ARTICLEdoi:10.1038/nature11701

High-resolution crystal structure ofhuman protease-activated receptor 1Cheng Zhang1, Yoga Srinivasan2, Daniel H. Arlow3, Juan Jose Fung1{, Daniel Palmer2, Yaowu Zheng2{, Hillary F. Green3,Anjali Pandey4, Ron O. Dror3, David E. Shaw3, William I. Weis1,5, Shaun R. Coughlin2 & Brian K. Kobilka1

Protease-activated receptor 1 (PAR1) is the prototypical member of a family of G-protein-coupled receptors that mediatecellular responses to thrombin and related proteases. Thrombin irreversibly activates PAR1 by cleaving theamino-terminal exodomain of the receptor, which exposes a tethered peptide ligand that binds the heptahelicalbundle of the receptor to affect G-protein activation. Here we report the 2.2-A-resolution crystal structure of humanPAR1 bound to vorapaxar, a PAR1 antagonist. The structure reveals an unusual mode of drug binding that explains how asmall molecule binds virtually irreversibly to inhibit receptor activation by the tethered ligand of PAR1. In contrast todeep, solvent-exposed binding pockets observed in other peptide-activated G-protein-coupled receptors, thevorapaxar-binding pocket is superficial but has little surface exposed to the aqueous solvent. Protease-activatedreceptors are important targets for drug development. The structure reported here will aid the development ofimproved PAR1 antagonists and the discovery of antagonists to other members of this receptor family.

Protease-activated receptors (PARs) are G-protein-coupled receptors(GPCRs) that mediate cellular responses to specific proteases1,2. Thecoagulation protease thrombin activates the prototypical PAR, PAR1,by specific cleavage of the N-terminal exodomain of the receptor togenerate a new N terminus. This new N terminus then functions as atethered peptide agonist that binds intramolecularly to the seven-transmembrane helix bundle of the receptor to affect G-proteinactivation1,3–8 (Fig. 1a). In adult mammals, the four members of thePAR family link tissue injury and local generation of active coagu-lation proteases to cellular responses that help to orchestrate hae-mostasis, thrombosis, inflammation and perhaps tissue repair2,9.PARs may also participate in the progression of specific cancers10,11.

In contrast to a typical receptor–agonist binding interaction, theinteraction of PAR1 with its activator, thrombin, is that of a proteasesubstrate, with thrombin binding transiently to the receptor, cleavingit, then dissociating1,3–7,12. Proteolytic unmasking of the tethered pep-tide agonist of the receptor is irreversible, and although a free syn-thetic hexapeptide with the amino acid sequence of the tetheredagonist (SFLLRN) can activate the receptor with half-maximum effec-tive concentration (EC50) values in the 3–10-mM range, the localconcentration of the tethered agonist peptide is estimated to be about0.4 mM. Accordingly, PAR signalling must be actively terminated13–15

and, unlike most other GPCRs that can go though many rounds ofactivation by reversible diffusible hormones and neurotransmitters,PARs are degraded after a single activation6,13–17. Identification of effec-tive PAR antagonists has been challenging because low molecular masscompounds must compete with the very high local concentration ofthe tethered agonist generated by proteolytic cleavage.

Vorapaxar is a highly specific, virtually irreversible PAR1 antago-nist18 (Supplementary Fig. 1). In a phase III trial, vorapaxar protectedpatients against recurrent myocardial infarction at a cost of increasedbleeding19,20. Given the latter, an antagonist that is reversible in thesetting of bleeding might be desirable. Although the very slow dissoci-ation rate of vorapaxar from PAR1 probably accounts for its ability to

1Department of Molecular and Cellular Physiology, Stanford University School of Medicine, Stanford, California 94305, USA. 2Cardiovascular Research Institute, University of California, San Francisco, 555Mission Bay Boulevard South, S452P, San Francisco, California 94158, USA. 3D. E. Shaw Research, New York, New York 10036, USA. 4Portola Pharmaceuticals, 270 East Grand Avenue, South SanFrancisco, California 94080, USA. 5Department of Structural Biology, Stanford University School of Medicine, 299 Campus Drive, Stanford, California 94305, USA. {Present addresses: ProNovusBioscience, LLC. 544 E Weddell Drive, Sunnyvale, California 94089, USA (J.J.F.); Northeast Normal University, Changchun, Jilin 130024, China (Y.Z.).

SFLLRN

Thrombin

Gi

GqG12/13

Extracellular

Intracellular

Inactive PAR1 Active PAR1

N

C C

N R L L F S

a

b

c60°

90°

ECL2

TM4

TM3

TM5

TM1

TM2

TM3

TM4

TM5TM6

TM7

ECL1

Disulphide

Extracellular

N

N

ECL2

TM6

TM2

TM7

Figure 1 | PAR1 activation and overall structure of human PAR1 complexwith antagonist vorapaxar. a, Thrombin cleaves the PAR1 N terminus andexposes a new N-terminal peptide, SFLLRN, which can bind to and activate thetransmembrane core of PAR1. PAR1 can activate several G proteins includingGi, G12/13 and Gq. b, Overall view of the human PAR1 structure and theextracellular surface. The receptor is shown as blue ribbon and vorapaxar isshown as green spheres. Monoolein is shown in orange, water in red. Thedisulphide bond is shown as a yellow stick. c, Surface view of the ligand-bindingpocket viewed from two different perspectives. The vorapaxar-binding pocket isclose to the extracellular surface but not well exposed to the extracellular solvent.

0 0 M O N T H 2 0 1 2 | V O L 0 0 0 | N A T U R E | 1

Macmillan Publishers Limited. All rights reserved©2012

inhibit receptor activation by its tethered agonist peptide, it may bepossible to develop a drug with an off rate slow enough to block sig-nalling but fast enough to allow useful reversal after cessation of drug.

In an effort to advance our understanding of PAR1 structure andfunction and to provide a foundation for discovery of new agents toadvance the pharmacology of PARs, we obtained the crystal structureof vorapaxar-bound human PAR1.

Crystallization of human PAR1To facilitate crystallogenesis, T4 lysozyme (T4L) was inserted intointracellular loop 3 (ICL3) in human PAR1, the N-linked glycosyla-tion sites in ECL2 were mutated21, and the N-terminal exodomain wasremoved by site-specific cleavage at a tobacco etch virus (TEV) prote-ase site introduced between amino acids 85 and 86 (ref. 4; Sup-plementary Fig. 2). The structure of human PAR1–T4L bound tovorapaxar was determined to 2.2 A by merging diffraction data setsfrom 18 crystals grown in the lipidic cubic phase (Supplementary Figs3 and 4). Details of data collection and structure refinement are listedin Supplementary Table 1.

PAR1 has the expected seven-transmembrane segment bundle(Fig. 1b). There are several lipid molecules assigned as monooleinfrom lipidic cubic phase in the structure (Fig. 1b), but no orderedcholesterol molecules were observed. The remaining N-terminal frag-ment Arg 86–Glu 90 and a part of the ICL2 from Gln 209 to Trp 213are not modelled in the structure because of the weak electron density.There is no clear electron density for residues after Cys 378, and nohelix 8 is observed after transmembrane segment 7 (TM7) in thestructure. Whether this reflects a lack of a helix 8 in PAR1 in its nativestate or conditions in the crystal is not known.

