Ghost authors revealed: The structure and function of human N⁶-methyladenosine RNA methyltransferases

Abstract

Despite the discovery of modified nucleic acids nearly 75 years ago, their biological functions are still being elucidated. N⁶-methyladenosine (m⁶A) is the most abundant modification in eukaryotic messenger RNA (mRNA) and has also been detected in non-coding RNAs, including long non-coding RNA, ribosomal RNA, and small nuclear RNA. In general, m⁶A marks can alter RNA secondary structure and initiate unique RNA–protein interactions that can alter splicing, mRNA turnover, and translation, just to name a few. Although m⁶A marks in human RNAs have been known to exist since 1974, the structures and functions of methyltransferases responsible for writing m⁶A marks have been established only recently. Thus far, there are four confirmed human methyltransferases that catalyze the transfer of a methyl group from S-adenosylmethionine (SAM) to the N⁶ position of adenosine, producing m⁶A: methyltransferase-like protein (METTL) 3/METTL14 complex, METTL16, METTL5, and zinc-finger CCHC-domain-containing protein 4. Though the methyltransferases have unique RNA targets, all human m⁶A RNA methyltransferases contain a Rossmann fold with a conserved SAM-binding pocket, suggesting that they utilize a similar catalytic mechanism for methyl transfer. For each of the human m⁶A RNA methyltransferases, we present the biological functions and links to human disease, RNA targets, catalytic and kinetic mechanisms, and macromolecular structures. We also discuss m⁶A marks in human viruses and parasites, assigning m⁶A marks in the transcriptome to specific methyltransferases, small molecules targeting m⁶A methyltransferases, and the enzymes responsible for hypermodified m⁶A marks and their biological functions in humans. Understanding m⁶A methyltransferases is a critical steppingstone toward establishing the m⁶A epitranscriptome and more broadly the RNome.

This article is categorized under:

RNA Interactions with Proteins and Other Molecules > Protein-RNA Recognition

RNA Interactions with Proteins and Other Molecules > RNA-Protein Complexes

RNA Interactions with Proteins and Other Molecules > Protein-RNA Interactions: Functional Implications

1 ∣. INTRODUCTION

About 75 years ago, the first nucleic acid modifications were reported when 5-methylcytidine (5mC) was discovered in DNA (Hotchkiss, 1948) followed by pseudouridine in RNA (Cohn & Volkin, 1951). Since the discovery of the first naturally occurring nucleic acid modifications, 58 DNA (Sood et al., 2019) and 143 RNA (Boccaletto et al., 2018; McCown et al., 2020) modifications across all domains of life have been discovered to date. The simplest and most abundant modification in nucleic acids is a methyl group (—CH₃), whose transfer from a methyl donor cofactor is catalyzed by a methyltransferase (MTase) or “writer” protein. The most common methyl donor is S-adenosylmethionine (SAM or AdoMet), which is the second most abundant metabolite in human cells after ATP (Cantoni, 1977; J. Gao et al., 2018; German et al., 1983). A single methylation of adenosine at the N⁶ position produces N⁶-methyladenosine (6mA) in DNA and (m⁶A) in RNA. 6mA is the most common DNA modification in prokaryotes, but there is a lack of evidence for 6mA being a bona fide human DNA modification and for the presence of a human 6mA MTase. Though some of the human m⁶A RNA MTases can methylate DNA in vitro (L. Q. Chen et al., 2022; Qi et al., 2022; Woodcock et al., 2019; Yu, Horton, et al., 2021), 6mA in mammals is currently thought to originate from misincorporation of recycled m⁶A that was generated by nucleotide-salvage pathway (X. Li, Zhang, et al., 2021; Musheev et al., 2020), a misidentification from a contamination such as bacterial plasmid (Kong et al., 2022), or due to antibody cross reacting with unmethylated adenosine (Douvlataniotis et al., 2020). Therefore, we focus this review on human m⁶A and the human m⁶A RNA MTases. We refer you to the following reviews to learn about 6mA in other organisms (O'Brown & Greer, 2016; C. Shen et al., 2022; K. J. Wu, 2020).

The m⁶A modification was first observed in eukaryotic messenger RNA (mRNA) nearly 50 years ago (Desrosiers et al., 1974) and, a few years later, shown to be present within a specific consensus motif that has since been verified to be DRACH (where D = A/G/U, R = A/G, H = A/C/U, underlined A denotes methylated A here and henceforth; Csepany et al., 1990; Dominissini et al., 2012; Harper et al., 1990; Linder et al., 2015; Meyer et al., 2012; Narayan & Rottman, 1988; Wei & Moss, 1977). For human mRNA, approximately 0.5% of all mRNA adenosines are m⁶A, which is ~3 m⁶A marks per transcript, and about half of the m⁶A marks are detected in multiple cell types (Dominissini et al., 2012, 2016; Körtel et al., 2021; Lavi et al., 1977; Meyer et al., 2012; Wei et al., 1975; Wei & Moss, 1977). Though most m⁶A marks fall within the DRACH motif, a notable fraction (~1–6%) are found elsewhere (Körtel et al., 2021; C. Liu, Sun, et al., 2022). It was not until the 1990s that the MTase responsible for the mRNA m⁶A modification was discovered to be MT-A70 (Bokar et al., 1997; Rottman et al., 1994), now known as the methyltransferase-like protein (METTL) 3/METTL14 complex. The expansion of methods for RNA sequencing and for detecting m⁶A led to the discovery of m⁶A throughout all domains of life. In addition to mRNA, m⁶A has been found in most human noncoding RNA (ncRNA) classes, including circular RNA (Y. Yang et al., 2017), chromatin-associated RNAs such as enhancer RNA and promoter upstream transcripts (J. Liu et al., 2020; Louloupi et al., 2018; Xiao et al., 2019; W. Xu et al., 2021; W. Xu, He, et al., 2022), long noncoding RNA (lncRNA; Dominissini et al., 2012; Meyer et al., 2012), micro RNA (miRNA; Alarcón et al., 2015), both the small and large subunits of ribosomal RNA (rRNA) (Taoka et al., 2018), small nuclear RNA (snRNA; Epstein et al., 1980; Harada & Kato, 1980), and small nucleolar RNA (Linder et al., 2015). Only within the past 6 years have more human m⁶A RNA MTases been confirmed: METTL16, which to date has been confirmed to methylate the methionine adenosyltransferase 2A (MAT2A) mRNA and U6 snRNA (Pendleton et al., 2017); METTL5, which methylates A1832 in 18S rRNA (Ignatova et al., 2020; Rong et al., 2020; Van Tran et al., 2019; Xing et al., 2020), and zinc-finger CCHC-domain-containing protein 4 (ZCCHC4), which methylates A4220 in 28S rRNA (H. Ma et al., 2019; Pinto et al., 2020; Ren et al., 2019; Van Tran et al., 2019; Figure 1). In addition to m⁶A MTases, there are also m⁶A demethylases (or “eraser” proteins) and m⁶A-binding proteins (or “reader” proteins). Human m⁶A demethylases include the fat mass and obesity-associated protein (FTO; Jia et al., 2011) and α-ketoglutarate-dependent dioxygenase alkB homolog 5, RNA demethylase (ALKBH5) (Zheng et al., 2013). However, it is important to note that the prevalence of m⁶A demethylation is debatable (Darnell et al., 2018; Mauer & Jaffrey, 2018; Murakami & Jaffrey, 2022; B. S. Zhao et al., 2018). Of the known human m⁶A-binding proteins, five contain a YT521-B homology (YTH) domain (YTHDC1,2 and YTHDF1,2,3), though the domain is not necessary for binding m⁶A, as other proteins, including the eukaryotic translation initiation factor 3 subunit A (eIF3A), bind to m⁶A (Y. Li, Bedi, et al., 2020; Meyer et al., 2015; C. Xu et al., 2014, 2015). Therefore, the m⁶A modification facilitates dynamic biological outcomes, with MTases writing the mark and binding proteins reading the mark. Though m⁶A exists in most RNA classes and throughout all domains of life, this review focuses on the structure and function of human m⁶A RNA MTases (summarized in Table 1). We also highlight m⁶A MTases in viruses and parasites, methods for detecting m⁶A in an MTase-specific manner, small molecules that target m⁶A RNA MTases, hypermodified m⁶A marks and the enzymes that catalyze them, and future directions of m⁶A MTase research.

FIGURE 1.

Schematics of known RNA substrates and targets of human m⁶A methyltransferases. RNA probes for METTL3/METTL14 were designed and used by J. Liu et al. (2014). MAT2A mRNA hairpins 1–6 are shown in their predicted “duckbill” structure, along with the structure of hairpin 1 observed when bound to METTL16 (PDB ID: 6DU4). Hairpins are arranged 5′ to 3′, whereby the dashed line represents missing intervening nucleotides between each hairpin. U6 snRNA is shown as the predicted structure along with a proposed METTL16-bound structure that mimics the MAT2A hairpin bound structure at the site of nonamer sequence (Aoyama et al., 2020; Doxtader et al., 2018). The MALAT1 triple helix structure is based on the solved crystal structure (PDB ID: 4PLX). 18S and 28S rRNAs are presented as the structure within the assembled human 80S ribosome (PDB ID: 4UG0). For all RNA secondary structures, a dash (—) represents a canonical Watson–Crick base pair, while a dot (•) represents any non-canonical nucleotide interaction. Red nucleotides show consensus motifs with the modified A in blue. C_m in 18S rRNA and U_m in 28S rRNA denote a nucleotide with a 2′-O-methyl modification. MALAT1, metastasis-associated lung adenocarcinoma transcript 1; MAT2A, methionine adenosyl transferase 2A; METTL3, methyltransferase-like protein 3; METTL14, methyltransferase-like protein 14; mRNA, messenger RNA; PDB ID, Protein Data Bank identifier.

TABLE 1.

Summary of features for each human m⁶A RNA methyltransferase.

	METTL3	METTL14	METTL16	METTL5	ZCCHC4
Phylogenetic conservation of MTase domain (C. Liu, Cao, et al., 2022)	Eukaryotes	Eukaryotes	Prokaryotes and eukaryotes	Eukaryotes	Multicellular eukaryotes
RNA targets for methylation	mRNA lncRNA miRNA	None, catalytically inactive	MAT2A mRNA U6 snRNA (A43)	18S rRNA (A1832)	28S rRNA (A4220, also known as A4190)
Motif	DRACH	DRACH	ACAGAR in stem loop or bulge	UAACA	UAACG in loop of stem-loop
Protein–protein interactions	Stable complex with HAKAI, KIAA1429/VIRMA, METTL14, RBM15, WTAP, and ZC3H13	Stable complex with HAKAI, KIAA1429/VIRMA, METTL3, RBM15, WTAP, and ZC3H13	Transient binding partners, for example, eIF3A/B, eIF4E2, La, LARP7, MePCE, and those associated with DNA replication and repair, 7SK RNP, spliceosome, rRNA processing and translation	Stable complex with TRMT112; transiently binds eIFs and eEFs (Rong et al., 2020)	Transient binding partners, for example, nucleolar-localized proteins and those involved in ribosome/RNA metabolism (Pinto et al., 2020)
Subcellular localization	Nuclear speckles, nucleus; cytoplasm for some cancers	Nuclear speckles, nucleus	Nucleolus, nucleus, cytoplasm	Nucleolus (enriched), nucleus, cytoplasm	Nucleolus (enriched), nucleus, cytoplasm
KO mice	Lethal	Lethal	Lethal	Subviable	Unknown

Note: Unless indicated in this table, please refer to Section 2 for references.

Abbreviations: eEF, eukaryotic translation elongation factor; eIF3A/B, eukaryotic translation initiation factor 3 subunit A/B; HAKAI, E3 ubiquitin-protein ligase Hakai; LARP7, La ribonucleoprotein 7, transcriptional regulator; lncRNA, long noncoding RNA; m⁶A, N⁶-methyladenosine; MePCE, methylphosphate capping enzyme; METTL3, methyltransferase-like protein 3; METTL5, methyltransferase-like protein 5; METTL14, methyltransferase-like protein 14; METTL16, methyltransferase-like protein 16; miRNA, micro RNA; mRNA, messenger RNA; RBM15, RNA-binding motif protein 15; RNP, ribonucleoprotein; VIRMA/KIAA1429, vir-like m⁶A methyltransferase associated protein; WTAP, Wilms' tumor 1-associating protein; ZCCHC4, zinc-finger CCHC-domain-containing protein 4; ZC3H13, zinc-finger CCCH domain-containing protein 13.

2 ∣. BIOLOGICAL FUNCTIONS OF HUMAN m⁶A RNA MTases

2.1 ∣. Methyltransferase-like proteins 3/14

METTL3 is a roughly 65-kDa MTase initially isolated in 1994 from HeLa nuclei (Rottman et al., 1994). A comparative sequence analysis suggested that the MTase domain of human METTL3 is conserved within over 40 proteins, including homologs in organisms such as Saccharomyces cerevisiae, Arabidopsis thaliana, and Drosophila melanogaster but not in bacteria (see Table 1; Bujnicki et al., 2002). Human METTL14 was also identified as a possible homolog of METTL3 (Bujnicki et al., 2002). METTL3 and METTL14 were originally thought to be independent, catalytically active m⁶A MTases that, together, have increased activity (J. Liu et al., 2014). However, it is now understood that METTL3 and METTL14 form a stable heterodimer and only METTL3 is the catalytically active MTase, as METTL14 has a degenerate active site motif of EPPL (see Sections 7.4 and 7.5; Śledź & Jinek, 2016; P. Wang, Doxtader, & Nam, 2016; X. Wang, Feng, et al., 2016; Y. Wang, Li, et al., 2014). Instead of a catalytic function, METTL14 enhances METTL3 methylation capacity by facilitating RNA binding (Śledź & Jinek, 2016; P. Wang, Feng, et al., 2016; X. Wang, Feng, et al., 2016). Though the METTL3/METTL14 heterodimer is catalytically active in a test tube, in vivo it is part of a ~1000-kDa holocomplex that includes other protein-binding partners such as Wilms' tumor 1-associating protein (WTAP), vir-like m⁶A methyltransferase-associated protein (VIRMA or KIAA1429), RNA-binding motif protein 15 (RBM15), the E3 ubiquitin-protein ligase Cbl proto-oncogene like 1 (CBLL1) more commonly referred to as E3 ubiquitin-protein ligase Hakai (HAKAI, named after “destruction” in Japanese), and zinc-finger CCCH domain-containing protein 13 (ZC3H13; J. Liu et al., 2014, 2018; Patil et al., 2016; Wen et al., 2018). WTAP is a notable adaptor protein because it is required for maintaining m⁶A abundance (Ping et al., 2014), and it mediates localization of METTL3/METTL14 to nuclear speckles in HeLa cells (Bokar et al., 1997; Ping et al., 2014). However, METTL3 and METTL14 have also been detected in the nucleoplasm and cytoplasm depending on cell type, cellular conditions (stress, cancer, etc.), stoichiometry of components in the holocomplex, and possibly some post-translational modifications (Arcidiacono et al., 2020; S. Lin et al., 2016; Schöller et al., 2018). METTL3 has 10 phosphorylation sites, most of which occur outside of the MTase domain (Matsuoka et al., 2007; Schöller et al., 2018; H. L. Sun et al., 2020). METTL3 phosphorylation is not crucial for MTase activity, interactions with WTAP, or subcellular localization (Schöller et al., 2018), however, phosphorylation of METTL3 by the extracellular signal-regulated kinases pathway allows for easier deubiquitination to stabilize the METTL3/METTL14 complex in mouse embryonic stem cells (mESCs; H. L. Sun et al., 2020). METTL14 appears to have only one phosphorylation site (i.e., S399), whose function is unclear because it is neither important for binding to METTL3 nor MTase activity (Schöller et al., 2018).

