Integrative microRNA and gene expression analysis identifies new epigenetically regulated microRNAs mediating taxane resistance in ovarian cancer

Taxane is a family of front-line chemotherapeutic agents against ovarian cancer (OC). The therapeutic efficacy is frequently counteracted by the development of chemoresistance, leading to high rates of relapse in OC patients. The role(s) of microRNAs (miRNAs) in cancer chemoresistance had been supported by many evidences Epigenetic regulation by miRNAs has been reported to influence cancer development and response to therapeutics, however, their role in OC resistance to paclitaxel (PTX) is unclear. Here, we conducted miRNA profiling in the responsive and PTX-resistant OC cell lines before and after treatment with epigenetic modulators. We reveal 157 miRNAs to be downregulated in the PTX-resistant cells compared to parental controls. The expression of five miRNAs (miRNA-7-5p, -204-3p, -501-5p, -3652 and -4286) was restored after epigenetic modulation, which was further confirmed by qPCR. In silico analysis of the signaling pathways targeted by the selected miRNAs identified the PI3K-AKT pathway as the primary target. Subsequent cDNA array analysis confirmed multiple PI3K-AKT pathway members such as AKT2, PIK3R3, CDKN1A, CCND2 and FGF2 to be upregulated in PTX-resistant cells. STRING analysis showed the deregulated genes in PTX-resistant cells to be primarily involved in cell cycle progression and survival. Thus, high throughput miRNA and cDNA profiling coupled with pathway analysis and data mining provide evidence for epigenetically regulated miRNAs-induced modulation of signaling pathways in PTX resistant OC cells. It paves the way to more in-depth mechanistic studies and new therapeutic strategies to combat chemoresistance.


Scientific Reports
| (2021) 11:562 | https://doi.org/10.1038/s41598-020-78596-5 www.nature.com/scientificreports/ action. PTX is a microtubule-stabilizing agent that selectively binds to microtubules and disrupts their dynamics, inhibiting chromosome segregation and subsequently causes mitotic arrest that leads to cell death 7 . Due to its high efficiency, PTX now is an FDA-approved drug as the first line treatment for OC in combination with platinum-based drugs (eg. Cisplatin). Despite the initial high response rates of OC to platinum/taxane treatment, around 80% of OC patients develop resistance to PTX 8 . The cellular and molecular mechanisms that lead to taxol resistance have been the subject of study for many years, yet still remains largely unknown. MicroRNAs (miRNAs) are short RNA molecules of about 22 nucleotides in length, modulate the expression of almost all genes mostly on the post-transcriptional level. Tumor-suppressor genes, as well as oncogenes, lie under miRNA control, which makes the dysregulation in miRNA expression a cornerstone in cancer progression 9 . A growing body of evidence demonstrate the involvement of miRNAs' dysregulation in the development of OC 10,11 . For example, high expression of miRNA-1207 was correlated with enhancing cancer stem cell-like features in OC through activation of the Wnt/beta-catenin signaling pathway 12 . In fact, many reports attributed the dysregulation of miRNA expression and the subsequent cancer progression to alterations in epigenetic marks at miRNA genes. The epigenetic landscape is known to be dramatically altered in cancer cells, affecting the expression of many coding and non-coding genes, including miRNAs 13 . For example, DNA hypermethylation was found to cause downregulation in miRNA-let-7a with subsequent upregulation of its oncogenic target IGF-II which led to poor prognosis in OC patients 14 . Interestingly, cancer cells acquiring chemoresistance are also known to epigenetically deviate away from their original chemosensitive precursors. The epigenetic silencing of miRNA-199b-5p has been reported to activate JAG1-Notch1 signaling leading to acquiring cisplatin resistance in OC cells 15 . Notably, the interrelation of epigenetics, miRNAs and PTX resistance triad has not been investigated in OC.
In this study, we used serous ovarian carcinoma cell line, KF-28, to map out the role/s of the epigenetically downregulated miRNAs in PTX resistance in OC. We merged data from high-throughput miRNAs and cDNA microarray profiling in both PTX-sensitive and PTX-resistant OC cells coupled with Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) database analysis and data mining to identify candidate miRNAs that can induce PTX resistance in OC with hyperfocus on those which are possibly epigenetically controlled. We also overexpressed or inhibited the miRNAs yielded by the pathway analysis and validated their correlation with their predicted signaling pathways and overall impact on OC response to PTX.