Cys 1753.25 in helix III and Cys 254 in extracellular loop 2 (ECL2)form a conserved disulphide bond (Figs 1b and 2a). Amino-terminalto Cys 254, ECL2 loops outwards in two anti-parallel b strands. Thisstructural feature is found in other peptide receptors, including theCXCR4 receptor and the opioid receptors22–25, despite absence ofamino acid sequence homology among these receptors in ECL2(Supplementary Fig. 5). In contrast to the open, solvent-exposedbinding pocket observed in the m-opioid receptor (MOR) and otherpeptide receptors, access to the vorapaxar-binding pocket is restrictedby the central location of ECL2 (Fig. 1 and Supplementary Figs 5 and6), which almost completely covers the extracellular-facing surface ofvorapaxar. ECL2 is anchored in this position by hydrogen bondsbetween His 255 in ECL2 and Tyr 3537.35 in TM7, and betweenAsp 256 in ECL2 and Tyr 95 in the N terminus (SupplementaryFig. 6), and by extensive interactions with vorapaxar (Fig. 2). Thecovered vorapaxar-binding pocket in PAR1 more closely resemblesrhodopsin and the lipid-activated sphingosine-1-phosphate receptor(S1P1) than other peptide-activated GPCRs (Supplementary Fig. 5).

Divergence of PAR1 from other family A GPCRsMembers of the family A GPCRs share a set of conserved amino acidsthat are thought to be important in signal transduction26,27. Specificresidues in the highly conserved FXXCWXP motif (in which Xdenotes any amino acid) in TM6, and the NPXXY motif in TM7undergo structural rearrangements during activation of the b2 adre-nergic receptor (b2AR) (Fig. 3b, d). However, based on a phylogeneticanalysis of amino acid sequences28, PAR1 is a more distant relative ofthe family A GPCRs that have been crystallized thus far. PAR1belongs to the d-subfamily, which includes the glycoprotein receptors,the purinergic receptors and the olfactory receptors28. The tryptophanresidue in FXXCW6.48XP proposed to act as a toggle switch duringactivation in some GPCRs28 is replaced by Phe6.48 in all PARs (Fig. 3a).Phe6.44, also highly conserved in family A GPCRs, is Phe6.44 in PAR1,but Tyr6.44 in PAR2 and Ala6.44 in PAR4. When comparing inactiveand active states of the b2AR, changes in packing interactions in-volving Pro5.50, Ile3.40 and Phe6.44 seem to have a role in structuralchanges needed to accommodate G-protein binding29–31. Packing

interactions of the corresponding residues Pro 2825.50, Ile 1903.40

and Phe 3226.44 in the PAR1 differ from those in both active andinactive b2AR structures (Fig. 3b). Taken together, these differencessuggest that PAR1 may differ from other family A GPCRs in themechanism by which signals propagate from the extracellular pep-tide-binding interface to the cytoplasmic domains that interact with Gproteins and other signalling molecules.

The NP7.50XXY motif at the end of helix VII observed in mostfamily A GPCRs is DP7.50XXY in PAR1. This region undergoesstructural rearrangement after activation of the b2AR. In PAR1,Asp 3677.49 and Tyr 3717.53 form hydrogen bonds with residues inTM2 and TM1 (Fig. 3c). The hydrogen-bonding network associatedwith Asp 3677.49 is extensive and includes several water molecules anda putative sodium ion. Na1, rather than a water molecule, wasassigned to this region of electron density as it has five oxygen neigh-bours and short distances to its oxygen ligands (average refined dis-tance 2.4 A), both consistent with known Na1–oxygen interactions32,and it interacts with two acidic side chains that, assuming deproto-nated states, would repel one another without charge neutralizationprovided by the Na1. The sodium ion also interacts with a conservedAsp 1482.50 in TM2 and Ser 1893.39 in TM3, with two water moleculesnearby (Fig. 3c). Na1 is an allosteric modulator for several family AGPCRs such as the a2A adrenergic receptor, A2A adenosine receptor,m- and d-opioid receptors and D2 dopamine receptor33–37. The con-served Asp2.50 is necessary for sodium sensitivity of thea2A adrenergicreceptor37 and D2 dopamine receptor36,38. In PAR1, Asp 3677.49 mightbe expected to form a stronger hydrogen-bonding network andsodium coordination site than asparagine residues found in mostother family A GPCRs. This more stable network may contributeto the unusual position of the cytoplasmic end of TM7 that is dis-placed inward towards TM2. This position is more similar to the

a

c

b

aL340L340

Y337Y337

F271F271Y183Y183

Y353Y353

A349A349

H336H336

L258L258

TM6TM6

TM5TM5

TM7TM7

TM4TM4TM3TM3

H255H255

ECL2ECL2

H255H255 Y337Y337

F271F271Y183Y183

Y353Y353L332L332

L333L333

TM6TM6

TM5TM5

TM7TM7

TM3TM3

L258L258

L262L262

Y350Y350

L263L263L262L262

H255H255

ECL2ECL2

TM5TM5

ECL2ECL2

TM2TM2

TM1TM1

TM6TM6

DisulphideDisulphide

F274F274

F278F278F271F271

L340

Y337

F271Y183

Y353

A349

H336

L258

TM6

TM5

TM7

TM4TM3

H255

ECL2

H255 Y337

F271F271Y183

Y353L332

L333

TM6

TM5

TM7

TM3

L258

L262

Y350

L263L262

H255

F271

ECL2

TM5

ECL2

Y337

TM2

TM1

TM6

Disulphide

F274

F278F271

Figure 2 | Binding interactions of vorapaxar with human PAR1.a, b, Ligand-binding pocket viewed from the extracellular surface (a) and fromthe side of the transmembrane helix bundle (b). ECL2 is coloured in orange ina and b. Ligand vorapaxar is shown as green sticks. Water molecules are shownas red spheres. Hydrogen bonds are shown as dotted lines. c, Two residues,Leu 262 and Leu 263 in ECL2 (shown as dot surface), which pack againstresidues His 255, Phe 2715.39 and Tyr 3376.59 (shown in Corey–Pauling–Koltun(CPK) representation), may contribute to the selectivity of vorapaxar forhuman PAR1. Also shown are Phe 2745.42 and Phe 2785.46 in TM5 (shown asdot surface), which may indirectly influence vorapaxar binding by packinginteractions with Phe 2715.39.

RESEARCH ARTICLE

2 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 2

Macmillan Publishers Limited. All rights reserved©2012

active b2AR bound with either nanobody 80 or heterotrimeric Gprotein29,31 (Fig. 3d).

Structural insights into binding properties of vorapaxarVorapaxar binds in an unusual location very close to the extracellularsurface of PAR1. By contrast, ligands for other GPCRs penetrate moredeeply into the transmembrane core (Fig. 1 and Supplementary Figs 5and 7). The vorapaxar-binding pocket, composed of residues fromTM3, TM4, TM5, TM6 and TM7 as well as ECL2 and ECL3, forms atunnel across the receptor with one end open between TM4 and TM5and the other between TM6 and TM7 occupied by the ethyl carbamatetail of vorapaxar (Figs 1 and 2). There is only a small opening in theextracellular surface between ECL2 and ECL3. Details of interactionsbetween vorapaxar and PAR1 are illustrated in Fig. 2 and Sup-plementary Fig. 8.

Vorapaxar shows high selectivity for human PAR1 over humanPAR2 and PAR4, and mouse PAR1 in functional assays (Sup-plementary Fig. 9A, B). The structural basis for this selectivity is notreadily apparent from the crystal structure. Nearly all the residues thatinteract with vorapaxar in human PAR1 are conserved in humanPAR2, human PAR4 and mouse PAR1 (Supplementary Fig. 10).Residues Leu 262 and Leu 263, which are involved in weak hydro-phobic interactions with vorapaxar in human PAR1, are alanineand asparagine, respectively, in human PAR4, and Leu 263 is amethionine in mouse PAR1 (Fig. 2c and Supplementary Fig. 10).These differences by themselves would not be expected to explainthe high selectivity of vorapaxar. However, Leu 262 and Leu 263 pack

against other amino acids that have more extensive interactions withvorapaxar. Leu 262 interacts with His 255 in ECL2 and Leu 263 inter-acts with Phe 2715.39 at the top of TM5 and Tyr 3376.59 at the top ofTM6 (Fig. 2c). These interactions may influence ligand-bindingselectivity indirectly by contributing to the overall structure andstability of the binding pocket. Amino acid differences betweenPAR1, PAR2 and PAR4 more distant from the ligand-binding pocketmay also contribute to subtype-specific binding of vorapaxar.Phe 2745.42 is Leu in PAR2 and PAR4, and Phe 2785.46 is Val inPAR2 and Gly in PAR4. Although neither Phe 2745.42 nor Phe 2785.46

directly contact vorapaxar, Phe 2785.46 packs against Phe 2745.42, whichin turn packs against Phe 2715.39 in the binding pocket (Fig. 2c).

