4. DEREGULATED EPITRANSCRIPTOME IN CANCER.
Given the importance of RNA modifications in regulating RNA life cycle and their role in cell fate and normal human development (Esteller and Pandolfi, 2017; Roundtree et al., 2017a; Morena et al., 2018), it is therefore not surprising that abnormal expression of the epitranscriptome lead to human diseases. In the last years, a number of studies have revealed that deregulated epitranscriptomes are associated with human pathologies, mainly cancer (but not limited to) (Table 1 ) through the deregulation of main tumourogenic pathways, including stem cell differentiation, cell invasion, immune responses, tissue renewal, viral infection or angiogenesis, among others. In this section, we will summarize the major evidences of altered epitranscriptome associated with cancer. This is a growth area with major interest, so it is foreseeable that it will open up a wide range of possibilities for oncology research.
4.1. Alterations of N6-Methyladenosine contents in cancer. The study of m6A dysregulation has been the main focus in oncology. The expression of m6A methyltransferases is frequently altered in cancer, and the functional consequences could be compatible with oncogenic but also tumour suppressor properties depending on the tumour type (Barbieri and Kouzarides, 2020; Rosselló-Tortella et al., 2020). Although the supporting evidences are still limited, this dual role in cancer seems to be determined by the cancer-specific downstream targets of m6A-related enzymes (Zheng et al., 2019).
An oncogenic role for METTL3 has been proposed in AML (Barbieri et al., 2017), clear cell renal cell carcinoma (Wang et al., 2020), gastric cancer (Yang et al., 2020) or pancreatic cancer (Taketo et al., 2017), among others. The role of METTL3-METTL14 complex in the AML model has been widely reported (Barbieri et al., 2017; Vu et al., 2017). m6A methylation is essential for the maintenance of crucial mRNAs involved in the self-renewal of haematopoietic stem/progenitor cells, so that METTL3 overexpression contributes to the maintenance of undifferentiated leukemic cells (Barbieri et al., 2017; Vu et al., 2017). Mechanistically, overexpression of METTL3 results in increased translation of oncogenic transcripts such as MYC, BCL2 or PTEN, as firstly demonstrated in in vitro studies performed in the MOLM-13 cell line (Vu et al., 2017). In vivo assays performed in METTL3- knockdown immunodeficient mice result in increased differentiation of leukemic cells and decreased anti-tumoural effect (Barbieri et al., 2017), confirming the oncogenic role of METTL3. A MYC-dependent oncogenic role of METTL14 overexpression has been also described in AML (Weng et al., 2018). Epithelial-mesenchymal transition (EMT), a crucial process for cancer metastasis, has been also associated with METTL3 dysregulation. m6A in Snail CDS causes polysome-mediated translation of Snail mRNA in liver cancer cells (Lin et al., 2019). Moreover, the upregulation of METTL3 and its reader YTHDF1 could be used as a prognosis factor for adverse overall survival of liver cancer patients (Lin et al., 2019). Overexpression of the EMT effectors metallopeptidase 2 (MMP2) and N-cadherin has been also observed in melanoma cells together with increased METTL3 expression (Dahal et al., 2019). Interestingly, the METTL3 mode of action also includes an effect on miRNA processing (Alarcón et al., 2015b). METTL3 promotes the maturation of miRNAs by interacting with the microprocessor protein DGCR8 (Alarcón et al., 2015b). In this model, METTL3 is able to dually modulate oncogenes or tumour suppressor genes by regulating the maturation of multiple miRNAs with pro- or anti-tumoural activity. For example, METTL3 overexpression promotes the maturation of pri-miR221/222, resulting in decreased expression of the tumour suppressor gene PTEN, and leading to the proliferation of bladder cancer (Han et al., 2019b). On the contrary, METTL14- DGCR8 interaction positively modulates the primary miRNA126 process in an m6 A-dependent manner leading to decreased metastatic potential in hepatocellular carcinoma (Ma et al., 2017). Tumour suppressor functions on METTL3- METTL14 complex have been identified (Cui et al., 2017; Liu et al., 2018; Deng et al., 2019). Human endometrial cancer carrying hotspot mutations in METTL14, and consequently reductions in m6A methylation, showed increased proliferation and tumorigenicity. Reductions in m6A methylation lead to decreased expression of the negative AKT regulator PHLPP2 and increased expression of the positive AKT regulator mTORC2 (Liu et al., 2018). METTL3 also acts as tumour-suppressor in colorectal cancer through p38/ERK pathways (Deng et al., 2019)
