Molecular recognition of itch-associated neuropeptides by bombesin receptors

Dear Editor, Itch, especially persistent itch (or chronic pruritus) in patients with allergic skin diseases and/or neuropathic problems, significantly affects sleep, mood and individual health. Extensive efforts have been made to develop novel therapeutic strategies to fight itch. Unfortunately, there is no FDA-approved treatment for chronic pruritus despite great efforts in antipruritic research in the past few decades. Bombesin (Bn) is an amphibian tetradecapeptide found in the skin of European Bombina. Two mammalian Bn-like peptides have been characterized, neuromedin B (NMB) and gastrin-releasing peptide (GRP), which are the endogenous ligands for neuromedin B receptor (NMBR or BB1 receptor) and gastrin-releasing peptide receptor (GRPR or BB2 receptor), respectively. 2 In addition, there is an orphan bombesin receptor, BRS3 or BB3R, which shares sequence and structural similarity with NMBR and GRPR but is not activated by NMB or GRP (Supplementary information, Fig. S1). Both NMB and GRP are involved in a variety of physiological and pathological processes, including itch induced by histaminergic or nonhistaminergic pruritogens (Fig. 1a). As itch-associated neuropeptide receptors in the spinal cord, both NMBR and GRPR are at the core of itch transmission and play a pivotal role in itch biology; accordingly, they are attractive targets for antipruritic intervention. To explore the mechanism of peptide recognition by itch receptors, we assembled agonist–bombesin receptor–Gq complexes for cryo-electron microscopy (cryo-EM) studies (Supplementary information, Fig. S2) using approaches of BRIL fusion, NanoBiT tethering, Gq engineering, and antibody scFv16, which had previously been used successfully in solving several GPCR/Gprotein complexes. The cryo-EM structures of NMBR and GRPR, both complexed with the engineered Gq and selective peptide agonists NMB30 and GRP(14–27), were solved at global resolutions of 3.15 Å and 3.30 Å, respectively (Fig. 1b–g; Supplementary information, Figs. S3–S5 and Table S1). In both structures, unambiguous electron densities were observed for the C-terminal 9 residues of both bombesin agonists, which are close to the minimal 10-residue peptide required for full potency. Compared to the shallow, solvent-exposed pockets in the other class of itch receptors, MRGPRX2/4, both bombesin receptors exhibit classically deep pockets that are commonly found in neuropeptide receptors (Supplementary information, Fig. S6). Each bombesin agonist has a dumbbell-shaped overall conformation and adopts a binding pose perpendicular to the membrane plane (Fig. 1d, g), with its C-terminus inserted deep into the transmembrane domain (TMD) bundle and its N-terminus (referring throughout the paper to the resolved N-terminal amino acids) pointing to the extracellular surface, partially explaining the conserved activation mechanism between these two bombesin receptors. The ligand-recognition region of both receptors can be divided into two major parts: (1) the extracellular loops around the N-terminal dumbbell end and (2) the bottom of the TMD pocket burying the C-terminal dumbbell end (Fig. 1h–k). All positions of bombesin peptides are numbered from the amino terminus of NMB or NMC for clarity (Supplementary information, Table S2). In the NMBR structure, the N-terminal dumbbell end comprises NLWAT (2–6) (the superscript N refers to NMB30) residues, forming extensive hydrophobic interactions with residues in the binding pocket (Fig. 1i). In addition, R103 forms a close hydrogen bond with T6 and plays a vital role in NMBR activation, which is consistent with our mutagenesis analysis showing that R103A decreased the potency of NMB30 to activate NMBR (Supplementary information, Fig. S7a and Table S3). At the bottom of the binding pocket is the C-terminal dumbbell end formed by the HFM motif (Fig. 1i). H8 is well sandwiched by P120 and P200, which is further stabilized by the conserved disulfide bond between C116 and C198. H8 also forms hydrogen bonds with E178, S182, C198 and I199 (Fig. 1i). Alanine mutations of P120, E178, and P200 had a great influence on receptor activation, especially the P200A and P200S mutations, suggesting the key role of ECL2 in peptide recognition (Supplementary information, Fig. S7b, c and Table S3). F9 forms a stabilizing π–π packing with H283, while F9 and M10 fit into a hydrophobic crevice formed by the residues C96, V127, W279, H283, and F314 (Fig. 1i). The hydrophobic network here is essential for NMBR activation, since replacing V127 with alanine entirely abolished the activity of NMB30 (Supplementary information, Fig. S7c and Table S3). The C-terminal amide forms a hydrogen bond with D100, and the C-terminal carbonyl group forms an additional hydrogen bond with R310 (Fig. 1i). Alanine replacements of D100 and R310 displayed a 10to 50-fold reduction in NMB30 potency (Supplementary information, Fig. S7a, e and Table S3). It is noteworthy that the remarkable conformational shift of the C-terminal amidated end plays a pivotal role in NMB30 potency, since our mutation of Q123 to alanine was associated with only a ~30-fold reduction in NMB30 potency, while the mutation of Q123 to arginine with a bulky side chain to clash with the C-terminal amidated end completely abolished the ability of NMB30 to stimulate NMBR (Supplementary information, Fig. S7c and Table S3). Together, these detailed structural analyses provide important information to better understand the mechanism of NMB30 recognition by NMBR. In the GRPR structure, the N-terminus adopts a more extended conformation. The side chains of R100 and R287 form three hydrogen bonds with the main-chain carbonyl groups of W4, A5, and V6 and drag the N-terminus of GRP(14–27) towards ECL1 and ECL2, where H3 forms a hydrogen bond with the backbone carbonyl of A103 and an additional hydrogen bond with the side chain of S195 (Fig. 1k). In addition, W4 forms

