Introduction
The epitranscriptome landscape is very complex, with more than 170 different types of chemical modifications of RNA described to date to decorate coding and non-coding RNAs (ncRNAs) [
1]. Their occurrence has been well documented for over 50 years, however their function remains still widely unknown [
2]. Thus, while known since the emergence of molecular biology, RNA modifications were only coined as the “epitranscriptome” in 2015. The study of the function of these modifications is now emerging and has shown to have big implications in human pathologies [
3,
4]. For example, the role of 6-methyladenosine (m
6A), the most abundant and better characterized internal modification in messenger RNA (mRNA), is to regulate embryonic stem cells and cancer cells self-renewal and to favour survival upon heat shock or DNA damage [
5‐
7]. In addition to the roles of m
6A modification in mRNAs, adenosine methylation is also found in non-coding RNAs, such as microRNAs (miRNAs), long non-coding RNAs (lncRNAs), and circular RNAs (circRNAs) regulating their biogenesis and function [
8‐
15]. We now begin to appreciate the plethora of molecular processes that are finely regulated by RNA modifications ranging from RNA metabolism, decay, splicing or translation, localization, stability, turnover, binding to RNA binding proteins (RBPs) or other RNAs, and thereby diversifying genetic information. Similar to epigenetics, groups of proteins have been identified that specifically “write” (catalyse the deposition of a specific modification), “erase” (catalyse the removal of a specific modification), and “read” (recognize and bind modified nucleotides) thereby affecting the fate of RNA. Other modifications have been recently documented in mRNA including N6,2′-O-dimethyladenosine (m
6Am), 5-methylcytosine (m
5C), 5-hydroxymethylcytosine (hm
5C), pseudouridine (Ψ), 1-methyladenosine (m
1A) or 2'-O-ribose methylation, although their molecular functions remain still widely unknown [
5,
16].
RNA modifications are also present in other regulatory ncRNAs, in fact the most modified RNAs are transfer RNA (tRNAs) and ribosomal RNAs (rRNAs) and their modifications shape protein synthesis efficiency and fidelity. More than 100 modifications have been described for tRNA, being the anticodon loop one hotspot of modifications and play key roles in accurate and efficient decoding in translation [
17]. In rRNA, most modifications cluster around functional sites including the decoding site and the peptidyl transfer centre (PTC), suggesting their functional relevance in regulating protein synthesis [
18,
19]. Studies in humans, yeast, and bacteria have shown that dynamic deposition of these modifications in rRNA regulate cell growth, and drug and stress sensitivity by fine-tuning translation and is a very conserved mechanism [
17,
20‐
22]. For instance, in yeast, flies, worms, and humans, alterations of m
5C levels in rRNA favours the translation of stress response-decoding transcripts in order to increase survival [
23,
24]. Occurrence of Ψ residues increases in mRNA in yeast under starvation and heat shock [
25‐
27]. And lack of 2′-O-ribose methylations in rRNAs decreases efficient translation and affects growth and sensitivity to antibiotics [
28]. Similarly, the overall levels of tRNA modifications change to reprogram protein translation by changing codon usage [
29‐
31].
