Advances in brain epitranscriptomics research and translational opportunities

Abstract

Various chemical modifications of all RNA transcripts, or epitranscriptomics, have emerged as crucial regulators of RNA metabolism, attracting significant interest from both basic and clinical researchers due to their diverse functions in biological processes and immense clinical potential as highlighted by the recent profound success of RNA modifications in improving COVID-19 mRNA vaccines. Rapid accumulation of evidence underscores the critical involvement of various RNA modifications in governing normal neural development and brain functions as well as pathogenesis of brain disorders. Here we provide an overview of RNA modifications and recent advancements in epitranscriptomic studies utilizing animal models to elucidate important roles of RNA modifications in regulating mammalian neurogenesis, gliogenesis, synaptic formation, and brain function. Moreover, we emphasize the pivotal involvement of RNA modifications and their regulators in the pathogenesis of various human brain disorders, encompassing neurodevelopmental disorders, brain tumors, psychiatric and neurodegenerative disorders. Furthermore, we discuss potential translational opportunities afforded by RNA modifications in combatting brain disorders, including their use as biomarkers, in the development of drugs or gene therapies targeting epitranscriptomic pathways, and in applications for mRNA-based vaccines and therapies. We also address current limitations and challenges hindering the widespread clinical application of epitranscriptomic research, along with the improvements necessary for future progress.

INTRODUCTION

In a classic view of molecular biology, coded genetic information flows from double-stranded DNA to single-stranded RNA during transcription, and then to protein during translation [1]. More than 170 chemically distinct modifications of all kinds of RNA transcripts, named epitranscriptomics, orchestrate the transcription and translation of genetic information and the dynamic and reversible RNA modifications are analogous to the epigenetic code formed by DNA and histone modifications [2–7]. Accumulative evidence has demonstrated diverse and important roles of epitranscriptomics in modulating nearly all aspects of RNA metabolism to regulate various biological processes, including development, aging, and diseases [8–11]. The translational potential of epitranscriptomics has been highlighted by the recent success of COVID-19 mRNA vaccines utilizing a N1-methylpseudouridine modification to reduce immunogenicity and enhance efficacy [12]. Notably, epitranscriptomics also play important roles in the nervous system, ranging from neural development, circuit activities, to pathology, including psychiatric disorders [13–18]. Here we review recent advances in our understanding of epitranscriptomic mechanisms in healthy and diseased mammalian brains. We also discuss translational opportunities and new directions of epitranscriptomic research of the nervous system. Interested readers can consult general reviews regarding epitranscriptomics [7, 8, 10, 19–22].

GENERAL INTRODUCTION OF EPITRANSCRIPTOMICS

Although RNA modifications were discovered about a century ago, only recently has our knowledge of modification sites and functions expanded significantly, mainly due to the development of mass-spectrometry and high-throughput sequencing methods for mapping modification sites, some at single-nucleotide resolution, and identification of writers, readers, and erasers for different modifications [2, 4, 6, 7, 23–33] (Table S1). RNA modifications have been found in different types of RNA, including messenger RNA (mRNA), transfer RNA (tRNA), ribosomal RNA (rRNA), microRNA (miRNA), long non-coding RNA, small nuclear RNA, small nucleolar RNA (snoRNA), Piwi-interacting RNA (piRNA), chromosome-related regulatory RNAs (CarRNAs), enhancer RNA (eRNA), circular RNA, and mitochondrial RNA [5, 7, 8, 19, 20, 22, 34–74] (Fig. 1 and Table S2). The types of RNA modifications are also very diverse, including N⁶-methyladenosine (m⁶A), pseudo-uridine (Ψ), C⁵- methylcytosine (m⁵C), N³-methylcytosine (m³C), N¹-methyladenosine (m¹A), 2′-O-methylation, 2’-O-methyladenosine (m⁶Am), N⁷-methylguanosine (m⁷G), A-to-I editing, glycosylated RNA and various tRNA modifications (Fig. 1 and Table S2). Beyond being the byproduct of passive incorporation of modified ribonucleotides, RNA modifications are often actively deposited on specific sites by “writer” proteins, potentially suppressed by “suppressor” factors, recognized and decoded by “reader” proteins, and eliminated by “eraser” proteins, rendering the epitranscriptome reversible and dynamic. RNA modifications are critically involved in regulating almost every aspect of RNA metabolism, including transcription, splicing, decay, stability, translation, translocation, catalytic function, interaction with proteins, charging, and RNA structure (Fig. 1 and Table S2). Thus, it is not surprising that RNA modifications have been shown to play vital roles in regulating the nervous system, ranging from neurodevelopment and physiological brain functions to mental disorders.

Fig. 1. Epitranscriptome landscape.

Schematic illustration of several important RNA modifications focusing on mRNA, tRNA and rRNAs, as well as their writers, erasers, readers, and general molecular functions. The nucleus or cytoplasm localization of RNA modifications and their regulators are not specified. m⁶A: N⁶-methyladenosine; Ψ: Pseudo-uridine; m⁵C: C⁵-methylcytosine; m³C: N³-methylcytosine; m¹A: N¹-methyladenosine; m⁶Am: 2’-O-methyladenosine; m⁷G: N⁷-methylguanosine; m¹G: N1-methylguanosine; m²₂G: N2, N2 dimethylguanosine; τm⁵U: 5-taurinomethyluridine; τm⁵s²U: 5-taurinomethyl-2-thiouridine; f⁵C: 5-formylcytidine; i⁶A: N⁶-isopentenyladenosine; t⁶A: N⁶-threonylcarbamoyladenosine; Q: queuosinylation.

