Structural Features for Functional Selectivity at Serotonin Receptors

See allHide authors and affiliations

Science  03 May 2013:
Vol. 340, Issue 6132, pp. 615-619
DOI: 10.1126/science.1232808

Dissecting Serotonin Receptors

Serotonin receptors are the targets for many widely used drugs prescribed to treat ailments from depression to obesity and migraine headaches (see the Perspective by Palczewski and Kiser). C. Wang et al. (p. 610, published online 21 March) and Wacker et al. (p. 615, published online 21 March) describe crystal structures of two members of the serotonin family of receptors bound to antimigraine medications or to a precursor of the hallucinogenic drug LSD. Subtle differences in the way particular ligands bind to the receptors cause substantial differences in the signals generated by the receptor and the consequent biological responses. The structures reveal how the same ligand can activate one or both of the two main serotonin receptor signaling mechanisms, depending on which particular receptor it binds.


Drugs active at G protein–coupled receptors (GPCRs) can differentially modulate either canonical or noncanonical signaling pathways via a phenomenon known as functional selectivity or biased signaling. We report biochemical studies showing that the hallucinogen lysergic acid diethylamide, its precursor ergotamine (ERG), and related ergolines display strong functional selectivity for β-arrestin signaling at the 5-HT2B 5-hydroxytryptamine (5-HT) receptor, whereas they are relatively unbiased at the 5-HT1B receptor. To investigate the structural basis for biased signaling, we determined the crystal structure of the human 5-HT2B receptor bound to ERG and compared it with the 5-HT1B/ERG structure. Given the relatively poor understanding of GPCR structure and function to date, insight into different GPCR signaling pathways is important to better understand both adverse and favorable therapeutic activities.

Apart from canonical G protein–mediated signaling, G protein–coupled receptors (GPCRs) also activate noncanonical G protein–independent pathways, frequently mediated by β-arrestins (1, 2). So-called “biased” GPCR agonists differentially activate signaling pathways with distinct efficacies and potencies as compared with unbiased agonists that activate both pathways equally (3). This preferential activation of one pathway over the other has been termed “functional selectivity” or “signaling bias” (25). Depending on the receptor, biased signaling patterns are key for mediating inflammation (6), apoptosis (7), and many other processes (2). Biased ligands have been proposed to stabilize receptor conformations that are distinct from those induced by unbiased ligands and selectively change the propensity of GPCR coupling to either G proteins or β-arrestins (2).

Agonist-induced changes in trigger motifs of GPCRs (8) near the binding pocket facilitate large-scale helical movements that are accompanied by rearrangements in highly conserved residues called “microswitches” (9) that prime GPCRs for subsequent G protein binding and activation (10). The structural features of a signaling-biased receptor state remain elusive, and although complexes of two β-arrestin–biased ligands with the β1-adrenergic receptor (β1AR) have recently been solved (11), they did not reveal activation-related changes in the receptor.

To elucidate molecular and structural details of biased signaling, we characterized G protein– and β-arrestin–mediated signaling at G protein–coupled serotonin [5-hydroxytryptamine (5-HT)] receptors with several representative ergolines, such as lysergic acid diethylamide (LSD) and ergotamine (ERG). Additionally, we solved the crystal structure of the 5-HT2B receptor in complex with ERG, which was identified as a highly biased agonist for the 5-HT2B receptor (12).

To investigate potential differences of ergoline signaling at 5-HT receptors, we examined three prototypical serotonin receptors that interact with distinct G proteins. The 5-HT1B receptor inhibits cyclic adenosine monophosphate (cAMP) production through Gi, the 5-HT2B receptor mediates phospholipase C activation through Gq, and the 5-HT7A receptor stimulates cAMP production through Gs (13). We compared G protein– and β-arrestin–mediated signaling at cloned human 5-HT1B and 5-HT2B receptors and G protein–mediated signaling at 5-HT7A receptors stimulated by selective and nonselective ligands in human embryonic kidney (HEK) 293 cells (Fig. 1 and table S1) (14).

Fig. 1 Distinct signaling properties of LSD and ERG at 5-HT1B, 5-HT7A, and 5-HT2B receptors.

