The Emerging Role of Epitranscriptomics in Cancer: Focus on Urological Tumors

Epitranscriptomics has gained ground in recent years, especially after the advent of techniques for accurately studying these mechanisms. Among all modifications occurring in RNA molecules, N6-methyladenosine (m6A) is the most frequent, especially among mRNAs. m6A has been demonstrated to play important roles in many physiological processes and several disease states, including various cancer models (from solid to liquid tumors). Tumor cells’ epitranscriptome is indeed disrupted in a way to promote cancer-prone features, by means of up/downregulating m6A-related players: the so-called writers, readers and erasers. These proteins modulate m6A establishment, removal and determine mRNAs fate, acting in a context-dependent manner, so that a single player may act as an oncogenic signal in one tumor model (methyltransferase like 3 (METTL3) in lung cancer) and as a tumor suppressor in another context (METTL3 in glioblastoma). Despite recent advances, however, little attention has been directed towards urological cancer. By means of a thorough analysis of the publicly available TCGA (The Cancer Genome Atlas) database, we disclosed the most relevant players in four major urogenital neoplasms—kidney, bladder, prostate and testicular cancer—for prognostic, subtype discrimination and survival purposes. In all tumor models assessed, the most promising player was shown to be Vir like m6A methyltransferase associated (VIRMA), which could constitute a potential target for personalized therapies.


RNA Modifications in Brief: From Epigenetics to Epitranscriptomics
In the past few years, RNA modifications have caught the scientific community's attention. Expanding the scope of epigenetics, which comprises a group of chromatin-based mechanisms including chemical and conformational modifications of DNA and/or histones [1], "epitranscriptomics" (also called "RNA Epigenetics" [2]) relates to modifications in RNA molecules, and has emerged as a pivotal player in several biologic and disease processes [3].

m 6 A Modification in Non-Urological Malignancies: Literature Review
The epitranscriptome of cancer cells has been demonstrated to be disrupted [40], and associations with dysregulation of expression of m 6 A-related proteins (i.e., their writer, readers and erasers) have been increasingly found in many neoplasms [41]. It is reasonable to think that, by modifying the epitranscriptome, tumor cells modify the fate of many target transcripts, which might influence many aspects of cancer progression, including growth and proliferation, invasiveness, migration and metastatic spread, stemness maintenance and differentiation, response to immune surveillance and to stress, among others [21]. Again, and in accordance with the context-dependent role of RNA modifications, evidence has shown that both writers and erasers can assume an oncogenic or tumor suppressor role in different tumor models (for instance, the writer METTL3 may act as oncogene in lung adenocarcinoma and as tumor suppressor in glioblastoma) [21,42].
Modifications in m 6 A levels and/or m 6 A-related proteins expression have been found in a broad spectrum of cancer types. Thus, targeting m 6 A regulatory mechanisms might constitute a new form of cancer treatment [43], especially for suppressing cancer stem cells [42,44]. Clinical trials with drugs targeting oncogenic regulators of the epitranscriptome (such as FTO inhibitors like Citrate and R-2-hydroxyglutarate (R-2HG)) are needed and expected for the near future [37,45].
Polymorphisms in intron 1 of FTO have been associated with a higher risk for development of many neoplasms; however, a metanalysis concluded that, except for pancreatic cancer, the risk was mainly due to body mass index (BMI) [46,47]. However, a single-nucleotide polymorphism (SNP) in FTO intron 8 was found to be associated with a higher risk for melanoma [48], and, as for breast cancer (BCa), another SNP in FTO intron 1 was identified as a susceptibility locus for estrogen-negative BCa [49], both not explained by BMI. FTO was also overexpressed in BCa (particularly in human epidermal growth factor receptor 2 (HER2)-positive tumors) and was also shown to play a role in triple-negative, pan-resistant, inflammatory breast cancer cell lines [50,51]. Also in BCa, a link between hypoxia, tumor invasiveness/metastasis and m 6 A has been proposed, with hypoxia-inducible factors (HIFs) leading to increased mRNA expression of the pluripotency factor homeobox transcription factor Nanog (NANOG) (and subsequent BCa stem cells specification) by means of m 6 A demethylation by the eraser ALKBH5 [52,53]. Furthermore, a positive feedback loop involving HBXIP/miR let-7g/METTL3 was reported to promote BCa progression and proliferation [54].
