Structural determinants of G protein subtype selectivity at the serotonin receptor 5-HT1A

Abstract

Activation of the serotonin receptor 5-HT1A has been shown to regulate mood and cognition, making 5-HT1A an important target in the treatment of anxiety, depression, and psychosis. Although the receptor signals through inhibitory G proteins, more work is necessary to understand differences in transducer coupling and its relation to functional activity. To develop a molecular understanding of the differences underlying transducer coupling and activation, we performed structure-activity relationship studies of 5-HT1A with distinct G proteins. Through a combination of in vitro assays, we identified a potent partial agonist that selectively engages a G protein subtype. We further investigated the differences in G protein engagement at 5-HT1A with cryo–electron microscopy, determining structures of 5-HT1A bound to distinct ligands and G protein subtypes. Combined with subsequent structure-guided mutagenesis and signaling assays, our studies uncover both orthosteric and allosteric determinants of agonist-specific stimulation of distinct transducers.

Structures reveal how a serotonin receptor binds different G proteins and how an antipsychotic selectively activates one pathway.

INTRODUCTION

The serotonin receptor 5-HT1A has been implicated in various psychiatric disorders, including major depressive disorder, schizophrenia, and anxiety disorder (1–3). The receptor is the primary target of buspirone, vilazodone, and gepirone—Food and Drug Administration–approved medications for the treatment of anxiety and depression (4–7). Even with decades of extensive study and drug development, 5-HT1A remains an emerging target in the treatment of various disorders, including Parkinson’s disease, chronic pain, attention deficiency hyperactivity disorder, autism spectrum disorder, sexual dysfunction, and eating disorders (8–10). In addition to therapeutics, 5-HT1A is also targeted by psychedelic compounds, including d-lysergic acid diethylamide (LSD), psilocin, and 5-MeO-DMT, and may play a role in mediating or modulating the physiological effects of these compounds (11, 12). Like other G protein–coupled receptors (GPCRs), 5-HT1A couples to canonical G protein and β-arrestin transducers (13), activating a variety of cellular signaling pathways. 5-HT1A, along with all 5-HT1 receptors, couples to the inhibitory class of G proteins, including G_i1-3, G_oa/b, and G_z, which inhibit adenylyl cyclases thereby reducing cellular cyclic adenosine 3′,5′-monophosphate (cAMP) levels (8, 14). Tissue distribution of these G protein subtypes varies greatly, with G_o being highest expressed in the brain followed by G_i3 (15). Splice variants of G_o also have distinct expression patterns with higher levels of G_oa and G_ob in neural and peripheral tissue, respectively (16). 5-HT1A–mediated activation of different G protein transducers in different brain circuitry has thus been associated with different physiological outcomes (17). Accordingly, innate receptor preference, tissue expression, and drug-induced G protein subtype selectivity may contribute to drug effects in a way that is not yet fully understood.

5-HT1A has been suggested to preferentially interact with different inhibitory G proteins; however, the results vary based on tissue type, model system, and experiment. In transfected HeLa cells, 5-HT1A preferentially couples to G_i3 in one study (18) and to G_i1 and G_i2 in another (19). Pull-down analyses of 5-HT1A receptors from rat brain samples suggest that G protein coupling varies by brain region; in the hippocampus and hypothalamus, 5-HT1A largely couples to G_o whereas the receptor couples to G_i3 in the cortex and anterior raphe (15). However, these results may be a function of G protein expression levels within different brain regions and not 5-HT1A preference. Studies of 5-HT1A receptors expressed in Sf9 cells indicate that the receptor lacks preference for a specific inhibitory G protein (20). Adding complexity to the innate preference of a receptor for a particular G protein, studies examining G protein subtype signaling at 5-HT1A, the D2 dopamine receptor, and other GPCRs indicate that different drugs may stimulate preferential G protein activation (21, 22). However, investigations into G protein subtype selectivity at 5-HT1A and other GPCRs have yet to identify common mechanisms underlying signaling differences (21, 23), which is likely further complicated by the use of different cellular systems and reporter assays (22, 24–26).

The following studies seek to expand on our understanding of the mechanisms underlying G protein subtype engagement at 5-HT1A. Specifically, our work identifies the antipsychotic asenapine as a 5-HT1A partial agonist that preferentially activates one inhibitory G protein subtype. By determining cryo–electron microscopy (cryo-EM) structures and structure-guided in vitro mutagenesis studies, we identify key residues within the receptor and G protein that contribute to activation of distinct G proteins. In addition, these structure-function studies suggest an important role for allosteric lipids in activating distinct G protein subtypes.

RESULTS

Activation of distinct G protein transducers by 5-HT1A

We first investigated whether commonly known 5-HT1A agonists can mediate the preferential activation of distinct G protein transducers downstream of the receptor. Using bioluminescence resonance energy transfer (BRET) G protein reporters (26), we analyzed the pharmacological activities of medications such as buspirone and asenapine (27), research chemicals including 8-OH-DPAT (28), 4-F, 5-MeO-PyrT (11, 29), tiospirone (30), as well as the psychedelic LSD in transfected human embryonic kidney (HEK) cells (Fig. 1A, fig. S1, and table S1).

Fig. 1. 5-HT1A–mediated G protein signaling stimulated by psychiatric medications at 5-HT1A and cryo-EM structures of asenapine- and buspirone-bound G protein complexes.

(A) 5-HT1A–mediated G_i1, G_oa, and G_z activation by psychiatric drugs determined by BRET. Concentration-response experiments were performed in triplicate and are averaged from two (G_z methiothepin) or at least three independent experiments. Data have been normalized against 5-HT, and errors bars denote SEM. (B) Cryo-EM structure of 5-HT1A receptor-Gα_i1 and Gα_oA/Gβ₁/Gγ₂ signaling complexes. Left: Representative structure of the 5-HT1A–G_i1 signaling complex. The 5-HT1A receptor, Gα_i1, Gβ₁, and Gγ₂ are shown in pink, dark blue, green, and yellow, respectively, with the binding poses of buspirone [teal; Protein Data Bank (PDB): 8FYX] and asenapine (gray) to the right. Right: Representative structure of the 5-HT1A–G_oa signaling complex. The 5-HT1A receptor, Gα_i1, Gβ₁, and Gγ₂ are shown in magenta, light blue, green, and yellow, respectively, with the binding poses of buspirone (aqua) and asenapine (purple) to the right.

