Epigenetic loss of the transfer RNA-modifying enzyme TYW2 induces ribosome frameshifts in colon cancer Margalida Rosselló-Tortellaa,b, Pere Llinàs-Ariasa, Yuriko Sakaguchic, Kenjyo Miyauchi (宮内健常)c, Veronica Davalosa, Fernando Setiena, Maria E. Calleja-Cervantesa, David Piñeyroa,d, Jesús Martínez-Gómeza, Sonia Guila,b, Ricky Joshia, Alberto Villanuevae, Tsutomu Suzukic, and Manel Estellera,d,f,g,1 a Cancer Epigenetics Group, Josep Carreras Leukaemia Research Institute, 08916, Badalona, Barcelona, Catalonia, Spain; bGermans Trias i Pujol Health Science Research Institute, 08916, Badalona, Barcelona, Catalonia, Spain; cDepartment of Chemistry and Biotechnology, Graduate School of Engineering, University of Tokyo, 113-8656 Tokyo, Japan; dCentro de Investigacion Biomedica en Red Cancer, 28029 Madrid, Spain; eTranslational Research Laboratory, Catalan Institute of Oncology, Bellvitge Biomedical Research Institute, 08908, L’Hospitalet de Llobregat, Barcelona, Catalonia, Spain; fInstitucio Catalana de Recerca i Estudis Avançats, 08010 Barcelona, Catalonia, Spain; and gPhysiological Sciences Department, School of Medicine and Health Sciences, University of Barcelona, 08907 Barcelona, Catalonia, Spain Edited by Nahum Sonenberg, McGill University, Montreal, Canada, and approved July 17, 2020 (received for review February 21, 2020) Transfer RNA (tRNA) activity is tightly regulated to provide a sequences (12, 15), which is essential for the correct expression of physiological protein translation, and tRNA chemical modifications viral proteins. Ribosome frameshifting also occurs in messenger control its function in a complex with ribosomes and messenger RNA (mRNA). When ribosomes slip and alter the reading frame, RNAs (mRNAs). In this regard, the correct hypermodification of a premature stop codon usually appears, and the mRNA is de- position G37 of phenylalanine-tRNA, adjacent to the anticodon, is graded via nonsense-mediated decay (NMD) (16). Therefore, critical to prevent ribosome frameshifting events. Here we report PRF events are postulated to be an additional layer of posttran- that the tRNA-yW Synthesizing Protein 2 (TYW2) undergoes pro- scriptional regulation of mRNA abundance. MEDICAL SCIENCES moter hypermethylation-associated transcriptional silencing in hu- An eye-opening stream of data indicating a role for the man cancer, particularly in colorectal tumors. The epigenetic loss hypermodification of position G37 in cancer occurred in the late of TYW2 induces guanosine hypomodification in phenylalanine- 1970s, when various researchers highlighted in prominent pub- tRNA, an increase in −1 ribosome frameshift events, and down- lications the tumor-specific loss of the Y-base of tRNAPhe and regulation of transcripts by mRNA decay, such as of the key cancer suggested that this phenomenon could provide a growth advantage gene ROBO1. Importantly, TYW2 epigenetic inactivation is linked to these cells (17–19). The underlying mechanism has remained to poor overall survival in patients with early-stage colorectal can- unknown for the last 45 y, however. To solve this enigma, we cer, a finding that could be related to the observed acquisition of wondered about the presence of cancer-specific defects in the six enhanced migration properties and epithelial-to-mesenchymal fea- enzymes, TRMT5 (20) and TYW1–5 (21, 22), that act sequentially tures in the colon cancer cells that harbor TYW2 DNA methylation- in humans to modify the original G37 nucleotide to finally obtain associated loss. These findings provide an illustrative example of the hypermodified yW in its hydroxylated (OHyW) or peroxydated how epigenetic changes can modify the epitranscriptome and fur- ther support a role for tRNA modifications in cancer biology. (o2yW) form. Since transcriptional silencing associated with epigenetics | transfer RNA | cancer Significance Defects in transfer RNA (tRNA) modifications occur in human C oding and noncoding RNA molecules harbor chemically modified nucleosides that together constitute the so-called “epitranscriptome” (1, 2). These chemical modifications are pathologies such as cancer; however, how these alterations contribute to the disease is poorly understood. One example is particularly abundant in transfer RNA molecules (tRNAs), with the tumor-specific hypomodification of position 37 of tRNAPhe, more than 50 different modifications described in eukaryotic which was first described 45 y ago, although its cause and tRNAs (3–6). Modified bases in tRNA are critical for its trans- consequences have remained unknown. Here we report that lational function at multiple levels, including amino acid loading, the tRNAPhe hypomodification is due to promoter CpG island wobbling or translation efficiency, and fidelity, among others (5, hypermethylation-associated transcriptional silencing of TYW2, a key enzyme in the synthesis of wybutosine derivatives. Fur- 7). tRNA modifications also serve as a translational regulation thermore, epigenetic loss of TYW2 in transformed cells provokes mechanism under stress conditions (4). Defects in tRNA modi- hypomodified tRNAPhe-mediated ribosome frameshifting, dys- fications or tRNA modifier enzymes are present in various hu- regulating mRNA abundance via nonsense-mediated decay. Im- man pathologies, including neurologic disorders, mitochondrial portantly, TYW2 silencing in cancer cells confers enhanced diseases, and cancers (8–10). migration and epithelial-to-mesenchymal features that are as- The causes and roles of the tRNA modification changes ob- sociated in early-stage colorectal cancer patients with poor served in cancer are poorly understood. One example of chemical clinical outcome. modifications found in tRNA that has not been studied in depth in the context of transformed cells is the existence of wybutosine Author contributions: M.R.-T. and M.E. designed research; M.R.-T., P.L.-A., Y.S., K.M., F.S., (yW)-derived nucleosides in position G37, adjacent to the anti- J.M.-G., S.G., R.J., A.V., and T.S. performed research; V.D., M.E.C.-C., and D.P. analyzed codon of phenylalanine tRNA (tRNAPhe). Highly modified pu- data; and M.R.