Structural basis for the activation of proteinase-activated receptors PAR1 and PAR2

Abstract

Members of the proteinase-activated receptor (PAR) subfamily of G protein-coupled receptors (GPCRs) play critical roles in processes like hemostasis, thrombosis, development, wound healing, inflammation, and cancer progression. Comprising PAR1-PAR4, these receptors are specifically activated by protease cleavage at their extracellular amino terminus, revealing a ‘tethered ligand’ that self-activates the receptor. This triggers complex intracellular signaling via G proteins and beta-arrestins, linking external protease signals to cellular functions. To date, direct structural visualization of these ligand-receptor complexes has been limited. Here, we present structural snapshots of activated PAR1 and PAR2 bound to their endogenous tethered ligands, revealing a shallow and constricted orthosteric binding pocket. Comparisons with antagonist-bound structures show minimal conformational changes in the TM6 helix and larger movements of TM7 upon activation. These findings reveal a common activation mechanism for PAR1 and PAR2, highlighting critical residues involved in ligand recognition. Additionally, the structure of PAR2 bound to a pathway selective antagonist, GB88, demonstrates how potent orthosteric engagement can be achieved by a small molecule mimicking the endogenous tethered ligand’s interactions.

Proteinase-activated receptors (PARs) regulate thrombosis, inflammation, and wound healing. Here, authors present PAR1 and PAR2 structures, revealing conserved activation mechanisms and ligand interactions within their orthosteric pockets, providing insights into PAR functionality.

Introduction

Proteinase-activated receptors (PARs) are a distinct subgroup within the broader family of G protein-coupled receptors (GPCRs) and are essential to a myriad of biological processes^1–3. These receptors are activated through proteolytic cleavage by various serine proteases, which exposes a ‘tethered ligand’ within the receptor’s N-terminal region. This tethered ligand subsequently binds intramolecularly to initiate signaling, distinguishing PARs from other GPCRs that are typically activated intermolecularly by soluble ligands⁴. Like many GPCRs, PAR1 and PAR2 play crucial roles in human physiology and pathology, particularly in processes such as inflammation, coagulation, and immune responses^4–6.

PAR1, the first identified member of the PAR family, has been widely studied for its critical roles in hemostasis and thrombosis^7,8. Predominantly expressed on platelets, smooth muscle cells and endothelial cells, PAR1 is primarily activated by thrombin, a key enzyme in the coagulation cascade^9–12. This highlights PAR1’s pivotal role in vascular biology and its potential as a therapeutic target for novel antithrombotic therapies designed to reduce thrombotic risk while minimizing the bleeding complications often linked to conventional anticoagulants¹³.

In contrast, PAR2 is predominantly activated by trypsin and tryptase, playing diverse roles in inflammation and immune response¹⁴. Unlike PAR1, PAR2’s functions are implicated in conditions such as chronic inflammation, pulmonary disorders, and cancer progression^15,16. Its ability to respond to multiple proteases underscores its broader physiological significance and complexity, making it an attractive target for pharmaceutical development aimed at treating a range of inflammatory and autoimmune disorders¹⁷.

Mutational scanning methods were first applied to investigate the mechanism of PARs, revealing that residues from the tethered ligands are in contact with extracellular loop 2 (ECL2) of the receptor¹⁸. Subsequent crystal structures of thrombin bound to PAR1’s N-terminus revealed their interactions and hypothesized how these interactions may cause allosteric effects on GPCR activation, in addition to proteolysis linked activation of PAR1^19–21. The crystal structure of PAR1 in complex with vorapaxar, a highly selective tricyclic himbacine-derived inhibitor of PAR1, marked a significant milestone in the structural study of PARs²². This work revealed unprecedented details of the interaction between PAR1 and the antagonistic compound. Separately, a study reported the crystal structures of PAR2 in complex with the allosteric antagonists AZ3451 and AZ8838²³. While these structures provided unparalleled detail, the activation mechanism of PARs and details of tethered ligand-GPCR interactions within the orthosteric pocket remain poorly understood.

Building on this foundational knowledge, our research delves further into the structural analysis of PAR1 and PAR2, revealing important details of their activation mechanisms. We have resolved the structures of PAR1 and PAR2 in complex with their endogenous tethered ligands and G proteins, providing details of receptor activation rearrangements. Our analysis reveals that both PAR1 and PAR2 feature shallow, constricted orthosteric binding pockets essential for ligand recognition and receptor activation. This structural characterization not only deepens our understanding of their molecular functionality, but also identifies residues that are critical to ligand binding and receptor activation. Surprisingly, our comparative studies with antagonist-bound structures reveal minimal conformational shifts in the TM6 helix, suggesting a subtle yet effective activation mechanism distinct from the dramatic transformations observed in other GPCRs. These insights are vital for comprehending the specificity of receptor activation and signaling and allow us to propose a unified model for the activation of PARs.