ROLES AND MECHANISMS OF RNA MODIFICATIONS REVEALED IN ANIMAL STUDIES

Neurogenesis and neuronal development

Neurogenesis involves proliferation and differentiation of neural progenitor cells (NPCs) and migration and maturation of immature neurons, and these processes are precisely coordinated to generate layered or nuclear structures and functional neuronal connectivity of the brain [75, 76]. An increasing number of studies via genetic manipulation of writers, readers, and erasers of various RNA modifications using animal and human cellular models has demonstrated extensive involvement of epitranscriptomics in regulating diverse aspects of neurogenesis (Fig. 2).

Fig. 2. Epitranscriptomics in the nervous system development.

Involvement of epitranscriptomics in regulating brain development, including embryonic neurogenesis (a), adult neurogenesis (b), gliogenesis (c), and synaptic transmission (d). Examples of brain development-related phenotypes in the mouse models with loss of function of several RNA modification regulators, like m⁶A, m⁵C, m³C, 2′-O-methylation, and A-to-I editing, are illustrated. A-to-I: A-to-I editing. LTP: Long-term potentiation.

The most well-studied example is m⁶A mRNA modification, the most abundant internal modification of mRNA. The level of m⁶A RNA modification in the brain is dynamic, which is low at embryonic stages and elevated in adulthood [23]. Deletion of m⁶A methyltransferases Mettl14 and Mettl3, or m⁶A reader Fmr1, or double knockout of m⁶A readers Ythdf1/2, lead to similar defects in mouse cortical development, with some exceptions, that are generally characterized by prolonged cell cycle of NPCs, decreased generation of neurons in the cortex, and a smaller brain size [18, 37, 77–80] (Fig. 2a). In contrast, depletion of m⁶A erasers, Fto and Alkbh5, does not appear to affect embryonic brain development [18, 77, 78] except under exposure to hypobaric hypoxia, which leads to aberrant NPC proliferation and differentiation in the postnatal cerebellum of mice with Alkbh5 deletion [81]. It has been proposed that mRNA m⁶A modification ensures the degradation of the old and pre-patterned mRNA and the translation of new mRNA in a fast and precise fashion to enhance the temporal resolution and accuracy of transcriptome/proteome changes during NPC differentiation [18]. With more m⁶A modification regulators being identified, such as methyltransferase Mettl16 and m⁶A readers Ythdc1/2, Prrc2a, and Igf2bp1/2/3, characterization of neurogenesis phenotypes caused by genetic manipulation of these genes in a spatiotemporal conditional fashion will greatly expand our knowledge of functions and mechanisms of m⁶A mRNA modification in regulating development of the nervous system.

In addition to m⁶A modifications on mRNA, modifications of other types of RNA also regulate neurogenesis. For example, knockdown of Mettl5, which encodes an 18S rRNA m⁶A methyltransferase, causes a microcephalic phenotype in Zebrafish [34]. Knockout of Mettl8, which encodes a m³C methyltransferase only for mitochondria tRNA^Thr/Ser(UCN), both in mice and in human induced pluripotent stem cell (iPSC)-derived forebrain organoids leads to impaired NPC maintenance and increased neuronal differentiation by attenuating mitochondrial protein translation and function [46]. Knockout of Nsun2, which encodes an m⁵C methyltransferase, leads to deficits in neural differentiation and a reduced number of upper-layer neurons in the mouse cerebral cortex [82]. Knockout of Nsun5, which encodes a 28S rRNA m⁵C methyltransferase, impairs radial neuronal migration and disrupts the lamination of deep/superficial layer neurons in the mouse cortex [83]. Knockout of Wdr4, which encodes a member of the METTL1/WDR4 complex responsible for RNA m⁷G modification, leads to embryonic lethality of mice at E9.5–10.5 with severe abnormality of brain development [84].

Beyond embryonic development, neurogenesis occurs in the dentate gyrus of adult mammalian brains, which is important for learning, memory, and mood regulation, and its dysregulation has been associated with psychiatric disorders [85, 86]. Mettl3 knockdown and Fto knockout in adult NPCs lead to similar phenotypes with decreased NPC proliferation and neuronal differentiation [87–89] (Fig. 2b). However, another study showed that conditional Fto knockout in adult NPCs in vivo transiently increases their proliferation and neuronal differentiation, leading to decreased adult neurogenesis over the long-term [90]. The disparities in the roles of FTO in regulating adult neurogenesis may be attributed to the use of conventional or conditional knockout mice in different studies. Besides, FTO appears to affect adult neurogenesis but largely not embryonic brain development, which may be due to the cell-type-specific effects and distinct m⁶A-modified mRNA targets of FTO in these two quite different systems in which embryonic brain development involves the dividing stem cells undergoing massive proliferation and differentiation, while adult neurogenesis involves the activation of long-term quiescent neural stem cells, warranting future clarification. Despite these advances, future functional studies in animal and human brain organoid models are needed to provide a comprehensive understanding of the roles of diverse RNA modifications beyond m⁶A in regulating embryonic and adult neurogenesis.

Gliogenesis

Glial cells are as numerous as neurons in the human brain and are important for providing metabolic support to neurons and regulating neural specification, synaptic plasticity, myelination, and cellular debris clearance [91, 92]. Compared to neurogenesis, the role of epitranscriptomics in regulating gliogenesis is much less known. Knockout of Mettl14 in oligodendrocyte progenitor cells (OPCs) and post-mitotic oligodendrocytes leads to a reduced number of mature oligodendrocytes in the mouse brain [93] (Fig. 2c). Knockout of m⁶A reader Prrc2a in oligodendroglia cells inhibits OPC generation and proliferation, while it promotes oligodendrocyte differentiation [36]. Knockout of Nsun5 impairs OPC proliferation and reduces numbers of OPCs and oligodendrocytes [94, 95]. All these mice exhibit deficits in myelination, highlighting the importance of epitranscriptomics in regulating myelination. Regarding astrocyte generation, conditional knockout of Mettl14 in embryonic neural stem cells causes deficits in astrocyte generation in the postnatal mouse cortex [18]. Knockout of Mettl3 in retinal progenitor cells of the retina leads to extended and elevated generation of Müller glial cells during late-stage retinogenesis [96]. Besides, deletion of Ythdf2 in neurospheres leads to a dramatic reduction of GFAP⁺ astrocytes upon differentiation [80]. Despite these advances, it will also be interesting to investigate potential roles of epitranscriptomics in other cell types in the nervous system, such as microglia, choroid plexus, blood vessels, tanycytes, and ependymal cells.

