Uveal melanoma (UM) is the most common primary intraocular tumor in adults. miR-16 peaks falling within those genes were called following the procedure described by Srandour et al (2012). The prediction of non-canonical pairs had lower accuracy, but followed the same trend (Fig 5). Structures composed of highly probable pairs had higher positive predictive value than structures predicted with the MFE method. Prediction using this model is implemented in the MC-Fold program [12]. Also shown is prediction accuracy using evolutionary couplings from the plmc program [16] (green). Four NN tables, W, WL, WMB, and WMBL, store components of multibranched structures [36]. This improved the prediction of non-canonical pairs with the MFE method to a PPV of 59.2% and a sensitivity of 76.8%, sufficiently accurate to develop hypotheses about structure (Fig 3A; S2 Table). UM carries mutually exclusive mutations that trigger overactivity of the Gq pathway (G protein subunit alpha q [GNAQ], G protein subunit alpha 11 [GNA11], Phospholipase C Beta 4 [PLCB4], or Cysteinyl Leukotriene Receptor 2 [CYSLTR2]) (Robertson et al, 2018). Our model is based on a constant level of miR-16. The absolute quantification (copy number) of miR-16 in 501Mel was determined by Northern blot in Gilot et al (2017). The formulation of the partition function algorithm is similar to the minimum free energy algorithm for the NCM model [33], with the restriction that care must be taken to count each structure exactly once. Data Availability: Software is available under the GNU GPL at http://rna.urmc.rochester.edu. Cell density was measured by methylene blue colorimetric assay (Gilot et al, 2017). We do not capture any email address. A complete repertoire of base pairing types was built from the detected H-bonds of all X-ray crystal structures of a resolution of 3.0 Aor better, including the large and small ribosomal subunits. Prior to the calculation, all array values are initialized to 0 except for W5(1) and W3(N), which are initialized to 1 to represent the equilibrium constant of the completely unfolded state (which is the reference state with folding free energy of zero). This result suggests that a bulge may be created when the seed sequence of miR-16 binds to the sponge RNAs. Altogether, these results indicate that miR-16 activity (assessed using miR-16 sponges and target expression levels) is a useful marker for clinicians, in contrast to miR-16 expression. To test this hypothesis, we first investigated the tumor suppressor activity of miR-16 in UM cells by elevating miR-16 expression levels. In the implementation of CycleFold, the previously determined weight of 0.3 was used. Given sufficient human expertise, manual structure modeling can produce high-quality predictions of structure. non-canonical pairs, such as mismatched nucleotides at the end of helices, but it does not explicitly make predictions of non-canonical pairs. The non-canonical base pairs describe all association of bases that are different from those between A-U and G-C base pairs [3,4,5,6,7,8,9,10]. (F) Basal expression of the 30 most expressed genes at the basal level (from the selected genes represents in the graph c [Fig S3B]). The authors thank La ligue Contre le Cancer. It is common to find a modest number of non-canonical base pairs in these structures in addition to the usual Watson-Crick pairs. Prediction accuracy is shown for structures composed of highly probable pairs using information from a single sequence (blue) or a TurboFold calculation with 10 sequences (red). In these benchmarks for predicting conserved pairs, base triples are considered two separate pairs, where one pair can be a canonical pair. DOI: 10.26508/lsa.202201643. Historically, the location of the canonical base pairs has been known with high accuracy before the tertiary structure was solved, so the canonical secondary structure can be used to help infer tertiary structure, including non-canonical base pairs [7,52]. (H) miR-16 quantification by RT-qPCR after miRNA transfection in two UM cell lines. In this work, a partition function algorithm is introduced that allows the estimation of base pairing probabilities for both canonical and non-canonical interactions. The lower left triangle shows the pairs that are present in the reference structure. Non-canonical pairs can be predicted using a knowledge-based model of structure that scores nucleotide cyclic motifs, or NCMs. Corrections, Expressions of Concern, and Retractions. Because the function of up-regulated miR-16 interactants is poorly studied, we first validated their miR-16mediated increase in other cell lines (Mel202, 92-1 and HCT116 KO DROSHA) as done for miR-16 targets. Additional experiments are needed to explain this phenomenon. A detailed description of the algorithm is provided in Materials and Methods, below. David H. Mathews, Affiliation: WL has the additional restriction that the nucleotide