RNA 5-methylcytosine modification and its emerging role as an epitranscriptomic mark

ABSTRACT

5-methylcytosine (m⁵C) is identified as an abundant and conserved modification in various RNAs, including tRNAs, mRNAs, rRNAs, and other non-coding RNAs. The application of high-throughput sequencing and mass spectrometry allowed for the detection of m⁵C at a single-nucleotide resolution and at a global abundance separately; this contributes to a better understanding of m⁵C modification and its biological functions. m⁵C modification plays critical roles in diverse aspects of RNA processing, including tRNA stability, rRNA assembly, and mRNA translation. Notably, altered m⁵C modifications and mutated RNA m⁵C methyltransferases are associated with diverse pathological processes, such as nervous system disorders and cancers. This review may provide new sights of molecular mechanism and functional importance of m⁵C modification.

KEYWORDS:

• 5-methylcytosine (m⁵C)
• RNA methylation
• RNA m⁵C methyltransferase
• epitranscriptome

1. Introduction

More than 170 types of RNA nucleotide modifications have been documented [1]. Among these modifications, m⁵C is an abundant and conserved post-transcriptional modification [2]. With the advances in RNA m⁵C detection techniques, including RNA bisulphite sequencing [3], immunoprecipitation-based detection methods, liquid chromatography-tandem mass spectrometry [4], and third-generation sequencing [5], more than 10,000 potential m⁵C modification sites were detected within the whole human transcriptome [6]. A great number of researches have been geared towards exploring m⁵C modification and its distribution, regulation, and functions.

The existence of m⁵C was found in diverse RNA species from various organisms, including plants, animals, and microorganisms [1]. tRNAs are the most enriched substrate of m⁵C modification and they have been most widely explored [7]. The modified regions in tRNA are often at the junction of the variable loop and the T stem-loop. m⁵C modifications in rRNAs are mainly located in 18S rRNAs and 25S rRNAs [8]. It remains less clear whether there exist m⁵C modifications in mRNAs and whether m⁵C is functional in mRNAs [9]. Detected m⁵C sites in mRNAs are mainly located within 3ʹ untranslated region (3ʹUTR) or near translation start sites, affecting mRNA translation efficiency [2]. Transcriptome-wide detection technique also uncovered m⁵C in 5ʹ untranslated region (5ʹUTR) and coding regions, with functions not completely defined [2]. In addition, m⁵C modifications also occur in long non-coding RNA [10], circular RNA [11], small nuclear RNAs and small nucleolar RNAs [12].

RNA m⁵C modifications are regulated by a series of proteins defined as ‘writers’, ‘readers’ and ‘erasers’. The ‘writers’ refer to RNA m⁵C methyltransferases (RCMTs), which consist of the NOL1/NOP2/SUN domain (NSUN) family of proteins (NSUN1 to 7 in humans) and the DNA methyltransferase (DNMT) homologue DNMT2 [13,14]. As ‘erasers’, ten-eleven translocation family proteins (TETs) can oxidize m⁵C into cytosine-5-hydroxymethylation (hm⁵C) in RNA [15], especially in mRNA [16]. At present, only two ‘readers’ have been identified. One is YBX1, recognizing m⁵C-modified mRNA and mediating mRNA stability [17], and the other is ALYREF, recognizing methylated cytosines in GC-rich regions of mRNA and mediating mRNA nuclear export [18,19].

The m⁵C modification is crucial for several biological processes, including tRNA stability [20], rRNA assembly [21] and functionally stress response [22,23] and embryonic and organic development [24]. Besides, altered m⁵C modifications or mutated RNA m⁵C methyltransferases are linked to a few pathological processes, including nervous system disorders [25] and cancers [26].

In this review, we outline the current detection methods of m⁵C in RNA, explore their advantages and disadvantages, and review the m⁵C distribution pattern and advances in RNA m⁵C methyltransferases. Additionally, we evaluate the physiological roles and pathological implications related to m⁵C modification.

2. Detection methods of m⁵C modification

2.1 RNA bisulphite sequencing (RNA-BisSeq)

Description: Bisulphite sequencing is originally applied to detect 5mC in DNA. Currently, it has become the most used technique for sequencing RNA m⁵C. Bisulphite treatment converts unmethylated cytosine to uracil in single-strand RNA, whereby methylated cytosines remain unchanged. Coupling with high throughput sequencing, it maps m⁵C in RNA at a single-nucleotide resolution. This technique has been successfully applied to provide a global view of m⁵C in Drosophila tRNAs [27], rRNAs, and human RNA [2].

Advantages: RNA-BisSeq can provide a transcriptome-wide landscape of m⁵C profile with high specificity at a single-nucleotide resolution [28]. Also, its protocol is relatively straightforward and mature.