-T. and M.E. wrote the paper. Downloaded at Universitat Autonoma de Barcelona (UAB) on February 10, 2021 rines in position 37 of various tRNAs are key for maintaining the Competing interest statement: M.E. serves as a consultant for Ferrer International and ribosome reading frame and ensuring translation fidelity. Con- Quimatryx. cretely, yW, first described in 1968 and originally called Y-base This article is a PNAS Direct Submission. (11), prevents −1 programmed ribosome frameshifting (PRF) This open access article is distributed under Creative Commons Attribution License 4.0 events (12) as a result of a proper stabilization of the codon- (CC BY). anticodon pairing through base-stacking interaction (13). Ribo- 1 To whom correspondence may be addressed. Email: mesteller@carrerasresearch.org. some frameshifts occur in sequences with a specific structure This article contains supporting information online at https://www.pnas.org/lookup/suppl/ comprising a slippery heptamer followed by a complex secondary doi:10.1073/pnas.2003358117/-/DCSupplemental. structure (14). The loss of yW enhances −1 PRF events in viral First published August 10, 2020. www.pnas.org/cgi/doi/10.1073/pnas.2003358117 PNAS | August 25, 2020 | vol. 117 | no. 34 | 20785–20793 promoter CpG island hypermethylation is a frequent mechanism SW48 and HT-29 minimally expressed the TYW2 RNA tran- of gene inactivation in transformed cells (23, 24), we interrogated script and protein, as determined by quantitative real-time PCR the presence of this type of epigenetic inactivation in the genes and Western blot analysis, respectively (Fig. 1H). In contrast, the encoding the enzymes responsible for yW derivative synthesis. We unmethylated cell lines HCT-116 and SW480 expressed the found that TYW2 undergoes tumor-specific silencing by promoter TYW2 transcript and protein (Fig. 1H). The use of the inhibitor CpG island hypermethylation, preventing formation of the hyper- of DNA methylation 5-aza-2-deoxycytidine in the TYW2 hyper- modified forms of G37. Those cancer cells with the hypomodified methylated cell lines restored its expression at the RNA and guanosine were prone to ribosome frameshifting events. Most im- protein levels (Fig. 1I), reinforcing the link between TYW2 portantly, the epigenetic defect in TYW2 was associated with poor promoter hypermethylation and transcriptional silencing. clinical outcome of the studied colorectal cancer, an observation that can be linked to the acquisition of enhanced cellular migration TYW2 Epigenetic Silencing in Cancer Cells Mediates Loss of Hypermodified features and epithelial-to-mesenchymal features on TYW2 loss. Guanosine in tRNA and Ribosomal Frameshifting. After detecting the presence of TYW2 CpG island hypermethylation-associated tran- Results scriptional loss in colorectal cancer cell lines, we studied its con- Promoter CpG Island Hypermethylation-Associated Transcriptional tribution to the chemical modification status of G37 using tRNA- Silencing of the tRNA-Modifying Enzyme TYW2 in Colon Cancer. To associated liquid chromatography-mass spectrometry (LC/MS) identify epigenetic defects in the enzymes involved in generating (26). We first assessed the profile of the modified nucleosides of the hypermodified guanosine nucleotide in the tRNAPhe, we G37 according to the TYW2 CpG island methylation status in the studied the DNA methylation status of the 5′-end–associated CpG studied colorectal cancer cell lines. We observed that the fully islands of TRMT5, TYW1, TYW2, TYW3, TYW4, and TYW5 genes. hypermodified residues OHyW and o2yW were present in the These six human enzymes were identified as involved in the dif- TYW2 unmethylated/expressing cells HCT-116 and SW480 (Fig. ferent chemical modifications that this nucleotide undergoes from 2A), whereas the intermediate modified nucleoside imG-14, the its initial G37 form (Fig. 1A) (21, 22). Data mining of thousands of direct substrate of TYW2 (Fig. 1A), was absent (Fig. 2A). The primary tumors corresponding to all cancer types available at The opposite scenario was observed in the TYW2 hypermethylated/si- Cancer Genome Atlas (TCGA) (https://www.cancer.gov/about-nci/ lenced cell lines HT-29 and SW48, which showed an accumulation organization/ccg/research/structural-genomics/tcga) did not indi- of the TYW2 substrate imG-14 and a loss of the hypermodified cate the presence of TRMT5, TYW1, TYW3, TYW4 or TYW5 OHyW and o2yW (Fig. 2A). promoter hypermethylation events (SI Appendix, Fig. S1A and To further interrogate the association between TYW2 ex- Dataset S1). In contrast, the TYW2 promoter CpG island was pression and the chemical modification status of G37, we gener- methylated in 19.03% (79 of 415) of the primary colorectal car- ated two stable TYW2 knockout models in the unmethylated cell cinomas available in the TCGA (Fig. 1B and Dataset S1). Beyond lines HCT-116 and SW480 to mimic the effects of the observed colorectal tumors, the TYW2 promoter CpG island was also transcriptional silencing. In this approach, we used the CRISPR/ commonly hypermethylated in cervical (17.37%; 45 of 259), gastric Cas9 system to generate a deletion within TYW2 gene body and (12.88%; 51 of 396). and uterine (12.12%; 60 of 495) carcinomas confirmed the removal of a 234-bp fragment by Sanger sequencing (Fig. 1B and Dataset S1). For TYW2, as well as the other five (SI Appendix, Fig. S3A) and by genomic PCR (SI Appendix, Fig. tRNA-modifying enzymes studied here, the promoter CpG island S3B). The effectiveness of the knockout was confirmed by Western was always unmethylated in all the normal human tissues studied blot analysis, which showed no TYW2 protein expression in HCT- (Dataset S2). Data mining of the TCGA RNA-sequencing (RNA- 116 and SW480 TYW2 KO cell lines (Fig. 2B). On CRISPR/Cas9- seq) data available in colorectal tumors showed that TYW2 mediated loss of TYW2 in HCT-116 and SW480, we found an hypermethylation was associated with transcript down-regulation accumulation of the TYW2 substrate imG-14 and loss of hyper- (Spearman rho = −0.47; P < 0.0001) (Fig. 1C). In this regard, modified OHyW and o2yW (Fig. 2C), mimicking the effects of the TYW2 was the only G37 tRNAPhe- modifying enzyme showing DNA methylation-associated transcriptional silencing in the HT-29 RNA down-regulation in the TCGA dataset (SI Appendix, Fig. S2). and SW48 cell lines (Fig. 2A). We also performed the reverse We also data mined the Sanger set of 1,001 human cancer cell experiment in which we restored expression of TYW2 in the lines, in which we had recently obtained the DNA methylation hypermethylated/silenced cell line HT-29 by stable transfection and expression profiles (25), and obtained results similar to those (Fig. 2B). The LC/MS approach resulted in the opposite scenario observed in the TCGA cohorts (SI Appendix, Fig. S1B and as seen in the CRISPR/Cas9 deletion model: the recovery of Dataset S3). In this regard, colorectal cancer cell lines were the TYW2 expression by transfection depleted the intermediate imG- most frequently TYW2-hypermethylated tumor type (Fig. 1D and 14 nucleoside and caused the appearance of OHyW and o2yW Dataset S3), and TYW2 hypermethylation was also associated hypermodified residues (Fig. 2C), mimicking the pattern observed with transcript down-regulation (Fig. 1E), mimicking the results in the TYW2 unmethylated/expressing cell lines HCT-116 and observed in the TCGA cohorts. Thus, the higher frequency of SW480 (Fig. 2A). The modified nucleoside yW-86 was undetect- aberrant TYW2 DNA methylation-associated transcriptional si- able in all the samples studied (SI Appendix, Fig. S4), supporting lencing observed in colorectal cancer prompted us to focus our the reported kinetics of the pathway that rapidly transforms this efforts on this tumor type. residue to the hypermodified OHyW and o2yW nucleosides (21). We analyzed in more detail the link between TYW2 promoter Hypermodified guanosine at position 37 of tRNA, such as CpG island hypermethylation and transcriptional inactivation of wybutosine in yeast, is known to help maintain the reading frame the gene at the RNA and protein levels in colon cancer cell lines. during translation by preventing −1 PRF events at mRNAs that We performed bisulfite genomic sequencing of multiple clones contain a slippery sequence (12). Thus, we wondered whether Downloaded at Universitat Autonoma de Barcelona (UAB) on February 10, 2021 from SW48, HT-29, HCT-116, and SW480 cells using primers the appearance of hypomodified yW on TYW2 epigenetic in- that encompassed the transcription start site (TSS)-associated activation could affect this mechanism, which preserves the proper CpG island. We observed that the 5′-end CpG island of TYW2 in mRNA-to-protein translation. To investigate this phenomenon in the SW48 and HT-29 cell lines was hypermethylated compared our human cancer cells, we used a dual luciferase reporter system with normal colon mucosa (Fig. 1F), whereas the HCT-116 and as described previously (27). In this model, Renilla and firefly lu- SW480 cells were unmethylated (Fig. 1F). These results mim- ciferases are separated by the widely known slippery sequence icked the DNA methylation patterns found by the microarray from HIV (28), and if ribosome slippage occurs, changing the approach used in the Sanger cell line cohort (25) (Fig. 1G). reading frame and inducing a premature stop codon, the firefly Importantly, the TYW2 hypermethylated colon cancer cell lines activity is abolished (Fig. 2D). Using this reporter to monitor the 20786 | www.pnas.org/cgi/doi/10.1073/pnas.2003358117 Rosselló-Tortella et al. MEDICAL SCIENCES Fig. 1. TYW2 promoter CpG island hypermethylation-associated transcriptional silencing in primary colorectal tumors and cancer cell lines. (A) Schematic representation of wybutosine derivative synthesis at position G37 of human tRNAPhe. The enzymes catalyzing these reactions are shown in blue. Abbrevi- ations for the modified G37 nucleoside are as follows: m1G, 1-methylguanosine; imG-14, 4-demethylwyosine; yW-86, 7-aminocarboxypropyl-demethylwyo- sine; yW-72, 7-aminocarboxypropylwyosine; OHyW*, undermodified hydroxywybutosine; OHyW, hydroxywybutosine; o2yW, peroxywybutosine. (B) Frequency of TYW2 hypermethylation in primary tumors derived from TCGA according to cancer type. (C) The presence of TYW2 methylation is significantly associated with loss of expression of the TYW2 transcript in primary colorectal tumors from TCGA. Spearman correlation, P < 0.001. (D) Frequency of TYW2 hypermethylation in cancer cell lines derived from the Sanger panel according tumor type. (E) The presence of TYW2 methylation is significantly associated with loss of expression of the TYW2 transcript in colorectal cancer cell lines from the Sanger panel. Spearman correlation, P < 0.001. (F) Bisulfite genomic sequencing of TYW2 promoter CpG island in colon cancer cell lines and in normal colon. CpG dinucleotides are represented as short vertical lines; the TSS is indicated with a black arrow. Single clones are shown for each sample. The methylation status of each CpG dinucleotide within the interrogated sequence is denoted with a black (methylated) or white (unmethylated) square. (G) DNA methylation profile of the TYW2 promoter CpG island analyzed by the Infinium 450K DNA methylation array. Single CpG absolute methylation β-values are shown (0 to 1). Red, methylated; green, unmethylated. Data from four colon cancer cell lines and two normal colon samples are shown. (H) TYW2 expression levels in methylated (SW48 and HT-29) and unmethylated (HCT-116 and SW480) cancer cell lines determined by qRT-PCR (Top) and Western blot analysis (Bottom). The qRT-PCR data shown represent the mean ± SD of at least three biological replicates and were analyzed using the unpaired two-tailed Student’s t test. ***P < 0.001. (I) Recovery of TYW2 transcript (Top) and protein (Bottom) expression in hypermethylated colon cancer cell lines SW48 and HT-29 on use of the demethylating agent 5-aza-2’-deoxycytidine (AZA). The qRT-PCR data shown represent the mean ± SD of at least three biological replicates and were analyzed using the unpaired two-tailed Student’s t test. ***P < 0.001. reading frame maintenance of the described sequence, we found detected by Western blot analysis, and frameshifting frequency that CRISPR/Cas9-mediated loss of TYW2 in HCT-116 and can be inferred by calculating the percentage of the large protein SW480 cells provoked a significant reduction in firefly luciferase production (frameshift product) over the total protein production activity (Fig. 2D). In the opposite experiment, transfection-mediated (the sum of the dual GST/large protein and the single GST/short restoration of TYW2 in the hypermethylated/silenced cell line protein) (Fig. 2E). Using this reporter, we observed a significant HT-29 increased firefly luciferase activity compared with empty increase in the frameshifting frequency of the TYW2 CRISPR/ Downloaded at Universitat Autonoma de Barcelona (UAB) on February 10, 2021 vector-transfected cells (Fig. 2D). Cas9 KO HCT-116 and SW480 cells compared with the TYW2 We also used a second programmed frameshift reporter based- unmethylated/expressing cells (Fig. 2E). Thus, losing TYW2 and assay to evaluate the production of proteins resulting from −1 hypermodified yW in colon cancer cells confers a phenotype prone PRF events (12). In this dual glutathione S-transferase (GST) to ribosomal frameshifting events. construct, two GST coding sequences are separated by the HIV slippery sequence and an in-frame stop codon before the second Transcriptomic Characterization of RNA Slippery Sequences and GST (Fig. 2E). In the event of ribosome slippage and alteration of Ribosome Frameshift on TYW2 Epigenetic Loss in Cancer Cells. the reading frame, this stop codon is ignored, and the second GST There is a balance between mRNA production and degrada- is translated, generating a larger protein. These proteins can be tion that confers cellular homeostasis and adaptation. Among Rosselló-Tortella et al. PNAS | August 25, 2020 | vol. 117 | no. 34 | 20787 Fig. 2. TYW2 epigenetic silencing in cancer cells mediates loss of hypermodified guanosine in tRNA and impairs ribosomal reading frame maintenance. (A) Nucleoside analysis of tRNAs by LC/MS showing that TYW2 promoter hypermethylation and transcriptional inactivation in SW48 and HT-29 cell lines are associated with complete depletion of the fully modified wybutosine derivatives OHyW and o2yW and induce an accumulation of TYW2 substrate imG-14. TYW2 unmethylated cell lines HCT-116 and SW480 present the fully modified residues and absence of the hypomodified intermediary imG-14. (B) Western blots showing efficient depletion of TYW2 expression in the unmethylated colon cell lines HCT-116 and SW480 on CRISPR/Cas9-mediated deletion (Top) and transfection-mediated recovery of TYW2 in the hypermethylated cell line HT-29 (Bottom). (C) LC/MS showing that CRISPR/Cas9 deletion of TYW2 in HCT-116 and SW480 cells provokes a complete depletion of the fully modified wybutosine (yW) derivatives OHyW and o2yW and accumulation of the TYW2 substrate imG-14. The opposite scenario is observed on transfection-mediated recovery of TYW2 expression in HT-29 cells. WT, wild-type cells; EV, empty vector transfected cells. (D) In vitro assay to measure PRF at an