The Domino Effect Triggered by the Tethered Ligand of the Protease Activated Receptors

Abstract

Protease activated receptors (PARs) are G-protein coupled receptors (GPCRs) that have a unique activation mechanism. Unlike other GPCRs that can be activated by free ligands, under physiological conditions, PARs are activated by the tethered ligand, which is a part of their N-terminus that is unmasked by proteolysis. It has been 30 years since the first member of the family, PAR1, was identified. In this review, we will discuss this unique tethered ligand mediate receptor activation of PARs in detail: how they interact with the proteases, the complex structural rearrangement of the receptors upon activation, and the termination of the signaling. We also summarize the structural studies of the PARs and how single nucleotide polymorphisms impact the receptor reactivity. Finally, we review the current strategies for inhibiting PAR function with therapeutic targets for anti-thrombosis. The focus of this review is PAR1 and PAR4 as they are the thrombin signal mediators on human platelets and therapeutics targets. We also include the structural studies of PAR2 as it informs the mechanism of action for PARs in general.

1. Introduction

Protease activated receptors (PARs) are the key mediators that link extracellular proteases to the intracellular signaling machinery. PARs are activated by cleavage of the N-terminus to generate a tethered ligand that intramolecularly binds to the ligand binding site (Figure 1). PARs are expressed on the surface of many cell types, including most cells that are associated with the blood vessel wall (endothelial cell, fibroblasts, myocytes) and the blood (platelets, neutrophils, macrophages).(1–5) Combining this expression profile and the essential role of proteases in many physiological processes, PARs have pivotal roles in coagulation, inflammation and cancer. In this review, we provide a brief historical context and highlight the most recent developments in PAR signaling related to alternative cleavage sites and biased signaling, sequence variants, and lessons learned from structural studies to inform strategies for therapeutics targeting PARs. This review focuses on PAR1 and PAR4. We also draw on a limited number of PAR2 studies to highlight general properties of this receptor family.

Figure 1. Tethered ligand mediated receptor activation mechanism of PARs.

PARs can be activated by enzymatic cleavage of their N-terminus by proteases. The newly exposed N-terminus serves as a tethered ligand to activate the receptor. PARs are the key mediators that relay the extracellular protease signal across the cell membrane.

2. PAR activation by proteases

PAR1 was originally identified as the “thrombin receptor” on platelets and other cells.(6) This led to the discovery of a family of receptors, PAR1–4.(7–10) It is now recognized that PARs can be cleaved by multiple proteases depending on the membrane environment and cofactors that are present. For PAR1 in particular, these non-canonical cleavage sites generate various tethered ligands, which can alter the cellular responses compared to thrombin-mediated PAR1 activation (Figure 2).(11–13) For example, canonical PAR1 signaling mediated by thrombin increases the permeability of endothelial cells. Non-canonical PAR1 signaling mediated by APC leads to endothelial barrier stabilization and cyto-protection. Functional selectivity of a receptor by distinct ligands is known as biased signaling in GPCRs.(14–16) Biased signaling from alternative cleavage sites provides a mechanism for context dependent signaling from PARs. Biased signaling downstream of PARs can also be mediated by posttranslational modification, cofactors, or pharmacologically.(11, 17–21)

Figure 2. The panel of activating proteases and their cleavage sites on PAR1 and PAR4.

Both PAR1 (top) and PAR4 (bottom) can be activated by multiple proteases. Different proteases cleave PAR1 at distinct sites, which generates unique tethered ligands to initiate both canonical and non-canonical signaling pathways. In contrast, all of the known proteases that activate PAR4 do so and the same cleavage site on its N-terminus (marked as a black triangle).

2.1. Thrombin

The interaction between thrombin and PAR1 is the most well understood PAR activation mechanism.(22) Thrombin specifically recognizes and cleaves the N-terminus of PAR1 at the Arg⁴¹. The newly exposed N-terminus (S⁴²FLLRN…) serves as a tethered ligand that interacts with a conserved sequence in extracellular loop 2 (ECL2) (Figure 3).(23, 24) No co-factor is needed to facilitate thrombin-mediated PAR1 activation due to the high specificity and affinity between the hirudin-like sequence of PAR1 and exosite I of thrombin (Figure 3).(6, 25, 26)

Figure 3. PAR1 and PAR4 interact with thrombin differently.

PAR1 has a longer N-terminus compared to PAR4 (81 amino acids versus 65 amino acids). Thrombin cleavage shortens the PAR1 (top left) exodomain by 20 amino acids to uncover the tethered ligand sequence SFLLRN, and the remaining N-terminus of PAR1 after activation is 61-amino acids. In contrast, thrombin cuts 30 amino acids off from PAR4 (top right) exodomain to expose the tethered ligand GYPGQV, and the remaining N-terminus is 35 amino acids. Thrombin interacts with PAR1 and PAR4 by different mechanisms. PAR1(bottom left) contains a hirudin-like sequence that interacts with exosite I of the thrombin and locks the enzyme in its active conformation. As a result, PAR1 is an excellent substrate that can be activated by sub-nanomolar concentrations of thrombin. PAR4 (bottom right) does not have a hirudin sequence and, as a result, is a less efficient thrombin substrate that requires >10-fold more thrombin than PAR1. The anionic region of PAR4 interacts with the autolysis loop of thrombin and contributes to PAR4 activation.