PTX resistant cells show differentially expressed miRNAs profile. MiRNAs microarray profiling
of the OC cell line, KF-28, and their taxol resistant counterpart, KF-Tx (diffrencial response was validated by viability assay; Fig. 1A and supplementary figure S.1), identified multiple upregulated and downregulated miRNAs whose expression is changed due to the acquired state of PTX resistance (Fig. 1B). In this study, we are primarily interested in those miRNAs whose expression was possibly controlled by epigenetic mechanism/s. Therefore, we focused on the most downregulated miRNAs in KF-Tx cells when compared with the KF-28 cells. This comparison yielded a considerable number, 157, of miRNAs that were downregulated in KF-Tx cells. We subsequently narrowed down the list by only selecting miRNAs whose expression levels were re-elevated again in KF-Tx cells after treatment with the epigenetic modulators 5-Aza-2-deoxycytidine (5′AZA) and trichostatin A (TsA). Importantly the combinatorial treatment with 5′AZA/TsA with PTX significantly sensitized the resistant cells to PTX (Same Fig. 1A). Comparing expression profiles of miRNAs in KF-Tx cells before and after co-treatment with the epigentic modulators only revealed that the expression levels of miRNA-7-5p, miRNA-204-3p, miRNA-365a, miRNA-501-5p, miRNA-3652, miRNA-4286, and miRNA-8071 were restored with relative expression changes more than two folds ( Fig. 1C and Table 1). These results suggest that these miRNAs might be candidates for hypermethylation and/or deacetylation at their promoter due to chemoresistance development.
Epigenetically regulated miRNAs modualte multiple signalling pathways in cancer. We next analysed the involvement of only the selected, epigenetically, regulated miRNAs in biological pathways with a specific focus on pathways involved in cancer development and progression. DIANA-miRNAPath tool was used with miRNA-7-5P, miRNA-501-3P, miRNA-3652 and miRNA-4286 as input. The validated miRNA targets from DIANA-TarBase v6.0 were pooled together and were linked to KEGG 16 to identify the enrichment of these targets in all the biological pathways enlisted by KEGG. The analysis yielded a number of biological pathways that lie under the control of the given miRNAs and showed enrichment of miRNAs targets in multiple cancer-related Epigenetically regulated miRNAs are involved in taxane resistance in OC. In order to link the results of the aforementioned miRNAs' targets, we set out to check the expression of the miRNA target genes indicated/validated in DIANA-miRNAPath in our cellular model of PTX resistance. We, therefore, performed a microarray analysis to compare the expression of over 25,000 genes in KF-28 and KF-Tx cells in addition to KF-Tx cells treated with a sublethal dose of PTX. The microarray analysis revealed hundreds of differentially Relative expression of the same set of miRNAs in KF-Tx cells before and after co-treatment with 5′AZA (10 µM; 3 days) and TsA (300 nM; one day). U6 was used as a control. Values in A, C and D are means ± SE from tree independent experiments. One way anlysis of variance (ANOVA) was performed to get the satsitical significance of diffrence at each point (*p < 0.05, **p < 0.01 and ***p < 0.001). www.nature.com/scientificreports/ expressed genes (Fig. 3A). We then merged the results from miRNA microarray with those from cDNA microarray, with a special focus on genes that are upregulated in KF-Tx compared to the KF-28 cell, as their upregulation could be attributed to the observed epigenetic downregulation of any of the identified miRNAs. This analysis revealed multiple oncogenes to be upregulated in KF-Tx cells compared to KF-28 (> 1.5 folds). The miRNAs target genes and their pathway in KF-Tx before and after co-treatment with 5′AZA and TsA were then categorized (Fig. 3B). This list includes key genes such as AKT2; V-akt murine thymoma viral oncogene homolog 2 (3.1 folds), CDKN1A; Cyclin D2 (5.68 folds), FGF2; Fibroblast growth factor 2 (5.2 folds) and PIK3R3; Phosphoinositide-3-kinase, regulatory subunit 3 (gamma) (1.5 folds) ( Table 3). Applying gene lists into the KEGG database revealed roles in detrimental pathways that control cancer cell proliferation and survival such as PI3K-AKT pathway and MAPK-pathway in chemoresistance of OC (Tables 4, 5). Figure 4 illustrates an example where the PI3K-AKT signaling pathway detected by DIANA-miRNAPath to be targeted by miRNA-7-5P, miRNA-501-3P, miRNA-3652 and miRNA-4286, collectively. Genes targeted by one miRNA (yellow boxes) and genes targeted by more than one miRNA (orange boxes) are shown. Interestingly, multiple key genes in this pathway are upregulated in KF-Tx cells (Red-framed boxes) as indicated from cDNA microarray results. Many of those are well-known to control cell cycle and cancer development and progression. Of course, the miRNAs used in the pathway analysis do not affect the pathways listed evenly. The five miRNAs have different number of targets and thus affect a different signaling pathways with different intensities (Fig. 4) as illustrated by the heat map (Fig. 3).