In human PAR2 and PAR4, ECL3 connecting TM6 and TM7 is oneresidue shorter than it is in PAR1 (Supplementary Fig. 9C).Tyr 3376.59 at the carboxy-terminal end of TM6 forms a strong hydro-gen bond with vorapaxar (Fig. 2 and Supplementary Fig. 9D).Another residue, Tyr 3537.35, which forms the base of the ligand-binding pocket together with Tyr 1833.33, is located at the N-terminal end of TM7 (Fig. 2 and Supplementary Fig. 9D). A shorterECL3 in human PAR2 and PAR4 may change the relative position ofthese amino acids, thereby altering the overall geometry of the bindingpocket. Although the length of ECL3 in mouse and human PAR1 isthe same, four of the eight amino acids are different (SupplementaryFig. 9C). These differences may affect the structure of ECL3 andthereby influence interactions between vorapaxar in Tyr 3376.59 andTyr 3537.35. Alternatively, these differences could have an effect on themechanism by which vorapaxar gains access to the binding pocket.

a

TM5

TM6

F3266.48

F3226.44

b

F3226.44

F3266.48I1903.40

P2825.50

c

N1853.35

S1893.39

D1482.50

N1201.50

D3677.49

Y3717.53 P3687.50TM3

TM2

TM7

TM1 TM7TM6

TM5

C-terminal

d

Figure 3 | Structure motifs in PAR1 compared with other family A GPCRs.a, Superimposition of TM5 and TM6 of human PAR1 (in blue) with those ofother GPCRs includingb2AR andb1AR, A2A adenosine receptor, dopamine D3

receptor, M2 muscarinic receptor, histamine H1 receptor, m-opioid receptor,S1P1 and CXCR4 (all in orange). Phe 3266.48 and Phe 3226.44 in theF6.44XXC(F)6.48XP motif in PAR1 are shown as sticks. This motif isFXXC(W)6.48XP in most other family A GPCRs. Phe 3266.48 and Phe 3226.44 areboth in different conformations compared to their counterparts in otherGPCRs. b, In the b2AR, rearrangements of three residues, Pro5.50, Ile3.40 andPhe6.44, are associated with receptor activation. Black arrows indicate changesof these residues going from inactive (cyan) to active (yellow) b2AR structures.The counterparts in the inactive state structure of PAR1 (Pro 2825.50, Ile 1903.40

and Phe 3226.44) are shown in blue. c, DP7.50XXY motif in TM7 and sodium-binding site in PAR1. Residues Asp 3677.49, Pro 3687.50 and Tyr 3717.53 in theDP7.50XXY motif are shown as cyan sticks. This motif is normally NPXXY inmost other family A GPCRs. Sodium is shown as a purple sphere and watermolecules are shown as red spheres. Polar interactions are shown as blackdashed lines. An Fo 2 Fc omit electron-density map for the putative sodium ionand water molecules contoured at 4s is shown as purple mesh.d, Superimposition of the C-terminal part of TM7 in the structure of humanPAR1 (blue), in the inactive structures of other GPCRs (all in orange)mentioned in a and in the active structure of b2AR (in magenta). TheC-terminal part of TM7 in PAR1 adopts a conformation more similar to thatobserved in the active state of the b2AR.

ARTICLE RESEARCH

0 0 M O N T H 2 0 1 2 | V O L 0 0 0 | N A T U R E | 3

Macmillan Publishers Limited. All rights reserved©2012

Structural insights into vorapaxar inhibition of PAR1Although this structure is compatible with the very slow dissociationrate of vorapaxar, it does not provide insight into the mechanism bywhich vorapaxar or an agonist peptide gains access to the bindingpocket. None of the three openings to the vorapaxar-binding pocket islarge enough to accommodate the passage of the ligand. We thuswondered whether the unliganded receptor might have a more openstructure, similar to that observed for opioid receptors, with uniqueinteractions between vorapaxar and PAR1 causing an otherwise openbinding pocket to close after vorapaxar binding. To investigate thisissue, we performed long-timescale molecular dynamics simulationsof PAR1 with and without vorapaxar bound. Intriguingly, removal ofthe ligand did not lead to a more open binding pocket, but instead toone that was even more closed (Fig. 4a, b, Supplementary Fig. 11 andSupplementary Table 2). The extracellular end of TM6 moved about4 A inward towards TM4, bringing ECL3 in full contact with ECL2and completely occluding the binding pocket. By contrast, in a similarstudy on the MOR, the binding pocket remained open when theligand was removed (Supplementary Fig. 11). The collapse of thevorapaxar-binding pocket may reflect the fact that both vorapaxarand its binding pocket are uncharged, whereas in the opioid receptorsand many other family A GPCRs, the charged residue Asp3.32 helps tokeep the binding pocket hydrated after the ligand is removed.

It is interesting to speculate that vorapaxar, a highly lipophilicmolecule, may access the binding pocket through the lipid bilayer,possibly between TM6 and TM7. This is similar to the bindingmode proposed for retinal to rhodopsin and the lipid sphingosine-1-phosphate (S1P) to the S1P1 receptor39,40 (Supplementary Fig. 5).

To understand the ability of vorapaxar to inhibit agonist bindingand activation, we examined the functional consequences of mutatingfour aromatic amino acids that form strong interactions with vora-paxar: Tyr 1833.33, Tyr 3537.35, Phe 2715.39 and Tyr 3376.59 (Fig. 2a, band Supplementary Fig. 3C). Three of these (Tyr 3537.35, Phe 2715.39

and Tyr 3376.59) assume substantially different positions in simula-tions of unliganded PAR1 (Fig. 4a). Tyrosine residues Tyr 1833.33 inTM3 and Tyr 3537.35 in TM7 are linked to each other by a hydrogenbond and form the base of the binding ‘tunnel’, and are part of a

hydrophobic cage that surrounds the ligand. Tyr 3537.35 also formsa hydrogen bond with His 255, the most deeply buried amino acid inECL2 (Fig. 2a and Supplementary Fig. 6). This interaction of His 255with Tyr 3537.35 contributes to the closed conformation of ECL2over the ligand-binding pocket. Phe 2715.39 interacts with the fluoro-phenyl ring and Tyr 3376.59 forms a strong hydrogen bond with thepyridine ring of vorapaxar.

Mutation of Tyr 3376.59 to phenylalanine and Tyr 3537.35 to alanineled to a reduction in cell surface expression, making it difficult tointerpret the associated reduction in agonist peptide activation(Fig. 4c, d). Mutation of Phe 2715.39 to alanine was associated withenhanced cell surface expression, but reduced activation by agonistpeptide. There was little effect of this mutation on maximal inhibitionby vorapaxar. Although not conclusive, this result suggests thatPhe 2715.39 may have a role in both peptide and vorapaxar binding.Of interest, mutation of Tyr 1833.33 exhibited enhanced response to theagonist peptide and loss of inhibition by vorapaxar (Fig. 4c). This resultsuggests a possible role for Tyr 1833.33 in maintaining the receptor in aninactive state and indicates that interactions between Tyr 1833.33 andvorapaxar may further stabilize an inactive conformation.

