Targeting biased signaling by PAR1: function and molecular mechanism of parmodulins

Extracellular proteases modulate cellular function through various receptor mechanisms.¹ Some proteases convey cellular effects through protease-activated receptors (PARs), which belong to the 7-transmembrane, G protein-coupled receptor (GPCR) superfamily. PARs are activated when proteases cleave the amino-terminus at specific sites to expose a tethered ligand that interacts with the shallow binding pocket (Figure 1). Among the 4 PARs identified, PAR1 and PAR4 are potential therapeutic targets for the prevention of cardiovascular disease because they mediate thrombin-dependent platelet activation in humans.² Because PAR1 and PAR4 are widely expressed, side effects upon inhibition are a potential limitation.² Indeed, the use of orthosteric PAR1 inhibitors was associated with an increased risk of hemorrhage, including intracranial hemorrhage.³ Whether the increased risk of hemorrhage simply reflected efficient platelet inhibition or unwanted side effects remains unknown. Side effects seem likely, because PAR1 is expressed by various cell types, including vascular cells and because PAR1 (like other GPCRs) mediates biased signaling.^2,4 Although thrombin-dependent PAR1 activation ordinarily induces overall cytodisruptive effects, the activation of PAR1 by the coagulation protease–activated protein C (aPC) conveys overall cytoprotective effects.^4,5 Hence, rather than orthosterically inhibiting all functions of PAR1, allosteric modulation of PAR1 signaling may be a superior therapeutic approach. The following different approaches to modulate PAR1 signaling have been pursued (Figure 2): (1) signaling-selective mutants of aPC with reduced anticoagulant but maintained cytoprotective activity,^4,6 (2) lipidated peptides (pepducins) mimicking intracellular loops of PAR1,^7-10 and (3) small chemical compounds thought to bind to the intracellular face of PAR1 (parmodulins).^11-13 The peptide sequence of PAR1-targeting pepducins corresponds to a region in PAR1 intracellular loops ICL1, ICL2, or ICL3 and is covalently linked to a lipid moiety (eg, palmitate).⁷ The lipid moiety facilitates transport through and anchoring in the lipid bilayer. The close proximity of the peptide to the GPCR-G protein interface interferes with G protein-mediated signaling.^7,9 However, the effect of the pepducins on biased signaling via PAR1 and G proteins remains to be established. Conversely, parmodulins have shown promising effects regarding biased PAR1 signaling. Hence, in this review, we focus on parmodulins, and interested readers are referred to other reviews discussing pepducins.^{8-10,12,14,15} 

The cloning and expression of the receptor mediating thrombin’s effects (later called PAR1) was reported in 1991 by Coughlin¹⁶ and shortly thereafter by Rasmussen et al.¹⁷ Subsequently, 3 further receptors of the same family, referred to as PAR2,¹⁸ PAR3,¹⁹ and PAR4,²⁰ were identified. PAR activation depends on proteolytic cleavage by proteases and binding of the newly generated N-terminal tethered ligand to a shallow binding site on the extracellular surface of the receptor, inducing conformational changes and initiating G protein-dependent signaling.^21-23 Like other GPCRs, PARs are capable of biased signaling. This is best described for PAR1, which conveys the effects of several proteases, including the coagulation proteases, thrombin, and aPC⁴ (Figure 1). Although thrombin is procoagulant and promotes cytodisruptive effects (eg, inflammation, endothelial barrier disruption, cell death), aPC is anticoagulant and conveys cytoprotective effects (eg, inhibition of inflammation, endothelial barrier protection, cell survival).⁴ Several nonexclusive mechanisms underlying PAR1- biased signaling have been proposed. (1) thrombin and aPC cleave the N-terminal end at different cleavage sites, generating different tethered peptide ligands, which interact with PAR1 and promote different conformational changes of PAR1; (2) PAR1 forms a heterodimeric complex with coreceptors, which modulate the conformation of PAR1; and (3) the coupling of PAR1 to G proteins is modulated by lipid-membrane composition.^4,21,24-27 The net result is expected to be a specific PAR1 conformation, favoring signaling through specific G proteins or β-arrestins. The structural dynamics of this PAR1-biased signaling remain incompletely understood.

