Epithelial Ovarian Cancer-Induced Angiogenic Phenotype of Human Omental Microvascular Endothelial Cells May Occur Independently of VEGF Signaling^,

Abstract

Epithelial ovarian cancer (EOC) metastasizes transcoelomically to the peritoneum and omentum, and despite surgery and chemotherapy, recurrent disease is likely. Metastasis requires the induction of proangiogenic changes in the omental microenvironment and EOC-induced omental angiogenesis is currently a key therapeutic target. In particular, antiangiogenic therapies targeting the vascular endothelial growth factor A (VEGFA) pathway are commonly used, although, with limited effects. Here, using human omental microvascular endothelial cells (HOMECs) and ovarian cancer cell lines as an in vitro model, we show that factors secreted from EOC cells increased proliferation, migration, and tube-like structure formation in HOMECs. However, EOC-induced angiogenic tube-like formation and migration were unaffected by inhibition of tyrosine kinase activity of VEGF receptors 1 and 2 (Semaxanib; SU5416) or neutralization of VEGFA (neutralizing anti-VEGFA antibody), although VEGFA₁₆₅-induced HOMEC migration and tube-like structure formation were abolished. Proteomic investigation of the EOC secretome identified several alternative angiogenesis-related proteins. We screened these for their ability to induce an angiogenic phenotype in HOMECs, i.e., proliferation, migration, and tube-like structure formation. Hepatocyte growth factor (HGF) and insulin-like growth factor binding protein 7 (IGFBP-7) increased all three parameters, and cathepsin L (CL) increased migration and tubule formation. Further investigation confirmed expression of the HGF receptor c-Met in HOMECs. HGF- and EOC-induced proliferation and angiogenic tube structure formation were blocked by the c-Met inhibitor PF04217903. Our results highlight key alternative angiogenic mediators for metastatic EOC, namely, HGF, CL, and IGFBP-7, suggesting that effective antiangiogenic therapeutic strategies for this disease require inhibition of multiple angiogenic pathways.

Introduction

Epithelial ovarian cancer (EOC) is the most lethal of all gynecological cancers. Symptoms are often vague, leading to advanced disease with widespread metastases at diagnosis. Although EOC can metastasize through the hematogenous, lymphatic, or transcoelomic route, it is the latter that most commonly leads to metastases, with spread occurring through peritoneal and omental dissemination [1]. Although the exact mechanisms of metastasis formation by this route are not fully understood, it is widely accepted that implantation of metastatic EOC cells on the peritoneal organs is followed by the induction of angiogenesis in the host organ, which facilitates metastatic cancer growth. Integral to this process is the “switch” of local microvascular endothelial cells (ECs) to an activated phenotype that supports tumor angiogenesis.

One of the major organs susceptible to transcoelomic metastatic spread of EOC is the omentum. The observation that vascular endothelial growth factor A (VEGFA) secretion is upregulated in EOCs suggested a role for this protein in omental metastasis [2,3] and prompted the investigation of anti-VEGFA therapy in clinical trials for patients with gynecological cancers [4]. However, to date, the most studied therapy, bevacizumab (anti-VEGFA monoclonal antibody), has shown little efficacy in patients with ovarian cancer, suggesting a complex metastatic pathway involving mediators other than VEGF alone. Therefore, an understanding of the proangiogenic signaling networks activated in the omental microvasculature during suppression of the VEGFA pathways in ovarian cancer is necessary to tailor accurate antiangiogenic therapy to this specific tumor type.

It is likely that the omental metastatic spread of EOC is driven, at least partially, by the intraperitoneal environment that constitutes a dynamic reservoir of growth stimulators and prosurvival factors. However, local manipulation of the microvasculature at the site of implantation by factors locally secreted by the migrant EOC cells is also likely to play a key role in the initiation and progression of the angiogenic process. Indeed, both primary and metastasized ovarian tumor cells are known to express and/or secrete a range of key proangiogenic proteins, including various forms of VEGFs, angiopoietin-2, basic fibroblast growth factor (bFGF), hypoxia-inducible factor 1, and heparin-binding epidermal growth factor-like growth factor, as well as cytokines involved in tumor immunosuppression and metastatic progression such as interleukins 6 and 8 and transforming growth factor-β1 (TGF-β1) [5–9]. It is now recognized that the EOC metastatic cascade also involves proteases, and proteins such matrix metalloproteinases (MMPs) and cathepsins have been implicated [10–12]. However, currently the main clinical focus is on manipulating the metastasizing ovarian cancer cells rather than studying the proangiogenic responses they initiate in their target microvasculature.

Here, we tested the hypothesis that EOC cells secrete an array of factors that facilitate angiogenesis in the microvasculature, specifically ECs, of the omentum during transcoelomic metastasis. It is now well recognized that ECs from different vascular beds display considerable phenotypic heterogeneity that is reflected not only in their morphology but also in their proteome and cellular responses. It is therefore essential to study ECs from relevant vascular beds when attempting to draw disease-specific conclusions. We have previously published a technique for isolating human omental microvascular ECs (HOMECs) [13]. In this report, we use these cells to examine the influence of potential angiogenesis-associated proteins identified in EOC secretome on HOMEC phenotype. We demonstrate that ovarian cancer cells induce HOMEC proliferation, migration, and tube-like structure formation. However, inhibition of VEGFA signaling either by blocking the activity of the VEGF receptors 1 and 2 (VEGFR1/2; using SU5416) or by anti-VEGFA neutralizing antibody had no inhibitory effect on ovarian cancer cell-induced HOMEC migration and tube-like structure formation. These data strongly suggest the involvement of factors other than VEGFA in the proangiogenic activation of HOMECs when the VEGFA-VEGFR(s) pathway is disrupted. Using a range of proteomic techniques, we have identified several other potential EOC-secreted proteins that influence the overall proangiogenic cellular responses of HOMECs including hepatocyte growth factor (HGF), cathepsin D (CD), cathepsin L (CL), and insulin-like growth factor binding proteins (IGFBPs). Furthermore, we have shown that the HGF receptor c-Met is highly expressed on HOMECs and that c-Met inhibition by an ATP-competitive inhibitor of c-Met kinase (PF04217903) prevented the induction of the HOMEC angiogenic phenotype. Clinically, the action of alternative angiogenic activators such as those identified could mediate HOMEC angiogenic responses even in the presence of VEGF pathway inhibition, highlighting a potential therapeutic strategy to circumvent the ineffectiveness of anti-VEGF therapies in metastatic ovarian cancer.