To further our understanding of PAR2 druggability, we determined the structure of GB88 bound to an activated PAR2 heterotrimeric G-protein complex. GB88 is a biased antagonist of PAR2 that selectively inhibits the PAR2/Gq/11/Ca²⁺/PKC signaling pathway, resulting in anti-inflammatory effects in vivo²⁴. At the same time, it acts as an agonist in three other PAR2-activated pathways (cAMP, ERK, Rho) in human cells²⁴. Our structure of PAR2 bound to this versatile molecule reveals it to bind deep within the constricted orthosteric pocket by effectively mimicking key interactions that we observe in our tethered ligand structure of PAR2.

Here, we report the structural basis of PAR1 and PAR2 activation, revealing key molecular interactions involved in tethered ligand binding and receptor activation. Our findings identify conserved residues critical for orthosteric pocket interactions and reveal minimal conformational shifts in TM6, paired with distinct TM7 movements, driving a unified activation mechanism for PAR1 and PAR2. Additionally, we present the structure of PAR2 bound to GB88, illustrating its mimicry of endogenous ligand interactions for potent orthosteric engagement. Overall, this work enhances our understanding of ligand-receptor and small molecule-receptor interactions, providing a foundation for designing therapies targeting the PAR1 and PAR2 orthosteric binding sites.

Results

PAR1 and PAR2 bound to their endogenous tethered ligands

To elucidate the agonist-binding and activation mechanisms of PAR1 and PAR2, we focused on their natural agonist, the tethered ligand. We employed consistent strategies in designing constructs for both receptors. Briefly, PAR1 (residues 42-425) and PAR2 (residues 37-397) were cloned into the pFastbac vector following the signal peptide (Supplementary Fig. 1A). This arrangement simulates protease cleavage, thereby exposing the first amino acid of the tethered ligand, which is crucial for receptor activation. To stabilize the PAR1 and PAR2 complexes with Gαq proteins, we implemented the NanoBiT tethering strategy by attaching the large BiT (LgBiT) to the C-terminus of each receptor, followed by a Flag tag for purification purposes²⁵.

Additionally, we utilized an engineered Gαq protein (GqiN), in which the N-terminus (residue 1–32) of Gαq was replaced with the N-terminus (residue 1–28) of Gαi²⁶. This modification significantly enhances the protein’s affinity for the scFv16 antibody, a tool that has been instrumental in elucidating numerous receptor/G-protein complexes. The Gβ subunit in both the PAR1 and PAR2 complexes features a C-terminal HiBiT, which pairs with the LgBiT²⁵ (Supplementary Fig. 1A).

To assemble the PAR1-Gαq and PAR2-Gαq complexes, receptor subunits, Gαq, Gβ, Gγ, and scFv16 were co-expressed in Hi5 insect cells. The complexes were then purified using Flag affinity resin (Supplementary Fig. 1B). The structures of the PAR1-Gαq and PAR2-Gαq complexes were resolved via cryo-EM to resolutions of 2.7 Å and 3.1 Å (Figs. 1,  2 and Supplementary Fig. 2–4), respectively. The clarity of the cryo-EM density maps enabled modeling of most side chains of the receptors, the Gαβγ heterotrimer, the tethered ligands and scFv16 (Figs. 1A–D,  2A–D, Supplementary Fig. 3). Both complexes exhibit a highly similar overall structural fold, adopting a canonical seven-transmembrane helix domain (TMD), with root mean square deviation (RMSD) values of 1.07 Å over 205 Cα atoms. The tethered ligands bind at the orthosteric pocket formed by transmembrane helices TM5, TM6, TM7, the N-terminal loop, and ECL2 (Figs. 1,  2).

Fig. 1. Cryo-EM structure of the PAR1-Gαq-scFv16 tethered peptide agonist complex.

A Cryo-EM density map of PAR1-Gαq-scFv16 with tethered ligand. (PAR1, blue; Gαq, teal; Gβ, yellow; Gγ, purple; scFv16, gray). B Cartoon representation of PAR1-Gαq-scFv16 with tethered ligand bound. Inset: tethered ligand (yellow sticks) is shown within the orthosteric pocket of PAR1 (slabbed surface). The tethered ligand cryo-EM map is shown as blue mesh at 5.0 sigma. C Cartoon representation of PAR1 (blue) with the tethered ligand bound (yellow sticks) viewing from the extracellular side of the membrane with TMs and ECLs labeled. D H-bond interactions between the tethered ligand (yellow sticks) and PAR1 orthosteric residues (orange sticks).

Fig. 2. Cryo-EM structure of the PAR2-Gαq-scFv16 tethered peptide agonist complex.