Because PAR1 conveys thrombin-dependent platelet activation, several drugs targeting PAR1 have been developed, the most advanced being vorapaxar. Vorapaxar has been evaluated in 2 large phase 3 studies in patients with the acute coronary syndrome (TRACER) and the prevention of secondary atherothrombotic events (TRA-2P).^28,29 Although reducing the composite end points comprising cardiovascular or ischemic events, these studies were prematurely discontinued and vorapaxar received a boxed warning by the US Food and Drug Administration to inform about an increased risk of major bleeding, including intracranial hemorrhage.²⁸ The plausible explanation of the increased hemorrhagic risk is the inhibition of thrombin-dependent platelet activation on top of standard care, which includes dual platelet inhibition.^3,28 However, other causes for an increased hemorrhage risk can currently not be excluded. PAR1 is expressed in various tissues and cell types, including endothelial cells, and the long-term consequences of PAR1 inhibition remain to be established. A meta-analysis of studies evaluating vorapaxar and atopaxar, a second orthosteric PAR1 inhibitor, confirmed the association of orthosteric PAR1 inhibitors with bleeding risk.^3,30 In addition, atopaxar showed increased liver toxicity and QTc prolongation when administered at higher doses in a phase II study.^31-33 Whether these side effects of atopaxar were related to PAR1 inhibition or off-target effects remains unknown.

The hemorrhagic risk associated with PAR1 inhibition may be related to the various effects of PAR1 in the vascular system. Thus, PAR1 conveys aPC-mediated cyto- and barrier-protective effects in endothelial cells.⁴ Endothelial PAR1–deficiency results in partial embryonic lethality,³⁴ highlighting an important function of PAR1 in endothelial cells at least during the development. In vitro, vorapaxar at nanomolar concentration induces endothelial barrier dysfunction and apoptosis,³⁵ corroborating detrimental effects secondary to orthosteric PAR1 inhibition on endothelial cells.

Aiming to identify new inhibitors of PAR1, Dowal et al¹¹ screened a small-molecule library using platelet activation by the PAR1-specific activation peptide SFLLRN and dense granule release as a read out. Using this screening, the authors identified a promising compound modulating PAR1 signaling, a quinolone derivative referred to as JF5 (parmodulin-1).¹¹ JF5 reversibly inhibited guanosine-5′-triphosphate binding and GTPase activity in platelet membrane and blocked SFLLRN-induced dense granule release without interfering with other agonists. Although JF5 inhibited dense granule release, it did not inhibit platelet shape change upon SFLLRN stimulation, suggesting that JF5 was an allosteric regulator of PAR1 signaling. Comparative structure-function analyses suggested that JF5 interacts with helix 8 (H8) and that its effect depends on the interactions of H8 with transmembrane (TM) helix 1 and 7.¹¹ However, JF5 lacked selectivity, as it also interfered with the α_2A-adrenergic receptor and other signaling pathways, indicating a need for further optimization.

Further compounds were identified in a library screening conducted by the Flaumenhaft group in collaboration with the Broad Institute through NIH's Molecular Libraries Probe Production Centers Network.^36-38 The aim was again to identify platelet inhibitors, possibly inhibiting some but not all pathways required for platelet activation and hence efficiently preventing thrombosis while reducing the hemorrhage risk. Of the 302 457 compounds screened in an assay for SFLLRN-induced dense granule release, 28 were selected for a secondary screening.^36-38 Eventually, 3 compounds were identified, which targeted cell-surface receptors and because SFLLRN was used as an activator, likely targeted PAR1.^36-38 In a subsequent study by Dockendorff et al³⁹ the lead compound ML161 (also referred to as parmodulin 2) was characterized as a small compound containing a 1,3-diaminobenzene core. ML161 prevented SFLLRN-induced dense granule release from the activated platelets without interfering with platelet shape change, suggesting that ML161 likewise acts as an allosteric PAR1 regulator.³⁹ An oxazole derivative of ML161, termed NRD-21, showed improved plasma stability and a slightly increased anti-inflammatory effect (tissue factor expression) in tumor necrosis factor α (TNF-α)-stimulated endothelial cells.⁴⁰ A comparison of orthosteric and allosteric PAR1 regulators demonstrated that both the types of compounds inhibited PAR1-mediated platelet activation, albeit through different mechanisms. Thus, G_α13-mediated platelet shape change is much less affected by allosteric PAR1 modulators.³⁵ Hence, this new drug class targeting PAR1 is now referred to as parmodulins.³⁵ 

