Emerging Role of m 6 A Methylome in Brain Development: Implications for Neurological Disorders and Potential Treatment
Journal: 2021/June - Developmental Biology
Abstract:
Dynamic modification of RNA affords proximal regulation of gene expression triggered by non-genomic or environmental changes. One such epitranscriptomic alteration in RNA metabolism is the installation of a methyl group on adenosine [N6-methyladenosine (m6A)] known to be the most prevalent modified state of messenger RNA (mRNA) in the mammalian cell. The methylation machinery responsible for the dynamic deposition and recognition of m6A on mRNA is composed of subunits that play specific roles, including reading, writing, and erasing of m6A marks on mRNA to influence gene expression. As a result, peculiar cellular perturbations have been linked to dysregulation of components of the mRNA methylation machinery or its cofactors. It is increasingly clear that neural tissues/cells, especially in the brain, make the most of m6A modification in maintaining normal morphology and function. Neurons in particular display dynamic distribution of m6A marks during development and in adulthood. Interestingly, such dynamic m6A patterns are responsive to external cues and experience. Specific disturbances in the neural m6A landscape lead to anomalous phenotypes, including aberrant stem/progenitor cell proliferation and differentiation, defective cell fate choices, and abnormal synaptogenesis. Such m6A-linked neural perturbations may singularly or together have implications for syndromic or non-syndromic neurological diseases, given that most RNAs in the brain are enriched with m6A tags. Here, we review the current perspectives on the m6A machinery and function, its role in brain development and possible association with brain disorders, and the prospects of applying the clustered regularly interspaced short palindromic repeats (CRISPR)-dCas13b system to obviate m6A-related neurological anomalies.
Keywords: N6-methyladenosine (m6A); clustered regularly interspaced short palindromic repeats (CRISPR)–dCas13b; cortical development; m6A editing; mRNA metabolism; mRNA methylation; neurological disorders.
Relations:
Content
Diseases
(1)
Drugs
(1)
Processes
(4)
Anatomy
(3)
Similar articles
Articles by the same authors
Discussion board

Emerging Role of m<sup>6</sup> A Methylome in Brain Development: Implications for Neurological Disorders and Potential Treatment

Introduction

Over 170 chemical modifications of RNA are known to exist in eukaryotes (Boccaletto et al., 2018). These RNA modifications, together referred to as the epitranscriptome, play essential roles in gene expression regulation via affecting RNA metabolism: RNA processing, decay, transport, and translation. N-methyladenosine (mA) is among the characterized adenosine methylations of messenger RNAs (mRNAs) (Engel and Chen, 2018) and the most occurring in mammalian cells (Roundtree et al., 2017a). The mA mRNA methylome is dynamically regulated by factors that install, remove, or bind the mA mark on mRNA. Such dynamism in the mA landscape is known to critically regulate mRNA metabolism to influence gene expression. In essence, mA modification is reported to modulate several biological events, including cell proliferation, differentiation, and embryonic development, and can also lead to disease conditions when dysregulated (Dominissini et al., 2012; Meyer et al., 2012; Ke et al., 2015; Linder et al., 2015; Zhao et al., 2017; Ries et al., 2019).

The impact of mA modification on cell biological processes is notable in nervous tissues (Widagdo et al., 2016; Li et al., 2019). This is because neural cells are known to be enriched with mA-tagged mRNAs. As a result, the developing and adult brain is reported to be ubiquitously enriched with mA modifications (Chang et al., 2017; Zhang F. et al., 2018). The mA level in the brain is temporally regulated in the course of its development such that the adult brain registers the highest level of mA (Meyer et al., 2012). The massive prevalence of mA in the developing and postnatal brain signifies the importance of mA modification in regulating brain morphogenesis and function. Indeed, a chunk of the expanding knowledge indicating the phenomenal role of mA in orchestrating neural development and function includes the proliferation of neural stem cells (NSCs) and other neural precursor cells, neuroprogenitor cell differentiation, gliogenesis, elaboration of neural processes, and synaptic transmission (reviewed in Widagdo and Anggono, 2018; Li et al., 2019; Chokkalla et al., 2020; Dermentzaki and Lotti, 2020; Livneh et al., 2020).

While maintenance of the general mA homeostasis is indispensable for proper brain development and activity, selective hypomethylation and hypermethylation of gene transcripts are critical mechanistic modalities for normal brain neurodevelopment and functionality. Moreover, the targeted binding of mA on transcription factor-encoding mRNAs and disease-risk gene transcripts in the developing brain (Zhang et al., 2020) reflects how mA signaling is critical for brain structure and function in health. It also means that disturbance of the mA RNA methylome can have implications for abnormal anatomy and physiology of the brain, leading to neurological disorders.

In this review, we present the role of the mA methylation machinery in mRNA metabolism, with discussion focused on how the mA landscape regulates brain development and function. Neurodevelopmental, neurodegenerative, and neuropsychiatric disorders of the brain attributable to mA signaling dysregulation are highlighted.

The N-Methyladenosine Modification Machinery

Early studies revealed that mA is the most substantial posttranscriptional modification in eukaryotic mRNAs (Desrosiers et al., 1974; Perry et al., 1975a, b). It was proposed that nearly 7,000 mRNAs from human and mouse transcriptome contain mA modification (Dominissini et al., 2012; Meyer et al., 2012). Currently, over 10,000 mA-modified mRNA transcripts have been identified in yeast and mammalian cells (Wang and Zhao, 2016). mA modification and recognition (binding) are achieved by three functional components of the mA machinery, namely, mA methyltransferases (“writers”), mA demethylases (“erasers”), and mA binding or interacting proteins (“readers”) (Figure 1A). Through the specific binding of mA-interacting proteins, mA mRNA methylation is able to play a central role in regulating several aspects of mRNA metabolism, such as transport, splicing, stability, translation, and phase separation (Figure 1). High-throughput data revealed that mA modification typically occurs within the consensus sequence RRACH (R stands for G or A; H stands for A, C, or U) (Dominissini et al., 2012; Meyer et al., 2012). The consensus sequence was recently redefined as DRACH, where D stands for G, A, or U (Linder et al., 2015). As depicted in Figure 1B, mA distribution along mRNA is asymmetric. In general, mA sites are concentrated in the protein coding region (CDS) near stop codons, followed by the 3′ untranslated region (UTR), and in the 5′ UTR (Figure 1B) (Dominissini et al., 2012; Meyer et al., 2012; Ke et al., 2015; Linder et al., 2015).

An external file that holds a picture, illustration, etc.
Object name is fcell-09-656849-g001.jpg

The N-methyladenosine (mA) machinery and modification of messenger RNA (mRNA). (A) An illustration showing the mA machinery. It is made of factors that can functionally be categorized as writers, erasers, and readers of mA. The mA marks are deposited by the methylation complex (writers), including METTL3, METTL14, and WTAP, which is counteracted by the mA demethylases (FTO and ALKBH5), leading to mA removal. The recognition of mA marks is done by the readers or binding proteins as indicated. (B) Diagram showing a typical mA distribution in regions of an mRNA located in the nucleus. (C) The mA readout affects mRNA fates, including trafficking, stability, decay, translation, and localization.

N-Methyladenosine Writers

The installation of mA is carried out by ∼1 MDa mA writer complex composed of the methyltransferase-like protein 3 (METTL3) and METTL14, which heterodimerize (METTL3-METTL14) to function as the enzymatic core of the writer complex (Bokar et al., 1994; Bujnicki et al., 2002; Liu et al., 2014; Iyer et al., 2016). Additionally, other factors are known to interact with the mA writer complex. These include Wilms tumor 1-associated protein (WTAP) (Ping et al., 2014), VIRMA/KIA1429 (Yue et al., 2018), RNA-binding protein 15 (RBM15) (Patil et al., 2016; Huang and Yin, 2018), ZC3H13 (Knuckles et al., 2018), and HAKAI (Yue et al., 2018) (Figure 1A). These cofactors are regulated by the binding of RNA and the catalytic activity of the enzymatic core of the mA writer complex (Bujnicki et al., 2002; Liu et al., 2014; Ping et al., 2014; Iyer et al., 2016; Yue et al., 2018).

METTL3 and METTL14

Bokar et al. (1994) partially purified the mA writer complex using an in vitro methylation system and identified MT-70, a 70-kDa sub-complex possessing S-adenosylmethionine-binding methyltransferase capacity (Bokar et al., 1994). Later, it was renamed METTL3 (Narayan and Rottman, 1988; Bokar et al., 1997). Knockout of METTL3 in cells effectively blocks mA modification of mRNAs (Zhong et al., 2008; Agarwala et al., 2012; Geula et al., 2015). On the other hand, METTL14 forms a stable heterodimer with METTL3 to form the methyltransferase core of the mA methylation machinery (Liu et al., 2014; Wang Y. et al., 2014). METTL14, however, lacks enzymatic function and instead acts as an RNA-binding scaffold to augment the enzyme activity of METTL3 by directing the location of SAM methyl group required for the reaction (Śledź and Jinek, 2016; Wang P. et al., 2016; Wang X. et al., 2016). Therefore, METTL3 is the primary enzyme responsible for mA installation on mRNA.

METTL3-METTL14-Associated Adaptors: WTAP, VIRMA (KIAA1429), RBM15/15B, ZC3H13 (KIAA0853), and HAKAI

The core mA writer complex METTL3-METTL4 interacts with other adaptor proteins. It was found that FIP37 (the plant homolog of WTAP) co-localized with MTA (Arabidopsis homolog of METTL3) in the nucleus through physical interaction (Zhong et al., 2008). Similar interaction between WTAP and METTL3 was observed in mammalian cells (Liu et al., 2014; Ping et al., 2014; Schwartz et al., 2014). WTAP is key in keeping the METTL3-METTL4 heterodimer in nuclear speckles (Ping et al., 2014). Loss of WTAP leads to the depletion of mA modification in mRNA, indicating that WTAP may orient METTL3-METTL14 onto targets (Ping et al., 2014). However, the detailed mechanism remains elusive. Of note, it was demonstrated that two classes of mA sites exist: WTAP-dependent and WTAP-independent sites (Schwartz et al., 2014). VIRMA is known to also interact with the WTAP-METTL3-METTL4 complex (Figure 1A; Schwartz et al., 2014) and indicates its essentiality for the mA writer complex functionality. Indeed, VIRMA deletion in human cells leads to a significant reduction in mRNA methylation levels, although not as intense as that resulting from WTAP knockdown (Schwartz et al., 2014). Biochemical studies from Yue et al. (2018) demonstrated that VIRMA preferentially regulates mA modification in the 3’ UTR proximal to stop codon.

Through proteomic studies, Horiuchi et al. (2013) observed that RBM15 and its paralog RBM15B, together with ZC3H13/KIAA0853, and MTA70 (METTL3), associate with WTAP (Horiuchi et al., 2013; Patil et al., 2016), which raises the possibility that RBM15 and RBM15B may also play role(s) in mA modification. Indeed, silencing of RBM15 and RBM15B led to a demonstrable decrease in mA levels of mRNA (Patil et al., 2016). Based on Individual-nucleotide resolution UV crosslinking and immunoprecipitation (iCLIP) data, it was proposed that RBM15/15B recruit the mA methylation machinery to perform mA modification through binding to uridine-rich regions near DRACH sites. That notwithstanding, it is not always the case that uridine-rich regions exist near mA sites; therefore, other methylation complex adaptors may mediate the complex binding to such variant mA sites (Patil et al., 2016; Meyer and Jaffrey, 2017).

ZC3H13/KIAA0853 is also an interactor of the mA machinery, and it is demonstrated to be crucial in linking RBM15/15B to WTAP (Horiuchi et al., 2013; Knuckles et al., 2018; Wen et al., 2018). Knockdown of ZC3H13 shifts the localization of the mA adaptors WTAP, Virilizer, and Hakai from nucleus to cytosol in embryonic stem cells and leads to a significant total reduction in mA level on mRNA (Wen et al., 2018). This reflects an essential role of ZC3H13 in the deposition of mA on mRNAs. The E3 ubiquitin ligase HAKAI (CBLL1) is another notable factor that interacts with the mA machinery (Figure 1A; Horiuchi et al., 2013; Rùžièka et al., 2017). However, its function in mA modification of mRNA in mammals is yet to be established.

Erasers (Demethylases) of N-Methyladenosine

mA modification is believed to be a reversible dynamic process premised on the identification of two demethylases: fat mass and obesity-associated protein (FTO) (Jia et al., 2011) and α-ketoglutarate-dependent dioxygenase alkB homolog 5 (ALKBH5) (Figure 1A). However, this important concept has been in controversy due to various supporting data from various studies (Mauer et al., 2017; Darnell et al., 2018; Wei et al., 2018) as discussed below.

Fat Mass and Obesity-Associated Protein

Following an in vitro assay, which demonstrated that FTO erases mA methylation of mRNA (Jia et al., 2008), it was further shown that downregulation (knockdown) of FTO in HeLa or 293FT cells caused reduction in mA methylation of mRNA (Jia et al., 2011). In support of this observation, it was identified that a small proportion of mA peaks of the whole transcriptome increased in Fto knockout mouse (Hess et al., 2013). These evidence consolidates the concept that mA modification can be reversed by FTO functionality. However, this idea was challenged by another study, in which no significant increase in mA level was observed in Fto knockout cells (Mauer et al., 2017). Instead, they noticed that FTO exhibits much higher catalytic capacity against mAm than mA. These studies indicate that the preferred substrate of FTO may be mAm (Meyer and Jaffrey, 2017). What could be the explanation behind the discrepancy between these findings? It is worth pointing out that several independent groups have reported that the Kcat/Km of FTO against mA is in the range of 0.6−0.7 min μM (Jia et al., 2011; Zhu and Yi, 2014; Zou et al., 2016), whereas that from the study of Mauer et al. (2017) is only 0.06 min μM, indicating that most likely there is a technique issue behind quantification of the Kcat/Km of FTO against mA. Additionally, both investigations used different methods to determine the level of mA, noting that the RNase T1 treatment of mRNA combined with thin-layer chromatography can only measure the mA in the case of RGACH, but not RAACH (Mauer et al., 2017). Moreover, a recent study further demonstrates that FTO not only demethylates internal mA but also caps mAm (Wei et al., 2018). The subcellular distribution of FTO varies among cultured cell lines, which indicates that the pattern of FTO demethylation of mA in cytosol or nucleus could be cell lineage-dependent. Consistent with the above studies, it was found that FTO plays a vital role in cell cycle and mitosis regulation in an mA demethylation-dependent manner during spermatogenesis (Huang T. et al., 2018).

Structural studies uncovered that FTO prefers mA-modified nucleobase, and its demethylase activity can be influenced by the primary and the tertiary structure of target RNA (Zhang X. et al., 2019), thus shedding light on the molecular mechanism behind the demethylation function of FTO. Recent findings show that the transcription of FTO is regulated by a transcriptional factor Zfp217 during adipogenesis, and Zfp217 is critical for FTO to associate with mA sites, albeit through competition with YTHDF2 for binding sites (Wei et al., 2019).

ALKBH5

ALKBH5 is another mA factor with demethylase capacity (Jia et al., 2011; Zheng et al., 2013). Manipulating ALKBH5 expression level leads to a slight but significant change in mA levels in the poly(A) region of mRNA. Compared with FTO, which demethylates mAm and mA, ALKBH5 shows specificity for mA demethylation (Wei et al., 2018). Importantly, mA-mediated conformational change facilitates distinction of substrates with minor sequence by FTO and ALKBH5 (Zou et al., 2016). As a nuclear protein, ALKBH5 is proposed to only erase the mA methylation in the nucleus (Meyer and Jaffrey, 2017). The demethylation capacity of ALKBH5 plays important roles in mRNA splicing, transport, stability, and processing. For instance, spermatogenic transcripts with increased mA levels exhibit increased splicing events in Alkbh5 KO mice (Tang et al., 2018). Recently, it was reported that METTL3 and ALKBH5 counteractively modulate the mA methylation of TFEB transcript to effect regulation of autophagy (Song et al., 2019). The demethylation activity of ALKBH5 can be regulated by DEAD-Box RNA helicase through physical interaction (Shah et al., 2017).

Readers (Binding Proteins) of N-Methyladenosine

The functional significance of mA modification also depends on mA-binding proteins also referred to as mA readers. As described below, we categorize the readers in mammalian cells into two groups: YTH domain-containing proteins, including YTHDC1, YTHDC2, and YTHDF1−3 (Hazra et al., 2019), and Non YTH domain-containing proteins, including eIF3 (Meyer et al., 2015), IGF2BPs (Huang H. et al., 2018), HuR (Dominissini et al., 2012; Wang Y. et al., 2014), FMRP (Zhang F. et al., 2018), hnRNPA2/B1/C (Dominissini et al., 2012; Alarcón et al., 2015), and METTLs (Wang et al., 2015).

YTH Domain Containing N-Methyladenosine-Binding Proteins

This group contains YTHDC1, YTHDC2, and YTHDF1−3 families in mammals. The common YTH domain defines members of this group of mA binding and determines the nature of mA reading (Zhang et al., 2010). However, they are not paralogs. This is because of the non-similarity of other aspects of proteins apart from the common YTH domain (Hazra et al., 2019).

YTHDC1 (YT521-B) is the first identified mA reader, which was found as a protein associated with splicing factors (Imai et al., 1998; Harfmann et al., 1999; Xiao et al., 2016; Hazra et al., 2019). Interestingly, human YTHDC1 shows much greater binding affinity for the mA-modified mRNA region in the context of GmAC (five-fold to six-fold difference) than AmAC, although the distribution of mA modification found in the consensus sequence is GmAC (70%) and AmAC (30%) (Xu et al., 2014, 2015). It localizes in various subnuclear bodies close to the nuclear splicing factor (SF) compartments and plays a role in mRNA splicing through physical interaction with splicing factor SRSF3 and SRSF10 (Figure 2A; Xiao et al., 2016). Furthermore, YTHDC1 works together with NXF1 and SRSF3 to regulate mA-modified mRNA nuclear export (Figure 2B; Roundtree et al., 2017b). Moreover, YTHDC1 has been reported to bind mA-modified MAT2A mRNA. The mA modification results in the degradation of MAT2A mRNA, although the detailed mechanism is not known (Shima et al., 2017).

An external file that holds a picture, illustration, etc.
Object name is fcell-09-656849-g002.jpg

Effects of N-methyladenosine (mA) methylation on messenger RNA (mRNA) fate. (A) mA modification regulates mRNA splicing and polyadenylation via YTHDC1 and its associating factors SRSF3, SRSF10, and CPSF6. (B) mA modulates mRNA nuclear export through YTHDC1, SRSF3, and NXF1. (C) mA regulates mRNA translation and stability via YTHDC2-mediated recruitment of the ribosome and the XRN1 exoribonuclease. (D) mA marks are bound by IGFBPs, which can regulate a subset of mRNA translation, decay in P-body, and storage in stress granules. (E) mA modification regulates mRNA degradation in P-body through associating with the YTHDF2-CCR4-NOT complex. (F) mA marks on circRNA modulate its translation via recruiting YTHDF3. (G) mA marks recruit YTHDF1/YTHDF3 to enhance translation in a Cap-dependent manner. (H) METTL3 serves as an mA reader and increases translation via recruiting translation initiation complex independent of its methyltransferase activity. (I) mA directly binds to eIF3 and increases translation in a Cap-independent manner.

On the other hand, YTHDC2 is a multi-domain protein and mainly localized in the cytoplasm, but it is also highly expressed in perinuclear compartment. It prefers to bind mA-containing RNAs through the YTH domain and enhances RNA degradation. Meanwhile, it also enhances mA-modified mRNA translation efficiency (Kretschmer et al., 2018).

Human YTHDF1−3 proteins contain a YTH domain in the C-terminus and a low-complexity domain in the N-terminus. These three members of the YTHDF family share high sequence identity and similarity (65%−80%) (Li et al., 2014; Wang X. et al., 2014; Hazra et al., 2019). As a characterized mA modification reader, the human YTHDF2 binds over 3,000 transcripts primarily in their 3′ UTRs and around the stop codon. The binding of YTHDF2 leads to degradation of the bound mRNAs in cytoplasmic processing bodies (P-bodies). Knockdown of YTHDF2 leads to an accumulation of mA-containing mRNAs (Wang X. et al., 2014). YTHDF2 was also found to associate with CNOT1, the scaffolding component of the CCR4-NOT mRNA deadenylation complex (Figure 2E). This interaction is required for YTHDF2 to localize in P-bodies (Du et al., 2016). Therefore, the main function of YTHDF2 is to control the degradation of mA-modified mRNAs (Kang et al., 2014; Hazra et al., 2019).

Unlike YTHDF2, YTHDF1 does not induce the degradation of associated mA-containing mRNAs. Instead, but arguably, YTHDF1 increases the translation efficiency of associated mRNAs (about 1,200) in an mA-dependent fashion (Wang et al., 2015). This function of YTHDF1 is further supported by the work of Wu et al. (2019), who showed that YTHDF1 targets mA-modified Jak2 and regulates its translation (Wu et al., 2019). Recently, another cytoplasmic mA reader protein YTHDF3 was found to interact with YTHDF1 to promote translation, whereas YTHDF3 interacts with YTHDF2 to reinforce mRNA decay (Li A. et al., 2017; Shi et al., 2017). Furthermore, biochemical studies showed that YTHDF3 shares greater than 50% of common mA-modified mRNA targets with YTHDF1 and also with YTHDF2 (Li A. et al., 2017; Shi et al., 2017). In addition, YTHDF3 might also function as mA-modification reader independent of YTHDF1 and YTHDF2 under certain conditions. Oxidative stress induces specific mA modifications in a set of transcripts, and the binding of YTHDF3 to the modifications triggers the mRNA–YTHDF3 complex localization in the stress granules, but without much influence on YTHDF1 and YTHDF2 (Anders et al., 2018). Importantly, YTHDF3 can also enhance translation independent of METTL3-mediated mA modification. For example, YTHDF3 functions together with eIF4G2 and Poly(A)-binding protein 1 (PABP1) to promote the translation of forkhead box protein O3 (FOXO3) (Zhang Y. et al., 2019).

