Functional miRNA Screening Identifies Wide-ranging Antitumor Properties of miR-3622b-5p and Reveals a New Therapeutic Combination Strategy in Ovarian Tumor Organoids

This article is featured in Highlights of This Issue, p. 1383

Epithelial ovarian cancers (EOC) are the leading cause of death from gynecologic cancer. More than 230,000 new cases are diagnosed each year worldwide leading to the death of about 140,000 women (1). The current first-line treatment for patients with EOC consists of cytoreductive surgery and platinum-based chemotherapy (2, 3). Despite a good primary chemotherapy response, the majority of patients with EOC will relapse and develop platinum resistance, explaining the poor 5-year survival rate below 30% for advanced FIGO (International Federation of Gynecologists and Obstetricians) stages (III-IV; ref. 4). The development of novel and effective therapies is critical for improving EOC outcome.

miRNAs constitute a class of endogenous small regulatory noncoding RNAs that repress gene expression post-transcriptionally through partial pairing with mRNAs, yielding a combination of translational repression and mRNA destabilization (5, 6). To date, there are 2,654 potential human miRNAs recorded in miRBase (version 22.1, October 2018), and it is estimated that more than 60% of protein-coding genes could be targeted by miRNAs (7). Importantly, a single miRNA can target several hundreds of mRNAs, and a single transcript can be targeted by multiple miRNAs (5, 8), leading to the formation of complex regulatory networks that have remained largely elusive until now. It is thus not surprising that miRNAs have been found to play a crucial role in all basic biological processes such as development, proliferation, and cell death, and that their dysregulation is implicated in cancer development and progression, including ovarian carcinoma (9–12). Expression profiling studies have identified miRNAs which are deregulated in EOC, and functional studies have uncovered miRNAs acting as oncogenes or tumor suppressors (13). Evasion from apoptosis is an important feature of malignant tumor cells, and the oncogenic role of miRNAs partly involves the regulation of apoptosis. miRNA deregulation is also a feature of chemoresistance in EOC and functional studies have documented that miRNAs may influence the response to chemotherapy (13).

As post-transcriptional regulators of gene expression, miRNAs add a level of functional complexity to regulatory gene networks. miRNA effects involve the targeting of multiple genes. This feature enables effects within highly connected signaling pathways, and the inhibition of a single target may not be sufficient to trigger a phenotype such as cell death, as other compensatory mechanisms may take over. Within these networks, some targeted genes are likely to play a major role in driving the phenotype regulated by a miRNA. miRNAs therefore hold promise for identifying critical regulated genes involved in apoptosis, and potentially contextual synthetic lethal interactions (14).

In this study, we used miRNA multiple targeting to discover vulnerabilities in ovarian cancer cells. We combined real-time label-free cell methodology (impedance measurement) and the functional genomic screening of a miRNA library (gain-of-function) to monitor the whole cellular response over time. We found a miRNA that induces various tumor suppressive effects in chemoresistant ovarian cancer cells and identified several key targets, therefore suggesting a combinatorial drug action.

The cisplatin-sensitive IGROV1 and OAW42 ovarian cancer cell lines were kindly provided by Dr J. Bénard (Institute G. Roussy) and purchased from ECACC, respectively. Their cisplatin-resistant counterpart IGROV1-R10 (15) and OAW42-R (16) cell lines were obtained as described previously. The cisplatin-resistant SKOV3 cell line was purchased from ATCC. IGROV1, IGROV1-R10, and SKOV3 cells were grown in RPMI 1640 medium supplemented with 2 mmol/L Glutamax, 10% FCS, 20 mmol/L HEPES, and 33 mmol/L sodium bicarbonate (Gibco). OAW42 and OAW42-R cells were grown in DMEM (Gibco), supplemented with 10% insulin (Novo Nordisk). All cell lines were maintained in a 5% CO₂ humidified atmosphere at 37°C. Cells were passaged for no longer than 2 months after thawing of early-passage stocks. Ovarian cancer cell lines were certified Mycoplasma-free on a regular basis (MycoAlert Mycoplasma Detection Kit, Lonza) and authenticated by the German Collection of Microorganisms and Cell Cultures and the Swiss Microsynth company. Short tandem repeat (STR) DNA typing was carried out for authentication according to the guidelines published recently (ANSI/ATCC ASN00022011. Authentication of Human Cell Lines: Standardization of STR Profiling. ANSI eStandards Store, 2012).

