The brain is the most complex and cellularly diverse organ in the body. Higher order brain functions, and those functions critical to sustain life are facilitated by the concerted activity of different cell types, each functionally driven by distinct, context dependent gene expression and gene expression regulation patterns (Hawrylycz et al., 2012; Mitra et al., 2021). Recently, the application of single cell RNA-sequencing technologies has begun to reveal at single cell resolution this dynamic and distinct gene expression patterns and how they respond to stimulus or activity (Agarwal et al., 2020; Pfisterer et al., 2020; Joglekar et al., 2021; Song et al., 2021).

Appropriate spatiotemporal gene expression in the brain involves several tightly monitored layers of regulation. Broad effector transcription factors direct transcription of genes via canonical binding sites throughout the genome, a process known to be critical for learning and memory and memory consolidation (Kaldun and Sprecher, 2019). Epigenetic mechanisms (DNA methylation, histone modifications etc.), often in response to transcription factor signals, modify DNA and associated histone structure to enhance or repress transcription (Kim et al., 2009; Conboy et al., 2021). Post-transcriptional mechanisms, including microRNAs, are also known to profoundly influence gene expression (or rather mRNA translation) in the brain adding an additional layer of regulatory complexity and fine-tuning of gene output (Bartel, 2018; Brennan and Henshall, 2020). The proteome is also tightly regulated and post-translational modifications such as phosphorylation, ubiquitylation and sumoylation dictate the efficacy or function of the resulting protein according to the needs of a given cell (Zhang et al., 2016; Czuba et al., 2018). These processes are not independent of each other and function cohesively, cumulatively forming a complex control over gene output.

RNA is now widely recognized to undergo complex post-transcriptional editing and modification (in addition to alternative splicing and polyadenylation) which confer additional information carrying capacity and profoundly influence its fate (Jain et al., 2019; Zaccara et al., 2019; He and He, 2021). Among the most well studied involves adenosine to inosine (A-to-I) conversion which is mediated by both ADAR I and II enzymes (Eisenberg and Levanon, 2018; Christofi and Zaravinos, 2019). A-I editing involves the hydrolytic deamination of adenosine to inosine. This recoding can result in non-synonymous amino acid substitutions, resulting in altered protein-coding sequences and potentially altered protein structure (Eisenberg and Levanon, 2018). Additionally, recoding of the RNA sequence within the 3′-UTR may alter mRNA-translational efficiency as it may affect microRNA-mediated targeting (Brummer et al., 2017; van der Kwast et al., 2020). Other modifications are more subtle than overt structural modification of bases, involving covalent modification of RNA such as the addition of methyl groups to specific nucleotides. Various forms of RNA were known to undergo extensive methylation as far back as the 1970s (Munns et al., 1977; Wei and Moss, 1977; Burke and Joh, 1982). However, recent advances in transcriptomic technologies and the development of antibodies, which specifically target these modifications have made it possible to reveal the extent and identity of these novel RNA modifications (Dominissini et al., 2012; Meyer et al., 2012; Dominissini et al., 2013). Critically, many of these modifications may be dynamic in nature undergoing context-dependent addition and removal.

The addition of methyl groups to the N6 position of adenosine [N6-methyladenosine (m⁶A)] is the most abundant internal modification in RNA and is prevalent in brain tissue. However, the function of m⁶A in brain is only beginning to emerge. In this review, we provide an overview of the mechanisms by which m⁶A is regulated and try to define its overall contribution to the gene expression landscape in brain cells. We also discuss recent reports, which detail the involvement of m⁶A in neurological diseases.

It was identified in the 1970s that certain forms of RNA such as rRNA undergo extensive methylation (Munns et al., 1977; Wei and Moss, 1977). However, it was thought to be mainly structural and further interrogation was limited by the technology at this time (Furuichi et al., 1975). The development of an antibody which recognizes the m⁶A mark prompted transcriptome-wide investigations of N6-methyladenosine. This modification was found to be widespread and enriched near stop codons, the 5′ and 3′ UTR and internal long exons of mRNA, as well as rRNA, tRNA, snoRNA and lncRNA (Meyer et al., 2012; Berulava et al., 2015). A canonical m⁶A motif was identified and consists of RRACH with R =G/A and H =A/C/U (Balacco and Soller, 2019; Dominissini et al., 2012). The methyl group is catalyzed from a donor substrate S-adenosylmethionine (SAM) to an adenosine residue of an RNA moiety along a specific sequence as stated above (Bokar et al., 1997). The m6A modification plays a role in several diverse RNA mechanisms, most notably RNA stability and translational efficiency (Meyer et al., 2015; Chen X.Y. et al., 2019). Other studies have implicated m6A in the control of mRNA dynamics including alternative splicing (although there is considerable debate regarding this Ke et al., 2017) and subcellular localization. Moreover, the role of m⁶A may be dictated by the subcellular localization of the m⁶A-tagged RNA. In the nucleus, m⁶A deposited on nascent pre-mRNA may influence alternative splicing (Dominissini et al., 2012), and microRNA biogenesis (Alarcon et al., 2015), while in the cytoplasm, it is thought to regulate RNA stability (Batista et al., 2014; Wang X. et al., 2014), translational efficiency and RNA decay (Wang et al., 2015).