at j is required to be one of the closing base pairs of the helix. Here, it is shown that the additional information from a partition function calculation can be used to identify the predicted pairs more likely to be correct. To further challenge the miR-16 sequestration hypothesis while preserving the stoichiometry between miR-16 and its interactome, we next depleted PYGB mRNA and quantified endogenous miR-16 targets (Fig 4G and H) selected as a function of: (i) a miR-16dependent mRNA decay (Fig 3), (ii) presence of a predicted MRE-16 in their 3UTR (Fig 2), and (iii) a decrease in MP41 cell density after their depletion (Fig 3D). To elucidate the molecular mechanism, explaining the translation up-regulation mediated by miR-16, we looked for miR-16binding sites on PYGB mRNA using the RNAhybrid webtool (Krger & Rehmsmeier, 2006). KM analyses have been performed for potential miR-16 sponges RNA and targets RNA according to the clusters 1 and 2 defined in Figs 5A and S2A and Table S1. (H) mRNA expression levels of miR-16 targets: AMOT, TACC1, NRBP1, DNAJB4, and WEE1 after PYGB mRNA depletion in MP41 cells (shPYGB relative to shCTR) (n = 3 biologically independent experiments). However, additional genetic events occur, such as BES alterations, BRCA1-associated protein 1 (BAP1), Eukaryotic Translation Initiation Factor 1A X-Linked (EIF1AX) and Splicing Factor 3b Subunit 1 (SF3B1), and recurrent copy number variations (CNVs). Prior work had emphasized the relative inaccuracy of the prediction of the NCM model when used with a dynamic programming algorithm to search the folding space [33]. CycleFold provides a dramatic improvement in accuracy over previously available methods, and its output could be used to refine three dimensional structure predictions from any modeling software. RNA plays a central role in all of life, acting as both a carrier of genetic information and as an active participant in numerous cellular processes, including pre-mRNA splicing and gene regulation via modulation of transcription and translation [1]. (B) Graph a: plot showing the miR-16 effects on miR-16interacting RNAs (up or down-regulated). Lentiviral particles carrying shRNA vectors targeting human PYGB mRNA (shPYGB, TL310025V), and scramble shRNA (shCTR, TR30021V) were purchased from Origen. No significant difference was found. n = 3, 4, and 3 biologically independent experiments, respectively. Consequently, the canonical miR-16 activity, involved in the RNA decay of oncogenes, such as cyclin D3, is impaired. [LSA-2022-01643_TableS3.xlsx]. All mimics were purchased from Dharmacon. where is a fitted parameter that quantifies the weight associated with the extrinsic information. A ceRNA hypothesis: The rosetta stone of a hidden RNA language? All statistical analyses were performed using Prism 8 software (GraphPad) or Microsoft Excel software. Nevertheless, experiments based on cross-linked immunoprecipitation (CLIP) and alternative methods, performed by different teams, have identified such interactions (Loeb et al, 2012; Helwak et al, 2013; Grosswendt et al, 2014; Luna et al, 2015). Accurate predictions of extended secondary structure as provided by this method could prove useful as additional modeling constraints for any tertiary structure modeling program. (B) The probability dot plot calculated using Cyclefold with the partition function algorithm. Briefly, cells were fixed for 30 min with 70% ethanol. The interpretation of ceRNA results is still a topic of debate because the stoichiometry between miRNA and RNAs (sponges) is not (or rarely) investigated. HCT116 WT and HCT116 KO DROSHA (Kim et al, 2016b) cell lines were obtained from Korean Collection for Type Cultures (KCTC), Microbial Resource Center. Overall survival after subdivision into low (blue) and high (red) expression groups (below or above median expression of the gene). Altogether, these results suggest that the miR-16 targets identified in the MP41 cell line are highly conserved in the same cancer models (UM), despite the diversity of driver mutations and chromosome abnormal copy number that characterize these three cell lines. The KaplanMeier method was used to estimate the survival distributions. Left heat map illustrating the down-regulated RNAs in response to miR-16 transfection in MP41 (0, 6, 12, 24, 48 h). An analogous calculation is used to calculate EBA, the extrinsic information about A from B. Graph c represents same genes of the graph c without those suspected to be false positive because of their detection with biotinylated miR-CTR (threshold 500 reads in the RNAseq for miR-CTR) (n = 403 genes). Each histogram represents the mean + SD; bilateral Student test (with non-equivalent variances) **P < 0.01; ***P < 0.001. Pair probability estimates can be further improved by predicting the structure conserved across multiple homologous sequences using the TurboFold algorithm. This represents a significant improvement in PPV as compared to unconstrained MFE prediction using CycleFold (p < 0.05; S5 Table). KaplanMeier analyses for each potential sponge from the Signature S4 (LIPA, FLNC, TSPAN14, and NLE1). (C) mRNA expression levels of potential miR-16 targets (down-regulated genes), 72 h after transfection of miR-16 or miR-CTR in HCT116 DROSHA KO. Analyses were carried out with the survival and survivalROC R packages. 