Angiotensin and biased analogs induce structurally distinct active conformations within a GPCR

See allHide authors and affiliations

Science  21 Feb 2020:
Vol. 367, Issue 6480, pp. 888-892
DOI: 10.1126/science.aay9813

Choosing the drug to fit the protein

Many approved drugs bind to G protein–coupled receptors (GPCRs). A challenge in targeting GPCRs is that different ligands preferentially activate different signaling pathways. Two papers show how biased signaling arises for the angiotensin II type 1 receptor that couples to two signaling partners (G proteins and arrestins). Suomivuori et al. used large-scale atomistic simulations to show that coupling to the two pathways is through two distinct GPCR conformations and that extracellular ligands favor one or the other conformation. Wingler et al. present crystal structures of the same receptor bound to ligands with different bias profiles. These structures show conformational changes in and around the binding pocket that match those observed in simulations. This work could provide a framework for the rational design of drugs that are more effective and have fewer side effects.

Science, this issue p. 881, p. 888


Biased agonists of G protein–coupled receptors (GPCRs) preferentially activate a subset of downstream signaling pathways. In this work, we present crystal structures of angiotensin II type 1 receptor (AT1R) (2.7 to 2.9 angstroms) bound to three ligands with divergent bias profiles: the balanced endogenous agonist angiotensin II (AngII) and two strongly β-arrestin–biased analogs. Compared with other ligands, AngII promotes more-substantial rearrangements not only at the bottom of the ligand-binding pocket but also in a key polar network in the receptor core, which forms a sodium-binding site in most GPCRs. Divergences from the family consensus in this region, which appears to act as a biased signaling switch, may predispose the AT1R and certain other GPCRs (such as chemokine receptors) to adopt conformations that are capable of activating β-arrestin but not heterotrimeric Gq protein signaling.

Agonist binding to an extracellular-facing pocket of G protein–coupled receptors (GPCRs) initiates conformational changes that are propagated to the intracellular regions of the receptor. GPCR activation by agonists typically not only activates heterotrimeric G proteins but also promotes receptor phosphorylation by G protein–coupled receptor kinases (GRKs) and subsequent binding of β-arrestins (1). In addition to promoting GPCR desensitization and endocytosis, β-arrestins initiate additional signaling cascades (2). Although most agonists activate both G protein and β-arrestin pathways, it has been well established that many GPCR ligands show pronounced bias and can preferentially—or, in the limiting case, exclusively—activate particular downstream pathways (3).

The angiotensin II type 1 receptor (AT1R) is a particularly compelling model system for investigating biased agonism because small structural modifications to the angiotensin II (AngII) octapeptide agonist lead to strongly biased signaling. In particular, alteration of the C-terminal F8 of AngII profoundly attenuates Gq-mediated signaling but not β-arrestin coupling (4, 5). For example, TRV026, which lacks an eighth residue, and TRV023, which has an F8→A (F8A) substitution, are deficient in Gq-dependent inositol phosphate generation; however, both of these β-arrestin–biased ligands robustly promote β-arrestin–dependent endocytosis (Fig. 1, A to C) (6, 7). These β-arrestin–biased ligands are of interest for the treatment of heart failure, as they have the same antihypertensive effects as clinically used AT1R antagonists (angiotensin receptor blockers) but also improve cardiac function through β-arrestin–mediated pathways (8, 9).

Fig. 1 Endogenous and biased AT1R ligands.

(A) Peptide ligands crystallized with the AT1R. Bold residues indicate mutations relative to AngII, the endogenous agonist. NMe, N-methyl. (B) Ligand activation of Gq-mediated inositol monophosphate increases and β-arrestin2 endocytosis. M, Molar. (C) Bias factors of ligands relative to AngII, determined from the data shown in panel (B) as described in the materials and methods. A bias factor of 1 represents a 10-fold difference in the ligand’s ability to activate the β-arrestin pathway compared with the Gq pathway. (D) Nanobody AT110i1 allosterically increases the binding of AngII, TRV026, and TRV023 to purified AT1R [Ki (inhibition constant) values are in table S1]. (E) AngII, TRV026, and TRV023 promote interaction of FLAG-AT1R and AT110S-His6, the lower-affinity, parental clone of AT110i1, by AlphaScreen. Data are normalized to signal without ligand. In (B), (D), and (E), the means ± SE from three independent experiments are shown. In (C), error bars represent the SE in bias factors derived from curve fit parameters from (B).