Protein-protein interaction network analysis. We next analyzed the upregulated genes in KF-Tx cells that were matched in KEGG and GO databases using the STRING software. The predicted protein-protein associations were queried through a vast number of databases derived in different ways, such as experimentally determined interactions, protein neighborhood data, or data acquired via text mining. For the predicted 29 upregulated proteins, two main networks of protein interactions were identified, including Growth Factorinduced PI3K-Akt signaling network (red circle) and anti-apoptotic network (Blue circle) (Fig. 5).
Epigenetically regulated miRNAs sensitize OC to PTX. In order to confirm that the expression of the shortlisted miRNAs is functionally correlated with resistance of OC cells to PTX, colony formation assay was performed. The mimic-based overexpression of the different miRNAs was confirmed by qPCR ( Fig. 6A-C; left panels). The overexpression of miRNA-204-3p, miRNA-3652 or miRNA-4286 resulted in significant sensitization of KF-Tx cells to PTX as identified by a reduction in the surviving fractions of the miRNA-transfected cells compared to controls ( Fig. 6A-C, respectively, right panels). In addition, the mimic-based overexpression of two representative miRNAs (miRNA-204-3p and miRNA-4286) in A2780-Tx cells has significantly restored the sensitization to PTX as proved by viability assay (Fig. 6D). Additionally, we validate our observation by www.nature.com/scientificreports/ Relative expression of the same set of microRs in A2780-Tx cells before and after co-treatment with 5′AZA (10 µM; three days) and TsA (300 nM; one day). U6 was used as a control. One way anlysis of variance (ANOVA) was performed to get the satsitical significance of diffrence at each point (*p < 0.05, **p < 0.01 and ***p < 0.001). www.nature.com/scientificreports/ inhibiting the selected three miRNAs (miRNA-3652, miRNA-204-3p. and miRNA-4286) in the parental, KF-28, cells. Transfection of the KF-28 cells with miRNA/s inhibitor conferred resistance to KF-28 cells against PTX and enhanced their viability, specifically at concentration of 25 nm (Fig. 6E). Similar results were obtained from A2780 cells, especially those transfected with inhibitor for miRNA-204-3p (Fig. 6F). Again, these results were confirmed by colony forming assay when KF-28 cells were transfected with each miRNA inhibitor individually (Fig. 6J, left panel), and when A2780 cells were transfected with mixture of the three miRNA inhibitors (Fig. 6J, right panel). The results revealed that the indicated miRNAs inhibition significantly induce PTX resistance in both responsive KF-28 and A2780 cells. Also, These results verify the involvement of those epigenetically controlled miRNAs in the PTX resistance in OC cells. Moreover, we chose one representative miRNA-4286, and confirmed that its overexpression in the KF-Tx cells significantly enhanced the annexin V positive population accumulation, as an indication for the early apoptotic cells, when cells were treated with PTX for 36 h at 500 nM (Supplementary figure S.2). To further determine the potential effect of each of those selected miRNAs on the target genes, qPCR analysis was performed after overexpression of the miRNAs in KF-Tx cells. Overexpression of miRNA-204-3p led to significant downregulation of AKT2, Fibroblast Growth Factor 2 (FGF2), IRS1 as well as Wnt5A mRNAs. In addition, overexpression of miRNA-3652 resulted in significant downregulation of FGF2, IRS2 and Wnt5a mRNAs. While overexpression of miRNA-4286 resulted in significant downregulation of AKT2, CRK, FGF2, IRS1 and Wnt5a mRNAs ( Fig. 7A-C, respectively). Similar to the cDNA data from KF-28/KF-Tx set, some of these genes (BCL2L1 and PIK3R3) were differentially expressed in the A2780/A2780-Tx set of cells (Fig. 7D).