M⁶A is a dynamic modification, catalyzed by a distinct enzymatic complex (writers), identified and processed by several “reader” proteins and potentially removed by “eraser” proteins. In the following sections, we summarize what is known about the proteins associated with the deposition, identification, and removal of m⁶A (Figure 1).

The methyltransferase complex which catalyzes m⁶A addition is composed of two distinct sub-complexes (Knuckles et al., 2018): the m⁶A-methyltransferase-like (METTL) complex (MAC) which is composed of METTL3 and METTL14 and the m⁶A-METTL associated complex (MACOM) which consists of RBM15, ZC3H13, WTAP, and VIRMA (Meyer and Jaffrey, 2014; Cao et al., 2016; Lence et al., 2019). Together MAC and MACOM function to catalyze the addition of methyl groups to adenosine.

Two proteins have been shown to possess m⁶A-demethylase activity. FTO (Fat mass and obesity-associated protein) and ALKBH5 (AlkB homologue 5 protein). These proteins are [(α-ketoglutarate (α-KG)-dependent dioxygenases which are inhibited by D2-hydroxyglutarate (D2-HG)] (Fedeles et al., 2015; Chen X.Y. et al., 2019).

FTO is an AlkB-like 2-oxoglutarate-dependent nucleic acid demethylase (Gerken et al., 2007), initially identified for its role in diabetes and obesity (Lindgren and McCarthy, 2008; Loos and Bouchard, 2008). Since then it has been shown that, in the nucleus, FTO catalyzes the removal of m⁶A and m⁶Am (N⁶, 2’-O-dimethyladenosine) marks from mRNA although importantly shows a 100-fold greater affinity for m⁶Am than m⁶A (Mauer et al., 2017; Mauer and Jaffrey, 2018; Balacco and Soller, 2019). The m⁶Am mark is a terminal modification distinct from the internal m⁶A mark and follows the m⁷G (N⁷-methylguanosine) cap at the N⁶ position (Wei et al., 1975), where A or m⁶A are not found (Wei et al., 1976). Some studies have identified FTO as an important regulator of m⁶A-tagged transcripts, important for mRNA alternative splicing and gene expression, contributing to the regulation of adipogenesis (Zhao et al., 2014; Bartosovic et al., 2017; Chen X.Y. et al., 2019). However, some recent emerging data does not support the role of FTO as an m⁶A eraser and suggests that FTO under normal physiological conditions does not remove m⁶A and solely functions on m⁶Am (Mauer et al., 2017). Indeed, FTO is present almost exclusively in the nucleus, which suggests it is unlikely to dynamically regulate cytosolic mRNAs. Furthermore, studies by the Darnell group have shown that the levels of m⁶A are perhaps stable once deposited. Together, this calls into question the extent of m⁶A demethylation in vivo, at least under normal physiological conditions (Mauer et al., 2017; Mauer and Jaffrey, 2018). M⁶A removal by FTO or others may therefore be context and cell-type specific and/or developmentally regulated and this warrants further investigation.

ALKBH5 is an α-ketoglutarate dependent oxidase and nuclear demethylase only found in vertebrates which can bind single stranded nucleic acids (Zheng et al., 2013; Balacco and Soller, 2019). In Ensfelder et al. (2018) identified ALKBH5 as a ribosomal RNA m⁶A eraser. It has since been shown that ALKBH5 is also required for correct splicing and production of transcripts in spermatocytes via selective removal of m⁶A (Tang et al., 2018). Moreover, ALKBH5 is involved in glioblastoma to repress tumorigenesis (Zhang et al., 2017), in spermatogenesis and in male fertility (Zheng et al., 2013). Again, the relevance of ALKBH5 as an m⁶A eraser in somatic cells remains unclear. The study by Darnell did note that there was loss of m⁶A on a minority of transcripts suggesting a small amount of selective demethylation may occur under normal physiological conditions (Ke et al., 2017).