Interestingly, very recent studies have shown evidence indicating functional redundancy of the YTHDFs during mRNA degradation and cellular differentiation. As such, it is only when all three YTHDF homologs (YTHDF1−3) are ablated that mRNA stability and cell differentiation regulation become evident (Kontur et al., 2020; Zaccara and Jaffrey, 2020). This may partly stem from the observations that all three YTHDFs are similar in sequence characteristics and usually have common mRNA binding targets (Zaccara and Jaffrey, 2020). Yet, it has been reported that probably due to variation in its expression, YTHDF2 dominates the mA reader function of all the YTHDFs (Lasman et al., 2020). It was also unraveled that YTHDFs are unable to induce translation in HeLa cells (Zaccara and Jaffrey, 2020). While these new findings present a unified model seeking to define the regulatory functions of YTHDFs in mA modification, they provoke questions that need to be addressed to reconcile the discrepancy between the recent findings and previous observations with respect to the precise role of YTHDFs in mRNA translation.

Non YTH Domain Containing N-Methyladenosine Readers (eIF3, IGF2BPs, HuR, FMRP, HNRNP Proteins, and PRRC2A)

Meyer et al. (2015) characterized the function of eIF3 as an mA reader. eIF3 is preferentially recruited by the mA-modified mRNA over unmethylated mRNA (Meyer et al., 2015). It was shown that about 35% of mA marks in the 5′ UTR are also eIF3-binding sites. Depletion of mA through METTL3 loss-of-function decreased the translation of mA-modified mRNA in the 5′ UTR, but not the mRNAs bearing mA marks elsewhere (Meyer et al., 2015). Notably, one of the two modes of mA-mediated Cap-independent translation is through direct association of mA in the 5′ UTR and eIF3 (Figure 2I; Meyer et al., 2015), while the other mode involves YTHDF1 association with mA mark followed by delivery of eIF3 to the 5′ UTR (Figure 2G; Wang et al., 2015). This indicates the correlation between the 5′ UTR mA and translation and highlights the involvement of eIF3 in the regulation of mRNA translation. Currently, it is not known what the detailed mechanisms are in determining the mode of eIF3−5′ UTR association.

Insulin-like growth factor-2-binding proteins (IGFBPs), including IGFBP1−3, have been reported as RNA-binding proteins (Bell et al., 2013). Recently, it was demonstrated that IGFBP1−3 bind mA-modified mRNAs with a three-fold to four-fold greater affinity than the mA-unmodified mRNAs (Huang H. et al., 2018). By means of RIP-Seq or PAR-CLIP-Seq, it was found that IGFBP1−3 share 55%−70% RNA targets with preference for binding to the “UGGAC” consensus motif, e.g., MYC, FSCN1, and TK1 (Huang H. et al., 2018). Knockdown of METTL14, a critical component of the methylation machinery, dramatically undermined IGFBP binding. Interestingly, knockdown of IGF2BPs reduces mRNA stability (Huang H. et al., 2018). Consistently, IGFBPs were found to associate with three mRNA stabilizing factors, including HuR, MATR3, and PABP1, which can support IGFBPs in stabilizing their mRNA targets (Huang H. et al., 2018).

HuR is an RNA-binding protein with multiple molecular functions. It was first described as a stabilizer of ARE-containing mRNAs (Fan and Steitz, 1998; Peng et al., 1998). It is also known to enhance translation, although it can also exert translation suppression (Hinman and Lou, 2008; Abdelmohsen and Gorospe, 2010). This portrays HuR as both a reader and anti-reader of mA (Dominissini et al., 2012; Wang Y. et al., 2014). However, the underlying mechanism that makes mA modification sites to recruit or block HuR binding is unknown. We think that a sequence-dependent context may be at play in determining the function of HuR in mA interaction. This speculation remains to be investigated.

FMR1 (also known as FMRP1) is an RNA-binding protein and known to associate with hundreds of transcripts to decrease their translation. It binds to mA-modified mRNA in an RNA sequence context-dependent manner. FMR1 selectively binds to the mA marks associated with GGACU RNA sequence (Edupuganti et al., 2017). Bioinformatic analysis revealed that FMR1 and YTHDF1 shared an abundant set of common mA-modified mRNAs, indicating that FMR1 might compete with YTHDF1 for binding of mA-modified mRNAs to downregulate translation (Ascano et al., 2012; Wang et al., 2015). It is possible that the mechanism may underlie the previously reported regulatory function on the translation of mRNA targets.

Heterogeneous nuclear ribonucleoproteins (hnRNPs: hnRNPA2/B1, hnRNPC, and hnRNPG) are RNA-binding proteins that play important roles in pre-RNA processing (Dominissini et al., 2012; Alarcón et al., 2015; Liu et al., 2015, 2017; Xiao et al., 2016). Alarcón et al. (2015) discovered that hnRNPA2B1 interacts with a group of mA-modified RNAs in the nucleus and regulates their splicing in a comparable pattern as for METTL3. However, the binding of hnRNPA2/B1 to mA is likely indirect and may require an hnRNPC-mediated switch mechanism to do so (Wu et al., 2018). hnRNPC can read mA-modified hairpin and mA-containing RNAs. mA-modification leads to a change in the regional RNA structure and increases the binding of hnRNPC (Liu et al., 2015). Consistently, general reduction in mA marks due to METTL3/L14 knockdown eliminates the association of hnRNPC to the aforementioned mA-mediated RNA structural modification (Liu et al., 2015). Furthermore, HNRNPG is known to bind mA-modified lncRNA through its C-terminal low-complexity domain (LCD), indicating that LCD domain might be used by some other readers to bind to mA modification (Liu et al., 2017).

Recently, PRRC2a was reported as an mA modification reader (Wu et al., 2019). Through RIP-seq and mA-seq, it was identified that PRRC2a binding peaks within over 2,800 genes in brain samples, and PRRC2a mainly binds to the consensus motif UGGAC in mA-modified transcripts (Wu et al., 2019). PRRC2A was found to be associated with YTHDF2 in granule-like organelles, which may be involved in the regulation of PRRC2A involvement in Olig2 mRNA stability (Wu et al., 2019). However, since PRRC2A has low tissue expression specificity, it is unclear whether PRRC2A serves as an mA modification reader in other tissues.

Reader Function of METTLs

Besides its role as an mA writer, METTL3 can also bind to mA-modified mRNAs to act as a reader. It was found that METTL3 regulates the translation of some oncogenic mA-modified mRNAs independent of its methyltransferase activity through eIF3 recruitment to the translation initiation complex (Lin et al., 2016). A study from the same group identified a physical interaction between mA-bound METTL3 near the stop codon and eIF3h, providing a mechanism to explain how METTL3 can enhance translation (Choe et al., 2018). The methyltransferase METTL16 also serves as an mA reader in a certain context. When SAM concentrations become low, METTL16 remains bound to mA-modified MAT2A in its 3′ UTR hp1 site to enhance MAT2A splicing, resulting in increased MAT2A levels in the cytosol. On the contrary, when SAM levels are high, METTL16 methylates MAT2A and facilitates intron retention (Pendleton et al., 2017).

Deposition of N-Methyladenosine Modification During Transcription

Mechanistically, how mA modification of transcripts is carried out needs elucidation. A recent study uncovered an insightful detail in the installation of mA. Specifically, it was found that H3K36me3 cooperates with METTL3/METTL14 to deposit mA on mRNA (Huang et al., 2019). The study showed that H3K36me3 physically interacts with METT14, thus recruits the mA methylation machinery to RNA Pol II, and allows the mA methylation machinery to effect mA modification during transcription. Decreasing the level of H3K36me3 through loss-of-function of SETD2, the specific enzyme that converts H3K36me2 or H3K36me0 to H3K36me3, significantly led to the reduction in mA level on RNAs, mimicking the impact of depletion of individual mA writer complex components (Huang et al., 2019).

Impact of N-Methyladenosine Modification on Gene Regulation

The reversible modification of mA exerts functional impact on several aspects of mRNA metabolism, including nuclear export, polyadenylation, splicing, degradation, and translation (Figure 2). By these means, the mA methylome affords an additional level of gene expression regulation to sculpt the transcriptome (Fu et al., 2014).

N-Methyladenosine Modification Regulates mRNA Splicing

Some factors involved in mA modification of mRNA are known to interact with pre-mRNA splicing factors (SRSFs), indicating a possible role for mA in mRNA splicing (Zhao et al., 2014; Xiao et al., 2016). It has been demonstrated that enrichment of mA modification promotes recruitment of SRSF2 and leads to enhanced exon inclusion of target mRNA (Zhao et al., 2014). It has been further suggested that the mA reader YTHDC1 regulates the association of mA and SRSFs. Indeed, mA-bound YTHDC1 enhances the recruitment of SRSF3 that favors exon inclusion but blocks the recruitment of SRSF10, an exon skipping-related splicing factor (Xiao et al., 2016). Moreover, hnRNPs may also be involved in the regulation of RNA splicing (Liu et al., 2015, 2017). For example, the modification of mA on pre-mRNA favors the binding of hnRNPC (Liu et al., 2015), which could further facilitate splicing through its known function in repressing exon inclusion (Zarnack et al., 2013). Therefore, it is possible that perturbation of the mA machinery components can impair mRNA alternative splicing. This idea is especially supported by the observation that knockdown of METTL3 can antagonize the association of SRSF2 or SRSF3 with mA-modified pre-mRNAs (Zhao et al., 2014; Xiao et al., 2016), and facilitates the expression of the long isoform of MyD88 (MyD88L) via exon skipping attenuation (Feng et al., 2018). Additional evidence is also based on the essential role played by METTL16 in MAT2A-mediated pre-mRNA alternative splicing (Pendleton et al., 2017).

N-Methyladenosine Controls Alternative Polyadenylation

Ke et al. (2015) found that mA modification peaks in the 3′ UTR, especially for transcripts that use alternative polyadenylation (APA), and longer last exons exhibit a higher mA density. By comparing the mA density of thousands of mRNA UTRs from liver and brain tissues, it was observed that greater amount mA marks in the last exons are linked to the usage of more distal polyA sites. Indeed, global reduction of mA levels via triple knockdown of METTL3, METTL14, and WTAP changed the polyA sites of one-sixth of the examined 661 mRNAs and promoted the usage of proximal APA sites, indicating that some mA marks inhibit proximal polyadenylation (Ke et al., 2015).

Recently, a mechanism through which mA controls alternative polyadenylation was proposed. VIRMA (Figure 1A) was found to interact with polyA cleavage factors F5 and CPSF6 (Yue et al., 2018). Consistent with an earlier report, knockdown of METTL3 or VIRMA was found to encourage the usage of distal APA sites, thus lengthening the 3′ UTR of mA-rich mRNAs. In contrast, CPSF5 knockdown elicits an opposite effect on the length of the 3′ UTR of mA-marked mRNAs (Yue et al., 2018).

N-Methyladenosine Promotes Nuclear Export

Considerable amount of nuclear export of mRNAs is regulated by the THO/TREX complex and the nuclear export factor heterodimer NXF1/P15 (Lesbirel and Wilson, 2019). Evidence is accumulating for the role of mA modification in mRNA nuclear export. Knockdown of METTL3 resulted in delayed nuclear export of specific mRNAs of clock genes (Fustin et al., 2013), indicating the requirement of mA methylation for specific mRNA nuclear export. Conversely, knockdown of ALKBH5 increased the cytoplasmic accumulation of polyA mRNAs (Zheng et al., 2013). Moreover, VIRMA was observed to interact with the mRNA export factor ALYREF, and its downregulation led to defective mRNA export (Masuda et al., 2005).

Interestingly, several TREX components associate with the components of the core mA machinery (METTL3-METTL14-WTAP-VIRMA), and TREX also enhances the association of mA reader YTHDC1 with the mRNA. Knockdown of YTHDC1 also resulted in reduced nuclear export of specific mRNAs (Lesbirel et al., 2018). Taken together, the abovementioned literature demonstrates that mA modification factors promote mRNA nuclear transport through physical interaction with the mRNA transport machinery.

N-Methyladenosine Enhances mRNA Degradation

Numerous recent studies suggest that impaired mA writer complex function reduces mA modification levels and raises mRNA stability, indicating that mA methylation drives mRNA degradation (Batista et al., 2014; Schwartz et al., 2014; Wang X. et al., 2014; Wang Y. et al., 2014; Park et al., 2019). Mechanistically, mA-containing mRNA recruits YTHDF2, which is followed by the translocation of the YTHDF2–mRNA complex from the translation machinery to P-bodies, leading to the degradation of YTHDF2-targeted mRNA. As a result, mRNA targets have increased half-life following YTHDF2 knockdown (Wang X. et al., 2014). It has been clearly demonstrated that YTHDF2 enhances mA-modified mRNA decay through recruiting CCR4-NOT deadenylase complex via the N-terminus of YTHDF2 and reveals an underlying mechanism by which YTHDF2 regulates degradation of mA-modified mRNAs (Du et al., 2016).

In a recent study, it was reported that some mA-modified mRNAs interact with YTHDF2 to undergo decay in an RNase P/MRP-dependent manner and in which HRSP12 serves as a bridge between YTHDF2 and RNase P/MRP (Park et al., 2019). The interaction of human YTHDF2 and HRSP12 was first hinted by the association between their respective yeast homologs Pho92 and Mmf1 (Krogan et al., 2006). It was found in an immunoprecipitation experiment that HRSP12 links YTHDF2 and RNase P/MRP and that the N-terminus of YTHDF2 is required to interact with HRSP12 (Park et al., 2019). Of note, the subset of mA-modified mRNAs, whose decay depends on YTHDF2–HRSP12–RNase P/MRP complex, contains a specific HRSP12-binding motif proximally upstream of the YTHDF2-binding motif, while the RNase P/MRP cleavage site is downstream and close to the YTHDF2-binding motif (Park et al., 2019). Therefore, this study discloses at least two mechanisms involved in the degradation of YTHDF2-associated mA RNAs: HRSP12-RNase P/MRP-dependent and CCR4-NOT complex-dependent.

N-Methyladenosine Modulates Translation

The mA reader YTHDF1 enhances translation efficiency via interaction with eIF3A/eIF3B, and the YTHDF1-regulated translation likely hinges on eIF4G-dependent loop formation (Wang et al., 2015). According to Meyer et al. (2015), 5′ UTR mA elevates cap-independent translation through recruiting the 43S complex to form 48S initiation complex in the absence of the cap-associating complex, eIF4F. This mechanism is important for cells to bypass 5′ cap-binding factors to enhance translation under stress conditions (Meyer et al., 2015). Moreover, heat stress-induced cytoplasmic-to-nuclear translocation of YTHDF2 is required for maintaining 5′ UTR mA levels via competing for binding of the demethylase FTO to mA sites, which further promotes cap-independent translation (Zhou et al., 2015). YTHDF1 preferentially binds to mA marks in 3′ UTR of the oncogene CDCP1 mRNA and promotes translation by increasing the amount of polysome-bound (translationally active) CDCP1 transcripts (Yang et al., 2019).

Of note, METTL3 is also involved in mA-enhanced mRNA translation through its role as an mA reader in several ways. It promotes mRNA translation via physical association with the translation initiation complex (Lin et al., 2016). It was found that promoter-associated METTL3 regulates mA methylation inside the coding region and improves mRNA translation through relief of ribosome stalling (Barbieri et al., 2017).

Besides promoting translation efficiency, mA modification also plays an important role in regulating alternative translation (Zhou J. et al., 2018). It has been reported that widespread alternative translation occurs under various nutrient conditions, but the underlying mechanism is unclear (Gao et al., 2015). Recently, Zhou J. et al. (2018) found that mA modification in the 5′ UTR modulates the selection of start codon globally, hence driving alternative translation. As representative examples, Atf4 depends on decreased mA modification of the upstream open reading frame 2 (uORF2) to improve the translation of the major isoform, and Gadd45g heightens the translation of the major isoform by lowering the mA modification of the 5′ UTR (Zhou J. et al., 2018).

N-Methyladenosine Methylation Increases the Phase Separation Capacity of mRNA

Only until recently has it become clearer how mA modification drives mRNA fate and why the consequence of mA modifications can vary in various scenarios. According to Ries et al. (2019), the mA readers, YTHDF1−3, experience liquid–liquid phase separation (LLPS). The mRNAs with multiple mA marks serve as a scaffold to bind with YTHDF readers via their low-complexity regions (LCRs). The mRNA–YTHDF complexes are then transported into various phase separators, like P-bodies, stress granule, and neuronal granules. The study suggests that the number and allocation of mA modifications in mRNAs remodel the transcriptome of different phase-separated compartments. The efficacy of mA modification-dependent modulation of an mRNA is likely governed by signals regulating the ability of YTHDF protein involved in LLPS formation (Ries et al., 2019).

N-Methyladenosine Writers

The installation of mA is carried out by ∼1 MDa mA writer complex composed of the methyltransferase-like protein 3 (METTL3) and METTL14, which heterodimerize (METTL3-METTL14) to function as the enzymatic core of the writer complex (Bokar et al., 1994; Bujnicki et al., 2002; Liu et al., 2014; Iyer et al., 2016). Additionally, other factors are known to interact with the mA writer complex. These include Wilms tumor 1-associated protein (WTAP) (Ping et al., 2014), VIRMA/KIA1429 (Yue et al., 2018), RNA-binding protein 15 (RBM15) (Patil et al., 2016; Huang and Yin, 2018), ZC3H13 (Knuckles et al., 2018), and HAKAI (Yue et al., 2018) (Figure 1A). These cofactors are regulated by the binding of RNA and the catalytic activity of the enzymatic core of the mA writer complex (Bujnicki et al., 2002; Liu et al., 2014; Ping et al., 2014; Iyer et al., 2016; Yue et al., 2018).

METTL3 and METTL14

Bokar et al. (1994) partially purified the mA writer complex using an in vitro methylation system and identified MT-70, a 70-kDa sub-complex possessing S-adenosylmethionine-binding methyltransferase capacity (Bokar et al., 1994). Later, it was renamed METTL3 (Narayan and Rottman, 1988; Bokar et al., 1997). Knockout of METTL3 in cells effectively blocks mA modification of mRNAs (Zhong et al., 2008; Agarwala et al., 2012; Geula et al., 2015). On the other hand, METTL14 forms a stable heterodimer with METTL3 to form the methyltransferase core of the mA methylation machinery (Liu et al., 2014; Wang Y. et al., 2014). METTL14, however, lacks enzymatic function and instead acts as an RNA-binding scaffold to augment the enzyme activity of METTL3 by directing the location of SAM methyl group required for the reaction (Śledź and Jinek, 2016; Wang P. et al., 2016; Wang X. et al., 2016). Therefore, METTL3 is the primary enzyme responsible for mA installation on mRNA.

METTL3-METTL14-Associated Adaptors: WTAP, VIRMA (KIAA1429), RBM15/15B, ZC3H13 (KIAA0853), and HAKAI

The core mA writer complex METTL3-METTL4 interacts with other adaptor proteins. It was found that FIP37 (the plant homolog of WTAP) co-localized with MTA (Arabidopsis homolog of METTL3) in the nucleus through physical interaction (Zhong et al., 2008). Similar interaction between WTAP and METTL3 was observed in mammalian cells (Liu et al., 2014; Ping et al., 2014; Schwartz et al., 2014). WTAP is key in keeping the METTL3-METTL4 heterodimer in nuclear speckles (Ping et al., 2014). Loss of WTAP leads to the depletion of mA modification in mRNA, indicating that WTAP may orient METTL3-METTL14 onto targets (Ping et al., 2014). However, the detailed mechanism remains elusive. Of note, it was demonstrated that two classes of mA sites exist: WTAP-dependent and WTAP-independent sites (Schwartz et al., 2014). VIRMA is known to also interact with the WTAP-METTL3-METTL4 complex (Figure 1A; Schwartz et al., 2014) and indicates its essentiality for the mA writer complex functionality. Indeed, VIRMA deletion in human cells leads to a significant reduction in mRNA methylation levels, although not as intense as that resulting from WTAP knockdown (Schwartz et al., 2014). Biochemical studies from Yue et al. (2018) demonstrated that VIRMA preferentially regulates mA modification in the 3’ UTR proximal to stop codon.

Through proteomic studies, Horiuchi et al. (2013) observed that RBM15 and its paralog RBM15B, together with ZC3H13/KIAA0853, and MTA70 (METTL3), associate with WTAP (Horiuchi et al., 2013; Patil et al., 2016), which raises the possibility that RBM15 and RBM15B may also play role(s) in mA modification. Indeed, silencing of RBM15 and RBM15B led to a demonstrable decrease in mA levels of mRNA (Patil et al., 2016). Based on Individual-nucleotide resolution UV crosslinking and immunoprecipitation (iCLIP) data, it was proposed that RBM15/15B recruit the mA methylation machinery to perform mA modification through binding to uridine-rich regions near DRACH sites. That notwithstanding, it is not always the case that uridine-rich regions exist near mA sites; therefore, other methylation complex adaptors may mediate the complex binding to such variant mA sites (Patil et al., 2016; Meyer and Jaffrey, 2017).

