RNA modifications in cancer

Abstract

Currently, more than 170 modifications have been identified on RNA. Among these RNA modifications, various methylations account for two-thirds of total cases and exist on almost all RNAs. Roles of RNA modifications in cancer are garnering increasing interest. The research on m⁶A RNA methylation in cancer is in full swing at present. However, there are still many other popular RNA modifications involved in the regulation of gene expression post-transcriptionally besides m⁶A RNA methylation. In this review, we focus on several important RNA modifications including m¹A, m⁵C, m⁷G, 2′-O-Me, Ψ and A-to-I editing in cancer, which will provide a new perspective on tumourigenesis by peeking into the complex regulatory network of epigenetic RNA modifications, transcript processing, and protein translation.

Background

Various types of RNA exist in the human transcriptome, including coding RNA (message RNA, mRNA) and non-coding RNA (including tRNA, rRNA, miRNA, lncRNA, circRNA, etc.). Each of these RNAs plays its own role to ensure cells function in an orderly manner [1–7]. There are a large number of chemical modifications on the bases and ribose of RNA that are involved in the post-transcriptional regulation of gene expression.

So far, more than 170 different types of modifications of RNA have been reported [8] and both mRNA and most non-coding RNAs can be modified [9, 10]. Similar to modifications on DNA and histone, RNA modifications can also be written, erased and recognised by corresponding enzymes (writer, eraser and reader). RNA modifications generally affect RNA splicing, stability, localisation, translation, and RNA–RNA/RNA–protein interactions, consequentially modulating life activities. Abnormalities in regulatory enzymes and changes in RNA modification patterns are closely associated with the occurrence and development of various diseases, including cancer [11, 12].

RNA modifications promote or inhibit tumourigenesis by regulating cell proliferation, differentiation, invasion, migration, stemness, metabolism, and drug resistance. Among these numerous RNA modifications, N⁶-methyladenosine (m⁶A), N¹-methyladenosine (m¹A), 5-methylcytidine (m⁵C), N⁷-methylguanosine (m⁷G), 2′-O-methylation (2′-O-Me), pseudouridine (Ψ), and adenosine-to-inosine (A-to-I) editing are notoriously linked to tumourigenesis. The functions, regulatory enzymes, detection methods of m⁶A modification and its role in tumours have been summarised in detail in previous reviews [13, 14]. Therefore, this review focuses on the molecular functions and regulatory enzymes of the other six RNA modifications (m¹A, m⁵C, m⁷G, 2′-O-Me, Ψ and A-to-I) in RNAs (Table 1) and their roles in cancer.

Table 1.

Regulatory enzymes and molecular functions of RNA modifications.

RNA modifications	RNA	Writer	Eraser	Reader	Function	References
m1A	mRNA	TRMT6/TRMT61A, TRMT10C, TRMT61B	ALKBH3	YTHDF2, YTHDF3	Promote/inhibit translation and degradation	[16–18, 97, 100]
	tRNA	TRMT6/TRMT61A, TRMT10C/SDR5C1, TRMT61B	ALKBH1, ALKBH3, FTO		Stabilisation, promote translation initiation and elongation and inhibit tDRs production	[23, 24, 103, 105, 108]
	rRNA	RRP8, TRMT61B			Promote ribosome assembly and translation	[28–30]
m5C	mRNA	DNMT2, NSUN2/4/6	TET1	YBX1, ALYREF, RAD52, LIN28B, FMRP	Nuclear export, stabilisation, promote/inhibit translation	[34, 36–41, 111, 118, 120]
	tRNA	DNMT2, NSUN2/3/6	TET2, ALKBH1		Stabilisation, promote translation efficiency and fidelity	[103, 112, 116, 117, 119, 123]
	rRNA	NSUN1/4/5		YTHDF2	Promote ribosome biogenesis, translation and pre-rRNA processing	[46–49]
m7G	mRNA	RNMT/RAM, METTL1/WDR4		eIF4E family, Ago2	Promote/inhibit translation and nuclear export	[53, 60, 131–133, 140, 143]
	tRNA	METTL1/WDR4			stabilisation, promote translation	[135–139]
	rRNA	WBSCR22/TRMT112			ribosome biogenesis	[141]
Ψ	mRNA	RPUSD2/3/4, PUS1, PUS7, TRUB1/2, H/ACA snoRNPs		MetRS	Promote/inhibit translation, stabilisation, pre-mRNA processing	[166, 167, 169, 171, 180]
	tRNA	RPUSD4, PUS10, PUS7, PUS3, TRUB1/2		MetRS	Stabilisation, promote/inhibit translation, tRFs generation	[168, 172, 174, 175, 180]
	rRNA	RPUSD4, H/ACA snoRNPs			Promote translation fidelity and IRES-dependent translation	[174, 215, 238]
	snRNA	H/ACA snoRNPs		Prp5	Processing of pre-mRNA, RNA–RNA/protein interactions	[181]
2′-O-Me	mRNA	CMTR1/2			Processing of pre-mRNA, translation, stabilisation	[162, 163]
	tRNA	hTrmt13, FTSJ1, TARBP1			Affect tRNA stability, cellular translation, and inhibits tRFs production	[152–156]
	rRNA	C/D-box snoRNPs, MRM1/2/3			Promote IRES-dependent translation	[157, 213, 239]
A-to-I	mRNA	ADAR1, ADAR2			Affect mRNA stability, nucleoplasmic localisation, circRNA formation, and amino acid changes	[76–87]
	miRNA	ADAR1, ADAR2			Affect miRNA biosynthesis and binding to targets	[88–95]

Chemical structures and molecular functions of RNA modifications

RNA methylation

N¹-methyladenosine (m¹A)

N¹-methyladenosine (m¹A) modification adds a methyl group to the N¹ position of adenosine, which blocks Watson–Crick base-pairing and introduces a positive charge on this position, affecting RNA secondary structure and RNA–RNA/RNA–protein interactions. In 1960s, the existence of m¹A modification was first discovered in mammalian and plant RNA [15]. Initially, m¹A was shown to be mainly present in tRNA and rRNA. With the progress of detection techniques, m¹A in mRNA has been gradually confirmed [16]. m¹A modification exists in the coding sequence (CDS), 5′UTR and 3′UTR of mRNA to regulate mRNA stability and translation [16–21]. m¹A is reported on positions 9, 14, 22, 57 and 58 in tRNA [22] to influence tRNA self-assembly, stability, and regulate translation [23–27]. Similarly, m¹A in rRNA is essential for ribosome assembly and regulates cellular translation [28–30].