Synaptic transmission

m⁶A methylated RNA and components of RNA modification machinery, such as FTO, METTL14, and YTHDF1/2/3, are localized near synapses [97–100]. Knockout of Mettl3 or Ythdf1 in the mouse hippocampus impairs synaptic transmission and long-term potentiation [101, 102], and knockout mice for Ftsj1 encoding a 2’-O-methyltransferase exhibit immature filopodia-like spines with fewer mushroom-type spines [103] (Fig. 2d). In addition, A-to-I editing of mRNAs occurs at high levels in the human brain and regulates synaptic transmission [104]. Pre-mRNAs of serotonin (5-HT) 2 C receptors (5-HT2CRs), AMPA-type glutamate receptors, and γ-aminobutyric acid type A receptors are heavily edited in a cell-type and brain-region-specific fashion and produce various protein isoforms that form receptors with different permeabilities and neurotransmission efficiencies [104]. As the most studied example, AMPA receptors lacking GluA2 subunits or containing GluA2 subunits encoded by unedited Grin2a are Ca²⁺-permeable, whereas GluA2 subunits encoded by an edited version of Grin2a form Ca²⁺-impermeable AMPA receptors [104] (Fig. 2d). Together, these studies point towards an important role of RNA modifications in modulating synaptic properties in the brain.

Learning and memory

Neuronal activity induces dynamic changes of m⁶A and m⁵C landscapes [13, 102, 105–107] and decreases A-to-I editing of Grik1 and CaV1.3 mRNA [108, 109] in various brain regions, affecting multiple steps of learning and memory formation [13, 97, 100–102, 109] (Figs. 2d, 3a). For example, postnatal knockout of Mettl3 or Ythdf1 in forebrain excitatory neurons in the mouse hippocampus leads to deficits in long-term memory formation [13, 101, 102], whereas Mettl14 knockout in the striatum alters dopamine signaling and impairs striatal learning [110]. In contrast, knockdown of Fto enhances fear memory [97, 111]. As one potential mechanism, the interaction between CYFIP2 and DPYSL2 RNA binding proteins and m⁶A-modified lncRNA metastasis-associated lung adenocarcinoma transcript 1 (Malat1) in synapses of the medial prefrontal cortex (PFC) is important for the consolidation of fear-extinction memory [100]. Moreover, a recent study of Qtrt1 (queuine tRNA-ribosyltransferase 1) knockout mice indicates that queuosine (Q) on tRNA has emerged as an essential component of learning and memory formation, exerting a greater influence in female mice [74] (Fig. 3a). In the future, systematic time-resolved characterization of all changes in RNA modifications and their signaling upon neuronal activation will provide a more complete picture on roles of RNA modifications and their combinations in learning and memory. Furthermore, achieving subcellular resolution is crucial, as one study discovered an upregulation of nuclear levels of METTL14 and FTO but a downregulation of postsynaptic FTO after a memory-inducing behavioral experience in mice [112].

Fig. 3. Functions of RNA modificaions in the adult nervous system.

Functions of RNA modifications in (a) learning and memory, (b) stress response, (c) neurodegeneration, (d) substance abuse disorders, (e) axon regeneration and brain injury, (f) post-ischemia recovery in the adult nervous system revealed by the gain- and loss-of-function experiments in mice and (g) in the adult human brain based on data from postmortem patient samples. TBI traumatic brain injury, CI-AMPAR calcium-impermeable AMPA receptor, CP-AMPAR calcium-permeable AMPA receptor; ↑↓: changes were observed in both directions depending on the brain region, environmental factors or specific transcripts.

Stress

Some stress responses in mice are also accompanied by dynamic m⁶A and m⁶Am level changes (antibodies used for the detection recognize both modifications (see Table S1)) in mRNA with hypomethylation in the PFC and hippocampus and hypermethylation in the amygdala [113, 114] (Fig. 3b). Chronic unpredictable stress-induced depressive-like behaviors in mice are attenuated by overexpression of circular RNA circSTAG1 that restores m⁶A levels on Faah mRNA via interaction with m⁶A RNA demethylase ALKBH5 [113]. RNA editing of 5-HT2CRs can produce 24 various isoforms that exhibit different activities and mice expressing the unedited isoform of 5-HT2CR mRNA showed anxiety-like and depression-like behavior and an increase in the behavior of despair [115, 116] (Fig. 3b). This phenotype is opposite in heterozygous Adar2 knockout mice, which are resistant to behavioral despair [117]. This is likely because multiple pre-RNAs of receptors and ion channels are edited in wild-type animals; thus, a decrease in ADAR2 levels affects not only mRNA of Serotonin (5-HT) 2 C receptors but all other neuronal receptors resulting in the different phenotypes observed in these two studies. In addition, Nsun2 neuronal deficient mice with alterations in m⁵C methylation show decreased translational efficiencies of glycine-rich proteins in neurons and altered glutamatergic synaptic signaling and antidepressant-like behavioral phenotypes [107] (Fig. 3b). Together, these studies revealed critical roles of various RNA modifications in stress-related behaviors and further studies can address the underlying molecular mechanisms.