2023 Life Science Alliance LLC. Next, we assessed the translational consequence of miR-16 binding on these up-regulated miR-16 interactants because miRNA binding on canonical binding sites commonly provokes a translation blockade and induces RNA decay via seed base-pairing (Bartel, 2004). For many RNA molecules, the . The size and the median survival of each group are specified (with 95% confidence interval between brackets). All unique/stable reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement. For example, the structure of the Group I self-splicing intron was modeled and subsequently many of the atomic interactions were confirmed by crystal structures [710]. The KaplanMeier method was used to estimate the survival distributions. All of these methods can use sparse information from experimental methods or low-resolution computational structure prediction, such as prediction of secondary structure, to reduce the search space of the tertiary structure problem and improve accuracy. Department of Biochemistry & Biophysics and Center for RNA Biology, University of Rochester Medical Center, Rochester, NY, United States of America, Two subsequent implementations of this calculation, MC-Fold-DP [33] and MC-Flashfold [34], use a dynamic programming algorithm, making structure calculations feasible for long sequences (>150 nucleotides). Non-canonical base pairs can be explicitly predicted using an alternative model of RNA structure that scores Nucleotide Cyclic Motifs, or NCMs [12,32], instead of thermodynamic nearest neighbor parameters. The native structures as shown in the crystal structures are provided as Supplementary S1 Fig. Sequences are available in Table S3. Although UM is considered to be a G proteincoupled receptor disease with BES alterations and CNVs, at least three other cellular processes seem to be frequently deregulated. CycleFold (as well as MC-Fold and MC-Fold-DP) is able to accept a set of input pairs, and predict structures that are required to contain those pairs. This term is evaluated for the junction of NCM a with NCM b, NCM b with NCM c, and NCM c with NCM d. In the NCM model, the pseudo-free energy of a particular structure is given by Table S4. Data are presented as mean SD unless otherwise specified, and differences were considered significant at a P-value of less than 0.05. miR-16 activity, assessment using our RNA signature, discriminates the patients OS as effectively as current methods. The biological function of up-regulated miRNA interactants remains unclear in the literature even though they have already been highlighted by different teams (Vasudevan et al, 2007). Alternatively, predicted non-canonical pairs could be used in order to help identify RNA folding modules, which typically have a set of known canonical and non-canonical interactions. Base-pairing was evaluated for wild-type (WT) MRE-16 (non-canonical MRE-16 #1 and #2) from PYGB mRNA. For our signature, all 57 genes are potential miR-16 interactants. The translation efficiency of these chimeric RNAs is estimated by assessing the luciferase activity. In accordance with this hypothesis, a recent study demonstrated that miRNAs loaded in Ago2 are enriched in the 3UTR of several RNAs without inducing their decay. We observed a down-regulation of these miR-16 targets in all cell lines except MYB in the Mel202 cell line. The picture is representative of n = 3, 2, and 3 biologically independent experiments, respectively. Consensus sequences of miRNA-binding site motifs (3-5) enriched in cluster of up-regulated miR-16 interactants after miRNA pull-down in MP41. (B) mRNA expression levels of potential miR-16 sponges 72 h after transfection of synthetic miR-16 relative to miR-CTR in HCT116 DROSHA knock-out cells. This method allows assessment of confidence in a predicted secondary structure [36], estimates of accessibility of a region of secondary structure [3739], prediction of pseudoknots [40,41], and identification of single-nucleotide polymorphisms that are expected to affect structure [4245]. (E) Quantification of miR-16 expression in three UM cell lines (MP41, Mel202, and 92-1) and primary human melanocytes by RT-qPCR, compared with cutaneous melanoma cell line (501Mel). From each set of sequences, 9 were chosen randomly to be used by TurboFold. (An American ophthalmological society thesis), CeRNA cross-talk in cancer: When ce-bling rivalries go awry, General rules for functional microRNA targeting, Re-evaluation of the roles of DROSHA, Exportin 5, and DICER in microRNA