STRUCTURAL BIOLOGY

Molecular basis for high-affinityagonist binding in GPCRsTony Warne1, Patricia C. Edwards1, Andrew S. Doré2,Andrew G. W. Leslie1, Christopher G. Tate1*

Gprotein–coupled receptors (GPCRs) in the G protein–coupled active state have higher affinityfor agonists as compared with when they are in the inactive state, but the molecular basisfor this is unclear.We have determined four active-state structures of the b1-adrenoceptor(b1AR) bound to conformation-specific nanobodies in the presence of agonists of varyingefficacy. Comparison with inactive-state structures of b1AR bound to the identical ligandsshowed a 24 to 42% reduction in the volume of the orthosteric binding site. Potentialhydrogen bonds were also shorter, and there was up to a 30% increase in the number ofatomic contacts between the receptor and ligand.This explains the increase in agonistaffinity of GPCRs in the active state for a wide range of structurally distinct agonists.

Gprotein–coupled receptors (GPCRs) existin an ensemble of conformations that canbe selectively stabilized by the binding ofa ligand and through interactions withsignaling molecules such as G proteins

(1, 2). Pharmacology has characterized at leasttwo distinct states of GPCRs, an active state withhigh affinity for agonists when coupled to G pro-teins and an inactive state with low affinity foragonists in the absence of G proteins (1), al-though a plethora of substates can also exist be-tween these two extremes (3–7). The reason whythe active state has a high affinity for agonistsis unclear because receptor structures in theinactive and active states have been deter-mined with different ligands bound, such asfor the b2-adrenoceptor (b2AR) (8–11). Here,we present structures of b1AR in the active stateand compare them with inactive-state struc-tures (12) bound to the identical ligand to definethe structural differences in the orthosteric bind-ing site.Four crystal structureswith overall resolutions

between 3.0 and 3.2 Å (table S1) were determinedof b1AR bound to either Nb80 or Nb6B9, and theoverall structures were virtually identical [0.2 to0.3 Å root mean square deviation (RMSD) for Caatoms] (Fig. 1). Nb80 and Nb6B9 are nanobodiesoriginally developed to stabilize the active stateof b2AR (8, 10) and bind to b1AR because ofthe high sequence conservation of the receptors.Structures were determined bound to a fullagonist (isoprenaline), partial agonists (salbuta-mol and dobutamine), and aweak partial agonist(cyanopindolol). Isoprenaline, salbutamol, anddobutamine showed an increase in affinity whenb1AR was coupled to the engineered G proteinmini-Gs, whereas cyanopindolol boundwith sim-ilar high affinity in both the presence and absenceof mini-Gs (Fig. 1). The overall structure of the

b1AR-nanobody complexes bound to either agonistor partial agonists is virtually identical to thatof the agonist-bound Nb6B9-b2AR complex (0.5 ÅRMSD of 1601 atoms), and the overall conforma-tional changes are virtually identical. These changesresult in the partial occlusion of the orthostericbinding pocket (Fig. 2), which is consistent withobservations on nanobody-bound b2AR (8, 10).Detailed comparisons were made between the

inactive-state structures of b1AR with the respec-tive active-state structures bound to the sameligand (Figs. 2 and 3). In all cases, there was adecrease in the volume (13) of the orthostericbinding site that varied depending on the ligand(Fig. 2 and fig. S1). The largest decrease was ob-served for the full agonist isoprenaline (42%),and the smallest decrease was observed for theweak partial agonist cyanopindolol (24%). Thedecrease in the volume of the orthosteric bindingsite when isoprenaline was bound was primarilydue to the inward movement of the extracellularends of H6 and H7, an inwardmovement and anincrease in the H5 bulge at Ser2155.46, and thereorientation of residuesPhe201ECL2 andPhe3257.35

