New directions for Ψ and m¹A decoding in mRNA: deciphering the stoichiometry and function

Abstract

Over the past decade, advancements in epitranscriptomics have significantly enhanced our understanding of mRNA metabolism and its role in human development and diseases. This period has witnessed breakthroughs in sequencing technologies and the identification of key proteins involved in RNA modification processes. Alongside the well-studied m⁶A, Ψ and m¹A have emerged as key epitranscriptomic markers. Initially identified through transcriptome-wide profiling, these modifications are now recognized for their broad impact on RNA metabolism and gene expression. In this Perspective, we focus on the detections and functions of Ψ and m¹A modifications in mRNA and discuss previous discrepancies and future challenges. We summarize recent advances and highlight the latest sequencing technologies for stoichiometric detection and their mechanistic investigations for functional unveiling in mRNA as the new research directions.

A BRIEF INTRODUCTION TO EPITRANSCRIPTOMICS

Epitranscriptomics, also known as “RNA epigenetics,” refers to posttranscriptional regulation of gene expression through chemical modifications that affect RNA functions (Frye et al. 2016). Over 170 chemical modifications that exist on diverse types of RNA have been identified (Boccaletto et al. 2022). Previous studies on RNA modifications have focused mainly on noncoding RNAs, including tRNA, rRNA, snRNA, and snoRNA. Subsequent studies unveiled various forms of modifications present in mRNA. To date, several types of mRNA modifications, including N⁶-methyladenosine (m⁶A), inosine (I), pseudouridine (Ψ), N¹-methyladenosine (m¹A), N⁶,2′-O-dimethyladenosine (m⁶Am), 5-methylcytidine (m⁵C), 5-hydroxyl-methylcytidine (hm⁵C), N⁷-methylguanosine (m⁷G), and N⁴-acetylcytidine (ac⁴C) have been identified. In the past 10 years, advancements in high-throughput sequencing technologies have significantly enhanced the profiling and identification of mRNA modifications (Li et al. 2017a). The precision, sensitivity, and quantitative capabilities of these methods continue to evolve. mRNA modifications, now recognized as key regulators of mRNA fate and function, impact stability, structure, localization, translation, and splicing (Zhao et al. 2017). While the functions of m⁶A in gene regulation have been intensely investigated, other epitranscriptomic marks, like Ψ and m¹A, have also made great progress in recent years (Sun et al. 2023a). The rapidly growing toolbox for tracking Ψ and m¹A modifications and their functions has led to discoveries about their roles in mRNA metabolism. We will review these recent developments, address the challenges in sequencing and functional studies, and suggest directions for future research.

A BRIEF INTRODUCTION TO PSEUDOURIDINE

Ψ was first discovered and also most abundant among posttranscriptional modifications in RNA, which is also referred to as the “fifth ribonucleotide” (Cohn and Volkin 1951; Li et al. 2016a). Ψ is the C5-glycoside isomer of uridine that results in the C5–C1′ bond between uracil and ribose rather than the usual N1–C1′ bond. It has an extra hydrogen-bond donor at its non-Watson–Crick edge, and the distinct chemical properties have been thought to contribute to base-pairing stability (Charette and Gray 2000). Initial intense studies have well documented the clustered distribution and regulatory roles in noncoding RNA (including rRNA, tRNA, snRNA, etc.) (Borchardt et al. 2020; Cerneckis et al. 2022). Mass spectrometry data of purified mammalian mRNA reveal Ψ is the second most abundant, with a Ψ/U ratio of ∼0.2% to 0.6% in human cells and mouse tissues (Li et al. 2015). Facilitated by the great achievement of transcriptome-wide sequencing technologies, the past decades have seen remarkable progress in the location, stoichiometries, and functions of Ψ in mRNA. Since no antibodies can selectively recognize and enrich Ψ modification, chemical-assisted approaches, from carbodiimide labeling to the more recent quantitative bisulfite-assisted labeling, have been integrated with next-generation sequencing methods (Zhang et al. 2023b). These advancements have heralded a new era in detecting Ψ modifications, enabling transcriptome-wide, base-resolution, and stoichiometric-level analysis.

