Platelets enhance tissue factor protein and metastasis initiating cell markers, and act as chemoattractants increasing the migration of ovarian cancer cells

Ovarian cancer is the deadliest gynecological cancer, causing around of 15,000 deaths in the USA this year [1]. Despite clinical advances, ovarian cancer continues to be a poorly understood disease with an unfavorable prognosis, principally due to late stage diagnosis when the cancer has already disseminated throughout the peritoneum [2]. The current theory of cancer progression proposes that from the total population of cancer cells within a primary tumor, only a small sub-population has the capacity to migrate, survive in isolation and establish secondary tumors within distant organs [3]. Metastasis Initiating Cells (MICs) are characterized by their enhanced chemoresistance, low metabolic rate, possessing “stemness”, for having undergone Epithelial-Mesenchymal Transition (EMT) and their enhanced capacity to generate metastatic foci. Cancer cell stemness (also known as cancer stem cells) refers to the presence of proteins on the cancer cells that are normally associated with physiological stem cells and whose presence associate with enhanced tumor-forming capacity in xenograft assays and increased chemoresistance [4] . These markers are numerous and differ depending on the tissue origin, however for ovarian cancer, one of the most reported examples is increased CD44 expression [5,6]. EMT, the conversion of epithelial cells to a more stromal or mesenchymal phenotype, is believed to be a fundamental event by which the cancer cell acquires migratory and invasive properties [7,8]. This event has been associated with lower E-Cadherin expression and enhanced N-Cadherin and vimentin levels [9]. The transcription factors Twist, Snail and Slug are known to promote EMT and cause the down regulation of E-Cadherin [10-12] As an in vitro model of MICs, under special culture conditions a heterogeneous population of cells can give rise to three-dimensional cancer spheres (cell clusters) which present enhanced expression of CD44 and display the capability of anchorage-independent growth [13].

Cancer patients have long been reported to present abnormal risk of thrombosis that correlates with the progression of the disease [14-16]. High platelet counts are prevalent in 31-42% of primary epithelial ovarian cancers and this correlated with significantly worse prognosis. [17-19].

It has been speculated that platelets may contribute to tumor metastasis through EMT induction [20], immune system evasion, adhesion to endothelial layer, and angiogenesis and vascular remodeling [21].

A further protein associated with cancer metastasis is Tissue factor (TF). This transmembrane receptor and initiator of the extrinsic coagulation pathway is not normally expressed in the vascular lumen, but makes contact with the circulatory system only upon vascular injury, resulting in clotting activation [22]. It is widely reported that many cancer types overexpress functional TF on the cell membranes and also in tumor derived microparticles, thus being responsible for enhanced coagulation and invasion [23-26].

Taken together, accumulating evidence is correlating platelet function with increased metastasis and poorer patient survival. However, to date, the effect of platelets on TF levels have not been described, neither are the levels of this protein during the acquisition of a MIC phenotype. Considering that ovarian cancer metastasis occurs mainly within the peritoneal cavity as a result of the accumulation of cancer cells in the ascites, the goal of the present work is to evaluate the effect of platelet interaction with ovarian cancer cells, regarding phenotype, TF and EMT associated protein levels and biological function. Herein, we present evidence that platelet addition brings about an increase in TF protein, a switch to a MIC phenotype and enhanced migration of ovarian cancer cells.

Immunofluorescence using a platelet specific CD42b antibody (that reacts with platelet GPIb) demonstrated that platelets are present together with cancer cells in peritoneal fluid extracted from an ovarian cancer patient (Figure 1). The merge with vimentin antibody and Hoechst stain shows that the platelets associate with cancer cells (see also immunofluorescence negative control in Additional file 1: Figure S1). Furthermore, flow cytometer analysis of ascites extracted from individual patients demonstrated the presence of free platelets in this fluid (see Additional file 2: Figure S2).

All patient information is confidential and thus coded for scientific use. We assigned patient numbers (e.g. patient 5) to cancer cell populations isolated from individual patients to allow comparison of results between the different assays used in the paper. Due to limited cell numbers, not all experiments could be performed for each patient. After confirming that platelets are present together with cancer cells, we moved to an in vitro model to determine the effect of platelets on primary cultures of advanced ovarian cancer cells. The co-culture of ascites from 5 separate advance ovarian cancer patients for 12 hours with 150,000 platelets/uL (physiological concentration) resulted in a marked change in cancer cell phenotype, with the occurrence of cell elongation and a more stromal appearance (Figure 2). It was noted that the majority of platelets become activated upon contact with cell culture medium (please refer to Additional file 3: Figure S3). The change in phenotype was present in each patient culture tested, however different degrees of phenotypic change were noted (compare patient 2 and 4 with patient cultures 3, 5 and 6) (Figure 2). This was also reflected in ovarian cancer cell lines treated in the same manner, where SKOV3 had subtle changes, whereas UCI101 showed a rapid and notorious phenotypic alteration.

