RNA m⁶A modification, signals for degradation or stabilisation?

N⁶-methyladenosine (m⁶A) was originally identified and partially characterised in the 1970s [1–3]. Since then, it has emerged as one of the most prevalent internal RNA modifications. Notably, m⁶A has been found in a wide range of organisms, spanning mammals, fish, insects, plants, and yeast. Furthermore, this modification is observed in nearly all types of RNA, encompassing mRNA, long noncoding (lnc)RNA, enhancer (e)RNA, promoter upstream transcript (PROMPT), repeat RNA, circular RNA, microRNA, rRNA, snRNA, and viral RNA.

METTL3, the enzyme responsible for catalysing m⁶A deposition, was identified in the mid-1990s [4]. Throughout the 2010s, several other proteins were found to form a complex with METTL3, acting together to facilitate the deposition of m⁶A onto RNA. This complex is now referred to as the m⁶A ‘writer' complex and consists of a heterodimeric enzymatic core, comprising METTL3 and METTL14, along with accessory subunits including WTAP, CBLL1/HAKAI, ZC3H13, VIRMA, and RBM15/15B (Figure 1A) (reviewed in [5]). m⁶A is particularly enriched near the stop-codon and internal long-exon regions of mRNAs in many species [6,7]. Structural studies revealed that within the ‘writer' complex, METTL3 serves as the sole catalytic subunit, whereas METTL14 possesses a degenerate active site, playing an essential role in preserving the structural integrity of the complex and facilitating substrate recognition [8–10]. Although in recent years other m⁶A-modifying enzymes, including METTL16, METTL5, and ZCCHC4, have been identified, the vast majority of m⁶A on mRNA and lncRNAs is METTL3 dependent [11,12] and hence form the primary focus of this review.

In 2011, the enzyme FTO was found to have the ability to demethylate m⁶A, indicating that this modification is dynamic [13]. Subsequently, a second m⁶A demethylase ALKBH5 was identified through homology search and functional characterisation [14]. These two proteins are now referred to as m⁶A ‘erasers' (Figure 1A).

Proteins directly recognising m⁶A are typically identified using synthetic m⁶A-containing RNA as a bait. RNA pull-down and subsequent mass spectrometry analysis are applied to isolate and identify the interacting proteins [6,15,16]. To date, two classes of proteins, collectively referred to as m⁶A ‘readers', have been identified using this method: canonical readers (YTH-domain-containing proteins: YTHDF1/2/3 and YTHDC1/2) (Figure 1B) and non-canonical readers. These non-canonical readers include insulin-like growth factor 2 mRNA-binding proteins (IGF2BPs) (IGF2BP1/2/3) (Figure 1C), along with numerous other RBPs identified in RNA pull-down analysis such as FMR1 [16,17], RBM45 [18], PRRC2A/B [19,20], and TDP-43 [21]. These proteins play crucial roles in conveying the functional impact of m⁶A modifications by interfacing with various downstream molecular pathways.

The deposition of m⁶A onto RNA is widely considered to be co-transcriptional, supported by multiple lines of evidence based on sequencing, imaging, and biochemical data [22–28]. m⁶A continues to play a multifaceted role in nearly all aspects of RNA metabolism throughout its life cycle, including splicing, stability, export, structure, and translation. Indeed, one of the most well-documented functions of m⁶A is its role in facilitating RNA degradation. However, there are many instances where m⁶A is also implicated in maintaining RNA stability, a phenomenon observed in various developmental and disease contexts. This review will specifically focus on the differing roles of m⁶A in the regulation of stability in different types of RNA, with the broader spectrum of m⁶A functions extensively reviewed elsewhere (reviewed in [12,29–32]).

The well-documented function of m⁶A is its role in promoting RNA degradation, which is predominantly facilitated by m⁶A ‘reader’ proteins that bridge with downstream molecular pathways. In the cytosol, the YTHDF family of proteins, particularly YTHDF2, preferentially recognise m⁶A-containing RNAs and subsequently direct them towards degradation pathways [33]. YTHDF2 has been discovered to engage with at least three prominent pathways to facilitate the degradation of m⁶A-containing mRNAs (Figure 2A).

