Thrombomodulin switches signaling and protease-activated receptor 1 cleavage specificity of thrombin

Abstract

Background

Cleavage of the extracellular domain of protease-activated receptor 1 (PAR1) by thrombin at Arg41 and by activated protein C (APC) at Arg46 initiates paradoxical cytopathic and cytoprotective signaling in endothelial cells. In the latter case, the ligand-dependent co-receptor signaling by endothelial protein C receptor (EPCR) is required for the protective PAR1 signaling by APC. Here, we investigated the role of thrombomodulin (TM) in determining the specificity of PAR1 signaling by thrombin.

Methods

We prepared a PAR1 knockout (PAR1^−/−) EA.hy926 endothelial cell line by CRISPR/Cas9 and transduced PAR1^−/− cells with lentivirus vectors expressing PAR1 mutants in which either Arg41 or Arg46 was replaced with an Ala. Furthermore, HEK293 cells were transfected with wildtype or mutant PAR1 cleavage reporter constructs carrying N-terminal NanoLuc luciferase and C-terminal enhanced yellow fluorescent protein tags.

Results

Characterization of transfected cells in signaling and receptor cleavage assays revealed that, upon interaction with TM, thrombin cleaves Arg46 to elicit cytoprotective effects by a β-arrestin-2 biased signaling mechanism. Analysis of functional data and cleavage rates indicated that thrombin-TM cleaves Arg46 >10-fold faster than APC. Upon interaction with thrombin, the cytoplasmic domain of TM recruited both β-arrestin-1 and −2 to the plasma membrane. Thus, the thrombin cleavage of Arg41 was also cytoprotective in TM expressing cells by β-arrestin-1 biased signaling. APC in the absence of EPCR cleaved Arg41 to initiate disruptive signaling responses like thrombin.

Conclusions

These results suggest that co-receptor signaling by TM and EPCR determines the PAR1 cleavage and signaling specificity of thrombin and APC, respectively.

Introduction

Thrombomodulin (TM) is an endothelial cell surface receptor protein that binds thrombin and switches the macromolecular substrate specificity of thrombin from a procoagulant to an anticoagulant protease.^1–4 It binds the basic exosite-1 of thrombin and blocks the interaction of procoagulant substrates with this site on thrombin.^5,6 In addition to blocking the interaction of procoagulant substrates with thrombin, TM also functions as obligatory cofactor to improve the catalytic activity of thrombin, rendering it capable of rapidly activating protein C to activated protein C (APC).^1,2,4 The activation of protein C by the thrombin-TM complex is further accelerated ∼20-fold by its binding to endothelial protein C receptor (EPCR).⁷ Following activation, APC can dissociate from EPCR and function as an anticoagulant protease by binding to the plasma protein cofactor, protein S, to downregulate thrombin generation by proteolytically inactivating factors Va and VIIIa on negatively charged membrane surfaces.⁴ In addition to its anticoagulant activity, APC also possesses anti-inflammatory properties.⁸ To function in the anti-inflammatory pathway, APC remains associated with EPCR, and the protease-receptor complex cleaves the extracellular domain of protease activated receptor 1 (PAR1), initiating cytoprotective signaling responses in endothelial cells.^8–11 The extracellular domain of PAR1 has two distinct cleavage sites which are specifically recognized by either APC (Arg46 site) or thrombin (Arg41 site). While the cleavage of Arg46 site by APC has been shown to initiate anti-inflammatory signaling responses, the cleavage of Arg41 site by thrombin culminates in proinflammatory signaling effects in endothelial cells.^8,12 The mechanisms through which the two proteases elicit paradoxical intracellular signaling responses through cleavage of two distinct cleavage sites have been extensively studied. It has been found that the thrombin cleavage of the Arg41 site leads to coupling of the cytoplasmic domain of PAR1 to members of the heterotrimeric G-proteins.^13,14 however, the cleavage of Arg46 site by APC leads to recruitment of β-arrestin-2 to the plasma membrane and initiation of biased PAR1 signaling.^12,15 Interestingly, we have demonstrated that this mechanism of PAR1 signaling is modulated by the EPCR co-receptor signaling since the occupancy of EPCR by the Gla-domain of protein C/APC also switches thrombin-mediated Arg41 cleavage-dependent signaling specificity of PAR1 from a proinflammatory to an anti-inflammatory response.¹⁶ Further studies have revealed that the EPCR occupancy by the Gla-domain of protein C/APC recruits β-arrestin-2, thereby allowing both thrombin and APC to elicit cytoprotective effects through biased PAR1 signaling in endothelial cells.¹⁷

It is known that a low concentration of thrombin (~50 pM) initiates a PAR1-dependent cytoprotective effect in endothelial cells by an unknown mechanism.¹⁸ Our recent results indicated that the signaling effect of the low concentration of thrombin may be mediated through the protease cleavage of PAR1 upon its interaction with TM since the cytoprotective signaling effect of the protease disappeared when TM was deleted from endothelial cells.¹⁹ In this study, we hypothesize that, like switching the procoagulant specificity of thrombin, TM may also switch the proinflammatory signaling specificity of the protease by enabling thrombin to cleave the Arg46 site of the receptor. To test this hypothesis, we generated PAR1 knockout EA.hy926 endothelial cells (PAR1^−/− cells) by CRISPR/Cas9 technology and transduced them with two mutant PAR1 expressing lentivirus constructs in which either Arg41 or Arg46 was replaced with an Ala. We also transfected HEK293 cells individually with PAR1 cleavage reporter constructs containing similar types of PAR1 mutants, but also carrying N-terminal NanoLuc luciferase and C-terminal enhanced yellow fluorescent protein tags.²⁰ Characterization of PAR1^−/− cells transfected with these constructs in different signaling and receptor cleavage assays support our hypothesis that TM switches signaling and cleavage specificity of thrombin and enables the protease to cleave Arg46 site of PAR1. Remarkably, the results indicate that both the cleavage rate and protective signaling efficiency of thrombin through Arg46 site are at least 10-fold greater than APC.

Materials and Methods

The authors declare that all supporting data are available within the article and in online supplementary file.

