Protease-activated-receptor-2 affects protease-activated-receptor-1-driven breast cancer

Abstract

Mammalian protease-activated-receptor-1 and -2 (PAR₁ and PAR₂) are activated by proteases found in the flexible microenvironment of a tumor and play a central role in breast cancer. We propose in the present study that PAR₁ and PAR₂ act together as a functional unit during malignant and physiological invasion processes. This notion is supported by assessing pro-tumor functions in the presence of short hairpin; shRNA knocked-down hPar2 or by the use of a truncated PAR₂ devoid of the entire cytoplasmic tail. Silencing of hPar2 by shRNA-attenuated thrombin induced PAR₁ signaling as recapitulated by inhibiting the assembly of Etk/Bmx or Akt onto PAR₁-C-tail, by thrombin-instigated colony formation and invasion. Strikingly, shRNA-hPar2 also inhibited the TFLLRN selective PAR₁ pro-tumor functions. In addition, while evaluating the physiological invasion process of placenta extravillous trophoblast (EVT) organ culture, we observed inhibition of both thrombin or the selective PAR₁ ligand; TFLLRNPNDK induced EVT invasion by shRNA-hPar2 but not by scrambled shRNA-hPar2. In parallel, when a truncated PAR₂ was utilized in a xenograft mouse model, it inhibited PAR₁–PAR₂-driven tumor growth in vivo. Similarly, it also attenuated the interaction of Etk/Bmx with the PAR₁-C-tail in vitro and decreased markedly selective PAR₁-induced Matrigel invasion. Confocal images demonstrated co-localization of PAR₁ and PAR₂ in HEK293T cells over-expressing YFP-hPar2 and HA-hPar1. Co-immuno-precipitation analyses revealed PAR₁-PAR₂ complex formation but no PAR₁-CXCR4 complex was formed. Taken together, our observations show that PAR₁ and PAR₂ act as a functional unit in tumor development and placenta-uterus interactions. This conclusion may have significant consequences on future breast cancer therapeutic modalities and improved late pregnancy outcome.

Electronic supplementary material

The online version of this article (doi:10.1007/s00018-013-1498-7) contains supplementary material, which is available to authorized users.

Introduction

G protein-coupled receptors (GPCRs) represent the largest family of cell surface receptors [1, 2] and are divided into six subfamilies (classes A–F) based on their ligand interaction and signature residues [3, 4]. Many class A subfamily are major drug targets [5] among which are mammalian protease-activated receptors (PARs). There are four PAR family members (PAR1–4), all of which are proteolytically activated at the N-terminal extra-cellular domain generating a new N-terminal site that functions as a tethered ligand by binding intramolecularly to the second loop of a given PAR [6]. Class A receptors are characterized by several features among which are conserved motifs in the third trans-membrane domain (TM3), three cytoplasmic loops (C1, C2, and C3), and a C-terminal tail (C4). X-ray structure of rhodopsin, a prototype of class A receptor, revealed the presence of a highly conserved amphipathic 8th helix (H8) within the C-tail that is likely to be physically interposed between the receptor and G protein [7]. Indeed, the H8 helix loop was also described in other class A receptors including PAR₁ and PAR₂ [8] and is formed by anchoring to the membrane by palmitoylation of the C-terminal cysteine residues.

Our view regarding the regulation and signaling of the large GPCR family has greatly evolved over the last several years. It has become apparent that while GPCRs switch between active conformations (capable of activating G proteins), and inactive conformations (that do not activate G-proteins) [9, 10]; additional options of activation exist. Adding to the complexity of GPCR signaling, it is now clear that GPCRs are capable of adopting more than one active conformation, in which each active state is capable of stimulating an overlapping yet distinct group of downstream effectors [9]. These properties are recapitulated by the discovery of biased-agonism/antagonism or as more frequently referred “biased-ligands”. This term has become synonymous with additional terms to indicate the more likely selective activation of GPCR signaling pathway for functional selectivity and including stimulus trafficking [11]. It allows an agonist to bind preferentially and stabilize one active signaling state over another, thus endowing the term activation in a more profound meaning, illustrating in-depth the dynamic nature of GPCR-stabilized signal activation.

Overall, the elucidation in recent years of the crystal structures of several archetypal GPCRs, such as rhodopsin [7], chemokine receptors, as well as others (such as: β₂ and β₁-adrenergic receptors; [12, 13], A_2A adenosine receptor; [14, 15], dopamine D3 receptor; [16], and chemokine CXCR4 receptor; [17]) has provided new insights into the common features of general structures. It has become more evident that the significant differences and selectivity of receptors in fact relies on the extracellular/transmembrane ligand binding site determining a receptor with its unique features. In addition, the technology of high-resolution crystal structure of class A GPCRs demonstrated homodimer formation (as seen by the opioid receptors [18] and by CXCR4 [19]) and heterodimers. Evidence is accumulating showing that chemokine receptors, for example, can dimerize [19, 20]. CCR2 and CXCR4, which are more distantly related, form heterodimers as revealed by the observation that specific antagonists of one receptor inhibit the binding of chemokines to the other receptor [21]. The current emerging prospect is that dimers or oligomers of GPCRs are the functional units that interact with heterotrimeric G proteins and arrestins for signal generation and receptor desensitization. This form of heterodimer generation adds another layer of complexity to the regulation of GPCR signaling. GPCR heterodimerization affects receptor conformation and contributes to the activation of selective signaling, adding an extra level to the multifaceted action of GPCRs.

Members of the mammalian PAR family have been shown to form heterodimers. PAR₁/PAR₃ heterodimerization is required in endothelial cells for regulation of cell permeability in response to thrombin [22]. PAR₁ and PAR₄ also form stable heterodimeric complexes on the surface of platelets and PAR₁ appears to assist in the activation of PAR₄ in platelet-mediated thrombosis [23]. Recent studies suggest that PAR₁ and PAR₂, relatively distant receptors, exist in close proximity on the cell surface and can signal together during acute vascular inflammation [24]. The molecular basis for this interaction has been elucidated in endothelial cells, fibroblasts, and human embryonic kidney cells, whereby thrombin-cleaved PAR₁ donates its tethered ligand to transactivate PAR₂ [25, 26]. Although PAR₂ is not activated directly by thrombin, the thrombin-generated PAR₁-tethered ligand, SFLLRN, serves as an agonist for PAR₂ [26, 27]. Sevigny et al. [28] have demonstrated that knocking-out PAR₂ expression reduced neointimal vasculature hyperplasia induced by PAR₁. Recently, it was demonstrated that PAR₁ drives the trafficking behavior of PAR₂, but not vice versa. Furthermore, thrombin activation of PAR₁–PAR₂ heterodimer results in co-internalization and β-arrestin recruitment, an event that is not observed with the activated PAR₁ alone. Therefore, a biased and specific signaling response is demonstrated by PAR₁–PAR₂ inducing a distinct signaling β-arrestin recruitment [29].

Here, we demonstrate that PAR₁-induced tumor-promoting processes depend on the presence of a PAR₂. This was shown by either shRNA silencing of PAR₂ expression or by the use of a truncated PAR₂ construct. The importance of the PAR₂ C-tail was evaluated in co-IP assays between PAR₁ and signal proteins, colony formation in soft agar, migration and invasion, in vitro as well as tumor growth in a xenograft mouse model in vivo. Co-localization of PAR₁–PAR₂ was established by confocal immunofluorescence and co-IP between PAR₁ and PAR₂ but not CXCR4. It is well recognized in a physiological invasion process that placenta trophoblasts anchor to the uterus deciduas, in a similar manner as tumor cell invasion, providing an attractive model to study a highly regulated invasion scheme [30–32]. We demonstrate that shRNA-hPar2 silencing significantly inhibits thrombin-activated extravillous trophoblast (EVT) invasion in an organ culture system, recapitulating cytotrophoblast invasion in vivo. This was shown in the first-trimester trophoblasts that act in the establishment and bridging into the maternal tissues. Hence, PAR₂ is also required for EVT invasion. Altogether, silencing of hPar2 expression or the use of a truncated PAR₂ form modulates PAR₁-induced tumor function. Our findings can be partially explained by the formation of PAR₁–PAR₂ functional unit, in PAR-driven breast tumor.