Firstly, through co-immunoprecipitation (co-IP) and mass spectrometry assays, YTHDF2 was found to interact with the CCR4–NOT deadenylation complex, marking the first described molecular pathway linking m⁶A-containing mRNAs to degradation [34]. Secondly, a more novel and recent study uncovered that m⁶A-containing RNAs which possess HRSP12-binding sites in close proximity to RNase P/MRP-directed cleavage sites are preferentially targeted for endoribonucleolytic cleavage through the YTHDF2–HRSP12–RNase P/MRP axis [35]. Notably, this axis is also responsible for the degradation of a subset of m⁶A-containing circular RNAs that lack a poly(A) tail. A recent third body of work demonstrated that YTHDF2 also interfaces with UPF1, an RNA helicase renowned for its involvement in nonsense-mediated decay (NMD). However, the YTHDF2–UPF1 axis is likely to operate by recruiting the RNA decapping enzyme DCP1A [36], rather than via NMD. The latter two mechanisms require further validation across various contexts.

The presence of multiple cellular pathways dedicated to m⁶A-mediated RNA degradation may be attributed to the abundance and diverse RNA species and/or sequence contexts, ensuring their efficient clearance. It is worth noting that recent studies have proposed a redundancy in the function of m⁶A reader proteins YTHDF1, YTHDF2, and YTHDF3. This redundancy is likely attributed to their striking similarities in protein sequence, domain architecture, interacting partners, and the RNAs they bind to [37–39]. Redundant or overlapping functions of YTHDF proteins are supported by observations in vivo (during mouse gametogenesis and postnatal viability) and in vitro (mouse embryonic stem cells (mESCs)) [40]. However, a recent study found dynamic levels of O-GlcNAc modifications appearing on YTHDF1/3 proteins, but not YTHDF2, dependent on background, indicating that they do have diverse functions [41]. Nevertheless, it is important to interpret RNA stability data from triple YTHDF1/2/3 knockdown/knockout experiments with care, as depletion of all three YTHDF proteins results in an increase in cellular P-body formation in HeLa cells, consequently leading to the global stabilisation of most mRNAs which is not strictly dependent on m⁶A [42].

YTHDF proteins have also been found to recognise m⁶A within viral RNA and play regulatory roles during viral life cycles [43–46]. However, the specific impact of YTHDFs on viral RNA stability, akin to their role in host cellular RNAs, remains controversial. For instance, Tsai et al. [43] reported that YTHDF2 binding to m⁶A sites on HIV-1 RNA significantly stabilises these viral RNAs, boosting viral infection, but conflicting studies have suggested inhibitory roles of YTHDFs in HIV-1 infection [44,45]. These findings underscore the need for further studies to elucidate the precise context-dependent mechanisms underlying the functions of each YTHDF protein.

In the nucleus, the YTH-domain-containing protein YTHDC1 stands out as one of the best characterised canonical m⁶A readers [47]. Co-IP experiments with YTHDC1 have found it associated with splicing regulators [27], as well as ZCCHC8 [48,49], a core component of the nuclear exosome targeting complex (NEXT). NEXT plays a critical role in the degradation of non-polyadenylated chromatin-associated regulatory (car) RNAs, including eRNAs and PROMPTs. In turn, ZCCHC8 has been reported to reciprocally interact with YTHDC1 in experiments employing stable isotope labelling in cell culture (SILAC) and mass spectrometry [50]. Furthermore, various studies employing distinct methodologies have identified the core components of NEXT, RBM7, or ZCCHC8, as proteins linked to m⁶A [51]. The YTHDC1–RNA exosome axis has also been implicated in the degradation of many nuclear m⁶A-containing RNAs (Figure 2A), including eRNAs and PROMPTs in mESCs [49], immunoglobulin heavy chain locus-associated lncRNA (SμGLT) in B cells [52], and C9ORF72 repeat RNA in ALS/FTD patient-derived induced pluripotent stem cell (iPSC)-differentiated neurons [53]. On the contrary, Lee et al. [54] found that YTHDC1 depletion did not alter eRNA stability in human cell lines (MCF7, K562, and HeLa). Instead, they found that m⁶A-eRNA recruits YTHDC1 to enhancer regions to stimulate enhancer activity and gene transcription by facilitating transcriptional condensate formation. The discrepancy between these studies may arise from differences in cell types (pluripotent vs differentiated), the technical methodologies (chronic knockout vs acute depletion) used, and the lack of m⁶A stoichiometry measurement on these eRNAs studied.