PAR4 does not have a hirudin-like sequence and primarily interacts with thrombin via the cleavage site at Arg⁴⁷(Figure 3).(27–29) PAR4 also uses an anionic cluster (D⁵⁷D⁵⁹E⁶²D⁶⁵) in the exodomain that interacts with thrombin’s autolysis loop to decrease the disassociation rate and facilitate the cleavage reaction (Figure 3).(30–32) Since PAR4 does not interact with thrombin’s exosite I, it likely does not modulate thrombin allosterically.(31–33) A recent study showed that thrombin-mediated PAR4 activation can be disrupted by blocking thrombin’s exosite II indicating a potential interaction between exosite II and PAR4.(34) The net result is that PAR4 is a less efficient thrombin substrate than PAR1.(27, 28) On cells, this is overcome by cofactors such as PAR1 or PAR3 that decrease the amount of thrombin required for PAR4 activation by 6–10 fold. (30, 35–38)

2.2. Activated Protein C (APC)

Activation of PAR1 by activated protein C (APC) was first reported in 2002 on endothelial cells.(39) APC has two cleavage sites on PAR1, Arg⁴¹ and Arg⁴⁶; the latter is the preferential APC site (Figure 2).(40, 41) APC-mediated PAR1 cleavage has a significantly lower catalytic efficiency compared to thrombin.(42) This raised an important question. Thrombin, along with thrombomodulin (TM), activates protein C (PC) to APC.(43) Given that thrombin cleaves PAR1 with a catalytic efficiency of nearly 3 orders of magnitude higher than APC (42), how can APC cleave PAR1 before thrombin does to elicit the cytoprotective response? Several studies have now mechanistically addressed this question. Under physiological conditions, the endothelial protein C receptor (EPCR) is an essential co-factor to assist the cleavage of PAR1 by APC (Figure 4).(39) These four proteins (thrombin, PC, TM and GPCR) colocalize with PAR1 in the lipid-raft/caveolae microenvironment on the surface of endothelial cells (Figure 4).(44) TM-bound thrombin does not induce a PAR1-mediated proinflammatory response.(45) Taken together, TM-bound thrombin-mediated PC activation and APC-EPCR mediated PAR1 biased signaling are two connected events, which requires a specific microenvironment location on the cell surface (Figure 4).(46, 47) In addition, a report by Schuepbach et al demonstrated in the presence of up to 1 nM thrombin, APC was able to mediate additional PAR1 activation.(48) Further, the internalization rate of APC-cleaved PAR1 is significantly lower than the thrombin-activated PAR1. This altered intracellular cellular trafficking of APC-activated PAR1 may also explain how PAR1 signaling by an inefficient protease (APC) can have physiologically relevant consequences in the presence of thrombin.(48) Griffin and colleagues have recently generated mice with PAR1 mutations at Arg⁴¹, Arg⁴⁶, or both.(49) These mice demonstrate the physiological importance of the alternative cleavage sites on PAR1 in vivo and will be important tools to examine thrombin and APC signaling in disease specific contexts.

Figure 4. APC mediates biased signaling of PAR1.

Thrombin cleaves PAR1 at the R⁴¹-S⁴² site, which generates the tethered ligand ⁴²SFLLRN and primarily initiates G-protein signaling pathways. This canonical pathway in endothelial cells leads to proinflammatory signaling and barrier disruption and increased cell permeability. In contrast, activated by thrombomodulin (TM)-bound thrombin, EPCR-bound APC cleaves PAR1 at a noncanonical site, R⁴⁶-N⁴⁷. The APC-cleaved tethered ligand preferentially initiates a β-arrestin pathways and anti-inflammatory response and endothelial barrier protection.

2.3. Matrix Metalloproteases (MMPs)

Matrix metalloproteases (MMPs) is a family of 28 zinc-dependent endopeptidases, which can be further divided into subgroups based on their, substrate specificity.(1, 50) MMPs have roles in platelet and endothelial cell function through activation of PAR1 (Figure 2).(1) MMP-1, MMP-2, and MMP-13 are capable of cleaving the PAR1 exodomain at noncanonical sites leading to MMP-specified PAR1 signaling. Similar to APC, this biased agonism may also be due to the distinct tethered ligands generated by MMPs versus thrombin (Figure 2).(1) Specifically, MMP-1 cleaves PAR1 after Asp³⁹ (D³⁹-P⁴⁰) generating a tethered ligand that is two amino acids (PRSFLLRN) longer than what is generated by thrombin cleavage (SFLLRN).(51) The signaling cascade elicited by PRSFLLRN includes RhoA activation, p38 phosphorylation and platelet shape change, but to a lesser degree than thrombin-mediated PAR1 activation.(51) On endothelial cells, MMP-1 triggers delayed MAPK signaling.(52) Similarly, the PAR1-dependent phospho-Akt signaling mediated by MMP-1 reaches the peak after 1 hour of activation; while the Akt signaling initiated by canonical PAR1 activation peaks after 5 minutes.(53) MMP-13 cleaves PAR1 at S⁴²-F⁴³ site generating the tethered ligand (FLLRN).(1) More recently, Sebastiano et al showed MMP-2 can enhance platelet activation by cleaving PAR1 after Leu³⁸ (L³⁸-D³⁹) to activate G_q and G_12/13.(54)

2.4. Other proteases

PARs can be activated by several other serine proteases, including proteinase 3, cathepsin G, plasmin, elastase, Factor VIIa, and Factor Xa. (Figure 2)(11, 55–57) Both platelet PAR1 and PAR4 can be cleaved by plasmin, however the signaling outcomes are distinct.(55) Plasmin has four cleavage sites (Arg⁴¹, Arg⁷⁰, Lys⁷⁶, and Lys⁸²) on the PAR1 exodomain.(58) Although plasmin is capable of cleaving PAR1 at its canonical thrombin site (Arg⁴¹), the primary result is inactivation of PAR1 by truncating the tethered ligand at the other sites.(58) Cathepsin G cleaves PAR1 at Phe⁵⁵, which also leads to inactivation of the receptor.(59) PAR4 can be activated by plasmin and cathepsin G at the canonical thrombin site (Arg⁴⁷), which yields the same PAR4 tethered ligand and downstream signaling.(60–62)

