Abstract
High mortality rates in ovarian cancer are due to late-stage diagnosis when extensive metastases are present, coupled with the eventual development of resistance to standard chemotherapy. There is, thus, an urgent need to identify targetable pathways to curtail this deadly disease. In this study, we show that the apelin receptor, APJ, is a viable target that promotes tumor progression of high-grade serous ovarian cancer (HGSOC). APJ is specifically overexpressed in tumor tissue, and is elevated in metastatic tissues compared with primary tumors. Importantly, increased APJ expression significantly correlates with decreased median overall survival (OS) by 14.7 months in patients with HGSOC. Using various ovarian cancer model systems, we demonstrate that APJ expression in cancer cells is both necessary and sufficient to increase prometastatic phenotypes in vitro, including proliferation, cell adhesion to various molecules of the extracellular matrix (ECM), anoikis resistance, migration, and invasion; and these phenotypes are efficiently inhibited by the APJ inhibitor, ML221. Overexpression of APJ also increases metastasis of ovarian cancer cells in vivo. Mechanistically, the prometastatic STAT3 pathway is activated downstream of APJ, and in addition to the ERK and AKT pathways, contributes to its aggressive phenotypes. Our findings suggest that the APJ pathway is a novel and viable target, with potential to curb ovarian cancer progression and metastasis.
The APJ pathway is a viable target in HGSOC.
Introduction
Patients with ovarian cancer are typically diagnosed at an advanced stage of the disease with widespread presence of metastases, and thus have an overall 5-year survival rate of around 30% (1). High-grade serous ovarian carcinoma (HGSOC) is the most commonly diagnosed ovarian cancer subtype, and although the initial response rate of patients with HGSOC to platinum and taxane-based drugs is around 70%, they often present with recurrences that are chemoresistant (1). Hence, there is an urgent need to identify novel and targetable pathways in this deadly disease.
The apelin receptor, APJ, is a G protein–coupled receptor (GPCR) involved in regulating physiologic processes such as angiogenesis, cardiovascular development, fluid homeostasis, and reprograming of energy metabolism, when it is bound by its endogenous ligand apelin (2, 3). Although specific regulators of these processes downstream of APJ remain largely unknown, multiple players including ERK, PLCβ-PKC, and AKT cascades have been implicated (2). The role of this pathway in cancer is now slowly being recognized. APJ/apelin expression is elevated in many malignant tissues (4), indicating that its misexpression/activation may be crucial in tumorigenesis. Studies have also shown that increased expression and/or activation of this axis in the cancer cells as well as in the stroma, contributes to cancer progression mainly by increasing tumor-related angiogenesis (5–7) and proliferation (8–10). However, the extent of APJ expression by cancer cells and its pathophysiologic roles in human cancers remain largely unknown.
Herein, we demonstrate that APJ promotes HGSOC tumor progression and metastasis. Using various model systems, we show that increased APJ expression is both required and sufficient to enhance numerous prometastatic phenotypes of ovarian cancer cells in vitro, and increases metastasis in vivo. The APJ pathway is clinically relevant, as increased APJ expression in tumor tissues significantly correlates with worsened overall survival (OS) in patients with HGSOC. Taken together, our data demonstrate that the APJ pathway plays tumor-promotional roles in ovarian cancer and, hence, presents the potential for a novel therapeutic strategy.
Materials and Methods
Reagents and cell culture
Human ovarian cancer cell lines, OVCAR-5 and OVCAR-8, were purchased from the DCTD Tumor Repository, NCI (Frederick, MD). Immortalized normal fallopian tube epithelial cells (FTE188) were a generous gift from Dr. Jinsong Liu (MD Anderson Cancer Center, Houston, TX). Human normal ovarian surface epithelial cells (HOSE), OVCAR-4, and TykNu were a kind gift from Dr. Danny Dhanasekaran [University of Oklahoma Health Sciences Center (OUHSC), Oklahoma City, OK]. SKOV-3 and OVCAR-3 were a generous gift from Dr. Youngjae You (OUHSC). The cell lines were profiled via short tandem repeat profiling to confirm their identity before receipt. The cell lines were cultured in RPMI (OVCAR-3, OVCAR-4, and OVCAR-5); ATCC-EMEM (TykNu), McCoy's (SKOV-3), and 1:1 MCDB-105 and M199 media (HOSE and FTE188). Media were supplemented with 10% to 20% FBS. Cells and media were periodically tested for Mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza), and if found positive, older freezes of Mycoplasma-free cells were used. Experiments were performed on cells within 15 passages post thaw. APJ-overexpressing cells were transduced as per established protocols with pLenti-CMV-APLNR or empty vector (EV) and selected with 1 μg/mL (OVCAR-3) or 1.5 μg/mL (OVCAR-5) puromycin. Two different short hairpin RNA (shRNA) constructs (shAPJ-1: 5′-GCGCTCAGCTGATATCTTCAT-3′ and shAPJ-2: 5′-GGCTTCTAGAAGGGAAGAAAT-3′) were inserted into RNAi-ready pSIREN-RetroQ vector at the BamHI and EcoRI restriction sites. OVCAR-8 cells were transduced and selected with 2 μg/mL puromycin. Cells were treated with apelin-13 peptide (Bachem, H-4568) for the duration of the experiment as specified. The inhibitors, ML221 (Tocris, 47-481-0), GSK2126458 (denoted as GSK458; MedChem Express, HY-10297), STATTIC (MedChem Express, HY-13818), and U0126 (MedChem Express, HY-12031) were resuspended in DMSO according to manufacturer's protocol.