Another category of m⁶A-interacting proteins is the insulin-like growth factor 2 mRNA-binding proteins (IGF2BPs; including IGF2BP1/2/3). These were initially identified through the use of synthetic single-stranded m⁶A-containing RNA (GGm⁶ACU) as a bait for RNA pull-down, followed by mass spectrometry analysis [15]. Unlike the YTH domain, the K homology (KH) domains of IGF2BPs are responsible for m⁶A recognition (Figure 1C). A recent structural study has further revealed that the hydrophobic groove within KH4 of IGF2BPs contains a dedicated and evolutionarily conserved structural element that conveys m⁶A specificity. While this m⁶A recognition is independent of the underlying RNA sequence context, there is a preference for GGAC, with double mutations of V523I/P524S shown to significantly reduce cognate binding [55]. Interaction between IGF2BPs and m⁶A is observed only at certain m⁶A sites, as revealed by cross-linking and immunoprecipitation followed by high-throughput sequencing (CLIP-seq) analysis [15,55], in contrast with YTH-domain-containing proteins. This suggests that m⁶A does not represent a universal layer of regulation in IGF2BPs target selection. Additionally, it is important to note that the binding affinity of IGF2BPs for m⁶A-containing RNAs is considerably weaker compared with that of YTH domains [47,55]. Moreover, RNA structural data suggest that IGF2BPs may recognise the structural changes induced by the so-called ‘m⁶A-switch [56]'. In this scenario, m⁶A alters local RNA structures, thereby facilitating the binding of RNA-binding proteins such as hnRNPC [57] and hnRNP A2B1 [58]. Further experiments are needed to elucidate this aspect regarding the precise mechanisms through which IGF2BPs execute RNA stabilisation.

In contrast with the destabilising role of the YTHDF–m⁶A axis, IGF2BPs play a role in promoting the stability of their target RNAs in an m⁶A-dependent manner (Figure 2B). This axis was determined by experiments where knockdown/knockout m⁶A writer complex led to reduced stability of m⁶A-containing RNAs, or knockdown/knockout m⁶A eraser proteins resulted in increased stability of certain m⁶A-containing RNAs, with further biochemical experiments functionally confirmed the stabilising role of IGF2BPs–m⁶A across multiple cellular, physiological, and pathological contexts, including human cancer cells (hepatocellular carcinoma: HepG2, cervical cancer: HeLa, prostate cancer: 22Rv1, acute myeloid leukaemia) [15,59], as well as mouse oocytes, during early embryonic development [60], and postnatal liver development [61].

Interestingly, while m⁶A plays a stabilising role for zygotic degraded m⁶A-containing maternal transcripts before fertilisation, these transcripts with m⁶A are degraded more quickly after zygotic genome activation than those without m⁶A. These opposing effects of m⁶A on maternal transcripts during the maternal-to-zygotic transition suggest the involvement of distinct readers or pathways [60,62–64].

YTHDC1 has also been implicated in stabilising m⁶A-containing RNAs. For instance, in acute myeloid leukaemia, YTHDC1 binding of nuclear m⁶A transcripts enables the formation of nuclear condensates that protect nuclear mRNAs such as MYC from poly(A) tail exosome targeting complex (PAXT)-mediated degradation (Figure 2B) [65]. Additionally, during mouse preimplantation development, YTHDC1 binds to a small group of m⁶A-containing genes, sustaining their RNA stability [66]. However, none of these studies demonstrated the effect of YTHDC1 on RNA stability regulation depends on the quantity of m⁶A sites or m⁶A levels present on these RNAs. Furthermore, these studies raised another question regarding the extent to which YTHDC1 binds m⁶A-containing transcripts for either enhanced stability or destruction in different cellular contexts. Specifically, for nuclear m⁶A-bearing transcripts, how does YTHDC1 distinguish between those targeted for degradation and those marked for protection?

In the parasitic kinetoplastid Trypanosoma brucei, remarkably approximately half of the m⁶A modifications in VSG RNA is situated within its poly(A) tail. These m⁶A modifications are then systematically eliminated from the VSG poly(A) tail prior to RNA deadenylation and degradation [67]. Intriguingly, a 16-mer cis-acting motif positioned in the 3′-UTR of VSG plays a crucial role in the deposition of the poly(A) tail m⁶A, with excision of the motif resulting in lack of m⁶A, and the subsequent acceleration in the rate of deadenylation and degradation [67]. Given that trypanosomes lack an orthologue of METTL3 and m⁶A is found in the poly(A) tail instead of the canonical DRACH motif [67], this is highly suggestive of the presence of an alternative m⁶A writing mechanism and possibly different reading mechanisms within this species.