A Cryo-EM density map of PAR2-Gαq-scFv16 with tethered ligand. (PAR2, salmon; Gαq, teal; Gβ, yellow; Gγ, purple; scFv16, gray) B Cartoon representation of PAR2-Gαq-scFv16 with tethered ligand bound. Inset: tethered ligand (green sticks) is shown within the orthosteric pocket of PAR2 (slabbed surface). The tethered ligand cryo-EM map is shown as blue mesh at 5.0 sigma. C Cartoon representation of PAR2 (salmon) with the tethered ligand bound (green sticks) viewing from the extracellular side of the membrane with TMs and ECLs labeled. D H-bond interactions between the tethered ligand (green sticks) and PAR2 orthosteric residues (yellow sticks).

Tethered ligand bound in the orthosteric binding pocket

PAR1 is uniquely activated by thrombin through a specialized proteolytic mechanism that sets it apart from most other G protein-coupled receptors. This activation begins when thrombin cleaves PAR1 at residue R41^N-term located within its extracellular N-terminal domain²⁷. This proteolytic event exposes a new sequence motif SFLLR, beginning at S42^TL, and acts as the tethered ligand. Our PAR1-Gαq cryo-EM map reveals density for these initial five residues (Fig. 1B). They adopt an extended conformation that binds into a shallow pocket at the center of the 7 TM bundle of PAR1, forming contacts with residues from TM1, TM5, TM6, TM7, and ECL2 (Fig. 1D). The tethered ligand is nestled into a shallow and narrow pocket with the deepest residue, S42^TL, in hydrogen bonds distance with residues H255^ECL2 of ECL2 and Y337^6.59 of TM6 (Fig. 1D). The adjacent residues, F43^TL, L44^TL, and L45^TL, contribute van der Waals interactions with Y350^7.32 of TM7, and additional H-bonds with D256^ECL2 and L258^ECL2, illustrating a complex network of interactions that facilitate the activation of the receptor (Fig. 1D).

In comparison, PAR2 activation involves a direct mechanism by trypsin, which cleaves the receptor at R36^N-term. This cleavage reveals a new N-terminal sequence starting with SLIG, beginning at S37^TL, that serves as a tethered ligand²⁸. This ligand then folds back onto the receptor itself, binding intramolecularly and inducing significant conformational changes that are essential for receptor activation. Our cryo-EM density map of the PAR2-Gαq complex has effectively resolved the initial three residues, SLI, showing a similar pattern of interactions (Fig. 2B) that are observed in PAR1. This tethered ligand also binds to a shallow pocket, engaging in key interactions with residues from TM1, TM5, TM6, TM7, and ECL2 (Fig. 2D). Notably, residues S37^TL and L38^TL play important roles in this binding. S37^TL reaches into the base of the pocket, forming putative hydrogen bonds with the main chain carbonyl of H227^ECL2 and the sidechain of Y311^6.59 while L38^TL is involved in a polar interaction with L230^ECL2 and forms van der Waals interactions with Y323^7.32, as well as forming a putative H-bond with D228^ECL2 (Fig. 2D), illustrating the detailed mechanism of ligand-receptor interactions that lead to the activation of PAR2. Alignment of the PAR1 and PAR2 structures reveal that they share highly similar binding modes of the tethered ligands and overall conformation (Supplementary Fig. 5).

Structural comparison of active and inactive states of PAR1

The structure of PAR1 bound to the antagonist vorapaxar (PDB 3VW7) illustrates how vorapaxar partially occupies the orthosteric site of PAR1, effectively inhibiting its activation by sterically preventing rearrangements necessary for activation²². Vorapaxar stabilizes PAR1 in its inactive conformation through multiple hydrogen bonds and hydrophobic interactions with multiple residues. These interactions hinder the conformational changes required for G protein coupling and physically block the tethered ligand binding site, with the tricyclic ring of vorapaxar occupying the same position as S42^TL when the tethered ligand is bound. In our cryo-EM structure, the tethered ligand is demonstrated to be positioned higher than vorapaxar’s methylpyridine core and distal chlorophenyl ring, that both penetrate deep within the receptor, sterically obstructing the large conformational rearrangements of bulky tyrosine residues Y350^7.32 and Y353^7.35, (Fig. 3A–D) and virtually irreversibly inhibiting receptor activation by PAR1’s tethered ligand²².

Fig. 3. Structural Comparison of Active and Inactive States of PAR1.

A Comparison of the tethered ligand-bound to active PAR1 (blue) and the antagonist vorapaxar-bound to inactive PAR1 (pink; PDB 3VW7). Inset: structural superposition illustrating the relative binding positions of vorapaxar (gray sticks) and the tethered peptide (yellow sticks) to PAR1. B Extracellular view of the comparison between tethered ligand-bound active PAR1 (blue tubes) and vorapaxar-bound inactive PAR1 (pink tubes). The tethered peptide is colored in yellow spheres, while vorapaxar is shown in gray spheres. C Key residues conformations in PAR1 bound to the antagonist vorapaxar. D Key residue conformations in PAR1 bound to the tethered ligand.