Patient tissue specimens
Patient samples were obtained from the Peggy and Charles Stephenson Cancer Center, OUHSC. The tissue microarray (TMA) comprised of 124 HGSOC samples from patients. Written informed consent was obtained from all women enrolled in the study. Institutional Review Board approval was provided by OUHSC. A gynecological pathologist (S. Husain) graded APJ staining (Abcam, ab84296) on a scale of 0 to 3 (0, no staining; 3, strong staining). The median score was recorded from three repeats of each tissue and association between APJ intensity and patient survival was assessed. Covariates (age at diagnosis, race, stage) were considered in the multivariate analyses using Cox model. Two-way interactions between APJ expression (high vs. low) and each of the covariates were assessed. Backward selection was used to obtain the final model.
Gene expression data were obtained from publicly available HGSOC datasets in Gene Expression Omnibus (GEO). GSE14407 (11) and GSE18520 (12) specifically provide gene expression signatures in serous ovarian cancer epithelial cells isolated by laser capture microdissection and normal ovarian surface epithelial cells (OSE). Differential APJ gene expression in primary tumor and metastases samples was analyzed with Oncomine (version 3.0; www.oncomine.org) in three publicly available ovarian cancer datasets (13, 14). APJ expression in 16 human ovarian cancer cell lines was obtained from the Cancer Cell Line Encyclopedia (CCLE).
qRT-PCR
Total RNA was extracted using HP E.Z.N.A Kit from Promega. cDNA synthesis was performed using a Maxima cDNA Synthesis Kit (Thermo Fisher Scientific). qRT-PCR assays were performed using ssoFast Evagreen supermix (Bio-Rad) and analyzed using Bio-Rad CFX96. Primer sequences are listed in Supplementary Table S1.
ELISA assay
Assay was performed following manufacturer's protocol for Extraction Free EIA Kit (Phoenix Pharmaceuticals). In brief, 50 μL of conditioned media was collected from ovarian cancer cell lines cultured under normoxic and hypoxic conditions (1% oxygen; 5% carbon dioxide, and 94% nitrogen), for 24 hours in duplicates. Concentration of apelin in the samples was determined from the standard curve of apelin ranging from 0.01 to 100 ng/mL, and according to manufacturer's protocol.
Immunoblot analysis
Whole-cell lysates (WCL) were generated as described previously (15). Briefly, RIPA (Sigma) lysis buffer and T-PER Tissue Protein Extraction Reagent (Thermo Fisher Scientific) were used to extract WCLs from cells in culture and tumor tissue, respectively. Protein concentration was determined using DC Protein reagents (Bio-Rad). Equal amounts of lysates (10–15 μg) were electrophoresed and transferred to nitrocellulose membranes. Ponceau S Stain (Sigma) was used to stain total protein on membrane to obtain a nonspecific band. Membranes were blocked in 5% BSA in TBST for 1 hour posttransfer, and incubated overnight at 4°C with primary antibody. After secondary antibody incubation, membranes were analyzed using FluorChemFC2. For cell lysates from suspended cells, 250,000 cells were plated in poly-HEMA (Sigma, 12 mg/mL in 95% ethanol)-coated plates for 24 to 48 hours. Cell clusters were centrifuged to obtain a pellet. Tumor tissue was weighed and homogenized prior to addition of lysis buffer. The antibody sources and dilutions are listed in Supplementary Table S2.
Human phospho-antibody array analysis
The phospho-antibody array analysis was performed using the Proteome Profiler Antibody Array Kit (R&D systems, ARY003B) as per manufacturer's protocols. Briefly, OVCAR-5–EV and APJ cells were plated for 48 hours, lysed with lysis buffer 6 (R&D systems), agitated for 30 minutes gently at 2°C to 8°C, and protein concentration was determined as described previously. The nitrocellulose membranes were blocked with 5% BSA in Tris-buffered saline (TBS) and Tween 20 and then treated with samples overnight on a rocking platform at 4°C. The membranes were washed with 1× wash buffer to remove the unbound protein and incubated with the mixture of biotinylated detection antibodies and streptavidin–horseradish peroxidase antibodies. Chemi-Reagent mix was applied for detection of the spot densities, and quantified using ImageJ.
