Abstract
The majority of women diagnosed with epithelial ovarian cancer eventually develop recurrence, which rapidly evolves into chemoresistant disease. Persistence of ovarian cancer stem cells (OCSC) at the end of therapy may be responsible for emergence of resistant tumors. In this study, we demonstrate that in OCSC, the tumor suppressor disabled homolog 2–interacting protein (DAB2IP) is silenced by EZH2-mediated H3K27 trimethylation of the DAB2IP promoter. CRISPR/Cas9-mediated deletion of DAB2IP in epithelial ovarian cancer cell lines upregulated expression of stemness-related genes and induced conversion of non-CSC to CSC, while enforced expression of DAB2IP suppressed CSC properties. Transcriptomic analysis showed that overexpression of DAB2IP in ovarian cancer significantly altered stemness-associated genes and bioinformatic analysis revealed WNT signaling as a dominant pathway mediating the CSC inhibitory effect of DAB2IP. Specifically, DAB2IP inhibited WNT signaling via downregulation of WNT5B, an important stemness inducer. Reverse phase protein array further demonstrated activation of noncanonical WNT signaling via C-JUN as a downstream target of WNT5B, which was blocked by inhibiting RAC1, a prominent regulator of C-JUN activation. Coadministration of EZH2 inhibitor GSK126 and RAC1 inhibitor NSC23766 suppressed OCSC survival in vitro and inhibited tumor growth and increased platinum sensitivity in vivo. Overall, these data establish that DAB2IP suppresses the cancer stem cell phenotype via inhibition of WNT5B-induced activation of C-JUN and can be epigenetically silenced by EZH2 in OCSC. Targeting the EZH2/DAB2IP/C-JUN axis therefore presents a promising strategy to prevent ovarian cancer recurrence and has potential for clinical translation.
These findings show that combining an epigenetic therapy with a noncanonical WNT signaling pathway inhibitor has the potential to eradicate ovarian cancer stem cells and to prevent ovarian cancer recurrence.
Introduction
Ovarian cancer is a chemoresponsive tumor with very high initial response rates to standard therapy consisting of platinum/paclitaxel. However, about 75% of women diagnosed with advanced stage high-grade serous (HGS) ovarian cancer experience chemoresistant relapse after first-line chemotherapy and have a 5-year survival rate of only 26% (1, 2). Recurrent ovarian cancer is essentially incurable. There is an urgent need to develop strategies and novel approaches to improve the clinical outcome for patients with ovarian cancer.
Although chemotherapy initially decreases tumor bulk, it may leave behind residual cancer stem cells (CSC) capable of regenerating tumors (3). Persistence of ovarian CSCs (OCSC) at the end of therapy has been shown to be a key contributor to resistant tumors (4). CSC possess properties such as a quiescent state, increased ABC transporter efflux and enhanced DNA damage response, rendering them “undruggable” by traditional cytotoxic interventions (5). Analogous to normal stem cells, CSC maintenance requires elaborate orchestration of the epigenome, and aberrant epigenetic alterations are common features of CSC, including OCSC (6–8), which may represent therapeutic vulnerabilities. Thus, it is of great interest to understand the (epi)genetic mechanism responsible for OCSC survival and develop strategies to target these recalcitrant cells.
Disabled homolog 2-interacting protein (DAB2IP) is a tumor suppressor frequently downregulated in multiple aggressive cancers (9, 10). In addition to possessing Ras-GTPase activity, DAB2IP serves as a multifunctional scaffold protein, and DAB2IP loss affects multiple oncogenic signaling pathways (9, 10). Recent studies in renal, prostate, and colorectal cancer demonstrated that DAB2IP expression inhibited CSC-like phenotypes, suggesting a role for DAB2IP in suppressing CSCs (11–13). Epigenetic silencing of DAB2IP by EZH2, the enzymatic subunit of the polycomb repressive complex 2 (PRC2), was reported in medulloblastoma (14) and prostate (15) cancers but whether EZH2 regulates DAB2IP regulation in CSC has not been examined.
In this study, with the goal of identifying new therapeutic targets of OCSCs to inhibit tumor relapse and overcome chemoresistance, we demonstrate that combining an epigenetic therapy with a noncanonical WNT signaling pathway has the potential to eliminate OCSCs. We show that DAB2IP was downregulated in OCSCs and could be derepressed by inhibiting EZH2-induced H3K27 trimethylation. By using transcriptomic [RNA sequencing (RNA-seq)] analysis combined with reverse phase protein array (RPPA), we establish that downregulation of DAB2IP enriched OCSC by promoting WNT5B-induced C-JUN activation in the noncanonical wingless-related integration site/planar cell polarity (WNT/PCP) pathway. Furthermore, preventing epigenetic-mediated silencing of DAB2IP with an EZH2 inhibitor and blocking WNT5B-induced noncanonical WNT signaling using a RAC1 inhibitor synergistically repressed OCSC survival in vitro and enhanced platinum sensitivity in a mouse xenograft model. We suggest that combining an epigenetic therapy with a noncanonical WNT pathway signaling pathway inhibitor may be required to eradicate ovarian cancer stem cells. We propose this promising strategy to prevent ovarian cancer recurrence has potential for clinical translation.
Materials and Methods
Cell culture
Ovarian cancer cell lines OVCAR3 and A2780 were obtained from the ATCC; COV362 was obtained from Sigma. These cell lines were authenticated by short tandem repeat (STR) analysis in 2017 (IDEXX BioAnalytics). Kuramochi cell line was obtained from the Japanese Collection of Research Bioresources and authenticated by STR analysis in 2018 (Clinical Molecular Oncology Laboratory, University of Kansas Cancer Center, KS). Cells were cultured as we have described previously (16). Cell lines were tested for Mycoplasma contamination (Lonza) every 6 months. See Supplementary Materials for more information.
Immunofluorescence and IHC
For Immunofluoresence staining, ovarian cancer cells were fixed, permeabilized, and stained (17). At least three random images from three independent experiments were taken using a light microscope. For IHC, ovarian cancer tissue microarray (TMA) OVC481 was purchased from Pantomics. See Supplementary Materials for a detailed description.
Flow cytometry
ALDH enzymatic activity was measured using the AldeRed ALDH Detection Assay Kit (Millipore) or Aldefluor Assay Kit (Stem Cell Technologies) following the manufacturer's instructions. ALDH (+) cells were determined by FACS Aria II flow cytometer (BD Biosciences) or LSRII flow cytometer (BD Biosciences). Diethylamino benzaldehyde (DEAB) was added to each sample and served as a negative control for gating purpose as we have described previously (8). For CD133 detection, 1 million cells were harvested and incubated 50:1 with CD133/1-PE-Vio615 antibody or REA Control Antibody (Miltenyi Biotec) for 10 minutes in the dark at 4°C. Cells were washed with ice-cold PBS and filtered before subjected to flow cytometry analysis. Data were collected with FACSDiva software (BD Biosciences) and analyzed with FlowJo software (FlowJo LLC).