The proline-rich coiled-coil 2 (PRRC2) family protein PRRC2A has been identified as a neuronal cell-specific m⁶A reader protein exhibiting a preference for m⁶A [19]. PRRC2A stabilises Olig2 mRNAs by binding to the m⁶A within the coding sequence region of Olig2 and this stabilisation effect is reversible through the action of the m⁶A demethylase FTO. However, another recent study showed that PRRC2A recognises spermatogonia-specific transcripts and down-regulate their RNA abundance [68], thereby maintaining the spermatocyte expression pattern during the meiosis prophase. It is important to note that this analysis was conducted on a transcriptome-wide scale rather than focusing on individual transcripts. The divergent roles observed for PRRC2A in different tissues underscore the tissue-specific nature of its function. Recently, another PRRC2 family protein PRRC2B has also been identified as an m⁶A reader and stabilises Sox2 mRNA in an m⁶A-dependent manner [20].

In addition, work in Drosophila has identified FMR1 as an m⁶A reader through a short m⁶A-modified RNA oligo-based approach [69]. Beyond its documented role in translational regulation, FMR1 is implicated in maternal RNA decay during Drosophila embryogenesis [69], mediated at least in part by m⁶A. In the mouse cerebral cortex, FMR1 regulates the stability of its m⁶A-marked mRNA targets through binding to YTHDF2 [70].

However, further experiments are warranted to fully characterise the precise mechanisms by which these non-canonical m⁶A readers recognise m⁶A (direct recognition, m⁶A-induced RNA structural alteration, or interaction with other m⁶A readers [19,71]) and execute their functional actions in RNA stability.

m⁶A appears on RNA from LINE1 (L1) elements, which are also transcribed by RNA polymerase II. The first study that identified m⁶A on L1 RNAs suggested a model wherein m⁶A promotes the degradation of L1 RNAs, exemplified by L1Md_F, through the YTHDC1–ZCCHC8/NEXT axis in mESCs [49]. A similar mechanism has also been observed in relation to other m⁶A-modified carRNAs in mESCs, but also worth noting that in this study Mettl3 knockout mESC lines used were demonstrated to be hypomorphic, due to alternative splicing events bypassing the CRISPR-mediated indels [11]. Unlike PROMPTs and eRNAs, full-length L1 RNAs have a poly(A) tail. It remains in question how ZCCHC8/NEXT, which primarily targets non-polyadenylated nuclear RNA, is responsible for degrading L1 RNAs [49,72].

In contrast with this finding, a study conducted in K562 cells, a human immortalised myelogenous leukaemia cell line, observed a preference for m⁶A on intronic L1 RNAs. These intronic L1 RNAs are evolutionarily young and oriented towards the hosting gene. By comparing the steady-state L1 RNA levels with nascent L1 RNAs, this study suggested that m⁶A positively influences the expression of both autonomous L1 RNAs and co-transcribed L1 relics, and promotes L1 retrotransposition activity [73]. Moreover, these m⁶A-marked intronic LINE1 elements appear to act as ‘roadblocks' to RNA polymerase II within their host-long genes. In line with this positive correlation, another study utilising mESCs found that the majority of L1 RNAs are down-regulated when one of the components of m⁶A writer complex is constitutively knocked out (i.e. Mettl3, Mettl14, Wtap, Zc3h13). This led to the conclusion that m⁶A regulation on L1 RNAs operates through a post-transcriptional mechanism i.e. a stability-based control, as no differences were observed in histone modifications (H3K4me3, H3K27ac, and H3K9me3) at LINE1 loci [74].

As opposed to post-transcriptional regulation [75], another recent study reported that m⁶A can transcriptionally suppress LINE1 transcription in mESCs. They found that m⁶A transcriptionally represses retrotransposons via the YTHDC1–SETDB1–H3K9me3 pathway in mESCs with the observation that steady-state L1 RNAs are up-regulated upon conditional knockout (cKO) of Ythdc1 or Setdb1, the enzyme that deposits H3K9me3 over LINE1 loci to repress their transcription. Ythdc1 cKO induces the 2-cell (2C)-like transcriptional programme, largely through indirect regulation of Dux, Zscan4, and MERVL. However, this study raised another conflict: how does the loss of METTL3 or YTHDC1 could lead to both L1 RNAs and 2C marker gene Dux up-regulation? Specifically, as Percharde et al. [76] reported that L1 RNAs repress Dux transcription by recruiting Nucleolin/KAP1, thus resulting in a reciprocal expression relative to Dux. These conflicting results may stem from variations in the methods used to generate knockouts or potential secondary effects resulting from chronic knockouts, or the sensitivities of the technology to measure RNA half-life/stability. To provide clarity on the role of m⁶A in L1 RNA stability, employing an experimental system with acute depletion and/or complementation assays, together with time-resolved sensitive measurement, will be instrumental [22].