With METTL3/METTL14 residing predominantly in the nucleus, this complex methylates nascent pre-mRNA and other RNA polymerase II-transcribed RNAs in a co-transcriptional manner in HeLa cells (Ke et al., 2017). Most m⁶A marks deposited by METTL3/METTL14 are detected within the DRACH consensus motif in either single- or double-stranded RNA (N. Liu et al., 2013; Meiser et al., 2020; X. Wang, Feng, et al., 2016; Figure 1) and are enriched near stop codons and in 3′ untranslated regions (UTRs; Dominissini et al., 2012; Ke et al., 2017; Meyer et al., 2012) due to the exon junction complex, namely eIF4A3, blocking METTL3 from binding to sites in the mRNA coding sequence (X. Yang et al., 2022). The METTL3/METTL14 complex methylates over 15,000 sites in ~25–60% of mRNAs (human embryonic kidney 293 cells with mutant Simian virus 40 large T antigen [HEK293T]) (Dominissini et al., 2012; He & He, 2021; Koh et al., 2018). Only ç5% of DRACH sites are methylated (Dominissini et al., 2012; He & He, 2021). Although the mechanism of site selection is not well understood, it appears that RBM15 and RBM15B bind to U-rich regions flanking DRACH sites, which subsequently recruits the METTL3/METTL14/WTAP complex (Patil et al., 2016). The plethora of m⁶A marks raises multiple questions: What is the function of each m⁶A mark? What percent of each site is methylated? Do all m⁶A sites on an entire population of a particular transcript need to be there to elicit a biological function? Although context-dependent answers are emerging for select m⁶A sites, mRNA metabolism may be altered at various steps, such as alternative splicing (Dominissini et al., 2012; Zhou et al., 2019), polyadenylation (J. Liu et al., 2018), nuclear export (Roundtree et al., 2017), decreased mRNA stability/half-life (H. Du, Zhao, et al., 2016; X. Wang, Lu, et al., 2014), and translation (S. Lin et al., 2016). In another study of METTL3 function, alternative splicing events in human liver cancer cell line (HepG2) cells are disrupted when METTL3 is silenced (Dominissini et al., 2012). The METTL3/METTL14 complex can also engage with other cellular machinery, such as the translation initiation factor elF3 to promote eIF4E-independent translation of mRNAs containing m⁶A in 5′-UTR (Choe et al., 2018; S. Lin et al., 2016; Meyer et al., 2015), the transcription factor CCAAT enhancer-binding protein zeta to aid translation (Barbieri et al., 2017), chromatin to silence retroviral elements in mammalian cells (W. Xu et al., 2021), or for nascent RNA to escape premature termination by the Integrator complex (W. Xu, He, et al., 2022). Further regulation is conferred by “reader” proteins that recognize m⁶A sites, including those that depend upon a “structural switch” (N. Liu et al., 2015, 2017) or facilitate phase separation (Y. Fu & Zhuang, 2020; Y. Gao et al., 2019; D. Han, Longhini, et al., 2022; Lee et al., 2021; Ries et al., 2019).

In mice, the Mettl3 and Mettl14 genes are both essential, as Mettl3^−/− knockout mice and Mettl14^−/− knockout mice are embryonic lethal (Geula et al., 2015; Meng et al., 2019; Y. Wang, Li, et al., 2018). Conditional knockout mice and knockout cell lines are viable, although numerous stem cells lacking Mettl3 or Mettl14 exhibit impaired differentiation (Batista et al., 2014; Geula et al., 2015; Meng et al., 2019; Y. Wang, Li, et al., 2018). However, a recent study reports that these so-called “knockouts” may simply be alternatively spliced variants of METTL3 that are still capable of methylation (Poh et al., 2022). Knockout of Mettl3 in mouse cells reduces m⁶A levels on mRNA by ~70–100% (Geula et al., 2015; Z. Lin et al., 2017; C. X. Wang, Cui, et al., 2018). The loss of Mettl3 and/or Mettl14 impacts a multitude of physiological processes, such as brain development (K. Du et al., 2021; Flamand & Meyer, 2019; Madugalle et al., 2020; C. X. Wang, Cui, et al., 2018; Yoon et al., 2017), the cardiovascular system (Dorn et al., 2019; Yao et al., 2020), immunity (W. Qiu, Zhang, et al., 2021; Tong et al., 2021), memory (Z. Zhang et al., 2018), spermatogenesis (Z. Lin et al., 2017), and ultraviolet-induced DNA damage response (Xiang et al., 2017). Thus, it is not surprising that the METTL3/METTL14 complex and global m⁶A levels are implicated in human health and disease. Reduced expression of METTL3 and/or METTL14 contribute to Type 2 diabetes (De Jesus et al., 2019), obesity (Y. Wang, Gao, et al., 2020), premature aging (Z. Wu et al., 2020), features of osteoporosis (Y. Wu et al., 2018), and select cancer types such as renal cell carcinoma (X. Li et al., 2017; Y. Zhang, Yao, Qi, et al., 2021). In contrast, the upregulation of METTL3 and/or METTL14 contribute to acute myeloid leukemia (AML; Barbieri et al., 2017; Sorci et al., 2018; Vu et al., 2017; Weng et al., 2018), lung cancer (Choe et al., 2018; S. Lin et al., 2016; Wanna-udom et al., 2020), and pancreatic cancer (M. Wang, Liu, et al., 2020; T. Xia et al., 2019), just to name a few. There is some ambiguity with functions of METTL3/METTL14 in cancers. For example, in human hepatocellular carcinoma, METTL3 upregulation promotes cancer proliferation, tumorigenicity, and metastasis (M. Chen et al., 2018) whereas METTL14 downregulation promotes cancer metastasis and invasion (J. Z. Ma et al., 2017). Likewise, conflicting results regarding whether METTL3 and METTL14 facilitate or suppress tumor growth in glioblastoma stem cells have been reported (Q. Cui et al., 2017; X. Jiang, Liu, et al., 2021; Visvanathan et al., 2018). The implications of METTL3 and METTL14 in various cancers are summarized in recent reviews (L. J. Deng, Deng, et al., 2022; Guan et al., 2022; C. Zeng et al., 2020). Depending upon the disease, both activation and inhibition of the METTL3/METTL14 complex may prove fruitful as possible therapeutics (see Section 6).

2.2 ∣. Methyltransferase-like protein 16

Compared to METTL3/METTL14, much less is known about the biological function of METTL16 since it was 2017 when METTL16 was first shown to have m⁶A MTase activity (Pendleton et al., 2017). There are various homologs of human METTL16 such as 23S rRNA m⁶A1618 methyltransferase (RlmF; Sergiev et al., 2008) in Escherichia coli, METT-10 in Caenorhabditis elegans (Dorsett & Schedl, 2009), and FIONA1 in A. thaliana (see Table 1; J. Kim et al., 2008). The cellular localization of endogenous METTL16 is cell-type and cell-cycle dependent for the protein has been detected in the nucleolus, nucleus, and cytoplasm of various human cell lines (Brown et al., 2016; Nance et al., 2020; Sjöstedt et al., 2020; P. Wang, Wang, et al., 2020; Weng et al., 2018). Currently, there are seven RNA substrates of human METTL16 that have been validated in vitro: the U6 snRNA and each of the six hairpins (hp1–6) in the 3′-UTR of MAT2A mRNA (Mendel et al., 2018; Pendleton et al., 2017; Shima et al., 2017; Warda et al., 2017). METTL16 preferentially methylates adenosines near structured RNA regions, such as the nonamer sequence UACAGARAA (R = A or G) present in both U6 snRNA and the MAT2A hp1–6 (Doxtader et al., 2018; Mendel et al., 2018; Pendleton et al., 2017; Figure 1). In mice, two additional mRNA substrates of METTL16 were identified: Fanconi anemia complementation group M and breast cancer 2 protein or BRCA2 (Yoshinaga et al., 2022). For human U6 snRNA, m⁶A43 is part of the ACAGAGA box that base pairs with the 5′-splice sites of pre-mRNAs during splicing (Wassarman & Steitz, 1992). Indeed, the m⁶A mark in both Schizosaccharomyces pombe and Arabidopsis thaliana U6 snRNA presumably forms a stable m⁶A•A₊₄ base pair with the intron to enable efficient splicing regardless of the 5′ exon sequence (Ishigami et al., 2021; Parker et al., 2022). METTL16 indirectly controls SAM concentrations because METTL16 regulates MAT2A protein levels via methylation of the MAT2A mRNA hp1 (Doxtader et al., 2018; Pendleton et al., 2017; Shima et al., 2017). METTL16 will remain on hp1 if SAM levels are low, allowing for proper splicing of the MAT2A mRNA and subsequently protein synthesis. However, if SAM levels are high, then METTL16 will methylate and dissociate from hp1, leading to intron-detained MAT2A mRNA and less MAT2A protein (Pendleton et al., 2017). This mechanism involves the cleavage factor I_m (CFI_m) complex; specifically a CFI_m25 dimer recruits CFI_m59 and CFI_m68 to mediate splicing of MAT2A while METTL16 acts as an upstream SAM sensor (Scarborough et al., 2021). However, the interplay between the CFI_m complex and METTL16 requires further elucidation. A similar mode of regulation exists in C. elegans, where METT-10, the ortholog of human METTL16, controls the splicing and subsequently abundance of mRNAs for SAM-producing enzymes (Mendel et al., 2021; Watabe et al., 2021). In this case, high levels of SAM lead to methylation of a 3′ splice site (AG) that interferes with splicing factor U2AF35 recognition of the splice site (Mendel et al., 2021; Watabe et al., 2021).

METTL16 has an RNA interactome that is over 4000 mRNAs and also binds to various ncRNA classes (e.g., lncRNA, miRNA, and snRNA) based on a crosslinking and analysis of complementary DNA assay (R. Su, Dong, et al., 2022; Warda et al., 2017). METTL16 binds to the triple helix at the 3′-end of mature metastasis-associated lung adenocarcinoma transcript 1 (MALAT1; Brown et al., 2016; Warda et al., 2017). Although m⁶A signals have been identified via m⁶A-seq analysis (methylation [or m⁶A]-individual-nucleotide resolution crosslinking and immunoprecipitation [miCLIP]; Linder et al., 2015) and nanopore direct RNA sequencing (Krusnauskas et al., 2023), a recent study suggests that METTL16 does not methylate the MALAT1 triple helix (Figure 1) under in vitro conditions (Breger & Brown, 2023). Thus, the methylome of METTL16 is smaller than its interactome. The m⁶A-crosslinking-exonuclease-sequencing (m⁶ACE-seq) method (described later in Section 4) identified 136 METTL16-dependent m⁶A sites (Koh et al., 2018) while methylated RNA immunoprecipitation sequencing (MeRIP-seq) identified 334 mRNAs with METTL16-dependent m⁶A sites (R. Su, Dong, et al., 2022).

Unlike the METTL3/METTL14 heterodimer, METTL16 engages in dynamic protein–protein interactions (Ignatova et al., 2019). For the methylation of U6 snRNA, METTL16 interacts with the proteins La, La ribonucleoprotein 7, transcriptional regulator (LARP7), and methylphosphate capping enzyme (MePCE) in an RNA-dependent manner (Covelo-Molares et al., 2021; Warda et al., 2017). Other notable proteins that interact with METTL16 include those associated with DNA replication and repair, 7SK ribonucleoprotein (RNP), spliceosome, rRNA processing, and translation (Covelo-Molares et al., 2021; Stixová et al., 2021; R. Su, Dong, et al., 2022; Warda et al., 2017). Interestingly, cytosolic METTL16 enhances translation efficiency in a methylation-independent manner, likely via interactions between the MTase domain of METTL16 and eIF3A/B and/or eIF4E2 proteins that are critical for the formation of 80S ribosomes (R. Su, Dong, et al., 2022; F. Wang et al., 2023). METTL16 was also shown to bind certain RNAs to associate with double-strand break repair protein MRE11 (MRE11) in an RNA-dependent manner, preventing MRE11-dependent DNA-end resection (X. Zeng et al., 2022).

With roles in SAM homeostasis, differentiation, and DNA-damage response, METTL16 is an essential protein in normal human cells (T. Wang et al., 2015) and genetic knockout of Mettl16 in mice results in embryonic lethality (Mendel et al., 2018; Yoshinaga et al., 2022). Recent analyses suggest that METTL16 may play a role in human health and disease because (i) METTL16 is upregulated in multiple cancers including AML (L. Han et al., 2023), breast cancer (F. Ye et al., 2023; B. Zhang et al., 2020), colon cancer (S. Wang, Fan, et al., 2021), esophageal cancer (H. Zhao, Xu, et al., 2021), hepatocellular carcinoma (R. Su, Dong, et al., 2022), osteosarcoma (Cheng et al., 2023), and stomach cancer (X. K. Wang, Zhang, et al., 2021), and (ii) METTL16 is downregulated in liver cancer (W. Wang, Sun, et al., 2020), patients with hepatocellular carcinoma (P. Wang, Wang, et al., 2020), cases of infertility (S. Zhao, Lu, et al., 2021), and instances involving the absence of gut microbiota (Jabs et al., 2020). Interestingly, a recent study found that METTL16 is the most crucial METTL protein for cancer survival out of 25 METTLs and that knockout of METTL16 in hepatocellular carcinoma, where METTL16 is overly abundant, led to the suppression of tumorigenesis (R. Su, Dong, et al., 2022). In some cases, METTL16 appears to mediate cancer independent of its “writer” role because METTL16 acts as a translational regulator via interactions with eIF3A/B, eIF4E2 and 18S rRNA (R. Su, Dong, et al., 2022; F. Wang et al., 2023). Future studies into METTL16 will likely focus on the function of METTL16 as a SAM sensor and its possible interactions with the splicing complex as well as elucidating whether the m⁶A within U6 snRNA aids base pairing for splicing in humans akin to S. pombe and A. thaliana.

