A Polysome-Based microRNA Screen Identifies miR-24-3p as a Novel Promigratory miRNA in Mesothelioma

It is well established that miRNAs are powerful regulators of gene expression and induce translational repression (1, 2). miRNA-mediated gene silencing, guided by sequence complementarity to their mRNA targets, consists of several steps: mRNA deadenylation, mRNA degradation, and translation inhibition (3). The temporal sequence of these steps is not totally clear. Some evidences show that translational repression is an early common event, later followed by mRNA destabilization (1, 3–6). Initiation of translation is a rate-limiting event, and most studies suggest that miRNAs repress translation at the initiation phase (3, 7). Translation initiation is controlled by eukaryotic initiation factors (eIF). Two eIFs exert a predominant role on tumorigenesis, the eIF4F complex (7, 8) and eIF6 (9, 10). eIF4F complex is composed by the cap-binding protein eIF4E, the RNA helicase eIF4A, and the scaffold protein eIF4G. eIF4E recruitment in the eIF4F complex is inhibited by dephosphorylated 4E-binding proteins (4E-BP). mTORC1 phosphorylates 4E-BPs and relieves their inhibition (7, 11, 12). Importantly, the availability of eIF4E is required for efficient translational repression by miRNAs (7). As for eIF6, the evidence that it may affect miRNA repression of translation is dubious (13).

Translated mRNAs are in large molecular complexes known as polysomes. Translational repression of a specific mRNA is accompanied by its loss from translating ribosomes. It is evident that a miRNA, which targets a specific coding mRNA that is recruited to polysomes for translation, must transiently associate with polyribosomes. In spite of this, studies have not taken advantage from the mechanistic feature of miRNA dynamic association with translated mRNAs, to isolate functional miRNAs that counteract the translation of their targets. Here, we evaluated whether polysomal association of miRNAs could allow us to identify oncogenic miRNAs relevant for cancer progression.

Malignant pleural mesothelioma (MPM) is an incurable cancer (14), characterized by clinical heterogeneity and variability at the transcriptional level with at least three subtypes that do not share extensive transcriptional signatures (15). This suggests that other mechanisms controlling gene expression, such as translational regulation, may be important in generating oncogenic properties. Previous works have shown that the translational machinery of MPM cells is deregulated, with aberrant initiation checkpoints (16, 17). The mTOR pathway, inhibited by rapamycin, drives oncogenic signaling to the translational machinery, and MPM is largely insensitive to rapamycin inhibition (16). Activation of eIF4F complex by mTOR kinase is a signal for cell growth and transformation (18). Because studies on MPM suggest that eIF4E activity is constitutive (19) and its availability is required for miRNAs translation inhibition (7, 8), it is likely that in MPM, miRNAs hijack the translational machinery for the repression of specific targets.

Previous works on MPM cells have shown a remarkable heterogeneity in miRNA expression. Different sample sources, variability of investigated histologic types, or heterogeneity of control groups led to divergent results (20). Alternatively, we hypothesize that miRNAs are oncogenic for their presence on the translational machinery, where they are in equilibrium with their targets, rather than for their simple expression or loss. By using an approach based on the definition of miRNA dynamically associated with polysomes, we discovered miR-24-3p, a new miRNA affecting MPM biology. Strikingly, manipulation of miR-24-3p levels in various cell types and conditions, led to a consistent biological effect on their migration properties. We suggest that miR-24-3p shows oncogenic properties in MPM, and hypothesize that our miRNA screen may unveil relevant connections in cancer studies.

All experiments involving mice were performed in accordance with national and EU regulations. Experimental protocols were reviewed and approved by the local Institutional Animal Care and Use Committees (IACUC no 688). Adult immunocompromised NOD-SCID male mice (Charles River Laboratories) were used for detecting tumor growth after intraperitoneal (i.p.) injection of REN cells manipulated for miR-24-3p expression level, as indicated. A cohort of these mice was used to identify the overall survival rate. The remaining experimental mice were sacrificed at 30 days postinjection to analyze tumor development and metastasis formation.

In this study, we used cell lines from human MPM: REN (21) and H28 (ATCC CRL-5820) for epithelioid subtypes, MM98 (22) for sarcomatoid subtypes, and MSTO-211H (ATCC CRL-2081) for biphasic subtypes. MeT-5A cells (ATCC CRL-9444) are immortalized mesothelial cells used as nontumorigenic control. MCF-7 cells (ATCC HTB22) are metastatic breast cancer cells. 293T cells (ATCC CRL-3216) were used for in vitro luciferase assay. Each cell line was grown in DMEM (Lonza) except for MSTO-211H and MeT-5A cells, which were grown in RPMI1640. All media were supplemented with 10% FBS and 1% penicillin, streptomycin, l-glutamine, and cells were maintained at 37°C and 5% CO₂. Meso 3T, 7T, 8T, 9T, 12T, 13T, 14T, and 17T are human primary mesothelioma cells established in the MRC Toxicology Unit (Leicester, United Kingdom). Primary cell lines were authenticated by short tandem repeat (STR) DNA profiling (23). All cell lines were tested by PCR for Mycoplasma detection and were not cultured for more than 2 months.