Disadvantages: The efficiency of bisulphite conversion is affected by many factors including the structure of RNA, the experiment temperature, pH, and so on. RNA species of which m⁵C positions are well known, such as tRNA^Asp(GTC) or 28S rRNA, can be used to determine the efficiency. Other cytosine modifications, such as hm⁵C, are also detectable leading to false-positive results. Recently reported alternative methods, such as oxBS-seq and TAB-seq, can distinguish hm⁵C [29,30]. Besides, the presence of random unconverted cytosine in the datasets would influence the analysis results. Moreover, its application requires a high amount of RNA molecules because of bisulphite treatment-induced degradation [2].

2.2. Immunoprecipitation-based detection methods

2.2.1. Methylated RNA immunoprecipitation followed by sequencing (MeRIP-seq)

Description: This method utilizes RNA m⁵C antibodies to pull down m⁵C-modified RNA, available for mapping m⁵C sites in low-abundance RNAs. Combined with high throughput sequencing, it provides a transcriptome-wide view of the m⁵C profile, but not at a single-nucleotide resolution [31]. Additionally, 5-methylcytosine binding domain pulldown followed by sequencing (MBD-seq) has also been applied to sequence m⁵C in RNA. MBD-seq captures m⁵C-modified RNA fragments by a protein with high affinity [32].

Advantages: It avoids the interference of other RNA modifications, and provides a transcriptome-wide view of the m⁵C profile even in low-abundance RNAs.

Disadvantages: The specificity of this method highly relies on anti-m⁵C antibodies. The specificity of anti-m⁵C antibodies is affected by various factors. Thus, it is important to include unmodified and modified control RNA to determine the specificity of the antibody in the pulldown.

2.2.2. 5-azacytidine-mediated RNA immunoprecipitation(5-azaIP)

Description: 5-azacytidine is a cytidine analog, which is randomly incorporated into a nascent RNA transcript. At methylation sites, 5-azacytidine binds to RNA m⁵C methylases forming a covalent adduct, which is immunoprecipitated by enzyme-specific or tag antibodies. When RNA is released from the RNA-methylase complex, 5-aza-C ring is hydrolysed, which is read as G during sequencing [33]. Therefore, it provides a transcriptome-wide landscape of m⁵C at a single-nucleotide resolution. This method has been adopted to reveal specific m⁵C sites of DNMT2 and NSUN2 in tRNA and non-coding RNA [33].

Advantages: It reduces the background of non-specific RNAs due to stable covalent binding. Most importantly, it identifies enzyme-specific methylation sites at a single-nucleotide resolution.

Disadvantages: 5-azacytidine is highly toxic for cells, because it can be incorporated into DNA and RNA, altering the normal transcription. There is also concern about the incorporation efficiency of 5-azacytidine.

2.2.3. Methylation-individual nucleotide resolution crosslinking immunoprecipitation (miCLIP)

Description: MiCLIP utilizes mutated RNA m⁵C methyltransferases, which contain a point mutation in conserved cysteines within its catalytic domain, to effectively form irreversible enzyme-RNA catalytic intermediates. These cross-linked covalent adducts are subsequently immunoprecipitated by enzyme-specific or tag antibodies. With RNA release and high-throughput sequencing, we can identify enzyme-specific m⁵C sites at a single-nucleotide resolution [34–36].

Advantages: It provides the same results as 5-azaIP, meanwhile avoiding 5-azacytidine toxicity.

Disadvantages: Relying on mutated enzymes, this approach requires a large amount of time and money [36].

2.3. Third generation sequencing

Description: Third-generation sequencing, including SMRT sequencing and nanopore direct RNA sequencing (DRS), can also be used for single-molecule sequencing and direct RNA modification location analysis. SMRT sequencing differentiates distinct RNA modifications by monitoring HIV reverse transcriptase in real time to distinguish the different dynamic signals of normal and modified bases on the template. Nanopore DRS distinguish distinct modifications by monitoring current changes when RNA molecular traverses through nanopores [37,38].

Advantages: Third-generation sequencing can directly detect RNA sequences and their modifications [5].

Disadvantages: This method is more costly and has a higher error rate.

2.4. Liquid chromatography-tandem mass spectrometry (LC-MS/MS)

Description: LC-MS/MS is a sensitive and quantitative method for detecting RNA modifications, m⁵C included. It provides a global abundance of m⁵C modification in samples, without specific modification sites [19,39,40]. This method has successfully been employed to evaluate the global RNA modification extent in rRNAs [6], tRNAs [41], coding RNAs and non-coding RNAs [42].

Advantages: This method provides a global abundance of m⁵C modification. Also, it can measure other RNA modifications meanwhile [43,44].

Disadvantages: The detection resolution of this method cannot reach a single-nucleotide level (Table 1) (Figure 1).

