m¹A RNA Modification in Gene Expression Regulation

Abstract

N¹-methyladenosine (m¹A) is a prevalent and reversible post-transcriptional RNA modification that decorates tRNA, rRNA and mRNA. Recent studies based on technical advances in analytical chemistry and high-throughput sequencing methods have revealed the crucial roles of m¹A RNA modification in gene regulation and biological processes. In this review, we focus on progress in the study of m¹A methyltransferases, m¹A demethylases and m¹A-dependent RNA-binding proteins and highlight the biological mechanisms and functions of m¹A RNA modification, as well as its association with human disease. We also summarize the current understanding of detection approaches for m¹A RNA modification.

1. Introduction

Cellular RNAs contain more than 170 different types of chemical modifications across species [1]. N¹-methyladenosine(m¹A) is a reversible methylation involving the addition of a methyl group at the N¹ position of adenosine in cellular transcripts [2]. The methyl group can block the normal Watson–Crick base pairing of A:T or A:U, resulting in an unstable mismatch with other nucleosides by forming Hoogsteen base pairs [3]. The secondary structure and RNA–protein interaction of m¹A-modified RNAs are also altered under physiological conditions [4]. As a dynamic and reversible post-transcriptional RNA modification, m¹A can be installed by methyltransferases, removed by demethylases and recognized by m¹A-dependent RNA-binding proteins [2,5]. m¹A RNA modification affects RNA metabolism, including RNA structure, stability and mRNA translation, thereby regulating gene expression and several fundamental cellular processes [6].

m¹A RNA modification has been found with high abundance in transfer RNAs (tRNAs) and ribosomal RNAs (rRNAs) but at low levels in messenger RNAs (mRNAs) [7,8,9,10,11,12]. It occurs in the tRNA of bacteria, archaea and eukaryotes at positions 9, 14, 16, 22, 57 and 58 (m¹A9, m¹A14, m¹A16, m¹A22, m¹A57, and m¹A58, respectively) [13]. In cytosolic (cyt) tRNAs, m¹A RNA modification occurs at five different positions (9, 14, 22, 57, and 58) [14,15]. Among them, m¹A14 has only been identified in cyt(tRNA)^Phe from mammals, m¹A22 has only been identified in bacteria tRNAs, and m¹A57 has been identified in archaea existing only transiently as an intermediate of 1-methylinosine (m¹I) [14,15]. In mitochondria, m¹A9 is quite abundant and found in 14 species of mt-tRNA, while m¹A58 is a minor modification with a 17% frequency found in four species of mt-tRNAs [16]. Additionally, m¹A16 is unique to human mt-tRNA^Arg, and its frequency is approximately 20% [16]. For rRNAs, the nuclear-encoded large subunit rRNA m¹A645 in 25S rRNA and m¹A1322 in 28S rRNA located in the peptidyl transfer center of the ribosome are conserved in budding yeast and humans, respectively [17,18,19], and m¹A is conserved at position 947 of 16S rRNA in the mitochondrial ribosome of vertebrates [20]. Regarding mRNAs, m¹A in mRNA accounts for approximately 0.015–0.054% of all adenosines in mammalian cell lines and 0.05–0.16% in mammalian tissues [9,10,21]. m¹A sites are usually located near the translation start site and the first splice site of mRNA, and they are associated with the translation of coding transcripts [9,10].

In this review, we describe mammalian m¹A RNA-modifying proteins that specifically install, remove, and bind to the m¹A base. We also discuss the recent progress in understanding the biological mechanisms involving m¹A in post-transcriptional gene expression regulation and the biological functions of m¹A. Further, we review the current approaches for transcriptome-wide and single-base resolution m¹A detection.

2. m¹A RNA-Modifying Proteins

Reversible m¹A methylomes in nuclear- and mitochondrial-encoded transcripts are achieved via the dynamic regulation of m¹A RNA-modifying proteins (m¹A methyltransferases, m¹A demethylases and m¹A-dependent RNA-binding proteins). The characterization of m¹A-modifying proteins is crucial for understanding the mechanisms underlying m¹A-mediated gene regulation and the biological roles of m¹A RNA modification. To date, several m¹A RNA-modifying proteins responsible for nuclear- and mitochondrial-encoded transcripts have been identified in humans (Figure 1).