Human fallopian tube (FT) tissues were collected from 10 women undergoing prophylactic surgery at the François Baclesse Cancer Center. The study was approved by the “North West III” ethical committee (IDRCB: 2018-A02152-53). Oncologists informed all patients included in the study that their biological samples could be used for research purposes and patients gave their written informed consent.

Tumor samples were collected from four patients with serous ovarian cancer at the François Baclesse Cancer Center and transported to the laboratory following surgery. The study was approved by the “North West III” ethical committee (IDRCB: 2018-A02152-53). Oncologists informed all patients included in the study that their biological samples could be used for research purposes and patients gave their written informed consent. Samples were confirmed as tumor on the basis of histopathologic assessment and the diagnosis of each case was confirmed by pathologists at the Cancer Center. Tumor samples were subjected to mechanical and enzymatic dissociation using the Human Tumor Dissociation Kit and a gentleMACS Dissociator according to the manufacturer's instructions (Miltenyi Biotech). The cells were resuspended in a small volume of organoid culture medium mixed with a 1:1 volume of growth factor–reduced Matrigel (Corning). Drops of Matrigel/cell suspension were placed in prewarmed 24-well plates (Eppendorf). Once the Matrigel was solidified, organoid culture medium was added to each well. The medium was changed twice a week and tumor organoids were passaged every 7 to 14 days by dissociation with TrypLE Express (Gibco; Supplementary Materials and Methods).

miRNA mimics and siRNA were purchased from Horizon Discovery or Eurogentec. miRNA (miRNA mimic negative control #1, CN-001000-01) or siRNA (control siRNA duplex negative control, SR CL000-005) were used as controls. siRNA guide sequences targeting Bcl-xL and BIM were described previously (17, 18). Growing cells were seeded at 350,000 (for IGROV1 and IGROV1-R10 cells) or 150,000 cells (for OAW42, OAW42-R, and SKOV3 cells) per 25 cm² flask. Twenty-four hours after seeding, cells were transfected with 20 nmol/L mi/siRNA using INTERFERin following manufacturer's protocol (Polyplus-Transfection).

DAPT (a γ-secretase inhibitor; ref. 19) and CRT-0105950 [a LIMK1 (LIM domain kinase 1) inhibitor; ref. 20] were purchased from Tocris Bioscience. Erlotinib [EGFR inhibitor (EGFRi); ref. 21] and ABT-737 (a BH3-mimetic molecule that binds with high affinity to Bcl-xL; ref. 22) were purchased from Selleck Chemicals. Z-VAD-FMK, a pan-caspase inhibitor of apoptosis was purchased from Promega (Supplementary Materials and Methods).

A total of 7 × 10³ IGROV1-R10 cells per well were plated in 96-well E-Plate View and transfected with each miRNA of human miRIDIAN miRNA Mimic Library (miRBase version 16.0, 1,233 miRNAs mimics, Dharmacon). The plates were placed onto the xCELLigence Real-Time Cell Analysis (RTCA MP, ACEA Bioscience) located inside a tissue culture incubator. Cells were grown for 24 hours before transfection and impedance was continuously measured until the end of the treatment. The impedance of each well was expressed as a cellular index (CI) value. SDs of well replicates were analyzed with the RTCA 2.1.0 Software. Candidate miRNAs were selected according to three complementary criteria: (i) the shape of the curve (downward slope after miRNA transfection) and two measurable parameters, (ii) the AUC, and (iii) the CI at the end of the experiment (closed to zero) in association with cell imaging (appearance of cell debris; Supplementary Materials and Methods).

The rate of cell migration and invasion of SKOV3 cells was monitored in a label-free real-time setting with the xCELLigence RTCA DP system (ACEA Biosciences) using CIM-Plates. The impedance value of each well was automatically monitored by the xCELLigence RTCA system for 24 hours and expressed as a CI value (Supplementary Materials and Methods).

IncuCyte S3 system (Essen BioScience) is an automated imaging platform providing realtime images and quantitative data generated throughout the wound-closing process. Twenty-four hours after transfection, a total of 3 × 10⁴ SKOV3 cells per well were seeded in 96-well IncuCyte ImageLock plates at a density of 70% to 80%. A single homogeneous scratch wound was created using the IncuCyte Woundmaker. Cells were placed into the device located inside a CO₂ incubator for 24 hours. Images were analyzed to quantify wound healing by the dedicated software.