Generation of PAR1 knockout endothelial cells and re-expression of PAR1 variants

EA.hy926 cell were maintained in complete Dulbecco’s Modified Eagle’s Medium (DMEM) containing 10% FBS, 100 μg/mL penicillin, 100 μg/mL streptomycin, 1X HAT supplement and 2 mM L-glutamine. PAR1 knockout EA.hy926 endothelial cells (PAR1^−/− cells) were generated by CRISPR/Cas9 technology. In brief, cells were seeded onto 100 mm tissue culture plates and were transfected with the CRISPR/Cas9 knockout plasmid containing puromycin resistance gene (VectorBuilder Inc. Chicago, IL) using Lipofectamine-2000 according to manufacturer’s protocol. The following guide RNAs: CATAAGCATTGACCGGTTTC and CTCAATGAAACCCTGCTCGA were used for the transfection. After 48h, cells were transferred to a complete growth medium containing puromycin (1μg/mL). After 3 weeks, single clone cells were isolated and once confluent were screened for PAR1 expression using phycoerythrin (PE)-conjugated PAR1 antibody (clone ATAP2) and clones negative for PAR1 were selected. The selected clones were subsequently screened for several passages by flow cytometry to confirm absence of PAR1 expression. Two clones (A4 and A6) were eventually selected as stable PAR1^−/− cells. Clone A6 was used for all experiments. For re-expression of PAR1 constructs (PAR1-WT, PAR1-R41A and PAR1-R46A), sub-confluent PAR1^−/− cells (70-80%) were transduced with lentivirus particles carrying PAR1 derivatives as described.¹⁹ After 5 days, cells were selected and maintained in selection medium containing neomycin G418 (400 μg/mL) and puromycin (1ug/mL). TM was over-expressed in cells using recombinant lentivirus particles as previously described.¹⁹

PAR1 expression vectors

Lentivirus-based wildtype PAR1 (PAR1-WT), Arg41 to Ala (PAR1-R41A) or Arg46 to Ala (PAR1-R46A) expression vectors were constructed by VectorBuilder Inc. (Chicago, IL, USA). Cleavage reporter vectors PAR1-WT, Arg41Gly, Arg46Gly, and Arg41Gly/Arg46Gly, all carrying an N-terminal NanoLuc luciferase, and a C-terminal enhanced yellow fluorescent protein (YFP) tags were generous gifts from Dr. Hollenberg (University of Calgary) and prepared as described.²⁰ The cleavage reporter vectors, which all carried a G418 resistant gene, were transfected to sub-confluent wildtype HEK293 cells and HEK293 cells over-expressing either TM or both TM and EPCR (obtained from Dr. Esmon, Oklahoma Medical Research Foundation) HEK293 cells using Lipofectamine-3000 according to the manufacturer’s protocol. The stable G418 resistant clones were identified and sorted by flow cytometry, based on YFP. FACS-sorted cells were expanded and several vials for each clone were prepared and frozen in liquid nitrogen for future use.

PAR1 exodomain cleavage rate measurement

The ability of proteases to cleave the exodomain of PAR1 at different cleavage sites was measured by a luciferase assay employing the Nano-Glo^®Luciferase assay kit (#N1110) as described.²⁰

Permeability assay

Cell permeability was assessed by spectrophotometric measurement of the flux of Evans blue-bound albumin across functional cell monolayer by a modified two-compartment chamber model as described.¹⁶ For siRNA transfection, endothelial cells (1 × 10⁵/well) were plated in trans-well plates (3-μm pore size and 12-mm diameter) and grown overnight. Then the cells were transfected with S1P₁ siRNA (QIAGEN, #SI00376201), β-arrestin-1 siRNA (Dharmacon) (5′-CAUAGAACUUGACACAAAU-3′), β-arrestin-2 siRNA (Dharmacon) (5′-GGACCGCAAAGUGUUUGUG-3′) and control siRNA, (Invitrogen #12935-200) using Lipofectamine-RNAiMAX (#13778075) according to the manufacturer’s instruction. The spectrophotometric absorbance (at 650 nm) of untreated cells (baseline permeability) was taken as a reference response and given a value of 1. The fold change in permeability was calculated based on the ratio of permeability in the presence of the stimuli (protease, TNFα, siRNA, blocking antibody, etc) to the baseline permeability.

Flow cytometry

The expression of cell surface receptors PAR1, TM, and EPCR on different cells were analyzed by flow cytometry as described.¹⁹ Cells were stained using FITC-conjugated anti-ICAM-1 antibody and PE-conjugated anti-E-selectin antibody. PAR1, TM and EPCR, endothelial cells were stained with PE-conjugated PAR1 antibody, PE-conjugated TM antibody and anti-EPCR antibody followed by Alexa Fluor 488 conjugated anti-mouse antibody.

SDS-PAGE and Western blotting

SDS-PAGE and western blotting of cell lysates from confluent cells were analyzed as described.¹⁹ Membrane-associated protein fractions were prepared from endothelial cells as described.²¹

Statistical analysis

Data are presented as mean ± standard error of mean (SEM) from ≥3 independent experiments. Data between two groups were analyzed by Student t-test, and multiple groups were analyzed by one-way Analysis of Variance (ANOVA) followed by Bonferroni multiple comparison test using Graph Pad Prism 8.3.0 (Graph Pad Prism, San Diego, CA).