5-methylcytidine (m⁵C)

5-methylcytidine (m⁵C) is a methylation modification on C⁵ of RNA cytosine ring [31]. This modification was reported as early as 1925, and subsequent studies proved that m⁵C indeed existed in animals and plants [32, 33]. m⁵C is enriched in 3′UTR, the GC-rich sequence, and around the start codon [34, 35] to modulate mRNA stability [36–38], nuclear export [34], translation [39], and DNA damage repair [40, 41]. m⁵C exists on positions 34, 38, 47–50 and 72 of tRNA [42] and mainly concentrated in the anticodon loop, between the variable loop and the TΨC loop, and near the amino acid accepting stem. m⁵C modification in tRNA helps to maintain tRNA stability [43], promote translation [44], and enhance translation accuracy [45]. In rRNA, m⁵C modification is associated with ribosome maturation [46–48] and translation efficiency [49].

N⁷-methylguanosine (m⁷G)

N⁷-methylguanosine (m⁷G) is the methylation of guanine N⁷ and adds a positive charge to the N atom, which affects the structure of RNA through electrostatic and spatial effects, but does not affect Watson–Crick base-pairing [50]. m⁷G is present in the 5′cap structure of mRNA and pre-tRNA [51, 52]. The m⁷G-cap is important for mRNA shearing and processing, nucleocytoplasmic transport, stability, and protein synthesis. m⁷G is also widely present in the interior of mRNA and tRNA [53–55] and on rRNA [56] and miRNA [57], affecting their processing and maturation [57], stability [58], nucleocytoplasmic transport [59], transcription extension and translation efficiency [53, 58, 60, 61].

2′-O-methylation (2′-O-Me)

2′-O-methylation (2′-O-Me) is adding a methyl group (-CH3) to the 2′-OH of RNA ribose. 2′-O-Me is not limited to bases and is collectively referred to Nm (Am, Um, Cm and Gm). The Nm sites are found in a variety of RNAs to affect RNA stability, flexibility, and RNA–protein interactions [62–66]. Nm is present in the mRNA cap structure and interior of mRNA to regulate mRNA stability, translation, and the interaction between mRNA and other RNAs [67–69]. A large number of sites on tRNA can be modified by 2′-O-Me, which is related to tRNA stability and translation efficiency. Among that Gm18, Cm32 and Nm34 are relatively conserved [70]. Similar to Ψ (see below), 2′-O-Me modification in rRNA are also mainly concentrated on important functional sites of the ribosome [71, 72], such as decoding centre (DC), intersubunit bridge and peptidyl transferase centre (PTC).

Other RNA modifications

Pseudouridine (Ψ)

With C³ and C⁶ as an axis, the pyrimidine ring of uridine (U) rotates 180° to form pseudouridine (Ψ). Compared with U, the base of Ψ is connected with ribose through C-C bond instead of N-C bond. Ψ has a more stable base arrangement than U and the N¹ position can act as a proton donor to form additional hydrogen bonds, which helps improve RNA stability [73]. Ψ is commonly present in various RNAs [74], affecting RNA stability, RNA–RNA/RNA–protein interactions, tRNA fragments (tRFs) generation, splicing and translation.

Adenosine-to-inosine (A-to-I)

C⁶ hydrolytic deamination of adenosine (A) produces inosine (I), also known as hypoxanthine which is similar to guanosine (G) in structure. Inosine, with 2 fewer amino groups than guanosine, can be recognised as guanosine (G) and pairs with cytidine (C), leading to misreading of genetic information. Rate of A-to-I editing increases with intracellular acidification [75]. In cancer, A-to-I editing occurs frequently on mRNAs and miRNAs. A-to-I editing on mRNA modulates its stability, nucleocytoplasmic localisation, translation and circRNA formation [76–82]. In addition, A-to-I editing in the coding region of mRNA may lead to amino acid change, resulting in conformational change of protein, further affecting protein stability, protein-protein interaction, and subcellular localisation [76, 83–87]. Of note, A-to-I editing on miRNA affects its production and downstream targets [88–95].

In summary, RNA modifications regulate RNA splicing, stability, translation, nucleoplasmic localisation, base-pairing and RNA–RNA/RNA–protein interactions, further regulating life activities. The functional diversity of RNA modifications is closely related to localisation of modifications, types of RNA and reader proteins. For example, m¹A and m⁷G occur on adenosine and guanosine, respectively, and although they both introduce positive charges, m¹A blocks Watson–Crick base-pairing, whereas m⁷G does not affect Watson–Crick base-pairing, but rather affects the structure of the RNA through electrostatic interactions and spatial effects. 2′-O-Me adds methyl group to ribose and can be present on any base, increasing the hydrophobicity of RNAs, protecting them from nuclease attack as well as limiting the rotational freedom of 3′-phosphate. The pseudouridine (Ψ) and A-to-I editing introduce new bases. Compared to uridine (U), Ψ alters the chemical bond between bases and ribose and has a more stable base stacking arrangement. Compared to adenosine (A), inosine (I) pairs with cytidine (C), leading to the misreading of genetic information.

Regulatory enzymes for RNA modifications

Most modifications are regulated by three types of enzymes, the “writer” which introduces the modification, the “eraser” for the removal of modification, and the “reader” for recognition of modification (Fig. 1). Balanced expression of these enzymes is necessary for the maintenance of normal cellular activities. Tumorigenesis and progression are usually due to dysregulation of these RNA-modifying enzymes, resulting in abnormal RNA modifications that further affect the expressions of oncogenes/tumour suppressor genes.

Fig. 1.

Chemical structures and regulatory enzymes for common RNA modifications

Regulatory enzymes for m¹A

Till now, five writers (TRMT6, TRMT61A, TRMT61B, TRMT10C and RRP8), three erasers (ALKBH1, ALKBH3 and FTO), and four readers (YTHDF1, YTHDF2, YTHDF3 and YTHDC1) have been identified for m¹A modification.

Writers

TRMT6, TRMT61A, TRMT61B and TRMT10C belong to the tRNA methyltransferase protein family. The TRMT6/TRMT61A complex was originally identified as a methyltransferase of cytoplasmic tRNA that recognises the motif GTTCRA and is responsible for m¹A modification at position 58 in the TΨC loop of tRNA [96]. In this complex, TRMT61A has catalytic activity, while TRMT6 binds to RNA. Later, studies found the TRMT6/TRMT61A complex could also catalyse m¹A modification on mRNA [97] and in the seed region of tRFs, disrupting the base-pairing of tRFs with their target genes [98]. TRMT61B resides in mitochondria. Initially, TRMT61B was shown to be responsible for m¹A58 in mt-tRNA [99], and it was later found that TRMT61B was also able to catalyse m¹A modification on mitochondrial mRNA [97] and 16 S rRNA [28]. TRMT10C is an m¹A catalytic enzyme for mt-tRNA and mt-mRNA [100–102]. Specifically, TRMT10C interacts with SDR5C1 to catalyse m¹A modification at position 9 of mt-tRNA [101, 102]. RRP8 (also known as nucleomethylin, NML), a nucleolar factor, mediates m¹A modification on 28 S rRNA of mammals to contribute to the formation of 60 S ribosomal subunit [29, 30].

