The miR–181a–SFRP4 Axis Regulates Wnt Activation to Drive Stemness and Platinum Resistance in Ovarian Cancer

Epithelial ovarian cancer (EOC) is currently the most lethal gynecologic malignancy with an estimated 22,530 new cases and 13,980 deaths in 2020. Thus, novel therapeutic interventions are needed to improve the survival of women affected by this disease (1). Tumor heterogeneity presents a major barrier toward gaining a better understanding of the molecular underpinnings driving tumor progression, chemoresistance and tumor-recurrence, thus the development of novel effective therapies to treat EOC is lacking (2, 3). Heterogeneity is a defining feature of EOC, given that the disease can arise from several distinct anatomic locations. The most common site of origin is the fallopian tube surface epithelium that in turn gives rise to the most common EOC subtype, high-grade serous ovarian tumors (HGSOC; refs. 4–8). HGSOC is characterized by a high degree of genomic instability rather than a small subset of driver mutations that greatly increases the heterogeneity of cancer cells within the tumor. This high degree of tumor heterogeneity ultimately translates to a much higher probability of developing chemoresistance and recurrence.

Understanding mechanisms driving EOC pathogenesis in the context of heterogeneity forms a crucial component of determining the different signaling pathways that regulate diverse subpopulations of tumor cells in primary ovarian tumors. Current experimental approaches have limited utility to explore this possibility. Classic overexpression or knockdown genetic approaches fail to account for tumor heterogeneity and do not enable analysis of tumor subpopulations. Hence, novel experimental methodologies that can enable isolation and analysis of tumor subpopulations can decipher mechanisms underlying EOC pathogenesis in the context of heterogeneity, thus contributing to the design of successful targeted therapies.

We have developed an miRNA 3′ untranslated region (UTR) sensor-based functional platform (9) to isolate and analyze tumor subpopulations based on miRNA activity (10). Given that a single miRNA can bind to hundreds of 3′UTRs thus regulating multiple pathways simultaneously, miRNAs constitute efficient targets to dissect tumor heterogeneity because they regulate multiple pathways and can be used to isolate and analyze multiple tumor-subpopulations as compared with single pathway reporter–based approaches. Using this miRNA sensor platform, we found that an miR-181a^high subpopulation of cells sorted from primary ovarian tumor cells exhibited stem-like properties in vivo, were enriched in response to continuous cisplatin treatment, and showed activation of major stem cell regulatory pathways (10). In the current study, we have identified Wnt/β-catenin signaling to be enriched in miR-181a^high primary ovarian tumor cells and show that miR-181a–driven Wnt signaling promotes stem-like properties. Furthermore, we show that miR-181a activates Wnt signaling by targeting the secreted WNT inhibitor SFRP4. Our study highlights the importance of using endogenous miRNA activity–based experimental approaches to study tumor subpopulations in cancers that can: (i) uncover endogenous regulators of relevant signaling pathways and, (ii) enhance the clinical translation of targeted therapies to overcome heterogeneity-driven chemoresistance and tumor-recurrence.

Cells were cultured in 10-mm plates in a humidified atmosphere (5% CO₂) at 37°C. At 70%–90% confluence, trypsin (0.25%)/EDTA solution was used to split the cells that were used until passage 20. Cell culture medium was supplemented with 10% FBS (Gibco) and 1% penicillin–streptomycin [Gibco; OV81.2-CP10 cells (DMEM), HEYA8 cells (RPMI)]. Primary HGSOC cells (OCI-P5X, OV236) were cultured in OCMI-L medium (Liver tumor culture core, University of Miami). OV81.2-CP10 and OV236 are primary HGSOC generated in the DiFeo Laboratory as previously described (10, 11). OCI-P5X cells were purchased from Liver Tumor Culture Core, University of Miami. HEYA8 was obtained from Dr. Anirban Mitra from University of Indiana. All cell lines were maintained below 25–30 passages for all experiments and tested monthly for Mycoplasma infection. In addition, the cell lines have been authenticated using genetic profiling via polymorphic short tandem repeat. The following reagents were used in the study: Cisplatin (Mount Sinai Hospital Pharmacy), Lithium chloride (Sigma), iCG-001 (Selleckchem), Sh-β-catenin and Sh-Control vectors (Addgene; ref. 11), Top-eGFP plasmid (Addgene), miR-181a overexpression and control vector (Biosettia), miR-181a antagomiR and control vector (Genecopoeia). Top-FLASH plasmid was a kind gift from Dr. William Scheimann (Case Western Reserve University, Cleveland, Ohio).