Substance abuse

A common feature of many substance abuse disorders is dysregulation of dopaminergic midbrain circuitry. Changes in RNA modifications due to substance use may contribute to the formation of an addiction cycle (Fig. 3c). FTO is downregulated, and m⁶A levels of thousands of transcripts, many of which are involved in synaptic functions, are altered in the hippocampus of cocaine-conditioned mice [118], whereas Fto-deficient mice exhibit an increased sensitivity toward cocaine-induced locomotor activity and reward-stimulatory actions of cocaine [119]. Cocaine-seeking behavior in rats is accompanied by increased synaptic expression of calcium-permeable AMPA receptor channels in the nucleus accumbens (NAc), a key brain structure related to the rewarding effect of substance abuse. After 7 days of cocaine abstinence, ADAR2 expression and editing of the GluA2 subunit of the AMPA receptor are reduced in the NAc. Editing of the GluA2 subunit leads to the formation of calcium-impermeable AMPA receptors, while a decrease in editing leads to a higher amount of calcium-permeable AMPA receptors, promoting the reinstatement of cocaine-seeking behavior [120]. In addition, chronic consumption of ethanol is accompanied by ADAR1 and ADAR2 regulation and increased levels of RNA editing of Htr2C pre-mRNA in the NAc [121]. NAc-specific Adar2 knockout mice showed increased preference for ethanol induced by chronic exposure [122]. The discovery of the role of RNA modifications in substance abuse disorders opens avenues for developing novel treatments for addiction by modulating RNA modification levels.

Neuronal degeneration

The m⁶A pathway is altered in mouse models of multiple neurodegenerative disorders (Fig. 3d). The global m⁶A level is increased in the hippocampus of mice modeling Huntington’s disease (HD) [112], spinal cord samples from ALS patients, and rat models of ALS [123], but decreased in the brains of Alzheimer disease (AD) patients [124, 125], mouse models of cognitive decline [125], and in the striatum in rat models of Parkinson’s disease [126] (Fig. 3d, g). Mettl3 knockdown in the adult mouse hippocampus leads to neurodegeneration, gliosis, and impairment in recognition memory that recapitulate AD symptoms in patients [124]. Mechanistically, Mettl3 knockdown in cultured mouse [125] and rat [124] primary neurons leads to impaired synaptic translation and decreased neuronal activity [125] and dysregulation of cell cycle genes and subsequent postsynaptic deficits and neurite degeneration, which can be rescued by FTO inhibition [124]. In the brain of 5xFAD mice modeling AD, global m⁶A and METTL3 levels are decreased, whereas FTO levels are increased (Fig. 3d). Neuronal knockout or knockdown of Fto in the dorsal hippocampal CA1 region rescues cognitive deficits in the 3xTg mice modeling AD [127] and mice modeling HD [112] correspondingly. In mice modeling Nasu-Hakola disease, a heritable neurodegenerative disease resembling AD, METTL3, METTL14, and WTAP levels are decreased [128]. In contrast, the APP/PS1 mouse model of AD exhibits an increased level in global m⁶A methylation in the cortex and hippocampus [129]. This difference in various mouse models for AD requires further investigation. Overall, dysregulation of the m⁶A pathway appears to be a common feature of neurodegenerative diseases and raises the possibility of modulating the FTO level to manage the development and symptoms of neurodegenerative diseases.

Neuronal injuries and axon regeneration

Axonal injury in the adult peripheral and central nervous systems in mice results in a transient increase of m⁶A levels on regeneration-associated genes and protein translation machinery components, which leads to increased translation of these mRNA in a Mettl14- and Ythdf1-dependent fashion [130] (Fig. 3e). On the other hand, the FTO level is increased after peripheral nerve injury in rats, leading to a decrease in m⁶A levels on euchromatic histone lysine methyltransferase 2 mRNA [131], a key regulator of transcription and neuropathic pain symptoms. Studies of traumatic brain injury in rodent models uncovered dynamic injury-induced changes in m⁶A pathway components, including METTL3 [132], METTL14 and FTO [133], as well as many hypermethylated and hypomethylated transcripts [132, 133]. The post-ischemic stroke recovery success is also linked to dynamic epitranscriptomic changes, including a decrease of FTO [134, 135] and increased Alkbh5 and global m⁶A/m⁶Am levels [134, 135]. Transient focal cerebral ischemia in rats leads to decreased protein levels of YTHDF1, YTHDF2, and YTHDF3 and an increase of YTHDC1 and YTHDC2 in the ipsilateral cortex after reoxygenation [136] (Fig. 3f). Ischemic injury is accompanied by a massive Ca²⁺ overload and death of the ischemic-vulnerable neurons, which is characterized by reduced expression of ADAR2 post-ischemia [137–139] (Fig. 3f). RNA editing by ADAR2 of Grin2a pre-mRNA leads to the formation of Ca²⁺-impermeable AMPA receptor channels, thus protecting neurons from the massive Ca²⁺ influx [140]. As a result, overexpression of Adar2 protects vulnerable CA1 neurons during an ischemic injury in rats [139]. Time-resolved characterization of epitranscriptomic changes occurring post-injury may provide insight into developing therapeutic interventions to improve neuronal regeneration and stroke recovery by temporarily altering patterns of RNA modifications in affected brain regions.