2.3 ∣. A pair of m⁶A rRNA MTases: METTL5 produces m⁶A1832 in 18S rRNA and ZCCHC4 produces m⁶A4220 in 28S rRNA

In human lymphoblast cell line (TK6) cells, 228 RNA modifications have been detected in human rRNA but only two of those modifications are m⁶A: m⁶A1832 in 18S rRNA and m⁶A4220 (also known as m⁶A4190 using alternative numbering system) in 28S rRNA (Taoka et al., 2018) (Figure 1). Although both of these m⁶A marks were discovered over 30 years ago (Maden, 1986, 1988), it is only recently that the writers of these marks have been identified: METTL5 methylates A1832 and ZCCHC4 methylates A4220 (Table 1). Both MTases have been rigorously validated using various knockout cell lines that all show a reduction in m⁶A levels for the appropriate rRNA (Ignatova et al., 2020; H. Ma et al., 2019; Pinto et al., 2020; Rong et al., 2020; Sepich-Poore et al., 2022; Van Tran et al., 2019; Xing et al., 2020). Furthermore, methylated rRNA levels can be restored in the knockout cell lines via expression of catalytically active, but not inactive mutants of, METTL5 (Sepich-Poore et al., 2022) or ZCCHC4 (H. Ma et al., 2019). Knockout cell lines are viable, indicating neither MTase is essential. However, as described below, mechanistic studies have revealed that each MTase contributes to critical roles in translation and human health.

m⁶A1832 resides in a UAACA motif of helix 44 (Figure 1), which is proximal to the decoding center in the mature human ribosome (Natchiar et al., 2017). However, METTL5, which forms a heterodimer with the protein tRNA methyltransferase activator subunit 112 (TRMT112) (Brūmele et al., 2021; Sepich-Poore et al., 2022; Van Tran et al., 2019), likely adds the m⁶A mark at a time when A1832 is accessible during the biogenesis of 18S rRNA, which is in accord with endogenous METTL5 being detected predominantly in the nucleoli of HEK293T, HeLa, epidermoid carcinoma (A-431), and human osteosarcoma (U2OS) cell lines (Rong et al., 2020; Sepich-Poore et al., 2022; Sjöstedt et al., 2020). Furthermore, METTL5 appears to methylate only A1832 because anti-METTL5 miCLIP and enhanced crosslinking and immunoprecipitation studies detected METTL5 binding to primarily 18S rRNA and detected no significant changes in m⁶A levels of mRNA in METTL5-knockout cells (Van Tran et al., 2019). In vitro MTase activity assays showed that METTL5 modifies short, synthetic 18S rRNA mimics and purified 18S rRNA but does not methylate RNAs with AACU, GAACU, GGACU, and UAACG motifs (Rong et al., 2020; Xing et al., 2020).

METTL5 is expressed in all human cell types examined thus far (Sjöstedt et al., 2020) and various Mettl5-knockout cell lines have been generated using human (near-haploid human cell line [HAP1], human colon cancer cell line [HCT116], HEK293T, HeLa, HepG2) and mouse (B16, mESC, mouse embryonic fibroblasts) cells. Although there are some cell-type specific observations, the general trends that have emerged upon loss of METTL5 are as follows: differentiation is impaired (Ignatova et al., 2020; Xing et al., 2020), pluripotency is delayed/lost (Ignatova et al., 2020; Xing et al., 2020), no change in 18S rRNA levels but the number of polysomes is reduced (Ignatova et al., 2020; Rong et al., 2020), and translation is inefficient (Ignatova et al., 2020; Rong et al., 2020). Notably, ribosome profiling or sequencing experiments have revealed that decreased translation occurs on differentiation-promoting genes like F-box/WD repeat-containing protein 7 (Fbxw7; Xing et al., 2020). In addition, translation initiation is attenuated because there are fewer translation initiation/elongation factors associated with the 80S ribosome and phosphorylated ribosomal protein S6 kinase B1, a critical signal for translation initiation, is reduced (Rong et al., 2020). For human cell lines examined to date, it is estimated that ~60–100% of A1832 sites are modified with m⁶A (Ignatova et al., 2020; N. Liu et al., 2013; Rong et al., 2020; Taoka et al., 2018), a finding that prompts further investigation into the relationship between ribosome heterogeneity and functionality, particularly with respect to the stress response. Ultimately, additional experiments are needed to understand how a single methyl group at the N⁶ position of A1832 controls both global and gene-specific translational effects.

At the organism level, Mettl5-deficient mice are subviable, displaying several craniofacial abnormalities, cardiac defects, hearing and vision problems, infertility, reduced size, and inactivity (Y. Han, Du, et al., 2022; Ignatova et al., 2020; Sepich-Poore et al., 2022). Such observations are consistent with microcephaly and other developmental disorders observed in human patients having mutations in the Mettl5 gene (Reuter et al., 2017; Riazuddin et al., 2017; Richard et al., 2019; Torun et al., 2022). Phylogenetic analysis shows that METTL5 is conserved in most eukaryotes, except for fungi/yeast (Leismann et al., 2020). Thus, it is not surprising that parallel functional observations have been reported for METTL5 homologs in C. elegans (Liberman et al., 2020) (m⁶A1717), D. melanogaster (Leismann et al., 2020), and Danio rerio (i.e. zebrafish; Richard et al., 2019). It has also been noted that METTL5 expression decreases in aged human tissues (Rong et al., 2020) and gastric cancer (X. K. Wang, Zhang, et al., 2021) but increases in various tumor types (TCGA Research Network), including breast cancer tissue and cell lines (Rong et al., 2020), liver (Peng et al., 2022; W. Xu, Liu, et al., 2022), lung, and pancreatic cancer (H. Huang et al., 2022; S. Sun et al., 2021; Yan et al., 2021; Z. Zhang, Zhang, Luo, et al., 2021). Because the proliferation of breast and liver cancer cell lines decreased upon short interfering RNA knockdown of METTL5 (Peng et al., 2022; Rong et al., 2020; W. Xu, Liu, et al., 2022), METTL5 is a potential target for anti-cancer therapy.

m⁶A4220 is located in a UAACG loop motif of helix 81, which is adjacent to the active center of the mature human ribosome (Natchiar et al., 2017; Pinto et al., 2020). A4220 appears to be the major target of ZCCHC4, for both miCLIP and photoactivatable ribonucleoside-enhanced crosslinking and immunoprecipitation against ZCCHC4 did not show enrichment for other RNA targets (H. Ma et al., 2019; Van Tran et al., 2019). ZCCHC4 exhibits the greatest in vitro MTase activity when given a stem-loop RNA substrate containing UAAC in the loop (Pinto et al., 2020; Ren et al., 2019). Depending on cell type, ZCCHC4 has been detected in the nucleolus, nucleus, and cytoplasm (H. Ma et al., 2019; Pinto et al., 2020; Sjöstedt et al., 2020) and the percent of A4220 that is methylated ranges from ~55–100% (N. Liu et al., 2013; H. Ma et al., 2019; Pinto et al., 2020; Taoka et al., 2018; Van Tran et al., 2019). Thus, studies of various ZCCHC4-deficient cell lines (HAP1, HeLa, HepG2, HCT116, mouse embryonic cell line [J1 ES], mouse fibroblast cell line [NIH 3T3]) have some cell-type differences but have converged on the following conclusions: loss of ZCCHC4 does not alter rRNA production (H. Ma et al., 2019; Pinto et al., 2020; Van Tran et al., 2019) but decreases global translation (H. Ma et al., 2019) and ribosome occupancy (Pinto et al., 2020). Although ZCCHC4 orthologs are encoded in a wide range of animals (Pinto et al., 2020), no ZCCHC4-knockout animals have been reported. However, ZCCHC4 may be implicated in cancer and chemotherapy resistance (Z. Zhang, Zhang, Yang, et al., 2021). For example, ZCCHC4 is overexpressed and higher levels of methylated 28S rRNA are detected in hepatocellular carcinoma tumor tissue compared to normal tissue samples (H. Ma et al., 2019). Importantly, xenograft mouse models using Zcchc4-knockout cells showed reduced tumor growth compared to wild-type cells (H. Ma et al., 2019) and similarly, HepG2 cells lacking ZCCHC4 exhibit reduced cell proliferation (H. Ma et al., 2019). We anticipate that future studies will address how the loss of ZCCHC4 affects various model organisms and will establish how the N⁶-methyl group of A4220 results in specific phenotypes.

3 ∣. ROLES OF m⁶A MTases IN INFECTIOUS AGENTS: VIRUSES AND PARASITES

N⁶-methylation is an important modification in the progression of infectious diseases and has been identified in many human pathogens. Although m⁶A in viral RNA was first identified in the mRNA of influenza virus (Krug et al., 1976) and adenovirus type-5 (Moss & Koczot, 1976; Sommer et al., 1976) in 1976, m⁶A marks have since been identified in the genomes and/or transcriptomes of various viruses, such as herpes simplex virus 1 (HSV-1; Feng et al., 2022; Moss et al., 1977), Simian virus 40 (SV40; Finkel & Groner, 1983; Tsai et al., 2018), Rous sarcoma virus (RSV; Kane & Beemon, 1985), human immunodeficiency virus 1 (HIV-1; Kennedy et al., 2016; Lichinchi, Gao, et al., 2016), flaviviruses such as Zika virus (ZIKV), West Nile virus, yellow fever, dengue, and others (Gokhale et al., 2016), hepatitis B (Imam et al., 2018) and C (HBV and HCV; Gokhale et al., 2016), Kaposi's sarcoma-associated herpes virus (KSHV; Baquero-Perez et al., 2019; Tan et al., 2017; F. Ye et al., 2017), Epstein–Barr virus (EBV; Lang et al., 2019), enterovirus 71 (EV71; Hao et al., 2019), porcine endemic diarrhea virus (PEDV; J. Chen, Jin, et al., 2020), human metapneumovirus (hMPV; Lu et al., 2020), severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2; Burgess et al., 2021; J. Liu et al., 2021), and human papillomavirus 16 (X. Cui et al., 2022). The percentage of adenosines with m⁶A marks in viral RNA genomes varies from 0.089% in EV71 to 3% in ZIKV (Lichinchi, Zhao, et al., 2016). Eighteen m⁶A marks were identified in respiratory syncytial viral mRNAs, the highest for any virus (Xue et al., 2019). During viral infection, m⁶A levels of host transcripts can increase or decrease depending on the particular m⁶A mark (Gokhale et al., 2016; Jansens et al., 2022; N. Li, Hui, et al., 2021; Lichinchi, Gao, et al., 2016; Lichinchi, Zhao, et al., 2016; Lichinchi & Rana, 2019; J. Liu et al., 2021; X. Qiu, Hua, et al., 2021; Selberg et al., 2021). Interestingly, cellular m⁶A methylation is almost completely inhibited in pseudorabies virus-infected cells due to phosphorylation of the METTL3/METTL14/WTAP complex by viral proteins (Jansens et al., 2022).

To date, no instances of virally encoded m⁶A MTases have been discovered; therefore, m⁶A modifications are likely mediated via expression and cellular localization changes of host writers and erasers. HIV-1, HCV, KSHV, and HBV all depend on the host m⁶A writer METTL3, which was confirmed by the knockdown or knockout of METTL3 (Gokhale et al., 2016; Imam et al., 2018; Lichinchi, Gao, et al., 2016; Tan & Gao, 2018; F. Ye et al., 2017). Viruses using host METTL3 are further supported by the retention of the DRACH motif, the consensus sequence of METTL3/METTL14 methylation, or at least RAC in viral genomic RNA and/or viral mRNA isolated from HIV-1, HCV, KSHV, influenza virus, HBV, SV40, respiratory syncytial virus, ZIKV, and hMPV (Bayoumi & Munir, 2021; Courtney et al., 2017; Gokhale et al., 2016; Imam et al., 2018; Lichinchi, Gao, et al., 2016; Lichinchi & Rana, 2019; Tsai et al., 2018; Xue et al., 2019; F. Ye et al., 2017). Conserved methylation sites were also mapped in many Flaviviridae genomes, particularly in the nonstructural protein 3 (NS3) helicase, nonstructural protein 5 (NS5) RNA-dependent RNA-polymerase (RdRp), and envelope genes (Gokhale et al., 2016). m⁶A marks in viral genomes are also frequently colocalized with hairpins and G-quadruplexes, suggesting that some m⁶A sites in viruses require RNA structure for recognition by the m⁶A writer (Fleming et al., 2019). In HBV, a viral protein known as HBV X protein directs METTL3/METTL14 to viral RNA as well as at least one host mRNA, phosphatase and tensin homolog or PTEN (G. W. Kim & Siddiqui, 2021).

m⁶A marks affect the viral life cycle at various stages. For example, m⁶A marks dampen viral replication for ZIKV, KSHV, EBV, HBV, HCV, PEDV, and SARS-CoV-2 by as much as fourfold (J. Chen, Jin, et al., 2020; G. W. Kim et al., 2020; Lang et al., 2019; Lichinchi, Zhao, et al., 2016; J. Liu et al., 2021; Tan et al., 2017) but stimulate viral replication for HIV-1, influenza A virus, human cytomegalovirus, SV40, EV71, respiratory syncytial virus, and hMPV by as much as 10-fold (Courtney et al., 2017; Hao et al., 2019; Lichinchi, Gao, et al., 2016; Lu et al., 2020; Rubio et al., 2018; Tsai et al., 2018; Xue et al., 2019). EV71 infection noticeably upregulates expression of METTL3/METTL14 (Hao et al., 2019). The EV71 RdRp 3D interacts with METTL3, which enhances sumoylation and ubiquitination of the RdRp 3D, thereby increasing viral replication (Hao et al., 2019; Y. Liu et al., 2016). For ZIKV, knockdown of METTL3/METTL14 decrease m⁶A levels in ZIKV RNA and lowered ZIKV RNA release by approximately twofold, while knockdown of FTO and ALKBH5 had the opposite effect (Lichinchi, Zhao, et al., 2016). An increase in m⁶A marks also increases viral RNA nuclear export (Casu et al., 2013; Lichinchi, Zhao, et al., 2016), enhances splicing (Price et al., 2020; F. Ye, 2017), differentially affects viral RNA stability and protein expression (Imam et al., 2018; Tsai et al., 2018), and allows the viral RNA to avoid detection by the host via pattern recognition receptors such as RIG-I-like receptors or Toll-like receptors (Durbin et al., 2016; Karikó et al., 2005; N. Li, Hui, et al., 2021; Lu et al., 2020; W. Qiu, Zhang, et al., 2021).