Caspase-3/7 activity was assessed using the IncuCyte caspase-3/7 Green Apoptosis Assay Reagent (Essen BioScience) as described previously (23). A total of 6 × 10³ IGROV1-R10 cells were cultured in 96-well plates and monitored in the IncuCyte acquiring images every hour in two separate regions per well after transfection with the indicated miRNAs. The live-cell phase contrast images were used to calculate confluence using the IncuCyte software.

miRNA were isolated from human FT samples and ovarian cancer cells using the NucleoSpin miRNA kit according to the manufacturer's instructions (Macherey-Nagel). The expression of miR-3622b-5p and RNU44 were determined by qRT-PCR. miRNAs were first retrotranscribed using miRNA Reverse Transcription Kit (Thermo Fisher Scientific). The ID references for stem-loop primers and TaqMan hydrolysis probes were the following: hsa-miR-3622b-5p (478839_miR) and RNU44 (001094). Using the RNU44 endogenous control, relative changes in miRNA expression were calculated using the 2^−ΔΔC_q (cycle quantitative) or ΔC_q method (24; Supplementary Materials and Methods).

Analysis of tumor samples was performed according to the relevant national law on the protection of people taking part in biomedical research. Tumor samples (N = 50) were retrospectively collected at the François Baclesse Cancer Center and stored in the OvaRessources Biological Resources Center (BRC; NF-S 96900 quality management, AFNOR N°2016:72860.5). The whole biological collection from BRC was declared to the MESR (Ministry of Education, Health and Research, France, N°DC-2010-1243). We obtained written informed consent from patients still alive under the agreement of the ethical committee “North-West III”. For patients deceased or lost to follow-up, authorization was obtained to use their samples in the context of this study. RNA extraction was performed with the Nucleospin miRNA extraction kit. Small RNA sequencing were run on a flowcell of Hiseq 4000 (Illumina) in paired-end (2 × 150 bp). Reads were aligned against miRBase release 22 with Bowtie (v1.2.2), and a read-count matrix normalized on library size was constructed using DESeq2 (Supplementary Materials and Methods).

Protein levels were analyzed by immunoblotting with indicated antibodies (Supplementary Table S1A) as described previously (18). Each immunoblot is representative of three distinct experiments. Protein expression was measured by quantifying the density of immunoblots bands adjusted to with β-actin or α-tubulin using ImageJ software (Supplementary Materials and Methods).

Morphologic characterization of apoptotic cells by nuclear staining with DAPI and DNA content after staining with propidium iodide using a Gallios flow cytometer (Beckman Coulter) were performed as described previously (18; Supplementary Materials and Methods).

OAW42-R cells were first transfected for 24 hours in 25 cm² flask using JetPRIME (Polyplus Transfection) with 1,000 ng of pCMV-Bcl-xL. Cells were then transfected for 24 hours with 20 nmol/L of miR-3622b-5p or miR-CTRL using INTERFERin. Finally, cells were exposed to cisplatin in serum-free medium for 2 hours, then were washed with PBS and incubated in the complete growth medium for an additional 24 hours.

SKOV3 cells were first transfected for 24 hours in 6-well plate using JetPRIME with 200 ng of pMIRGLO plasmid empty or containing each the Targetscan predictive miR-3622b-5p binding site (Supplementary Table S1B). SKOV3 cells were transfected 24 hours with 20 nmol/L of miR-3622b-5p or miR-CTRL using INTERFERin. Firefly and Renilla luciferase activities were assayed 24 hours after miRNAs transfection with Dual-Luciferase Reporter Assay System (Promega) and measured with a luminometer Centro LB 960 (Berthold; Supplementary Materials and Methods).

White clear bottom 96-well plates (Greiner) were coated with growth factor–reduced Matrigel diluted 1:1 with organoid culture medium for treatment. Tumor organoids were collected before being resuspended in 2% Matrigel/organoid culture medium for treatment and plated in 96-well plates. Two hours later, tumor organoids were treated with the indicated molecules and viability was assessed 48 hours after the beginning of treatments using the CellTiter-Glo 3D cell viability assay according to the manufacturer's instructions (Promega).

Experiments were carried out in triplicate and repeated independently at least three times (unless indicated otherwise). The normality of variables was first verified by the Shapiro test. If samples were not considered to be normally distributed, we used data log transformation and normality was reassessed. The analyses of variables considered to be taken from a normal distribution was performed using Student t test (NS P > 0.05; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001). Statistical analyses were performed using Microsoft Excel 2010 (Microsoft) and R [version 3.5.0 (2018-04-23)].