The following antibodies were used: rabbit polyclonal anti-cingulin (HPA027587, Sigma), mouse mAbs against RhoA (240302, Cell Biolabs), and β-actin (A5441, Sigma). For apoptosis analysis, staurosporine was bought from Sigma. For mRNA stability experiments, Actinomycin D was bought from Thermo Fisher Scientific.

REN, MM98, and MSTO-211H cells were stably transduced either with a miR-Zip construct expressing a short hairpin inhibiting miR-24-3p or an empty vector, used as control. MeT-5A cells were permanently infected with a miR construct expressing a miR-24-3p mimic or a scramble vector used as control. All lentiviral vectors used in this study were provided by System Biosciences.

Total RNA was prepared using mirVana (Thermo Fisher Scientific) and treated with RNase-free Dnase I (Promega). cDNA was generated using the SuperScript III First-Strand Synthesis kit (Thermo Fisher Scientific). Real-time PCR amplification was conducted in StepOnePlus Real-Time PCR System (Thermo Fisher Scientific) using SYBR Green technique. The following primers were used to amplify selected genes: CGN (forward: CCCGGCTAGGGACTCCTC; reverse: TGGACTCCATGGTCTACGGG), CXADR (forward: GCTGTCGTAAAAAGCGCAGA; reverse: AGTGGACGTACGGCTCTTTG), BCAM (forward: GAGACGCCAGCTTCCACTG; reverse: GTGGGATAGTGCAGGGTGAG). ACTIN (forward: AGAGCTACGAGCTGCCTGAC; reverse: CGTGGATGCCACAGGACT) was used as endogenous control to normalize samples using the ΔΔC_t method. For the analysis of miRNA expression levels, cDNA was synthesized using the Universal cDNA Synthesis (Exiqon) and qRT-PCR was performed with miRCURY LNA Universal RT microRNA PCR (Exiqon) on an ABIPRISM 7900HT Sequence Detection System (Applied Biosystems). U6 snRNA was used as internal control to normalize samples by ΔΔC_t method. miR-24-3p expression level in primary mesothelioma cells was established by TaqMan Assay (Thermo Fisher Scientific) following the standard procedure. miR-24-3p assay ID is 000402. The internal control U6B snRNA assay ID is 001093.

Growing cells were lysed in 50 mmol/L Tris HCl pH 7.5, 100 mmol/L NaCl, 30 mmol/L MgCl₂, 0.1% Nonidet P-40, 5 mmol/L DTT, 100 μg/mL cycloheximide and 30 U/ml RNasin, and cytoplasmic lysates were clarified at 15,000 × g for 15 minutes at 4°C. To distinguish miRNAs that sedimented with heavy polysomes, but were not physically bound to them, we disrupted polysomes by adding 30 mmol/L EDTA after lysate preparation.

Equal amounts of RNA were loaded on 10%–50% sucrose gradient dissolved in 25 mmol/L Tris HCl pH 7.4, 25 mmol/L NaCl, 5 mmol/L MgCl₂,1 mmol/L DTT, with or without 10 mmol/L EDTA, and centrifuged at 39,000 rpm for 3 hours in a SW41Ti Beckman rotor at 4°C. The gradient was then analyzed by continuous flow absorbance at 254 nm, recorded by BioLogic LP software (Bio-Rad), and fractions were collected and kept on ice to avoid RNA degradation. The following fractions were collected: (i) subpolysomal pool, from the top of the gradient to the 80S peak; (ii) light polysomes and (iii) heavy polysomes. Samples were incubated with proteinase K and SDS 1% for 1 hour at 37°C. RNA was extracted by phenol/chloroform/isoamyilic alcohol method. The same defined amounts of synthetic RNA spike-ins (osa-miR-414, ath-miR-159a, cel-miR-248) are added to all samples during preparation, to normalize RNA measurement.