The deposition of RNA modifications is dynamic, and thereby allows rapid cellular responses to environmental signals [
16,
25,
31‐
34]. The ability to adapt to changing microenvironments such as that of stress or chemotherapeutic drugs is crucial to ensure survival of tumour cells, indicating that RNA modifications could play important roles in cancer. Historically, cancer has been considered fundamentally as a disease characterized by stepwise accumulation of genetic or epigenetic alterations of different oncogenes and tumour suppressor genes. However, compelling evidence indicates that epitranscriptomics could also play a fundamental role in this pathology. Through its ability to modulate many processes of RNA metabolism, dynamic RNA modifications have been shown to be important emerging regulators in cancer [
3,
33,
35‐
38]. Although RNA modifications are not generally considered cancer drivers, cumulative evidence shows that their aberrant expression is functionally related to survival, proliferation, self-renewal, differentiation, stress adaptation, invasion, and resistance to therapy, all of which are hallmarks of cancer [
24,
33,
35,
37,
39‐
43]. For example, dynamic changes for multiple RNA modifications can be observed in the urine of cancer patients [
44]. Most striking it has been the extraordinary enlargement of experimental evidence that implicates alterations in the expression of m
6A writers, erasers or readers are associated with increased risk of obesity and diabetes, infertility and with tumour-suppressive or tumour-promoting scenarios [
3,
45]. Other RNA modifying enzymes have been found to be altered in cancer. For example, in an aggressive breast cancer cell lines, 2′-O methylation appeared to be hypermodified in rRNA and correlated with altered translation [
46]. Mutations in the rRNA pseudouridine synthase DKC1, cause X-linked dyskeratosis congenita (X-DC) characterized by impaired translation, hematopoietic stem cells differentiation failure and increased cancer susceptibility [
47]. Alterations in tRNA modifications have been also reported in cancer including m
5C or 5-methoxycarbonylmethyluridine (mcm
5U) and correlate with altered protein translation [
33,
48‐
51]. All these studies show that aberrant RNA modifications contribute to proliferation, self-renewal, migration, stress adaptation and survival of cancer cells and suggest that targeting aberrant posttranscriptional modifications in cancer cells may hold promise as an efficient therapy for tumours [
52].
In this review we will discuss the molecular and cellular functions of RNA modifications in modulating gene expression programmes, with a focus on their roles in cancer. We further summarize here recent studies that elucidate the therapeutic potential of targeting their aberrant deposition in cancer. We will focus our review article on m6A, m5C and Ψ in coding and non-coding RNAs as notable examples due to the advances in our understanding of the role of these epitranscriptomic marks in cellular functions including proliferation, self-renewal, survival to stress or migration. In addition, expression alterations or mutations in m6A, m5C and Ψ depositing machineries have been documented in cancer.
Conclusion
RNA modifications have emerged as critical posttranscriptional regulators of gene expression programmes. We start to appreciate the functional networks that the epitranscriptome interacts with, ranging from metabolisms [
31] to epigenetics and chromatin remodelling [
24,
216] or the immune system [
116]. Despite the progress, most studies have focused on the molecular and physiological functions of only one mark, 6-methyladenosine on mRNA, however the epitranscriptome embraces over 170 RNA chemical modifications that decorate coding and ncRNAs, several other posttranscriptional RNA processing events, and RNA binding proteins that may be as well modified as histones in DNA [
1]. Thus, association studies of hundreds of other RNA marks on coding but also ncRNAs such as tRNA or rRNA remain to be explored.
To achieve this, we first need to develop system-wide methods and tools for rapid and quantitative detection of RNA modifications. Most of the stablished methods rely on next-generation sequencing and, as such, they are typically blind to nucleotide modifications. Consequently, indirect methods are required that are based on immunoprecipitation techniques using specific antibodies, or enzymatic methods and chemical labelling and unique base-modification properties of RNA pairing [
305]. These methodologies have allowed us to catalogue and identify endless modifications with high precision and at nucleotide resolution, yet these strategies have some limitations, reproducibility rate is low due to technical limitations and poor computational methods. For example, antibodies used to recognize modifications such as m
6A still exhibit non-specific binding and can bind to N
6,2′-O-dimethyladenosine (m
6Am) too [
59]. In fact, all antibody-related approaches suffer from poorly characterized, and thus unpredictable specificity of the antibody used for enrichment [
306]. To overcome these limitations, novel detection methods have been introduced such as DART-seq (deamination adjacent to RNA modification targets), an antibody-free method for detecting m
6A sites, where the cytidine deaminase APOBEC1 is fused to the m
6A-binding YTH domain [