The supernatant was collected by centrifugation at 65,000 × g for 40 min and then incubated with dextrin resin (Dextrin Beads 6FF, Smart Life Sciences) at 4˚C for 4 h.

Cryo-EM data collection
Cryo-EM grids were prepared with the Vitrobot Mark IV plunger (FEI) set to 4 ˚C and 100% humidity. Three-microliter of the NMB30-NMBR-Gq complex was applied to the glow discharged Au R1.2/1.3 holey carbon grids. The sample was incubated for 5 s on the grids before blotting for 3 s (double-sided, blot force 2) and flash-frozen in liquid ethane immediately. The same condition was used for the sample GRP (14-27)-GRPR-Gq complex.
For NMB30-NMBR-Gq complex dataset, 4,858 movies were collected on a Titan Krios equipped with a Gatan K3 direct electron detection device at 300 kV with a magnification of 81,000, corresponding to a pixel size 1.04 Å. Image acquisition was performed with EPU Software (FEI Eindhoven, Netherlands). We collected a total of 36 frames accumulating to a total dose of 50 e -Å -2 over 2.5 s exposure.
For GRP (14-27)-GRPR-Gq complex dataset, 9,002 movies were collected on a Titan Krios equipped with a Falcon4 direct electron detection device at 300 kV with a magnification of 96,000, corresponding to a pixel size 0.8 Å. Image acquisition was performed with EPU Software (FEI Eindhoven, Netherlands). We collected a total dose of 50 e -Å -2 over 2.5 s exposure on each EER format movie 8 . Each movie was divided into 36 frames during motion correction.