Results

Generation of PAR1^−/− cells and transfection with PAR1 constructs

To investigate the mechanism and cleavage specificity of PAR1 signaling by APC and thrombin, we prepared PAR1 null EA.hy926 cells (PAR1^−/− cells) by CRISPR/Cas9 technology as described under the Methods’ section and identified a puromycin resistant single clone and characterized it by the FACS analysis using PE-conjugated anti-PAR1 mouse monoclonal antibody (ATAP2) demonstrating that PAR1 has been deleted in this clone (Fig. 1A). This result was confirmed by immunofluorescence analysis (Fig. 1B). PAR1-R41A and PAR1-R46A mutants were re-expressed in PAR1^−/− cells using lentivirus-based expression vectors and the cell surface expression of the mutant receptors was confirmed by the FACS analysis using the same PE-conjugated anti-PAR1 antibody (Fig. 1C). Untransfected normal EA.hy926 cells were used as positive and PAR1^−/− cells were used as negative controls (Fig. 1C). FACS analysis was also carried out to evaluate the cell surface expression of EPCR and TM receptors on different PAR1 mutant cells using specific FITC- and/or PE-conjugated antibodies to each receptor. Results indicated that untransfected PAR1^−/− cells as well as cells transfected with either PAR1-R41A or PAR1-R46A constructs express similar levels of EPCR (Fig. 1D, and Fig. S1A) and significantly higher levels of TM (Fig. 1E, and Fig. S1B) on their surfaces. It is well established that APC elicits a barrier-protective effect in response to pro-inflammatory cytokines through cleavage of Arg46 of PAR1, but thrombin cannot cleave this site, but rather cleaves the canonical Arg41 site of PAR1 to initiate a barrier-disruptive effect in endothelial cells independent of cytokines.^12–14 The characterization of PAR1^−/− cells in the permeability assay induced by TNFα indicated that APC, as expected, exhibits similar barrier protective effects in normal EA.hy926 cells and PAR1^−/− cells transfected with PAR1-R41A but exerts insignificant effect in cells transfected with PAR1-R46A, confirming previous observations that the barrier-protective activity of APC in response to pro-inflammatory stimuli is primarily mediated via cleavage of the Arg46 site (Fig. 1F). As expected, thrombin exhibited a barrier-disruptive effect in both normal EA.hy926 cells and PAR1^−/− cells transfected with PAR1-R46A, but not in PAR1^−/− and PAR1-R41A cells, confirming that the barrier-disruptive signaling function of thrombin is mediated through cleavage of the Arg41 site of PAR1 (Fig. 1G).

Figure 1. Generation and characterization of PAR1^−/− cells expressing PAR1 mutants.

(A) PAR1^−/− EA.hy926 cells were characterized by flow cytometry using PE-conjugated anti-PAR1 mouse monoclonal antibody (ATAP2) as described in Materials and methods. PE-conjugated normal mouse IgG1 was used as an isotype control (n=3). (B) Confluent wildtype and PAR1^−/− cells were fixed permeabilized and stained with ATAP2 followed by Alexa Fluor 568-conjugated goat anti-mouse IgG. The nucleus was stained with DAPI. Immunofluorescence images were obtained with confocal microscopy. Scale bar=20μm (n=3). (C) PAR1^−/− cells were transduced with lentivirus vectors carrying Arg41A and Arg46A mutants of PAR1 and their cell surface expression was analyzed using PE-conjugated ATAP2 (n=3). (D) Cell surface expression of EPCR in EA.hy926, PAR1^−/−, PAR1-R41A, and PAR1-R46A cells were analyzed by flow cytometry using anti-EPCR (JRK1535) mouse monoclonal antibody, followed by FITC conjugated anti-mouse secondary antibody. Normal mouse IgG1 was used as isotype control (n=3). (E) The same as (D), except that the cell surface expression of TM was analyzed using a PE-conjugated anti-TM antibody. PE-conjugated normal mouse IgG1 was used as isotype control (n=3). (F) The permeability of confluent EA.hy926, PAR1-R41A and PAR1-R46A cells in response to TNFα (10 ng/mL for 16h) with or without pretreatment with APC (25nM for 4h) was monitored (n=6). (G) The same as (F), except that cells were treated with thrombin (10nM) for 10 min (n=5). The amount of Evans blue dye, leaked into the lower chamber of the trans-well assay plates, was measured as described in Materials and methods. Data are mean ± SEM. P values were determined by One-way ANOVA, followed by a Bonferroni’s multiple comparison test.

Thrombin in complex with TM elicits cytoprotective signaling through cleavage of Arg46

The characterization of PAR1-R41A cells in a permeability assay induced by TNFα unexpectedly indicated that both APC (25 nM) and thrombin (1 nM) exhibit similar barrier-protective effects (Fig. 2A). To determine the mechanism by which thrombin elicits a protective effect in cells that express a PAR1 mutant that contains only the Arg46 site and lacks the Arg41 site, we expressed the PAR1-R41A construct in the TM^−/− cells in which the TM gene has been deleted by the CRISPR/Cas9 methods as described (19). The results indicated that while APC was able to elicit a barrier-protective effect in response to TNFα in TM^−/− cells transfected with the PAR1-R41A construct, however, thrombin did not initiate a protective effect in these cells (Fig. 2A), suggesting that the protective effect of thrombin is likely mediated through the cleavage of Arg46 when the protease binds to TM on PAR1-R41A transfected cells. The protease concentration dependence of barrier-protective effects of APC (Fig. 2B) and thrombin (Fig. 2C) revealed that the extent of the protective effect of 25 nM APC through cleavage of Arg46 is nearly equal to that of 1 nM thrombin in PAR1-R41A transfected cells. The barrier-protective effect of thrombin through cleavage of the Arg46 site was recapitulated if instead of TNFα other proinflammatory agents including poly I:C (Fig. 2D) or histone H3 (Fig. 2E) were used in the permeability assays. We have previously demonstrated that these two molecules exert barrier-disruptive effects in endothelial cells.^22,23

Figure 2. Thrombin induces a TM-dependent barrier protective effect in PAR1-R41A cells.

(A) Confluent normal EA.hy926 cells (orange) and TM^−/− EA.hy926 cells expressing the PAR1-R41A mutant (violet) were pre-treated with APC (25nM) or thrombin (1nM) for 4h followed by treatment with TNFα (10ng/mL) for 16h. The cell permeability was measured as described in Materials and methods (n≥4). (B) The same as (A) except that the barrier protective effect of increasing concentrations of APC (4h) in response to TNFα (10ng/mL for 16h) in confluent TM^+/+ EA.hy926 cells expressing PAR1-R41A (orange) was monitored (n≥5). (C) The same as (B) except that the concentration dependence of the barrier protective effect of thrombin in confluent PAR1-R41A cells (orange) was monitored (n≥6). (D) The same as (C) except that the barrier protective effect of thrombin (1nM for 4h) in response to Poly(I:C) (10μg/ml) was measured in PAR1-R41A cells (n=5). (E) The same as (D) except that the barrier protective effect of thrombin in response to histone H3 (1μM) was measured (n=5). Data are mean ± SEM. P values were determined by One-way ANOVA, followed by a Bonferroni’s multiple comparison test.

