Crystal Structure of an LSD-Bound Human Serotonin Receptor

This crystallography study analyzed the structure of LSD bound to a serotonin receptor and found that a branch of the receptor folds over the molecule while it is lodged into the binding pocket, and this lid-like structure secures LSD in place. This contributes to a slow dissociation rate of LSD, which forms the basis for its long-lasting effect. The authors suggest ways of introducing molecular mutations to selectively alter receptor signaling by increasing the mobility of this lid structure.

Abstract

“Introduction: The prototypical hallucinogen LSD acts via serotonin receptors, and …

Methods: … here we describe the crystal structure of LSD in complex with the human serotonin receptor 5-HT_2B.

Results: The complex reveals conformational rearrangements to accommodate LSD, providing a structural explanation for the conformational selectivity of LSD’s key diethylamide moiety. LSD dissociates exceptionally slow from both 5-HT_2BR and 5-HT_2AR—a major target for its psychoactivity. Molecular dynamics (MD) simulations suggest that LSD’s slow binding kinetics may be due to a “lid” formed by extracellular loop 2 (EL2) at the entrance to the binding pocket. A mutation predicted to increase the mobility of this lid greatly accelerates LSD’s binding kinetics and selectively dampens LSD-mediated β-arrestin2 recruitment.

Discussion: This study thus reveals an unexpected binding mode of LSD; illuminates key features of its kinetics, stereochemistry, and signaling; and provides a molecular explanation for LSD’s actions at human serotonin receptors.”

Authors: Daniel Wacker, Sheng Wang, John D. McCorvy, Robin M. Betz, A. J. Venkatakrishnan, Anat Levit, Katherine Lansu, Zachary L. Schools, Tao Che, David E. Nichols, Brian K. Shoichet, Ron O. Dror & Bryan L. Roth

Summary

INTRODUCTION

Lysergic acid diethylamide (LSD) is a prototypical human hallucinogen that alters human perception and mood. LSD is a semi-synthetic member of a larger class of ergolines that have long been recognized as therapeutics for many conditions, including migraine headaches, post-partum hemorrhage, and Parkinson’s disease.

LSD activates both G-protein and b-arrestin pathways at many GPCRs, but it also activates the non-canonical b-arrestin pathway at all biogenic amine GPCRs, including all but one serotonin receptor. This phenomenon has been termed ”functional selectivity” or ”biased agonism”.

We wanted to investigate the molecular mechanisms responsible for LSD’s activity at serotonin receptors, so we studied the structure of 5-HT2BR and used this structure to model LSD’s activity at 5-HT2AR.

Insights from 5-HT2BR/LSD Structure

We crystallized a 5-HT2BR construct bound to LSD and solved the X-ray structure to a resolution of 2.9 A. LSD is bound in the orthosteric binding site while also engaging the previously described extended binding site of the receptor. LSD binds to 5-HT2BR by a conserved salt bridge, and forms edge-to-face aromatic contacts with conserved phenylalanines in helix VI, and hydrogen bonds with the backbone of G2215.42 in helix V. Its diethylamide group binds in a crevice between helices II, III, and VII.

LSD’s Distinct Binding Pose

Although structurally and chemically related, ergolines have different in vivo activities, including the anti-migraine effects of ERG and the hallucinogenic actions of LSD. This is likely due to differences in blood-brain barrier permeability and in the pharmacology of ergolines at serotonin receptors.

ERG is located deeper in the pocket, with its ergoline moiety hydrogen bonding to helix III, whereas LSD is located higher in the pocket, closer to EL2 and the extracellular space.

When comparing the structures of the 5-HT2BR-ERG and 5-HT2BR-LSD complexes, we observe that several important orthosteric pocket residues change their rotamer states between the two structures. These changes likely reflect distinct ligand-receptor interactions and an unexpected plasticity of the receptor for these structurally related compounds.

LSD’s smaller amide substituent accounts for an overall contraction of the extended binding site relative to the ERG-bound structure, and the amide substituents of LSD and ERG are differentially arranged with respect to the ionic bond with D1353.32.

These observed rotamer changes, helical movements, and differential positioning of the ergoline moiety could reflect different receptor conformational and dynamic states, similar to the changes observed in the ligand-binding pocket of the b2 adrenergic receptor.

LSD Diethylamide Stereoselectivity and Function

We used sterically constrained LSD analogs to investigate the functional significance of different diethylamide conformations. We found that the SSAz conformation more closely resembled the diethylamide conformation observed in the 5-HT2BR bound LSD conformation, whereas RRAz was more similar to the diethylamide conformation observed in the small molecule LSD crystal structure.

To investigate the role of these ergoline substituents in LSD’s potency and activity at 5-HT2AR, we built a homology model of 5-HT2AR based on our 5-HT2BR/ LSD crystal structure and docked LSD, SSAz, RRAz, and LSA into the binding pockets of both the 5-HT2BR and 5-HT2AR models.

and Arrestin Translocation

We observed that the EL2 residues form a ”lid” over LSD in the 5-HT2BR structure, which contributes to its slow dissociation rate. We hypothesized that fluctuations in the position of the lid are necessary for LSD to exit or enter the binding pocket.

