Chromatin and transcriptional regulation by reversible RNA methylation

Abstract

Dynamic RNA modifications have been a burgeoning area in the last decade since the concept of “RNA epigenetics” was proposed [1]. N⁶-methyladenosine (m⁶A) is the most abundant mRNA modification in eukaryotic cells. It can be installed by “writers”, removed by “erasers”, recognized by “readers”, and dynamically regulate the fate of methylated RNA. Until recently, the roles of reversible RNA methylation in chromatin and transcriptional regulation were not adequately studied. We discuss the new discoveries and insights into the chromatin and transcriptional regulation by m⁶A through two pathways: i) effects of m⁶A on mRNAs encoding histone modifiers and transcriptional factors; ii) m⁶A regulation of chromatin-associated regulatory RNAs. Additionally, we provide an outlook on how the transcriptional regulation by RNA m⁶A could add an additional critical layer to transcriptional regulation.

N⁶-methyladenosine (m⁶A)

m⁶A is the most prevalent and well-characterized internal mRNA modification in higher eukaryotes [2]. Besides mRNA, m⁶A has been shown to widely exist in ribosomal RNAs (rRNAs), long non-coding RNAs (lncRNAs), microRNAs (miRNAs), small nuclear RNAs (snRNAs), and circular RNAs (circRNAs) [3••].

“Writers”, “erasers”, and “readers”, which respectively deposit, remove, and recognize the m⁶A on RNA, are collectively known as m⁶A effectors [3••]. A large portion of mRNA m⁶A is installed co-transcriptionally by a methyltransferase complex (MTC) composed of the core methyltransferase-like 3 (METTL3) and methyltransferase-like 14 (METTL14) [4–6]. MTC is associated with additional subunits, including Wilms tumor 1-associating protein (WTAP) [4], Vir like m⁶A methyltransferase associated (VIRMA) [7], zinc finger CCCH-type containing 13 (ZC3H13) [8], and RNA binding motif protein 15/15B (RBM15/15B) [9]. More recently, methyltransferase-like 16 (METTL16) [10,11], methyltransferase-like 5 (METTL5) [12], and Zinc finger CCHC-type-containing 4 (ZCCHC4) [13] have been identified as m⁶A methyltransferases that can install internal m⁶A on RNA species possessing certain structures, such as small nucleolar RNAs, pre-mRNAs, and rRNA. Fat mass and obesity-associated protein (FTO) [14,15] and AlkB homolog 5 (ALKBH5) [16] are the only two known demethylases that can actively remove the m⁶A on RNA.

m⁶A can be preferentially “read” via three mechanisms: i) m⁶A directly recruits readers such as the YTH domain family proteins and the associated regulatory machinery [17–19]; ii) m⁶A repels certain RNA binding proteins (RBPs) that prefer unmodified A over m⁶A [20]; iii) m⁶A changes the secondary structures to alter protein-RNA interactions [21]. Numerous m⁶A readers have been reported that include YTHDF1–3 [18,19,22], YTHDC1–2 [23–26], insulin-like growth factor 2 mRNA-binding proteins (IGF2BPs) [27], heterogeneous nuclear ribonucleoproteins (HNRNPs) [21,28–30], Fragile X mental retardation 1 (FMR1) [31,32], and proline rich coiled-coil 2A (Prrc2a) [33]. The list of potential m⁶A readers is still growing.