Cell proliferation assay
Proliferation assays were performed by cell counting and colony formation assays. Total cell numbers were counted using the LUNA-II cell counter. Briefly, 10,000 OVCAR-4 or OVCAR-5 cells and 7,500 OVCAR-8 cells were plated in 6-well plates and treated with 10 to 100 ng/mL apelin-13 and/or 15 to 50 μmol/L ML221. After incubation for 24 to 96 hours, cells were trypsinized and suspended in equal volumes of media. An average of two replicates per sample was used for the analysis. For colony formation assays, 2,500 cells/well were seeded in 6-well plates and cultured for approximately 11 days with media replacement every 3 to 4 days. Apelin-13 and drugs were added at the time of plating as indicated in the text. Approximately on day 11, the colonies were fixed using 70% ethanol and stained using 0.4% crystal violet (CV). Images were taken using a bright field microscope (Leica, CFD365-FX). The assay was analyzed by counting colonies using ImageJ or measuring absorbance at 570 nm.
Cell adhesion assay
Cell adhesion assays were performed on 96-well fibronectin/laminin I- and collagen IV–coated plates (Biocoat, Thermo Fisher Scientific). The plates were blocked with 1% BSA in PBS for 1 hour at 37°C, followed by washes with 1× PBS. Five replicates of 40,000 cells/well were plated and incubated at 37°C for 2 to 2.5 hours. Apelin-13 and drugs were added to cells at the time of plating. The wells were washed thrice with 1× PBS to remove nonadherent cells, fixed with ice-cold methanol for 10 minutes at room temperature, and stained with 0.05% CV. The cells were destained using 10% glacial acetic acid, followed by measurement of absorbance at 570 nm corresponding to the “adhered” cells.
Anoikis resistance assay
Cells (50,000–150,000/well) were plated on poly-HEMA coated 12-well plates and treated with apelin-13 or drugs. Cells were retrieved after 24 hours (48 hours for OVCAR-8 cells), centrifuged for 5 minutes at 1,000 rpm, and replated on adherent 24-well tissue culture plates for 5 to 6 hours. Post adherence, cells were fixed with 10% formalin and stained with 0.05% CV. Bright-field images were taken using Leica either before or after CV staining, followed by destaining. Absorbance corresponding to “cells with increased anoikis resistance” was measured at 570 nm.
Cell migration assay
Migration assays were performed using Transwell 8-μm cell culture inserts (BD Falcon, 353097). 40,000 cells/well were plated in serum-free medium on Transwell filter, and allowed to migrate to medium containing 10% FBS. After 6 to 8 hours, cells from above the membrane were wiped with cotton swabs, and cells at the bottom were fixed in 10% formalin and stained with 0.05% CV. Cell migration was analyzed by counting cells using a bright field microscope (Leica) and ImageJ, or measuring absorbance at 570 nm. Apelin-13 and drugs were added at the time of plating.
Cell invasion assay
Invasion assays were performed using 8-μm Transwell cell culture inserts (BD Falcon), after coating the filters with 1:20 diluted Matrigel (Thermo Fisher Scientific, CB40230) in serum-free medium. Cells (100,000–200,000/well) were plated on the Matrigel and allowed to invade the 10% FBS-medium for 16 hours. Apelin-13 was added to the cells and to the medium at the bottom. Inhibitors were added to the cells at the time of cell plating. Cell invasion was analyzed similar to Transwell migration assays.
Dose–response assay for ML221
Dose–response assays were performed using colorimetric MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium] Assay Kit (PR G3580). Briefly, 10,000 cells were plated in 96-well plates overnight, treated with different concentrations of ML221 and incubated for 48 hours. MTS assay was analyzed according to manufacturer's protocol, with absorbance read at 490 nm and cell survival calculated as the percentage of control (DMSO-treated) group.
Each in vitro experiment was independently and successfully repeated more than three times.
Animal experiments
All animal studies were performed according to protocols reviewed and approved by the Institutional Animal Care and Use Committee at OUHSC.
Orthotopic model: OVCAR-3 cells (1 × 107) and OVCAR-5 cells (5 × 106) in PBS were intraperitoneally injected in 6-to-8-week-old female athymic nude mice (Charles River). Mouse weights were measured weekly, and mice were checked for ascites formation every 4 days. Although in some cases, mice that received APJ-overexpressing cells appeared to be moribund with extreme weight loss, no statistical trends were observed in either orthotopic model. Mice were euthanized after 55 days of injection for the OVCAR-3 model (n = 4–5 per group) and 22 days for the OVCAR-5 model (n = 9 per group), when moribund, which was similar to the time lines previously established in the literature (16). The tumor colonies were counted and collected for further analyses.