Protective signaling by thrombin-TM through Arg46 requires β-arrestin-2 and crosstalk with S1P₁ receptor

It is known that the EPCR-dependent cytoprotective signaling function of APC through cleavage of Arg46 is mediated through β-arrestin-2 biased PAR1 signaling and requires crosstalk with G_i-protein coupled sphingosine 1-phosphate receptor 1 (S1P₁).^24,25 Moreover, analysis of the cytoprotective signaling mechanism of APC in response to pro-inflammatory stimuli in endothelial cell culture systems requires prior pre-incubation of cells with the protease (~20 nM optimal) for at least three hours. Analysis of the time course of the barrier-protective effect of thrombin (1 nM) through cleavage of Arg46 in response to TNFα indicated that, like APC, the protective effect of thrombin requires prior incubation of PAR1-R41A cells with the protease for at least three hours (Fig. 3A). Like APC, crosstalk with S1P₁ receptor was also required for the barrier-protective effect of thrombin in response to TNFα since the siRNA knockdown of the receptor (Fig. 3B, 3C, and Fig. S2A) and the S1P₁ receptor antagonist W146 (Fig. 3D) both effectively inhibited the barrier-protective activity of both APC and thrombin in PAR1-R41A cells. Moreover, like APC, the barrier-protective effect of thrombin was mediated through the β-arrestin-2 biased PAR1 signaling since the siRNA knockdown of β-arrestin-2, but not β-arrestin-1, abrogated the protective effects of both APC and thrombin in PAR1-R41A cells (Fig. 3E, 3F, and Figs. S2B, and S2C). These results clearly suggest that both thrombin and APC, when in complex with their respective receptors (TM and EPCR) elicit cytoprotective signaling responses in endothelial cells by a similar mechanism through cleavage of the Arg46 site of PAR1. Interestingly, the extent of the protective effect of thrombin (1 nM) is at least 20-fold higher than that of APC (25 nM) in these functional assays (see also Figs. 2B and C).

Figure 3. Mechanism of the barrier-protective effect of thrombin through cleavage of Arg46.

(A) Time course of barrier-protective effect of thrombin (1nM) in PAR1-R41A cells (orange) in response to TNFα (10ng/mL for 16h) was monitored (n≥6). (B) The same as (A) except that the barrier-protective effect of either thrombin (1nM, 4h) or APC (25nM, 4h) in confluent PAR1-R41A cells in response to TNFα (10 ng/mL for 16h) was conducted by the permeability assay in cells transfected with either the control siRNA (orange) or siRNA specific for the S1P₁ receptor (green) (n=4). (C) PAR1-R41A cells transfected with either the control siRNA or the S1P₁ receptor specific siRNA for 48h. Cells were lysed and immunoblotted for S1P₁ receptor. GAPDH was used as loading control (n=3). (D) The same as (B) except that the requirement for crosstalk with the S1P₁ receptor for the barrier-protective effect of either thrombin or APC in PAR1-R41A cells was monitored after cells were pre-treated with the S1P₁ receptor antagonist W146 for 30min (orange) (n=4). (E) Similar to panel (B), except that PAR1-R41A cells were transfected with control siRNA (orange), β-arrestin-1 siRNA (pink), and β-arrestin-2 siRNA (blue) before measuring the permeability (n≥6). (F) PAR1-R41A cells transfected with the control siRNA, β-arrestin-1 siRNA, and β-arrestin-2 siRNA for 48h. Cells were lysed and immunoblotted for β-arrestin-1 and β-arrestin-2. GAPDH was used as loading control (n=3). Data are mean ± SEM. P values were determined by One-way ANOVA, followed by a Bonferroni’s multiple comparison test.

Thrombin-TM inhibits TNFα-mediated CAM expression through cleavage of Arg46

The PAR1 cleavage-dependent inhibitory effect of APC on the cytokine-mediated upregulation of cell adhesion molecules (CAMs) has been shown to contribute to the mechanism of the protective function of APC in endothelial cells.²⁶ This assay was used to determine whether, like APC, thrombin can downregulate TNFα-mediated upregulation of CAMs in PAR1-R41A cells. The flow cytometry results indicated that thrombin effectively downregulates the expression of ICAM-1 (Fig. 4A, and Fig. S3A) and E-selectin (Fig. 4B, and Fig. S3B) through cleavage of Arg46 in PAR1-R41A transfected endothelial cells. Moreover, western-blot analysis of cell lysates revealed that thrombin also inhibits TNFα-mediated phosphorylation of VE-cadherin and activation of the NF-κB pathway through cleavage of Arg46 in PAR1-R41A cells (Fig. 4C, and Figs. S4A–C), accounting for the barrier-protective effect of thrombin in the permeability assays. Unlike in PAR1-R41A cells, thrombin exhibited no protective effect in TNFα-stimulated PAR1-R46A cells in which Arg46 has been mutated (Figs. 4D–F, and Figs S3C,D, and Figs. S4D–F). To confirm that the protective effect of thrombin through cleavage of Arg46 requires the interaction of the protease with TM, the flow cytometry experiments were also repeated with the TM^−/− cells, which were transfected with the PAR1-R41A construct. Thrombin exhibited no protective effect in downregulating TNFα-mediated expression of ICAM-1 (Fig. 5A, and Fig. S3E), E-selectin (Fig. 5B, and Fig. S3F) and VE-cadherin/NF-κB p65 phosphorylation (Fig. 5C, and Figs. S5A–C) in the absence of TM, suggesting that like APC, which requires EPCR for its protective signaling function through cleavage of Arg46, thrombin requires TM for its protective function by the same mechanism. Thus, both APC and thrombin exert protective effects through cleavage of Arg46 only when they are bound to their specific receptors (EPCR and TM, respectively). Indeed, like thrombin, APC in the absence of APC-EPCR interaction cleaved Arg41 to elicit barrier-disruptive effect in PAR1-R46A cells. This was evidenced by the observations that a Gla-domainless APC (aGDPC), which cannot bind EPCR, cleaved Arg41 and disrupted the barrier permeability function of PAR1-R46A cells (Fig. 5D). This result was further confirmed in the presence of a function-blocking anti-EPCR antibody that blocks the interaction of Gla-domain of APC with EPCR. In the presence of this antibody, APC also induced a barrier-disruptive effect in PAR1-R46A cells (Fig. 5E). The APC-EPCR interaction is indispensable for APC-mediated Arg46 cleavage and its subsequent barrier protective effects, as the inhibition of the APC-EPCR interaction by the anti-EPCR antibody resulted in abrogation of APC-mediated protective effects in PAR1-R41A cells (Fig. 5F). The APC cleavage of Arg41 is required to induce permeability since APC failed to induce permeability in PAR1-R41A cells (10 minutes incubation) in presence of EPCR-blocking antibody (Fig. 5G).