Cells were plated in 12-well plate at 50,000 cells/well and treated the next day with the drugs. After 48 hours, cells were then incubated with 3-(4,5-Dimethylthiazolyl) for 2 hours and absorbance was measured at 600 nm.

BD LSR II was used for flow cytometry analysis with miR-181a sensor and BD FACS Aria was used for miR-181a sensor-based sorting.

Total protein lysates were prepared from cell pellets using RIPA buffer supplemented with PhosSTOP phosphatase inhibitor and cOmplete protease inhibitor cocktail (Roche). Protein concentration of cell lysates was determined using BCA assay. Western blots were run as described previously (12, 13). Primary and secondary antibodies used are listed below. List of antibodies used in western blots are presented in Supplementary Table S1.

Immunocytochemistry of cells was performed as previously described (14), wherein cells were seeded in 4-well chamber slides at a density of 70,000 cells per well (approximately 70% confluent) and were allowed to grow for 48 hours. The cells were washed with 1X PBS and then fixed with 4% paraformaldehyde for 10 minutes. Cells were washed with 1X PBS three times for 5 minutes each followed by permeabilization with 0.01% (v/v) Triton X100 (Sigma) for 10 minutes. The cells were washed 3X with 1X PBS for 5 minutes each followed by blocking in 3% BSA for 1 hour. The cells were then incubated with anti-active β-catenin primary antibody (1:500 dilution) overnight. The cells were then washed 3X with PBS for 5 minutes each followed by incubation with anti-rabbit Dylight488–conjugated secondary antibody (Thermo Fisher cat. #35553 1:100 dilution) in 3% BSA for 1 hour. After incubation with the secondary antibody, the cells were then washed 3X with PBS for 5 minutes each followed by counterstaining with ActinRed 555 ReadyProbes reagent (Invitrogen) for 30 minutes. After counterstaining, the cells were then washed 3X with PBS for 5 minutes each followed by mounting with Prolong Diamond mountant w/DAPI. Cells were imaged using a Leica DMI6000 inverted microscope. LAS-X imaging software was used to analyze images for fluorescence intensity. Cells were imaged in biological triplicate with each biological replicate consisting of a minimum of 30 cells analyzed.

All in vivo experiments were conducted according to protocols approved by the Animal Research Committee at University of Michigan.

For the HEYA8 miR-181a sensor study, 2 × 10⁴ cells were injected subcutaneously in a 1:1 mixture of DMEM media and Matrigel matrix. For each group (HEYA8 miR-181a^Low, HEYA8 miR-181a^High), a total of 10 mice were used per group with one injection site per mouse for a total of 10 tumor injections per group. Tumors were measured every other day for up to 40 days. The tumor formation endpoint was predetermined as the time point where at least one group had a majority (>5/10 mice) of mice with palpable tumors. At 40 days, all mice were euthanized, and tumors were recovered for endpoint measurements, RNA, protein, and histological analysis. For the HEYA8 ICG study, 2.5 × 10⁵ cells were injected subcutaneously in a 1:1 mixture of DMEM media and Matrigel matrix. After injection, mice were randomized into 4 treatment groups; vehicle, cisplatin, ICG-001, and cisplatin + ICG-001 with at least 6 mice per group. Mice were enrolled in the study and treatment was begun when measured tumor volume was ≥100 mm³. Vehicle mice were injected with PBS solution twice a day. Cisplatin mice were injected intraperitoneally with 2-mg/kg cisplatin every other day. ICG-001 mice were injected intraperitoneally with 100-mg/kg ICG-001 twice per day. Cisplatin + ICG-001 mice were injected intraperitoneally with 2-mg/kg cisplatin every other day and were injected intraperitoneally with 100-mg/kg ICG-001 twice per day. Mice were treated for 19 days with the primary endpoint reached when all of the mice enrolled in the vehicle control group reached a tumor volume of ≥1,600 mm³. Mice with tumors ≥2,500 mm³ before the primary endpoint were euthanized for animal welfare reasons. Once the primary endpoint was reached, all mice were euthanized, and tumors were recovered for endpoint measurements, RNA, protein, and histological analysis.

