MS uses nucleases to catalytically degrade RNA sequences into nucleosides, which are then separated and identified based on their mass-to-charge ratio; this approach enables the detection of modified nucleosides. MS is widely used for detecting various RNA modifications; however, it requires markedly purified and concentrated target RNA and has relatively low sensitivity (43). LC-MS/MS partially addresses these limitations, as it can determine the overall abundance of RNA m⁵C modifications in a sample; however, its resolution does not reach the single-nucleotide level (44,45).

RNA-BisSeq is a widely applicable and cost-effective method for the specific analysis of m⁵C RNA. Bisulfite treatment converts unmethylated cytosines in single-stranded RNA to uracil, whereas methylated cytosines remain unchanged. When combined with high-throughput sequencing, this technique can generate a transcriptome-wide m⁵C map with single-nucleotide resolution. However, this method can be influenced by factors such as RNA structure, experimental temperature and other variables, which may lead to false-positive results (16,46).

TAWO sequencing combines peroxotungstate oxidation with TET enzyme oxidation, using peroxotungstate to oxidize f⁵C to trihydroxylated thymine (thT), and thermostable group II intron reverse transcriptase to convert thT to T during cDNA synthesis. TAWO sequencing allows for the direct detection of modified cytosines without affecting unmodified cytosines, thus overcoming the false-positive issue associated with BisSeq (47). However, because TAWO sequencing relies on the conversion of m⁵C, the conversion efficiency in mRNA samples requires further optimization.

Methylated RNA immunoprecipitation sequencing (MeRIP-seq) identifies markedly methylated RNA fragments by using an RNA m⁵C antibody to bind m⁵C-modified RNA, combined with high-throughput sequencing. This method can detect low-abundance m⁵C modification sites and avoids interference from other RNA modifications, although it is notably dependent on the specificity of the m⁵C antibody (48,49). Methylation individual-nucleotide resolution crosslinking and immunoprecipitation (miCLIP) uses an NSUN2 antibody instead of an m⁵C antibody to target the NSUN2 m⁵C transcriptome landscape (50,51). Moreover, 5-azacytidine-mediated RNA immunoprecipitation (5-azaIP) utilizes 5-azacytidine, a cytidine analogue that forms a reversible covalent bond with RNA m⁵C methyltransferases; methyltransferase antibodies are then used to pull down the target RNA. 5-azaIP reduces the non-specific RNA background and can identify enzyme-specific methylation sites with single-nucleotide resolution; however, 5-azacytidine is markedly toxic, and its incorporation efficiency into RNA can influence the results (52,53).

Nanopore direct RNA sequencing (DRS) enables direct interrogation of RNA molecules without the need for cDNA conversion, allowing for the capture of transcript isoforms and the preservation of epitranscriptomic modifications (54,55). As RNA passes through the nanopore, the bases cause characteristic changes in the electrical current, which are then used to infer the corresponding bases from the signal data (56). DRS provides a powerful platform for parallel profiling of various RNA modifications, revealing the complexity of the epitranscriptome (57). However, its implementation poses challenges due to high costs and a reliance on machine learning-based analysis, where methodological variability can limit reproducibility (55); therefore, the development of more robust and integrative analytical tools remains a critical hurdle.

Experimental detection methods provide precise information about RNA modifications; however, they are not without limitations, including operational complexity and high costs. As a result, several models have been developed to predict RNA m⁵C modifications. These include PEA-m⁵C (58), m⁵C-PseDNC (59) and m⁵C-HPCR (60). Additionally, some models, such as m⁵C-Pred-SVM (61) and DeepMRMP (62), are designed to predict specific m⁵C sites. However, the predictive accuracies of these models vary notably across different species, highlighting the need to improve the accuracy and specificity of traditional prediction models. The integration of experimental detection tools with predictive models can lead to more accurate experimental data, ultimately enhancing the precision of research outcomes.