The writer METTL3 was also shown to be upregulated in various solid tumors, including hepatocellular carcinoma, associated with poor prognosis. In this tumor model, METTL3-mediated m 6 A modification targets suppressor of cytokine signaling 2 (SOCS2), promoting its degradation, in a process dependent of YTHDF2 reader [55]. More recently, YTHDF1 proved also to be upregulated in hepatocellular carcinoma, associated with more advanced stages and poorer survival, contrarily to METTL14, which promotes metastatic potential when downregulated [56,57]. In addition, m 6 A and related proteins are implicated in treatment resistance, as shown in pancreatic cancer cells, in which knockdown of the writer METTL3 improved sensitivity to both chemo-and radiation therapy [58], clearly demonstrating the rationale for using treatments targeting m 6 A modulators. Finally, the reader YTHDF2 was shown to display both diagnostic and prognostic value in pancreatic cancer and to regulate the epithelial-to-mesenchymal transition (EMT) phenomenon [59], whereas WTAP was found to promote migration and invasion in cholangiocarcinoma [60].
Concerning colorectal cancer, the reader YTHDF1 seems to be of paramount importance in disease progression, with immunoexpression associating with unfavorable prognosis disease parameters and poorer survival. Again, the knockdown of YTHDF3 sensitized cancer cells to chemotherapy and, additionally, oncogene c-Myc was found to drive YTHDF1 expression [61]. Moreover, and besides the several RNA editing modifications reported, increased expression of the reader YTHDC2 in colorectal cancer promotes metastatic spread by upregulating hypoxia inducible factor 1 subunit α (HIF-1α) [62,63]. FTO overexpression was shown to impact on prognosis in gastric cancer patients, associated with poor differentiation, lymph node metastases, tumor stage and poor survival [64].
In cervical cancer, a lower amount of m 6 A mRNA modification was associated with disease progression and poor prognosis (higher International Federation of Gynecology and Obstetrics (FIGO) stage, recurrence, metastases and survival), and further manipulation of m 6 A levels in cell lines by altering the expression of respective writers and erasers resulted in increasing or decreasing disease aggressiveness, respectively [65]. More recently, it was demonstrated that the eraser FTO is also upregulated in cervical cancer and leads to chemo-and radiation therapy resistance by demethylating the mRNA transcripts of its target, β-catenin [66]. In addition, a recent study in endometrial cancer has elegantly shown that decreased m 6 A caused by a mutation in METTL14 or downregulation of METTL3 ultimately leads to increased proliferation by activating the AKT signaling pathway [67]. m 6 A modification in mRNA of glioblastoma stem cells regulates their capacity of self-renewal and tumorigenesis, with overexpression of writers (METTL3 and METTL14) and downregulation of erasers (FTO and ALKBH5) inhibiting tumor growth. In addition, high levels of the eraser ALKBH5 associated with poor prognostic features and METTL3 associated with radiation therapy resistance [68][69][70]. This finding may be explored as a potential therapeutic target. Moreover, in lung cancer, another aggressive neoplasm, METTL3 was shown to act as an oncogene, inducing tumor growth and proliferation, also promoting translation of important genes such as epidermal growth factor receptor (EGFR) and tafazzin (TAZ) [71]. An interaction between molecules like microRNAs and m 6 A alterations was also depicted, with miR-33a inhibiting lung cancer cells proliferation by targeting METTL3 [72]. m 6 A has been demonstrated to have an impact in biogenesis of hematolymphoid neoplasms, as well. It was shown that mutations in m 6 A-related proteins confer poor prognosis in acute myeloid leukemia (AML) [73]. Mutations in writers (METTL3, METTL14, WTAP, RBM15) promote and maintain leukemogenesis in AML [74][75][76][77][78][79], whereas overexpression of the eraser FTO in AML cell lines also promoted proliferation and decreased apoptosis [80]. Moreover, FTO plays a role in response to all-trans-retinoic acid (ATRA) and, interestingly, D-2-hydroxyglutarate (D2-HG) (the metabolite accumulated in isocitrate dehydrogenase 1 and 2 (IDH1/2)-mutant leukemias (20% of AMLs)) functions as an inhibitor of FTO demethylase, meaning that FTO expression is context-dependent and has to be interpreted according to IDH mutational status [81,82]. A summary of the findings presented in this section is depicted in Table 1.