[superscripts refer to the Ballesteros-Weinsteinnumbers (14)]. The magnitude of these changeswas greatest for the full agonist isoprenaline andsmallest for the weak partial agonist cyanopindolol.The pincerlike movement of Phe201ECL2 andPhe3257.35 toward the ligand has the largesteffect on reducing the volume of the orthos-teric binding pocket, with the maximal shift ob-served in the isoprenaline structure of 3.1 Å forPhe201ECL2 and 2.5 Å for Phe3257.35 (measuredat the CZ atom of the side chain). The movementof Phe201ECL2 appeared to correlate with thestructure of the ligand bound because in allcases it formed van der Waals contacts withthe ligand. By contrast, Phe3257.35 was not with-in van der Waals contact with any of the fourligands and moved as a consequence of the in-ward tilt of H7.The reduction in the volume of the orthosteric

binding pocket correlated with an overall reduc-tion in the average distance between atoms inthe ligand and receptor by 0.1 to 0.3 Å. Amino

acid residues in H3, H5, H6, H7, and ECL2 (andH2 for dobutamine) were all involved in contrib-uting to ligand-receptor contacts, but there wasno clear pattern as to which regions of the re-ceptor changedmost substantially (Fig. 3 and fig.S2). Side chains were up to 1.2 Å closer to theligand in the active state compared with the in-active state. In a number of instances, changesresulted in the strengthening of hydrogen bonds.For example, Asn3106.55 was predicted to makea weak hydrogen bond to the para-hydroxylgroup of isoprenaline (3.5 Å between donor andacceptor) in the inactive state, which changedto 2.8 Å in the active state. In the active-statestructures containing dobutamine and salbuta-mol, the distance to Ser2155.46 was 0.8 Å shorterfor both ligands, allowing hydrogen bond forma-tion; the hydrogen bond to Ser2115.42 also short-ened by 0.7 Å to the para hydroxyl in salbutamolbut remained unchanged to themeta hydroxyl indobutamine. Most of the observed differencesare due to the contraction of the binding pocket,whereas the substantial shortening of the hydro-gen bond between Ser2115.42 and salbutamol isdue to a rotamer change. Although all the ligandbinding pockets contracted upon receptor acti-vation, the changes in ligand-receptor contactswere not conserved, despite the similarity inchemotypes among the four ligands studied.There was a weak correlation between the de-crease in volume of the orthosteric bindingsite and ligand efficacy (fig. S3), particularly ifthe most similar chemtoypic ligands were com-pared (cyanopindolol, salbutamol, and isoprena-line). However, there was no correlation betweenefficacy and the magnitude of ligand affinity in-crease on receptor activation or the increase inthe number of ligand-receptor atomic contacts(fig. S3). Cyanopindolol bound to b1AR with sim-ilar affinity in both the presence and absence ofa coupled G protein (Fig. 1) despite the contrac-tion of the binding pocket and increase in re-ceptor ligand contacts upon activation. This maybe a consequence of constraints on the possibleconformation change imposed by the rigidityof cyanopindol that prevents the full contrac-tion of the ligand binding pocket by preventingthe movement of H7 and the bulge in H5 ob-served in the other structures (fig. S1).The role of the partial occlusion of the ortho-

steric binding site upon activation of b1AR wastested by mutagenesis inspired from the activestate structure of b2AR (8, 10). In b2AR, it wasproposed that the occlusion of the binding sitewas a major factor in increasing agonist affinityupon G protein coupling (15). In particular,Tyr3087.35 was within van der Waals distance ofPhe193ECL2 on the opposite side of the entranceto the orthosteric binding pocket and had amajor effect on decreasing the rates of associa-tion and dissociation of ligands in the active statecompared with the inactive state (15). The b2ARresidues Phe193ECL2/Tyr3087.35 are equivalentto Phe201ECL2/Phe3257.35 in b1AR, which are notin van der Waals contact (Fig. 4). Thus, the mu-tation F325Y7.35 in b1ARwas predicted to occludethe entrance to the orthosteric binding pocket