In our assays, most of the compounds tested did not show any preference for activation of G_i1, G_oa, or G_z transducers when compared with serotonin, with the notable exception of asenapine and, to some extent, tiospirone. For instance, buspirone appears both efficacious and nearly a full agonist of 5-HT1A–mediated G_i1, G_oa, and G_z activation as compared to serotonin [buspirone G_i1 median effective concentration (EC₅₀) = 9.42 nM, E_max = 93.3%], with slightly more pronounced partial agonist activity in G_i3 signaling (buspirone G_i3 E_max = 83.7%) (Fig. 1A and table S1). By comparison, in our hands, LSD, 8-OH-DPAT, and 4-F, 5-MeO-PyrT showed near indistinguishable efficacy from serotonin and exhibited similar potencies in activating the different G proteins. Tiospirone appears to be a partial agonist in our assays that shows a slight preference for G_oa activation at 5-HT1A (G_i1 E_max = 50.5%, G_oa E_max = 82.4%, G_z E_max = 55.7%). Asenapine, on the other hand, revealed a different pattern of activation, stimulating no significant activity at G_i1 when compared to the pan serotonergic antagonist methiothepin (Fig. 1A, fig. S1, and table S1). However, asenapine’s efficacy in activating G_oa was 38.8% that of serotonin. Asenapine may stimulate low levels of G_z activity at high concentrations when compared against the control dimethyl sulfoxide (DMSO) and methiothepin; however, the low level of activation makes the exact efficacy of asenapine-stimulated G_z activation difficult to quantify (Fig. 1A and fig. S1B). These results suggest that asenapine predominately stimulates G_o transducers when acting at 5-HT1A, which we also observe for tiospirone, although to a lesser extent.

Moreover, we further confirmed that asenapine and tiospirone are partial agonists of 5-HT1A–mediated cAMP inhibition in HEK293T cells (fig. S1C), which is in line with previous reports on the drugs in electrophysiological (31) and GTPγS assays (32). Our previous work on 5-HT1A has shown that most agonists, even those reported to be partial agonists, such as the anxiolytic buspirone (32), approach full efficacy in our assays when compared to serotonin (table S1). The partial agonism of 5-HT1A–mediated cAMP inhibition or G_oa activation by asenapine in particular indicates that the drug may be selectively stimulating distinct inhibitory G protein subtypes, in a manner unlike most 5-HT1A agonists.

Determination of 5-HT1A–G protein complex structures via cryo-EM

The selective 5-HT1A–mediated activation of G_oa by asenapine prompted an evaluation of the fundamental basis of 5-HT1A–mediated G protein activation by examining both ligand and G protein binding in molecular detail. To better understand (i) how ligands may facilitate 5-HT1A–mediated activation of distinct G protein transducers and (ii) how distinct 5-HT1A conformations engage different G proteins, we determined four cryo-EM structures of 5-HT1A signaling complexes: 5-HT1A bound to buspirone and coupled to heterotrimeric G_oa and G_z as well as 5-HT1A bound to asenapine and coupled to G_i1 and G_oa (Fig. 1B, figs. S2 and S3, and table S2). Toward this end, we used previously established approaches leading to the generation of several 5-HT1A signaling complex structures, including that of the buspirone-bound 5-HT1A–G_i1 complex (11).

The 5-HT1A–G_i1 asenapine signaling complex was determined to a global resolution of 2.74 Å, comparable to the resolution of the 5-HT1A–G_i1 buspirone complex (2.62 Å) (11). Structures of 5-HT1A–G_oa signaling complexes were also determined to sub-3-Å resolution (buspirone: 2.90 Å; asenapine: 2.96 Å). Generally, each of these complexes showed the highest resolution in the core of the receptor, including the orthosteric binding site (OBP) and the Gβ subunit with resolutions below 2.8 Å (fig. S2, A to C). Sub-3-Å resolution allowed for unambiguous assignments of ligands and side chains within the OBP. Structure determination of buspirone-bound 5-HT1A–G_z complexes curiously revealed predominantly antiparallel dimers (fig. S3), which has also been observed in previous opioid receptor structures (33). Although the buspirone-bound 5-HT1A–G_z complex structure was solved to an overall resolution of 3.03 Å, the resolution of the G protein(s) is considerably lower, limiting our interpretation to ligand-receptor interactions as the OBP shows a local resolution of ~2.8 Å.

Drug binding across 5-HT1A signaling complexes

Drug binding poses across each receptor–G protein complex appear largely conserved. Across 5-HT1A–G_i1, G_oa, and G_z complex structures, buspirone assumed a kinked binding pose previously described (11) with the azaspirodecanedione moiety pointing toward a crevice between TM2 and TM3 made accessible by a rotameric shift in F112^3.28 (Fig. 2A). All 5-HT1A–asenapine structures along with previously published 5-HT1A structures (11, 34) show F112^3.28 pointing toward the core of the orthosteric binding pocket, whereas buspirone induces a shift in F112^3.28 away from the orthosteric core toward the membrane as we have previously reported (11).

Fig. 2. Comparison of agonist and antagonist ligand binding in 5-HT1 serotonin receptor structures.

(A) Zoom in of 5-HT1A OBP bound to buspirone (left; PDB: 8FYX) and asenapine (right) when coupled to G_i1 or G_oa, with ionic interactions indicated by dashed lines. (B) Solvent accessible surface area of the orthosteric binding pockets of 5-HT1A–G_i1 bound the agonist buspirone (left, teal), the G_oa selective partial agonist asenapine (middle, gray), and 5-HT1B bound to the antagonist methiothepin (right, sage; PDB: 5V54) shown from the top down (top) and side (bottom).

Buspirone is a highly potent agonist in all G protein pathways tested (table S1). The rotameric shift of F112^3.28 present in each buspirone structure suggests that this residue may play an important role in buspirone-mediated signaling. Mutation of F112^3.28 to an alanine reduces potency in G protein signaling stimulated by 5-HT by more than 100-fold at G_z and nearly 300-fold at G_i1 and G_oa (fig. S4 and table S2). The mutation seemingly ablates efficacy of buspirone at G_i1, G_oa, and G_z and asenapine’s partial agonism in G_oa. Because 5-HT still activates 5-HT1A F112^3.28A, albeit with much reduced potency, this mutation thus disproportionally affects asenapine- and buspirone-mediated signaling. As previously described, buspirone occupies a unique crevice of the 5-HT1A extended binding pocket (11). Combined with structural observations, these results suggest that this residue may contribute to the stability of buspirone’s binding in the extended binding pocket.

In the 5-HT1A–G_z structure, buspirone can be modeled in two distinct binding poses—one in which the ligand assumes the aforementioned “kinked” conformation and the other in a “straight” conformation (fig. S3, D to F), which is reminiscent of the binding pose of aripiprazole and vilazodone at 5-HT1A–G_i1 (11, 34). In the straight conformation, F112^3.28 points toward the OBP, as observed in the other 5-HT1A structures. This alternate ligand binding pose may explain why the F112A mutation affects activation of G_z less than activation of G_i1 or G_oa, but we cannot rule out that the antiparallel dimer architecture in our structure affects the buspirone binding pose in the G_z complex. Additional experiments are thus necessary to further investigate the pharmacological and potentially physiological relevance of these two potential binding poses.