In parallel to the well-known reversible epigenetic modifications of histone and DNA, studies have demonstrated the dynamic regulatory roles of m⁶A on mRNA in response to development signaling and environmental stimulation and stresses. m⁶A impacts the fate of modified RNA in many aspects, including pre-mRNA processing, nuclear export, stability, translation, and RNA-protein granule formation [3••], thereby leading to profound impacts on mammalian development and human diseases [34,35••]. m⁶A on mRNA is enriched around the stop codon and in 3’ UTRs; functions of these m⁶A marks on mRNA can be context-dependent (see review [3••]). For instance, 5’ UTR m⁶A is linked to cap-independent translation; while 3’ UTR m⁶A can mediate cap-dependent translation. As for m⁶A readers, IGF2BPs can bind to m⁶A at both CDS region and 3’ UTR; whereas YTHDF2 and YTHDF1 preferentially bind m⁶A at CDS and 3’ UTR, respectively. A systematic profiling of the binding sites of m⁶A effectors and the altered m⁶A sites by each effector can help understand the mRNA region-specific m⁶A function and regulation. Site-specific m⁶A modulation could further establish functional importance of the specific m⁶A sites on mRNA [36].

The functional studies during the past decade mainly focused on how m⁶A plays regulatory roles in gene expression at the post-transcriptional level. However, there have been emerging studies in recent years which also revealed critical roles of RNA m⁶A methylation on chromatin regulation. RNA m⁶A can modulate transcription and chromatin state (Figure 1). The currently known pathways are: i) mRNA m⁶A impacts translation and stability of the methylated transcripts that encode histone modifiers, transcriptional factors, or key proteins in transcription regulation; ii) m⁶A directly regulates the cellular fate of critical nuclear RNAs, especially chromatin-associated regulatory RNAs (carRNAs). In this review, we summarize these recent advances in our understanding of how m⁶A functions in transcriptional regulation, highlight the novel mechanisms by which m⁶A affects transcription and chromatin structure, and the key questions that remain to be investigated in the future.

Figure 1.

Divergent roles of RNA m⁶A in R-loop formation under different cellular contexts. (a) RNA m⁶A in RNA:DNA hybrids may be recognized by the reader YTHDF2, which facilitates the degradation of RNA in RNA:DNA hybrids and subsequently destabilizes RNA:DNA hybrids in hPSCs cells. (b) At DSBs, YTHDC1 is recruited to RNA m⁶A sites to protect the methylated RNA, leading to the accumulation of RNA:DNA hybrids in U2OS cells.

Transcriptional regulation by mRNA m⁶A

m⁶A can be deposited on mRNAs encoding histone modifiers and transcription factors to affect transcription regulation. In mouse embryonic neural stem cells (NSCs), Mettl14 knockout leads to genome-wide changes in specific histone modifications in an m⁶A-dependent manner. CBP and p300 transcripts, which encode acetyltransferases for the active transcription mark H3K27Ac are m⁶A methylated. Upon loss of m⁶A, these transcripts are stabilized and lead to increased H3K27Ac, which is associated with decreased proliferation and premature differentiation of NSCs [37•]. Likewise, in adult neural stem cells (aNSCs), METTL3 knockdown leads to reduced levels of both m⁶A and expression of EZH2 mRNA. EZH2 is a functional enzymatic component catalyzing the methylation of the repressive histone mark H3K27Me3. Overexpression of EZH2 could rescue the defects of neurogenesis and neuronal development induced by loss of Mettl3 [38]. In bacteria-induced inflammatory response, mRNA of KDM6B has been shown to be m⁶A-modified and undergo YTHDF2-mediated degradation. Loss of YTHDF2 stabilizes the KDM6B transcript to induce H3K27Me3 demethylation of proinflammatory cytokines thereby enhancing their transcription [39]. All three cases indicate mRNA m⁶A could further modulate gene expression through transcriptional regulation by histone modifications.

In acute myeloid leukaemia (AML) cells, METTL3 can associate with chromatin and localize to the transcriptional start site (TSS) of active genes. Promoter-bound METTL3 deposits m⁶A in the coding region of the associated mRNA to promote its translation. The transcripts encoding transcription factors SP1 and SP2 are two key substrates of METTL3. METTL3 depletion leads to significantly reduced protein expression of SP1 and SP2, indicating a potential activating role of mRNA m⁶A in SP1- and SP2-dependent transcription [40•]. In addition, in certain cancer cells, the m⁶A reader IGF2BP1 could increase the transcription factor SRF abundance by stabilizing its 3’UTR m⁶A, thereby promoting SRF-dependent transcriptional activity [41].

mRNA m⁶A has also been shown to play important roles in circadian rhythm maintenance [42] and cell cycle regulation [43]. Considering circadian rhythm could involve a negative feedback loop on gene expression through transcription repression [44], and cell cycle regulation is closely connected to histone mRNA metabolism, DNA replication, and transcription [45,46], one can speculate that m⁶A on mRNAs that encode clock proteins and cell cycle markers could also indirectly regulate transcription and result in prolonged effects.