Activation of PAR1 by the agonist peptideOur structure is consistent with data from mutagenesis studies that sug-gest that the PAR1-tethered agonist peptide may activate the heptahe-lical bundle of the receptor by interacting with superficial structuresrather than penetrating deeply into the transmembrane core8,41–45.Glu 260, a solvent-exposed residue in ECL2 in both vorapaxar-boundand unliganded PAR1 (Fig. 5), is of particular interest, as evidence frommutagenesis studies suggests an interaction with Arg 46 in the tetheredpeptide SFLLRN8,44. Substitution of Glu 260 with arginine markedlyreduces activation of PAR1 by a peptide with the native tethered ligandsequence (SFLLRN) but facilitates activation by SFLLEN. Argininesubstitution of Glu 264, which is surface-exposed and near Glu 260 inthe structure, also facilitates activation by SFLLEN.

Mutation of other residues near the extracellular surface, includingLeu96Ala (N terminus), Asp256Ala (ECL2) and Glu3477.29Ala/Gln(ref. 45) (Fig. 5a), markedly reduces activation of PAR1 by the peptide

Y353

H336

Y337

L340

TM3

TM7

TM6

F271

TM5

cba

Y183

TM4

TM1

PAR1–vorapaxar PAR1 no ligand

Vorapaxar

PAR1

PAR1(

Y183A

)

PAR1(

F271

A)

PAR1(

Y337F

)

PAR1(

Y353A

)0

5,000

10,000

15,000

20,000

25,000

VehicleSFLLRN

SFLLRN + VOR (100 nM)

**

*

* NS NS

[3H

]-in

osito

l p

ho

sp

hate

s (c.p

.m.)

Unt

rans

fect

ed

PAR1

PAR1(

Y183A

)

PAR1(

F271

A)

PAR1(

Y337F

)

PAR1(

Y353A

)0.0

0.5

1.0

1.5

2.0

Surf

ace r

ecep

tor A

405 n

md

Figure 4 | Collapse of ligand-binding pocket in long-timescale moleculardynamics simulations of unliganded PAR1. Molecular dynamics simulationswere performed on PAR1 from which vorapaxar had been removed. Thevorapaxar-bound PAR1 crystal structure is shown in blue and the unligandedstructure obtained from molecular dynamics simulation is shown in grey.a, The largest differences between vorapaxar-bound and unliganded PAR1 areat the extracellular end of TM6 and in ECL3. Residues involved in vorapaxarbinding are shown as sticks. b, Surface view showing collapse of the ligand-binding pocket during molecular dynamics simulation in the absence ofvorapaxar. c, d, Signalling (c) and cell surface (d) expression for wild-type

human PAR1 and binding-site mutants. Cos7 cells expressing the indicatedreceptor constructs were labelled with [3H]-myoinositol, pretreated withvehicle or 100 nM vorapaxar in DMEM medium containing 0.1% BSA, 20 mMHEPES and 0.2% b-hydroxy cyclodextrin (to retain vorapaxar in solution) for1 h, then incubated with vehicle or PAR1 agonist (100mM SFLLRN) for 1 h at37 uC. Total [3H]-inositol-phosphate accumulation was measured. Surfaceexpression of receptors in cells transfected in parallel was assessed by measuringbinding of anti-Flag antibody to an epitope displayed at the N terminus of thereceptor. Error bars, mean 6 s.e. Results are representative of three separateexperiments. NS, not significant.

RESEARCH ARTICLE

4 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 2

Macmillan Publishers Limited. All rights reserved©2012

agonist. However, these mutations have only a small effect on agonistpeptide binding, with only Asp256Ala resulting in a more than tenfoldloss of binding affinity43. The positions of these residues do not change

substantially when comparing vorapaxar-bound and unliganded PAR1(Fig. 5b). Of interest, only Glu 3477.29 is surface exposed (Fig. 5a, b),suggesting that Leu 96 and Asp 256 may not interact directly with theagonist peptide or that these amino acids are more exposed than wouldappear from the molecular dynamics model of the unliganded receptor.In the inactive structure, Asp 256 forms a hydrogen bond with Tyr 95and helps to stabilize interactions between the C-terminal end of theN terminus and ECL2 (Fig. 5a).

Interestingly, substitution of human PAR1 sequence Asn 259–Ala 268, the region of ECL2 implicated in tethered ligand binding,with the cognate Xenopus ECL2 sequence results in an approximatelytenfold increase in basal activity46. Figure 5c shows the position ofamino acids that differ between human and Xenopus receptors inECL2 in both the crystal structure and the unliganded moleculardynamics simulation. The superficial location of these activatingmutations suggests that very superficial interactions between thetethered agonist peptide and the extracellular loops may be sufficientto activate PAR1. Taken together, these mutagenesis studies suggestthat the PAR1 agonist peptide may activate PAR1 by binding moresuperficially than do agonist peptides for opioid receptors. Alterna-tively, the tethered peptide may bind in a sequential manner, initiallyto the extracellular loops but penetrating more deeply into the core ofthe receptor through a sequence of conformational intermediates.

ConclusionThe unusual mode of activation and the paucity of pharmacologicaltools have made PAR1 one of the more challenging GPCRs to cha-racterize and a difficult target for drug development. The crystalstructure offers insights into the very high affinity interaction withthe antagonist vorapaxar. This structure will provide a template forthe development of PAR1 antagonists with better drug properties andthe development of antagonists for other PAR subtypes to probe theirbiological roles. The mechanism of activation of PAR1 remains poorlyunderstood. Molecular dynamics simulations of an unliganded recep-tor together with the location of amino acids known to influenceagonist peptide activity suggest that activation of PAR1 by its agonistpeptide may involve superficial interactions with extracellular loops.Future efforts will focus on an active-state structure of PAR1 bound toits tethered agonist peptide.

METHODS SUMMARYThe human PAR1–T4L fusion protein was expressed in Sf9 insect cells andpurified by nickel-affinity chromatography, Flag M1 antibody affinity chromato-graphy followed by size exclusion chromatography. PAR1–T4L crystals weregrown using the in meso crystallization method. The diffraction data were col-lected from 18 crystals at the GM/CA@APS beamline in the Argonne NationalLaboratory. The structure was solved by molecular replacement and refined inPhenix. Refinement statistics are provided in Supplementary Table 1. All-atommolecular dynamics simulations were performed on Anton47 with lipids andwater molecules represented explicitly. Phosphoinositide hydrolysis assays weredone in Cos7 cells transfected with wild-type and mutant PAR1. More details areprovided in Methods.

Full Methods and any associated references are available in the online version ofthe paper.

Received 6 August; accepted 22 October 2012.

Published online 9 December 2012.

1. Vu, T. K., Hung, D. T., Wheaton, V. I. & Coughlin, S. R. Molecular cloning of afunctional thrombin receptor reveals a novel proteolytic mechanism of receptoractivation. Cell 64, 1057–1068 (1991).

2. Coughlin, S. R. Thrombin signalling and protease-activated receptors. Nature 407,258–264 (2000).

3. Vu, T. K., Wheaton, V. I., Hung, D. T., Charo, I. & Coughlin, S. R. Domains specifyingthrombin-receptor interaction. Nature 353, 674–677 (1991).

4. Chen, J., Ishii, M., Wang, L., Ishii, K. & Coughlin, S. R. Thrombin receptor activation.Confirmationof the intramolecular tethered liganding hypothesis and discoveryofan alternative intermolecular liganding mode. J. Biol. Chem. 269, 16041–16045(1994).