Vorapaxar is a highly lipophilic molecule, speculated to enter the receptor through the lipid bilayer between TMs 6 and 7. In the vorapaxar-bound inactive state, TM6 and TM7 are spaced apart; upon receptor activation, these helices move towards each other by ~3 Å (Fig. 3A, B), eliminating the space necessary to accommodate vorapaxar’s ethyl carbamate tail, and possibly vorapaxar’s hypothesized entry into the helical bundle via the membrane.

The most significant structural change during activation is an inward movement of the extracellular side of TM5 by ~5.6 Å, narrowing the pocket between TM4 and TM5 that accommodates vorapaxar’s fluorophenyl pyridine tail (Fig. 3B, C). In contrast, residues of the tethered ligand bind within the relatively superficial orthosteric pocket formed by TM1, TM5, TM6, TM7, and ECL2 (Fig. 3D).

Unlike other Class A and B GPCRs, which typically exhibit significant activation-associated movements in the intracellular end of TM6 (such as an ~9 Å outward shift in rhodopsin²⁹ and ~16 Å in GLP1R when engaged with G proteins³⁰), PAR1 shows a more modest 4.7 Å outward movement of TM6 in the active state compared to the inactivated vorapaxar-bound state (Fig. 3A). Furthermore, the intracellular end of TM5 exhibits an ~2.6 Å outward movement in the ligand-bound state compared to the inactive state, underscoring the unique activation dynamics of PAR1 (Fig. 3A).

Significant conformational shifts induced by the tethered peptide are observed in Y183^3.33, F271^5.39, Y337^6.59, Y350^7.32, and Y353^7.35 are also observed. Mutagenesis of these residues to alanine significantly diminishes the potency of the tethered ligand in activating PAR1³¹. In the vorapaxar-bound structure, Y350^7.32 is located at a position that is poised to form a hydrogen bond with H255^ECL2 on ECL2, stabilizing Y350^7.32 and helping to maintain TM7 in the inactive conformation (Fig. 3C). However, when bound, the tethered ligand forces these residues downward (Fig. 3D). The sidechain of Y350^7.32 on TM7 breaks the H-bond with H255 ^ECL2 and forms an H-bond with Y183^3.33 of TM3, leading to the downward movement of TM6 and subsequent receptor activation (Fig. 3A, B). Mutation of Y183^3.33 exhibited loss of inhibition by vorapaxar and indicates that interactions between Y183^3.33 and Y350^7.32 may stabilize an inactive conformation²².

Structural comparison of active and inactive states of PAR2

Previous structures of PAR2 have revealed an inactive conformation and detailed, distinct allosteric binding mechanisms of PAR2 antagonists. The antagonist AZ8838 (PDB 5NDD) demonstrates slow binding kinetics and binds to an occluded pocket²³ below the tethered ligand orthosteric pocket we describe here (Fig. 4A–C). Additionally, AZ3451 (PDB 5NDZ), was found to bind to a distant allosteric site on the outside the helical bundle, preventing necessary structural rearrangements for receptor activation and signaling^23,32.

Fig. 4. Structural Comparison of Active and Inactive States of PAR2.

A Overlay of the tethered ligand-bound active PAR2 (coral) and the antagonist AZ8838-bound inactive PAR2 (yellow; PDB 5NDD). The tethered peptide is colored in green spheres, while AZ8838 is shown in blue spheres. B Rotated extracellular view of the comparison between tethered ligand-bound active PAR2 (coral) and AZ8838-bound inactive PAR2 (yellow). C Conformations of key residues in inactive PAR2 (yellow) bound to the antagonist AZ8838 (blue) compared to active PAR2 (coral). D Detailed view of specific residues involved in H-bond rearrangements in the inactive (yellow) to active conformation (coral) of PAR2. E Detailed view of TM6 and TM7 highlighting residues that propagate agonist binding rearrangements in the active state (coral) compared to relative position in the inactive state (yellow). Red arrows denote steric forces applied by moving residues on neighboring residues.

In our tethered ligand bound PAR2 structure, we also observe a superficial orthosteric site that is comparable to PAR1 (Supplementary Fig. 5) and dissimilar to other Class A or Class B GPCRs. Analogous to our PAR1-Gαq complex, activated PAR2 exhibits minimal movement of TM6, with a 3.4 Å outward tilt and a 2 Å downwards movement towards the intracellular side. Additionally, TM5 exhibits an ~2.6 Å outward movement to accommodate the helix-5 of the G protein (Fig. 4A, B).