Tumorsphere assay
Briefly, monolayer culture cells were trypsinized, and 500 or 1,000 cells were seeded into ultralow attachment plates (Corning). Cells were cultured for indicated days, observed and counted using a Zeiss Axiovert 40 inverted microscope with Axio-Vision software as we have described previously (8, 18). See Supplementary Materials for a detailed description.
CRISPR/Cas9
Ovarian cancer cells were cotransfected with DAB2IP CRISPR/Cas9 KO (sc-406961-KO-2) and DAB2IP HDR (sc-406961-HDR-2), or control CRISPR/Cas9 (sc-418922) plasmids from Santa Cruz Biotechnology using UltraCruz transfection reagent (sc-395739) and plasmid transfection medium (ssc-108062) according to manufacturer's protocol. DAB2IP CRISPR/Cas9 KO plasmid consists of a pool of three plasmids designed to disrupt gene expression by causing a double-strand break in exons within the DAB2IP gene, while the control CRISPR/Cas9 plasmid contains a nontargeting 20 nt scramble guide RNA (gRNA). Selection procedure for stably transfected clones is described in Supplementary Materials.
Immunoblotting
Cell lysates were prepared with RIPA lysis buffer (Thermo Fisher Scientific) containing Pierce protease inhibitor cocktails and Pierce phosphatase inhibitor. See Supplementary Materials for a detailed description.
Chromatin immunoprecipitation and real-time qPCR
See Supplementary Materials for a detailed description. Primer sequences were listed in Supplementary Tables S1 and S2.
MTT and migration assay
For MTT, 2,000 cells were seeded in 96-well plates in triplicates and allowed to grow for 1 day. IC50 values were calculated using Prism 6 (GraphPad Software). For migration assay, a total of 200,000 cells were seeded inside migration Boyden chamber inserts (8 mm pore; Corning) in serum-free media supplemented with 0.1% BSA as we have described previously (19). See Supplemenatry Materials for detailed descriptions of these assays.
RNA-seq analysis
RNA-seq was performed essentially as we have described previously (19, 20). Biological triplicates of OVCAR3 cells transfected with DAB2IP-overexpressing vector and empty vector were lysed with RLT, and RNA was extracted with RNeasy Mini Kit (Qiagen) according to the manufacturer's protocol. The RNA-seq results are available for download at Gene Expression Omnibus (GEO) data repository at the National Center for Biotechnology Information (NCBI) under the accession number GSE144948. See Supplementary Materials for a detailed description and bioinformatic analysis.
RPPA
OVCAR3-DAB2IP–overexpressing cells with or without recombinant WNT5B treatment, as well as empty vector control cells, were washed twice in ice-cold PBS and harvested (three biological replicates per sample). Samples were sent to the RPPA Core Facility at MD Anderson Cancer Center (Houston, TX) for further analysis. See Supplementary Materials for analysis results and bioinformatic analysis.
Cell transfection and plasmids
For EZH2 knockdown, cell lines were transfected with two independent lentivirus-based short hairpin RNAs (shRNA; MISSION shRNA, Sigma-Aldrich) targeting human EZH2 gene. Sequence of shEZH2s can be found in Supplemenatry Materials.
Mouse xenograft experiments
All animal studies were conducted according to ethical guidelines approved by the Institutional Animal Care and Use Committee of Indiana University (Bloomington, IN). A total of 2 × 106 OVCAR3 cells were mixed with matrigel at 1:1 ratio and injected subcutaneously into right flanks of NOD/SCID 5-week-old female mice (Envigo). Tumor size was measured twice a week with a caliper, and tumor volume was determined using the formula V = ½ × L × W2. When average tumor volumes reached 100 mm3, mice were randomized into eight groups (N = 6 mice/group), and corresponding treatment was initiated by intraperitoneal injection. Xenograft tumors were collected at the end of study, followed by tumor dissociation procedures. Single-cell suspension was obtained using Tumor Dissociation Kit (Miltenyi Biotec) in combination with the gentleMACS Dissociator, prior to flow cytometry analysis. Establishment of Kuramochi xenograft model can be found in Supplementary Materials.
Statistical analysis
All data are presented as mean values ± SD of at least three biological experiments unless otherwise indicated. The estimate variation within each group was similar therefore Student t test was used to statistically analyze the significant difference among different groups. Combination index (CI) was determined as reported previously (21) and see Supplementary Materials for more details.
Results
DAB2IP is downregulated in patients with ovarian cancer and correlates with poor patient survival
DAB2IP has been shown to play a key role in tumor growth inhibition in multiple types of aggressive cancers, including prostate and colorectal cancer (9). To determine whether DAB2IP is a tumor suppressor in ovarian cancer, we queried publicly available databases, including gene expression profiles in The Cancer Genome Atlas (TCGA) data portal. DAB2IP expression in ovarian tumor samples was lower compared with paired normal tissues using Gene Expression Profiling Interactive Analysis of the TCGA (Fig. 1A). A similar downregulated expression pattern was observed in several cancers, compared with corresponding normal tissues (Supplementary Fig. S1A). To determine whether DAB2IP contributes to ovarian cancer disease progression, Ovarian Cancer Database of Cancer Science Institute Singapore (CSIOVDB; ref. 22) was analyzed. Low DAB2IP expression correlated with poor disease-free survival rate (n = 707, P = 0.0105) and high-grade disease (Fig. 1B; Supplementary Fig. S1B). In addition, differential expression of DAB2IP was observed across ovarian cancer histologic subtypes, with significantly lower (P = 8.77e-12) DAB2IP expression in HGS ovarian cancer compared with low malignant potential (LMP) serous ovarian cancer (Supplementary Fig. S1C). Taken together, these results indicated that DAB2IP negatively correlated with disease progression and that low DAB2IP level was associated with poor prognosis of ovarian cancer.
