Background
The Eph receptors are the largest family of receptor tyrosine kinases (RTKs) in humans with 14 members. Eph receptors are activated by cell–bound ephrin ligands, and the Eph-ephrin system governs contact-dependent intercellular communication controlling a wide array of biological processes such as development, tissue organization and cell migration [
1,
2]. EphA2 of the A type Eph subclass is expressed at low levels in differentiated tissues but expression frequently increases in advanced cancers, implicating EphA2 in tumor progression [
3]. The preferred ligand for EphA2 is ephrin-A1 [
4], and ligation of EphA2 by ephrin-A1 leads to the formation of multimeric receptor-ligand clusters that activate a signaling response that controls cytoskeletal dynamics and cell morphology. While ligand-dependent EphA2-activation has been considered tumor suppressive, recent reports have highlighted a role for EphA2-ephrin-A1 signaling in tumor cell plasticity and a shift from mesenchymal to amoeboid morphology [
5,
6] and increased single cell invasion [
7]. In addition, oncogenic EphA2 signaling has been proposed to be ligand-independent, drawing from the observations of decreased expression of the ephrin-A1 ligand paralleling increased EphA2 expression in human cancers [
8]. Miao et al. showed that EphA2 is a substrate and effector of PI3 kinase/Akt signaling through phosphorylation of serine 897 in the EphA2 cytoplasmic domain, a pathway by which EphA2 controls cancer cell motility and invasion independently of ephrin-A1 [
9,
10].
Tissue Factor (TF) is the receptor and co-factor for coagulation factor VII/VIIa (FVII/FVIIa), a circulating serine protease. The proteolytic TF/FVIIa complex functions as the physiological trigger of blood coagulation and in addition activates cell signaling through mechanisms dependent or independent of protease-activated receptors (PARs) and the TF cytoplasmic domain [
11]. TF expression is found in tumor cells [
12], and in preclinical models, TF/FVIIa signaling has been implicated in tumor progression through effects on processes such as cell migration and angiogenesis [
13,
14]. Furthermore, a clinically relevant role of the coagulation system in malignancies is evidenced by the increased risk of thrombosis in cancer patients. In contrast, anticoagulant treatment only modestly influences cancer incidence and survival in humans, and the effect seem to differ between cancer types [
15].
We previously reported on a direct cleavage by TF/FVIIa in the ligand binding domains (LBD) of the Eph receptors EphB2 and EphA2. We also identified a conserved disulfide bond that kept the N-terminal fragment tethered to the receptors after cleavage [
16]. In this study we set out to further explore how TF/FVIIa influences EphA2 signaling and activity. We report herein that TF and EphA2 co-localizes in MDA-MB-231 breast cancer cells with constitutive high TF expression and in TF transfected U251 glioblastoma cells, and that FVIIa sensitizes MDA-MB-231 cells to ephrin-A1-mediated cytoskeletal reorganization and cell rounding independently of PAR2-activation through a RhoA/ROCK pathway. EphA2 and TF were co-expressed in a cohort of human colorectal cancer specimens, providing evidence that the prerequisites for TF–EphA2 cross-talk in vivo are present.
Methods
Reagents
Antibodies towards EphA2 (6997), pS897-EphA2 (6347), pY588-EphA2 (12677) and GAPDH (2118) were from Cell Signaling Technology. The RhoA antibody (ARH04) was from Cytoskeleton. The TF antibody (clone 10H10) was a kind gift from Professor J Morrissey (University of Illinois) and the PAR2 blocking antibody was a kind gift from Professor W Ruf (Scripps Institute). PI3 kinase inhibitor LY294002 was from Calbiochem and ROCK inhibitor Y-27632 from Sigma.
Cell culture
MDA-MB-231 cells were obtained from the American Type Culture Collection and cultured in complete RPMI 1640 medium. For experiments, cells were seeded in individual wells and left to attach over night. Cells were then switched to medium containing 0.1 % FBS, for the FVIIa groups supplemented with 10 nM FVIIa for 1 h and then stimulated with ephrin-A1 or Fc control as indicated in the figure legends. Prior to experiments, ephrin-A1 was preclustered for 1 h at room temperature with anti-human Fc goat IgG (Jackson ImmunoResearch) at 1:10 concentration. Preclustered Fc fragments (Jackson Immunoresearch) were used as controls. In some cases, cells were pretreated for 30 min with inhibitors prior to stimulations.
U251 cells were from Cell Line Services and were cultured in complete DMEM medium. Experiments were performed in DMEM with 0.1 % FBS.