RESEARCH

Warne et al., Science 364, 775–778 (2019) 24 May 2019 1 of 4

1Medical Research Council (MRC) Laboratory of MolecularBiology, Francis Crick Avenue, Cambridge CB2 0QH, UK.2Sosei Heptares, Steinmetz Building, Granta Park, GreatAbington, Cambridge CB21 6GT, UK.*Corresponding author. Email: cgt@mrc-lmb.cam.ac.uk

on Septem

ber 25, 2020

http://science.sciencemag.org/

Dow

nloaded from

and decrease the rate of ligand association, andconversely, F325A7.35 was predicted to makethe entrance wider and increase the rate of lig-and association. When the initial rate of 3H-dihydroalprenolol (3H-DHA) association wasmeasured (Fig. 4), b1AR(F325A) had the samerate as b1AR, but b1AR(F325Y) had a considerablyslower rate of association. However, the affinities(Fig. 4) of epinephrine and isoprenaline forthe high-affinity state of b1AR and b1AR(F325Y)

were identical, and there was only a minor dif-ference with norepinephrine. The inactive stateof b1AR and the inactive state of b1AR(F325Y)were compared for their affinity to agonists (Fig.4), and it was found that b1AR had higher affinitythan b1AR(F325Y) for norepinephrine (8.1 nMversus 120nM), epinephrine (56nMversus 630nM),and isoprenaline (6.9 nM versus 52 nM). Thisimplied that the greater agonist affinity shift ob-served in b1AR(F325Y) compared with b1AR was

due to destabilization of the inactive state andnot stabilization of the active state. This suggestedthat partial occlusion of the ligand binding pocketin b1AR(F325Y) during formation of the activestate played little role in the increase of agonistaffinity on G protein coupling.The destabilizing effect of the F325Y mutation

in b1AR on the agonist-bound inactive state sug-gested that converting the extracellular surface ofb2AR to make it similar to b1AR would increase

Warne et al., Science 364, 775–778 (2019) 24 May 2019 2 of 4

Fig. 2. Conformational changes in isoprenaline-bound b1AR. (A) Superposition of isoprenaline-bound b1AR in the inactive state (gray, PDB ID 2Y03)with isoprenaline-bound b1AR in the active state(rainbow coloration). Arrows (magenta) indicatethe transitions from the inactive to active state.Alignment was performed based on the isoprenalinemolecules by using PyMol (magenta, isoprenalinebound to active state b1AR). (Single-letter abbrevia-tions for the amino acid residues are as follows: A,Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile;K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg;S, Ser; T,Thr; V,Val; W,Trp; and Y,Tyr. In the mutants,other amino acidswere substituted at certain locations;for example, F325Y indicates that phenylalanine atposition 325 was replaced by tyrosine.) (B) View ofthe orthosteric binding site from the extracellularsurface, with atoms shown as space filling models.For isoprenaline, magenta, carbon atoms. For b1AR,H1, dark blue; H2, light blue; H5, yellow; ECL2, green;ECL3 and parts of H6 and H7, red. (C) Volumes of theorthosteric binding site in the low-affinity inactivestate (L, pink bars) compared with the high-affinityactive state (H, green bars). (D) Number of atomic contacts (data file S1) between the respective ligands and b1AR in the low-affinity inactive state (L, pink bars)compared with the high-affinity active state (H, green bars).The dark shades represent the number of polar interactions. Ligand abbreviations are shown in Fig. 1.

Fig. 1. Structure of the active state ofagonist-bound b1AR-nanobody complex.(A) Superposition of four structuresof b1AR-nanobody complexes boundto ligands shown in (C). (B) Affinitiesof b1AR in the low-affinity state (L)and high-affinity state coupled tomini-Gs (H) for the ligands co-crystallizedwith the receptor. Data are in tablesS2 and S3 and fig. S6. Results arethe mean of two to four experimentsperformed in duplicate, with errorbars representing the SEM. (C) Structuresof the ligands cocrystallized in the b1ARcomplexes. (D) Disposition of theligands after superposition of thereceptors, using the same colorcoding as in (B). Ligand densitiesare provided in fig. S9.