Much of the viral-centric research has examined effects upon METTL3 knockout/knockdown, yet only recently have METTL16, METTL5, and ZCCHC4 been validated as m⁶A MTases. This point presents interesting questions for future research: Are there roles for m⁶A MTases other than METTL3/METTL14 in methylating viral RNAs? Are other writers, such as METTL16 under SAM-limiting conditions, playing any regulatory roles? The involvement of other m⁶A writers seems possible given (i) their ability to interact with highly structured RNAs, which are common in viral genomes and transcripts, and (ii) some m⁶A marks survive disruption of the DRACH motif in SV40 (Tsai et al., 2018). When m⁶A levels increase or decrease in viral RNAs, what is the regulatory interplay among the m⁶A machinery of writers, readers, and erasers? Are there instances where methylation at specific sites triggers structural switches (Imam et al., 2018)? Would drugs directly targeting specific m⁶A MTases be effective antivirals? Although no MTase-specific inhibitors have been used as an antiviral therapeutic, it is worth mentioning that 3-deazaadenosine, a nucleoside analog that inhibits S-adenosyl-L-homocysteine (SAH) hydrolase, has been successful as an antiviral for RSV, HSV-1, reovirus, measles, vesicular stomatitis virus, parainfluenza virus, respiratory syncytial virus, and HIV-1 in various model systems (Bader et al., 1978; De Clercq & Montgomery, 1983; Mayers et al., 1995; Montgomery et al., 1982; Wyde et al., 1990), suggesting that changes to the cellular methylation potential via MTases may be effective.

In addition to viruses, m⁶A impacts the biology of human parasitic infections in the parasite Plasmodium falciparum, which causes malaria, and the obligate intracellular parasite Toxoplasma gondii, which causes toxoplasmosis. m⁶A regulates the parasitic life cycle by controlling the intra-erythrocytic development stage in P. falciparum and the growth to the bradyzoite stage in T. gondii (Baumgarten et al., 2019; Holmes et al., 2021). Interestingly, these parasites have their own putative METTL3 ortholog MTases as well as associated METTL14 and WTAP orthologs (Baumgarten et al., 2019; Holmes et al., 2021). Additional binding partners were co-precipitated for both parasites, although, they did not match proteins that are analogous to those in the human METTL3/METTL14 complex. The METTL3 orthologs have their own recognition sequence: GGACA in P. falciparum and YGCAUGCR (Y = pyrimidine and R = purine) in T. gondii (Baumgarten et al., 2019; Holmes et al., 2021). Trypanosoma brucei, which causes trypanosomiasis, does not have a METTL3 ortholog. There is likely another MTase that installs m⁶A marks, for a CAU motif was identified as a potential methylation sequence in coding sequences, UTRs and intergenic regions (L. Liu, Zeng, et al., 2019; Viegas et al., 2022). T. brucei has a reported threefold increase in m⁶A sites primarily within its coding sequences between T. brucei life stages: the bloodstream form and procyclic form (L. Liu, Zeng, et al., 2019). T. brucei has also been shown to use m⁶A marks in poly(A) tails to protect against deadenylation and degradation (Viegas et al., 2022). For parasitic infections involving Cryptosporidium parvum, there is a 30% global increase in the m⁶A methylome of host cells eight hours postinfection due to down regulation of ALKBH5, FTO, and a nuclear factor kappa-light-chain-enhancer of activated B signaling pathway for innate antimicrobial immune defense (Z. Xia, Xu, et al., 2021). Important future areas of study include how widespread METTL3 orthologs are in human parasites and the prevalence of m⁶A methylome changes for host and parasitic RNAs. Please refer to a review by Catacalos et al for more details on the biology of m⁶A in parasites (Catacalos et al., 2022).

4 ∣. WHO MADE THE MARK: TRANSCRIPTOME-WIDE m⁶A DETECTION METHODS

Due to their improved precision and quantitation, detection methods for m⁶A sites have played a pivotal role in elucidating the biological functions of m⁶A MTases (O’Brown & Greer, 2016; Zaccara et al., 2019; Z. Zhang et al., 2019; Zhu et al., 2019). Early methods were limited to single-RNA strategies such as mass spectrometry and 2D thin layer chromatography (Mongan et al., 2019). Current methods can now site specifically detect m⁶A throughout the entire transcriptome (see the review article by Y. Wang & Jia, 2020, for more information on detection methods; Y. Wang & Jia, 2020). Herein we focus on studies that aim to reveal which m⁶A RNA MTase, namely METTL3/METTL14 versus METTL16, is responsible for the respective marks.

Antibody based m⁶A-seq methods are fairly common; however, the anti-m⁶A antibodies bind to m⁶A introduced by any MTase. A means to determine which m⁶A marks are introduced by certain MTases is to knockdown or knockout (if cell viability is not disrupted) the particular MTase of interest. This approach has been coupled with a method called m⁶ACE-seq, which utilizes 5′-3′ exoribonuclease 1 (XRN1), to achieve single-nucleotide resolution (Koh et al., 2019). The antibody photocrosslinked to an m⁶A-containing RNA protects the RNA at the m⁶A site from XRN1 digestion, leaving the reads that start at the m⁶A site (Koh et al., 2019). Knockout of METTL3 and phosphorylated CTD interacting factor 1 (PCIF1), an enzyme that converts 2’-O-methyladenosine (Am) to N⁶,2′-O-dimethyladenosine (m⁶Am; see Section 8 for more details about PCIF1), indicated 14,508 m⁶A sites and 3798 m⁶Am sites introduced by the respective MTases in HEK293T cells. In contrast, METTL16 knockdown generated ambiguous results because m⁶A levels fluctuated for 602 m⁶A sites, including those assigned to be installed by METTL3 or PCIF1. METTL16 regulates SAM homeostasis; therefore, fluctuations in m⁶A sites not directly targeted by METTL16 were likely due to the dysregulation of SAM/SAH levels. Only 136 m⁶A sites were deduced to be solely METTL16 dependent. Another study performed m⁶A MeRIP-seq on HEK293T-METTL16 knockout cells to further identify m⁶A marks installed by METTL16 (R. Su, Dong, et al., 2022). This study concluded that METTL16 methylates sites in 334 mRNAs. Similarly, methylation CLICK-degradation sequencing (meCLICK-seq) was performed using short hairpin RNA-mediated knockdown of METTL3 and METTL16 in MOLM-13 cells (Mikutis et al., 2020). For meCLICK-seq, cell cultures are grown in the presence of a methionine analog, propargyl-L-selenohomocysteine, which the cell converts into the SAM surrogate propargylic Se-adenosyl-L-selenomethionine, so that MTases will introduce a propargyl, rather than a methyl, modification onto RNA targets (Mikutis et al., 2020). Then, the Cu(I)-catalyzed azide—alkyne cycloaddition degrades methylated RNA, but not propargyl-modified RNA, to enable measurement of m⁶A modifications through the depletion of RNA (Mikutis et al., 2020). meCLICK-seq findings were reported with respect to RNAs rather than m⁶A sites: 6130 RNAs had METTL3-dependent m⁶A marks, 8545 RNAs had METTL16-dependent m⁶A marks, and 5721 RNAs had both METTL3- and METTL16-dependent m⁶A marks.

Some recently developed methods do not use antibodies, thereby mitigating antibody cross reactivity to m⁶Am (Linder et al., 2015) while enriching for m⁶A marks catalyzed by certain MTases, usually METTL3. Many of these methods operate under the assumption that a METTL3-dependent m⁶A mark occurs in the DRACH motif. One such method is the deamination adjacent to RNA modification targets (DART-seq) method (Meyer, 2019). Apolipoprotein B mRNA editing enzyme, catalytic polypeptide 1 (APOBEC1), a cytidine deaminase, is fused to a YTH domain that binds to m⁶A (Meyer, 2019). Because a majority of m⁶A marks are upstream of C in the DRACH motif, the YTH domain can bind to m⁶A and then APOBEC1 will convert the adjacent “C” to “U”, providing single-nucleotide resolution of the DRACH-to-DRAUH motif (Meyer, 2019). DART-seq identified 40,263 DRAUH motifs in 7707 RNAs. It is worth pointing out that METTL16 methylates UACAGARAA motifs; therefore, it would be interesting to know if UAUAGARAA motifs were present in the DART-seq results. Another antibody-free method, MAZTER sequencing, uses the enzyme endoribonuclease toxin MazF, which cleaves unmethylated ACA sites to detect unmethylated and methylated ACA locations at single-nucleotide resolution. However, this method is limited to m⁶A marks occurring in an ACA context so METTL3 sites would be targeted (Garcia-Campos et al., 2019; Pandey & Pillai, 2019).

Third-generation sequencing technologies, such as single-molecule real-time sequencing and Oxford Nanopore Technologies, are now being adapted for the direct detection of modified nucleotides, such as m⁶A, at single-nucleotide resolution (Lucas & Novoa, 2023). Traditionally, the DNA or RNA is pulled through a nanopore inside a membrane which causes characteristic disruptions in a measured current. These disruptions are unique to specific nucleotides, including modifications like m⁶A. Although deconvoluting disruptions of modified nucleotides has been challenging (Garalde et al., 2018; H. Liu, Begik, et al., 2019), recent advances in algorithms have improved the accuracy of m⁶A detection to roughly 90% for in vitro transcribed RNA and 87% from polyA-selected RNA isolated from S. cerevisiae (H. Liu, Begik, et al., 2019). This methodology could also be applied to cell lines with knockdown/knockouts of specific MTases.

In summary, many detection methods have been developed over the years to identify and to quantitate m⁶A in RNA (Y. Wang & Jia, 2020), although none of the methods can track m⁶A marks catalyzed by a specific MTase akin to what is possible for 5mC (Dai et al., 2021). Such technology might provide stronger evidence for whether there are other m⁶A RNA MTases.

5 ∣. HOW TO WRITE m⁶A: CATALYTIC AND KINETIC MECHANISMS

A majority of 6mA and m⁶A MTases possess a similar catalytic mechanism of methyl transfer due to the conserved (D/N/S)PP(F/W/Y) catalytic core (see Section 7.5 for more details; Bheemanaik et al., 2006). As shown in Figure 2a, the side chain of D/N/S within the catalytic core forms a hydrogen bond with the amine of the N⁶ and the other hydrogen of the amine interacts with the carbonyl of the proline-proline peptide backbone. These interactions enhance the nucleophilic capacity of the nitrogen at N⁶. During the nucleophilic attack, the amine is deprotonated, leading to a C-N bond with the methyl group of SAM. Once the methyl group is fully transferred to adenosine, the other resulting product is SAH (or sometimes referred to as AdoHcy) with a lone pair of electrons on the sulfur. Please note that the methyl group is directly transferred to N⁶ of adenosine and does not occur via an intermediate such as an N¹-intermediate (Pogolotti et al., 1988). The above-described catalytic mechanism is favored by many, although an alternative mechanism involving amino-imino tautomerization has been proposed (Mashhoon & Reich, 1994).

FIGURE 2.

Chemical mechanism of methylation catalyzed by m⁶A RNA methyltransferases and the possible substrate-binding mechanisms for SAM-dependent methyltransferases. (a) Schematic shows the transfer of the methyl group (blue) from SAM, the methyl donor, to the acceptor adenosine, N⁶-methyladenosine. SAH is a byproduct. Catalysis depends on a conserved (D/N/S)PP(F/W/Y) motif. Note that the F/W/Y portion orients the acceptor adenosine and is not depicted for clarity. (b) The schemes depict binding order for SAM and RNA for random-order binding mechanism, ordered-sequential binding mechanisms, and a double-displacement reaction or ping-pong binding mechanism. m⁶A, N⁶-methyladenosine; SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine.

Little is confirmed about the transition state of m⁶A MTases. For MTases in general, the transition state tends to involve the proper arrangement of the methyl acceptor, whether it is an adenosine, cytidine, or other substrate, with the SAM methyl donor in the active site (Lopez et al., 2016). The methyl group of adenosine forms a trigonal bipyramidal arrangement in which the hydrogens of the methyl group are in a planar arrangement and the methyl group is shared between the sulfur of SAM and the recipient nitrogen atom of the methyl acceptor (S. Chen, Kapilashrami, et al., 2020; Q. Du, Wang, & Schramm, 2016; Mahmoodi et al., 2020; Newby et al., 2002). Computational modeling suggests that a hydrogen on the N⁶ may remain as the nitrogen-methyl group forms a bond; the hydrogen will be subsequently removed by a basic group as the methyl-adenosine intermediate has improved acidity relative to the reactant (Newby et al., 2002). For m⁶A MTases, structural, biochemical, and computational studies have not yet revealed the actual transition state, distances from the methyl group of SAM to the methyl acceptor, conformational rearrangement of the SAM ligand, and other perturbations to the adenosine and/or RNA in 3D space. Nonetheless, it seems likely that m⁶A MTases may share transition state features with other MTases.

m⁶A RNA MTases have two different substrates; therefore, the binding order of SAM and RNA may vary depending upon the MTase and RNA substrate. Most characterized MTases employ either a random sequential mechanism, where SAM and DNA/RNA bind in any order (Figure 2b), or an ordered-sequential mechanism, where SAM binds prior to DNA/RNA or vice versa (Figure 2b; Bheemanaik et al., 2006). A few MTases, such as 23S rRNA m²A2503 methyl-transferase/tRNA m²A37 methyltransferase and 23S rRNA adenosine-C⁸-methyltransferase Cfr, which methylate positions C² and C⁸ of A2503 in 23S rRNA, respectfully, use a ping-pong mechanism (Figure 2b), whereby the MTase will receive the methyl group from SAM, forming a methyl-MTase intermediate, then eject SAH before binding and methylating the methyl acceptor DNA/RNA (Grove et al., 2011). Substrate binding order for METTL16 methylating MAT2A hp1 is reported to be a random-sequential binding mechanism (Figure 2b) based on similar rates obtained from a preincubation assay (Yu, Kaur, et al., 2021); however, a solved structure of METTL16 bound to a minimal MAT2A hp1 RNA suggested that METTL16 requires SAM to bind first because the METTL16•RNA complex showed no possible entry for SAM to enter the active site (Doxtader et al., 2018). For U6 snRNA, METTL16 displays an ordered-sequential binding mechanism with U6 snRNA binding prior to SAM (Figure 2b), suggesting the possibility that binding mechanisms may depend on the substrate (Breger & Brown, 2023). The binding order for the other human m⁶A RNA MTases remains unknown.