E260

D256

L96

Y95

E347

ECL2

ECL3

TM6

TM7

TM1

a

E260

D256

TM6

TM1

L96

TM4

TM5

TM3

ECL2

ECL3

Y266

E264T261

L262

E260

S259

TM4 TM5

ECL2

b

ECL2

E3477.29

E264

PAR1–vorapaxar PAR1 no ligand

c

Figure 5 | Residues important for agonist peptide binding and receptoractivation. a, Mutations of residues Glu 260, Asp 256, Leu 96 and Glu 3477.29

near the extracellular surface have been shown to reduce activation of PAR1 bythe free agonist peptide. Among them only Glu 260 is completely exposed to thesolvent, whereas Asp 256 is the most deeply buried, forming a hydrogen bondwith residue Tyr 95. Although none of these amino acids forms part of thevorapaxar-binding pocket, Asp 256 forms a hydrogen bond with Tyr 95 thatmay stabilize ECL2 over the vorapaxar-binding pocket. Vorapaxar is shown ingreen. b, c, Superimposition of the unliganded molecular dynamics simulationmodel (grey) with the ligand-bound crystal structure (blue). In b, ResiduesGlu 260, Asp 256, Leu 96 and Glu 3477.29, which are important for agonistpeptide signalling, are in similar positions in both structures. c, The positions ofresidues that differ between human and Xenopus PAR1 in ECL2. Substitutionof these residues in human PAR1 with corresponding residues from XenopusPAR1 results in increased basal activity.

ARTICLE RESEARCH

0 0 M O N T H 2 0 1 2 | V O L 0 0 0 | N A T U R E | 5

Macmillan Publishers Limited. All rights reserved©2012

5. Liu, L. W., Vu, T. K., Esmon, C. T. & Coughlin, S. R. The region of the thrombinreceptor resembling hirudin binds to thrombin and alters enzyme specificity.J. Biol. Chem. 266, 16977–16980 (1991).

6. Ishii, K., Hein, L.,Kobilka,B.& Coughlin, S. R.Kinetics of thrombin receptor cleavageon intact cells. Relation to signaling. J. Biol. Chem. 268, 9780–9786 (1993).

7. Mathews, I. I. et al. Crystallographic structures of thrombin complexed withthrombin receptor peptides: existence of expected and novel binding modes.Biochemistry 33, 3266–3279 (1994).

8. Gerszten, R. E. et al. Specificity of the thrombin receptor for agonist peptide isdefined by its extracellular surface. Nature 368, 648–651 (1994).

9. Coughlin, S. R. Protease-activated receptors in hemostasis, thrombosis andvascular biology. J. Thromb. Haemost. 3, 1800–1814 (2005).

10. Garcıa-Lopez, M. T., Gutierrez-Rodriguez, M. & Herranz, R. Thrombin-activatedreceptors: promising targets for cancer therapy? Curr. Med. Chem. 17, 109–128(2010).

11. Ramachandran, R., Noorbakhsh, F., Defea, K. & Hollenberg, M. D. Targetingproteinase-activated receptors: therapeutic potential and challenges. Nature Rev.Drug Discov. 11, 69–86 (2012).

12. Ishii, K. et al. Determinants of thrombin receptor cleavage. Receptor domainsinvolved, specificity, and role of the P3 aspartate. J. Biol. Chem. 270,16435–16440 (1995).

13. Shapiro, M. J., Trejo, J., Zeng, D. & Coughlin, S. R. Role of the thrombin receptor’scytoplasmic tail in intracellular trafficking. Distinct determinants for agonist-triggered versus tonic internalization and intracellular localization. J. Biol. Chem.271, 32874–32880 (1996).

14. Trejo, J. & Coughlin, S. R. The cytoplasmic tails of protease-activated receptor-1and substance P receptor specify sorting to lysosomes versus recycling. J. Biol.Chem. 274, 2216–2224 (1999).

15. Trejo, J., Hammes, S. R. & Coughlin, S. R. Termination of signaling by protease-activated receptor-1 is linked to lysosomal sorting. Proc. Natl Acad. Sci. USA 95,13698–13702 (1998).

16. Shapiro, M. J. & Coughlin, S. R. Separate signals for agonist-independent andagonist-triggered trafficking of protease-activated receptor 1. J. Biol. Chem. 273,29009–29014 (1998).

17. Hein, L., Ishii, K., Coughlin, S. R. & Kobilka, B. K. Intracellular targeting andtrafficking of thrombin receptors. A novel mechanism for resensitization of a Gprotein-coupled receptor. J. Biol. Chem. 269, 27719–27726 (1994).

18. Chackalamannil, S. et al. Discovery of potent orally active thrombin receptor(protease activated receptor 1) antagonists asnovel antithrombotic agents. J.Med.Chem. 48, 5884–5887 (2005).

19. Morrow, D. A. et al. Vorapaxar in the secondary prevention of atherothromboticevents. N. Engl. J. Med. 366, 1404–1413 (2012).

20. Kosoglou, T. et al. Pharmacodynamics and pharmacokinetics of the novel PAR-1antagonist vorapaxar (formerly SCH 530348) in healthy subjects. Eur. J. Clin.Pharmacol. 68, 249–258 (2012).

21. Soto, A. G. & Trejo, J. N-linked glycosylation of protease-activated receptor-1second extracellular loop: a critical determinant for ligand-induced receptoractivation and internalization. J. Biol. Chem. 285, 18781–18793 (2010).

22. Wu, B. et al. Structures of the CXCR4 chemokine GPCR with small-molecule andcyclic peptide antagonists. Science 330, 1066–1071 (2010).

23. Manglik, A. et al. Crystal structure of the micro-opioid receptor bound to amorphinan antagonist. Nature 485, 321–326 (2012).

24. Wu, H. et al. Structure of the human k-opioid receptor in complex with JDTic.Nature 485, 327–332 (2012).

25. Granier, S. et al. Structure of the d-opioid receptor bound to naltrindole. Nature485, 400–404 (2012).

26. Smit, M. J. et al. Pharmacogenomic and structural analysis of constitutive gprotein-coupled receptoractivity.Annu.Rev. Pharmacol. Toxicol. 47, 53–87 (2007).

27. Schwartz, T. W., Frimurer, T. M., Holst, B., Rosenkilde, M. M. & Elling, C. E. Molecularmechanism of 7TM receptor activation—a global toggle switch model. Annu. Rev.Pharmacol. Toxicol. 46, 481–519 (2006).

28. Fredriksson, R., Lagerstrom, M. C., Lundin, L. G. & Schioth, H. B. The G-protein-coupled receptors in the human genome form five main families. Phylogeneticanalysis, paralogon groups, and fingerprints. Mol. Pharmacol. 63, 1256–1272(2003).

29. Rasmussen, S. G. et al. Structure of a nanobody-stabilized active state of the b2adrenoceptor. Nature 469, 175–180 (2011).

30. Rosenbaum, D. M. et al. Structure and function of an irreversible agonist-b2adrenoceptor complex. Nature 469, 236–240 (2011).

31. Rasmussen, S. G. et al. Crystal structure of the b2 adrenergic receptor–Gs proteincomplex. Nature 477, 549–555 (2011).

32. Harding, M. M. Metal-ligand geometry relevant to proteins and inproteins: sodiumand potassium. Acta Crystallogr. D 58, 872–874 (2002).

33. Horstman, D. A. et al. An aspartate conserved among G-protein receptors confersallosteric regulation of a2-adrenergic receptors by sodium. J. Biol. Chem. 265,21590–21595 (1990).

34. Costa, T., Lang, J., Gless, C. & Herz, A. Spontaneous association between opioidreceptors and GTP-binding regulatory proteins in native membranes: specificregulation by antagonists and sodium ions. Mol. Pharmacol. 37, 383–394 (1990).

35. Gao, Z. G. & Ijzerman, A. P. Allosteric modulation of A2A adenosine receptors byamiloride analogues and sodium ions. Biochem. Pharmacol. 60, 669–676 (2000).

36. Selent, J., Sanz, F., Pastor, M. & De Fabritiis, G. Induced effects of sodium ions ondopaminergic G-protein coupled receptors. PLOS Comput. Biol. 6, e1000884(2010).