We performed functional screening of a genome-wide library of 1,233 miRNAs mimics in the IGROV1-R10 chemoresistant ovarian cancer cell line. The use of an impedance-based real time assay enables continuous monitoring of dynamic cell behavior (adhesion, proliferation and survival), in association with cell imaging, to characterize the phenotypic response induced by each miRNA. This cellular kinetic impedance-based information makes it possible to pinpoint the optimal time window for performing downstream endpoint assays to examine cell death. To select the most potent cytotoxic miRNAs, we considered three complementary parameters (the shape of the curve, the AUC, and the CI at the end of the experiment) and morphologic observations (Fig. 1A). miR-491-5p, an “apoptomiR” recently identified by our group (18), was used as a positive control for cytotoxicity. The combination of all these criteria revealed 15 miRNAs of interest inducing a potential cytotoxic effect (Fig. 1B; Supplementary Fig. S1).

We next studied the cytotoxicity of the 15 selected miRNAs candidates by a functional individual study. Three of them (miR-648, miR-3165, and miR-3622b-5p) induced apoptotic cell death in IGROV1-R10 cells, as shown by the marked cleavage of caspase-3 and poly(ADP-ribose) polymerase (PARP) visualized by Western blot analysis (Fig. 1C), similar to that observed for miR-491-5p. Live-cell imaging experiments also showed an increase in the percentage of caspase-3/7–positive cells up to 49.6% (P < 0.01) and 48.7% (P < 0.01) for miR-648 and miR-3622b-5p, respectively. Of note, no significant result was obtained with miR-3165 (Fig. 1D). A significant increase in the percentage of sub-G₁ events up to 40.6% (P < 0.05), 21% (P < 0.05), and 31.3% (P < 0.05) was also induced by miR-648, miR-3165, and miR-3622b-5p, respectively, whereas miR-CTRL had no such effect (Fig. 1E). Altogether, the apoptotic effect of miR-648, miR-3165, and miR-3622b-5p was validated on the IGROV1-R10 ovarian cancer cell line.

We focused our study on miR-3622b-5p because its coding gene is located in a chromosomal region (chr8p21) which is frequently deleted in ovarian carcinoma (25, 26). Accordingly, RNA sequencing performed on 50 ovarian tumor cancer samples revealed that miR-3622b-5p is poorly expressed in comparison with clinically relevant underexpressed (miR-125b-1-3p, let-7a-3p) or overexpressed (miR-200a-3p, miR-21-5p) miRNAs (11, 27; Supplementary Fig. S2A). We also showed that miR-3622b-5p expression in IGROV1-R10 and a panel of ovarian cancer cell lines (IGROV1, OAW42, OAW42-R, and SKOV3) is low and attenuated compared with human FT epithelial cells by qRT-PCR (Supplementary Fig. S2B and S2C).

miR-3622b-5p effects were analyzed in these ovarian cancer cell lines. Seventy-two hours after transfection, miR-3622b-5p induced apoptotic cell death in IGROV1 and OAW42 cancer cell lines to the same extent as in IGROV1-R10 cancer cells, as shown by massive pools of cell debris and numerous condensed and fragmented nuclei, the cleavage of caspase-3 in Western blot analysis and an increase in the sub-G₁ population in flow cytometry, up to 24.8% and 26.8%, for IGROV1 and OAW42 cells, respectively (Fig. 2A–C; Supplementary Fig. S2D and S3A). In addition, treatment with the pan-caspase inhibitor Z-VAD-FMK protected IGROV1-R10, IGROV1, and OAW42 cells from miR-3622b-5p–induced apoptosis (Supplementary Fig. S3B and S3C). In contrast, no cytotoxic effect was observed in the two chemoresistant OAW42-R and SKOV3 cell lines (Fig. 2A–C). However, miR-3622b-5p–induced morphologic changes (thinness and star shape phenotype) in SKOV3 cancer cells.

Although miR-3622b-5p had no effect alone on OAW42-R cancer cells, it sensitized OAW42-R cancer cells to cisplatin, as shown by the accumulation of cell debris, the increase in sub-G₁ events up to 26.7% (P < 0.01) and the cleavage of PARP and caspase-3 in comparison with the association of miR-CTRL and cisplatin (Fig. 2D–F). In contrast, no chemosensitizing effect was observed for SKOV3 cancer cells (Supplementary Fig. S4A–S4C).