Given that the fibroblastic phenotype has been correlated with migratory potential, we tested if platelet presence enhanced cell motility. The scratch assay disclosed an enhanced ovarian cancer cell migration at 12 and 24 hours post-addition of platelets (Figure 3). Cell cycle analysis by flow cytometry demonstrated that the platelet-mediated increase in migration was not a result of increased cancer cell proliferation (please refer to Additional file 4: Figure S4). Interestingly, during the course of these migration assays we visually observed that the cancer cells (especially in the UCI101 cell line) appeared to be attracted to platelet aggregates that had formed in the culture plate. This is shown at different magnifications (Figure 3E), where the ovarian cancer cells nearest to the platelet aggregate are more elongated. To test the hypothesis that platelets have chemoattractant capacity, we performed migration assays using Boyden chambers. In this experiment ovarian cancer cells SKOV3 (Figure 4A), UCI101 (Figure 4B) or primary cultured ascites cells (Patient 5, Figure 4C), were placed with platelets in either the upper chamber or in the lower chamber below the polycarbonate membrane and not in contact with the cancer cells. As shown in each of the three experiments, an increase in migration to the lower side of the polycarbonate membrane was only observed when the platelets were used as a chemoattractant, independent of platelet-induced phenotype changes (Figure 2).

To analyze further the phenotype change we examined at the transcript and protein level genes known to be associated with a MIC and EMT phenotypes. In both cell lines tested, platelet co-culture for 6 hours increased CD44 and N-Cadherin mRNA levels. E-Cadherin was barely detectable in UCI101, but was decreased in SKOV3 (Figure 5A and B). This mRNA change was also reflected at the protein level, as demonstrated by representative blots (Figure 5C), and in its quantification (Figure 5D). A platelet-induced increase in the EMT markers Snail and Twist was also observed in SKOV3 cells and primary cultured cancer cells (please refer to Additional file 5: Figure S5).

Culture of the ovarian cancer cell line UCI101 in basal medium was used to form cancer spheres in the presence and absence of platelets. Isolated spheres were selected on the basis of size and were confirmed to demonstrate increased expression of CD44 by immunocytochemistry and expression of EMT protein markers by Western blotting (please refer to Additional file 5: Figure S5). Spheres formed under both conditions; however their number was greatly enhanced in the presence of platelets (Figure 5E and F).

Platelet interaction with tumor cells increased TF mRNA at 6 hours (Figure 6A) and the corresponding protein levels at 12 and 24 hours (Figure 6B and C, for SKOV3 and UCI101, respectively). β-Actin expression in these conditions is also increased due to the presence of platelets (Figure 6B and C). To control for this we evaluated the nuclear protein Histone 3 which is not expressed by platelets (Figure 6B and C) as a loading control for the levels of cancer cell proteins. This increase in TF protein upon platelet interaction was also observed in cells isolated form spheres (Figure 6D). By flow cytometry we observed that in the whole cell UCI101 population 40% of the cells expressed TF. Of the population of UCI101 cells that formed spheres, only 8% expressed this protein, however this percentage increased to 22% in sphere cells that formed in the presence of platelets. The increase of TF protein in tumor cells was associated with an enhanced TF-mediated FXa generation, both in SKOV3 and UCI101 cell lines (Figure 6E). Despite platelets having been reported to express functional TF [31] our antibody did not detect TF protein in the platelet lysates (Figure 6B and C). Analyzing the ovarian cancer line A2780 that does not express TF, we observed no changes in procoagulant activity (Figure 6E) after incubation with platelets. The tissue factor pathway inhibitor (TFPI-1) inhibited the increase in coagulation in SKOV3 and UCI101 cells demonstrating that TF was responsible for the platelet-mediated increase in coagulation (Figure 6E).