mRNA containing a slippery sequence. (Left) Schematic representation of the dual luciferase reporter used to monitor −1 PRF events in the colon cell line models in which Renilla and firefly luciferases are separated by the slippery sequence from HIV and if ribosome slippage occurs, the reading frame changes and induces a premature stop codon detected by diminished firefly activity. (Right) CRISPR/Cas9-me- diated loss of TYW2 in HCT-116 and SW480 cells caused a significant reduction in firefly luciferase activity, whereas transfection-mediated recovery of TYW2 in the hypermethylated/silenced cell line HT-29 increased firefly luciferase activity. Firefly luciferase activity was compared using an unpaired two-tailed Student’s t test. **P < 0.01. (E, Left) Schematic representation of the dual-GST reporter used to evaluate −1 PRF events in colon cancer cell lines on CRISPR/ Cas9-mediated TYW2 depletion. Two GST coding sequences are separated by the slippery sequence from HIV, and the second GST is translated only if ri- bosome slippage occurs. (E, Middle) Representative Western blot of the reporter construct using anti-HA antibody in HCT-116 and SW480 cell line models. Western blots were performed in triplicate. (E, Right) Quantification of three independent replicates to evaluate the percentage of −1 PRF in each cell line model. The data represent the fraction between the intensity of the dual GST protein band (upper band) against the sum of intensities of the dual GST Downloaded at Universitat Autonoma de Barcelona (UAB) on February 10, 2021 protein and single GST protein (lower band). Statistical differences among proportions were calculated using an unpaired two-tailed Student’s t test. *P < 0.05. the different mechanisms involved in this equilibrium, PRF may by generation of hypomodified guanosine and −1 PRF could serve as an mRNA abundance control mechanism by inducing a induce aberrant mRNA degradation in colon cancer. premature stop codon and mRNA degradation via NMD (16). To explore this, we performed RNA-seq in the TYW2 unme- The physiological status quo for mRNA levels is distorted in thylated/expressing HCT-116 cells and its derived cell line har- human tumors, and we wondered whether TYW2 epigenetic loss boring CRISPR/Cas9-mediated loss of TYW2. We observed that 20788 | www.pnas.org/cgi/doi/10.1073/pnas.2003358117 Rosselló-Tortella et al. the experimentally induced depletion of TYW2 in HCT-116 cells ROBO1 levels (Fig. 3F), indicating that NMD contributes to the altered the levels of 2,370 transcripts, mostly inducing transcript loss of ROBO1 in TYW2-deficient cells. As endogenous positive down-regulation (in 86%; 2,046 of 2,370) (Fig. 3A). Gene set controls for the assay, we used transcripts reportedly targeted by enrichment analysis (GSEA) using Gene Ontology (GO) signa- NMD inhibition (36) (SI Appendix, Fig. S9). As an internal con- ture collections in this set of down-regulated transcripts dem- trol, we did not observe any changes in RNA stability and NMD onstrated an overrepresentation of GO biological processes assays for transcripts that did not change on TYW2 loss in the related to cell migration (e.g., “locomotion,” “cell motility,” RNA-seq analyses (SI Appendix, Fig. S10 A–C). Finally, we further “biological adhesion”) (Fig. 3B). The transcripts down-regulated assessed the impact of TYW2 loss in the generation of ROBO1 on TYW2 CRISPR/Cas9 deletion were also commonly down- frameshifting events by cloning the predicted slippery and stimu- regulated in those colorectal cell lines in which the TYW2 en- latory sequences of ROBO1 in the produced dual luciferase vector zyme was naturally silenced by promoter CpG island hyper- (Fig. 2D), as described in SI Appendix, Fig. S11. We observed that methylation (SI Appendix, Fig. S5). Among the down-regulated the CRISPR/Cas9-mediated loss of TYW2 caused decreased transcripts, 109 (SI Appendix, Table S1) were included in a da- firefly luciferase activity for ROBO1 compared with the wild-type tabase of computationally predicted eukaryotic programmed −1 cells (Fig. 3G). In the opposite model, TYW2 restoration by PRF signals (PRFdb) (29) and contained at least one predicted transfection in the hypermethylated HT-29 cells increased firefly slippery sequence with UUUU/C, the codon decoded by tRNAPhe activity for ROBO1 (Fig. 3G). Importantly, and to ensure that and thus regulated by the hypermodification of G37. Importantly, differences in reading frame maintenance between conditions these transcripts with the −1 PRF signal and the UUUU/C se- were due to specific slippage of tRNAPhe, we also generated a quence were significantly enriched among all down-regulated luciferase reporter for ROBO1 harboring a mutation in the phe- transcripts compared with up-regulated transcripts and all the nylalanine codons in the slippery heptamer (SI Appendix, Fig. genes included in the PRFdb database (Fig. 3C). Serving as an S11). This mutation abolished the effect of TYW2 CRISPR/Cas9 important negative control, the transcripts with slippery sequences deletion (HCT-116 and SW480 KO cells) or TYW2 epigenetic containing AAAA/G, the codon for another frameshift-related silencing (HT-29 hypermethylated cells) on firefly activity tRNA (tRNALys), were not enriched in the transcripts down- MEDICAL SCIENCES (Fig. 3G), similar to the action observed in the TYW2-expressing regulated on TYW2 loss (Fig. 3C). In a similar manner, and as cells such as the unmethylated wild-type HCT-116 and SW80 cells an additional negative control, transcripts that did not contain the or the methylated HT-29 cells in which TYW2 expression was UUU or UUC codons were not enriched in this down-regulated restored by transfection (Fig. 3G). Thus, the loss of the multi- group (SI Appendix, Fig. S6). Interestingly, the proteins encoded faceted cancer gene ROBO1 on TYW2 epigenetic inactivation by the down-regulated transcripts with the −1 PRF signal fre- provides an illustrative example of how alterations of tRNA quently interacted with proteins derived from the down-regulated chemical modifications might contribute to tumorigenesis. transcripts without the motif (SI Appendix, Fig. S7), an event po- tentially affecting RNA fate. Thus, the 109 transcripts identified TYW2 Epigenetic Inactivation Highlights Early-Stage Primary Colorectal constitute bona fide candidates for direct targeting by TYW2 Tumors with Poor Clinical Outcome. Having demonstrated the pres- epigenetic loss in the colon cancer cells. ence of TYW2 promoter CpG island hypermethylation-associated Among the identified TYW2 targets, we selected the round- transcriptional silencing and its role in the generation of hypo- about guidance receptor 1 (ROBO1) for further validation and modified guanosine in tRNAPhe and ribosome frameshift-associated study. ROBO1 was selected because it is reportedly lost in some RNA degradation, we studied if TYW2 epigenetic loss in primary colorectal tumors (30, 31) and because of its proposed role as an colorectal carcinoma had any impact on outcome in these pa- inhibitor of cell migration and epithelial-to-mesenchymal tran- tients. To do so, we data-mined the colorectal