Select kinetic parameters of the known human m⁶A MTases have been measured and the values revealed a wide range despite the high structural similarity of MTases. The turnover numbers (k_cat) for METTL3/METTL14 and METTL5 are rather close at 18 and 17.6 h⁻¹, respectively, while METTL16 has turnover rates of 4.2 h⁻¹ for U6 snRNA (Breger & Brown, 2023) and 30 h⁻¹ for the MAT2A hp1 (Table 2; F. Li et al., 2016; Yu, Kaur, et al., 2021). Substrate binding affinity can be evaluated based on the various equilibrium dissociation constant (K_D) or Michaelis–Menten constant (K_M) values. METTL3/METTL14, METTL5/TRMT112, and ZCCHC4 all have binding values (either K_D or K_M) for SAM in the range of 1–7 μM (Table 2) (Ren et al., 2019; X. Wang, Feng, et al., 2016; Yu, Kaur, et al., 2021). In contrast, METTL16 has a K_D of 126 μM for the METTL16•U6 snRNA•SAM complex (Breger & Brown, 2023). This unusually weak affinity for SAM may be due to METTL16 regulating intracellular SAM concentrations, which are reported to be approximately 10 μM (Breger & Brown, 2023; Doxtader et al., 2018; Pendleton et al., 2017; Shima et al., 2017; C. Ye & Tu, 2018). For RNA substrate binding affinity, METTL16 displays K_D values of 16–18 and 42 nM for U6 snRNA and MAT2A hp1, respectively (Table 2; Aoyama et al., 2020; Breger & Brown, 2023), whereas METTL3/METTL14 has a K_D at 23 μM and METTL5/TRMT112 has a K_M for its rRNA substrate at 1.1 μM (Table 2; Yu, Kaur, et al., 2021). Interestingly, METTL3/METTL14 exhibits a 75-fold greater catalytic efficiency on damaged dsDNA than ssRNA (0.02 vs. 1.5 μM⁻¹ min⁻¹), raising the intriguing possibility that METTL3/METTL14 functions as a 6mA DNA MTase (L. Q. Chen et al., 2022; Woodcock et al., 2019; Yu, Horton, et al., 2021). In addition, protein-binding partners further impact RNA-binding affinity and MTase activity (S. Su, Li, et al., 2022). Most kinetic parameters are yet to be determined for ZCCHC4.

TABLE 2.

Kinetic parameters of the human m⁶A RNA methyltransferases.

MTase (RNA substrate)	k cat	KD (E•SAM) or KM	KD (E•RNA) or KM
METTL3/METTL14 (RNA with DRACH motifa)	18 ± 2 h−1 (F. Li et al., 2016)	KD = 1.5 ± 0.2 μM (X. Wang, Feng, et al., 2016)	KD = 23.0 ± 2.3 μM (J. Huang et al., 2019)
METTL16 (U6 snRNA)	~3 h−1 (Aoyama et al., 2020) 4.2 ± 1.2 h−1 (Breger & Brown, 2023)	KD = 126 ± 6 μMb (Breger & Brown, 2023) KM = 132 ± 8 μM (Breger & Brown, 2023)	KD = 16 ± 2 nM (Aoyama et al., 2020) KD = 18 ± 7 nM (Breger & Brown, 2023) KM = 25 ± 10 nM (Aoyama et al., 2020)
METTL16 (MAT2A hp)	~30 h−1 (Yu, Horton, et al., 2021)	KM >0.4 mM (Yu, Horton, et al., 2021)	KD = 42 ± 6 nM (Aoyama et al., 2020) KM = 29 ± 6 nMc (Aoyama et al., 2020) KM = ~10 μMc (Yu, Horton, et al., 2021)
METTL5/TRMT112 (UCGUAACAAGGUUU)	17.6 ± 0.6 h−1 (vary [SAM]) (Yu, Horton, et al., 2021) 13.1 ± 0.8 h−1 (vary [RNA]) (Yu, Horton, et al., 2021)	KM = 1.0 ± 0.2 μM (Yu, Horton, et al., 2021)	KM = 1.1 ± 0.2 μM (Yu, Horton, et al., 2021)
ZCCHC4	Unknown	KD = 6.7 ± 0.4 μM (Ren et al., 2019)	Unknown

Abbreviations: K_D, equilibrium dissociation constant; K_M, Michaelis–Menten constant; m⁶A, N⁶-methyladenosine; METTL3, methyltransferase-like protein 3; METTL5, methyltransferase-like protein 5; METTL14, methyltransferase-like protein 14; METTL16, methyltransferase-like protein 16; MTase, methyltransferase; SAM, S-adenosylmethionine; ZCCHC4, zinc-finger CCHC-domain-containing protein 4.

^aTurnover number, k_cat, value measured using UACACUCGAUCUGGACUAAAGCUGCUC; K_D value measured using AAAAGGACUAAAA.

^bValue reflects METTL16•U6 snRNA•SAM complex.

^cReason for K_M values differing by ~345-fold is unclear but could be due to shortened hp RNA, SAM, or reaction conditions.

Currently, METTL16 is the only human m⁶A RNA MTase with information regarding rate-limiting steps. For the methylation of U6 snRNA, the steady-state rate of 0.07 min⁻¹ was eightfold less than the single-turnover rate of 0.56 min⁻¹, indicating that steps following methylation were rate-limiting, such as product release or conformational rearrangements (Breger & Brown, 2023). Although the rate-limiting steps of other human m⁶A RNA MTases are unknown, other MTases exhibit different rate-limiting steps during methylation. METTL3/METTL14 displays slower methylation activity on higher-affinity substrates, which might suggest that product release is rate limiting (Qi et al., 2022). Metal ions, such as zinc ions, and protein cofactors, such as METTL14 and TRMT112, are another consideration for certain m⁶A MTases. METTL3 has two CCCH domains (i.e., zinc-finger domain [ZnF] 1 and 2) that bind zinc and are needed for MTase activity in vitro (Huang et al. 2019). Similarly, ZCCHC4 requires zinc due to the GRF, C2H2, and CCHC zinc-finger domains (all named after the conserved residues) (Ren et al., 2019). In contrast, zinc ions inhibit the MTase activity of METTL16 and METTL5 (Breger & Brown, 2023; Yu, Kaur, et al., 2021). METTL16 does not appear to have any zinc fingers (Yu, Kaur, et al., 2021) and neither does METTL5, although its binding partner TRMT112 does have a zinc-binding domain (Van Tran et al., 2019). In theory, methyl transfer could also be limited by conformational changes. An interesting distinction between the 6mA and m⁶A MTases is base flipping. Because DNA is usually double-stranded, 6mA MTases often need to “flip” the adenosine acceptor into an extra-helical position to receive the methyl group (Bheemanaik et al., 2006). In general, it is thought that m⁶A RNA MTases do not employ base flipping because the targeted adenosines are not base paired or have the rotational freedom to orient N⁶ for methylation (J. Liu et al., 2014; Pendleton et al., 2017; Van Tran et al., 2019). However, METTL5 appears to have the A1832 more exposed, which suggests the MTase may be capable of base flipping (Van Tran et al., 2019).

6 ∣. SMALL MOLECULES AND OTHER APPROACHES TARGETING m⁶A MTases

DNA and RNA MTases have been targeted with an array of small molecules, acting as either inhibitors or activators, for both research and therapeutic purposes. Some of the first inhibitors were products of the methylation reaction: methylated DNA/RNA and SAH (Figure 3). Dead-end analog inhibitors for m⁶A MTases include a DNA/RNA, whereby the adenosine acceptor is substituted with another nucleobase or the use of SAM analogs like sinefungin (Figure 3). Sinefungin was first isolated from Streptomyces incarnatus and Streptomyces griseolus, where it was used as an antibiotic, but due to its similarity in structure to SAM, it is often used as an MTase competitive inhibitor in kinetic assays (Horiuchi et al., 2013; Nolan, 1987; Selberg et al., 2019). Thus, both product and dead-end analog inhibitors have been useful research tools for kinetic assays. There have also been advancements in bisubstrate analogs, molecules that generally mimic both the SAM and targeted adenosine moieties, which inhibit MTases, such as the bacterial rRNA m⁶A methyltransferase 23S rRNA m⁶A2030 methyltransferase (RlmJ, Figure 3; Oerum et al., 2019). The bisubstrate analogs may provide mechanistic insights about RNA recognition and the catalytic mechanism, such as the transition state, utilized by the human m⁶A MTases.

FIGURE 3.

Chemical structures of small molecules that target methyltransferases. Where indicated, IC₅₀ and EC₅₀ values have been determined for the METTL3/METTL14/WTAP complex; however, IC₅₀ values for METTL3 inhibitors STM2457 and UZH1a were determined in the presence of METTL3/METTL14. EC₅₀, half-maximal effective concentration; IC₅₀, half-maximal inhibitory concentration; METTL3, methyltransferase-like protein 3; METTL14, methyltransferase-like protein 14; WTAP, Wilms′ tumor 1-associating protein.

Interest in developing small molecules to target MTases, particularly METTL3 (see reference Fiorentino et al., 2023 for review), for therapeutic purposes is growing. Artificial intelligence and high-throughput screening methods, such as the high-throughput-methyl-reading assay, will enable fast and easy means of testing libraries of compounds for potential activation/inhibition of MTases (Xiao et al., 2022) in addition to virtual high-throughput screening methods that have yielded promising hits for METTL3 (Selberg et al., 2019) and METTL16 (Mitra et al., 2023). Inhibitors of METTL3 would be therapeutically useful because inhibition limits the growth of AML (Barbieri et al., 2017). A few companies, such as STORM therapeutics, have discovered inhibitors for METTL3 and other RNA-modifying enzymes (Cully, 2019; Lee et al., 2022). Inhibitors, such as STM2475 and UZH1a (Figure 3), have been observed to occupy the SAM-binding site of METTL3 based on X-ray crystal structures (Table 3) (Moroz-Omori et al., 2021; Yankova et al., 2021). STM2475 and UZH1a exhibit half-maximal inhibitory concentration (IC₅₀) values of 2.2 and 4.6 μM, respectively, in MOLM-13 cells and prevent proliferation of AML cell lines (Moroz-Omori et al., 2021; Yankova et al., 2021). Other molecules are also currently being investigated for their inhibition of METTL3 in the context of AML (Sabnis, 2021). Currently, the drug STC-15, developed by STORM therapeutics to target METTL3 in a solid tumor, is in Phase 1 clinical trials.

TABLE 3.

All currently available structures of the human m⁶A methyltransferases with their associated Protein Data Bank identifiers, resolutions, amino acid numbers of recombinant protein, and ligand bound (if any).

Protein name	PDB ID	Resolution (Å)	Amino acids	Ligand
METTL3/METTL14	5IL0	1.88	MTase (369–580/109–408)	Apo
	5IL1	1.71	MTase (369–580/109–408)	SAM
	5IL2	1.61	MTase (369–580/109–408)	SAH
	5K7M	1.65	MTase (356–580/108–456)	Apo
	5K7U	1.70	MTase (356–580/108–456)	SAM
	5K7W	1.65	MTase (356–580/108–456)	SAH
	5L6D	1.85	MTase (354–580/107–395)	SAH
	5L6E	1.90	MTase (354–580/107–395)	SAM
	5TEY	1.80	MTase (407–568/186–363)	SAH
	5YZ9	(NMR)	METTL3 ZnF1-ZnF2 (259–357)	Apo
	6TTP	2.0	MTase (354–580/107–395)	Adenosinea
	6TTT	2.30	MTase (354–580/107–395)	ASI_M3M_140a
	6TTV	2.14	MTase (354–580/107–395)	ASI_M3M_138a
	6TTW	2.20	MTase (354–580/107–395)	ASI_M3M_047a
	6TTX	2.00	MTase (354–580/107–395)	ASI_M3M_051a
	6TU1	2.31	MTase (354–580/107–395)	ASI_M3M_091a
	6Y4G	1.90	MTase (354–580/107–395)	Sinefungina
	7ACD	2.50	MTase (354–580/107–395)	T30a
	7NHG	2.5	MTase (354–580/107–395)	ASI_M3M_041a
	7NHH	2.1	MTase (354–580/107–395)	UOZ002a
	7NHI	1.85	MTase (354–580/107–395)	UOZ004a
	7NHJ	2.16	MTase (354–580/107–395)	DHU_M3M_154a
	7NHV	1.91	MTase (354–580/107–395)	UOZ016a
	7NI7	2.5	MTase (354–580/107–395)	UOZ031a
	7NI8	2.2	MTase (354–580/107–395)	UOZ040aa
	7NI9	2.2	MTase (354–580/107–395)	UOZ058a
	7NIA	2.3	MTase (354–580/107–395)	UOZ059aa
	7NID	2.3	MTase (354–580/107–395)	UOZ078a
	7O08	2.0	MTase (354–580/107–395)	ADO_AB_075a
	7O09	1.8	MTase (354–580/107–395)	ADO_AC_074a
	7O0L	1.9	MTase (354–580/107–395)	ADO_AC_093a
	7O0M	2.39	MTase (354–580/107–395)	ADO_AD_023a
	7O0P	2.7	MTase (354–580/107–395)	ADO_AD_022a
	7O0Q	2.49	MTase (354–580/107–395)	ADO_AD_066a
	7O0R	2.3	MTase (354–580/107–395)	ADO_AE_026a
	7O27	2.4	MTase (354–580/107–395)	ADO_AE_005a
	7O28	2.47	MTase (354–580/107–395)	ADO_AE_009a
	7O29	2.75	MTase (354–580/107–395)	ADO_AD_044a
	7O2E	2.5	MTase (354–580/107–395)	ADO_AD_089a
	7O2F	2.1	MTase (354–580/107–395)	UZH2a
	7O2H	2.5	MTase (354–580/107–395)	ADO_AD_091a
	7O2I	3	MTase (354–580/107–395)	STM2457a
	7O2X	2.80	MTase (354–580/107–395)	T180a
	7OED	2.0	MTase (354–580/107–395)	UOZ019aa
	7OEE	2.7	MTase (354–580/107–395)	UOZ019ba
	7OEF	2.03	MTase (354–580/107–395)	UOZ038a
	7OEG	2.79	MTase (354–580/107–395)	UOZ040ba
	7OEH	2.01	MTase (354–580/107–395)	UOZ059ba
	7OEI	2.48	MTase (354–580/107–395)	UOZ083a
	7OEJ	2.3	MTase (354–580/107–395)	UOZ090a
	7OEK	1.9	MTase (354–580/107–395)	UOZ091a
	7OEL	1.86	MTase (354–580/107–395)	UOZ097a
	7OEM	2.2	MTase (354–580/107–395)	UOZ120a
	7OQL	2.5	MTase (354–580/107–395)	UOZ094a
	7OQO	3.35	MTase (354–580/107–395)	UOZ111a
	7OQP	2.0	MTase (354–580/107–395)	UOZ113a
	34169b	4.4	MTase (1–580/1–456); no residues modeled	HWVZb
METTL16	6B91	1.94	MTase (1–291)	Apo
	6B92	2.10	MTase (1–291)	SAH
	6DU4	1.70	MTase (2–310)	MAT2A HP1
	6DU5	3.01	MTase (2–310)	MAT2A HP6
	6GFK	2.30	MTase (1–251)	Apo
	6GFN	2.86	MTase (1–291)	Apo
	6GT5	2.45	MTase (1–291)	Apo
	6M1U	2.79	VCR1 (310–410); VCR2 (509–562)	Apo
METTL5/TRMT112	6H2U	1.60	MTase (1–209/1–118)	SAM
	6H2V	2.49	MTase (1–209/1–125)	SAM
ZCCHC4	6UCA	3.10	GRF-C2H2-MTase-CCHC (23–464)	SAH