Subcutaneous model: A total of 5 × 106 OVCAR-3 cells (in sterile PBS) were injected in the left flanks of 6-to-8-week-old female athymic nude mice. OVCAR-3–APJ cells were pretreated with 100 ng/mL apelin-13 for 48 hours prior to injection, and 100 ng/mL apelin-13 was added to the cell suspension at time of injection. Tumor sizes and mouse weights were measured weekly. Tumor volume (mm3) was calculated using the formula:|\ length\ \times \ ( {widt{h^2}} )\ \times \ 0.5$| and tumor masses were counted when mice were euthanized according to IACUC protocols.
IHC staining for pSTAT3 and STAT3 and hematoxylin and eosin (H&E) stains were performed in tumor tissues isolated from mice. Quantification was done using ImageJ software. The antibody sources and dilutions are listed in Supplementary Table S2.
Statistical analysis
GraphPad Prism version 7.0 for Windows (GraphPad Software) was used for all statistical analyses. Two-tailed unpaired Student t test was used to compare pairs of conditions. One-way ANOVA nonparametric followed by Tukey/Dunnett post hoc test was used to compare more than two conditions and two-way ANOVA was used to analyze animal experiments. The mRNA expression in normal OSE and serous ovarian cancer epithelial cells was compared with the unpaired Mann–Whitney test. OS outcome was summarized using Kaplan–Meier curves and compared between groups using the log-rank test. The association of APJ expression with patient OS was evaluated using the Cox model. A P value of <0.05 denoted statistical significance.
Results
Increased APJ expression correlates with worsened prognosis in patients with HGSOC
To interrogate the specific role of apelin receptor APJ in ovarian cancer, we first screened a panel of human ovarian cancer cell lines for APJ expression. We found that on the mRNA and protein levels, ovarian cancer cells differentially express APJ independent of their classification (e.g., HGSOC carcinomas, p53 mutation status), but at a similar or higher level than that in human ovarian surface epithelial (HOSE) cells or fallopian tube epithelial (FTE188) cells (Fig. 1A). An ELISA assay showed that ovarian cancer cells secreted variable levels of the pathway ligand, apelin (Supplementary Fig. S1A), indicating that ovarian cancer cells coexpress the receptor APJ and its ligand. We also observed elevated expression of apelin in response to hypoxia, akin to what has been shown in other systems where HIF-1 regulates expression of apelin (17, 18). Analysis of APJ expression in HGSOC using publicly available datasets revealed that APJ expression was significantly higher in tumor tissues compared with in nonmalignant tissues (Fig. 1B). These studies (11, 12) were performed on cancer cells microdissected from tumor tissues, indicating that APJ is specifically upregulated in cancer cells, and not the surrounding tumor microenvironment. Further analysis in 16 human ovarian cancer cell lines using the CCLE showed that APJ expression in immortalized cell lines cultured in vitro was lower than the levels expressed by malignant cells in the tumor tissues (Fig. 1B). This suggests that APJ expression may increase in vivo. Furthermore, using Oncomine, we found that APJ expression was significantly increased in metastases compared with primary tumors in multiple human ovarian cancer patient datasets (Supplementary Fig. S1B–S1D). A meta-analysis (19) further showed that increased APJ expression correlated with worsened progression-free survival and postprogression survival in patients with serous ovarian cancer (Supplementary Fig. S1E and S1F).
We independently assessed the protein expression of APJ by performing IHC analysis in TMAs consisting of 124 HGSOC tissues (Fig. 1C). Seventy-five samples (60%) had moderate to high APJ expression (APJ high). This high expression of APJ protein (represented by both membranous and diffused staining in the cytoplasm of cancer cells) correlated with significantly reduced median OS by 14.7 months (38.5 vs. 53.2 months, P = 0.006; Fig. 1D). In addition, a multivariate Cox model revealed that high APJ expression was significantly associated with shorter OS (HR 2.1; 95% CI 1.3–3.3; P = 0.0016), after adjusting for age at diagnosis and stage. Together, our data demonstrate the pathologic significance of APJ in ovarian cancer.