Materials and Methods

Detailed reagents and equipment are listed in the Supplementary Materials and Methods section. For abbreviations, see Figure W1. All experiments were performed with at least two separate isolations of HOMECs.

Cell Culture and Collection of Conditioned Medium

Nonmalignant omental tissue samples were collected from patients at the Royal Devon and Exeter NHS Foundation Trust (Exeter, United Kingdom) with ethical approval and informed written consent. HOMECs were isolated and cultured as previously described [13]. Ovarian cancer cell lines, SKOV3 and A2780, and human dermal fibroblasts (HDFs) were purchased from the European Collection of Cell Cultures (ECACC; Salisbury, United Kingdom) and PromoCell (Heidelberg, Germany), respectively. Cancer cell lines were cultured in RPMI 1640 supplemented with gentamicin (50 µg/ml) and 10% (vol/vol) FBS and HDFs in fibroblast growth medium. HOMECs were incubated overnight in growth factor (GF)-deprived MV2 medium before treatments [2% FCS (vol/vol) only]. Collection of tumor and endothelial conditioned media (CM) is described in the Supplementary Materials and Methods section.

Proliferation, Migration, and Fibrin Matrix/Geltrex Tube Formation Assays

For assay details, see Table W1.

Proliferation was assessed using the water soluble tetrazolium salt-1 (WST-1) assay in 96-well plates. HOMECs were incubated overnight in GF-deprived MV2 medium before treatment. After times indicated, WST-1 reagent was added in a 1:10 dilution to the assay medium for a 2-hour incubation and absorbance was measured against the blank on a microplate reader.

The ThinCert migration assay was carried out according to the manufacturer's protocol, with minor changes. SKOV3 and A2780 cells and HDFs were plated on the bottom of a 24-well plate. After 24 hours, HOMECs were seeded into the inserts and the assay was assembled. Some wells with cancer cells were supplemented with neutralizing anti-VEGFA antibody (500 ng/ml), and some with SU5416 (10 µM). The concentration of anti-VEGFA antibody required to effectively neutralize VEGFA secreted by SKOV3 and A2780 cancer cells over 72 hours was empirically established before the assay (data not shown). Control wells contained no cells. After a 24-hour incubation at 37°C, migrated cells were labeled with calcein AM and the inserts were transferred to a fresh 24-well plate containing prewarmed trypsin for incubation with shaking. Detached, migrated HOMECs were allowed to settle, and the fluorescent signal was measured on a plate reader.

The Oris Cell migration assay was carried out according to the manufacturer's recommendations in a 96-well plate. Briefly, HOMECs were incubated overnight in GF-deprived MV2 medium before treatments. The assay medium was supplemented with different proteins (as indicated), SU5416 (10 µM), or with fractions of CM from SKOV3 cells (TCM) or A2780 cells (TCMA; see Fast Protein Liquid Chromatography section). Controls included GF-deprived MV2 medium. After 24 or 48 hours at 37°C, the detection mask was attached to the bottom of the plate, and calcein AM-labeled, migrated HOMECs were quantified using a fluorescent plate reader.

Two tube-forming assays were used. Initial studies were carried out using a fibrin matrix assay (indirect co-culture contact through common medium) as previously described [14]. HOMECs were seeded onto fibrin matrices in 10-mm diameter rings or on matrices prepared in 24-well plates and incubated overnight in GF-deprived MV2 medium. For co-culture experiments, ovarian cancer cells (SKOV3 and A2780) were plated onto the bottom of the wells (around the rings containing the matrices) and incubated for 4 hours until attached. Fresh GF-deprived MV2 medium was added (to cover the wells above the levels of the rings) either alone (control) or supplemented with VEGFA₁₆₅ (positive control) or cancer-secreted proteins (including 40 ng/ml CL, 50 ng/ml HGF, and 50 ng/ml IGFBP-7). SU5416 (Semaxanib) was included in wells with inhibitory treatments. After 72 hours at 37°C, tubule structure formation was analyzed using bright-field microscopy or fluorescent microscopy for cells labeled with calcein AM. ImageJ software was used for quantification. The results are presented as a percentage of control (tube-like structure index).

Further studies used a reduced GF basement membrane extract (Geltrex) tube formation assay carried out in 96-well microplates with 50 µl of Geltrex per well. For co-culture experiments, ovarian cancer cells (A2780) were mixed with Geltrex before loading into wells. Geltrex was allowed to gel for 30 minutes at 37°C, and then calcein AM-labeled HOMECs were seeded onto matrices. Fresh GF-deprived MV2 medium (100 µl/well) was added to control and co-culture control wells. Assay medium was supplemented with HGF (50 ng/ml) ± PF04217903 (20 nM) or SU5416 (10 µM) as indicated. Assays were developed for approximately 4 hours in a 37°C humidified incubator with 5% CO₂ followed by fluorescent image acquisition (one picture per well; total magnification, x40). Images were then processed with Corel Paint Shop Pro Photo X2 to ensure a format accessible to AngioSys software (Caltag Medsystems, Little Balmer, United Kingdom). Automated quantification was then performed using this software to determine the number of junctions, number of tubules, and total tubule length.

Fast Protein Liquid Chromatography

Whole CM from A2780 or SKOV3 cells (15.5 ml) were sequentially fractionated using Amicon Ultra Centrifugal Filter Units (Merck Millipore, Watford, United Kingdom) with 100,000 and then 10,000 molecular weight (MW) cutoff to produce a 100 ≤ 10 kDa prefraction reduced to a volume of ∼400 µl. The concentrated prefractions were then injected through a 0.5-ml loop onto a Superdex 75 10/300 GL column (GE Healthcare, Little Chalfont, United Kingdom) of an ÄKTApurifier. The flow rate was set at 1 ml/min, and 1-ml fractions were collected using basal MV2 medium (for bioassays) or phosphate-buffered saline (PBS; for ELISA). The relatively large fraction volume facilitated fraction analysis but reduced protein molecular weight resolution. Approximate size of unknown proteins in eluted fractions was assessed by gel filtration molecular weight markers (weight range of 6500–2,000,000 Da; Sigma, Gillingham, United Kingdom; see Figure W1A) run through the same system. Fractions were stored at -80°C until used.