Abbreviations: HAKAI, E3 ubiquitin-protein ligase Hakai; METTL3, methyltransferase-like protein 3; METTL5, methyltransferase-like protein 5; METTL14, methyltransferase-like protein 14; METTL16, methyltransferase-like protein 16; MTase, methyltransferase; SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; WTAP, Wilms' tumor 1-associating protein; VCR1/2, vertebrate conserved region 1/2; VIRMA, vir-like m⁶A methyltransferase associated protein; ZCCHC4, zinc-finger CCHC-domain-containing protein 4; ZC3H13, zinc-finger CCCH domain-containing protein 13.

^aLigands designed as METTL3/METTL14 inhibitors; please refer to original published source for details on each ligand.

^bID for Electron Microscopy Data Bank (EMDB) because there is no model deposited in PDB. HWVZ represents the protein-binding partners HAKAI, WTAP, VIRMA, and ZC3H13, respectively.

In addition to inhibitors, molecules that activate or enhance the activity of MTases represent a potential therapeutic approach; activation of METTL3/METTL14 could increase m⁶A RNA levels and thereby mitigate symptoms of hypomethylation, such as Type 2 diabetes or tumorigenesis (Q. Cui et al., 2017; De Jesus et al., 2019; J. Li & Gregory, 2021; Z. Zhang, Zhang, Luo, et al., 2021). To identify activators, one study employed a virtual drug screening approach and reported four compounds (methyl piperidine-3-carboxylate hydrochloride, tert-butyl 6-methylpiperidine-3-carboxylate, methyl 6-methylpiperidine-3-carboxylate, and methyl piperazine-2-carboxylate; Figure 3) that putatively target the METTL3/METTL14/WTAP complex (Selberg et al., 2019). Subsequent testing of these compounds, methyl piperidine-3-carboxylate hydrochloride and methyl piperazine-2-carboylate, showed that they enhanced methylation activity by approximately 1.2- and 1.5-fold, respectively (Selberg et al., 2019). The K_D for METTL3•SAM is 1.92 μM but improves to 4.7 and 13.7 nM in the presence of 1–100 nM methyl piperidine-3-caroxylate hydrochloride and 25 μM methyl piperazine-2-carboylate, respectively (Selberg et al., 2019). A 16–20% increase in m⁶A levels was detected after treating HEK293 cells for two hours with three of the four compounds (Selberg et al., 2019). Moreover, these activators demonstrated that enhanced methylation stimulated production of HIV-1 particles (Selberg et al., 2021). Computational molecular docking and other in silico approaches have suggested that the drugs nilotinib and lumacaftor (VX-809) might be potential inhibitors of METTL16; however, no empirical evidence of their inhibition of METTL16 activity has been reported yet (Mitra et al., 2023). Another novel design is the incorporation of a temporal photocage modality via conjugation to the METTL3 activator, methyl piperidine-3-carboxylate (Figure 3; Lan et al., 2021). This photocage is removed upon exposure to light, thereby releasing activator and enhancing the levels of methylation at a specific time (Lan et al., 2021). Activators for other MTases, such as METTL16 decreasing intracellular SAM concentrations, may be possible and their use as possible therapeutics is exciting. Integrating the dynamics of the m⁶A methylome in the context of specific diseases and drug responses is complex; therefore, resources like m⁶A-centered regulations of disease development and drug response database (M6AREG), which is a database curating m⁶A regulators, their RNA targets, and drugs associated with different disease states, may be useful (S. Liu et al., 2023).

Lastly, it is also important to note that macromolecular options for manipulating methylation sites, rather than MTases, are beginning to emerge. One prime example is a Cas13b-METTL14 fusion protein that utilizes a guide RNA (gRNA) to target a particular DRACH motif upon endogenous METTL3 binding to the fusion protein and methylating the target site (M. Gao et al., 2022). Other clustered regularly interspaced short palindromic repeats (CRISPR)-associated protein 9/Cas 13-based systems that enable site-specific methylation exist for demethylation, which similarly uses a fusion protein system except the MTase is replaced with either the ALKBH5 or FTO demethylase (Chang et al., 2022; J. Li, Chen, et al., 2020; Wilson et al., 2020). Clever temporal-controlled strategies have emerged, such as catalytically inactive mutant of Cas13b (dCas13b) in combination with abscisic acid-induced proximity technologies. Here, a gRNA directs dCas13b fused to a pyrabactin-resistance like protein (PYL) to a particular RNA site. Then, the addition of abscisic acid enables PYL to interact with an abscisic acid insensitive domain fused to METTL3, facilitating temporal control of m⁶A sites through the application of abscisic acid (Shi et al., 2022). We recommend the review paper by X. Sun et al. (2022) to learn more about m⁶A manipulation via CRISPR-Cas methods. The CRISPR-Cas tools can be delivered via engineered viruses for gene therapy, revealing the potential for these tools to be used for therapeutic purposes in humans (Z. Xia, Tang, et al., 2021).

7 ∣. 3D STRUCTURES OF THE HUMAN m⁶A MTases AND THEIR SUBSTRATE RECOGNITION

High-resolution 3D structures have provided insights into the functions of MTases. All known human m⁶A MTases have at least one X-ray crystal structure of the MTase domain in complex with either SAM or SAH (Figure 4 and Table 3). The first of the human m⁶A MTase structures to be solved was the MTase domains of the METTL3/METTL14 complex, in apo form, bound to SAM, and bound to SAH (Šledź & Jinek, 2016; P. Wang, Feng, et al., 2016; X. Wang, Feng, et al., 2016). Since then, slight variations of the METTL3/METTL14 MTase domains have been solved, as well as multiple structures with SAM-analog inhibitors (see Table 3). The next structure solved was the MTase domain of METTL16 in its apo form, bound to SAH (Ruszkowska et al., 2018), and shortly after, bound to a minimal MAT2A hp substrate (Doxtader et al., 2018). More recently, the crystal structure of the vertebrate conserved regions (VCR) in the C-terminal domain was solved, lending a more complete structural view of METTL16 (Aoyama et al., 2020). Next, the structure of the METTL5/TRMT112 complex bound to SAM was solved (Van Tran et al., 2019). Finally, the MTase domain of ZCCHC4 bound to SAH, along with its three zinc-finger domains, was solved (Ren et al., 2019). All five known human m⁶A MTases are a part of the Class I SAM-dependent MTases, which contain a Rossmann fold and utilize SAM as the methyl donor. In this section, a comparative structural analysis is presented and the unique structural features for each of the human m⁶A RNA MTases are highlighted.

FIGURE 4.

Domain organization, representative X-ray crystal structure, and Rossmann fold schematic for each human m⁶A MTase. On the left side of figure, MTase domains (orange/pink) contain the Rossmann fold, NLS indicates nuclear localization signal, RGG indicates arginine-glycine repeats, VCR indicates vertebrate conserved regions, ZnF indicates zinc-finger domains, and GRF, C2H2, and CCHC are specific types of zinc-finger domains. The MTase domain of METTL14 is catalytically inactive (pink), however, it still contains a Rossmann fold. Zinc ions are shown as green spheres. All structures are oriented so that the Rossman fold is displayed to match the same orientation. Displayed on the right side of figure are schematics, which are not drawn to scale, depicting the Class Ia and Class Ig Rossmann folds. Arrows and rectangles represent β-strands and helices, respectively, with colored ones showing the Rossmann fold and gray ones showing other known secondary structures outside of the Rossmann fold. Green circles represent zinc ions bound to the zinc-finger domains. Dashed lines represent unknown structures or disordered regions. The blue “N” and red “C” respectively denote the N- and C-termini while the parenthetical superscript denotes residue number. m⁶A, N⁶-methyladenosine; METTL14, methyltransferase-like protein 14; MTase, methyltransferase.

7.1 ∣. Rossmann fold

Rao and Rossmann first observed the Rossmann fold, a well conserved “super-secondary structure” found within nucleoside-binding proteins, 50 years ago (Rao & Rossmann, 1973). Most MTases, including all known human m⁶A RNA MTases, adopt the Rossmann fold composed of at least six β-strands sandwiched by α-helices (Figure 4). X-ray crystal structures of the human m⁶A MTases exhibit structural similarities, but also a few unique features in their Rossmann folds and in their SAM/SAH-binding pockets (Table 3 and Figures 4 and 5). Though all known human m⁶A RNA MTases contain a Rossmann fold, the connectivity of the β-strands and α-helices vary among the five structurally characterized human m⁶A RNA MTases (Figure 4). METTL16, METTL5, and ZCCHC4 have a Class Ia Rossmann fold, where β-strands interact as 3214576 and all β-strands are unidirectional except β7, while METTL3 and METTL14 contain a Class Ig Rossmann fold, where β-strands interact as 1762354 and all β-strands are unidirectional except β5 (Figure 4; Gana et al., 2013). Throughout the rest of this review, β-strand nomenclature follows the numbering system of the Class Ia Rossmann fold for simplicity (i.e., 3214576). Among all the Rossmann folds of amino-MTases, there are nine conserved sequence motifs (Motifs I–VIII and X; Malone et al., 1995). These nine motifs occur within the human m⁶A MTases, albeit to varying degrees of conservation (Figure 5a). Each of the nine sequence motifs and their occurrence in human m⁶A RNA MTases are described below.

FIGURE 5.

Sequence alignment of five of the conserved Rossmann fold motifs and visualization of motifs within SAM-binding pocket of the human m⁶A MTases. (a) Gray-shaded boxes show residues that are structurally conserved between the m⁶A MTases with consensus sequence above the sequences and the structural elements below. As defined in Figure 4, arrows and rectangles represent β-strands and helices, respectively. Structural elements are labeled based on a Class Ia Rossmann fold. (b) Schematics of SAM-binding pockets showing relative location and residues of Motifs I, II, III, IV, and X. Red-dashed lines denote hydrogen bonds (i.e., side chain and/or backbone interactions) and red arcs denote van der Waals interactions between SAM and specific amino acids. Schematics are based on the following X-ray crystal structures: METTL3•SAM (PDB ID: 5IL1), METTL16•SAH (PDB ID: 6B92), METTL5•SAM (PDB ID: 6H2U), and ZCCHC4•SAH (PDB ID: 6UCA). Water molecules are shown as blue circles. Water molecules were not modeled in X-ray crystal structure of ZCCHC4•SAH. m⁶A, N⁶-methyladenosine; MTase, methyltransferase; METTL3, methyltransferase-like protein 3; METTL5, methyltransferase-like protein 5; METTL14, methyltransferase-like protein 14; METTL16, methyltransferase-like protein 16; PDB ID, Protein Data Bank identifier; SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; ZCCHC4, zinc-finger CCHC-domain-containing protein 4.

7.2 ∣. Motif I

Motif I is located in the β1-loop-αA region, whereby β1 terminates with a conserved hydrophobic amino acid connected to a GxGxG (or GxG) sequence in the loop (Figures 4 and 5a). The hydrophobic amino acid (leucine in METTL3, METTL14, and METTL5; isoleucine in METTL16; and valine in ZCCHC4) packs against αC and αD within the hydrophobic core of the protein via van der Waals interactions. The GxGxG sequence canonically interacts with the carboxylic acid of the methionine unit of SAM/SAH either directly and/or via water molecule(s) (Figure 5b). Of the five human m⁶A MTases, only METTL5 (GCGCG) has the full GxGxG sequence while METTL16 (GTGAS) is missing the final glycine. Though METTL3 and ZCCHC4 interact with the carboxylic acid of SAM/SAH with the amino acids following β1, their sequences contain only one glycine (GRPHN and GTPRL, respectively) and the second glycine is replaced with a less conserved proline. METTL14, which has not been shown to bind to SAM/SAH, also contains only one glycine (GRDST). It is unlikely that the missing glycine is responsible for the inability of METTL14 to bind SAM/SAH, as other m⁶A MTases have been successful catalysts without the entire GxGxG sequence. Beyond interacting with the carboxylic acid of SAM/SAH, the GxGxG sequence sometimes interacts with the amino moiety of the methionine unit of SAM/SAH (METTL3, METTL5, and METTL16) and the 2′- and 3′-hydroxyls of the SAM/SAH ribose (METTL3; Figure 5b).

7.3 ∣. Motif II

Motif II resides in the β2-loop region (Figures 4 and 5a). The loop begins with an acidic amino acid followed by a hydrophobic amino acid (Figure 5a). The negative charge on the acidic amino acid interacts with both the 2′- and 3′-hydroxyls of the SAM/SAH ribose while the hydrophobic amino acid interacts with the SAM/SAH adenine base via van der Waals interactions (Figure 5b). The four catalytically active human m⁶A MTases exhibit the canonical interactions: METTL5 and ZCCHC4 interact with the ribose hydroxyls using an aspartate (D81 and D225, respectively) and METTL16 uses a glutamate (E133) (Figure 5b). In contrast, METTL3 uses a glutamine (Q550) at this position, and its side chain still seems capable of forming hydrogen bonds with both the 2′- and 3′-hydroxyls (2.2 and 3.2 Å, respectively; Protein Data Bank identifier [PDB ID]: 5IL1) (Figure 5b). METTL14 contains a threonine (T363), which may partially explain why METTL14 does not bind to SAM/SAH (Figure 5a). The hydrophobic amino acid is isoleucine in METTL5 and ZCCHC4 (I82 and I226, respectively), valine in METTL16 (V134), and leucine in METTL3 and METTL14 (L551 and L364, respectively). However, in METTL3, the amino acid that has van der Waals interactions with the SAM/SAH adenosine base is not the leucine (L551) but rather an asparagine (N549) that precedes the glutamine (Q550; Figure 5b).

