Platelet Adhesion and Degranulation Induce Pro-Survival and Pro-Angiogenic Signalling in Ovarian Cancer Cells

Thrombosis is common in ovarian cancer. However, the interaction of platelets with ovarian cancer cells has not been critically examined. To address this, we investigated platelet interactions in a range of ovarian cancer cell lines with different metastatic potentials [HIO-80, 59M, SK-OV-3, A2780, A2780cis]. Platelets adhered to ovarian cancer cells with the most significant adhesion to the 59M cell line. Ovarian cancer cells induced platelet activation [P-selectin expression] in a dose dependent manner, with the most significant activation seen in response to the 59M cell line. The platelet antagonists [cangrelor, MRS2179, and apyrase] inhibited 59M cell induced activation suggesting a P2Y12 and P2Y1 receptor mediated mechanism of platelet activation dependent on the release of ADP by 59M cells. A2780 and 59M cells potentiated PAR-1, PAR-4, and TxA2 receptor mediated platelet activation, but had no effect on ADP, epinephrine, or collagen induced activation. Analysis of gene expression changes in ovarian cancer cells following treatment with washed platelets or platelet releasate showed a subtle but valid upregulation of anti-apoptotic, anti-autophagy pro-angiogenic, pro-cell cycle and metabolic genes. Thus, ovarian cancer cells with different metastatic potential adhere and activate platelets differentially while both platelets and platelet releasate mediate pro-survival and pro-angiogenic signals in ovarian cancer cells.


Introduction
Ovarian cancer is the fifth leading cause of cancer related deaths in women [1]. It is the most common gynaecologic malignancy and has the highest fatality to case ratio of all gynaecologic malignancies. The poor survival rate is the result of late stage diagnoses. Most patients are asymptomatic until the disease has metastasised [2]. Spread of ovarian cancer has been considered to occur primarily in the peritoneum [3]. However, autopsy and imaging studies [4], as well as evidence for the presence of micrometastases in the bone aspirates of early stage ovarian cancer patients [5] suggest that hematogenous metastasis is more common than previously thought.
During hematogenous dissemination, the ability of circulating tumour cells to interact with platelets is believed to promote their survival within the circulation and therefore facilitate metastasis. Pre-clinical animal experiments have demonstrated that pharmacologically [6] or genetically [7] induced thrombocytopenia, as well as defects in platelet function [7][8][9][10] are associated with reduced metastasis. The interaction of cancer cells with platelets is believed to confer a number of advantages that promote successful metastasis, including protection from immunological assault and evasion of immune surveillance [11,12], the release of growth, angiogenic, and vascular permeability factors during activation and degranulation [13].
Thrombosis and thrombocytosis are frequent complications of ovarian cancer and are associated with poor prognosis [14][15][16], highlighting the importance of platelets in the pathology of ovarian cancer. However, the interaction between platelets and ovarian cancer cells has not been well studied. In this study, we aimed to characterise the interaction of platelets with ovarian cells, using a normal ovarian cell line  and ovarian cancer cells lines with different biological properties and metastatic potentials [59M, SK-OV-3, A2780, and A2780cis].
Firstly, we studied platelet adhesion to ovarian cancer cells under static conditions to determine if an adhesive interaction between platelets and ovarian cancer cells exists. Secondly, we assessed the ability of ovarian cancer cells to induce platelet activation and degranulation [P-selectin expression]. After establishing that platelets adhere to ovarian cancer cells, and ovarian cancer cells are capable of inducing platelet activation and degranulation, we next assessed gene expression changes at the transcriptome level in ovarian cancer cells treated with platelets or platelet releasate. Our results show differential interactions between platelets and ovarian cancer cell lines, not only in terms of platelet adhesion and activation, but also in gene expression changes in cancer cells treated with washed platelets or platelet releasate. Multiple interactions occur between platelets and ovarian cancer cells involving factors released by platelets and cancer cells, as well as direct platelet-ovarian cell interactions. This interaction results in a pro-survival, pro-angiogenic signal for the ovarian cancer cell.

Ethics statement
Blood collection for this study was approved by the Royal College of Surgeons in Ireland ethics committee and written informed consent was obtained from all donors prior to phlebotomy.

