Results and Discussion
Shifts in the DNA methylation patterns of tDNAs occur in human tumorigenesis
To identify DNA methylation defects in tDNA, we retrieved the Infinium HumanMethylation450 (HM450) methylation microarray probes’ identity that mapped into the loci corresponding to the 416 high confidence tDNA that are found in the human nuclear genome . 138 tDNAs were represented in the HM450 microarray, and 95 of them were located further than 2 kb from RNAPII-transcribed genes (Supplementary Figure S1A). We established this exclusion criterion because the interferences mediated by elongating RNAPII on RNAPIII transcription of neighboring genes could mask the effects that DNA methylation would induce on tRNA expression. We identified 71 tDNA genes that were represented by at least one CpG in the HM450 microarray whose methylation status could be efficiently interrogated (Supplementary Figure S1A).
The data mining of HM450 microarray-derived data of thousands of samples from various cancer types and normal tissues available at The Cancer Genome Atlas (TCGA) and of a panel encompassing approximately 1000 cell lines revealed global differences in tDNA methylation in TCGA primary tumors and cancer cell lines compared to TCGA normal samples, mostly tending towards hypermethylation in malignant samples (Supplementary Figure S1B). Specifically, 60 of the 71 tDNA presented statistically significant differences between normal and tumoral TCGA sets of samples (Supplementary Figure S1B). The separation of TCGA and cell line samples according to their origin showed that the different normal and tumoral tissues also presented variations in tDNA methylation levels according to the cancer type (Fig. 1A). These findings indicate that tumorigenesis involves epigenetic lesions in tDNA genes that are characteristic among different tissues.
tDNA methylation is associated with reduced tRNA expression in cancer cells
The differences in tDNA methylation status between normal and malignant samples and among tissues prompted us to investigate whether this phenomenon could drive cancer-associated and tissue-specific variations in the expression of specific tRNAs. To address this, we evaluated the in silico correlation between tDNA methylation and the expression of their cognate tRNA using a publicly available dataset of tRNA expression in TCGA samples . Three tRNAs presented a statistically significant Spearman’s correlation coefficient (ρ ≤ -0.2) when considering all the samples: tRNA-Arg-TCT-4-1, tRNA-Ile-AAT-8-1, and tRNA-Val-CAC-2-1 (Fig. 1B). Since tDNA methylation presented differences depending on the origin of the sample (Fig. 1 A), we also investigated if the association between tDNA methylation and tRNA expression was dependent on the tissue (Fig. 1B). From the total 1775 computed Spearman’s correlations, 1073 revealed a negative association between tDNA methylation and tRNA expression (Fig. 1B). 114 of the computed correlations displayed a significant anticorrelation (ρ ≤ -0.2; Supplementary Figure S2A). tRNA-Arg-TCT-4-1, tRNA-Ile-AAT-8-1, and tRNA-Val-CAC-2-1 were those tRNAs with the highest frequency of tissues displaying a negative association between tDNA methylation and tRNA expression (Supplementary Figure S2B). This negative correlation between tDNA methylation and tRNA levels in TCGA suggests a plausible link between DNA methylation and tRNA expression.
Following the above described findings, we aimed to analyze in more detail the relationship between DNA methylation and tRNA expression by providing additional experimental evidence of the epigenetic regulation of two tDNA whose methylation is susceptible to impact tRNA expression according to our analysis: tRNA-Arg-TCT-4-1 and tRNA-Ile-AAT-8-1. tRNA-Arg-TCT-4-1 is methylated in most normal tissues, whereas tumoral samples show hypomethylation events when compared to their matched healthy counterparts. (Figure 1A). Conversely, tRNA-Ile-AAT-8-1 is unmethylated in normal tissues and hypermethylated in cancer (Fig. 1A). We selected two cell lines that were hypermethylated for both genes according to the HM450 microarray data: DND41 and SW48 (Supplementary Figure S3A). Both tDNAs are represented in the HM450 microarray with a single CpG that is located within the internal promoter of the gene –the conserved A and B boxes where TFIIIC binds to recruit the RNAPIII machinery . We confirmed by bisulfite genomic sequencing of multiple clones the methylation status of the two selected tDNA (Supplementary Methods) in these cell lines and obtained a comprehensive view of the whole methylation landscape for each of them (Supplementary Figure S3B). DNA demethylation upon the use of the drug 5’-azacytidine on these cell lines increased the binding of the subunit 1 of TFIIIC (GTF3C1) and the subunit A of RNAPIII (POLR3A) to the two tDNA genes as measured by chromatin immunoprecipitation (ChIP)-qPCR (Supplementary Methods), indicating that DNA methylation prevents the binding of the tDNA-associated transcriptional machinery to the internal promoter of the tDNAs (Fig. 1 C). Moreover, the treatment with the demethylating agent 5’-azacytidine also restored the expression of both tRNAs in the two interrogated hypermethylated cell lines (Fig. 1D) as determined by qRT-PCR (Supplementary Methods), confirming that DNA methylation can repress RNAPIII-mediated gene expression. The use of the DNA methylation inhibitor, and its associated restoration of tRNA expression upon demethylation, reduced cell growth (Supplementary Figure S4A) and induced apoptosis (Supplementary Figure S4B). Thus, our in vitro results showing an increase in tRNA expression in hypermethylated cell lines upon DNA demethylation are in accordance with the negative relationship between tDNA methylation and tRNA expression that arose from the in silico analysis in TCGA (Fig. 1B).