Table 1. Detection methods of RNA m⁵C

technique		description	advantages	disadvantages	Resolution	references
RNA-BisSeq		sodium bisulphite deaminating unmethylated cytosine into uracil while methylated cytosines unchanged	high specificity	efficiency of bisulphite conversion; presence of random unconverted cytosine in the datasets	single nucleotide resolution	[2,27–30]
Immunoprecipitation-based detection methods	MeRIP-seq	m^5C antibody Immunoprecipitation	avoiding the interference of other RNA modifications; available for low-amount RNAs	highly dependent on the specificity of anti-m^5C antibodies	100–150nt	[31]
	5-azaIP	enzyme-specific antibody Immunoprecipitation	enzyme-specific methylation sites	toxic to cells; incorporation efficiency of 5-azacytidine	enzyme-specific single nucleotide resolution	[33]
	miCLIP	enzyme-specific antibody Immunoprecipitation	enzyme-specific methylation sites; avoiding the toxicity of 5-azacytidine	requiring more time and money	enzyme-specific single nucleotide resolution	[34–36]
Third generation sequencing	SMRT sequencing; nanopore DRS	single-molecule sequencing and direct RNA modification location analysis	directly detecting RNA sequences and their modifications	more costly; a higher error rate	single nucleotide resolution	[37,38]
LC-MS/MS		evaluating a global abundance of RNA modifications	providing a global abundance	not at single-nucleotide resolution	a global abundance	[19]

Figure 1. Detection methods of RNA m⁵C. RNA-BisSeq relied on bisulphite conversion and high-throughput sequencing. Immunoprecipitation-based detection methods are based on specific antibodies targeting m⁵C or enzymes, which include MeRIP-seq, 5-azaIP and miCLIP. Third generation sequencing directly detects RNA sequences and modifications. LC-MS/MS combines high pressure liquid chromatography (HPLC) separation and mass spectrometry (MS) measurement, providing a global abundance of RNA m⁵C modification

3. Distribution and regulation of m5C modification

m⁵C is introduced by RNA m5C methyltransferases, which include NSUN1 to NSUN7 and DNMT2, which are all evolutionarily conserved from archaea to eukaryotes [14]. Their substrates are gradually identified, but not completely (Table 2). The NSUN family all include an S-adenosyl methionine binding-domain and two conserved catalytic cysteines in the active site [45]. DNMT2 possesses a catalytic cysteine in the active site, similar with DNA methyltransferases [46–48].

Table 2. RNA m⁵C methyltransferases and their substrates

methyltransferase	RNA substrates	m^5C sites	functions in RNA metabolism	physiological and pathological functions	references
NSUN1	25S rRNA	m^5C2870 in domain V	ribosome assembly; oxidative phosphorylation	tumorigenesis	[65,66]
NSUN2	tRNA^Leu (CAA)	m^5C34	tRNA stability	stress response, embryonic development, nervous system disorders, intellectual disability
	tRNA^Gly (GCC)	m^5C40,48,49,50			[58–61]
	tRNA^Asp (GTC) tRNA^Val (AAC)	m^5C48,49
	mt-tRNA^Glu mt-tRNA^Phe mt-tRNA^His, mt-tRNA^Leu(UUR) mt-tRNA^Ser(AGY) mt-tRNA^Tyr	m^5C48,49	oxidative phosphorylation	mitochondrial respiratory chain complex deficiency	[62,63]
	mRNAs	various	protein translation, nuclear-cytoplasmic shuttling, stability	cell proliferation and senescence, tumorigenesis,	[2,19]
	non-coding RNAs	various	/	potentially tumorigenesis	[76–78]
	vtRNA1.1	m^5C69	/	epidermal differentiation	[75]
	vtRNA1.2	m^5C27,59	/	/
	vtRNA1.3	m^5C15,27,59	/	/
NSUN3	mt-tRNA^Met	m^5C34	oxidative phosphorylation	mitochondrial respiratory chain complex deficiency, embryonic development	[36,62]
NSUN4	mt-12S rRNA	m^5C911	ribosome assembly, oxidative phosphorylation	mitochondrial respiratory chain complex deficiency	[9,67]
	mt-18S rRNA	m^5C628,631,632
	mt-tRNA^Met(CAU)	m^5C34	tRNA stability	stress response
NSUN5	25S rRNA	m^5C2870 in domain V	ribosome assembly, protein translation	stress response, Williams-Beuren syndrome	[65,89]
NSUN6	tRNA^Thr(CGU/GGU/UGU) tRNA^Cys(GCA)	m^5C72	tRNA stability	stress response	[64,109–111]
NSUN7	eRNAs	various	/	/	[112]
DNMT2	tRNA^Asp(GUC/GTC) tRNA^Glu(CUC/UUC) tRNA^Gly(GCC) tRNA^Val (AAC)	m^5C38	tRNA stability	stress response, embryonic development, tumorigenesis	[7,55–57]