Erasers

AlkB family proteins ALKBH1, ALKBH3, and FTO act as erasers for m¹A. ALKBH1 depletes m¹A on tRNA to regulate protein translation [23, 103], affect mitochondrial unfolded protein response (UPR^mt) [104], and enhance tRNA cleavage leading to decreased tRNA stability [105]. ALKBH3 is an m¹A demethylase for mRNA [16, 17, 19, 20] and tRNA [24], regulating RNA stability and translation. ALKBH3 removes not only m¹A on RNA, but also 1-meA on DNA to repair these abnormally methylated DNA/RNA [106, 107]. FTO has a wide range of substrates and can remove m⁶A and m⁶Am on mRNA and snRNA as well as m¹A on tRNA [108]. As a demethylase for m¹A, FTO displays strong substrate selectivity and can effectively remove m¹A in loop structures (such as m¹A58 in the TΨC loop of tRNA) rather than m¹A in linear transcripts [108].

Readers

YTH domain family proteins recognise multiple methylation modifications. YTHDF1–3 and YTHDC1, key members of the YTH domain family, are identified as m¹A readers [109]. YTHDF1 promotes translation of m¹A-modified transcripts [109]. YTHDF2 and YTHDF3 recognise m¹A on mRNA and reduce transcript stability [17, 18].

Regulatory enzymes for m⁵C

Writers for m⁵C include NSUN1–7, members of the NOL1/NOP2/SUN (NSUN) domains protein family and DNA methyltransferase (DNMT) homologue DNMT2. Erasers for m⁵C include the TET family (TET1–3) and ALKBH1, which oxidises m⁵C to 5-hydroxymethylcytidine (hm⁵C) or 5-formylcytidine (f⁵C). m⁵C in RNA is recognised by six readers, ALYREF, YBX1, LIN28B, YTHDF2, RAD52 and FMRP.

Writers

NSUN1–7 and DNMT2 utilise S-adenosyl-L-methionine (SAM) as methyl donor to catalyse m⁵C modification. Unlike other members of the family, DNMT2 does not methylate DNA, but mediates m⁵C modification in the anticodon loop of tRNA at position 38 [110] and on mRNA [40, 111]. NSUN1 (NOP2) and NSUN5 (Rcm1) catalyse m⁵C modification on rRNA to promote ribosome synthesis, processing, and global protein synthesis [46, 49]. NSUN2 (Trm4) mediates the methylation of various RNAs, including tRNA [112], mRNA [34, 36] and other ncRNAs [113–115] to regulate RNA maturation, stabilisation, nuclear export, translation, and affect mitochondrial function. NSUN3, a mitochondrial methylase, induces m⁵C modification at position 34 in the anticodon loop of mt-tRNA which is further oxidised into f⁵C34, promoting protein synthesis in mitochondria [116, 117]. NSUN4 mediates m⁵C modification on mt-rRNA and mRNA to enhance mitochondrial ribosome assembly and maturation, and protein translation [47, 118]. NSUN6 locates in the cytoplasm and catalyses cytoplasmic tRNA mC72 modification [119] as well as m⁵C modification in the 3′UTR of mRNA, implying higher translation efficiency [39, 120]. NSUN7 introduces m⁵C modification on enhancer RNA (eRNA), modulating cellular energy metabolism [121].

Erasers

Reported m⁵C erasers include TET1–3 and ALKBH1. TET1–3 belong to the Tet family, a group of DNA dioxygenases that oxidise 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC). Tet enzymes can oxidise m⁵C on RNA to hm⁵C [122]. For example, TET1 has been shown to remove m⁵C in DNA:RNA hybrids at DNA damage sites to promote homologous recombination (HR) [41]. TET2 oxidises m⁵C to hm⁵C in tRNA to promote cellular translation [123]. ALKBH1 is a member of the AlkB family, which functions on demethylation. ALKBH1 shows substrate diversity and can remove m¹A, m⁶A, m⁵C, m³C, etc. [124]. As a dioxygenase, ALKBH1 oxidises m⁵C to hm⁵C or f⁵C on mt-RNA, which is involved in the regulation of mitochondrial translation and the activity of the respiratory chain [103].

Readers

Several proteins including ALYREF, YBX1, LIN28B, YTHDF2, RAD52 and FMRP can recognise m⁵C in RNA. Aly/REF export factor (ALYREF) is the first identified nuclear m⁵C reader, relying on its m⁵C recognition site K171 [34]. ALYREF recognises m⁵C in mRNA and viral RNA to promote mRNA export [60, 125] and improve mRNA stability [126]. Y-box binding protein 1(YBX1) relies on its cold-shock domain (CSD) to recognise m⁵C-modified RNAs in a sequence-dependent manner and enhance the mRNA stability by recruiting RNA stabilisers ELAVL1 or Pabpc1a [36, 37]. YBX1 can also directly improve mRNA stability by recognising m⁵C [127–129]. Lin-28 homologue B (LIN28B) has a conserved cold-shock domain (CSD) similar to YBX1, which recognises m⁵C in mRNA and improves the stability of mRNA [38]. YTHDF2 recognises m⁵C in RNA and participates in pre-rRNA processing and maturation [48]. DNA repair protein RAD52 and Fragile X mental retardation protein (FMRP) recognise m⁵C in mRNA at DNA damage sites to promote HR [40, 41].

Regulatory enzymes for m⁷G

The reported m⁷G writers include three protein complexes, RNMT/RAM, METTL1/WDR4 and WBSCR22/TRMT112. Notably, there is no report that m⁷G is a reversible modification. As for m⁷G readers, the eukaryotic translation initiation factor 4E (eIF4E) family proteins (eIF4E1/eIF4E, eIF4E2/4EHP and eIF4E3) and Ago2 are able to recognise m⁷G in the 5′UTR cap of mRNA.