AKT2 expression mediate the PTX response through indicated miRNAs. Previous reports
showed evidence that supports our qPCR data and indicates that the upregulation on AKT2 expression induces OC cells to resist PTX 17 . To further support our findings, we investigated the expression of AKT2 protein by western blotting in both KF-28 and KF-Tx cells treated with different concentrations of PTX (starting from concentration equivalent to IC50 of each cell line) for 36 h. In accordance with cDNA microarray data, our pro- www.nature.com/scientificreports/ www.nature.com/scientificreports/ tein analysis shows differential basic expression of AKT2 at zero (0;vehicle) concentration of PTX. In addition, AKT2 expression was enhanced in response to PTX in both KF-28 and KF-Tx cells in a dose-dependent manner (Fig. 8A). Whereas the co-treatment by the epigenetic modulators; 5′AZA and TsA suppressed AKT2 expression in KF-Tx cells (Fig. 8B). Interestingly, we found that the inhibition of the shortlisted miRNAs individually or in combination restored AKT2 expression in KF-28 cells (Fig. 8C). Moreover, siRNA-based silincing of AKT2 enhanced the response to PTX as indicated by annexin/PI staining (Fig. 8D).

Discussion
Expression profiling has been widely used to reveal the abnormal gene regulation associated with OC 18 and has enabled the identification of potential targets for improved therapeutic strategies. Understanding the mechanism behind such expression profile changes would help to improve treatment strategies to decrease the incidence of chemoresistance and thus improve patients' prognosis. Here, we hypothesized that the upregulation of some genes involved in chemoresistance is due to epigenetic-mediated downregulation of key cellular miRNAs. The analysis of miRNA expression profile as well as mRNA profile of a representative chemoresponsive versus chemoresistant OC cell line before and after treatment with global epigenetic modulators not only helped us to map out the major mechanisms involved in chemoresistance but also shed light to those miRNAs which might be epigenetically controlled during resistance development. The putative target mechanism/s was then analyzed in silico, which provided new insights into OC resistance. It is well established that cancer epigenome significantly contributes to the pathogenesis of many cancers including OC 19 . DNA methylation and/or histone deacetylation are regulators of miRNAs expression and may contribute to the development of chemoresistance. The silencing through DNA methylation can be reversed by demethylating drugs like 5′AZA while silencing through histone deacetylation can be reversed by global HDAC inhibitors like TsA. Both drugs act in synergy to reverse the transcriptionally/epigenetically repressed genes, a strategy previously described in many tumor types 20 .
The fact that combinatorial treatment of 5′AZA and a low dose of PTX was well tolerated and led to favorable clinical outcomes in chemoresistant OC patients 21 supports our experimental approach of targeting the druginduced variations in epigenomic landscape in PTX-resistant OC cells. At first, we assessed the differential expression of miRNAs in the OC cell line, KF-28, and its PTX resistant counterpart, KF-Tx, before and after co-treatment with 5′AZA and TsA. We focused on miRNAs that were downregulated in KF-Tx cells when compared with the parental one proposing that this downregulation might be due to epigenetic reasons. Microarray analysis revealed 157 miRNAs to be downregulated in KF-Tx cells. Among the top 25 downregulated miRNAs, 7 miRNAs were restored to significantly higher expression levels when this resistant, KF-Tx, cells were co-treated with 5′AZA and TsA. Results from qPCR analysis confirmed the array data and narrowed down the list to five miRNAs; miRNA-7-5p, miRNA-204-3p, miRNA-501-5p, miRNA-3652 and miRNA-4286. The fact that these five miRNAs were downregulated in PTX resistant cells suggests that they contribute to the sensitization of OC cells to PTX. In addition, the restoration of their expression after drug-induced epigenetic modification suggests that they are under tight epigenetic control in KF-Tx cells. These data are consistent with the large scale differential expression of miRNAs in PTX-resistant OC cells 22 . Notably, our study is the first to report a large-scale expression profile of epigenetically regulated miRNAs in a resistant OC model, which subsequently identifies five novel miRNAs as new players in OC resistance.