TETs are conventionally responsible for DNA demethylation, oxidizing 5-methyl-2ʹ-deoxycytidine (5mdC) to 5-hydroxymethyl2ʹ-deoxycytidine (5hmdC), and then to 5-formyl-2ʹ-deoxycytidine (5fdC), and eventually to 5-carboxyl-2ʹ-deoxycytidine (5cadC) [29,49]. Recent studies have confirmed that TETs participated in the oxidation of RNA m⁵C into hm⁵C in vitro and in vivo. However, their oxidative activity in RNA is less effective than that in DNA. About 0.02% of m⁵C is modified to hm⁵C in different types of tissues, including brain, lung, and so on, and cellular RNA samples, including 293 T cells, embryonic cells and tumour cells [15]. The extent of hm⁵C was found to be much lower than that of m⁵C in mammals cellular mRNAs [50]. Tet1 could oxidize 5-formylcytosine (f⁵C) to 5-carboxycytidine (ca⁵C) in vitro [51]. Delatte et al. provided the distribution and function of hm⁵C in Drosophila. They found that hm⁵C was preferentially present in polyadenylated RNAs, deposited by TETs [16]. Functionally, TETs-mediated m⁵C oxidization participates in numerous physiological processes. Tet2 can increase infection-induced myelopoiesis by oxidizing m⁵C into hm⁵C in Socs3 mRNA and repressing Socs3 expression [50]. Huang et al. found significantly lower RNA hm⁵C in human colorectal carcinoma and hepatocellular carcinoma tissues compared to tumour-adjacent normal tissues [49]. This may indicate that TETs-mediated m⁵C oxidation may contribute to tumorigenesis.

At present, only two ‘readers’ have been discovered. ALYREF recognizes m⁵C modified mRNAs and mediates their nuclear-cytoplasmic shuttling [18,19]. YBX1 recognizes m⁵C modified mRNAs and mediates their stabilization [17,52].

3.1. m⁵C in tRNA

The modification m⁵C is highly abundant in tRNAs. Three m⁵C methyltransferases, DNMT2, NSUN2, and NSUN6, have been revealed to methylate cytoplasmic tRNAs [53], whereas three m⁵C methyltransferases, NSUN2, NSUN3, and NSUN4, have been reported to modify mitochondrial tRNAs [9,36,54].

tRNA^Asp, as the most common substrate, is methylated by DNMT2 at cytosine 38 at the anticodon loop in mouse, Arabidopsis thaliana, Drosophila melanogaster [7], and Dictyostelium discoideum [55]. In addition to tRNA^Asp(GTC), tRNA^Val(AAC) and tRNA^Gly(GCC) are methylated by DNMT2 in Drosophila [56]. Pmt1, a DNMT2 homolog in Schizosaccharomyces pombe, exerts in vitro and in vivo methylation activities on cytosine 38 of tRNA^Asp and tRNA^Glu [57].

Substrates methylated by NSUN2 include cytosine 34 (pre-tRNA^Leu(CAA)), cytosine 40 (tRNA^Gly(GCC)), cytosines 48 and 49 (tRNA^Asp(GTC), tRNA^Val(AAC), and tRNA^Gly(GCC)), and cytosine 50 (tRNA^Gly(GCC)) in mice [58]. In mammals, the methylated cytosines are located at positions 34, 38, 48, 49, 50, and 72 of tRNAs, and advances in detection techniques enable detection of new sites 19, 27, 40, 55, 61, 66, 70, 73, and 75 [59]. Homologues of NSUN2 include Trm4p in Saccharomyces cerevisiae [60], Trm4 in Saccharomyces cerevisiae, and Trm4a/Trm4b in fission yeast Schizosaccharomyces pombe [61]. Besides cytoplasmic tRNAs, mitochondrial tRNAs are also substrates of NSUN2. Shinoda et al. revealed the localization of NSUN2 within mitochondria through structured illumination microscopy and revealed m⁵C sites 48–50 in mouse and human mitochondrial tRNAs samples. Moreover, in NSUN2 knockout (KO) mice and human cells, m⁵C was absent in mt-tRNAs [62,63].

Another research uncovered that NSUN6 methylates tRNA^Cys(GCA) and tRNA^{Thr(CGU/GGU/UGU)} at C72 within the 3′ end of the tRNA acceptor stem in hyperthermophilic archaea [64].

Haute et al. discovered for the first time that NSUN3 is an RNA methyltransferase, which methylates mitochondrial tRNA^Met at cytosine 34 in the anticodon arm. NSUN3-mediated m⁵C modification initiates f⁵C biogenesis, further affecting mitochondrial proteins translation [36]. Moreover, NSUN4 methylates mt-tRNA^Met(CAU) at cytosine 34 [9]. Shinoda et al. mapped mitochondrial tRNA m⁵C modifications in HEK293T cells. The methylated cytosines are located at position 48 of mt-tRNAs^Phe, mt-tRNAs^His, mt-tRNAs^Leu(UUR), mt-tRNAs^Ser(AGY), and mt-tRNAs^Tyr, at position 49 of mt-tRNAs^Glu and mt-tRNAs^Ser(AGY), and at position 50 of mt-tRNAs^Ser(AGY) [62].