m 6 A Modifications in Urological Tumors: Analysis of The Cancer Genome Atlas Database
Although RNA modifications have been analyzed in several tumor models in recent years, little attention has been paid to urological cancer. One of our main research goals is to uncover and characterize new epigenetic modifiers in urological malignancies, to be applied in diagnosis, prognosis and disease monitoring. In this line, we performed an in silico analysis of the publicly available The Cancer Genome Atlas (TCGA) database regarding m 6 A-related proteins (writers, erasers and readers) in the four main urological cancers: bladder (BlCa), kidney (KCa), prostate (PCa) and testicular cancer. For that purpose, the online resource cBioPortal for Cancer Genomics [83] was used, with the user-defined entry gene set "METTL3, METTL14, METTL4, METTL16, WTAP, VIRMA, RBM15, RBM15B, FTO, ALKBH5, YTHDF1, YTHDF2, YTHDF3, YTHDC1, YTHDC2, EIF3A, HNRNPC and HNRNPA2B1". Statistical analysis with the available data was performed with Microsoft Excel 2016, (Microsoft, Redmond, Washington, USA), GraphPad Prism 6 (Prism, San Diego, California, USA) and IBM SPSS Statistics v.24 (Armonk, NY, USA). Distribution of continuous variables between groups was compared using the nonparametric Mann-Whitney test. Correlations between continuous variables were assessed with Spearman's non-parametric correlation test. Co-occurrence/mutual exclusivity of alterations in pairs of genes was estimated with odds ratio (OR). Biomarker performance was assessed through receiver operating characteristics (ROC) curve construction. In brief, for each transcript, an ROC curve was constructed plotting sensitivity (true positive) against 1-specificity (false positive). A cut-off was established by ROC curve analysis (sensitivity + (1-specificity)), to maximize both sensitivity and specificity. In addition, area under the curve (AUC) and biomarker performance parameters, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy, were ascertained. Survival curves were plotted using the Kaplan-Meier method and log rank test was used for survival analysis. A p-value equal or inferior to 0.05 was considered statistically significant.

Prostate Cancer
PCa is a major public health concern in male gender mainly due to the growth and aging of the global population [84]. It is a highly prevalent malignancy, being the second most common cancer and the fifth leading cause of death from cancer in men, mostly due to aggressive and metastatic disease [85]. This neoplasia is usually clinically silent until extra-prostatic invasion or metastization occur, being a complex and heterogeneous disease, ranging from clinically indolent to highly aggressive [86,87]. At its earliest stages, PCa is sensitive to androgen-deprivation therapy, which is the mainstay treatment for advanced disease. Nevertheless, patients eventually develop castration-resistance and progress to lethal PCa [88].
Concerning patient management, clinicians face three major challenges: to distinguish PCa from benign prostatic hyperplasia and other cancer mimickers; to discriminate indolent from aggressive disease; and to foresee patients that will undergo disease progression and develop metastatic disease [89]. Epigenetic alterations are a common trait in PCa and are involved in disease onset and progression. Despite their exact roles are still not fully understood, the fact that they occur at a higher rate and in an earlier point than mutations makes them very attractive biomarkers for diagnosis, prognosis and follow-up purposes [90].
Deregulation of VIRMA, a component of the methylation complex, and of the readers YTHDF3 and YTHDC2 are of particular interest, constituting the most commonly altered genes in the pathway (18%, 13% and 11% of the samples, respectively). The remainder genes analyzed depicted alterations in less than 10% of samples. In addition, no mutations are described for YTHDF3 and YTHDC2 that may explain the deregulation, and only three samples disclosed a missense mutation in VIRMA. There was also a modest correlation between VIRMA and YTHDF3 mRNA expression in PCa samples (correlation coefficient: 0.62).
Analysis of alterations in the various pairs of genes showed two gene pairs with significant co-occurrent alterations. The strongest associations, with Bonferroni correction, included VIRMA and YTHDF3 (log OR > 2 and p < 0.001).