Proteome Profiling (Mass Spectrometry and Antibody Array)

Proteins from a mixed lot of whole CM from SKOV3 cells and control Dulbecco's modified Eagle's medium (DMEM)-Bottenstein-Sato (BS) medium were affinity purified using StrataClean resin according to the manufacturer's protocol (StrataGene, Wokingham, United Kingdom). Proteins bound to the beads were sent to Robert Jones and Agnes Hunt Orthopedic and District Hospital for mass spectrometry (MS) analysis. For details of MS protocol, see Supplementary Materials and Methods section.

Protein content of fractions 11 and 12 of A2780 and SKOV3 CM after fast protein liquid chromatography (FPLC) was analyzed using the R&D Systems (Abingdon, United Kingdom) Proteome Profiler antibody array (human angiogenesis) according to the manufacturer's protocol. Array data on developed X-ray film were quantitated by densitometry using Quantity One software (Bio-Rad, Hemel Hempstead, United Kingdom).

Enzyme-Linked Immunosorbent Assays

All ELISAs were carried out according to the manufacturer's recommendations, except IGFBP-7, which was developed in-house (for details, see Table W2). For this ELISA, microplates were coated with capture antibody and incubated overnight at room temperature. Between all stages of the ELISA, the wells were washed three times with 0.05% (vol/vol) Tween-20/PBS. Blocking was in 1% (wt/vol) BSA/PBS. Standards, blanks, detection antibody, and streptavidin-HRP conjugate were prepared in 0.1% (wt/vol) BSA/PBS. All samples and standards were added in duplicate followed sequentially by detection antibody, streptavidin-HRP conjugate, 3,3′,5,5′-tetramethylbenzidine (TMB) substrate, and stop solution as described in Table W2. Following measurement of absorbance, the standard curve and concentrations of measured IGFBP-7 were obtained using data analysis software supplied by BMG Labtech (Aylesbury, United Kingdom).

Immunocytochemistry

HOMECs were cultured on 2% (wt/vol) gelatin-coated fluorodishes until confluent. Staining for CD31 and c-Met was performed after fixation with 4% (wt/vol) paraformaldehyde and blocking with goat serum (1:10 in PBS). Mixed primary antibodies diluted in PBS were applied for 60 minutes. Nonspecific staining controls received PBS alone. After washing, mixed secondary antibodies were applied for 30 minutes. All incubations were performed at room temperature. Cells were counterstained with 4′,6-diamidino-2-phenylindole (DAPI), mounted, and analyzed using a fluorescent microscope and imaging software (Openlab 4.0.4, Corel Paint Shop Pro Photo X2). For procedure and antibody details, see Table W3.

Results

Factors Released from Ovarian Cancer Cells Increase HOMEC Proliferation, Migration, and Tube-Like Structure Formation

Angiogenesis requires phenotypic changes in normal relatively quiescent ECs. In tumor-induced angiogenesis, the factors activating ECs and initiating these cellular changes originate in cancer cells and tumor-associated cells. Therefore, our initial experiments assessed the mitogenic, migratory, and angiogenic potential of the total secretome of ovarian cancer cells on HOMECs using CM or co-culture techniques. CM from two different ovarian cancer cell lines (SKOV3 and A2780) induced a significant time-dependent increase in proliferation of HOMECs (Figure 1A; **P ≤ .01 and ***P ≤ .001, n = 4). Furthermore, HOMECs actively migrated in response to VEGFA165 or when co-cultured with SKOV3 and A2780 cells, but this effect was not observed with the normal HDFs (Figure 1, B and C; **P ≤ .01 and ***P ≤ .001, n = 12). Finally, we showed that HOMEC can undergo VEG-FA₁₆₅-induced (positive control) and EOC-mediated tube-like structure formation in an in vitro tube formation assay (Figure 1E; *P ≤ .05, **P ≤ .01, and ***P ≤ .001, n = 8–16).

Figure 1.

Ovarian cancer cell secretome increases HOMEC cellular responses and contains a range of potential proangiogenic proteins. (A) Proliferation of HOMECs treated with CM from SKOV3 and A2780 cells for 24, 48, or 72 hours assessed using the WST-1 assay. Controls received basal medium only. **P ≤ .01 and ***P ≤ .001 versus control levels (100%). (B) Migration of HOMECs in co-culture with ovarian cancer cells and HDFs (co-culture control) assessed by the ThinCert migration assay. After 24 hours, fluorescence of migrated cells was quantified. ***P ≤ .001 versus control (100%) and ^###P ≤ .001 versus HDFs. (C) HOMEC Oris cell migration after 48 hours in the presence of VEGFA165 (20 ng/ml) ± SU5416 (10 µM). Controls received medium alone. ***P ≤ .001 versus control (100%) and ^###P ≤ .001 versus VEGFA165. (D) Migration of HOMECs in co-culture with ovarian cancer cells ± anti-VEGFA antibody (500 ng/ml) or SU5416 (10 µM) assessed by the ThinCert migration assay. ***P ≤ .001 versus control (100%), **P ≤ .01 versus control (100%), and NS versus A2780 or SKOV3. (E) VEGF- and EOC cell-induced tube-like structure formation of HOMECs and the effects of the SU5416 inhibitor. HOMECs were plated onto fibrin matrices and exposed to either VEGFA165 or EOC cells in co-culture ± SU5416 (10 µM). Negative controls received medium alone. Tube-like structure formation was quantified. *P ≤ .05, **P ≤ .01, and ***P ≤ .001 versus control (100%), ^###P ≤ .001 versus VEGF, and NS versus SKOV3/A2780. For A to E, data are presented as means ± SD. P values were calculated by one-way analysis of variance (ANOVA), followed by Tukey's post hoc test. (F) Shotgun proteomics of SKOV3 CM by the 4800 MALDI TOF/TOF analyzer. Significant results were determined by selecting proteins that matched with two or more peptides with total ion CIs of >95%, followed by a literature search to determine proteins with a possible role in angiogenesis. (G) Analysis of CL, CD, IGFBP-7, and VEGF concentration in CM from SKOV3 and A2870. Commercially available ELISAs were used for determination of VEGF, CD, and CL concentrations, whereas the IGFBP-7 ELISA protocol was developed in-house. All experiments were carried out in duplicate on three separate samples. B/D, below detection limit.