LSD’s long residence time on the 5-HT2BR receptor can modulate kinetically sensitive patterns of intracellular signaling. The L209AEL2 mutation strongly reduces LSD’s b-arrestin2 recruitment potency and efficacy.

We wondered whether this model of LSD’s actions at a molecular level held true for the 5-HT2AR, which represents LSD’s principal molecular target in vivo. The results show that LSD has a slower off rate and longer residence time at 5-HT2AR compared to 5-HT2BR.

To investigate the hypothesis that LSD’s slow binding kinetics are important for its signaling, we modified a bioluminescence resonance energy transfer (BRET) assay to measure b-arrestin2 recruitment at 5-HT2AR and 5-HT2BR. We found that LSD’s potency is weak at both mutants.

We plotted the transduction coefficients of Gq (IP accumulation) and arrestin (b-arrestin2 recruitment) for LSD’s actions at 5-HT2AR and 5-HT2BR. The L209AEL2 and L229AEL2 mutations abrogated the time-dependency for b-arrestin2 translocation.

DISCUSSION

LSD, one of the most prominent psychoactive drugs, has a constrained conformation in its binding site that is crucial for its activity, kinetics, and signaling. This conformation contributes to LSD’s relatively potent ability to promote b-arrestin translocation.

The crystal structure of the 5-HT2BR/LSD complex reveals that the amide substituents largely determine the positioning of the ergoline system within the orthosteric pocket, and that the diethylamide of LSD must adopt a specific conformation for activity.

LSD’s long residence time at 5-HT2BR and 5-HT2AR may be due to a lid formed by EL2 covering the binding pocket. A substitution to a key residue identified structurally attenuates b-arrestin2 recruitment while minimally affecting Gq signaling.

Crystal structures and molecular simulations can never fully explain CNS drug efficacy, but our observations provide the first structure-informed insights into the molecular actions for any hallucinogen. These insights may also template future structure-based efforts to discover new chemotypes at 5-HT2A and 5-HT2B receptors.

STAR+METHODS

5-HT2BR receptor crystallization construct, expression and purification, lipidic cubic phase crystallization, data collection, structure solution and refinement, MD simulations, homology modeling, and molecular docking of LSD and its derivatives.

METHOD DETAILS

The 5-HT2BR receptor crystallization construct was generated based on a previously engineered receptor construct that was edited by Quickchange PCR. The final construct contains a thermostabilizing M144W3.41 mutation and a haemagglutinin signal sequence followed by a FLAG tag at the N terminus.

Expression and purification of 5-HT2BR

High-titer recombinant baculovirus was generated using the Bac-to-Bac Baculovirus Expression System (Invitrogen). The virus was induced in Spodoptera frugiperda cells and the virus titer was determined by flow-cytometric analysis of cells stained with gp64-PE antibody. 5-HT2BR was expressed in Sf9 cells by infection with P1 or P2 virus at a MOI of 3-5. Membranes were purified by repeated centrifugation in a high osmolarity buffer containing 1.0 M NaCl, 10 mM MgCl2, 20 mM KCl and protease inhibitors.

After purification, membranes were resuspended in buffer containing 10 mM HEPES, pH 7.5, 10 mM MgCl2, 20 mM KCl, 150 mM NaCl, 50 mM LSD, and protease inhibitors. Proteins were purified using TALON IMAC resin (Clontech) and then concentrated using a Vivaspin 20 centrifuge concentrator. The C-terminal 10 3 His-tag was removed by addition of His-tagged PreScission protease (GenScript) and incubation overnight at 4 C.

Lipidic cubic phase crystallization

5-HT2BR/LSD complexes were reconstituted into lipidic cubic phase (LCP) using the twin-syringe method and crystallized on 96-well glass sandwich plates in 100 mM Tris/HCl pH 7.5-8.0, 90-130 mM potassium phosphate monobasic, 28% – 30% PEG400.

LSD synthesis

LSD was synthesized by the method of Johnson et al. (Johnson et al., 1973) as follows: 315 mg d-lysergic acid monohydrate was stirred under N2 and heated to reflux on a 90 C oil bath, then diethylamine and POCl3 were added simultaneously from separate dropping funnels over about 2 min. The crude product was purified by centrifugal thin layer chromatography using a 2 mm silica plate, under a N2/NH3 atmosphere, and eluting with 100% CH2Cl2. The LSD tartrate was made by dissolving the crude base in reagent MeOH and adding 0.5 equivalent of L-(+)-tartaric acid. The tartrate salt crystallized as fine needles and was collected by suction filtration.

Calcium flux assay

Cells were grown in a Flp-In 293 T-Rex Tetracycline inducible system and were stimulated with 5-HT2BR and 5-HT2AR constructs using a Fluo-4 Direct dye. The cells were then placed in a FLIPRTETRA fluorescence imaging plate reader. The FLIPRTETRA was programmed to read baseline fluorescence for 10 s, then add 10 ml of drug/well and read for 5 min. The maximum-fold increase was determined and plotted as a function of drug concentration.

