Structure-based discovery of nonhallucinogenic psychedelic analogs

Cao and colleagues frame the problem around the human serotonin 2A receptor (5-HT2A R), a receptor targeted by many neuropsychiatric drugs and by classical psychedelics such as LSD and psilocybin. Earlier research shows these psychedelics can produce rapid and persistent antidepressant effects, but their hallucinogenic properties limit wider clinical use. At the same time, lipid modulation of 5-HT2A R and the structural determinants of ligand bias at this receptor remained incompletely understood, constraining rational design of therapeutics that separate beneficial effects from hallucinations. This study set out to provide high-resolution structural information on 5-HT2A R bound to multiple ligands (serotonin, psilocin, LSD, lisuride and several drug-like scaffolds) and to use those structures to design and test nonhallucinogenic, arrestin-biased 5-HT2A R agonists with antidepressant-like activity in mice. The investigators aimed to characterise lipid interactions, reveal alternative ligand binding modes, and exploit those insights to create small-molecule analogs that preferentially engage β-arrestin pathways without eliciting hallucinatory behaviours in preclinical models.

The researchers used x-ray crystallography to determine multiple 5-HT2A R complex structures, including serotonin-, psilocin-, LSD-, lisuride-, lumateperone-, and IHCH-7086-bound receptors. Several structures were reported at near-atomic resolution (for example, LSD and lisuride at 2.6 Å, lumateperone at 2.45 Å, and IHCH-7086 at 2.5 Å), and electron density maps and Fo-Fc omit maps were used to assign ligand and lipid poses. Structural comparisons were made against previously reported 5-HT receptor structures to identify the orthosteric binding pocket (OBP), an extended binding pocket (EBP), and a side-extended pocket (SEP) that accommodated lipid-like density. Complementary biochemical and pharmacological assays assessed ligand function. The team performed calcium flux assays and β-arrestin2 recruitment assays to measure G protein and arrestin signalling, and used various bioluminescence resonance energy transfer (BRET) formats to probe G protein dissociation (Gq, G11, G12, G13, G15, Gz) and β-arrestin association. Binding affinities (Ki) and bias factors were calculated; transduction efficiency across pathways was summarised using relative log(t/KA) values. Site-directed mutagenesis (including alanine scans and targeted substitutions such as G2385.42S, S2395.43A, S2425.46A, W1513.28 variants, L3627.35F, and Y3707.43W) was used to test structural hypotheses about ligand contacts and functional bias. The drug-discovery component combined structure-guided medicinal chemistry and synthesis of analogs derived from antipsychotic substructures (IHCH series). The investigators characterised the new compounds in the in vitro assays described above, solved the structure of a selected analog bound to the receptor, and measured pharmacokinetic properties (half-life and brain penetration) in mice. Behavioural pharmacology used mouse models: head twitch response (HTR) detected by an automated magnetometer-based detection system, forced swimming test (FST) and tail suspension test (TST) in acute restraint stress (ARS) and corticosterone-induced depression paradigms. Antagonist blockade (MDL100907) and genetically modified mice carrying the Y3707.43W mutation were employed to probe mechanism. The extracted text does not present complete procedural details such as sample sizes, randomisation, or statistical models used for behavioural analyses.