NEW DIRECTIONS FOR Ψ: STOICHIOMETRIC DETECTION IN mRNA

Initially reported Ψ landscapes utilized Ψ reactivity with carbodiimide. The selective chemical labeling of Ψ by N-cyclohexyl-N0-(2-morpholinoethyl)-carbodiimide metho-p-toluenesulfonate (CMC) was adopted for transcriptome-wide Ψ detection (Ho and Gilham 1971). CMC can specifically and irreversibly bind at the N3 position of Ψ, while CMC adducts to G and U are susceptible to alkaline hydrolysis. Ψ-CMC adducts cause reverse transcription (RT) termination of one base downstream from Ψ, leaving a stop signature in the cDNA product, thus enabling Ψ detection at single-base resolution. Three high-throughput sequencing methods, namely Pseudo-seq, Ψ-seq, and PSI-seq, utilized CMC chemistry and identified dozens to hundreds of modified sites for the first time in yeast and human cells (Carlile et al. 2014; Lovejoy et al. 2014; Schwartz et al. 2014). CeU-seq pre-enriches Ψ-containing fragments using an azide-modified CMC derivative, allowing the detection of thousands of Ψs (Li et al. 2015). These previous sequencing approaches allow the global detection of Ψ modifications transcriptome-wide and open the door to biological functions of Ψ in mRNA. However, the low overlap of identified Ψs in mRNA by different methods showed the limitation of CMC labeling and RT stop signatures (Zaringhalam and Papavasiliou 2016). The sources of low reproducibility were speculated to be false positives resulting from RNA secondary structures or false negatives due to unequal CMC derivatizing to each Ψ and alkaline-induced RNA degradation. Another major challenge of CMC sequencing is that both inefficient labeling and pre-enrichment could lead to missing Ψ stoichiometries. Quantifying stoichiometries is important as it provides the percentage of modified transcripts, indicating the potential biological significance of modification sites.

One recent achievement in the Ψ field was developing quantitative sequencing methods, namely PRAISE (Zhang et al. 2023a) and BID-seq (Dai et al. 2023). These methods utilized bisulfite labeling reactions and Ψ-bisulfite adducts, which induced deletion signatures during RT. By increasing the sulfite molar ratio and pH compared to traditional RBS-seq conditions (Khoddami et al. 2019), they achieved selective and quantitative reactivity with Ψ, with minimal reactivity to cytidine. Both methods have demonstrated their quantitative capacity using synthetic spike-in RNA and rRNA. These studies have expanded the list of Ψ sites in mRNA with stoichiometries. For example, PRAISE identified 2209 Ψ sites in nuclear-encoded mRNA and ncRNA, as well as four sites in mitochondrial-encoded mRNA in HEK293T cells. The median modification level of global mRNA Ψ sites is ∼10%, with over 400 sites being moderately or highly modified (Ψ level >20%). BID-seq reported abundant Ψ sites with modification stoichiometry information in three human cell lines and 12 different mouse tissues using only 10–20 ng poly(A)⁺ RNA. The results of PRAISE and BID-seq are very consistent with each other, with a 53% overlap in transcriptome-wide mRNA Ψ sites in HEK293T cells. Additionally, they demonstrated high reproducibility of Ψ modification sites and stoichiometry across biological replicates. Their robustness extends to various library constructions, enhancing their suitability for widespread use in quantitative Ψ detection. These methods affirm their reliability and potential for broader applications in Ψ research. However, challenges remain: PRAISE and BID-seq struggle to precisely define Ψ sites within constitutive “U” contexts, and Ψ calling in low-abundance transcripts may depend more on sequencing depth. Validation of these modification sites might require targeted sequencing. Nevertheless, these studies offer novel methodologies for investigating the biological relevance and regulatory mechanisms of variable Ψ stoichiometries.

Recently, nanopore-based direct RNA sequencing has emerged as an alternative technology, which has the potential to detect and quantify multiple RNA modifications in native RNA molecules. As an RNA strand passes through the pore, the ion current changes for different k-mer sequences within the pore (k = 5 nt), allowing distinction between modified and unmodified nucleosides based on ionic current, dwell time, and base-calling errors (Pratanwanich et al. 2021). Machine learning algorithms have been developed to infer modification sites and fractions. Utilizing this approach, Ψ in Saccharomyces cerevisiae rRNA, mRNA, and viral RNAs have been quantified (Begik et al. 2021; Fleming et al. 2021). A similar detection approach has also been used to identify interferon-inducible Ψ modifications in human mRNA (Huang et al. 2021). A recent method using nanopore sequencing detects Ψ modifications through systematic U-to-C base-calling errors, revealing both known and novel sites in a HeLa transcriptome (Tavakoli et al. 2023). Despite these advances, nanopore sequencing faces limitations in accurately and quantitatively detecting Ψ, such as a low signal-to-noise ratio, especially in clustered Ψ sites like rRNA, which can lead to false positives. Additionally, base-calling errors are highly sequence-context dependent. Therefore, synthetic controls, including negative controls without RNA modifications and positive controls with all possible 5-mers containing modified nucleotides, should be constructed. Considering these challenges, other nanopore sequencing strategies used chemical tools to selectively modify RNA modifications and alter the nanopore signatures, like bisulfite-assisted nanopore sequencing (Fleming et al. 2023), for reliable base-calling error profiles.