Ovarian Cancer-Mediated Migration and Tube-Like Structure Formation Is Unaffected by Inhibition of the VEGF Pathway

Given the accepted importance of VEGFA to the angiogenic process, we examined whether inhibition of VEGFR1/2 in HOMECs would alter in vitro tube-like structure formation and migration. SU5416 is a potent synthetic inhibitor of the tyrosine kinase activity of VEGFR1/2, previously shown to effectively inhibit VEGFA₁₆₅- and placenta growth factor 1 (PlGF-1)-induced autophosphorylation of both receptors at concentrations ≥0.5 µM [15]. Our previous data showed that in human umbilical vein ECs (HUVECs) SU5416 abolished VEGFA₁₆₅-induced in vitro tube-like structure formation at a concentration of 10 µM [16]. Figure 1 indicates that VEGFA₁₆₅ induced migration and tube-like structure formation in HOMECs were significantly inhibited by SU5416 but that EOC-dependent migration and tube-like structure formation were unaffected by the presence of the inhibitor (Figure 1, C–E; ^###P ≤ .001, n = 12; **P ≤ .01 and ***P ≤ .001, n = 9; ^###P ≤ .001, n = 8–16, respectively). To examine this further, we investigated whether depletion of VEGFA from the co-culture system by neutralizing anti-VEGFA antibody influenced HOMEC migration. VEGFA neutralization had no significant effect on migration of HOMECs (Figure 1D; NS for SKOV3 + antibody/SKOV3 + SU5416 and A2780 + antibody/A2780 + SU5416 vs SKOV3 and A2780, respectively). These results indicated that HOMECs can undergo formation of tubular networks, proliferation, and migration in response to factors secreted by ovarian cancer cells but that suppression of the VEGFA pathway was not sufficient to prevent EOC-mediated HOMEC migration and tube-like structure formation. This suggests the involvement of additional proangiogenic mediators in EOC-induced metastatic angiogenic responses in the omental microvasculature.

Ovarian Cancer Cells Secrete a Range of Factors with a Potential Role in Omental Metastatic Angiogenesis

To further investigate alternative mechanisms for EOC-induced HOMEC tube-like formation and migration, we initially screened, using MS, whole CM from SKOV3 cells for the presence of angiogenesis-related proteins. Although the MS search identified a range of proteins in the CM (see Table W4), only the following three prospective candidates were identified: CD, CL, and IGFBP-7 (Figure 1F). Surprisingly, no GFs or cytokines were detected using this approach, presumably due to limitations of MS in identifying proteins in complex mixtures, such as CM, and instrument limitation. Thus, ELISAs were carried out to confirm the presence of VEGFA as well as to quantify MS-identified factors in tumor CM of SKOV3 and A2780 cancer cell lines. Figure 1G shows the concentrations of VEGFA (mixed isoforms), CD, CL, and IGFBP-7 in CM from both cancer cell lines. bFGF was not detected (data not shown). Clearly, concentrations of measured proteins varied between cancer cell lines, favoring A2780 as a more potent inducer of HOMEC cellular changes (see Figure 1).

Fractionation of CM from EOC Cells Reveals Specific Bioactive Fractions that Stimulate HOMEC Migration

Having identified several angiogenesis-related proteins secreted from EOC cells, we next carried out a series of experiments to examine the possible presence of additional, previously undetected potential mediators. In initial studies, we performed FPLC size exclusion fractionation of the CM from both cancer cell lines and screened the fractions for their promigratory bioactivity on HOMECs. Initially, a 100 ≤ 10 kDa prefraction was prepared from the CM because most bioactive factors were likely to be in this size range. SKOV3 and A2780 cells displayed distinct profiles of secreted factors, although all fractionated proteins were found in fractions 8 to 21 for both cell lines (Figures 2, A and C, and W1A). Functional screening of HOMEC migration in response to these fractions using the Oris cell assay indicated significant promigratory biologic activity in fractions 11 and 12 from both SKOV3 and A2780 cells (Figure 2, B and D; **P ≤ .01, *P ≤ .05, and ***P < .001, n = 21, respectively). These data suggested that the key promigratory factors secreted from the EOC cells were enriched in these fractions, and therefore, the protein content of these fractions was investigated further to examine 1) whether the proteins already identified above, i.e., CL, CD, and IGFBP-7, were present in these fractions and 2) whether additional, as yet unidentified targets could be discovered.

Figure 2.

Bioactive effects of fractionated A2780 and SKOV3 CM on HOMEC migration. (A and C) Relative protein content of fractionated CM from SKOV3 (A) and A2780 (C) assessed by absorbance at 280 nm. A 100 ≤ 10 kDa prefraction of both tumor CM was separated by size exclusion gel filtration, and 1-ml fractions were collected. Distinct protein peaks were observed in both profiles. (B and D) Fractions 11 and 12 of CM from SKOV3 (B) and A2780 (D) obtained by FPLC increase HOMEC migration. The bioactivity of protein-rich fractions (11 and 12) produced by FPLC was assessed in HOMECs using the Oris Cell migration assay. The experiment was performed on three separate cell isolations in heptaplicate. Data are presented as means ± SD. *P ≤ .05, **P ≤ .05, and ***P < .001 versus control (100%); P values were calculated by the Mann-Whitney U test.