Although miR-3622b-5p–induced morphologic changes in SKOV3 cancer cells, suggesting a potential epithelial-to-mesenchymal transition, we did not observe any switch between E-cadherin and N-cadherin or a change in Vimentin expression in response to miR-3622b-5p (Supplementary Fig. S4D). Next, we assessed the migration and invasion effects of miR-3622b-5p using an electronically integrated Boyden chamber based on impedance measurement. We observed that miR-3622b-5p significantly decreased the migratory (P < 0.01) and the invasive (P < 0.05) capacities of SKOV3 cancer cells in comparison with control conditions (Fig. 3A and B). Similarly, overexpression of miR-3622b-5p decreased the ability of SKOV3 cancer cells to reinvest the wound area, as assessed by wound-healing assays based on live-cell imaging (P < 0.01; Fig. 3C and D).

To decipher the mechanisms underlying the apoptotic effect of miR-3622b-5p in ovarian cancer cells lines, we focused on ERBB2 (erb-b2 receptor tyrosine kinase 2; ref. 28) and EGFR (29). Both were previously shown to be direct targets of miR-3622b-5p. Moreover, their modulation by this miRNA could be partially explained by the proapoptotic effect observed in other settings such as breast, gastric, and prostate cancer cells. First, we found that ERBB2 was not expressed in the panel of ovarian cancer cell lines (except SKOV3 used as positive control) (Supplementary Fig. S5A). Next, we observed that miR-3622b-5p directly targeted EGFR in ovarian cancer cells, as shown by Western blot analysis and luciferase reporter assays (Fig. 4A; Supplementary Fig. S5B). We also found that miR-3622b-5p downregulated the activity of proteins involved downstream in the EGFR signaling pathway, p-AKT^Ser473 and p-ERK (except for OAW42), leading to BIM upregulation at the protein level in ovarian cancer cells (Fig. 4A; Supplementary Fig. S5C). Bcl-xL protein expression was reduced while MCL-1 protein expression was increased in response to miR-3622b-5p in the three cancer cell lines, where it had cytotoxic effects (Fig. 4A; Supplementary Fig. S5C). We also observed that EGFR and Bcl-xL protein expression was reduced in response to miR-3622b-5p transfection in OAW42-R and SKOV3 cancer cells lines, although the inhibition of EGFR protein expression did not lead to BIM upregulation in these two ovarian cancer cells lines (Supplementary Fig. S5D). In silico analysis identified three putative target sites for miR-3622b-5p in the 3′-untranslated region (UTR) of human Bcl-xL. Cotransfection of the wild-type 3′-UTR luciferase reporter construct for each predicted site together with miR-3622b-5p mimic substantially decreased the luciferase activity in transfected cells only for site 2, demonstrating for the first time the direct effect of this miRNA on Bcl-xL mRNA (Fig. 4B; Supplementary Fig. S5E).

We next determined whether BIM upregulation is critical for miR-3622b-5p-induced apoptosis. BIM silencing conferred a significant resistance of ovarian cancer cells to miR-3622b-5p-induced apoptosis, as shown by the significant decrease in the sub-G₁ population for IGROV1-R10 (P < 0.01), IGROV1 (P < 0.05), and OAW42 (P < 0.01) cancer cell lines (Fig. 4C) and the absence of PARP and caspase-3 cleavage (Fig. 4D). Overall, these results demonstrated that the induction of the BH3-only protein BIM by miR-3622b-5p is critical for the induction of apoptosis in IGROV1-R10, IGROV1, and OAW42 cancer cell lines.

As previously demonstrated in IGROV1-R10 cancer cells (18), we next wondered whether the combination of erlotinib, an EGFRi with ABT-737 (a BH3-mimetic molecule which neutralizes Bcl-xL) could mimic the proapoptotic effect of miR-3622b-5p in a panel of ovarian cancer cell lines. Indeed, whereas ABT-737 induced mild apoptosis, the combination of an EGFRi and ABT-737 led to massive apoptosis, with an increase in sub-G₁ events up to 42.7% (P < 0.01) and 33.4% (P < 0.01) in IGROV1 and OAW42 cancer cell lines, respectively (Fig. 4E and F). However, as expected from our previous results obtained with miR-3622b-5p, the combination of erlotinib and ABT-737 did not induce cell death in OAW42-R (Supplementary Fig. S5F and S5G) and SKOV3 cancer cells (18).