Besides the high-throughput sequencing technologies, researchers can also detect or validate the interest of Ψ sites using site-specific detection methods. For instance, Ψ can be detected and quantitated utilizing a combination of RNase digestion, radiolabeling, and 2D thin-layer chromatography (Gupta et al. 1979). A relevant method was developed, termed Site-specific Cleavage And Radioactive-labeling followed by Ligation-assisted Extraction and Thin-layer chromatography (SCARLET), which has been used to validate and quantify putative Ψ sites (Liu et al. 2013). For targeted Ψ sites, the CMC-Ψ adduct can be also detected by coupling with qPCR analysis or ligation-assisted PCR analysis at locus-specific Ψ sites of interest (Lei and Yi 2017; Zhang et al. 2019). But note that these methods were best suited to highly abundant transcripts and required knowledge of the specific Ψ-containing sequences. In principle, bisulfite-based methods facilitate the readthrough of Ψ adducts during RT, making it readily applicable for site-specific detection.

NEW DIRECTIONS FOR Ψ: MOLECULAR FUNCTIONS AND THERAPEUTIC APPLICATIONS

Site-specific pseudouridylation is catalyzed by either stand-alone or H/ACA box snoRNA-guided pseudouridine synthases (PUS) (Borchardt et al. 2020). In humans, 13 PUS proteins have been found and classified into five families: TruA, TruB, TruD, RluA, and PUS10. Mutation or aberrant expression of the PUSs has also been implicated in human genetic diseases and cancers (Heiss et al. 1998; Fernandez-Vizarra et al. 2007; Shaheen et al. 2016, 2019). Combining writer perturbations (knockout or knockdown) with developed Ψ profiling methods has identified several PUSs that modify Ψs in mRNA, including TRUB1, PUS1, PUS7, PUSL1, PUS3, PUS7L, TRUB2, and DKC1 (Fig. 1A; Dai et al. 2023; Zhang et al. 2023a). Notably, TRUB1 mRNA targets possess the highest modification stoichiometries with the GUUCNANNC sequence motif, whereas PUS7 targets harbor the UGUAR core motif. Other enzymes lack a strong sequence or secondary structure bias for mRNA Ψ recognition. Previous studies have attributed most mRNA targets to specific PUS enzymes without observed redundancy in Ψ deposition in vivo. While similarities in Ψ modifications between mRNA and tRNA might suggest potential off-target modification sites, a series of functional studies have demonstrated the distinct regulatory roles of mRNA targets. These included influences on pre-mRNA splicing, translation, and stability, affirming the biological significance of depositing pseudouridine on target mRNAs.

FIGURE 1.

Molecular functions and therapeutic applications of Ψ modification in mRNAs. (A) mRNA Ψ modifications are catalyzed by various PUSs, including PUS1, PUS1L, PUS3, PUS7, PUS7L, TRUB1, TRUB2, RPUSD4, and DKC1. (B) Ψ impacts multiple facets of mRNA processing and metabolism, such as the local mRNA secondary structure, translation, pre-mRNA splicing, and mRNA stability. Ψ modifications in UAG, UGA, and UAA stop codons lead to stop codon readthrough. Ψ in pre-mRNA can affect the splicing machinery, thereby regulating alternative splicing. Ψs in human mRNAs have been shown to stabilize mRNAs. (C) snoRNA-guided pseudouridylation is emerging as a promising RNA-editing technique for mediating premature termination codon (PTC) readthrough in contexts relevant to diseases. The incorporation of m¹Ψ into mRNA vaccines is utilized to enhance their effectiveness as a therapeutic tool.

Pseudouridylation can directly or indirectly alter RNA–protein interactions by impacting RNA structure (Fig. 1B). The extra hydrogen bond donor in Ψ can improve water bridge formation, thereby rigidifying the RNA backbone (Davis 1995). While specific mRNA pseudouridine readers are yet to be identified, several RNA-binding proteins (RBPs) have been shown to bind RNA with higher or lower affinity due to Ψ modifications. Artificial introduction of Ψ into the UNUAR motif within mRNA significantly reduced PUM2 binding affinity, approximately threefold (Vaidyanathan et al. 2017). Similarly, the substitution of pseudouridine within CUG repeats or in a polypyrimidine tract has a notable impact on splicing factors: it greatly reduced MBNL1 binding by ∼4 to 20 times and inhibited U2AF2 binding, respectively, in in vitro experiments (Chen et al. 2010; deLorimier et al. 2017). However, the effect of Ψ on RNA/protein interactions is context dependent. To fully understand its effect on endogenous mRNA, it is essential to measure these stoichiometric changes. The integration of advanced, newly developed quantitative Ψ detection techniques will provide additional information to uncover the complexity of the posttranscriptional regulation of gene expression.