Factor VIIa and Xa were first identified to cleave PAR1 nearly 20 years ago.(59, 63) Two recent studies highlight the physiological relevance of PAR1 activation by these proteases.(56, 57) FVIIa induces an EPCR-mediated cytoprotective effect via cleavage of PAR1 at its canonical Arg⁴¹ site, which demonstrates that cofactors can diversify PAR1 signaling even at canonical cleavage sites.(56) Importantly, Kondreddy et al. demonstrated this in vivo using PAR1 mutant mice. In another study, FXa antagonists inhibited PAR1-dependent platelet activation.(57) This finding showed an unexpected direct anti-platelet effect of a FXa antagonist.(57)

3. Tethered ligand activation mechanism

The simplified conventional diagram of PAR activation (Figure 1) depicts the protease solely interacting with the N-terminus of the receptor for the proteolysis reaction.(10, 25, 28, 31, 64) The newly exposed tethered ligand swings in an energy-unfavored fashion to interact with the ligand binding site. However, many studies have shown the interaction between the tethered ligand and the receptor is a complicated structural rearrangement that requires participation of several extracellular loops. Early studies with PAR1 by Gerzsten and colleagues highlighted the coordination between ECL2 and the tethered ligand using a panel of human-Xenopus PAR1 chimeric receptors.(23, 24) Studies with PAR2 polymorphisms and site specific mutations also identified key residues in ECL2 and unmasked its critical role in trypsin-mediated PAR2 activation.(65, 66) ECL2 also participates in the ligand specificity and potency of both the tethered ligand and PAR2 activation peptides.(65–68) Our recent study with histidine-hydrogen deuterium exchange (His-HDX) also supports a role for ECL2 in thrombin-mediated PAR4 activation.(69) Thibeault et al. further zoomed in on two key residues in ECL2, His²²⁹ and Asp²³⁰, that are critical for PAR4 activation.(70) We have recently used amide-hydrogen deuterium exchange (amide-HDX) to monitor the global structural dynamics of PAR4 upon thrombin activation.(71) This study revealed that ECL3 undergoes a dramatic rearrangement following activation and that Pro³¹⁰ in this region is essential for proper PAR4 signaling. (71) This is the first study to show a direct role for ECL3 in PAR signaling. However, the length of ECL3 in PAR1 may also participate in maintaining a proper geometry of the ligand binding pocket, which is critical to the high selectivity of vorapaxar to PAR1 over PAR2 and PAR4.(72) Altogether, these studies point to a more complicated mechanism for the tethered ligand mediated activation of PARs.

The structural features of proteins are tightly connected to their function. PAR4 has distinct signaling kinetics from PAR1 and PAR2. Both thrombin-mediated PAR1 signaling and trypsin-mediated PAR2 signaling are rapid and transient.(65, 73, 74) In contrast, thrombin-mediated PAR4 signaling is prolonged in platelets and other cells.(73–75) Molecular modeling based on the PAR1 structure suggests a shallow endogenous ligand binding site that locates on the surface of the receptor.(72) This is also consistent with another study mapping the ligand binding site of PAR1 to 4 specific residues (Leu⁹⁶, Asp²⁵⁶, Glu²⁶⁰, and Glu³⁴⁷) on the surface.(23, 72) In contrast, our recent amide-HDX study showed a deep binding pocket within the transmembrane bundle of PAR4.(71) This distinct feature between the ligand binding sites of PAR1 and PAR4 may explain the transient versus prolonged signaling of those receptors.

PARs can also be activated independent of cleavage by PAR activation peptides that are the derivatives the tethered ligand sequence. However, the signaling response triggered by those synthesized agonist peptides is different from that of the endogenous tethered ligand generated by proteolysis.(76, 77) In addition, there are several lines of experimental evidence indicating that PARs may undergo distinct structural rearrangements upon activation by the activation peptides compared to the activation by its endogenous tethered ligand. We compared the structural rearrangement of thrombin-activated PAR4 versus PAR4-AP-activated PAR4 by limited proteolysis with thermolysin. PAR4 globally became more compact upon thrombin activation while PAR4-AP does not induce the same conformational rearrangement in the receptor.(69) In addition, point mutations introduced to the ligand binding site of PAR4 have a greater impact PAR4-AP stimulation compared to thrombin.(71) Further, Thiebeault et al. recently used an extensive panel of activation peptides to re-direct the biased signaling of PAR4 pharmacologically.(70) This suggests that different PAR4-APs have the potential to elicit distinct structural conformations of PAR4 and influence downstream signaling. A direct comparison of the high-resolution structures of thrombin-activated PAR4 versus PAR4-AP activated PAR4 is still needed.