In the case of altered m6A RNA demethylation in cancer, the first studies provide an oncogenic role for FTO in melanoma (Iles et al., 2013; Yang et al., 2019). FTO decreases m6A methylation and increases the stability of pro-tumorigenic melanoma genes such as PD-1 (PDCD1), CXCR4, and SOX10, in a mechanism dependent on the m6A reader YTHDF2. A role for FTO in the promotion of resistance to immunotherapy (i.e., anti-PD-1) in melanoma therapy has been also demonstrated in mice models (Yang et al., 2019). FTO also promotes tumour progression in AML with t(11q23)/MLL rearrangements, t(15;17)/PML-RARA, FLT3-ITD, and/or NPM1 mutations. FTO demethylase m6A levels in ASB2 and RARA mRNA transcripts leading to inhibition of all-trans-retinoic acid (ATRA)-induced AML cell differentiation and promotion of leukemogenesis (Li et al., 2017b). In contrast, it could have a suppressive effect in IDH1/2-mutant AML tumours. FTO is an α-ketoglutarate dependent dioxygenase that is competitively inhibited by the structurally related oncometabolite R-2-hydroxyglutarate, which aberrantly accumulates in IDH1/2-mutant AML tumours. This FOT inhibition results in an increase of m6A content at specific targets that contribute to leukaemia suppression (Elkashef et al., 2017). In sum, the FTO effect on tumorigenesis strongly depends on the genomic context and the down-stream pathways that are involved.
The role of the m6A demethylase ALKBH5 has been well characterized in a glioblastoma model (Zhang et al., 2017c). ALKBH5 is highly expressed in glioblastoma stem-like cells. ALKBH5 target in glioblastoma is the transcription factor FOXM1, which is crucial for the maintenance of glioblastoma stem-cells properties and self-renewal. As a result of loss of m6A in FOXM1 mRNA transcript, its stability is increased, the FOXM1 expression is enhanced and cell differentiation diminished. Interestingly, ALKBH5 expression is correlated with poor prognosis in glioblastoma (Zhang et al., 2017c). Similarly, ALKBH5 overexpression (stimulated by hypoxia-inducible factor (HIF)-1α- and HIF-2α) in breast cancer demethylated the mRNA transcript of the pluripotency factor Nanog. The gain of Nanog stability favours a breast cancer stem cell phenotype (Zhang et al., 2016).
Finally, as previously mentioned, the m6A signal is interpreted by a set of reader proteins that exert their function in multiple biological pathways. Although the dysregulation of m6A readers should not result in changes in m6A patterns, alterations in the expression levels of these effector proteins could result in changes in the molecular function of the RNA modification. Furthermore, it still needs to be elucidated whether these readers have a crucial role in cancer independently of the m6A-signal. The reader YTHDF1 has a suppression effect of the antigen presentation in dendritic cells facilitating stable neoantigen-specific immunity (Han et al., 2019a). In vivo studies using mice models demonstrated that a loss of YTHDF1 in classical dendritic cells enhanced the cross-presentation of tumour antigens and the cross-priming of CD8+T cells through \souta mechanisms involving the control of mRNA translation of lysosomal cathepsins. Most important, the therapeutic efficacy of PD-L1 checkpoint blockade is enhanced after YTHDF1 abolition, highlighting the potential therapeutic application of YTHDF1 expression in immunotherapy (Han et al., 2019a). YTHDF2 reader is overexpressed in metastatic colorectal cancer, leading to gain of expression of the metastasis-related gene HIF-1α, of tumour cells bothin vitro and in vivo . A potential biomarker role for predicting metastasis has been proposed (Tanabe et al., 2016). Although based on preliminary results, the prediction potential is extended to HNRNPC reader (Liu et al., 2019). YTHDF2 is overexpressed in several subtypes of AML and is required for disease initiation as well as propagation in mouse and human AML (Paris et al., 2019). YTHDF2 decreases the half-life of tumour necrosis factor receptor Tnfrsf2 transcript avoiding apoptosis in leukaemia stem cells promoting their expansion.
4.2. Alterations of N1-Methyladenosine contents in cancer. Most of the efforts for the elucidation of the role of m1A dysregulation in cancer mainly refer to tRNA demethylation. Recently, it was described that loss of m1A contents mediated by ALKBH3 increased the mRNA transcript abundance of colony-stimulating factor (CSF1) promoting cell invasion without affecting cell proliferation or migration in ovarian and breast cancer cells (Woo and Chambers, 2019). This study anticipates a pathological role of m1A dysregulation in mRNA species.