ZC3H13/KIAA0853 is also an interactor of the mA machinery, and it is demonstrated to be crucial in linking RBM15/15B to WTAP (Horiuchi et al., 2013; Knuckles et al., 2018; Wen et al., 2018). Knockdown of ZC3H13 shifts the localization of the mA adaptors WTAP, Virilizer, and Hakai from nucleus to cytosol in embryonic stem cells and leads to a significant total reduction in mA level on mRNA (Wen et al., 2018). This reflects an essential role of ZC3H13 in the deposition of mA on mRNAs. The E3 ubiquitin ligase HAKAI (CBLL1) is another notable factor that interacts with the mA machinery (Figure 1A; Horiuchi et al., 2013; Rùžièka et al., 2017). However, its function in mA modification of mRNA in mammals is yet to be established.

METTL3 and METTL14

Bokar et al. (1994) partially purified the mA writer complex using an in vitro methylation system and identified MT-70, a 70-kDa sub-complex possessing S-adenosylmethionine-binding methyltransferase capacity (Bokar et al., 1994). Later, it was renamed METTL3 (Narayan and Rottman, 1988; Bokar et al., 1997). Knockout of METTL3 in cells effectively blocks mA modification of mRNAs (Zhong et al., 2008; Agarwala et al., 2012; Geula et al., 2015). On the other hand, METTL14 forms a stable heterodimer with METTL3 to form the methyltransferase core of the mA methylation machinery (Liu et al., 2014; Wang Y. et al., 2014). METTL14, however, lacks enzymatic function and instead acts as an RNA-binding scaffold to augment the enzyme activity of METTL3 by directing the location of SAM methyl group required for the reaction (Śledź and Jinek, 2016; Wang P. et al., 2016; Wang X. et al., 2016). Therefore, METTL3 is the primary enzyme responsible for mA installation on mRNA.

METTL3-METTL14-Associated Adaptors: WTAP, VIRMA (KIAA1429), RBM15/15B, ZC3H13 (KIAA0853), and HAKAI

The core mA writer complex METTL3-METTL4 interacts with other adaptor proteins. It was found that FIP37 (the plant homolog of WTAP) co-localized with MTA (Arabidopsis homolog of METTL3) in the nucleus through physical interaction (Zhong et al., 2008). Similar interaction between WTAP and METTL3 was observed in mammalian cells (Liu et al., 2014; Ping et al., 2014; Schwartz et al., 2014). WTAP is key in keeping the METTL3-METTL4 heterodimer in nuclear speckles (Ping et al., 2014). Loss of WTAP leads to the depletion of mA modification in mRNA, indicating that WTAP may orient METTL3-METTL14 onto targets (Ping et al., 2014). However, the detailed mechanism remains elusive. Of note, it was demonstrated that two classes of mA sites exist: WTAP-dependent and WTAP-independent sites (Schwartz et al., 2014). VIRMA is known to also interact with the WTAP-METTL3-METTL4 complex (Figure 1A; Schwartz et al., 2014) and indicates its essentiality for the mA writer complex functionality. Indeed, VIRMA deletion in human cells leads to a significant reduction in mRNA methylation levels, although not as intense as that resulting from WTAP knockdown (Schwartz et al., 2014). Biochemical studies from Yue et al. (2018) demonstrated that VIRMA preferentially regulates mA modification in the 3’ UTR proximal to stop codon.

Through proteomic studies, Horiuchi et al. (2013) observed that RBM15 and its paralog RBM15B, together with ZC3H13/KIAA0853, and MTA70 (METTL3), associate with WTAP (Horiuchi et al., 2013; Patil et al., 2016), which raises the possibility that RBM15 and RBM15B may also play role(s) in mA modification. Indeed, silencing of RBM15 and RBM15B led to a demonstrable decrease in mA levels of mRNA (Patil et al., 2016). Based on Individual-nucleotide resolution UV crosslinking and immunoprecipitation (iCLIP) data, it was proposed that RBM15/15B recruit the mA methylation machinery to perform mA modification through binding to uridine-rich regions near DRACH sites. That notwithstanding, it is not always the case that uridine-rich regions exist near mA sites; therefore, other methylation complex adaptors may mediate the complex binding to such variant mA sites (Patil et al., 2016; Meyer and Jaffrey, 2017).

ZC3H13/KIAA0853 is also an interactor of the mA machinery, and it is demonstrated to be crucial in linking RBM15/15B to WTAP (Horiuchi et al., 2013; Knuckles et al., 2018; Wen et al., 2018). Knockdown of ZC3H13 shifts the localization of the mA adaptors WTAP, Virilizer, and Hakai from nucleus to cytosol in embryonic stem cells and leads to a significant total reduction in mA level on mRNA (Wen et al., 2018). This reflects an essential role of ZC3H13 in the deposition of mA on mRNAs. The E3 ubiquitin ligase HAKAI (CBLL1) is another notable factor that interacts with the mA machinery (Figure 1A; Horiuchi et al., 2013; Rùžièka et al., 2017). However, its function in mA modification of mRNA in mammals is yet to be established.

Erasers (Demethylases) of N-Methyladenosine

mA modification is believed to be a reversible dynamic process premised on the identification of two demethylases: fat mass and obesity-associated protein (FTO) (Jia et al., 2011) and α-ketoglutarate-dependent dioxygenase alkB homolog 5 (ALKBH5) (Figure 1A). However, this important concept has been in controversy due to various supporting data from various studies (Mauer et al., 2017; Darnell et al., 2018; Wei et al., 2018) as discussed below.

Fat Mass and Obesity-Associated Protein

Following an in vitro assay, which demonstrated that FTO erases mA methylation of mRNA (Jia et al., 2008), it was further shown that downregulation (knockdown) of FTO in HeLa or 293FT cells caused reduction in mA methylation of mRNA (Jia et al., 2011). In support of this observation, it was identified that a small proportion of mA peaks of the whole transcriptome increased in Fto knockout mouse (Hess et al., 2013). These evidence consolidates the concept that mA modification can be reversed by FTO functionality. However, this idea was challenged by another study, in which no significant increase in mA level was observed in Fto knockout cells (Mauer et al., 2017). Instead, they noticed that FTO exhibits much higher catalytic capacity against mAm than mA. These studies indicate that the preferred substrate of FTO may be mAm (Meyer and Jaffrey, 2017). What could be the explanation behind the discrepancy between these findings? It is worth pointing out that several independent groups have reported that the Kcat/Km of FTO against mA is in the range of 0.6−0.7 min μM (Jia et al., 2011; Zhu and Yi, 2014; Zou et al., 2016), whereas that from the study of Mauer et al. (2017) is only 0.06 min μM, indicating that most likely there is a technique issue behind quantification of the Kcat/Km of FTO against mA. Additionally, both investigations used different methods to determine the level of mA, noting that the RNase T1 treatment of mRNA combined with thin-layer chromatography can only measure the mA in the case of RGACH, but not RAACH (Mauer et al., 2017). Moreover, a recent study further demonstrates that FTO not only demethylates internal mA but also caps mAm (Wei et al., 2018). The subcellular distribution of FTO varies among cultured cell lines, which indicates that the pattern of FTO demethylation of mA in cytosol or nucleus could be cell lineage-dependent. Consistent with the above studies, it was found that FTO plays a vital role in cell cycle and mitosis regulation in an mA demethylation-dependent manner during spermatogenesis (Huang T. et al., 2018).

Structural studies uncovered that FTO prefers mA-modified nucleobase, and its demethylase activity can be influenced by the primary and the tertiary structure of target RNA (Zhang X. et al., 2019), thus shedding light on the molecular mechanism behind the demethylation function of FTO. Recent findings show that the transcription of FTO is regulated by a transcriptional factor Zfp217 during adipogenesis, and Zfp217 is critical for FTO to associate with mA sites, albeit through competition with YTHDF2 for binding sites (Wei et al., 2019).

ALKBH5

ALKBH5 is another mA factor with demethylase capacity (Jia et al., 2011; Zheng et al., 2013). Manipulating ALKBH5 expression level leads to a slight but significant change in mA levels in the poly(A) region of mRNA. Compared with FTO, which demethylates mAm and mA, ALKBH5 shows specificity for mA demethylation (Wei et al., 2018). Importantly, mA-mediated conformational change facilitates distinction of substrates with minor sequence by FTO and ALKBH5 (Zou et al., 2016). As a nuclear protein, ALKBH5 is proposed to only erase the mA methylation in the nucleus (Meyer and Jaffrey, 2017). The demethylation capacity of ALKBH5 plays important roles in mRNA splicing, transport, stability, and processing. For instance, spermatogenic transcripts with increased mA levels exhibit increased splicing events in Alkbh5 KO mice (Tang et al., 2018). Recently, it was reported that METTL3 and ALKBH5 counteractively modulate the mA methylation of TFEB transcript to effect regulation of autophagy (Song et al., 2019). The demethylation activity of ALKBH5 can be regulated by DEAD-Box RNA helicase through physical interaction (Shah et al., 2017).

Fat Mass and Obesity-Associated Protein

Following an in vitro assay, which demonstrated that FTO erases mA methylation of mRNA (Jia et al., 2008), it was further shown that downregulation (knockdown) of FTO in HeLa or 293FT cells caused reduction in mA methylation of mRNA (Jia et al., 2011). In support of this observation, it was identified that a small proportion of mA peaks of the whole transcriptome increased in Fto knockout mouse (Hess et al., 2013). These evidence consolidates the concept that mA modification can be reversed by FTO functionality. However, this idea was challenged by another study, in which no significant increase in mA level was observed in Fto knockout cells (Mauer et al., 2017). Instead, they noticed that FTO exhibits much higher catalytic capacity against mAm than mA. These studies indicate that the preferred substrate of FTO may be mAm (Meyer and Jaffrey, 2017). What could be the explanation behind the discrepancy between these findings? It is worth pointing out that several independent groups have reported that the Kcat/Km of FTO against mA is in the range of 0.6−0.7 min μM (Jia et al., 2011; Zhu and Yi, 2014; Zou et al., 2016), whereas that from the study of Mauer et al. (2017) is only 0.06 min μM, indicating that most likely there is a technique issue behind quantification of the Kcat/Km of FTO against mA. Additionally, both investigations used different methods to determine the level of mA, noting that the RNase T1 treatment of mRNA combined with thin-layer chromatography can only measure the mA in the case of RGACH, but not RAACH (Mauer et al., 2017). Moreover, a recent study further demonstrates that FTO not only demethylates internal mA but also caps mAm (Wei et al., 2018). The subcellular distribution of FTO varies among cultured cell lines, which indicates that the pattern of FTO demethylation of mA in cytosol or nucleus could be cell lineage-dependent. Consistent with the above studies, it was found that FTO plays a vital role in cell cycle and mitosis regulation in an mA demethylation-dependent manner during spermatogenesis (Huang T. et al., 2018).

Structural studies uncovered that FTO prefers mA-modified nucleobase, and its demethylase activity can be influenced by the primary and the tertiary structure of target RNA (Zhang X. et al., 2019), thus shedding light on the molecular mechanism behind the demethylation function of FTO. Recent findings show that the transcription of FTO is regulated by a transcriptional factor Zfp217 during adipogenesis, and Zfp217 is critical for FTO to associate with mA sites, albeit through competition with YTHDF2 for binding sites (Wei et al., 2019).

ALKBH5

ALKBH5 is another mA factor with demethylase capacity (Jia et al., 2011; Zheng et al., 2013). Manipulating ALKBH5 expression level leads to a slight but significant change in mA levels in the poly(A) region of mRNA. Compared with FTO, which demethylates mAm and mA, ALKBH5 shows specificity for mA demethylation (Wei et al., 2018). Importantly, mA-mediated conformational change facilitates distinction of substrates with minor sequence by FTO and ALKBH5 (Zou et al., 2016). As a nuclear protein, ALKBH5 is proposed to only erase the mA methylation in the nucleus (Meyer and Jaffrey, 2017). The demethylation capacity of ALKBH5 plays important roles in mRNA splicing, transport, stability, and processing. For instance, spermatogenic transcripts with increased mA levels exhibit increased splicing events in Alkbh5 KO mice (Tang et al., 2018). Recently, it was reported that METTL3 and ALKBH5 counteractively modulate the mA methylation of TFEB transcript to effect regulation of autophagy (Song et al., 2019). The demethylation activity of ALKBH5 can be regulated by DEAD-Box RNA helicase through physical interaction (Shah et al., 2017).

Readers (Binding Proteins) of N-Methyladenosine

The functional significance of mA modification also depends on mA-binding proteins also referred to as mA readers. As described below, we categorize the readers in mammalian cells into two groups: YTH domain-containing proteins, including YTHDC1, YTHDC2, and YTHDF1−3 (Hazra et al., 2019), and Non YTH domain-containing proteins, including eIF3 (Meyer et al., 2015), IGF2BPs (Huang H. et al., 2018), HuR (Dominissini et al., 2012; Wang Y. et al., 2014), FMRP (Zhang F. et al., 2018), hnRNPA2/B1/C (Dominissini et al., 2012; Alarcón et al., 2015), and METTLs (Wang et al., 2015).

YTH Domain Containing N-Methyladenosine-Binding Proteins

This group contains YTHDC1, YTHDC2, and YTHDF1−3 families in mammals. The common YTH domain defines members of this group of mA binding and determines the nature of mA reading (Zhang et al., 2010). However, they are not paralogs. This is because of the non-similarity of other aspects of proteins apart from the common YTH domain (Hazra et al., 2019).

YTHDC1 (YT521-B) is the first identified mA reader, which was found as a protein associated with splicing factors (Imai et al., 1998; Harfmann et al., 1999; Xiao et al., 2016; Hazra et al., 2019). Interestingly, human YTHDC1 shows much greater binding affinity for the mA-modified mRNA region in the context of GmAC (five-fold to six-fold difference) than AmAC, although the distribution of mA modification found in the consensus sequence is GmAC (70%) and AmAC (30%) (Xu et al., 2014, 2015). It localizes in various subnuclear bodies close to the nuclear splicing factor (SF) compartments and plays a role in mRNA splicing through physical interaction with splicing factor SRSF3 and SRSF10 (Figure 2A; Xiao et al., 2016). Furthermore, YTHDC1 works together with NXF1 and SRSF3 to regulate mA-modified mRNA nuclear export (Figure 2B; Roundtree et al., 2017b). Moreover, YTHDC1 has been reported to bind mA-modified MAT2A mRNA. The mA modification results in the degradation of MAT2A mRNA, although the detailed mechanism is not known (Shima et al., 2017).

An external file that holds a picture, illustration, etc.
Object name is fcell-09-656849-g002.jpg

Effects of N-methyladenosine (mA) methylation on messenger RNA (mRNA) fate. (A) mA modification regulates mRNA splicing and polyadenylation via YTHDC1 and its associating factors SRSF3, SRSF10, and CPSF6. (B) mA modulates mRNA nuclear export through YTHDC1, SRSF3, and NXF1. (C) mA regulates mRNA translation and stability via YTHDC2-mediated recruitment of the ribosome and the XRN1 exoribonuclease. (D) mA marks are bound by IGFBPs, which can regulate a subset of mRNA translation, decay in P-body, and storage in stress granules. (E) mA modification regulates mRNA degradation in P-body through associating with the YTHDF2-CCR4-NOT complex. (F) mA marks on circRNA modulate its translation via recruiting YTHDF3. (G) mA marks recruit YTHDF1/YTHDF3 to enhance translation in a Cap-dependent manner. (H) METTL3 serves as an mA reader and increases translation via recruiting translation initiation complex independent of its methyltransferase activity. (I) mA directly binds to eIF3 and increases translation in a Cap-independent manner.

On the other hand, YTHDC2 is a multi-domain protein and mainly localized in the cytoplasm, but it is also highly expressed in perinuclear compartment. It prefers to bind mA-containing RNAs through the YTH domain and enhances RNA degradation. Meanwhile, it also enhances mA-modified mRNA translation efficiency (Kretschmer et al., 2018).

Human YTHDF1−3 proteins contain a YTH domain in the C-terminus and a low-complexity domain in the N-terminus. These three members of the YTHDF family share high sequence identity and similarity (65%−80%) (Li et al., 2014; Wang X. et al., 2014; Hazra et al., 2019). As a characterized mA modification reader, the human YTHDF2 binds over 3,000 transcripts primarily in their 3′ UTRs and around the stop codon. The binding of YTHDF2 leads to degradation of the bound mRNAs in cytoplasmic processing bodies (P-bodies). Knockdown of YTHDF2 leads to an accumulation of mA-containing mRNAs (Wang X. et al., 2014). YTHDF2 was also found to associate with CNOT1, the scaffolding component of the CCR4-NOT mRNA deadenylation complex (Figure 2E). This interaction is required for YTHDF2 to localize in P-bodies (Du et al., 2016). Therefore, the main function of YTHDF2 is to control the degradation of mA-modified mRNAs (Kang et al., 2014; Hazra et al., 2019).

Unlike YTHDF2, YTHDF1 does not induce the degradation of associated mA-containing mRNAs. Instead, but arguably, YTHDF1 increases the translation efficiency of associated mRNAs (about 1,200) in an mA-dependent fashion (Wang et al., 2015). This function of YTHDF1 is further supported by the work of Wu et al. (2019), who showed that YTHDF1 targets mA-modified Jak2 and regulates its translation (Wu et al., 2019). Recently, another cytoplasmic mA reader protein YTHDF3 was found to interact with YTHDF1 to promote translation, whereas YTHDF3 interacts with YTHDF2 to reinforce mRNA decay (Li A. et al., 2017; Shi et al., 2017). Furthermore, biochemical studies showed that YTHDF3 shares greater than 50% of common mA-modified mRNA targets with YTHDF1 and also with YTHDF2 (Li A. et al., 2017; Shi et al., 2017). In addition, YTHDF3 might also function as mA-modification reader independent of YTHDF1 and YTHDF2 under certain conditions. Oxidative stress induces specific mA modifications in a set of transcripts, and the binding of YTHDF3 to the modifications triggers the mRNA–YTHDF3 complex localization in the stress granules, but without much influence on YTHDF1 and YTHDF2 (Anders et al., 2018). Importantly, YTHDF3 can also enhance translation independent of METTL3-mediated mA modification. For example, YTHDF3 functions together with eIF4G2 and Poly(A)-binding protein 1 (PABP1) to promote the translation of forkhead box protein O3 (FOXO3) (Zhang Y. et al., 2019).

Interestingly, very recent studies have shown evidence indicating functional redundancy of the YTHDFs during mRNA degradation and cellular differentiation. As such, it is only when all three YTHDF homologs (YTHDF1−3) are ablated that mRNA stability and cell differentiation regulation become evident (Kontur et al., 2020; Zaccara and Jaffrey, 2020). This may partly stem from the observations that all three YTHDFs are similar in sequence characteristics and usually have common mRNA binding targets (Zaccara and Jaffrey, 2020). Yet, it has been reported that probably due to variation in its expression, YTHDF2 dominates the mA reader function of all the YTHDFs (Lasman et al., 2020). It was also unraveled that YTHDFs are unable to induce translation in HeLa cells (Zaccara and Jaffrey, 2020). While these new findings present a unified model seeking to define the regulatory functions of YTHDFs in mA modification, they provoke questions that need to be addressed to reconcile the discrepancy between the recent findings and previous observations with respect to the precise role of YTHDFs in mRNA translation.

Non YTH Domain Containing N-Methyladenosine Readers (eIF3, IGF2BPs, HuR, FMRP, HNRNP Proteins, and PRRC2A)

Meyer et al. (2015) characterized the function of eIF3 as an mA reader. eIF3 is preferentially recruited by the mA-modified mRNA over unmethylated mRNA (Meyer et al., 2015). It was shown that about 35% of mA marks in the 5′ UTR are also eIF3-binding sites. Depletion of mA through METTL3 loss-of-function decreased the translation of mA-modified mRNA in the 5′ UTR, but not the mRNAs bearing mA marks elsewhere (Meyer et al., 2015). Notably, one of the two modes of mA-mediated Cap-independent translation is through direct association of mA in the 5′ UTR and eIF3 (Figure 2I; Meyer et al., 2015), while the other mode involves YTHDF1 association with mA mark followed by delivery of eIF3 to the 5′ UTR (Figure 2G; Wang et al., 2015). This indicates the correlation between the 5′ UTR mA and translation and highlights the involvement of eIF3 in the regulation of mRNA translation. Currently, it is not known what the detailed mechanisms are in determining the mode of eIF3−5′ UTR association.

Insulin-like growth factor-2-binding proteins (IGFBPs), including IGFBP1−3, have been reported as RNA-binding proteins (Bell et al., 2013). Recently, it was demonstrated that IGFBP1−3 bind mA-modified mRNAs with a three-fold to four-fold greater affinity than the mA-unmodified mRNAs (Huang H. et al., 2018). By means of RIP-Seq or PAR-CLIP-Seq, it was found that IGFBP1−3 share 55%−70% RNA targets with preference for binding to the “UGGAC” consensus motif, e.g., MYC, FSCN1, and TK1 (Huang H. et al., 2018). Knockdown of METTL14, a critical component of the methylation machinery, dramatically undermined IGFBP binding. Interestingly, knockdown of IGF2BPs reduces mRNA stability (Huang H. et al., 2018). Consistently, IGFBPs were found to associate with three mRNA stabilizing factors, including HuR, MATR3, and PABP1, which can support IGFBPs in stabilizing their mRNA targets (Huang H. et al., 2018).

HuR is an RNA-binding protein with multiple molecular functions. It was first described as a stabilizer of ARE-containing mRNAs (Fan and Steitz, 1998; Peng et al., 1998). It is also known to enhance translation, although it can also exert translation suppression (Hinman and Lou, 2008; Abdelmohsen and Gorospe, 2010). This portrays HuR as both a reader and anti-reader of mA (Dominissini et al., 2012; Wang Y. et al., 2014). However, the underlying mechanism that makes mA modification sites to recruit or block HuR binding is unknown. We think that a sequence-dependent context may be at play in determining the function of HuR in mA interaction. This speculation remains to be investigated.

