Mechanistic Basis for 5-HT2AR Over 5-HT2BR Activation Authors: Lingjie Tang, Xin Ji, Yumeng Wang, Dongmei Cao, Huiqiong Li, Licong He, Jianjun Cheng, Sheng Wang Published: 2026 ABSTRACT AbstractSelective 5-HT2AR agonism without 5-HT2BR activation is critical for safe psychedelic-inspired therapeutics. Here, we systematically probe the ligand-binding pockets of 5-HT2AR and 5-HT2BR and identify steric and conformational constraints at the side-extended pocket (SEP) and extended binding pocket (EBP) that dictate ligand orientation and receptor signaling. Guided by these insights, we design derivatives across tryptamine and phenethylamine scaffolds that integrate SEP and EBP engagement with scaffold modification to reinforce 5-HT2AR activation while limiting 5-HT2BR activity. Cryo-EM structures of 5-HT2AR and 5-HT2BR bound to IHCH-2330, a selective 5-HT2AR agonist and 5-HT2BR antagonist, confirm the design principle. These findings show differential activation mechanisms between 5-HT2AR and 5-HT2BR and provide a rational framework for engineering safer, functionally selective serotonergic compounds, mitigating 5-HT2BR–mediated side effects while preserving therapeutic efficacy.Subject terms: Pharmacology, Cryoelectron microscopy, Computational biophysics Introduction Psychedelics, including psilocybin and lysergic acid diethylamide (LSD), have attracted sustained interest for their potential in treating psychiatric disorders1–6. Clinical evidence continues to build: psilocybin (the prodrug of psilocin) has shown durable antidepressant effects in clinical trials7–10, and LSD demonstrating potential in alleviating substance abuse and anxiety disorders3,11,12. Although the precise mechanisms remain under active investigation, activation of 5-HT2AR is considered one of the shared mechanisms contributing to the psychoactive and neuroplastic effects of classical psychedelics13–17. This concept has inspired the development of 5-HT2AR-targeted agonists with antidepressant-like potential18–22. However, classical psychedelics often display broad receptor engagement, underscoring the importance of defining subtype-specific mechanisms for both efficacy and safety23. Of particular concern is agonist activity at 5-HT2BR, whose activation is firmly linked to valvular heart disease (VHD)24–29. Historical safety signals, including the withdrawal of fenfluramine-based anti-obesity medications, have led the U.S. Food and Drug Administration (FDA) to caution against 5-HT2BR activation in psychedelic drug development30. The high structural homology between 5-HT2AR and 5-HT2BR presents a major challenge for designing ligands that selectively activate 5-HT2AR while avoiding 5-HT2BR agonism. Nonetheless, recent structural studies have identified tractable determinants of subtype selectivity. These include: (i) transmembrane helix (TM) 7 and TM5, which influence ligand selectivity for compounds such as lisuride31 and 2-Bromo-LSD32; (ii) extracellular loop 2 (ECL2), which shapes ligand binding kinetics31,33–35; and (iii) the side-extended pocket (SEP) in 5-HT2AR, defined by the conformational flexibility of F5.38 and the presence of G5.42 (Ballesteros–Weinstein numbering shown as superscript)20,36. Although 5-HT2BR also contains G5.42, a downward conformation of F5.38 in 5-HT2BR sterically occludes the SEP, creating a potential selectivity handle that is accessible in 5-HT2AR but restricted in 5-HT2BR. Supporting this concept, our recent work showed that membrane lipids can modulate 5-HT2AR activation through SEP interactions, with minimal effects on 5-HT2BR20, underscoring a functional divergence at this site. In this work, we systematically investigate the mechanistic basis of SEP-mediated activation in 5-HT2AR versus 5-HT2BR, and report rational design principles for ligands that act as 5-HT2AR-selective agonists while avoiding—or ideally antagonizing—5-HT2BR, thereby minimizing the risk of VHD. Results Structural determinants of differential activation in 5-HT2AR and 5-HT2BR To elucidate the molecular basis for the distinct activation mechanisms of 5-HT2AR and 5-HT2BR, we conducted a comprehensive structural comparison of all available receptor conformations (Supplementary Fig. 1a). This analysis highlighted a key regulatory region—designated Reg1—positioned above the orthosteric binding pocket (OBP) and spanning from ECL2 to residue F5.38 (Fig. 1a and b). Reg1 shows subtype divergence. In 5-HT2AR, ECL2 forms a short three-residue helix near TM4, whereas in 5-HT2BR it forms a helix adjacent to TM5. Additionally, F5.38 exhibits distinct conformational dynamics: in 5-HT2AR, it adopts a horizontal orientation primarily in active states but shifts downward in inactive states, while in 5-HT2BR, it remains consistently downward in both active and inactive structures (Fig. 1b, Supplementary Fig. 1b). The difference in activation mechanisms between 5-HT2AR and 5-HT2BR. a Overall structural comparison highlights distinct conformations of TM4, ECL2, and TM5 in 5-HT2AR (PDB: 7WC6, gold) and 5-HT2BR (PDB: 5TVN, grey). Residues are labeled according to the Ballesteros–Weinstein (BW) numbering scheme. b Comparing of differences in ECL2 and SEP between 5-HT2AR (left) and 5-HT2BR (right). c Heatmap of Gαq activation and β-arrestin2 recruitment BRET responses, presented as Δlog (Emax/EC50). Data are means of n = 3 biological replicates. d Side view comparing SEP in 5-HT2AR (PDB: 22TA, gold, up left) with 5-HT2BR (grey, up right). Influence of residues constituting SEP for activity of 5-HT2AR and 5-HT2BR induced by IHCH-8152, as tested by BRET (down). e Comparing different residues surrounding SEP with potential for activation modulation of 5-HT2AR (up left) and 5-HT2BR (up right), validated by corresponding mutation function test using BRET (down). f Schematic comparing of activation mechanism in 5-HT2AR (left) and 5-HT2BR (right) through TM5. Residues and TM contributing to activation were colored in orange and for inhibition in grey. Data are mean ± SEM of n = 3 biological replicates. Source data are provided as a Source Data file. To probe the functional contributions of helix at ECL2 and F5.38 in receptor activation, we employed a dual mutagenesis strategy: (1) substitution of F5.38 with alanine and (2) disruption of the ECL2 helix through triple glycine replacement (‘ECL2-GGG’). Both mutations individually attenuated receptor activation by 5-HT and by three representative psychedelics, while the combined mutation produced the strongest impairment (Fig. 1c, Supplementary Fig. 2). These results highlight cooperative regulation by ECL2 and F5.38 in both receptors. Notably, psychedelic-specific responses further underscored these differences: the tryptamine compound psilocin showed the greatest loss of activity in 5-HT2AR mutants (Fig. 1c, Supplementary Fig. 2b), whereas LSD showed modest signaling changes across these mutations, possibly reflecting its rigid ergoline scaffold and limited engagement of the SEP-facing region, resulting in reduced dependence on SEP–ECL2 interactions for receptor activation (Fig. 1c, Supplementary Fig. 2c). Interestingly, the phenethylamine DOI was most sensitive to 5-HT2BR mutations (Fig. 1c, Supplementary Fig. 2d). Thus, while the SEP differs structurally between the two receptors, ligand-specific interactions are required to exploit this divergence productively. Structural analysis further showed that the horizontal F2345.38 conformation in 5-HT2AR cooperates with G2385.42 to form a distinctive side-extended pocket (SEP)20. Using our designed probe IHCH-8152, we validated this SEP as functional relevance: in the resolved IHCH-8152-bound structure, the ligand’s lipid tail engages the SEP while maintaining potent 5-HT2AR agonism (Fig. 1d). SEP closure via a G2385.42S mutation abolished IHCH-8152 activity in 5-HT2AR (Fig. 1d, Supplementary Fig. 3a). In 5-HT2BR, the downward F2175.38 precludes stable SEP formation, but IHCH-8152 retained weak agonism that was abolished by G2215.42S, suggesting transient access to a horizontal F2175.38 conformations (Fig. 1d, Supplementary Fig. 3a). Notably, the F5.38A mutation preserved IHCH-8152’s 5-HT2AR agonism while reducing 5-HT2BR activation, indicating receptor-specific regulatory roles for this residue. These findings imply that while F2175.38 contributes to modulation of 5-HT2BR signaling, it alone does not dictate agonist versus antagonist outcomes. Instead, its dynamic behavior complicates selective avoidance of 5-HT2BR activation. To dissect additional modulators of F5.38 and SEP, we extended our analysis to neighboring divergent residues within Reg1. Two sites—4.63/5.35—differ between subtypes (F/D in 5-HT2AR vs. K/F in 5-HT2BR) and are positioned at the top of the SEP (Fig. 1e). In 5-HT2AR, phenylalanine (F2134.63) sits on TM4 to the left of the SEP (extracellular view), while in 5-HT2BR, phenylalanine (F2145.35) lies on TM5 to the right, consistent with the ECL2 helix orientation. We hypothesized that these phenylalanines contribute indirectly to the regulation of ECL2 and F5.38. With these findings, F2134.63K mutation mildly reduced IHCH-8152 activity in 5-HT2AR, whereas K1934.63F markedly enhanced β-arrestin2 recruitment in 5-HT2BR (Supplementary Fig. 3b). D2315.35F severely impaired 5-HT2AR activation, while F2145.35A enhanced 5-HT2BR activation. We also tested residue 5.46, located at the base of the SEP, given IHCH-8152’s extended interactions with both ends of this region. As expected, S2425.46A and F2134.63K/S2255.46A mutants slightly decreased the potency of IHCH-8152 at 5-HT2AR while modestly increasing its maximal efficacy, while A2255.46S/K1934.63F enhanced potency at 5-HT2BR (Fig. 1e, Supplementary Fig. 3b). Most mutants exhibited expression levels comparable to wild type (Supplementary Fig. 3c). Collectively, these results show that the coordinated structural features within Reg1 define distinct activation landscapes in 5-HT2AR and 5-HT2BR (Fig. 1f). Reg1, together with residue 5.46, forms a continuous activation chain shared by both receptors. In 5-HT2AR, this chain favors conformational transitions toward activation due to the dynamic flexibility of Reg1, whereas in 5-HT2BR, the same region constrains F5.38 via coupling with ECL2, thereby functioning as an ‘inactive anchor’. Therefore, selective ligand design for 5-HT2AR requires perturbing this shared regulatory network in a manner that reinforces activation chain function in 5-HT2AR, while exploiting its anchoring role to suppress activation in 5-HT2BR. Multiple psilocin binding poses inspire the design of a selective agonist The subtle structural distinction between 5-HT2AR and 5-HT2BR within Reg1 necessitates precise control of ligand binding poses to reinforce activation in 5-HT2AR while stabilizing the inactive conformation in 5-HT2BR. Given that different psychedelic chemotypes engage Reg1 in distinct manners, a rational strategy is to re-engineer these ligands into 5-HT2BR antagonists that retain 5-HT2AR agonism by strengthening interactions with inactive anchor in Reg1. However, the small and highly flexible frameworks of psychedelics hinder such selective interactions. To overcome this challenge, we introduced additional substituents to rigidify the scaffold and constrain binding poses. Traditional drug design often treats sub-pocket pharmacophores independently, neglecting how binding poses of distinct moieties constrain one another across sub-pockets. Our previous inactive-state 5-HT2AR–psilocin structure showed that lipid insertion into the SEP displaced psilocin from the OBP toward the extended binding pocket (EBP), suggesting that SEP engagement can trigger ligand rotation within 5-HT2 subfamily receptors20. To probe psilocin’s pose without lipid interference, we resolved the cryo-EM structure of a psilocin-bound 5-HT2AR/miniGq/β1/γ2 complex at 3.20 Å resolution. Local refinement focused on the receptor yielded a resolution of 3.33 Å, enabling visualization of the overall psilocin density and its placement within the binding pocket (Fig. 2a, Supplementary Fig. 4, Supplementary Table 1). In contrast to the lipid-displaced complex, psilocin here remained in the OBP (Fig. 2b, Supplementary Fig. 5a). Even compared to the recently reported psilocin-bound 5-HT2AR structure32, our pose rotated 20.3° toward TM5 (away from the SEP) (Supplementary Fig. 5b and c). This observed pose variability underscores the conformational plasticity of tryptamine-like ligands and provides a structural rationale for introducing pose-stabilizing substituents. The variable binding pose of psilocin inspiring design of selective 5-HT2AR agonist. a Global Cryo-EM map of 5-HT2AR/psilocin, a local refined map of psilocin and receptor, and corresponding locally refined model showing the ligand-binding regions, psilocin and map colored by subunits (5-HT2AR, green; miniGq/γ2, light grey; Gβ1, blue; scFv16, pink; psilocin, deep green). b Psilocin applies OBP-binding pose (left with side view, green), compared with previously resolved EBP-binding inactive constructure (PDB, 7WC5, right with side view, yellow). c IHCH-2381 mimics a different binding pose of psilocin. d Gαq activation of IHCH-2381 in 5-HT2AR and 5-HT2BR as measured by BRET. e Schematic revealing the development of the IHCH-1896 series, referring to the binding pose of risperidone (PDB: 6A93). f Gαq activation of the IHCH-1896 series in 5-HT2AR and 5-HT2BR as measured by BRET, showing increased selectivity. Data are mean ± SEM of n = 3 biological replicates. Source data are provided as a Source Data file. To test this concept, we