First, we studied whether cisplatin could induce the expression of miR-3622b-5p in OAW42-R cancer cells. However, it did not (Supplementary Fig. S6A). Next, because we had previously demonstrated that the downregulation of Bcl-xL can sensitize the IGROV1-R10 cancer cell line to cisplatin and that miR-3622b-5p directly targets Bcl-xL not only in IGROV1-R10 cells but also in OAW42-R cancer cells, we hypothesized that the chemosensitizing effect of miR-3622b-5p could be related to Bcl-xL downregulation. Using an siRNA targeting Bcl-xL, we found that Bcl-xL silencing sensitizes OAW42-R cancer cells to cisplatin, as shown by an increase in the proportions of both sub-G₁ events (20.6%, P < 0.01) and floating cells, and by the cleavage of PARP and caspase-3 (Supplementary Fig. S6B–D). The overexpression of Bcl-xL rescued the apoptotic effects induced by miR-3622b-5p in combination with cisplatin in OAW42-R cancer cells, underlining the pivotal role of Bcl-xL in the chemosensitizing effect of miR-3622b-5p (Supplementary Fig. S6E).

Because miR-3622b-5p affects the motility of SKOV3 cells, we sought to determine the targets involved. We focused on LIMK1, ROCK1 (Rho-associated coiled-coil containing protein kinase 1), PAK4 [p21 (RAC1) activated kinase 4; three members of the same signaling pathway) and NOTCH1 because the TargetScan algorithm predicted that they could be direct targets of miR-3622b-5p and because they have already been found to be involved in SKOV3 cell migration (30, 31). Western blot analysis after miR-3622b-5p transfection revealed that LIMK1, NOTCH1, and cleaved NOTCH1 levels were decreased, whereas ROCK1 and PAK4 levels remained unchanged (Fig. 5A; Supplementary Fig. S7A). Cotransfection of the wild-type 3′-UTR luciferase reporter construct for LIMK1 together with the miR-3622b-5p mimic showed a trend to a decrease in luciferase activity, but nonsignificantly (P = 0.068; Supplementary Fig. S7B and S7C). The same experiment with NOTCH1 3′-UTR did not decrease the luciferase activity in transfected cells, suggesting that the decreased expression of these two proteins upon miR-3622b-5p transfection is due to an indirect mechanism.

We next wondered whether the association of an LIMK1 inhibitor (CRT-0105950) and a γ-secretase inhibitor (DAPT), which prevents NOTCH1 cleavage and therefore activation (32), could have the same effect as miR-3622b-5p transfection (Fig. 5B). LIMK1 inhibition affected SKOV3 wound-healing, whereas NOTCH1 inhibition had no obvious effect. Interestingly, the combination of both inhibitors was more potent than any inhibitor alone, suggesting a possible synergistic mechanism (Fig. 5C and D).

Patient-derived tumor organoids (PDO) have recently emerged as robust preclinical models of patient responses to drugs in clinical practice (33–36), including in ovarian carcinoma (37, 38). To support the therapeutic interest of our results, we then investigated the treatment effects of an EGFRi in combination with a BH3-mimetic molecule in four ovarian PDOs. Morphologic observation as well as cell viability assays showed that the combination of erlotinib with ABT-737 led to a considerable decrease in cell viability for all PDOs, whereas they were only slightly sensitive to each drug alone (Fig. 6A and B).

Ovarian cancer is the most aggressive gynecologic malignancy. If therapeutic care for ovarian cancer is to improve, innovative treatments likely to overcome chemoresistance are needed. The initial objectives of this study were to identify miRNAs with cytotoxic effects on ovarian cancer cells and to characterize their mode of action to establish new targets that could form the basis of innovative pharmacologic strategies.

The combination of real-time continuous cell monitoring with miRNA screening was found to be a very efficient and convenient way to discriminate subtle and robust phenotypic changes as well as to identify the best time points for assaying changes in either mRNA or protein expression. However, only a few cytotoxic miRNAs were identified by our stringent criteria in the IGROV1-R10 cancer cell line. First, it would be interesting to find out how many miRNAs among those now known (2,654 mature miRNAs) would also have a cytotoxic effect on these ovarian cancer cell lines. Second, out of the 15 candidates we selected, we efficiently retrieved four miRNAs previously described by other groups (miR-193a-3p, miR-193b-3p, and let-7g-3p; refs. 39–43) or us (miR-491-5p; ref. 18) that display cytotoxic effects in ovarian cancer cells, demonstrating the efficiency of our screening technique. Third, besides miR-3622b-5p, we also identified two novel apoptotic candidates (miR-648 and miR-3165). Finally, further validation is also required to ascertain the potential apoptotic—or other forms of—cell death triggered by the other miRNA candidates.