carcinoma cases sition (EMT) (32, 33). We first validated the RNA expression from the TCGA effort in which we had previously identified the levels observed in the RNA-seq approach by quantitative real- presence of TYW2 hypermethylation (Fig. 1B) and associated si- time RT-PCR, and found that the TYW2 unmethylated/expressing lencing (Fig. 1C) for the 371 available tumors with complete HCT-116 and SW480 cells expressed the ROBO1 transcript, clinical information. For the entire population of colorectal tu- whereas CRISPR/Cas9-mediated inactivation of TYW2 induced mors, TYW2 promoter hypermethylation showed a trend toward the loss of ROBO1 transcript levels (Fig. 3D). This finding was also an association with poor overall survival, but this was not statis- mirrored at the protein level by Western blot analysis showing that tically significant (P = 0.460, log-rank test) (SI Appendix, Fig. S12). ROBO1 protein was present in the wild-type HCT-116 and However, when subdivided by stages, colorectal cancer patients SW480 cells, whereas it was lost on experimental depletion of with early-stage disease (n = 184, stage I and II) harboring TYW2 TYW2 (Fig. 3D). The reverse profile of ROBO1 was observed in epigenetic inactivation showed significantly reduced overall sur- TYW2 hypermethylated HT-29 cells; the recovery of TYW2 ex- vival (hazard ratio [HR] = 2.984; 95% confidence interval [CI] = pression by transfection raised the RNA and protein levels of 1.26 to 7.08; P = 0.009, log-rank test) (Fig. 4A). Most importantly, ROBO1 compared with cells transfected with the empty vector multivariate Cox regression analysis identified TYW2 hyper- (Fig. 3D). Remarkably, using actinomycin D (Fig. 3E) and alpha- methylation as an independent predictor of shorter overall survival amanitin (SI Appendix, Fig. S8) chase assays, we demonstrated a in early-stage colorectal cancer (HR = 2.69; 95% CI = 1.13 to 6.41; loss of ROBO1 transcript stability on TYW2 CRISPR/Cas9 dele- P = 0.026) (Fig. 4B) compared with other patient characteristics tion in HCT-116 and SW480 cells (Fig. 3E and SI Appendix, Fig. that also have been associated with the clinical outcome (Fig. 4B). S8), but increased ROBO1 transcript stability on TYW2 trans- Reinforcing these data, similar results were obtained when we fection in HT-29 cells (Fig. 3E and SI Appendix, Fig. S8), thus interrogated TYW2 expression levels in the same TCGA colo- supporting the role of TYW2 in the regulation of ROBO1 mRNA. rectal cohort. For the entire population of colorectal cancer Downloaded at Universitat Autonoma de Barcelona (UAB) on February 10, 2021 We considered that the reduced stability of ROBO1 mRNA patients, low TYW2 expression levels showed a trend toward its might be associated with NMD (34, 35) induced by the ribosome association with poor overall survival, but this was not statistically frameshifting originating at a premature stop codon located 52 significant (P = 0.320, log-rank test) (SI Appendix, Fig. S13A). nt upstream of the subsequent exon-exon junction. To test this However, when subdivided by stage, early-stage colorectal cancer idea, we used siRNA targeting UPF1 to deplete the RNA helicase patients with TYW2 low expression levels showed significantly responsible for initiating NMD (34, 35). We observed that the reduced overall survival (HR = 4.78; 95% CI = 2.03 to 11.27; P < inhibition of NMD activity in all the cell line models with TYW2 0.001, log-rank test) (SI Appendix, Fig. S13B). As was shown for loss (i.e., the CRISPR/Cas9-depleted HCT-116 and SW480 cells TYW2 DNA methylation (Fig. 4B), multivariate Cox regression and the hypermethylated HT-29 cells) resulted in up-regulation of analysis identified low TYW2 expression as an independent Rosselló-Tortella et al. PNAS | August 25, 2020 | vol. 117 | no. 34 | 20789 A B C D E F G Fig. 3. Transcriptomic identification of transcripts undergoing degradation on ribosome frameshift-mediated RNA decay and characterization of the ROBO1 candidate. (A) Volcano plot summarizing the results of the RNA-seq experiment to find differentially expressed transcripts in TYW2 CRISPR/Cas9-depleted HCT-116 cells compared with wild-type HCT-116 cells. (B) GO analysis of Biological Process categories in the transcripts down-regulated on TYW2 depletion in HCT-116 cells shows enrichment of GO Biological Process categories related to cell migration, including “locomotion,” “cell motility,” and “biological adhesion.” (C) Down- regulated transcripts in TYW2 KO cells are enriched in −1 PRF sites containing at least one UUUU/C slippery sequences that are present in tRNAPhe (Left); however, no enrichment is observed for transcripts containing AAAA/G slippery sequences, the codon for tRNALys (Right). Statistical differences among proportions were cal- culated using Fisher’s exact test. ns, not significant; *P < 0.05. (D) ROBO1 mRNA and protein levels in the studied experimental models. (Left) TYW2 CRISPR/Cas9 depletion in HCT-116 and SW480 cells down-regulated ROBO1 mRNA as determined by qRT-PCR. (Right) Western blot assay showing the ultimate loss of ROBO1 protein in the TYW2 KO cells. In the opposite model, TYW2 transfection-mediated recovery in HT-29 cells raised ROBO1 transcript and protein levels. For all cases, the qRT-PCR data shown represent the mean ± SD of at least three biological replicates analyzed using an unpaired two-tailed Student’s t test. ***P < 0.001. (E) RNA stability determination by the actinomycin D chase assay. A reduction in ROBO1 transcripts levels is observed on CRISPR/Cas9-mediated deletion of TYW2 HCT-116 and SW480 cells, but not in wild-type cells. In the reverse experiment, TYW2 transfection-mediated restoration in HT-29 cells stabilized ROBO1 transcript levels. For all cases, data shown represent the mean ± SD of at least four biological replicates analyzed using the unpaired two-tailed Student’s t test at each time point. (F) Nonsense-mediated mRNA decay inhibition using a siRNA against UPF1 in TYW2 CRISPR/Cas9-depleted HCT-116 and SW480 cells and in empty vector (EV) transfected HT-29 cells resulted in ROBO1 mRNA up-regulation as assessed by qRT-PCR. Data shown represents the mean ± SD of biological triplicates analyzed by unpaired two- tailed Student’s t test. *P < 0.05; **P < 0.01. (G) In vitro assay to measure PRF events at the cloned ROBO1 mRNA containing the slippery and stimulatory sequences in the generated dual luciferase vector. HCT-116 and SW480 cells harboring CRISPR/Cas9 deletion of TYW2 showed a reduction in firefly luciferase activity compared with wild-type cells. In the reverse model, TYW2 transfection-mediated recovery in HT-29 cells increased firefly activity. No differences in firefly activity between conditions were observed when phenylalanine codons of the slippery heptamer were mutated to leucine. ROBO1 WT, wild-type phenylalanine codons; ROBO1 MUT, mutant-introduced leucine codons. P values correspond to two-tailed unpaired Student’s t test. ns, not significant; *P < 0.05; **P < 0.01. predictor of shorter overall survival in early-stage colorectal can- Using a Transwell assay, we found that CRISPR/cas9-mediated cer (HR = 4.09; 95% CI = 1.71 