The m⁵C modification in tRNA is among the earliest and most extensively studied aspects of RNA biology. Methylation typically occurs at the junction of the variable loop and the T stem, often involving positions 47–50, with 1–3 cytosine residues (51). In mice, cytosine at position 38 within the anticodon loop of tRNA is another frequent modification site (86). The m⁵C sites in cytoplasmic tRNA are relatively conserved and play a vital role in maintaining tRNA secondary structure stability and regulating translation efficiency. Moreover, the m⁵C modification regulates codon/anticodon pairing, which is essential for the correct loading of amino acids and the prevention of misloading (25).

Among the tRNA methyltransferases, DNMT2 and NSUN enzymes have been extensively studied (23). The deletion of DNMT2 in mouse cells increases the proportion of uncharged tRNA^Asp, resulting in decreased translation efficiency of proteins containing polyaspartic acid (86); this deletion also triggers tRNA fragmentation and decreases steady-state levels (87,88). Additionally, the m⁵C modification safeguards tRNAs against stress-induced endonuclease-mediated fragmentation, thereby ensuring the accurate translation of near-cognate codons. For example, DNMT2-mediated m⁵C modification in tRNA^Asp helps to distinguish it from tRNA^Glu, preventing amino acid misincorporation (89). Studies have also shown that a reduction in DNMT2 methyltransferase activity due to somatic cancer mutations strongly associated with decreased tRNA levels (90,91). Furthermore, the TET2-mediated oxidation of m⁵C in tRNA disrupts the binding of readers to RNA and produces 5-hydroxymethylcytosine (hm⁵C), altering tRNA methylation and thus impacting translation (25). In anaplastic thyroid cancer (ATC), NSUN2 knockdown markedly reduces tRNA m⁵C modification, suggesting it has a role in stabilizing tRNA in ATC, facilitating the transport of amino acids such as leucine and enhancing translation efficiency. Moreover, NSUN2 also contributes to the formation, proliferation and drug resistance of ATC cells (Table I) (92). In addition to influencing codon translation, NSUN2 also regulates the cleavage site of tRNA^Arg in colorectal cancer in an m⁵C-dependent manner under hypoxic conditions and serves a critical role in tumor metastasis (Table I) (93).

Mt-tRNA^Met also undergoes m⁵C modification, which is essential for the expression of the mitochondrial genome. This modification supports cellular energy metabolism and various metabolic pathways and plays notable roles in tumorigenesis and metastasis (94). The mt-tRNA^Met not only decodes the conventional AUG codon but also mediates the incorporation of methionine into the AUA codon during translation initiation (32,95,96). These modifications of mt-tRNA^Met predominantly occur in the anticodon and its surrounding regions, especially at the wobble position. Additionally, f⁵C, an oxidation product of m⁵C, plays a critical role in enabling mt-tRNA^Met to decode the AUA methionine codon during mitochondrial translation (32,97,98). A previous study highlighted that the α-ketoglutarate-dependent dioxygenase ALKBH1 is involved in the biosynthesis of f⁵C at the first position of the anticodon (position 34 of tRNA) in mt-tRNA^Met, thus facilitating this decoding process. ALKBH1 also promotes the formation of hm⁵Cm and f⁵Cm at the same position in cytoplasmic tRNA^Leu (99). The ALKBH1-mediated oxidation of m⁵C has been shown to be vital for translation and mitochondrial function, underscoring its notable role in cellular processes (97). Furthermore, m⁵C modification of mt-tRNA^Met dynamically regulates tumor cell metastasis and invasion by modulating the initiation and maintenance of mt-mRNA translation (100). Depletion of NSUN3 results in marked reductions in the levels of m⁵C and f⁵C at C34 in mt-tRNA^Met, altering mitochondrial morphology and reducing the number of mitochondrial cristae, thereby regulating cellular energy metabolism (101).

In conclusion, the m⁵C modification of tRNA enhances translation efficiency and helps prevent amino acid mismatches. Studies have also revealed that m⁵C modification contributes to the site-specific cleavage of tRNA. Targeting tRNA methylation modifications may serve as a therapeutic target for certain malignancies with poor prognosis, such as ATC. Moreover, due to the aberrant energy metabolism in malignant tumors, especially the increased metabolic plasticity in aggressive and metastatic cancers, inhibiting the m⁵C modification of mt-tRNA^Met presents a promising therapeutic strategy to effectively curb the spread of malignant tumors (94).