One of the well-established molecular functions of Ψ modifications in mRNA is to affect translation fidelity (Fig. 1B). Early work showed that pseudouridylation of stop codons (ΨAA, ΨAG, ΨGA) promoted stop codons readthrough in yeast cells (Karijolich and Yu 2011). Interestingly, Ψs also exist in endogenous stop codons of mammalian mRNA, and they could promote readthrough at modified stop codons in vivo (Dai et al. 2023). Recent studies have demonstrated that targeted pseudouridylation of PTCs enabled efficient PTC readthrough in endogenous disease-relevant genes, and restored full-length functional protein production (Fig. 1C; Adachi et al. 2023; Song et al. 2023). To date, the mechanism behind Ψ-mediated stop codon readthrough remained elusive. However, these studies shed light on a promising approach involving mRNA pseudouridylation using artificial snoRNA, with significant potential for therapeutic applications in treating genetic diseases. Another crucial application is incorporating Ψ derivates (m¹Ψ) into mRNA vaccines to accelerate mRNA vaccines and therapy development (Fig. 1C; Nance and Meier 2021). The incorporation of m¹Ψ into mRNA vaccines enhances translation by diminishing PKR activation (Anderson et al. 2010). However, a recent study reported that m¹Ψ could affect mRNA translation fidelity. They found that m¹Ψ-modified mRNA translation can lead to +1 ribosomal frameshifting in vitro and in cultured cells (Mulroney et al. 2024). These findings are important for designing mRNA-based therapeutics, for instance, through synonymous targeting to avoid mistranslation events.

Recent work has demonstrated that Ψ can be cotranscriptionally installed in human cells, with thousands of Ψ sites identified in pre-mRNA using Pseudo-seq (Martinez et al. 2022). PUS1, PUS7, and RPUSD4 have been identified as pre-mRNA-modifying enzymes. The majority of Ψ sites were found in intronic regions, notably enriched at splice sites, splicing regulatory elements, and regulatory RBP binding sites. In vitro and in vivo splicing assays further highlighted the role of Ψ in regulating pre-mRNA splicing (Fig. 1B). However, it remains unclear whether Ψ’s effect on splicing is mediated by modifying RBP binding sites or influencing pre-mRNA secondary structure.

Intriguing evidence suggested that Ψ might stabilize mRNA transcripts (Fig. 1B). In yeast, upon heat shock, mRNA targets mediated by Pus7p showed decreased expression levels in a Pus7p deletion strain, indicating a stabilizing role for Ψ in mRNA (Schwartz et al. 2014). This, however, does not rule out the possibility of an indirect result of changes in mRNA splicing. Further research revealed that TRUB1-targeted mRNAs exhibited shorter half-lives upon TRUB1 knockdown (Dai et al. 2023). The site-specific Ψ deposition by a fused dCas13d-TRUB1 system significantly prolonged mRNA lifetime, suggesting that TRUB1-installed Ψ stabilized the target mRNA. Nonetheless, no effects on RNA stability were observed through PUS7 and TRUB1 knockout (Zhang et al. 2023a). The discrepancy may stem from differential biological contexts or experimental approaches used in the studies. Additionally, in certain contexts, such as in Toxoplasma gondii, pseudouridine synthase TgPUS1 modestly destabilizes mRNAs, indicating diverse roles of Ψ in mRNA stability (Nakamoto et al. 2017).

While previous research on Ψ modification mainly focused on yeast and human models, recent findings have provided strong evidence for its critical role across various biological contexts, including viruses and plants. Pseudouridylation of the Epstein–Barr virus is required for efficient viral lytic replication (Henry et al. 2022). Ψ modifications are also present in HIV-1 transcripts and subgenomic RNAs of the SARS-COV-2 virus, although their regulatory functions are yet to be elucidated (Fleming et al. 2021; Martinez Campos et al. 2021). Studies have suggested significant roles for Ψ in plant development and stress responses, with Arabidopsis and maize having 20 and 22 PUSs, respectively (Niu and Liu 2023). Ψ content in Arabidopsis mRNA increased under environmental stimuli (Sun et al. 2019), indicating a broad presence and functional impact of Ψ in diverse cellular contexts.