RNAs were processed with the NanoString nCounter Human v2 miRNA Expression Assay (NanoString) in the Genomics Shared Resource at The Ohio State University (Columbus, OH). The miRNA v2 panel can detect 699 endogenous miRNAs (with 654 probes), 5 housekeeping transcripts, and 5 Spike-in specific for non-mammalian small RNA transcript used as controls when processing RNAs. One-hundred nanograms of total RNA were annealed with multiplexed DNA tags (miR-tag) and target-specific bridges. Mature miRNAs were bound to specific miR-tags using a ligase enzyme and all the tags in excess were removed by enzymatic clean-up step. The tagged miRNA product was diluted 1 to 5 and 5 μL was combined with 20 μL of reported probes in hybridization buffer and 5 μL of capture probes. The overnight hybridization (16–20 hours) at 65°C allowed for complex sequence-specific probes with targets, and probe excess was removed using two-step magnetic beads–based purification on an automated fluidic handling system (nCounter Prep Station). Target/probe complexes were immobilized on the cartridge for data collection. The nCounter Digital Analyzer collected data by taking images of immobilized fluorescent reporters in the sample cartridge with a CCD camera through a microscope objective lens. For each cartridge, a high-density scan encompassing 600 fields of view was performed.

Proliferation rate of MPM cells was analyzed by CellTiter-Blue assay (G8080, Promega). Briefly, 4 × 10⁴ cells were plated at 100 μL/well in a 96-well plate, and assayed after 24, 48, and 72 hours. CTB was added and left on cells for 3 hours at 37°C and 5% CO₂. The number of viable cells was measured recording fluorescence at 560Ex/590Em nm in a Victor3 Multilabel Counter (PerkinElmer). Cell-cycle analysis was performed on 48-hour serum-starved MPM cells, restimulated with serum. At the indicated time points, cells were fixed, stained with propidium iodide (PI, P1304MP, Thermo Fisher Scientific), and acquired on a BD FACSCANTO II flow cytometer (BD Biosciences). Cell-cycle analysis was performed using the FCS Express software. Cell death detection was performed using APC-Annexin V (BioLegend) and recorded on a BD FACS ARIA flow cytometer (BD Biosciences).

Cells were plated in a 12-well plate at 6 × 10⁵ cells/well. Cells were grown to a confluency of 90% and then a thin wound was introduced by scratching with a sterile pipette tip. The wound closure was monitored on a Nikon Eclipse Timelapse every 4 hours and analysis was performed using NIS-Elements Viewer.

Total RNA from miR-24-3p knockdown REN and MM98 cells was isolated using mirVana Isolation Kit (Thermo Fisher Scientific). RNA quality was controlled with BioAnalyzer (Agilent Technologies). Random primed cDNA libraries were constructed from 100 ng of total RNA. A paired-end (2 × 125) run was performed on a Genome Sequencer Illumina HiSeq 2500 (GATC Biotech) with 50 million reads per sample.

Three biological replicates were analyzed for REN and MM98 cell lines. Raw reads were checked for quality by FastQC software (version 0.11.2, S., A.FastQC: a quality control tool for high throughput sequence data. 2010; available from: http://www.bioinformatics.babraham.ac.uk/projects/fastqc), and filtered to remove low-quality calls by Trimmomatic (version 0.32) using default parameters and specifying a minimum length of 50. Processed reads were then aligned to Homo sapiens genome assembly GRCm38 (Ensembl version 83) with STAR software (version 2.4.1c).

The number of reads per gene was counted while mapping using the option quantMode GeneCounts from the STAR software. A read is counted if it overlaps (1 nt or more) one and only one gene. Both ends of the paired-end read are checked for overlaps. The counts coincide with those produced by HTSeq-count algorithm with default parameters (gene annotation release 83 from Ensembl). To estimate differential expression, the matrix of gene counts produced by HTSeq was analyzed by DESeq2 (version DESeq2_1.12.4). The differential expression analysis by DeSeq2 algorithm was performed on the entire dataset composed by both REN and MM98 samples. The two following comparisons were analyzed: REN miR-Zip24 versus REN miR-empty and MM98 miR-Zip24 versus MM98 miR-empty. Reads counts were normalized by calculating a size factor, as implemented in DESeq2. Independent filtering procedure was then applied, setting the threshold to the 60 percentile; 24,178 genes were therefore tested for differential expression. Significantly modulated genes in REN miR-Zip24 and MM98 miR-Zip24 conditions were selected by considering a false discovery rate lower than 10%. Regularized logarithmic (rlog) transformed values were used for heatmap representation of gene expression profiles. Analyses were performed in R version 3.3.1 (2016-06-21, Computing, T.R.F.f.S. R: A Language and Environment for Statistical Computing; available from: http://www.R-project.org/).

The Gene Ontology enrichment analysis was performed using topGO R Bioconductor package (version topGO_2.24.0) as described in ref. 24. The annFUN.db function is used to extract the gene-to-GO mappings from the genome-wide annotation library org.Hs.eg.db for H. Sapiens.