M⁶A readers elicit different actions upon recognition and binding of the methylated RNA transcripts. Several reader proteins have now been identified.

HNRNPC & HNRNPA2B1 (Heterogeneous Nuclear Ribonucleoprotein C & A2B1) are both nuclear-confined m⁶A readers and are particularly involved in pre-mRNA processing and in maturation of pri-miRNA to pre-miRNA (Cao et al., 2016). HNRNPA2B1 promotes processing of METTL3-dependent microRNAs and plays a role in the regulation of the alternative splicing of exons (Alarcon et al., 2015).

YTH domain containing proteins (YTHDC1 and YTHDC2) and YTH domain-containing family proteins (YTHDF1, YTHDF2 and YTHDF3), a highly conserved family of proteins, have been shown to detect and bind m⁶A (Berlivet et al., 2019). The DC proteins are usually confined to the nucleus (Balacco and Soller, 2019) while the DF proteins are found primarily in the cytoplasm. The structure of m⁶A-bound by YTH-domain containing proteins was elucidated several years ago. The YTH domain is critical for this binding and is involved in the development of a tryptophan cage around the m⁶A mark (Theler et al., 2014; Xu et al., 2014). Recent iCLIP data from the Jaffrey lab, identified the binding sites for all YTH proteins. They found that DC1 binds m⁶A sites in mRNA and nuclear non-coding RNAs, while all three DF proteins bind most m⁶A sites (Patil et al., 2018). It has been found that DC1 contributes to extensive alternative polyadenylation and 3′UTR length alteration (Chen X.Y. et al., 2019) while DC2 is primarily localized in the testes where it has been found to have weak m⁶A affinity (Wojtas et al., 2017). In the mouse hippocampus, upon binding, YTHDF1 promotes the translation of m⁶A-tagged transcripts via interaction with a translation initiation factor (Wang et al., 2015) and the help of YTHDF3 (Shi et al., 2017) and of eIF1 (Chen X.Y. et al., 2019) to facilitate learning and memory (Shi et al., 2018). YTHDF2 recognition of m⁶A sites, contrarily, was reported to regulate the degradation of specific mRNAs via the recruitment of CCR4-NOT (Alarcon et al., 2015; Cao et al., 2016; Du et al., 2016; Chen X.Y. et al., 2019). This protein may also play a role in mRNA stability and translation (Wang X. et al., 2014; Chen X.Y. et al., 2019), targeting stop codon regions, 3′UTR and coding regions (Wang X. et al., 2014). Recent studies however, including iCLIP-sequencing datasets suggest that all three DF proteins do not selectively bind to m⁶A-tagged transcripts and all likely lead to degradation of bound transcripts. It is possible that context dependence is critical toward understanding the function of m⁶A-recognition by YTH-domain containing proteins.

METTL16, usually a writer enzyme, may also function as a reader when SAM levels are low, in order to stimulate the SAM synthetase MAT2A mRNA translation to increase the level of SAM. In contrast, when SAM levels are high, METTL16 methylates MAT2A mRNA, which promotes nuclear mRNA decay (Pendleton et al., 2017; Balacco and Soller, 2019).

elF3 (eukaryotic initiation Factor 3) is a component of the 43s translation pre-initiation complex. It has been found that this protein is an m⁶A reader, which promotes translation via two mechanisms. First, eIF3 can directly bind the 5′UTR-localised m⁶A to allow the recruitment of the 43s complex (Meyer et al., 2015; Cao et al., 2016). The second method is via its interaction with YTHDF1 which facilitates cap-dependent translation (Wang et al., 2015).

FMRP (Fragile X mental retardation protein) is an RNA-binding protein, loss of which is implicated in the neurodevelopmental condition Fragile X syndrome (FXS). It has been shown to be important for the stability of m⁶A-tagged mRNA (Zhang et al., 2018) and for their nuclear export (Hsu et al., 2019). FMRP has been found to interact with WTAP and recently suggested to bind m⁶A sites by competing with YTHDF2 to regulate specific mRNA turnover (Horiuchi et al., 2013; Balacco and Soller, 2019). Indeed, another study found that FMRP modulates the stability of its target through YTHDF2 (Zhang et al., 2018). FMRP was recently shown to directly bind to m⁶A-marked mRNA (Westmark et al., 2020) and negatively regulate their translation (Edupuganti et al., 2017). How m⁶A processing in FXS is affected has yet to be analyzed.