Asenapine binding across signaling complexes

As with buspirone, binding of asenapine across the 5-HT1A–G_i1 and G_oa signaling complexes remains highly conserved. The ligand adopts an overall c-shaped conformation arcing toward TM6. Asenapine forms an interaction with D116^3.32, conserved in all aminergic ligands (Fig. 2A). Although asenapine adopts a radically different binding pose relative to buspirone, the receptor conformation and overall OBP structure appear nearly identical in the presence of both ligands [root mean square deviation (RMSD) = 0.289 to 1.393 Å between all complexes in Fig. 2A], except for F112^3.28. Despite differential signaling, both 5-HT1A–asenapine complexes display conserved features of active state aminergic receptors (35, 36) (fig. S5). These conclusions are based on comparison to the inactive state structure of 5-HT1B bound to methiothepin (37) as there is currently no structure of the 5-HT1A inactive state available. Both asenapine-bound structures show activation-related changes in the CWxP, PIF, and NPxxY motifs (38, 39). The ionic lock (40) formed by the D^3.49 and R^3.50 is broken in both asenapine structures—a key feature of the active state critical for G protein binding (fig. S5).

The binding pose of asenapine is distinct from most other 5-HT1A agonists as asenapine sits deep within the OBP—deeper than observed for other agonists bound to 5-HT1A (Fig. 2B and fig. S6). The binding pose is reminiscent of the binding of the pan-serotonergic antagonist methiothepin that also extends further toward the receptor core in the inactivate state crystal structure of 5-HT1B (37). The tricyclic moiety of methiothepin forms a slight arc toward TM6 and even sits notably deeper in the OBP compared to asenapine, with its tricyclic moiety seated closer to TM6. Specifically, the methylthio group extends further toward TM6 than the chlorine in asenapine, thereby potentially preventing activation-related changes. Neither asenapine nor methiothepin form interactions within the extended binding pockets toward the extracellular face; however, this is not a unique feature of antagonists and partial agonists as it is shared with the smaller tryptamine agonists 5-HT, 5-MeO-DMT, and 4-F, 5-MeO-PyrT (fig. S6).

Unexpectedly, asenapine’s binding pose does not mimic that of the binding of the structurally related tetracyclic compound mianserin bound to the 5-HT1e–G_i1 complex (41). As observed for asenapine here, mianserin also forms a c-shaped arc but bends toward TM3 and TM4 instead of TM6. Mianserin does not sit as deeply in the binding pocket of 5-HT1e compared to either asenapine at 5-HT1A or methiothepin at 5-HT1B (fig. S6). The shallow binding of mianserin compared to asenapine suggests that mianserin may clash with ECL2. Although mianserin is nearly a full agonist in 5-HT1A G protein assays, its potency is low (cAMP EC₅₀ = 669.88 nM) (fig. S1C), possibly due to an unfavorable binding pose. It should, however, be noted that the current absence of an inactive state 5-HT1A structure complicates interpretation of asenapine’s binding pose in the context of the drug’s pharmacological profile.

Orthosteric binding pocket contributions to G protein activity

To determine the contribution of interactions in the 5-HT1A OBP to the observed preferential G protein activation by asenapine, we mutated select residues to study their impact on G_i1, G_oa, and G_z activation (Fig. 3).

Fig. 3. Contributions of the orthosteric binding pocket residues to G protein signaling at 5-HT1A.

(A) Detailed view of OBP residues surrounding asenapine (gray) in the 5-HT1A–G_i1 signaling complex. (B) pEC50 of 5-HT–mediated signaling at 5-HT1A WT and receptor mutants according to G_i1, G_oa, and G_z BRET assays. 5-HT1A was mutated to alanine or corresponding residues in 5-HT2 receptors. Points represent independent experiments performed in triplicate with the bar and error bars denoting mean and SEM, respectively. (C) Concentration-response experiments of 5-HT–stimulated signaling of 5-HT1A WT and 5-HT1A S199A according to G_oa assays. Experiments were performed in triplicate and are averaged from at least three independent experiments. Data have been normalized against 5-HT, and errors bars denote SEM.

We observed that mutations within the OBP largely ablate asenapine signaling (fig. S7) and are detrimental to serotonin potency without affecting the signaling profile of 5-HT through G_i1, G_oa, and G_z (Fig. 3B and fig. S7). For instance, mutation of I189^ECL2, which was reported to aid in stabilizing ligand binding in serotonin receptors—e.g., serving as a lid to stabilize binding of LSD at 5-HT2A and 5-HT2B receptors (42)—reduces the potency of 5-HT more than 10-fold at all tested G protein transducers and ablates asenapine activity. Similarly, mutation of A365^6.55, a unique 5-HT1A residue among serotonin receptors that is implicated in the potency of various tryptamines (11), reduces the potency of serotonin at G_i1, G_oa, and G_z, and entirely ablates the ability of asenapine to activate G_oa (Fig. 3B, fig. S7, and table S3). Position 6.55 is, however, not the sole determinant of asenapine agonism in aminergic GPCRs as we previously identified potent asenapine agonism at the TA1 receptor, which contains a threonine at this position (43).

The only mutation tested that did not produce comparable reduction of 5-HT potency across G proteins and ablation of asenapine activity was S199^5.42A. This mutation reduced serotonin potency comparably across inhibitory G proteins (Fig. 3C and tables S1 and S3). However, at the same time, S199^5.42A marginally increased asenapine’s potency in G_oa [EC₅₀ wild type (WT) = 9.12 nM; EC₅₀ (S199^5.42A) = 6.84 nM] (tables S1 and S3). In the apo state of 5-HT1A, this residue has been proposed to play an important role in water networks within the binding pocket (34). If anything, this alanine substitution may thus make the receptor more amenable to stabilizing the hydrophobic asenapine by increasing the available space and/or hydrophobicity of the OBP. Despite the observed marginal increase in asenapine potency in G_oa activation, no significant activity was observed for asenapine-mediated activation of 5-HT1A S199^5.42A in G_i1 and G_z assays, suggesting that this residue plays a greater role in ligand potency over pathway-specific efficacy.

In aggregate, these results suggest that, although individual residues are critical for receptor activation, no single residue is responsible for asenapine’s partial agonism or G_oa activity, and no single binding pocket residue contributes to the preferential activation of one inhibitory G protein over another.

Conformational changes contributing to G protein subtype selectivity

Although the OBP does not display conformational or rotameric changes in the presence of different ligands and G proteins—with the exception of F112^3.28—the intracellular face of the receptor shows conformational differences in the presence of different G proteins (Fig. 4). Of note, we observe a higher resolution and better defined density near the intracellular tip of TM6 in the G_i1- versus the G_oa- and G_z-bound complexes (fig. S8). Because of the absence of interpretable density for parts of the Gα_z subunit, we will focus on the comparison of our G_i1- and G_oa-bound complexes.

Fig. 4. Conformational differences in 5-HT1A–G_i1 and G_oa complexes.

(A) Comparison of G_i1- and G_oa-coupled receptors reveals a 3.1-Å shift in the position of TM6 (as measured at Cα of N328^6.15), which is in close proximity to Gα^h4s6 (left). Detailed view of the Gα_i1^h4s6 (middle) and Gα_oa^h4s6 (right) loops with adjacent 5-HT1A TM6 residues. (B) 5-HT1A–mediated signaling through WT and mutant G_i1 determined by BRET. Concentration-response experiments were performed in triplicate and are averaged from at least three independent experiments. Data have been normalized against 5-HT, and errors bars denote SEM.