Cryo-EM image processing
MotionCor2 was used to perform the frame-based motion-correction algorithm to generate drift-corrected micrograph for further processing and CTFFIND4 provided the estimation of the contrast transfer function (CTF) parameters 9,10 .
For NMB30-NMBR-Gq complex dataset, 480 aligned micrographs were deleted because of contaminations or bad ice quality. After selection, approximately 800 particles were manually picked and two-dimensional classes were calculated and used as references for automatic picking. All subsequent steps of particle picking, extraction, classification and post processing of refined models were performed with Relion3.0 11 . A total of 3,034,736 particles were extracted from the cryo-EM micrographs and followed by reference-free 2D classification, yielding 619,210 particles after clearance. Mask three-dimensional (3D) classification on the receptor part was used to separate out 355,509 particles that resulted to a clearer density of NMBR. We refined this portion of particles, which led to a structure at 3.52 Å global resolution. After CTF refinement, Bayesian polishing, and postprocessing with DeepEMhancer 12 , then the particles were reconstituted to a 3.15 Å structure (Supplementary information, Fig. S3).
For GRP (14-27)-GRPR-Gq complex dataset, 762 aligned micrographs were deleted because of contaminations or bad ice quality. After selection, NMBR was used as 3D reference for automatic picking. All subsequent steps of particle picking, extraction, classification and post processing of refined models were performed with Relion3.0 11 . A total of 3,365,839 particles were extracted from the cryo-EM micrographs and followed by reference-free 2D classification, yielding 577,108 particles after clearance. Mask 3D classification on the receptor part was used to separate out 301,192 particles that resulted to a clearer density of GRPR. The second round of 3D classification was performed without mask and separated out 55,286 particles. We refined the remained particles, which led to a structure at 3.72 Å global resolution.
After the postprocessing with DeepEMhancer 12 , then the particles were reconstituted to a 3.3 Å structure (Supplementary information, Fig. S4).

Model building
NMBR and GRPR structures predicted from Alphafold2 were used as the starting reference models for receptors building 13 . Structures of Gαq, Gβ, Gγ and the scFv16 were derived from PDB entry 7WKD (unpublished) were rigid body fit into the density.
All models were fitted into the EM density map using UCSF Chimera 14 followed by iterative rounds of manual adjustment and automated rebuilding in COOT 15 and PHENIX 16 , respectively. The model was finalized by rebuilding in ISOLDE 17 followed by refinement in PHENIX with torsion-angle restraints to the input model. The final model statistics were validated using Comprehensive validation (cryo-EM) in PHENIX 16 and provided in the supplementary information (Supplementary information, Table S1). All structural figures were prepared using Chimera 14 , Chimera X 18 , and PyMOL (Schrödinger, LLC.).

Function essay
AD293 cells (Agilent) were cultured in DMEM/high Glucose medium (GE healthcare) supplemented with 10% (v/v) fetal bovine serum (FBS, Gemini) and 1% penicillin/streptomycin and maintained at 37°C in 5% CO2 incubator. Inositol phosphate 1 (IP1) production was measured using the IP-One HTRF kit (Cisbio, 621PAPEJ) 19 . Briefly, cells were seeded onto 12-well cell culture plates for 16 h before transfection. The cells were then transiently with different NMBR or GRPR constructs using FuGENE HD transfection reagent. After 24 h, cells were harvested and resuspended in IP1 stimulation buffer at a density of 7 × 10 5 cells/mL. Cells were then plated onto 384-well assay plates at 4900 cells/7 μL/well. Another 7 μL IP1 Stimulation Buffer 2 containing ligand was added to the cells, and the incubation lasted for 1 h at 37 °C. Intracellular IP1 measurement was carried with the IP-One HTRF kit and EnVision multiplate reader according to the manufacturer's instructions.
The HTRF ratio was converted to a response (%) using the following formula: response (%) = ratio of sample/WT×100. Data presented are mean±S.E.M. of at least three biologically independent experiments.

Cell-surface expression assay
Cell-surface expression for each NMBR and GRPR mutant was monitored by a fluorescence-activated cell sorting (FACS) assay. The mutants were cloned into pcDNA6.0 vector (Invitrogen) with a N-terminal FLAG tag. The cell seeding and transfection follow the same method as function assay. After 24h of transfection, cells were washed once with PBS and digested with 0.2% (w/v) EDTA in PBS. Thereafter, the expressed cells were incubated with Monoclonal anti-FLAG M2-FITC (Sigma-Aldrich) at a dilution of 1:100 for 15 min at 4 °C, and then a 9-fold excess of PBS was added to cells. After cells were resuspended, fluorescence intensity was quantified in a BD Accuri C6 flow cytometer system (BD Biosciences) at excitation 488 nm and emission 519 nm. The FACS data were analyzed by BD Accuri C6 software 1.0.264.21 and data were normalized to WT.