Experimental procedures

Cells

Breast cancer cells; MCF7, MDA-MB-231, and MDA-MB-468 (ATCC VA, USA) and HEK293T cells were maintained in DMEM with 10 % fetal calf serum. HU breast epithelial cells were generated by the late Dr. Aviva Horowitz of the Sharett-Institute of Oncology, Hadassah-Hebrew University Medical Center, Jerusalem.

Plasmids and transfection

A cDNA encoding wild-type human Par2 was kindly provided by Prof. Morley D. Hollenberg (Faculty of Medicine, University of Calgary, Canada). Truncated hPar2 were generated by insertion of a stop codon at position K348Z of hPar2. This was prepared by using QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies Stratagene, Santa Clara, CA, USA) according to the manufacturer’s instructions. The designated primer used for inserting the stop codon TCA in hPar2 at position K348 is S348Z-5′-CGACCCCTTTGTCTAT TACTTTGTTTCACATGATTTCAGGG-3′. MCF7 (of endogenous hPar2) or HU (do not express either hPar1 or hPar2) cells were transfected with 1–2 μg of cDNA encoding wt human hPar1 or HA-tag-hPar1 or hPar2 or truncated hPar2 (devoid of the cytoplasmic tail), or T7-tag-etk/bmx or with a control pcDNA3 vector (Invitrogen, Carlsbad, CA, USA) using FuGene transfection reagent (Roche Molecular Biochemicals, Indianapolis, IN, USA). Transfected cells were selected with G418 (800 μg/ml) to obtain stable populations of cells expressing hPar1 and hPar2 or hPar1 and truncated hPar2.

RNA isolation and RT-PCR

RNA was isolated with Tri-Reagent (MRC, Cincinnati, OH, USA) according to the manufacturer’s instructions. After reverse transcription of 1 μg total RNA by oligo (dT) priming, cDNA was amplified using Taq DNA polymerase (Promega, Madison, WI, USA). Comparative semi-quantitative PCR was performed using the following primers: GAPDH sense: 5′-CCACCCATGGCAAATTCCATGGC-3′ and antisense: 5′-TCTAGACGGCAGGTCAGGTCC ACC-3′ primers. PAR₁ N-terminus primers were as follows: hPar1-sense: 5′- CTCGTCCTCAAG GAGCAAAC-3′, antisense orientation: 5′-TGGGATCGGAACTTTCTTTG-3′ (resulting with a 564-bp PCR product). PAR₂ primers: F: 5′-GGCCAATCTGGCCTTGGCTGA C-3′; R: 5′-GGC AGGAATGAAGATGGTCTG-3′.

Generation of shRNA for hPar2

To prepare shRNA constructs, we used a U6 promoter—driven and lentivirus (pLentiLox 3.7)—mediated delivery cassette of shRNA containing a stem-and-loop structure based on a 19-nucleotide sequence from the hPar2 coding region. shRNA cassette sequences were then ligated into pLentiLox 3.7 vector that encodes GFP (gift of Van Parij Laboratory, Massachusetts Institute of Technology, Cambridge, MA, USA). The sequence of shRNA for the hPar2 target gene is 5′- GGAAGAAGCCTTATTGGTA-3′. Scrambled sh-RNA hPar2 is 5′-GCAA GGTA AGCGTATGTTA-3′. The lentivirus particles were generated by a three-plasmid expression system, in which 293T cells were co-transfected with the following three vectors: packaging (CMVD R8.91), envelope (CMV-VSV-G), and the transfer vector pLentiLox 3.7. 293T cells were plated to 60 % confluency 1 day before transfection. Cells were transfected in fresh medium the following day with the three plasmids using Fugene 6 transfection reagent. Medium was replaced with fresh medium 24 h later. On days 2 and 3 after transfection, medium was collected and the viral particles were concentrated 100-fold by centrifuging for 1 h at 40,000 rpm.

Generation of a mutant R36A PAR₂. In the extra-cellular portion of PAR₂

GR/SLIGK whereby R36 was changed to A to generate R36A mutant by using QuikChange site-directed mutagenesis (Agilent Technologies, Santa Clara, CA, USA) using the following primer: 5′-GA ACC AAT AGA TCC TCT AAA GGA GCA AGC CTT ATT GGT AAG GTT GAT-3′. The inserted mutant R36A was confirmed by didoxy sequencing.

Western-blot and immunoprecipitation analysis

Cells were activated with agonist peptide TFLLRNPNDK for the indicated periods of time and solubilized in lysis buffer containing10 mM Tris–HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 % TritonX-100, and protease inhibitors (5 mg/ml aprotinin, 1 mM phenylmethylsulfonylfluoride, and 10 mg/ml leupeptin) at 4 °C for 30 min. The cell lysates were subjected to centrifugation at 12,000 rpm at 4 °C for 20 min. Supernatant (400 μg) was incubated with on of the following antibodies: anti-PAR₁ (ATAP, Santa Cruz, CA, USA; 1 μg/ml), anti-HA (HA.11 mAb COVANCE, Berkeley, CA, USA), anti-PAR₂ (SAM11; Santa Cruz, CA, USA; 5 μg/ml), anti-GFP (mAb 6556; Abcam, Cambridge, England) or anti-T7 (T7-Tag antibody; Novagen, Madison, WI, USA). After overnight incubation, protein A-sepharose beads (Amersham Pharmacia Biotech, Buckinghamshire, UK) were added to the suspension (50 μl), which was rotated at 4 °C for 1 h. Extensive washes with lysis buffer then followed to specifically obtain the immunocomplexes. The immunocomplexes were eluted and separated on a 10 % SDS-PAGE gel, followed by transfer to an Immobilon-P membrane (Millipore, USA). Membranes were blocked and probed with 1 μg/ml amounts of the appropriate antibodies as follows: anti-PAR₁ thrombin receptor mAb, (ATAP, from Santa Cruz, 1:1,000); anti-T7 (T7-Tag antibody; Novagen, Madison, WI, USA, 1:1,000) or a polyclonal anti-phospho-Akt (ser 473; 1:1,000; Cell Signaling, Boston, MA, USA), suspended in 3 % BSA in 20 mM Tris–HCl, pH 7.5, 15 mM NaCl, and 0.1 % Tween-20. After washing, the blots were incubated with secondary antibodies conjugated to horseradish-peroxidase. Immunoreactive bands were detected by enhanced chemiluminescence (ECL). Membranes were stripped and incubated with anti-IP antibodies to ensure equal protein load.

PAR₁ and PAR₂ and PAR₄ activation and antagonist

Thrombin as an activator was used at 1 U/ml (e.g., 10 nM). Selective PAR₁ activation was carried out by either TFLLRN or TFLLRNPNDK peptide and PAR₂ activation by SLIGKV. PAR₁ antagonist; SCH79797 (sc-203693, Santa Cruz Biotechnology, Inc. Dallas, TX, USA). PAR₄ ligand; AYPGKF (GenSript, Piscataway, NJ, USA). tcY-NH₂; PAR4 antagonist (Tocris Bioscience Inc, Minneapolis, MN, USA).