We used luminescence-based assays to measure 5-HT1B receptor–mediated Gi activation and cAMP production, fluorescence-based calcium mobilization assays to measure 5-HT2B receptor–mediated Gq activation, and β-arrestin translocation-dependent luciferase reporter assays to measure 5-HT1B and 5-HT2B receptor–mediated β-arrestin recruitment, all in HEK293 derived cells. (A) Normalized concentration-response studies for LSD and ERG at human cloned 5-HT1B receptor–mediated activation of Gi and noncanonical (arrestin) signaling. (B) Normalized concentration-response studies for LSD and ERG at human cloned 5-HT7A receptor–mediated activation of Gs signaling in the presence and absence of 5-HT. The solid red circles for LSD-Gs signals are superimposed by the solid blue squares for ERG-Gs signals and, thus, not visible. (C) Normalized concentration-response studies for LSD and ERG at human cloned 5-HT2B receptor–mediated activation of Gq and noncanonical (arrestin) signaling. (D) Mean β-arrestin bias factors were calculated for serotonergic agonists at 5-HT2B and 5-HT1B receptors. Concentration-responses curves were fit to the Black and Leff operational model to obtain transduction coefficients [Log(τ/KA)] (where τ is agonist efficacy and KA is the equilibrium dissociation constant) for each ligand at each corresponding pathway. The ΔLog(τ/KA) was then calculated with 5-HT as a reference agonist for each pathway, and the ΔΔLog(τ/KA) was calculated between two pathways for each ligand. The bias factor is unitless and defined as Embedded Image (28). Compounds with values close to one represent unbiased agonists, whereas compounds with large numerical values, typically >100, represent extremely biased agonists. *P < 0.0001 via two-way analysis of variance comparing 5-HT2B versus 5-HT1B bias factors; n = 3 to 6 separate experiments. ERG, DHE, and, to a lesser extent, LSD, MTE, and PER show strong β-arrestin bias at the 5-HT2B receptor, but not the 5-HT1B receptor. Error bars in (A) to (C) denote SEM from a minimum of three assays.

LSD and, especially, ERG displayed bias for β-arrestin signaling at 5-HT2B (bias factors 101 and 228, respectively) (Fig. 1D), minimal bias at 5-HT1B (bias factors 5 and 25, respectively) (Fig. 1D), and G protein antagonism at 5-HT7A receptors (Fig. 1B and table S1). We also found significant β-arrestin signaling bias for other ergolines—such as dihydroergotamine (DHE), methylergonovine (MTE), pergolide (PER), and cabergoline (CAB)—at the 5-HT2B receptor, whereas all other evaluated compounds showed no significant bias (Fig. 1D). ERG and DHE, both of which contain a large tripeptide moiety substitution at the amide scaffold, displayed more extreme signaling bias at the 5-HT2B receptor compared with LSD.

To investigate the molecular details responsible for biased signaling, we crystallized an engineered 5-HT2B receptor construct in complex with ERG, solved its structure at 2.7 Å (fig. S1 and S2 and table S3), and compared it with the structure of 5-HT1B/ERG reported in the companion manuscript (15), as well as to other known unbiased active-state GPCR structures. Residues P5.50, I3.40, and F6.44 (16, 17), the “P-I-F” motif (P, Pro; I, Ile; F, Phe), form an interface between helices V, III, and VI near the base of the ligand binding pocket in β2AR and many other aminergic receptors, including all 5-HT GPCRs. In the active-state structures of β2AR (8, 18), a chain of conformational rearrangements occurs in the P-I-F residues, in which an inward shift of helix V residue P2115.50 is coupled with: (i) a rotamer switch in I1213.40, (ii) a large movement of the F2826.44 side chain, and (iii) a corresponding rotation of helix VI on the cytoplasmic side (8). The 5-HT1B and 5-HT2B receptor structures display two different conformations of the P-I-F motif (Fig. 2). For the 5-HT1B receptor, we observe that the P-I-F configuration is essentially identical to that of the active-state of β2AR [β2AR-R*, Protein Data Bank identification number (PDB ID): 3SN6 (18)] (Fig. 2B). Whereas the 5-HT2B receptor adopts a similar active-like conformation of P2295.50 and I1433.40, the side-chain conformation of F3336.44 was similar to that observed in the inactive β2AR [β2AR-R, PDB ID: 2RH1 (19)] (Fig. 2C and fig. S2B). The P-I-F motif, therefore, appears to be in an active-like conformation in the 5-HT1B structure, but only in an intermediate active conformation in the 5-HT2B receptor structure.

Fig. 2 Trigger motif P-I-F displays active and intermediate active states for 5-HT1B and 5-HT2B receptors, respectively.

