m⁶A RNA modification and its writer/reader VIRMA/YTHDF3 in testicular germ cell tumors: a role in seminoma phenotype maintenance

Testicular cancer is the most common neoplasia among Caucasian men aged 15–44 years, with rising incidence due to widespread adoption of Western lifestyle, with 65,827 new cases expected worldwide in 2030 [1–3]. More than 95% of cases are derived from germ cells—testicular germ cell tumors (TGCTs)—and the vast majority of these correspond to germ-cell neoplasia in situ (GCNIS)-related tumors, according to the most recent World Health Organization (WHO) classification [4]. Furthermore, this category comprises two major subtypes—seminomas (SEs) and non-seminomatous tumors (NSTs)—and discrimination between them is of paramount clinical importance, entailing different prognosis and treatment algorithms [5, 6].

TGCTs are fascinating tumors, in part because they are truly developmental cancers [7]. Heterogeneity is the hallmark of GCNIS-related TGCTs, reflecting this complex tumor model, although they share a common cytogenetic background, i.e., isochromosome 12p [4, 8]. Thus, biological, morphological and clinical heterogeneity might also be related with dissimilar epigenetic backgrounds, which might be surveyed using novel biomarkers. Indeed, there is an increasing need for reliable and clinically validated TGCT biomarkers that might improve diagnosis, subtype discrimination, prognostication and patient monitoring, overcoming the limitations of classical serum markers currently employed in the clinical setting [9–14].

Recently, post-translational RNA modifications (so-called “Epitranscriptomics”) have emerged as fundamental modulators of many biological and disease processes [15, 16]. The most abundant of these modifications, N6-methyladenosine (m⁶A), is dynamically regulated by a variety of m⁶A-related proteins, organized as writers (which catalyze the methyl code), erasers (which delete the methyl code) and readers (m⁶A-binding proteins that target RNAs to their final destiny) [17–22]. The amount of m⁶A modification in messenger RNAs (mRNAs) has been implicated in diverse biological mechanisms and related diseases, such as immune response, metabolism, gametogenesis, embryogenesis, neurodevelopment and cancer [23–27].

Deregulation of m⁶A-related proteins has been shown to impact tumorigenesis and progression of several neoplasms, including breast, lung, liver and colorectal carcinomas, leukemias and glioblastoma, among others [24, 28–31]. Recently, we have thoroughly reviewed the available literature concerning these players in all tumor models [32]. There have been several studies, focusing both on the writer METTL3 [33] and other components of the methylation complex (such as METTL14 [34]), on erasers (such as ALKBH5 [35]) and readers (such as YTHDF1 [36]), seeking for clinicopathological correlates or further upstream or downstream (de)regulation mechanisms in cancer. While most studies seem to imply that overexpression of writers associates with poor prognostic features, many exceptions were depicted. However, no studies have focused on VIRMA and YTHDF3 as a writer/reader pair, which is surprising since our in silico analysis of publicly available databases pointed out these players as being preferably deregulated in urological cancers, including prostate, kidney and bladder cancer and also in TGCTs.

Moreover, and to the best of our knowledge, no studies on TGCTs have been reported, despite the importance of m⁶A modification in germ cell differentiation [37]. Given this link between m⁶A modification and embryogenesis and spermatogenesis, and since TGCTs are developmental-related, we hypothesize that alterations in m⁶A amount may also have a role in these tumors, with possible differences among more undifferentiated forms such as SE and more differentiated subtypes such as Teratoma (TE).

Given this rationale, we set out to assess the differential expression of m⁶A writer VIRMA and reader YTHDF3, in a cohort of TGCT patients, comparing with m⁶A abundance and establishing associations with clinicopathological data, looking for potential biological and clinical relevance of these findings.