307]. Yet, this methodology is also limited since it allows to simultaneously map only RNAs bound to YTH domain containing readers. For m
5C detection, bisulphite-RNA sequencing is the gold standard, yet the reproducibility is low, especially in low abundant and unstable RNA such as mRNAs [
16]. The large initial amount of RNA required to compensate for the high losses caused by the treatment, the resistance to C-U conversion from neighbouring modifications or double-stranded sequences, the inability to differentiate from other modifications protecting the C from the conversation such as 5-hydroxymethylation or 3-methylcytosine, and poor computational analysis are the most common found difficulties [
16]. Yet, careful assessment of C-U conversion together with development of statistically robust bioinformatic tools that highly refined for data analysis are still generating contradictory results [
155,
156]. Regarding the detection of pseudouridine, several labs have developed methodologies that rely only the chemical treatment of RNAs with CMCT, resulting in little overlap of pseudouridine sites on mRNA from the different studies [
25,
26,
221]. Thus, despite to the variety of the established techniques, the reality is that there is currently no generic and precise method for mapping and quantifying modifications in RNA. In addition, current methods are complex and lack of single molecule resolution. In this regard, the emerging third-generation sequencing technologies, such as the platforms provided by Oxford Nanopore Technologies (ONT) and Pacific Biosciences (PacBio) have been proposed as a new means to directly detect RNA modifications [
308]. RNA modifications can be detected by kinetic changes of reverse transcriptases when encountering a modified nucleotide (PacBio) [
309]. Or by current changes as the native RNA molecule is pulled through a membrane pore (ONT) [
310]. Although the detection of modifications using ONT direct RNA sequencing is already a reality [
311], yet current efforts have not yielded an efficient and accurate RNA modification detection algorithm, largely due to the challenges in the alignment and re-squiggling of RNA current intensities. But emerging alternative base-calling strategies such as
EpiNano algorithms which identifies m
6A from RNA reads with an overall accuracy of ~90%, open new avenues to explore additional RNA modifications in the future [
312].
The dynamic expression patterns of writer, reader and eraser proteins complicate the identification of the precise functional consequences of aberrant deposition of modifications on RNA metabolism. Thus uncovering the complete repertoire of cellular RNA substrates and the writers, readers, and erasers will unveil how the intricate network of epitranscriptomic events can converge into similar cellular processes and showing how their unbalanced deposition may lead to pathologies. For example,
HIF1A mRNA is stabilized by m
6A deposition, while in melanoma cells high levels of HIF1α protein are maintained by modifications at the U34 wobble position of tRNAs [
51,
143]. Furthermore, it will be essential to understand the factors or signals that determine the specificity of the RNA modification writers, readers, and erasers and how these proteins are regulated in different cell types. For instance, METTL16 is sensitive to SAM levels and can regulate its synthesis by modifying the SAM synthase gene MAT2A as a feedback loop mechanism [
313]. In addition, we need to develop innovative technologies for precise manipulation of the epitranscriptome and functional assays that enables to understand their dynamic mechanisms of action of each modified RNA, since depletion of individual RNA modifier may not be sufficient to comprehensively understand their roles. For example, while the main target for NSUN2 are tRNAs, it still remains unclear whether the phenotypic changes seeing upon
Nsun2 deletion are caused by decrease methylation of tRNAs, or other RNA substrates may as well contribute to the observed phenotype [
42,
168,
179,
181,
182].
We start to appreciate the wide range of functional consequences of the aberrant deposition of RNA modifications in human diseases including cancer. For example, aberrant deposition of tRNA modifications, including Ψ and m
5C, leads to perturbed accumulation of tRNA fragments, a novel class of functional ncRNAs whose role is associated with aberrant protein synthesis rates and reprograming of the translational machinery in tissue and cancer stem cell populations [
33,
230]. These findings support the current view that balanced protein synthesis and tight control of mRNA translation is central to cellular processes involved in tumorigenesis [
314] and highlights a key role for tRNAs and tRNAs fragments in tumourigenic processes [
315]. Yet, our understanding on how specific species of tRNA fragments govern these processes and the clinical implications of their aberrant expression are still very limited. Future studies will be necessary to decipher the molecular bases of the translation reprograming driven by tRNA fragments. Additional work will be necessary to differentiate the contribution of global changes in tRNA pools versus specific population tRNA-derived fragments in the tumour promoting effect of aberrant protein synthesis [
315]. More importantly, given their association to cancer progression, their therapeutic potential must be explored exploiting the current advent in miRNA-based therapeutics [
316].