Although the overall receptor conformations are nearly identical, we observe the largest differences in the form of a 3.1-Å shift at the intracellular end of TM6 in the 5-HT1A–G_i1 relative to the 5-HT1A–G_oa signaling complexes (Fig. 4A). This effect was observed when comparing all 5-HT1A–G_i1 and G_oa complexes regardless of the ligand bound, suggesting this shift to be a result of the G protein interaction.

We next mutated both receptor and G protein residues in the interface to interrogate which interactions (i) might drive the overall more potent activation of G_oa by 5-HT1A and (ii) confer asenapine’s partial agonism in G_oa activation (figs. S9 and S10). Mutation of R339^6.29—which interacts with h5.26, the terminal residue of Gα^h5 [numbering according to Flock et al. (44)]—to alanine reduces serotonin potency between 3.4- and 8.8-fold at G_i1, G_oa, and G_z and effectively eliminates asenapine efficacy at G_oa (fig. S9, A and D, and table S4). Because previous studies suggest an interaction between GPCR residues in position 5.68 and Gα subunits (45, 46), we mutated R225^5.68 to an alanine; however, R225^5.68A does not appear to affect either 5-HT or asenapine activity (fig. S9, B and D). G_i1, G_oa, and G_z have a conserved isoleucine at position h5.16 that is in close proximity to 5-HT1A’s A222^5.65. Because studies reported that 5.56-h5.16 interactions are enriched in G_i/o/z-coupled receptors (47), we hypothesized that adding steric bulk to 5.56 via mutation to a valine might produce transducer specific effects on signaling. Although serotonin potency remains largely unchanged across transducers, the receptor mutant selectively increases asenapine efficacy at G_oa (WT E_max = 38.8%, A222V E_max = 64.7%) but not G_i1 or G_z (fig. S9, C and D, and table S4). Conceivably, introduction of a valine may enable hydrophobic contacts that stabilize the interaction between TM6 and Gα^h5 in a manner that enhances G_oa activation, thereby increasing the efficacy of asenapine in this pathway. The A222^5.65V mutation, however, did not increase asenapine efficacy at G_i1 and G_z, suggesting differences in allosteric communication between 5-HT1A and the different Gα subunits.

The intracellular tip of TM6, which is unusually long in 5-HT1A (48), is near the h4s6 loop of the Gα subunit (Fig. 4A). The G_i1^h4s6 loop in these and our previous complex structures is well defined with local resolutions near 2.8 Å and side chains readily apparent from the density map (fig. S8). Although the h4s6 loop in the 5-HT1A–G_oa complexes is apparent, the local resolution of this region is slightly lower but sufficient to unambiguously model the overall conformation of the loop (fig. S8). Comparing the sequence of the h4s6 loop shows that G_i1-3 are entirely conserved, and G_z is otherwise identical to G_i1 except for two substitutions that maintain the same polar interactions (fig. S10A). However, G_oa does not show the same degree of sequence conservation as the other inhibitory Gα subunits. G_oa^h4s6 is one residue shorter, and G_oa has a serine at position h4s6.08, whereas all other inhibitory subunits have a lysine. G_oa also has a proline at position h4s6.09 where the other inhibitory G proteins have a negatively charged aspartic or glutamic acid residue (fig. S10A). These differences result in divergent charge distributions across h4s6 in G_i1-3 and G_z relative to G_oa. Our observed structural differences in this interface suggest a fundamentally different interaction between different Gα^h4s6 loops and 5-HT1A’s TM6 that may contribute to the functional differences observed between G_i1, G_z, and G_oa.

To explore the impact of differences between G_i1 and G_oa h4s6, we mutated G_i1 D315^h4s6.09 to the proline found at the equivalent position in G_oa. G_i1 D315^h4s6.09P results in a loss of concentration-dependent agonist signaling, displaying comparable level of activities in the presence of serotonin, asenapine, or vehicle (fig. S10, B and C). However, the corresponding mutation in G_oa does not result in any aberrant activity with responses to serotonin and asenapine stimulation equivalent to WT G_oa. Our results thus indicate that conservation of this loop may be less important for 5-HT1A–mediated activation of G_oa than G_i1 (49).

Next, we mutated G_i1 K314^h4s6.08 to the corresponding serine in G_oa and vice versa to further explore the functional impact resulting from these divergent Gα^h4s6 sequences. Once again, mutagenesis of G_oa S314^h4s6.08 to the corresponding lysine in G_i1 causes no apparent functional difference relative to WT G_oa (fig. S10C). However, a G_i1 K314^h4s6.08S mutation results in a gain of function: Stimulation of 5-HT1A by asenapine activates G_i1 K314^h4s6.08S to levels comparable to that of WT G_oa (Fig. 4B and table S5). Consistent with h4s6.09 mutagenesis, G_oa is insensitive to mutation, whereas G_i1 is sensitive to potential changes in structure or electrostatics of the h4s6 loop. We next tested chimera constructs, in which we inserted the h4s6 loop (from Gα^h4.16 to Gα^h4s6.12) of G_oa into G_i1 and vice versa. Replacing the h4s6 loop in G_i1 with that of G_oa did not produce any notable asenapine efficacy, and introducing the G_i1 h4s6 loop into G_oa did not alter asenapine’s activity when compared to 5-HT and vehicle (fig. S10C and table S5). We would like to note that signaling differences, or lack thereof, in chimeric G proteins are challenging to interpret as a large section of the protein has been modified, potentially affecting the packing of the receptor–G protein interface and interactions within Gα. These data nonetheless suggest that, at 5-HT1A, the TM6-h4s6 interface plays a more important role in the activation of G_i1 than G_oa but also indicate that additional features contribute to 5-HT1A–mediated G protein activation outside of Gα^h4s6.

Impacts of allosteric modulation at the G protein interface

Our structural and functional studies of the 5-HT1A–Gα interface provide key insights into how residue differences can drive signaling through distinct inhibitory Gα subunits. However, it needs to be noted that the 5-HT1A G protein binding interface is unique as we and others have previously reported on a negatively charged phospholipid [phosphatidylinositol 4-phosphate (PtdIns4P)] that binds to an allosteric site in this interface and coordinates G_i1 (11, 34). PtdIns4P from insect cell expression systems binds stably in both our and previously published structures such that no additional PtdIns4P supplementation is required throughout the purification process (11, 34, 50). To our knowledge, no other GPCR–G protein complex structure reports the binding of a lipid at the TM6-TM7 cleft. Given that allosteric modulators play a key role in GPCR function, interactions at this interface could modulate interactions with different inhibitory Gα subunits and thus contribute to both 5-HT1A’s G protein selectivity profile and asenapine’s G_oa-selective activity.