EPITRANSCRIPTOMIC DYSREGULATION IN HUMAN BRAIN DISORDERS

Neurodevelopmental disorders

Mutations or aberrant expression of some genes encoding writers, readers, and erasers of different RNA modifications are correlated with microcephaly, intellectual disability, developmental delay, autism spectrum disorders, and other neurodevelopmental disorders in human patients [19, 34, 51, 60, 83, 141–159] (Fig. 4 and Table S3). Notably, most of these mutant genes encode regulators of cytosolic/mitochondrial tRNA modifications, which are usually critically involved in modulating the folding and codon-anticodon recognition of tRNA to tune cytosolic/mitochondrial protein translation, reflecting the importance of tRNA modifications and protein translation in brain development. Besides, the necessity of depositing methyl groups to diverse RNAs in brain development reemphasizes the importance of methyl donor-folic acid supplement during pregnancy. Indeed, compared with control samples, significantly reduced mRNA m⁶A modification levels and Mettl3 expression levels were observed in the brain tissues of mouse models with neural tube defects (NTD) induced by administration of retinoic acid and ethionine [160, 161]. Interestingly, treatment of S-adenosylmethionine (SAM), a methyl donor, can restore reduced mRNA m⁶A modification levels in the NTD mouse model, while there is no evidence indicating that SAM administration can also rescue neural tube defects [161], which needs further examination. Despite these observed connections between human neurodevelopmental disorders and mutations in epitranscriptomic regulators, causal links remain to be established in animal or human brain organoid models for most cases.

Fig. 4. Alterations in RNA modification pathways associated with brain disorders in humans.

mcm⁵U: 5-methoxycarbonylmethyl uridine; mcm⁵s²U: 5-methoxycarbonylmethyl-2-thiouridine; ncm⁵U: 5-carbamoylmethyl uridine.

Glioblastoma and neuroblastoma

The epitranscriptome is also critically involved in the pathogenesis of primary brain tumors, including glioblastoma multiforme (GBM) and neuroblastoma [22, 51, 162–186] (Fig. 4 and Table S4). GBM is often associated with high levels of expression of both m⁶A methyltransferase-related proteins, METTL3 and WTAP, and ALKBH5 (Fig. 3g), while higher glioma grades and poorer clinical outcomes are found to be correlated with decreased FTO expression and increased expression of YTHDF2 and YTHDF3 [174–179]. Functionally, knockdown of METTL3, METTL14, YTHDC1 and FTO, or overexpression of METTL14, WTAP, YTHDF1 and IGF2BP3 promotes glioma stem cell (GSC) proliferation, self-renewal, and tumorigenesis; on the contrary, overexpression of METTL3 and FTO, silencing ALKBH5, WTAP, METTL3, YTHDF1, YTHDF2 and IFG2BP3, or treatment with FTO inhibitors (including MA2, R-2HG, and FTO-04) inhibits proliferation and self-renewal of GSCs and GSC-induced tumorigenesis [174, 175, 178–187]. The variable effects of manipulating different m⁶A regulators on GSC proliferation and tumorigenesis in different studies probably reflects different cancer cell lines examined. Overall, WTAP and YTHDF proteins seem to play a consistently oncogenic role in promoting GBM tumorigenesis, raising the possibility of developing inhibitors to treat GBM.

In addition, higher expression levels of regulators including DKC1 encoding a snoRNA-dependent Ψ synthase, PUS7 encoding a RNA-independent Ψ synthase, several m⁵C RNA methyltransferases, FTSJ2 encoding a rRNA 2′-O-methyltransferase, TRMT6 encoding a m¹A RNA methyltransferase, and METTL1 encoding a m⁷G tRNA methyltransferase, have been reported to be associated with patients with glioma and neuroblastoma and often predict a poorer prognosis [22, 51, 165–172] (Table S4). At the molecular level, overexpression of PUS7 and NSUN5 promotes GSC proliferation, self-renewal and growth, whereas knockdown of PUS7, DKC1, NSUN5, TRMT6, and METTL1 and treatment with PUS7 inhibitors attenuates glioma cell proliferation, migration, and tumorigenesis [22, 165, 166, 168, 170–172]. Collectively, aberrant expression levels of multiple regulators of RNA modifications, including m⁶A, Ψ, m⁵C, 2’-O-methylation, m⁷G, and m¹A, have the potential to be biomarkers for predicting the occurrence and prognosis of brain tumors as well as potential therapeutic targets of brain tumors. As oncogenes or tumor suppresser genes are often master genes regulated by RNA modifications, inhibitors and activators of these RNA modification regulators may be effective in attenuating brain tumors, such as METTL3 inhibitors (STM2457, UZH1a), FTO inhibitors (MA2, R-2HG, and FTO-04), YTHDF1 inhibitors (Ebselen), PUS7 inhibitors (C17), and DKC1 inhibitors (pyrazofurin) [165, 183, 185, 188–191]. In contrast to disparate roles of m⁶A mRNA modifications in regulating glioma tumorigenesis, which may be attributed to the fact that significantly more studies using various glioma samples have been conducted on m⁶A compared to other types of RNA modifications, the oncogenic roles of some RNA modifications, such as m⁷G and their modifiers, have been relatively consistent to date in glioma, thus deserving more effort to develop small molecule inhibitors to test their therapeutic efficacy.