SDS-PAGE and Western blot
SDS-PAGE and Western blot was performed using the Novex Bis-Tris gel system (Life Technologies) as previously described [
14].
mRNA analyses
Total RNA was extracted from cells by Trizol® (Life Technologies) using standard protocols, and converted to cDNA using oligoDT primers. Quantitative real-time PCR (qPCR) was performed using Assays on demand (Applied Biosystems) for IL8 with β2-microglobulin as housekeeping gene on an ABI prism 7500 system. Results were calculated using the comparative CT method for separate tubes.
siRNA knock-down of RhoA
MDA-MB-231 cells were transfected with 10 nM RhoA siRNA (Silencer select, Ambion) using Lipofectamine RNAiMAX (Life Technologies) according to instructions from the manufacturer. Cells were assayed 48 h after transfection. Efficiency of protein knock-down was analyzed by Western blot.
In situ proximity ligation assay
The assay was performed with reagents supplied by the manufacturer (Olink Bioscience) and according to instructions supplied with the kit. In brief, MDA-MB-231 cells were grown on chamber slides (Lab Tek), fixed in 4 % PFA/PBS, blocked and incubated with antibodies towards TF and EphA2, which were bound by secondary antibodies connected to specific oligonucelotides that ligate when they are in close proximity. Ligated oligonucleotides then serve as template for a rolling-circle amplification and the amplification products were visualized by fluorescently labeled probes. Cell nuclei were stained with DAPI and images captured with a 40x objective using a Zeiss Axioimager fluorescence microscope.
Confocal microscopy
For microscopy experiments, cells were grown on 8-well chamber slides (Lab Tek), and then washed with PBS and fixed in 4 % PFA. Cells were permeabilized with 0.2 % Triton X-100 and subsequently blocked in 2 % BSA for 30 min. Antibody incubations were performed for 1 h at room temperature for primary antibodies (EphA2 1:200, TF 1:500) and 30 min in the dark at room temperature for secondary antibodies (1:1000) (Molecular Probes). Staining of the actin cytoskeleton with phalloidin-FITC (Sigma) diluted 1:500 was performed together with the secondary antibodies. Slides were washed with PBS between all steps, and mounted in Vectashield mounting medium with DAPI and sealed with nail varnish. Confocal images were captured with a Zeiss LSM710 confocal microscope using the 40x or 63x objectives.
Assay for cell rounding and retraction fiber formation
MDA-MB-231 cells were seeded on 8-well chamber slides (Lab-Tek) coated with 10 μg/ml collagen IV (Sigma) and left to attach over night. After treatments, cells were washed with PBS and fixed in 4 % PFA/PBS. The cytoskeleton was stained by FITC-conjugated phalloidin (Sigma) for 30 min at room temperature protected from light. Slides were then washed, mounted in mounting media with DAPI and sealed with nail varnish. Images were captured using an Axiovert 40 CFL inverted epifluorescence microscope (Zeiss) and the 40x objective. 3–4 images per well were taken at random locations and the percentage of cells with rounded morphology and retraction fibers was quantified according to Taddei et al. [
17].
TF overexpression
U251 cells were transfected to transiently overexpress TF by using Lipofectamine 3000 (Life Technologies) and a plasmid encoding untagged human TF (Origene). Briefly, cells were seeded in 24-well plates, left to attach over night and transfected with 400 ng DNA. Lipofectamine without DNA was used as control, and stimulations were performed 24 h post transfection in low serum DMEM.
Immunohistochemistry (IHC) of colorectal cancer specimens
The patient cohort used in the study contained non-consecutive cases diagnosed with colorectal cancer between 1990 and 2003 in the Uppsala region in Sweden, collected with the purpose to screen for protein expression differences between disease stages and between normal colorectal tissue, cancer and metastases. The cohort included 60 patients, with 20 (33 %) patients each in stages I, II and III. 41 (68 %) cases were colon cancers and 19 (32 %) cases were rectal cancers. 10 cases each of colorectal normal tissues and adenomas as well as 20 cases of lymph node or distant metastases were also included. Tumor and patient characteristics were based on the original histopathology reports and the patients’ clinical records.