Low DAB2IP links to poor patient prognosis in ovarian cancer. A, DAB2IP expression in ovarian tumor samples compared with paired normal tissues from TCGA data portal using Gene Expression Profiling Interactive Analysis. B, Kaplan–Meier survival curves of patients with high DAB2IP expression (red) and low DAB2IP expression (blue) in in CSIOVDB. C, Inverse correlation between EZH2 and DAB2IP expression at mRNA level in the ovarian cancer TCGA database. D, Representative pictures of IHC staining of DAB2IP and EZH2 on an ovarian cancer TMA. E, An inverse correlation between EZH2 staining and DAB2IP staining is shown based on quantification of TMA analyses. F, DAB2IP expression was determined using quantitative real-time PCR (qPCR) in ALDH(+) cells versus ALDH(−) cells isolated from human epithelial ovarian cancer cells (Kuramochi, OVCAR3, COV362, and A2780). G, Western blot analysis of DAB2IP and ALDH1 in ALDH(+) versus ALDH(−) cells isolated from human epithelial ovarian cancer cells. GAPDH was used as the loading control. H, Enrichment of H3K27me3 at the DAB2IP promoter was quantified by qPCR after ChIP in ALDH(+) cells versus ALDH(−) cells isolated from Kuramochi cells. I, DAB2IP expression was determined using qPCR in EZH2 knockdown ALDH(+) compared with ALDH(−) Kuramochi cells. J, Western blot analysis of DAB2IP, EZH2, and H3K27me3 in EZH2 knockdown ALDH(+) compared with ALDH(−) cells in Kuramochi cells. GAPDH and H3 were used as the loading controls. Representative data of at least three replicates. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Low DAB2IP links to poor patient prognosis in ovarian cancer. A, DAB2IP expression in ovarian tumor samples compared with paired normal tissues from TCGA data portal using Gene Expression Profiling Interactive Analysis. B, Kaplan–Meier survival curves of patients with high DAB2IP expression (red) and low DAB2IP expression (blue) in in CSIOVDB. C, Inverse correlation between EZH2 and DAB2IP expression at mRNA level in the ovarian cancer TCGA database. D, Representative pictures of IHC staining of DAB2IP and EZH2 on an ovarian cancer TMA. E, An inverse correlation between EZH2 staining and DAB2IP staining is shown based on quantification of TMA analyses. F, DAB2IP expression was determined using quantitative real-time PCR (qPCR) in ALDH(+) cells versus ALDH(−) cells isolated from human epithelial ovarian cancer cells (Kuramochi, OVCAR3, COV362, and A2780). G, Western blot analysis of DAB2IP and ALDH1 in ALDH(+) versus ALDH(−) cells isolated from human epithelial ovarian cancer cells. GAPDH was used as the loading control. H, Enrichment of H3K27me3 at the DAB2IP promoter was quantified by qPCR after ChIP in ALDH(+) cells versus ALDH(−) cells isolated from Kuramochi cells. I, DAB2IP expression was determined using qPCR in EZH2 knockdown ALDH(+) compared with ALDH(−) Kuramochi cells. J, Western blot analysis of DAB2IP, EZH2, and H3K27me3 in EZH2 knockdown ALDH(+) compared with ALDH(−) cells in Kuramochi cells. GAPDH and H3 were used as the loading controls. Representative data of at least three replicates. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Inverse correlation of DAB2IP and EZH2 in ovarian cancer tumors and cell lines
As DAB2IP inactivation due to EZH2-mediated H3K27 methylation has been reported previously (14, 15), it was of interest to investigate whether DAB2IP was associated with EZH2 transcriptional silencing in patient tumors. Analysis of transcriptomic profiles of 308 ovarian cancer tumors from TCGA data portal showed an inverse correlation between EZH2 and DAB2IP expression (Fig. 1C). Furthermore, protein expression of DAB2IP and EZH2 by IHC analysis of an ovarian cancer TMA demonstrated a significant inverse correlation (r = −0.37, P < 0.05) between DAB2IP and EZH2 (Fig. 1D and E). Expression of EZH2 in patients with ovarian cancer with different subtypes and disease grades was analyzed. Greater expression of EZH2 in high-grade disease was observed, with DAB2IP showing the opposite expression pattern (Supplementary Fig. S1B and S1D). In contrast to DAB2IP, EZH2 demonstrated significantly higher (P = 5.65e-18) expression in HGS ovarian cancer compared with the serous subtype of LMP ovarian cancer (Supplementary Fig. S1E). On the basis of the above observations in patients with ovarian cancer tumor data, it was of interest to investigate whether EZH2 negatively regulated expression of DAB2IP in ovarian cancer cell models. EZH2 knockdown increased DAB2IP expression compared with empty vector transfected cells at both mRNA and protein levels (Supplementary Fig. S1F and S1G). Furthermore, we examined H3K27me3 occupancy on the DAB2IP promoter by using four different primer sets spanning 2 kb upstream of the DAB2IP transcription start site and chromatin immunoprecipitation (ChIP) analysis followed by qPCR. H3K27me3 occupancy at the DAB2IP promoter was significantly reduced after EZH2 knockdown (Supplementary Fig. S2A and S2B). Collectively, the data indicated that DAB2IP was regulated by EZH2 in ovarian cancer and the inverse relationship was associated with patient outcomes.
DAB2IP is downregulated by EZH2 in OCSCs
Persistence of CSCs and disease progression has been reported in experimental models of ovarian cancer and in patient tumors (23, 24). We next tested whether downregulation of DAB2IP was associated with OCSC phenotypes. ALDH, a well-accepted CSC marker, was used to isolate OCSCs in our study due to its enhanced ability to initiate tumor growth as reported previously (8, 25–27). Subpopulations of CSC and non-CSC were isolated by FACS, based on ALDH activity from ovarian cancer cell lines Kuramochi, OVCAR3, COV362, and A2780. We first confirmed that the ALDH(+) subpopulation of different ovarian cancer cell lines was enriched for expression of stemness-related genes, including TWIST, NANOG, OCT4, and PROM1 (Supplementary Fig. S2C) and then investigated whether DAB2IP played a role in regulating ALDH(+) OCSCs. Expression of DAB2IP in ALDH(+) cells was significantly lower compared with non-CSC ALDH(-) cells at both mRNA and protein levels (Fig. 1F and G), indicating that DAB2IP was downregulated in OCSC population and associated with regulating stemness. Similarly, a significant increase in EZH2 expression was detected in ALDH(+) cells versus ALDH(-) cells in three ovarian cancer cell lines (OVCAR3, COV362, and A2780; Supplementary Fig. S2D). To investigate whether the underlying mechanism of DAB2IP downregulation in OCSCs was due to EZH2-H3K27me3, we examined H3K27me3 occupancy on the DAB2IP promoter. Enrichment of H3K27me3 at DAB2IP promoter loci was greater in CSCs than non-CSCs (Fig. 1H), and knocking down EZH2 in OCSCs increased DAB2IP mRNA (Fig. 1I) and reexpressed DAB2IP protein (Fig. 1J). Overall, these findings demonstrated that downregulation of DAB2IP in OCSCs was mediated by EZH2-induced H3K27me3 at the DAB2IP promoter.
EZH2-regulated DAB2IP in OCSCs inhibits spheroid formation
Although upregulation of EZH2 in side population of cells isolated from ascites of patients with ovarian cancer has been reported previously (7), the function of EZH2 in OCSC maintenance remains incompletely understood. To explore the role of EZH2 in supporting spheroid formation ability, we first examined EZH2 levels in spheroids. Immunofluorescent staining showed nuclear EZH2 expression in ovarian tumor spheroids (Fig. 2A; Supplementary Fig. S2E). To further assess the role of EZH2 in spheroid formation, ovarian cancer cell lines were stably transfected with shRNA targeting EZH2 or with nonsilencing control shRNA vector and then seeded into ultralow attachment plates. EZH2 downregulation significantly inhibited spheroid formation (Supplementary Fig. S2F–S2I), demonstrating a key role for EZH2 in spheroid formation, a hallmark of the CSC phenotype. To test whether decreased spheroid number was due to a change in cell proliferation, proliferation rate was assessed after EZH2 knockdown. The proliferation curve showed a 10%–20% decrease (Supplementary Fig. S3A and S3B), compared with an approximately 50% decrease in spheroid formation (Supplementary Fig. S2F–S2I), indicating a greater impact of EZH2 knockdown on reduced stemness to inhibit spheroid formation compared with a decrease in cell number.