Psychiatric disorders

Multiple studies have revealed alterations in RNA modifications in patients with psychiatric disorders (Figs. 3 and 4). Global m⁶A/m methylation levels in whole blood were transiently reduced after an acute stressful challenge and glucocorticoid stimulation in healthy subjects, but not in patients with major depressive disorders (Fig. 5a) [114]. In addition, the level of circular RNA circSTAG1, which regulates the m⁶A methylation level of Faah mRNA in the mouse hippocampus, was significantly decreased in the plasma and whole blood of patients with depression [113]. A number of genetic variants have been found to be associated with an increased risk for major depressive disorders, such as FTO (rs9939609) [192] and ALKBH5 (rs12936694) [193]. A pilot study also found changes in m⁶A methylation of mRNAs involved in inflammation and immune responses in postmortem NAc from three males with alcohol use disorder compared to three matched control subjects [194]. In another study, increased expression and decreased m⁶A methylation of BDNF-AS lncRNA was found in the postmortem amygdala in early-onset alcohol use disorders [195]. An association study of patients with major depressive disorder, bipolar disorder, and schizophrenia found rs9983925 and rs4819035 polymorphisms in ADARB1 to be associated with suicide risk [196, 197], indicating that alterations in RNA editing may also be a risk factor for suicidal behavior. There are region-specific alterations of RNA editing and expression levels of cyclic nucleotide phosphodiesterase 8 A mRNA in the cortex [197] and 5-HT2CR pre-mRNA in the neocortex [198] of patients of major depressive disorders with suicide risk. ADAR2 expression is often decreased in the postmortem brains of patients with schizophrenia and bipolar disorder and an RNA editing deficit was found in many cases of schizophrenia [117, 199]. To note, there is a trend for RNA hypoediting in schizophrenia patients of European descent [200]. These studies showed a promising direction for developing biomarkers for early diagnostics of psychiatric disorders and prediction of suicide behavior (Fig. 5a).

Fig. 5. Potential translational opportunities of epitranscriptomics in the nervous system.

a Application of epitranscriptomics as biomarkers for human brain disorders. The levels of RNA modifications, expression levels of RNA modification regulators, and mutations of RNA modification regulators can potentially be established as biomarkers for diagnostics or prognosis predictors of brain disorders, such as glioma (a1) or psychiatric disorders (a2). b Drugs and gene therapy for RNA modifications. Inhibitors of RNA modification regulators or CRISPR/Cas13-based RNA modification editing can be developed and used to manipulate the levels of RNA modifications and alleviate symptoms of brain disorders. For instance, FTO (the m⁶A eraser) and PUS7 (Ψ writer) inhibitors have the potential of attenuating the growth of GSCs (b1), and FTO inhibitors have the potential of restoring the cognitive deficits in AD and HD patients based on Fto KD and KO studies in mouse models (b2). c Mechanism-based downstream intervention. Deciphering dysregulated downstream molecular pathways mediating the pathogenesis of epitranscriptomics-related brain disorders can help to attenuate diseases by targeting and restoring specific dysregulated downstream molecular pathways or cellular processes. For example, impaired NPC maintenance caused by attenuated mitochondria activity in Mettl8 (mt-tRNA m³C writer) knockout mice can be rescued by pharmacologically enhancing mitochondria function with piracetam treatment (c1). d mRNA vaccines and mRNA therapy. Application of RNA modifications can enhance the protein expression and reduce the immunogenicity of mRNA introduced into the cells, potentially increasing the efficacy of mRNA vaccines (d1) and mRNA therapy (d2) targeting various brain disorders.

Neurodegenerative disorders

The expression of m⁶A machinery components is altered in patients with multiple sclerosis (MS) compared to matched controls (Fig. 4), leading to increased m⁶A levels in cerebrospinal fluid from patients with relapsing–remitting MS and progressive MS [201]. Similarly, an increase in m⁶A levels and hypermethylation of mRNAs and lncRNAs were found in postmortem spinal samples from patients with sporadic ALS [123]. In contrast, reduced levels of m⁶A in mRNA and m⁶A regulators were found in the postmortem brains of patients with AD, even at early stages of the disease (Fig. 3g) [124]. Genetic studies have also revealed an association of FTO (rs9939609) with a higher risk for AD and dementia [202]. Alterations in A-to-I RNA editing levels were also found in postmortem brain samples of patients with AD, mainly in the hippocampus and, to a lesser degree, in the temporal and frontal lobes [203, 204]. In addition, decreased ADAR2 expression was found in motor neurons of patients with sporadic ALS [205] (Fig. 3g). Together, these preliminary studies have identified genetic associations among RNA modification regulators with different neurodegenerative disorders and dysregulated RNA modifications in patient samples (Fig. 4). Future large cohort studies are needed to validate these initial findings to establish potential biomarkers for disease detection and to identify therapeutic targets for intervention.

TRANSLATIONAL OPPORTUNITIES

Aberrant expression or mutations of RNA modification regulators and dysregulated RNA modification levels in various brain disorders, such as neurodevelopmental diseases, psychiatric disorders, and brain tumors, highlights not only the biological significance of RNA modifications in regulating the nervous system, but also the potential translational opportunities of applications of epitranscriptomics in the diagnosis and treatment of brain disorders.

Epitranscriptomics as biomarkers for human brain disorders

Aberrant levels of RNA modifications and RNA modification regulators that accompany the progression of brain disorders can potentially serve as diagnostic biomarkers and prognosis predictors (Fig. 5a). For example, patients with relapsing–remitting MS and progressive MS have increased m⁶A levels in cerebrospinal fluid compared to healthy controls [201]. The m⁶A/m levels in blood after glucocorticoid stimulation is also a promising biomarker of major depressive disorder [114]. An RNA editing signature in whole blood was successfully used to identify patients with bipolar disorder among depressed patients to facilitate appropriate treatment in a timely manner [206]. Also, aberrant expression levels of multiple regulators of RNA modifications, including pseudouridine, m⁷G, and m¹A, have the potential to be biomarkers for predicting the occurrence and prognosis of brain tumors. The main challenge for developing epitranscriptomics-based biomarker panels is establishing and expanding the strong association between disease conditions and epitranscriptomic signatures in biosamples, such as blood or cerebrospinal fluid, from a large cohort of patients with diverse genetic backgrounds.