Writers

m⁷G writers use S-adenosylmethionine (SAM) as methyl donor to add methyl group to N⁷ of guanine. ABD1 of Saccharomyces cerevisiae was the first identified m⁷G-cap writer [130]. Human RNMT, with the same catalytic subunit as ABD1, can form a complex with the cofactor RAM to catalyse methylation of the mRNA 5′cap structure to affect mRNA stability and translation [131–133]. There are specific writers to introduce m⁷G onto mRNAs and other RNAs. Trm8 and Trm82 are reported to be methyltransferases of tRNA m⁷G in yeast, and their homologues in humans are METTL1/WDR4 complexes [134]. METTL1/WDR4 mediate m⁷G modification in tRNA [135–139], mRNA [53, 140] and miRNA [57] to promote tRNA expression, mRNA translation efficiency, and miRNA processing. The Bud23–Trm112 complex in Saccharomyces cerevisiae, catalysing m⁷G at site 1575 of 18 S rRNA, is essential for 18 S rRNA processing and maturation [141]. The human homologue of Bud23 is WBSCR22 (MERM1), which mediates m⁷G modification at site 1639 in human 18 S rRNA [142].

Readers

The eIF4E family contains three members: eIF4E1 (eIF4E), eIF4E2 (4EHP) and eIF4E3, which bind to m⁷G in the 5′UTR cap structure of mRNA [143]. In general, eIF4E localised in the cytoplasm enhances translation, while eIF4E localised in the nucleus promotes nuclear export [144, 145]. 4EHP has been considered as a translation repressor [146]. However, under some specific environmental stimuli (such as hypoxia) or in combination with some cofactors (threonyl-tRNA synthetase (TRS), ariadne homologue 1 (ARIH1)), 4EHP can also activate translation [147–149]. Under normal physiological conditions, eIF4E3 competes with eIF4E1 for the same transcript and inhibits mRNA export and translation [150]. However, eIF4E3 can also regulate translation reprogramming during stress [151]. Ago2 contains a Cap-binding-like domain to compete with eIF4E for the m⁷G-cap and inhibits translation initiation [60].

Regulatory enzymes for 2′-O-Me

2′-O-Me writers include two types of enzymes: independent proteases and the C/D-box snoRNPs complex (SNORNP). The independent proteases in humans are CCDC76 (hTrmt13), FTSJ1, TARBP1, MRM1, MRM2 (FTSJ2), MRM3 (RNMTL1), FTSJ3, HENMT1 and CMTR1/2. But, no reader or eraser for 2′-O-Me has yet been reported.

Writers

The 2′-O-Me modification in tRNAs is mainly mediated by independent proteases, whereas 2′-O-Me of most rRNAs and snRNAs is mainly catalysed by C/D-box snoRNPs. C/D-box snoRNPs is a complex composed of small RNAs and proteins. Small RNAs have two highly conserved sequences: C-box (RUGAGA) and D-box (CUGA), which target proteins and modified RNAs, respectively [70]. Proteins include SNU13 (NHP2L1), NOP56/58, and fibrillin (FBL), among which FBL possesses catalytic activity [70].

hTrmt13 is a 2′-O-methyltransferase for position 4 of tRNA and may promote translation by inhibiting the production of tRFs [152]. FTSJ1 is the human homologue of S. cerevisiae Trm7 and is responsible for 2′-O-methylation at positions 32 and 34 on some tRNAs [153–155]. TARBP1, the human homologue of trmH (responsible for Gm18 in bacterial tRNA), introduces Gm18 in tRNA and is associated with immune stimulatory properties of tRNA [156]. MRM1, MRM2, and MRM3 are responsible for 2′-O-methylation of 16 S rRNA in mitochondria to maintain normal mitochondrial function and homeostasis [157]. FTSJ3 contains a RNA-methyl-transferase domain (FtsJ) in the N-terminal region [158], which introduces 2′-O-methylation and can be recruited to HIV-1 RNA by TRBP to introduce internal 2′-O-methylation on HIV-1 RNA to evade host immune surveillance [159]. HENMT1 catalyses 2′-O-methylation in miRNA and Piwi-interacting RNA (piRNA), which protects miRNA from exonuclease cleavage, enhances binding of miRNA to AGO2 [160], promotes piRNA stability, and maintains the length and quantity of piRNA [161]. Notably, there are also some Nm modifications that occur in the cap structure, catalysed by CMTR1 and CMTR2 [162, 163] to stabilise RNA and facilitate RNA processing and protein translation.

Regulatory enzymes for Ψ

Ψ modification is catalysed by two types of enzymes, independent proteases or a complex composed of snoRNA/scaRNA and proteins. Pseudouridine synthases include at least six families: TruA, TruB, TruD, RsuA, RluA and PUS10. The RsuA family has not been reported in eukaryotes [164]. There is no study indicating that Ψ is a reversible modification. Reported Ψ readers include the yeast methionine aminoacyl tRNA^Met synthetase (MetRS) and the RNA ATPase Prp5 in yeast.

Writers

There are 12 independent pseudouridine synthases identified in humans: PUS1, PUSL1, PUS3, TRUB1, TRUB2, RPUSD1, RPUSD2, RPUSD3, RPUSD4, PUS7, PUS7L and PUS10 [165]. PUS1/L1/3 belong to the TruA family, TRUB1/2 to the TruB family, PUS7/7 L to the TruD family, and RPUSD1/2/3/4 to the RluA family. PUS1 is a pseudouridine synthase of mRNA that is associated with the processing and maturation of pre-mRNA [166, 167]. PUS3 catalyses pseudouridylation at positions 38 and 39 in the tRNA anticodon loop and is associated with intellectual disability in humans [168]. TruB1 catalyses Ψ modifications in tRNA and mRNA [169, 170]. TruB1 mediates nuclear tRNA Ψ55 modification, contributing to nuclear export of tRNA [170]. TruB2 mainly catalyses Ψ modification at position 55 of mt-tNRA and in mt-mRNA, which contributes to the assembly of mitochondrial ribosomes and protein synthesis [166, 170, 171]. PUS7 is a pseudouridine synthase for tRNA and mRNA that affects tRFs production, cellular translation, and pre-mRNA processing [167, 172, 173]. RPUSD2 catalyses Ψ modification in mRNA [166]. RPUSD3 has been reported as a pseudouridine synthase for mt-mRNA and is associated with mitochondrial protein synthesis [171]. RPUSD4 mediates pseudouridine in mt-tRNA, 16 S rRNA, and pre-mRNA [167, 171, 174] to facilitate pre-mRNA processing and maturation and maintain normal mitochondrial function. PUS10, localised in the cytoplasm, is responsible for pseudouridylation of tRNA at positions 54 and 55 to promote cell growth [175].

H/ACA snoRNPs are the complexes of box H/ACA snoRNA and four proteins, dyskerin (NAP57/DKC1), NOP10, NHP2 and GAR1 to target rRNA and snRNA for pseudouridylation. Only DKC1 is catalytically active and catalyses U-to-Ψ isomerization, and belongs to the TruB family. NOP10 and GAR1 enhance the catalytic activity of DKC1 [176]. In addition, GAR1 contributes to substrate recruitment and product release [177]. NHP2 and Box H/ACA help target RNAs to enter the catalytic centre of H/ACA RNPs [178]. scaRNA is small Cajal body-specific RNA, a specialised snoRNA in Cajal bodies (CBs), which can also guide Ψ modification in conjunction with the above four proteins [179].