Figure 1.

m¹A-modifying proteins for different types of RNAs. The nuclear-encoded (top panel) and mitochondrial (bottom panel) RNAs are reversibly methylated by m¹A methyltransferases (blue; dark blue represents catalytic core of the methylase complex), demethylased by m¹A demethylases (pink), and bound by m¹A-dependent RNA-binding proteins (green). A, adenosine; m¹A, N¹-methyladenosine; TRMT, tRNA (adenine (58)-N (1))-methyltransferase subunit; ALKBH, α-ketoglutarate-dependent dioxygenase alkB homolog; FTO, α-ketoglutarate-dependent dioxygenase alkB homolog FTO; NML, nucleomethylin; YTHDF, YTH domain-containing family protein; YTHDC1, YTH domain-containing protein 1; SDR5C1, 3-hydroxyacyl-CoA dehydrogenase type-2.

2.1. m¹A-Modifying Proteins for Nuclear-Encoded Transcripts

m¹A is highly abundant in tRNAs, and data on the modifying proteins for m¹A58 are clearer than those for other m¹A sites in cytoplasmic tRNAs. The tRNA m¹A58 methyltransferase complex was first identified in Saccharomyces cerevisiae, which contains a non-catalytic subunit and a catalytic subunit encoded by tRNA (adenine(58)-N(1))-methyltransferase non-catalytic subunit TRM6 (Trm6) and Trm61, respectively [22]. Trm61 is critical for AdoMet binding and catalytic function, while Trm6 is responsible for tRNA binding [23]. The conserved homolog of this complex in eukaryotes comprises TRMT61A and TRMT6, which mediate the m¹A58 modification in cytoplasmic tRNAs [24]. Demethylases for nuclear-encoded tRNAs include α-ketoglutarate-dependent dioxygenase alkB homolog 1 (ALKBH1), ALKBH3 and α-ketoglutarate-dependent dioxygenase FTO (FTO), all belonging to the AlkB family of Fe(II)/α-ketoglutarate-dependent dioxygenases [25,26].

rRNAs act as scaffolds for ribosomal proteins and as ribozymes for peptide bond formation in ribosomes; these functions are partly regulated by post-transcriptional modifications [26]. The m¹A modification has been founded in the large subunit of nuclear-encoded rRNAs in yeast and humans [27]. Yeast 25S rRNA (adenine(645)-N(1))-methyltransferase (Rrp8) and 25S rRNA (adenine(2142)-N(1))-methyltransferase (Bmt2) were reported to be the methyltransferases responsible for the m¹A645 and m¹A2142 modifications, respectively, of 25S rRNA [28,29]. The homolog of yeast Rrp8 in humans is nucleomethylin (NML), a nucleolar factor that catalyzes the m¹A modification of 28S rRNA in human and mouse cells [30,31]. NML contains a Rossmann-fold methyltransferase-like domain that binds to S-adenosyl-L-methionine, which acts as a methyl donor [30]. YTH domain-containing family protein 3 (YTHDF3) was identified as an m¹A-dependent RNA-binding protein in HEK293T and RAW264.7 cells, because it showed the highest ability to bind to a human 28S rRNA-derived m¹A probe. A subsequent RNA immunoprecipitation analysis confirmed that m¹A-modified RNA was bound by YTHDF3, but not YTHDF1/2, in trophoblast HTR8/SVneo cells [32].

Studies have shown that tRNA m¹A-modifying proteins are also responsible for the m¹A modification of mRNAs. For example, the TRMT6/TRMT61A complex mediates the m¹A addition to mRNA. TRMT6/61A also requires a consensus GUUCRA motif and a tRNA T-loop-like structure for substrate recognition in mRNA, which is similar to tRNA recognition [12]. In contrast, m¹A in mRNA can be demethylated by ALKBH3 [9,10,33]. In total, 774 m¹A peaks were reported to be specifically detected in ALKBH3-knockout HEK293T cells, and these peaks were enriched in the 5′UTR [9]. Stable isotope labeling by amino acid in cell culture (SILAC)-based quantitative proteomics have identified that YTH domain-containing proteins (YTHDF1-3 and YTHDC1) act as m¹A-dependent RNA-binding proteins in mRNA; these proteins can directly bind to an m¹A-modified RNA probe (sequence from the human transcription factor SOX-18 mRNA) in HEK293T cells. It was also found that Trp432 in YTHDF2 is a key binding residue [34]. m¹A RNA modification was reported to favor a GA-rich motif, and transcripts containing this motif were increased after ALKBH3 depletion [9,11]. Therefore, the researchers of another study designed a biotin-labeled RNA probe in which m¹A RNA modification was placed in a GAGGm¹AG sequence. It could be recognized by YTHDF1 and YTHDF2 in HeLa cells, but YTHDC1 was not detected in this system [35].