4.3. Alteration of 5-Methylcytosine content in cancer.Consistent with its role in cellular differentiation, alterations in NSUN2 expression has been associated with human cancer progression. Gain of protein expression of NSUN2 was quantified in multiple cancer types, including oesophageal squamous cell carcinoma (ESCC), stomach, liver, pancreas, uterine cervix, prostate, kidney, bladder, thyroid, and breast cancer (Chellamuthu and Gray, 2020). In some cases, NSUN2 expression has potential biomarker applications. Ovarian cancer patients with high NSUN2 expression and low IGF-II expression exhibit higher overall and disease progression-free survival (Yang et al., 2017a). In contrast, NSUN2 upregulation in Head and Neck Squamous Carcinoma (HNSCC) was associated with shorter overall survival and a higher mortality risk (Lu et al., 2018). Interestingly, NSUN2 expression have been proposed as a biomarker for the prediction of response to immunotherapy in HNSCC. The effect could be mediated by an association of NSUN2 expression and T-cell activation (Lu et al., 2020). In a similar manner to m6A or m1A RNA modifications, we are still far away from the understanding of the molecular pathways governing tumorigenesis. An elegant work aimed at identifying 5mC mRNA modifications at single-nucleotide resolution in human bladder carcinoma showed that hypermethylation of m5C mRNAs is highly enriched in well-known cancer-related pathways, including PI3K–AKT35 and ERK–MAPK36, and the oncogene heparin binding growth factor (HDGF), resulting in enhanced mRNAs stabilization (Chen et al., 2019). In addition, authors found a 5mC reader, named Y-box binding protein 1 (YBX1), whose expression is aberrantly increased in bladder cancer, providing new molecular clues on the 5mC dynamism (Chen et al., 2019). A separate study described a role of NSUN2-dependent methylation in the stabilization of the oncogenic lncRNA NMR (LINC01672) in ESCC. As a result of its increased stability, NMR transcript could directly bind to chromatin regulator BPTF, and potentially promote the expression of the metalloproteinase MMP3 and MMP10 by ERK1/2 activation (Li et al., 2018a). Conversely, other NSUN familiy members, such as NSUN5, exert tumor suppressor roles, being epigenetically inactivated in human brain tumors (Janin et al., 2019).
4.4. Pseudouridyne alterations associated with cancer .
Despite pseudouridylation biogenesis not being well understood, it is likely ψ plays a role in various physiological and pathological contexts. Unfortunately, its implication in disease and mode of action has been only partially explored up to now. Furthermore, most of the defects of Ψ modification linked to cancer are mediated by its control of non-mRNA species, mainly rRNA or tRNA.
Altered dyskerin pseudouridine synthase (DKC1) activity has been recognized as a potential trigger for cancer onset in hereditary syndrome-associated tumours and sporadic cancers. DKC1 expression levels have been correlated with tumour progression and poor overall survival in breast cancer, hepatocellular carcinomas, lung, and prostate cancers (Montanaro et al., 2006; Sieron et al., 2009; Penzo et al., 2017). Interestingly, the molecular consequences of dyskerin overexpression in cancer has been linked to the stabilization of the telomerase RNA component (TERC) (Penzo et al., 2015). Additionally, impairment of DKC1 function has been associated with aberrant p53 mRNA translation and p53 inactivation in human breast cancer cells (Montanaro et al., 2010).
The pseudouridine synthase PUS1 has been associated with melanoma and breast cancer through the pseudouridylation activity on its target ncRNA, the steroid receptor RNA activator 1 (SRA1). SRA1-associated PUS1 then binds the nuclear receptor domains of target genes (e.g., retinoic acid receptor-γ) to help establish the transcriptional pre-initiation complex (Zhao et al., 2004). Nevertheless, the molecular mechanisms underlying this phenomenon have not yet been clarified.
On the other hand, a mechanism involving the pseudouridine synthase PUS10 in TRAIL-induced apoptosis was discovered (Jana et al., 2017). PUS10 is exported from the nucleus to the mitochondria in the early stages of TRAIL-induced apoptosis. A feedback loop between PUS10 and caspase 3 is involved: active caspase-3 is required for PUS10 export whilst the movement of PUS10 reciprocally amplifies caspase-3 activity (Jana et al., 2017). Whether the pseudouridine synthase is involved it is still uncertain. Recently, it has been described that PUS10 binds to pre-miRNAs and interacts with the microprocessor DROSHA-DGCR8 complex to promote miRNA biogenesis in multiple cell types. Mechanistically, this process is also independent of the catalytic activity of PUS10 (Song et al., 2020).