EZH2-regulated DAB2IP is essential for spheroid formation. A, Immunofluorescence staining of EZH2 and DAPI in OVCAR3 spheroids. B, IC50 of OVCAR3 for EZH2 inhibitor (GSK126) was determined by MTT assay. IC50 was determined using GraphPad. C, Representative images of spheroid formation (500 or 1,000 OVCAR3 cells treated with GSK126 or vehicle control) in 24-well nonadherent conditions. D, Quantification is shown. E, OVCAR3 spheroids were treated with GSK126 or vehicle control for 5 days. ALDH(+) cell population was determined by ALDEFLUOR assay using flow cytometry. F, Quantification of E. G and H, Representative images of spheroid formation (500 or 1,000 OVCAR3 cells transfected with shEZH2s, either alone or together with shDAB2IP, and control vector; G) and quantification (H) are shown. Representative data of at least three replicates. Scale bar, 100 μm. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
EZH2-regulated DAB2IP is essential for spheroid formation. A, Immunofluorescence staining of EZH2 and DAPI in OVCAR3 spheroids. B, IC50 of OVCAR3 for EZH2 inhibitor (GSK126) was determined by MTT assay. IC50 was determined using GraphPad. C, Representative images of spheroid formation (500 or 1,000 OVCAR3 cells treated with GSK126 or vehicle control) in 24-well nonadherent conditions. D, Quantification is shown. E, OVCAR3 spheroids were treated with GSK126 or vehicle control for 5 days. ALDH(+) cell population was determined by ALDEFLUOR assay using flow cytometry. F, Quantification of E. G and H, Representative images of spheroid formation (500 or 1,000 OVCAR3 cells transfected with shEZH2s, either alone or together with shDAB2IP, and control vector; G) and quantification (H) are shown. Representative data of at least three replicates. Scale bar, 100 μm. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
To complement the genetic inhibition studies and explore whether the inhibitory effect was dependent on the enzymatic activity of EZH2, the selective EZH2 inhibitor GSK126 was utilized (Supplementary Fig. S3C). Treatment with sub IC50 concentrations of GSK126 (5 or 10 μmol/L; IC50 = 14 μmol/L) disrupted spheroid formation under nonadherent conditions (Fig. 2B–D; Supplementary Fig. S3D–S3F). Moreover, GSK126 treatment of ovarian cancer cells under nonadherent conditions resulted in decreased ALDH(+) activity based on ALDEFLUOR assay compared with negative control cells treated with DEAB (Fig. 2E and F; Supplementary Fig. S3G), supporting that enzymatic inhibition of EZH2 suppressed CSC enrichment during spheroid formation. To further explore EZH2 regulation of DAB2IP in the context of spheroid formation, DAB2IP knockdown was performed in stable EZH2 knockdown cells (Supplementary Fig. S3H). Spheroid formation ability was restored (Fig. 2G and H), indicating that DAB2IP plays a critical role as a downstream effector of EZH2 suppression of tumorsphere formation in OCSCs.
DAB2IP expression suppresses the CSC phenotype
As our findings indicated that OCSCs were deficient in DAB2IP and pointed to a previously unidentified role for DAB2IP in opposing EZH2-induced tumorsphere formation, we hypothesized that DAB2IP is a suppressor of the CSC phenotype. To test this hypothesis, we generated DAB2IP knockout (DAB2IP-KO) cells lines using CRISPR/CAS9 system (Fig. 3A). Increased expression of stemness-related genes, including OCT4, NANOG, BMI1, and PROM1, was observed in the DAB2IP-KO OVCAR3 clones (Fig. 3B). PROM1 encodes promomin-1, also known as CD133, and previous reports have demonstrated greater tumor-initiating ability in CD133(+) cells versus CD133(-) cells isolated from primary ovarian cancer tumors and cell lines (27–29). Consistent with the marked increase in PROM1 expression (Fig. 3B), the CD133(+) population, another well-accepted OCSC marker, significantly increased from 0.4% to over 20% in the DAB2IP-KO clones (Fig. 3C). In addition, enhanced tumor spheroid formation by DAB2IP-KO clones was observed (Supplementary Fig. S4A and S4B).
CRISPR knockout of DAB2IP confers cancer stemness. A, Western blot analysis of DAB2IP in OVCAR3 DAB2IP knockout (KO) and vector control cells. GAPDH was used as the loading control. B, Expression of stemness-related genes was determined using qPCR in OVCAR3 DAB2IP KO versus vector control cells. C, Flow cytometry analysis of CD133(+) cell population in OVCAR3 DAB2IP KO cells and vector control cells. D, Flow cytometry analysis of ALDH(+) cell population in OVCAR3 DAB2IP KO cells and vector control cells. E, Quantification of flow cytometry showing ALDH(+) cell population in Kuramochi and A2780 DAB2IP KO and vector control cells. F, Western blot analysis showing expression of DAB2IP and ALDH1 in Kuramochi and A2780 DAB2IP KO and vector control cells. GAPDH was used as the loading control. *, P < 0.05; ***, P < 0.001.
CRISPR knockout of DAB2IP confers cancer stemness. A, Western blot analysis of DAB2IP in OVCAR3 DAB2IP knockout (KO) and vector control cells. GAPDH was used as the loading control. B, Expression of stemness-related genes was determined using qPCR in OVCAR3 DAB2IP KO versus vector control cells. C, Flow cytometry analysis of CD133(+) cell population in OVCAR3 DAB2IP KO cells and vector control cells. D, Flow cytometry analysis of ALDH(+) cell population in OVCAR3 DAB2IP KO cells and vector control cells. E, Quantification of flow cytometry showing ALDH(+) cell population in Kuramochi and A2780 DAB2IP KO and vector control cells. F, Western blot analysis showing expression of DAB2IP and ALDH1 in Kuramochi and A2780 DAB2IP KO and vector control cells. GAPDH was used as the loading control. *, P < 0.05; ***, P < 0.001.
Despite increased cancer stemness properties and CD133(+) population, the ALDH(+) population was not increased by DAB2IP-KO in OVCAR3 cells (Fig. 3D), perhaps due to the relatively high basal ALDH activity in the OVCAR3 cell line. To address this possibility, DAB2IP knockout was performed in ovarian cancer cell lines with low basal population of ALDH(+) cells, Kuramochi and A2780 [approximately 2% ALDH(+) cells]. Enrichment of the ALDH(+) population in DAB2IP-KO Kuramochi and A2780 cells (by 1.6-fold and 3.3-fold, respectively; Fig. 3E; Supplementary Fig. S4C and S4D) was observed. ALDH1 protein level increased (Fig. 3F) and the CD133(+) population also increased compared with control cells (Supplementary Fig. S4E).