2.2. m¹A-Modifying Proteins for Mitochondrial RNAs

tRNA methyltransferase 10 homolog C (TRMT10C) acts as a catalytic core in the TRMT10C/SDR5C1 (also known as HSD17B10, hydroxysteroid 17-β dehydrogenase 10) methylase complex, which is responsible for the m¹A9 of mt tRNA^Lys and m¹A1374 of mitochondrial NADH–ubiquinone oxidoreductase chain 5 (ND5) mRNA [12,36]. In addition, TRMT10C is also able to catalyze the m¹G9 modification in mt tRNAs [36]. TRMT61B catalyzes the m¹A58 of mt tRNA^{Leu (UUR)} and m¹A947 of mt 16S rRNA [20,37]. Two demethylases, ALKBH1 and ALKBH7, have been shown to remove m¹A sites from mt RNAs. The level of m¹A16 of mt-tRNA^Arg and m¹A58 of mt-tRNA^Lys was found to be increased upon ALKBH1 knockout, indicating the role for ALKBH1 in the m¹A of mt RNAs [38]. The human ALKBH7 has been reported to demethylate m¹A in Ile and Leu1 pre-tRNA in the mitochondria [39].

3. Biological Functions of m¹A RNA Modification

Since the discovery of m¹A RNA modification as a chemical modification of RNAs, efforts have been taken to understand the functional characterization of this dynamic methylation in RNA metabolism and gene expression regulation.

3.1. m¹A RNA Modification in RNA Metabolism

m¹A RNA modification is a pivotal regulator of RNA metabolism, including RNA structure alteration, decay and translation (Figure 2).

Figure 2.

Action mechanisms of m¹A in RNA metabolism. m¹A RNA modification regulates RNA metabolism in multiple layers (from top to bottom: (1) m¹A RNA modification stabilizes tRNAs to promote translation initiation; (2) m¹A-modified mRNAs interfere with Watson–Crick base-pairing with tRNA to suppress translation; (3) m¹A-modified tRNAs are coupled with eEF1α to polysomes to promote translation elongation; (4) m¹A-modified mRNAs are subjected to degradation by interacting with YTHDF2; (5) m¹A-modified mRNAs become stable when they bind to YTHDF3). m¹A, N¹-methyladenosine; eEF1α, eukaryotic elongation factor 1-α; YTHDF, YTH domain-containing family protein.

The chemical properties of m¹A RNA modification enable changes in RNA secondary structure. For instance, m¹A9 and m¹A58 in tRNAs are required for the conformational shift of mitochondrial tRNA^Lys and tRNA^iMet, respectively, which contribute to the stabilization of alternative native structures [40,41,42,43]. The loss of m¹A645 has been shown to affect the topological structure of 28S rRNA and alter the RNA interactome [31]. m¹A was also found to favor the hairpin structure of palindromic RNA sequences, wherein m¹A can stably localize within apical loops [44]. A recent study revealed that m¹A RNA modification controlled RNA conformational equilibrium by blocking base-pairing to modulate the RNA duplex [3].

The regulation of m¹A-modified mRNA decay is mediated by m¹A-dependent RNA-binding proteins. Limited evidence suggests that the knockdown of YTHDF2 increases the abundance of 7 out of 8 m¹A-modified transcripts and 2 out of 3 transcripts that bear only the m¹A but not m⁶A (N⁶-methyladenosine) modification [35]. In addition to YTHDF2, YTHDF3 overexpression has been reported to decrease the abundance and decay rate of insulin like growth factor 1 receptor (IGF1R) mRNA [32].