Cell lines
A selection of ovarian cell lines of epithelial origin were chosen for inclusion in this study as epithelial ovarian cancers are the most common histological type. HIO-80 [a gift from the Fox Chase Cancer Center, Philadelphia, PA] represents a non-tumorigenic normal human ovarian epithelial cell line, which has been immortalised by transfection with a plasmid encoding for the SV40 large T gene. The HIO-80 cells were maintained in a 1:1 mixture of medium 199 and MCDB-105 supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ml penicillin, 100 ug/ml streptomycin and 0.2 IU/ml of recombinant human insulin as recommended by Fox Chase. 59M [European collection of cell cultures (ECACC), Salisbury, UK] represents a tumorigenic human ovarian epithelial carcinoma of endometrioid type with clear cell components, originally isolated from ascites of a 65 year old patient with metastatic ovarian cancer. 59M cells were maintained in DMEM, 1 g glucose/L, supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 ug/ml streptomycin as recommended by ECACC. SK-OV-3 [American Type Culture Collection (ATCC), Manassas, VA, USA] represents a tumorigenic human ovarian epithelial adenocarcinoma, originally isolated from ascites of a 64 year old patient with metastatic ovarian cancer. SK-OV-3 was cultured with McCoy's 5A media, supplemented with 10% fetal bovine serum [Invitrogen, The Netherlands], 100 units/ml penicillin and 100 ug/ml streptomycin as recommended by ATCC. A2780 [ECACC, Salisbury, UK] is derived from an ovarian serous epithelial tumour tissue in an untreated patient. The cisplatinresistant cell line A2780cis was developed by chronic exposure of the parent cisplatin-sensitive A2780 cell line to increasing concentrations of cisplatin. Both cell lines were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ml penicillin and 100 ug/ml streptomycin as recommended by ECACC. All cell lines were grown in standard conditions in a humidified atmosphere containing 5% CO 2 at 37uC.

Platelet preparation
Blood was obtained from healthy donors who had not taken medications known to affect platelet function for at least 10 days. Blood was collected by venipuncture through a 19-gauge butterfly needle without a tourniquet, to avoid platelet activation. Platelets were prepared as previously described [17]. In brief, for the preparation of platelet-rich plasma [PRP], blood was collected into a syringe containing 3.2% trisodium citrate as anticoagulant [10% vol/vol, then centrifuged at 170 g for 10 minutes at room temperature. For the preparation of washed platelets, blood was collected into Acid-Citrate-Dextrose [ACD: 38 mM citric acid, 75 mM sodium citrate, 124 mM D-glucose] as anticoagulant [15% vol/vol]. Blood was centrifuged at 170 g for 10 minutes at room temperature. PRP was acidified to pH 6.5 with ACD, and PGE 1 [1 mM] was added to avoid platelet activation during centrifugation. Platelets were pelleted by centrifugation at 720 g for 10 minutes. The supernatant was removed and the platelet pellet was resuspended in JNL buffer [130 mM NaCl, 10 mM sodium citrate, 9 mM NaHCO 3 , 6 mM D-glucose, and 0.9 mM MgCl 2 , 0.81 mM KH 2 PO 4 , and 10 mM Tris, pH 7.4] to a concentration of 2.5610 8 /mL and supplemented with 1.8 mM CaCl 2 .

Platelet adhesion assay
The platelet adhesion assay was performed as previously described [18]. In brief, ovarian cells were seeded in a 96 well plate [Nunc, Thermo Fisher Scientific, Braunschweig, Germany] at a concentration of 5610 4 /ml and grown until confluent.