The tethered ligand occupies a non-overlapping site to that of AZ8838, which is completely solvent inaccessible in a small, deep binding pocket below the orthosteric site of the tethered ligand (Fig. 4C). Concomitant with this key difference, the tethered ligand induces a different conformation of TM6 around the binding pocket, causing an ~2.1 Å shift on the extracellular side of TM6 (Fig. 4B). Upon binding at the orthosteric pocket, the sidechain of L38^TL sterically pushes Y323^7.32 of TM7 downwards, forming a stabilizing H-bond with D228^ECL2 via the sidechain hydroxyl group (Fig. 4D). This movement causes another sterically driven downward movement of Y326^7.35 of TM7. Commensurate with activation is the breaking of an H-bond between H227^ECL2 and Y156^3.33 and the formation of a stabilizing H-bond between the sidechains of Y326^7.35 and Y156^3.33 (Fig. 4D). Mutations in Y323^7.32, D228 ^ECL2, Y156^3.33, and Y326^7.35 each were shown to have critical impacts on PAR2 activation^33,34, which correlates with our structural observations. Once stabilized, Y326^7.35 of TM7 sterically pushes TM6 outwards through interactions with L307^6.55 of TM6, as well as further propagating the shift of TM7 downward via L330^7.39 to L332^7.41 leading to movement of I298^6.46 further down TM6 (Fig. 4E).

Our structures reveal how tethered ligand induced reorientations of TM6 and TM7 lead to the transmission of the activation signal through a series of conserved tyrosine residues. They further reveal how binding at the orthosteric site results in coupled activation of G-protein binding on the intracellular side of PARs.

G protein coupling and activation mechanism

PAR1 and PAR2 exhibit similar G-protein coupling profiles, both leading to strong Gαq-mediated signaling pathways^35,36. Nevertheless, structural alignments between Gαq in PAR1 and PAR2 complexes reveal notable conformational differences. The α5 helix of Gαq in both receptors shows a tilt of ~7°, and a spatial separation of ~6.4 Å is observed in their respective αN helices (Fig. 5). These findings highlight the intricate and receptor-specific nature of GPCR-G protein interactions, emphasizing that insights from one receptor subtype cannot be directly transferred to another. Our study elucidates G-protein coupling specificities in PAR1 and PAR2, revealing the complex interplay of structural elements that govern these interactions.

Fig. 5. Interaction between PAR1 and PAR2 with Gq.

A Structural based alignment (using the receptor chains) of active PAR1-Gαq (blue and green cartoons) and PAR2-Gαq (coral and beige). Comparative analysis of the Gαq conformation in the α5 helix (B) and αN domain (C).

Our agonist bound structures and comparison with antagonists bound structures of PAR1 and PAR2 reveals a shared activation mechanism involving the downward movement of TM7 and subsequent displacement of TM6 (Fig. 6). The key ligand-induced driver of TM7 movement is the downward force exerted by the second residue of the tethered ligand (L38^TL in PAR1 and F43^TL in PAR2) directly on Y350^7.32 in PAR1 and Y323^7.32 in PAR2. This then promotes the sliding downward movement of TM6 through concomitate downward force on Y353^7.35 and Y326^7.35 in PAR1 and 2 respectively. The downward shift of TM6 and TM7 sets in motion a cascade of rearrangements of other bulky sidechains, most notably F182^3.32, Y183^3.33, M186^3.36 and the PIF motif residue I190^3.40 in PAR1 and similarly the corresponding conserved residues F155^3.32, Y156^3.33, M159^3.36 and I163^3.40 in PAR2. These rearrangements proliferate agonist binding signaling to canonical GPCR motifs, that are crucial for G protein coupling, and consisting of residues (PAR1/PAR2, respectively) P282/254, I190/I163 and F322/Y296 (P^5.50 I^3.40 F/Y^6.44; PIF/Y motif), D367/D340, P368/P341 and Y371/Y344 (D^7.49 P^7.50xxY^7.53; DPxxY motif), and D199/Q172, R200/R173 and F201/Y174 (D/N^3.49 R^3.50 F^3.51; DRF/NRY motif). The rearrangements of these motifs are highly analogous between PAR1 and PAR2, concomitant with their critical role in stabilizing the G-protein binding interface. Correspondingly, in PAR1, mutations of these residues were shown to significantly impede the activation of PAR1 and disrupted G protein signaling³¹.

Fig. 6. PAR1 and PAR2 activation and inactivation.

Schematic representation of PAR1 and PAR2 in antagonist-bound (left; A), ligand-free (middle; B), and agonist-bound (right; C) states. The sidechains of two mechanistically important tyrosines are shown in stick representation. Conformational flexibility in TM7, TM5 and TM6 in the three states are indicated with red arrows, while conformational movements of the tyrosines upon agonist and antagonist binding are highlighted with black arrows in (C). The relative location of the PIF/Y, DRF/NRY and DPxxY motifs are indicated with blue arrows.