37. Wilson, M. H., Highfield, H. A. & Limbird, L. E. The role of a conserved inter-transmembrane domain interface in regulating a2a-adrenergic receptorconformational stability and cell-surface turnover. Mol. Pharmacol. 59, 929–938(2001).

38. Neve, K. A. et al. Modeling and mutational analysis of a putative sodium-bindingpocket on the dopamine D2 receptor. Mol. Pharmacol. 60, 373–381 (2001).

39. Palczewski, K. et al. Crystal structure of rhodopsin: A G protein-coupled receptor.Science 289, 739–745 (2000).

40. Hanson, M. A. et al. Crystal structure of a lipid G protein-coupled receptor. Science335, 851–855 (2012).

41. Lerner, D. J., Chen, M., Tram, T. & Coughlin, S. R. Agonist recognition by proteinase-activated receptor 2 and thrombin receptor. Importance of extracellular loopinteractions for receptor function. J. Biol. Chem. 271, 13943–13947 (1996).

42. Nanevicz, T. et al. Mechanisms of thrombin receptor agonist specificity. Chimericreceptors and complementary mutations identify an agonist recognition site.J. Biol. Chem. 270, 21619–21625 (1995).

43. Blackhart, B. D. et al. Extracellular mutations of protease-activated receptor-1result in differential activation by thrombin and thrombin receptor agonistpeptide. Mol. Pharmacol. 58, 1178–1187 (2000).

44. Seeley, S. et al. Structural basis for thrombin activation of a protease-activatedreceptor: inhibition of intramolecular liganding. Chem. Biol. 10, 1033–1041(2003).

45. Bahou,W. F., Kutok, J. L.,Wong, A., Potter, C. L. & Coller, B. S. Identification of a novelthrombin receptor sequence required for activation-dependent responses. Blood84, 4195–4202 (1994).

46. Nanevicz, T., Wang, L., Chen, M., Ishii, M. & Coughlin, S. R. Thrombin receptoractivating mutations. Alteration of an extracellular agonist recognition domaincauses constitutive signaling. J. Biol. Chem. 271, 702–706 (1996).

47. Shaw, D. E.et al.Millisecond-scalemoleculardynamics simulationsonAnton.Proc.Conf. High Performance Computing Networking, Storage Analysis Article no. 39(2009).

Supplementary Information is available in the online version of the paper.

Acknowledgements We acknowledge support from the National Institutes of Health,grants NS028471 (B.K.K.), and HL44907 and HL65590 (S.R.C.), and from the MathersFoundation (B.K.K. and W.I.W.).

Author Contributions C.Z. optimized the construct, expressed and purified humanPAR1–T4L for crystallization, developed the purification procedure, performedcrystallization trials, optimizedcrystallizationconditions, collecteddiffraction data, andsolved and refined the structure. Y.S. helped design and make constructs forbaculoviral expression, and executed and analysed cell-based functional assays ofwild-type and mutant PAR1. D.H.A. designed, performed and analysed moleculardynamics simulations. J.J.F. developed the initial expression and purification protocolfor PAR1. D.P. helped design, execute and analyse platelet function studies. Y.Z. helpeddesign and make constructs for baculoviral PAR expression. H.F.G. performed andanalysed molecular dynamics simulations, and assisted with manuscript preparation.A.P. made vorapaxar. R.O.D. oversaw, designed and analysed molecular dynamicssimulations, and assisted with manuscript preparation. D.E.S. oversaw moleculardynamics simulations and analysis. W.I.W. assisted with X-ray diffraction dataprocessing and crystal structure refinement. S.R.C. and B.K.K. initiated the project,planned and analysed experiments, and supervised the research and wrote themanuscript with C.Z.

Author Information Atomic coordinates and structure factors for PAR1 are depositedin the Protein Data Bank under accession code 3VW7. Reprints and permissionsinformation is available at www.nature.com/reprints. The authors declare nocompeting financial interests. Readers are welcome to comment on the online versionof the paper. Correspondence and requests for materials should be addressed to S.R.C.(shaun.coughlin@ucsf.edu) or B.K.K. (kobilka@stanford.edu).

RESEARCH ARTICLE

6 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 2

Macmillan Publishers Limited. All rights reserved©2012

METHODSPAR1–T4L expression and purification. To facilitate crystallogenesis, a humanPAR1 construct was generated with several modifications. A TEV protease-recognition site was introduced after residue Pro 85, two N-linked glycosylationsites in ECL2 were removed by mutation (Asn250Gly and Asn259Ser), and the Cterminus was truncated after residue Ser 395. T4L residues 2–161 (ref. 48) wereinserted into the third intracellular loop between residues Ala 301 and Ala 303,with only one residue Val 302 removed. To facilitate purification, an N-terminalFlag epitope was inserted after a signal peptide and a C-terminal deca-histidinetag was introduced. The final crystallization construct PAR1–T4L is shown inSupplementary Fig. 2.

The modified PAR1 was expressed in Sf9 cells using the pFastBac baculovirussystem (Invitrogen). The ligand vorapaxar was added at 100 nM to the cellsduring expression. The cells were infected with baculovirus at 27 uC for 48 hbefore collection. To purify the receptor, infected cells were lysed by osmoticshock in low-salt buffer containing 10 mM Tris-HCl, pH 7.5, 1 mM EDTA,100 nM vorapaxar and 2 mg ml21 iodoacetamide. Iodoacetamide was used toalkylate reactive cysteines to prevent nonspecific oligomerization. The proteinwas further extracted from cell membranes using a glass dounce homogenizer inbuffer containing 20 mM Tris-HCl, pH 7.5, 350 mM NaCl, 1% dodecyl maltoside(DDM), 0.03% cholesterol hemisuccinate (CHS), 0.2% sodium cholate, 10% gly-cerol, 2 mg ml21 iodoacetamide and 100 nM vorapaxar. Cell debris was removedby high-speed centrifugation. From this point, 1mM vorapaxar was added in allthe following buffers used for purification. Nickel-NTA agarose resin was addedto the supernatant after homogenization and stirred for 1 h at 4 uC. The resin wasthen washed three times in batch with buffer comprised of 20 mM HEPES,pH 7.5, 350 mM NaCl, 0.1% DDM, 0.02% CHS and 1 mM vorapaxar, and trans-ferred to a glass column. The bound receptor was eluted with buffer containing300 mM imidazole and loaded onto an anti-Flag M1 affinity column. Afterextensive washing with buffer comprised of 20 mM HEPES, pH 7.5, 350 mMNaCl, 0.1% DDM, 0.02% CHS, 1mM vorapaxar and 2 mM Ca21, the receptorwas eluted from M1 resin using the same buffer without Ca21 but with 200mgml21 Flag peptide and 5 mM EDTA. To remove extra N-terminal residues andthe Flag epitope, TEV protease was added to the receptor and the cleavagereaction run at room temperature overnight. Size exclusion chromatographywas used to obtain the final monodisperse receptor preparation. PurifiedPAR1–T4L was concentrated to 40–50 mg ml21 using 100 kDa cut-off Vivaspinconcentrators for crystallization.Crystallization. As for other T4L-fused GPCRs crystallized so far, in meso crys-tallization was used to obtain PAR1–T4L crystals49,50. The protein was mixed withmonoolein and cholesterol (10:1 by mass) using the two syringe mixing methodby weight of 1:1.5 (protein:lipid). After a clear lipidic cubic phase formed, the mixwas dispensed onto glass plates in 20–40 nl drops overlaid with 700 nl precipitantsolution using a Gryphon LCP robot. Crystals appeared in two days in 0.1–0.2 Msodium chloride, 100 mM sodium phosphate, pH 6.0–6.5, 25–35% PEG300, andgrew to full size after 1 week (Supplementary Fig. 3).Data collection and structure determination. Crystals were collected and frozenin liquid nitrogen. Date collection was performed at beamline 23-ID of GM/CA@APS at the Advanced Photon Source. Microbeams of 10 or 20 mM diameterwere used to acquire all diffraction data. Owing to radiation damage, only 5–20degrees of rotation data were collected from each crystal. All data were processedwith the HKL2000 package51. A 2.2-A data set was obtained by merging diffrac-tion data from 18 crystals. The space group was determined to be P212121.Molecular replacement was performed using the program Phaser52 in Phenix53,with the CXCR4 structure (PDB accession 3ODU) as the search model. Theseven-transmembrane helices without any loops, and the T4L in the CXCR4structure, were used as independent search models. The initial structure modelwas completed and improved through iterative refinement in Phenix53 and man-ual rebuilding of all the loops and several parts in the transmembrane region inCoot54. Model refinement in Phenix and manual adjustment in Coot was per-formed to improve the model. The final structure was determined at 2.2-A reso-lution. The quality of the structure was assessed using Molprobity55. Dataprocessing and structure refinement statistics are shown in SupplementaryTable 1. Figures were prepared with PyMol56.Phosphoinositide hydrolysis assays and cell-surface expression level. TheQuikChange site-directed mutagenesis kit (Agilent) was used to generate humanPAR1 mutants and all mutants were fully sequenced. Cos7 cells were transientlytransfected using Fugene HD with either empty vector or wild-type human PAR1and mutants in the mammalian expression vector pBJ1 and signalling assays wereperformed as described in ref. 46. In brief, Cos7 cells expressing wild-type ormutant human PAR1 were labelled with [3H]-myoinositol, then incubated withvehicle or 100 nM vorapaxar in DMEM medium containing 0.1% BSA, 20 mMHEPES, 0.2% 2-hydroxypropyl-b-cyclodextrin (to retain vorapaxar in solution)