ELAVL1 (Drosophila homologue-like 1), also known as human antigen R (HuR), and IGF2BPs (Insulin-like growth factor 2 mRNA-binding proteins) has been identified as an m⁶A reader (Balacco and Soller, 2019). Several studies show that ELAVL1 increases the stability of transcripts via m⁶A-mRNA-binding. Indeed, in prostatic carcinoma cells, ELAVL1 plays a role in stabilizing integrin β1 mRNA (Li et al., 2020). It may also promote SOX2 mRNA stabilization to stimulate the maintenance of glioma stem-like cells (Visvanathan et al., 2018) as well as DRG1 in osteosarcoma human tissue (Ling et al., 2020).

The m⁶A mark influences the behavior of RNA in several biological pathways and subsequently biological processes like cell differentiation and proliferation, development, sex determination and circadian rhythms (Chen et al., 2017). Here, we review the most important and studied cerebral processes impacted by m⁶A methylation such as learning and memory as well as neurogenesis, neurodevelopment, the stress response, myelination, and axon plasticity (Table 1).

As described above, m⁶A RNA methylation is involved in many essential cerebral processes, so, unsurprisingly, this process is found to be altered in many brain disorders. Here we describe neurological disorders in which m⁶A, or m⁶A-associated proteins are affected such as Alzheimer’s disease, Parkinson’s disease, glioblastoma, schizophrenia, depression, transient focal ischemia, and traumatic brain injury (Table 2).

As discussed, m⁶A was identified as a prevalent RNA modification as far back as the 1970s yet was largely ignored until recently due to a lack of resources available to profile and probe its function (Munns et al., 1977; Wei and Moss, 1977; Burke and Joh, 1982). The development of an m⁶A specific antibody coupled with the growing use of next generation sequencing in the last decade led to the development of methylated RNA immunoprecipitation sequencing (meRIP-Seq/m⁶A-Seq) approaches (Dominissini et al., 2013). This technology is an adaptation of a standard RIP-Seq approach whereby RNA is isolated from a sample and sheared to a specific length. RNA is ribosome depleted and/or polyA-selected and then subjected to immunoprecipitation to isolate m⁶A-tagged mRNAs. The resulting isolated m⁶A-tagged RNA is then used to prepare RNA-Seq libraries and subjected to next generation sequencing and whole transcriptome analysis. The development of this technique enabled widespread analysis of this modification in various tissues, organisms and disease settings and highlighted the extent of the epitranscriptomic modification as well as identify how it may contribute to disease. Certain limitations existed however including high RNA input requirements (300 (μg total RNA) and it also suffers from limited resolution although bioinformatic predictions do help to identify exact sites of m6A. These limitations have been overcome to some degree and meRIP-seq can now be performed with as little as 3 (μg starting material (Zeng et al., 2018). Furthermore, single base resolution of m6A can now be mapped using crosslinking and immunoprecipitation (miCLIP) approaches (Linder et al., 2015).

M⁶A can now be mapped in native full length RNA moieties without the need for reverse transcription, shearing or immunoprecipitation steps using nanopore long read sequencing technologies (Liu et al., 2019; McIntyre et al., 2019; Jenjaroenpun et al., 2021). Several robust bioinformatic pipelines to identify methylated bases from long sequencing reads have now also been established.

The complex functions of the mammalian brain are dependent upon precise control of gene expression and regulation in a temporospatial fashion. M⁶A represents an additional layer of gene regulation, which has now been shown to critically regulate neuronal development and function. Technically difficult to study, tools are now available to analyze m⁶A RNA methylation at a global level and at single base resolution and there is huge scientific interest in this modification at present (Dominissini et al., 2012; Linder and Jaffrey, 2019). The next hurdle lies in developing and adapting current applications to profile m⁶A profiles in discrete cell types and populations/brain regions to understand how it contributes to neurodevelopment at the single cell level as well as specific brain functions such as learning and memory. Further challenges lie in understanding the precise function of m⁶A-eraser proteins in a context-dependent manner as well as probing the regulatory mechanisms governing m⁶A-associated protein expression and function.

Understanding the role of m⁶A in physiological homeostasis and disease is critical as components of these pathways may represent therapeutic targets for the treatment of neurological disease including many underserved conditions like TBI and stroke. The development of specific pharmacological inhibitors and antisense oligonucleotide approaches will also enable more precise interrogation of transient manipulations of m⁶A-associated proteins and therapeutic viability.

JM, DH, and GB researched and wrote the article. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.