4. Termination of PAR signaling

Upon activation, GPCRs are rapidly internalized from the cell membrane via clathrin-mediated endocytosis, which is controlled by the intracellular loops and C-terminal tail.(78) The internalization starts from the phosphorylation of the receptor, which increases the affinity for β-arrestin.(79–81) The binding of the β-arrestin prevents the formation of additional GPCR-G protein complexes, which promotes the uncoupling of GPCRs from their downstream effectors within seconds.(79, 80, 82) The desensitized receptors are internalized within minutes via β-arrestin and other adaptor proteins, such as clathrin and adaptor protein 2 (AP-2).(79, 80, 83) The adaptor proteins facilitate the formation of clathrin-coated pits that enclose the desensitized GPCRs.(80) The GTPase dynamin coordinates the budding-off process of the clathrin-coated vesicle from the cell membrane and the fusion of the vesicle to early endosome(80). Once the receptor is in the endosomes they can either be dephosphorylated and recycle back to the cell surface, or be sorted to lysosome and degraded.(84)

The activation of PARs is irreversible due to their unique activation mechanism. Following cleavage, the exposed ligand is tethered to the receptor and cannot simply diffuse away to terminate the PAR signaling. Thus, for activated PARs, desensitization and internalization are the major ways to mediate PAR trafficking and to discontinue PAR signaling, which are important for appropriate cellular responses (Figure 5). In the absence of stimulus, PARs are continuously circulated between the cell membrane and the intracellular compartment. This maintains a stable pool of the receptor on the surface.

Figure 5. Termination of PAR signaling.

PARs were quickly desensitized and internalized following activation to discontinue signaling. Like other GPCRs, PAR internalization requires both clathrin and dynamin for endocytosis. The vesicles then undergo a series of sorting and the proteins are degraded in lysosome.

In endothelial cells, the termination of PAR signaling is mediated by endocytosis. Unlike other GPCRs, activated PARs do not get recycled back to the plasma membrane after internalization.(85) Instead, they are directly sorted to the lysosome and degraded.(83, 85) Upon activation, PAR1 is quickly phosphorylated by protein kinases, including G-protein coupled receptor kinase (GRKs), and second messenger kinases, such as PKA and PKC, to initiate clathrin dependent internalization(86–90). GRK5 is critical for this primary mechanism of termination of PAR1 signal in endothelial cells(89). Ubiquitination and glycosylation are also involved in this desensitization process. In contrast to most GPCRs, PAR1 internalization follows a β-arrestin independent mechanism.(87) For PAR1, tyrosine-based motif in the C-terminal tail is essential for AP-2 recognition and formation of clathrin-coated pits.(87, 91) The clathrin-coated vesicle fuses with the early endosome and is then sorted to lysosome. SNX1 is required for to prevent PAR1 from being recycled back to cell membrane (Figure 5). (92, 93)

Similar to PAR1, the internalization of activated PAR4 is also independent of β-arrestin and requires both clatherin and dynamin for endocytosis.(94) However, in contrast, PAR4 endocytosis requires phosphorylation of the C-terminal tail and AP-2 binds to a highly-conserved tyrosine-based sorting motif at intracellular loop III (ICL3) on PAR4 to start the receptor internalization.(94)

5. Crystal Structures of PAR1 and PAR2

Solving the structure of membrane proteins has long been a challenge. Twenty years ago, Palczewski et al solved the structure of bovine rhodopsin purified from rod outer segment membranes. This was the first structure of a mammalian GPCR using crystallography.(95) Since then, the strategies for the expression and purification of membrane proteins have continuously evolved to yield large quantities of functional proteins to facilitate structural studies. The artificial expression systems for GPCRs varies from mammalian cell lines (HEK293, COS-1), insect cells (High Five, Sf9, Sf21), yeast (P. pastoris), bacteria (E. coli BL21), to cell-free expression.(96) The sequence of GPCRs are usually engineered for crystallization. The N- and C-termini of the receptors are commonly removed to decrease the flexibility and improve protein homogeneity.(96) Several frequently used fusion partners (T4 lysozyme, mT4L, dsT4L, nanobody, b562RIL, Fab fragment, Rd, PGS) are inserted in unstructured ICL3, N- and/or C-termini to increase the solvent-exposed surface area, which significantly improves lattice formation in the crystal.(96) Additionally, point mutations are commonly introduced into the receptor to increase thermostability, detergent-stability, increase expression yield, or remove heterogeneous glycosylation sites.(96) Further, detergent-micelles, nanodiscs, liposomes, bicelles, lipidic mesophases or SMALPs are often used as the solubilization solvent to extract the protein from the membrane while maintaining their function.(96, 97) Using this mature expression, purification, and crystallization procedure, the structures of many GPCRs were solved within the past couple years, including PAR1 and PAR2.(72, 98)

The structure of PAR1 in complex with vorapaxar, an FDA-approved PAR1 antagonist, at 2.2 Å resolution was solved in 2012.(72) As expected, PAR1 structure showed a conserved 7TM core structure.(72) The PAR family belongs to the δ-subfamily of the class A GPCR. PAR1 is a distant relative to other class A GPCRs that have been solved, which implies that this receptor may have a unique structural rearrangement and signal propagation compared to other GPCRs.(72) In addition to the general architecture of the PAR family, the crystal structure of the PAR1-vorapaxar complex also revealed the distinct binding properties of vorapaxar.(72) Vorapaxar interacts with PAR1 at an unusual location that is very close to the extracellular surface of the receptor. Interestingly, the precise location of the vorapaxar binding site did not provide a structural based explanation for the high selectivity of vorapaxar for human PAR1 over mouse PAR1, human PAR2, and human PAR4.(72) Nearly all of the residues that interact with vorapaxar are conserved across these proteins. Other regions, such as ECL3, may contribute to maintain the overall geometry of binding pocket of vorapaxar on PAR1 and further impact the selectivity of the compound.(72)