FMR1 (also known as FMRP1) is an RNA-binding protein and known to associate with hundreds of transcripts to decrease their translation. It binds to mA-modified mRNA in an RNA sequence context-dependent manner. FMR1 selectively binds to the mA marks associated with GGACU RNA sequence (Edupuganti et al., 2017). Bioinformatic analysis revealed that FMR1 and YTHDF1 shared an abundant set of common mA-modified mRNAs, indicating that FMR1 might compete with YTHDF1 for binding of mA-modified mRNAs to downregulate translation (Ascano et al., 2012; Wang et al., 2015). It is possible that the mechanism may underlie the previously reported regulatory function on the translation of mRNA targets.

Heterogeneous nuclear ribonucleoproteins (hnRNPs: hnRNPA2/B1, hnRNPC, and hnRNPG) are RNA-binding proteins that play important roles in pre-RNA processing (Dominissini et al., 2012; Alarcón et al., 2015; Liu et al., 2015, 2017; Xiao et al., 2016). Alarcón et al. (2015) discovered that hnRNPA2B1 interacts with a group of mA-modified RNAs in the nucleus and regulates their splicing in a comparable pattern as for METTL3. However, the binding of hnRNPA2/B1 to mA is likely indirect and may require an hnRNPC-mediated switch mechanism to do so (Wu et al., 2018). hnRNPC can read mA-modified hairpin and mA-containing RNAs. mA-modification leads to a change in the regional RNA structure and increases the binding of hnRNPC (Liu et al., 2015). Consistently, general reduction in mA marks due to METTL3/L14 knockdown eliminates the association of hnRNPC to the aforementioned mA-mediated RNA structural modification (Liu et al., 2015). Furthermore, HNRNPG is known to bind mA-modified lncRNA through its C-terminal low-complexity domain (LCD), indicating that LCD domain might be used by some other readers to bind to mA modification (Liu et al., 2017).

Recently, PRRC2a was reported as an mA modification reader (Wu et al., 2019). Through RIP-seq and mA-seq, it was identified that PRRC2a binding peaks within over 2,800 genes in brain samples, and PRRC2a mainly binds to the consensus motif UGGAC in mA-modified transcripts (Wu et al., 2019). PRRC2A was found to be associated with YTHDF2 in granule-like organelles, which may be involved in the regulation of PRRC2A involvement in Olig2 mRNA stability (Wu et al., 2019). However, since PRRC2A has low tissue expression specificity, it is unclear whether PRRC2A serves as an mA modification reader in other tissues.

Reader Function of METTLs

Besides its role as an mA writer, METTL3 can also bind to mA-modified mRNAs to act as a reader. It was found that METTL3 regulates the translation of some oncogenic mA-modified mRNAs independent of its methyltransferase activity through eIF3 recruitment to the translation initiation complex (Lin et al., 2016). A study from the same group identified a physical interaction between mA-bound METTL3 near the stop codon and eIF3h, providing a mechanism to explain how METTL3 can enhance translation (Choe et al., 2018). The methyltransferase METTL16 also serves as an mA reader in a certain context. When SAM concentrations become low, METTL16 remains bound to mA-modified MAT2A in its 3′ UTR hp1 site to enhance MAT2A splicing, resulting in increased MAT2A levels in the cytosol. On the contrary, when SAM levels are high, METTL16 methylates MAT2A and facilitates intron retention (Pendleton et al., 2017).

YTH Domain Containing N-Methyladenosine-Binding Proteins

This group contains YTHDC1, YTHDC2, and YTHDF1−3 families in mammals. The common YTH domain defines members of this group of mA binding and determines the nature of mA reading (Zhang et al., 2010). However, they are not paralogs. This is because of the non-similarity of other aspects of proteins apart from the common YTH domain (Hazra et al., 2019).

YTHDC1 (YT521-B) is the first identified mA reader, which was found as a protein associated with splicing factors (Imai et al., 1998; Harfmann et al., 1999; Xiao et al., 2016; Hazra et al., 2019). Interestingly, human YTHDC1 shows much greater binding affinity for the mA-modified mRNA region in the context of GmAC (five-fold to six-fold difference) than AmAC, although the distribution of mA modification found in the consensus sequence is GmAC (70%) and AmAC (30%) (Xu et al., 2014, 2015). It localizes in various subnuclear bodies close to the nuclear splicing factor (SF) compartments and plays a role in mRNA splicing through physical interaction with splicing factor SRSF3 and SRSF10 (Figure 2A; Xiao et al., 2016). Furthermore, YTHDC1 works together with NXF1 and SRSF3 to regulate mA-modified mRNA nuclear export (Figure 2B; Roundtree et al., 2017b). Moreover, YTHDC1 has been reported to bind mA-modified MAT2A mRNA. The mA modification results in the degradation of MAT2A mRNA, although the detailed mechanism is not known (Shima et al., 2017).

An external file that holds a picture, illustration, etc.
Object name is fcell-09-656849-g002.jpg

Effects of N-methyladenosine (mA) methylation on messenger RNA (mRNA) fate. (A) mA modification regulates mRNA splicing and polyadenylation via YTHDC1 and its associating factors SRSF3, SRSF10, and CPSF6. (B) mA modulates mRNA nuclear export through YTHDC1, SRSF3, and NXF1. (C) mA regulates mRNA translation and stability via YTHDC2-mediated recruitment of the ribosome and the XRN1 exoribonuclease. (D) mA marks are bound by IGFBPs, which can regulate a subset of mRNA translation, decay in P-body, and storage in stress granules. (E) mA modification regulates mRNA degradation in P-body through associating with the YTHDF2-CCR4-NOT complex. (F) mA marks on circRNA modulate its translation via recruiting YTHDF3. (G) mA marks recruit YTHDF1/YTHDF3 to enhance translation in a Cap-dependent manner. (H) METTL3 serves as an mA reader and increases translation via recruiting translation initiation complex independent of its methyltransferase activity. (I) mA directly binds to eIF3 and increases translation in a Cap-independent manner.

On the other hand, YTHDC2 is a multi-domain protein and mainly localized in the cytoplasm, but it is also highly expressed in perinuclear compartment. It prefers to bind mA-containing RNAs through the YTH domain and enhances RNA degradation. Meanwhile, it also enhances mA-modified mRNA translation efficiency (Kretschmer et al., 2018).

Human YTHDF1−3 proteins contain a YTH domain in the C-terminus and a low-complexity domain in the N-terminus. These three members of the YTHDF family share high sequence identity and similarity (65%−80%) (Li et al., 2014; Wang X. et al., 2014; Hazra et al., 2019). As a characterized mA modification reader, the human YTHDF2 binds over 3,000 transcripts primarily in their 3′ UTRs and around the stop codon. The binding of YTHDF2 leads to degradation of the bound mRNAs in cytoplasmic processing bodies (P-bodies). Knockdown of YTHDF2 leads to an accumulation of mA-containing mRNAs (Wang X. et al., 2014). YTHDF2 was also found to associate with CNOT1, the scaffolding component of the CCR4-NOT mRNA deadenylation complex (Figure 2E). This interaction is required for YTHDF2 to localize in P-bodies (Du et al., 2016). Therefore, the main function of YTHDF2 is to control the degradation of mA-modified mRNAs (Kang et al., 2014; Hazra et al., 2019).

Unlike YTHDF2, YTHDF1 does not induce the degradation of associated mA-containing mRNAs. Instead, but arguably, YTHDF1 increases the translation efficiency of associated mRNAs (about 1,200) in an mA-dependent fashion (Wang et al., 2015). This function of YTHDF1 is further supported by the work of Wu et al. (2019), who showed that YTHDF1 targets mA-modified Jak2 and regulates its translation (Wu et al., 2019). Recently, another cytoplasmic mA reader protein YTHDF3 was found to interact with YTHDF1 to promote translation, whereas YTHDF3 interacts with YTHDF2 to reinforce mRNA decay (Li A. et al., 2017; Shi et al., 2017). Furthermore, biochemical studies showed that YTHDF3 shares greater than 50% of common mA-modified mRNA targets with YTHDF1 and also with YTHDF2 (Li A. et al., 2017; Shi et al., 2017). In addition, YTHDF3 might also function as mA-modification reader independent of YTHDF1 and YTHDF2 under certain conditions. Oxidative stress induces specific mA modifications in a set of transcripts, and the binding of YTHDF3 to the modifications triggers the mRNA–YTHDF3 complex localization in the stress granules, but without much influence on YTHDF1 and YTHDF2 (Anders et al., 2018). Importantly, YTHDF3 can also enhance translation independent of METTL3-mediated mA modification. For example, YTHDF3 functions together with eIF4G2 and Poly(A)-binding protein 1 (PABP1) to promote the translation of forkhead box protein O3 (FOXO3) (Zhang Y. et al., 2019).

Interestingly, very recent studies have shown evidence indicating functional redundancy of the YTHDFs during mRNA degradation and cellular differentiation. As such, it is only when all three YTHDF homologs (YTHDF1−3) are ablated that mRNA stability and cell differentiation regulation become evident (Kontur et al., 2020; Zaccara and Jaffrey, 2020). This may partly stem from the observations that all three YTHDFs are similar in sequence characteristics and usually have common mRNA binding targets (Zaccara and Jaffrey, 2020). Yet, it has been reported that probably due to variation in its expression, YTHDF2 dominates the mA reader function of all the YTHDFs (Lasman et al., 2020). It was also unraveled that YTHDFs are unable to induce translation in HeLa cells (Zaccara and Jaffrey, 2020). While these new findings present a unified model seeking to define the regulatory functions of YTHDFs in mA modification, they provoke questions that need to be addressed to reconcile the discrepancy between the recent findings and previous observations with respect to the precise role of YTHDFs in mRNA translation.

Non YTH Domain Containing N-Methyladenosine Readers (eIF3, IGF2BPs, HuR, FMRP, HNRNP Proteins, and PRRC2A)

Meyer et al. (2015) characterized the function of eIF3 as an mA reader. eIF3 is preferentially recruited by the mA-modified mRNA over unmethylated mRNA (Meyer et al., 2015). It was shown that about 35% of mA marks in the 5′ UTR are also eIF3-binding sites. Depletion of mA through METTL3 loss-of-function decreased the translation of mA-modified mRNA in the 5′ UTR, but not the mRNAs bearing mA marks elsewhere (Meyer et al., 2015). Notably, one of the two modes of mA-mediated Cap-independent translation is through direct association of mA in the 5′ UTR and eIF3 (Figure 2I; Meyer et al., 2015), while the other mode involves YTHDF1 association with mA mark followed by delivery of eIF3 to the 5′ UTR (Figure 2G; Wang et al., 2015). This indicates the correlation between the 5′ UTR mA and translation and highlights the involvement of eIF3 in the regulation of mRNA translation. Currently, it is not known what the detailed mechanisms are in determining the mode of eIF3−5′ UTR association.

Insulin-like growth factor-2-binding proteins (IGFBPs), including IGFBP1−3, have been reported as RNA-binding proteins (Bell et al., 2013). Recently, it was demonstrated that IGFBP1−3 bind mA-modified mRNAs with a three-fold to four-fold greater affinity than the mA-unmodified mRNAs (Huang H. et al., 2018). By means of RIP-Seq or PAR-CLIP-Seq, it was found that IGFBP1−3 share 55%−70% RNA targets with preference for binding to the “UGGAC” consensus motif, e.g., MYC, FSCN1, and TK1 (Huang H. et al., 2018). Knockdown of METTL14, a critical component of the methylation machinery, dramatically undermined IGFBP binding. Interestingly, knockdown of IGF2BPs reduces mRNA stability (Huang H. et al., 2018). Consistently, IGFBPs were found to associate with three mRNA stabilizing factors, including HuR, MATR3, and PABP1, which can support IGFBPs in stabilizing their mRNA targets (Huang H. et al., 2018).

HuR is an RNA-binding protein with multiple molecular functions. It was first described as a stabilizer of ARE-containing mRNAs (Fan and Steitz, 1998; Peng et al., 1998). It is also known to enhance translation, although it can also exert translation suppression (Hinman and Lou, 2008; Abdelmohsen and Gorospe, 2010). This portrays HuR as both a reader and anti-reader of mA (Dominissini et al., 2012; Wang Y. et al., 2014). However, the underlying mechanism that makes mA modification sites to recruit or block HuR binding is unknown. We think that a sequence-dependent context may be at play in determining the function of HuR in mA interaction. This speculation remains to be investigated.

FMR1 (also known as FMRP1) is an RNA-binding protein and known to associate with hundreds of transcripts to decrease their translation. It binds to mA-modified mRNA in an RNA sequence context-dependent manner. FMR1 selectively binds to the mA marks associated with GGACU RNA sequence (Edupuganti et al., 2017). Bioinformatic analysis revealed that FMR1 and YTHDF1 shared an abundant set of common mA-modified mRNAs, indicating that FMR1 might compete with YTHDF1 for binding of mA-modified mRNAs to downregulate translation (Ascano et al., 2012; Wang et al., 2015). It is possible that the mechanism may underlie the previously reported regulatory function on the translation of mRNA targets.

Heterogeneous nuclear ribonucleoproteins (hnRNPs: hnRNPA2/B1, hnRNPC, and hnRNPG) are RNA-binding proteins that play important roles in pre-RNA processing (Dominissini et al., 2012; Alarcón et al., 2015; Liu et al., 2015, 2017; Xiao et al., 2016). Alarcón et al. (2015) discovered that hnRNPA2B1 interacts with a group of mA-modified RNAs in the nucleus and regulates their splicing in a comparable pattern as for METTL3. However, the binding of hnRNPA2/B1 to mA is likely indirect and may require an hnRNPC-mediated switch mechanism to do so (Wu et al., 2018). hnRNPC can read mA-modified hairpin and mA-containing RNAs. mA-modification leads to a change in the regional RNA structure and increases the binding of hnRNPC (Liu et al., 2015). Consistently, general reduction in mA marks due to METTL3/L14 knockdown eliminates the association of hnRNPC to the aforementioned mA-mediated RNA structural modification (Liu et al., 2015). Furthermore, HNRNPG is known to bind mA-modified lncRNA through its C-terminal low-complexity domain (LCD), indicating that LCD domain might be used by some other readers to bind to mA modification (Liu et al., 2017).

Recently, PRRC2a was reported as an mA modification reader (Wu et al., 2019). Through RIP-seq and mA-seq, it was identified that PRRC2a binding peaks within over 2,800 genes in brain samples, and PRRC2a mainly binds to the consensus motif UGGAC in mA-modified transcripts (Wu et al., 2019). PRRC2A was found to be associated with YTHDF2 in granule-like organelles, which may be involved in the regulation of PRRC2A involvement in Olig2 mRNA stability (Wu et al., 2019). However, since PRRC2A has low tissue expression specificity, it is unclear whether PRRC2A serves as an mA modification reader in other tissues.

Reader Function of METTLs

Besides its role as an mA writer, METTL3 can also bind to mA-modified mRNAs to act as a reader. It was found that METTL3 regulates the translation of some oncogenic mA-modified mRNAs independent of its methyltransferase activity through eIF3 recruitment to the translation initiation complex (Lin et al., 2016). A study from the same group identified a physical interaction between mA-bound METTL3 near the stop codon and eIF3h, providing a mechanism to explain how METTL3 can enhance translation (Choe et al., 2018). The methyltransferase METTL16 also serves as an mA reader in a certain context. When SAM concentrations become low, METTL16 remains bound to mA-modified MAT2A in its 3′ UTR hp1 site to enhance MAT2A splicing, resulting in increased MAT2A levels in the cytosol. On the contrary, when SAM levels are high, METTL16 methylates MAT2A and facilitates intron retention (Pendleton et al., 2017).

Deposition of N-Methyladenosine Modification During Transcription

Mechanistically, how mA modification of transcripts is carried out needs elucidation. A recent study uncovered an insightful detail in the installation of mA. Specifically, it was found that H3K36me3 cooperates with METTL3/METTL14 to deposit mA on mRNA (Huang et al., 2019). The study showed that H3K36me3 physically interacts with METT14, thus recruits the mA methylation machinery to RNA Pol II, and allows the mA methylation machinery to effect mA modification during transcription. Decreasing the level of H3K36me3 through loss-of-function of SETD2, the specific enzyme that converts H3K36me2 or H3K36me0 to H3K36me3, significantly led to the reduction in mA level on RNAs, mimicking the impact of depletion of individual mA writer complex components (Huang et al., 2019).

Impact of N-Methyladenosine Modification on Gene Regulation

The reversible modification of mA exerts functional impact on several aspects of mRNA metabolism, including nuclear export, polyadenylation, splicing, degradation, and translation (Figure 2). By these means, the mA methylome affords an additional level of gene expression regulation to sculpt the transcriptome (Fu et al., 2014).

N-Methyladenosine Modification Regulates mRNA Splicing

Some factors involved in mA modification of mRNA are known to interact with pre-mRNA splicing factors (SRSFs), indicating a possible role for mA in mRNA splicing (Zhao et al., 2014; Xiao et al., 2016). It has been demonstrated that enrichment of mA modification promotes recruitment of SRSF2 and leads to enhanced exon inclusion of target mRNA (Zhao et al., 2014). It has been further suggested that the mA reader YTHDC1 regulates the association of mA and SRSFs. Indeed, mA-bound YTHDC1 enhances the recruitment of SRSF3 that favors exon inclusion but blocks the recruitment of SRSF10, an exon skipping-related splicing factor (Xiao et al., 2016). Moreover, hnRNPs may also be involved in the regulation of RNA splicing (Liu et al., 2015, 2017). For example, the modification of mA on pre-mRNA favors the binding of hnRNPC (Liu et al., 2015), which could further facilitate splicing through its known function in repressing exon inclusion (Zarnack et al., 2013). Therefore, it is possible that perturbation of the mA machinery components can impair mRNA alternative splicing. This idea is especially supported by the observation that knockdown of METTL3 can antagonize the association of SRSF2 or SRSF3 with mA-modified pre-mRNAs (Zhao et al., 2014; Xiao et al., 2016), and facilitates the expression of the long isoform of MyD88 (MyD88L) via exon skipping attenuation (Feng et al., 2018). Additional evidence is also based on the essential role played by METTL16 in MAT2A-mediated pre-mRNA alternative splicing (Pendleton et al., 2017).

N-Methyladenosine Controls Alternative Polyadenylation

Ke et al. (2015) found that mA modification peaks in the 3′ UTR, especially for transcripts that use alternative polyadenylation (APA), and longer last exons exhibit a higher mA density. By comparing the mA density of thousands of mRNA UTRs from liver and brain tissues, it was observed that greater amount mA marks in the last exons are linked to the usage of more distal polyA sites. Indeed, global reduction of mA levels via triple knockdown of METTL3, METTL14, and WTAP changed the polyA sites of one-sixth of the examined 661 mRNAs and promoted the usage of proximal APA sites, indicating that some mA marks inhibit proximal polyadenylation (Ke et al., 2015).

Recently, a mechanism through which mA controls alternative polyadenylation was proposed. VIRMA (Figure 1A) was found to interact with polyA cleavage factors F5 and CPSF6 (Yue et al., 2018). Consistent with an earlier report, knockdown of METTL3 or VIRMA was found to encourage the usage of distal APA sites, thus lengthening the 3′ UTR of mA-rich mRNAs. In contrast, CPSF5 knockdown elicits an opposite effect on the length of the 3′ UTR of mA-marked mRNAs (Yue et al., 2018).

N-Methyladenosine Promotes Nuclear Export

Considerable amount of nuclear export of mRNAs is regulated by the THO/TREX complex and the nuclear export factor heterodimer NXF1/P15 (Lesbirel and Wilson, 2019). Evidence is accumulating for the role of mA modification in mRNA nuclear export. Knockdown of METTL3 resulted in delayed nuclear export of specific mRNAs of clock genes (Fustin et al., 2013), indicating the requirement of mA methylation for specific mRNA nuclear export. Conversely, knockdown of ALKBH5 increased the cytoplasmic accumulation of polyA mRNAs (Zheng et al., 2013). Moreover, VIRMA was observed to interact with the mRNA export factor ALYREF, and its downregulation led to defective mRNA export (Masuda et al., 2005).

Interestingly, several TREX components associate with the components of the core mA machinery (METTL3-METTL14-WTAP-VIRMA), and TREX also enhances the association of mA reader YTHDC1 with the mRNA. Knockdown of YTHDC1 also resulted in reduced nuclear export of specific mRNAs (Lesbirel et al., 2018). Taken together, the abovementioned literature demonstrates that mA modification factors promote mRNA nuclear transport through physical interaction with the mRNA transport machinery.

N-Methyladenosine Enhances mRNA Degradation

Numerous recent studies suggest that impaired mA writer complex function reduces mA modification levels and raises mRNA stability, indicating that mA methylation drives mRNA degradation (Batista et al., 2014; Schwartz et al., 2014; Wang X. et al., 2014; Wang Y. et al., 2014; Park et al., 2019). Mechanistically, mA-containing mRNA recruits YTHDF2, which is followed by the translocation of the YTHDF2–mRNA complex from the translation machinery to P-bodies, leading to the degradation of YTHDF2-targeted mRNA. As a result, mRNA targets have increased half-life following YTHDF2 knockdown (Wang X. et al., 2014). It has been clearly demonstrated that YTHDF2 enhances mA-modified mRNA decay through recruiting CCR4-NOT deadenylase complex via the N-terminus of YTHDF2 and reveals an underlying mechanism by which YTHDF2 regulates degradation of mA-modified mRNAs (Du et al., 2016).