Figure 4. Thrombin downregulates TNFα-mediated upregulation of cell adhesion molecules in PAR1-R41A but not in PAR1-R46A cells.

Confluent PAR1-R41A cells were pre-treated with thrombin (5nM for 4h) followed by treatment with TNFα (10ng/mL) for 4h. The cell surface expression of ICAM-1 (A) and E-selectin (B) were measured by the flow cytometry (n=5). (C) Confluent PAR1-R41A cells were pre-treated with thrombin (5nM, 4h) followed by treatment with TNFα (10ng/mL) for 15min. Cells were lysed, immunoblotted for VE-Cadherin (phosphorylated at Tyr658 and the total), NF-κB p65 (phosphorylated at Ser536 and the total). GAPDH was used as loading control (n=3). (D, E) The same as (A, B) expect that PAR1-R46A cells were used for measuring the effect of thrombin in TNFα-mediated expression of ICAM-1and E-selectin (n=5). (F) The same as (C) expect that immunoblots for phosphorylation of VE-cadherin and NF-κB p65 were monitored in PAR1-R46A cells. GAPDH was used as loading control (n=3). Quantitation of the flow cytometry data are presented in Supplementary Figs. 3A–D.

Figure 5. Receptor-dependent anti-inflammatory effects of thrombin and APC.

Confluent TM^−/− endothelial cells expressing the R41A mutant of PAR1 (PAR1-R41A-TM^−/−) were pre-treated with thrombin (5nM for 4h) followed by treatment with TNFα (10ng/ml) for 4h. The cell surface expression of ICAM-1 (A) and E-selectin (B) were measured by the flow cytometry (n≥4). Quantitation of the flow cytometry data are presented in Supplementary Fig. 3E and F. (C) Confluent PAR1-R41A-TM^−/− cells were pre-treated with thrombin (5nM, 4h) followed by treatment with TNFα (10ng/mL) for 15min. Cells were lysed, immunoblotted for VE-cadherin (phosphorylated at Tyr658 and the total), NF-κB p65 (phosphorylated at Ser536 and the total). GAPDH was used as loading control (n=3). (D) Confluent PAR1-R46A cells (blue) were treated with increasing concentrations of activated Gla-domain less protein C (aGDPC) for 10min followed by measuring permeability. (E) The same as (D) except that PAR1-R46A cells were first pre-treated with anti-EPCR blocking antibody or control IgG for 30min followed by incubation with increasing concentrations of APC for 10min. (F) Confluent PAR1-R41A cells (orange) were pre-treated with anti-EPCR blocking antibody or control IgG for 30min followed by incubation with APC (25nM) for 4h and treatment with TNFα (10ng/mL) for 16h (n=4). (G) The same as (E) except that confluent PAR1-R41A cells (orange) were first pre-treated with control IgG or anti-EPCR blocking antibody for 30min followed by incubation with 250 nM APC for 10min (n=4). The permeability was calculated as described in Materials and methods. Data are mean ± SEM. P values were determined by One-way ANOVA, followed by a Bonferroni’s multiple comparison test.

It should be noted that, unlike the cytoprotective function of APC and thrombin through cleavage of Arg46, which is dependent on β-arrestin-2 biased signaling and crosstalk with the S1P₁ receptor, the barrier-disruptive function of thrombin through cleavage of Arg41 is direct, rapid and mediated through the cleavage-dependent coupling of the PAR1 cytoplasmic domain to the G-proteins (G_12/13), thereby leading to rapid activation of RhoA GTPase, phosphorylation of the myosin light chain, stress fiber formation and increased barrier permeability.^13,14,27 The permeability studies presented in Figs. 5D&E in PAR1-R46A cells have all been conducted by treating them with the proteases for 10 min. Thus, like thrombin, free APC can only cleave Arg41 to elicit a barrier-disruptive response in endothelial cells, however, when the proteases bind to their specific receptors (TM and EPCR) they both initiate cytoprotective responses through cleavage of Arg46. It should be noted that when EPCR is occupied by the Gla-domain of APC/protein C, the cleavage of Arg41 by either thrombin or APC is cytoprotective by β-arrestin-2 biased signaling.¹⁷

TM switches Arg41-cleavage dependent signaling specificity of thrombin from a barrier-disruptive to a barrier-protective effect by recruiting β-arrestins to the plasma membrane