Transcriptional regulation by m⁶A on Pre-mRNA

It is not surprising that many mRNAs encoding key proteins in transcriptional regulation are m⁶A methylated, therefore m⁶A could affect transcription through tuning the levels of these proteins. In addition to the transcriptional regulation of the m⁶A on mature mRNA at the post-transcriptional level, m⁶A can also modulate transcription at the pre-mRNA stage.

m⁶A has been shown to affect the stability of R-loops, which affects transcription initiation, elongation, and termination. In human pluripotent stem cells, the presence of m⁶A is found on the RNA moiety in the formation of most RNA:DNA hybrids. In human pluripotent stem cells (hPSCs), the reader protein YTHDF2 can be recruited to these m⁶A sites where it subsequently facilitates the degradation of RNA:DNA hybrids. METTL3 depletion leads to the accumulation of m⁶A-methylated R-loops and elevated DNA double-strand breaks, implying a critical role for m⁶A processing in safeguarding genomic stability (Figure 1a) [47•]. In Hela cells, m⁶A has been shown to exist in a subset of R-loops. Instead of accelerating the removal of R-loops, m⁶A promotes R-loop formation to facilitate transcription termination. Therefore, METTL3 modulates transcription termination through m⁶A methylation of RNA involved in R-loop formation [48]. More recently, a study of DNA double-strand break (DSB) repair provided evidence that RNA m⁶A leads to the accumulation of RNA:DNA hybrids. In human sarcoma U2OS cells, methylated RNA by METTL3 is enriched at DSBs, which could be recognized and protected by YTHDC1. The accumulation of RNA:DNA hybrids at DSBs induced by the METTL3-m⁶A-YTHDC1 axis could further facilitate homologous recombination (HR)-mediated DSB repair (Figure 1b) [49•].

Emerging evidence suggests a connection between m⁶A and the transcriptional process [40,50–53], which is supported by the genetic analyses of m⁶A quantitative trait loci (QTLs) [54••]. m⁶A deposition has been shown to be affected by the progression rate of RNA polymerase II (RNAP II) [51], and m⁶A has also been proposed to regulate the occupancy and promoter-proximal pausing of RNAP II. m⁶A sites in pre-mRNA could promote the binding of an m⁶A reader hnRNPG to modulate RNAP II occupancy thereby affecting alternative splicing likely through RNAP II pausing [55•]. In Drosophila, MTC interacts with chromatin and binds to promoters to aid the release of RNAP II from the paused state. Knockdown of METTL3 or METTL14 hinders transcriptional elongation, while tethering METTL3 to a heterologous gene promoter is sufficient to increase RNAP II pause release, which reveals positive feedback of m⁶A on the transcription machinery [56]. It is known that METTL14 of MCT can recognize H3K36Me3, which co-transcriptionally facilitates the deposition of m⁶A on mRNA [53]. The model of the co-transcriptional interplay between m⁶A and histone modifications also reveals the transcriptional induction role of RNA m⁶A. m⁶A could co-transcriptionally direct KDM3B by the reader YTHDC1 to target chromatin to demethylate H3K9Me2. Since H3K9me2 generally leads to gene repression, the existence of RNA m⁶A promotes the demethylation of H3K9me2 and consequently promotes the overall transcriptional activity in gene bodies [57].