Bioactive Fractions 11 and 12 Contain Additional Proteins Able to Influence HOMEC Proliferation and Migration

Two different approaches were undertaken to identify the secreted proteins in the promigratory fractions of EOC cell CM. Fractions 11 and 12 were firstly analyzed using a Proteome Profiler human angiogenesis kit. Although this profiler does not include all of the proteins previously identified, it did confirm the presence of VEGFA in both fractions of SKOV3 and A2780 EOC cells (Figure 3A). For a full description of the proteins analyzed and representative images of the profiler X-ray films, see Supplementary Materials and Methods section (Figure W1, B–D). Again, no secreted bFGF was detected. The presence of CL and CD in the bioactive fractions from both ovarian cancer cell lines was confirmed by ELISA, although fraction 11 of both cancer cell lines contained greater concentrations of CD and CL than fraction 12 (Figure 3B). IGFBP-7 was only detected in fraction 11 of SKOV3 CM. Additionally, both cancer cell lines secreted other potent modulators of angiogenesis/proliferation, i.e., endocrine gland derived (EG)-VEGF, HGF, PlGF-1, PDGF-AA, neuregulin 1-β 1 (NRG1-β1), serpin F1 [pigment epithelium-derived factor (PEDF)], and angiogenin. Interestingly, results indicated distinct qualitative and quantitative differences between the two cell lines. For instance, VEGFA and HGF were found in both tumor CM; however, levels of both were higher in A2780 cells, particularly HGF. Regulators of extra-cellular matrix (ECM) remodeling, including MMP8, MMP9, serpin E1 [plasminogen activator inhibitor type 1 (PAI-1)], tissue inhibitor of metalloproteinases-1 (TIMP-1), and urokinase-type plasminogen activator (uPA) were also secreted by both ovarian cancer cell types, as were additional IGFBPs (implicated in vascular repair and homeostasis, cell proliferation, and apoptosis).

Figure 3.

Identification of angiogenesis-associated proteins in fractions 11 and 12 of CM from SKOV3 and A2780 ovarian cancer cells. (A) Fractions 11 and 12 of both CM were profiled using a Proteome Profiler antibody array (human angiogenesis). The results were quantified by densitometric analysis and are presented as a mean of duplicates corresponding to one target protein. Background and adjustment factors were subtracted, and angiogenesis-associated proteins with values above the adjustment level are shown. (B) CD, CL, and IGFBP-7 in fractions 11 and 12 of both CM were examined by ELISA because these targets were not included in the antibody array. For CD and CL, commercially available ELISAs were employed, whereas for IGFBP-7 an in-house ELISA was used. Each ELISA was carried out in duplicate on two or three separately collected lots of fractions 11 and 12. B/D, below detection limit of ELISA.

Taken together, these findings indicate that ovarian carcinoma cells synthesize and secrete a variety of factors that may influence HOMEC angiogenic phenotype. These include ligands of VEGFRs/neuropilins (i.e., VEGF and PlGF-1), ligand of c-Met (HGF), proteases (MMP9, MMP8, and uPA), protease inhibitors (TIMP-1 and serpin E1), chemoattractants (CXCL16 and PF4), antiangiogenic factor PEDF, and IGFBPs (modulators of IGF/insulin signaling).

To examine which of the identified proteins may have been responsible for the original biologic effects of the whole CM, relevant individual proteins present in the bioactive fractions were assessed for their proliferative and migratory potential (Figure 4, A and B). The results confirmed significant induction of proliferation by HGF and IGFBP-7 and a weak (borderline) proliferative effect of PlGF-1 on HOMECs (Figure 4A; *P ≤ .05, **P ≤ .01, and ***P ≤ .001, n = 18–24). Significant migratory effects of CD, CL, IGFBP-7, and HGF were also detected (Figure 4B; **P ≤ .01 and ***P ≤ .001, n = 15–19). No increase in HOMEC migration by PlGF-1 was observed.

Figure 4.

Individual proteins identified in the EOC cell secretome display mitogenic, migratory, and angiogenic activities in HOMECs. (A) HGF, PlGF-1, and IGFBP-7 increase HOMEC proliferation. HOMECs were treated with HGF (25 ng/ml), PlGF-1 (25 ng/ml), or IGFBP-7 (25 ng/ml) for 72 hours, and proliferation was assessed by the WST-1 assay. Controls received medium alone, *P ≤ .05, **P ≤ .01, and ***P ≤ .001 versus control (100%). (B) HOMEC migration induced by ovarian cancer-secreted proteins. HOMECs were treated with different proteins for 48 hours (CD, 80 ng/ml; CL, 40 ng/ml; MMP9, 40 ng/ml; IGFBP-7, 20 ng/ml; HGF, 25 ng/ml; and PlGF-1, 25 ng/ml), and migration was assessed using the Oris migration assay. Controls received medium alone, **P ≤ .01 and ***P ≤ .001 versus control (100%). (C) HGF, CL, and IGFBP-7 induced formation of tube-like structures in HOMECs. HOMECs were seeded onto fibrin matrices in 24-well plates, incubated overnight, and then treated with HGF (50 ng/ml), CL (40 ng/ml), and IGFBP-7 (50 ng/ml). After a maximum of 72 hours, tubule structures were quantified and expressed as a percentage of control, ***P ≤ .001 versus control (100%). For A to C, data are presented as means ± SD. P values were calculated by one-way ANOVA, followed by Tukey's post hoc test. (D) Representative images of the tube-like structure formation assessed in C. HOMECs are labeled with calcein AM for quantitation; original magnification, x100.

CL, HGF, and IGFBP-7 Induced Tube-Like Structure Formation in HOMECs

Next, we examined whether the proteins secreted by EOC cells were able to induce tube-like structure formation in HOMEC. Figure 4 indicates that CL, HGF, and IGFBP-7 significantly induced formation of tube-like structures. CL and HGF induced extensive lacunae formation with cells reorganizing themselves in characteristic elongated tube-like structures (Figure 4, C and D; ***P ≤ .001, n = 12). IGFBP-7 also displayed angiogenic activity on HOMECs, however, to a lesser extent (Figure 4, C and D; ***P ≤ .001, n = 12). This lack of well-defined tube-like structure formation may suggest that proangiogenic signaling of IGFBP-7 could be involved in the later stages of angiogenesis. These data suggest that CL, HGF, and IGBP-7 trigger a proangiogenic phenotype in HOMECs and that this could occur independently of VEGFA, because none of them have been reported to directly bind to VEGFRs.

