing

DART-seq

deamination adjacent to RNA modification targets sequencing

rRNA

ribosomal RNA

lncRNA

long non-coding RNA

miRNA

microRNA

snRNA

small nulear RNA

circRNA

circular RNA

MTC

methyltransferase complex

SAM

S-adenosyl methionine

H3K36me3

histone H3 lysine 36 trimethylation

αKG

alpha-ketoglutarate

m⁶A_m

N⁶,2'-O-dimethyladenosine

AML

acute myeloid leukemia

pre-mRNA

precursor mRNA

mRNP

messenger ribonucleoprotein

P-body

processing body

YTH

YT521-B homology

eIF

eukaryotic translation initiation factor

nt

nucleotides

SCARLET

site-specific cleavage and radioactive-labeling followed by ligation-assisted extraction and thin-layer chromatography

pri-miRNAs

the primary transcripts of miRNAs

pre-miRNAs

precursor miRNAs

TKI

tyrosine kinase inhibitor

HCC

hepatocellular carcinoma

GC

gastric cancer

CRC

colorectal cancer

NSCLC

non-small cell lung cancer

BLC

bladder cancer

EMT

epithelial-mesenchymal transition

GBM

glioblastoma

MTD

methyltransferase domain

LSC

leukemia stem cells

LIC

leukemia initiating cells

HSC

hematopoietic stem cells

TIC

tumor initiating cells

ATRA

all-trans-retinoic acid

R-2HG

R-2-hydroxyglutarate

BRC

breast cancer

MLM

melanoma

GSC

glioblastoma stem cell

BCSC

breast cancer stem cell

HBE

human bronchial epithelial

Cd

cadmium

Ni

nickel

PDAC

pancreatic ductal adenocarcinoma

EBV

Epstein–Barr virus

MA

meclofenamic acid

CSC

cancer stem cell

5-Aza

5-azacytidine

DAC

decitabine

DNR

daunorubicin

Ara-C

cytosine arabinoside

TMZ

temozolomide

5-FU

5-fluorouracil

HCMV

human cytomegalovirus

PROTAC

proteolysis targeting chimera