Structure determination and overall structure of PAR2 bound to GB88

GB88 is an intriguing molecule, that was initially reported as a PAR2 selective antagonist³⁷. The true functional profile of the compound has recently been shown to be nuanced, demonstrating partial agonist properties against several PAR2 signaling arms, including Gαq/11 directed pathways, and the ability to internalize the receptor post activation²⁴. To gain insight into GB88’s ability to act as a small molecule PAR2 agonist, we determined the structure of the co-complex formed between GB88, PAR2, and heterotrimeric G-proteins using a chimeric Gαq (Fig. 7A, B).

Fig. 7. Cryo-EM structure of the GB88:PAR2-Gαq-scFv16 complex.

A Cryo-EM density map of the GB88:PAR2-Gαq-scFv16 complex perpendicular to the membrane (B). Model of GB88:PAR2-Gαq-scFv16 complex shown in cartoon representation. C Cartoon representation of PAR2 (green) with GB88 bound (cyan sticks) viewing from the extracellular side of the membrane with TMs and ECLs labeled. D GB88 (cyan sticks) is shown within the orthosteric pocket of PAR2 (slabbed surface). The GB88 cryo-EM map is shown as blue mesh at 5.0 sigma (E). GB88 (cyan) bound to the orthosteric pocket of PAR2 highlighting key residues (yellow) and H-bonds formed between the ligand and receptor. F Structural overlay (receptor not shown) of the GB88 and the tethered agonist structures to demonstrate GB88’s mimicry of the tethered ligand within the orthosteric site. GB88 is shown in cyan with chemical substituents labeled also in cyan. The tethered ligand is shown in green with residues labeled also in green.

Our structure reveals that GB88 binds with its isoxazole group deeply situated within the orthosteric pocket of PAR2 (Fig. 7C–E), burying 73.2% of the molecule’s accessible surface area (590 Ų) and interfacing with 22 residues from ECL2, TM1, TM6 and TM7 of PAR2. Many of these interactions are contributed by ECL2 residues in the region 217-236. Intriguingly, GB88 overlays very closely with the bound tethered ligand (Fig. 7F) forming several common interactions, most notably with ECL2 residues D228^ECL2 and L230 ^ECL2 and residues H310^6.58, S320^ECL3 and Y311^6.59 from ECL3 and TM6 (Fig. 7E).

The isoxazole group of GB88 has been shown to confer antagonist properties³⁸. In our structure, the isoxazole oxygen accepts an H-bond from H310^6.58 with the nitrogen accepting an H-bond from Y311^6.59. Adjacent to this, the bulky cyclohexane is positioned where L38^TL of the tethered agonist resides, pushing L69^N-term and shifting the top of TM1 outwards by ~2-3 Å when compared to the tethered ligand bound structure, to accommodate the bulky C-terminus spiroindene-piperdidine. Of note is the almost identical interaction with the activation residue Y323^7.32, which propagates TM7 movement during receptor activation. Similar to interactions formed by the tethered ligand agonist, the backbone amides of GB88 form H-bond interactions by donating to the carbonyl oxygen of D228^ECL2 and accepting H-bonds from both L230 ^ECL2 and S320 ^ECL3. Previous studies demonstrated key SAR around both of GB88’s backbone amides³⁸. Our structure supports the proposed hypothesis that this is due to preference for maintaining a preferred conformation for hydrogen bonding to receptor residues. Next, GB88’s isoleucine moiety mimics the position of I39^TL in the activated PAR2 structure, partially filling in the hydrophobic pocket formed between V236^5.32 and I314^6.62.While the structure does not readily reveal the basis for GB88 selectivity for PAR2 over PAR1, it is of note that two regions of the molecule appear to be less compatible with PAR1’s orthosteric pocket, namely in the bulky regions of the cyclohexane and the C-terminal spiroindene -piperdidine.

Discussion

A robust understanding of the molecular basis of GPCR agonism and antagonism, as well as orthosteric pocket information and ligand-receptor interactions are critical for leveraging rational SBDD during drug discovery campaigns. Until now, such information has largely not been available for the PARs. In this study, we first sought to advance our understanding of the structural mechanisms underlying the activation of PAR1 and PAR2 by their endogenous tethered ligand. Our cryo-EM structures provide a detailed view of the ligand-binding sites in PAR1 and PAR2, outlining residues that define the location and dimensions of their relative orthosteric binding pockets. Unlike previous findings of PAR1 and PAR2 in complex with small molecules^17,18 the binding pockets of these proteins for endogenous ligands are shallow and much closer to the extracellular side of the receptor than previous hypothesized³². Unlike most Class A and Class B GPCRs, the movement of TM6 in PAR1 and PAR2 is surprisingly minimal, indicating a G protein coupling mechanism that is so far unique to the PARs. Both our endogenous peptide bound structures reveal a common activation mechanism, where the tethered ligand sterically pushes two key tyrosine residues, which are conserved across all 4 human PARs, downward, leading to TM6 movement and subsequent activation. Given this, it is highly likely that all four PARs share a conserved orthosteric activation mechanism triggered by downward movement of conserved tyrosine residues on TM7 and TM6.