To date, the molecular mechanisms of biased agonist action remain unclear. Although several structures of GPCRs bound to ligands exhibiting varying degrees of bias have been reported (10, 11), it has been challenging to trace how the conformational changes induced by biased agonists are transmitted from the extracellular ligand-binding site to the intracellular transducer-binding pocket. This question is best addressed through the use of several complementary techniques. Using double electron-electron resonance spectroscopy, we have previously demonstrated that AngII, TRV023, and TRV026 each promote distinct sets of intracellular conformations of the AT1R (12). Because this approach can only report on conformational changes in the intracellular regions that are accessible to chemical labeling, we turned to crystallography to delineate, at high resolution, the structural rearrangements induced by AngII and biased ligands elsewhere in the receptor. Molecular dynamics simulations, presented in a companion manuscript (13), capture the same structural rearrangements and relate them to distinct intracellular conformations that affect transducer binding.

We recently reported an active-state crystal structure of the AT1R bound to the high-affinity partial agonist S1I8 (Fig. 1A) in complex with an active state–stabilizing synthetic nanobody, AT110i1 (14). Compared with the strongly β-arrestin–biased ligands TRV026 and TRV023, S1I8 activates Gq signaling more potently and, thus, shows less bias toward β-arrestin pathways (Fig. 1, B and C). AT110i1 shows a positive allosteric interaction with AngII, TRV026, and TRV023, increasing their affinity for the AT1R, as assessed by radioligand binding (Fig. 1D and table S1), and stabilizing the receptor-ligand complex, as measured by thermal shift assays (fig. S1 and table S2). Likewise, AngII and the β-arrestin–biased ligands promote interaction of the AT1R with a lower-affinity variant of AT110i1 (Fig. 1E), which suggests that AT110i1 would be a suitable crystallographic chaperone for AT1R bound to all of these ligands. Consistent with these data, when the nanobody is removed from the AT1R-S1I8-AT110i1 structure in molecular dynamics simulations, the intracellular side of the AT1R relaxes to a conformation that is highly similar to that observed crystallographically, even when the ligand is replaced with AngII, TRV026, or TRV023 (13).

For crystallography efforts, we employed a previously described AT1R construct (14). The construct contains an insertion of thermostabilized apocytochrome b562RIL (BRIL) in the third intracellular loop and a C-terminal truncation, which is compatible with AT110i1 complex formation in the presence of AngII and β-arrestin–biased ligands (table S1). Diffraction-quality crystals of AT1R-AT110i1 with AngII (2.9 Å), TRV026 (2.7 Å), and TRV023 (2.8 Å) (table S3) were obtained in several distinct crystal forms by lipidic mesophase crystallization (15) under similar conditions to those used for AT1R-AT110i1-S1I8 (fig. S2). Globally, all three AT1R structures are similar to each other and to the S1I8 structure in their backbone conformations [0.39 Å root mean square deviation (RMSD) to AT1R-AngII for 271 Cα atoms in the seven transmembrane (TM) helices]. In the extracellular regions, we observe a marked contraction of the ligand-binding pocket relative to small molecule antagonist-bound AT1R structures (fig. S3) (14, 16, 17). AT110i1 binding to the transducer pocket stabilizes an intracellular conformation with features characteristic of GPCR activation, such as the outward displacement of TM6 (Fig. 2A). Extracellular and intracellular conformational changes are linked by the rearrangement of conformational locks in the receptor (Fig. 2A). Notably, as reported in our companion manuscript (13), the crystallographically observed conformation is almost identical to one of the two major conformations to which the AT1R relaxes when the nanobody is removed from the AT1R-AT110i1-S1I8 structure in molecular dynamics simulations. The conformation seen in the simulation differs only in the rotameric state of Y3027.53 [superscripts indicate Ballesteros-Weinstein numbering for conserved GPCR residues (18)] at the intracellular end of TM7 and, notably, appears to be able to accommodate β-arrestin but not Gq binding in models of AT1R-transducer complexes (13). This suggests that although AT110i1 constrains the intracellular conformation of the receptor in our structure, the crystal structures reflect a conformation sampled by the agonist-bound receptor in the absence of the nanobody.

Fig. 2 Comparison of AngII- and β-arrestin–biased ligand-bound AT1R.

(A) Activated AT1R (orange) bound to AngII (yellow) displays characteristic TM6 and TM7 movements stabilized by TM6 conformational locks (shown in stick representation). AT1R bound to TRV023 (0.43 Å RMSD for 264 Cα atoms) and TRV026 (0.42 Å RMSD for 264 Cα atoms) is nearly identical in overall conformation. ECL, extracellular loop. (B to D) Similar binding modes of TRV023 (B), TRV026 (C), and AngII (D). Dashed lines indicate hydrogen bonds. (E) AngII-AT1R ligand-binding pocket colored by Cα B-factors, highlighting the mobility of F8. (F to H) Electron density (gray mesh shows 2Fo-Fc electron density map contoured at 1σ) of C-terminal regions of TRV023 (F), TRV026 (G), and AngII (H) and the surrounding AT1R residues. Weak density for AngII F8 and Y292 (H) indicates that the residues are structurally heterogeneous within the crystal lattice, likely because they are dynamic (13). Single-letter abbreviations 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.