Although relatively infrequent, TGCTs are highly curable, carrying a generally good prognosis. Nevertheless, and despite continuous improvement of multimodal treatments, reduced fertility and second neoplasms remain important side effects. On the other hand, 15–20% of patients with disseminated disease eventually relapse, entailing poor prognosis (especially late relapses) whereas others develop cisplatin resistance, by still elusive mechanisms, and the therapeutic strategy for these patients is suboptimal. Thus, novel disease biomarkers that may improve TGCT diagnosis and subtyping, prediction of disease progression and patient monitoring are needed, and their clinical implementation might aid in tailoring and individualizing therapeutic strategies [9, 11, 13, 46–52]. Over the past few years, more than 140 RNA modifications have been uncovered, as well as their context-dependent role in regulating target RNAs fate, namely in stability, translation and splicing [53]. Among these modifications, m⁶A seems, by far, the most abundant in mammalian mRNAs and a handful of enzymes that regulate this covalent modification have been characterized, including adenosine methyltransferases (writers), demethylating enzymes (erasers) and m⁶A-binding proteins (readers). Recently, the amount of m⁶A in RNAs, fine-tune regulated by those proteins, has been suggested to impact in cancer onset and progression. These players may also constitute therapeutic targets [19, 22, 24, 25, 28, 54], thus representing potential cancer biomarkers.

Considering the role of m⁶A modification in germ cell development, and being TGCTs considered development-related cancers [7, 55], we hypothesized that changes in abundance of this mark and/or altered expression of the enzymes involved in its writing, reading or erasing might be potential TGCT biomarkers. For selection of the best candidates among all m⁶A-related players, the TCGA dataset was surveyed, and VIRMA (a component of the m⁶A writer complex) and YTHDF3 (reader) surfaced as the most commonly altered genes involved in m⁶A dynamics in TGCTs, prompting validation in an independent tissue cohort. Remarkably, SEs displayed higher VIRMA and YTHDF3 transcript levels compared to NSTs, in both cohorts (TCGA’s and IPO Porto’s), notwithstanding some differences in composition. Indeed, although the proportion of SEs was similar (46% vs. 42%, in IPO Porto’s and TCGA cohorts, respectively), IPO Porto’s cohort globally displayed a lower proportion of stage I tumors (64% in IPO Porto’s vs. 79% in TCGA). Nonetheless, most metastatic tumors were within the Good IGCCCG Prognosis group in both series (71% and 74%, respectively). Moreover, median age at diagnosis was 28 years in our cohort and 31 years in TCGA’s, slightly lower than the 35 years reported in most studies, which might be due to the lower proportion of SEs in both series, as SEs tend to be diagnosed a decade later that NSTs [3, 4]. Thus, the overall characteristics of both cohorts might be considered globally representative of TGCT patients.

TGCTs tumorigenesis relates to biological processes of spermatogenesis and stem-cell differentiation [55], with SE constituting the so-called default pathway for GCNIS cells, whereas reprogramming originates NST components, including more undifferentiated and embryonal forms like EC, but also extra-embryonal forms like YST and CH, and more differentiated and somatic forms like TE [14, 51]. Because RNAs m⁶A chemical modification has been implicated in stemness state maintenance [37, 56–59], changes in abundance of m⁶A mark and differential expression of respective regulating proteins might be expected along differentiation of the various TGCT components. SEs displayed the highest VIRMA and YTHDF3 mRNA levels, which were further confirmed at protein level for VIRMA, with no significant differences among NST subtypes. Considering their roles as m⁶A writer and reader, respectively, it is tempting to speculate whether m⁶A might contribute to the emergence and maintenance of the SE phenotype. Interestingly, knockdown of m⁶A writer complexes such as METTL3, METTL14, VIRMA, Hakai and WTAP lead to mouse embryonic stem cells (mESCs) self-renewal capacity loss, triggering differentiation [37, 56]. Hence, it has been proposed that m⁶A modification preferentially targets transcripts involved in development regulation, acting predominantly by reducing their stability and/or promoting degradation, maintaining mESCs at their ground state [37]. This observation is also in line with higher VIRMA expression observed in SEs compared to TEs, which represent the more differentiated TGCTs. TEs, however, showed the stronger m⁶A immunostaining intensity, suggesting that other writer complexes might be responsible for establishing m⁶A in this tumor subtype and/or that m⁶A modification might target other RNAs and even impart them a different fate. Intriguingly, this might parallel previous observations in mESCs, in which m⁶A displays conflicting results depending on the cell state: preserving stability of primed epiblast stem cells—EpiSCs—which are primed for differentiation, but inducing differentiation of naïve stem cells by degrading pluripotency factors [23, 58].