Recent advancements in the rapidly evolving field of epitranscriptomics have linked the reprogramming of components of the epitranscriptomic machinery including writers, erasers or readers of the m
6A, m
5C or Ψ to cancer. The extensive number of RNA modifications that constitute the epitranscriptome and the reversible nature of epitranscriptomic aberrations hold promise to the emergence of a promising field of epitranscriptomic therapy, which is already making progress with the recent development of effective inhibitors against m
6A modifying proteins. In the last years, several seminal studies have shed light onto the diagnostic and therapeutic potential of targeting the cancer epitranscriptomic code [
3,
33,
317], yet we are far from understanding whether spatio-temporal modifications of the epitranscriptome can drive tumour initiation [
318]. In addition, their use as therapeutic agents remains a great challenge due to the lack of consistent and consolidated evidence on the oncogenic or tumour suppressive nature of for example the aberrant deposition of m
6A. The inconsistent evidence may reflect the precise functional outcome of the RNA modification on each modified RNA type, their crosstalk with other active signalling processes and the dynamic nature of the epitranscriptome. Identifying accurate epitranscriptome biomarkers and defining the oncogenic or tumour suppressive effect of a given aberrant modification within a specific cellular context, cell type, cellular proliferative capacity or tumour microenvironment will guide finding the exact molecular targets to develop selective and effective therapies for a given tumour type.
Though aberrant expression of several methylases and Ψ synthetases has now been described in cancer, it remains unclear whether they could be efficient targets for cancer therapy. Thus, the precise contribution of methylases and Ψ synthetases to tumour initiation, growth, metastasis and resistance need to be further investigated. It remains unclear the RNA target specificity of each enzyme and how specific modified targets can contribute to the malignant phenotype. In addition, little is known on the dynamics of the deposition of these modifications, their erasers and how the binding to their readers influence the RNA metabolism and tumour cell fate. Yet, the availability of the 3D structure of most of these enzymes and the fact that potent and selective inhibitors have been found [
52], it is reasonable to expect that inhibition of these enzymes is achievable. These structures can provide the basis for structure-guided drug design which in combination with computational tools can be powerful resources for the development of RNA modifying enzymes inhibitors. Few small molecule inhibitors have been described that can target specifically m
6A erasers, yet none of them have reached clinical stages [
52]. For m
5C methylases, azacytidine and decitabine (5-aza2′-deoxycytidine), which are cytidine analogues and can inhibit any cytosine-5 methylase, have been approved for clinical use in haematological malignancies [
319]. However, their use should be taken carefully due to the lack of specificity of these inhibitors that can inhibit both RNA or DNA cytosine methylases. The validation of these enzymes and their modifications as good pharmacological targets will require the discovery of potent, selective, cell-permeable inhibitors to determine the therapeutic benefit and potential risks associated with inhibiting these enzymes.
Treatment failure in certain settings has been attributed to the presence of sub-populations of cancer stem cells or persister cells which are intrinsically resistant to many therapeutic approaches [
320]. Given that m
6A, m
5C or Ψ regulate cell survival to stress in many settings and stem cell functions, targeting them represent a very promising opportunity to specifically target these cells populations and reduce chemoresistance and recurrence. Whether other epitranscriptomics changes can lead to drug resistance is not yet understood. This clearly opens the opportunity to explore novel avenues to develop diagnostic tools based on epitranscriptomic signatures that allow for better patient stratification. In addition, combinatorial therapeutic strategies with the potential to re-establish the normal epitranscriptomic landscape or inhibit survival signalling pathways are promising strategies to specifically eliminate cancer stem cells. Combinatorial therapies that target independent pathways may be a better option as the possibilities for the development of tumour drug resistance may be more limited. Understanding the machineries and factors that introduce, remove, or read RNA modifications will allow the development of novel drugs with pharmaceutical value, not only for cancer but other complex human pathologies that have been linked to aberrant deposition of RNA modifications such as diabetes, neurological, immune and mitochondrial-linked disorders [
47].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.