Platelet activation
The ability of ovarian cancer cells to induce platelet activation (P-selectin expression) was assessed by a flow cytometry based assay modified from Nylander et al [19]. In a total reaction volume of 100 ml, 10 ml of PRP was incubated with ovarian cancer cells

MTT assay
To optimise the concentration of platelet releasate for tumour cell exposure, a series of MTT cell proliferation assays [Roche Diagnostics Ltd, United Kingdom] were performed according to the manufacturer's instructions to determine the maximum concentration that could be applied to cells without negatively impacting on cell growth or survival. Cells were seeded into 96well cell culture plates at 2610 4 cells/well and cultured for 24 hours to allow cell attachment. Once attached to the plates, growth media was aspirated and cells were briefly washed with 200 ml of pre warmed PBS. Cells were then exposed to serial dilutions of platelet releasate from 1:10 to 1:10,000 in full media, serum free media or JNL buffer to allow optimisation of concentration to be used in subsequent gene expression based studies.

Washed platelet/platelet releasate exposure for gene expression analysis
The optimal concentration of platelet releasate [determined by MTT] was applied to the cell lines and the levels of apoptosis were compared against cells grown in full media only using a Roche Apodirect TUNEL/Propidium iodide kit. As the 1:1000 dilution of platelet releasate in full growth media was observed to be the highest concentration of platelet releasate in full media that did not impact on cell growth/viability for all cell lines and was subsequently shown to result in no significant difference in apoptosis compared to cells grown in full media alone it was determined that it was suitable to proceed with this concentration. Given that the platelet releasate was prepared from identical washed platelet preparations, the 1:1000 dilution of platelet releasate directly informs the use of a 1:1000 dilution of washed platelets for comparative study. Optimal concentrations of platelet releasate and washed platelets [1:1000] were suspended in full culture media and applied to cultured cells in 75 cm 2 flasks in triplicate. Cells were exposed to releasate or washed platelets for 6 hours after which total RNA was extracted from cells. Transcriptome analysis was performed using Affymetrix Human Exon Arrays.

RNA extraction
Cells were washed briefly in PBS, trypsinised and centrifuged to remove supernatant. RNA was extracted using RNeasy mini kit [Qiagen Ltd., West Sussex, UK] according to the manufacturer's protocol. RNA quantity was assessed using a Nano-Drop ND-1000 Spectrophotometer [Wilmington, USA] and quality by an Agilent Bioanalyser 2100 and RNA 6000 Nano microfluidic chip assay [Santa Clara, USA]. RNA was stored at 280uC.

Affymetrix array
100 ng of total RNA extracted from control cells and cells exposed to washed platelets/platelet releasate was labelled using the Ambion WT Expression kit [Ambion/Life Technologies, Austin, TX, USA] including the labelling controls from the Affymetrix Gene Chip Poly-A RNA Control Kit. As suggested by Affymetrix/Ambion, at each step of the sample preparation protocol, progress was monitored using both the Agilent 2100 Bioanalyzer and the Nanodrop spectrophotometer. Quality control [QC] required assessment after the first cycle RNA cleanup, after the second cycle single-strand cDNA cleanup and following ssDNA fragmentation. Prepared fragmented ssDNA was hybridised to the Affymetrix Human Exon 1.0 ST Array at 45uC for 16 hours following Affymetrix protocols for their GeneChip WT Terminal Labeling, GeneChip Hybridization Control and GeneChip Hybridization, Wash, and Stain kits [Affymetrix, Santa Clara, USA]. Following hybridization, the chips were washed and stained using the Affymetrix GeneChip Fluidics Station with appropriate 64 format assay protocol. Following staining and washing steps the Affymetrix GeneChip Scanner 3000 and Affymetrix GeneChip Operating Software was used for the management and initial processing of the expression data. The data from 45 exon arrays was subject to array quality control performed using the Affymetrix Expression Console. All controls were within the parameters suggested by Affymetrix. Following successful quality control standards assessment, the chip data was then analysed in depth using Biotique Systems XRAY analysis software. Each cell line was examined under three different conditions; resting cells, cells exposed to platelet releasate, and cells exposed to washed platelets. Each cell line and condition was assayed in triplicate. The software was used to compare the two exposure cohorts against the resting control and genes exhibiting a positive or negative fold change of greater than 1.5 and a significance of p#0.05 examined.