Immunofluorescence

HEK293T cells were plated on cover-slips that were pre-coated with poly-l-lysine, at least 24 h prior to transfections. Cells were transfected either with YFP-hPar2 (kindly provided by Dr. Morely D. Hollenberg, Department of Pharmacology and Therapeutics and Medicine, University of Calgary, Calgary, Canada; 0.5 μg/ml) or HA-hPar1 (0.5 μg/ml) or both for 48 h and then activated. Twenty-four hours before activation, the cells were starved in medium with 0.2 % BSA. The cells were activated by thrombin for brief periods (e.g., 2′, 5′ and 10 min) and then the medium was aspirated and the samples were washed twice with cold PBS and fixed with absolute cold methanol. Fixed cells were incubated with anti-HA antibodies (10 μg/ml, Covance Inc., Princeton, NJ, USA) followed by Cye3-conjugated anti-mouse IgG (4 μg/ml; Jackson Immuno-research Laboratories, PA, USA) as secondary antibodies. PAR₂ was detected directly by fluorescence imaging (green) of the YFP (at 514 wave length; 514–542). Nuclear staining was performed using DRAQ5 (4 μM; Cell Signaling; Boston, MA, USA). Images were obtained using a Zeiss LSM 5 confocal microscope and analyzed with Zen software (Carl Zeiss).

Colony formation in soft agar

A total of 5 × 10³ cells per well were plated on a 0.5 % agarose base in six-well plates in growth medium containing 0.35 % agarose. Cells were either thrombin activated (1 U/ml) or not (control) added every 48 h. After 12 days of incubation, images of the colonies were either taken with phase microscopy or after staining with 0.005 % crystal violet at room temperature for 1 h.

Placenta tissue collection and preparation

Placental tissues were prepared as previously described [32]. Briefly, the placental tissues were extracted from discarded material provided by patients who voluntarily and legally chose to terminate pregnancy during the first trimester (between 5 and 12 weeks gestation). Gestational age was determined by the date of the last menstrual period and ultrasound measurement of the crown-rump length. All specimens were obtained strictly in adherence with the Hadassah Hospital Institutional Ethics Committee Guidelines.

First-trimester villous explants in cultures

The preparation and cultivation of villous explants of different first-trimester placentas were performed as described elsewhere [32–34]. Briefly, the placental tissue originated from samples between 7 and 9 weeks gestation were immediately rinsed in sterile cold phosphate-buffered saline (PBS) and processed within 2 h of collection. Following dissection under microscope, approximately 2–5 mg weight specimens were carefully laid on 200 μl of solid reconstituted undiluted Matrigel substrate (Becton–Dickinson, Bedford, MA, USA) in 0.4 μm pore-culture Millicell-CM culture dish inserts (pore size 0.4 μm, Millipore Corp., Bedford, MA, USA). Explants were cultured in DMEM/F-12 supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin, and 0.25 μg/ml ascorbic acid, pH 7.4. Villous explants were maintained in culture for 3 days. Viability of the explants was assessed by adherence to Matrigel and emerging EVTs breaking from the tips as observed under a phase microscope. At 24 h in culture, explants were treated with thrombin (1 U/ml), TFLLRN (100 μM), and SLIGKV (100 μM) and infected or not with shRNA hPar2- GFP- or scrambled shRNAhPar2 viral vector. After 72 h in culture, the Matrigel cast cylinders with the EVTs were detached from the base of the insert cells by fine dissection under the microscope, fixed in 4 % paraformaldehyde, embedded in paraffin and further processed for histochemical tissue analysis. Each experimental condition was carried out in triplicate originating from eight different placental sets samples.

EVT invasion assessment

We assessed the H&E-stained serial sections of the EVT. We limited our study to the trophoblasts that penetrated through the Matrigel but remained on the surface of the polycarbonate membrane, that is, within the cast [33, 35, 36]. These were visualized on sequential slices of the entire Matrigel cast by hematoxylin staining. Slides were photographed by digital image capture software and all specimens were evaluated by two of the authors (S.G–G. and M.M.). Depth of invasion was assessed for each experiment (disappearance of trophoblast cells on the serial sections). At the deepest common level of invasion the total EVT cells at five randomly selected microscope fields (magnification 40×) were counted, for each experimental condition and the mean ± SEM/HPF calculated.

Matrigel invasion assay

Blind-well chemotaxis chambers with 13-μm-diameter filters were used for this assay. Polyvinylpyrrolidone-free polycarbonate filters, 8 μm pore size (Costar Scientific Co. Cambridge, MA, USA), were coated with basement membrane Matrigel (50 μl of 10 mg/ml Matrigel applied per blind well; 25 μg/filter). Briefly, the Matrigel was diluted to the desired final concentration with cold distilled water, applied to the filters and dried under a hood. Cells (2 × 10⁵) suspended in DMEM were added to the upper chamber. Conditioned medium of NIH-3T3 fibroblasts was added to the lower compartment of the Boyden chamber as a chemoattractant. Assays were carried out at 37 °C in 5 % CO₂. Following incubation, cells on the upper surface of the filter were removed by wiping with a cotton swab. The filters were fixed and stained with DifQuick System (Dade Behring Inc, Newark, NY, USA). Cells from various areas of the lower surface were counted and each assay was performed in triplicate.

Animal studies

Female athymic nude mice at 6–8 weeks of age were implanted subcutaneously with 2 × 10⁶ HU cells stably transfected with either hPar1 and hPar2, hPar2 alone, or hPar1 and a truncated form of hPar2. Mice were monitored for tumor size by external caliber measurements (length and width) at the time points indicated. Tumor volume (V) was calculated by V = L × W ² × 0.5, where L is length and W is width. After 3 weeks, the mice were killed and the tumors were excised, measured, and weighed (n = 5). All animal experiments were strictly in adherence with the Hadassah-Hospital Institutional Ethics Committee guidelines and approved by the animal committee of the Hebrew University, Jerusalem, Israel (MD-107.05-4).

Results

The silencing of hPar2 inhibits the association of signaling partners on PAR₁ C-tail

Etk/Bmx specifically associates with the PAR₁-C-tail via its pleckstrin homology (PH) domain and forms a scaffold assembly site that initiates the cell-signaling network [37]. To examine whether the association between the PAR₁-C-tail and Etk/Bmx is modulated by PAR₂, we utilized MCF7 clones expressing either HA-hPar1 and T7-tag-Etk/Bmx, HA-hPar1 7A mutant, which fails to bind Etk/Bmx (see supplementary data, Fig. S1A&B) or scrambled short hairpin (sh) RNA hPar2 in the presence of HA-hPar1 and T7-Etk/Bmx. Co-immuno precipitation analyses showed that while Etk/Bmx is immobilized onto PAR₁-C-tail following thrombin activation, this complex formation is completely abrogated when hPar2 is knocked down (MCF7 cells express endogenously moderate levels of PAR₂) (Fig. 1Ai and Supplementary data S4C). The complex formation is observed in the presence of scrambled shRNA-hPar2, similar to the association between control, non-treated cells of Etk/Bmx and PAR₁-C-tail in MCF7 clone over-expressing HA-hPar1 and T7-tag-Etk/Bmx (Supplementary data Figs. S1 and S2). No co-immunoprecipitation was seen following empty vector transfection (data not shown). Strikingly, in the presence of shRNA-hPar2 inhibition of the specific TFLLRN PAR₁ ligand [38] induced Etk/Bmx-PAR₁ complex is observed as compared with the scrambled shRNA-hPar2 following TFLLRN activation (Fig. 1Aii). Interestingly, SLIGKV selective PAR₂ activation also results with the immobilization of Etk/Bmx onto PAR₂ C-tail (Kancharla A et al, Manuscript in preparation). Therefore, it appears that both PAR₁ and PAR₂ follow at least partly, a similar pattern of signal activation.

Fig. 1.