7.4 ∣. Motif III

Motif III corresponds to β3 and includes two conserved amino acids (Figures 4 and 5a). The first is an amino acid with a side chain oxygen that forms a hydrogen bond with one of the N⁶ hydrogens of the SAM/SAH adenosine (Figure 5b). METTL3, METTL14, and METTL5 contain an aspartate (D377, D173, and D108, respectively), METTL16 contains a threonine (T164, which interacts via main chain oxygen rather than side chain), and ZCCHC4 contains an asparagine (N243; Figure 5). Immediately after the oxygen-containing amino acid is a hydrophobic amino acid whose side chain interacts with the hydrophobic core of the Rossmann fold while the backbone nitrogen forms a hydrogen bond to SAM/SAH at the N¹ position (Figure 5). METTL3 and METTL14 contain an isoleucine (I378 and I174, respectively), METTL16 contains a leucine (L165), METTL5 contains a valine (V109), and ZCCHC4 contains a methionine (M244, Figure 5). Most of the backbone nitrogens directly interact with N¹, but the X-ray crystal structure of METTL16 bound to SAH reveals that this interaction is indirect through a water molecule (Figure 5b). Though sequence alignment alone cannot identify Motif III, the amino acid identities and functions are supported by the crystal structures of the individual MTases, except for METTL14, which has not been shown to bind to SAM/SAH (Figure 5).

7.5 ∣. Motif IV

Class I SAM-dependent MTases have a highly conserved catalytic core of (D/N/S)PP(F/W/Y) following β4; this catalytic core constitutes Motif IV (Figures 4 and 5a). It is thought the side-chain oxygen in the D/N/S forms a hydrogen bond with one of the amino hydrogens at the N⁶ position of the methyl-accepting adenosine while the backbone carbonyl between the two prolines forms a hydrogen bond with the other amino hydrogen (Figures 2a and 5b). These hydrogen bonds are directly contributing to the catalysis of methyl transfer, as one of these interactions results in the deprotonation of the adenosine acceptor at the N⁶ position, which is necessary for methyl transfer to occur. To date, the amino acid responsible for the deprotonation of the N⁶-amino is unknown and requires further investigation; however, it is thought to be the D/N/S residue (Bheemanaik et al., 2006). Based on X-ray crystal structures of METTL16 bound to MAT2A hairpin RNAs (PDB ID: 6DU4, 6DU5), the backbone carbonyl between the two prolines is the only electrophile close enough to deprotonate N⁶, while the asparagine interacts with N⁶ indirectly through a water molecule. However, the METTL16•MAT2A crystal structures lack SAM/SAH in the active site; therefore, the active site of the ternary complex may still undergo further remodeling that would move the asparagine into position to form a hydrogen bond with the N⁶ acceptor. The aromatic F/W/Y amino acid interacts with the methyl-accepting adenosine via parallel π–π interactions (i.e., π-stacking) to properly orient the adenosine for methyl transfer. Although all known human m⁶A MTases contain the central PP, they utilize different amino acids as the N⁶ hydrogen bond acceptor and aromatic amino acids for stacking (Figure 5a). ZCCHC4 and METTL3 use aspartate (D276 and D395, respectively) while METTL16 and METTL5 use asparagine (N184 and N126, respectively). For the aromatic amino acid, METTL16, METTL5, and ZCCHC4 use a phenylalanine (F187, F129, and F279, respectively) and METTL3 uses a tryptophan (W398). METTL14 has an inactive EPPL catalytic core. Though a glutamate (E192) has hydrogen bond acceptor capabilities, no known catalytically active MTase contains a glutamate at this position in motif IV. Furthermore, leucine (L195) is unable to participate in π–π interactions. Overall, even if METTL14 does bind to SAM, it is unlikely that it has any catalytic activity. In addition to catalysis, the crystal structures of the human m⁶A MTases in complex with SAM/SAH reveal that the D/N side chain forms a hydrogen bond with the amino moiety of the methionine unit of SAM/SAH in the absence of its RNA substrate (Figure 5b). To date, METTL16 is the only m⁶A MTase with a solved 3D structure in complex with an RNA substrate and these structures lack SAM/SAH in the binding pocket. The field would benefit from 3D structures of the other m⁶A MTases in their binary (E•RNA) and ternary (E•RNA•SAM) states.

7.6 ∣. Motifs V–VIII

Although Motifs I–IV are conserved in their active site interactions, Motifs V–VIII are much less conserved among all m⁶A MTases. This lack of sequence conservation is perhaps expected because these motifs, though structurally similar, have different functions among the MTases. Structurally, Motifs V–VIII include αD, β5, αE, β6, β7, and all loops linking these secondary structures (Figure 4). Characterized functions of these motifs include dimerization interface in METTL3/METTL14 and METTL5/TRMT112, interactions with other intramolecular features of the protein such as the C-terminal zinc-finger domains in ZCCHC4, and RNA recognition by METTL16. These unique structural features are discussed in more detail (see Section 7.8).

7.7 ∣. Motif X

Motif X spans αZ and the loop preceding it (Figure 4). Motif X forms the last side of the SAM/SAH binding pocket by forming a hydrogen bond to the methionine carboxyl- and/or amino-groups in SAM/SAH (Figure 5b). Each m⁶A MTase has a unique mechanism for interacting with the aforementioned moieties of SAM/SAH, therefore, little sequence conservation among m⁶A MTases is observed in motif X (Figure 5). To interact with the methionine carboxyl, METTL5 and ZCCHC4 use the backbone amine of T31 and F175, respectively, as a hydrogen bond donor; METTL3 indirectly uses the backbone amine of K510 and the side chain of S508 through a water molecule; and METTL16 indirectly uses the side chain of R82 and the backbone carbonyl of L75 via two water molecules (Figure 5b). Although METTL3 and METTL16 make no interactions with the methionine amino-group within Motif X, METTL5 uses the side chain of T31 as the hydrogen bond acceptor and ZCCHC4 uses the backbone carbonyl of Y173 (Figure 5b). Because of the variations of Motif X within the catalytically active MTases, it is difficult to conclude whether this area of METTL14 is responsible for its inability to bind SAM/SAH (Figure 5a).

7.8 ∣. Structural features unique to each MTase

The MTase domains of the METTL3/METTL14 complex were crystallized and published by multiple labs within 4 months of each other (Table 3; Šledź & Jinek, 2016; P. Wang, Feng, et al., 2016; X. Wang, Feng, et al., 2016). All 3D structures showed a “pseudo-symmetric heterodimer,” where the two Rossmann folds interact through an extensive dimerization interface (over 2000 Ų) from αC to αD (Figure 4). Though it was originally thought that METTL14 was the active MTase in the METTL3/METTL14 complex (J. Liu et al., 2014), structural data and further biochemical analysis established METTL3 as the active MTase (Šledź & Jinek, 2016; P. Wang, Feng, et al., 2016; X. Wang, Feng, et al., 2016). Not only does METTL3 have the conserved catalytic core observed in MTases (i.e., Motif IV) while METTL14 does not, only the METTL3 MTase domain contains electron density for SAM/SAH (Šledź & Jinek, 2016; P. Wang, Feng, et al., 2016; X. Wang, Feng, et al., 2016). Furthermore, modeling a nucleic acid substrate into these catalytic cores reveals that METTL14 lacks the positively charged surface necessary to bind the negatively charged phosphate backbone (Doxtader et al., 2018; Šledź & Jinek, 2016; P. Wang, Feng, et al., 2016). However, biochemical assays show that METTL3 alone is insufficient for methylation, as METTL14 is necessary for METTL3 stability and, therefore, catalysis (Šledź & Jinek, 2016; P. Wang, Feng, et al., 2016). The minimized MTase domains of the METTL3/METTL14 complex used for crystallization of the complex are also inactive, though the addition of the METTL3 zinc-finger domains, which are adjacent to the N-terminus of the MTase domain (Figure 4), restores activity to that of the wildtype METTL3/METTL14 complex (J. Huang et al., 2019; P. Wang, Feng, et al., 2016). Together, the two METTL3 zinc-finger domains, whose structure was solved using NMR (Table 3), are necessary and sufficient to bind GGACU-containing RNA (J. Huang et al., 2019). How the zinc-finger and MTase domains work together to specifically bind and methylate RNA is an exciting structural question that remains unanswered. Recently, a 4.4 Å-resolution structure of the full-length METTL3/METTL14 complex, along with other proteins (i.e., HAKAI, WTAP, VIRMA, and ZC3H13) in the holoenzyme, was solved using cryogenic electron microscopy (cryo-EM; S. Su, Li, et al., 2022). Using this model, protein–protein crosslinking mass spectrometry, and AI-structure characterization, the METTL3/METTL14 complex appears to interact with the rest of the holoenzyme through the N-terminal region of METTL3, including the zinc-finger domains, and the N-terminal region of METTL14 (S. Su, Li, et al., 2022; Yan et al., 2022). These studies reveal the power of integrative structural biology, which is likely to uncover more exciting structure–function relationships for all human m⁶A RNA MTases.

In METTL16, the N-terminal domain (amino acids 1–288) is conserved from E. coli to humans but the C-terminal domain (amino acids 307–562) is conserved only through vertebrates at the sequence level (Sergiev et al., 2008; Figure 4 and Table 1). The N-terminal domain consists of the highly conserved Rossmann fold, a structured N-terminal region that is needed for RNA binding (amino acids 1–77) (Doxtader et al., 2018; Mendel et al., 2018), a 3₁₀ helix preceding β1, and a structured loop following β4 (Figure 4). The C-terminal domain consists of VCR1 and VCR2 separated by a predicted disordered region (Figure 4). The two VCRs fold into a β-sheet with five β-strands (15432), which is structurally and functionally similar to the kinase-associated 1 (KA1) domain of the U6 snRNA-specific terminal uridylyl transferase 1 (Aoyama et al., 2020; Figure 4). Interestingly, AlphaFold indicates that the C-terminal domain of the C. elegans homolog METT-10 folds into a KA1-like domain, suggesting the structure and function of the C-terminal domain of METTL16 is more widely conserved than the sequence (Ju et al., 2023). The VCRs enhance the MTase activity and binding affinity of METTL16 to U6 snRNA by up to ~100-fold compared to the MTase domain alone (Aoyama et al., 2020; Breger & Brown, 2023), while the VCRs are not required for methylation of the MAT2A mRNA hairpins but do improve RNA binding by ~30-fold (Aoyama et al., 2020; Pendleton et al., 2017). Unfortunately, there is no structure of the METTL16 MTase domain and VCRs together. A structure of the full-length METTL16 protein, both apo and RNA bound, will likely lead to a better understanding of how the VCRs differentially contribute to catalysis of different RNA substrates. Unlike the other human m⁶A MTases, a structure of METTL16 has been solved in complex with an RNA substrate: the minimal MAT2A mRNA hairpin (Doxtader et al., 2018; Table 3). Interestingly, the MTase domain of the METTL16•RNA bound complex is extraordinarily similar to that of the apo and SAH-bound structures (root mean square deviation of α-carbons is ~0.3 Å among the structures). However, the major difference among these structures is that the apo and SAH-bound structures lack electron density for amino acids (188–214) between β4 and αD while the RNA-bound structure shows this region interacting with the RNA, suggesting a conformational change or stabilization (Figure 4). The RNA-bound structure also revealed that K163, a residue located in the K-loop following β3, blocks the SAM binding site in the presence of RNA but is solvent exposed in the absence of RNA (Doxtader et al., 2018; Ruszkowska et al., 2018). The METTL16 K163A mutant increases the binding affinity to SAM and increases METTL16 catalytic activity on MAT2A hp1 by approximately sixfold (Doxtader et al., 2018). METTL16 may have evolved to prefer the weaker binding and methylation activity due to its role in SAM homeostasis. It will be interesting to see how these dynamic regions of METTL16 as well as the C-terminal domain interacts with its other RNA-binding partners.

Composed of only an MTase domain, METTL5 is the simplest validated human m⁶A MTase (Figure 4). Within its MTase domain, METTL5 has the well-conserved Rossmann fold, a short (25 amino acid) N-terminal structure containing two α-helices, and a structured loop following β4 (Figure 4). However, knowing the sequence and structure of METTL5 alone is not sufficient to understand its activity, as it was shown to be correctly folded and functional only when co-purified with TRMT112, which interacts via a highly hydrophobic surface of METTL5 (1160 Ų) (Van Tran et al., 2019; Figure 4). Though TRMT112 is not known to be necessary for the activity of any other human m⁶A MTase, TRMT112 is necessary for the functioning of at least seven other human RNA and protein MTases (Brūmele et al., 2021), including metastasis-related methyltransferase 1 (originally WBSCR22 due to it being deleted in Williams-Beuren syndrome), which methylates rRNA G1575 to N⁷-methylguanosine (Zorbas et al., 2015); AlkB homolog 8, tRNA methyltransferase, which methylates tRNA at the wobble position to 5-methylcarboxymethyluridine (D. Fu et al., 2010); and lysine methyltransferase 9 (originally N6AMT1), which methylates histone H4 lysine 12 (Figaro et al., 2008; W. Li et al., 2019; Metzger et al., 2019).

ZCCHC4 is an evolutionarily conserved MTase within multicellular organisms (H. Ma et al., 2019; Table 1). The Rossmann fold of ZCCHC4 is centralized within three different canonical zinc-finger domains, which collectively bind to six zinc ions (Ren et al., 2019; Figure 4). The two N-terminal zinc-finger domains, GRF- and C2H2-type zinc-fingers (named after the conserved residues), are a part of the putative RNA-binding site (Ren et al., 2019). The C-terminus contains a CCHC-type zinc-finger domain, which is found in proteins that bind to single-stranded nucleic acids. All three zinc-finger domains, in addition to the MTase domain, are necessary for ZCCHC4 to methylate its 28S rRNA substrate. Within the Rossmann fold is a unique auto-inhibitory loop, known as the regulatory loop, that interacts with the cofactor loop so that apo ZCCHC4 adopts a fully closed, inactive conformation (Ren et al., 2019). Although the trigger to activate ZCCHC4 is unknown, Ren et al. propose conformational changes induced by RNA and/or SAM binding. It will be interesting to see how the zinc-finger domains and regulatory loop interact with the RNA to aid in the methylation of the 28S rRNA.