designed IHCH-2381, merging two 4-MeO-DMT molecules to mimic bidirectional OBP and EBP orientations, with the N-methyl group to bias binding toward Reg1 (Fig. 2c). IHCH-2381 exhibited potent 5-HT2AR agonism in both Gαq and β-arrestin signals, confirming that psilocin-derived ligands can access the EBP (Fig. 2d, Supplementary Fig. 5d, Supplementary Tables 2–4). However, it also strongly activated 5-HT2BR, underscoring the need for improved selectivity. Inspired by risperidone’s EBP-binding scaffold, we synthesized IHCH-1896 by attaching a tetrahydropyridopyrimidinone moiety to 4-MeO-DMT (Fig. 2e). This compound showed reduced potency at 5-HT2BR (Fig. 2f, Supplementary Fig. 5e, Supplementary Tables 2 and 3). Further scaffold refinement yielded IHCH-1897, bearing a pyridopyrimidinone moiety, which enhanced 5-HT2AR agonism and selectivity over 5-HT2BR. We propose that IHCH-1897’s pyridopyrimidinone group engages the EBP while its 4-methoxyindole moiety occupies the OBP. Nonetheless, OBP-bound methoxy orientation limited SEP exploitation and thus full 5-HT2BR selectivity. To address this, we modified the OBP-bound moiety by shifting the indole methoxy from the 4- to the 7-position, producing IHCH-2389, which promoted SEP-inactive anchor engagement through upward indole displacement. IHCH-2389 preserved strong 5-HT2AR while further reducing 5-HT2BR activation in both Gαq and β-arrestin signals (Fig. 2f, Supplementary Fig. 5e). Collectively, these results demonstrate that dual engagement of the OBP and EBP is critical for activation and that strategic incorporation of SEP-targeting substructures provides a viable route to 5-HT2AR-selective agonism while minimizing 5-HT2BR activity. Customized strategy achieves a selective 5-HT2AR agonist utilizing SEP We next explored phenethylamine scaffolds, as DOI is among the most potent 5-HT2AR agonists and the 25CN-NBOH series exhibits relative 5-HT2AR selectivity34. Consistent with our hypothesis, SEP mutations differentially modulated DOI activity at 5-HT2AR and 5-HT2BR (Fig. 1c), implicating SEP as a key discriminator. In a previously resolved 25CN-NBOH-bound 5-HT2AR structure34, the 2-hydroxyphenyl group adopts a downward orientation that pushes the 2, 4, 5-substituted phenyl group upward toward the SEP anchor, thereby preventing interaction with S2425.46 and contributing to selectivity (Fig. 3a). To exploit this mechanism, we introduced the pyridopyrimidinone group to the basic nitrogen atom and synthesized compound IHCH-2302 (Fig. 3b). IHCH-2302 behaved as a 5-HT2BR partial agonist similar to 25CN-NBOH. To better anchor the EBP-binding moiety, we further introduced an N-methyl substitution as in IHCH-2308, which maintained identical 5-HT2AR signaling but reduced 5-HT2BR agonism (Fig. 3c, Supplementary Fig. 5f, Supplementary Tables 2–4). These results highlight the restrictor’s role in orienting ligands toward the SEP anchor in 5-HT2BR. Strategy for selective 5-HT2AR agonist design utilizing TM5. a Regulation of the binding pose of 25CN-NBOH (PDB: 6WHA) by restriction of 2-hydroxyphenyl group compared with mescaline (PDB: 9AS6). b Schematic of design strategy for IHCH-2330 series from phenethylamine psychedelic DOI. c Gαq activation of reference. 25CN-NBOH and IHCH-2330 series in 5-HT2AR and 5-HT2BR as measured by BRET. Data are mean ± SEM of n = 3 biological replicates. Source data are provided as a Source Data file. Based on these insights, we further optimized SEP interactions by replacing the bromo substituent with a trifluoromethyl group, yielding IHCH-2330 (Fig. 3b). This compound displayed potent 5-HT2AR agonism with concurrent 5-HT2BR antagonism (Fig. 3c, Supplementary Fig. 5f, Supplementary Table 2-4). IHCH-2302, IHCH-2308, and IHCH-2330 showed a pronounced reduction in β-arrestin activity at 5-HT2BR, mirroring their effects on Gαq signaling, demonstrating a shared mechanism of selectivity for 5-HT2AR over 5-HT2BR at SEP. Functional profiling across aminergic GPCRs showed that IHCH-2330 achieved 9-, 19-, 24-, 40- and 62-fold selectivity over 5-HT1DR, 5-HT1BR, 5-HT6R, β2-AR and 5-HT1FR, respectively. Apart from the 5-HT2 subfamily, IHCH-2330 exhibited negligible agonist activity at the remaining 5-HT receptors. It displayed only weak antagonist activity at α1A-AR, α1B-AR, α2C-AR, and M1R (IC50 > 1 μM), no detectable activity at dopamine, histamine, other adrenergic and muscarinic receptors (Supplementary Table 5). Notably, the strong activity at 5-HT2CR suggests a high degree of structural similarity in the SEP between 5-HT2AR and 5-HT2CR. Overall, these results identify SEP as a critical structural determinant for designing functionally selective 5-HT2AR agonists. By integrating scaffold restrictors, EBP-binding modules, and SEP-targeting substituents, we demonstrate a generalizable strategy for engineering ligands that promote 5-HT2AR activation while suppressing 5-HT2BR activity. Structural insights into the selectivity mechanism of IHCH-2330 Encouraged by the potent 5-HT2AR activity and selectivity of IHCH-2330, we determined cryo-EM structures of IHCH-2330 bound to active 5-HT2AR and inactive 5-HT2BR at 3.13 and 2.98 Å resolution, respectively (Supplementary Figs. 6, 7, Supplementary Table 1). To better assess receptor features, local refinement focused on the receptor was performed, yielding resolutions of 3.20 Å for 5-HT2AR and 3.21 Å for 5-HT2BR. Local resolution heatmaps from the local refinement show improved resolution within the orthosteric binding pocket, while extracellular loop regions display lower local resolution consistent with their intrinsic flexibility (Supplementary Figs. 6b, 6d and 7b, 7d). In the 5-HT2AR structure, the density map clearly resolves the overall binding pose of IHCH-2330. In contrast, the density corresponding to the pyridopyrimidinone moiety of IHCH-2330 is less well resolved in the 5-HT2BR structure. Notably, the OBP—engaging 2,4,5-substituted phenyl group of IHCH-2330 is well defined in both structures, allowing reliable analysis of its interactions with the receptor and the SEP region (Fig. 4a and b). The tertiary nitrogen atom on the linker between both aromatic pharmacophores forms the canonical ionic interaction with D3.32, conserved across aminergic receptors (Fig. 4c). The N-methyl substituent adopts a downward orientation into a cavity created by the absent side chain of G3697.42, thereby restricting the relative conformation of the two pharmacophore groups (Fig. 4d). This restriction is critical, as the analog IHCH-2302—lacking this methyl group—displayed 5-HT2BR agonism. Thus, the linker N-methyl substituent acts as a structural restrictor that stabilizes the subtle binding pose necessary for selective engagement of EBP and SEP. The structure of 5-HT2AR and 5-HT2BR complexed with IHCH-2330. a