Silencing of hPar2 inhibits association of key signaling partners with the PAR₁ C-tail. A i MCF7 cells stably over-expressing HA-hPar1 and T7-Etk/Bmx were infected with a lentiviral vector driving either shRNA- hPar2 expression or a scrambled shRNA-hPar2 and cultured with or without thrombin activation (1 U/ml) (A i) or the PAR₁ selective ligand, TFLLRN (100 μM) (A ii). Cell lysates were collected at the indicated times and immunoprecipitated using anti-HA antibodies. Immunoprecipitates were separated on SDS-PAGE and Western blotted with anti-T7 antibody to detect T-7-tagged Etk/Bmx-T7. Etk/Bmx is seen after 15 and 20 min of thrombin treatment in cells without hPar2 silencing (scrambled shRNA-hPar2); but is not detectable in cells infected with shRNA-Par2. A specific, Etk/Bmx is seen associated after 5′, 10′, 15′, and 20 min following TFLLRN activation in cells expressing both HA-hPar1 and endogenous hPar2 (scrambled shRNA-hPar2). In-contrast, the Etk/Bmx association is not detected in the presence of shRNA-hPar2. B Silencing hPar2 inhibits thrombin-induced association between PAR₁ and Akt. MCF7 clones overexpressing HA-hPar1, with shRNA hPar2 silencing (right), or scrambled shRNA hPar2 (left), were treated with thrombin for the indicated times. Cell lysates were immunoprecipitated with anti-HA to precipitate HA-PAR₁ and analyzed by Western blot with anti-Akt antibodies, followed by antibodies to phosphorylated Akt (pAkt). HA-PAR₁ serves as a loading control. C PAR₂ expression in the MCF7 clone (expressing hPar1, hPar2, and etk/bmx) with and without shRNA silencing as compared with a housekeeping gene GAPDH. D Silencing hPar2 inhibits TFLLRN PAR₁ activation in MDA-MB-231 cells. MDA-MB-231 cells expressing high endogenous levels of both PAR₁ and PAR₂ show a similar pattern (to MCF7 cells) of selective PAR₁ inhibition in the presence of sh-RNA hPar2. TFLLRN activation induces the co-IP between PAR₁ and Akt immediately after 2′, 10′, and 20′ activation (scrambled sh-RNA hPar2). In the presence of sh-RNA hPar2, this association is markedly inhibited. β-actin serves as a loading control

We also examined in these clones the interactions between the PAR₁-C-tail and Akt, a prime signaling partner in cancer cells. For this purpose, anti-HA antibodies were used to immunoprecipitate the HA-tagged PAR₁ protein, following thrombin activation. Association of Akt with the PAR₁ C-tail increased with time. Furthermore, using anti-phospho-Akt antibodies, it is demonstrated that PAR₁-bound Akt undergoes phosphorylation, indicating that Akt is functional in the PAR₁-immobilized state. Knocking down hPar2 abrogated the association of Akt with PAR₁ C-tail, as well as its phosphorylation following thrombin activation. A scrambled shRNA-hPar2 had no effect on the complex formation between PAR₁-C-tail and Akt as also p-Akt (Fig. 1b see also Fig. S1). Similarly, empty vector transfection did not result in any co-immunoprecipitation (MCF7 cells do not express PAR₁; data not shown). The levels of hPar1, hPar2, Etk/Bmx, and degree of hPar2 silencing in these clones are shown in Fig. 1c. A similar pattern of selective PAR₁ inhibition (by shRNA- hPar2) is also observed in MDA-MB-231 cells expressing endogenously high levels of PAR₁, PAR₂, and Akt. In the presence of scrambled shRNA-hPar2, a potent co-association between PAR₁ and Akt is observed following TFLLRN activation. This complex formation is completely abrogated in the presence of shRNA- hPar2 (Fig. 1d).

Silencing hPAR2 significantly reduces soft agar colony formation, Matrigel invasion, and wound migration

Anchorage-independent growth in soft agar as manifested by colony formation is a well-established indicator of tumorigenicity [39, 40]. Normal cells are incapable of multiplying in a semisolid environment. Furthermore, there is a good correlation between soft-agar colony formation and the extent of tumor generation in vivo in animal models. While MDA-MB-231 cells formed very small colonies prior to activation (Fig. 2Ai, a, b; control), large colonies in soft agar were observed following treatment with thrombin (Fig. 2Ai, panel c) or with the PAR₁-specific ligand TFLLRNPNDK (Fig. 2Ai, panel d). In contrast, cells expressing shRNA-GFP-hPar2 were incapable of generating colonies following the addition of either thrombin (Fig. 2Ai, panel e, cells expressing shRNA-hPar2 visualized by GFP fluorescence) or TFLLRNPNDK (Fig. 2Ai, panel f; the small-size colonies shown, all express shRNA-hPar2 due to high lentiviral infection levels). Scrambled hPar2 had no effect on the formation of large colonies in the presence of thrombin activation (Fig. 2Ai panel g). Cells infected with shRNA- hPar1 and thrombin activated formed small-size colonies (Fig. 2Ai panel h, as compared with panel c; for control non-silenced cells after thrombin activation). SLIGKV PAR₂-specific activation showed the formation of rather large colonies (Fig. 2Ai panels i, j). Overall, when grown on soft agar, MDA-MB-231 sh-hPar2 cells formed significantly smaller foci and a smaller number of foci than control MDA-MB-231-activated cells. In-contrast, MDA-MB-231 cells formed significantly larger foci following PAR activation than control cells, while the number of foci remained the same. Figure 2Aii shows that infection with the shRNA-hPar2 vector decreased hPAR₂ expression to undetectable levels without altering hPar1 expression. Histograms represent the mean colony area size/cm² under the different treatments (Fig. 2Aiii). Thus, reducing PAR₂ expression alone not only attenuates PAR₁ and PAR₂-driven cell transformation but also considerably inhibits transformation induced by the selective activation of PAR₁. The data were statistically evaluated by ANOVA Tukey HSD of multiple comparison, showing a p value of 0.004 within groups. The mean difference is significant at the 0.05 level.

Fig. 2.

A i Silencing hPar2 in a breast cancer cell line inhibits thrombin- or TFLLRNPNDK-induced colony formation in soft agar. 5 × 10³ MDA-MB-231 cells, either uninfected (a, b) or infected with virus expressing shRNA-GFP-hPar2 (e, f), were plated in soft agar and activated by thrombin (e) or TFLLRNPNDK (f). Uninfected cells were treated with thrombin (c) or TFLLRN (d) as also with SLIGKV (i, j). Controls of either scrambled shRNA hPar2 with thrombin activation (g) and shRNA hPar1 with thrombin activation (h) were shown as well. After 12 days, live images were collected using a Zeiss microscope at ×20 magnification. In uninfected cells, large colonies formed following treatment with either thrombin (c) and somewhat smaller with TFLLRNPNDK (d). MDA-MB-231 cells infected with sh-GFP-hPar2 formed only very small colonies after thrombin (e) or TFLLRNPNDK (f) treatment. Cells containing sh-hPar2 are visualized by GFP fluorescence (e). The colonies formed under different conditions were compared with control non-activated, non-treated cells (control; a, b). The colonies formed in the presence of shRNA-hPar1 and after thrombin activation is significantly smaller (panel h) than colonies formed after thrombin activation (c). Scrambled shRNA-hPar2 had no effect on the colonies formed (panel g). Comparable colony size is obtained by SLIGKV PAR₂ activation (i, j) as found by TFLLRNPNDK (d). Images shown are representative of three independent experiments. A ii RT-PCR analysis of hPar1 and hPar2 mRNA expression before and after shRNA silencing. GAPDH levels were analyzed as a control. A iii Histogram shows mean ± SE of triplicate values from three independent experiments. Post hoc evaluation of multiple comparison (ANOVA Tukey HSD) showed a p value of 0.004 within groups. The mean difference is significant at the 0.05 level. For ANOVA evaluation we used IBM SPSS 20.0 software. B i Silencing hPar2 in MDA-MB-231 cells inhibits TFLLRN PAR₁ invasion. Matrigel invasion in the presence of shRNA-hPar2-infected cells as compared with scrambled shRNA. While TFLLRN specific activation of PAR₁ induces Matrigel invasion in MDA-MB-231 cells, this was attenuated in the presence of shRNA-hPar2-infected cells. In-contrast, no effect was observed when a scrambled shRNA was utilized to infect the cells, demonstrating a markedly induced Matrigel invasion similar to non-treated activated parental MDA-MB-231 cells. B ii Histograms represents quantification of the cells/HPF invaded the Matrigel layer. Unpaired Student’s t test was used. This experiment is a representative of three independent experiments performed in triplicates. B iii Silencing hPar1 in MDA-MB-231 cells did mot inhibit PAR₂ function. In MDA-MB-231 cells silenced for hPar1, thrombin activation resulted with a low level of Matrigel invasion. In-contrast, SLIGKV PAR₂ activation under conditions of sh-hPar1 silencing, resulted with potent Matrigel invasion. B iv Histograms represents quantification of the cells/HPF invaded the Matrigel layer in shRNA hPar1 silenced cells. Unpaired Student’s t test was used. This experiment is a representative of three independent experiments performed in triplicates