The m⁶A methylation on pre-mRNA can be reversible and may play roles in transcription and pre-mRNA processing as well as nuclear decay of certain RNA species to affect transcription outcome. A recent study also showed that the stability of DNA:RNA hybrid structures could be regulated by METTL8-mediated RNA 3-methylcytosine (m³C) methylation [58].

Transcriptional regulation by m⁶A on carRNAs

Phenotypes of cytoplasmic m⁶A reader knockout mice may not fully recapitulate the severe phenotypes of the writer KO mice in many development processes. These observations suggest critical nuclear regulation events mediated through RNA m⁶A methylation. Our research revealed that in mouse embryonic stem cells (mESCs) the knockout of the writer or a nuclear m⁶A reader led to substantial changes in the chromatin accessibility. We further found that carRNAs, including promoter-associated RNAs (paRNAs), enhancer RNAs (eRNAs), and repeat RNAs, undergo m⁶A methylation which regulates their levels and downstream transcription (Figure 2a) [59••].

Figure 2.

m⁶A destabilizes carRNAs and suppresses transcription. (a) m⁶A on carRNAs can be deposited by METTL3. At least a subset of m⁶A sites can be recognized by YTHDC1, resulting in the recruitment of the NEXT complex to promote the degradation of the methylated carRNAs, which ultimately leads to less active transcription. (b) FTO can demethylate m⁶A within repeat RNAs, especially LINE1 RNA, thereby affecting their abundance. Loss of FTO leads to more closed chromatin and reduced transcriptional activity in mESCs and mouse tissues.

In mESCs, knockout of Mettl3 or Ythdc1 increases global chromatin accessibility and activates nascent transcription. A portion of m⁶A on carRNAs is installed by METTL3 and negatively correlates with carRNA expression. YTHDC1 recognizes and recruits the nuclear exosome targeting (NEXT) complex to some of these m⁶A sites, which ultimately facilitates the degradation of at least a subset of methylated carRNAs, in particular repeat RNAs such as LINE1 RNA. Upon METTL3 depletion, m⁶A-methylated carRNAs display reduced m⁶A levels and increased abundance. Meanwhile, the downstream genes of these methylated carRNAs exhibit elevated transcription rates. Moreover, loss of m⁶A on carRNAs leads to enrichment of active histone markers and open chromatin at carRNA loci. Both transcription and chromatin state regulation can be recapitulated by site-specific m⁶A demethylation on target carRNAs, further validating the necessity of the m⁶A-dependent regulation. Note that in mESCs and endometrial cancer cells this chromatin regulation through carRNA m⁶A methylation dominates; however, in other cancer cells the mRNA m⁶A methylation on transcripts encoding histone modifiers and transcriptional factors can dominate. Cell context is important when dissecting the exact regulatory pathways.

Reversible m⁶A methylation on repeat RNAs and caRNAs

FTO demethylates m⁶A on chromatin-associated RNAs (caRNA) that include pre-mRNA and carRNAs in mESCs [60••]. As the first RNA m⁶A demethylase, FTO has been shown to impact a range of biological processes. It plays vital roles during mammalian development; the Fto^−/− mice display intriguing phenotypes, including a notable proportion of embryo and postnatal lethality as well as postnatal growth retardation [61–64]. However, the physiologically relevant substrate(s) of FTO which contribute to these phenotypes has not yet been identified.

Our recent study found that FTO demethylates m⁶A within repeat RNAs, especially LINE1 RNA, which affects their abundance and consequently regulates chromatin state in mESCs, mNSCs, and mouse tissues (Figure 2b). A time-course inhibition of FTO in mESCs further establishes the causal relationship between m⁶A demethylation on LINE1 RNA and chromatin state regulation. Moreover, the FTO-mediated m⁶A demethylation of LINE1 RNA plays important roles in regulating mESC differentiation and proliferation. The dysregulated differentiation and proliferation of mESCs caused by Fto KO can be recapitulated by knocking down of LINE1 RNA, and can be partially rescued in Fto KO cells by targeted delivery of dCAS13b-wtFTO to LINE1 RNA to mediate LINE1-specific m⁶A demethylation. Our further analysis reveals LINE1 RNA as a general substrate of FTO across normal mouse and human tissues, indicating LINE1 RNA demethylation is likely a prevalent process in mammals. Depletion of FTO leads to ovary and oocyte defects with a dramatic reduction of LINE1 RNA abundance in GV oocytes. In GV oocytes, Fto KO also leads to more closed chromatin, suggesting a critical role of this process during oocyte development as an in vivo model.