Parmodulins have also been characterized in cultured endothelial cells by their inhibition of PAR1-driven calcium mobilization.^13,40 Functionally, both orthosteric PAR1 inhibitors and parmodulins prevented SFLLRN-induced impairment of endothelial barrier and endothelial release of von Willebrand factor.³⁵ However, orthosteric PAR1 inhibitors, but not ML161, blocked the antiapoptotic effect of aPC in TNF-α stimulated endothelial cells.³⁵ Even in the absence of an agonist, prolonged incubation of endothelial cells with vorapaxar (>24 hours) impaired endothelial barrier function and induced endothelial cell apoptosis, an effect not observed with ML161.³⁵ 

These studies established that the allosteric regulation of PAR1 signaling can be achieved by parmodulins, raising the intriguing possibility that the selective inhibition of PAR1-mediated cytodisruptive effects while sparing its cytoprotective effects may be pharmacologically feasible.

The initial screening to identify allosteric modulators of PAR1 used platelets and platelet inhibition as a readout. Accordingly, newly discovered parmodulins (JF5, ML161) were tested in the in vivo models of laser injury–induced thrombus formation, which was efficiently inhibited by both compounds (see Table 1).^11,35 Yet, ML161 less efficiently inhibits platelet activation by thrombin in comparison with orthosteric PAR1 inhibitors.³⁵ This most likely reflects the circumstance that parmodulins selectively inhibit some but not all G protein–signaling pathways (refer to the molecular mode of action of parmodulins), whereas orthosteric PAR1 inhibitors block all signaling pathways. In addition, as mouse platelets do not require PAR1 for their activation, the antithrombotic effect of parmodulins in mice must be independent of platelet inhibition. Owing to structural similarities, parmodulins may interfere with mouse PAR4 but not human PAR4.³⁵ Hence, PAR4 inhibition could possibly explain the antithrombotic effect of ML161 in mice. However, ML161 inhibits the activation of mouse platelets only by PAR4 agonists but not thrombin.⁴¹ Accordingly De Ceunynck et al⁴¹ could not confirm an inhibitory effect of ML161 on the thrombin-induced aggregation of mouse platelets, suggesting that ML161 has an antithrombotic effect independent of platelet inhibition in mice as well. Parmodulins mimic aPC’s endothelial protective effect, including the induction of anti-inflammatory and anticoagulant effects. Thus, the parmodulins ML161 and NRD-21 reduce the expression of tissue factor in TNF-α–stimulated endothelial cells⁴⁰ and also decrease thrombin generation on lipopolysaccharide (LPS) or TNF-α–stimulated endothelial cells without directly interfering with plasmatic coagulation.⁴¹ Furthermore, parmodulin reduces the adhesion of platelets to activated endothelial cells under flow conditions. Using a lung-on-a-chip model comprising lung alveolar epithelium and endothelial cells, Jain et al⁴² demonstrated the efficient prevention of LPS-induced platelet accumulation by ML161. Interestingly, the inhibitory effect was slightly stronger upon the preincubation of endothelial cells than blood, corroborating that ML161 may act primarily on endothelial cells.⁴² Therefore, parmodulins seem to dampen coagulation by maintaining the antithrombotic properties of endothelial cells.⁴¹ 

Possible effects of parmodulins in other cellular systems have been proposed too. In retinal pigment epithelial cells, the coagulation factor Xa via PAR1 induces epithelial-to-mesenchymal transition, and this effect can be prevented by pretreatment with ML161.⁴³ These in vitro experiments raise the intriguing possibility that ML161 modulates PAR1 signaling upon factor Xa activation, but direct evidence for this scenario was not provided.

Differential effects of vorapaxar vs ML161 on thrombin-stimulated endothelial cells were confirmed using unbiased phosphoproteomics. Neither PAR1 ligand alone induced changes in the phosphoproteome in the short term (after 1 hour). In the thrombin-stimulated endothelial cells, vorapaxar completely but ML161 only partially blocked phosphoproteomic changes.⁴⁴ These comparative studies confirmed the differential effects of orthosteric PAR1 inhibition vs allosteric PAR1 modulation.