Regarding clinicopathologic correlates, VIRMA and YTHDF3 mRNA expression levels were significantly higher in stage III/IV compared to stage II tumors (p ≤ 0.0001 and p = 0.0454, respectively). In the same line, higher VIRMA and YTHDF3 transcript levels associated with higher GG (GG2-5 vs. GG1, p = 0.0198 and p = 0.0215, respectively), again suggesting higher expression of these players in more aggressive diseases. None of the genes tested impacted on overall survival (OS) or disease-free survival (DFS).
Although still largely unexplored, there is already a study (using both cell lines and human tissues from 35 patients) reporting m 6 A alterations in PCa. Specifically, the authors report that YTHDF2, an m 6 A reader, is regulated by miR-493-3p and its upregulation is involved in the m 6 A modification and malignant progression [91].

Testicular Cancer
Testicular germ cell tumors (TGCTs) comprise more than 95% of all testicular neoplasms, and are grouped into two major families according to the most recent World Health Organization (WHO) classification: the germ-cell neoplasia in situ (GCNIS)-related tumors (the most frequent, which include seminomas (SEs) and non-seminomatous Tumors (NSTs), two subgroups with very distinct behavior and clinical impact), and the GCNIS-unrelated ones [92].
Despite representing only 1% of male cancer worldwide, they constitute the most common cancer afflicting Caucasian men between 15-44 years old, with Western lifestyle contributing to a rising incidence worldwide. They also exhibit outstanding cure rates and a dropping mortality trend in response to multimodal treatments. However, many issues are left unresolved and deserve our attention, namely the substantial proportion of patients with disseminated disease that relapse with poor prognosis, the emergence of cisplatin resistance and the considerable morbidity induced by chemo-and radiotherapy in such young patients with long survival expectancy [85,[93][94][95].
Testicular germ cell tumors are remarkably heterogeneous (reflecting the complexity of this tumor model) [96] but mainly share a unifying cytogenetic background. In this line, it is only natural that various Epi-phenomena might play a fundamental role in these neoplasms. Therefore, the study of new Epi-markers might aid in tumor subtype discrimination, prognosis assessment and disease monitoring, as no accurate validated biomarkers exist for these purposes. In addition, the manipulation of these Epi-markers might provide ways of uncovering therapies with improved antitumor activity, less toxicity and that may overcome cisplatin resistance [97][98][99][100][101].
Three patients died and 33 experienced disease recurrence/progression, resulting in an OS and DFS at five years of 98% and 76%, respectively.
Paralleling our analysis on PCa, deregulation of VIRMA and the reader YTHDF3 is particularly interesting in TGCTs as well, being the two most commonly altered genes in the pathway (52% and 48% of samples, respectively). Following these two major deregulated genes, the reader HNRNPA2B1 and the writer METTL3 were also altered in 13% and 10% of the samples, respectively. The remaining genes analyzed disclosed alterations in less than 10% of samples. VIRMA and YTHDF3 are differently deregulated in SEs and NSTs (depicting alterations in 80% and 72% of SEs and in only 31% and 31% of NSTs, respectively), again mainly by mRNA upregulation. In addition, no mutations have been described for YTHDF3 that can explain its deregulation, and only one sample disclosed a missense mutation in VIRMA. There was also a strong correlation between VIRMA and YTHDF3 mRNA expression in TGCT samples (correlation coefficient 0.77).
Regarding subtype discrimination, mRNA expression levels of VIRMA and YTHDF3, but also the writer METTL4, the eraser ALKBH5 and the reader YTHDC1, were significantly higher in SEs compared to NSTs (p < 0.0001 for all). On the contrary, the writer METTL14 was significantly downregulated in SEs vs. NSTs (p < 0.0001). Of these genes, the best discriminative power assessed by ROC curve analysis was METTL4 (AUC = 0.91), followed by VIRMA (AUC = 0.83). Using the mRNA expression level that maximizes both sensitivity and specificity (228.04925) as cutoff, METTL4 discriminated between SEs and NSTs with 92.3% sensitivity, 82.4% specificity, 78.9% positive predictive value, and 93.8% negative predictive value, resulting in overall accuracy of 86.5%. Remarkably, METTL4 outperforms the serological markers commonly used in clinical practice (α-fetoprotein, subunit β of the human chorionic gonadotropin and lactate dehydrogenase) [95].