Drugs or gene therapies that target epitranscriptomic pathways

Realization of clinical applications of RNA modifications in treating diseases requires more comprehensive understanding of key dysregulated RNA modifications and molecular pathways involved in pathogenesis and the development of efficient and specific strategies for manipulating RNA modifications, such as agonists or inhibitors of RNA modification regulators and CRISPR/Cas13-based or engineered snoRNA-mediated RNA modification editing [207, 208] (Fig. 5b). Inhibitors of RNA modification regulators, which can be screened from various natural products or synthetic chemicals, are gaining more interest for treating brain disorders, including glioma and neurodegenerative diseases [127, 183, 185, 188–191]. For example, inhibitors of m⁶A eraser-FTO and pseudouridine synthase PUS7 have been reported to attenuate proliferation and tumorigenesis of GSCs [165, 183, 185, 187], suggesting their therapeutic potential in treating brain tumors. Fto knockdown in the dorsal hippocampal CA1 region improved behavior and cognitive defects in a mouse model of HD [112], and the neuronal Fto knockout rescued the cognitive deficits in 3xTg mouse model of AD [127], suggesting the potential of FTO inhibitors to mitigate symptoms in human neurodegenerative diseases. Overexpression of circular RNA circSTAG1 prevents ALKBH5 translocation into the nucleus, restoring m⁶A levels on Faah mRNA and attenuating depressive-like behaviors in chronic unpredictable stress-treated mice [113], suggesting the potential of ALKBH5 inhibitors for depression treatment. In addition, two recent studies showed that directed site-specific pseudouridylation of premature termination codons (PTCs) mediated by engineered small nucleolar RNAs (snoRNAs) enhances nonsense suppression and PTC readthrough to produce full-length functional proteins in mammalian cells, revealing the therapeutical potential of targeted pseudouridylation in treating PTC-associated brain disorders, such as spinal muscular atrophy caused by nonsense mutations in the SMN1 gene [208, 209].

Mechanism-based therapies

A better understanding of cellular mechanisms underlying deficits due to mutations of epitranscriptomic pathways can suggest mechanism-based downstream interventions (Fig. 5c). For example, in MELAS disease affecting mitochondria, mutations in U34 of mitochondrial tRNA^Leu(UAA) leads to deficiency of its 5-taurinomethyluridine (τm⁵U) modification and impaired mitochondrial function [68, 210]. Deficiency in mitochondrial function in MELAS could be restored by expression of the other isoacceptor-mitochondrial-tRNA^Leu(UAG) with a mutated anticodon to (UAA), which can be τm⁵U34 modified [68, 211], indicating that introducing tRNA or other RNAs with proper modification can restore defects in diseases caused by hypomodified tRNA or other RNAs. In another example, impaired NPC maintenance in mice or human forebrain organoids with knockout of Mettl8 encoding mitochondrial tRNA^Thr/Ser(UCN) m³C methytransferase is due to attenuated mitochondrial activity, whereas pharmacologically enhancing mitochondrial activity with piracetam treatment can largely restore NPC maintenance [46].

mRNA vaccines and therapy

One of the best examples of a clinical application of epitranscriptomics is the recent success of COVID-19 mRNA vaccines utilizing RNA modifications to enhance efficacy and reduce immunogenicity [12, 212]. In addition, several RNA-based drugs have been approved for various neurological diseases by the FDA, mainly focused on siRNA and antisense oligonucleotides (AONs) to reduce expression of genes with pathogenic mutations or correct splicing defects of mutated pathogenic genes [213]. Utilization of RNA modifications when designing antisense oligonucleotides improves their binding affinity to target RNAs and their resistance to nuclease degradation [213]. Furthermore, the profound success of COVID-19 mRNA vaccines led the field to re-evaluate potential applications of RNA modifications in mRNA vaccines and mRNA therapy for attenuating brain disorders, such as neurodegenerative disorders and brain tumors [12, 212] (Fig. 5d). For example, given the inefficiency of Aβ vaccine in ameliorating symptoms of AD [214], mRNA vaccines against toxic Aβ protein could be an alternative. Similarly, it will also be promising to develop mRNA vaccines against brain tumor-specific antigens or viruses infecting brain tissue, such as ZIKA virus and viral encephalitis [215–217]. Although transdifferentiation from astrocytes to neurons in vivo through infecting astrocytes with AAV virus expressing factors such as NeuroD1, remains controversial [218, 219], this strategy may be improved by NeuroD1 mRNA therapy with an astrocyte-specific delivery strategy. In the case of expressing exogenous proteins, such as neurotrophic factors to improve neuronal function and activity in neurodegenerative disorders, reversible and transient mRNA therapy could be a competitive and safe choice compared to AAV virus-based gene therapies [220, 221].

FUTURE DIRECTIONS

To date, clinical applications of epitranscriptomics for treating brain disorders is still in a nascent stage. Despite many correlations between mutations or aberrant expression of RNA modification regulators and brain disorders in patients, causal links need to be validated in animal or human brain organoid models with corresponding genetic manipulations. It will also be useful to systematically evaluate levels of various RNA modifications and different RNA modification regulators in biosamples from patients with neurodevelopmental, psychiatric disorders or brain tumors versus healthy people to identify potential biomarkers. These advances will lead to a more comprehensive understanding of pathogenesis mechanisms of brain disorders and lay the foundation for unraveling the therapeutic potential of manipulating levels of some specific RNA modifications or their regulators.