Residues of the P-I-F motif are highlighted in a dashed red circle in each panel. (A) Overall architecture of the 5-HT2B receptor (green) bound to ERG (magenta); residues of the P5.50 I3.40 F6.44 motif are illustrated in space-filling representation. (B) Alignment between 5-HT1B receptor (gray), β2AR-R (magenta; PDB ID: 2RH1), and β2AR-R*(yellow; PDB ID: 3SN6) indicates an activated P-I-F motif in the 5-HT1B/ERG structure. Black arrows denote likely rearrangements upon 5-HT1B receptor activation, according to analysis of β2AR. (C) Alignment between 5-HT2B receptor (green), β2AR-R (magenta), and β2AR-R* (yellow) suggests an intermediate active state of the P-I-F motif in the 5-HT2B/ERG structure, with F6.44 in an inactive conformation. Black arrows indicate likely rearrangements upon 5-HT2B receptor activation, according to analysis of β2AR.

At the cytoplasmic side of the receptors, GPCR activation is generally characterized by the displacement of helices V, VI, and VII (10). The magnitude of the helical motions depends on the activation-state; for example, the outward displacement of the intracellular tip of helix VI ranges between 3 and 14 Å, whereas helix VII shifts between 3 to 5 Å toward the receptor core (10, 1822). These concerted rearrangements induce an opening of the helical bundle at the cytoplasmic side, which facilitates the binding and subsequent activation of G proteins. Analysis of the 5-HT1B and 5-HT2B receptor structures shows that the conformation in the intracellular half of the helical bundle is notably shifted toward that seen in active-state GPCR structures (Fig. 3 and figs. S3 and S4) and is distinct from those of inactive-state GPCR structures (Fig. 3 and figs. S5 and S6). In the 5-HT1B and 5-HT2B receptor structures, the helix VI intracellular part is located at least 2 to 4 Å further away from the receptor core than in the inactive-state structures of other aminergic GPCRs (fig. S5). This conformation of helix VI is close to the one observed in active-state structures of the A2A adenosine receptor (A2AAR) (fig. S3) and rhodopsin (Rho) (fig. S4), though the outward shift of the helix is smaller in magnitude compared with that of the G protein–bound β2AR (Fig. 3, A and C). The only difference that we observed in helix VI between 5-HT receptor subtypes was a small clockwise rotation in the 5-HT1B receptor toward the active state that is absent in the 5-HT2B structure (Figs. 2, B and C, and 3, A and C). Helix VII in both 5-HT receptor structures also displays intermediate active states when compared with β2AR (Fig. 3, B and D), where it is shifted toward the receptor core as compared with inactive-state structures of other aminergic GPCRs (fig. S6). Whereas the 5-HT2B/ERG receptor structure shows less pronounced active-like changes in helix VI, helix VII appears to be in a more active conformation than it is in the 5-HT1B/ERG receptor structure (Fig. 3, B and D).

Fig. 3 Structural alignment with β2AR-R and β2AR-R* reveals distinct active-state seven-TM conformations of the 5-HT1B and 5-HT2B receptor structures.

All structures were analyzed based on the last membrane-embedded residue to minimize the effect of G proteins and fusion partners on the relative helix positions (see supplementary materials). The center panel shows the overall seven-TM configuration of β2AR-R (magenta; PDB ID: 2RH1), β2AR-R*(yellow; PDB ID: 3SN6), 5-HT1B (gray), and 5-HT2B (green) receptors aligned through helices I to IV. Membrane boundaries are indicated by gray dots according to the Orientations of Proteins in Membranes database (29). (A and C) Intracellular view of helices V and VI in β2AR-R, β2AR-R* compared with (A) the 5-HT1B receptor or (C) the 5-HT2B receptor. Residues on the intracellular side of the membrane have been removed for better comparison of ligand-induced helical rearrangements (see main text). Helix VI of the 5-HT1B and 5-HT2B receptors is in an intermediate active state compared with β2AR. (B and D) Side view of helices VII and VIII in β2AR-R, β2AR-R* and the (B) 5-HT1B receptor or (D) 5-HT2B receptor. Helix VII of the 5-HT1B and 5-HT2B receptors is in an intermediate active state compared with β2AR, with more pronounced activation features for the 5-HT2B receptor.