Silencing of hPar2 also inhibited significantly TFLLRN-induced invasion of MDA-MB-231 cells in a Matrigel invasion assay, while scrambled shRNA-hPar2 had no effect (Fig. 2Bi, ii). In parallel, when we analyzed Matrigel invasion in sh hPar1 MDA-MB-231-silenced cells, we observed a marked inhibition following thrombin activation. This was not seen following SLIGKV activation, whereby potent Matrigel invasion is observed (Fig. 2Biii, iv). Therefore, the same pattern as in MCF7 cells is observed inhibiting Matrigel invasion in the presence of silenced hPar2 but not vice versa by sh hPar1. We demonstrate specificity toward PAR₁ by inhibition in the presence of PAR₁ selective antagonist SCH79797 (Supplementary data Fig S3). In order to exclude the possibility that PAR₄ is involved in MDA-MB-231 cell invasion (since these cells express also PAR₄, see Supplementary data Fig. S4, C), we performed Matrigel invasion in the presence of a PAR₄ agonist, AYPGKF. As can be observed, a low level of invasion was obtained (Fig. S4, Ai, Aii). Next, we performed Matrigel invasion following thrombin activation, in the presence of tcY-NH2 (Tocris Bioscience Inc, Minneapolis, MN, USA), a PAR₄-specific antagonist. Thrombin activation induced Matrigel invasion with no significant impact of PAR₄ antagonist on the invasion level Fig. S4, Ai, Aii). Similarly, we show that the PAR₄ antagonist does not have a significant impact on Thrombin-induced co-association between PAR₁ C-tail and Akt, in MDA-MB-231 cells (Fig. S4, B). We thus conclude that PAR₂ affects PAR₁ function and the involvement of PAR₄ in this process is marginal.

Next, we looked at the effect of silencing by shRNA-hPar2 or shRNA-Par1 on the migration of tumor cells using a tissue monolayer-wound assay. Cells of the aggressive breast cancer line MDA-MB-468 (expressing high levels of PAR₁ and PAR₂; Supplementary data Fig. S5A) or PC3 prostate cancer cells (Supplementary data Fig. S5B) were infected with lentiviral vectors expressing either shRNA-hPar2 or shRNA-hPar1. A carefully defined wound (1-cm-wide scratch) was made in confluent MDA-MB-468 cells or PC3 cells following infection and in uninfected confluent cultures. Cells were then treated with thrombin and migration of cells into the empty space was assessed over time. There was extensive cell migration in the non-infected cells, closing the gap within 24 h. In-contrast, migration was clearly attenuated in cells infected with either shRNA-hPar1 or shRNA-hPar2; shRNA-hPar2 had a more pronounced effect than shRNA-hPar1 (Fig. S5Ai, iii and S3Bi, iii, Supplementary data). When either PC3 or MDA-MB-468 cells were either infected with shRNA-hPar1 or with shRNA-hPar2, there was no detectable expression of PAR₁ or PAR₂, respectively (Figs. S5Aii and S5Bii, Supplementary data).

Silencing of hPar2 significantly inhibits thrombin-induced EVT cell invasion

A decisive key step in the establishment of human pregnancy is the well-orchestrated invasion of the uterus wall by specialized cells termed extravillous trophoblast (EVT). These cells are in fact a subpopulation of placenta cytotrophoblast capable of forming anchoring villi that invade and reach the uterine wall thereby allowing direct contact with the maternal blood. A well-established in vitro model system based on the isolation of villi from early gestation placentas plated on Matrigel as a substitute for the extra cellular matrix (ECM) is utilized [33–36]. The advantage of this system is that cytotrophoblasts that differentiate along the invasive pathway mimic at large the development of normal early placenta villi [36]. We initiated our evaluation by comparing the depth of invasion under the different experimental conditions: infection of shRNA-hPar2- viral vector as compared with scrambled sh-RNA-hPar2 before and after either thrombin or TFLLRN or SLIGKV activation. The morphology of the villous trophoblast as well as the extra villous cell outgrowth columns in the various treatment groups is illustrated.

The EVT invasion into the Matrigel reached as deep as 110 μm, as observed by representative 5-μm serial sections of the Matrigel cast. Histological analyses of H&E staining are shown in Fig. 3a. The preparations of shRNA-hPar2 viral vector following activation of the PARs (e.g., thrombin, TFLLRN, SLIGKV) showed reduced EVT invasion, reaching levels of maximum 70 and 80 μm. In contrast, activation of PARs (e.g., thrombin and/or TFLLRN and SLIGKV), showed deeper invasion (e.g., 110 μm). The number of cells per high-power field for each of the treatments was evaluated at an equal level of invasion (60 μm), as shown in a representative histogram (Fig. 3a). A high level of invading cells was observed after either thrombin, TFLLRN, or SLIGKV activation, regardless of either before or after scrambled the shRNA hPar2 treatment. In conclusion, a significant inhibition was observed in the presence of shRNA-hPar2 lentiviral vector following the different activation treatments (e.g., thrombin, TFLLRN, and SLIGKV). The relative levels of PAR₁ and PAR₂ in the presence and absence of either sh hPar2 or scrambled sh hPar2 in the EVT organ culture is shown in Fig. 3c. The count/HPF data was statistically evaluated by ANOVA Tukey HSD of multiple comparison showing a p value of 0.0001 within groups. The mean difference is significant at the 0.05 level.

Fig. 3.

Placenta-EVT organ culture invasion. A Morphology of EVT column growth. Invasion of placental explants of EVT-Matrigel cylinder cast after H&E staining (magnification ×20). Serial sections (5 μm each) of the Matrigel cylinder casts were embedded in paraffin blocks prepared for each of the experimental treatment conditions. Images of representative H&E-stained sections at 60-μm depth are presented. a Non-treated (control); b Thrombin (1 U/ml) activation; c TFLLRN (100 μM) activation; d SLIGKV (100 μM) activation; e shRNA hPar2 lentiviral infected EVT and thrombin (1 U/ml) activated; f shRNA hPar2 lentiviral-infected EVT and TFLLRN (100 μM) activation; g Scrambled shRNA hPar2 lentiviral infected EVT and thrombin (1 U/ml) activation; h shRNA hPar2 lentiviral-infected EVT and SLIKGV (100 μM) activation; i shRNA hPar2 lentiviral-infected EVTs. Note the increase in the EVT migrating cells and cell mass at the tip of the villi following either of PAR activation. In contrast, a blunt end of the villi with no sprouting cells is seen at the villi tip column in the presence of shRNA hPar2 treatment following the different activation treatment. The experiment was terminated 3 days (72 h) after EVTs treatment. Each experiment was performed using three different placentas, in triplicates. Arrowheads show sprouting cytotrophoblasts. B Quantification of EVT outgrowth cell number at 60-μm depth of invasion as shown by histogram mean values. Cells (not the villous compartment) were counted per high-power field and expressed as mean ± SEM. C RT-PCR analyses of PAR₁ and PAR₂ in the cultured EVTs. GAPDH serves as a control. Post hoc evaluation of multiple comparison (ANOVA Tukey HSD) showed a p value of 0.0001 within groups. The mean difference is significant at the 0.05 level. For ANOVA evaluation, we used IBM SPSS 20.0 software. D Schematic representation of EVT experimental model for invasion