Crystal structures of PAR2 in complex with two antagonists that have distinct mechanisms and a blocking antibody have been solved.(98) Since PAR1 and PAR2 share 36% sequence identity, their overall 7TM structures are highly similar. The structure of the PAR2-AZ8838 complex at 2.8 Å resolution revealed the binding site of this competitive antagonist.(98) This small compound is completely buried in a small binding pocket formed by residues from TM1–3, TM7 and ECL2 on the extracellular side of the receptor.(98) The structure of PAR2 with an allosteric inhibitor, AZ3451, was determined at 3.6 Å resolution. AZ3451 bound to an allosteric site outside of the 7TM bundle, which is different from the allosteric site found in P2Y1 and glucagon receptor.(98) This observation highlighted the potential of specifically regulating PAR function using allosteric inhibitors. A commercial antibody, MAB3949 showed a high binding affinity to PAR2. The structure of PAR2- FAB3949-AZ7188 was solved at 4 Å resolution.(98) At this resolution, the structures of PAR2, antibody and b₅₆₂RIL were well resolved but the electron density of the weak PAR2 antagonist, AZ7188, was not observed. The FAB3949 interacts with PAR2 on its extracellular side, including the residues 59–63 from the N-terminus, the ECL2, TM6 and ECL3, which is different from the antibodies targeting PAR2 trypsin cleavage site.(98)

Collectively, the structures of PAR1 and PAR2 in complex with their antagonists or inhibitory antibody are essential for the overall architecture of the PAR family, how their antagonist bind, and how the receptors are allosterically modulated. However, there are remaining questions on the tethered ligand activation mechanism that cannot be resolved based on these structures alone for several reasons. First, the N-terminus of both PAR1 and PAR2 were truncated to increase their conformational stability.(72, 98) As a result, the structures are missing the critical information of the endogenous ligand binding site of PAR1 and PAR2, which prevents the understanding of the tethered ligand mediated receptor activation. Second, recombinant PAR1 and PAR2 were heavily engineered for crystallization. For the recombinant PAR1, a T4 lysozyme was engineered into ICL3 in human PAR1 sequence.(72) For PAR2, the N-terminus was replaced by T4L, and ICL3 was replaced by cytochrome b₅₆₂RIL.(98) In addition, nine point mutations were introduced in PAR2 sequence to increase its thermostability.(98) It is difficult to predict the impact of these thermostabilizing approaches. Therefore, future studies on unmodified full-length PARs using recent technological breakthroughs in single particle cryo-electron microscopy will untangle the remain questions.

6. Single nucleotide polymorphisms (SNPs) in PARs

PAR1 is encoded by the gene F2R. An intronic single nucleotide polymorphism F2R IVS-14 A/T affects PAR1 receptor density on platelets and receptor function.(99, 100) The T allele (allelic frequency, 0.14) led to decreased expression of PAR1 and reduced PAR1-mediated platelet responses compared to the A allele.(100) Interestingly, the reduced platelet activity by the F2R IVS-14 T allele was not associated with a lower risk for thrombotic events or a higher risk for bleeding following percutaneous coronary intervention (PCI).(99)

PAR4, encoded by F2LR3, has four characterized sequence variants that alters receptor reactivity (Figure 6). rs773902 leads to an alanine (Ala) to threonine (Thr) substitution at position 120 in the middle of the transmembrane domain 2 (TM2, 2.48, Ballesteros and Weinstein numbering, (101)). The minor allele frequency (MAF) for the Thr¹²⁰ at rs773902 is 0.19 in individuals with European ancestry (EA) and 0.45 for those with African ancestry (AA).(71) A Thr at 120 (PAR4–120T) leads to a hyperreactive PAR4 compared to the Ala at the same position, which impacts platelet function.(102–105) Platelets with the hyperreactive PAR4 (PAR4–120T) showed an increase in ex vivo thrombus formation, increased sensitivity to low dose thrombin, and were resistant to desensitization.(105, 106) In addition, this genetic variant results in decreased resistance to antiplatelet therapies targeting P2Y12 receptor.(105, 106) Further, PAR4–120T is resistant to inhibition by YD-3, a selective PAR4 antagonist, compared to PAR4–120A.(102) Finally, PAR4–120T was associated with an increased risk of stroke in the SiGN data but was not associated with the risk of venous thromboembolism (VTE) in the INVENT data.(71, 106) Although the precise mechanism is unknown, these data strongly suggest that the Ala to Thr change in TM2 impacts the ligand binding site. In support of this hypothesis, we recently used amide hydrogen deuterium exchange (amide-HDX) to determine that the PAR4 tethered ligand binding site is formed by TM3 and TM7. Interestingly, Thr¹²⁰ is near this region, which raises the intriguing possibility that the change from Ala to Thr impacts how the tethered ligand interacts with its binding site.(71) This needs to be formally tested.

Figure 6. Location of PAR4 polymorphisms.

The location of the four single nucleotide polymorphisms (SNPs) of PAR4 are highlighted. The red blocks indicate the location of the endogenous ligand binding site of PAR4. The extracellular loop 3 (ECL3) is highlighted in orange.