Furthermore, mRNA expression levels of METTL4, VIRMA and YTHDF3 were also significantly higher in stage I compared to stage II/III TGCTs (p = 0.0234, p = 0.0065 and p = 0.0165, respectively). Regarding survival analysis, the only genes with impact on survival were METTL4 (cases with alterations showing worse DFS, p = 0.0249), WTAP (cases with alterations showing worse DFS, p = 0.0402) and YTHDF1 (cases with alterations showing worse OS, p = 0.0440).

Kidney Cancer
Kidney cancer is the 14th most common malignancy worldwide and the 8th most prevalent cancer in Europe representing 3.5% of all adult malignancies. It is the most lethal among common urological cancers and, in 2012, there were 143,406 deaths attributable to this malignancy worldwide. Furthermore, incidence varies by gender, with men having twice the risk of women [85,103]. Due to its retroperitoneal topography, many renal masses remain asymptomatic until late stages. However, widespread use and improvement of imaging methods led to increased incidental detection of small renal masses, emphasizing the need for accurate discrimination among KCa subtypes, specifically between those which will be more aggressive and develop metastases and those that will have a more indolent growth and may be managed more conservatively [104].
There were two gene pairs with significant co-occurrent alterations, after Bonferroni correction, including VIRMA + YTHDF3 and RBM15B + YTHDC2 (both with log OR > 3 and p < 0.001). No significant gene pairs with mutually exclusive alterations were found.
VIRMA, RBM15B and YTHDC2 mRNA expression levels discriminated among ccRCC, chRCC and pRCC; transcript levels of VIRMA and YTHDC2 were lower in chRCC (p < 0.0001 for both) and in pRCC (p < 0.0001 and p = 0.0006, respectively) compared to ccRCCC. Contrarily, RBM15B was significantly upregulated in chRCC (p < 0.0001) and in pRCC (p < 0.0001) compared to ccRCC. Furthermore, RBM15B and YTHDC2 higher expression levels associated with advanced AJCC tumor stage (Stages II-IV vs. Stage I, p = 0.0361 and p = 0.0045, respectively). Regarding survival analysis, the only genes with impact on survival were VIRMA and YTHDC2 in ChRCC in both OS (mRNA upregulation conferring worse OS, p = 0.0280 and p = 0.0497, respectively) and DFS (mRNA upregulation conferring worse DFS, p = 0.0203 and p = 00152, respectively), and RBM15B in pRCC only in DFS (mRNA downregulation conferring worse DFS, p = 0.0082).
Although there is substantial lack of information regarding m 6 A alterations in KCa, Li and Tang et al. reported that higher expression of METTL3 might indicate better survival among RCC patients. Their study, which included both cell lines and human tissues from 145 patients (127 ccRCC and 18 designated as "others"), showed that this m 6 A writer might act as a tumor suppressor and may have impact on tumorigenesis and survival of KCa patients [105].

Bladder Cancer
Bladder cancer is the 9th most common cancer worldwide, with an estimated 430,000 new cases diagnosed in 2012. It is more prevalent in males (3/4 of all BlCa cases), representing the 6th most incident neoplasm in this group. Importantly, it is an important health issue, as recent trends follow tobacco smoking prevalence and because it was responsible for 165,000 deaths in 2012 (75% of which in men) [106,107].
The major histological type of BlCa is urothelial carcinoma. There are two major clinical, pathological and molecular forms of the disease: the non-papillary muscle-invasive tumors (with carcinoma in situ as the major precursor, being the most aggressive, more prone to progress and metastasize-25% of newly diagnosed cases) and the papillary non-muscle-invasive ones (with papillary urothelial lesions as precursors, being mainly characterized by multiple local recurrences, associated morbidity and, finally, increased risk of muscle-invasion over time-75% of newly diagnosed cases) [108].