Reversible m⁶A on carRNAs serves as a switch that can globally tune chromatin state and directly modulate transcription of hundreds to thousands of genes in mESCs. The discovery of quite prevalent reversible carRNA methylation in transcriptional regulation adds another layer of chemical modifications on bio-macromolecules to chromatin regulation in addition to DNA and histone modifications. Reversible carRNA methylation may have broad implications in the development and normal function of mammalian tissues and in human diseases.

Concluding remarks

During a decade of the rapid growth of post-transcriptional gene regulation by reversible RNA modifications (in particular RNA m⁶A), extensive efforts have been made to profile the m⁶A methylome, identify the m⁶A effectors, and investigate biological functions of m⁶A particularly on mRNA. These studies reveal effects on transcription through modulating m⁶A on transcripts that encode regulatory proteins. Moving forward, studies of critical roles of m⁶A on carRNAs provide numerous opportunities for both basic studies of gene expression regulation as well as their potential involvement in human diseases.

There are still many questions that remain to be answered. We currently do not know all nuclear m⁶A effectors, how these effectors interact with transcriptional factors, chromatin remodelers, histone modifiers, and DNA methylation systems, and the pathways in which these effectors and RNA m⁶A methylation integrate into the transcriptional regulation network. There is very likely involvement of methyltransferases other than METTL3 and METTL14. METTL3 and METTL14 may have methylation independent roles as suggested previously [65,66]. They could function as transcription factors and may serve as scaffold proteins to establish certain chromatin conformation or chromatin states. There are clearly other proteins involved in the decay of m⁶A methylated eRNAs and paRNAs besides YTHDC1. While promoting decay of some carRNAs YTHDC1 also plays a role in other nuclear processes dependent on RNA m⁶A methylation. Its post-translational modification status could be a key factor in its biological function and this can be very much cell context-dependent. YTHDC1 is prone to granule formation and is known to form the YTH nuclear body [67]. YTHDC1 in the granule and outside of the granule may perform very different functions.

It is interesting that repeat RNAs are quite heavily methylated. One can speculate that the initial methylation is a way to control activation of retrotransposon RNAs, leading to their decay during evolution. Additional regulatory functions might have evolved later on. It is not known how reversible methylation on repeat RNAs affects development and normal tissue function in general? We also do not know how cellular signaling regulates carRNA methylation and its overall impacts on transcription. We have shown that different cellular contexts would lead to opposite outcomes and we do not currently know what determines the dominant effects and effectors. Lastly, do other RNA modifications also exist on carRNAs or RNA:DNA hybrids and contribute to transcriptional regulation?

The m⁶A methylation of nuclear RNA is involved in R-loop formation and DNA damage repair response. This could present therapeutic opportunity in cancer cells with a defect in the DNA damage response. Could tuning of carRNA methylation synergize with existing therapies that target epigenetic pathways? Lastly, could some of the RNA modification information be inheritable, either through maternal effector proteins or through inheritable chromatin state that was set up by nuclear RNA methylation? These are all intriguing questions in the future.

Highlights.

• m⁶A on chromatin-associated regulatory RNAs (carRNAs) regulates chromatin state and transcription.

• m⁶A on repeats RNA can be demethylated by FTO to affect transcription.

• mRNA m⁶A regulates transcription by controlling transcripts encoding histone modifiers and transcriptional factors.

• Pre-mRNA m⁶A affects transcriptional regulation through divergent co-transcriptional RNA processing.