Pull-down assay to detect RhoA-GTP was performed using RhoA activation assay kit (Cell Biolabs). REN and MeT-5A cells were transduced with lentiviral vectors to inhibit or overexpress miR-24-3p, as before. Cells in growing conditions were lysed for 24 hours after plating and GTP pulldown assay was performed for 1 hour at 4°C. After washing, samples were resuspended in SDS-PAGE sample buffer and loaded in a polyacrylamide gel.

The plasmids containing wild-type (catalog no. HmiT088611-MT06) or mutant (catalog no. CS-HmiT088611-MT06) Luc-CGN 3′ UTR, wild-type (catalog no. HmiT064331-MT06) or mutant (catalog no. CS-HmiT064331-MT06) Luc-CXADR 3′ UTR and wild-type (catalog no. HmiT010922-MT06-01) or mutant (catalog no. CS-HmiT010922–MT06-01) Luc-BCAM 3′ UTR were synthesized (GeneCopoeia) and used in luciferase reporter assays. These plasmids contain firefly luciferase, fused to the 3′ UTR of analyzed genes, and Renilla luciferase used as a reporter gene. Luciferase activity assays were performed following the manufacturer's protocol. In detail, 293T cells were seeded in 12-well plates and cotransfected with miRNA mimic for hsa-miR-24-3p (MC10737, Ambion) and either wild-type or 3′ UTR-mutated genes. Cells were harvested after 48 hours and the firefly and Renilla luciferase activities were measured sequentially using the Dual Luciferase Assay Kit (GeneCopoeia). Luciferase activities were normalized with Renilla luciferase. Experiments were run in triplicate.

Each experiment was repeated at least three times, as biological replicates; means and SDs between different experiments were calculated. Statistical P values obtained by two-tailed Student t test were indicated: ***, P < 0.001; **, P < 0.01; *, P < 0.05.

The RNA seq data are available at www.ebi.ac.uk/arrayexpress with accession number ID: E-MTAB-7079.

We hypothesized that subcellular miRNA localization is essential to mediate an oncogenic effect. Considering the relevance of translational control in cancer biology (8), we aimed at the specific isolation of miRNAs physically associated with the translational machinery. We used as a model system the malignant mesothelioma cell line REN, in which we previously observed dysregulated translation characterized by insensitivity to mTOR inhibition (16). In this cell line, eIF4E, necessary for translational repression by miRNAs at initiation (8), is at higher levels compared with its inhibitors 4E-BPs, explaining the insensitivity to mTOR inhibition. We expected that miRNAs that act by repressing translation were in continuous equilibrium with their target mRNAs and could cosediment with polysomes. Growing REN cells had well-defined polysomes (Supplementary Fig. S1A). We used fractionation of polysomes by sucrose density-gradient sedimentation, followed by high-throughput analysis of associated miRNAs (Fig. 1A). We collected subpolysomal, light polysomal, and heavy polysomal fractions into three separate pools (Supplementary Fig. S1A). Subpolysomal fractions contain untranslated mRNAs and soluble miRNAs. Heavy polysomal fractions contain highly translated mRNAs; light polysomes may contain a mixture of the two, and we considered them as a cushion phase (25, 26). Next, we performed a high-throughput analysis of miRNAs by the NanoString nCounter System (NanoString), and measured the expression levels of 699 miRNAs. Only 66 miRNAs were associated with the translation machinery of REN cells, indicating they are in equilibrium with their targets (Fig. 1B). Because of the fact that the presence of miRNAs in the polysomal fraction could be due to interactions with high molecular weight structures, we refined the analysis on EDTA-treated polysomes. EDTA treatment disrupts polysomes, and leads to their dissociation in free 40S and 60S ribosomal subunits (Supplementary Fig. S1B). Concomitant with polysomes’ disaggregation, few miRNAs were EDTA-insensitive, indicating that they were not associated with polysomes, whereas several miRNAs showed an expected shift from polyribosomes to monosomes. The latter were further analyzed. To evaluate the preference of a miRNA for polysomal association, we calculated the miRNA polysome occupancy (25). We observe that the bulk of miRNAs are characterized by a medium occupancy, which shrinks after EDTA treatment (Supplementary Fig. S1C). Finally, ratio of EDTA-treated/EDTA untreated miRNA polysome occupancy indicates that a restricted set of miRNAs dramatically changes its distribution, like miR-15a-5p, miR-16-5p, miR-24-3p, miR-191-5p, and let-7g-5p (Fig. 1C). Notably, three of these miRNAs were already linked to MPM through previous analysis, miR-15/16 (27, 28) and miR-191-5p (29). These miRNAs represent bona fide highly active miRNAs that act in a dynamic equilibrium with polysomes and their target mRNAs.