For lentiviral transfection, Lenti Starter Kit (System Biosciences) was used. Briefly, 293T cells (ATCC) were transduced with 2-μg plasmid and 10 μg of pPACKH1-plasmid mix with Lipofectamine 2000 (Life Technologies). 48 hours later, virus particles were harvested and precipitated. Target cells were transduced by plating 100,000 cells/well in a 6-well plate with virus particles and 4-μg/mL polybrene (Santa Cruz Biotechnologies).

For limiting dilution sphere assays, BD FACSAria II sorter was used to sort cells directly into 96-well ultra-low attachment plates in 200-μL mammocult media (STEMCELL Technologies) per well. After 14 days, the number of wells with tumor spheres was counted and the data were analyzed by Extreme Limited Dilution Analysis [ELDA] platform to determine the stem cell frequency.

Top-Flash plasmid (1 μg) or HES1 plasmid (1 μg) and Renilla (150 ng) were cotransfected using lipofectamine 2000 (Life Technologies). SFRP4 3′UTR reporter plasmid (1 μg) was transfected using Lipofectamine 2000. Luciferase activities were analyzed using the Dual-Luciferase Reporter Assay System (Promega) with data normalization to the corresponding Renilla values.

Total RNA was extracted using the Total RNA Purification Plus Kit (NorgenBiotek). For mRNA analysis, cDNA synthesis from1 μg of total RNA was done using the Transcriptor Universal cDNA Master kit (Roche). SYBR Green-based real-time PCR was subsequently performed in triplicate using SYBR Green Master Mix (Roche) on the Light Cycler 480 II real-time PCR machine (Roche). For miRNA expression assays, cDNA synthesis was done using TaqMan gene expression assays (Life Technologies) and subsequent real-time PCR was performed using TaqMan Universal PCR Master Mix, no AmpErase UNG buffer (Life Technologies) with corresponding probes and primer mix (TaqMan gene expression assays).

Cisplatin-sensitive parental OVCAR4 cells were cultured in RPMI medium supplemented with 2% serum. To generate, cisplatin-resistant OVCAR4-CP cells, parental cells were initially treated with 10 μmol/L cisplatin, and then dose was gradually increased to 500 μmol/L in the course of 14 days.

Cisplatin-sensitive parental OVCAR4 cells were cultured in RPMI medium supplemented with 2% serum. To generate, cisplatin-resistant OVCAR4-CP cells, parental cells were initially treated with 10 μmol/L cisplatin, and then dose was gradually increased to 500 μmol/L in the course of 14 days. To compare cisplatin sensitivity of OVCAR4-CP and parental cells, cell viability assay was performed (Supplementary Fig. S7) in 96-well format. 1,000 cells were dispensed in 100-μL RPMI medium supplemented with 2% serum. Cells were treated with varying doses of cisplatin (10–600 μmol/L), incubated for 72 hours, after which, viability was measured using the Alamar blue method.

Subsequently, cells were tested for WNT modulators LGK-974 and ICG-001 in dose–response manner (20–40,000 nmol/L) as described above. Hence, IC₅₀ value for these drugs was calculated (Supplementary Fig. S7). Later, parental and cisplatin-resistant cells were treated DMSO, LGK-974 (3 μmol/L) and ICG-001 (3 μmol/L) for 72 hours, then harvested for RNA extraction.

Unless otherwise noted, data are presented as mean ± SD from three-independent experiments, and Student t test (two-tailed) was used to compare two groups (P<0.05 was considered significant) for independent samples.

Wnt/β-catenin signaling has consistently been shown to be a critical driver of EOC recurrence and cisplatin resistance (11, 12, 15–18). Unlike other cancers, Wnt/β-catenin signaling components are not known to exhibit mutations in EOC, particularly in high-grade serous ovarian cancer (HGSOC). In addition, upstream regulators of this pathway are not well understood in the context of HGSOC (19, 20). Hence, we set out to explore the utility of our miRNA sensor platform to study the potential regulation of Wnt/β-catenin signaling by miR-181a. miR-181a is reported to induce Wnt/β-catenin signaling in acute lymphoblastic leukemia (21) but the regulation of Wnt/β-catenin signaling by miR-181a has not explored in EOC.