tDNA methylation can predict the patients’ clinical outcome in TCGA cohorts
Apart from modulating gene expression to promote the malignant transformation of the cell and facilitate cancer progression, tumor-specific alterations in DNA methylation can be translated into the clinical practice as potential biomarkers to predict the evolution of the disease . The expression of some tRNA has also been associated to different clinical outcomes of cancer patients [9, 12]. For this, because we observed that the changes in tDNA methylation that occur in tumorigenesis modulate tRNA expression, we sought to evaluate if these events had any impact on the outcome of cancer patients and therefore could serve as predictive biomarkers for this disease. To address this hypothesis, we data mined the clinical information from 31 TCGA projects to assess if differential tDNA methylation could inform patients’ survival. We identified 86 cases in which the methylation status of a specific tDNA was significantly associated with differences in patients’ overall survival (Fig. 2A). 56 of these events were confirmed to be prognostic predictive factors for the disease according to univariate Cox regression analyses (Fig. 2 A, Supplementary Table S1). Since the tRNA pool composition is reported to vary among tissues to match the specific cellular needs and permit the expression of different sets of proteins , whether each differential tDNA methylation event results in an increased or a reduced risk of the disease should be evaluated individually for every gene and tumor (Fig. 2 A, Supplementary Table S1).
The silencing of tRNA-Arg-TCT-4-1 reduced endometrial cancer cell growth
Our analyses identified tRNA-Arg-TCT-4-1 as a promising candidate to participate in tumorigenesis. Interestingly, high levels of tRNA-Arg-TCT-4-1 are linked to a proliferative status of the cell , which fits with the cancer-associated demethylation of this tDNA in comparison with the methylated status of the normal tissues (Fig. 1 and Supplementary Figure S1A) and its resulting overexpression in 12 cancer types represented in TCGA . Importantly, tRNA-Arg-TCT-4-1 demethylation in kidney renal papillary cell carcinoma (KIRP) and in uterine corpus endometrial carcinoma (UCEC) cohorts of TCGA primary tumors is associated to a shorter overall survival (Fig. 2B). This hypomethylation change in KIRP and UCEC TCGA cohorts accounts for an 8% and 29% of cases, respectively (Supplementary Figure S5A). In both cases, the hypomethylation of the gene was associated to an increased expression of the tRNA (Supplementary Figure S5B). This is consistent with the overexpression of this tRNA in a subgroup of tumoral samples compared to the normal matched tissue  that is methylated (Fig. 1 A and Supplementary Figure S1A). The higher percentage of cases presenting tRNA-Arg-TCT-4-1 demethylation in the UCEC cohort of primary tumors (Supplementary Figure S5A) prompted us to investigate the role of this tRNA in endometrial cancer and confirm the proposed oncogenic role of its demethylation. The prognosis for advanced forms of this disease is poor, and therapeutic options beyond surgery and first-line chemotherapy are limited . Therefore, there is an urgent need to investigate this malignancy and describe biomarkers that predict its aggressiveness to spot high-risk patients that would need a more comprehensive vigilance or that would benefit from receiving adjuvant therapy after the surgical resection of the tumor .
First, we verified by bisulfite genomic sequencing of multiple clones the unmethylated status of tRNA-Arg-TCT-4-1 gene in the endometrial cancer cell line HEC1 (Fig. 2 C), as previously noted in the overall screening of cancer cell lines (Fig. 1 and Supplementary Figure S1B). We also confirmed that expression of this tRNA was higher in HEC1 than in the hypermethylated DND41 and SW48 cell lines (also derived from our in silico screening, Fig. 1and Supplementary Figures S1B and S3A) to support that tDNA hypermethylation represses tRNA expression (Fig. 2D). Changes in tRNA expression can drive tumorigenesis as they can confer advantages to the malignant cell to promote cancer progression and dissemination . To further characterize the role of tRNA-Arg-TCT-4-1 in endometrial carcinoma, we eliminated this gene from HEC1 cell line using the CRISPR/Cas9 system (Supplementary Methods) to mimic the effects of its hypermethylation-associated silencing (Supplementary Figure S6A). This deletion was confirmed by genomic PCR and Sanger sequencing in two HEC1 knockout clones (Supplementary Figure S6B), which showed minimal tRNA-Arg-TCT-4-1 expression compared to the wild-type cell line (Supplementary Figure S6C). tRNA-Arg-TCT-4-1 CRISPR/Cas9-mediated deletion delayed the progression of the cell cycle, as seen by an accumulation of cells in the G0/G1 phase in detriment of the S and G2/M phases in these cells (Fig. 2E). This resulted in a reduced growth of HEC1 knockout cells compared to the wild-type cell line, as assessed by the SRB assay (Fig. 2 F). Additionally, tRNA-Arg-TCT-4-1 silencing also reduced cell migration in the two HEC1 knockout clones (Fig. 2G). Overall, our results indicate that tRNA-Arg-TCT-4-1 demethylation and increased expression exerts oncogenic-like roles in endometrial cancer by promoting cell growth, in agreement with the observed reduced overall survival of those cancer patients that undergo hypomethylation-associated activation of tRNA-Arg-TCT-4-1 (Fig. 2B and Supplementary Figure S5B).