In a recent study, it was reported that some mA-modified mRNAs interact with YTHDF2 to undergo decay in an RNase P/MRP-dependent manner and in which HRSP12 serves as a bridge between YTHDF2 and RNase P/MRP (Park et al., 2019). The interaction of human YTHDF2 and HRSP12 was first hinted by the association between their respective yeast homologs Pho92 and Mmf1 (Krogan et al., 2006). It was found in an immunoprecipitation experiment that HRSP12 links YTHDF2 and RNase P/MRP and that the N-terminus of YTHDF2 is required to interact with HRSP12 (Park et al., 2019). Of note, the subset of mA-modified mRNAs, whose decay depends on YTHDF2–HRSP12–RNase P/MRP complex, contains a specific HRSP12-binding motif proximally upstream of the YTHDF2-binding motif, while the RNase P/MRP cleavage site is downstream and close to the YTHDF2-binding motif (Park et al., 2019). Therefore, this study discloses at least two mechanisms involved in the degradation of YTHDF2-associated mA RNAs: HRSP12-RNase P/MRP-dependent and CCR4-NOT complex-dependent.

N-Methyladenosine Modulates Translation

The mA reader YTHDF1 enhances translation efficiency via interaction with eIF3A/eIF3B, and the YTHDF1-regulated translation likely hinges on eIF4G-dependent loop formation (Wang et al., 2015). According to Meyer et al. (2015), 5′ UTR mA elevates cap-independent translation through recruiting the 43S complex to form 48S initiation complex in the absence of the cap-associating complex, eIF4F. This mechanism is important for cells to bypass 5′ cap-binding factors to enhance translation under stress conditions (Meyer et al., 2015). Moreover, heat stress-induced cytoplasmic-to-nuclear translocation of YTHDF2 is required for maintaining 5′ UTR mA levels via competing for binding of the demethylase FTO to mA sites, which further promotes cap-independent translation (Zhou et al., 2015). YTHDF1 preferentially binds to mA marks in 3′ UTR of the oncogene CDCP1 mRNA and promotes translation by increasing the amount of polysome-bound (translationally active) CDCP1 transcripts (Yang et al., 2019).

Of note, METTL3 is also involved in mA-enhanced mRNA translation through its role as an mA reader in several ways. It promotes mRNA translation via physical association with the translation initiation complex (Lin et al., 2016). It was found that promoter-associated METTL3 regulates mA methylation inside the coding region and improves mRNA translation through relief of ribosome stalling (Barbieri et al., 2017).

Besides promoting translation efficiency, mA modification also plays an important role in regulating alternative translation (Zhou J. et al., 2018). It has been reported that widespread alternative translation occurs under various nutrient conditions, but the underlying mechanism is unclear (Gao et al., 2015). Recently, Zhou J. et al. (2018) found that mA modification in the 5′ UTR modulates the selection of start codon globally, hence driving alternative translation. As representative examples, Atf4 depends on decreased mA modification of the upstream open reading frame 2 (uORF2) to improve the translation of the major isoform, and Gadd45g heightens the translation of the major isoform by lowering the mA modification of the 5′ UTR (Zhou J. et al., 2018).

N-Methyladenosine Methylation Increases the Phase Separation Capacity of mRNA

Only until recently has it become clearer how mA modification drives mRNA fate and why the consequence of mA modifications can vary in various scenarios. According to Ries et al. (2019), the mA readers, YTHDF1−3, experience liquid–liquid phase separation (LLPS). The mRNAs with multiple mA marks serve as a scaffold to bind with YTHDF readers via their low-complexity regions (LCRs). The mRNA–YTHDF complexes are then transported into various phase separators, like P-bodies, stress granule, and neuronal granules. The study suggests that the number and allocation of mA modifications in mRNAs remodel the transcriptome of different phase-separated compartments. The efficacy of mA modification-dependent modulation of an mRNA is likely governed by signals regulating the ability of YTHDF protein involved in LLPS formation (Ries et al., 2019).

N-Methyladenosine Modification Regulates mRNA Splicing

Some factors involved in mA modification of mRNA are known to interact with pre-mRNA splicing factors (SRSFs), indicating a possible role for mA in mRNA splicing (Zhao et al., 2014; Xiao et al., 2016). It has been demonstrated that enrichment of mA modification promotes recruitment of SRSF2 and leads to enhanced exon inclusion of target mRNA (Zhao et al., 2014). It has been further suggested that the mA reader YTHDC1 regulates the association of mA and SRSFs. Indeed, mA-bound YTHDC1 enhances the recruitment of SRSF3 that favors exon inclusion but blocks the recruitment of SRSF10, an exon skipping-related splicing factor (Xiao et al., 2016). Moreover, hnRNPs may also be involved in the regulation of RNA splicing (Liu et al., 2015, 2017). For example, the modification of mA on pre-mRNA favors the binding of hnRNPC (Liu et al., 2015), which could further facilitate splicing through its known function in repressing exon inclusion (Zarnack et al., 2013). Therefore, it is possible that perturbation of the mA machinery components can impair mRNA alternative splicing. This idea is especially supported by the observation that knockdown of METTL3 can antagonize the association of SRSF2 or SRSF3 with mA-modified pre-mRNAs (Zhao et al., 2014; Xiao et al., 2016), and facilitates the expression of the long isoform of MyD88 (MyD88L) via exon skipping attenuation (Feng et al., 2018). Additional evidence is also based on the essential role played by METTL16 in MAT2A-mediated pre-mRNA alternative splicing (Pendleton et al., 2017).

N-Methyladenosine Controls Alternative Polyadenylation

Ke et al. (2015) found that mA modification peaks in the 3′ UTR, especially for transcripts that use alternative polyadenylation (APA), and longer last exons exhibit a higher mA density. By comparing the mA density of thousands of mRNA UTRs from liver and brain tissues, it was observed that greater amount mA marks in the last exons are linked to the usage of more distal polyA sites. Indeed, global reduction of mA levels via triple knockdown of METTL3, METTL14, and WTAP changed the polyA sites of one-sixth of the examined 661 mRNAs and promoted the usage of proximal APA sites, indicating that some mA marks inhibit proximal polyadenylation (Ke et al., 2015).

Recently, a mechanism through which mA controls alternative polyadenylation was proposed. VIRMA (Figure 1A) was found to interact with polyA cleavage factors F5 and CPSF6 (Yue et al., 2018). Consistent with an earlier report, knockdown of METTL3 or VIRMA was found to encourage the usage of distal APA sites, thus lengthening the 3′ UTR of mA-rich mRNAs. In contrast, CPSF5 knockdown elicits an opposite effect on the length of the 3′ UTR of mA-marked mRNAs (Yue et al., 2018).

N-Methyladenosine Promotes Nuclear Export

Considerable amount of nuclear export of mRNAs is regulated by the THO/TREX complex and the nuclear export factor heterodimer NXF1/P15 (Lesbirel and Wilson, 2019). Evidence is accumulating for the role of mA modification in mRNA nuclear export. Knockdown of METTL3 resulted in delayed nuclear export of specific mRNAs of clock genes (Fustin et al., 2013), indicating the requirement of mA methylation for specific mRNA nuclear export. Conversely, knockdown of ALKBH5 increased the cytoplasmic accumulation of polyA mRNAs (Zheng et al., 2013). Moreover, VIRMA was observed to interact with the mRNA export factor ALYREF, and its downregulation led to defective mRNA export (Masuda et al., 2005).

Interestingly, several TREX components associate with the components of the core mA machinery (METTL3-METTL14-WTAP-VIRMA), and TREX also enhances the association of mA reader YTHDC1 with the mRNA. Knockdown of YTHDC1 also resulted in reduced nuclear export of specific mRNAs (Lesbirel et al., 2018). Taken together, the abovementioned literature demonstrates that mA modification factors promote mRNA nuclear transport through physical interaction with the mRNA transport machinery.

N-Methyladenosine Enhances mRNA Degradation

Numerous recent studies suggest that impaired mA writer complex function reduces mA modification levels and raises mRNA stability, indicating that mA methylation drives mRNA degradation (Batista et al., 2014; Schwartz et al., 2014; Wang X. et al., 2014; Wang Y. et al., 2014; Park et al., 2019). Mechanistically, mA-containing mRNA recruits YTHDF2, which is followed by the translocation of the YTHDF2–mRNA complex from the translation machinery to P-bodies, leading to the degradation of YTHDF2-targeted mRNA. As a result, mRNA targets have increased half-life following YTHDF2 knockdown (Wang X. et al., 2014). It has been clearly demonstrated that YTHDF2 enhances mA-modified mRNA decay through recruiting CCR4-NOT deadenylase complex via the N-terminus of YTHDF2 and reveals an underlying mechanism by which YTHDF2 regulates degradation of mA-modified mRNAs (Du et al., 2016).

In a recent study, it was reported that some mA-modified mRNAs interact with YTHDF2 to undergo decay in an RNase P/MRP-dependent manner and in which HRSP12 serves as a bridge between YTHDF2 and RNase P/MRP (Park et al., 2019). The interaction of human YTHDF2 and HRSP12 was first hinted by the association between their respective yeast homologs Pho92 and Mmf1 (Krogan et al., 2006). It was found in an immunoprecipitation experiment that HRSP12 links YTHDF2 and RNase P/MRP and that the N-terminus of YTHDF2 is required to interact with HRSP12 (Park et al., 2019). Of note, the subset of mA-modified mRNAs, whose decay depends on YTHDF2–HRSP12–RNase P/MRP complex, contains a specific HRSP12-binding motif proximally upstream of the YTHDF2-binding motif, while the RNase P/MRP cleavage site is downstream and close to the YTHDF2-binding motif (Park et al., 2019). Therefore, this study discloses at least two mechanisms involved in the degradation of YTHDF2-associated mA RNAs: HRSP12-RNase P/MRP-dependent and CCR4-NOT complex-dependent.

N-Methyladenosine Modulates Translation

The mA reader YTHDF1 enhances translation efficiency via interaction with eIF3A/eIF3B, and the YTHDF1-regulated translation likely hinges on eIF4G-dependent loop formation (Wang et al., 2015). According to Meyer et al. (2015), 5′ UTR mA elevates cap-independent translation through recruiting the 43S complex to form 48S initiation complex in the absence of the cap-associating complex, eIF4F. This mechanism is important for cells to bypass 5′ cap-binding factors to enhance translation under stress conditions (Meyer et al., 2015). Moreover, heat stress-induced cytoplasmic-to-nuclear translocation of YTHDF2 is required for maintaining 5′ UTR mA levels via competing for binding of the demethylase FTO to mA sites, which further promotes cap-independent translation (Zhou et al., 2015). YTHDF1 preferentially binds to mA marks in 3′ UTR of the oncogene CDCP1 mRNA and promotes translation by increasing the amount of polysome-bound (translationally active) CDCP1 transcripts (Yang et al., 2019).

Of note, METTL3 is also involved in mA-enhanced mRNA translation through its role as an mA reader in several ways. It promotes mRNA translation via physical association with the translation initiation complex (Lin et al., 2016). It was found that promoter-associated METTL3 regulates mA methylation inside the coding region and improves mRNA translation through relief of ribosome stalling (Barbieri et al., 2017).

Besides promoting translation efficiency, mA modification also plays an important role in regulating alternative translation (Zhou J. et al., 2018). It has been reported that widespread alternative translation occurs under various nutrient conditions, but the underlying mechanism is unclear (Gao et al., 2015). Recently, Zhou J. et al. (2018) found that mA modification in the 5′ UTR modulates the selection of start codon globally, hence driving alternative translation. As representative examples, Atf4 depends on decreased mA modification of the upstream open reading frame 2 (uORF2) to improve the translation of the major isoform, and Gadd45g heightens the translation of the major isoform by lowering the mA modification of the 5′ UTR (Zhou J. et al., 2018).

N-Methyladenosine Methylation Increases the Phase Separation Capacity of mRNA

Only until recently has it become clearer how mA modification drives mRNA fate and why the consequence of mA modifications can vary in various scenarios. According to Ries et al. (2019), the mA readers, YTHDF1−3, experience liquid–liquid phase separation (LLPS). The mRNAs with multiple mA marks serve as a scaffold to bind with YTHDF readers via their low-complexity regions (LCRs). The mRNA–YTHDF complexes are then transported into various phase separators, like P-bodies, stress granule, and neuronal granules. The study suggests that the number and allocation of mA modifications in mRNAs remodel the transcriptome of different phase-separated compartments. The efficacy of mA modification-dependent modulation of an mRNA is likely governed by signals regulating the ability of YTHDF protein involved in LLPS formation (Ries et al., 2019).

N-Methyladenosine Modification Prominently Regulates Brain Development and Function

Evidence for the role of mA signaling in modulating the development of the brain and its functions has accumulated in recent years, and the quest for extending the frontier is of great interest. Several investigations have revealed that the various factors that come together to form the mA methylation machinery exert notable effect(s) on specific aspects of brain morphogenesis to permit optimal neural function, as summarized in Table 1. Conversely, the dysregulation of the mA methylation machinery is known to elicit perturbations in the neural transcriptome, which have implications for defective development and dysfunction of the brain. The integrity of the mA machinery functionality is of high priority in cells to the extent that simply ablating its cofactors can have significant consequences for brain development disturbance, as exemplified by the importance of Exosc10-mediated regulation of mRNA stability in forebrain development (Ulmke et al., 2021). The sections below discuss how specific factors associated with the mA methylation machinery drive neural development, functional adaptation, and plasticity of the brain (Figure 3).

TABLE 1

mA mRNA methylation factors and their role in brain development and function.

EffectorExperimental manipulationPhenotypeMechanismsReferences
Neurogenesis
METTL3Mettl3; Nestin-CreProlongation of the cell cycle of RGCs and protraction of embryonic cortical neurogenesismA depletion caused increased stability of NSC transcriptsYoon et al. (2017)
METTL14Mettl14; Nestin-CreReduced NSC proliferation and precocious NSC differentiation; loss of late-born neurons during cortical neurogenesisStabilization of CBP and p300 transcripts; H3K27me3-mediated transcription suppression of NSC proliferation genes; upregulation of H3K27ac in differentiation-related genesWang Y. et al. (2018)
RBM15OE of RBM15NSC delaminationSuppression of BAF155-dependent gene expressionXie et al. (2019)
FTOKO of FTODecrease in adult NSC proliferation and defective hippocampal neurogenesisImpairment of BDNF and MAPK signalingLi L. et al. (2017); Spychala and Ruther (2019)
YTHDF2Ythdf2 ; Cre (ubiquitously)Decreased proliferation and differentiation capabilities of NSCs; less Tbr2+ bIPs; Reduced CP thicknessPromotion of mA-dependent degradation of neurodevelopment-related transcriptsLi M. et al. (2018)
FMRPKO of Fmr1Nuclear retention of neurogenic mRNAs; prolonged cell cycle progression in the postnatal mouse brainUnknownEdens et al. (2019)
Exosc10Exosc10; Foxg1-Cre; Emx1-CreApoptosis-mediated cortical agenesisMediates degradation of Bbc3 and Aen mRNAs, which are effectors of apoptosisUlmke et al. (2021)
Gliogenesis
METTL14Mettl14; Olig2-CreDecrease in oligodendrocytes maturation; cortical hypomyelinationAlters alternative splicing and expression of Nfasc155Xu et al. (2020)
METTL14Mettl14; Nestin-CreReduced number of s100β+ astrocytesUnknownYoon et al. (2017)
FTOFTO; Olig2-Cre; Nestin-CreLoss of OPCs and Sox10+ cells; cortical hypomyelinationPromotes Olig2 mRNA degradationWu et al. (2019)
PRRC2APrrc2a; Nestin-Cre; Olig2-CreLoss of OPCs and mature oligodendrocytes; cortical hypomyelinationGene targeting of Prrc2a by olig2 mRNAWu et al. (2019)
PRRC2APrrc2a; Nestin-Cre;Reduced proliferation capacity and number of astrocytesCompetitive expression with YTHDF2Wu et al. (2019)
Axonogenesis, dendritogenesis, synaptic plasticity
YTHDF1 YTHDF3KD of Ythdf1 and Ythdf3Abnormal dendritic spine morphologyInhibition of Apc mRNA translationMerkurjev et al. (2018)
METTL14Mettl14; D1R-CreAbnormal excitability of striatal neuronsUnknownKoranda et al. (2018)
FTOKO of FTODefective synaptic plasticityUnknownHess et al. (2013)
YTHDF1 YTHDF3KD of Ythdf1 and Ythdf3; CRISPR/Cas9-based KO of Ythdf1Suppression of neuronal excitabilityNot clearMerkurjev et al. (2018); Shi et al. (2018)
Learning and behavior
METTL3OE of METTL3Improved long-term memory consolidationUnknownZhang Z. Y. et al. (2018)
METTL14Mettl14; D1R-CreImpaired striatum-mediated behavior patternsUnknownKoranda et al. (2018)
FTOCRISPR/Cas9 or shRNA-mediated KD of FTOLearning disabilities; defective memory processing and verbal fluencyNot clearHo et al. (2010); Benedict et al. (2011); Widagdo et al. (2016); Li L. et al. (2017); Walters et al. (2017); Sun et al. (2019)
PRRC2APrrc2a; Nestin-Cre; Olig2-CreCognitive defects due to cortical hypomyelinationUnknownWu et al. (2019)
YTHDF1CRISPR/Cas9-based KO of Ythdf1; KD of Ythdf1Defective long-term potentiation and synaptic transmission in hippocampus; behavioral defectsUnknownShi et al. (2018)
Circadian clock
METTL3KD of Mettl3Elongation of circadian periodDefective processing of Per2 and Arntl (clock genes) mRNAsFustin et al. (2013)
Stress response
FTOFTO; Camk2a-Cre; Nex-CreERT2Reduced ability to cope with stressUnknownEngel et al. (2018)
METTL3Mettl14; Camk2a-Cre; Nex-CreERT2Reduced ability to cope with stressUnknownEngel et al. (2018)
KO, knockout; KD, knockdown; OE, overexpression; fl/fl, double floxed; RGCs, radial glial cells; OPCs, oligodendrocyte precursor cells; NSCs, neural stem cells; CRISPR, clustered regularly interspaced short palindromic repeats; mA, N-methyladenosine.
An external file that holds a picture, illustration, etc.
Object name is fcell-09-656849-g003.jpg

An illustration summarizing the role of N-methyladenosine (mA) in brain development and function. The functions of the various mA-related factors involved in the proliferation and differentiation of neural precursors, neuronal maturation, production of glia, synapse formation during brain development, and common brain physiology are shown.

N-Methyladenosine Modification Is Indispensable for Neurogenesis in the Brain

Neurons are produced through the process of neurogenesis, which entails specification and proliferation of NSCs, and the differentiation of such neural progenitors into neuroblasts, which undergo maturation to become functional neurons. It has been shown that the dynamic addition of mA to gene transcripts in the multipotent NSCs greatly influences cortical neuroprogenitor competence and the generation of neurons during brain development (Yao et al., 2016; Boles and Temple, 2017; Yoon et al., 2018; Zhou H. et al., 2018; Rockwell and Hongay, 2019). Dysregulation of writers, erasers, and readers of mA has been reported to cause notable perturbations in the cell cycle progression, proliferation, and differentiation of NSCs in the developing and adult brain.

Effect of N-Methyladenosine Writers on Neurogenesis in the Brain

So far, it has been shown that ablation of the mA writer METTL3 or its cofactor METTL14 in cortical neuroepithelium or isolated cortical NSC results in prolonged cell cycle dynamics of cortical neuroprogenitors and their precocious differentiation into neuronal or neurogenic cells (Batista et al., 2014; Yoon et al., 2017; Wang Y. et al., 2018). Detailed analysis through mA sequencing revealed that gene transcripts involved in the cell cycle of neural cells, production of neurons, and neuronal differentiation are enriched with mA tags. Interestingly, the decay of such mRNAs is promoted in the absence of METTL3 and METTL14 (Yoon et al., 2017), meaning that METTL3 and METTL14 are key players in driving neurogenesis via the stabilization of gene transcripts critical for neurogenesis in the brain. For example, loss of mA due to deletion of METTL3 in mouse cerebellum resulted in overt hypoplasia partly attributable to apoptosis of cerebellar granule cells (Wang C. X. et al., 2018). Key downstream effects of mA on genes important for neurogenesis include the modulation of histone modification in the promoter environment of NSC proliferation- and differentiation-related gene loci (Wang Y. et al., 2018). In the absence of METTL14, the transcription-suppressing histone mark H3K27me3 is upregulated on genes involved in cell proliferation, whereas differentiation-related genes show an increase in the transcription activation histone mark H3K27ac when METTL14 is deficient (Wang Y. et al., 2018). Lack of METTL3 in the developing brain can also cause the aforementioned histone alterations, at least in terms of H3K27me3 enhancement, which can cause transcription repression (Chen J. et al., 2019). This is possible because in the absence of METTL3, which leads to a reduction in mA levels, the polycomb repressor complex becomes hyperactive due to derepression of its core methlytransferase factor Ezh2 (Chen J. et al., 2019).

RBM15, a core component of the mA writer complex (Figure 1A), is a potential regulator of cortical neurogenesis due to its distinctive expression in the cortical germinative zone and cortical plate of the developing mouse cortex (Xie et al., 2019). Knockdown of RBM15 in neurons in vitro promoted endogenous expression of the chromatin remodeling factor BAF155 (Xie et al., 2019), which is a known key regulator of cortical development (Nguyen et al., 2016, 2018; Narayanan et al., 2018). This profound effect can be linked to a significant reduction in cellular levels of mA due to the inactivation of RBM15 (Knuckles et al., 2018). However, overexpression of RBM15 in vivo was found to promote delamination of radial glial cells in the cortical ventricular zone by suppressing the expression BAF155 and, hence, BAF155-dependent gene expression program supportive for cortical development (Xie et al., 2019). The role of RBM15 in cortical neurogenesis further highlights the contribution of mA methyltransferase in brain development.