A BRIEF INTRODUCTION TO N¹-METHYLADENOSINE

N¹-methyladenosine (m¹A) is a prevalent and reversible posttranscriptional RNA modification, that was first found in 1961 (Dunn 1961). m¹A modification hosts an additional methyl group at the N1 position of adenosine, which is located at the Watson–Crick base-pairing interface and interferes with base-pairing, changing the Watson–Crick to Hoogsteen base-pairing (Zhou et al. 2016). In addition, the extra methyl group makes the m¹A with a positive electrostatic charge, indicating that it could make a significant contribution to tRNA structure stability through an electro-chemical interaction (Anderson and Droogmans 2005). Mass spectrometry data of purified mammalian mRNA reveal that m¹A is relatively low, with a m¹A/A ratio ranging from 0.015% to 0.054% in cell lines and up to 0.16% in tissues (Dominissini et al. 2016; Li et al. 2016b). m¹A is a prevalent modification in tRNA, rRNA, and also deposits in mRNA (Fig. 2). In mammals, m¹A modification is found at positions 9,14,58 of tRNA, and m¹A58 is a conversed modification across the three domains of life, and m¹A58 is globally present in cytosolic tRNAs. In mRNA, m¹A sites are enriched in the 5′-UTR region of nuclear-encoded mRNAs, and also prevalent in the mitochondrial-encoded transcripts; 10/13 mt-mRNAs are m¹A modified, with the m¹A sites mainly located in the coding regions (CDS) of mt-mRNAs (Dominissini et al. 2016; Li et al. 2016b, 2017b; Safra et al. 2017).

FIGURE 2.

Molecular functions of m¹A modification in tRNA and mRNAs. (A) m¹A in tRNAs is modified by the TRMT6/TRMT61A complex and removed by ALKBH1/ALKBH3/FTO. The presence of m¹A58 in tRNA promotes translation. (B) The abundance of tRNA, tRNA-m¹A58, and tRNA-related effectors dynamically respond to various physiological and pathological processes, such as T cell activation. (C) tRNA-m¹A mediates the generation of tRFs (tRNA-derived fragments). (D) m¹A in the 5' UTR of nuclear-encoded mRNA enhances translation, while m¹A in the CDS region of mitochondrial-encoded mRNA inhibits translation. (E) YTH domain-containing proteins may also recognize m¹A in mRNA in vitro, and YTHDF2 mediates the decay of m¹A-containing mRNAs; however, further in vivo experiments are necessary to elucidate the direct or indirect interactions. (F) TDP-43 is a reader of m¹A. The presence of m¹A induces TDP-43 accumulation in stress granules (SGs) and disrupts the phase separation of TDP-43.

NEW DIRECTIONS FOR m¹A: NEW EFFECTORS FOR m1A

The effectors in m¹A pathways include “writers” and “erasers” that respectively install and remove the methylation, and “readers” that recognize it. The TRMT6–TRMT61A complex is responsible for m¹A methylation in nuclear-encoded tRNA and a subset of m¹A in mRNA located within a tRNA T-loop structure and a GUUCNA consensus sequence in humans (Ozanick et al. 2005; Li et al. 2017b). In this complex, TRMT61A serves as the catalytic subunit, while TRMT6 plays a role in substrate selection (Fig. 2A). For mitochondrial RNAs, different from the nuclear-encoded RNAs, TRMT61B and TRMT10C are responsible for catalyzing m¹A modifications (Chujo and Suzuki 2012; Vilardo et al. 2012; Li et al. 2016b; Safra et al. 2017). In addition, the m¹A1322 in 28S rRNA is catalyzed by the NML (Waku et al. 2016). Apart from the nature of methyltransferase protein, an artificially evolved ribozyme (named MTR1) could catalyze site-specific m¹A in the designed target RNA (Scheitl et al. 2020). Those natural or artificial m¹A writers improve the fundamental aspects of m¹A methylation. As a reversible RNA modification, m¹A is removed by the AlkB family proteins, ALKBH1, ALKBH3, and FTO (Li et al. 2016b; Liu et al. 2016a; Wei et al. 2018). The ALKBH1-mediated and FTO-mediated reversible tRNA-m¹A methylation dynamically regulates translation (Liu et al. 2016a; Wei et al. 2018), and ALKBH3 is responsible for m¹A demethylation on mRNA and tRNA (Fig. 2A; Li et al. 2016b). The reader of RNA modification enhances the understanding of its functional intricacies by providing valuable insights into diverse molecular mechanisms; m⁶A has diverse regulatory or functional machinery through its abundant readers (Shi et al. 2019). Similarly, m¹A readers have been explored. In vitro data indicate that the m⁶A reader proteins with YTH domain-containing proteins, YTHDF1–3 and YTHDC1, can also recognize and interact with m¹A in mRNA, providing m⁶A reader-like function, such as YTHDF2-mediated decay in the m¹A-containing mRNAs (Fig. 2E; Dai et al. 2018; Boo and Kim 2020). Because of physicochemical property differences in m⁶A and m¹A, further validation of this interaction in cellular contexts remains imperative to consolidate its significance. Recently, TDP-43 has been identified as a novel reader of m¹A in mRNA, exhibiting a particularly strong affinity for m¹A-containing CAG repeat RNA. Dysregulated m¹A modification levels in CAG repeat RNA induce the sequestration of TDP-43 into SGs, consequently disrupting the phase separation of TDP-43. This leads to cytoplasmic redistribution and aggregation of TDP-43, ultimately contributing to the development of neurological diseases (Fig. 2F; Sun et al. 2023b).