HOMEC c-Met Inhibition Reduces HGF-Mediated Mitogenic and Angiogenic Effects

HGF signals through its tyrosine kinase receptor c-Met. Because HGF was secreted from EOC cells and influenced HOMEC migration, proliferation, and tube-like structure formation, we investigated c-Met expression on HOMECs. Figure 5A shows that HOMECs highly express c-Met along with the endothelial marker CD31. Functional experiments indicated that an ATP-competitive inhibitor of c-Met kinase, PF04217903, abolished the significant increase in HOMEC proliferation induced by both HGF and CM from A2780 ovarian cancer cells (Figure 4B; ***P ≤ .001 and ^###P ≤ .001, n = 24). Published studies in preclinical models indicate that activation of alternative angiogenic pathways during anti-VEGFA therapy may potentially lead to unresponsiveness of the tumors to therapy [17]. Therefore, we further investigated the role of the HGF pathway by examining whether HGF- and A2780-stimulated HOMEC tube structure formation could be inhibited by PF04217903. We chose to study the A2780 EOC cell line because it secretes HGF and VEGF to a similar extent and, also, does not express c-Met [18], ensuring our research design would not influence the cancer cells. The data presented in Figure 5 show that both HGF- and A2780-mediated in vitro angiogenic changes of HOMECs (number of tubules, total tubule length, and number of junctions) were reduced to control levels by PF04217903 (Figure 5, C–E; ***^/###P ≤ .001, **P ≤ .01, and *^/#P ≤ .05, n = 10). Furthermore, lack of antiangiogenic effect of SU5416 in co-culture with A2780 cells was confirmed.

Figure 5.

Inhibition of c-Met in HOMECs abolishes HGF- and EOC-mediated mitogenic and angiogenic responses. (A) HOMECs express the HGF receptor c-Met. HOMECs were co-stained for CD-31 (red) and c-Met (green). Nuclei were stained with DAPI (blue). (a) Membranous expression of endothelial marker CD31, which coincided with diffused membrane expression of c-Met. (b) Negative control without primary antibody. Scale bar, 40 µm. (B) Proliferation of HOMECs (48 hours) treated with either HGF or CM from A2780 cells ± the c-Met inhibitor (PF04217903) assessed using the WST-1 assay. Controls received basal medium only. ^###P ≤ .001 versus control levels (100%) and ***P ≤ .001 versus HGF or A2780. (C–E) Inhibition of HGF- and EOC-induced HOMEC tube formation (using basement membrane extract) by the c-Met kinase inhibitor (PF04217903; 20 nM). Quantification of number of tubules (C), total tubule length (D), and number of junctions (E) of HOMEC was performed by AngioSys software. ^###P ≤ .001 and ^#P ≤ .05 versus control levels (100%), ***P ≤ .001, **P ≤ .01, and *P ≤ .05 versus A2780 or HGF (50 ng/ml). For A to D, data are presented as means ± SD. P values were calculated by one-way ANOVA, followed by Tukey's post hoc test.

Discussion

Despite improvements in clinical management, and substantial investment in clinical trials, the prognosis for patients with ovarian cancer remains poor. The urgent search for effective treatment has led to trials of antiangiogenic therapy, with the VEGF-VEGFR axis offering an attractive target, e.g., VEGF traps, a soluble receptor decoy, small tyrosine kinase inhibitors of VEGFRs, and also antibodies (anti-VEGFA or anti-VEGFRs), particularly bevacizumab [4,19,20]. Despite the potential benefits of antiangiogenic therapy in EOC trials, results have not been encouraging, with a range of problems encountered including 1) small benefits in progression-free survival, 2) numerous adverse effects, and 3) lack of significant impact on overall survival [20,21].

It is now well recognized that metastasized primary tumor cells secrete factors that alter their microenvironment and initiate angiogenic responses in the microvessel endothelium of the host organ, i.e., proliferation, migration, and ultimately angiogenesis. However, the array of tumor factors secreted, and EC signaling pathways activated, to achieve these changes is likely to be highly specific to the tumor type and the target endothelium, and further complexity might be achieved by administration of antiangiogenic therapy. In this study, we specifically examined the influence of EOC-secreted proteins on phenotypic changes in the omental microvasculature using an in vitro model of HOMECs and EOC cells. Initially, we demonstrated that the total secretome of EOC increased HOMEC proliferation, migration, and tube-like structure formation. However, investigation of the role of the VEGF-VEGFR axis showed that although inhibition of the VEGFR1/2 tyrosine kinases (SU5416) abolished VEGFA₁₆₅-mediated HOMEC migration and tube-like structure formation, EOC cell-induced migration and tube-like structure formation were refractory to such inhibition. Similarly in the co-culture migration assay EOC-induced HOMEC migration was not inhibited by the depletion of VEGFA ligand (anti-VEGFA antibody). These data strongly imply that EOC cells secrete factors other than VEGFA that are able to contribute to the induction of a proangiogenic phenotype in HOMECs during suppression of VEGFA-VEGFR signaling. This perhaps could provide an explanation for the disappointing outcomes of VEGFA-targeted therapies in EOC. Although a discussion of EC heterogeneity is not within the remit of this report, it is important to note that when we carried out identical studies using HUVECs, VEGFR1/2 inhibition with SU5416 significantly inhibited EOC-induced tubule formation (data not shown), highlighting the importance of studying disease-relevant in vitro model systems.

In light of our initial findings, we investigated, using various proteomic approaches, other potential activators secreted by EOC cells that could induce proangiogenic responses in HOMEC. Interestingly, the secretomes of A2780 and SKOV3 differed qualitatively and quantitatively. This was reflected in their ability to induce HOMEC proliferation, tube-like structure formation, and migration. This may be explained by the presence of CD133⁺ cells in the A2870 cell population. CD133⁺ ovarian cells were reported recently and termed cancer-initiating cells [22]. Of 40 screened ovarian cancer cell lines, 15 were positive for CD133 expression, with A2780 having strong heterogeneous positivity and SKOV3 being negative. The CD133⁺ A2780 cancer cells exhibited enhanced aggressiveness, closely resembling metastatic ascites-derived primary ovarian cancer cells, compared to CD133^- A2780 cells. It is now recognized that cancer-initiating cells produce larger quantities of proangiogenic factors (reviewed in [23]); therefore, the observed effect of A2780 cells may arise from the presence of CD133⁺ A2780 cells in this cancer cell population.

A range of mediators implicated in angiogenesis and antiangiogenic factors was identified in the secretome of EOC cells. These fell into several distinct groups: first, regulators of ECM, including proteases such as MMP8 and MMP9, CD and CL, and uPA, and protease inhibitors such as TIMP-1 and PAI-1. Because the process of angiogenesis is characterized by temporal regulation of the equilibrium between proangiogenic and antiangiogenic factors, the presence of antiangiogenic TIMP-1 and PEDF in the EOC CM is not surprising. It has been reported that the proteases CL and CD have proangiogenic roles as upstream stimulators of MMP secretion/activation and endothelial migration [24–26]. However, the mechanism involved in CL-mediated regulation of HOMEC migration and tube-like structure formation needs further investigation. The presence of MMPs in the EOC secretome agrees with the general concept of MMP involvement in cancer-mediated angiogenesis. Specifically, tumor-secreted MMP9, but not membrane-anchored MMP9, has been implicated in MMP9-induced tumor angiogenesis, associating with increased VEGFA/VEGFR2 expression [27].