To assess potential functional interaction between miR-181a and Wnt/β-catenin signaling in EOC, we first assessed miR-181a 3′UTR activity in multiple EOC cell lines, including HEYA8 cells and two primary cell lines developed from primary HGSOC OCI-P5X cells and cisplatin-resistant OV81.2-CP10 cells (11, 22). Using our miR-181a sensor, we dichotomized each cell line into miR-181a^High and miR-181a^Low subpopulations (Supplementary Fig. S1) and confirmed that the miR-181a^High subpopulations exhibited increased stem cell capacity using an in vitro extreme limiting dilution sphere formation assay (ELDA; Fig. 1A). The HGSOC sensor miR-181a^High cells exhibited a 3–6-fold increase in miR-181a expression versus miR-181a^Low cells (Fig. 1B) and western blot analysis showed that miR-181a^High cells exhibited increased β-catenin protein expression (Fig. 1C). We also observed that the miR-181a^High HGSOC cells exhibited higher IC₅₀ values in response to cisplatin treatment as compared with the miR-181a^low cells (approximately increase of 2–3-fold in IC₅₀ value between miR-181a^High versus miR-181a^Low HGSOC cells; Fig. 1D and E). Furthermore, miR-181a functional knockdown in OVCAR3 cells using stable miR-181a decoy vectors (23) confirmed that targeted inhibition of miR-181a results in decreased survival, increased sensitivity to cisplatin, and reduced tumor-sphere formation ability (Supplementary Fig. S2A–S2E).

To determine whether miR-181a is a novel regulator of Wnt/β-catenin signaling in EOC, we next assessed the status of this pathway in miR-181a^high and miR-181a^low subpopulations. Examination of downstream Wnt transcriptional targets showed increase in stemness markers such as ALDH, CD133, and SOX2, as well as canonical Wnt targets such as CTNNB1, TCF7, LGR5, and LEF1 in the miR-181a^high HGSOC (Fig. 2A). Using immunofluorescence, we examined the levels of active β-catenin. We observed a 3–5-fold increase in active β-catenin levels in the HGSOC miR-181a^high subpopulations (Fig. 2B and C). Next, we evaluated nuclear β-catenin activity using TOP-eGFP and TOPFLASH reporter assays. A 2–6-fold increase in WNT signaling transcriptional activation was observed in the miR-181a^high HGSOC (Fig. 2D). We further assessed the effects of miR-181a upregulation on Wnt signaling by overexpressing miR-181a in a high-grade serous PDX-derived cell line (OV81.2; Supplementary Fig. S3A and S3B; ref. 11). We found that miR-181a overexpression increased sphere formation by 4-fold, led to cisplatin resistance, and significantly enhanced the expression of both total and active β-catenin further confirming that miR-181a activates Wnt signaling. Collectively, these results reveal that miR-181a activity is significantly associated with increased Wnt/β-catenin signaling in EOC cells, and suggest that the miR-181a/β-catenin axis is functionally important in the regulation of stem-like properties.