to 9.79; P = 0.002) compared with Downloaded at Universitat Autonoma de Barcelona (UAB) on February 10, 2021 deletion of TYW2 in the unmethylated and expressing HCT-116 other patient characteristics that also have been associated with and SW480 colon cancer cell lines induced a significant increase in clinical outcome (SI Appendix, Fig. S13C). the migration potential of the obtained TYW2-deficient cells These results could suggest that TYW2 hypermethylation in (Fig. 4C). Most importantly, we wondered whether this type of patients with early-stage colorectal cancer might pinpoint those cellular reprogramming was occurring on TYW2 loss in epithelial tumors that, even at this early stage, contain transformed cells tumors such as colon cancer, in which the migration, cellular de- that are more prone to escape from the primary site and dissem- tachment, and invasion capacities involve the EMT (37, 38). inate the disease. To check this, we went back to our colorectal Most molecular pathways controlling the EMT converge in cancer cell line models to assess their cellular migration capacities. E-cadherin, a cell adhesion protein crucial for maintaining epithelial 20790 | www.pnas.org/cgi/doi/10.1073/pnas.2003358117 Rosselló-Tortella et al. MEDICAL SCIENCES Fig. 4. TYW2 epigenetic loss is associated with early-stage primary colorectal tumors with poor clinical outcome. (A) Kaplan–Meier curve showing that the presence of TYW2 hypermethylation in patients with early-stage colorectal cancer (n = 184) was significantly associated with shorter overall survival (P = 0.009). Green line, unmethylated cases; red line, methylated cases. **P < 0.01. (B) Forest plot representation of the Cox proportional hazard regression models demonstrating that TYW2 hypermethylation is an independent prognostic factor of poor overall survival in patients with early-stage colorectal cancer (HR = 2.69; 95% CI = 1.13 to 6.41; P = 0.026). ns, not significant; *P < 0.05; (C) Transwell assay to determine cell migration capability in the studied models. HCT-116 and SW480 cells carrying the CRISPR/Cas9 deletion of TYW2 showed a significant increase in the number of migrated cells after 48 h compared with wild-type cells. Representative images of the Transwell membrane staining are shown. P values are those associated with an unpaired two-tailed Student’s t test. *P < 0.05. (D) Assessment of EMT features in the studied models using qRT-PCR levels of the E-cadherin (epithelial) and vimentin (mesenchymal) markers. The KOs of TYW2 in the HCT-116 and SW480 cell lines showed a combined decrease in E-cadherin expression and enhancement Vimentin levels compared with wild- type cells, a transcript profile suggestive of the acquisition of EMT features. Data shown represent the mean ± SD of at least three biological replicates and were analyzed using an unpaired two-tailed Student’s t test. *P < 0.05; **P < 0.01; ***P < 0.001. structure (37, 38), whereas expression of vimentin is a hallmark of increased migration capacity and the emergence of EMT features, mesenchymal features (37, 38). In this regard, we found that TYW2 both of which have central roles in metastasis development, the CRISPR/Cas9-mediated deletion in both HCT-116 and SW480 leading cause of cancer-associated mortality. These observations are colon cancer cells induced down-regulation of E-cadherin and up- in line with the identified poor clinical outcomes of colorectal cancer regulation of vimentin (Fig. 4D). Interestingly, and in line with the patients with TYW2 hypermethylation-associated inactivation. role of TYW2 loss in ROBO1 down-regulation (Fig. 3 D–G) and in Discussion Downloaded at Universitat Autonoma de Barcelona (UAB) on February 10, 2021 the proposed EMT inhibitory activity of this protein (32, 33), res- toration of ROBO1 expression by transfection in the TYW2 tRNAs are extremely critical biological molecules that, working CRISPR/Cas9-deleted HCT-116 and SW480 cells rescued the phe- together with the ribosomes, allow translation of the genetic notype, inducing the opposite effect with E-cadherin up-regulation code into amino acids. In humans, there are 47 different types of and vimentin down-regulation (SI Appendix, Fig. S14 A and B). tRNAs for 20 standard amino acids, in addition to selenocysteine CRISPR/cas9-mediated deletion of TYW2 did not change the cell (39). The amount of each tRNA is tightly controlled to match the cycle, as assessed by BrdU and 7AAD incorporation, p21 expression codon usage of mRNAs to optimize protein expression according as a marker of cell arrest, or apoptosis measured by PARP cleavage to the cellular necessities. For many years, it was thought that (SI Appendix, Fig. S15 A–C). Thus, TYW2 loss is associated with tRNAs were passive bystanders in cellular transformation, and Rosselló-Tortella et al. PNAS | August 25, 2020 | vol. 117 | no. 34 | 20791 that changes in tRNA levels were simply a consequence of the rescue TYW2 deficiency, restoring ribosome and RNA base high proliferative status of tumoral cells. However, an increasing pairing, and avoid the migration-prone features of these cells, amount of new data show that tRNA might play an active role in thereby reducing the mortality of the identified high-risk patients tumorigenesis (40, 41). These results include, among others, the that present with an otherwise apparently localized disease. presence of mitochondrial tRNA mutations in cancer cells (42), the antiapoptotic activity of tRNAs (43), and the detection of Materials and Methods tRNA derivatives in transformed cells that exert oncogenic or The materials and methods used in this study are summarized here; more tumor-suppressor functions (44, 45). detailed information is available in SI Appendix, Materials and Methods. The Another important twist involving tRNAs in cancer origin and sequences of all primers, as well as references for all antibodies, reagents development is the recognition that an aberrant pattern of chem- and commercial assays, are provided in SI Appendix, Table S2. ical modifications in the tRNA molecule occurs in neoplasia, causing altered protein translation (1, 10). These findings can now DNA Methylation Analyses. Gene promoter DNA methylation status was de- be categorized within the emerging field of epitranscriptomics, termined by DNA methylation microarrays and bisulfite genomic sequencing. defined as the chemical modifications that affect RNA activity. This area is booming because RNA modifications seem to play an Expression Analyses. RNA expression was assessed by qRT-PCR, and protein expression was determined by Western blot analysis. essential role in cellular differentiation (46) as well as in tumori- genesis (1, 47). Human RNAs are modified in many ways (e.g., TYW2 Cellular Models. In unmethylated HCT-116 and SW480 cell lines, TYW2 5-methylcytosine, N6-methyladenosine, N1-methyladenosine, pseu- expression was abolished using the CRISPR/Cas9 system as described previ- douridine, etc.) (48), and disruptions of the proteins involved in ously (57). For stable TYW2 restoration in the hypermethylated HT-29 cell adding, removing, or binding to these chemical marks are