To further examine the role of DAB2IP as a suppressor of cancer stemness, GFP-tagged DAB2IP-expression vector was transfected into cells and ALDH activity was determined using the Aldered assay. Ovarian cancer cells with high level of ectopic DAB2IP expression, marked by high GFP expression, had significantly reduced ALDH(+) population (Fig. 4A) and altered stemness characteristics, including increased sensitivity to platinum (IC50 from 15.63 μmol/L to 8.27 μmol/L in OVCAR3 cells; Fig. 4B) and decreased cell motility toward attractant in migration assays (Fig. 4C). Furthermore, reexpressing DAB2IP in OCSCs decreased ALDH activity (Supplementary Fig. S5A), inhibited ALDH1 expression (Supplementary Fig. S5B), decreased cell migration capacity (Supplementary Fig. S5C), and sensitized cells to cisplatin (from 62.17 μmol/L to 30.84 μmol/L; Supplementary Fig. S5D). In addition, repressed expression of genes strongly associated with stemness was observed in OCSCs reexpressing DAB2IP (Supplementary Fig. S5E). Collectively, these results indicated that DAB2IP reprogramed CSCs to a less stem-like state.
DAB2IP overexpression inhibits stemness phenotype in ovarian cancer cells. A, Flow cytometry analysis showing ALDH(+) cell population in OVCAR3 DAB2IP-GFP–expressing and vector control cells. B, The IC50 values for cisplatin were determined by treating OVCAR3 and OVCAR3 DAB2IP-expressing cells with increasing doses of cisplatin (CDDP) for 3 hours, followed by a 3-day recovery and MTT assay to measure viability. IC50 was determined using GraphPad. C, OVCAR3 cells expressing empty vector and DAB2IP were seeded into migration chambers and cell migration was analyzed 16 hours later. Representative images (left) and quantification (right) are shown. D, Heatmap of gene expression related to stemness and ABC transporters identified by RNA-seq using OVCAR3 DAB2IP-expressing cells and control. E, Networks of biological processes constructed using significantly altered genes between OVCAR3 DAB2IP-expressing and control cells. Representative data of at least three replicates. **, P < 0.01.
DAB2IP overexpression inhibits stemness phenotype in ovarian cancer cells. A, Flow cytometry analysis showing ALDH(+) cell population in OVCAR3 DAB2IP-GFP–expressing and vector control cells. B, The IC50 values for cisplatin were determined by treating OVCAR3 and OVCAR3 DAB2IP-expressing cells with increasing doses of cisplatin (CDDP) for 3 hours, followed by a 3-day recovery and MTT assay to measure viability. IC50 was determined using GraphPad. C, OVCAR3 cells expressing empty vector and DAB2IP were seeded into migration chambers and cell migration was analyzed 16 hours later. Representative images (left) and quantification (right) are shown. D, Heatmap of gene expression related to stemness and ABC transporters identified by RNA-seq using OVCAR3 DAB2IP-expressing cells and control. E, Networks of biological processes constructed using significantly altered genes between OVCAR3 DAB2IP-expressing and control cells. Representative data of at least three replicates. **, P < 0.01.
DAB2IP alters multiple cancer-associated pathways
Transcriptomic analysis of DAB2IP-overexpressing cells and empty vector control cells was carried out using RNA-seq and bioinformatics analysis. Altered expression of 449 genes was observed (FDR < 0.05, fold change > 2; Fig. 4D; Supplementary Table S3), including important stem cell markers (ALDH1A1, LGR5, PROM1, TWIST1) and ATP-binding cassette transporters (ABCB11, ABCG1, ABCA2, ABCA3, ABCA5), closely associated with multidrug resistance in ovarian cancer (30) and other cancers (31). Ingenuity Pathway Analysis (IPA) analysis and Gene Ontology (GO) annotations of differentially expressed genes regulated by DAB2IP revealed overlapping molecular functions and predicted the inhibition of a number of key biological processes associated with tumor progression, including cell movement of tumor cell lines, angiogenesis, development of epithelial tissue, growth of malignant tumor and synthesis of DNA (red and green circles represent upregulated or downregulated genes; Fig. 4E). Moreover, gene networks revealed the potential functional impact of DAB2IP on embryonic/organismal development and organ/tissue morphology (Supplementary Fig. S6A and S6B), supporting a role for DAB2IP in modulating stem cell biology. Furthermore, DAB2IP overexpression altered malignancy-associated signaling pathways, including G protein–coupled receptor signaling, cAMP-mediated signaling, and inhibition of matrix metalloproteases (MMP; involving MMP7, MMP16, MMP10; Supplementary Fig. S6C). These results demonstrated that DAB2IP expression induced transcriptomic changes that decreased aggressiveness of ovarian cancer.
WNT5B is a direct target of DAB2IP and key regulator of OCSCs
The above differentially expressed genes were subjected to IPA to identify upstream regulators. The top upstream regulators associated with DAB2IP overexpression included mediators of the WNT/PCP pathway (JUN; P = 1.15 × 10−7) and canonical WNT/β-catenin signaling pathway (CTNNB1; P = 1.19 × 10−5; Fig. 5A). As a key role for WNT signaling in stem cell functionality has been reported previously (32, 33), it was of interest to identify potential WNT ligands altering WNT signaling. Among the WNT family members, DAB2IP overexpression significantly (P < 0.05) decreased WNT5B (fold change = –3.83; Supplementary Fig. S7A). Western blot analysis of DAB2IP-overexpressing cells demonstrated that expression of ALDH1 and WNT5B at the protein level were decreased (Fig. 5B). On the other hand, DAB2IP-KO increased WNT5B mRNA in ovarian cancer cells (Supplementary Fig. S7B), further demonstrating that DAB2IP represses WNT5B and indicating that WNT5B contributes to the stemness phenotype. To test this possibility, ovarian cancer cells were treated with recombinant WNT5B and analyzed using the Aldefluor assay. The ALDH(+) population was increased by WNT5B stimulation (Supplementary Fig. S7C). Furthermore, the addition of recombinant WNT5B reversed the anti-spheroid formation ability of DAB2IP, indicating WNT signaling pathway as a direct downstream regulator of DAB2IP (Supplementary Fig. S7D and S7E). Additional pathways altered by WNT5B treatment of OVCAR3 DAB2IP-overexpressing cells were then examined using RPPA, a high-throughput antibody-based technique (34), and subsequent IPA analysis of the significantly altered proteins was performed. Pathways related to ovarian cancer pluripotency were highly enriched because of WNT5B treatment (Supplementary Fig. S7F; Supplementary Table S4). Network analysis of significantly altered phospho-proteins identified top prediction score for biological processes of embryonic development, cell cycle, and cell signaling (Fig. 5C). In this signaling network, C-JUN was predicted as a master regulator (Fig. 5C), indicating that WNT5B induces C-JUN activation, which was confirmed by increased nuclear phosphorylated C-JUN after recombinant WNT5B treatment (Fig. 5D and E). Moreover, based on RPPA analysis, WNT5B significantly activated JNK; however, phosphorylation level of β-catenin remained unchanged (Fig. 5F and G), further demonstrating that WNT5B activates JNK/C-JUN–induced noncanonical WNT/PCP signaling.