We next looked at the relative expression of several Wnt inhibitors that are predicted to be targets of miR-181a (SFRP4, BTRC, APC, and NLK1) to ascertain what could be driving the upregulation of Wnt/β-catenin signaling in miR-181a^high EOC. Interestingly, SFRP4 was the only candidate gene whose mRNA and protein expression was consistently low in the miR-181a^high EOC cells, suggesting that it could be a potential target of miR-181a (Fig. 3A and B). Because SFPR4 is reported to inhibit Wnt/β-catenin signaling in EOC (20), downregulation of SFRP4 could be one of the mechanisms by which miR-181a activates Wnt/β-catenin signaling. To confirm direct binding of miR-181a on the 3′UTR of SFRP4, we performed 3′UTR luciferase assays and found that the SFRP4 3′UTR expression in the miR-181a^high cells was significantly decreased (Fig. 3C). Consistent with these findings, we also found that miR-181a overexpression decreased SFRP4 3′UTR activity in OV81.2 cell line (Supplementary Fig. S3C). Next, we assessed whether miR-181a targeting of SFRP4 affected in vivo tumor formation using the HEYA8 sensor cells. We have previously shown that miR-181a^High cells have increased tumorgenicity potential, increased tumor kinetics, and total tumor burden compared with miR-181a^Low cells. Therefore, we next wanted to assess whether these tumors had differential expression of β-catenin and SFRP4 (10). We were able to confirm our previous findings, and also found that the miR-181a^Low tumors had significantly decreased expression of β-catenin and increased expression of SFRP4, indicating repression of WNT signaling in the tumors by SFRP4 (Fig. 3D–H). To confirm the clinical relevance of the miR-181a/SFRP4 axis in EOC patient tumors, we next assessed whether we could use the miR-181a sensor platform in primary patient tumor cells. We found that in tumor cells isolated from primary recurrent HGSOC (OV236), SFRP4 expression was low in the miR-181a^High subpopulation as compared with miR-181a^Low subpopulation (Supplementary Fig. S4A and S4B). Furthermore, we found that in a cohort of 23 matched primary and recurrent samples from stage III to IV patients with HGSOC, miR-181a expression was increased (4-fold increase primary vs. recurrent) and SFRP4 expression was decreased (−0.5-fold decrease primary vs. recurrent) in the primary versus recurrent tumors (Fig. 3I). Interestingly, we also found that the recurrent tumors had a significant upregulation of CD44 and β-catenin (Fig. 3I). Collectively, these results identify SFRP4 as a potentially clinically relevant target of miR-181a in EOC that leads to the activation of the Wnt/β-catenin signaling pathway.

Next, to determine whether SFRP4 was driving the stem-like properties observed in the miR-181a^High, we re-expressed SFRP4 lacking a 3′UTR in these cells and ascertained whether it could reverse these properties. SFRP4 re-expression in the OCI-P5X miR-181a^High cells (Fig. 4A and B) resulted in significant decrease in active β-catenin expression and a robust decrease in tumor-sphere formation capacity (Stem Cell Frequency Ratio of 1:26 vs. 1:3 in OCI-P5X miR-181a^High pSFRP4 vs. pTOL cells; Fig. 4C and D). Primary HGSOC OCI-P5X miR-181a^High pSFRP4 cells also had decreased expression of CSC markers, including ALDH1, CD44, and CD133 (Fig. 4E). Wnt signaling markers and TOPFLASH luciferase reporter activity were also significantly decreased in the miR-181a^High pSFRP4 cells (Fig. 4E and F). Taken together, these data show that the increase in tumor-initiating cell (TIC) properties driven by miR-181a Wnt activation is mediated by miR-181a targeting of SFRP4.

We next wanted to confirm that miR-181a activation of Wnt/β-catenin signaling was responsible for the increase in stem-like properties observed in the HGSOC miR-181a^High cells. ShRNA-mediated downregulation of β-catenin in HEYA8 miR-181a^High and OV81.2-CP10 miR-181a^High cells decreased the expression of β-catenin–driven stem cell markers that correlated with decreased stem-cell frequency in these cells, suggesting that β-catenin functions downstream of miR-181a (Fig. 5A–C). Next, we employed the use of the ICG-001 compound to specifically inhibit β-catenin/CBP transcriptional activity and assess whether suppression of β-catenin/CBP would selectively kill miR-181^high/Wnt^high cells or promote differentiation to a miR-181 low population (Fig. 5D). We found that ICG-001 treatment enriched for miR-181^low and eliminated the majority of miR-181^high cell and significantly decreasing miR-181a expression (Fig. 5E and F). In addition, ICG-001 decreased the expression of stem cell markers and downstream Wnt activation markers (Fig. 5G). Furthermore, ICG-001 also decreased tumor sphere formation and stem-cell frequency (stem cell frequency ratio of 1:2.26 vs. 1:100 in HEYA8 miR-181a^high DMSO vs. ICG-001 treated cells) further confirming that β-catenin functions downstream of miR-181a (Fig. 5H and I).