APJ enhances ovarian cancer cell proliferation
To determine the role of increased APJ expression in ovarian cancer and based on the lower expression of APJ in ovarian cancer cell lines compared with the human tumor tissue, we first stably overexpressed APJ in OVCAR-5 (OVCAR-5–EV, APJ) and OVCAR-3 (OVCAR-3–EV, APJ) cells [Fig. 2A (inset); Supplementary Fig. S2A]. To determine the necessity of APJ to increase metastatic properties of ovarian cancer cells, we stably knocked down (KD) APJ using two different shRNA constructs (shAPJ-1 and shAPJ-2) in OVCAR-8 cells expressing high endogenous APJ (Fig. 2C, inset). In addition, we used OVCAR-4 cells that have high endogenous expression of APJ and secrete relatively lower levels of apelin (Supplementary Fig. S1A) to determine whether the exogenous addition of apelin-13 (the most biologically active form; ref. 20) can enhance APJ-induced activity in the cells.
Using cell counting assays (as a surrogate for proliferation), we found significantly increased cell numbers in OVCAR-5–APJ cells compared with EV cells (Fig. 2A). We observed similar effects of increased proliferation in OVCAR-3–APJ compared with EV cells, using colony formation assays (Supplementary Fig. S2B). OVCAR-4 cells treated with apelin-13 also exhibited increased cell counts compared with control (Fig. 2B); together indicating that APJ expression is sufficient to increase proliferation in ovarian cancer cells. In contrast, APJ KD in OVCAR-8 resulted in decreased cell numbers compared with control (Fig. 2C). When treated with apelin-13, only the cells with high APJ expression (OVCAR-8–shNT cells) and not the APJ KD cells, showed increases in proliferation (Supplementary Fig. S2C), further indicating the specificity of this phenotype to APJ activation. We additionally used ML221, a small-molecule inhibitor of APJ (21), to determine whether the increases in proliferation of overexpression cell lines were specific to APJ, and dose–response studies for ML221 were first performed to determine appropriate drug concentrations for use in the cells (Supplementary Fig. S3A and S3B). In both OVCAR-5 and OVCAR-4 cells, ML221 efficiently suppressed increased cell proliferation (Fig. 2A and B), and to a greater extent when APJ was overexpressed and/or activated compared with corresponding controls, especially at the later time points. Together, these data demonstrate that APJ specifically enhances the proliferation of ovarian cancer cells in vitro.
APJ increases prometastatic phenotypes of anoikis resistance and cell adhesion
Due to their prime position in the peritoneal cavity, ovarian cancer cells “slough off” from the primary tumor site as multicellular structures, and survive in suspension to metastasize (22). We, thus, examined whether APJ expression affected anoikis resistance, which is the ability of epithelial cells to survive detached from the basement membrane, a property that would be crucial for survival in the peritoneum. We found that OVCAR-5–APJ cells had significantly increased survival in suspension compared with EV cells (Fig. 3A and D), which corresponded with significantly decreased levels of cleaved-PARP (used as a marker for apoptosis; Supplementary Fig. S4A) indicating that APJ protects ovarian cancer cells from anoikis-induced cell death. We observed the same phenomenon of increased survival in suspension in OVCAR-3–APJ cells (Supplementary Fig. S4B and S4C), as well as in apelin-13–treated OVCAR-4 cells (Fig. 3B and D), compared with their respective controls. Both pharmacologic inhibitions using ML221 in the overexpression cell lines (Fig. 3A, B and D), and genetic inhibition via knockdown of APJ in OVCAR-8 cells (Fig. 3C and D) were able to restore anoikis sensitivity to different extents.
Modelling in vivo adhesion of ovarian cancer cells to the peritoneal mesothelium (22), a phenotype that would aid in more efficient formation of metastases, we next tested whether the APJ-expressing ovarian cancer cells displayed increased cell adhesion to extracellular matrix (ECM) molecules in vitro. OVCAR-5–APJ had significantly increased cell adhesion to fibronectin (FN1)/laminin-coated plates compared with EV cells (Fig. 3E and H), but not to collagen IV (Supplementary Fig. S4D and S4G). Treatment of APJ-overexpressing cells with ML221 reverted the increased cell adhesion to FN1/laminin plates to control levels (Fig. 3E and H). Interestingly, apelin-13–treated OVCAR-4 cells exhibited significant increases in adhesion to both FN1/laminin (Fig. 3F and H) and collagen IV (Supplementary Fig. S4E and S4G), which was suppressed by ML221 treatment (Fig. 3F and H; Supplementary Fig. S4E and S4G). In OVCAR-8 cells, KD of APJ significantly decreased adhesion to FN1/laminin (Fig. 3G and H) and collagen IV (Supplementary Fig. S4F and S4G) compared with control cells. Together, these data indicate that increased APJ expression/activation confers increased anoikis resistance and cell adhesion to ovarian cancer cells.