Overall, though some 3D structures are available for each of the human m⁶A RNA MTases, structural models for all steps in the kinetic pathway (Figure 2b) have not been reported for each of the MTases. Both METTL5 and ZCCHC4 lack structures without SAM/SAH bound in the active site, though we suspect there will be little difference between the apo and SAM/SAH bound structures, as both METTL3 and METTL16 show little differences between apo and SAM/SAH-bound structures. Conversely, the METTL16•MAT2A RNA hp complex does suggest conformational changes between apo and RNA-bound structures. Therefore, the MTases in complex with their RNA substrates is one knowledge gap that needs to be filled. Furthermore, high-resolution structural models of the full-length METTL3/METTL14 complex and of full-length METTL16 are not available. Perhaps the use of cryo-EM, which does not depend on crystallization and is able to provide atomic resolution of large macromolecules, will lead to a better understanding of the functional interplay between the MTase domain, other regions of the MTases, and protein–protein interactions. In addition, AlphaFold2 and the like may provide early insights (Yan et al., 2022). We look forward to learning more about the structures of MTases and the MTase•RNA complexes.

8 ∣. HYPERMODIFIED m⁶A-CONTAINING RIBONUCLEOTIDE VARIANTS IN HUMANS

Besides m⁶A, there are other known naturally occurring adenosine modifications that contain an N⁶-methyl group, including m⁶Am, N⁶,N⁶-dimethyladenosine (m^6,6A), N⁶,N⁶,2′-O-trimethyladenosine, and 2-methylthio-N⁶-methyladenosine. To date, only m⁶Am and m^6,6A have been identified in humans. These two hypermodifications will be discussed along with their related MTases, providing a brief comparative analysis with the five m⁶A MTases.

There are two known human SAM-dependent MTases that catalyze the methyl transfer to the N⁶ position of 2’-O-methyladenosine (Am) to form m⁶Am: METTL4 and PCIF1, also known as cap-specific adenosine methyltransferase. METTL4 is conserved from yeast to humans. The C. elegans and mouse homologs of METTL4 methylate DNA to make 6mA marks; however, human METTL4 has shown no evidence for DNA methylation, but instead methylates the N⁶ position of an internal Am (but not A) in U2 snRNA (H. Chen, Gu, et al., 2020; Goh et al., 2020). Early evidence suggests that the m⁶Am mark in U2 snRNA plays a role in differential splicing of select mRNAs, though further studies are needed to understand the mechanism (H. Chen, Gu, et al., 2020; Goh et al., 2020). Sequence conservation and AlphaFold predict human METTL4 to adopt a Class Ig Rossmann fold with the conserved DPPW catalytic core, similar to METTL3 (Jumper et al., 2021). Using X-ray crystallography, a 3D structure of Arabidopsis METTL4•Am-containing RNA•SAH complex has been solved, revealing that a Phe residue conserved in METTL4 creates a more hydrophobic binding pocket to favor binding a 2′-O-methyl rather than 2′-OH (Luo et al., 2022). PCIF1 catalyzes the N⁶-methylation of Am to m⁶Am at the first position after the 5’-N⁷-methylguanosine (m⁷G) cap. The m⁷G cap alone (also known as cap 0) is necessary for eukaryotic translation, whereas higher eukaryotes also add a N⁷-methylguanosine followed by a 2′-O-methylnucleotide (Cap 1) to distinguish self from non-self mRNA (Ramanathan et al., 2016). Sometimes, when Cap 1 contains an adenosine (m⁷G-Am), PCIF1 will methylate Am at the N⁶-position (m⁷G-m⁶Am; Ramanathan et al., 2016). Though the function of the m⁷G-m⁶Am cap is not fully understood, it may have an effect on translation, mRNA stability, and stress pathways (Akichika et al., 2019; Boulias et al., 2019; Sendinc et al., 2019; Sikorski et al., 2020). Like most of the human m⁶A MTases, PCIF1 adopts a Class 1a Rossmann fold and has the conserved NPPF catalytic core (Akichika et al., 2019; PDB ID: 6IRV, 6IRW). The crystal structure also reveals that PCIF1 has a small, positively charged groove that can accommodate the 5′ cap, which contributes to its substrate specificity (Akichika et al., 2019). Readers can learn more about m⁶Am from this recent review (Cesaro et al., 2023).

There are two known human m^6,6A MTases: dimethyladenosine transferase 1 (DIMT1) and the mitochondrial transcription factor B1 (TFB1M, also mtTFB1). DIMT1 is thought to place the m^6,6A mark on two neighboring adenosines, A1850 and A1851, in the 18S rRNA, analogous to the yeast homolog Dim1 (Zorbas et al., 2015). DIMT1 has catalytic activity in vitro on a short, synthetic 12-nt hairpin having the same sequence and secondary structure as the predicted m^6,6A sites in the 18S rRNA (H. Shen et al., 2020). However, the methylation order of the two adenosines and the binding order of the substrates (RNA and SAM) remains unknown. The MTase activity of DIMT1 is necessary for efficient protein translation (H. Shen et al., 2020). In addition to MTase activity, DIMT1 is a critical scaffold for 40S rRNA assembly, as heterologous expression of a catalytically inactive mutant in a DIMT1^+/− heterozygous knockdown cell line (HEK293T) restores the abundance of 40S rRNA to normal levels (H. Shen et al., 2020). Three crystal structures of full-length human DIMT1 (PDB ID: 1ZQ9, 6W6C, 6W6F) have been solved (H. Shen et al., 2020) and show that DIMT1 also contains a Class Ia Rossmann fold but with a noncanonical NLPY catalytic core. TFB1M also catalyzes m^6,6A marks in rRNA, specifically on A936 and A937 in helix 45 of mitochondrial 12S rRNA (McCulloch et al., 2002; Metodiev et al., 2009; Seidel-Rogol et al., 2003). Knockout of Tfb1m in mice leads to a destabilization of 12S rRNA and eliminates translation in the mitochondria (Metodiev et al., 2009). Two crystal structures of TFB1M have been solved (PDB ID: 6AAX, 6AJK), showing a Class Ia Rossmann fold with a noncanonical NLPF (141–144) catalytic core in complex with a minimized hairpin substrate (X. Liu, Shen, et al., 2019). More recently, cryo-EM structures showing the formation of the mature mitochondrial small ribosomal subunit shows TFB1M binding to its rRNA substrate (PDB ID: 8CSP, 8CSQ, 8CSR, 8CSU) and the dimethylated product without TFB1M bound (PDB ID: 8CSS, 8CST; Burnside et al., 2023). Both the crystallographic and cryo-EM models show the second adenosine in the MTase active site but not any methylation when bound to TFB1M, so the adenosine methylation order is still not fully understood. It will be interesting to discover how these m^6,6A MTases can methylate two adjacent sites with such high specificity.

9 ∣. CONCLUSION

Within the past 7 years, three more proteins (i.e., METTL16, METTL5, and ZCCHC4) have been identified as human m⁶A MTases, yet many exciting discoveries remain such as a deeper understanding of their biological functions, their roles in rare and neglected diseases, catalytic mechanisms, 3D structures of binary and ternary complexes and perhaps the identification of more m⁶A RNA MTases. Even E. coli has a large and dynamic methylome with m⁶A marks installed in mRNA by an unknown m⁶A MTase (X. Deng et al., 2015). With about 200 SAM-dependent MTases encoded in the human genome, one wonders: are there other MTases that methylate adenosine to produce m⁶A (Petrossian & Clarke, 2011)? A new twist on RNA methylation is the existence of ribozymes that catalyze methyl transfer reactions, a finding that begs the question: is there a naturally occurring counterpart (Scheitl et al., 2020)? There is a SAM-dependent m⁷G MTase ribozyme that was created by in vitro selection and similar ribozymes appear to exist in organisms from bacteria to mammals (H. Jiang, Gao, et al., 2021). Recently the first natural RNA MTase was discovered and aptly named methyltransferase ribozyme 1, which utilizes O⁶-methylguanosine as methyl donor to create N¹-methyladenosine as well as a riboswitch that can self-methylate to install 3-methylcytidine using prequeuosine (preQ₁) as a methyl donor (J. Deng, Wilson, et al., 2022; Flemmich et al., 2021; Hiller & Strobel, 2022; Scheitl et al., 2022). These recent findings may hint at a number of new RNA ribozymes waiting to be discovered and elucidated.

Among the known human m⁶A MTases, several key unknowns remain, such as their druggability, the structural basis of their substrate specificity, how m⁶A RNA MTases impact phase separation (D. Han, Longhini, et al., 2022; Ries et al., 2019), and the physiological relevance of m⁶A RNA MTases methylating DNA to form 6mA marks. Substrate specificity extends beyond the MTase domain; therefore, more studies will be required to elucidate the unique structural elements that diversify and differentiate these MTases, such as the disordered region in METTL16 or protein-binding partners. Of course, the structure and function of m⁶A writers must be understood in the context of an m⁶A methylome that is also regulated by m⁶A readers and possibly erasers. Therefore, having an improved quantitative view of the m⁶A methylome is paramount, emphasizing the need for direct, transcriptome-wide readout of m⁶A marks at single-nucleotide resolution (Alfonzo et al., 2021). By filling-in the aforementioned knowledge gaps, m⁶A biology and its related MTases will penetrate into new realms of cellular importance and human health.

Abbreviations:

5mC

5-methylcytidine

6mA

N⁶-methyladenosine (for DNA)

A-431

epidermoid carcinoma cell line

ALKBH5

AlkB homolog 5, RNA demethylase

Am

2′-O-methyladenosine

AML

acute myeloid leukemia

APOBEC1

apolipoprotein B mRNA editing enzyme, catalytic polypeptide 1

B16

mouse melanoma cell line

Cas9

CRISPR-associated protein 9

CFI_m

cleavage factor Im

CRISPR

clustered regularly interspaced short palindromic repeats

cryo-EM

cryogenic electron microscopy

DART-seq

deamination adjacent to RNA modification targets

dCas

catalytically inactive mutant of Cas

DIMT1

dimethyladenosine transferase 1

dRNA-seq

direct RNA sequencing

EBV

Epstein–Barr virus

eEF

eukaryotic translation elongation factor

eIF

eukaryotic translation initiation factor

ERK

extracellular signal-regulated kinases

EV71

enterovirus 71

FTO

fat mass and obesity-associated protein

gRNA

guide RNA

HAKAI

E3 ubiquitin-protein ligase Hakai

HAP1

near-haploid human cell line

HBV/HCV

hepatitis B and C viruses

HCT116

human colon cancer cell line

HEK293T

human embryonic kidney 293 cells with mutant SV40 large T antigen

HepG2

human liver cancer cell line

HIV-1

human immunodeficiency virus 1

hMPV

human metapneumovirus

hp

hairpin

HSV-1

herpes simplex virus-1

IC₅₀

half-maximal inhibitory concentration

J1 ES

mouse embryonic cell line

KA1

Kinase-associated 1 domain

k _cat

turnover number

K _D

equilibrium dissociation constant

K _M

Michaelis–Menten constant

KSHV

Kaposi's sarcoma-associated herpes virus

LARP7

La ribonucleoprotein 7, transcriptional regulator

lncRNA

long noncoding RNA

m^6,6A

N⁶,N⁶-dimethyladenosine

m⁶A

N⁶-methyladenosine (for RNA)

m⁶ACE-seq

m⁶A-crosslinking-exonuclease-sequencing

m⁶Am

N⁶,2′-O-dimethyladenosine

M6AREG

m⁶A-centered regulations of disease development and drug response database

m⁷G

N⁷-methylguanosine

m⁷G-m⁶Am

N⁷-methylguanosine followed by N⁶,2^′-O-dimethyladenosine

MALAT1

metastasis-associated lung adenocarcinoma transcript 1

MAT2A

methionine adenosyl transferase 2A

MazF

endoribonuclease toxin MazF

meCLICK-seq

methylation CLICK degradation-sequencing

MEF

mouse embryonic fibroblasts

MePCE

methylphosphate capping enzyme

MeRIP-seq

methylated RNA immunoprecipitation-sequencing

Merm1

metastasis-related methyltransferase 1

mESC

mouse embryonic stem cell

METTL

methyltransferase-like protein

miCLIP

methylation (or m⁶A)-individual-nucleotide resolution crosslinking and immunoprecipitation

miRNA

micro RNA

MjDim1

Methanocaldococcus jannaschii dimethyltransferase 1

MOLM-13

human acute myeloid leukemia cell line

MRE11

MRE11 homolog, double-strand break repair nuclease

mRNA

messenger RNA

MTase

methyltransferase

ncRNA

noncoding RNA

NIH 3T3

mouse fibroblast cell line

NS

nonstructural protein

nt

nucleotide

PCIF1

phosphorylated CTD interacting factor 1

PDB ID

Protein Data Bank identifier

PEDV

porcine epidemic diarrhea virus

PYL

pyrabactin resistance like

RBM

RNA-binding motif

RdRp

RNA-dependent RNA polymerase

RNP

ribonucleoprotein

rRNA

ribosomal RNA

RSV

Rous sarcoma virus

SAH/AdoHcy

S-adenosyl-L-homocysteine

SAM/AdoMet

S-adenosyl-L-methionine

SARS-CoV-2

severe acute respiratory syndrome coronavirus 2

STM2457

METTL3 inhibitor

SV40

Simian virus 40

TK6

human lymphoblast cell line

TRMT112

tRNA methyltransferase activator subunit 11-2

UTR

untranslated region

UZH1a

METTL3 inhibitor

U2OS

human osteosarcoma cell line

VCR

vertebrate conserved region

VIRMA/KIAA1429

vir-like m⁶A methyltransferase associated protein

WTAP

Wilms' tumor 1-associating protein

XRN1

5′-3′ exoribonuclease 1

YTH

YT521-B homology domain

YTHDC

YTH domain-containing protein

YTHDF

YTH domain-containing family protein

ZC3H13

zinc-finger CCCH domain-containing protein 13

ZCCHC4

zinc-finger CCHC-domain-containing protein 4

ZIKV

Zika virus

ZnF

zinc-finger domain

DATA AVAILABILITY STATEMENT

Data sharing is not applicable to this article as no new data were created or analyzed in this study.