Co-localization of PAR₁ and PAR₂

It has been proposed previously that PAR₁ and PAR₂ are localized near each other on the cell membrane, enabling their activation as an intact functional unit. We examined their relative localization on breast cancer cells in situ by confocal microscopy. HEK293T cells were transfected with both HA-hPar1 and YFP-hPar2. Twenty-four hours later, the cells were serum deprived for an additional 24 h and then activated with thrombin for 5 min. PAR₁ was visualized by anti-HA antibodies and Cy3-conjugated secondary antibodies or direct fluorescence detection for YFP-PAR₂. Activation showed merge co-localization of both PAR₁ and PAR₂ in the cell membranes (Fig. 4A, panels d, iv) as compared with either PAR₂ (Fig. 4A panels a, i) or HA-PAR₁ (Fig. 4A, panels b, ii). Control cells prior to activation exhibited abundant PAR₁ and PAR₂ on the cell surface, indicating that the transfection was efficient (data not shown). Cell nuclei were detected by draq 5 (Fig. 4A, panels c, iii) and gross cell morphology was detected by phase-contrast microscopy (Fig. 4A, v).

Fig. 4.

Co-localization of PAR₁ and PAR₂. A Confocal immunofluorescence analysis. HEK-293T cells were transfected with both HA-hPar1 and YFP-hPar2. After 24 h, the cells were serum deprived for an additional 24 h then activated with thrombin for 5 min followed by fixation with cold methanol. PAR₂ was visualized by direct fluorescence (a) and PAR₁ was visualized by immunostaining with anti HA-antibodies followed by Cy3-conjugated IgG secondary antibodies (b). Merge staining for both PAR₁ and PAR₂ revealed co-localization confined to the cell membrane (d). For reference, staining of cell nuclei with Draq5 is shown (c). The bottom panel highlights staining in one cell exhibiting the expression of YFP-hPar2 and HA-hPar1 on the cell membrane prior to activation (i, ii). Significant co-localization recapitulated by merge fluorescence is observed following five minutes thrombin activation (iv). This is detected as compared with cell nuclei staining (iii) and phase-contract analysis (v). The percent of positive cells for HA-hPar1 was 58 % ± 1.2 and for YFP-hPar2 were 55 % ± 1.86; PAR₁-PAR₂ merge was observed in 45 % ± 2.1. B PAR₁ and PAR₂ co-immunoprecipitate. HEK 293T cells expressing either HA-hPar1, YFP-hPar2 or GFP-CXCR4 were treated with thrombin for 10 min and lysed. Cell lysates were then immunoprecipitated before (−) and after (+) thrombin activation by anti HA or IgG, resolved by SDS-PAGE and immunoblotted as indicated (by anti-GFP). While no specific complex is formed when CXCR4 and PAR₁ were co-IP (following 10-min thrombin activation), a specific complex is observed between PAR₁ and PAR₂ following activation. No specific band is seen when IgG were applied in the immunoprecipitation assay. Expression of CXCR4, PAR₂ (as indicated by anti-GFP, first lane), and PAR₁ (as shown by anti-HA) were shown as controls for transfection efficiency. The figure shown is a representative of three independent experiments

Co-localization of PAR₁ and PAR₂ was quantitated with the “Image-Pro Plus” (version 4.5) program, which can specifically calculate the extent of co-localization within a defined field (co-localization is expressed using a scoring range from a low of 0.2 up to 1.0, which indicates maximal co-localization). When we used “Image Pro Plus” to analyze the HEK293T cell membrane compartment, scores of 0.97–0.98 were obtained throughout the field, indicating high co-localization between PAR₁ and PAR₂ following activation by thrombin.

To obtain direct evidence for PAR₁–PAR₂ heterodimer formation, we carried out co-immunoprecipitation analyses. For this purpose, we ectopically over-expressed HA-tagged-hPar1 with YFP-tagged-hPar2. As a control, we used CXCR4, another member of the GPCR family. For this purpose, we expressed GFP-cxcr4 in HEK293T cells along with HA-hPar1. The cells were treated with thrombin for 10 min and then lysed. Next, we performed immunoprecipitation using either anti-HA antibodies, or IgG (as control), separated on SDS-PAGE and Western blotted with anti-GFP antibodies. PAR₂ were detected in the immunoprecipitates after thrombin activation; (+) as compared to PAR₂ levels prior to activation; (−) and similar to the position of PAR₂ in cell lysates expressing only PAR₂ (PAR₂) (Fig. 4B). No association was seen in cells expressing GFP-cxcr4 or following immunoprecipitation with IgG. These results provide supporting evidence that a specific complex is formed between PAR₁ and PAR₂ following activation, but not with PAR₁ and CXCR4.

Modulation of PAR₁-PAR₂ induced breast tumors by a truncated PAR₂

The basis for this study is the concept that PAR₁ and PAR₂ act together as a functional unit in response to thrombin application, based on the close proximity position enabling the formation of heterodimers or hetero-oligomers. Knocking down of PAR₂ considerably attenuated the response to pro-tumor processes induced either by thrombin or the specific PAR₁ ligand. We therefore sought to examine the effect of a truncated form of PAR₂ (PAR₂ lacking the cytoplasmic tail) both in vitro and in vivo on PAR₁ tumor growth (for characterization of truncated hPar2 cell surface expression, see Supplementary data, Fig. S6). For this purpose, we have utilized HU cells, nearly normal breast epithelial cells that do not express neither hPar1 nor hPar2 (see Supplementary data; Fig. S4, C). Toward this, we generated stable clones of HU cells over-expressing the following plasmids. A clone expressing HA-tagged hPar1, T7-tagged-Etk/Bmx and wt hPar2; another clone expressing HA-tagged hPar1, T7-tagged-Etk/Bmx and a truncated hPar2; a clone of hPar2 alone or a clone expressing empty vector alone. The association between Etk/Bmx and PAR₁ C-tail in these clones was first analyzed. Cell lysates were processed and immunoprecipitated with anti-HA antibodies, followed by Western-blot detection of the T7-tagged-Etk/Bmx immunocomplex with anti-T7 antibody. Etk/Bmx was detected with HA-tagged PAR₁ in cells co-transfected with wt hPar1 and hPar2 following 10 min of thrombin activation. In contrast, in the clone expressing hPar1 and a truncated form of hPar2, thrombin treatment elicited a markedly less association between PAR₁ C-tail and Etk/Bmx (Fig. 5A). The levels of Etk/Bmx, truncated PAR₂, and HA-hPar1 expression is shown (Fig. 5Aii). This finding suggests that the presence of truncated PAR₂, lacking the cytoplasmic C-tail inhibits PAR₁-PAR₂ functional activity. We further demonstrate that the truncated form of PAR₂ co-immunoprecipitates with PAR₁. For this purpose, we have used HEK293T cells expressing HA-hPar1 and truncated form of PAR₂. The cells were treated for 10 min with thrombin (+), lysed, and immunoprecipitated using anti-PAR₂ or IgG (as control). Western-blot analysis using anti-HA antibodies showed the presence of PAR₁ in the PAR₂ immunoprecipitates following thrombin activation but not prior to activation (−), or by the use of IgG. The position of PAR₁ in the blot is compared with PAR₁ size obtained in the cell lysates (Fig. 5Aiii). Hence, the truncated form of PAR₂ is well expressed on the cell-surface (Supplementary data, Fig. S6) presumably in a close proximity with PAR₁, and upon activation forms a complex with PAR₁.