To assess whether miRNA expression levels influenced miRNA polysome occupancy, we performed a correlation analysis between miRNAs expression level and miRNAs occupancy (Supplementary Fig. S1D). Importantly, miRNA occupancy and expression were not associated (R² = 0.35182), indicating that the presence of miRNAs on the polyribosomes is due to other variables, such as the mRNA target level. If this last hypothesis was correct, we reasoned that in silico analysis of miRNA potential targets (http://mirsystem.cgm.ntu.edu.tw/) may lead to a common signature pathway. The most significant polysome-enriched miRNAs (miR-15a-5p, miR-16-5p, miR-24-3p, miR-191-5p and let-7g-5p) were commonly predicted to regulate genes in the cell cycle and in the TCR pathway (Supplementary Fig. S1E). We conclude that our screen identifies a minimal set of miRNAs that, by being in active equilibrium with polysomes, may dynamically control at the posttranscriptional level important biological functions in REN cells.

Establishing that only a restricted number of miRNAs is associated with the polysomal machinery, we evaluated miRNAs expression in multiple MPM samples. We performed qRT-PCR on MPM cell lines, representative of the different MPM histologic subtypes: REN and H28 for epithelial type, MM98 for sarcomatoid type, and MSTO-211H for biphasic one. Each analyzed miRNA, except miR-24-3p, displayed a variable expression. miR-24-3p was found upregulated in all MPM cell lines analyzed, compared with MeT-5A control cells (Fig. 2A). miR-24-3p was the only one not previously identified in MPM tumors by other approaches and was further analyzed. Patient-derived primary mesothelioma cells well represent the native state of the tumor and reflect the disease heterogeneity observed in patients (23). We used as control, a primary untransformed human mesothelium. We found that miR-24-3p is expressed above the control in all primary mesothelioma cells analyzed (Fig. 2B), as well in other tumor cell lines (Supplementary Fig. S2). By qRT-PCR, we validated the association of miR-24-3p on polyribosomes of cells. We confirmed that EDTA causes a 5-fold loss of miR-24-3p expression from highly translating ribosomes, from 30% to 6%, and an increase of its expression in monosomes (Fig. 2C). In conclusion, we show the upregulation of miR-24-3p in MPM and its association with actively translated mRNAs, suggesting that a polysome-enrichment analytic approach efficiently identifies biologically active miRNAs that may have escaped previous works.

To investigate the biological role of miR-24-3p in MPM, we stably knocked down miR-24-3p in three MPM cell lines, representative of the main histopathologic subtypes (REN cells for the epithelioid subtypes, MM98 cells for the sarcomatoid, and MSTO-211H cells for mixed type). All cell lines were sorted for GFP coreporter, and miR-24-3p knockdown was confirmed by quantitative RT-PCR. The efficiency of inhibition was around 70% in all cell lines analyzed (Supplementary Fig. S3A). Next, we analyzed MPM cell growth at 72 hours after plating and upon miR-24-3p depletion. The reduction of miR-24-3p slightly reduces cell number relative to control (Supplementary Fig. S3B), does not affect the apoptotic rate (Supplementary Fig. S3C), and causes a mild reduction of cells in S-phase (Supplementary Fig. S3D). Next, we investigated migratory properties of MPM cells by performing wound-healing assays. All MPM cell lines depleted of miR-24-3p had a lowered migratory capability, as compared with controls (Fig. 3A–D; Supplementary Fig. S4A). To further challenge the promigratory role of miR-24-3p, we used the mesothelial control cell line MeT-5A that expresses little amounts of miR-24-3p. We transduced MeT-5A cells with a Lenti-miR vector containing miR-24-3p, obtaining 3.5-fold expression, compared with control (Supplementary Fig. S4B). Scratch assay analysis showed that MeT-5A cells transduced with miR-24-3p acquired higher migratory capability (Fig. 3E and F). Next, we investigated the migratory role of miR-24-3p in MCF-7 cells, a metastatic breast cancer cell line, analyzing both miR-24-3p knockdown and overexpression. We found that in MCF-7 cells, the migration capability is positively correlated with miR-24-3p levels (Supplementary Fig. S4C and S4D). In conclusion, our analysis demonstrates that a miRNA isolated by its association with polysomes has a consistent biological effect in all tested cells.

The effect of miR-24-3p on MPM cells’ proliferation and migration led us to investigate the role of miR-24-3p on tumor growth in vivo. We injected intraperitoneally REN miR-Zip24 and REN miR-empty control cells into immunocompromised NOD-SCID mice and evaluated overall survival, tumor volumes, and metastasis formation. As reported in a Kaplan–Meier curve, a negative correlation between miR-24-3p expression and mice overall survival was found (Fig. 4A). A subset of mice were sacrificed at 30 days postinjection and analyzed. At autopsy, we observed a consistent reduction of metastasis formation and intraperitoneal hemorrhage in mice injected with REN cells depleted of miR-24-3p (Fig. 4B). We measured total body, tumor mass, and diaphragm weight and found that knockdown of miR-24-3p causes a significant reduction in tumor volume in vivo (Fig. 4C; Supplementary Table S1). These results confirm that miR-24-3p may function as a pro-oncogenic miRNA also in vivo.