WNT5B-induced C-JUN activation is a direct target of DAB2IP. A, Differentially expressed genes in control vector and DAB2IP-expressing OVCAR3 cells identified by RNA-seq were subjected to IPA for upstream regulators. B, Western blot analysis of DAB2IP, ALDH1, and WNT5B in DAB2IP-overexpressing and vector control cells. GAPDH was used as the loading control. C, Top network identified by IPA analysis of proteins was significantly altered in phosphorylation status after recombinant WNT5B stimulation in RPPA. D, Immunofluorescent staining of DAPI and phospho-C-JUN in OVCAR3 cells after treatment with human recombinant WNT5B treatment for one hour. E, Quantification of D. F, Heatmap of RPPA analysis. G, Plotted RPPA signal for phospho-C-JUN, phospho-JNK, and phospho-β-catenin. ***, P < 0.001.
WNT5B-induced C-JUN activation is a direct target of DAB2IP. A, Differentially expressed genes in control vector and DAB2IP-expressing OVCAR3 cells identified by RNA-seq were subjected to IPA for upstream regulators. B, Western blot analysis of DAB2IP, ALDH1, and WNT5B in DAB2IP-overexpressing and vector control cells. GAPDH was used as the loading control. C, Top network identified by IPA analysis of proteins was significantly altered in phosphorylation status after recombinant WNT5B stimulation in RPPA. D, Immunofluorescent staining of DAPI and phospho-C-JUN in OVCAR3 cells after treatment with human recombinant WNT5B treatment for one hour. E, Quantification of D. F, Heatmap of RPPA analysis. G, Plotted RPPA signal for phospho-C-JUN, phospho-JNK, and phospho-β-catenin. ***, P < 0.001.
As WNT/PCP signaling was shown to be essential for maintaining cell polarization and proper asymmetric division of stem cells (35), and previous studies have reported that JNK/C-JUN phosphorylation was due to RAC1 (36), we tested the effects of RAC1 inhibitor on reversing C-JUN phosphorylation in ovarian cancer. Western blot analysis was performed on cells transiently transfected with DAB2IP-expression vector, followed by WNT5B induction and RAC1 inhibitor NSC23766 treatment. WNT5B reactivated C-JUN phosphorylation, which was suppressed by DAB2IP overexpression; however, this activation was suppressed by RAC1 inhibitor (Fig. 6A; Supplementary Fig. S8A), demonstrating that RAC1 inhibition prevented WNT5B-induced C-JUN activation in the absence of β-catenin or AKT activation (Supplementary Fig. S8B). In addition, of the WNT signaling inhibitors tested (IWP2 inhibitor, IWR-endo inhibitor, RAC1 inhibitor, and RHO inhibitor), RAC1 inhibitor (NSC23766) was the most effective at suppressing the ALDH(+) cell population and ALDH1 expression (Supplementary Fig. S8C–S8E), suggesting a potential OCSC-targeted strategy.
Targeting RAC1 inhibits stemness phenotype and inhibiting the EZH2/DAB2IP/C-JUN axis reduces survival of OCSC. A, Western blot analysis of DAB2IP, P-C-JUN, and C-JUN in transient DAB2IP-overexpressing OVCAR3 cells treated with human recombinant WNT5B and Rac1 inhibitor (NSC23766). GAPDH and H3 were used as the loading controls. B, Quantification of flow cytometry showing CD133(+) population in OVCAR3 DAB2IP KO cells treated with Rac1 inhibitor (NSC23766) and vehicle control. C, Relative mRNA expression of PROM1 in OVCAR3 DAB2IP KO cells treated with Rac1 inhibitor (NSC23766) and vehicle control by qPCR analysis. D, Normalized isobologram for combination treatment of EZH2 inhibitor (GSK126) and Rac1 inhibitor (NSC23766) at different dosages using OVCAR3 spheroids in vitro. E, Representative images of spheroid formation at day 5 are shown. F and G, Flow cytometry showing ALDH(+) cells in OVCAR3 with different treatments (F) and quantification (G). Representative data of at least three replicates. Scale bar, 100 μm. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Targeting RAC1 inhibits stemness phenotype and inhibiting the EZH2/DAB2IP/C-JUN axis reduces survival of OCSC. A, Western blot analysis of DAB2IP, P-C-JUN, and C-JUN in transient DAB2IP-overexpressing OVCAR3 cells treated with human recombinant WNT5B and Rac1 inhibitor (NSC23766). GAPDH and H3 were used as the loading controls. B, Quantification of flow cytometry showing CD133(+) population in OVCAR3 DAB2IP KO cells treated with Rac1 inhibitor (NSC23766) and vehicle control. C, Relative mRNA expression of PROM1 in OVCAR3 DAB2IP KO cells treated with Rac1 inhibitor (NSC23766) and vehicle control by qPCR analysis. D, Normalized isobologram for combination treatment of EZH2 inhibitor (GSK126) and Rac1 inhibitor (NSC23766) at different dosages using OVCAR3 spheroids in vitro. E, Representative images of spheroid formation at day 5 are shown. F and G, Flow cytometry showing ALDH(+) cells in OVCAR3 with different treatments (F) and quantification (G). Representative data of at least three replicates. Scale bar, 100 μm. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
To further investigate whether RAC1 inhibition was capable of reversing stemness caused by DAB2IP knockout, the effect of RAC1 inhibition on DAB2IP-KO cells was tested. NSC23766 treatment inhibited the increase of CD133(+) cells (Fig. 6B; Supplementary Fig. S8F) and PROM1 mRNA expression (Fig. 6C) induced by DAB2IP-KO, demonstrating that RAC1 inhibition reversed cancer stemness downstream of DAB2IP. Furthermore, DAB2IP overexpression resulted in decreased RAC1 levels, suggesting that DAB2IP acts as a negative regulator upstream of RAC1 (Supplementary Fig. S9A). In addition, we analyzed ALDH(+) cell population after recombinant WNT5B treatment followed by RAC1 inhibitor treatment. OCSC enrichment upon WNT5B stimulation was reduced by the RAC1 inhibitor, to a level similar to RAC1 inhibitor alone (Supplementary Fig. S9B and S9C). To investigate the EZH2/DAB2IP/WNT5B/C-JUN axis, the level of WNT5B and C-JUN activation after EZH2 inhibition was examined. EZH2 knockdown downregulated WNT5B and decreased C-JUN activation and recombinant WNT5B increased P-C-JUN in EZH2-knockdown cells (Supplementary Fig. S9D and S9E), suggesting WNT5B mediates EZH2 activation of C-JUN.