PtdIns4P appears to be anchored by contacts with TM3, TM6, and TM8 (Fig. 5A), and previous mutagenesis of R134^3.50, K345^6.35, and K405^8.38 together severely diminishes 5-HT1A–mediated G_i1 signaling (34). However, these experiments only examined the impact of a triple mutant on G_i1 signaling, which showed reduced basal activity relative to WT receptors and insensitivity to addition of PtdIns4P. Examining each of these residues individually reveals that they are not equally important for G protein signaling: Mutagenesis of K345^6.35 and K405^8.38 to an alanine alone or in combination stimulates activity comparable to WT receptors at G_i1, G_oa, and G_z (Fig. 5B, fig. S11, and table S4). We observe only slight reductions in serotonin potency and no notable effect on asenapine’s ability to activate G_oa. In contrast to K345^6.35 and K405^8.38, the side chain of R134^3.50 forms a strong ionic interaction with the phosphate head group of PtdIns4P (Fig. 5A). Mutation of R134^3.50 to an alanine appears to drastically reduce activation of G_i1, G_oa, and G_z, which is not a result of aberrant expression or trafficking as the cell surface expression of 5-HT1A WT and R134^3.50 remain comparable (Fig. 5C). Specifically, whereas the 5-HT1A R134^3.50A mutation appears to be unable to activate G_i1 or G_z following stimulation with serotonin, the mutant receptor is still capable of activating G_oa. At high concentrations of serotonin, 5-HT1A R134^3.50A can stimulate G_oa to levels near that of WT receptors (E_max = 72.1%), albeit with reduced potency (EC₅₀ = 51.5 nM) (Fig. 5C, fig. S11, and table S4). However, we simultaneously observe a considerable reduction of receptor basal activity for the R134^3.50A mutant by ~50% compared to WT. This finding agrees with a prior work observing reduced basal activity in guanosine triphosphate (GTP) turnover for the 5-HT1A R134/K345/K405 triple mutant (34) (Fig. 5C).

Fig. 5. Lipid mediated control of G protein subtype signaling.

(A) Detailed view of the 5-HT1A–G_i1 G protein binding interface showing the head group of PtdIns4P interacting with 5-HT1A R134^3.50. Ionic interactions are indicated by dashed lines. (B and C) pEC50 of 5-HT mediated signaling at 5-HT1A WT, as well as K345A, K405A, K345A + K405A (B), and R134A mutants (C) according to G_i1, G_oa, and G_z BRET assays. Points represent independent experiments performed in triplicate with the bar and error bars denoting mean and SEM, respectively. (C) 5-HT1A–mediated signaling through 5-HT1A WT and R134A determined by G_oa BRET (middle). Concentration-response experiments were performed in triplicate and are averaged from at least three independent experiments. Data have been normalized against 5-HT and WT 5-HT1A, and error bars denote SEM. Right: ELISA measuring the cell surface expression of 5-HT1A WT and R134A. Experiments were performed twice in triplicate. Data have been normalized according to cell count and background fluorescence. ND, not determined.

Building on previous studies of the G protein interface, these results suggest that R134^3.50 serves as the primary anchor point for PtdIns4P. Together, our results suggest that R134^3.50 is required for the receptor to interact with PtdIns4P and is necessary for 5-HT1A–mediated G_i1 and G_z activation. PtdIns4P may be less important for 5-HT1A–mediated G_oa activation, further supporting our notion that this transducer may have fewer activation requirements at 5-HT1A compared to the other inhibitory Gα subunits.

DISCUSSION

Through in vitro assays, we reveal that the antipsychotic asenapine is a potent partial agonist of 5-HT1A–mediated G_o signaling with no significant activity in other G protein pathways. These observations consequently prompted a structural investigation to elucidate the molecular basis by which asenapine binds 5-HT1A, activates the receptor, and selectively stimulates G_o signaling pathways.

High-resolution cryo-EM structures of 5-HT1A–G_i1 and 5-HT1A–G_oa signaling complexes demonstrate that asenapine binds to the OBP forming interactions with conserved pocket residues previously described. Because there is currently no structural information on 5-HT1A in the inactive state, we compared our structures to that of 5-HT1B bound to the inverse agonist/antagonist methiothepin (37), which most closely resembles the chemical structure of asenapine. Curiously, asenapine sits deeper in the OBP than other 5-HT1A agonists, and its pose in the OBP is similar to methiothepin bound to 5-HT1B. On the basis of these similarities, asenapine assumes a binding pose that exhibits features observed for antagonists, particularly when compared to the shallower binding pose of the chemically related mianserin, which is a full agonist at 5-HT1e and 5-HT1F receptors (41). These findings could explain asenapine’s likely antagonism in G_i and G_z activation assays and suggest that asenapine is able to stabilize select 5-HT1A conformations that are conducive to activating G_oa, albeit insufficient to achieve full efficacy.

In an effort to delineate the mechanistic basis of asenapine’s G protein bias at 5-HT1A, we performed structural and functional studies. Our studies uncover considerable structural differences in both 5-HT1A’s TM6 positioning and the conformation of the h4s6 loop in different 5-HT1A–G protein complexes—a region previously suggested to interact with ICL3 of GPCRs (24, 49, 51). Overall, G_oa binding appears to require less displacement of TM6 and a smaller opening of the transducer binding pocket compared to G_i1, something we also previously noted for G_s and G_o complexes (52). 5-HT1A’s TM6 is tilted 3.1 Å closer to G_oa^h4s6 than G_i1^h4s6, with the different loops forming distinct conformations in the receptor’s transducer interface. Moreover, our mutational studies show that mutating G_i1’s K314^h4s6.08 to a serine found in G_oa produces asenapine-mediated partial agonism in G_i1 activation assays. Because none of the OBP residues probed in this study were solely responsible for asenapine’s G_oa activity and similar G_oa^h4s6 mutagenesis did not ablate asenapine activity, our findings collectively argue that, at 5-HT1A, G_oa has fewer activation requirements and relies on distinct allosteric networks connecting G protein transducer and receptor OBP. However, it should be considered that differences in (i) G protein binding kinetics (24, 53, 54), (ii) stabilities and kinetics of distinct G protein heterotrimers (55), and (iii) affinities for nucleotide-bound G protein heterotrimers and other states (56, 57) may also play a role in the observed differences. Despite our intriguing findings, conclusions drawn about general mechanisms of preferential G protein activation need to be carefully examined for distinct receptors. For instance, studies of different mTAAR9 and kappa opioid receptor signaling complexes showed largely conserved receptor–G protein interfaces (23, 58), and mu opioid receptors appear to preferentially activate G_z over other G protein transducers (59). However, it should be noted that studies of G_s revealed a key allosteric connection between h4s6 and the G protein’s nucleotide binding site, indicating that receptor-h4s6 interactions contribute to G protein activation (60).