3.2. m⁵C in rRNA

Nuclear, cytoplasmic, and mitochondrial rRNAs are all characterized by m⁵C modification, which mainly influences protein synthesis. In nucleolus, 25S rRNA is methylated in domain V by NSUN1(p120) and NSUN5 [65]. In cytoplasm, 25S rRNA is methylated by NSUN1 [66]. And in mitochondria, 12S rRNA, and 18S rRNA are methylated by NSUN4 [9].

NSUN4 methylates 12S rRNA at C911 in the small ribosomal subunit and methylates 18S rRNA at C628, 631 and 632 in mitochondria [67]. In yeast, Sharma et al. identified two putative methyltransferases, Rcm1 and Nop2, methylate 25S rRNA at C2278 and 2870 [68]. Human proliferation-associated nucleolar antigen p120, the homologue of yeast Nop2, also displays an RNA m5C methyltransferase activity. It is also called NSUN1 in humans [66]. It can restore m⁵C formation in domain V in endogenous yeast 25S rRNA when expressed in a Nop2-deficient yeast strain [6]. NSUN5, a homologue of Rcm1, methylates 28S rRNA at C3782 in human and at C3438 in mice [65].

3.3. m⁵C in mRNA

Transcriptome-wide mapping of m⁵C has been conducted in mammals, plants, bacteria, archaea, and yeast [31,69–71]. m⁵C modification in mRNA is rare or even undetectable [9,72,73]. Advances in m⁵C detection techniques have promoted more extensive analysis. The present identified mRNA methyltransferase is NSUN2 [2].

Through RNA-BisSeq, Squires et al. discovered 10,275 methylation sites in mRNAs and other non-coding RNAs in Hela cells, and these sites were enriched in untranslated regions or near Argonaute binding regions within mRNAs [2]. Yang et al. also revealed that m⁵C modification was enriched in regions immediately downstream of translation initiation sites in Hela cells [19]. Further, Amort et al. demonstrated the landscape of RNA m⁵C modifications in mouse embryonic stem cells (ESC) and brain tissues, in total poly(A) RNA and nuclear poly(A) RNA. They revealed that the m⁵C modification extent and distribution pattern were different in ESC and brain tissue, with a higher extent in ESC. Besides, m⁵C sites were mainly located within introns, CDS and 3ʹUTR in nuclear poly(A) RNA, whereas in total poly(A) RNA, they are mainly located within 3ʹUTR, CDS and around translational start sites [74].

3.4. m⁵C in other non-coding RNAs

m5C modification also occurs in non-coding RNA [2]. The main methyltransferase is NSUN2.

NSUN2 methylates non-coding vault RNA VTRNA1.1 at cytosine 69, determining the transition of VTRNA1.1 into small-vault RNAs, involved in epidermal differentiation. VTRNA1.3 was also reported as another substrate of NSUN2 [75]. Moreover, microRNAs can be methylated by NSUN2, such as micro126b [76]. lncRNA is also modulated by NSUN2. High-throughput sequencing has mapped the landscape of lncRNA regulated by NSUN2, whereas functional analysis has revealed some pathways in tumorigenesis. However, the mechanism via which NSUN2 modulates lncRNA and the specific modification sites are unknown [77]. Amort et al. demonstrated for the first time that the two well-studied regulatory lncRNAs HOTAIR and XIST have m⁵C modifications and the methylated position of HOTAIR is C1683 in fragment H1 and the methylated region of XIST is A-region [78].

4. Biological Functions of m⁵C modification

4.1. m⁵C Modification and RNA processing

4.1.1. tRNA metabolism

The functional characterization of DNMT2 or NSUN2‑mediated tRNA methylation in tRNA metabolism remains unclear. However, present studies have revealed that tRNA methylation can protect tRNAs against endonucleolytic cleavage, maintaining tRNA stability [20]. Under stress stimuli, angiogenin is activated to cleave tRNA within anticodon loops into 5ʹ-tRNA fragments, inhibiting protein translation initiation and inducing stress granules assembly to activate stress response programme [79]. Angiogenin has a higher affinity to unmethylated tRNAs, generating a larger number of angiogenin-induced tRNA fragments [22,23]. The released tRNA fragments repress cap-dependent translation by displacing translation initiation and elongation factors from mRNAs or by altering efficient transpeptidation, therefore leading to activated stress responses [20]. The released tRNA fragments can mediate siRNA pathway downregulation, especially during the heat-shock response [21]. DNMT2-mediated methylation can protect tRNAs from ribonuclease cleavage under stress conditions in Drosophila [56]. Notably, global protein synthesis is reduced in the absence of NSUN2 in vitro and in vivo [80].

Eukaryotic genome stability is associated with repeat regions and transposable elements. It was found that mutation of two RNA m⁵C methyltransferases, DNMT2 and NSUN2, reduced tRNA stability, and altered the expression of the mobile elements, mediating genome integrity in Drosophila [81].