Global Cryo-EM map of 5-HT2AR/IHCH-2330, local refined map of IHCH-2330 and receptor, and corresponding locally refined model showing the ligand-binding regions, colored by subunits (5-HT2AR, salmon; miniGq/γ2, pink; Gβ1, grey; scFv16, light yellow, IHCH-2330, orange). b Global Cryo-EM map of 5-HT2BR/IHCH-2330, local refined map of IHCH-2330 and receptor, and corresponding locally refined model showing the ligand-binding regions, colored by subunits (5-HT2AR, light blue; P2C2-HL, light yellow; P2C2-LL, light green, IHCH-2330, purple). c Binding pocket of IHCH-2330 in 5-HT2AR and 5-HT2BR. The salt bridge interaction with D3.32 is shown as yellow dashed lines. d Side view of IHCH-2330 binding pocket in 5-HT2AR and 5-HT2BR. Cavity provided by G7.42 allows binding of the methyl of IHCH-2330. e Comparing the binding pose of IHCH-2330 in 5-HT2AR and 5-HT2BR. f Comparing of binding pose of IHCH-2330 in 5-HT2AR with mescaline (PDB: 9AS6, left) and 25CN-NBOH (PDB: 6WHA, right). IHCH-2330 binds closer to TM5 than these two phenethylamine analogs. The pyridopyrimidinone group of IHCH-2330 occupies the EBP and OBP in both receptors (Fig. 4c–e). At SEP, IHCH-2330 sits closer to TM5 in 5-HT2AR, where its trifluoromethyl group penetrates deeply into a hydrophobic “clamp” formed by V2355.39 and the flexible F2345.38, making tight contact with TM5 (Fig. 4e). In contrast, in 5-HT2BR, the larger M2185.39 and the more rigid F2175.38—stabilized by the inactive anchor—push the 4-substituted phenethylamine group slightly away from TM5. Together with the distinct roles of the activation chain in the two receptors, these alternative poses enable IHCH-2330 to promote 5-HT2AR activation while simultaneously antagonizing 5-HT2BR. To contextualize this hypothesis, we compared the IHCH-2330–bound 5-HT2AR structure with previously resolved complexes of 25CN-NBOH and mescaline32,34 (Fig. 4f). Both reference ligands, lacking an EBP-binding group, position their protonated amines more shallowly within the orthosteric pocket. Mescaline, bearing a 3-methoxy rather than the 4-methoxy present in 25CN-NBOH and IHCH-2330, is sterically displaced 1.0 Å away from TM5 (relative to IHCH-2330). By contrast, in 25CN-NBOH, the bulky N-benzyl substituent forces the 4-substituted phenethylamine moiety to rotate upward by 25° away from TM5 (again contrasting with IHCH-2330). Collectively, these deviations highlight that without a strong EBP anchor, ligands become misaligned around the SEP, explaining their 5-HT2BR agonism, whereas IHCH-2330 maintains precise SEP engagement that confers functional selectivity. Distinct modulation mechanism of SEP provides an approach for rational compound design The tight attachment of IHCH-2330 in 5-HT2AR and 5-HT2BR suggests a crucial contribution of SEP to ligand modulation in the two receptors (Fig. 5a, Supplementary Fig. 8a). To test the role of the inactive anchor in selectivity, we introduced the K1934.63F mutation into 5-HT2BR. Consistent with our hypothesis, IHCH-2330 switched from antagonist to agonist in this mutant (Supplementary Fig. 8b). We further examined the role of EBP by mutating V3667.39, located at the base of the EBP, to tyrosine, which elevates the pyridopyrimidinone group of IHCH-2330 (Supplementary Fig. 8a). The K1934.63F/V3667.39Y double mutant markedly enhanced 5-HT2BR activity (Supplementary Fig. 8b and c). Additional combination with F2145.35A and A2255.46S mutations, which weaken the inactive anchor and displace the ligand downward, yielded maximal agonism in the K1934.63F/F2145.35A/A2255.46S/V3667.39Y quadruple mutant (Fig. 5b, Supplementary Fig. 8b and c). Conversely, stepwise introduction of reciprocal mutations in 5-HT2AR (F2134.63K, D2315.35F, S2425.46A, V3667.39A) progressively diminished IHCH-2330–induced agonism activity, with the quadruple mutant showing the strongest loss of activity. These results confirm that EBP stabilizes binding conformations in SEP/OBP, but it is the SEP configuration that primarily drives 5-HT2AR selectivity of IHCH-2330. The selectivity mechanism of IHCH-2330. a Contribution of the combined regulation of IHCH-2330 binding posture by EBP (left) and SEP (right) to 5-HT2AR selectivity. Modulation in 5-HT2AR (up) and 5-HT2BR (down) was shown, respectively. b Decrease of 5-HT2AR agonism activity and rescue of 5-HT2BR agonism activity of IHCH-2330 by EBP mutation, combining with SEP mutation, as measured by BRET. c Schematic of strategy for rational design of functional 5-HT2AR selectivity agonist utilizing distinct modulation mechanism of TM5 between 5-HT2AR and 5-HT2BR. Data are mean ± SEM of n = 3 biological replicates. Source data are provided as a Source Data file. To further evaluate our hypothesis, molecular dynamics simulations were performed for the 5-HT2AR/miniGαq/IHCH-2330 and 5-HT2BR/IHCH-2330 complexes. Statistical analysis showed receptor-dependent differences in ECL2 behavior. In 5-HT2AR, a short segment of ECL2 (residue 218–220) transiently adopted a helical conformation for ~14% of the simulation time, indicative of substantial conformational flexibility (Supplementary Fig. 9a). In contrast, a different segment within ECL2 in 5-HT2BR (residue 211-–213) remained stably helical for ~81% of the trajectory, consistent with a more rigid configuration closely associated with TM5. Further analysis showed that F2175.38 in 5-HT2BR remained in close and persistent contact with IHCH-2330 throughout the simulation, whereas the corresponding F2345.38 in 5-HT2AR displayed greater positional variability relative to the ligand (Supplementary Fig. 9b). These observations suggest that the stable ECL2 helix and its surrounding residues in 5-HT2BR may restrict the conformational flexibility of F2175.38, potentially limiting SEP-adjacent rearrangements associated with receptor activation. Notably, the overall binding pose of the OBP-engaging moiety of IHCH-2330 was comparable in the two receptors, suggesting that the divergent pharmacological outcomes are unlikely to arise from fundamentally distinct ligand poses. Rather, the data are more consistent with receptor-intrinsic differences in extracellular and TM5-associated dynamics that modulate how SEP-engaging ligands are accommodated. Together, these results identify SEP as the key determinant of 5-HT2AR selectivity. Structural divergence at TM4–TM5, originating from helix positioning in ECL2 and transmitted via hydrophobic networks, leads to different flexibilities of F5.38 and creates the inactive anchor at the top of SEP in 5-HT2BR (Fig. 5c). Leveraging this divergence provides a generalizable design strategy: direct engagement of the SEP inactive anchor, reinforced by an EBP-stabilizing group (Fig. 5c). Applied in a scaffold-dependent manner, this framework enables the rational design of