To determine the effect of TM on the barrier-permeability function of endothelial cells through cleavage of Arg41, we compared the signaling effect of thrombin in PAR1-R46A cells with or without transduction with the lentivirus vector carrying the TM construct. As expected, a short exposure (10 min) of cells to thrombin resulted in a concentration-dependent barrier-disruptive effect through cleavage of Arg41 in PAR1-R46A cells (Fig. 6A). However, after transduction with TM, cells were resistant to barrier-disruptive function of thrombin even at high concentrations (Fig. 6A), suggesting that TM abrogates the proinflammatory function of thrombin in PAR1-R46A-TM^high cells. This result is consistent with our previous study showing that thrombin does not induce a barrier-disruptive effect in cells expressing ~10-fold higher TM.¹⁹ Interestingly, thrombin effectively induced a barrier-disruptive effect in PAR1-R46A cells overexpressing a TM construct in which the cytoplasmic domain of the receptor was deleted (PAR1-R46A-TM-des-Cyto^high), suggesting a role for the cytoplasmic domain of TM in changing the Arg41 cleavage-dependent specificity of PAR1 signaling by thrombin (Fig. 6A). A high expression level for both wildtype TM and TM-des-Cyto^high on PAR1-R46A cells were confirmed through flow cytometry, showing higher cell surface expression of TM on these cells (Fig. S6A) that is consistent with our previous results.¹⁹ To test the hypothesis that interaction with thrombin results in TM recruiting β-arrestins to the plasma membrane and that the inability of TM-des-Cyto to recruit β-arrestins is responsible for the barrier-disruptive effect of thrombin in PAR1-R46A-TM-des-Cyto^high cells, we conducted the thrombin-induced permeability assay in PAR1-R46A-TM^high cells following the siRNA knockdown of either β-arrestin-1 or β-arrestin-2. In support of our hypothesis, similar to PAR1-R46A-TM-des-Cyto^high cells, thrombin induced a barrier-disruptive effect in cells overexpressing wildtype TM if β-arrestin-1, but not β-arrestin-2 expression was knocked down (Fig. 6B), indicating that the cytoplasmic domain of TM is involved in recruiting β-arrestin-1 to the plasma membrane, thereby inhibiting the coupling of PAR1 to G-proteins after Arg41 is cleaved by thrombin. Although it was not statistically significant, the siRNA knockdown of β-arrestin-2 also enhanced the barrier-permeability effect of thrombin in PAR1-R46A-TM^high cells, suggesting that TM recruits both β-arrestins for biased signaling which is consistent with the siRNA knockdown of β-arrestin-2 inhibiting the TM-dependent barrier-protective effect of Arg46 cleavage by thrombin as shown in Fig. 3E. Further studies revealed that TM endows a barrier-protective effect for thrombin in TNFα-stimulated PAR1-R46A-TM^high cells via cleavage of Arg41 by biased β-arrestin-1 signaling mechanism (Fig. 6C), and unlike the biased protective signaling through Arg46 which is mediated by β-arrestin-2, the TM cytoplasmic domain-dependent protective signaling function of thrombin through cleavage of Arg41 is mediated by β-arrestin-1 biased signaling. Consistent with an important role for the cytoplasmic domain of TM in modulating the thrombin-dependent signaling specificity of PAR1 through both cleavage sites, thrombin did not induce a barrier-protective effect in either PAR1-R46A-TM-des-Cyto^high cells (Fig. 6D) or in PAR1-R41A-TM-des-Cyto^high cells (Fig. 6E). In further support of these results, western blot analysis of membrane fractions derived from thrombin stimulated PAR1-R41A cells and PAR1-R41A-TM-des-Cyto^high cells indicated that only thrombin stimulation of the former cells results in enhanced β-arrestin-2 recruitment to the plasma membrane (Fig. 6F). Similarly, analysis of membrane fractions derived from PAR1-R46A-TM^high and PAR1-R46A-TM-des-Cyto^high cells indicated that only thrombin cleavage of the Arg41 site in the former cells leads to enhanced β-arrestin-1 recruitment to the plasma membrane (Fig. 6F). An antibody to the C-terminal cytoplasmic domain of TM was used to validate the expression of wildtype and TM-des-Cyto constructs on transfected cells. The 3 repeats for Fig. 6F together with their quantitation are presented in Fig. S7.

Figure 6. Cytoplasmic domain of TM is required for the PAR1-dependent barrier-protective effect of thrombin.

(A) Confluent PAR1-R46A cells (blue), PAR1-R46A cells overexpressing either TM (TM^high) (light blue) or TM-des-Cyto (TM-des-Cyto^high) (green) were treated with increasing concentrations of thrombin for 10min followed by measuring the permeability (n=4). (B) The barrier permeability function of PAR1-R46A overexpressing TM (TM^high) in response to thrombin (1nM for 10min) was monitored after transfecting cells with the control siRNA (blue), β-arrestin-1 siRNA (pink), or β-arrestin-2 siRNA (light blue) for 48h as described in Materials and methods (n=5). (C) The barrier-protective effect of thrombin (1 nM for 4h) in response to TNFα (10ng/mL for 16h) in PAR1-R46A cells overexpressing TM (TM^high) was monitored after transfecting cells with control siRNA (blue), β-arrestin-1 siRNA (pink), or β-arrestin-2 siRNA (light blue) as described in Materials and methods (n=5). (D) The same as (C) except that the barrier-protective effect of thrombin in response to TNFα was monitored in PAR1-R46A cells overexpressing either TM (TM^high) (blue) or TM-des-Cyto (TM-des-Cyto^high) (green) (n=6). (E) The same as (D) except that the barrier-protective effect of either thrombin or APC in response to TNFα was monitored in either normal PAR1-R41A cells (orange, not transfected by TM) or cells overexpressing TM-des-Cyto (TM-des-Cyto^high) (magenta) (n=5). The amount of Evans blue dye that leaked into the lower chamber in the trans-well assay plates was measured and represented as fold change over control as described in Materials and methods. (F) Western blot analysis of membrane fractions of unstimulated and thrombin (1 nM) stimulated PAR1-R41A cells, PAR1-R41A-TM-des-Cyto^high cells, PAR1-R46A-TM^high and PAR1-R46A-TM-des-Cyto^high cells. TM-C-terminal Ab is an antibody that recognizes the C-terminal domain of TM (n=3). Data are mean ± SEM. P values were determined by One-way ANOVA, followed by a Bonferroni’s multiple comparison test.

It is known that thrombin at the low concentration of 50 pM elicits a PAR1-dependent partial cytoprotective effect in EA.hy926 cells and we demonstrated that this effect of thrombin is TM-dependent.¹⁹ To determine whether this effect of thrombin is mediated through cleavage of Arg41 or Arg46, we compared this effect of thrombin in PAR1-R46A and PAR1-R41A cells. Results indicated that the TM-dependent protective effect of 50 pM thrombin is primarily mediated through cleavage of Arg41 (Fig. S8A). Relative to microvascular endothelial cell, EA.hy926 cells express a lower cell surface TM level, thus we analyzed the protective effect of thrombin in primary human pulmonary microvascular endothelial cells (HPMEC). Time course analysis indicated that thrombin (1 nM) elicits a protective effect in HPMEC at both 4 hours and 16 hours incubation times tested (Fig. S8B).