Another sequence variant, Tyr157Cys, is located in TM3 (position 3.33) (Figure 6). The MAF of this variant is less than 10⁻⁴ determined from the Exome Aggregation Consortium Browser.(107, 108) Platelets with PAR4–157C showed reduced responses to the stimulations of PAR4-AP and thrombin compared with the PAR4–157Y.(107) The PAR4–175C platelets showed lower residual thrombin responses when pretreated with vorapaxar, a PAR1 antagonist, which indicated a greater inhibitory effect of vorapaxar on PAR4–157C platelets than on the PAR4–157Y platelets. The reduced function of PAR4-Y157C is caused by the aberrant anterograde receptor trafficking resulting in cells having an unchanged total PAR4 expression but reduced surface expression of the receptor.(107)

rs2227346 is located in TM6 at position 6.48 (Figure 6). The more common “T” allele encodes phenylalanine (Phe) at 296 position, and the less common “G” allele encodes valine (Val).(102) The MAF of this variant is 3.95 × 10⁻³ determined from the Exome Aggregation Consortium Browser.(108) The Val²⁹⁶ showed less PAR4 reactivity than Phe²⁹⁶ variant as measured by a decrease IP₃ generation.(102) This decrease in reactivity for PAR4–296V was also directly associated with a dramatic reduction in PAR4-mediated platelet aggregation compared to PAR4–296F.(102) This SNP locates at a topologically critical residue in TM6 (6.48), which is a part of a consensus scaffold of the ligand-biding pocket that contacts the ligand in nearly all Class A GPCRs.(109) This site is one of the transmission switches (3.40, 5.51, 6.44 and 6.48) that undergo an identical structural alteration upon agonist binding on the extracellular surface of GPCRs.(109) In addition, the position 6.48 is a conserved Na⁺ pocket microswitch. A proper rotation of this site is essential for receptor activation.(109) Therefore, changing the Phe at 296 to Val potentially alters the interaction between the tethered ligand and the receptor, disturbs the collapse of Na⁺ pocket, or both by altering the rotation of this residue to impact PAR4 function.

rs2227376 is the only SNP of PAR4 that is located on the extracellular surface of the receptor (Figure 6). The MAF for the less common leucine (Leu) allele was 0.015 for individuals with European ancestry (EA); this polymorphism was not present in African-ancestry (AA) participants.(71) The proline (Pro) at 310 position is located on extracellular loop 3 (ECL3) and is conserved across species (human, mouse and rat).(71) The change from Pro to Leu reduced the receptor reactivity in response to both PAR4-AP and thrombin. Further, a Leu at 310 was associated with a 15% relative risk reduction for VTE compared to the Pro in the International Network Against Venous Thrombosis (INVENT) Consortium.(71) The protection from VTE is consistent with the Leu having decreased PAR4 reactivity. ECL3 acts as a gatekeeper for PAR4 activation. It blocks the interaction between the tethered ligand and the ligand binding site in the receptor inactive state. Upon thrombin activation, this loop swings out and opens the access to allow the tethered ligand bind to the ligand binding site. The Pro at 310 position is critical for maintaining the rigidity of ECL3, which is important for its role as the gatekeeper. Changing the Pro to Leu potentially increase the flexibility of the ECL3 which further impacts PAR4 reactivity.(71)

7. Antagonists targeting PAR1 and PAR4

As a primary thrombin receptor on the surface of human platelets, PAR1 has been a major target for developing antiplatelet and antithrombotic therapies. Based on their chemical properties and mechanisms of action, PAR1 antagonists can be divided into 4 groups: 1.) Blocking antibodies that inhibit enzymatic cleavage and/or activation of the receptor;(75, 110, 111) 2.) Small molecules or peptide mimetics that orthosterically block the interaction between the tethered ligand and the endogenous ligand binding site;(75, 112–127) 3.) Pepducins which block the recruitment of G-proteins to ICL3 of PAR1;(128, 129) and 4.) Parmodulins that target the cytoplasmic face of PAR1 to selectively block G_q signaling (Figure 7).(130, 131) In addition to inhibiting thrombosis in both in vitro and in vivo models, these antagonists can also be used as probes to enhance our understanding of the tethered ligand mediated PAR1 activation mechanism. Hamilton and Trejo have an excellent review on this topic, in which the details of PAR antagonists are comprehensively described.(132)

Figure 7. Mechanisms of PAR antagonists.

(A) Physiologically, PARs are activated following the protease cleavage. Using PAR1 as the example, the hirudin-like sequence (yellow oval) interacts with the exosite I of thrombin (green sphere), which locks the enzyme in its active conformation leading to efficient PAR1 cleavage. Once cleaved, the newly exposed N-terminus serves as the tethered ligand (pink star), which interacts with the extracellular loop 2 (ECL2) (pink) to initiate a global structural rearrangement of PAR1 to recruit G-proteins to intracellular loop 3 (ICL3). The G-proteins then interact with their effectors to relay the signal. (B) Three strategies to inhibit PAR signaling: 1) blocking the enzymatic cleavage using antibodies that target either the hirudin-like sequence or thrombin cleavage site. 2) interrupting the interaction between the ECL2 and the tethered ligand using peptide mimetics or small molecules. or 3) blocking the recruitment site of G-proteins using pepducins that mimic ICL3 of the PARs. (C) Parmodulin is a fourth PAR1 inhibitory strategy that selectively blocks the signaling G_q-axis but does not affect the G_12/13-axis.