The endothelial protective effects of parmodulins contrast with the effects of orthosteric PAR1 inhibitors. Aisiku et al³⁵ reported that vorapaxar, but not ML161, significantly impaired the endothelial barrier upon prolonged exposure. Aksoyalp et al⁴⁵ showed that vorapaxar, but not ML161, induced human left internal mammary artery endothelial cell dysfunction, as reflected by the impaired endothelial-dependent vascular relaxation. Intriguingly, parmodulins do not only protect endothelial cells from thrombin-PAR1–induced cell death but also from cell death induced by other stimuli (TNF-α or staurosporine).^41,45 Parmodulins induce a genetic program hallmarked by reduced NF-kB expression but increased cytoprotective transcripts,⁴¹ reflecting changes observed with aPC.^5,46,47 Small interfering RNA studies also corroborated that parmodulins convey endothelial-protective effects via PAR1, mimicking aPC-mediated cytoprotective signaling.^41,45 The cytoprotective transcriptional response by parmodulins may also depend in part on G_βγ and PI3K/Akt signaling as the exposure of human endothelial cells to ML161 is sufficient to induce Akt phosphorylation even in the absence of further stimuli.⁴¹ Akt-activation by ML161 and aPC depends on the G_βγ–PI3K interaction and PI3K activity.⁴¹ Cytoprotective effects of aPC have been linked to the modulation of the PI3K-signaling pathways, involving improved cardiac function in a myocardial or neuronal ischemia model or inducing an adaptive endoplasmic stress response in renal epithelial cells.^27,48 The relevance of G_βγ signaling and the activation of the PI3K/Akt for the cytoprotective effects of ML161 and whether other parmodulins have the same effect on G_βγ signaling remains unknown.

The cytoprotective effects of ML161 were investigated in several in vivo disease models. In a model of myocardial ischemia–reperfusion injury, ML161 (5 mg/kg intravenously 30 minutes before ischemia induction) markedly reduced infarct size and mTOR-dependent NLRP3 inflammasome activation, mimicking the cytoprotective effects of aPC.⁴⁹ In these studies, aPC inhibited inflammasome activation in various nonendothelial cells, but whether parmodulin directly affects cell types other than endothelial cells was not evaluated.⁴⁹ Another study evaluated the effect of protein C signaling on p21-induced metabolic memory in diabetic kidney disease.⁵⁰ Like wild-type aPC and an aPC mutant with minimal anticoagulant effect (3K3A-aPC⁵¹), ML161 reversed preestablished albuminuria in hyperglycemic mice, improved renal pathohistological changes, and reduced senescence of tubular cells. Although in vitro studies suggest that aPC acts directly on tubular cells, the same was not demonstrated for ML161.^50,52 Hence, it remains currently unknown whether parmodulins directly modulate PAR signaling in renal tubular cells. Collectively, these studies suggest that parmodulins mimic the cytoprotective effects of aPC both in acute and chronic in vivo disease models.^49,50,52 Yet, we lack insights into whether parmodulins may likewise mimic aPC-PAR1–dependent cytoprotective signaling in humans. Further studies are needed to define the cell type(s) parmodulins act on in vivo and the extent to which these observations can be extended to other disease and animal models.

The mode of action of parmodulins is thought to involve the biased modulation of PAR1 coupling with intracellular G proteins. Parmodulins inhibit PAR1 signaling through Gα_q, leading to the blockage of Gα_q downstream responses like cytosolic Ca²⁺ influx, but have not been reported to inhibit signaling through Gα_12/13.^11,13,35,40 Experiments with mutant PAR1 indicated that H8 of the receptor plays a role in the mode of action of parmodulins. The activity of the 2 structurally different parmodulins JF5 and ML161 was blocked when the residues S376^8.48 to L386^8.58 of H8 were replaced with 3 Ala residues^11,35 (Table 2). Similarly, chimeric human PAR1/PAR4 receptors in which H8, the last 9 residues in TM7, and the C-terminal tail of PAR1 were replaced with the corresponding sequence of PAR4, were insensitive to parmodulins JF5 and ML161³⁵ (Table 2). Comparison of the amino acid sequences of PAR1 and other JF5-sensitive GPCRs suggests that the activity of JF5 could depend on PAR1 structural features that stabilize the conformation of H8, such as a possible ionic interaction of position 1 in H8 with ICL1, a hydrophobic interaction of position 2 of H8 with the NPXXY motif in TM7, and the palmitoylation sites in H8.¹¹ Experiments with chimeric receptors also indicated that parmodulin activity is dependent on ICL3, whereas no influence could be found for ICL2.³⁵ Whether the action of parmodulins is based on a physical interaction with H8 and ICL3 or an allosteric mechanism involving these segments of the PAR1 structure has not been clarified. Further experiments, such as site-directed mutagenesis or X-ray crystallography are needed to obtain high-resolution information on the binding sites of parmodulins.