being line, the TYW2 cDNA sequence was cloned into pLVX-IRES-ZsGreen plasmid identified in human cancers (49), such as the ribosomal RNA (Clontech) and delivered using lentivirus containing this construct. modifiers DKC1 (50) and NSUN5 (51). In this regard, tRNA itself constitutes one of the most heavily modified molecules, including tRNA Nucleosides LC-MS. tRNA from cellular models were extracted and some highly specific nucleotides (6). Examples of tRNA modifiers digested, and nucleosides were analyzed by liquid chromatography-mass involved in carcinogenesis include ELP3, which is up-regulated in spectrometry as described previously (26). different tumor types and acts on the Wnt pathway (52) and also has been related to metastasis formation (53); the tRNA methyl- −1 PRF Evaluation. −1 PRF events were evaluated using a dual luciferase transferase 9B, which has been proposed to exert a role in tumor reporter as described previously (27). In brief, Renilla and firefly luciferases suppression (54); and the tRNA methyltransferase NSUN2, which were cloned into the pcDNA4 T/O vector, separated by the slippery sequence undergoes gene amplification-associated overexpression in human of interest. In the event of a −1 PRF event, the new reading frame will tumors (55). Interestingly, modifications within the anticodon stem- generate a premature stop codon, and firefly activity will be abolished. loop at position 37 of tRNA is crucial for a correct base pairing interaction (56), and deficient hypermodification of G37 in tRNA- RNA-Seq. Total RNA from wild-type and CRISPR/Cas9-mediated TYW2 KO Phe HCT-116 cells was used for RNA-seq analyses. Genes were considered to be , without any evident cause, was reported in cancer models de- differentially expressed when the log2 fold change was <−1.0 or >1.0 and the cades ago (17–19). Here we demonstrate that the generation of adjusted P value was < 0.05. GO biological processes gene sets included in the hypomodified G37 in tRNAPhe is due to the cancer-specific epige- GSEA signature database were used to perform an overrepresentation analysis netic inactivation of the tRNA- modifying enzyme TYW2. on the down-regulated transcripts. Coding genes were queried to the PRFdb Promoter CpG island hypermethylation-associated silencing (29) and interrogated for the presence of putative slippery sequences. of TYW2 was observed across a wide spectrum of human cancer types, with colorectal tumors the most common type of tumor in Migration Assays. Cell migration capacity was assessed by a Transwell assay. which this epigenetic defect was identified. Most importantly, the loss of TYW2 was associated not only with the lack of hyper- Cell Cycle and Apoptosis Assays. The cell cycle was assessed by measuring BrdU modified forms of G37 (e.g., OHyW, o2yW), but also with in- incorporation and 7AAD staining, by immunoblotting of p21 and PARP cleavage. creased ribosome frameshift, leading to transcript degradation. The transcriptomic assay showed that numerous RNAs were Statistics. Spearman’s correlation, unpaired Student’s t test, or Fisher’s exact targeted by this tumor-specific event induced on TYW2 silenc- test were used as appropriate. Kaplan–Meier plots and the log-rank ing. The tumor suppressor ROBO1 is a robust example of key (Mantel–Cox) test were used to estimate overall survival. P < 0.05 was con- genes that can be hampered by this process. Even more impor- sidered to indicate statistical significance. tant from the clinical side, the DNA methylation-linked loss of TYW2 was associated with poor clinical outcome in colorectal Data Availability. All study data are included in the main text and SI Appendix. cancer patients at early stages of the disease. Thus, TYW2 could ACKNOWLEDGMENTS. Funding for this research was provided by the be proposed as a biomarker of early dissemination for those Secretariat for Universities and Research of the Ministry of Business and tumors that at first glance seem confined to the primary site but Knowledge of the Government of Catalonia (2017SGR1080); Ministerio de in reality might have escaped from its original niche. These re- Ciencia, Innovación y Universidades, Agencia Estatal de Investigación, Fondo sults could be linked to the observed appearance of higher mi- Europeo de Desarrollo Regional (RTI2018-094049-B-I00 and SAF2014-55000); gration rates and mesenchymal features in those cancer cells the Asociación Española Contra el Cáncer Scientific Foundation (Accelerator Award A26825); and the Olga Torres Foundation. Institutional support was harboring TYW2 epigenetic silencing. Given the emergence of provided by Centres de Recerca de Catalunya Programme/Generalitat de many epigenetic drugs that restore the activity of genes inacti- Catalunya. M.R.-T. is a fellow of the Instituto de Salud Carlos III (contratos vated by DNA hypermethylation (24), it is also tempting to Predoctorales de Formación en Investigación y Salud IFI17/00006). M.E. is an Downloaded at Universitat Autonoma de Barcelona (UAB) on February 10, 2021 speculate that the use of inhibitors of DNA methylation would Institució Catalana de Recerca i Estudis Avançats Research Professor. 1. M. Esteller, P. P. Pandolfi, The epitranscriptome of noncoding RNAs in cancer. Cancer 5. M. Duechler, G. Leszczyn ska, E. Sochacka, B. Nawrot, Nucleoside modifications in the Discov. 7, 359–368 (2017). regulation of gene expression: Focus on tRNA. Cell. Mol. Life Sci. 73, 3075–3095 2. B. Linder, S. R. Jaffrey, Discovering and mapping the modified nucleotides that comprise (2016). the epitranscriptome of mRNA. Cold Spring Harb. Perspect. Biol. 11, a032201 (2019). 6. T. Pan, Modifications and functional genomics of human transfer RNA. Cell Res. 28, 3. M. A. Machnicka et al., MODOMICS: A database of RNA modification pathways–2013 395–404 (2018). update. Nucleic Acids Res. 41, D262–D267 (2013). 7. T. Suzuki, Biosynthesis and function of tRNA wobble modifications in Fine-tuning of 4. C. Gu, T. J. Begley, P. C. Dedon, tRNA modifications regulate translation during cel- RNA Functions by Modification and Editing, H. Grosjean, Eds. (Springer, 2005), pp. lular stress. FEBS Lett. 588, 4287–4296 (2014). 23–69. 20792 | www.pnas.org/cgi/doi/10.1073/pnas.2003358117 Rosselló-Tortella et al. 8. T. Suzuki, A. Nagao, T. Suzuki, Human mitochondrial tRNAs: Biogenesis, function, 31. G. A. Rezniczek et al., ROBO1 expression in metastasizing breast and ovarian cancer: structural aspects, and diseases. Annu. Rev. Genet. 45, 299–329 (2011). SLIT2-induced chemotaxis requires heparan sulfates (heparin). Anticancer Res. 39, 9. A. G. Torres, E. Batlle, L. Ribas de Pouplana, Role of tRNA modifications in human 1267–1273 (2019). diseases. Trends Mol. Med. 20, 306–314 (2014). 32. P. Nguemgo Kouam et al., Robo1 and vimentin regulate radiation-induced motility of 10. L. Endres, M. Fasullo, R. Rose, tRNA modification and cancer: Potential for therapeutic human glioblastoma cells. PLoS One 13, e0198508 (2018). prevention and intervention. Future Med. Chem. 11, 885–900 (2019). 33. Y. Xia et al., Reduced USP33 expression in gastric cancer decreases inhibitory effects 11. U. L. RajBhandary, S. H. Chang, Studies on polynucleotides. LXXXII yeast phenylala- of Slit2-Robo1 signalling on cell migration and EMT. Cell Prolif. 