Fluidigm validation
Gene expression was validated using Fluidigm's high throughput qPCR 48.48 dynamic array system. Genes displaying the most significant fold changes in addition to genes involved in biologically relevant pathways were selected for validation. TaqManH real-time PCR expression assays were used in conjunction with Fluidigm's microfluidic Biomark system. Samples were analysed in triplicate and results were compared with Affymetrix expression data. RNA

Epithelial mesenchymal transition [EMT] expression analysis in platelet cloaked/releasate treated cells
The process of EMT is critical in the metastatic cascade, by facilitating entry of cells into the vascular system. RNA was reverse transcribed to single stranded cDNA using a High Capacity cDNA RT Kit [Applied Biosystems, CA, USA] in 100 ml reactions as above. A combination of the primary players in the EMT process [20] and additional genes identified in our laboratory as integral to the EMT process [data not shown] were examined as part of the profiling for EMT in ovarian cancer cells:

Adhesion of platelets to ovarian cancer cells under static conditions
Platelet adhesion to ovarian cancer cell lines under static conditions was quantified based on the fluorescent detection of labelled platelets [ Figure 1A].  Figure 1B and C]. Consistent with the platelet adhesion assay, figure S1 demonstrates the minimal platelet adhesion to the HIO-80 cell line.

Ovarian cancer cells induce platelet activation in a dose dependent manner
The ability of ovarian cancer cells to induce platelet activation was quantified by flow cytometry. Increasing concentrations of ovarian cancer cells [0 -1.5610 6 cells/ml] were added to PRP and platelet activation was assessed based on P-selectin expression. Ovarian cancer cells induced a dose-dependent increase in platelet activation [ Figure 2]. The two metastatic ovarian cancer cell lines; SK-OV-3 [1.5610 6 cells/ml, 47610.2% of platelets P-selectin positivity, n = 3] and 59M [1.5610 6 cells/ml, 51618% P-selectin positivity n = 6] induced the most significant platelet activation.

Ovarian cancer cells potentiate TRAP [thrombin receptor activating peptide], PAR 4 agonist, and arachidonic acid induced platelet activation in a dose dependent manner
Since thrombosis is a complex process that involves multiple agonists in vivo, we next asked if ovarian cancer cells modulated agonist [TRAP, PAR 4 agonist, Arachidonic Acid, ADP, epinephrine, and Collagen] induced platelet activation. To  Other platelet antagonists such as hirudin, EDTA, abciximab, RGDS, and aspirin had no effect on 59M cell induced platelet activation. doi:10.1371/journal.pone.0026125.g003 determine this, we assessed activation in platelets co-incubated with low ovarian cancer cell concentrations and low agonist concentrations. Platelet agonist concentrations that gave #20% platelet activation [P-selectin expression] were used. At low cellular concentrations [0.5-1610 5

Expression array analysis of ovarian cancer cells treated with washed platelets or platelet releasate using Affymetrix exon arrays
All arrays passed QC using Affymetrix QC software. Biotiques X-Ray software plug-in for Microsoft Excel was used to interrogate gene expression changes between treatments and cell types [Fold change .1.5 and p,0.05]. Analytical stringency was relaxed to fold changes of .1.5 to accommodate subtle but meaningful biological variation [p,0.05] that was contingent upon treatment. Tables 1, 2

Validation of gene expression signature by the Fluidigm 48648 dynamic array
Gene expression array results were validated using Fluidigm high throughput qPCR technology. Twenty five genes were selected along with two endogenous controls for analysis using Fluidigm's 48648 Dynamic Array. The genes selected and the cell lines and conditions in which significant alteration of expression were observed by Affymetrix analyses are highlighted in bold in Tables 1, 2  close correlation between Affymetrix and Fluidigm data sets for those genes observed to have undergone altered expression in the 59M, SK-OV-3 and A2780cis cell lines [ Figure 5A, B, C, D]. Discrepancies between expression calculated by Affymetrix and TaqMan analyses are likely due to normalisation and summarisation methods. The expression score for each probeset for the arrays is derived using the standard Robust Multiarray Averaging [RMA] normalisation and summarisation methods. The RMA algorithm was used because compared to other approaches, it increases sensitivity to small changes between experiment and control samples and minimses variance across dymanic range, but in doing so it does compress calculated fold change values. This means generally the fold change value obtained with the affmetrixy array is smaller than that obtained with the TaqMan assay. The correlation for genes observed to undergo significant alteration of expression in both the normal [immortalised] HI0-80 and malignant 59M ovarian carcinoma cell lines was high [.0.8] and was statistically significant [p,0.05]. Figure 6 summarises the findings of the gene expression analysis.
Expression analysis of EMT associated genes in ovarian cancer cells exposed to platelet releasate EMT profiling was performed to see if interaction with platelets would induce EMT. Expression analysis of EMT associated genes demonstrated constitutive expression of the majority of EMT associated genes [ Figure 7]. Importantly, platelet releasate did not induce/suppress EMT.