biogenesis, A Network of noncoding regulatory RNAs acts in the mammalian brain, RNAhybrid: MicroRNA target prediction easy, fast and flexible, Identification of novel genes coding for small expressed RNAs, Capture of microRNA-bound mRNAs identifies the tumor suppressor miR-34a as a regulator of growth factor signaling, High PTP4A3 phosphatase expression correlates with metastatic risk in uveal melanoma patients, Analysis of the proteasome activity and the turnover of the serotonin receptor 2B (HTR2B) in human uveal melanoma, MiR-16 and miR-26a target checkpoint kinases Wee1 and Chk1 in response to p53 activation by genotoxic stress, An 11-gene-based prognostic signature for uveal melanoma metastasis based on gene expression and DNA methylation profile, miR-16 family induces cell cycle arrest by regulating multiple cell cycle genes, Cistrome: An integrative platform for transcriptional regulation studies, Transcriptome-wide miR-155 binding map reveals widespread noncanonical microRNA targeting, Hepatitis C virus RNA functionally sequesters miR-122, Prognostic implications of novel ten-gene signature in uveal melanoma, Molecular characteristics of liver metastases from uveal melanoma, Definition and identification of small RNA sponges: Focus on miRNA sequestration, miRNPs: A novel class of ribonucleoproteins containing numerous microRNAs, Integrated analyses identify potential prognostic markers for uveal melanoma, Targeting primary and metastatic uveal melanoma with a G protein inhibitor, Single-cell RNA sequencing reveals intratumoral heterogeneity in primary uveal melanomas and identifies HES6 as a driver of the metastatic disease, Fast and effective prediction of microRNA/target duplexes, Integrative analysis identifies four molecular and clinical subsets in uveal melanoma. Atomic resolution RNA structures are being published at an increasing rate. Support was provided by a Ligue Nationale Contre le Cancer (LNCC) fellowship and French Ministry of Research (Ministre franais de lEnseignement suprieur, de la Recherche et de lInnovation) fellowship (AM Qumner). Analyses were performed using Cistrome SeqPos. Cluster 1 is associated with poor clinical outcome (chromosome 3 monosomy, metastasis). Using the NCM partition function implemented in CycleFold, structures were assembled of canonical pairs that had probabilities higher than a threshold [36]. Base pairs were considered to be correct if they are predicted within one nucleotide for one index. UM is usually a diploid tumor with recurrent CNV of whole chromosomes or arms. To discard artefactual interactants due to nonspecific background inherent to all biotinylated experiments (Lal et al, 2011; Tan & Lieberman, 2016; Dash et al, 2018), we analyzed our data in two steps. This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Crystal structures, to which the predictions are compared, generally capture only one pairing state, and therefore requiring exact matches the crystal structure would not be a good assessment of performance. The pair probabilities for non-canonical pairs, however, can still be high, i.e. W and WL store the partition function for the fragment from i to j where i and j are not required to be paired to one another, and the fragment contains one helix. The CycleFold partition function calculation can also be constrained to require specified base pairs. Non-canonical UU base pair is ubiquitous in ncRNA, and Watson-Crick U-A pair can often be replaced with U-G wobble pair without significant duplex destablization, which increases structure and . However, additional biochemical experiments are need to further characterize the miR-16 interactions with these potential miRNA sponges as well as the potential miRNA targets identified here. We summarized the criteria describing a miRNA sponge in 2017 (Migault et al, 2017). F Mouriaux: writingoriginal draft, review, and editing. The base pairing types are labeled using an extension of the nomenclature recently introduced by Leontis and Westhof. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. This method implicitly considers the free energy change of some non-canonical pairs, such as mismatched nucleotides at the end of helices, but it does not explicitly make predictions of non-canonical pairs. Experimental methods to determine RNA tertiary structures at atomic resolution, such as NMR, X-ray crystallography, and cryo-electron microscopy, remain expensive, difficult, and time-consuming, so computational structure prediction methods, which have proven effective in predicting unknown protein structures [5,6], are desirable to help investigators understand RNAs of unknown structure. Evolutionary information was used recently to infer non-canonical and long-distance contacts using evolutionary coupling with the plmc program, and these contacts were then used to generate structure models [16]. Base pairs were found from the coordinates using 3DNA-DSSR