NEW DIRECTIONS FOR m¹A: DETECTION OF m¹A IN mRNA

To comprehensively explore m¹A methylation in the transcriptome, several high-throughput sequencing methods for m¹A have been developed. Due to the low abundance of m¹A in mRNA molecules, highly sensitive detection techniques are required. The unique characteristics of m¹A have been leveraged in the development of m¹A detection methodologies. For instance, the presence of the m¹A methyl group disrupts Watson–Crick base-pairing, leading to misincorporated or truncated RT products. Additionally, m¹A can undergo rearrangement to m⁶A under alkaline conditions through the Dimroth rearrangement (Dominissini et al. 2016). Taking inspiration from MeRIP-seq (methylated RNA immunoprecipitation sequencing) methods used for m⁶A mapping (Dominissini et al. 2012), two groups independently developed the first-generation m¹A sequencing methods, which are dependent on the m¹A-antibody (Dominissini et al. 2016; Li et al. 2016b). These approaches used anti-m¹A antibodies to enrich m¹A-containing fragments and take advantage of the characteristics of m¹A to induce mismatch or truncation during the RT process. The specificity of m¹A antibodies and SNP sites also needs to be taken into consideration, to eliminate false signals, both groups used control groups. Dominissini et al. (2016) used the Dimroth rearrangement to convert m¹A to m⁶A, enabling the analysis of m¹A sites based on mismatch rates; on the other hand, Li et al. (2016b) utilized Escherichia coli AlkB to demethylate m¹A to its unmodified A, allowing the identification of high-confidence m¹A sites based on the demethylase-sensitive regions. First-generation m¹A profiling technologies have a resolution of about tens of nucleotides to several hundred nucleotides. To identify m¹A sites more precisely for a more accurate picture of m¹A distribution, two groups developed the second-generation m¹A sequencing method to profile m¹A at a single-base resolution. Taking advantage of the ability of m¹A to disrupt Watson–Crick base-pairing, a comprehensive screen of different reverse transcriptases was carried out, and TGIRT (thermostable group II intron reverse transcriptase) demonstrated excellent readthrough efficiency and high mutation frequency at m¹A sites. By comparing with AlkB-treated control samples or Dimroth rearrangement samples, the precise position of m¹A can be determined based on the higher signal-to-noise ratio observed at the mutation sites (Li et al. 2017b; Safra et al. 2017). Two groups independently identified high confidence m¹A sites and reported some similar findings, such as the presence of a subset of m¹A in mRNA sites exhibiting a GUUCRA tRNA-like motif. However, diverging opinions arise regarding the abundance of m¹A sites in human mRNAs. Safra et al. (2017) propose that m¹A is a rare modification in human mRNAs, as they identified only a limited number of m¹A sites using their method, totaling around a dozen. In contrast, Li et al. (2017b) have identified hundreds of m¹A sites, leading them to consider m¹A as a prevalent modification in human mRNAs. Another independent research team successfully engineered a highly sensitive reverse transcriptase HIV-1, which exhibited strong readthrough capabilities and high mutation rates specifically at m¹A sites (Zhou et al. 2019). The evolved reverse transcriptase enabled the detection of well-characterized m¹A sites and revealed hundreds of m¹A sites in human mRNA, aligning with the findings reported by Li et al. The limited specificity of m¹A-antibody potentially introduces noise into the m¹A maps, highlighting the necessity of using demethylation controls. Additionally, ensuring high-quality RNA and utilizing suitable bioinformatic tools are also crucial for m¹A identification (Dominissini and Rechavi 2017; Xiong et al. 2018a; Grozhik et al. 2019).