A growing body of evidence links alterations in rRNA modification levels and defects in components of the rRNA modification machinery to tumors. rRNAs are extensively modified during their transcription and subsequent maturation in the nucleolus, nucleus and cytoplasm (102,103). The m⁵C modification in rRNA is associated with ribosome synthesis and protein translation; the m⁵C modification stabilizes the RNA structure by promoting base stacking and increasing the thermal stability of hydrogen bonding with guanine. In functionally crucial areas of the ribosome, m⁵C modification helps stabilize rRNA folding, which is essential for efficient ribosome function (42,104,105). Studies have shown that NSUN4 is crucial for the biosynthesis of mitochondrial rRNA; the knockout of NSUN4 results in the loss of methylation at position C911 of mitochondrial rRNA in mouse hearts, suggesting its key role in coordinating mitochondrial ribosome biosynthesis (106,107). Similarly, NSUN5 is necessary for the specific methylation of 28S rRNA in humans and mice, and its absence reduces protein synthesis, underscoring the importance of NSUN5 in ribosome function and translation regulation (108). In hepatocellular carcinoma, overexpression of NSUN5 upregulates the m⁵C level of 28S rRNA, which in turn promotes growth and metastasis (Table I) (109). Additionally, the depletion of YTHDF2, a multifunctional reader, results in a marked increase in m⁵C levels at various sites within rRNA. This alteration may influence translation fidelity, suggesting a potential link between the role of YTHDF2 in recognizing m⁵C and its impact on translation accuracy (42). A number of studies have shown that YTHDF2 is dysregulated in malignant tumors such as bladder cancer, liver cancer, gastric cancer, osteosarcoma and blood system by m⁶A methylation modification (110–115), while, to the best of our knowledge, no studies have investigated the specific mechanism of m⁵C methylation modification in malignant tumors. Further research is expected to uncover the potential biological functions of YTHDF2.

The sncRNAs, which include microRNAs (miRNAs), PIWI-interacting RNAs, small interfering RNAs and tRNA-derived small RNAs (tsRNAs), play vital roles within cells by binding to proteins, regulating transcription factors and participating in genome stability (24). High-throughput next-generation sequencing-based methods, such as bisulphite miRNA sequencing, and an analysis pipeline such as methylation assessment of miRNAs after bisulphite analysis, have revealed widespread m⁵C modifications in miRNAs (116). The miRNAs regulate gene transcription and posttranscriptional regulation through partial base pairing with sequences mainly in the 3′UTRs of target mRNAs, guiding the RNA-induced silencing complex to suppress mRNA translation. The m⁵C modification in miRNA affects its pairing with mRNA, potentially disrupting its gene-silencing activity (117). To the best of our knowledge, few studies have investigated the role of RNA m⁵C methylation in miRNAs in cancer, although some studies have explored DNA methylation. In glioblastoma, miRNA is methylated at cytosine residues through complexes such as DNMT3A/AGO4, which abrogates its inhibitory effect on gene expression and is associated with poor prognosis (118). Additionally, a pivotal study has successfully developed a serum diagnostic signature using m⁵C-modified miRNAs (119). Due to the potential crosstalk between DNA methylation and RNA methylation, studying the role of RNA methylation in miRNAs may lead to notable breakthroughs in tumor treatment and diagnosis (119).

tsRNAs, derived from tRNA, interact with cytoplasmic ribonucleoproteins (RNPs) to form tsRNA-RNP complexes. These complexes can bind to Argonaute family proteins, effectively mediating posttranscriptional gene silencing through mechanisms such as RNA interference or direct inhibition of mRNA translation (120). DNMT2-mediated m⁵C modifications in mice markedly affect sperm tsRNA levels, with the loss of DNMT2 disrupting the transmission of high-fat diet-induced metabolic disorders mediated by sperm sncRNA to offspring (24). It has been demonstrated that tsRNAs can influence translation rates through YBX1. In breast cancer cells, tsRNAs can induce tumor-suppressive effects by replacing YBX1 with oncogenic mRNAs, thereby promoting the degradation of such transcripts (121–123). Thus, the m⁵C modifications in tRNA notably influence tsRNA functionality, and the regulation of protein translation mechanisms by tsRNA represents a promising new avenue of investigation in cancer studies.