Detection and mapping methods

In recent years, a plethora of innovative methods for detecting RNA modifications [6, 23–33] (Table S1) has played a pivotal role in enhancing our understanding of their roles in the pathogenesis of brain disorders. However, to pave the way for the creation of safe and efficacious epitranscriptome-based therapeutics, there remains a pressing need for continued advancements in detection methodologies. Considering the high complexity and tissue specificity of RNA modifications’ repertoire, development of RNA modification detection methods that retain spatial and cell-type information in the brain will provide an exciting opportunity to understand the complexity of brain region- and cell type-specific roles of RNA modifications in brain tissues, including human postmortem brains. A combination of spatial transcriptomics with sequencing-based methods for RNA modification detection and single-cell RNA modification analysis is the next logical step in method development. In addition, understanding the crosstalk between different RNA modifications and epitranscriptomic codes on one RNA molecule is an exciting new direction that might help to decipher the complexity of RNA modification-based regulation in the nervous system. Methods are needed to simultaneously detect different RNA modifications on one RNA molecule. Direct RNA sequencing, such as Nanopore sequencing, has great potential to become a gold standard in the RNA modification detection field should technical and computational challenges be addressed. Moreover, for diagnostic purposes, working with human samples requires the development of RNA modification detection methods with clinical feasibility that do not rely on transgenic protein expression or culturing cells.

Understanding RNA modification mechanisms of action in disease progression

Despite recent significant advances, only a handful of the over 170 RNA modifications have been investigated for their roles in regulating the nervous system, while the function of many others, such as RNA glycosylation on the cell surface, remains ambiguous. In addition, the roles of RNA modifications in other types of cells in the brain beyond neurons, such as oligodendrocytes, astrocytes, microglia, choroid plexus, brain blood vessels, tanycytes, and ependymal cells, are much less clear. Comprehensive knowledge of the epitranscriptomic landscapes in diverse types of cells in the brain would help us better understand their roles in the pathogenesis of brain disorders involving different cell types and develop cell-type-specific therapeutic interventions to treat brain disorders.

While mouse models for human brain disorders are very useful for developing new strategies for managing disease symptoms and progression, caution should be taken while transferring knowledge gained from mice models to patients, especially in the case of multifactorial diseases, such as AD. For instance, Fto downregulation mitigates symptoms in mouse models of HD [112] and AD [127], yet the APP/PS1 mouse model of AD is characterized by an increase in global m⁶A methylation in the cortex and hippocampus [129]. This warrants more research to better understand the molecular mechanisms of how specific RNA modifications and their combinations affect disease progression, in which human cell-based models, such as brain organoids, will be particularly useful.

Developing novel therapeutics

Regarding the clinical use of the small molecule inhibitors of RNA modification regulators in attenuating brain disorders and taking glioma as an example [183, 185, 188–191], manipulating the same RNA modifications can have opposite effects in different glioma cells, either promoting or attenuating tumorigenesis. Therefore, personalized medicine should be a priority, depending on the directionality of the change of RNA modifications and affected key downstream genes in diseases. Other issues also need to be considered, such as blood-brain-barrier (BBB) penetration, and the specificity and side effects of small molecules. For example, some drugs designed for inhibiting m⁵C RNA methyltransferases can often have cross-activity on other targets, like m⁵C DNA methyltransferases, which may have unexpected consequences [22]. Interestingly, photoactivated compounds, such as a caged molecule activator of METTL3/14, can potentially increase the spatiotemporal specificity and reduce side effects of small molecule drugs of RNA modifiers in vivo [222, 223]. Furthermore, the specificity of manipulating RNA modifications can be improved by taking advantage of CRISPR/Cas13-based RNA modification editing. Like CRISPR/Cas9-mediated genome-editing, use of the CRISPR/Cas13 system makes it possible to target specific key RNA molecules and manipulate their specific RNA modifications [207, 224]. CRISPR/Cas13 fused with RNA modification regulators, such as FTO, METTL3/14, METTL1, and YTHDFs, to specifically remove, deposit, or target the modifications of specific RNAs, especially for key master regulator genes of signaling pathways, genes with mutations or aberrant expression, and tumor suppresser genes or oncogenes, would potentially reduce side effects of manipulating all RNA modifications.

Currently, applications of RNA modifications in mRNA vaccines and therapy mainly focus on the 5′ cap analog and uridine moieties. Whether combining other types of RNA chemical modifications with varying compositions and depositing RNA modifications to different regions of mRNA (5’ UTR, 3’ UTR, and open reading frame) can further improve the expression efficacy and reduce the immunogenicity of mRNA-based vaccines and therapy warrants more investigation in the future [212]. Moreover, the lack of safe and efficient delivery methods of mRNA therapy into the nervous system still greatly hinders its clinical applications and requires significant improvement [212, 213].

CONCLUSION

Epitranscriptomics holds great promise in revolutionizing diagnosis and treatment of brain disorders. RNA modifications, which influence various biological processes in the nervous system, have been linked to a wide range of neurodevelopmental, neurological, and psychiatric disorders. Epitranscriptomics-based therapies may potentially offer advantages such as flexibility, safety, and robust therapeutic effects, making them an attractive approach for treating brain disorders. Moreover, RNA modifications can be directly utilized to enhance the effectiveness of mRNA vaccines and may serve as valuable biomarkers for brain tumors and psychiatric disorders. However, widespread clinical application of epitranscriptomics-based therapies necessitates a deeper understanding of the causal relationship between altered RNA modifications and levels of their regulators and development of brain disorders. It is crucial to systematically evaluate RNA modifications and their regulators in both healthy and diseased brains, while also developing more sensitive detection methods capable of preserving spatial information and being readily implemented in clinical settings. Development of novel animal models and human brain organoids specific to brain disorders will be essential in identifying new strategies for managing symptoms and disease progression through epitranscriptomic modulation. Finally, gaining a comprehensive understanding of the epitranscriptome functions across diverse types of brain cells will expand our knowledge on the role of neuroepitranscriptomics in the pathogenesis of brain disorders involving multiple cell types, contributing to the development of cell-type-specific therapeutic interventions. Together, these research efforts hold the potential to transform our understanding of molecular mechanisms underlying brain disorders and pave the way for the development of new and more effective therapies.