Discussion
The results of the present investigation demonstrate a differential interaction between platelets and ovarian cancer cell lines, not only in terms of the effect of ovarian cancer cells on platelets, but also in the effect of platelets on ovarian cancer cells. Firstly, we used a panel of ovarian cell lines to assess platelet adhesion under static conditions. Adhesion across the 5 cells lines was extremely heterogeneous with significant platelet adhesion to the 59M, A2780, and A2780cis cell lines, while adhesion to the HI0-80 and SK-OV-3 cell lines was not significant compared to the BSA negative control [ Figure 1]. The adhesion of platelets to all ovarian cancer cell lines was significantly less than to fibrinogen, suggesting that the adhesion is mediated by either a lower copy number receptor-ligand interaction, or a lower affinity interaction than platelet integrin aIIbb3 mediated adhesion to fibrinogen. The molecular mechanisms mediating platelet adhesion to ovarian cancer cells remains to be elucidated. Secondly, we assessed the ability of ovarian cancer cells to induce platelet activation. We found that ovarian cancer cells induce platelet activation and degranulation in a dose dependent manner with the most significant platelet activation seen in response to the 59M and SK-OV-3 ovarian cancer cell lines [ Figure 2].
Using 59M cells we then investigated the effect of different platelet inhibitors to determine the mechanism of ovarian cancer cell induced platelet activation. Following treatment with cangrelor [P2Y12 antagonist], MRS2179 [P2Y1 antagonist], or apyrase [ADP/ATPase], platelet activation in the presence of 59M ovarian cancer cells was greatly diminished, suggesting a P2Y12/P2Y1 dependent mechanism of activation, mediated by the release of ADP by 59M cells, as the 59M cell supernatant induced comparable platelet activation [ Figure 3]. Consistent with the literature, other studies have also demonstrated ADP dependent platelet aggregation induced by cancer cells [21]. Of note, Uluckan et al have shown that treatment with APT102 [an ADPase] and aspirin in combination, inhibited B16 melanoma cell induced aggregation, and decreased metastasis in a murine B16 melanoma model of bone metastasis [22]. Italiano et al have previously shown that platelets differentially package angiogenic regulatory proteins in separate alpha granules that are subject to differential release depending on the platelet agonist [23]. Interestingly, Bambace et al have demonstrated that ADP induced platelet activation causes the release of VEGF [pro-angiogenic] but not endostatin [anti-angiogenic] in vitro [24], indicating a potential mechanism by which ovarian cancer cells could induce the release of pro-angiogenic but not anti-angiogenic factors by platelets in vivo.
Ovarian cancer cells alone induce activation of platelets. However, thrombosis is a complex process that involves many different agonists in vivo. Therefore, we examined how ovarian cancer cells modulate agonist induced platelet activation. At low cellular concentrations, both A2780 and 59M cells potentiate TRAP, PAR-4 agonist, and arachidonic acid induced platelet activation, but not ADP, epinephrine, or collagen induced platelet activation [ Figure 4]. This suggests a synergistic relationship between PAR-1/PAR-4/TxA2 receptor mediated activation and the mode of ovarian cancer cell induced activation. Thrombin is the physiological ligand for the PAR-1 and PAR-4 receptors, and its generation is critical for coagulation. The expression of thrombin has been identified in ovarian cancer tissue [25] and is known to potentiate invasion in an in vitro model of ovarian cancer [26]. Holmes et al have also shown that TRAP treated platelets increase ovarian cancer cell invasion [27]. Prostaglandin synthesis is also increased in ovarian cancer, with a trend towards higher fold increases in pro-aggregatory TxA2 levels compared to antiaggregatory PGI 2 levels [28].