RESEARCH | REPORTon S

eptember 25, 2020

http://science.sciencem

ag.org/D

ownloaded from

the affinity of the inactive state and leave theaffinity of the G protein–coupled activated stateapproximately unchanged. The b2ARmutant con-structed, b2AR(b1LBP), did indeed show thesecharacteristics (fig S4; the rationale of the fourmutations used—Y174WECL2, H296N6.58, K305D7.32,and Y308F7.35—is provided in the supplementarymaterials, materials and methods). In addition,the accessibility of the b1AR orthosteric bindingpocket in the G protein–coupled state to 125I-cyanopindolol was greater than that observedfor b2AR (fig. S5). The four mutations in b2AR(b1LBP) converted the behavior of b2AR to thatof b1AR. Adding the converse residues from b2ARinto b1AR, to make the mutant b1AR(b2LBP),converted the accessibility of the orthosteric bind-ing site in b1AR to that of b2AR (fig. S5).The increase in affinity of agonist binding to

the active state of b1AR arises from the increasein the number and/or strength of ligand-receptorcontacts (a thermodynamic effect). This is conse-quently associated with a decrease in the rateof ligand dissociation. Where a native ligand isa large peptide that interacts with extracellulardomains, this thermodynamic effect may be themajor contributor to the decrease in the rate ofligand dissociation. However, where the ligandis small, such as adrenaline or acetylcholine, therate of ligand dissociation may also be decreasedbecause of a purely steric blockage of the en-trance to the ligand binding pocket (a kinetic ef-fect) (14). Given the high conservation of both thestructure and function of GPCRs (16, 17), theseconsiderations are likely to apply to other GPCRsthat bind diffusible ligands.

REFERENCES AND NOTES

1. U. Gether, Endocr. Rev. 21, 90–113 (2000).

2. D. M. Rosenbaum, S. G. Rasmussen, B. K. Kobilka, Nature 459,356–363 (2009).

3. G. Vauquelin, I. Van Liefde, Fundam. Clin. Pharmacol. 19, 45–56(2005).

4. G. G. Gregorio et al., Nature 547, 68–73 (2017).

5. X. J. Yao et al., Proc. Natl. Acad. Sci. U.S.A. 106, 9501–9506(2009).

6. A. Manglik et al., Cell 161, 1101–1111 (2015).

7. L. Ye, N. Van Eps, M. Zimmer, O. P. Ernst, R. S. Prosser, Nature533, 265–268 (2016).

8. S. G. Rasmussen et al., Nature 469, 175–180 (2011).

9. S. G. Rasmussen et al., Nature 477, 549–555 (2011).

10. A. M. Ring et al., Nature 502, 575–579 (2013).

11. D. M. Rosenbaum et al., Nature 469, 236–240 (2011).

12. T. Warne et al., Nature 469, 241–244 (2011).

13. J. D. Durrant, L. Votapka, J. Sørensen, R. E. Amaro, J. Chem.Theory Comput. 10, 5047–5056 (2014).

14. J. A. Ballesteros, H. Weinstein, Methods Neurosci. 25, 366(1995).

15. B. T. DeVree et al., Nature 535, 182–186 (2016).16. A. J. Venkatakrishnan et al., Nature 536, 484–487 (2016).

ACKNOWLEDGMENTS

We thank the beamline staff at the European SynchrotronRadiation Facility (beamlines ID23-2, ID30-A3, ID29, ID30B,and MASSIF-1) and at Diamond Light Source (beamline I24).Funding: This work was supported by core funding fromthe Medical Research Council (MRC U105197215 andU105184325) and a grant from the European ResearchCouncil (EMPSI 339995). Author contributions: T.W. performedreceptor and nanobody expression, purification, crystallization,cryo-cooling of the crystals, data collection, data processing, andstructure refinement. T.W. also performed the pharmacologicalanalyses. P.C.E. purified mini-Gs, and A.S.D. assisted with