Aisiku et al³⁵ demonstrated that the binding affinity of orthosteric PAR1 ligands was unchanged in the presence of parmodulins JF5 and ML161, which indicates that parmodulins have no significant allosteric effect on the extracellular binding site. No positive or negative cooperative effect of parmodulins on the orthosteric ligand could be detected. Together, these findings suggest that the sites mediating the action of parmodulins are localized to the cytosolic side of PAR1 and possibly involve interactions between PAR1 and G proteins.

Thus, the mode of action of parmodulins is clearly different from that of orthosteric PAR1 inhibitors, like vorapaxar and atopaxar that target the extracellular ligand binding pocket. Intriguingly, ligand occupancy of the endothelial protein C receptor (EPCR), a coreceptor required for cytoprotective aPC-PAR1 signaling in endothelial cells, likewise switches PAR1 signaling toward cytoprotection.^53-55 Whether the parmodulin-induced and the EPCR-induced switch in PAR1 signaling are mechanistically related remains unknown. Addressing these questions depends on a better understanding of the mechanism(s) of biased PAR1 signaling as well as obtaining structural insight into PAR1’s interaction with G proteins, β-arrestins, parmodulins, and EPCR.

It is intriguing to see that parmodulins have very different chemical structures (Figure 3A) but induce the same functional outcome of blocking thrombin-induced proinflammatory signaling in endothelial cells while maintaining aPC-induced cytoprotective signaling.³⁵ The fraction of matched atoms (Tanimoto coefficient, TC) for the 4 first-generation parmodulins (JF5, ML159, ML160, ML161) ranges between 16% and 28% (Figure 3B), which shows that they are chemically diverse. ML161 and NRD-21 have a high TC (80%), which reflects that they are derivatives. JF5 (parmodulin 1) is a quinolin-imine derivative with a pentyl tail group. Interestingly, the inhibitory activity of JF5 and JF5 analogs depends on the length of their alkyl tail group.¹¹ Compounds with tail groups of 3 to 5 carbon atoms have the strongest inhibitory potency, whereas analogs with longer alkyl tails (7-16 carbon atoms) fail to inhibit PAR1-mediated platelet aggregation.¹¹ ML161 has a 1,3-diaminobenzene scaffold and is the most potent and selective first-generation parmodulin.^35,39 ML161 demonstrated relatively little off-target inhibition when tested in GPCR profiling studies^35,39 in contrast to JF5, which demonstrated inhibition also of other platelet GPCRs, including α2A, 5-HT_2A, and CCR4 receptors.³⁹ Because of the promising potency and selectivity of ML161, structure-activity relationship studies were undertaken which helped derive the requirements for the optimal potency of this scaffold acting on PAR1.³⁹ The requirements for activity are (1) a secondary benzamide on the eastern side of the molecule with halogen atoms or small lipophilic groups in the ortho position, (2) a secondary amine or amide on the western side with a linear or branched alkyl chain with fewer than 4 carbons, and (3) a 1,3-substitution pattern of the central benzene ring.³⁹ Replacement of the central ring with N-methyl amides made the molecule inactive, which suggests that both amide groups interact via hydrogen bonds with PAR1. Recently, the group of Dockendorff expanded their structure-activity relationship studies on the ML161 scaffold,⁴⁰ which yielded the profiling of oxazole derivative NRD-21 in additional detail (Figure 3A). NRD-21 showed slightly improved potency in Ca²⁺ mobilization assay compared with ML161 but had much improved stability in mouse plasma.⁴⁰ 