Fig. 5.

A A truncated form of hPar2 (devoid of C-tail) impairs PAR₁ C-tail tail ability to associate with Etk/Bmx following thrombin activation. i HU stable clones over-expressing either hPar1&hPar2 or hPar1 and a truncated form of hPar2, also expressing T7-etk/bmx, were immunoprecipitated using anti-HA antibodies to immunoprecipitate HA-PAR₁. The PAR₁ immunoprecipitates were analyzed by Western blot with anti-T7 antibodies to detect T7-tagged Etk/Bmx. Anti-HA blotting is shown as a loading control. Plasmid expression levels in the various clones are indicated (ii). Clones expressing HA-hPar2 and T7-etk/bmx were analyzed for transfection efficiency (iii). PAR₁ co-immunoprecipitates with a truncated PAR₂ form. HEK 293T cells expressing HA-hPar1 and a truncated PAR₂ form were treated with thrombin for 10 min and then lysed. Cell lysates before (−) and after (+) thrombin activation were then immunoprecipitated by either anti-PAR₂ antibodies (SAM11; Santa Cruz, CA, USA) or IgG, resolved by SDS-PAGE and immunoblotted as indicated (by anti-HA abs). While no specific band is seen when IgG was used as a control, increased PAR₁ is observed following thrombin activation (+) as compared to levels before activation (−). In lysates expressing hPar, blotting with anti-HA antibodies indicate the position of PAR₁ (e.g., PAR₁). Levels of truncated PAR₂ are indicated by anti PAR₂ antibodies (directed to the N-terminus of PAR₂). Anti-HA was shown for PAR₁ transfection levels. This experiment is a representative of three independent experiments. B Matrigel invasion induced by selective PAR₁ activation. Matrigel invasion is induced by TFLLRN treatment of HU clones over-expressing hPar1 and hPar2. In contrast, TFLLRN fails to induce Matrigel invasion in cells over-expressing hPar1 and a truncated form of PAR₂. Images are taken at ×20 magnification. Histograms representing the extent of Matrigel invasion are shown. Unpaired Student’s t test was used. This experiment is a representative of three independent experiments performed in triplicates. Error bars, ± SD; **p < 0.007

In a Matrigel invasion assay, selective activation of PAR₁ with TFLLRN considerably enhanced the invasion of HU cells over-expressing both PAR₁ and PAR₂. This PAR₁-induced invasion was noticeably inhibited in clones stably expressing PAR₁ and the truncated form of PAR₂ (Fig. 5Bi, ii).

Noticeably, TFLLRN activates specifically PAR₁, but not PAR₂ (see Supplementary data Figs. S8, S9). When the PAR₂ mutant R36A, incapable of extra-cellular cleavage by trypsin for activation was silenced, a marked inhibition of TFLLRN PAR₁ induced Matrigel invasion (Supplementary data Fig. S9) was observed. Hence, the physical presence of PAR₂ in a non-activated state is sufficient and required in close proximity for PAR₁ function. When the clone expressing wt hPar1 and a truncated form of hPar2 were re-infected with wt hPar2, a complete rescue was observed after TFLLRN activation showing a potent Matrigel invasion properties (Supplementary data, Fig. S7).

The various clones (e.g., wthPar2, hPar1&hPar2, hPar1, and truncated hPar2), were then inoculated subcutaneously into nude mice. The mice were monitored for 35 days (at which time the tumors became noticeable), killed, and tumors were collected and embedded in paraffin. While minimal to nearly no tumors were seen following injection of the parental HU cells, large tumors were observed after injection of HU clones over-expressing hPar1 and hPar2 (as well as the prime signal protein; T7-tagged-Etk/Bmx), similar to the tumors obtained following inoculation of cells expressing hPar2 alone. In the presence of hPar1 and a truncated form of hPar2, considerably smaller tumors were seen (Fig. 6A, B). These tumors were also significantly smaller than PAR₁-driven tumors (data not shown). The levels of tag-PAR₁, wt PAR₂ and truncated PAR₂ expression in two clones are shown (Fig. 6C).

Fig. 6.

Impaired tumor development in the presence of PAR₁ and truncated PAR₂. Stable clones of HU cells expressing either hPar2 or hPar1 and hPar2, or hPar1 and truncated hPar2 were injected subcutaneously into nude mice (2 × 10⁶ cell/mouse). Parental non-transfected HU cells were used as a control. A Tumor morphological appearance. Mice injected either with hPar1 and hPar2 or hPar2 developed tumors at the sites of injection. In-contrast, mice inoculated with wt hPar1 and a truncated form of hPar2 did not develop or developed very little tumors (a). By the end of the experiment (e.g., 35 days) the mice were killed and the tumors were excised, measured, and weighed (n = 4/per group). B Mouse tumor growth curve in nude mice. Tumors were excised, weighed and measured at the indicated times and tumor volume (mm³) was calculated. Error bars show SD; * indicates p < 0.006. Data shown are representative of three experiments performed. C i, ii RT-PCR analyses of the indicated clones (a and b; two different clones) show the expression of PAR₁ and PAR₂, truncated PAR₂ and the prime signal protein Etk/Bmx

Together, the in vitro and in vivo effects of introducing a truncated PAR₂ indicate that truncated PAR₂ has an inhibitory effect on PAR_1/PAR₂-induced tumor-promoting processes similar to the inhibition seen following shRNA silencing.

Discussion

It has been shown that PAR₁ transactivates PAR₂ donating its cleaved new N-terminal domain to bind inter-molecularly and activate in trans PAR₂ [26, 27]. Ample evidence has been gathered since the first demonstration by O’Brien et al. [26] showing conclusively that PAR₁-tethered ligands can transactivate PAR₂ and provide a full agonist for its function. This takes place in vivo as shown by Kaneider et al. [24] that PAR₂ associates with PAR₁ and switches thrombin-induced endothelial cell from a barrier disruption function to a protective role. In another work, PAR₂ expression was shown necessary for PAR₁-induced hyperplasia in mouse models [28]. Recently, an in-depth work by the group of Trejo JA showed that the internalization trafficking machinery of PAR₁–PAR₂ heterodimer differs from PAR₁ trafficking [29]. Albeit, the molecular mechanisms that regulate PAR₁–PAR₂ heterodimer formation, signaling, and trafficking are not yet known.

In the present study, we show that PAR₁ and PAR₂ act as a functional unit in promoting breast cancer. We found that PAR₁-induced breast tumor development and the corresponding signaling events are markedly inhibited when PAR₂ expression is either knocked-down or lacking the C-tail portion. Hence, PAR₂ must be present for a wide spectrum of tumorigenic activities and for the initiation of PAR₁ early signaling events but not vice versa.

Our studies confirm the close proximity of PAR₁ and PAR₂, which leads to co-localization following appropriate activation. This was demonstrated by both confocal image analysis and co-immunoprecipitation of PAR₁–PAR₂ forming a complex as shown following the ectopic over-expression of either GFP-cxcr4, HA-hPar1, and YFP-hPar2 plasmids. While specific co-immunoprecipitation is seen between HA-hPar1 and YFP-hPar2, no such complex is observed when GFP-cxcr4 and HA-hPar1 are used. Certainly, the truncated form of PAR₂ is also found localized in close proximity with PAR₁ allowing a potent complex formation. The physiological significance of the presence of PAR₂ cytoplasmic C-tail is demonstrated in a xenograft mouse model for tumor development, showing a markedly reduced tumor formation when a truncated hPar2 is present.

