G protein-coupled receptors (GPCRs): advances in structures, mechanisms and drug discovery

3308 sigtrans Signal Transduction and Targeted Therapy Signal Transduct Target Ther Nature Publishing Group PMC11004190 11004190 11004190 38594257 10.1038/s41392-024-01803-6 G protein-coupled receptors (GPCRs): advances in structures, mechanisms and drug discovery Zhang Mingyang 1 2 # Chen Ting 3 # Lu Xun 2 # Lan Xiaobing 1 Chen Ziqiang 4 ✉ Lu Shaoyong 1 2 ✉ 1 Key Laboratory of Protection, Development and Utilization of Medicinal Resources in Liupanshan Area, Ministry of Education, Peptide & Protein Drug Research Center, School of Pharmacy, Ningxia Medical University, Yinchuan, 750004 China 2 Medicinal Chemistry and Bioinformatics Center, Shanghai Jiao Tong University School of Medicine, Shanghai, 200025 China 3 Department of Cardiology, Changzheng Hospital, Affiliated to Naval Medical University, Shanghai, 200003 China 4 Department of Orthopedics, Changhai Hospital, Affiliated to Naval Medical University, Shanghai, 200433 China ✉ Corresponding author. # Contributed equally. 10 4 2024 9 88 88 10 4 2024 © The Author(s) 2024 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . Abstract G protein-coupled receptors (GPCRs), the largest family of human membrane proteins and an important class of drug targets, play a role in maintaining numerous physiological processes. Agonist or antagonist, orthosteric effects or allosteric effects, and biased signaling or balanced signaling, characterize the complexity of GPCR dynamic features. In this study, we first review the structural advancements, activation mechanisms, and functional diversity of GPCRs. We then focus on GPCR drug discovery by revealing the detailed drug-target interactions and the underlying mechanisms of orthosteric drugs approved by the US Food and Drug Administration in the past five years. Particularly, an up-to-date analysis is performed on available GPCR structures complexed with synthetic small-molecule allosteric modulators to elucidate key receptor-ligand interactions and allosteric mechanisms. Finally, we highlight how the widespread GPCR-druggable allosteric sites can guide structure- or mechanism-based drug design and propose prospects of designing bitopic ligands for the future therapeutic potential of targeting this receptor family. Subject terms: Target identification, Target identification status released display-pdf yes is-olf no is-manuscript no is-preprint no is-journal-matter no is-scanned no is-retracted no Received 2023 Aug 15; Revised 2024 Feb 19; Accepted 2024 Mar 13; Collection date 2024. Introduction G protein-coupled receptors (GPCRs) are the largest superfamily of cell surface membrane receptors and are encoded by approximately 1000 genes, sharing conserved seven-transmembrane (7TM) helices connected by three intra- and three extra-cellular loops. 1 – 3 GPCRs are conformationally dynamic proteins that mediate vital biological functions of signal transduction triggered by various extracellular signals such as photons, ions, lipids, neurotransmitters, hormones, peptides, and odorants. 4 – 8 Due to the distinct topography between the binding sites of extracellular stimuli and the subsequent signaling events at the intracellular site (approximately 40 Å), GPCR signal transduction is allosteric. 9 – 13 Advances in protein engineering, X-ray crystallography, and cryo-electron microscopy (cryo-EM), coupled with innovative technologies such as X-ray free electron lasers (XFELs) and nuclear magnetic resonance (NMR) spectroscopy, have revolutionized our understanding of GPCR structures and dynamics. These studies provide insights into ligand-receptor interactions, conformational changes, and signaling complexes, offering unprecedented opportunities for in-depth investigations into receptor activation, orthosteric/allosteric modulation, biased signaling, and dimerization. Once activated by exogenous stimuli, GPCRs primarily employ heterotrimeric G-proteins and arrestins as transducers to produce second messengers and further initiate the downstream signaling, resulting in promiscuous signaling profiles within cells. 11 Such spectrum of signaling is the prerequisite for function diversity of GPCRs and is fundamental in regulating physiological processes, including sensory perception, neurotransmission, and endocrine processes. 14 , 15 The mutations and truncation of GPCRs; however, can dysregulate GPCR functionality by altering constitutive activity, influencing membrane expression and affecting post-translational behaviors. 