In addition to 5-HT1A’s uniquely long TM6, which makes inferences about h4s6 interactions difficult to generalize to other receptors, structures of 5-HT1A reveal phospholipids bound to the transducer interface that modulate G protein activation. Lipid binding has generally been proposed to stabilize the active state of GPCRs. For instance, studies of rhodopsin suggests that negatively charged phospholipids are necessary for G_t activation rates comparable to those in native membranes (61), and NTS1-mediated G protein activation requires negatively charged lipids (62). However, 5-HT1A signaling complex structures are, to our knowledge, the only studies to date that provide direct structural evidence for phospholipids mediating the interaction between G proteins and receptors (34). Although previous studies proposed that a triad of positively charged 5-HT1A residues is critical for the allosteric effects of phospholipids (34), we herein show that only R134^3.50 appears critical for these effects. We further find that mutation of this residue, likely concomitant with loss of phospholipid binding, seemingly eliminates 5-HT1A signaling via G_i1 and G_z pathways but retains considerable G_oa signaling capabilities. These data add to our overall hypothesis that there are fewer requirements for 5-HT1A–mediated G_oa activation compared to other G protein transducers. On the G protein side, the lipid’s phospho head group appears to only interact with the backbone of the Gα subunit C terminus, whose structural positioning is largely conserved between G_oa and G_i1. However, different sequences and subtle structural differences between G_oa and other G proteins in this area could conceivably explain the G protein selective effects of the R134^3.50A mutation. It should be noted that R^3.50 is a residue of the highly conserved DRY motif that was reported to maintain the receptor in an inactive state (63) and directly interact with G proteins (40). Mutation of this residue could thus also directly affect the stability of distinct receptor states. We find that mutating R^3.50 decreases 5-HT1A basal activity, which is line with the finding that mutating R^3.50 can either increase or decrease activity, depending on the receptor (63). Similarly, our findings on R^3.50–dependent lipid effects are not without precedent. Molecular dynamics (MD) studies of the β₂ adrenergic receptor suggest that a phospholipid can form a salt bridge with R^3.50 in the active state but not the inactive state (64). Recent simulations of the A2A adenosine receptor suggest that a phosphatidylinositol 4,5-biphosphate (PIP2) molecule binds on the intracellular face of the receptor near TM6/TM7 and enhances association of the receptor with G proteins (65). Together, these studies further support our notion that the observed lipid directly interacts with 5-HT1A’s R134^3.50, stabilizes the receptor’s active state, and promotes G protein activation with distinct effects on the activation of different G protein transducers. Collectively, our studies on 5-HT1A G protein binding, ligand pharmacology, and lipid modulation thus argue that 5-HT1A inherently prefers activation of G_o over other G protein transducers, and asenapine’s low intrinsic efficacy as a partial agonist is insufficient to activate pathways other than G_o.

In sum, our studies shed light on the mechanisms by which 5-HT1A selectively engages different G protein pathways, and by which the antipsychotic asenapine can stimulate receptor activation. 5-HT1A activates all inhibitory G_i/o/z proteins (8) and regulates processes including, but not limited to, vision, taste, cardiac contractility, developmental processes, lipid metabolism, and immune cell regulation (14). Activation of these distinct transducers via 5-HT1A in different brain regions has already been linked to distinct physiological effects (8, 17). This suggests that 5-HT1A–mediated G_o activation likely controls specific neuronal processes, and thus also contributes to select aspects of asenapine’s pharmacological effects. In addition to modulation by exogenous drugs, our studies further implicate endogenous lipids in the activation of distinct signaling pathways at 5-HT1A—and possibly other GPCRs. Although PtdIns4P is broadly distributed (66), lipid composition at the plasma membrane differs across tissues and cell types (67). Lipid distribution and its regulation may thus play a key role in the activity of 5-HT1A and potentially other receptors (64, 65, 68).

Given 5-HT1A’s major role as drug target for antipsychotics, antidepressants, anxiolytics, psychedelics, and other drugs, our work should greatly facilitate the development of novel probes to uncover the (patho)physiology associated with distinct 5-HT1A signaling pathways.

MATERIALS AND METHODS

BRET signaling assays

5-HT1A–mediated activation of heterotrimeric G_i1 (Gα_i1/Gβ₃/Gγ₉), G_oa(Gα_oa/Gβ₃/Gγ₈), and G_z (Gα_z/Gβ₃/Gγ₁) was determined using previously published TRUPATH reporter constructs (26). Assays were essentially performed as previously described (11). In brief, the receptor as well as RLuc-Gα, Gβ, and eGFP-Gγ constructs were transfected with polyethyleneimine (PEI; Alfa Aesar) into HEK293T cells using a ratio of 1:1:1:1 (5-HT1A:Gα:Gβ:Gγ). Typically, 6 million to 7 million cells in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 1% (v/v) dialyzed fetal bovine serum (dFBS) in a 10-cm plate were transfected with a total of 8 μg of DNA. The next day, 20,000 cells were plated into each well of a poly-l-lysine–coated white-bottom 384-well plate (Greiner Bio One) in DMEM supplemented with 1% (v/v) dFBS. The following day, the day of the experiment, medium was aspirated and the cells were washed with Hanks’ balanced salt solution (HBSS) supplemented with 20 mM Hepes (pH 7.4). Fifteen microliters of drug solutions at 3× concentration was then added to 30 μl of HBSS supplemented with 20 mM Hepes (pH 7.4), 0.1% (w/v) bovine serum albumin (BSA), and 0.01% (w/v) ascorbic acid. Cells were incubated for 30 min at 37°C followed by addition of 15 μl of freshly prepared 30 μM coelenterazine 400a (Gold Biotechnology). Plates were then immediately read in a Victor NIVO plate reader (PerkinElmer) with 395-nm (RLuc8-coelenterazine 400a) and 510-nm (GFP2) emission filters at integration times of 1 s per well. BRET2 ratios were computed as the ratio of the GFP2 emission to RLuc8 emission. Data were plotted as a function of drug concentration, normalized to %5-HT stimulation, and analyzed using “log(agonist) versus response” in GraphPad Prism 8.0.

5-HT1A–mediated cAMP inhibition assay

Reduction in cellular cAMP levels mediated by 5-HT1A was determined in HEK293T cells (American Type Culture Collection) using a cAMP biosensor (GloSensor, Promega) (35, 69). First, HEK293T cells were cultured in DMEM supplemented with 10% (v/v) FBS and penicillin-streptomycin (Invitrogen). For transfection, cells were plated in 10-cm plates and medium was exchanged to DMEM supplemented with 1% (v/v) dFBS. Once cells reached 6 million to 7 million in a 10-cm plate, transfection was done forming particles in 500 μl of Opti-MEM (Invitrogen) using 4 μg of human 5-HT1A with an N-terminal HA (hemagglutinin) signal sequence and FLAG tag cloned into a pcDNA3 vector, 4 μg of GloSensor DNA, and 16 μl of PEI (Alfa Aesar). The following day, medium was aspirated, and cells were trypsinized and plated into each well of a poly-l-lysine–coated clear-bottom 384-well plate (Greiner Bio One) in DMEM supplemented with 1% (v/v) dFBS. Forty microliters of cells at a density of 500,000/ml was plated per well. The next day, medium was aspirated and d-luciferin (Gold Biotechnology) was loaded into cells by adding 30 μl of a 1.2 mM solution in HBSS supplemented with 20 mM Hepes (pH 7.4), 0.1% (w/v) BSA, and 0.01% (w/v) ascorbic acid. Loading was performed at 37°C for an hour, followed by addition of 15 μl of drug solutions at 3× concentration for 15 min at room temperature. To measure agonist activity, 15 μl of isoproterenol in HBSS supplemented with 20 mM Hepes (pH 7.4) was added at a final concentration of 400 nM (to activate G_s via endogenous adrenergic receptors). Luminescence intensity was determined 15 min later in a MicroBeta TriLux liquid scintillation counter (PerkinElmer). Data were plotted and analyzed in GraphPad Prism 8.0.