We also found that miR-3622b-5p displays wide-ranging tumor suppressor effects in ovarian cancer cells. The miR-3622b-5p coding gene is located in a chromosomal region (chr8p21) that is frequently deleted in ovarian carcinoma (25, 26). Accordingly, miR-3622b-5p expression is low in patients with ovarian cancer and in several ovarian cancer cell lines compared with human FTs, highlighting the widespread attenuated expression of miR-3622b-5p in ovarian carcinoma. Further studies are required to demonstrate that miR-3622b-5p is a “bona fide” tumor suppressor in ovarian cancer.

We also demonstrated that the apoptotic effect of miR-3622b-5p in ovarian cancer cells is related to both Bcl-xL inhibition and BIM stabilization through inhibition of EGFR signaling. During the course of this study, another group reported that miR-3622b-5p directly targets EGFR in prostate cancer cells (29). We also found that EGFR is a direct target in the context of ovarian cancer. We now show for the first time that Bcl-xL is a direct target of miR-3622b-5p in ovarian carcinoma. In the three ovarian cancer cell lines where EGFR inhibition led to BIM activation, the concomitant inhibition of Bcl-xL by miR-3622b-5p led to apoptosis. While we found that BIM is critical for the induction of apoptosis by miR-3622b-5p in these cancer models, the absence of cell death in OAW42-R cells where BIM is not upregulated in response to miR-3622b-5p further underlines its central role in ovarian cancer cells. Indeed, it is now widely accepted that the balance between proapoptotic and antiapoptotic proteins, specifically BIM, Bcl-xL, and MCL-1, is a major determinant of the survival or death fate of ovarian cancer cells (44). Accordingly, our group and others have demonstrated the value of exploiting the apoptotic priming induced by MEK inhibition, leading to upregulation of BIM to promote the cytotoxicity of in vitro (44), ex vivo (45), and patient-derived xenografts ovarian cancer models through combined Bcl-2/Bcl-xL inhibition by a BH3-mimetic molecule (ABT-263; ref. 46). Our observation with miR-3622b-5p echoes our previous finding regarding miR-491-5p, which exhibits a similar cytotoxic effect on several ovarian cancer cell lines and also involves the concomitant inhibition of Bcl-xL and EGFR, also leading to BIM upregulation (18). Interestingly, in the prostate cancer cells where miR-3622b-5p triggers apoptosis, EGFR inhibition alone had no effect (29). Our results suggest that Bcl-xL inhibition by miR-3622b-5p could be involved in prostate cancer cell death. This raises the issue regarding the possible alternative ways to induce cell death in cancer cells. The rather low number of cytotoxic miRNAs we identified in ovarian cancer cells could be the reflection of the limited number of gene combinations whose inhibition is able to trigger cell death. This miRNA was also shown to induce apoptosis in breast and gastric cancer cells by targeting ERBB2 (28). However, the modulation of this target cannot explain the proapoptotic effect of miR-3622b-5p observed in our study because of the absence of ERRB2 expression in the ovarian cancer cell lines we studied. Finally, in the SKOV3 ovarian cancer cell line, miR-3622b-5p repressed migration and invasion through the indirect inhibition of NOTCH1 and possibly by directly targeting LIMK1. Overall, miR-3622b-5p appears to have broad tumor suppressor effects through several pathways and on different hallmarks of cancer cells.

While RNAi-based therapeutics for cancer therapy remain challenging, the use of pharmacologic inhibitors to mimic the action of a miRNA of interest presents a realistic alternative. Indeed, although the NOTCH1 inhibitor did not show a significant effect, its association with a LIMK1 inhibitor produced a strengthened antimigratory effect, suggesting a synergistic mode of action. A clinical trial has demonstrated that RO4929097, a gamma secretase inhibitor of the Notch signaling pathway, leads to prolonged stable disease in patients with ovarian cancer (47), suggesting that RO4929097 could be combined with LIMK1 inhibitors to improve the therapeutic effect observed with a NOTCH inhibitor alone.