All tumor material was present as formalin-fixed paraffin-embedded tissue in duplicate cores on tissue microarrays (TMAs), constructed at the SciLife Laboratory Tissue Profiling Facility at Uppsala University as described previously [
18]. Expression of TF and EphA2 was detected by IHC using a rabbit polyclonal TF antibody developed by the Human Protein Atlas project [
19] and a rabbit monoclonal EphA2 antibody (6997, Cell Signaling Technology). Automated IHC staining were performed as previously described [
20], using a LabVision Autostainer 480S (ThermoFisher Scientific). Microarray sections were baked over night, deparaffinized, hydrated in graded alcohols and blocked for endogenous peroxidase using 0.3 % hydrogen peroxide. Following antigen retrieval, sections were stained with primary antibody (30 min) and secondary dextran polymer visualization system (30 min), followed by the addition of diaminobenzidine as chromogen. All solutions except primary antibodies were obtained from Laboratory Vision (Laboratory Vision). Sections were counterstained in Mayer’s hematoxylin (Histolab), before dehydration and mounting of coverslip. Stained slides were digitalized by scanning, using an Aperio ScanScope XT Slide Scanner (Aperio Technologies). Tumor cell staining was annotated semi-quantitatively with respect to staining intensity and fraction of positive cells. Intensity was graded as negative, weak, moderate or strong, and graded in six fractions (0–1 %; 2–10 %; 11–25 %; 26–50 %, 51–75 %; >75 %) and specimens with moderate or strong staining in more than 2 % of tumor cells were considered positive.
During TMA sectioning 6 primary cancer cases and 3 metastases were lost or compromised, resulting in TMA cores with no representative tumor tissue present. These cases were not annotated and excluded from all statistical analyses.
Statistics
Statistics were performed using the GraphPad Prism (GraphPad Software) and Statistica (Statsoft Scandinavia) softwares. Unpaired two-tailed t-test was used to compare experimental groups. For comparison of TF and EphA2 expression in colorectal cancer non-parametric correlation was calculated according to Spearman, and Chi-squared tests were used to compare groups defined by TF and/or EphA2. P-values equal to or below 0.05 were considered statistically significant.
Discussion
We report herein on a close cross-talk between TF and the tyrosine kinase receptor EphA2 and present evidence of a role for the TF/FVIIa complex as a co-receptor and signaling partner of EphA2 with possible implications in human cancer. We observed that TF and EphA2 co-localized in MDA-MB-231 breast cancer cells with high endogenous TF expression, and in U251 glioblastoma cells with forced overexpression of TF. EphA2 and TF appeared to cluster at cell-cell contacts and subcellular compartments with an accumulation of dynamic actin cytoskeleton, in agreement with literature documenting an important role for EphA2 in regulating cytoskeletal dynamics [
29,
30]. Importantly, we found that FVIIa potentiated the cellular response to ephrin-A1 as measured by increased cell rounding and retraction fiber formation upon stimulation, demonstrating that FVIIa and ephrin-A1 act synergistically to enhance ligand-dependent EphA2 signaling. By antibody blocking experiments, we show that this is an event uncoupled from PAR2-activation, in line with biochemical data demonstrating direct cleavage of EphA2 by TF/FVIIa, and supporting a role of the TF/FVIIa complex acting as a co-receptor in EphA2 signaling.
EphA2 is cleaved by FVIIa after a conserved arginine residue in the J-K loop of the LBD, and we previously showed that the cleaved fragment remains associated to the truncated EphA2 by a conserved disulfide bond (Cys70-Cys188), and the LBD is also stabilized by an additional disulfide (Cys105-Cys115). Since the Cys70-Cys188 disulfide will prevent dissociation of the N-terminal fragment we predict that the structure of the EphA2 LBD is largely retained after cleavage, with the cleavage leading to a local conformational change in the J-K loop. We hypothesize that cleavage by TF/FVIIa might, by a yet unidentified exact mechanism, enhance EphA2 activation by its ligand. As it was tyrosine phosphorylated and rapidly underwent ligand-induced degradation, our data indicate that the cleaved fragment indeed contributes to ephrin-A1-dependent signaling and that the cleavage does not results in a ligand-unresponsive form of EphA2. Of note, as the synergism between FVIIa and ephrin-A1 was PAR2-independent in line with the cleavage mechanism, it appears not to be an unrelated event resulting from PAR2 activation by TF/FVIIa.
EphA2 tyrosine phosphorylation was very low in unstimulated cells, which was expected since MDA-MB-231 cells are reported to express very low amounts of the ephrin-A1 ligand [
8]. We observed a slight increase of phosphorylation at the Y588 site by FVIIa, but since this effect was negligible compared to the response induced by ephrin-A1 the relevance of this observation with regards to synergism between FVIIa and ephrin-A1 is not clear at the moment. Of note, we previously analyzed this cell line using antibody arrays [
16], where FVIIa decreased the phosphotyrosine signal for EphA2 and as we noted, since those experiments were run with native samples a decrease in signal can equally well correspond to masking of the phosphotyrosine epitope by proteins recruited to the activated receptor [
31].