4.1.2. rRNA metabolism

The methylated cytosines in rRNA are fundamental for ribosome biogenesis, thereby influencing protein synthesis. Loss of NSUN4-mediated mitochondrial 12S rRNA methylation alters ribosome subunits assembly [67]. Rcm1 and Nop2 function as above in yeast [68]. NSUN5 deficiency induced decreased proliferation of mammalian cells and reduced body weight in mice, all of which were attributed to a reduction of total protein synthesis by altered ribosomes [65].

4.1.3. mRNA metabolism

m⁵C modifications also participate in mRNA metabolism, including its translation, nuclear-cytoplasmic shuttling, and stability [82]. However, its specific mechanism remains elusive.

m⁵C modifications can directly regulate mRNA translation and indirectly regulate mRNA turnover. NSUN2 methylates IL17a mRNA at cytosine 466 in the coding region, promoting IL17a translation without effects on mRNA half-life [83]. NSUN2 methylates CDK1 at cytosine 1733 in the 3ʹUTR region, upregulating CDK1 translation via elevated translation initiation [84]. NSUN2 methylates p27 at cytosine 64 in the 5ʹ UTR region, repressing p27 translation [85]. NSUN2 methylates SHC mRNA at multiple sites across 5ʹUTR, coding region and 3ʹUTR, elevating SHC expression without effects on mRNA turnover [86]. NSUN2 potentially regulates p53 expression through methylating miR-125b and mRNA 3ʹUTR region [76]. m⁵C modification can also interact with other RNA modifications, mediating mRNA turnover and subsequent translation. Combined m⁵C modification by NSUN2 and m⁶A modification by METTL3/METT14 at the 3ʹUTR of p21 mRNA enhances p21 expression in a model of oxidative stress-induced cellular senescence. Particularly, m⁶A modification facilitates m⁵C modification, indicating that there exists an interaction between m⁶A and m⁵C [39].

As to mRNA nuclear-cytoplasmic shuttling, an important discovery is an m⁵C reader ALYREF, which functions as an mRNA export adaptor. ALYREF-mediated nuclear-cytoplasmic shuttling of mRNA is dependent on m⁵C modification by NSUN2 [18,19]. m⁵C is also essential for mRNA stability. YBX1, an m⁵C reader, recruits ELAVL1 to stabilize HDGF mRNA, therefore promoting the pathogenesis of bladder cancer [17]. In addition, m⁵C modifications stabilize mRNA during the maternal-to-zygotic transition, indicating m⁵C modifications in mRNA may participate in embryonic development. The ‘readers’ include Ybx1 and Pabpc1a in this process (Figure 2)[52].

Figure 2. Distribution patterns and roles of m⁵C in mRNA. m⁵C is distributed among all the regions in mRNA, mainly enriched within 3ʹUTR and around translation start sites. m⁵C plays various roles in mRNA metabolism, including mRNA stability and mRNA nuclear-cytoplasmic shuttling and mRNA translation

4.2. m⁵C Modification and stress response

RCMTs, tRNA stability, and stress response interact with one another. Notably, DNMT2-mediated methylation can protect tRNAs against ribonuclease cleavage under stress conditions in Drosophila [56]. The accumulation of tRNA-derived small RNA fragments was found to activate stress response because of loss of m⁵C [22,23]. Consequently, exposure to oxidative stress lowered NSUN2 expression, reduced methylation at specific tRNA sites and resulted in lower protein synthesis. Based on these facts, NSUN2 has been regarded as a sensor of external stimuli (Figure 3) [80]. Besides, Shanmugam et al. found that stress inhibited DNMT2 expression. The reduced DNMT2 decreased C38 methylation in tRNA^Asp and diminished the translation of Asp-rich proteins. Besides, stress potentially aggravated the effects of RCMTs deletion [87]. In addition to animals, m⁵C modifications in RNA enhance heat stress adaptation in rice. Rice with mutated osnsun2, an RNA m⁵C methyltransferase in rice, displayed heat-stress hypersensitivity, and impaired repair reactions [88].

Figure 3. m⁵C in tRNA metabolism and its association with stress. Loss of NSUN2-mediated m⁵C in tRNA increases the angiogenin-mediated tRNAs cleavage, results in an accumulation of tRNA-derived small RNA fragments, reduces protein translation rates and thus activates stress response

NSUN5-mediated rRNA methylation has been reported to be associated with stress resistance. Lower NSUN5 could elevate stress resistance in yeast, worms, and flies along with increased lifespan. Mechanically, loss of Rcm1 disrupted the structural ribosome conformation and translational fidelity, recruited a distinct group of oxidative stress-responsive mRNAs into polysomes and increased the lifespan and stress resistance [89].