Several properties of malignant progression and migration have been attributed to aberrant signaling through RHO-GTPase proteins. The Rho-GTPases play crucial role in cell migration, tumor growth, and metastasis (30). We investigated RhoA activity in cells depleted for miR-24-3p or overexpressing it. RhoA-GTP pulldown assays showed that miR-24-3p knockdown causes a reduction of RhoA activity, in REN cells (Fig. 5A), whereas miR-24-3p overexpression in MeT-5A cells caused an increment of RhoA activity (Fig. 5B). We noted that the modulation of RhoA activity is independent from RhoA mRNA and protein levels (Fig. 5C and D). These data suggest that miR-24-3p indirectly regulates RhoA activity in MPM.

Next, we searched for miR-24-3p direct targets that explain the promigratory phenotype. We aimed at defining genes regulated at the translation level by miR-24-3p in an unbiased way. Technical issues complicate a direct analysis of translation targets. We exploited the well-known observation that miRNA translational repression of mRNA targets is followed by their degradation (4, 28). We reasoned that if translational control is involved in a dynamic fashion (equilibrium between a miRNA on polysomes and continuous transcription of the targeted mRNA), we may unveil situations in which a reduction in the target mRNA is present, but weaker than the reduction in the encoded protein. Thus, we performed RNA-seq analysis of REN and MM98 in which miR-24-3p was knocked down. We observed that the genes up and downregulated in the two cell lines analyzed were highly dissimilar at the steady-state mRNA level. Nevertheless, most miR-24-3p–modulated genes were protein-coding genes (Supplementary Fig. S5A).

To bring order in this scenario we clustered, by Gene Ontology analysis, differentially expressed genes in functional groups and we observed that miR-24-3p modulates several mRNAs involved in cell adhesion and migration in REN and MM98 cells (Fig. 6A). Thus, in spite of an apparent heterogeneous signature, the biological landscape evoked by miR-24-3p is similar, with a contribution of targets involved in migration. Next, we overlapped all genes upregulated by mir-24-3p depletion in REN and MM98 cells and found that the two MPM cell lines share 194 genes (Supplementary Table S2), which we arbitrarily considered biologically relevant for explaining the consistent biological results obtained by miR-24-3p manipulation in multiple cell lines. To further restrict targets, we performed a miRNA target prediction analysis using 5 different miRNA prediction software (PITA, MiRWalk, RNAHybrid, Targtescan, miRanda) and we matched the obtained list with the 194 common genes derived from RNA-seq analysis. Thus, we shortlisted 33 genes that are bona fide direct targets of miR-24-3p (Supplementary Fig. S5B). We then asked whether these genes are involved in cancer progression and are shared among different tumor types. We interrogated PRECOG (PREdiction of Clinical Outcomes from Genomic profiles) system to analyze the association between the 33 miR-24-3p–regulated genes and cancer outcomes. Many genes identified as miR-24-3p direct targets correlate with favorable survival across different cancer types (Supplementary Fig. S5C).

Our RNA-seq data indicate that miR-24-3p hits relevant biological pathways in cancer, including cell motility and migration. To validate NGS results and identify miR-24-3p direct target mRNAs, we analyzed genes with specific involvement in cell migration and tight junction structures. We scanned the 3′ UTRs of three selected genes, CGN, BCAM, and CXADR, for potential miR-24-3p–binding sites, and identified in their 3′ UTR sequence elements complementary to the seed sequence of human miR-24-3p (Fig. 6B–D, left). We hypothesized that miR-24-3p binding to CGN, BCAM, and CXADR mRNAs may inhibit their expression. In support of this idea, miR-24-3p expression in 293T cells is able to specifically decrease the activity of a luciferase reporter gene fused to the wild-type 3′ UTR of CGN, whereas the presence of a mutant form of CGN 3′ UTR has no effect on the luciferase activity (Fig. 6B, right). We conclude that the effect of miR-24-3p on the luciferase mRNA translation is dependent on the presence of miR-24-3p–binding site within the 3′ UTR of the gene. Similar results were obtained for BCAM and CXADR genes (Fig. 6C and D).