Even though a large number of studies have confirmed an important role of EphA2 in human cancers, there are controversies regarding the contributions of ligand dependent and ligand independent signaling. Ephrin-A1 ligation of EphA2 and subsequent receptor activation has been proposed to be tumor suppressive, although other studies point to an important role for kinase dependent EphA2 activity in cancer development. Cell rounding and retraction is a recognized event in Eph-ephrin signaling, and requires RhoA activation [
6,
7], in accordance with our results. Interestingly, recent work suggest that EphA2-ephrin-A1 signaling may support cell migration and invasion in prostate cancer [
5] and melanoma cells [
6] in this manner. Here, RhoA-mediated mesenchymal-to-amoeboid transition has been implicated, indicating a qualitative shift in cell migration characterized by cell rounding, single cell invasion and a squeezing movement style, leading to tissue invasion and tumor dissemination [
32,
33]. Also, other studies demonstrated that repulsive EphA2-RhoA signaling among cancer cells may serve to allow tumor cells to disperse from the main tumor mass [
26]. Importantly, contact inhibition of locomotion which may occur as a consequence of repulsive interactions between Eph and ephrin expressing cells does not need to lead to a complete inhibition of motility but rather a change in direction, in certain contexts contributing to a shift from collective to single cell cancer invasion [
34]. In this context, our results demonstrating a synergism of FVIIa and ephrin-A1 in EphA2 activation lead us to hypothesize that TF/FVIIa may serve to increase EphA2-mediated cell dispersion upon contact with ephrin-A1 expressing cells, and that FVIIa may facilitate ephrin-A1 induced mesenchymal-to-amoeboid transition in this context.
EphA2 signaling is characterized by a high level of complexity and several seemingly independent activation mechanisms, as demonstrated by the present study where FVIIa caused both a direct cleavage and S897 phosphorylation of EphA2. While cleavage was insensitive to PI3 kinase inhibition, S897 phosphorylation was completely abolished in agreement with the notion that this is a common event downstream of activation of PI3 kinase. Cell rounding and retraction fiber formation was not affected by PI3K inhibition, showing a dissociation between EphA2-pS897 signaling and ligand-dependent EphA2 activation as has been proposed elsewhere [
9]. The role and function of FVIIa-induced EphA2 S897-phosphorylation was not elucidated in this study, and its role as an effector downstream of FVIIa-induced PI3K activation will be an important issue for further studies.
In support of the relevance of TF/FVIIa-EphA2 cross-talk we present descriptive data that TF and EphA2 were co-expressed to a high extent in a human cancer material, and TF positivity was with a few exceptions only found in EphA2 expressing tumors. TF and EphA2 have separately been reported to be expressed in various cancers, but our study is the first to demonstrate co-expression in the same tumor specimens. In contrast to our experimental work that was performed in breast cancer and glioblastoma cell lines, we used another cancer type, colorectal cancer, to demonstrate the association between TF and EphA2 in human tumor tissue. Although these are distinct cancer types, both TF and EphA2 expression appears to foremost be a feature of advanced cancers, and TF expression is regulated by common oncogenic mutations [
35]. Indeed, EphA2 has been linked to colorectal cancer development as targeted EphA2 gene disruption was found to reduce intestinal tumorigenesis in the APC/Min + mouse model [
36]. We found that co-expressing tumors were enriched among high-grade cases, supporting the view that these results may be a feature of poorly differentiated human cancers rather than a tissue specific event in colorectal neoplasms. TF expression was, for both primary cancers and metastases, found in a minority of cases, and our results indicate that TF and EphA2 co-expression is an event that occurs in a subset of colorectal cancers that are characterized by low differentiation. As a large number of cases were TF negative, this might be a mechanism that contributes to tumorigenesis in a subset of tumors, whereas other oncogenic pathways drive more differentiated cancers. Here, future studies on larger clinical materials will have to provide answers on the possible impact on patient survival and outcome.
We previously noted that higher concentrations of FVIIa were needed for cleavage of EphA2 compared to cleavage of another Eph receptor, EphB2. In the present study, overexpression of TF was required to achieve EphA2 cleavage by FVIIa in U251 cells and it thus appears that very high levels of TF, such as may be found in transformed cells are necessary for this to occur. However, high levels of both TF and EphA2 are recognized events in malignant cells, and we used a colorectal cancer material to show that co-expression of TF and EphA2 indeed occurs in vivo. In addition, cleavage of cellular subpopulations with EphA2 co-localizing with TF may also play important biological roles even though these are events that are below the detection limit of Western blot assays.
Acknowledgements
The authors would like to acknowledge researchers in the Human Protein Atlas Project, Uppsala University, for skillful technical assistance.