Readers

MetRS can recognise Ψ in mRNA and tRNA to regulate the translation level in yeast cells [180]. The RNA ATPase Prp5 in yeast binds to Ψ42 and Ψ44 in U2 snRNA to facilitate splicing processing of pre-mRNA [181].

Regulatory enzymes for A-to-I editing

The known enzymes responsible for A-to-I editing are ADAR1 and ADAR2 of the ADAR family and ADAT family proteins [182]. To our knowledge, till now there is no specific enzyme that recognises this modification, and no report showing that A-to-I editing is reversible.

Writers

Adenosine deaminase includes ADAR family proteins and ADAT family proteins. The ADAR family contains three members: ADAR1, ADAR2 (ADARB1), and ADAR3 (ADARB2), which act on double-stranded RNA (dsRNA). These three proteins are highly conserved during evolution, and all contain a C-terminal catalytic deaminase domain and variable numbers of dsRNA binding domains (dsRBD) [183, 184]. However, only ADAR1 and ADAR2 exhibit catalytic activity and mediate A-to-I editing on dsRNAs. ADAR3 does not have deaminase activity and cannot catalyse A-to-I editing [185], and even competes with ADAR2 for target binding, inhibiting RNA editing [186]. ADAT family proteins, including ADAT1, ADAT2 and ADAT3, are deaminases that catalyse A-to-I editing in tRNA [187].

Roles of RNA modifications in cancer

RNA modifications directly promote/suppress the expression of oncogenes/tumour suppressor genes by affecting the transcription, splicing, stability, translation, nuclear export, and other processes of mRNA, and eventually affects the occurrence and development of cancer. In addition, modifications on non-coding RNAs can consequentially modulate the expression of oncogenes/tumour suppressor genes by affecting ncRNAs splicing, stability, translation, and interaction between RNA. For example, m¹A, m⁵C, m⁷G, 2′-O-Me and Ψ modifications on tRNAs and rRNAs typically regulate the translation of downstream target genes, while m⁵C on lncRNAs, m⁷G, 2′-O-Me and A-to-I editing of miRNAs regulate the expression of target genes by affecting their own synthesis as well as stability or miRNA-mRNA interactions. Notably, A-to-I editing also affects tumourigenesis and cancer progression by directly altering amino acid sequence without affecting RNA expression. Of course, abnormalities in RNA modifications in cancer are closely associated with up- or downregulation of their regulatory enzymes.

m¹A modification in cancer

m¹A and its regulatory enzymes are abnormally expressed in cancer, and could be used as good diagnostic markers [188]. For instance, different m¹A modification patterns are distinguished in ovarian, colon, and oral squamous cell carcinomas based on differences in expression of these regulatory enzymes [189–191]. Individual tumour m¹A modification pattern predicts patient prognosis, cancer stage and grade, and has different immune infiltration characteristics, which will be used for individualised treatment in the future.

Specifically, m¹A modification and its regulatory enzymes play important roles in tumour cell proliferation, apoptosis, invasion and migration, cancer stem cell (CSCs) self-renewal, and metabolism by regulating the stability or translation of oncogenes or tumour suppressor genes [20, 24, 25, 192–194] (Fig. 2a). For example, the TRMT6/TRMT61A complex enhances the translation of peroxisome proliferator-activated receptor (PPARδ) by catalysing m¹A modification on tRNA and promoting cholesterol biosynthesis to initiate the Hedgehog signalling, resulting in self-renewal of liver CSCs and tumourigenesis of hepatocellular carcinoma [25]. ALKBH3, an eraser of m¹A, raises the sensitivity of tRNA to ANG cleavage by removing m¹A, thereby promoting the generation of tDRs to enhance ribosomal assembly and translation or interact with cytochrome c (Cytc) to inhibit apoptosis [24]. In ovarian and breast cancers, upregulation of ALKBH3 leads to m¹A demethylation of cytokine CSF-1 mRNA to enhance its stability, and ultimately boosts tumour invasive capacity [20].

Fig. 2. Roles of m¹A and m⁵C modifications in cancer.

a Role of m¹A modifications in cancer. m¹A on mRNA (e.g., CSF-1, COL1A1/2) affects tumour invasion and metastasis by regulating mRNA stability. m¹A on tRNA affects tumourigenesis and apoptosis by promoting the translation of downstream target genes (such as PPARδ) and inhibiting the generation of tDRs. b Role of m⁵C modifications in cancer. m⁵C modifications on oncogenes (FTH1/FTL, PKM2, TEAD1, E2F5, YY1, RCC2, FOXC2, HDGF, KRT13, PIK3R1, PCYT1A and GRB2) enhance mRNA stability or translation, and then inhibit ferroptosis and promote tumour cell metabolic reprogramming, proliferation, invasion, and migration. The m⁵C modification of CDKN1C mRNA promotes CDKN1C degradation and induces cell proliferation. The m⁵C modifications of lncRNA (lncRNA H19 and lncRNA NKILA) enhance their stability, which further influences tumourigenesis and cancer progression. m⁵C on mRNA of DNA:RNA hybrids at DNA damage sites promotes DNA damage repair by recruiting DNA repair proteins and regulating the decomposition of R-loop.

m⁵C modification in cancer

RNA m⁵C modification is usually highly expressed in cancer tissues and circulating cancer cells [195, 196]. m⁵C affects tumour occurrence and development by regulating RNA function (Fig. 2b). m⁵C modification enhances oncogene mRNA stability and expression, promoting metabolic reprogramming, leading to proliferation, invasion, and migration of tumour cells [36, 38, 126, 128, 129, 197–200]. For example, NSUN2 catalyses m⁵C modification of HDGF [36] and KRT13 [197], which is recognised by YBX1 and stabilises HDGF and KRT13 to promote migration and invasion of bladder cancer cells and ovarian cancer cells, respectively. ALYREF is upregulated in bladder cancer and promotes glycolytic metabolism and cell proliferation by recognising m⁵C in 3′UTR of PKM2 and stabilising PKM2 [126]. NSUN5 promotes FTH1/FTL expression through mediating m⁵C modification in FTH1/FTL, resulting in decreased intracellular Fe²⁺ concentration and ultimately inhibiting cell ferroptosis [200]. Meanwhile, m⁵C also modifies the tumour suppressor genes to prompt their degradation, resulting in lower levels during tumourigenesis. For example, NSUN2 promotes gastric cancer cell proliferation by reducing CDKN1C (P57 Kip2) in an m⁵C-dependent manner [201].