Another important aspect of GPCR activation is the rearrangement of side chains in highly conserved motifs D(E)/RY (helix III) and NPxxY (helix VII) (D, Asp; E, Glu; R, Arg; Y, Tyr; N, Asn; x, any amino acid), which are referred to as microswitches (9). Thus, the D(E)/RY motif of all inactive-state and most active-state GPCR structures shows an intact salt bridge between the side chains of D(E)3.49 and R3.50. Importantly, this salt bridge is broken only in the active-state structures of β2AR-Gαβγ (18) and Opsin-GαCT (21), where R3.50 interacts instead with the G protein and Gα peptide, respectively. The salt bridge is preserved between the side chains of R1533.50 and D1523.49 in the 5-HT2B receptor structure, but it is broken in the 5-HT1B receptor structure (Fig. 4, A, B, and D; and fig. S2C). In the 5-HT1B receptor structure, the D1463.49 side chain forms a hydrogen bond to Y157 in intracellular loop 2, and the R1473.50 side chain interacts with the main-chain carbonyl in a loop of the fusion protein BRIL (residue L1048; L, Leu), which is partially inserted into the G protein binding crevice (Fig. 4A). Thus, the conformation of the D(E)RY motif mimics the active state of β2AR in the 5-HT1B structure but resembles the inactive state in the 5-HT2B structure (Fig. 4, A and C; and fig. S7).

Fig. 4 Activation state of the D(E)/RY and NPxxY motifs in 5-HT1B and 5-HT2B receptors compared with β2AR-R and β2AR-R*.

Configuration of the D(E)/RY motif in (A) the 5-HT1B receptor (gray), (B) the 5-HT2B receptor (green), (C) β2AR-R* (yellow; PDB ID: 3SN6), and (D) β2AR-R (magenta; PDB ID: 2RH1). Residues of the G protein in β2AR-R* (C) and the G protein–mimicking BRIL loop (A) are highlighted in orange. The conformation of the D(E)/RY motif in the 5-HT1B receptor is similar to that observed in β2AR-R*, whereas the configuration of the 5-HT2B receptor compares to that of β2AR-R. (E and F) Conformational states of Y7.53 of the NPxxY motif and the proceeding residue 7.54 in β2AR-R, β2AR-R*, and (E) the 5-HT1B receptor or (F) 5-HT2B receptor. When compared with β2AR, the conformation of the NPxxY motif in the 5-HT1B receptor is in an intermediate active state, whereas the configuration of the 5-HT2B receptor is similar to β2AR-R*. T, Thr; C, Cys.

The highly conserved NPxxY motif at the cytoplasmic end of helix VII is another key microswitch of GPCR activation (9). Upon GPCR activation, the intracellular end of helix VII moves toward the receptor core, and a rotation of Y7.53 around the helical axis moves the side chain further into the seven-transmembrane (TM) bundle (10). Both 5-HT receptor structures show active-state conformations of the NPxxY motif when compared with β2AR, A2AAR, and Rho (Fig. 4, D and E; and figs. S2D and S7), with more pronounced activation features in the 5-HT2B receptor.

Our analysis indicates that the 5-HT1B/ERG structure has most of the attributes of a classical agonist-induced, active-like state, consistent with our biochemical findings that ERG is a comparatively unbiased agonist at the 5-HT1B receptor. In contrast, the 5-HT2B/ERG structure exhibits conformational characteristics of both the active and inactive states. The structure of β2AR-Gαβγ complex, along with recent nuclear magnetic resonance and fluorescence studies of β2AR, implicates helix VI predominantly in G protein signaling, whereas conformational changes in helix VII are associated with enhanced β-arrestin signaling (18, 23, 24). Thus, an active-like state in the helix VII conformation of the 5-HT2B receptor, but only partial changes in helix VI, mirrors the strong β-arrestin bias of ERG at 5-HT2B receptors observed in pharmacological assays.

A likely structural explanation for the distinct conformational features and biased pharmacology of ERG between 5-HT1B and 5-HT2B receptors can be found in the region of the extracellular loop 2 (ECL2) junction with helix V. In the 5-HT2B receptor structure, E212-R213-F214 forms an additional helical turn stabilized by a structured water molecule at the extracellular tip of helix V (Fig. 5). As a result, the segment of ECL2 that connect helices III and V via the conserved disulfide bond is shortened in the 5-HT2B receptor, inducing an inward shift and creating a conformational constraint on the position of the extracellular tip of helix V. Because of these rearrangements, ERG forms additional hydrophobic contacts with M2185.39, L3476.58, V3486.59, L3627.35, and K211ECL2 (M, Met; V, Val; K, Lys) (fig. S8A) and stabilizes a closer distance between helices V and VI, which form hydrophobic contacts between L2195.40, V3486.59, and L3496.60, along with a hydrogen bond between S2225.43 and N3446.55 (S, Ser) (fig. S8B). These extensive ligand-mediated interactions between helices V and VI in the 5-HT2B/ERG complex may be inferred to prevent rearrangements in helix VI and the corresponding rotation of F6.44 (8), observed in the 5-HT1B receptor and structures of other active-state GPCRs. The strengthened interactions of helix V and VI through ligand-mediated hydrogen bonds to both helices have also been linked to inhibition of G protein signaling at β2AR by the most efficacious inverse agonist (25).