The observation that all parmodulins (JF5, ML159, ML160, ML161) require H8 for activity and cause the same functional outcome of selective PAR1 inhibition through Gα_q but not Gα_12/13 signaling, albeit with different potencies and selectivities, suggests that these parmodulins act via a common molecular mechanism. Indeed, different parmodulins (JF5, ML161) induce platelet shape change, which is sensitive to Rho kinase inhibitors,^11,35 indicating that JF5 and ML161 modulate RhoA signaling in the platelets despite their different structure. Whether parmodulins target the same signaling pathways in other cells is an interesting question, given that parmodulins act on platelets, endothelial cells, and possibly other cells. Moreover, parmodulins convey their effects through (1) the reduction (not inhibition) of platelet activation and (2) the induction of an anti-inflammatory or cytoprotective phenotype in endothelial cells. These effects are expected to work in concert to reduce thrombo-inflammatory blood disorders but may be substance and context-dependent. Thus, in an acute setting such as ischemia-reperfusion injury, both features, inhibition of platelet activation and endothelial cytoprotection, are expected to improve the outcome. In contrast, in a chronic disease not linked to thrombotic vascular occlusion, such as diabetic kidney disease, the cytoprotective effects may be more important. Further preclinical work is required to define the mechanisms of action and functional consequences of parmodulins.

A similar mode of action despite different chemical structures may not be completely unexpected. Studies on other class A GPCRs in complex with chemically diverse ligands provided evidence for a common inhibitory mechanism involving a conserved intracellular facing binding pocket of the GPCR.⁵⁶ This concept is best studied for chemokine receptors and the β₂AR for which a handful of X-ray structures of receptors bound to small molecules at the cytosol facing side have been determined.^57-61 The chemical structures of these ligands are highly diverse, as reflected by the TC values ranging from 10% to 21%, despite their common binding site on the GPCR and inhibitory mechanism (Figure 3). Similarly, TC values for different parmodulins range from 16% to 28% yet convey comparative modulatory effects on PAR1 signaling. As the potential application cases of intracellular GPCR ligands are more and more realized,⁵⁶ a comparison of parmodulins with other intracellular GPCR ligands may provide further insights into their mode of action.

Structural studies on β₂AR⁵⁷ and chemokine receptors^58-61 have demonstrated the binding of allosteric ligands to a pocket on the intracellular receptor surface. The pocket occupied by Cmpd-15PA in β₂AR,⁵⁷ CCR2-RA-[R] in CCR2,⁵⁸ Cmp2105 in CCR7,⁶⁰ vercirnon in CCR9,⁵⁹ and compound 00767013 in CXCR2⁶¹ is structurally highly similar and may be conserved among GPCRs. In the solved structures, the ligand is caged by TMs 1, 2, 3, 6, and 7, and H8 and ICL1. One common chemical feature shared by the intracellular ligands of these GPCRs is the presence of oxygen-donating functional groups (eg, a thiadiazole-dioxide in Cmp2105, hydroxypyrrolinone in CCR2-RA-[R], and sulfonamide in vercinon). The oxygen atoms form hydrogen bonds to the residues in the TM7-H8 connecting turn and thus act as a helix cap for H8.

The intracellular ligands of these GPCRs are allosteric antagonists. The following 3 main explanations have been suggested for their allosteric antagonism: (1) The ligands stabilize the receptor in an inactive conformation and block structural changes required for the activation, for example, of TMs 6 and 7. The stabilization of the inactive state also contributes to increasing the thermal stability of the receptor as seen for CCR2-RA-[R] on CCR2⁵⁸ and Cmp2105 on CCR7.⁶⁰ (2) The ligands exert a negative cooperative effect on the orthosteric agonist, which weakens or prevents agonist binding, and a positive cooperative effect on the antagonists, owing to the stabilization of the inactive state. (3) The position of the intracellular ligands is sterically incompatible with the G protein and arrestin binding sites, thus preventing these transducers from engaging in stable interactions with the receptor.

The amino acids in PAR1 that are homologs of the intracellular binding pocket residues in β₂AR, CCR2, CCR7, CCR9, and CXCR2 are conserved above average (Figure 4B) and are 40% to 57% similar (17%-30% identical) to their homologs in the other GPCRs. The parmodulin sites of action, shown in Figure 4A, partially overlap with those regions that surround the intracellular ligands in β₂AR, CCR2, CCR7, CCR9, and CXCR2; particularly H8 is a shared target site. Moreover, the PAR1 receptor structure²² reveals accessible pockets on the intracellular side (Figure 4C), which could be large enough for a small molecule to fit in and where a small molecule could be poised to modify transducer protein interactions. Figure 4C shows surface representations of these intracellular pockets observed in the PAR1 structure. We note that a refined PAR1 structure from the GPCR database⁶² has been used to create the visualizations.