In addition, m⁵C on mRNA of DNA:RNA hybrids at DNA damage sites promotes DNA damage repair by recruiting DNA repair proteins or regulating the decomposition of R-loop. TRDMT1 (DNMT2) induces the formation of m⁵C on mRNA at DNA damage sites, which is recognised by RAD52 or FMRP, and finally promotes HR repair. RAD52 recognises m⁵C and then recruits RAD51 to promote HR [40]. In the late stage of DNA repair, FMRP binds to m⁵C-modified mRNA and recruits TET1 to remove m⁵C, resulting in accelerated decomposition of R-loop (easy to break, causing genome instability), promoting the completion of DNA repair and maintaining genomic stability [41].

The tumour-promoting effect of m⁵C is not limited to mRNA, and many m⁵C-modified ncRNAs are also important for tumourigenesis [127, 202]. For example, NSUN2 enhances the stability of lncRNA NKILA in a YBX1-m⁵C-dependent manner. As a sponge for miR-582-3p, NKILA promotes the expression of YAP1 to promote the development of cholangiocarcinoma [127].

m⁷G modification in cancer

m⁷G modification and its regulatory enzymes affect the development of tumours by modulating cellular translation and miRNA biosynthesis (Fig. 3a). m⁷G-cap is associated with tumourigenesis. Cap methylation of some oncogene mRNAs boosts their nuclear export and translation [203–205]. For example, RNMT enhances the translation of cyclin D1 mRNA by promoting its cap methylation and ultimately promotes mammary epithelial cell transformation [204]. c-Myc activates the WNT/β-catenin signalling pathway by promoting cap methylation of transcripts of the WNT/β-catenin signalling pathway by RNMT [205].

Fig. 3. Role of m⁷G, 2′-O-Me, and Ψ modifications in cancer.

a Role of m⁷G modifications in cancer. mRNA (such as cyclin D1) cap methylation promotes mRNA translation and affects tumourigenesis. The increase of m⁷G modification on tRNA upregulates the translation of some oncogenes (e.g., EGFR/EFEMP1, RPTOR and relevant genes in PI3K/AKT/mTOR and WNT/β-catenin signalling pathways) to promote tumourigenesis, proliferation, invasion, and metastasis, and inhibit cell death. The m⁷G modification on miR-149-3p upregulates miRNA levels and inhibits the expression of oncogene S100A4. The m⁷G modification on let-7e pri-miRNA promotes its processing and further represses some genes associated with migration. b Role of 2′-O-Me modifications in cancer. Upregulation of C/D-box snoRNAs (e.g., SNORD14D, SNORD35A, SNORD42A, SNORD12C and SNORD78) in tumours enhances the translation of some oncogenes (e.g., VEGF, XIAP, EIF4A3 and LAMC2) by inducing 2′-O-Me modifications of rRNA. 2′-O-Me modification of miR-21-5p enhances miRNA stability and interaction with AGO2, which inhibits the expression of oncogene PDCD4 and ultimately promotes apoptosis. c Role of Ψ modifications in cancer. Ψ modifications on ribosomal protein mRNAs (e.g., RPL10A, RPL22L1, RPL34 and RPS3) enhance RNA stability, leading to increased ribosomal protein abundance, and promoting tumour cell proliferation. Ψ modifications on rRNA and tRNA affect tumourigenesis and cancer progression by regulating translation. The Ψ modifications on rRNA enhance IRES-dependent translation of the antioncogene p27 to suppress tumours, and also affects tumourigenesis by enhancing translation efficiency of ribosomes, improving translation fidelity, and enhancing mitochondrial function. Ψ modifications on tRNA affect tumourigenesis and cancer progression by promoting the production of 5′ tRFs or by inhibiting the translation of target genes (e.g., TYK2).

m⁷G modifications in mRNA interior and in some ncRNAs play an increasingly prominent role in cancer. m⁷G modification in tRNA upregulates translations of several oncogenes to promote tumourigenesis [135–139, 206, 207]. For example, METTL1 promotes the translation of EGFR/EFEMP1 by introducing m⁷G modification onto tRNA to promote proliferation, migration, and invasion in bladder cancer [139]. In esophageal squamous cell carcinoma (ESCC), METTL1/WDR4 prompts the expression of tRNA in an m⁷G-dependent manner, leading to abnormal activation of RPTOR, which further promotes the phosphorylation of ULK1 to inhibit autophagy [206]. m⁷G modifications in tumour suppressor miRNA upregulate the expression of miRNA by promoting the processing and maturation of pri-miRNA, resulting in decreased expression of downstream oncogenes [57, 208]. For example, METTL1 inhibits migration of A549 cells by inducing m⁷G in let-7e pri-miRNA to destroy the G-quadruplex structure and promote miRNA processing, which further inhibits several migration-related oncogenes [57].

2′-O-Me modification in cancer

2′-O-Me modification in rRNA and miRNA are closely linked to tumourigenesis (Fig. 3b). 2′-O-Me in rRNA is an important source of ribosomal heterogeneity in human cells, which is a stress response to external signals and dynamically regulates translation during tumourigenesis [209]. For example, upregulation of C/D-box snoRNA (SNORD14D, SNORD35A, and SNORD42A) enhances cellular translation by inducing 2′-O-Me on rRNA, which further promotes proliferation and self-renewal of leukaemia [210, 211]. Under hypoxic conditions, 2′-O-Me modified rRNA undergoes reprogramming and assembles into ribosomes that facilitates IRES recognition, leading to increased IRES-dependent translation of VEGF-C [212]. Many oncogenes such as VEGF and XIAP, have IRES sites that enhance translation in an IRES-dependent manner in the presence of increased 2′-O-Me on rRNA to promote tumourigenesis [212, 213]. The 2′-O-Me in rRNA is mainly introduced by box C/D snoRNP, and the assembly of box C/D snoRNP is a dynamic process which is regulated by a variety of factors. 2′-O-Me in miRNA helps to enhance its own stability which plays a role in cancer. For example, HENMT1 catalyses 2′-O-Me on miR-21-5p in lung cancer, which enhances miRNA stability through interacting with AGO2. Stabilised miR-21-5p inhibits the expression of programmed cell death protein 4 (PDCD4) and hinders apoptosis [160].