Translational regulation by m¹A modification varies among different RNA types. The m¹A demethylases ALKBH1 and FTO have been reported to control specific tRNA m¹A demethylation and decrease translation initiation [45,46]. Eukaryotic elongation factor 1-α (eEF1α) immunoprecipitation was used to reveal that m¹A-methylated tRNAs are enriched in polysomes, indicating the role of m¹A RNA modification in translation activation [45]. During retroviral reverse transcription in early human immunodeficiency virus 1 (HIV-1) replication, TRMT6-mediated m¹A58 of tRNA₃^Lys acted as a stop site that contributed to genome integration [47]. Further, mRNAs carrying m¹A undergo translation repression because of interfered Watson–Crick base pairing [8,12,48].

3.2. m¹A RNA Modification in Biological Processes

Post-transcriptional modifications are involved in various biological processes, and recent evidence showed the importance of m¹A RNA modification in this field. In a high-temperature-sensitive Thermococcus kodakarensis strain, decreased m¹A58 and melting temperature of tRNA were observed, suggesting the relevance of m¹A58 and the growth ability of this strain at high temperatures [49]. m¹A RNA modification was found to exhibit its protective ability of RNAs under stress conditions. During heat shock, m¹A-harbouring transcripts were found to preferentially accumulate in stress granules, subsequently resulting in a shorter time to restore the translation state during recovery [50]. Alkylating agents induced m¹A modification in RNAs and orchestrated translational suppression by recruiting the ASCC damage repair complex (activating signal cointegrator 1 complex) [51]. The tRNA modification profiles of the Aplysia central nervous system showed increased m¹A RNA modification levels in animals after behavioral training [52]; this was the first study to characterize the variable pattern of m¹A RNA modification during defensive reflex-associated behavioral sensitization. Petunia TRMT61A catalyzed m¹A RNA modification in mRNAs, and the knockdown of TRMT61A decreased the chlorophyll content and changed chlorotic and wrinkled leaf phenotype [53]. A recent study showed that the m¹A demethylase ALKBH3 functioned as a negative regulator of ciliogenesis by removing the m¹A sites on Aurora A mRNA (a key regulator of cilia disassembly) in mammalian cells, which was further involved in cilia-associated developmental processes in zebrafish [54].

4. m¹A RNA Modification in Diseases

The limited exploration of m¹A RNA modification as a pathological feature has mainly focused on tumor progression (Table 1). It was reported that the knockdown of m¹A demethylase ALKBH3 increased the abundance of m¹A RNA modification in small RNAs (< 200 nucleotides) along with suppressed nascent protein in pancreatic cancer cells [55]. The ALKBH3-dependent m¹A demethylation of macrophage colony-stimulating factor 1 (CSF1) mRNA enhanced its mRNA stability and thus promoted the invasion of breast and ovarian cancer cells [56]. In addition, ALKBH3 removed the m¹A RNA modification of tRNA^GlyGCC to promote tRNA cleavage by angiogenin. The generation of excessive tRNA-derived small RNAs may affect ribosome assembly and apoptosis in HeLa cells [57]. Furthermore, ALKBH3 promoter CpG island hypermethylation and transcriptional silencing were found in Hodgkin lymphoma cells, which were identified as a potential prognostic biomarker associated with poor clinical outcomes in patients with Hodgkin lymphoma [58]. A recent study found that levels of tRNA m¹A modification were upregulated in hepatocellular carcinoma (HCC) tissues. The TRMT6/TRMT61A complex mediated increased m¹A58 levels in tRNA, which then triggered peroxisome proliferator-activated receptor delta (PPARδ) mRNA translation in HCC stem cells. PPARδ promoted cholesterol biogenesis to activate the Hedgehog pathway, thereby initiating the self-renewal of HCC stem cells [59].

Table 1.

Dysregulation of m¹A RNA modification in human cancers.