Enzyme-linked immunosorbent assay

To determine 5-HT1A expression levels, we performed enzyme-linked immunosorbent assay (ELISA) essentially as previously described (70). Forty-eight hours following culturing and transfection procedures as described above, HEK293T cells transfected with 5-HT1A WT or 5-HT1A R134A were plated in 96-well plates and fixed in 4% (w/v) paraformaldehyde (Thermo Fisher Scientific) for 15 min at room temperature. After washing, a portion of the cells were permeabilized with 1% (v/v) Triton X-100 (Acros Organics) as a control. Blocking was performed in a buffer containing 3% (w/v) BSA and 0.1% (v/v) Tween 20 (Thermo Fisher Scientific). Fixed cells were incubated overnight at 4°C in the blocking solution with anti-FLAG FITC (fluorescein isothiocyanate) antibody (GeneScript) targeting the N-terminal FLAG tag on the receptor construct. The following day cells were washed 4× with 0.1% (v/v) Tween 20 in phosphate-buffered saline. Measurements were performed in a Victor Nivo Multimode microplate reader at a fluorescent excitation/emission wavelength of 480/530 nm. Normalization was performed by staining cells with 1% (w/v) Janus Green (Thermo Fisher Scientific) and recording the absorbance at a wavelength of 595 nm. The data were processed in GraphPad Prism 8.0. Raw counts were normalized using absorbance data from Janus Green staining by dividing the raw counts with the absorbance values. One well was not treated with antibodies to account for autofluorescence from cells in each condition tested. The normalized counts were then subtracted by the wells that were not incubated with antibodies. Normalized, background adjusted counts for transfected cells were divided by nontransfected cells to create a ratio of counts over baseline that can be attributed to surface level expression of 5-HT1A.

Constructs and expression

Structural studies reported herein were performed with a 5-HT1A construct based on previous works (11, 34). Briefly, human 5-HT1A (UniprotKB-P08908) was cloned into a modified pFastBac vector and a cleavable HA signal sequence followed by a FLAG tag, a 10xHis tag, and a TEV protease site was introduced at the N terminus. In addition, the first 24 residues were replaced with BRIL (71) and an L125^3.41W mutation was introduced to increase expression levels (72). Heterotrimeric G protein was expressed from a single multibac virus following previous construct design (73). In brief, N-terminally 6xHis-tagged human Gβ₁ was cloned under control of a polyhedrin promoter, whereas Gγ₂-Gα_i1, Gγ₂-Gα_oa, and Gγ₂-Gα_z fusion constructs were cloned under control of a P10 promoter. Gγ₂ and Gα_i1/oa/z were fused with a GSAGSAGSA linker. The asenapine/5-HT1A–Gα_i1-Gβ₁-Gγ₂ structure used a Gα_i1 protein containing four individual point mutations (S47N, G203A, E245A, and A326S) to enhance the stability of the heterotrimeric G protein (34, 74). Protein expression was done in Sf9 cells (Expression Systems) using the Bac-to-Bac Baculovirus expression system (Invitrogen), for which bacmid DNA was generated in DH10Bac cells (Invitrogen). Initial P0 virus was obtained by addition of ~3 μg of recombinant bacmid DNA with 3 μl of FuGENE HD Transfection reagent (Promega) in 100 μl of Sf900 II media (Invitrogen) to 500,000 Sf9 cells plated in 2 ml of SF900 II media in wells of a 12-well plate. After 5 days at 27°C, the supernatant was harvested as viral stock, and high-titer recombinant P1 baculovirus (>10⁹ viral particles/ml) was generated by adding 300 μl of P0 to 30 ml of 3 × 10⁶ cells/ml and incubating cells at 27°C for 3 days. Approximate titers were estimated by flow cytometry analysis staining P1 infected cells with gp64-PE antibody (Expression Systems). Expression of both 5-HT1A and heterotrimer was done separately by infection of Sf9 cells at a cell density of 2 × 10⁶ to 3 × 10⁶ cells/ml with P1 virus at a multiplicity of infection of 5. After 48 hours of shaking at 27°C, cells expressing either the receptor or G protein were harvested by centrifugation (2500g, 15 min, 4°C) at 48 hours postinfection and stored at −80°C until use.

5-HT1A purification

5-HT1A purification was performed as previously described (11). Insect cells were lysed by thawing frozen cell pellets in a hypotonic buffer containing 10 mM Hepes (pH 7.5), 10 mM MgCl₂, 20 mM KCl, and homemade protease inhibitor cocktail [500 μM 4-(2-Aminoethyl)-benzenesulfonylfluoride (AEBSF), 1 μM E-64, 1 μM leupeptin, and 150 nM aprotinin] (Gold Biotechnology). Total cellular membranes were spun down and washed twice by homogenization in a manual Dounce homogenizer using ~30 strokes and followed by centrifugation (125,000g, 40 min, 4°C) in a high osmotic buffer containing 1 M NaCl, 10 mM Hepes (pH 7.5), 10 mM MgCl₂, 20 mM KCl, and homemade protease inhibitor cocktail. Purified membranes were directly flash frozen in liquid nitrogen and stored at −80°C. For formation of 5-HT1A–drug complexes, thawed membranes were resuspended in buffer containing 10 mM Hepes (pH 7.5), 10 mM MgCl₂, 20 mM KCl, 150 mM NaCl, homemade protease inhibitor cocktail, and 20 μM asenapine (Acros Organics) or buspirone (Alfa Aesar). Drug complexation was initiated by agitation for 1 hour at room temperature followed by equilibration at 4°C, addition of a 2 mg/ml final concentration of iodoacetamide (Acros Organics), and subsequent solubilization by addition of a final concentration of 1% (w/v) n-Dodecyl-β-D-Maltopyranoside (DDM; Anatrace), 0.2% (w/v) cholesteryl hemisuccinate (CHS; Anatrace), and homemade protease inhibitor cocktail for 2 hours at 4°C under agitation. The unsolubilized material was then removed by centrifugation (200,000g, 30 min, 4°C), and imidazole was added to the supernatant to a final concentration of 20 mM. Proteins were bound to TALON IMAC resin (Clontech). Protein-bound TALON resin was washed with 10 column volumes (CV) of Wash Buffer I [50 mM Hepes (pH 7.5), 800 mM NaCl, 0.1% (w/v) DDM, 0.02% (w/v) CHS, 20 mM imidazole, 10% (v/v) glycerol, and 10 μM drug]. The detergent was then exchanged for lauryl maltose neopentyl glycol (LMNG) by incubating the protein-bound TALON with Wash Buffer I supplemented with 0.1% LMNG for 1 hour at 4°C followed by successive washes with the following buffers: Wash Buffer II [50 mM Hepes (pH 7.5), 800 mM NaCl, 0.05% (w/v) LMNG, and 0.01% (w/v) CHS], Wash Buffer III [50 mM Hepes (pH 7.5), 800 mM NaCl, 0.01% (w/v) LMNG, and 0.002% (w/v) CHS], and Wash Buffer IV [25 mM Hepes (pH 7.5), 500 mM NaCl, 0.005% (w/v) LMNG, and 0.001% (w/v) CHS]. After the final incubation, the receptor was eluted with 25 mM Hepes (pH 7.5), 500 mM NaCl, 0.005% (w/v) LMNG, 0.001% (w/v) CHS, 10 μM drug, and 250 mM imidazole. The eluted receptor was concentrated using Vivaspin 6 Centrifugal Concentrators (Sartorius), and imidazole was removed from the protein solution by applying the sample to a PD MiniTrap Sample Column (Cytiva) according to the manufacturer’s protocol. The resulting protein sample was incubated with 200 μg of TEV protease overnight. The following day, the protein mixture was incubated with TALON resin for 10 min at 4°C to remove TEV and the uncleaved receptor from the solution. The cleaved receptor samples were further concentrated and immediately used for complexation. General protein purity and monodispersity was assessed by SDS–polyacrylamide gel electrophoresis and analytical size exclusion chromatography.