4.3. m⁵C modification and organ development and regeneration

Cytosine methylation in RNA is fundamental for embryonic and organic development at the post-transcriptional level. Double mutants of NSUN2 and DNMT2 showed a synthetic lethal interaction in mice, which is due to impaired protein synthesis [58]. DNMT2 promotes liver, brain, and retina development in zebrafish based on its methylation in cytoplasmatic tRNAs [24]. DNMT2 also plays an important role in haematopoiesis through its methylation on tRNAs. Newborn DNMT2-deficient mice demonstrated delayed endochondral ossification with reduced haematopoietic stem and progenitor cell population. RNA bisulphite sequencing revealed that DNMT2 methylates tRNA^Asp/Gly/Val at C38 in the bone marrow and modulates specific protein synthesis. Furthermore, DNMT2-mediated methylation on tRNA^Asp has a role in the discrimination between Asp and Glu near-cognate codons, contributing to accurate polypeptide synthesis. In DNMT2 deficient cells, the lack of tRNA^Asp methylation reduces the capacity of tRNAs to properly discriminate between Asp and Glu codons, leading to decreased translational fidelity, the production of unfolded proteins, and an aberrant phenotype cell population [90]. DNMT2 also participates in intergenerational transmission of paternally acquired metabolic disorders through methylating sperm small non-coding RNAs in mice [91]. Mutations in NSUN2 cause microcephaly and other neurological abnormalities in mice and human. In neurological abnormalities, loss of m⁵C results in an accumulation of 5′ tRNA fragments, which reduces protein translation rates, activates stress pathways leading to reduced cell size and increases apoptosis of cortical, hippocampal, and striatal neurons [22,23]. NSUN2-mediated tRNA methylation also participates in male germ cell differentiation [92] and balances stem cell self-renewal and differentiation in skin [93]. NSUN3-mediated mitochondrial tRNA^Met methylation is essential for oxidative phosphorylation in mitochondria. However, loss of NSUN3 was revealed to influence ESC differentiation rather than its proliferation, impairing neuroectoderm developments [94]. In addition to animals, m⁵C modification in RNA is essential for plant growth and development. David et al. discovered more than 1,000 m⁵C modification sites in RNA of Arabidopsis thaliana [71]. Deletion of TRM4B, homologue of mammal NSUN2, resulted in loss of m⁵C modification and impaired root development in Arabidopsis thaliana. Besides, Yang et al. found enriched m⁵C modifications in mobile mRNAs in Arabidopsis thaliana, especially mobile mRNA TCTP1, facilitated their transport in distinct root cells and affected root growth [95].

4.4. m⁵C modification and cell proliferation and senescence

Cell proliferation and senescence are regulated by several cell cycle factors. NSUN2 was reported as a downstream of MYC. Its expression was enriched in the S phase, essential for mitosis assembly [96]. NSUN2 can also regulate cell proliferation and senescence through mediating p16 [97], p27, p53, SHC, CDK1 mRNA methylation, and subsequent translation. NSUN2 knockdown increased p27 expression, meanwhile reduced CDK1 expression, and thus repressed cell proliferation and accelerated replicative senescence in human diploid fibroblasts [85]. NSUN2-mediated SHC mRNA methylation elevated its expression and promoted human vascular endothelial cells senescence during oxidative stress or high glucose, probably by increasing ROS levels and activating p38 MAPK [86]. RNA m⁵C modifications also mediated ageing. In human, mouse, and yeast ageing models, NSUN5 was less abundant in aged donors. Specifically, Rcm1, yeast homologue of NSUN5, mediated 28S/25S rRNA methylation, loss of which induced structural changes and altered translating ribosomes. Rcm1-deficient yeast cells showed a decrease in translational fidelity and an increase in recruitment of stress-specific mRNAs to translating ribosomes, which might explain the increased stress resistance and lifespan [89].

4.5. m⁵C modification and nervous system disorders

Mutated NSUN2 was found in some nervous system disorders and intellectual disability, including autosomal recessive intellectual disabilities [25]. Blanco et al. revealed that loss of NSUN2 induced smaller cell size in vitro and caused neuro-developmental defects in vivo due to increased tRNA fragments and higher-sensitivity to stress stimuli [22]. In Dubowitz syndrome, there exists a homozygous splice mutation in NSUN2, which results in decreased m⁵C in tRNA^Asp at C47 and C48 [98].

NSUN5 is deleted in about 95% of patients with Williams-Beuren syndrome (WBS), which is characterized by cognitive disorder and hypotrophic corpus callosum. NSUN5 knockout (NSUN5-KO) mice showed deficits in spatial cognition, impaired cerebral cortex developments [17], and agenesis of corpus callosum [99]. Janin et al. reported DNA methylation-associated epigenetic silencing of NSUN5 as a marker of glioma with long-term survival. Loss of NSUN5 reduced methylation of 28S rRNA at C3782, resulted in impaired protein translation, and stimulated an adaptive translational program for survival against cellular stress [100].