Cancers	m1A-Modifying Proteins	Roles	Targets	Mechanisms	Refs
Pancreatic cancer	ALKBH3	Oncogene	small RNAs	Unknown	[55]
Breast and ovarian cancer	ALKBH3	Oncogene	CSF1	mRNA decay	[56]
Cervical cancer	ALKBH3	Oncogene	tRNAs	tRNA cleavage	[57]
Hodgkin lymphoma	ALKBH3	Tumor suppressor	COL1A1, COL1A2	Unknown	[58]
Hepatocellular carcinoma	TRMT6/TRMT61A	Oncogene	tRNAs	Unknown	[59]

ALKBH, α-ketoglutarate-dependent dioxygenase alkB homolog; TRMT, tRNA (adenine(58)-N(1))-methyltransferase subunit; CSF-1, macrophage colony-stimulating factor 1; COL1A1, collagen α-1(I) chain; COL1A2, collagen α-2(I) chain.

5. Approaches for m¹A RNA-Modification Detection

Several methods have been developed for the detection of m¹A modifications in RNAs. m¹A can be detected by bi-dimensional thin-layer chromatography (2D-TLC) or high-pressure liquid chromatography (HPLC), both of which utilize the differential retention properties for the separation of various nucleosides [4]. Furthermore, the coupling of liquid chromatography to mass spectrometry (LC–MS/MS) can provide a more sensitive approach for quantifying the m¹A level in RNAs [4]. The drawback of these methods is that the location information of m¹A is generally lost.

Recent advances in high-throughput technologies have revealed the global m¹A distribution map and dynamics in the transcriptome, and developed tools provide easy ways for m¹A site-specific detection. Here, we summarize the high-throughput sequencing approaches for m¹A RNA-modification detection (Table 2).

Table 2.

Detection methods for mapping m¹A RNA modification.

Methods	Features	RNA Substrate	Antibody Dependent	Resolution	Cell Line
ARM-seq [7]	AlkB treatment	tRNA	No	Fragment	BY4741, GM05372, GM12878
m1A-quant-seq [11]	Spike-in m1A RNA; AlkB treatment; RT-1306-mediated RT mutation	tRNA, rRNA, mRNA, lncRNA	No	Single base	HEK293T
m1A-seq [10]	m1A-RNA immunoprecipitation; Dimroth rearrangement	rRNA, mRNA	Yes	Fragment	BY4741, Sp1, MEF, mESC, HeLa
m1A-ID-seq [9]	m1A-RNA immunoprecipitation; AlkB treatment; AMV-mediated RT truncation	rRNA, mRNA, lncRNA	Yes	Fragment	HEK293T
m1A-seq-SS [12]	m1A-RNA immunoprecipitation; Dimroth rearrangement; SS-mediated RT truncation	tRNA, rRNA, mRNA, lncRNA	Yes	Single base	HEK293T
m1A-seq-TGIRT [12]	m1A-RNA immunoprecipitation; Dimroth rearrangement; TGIRT-mediated RT readthrough	tRNA, rRNA, mRNA, lncRNA	Yes	Single base	HEK293T
m1A-MAP [8]	m1A-RNA immunoprecipitation; AlkB treatment; TGIRT-mediated RT read-through	tRNA, rRNA, mRNA, lncRNA	Yes	Single base	HEK293T
m1A-IP-seq [11]	m1A-RNA immunoprecipitation; AlkB treatment; RT-1306-mediated RT mutation	tRNA, rRNA, mRNA, lncRNA	Yes	Single base	HEK293T

AlkB, α-ketoglutarate-dependent dioxygenase; RT, reverse transcription; AMV, avian myeloblastosis virus; SS, SuperScript III; TGIRT, thermostable group II intron reverse transcriptase; BY4741, saccharomyces cerevisiae BY4741 strains; GM05372, human B lymphocyte-derived GM05372 cell; GM12878, human B lymphocyte-derived GM12878 cell; HEK293T, human embryonic kidney 293T cell; Sp1, schizosaccharomyces pombe Sp1 cells; MEF, mouse embryonic fibroblasts; mESC, mouse embryonic stem cell; HeLa, human HeLa cell.