To demonstrate that the changes in TF and MIC/EMT markers with platelets were not a cell line artifact, we obtained primary cultures of ovarian cancer cells from eight separate patients. After 6 hours of platelet addition to these cancer cells we observed an increase in TF and CD44, with a decrease in E-Cadherin in all patients. An increase in N-Cadherin was observed in 5 of the 8 patients (Figure 7A). Although in many primary cultures not enough cells were recovered for western blotting, we demonstrate that these changes at the level of mRNA were also reflected at the protein level (Figure 7B). Interestingly, primary culture of a benign ovarian fibrothecoma (Patient 7, Figure 7A), a benign ovarian mucinous cystadenoma (Patient 10, Figure 7A) and a primary peritoneal carcinomatosis (Patient 4, Figure 7A), demonstrated that platelet-induced regulation of E-Cadherin, TF and CD44 was not limited to only malignant ovarian cancers. Furthermore, in three of three patients analyzed we observed a platelet-mediated increase in the EMT-related proteins Twist and Snail (Figure 7B). Finally, to evaluate if the effects observed by platelets are exclusively due to the presence of platelet membrane or whether factors released from the platelets (or membrane associated factors) are required, we denatured the platelets by heating to 95°C before addition to the culture medium. As observed in Figure 8A, denatured platelets are incapable elevating CD44 and N-Cadherin, or decreasing E-Cadherin protein levels in ovarian cancer cells. The platelet-induced change in phenotype of ovarian cancer cells was also lost (Figure 8B).

Metastasis is not solely a function of the tumor cells. The cancer cell requires an intricate coordination with its immediate environment, the site of extravasation and the host tissue (reviewed in [32]). The observation that cancer cells come into contact with platelets has been already established. Confocal microscopy analysis of leiomyosarcoma and histiocytoma tumors has demonstrated the presence of platelets and activated platelets within the tumor mass [33]. In a mouse cancer model, platelets expressing yellow fluorescent protein (YFP) isolated from female transgenic C57BL/6 mice were transfused by tail-vein injections into ovarian tumor-bearing mice. The presence of extravascular YFP platelets was observed in both ascites and tumor specimens [17]. Furthermore, once the cancer cell leaves the solid tumor and enters the blood stream platelet-cancer cell interaction becomes inevitable. Preclinical data suggest that cancer cell-platelet interaction in the blood facilitates tumor metastasis. These platelets may form aggregates with circulating tumor cells and physically act as a shield for the host immune surveillance [34]. A report by Stone and colleagues (2012) demonstrated a median time to disease progression of 4.65 years in patients without thrombocytosis, as opposed to 2.62 years in patients presenting more than 450,000 platelets/uL [17].

The platelet concentration present in ascites is difficult to ascertain, as platelets entering the peritoneal cavity would be expected to adhere to collagen in the peritoneal wall. In fact, in mouse models it has been shown that circulation-derived platelets adhere to the peritoneum surface [17]. We also see that a high proportion of the platelets that enter the ascites are bound to cancer cells (Figure 1). For these reasons we utilized normal circulating concentrations of platelets in our studies. Furthermore, this concentration of platelets allows us to extrapolate our proposed model to cancer cells entering the blood stream.

It has been convincingly demonstrated that the platelet addition to tumor cells can induce an invasive mesenchymal-like phenotype and that platelets can prime the tumor cells for metastasis [20]. Herein, we take this story a step further, demonstrating that this also occurs in ovarian cancer cell lines and in primary cultured cancer cells extracted directly from patients. In accordance with previous reports in the literature, our platelets underwent activation upon contact with the cancer cell cultures and we also detected high levels of platelet activation in patient ascites (Additional file 2: Figure S2 and Additional file 3: Figure S3) [35]. In all patient-derived cancer cells (8 patients) we found a reduction in E-Cadherin and an increase in the stem cell marker CD44 and TF protein upon exposure to platelets. This loss in E-Cadherin may be due to the platelet-mediated increase in Twist and Snail observed in Figure 7 [10-12]. We are not aware of previous reports describing increases of CD44 and TF after platelet exposure in ovarian cancer. We have previously demonstrated in breast cancer cells that increased TF expression is associated with enhanced cancer cell migration [25]. CD44 and a stemness phenotype have also been reported by others to increase migration and invasion [36-40]. Accordingly, in this manuscript we demonstrated statistical differences in migration of ovarian cancer cells upon exposure to platelets (Figure 3). An increase in migration by platelets has been observed previously in pancreatic cancer cell lines and this effect was reduced by the antiplatelet cilostazol [41]. However, a strikingly novel observation of our study was the attraction of ovarian cancer cells to platelet aggregates. In Boyden chambers we showed that the presence of platelets below the tumor cells caused enhanced migration across the polycarbonate membrane, thus a physical movement of tumor cells towards the platelets. This suggests that along with a change in EMT markers, the observed change in phenotype may be a result of cells migrating towards platelets.