functionally selective 5-HT2AR agonists over 5-HT2BR. Discussion Psychedelics show promising therapeutic efficacy in psychiatric disorders, but their polypharmacology raises the risk of side effects. In particular, cardiovascular toxicity from 5-HT2BR agonism led to the withdrawal of fenfluramine-like drugs and was highlighted in the 2023 FDA guidance on psychedelic drug development. Although multiple 5-HT2AR and 5-HT2BR structures have been resolved, the mechanistic distinction between their activation modes has remained elusive, limiting rational development of 5-HT2AR-selective agonists and leaving discovery largely serendipitous30. By systematically evaluating the contributions of ECL2 and TM5—the most divergent regions between the two receptors—we identified a conserved “activation chain” extending from ECL2 through F5.38 to S/A5.46. In 5-HT2AR this chain promotes activation, whereas in 5-HT2BR an “inactive anchor” embedded in the same region restrains activation. This subtle but crucial difference provides a structural basis for functional selectivity. To exploit this, we introduced EBP-binding elements to modulate ligand pose within the OBP and SEP. This strategy enabled selective modulation across two major psychedelic scaffolds—tryptamines and phenethylamines. Among these, the phenethylamine derivative IHCH-2330 displayed 5-HT2BR antagonism while retaining potent 5-HT2AR agonism. Structural studies of IHCH-2330 confirmed reciprocal interactions across OBP, EBP, and SEP that regulate the activation chain, while functional assays showed that IHCH-2330 relies on the inactive anchor for 5-HT2BR antagonism, albeit through scaffold-specific mechanisms. Our findings also help rationalize the activity of previously reported selective agonists. 2-Bromo-LSD22,32, likely engages the 5-HT2BR inactive anchor via an upward-shifted ergoline ring (Supplementary Fig. 9c). Q211837, with a disassembled ergoline scaffold, positions its fluorine substituent comparably to 2-Bromo-LSD (Supplementary Fig. 9d). Z7757 and Z250437 resemble IHCH-2330 in adopting EBP-stabilized poses that strengthen SEP interactions (Supplementary Fig. 9e and f). TBG18 achieves selectivity without an explicit EBP group, instead using an extended methoxy substituent to engage SEP (Supplementary Fig. 9g). Collectively, these cases illustrate a convergent principle: selective 5-HT2AR agonism emerges from ligand modulation of SEP, either by directly or indirectly perturbing the inactive anchor unique to 5-HT2BR. In conclusion, our findings show that a shared regulatory region encompassing ECL2 and adjacent transmembrane residues can act as either an “activation chain” in 5-HT2AR or an “inactive anchor” in 5-HT2BR. This dual role provides a structural explanation for subtype divergence and provides a conceptual framework for rational ligand design. By leveraging this asymmetry, safer and functionally selective 5-HT2AR agonists can be developed, advancing the next generation of psychedelic therapeutics. Methods Bioluminescence resonance energy transfer assay (BRET) The BRET assay was performed as previously described38. Human embryonic kidney (HEK) 293T cells (ATCC CRL-11268; mycoplasma-free) were transfected in 6-cm dishes at 60–80% confluency. Cells were transfected with 400 ng of 5-HT2AR or 5-HT2BR, 1000 ng of GFP2-krasCAAX, 1000 ng of Gαq, 40 ng of GRK2RH-RLuc8 and 1560 ng of pcDNA3.1 vector per 6-cm dish for Gαq activation39; 500 ng of receptor-RLuc8, 500 ng of GRK2 and 3000 ng of GFP2-β-arrestin2 per 6-cm dish for β-arrestin2 recruitment; 1000 ng of receptors, 1000 ng of Gα-Rluc, 1000 ng of Gβ and 1000 ng of Gγ-GFP2 corresponding to special receptors for G protein dissociation, and 800 ng of receptor-Rluc8 and 3200 ng of miniGs-GFP2 for miniGs recruitment. After 24 h, the cells were placed in a white opaque-bottom 96-well assay plate (Beyotime) containing DMEM (Thermo) supplemented with 2% dialyzed FBS (Omega Scientific). The next day, the medium was decanted, and 40 μl of 7.5 μM coelenterazine 400a (Goldbio) in drug buffer (1x Hank’s balanced salt solution (HBSS) and 20 mM HEPES, pH 7.4, components of all the drug buffers hereinbelow are the same unless stated specially) was added. After a 2-min incubation, cells were stimulated with 20 μl of serially diluted compounds in drug buffer, and 410 and 515 nm emission were measured in a Tristar 3 Multimode Reader (Berthold). BRET ratios were calculated by dividing GFP2 emission (515 nm) by Rluc8 emission (410 nm) and analyzed in GraphPad Prism 10 (GraphPad Software Inc., San Diego, CA). Receptor expression assay HEK 293 T cells (ATCC CRL-11268; mycoplasma-free) were transfected in 24-well plates at 60–80% confluency. Cells were transfected with 50 ng of wild-type or mutated receptor-RLuc8 and 450 ng of pCDNA3.1 vector per well. After 24 h, the cells were placed in a white opaque-bottom 96-well assay plate (Beyotime) containing DMEM (Thermo) supplemented with 2% dialyzed FBS (Omega Scientific). The next day, the medium was decanted and 40 μl of 7.5 μM coelenterazine 400a (Goldbio) in drug buffer. After a 2-min incubation, 515 nm emission was measured in a Tristar 3 Multimode Reader (Berthold). Expression level was normalized as % of wild-type receptor expression. Split-luciferase-based cAMP reporter assay The cAMP reporter assay was performed as previously described40. Human embryonic kidney (HEK) 293T cells (ATCC CRL-11268; mycoplasma-free) were transfected in 6-cm dishes at 60–80% confluency. Cells were transfected with 2000 ng of receptor and 2000 ng of split-luciferase-based cAMP biosensor (GloSensor; Promega) per 6-cm dish. After 24 h, the cells were placed in a white opaque-bottom 96-well assay plate (Beyotime) containing DMEM (Thermo) supplemented with 2% dialyzed FBS (Omega Scientific). The next day, the medium was decanted, and 20 μl of drug buffer was added, followed by a 5-min incubation. Then cells were stimulated with 10 μl of serially diluted compounds in drug buffer for 10 min, before adding 10 μl of 25 μM luciferin, and luminescence intensity was measured 8 min later in Ensight (PerkinElmer). Data was analyzed in GraphPad Prism 10. Radioligand-binding assay HEK 293 T cells were transfected with 10 μg of plasmids encoding the tested 5-HT receptors per 10-cm dish. The next day, the medium was transformed with DMEM containing 2% dialyzed FBS and incubated for 24 h before harvesting. The membrane with expressed receptors was harvested with HEK 293T lysis buffer (1 mM HEPES, pH 7.5 and 2 mM EDTA, adjusted to pH 7.4) and separated by centrifugation. Membrane was then resuspended into binding buffer (50 mM HEPES (pH 7.5), 0.5 mM EDTA and 5 mM MgCl2, pH 7.4) and distributed in 96-well plates with 30 μl per well and incubated with 30 μl 0.5–2 nM of [3H]-LSD (PE, for serotonin