All of the peptide ligands assume a similar binding pose, with their N termini positioned at the extracellular face and C termini positioned at the base of the ligand-binding pocket (Fig. 2, B to D). Although residues 1 and 5 vary among the ligands, these substitutions do not substantially alter interactions with the receptor, which is consistent with data showing that these changes are not required to convert AngII into a biased ligand (7, 19). In contrast to the other structures of GPCRs bound to agonists exhibiting varying degrees of bias that have been reported to date—which tend to differ in how ligands engage the extracellular receptor face (10, 11, 20)—pronounced differences between the AngII-bound and β-arrestin–biased ligand-bound structures occur only at the base of the ligand-binding pocket, around the C-terminal residues of the ligands (Fig. 2, E to H). Clear and continuous density is observed for each residue of TRV026 and TRV023 and the surrounding AT1R residues (Fig. 2, F and G, and fig. S4). However, AngII F8 is poorly resolved and displays markedly high B-factors for its side-chain and main-chain atoms—despite residing in the core of the receptor (Fig. 2, E and H, and fig. S4)—suggesting that it is conformationally heterogeneous. In support of this observation, simulations of the AngII-bound receptor show that the F8 side chain visits multiple distinct conformations that are consistent with the observed density (13). Weak electron density is also observed for the nearby L1123.36 side chain, and no density is observed for the side chain of the nearby Y2927.43 (Fig. 2H). In the AngII-bound structure, TM3 rotates on-axis, and L1123.36 moves past W2536.48 and into the position previously occupied by Y2927.43 (Fig. 3, A and B). Despite containing a hydrophobic I residue at position 8, the partial agonist S1I8 does not induce rotation of TM3 (fig. S5).

Fig. 3 AngII-AT1R exhibits distinct configurations of conformational locks.

(A) View of AngII-AT1R, with key residues highlighted. (B) Movement of L1123.36 accommodates the position of AngII F8. The concomitant rotation of TM3 repositions N1113.35 outside the receptor core. TRV026 binding does not have this effect. (C and D) Change in affinity of AT1R ligands for AT1R L112A3.36 (C) and Y292A7.43 (D) versus wild-type AT1R. Losartan is a small-molecule antagonist; TRV055 is a Gq-biased agonist with a C-terminal phenylalanine (7). Error bars represent the SE of the difference in the log Ki values determined from three to four independent experiments. See also fig. S6 and table S4. For (C) and (D), the asterisk symbol indicates statistically significant difference (*P < 0.05) for mutant versus wild-type Ki values, as determined by a t test with Holm-Sidak correction for multiple comparisons. (E) Both AngII-bound and TRV026-bound structures exhibit the outward rotation of TM6 and inward rotation of TM7, rearranging the conformational locks. (F) AngII-induced rotation of N1113.35 alters the polar network at the canonical sodium-binding site in the receptor core, involving D742.50, N2957.46, and N2987.49 (of the NPXXY motif).

Consistent with these crystallographic observations, an L112A3.36 substitution, which is expected to better accommodate the presumed flexibility of AngII F8, increases AT1R affinity for AngII but decreases affinity for TRV026, TRV023, and S1I8 (Fig. 3C, fig. S6, and table S4). A Y292A7.43 substitution designed to mimic the disordered side chain induced by AngII increases AT1R affinity for Gq-active ligands with a C-terminal F8, but it has relatively minor effects on TRV026, TRV023, and S1I8 affinity (Fig. 3D, fig. S6, and table S4). Despite having enhanced affinities for AngII, L112A3.36 and Y292A7.43 do not show increased efficacy or potency in activating AngII-dependent Gq signaling (fig. S7 and table S5). This supports the hypothesis that L112A3.36 and Y292A7.43 play a key role in coupling AngII binding to Gq activation. The L112A3.36 variant reduces affinity but increases Gq efficacy for the partial agonist S1I8 (Fig. 3C, figs. S6 and S7, and tables S4 and S5). This suggests that the alanine substitution is sufficient to permit rotation of TM3 in the presence of I8 by reducing the hydrophobic interaction between L1123.36 and W2536.48.