Ψ modification in cancer

Ψ modification affects tumourigenesis by regulating translation or RNA stability (Fig. 3c). Ψ modification in rRNA enhances the translation efficiency of ribosome to boost protein synthesis and ultimately promotes cell proliferation and invasion [214]. However, Ψ modification on rRNA can also prompt IRES-dependent translation of some antioncogenes (e.g., p27) and suppress tumourigenesis [215]. Ψ also helps to maintain the accuracy of translation. Abnormal Ψ modifications of certain parts of rRNA lead to mistranslation and tumourigenesis. In addition, Ψ modification in tRNA affects tumourigenesis by modulating cellular translation. For example, overexpression of PUS7 results in increased Ψ modification on tRNA to inhibit TYK2 translation and promote glioblastoma tumourigenesis [216]. Ψ8 in tRNA leads to the generation of 5′tRF (mTOG) and Ψ in mTOG inhibits translation initiation and maintains the balance of stem cell proliferation and differentiation [172]. Ψ modification in mRNA and tRNA also affects tumour progression by regulating self-stability. For example, Ψ-modified mRNAs of ribosomal proteins (RPL10A, RPL22L1, RPL34 and RPS3) are more stable, resulting in increased ribosomal protein abundance, which further promotes tumour cell proliferation [217].

Notably, RNA Ψ modification is critical for mitochondrial function. Ψ on mitochondrial 16 S rRNA and mRNA is essential for mitochondrial protein synthesis as well as oxidative phosphorylation reactions [171]. The enhanced mitochondrial function may help to meet the energy demands of rapid growth of tumour cells.

A-to-I editing in cancer

Compared with normal tissues, A-to-I editing in cancer is diverse, resulting in the heterogeneity of tumour proteomics [218]. Some specific editing sites are associated with cell viability and drug sensitivity [219]. Currently, the GPEdit website provides a global map of the genetics and pharmacogenomics of A-to-I editing in cancer, which helps A-to-I editing to be better utilised in cancer therapy [220]. In cancer, aberrant A-to-I editing mainly occurs on mRNA and miRNA (Fig. 4).

Fig. 4. Role of A-to-I editing in cancer.

a Role of A-to-I editing on mRNA in cancer. In mRNA, A-to-I editing regulates the expression of mRNA (such as GM2A, DHFR, SLC22A3, and CDC14B) to affect tumourigenesis. Meanwhile, A-to-I editing affects protein stability, binding to partners, and subcellular localisation through facilitating conformational changes in proteins such as CDK13, GLI1, Gabra3, AZIN1, COPA, BLCAP and IGFBP7 to regulate cell cycle, angiogenesis, proliferation, invasion, and migration. b A-to-I editing affects tumourigenesis and cancer progression by affecting pri-miRNA processing and maturation, and downstream targets of miRNA. For example, edited pri-miR-142-3p hinders its own processing and maturation, leading to upregulation of STAU1 and ultimately inhibiting cell migration and invasion. Unedited miR-376a* and miR-589-3p and edited miR-200b promote tumourigenesis, while edited miR-376a* and miR-589-3p and unedited miR-200b display tumour suppressor effects. Edited miR378a-3p and miR-379-5p suppressed the expression of oncogenes PAPVA and CD97, respectively. Edited miR-455-5p blocked the inhibitory effect of wild-type miR-455-5p on the tumour suppressor gene CPEB1.

A-to-I editing affects protein synthesis through regulating the stability of mRNA in cancer [79, 80, 82, 221]. For example, A-to-I editing of oncogenes GM2A and DHFR enhances their mRNA stability, leading to elevated protein synthesis and cell proliferation [79, 82]. Meanwhile, A-to-I editing also stabilises the expression of tumour suppressor genes and inhibit tumour progression under certain circumstances. For example, A-to-I edited CDC14B, with increased mRNA and protein expression, promotes the dephosphorylation and degradation of Skp2, leading to p27 and p21 accumulation, which inhibits the transition of cells from G1 phase to S phase and cell proliferation [221].

A-to-I editing may also cause amino acid change, leading to protein conformational change to regulate protein stability, partner binding, and subcellular localisation. The stability of the edited protein is altered, to affect tumourigenesis [85, 86, 222]. For example, in hepatocellular carcinoma, A-to-I edited COPA reduces CAV1 stability, resulting in declined PI3K/AKT/mTOR signalling pathway [85]. A-to-I editing alters the subcellular localisation of some edited proteins [83, 223]. For example, in breast cancer, edited Gabra3 reduces its membrane localisation and inhibits the activation of the AKT signalling pathway, which plays a tumour suppressor role [223]. In addition, A-to-I editing-induced changes in protein conformation may directly regulate expressions of downstream target genes. For example, A-to-I editing of transcriptional activator GLI1 leads to protein conformation change which activates the transcription of some target genes that positively regulate the cell cycle (e.g., Cyclin D1/CDKs) and the EMT pathway (e.g., Snail), resulting in proliferation and metastasis of pancreatic ductal adenocarcinoma [76].

Furthermore, A-to-I editing affects tumourigenesis and progression by influencing pri-miRNA processing and maturation, and the downstream targets of miRNAs. First, A-to-I editing in pri-miRNA hinders its processing and maturation [89, 224]. For example, A-to-I editing of let-7 pri-miRNA hinders its own processing and maturation, resulting in reduction of tumour suppressor let-7 miRNAs [89]. Second, A-to-I editing of miRNAs affects their recognition of downstream targets [91–94]. For example, wild-type miR-376a* plays a tumour-promoting role by targeting tumour suppressor gene RAP2A and promotes glioma cells migration and invasion, whereas A-to-I edited miR-376a* become a tumour suppressor to inhibit the expression of oncogene AMFR [94]. miR-200b plays a tumour suppressor role in a variety of tumours to inhibit the EMT progress by targeting ZEB1 and ZEB2, while the edited miR-200b inhibits the expression of metastasis suppressor LIFR [91, 92].

Conclusions and outlook

Hanahan and Weinberg characterised some common hallmarks of tumour cells that give them an advantage in growth [225–234]. In this review, we summarised six RNA modifications that regulate the proliferation, invasion, migration, metabolism, and apoptosis of tumour cells through affecting RNA processing, stability, nuclear export, and protein translation. In Table 2, we list the promoting or inhibitory effects of writers, readers, and erasers for RNA modification on these hallmarks.

Table 2.

Roles of RNA modification in human cancers.