receptors), diluted in binding buffer, and 30 μl concentration gradients of compounds diluted in drug buffer containing 0.1% BSA for 2 h at room temperature. Reaction was terminated with rapid vacuum filtration onto chilled 0.3% polyethyleneimine-soaked GF/A filters, followed by three times of quick washes with cold washing buffer (50 mM Tris–HCl, pH 7.4). Counts per minute (CPM) were measured using a MicroBeta TriLux reader (PerkinElmer). Generation of constructs for structure determination The 5-HT2AR and 5-HT2BR constructs were generated as previously described34,41. In brief, for the 5-HT2AR active complex, a truncated wild-type human ΔN65-5-HT2AR-ΔC405 construct, including an N-terminal FLAG tag, a 10x His tag, a TEV protease site, a Maltose Binding Protein (MBP) and an HRV 3C protease site, was subcloned into a pFASTBac1 vector for expression in Spodoptera frugiperda (Sf9) cells. The miniGαq-γ2 and Gβ1 were subcloned into a pFASTBac-Dual vector. The human Gγ2 was fused to the N-terminal of the miniGαq subunit. The minGαq was derived from the miniGs/q58 system reported previously42, with the first 27 amino acids (AA) replaced by the first 30 AA of Gαi2, facilitating the binding of scFv16, and a K321R mutant at the C-terminal. Additionally, the N-terminal of the human Gβ1 subunit was tagged with a 6x His tag and a TEV protease site for purification purposes. For the 5-HT2BR inactive complex, M1-Q35 at the N-terminal and K406-V481 at the C-terminal of 5-HT2BR were truncated, with the ICL3 at A248-V313 being replaced with a linker ‘RLLSGSR’. M1443.41W mutation was applied for stabilization. A FLAG tag, a 10x His tag, a TEV protease site, a modified thermostabilized apocytochrome b562RIL (BRIL), an HRV 3 C protease site, and a ‘GSGSG’ linker were fused into the N-terminal of this truncated 5-HT2BR. The whole construction was subcloned into a pFASTBac1 vector. For Fab P2C2, the ST II single peptide LC was fused to the N-terminal of the heavy chain and light chain, respectively, while a TEV protease site and an 8x His tag were incorporated at the C-terminal of the heavy chain. The Fab expression cassette was subcloned into a pRH2.2 vector for expression in E. coli. Proteins expression To generate high-titer recombinant baculovirus (>109 virus particles/ml), the Bac-to-Bac baculovirus expression system (Invitrogen) was applied. 1 ml/well 0.5 million/ml Spodoptera frugiperda Sf9 cells (Expression Systems) were transfected with ~5 μg recombinant bacmid in a 12-well plate using 3 μl of Cellfectin II reagent (Invitrogen). Medium was replaced with 500 μl Sf-900 II SFM medium (Invitrogen) before incubating with the transfection mixture. After 12 h, the medium was transformed to 1 mL ESF 921 (Expression Systems) and the cells were cultured for 4 days at 27 °C to produce the P0 virus. The P1 virus was obtained by diluting the P0 virus supernatant into 50 ml 3 million/ml Sf9 cells and incubating for 3 days. For expression of 5-HT2AR and 5-HT2BR, 3 million/ml of Sf9 cells was infected with P1 virus at a multiplicity of infection = 5 and cultured for 48 h. For miniGαq-γ2/Gβ1 complex and scFv16, Sf9 cells were infected at a density of 3 and 2 million/ml, respectively, and expressed for 72 h. Sf9 cells expressing 5-HT2AR, 5-HT2BR and miniGαq-γ2/Gβ1 complex were collected and stored at −80 °C until purification. Supernatant containing scFv16 was harvested for immediate purification. E. coli BL21(DE3) cells were transformed with the pRH2.2 vector encoding Fab P2C2 and grown in LB medium with 1‰ ampicillin. Fab P2C2 expression was induced with 1 mM IPTG (Amresco) at OD600 = 0.6–0.8, followed by incubation at 20 °C for 20 h. Then, the cells were harvested and stored at −80 °C until purification. Purification of receptor/compound complex Thawed Sf9 cells expressing receptors were homogenized by dounce in hypotonic buffer (10 mM HEPES pH 7.5, 10 mM MgCl2, 20 mM KCl, 1x protein inhibitor (PI) containing 500 μM AEBSF (Goldbio), 1 μM E-64 (Selleck), 1 μM leupeptin (Sigma), and 0.15 μM aprotinin (Goldbio)), followed by centrifugation at 47,000×g and 4 °C for 15 min. The soluble and membrane-associated proteins were further eliminated by repeated high osmotic buffer (10 mM HEPES, pH 7.5, 10 mM MgCl2, 20 mM KCl, 1000 mM NaCl, 1x PI). The purified membrane was resuspended with resuspension buffer (100 mM HEPES pH 7.5, 150 mM NaCl, 10 mM MgCl2, 20 mM KCl, 2x PI). 20 μM psilocin or IHCH-2330 was added and incubated for 1 h at room temperature. The compounds-bound receptors were then extracted in solubilization buffer (2 mg/ml iodoacetamide (Sigma), 50 mM HEPES pH 7.5, 150 mM NaCl, 1% (w/v) lauryl maltose neopentylglycerol (LMNG, Anatrace), 0.2% (w/v) cholesterol hemisuccinic acid (CHS, Anatrace), 10 μM compounds and 1x PI) in 4 °C for 2 h. After solubilization, the supernatant containing extracted receptors was isolated by centrifugation at 90,000×g and 4 °C for 30 min and incubated overnight at 4 °C with 500 μl TALON IMAC resin (Clontech) in the presence of 20 mM imidazole and 800 mM NaCl to reduce non-specific adhesion. The resin was collected by centrifugation and washed with 15 columns of volume (cv) wash buffer I (50 mM HEPES (pH 7.5), 800 mM NaCl, 0.1% (w/v) LMNG, 0.02% (w/v) CHS, 20 mM imidazole, 10% glycerol, 10 μM psilocin or IHCH-2330) and following wash buffer II (50 mM HEPES (pH 7.5), 500 mM NaCl, 0.1% (w/v) LMNG, 0.02% (w/v) CHS, 10% glycerol, 10 μM psilocin or IHCH-2330). Proteins were eluted by 3 cv elution buffer (50 mM HEPES (pH 7.5), 500 mM NaCl, 0.04% (w/v) LMNG, 0.008% (w/v) CHS, 250 mM imidazole, 10% glycerol, 10 μM psilocin or IHCH-2330) and concentrated to 500 μl in the 100 kDa molecular weight threshold Vivaspin 20 concentrator (Sartorius Stedim) for desalting on a PD MiniTrap G-25 colum (GE Healthcare) with desalting buffer (25 mM HEPES (pH 7.5), 500 mM NaCl, 0.04% (w/v) LMNG, 0.008% (w/v) CHS, 10% glycerol, 10 μM psilocin or IHCH-2330). The desalted proteins were incubated with 5 IU His-tagged HRV-3C protease and 500 IU PNGase F (NEB) at 4 °C for 6 hours to remove N-terminal His tag, BRIL, and amino-linked glycosylation. Cleaved proteins were passed through equilibrated TALON IMAC resin (Clontech) and concentrated to 500 μl for further purification by size exclusion chromatography on a Superdex 200 10/300 GL column (GE Healthcare) with SEC buffer (20 mM HEPES (pH 7.5), 100 mM NaCl, 0.001% (w/v) LMNG, 0.0002% (w/v) CHS, 0.00025% (w/v) glyco-diosgenin (GDN, Anatrace), 100 μM TCEP (Thermo), 10 μM psilocin or IHCH-2330). Fractions containing the receptor/compound complex were collected for structural studies. Purification of miniGαq-γ2/Gβ1 heterotrimeric complex Thawed Sf9 Cells expressing miniGαq-γ2/Gβ1 protein complex were disrupted by dounce in lysis buffer containing 20 mM HEPES (pH 7.5), 