Warne et al., Science 364, 775–778 (2019) 24 May 2019 3 of 4

A

C D

B

Fig. 4. Comparisons between b1AR and b2AR. (A) Alignment of the active-state structures of b1AR (rainbowcoloration) and b2AR (gray, PDB ID 4LDO). Ligands are shown as sticks. Isoprenaline, yellow; adrenaline,grey. (B) Rate of association of the radioligand 3H-DHA on to b1AR (blue circles), b1AR(F325Y) (red triangles),and b1AR(F325A) (orange circles). (Inset) The affinities of 3H-DHA (same color code). (C and D) Affinitiesof b1AR, b2AR, and their respective mutants in the low-affinity state (L) and high-affinity state coupled tomini-Gs (H). All data are in tables S2 and S3, and representative graphs of affinity shifts are in fig. S6. Resultsare the mean of two to six experiments performed in duplicate, with error bars representing the SEM.

Fig. 3. Changes in b1AR-ligand contact distances. The maximal changes in contact distancesbetween ligands and atoms in b1AR from the inactive to active states are depicted. Amino acid sidechains making contact with the ligands are indicated and colored according to where they are inb1AR (blue, H2; red, H3; orange, ECL2; gray, H5; green, H6; purple, H7), with the diameter of thecircle representing the magnitude of the distance change (shown as numbers below the amino acidresidue). Numbers next to the lines indicate the change in length of polar contacts (blue dashedlines) and hydrogen bonds (red dashed lines; determined by using HBPLUS). Negative numbersimply a decrease in distance between the ligand and receptor in the transition from the inactive stateto the active state. An asterisk indicates a rotamer change between the inactive and active states.The details of additional contacts made by each side chain are provided in fig. S2.

RESEARCH | REPORTon S

eptember 25, 2020

http://science.sciencem

ag.org/D

ownloaded from

structure solution. A.G.W.L. was involved in data processing andstructure solution, refinement, and analysis. Manuscriptpreparation was performed by T.W., A.G.W.L., and C.G.T. Theoverall project management was by C.G.T. Competing interests:C.G.T. is a shareholder, consultant, and member of the ScientificAdvisory Board of Heptares Therapeutics, who also partlyfunded this work. Data and materials availability: Thecoordinates and structure factors for all the structures determined

have been deposited at the Protein Data Bank (PDB) with thefollowing accession codes (ligand cocrystallized in parentheses):6H7J (isoprenaline), 6H7L (dobutamine), 6H7M (salbutamol), and6H7O (cyanopindolol).

SUPPLEMENTARY MATERIALS

science.sciencemag.org/content/364/6442/775/suppl/DC1Materials and Methods

Figs. S1 to S9Tables S1 to S4References (17–29)Data Files S1 and S2

22 June 2018; accepted 29 April 2019Published online 9 May 201910.1126/science.aau5595

Warne et al., Science 364, 775–778 (2019) 24 May 2019 4 of 4

RESEARCH | REPORTon S

eptember 25, 2020

http://science.sciencem

ag.org/D

ownloaded from

Molecular basis for high-affinity agonist binding in GPCRsTony Warne, Patricia C. Edwards, Andrew S. Doré, Andrew G. W. Leslie and Christopher G. Tate

originally published online May 9, 2019DOI: 10.1126/science.aau5595 (6442), 775-778.364Science 

, this issue p. 775Scienceinactive state elucidates how agonist binding is altered in the active conformation.receptor bound to a full agonist, two partial agonists, and a weak partial agonist. Comparison of these structures to the

1-adrenergicβ(recombinant variable domains of heavy-chain antibodies) and engineered G protein to stabilize the in its active state. They used nanobodies−−1-adrenergic receptorβthe −−four crystal structures of complexes of a GPCR

describe et al.coupled receptors (GPCRs) are exceptionally good targets for drug development. Warne −G proteinA GPCR seen in the active state

ARTICLE TOOLS http://science.sciencemag.org/content/364/6442/775

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2019/05/08/science.aau5595.DC1

REFERENCES

http://science.sciencemag.org/content/364/6442/775#BIBLThis article cites 29 articles, 2 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Science. No claim to original U.S. Government WorksCopyright © 2019 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of

on Septem

ber 25, 2020

http://science.sciencemag.org/

Dow

nloaded from