N-Methyladenosine Erasers Regulate Neurogenesis in the Brain

Erasers of the mA mark (FTO and ALKBH5) can also exert a regulatory effect on the process of neurogenesis given their prominent expression in neurons (Li L. et al., 2017; Yoon et al., 2017; Spychala and Ruther, 2019; Du et al., 2020). Whereas FTO displays the highest expression level late in brain neurogenesis (Li L. et al., 2017; Yoon et al., 2017), ALKBH5 expression decreases in the course of brain development (Du et al., 2020). This may have implications for their roles in the spatiotemporal regulation of neurogenesis during brain development. Indeed, it was reported that FTO deficiency in the adult mouse brain induces signal transducer and activator of transcription (STAT)3 pathway activation via its modulators platelet-derived growth factor receptor (PDGFR)α and suppressor of cytokine signaling (SOCS)5 in an mA-dependent manner (Cao et al., 2020). As a result, a transient increase in the proliferation and differentiation of adult NSCs was observed in the FTO mutant brain, with implications for adult neurogenesis inhibition in the long term (Cao et al., 2020). It was also observed that FTO deletion in adult mouse brain impairs brain-derived neurotrophic factor (BDNF) and mitogen-activated protein kinase (MAPK) signaling pathways, leading to a reduction in adult NSC proliferation and neurogenesis in the hippocampal formation (Li L. et al., 2017; Spychala and Ruther, 2019). Although these studies report diverging effects of FTO loss on adult NSC proliferation, they both show a resultant effect of adult neurogenesis reduction. We think that, while being mindful of the low level of FTO expression in the early developing cortex, conducting an investigation on how FTO regulates corticogenesis in the course of development may lend clarity to how it mechanistically impacts cortical neurogenesis.

Notable N-Methyladenosine Readers in Cortical Neurogenesis

Protein factors that act as readers of the mA mark have also been shown to have a profound effect on neurogenesis in the brain. For instance, the mA reader YTHDF2 has been reported to be indispensable for corticogenesis in mouse. Conditional knockout of YTHDF2 in the mouse neocortical neuroepithelium resulted in a reduction in the proliferation and differentiation of the Ythdf2/ neuroprogenitor cells (Li M. et al., 2018). This phenotype may have mechanistic underpinnings, including abnormal upregulation of genes that inhibit the JAK–STAT signaling pathway, due to increased stability of such gene transcripts in the absence of YTHDF2 (Li M. et al., 2018). Yet, it seems that the induction of neural fate in pluripotent stem cells requires downregulation of YTHDF2, leading to the stability and expression of neural gene transcripts (Heck et al., 2020). We are of the opinion that the functional consequence of the mA reading by YTHDF2 may be contextually variable along the cortical development axis such that reduced dosage may support neural cell fate specification, whereas its increased activity/expression is necessary for later cortical neurodevelopment.

Another mA reader, FMRP, was identified to be critical for neural progenitor cell proliferation. Mice lacking Fmr1 displayed prolonged cell cycle progression. As a result, proliferation of neural progenitors extended into postnatal stages of brain development (Edens et al., 2019). Of note, it was observed that nuclear export of mA-modified neurogenic mRNAs readable by FMRP is defective, leading to retention of such neurodifferentiation gene transcripts in the nucleus of the Fmr1-deficient neural progenitor cells (Edens et al., 2019). Lastly, the mA reader protein Imp (IGF2BP) was identified as a key regulator of NSC proliferation rate through the stabilization of Myc mRNA in Drosophila brain neuroblasts (Samuels et al., 2020).

Together, the mA machinery has been identified to play critical roles in brain morphogenesis by regulating the proliferation of neural progenitor cells and the production of neurons. As such, hypomethylation due to METTL3 or METTL14 deficiency and aberrant mA reading or erasure in the embryonic or adult brain can precipitate phenotypes, including defective transcriptional prepatterning, abnormal neuroprogenitor pool, impaired neurogenesis, and cortical hypoplasia (Yoon et al., 2017), which can engender deficits in brain structure and function.

N-Methyladenosine Signaling Is Essential for Gliogenesis in the Brain

The process of generating glial cells constitutes gliogenesis. Brain neuroglia include astrocytes and oligodendrocytes, which are derived from the neuroepithelium. During cortical development, a switch from neurogenesis to gliogenesis coincides with a decrease in mA modification of proneural genes (Donega et al., 2018). Although mA enrichment in glial cells is less than that observed in neurons (Chang et al., 2017), a few studies have uncovered the importance of the mA methylome in brain gliogenesis, at least for astrocyte production (astrogenesis) and oligodendrocyte generation (oligodendrogenesis).

Regulation of Glia Production in the Brain by an N-Methyladenosine Writer-Related Factor

It was observed that loss of METTL14-mediated mA writing in the mouse cortex leads to hypomyelination that can be linked to a reduction in the number of (mature) oligodendrocytes (Xu et al., 2020). The loss of oligodendrocytes caused by the absence of METTL14 in the brain likely did not emanate from abnormal specification or proliferation of oligodendrocyte precursor cells (OPCs) (Xu et al., 2020). Notably, the transcriptome of OPCs and oligodendrocytes is altered following METTL14 deletion, with possible impact on gene expression programs critical for oligodendrocyte lineage progression (Xu et al., 2020). Lack of METTL14 has also been reported to disrupt astrogenesis. Indeed, s100β-expressing astrocytic progenitors were found to be reduced in the METTL14 knockout mouse cortex at postnatal stage 5 (Yoon et al., 2017). It would be interesting to investigate whether other mA writer-related factors, including METTL3, have roles to play in cortical gliogenesis.

The N-Methyladenosine Eraser FTO Regulates Glia Production in the Brain

mA-mediated RNA methylation dynamics under the guild of FTO is known to influence oligodendrogenesis via modulation of the half-life of Olig2 mRNA (Wu et al., 2019). Olig2 is a central factor indispensable for oligodendrocyte lineage progression (Liu et al., 2007). Specifically, FTO was reported to regulate the degradation of Olig2 transcripts via removal of mA tags installed on the Olig2 mRNA. The stability of Olig2 transcripts in OPCs deficient in FTO was thus seen to increase. In effect, the white matter in FTO mutant mouse brain was characterized by hypomyelination (Wu et al., 2019).

Involvement of N-Methyladenosine Readers in Glia Production in the Brain

The mA reader PRRC2A is known to be essential for oligodendrogenesis. It prominently regulates the specification, proliferation, and differentiation of oligodendroglia and the ability of oligodendrocytes to carry out myelination in the brain (Wu et al., 2019). More specifically, abolishing PRRC2A function in cortical NSCs or precisely in oligodendroglial lineage caused significant loss of OPCs (PDGFRα+ cells), Sox10+ cells, and mature oligodendrocytes (CC1+Oilg2+), which culminated in hypomyelination in the PRRC2A mutant brain (Wu et al., 2019). Interestingly, deletion of PRRC2A also affects astrogenesis, although slightly. Deficiency of PRRC2A in mouse brain caused a reduction in the proliferative capacity of astrocytes, leading to a reduced number of astrocytes in the mutant mouse brain (Wu et al., 2019). The additional role of PRRC2A in regulating the production of astrocytes in the brain during development may hinge on its interaction with YTHDF2, another mA-binding protein, such that lack of either mA reader augments the expression of the other to influence gliogenesis (Wu et al., 2019).

The competitive relationship between PRRC2A and YTHDF2 makes it complex to explain or reconcile the observation that glial fibrillary acidic protein (GFAP) expression, which can indicate astrocytic cells, was found to be dramatically reduced in neurospheres derived from the E14.5 Ythdf2–/– forebrain NSC. Such GFAP+ Ythdf2–/– cells also displayed abnormally branched processes (Li M. et al., 2018). Thus, further investigation is required to elucidate the role of YTHDF2 in brain gliogenesis and how the function of PRRC2A features in the regulatory pathway.

N-Methyladenosine Effectors Regulate the Formation of Neural Processes and Synapses

The developing and adult brain is characterized by the outgrowth of dendrites and axons of neurons known to form neural connections called synapses. Interestingly, synapses are enriched with mA, which modulates dendrite formation (dendritogenesis), axonogenesis, and synaptic growth (synaptogenesis) and activity (reviewed in Li et al., 2019; Dermentzaki and Lotti, 2020). mA-based transcriptome profiling of the mouse brain (cortex and cerebellum) showed enrichment of mA modification linked to dendrite and dendritic spine, axon and axon guidance, and synaptogenesis and synaptic transmission (Chang et al., 2017).

Distinctive localization of the YTHDFs, FTO, and METTL14 in dendrites of hippocampal neurons in culture and cortical neurons suggests the involvement of these mA-regulatory factors in the development of neural dendrites. Indeed, Ythdf1 and Ythdf3 knockdown in such cultured neurons resulted in abnormal dendritic spine (Merkurjev et al., 2018). Axons are also enriched with FTO, which can be translated locally. As such, FTO ablation in axons resulted in upregulation of mA levels, leading to a reduction in Gap-43 mRNA translation in axons of cultured dorsal root ganglion neurons (Yu et al., 2018). Yet, GAP-43 is a key factor involved in axon growth in neural tissues (Skene et al., 1986). In effect, the neurons lacking FTO displayed axon elongation repression (Yu et al., 2018). The mA reader YTHDF1 was also reported to influence axon formation by binding and promoting the translation of the axon guidance receptor Robo3.1, which directs spinal commissural axons in crossing the midline, in an mA modification-dependent manner (Zhuang et al., 2019). Together, these observations may have implications for perturbed axonogenesis in the brain lacking optimal mA modification due to ablation of FTO or YTHDF1. At least in the case of the mA-regulatory protein PRRC2A, it was found that axons that form the corpus callosum, a brain midline structure, are hypomyelinated and appeared hypoplastic in the PRRC2A-deleted mouse brain (Wu et al., 2019).

Given the enrichment of mA marks and related proteins in neural processes, it is not surprising that synapses are endowed with mA-modified mRNAs, especially postsynaptic transcripts in the mouse brain (Chang et al., 2017). The high localization of mA-modified mRNAs in synapses reflects the possible impact of the mA epitranscriptome on the structure, maturation, and function of synapses (Chang et al., 2017; Merkurjev et al., 2018; Yu et al., 2018; Zhuang et al., 2019). As a result, selective ablation of YTHDF1 and YTHDF3 in the cultured hippocampal neurons caused excitatory synaptic transmission suppression (Merkurjev et al., 2018; Shi et al., 2018). In addition, synapses formed by neurons lacking YTHDF2 appeared abnormal (Li M. et al., 2018), and synaptic transmission-related transcripts are hypermethylated in dopaminergic neurons with defective synaptic plasticity implication in the FTO-deficient mouse midbrain (Hess et al., 2013). Another indication of synapse malformation and synaptic plasticity impairment due to mA dysregulation was observed in METTL14-deleted striatal neurons, in which METTL14 abrogation resulted in aberrant neuronal excitability (Koranda et al., 2018). Given that Nito, the Drosophila version of RBM15, also regulates synaptic growth through regulation of axonogenesis (Gu et al., 2017), it would be interesting to investigate whether indeed RBM15 is involved in synaptogenesis in the mammalian brain.

Cognition and Behavior Are Modulated by N-Methyladenosine Signaling

The brain’s ability to process and store information and form or control behavior patterns has been shown to be greatly regulated by posttranscriptional modification of mRNA involved in brain development (reviewed in Jung and Goldman, 2018; Leighton et al., 2018; Noack and Calegari, 2018). Prominently emerging among these new (epitranscriptomic) levels of brain function regulation is mA modification of mRNA in the brain. Various studies in mouse models have revealed the involvement of the mA machinery-related factors in cognition and behavior (reviewed in Nainar et al., 2016; Chokkalla et al., 2020). The role of mA in the regulation of learning and behavior may be partly explained by the previously discussed role of mA in synaptogenesis and synaptic transmission (Weng et al., 2018).

N-Methyladenosine Writers Involved in Memory and Behavior

In a recent study by Zhang F. et al. (2018), it was found that the enrichment of METTL3 in the mouse hippocampus is supportive for memory consolidation via the promotion of neuronal early-response gene translation. Therefore, mice lacking METTL3 in the hippocampus displayed impaired long-term potentiation with attendant reduced ability to consolidate memory. Interestingly, long-term memory consolidation is demonstrably augmented following METTL3 overexpression in the dorsal hippocampus of the wild-type mouse brain (Zhang Z. Y. et al., 2018). The mA writer function of METTL14 is reported to be important for learning and behavior mediated by the striatum. Without affecting the number or morphology of striatal neurons, loss of METTL14 in striatopallidal and striatonigral neurons caused alterations in the transcriptome, eliciting increased neuronal excitability and spike frequency adaptation reduction, which possibly culminated in impairment of striatum-dependent behavior patterns (Koranda et al., 2018).

The N-Methyladenosine Eraser FTO Regulates Learning and Behavior

Accumulation of mA in the brain can affect its learning capacity and behavior. By regulating adult neurogenesis in the mouse hippocampus, FTO has been identified to play a pivotal role in learning (Li L. et al., 2017). Hypermethylation in the mouse brain or hippocampus caused by FTO functional loss was observed to call forth learning disabilities in mice, including increased fear memory consolidation (Widagdo et al., 2016; Walters et al., 2017). Additional evidence indicating the role of FTO in learning and behavior includes a study in which mice deficient in FTO were reported to exhibit behaviors consistent with depression and anxiety (Sun et al., 2019). Moreover, available data show that memory processing and verbal fluency may be affected in individuals with FTO ablation in the brain (Ho et al., 2010; Benedict et al., 2011).

Readers of N-Methyladenosine Modulate Learning and Memory

Cognitive deficits have been implicitly linked to lack of function of the mA reader PRRC2A, whose absence caused hypomyelination, leading to the cognitive anomalies in the mouse brain (Wu et al., 2019). Evidence indicating a more direct importance of an mA reader in learning and memory was obtained when YTHDF1 was deleted in the adult mouse brain. It was found that neuronal stimuli can evoke translation of gene transcripts readable by YTHDF1 to facilitate learning and memory (Shi et al., 2018). Hence, silencing of YTHDF1 in the mouse hippocampus resulted in defective long-term potentiation and impaired synaptic transmission in the hippocampus, which did not allow normal learning and memory processing, and the defects were rescuable by YTHDF1 re-expression in the YTHDF1 mutant brain (Shi et al., 2018).

Stress Response Is Regulated by Factors of the N-Methyladenosine Machinery

The brain plays a central role in stress response. In responding to stress, a host of gene expression programs is activated in the brain, leading to the secretion of several neuropeptides (de Kloet et al., 2005). Vulnerability to stressful stimuli and the response mechanism can have implications for neuropsychiatric anomalies under abnormal regulatory conditions. Thus, the transcriptomic stress response system is particularly crucial in maintaining homeostasis following exposure to stress.

Epigenetic mechanisms are known to play central roles in stress response (McEwen et al., 2015), and the epitranscriptome is an emerging gene expression regulation domain for stress modulation (Harvey et al., 2017). A putative role for mA in the regulation of stress response is evidenced by the presence of glucocorticoid response elements upstream the transcription start site of genes that encode for enzymes involved in mA modification (Engel et al., 2018). Additionally, nuclear localization of YTHDF2 precipitated by heat stress results in dynamic methylation of the 5′ UTR of newly synthesized mRNAs (Zhou et al., 2015). By limiting FTO, YTHDF2 is able to preserve methylation in the 5′ UTR of heat stress-induced mRNAs (Zhou et al., 2015).

In chick, upregulation of FTO in the brain (hypothalamus) may be a mechanism to afford thermoregulation in heat stress conditions (Kisliouk et al., 2020). However, following acute restraint stress, the mouse prefrontal cortex and amygdala displayed mA hypomethylation and hypermethylation, respectively (Engel et al., 2018). Fear-induced stress can cause downregulation of FTO, leading to elevation of mA in the prefrontal cortex and hippocampus of the mouse (Walters et al., 2017). Mice lacking METTL3 or FTO are unable to cope with stress (Engel et al., 2018). A general effect that may be caused by stress-induced alteration in mA modification is the suppression of mRNAs involved in synaptic plasticity and brain morphogenesis (Engel et al., 2018). Together, the above observations indicate a putative role for mA modulation in the human brain during stressful insults.

N-Methyladenosine Modification Is Indispensable for Neurogenesis in the Brain

Neurons are produced through the process of neurogenesis, which entails specification and proliferation of NSCs, and the differentiation of such neural progenitors into neuroblasts, which undergo maturation to become functional neurons. It has been shown that the dynamic addition of mA to gene transcripts in the multipotent NSCs greatly influences cortical neuroprogenitor competence and the generation of neurons during brain development (Yao et al., 2016; Boles and Temple, 2017; Yoon et al., 2018; Zhou H. et al., 2018; Rockwell and Hongay, 2019). Dysregulation of writers, erasers, and readers of mA has been reported to cause notable perturbations in the cell cycle progression, proliferation, and differentiation of NSCs in the developing and adult brain.

Effect of N-Methyladenosine Writers on Neurogenesis in the Brain

So far, it has been shown that ablation of the mA writer METTL3 or its cofactor METTL14 in cortical neuroepithelium or isolated cortical NSC results in prolonged cell cycle dynamics of cortical neuroprogenitors and their precocious differentiation into neuronal or neurogenic cells (Batista et al., 2014; Yoon et al., 2017; Wang Y. et al., 2018). Detailed analysis through mA sequencing revealed that gene transcripts involved in the cell cycle of neural cells, production of neurons, and neuronal differentiation are enriched with mA tags. Interestingly, the decay of such mRNAs is promoted in the absence of METTL3 and METTL14 (Yoon et al., 2017), meaning that METTL3 and METTL14 are key players in driving neurogenesis via the stabilization of gene transcripts critical for neurogenesis in the brain. For example, loss of mA due to deletion of METTL3 in mouse cerebellum resulted in overt hypoplasia partly attributable to apoptosis of cerebellar granule cells (Wang C. X. et al., 2018). Key downstream effects of mA on genes important for neurogenesis include the modulation of histone modification in the promoter environment of NSC proliferation- and differentiation-related gene loci (Wang Y. et al., 2018). In the absence of METTL14, the transcription-suppressing histone mark H3K27me3 is upregulated on genes involved in cell proliferation, whereas differentiation-related genes show an increase in the transcription activation histone mark H3K27ac when METTL14 is deficient (Wang Y. et al., 2018). Lack of METTL3 in the developing brain can also cause the aforementioned histone alterations, at least in terms of H3K27me3 enhancement, which can cause transcription repression (Chen J. et al., 2019). This is possible because in the absence of METTL3, which leads to a reduction in mA levels, the polycomb repressor complex becomes hyperactive due to derepression of its core methlytransferase factor Ezh2 (Chen J. et al., 2019).

RBM15, a core component of the mA writer complex (Figure 1A), is a potential regulator of cortical neurogenesis due to its distinctive expression in the cortical germinative zone and cortical plate of the developing mouse cortex (Xie et al., 2019). Knockdown of RBM15 in neurons in vitro promoted endogenous expression of the chromatin remodeling factor BAF155 (Xie et al., 2019), which is a known key regulator of cortical development (Nguyen et al., 2016, 2018; Narayanan et al., 2018). This profound effect can be linked to a significant reduction in cellular levels of mA due to the inactivation of RBM15 (Knuckles et al., 2018). However, overexpression of RBM15 in vivo was found to promote delamination of radial glial cells in the cortical ventricular zone by suppressing the expression BAF155 and, hence, BAF155-dependent gene expression program supportive for cortical development (Xie et al., 2019). The role of RBM15 in cortical neurogenesis further highlights the contribution of mA methyltransferase in brain development.

N-Methyladenosine Erasers Regulate Neurogenesis in the Brain

Erasers of the mA mark (FTO and ALKBH5) can also exert a regulatory effect on the process of neurogenesis given their prominent expression in neurons (Li L. et al., 2017; Yoon et al., 2017; Spychala and Ruther, 2019; Du et al., 2020). Whereas FTO displays the highest expression level late in brain neurogenesis (Li L. et al., 2017; Yoon et al., 2017), ALKBH5 expression decreases in the course of brain development (Du et al., 2020). This may have implications for their roles in the spatiotemporal regulation of neurogenesis during brain development. Indeed, it was reported that FTO deficiency in the adult mouse brain induces signal transducer and activator of transcription (STAT)3 pathway activation via its modulators platelet-derived growth factor receptor (PDGFR)α and suppressor of cytokine signaling (SOCS)5 in an mA-dependent manner (Cao et al., 2020). As a result, a transient increase in the proliferation and differentiation of adult NSCs was observed in the FTO mutant brain, with implications for adult neurogenesis inhibition in the long term (Cao et al., 2020). It was also observed that FTO deletion in adult mouse brain impairs brain-derived neurotrophic factor (BDNF) and mitogen-activated protein kinase (MAPK) signaling pathways, leading to a reduction in adult NSC proliferation and neurogenesis in the hippocampal formation (Li L. et al., 2017; Spychala and Ruther, 2019). Although these studies report diverging effects of FTO loss on adult NSC proliferation, they both show a resultant effect of adult neurogenesis reduction. We think that, while being mindful of the low level of FTO expression in the early developing cortex, conducting an investigation on how FTO regulates corticogenesis in the course of development may lend clarity to how it mechanistically impacts cortical neurogenesis.

Notable N-Methyladenosine Readers in Cortical Neurogenesis

Protein factors that act as readers of the mA mark have also been shown to have a profound effect on neurogenesis in the brain. For instance, the mA reader YTHDF2 has been reported to be indispensable for corticogenesis in mouse. Conditional knockout of YTHDF2 in the mouse neocortical neuroepithelium resulted in a reduction in the proliferation and differentiation of the Ythdf2/ neuroprogenitor cells (Li M. et al., 2018). This phenotype may have mechanistic underpinnings, including abnormal upregulation of genes that inhibit the JAK–STAT signaling pathway, due to increased stability of such gene transcripts in the absence of YTHDF2 (Li M. et al., 2018). Yet, it seems that the induction of neural fate in pluripotent stem cells requires downregulation of YTHDF2, leading to the stability and expression of neural gene transcripts (Heck et al., 2020). We are of the opinion that the functional consequence of the mA reading by YTHDF2 may be contextually variable along the cortical development axis such that reduced dosage may support neural cell fate specification, whereas its increased activity/expression is necessary for later cortical neurodevelopment.