In addition, we found that the combination of Erlotinib with ABT-737 induced a cytotoxic effect in a panel of ovarian cancer cells and very interestingly in several ovarian PDOs. This suggests that it would be interesting to combine EGFRis already used in clinical practice with ABT-263, the orally bioavailable analogue of ABT-737 which is currently in the clinical trial phase, in an attempt to recapitulate the cytotoxic effects of miR-3622b-5p. In addition, our results may explain why EGFR-directed therapies used as single agents in clinical trials have yielded disappointing results. In this case, although EGFR inhibition may lead to BIM induction, the induction of cell death would require the release of BIM from Bcl-xL obtained by the use of a BH3-mimetic molecule such as ABT-737. Accordingly, recent preclinical studies indicate that combining ABT-263 with targeted therapies could be very efficient for treating various solid tumors (48, 49). It would be also interesting to explore in ovarian cancer cells escaping BIM induction which molecular alterations might impede the activity of EGFRis and whether it is possible to bypass them to propose efficient strategies. Moreover, it is essential to identify the predictive factors allowing appropriate patient selection for such therapeutic strategies to succeed. The importance of BIM expression as a predictive factor of response to various treatments including targeted therapies has been suggested in ovarian carcinoma (45) and other tumor types (50). It would therefore be interesting to study the importance of BIM expression as a predictive factor in the success of such strategies in PDOs, as they have recently emerged as preclinical models of patient responses to drugs in clinical practice (33–36), including in ovarian carcinoma (37, 38).

In conclusion, by studying miR-3622b-5p and the effects of miRNAs in general, it should become possible to identify new therapeutic targets, irrespective of their being directly or indirectly affected by the miRNA. Most importantly, mimicking miRNA effects by using a combination of pharmacologic inhibitors effects holds promise for improving the effects of individual drugs through synthetic efficiency, eventually leading to new clinical trials.

No potential conflicts of interest were disclosed.

Conception and design: B. Lambert, M. Meryet-Figuière, L. Poulain, C. Denoyelle

Development of methodology: M. Vernon, E. Brotin, L.-B. Weiswald, N. Vigneron, C. Denoyelle

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Vernon, B. Lambert, E. Brotin, L.-B. Weiswald, H. Paysant, E. Abeilard, F. Giffard, M.-H. Louis

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Vernon, B. Lambert, E. Brotin, L.-B. Weiswald, H. Paysant, N. Vigneron, E. Abeilard, F. Giffard, M.-H. Louis, C. Blanc-Fournier, P. Gauduchon, L. Poulain, C. Denoyelle

Writing, review, and/or revision of the manuscript: M. Vernon, B. Lambert, M. Meryet-Figuière, E. Brotin, L.-B. Weiswald, H. Paysant, N. Vigneron, A. Wambecke, E. Abeilard, F. Giffard, M.-H. Louis, P. Gauduchon, L. Poulain, C. Denoyelle

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Vernon, B. Lambert

Study supervision: B. Lambert, L. Poulain, C. Denoyelle

This work was supported by the French State, the ‘Normandy County Council’ (to C. Denoyelle), Inserm, the ‘Ligue Contre le Cancer’ (Normandy confederation; to C. Denoyelle), the ‘Rose sur Green’ Association (C. Denoyelle), the University of Caen Normandy (UNICAEN), and the ‘Canceropole North-West’ (C. Denoyelle). We thank Marilyne Duval (Flow Cytometry Core Facility, Federative Structure 4206 ICORE, University of Caen Normandy) for her technical help. M. Vernon and H. Paysant are a recipient of a doctoral fellowship from the French Ministry for Higher Education and Research. A. Wambecke is a recipient of a doctoral fellowship from the “Normandy County Council”. L.-B. Weiswald is a recipient of the Emergence Cancéropôle Nord-Ouest fellowship “Orgraft”. This work was supported by a grant from l'Ecole de l'Inserm Liliane Bettencourt (to N. Vigneron). The xCELLigence devices were acquired thanks to the European Regional Development Fund and the François Baclesse Cancer Center (Caen). The IncuCyte S3 device was acquired thanks to the support of the French State and the ‘Normandy County Council’ (Contrat de Plan Etat Région—CPER INNOVONS). The authors also thank Dr. Laurent Maillet and Dr. Philippe Juin (UMR 1232, Nantes, France) who kindly provided the pCMV-Bcl-xL plasmid.

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.