The constricted nature of the orthosteric pocket in the region of these bulky agonist switches rationalizes the limited successes in discovery of a clinically successful small molecule antagonist against PAR2 activity. Vorapaxar appears to circumvent triggering orthosteric activation of PAR1, as it is hypothesized to bind the receptor laterally through the membrane²². The crystal structure demonstrates that vorapaxar binds to PAR1 in an atypical manner, using a superficial binding pocket with minimal solvent exposure. Once bound, it sterically prevents the movement of tyrosine switches to initiate activation, explaining how a small molecule can achieve receptor inactivation²².

Unlike PAR1, the search for a PAR2 selective orthosteric antagonist has proven extremely difficult¹⁷. Here, we present the structure of PAR2 bound to the small molecule GB88, an intriguing orthosteric engager that selectively inhibits the PAR2/Gq/11/Ca2 + /PKC signaling pathway, resulting in anti-inflammatory effects in vivo²⁴. Potent orthosteric engagement is achieved by GB88’s ability to mimic many of the same interactions that the tethered agonist forms with orthosteric pocket residues located on ECL2, TM1, TM6 and TM7. A structural understanding of GB88’s unique pharmacology holds the potential to rationally develop improved small molecule engagers that can target specific PAR2-linked signaling pathways implicated in diseases without disrupting beneficial PAR2 signaling in normal physiology.

During review of this manuscript, two cryo-EM structures of PAR1 bound to the tethered ligand agonist in complex with Gq and Gi heterotrimers were reported³¹. The overall structure of the PAR1–G_q complex is like that of the PAR1–G_i complex; suggesting the activation mechanism of PAR1 is not likely dependent on the G protein. Overall, these structures are in strong agreement with this study in several key findings, including similar overall structure, proposed activation mechanism, and key residues involved in tethered ligand binding.

Methods

Constructs

The human PAR1 and PAR2 gene were subcloned into the pFastBac plasmid with a prolactin-signal peptide sequence on its N-terminus and the LgBiT fused to its c-terminus followed by a Flag tag for purification. The first 41 residues (residue 1-41) of PAR1 and the first 36 residues (residue 1-36) of PAR2 were truncated to mimic the protease cleavage. HiBiT was fused to the c-terminus of human Gβ₁ and cloned into pFastBac plasmid as described in the VIP1R paper²⁵ The N-terminus (residue 1-32) of human Gα_q was replaced by the N-terminus of G_i (residue 1–28) and subcloned into pFastBac plasmid. The wild-type human Gγ₂ was cloned into pFastBac plasmid. The scFv16 that encodes the single-chain variable fragment of mAb16 was subcloned into pFastBac plasmid.

Expression and purification of PAR1-Gαq, PAR2- Gαq and PAR2-GB88- Gαq complex

Bacmid preparation and virus production were performed according to the Bac-to-Bac baculovirus system manual (Gibco, Invitrogen). For expression, the Spodoptera frugiperda (Sf9) cells at density of 2 × 10⁶ cells per ml were co-infected with baculovirus encoding the PAR1-LgBiT-Flag, G_qiN, Gβ_, Gγ, scFv16 and Ric8A protein at a ratio of 1:500 (virus volume vs cells volume). Cells were harvested 48 h after infection. Cell pellets were resuspended in 20 mM Hepes buffer (pH 7.5), 100 mM NaCl, and homogenized by douncing ~30 times. Apyase was added to the lysis at a final concentration of 0.5 mU/ml. To keep the complex stable, the lysate was incubated at room temperature for 1 h with flipping. Then, lauryl maltose neopentyl glycol (LMNG, Anatrace) was added at the final concentration of 0.5% to solubilize the membrane at 4 °C for 2 h. Then the lysis was ultracentrifuged at 56,000 g (45,000 rpm) at 4 °C for 40 min. The supernatant was collected and incubated with anti-FLAG M2 resin for 2 h. The resin was washed with a buffer of 25 mM Hepes (pH 7.5), 100 mM NaCl and 0.001% LMNG, and 0.0005% cholesteryl hemi-succinate (CHS), then eluted with the same buffer plus 100 µb/ml FLAG peptide. The elution was concentrated and separated on a Superdex 200 Increase 10/300 GL (GE health science) gel infiltration column with a buffer of 25 mM Hepes (pH 7.5), 100 mM NaCl, and 0.001% LMNG. The peak corresponding to the PAR1-Gαq and PAR2-Gαq complex was concentrated at about 3 mg/ml for later cryo-EM grid preparation.