16 Unraveling the mechanisms of stimuli-GPCR-effector coupling, as well as the concise regulation of GPCR dysfunction will bring about valuable therapeutic potentials and inspire the design of modulators with high potency, selectivity, or biased signaling. Till date, approximately 34% of the US Food and Drug Administration (FDA)-approved drugs are targeted to GPCRs, with modulators in clinical trials or preclinical stages experiencing exponential growth. 17 , 18 Among them, orthosteric ligands impose an effective alteration on GPCR activity and signaling process by competitively preventing the binding of endogenous ligands. 19 However, due to the sequence conservation of orthosteric sites, in most cases, subtype selectivity remains an intractable issue, which implies the inevitable side effects of orthosteric drugs. 20 As an alternative or complementary option, targeting allosteric sites alone or targeting both orthosteric and allosteric sites can overcome these major hurdles. 21 – 25 Allosteric modulators are highlighted for their high subtype selectivity and low side effects. A progressive structural understanding of the detailed receptor-ligand interactions is paving the way for fragment-to-lead optimization in structure-based drug design (SBDD) (Fig. 1 ). Moreover, the knowledge of allosteric sites is useful for the design of bitopic ligands by creating a molecule attached to both an allosteric and orthosteric site. Bitopic ligands have several advantages of improved affinity and enhanced selectivity over a single allosteric or orthosteric ligand. In addition, elucidating allosteric mechanisms of GPCRs provides a viable strategy to develop biased ligands such as G protein- or β-arrestin-based allosteric modulators. 26 Bitopic modulators have higher selectivity to reduce side effects since they exert pathway-specific effects on GPCR signaling. 26 , 27 Fig. 1 Phylogenetic tree of GPCRs indicating GPCR structures that have been solved in complex with modulators. Nodes represent GPCRs named according to their UniProt gene name and are organized according to the GPCR database. GPCR structures bound to modulators are highlighted by color In this review, we first summarize the structural progression, activation mechanisms, and functional diversity of GPCRs. To delve into the advancement of GPCR drug discovery, we investigate the detailed drug-target interactions at the orthosteric sites, focusing on GPCR structures in complex with recent FDA-approved orthosteric drugs. Subsequently, allosteric modulators are extensively discussed, with a focus on recent breakthroughs in GPCR structures that bind to synthetic small molecules. Notably, peptides and antibodies are excluded from our analysis. Such investigation systematically clusters the location of allosteric sites in the extracellular vestibule, transmembrane domain, and intracellular surface, highlighting the key binding modes with their target receptors and allosteric mechanisms. This review aims to provide a deeper understanding of GPCR structures, mechanisms, and drug discovery, which has important implications for structure- or mechanism-based drug design and the design of bitopic ligands for the future therapeutic potential of targeting this receptor family. Structure advances in GPCRS The low expression of membrane protein GPCRs, combined with their conformational flexibility, initially posed great challenges for high-resolution diffraction. 28 The initial crystal structures of rhodopsin and the ligand-activated β2 adrenergic receptor (β2AR) were resolved in 2000 and 2007, respectively. 29 , 30 Over the past two decades, considerable progress has been made in the engineering of proteins and the technique of X-ray crystallography. 31 Notably, the use of GPCR engineering with fusion proteins, 32 , 33 antibody fragment crystallization 34 , 35 , and thermostabilizing mutations 36 , has produced numerous antagonist- or agonist-bound GPCR structures. However, only agonist-bound GPCRs frequently exist in an intermediate conformation because the fully active conformation requires stabilizing chaperones, including G proteins, G protein mimetics, conformationally specific nanobodies, and mini-G proteins. 37 The first GPCR-G protein complex was determined in 2011 using X-ray diffraction; 38 however, the demanding nature of X-ray crystallography has rendered GPCR-G protein complex crystallization a difficult undertaking. Cryo-EM has developed to be an alternative technique, driving a novel trend in GPCR structural biology. Unlike X-ray crystallography, cryo-EM does not rely on crystals and has considerably superior potential to directly visualize detergent- or nanodisc-solubilized GPCRs. This capability enables the determination of previously intractable fully active states and larger protein complexes, including GPCR-G protein complexes. 