Structural studies revealed consistent lipid-like electron density occupying a side-extended pocket (SEP) in all 5-HT2A R complexes; the density was best modelled as monoolein (the crystallisation lipid) and the glycerol moiety inserted into the SEP. Functional assays showed monoolein to be a modest G protein partial agonist at 5-HT2A R without detectable β-arrestin2 activity, and its G protein activity was blocked by the selective antagonist MDL100907. Endogenous lipids oleamide, oleylethanolamide (OEA), and 2-oleoyl glycerol (2OG) also activated G protein signalling but not β-arrestin, whereas oleic acid and oleoyl-LPA did not activate signalling. A G2385.42S substitution—mimicking other aminergic receptors—abolished lipid agonism, providing a structural explanation for lipid sensitivity that depended on a conserved glycine at position 5.42. The serotonin- and psilocin-bound structures revealed a second binding pose: the indole core sits higher in the receptor, engaging an extended binding pocket (EBP) near extracellular loop 2 rather than the canonical bottom orthosteric pocket (OBP). Both ligands maintain a salt bridge with D1553.32 and an additional hydrogen bond to N3526.55. Mutagenesis of residues lining the EBP and OBP reduced ligand affinity and/or agonist efficacy, consistent with two compatible binding poses. Notably, mutations at TM7 (including L3627.35 substitutions) affected β-arrestin association without uniformly altering Gq potency; L3627.35F abolished psilocin's and lisuride's β-arrestin association while preserving Gq potency, indicating that interactions in the EBP, particularly with TM7, influence signalling bias. Further structural comparison of LSD- and lisuride-bound receptors showed that differences in the conformation of the diethylamide/diethylurea moiety and contacts with Y3707.43 correlated with β-arrestin recruitment. The Y3707.43W mutation increased lisuride's β-arrestin efficacy and altered LSD signalling, and in mice the Y3707.43W mutation reduced or abolished head twitch responses (HTR) induced by LSD and DOI in a gene-dosage-dependent manner. Monoolein could dose-dependently induce β-arrestin signalling only in the presence of serotonin and not in the G2385.42S mutant; monoolein did not induce β-arrestin signalling in the presence of LSD. These data link ligand pose, lipid interaction, and β-arrestin activation. Using a structure-oriented synthesis strategy, the investigators identified and synthesised several rigid substructures (the IHCH series) derived from antipsychotic scaffolds that target the EBP. IHCH-7113 was a 5-HT2A R agonist with modest arrestin bias (bias factor = 1.52; Ki ≈ 758.6 nM). Solving the lumateperone–5-HT2A R structure confirmed that a tetracyclic core occupies the EBP. Modifications that reduced contacts of a 4-fluorophenyl group with the deep hydrophobic pocket converted antagonists into modest agonists; IHCH-7112 and IHCH-7120 were arrestin-biased (bias factors 6.70 and 12.76). Further hybrid analogues with 2-methoxy or 2-hydroxy substitutions produced arrestin-biased ligands with higher affinity and potency; IHCH-7079 and IHCH-7086 had Ki values of 16.98 and 12.59 nM and preferentially bound 5-HT2 receptors. The IHCH-7086–5-HT2A R structure showed that a 2-methoxyphenyl moiety avoids contacts with conserved TM5 serines, which is consistent with detectable β-arrestin signalling and minimal Gq activity. Behaviourally, classical psychedelics (LSD, DOI, psilocin) induced robust HTR in mice, whereas lisuride and the arrestin-biased analogs IHCH-7079 and IHCH-7086 did not elicit HTR. IHCH-7113 produced an HTR that was abolished by MDL100907. Interestingly, LSD-induced HTR could be abolished by nonhallucinogenic analogs lisuride, IHCH-7079, and IHCH-7086. Analysis of transduction efficiency (relative log(t/KA) values) showed that hallucinogenic psychedelics had higher transduction efficiency across both G protein and β-arrestin pathways than nonhallucinogenic analogs. In assays of antidepressant-like behaviour, acute LSD administration reduced acute restraint stress–induced immobility in the forced swim and tail suspension tests, and similar antidepressant-like effects were seen with lisuride, IHCH-7079, and IHCH-7086. The antidepressant-like actions of IHCH-7079 and IHCH-7086 were blocked by MDL100907. In a corticosterone-induced depression model, IHCH-7079 and IHCH-7086 decreased immobility relative to corticosterone-treated controls, again in an MDL100907-sensitive manner. Pharmacokinetic studies indicated that IHCH-7113, IHCH-7079, and IHCH-7086 had reasonable half-lives and good brain penetration according to the extracted text.

The authors interpret their results as providing a structural basis for lipid modulation of 5-HT2A R and for a second binding mode of serotonin and psilocin that engages an extended binding pocket. They argue that these structural insights permitted the rational design of β-arrestin-biased 5-HT2A R agonists that display antidepressant-like activity in mice without producing hallucinatory behaviours in the head twitch assay. Cao and colleagues emphasise that ligand engagement with TM7 residues in the EBP, particularly interactions involving Y3707.43, correlates with β-arrestin recruitment; however, robust hallucinogenic effects appear to require high transduction efficiency across both G protein and β-arrestin pathways. By contrast, low-efficiency, arrestin-biased agonists produced antidepressant-like effects in rodent models without eliciting HTR, suggesting that hallucinogenic actions are not strictly necessary for antidepressant efficacy in these preclinical paradigms. The authors further note that the identified SEP and the conserved glycine at position 5.42 explain lipid sensitivity of 5-HT2A R and that lipid binding can modulate signalling depending on ligand occupancy of the EBP or OBP. Limitations and uncertainties are acknowledged implicitly in the Discussion: the precise mechanisms of psychedelic therapeutic action in humans remain unclear, and animal models such as HTR and forced-swim paradigms are imperfect surrogates for human subjective experiences and clinical antidepressant effects. The extracted text does not provide a detailed catalogue of experimental limitations, sample sizes, or translational caveats beyond these general remarks. Finally, the investigators propose that the new crystal structures will accelerate the structure-based discovery of both classical psychedelics and nonhallucinogenic psychedelic analogs for neuropsychiatric indications, by enabling rational design that separates β-arrestin-mediated therapeutic signalling from the high-efficiency signalling associated with hallucinations.