100 mM NaCl, 1 mM MgCl2, 30 mM imidazole, 5 mM β-mercaptoethanol (Sigma), 0.2% (w/v) Triton X-100 (Beyotime), and 1x PI. Supernatant was collected by centrifugation at 117,500×g and 4 °C for 50 min and then incubated with 500 μl Ni NTA beads 6FF (Smart-Lifesciences) at 4 °C for 2 h. After washing with 10 cv wash buffer I (20 mM HEPES (pH 7.5), 100 mM NaCl, 5 mM β-mercaptoethanol, 30 mM imidazole, and 1x PI), the proteins were eluted by 3 cv elution buffer (20 mM HEPES (pH 7.5), 100 mM NaCl, 5 mM β-mercaptoethanol, 300 mM imidazole, and 1x PI). The eluate was concentrated to 500 μl using the 10 kDa molecular weight threshold Vivaspin 20 concentrator (Sartorius Stedim) and further purified by size exclusion chromatography on a Superdex 75 10/300 GL column (GE Healthcare) with SEC buffer (20 mM HEPES (pH 7.5), 100 mM NaCl, 0.001% LMNG, 0.0002% CHS, 0.00025% GDN, 100 μM TCEP). The target protein fractions were collected and concentrated to ~10 mg/ml for the following 5-HT2AR/miniGαq-γ2/Gβ1/scFv16 complex formation. Purification of scFv16 Supernatant containing scFv16 was added with 20 mM HEPES (pH 7.5), 100 mM NaCl, 1x PI and 500 μl Ni NTA beads 6FF (Smart-Lifesciences) and incubated at 4 °C overnight. Beads absorbing proteins were washed by 10 cv wash buffer I (20 mM HEPES (pH 7.5), 100 mM NaCl, 30 mM imidazole, and 1x PI) and 10 cv wash buffer II (20 mM HEPES (pH 7.5), 100 mM NaCl, and 1x PI). Absorbed proteins were eluted by 3 cv elution buffer (20 mM HEPES (pH 7.5), 100 mM NaCl, 300 mM imidazole, and 1x PI) and further purified with size exclusion chromatography on a Superdex 75 10/300 GL column. Collections were concentrated to ~5 mg/ml for 5-HT2AR/miniGαq-γ2/Gβ1/scFv16 complex formation. Purification of P2C2 Thawed E. coli cells were resuspended with lysis buffer containing 20 mM HEPES (pH 7.5), 100 mM NaCl, and 1x PI and disrupted by sonication. Supernatant containing proteins was collected by centrifugation at 47,000×g for 30 min at 4 °C and incubated with 500 μl Ni NTA beads 6FF at 4 °C for 2 h. Beads were washed with 10 cv wash buffer I (20 mM HEPES (pH 7.5), 20 mM imidazole, 100 mM NaCl, and 1x PI) and 10 cv wash buffer II (20 mM HEPES (pH 7.5), 100 mM NaCl, and 1x PI), followed by 3 cv elution buffer (20 mM HEPES (pH 7.5), 300 mM imidazole, 100 mM NaCl, and 1x PI). The eluted fractions were concentrated to ~5 mg/ml for 5-HT2BR/P2C2 complex formation. Formation of 5-HT2AR and 5-HT2BR complex for cryo-EM For assembly of 5-HT2AR complex, purified 5-HT2AR/psilocin or 5-HT2AR/IHCH-2330 complex was mixed with a 1.2 molar excess of miniGαq-γ2/Gβ1 and scFv16. For the 5-HT2BR complex, purified 5-HT2BR/IHCH-2330 complex was mixed with a 1.2 molar excess of Fab P2C2. The mixture was incubated at 4 °C overnight on a wheel and further purified by size exclusion chromatography on a Superdex 200 10/300 GL column (GE Healthcare) with SEC buffer (20 mM HEPES (pH 7.5), 100 mM NaCl, 0.001% (w/v) LMNG, 0.0002% (w/v) CHS, 0.00025% (w/v) GDN, 100 μM TCEP, 10 μM psilocin or IHCH-2330). The purified complex was concentrated to 5 mg/ml for cryo-EM. Sample preparation and Cryo-EM data collection About 3.5 μl of purified protein complex at corresponding concentration (∼5 mg/ml for 5-HT2AR complex and ∼15 mg/ml for 5-HT2BR complex, respectively) was applied onto a glow-discharged (40 s) EM grid (Quantifoil R1.2/1.3). The sample was left to stand for 15 s, following an immediate blotting for 3 s under 100% humidity at 8 °C using a Vitrobot Mark IV (Thermo Fisher Scientific), and plunged into liquid ethane. Prepared samples were stored in liquid nitrogen. Titan Krios electron microscope (Thermo Fisher Scientific) at 300 kV accelerating voltage was applied for Cryo-EM movies collection using a K3 Summit direct electron camera (Gatan) with a Quantum energy filter (Gatan) in super-resolution mode (Supplementary Table 1). The defocus ranged from −1.0 to −2.5 μm at a nominal magnification of ×105,000, with a pixel size of 0.832 Å. Total dose was 60 e− Å−2 with each movie consisting of 40 frames. 2 s of the exposure time with a dose rate of 20 e− pixel−1 s−1 was applied using SerialEM software (v4.2.4)43 (Supplementary Table 1). Cryo-EM data processing Processing of Cryo-EM data was performed using CryoSPARC (v.4.5.3)44. Beam-induced motion correction of movies was subjected to using patch motion correction45, and locally variable CTF landscapes were fitted using patch CTF estimation46. Particle picking was proceeded with blob picker and Topaz training. Contaminative particles were removed through multiple 2D classification rounds. The remaining particles were subjected to several rounds of ab initio reconstruction, and heterogeneous refinement served as three-dimensional (3D) classification, following a final homogenous refinement, non-uniform refinement, and local refinement with the best class for generating the final density map with an indicated global resolution. A soft mask was applied to the receptor of 5-HT2AR or 5-HT2BR, with the threshold set to 0.05, the dilation radius set to 5, and the soft padding width set to 17 in the Volume Tools task of CryoSPARC. Subsequently, the final particles underwent another round of local refinement with this mask, aiming to further enhance the map quality of the receptor region. Visualization was assisted by half-maps sharpened using DeepEMhancer. Subsequent model building and structural analysis were performed using sharpened maps generated by the cryoSPARC local refinement job. Model building and refinement The structure of 5-HT2AR-(R)-69-miniGq/i complex (PDB: 7RAN47) was chosen as an initial model for the constitution of 5-HT2AR complex and structure of 5-HT2BR-ERG-P2C2 (PDB: 5TUD48) and 5-HT2BR-LSD-P2C2 (PDB: 7SRQ41) for 5-HT2BR complex for fitting into the EM density map using ChimeraX v.1.749. The starting models underwent iterative rounds of manual adjustment, restructuration, and rebuilding in Coot v.0.9.8.9350 and refined by real-space refinement in Phenix v.1.20.151. Supplementary Table 1 concluded final refinement statistics. ChimeraX v.1.749 was applied for figure preparation. Molecular dynamics simulation 5-HT2AR/miniGαq/IHCH-2330 complex (3 independent simulations, 200 ns each) and 5-HT2BR/IHCH-2330 complex (3 independent simulations, 200 ns each) were performed with simulations. 5-HT2AR/miniGαq/IHCH-2330 complex consists of a 5-HT2AR, a miniGαq and a IHCH-2330 molecular. 