39 Since then, the number of cryo-EM structures depicting GPCRs in complex with intracellular partners has experienced exponential growth (Fig. 2 ). As of November 2023, the Protein Data Bank has accumulated 554 complex structures, of which 523 are resolved using cryo-EM. 40 However, both crystallography and cryo-EM are limited to capturing the most stable and lowest energy conformations under crystallization conditions. 4 Moreover, the comprehensive characterization of intermediate states and transition kinetics remains elusive. Crystallographic, spectroscopic, and simulation techniques have offered complementary information on the conformational dynamics of GPCRs. Fig. 2 Timeline of major advancements in GPCR structure study using X-ray crystallography and cryo-EM The advanced XFELs possess the potential to solve the missing information. The exceptional properties of XFELs, characterized by extreme brilliance and femtosecond short pulses, allow them to overcome radiation damage, facilitating the determination of GPCR structures with atomic-level information at femtosecond timescales. 41 NMR spectroscopy offers a valuable technique to detect dynamic features of GPCRs in liquid environments. 42 , 43 The number, position, and shape of signals in the NMR spectra are sensitive to changes in the micro-environment of stable-isotope “probes” incorporated into receptors. Double electron-electron resonance (DEER) spectroscopy enables the assessment of a distance distribution between two different probes. Fluorescence resonance energy transfer (FRET), a technique based on fluorescence, functions as an “atomic ruler” to detect the proximity between two labels, providing valuable data about the number of states and their relative populations. 44 , 45 Among these, DEER and FRET provide only localized details regarding the chemical probes that have been inserted. In addition, molecular dynamics (MD) simulations offer a comprehensive, time-resolved view of complete protein structures, capturing intermediate states along the transition pathway. 46 – 48 Advances in the structural biology of GPCRs have revealed key information on ligand-receptor interactions, conformational changes, and signaling complexes, opening the opportunity for exploration of receptor activation, orthosteric/allosteric modulation, biased signaling, and dimerization. Mechanism of GPCR activation and signaling Although the nature of GPCRs and activating stimuli may vary significantly, GPCRs primarily coordinate distinct downstream signaling responses through two types of transducers: heterotrimeric G proteins and arrestins. Human G proteins comprise four major families (G s , G i/o , G q/11, and G 12/13 ) and more than half of GPCRs activate two or more G proteins, each of which exhibits distinct efficacies and kinetics. 49 , 50 The promiscuous coupling leads to fingerprint-like signaling profiles inside the cell, which contributes to the complexity of GPCR signaling. When bound to GDP, the Gαβγ heterotrimer is inactive. Agonist binding leads to the formation of an active conformation of GPCRs, which initiates signaling cascades involving the recruitment and activation of G-proteins. The activated GPCR catalyzes the GDP/GTP exchange on the Gα subunit, causing the dissociation of Gα from the Gβγ dimer. Due to high cellular concentrations of GTP, Gα rapidly binds a molecule of GTP at the nucleotide-binding site. Both Gα-GTP and Gβγ can modulate subsequent effector proteins. Gα-GTP can activate or inhibit enzymes such as adenylyl cyclase (AC), phospholipase C (PLC), or ion channels, depending on the specific G protein type. Gβγ can also modulate various signaling pathways and interact with target proteins. Activation of effector proteins by Gα-GTP or Gβγ generates second messengers, such as cyclic AMP (cAMP). The cellular response concludes with the Gα subunit hydrolyzing GTP to GDP, leading to its reassociation with Gβγ and G protein inactivation. Subsequently, the Gα subunit completes the G-protein activation cycle by reassociating with Gβγ. To prevent sustained signaling, activated GPCRs may also undergo C-terminal phosphorylation facilitated by G-protein-coupled receptor kinases (GRKs). This multi-site GPCR phosphorylation determines β-arrestin binding affinity and induces receptor desensitization via steric hindrance, followed by clathrin-mediated endocytosis and ubiquitination of the receptor (Fig. 3a ). 11 , 51 , 52 The receptor-arrestin complex also serves as a scaffold for over 20 different kinases, including mitogen-activated protein (MAP) kinases, ERK1/2, p38 kinases, and c-Jun N-terminal kinases, activating G-protein independent signaling pathway. Four isoforms of arrestin (arrestins 1-4) and multiple GRK isoforms were discovered, with arrestins 1 and 4 being only found in the visual system. β-arrestins 1 and 2, also referred to as arrestins 2 and 3, interact with and regulate numerous non-visual GPCRs. 