G protein purification

G_i1, G_oa, and G_z were purified according to a previous work (11). Sf9 insect cells expressing G proteins were homogenized in the following Lysis Buffer: 20 mm Hepes (pH 7.5), 100 mM NaCl, 1 mM MgCl₂, 0.01 mM GDP (guanosine diphosphate), 10% glycerol, 5 mM β-mercaptoethanol, 30 mM imidazole, 0.2% Triton X-100, and homemade protease inhibitor cocktail (500 μM AEBSF, 1 μM E-64, 1 μM leupeptin, and 150 nM aprotinin). The cytoplasmic and membrane fractions were separated via centrifugation (50,000g, 20 min, 4°C). The resulting supernatant was subjected to an additional centrifugation step (200,000g, 45 min, 4°C) to remove the insoluble material and protein aggregate. The final supernatant was bound to HisPur Ni-NTA resin (Thermo Fisher Scientific) overnight at 4°C. Protein-bound Ni-NTA resin was washed with 20 CV Lysis Buffer lacking 0.2% Triton X-100 followed by 20 CV Lysis Buffer lacking 0.2% Triton X-100 and 30 mM imidazole. Protein was eluted from the resin with Lysis Buffer lacking Triton X-100 and supplemented with 300 mM imidazole. The eluent was concentrated using Vivaspin 6 Centrifugal Concentrators (Sartorius). Imidazole was removed from the concentrated eluent using PD MiniTrap Sample Preparation Columns (Cytiva) according to the manufacturer’s protocol. Eluted protein was further processed by addition of PreScission Protease (GeneScript) to remove the N-terminal His tag from the Gβ₁ subunit. Purified G protein and PreScission Protease were incubated overnight at 4°C to facilitate cleavage. Cleaved G proteins were separated from PreScission Protease and uncleaved proteins by 15-min incubation with TALON resin at 4°C. Cleaved G proteins were concentrated for immediate use or flash frozen and stored at −80°C.

Receptor–G protein complexation, purification, and grid preparation

To assemble drug-bound 5-HT1A–G protein complexes, the purified receptor was combined with the purified heterotrimer at a molar ratio of 1:1.2 in a total volume of 110 μl. The final buffer was adjusted to 20 mM Hepes (pH 7.5), 120 mM NaCl, 5 mM CaCl₂, 2.5 mM MgCl₂, 0.001% (w/v) LMNG, 0.0002% (w/v) CHS, and 10 μM asenapine or buspirone and supplemented with apyrase (25 mU/ml) before incubation overnight at 4°C. The next day, samples were concentrated and lastly purified over a S200 size exclusion chromatography column equilibrated in 20 mM Hepes (pH 7.5), 100 mM NaCl, 0.001% (w/v) LMNG, 0.0002% (w/v) CHS, 0.00025% glyco-diosgenin (GDN), and 10 μM drug. Peak fractions were then pooled, concentrated to ~15 to 30 mg/ml, and immediately used to prepare grids for cryo-EM data collection. To prepare cryo-EM grids for imaging, 3 μl of the samples was applied to glow-discharged holey carbon EM grids (Quantifoil 300 copper mesh, R1.2/1.3) in an EM-GP2 plunge freezer (Leica). An EM-GP2 chamber was set to 95% humidity at 12°C. Sample-coated grids were blotted for 3 to 3.3 s before plunge-freezing into liquid ethane and stored in liquid nitrogen for data collection.

Cryo-EM data collection and processing

All automatic data collection was performed on FEI Titan Krios instruments equipped with Gatan K3 direct electron detectors operated by the Simons Electron Microscopy Center at the New York Structural Biology Center (New York, NY). The microscopes were operated at 300-kV accelerating voltage, at a nominal magnification of 64,000 to 81,000 corresponding to pixel sizes of 0.41 to 0.538 Å. For each dataset, at least 4000 movies were obtained at a dose rate of 25 to 30 electrons/Ų per second with a defocus ranging from −0.5 to −1.8 μm. The total exposure time was 2 s, and intermediate frames were recorded in 0.05-s intervals, resulting in an accumulated dose of 50 to 60 electrons/Ų and a total of 40 frames per micrograph. Movies were motion corrected using MotionCor2 (75) and imported to cryoSPARC (76) for further processing. Contrast transfer functions (CTFs) were estimated using patchCTF in cryoSPARC. For the 5-HT1A–G_z complex, an initial model was produced from a subset of micrographs using blob picking, followed by extraction, two-dimensional (2D) classification, selection of key classes, and generation of a model ab initio. Subsequent map models were produced from a curated micrograph set using particles found by picking using the initial map model as a template. For the monomeric 5-HT1A–G_i1/oa complexes, previous 5-HT1A structures were used to generate templates for particle picking. It should be noted that we observe antiparallel dimers in all datasets, but whereas the 5-HT1A–G_i1/oa datasets predominantly showed monomeric complexes, the 5-HT1A–G_z dataset repeatedly showed predominantly antiparallel dimers. In all cases, particles were extracted and subjected to 2D classification, and a final particle stack was obtained by iterative rounds of 3D classification generating several bad models from rejected particles as a sink in heterogeneous refinement. Final maps were obtained by NU-refinement in cryoSPARC (77), and volumes were postprocessed using DeepEMhancer (78). For the dimeric 5-HT1A–G_z structure, refinement using C₂ symmetry was attempted and did not improve the G protein resolution. A combination of masking and local refinement of a receptor-G_z monomer yielded slight improvements in resolution still inadequate for modeling the full heterotrimeric G protein. The local resolution of the dimeric receptors, however, is below 2.8 Å, enabling modeling of the ligand and receptor with high confidence. Kinked and straight 5-HT1A–G_z buspirone complex particles could not be separated during cryo-EM data processing and refinement, possibly due to the otherwise high similarity of the receptor complexes. Structures were built in Coot (79) based on previous 5-HT1A–G_i1 complex structures (11) and further refined in PHENIX (80). Final maps were imported to PyMOL and UCSF Chimera (81) for generating the figures shown in this chapter.

Supplementary Materials

This PDF file includes:

Figs. S1 to S11

Tables S1 to S5