Next, we validated the changes of cingulin, BCAM, and CXADR mRNAs in REN, MM98, and MSTO-211H cells, after miR-24-3p partial depletion (Fig. 6E). In summary, we observed an increment of the expression levels of these gene transcripts in miR-24-3p knockdown MPM cell lines (Fig. 6E; Supplementary Fig. S5D). We tested in parallel the protein levels of cingulin, BCAM, and CXADR. Unfortunately, commercial antibodies against BCAM and CXADR did not perform well and were discarded. Data are therefore shown only for cingulin target. As expected, we found that miR-24-3p reduction induces an increase in cingulin protein levels in REN, MM98, and MSTO-211H (Fig. 6F). Next, we explored cingulin, BCAM, and CXADR mRNAs and cingulin protein levels in miR-24-3p–overexpressing MeT-5A cells and found a partial reduction of target mRNAs expression (Fig. 6G; Supplementary Fig. S5E) accompanied by a total loss of cingulin protein (Fig. 6H). The total loss of cingulin protein in miR-24-3p overexpression conditions (Fig. 6H), combined with a partial reduction of cingulin mRNA transcript levels, suggests that, as hypothesized, miR-24-3p is able to regulate cingulin expression first at the translation level, and subsequently at the mRNA stability level (Supplementary Fig. S5F).

Next, we addressed the relative contributions of translation and mRNA stability. We first evaluated the association of cingulin mRNA with polysomes, when miR-24-3p is expressed at either low or high levels. REN cells with normal or reduced miR-24-3p expression levels were characterized by similar polysomal profiles, indicating that miR-24-3p has no effect on global translation (Fig. 7A). We analyzed cingulin mRNA distribution along the polysomal profile of normal or miR-24-3p knockdown REN cells and found that reduced levels of miR-24-3p cause an accumulation of cingulin mRNA in highly translating polysomes (Fig. 7B). The effect was consistently accompanied by a reduction of cingulin mRNA in the soluble fraction (Fig. 7B). Identical results were obtained for BCAM and CXADR mRNAs (Supplementary Fig. S6A and S6B). Next, we investigated the distribution of cingulin mRNA in MeT-5A cells that expressed miR-24-3p, compared with control. Strikingly, when miR-24-3p is overexpressed, the residual transcript moves from heavy polysomes to light ribosomes, confirming that its translation is inhibited (Fig. 7C). We then evaluated whether the loss of polysomal association of cingulin mRNA with polysomes in REN, MM98, and MSTO-211H cells, also results in its destabilization. Data show that cingulin mRNA has a shorter half-life in cells with increased miR-24-3p. In summary, our data demonstrate that miR-24-3p strongly acts at the translational level by its association with target mRNAs on polysomes, followed by mRNA destabilization.

Finally, the striking regulation of cingulin driven by miR-24-3p led us to address whether cingulin expression is related to the survival of patients with MPM. Analysis of CGN expression in MPM tissues, according to the SynTarget tool (27, 29), predicts that high levels of cingulin are a predictor for survival (Fig. 7D).

Several works have identified miRNAs involved in cancer biology by their differential expression in tumor versus control, and correlation between miRNA expression and prognosis. In some cases, miRNAs have been identified for their cellular localization, as in exosomes. One prominent effect of miRNAs is their effect on translational repression. Starting from this observation, we reasoned that miRNAs that dynamically modify gene expression counteract the translation of continuously transcribed genes and associate with the polysomal machinery. In this work, we suggest a screening approach for the identification of biologically active miRNAs in cancer through a rapid analysis of polysome-associated miRNAs. In this study, we confirm miRNAs previously associated with MPM. In addition, we identified miR-24-3p, a polysome-associated miRNA, which is expressed in MPM and in other solid tumors. Strikingly, our data establish that miR-24-3p expression consistently affects similar cellular properties in multiple cell lines. We conclude that our approach is suitable for the identification of biologically relevant miRNAs.

The observed localization of miRNAs with actively translating ribosomes has not been systematically exploited for the identification of active miRNAs. It is often not fully appreciated that many mRNAs are transcribed, but not translated, and that translational control exerts a fundamental role in gene expression (8). Indeed, initiation factors like eIF4E or eIF6 are rate limiting in translation (31), can regulate the translation of specific mRNAs (9) and control tumor progression (17). In summary, a miRNA, to modulate gene expression at the protein level, must target a mRNA that is continuously transcribed, and brought to ribosomes for translation. miRNAs in D. melanogaster and in C. elegans sediment with polysomes (32, 33), where miRNA-mRNA target interactions occurs (34). We have found that the association of a miRNA with polysomes is independent from miRNA abundance, in line with a model in which the affinity and abundance of target mRNA sequences regulate miRNA association with ribosomes (25). Our experiments identified, beside miR-24-3p, let-7 (35) and miR-191-5p (36), which are well established in connection to cancer. Furthermore, we observed that miR-15 and miR-16 cosediment with polysomes. TargomiR targeting miR-16 (37) are in phase I clinical trial in patients with MPM (clinicaltrials.gov NCT02369198). It is therefore evident that the identification of miRNAs bound to ribosomes is predictive of physiological relevance.