Next, to further assess whether miR-181a acts upstream or downstream of β-catenin, we activated the Wnt signaling pathway using LiCl in both the miR-181a^high and miR-181a^low cells, interestingly we found that only in the miR-181a^low cells did Wnt activation lead to a significant increase in tumor spheres (Supplementary Fig. S5A–S5E), indicating that Wnt activation in these cells was still able to induce stem-like properties. Conversely, in the miR-181a^high cells there was no increase in tumor spheres; however, the sphere size did significantly increase, suggesting that LiCl-induced Wnt activation in miR-181a^high cells was most likely still able to induce proliferative genes but not stem cell signatures.

Finally, to expand on these findings, we treated the OV81.2 miR-181a–overexpressing cells with ICG-001 and found that it was able to decrease tumor sphere formation in these cisplatin-resistant tumor spheres confirming that miR-181a–induced platinum resistance and increased stem-like properties were mediated via the upregulation of Wnt//β-catenin pathway (Supplementary Fig. S6). These results suggest that Wnt/CBP/β-catenin inhibitors such as ICG-001/PRI-724 are able to therapeutically target miR-181a/Wnt high EOC cells responsible for tumor progression.

Finally, we assessed the therapeutic efficacy of ICG-001 treatment in vivo using HEYA8 cells (which have endogenously high expression of miR-181a and Wnt activity). We found that mice treated with either ICG-001 or ICG-001/cisplatin combination showed a significant reduction in tumor volume compared with vehicle-treated mice, whereas mice treated only with cisplatin did not show a significant decrease in tumor volume (Fig. 6A). Importantly, the tumors treated with ICG-001 or ICG-001/cisplatin combination showed decreased levels of miR-181a compared with vehicle control tumors, whereas the cisplatin-treated tumors did not and no overt toxicities were noted (Fig. 6B; Supplementary Fig. S7A). We have also confirmed that other Wnt inhibitors, including LGK-974 can decrease the expression of miR-181a in OVCAR4 cisplatin-resistant models (Supplementary Fig. S8A–S8C). We then profiled the correlation between miR-181a or SFRP4 expression and tumor volumes from this in vivo tumor study and found that miR-181a expression positively correlated with tumor volume (Spearman correlation coefficient: r = 0.4976, P = 0.0083) whereas SFRP4 expression negatively correlated with tumor volume (Spearman correlation coefficient: r = −0.3823, P = 0.046; Fig. 6C and D). We also found that ICG-001 treatment led to a decrease in several Wnt and stem cell markers (Supplementary Fig. S9A). Importantly, matched miR-181a and SFRP4 expressions within the same tumor were also inversely correlated (Spearman correlation coefficient: r = −0.4656, P = 0.0114; Fig. 6E). In addition, the tumors treated with ICG-001 (which targets the miR-181a–Wnt axis) had lower expression levels of miR-181a and higher expression levels of SFRP4 (Fig. 6E). This indicated that ICG-001 was able to target the miR-181a–SFRP4 axis and miR-181a may serve as a surrogate biomarker for Wnt-targeted therapies.

Disease recurrence remains one of the primary barriers in successfully treating patients with ovarian cancer. Although HGSOCs are initially sensitive to platinum-based therapies and can remain responsive even after multiple rounds of chemotherapy, eventually 80%–90% of patients will develop recurrent disease that is cisplatin resistant and no longer responds to chemotherapy. One of the main reasons for the high rate of recurrence is that HGSOC progression is not defined by recurrent driver mutations, but rather by large-scale genomic instability involving frequent copy-number gains and losses across the entire genome. As such, a high degree of clonal heterogeneity within the tumor is present before initial treatment with chemotherapy. The abundance of this intratumor heterogeneity dramatically increases the probability of recurrence due to the increased frequency of resistant subclones.

Studying tumor subpopulations in cancers offers an approach to understand heterogeneity, and the miRNA sensor model that we have developed attempts to provide such an approach. The miRNA sensor platform can be successfully used toward deciphering miRNA-signaling pathway interactions in tumor subpopulations based on endogenous activity of the miRNA of interest. In the present study, we used an miR-181a sensor to reveal how this sensor platform can not only identify key miRNAs driving tumor phenotype, but also identify the downstream pathways that can be targeted. We have shown that miR-181a^high cell populations are able to promote recurrence and cisplatin resistance through the upregulation of Wnt/β-catenin signaling. Further analysis identified miR-181a as a novel upstream activator of Wnt/β-catenin signaling via the inhibition of SFRP4. Both genetic and chemical inhibition of this signaling pathway using shRNA-mediated β-catenin downregulation or ICG-001 in miR-181a^high cells decreased stem-like properties, thus identifying β-catenin as a functional component of miR-181a^high tumor cells and as a downstream effector of miR-181a in driving stem-like properties in HGSOC. This is further confirmed when we observed that activation of the Wnt signaling pathway in miR-181^high cells did not lead to an increase in tumor spheres as it did in miR-181^low cells, suggesting that miR-181a–induced activation of Wnt had already enriched for stem-like cells.