5.1. m¹A Antibody-Independent Detection Methods

Several enzyme-based strategies have been developed for detecting m¹A RNA modification. Escherichia coli α-ketoglutarate-dependent dioxygenase, AlkB, can demethylase several post-translational modifications to avoid unexpected termination during reverse transcription. In ARM-seq (AlkB-facilitated RNA methylation sequencing), the removal of m¹A, m³C (N³-methylcytidine), or m¹G (N¹-methylguanosine) modifications by AlkB treatment facilitates the production of full-length cDNAs from previously modified templates, producing a ratio of reads in treated versus untreated samples that can be used to identify methylated RNAs [7]. RT-1306 is an evolved HIV-1 reverse transcriptase that can induce m¹A-specific base mutations, thus providing a direct and rapid method for mapping m¹A modification [11]. One possible drawback is that the accuracy is dependent on the fidelity and specificity of the enzymes used in these approaches.

5.2. m¹A Antibody-Dependent Detection Methods

m¹A antibody-dependent RNA immunoprecipitation has several advantages in m¹A-modification detection. m¹A antibody-dependent methylated RNA immunoprecipitation sequencing (m¹A-MeRIP-seq) combined with AlkB demethylation (m¹A-ID-seq) has been used to describe m¹A dynamics in the human transcriptome [9]. Dominissini and colleagues also developed a method termed ‘m¹A-seq’ that combined MeRIP-seq and Dimroth rearrangement to obtain the mRNA m¹A methylome in human cells [10]. Moreover, combining advanced reverse transcriptase TGIRT (thermostable group II intron reverse transcriptase) and SuperScript III—which enable efficient RT misincorporation and truncation at m¹A sites—with m¹A antibody was shown to provide a feasible way to map m¹A at a single-base resolution (m¹A-seq-TGIRT and m¹A-seq-SS) [12]. The misincorporation-assisted profiling of m¹A (m¹A-MAP) has improved the previous m¹A-ID-seq by utilizing TGIRT and AlkB treatment. m¹A-MAP has been used to identify a total of 740 m¹A sites in the transcriptome and showed a GUUCRA consensus motif located in a subset of tRNA substrates of the TRMT6/61A complex [8]. A major drawback of these approaches is the non-specific binding of m¹A antibodies, which may cause false positives.

5.3. Site-Specific m¹A Detection

Based on m¹A-induced Watson–Crick base-pairing disruption, some approaches have been developed for site-specific m¹A evaluation. m¹A induces the truncation or mutation of cDNA during reverse transcription, so primer extension can be applied to compare m¹A modification levels based on the intensity of the truncated and full-length cDNA bands [60]. The clustered regularly interspaced short palindromic repeat (CRISPR)-associated endoribonuclease Cas13 can mediate the cleavage of both target and collateral RNAs after the formation of the Cas13/crRNA/target RNA complex. Based on this phenomenon, a reporter system with CRISPR/Cas13a can produce a high-fluorescence signal by cleaving quenched fluorescent RNAs with the correct Watson–Crick base-pairing of crRNA or target RNA but not m¹A-modified RNAs [61]. Another method takes advantage of the different nick ligation efficiencies by T3 DNA ligases at the adenosine and m¹A base. It can easily detect the m¹A state at a specific site by analyzing the PCR amplification products from nick ligation [62]. These methods are suitable for RNA with high abundance and known sequence of m¹A sites rather than the identification of novel modification sites.

6. Future Perspectives

Although findings related to the biological functions of the m¹A modification have been reported, some key questions remain to be addressed. For instance, studies on m¹A-dependent RNA-binding proteins are limited. With the advances in the sensitivity and throughput of proteomics techniques, it is critical to further characterize m¹A-dependent RNA-binding proteins involved in m¹A recognition and RNA metabolism, as well as to determine how they recognize the m¹A sites in their target transcripts. In addition, the reported methyltransferases and demethylases may target different transcripts in different cell types or biological processes; therefore, the context-dependent roles and regulatory mechanisms of m¹A-modifying proteins are also interesting and important to explore. Further, optimized technologies are still needed for transcriptome-wide and single-base resolution m¹A detection. These will greatly improve our understanding of the biological roles of m¹A in RNA. Finally, more work is needed to understand the roles of the m¹A RNA modification in human diseases.

Author Contributions

H.J. wrote the manuscript and prepared the figures; C.H. prepared the figures and tables; T.Z. revised the manuscript; S.X. conceptualized and supervised the manuscript. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Not applicable.

Conflicts of Interest

The authors declare no conflict of interest.

Funding Statement

This research was funded by the Chinese National Natural Scientific Foundation Grants, grant number U21A20197.