High grade tumors showed significantly higher mRNA expression of VIRMA, METTL4 and YTHDF3 compared to low grade tumors (p = 0.003, p < 0.001 and p = 0.041, respectively), but the discriminative power was limited, the best disclosed by METTL4 (AUC 0.80). METTL4 and YTHDF1 mRNA expression levels did not, however, discriminate between papillary and non-papillary derived tumors (p = 0.1622 and p = 0.4321, respectively), but VIRMA was significantly upregulated in non-papillary tumors (p = 0.022), contrarily to YTHDC1 which was upregulated in papillary tumors (p = 0.0038). Nonetheless, the discriminative power was poor (AUC 0.59 for both).
YTHDC1 was upregulated in stage I/II compared to stage III/IV disease (p = 0.0089). Regarding survival analysis, the only gene with impact on survival was WTAP (cases with alterations showing better OS, p = 0.0261). A summary of these findings is depicted in Table 2 and illustrated in Figure 2.  Abbreviations: ccRCC-clear cell renal cell carcinoma; chRCC-chromophobe renal cell carcinoma; GG-grade groups; NST-non-seminomatous tumors; OR-odds ratio; pRCC-papillary renal cell carcinoma; SE-seminoma.

Discussion
In PCa, higher expression levels of VIRMA and YTHDF3 appear to be associated with advanced disease. The positive correlation between VIRMA and the reader YTHDF3 suggests cooperation for the establishment of m 6 A modification in these tumors. YTHDF2 expression also has an impact on disease progression as shown by Li and Meng et al. [91].
On the other hand, in TGCTs, expression levels of METTL4 and VIRMA seem promising biomarkers for discrimination between SEs and NSTs. In this tumor model, a positive strong correlation between VIRMA and the reader YTHDF3 was also observed, supporting again a cooperation between these two players for establishing m 6 A modification in urologic tumors. METTL4 expression had also an impact on DFS and associated with disease stage, as did VIRMA and YTHDF3 expression.
In KCa, VIRMA, RBM15B and YTHDC2 expression levels are auspicious biomarkers for discrimination among RCC subtypes, having impact on OS and DFS. Specifically, RBM15B and YTHDC2 associate with the advanced disease stage. Moreover, METTL3 plays a tumor suppressor role in this malignancy possibly acting as a novel marker for kidney tumorigenesis, as suggested by Li and Tang et al. [105].
Finally, in BlCa, VIRMA and METTL4 seem to be the most useful markers, as they are amongst the most commonly deregulated and they are significantly upregulated in high grade tumors. VIRMA was significantly upregulated in non-papillary tumors, but discrimination of the two major BlCa pathways using these markers remains a challenge.
Overall, regarding non-urological malignancies, upregulation of writers and/or writer-related players, tend to associate with more aggressive cancer features (poor prognosis, invasiveness, metastases and even treatment resistance). Mechanistically, this seems to imply that higher amounts of m 6 A modification in target RNAs might result in the development of cancer-prone features. Urological cancers tend to follow the same pattern, with upregulation of methylating enzymes associated with higher tumor grade and stage. The finding seems to contrast with the idea that m 6 A introduction is necessary for differentiation and that decreased m 6 A amount results in resistance to differentiation [29]. Nonetheless, exceptions exist both in urological and non-urological cancers. For instance, in TGCTs, higher VIRMA expression was found to be associated with low disease stage; in addition, in KCa, RBM15B overexpression (an eraser) associated with advanced disease at diagnosis, whereas the writer METTL3 was reported to act as a tumor suppressor. Ultimately, this might be interpreted in several ways: either the reader dictates the overall final destination of the target RNA (which can vary from degradation to increased translation), or m 6 A target RNAs may function as tumor suppressors or oncogenes. Overall, one has to take into account tumor subtype, relative expression of writers, erasers and readers, as well as the exact transcripts that are m 6 A-targeted.
Considering the ensemble of urological cancers, VIRMA upregulation stands as a common and shared trait, although in a variable proportion of cases. Considering the dissimilarity of age groups affected, as well as of risk factors, this is an intriguing observation. Nevertheless, it emphasizes the relevance of epitranscriptomics, and of m 6 A alteration in particular, in the genesis and progression of urological cancers.

Conflicts of Interest:
The authors declare no conflict of interest.