NEW DIRECTIONS FOR m¹A: MOLECULAR FUNCTIONS AND THERAPEUTIC APPLICATIONS

Aberrant expression of m¹A levels or its related enzymes has also been implicated in various physiological and pathological processes (Xiong et al. 2018b; Zhang and Jia 2018). tRNA is a central component of protein synthesis, decoding the mRNA information, and tRNA is heavily modified to regulate the translation process meticulously. m¹A58 in tRNA plays an important role in maintaining the general tertiary structure of most tRNAs. In yeast initiator tRNA^iMet, a distinctive “substructure” is stabilized by hydrogen bonds formed between adenosines A20, A54, and A60. Notably, m¹A58 makes a significant contribution to this substructure by hydrogen bonding with A54 and A60 (Basavappa and Sigler 1991). The presence of m¹A58 in all eukaryotic initiator tRNAs implies that m¹A58 ensures the correct folding of tRNA molecules. When ALKBH1 is depleted, the m¹A58 level in most tRNA is increased as well as the tRNA^iMet expression level is up-regulated, indicating that m¹A58 is important for maintaining tRNA stable structure and mature tRNA expression level, consistent with previous studies. Moreover, the global translation initiation is impacted by tRNA^iMet, up-regulated tRNA^iMet affects translation initiation, and the m¹A-methylated tRNAs are also preferentially recruited to polysomes to promote translation elongation (Fig. 2A). This dynamic mechanism is utilized by human cells during glucose deprivation to cope with stressful conditions (Liu et al. 2016b). Besides the glucose deprivation, the process of T cell activation can also be regarded as a stress; upon antigen stimulation, naive T cells undergo rapid and profound transitions from a quiescent state to an activated state, triggering extensive protein translation to support cell growth and proliferation by synthesizing millions of new proteins. Upon the dramatic activation, the mRNA level and protein level of m¹A writer TRMT6/TRMT61A complex is significantly increased at an early stage, and TRMT61A deficiency affects T cell activation and proliferation, which is a catalyzed-dependent phenotype, indicating that TRMT61A plays an important role in T cell activation. By performing tRNA-seq and tRNA-m¹A-seq at different time points (0, 6, 18, and 48 h) after antigenic stimulation, tRNA transcripts display differential dynamic expression patterns and are up-regulated sequentially, with the early response tRNAs having higher m¹A modification levels. Furthermore, the regulator Myc, which can promote cell proliferation and is rapidly increased upon T-cell activation, is preferred to usage of the early response tRNA with a higher m¹A modification level (Fig. 2B). In summary, to satisfy the need for protein synthesis, TRMT6/TRMT61A is up-regulated, leading to an increase in tRNA abundance with m¹A58 modification, ensuring Myc translation, which is critical for mitosis and cell cycle progression (Liu et al. 2022). Besides the global m¹A methylation level change, a subset of tRNA aberrant m¹A modification is also accounted for in regulation of translation. A study reported that in four tRNAs with higher m¹A58 methylation levels in hepatocellular carcinoma (HCC) liver cancer stem cells (CSCs) compared to non-CSCs, the abnormally higher m¹A levels in a subset of tRNA increases PPARδ translation, which in turn triggers cholesterol synthesis to activate Hedgehog signaling, eventually driving self-renewal of liver CSCs and liver tumorigenesis (Wang et al. 2021). Together, tRNA-m¹A58 stabilizes tRNA structure, and promotes the translation initiation and translation elongation, affecting gene expression in many ways. In addition to m¹A modification in intact tRNA, tRFs can also inherit modifications from their parental tRNAs. Specifically, m¹A modification occurs at the fourth position of a 22-nt long 3′-fragment derived from tRNAs (named tRF-3bs), which corresponds to the m¹A58 position on mature tRNAs. This modification is dependent on TRMT6/TRMT61A. tRF-3bs exhibits miRNA-like functionality. In bladder cancer, elevated levels of TRMT6/TRMT61A lead to increased m¹A levels in tRF-3bs (Fig. 2C). The m¹A modification on tRF-3bs prevents it from silencing the expression of various genes in the unfolded protein response pathway within cancer cells, resulting in increased expression of those genes. This may be due to the disruption of regular Watson–Crick base-pairing caused by the m¹A modification (Su et al. 2022). This study explores the m¹A function on tRF-3bs, but the function of tRNA-m¹A58 modification is not included in this study, and the mechanism of elevated m¹A modification level impact on tRF-3bs generation is not clear. In another study, ALKBH3-mediated m¹A demethylated tRNAs are more sensitive to ANG treatment and tend to generate tRF around their anticodon regions (Chen et al. 2019). In the future, further investigation will be needed to elucidate the regulatory role of m¹A modification in tRF generation.