APJ increases migration and invasion of ovarian cancer cells in vitro
Using Transwell chamber assays, we examined the effect of APJ on ovarian cancer cell migration and invasion, phenotypes that mimic ovarian cancer cell movement to secondary sites, and invasion into peritoneal walls to establish metastases (22). Migration assays were carried out for 6 to 8 hours, and invasion assays for 12 to 16 hours to exclude the effects of APJ on cell proliferation. We found that OVCAR-5–APJ and OVCAR-4 cells (in response to apelin-13), had significantly increased migration (Fig. 4A and B) and invasion (Fig. 4D and E) compared with their corresponding controls. The migratory and invasive phenotypes were specific to increased APJ expression and/or activation because treatment with ML221 was able to efficiently reverse the phenotypes in vitro (Fig. 4A, B, D and E). Similarly, compared with control shNT cells, OVCAR-8 cells with APJ KD showed significantly decreased migration and invasion (Fig. 4C and F). Thus, together these data demonstrate that APJ expression is both required and sufficient to increase migration and invasion of ovarian cancer cells.
APJ functions via upregulation of STAT, AKT, and ERK signaling in ovarian cancer cells
Because APJ is a GPCR, we first examined the mechanisms involved in APJ signaling via phospho-kinase arrays using lysates from OVCAR-5–EV and APJ cells (Fig. 5A). The kinase array analyzed the levels of 43 individual phospho-proteins and 2 total proteins involved in cellular proliferation and survival. The relative phosphorylation levels of several proteins were higher in OVCAR-5–APJ cells compared with EV cells, and confirmed some known signaling pathways activated downstream of APJ such as ERK, AKT, and AMPKα1 (23). Interestingly, we found increased phosphorylation of Src family kinase members such as Src, Hck, and Fyn, and of some key STAT transcription factors including STAT5 and STAT3 (Fig. 5A).
Western blot analyses were performed to confirm a subset of phosphorylated proteins, including pERK, pAKT, and pSTAT, in OVCAR-5 (Fig. 5B) and OVCAR-3 cells (Supplementary Fig. S5A and S5B). Cancer cells grown in suspension often have distinct gene and protein expressions compared with adherent cells, which can aid in their increased survival (24). Since we observed that APJ mediates such anoikis resistance, we also examined whether those phospho-proteins had differential expression under suspended conditions. In OVCAR-5–APJ cells compared with EV cells (Fig. 5B), increases in pSTAT3 were most substantial for cells in suspension cultures in comparison with pERK, whereas pAKT was significantly upregulated in both adherent and suspended conditions. In OVCAR-3–APJ cells compared with EV cells (Supplementary Fig. S5A and S5B), we found that pERK were elevated both in adherent and suspended conditions, whereas increases in pAKT and pSTAT3 were manifested mainly under adherent conditions. Interestingly, the overall levels of the phospho-proteins were higher in the OVCAR-3–suspended cells, compared with adherent conditions. Among the tested phospho-STAT proteins, only pSTAT3 was consistently regulated across all our model systems.
To interrogate which upregulated pathways downstream of APJ contributed to its prometastatic phenotype, we used pharmacologic inhibitors, U0126 (MEK/ERK inhibitor), GSK458 (AKT inhibitor), and Stattic (STAT3 inhibitor) in OVCAR-5 cells. The specificity of the drugs and appropriate drug concentrations were first confirmed by Western blot analyses for the respective phospho-proteins in adherent conditions (Supplementary Fig. S5C). Using these inhibitors in OVCAR-5–APJ cells (Fig. 5C–G), we observed that (i) ERK inhibitor was able to abrogate APJ-induced migration but no other phenotypes; (ii) AKT inhibitor reversed the increased migration, invasion, cell adhesion, and anoikis resistance, but not clonal growth; and (iii) STAT3 inhibition efficiently suppressed all five phenotypes downstream of APJ. In OVCAR-5-EV cells, these inhibitors showed either no effect, or a similar trend but to a lesser extent (Fig. 5C–G); the latter possibly due to their basal expression of APJ. These findings indicate that although multiple prometastatic signaling cascades are activated downstream of APJ, STAT3 consistently regulates the prometastatic phenotypes downstream of APJ.
Increased APJ expression increases intraperitoneal metastasis of HGSOC cells in vivo
Because APJ expression in ovarian cancer cells increased metastasis-related phenotypes in vitro, we determined whether APJ promotes later stages of metastasis of ovarian cancer cells in vivo. Using OVCAR-3 cells, we found that compared with control mice injected with EV cells, mice injected with OVCAR-3–APJ cells had a significantly higher number of distinct large tumor masses (>1 mm in diameter) on the peritoneal wall, gastrointestinal organs, and ovaries (Fig. 6A and B). No overt ascites formation was present in either group, and hence, no differences were observed in the weights of the mice (Supplementary Fig. S6A). During a pilot study, we observed that when subcutaneously injected into the flank of mice, OVCAR-3-APJ cells produced modest increases in tumor volumes compared with control (Supplementary Fig. S6B), albeit with no statistical difference.