However, several considerations make the possibility of a shared binding site between parmodulins and the other GPCR ligands unlikely. Parmodulins differ from the intracellular antagonists of β₂AR and chemokine receptors in the following important aspects: (1) Parmodulins act as biased modulators of PAR1 not as full antagonists, as they inhibit PAR1 signaling through Gα_q but maintain signaling through Gα_12/13. Thus, they do not prevent the formation of the active receptor conformation, at least not completely and not in the case of all G proteins. Interactions of PAR1 with Gα_12/13 protein and arrestin need to be functional to elicit downstream signaling. (2) Parmodulins failed to show significant negative cooperativity on orthosteric PAR1 agonists,³⁵ indicating that they have no significant allosteric effect on the extracellular side of the receptor, in contrast to the intracellular antagonists of β₂AR⁵⁷ and chemokine receptors.⁶³ Overall, this comparison suggests that parmodulins have a different target site on PAR1 and act via a different mode of action than the intracellular ligands acting on β₂AR and chemokine receptors. Parmodulins’ mode of action may involve, among other things, modulation of the binding pathway on which G proteins approach the PAR1 cytosolic face or an allosteric fine-tuning of the strength and dynamics of PAR1-G protein interactions.

The biased signaling properties of PAR1 are well established, but therapeutically harnessing these remains a challenge. Allosteric PAR1 regulators have emerged as promising agents, as they mimic barrier-protective and cytoprotective signaling of aPC in endothelial cells, yet convey antithrombotic and platelet-inhibitory effects. The platelet-inhibitory effect of parmodulins is less potent (eg, reversible, some pathways spared) compared with the orthosteric PAR1 inhibitors,³⁵ and whether this differential response translates into a clinical benefit (lower risk of hemorrhage owing to platelet activation at high thrombin concentrations) or disadvantage (less protection from thrombotic vascular occlusion) remains to be shown. Considering the lethal phenotype in mice is because of endothelial-PAR1 deficiency,³⁴ sustained endothelial function upon exposure to parmodulins make these compounds interesting drug targets. Yet, whether the allosteric modulation of PAR1 signaling translates into a clinical benefit in all tissues remains to be explored. Early preclinical studies in mouse models confirm their antithrombotic effect paired with cytoprotective effects in both acute (myocardial ischemia–reperfusion injury) and chronic (diabetic kidney disease) injury models. Other diseases which have been linked to detrimental thrombin-PAR1 signaling may also be potentially treated with parmodulins. This includes neurological diseases, as signaling-specific aPC variants demonstrated protective effects in the preclinical models of stroke, experimental autoimmune encephalitis, and amyotrophic lateral sclerosis.^64,65 Considering the partial beneficial effect of aPC in the clinical studies of sepsis,⁶⁶ signaling-selective aPC variants (3K3A-aPC and 5A-aPC) in animal models of sepsis,⁶⁷ as well as the barrier-protective effects of parmodulins,³⁵ sepsis may be another possible indication for parmodulins. To take the matter further, more insights into the mode of action through which parmodulins regulate PAR1 dynamics and PAR1-dependent signaling need to be obtained. This will provide mechanistic insights and may allow the optimization of currently available parmodulins. In addition, further preclinical studies to define potential indications for allosteric PAR1 regulators are required. The use of small molecules to allosterically modulate signaling in a biased fashion is a promising therapeutic approach worthy of deeper study⁶⁸ for both PARs and GPCRs more generally.

The authors thank Chris Dockendorff for his careful reading of the manuscript and for his helpful comments.

This work was supported by the following grants from the Deutsche Forschungsgemeinschaft: IS 67/16-1, IS 67/22-1, IS-67/25-1, IS 67/26-1, Projektnummer 236360313/SFB 1118.

Contribution: G.K. and B.I. conducted the literature research and wrote the manuscript.

Conflict-of-interest disclosure: B.I. is a member of the scientific advisory board of Function Therapeutics, Inc, which develops parmodulins. G.K. declares no competing financial interests.

Correspondence: Berend Isermann, Institute of Laboratory Medicine, Clinical Chemistry and Molecular Diagnostic, University Hospital, Paul-List-Str 13/15, Haus T, 04103 Leipzig, Germany; e-mail: berend.isermann@medizin.uni-leipzig.de.