Antibodies to specifically designed to block activation by thrombin was one of the earliest strategies to inhibit PAR1.(75, 133) Two epitopes on the PAR1 exodomain were targeted: the hirudin-like sequence and the thrombin cleavage site.(75, 110, 111) Both the hirudin-like domain IgG generated by the Hung and colleagues (111) and IgG 9600 generated Merck Research Laboratory (110) inhibit PAR1-mediated platelet aggregation via blocking the interaction between thrombin’s exosite I and PAR1. A rabbit polyclonal antibody targeting the cleavage site blocks PAR1 activation and can be used experimentally to monitor receptor cleavage.(75)

A second strategy for designing PAR1 antagonists is to target the interaction between the tethered ligand and the ligand binding site. Orthosteric antagonists of PAR1 can be further divided into two sub-groups: 1.) peptidomimetics based on the tethered ligand and 2.) non-peptide small molecules. Modifications of the activation peptide (SFLLRN) at the R1, R2, R3 or R4 position disrupt the pivotal interaction between the tethered ligand and the ligand binding site. This design yielded a series of high affinity PAR1 antagonists: BMS-200261(75, 114), RWJ-56110(113) and RWJ-58259(116, 121), which are widely used for in vitro and in vivo experiments.(133) Small molecule inhibitors typically have better pharmacodynamic and pharmacokinetic profiles. FR171113 was one of the earliest non-peptide PAR1 antagonist with an IC₅₀ of 0.29 μM for human washed platelets.(123) FR17113 inhibited thrombin- or PAR1-AP-induced ERK1/2 activation.(120) FR171113 also inhibited PAR1-AP- and thrombin-induced guinea pig platelet aggregation in vitro (IC₅₀ = 1.5 and 0.35 μM, respectively).(118) SCH-79797 and its N-methyl analog (SCH-203099) are commonly used in both in vitro and ex vivo experiments due to the high selectivity and specificity for PAR1. Notably, this compound also works in the in vivo rodent models. (112, 119, 133) However, Lee and Hamilton reported that SCH-79797 has significant off-target effects that altered the morphology and function of platelets independently of PARs when used at concentrations required for a full inhibitory effect of PAR1.(134) This brought a red flag for using SCH79797 as a PAR1 antagonist.

Vorapaxar (SCH-530348) is a fourth generation PAR1 antagonist in this series with an outstanding oral antiplatelet effect.(115) In a series of preclinical in vitro experiments, vorapaxar inhibited human platelet aggregation induced by 10 nM thrombin or 15 μM of high affinity thrombin receptor activating peptide (haTRAP) with an IC₅₀ of 47 nM and 25 nM, respectively.(115) Vorapaxar has a high oral bioavailability in multiple species, including rat and monkey.(115) The FDA approved vorapaxar (brand name: Zontivity) as the first-in-class PAR1 antagonist as a novel antiplatelet agent in May 2014 to treat the patients with a history of myocardial infarction or peripheral arterial disease (PAD).(117) It can be administrated with other standards of care, including aspirin and/or clopidogrel, to further reduce the risk of thrombotic cardiovascular events in those patients.(117) However, the search for novel PAR1 antagonists continues because PAR1 is still considered a promising drug target and vorapaxar has limitations. It is contraindicated for patients with a history of stroke, transient ischemic attack, intracranial hemorrhage, or bodyweight <60 kg due to the increased risk of bleeding in the brain in these patients that outweighs the potential therapeutic benefits.(135, 136) The discovery of new PAR1 antagonists with different mechanisms of action is also essential for basic structure-functions studies of PAR1.(122, 124–127)

The third class of PAR1 inhibitors, pepducins, target at the intracellular loop 3 (ICL3) of PAR1 and disrupts the interaction between the receptor and G-proteins. These cell-penetrating palmitoylated peptides can be designed to match the ICL3 sequence of GPCRs, such as PAR1 (P1pal-12) and PAR4 (P4pal-10).(129) Upon incubation with the cells, pepducins insert into the cell membrane and mimic the interaction region between G-proteins and GPCRs to inhibit signaling. In the case of PAR1, P1pal-12 scavenged the free G-proteins in the cytosol to disrupt the recruitment of the G-proteins to PAR1.(129) The PAR1 pepducin, PZ128 (P1pal-7), has completed 2 clinical trials: NCT01806077 and NCT02561000. The former is a Phase I study to evaluate PZ128 in patients with vascular disease or with multiple coronary artery disease risk factors.(137) The second clinical trial was completed on September 24, 2019, in which PZ128 was used as an anti-thrombotic treatment in the non-acute setting in the Thrombin Receptor Inhibitory Pepducin-Percutaneous Coronary Intervention (TRIP-PCI) study.(138)

The fourth class of PAR1 antagonists, parmodulins, selectively target the cytoplasmic face of PAR1 without modifying the endogenous ligand binding site on the extracellular side. Parmodulins block the intracellular calcium flux mediated by PAR1-G_q axis, which is critical for platelet aggregation, but has no effect on RhoA activity by PAR1-G₁₃ axis, which is essential for platelets shape change.(130) In the case of endothelium, parmodulins inhibit PAR1 mediated proinflammatory effect without affecting APC-mediated cyto-protective pathways. (130, 131)

For decades, PAR4 had been considered a backup receptor for thrombin signaling and its cellular importance had been largely overlooked compared to PAR1. However, evidence now supports unique roles for PAR4 in the dual-receptor system of thrombin signaling.(139) Compared to PAR1, PAR4 activation requires a higher concentration of thrombin and leads to prolonged signaling. The sustained kinetics and duration of Ca²⁺ signaling, mediated by PAR4 coupled G_q, is critical for stable thrombus formation and full spreading on fibrinogen in a p38 and ERK1/2 dependent manner (55, 140, 141). Moreover, the prolonged PAR4 signaling also impacts phosphatidyl serine (PS) exposure, Factor V and P-selectin expression, microparticle release, procoagulant platelets, and thrombin generation.(55, 142) Duverney and colleagues also revealed that PAR4, not PAR1, is responsible for the majority of thrombin-mediated GPIIb/IIIa activation on human platelets.(55). The recently appreciated individual contributions of PAR1 and PAR4 to thrombin signaling in platelets has been recognized as an opportunity to exploit therapeutically.(143) This dual receptor system offers the intriguing possibility of tuning thrombin signaling pharmacologically. For example, blocking the sustained signaling from PAR4 may limit thrombosis, while leaving the transient PAR1 signaling to initiate hemostasis and limit bleeding.