We confirmed the prometastatic role of APJ in another intraperitoneal model using OVCAR-5 cells. The most noticeable difference was increased area of miliary deposits on the peritoneum, and implantation into peritoneal organs (which are hallmarks of HGSOC) in mice injected with OVCAR-5–APJ cells (Fig. 6D) compared with control mice. Although miliary metastases on the abdominal walls were difficult to quantify, mice injected with OVCAR-5–APJ cells also had a significantly higher number of large nonmiliary masses (>1 mm) compared with the control mice (Fig. 6D and E). Akin to the OVCAR-3 model, the weights of mice were fairly comparable in both groups (Supplementary Fig. S6C).
Further IHC analyses of tumor tissue taken from mice injected with APJ-overexpressing cells revealed significantly increased expression of pSTAT3 protein compared with control tumors, indicating that the STAT3 pathway is activated downstream of APJ in vivo as well (Fig. 6C and F). Notably, we observed massive upregulation of APJ protein expression in those excised tumor tissues compared with the EV control tumors, the extent of which was additionally multifold higher than that in cells grown in culture (Supplementary Fig. S6D), indicating that APJ expression is indeed increased in vivo. Taken together, these data demonstrate that increased APJ expression promotes the later stages of metastasis in ovarian cancer, which includes peritoneal implantation and growth within the peritoneum cavity.
Discussion
HGSOC is the most common and lethal subtype of ovarian cancer, accounting for around 80% of diagnosed cases in the United States (1). The high mortality rate associated with this disease is largely due to late-stage diagnosis of the disease when tumors are widely metastasized, and the eventual development of resistance to conventional chemotherapeutics (22). Advanced HGSOCs are highly metastatic, and disseminate and seed extensively in the peritoneum. Hence, OS remains low, and long-term survival is significantly worse (25). Thus, the identification of novel, targetable pathways to inhibit metastatic potential of HGSOC is urgently needed, and may lead to improvements in the outcomes of patients with this disease.
Herein, we demonstrated that the APJ pathway is protumorigenic in HGSOC (Fig. 6G). We showed that APJ expression is significantly higher in malignant cells within HGSOC tissues than in normal ovarian epithelial cells, and particularly higher in metastasized tumors compared with primary tumors. Ovarian cancer cells expressed variable levels of APJ, and the observed disparity of APJ expression on the mRNA and protein may be due to either posttranslational modifications or APJ protein stability in the individual cell lines, but remains to be determined. In vitro, using various model systems, we showed that APJ was both necessary and sufficient to increase aggressive phenotypes of ovarian cancer cell lines including migration, invasion, and proliferation. We observed that not only did APJ promote anoikis resistance, cells with increased APJ expression also expressed high basal levels of PARP (Supplementary Fig. S4A). There have been reports (26, 27) that show correlations between increased PARP levels and increased chemoresistance, and studies to elucidate this interplay are ongoing. The APJ-expressing cells also had differences in adhesion to the ECM proteins that may be attributed to the differences in cell types and/or the levels of APJ in the cells. Nevertheless, the consistent increase in adhesion to FN1/laminin indicates cross-talk between integrin (28) and APJ pathways, which remains to be explored.
In our in vitro assays, while OVCAR-4 cells exhibited increased metastatic properties with exogenous addition of the ligand, such addition was not required for OVCAR-5- and OVCAR-3–APJ-expressing cells. Because ovarian cancer cells secrete apelin, we speculate that the endogenous apelin secreted by the OVCAR-3 and OVCAR-5 cells is sufficient to activate the pathway. Alternatively, it is possible that the APJ pathway is constituently active in the APJ-overexpressing cells, which negates the requirement for additional ligand. Nonetheless, both autocrine (in the case of APJ-overexpressing cells) and paracrine (in OVCAR-4 cells) activation of the pathway in ovarian cancer cells increased prometastatic phenotypes. These phenotypes were suppressed by both APJ KD and pharmacologic inhibition using ML221, further indicating the necessity and specificity of APJ for increasing aggressive phenotypes. Although we observed some effects of ML221 in the EV cells, we speculate that this may be due to the basal level of APJ expressed in those cells. Moreover, ML221 effectively inhibited some phenotypes such as cell migration and invasion at low concentrations, but higher concentrations were required to suppress others, indicating the varying sensitivity of phenotypes to the drug. Coupled with poor solubility and metabolic stability issues (29), our data further indicate the need for better APJ antagonists. Nevertheless, our in vitro data demonstrate that expression/activation of APJ by cancer cells is functional, and protumorigenic in ovarian cancer.