Another mA reader, FMRP, was identified to be critical for neural progenitor cell proliferation. Mice lacking Fmr1 displayed prolonged cell cycle progression. As a result, proliferation of neural progenitors extended into postnatal stages of brain development (Edens et al., 2019). Of note, it was observed that nuclear export of mA-modified neurogenic mRNAs readable by FMRP is defective, leading to retention of such neurodifferentiation gene transcripts in the nucleus of the Fmr1-deficient neural progenitor cells (Edens et al., 2019). Lastly, the mA reader protein Imp (IGF2BP) was identified as a key regulator of NSC proliferation rate through the stabilization of Myc mRNA in Drosophila brain neuroblasts (Samuels et al., 2020).

Together, the mA machinery has been identified to play critical roles in brain morphogenesis by regulating the proliferation of neural progenitor cells and the production of neurons. As such, hypomethylation due to METTL3 or METTL14 deficiency and aberrant mA reading or erasure in the embryonic or adult brain can precipitate phenotypes, including defective transcriptional prepatterning, abnormal neuroprogenitor pool, impaired neurogenesis, and cortical hypoplasia (Yoon et al., 2017), which can engender deficits in brain structure and function.

Effect of N-Methyladenosine Writers on Neurogenesis in the Brain

So far, it has been shown that ablation of the mA writer METTL3 or its cofactor METTL14 in cortical neuroepithelium or isolated cortical NSC results in prolonged cell cycle dynamics of cortical neuroprogenitors and their precocious differentiation into neuronal or neurogenic cells (Batista et al., 2014; Yoon et al., 2017; Wang Y. et al., 2018). Detailed analysis through mA sequencing revealed that gene transcripts involved in the cell cycle of neural cells, production of neurons, and neuronal differentiation are enriched with mA tags. Interestingly, the decay of such mRNAs is promoted in the absence of METTL3 and METTL14 (Yoon et al., 2017), meaning that METTL3 and METTL14 are key players in driving neurogenesis via the stabilization of gene transcripts critical for neurogenesis in the brain. For example, loss of mA due to deletion of METTL3 in mouse cerebellum resulted in overt hypoplasia partly attributable to apoptosis of cerebellar granule cells (Wang C. X. et al., 2018). Key downstream effects of mA on genes important for neurogenesis include the modulation of histone modification in the promoter environment of NSC proliferation- and differentiation-related gene loci (Wang Y. et al., 2018). In the absence of METTL14, the transcription-suppressing histone mark H3K27me3 is upregulated on genes involved in cell proliferation, whereas differentiation-related genes show an increase in the transcription activation histone mark H3K27ac when METTL14 is deficient (Wang Y. et al., 2018). Lack of METTL3 in the developing brain can also cause the aforementioned histone alterations, at least in terms of H3K27me3 enhancement, which can cause transcription repression (Chen J. et al., 2019). This is possible because in the absence of METTL3, which leads to a reduction in mA levels, the polycomb repressor complex becomes hyperactive due to derepression of its core methlytransferase factor Ezh2 (Chen J. et al., 2019).

RBM15, a core component of the mA writer complex (Figure 1A), is a potential regulator of cortical neurogenesis due to its distinctive expression in the cortical germinative zone and cortical plate of the developing mouse cortex (Xie et al., 2019). Knockdown of RBM15 in neurons in vitro promoted endogenous expression of the chromatin remodeling factor BAF155 (Xie et al., 2019), which is a known key regulator of cortical development (Nguyen et al., 2016, 2018; Narayanan et al., 2018). This profound effect can be linked to a significant reduction in cellular levels of mA due to the inactivation of RBM15 (Knuckles et al., 2018). However, overexpression of RBM15 in vivo was found to promote delamination of radial glial cells in the cortical ventricular zone by suppressing the expression BAF155 and, hence, BAF155-dependent gene expression program supportive for cortical development (Xie et al., 2019). The role of RBM15 in cortical neurogenesis further highlights the contribution of mA methyltransferase in brain development.

N-Methyladenosine Erasers Regulate Neurogenesis in the Brain

Erasers of the mA mark (FTO and ALKBH5) can also exert a regulatory effect on the process of neurogenesis given their prominent expression in neurons (Li L. et al., 2017; Yoon et al., 2017; Spychala and Ruther, 2019; Du et al., 2020). Whereas FTO displays the highest expression level late in brain neurogenesis (Li L. et al., 2017; Yoon et al., 2017), ALKBH5 expression decreases in the course of brain development (Du et al., 2020). This may have implications for their roles in the spatiotemporal regulation of neurogenesis during brain development. Indeed, it was reported that FTO deficiency in the adult mouse brain induces signal transducer and activator of transcription (STAT)3 pathway activation via its modulators platelet-derived growth factor receptor (PDGFR)α and suppressor of cytokine signaling (SOCS)5 in an mA-dependent manner (Cao et al., 2020). As a result, a transient increase in the proliferation and differentiation of adult NSCs was observed in the FTO mutant brain, with implications for adult neurogenesis inhibition in the long term (Cao et al., 2020). It was also observed that FTO deletion in adult mouse brain impairs brain-derived neurotrophic factor (BDNF) and mitogen-activated protein kinase (MAPK) signaling pathways, leading to a reduction in adult NSC proliferation and neurogenesis in the hippocampal formation (Li L. et al., 2017; Spychala and Ruther, 2019). Although these studies report diverging effects of FTO loss on adult NSC proliferation, they both show a resultant effect of adult neurogenesis reduction. We think that, while being mindful of the low level of FTO expression in the early developing cortex, conducting an investigation on how FTO regulates corticogenesis in the course of development may lend clarity to how it mechanistically impacts cortical neurogenesis.

Notable N-Methyladenosine Readers in Cortical Neurogenesis

Protein factors that act as readers of the mA mark have also been shown to have a profound effect on neurogenesis in the brain. For instance, the mA reader YTHDF2 has been reported to be indispensable for corticogenesis in mouse. Conditional knockout of YTHDF2 in the mouse neocortical neuroepithelium resulted in a reduction in the proliferation and differentiation of the Ythdf2/ neuroprogenitor cells (Li M. et al., 2018). This phenotype may have mechanistic underpinnings, including abnormal upregulation of genes that inhibit the JAK–STAT signaling pathway, due to increased stability of such gene transcripts in the absence of YTHDF2 (Li M. et al., 2018). Yet, it seems that the induction of neural fate in pluripotent stem cells requires downregulation of YTHDF2, leading to the stability and expression of neural gene transcripts (Heck et al., 2020). We are of the opinion that the functional consequence of the mA reading by YTHDF2 may be contextually variable along the cortical development axis such that reduced dosage may support neural cell fate specification, whereas its increased activity/expression is necessary for later cortical neurodevelopment.

Another mA reader, FMRP, was identified to be critical for neural progenitor cell proliferation. Mice lacking Fmr1 displayed prolonged cell cycle progression. As a result, proliferation of neural progenitors extended into postnatal stages of brain development (Edens et al., 2019). Of note, it was observed that nuclear export of mA-modified neurogenic mRNAs readable by FMRP is defective, leading to retention of such neurodifferentiation gene transcripts in the nucleus of the Fmr1-deficient neural progenitor cells (Edens et al., 2019). Lastly, the mA reader protein Imp (IGF2BP) was identified as a key regulator of NSC proliferation rate through the stabilization of Myc mRNA in Drosophila brain neuroblasts (Samuels et al., 2020).

Together, the mA machinery has been identified to play critical roles in brain morphogenesis by regulating the proliferation of neural progenitor cells and the production of neurons. As such, hypomethylation due to METTL3 or METTL14 deficiency and aberrant mA reading or erasure in the embryonic or adult brain can precipitate phenotypes, including defective transcriptional prepatterning, abnormal neuroprogenitor pool, impaired neurogenesis, and cortical hypoplasia (Yoon et al., 2017), which can engender deficits in brain structure and function.

N-Methyladenosine Signaling Is Essential for Gliogenesis in the Brain

The process of generating glial cells constitutes gliogenesis. Brain neuroglia include astrocytes and oligodendrocytes, which are derived from the neuroepithelium. During cortical development, a switch from neurogenesis to gliogenesis coincides with a decrease in mA modification of proneural genes (Donega et al., 2018). Although mA enrichment in glial cells is less than that observed in neurons (Chang et al., 2017), a few studies have uncovered the importance of the mA methylome in brain gliogenesis, at least for astrocyte production (astrogenesis) and oligodendrocyte generation (oligodendrogenesis).

Regulation of Glia Production in the Brain by an N-Methyladenosine Writer-Related Factor

It was observed that loss of METTL14-mediated mA writing in the mouse cortex leads to hypomyelination that can be linked to a reduction in the number of (mature) oligodendrocytes (Xu et al., 2020). The loss of oligodendrocytes caused by the absence of METTL14 in the brain likely did not emanate from abnormal specification or proliferation of oligodendrocyte precursor cells (OPCs) (Xu et al., 2020). Notably, the transcriptome of OPCs and oligodendrocytes is altered following METTL14 deletion, with possible impact on gene expression programs critical for oligodendrocyte lineage progression (Xu et al., 2020). Lack of METTL14 has also been reported to disrupt astrogenesis. Indeed, s100β-expressing astrocytic progenitors were found to be reduced in the METTL14 knockout mouse cortex at postnatal stage 5 (Yoon et al., 2017). It would be interesting to investigate whether other mA writer-related factors, including METTL3, have roles to play in cortical gliogenesis.

The N-Methyladenosine Eraser FTO Regulates Glia Production in the Brain

mA-mediated RNA methylation dynamics under the guild of FTO is known to influence oligodendrogenesis via modulation of the half-life of Olig2 mRNA (Wu et al., 2019). Olig2 is a central factor indispensable for oligodendrocyte lineage progression (Liu et al., 2007). Specifically, FTO was reported to regulate the degradation of Olig2 transcripts via removal of mA tags installed on the Olig2 mRNA. The stability of Olig2 transcripts in OPCs deficient in FTO was thus seen to increase. In effect, the white matter in FTO mutant mouse brain was characterized by hypomyelination (Wu et al., 2019).

Involvement of N-Methyladenosine Readers in Glia Production in the Brain

The mA reader PRRC2A is known to be essential for oligodendrogenesis. It prominently regulates the specification, proliferation, and differentiation of oligodendroglia and the ability of oligodendrocytes to carry out myelination in the brain (Wu et al., 2019). More specifically, abolishing PRRC2A function in cortical NSCs or precisely in oligodendroglial lineage caused significant loss of OPCs (PDGFRα+ cells), Sox10+ cells, and mature oligodendrocytes (CC1+Oilg2+), which culminated in hypomyelination in the PRRC2A mutant brain (Wu et al., 2019). Interestingly, deletion of PRRC2A also affects astrogenesis, although slightly. Deficiency of PRRC2A in mouse brain caused a reduction in the proliferative capacity of astrocytes, leading to a reduced number of astrocytes in the mutant mouse brain (Wu et al., 2019). The additional role of PRRC2A in regulating the production of astrocytes in the brain during development may hinge on its interaction with YTHDF2, another mA-binding protein, such that lack of either mA reader augments the expression of the other to influence gliogenesis (Wu et al., 2019).

The competitive relationship between PRRC2A and YTHDF2 makes it complex to explain or reconcile the observation that glial fibrillary acidic protein (GFAP) expression, which can indicate astrocytic cells, was found to be dramatically reduced in neurospheres derived from the E14.5 Ythdf2–/– forebrain NSC. Such GFAP+ Ythdf2–/– cells also displayed abnormally branched processes (Li M. et al., 2018). Thus, further investigation is required to elucidate the role of YTHDF2 in brain gliogenesis and how the function of PRRC2A features in the regulatory pathway.

Regulation of Glia Production in the Brain by an N-Methyladenosine Writer-Related Factor

It was observed that loss of METTL14-mediated mA writing in the mouse cortex leads to hypomyelination that can be linked to a reduction in the number of (mature) oligodendrocytes (Xu et al., 2020). The loss of oligodendrocytes caused by the absence of METTL14 in the brain likely did not emanate from abnormal specification or proliferation of oligodendrocyte precursor cells (OPCs) (Xu et al., 2020). Notably, the transcriptome of OPCs and oligodendrocytes is altered following METTL14 deletion, with possible impact on gene expression programs critical for oligodendrocyte lineage progression (Xu et al., 2020). Lack of METTL14 has also been reported to disrupt astrogenesis. Indeed, s100β-expressing astrocytic progenitors were found to be reduced in the METTL14 knockout mouse cortex at postnatal stage 5 (Yoon et al., 2017). It would be interesting to investigate whether other mA writer-related factors, including METTL3, have roles to play in cortical gliogenesis.

The N-Methyladenosine Eraser FTO Regulates Glia Production in the Brain

mA-mediated RNA methylation dynamics under the guild of FTO is known to influence oligodendrogenesis via modulation of the half-life of Olig2 mRNA (Wu et al., 2019). Olig2 is a central factor indispensable for oligodendrocyte lineage progression (Liu et al., 2007). Specifically, FTO was reported to regulate the degradation of Olig2 transcripts via removal of mA tags installed on the Olig2 mRNA. The stability of Olig2 transcripts in OPCs deficient in FTO was thus seen to increase. In effect, the white matter in FTO mutant mouse brain was characterized by hypomyelination (Wu et al., 2019).

Involvement of N-Methyladenosine Readers in Glia Production in the Brain

The mA reader PRRC2A is known to be essential for oligodendrogenesis. It prominently regulates the specification, proliferation, and differentiation of oligodendroglia and the ability of oligodendrocytes to carry out myelination in the brain (Wu et al., 2019). More specifically, abolishing PRRC2A function in cortical NSCs or precisely in oligodendroglial lineage caused significant loss of OPCs (PDGFRα+ cells), Sox10+ cells, and mature oligodendrocytes (CC1+Oilg2+), which culminated in hypomyelination in the PRRC2A mutant brain (Wu et al., 2019). Interestingly, deletion of PRRC2A also affects astrogenesis, although slightly. Deficiency of PRRC2A in mouse brain caused a reduction in the proliferative capacity of astrocytes, leading to a reduced number of astrocytes in the mutant mouse brain (Wu et al., 2019). The additional role of PRRC2A in regulating the production of astrocytes in the brain during development may hinge on its interaction with YTHDF2, another mA-binding protein, such that lack of either mA reader augments the expression of the other to influence gliogenesis (Wu et al., 2019).

The competitive relationship between PRRC2A and YTHDF2 makes it complex to explain or reconcile the observation that glial fibrillary acidic protein (GFAP) expression, which can indicate astrocytic cells, was found to be dramatically reduced in neurospheres derived from the E14.5 Ythdf2–/– forebrain NSC. Such GFAP+ Ythdf2–/– cells also displayed abnormally branched processes (Li M. et al., 2018). Thus, further investigation is required to elucidate the role of YTHDF2 in brain gliogenesis and how the function of PRRC2A features in the regulatory pathway.

N-Methyladenosine Effectors Regulate the Formation of Neural Processes and Synapses

The developing and adult brain is characterized by the outgrowth of dendrites and axons of neurons known to form neural connections called synapses. Interestingly, synapses are enriched with mA, which modulates dendrite formation (dendritogenesis), axonogenesis, and synaptic growth (synaptogenesis) and activity (reviewed in Li et al., 2019; Dermentzaki and Lotti, 2020). mA-based transcriptome profiling of the mouse brain (cortex and cerebellum) showed enrichment of mA modification linked to dendrite and dendritic spine, axon and axon guidance, and synaptogenesis and synaptic transmission (Chang et al., 2017).

Distinctive localization of the YTHDFs, FTO, and METTL14 in dendrites of hippocampal neurons in culture and cortical neurons suggests the involvement of these mA-regulatory factors in the development of neural dendrites. Indeed, Ythdf1 and Ythdf3 knockdown in such cultured neurons resulted in abnormal dendritic spine (Merkurjev et al., 2018). Axons are also enriched with FTO, which can be translated locally. As such, FTO ablation in axons resulted in upregulation of mA levels, leading to a reduction in Gap-43 mRNA translation in axons of cultured dorsal root ganglion neurons (Yu et al., 2018). Yet, GAP-43 is a key factor involved in axon growth in neural tissues (Skene et al., 1986). In effect, the neurons lacking FTO displayed axon elongation repression (Yu et al., 2018). The mA reader YTHDF1 was also reported to influence axon formation by binding and promoting the translation of the axon guidance receptor Robo3.1, which directs spinal commissural axons in crossing the midline, in an mA modification-dependent manner (Zhuang et al., 2019). Together, these observations may have implications for perturbed axonogenesis in the brain lacking optimal mA modification due to ablation of FTO or YTHDF1. At least in the case of the mA-regulatory protein PRRC2A, it was found that axons that form the corpus callosum, a brain midline structure, are hypomyelinated and appeared hypoplastic in the PRRC2A-deleted mouse brain (Wu et al., 2019).

Given the enrichment of mA marks and related proteins in neural processes, it is not surprising that synapses are endowed with mA-modified mRNAs, especially postsynaptic transcripts in the mouse brain (Chang et al., 2017). The high localization of mA-modified mRNAs in synapses reflects the possible impact of the mA epitranscriptome on the structure, maturation, and function of synapses (Chang et al., 2017; Merkurjev et al., 2018; Yu et al., 2018; Zhuang et al., 2019). As a result, selective ablation of YTHDF1 and YTHDF3 in the cultured hippocampal neurons caused excitatory synaptic transmission suppression (Merkurjev et al., 2018; Shi et al., 2018). In addition, synapses formed by neurons lacking YTHDF2 appeared abnormal (Li M. et al., 2018), and synaptic transmission-related transcripts are hypermethylated in dopaminergic neurons with defective synaptic plasticity implication in the FTO-deficient mouse midbrain (Hess et al., 2013). Another indication of synapse malformation and synaptic plasticity impairment due to mA dysregulation was observed in METTL14-deleted striatal neurons, in which METTL14 abrogation resulted in aberrant neuronal excitability (Koranda et al., 2018). Given that Nito, the Drosophila version of RBM15, also regulates synaptic growth through regulation of axonogenesis (Gu et al., 2017), it would be interesting to investigate whether indeed RBM15 is involved in synaptogenesis in the mammalian brain.

Cognition and Behavior Are Modulated by N-Methyladenosine Signaling

The brain’s ability to process and store information and form or control behavior patterns has been shown to be greatly regulated by posttranscriptional modification of mRNA involved in brain development (reviewed in Jung and Goldman, 2018; Leighton et al., 2018; Noack and Calegari, 2018). Prominently emerging among these new (epitranscriptomic) levels of brain function regulation is mA modification of mRNA in the brain. Various studies in mouse models have revealed the involvement of the mA machinery-related factors in cognition and behavior (reviewed in Nainar et al., 2016; Chokkalla et al., 2020). The role of mA in the regulation of learning and behavior may be partly explained by the previously discussed role of mA in synaptogenesis and synaptic transmission (Weng et al., 2018).

N-Methyladenosine Writers Involved in Memory and Behavior

In a recent study by Zhang F. et al. (2018), it was found that the enrichment of METTL3 in the mouse hippocampus is supportive for memory consolidation via the promotion of neuronal early-response gene translation. Therefore, mice lacking METTL3 in the hippocampus displayed impaired long-term potentiation with attendant reduced ability to consolidate memory. Interestingly, long-term memory consolidation is demonstrably augmented following METTL3 overexpression in the dorsal hippocampus of the wild-type mouse brain (Zhang Z. Y. et al., 2018). The mA writer function of METTL14 is reported to be important for learning and behavior mediated by the striatum. Without affecting the number or morphology of striatal neurons, loss of METTL14 in striatopallidal and striatonigral neurons caused alterations in the transcriptome, eliciting increased neuronal excitability and spike frequency adaptation reduction, which possibly culminated in impairment of striatum-dependent behavior patterns (Koranda et al., 2018).

The N-Methyladenosine Eraser FTO Regulates Learning and Behavior

Accumulation of mA in the brain can affect its learning capacity and behavior. By regulating adult neurogenesis in the mouse hippocampus, FTO has been identified to play a pivotal role in learning (Li L. et al., 2017). Hypermethylation in the mouse brain or hippocampus caused by FTO functional loss was observed to call forth learning disabilities in mice, including increased fear memory consolidation (Widagdo et al., 2016; Walters et al., 2017). Additional evidence indicating the role of FTO in learning and behavior includes a study in which mice deficient in FTO were reported to exhibit behaviors consistent with depression and anxiety (Sun et al., 2019). Moreover, available data show that memory processing and verbal fluency may be affected in individuals with FTO ablation in the brain (Ho et al., 2010; Benedict et al., 2011).

Readers of N-Methyladenosine Modulate Learning and Memory

Cognitive deficits have been implicitly linked to lack of function of the mA reader PRRC2A, whose absence caused hypomyelination, leading to the cognitive anomalies in the mouse brain (Wu et al., 2019). Evidence indicating a more direct importance of an mA reader in learning and memory was obtained when YTHDF1 was deleted in the adult mouse brain. It was found that neuronal stimuli can evoke translation of gene transcripts readable by YTHDF1 to facilitate learning and memory (Shi et al., 2018). Hence, silencing of YTHDF1 in the mouse hippocampus resulted in defective long-term potentiation and impaired synaptic transmission in the hippocampus, which did not allow normal learning and memory processing, and the defects were rescuable by YTHDF1 re-expression in the YTHDF1 mutant brain (Shi et al., 2018).