Inhibiting both RAC1 and EZH2 suppresses OCSC
DAB2IP upregulation represented a key vulnerability of OCSCs, but no pharmacologic inhibitors of the tumor suppressor have been described. Therefore, with the goal of eradicating OCSCs, we chose to simultaneously target both upstream (EZH2 inhibitor; GSK126) and downstream (RAC1 inhibitor; NSC23766) regulators of DAB2IP. By using various doses of GSK126 and NSC23766 in combination (0–10 μmol/L and 0–100 μmol/L, respectively), synergistic inhibitory effects (CI < 1) on tumorsphere survival under ultralow attachment conditions were observed (Fig. 6D; Supplementary Fig. S9F). Furthermore, as tumorsphere size (Fig. 6E; Supplementary Fig. S9G) and OCSC population (Fig. 6F and G) were both significantly reduced by the combination treatment in vitro, we examined the combination treatment in vivo. We first performed a drug tolerability study in nude mice (n = 4 NOD/SCID female mice per group). Mice were treated with GSK126 (150 mg/kg, i.p.) and/or NSC23766 (2.5 mg/kg, i.p.) daily (QD5) plus cisplatin (2 mg/kg, i.p.) weekly for four cycles, and body weights were determined. Mice (all groups) gained weight during the treatment period (Supplementary Fig. S10A), demonstrating that the drugs were well tolerated.
The treatment regimen was next examined in HGSOC xenograft models. Mice were injected with OVCAR3 cells (2 × 106 cells) subcutaneously, and once tumor size reached 100 mm3, mice were randomized (n = 6 NOD/SCID female mice per group) and treated for 4 weeks (Fig. 7A). Tumor volume was measured twice a week and the experiments were terminated one week after the last cycle of treatment (day 38 after transplantation). Although groups that received single/double drug treatment showed moderate decreases in tumor volume, those decreases were not significant compared with the vehicle control-treated group. However, the three-drug combination (GSK126, NSC23766, and cisplatin) from day 24 onward significantly inhibited tumor volume compared with other treatments (Fig. 7B) and significantly reduced endpoint tumor weight (Fig. 7C and D). At the end of the study, tumors were harvested, dissociated to single-cell suspension for flow cytometry analysis, and the ALDH(+) population of cells was determined. The percentage of ALDH(+) cells was significantly decreased by the three-drug combination only (Fig. 7E; Supplementary Fig. S10B).
Combining the RAC1 inhibitor NSC23766 with the EZH2 inhibitor GSK126 inhibits tumor growth in vivo. A, A diagram illustrating the experimental design including GSK126 (150 mg/kg; daily, QD5) and NSC23766 (2.5 mg/kg; daily, QD5), plus cisplatin (2 mg/kg; weekly), followed by randomization of mice after tumor inoculation. B, Xenograft tumor growth curve of different treatment groups. C, Representative images of tumors collected at the end of the experiment. D, Tumor weight at the time of harvest. E, Quantification of flow cytometry analysis showing ALDH(+) cells in xenograft tumors from different treatment groups. F, A schematic diagram of EZH2/DAB2IP/C-JUN axis signaling in OCSCs. Representative data of at least three replicates. *, P < 0.05; **, P < 0.01.
Combining the RAC1 inhibitor NSC23766 with the EZH2 inhibitor GSK126 inhibits tumor growth in vivo. A, A diagram illustrating the experimental design including GSK126 (150 mg/kg; daily, QD5) and NSC23766 (2.5 mg/kg; daily, QD5), plus cisplatin (2 mg/kg; weekly), followed by randomization of mice after tumor inoculation. B, Xenograft tumor growth curve of different treatment groups. C, Representative images of tumors collected at the end of the experiment. D, Tumor weight at the time of harvest. E, Quantification of flow cytometry analysis showing ALDH(+) cells in xenograft tumors from different treatment groups. F, A schematic diagram of EZH2/DAB2IP/C-JUN axis signaling in OCSCs. Representative data of at least three replicates. *, P < 0.05; **, P < 0.01.
To strengthen these observations, a second xenograft model using Kuramochi cells was established and probability of survival was assessed using the same study design. A significant increase of probability of survival was observed using combination treatment compared with vehicle control (P < 0.01) or NSC23766 alone (P < 0.05) (Supplementary Fig. S10C). As proposed in Fig. 7F, inhibiting both EZH2 and RAC1 in combination with platinum chemotherapy effectively targeted the OCSC population and further decreased tumor burden compared with control or all the other single/double treatment conditions.
Discussion
Ovarian cancer is the deadliest gynecologic malignancy in the United States and more effective treatments for HGSOC are clearly needed. Toward developing new therapeutic targets for chemoresistant disease due to OCSCs, here we identify DAB2IP as an important regulator of OCSCs. We show that EZH2, the histone methyltransferase of PRC2, is responsible for epigenetic silencing of DAB2IP in OCSCs and further targeting DAB2IP reverses chemoresistance. We identify the EZH2/DAB2IP/C-JUN axis as a targetable signaling pathway to potentially eradicate OCSCs from residual tumors. We demonstrate for the first time that combining an epigenetic drug with a noncanonical WNT signaling pathway inhibitor eliminates aggressive CSC population in vitro and further enhances tumor reduction upon cisplatin in vivo. By using multiple databases, including the TCGA, we show that DAB2IP is a prognostic indicator in ovarian cancer, and patient TMA data and publicly available datasets further demonstrate a significant negative correlation between EZH2/DAB2IP and patient outcome. We propose a mechanism (Fig. 7F) based on combining an epigenetic therapy with a noncanonical WNT pathway signaling pathway inhibitor to eradicate ovarian cancer stem cells and a promising strategy to prevent ovarian cancer recurrence with potential for clinical translation.
While DAB2IP is an important multifunctional tumor suppressor in cancers (9, 10), the role of DAB2IP in ovarian cancer is limited to one study showing DAB2IP repression by a ubiquitin ligase SMURF1, resulting in altered ovarian cancer cell proliferation and gain of EMT through AKT/SKP2 signaling loops (37). By using multiple approaches including ectopic overexpression and CRISPR/CAS9 knockout in a panel of ovarian cancer cells, we demonstrate that DAB2IP suppression functionally maintains the phenotype and malignant potential of OCSCs, and re-expressing DAB2IP in OCSCs reprograms cells to a less stemness state. This finding is in agreement with other independent studies of renal, prostate, and colorectal cancer where DAB2IP is shown to be a CSC suppressor, however, through different downstream signaling pathways from our study, such as NFκB, c-kit-PI3K-AKT-mTOR, and WNT/β-catenin signaling pathway (11–13). Moreover, transcriptomic profiling and bioinformatic analysis further reveals that beyond modulating stemness-related phenotypes, DAB2IP plays previously undescribed roles in angiogenesis, cell movement, synthesis of DNA, and development of epithelial tissue in ovarian cancer, and functional alterations of the pathways (Supplementary Fig. S6C) may also contribute to reducing tumor malignancy in ovarian cancer and other malignancies.