APJ as a GPCR is known to activate several pathways, such as the ERK, AKT, and AMPK cascades to regulate a multitude of physiologic processes (23). Notably, we found that STAT3 signaling was activated in APJ-overexpressing ovarian cancer cells, the inhibition of which had the most profound effects on APJ-mediated prometastatic phenotypes in vitro. Importantly, we demonstrated that pSTAT3 protein expression was significantly increased in APJ-overexpressing tumors indicating that the STAT3 pathway is also activated downstream of APJ in vivo. Although the mechanism by which APJ regulates STAT3 remains to be elucidated, to our knowledge, this is the first study to demonstrate STAT3 activation downstream of APJ. Thus, together, our data indicate that a combination of inhibitors against STAT3 (which itself is emerging as an important player in ovarian cancer tumorigenesis; ref. 30), and other efficacious APJ pathway antagonists such as F13A and MM54 (31–33), will aid in achieving increased tumor inhibition in ovarian cancer.
Our in vivo models revealed that high APJ expression promotes peritoneal dissemination of HGSOC, which includes seeding and growth of cancer cells within the peritoneal cavity. The most striking difference was in the number of large tumor masses in mice that were injected with APJ-overexpressing cells compared with control. This indicates that APJ confers survival and/or proliferative advantages to cancer cells (which is in line with our in vitro data) and thus promotes peritoneal dissemination. Notably, we observed that APJ protein expression was significantly increased in the cells grown in vivo compared with those grown in vitro, which is in line with the observed patient data (Fig. 1B), demonstrating that APJ expression is indeed increased in vivo. Although apelin is secreted by cancer cells themselves, and by endothelial cells (34, 35) in the tumor microenvironment, apelin is also an adipokine secreted by adipocytes, and its serum level is elevated in some noncancer-related disease states (3, 36). Thus, in vivo, it is possible that there are multiple additional sources for apelin, given that the tumor microenvironment for ovarian cancer is replete with fat reserves consisting of adipocytes. In vivo activation of the pathway may also occur via other forms of apelin, including apelin-12 and apelin-36 (37), which are not well-studied in cancer, or via elabela (38), an additional ligand for APJ, which has recently been shown to play an important role in clear-cell ovarian carcinoma (39), but this remains to be determined.
Importantly, we showed that high expression of APJ in HGSOC patient tumors was significantly associated with shorter OS by 14.7 months, suggesting a pathologic significance of APJ pathway in ovarian cancer. Although the APJ pathway has been shown to increase prometastatic phenotypes in vitro in prostate cancer and cholangiocarcinoma (10, 40) and correlate with worsened prognosis in oral squamous cell carcinoma (18), many studies have focused on the role of this axis in tumor angiogenesis (41–43). Thus, to our knowledge, our study is the first to demonstrate a prometastatic role of APJ in HGSOC, a malignancy where no effective therapies are currently available, and are hence desperately needed to improve patient outcome.
In summary, we have identified that increased APJ expression in HGSOC promotes tumor progression and metastasis. Although not examined in this study, given the role of APJ in increasing angiogenesis (44, 45), it is plausible that inhibition of the APJ pathway will cause more complete suppression of ovarian cancer tumor progression by curbing both the metastatic phenotypes in the tumor cells and angiogenesis. Thus, our studies present the APJ pathway as a universal, novel therapeutic target in this deadly and highly metastatic disease.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: D. Neelakantan, S. Dogra, S. Woo
Development of methodology: D. Neelakantan, S. Dogra, M.C. Mukashyaka, R. Janknecht
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Neelakantan, S. Dogra, B.R. Devapatla, P. Jaiprasart, S.K.D. Dwivedi, R. Bhattacharya, S. Woo
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Neelakantan, S. Dogra, R. Janknecht, S. Husain, K. Ding, S. Woo
Writing, review, and/or revision of the manuscript: D. Neelakantan, S. Dogra, M.C. Mukashyaka, R. Janknecht, R. Bhattacharya, S. Husain, K. Ding, S. Woo
Study supervision: S. Woo
Acknowledgments
The authors thank Drs. Kar-Ming Fung and Muralidharan Jayaraman, and Sheeja Aravindan for their help with the IHC experiments, as well as the OUHSC Histology and Molecular Imaging Cores for their service and technical assistance. This work was supported in part by research grants P20GM103639 (to S. Woo) from the National Institute of General Medical Sciences, NIH, and DHHS; Research Scholar Grant RSG-16-006-01-CCE (to S. Woo) from the American Cancer Society; and Gynecology Oncology Drug Development Fund (to S. Woo) from Stephenson Cancer Center.
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.