The initial compounds targeting PAR4, YD-3 and ML354, showed promise but had low efficacy and low selectivity for PAR4 over PAR1, which limited their application to pre-clinical studies. In early 2017, Bristol-Myers-Squibb reported a potent and reversible PAR4 specific antagonist, BMS-986120.(144) This highly efficacious and specific PAR4 antagonist demonstrated saturable and reversible binding to human PAR4. The most compelling findings in this study were from a direct comparison of BMS-986120 to clopidogrel in a nonhuman primate thrombosis model. BMS-986120 administered orally at 1 mg/kg decreased thrombus weight by 80% with limited bleeding risk. In contrast, the dose of clopidogrel that achieved >80% reduction in thrombus weight (1 mg/kg) led to >8-fold increase in bleeding.(144) This comprehensive study showed BMS-986120 to be an effective PAR4 antagonist with a lower bleeding risk and a wider therapeutic window compared to the standard antiplatelet care. The follow-up Phase I clinical trials of BMS986120 (NCT02208882) and its derivative compound, BMS-986141 (NCT02341638) further confirmed the efficacy and the safety of both PAR4 antagonists as anti-platelet therapies. In addition, this study further supported that inhibiting PAR4 reactivity while preserving PAR1 function is a safer antithrombotic strategy with low bleeding risk. With a better pharmacodynamic and pharmacokinetic profile, BMS-986141 currently finished a Phase II clinical trial (NCT02671461) and showed promising antithrombotic effects with low bleeding risk.

Similar to PAR1, there are several antibodies designed to inhibit PAR4 activity.(145) We have generated a polyclonal antibody targeted to the anionic region of PAR4 that blocked both human and mouse platelet activation. This antibody did not result in increased bleeding in mouse models of hemostasis.(145) In 2016, French et al developed a rabbit polyclonal antibody against the thrombin cleavage site of PAR4.(146) Their antibody showed a highly specific inhibitory effect on PAR4 reactivity measured by intracellular calcium mobilization and PS-exposure on human platelets.(146)

8. Summary and Future Directions

In conclusion, we dissected the tethered ligand mediated receptor activation mechanism of PARs step-by-step: from the interaction between the proteases and PARs, to the overall structural rearrangement of the receptors following proteolysis, and finally the termination of the cellular signaling. There are still many unknowns even after 30 years of study on this receptor family. New proteases have been identified that cleave PARs and new co-factors and mediators that direct biased PAR1 signaling have been uncovered, which makes this an exciting time for PAR research.

We also covered the PAR1 and PAR2 structural studies. These enriched our knowledge on the overall architecture of PARs and how they interact with their antagonists. Due to the limitations of the crystallography, however, we cannot elucidate the tethered ligand activation mechanism from these crystal structures. There is a need to introduce new structural approaches to this field for a better understanding of this unique receptor activation mechanism. Our recent success using amide-HDX and histidine-HDX to monitor the structural rearrangement of PAR4 following activation by thrombin identified novel regions on the receptor that are critical for PAR4 function.(69, 71) These techniques can be applied to the analysis of alternative cleavage sites on PAR1 to determine the structural basis for biased signaling. Recent breakthroughs in cryo-electron microscopy (cryo-EM) also provide opportunities to solve the high-resolution structures of full-length PARs without structural modifications, which will further our understanding of this receptor family.

The polymorphisms of PARs that impact their function offers challenges to translate these findings from the bench back to the bedside. PAR polymorphisms have been associated with some diseases or risk factors in the genome-wide association studies (GWAS). Will these large data sets be clinically useful to guide us in making decisions for individual patients? Are we opening a new era to precision medicine or a Pandora’s box that seems valuable but turns out to be useless? Future studies with the appropriate in vivo models are essential to guide these efforts. Mice are the most commonly used animal models in biomedical research. The SNPs are located in highly conserved regions of PAR4. Mutant mice can be generated to study the impact of the SNPs on PAR4 function iv vivo. However, it is important to keep in mind that human and mouse have different PAR expression profiles on their platelets and efforts to generate PAR1 expressing mouse platelets have been unsuccessful.(55, 147, 148) For platelet specific studies we are limited to transfusion of human platelets into mice or bioengineering approaches using platelets from donors carrying specific SNPs.(149, 150)

PAR1 and PAR4 are important therapeutic targets for anti-platelet and anti-thrombotic therapies. Here, we covered the mechanisms and strategies used target PAR1 and PAR4 to date. Pharmacologically competing with the endogenous tethered ligand brings unique challenges for drug discovery. The progress from this angle has been slow since little is known about the precise location of the tethered ligand binding sites of PARs.(144, 151) Again, high resolution structures of PARs in their natural lipid environment will be useful for drug discovery. Future studies combining multiple biophysical and structural approaches will identify the location of the tethered ligand binding site and the structural basis for how each SNP contributes to PAR function. In closing, PAR1 and PAR4 are promising anti-platelet therapeutic targets, however we must keep in mind that both PAR1 and PAR4 are widely expressed in many cell types and organs. This creates challenges for managing side-effects, but also opportunities for targeting this family of receptors for other disease contexts.

Highlights.

• PARs have a unique tethered ligand mediate activation mechanism

• The structural studies of PARs enrich our knowledge of the architecture of PARs

• Single nucleotide polymorphisms impact the reactivity of PARs

• The strategies for inhibiting PAR function are promising antithrombosis approaches