Tango arrestin recruitment assay

The 5-HT2BR and 5-HT2AR Tango constructs were designed and assays were performed as previously described. HTLA cells were grown in DMEM containing 10% dialyzed FBS and transfected with 15 mg per 15-cm of either 5-HT2BR or 5-HT2AR Tango construct using the calcium phosphate transfection method. The next day, media and transfection reagents were removed from cells, and they were plated onto poly-L-lysine-coated 384-well white clear bottom cell culture plates at a density of 10,000 cells/well in a total of 40 ml. Drug solutions were added to cells for overnight incubation.

A 1:20 dilution of the drug was added per well, and the results were plotted as a function of drug concentration.

Phosphoinositide hydrolysis assay

PI hydrolysis assays were performed using the scintillation proximity assay. Cells were seeded into 96-well poly-lysine coated plates at a density of 40-50,000 cells/well in 100 mL inositol-free DMEM containing 1% dialyzed FBS and incubated overnight for 16-18 hr at 37 C and 5% CO2. After overnight incubation at 4 C, lysates were added to 96-well flexible, clear microplates containing 75 mL of 0.2 mg/well RNA binding yttrium silicate beads, and incubated for 1 hr on a shaker.

Bioluminescence resonance energy transfer arrestin assay

HEK293T cells were co-transfected with human 5-HT2BR containing C-terminal Renilla luciferase (RLuc8), GRK2, and Venus-tagged N-terminal b-arrestin2 and plated in poly-lysine coated 96-well white clear bottom cell culture plates. For kinetic experiments, 30 mL of drug (3X) was added per well, incubated for designated time points, and 10 mL of RLuc substrate (5 mM final concentration) was added per well. The ratio of eYFP/RLuc was calculated per well, and the net BRET ratio was plotted.

Ligand association and dissociation radioligand binding assays were performed utilizing the same concentrations of radioligand, membrane preparations, and binding buffer. The binding of unlabeled ergotamine to 5-HT2BR and 5-HT2AR was determined using a 96-well Filtermate harvester. The [3H]-LSD kon and koff rates of 5-HT2BR and 5-HT2BR L209AEL2 were used to estimate the kon and koff rates of ergotamine using Graphpad Prism 5.0.

MD simulations set-up

The 5-HT2BR receptor was simulated in five distinct conditions, including the LSD-bound crystal structure described in this manuscript, the same structure with the ligand removed, the same structure with the L209AEL2 mutation, and the same structure with the ligand removed.

In the simulations reported in this paper, all aspartate residues were deprotonated. The LSD tertiary amine nitrogen was protonated in the liganded simulations, enabling formation of the conserved salt bridge with neighboring D1353.32.

The protein structures were aligned on the transmembrane helices of the OPM structure of PDB 4NC3, and internal waters added with Dowser. They were then inserted into a pre-equilibrated POPC bilayer and neutralized with NaCl.

MD simulation force field parameters

We used the CHARMM36 parameter set for protein molecules, the CHARMM TIP3P model for water, and the CGenFF for LSD. Full parameter sets are available upon request.

MD simulation protocol

Simulations were performed on GPUs using the CUDA version of PMEMD in Amber15 (Case et al., 2015). The systems were minimized, then equilibrated by heating from 0 to 100K in the NVT ensemble over 12.5 ps, and then to 310K over 125 ps in the NPT ensemble.

Simulations were performed with periodic boundary conditions, a time step of 4.0 fs, and hydrogen mass repartitioning. Bond lengths to hydrogen atoms were constrained using SHAKE.

Root mean square fluctuation (RMSF) values were calculated for each Ca atom in a transmembrane helix using VMD’s Python scripting functionality. The RMSF values were then plotted against the average structure for each simulation condition.

Homology modeling of 5-HT2AR

Sequence alignment was generated with PROMALS3D using human 5-HT2AR, 5-HT2BR, and 5-HT2BR/LSD complex sequences. The alignment was manually edited to remove the engineered apocytochrome b562 RIL (BRIL) from the template structure. 1000 homology models were built using MODELER-9v15, and then evaluated for their ability to enrich known 5-HT2AR ligands over property-matched decoys through docking to the orthosteric binding site, using DOCK 3.7. The selected best scoring model was further optimized through minimization with the AMBER protein force field.

Molecular docking of LSD and its derivatives

LSD, its derivatives (SSAz, RRAz, and LSA) and lysergamide were docked to the orthosteric binding pocket of the 5-HT2AR homology model and the 5-HT2BR crystal structure using DOCK3.7. The docked ligand poses were scored by summing the receptor-ligand electrostatics and van der Waals interaction energies.

QUANTIFICATION AND STATISTICAL ANALYSIS

Normalized dose-response data was fitted using the Black and Leff operational model in Graphpad Prism 5.0, and transduction coefficients were calculated using the Black and Leff operational model in Graphpad Prism 5.0. Bias factors were calculated using the method by Kenakin et al. (2012).

Data Resources

HKL2000 was used to process the raw diffraction data and phenix and ccp4 software suites were used to determine, refine, and build the structural model.