Our next aim was to analyze the possible role(s) of the five epigenetically regulated miRNAs in our chemoresistance model of OC. To achieve this, a complex miRNA pathway analysis was performed. We made use of the data-mining resource DIANA-miRNAPath software. DIANA-miRNAPath is a miRNA pathway analysis webserver that provides accurate statistics and can accommodate advanced search pipelines. miRNAPath can utilize experimentally validated miRNA interactions derived from DIANA-TarBase v6.0 and subsequently combine this  BCL2L1,CASP8,CDKN1A,CFLAR,CRK,CSNK1E,E2F2, FADD,FGF2,F  ZD3,GNB5,GPC4,IFNAR2,IRS1,PAK1, PIK3R3,PLCG2,PVRL1,RELA,  RUNX2,TEAD1,WNT5A   0050776  Regulation of immune response  12  5.70E−07  CASP8,CDKN1A,CRK,FADD,FGF2,FGFR3,IFNAR2,  IRS1,PAK1, www.nature.com/scientificreports/ data with sophisticated merging and meta-analysis algorithms 23 . A collective pathway analysis of the five shortlisted miRNAs yielded a list of over validated 2000 genes by DIANA-TarBase v6.0. This large number of genes is mainly accounted for by miRNA-7-5p which alone yielded 1,830 validated genes as it is one of the most heavily studied miRNA. The validation techniques include microarray, qPCR, western blot, PAR-CLIP and HITS-CLIP 23 . Next, we used the tool available by KEGG database as well as Gene Ontology (GO) analysis to cluster the list of validated miRNAs and their target genes into cellular pathways and biological processes ( Table 2). The analysis shows that the PI3K/AKT pathway is the most affected signaling pathway by the five miRNAs where 70 genes, mediating this pathway, were significantly affected. The cDNA microarray data shows the expression profile of over 25,000 genes in KF-Tx and KF-28 cells, to check the status of validated miRNA targets in this model. Interestingly, we found key genes in PI3K/AKT signaling pathway to be upregulated in KF-Tx cells ( Table 3). The upregulated genes included PIK3Rs, AKT2, CDKN1A and CCND1, among others. It is thus reasonable to believe that the proposed miRNAs targeting these genes which are downregulated in PTX resistant cells allowing these target genes to be upregulated and to proceed with their oncogenic effects. Thus, our analysis proposes an indirect epigenetic regulation of this pathway through key upstream miRNAs. Furthermore, analysis of protein-protein association clustered the 29 upregulated genes into two major networks; PI3K/AKT signaling and an anti-apoptotic network that includes BCL2L1, CASP8, CDKN1A and RELA.
PI3K/AKT pathway is activated in approximately 70% OC 24 . Its activation is associated with higher invasive and migratory capacities 24 , making this pathway a potential predictor of invasiveness for ovarian tumor cells 25 . A prominent key member in this pathway, PIK3R3, phosphorylates and activates AKT2 which leads to www.nature.com/scientificreports/ propagation of the pathway 26 . Our findings go in line with previous studies which showed that silencing PIK3Rs hinders OC SKOV3 cells proliferation, migration, and invasion 27 . The protein-protein association analysis sets AKT2 at the center of the network. AKT2 is well known to induce anti-apoptotic processes. For example, AKT2 blocks the apoptotic function of BAD via phosphorylation 28 . Supporting to our data, a previous study revealed that overexpression of AKT2 stimulates the OC cells to override the PTX treatment 17 while targeting AKT2 by siRNA sensitized OC cells to PTX 29 . It was also previously reported that AKT2 inactivation by siRNAs lead to anti-proliferative effect in OC cells OVCAR-3 and SKOV-3 30 . It was of great importance to show that all these in silico predicted alterations in key pathways would functionally reflect on the resistance phenotype of the PTX resistant cells. The ability of assessed miRNAs to reverse the resistance phenotype of the KF-Tx and A2780-Tx cells (Fig. 5A-D) through modulating AKT2 expression (Fig. 8A-D) proves the efficacy of the in silico tools used to track the targeted downstream pathways of those miRNAs.