4.6. m⁵C modification and cancer and cancer therapies

Overexpression or higher copy number of NSUN2 is common in various human cancers, especially in oral and colorectal cancers [26]. Since tRNA methylation can protect itself from rapid decay, overexpressed tRNA methyltransferases and its consistent tRNA stability are a potential mechanism for tumour development [101]. NSUN2, reported as a novel downstream of MYC and named as MYC-induced SUN domain-containing protein (Misu), mediated breast cancer cell proliferation and therefore could be utilized as a potential therapy target [96,102]. Similar events were observed in head and neck squamous carcinoma and NSUN2 was related to patient survival [103]. Overexpressed NSUN2 has also been detected in gallbladder carcinoma tissues and cells [104]. Methylation of HDGF mRNA at 3ʹUTR by NSUN2 was proved to enhance its expression and promote pathogenesis in bladder cancer [17]. Chen et al. provided a whole landscape of mRNA methylation in human urothelial carcinoma of the bladder and adjacent normal tissues and found hypermethylated mRNAs in cancer. Hypermethylated mRNAs were mainly enriched in cancer-related pathways [17]. Elsewhere, the NSUN2^lowGF-II^high signature increased the mortality risk in ovarian cancer [19].

The cytosine analogues azacytidine and decitabine, drugs for epigenetic cancer therapy, have always been demonstrated to impede DNA methylation through covalent trapping DNMT. Moreover, azacytidines also inhibit m⁵C modifications in tRNA^Asp at C38 potentially by inhibiting DNMT2 activity. This finding uncovered a novel mechanism of azacytidine in cancer therapy and tRNA methylation may be adopted as a biomarker of azacytidine response in patients [105]. Also, there exists a correlation between T cell activation and NSUN2 mRNA level in head and neck squamous carcinoma based on TCGA database. Higher level of NSUN2 may depict positive effects in immune-checkpoint therapy [106]. Moreover, NSUN2 deletion mediated lower protein translation and increased tRNAs cleavage could increase stem cell and tumour-initiating cell activity and paradoxically render them hypersensitive to stress stimuli, such as chemotherapeutic drugs therapy [107]. Combined deletion of NSUN2 and METTL1, another tRNA methyltransferase, potentiated Hela cells sensitive to 5-fluorouracil (5-FU) due to decreased methylation in tRNA^Val(AAC) and increased destabilization [108].

5. Conclusions and perspectives

RNA modification is a new aspect of post-transcriptional gene regulation essential for RNA metabolism and participates in numerous diseases, including cancer. Although many studies have focused on RNA m⁵C modifications, our knowledge about it remains limited. Existing approaches have identified m⁵C sites in different RNA species and this modification is important for tRNA stability, mRNA translation, ribosome assembly, and other RNA metabolism processes. Altered m⁵C distribution and mutated RNA methyltransferases have been linked to physiological processes, including stress response, organ development, cell proliferation, senescence, and pathological processes, such as nervous system disorders and various cancers. Still, there exist many questions and challenges.

There exists a lack of consistency when mapping m⁵C sites. When mapping m⁵C sites using RNA-BisSeq, the number of putative mRNA m⁵C sites varies by 1,000-fold between studies [69], reflecting defects of existing approaches. Also, the low amount of mRNA poses challenges. Therefore, we need a novel approach to overcome these problems with higher specificity and sensitivity. Moreover, data analysis also needs further improvement. m⁵C modifications are rarely found in some types of RNA including mRNA, lncRNA and so on. Thus, it is very important to develop a better data analysis method to distinguish true m⁵C sites form noise disturbance.

Recent studies have focused on the discovery of RNA m⁵C methyltransferases. YBX1 and ALYREY are the only two ‘readers’ discovered. The biological roles of m⁵C, such as mRNA export and tRNA metabolism, require mediation by ‘readers’. Therefore, discovering more m⁵C readers is essential for further elucidating its biological roles.

The biological role of m⁵C modification in RNA metabolism is still unclear. Recent studies only confirmed that m⁵C in tRNA can protect tRNA from endonucleolytic cleavage, that m⁵C in rRNA is essential for ribosome assembly, and that m⁵C in mRNA may influence mRNA stability and subsequent protein translation. However, the different functions of m⁵C modifications in different regions of mRNA, such as 3ʹUTR, CDS, and 5ʹUTR, are still unknown. Furthermore, RNA-BisSeq has identified thousands of m⁵C sites in mRNAs, whose roles remain elusive.

RNA m⁵C modification participates in several physiological processes and pathological diseases. However, these findings are mainly based on correlation analysis, some of which rely on in vitro researches. Therefore, more researches are required to confirm this correlation in vivo, and further mechanism is also an important part to elucidate. Besides, m⁵C modifications in RNA is a potential marker of drug therapies, especially nucleoside drugs. Altogether, RNA m⁵C modification is a potential parameter in elucidating disease development and may be utilized as a new target or marker of diseases therapies.

Acknowledgments

Author Ya-Qi Gao is responsible for collecting and arranging references, writing, and revising for this manuscript. Corresponding author Jing-Yuan Fang helps organize the structure and main contents of this manuscript. This project was supported in part by grants from the National Natural Science Foundation of China (81421001, 81830081).

Disclosure statement

The authors have declared no competing interests.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [81421001, 81830081].