For the GB88-bound PAR2-Gαq complex, the same procedure as the tethered ligand PAR2 complex described above was followed, with the incorporation of 10 µM GB88 into all purification steps.

Grid preparation and cryo-EM data collection

Three microliters of PAR1-Gαq and PAR2-Gαq complex (including the GB88 preparation) sample at ~3 mg/ml was applied to a glow discharged UltraFoil R1.2/1.3 grids (Quantifoil GmbH). The grids were vitrified in liquid ethane using Vitrobot Mark IV (Thermo Fisher Scientific) instrument with the settings of blot force 5, blot time of 5 s, relative humidity of 100%, chamber temperature of 4 °C. Single-particle cryo-EM data were collected using EPU package on a Thermo Fisher Glacios transmission electron microscope operating at 200 keV using parallel illumination conditions. Micrographs were collected using a Falcon 4i direct electron detector with corrected pixel size of 0.69 Å/pix, a total electron exposure of 45 e^– /Å^–2, with a defocus range of −0.8 to −2 μm.

Cryo-EM data processing

The cryoSPARC Live application of cryoSPARC v. 4 (Structura Biotechnology) was used to streamline the movie processing, CTF estimation, particle picking, and 2D classification. Preprocessing involved anisotropic motion correction and local CTF estimation. The data were curated by only retaining images with better than 4.5 Å CTF fit resolution. Particle picking started with blob picking ~150 Å in diameter. Once a small set of particles was extracted and 2D class averages were obtained, the good 2D classes were used as templates for picking on the entire dataset. From the ~7 million (PAR1), ~6 million (PAR2) particles initially picked, ~3.1 million, ~1.9 million particles were kept from the good 2D classes, respectively. Then ab initio 3D reconstruction was carried out in cryoSPARC v. 4 and one out of three classes was giving the reconstruction of the designed complex from ~1.9 million and ~0.4 million particles for PAR1 and PAR2, respectively. Multiple rounds of iterative homogenous refinement, heterogenous refinement, and nonuniform, coupled with global CTF refinements gave the final 3D reconstructions at 2.74 Å (PAR1) and 3.1 Å (PAR2) resolution, respectively (0.143 gold standard FSC with correction of masking effects)³⁹. DeepEMHancer was used to improve the map quality⁴⁰. A similar strategy was utilized for the PAR2-GB88 complex, where ~4 million particles were initially picked and ~1 million particles were kept from the suitable 2D classes (Supplementary Table 1). During pre-processing with cryoSPARC Live, templated from the PAR2 dataset were used for particle picking. These picks were used in 2 cycles of heterogeneous refinement using models from the PAR1 dataset as seed. The best set of particles was then used for an ab initio reconstruction. Particles from the best model were used in homogeneous and non-uniform refinement, resulting in a 2.9 Å resolution map. DeepEMHancer was further utilized to help with the map quality.

Model building and refinement

Starting models for PAR1, PAR2, Gαq, Gβ₁γ₂, and scFv16 were based on Protein Data Bank (PDB) entries 3VW7²², 5NDD²³, 6OIJ⁴¹, respectively. All models were docked into the electron microscopy density map using UCSF ChimeraX⁴². The resulting model was subjected to iterative manual adjustment using Coot⁴³, followed by a Rosetta cryo-EM refinement at relax model and Phenix real_space refinement⁴⁴. The model statistics were validated using MolProbity. All structure alignments described were done by aligning the GPCR only instead of the whole complex in UCSF Chimera. Structural figures were prepared in UCSF ChimeraX and PyMOL (https://pymol.org/2/). The statistics for data collection and refinement are included in Supplementary Table 1.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Z.L., X.L., and Y.K. designed the research. Z.L., X.L., and Y.K. performed experiments. Z.L., X.L., G.X., A.P.M., D.C., V.A., M.F., N.A., H.Z., A.G.M., and Y.K. analyzed data. Z.L., X.L., A.P.M., and Y.K. wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Fei Xu and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The cryo-EM maps have been deposited in the Electron Microscopy Data Bank (EMDB) under accession codes under accession codes EMD-46571 (PAR1-Gαq), EMD-46448 (PAR2-Gαq) and EMD-47687 (PAR2-GB88). The coordinates have been deposited in the Protein Data Bank (PDB) under accession codes 9D4Z (PAR1-Gαq), 9D0A (PAR2-Gαq) and 9E7R (PAR2-GB88).

Competing interests

All authors are current or former employees of Takeda and own stock/stock options in the company.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-59138-x.