We decided to characterize miR-24-3p, among the miRNAs associated with the polysomal machinery for several reasons. First, it was not previously identified in MPM studies; second, it was expressed in both sarcomatoid and epithelial tumors; third, its putative function was puzzling. According to the literature, miR-24-3p has been described either as a pro-oncogenic (38–40) or as a tumor suppressor (41, 42). The two-faced role of a miRNA is not uncommon in miRNA biology. Whether the two-faced role reflects a biological role or the product of different testing conditions is often not clear. Many models claim a miRNA context-dependent function, stressing the fact that evolutionary pressure acts on the 3′ UTR of the target gene, rather than to the miRNA itself. Alternatively, methodological approaches may often be the cause. Most studies, for instance, rely on the overexpression of miRNA in pathological conditions, without addressing its localization. By our approach, we first pinpointed the potential biological relevance of miR-24-3p by its association with polysomes, independently from its expression levels. Next, we confirmed that miR-24-3p acts as a promigratory miRNA in multiple cell types. In the past, reports of a tumor suppressor role for miR-24-3p relied on the use of a mimic miRNA in cells normally devoid of it, without parallel down-regulation experiments (43). We conclude that in physiological conditions, miR-24-3p positively regulates migratory properties, rather than suppressing cancer. Supporting the conclusion of miR-24-3p acting as an oncogenic facilitator, PRECOG analysis showed that the expression of the vast majority of the common targets of miR-24-3p in MPM, is down-regulated in the progression of most tumors. Among genes identified, ZNF185 may function as tumor suppressor gene (44) whilst SALL3 governs methyltransferase activity in cancer and its silencing predicts poor survival (45). One drawback of our approach is that it is not fully suitable for antiassociation studies between target mRNAs and microRNAs. As we have shown in Figure 7, cingulin protein is annihilated by miR-24-3p, but its mRNA is not. This can be mechanistically explained by the fact that ongoing transcription may compensate for the loss of mRNA, but as soon as the mRNA reaches polysomes, its translation is first blocked and then it is degraded.

Mechanistically, the activity of RhoA was significantly reduced in a miR-24-3p dependent manner. RhoA signaling pathways normally correlate with the ability of tumor cells to invade and establish metastases (46). Interfering with the Rho-kinase pathway may inhibit migration and invasion of NSCLC (47). Hyperactivation of the Rho pathway is generally associated with more aggressive tumor properties (30) and RhoA expression is associated with a significant reduction of survival of patients with MPM (48). miR-24-3p controls RhoA activity through at least one target, cingulin. Indeed, it has been reported that the binding of GEF-H1 to the tight junction-associated adaptor cingulin results in inhibition of Rho signaling (49), consistently with our data. Besides cingulin, discussed above, also CXADR, a transmembrane protein involved in tight junction structure and function, is regulated by miR-24-3p and is essential for tight junction integrity. Unfortunately, we could not validate more targets due to the lack of good antibodies.

Even if miR-24-3p affects migration in all tested cells, including the MCF-7 metastatic breast cancer cells, its primary identification in mesothelioma is intriguing. In MPM, most patients succumb to the effect of the primary tumor mass rather than to metastatic burden. At the time of diagnosis most patients are described as late staging, presenting a highly infiltrative disease. However, the landscape of cancer diagnosis and treatment confirm the presence of unusual metastases for patients with MPM (50).

In summary, with our study we propose an original approach for identifying novel miRNAs with relevant function in cancer biology and we suggest that miR-24-3p expression is relevant for cancer progression and metastasization.

No potential conflicts of interest were disclosed.

Conception and design: S. Oliveto, L. Mutti, S. Biffo

Development of methodology: S. Oliveto

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Oliveto, A. Miluzio, A. Scagliola, R.S. Secli, P. Gasparini, S. Grosso

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Oliveto, R. Alfieri, A. Scagliola, L. Cascione, S. Biffo

Writing, review, and/or revision of the manuscript: S. Oliveto, R. Alfieri, P. Gasparini, S. Biffo

Study supervision: L. Mutti, S. Biffo

This work was supported by Gruppo Italiano Mesothelioma (GIMe), AIRC IG 2014 number 19793, and European Research Council (ERC) Translate number 338999 (to S. Biffo). We thank Dr. Raphael Bueno and his collaborator David Severson for helpful insights. We also thank Riccardo Rossi for bioinformatic support.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.