N-Methyladenosine Writers Involved in Memory and Behavior

In a recent study by Zhang F. et al. (2018), it was found that the enrichment of METTL3 in the mouse hippocampus is supportive for memory consolidation via the promotion of neuronal early-response gene translation. Therefore, mice lacking METTL3 in the hippocampus displayed impaired long-term potentiation with attendant reduced ability to consolidate memory. Interestingly, long-term memory consolidation is demonstrably augmented following METTL3 overexpression in the dorsal hippocampus of the wild-type mouse brain (Zhang Z. Y. et al., 2018). The mA writer function of METTL14 is reported to be important for learning and behavior mediated by the striatum. Without affecting the number or morphology of striatal neurons, loss of METTL14 in striatopallidal and striatonigral neurons caused alterations in the transcriptome, eliciting increased neuronal excitability and spike frequency adaptation reduction, which possibly culminated in impairment of striatum-dependent behavior patterns (Koranda et al., 2018).

The N-Methyladenosine Eraser FTO Regulates Learning and Behavior

Accumulation of mA in the brain can affect its learning capacity and behavior. By regulating adult neurogenesis in the mouse hippocampus, FTO has been identified to play a pivotal role in learning (Li L. et al., 2017). Hypermethylation in the mouse brain or hippocampus caused by FTO functional loss was observed to call forth learning disabilities in mice, including increased fear memory consolidation (Widagdo et al., 2016; Walters et al., 2017). Additional evidence indicating the role of FTO in learning and behavior includes a study in which mice deficient in FTO were reported to exhibit behaviors consistent with depression and anxiety (Sun et al., 2019). Moreover, available data show that memory processing and verbal fluency may be affected in individuals with FTO ablation in the brain (Ho et al., 2010; Benedict et al., 2011).

Readers of N-Methyladenosine Modulate Learning and Memory

Cognitive deficits have been implicitly linked to lack of function of the mA reader PRRC2A, whose absence caused hypomyelination, leading to the cognitive anomalies in the mouse brain (Wu et al., 2019). Evidence indicating a more direct importance of an mA reader in learning and memory was obtained when YTHDF1 was deleted in the adult mouse brain. It was found that neuronal stimuli can evoke translation of gene transcripts readable by YTHDF1 to facilitate learning and memory (Shi et al., 2018). Hence, silencing of YTHDF1 in the mouse hippocampus resulted in defective long-term potentiation and impaired synaptic transmission in the hippocampus, which did not allow normal learning and memory processing, and the defects were rescuable by YTHDF1 re-expression in the YTHDF1 mutant brain (Shi et al., 2018).

Stress Response Is Regulated by Factors of the N-Methyladenosine Machinery

The brain plays a central role in stress response. In responding to stress, a host of gene expression programs is activated in the brain, leading to the secretion of several neuropeptides (de Kloet et al., 2005). Vulnerability to stressful stimuli and the response mechanism can have implications for neuropsychiatric anomalies under abnormal regulatory conditions. Thus, the transcriptomic stress response system is particularly crucial in maintaining homeostasis following exposure to stress.

Epigenetic mechanisms are known to play central roles in stress response (McEwen et al., 2015), and the epitranscriptome is an emerging gene expression regulation domain for stress modulation (Harvey et al., 2017). A putative role for mA in the regulation of stress response is evidenced by the presence of glucocorticoid response elements upstream the transcription start site of genes that encode for enzymes involved in mA modification (Engel et al., 2018). Additionally, nuclear localization of YTHDF2 precipitated by heat stress results in dynamic methylation of the 5′ UTR of newly synthesized mRNAs (Zhou et al., 2015). By limiting FTO, YTHDF2 is able to preserve methylation in the 5′ UTR of heat stress-induced mRNAs (Zhou et al., 2015).

In chick, upregulation of FTO in the brain (hypothalamus) may be a mechanism to afford thermoregulation in heat stress conditions (Kisliouk et al., 2020). However, following acute restraint stress, the mouse prefrontal cortex and amygdala displayed mA hypomethylation and hypermethylation, respectively (Engel et al., 2018). Fear-induced stress can cause downregulation of FTO, leading to elevation of mA in the prefrontal cortex and hippocampus of the mouse (Walters et al., 2017). Mice lacking METTL3 or FTO are unable to cope with stress (Engel et al., 2018). A general effect that may be caused by stress-induced alteration in mA modification is the suppression of mRNAs involved in synaptic plasticity and brain morphogenesis (Engel et al., 2018). Together, the above observations indicate a putative role for mA modulation in the human brain during stressful insults.

Neurological Disorders Attributable to Defective N-Methyladenosine Modification in the Brain

Emerging evidence shows that a number of syndromic and non-syndromic neurological disturbances can be linked to mA methylome dysregulation in the brain (Engel and Chen, 2018). This is not surprising, given the previously discussed extensive role of mA in brain neurodevelopment (Figure 3). The mA ubiquity in the brain implies that neural perturbations due to mA dysregulation are likely to be complex and multifactorial in terms of downstream causatives. Neurologic problems so far identified to be caused by genetic variants of mA modification factors can be broadly characterized as neurodevelopmental, neurodegenerative, or neuropsychiatric. Specifically, these include Parkinson’s disease (PD), Alzheimer’s disease (AD), autism, Smith–Magenis syndrome, schizophrenia, and depression (Table 2). The following subsections discuss the role of mA and associated factors in neurological disorders of the brain.

TABLE 2

Brain disorders associated with mA dysregulation.

Neurological disordersExperimental systemmA factor(s) implicatedReferences
Neurodevelopmental disorders
MicrocephalyGWAS; KO miceFto deletion; METTL5 frameshiftRichard et al. (2019)
Fragile X SyndromeGWASSNP in FMRPVerkerk et al. (1991); Dictenberg et al. (2008)
Cerebellar ataxiaKO mice; KO DrosophilaDeletion of Ythdc1, Mettl3, Alkbh5Fernandez-Funez et al. (2000); Ma et al. (2018); Wang C. X. et al. (2018)
Smith–Magenis syndromeGenetic analysis in mouseAlkbh5 deletionRicard et al. (2010)
Intellectual disabilityGWASMETTL5 frameshiftRichard et al. (2019)
Autism spectrum disorderGWASMutations in FMR1Reddy (2005); Edupuganti et al. (2017); Kaufmann et al. (2017)
Neurodegenerative disorders
Parkinson’s disease6-OHDA treatment of PC12 cells and rats; KO miceFto deletion or inhibitionHess et al. (2013); Chen X. C. et al. (2019)
Alzheimer’s diseaseGWASSNP in FTOHo et al. (2010); Keller et al. (2011); Reitz et al. (2012)
Amyotrophic lateral sclerosisGWASSNP in FTO; SNP in HNRNP (A2B1 and A1); SNP in RBM15Kim et al. (2013); Cooper-Knock et al. (2017); Mitropoulos et al. (2017)
Cerebellar ataxiaKO mice; KO DrosophilaDeletion of Ythdc1, Mettl3, Alkbh5Fernandez-Funez et al. (2000); Ma et al. (2018); Wang C. X. et al. (2018)
Multiple sclerosisGWASSNP in METTL1Mo et al. (2019)
Neuropsychiatric disorders
Major depressive disorderGWASSNP in ALKBH5; SNP in FTOSamaan et al. (2013); Milaneschi et al. (2014); Du et al. (2015)
SchizophreniaGWASSNP in ZC3H13Oldmeadow et al. (2014)
Attention-deficit/hyperactivity disorderGWASSNP in FTOVelders et al. (2012); Choudhry et al. (2013)
GWAS, genome-wide association studies; 6-OHDA, 6-hydroxydopamine; SNP, single-nucleotide polymorphism; KO, knockout; mA, N-methyladenosine.

Fragile X Syndrome

It has been identified that Fragile X syndrome (FXS) is the most common cause of inherited intellectual disorders and usually co-occurs with autism spectrum disorder (ASD). Patients present with features such as poor language development, abnormal behavior, and seizures, which are mainly clinical manifestations of neuronal excitation–inhibition imbalance (Hagerman et al., 2017; Kaufmann et al., 2017). Silencing of the FMR1 gene, which leads to lack of FMRP expression, is the cause of FXS (Brown et al., 2001). The role of FMRP in multiple gene expression programs partly accounts for the syndromic nature of FXS (Hagerman et al., 2017). Synaptic abnormalities or loss of neuroplasticity caused by FMRP loss-of-function and perhaps YTHFC2 deficiency is a critical underlying mechanism that contributes to the etiology of FXS and associated ASD (reviewed in Liu et al., 2016; Bagni and Zukin, 2019).

Parkinson’s Disease

Parkinson’s disease is a complex progressive neurodegenerative disorder mainly associated with death of dopamine-producing neurons in the midbrain (substantia nigra pars compacta) and aggregation of Lewy bodies in various brain regions. The main symptoms of PD include tremor and bradykinesia. Until now, the cause of PD is unknown, as many genetic and environmental risks are involved, making definitive diagnosis and treatment challenging (Kalia and Lang, 2015; Hayes, 2019).

Interestingly, mA methylation deregulation caused by FTO abrogation, in the midbrain or in dopaminergic neurons, has been implicated in PD pathogenesis via impairment of neuronal activity and behavior response dependent on dopamine receptor types 2 and 3 (Hess et al., 2013). mRNAs involved in dopaminergic signaling are hypermethylated in the FTO-deficient mouse midbrain and striatum, leading to their decreased translation (Hess et al., 2013). It was found in another study that mA may play a role in loss of dopaminergic neurons, which characterizes PD (Chen X. C. et al., 2019). The study reported that PC12 cells treated with 6-hydroxydopamine (6-OHDA) and the striatum of rat brain with 6-OHDA-induced PD display mA modification downregulation, which is capable of inducing N-methyl-D-aspartate (NMDA) receptor 1 expression, alongside elevated oxidative stress and influx of Ca, culminating in cell death of dopaminergic neurons. Notably, FTO inhibition, and perhaps inhibition of ALKBH5, can attenuate 6-OHDA-induced PC12 cells apoptosis (Chen X. C. et al., 2019).

Alzheimer’s Disease

The commonest cause of dementia worldwide is AD. It is mainly characterized by progressive (age-dependent) neurodegeneration in brain regions (especially in the temporal and frontal lobes), with key clinical features, including memory loss, behavior abnormalities, and eventual cognitive decline (reviewed in Weller and Budson, 2018; Soria Lopez et al., 2019). Errors in RNA metabolism can have implications for AD. As discussed further, studies in human populations and in mouse models have shown that specific dysregulations in mA mRNA methylation contribute to AD pathogenesis.

Typically, mA levels in various brain regions increase with aging, and this disposition was shown to likely have relevance for AD development (Shafik et al., 2021). Interestingly, while METTL3 is downregulated in AD brain (hippocampus), it was observed to have accumulated in the postmortem AD brain at levels comparable to the insoluble Tau protein therein (Huang H. et al., 2020). Immunohistochemistry of the entorhinal cortex of patients with AD showed selective deficiency in the expression of another mA factor hnRNP-A/B, which probably underscores the alteration in alternative splicing in the AD brain (Berson et al., 2012). Moreover, FTO mis-expression is implicated in the development of AD. Carriers of the FTO variant rs9939609 were reported to display systematic deficits in brain volume consistent with brain atrophy in the elderly (Ho et al., 2010). Indeed, a population-based study found an association between the FTO variant rs9939609 and increased risk of AD (Keller et al., 2011). Reitz et al. (2012) reported an increased risk caused by some polymorphisms (rs11075997, rs11075996, rs17219084) in the FTO gene in AD cases among some investigated Caribbean Hispanics and Caucasians (Reitz et al., 2012). Reduced verbal fluency in obese and overweight elderly men, with unaffected general cognitive function, was attributed to bearing of the FTO A allele. Thus, the (dys)functional effect of FTO A allele mainly manifests in the frontal lobe of the brain to constitute AD (Benedict et al., 2011). These observations indicate perturbation of mA signaling as a notable underlying factor in the pathophysiology of AD in humans.

In vitro and in vivo experimentations using mouse models have yielded results that further support the involvement of mA mRNA methylation in AD. In one study, it was observed that knockdown of hnRNP A/B impaired alternative splicing in cultured neurons, which resulted in loss of dendrites, and caused memory impairment in mice that can be ascribed to aberrance in the cortical connectome (Berson et al., 2012). The level of hnRNP A/B increases with cholinergic excitation, whereas loss of cholinergic signaling was found to induce AD-like reduction in hnRNP levels in the cortex (Berson et al., 2012).

The AD brain of the APP/PS1 transgenic mouse has elevated levels of mA in the hippocampus and cortex, which may be due to the increased expression of METTL3 and concurrent downregulation of FTO expression (Han et al., 2020). However, the expression of FTO was identified to be increased in the brain of the triple transgenic AD mouse (Li H. et al., 2018). This gives an impression of the complex nature of the mechanism through which FTO or other mA-associated factors may drive the development of AD. In the case of FTO, a proposed mechanism is that it may promote the phosphorylation of Tau protein by encouraging a methylation scheme leading to stabilization of tuberous sclerosis complex 1 (TSC1) mRNA, which activates the kinase activity of the mammalian target of rapamycin (mTOR) (Li H. et al., 2018).

Interestingly, cognition in a mouse model of AD was observed to improve when FTO was conditionally deleted in neurons in the mouse brain with AD (Li H. et al., 2018). This makes FTO a prospective therapeutic candidate worth further investigation for its potential in slowing down the progression of AD or in remedying related symptoms.

Amyotrophic Lateral Sclerosis

Amyotrophic lateral sclerosis (ALS) is a debilitating neurodegenerative disorder hallmarked by loss of motor neurons leading to skeletal muscle dysfunction and other clinical features, including psychological disorders and respiratory distress (Rowland and Shneider, 2001). It is believed to be idiopathic, with a greater percentage (∼90%) of cases being sporadic, while 5%−10% of cases are familial or inheritable (Kiernan et al., 2011). Studies have revealed the prominent role played by pathogenic mutation of factors associated with the RNA methylation machinery (Kim et al., 2013; Cooper-Knock et al., 2017; Mitropoulos et al., 2017).

By means of whole-genome sequencing, it became evident that mA may be involved in the pathogenesis of ALS through FTO function alteration (Mitropoulos et al., 2017). Variants of FTO gene were thus associated with sporadic cases of ALS, which appears to be a founder effect among Greeks (Mitropoulos et al., 2017). In another key study, mutation in the prion-like domain of the mA reader HNRNP (A2B1 and A1) was implicated in the pathogenesis of a familial ALS case (Kim et al., 2013). The work of Cooper-Knock et al. (2017) supports the involvement of RNA-binding protein mutations in ALS. Deleterious variants of RBM15 gene or its paralog RMB15B were found to contribute to the pathogenesis of ALS (Cooper-Knock et al., 2017).

Major Depressive Disorder

Major depressive disorder (MDD) is a common neuropsychiatric condition that is considered a biobehavioral syndrome with clinical characteristics including depressed mood, cognitive dysfunction, neurovegetative disturbance, and diminished interests. Females are known to be more affected by MDD than males. Multiple factors are known to cause MDD. Notable underlying causatives include genetic and environmental factors leading to alteration in the volume of the hippocampus and aberrant brain circuitry (Fava and Kendler, 2000; Flint and Kendler, 2014; Otte et al., 2016).

The mA RNA methylome plays a role in the development of MDD (Engel et al., 2018). Genetic variants of FTO have been implicated in MDD, although heterogeneity in the associated phenotype is noteworthy (Milaneschi et al., 2014). In particular, it was found in a genome-wide association study that the FTO rs9939609 A variant is associated with a reduced risk of MDD (Samaan et al., 2013). A single-nucleotide polymorphism (rs12936694) in ALKBH5 was also found to likely be the cause of MDD among the Chinese Han population in an association study (Du et al., 2015). Interestingly, by blocking the translocation of ALKBH5 into the nucleus, it was possible to attenuate depression-like behavior in the mouse due to attendant hypermethylation and subsequent degradation of fatty acid amide hydrolase mRNA in astrocytes (Huang R. R. et al., 2020).

Fragile X Syndrome

It has been identified that Fragile X syndrome (FXS) is the most common cause of inherited intellectual disorders and usually co-occurs with autism spectrum disorder (ASD). Patients present with features such as poor language development, abnormal behavior, and seizures, which are mainly clinical manifestations of neuronal excitation–inhibition imbalance (Hagerman et al., 2017; Kaufmann et al., 2017). Silencing of the FMR1 gene, which leads to lack of FMRP expression, is the cause of FXS (Brown et al., 2001). The role of FMRP in multiple gene expression programs partly accounts for the syndromic nature of FXS (Hagerman et al., 2017). Synaptic abnormalities or loss of neuroplasticity caused by FMRP loss-of-function and perhaps YTHFC2 deficiency is a critical underlying mechanism that contributes to the etiology of FXS and associated ASD (reviewed in Liu et al., 2016; Bagni and Zukin, 2019).

Parkinson’s Disease

Parkinson’s disease is a complex progressive neurodegenerative disorder mainly associated with death of dopamine-producing neurons in the midbrain (substantia nigra pars compacta) and aggregation of Lewy bodies in various brain regions. The main symptoms of PD include tremor and bradykinesia. Until now, the cause of PD is unknown, as many genetic and environmental risks are involved, making definitive diagnosis and treatment challenging (Kalia and Lang, 2015; Hayes, 2019).

Interestingly, mA methylation deregulation caused by FTO abrogation, in the midbrain or in dopaminergic neurons, has been implicated in PD pathogenesis via impairment of neuronal activity and behavior response dependent on dopamine receptor types 2 and 3 (Hess et al., 2013). mRNAs involved in dopaminergic signaling are hypermethylated in the FTO-deficient mouse midbrain and striatum, leading to their decreased translation (Hess et al., 2013). It was found in another study that mA may play a role in loss of dopaminergic neurons, which characterizes PD (Chen X. C. et al., 2019). The study reported that PC12 cells treated with 6-hydroxydopamine (6-OHDA) and the striatum of rat brain with 6-OHDA-induced PD display mA modification downregulation, which is capable of inducing N-methyl-D-aspartate (NMDA) receptor 1 expression, alongside elevated oxidative stress and influx of Ca, culminating in cell death of dopaminergic neurons. Notably, FTO inhibition, and perhaps inhibition of ALKBH5, can attenuate 6-OHDA-induced PC12 cells apoptosis (Chen X. C. et al., 2019).

Alzheimer’s Disease

The commonest cause of dementia worldwide is AD. It is mainly characterized by progressive (age-dependent) neurodegeneration in brain regions (especially in the temporal and frontal lobes), with key clinical features, including memory loss, behavior abnormalities, and eventual cognitive decline (reviewed in Weller and Budson, 2018; Soria Lopez et al., 2019). Errors in RNA metabolism can have implications for AD. As discussed further, studies in human populations and in mouse models have shown that specific dysregulations in mA mRNA methylation contribute to AD pathogenesis.

Typically, mA levels in various brain regions increase with aging, and this disposition was shown to likely have relevance for AD development (Shafik et al., 2021). Interestingly, while METTL3 is downregulated in AD brain (hippocampus), it was observed to have accumulated in the postmortem AD brain at levels comparable to the insoluble Tau protein therein (Huang H. et al., 2020). Immunohistochemistry of the entorhinal cortex of patients with AD showed selective deficiency in the expression of another mA factor hnRNP-A/B, which probably underscores the alteration in alternative splicing in the AD brain (Berson et al., 2012). Moreover, FTO mis-expression is implicated in the development of AD. Carriers of the FTO variant rs9939609 were reported to display systematic deficits in brain volume consistent with brain atrophy in the elderly (Ho et al., 2010). Indeed, a population-based study found an association between the FTO variant rs9939609 and increased risk of AD (Keller et al., 2011). Reitz et al. (2012) reported an increased risk caused by some polymorphisms (rs11075997, rs11075996, rs17219084) in the FTO gene in AD cases among some investigated Caribbean Hispanics and Caucasians (Reitz et al., 2012). Reduced verbal fluency in obese and overweight elderly men, with unaffected general cognitive function, was attributed to bearing of the FTO A allele. Thus, the (dys)functional effect of FTO A allele mainly manifests in the frontal lobe of the brain to constitute AD (Benedict et al., 2011). These observations indicate perturbation of mA signaling as a notable underlying factor in the pathophysiology of AD in humans.

In vitro and in vivo experimentations using mouse models have yielded results that further support the involvement of mA mRNA methylation in AD. In one study, it was observed that knockdown of hnRNP A/B impaired alternative splicing in cultured neurons, which resulted in loss of dendrites, and caused memory impairment in mice that can be ascribed to aberrance in the cortical connectome (Berson et al., 2012). The level of hnRNP A/B increases with cholinergic excitation, whereas loss of cholinergic signaling was found to induce AD-like reduction in hnRNP levels in the cortex (Berson et al., 2012).

The AD brain of the APP/PS1 transgenic mouse has elevated levels of mA in the hippocampus and cortex, which may be due to the increased expression of METTL3 and concurrent downregulation of FTO expression (Han et al., 2020). However, the expression of FTO was identified to be increased in the brain of the triple transgenic AD mouse (Li H. et al., 2018). This gives an impression of the complex nature of the mechanism through which FTO or other mA-associated factors may drive the development of AD. In the case of FTO, a proposed mechanism is that it may promote the phosphorylation of Tau protein by encouraging a methylation scheme leading to stabilization of tuberous sclerosis complex 1 (TSC1) mRNA, which activates the kinase activity of the mammalian target of rapamycin (mTOR) (Li H. et al., 2018).

Interestingly, cognition in a mouse model of AD was observed to improve when FTO was conditionally deleted in neurons in the mouse brain with AD (Li H. et al., 2018). This make