RNA modification	Cancer	Mechanism	References
m1A	Glioma	TRMT6/TRMT61A upregulation, promote proliferation, migration and invasion	[20, 25, 192, 240, 241]
	Hepatocellular carcinoma	TRMT6/TRMT61A upregulation, drive self-renewal of liver CSCs and tumourigenesis
	Ovarian and cervical cancer	TRMT10C upregulation, promote proliferation and migration
	Breast and ovarian cancer	ALKBH3 upregulation, promote invasion
	Hodgkin lymphoma	ALKBH3 downregulation, promote migration
m5C	Hepatocellular carcinoma	NSUN2 upregulation, promote proliferation, migration, invasion and angiogenesis/ALYREF upregulation, lymph node metastasis	[36, 38, 111, 126, 198, 199, 201, 202, 242]
	Gastric carcinoma	NSUN2 upregulation, promote proliferation, migration and invasion
	Oesophageal squamous cell carcinoma	NSUN2 upregulation, promote proliferation, migration, invasion and poor prognosis
	Squamous cell carcinoma of the hypopharynx	NSUN2 upregulation, promote proliferation and invasion
	Colorectal cancer	NSUN5 upregulation, promote proliferation
	Pancreatic carcinoma	NSUN6 downregulation, inhibit migration
	Bladder cancer	ALYREF upregulation, promote glycolysis and proliferation/YBX1 upregulation, promote proliferation and metastasis
m7G	Intrahepatic cholangiocarcinoma	METTL1/WDR4 upregulation, promote proliferation, migration, invasion, and inhibit apoptosis	[58, 136–139, 206, 208, 236, 237, 243–249]
	Hepatocellular carcinoma	METTL1/WDR4 upregulation, promote proliferation, migration, invasion, and sorafenib resistance
	Lung carcinoma	METTL1/WDR4 upregulation, promote proliferation, migration and invasion, inhibit autophagy
	Oesophageal squamous cell carcinoma	METTL1/WDR4 upregulation, promote proliferation and inhibit apoptosis
	Head and neck squamous cell carcinoma	METTL1/WDR4 upregulation, promote tumour growth and metastasis
	Nasopharyngeal carcinoma	METTL1/WDR4 upregulation, promote tumour growth and metastasis
	Bladder cancer	METTL1 upregulation, promote proliferation, migration and invasion
	Glioma	WBSCR22 upregulation, promote proliferation, migration and invasion
	Pancreatic carcinoma	WBSCR22 downregulation, inhibit proliferation, migration and invasion
	Colon cancer	METTL1 downregulation, inhibit proliferation, migration, invasion and cisplatin resistance/WBSCR22 upregulation, oxaliplatin resistance and stemness maintenance
Ψ	Colorectal carcinoma	DKC1 upregulation, promote proliferation, migration, invasion and angiogenesis/PUS7 upregulation, promote proliferation, migration and invasion	[214, 216, 217, 250–255]
	Lung adenocarcinoma	DKC1 upregulation, promote proliferation
	Suprarenal epithelioma	DKC1 upregulation, promote proliferation, migration, invasion and angiogenesis
	Neuroturbo chargeoma	DKC1 upregulation, promote proliferation
	Glioblastoma	PUS7 upregulation, promote GSC growth and self-renewal
2′-O-Me	Non-samll cell lung carcinoma	FTSJ1/FTSJ2 downregulation, inhibit proliferation, migration and promote cell apoptosis	[213, 256–259]
	Breast cancer	FTSJ3 upregulation, promote proliferation
	Colorectal cancer	NOP58 upregulation, promote proliferation and inhibit cell apoptosis
	Prostatic carcinoma	FBL upregulation, promote proliferation
A-to-I	pancreatic ductal adenocarcinoma	ADAR1 upregulation, promote epithelial–mesenchymal transition and G2/M phase	[76, 79, 85, 92, 93, 106, 186, 222, 223, 260–263]
	Melanoma	ADAR1 downregulation, inhibit tumour growth and metastasis
	Triple-negative breast cancer	ADAR1 upregulation, promote proliferation and tumourigenesis
	Breast cancer	ADAR1 downregulation, inhibit invasion and metastasis
	Oral carcinoma	ADAR1 upregulation, promote proliferation, migration, invasion
	Thyroid carcinoma	ADAR1 upregulation, promote proliferation, migration, invasion
	Hepatic carcinoma	ADAR2 downregulation, inhibit proliferation
	Oesophageal squamous cell carcinoma	ADAR2 downregulation, inhibit tumour growth and promote apoptosis
	Glioma	ADAR3 downregulation, inhibit proliferation and angiogenesis
	Glioblastoma	ADAR1 upregulation, promote proliferation and self-renewal/ ADAR2 downregulation, inhibit proliferation, migration, invasion/ADAR3 upregulation, promote migration and invasion

Although RNA modification and its regulatory enzymes are abnormally expressed in a variety of cancers, most of the researches of these writers, erasers, and readers remains on tumour proliferation, invasion, and migration. Few studies investigate cancer immune escape, angiogenesis, and metabolism, which should be the focus of future work. In addition, the mechanism for up/downregulation of these regulatory enzymes remains elusive. It would be of great interest to know how hypoxia, virus infection, drug stimulation and other stress conditions [126, 200, 235] modulate the expression of these enzymes, which further affects gene expression through RNA modification, and ultimately impacts tumourigenesis. Elucidating the mechanism of up-/downregulation of these enzymes will help us to better prevent tumourigenesis. Furthermore, RNA modification is introduced by writer, removed by eraser, and recognised by reader to play a corresponding role. Writers for various RNA modifications have been extensively studied, but eraser and reader for some modifications have not been identified yet, which warrants further investigation.

Of note, due to the existence of tumour heterogeneity, it may not be accurate to designate a certain RNA modification or regulatory enzyme as a biomarker for a specific cancer. For example, m⁷G writer WBSCR22 promotes glioma growth whereas inhibits the proliferation of pancreatic carcinoma cells [236, 237]. In this direction, single-cell sequencing technology may aid in understanding the anomaly of RNA modifications and regulatory enzymes in different tumour subpopulations or tumour microenvironment immune cells for more accurate tumour diagnosis and treatment.

Studying the specific mechanism of RNA modification in cancers is only the first step. Importantly, it is promising to diagnose and treat cancers through exploring abnormal RNA modifications and its regulatory enzymes. For instance, thiram (TRMT6/TRMT61A complex inhibitor) [25] has been found effectively inhibits liver cancer growth. However, the research on small molecule drugs targeting RNA regulatory enzymes and modified RNA is still in its infancy. RNA modification targeting molecules in conjunction with other cancer therapies such as chemotherapy or immunotherapy may have prominent prospects for the treatment of cancers resistant to current therapeutics.

Author contributions

TQL, LLY, WYM, WP, HXC, OYJW, FCM, LZ, WFY, GC, ZM, LQJ and WH collected the relevant literature and drafted the manuscript. XB, JWH, LGY, ZZY and XW participated in the inception and revision of the manuscript. All authors read and approved the final manuscript.

Funding

This study was supported by grants from the National Natural Science Foundation of China (U21A20382, U20A20367 and 82072374), the Overseas Expertise Introduction Project for Discipline Innovation (BP1221008), the Natural Science Foundation of Hunan Province (2021JJ30897).

Data availability

Not applicable.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Not applicable.

Consent for publication

All authors agreed to publication.