In conclusion, this study sheds light on the role of epigenetic manipulation to reverse PTX resistance in OC. We made use of high throughput miRNA and cDNA array profiling coupled with pathway analysis and data mining to show that epigenetically regulated miRNAs can modulate major signaling pathways in PTX resistant OC cells. This work paves the way to more in-depth mechanistic studies, which is running in our laboratory now, for better understanding of each miRNA's specific target, better illustration of the molecular players in PTX resistance, and to identify potential targets for therapy. Such future experiment will also assess whether epigenetic silencing of these genes is essential or only individual silencing is enough to develop resistance phenotype to representative from taxane family, PTX.    www.nature.com/scientificreports/ RNA extraction and microarray processing. For microarray and qPCR experiments, cells were grown in 10 cm 2 culture dishes to a monolayer with confluence of 90% in triplicates. After different treatments, cells were washed twice with PBS and then pelleted by centrifugation for 5 min at 1800 rpm at 4 °C. Cells were lysed by Trizol reagent (Lonza, Belgium) and then total RNA was purified with phenol/chloroform extraction method. Concentration and purity of RNA samples were measured using Nanodrop (Qiagen; Germany). The RNA purity was assessed OD260/OD230 (≥ 1.5) using NanoDrop. gDNA contamination was evaluated by gel electrophoresis. The RNA from three biological samples was then pooled for miRNA analysis. The 2.5 μg and 13 μg of total RNA were individually determined for miRNA profiles and mRNA profiles using Human miRNAOneArray v4 chip andGenechip array, respectively(Affymetrix; USA). The chip uses two groups of probes respectively labeling with Cy3 and Cy5 fluorescent dyes to generate 532 nm and 635 nm excitation. The intensities of fluorescent were analyzed (Molecular Devices, Sunnyvale, CA). The miRNAs raw intensity of each probe was passed the criteria normalized by 75% median scaling normalization method. Target scan and miRNADB online database was further used in the annotation analysis for miRNAs target prediction and functional annotations.
Quantification of expression using qPCR. To  MiRNA pathway analysis. A pathway analysis was performed for shortlisted miRNAs to predict the molecular mechanisms by which those miRNAs can alter the OC cells response to PTX. The analysis was done using the DIANA LAB tool miRNAPath. DIANA-miRNAPath incorporates an updated database of experimentally validated miRNA targets (DIANA-TarBase v6.0) which is able to combine all validated targets of the selected input miRNAs and perform a hierarchical clustering of miRNAs and their affected pathways based on their interaction levels. DIANA-miRNAPath is also linked to the Kyoto Encyclopedia of Genes and Genomes (KEGG) which classifies biological pathways 16,23 .

Analysis of miRNAs target proteins interactions.
In order to evaluate the protein-protein interactions of shortlisted miRNAs' targets, Search Tool for the Retrieval of Interacting Genes (STRING) 10.5 database was used. The database includes both known and predicted direct physical or functional interactions between proteins. STRING constructed a network model showing all protein interactions in our protein set. STRING is also linked to KEGG and a Gene Ontology (GO) databases to cluster the shortlisted genes used as the input into pathways and biological processes.