version 1.1.2, all cis-Watson-Watson pairs are called canonical. At higher thresholds, too few pairs are predicted, and there is insufficient statistical power to for a significant result (Fig 3B; S7 Table). This is permitted because these alternative pairs are thermodynamically accessible according to solution NMR experiments, which show sampling of alternative base pairing of this kind [60,61], and optical melting experiments of bulges, which are consistent with multiple pairing states [62]. In other words, miRNA can be inactivated when sponged or sequestered in RNA via imperfect binding (also known as non-canonical binding). (D) Schematic representation of the 3UTR of PYGB mRNA containing two non-canonical MRE-16 predicted by RNAHybrid. The bolt region on the miR-16 sequence corresponds to the seed region of the miRNA. miRNA was isolated using mirVana miRNA isolation kit (Ambion; Life Technologies). At equilibrium, the probability of structure, s, is given by. Instead of promoting RNA decay of miR-16 targets such as cell cycle regulators (CCND3, CCND1, and WEE1), non-canonical miR-16 activity mediates the expression of pro-tumoral genes (acceleration effect). Determining the structure of an RNA is crucial to understand its function, and provides opportunities to design drugs that target the molecule, but the structure of the vast majority of RNA molecules is entirely unknown. Annotated genes associated with de novo motifs were identified. Dynamic programming algorithms exist that are able to explicitly predict pseudoknots, albeit at higher computational cost [41,57], and a dynamic programming method performed slightly better in benchmarks of pseudoknot prediction [40]. In general, predicting only probable pairs or predicting conserved pairs with TurboFold increases the PPV, at a cost of sensitivity. However, they do not reflect physical properties of the sequence, rather the log of the frequency with which these NCMs are observed in the training set. [LSA-2022-01643_TableS1.xlsx]. Structures were chosen that were derived from coordinates in the Protein Data Bank [2], and that were not used as part of the MC-Fold training data [12]. miR-16 interactome and signatures. A significant effort on a systematic analysis of the non-canonical base pairs is still far from sufficient both in base pair and base pair step level. Predictions of RNA tertiary structure tend to be least accurate in loop regions, where non-canonical pairs are important for determining the details of structure. Life Science Alliance is registered as a trademark in the U.S. Patent and Trade Mark Officeand in the European Union IntellectualProperty Office. Counterintuitively, increasing evidence indicates that miRNAs bind to a subset of RNA without inducing their decay. All siRNAs were purchased from IDT DNA. University of Rennes, Centre National de la Recherche Scientifique (CNRS), Institut de Gntique et Dveloppement de Rennes (IGDR) - UMR 6290, Rennes, France, INSERM U1242, University of Rennes, Rennes, France, SPARTE, University of Rennes, CNRS, IGDR - UMR 6290, Rennes, France, Service dOphtalmologie, CHU de Rennes, Rennes, France, NSERM U1241, Universit Rennes, INRAE, Institut NuMeCan (Nutrition, Metabolisms and Cancer), Rennes, France, Laboratoire de Cytogntique et Biologie Cellulaire, CHU Rennes, Rennes, France, Laboratory of Preclinical Investigation, Translational Research Department, Institut Curie, PSL Research University, Paris, France, Curie, Department of Medical Oncology, PSL Research University, Paris, France, Fdration Hospitalo Universitaire-OncoAge, CNRS, Institut de Pharmacologie Molculaire et Cellulaire, Universit Cte dAzur, Valbonne, France, CHU Rennes, Service de Gntique Molculaire et Gnomique, Rennes, France. The authors thank the Gene Expression and Oncogenesis team from the Centre National de la Recherche Scientifique (CNRS) UMR6290, Dr Pascal Loyer from NuMeCan (INSERM U1241), BIOSIT core facilities of Rennes 1 University (SFR UMS CNRS 3480 INSERM 018, especially P Gripon for the BSL3), the UCA GenomiX platform of Institut de Pharmacologie Molculaire et Cellulaire, and the Centre de Ressources Biologiques humaines Sant (especially C Pangault) for their help. An example of a pseudo-energy calculation using the NCM model. To increase robustness, this package also selects survival-associated genes by repetition (1,000 times) of a separation between the training and validation sets of samples. The modeled putRNA from U2* to U74* contains twenty-one Watson-Crick (WC) base pairs and five non-canonical base pairs, A9*-G35* (Saenger class VIII), G12*-U32* (Saenger class XXVIII), G43*-A64 . In V(i,j,), W(i,j), WL(i,j), WMB(i,j), and WMBL(i,j), entries where ij refer to exterior fragments, containing the ends of the sequence. where ncms is the set of all the NCMs in the structure, and junctions is the set of all junctions of NCMs in the structure such that two NCMs share a base pair.