for 1 h at 37 uC. Agonist (100mM SFLLRN or 10 nM thrombin for PAR1 or otherPAR agonists as indicated) was added and incubation continued for 1 h. The totalamount of accumulated [3H]-inositol phosphates accumulated was determinedas in ref. 46. Cos7 cells transfected with empty vector had little response to PARagonists, and treatment with vorapaxar alone did not affect phosphoinositidehydrolysis (Supplementary Fig. 9A).

Surface expression of receptors was measured as described in refs 13 and 16.Cos7 cells were transiently transfected with empty pBJ1 or pBJ1 directing expres-sion of N-terminal Flag-tagged versions of wild-type human PAR1 or mutants.After 48 h, cells were washed once with serum-free medium containing 0.1% BSAand 20 mM HEPES, then incubated with 3 mg ml21 Flag M1 antibody (Sigma) for1 h at 4 uC in the same medium. The cells were then washed twice with PBScontaining Ca21 and Mg21 to remove unbound antibody and fixed with 2%paraformaldehyde for 5 min. The cells were then washed twice with PBS withCa21 and Mg21, and incubated with goat anti-mouse horseradish peroxidase(HRP)-conjugated secondary antibody, washed and developed with one-stepABTS HRP substrate (Pierce). The absorbance at 405 nm was measured as indica-tion of cell surface receptor expression levels.Platelet signalling assays. Washed human platelets were prepared and PAR1-dependent responses were measured as described in ref. 57. In brief, acid-citrate-dextrose anti-coagulated human blood samples (60 ml per donor) were obtainedfrom AllCells. Blood was centrifuged without braking at 250g at 37 uC for 15 min.The upper platelet-rich plasma phase was collected, incubated at 37 uC for 10 minin the presence of prostacyclin (PGI2, 0.5 mM), and centrifuged at 2,200g for15 min. The pellet was resuspended in complete Tyrode’s solution (134 mMNaCl, 12 mM NaHCO3, 2.9 mM KCl, 0.34 mM Na2HPO4, 1.0 mM MgCl2,10 mM HEPES, 0.9% (w/v) dextrose, pH 7.4) containing 2 mM CaCl2, 0.35%(w/v) bovine serum albumin (BSA), 10 U ml21 heparin, and 0.5mM PGI2. Theplatelet suspension was incubated for 10 min at 37 uC then centrifuged at 1,900gfor 8 min. This wash step was repeated and the final pellet resuspended inTyrode’s buffer supplemented with BSA and 0.02 U ml21 apyrase. Platelets wereincubated at 37 uC for 30 min to allow recovery from the effects of PGI2, countedusing a Hemavet FS950 (Drew Scientific) and diluted to 300,000 cells per micro-litre in Tyrode’s solution.

To antagonize PAR1, vorapaxar or vehicle (2% (w/v) 2-hydroxypropyl-b-cyclodextrin in dimethylsulphoxide (DMSO)) were added to platelet suspensionsthat were then incubated for 1 h at 37 uC before addition of agonists. The finalconcentrations of 2-hydroxypropyl-b-cyclodextrin and DMSO in platelet sus-pensions were 0.002% and 0.1%, respectively. Where reversibility was evaluated,platelets were washed twice with Tyrode’s buffer containing BSA and PGI2 aftervorapaxar treatment then diluted for cell activation assays as above.

For flow cytometric analysis of platelet activation, platelets suspended inTyrode’s solution containing 2 mM CaCl2, 0.35% BSA and 0.02 U ml21 apyrasewere incubated with agonist and antibody simultaneously. Fluor-conjugatedantibodies directed against P-selectin (phycoerythrin (PE)-conjugate of AK-4;Ebiosciences; 1:25 dilution) and the activated conformation of integrin aIIbb3

(FITC-conjugate of PAC-1, BD Biosciences; 1:25 dilution) were used. After15 min at 37 uC, the platelet suspension was diluted with PBS and platelet-boundantibody measured using an Accuri C6 flow cytometer (Accuri). Samples from atleast three different donors were analysed, each in triplicate.Molecular dynamics simulation methods. In all simulations, the receptor wasembedded in a hydrated lipid bilayer with all atoms, including those in the lipidsand water, represented explicitly. Simulations were performed on Anton47, aspecial-purpose computer designed to accelerate molecular dynamics simula-tions by orders of magnitude.System set-up and simulation protocol. Simulations of PAR1 were based on thecrystal structure of the PAR1–vorapaxar complex. The crystallized construct hasT4L inserted into ICL3 in place of residue 302. For the simulations, the T4Lportion was omitted, and residue 302 was modelled in. The unresolved segmentof ICL2 (residues 209–213) was also modelled in. Residues 209 and 213 wereadded manually, and residues 210–212 were modelled in using Prime(Schrodinger LLC). The Refine Loops tool in Prime, with default settings, wasthen used to refine residues 209–213.

The simulation of the MOR dimer was based on the crystal structure of MORbound to the irreversible antagonist b-funaltrexamine (PDB accession 4DKL).Both monomers of the crystallographic dimer were included in the simulation,but b-funaltrexamine was deleted from the binding pocket. As with PAR1, theT4L sequence was omitted in our simulations. Side chains for residue Met 651.29,Thr 671.31, Lys 260ICL3 and Arg 263ICL3 were not fully resolved in the crystalstructure, so they were modelled in by hand, with rotamers chosen to avoidany clashes with resolved residues.

For both PAR1 and MOR, hydrogens were added to the crystal structures usingMaestro (Schrodinger LLC), as described in previous work58. Histidines were

ARTICLE RESEARCH

Macmillan Publishers Limited. All rights reserved©2012

singly protonated on the epsilon nitrogen. All other titratable residues were left intheir dominant protonation state for pH 7.0, except for Asp 3677.49 in PAR1 andAsp 1142.50 in MOR, which were protonated, and Asp 1482.50 in PAR1, which wasprotonated in certain simulations. PAR1 was simulated both with and without thecrystallographic sodium ion by Asp 1482.50 (Supplementary Table 2); Asp 1482.50

was not protonated in simulations that included this ion, but was protonatedotherwise. The conserved aspartate at position 2.50 is known to be protonated inrhodopsin59, and the residue at position 7.49 is most often an (uncharged) aspara-gine residue in family A GPCRs (the ‘N’ of the NPXXY motif).