PAR1 cleavage assay

To investigate the cleavage specificity of PAR1 by thrombin when the protease is in free form or in complex with TM, we monitored the ability of thrombin to cleave the N-terminus peptide of PAR1 at either Arg41 or Arg46 site using HEK293 cells stably expressing different PAR1 cleavage reporter constructs tagged with an N-terminus NanoLuc luciferase. As expected, increasing concentrations of thrombin effectively cleaved PAR1-WT and PAR1-R46G in HEK293 cells transfected with these constructs, however, thrombin exhibited no detectable activity toward cleavage of PAR1 in HEK293 cells expressing either PAR1-R41G or the double-mutant PAR1-R41G/R46G (Fig. 7A). HEK293 cells do not express TM or EPCR (28), and the APC concentration-dependence of cleavage rates in HEK293 cells transfected with the two reporter constructs, PAR1-R46G and PAR1-R41G, indicated that, like thrombin, APC in the absence of EPCR cleaves Arg41, but exhibits insignificant activity toward Arg46 (Fig. 7B), supporting the functional data (Fig. 5D and E) that APC cleaves Arg41 when its interaction with EPCR is lost. We then transfected the PAR1-R41G reporter construct to HEK293 cells overexpressing both TM and EPCR and monitored the cleavage rate of Arg46 site by both thrombin and APC when these proteases are bound to their specific cofactors. The protease concentration-dependence of cleavage rates indicated that thrombin in complex with TM cleaves the Arg46 site of PAR1 ~10-fold faster than the APC-EPCR complex (Fig. 7C, Figs. S6B, S6C), accounting for a similar extent of higher cytoprotective effect for thrombin in comparison to APC in the functional assays (Fig. 2B and C). On the other hand, in the absence of EPCR, APC failed to cleave Arg46 (Fig. S9A). In absence of TM and EPCR, both thrombin and APC cleaved Arg41 and the rate of cleavage by thrombin at this site was ~1000-fold higher than that of APC (Fig. S9B). These results suggest that, like APC, which requires its cofactor EPCR to cleave Arg46, thrombin also requires its cofactor TM to cleave Arg46 to initiate cytoprotective effects in endothelial cells.

Figure 7. The efficiency of PAR1 cleavage by thrombin and APC in the absence and presence of receptors.

(A) Stable HEK293 cells expressing wildtype and mutant PAR1 reporter constructs were incubated with increasing concentrations of thrombin and the cleavage of Arg41 or Arg46 in different constructs (, PAR1-WT; , PAR1-R41G; , PAR1-R46G and , PAR1-R41/46G) was monitored. The rate of cleavage of the PAR1 N-terminus was measured from the release of NanoLuc luciferase into the cell supernatant as described in Materials and methods. The activity of 10nM thrombin toward PAR1-WT is presented as 100%. (B) The same as (A) except that the cleavage of PAR1 mutants (), PAR1-R41G and , PAR1-R46G) by increasing concentrations of APC was monitored. (C) The same as (A) except that the cleavage of Arg46 in the PAR1-R41G reporter plasmid by increasing concentrations of either thrombin () or APC () was monitored in HEK293 cells overexpressing both TM and EPCR.

Discussion

In this study, we investigated the signaling and cell surface cleavage specificity of PAR1 by thrombin and discovered that interaction with TM endows an APC-like specificity for thrombin in endothelial cells. When in complex with TM, like APC, thrombin activates PAR1 through cleavage of Arg46 to elicit cytoprotective signaling responses in endothelial cells. Remarkably, based on both cleavage rate and functional data, the catalytic activity of thrombin toward cleavage of Arg46 is markedly (>10-fold) higher than that of APC. It was interesting to note that APC, like thrombin, did not cleave Arg46 of PAR1 on cells lacking EPCR, but cleaved Arg41 to elicit a barrier-disruptive effect. These results suggest that between the two PAR1 cleavage sites, Arg41 is the preferred cleavage site for both proteases when they are free and not bound to their cell surface receptors. However, when APC and thrombin bind to their specific cell surface receptors, they become catalytically competent of cleaving Arg46 to elicit cytoprotective signaling effects. This mechanism of cytoprotective PAR1 signaling was found to be similar for both APC and thrombin when they are in complex with their receptors, thus requiring both β-arrestin-2 biased signaling and crosstalk with the S1P₁ receptor. The mechanism by which interaction with receptors changes the specificity of thrombin and APC so that these proteases cleave the Arg46 site of PAR1 is not known. However, based on the molecular basis of coagulation protease specificity, the P3-P3’ residues of substrates and inhibitors play key roles in specificity determination of coagulation proteases.^29,30 Thus, we postulate that unlike the favorable P2-Pro residue of the P1-Arg41 cleavage site of PAR1,^29,30 the P2-Leu residue of the exodomain of PAR1 on which P1-Arg46 cleavage site is located is a poor recognition site for both thrombin and APC and therefore neither protease can accommodate this residue in their active-site pockets to effectively cleave the P1-Arg46 peptide bond. We hypothesize that interaction with TM and EPCR positions the catalytic domain of APC and thrombin at a proper distance and orientation on the membrane surface, to facilitate the docking of the Arg46 site into the catalytic pocket of these proteases. In support of this hypothesis, it has been demonstrated that when exosite-1 of thrombin binds to epidermal growth factor-like domains 5 and 6 of TM,⁵ the active-site of thrombin is located 66 Å above the membrane,³¹ implicating that the exodomain of PAR1 is an extended conformation and Arg46 is located at similar distance above the cell surface in the vicinity of the TM-thrombin complex. A similar upright conformation has also been demonstrated for the membrane-bound active-site pocket of APC.³² It should also be noted that the receptor-dependent cleavage of PAR1 by thrombin (also by APC) is facilitated by their colocalization of all three receptors PAR1, TM and EPCR in the lipid-raft microenvironments of endothelial cells.³³ Thus, these receptors function as cofactors by a mechanism that can be referred to as “substrate presentation” to enable these proteases to cleave Arg46 site of PAR1 on the surface of endothelial cells (Fig. 8 model, shown for TM only). This is a recurring theme in the protease-substrate recognition mechanism in the coagulation field since coagulation proteases exhibit no significant reactivity toward target physiological substrates primarily because they cannot recognize and interact with them in the absence of cofactors.³⁴ The hirudin-like site of PAR1 plays a similar role for the thrombin cleavage of the Arg41 site; in this case however, the cofactor function is mediated by an intramolecular binding site on the PAR1 receptor itself (Fig. 8) that binds to exosite-1 of thrombin to promote the rate of cleavage reaction. This mechanism of cofactor function is reminiscent of the activation of procofactors V and VIII by thrombin through exosite dependent interaction of the protease with complementary binding sites on the procoagulant cofactors.^34,35

Figure 8. Hypothetical model of PAR1 cleavage by thrombin.