The utilization of innovative detection technologies has notably advanced research into m⁵C modifications of sncRNAs. These advancements enhance the understanding of intracellular signaling mechanisms associated with various diseases. These findings open new avenues for targeted sncRNA therapies, offering more precise and effective treatment options.

lncRNAs, a category of ncRNAs >200 nucleotides in length, constitute a substantial part of the mammalian transcriptome (124). lncRNAs participate in various physiological and pathological processes in normal tissues and tumors by interacting with RNA-binding proteins. These processes include chromatin modification, transcriptional activation, transcriptional interference and nuclear transport (125,126). One of the enzymes involved in modifying lncRNAs is NSUN2, which has been linked to the development of various cancers. In gastric cancer, NSUN2 modifies the lncRNA NR_033928, which in turn affects the stability of glutaminase mRNA and participates in metabolic reprogramming (127). Similarly, in cholangiocarcinoma, NSUN2 interacts with the lncRNA NKILA, promoting disease progression (Table I) (128). In the nervous system, lncRNAs are involved in ischemic stroke through mechanisms such as calcium overload, oxidative stress, hypoxia and inflammatory responses (129). Notably, MeRIP-Seq analysis revealed increased methylation levels and frequency in lncRNAs in the middle cerebral artery occlusion model, which simulates human ischemic stroke. These differentially methylated lncRNAs are associated with several notable pathways, including mTOR signaling, Rap1 signal transduction, pyrimidine metabolism, dopamine receptor binding and phosphatidylinositol phosphate kinase activity (130).

The study of m⁵C modifications in lncRNAs within tumors is still in its early stages, with NSUN2 currently being the only gene known to function as a methyltransferase for lncRNAs. As bioinformatics analyses continue to accumulate, a high m⁵C score has been closely associated with the activation of malignant tumor-related pathways and the impairment of immune microenvironment functions. Furthermore, m⁵C-related lncRNAs show promise as biomarkers for malignant tumors and potential therapeutic targets, making this a notable prospective research direction (131).

circRNAs are a class of covalently closed RNA molecules lacking free 5′ and 3′ ends, and most circRNAs are non-coding; however, growing evidence indicates that some can encode functional proteins (132). Although circRNAs play marked roles in the pathogenesis of various diseases, the functional implications of m⁵C methylation on circRNAs remain largely unexplored. The emerging role of m⁵C-modified circRNAs in lung cancer pathogenesis has attracted increasing attention in recent studies. Chen et al (133) reported that NSUN2-mediated m⁵C modification of circFAM190B enhances its stability, suppresses autophagy and ultimately drives tumorigenesis. The study by Wu et al (134) revealed that NSUN4-catalyzed m⁵C modification of circERI3 facilitates its nuclear export, which in turn promotes the development and progression of lung cancer by enhancing mitochondrial energy metabolism. Cai et al (135) identified circRREB1 as an m⁵C-modified circRNA in lung cancer. This modification, mediated by the methyltransferase NSUN2 and recognized by the reader protein ALYREF, activates mitophagy to promote tumor progression. The functional relevance of circRNA m⁵C modification has also been implicated in other malignancies, including esophageal squamous cell carcinoma and breast cancer (Table I) (136,137). These findings demonstrate that m⁵C methylation can regulate broader processes such as autophagy and mitochondrial metabolism through the modification of circRNAs. Given the fundamental role of RNA modifications in physiology, research on m⁵C methylation in ncRNAs has been increasing, extending beyond mRNA, tRNA and rRNA to include insights into the mechanisms of lncRNA and circRNA modification. Future studies should further expand the scope from cancer to other pathological and functional fields, providing new perspectives for diagnosing diseases through non-invasive examinations.