Overall, IHC results were concordant with transcript analysis, as tumors with strong VIRMA immunoexpression showed significantly higher VIRMA mRNA levels and a similar trend was depicted for YTHDF3. Moreover, a positive correlation between the writer and reader transcript levels was demonstrated in both cohorts, indicating they may cooperate in catalyzing and recognizing m⁶A modification in target RNAs. Furthermore, we found an association between higher expression of the writer (both at the transcript and immunohistochemical levels) and stronger m⁶A immunoexpression, suggesting that changes in m⁶A abundance might be due to VIRMA upregulation. Importantly, the observed immunoexpression patterns are coincidental with the expected subcellular localization. Indeed, VIRMA is mainly localized in nuclear bodies and nucleoplasm (correlating with granular/dot-like immunoexpression), whereas YTHDF3 is mostly cytoplasmic (especially in processing-bodies, imparting the granular staining observed in our samples) but it may also localize to nuclear speckles and nuclear membrane [19, 60–62]. Furthermore, m⁶A may be localized inside or outside the nucleus, imparting different biological significance [27], which may underlie the differences in immunoexpression depicted among distinct TGCT subtypes. On the other hand, immunostaining results in primary TGCTs and respective metastases was somewhat unexpected. Paradoxically, VIRMA and YTHDF3 immunostaining were increased in the metastatic context, whereas m⁶A modification displayed reduced intensity, suggesting that m⁶A erasure may occur during the metastatic process. In this context, a role for m⁶A erasers FTO and ALKBH5 in TGCT metastization is suggested. WTAP is fundamental for maintaining the writer complex inside the nuclear speckles, and it has been reported that WTAP knockdown results in m⁶A reduction mediated by sequestering of the writer complex in the cytoplasm; however, in our study, VIRMA immunoexpression in metastatic samples still occurred mainly in the nucleus [19, 63]. Nonetheless, our results should be interpreted with caution considering the limited number of metastatic samples and that most metastases were obtained following chemo- and/or radiotherapy, which may impact in m⁶A maintenance and detection.

Accurate discrimination of SEs from NST is key for patient management because SEs are particularly susceptible to chemo- and radiotherapy agents and display generally good prognosis, whereas NSTs show variable degrees of chemo- and radio-resistance and are more commonly associated with unfavorable prognostic features [6, 64, 65]. Thus, the ability of VIRMA transcript levels to discriminate SEs from NSTs is also of clinical significance and displayed similar results in both cohorts (AUC of 0.85 and 0.83 in IPO Porto’s and TCGA cohorts, respectively). Moreover, a gene panel comprising VIRMA and YTHDF3 further increased sensitivity and specificity. If confirmed in liquid biopsies in the future, these results outperform those of classical TGCT serum markers [66, 67] and might constitute promising biomarkers for patient monitoring.

The retrospective nature of this study and the validation cohort size are important limitations of this study. Nevertheless, our cohort is similar in many respects to that of TCGA and reflects many aspects of TGCTs epidemiology. Furthermore, all patients were evaluated and treated in a single institution by the same multidisciplinary team, entailing homogeneity in patients’ staging and clinical management. Although m⁶A IHC assay is not ideal and qualitative analysis carries a considerable inter- and intra-observer variation, it is a widespread and accessible technique that may corroborate transcript findings.

In summary, in this work we have used a bioinformatic tool to perform an in silico analysis, which allowed us to identify the most promising players to be further validated in a patient tissue cohort. We have shown that m⁶A writer VIRMA and reader YTHDF3 are differentially expressed among TGCTs subtypes, with significant overexpression in SEs compared to NSTs, suggesting a contribution to stemness maintenance. Furthermore, a positive correlation between VIRMA and YTHDF3 expression and an association with m⁶A abundance was also disclosed. Because VIRMA and YTHDF3 transcript levels accurately discriminate between SEs and NSTs, they might constitute novel biomarkers for patient management. Hence, our results confirm in silico findings and further enlighten the differentiation biology of TGCTs, which are development-related neoplasms. If confirmed in liquid biopsies, these players may also prove useful in the clinical setting.

To the best of our knowledge this is the first study assessing the value of m6A and related proteins in TGCTs. Additional research on m⁶A modification dynamics might further illuminate TGCTs biology and clinical behavior.