4.5. Adenosine- to- inosine editing in cancer .
The overall biological significance of ADARs by affecting both the base pairing properties of mRNAs transcripts and ncRNAs but also altering codons after translation (and proteins) explains why their dysregulation result in multiple human diseases, including cancer. Alterations in ADAR activity has been described for multiple cancers, including lung cancer, hepatocellular carcinoma, chronic myelogenous leukaemia, glioblastoma and melanoma, among others (Chen et al., 2013; Jiang et al., 2013a; Chan et al., 2014; Qin et al., 2014).
In hepatocellular carcinoma (HCC), ADAR1 is fundamental in the earlier stages of tumorigenesis. A-to-I editing of antizyme inhibitor 1 (AZIN1) transcripts lead to a substitution of serine to glycine at residue 367, facilitating the tumorigenic phenotype and increased invasion properties (Chen et al., 2013). This effect is mediated by the prevention of the degradation of the ornithine decarboxylase ODC and cyclin D1 oncoproteins. Additionally A-to-I editing targets for ADAR1 in HCC has been proposed, including FLNB and COPA (Chan et al., 2014). AZIN1- dependent editing, together with FLNB A-to-I edition, is also involved in the pathogenesis of ESCC (Qin et al., 2014). Interestingly, type I interferon, and its associated JAK/STAT pathway, upregulates ADAR1 expression resulting in aberrant A-to-I editing process in ESCC (Zhang et al., 2017b). In contrast, a tumour suppressor role for ADAR2 has been described and downregulation of ADAR2 enzyme has been linked to ESCC progression. In this work, authors demonstrated that ADAR2 catalytic activity is necessary for the edition and stabilization of insulin-like growth factor binding protein 7 (IGFBP7) leading to cell apoptosis and inhibition of tumour growth (Chen et al., 2017). The mechanism by which ADAR1 expression is associated with CML involved the inflammatory pathway. In CML carrying BCR-ABL fusion gene, the expression of IFN-γ pathway genes promotes ADAR1 expression and editing activity (Jiang et al., 2013a). Through in vitro genetic assays, authors demonstrated that over-expression of ADAR1 positively correlated with expression of PU.1 (myeloid transcription factor) inducing a malignant reprogramming of embryonic stem cells (Jiang et al., 2013b). In lung cancer, ADAR1 gene amplification and overexpression has been observed in HNSCC cell lines and primary tumours, and it has been proposed as a biomarker for prediction of poor outcome (Anadón et al., 2016a). Mechanistically, ADAR1 overexpression enhances the editing frequencies of target transcripts such as NEIL1 (a DNA repair gene) and the oncogenic miR-381 (Anadón et al., 2016b). ADAR1- mediated editing of miRNAs (e.g., miR-455-5p) has been also described for metastatic melanoma by a mechanism involving the inhibition of the tumour suppressor gene CPEB1 (Shoshan et al., 2015). In gastric cancer, ADAR1 expression influences the phosphorylation level of crucial players of mTOR signalling pathway (i.e., mTOR, p70S kinase, S6 ribosomal protein) enhancing oncogenesis (Dou et al., 2016), while ADAR2 exerts a tumour suppressor role through the A-to-I editing of the PODXL (podocalyxin-like) transcript (Chan et al., 2016).
Indeed, ADAR2 seems to be more associated with tumour-suppressor properties than ADAR1 in several types of cancers as it was widely described in highly aggressive brain tumours. In high-grade astrocytoma or glioblastoma multiforme (GBM), ADAR2 regulates key cell cycle proteins, such as Skp2, p21 and p27, which control the G1/M phase and inhibits cellular growth (Galeano et al., 2013). A role of ADAR2 in the editing of oncogenic and tumour suppressors (e.g., miRNAs miR221, miR222 and miR-21) in GMB has been revealed (Tomaselli et al., 2015). The contribution of ADAR3 in glioblastoma revealed a potential binding competition between ADAR2 and ADAR3 to target specific transcripts and subsequently regulate their editing activity (Oakes et al., 2017a). Overexpression of ADAR3 in astrocyte and astrocytoma cell lines inhibits RNA editing at the Q/R site of the transcript GRIA2. Most importantly, the relation between ADAR2 and ADAR3 expression contributes to the relative level of GRIA2 editing in primary tumour samples taken from glioblastoma patients (Oakes et al., 2017b).