In addition to showing differences in the ligand-binding pocket compared with structures with β-arrestin–biased ligands, AngII-AT1R displays more-substantial changes in conformational locks located deeper in the receptor core (Fig. 3, A, E, and F). In all of the active-state AT1R structures, the outward rotation of TM6 and associated inward rotation of TM7 (Fig. 2A and Fig. 3E) disrupt an inactive state–stabilizing interaction between N1113.35 and N2957.46 underneath the ligand-binding pocket (Fig. 3B and fig. S5). Rotation of N2957.46 rearranges a central hydrogen bonding network of polar residues in the receptor core that also involves D742.50 and N2987.49, the first residue of the NPXXY motif (where X denotes any residue), which is involved in stabilizing the active state of many GPCRs (Fig. 3F and fig. S5). In the S1I8 and β-arrestin–biased ligand structures, N1113.35 remains in proximity to the reoriented N2957.46 (Fig. 3, B and F, and fig. S5) and may restrict the orientations of TMs 3 and 7. However, the rotation of TM3 in the AngII-bound structure, which is associated with L1123.36-repositioning in the ligand-binding pocket, is also accompanied by the side chain of N1113.35 flipping outside the TM bundle (Fig. 3, B and F). This results in a configuration of the polar network in the core of the AngII-bound structure that is distinct from the configurations observed in both antagonist-bound and β-arrestin–biased ligand-bound AT1R (Fig. 3, B and F). Notably, coupled movements of L1123.36 and N1113.35 also occur in molecular dynamics simulations of AT1R—in the absence of the stabilizing nanobody—and are frequently accompanied by changes in intracellular conformation (13).

Our structures suggest that the binding of partial or β-arrestin–biased peptide agonists triggers a shift of TM6, a hallmark of GPCR activation, and a conformational change in N2957.46, whereas the binding of AngII induces the additional outward movement of N1113.35. Based on the functional profiles of these ligands, the rearrangement of N2957.46 is likely sufficient to promote the conformational changes that are needed for β-arrestin coupling, whereas the flipping of N1113.35 is essential for Gq signaling. Consistent with this model, the balance of Gq and β-arrestin activation at the AT1R is acutely sensitive to mutagenesis of the polar network beneath the ligand-binding pocket. N111G3.35, releasing the constraints on TM3, exhibits high levels of ligand-independent activity, particularly with respect to Gq signaling (2123). Furthermore, the partial agonist S1I8 and the β-arrestin–biased agonist SII (sarcosine1, I4, I8-AngII) behave as full agonists toward Gq signaling at AT1R N111G3.35 (23). An important role for D742.50 is supported by the fact that AT1R D74N2.50 shows impaired Gq activity but still robustly activates β-arrestin pathways (24, 25).

In most family A GPCRs, the corresponding conserved polar residues form a sodium-binding site, which allosterically stabilizes TMs 3 and 7 in the inactive conformation and collapses upon receptor activation (26) (Fig. 4, A and B). However, sodium has almost no effect on AngII affinity for the AT1R (Fig. 4C) (16), and there is no electron density for sodium in the inactive AT1R structures (16, 17) (Fig. 4D). Substitution of the highly conserved S7.46 with N2957.46 in the AT1R (Fig. 4, A and D) disrupts the sodium coordination sphere and precludes sodium binding. Instead, a hydrogen bond between N2957.46 and N1113.35 replaces the need for sodium to stabilize the inactive state (Fig. 4D). This substitution appears to have helped create a bipartite activation mechanism that renders the AT1R particularly prone to biased signaling (Fig. 4, D and E). Deviations from the conserved sodium-binding motif are particularly enriched in peptide- and protein-binding family A GPCRs that are closely related to the AT1R (Fig. 4A). For example, only 3 out of 22 chemokine receptors possess all of the conserved residues typically required for tight sodium binding (27). Notably, the chemokine receptor family also contains a number of atypical receptors that do not signal through G proteins, some of which have been shown to activate β-arrestins (28), and endogenous chemokines function as biased ligands at many chemokine receptors (29, 30). Furthermore, substitutions in and immediately surrounding the sodium-binding site of many GPCRs [e.g., δ-opioid receptor (31), CCR5 (32), and NK1 receptor (33)] can bias receptor signaling either toward G proteins or β-arrestins, which demonstrates that transducer coupling is exquisitely sensitive to manipulation of this region. A pathogenic mutation associated with cancer is located in the canonical sodium-binding region of CysLTR2 and yields a constitutively active G protein–biased receptor (34). We provide the first direct, experimental evidence that this critical polar network assumes at least three configurations—the inactive configuration and two distinct active conformations—which provides a molecular explanation for how it contributes to differential transducer activation.