5-HT2BR/IHCH-2330 complex consists of a 5-HT2BR and a IHCH-2330 molecular. The initial structures were based on the resolved cryo-EM constructions and prepared using Maestro (Schrödinger, version 14.0.132, LLC). For the preparation of the 5-HT2AR/miniGαq/IHCH-2330 complex, Gγ2, Gβ1, and scFv16 were removed from the initial structure. Unresolved residues and sidechains were modeled using the ‘homology’ tool in Maestro with the predicted 5-HT2AR/miniGαq/IHCH-2330 complex using Boltz-2 version 2.2.152. For the preparation of the 5-HT2BR/IHCH-2330 complex, P2C2 Fab was removed from the initial structure. ‘homology’ tool in Maestro was applied again for modeling unresolved residues and sidechains with the predicted 5-HT2AR/IHCH-2330 complex using Boltz-2 version 2.2.1. The mutation M1443.41W in the initial structure was reverted following the same protocol. Prepared proteins were performed in a rectangular periodic membrane system using CHARMM-GUI53. The disulfide bonds C148–C227, C349–C353 in 5-HT2AR and C128–C207, C350–C353 in 5-HT2BR were added. PH was set to 7.4, along with protonation of the basic amine in D1553.32 in 5-HT2AR and D1353.32 in 5-HT2BR, respectively. The upper membrane consists of 38 cholesterol, 70 POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) and 70 POPE (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine) in 5-HT2AR/miniGαq/IHCH-2330 complex, 25 cholesterol, 46 POPC and 46 POPE in 5-HT2BR/IHCH-2330 complex, with 30 cholesterol, 75 POPC and 75 of POPE for lower membrane in 5-HT2AR/miniGαq/IHCH-2330 complex, and 19 cholesterol, 48 POPC and 48 POPE in 5-HT2BR/IHCH-2330 complex, respectively. The membrane was in a 0.15 M NaCl solution, and the temperature was set to 310.15 K. Gromacs version 2025.3 was applied for all simulations. Energy minimization, equilibration, and production were performed using standard protocol from CHARMM-GUI. The trajectory was analyzed using GROMACS. The statistics were visualized with GraphPad Prism 10. Molecular docking Molecular docking was performed using Boltz-2 version 2.2.152. N-terminal-truncated human wild-type or mutated 5-HT2AR sequence (C60-V471) and 5-HT2BR sequence (T36-V481) were applied for the prediction. The same miniGαq sequence in the structure of 5-HT2AR-(R)-69-miniGq/i complex (PDB: 7RAN47) was added for prediction in all wild-type or mutated 5-HT2AR structures and 5-HT2BR structure with K1934.63F/F2145.35A/A2255.46S/V3667.39Y mutations. Default Boltz-2 settings were applied while the MSA server was set to true. Quantification and statistical analysis For compounds function tested by BRET, the BRET ratio responding to concentration and heatmap was plotted using GraphPad Prism 10 (GraphPad Software Inc., San Diego, CA). Data was normalized to % dopamine response in dopamine receptor’s function tests, % 5-HT response in wild-type serotonin receptors tests, or % wild-type 5-HT in mutation serotonin receptors tests, and analyzed by nonlinear regression “log(inhibitor) vs. response–variable slope (four parameters)” to gain EC50 and Emax. For competitive radioligand binding assays, data were analyzed using “one site-fit Ki” for calculation of affinity. Each test was performed with three biological replicates. Chemical synthesis Detailed methods, schemes, and conditions for the synthesis of all new compounds, related nuclear magnetic resonance analysis, and mass spectral data are provided in Supplementary Information. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Acknowledgements The psychedelics involved in this paper were provided by the drug reference materials laboratory of the Third Research Institute of the Ministry of Public Security (China), Shanghai Research Institute of Criminal Science and Technology (China), and Shanghai Yuansi Standard Science and Technology Co., Ltd (China). The cryo-EM data were collected at the Bio-Electron Microscopy Facility of ShanghaiTech University. We thank all staff at the facility for their assistance in cryo-EM data collection. We are grateful to Dr. Lingyun Yang at the NMR core facility for assistance in obtaining NMR data of all compounds. Author contributions L.T. performed pharmacological assays and contributed to cryo-EM structure. X.J. synthesized psychedelic analogs and contributed to cryo-EM structure. Y.W., D.C., and H.L. contributed to the cryo-EM structure. L.H. participated in project discussion and edited the manuscript. J.C. supervised ligand design, synthesis, and edited the manuscript. S.W. conceived and supervised the project and wrote the manuscript with input from all authors. Peer review Peer review information Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding S.W. discloses support for the research of this work from the National Natural Science Foundation of China [82225045, 82530106], Strategic Priority Research Program of the Chinese Academy of Sciences [XDB0990000], and the Ministry of Science and Technology of China [2025YFA1309000]. S.W. is a SANS Exploration Scholar. J.C. discloses support for publication of this work from the National Natural Science Foundation of China [82525061, 22377077], Science and Technology Commission of Shanghai Municipality [22XD1421900], and the Shanghai Municipal Government and ShanghaiTech University. L.T., X.J., Y.W., D.C., H.L., and L.H. declare no relevant funding. Data availability The accession number and code for cryo-EM density maps and corresponding atomic coordinates models (PDB/EMDB-ID) provided in this study are: 5-HT2AR/psilocin global 9WND/EMD-66114, 5-HT2AR/psilocin local 22WP/EMD-68735, 5-HT2AR/IHCH-2330 global 9WNF/EMD-66116, 5-HT2AR/IHCH-2330 local 22WR/EMD-68737 [[https://www.ebi.ac.uk/pdbe/entry/emdb/EMD-68737], 5-HT2BR/IHCH-2330 global 9WNG/EMD-66117 and 5-HT2BR/IHCH-2330 local 22WT/EMD-68739, respectively. The MD trajectories and predicted structures are accessible in Zenodo under https://doi.org/10.5281/zenodo.19765313. Source data are provided with this paper as a Source Data file. Competing interests S.W., J.C., L.T., X.J., and H.L. have filed a patent (CN202610216991X) through their institutions for the compounds reported in this study. The other authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Lingjie Tang, Xin Ji. Contributor Information Jianjun Cheng, Email: chengjj@shanghaitech.edu.cn. Sheng Wang, Email: wangsheng@sibcb.ac.cn. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77659-x. References Associated Data Supplementary Materials Supplementary Information Reporting Summary Transparent Peer Review file Data Availability Statement The accession number and code for cryo-EM density maps and corresponding atomic coordinates models (PDB/EMDB-ID) provided in this study are: 5-HT2AR/psilocin global 9WND/EMD-66114, 5-HT2AR/psilocin local 22WP/EMD-68735, 5-HT2AR/IHCH-2330 global 9WNF/EMD-66116, 5-HT2AR/IHCH-2330 local 22WR/EMD-68737 [[https://www.ebi.ac.uk/pdbe/entry/emdb/EMD-68737], 5-HT2BR/IHCH-2330 global 9WNG/EMD-66117 and 5-HT2BR/IHCH-2330 local 22WT/EMD-68739, respectively. The MD trajectories and predicted structures are accessible in Zenodo under https://doi.org/10.5281/zenodo.19765313. Source data are provided with this paper as a Source Data file.