The second major group of secreted factors identified consisted of chemokines and GFs, i.e., HGF, VEGFA, PlGF-1 and CXC-motif, and CC-motif chemokines. We hypothesized that ovarian cancer cell-derived HGF may be a surrogate proangiogenic factor since not only has it recently been reported that EC HGF/c-Met activation promotes tumor angiogenesis in sunitinib-resistant cancers [28] but also HGF has been implicated in activation of the uPA receptor (uPAR) pathway. Activation of the uPAR pathway (uPA-uPAR) has been reported to be a critical step in VEGFA-mediated angiogenesis in other systems [29]. Because HOMEC tube-like structure formation and migration can be induced in the presence of VEGFR inhibition or VEGFA depletion, it is possible that HGF acts as an alternative activator of the uPAR pathway. Indeed, HGF has been shown to stimulate the expression of uPA and uPAR in ECs, promoting their migration and invasiveness [30]. Paradoxically, PAI-1, a serine protease inhibitor of uPA, which we detected in EOC CM and which controls uPA-uPAR-plasmin fibrinolysis, has long been implicated in tumor progression and angiogenesis. Despite illusive mechanisms explaining this phenomenon, a conceptual link suggests that PAI-1 protects ECs from plasmin-induced FasL-mediated apoptosis [31] and may regulate cellular motility in the surrounding matrices [32]. Several reports have demonstrated a synergistic effect of VEGFA₁₆₅ and either PlGF-1 or HGF on EC survival, chemotaxis, and angiogenic response [33,34]. In our studies, HGF displayed mitotic, chemotactic, and proangiogenic activity on HOMECs, whereas PlGF-1 only marginally increased HOMEC viability. Furthermore, we demonstrated that HOMECs strongly express the HGF receptor c-Met and that c-Met inhibition by PF04217903 is more effective than standard anti-VEGFA approaches in inhibiting the induction of HOMEC angiogenic phenotypes. These data again suggest that HGF could contribute to the EOC-induced proangiogenic HOMEC responses observed during suppression of the VEGFA-VEGFR axis. Interestingly, the VEGFA₁₆₅ and HGF pathways converge downstream, activating the same mitogen-activated protein kinases (MAPKs) and regulating components of cytoskeletal signaling in HUVECs [35]. The putative link between the pathways may be the scaffolding adaptor proteins Gab1/Gab2, which, upon autophosphorylation, bind to various Src homology 2 domain-containing transducers, such as protein tyrosine phosphatase 2, phosphatidylinositol 3-kinases and phospholipase C-γ [36]. Signal transduction through Gab1/Gab2 is essential in VEGFR2-mediated MAPK activation, e.g., during angiogenesis, and has recently also been reported to be downstream of c-Met activation in ECs, again promoting EC migration and in vitro angiogenic phenotype [37,38].

The observation that chemokines were secreted by EOC contributes to the growing literature highlighting their role in the proangiogenic tumor microenvironment. Two of the chemokines identified in the EOC secretome, PF4 and CXCL16, have been reported to have proangiogenic effects. For instance, tumor-secreted PF4 has been shown to chemoattract monocytes, neutrophils, activated T cells, natural killer cells, and fibroblasts [39–41]. Additionally, CXCL16 is strongly mitotic for fibroblasts [42] and mediates proliferation and migration of CXCR6-expressing T cells [43]. Interestingly, it has been reported that tumor-associated fibroblasts are able to promote tumor angiogenesis during inhibition of VEGFA signaling [44].

The third group of factors identified includes the IGFBPs. We particularly focused on IGFBP-7, which was present in the secretome of ovarian cancer cells and enhanced HOMEC proliferation, migration, and tube-like structure formation. Compelling evidence has described a strong up-regulation of IGFBP-7 in tumor microvasculature and in vitro capillary tube-like structures [45]. Indeed, IGFBP-7 has recently been implicated in vascular remodeling through modulation of VEGF bioavailability and, consequently, vascular patterning [46]. Additionally, IGFBP-7 induced tube-like structure formation in brain ECs, through the TGF-1β/ALK5/Smad-2 pathway that is known to be involved in the regulation of late-stage angiogenesis [47]. This suggests that cancer-derived IGFBP-7 may contribute toward the formation of functional nonpermeable tumor vasculature. Another member of the IGFBP family identified in the ovarian cancer secretome, IGFBP-2, has been shown to be associated with cancer-mediated endothelial recruitment through IGFBP-2/IGF-1/IGF receptor 1 and growth arrest-specific 6/c-mer proto-oncogene tyrosine kinase (GAS6/MERTK) signaling pathways [48].

To conclude, we have shown that EOC cells secrete an array of angiogenesis-associated proteins, some of which have profound proangiogenic effects on HOMECs in our in vitro model. Our data support the hypothesis that EOC-induced omental angiogenesis during transcoelomic EOC metastasis is driven by the interplay of a range of proangiogenic factors and that selective disruption of one pathway, such as during anti-VEGF angiogenic therapy, simply allows the compensatory action of other mediators, e.g., HGF in concert with supportive factors such as cathepsins, chemokines, and IGFBPs (overview in Figure 6). We believe that our findings may explain the poor efficacy of anti-VEGFA therapies in ovarian cancer and critically highlight the need to direct antiangiogenic therapies to a broader spectrum of targets than VEGFA in patients with EOC.

Figure 6.