In addition to transcriptional and post-transcriptional roles of m⁶A on L1 RNA, another study proposed that m⁶A on L1 RNAs recognised by YTHDC1 regulates the formation of L1 RNA–Nucleolin in the nucleus, thus facilitating KAP1 recruitment on the 2C-related transcriptional programme in early mouse embryonic development [77]. Hwang et al. [78] using a reporter assay found that the m⁶A cluster in the L1 5′-UTR region serves as a docking site for eIF3 to enhance translational efficiency and promote the formation of L1 RNPs. Indeed, further investigations are required to elucidate how m⁶A alters the structure or scaffold of L1 RNAs and whether this structural change contributes to RNA stability. It remains an open question whether different m⁶A sites or clusters on L1 RNAs exert distinct roles, contributing to the observed functional diversity. This area of study holds great potential for advancing our understanding of m⁶A-mediated regulation in the context of LINE1 elements and early mammalian development.

This review considers the current understanding in the field of how RNA modification m⁶A impacts on the regulation of RNA stability (Table 1). This layer of gene regulation is primarily orchestrated by m⁶A reader proteins along with their interacting partners. The reader-based recognition of m⁶A and its resultant impact on stability is influenced by developmental stages, cellular environments, and the RNA sequence/structure itself. The dual functions of m⁶A in RNA stability regulation underscore a fundamental question: are there any factors determining whether m⁶A recognition by different readers leads to RNA degradation versus stabilisation? Whilst it is established that YTHDFs and IGF2BPs contribute to m⁶A-dependent RNA degradation and stabilisation respectively, the precise features that dictate recognition by distinct readers of m⁶A sites or m⁶A-containing RNAs remain elusive. Exploring the phenomenon where distinct readers can bind to the same m⁶A moiety on RNAs raises interesting questions. It would be worthwhile to delve into the potential implications of binding competition as a mechanism for RNA stabilisation. This competition could function by preventing YTHDFs binding, offering a unique perspective on the complex interplay of different readers in regulating RNA stability. Furthermore, the relationship between the kinetics of stability and m⁶A stoichiometry remains to be clearly delineated. Considering the intricate regulation of RNA stability by various factors like microRNAs, RNA-binding proteins, poly(A) tails, and RNA structure, exploring how RNA modifications interact with these elements to govern stability demands deeper investigation.

Further exploration is also warranted to delve into the causal relationship between m⁶A and developmental or disease-related phenotypes, considering the broad impact of m⁶A on RNA metabolism that often results in substantial secondary effects when m⁶A pathways are chronically disrupted. Caution must be taken during the generation of Mettl3 knockout, as alternative splicing events may bypass these deletions, potentially yielding a truncated METTL3 protein that retains catalytic activity [11]. These occurrences could underlie several conflicting findings. Acute depletion systems offer a solution by largely eliminating those unwanted secondary effects, thereby providing a clearer delineation of direct causality [22,74]. Additionally, the targeted editing of individual m⁶A modifications using RNA-guided RNA-targeting CRISPR systems [79–81], such as CRISPR–Cas13, presents another avenue to investigate the specific functions of individual m⁶A modifications without disturbing the entire m⁶A writing or erasing systems.

• As one of the most prevalent internal RNA modifications, m⁶A influences various aspects of RNA metabolism, thereby regulating gene expression crucial to development and tissue functions. m⁶A dysregulation leads to diseases, including cancer.

• The modulation of RNA stability by m⁶A is largely driven by reader proteins that bridge to downstream molecular pathways, further influenced by RNA sequence context, developmental stage, and cellular environment.

• The intricate interplay between m⁶A and various RNA regulatory elements, including microRNAs, RNA-binding proteins, and RNA structure, forms a complex network that significantly influences RNA stability. A comprehensive understanding of these relationships is pivotal to fully elucidate the mechanisms governing RNA stability regulation.

The author declares that there are no competing interests associated with this manuscript.

Open access for this article was enabled by the participation of University of Oxford in an all-inclusive Read & Publish agreement with Portland Press and the Biochemical Society under a transformative agreement with JISC.

I thank Heather Coker and Neil Brockdorff for critical reading of this manuscript. This work is supported by a pump-priming award from the University of Oxford Medical Sciences Division.

cKO

conditional knockout

co-IP

co-immunoprecipitation

KH

K homology

m⁶A

N⁶-methyladenosine

mESCs

mouse embryonic stem cells

PROMPT

promoter upstream transcript

PRRC2

proline-rich coiled-coil 2; nuclear exosome targeting complex (NEXT)