52, e12606 (2019). nine transfer ribonucleic acid: Partial digestion with ribonuclease T1 and derivation of 34. F. Usuki et al., Specific inhibition of nonsense-mediated mRNA decay components, the total primary structure. J. Mol. Biol. 243, 598–608 (1968). SMG-1 or Upf1, rescues the phenotype of Ullrich disease fibroblasts. Mol. Ther. 14, 12. B. A. Carlson et al., Transfer RNA modification status influences retroviral ribosomal 351–360 (2006). frameshifting. Virology 255, 2–8 (1999). 35. N. Hug, D. Longman, J. F. Cáceres, Mechanism and regulation of the nonsense- 13. A. L. Konevega et al., Purine bases at position 37 of tRNA stabilize codon-anticodon mediated decay pathway. Nucleic Acids Res. 44, 1483–1495 (2016). 36. J. T. Mendell, N. A. Sharifi, J. L. Meyers, F. Martinez-Murillo, H. C. Dietz, Nonsense interaction in the ribosomal A site by stacking and Mg2+-dependent interactions. surveillance regulates expression of diverse classes of mammalian transcripts and RNA 10, 90–101 (2004). mutes genomic noise. Nat. Genet. 36, 1073–1078 (2004). 14. J. D. Dinman, Mechanisms and implications of programmed translational frame- 37. H. Peinado, D. Olmeda, A. Cano, Snail, Zeb and bHLH factors in tumour progression: shifting. Wiley Interdiscip. Rev. RNA 3, 661–673 (2012). An alliance against the epithelial phenotype? Nat. Rev. Cancer 7, 415–428 (2007). 15. B. A. Carlson et al., 1-Methylguanosine in place of Y base at position 37 in phenyl- 38. T. Brabletz, R. Kalluri, M. A. Nieto, R. A. Weinberg, EMT in cancer. Nat. Rev. Cancer 18, alanine tRNA is responsible for its shiftiness in retroviral ribosomal frameshifting. 128–134 (2018). Virology 279, 130–135 (2001). 39. P. P. Chan, T. M. Lowe, GtRNAdb 2.0: An expanded database of transfer RNA genes 16. V. M. Advani, J. D. Dinman, Reprogramming the genetic code: The emerging role of identified in complete and draft genomes. Nucleic Acids Res. 44, D184–D189 (2016). ribosomal frameshifting in regulating cellular gene expression. BioEssays 38, 21–26 40. S. Q. Huang et al., The dysregulation of tRNAs and tRNA derivatives in cancer. J. Exp. (2016). Clin. Cancer Res. 37, 101 (2018). 17. D. Grunberger, I. B. Weinstein, J. F. Mushinski, Deficiency of the Y base in a hepatoma 41. M. Santos, A. Fidalgo, A. S. Varanda, C. Oliveira, M. A. S. Santos, tRNA deregulation phenylalanine tRNA. Nature 253, 66–67 (1975). and its consequences in cancer. Trends Mol. Med. 25, 853–865 (2019). 18. J. F. Mushinski, M. Marini, Tumor-associated phenylalanyl transfer RNA found in a 42. C. Florentz, B. Sohm, P. Tryoen-Tóth, J. Pütz, M. Sissler, Human mitochondrial tRNAs in wide spectrum of rat and mouse tumors but absent in normal adult, fetal, and re- health and disease. Cell. Mol. Life Sci. 60, 1356–1375 (2003). generating tissues. Cancer Res. 39, 1253–1258 (1979). 43. Y. Mei et al., tRNA binds to cytochrome c and inhibits caspase activation. Mol. Cell 37, 19. Y. Kuchino, E. Borek, D. Grunberger, J. F. Mushinski, S. Nishimura, Changes of post- 668–678 (2010). MEDICAL SCIENCES transcriptional modification of wye base in tumor-specific tRNAPhe. Nucleic Acids Res. 44. R. L. Maute et al., tRNA-derived microRNA modulates proliferation and the DNA 10, 6421–6432 (1982). damage response and is down-regulated in B cell lymphoma. Proc. Natl. Acad. Sci. 20. L. Droogmans, H. Grosjean, Enzymatic conversion of guanosine 3′ adjacent to the U.S.A. 110, 1404–1409 (2013). anticodon of yeast tRNAPhe to N1-methylguanosine and the wye nucleoside: De- 45. V. Balatti et al., tsRNA signatures in cancer. Proc. Natl. Acad. Sci. U.S.A. 114, pendence on the anticodon sequence. EMBO J. 6, 477–483 (1987). 8071–8076 (2017). 21. A. Noma, Y. Kirino, Y. Ikeuchi, T. Suzuki, Biosynthesis of wybutosine, a hyper- 46. M. Frye, B. T. Haranda, M. Behm, C. He, Expression during development. Science 361, modified nucleoside in eukaryotic phenylalanine tRNA. EMBO J. 25, 2142–2154 1346–1349 (2018). (2006). 47. S. Delaunay, M. Frye, RNA modifications regulating cell fate in cancer. Nat. Cell Biol. 22. A. Noma et al., Expanding role of the jumonji C domain as an RNA hydroxylase. J. Biol. 21, 552–559 (2019). Chem. 285, 34503–34507 (2010). 48. V. Davalos, S. Blanco, M. Esteller, SnapShot: Messenger RNA modifications. Cell 174, 23. S. B. Baylin, P. A. Jones, Epigenetic determinants of cancer. Cold Spring Harb. Per- 498–498.e1 (2018). 49. R. Esteve-Puig, A. Bueno-Costa, M. Esteller, Writers, readers and erasers of RNA spect. Biol. 8, a019505 (2016). modifications in cancer. Cancer Lett. 474, 127–137 (2020). 24. M. Berdasco, M. Esteller, Clinical epigenetics: Seizing opportunities for translation. 50. D. Ruggero et al., Dyskeratosis congenita and cancer in mice deficient in ribosomal Nat. Rev. Genet. 20, 109–127 (2019). RNA modification. Science 299, 259–262 (2003). 25. F. Iorio et al., A landscape of pharmacogenomic interactions in cancer. Cell 166, 51. M. Janin et al., Epigenetic loss of RNA-methyltransferase NSUN5 in glioma targets 740–754 (2016). ribosomes to drive a stress adaptive translational program. Acta Neuropathol. 138, 26. Y. Sakaguchi, K. Miyauchi, B. I. Kang, T. Suzuki, “Nucleoside analysis by hydrophilic 1053–1074 (2019). interaction liquid chromatography coupled with mass spectrometry” in Methods in 52. A. Ladang et al., Elp3 drives Wnt-dependent tumor initiation and regeneration in the Enzymology, C. He Eds. (Elsevier, ed. 1, 2015), pp. 19–28. intestine. J. Exp. Med. 212, 2057–2075 (2015). 27. G. Grentzmann, J. A. Ingram, P. J. Kelly, R. F. Gesteland, J. F. Atkins, A dual-luciferase 53. S. Delaunay et al., Elp3 links tRNA modification to IRES-dependent translation of LEF1 reporter system for studying recoding signals. RNA 4, 479–486 (1998). to sustain metastasis in breast cancer. J. Exp. Med. 213, 2503–2523 (2016). 28. C. Penno, R. Kumari, P. V. Baranov, D. van Sinderen, J. F. Atkins, Specific reverse 54. U. Begley et al., A human tRNA methyltransferase 9-like protein prevents tumour transcriptase slippage at the HIV ribosomal frameshift sequence: Potential implica- growth by regulating LIN9 and HIF1-α. EMBO Mol. Med. 5, 366–383 (2013). tions for modulation of GagPol synthesis. Nucleic Acids Res. 45, 10156–10167 (2017). 55. M. Frye et al., Genomic gain of 5p15 leads to over-expression of Misu (NSUN2) in 29. A. T. Belew, N. L. Hepler, J. L. Jacobs, J. D. Dinman, PRFdb: A database of computa- breast cancer. Cancer Lett. 289, 71–80 (2010). tionally predicted eukaryotic programmed −1 ribosomal frameshift signals. BMC 56. P. F. Agris, Bringing order to translation: The contributions of transfer RNA anticodon- Genom. 9, 339 (2008). domain modifications. EMBO Rep. 9, 629–635 (2008). 30. J. V. Tricoli et al., A mutational comparison of adult and adolescent and young adult 57. F. A. Ran et al., Genome engineering using the CRISPR-Cas9 system. Nat. Protoc. 8, (AYA) colon cancer. Cancer 124, 1070–1082 (2018). 2281–2308 (2013). Downloaded at Universitat Autonoma de Barcelona (UAB) on February 10, 2021 Rosselló-Tortella et al. PNAS | August 25, 2020 | vol. 117 | no. 34 | 20793