Our findings are in line with the elegant studies by Sevigny et al. on PAR₁/PAR₂-driven hyperplasia in medial and neointimal arterial SMC growth [28, 41], and support preceding work from this group and others on PAR₁ and PAR₂ trans-activation [25, 27]. It is not surprising that silencing of PAR₂ reduces functional outcomes of PAR₁/PAR₂-driven tasks. However, the marked inhibition of a selective PAR₁-initiated activities, previously shown by the PAR₁-P1pal-13 intracellular specific activation [28, 41] and in the present study by TFLLRNPNDK, is not fully understood and needs further exploration. PAR₂ was found inactivated in the VSMC system, suggesting that the physical presence of PAR₂ is important even in the non-activated state. We present now data showing that the physical close proximity of PAR₂ is sufficient and required for PAR₁ selective function. When a mutant of PAR₂ (R36A) that is not activated by trypsin was silenced by shRNA-hPar2, a marked inhibition of TFLLRN induced Matrigel invasion was obtained. In parallel, we demonstrate that while trypsin effectively induced PAR₂-driven Matrigel invasion, TFLLRN had no effect resulting with null to very little PAR₂-induced Matrigel invasion (see Supplementary data Figs S8, S9). Hence, the presence of PAR₂ is required for PAR₁ function also in a non-activated state. More specifically, it is suggested that the presence of PAR₂ C-tail is essential, although this role needs yet to be further elucidated. The functional dependence of PAR₁ on PAR₂ may be due to the formation of heterodimers or oligomers, as shown in other cell types. However, it is also possible that PAR₂ has a dominating regulatory effect on PAR₁ in a manner that has not yet been discovered. Interestingly, recently it has been demonstrated that PAR₁ drives the trafficking behavior of PAR₂ [29]. In addition, the current data do not rule out the possibility that PAR₂ acts through a third, yet unknown, partner that is necessary for PAR₁ efficient activity. The fact that a truncated PAR₂ inhibits PAR₁ suggests that there is essential signaling cross-talk between PAR₁ and PAR₂.

Along this line of evidence, a different signaling interface was shown by PAR₁-PAR₂ heterodimers [29]. Thrombin-induced internalization trafficking indicates that PAR₁–PAR₂ heterodimer stability leads to a conformational change that allows co-internalization and β-arrestin recruitment, an event that is not observed by activated PAR₁ (which rather binds the clathrin adaptor; AP-2) [42]. The β-arrestin recruitment to PAR₁-PAR₂ heterodimer occurs through a distinct interface and elicits a unique signaling response that is different from signaling by PAR₁ alone. Therefore, thrombin PAR₁-PAR₂ dimer formation results in a biased response elicited by β-arrestins. We have previously demonstrated that PAR₁ promotes the binding of β-arrestin-2 to disheveled (DVL), suggesting a role for β-arrestin-2 in PAR₁-induced DVL phosphorylation dynamics and β-catenin stabilization [43]. It remains to elucidate whether thrombin PAR₁–PAR₂ formation evokes a biased pro-tumor signaling event that represents a different interface than either PAR₁ or PAR₂ alone.

In fact, it was demonstrated [29] that activated PAR₂ requires its C-tail for the stable association with β-arrestins but not for internalization, whereby a truncated form of PAR₂ effectively internalizes [44]. Indeed, the PAR₂ C-tail is not required or necessary for thrombin-induced β-arrestin recruitment to the PAR₁–PAR₂ on the endosomes. Hence, it appears that the truncated PAR₁–PAR₂ dimer internalizes similar to PAR₁–PAR₂ and recruits β-arrestins. Notably, we provide the first demonstration in the context of cancer that the use of a truncated form of PAR₂ attenuates PAR₁-driven pro-tumor function in vitro and in vivo.

It is also not clear whether activation of PARs leads to homodimer as well as heterodimer formation is not clear. It was previously suggested, based on BRET2 analysis, that both PAR₁–PAR₃ heterodimers and PAR₁–PAR₁ homodimers are formed in vascular endothelial cells and induce distinct signaling repertoires: PAR₁–PAR₃ heterodimers appeared to be selectively coupled to Gα13 as compared with the PAR₁–PAR₁ homodimer, which coupled to both G_αq and G_α13 [22].

Formation of heterodimeric receptors most likely allows a broader range of activation outcomes and more complex regulation than obtained with a monomer. In the case of PAR₁ and PAR₂, a broader G-protein response may be observed as PAR₁ functions via G_α12/13 and PAR₂ via G_αq, G_α11, G_α19, G_α12, and G_α13 [45]. However, Etk/Bmx is immobilized onto the PAR₁ C-tail as an early key partner [37], and we find a similar association of Etk/Bmx with PAR₂-C-tail following activation (Kancharla et al., manuscript in preparation). This suggests that in the heterodimeric status when the two C-tails are in close proximity, the amplification of an early signal associating partner rather takes place and not expansion to additional signaling partners. At the present time, the possibility that the heterodimers affect receptor internalization and trafficking cannot be ruled out [46] and was recently addressed [29].

Emerging data appointed PAR₂ with a major role in breast cancer development. This is based among others, on the observed delay in the onset of tumors, following intercross between PAR₂ knock-out mice and polyoma middle T (PyMT) mammary gland mouse model for spontaneous tumor growth but not via the intercross with PAR ^−/−₁mouse [47]. Although this was observed in one mouse model, it may suggest that PAR₂ plays a dominant role and drives pro-tumor function. This PAR₂ pro-tumor function may be initiated by another coagulation component, tissue factor (TF), since TF cytoplasmic domain-deleted mice have been shown also to delay the spontaneous breast cancer development in the polyoma middle T model [48].

A similar approach to assess the relative contribution of PAR₁ and PAR₂ in breast cancer was taken previously [49]; those studies used shRNA silencing of PAR₁ or PAR₂ and revealed a significant role of PAR₂ in the process. Like PAR₁, PAR₂ was first identified as a cell receptor activated by FVIIa and Xa enzymes of the coagulation/hemostasis system. The location of FVIIa and Xa upstream of thrombin in the coagulation cascade suggests cross-talk and sequential trans-activation of PAR₁ and PAR₂ in breast cancer [49]. Our study is novel in demonstrating the significance of PAR₂ C-tail rather than silenced expression. This was shown both in tumor-associated processes as well as in a physiological invasion procedure.

The placenta is characterized by extensive invasion of cytotrophoblasts into the uterus wall, thereby allowing a direct contact of cytotrophoblasts with the maternal blood [30–32]. It is well recognized that trophoblast invasion during the first trimester is a critical step in the establishment of human pregnancy outcome. We as well as others have demonstrated that PAR family members are spatially expressed along the invasive process of early human trophoblast development [50, 51]. While previously we have assessed PAR₁ and PAR₂ activation, β-catenin stabilization and invasion in the EVT organ-culture system [33], we now demonstrate that shRNA hPar2 silencing, following lentiviral infection significantly inhibits both the thrombin-induced and selective TFLLRN PAR₁ activated EVT invasion. Hence, also in a highly regulated, spatial and time-limited physiological invasion process, the presence of hPar2 is essential. The outcome of these findings points out the importance of PAR₂ during placenta implantation to the uterus deciduas. This may suggest that proper anchoring of the placenta to uterus deciduas may require PAR₂.

As the number of reports demonstrating GPCR dimerization has recently increased enormously, it is becoming well established that most GPCRs form either homo- or hetero-oligomers. PARs are no exception to this principle, acting predominantly as heterodimers in several combinations (PAR₁ and PAR₃, PAR₁ and PAR₄, as well as PAR₁ and PAR₂). Other GPCRs, including the chemokine receptors CCR2 and CCR5, have been shown to form both homo- and heterodimers [21], as have CCR2 and CXCR4.

In the present study, we demonstrate that PAR₁ and PAR₂ act together as one functional unit in tumor biology and show that PAR₂ C-tail is important for PAR₁ activities by a mechanism that remains to be elucidated. Our findings suggest that the PAR₁-PAR₂ system will be the focus of future novel therapies in breast cancer treatment.

Electronic supplementary material

Below is the link to the electronic supplementary material.