Fig. 5 Structural differences in extracellular configuration of helix V between 5-HT1B and 5-HT2B receptors likely explain β-arrestin functional selectivity at the 5-HT2B receptor.

Side view (A) and top view (B) of 5-HT1B (gray) and 5-HT2B (green) receptors show a kink in the extracellular end of helix V in the 5-HT2B receptor. A water molecule (red sphere) was found to stabilize the kink through hydrogen bonds (gray dashed lines) with the E2125.33 main-chain carbonyl oxygen and the main-chain nitrogens of D2165.37 and G2155.36 (G, Gly). Main-chain atoms of both residues are shown as lines.

Ergolines predominantly signal through β-arrestin pathways at 5-HT2B receptors, whereas signaling at 5-HT1B receptors appears nonbiased. The differential signaling patterns are mirrored in the crystal structures, which show features of an intermediate active state for the 5-HT1B receptor and a β-arrestin–biased activation state for the 5-HT2B receptor. We propose a mechanism by which ERG stabilizes a conformation of the 5-HT2B receptor that selectively interferes with G protein signaling (fig. S9). The tripeptide moiety of ERG appears to interfere with G protein signaling at the 5-HT2B receptor, as ERG exhibits strongly increased β-arrestin signaling bias compared with LSD and MTE. Because both therapeutic (26) and adverse (27) drug effects have been associated with β-arrestin recruitment by GPCRs, identifying the features of biased GPCR states by known and yet to be discovered intracellular signaling proteins may facilitate the development of safer and more effective therapeutics with selective signaling profiles.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Tables S1 to S3

References (3042)

References and Notes

  1. Materials and methods and supplementary figures and tables are available as supplementary materials on Science Online.
  2. Superscripts refer to the Ballesteros-Weinstein numbering, in which the most conserved among class A GPCRs residues in each TM helix are designated x.50, where x is the helix number.
  3. Acknowledgments: This work was supported by the National Institute of General Medical Sciences (NIGMS) Protein Structure Initiative:Biology grant U54 GM094618 for biological studies and structure production (target GPCR-4) (V.K., V.C., and R.C.S.); NIH Common Fund in Structural Biology grant P50 GM073197 for technology development (V.C. and R.C.S.); the Jay and Betty Van Andel Foundation, Amway (China), grant R01 DK071662; Ministry of Science and Technology (China) grants 2012ZX09301001-005 and 2012CB910403 (H.E.X.); grants U19 MH82441, R01 MH61887, and the National Institute of Mental Health Psychoactive Drug Screening Program (X.-P.H., E.V., and B.L.R.); and the Michael Hooker Chair of Pharmacology (B.L.R.). D.W. is supported by a Boehringer Ingelheim Fonds Ph.D. Fellowship. R.C.S. is a founder and paid consultant for Receptos, a GPCR structure–based drug discovery company. We thank J. Velasquez for help on molecular biology; T. Trinh, K. Allin, and M. Chu for help on baculovirus expression; L. N. Collins for help on initial construct selection; K. Kadyshevskaya for assistance with figure preparation; A. Walker for assistance with manuscript preparation; I. Wilson for careful review and scientific feedback on the manuscript; T. Kenakin (University of North Carolina) for helpful discussions regarding the quantification of ligand bias; J. Smith, R. Fischetti, and N. Sanishvili for assistance in development and use of the minibeam and beamtime at GM/CA-CAT beamline 23-ID at the Advanced Photon Source, which is supported by National Cancer Institute grant Y1-CO-1020 and NIGMS grant Y1-GM-1104. Use of the Advanced Photon Source was supported by the Office of Science of the U.S. Department of Energy. Coordinates and the structure factors of the 5-HT2B/ERG complex have been deposited in the Protein Data Bank under the accession code 4IB4.
View Abstract

Stay Connected to Science

Navigate This Article