Colony formation and proliferation assay. The ability of the identified miRNAs to reverse the state of resistance of KF-Tx to PTX was assessed using colony formation assay. MiRNA mimics (Qiagen, Germany) for miRNA-204-3p, miRNA-3652 and miRNA-4286 were transfected into KF-Tx cells using Lipofectamine 3000 Transfection Reagent (Thermo Fisher Scientific, USA) according to manufacturer's manual. In brief, 7.5 × 10 4 KF-Tx cells were seeded in 24-well plates and allowed to attach. After 24 h, transfection complex including mimic of any of the three assessed miRNAs or scrambled RNA was mixed with cells to final RNA concentration of 5 nM. Cells were collected after 48 h and 750 cells were seeded in 6-well plates and treated the next day with DMSO, 5 nM, 10 nM, 20 nM, 40 nM, and 80 nM PTX followed by incubation for 8 days to allow colonies formation. Colonies were fixed with 70% methanol for 20 min then stained with 0.5% Giemsa stain (Sigma-Aldrich, USA) in 70% methanol for 30 min. The number of colonies, each colony more than 50 cells, was counted using a microscope. The surviving fraction was calculated as the ratio of the plating efficiency of the miRNA-transfected cells to that of scrambled RNA-transfected cells. In order to assess the effect of miRNAs' mimics on cell viability, KF-Tx or A2780, 2.5 × 10 5 , cells were seeded in 6-well plates and left to attach for 24 h. Then, cells were transfected either with scrambled RNA, miRNA-204-3p, miRNA-3652 or miRNA-4286 mimics. After 48 h, cells were collected and seeded in 96-well plate (5 × 10 3 cell/well). The next day, cells were treated with PTX with doses ranging from 0.030 to 4 µM for KF-Tx cells and from 0.016 to 1 µM for A2780-Tx cells. After 72 h, cell www.nature.com/scientificreports/ proliferation was assessed using the Cell Proliferation Kit I (MTT) (Sigma, USA) according to manufacturer's instructions.
Western blot. Cell lysates were prepared by lyzing cells in RIPA buffer (10 mM Tris (pH 7.4), 150 mM NaCl, 1% Triton X100, 1% Na.deoxycholate), supplemented with protease inhibitors cocktail (Sigma). Protein concentration of whole cell lysates was determined by BSA assay using the BSA kit (Pierce), then equal protein amounts were heated to 95 °C for 5 min with sodium dodecyl sulfate (SDS) sample buffer (25 ml glycerol, 31.2 ml Tris buffer, 7.5 ml SDS, a dash of bromophenol blue/100 ml) and run on 12% SDS polyacrylamide gel electrophoresis (SDS-PAGE). Protein samples were then blotted onto PVDF membranes (Immobolin P, Watford, UK). The membranes were incubated in blocking solution (5% non-fat milk in PBS) for 1 h then in primary antibody (anti-human AKT2 mAb (at dilution of 20:1000) or GAPDH mAb (at dilution of 20:1000) overnight. After 3 × 10 min washes in TBS (0.1% Tween-20 in PBS) the membrane was incubated for 1 h at room temperature with horseradish-peroxidase-linked IgG (1:2000 dilution in T-TBS) followed by three washes (10 min each) with TBS. Signals on membrane were developed using ECL reagent (Amersham, CA, USA) and then was imaged with Chemidoc system (BioRad, USA).
Annexin V staining. After treatment, cells were harvested and washed in cold PBS. Then, cells were stained directly with PI at final concentration of 10 μg/ml and 2% Annexin V Flous (BioRad) in incubation buffer (10 mM HEPES/NaOH, pH 7.4, 140 mM NaCl, 5 mM CaCl2) for 10 min. Cells were then acquired with the FACS system (Perkin Elmer) after setting the instrument with controls (nontreated stained cells) after two washes in PBS. In this experiment, both cells with early apoptotic signals, stained with annexin V, and cells with late death signals, stained with PI, are all considered as apoptotic cells. Apoptotic cells were analyzed using the CellExpert software (Perkin Elmer).

Statistical analysis.
Statistical analysis was performed using Minitab Release (Ver.12). Data were subjected to one-way analysis of variance, ANOVA, followed by comparison using student t test to evaluate the difference between means. Differences between means were considered significant if p-values < 0.05 which is indicated in the figures by (*), highly significant if p-values < 0.01 which is indicated by (**) and very highly significant if p-values < 0.001 which is indicated by (***).