The residues surrounding the P1-Arg46 cleavage site (i.e., P2-Leu) of PAR1 are poor recognition sites for both thrombin and APC. We hypothesize that TM by maintaining the catalytic domain of thrombin at a proper distance and orientation above the membrane surface facilitates the docking of Arg46 into the active-site pocket of the protease by a mechanism referred to as substrate presentation. The hirudin-like site of PAR1 plays a similar role to facilitate the thrombin cleavage of Arg41 site. In the presence of high and saturating concentration of TM (right cartoon), the thrombin cleavage of PAR1 at either Arg41 or Arg46 site initiates a barrier protective effect (tested by a permeability assay requiring 3 hours incubation with thrombin) in response to proinflammatory cytokines by β-arrestin-1 (Arg41) and β-arrestin-2 (Arg46) biased signaling that leads to activation of the Rac1 small GTPase.⁴⁷ In the absence of TM (left cartoon), thrombin by interaction with the hirudin-like binding site of PAR1 rapidly activates PAR1 and induces a barrier disruptive effect (<10min time scale) through coupling of PAR cytoplasmic domain to a G-protein (i.e., G_12/13) that leads to activation of RhoA GTPase.^13,14,27 See the text for more details. The image was created with BioRender.com.

Another interesting finding in this study is the observation that thrombin did not induce a barrier-disruptive effect but exhibited a cytoprotective effect through cleavage of Arg41 in the PAR1-R46A cells overexpressing TM. The mechanism by which TM switched the signaling specificity of thrombin from a barrier-disruptive effect to a barrier-protective effect through cleavage of Arg41 was investigated and determined to be mediated through the cytoplasmic domain of TM recruiting β-arrestins to the plasma membrane, thereby inducing biased PAR1 signaling rather than coupling of PAR1 to G-proteins. This was evidenced by the observations that the protective effect of thrombin was eliminated in PAR1-R46A cells overexpressing TM-des-Cyto as well as siRNA knockdown of both β-arrestin-1 and 2 in these cells, though β-arrestin-1 was more effective. Interestingly, the siRNA knockdown of β-arrestin-2, but not β-arrestin-1 also inhibited the protective effect of thrombin through cleavage of Arg46 in PAR1-R41A, suggesting that cleavage of Arg46 is primarily associated with recruitment of β-arrestin-2 which is also observed with the EPCR-dependent cleavage of PAR1 by APC.^12,15 The protective biased PAR1 signaling by the factor VIIa-EPCR complex through cleavage of Arg41 is mediated through β-arrestin-1,³⁶ possibly suggesting that the conformational changes in the PAR1 structure that are induced by the cleavage of Arg46 and Arg41 sites are distinct so that the PAR1 C-terminal domain preferentially binds either β-arrestin-1 or β-arrestin-2 for the biased signaling. The mechanism by which the cytoplasmic domain of TM recruits β-arrestins to the plasma membrane was not investigated. However, in a recent study we demonstrated that the cytoplasmic domain of TM is involved in recruiting the tumor suppressor phosphatase and tensin homolog (PTEN) to the plasma membrane in endothelial cells.³⁷ Interestingly, it has also been found that the phosphatase activity of PTEN is modulated by β-arrestins binding directly to the C-terminal domain of PTEN.³⁸ Thus, upon stimulation, PTEN and β-arrestins are co-recruited to the plasma membrane. It appears that the cytoplasmic domain of TM recruits both PTEN and β-arrestins to the plasma membrane to regulate the PAR1-dependent signaling specificity of thrombin. The PAR1-dependent cytoprotective signaling function of the thrombin-TM complex can play a key physiological role in regulating inflammatory responses in microvasculature.

Previous studies have indicated that EPCR is abundantly expressed on large vessels and no significant expression of the receptor is detected in micro vessels such as lung capillaries.²⁸ On the other hand, because of the dramatically higher ratio of endothelial cell surface to blood volume in microvasculature, the effective local concentration of TM can reach as high as 500 nM.^39–41 The novel discovery in this manuscript provides insight into the mechanism by which thrombin through interaction with TM can effectively regulate procoagulant and proinflammatory pathways in microcirculation. The TM-thrombin complex under these conditions not only rapidly activates protein C independent of EPCR, but also induces a direct cytoprotective effect through biased activation of PAR1 on endothelial cells. Noting that the interaction of APC with EPCR is required for its cytoprotective function but that the receptor inhibits the anticoagulant activity of the protease, our results strongly indicate that the PAR1-dependent cytoprotective signaling in microcirculation is primarily mediated by thrombin and APC primarily exerts an anticoagulant function through interaction with protein S and degradation of procoagulant cofactors, Va and VIIIa.⁴ Thus, the thrombin-TM complex and the APC-protein S complex collaboratively protect microvasculature from the pathological effects of immunothrombosis which is mediated by activation of coagulation and inflammation in microcirculation. Through activation of the anticoagulant protein C and cleavage of Arg46 of PAR1, thrombin can play an important role in downregulating immunothrombosis, which has been identified as a key factor in inducing endothelial dysfunction and organ failure in severe inflammatory disorders including severe sepsis and coronavirus disease 2019 infection.^42–44 Thus, under pathological conditions if there is a substantial decrease in the cell surface level of TM and an increase in the local concentration of thrombin, unbound free thrombin in microcirculation may exert a detrimental effect through cleavage of Arg41 of PAR1 at vascular sites where TM is absent. It should be noted that in addition to inhibiting coagulation, APC through EPCR-independent cleavage and inhibition of histones H3 and H4,⁴⁵ and through macrophage antigen-1 (Mac-1, CD11b/CD18)-dependent activation of PAR1 on macrophages and neutrophils,⁴⁶ can also contribute to downregulation of immunothrombosis in microcirculation.

Highlights:

1. The cleavage of PAR1 by thrombin at Arg41 and by APC at Arg46 elicits cytopathic and cytoprotective signaling, respectively.

2. We discovered that when in complex with TM, thrombin cleaves Arg46 of PAR1 to induce cytoprotective signaling.

3. Like APC, the cytoprotective effect of thrombin through Arg46 requires β-arrestin-2 biased signaling and crosstalk with S1P₁ receptor.

4. The cleavage and cytoprotective signaling efficiency of thrombin in the presence of TM is >10-fold higher than APC.

Nonstandard Abbreviations and Acronyms

TM

Thrombomodulin

APC

Activated protein C

PAR1

Protease-activated receptor 1

EPCR

Endothelial protein C receptor

YFP

Yellow fluorescent protein

HEK293

Human Embryonic Kidney 293

Nluc

NanoLuc luciferase

BSA

Bovine serum albumin

FACS

Fluorescence-activated cell sorting

SEM

Standard error of mean

ANOVA

Analysis of variance

S1P1

sphingosine 1-phosphate receptor 1