As a reversible mRNA modification, compared to m⁶A modification, the physiological and pathological understanding of m¹A modification in mRNA has lagged. This is primarily due to the molecular function of m¹A in mRNA that remains elusive and the absence of identified specific readers for m¹A. m¹A modification located in the 5′-UTR region of nuclear-encoded mRNAs, but not those in CDS or 3′ UTR, is found to correlate with higher translation efficiency, whereas m¹A sites in the CDS of mt-mRNAs show an inhibitory role in translation regulation (Fig. 2D; Dominissini et al. 2016; Li et al. 2017b). Thus, the location of m¹A may dictate its function: it is speculated that m¹A disrupts base-pairing at the Watson–Crick interface. In the 5′ UTR, m¹A might aid in untangling complex structures, facilitating protein translation initiation, whereas in the CDS region, it could impede translation elongation by disrupting base-pairing. During heat shock conditions, m¹A serves as a protective mechanism by safeguarding mRNAs into SGs, preventing their irreversible aggregation and preserving the integrity of the transcripts (Alriquet et al. 2021). This observation strongly suggests that m¹A modification is dynamically regulated in response to physiological conditions. YTH domain-containing proteins, well known as readers of m⁶A, are also considered as readers of m¹A in RNA, and YTHDF2 plays a role in destabilizing m¹A-containing RNAs, ultimately leading to mRNA degradation. However, YTH domain-containing proteins are primarily regarded as readers of m⁶A, whereas their involvement in the functional aspects of m¹A is relatively limited. While interactions between YTH domain-containing proteins and m¹A have been identified, primarily in vitro, additional experiments in vivo are required to clarify whether these interactions involve direct m¹A:YTHDF2 interactions (Fig. 2E; Seo and Kleiner 2020). A study showed that the ALKBH3 removes m¹A sites on Aurora A mRNA, a master suppressor of ciliogenesis. Depletion of ALKBH3 enhances the decay of Aurora A mRNA and inhibits its translation (Kuang et al. 2022). It is suggested that this effect may be mediated through m¹A-dependent mRNA degradation. Recently, a study discovered the degree of m¹A methylation in the CAG repeat increased significantly with the increase of the number of repeats, and TRMT6/TRMT61A complex is required for the catalyze A in the CAG repeat RNA to m¹A. The ALKBH3 is responsible for the demethylation of m¹A. Longer CAG repeat lengths in RNA are associated with neurodegenerative diseases, and TDP-43 is extensively studied as the clinical hallmark of many neurological diseases; its cytoplasmic mislocalization and aggregation are recognized as contributing factors to neurodegeneration. Surprisingly, TDP-43 is identified as a newly discovered reader protein for m¹A in mRNA; it directly interacts with m¹A in mRNA; displaying a notably strong affinity for m¹A-containing CAG repeat RNA. Mechanistically, m¹A induces the sequestration of TDP-43 into SGs, consequently disrupting the phase separation of TDP-43, leading to cytoplasmic redistribution and aggregation of TDP-43, ultimately contributing to the development of neurological diseases (Fig. 2F; Sun et al. 2023b). Inhibitors targeting the TRMT6/TRMT61A complex have been identified, demonstrating significant therapeutic efficacy for m¹A-related diseases (Wang et al. 2021). These results revealed a novel pathological function of m¹A, eliciting its aberrant biochemical and biophysical properties.

CONCLUSION

The advancement of transcriptome-wide sequencing technologies has facilitated the global mapping and dynamic study of Ψ and m¹A modifications in mRNA. Recently, a method for the absolute quantification of single-base m⁶A methylation, GLORI, has been developed, enabling the detection of m⁶A modifications as well as m⁶Am modifications located specifically at transcription start sites. These technologies have gone through low-resolution and qualitative detection to sensitive, single-base, and quantitative assessments. They provide essential tools for functional decoding in the context of biological processes, including human development, diseases like cancer, and neurological deficits. However, achieving single-cell-level detection of Ψ and m¹A modifications with quantitative data remains a future challenge, which will unveil the spatial and temporal heterogeneity of these modifications. Another challenge is using long-read RNA sequencing to study RNA modifications, including Ψ and m¹A, at a single-molecule resolution. This approach will allow for the simultaneous study of multiple modifications, providing comprehensive insights into the interplay and overall profiles of combinatorial mRNA modifications.

Modification patterns are determined by the interplay of writers and erasers, with specific readers executing their functions (Flamand et al. 2023). Functional studies of mRNA modifications mainly depend on perturbations of these effector proteins. However, their regulatory mechanisms require careful investigation. On one hand, it is essential to determine whether RNA-modifying enzymes regulate mRNA metabolism in a catalytically dependent or independent manner. For instance, PUS10 and TRUB1 proteins have roles in miRNA biogenesis independent of their catalytic activity (Kurimoto et al. 2020; Song et al. 2020). On the other hand, many mRNA-modifying enzymes also target different RNA species, such as tRNA and rRNA. To understand how gene- and position-specific modification sites impact RNA metabolism, future studies may need to use CRISPR/dCas13b-tethered effectors targeted at Ψ or m¹A sites on transcripts. Despite some inconsistencies in the reported number of m¹A sites in human mRNA, the characterization of effector proteins underscored their significance in regulating RNA metabolism and various biological processes. For Ψ, several RBPs are known for differential binding affinity depending on modifications, but the readers that mediate effects in endogenous contexts are yet to be determined. In sum, further research into deciphering modification stoichiometry and functions holds great promise for biological decoding and therapeutic applications of Ψ and m¹A modifications.

COMPETING INTEREST STATEMENT

C.Y. is a founder of Modit Therapeutics. M.Z. and X.Z. declare no competing interests.