Fig. 4 Structural diversity in the polar core of GPCRs.

(A) Conservation of sodium coordination residues (26), with family A GPCR consensus residues shown in red. Many chemokine receptors (36) deviate from the consensus sequence. (B) Sodium binding in the δ-opioid receptor. N3.35 and S7.46 are found in the sodium coordination sphere. (C) Effect of sodium on AngII binding to wild-type AT1R in membranes. Competition radioligand binding was performed in buffer containing 150 mM NaCl (log KiAngII = −7.54 ± 0.06, Ki = 29 nM; Kd[3H]-olmesartan = 1.2 ± 0.1 nM) or lacking sodium and containing 150 mM KCl (log KiAngII = −7.73 ± 0.02, Ki = 19 nM; Kd[3H]-olmesartan = 1.2 ± 0.3 nM). The means ± SE from three independent experiments are shown. (D and E) Residues associated with sodium binding make up the AT1R polar network. (D) The N1113.35-N2957.46 hydrogen bond stabilizes the inactive state in antagonist (ZD7155)–bound AT1R. Upon TRV026 binding, TM7 movement disrupts the hydrogen bond. (E) AngII binding rotates N1113.35 away from the polar core, yielding a second active conformation.

Our structures of the AT1R define a mechanism for biased agonist action that undoubtedly has parallels in other GPCRs. AngII—but not β-arrestin–biased ligands—disrupts the receptor core, yielding a conformation that is structurally distinct from that induced by β-arrestin–biased ligands. Although allosteric effects of the intracellular nanobody required for crystallogenesis could mask additional ligand-specific changes, even beyond those observed in this study, the notable conformational differences that persist beyond the ligand-binding site attest to the loose allosteric coupling known to exist in GPCRs (35). Double electron-electron resonance spectroscopy experiments and molecular dynamics simulations confirm that the conformational differences initiated by AngII and β-arrestin–biased agonist binding are propagated all the way to the intracellular surface when the AT1R is structurally unconstrained (12, 13). Consistent with our crystallographic observations of the ligand-binding site and the receptor core, these complementary methods demonstrate that AngII promotes a more fully activated receptor conformation than that promoted by β-arrestin–biased ligands. Given the similarity of conformations of GPCRs once they have fully engaged G proteins and β-arrestin, these mounting data suggest that the initial recognition of the AT1R by GRKs and β-arrestins could involve the conformations that are stabilized preferentially by β-arrestin–biased ligands. Our work provides a structural explanation for the complex signaling pharmacology of this family of drug targets.

Supplementary Materials

Materials and Methods

Figs. S1 to S7

Tables S1 to S5

References (3748)

References and Notes

Acknowledgments: We thank V. Brennand, Q. Lennon, and J. Taylor for administrative assistance. We thank the staff at Advanced Photon Source GM/CA beamlines for technical assistance and support of data collection. GM/CA@APS is supported by the NIH National Institute of General Medical Sciences (AGM-12006) and the National Cancer Institute (ACB-12002). Funding: Funding was provided by the Sigrid Jusélius Foundation (C.-M.S.); the International Human Frontier Science Program Organization (LT000916/2018-L) (C.-M.S.); the Mandel Center for Hypertension and Atherosclerosis at Duke (R.J.L.); the Vallee Foundation (A.C.K.); the Smith Family Foundation (A.C.K.); and NIH grants R01GM127359 (R.O.D.), R01HL16037 (R.J.L.), and DP5OD021345 (A.C.K.). A.L.W.K. is a Howard Hughes Medical Institute Medical Research Fellow. R.J.L. is an investigator with the Howard Hughes Medical Institute. Author contributions: L.M.W., D.P.S., and A.L.W.K. performed signaling and functional characterization experiments. L.M.W. crystallized the complexes. L.M.W., C.M., and A.C.K. collected x-ray diffraction data. M.A.S., C.M., and A.C.K. performed x-ray data processing and refinement. C.-M.S., N.R.L., and R.O.D. proposed experiments on the basis of structural analysis. All authors interpreted data. R.J.L. and A.C.K. supervised the project. All authors wrote the manuscript. Competing interests: R.J.L. is a founder and stockholder of Trevena and is a director of Lexicon Pharmaceuticals. A.C.K. is an advisor for the Institute for Protein Innovation, a nonprofit research institute. Data and materials availability: Coordinates and structure factors for the AT1R-AT110i1 complexes with AngII, TRV023, and TRV026 ligands are deposited in the Protein Data Bank under accession codes 6OS0, 6OS1, and 6OS2, respectively.

Stay Connected to Science

Navigate This Article