Wnt/β-catenin–targeted therapies have been and are still currently being explored in cancer treatment (24); however, identifying biomarkers predictive of response to Wnt/β-catenin–targeted therapy has still remained a challenge given the complexity of this pathway. Our results show that pharmacologic targeting of β-catenin resulted in robust tumor growth inhibition that was associated with decreased miR-181a and increased SFRP4. Given that circulating miR-181a can be detected in ovarian cancer patient serum (25, 26) and SFRP4 is a secreted factor, they may represent potential biomarkers for β-catenin–targeted therapies. It has been shown that low SFRP4 expression is reported to correlate with poor prognosis in ovarian cancer (27), which correlates with numerous previous observations that high miR-181a expression is indicative of poor prognosis in ovarian cancer (8, 13, 14, 28, 29). Furthermore, inverse correlation between miR-181a and SFRP4 in primary versus recurrent ovarian tumors further asserts the clinical importance of miR-181a/SFRP4 axis as a potential biomarker for Wnt-targeted therapy in HGSOC. This clinical correlation was further confirmed by miR-181a sensor sorting in primary recurrent ovarian tumor cells. High miR-181a expression correlated with low SFRP4 expression in recurrent ovarian tumor subpopulations differing in miR-181a activity. These results highlight the importance of studying miRNA pathway interactions in tumor subpopulations to enable identification of novel biomarkers for targeted therapies in cancers. Interestingly, downregulation of miR-181a by pharmacologic inhibition of β-catenin suggests a potential upstream regulation of miR-181a by β-catenin itself. Further experiments are necessary to understand whether there is a double-positive feedback loop between miR-181a and Wnt/β-catenin signaling in HGSOC.

To summarize, using a novel experimental approach that enables studying tumor subpopulations based on miRNA activity in the context of heterogeneity, we have identified miR-181a as a novel driver of Wnt activity in HGSOC. Our results suggest that miR-181a–SFRP4 axis can be evaluated as a potential biomarker to predict response to β-catenin–targeted therapy in HGSOC.

L. Beltrame reports personal fees from Eli Lilly outside the submitted work. M. Kahn reports grants from NIH during the conduct of the study and grants and other funding from 3+2 Pharma outside the submitted work. No disclosures were reported by the other authors.

A. Belur Nagaraj: Conceptualization, writing–review and editing. M. Knarr: Conceptualization, writing–original draft, writing–review and editing. S. Sekhar: Data curation. R.S. Connor: Data curation, writing–review and editing. P. Joseph: Data curation. O. Kovalenko: Data curation. A. Fleming: Data curation. A. Surti: Data curation. E. Nurmemmedov: Resources, data curation. L. Beltrame: Data curation. S. Marchini: Data curation. M. Kahn: Resources, writing–review and editing. A. DiFeo: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, writing–review and editing.

The authors thank several core facilities within the Case Comprehensive Cancer Center (P30CA043703), including the Athymic Animal and Preclinical Therapeutics Core. The authors also thank the Rogel Cancer Center at the University of Michigan (P30CA046592) for pilot funds in support of this project. This work was supported by grants from The National Cancer Institute, R01CA197780, Department of Defense, OC150553, The Young Scientist Foundation (to A. DiFeo), and CWRU Pharmacology Department MTTP Training grant (T32GM008803-11A1). In addition, the authors thank all the generous donors and foundations who have supported the DiFeo laboratory and strive to improve the outcomes of patients with ovarian cancer, including Norma C. Geller and Albert I. Geller, The Silver Family Foundation, and Jacqueline E. Bayley (JEB) Foundation.

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.