Having established that ovarian cancer cells interact with platelets, we next assessed the effect of platelet adhesion and platelet granule release on ovarian cancer cells. Overall analysis of gene expression changes in ovarian cancer cells following treatment with washed platelets or platelet releasate showed an upregulation of anti-apoptotic, anti-autophagy, pro-angiogenic, pro-cell cycle and metabolic genes in the treated ovarian cancer cells [Tables 1, 2, 3, 4]. However, ovarian cancer cells of variable phenotype showed differential gene expression profiles, possibly related to their underlying biology.
In HIO-80 cells, there was significant upregulation of genes encoding for proteins associated with ovarian cancer metastasis [SERPINB2/PAI2], metabolic activities [IDI1, PMM2] and gene expression/transcription [PCGF6, ZNF267] [29]. LRP8 [ApoER2] which showed significantly increased expression in response to platelet releasate exposure was also observed to be up regulated in response to washed platelet exposure and its activation has been observed to alter migratory capability in a non-tumorigenic breast epithelial [MCF 10A] cell model [30]. It also has a key role in mediating increased platelet activation and adhesion in association with other glycoproteins or clotting factors [31][32][33][34] and may promote signalling in cells via ApoER2. HIO-80 are a non-tumorigenic normal human ovarian surface epithelial cell line, which has been immortalised by transfection with a plasmid encoding for the SV40 large T gene. SERPINB2/PAI2 was the most significantly upregulated gene on validation and this upregulation may be explained by the fact that platelets can release TNFa [35] which in turn has been shown to induce matrix   proteolytic enzyme production and basement remodelling by human ovarian surface epithelial cells providing a molecular mechanism linking ovulation and ovarian cancer risk [36] 59M cells exhibited aberrant upregulation of genes involved in anti-autophagy, anti-apoptotic and pro-angiogenic signalling [TRAF2, CCL2, TNFAIP2] [37][38][39]. Among the other pathways showing altered expression were pro-angiogenic signalling through increased PDGFb transcription in response to washed platelet exposure with further pro-proliferative [ [40][41][42][43][44].
SK-OV-3 cells exhibited significant overexpression in the Cardiolipin Synthase [CRLS1] gene, which is responsible for cardiolipin [CL] production. Anti-cardiolipin antibodies are associated with both solid and non-solid tumours and are associated with increased thrombocytosis [45]. This gene product [CRLS1] may also have an anti-apoptotic role as reduced CL expression is associated with increased apoptosis [46]. Genes associated with increased cell invasion/motility in breast and pancreatic cancers [ANXA2] were also dysregulated. Following treatment with platelet releasate, SK-OV-3 cells also showed expression changes in genes involved in, anti-autophagy in endometrial and ovarian cancer [KIAA1324/EIG121], as well as transcriptional regulation [ACTL6A], cell cycle[MUS81], cytoskeletal [TPM4] and homeostatic [TPM4] processes [47,48].
The A2780 cell line and its cisplatin resistant daughter cell line A2780cis displayed different responses to the exposure treatments. The A2780 cell line did not show significant alteration of gene expression following treatment with either platelet releasate or washed platelets. Conversely, the A2780cis cell line revealed a panel of dysregulated genes following treatment with washed platelet but none following treatment with platelet releasate.  [49,50].
TaqMan expression analysis of EMT associated genes demonstrated constitutive expression of the majority of EMT associated genes [ Figure 7]. It appears that EMT related genes are effectively 'primed' and that interaction with platelets does not significantly alter the capacity of cancer cells to mount this type of response. Other studies have reported similar findings in breast cancer [51,52].
In summary, our data shows for the first time that there is a potent dynamic interaction between ovarian cancer cells and platelets in vitro. This interaction involves platelet adhesion, platelet activation and degranulation and a resultant pro-survival and proangiogenic signal for the ovarian cancer cell that could potentially promote ovarian cancer cell metastasis. Further work is required to determine the significance of this interaction in vivo.