Interestingly, the ability of platelets to mediate cancer cell proliferation has been demonstrated in the literature, however the majority of our observations were performed at time points below the doubling time of these cancer cell lines [42]. We demonstrate at the 12 hours time point used in the migration assay that there was no change in cell cycle upon platelet addition (Additional file 4: Figure S4) and thus we speculate that platelets increase the migration of ovarian cancer cells through chemoattraction. Whether the expression or EMT and/or TF are involved in this chemoattraction is still unknown. Interestingly, the presence of the platelets above the cancer cells did not increase a downward migration, despite an increase in lateral migration being observed in the scratch assay. This may add evidence to our theory that the movement is towards the platelets (chemoattraction). In the scratch assay we did notice enhanced migration when a platelet aggregate was in the scratch area. The possibility that platelets can be chemoattractant to cancer cells in vivo raises several questions. Could platelet accumulation at sites of vascular injury attract or capture cancer cells in circulation? Once at this site, could the enhancement of EMT markers and TF promote extravasation? These questions may be also valid in the opposite direction; platelets at areas of vascular injury or poor vascular integrity (as often observed in tumors), could attract cancer cells to escape from the primary tumor and migrate towards the vasculature and thus become metastatic. These scenarios seem attractive in the setting of ovarian cancer where it is known that ascites derived platelets adhere to the peritoneal membrane, where the metastatic foci of ovarian cancer occurs.

Platelets contain many molecules with attractant properties, such as PDGF [43] or TGFβ [44], among others, that may play a role in this effect. Platelets have also been shown to liberate microparticles that could be playing a direct role upon the cancer cells [45]. Although we demonstrate that denatured platelets cannot bring about a change in phenotype and EMT marker protein expression, further experiments are required to elucidate the pathophysiological contribution of platelet-induced attraction in the development of metastasis. However, platelet-derived TGFβ1, together with other platelet-bound factors have been shown to cooperate to increase metastasis. Platelets can activate the NF-κB signaling pathway in a contact-dependent manner. Interestingly, it appears that platelet-induced effects on EMT and invasion are not mediated by secreted TGFβ alone, but require additional platelet-bound factors [20].

An approach to identify MICs in vitro is through the use of sphere-forming assays. Under special culture conditions, a subset of a cancer cell line or primary tumor population cells are able to form three-dimensional spheres, demonstrate enrichment in stem cell markers and display the capability of anchorage-independent growth. CD44-positive spheres isolated from ovarian serous adenocarcinomas by this technique have been shown to form tumors more efficiently in animal models [13]. Using this method we noted that the presence of platelets enhanced the number of spheres formed. In the whole cell population of the UCI101 ovarian cancer cell line, 40 % of these cells expressed TF. Only 8% for the spheres that formed expressed TF, however in the presence of platelets the fraction of TF-expressing spheres increased to 22%. The representation of a “metastasis initiating cell” with a rigid definition of protein markers is a concept that is mostly likely going to change. Of the numerous MICs that leave the primary tumor, most probably only a few of these are capable of forming secondary tumors. It is also likely that the MICs that survive to form metastatic foci change the expression of their cell surface proteins numerous times before the establishment of metastatic foci. For example, the presence of TF may or may not be necessary for the MIC to leave the primary tumor, but it may be necessary to cause localized coagulation to recruit more platelets to shield the cancer cell from the immune system [46]. Strengthening this theory, it has been reported that pre-operative serum TF levels of ovarian cancer patients are around 85.2 pg/mL, significantly higher than levels found in patients with low malignant potential tumors (12.8 pg/mL) [47]. A further role of TF, either alone or in its complex, maybe to act as a ligand or tether for endothelial-anchored TFPI-1. The interaction of TF-TFPI-1 may act to halt the cancer cell in circulation and favor extravasation [48]. Although the consequence of a platelet-induced gain of TF in MICs (in this case 8% to 22%) is unknown, it may contribute to the clinical observation correlating platelet levels to an increase in cancer metastasis [21].

In summary, we report for the first time that platelets are present together with cancer cells in peritoneal ascites from ovarian cancer patients. We show that platelets can increase the levels of functional TF expression and MIC markers in ovarian cancer cells. Finally we demonstrate that platelets can act as a chemoattractant to cancer cells. This data complements the existing concept that platelets may be an interesting target for future cancer therapies.