Prepared protein structures were inserted into an equilibrated POPC bilayer asdescribed previously32. Sodium and chloride ions were added to neutralize the netcharge of the system and to create a 150-mM solution.

Simulations of the PAR1 receptor initially measured 88.9 3 88.9 3 88.7 A andcontained 174 lipid molecules, and approximately 13,152 water molecules, for atotal of ,67,500 atoms. When the crystallographic sodium ion near Asp 1482.50

was included, the simulation contained 32 sodium ions and 36 chloride ions.When the crystallographic sodium ion was not included, the system contained 31sodium ions and 36 chloride ions. To simulate the unliganded PAR1 receptor,vorapaxar was deleted from the binding pocket. Simulations of the MOR dimerinitially measured 100.0 3 100.0 3 89.0 A and contained 204 lipid molecules, 19sodium ions, 43 chloride ions, and approximately 16,654 water molecules, for atotal of ,86,700 atoms.

All simulations were equilibrated using Anton in the NPT ensemble at 310 K(37 uC) and 105 Pa with 5 kcal mol21 A22 harmonic position restraints applied toall non-hydrogen atoms of the protein and the ligand; these restraints weretapered off linearly over 50 ns. All bond lengths to hydrogen atoms were con-strained using M-SHAKE60. A RESPA integrator61 was used with a time step of2 fs, and long-range electrostatics were computed every 6 fs. Production simula-tions were initiated from the final snapshot of the corresponding equilibrationruns, with velocities sampled from the Boltzmann distribution at 310 K, using thesame integration scheme, long-range electrostatics method, temperature andpressure. For PAR1, Van der Waals and short-range electrostatic interactionswere cut-off at 10.3 A and long-range electrostatic interactions were computedusing the k-space Gaussian split Ewald method62 with a 32 3 32 3 32 grid,s 5 2.27 A, and ss 5 1.59 A. For MOR, van der Waals and short-range electro-static interactions were cut-off at 10.16 A and long-range electrostatic interac-tions were computed using the k-space Gaussian split Ewald method with a64 3 64 3 64 grid, s 5 2.25 A, and ss 5 1.55 A.

We performed two vorapaxar-bound PAR1 simulations and four unligandedPAR1 simulations, and results were consistent across each set. The two receptorsin our MOR dimer simulation also exhibited consistent behaviour. The simu-lation protocol we followed has been validated in previous simulations ofGPCRs63,64. Nevertheless, it is possible that different behaviour might have beenobserved in even longer simulations, with different force field parameters, or witha different choice of simulation conditions.Force field parameters. We used the CHARMM27 parameter set for proteinmolecules and salt ions, with the CHARMM TIP3P water model65; protein para-meters incorporated CMAP terms66 and modified charges on the Asp, Glu andArg side chains67. We used a modified CHARMM lipid force field68. Force fieldparameters for vorapaxar were obtained from the CHARMM ParamChem webserver69, version 0.9.6 b.Analysis protocols. Trajectory snapshots, each containing a record of allatom positions at a particular instant in time, were saved every 180 ps during

production simulations. Time series data shown in Supplementary Fig. 11 weresmoothed by applying a 9.9-ns (55-snapshot) running average. VMD was used tovisualize trajectories70.

48. Rosenbaum, D. M. et al. GPCR engineering yields high-resolution structuralinsights into b2-adrenergic receptor function. Science 318, 1266–1273 (2007).

49. Caffrey, M. & Cherezov, V. Crystallizing membrane proteins using lipidicmesophases. Nature Protocols 4, 706–731 (2009).

50. Caffrey, M. Crystallizing membrane proteins for structure determination: use oflipidic mesophases. Annu. Rev. Biophys. 38, 29–51 (2009).

51. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected inoscillation mode. Methods Enzymol. 276, 307–326 (1997).

52. McCoy, A. J.et al.Phasercrystallographic software. J. Appl. Crystallogr.40, 658–674(2007).

53. Adams, P. D. et al. PHENIX: a comprehensive Python-based system formacromolecular structure solution. Acta Crystallogr. D 66, 213–221 (2010).

54. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. ActaCrystallogr. D. 60, 2126–2132 (2004).

55. Chen, V. B. et al. MolProbity: all-atom structure validation for macromolecularcrystallography. Acta Crystallogr. D 66, 12–21 (2010).

56. The. PyMOL Molecular Graphics System v.1.5.0.4 (Schrodinger, LLC, 2010).57. Cornelissen, I. et al. Roles and interactions among protease-activated receptors

and P2ry12 in hemostasis and thrombosis. Proc. Natl Acad. Sci. USA 107,18605–18610 (2010).

58. Dror, R. O. et al. Identification of two distinct inactive conformations of the b2-adrenergic receptor reconciles structural andbiochemical observations. Proc. NatlAcad. Sci. USA 106, 4689–4694 (2009).

59. Fahmy, K. et al. Protonation states of membrane-embedded carboxylic acidgroups in rhodopsin and metarhodopsin II: a Fourier-transform infraredspectroscopy study of site-directed mutants. Proc. Natl Acad. Sci. USA 90,10206–10210 (1993).

60. Krautler, V., Gunsteren, W. F. v. & Hunenberger, P. H. A fast SHAKE algorithm tosolve distance constraint equations for small molecules in molecular dynamicssimulations. J. Comput. Chem. 22, 501–508 (2001).

61. Tuckerman, M., Berne, B. J. & Martyna, G. J. Reversible multiple time scalemolecular dynamics. J. Chem. Phys. 97, 1990–2001 (1992).

62. Shan, Y., Klepeis, J. L., Eastwood, M. P., Dror, R. O. & Shaw, D. E. Gaussian splitEwald: A fast Ewald mesh method for molecular simulation. J. Chem. Phys. 122,54101 (2005).

63. Dror, R. O. et al. Activation mechanism of the b2-adrenergic receptor. Proc. NatlAcad. Sci. USA 108, 18684–18689 (2011).

64. Dror, R. O. et al. Pathway and mechanism of drug binding to G-protein-coupledreceptors. Proc. Natl Acad. Sci. USA 108, 13118–13123 (2011).

65. MacKerell, A. D. Jr, Bashford, D. & Bellott, M. All-atom empirical potential formolecular modeling and dynamics studies of proteins. J. Phys. Chem. B 102,3586–3616 (1998).

66. Mackerell, A. D. Jr, Feig, M. & Brooks, C. L. III. Extending the treatment of backboneenergetics in protein force fields: limitations of gas-phase quantum mechanics inreproducing protein conformational distributions in molecular dynamicssimulations. J. Comput. Chem. 25, 1400–1415 (2004).

67. Piana, S., Lindorff-Larsen, K. & Shaw, D. E. How robust are protein foldingsimulations with respect to force field parameterization? Biophys. J. 100, L47–L49(2011).

68. Klauda, J. B. et al. Update of the CHARMM all-atom additive force field for lipids:validation on six lipid types. J. Phys. Chem. B 114, 7830–7843 (2010).

69. Vanommeslaeghe, K. et al. CHARMM general force field: a force field for drug-likemolecules compatible with the CHARMM all-atom additive biological force fields.J. Comput. Chem. 31, 671–690 (2010).

70. Humphrey, W., Dalke, A. & Schulten, K. VMD: visual molecular dynamics. J. Mol.Graph. 14, 33–38 (1996).

RESEARCH ARTICLE

Macmillan Publishers Limited. All rights reserved©2012