Potential proangiogenic mechanisms activated in the omentum during transcoelomic EOC metastasis in the presence of anti-VEGF therapy. Malignant ovarian cancer cells create aggregates as a result of an adaptive metastatic process. Once omental implantation is established, the micrometastasis starts to secrete proinvasive factors. This leads to a “switching” of the omental microvasculature toward a proangiogenic phenotype. Secreted proteases and protease inhibitors [MMPs, CD and CL, urokinase (uPA), and PAI-1] interplay to remodel the ECM allowing the omental ECs to migrate, whereas cancer-derived VEGF and HGF promote vascularization. In the presence of anti-VEGF therapy (selective inhibition of the VEGFA proangiogenic pathway; represented by SU5416 = Semaxanib), c-Met signaling may be favored, contributing to the omental angioarchitecture. IGFBP-7 and IGFBP-2 intensify the EC angiogenic response, potentially facilitating the formation of functional tumor vasculature and EC recruitment, respectively. Furthermore, cancer-secreted chemokines (PF4 and CXCL16) fuel protumorigenic inflammatory responses in the malignant omentum caused by the migration of distinct proangiogenic cell populations. Therefore, the presence of anti-VEGF therapy may amplify the impact of alternative proangiogenic transducers, e.g., c-Met in concert with supportive factors such as cathepsins, MMPs, chemokines, and IGFBPs, possibly explaining the poor efficacy of anti-VEGFA therapies in ovarian cancer patients with transcoelomic omental metastasis.

Supplementary Materials and Methods

Reagents and Equipment

The following reagents and equipment were used: ELISA-grade BSA, human thrombin, EC growth supplement from bovine neural tissue, CD, CL, MMP9, thrombin, and human plasma (Cat. No. T7009) from Sigma; DMEM-BS from Gibco (Paisley, United Kingdom); WST-1 from Roche (Burgess Hill, United Kingdom); PF04217903 and PlGF-1 from R&D Systems; VEGFA₁₆₅, Proteome Profiler human angiogenesis array, CD DuoSet ELISA, and mouse anti-human monoclonal IGFBP-7 antibody (clone 192520; capture antibody); streptavidin-HRP conjugate and ELISA substrate reagent; HDFs, fibroblast growth medium 2, MV2 endothelial medium kit, and HGF from PromoCell/PromoKine (Heidelberg, Germany); calcein AM from eBioscience (Hatfield, United Kingdom); MaxiSorp microplates (Nunc); IGFBP-7, rabbit polyclonal anti-human IGFBP-7 (biotinylated, detection antibody), and CD ELISA from Abcam (Cambridge, United Kingdom); SKOV3 and A2780 cell lines from the European Collection of Cell Cultures; ThinCerts (8.0-µm pore diameter) and 24-well plates (Cat. No. 662638) from Greiner Bio One (Stonehouse, United Kingdom); Oris Cell migration assay (Cat. No. cma1.101) from Platypus Technologies (Fitchburg, WI); fibrinogen, human plasma (Cat. No. 341578), and Amicon centrifugal filters (Cat. No. UFC800324) from Merck Millipore; anti-VEGFA antibody (Cat. No. 07-1419; recognizes isoforms 121, 165, and 189 of VEGFA); VectaSpin Micro (Cat. No. 6835-3001) from GE Healthcare/Whatman.

Preparation and Collection of Tumor and Endothelial CM

For collection of tumor CM, SKOV3 (TCM) or A2780 (TCMA) cells at 80% to 90% confluence were washed with PBS (x3) and incubated with chemically defined DMEM-BS medium or MV2 basal medium (1 ml of medium per 9 cm² of growth area) for 4 or 24 hours at 37°C. Media were then centrifuged (600g, 10 minutes, 4°C) and stored at -80°C for further experiments.

For EC CM, subconfluent cultures of HOMECs were incubated overnight in GF-deprived MV2 medium, washed with PBS (x3), and incubated in fresh MV2 basal medium for 24 hours (1 ml of medium per 9 cm² of growth area). Some EC CM were supplemented, before use, with CD (80 ng/ml), CL (40 ng/ml), VEGF₁₆₅ (20 ng/ml), and HGF, IGFBP-7, and PlGF-1 at a concentration of 25 ng/ml. EC CM was processed as for TCM.

Proteome Profiling (MS)

Samples (affinity purified on StrataClean resin) were digested overnight at 30°C with 20 µl of trypsin (20 µg/ml; Promega) and then separated by liquid chromatography on an Ultimate 3000 Dionex system. Each sample was first loaded onto a C18 trapping column (10 µl) and then eluted onto a C18 analytical column (PepMap). The column was washed for 5 minutes, and the peptides were then eluted using a 40-minute MeCN gradient (2%–50% MeCN/H₂0) with further elution for 10 minutes at 90% MeCN. Samples were spotted at 10-second intervals using a Probot with α-cyano-4-hydroxycinnamic acid at 3 mg/ml [70% (wt/vol) MeCN and 0.1% (vol/vol) trifluoroacetic acid] at a flow rate of 1.2 µl/min. A 4800 MALDI TOF/TOF analyzer was used for MS and MS/MS spectra generation. MS settings are given as follows: mass range set to 700 to 4000 Da; focus mass of 25,000 Da; 50 shots per subspectrum and 1000 total shots per spectrum (no stop conditions and every subspectrum was accepted); laser intensity set to 3950. MS/MS settings are given as follows: 50 shots per subspectrum and 4000 shots per spectrum; stop conditions: when accumulated spectrum reaches estimated signal-to-noise ratio (S/N) of 30; minimum number of peaks above S/N; threshold 10 and subspectra accepted before S/N test stops 15; every subspectrum was accepted; laser intensity set to 3950. Maximum of 12 precursors per spot was chosen for MS/MS (minimum S/N filter 30; weakest precursors acquired first). GPS explorer software (Applied Biosystems, Paisley, United Kingdom) was used to search the NCBI database using Mascot server. Significant results were determined by selecting proteins that matched with two or more peptides with total ion confidence intervals (CIs) of >95%.

Proliferation, Migration, and Fibrin Matrix/Geltrex Tube Formation Assays

The parameters used in cell-based assays are shown in Table W1.

Enzyme-Linked Immunosorbent Assays

The parameters used in the final protocol of the in-house developed IGFBP-7 ELISA are shown in Table W2.

Immunocytochemistry

The antibodies and parameters of immunocytochemistry (ICC) are shown in Table W3.

Supplementary Results

The FPLC standard curve and the Proteome Profiler (human angiogenesis kit) analysis of FPLC fractions 11 and 12 of SKOV3 and A2780 CM are shown in Figure W1. The list of proteins identified by MS in whole CM from SKOV3 ovarian cancer cell line is shown in Table W4.