EZH2 maintains self-renewal of adult/embryonic stem cells, and overexpression of EZH2 can induce more differentiated cells to a pluripotent state (38, 39). Here we demonstrate that spheroid formation by OCSCs is highly dependent on EZH2. Mechanistically, we show that EZH2 by suppressing DAB2IP supports OCSC maintenance and inhibiting DAB2IP in EZH2 knockdown cells rescues stemness. Our unbiased screen for EZH2/DAB2IP downstream targets reveals WNT5B-induced WNT signaling as a primary regulator downstream of DAB2IP. WNT5B is a member of the secreted WNT protein family, but how DAB2IP regulates expression of WNT5B is unknown. Transcription factors potentially regulating WNT5B, identified by PROMO (prediction of transcription factor binding sites) analysis of the WNT5B promoter, includes the period-like domain for a DAB2IP–TRAF2 direct interaction to negatively regulate NFκB (15), a possible mechanism underlying DAB2IP regulation of WNT5B.
WNT5B is a marker of poorly differentiated, mesenchymal-type cancer in diverse cell lines and in human tumor tissue samples (40). However, compared with other WNT ligands, the function of WNT5B remains largely unknown. Specifically, the WNT signaling pathway activated by WNT5B is unclear but likely to be cell-context specific. For example, in triple-negative breast cancer, WNT5B acts through WNT/β-catenin responsive genes to modulate mitochondrial biogenesis (41), whereas in oral squamous cell carcinoma, WNT5B uses both canonical and noncanonical WNT signaling pathways to induce Snail and Slug expression (42). By utilizing RPPA to identify protein level changes upon WNT5B stimulation, we demonstrate C-JUN phosphorylation independent of WNT/β-catenin, supporting a context-dependent feature of WNT signaling pathways in different types of tissue. Although there is no reported inhibitor of WNT5B, in theory RAC1 inhibitor can be used to target WNT5B-induced C-JUN activation, which phenocopied the effect of DAB2IP. Importantly, NSC23766 in combination with the EZH2 inhibitor (GSK126) synergistically eliminated OCSC, perhaps due to reinforced inhibitory effects on the same signaling pathway. EZH2 has been shown to directly repress WNT genes, with acetylation and trimethylation exerting opposing roles on H3K27 (43, 44), and inhibiting EZH2 may enhance WNT signaling unexpectedly, providing additional rationale for the greater efficacy observed using the combination treatment.
WNT/PCP signaling pathway is known to control tissue polarity and cell movement during vertebrate development, and a growing body of evidence indicates that this branch of WNT signaling is strongly implicated in cancer development at both early and late stages (45). However, compared with canonical WNT signaling pathway, WNT/PCP signaling remains incompletely understood. Correct cell polarity is responsible not only for the diversification of cell shapes, but also for regulation of the asymmetric cell divisions of stem cells, including differentiation and self-renewal (46, 47). In line with this, a subset of WNT/PCP pathway genes are highly expressed in Stem-A subtype of ovarian cancer, driving pluripotency and aggressiveness of tumor cells (48). Moreover, elevated expression of RAC1, an effector of WNT/PCP pathway, in ovarian cancer has been reported (49). In this study, by upregulating DAB2IP and suppressing WNT5B effects, NSC23766 in combination with GSK126 eliminates the OCSC population. Overall, we propose that targeting WNT/PCP signaling pathway components holds promise in ovarian cancer treatment and warrants further investigation. In this regard, ketorolac, a pan-inhibitor of the Rho family members RAC1 and CDC42, has been tested in patients with ovarian cancer (49). Ketorolac confers pharmacologic activity that inhibits RAC1 and CDC42 in patients with ovarian cancer, potentially contributing to the observed survival benefit in women treated with the drug (49).
Clinically, OCSC targeting approaches in general remain to be firmly established. Strategies such as upfront treatment to potentially eradicate OCSCs or second-line treatment to prevent the emergence of OCSC and thus recurrent, chemoresistant disease should be strongly considered (50). In our xenograft model, administration of OCSC-targeting therapy plus cisplatin demonstrates efficacy (e.g., inhibited tumor growth and reduced the OCSC population), supporting an approach of adjuvant treatment together with first-line therapy with the goal of improving disease-free survival. Whether tumors with low basal DAB2IP level or high EZH2 level respond better to this regimen remains to be determined. In summary, we have shown that EZH2 downregulation of DAB2IP in OCSCs results in WNT5B induced C-JUN activation. Our study for the first time demonstrates a functional role and molecular mechanisms of DAB2IP in OCSCs and provides strong rationale for targeting the EZH2/DAB2IP/C-JUN network in conjunction with platinum-based therapy in patients with ovarian cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
X. Zong: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. W. Wang: Investigation, methodology. A. Ozes: Conceptualization, writing-review and editing. F. Fang: Data curation, formal analysis, visualization. G.E. Sandusky: Resources, methodology. K.P. Nephew: Conceptualization, resources, data curation, supervision, funding acquisition, validation, investigation, project administration, writing-review and editing.
Acknowledgments
This work was funded by a Collaborative Research Development grant from the Ovarian Cancer Research Alliance and the Congressionally Directed Medical Research Programs (CDMRP), Department of Defense, Ovarian Cancer Research Program (OCRP) Award Number W81XWh-17-1-0076, Indiana Clinical and Translational Sciences Institute Grant Number UL1TR002529 from the NIH, National Center for Advancing Translational Sciences, Clinical and Translational Sciences Award (the content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH), the Van Andel Institute through the Van Andel Institute-Stand Up To Cancer Epigenetics Dream Team. Stand Up to Cancer is a division of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the Scientific Partner of SU2C.
We thank Dr. Anirban Mitra (Indiana University School of Medicine) and Dr. Heather O'Hagan (Indiana University School of Medicine) for helpful discussions. We thank Christiane Hassel (Flow Cytometry Core Facility, Indiana University, Bloomington, IN) for technical assistance with flow cytometry. We thank Dr. Doug Rusch and Aaron Buechlein (Center for Genomics and Bioinformatics, Indiana University, Bloomington, IN) for help with bioinformatic analysis of RNA-seq data. We thank the RPPA Core (MD Anderson Cancer Center) for RPPA experiments and bioinformatics analysis. We thank the Immunohistochemical Core Lab (Indiana University School of Medicine, Indianapolis, IN) for performing immunohistochemical staining of TMA samples.
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.
References
Supplementary data
Supplementary Table S1
Supplementary Table S2