34 Fig. 3 a Schematic representation of GPCR activation process. Upon agonist (red circle) binding, the receptor proceeds into a pre-activation state coupling with the G protein heterotrimer, where the exchange of GDP and GTP in G protein α subunit leads to G protein dissociation and mediate G protein signaling pathway. The phosphorylation of the receptor C-terminal tail by GRK binding promotes arrestin recruitment and signaling. When the antagonists (blue circle) bind, the receptor stabilizes in an inactive state. b Crosstalk of downstream pathway of Gs, Gq, Gi and arrestin Originally classified as monomers, GPCRs were subsequently recognized to engage in homo- or hetero-dimerization, displaying distinct properties in receptor activation, pharmacological cascades, and biological functions. 53 , 54 Recent research indicates that GPCRs can bind to various single transmembrane accessory proteins to regulate their biological functions such as ligand binding, transducer coupling, and intracellular signaling. 55 , 56 Prominent examples include the family of receptor activity-modifying proteins (RAMPs) that majorly regulate the glucagon receptor (GCGR) and the melanocortin receptor accessory proteins (MRAPs) that regulate the melanocortin receptors (MC1R-MC5R). 57 , 58 Currently, the interactions between the negative allosteric modulator RAMP2 and GCGR as well as the positive allosteric modulator MRAP1 and MC2R have been elucidated by cryo-EM. 59 , 60 Structural changes within GPCRs facilitate their function as molecular conduits that transmit extracellular signals across membranes to elicit cellular responses. A distinctive feature of GPCR activation involves notable outward movement of the cytoplasmic end of TM6, creating an intracellular pocket to accommodate the downstream transducers. GPCRs contain several conserved structural motifs relevant to their activation, including the CWxP motif of TM6, the NPxxY motif of TM7, and the ionic lock that involves TM3-TM6, as well as TM3-TM7. 61 – 63 Additionally, Na + acts as an endogenous negative allosteric modulator (NAM) of class A GPCR activation, stabilizing the inactive state through direct interactions. 64 , 65 High resolution structures reveal that Na + interacts mainly with residues from TM1, TM2, TM3, and TM7 and these interactions vary across GPCRs. 66 , 67 Ligands can regulate receptor activity by stabilizing distinct conformations. Since the diverse signaling pathways elicit distinct physiological effects, ligands that selectively induce beneficial pathways hold promising therapeutic value. These drugs are commonly referred to as “biased ligands.” For instance, G protein-biased μ-opioid receptor (μOR) agonists are of remarkable clinical relevance as they enhance analgesia effects and reduce adverse reactions associated with the activation of β-arrestin pathways, in contrast to morphine. Several novel biased ligands are currently in clinical use or under investigation, such as TRV130, PZM21, and SR-17018. 68 – 70 Hence, unraveling the coupling mechanisms governing G proteins, GRKs, and arrestins will establish a robust foundation for designing biased ligands tailored to selectively activate or inhibit specific pathways. In the absence of agonists, GPCRs may display different levels of constitutive activities. The efficacy of diverse ligands acting on a single GPCR in terms of activation or inactivation also varies widely. Considering both receptor constitutive activity and drug efficacy, GPCR ligands are categorized as (full) agonists, partial agonists, antagonists, and inverse agonists. These variations in efficacy significantly influence their therapeutic properties. Functional diversity of GPCRS Overview of GPCR subfamilies and their physiological functions GPCRs can be categorized into class A, class B, class C, class F, and class T according to their structural and functional characteristics. Class A GPCRs, namely the rhodopsin-like family, is the superfamily with the largest proportion and the most extensive research. 71 Class A GPCRs can further be divided by function into aminergic, peptide, protein, lipid, melatonin, nucleotide, steroid, dicarboxylic acid, sensory, and orphan subgroups, 72 with their corresponding indications ranging from hypertension, cardiovascular diseases, and pulmonary diseases, to depression and psychiatric disorders. 17 Class B GPCRs are divided into secretin (B1) and adhesion (B2) subfamilies, with the former characteristic of large extracellular domains (ECD) and the latter possessing a unique long N-terminal motif and autoproteolysis-inducing domain. 73 While glucagon-like peptide-1 receptor (GLP-1R) and glucagon receptor (GCGR) are emerging as the famous B1 GPCR targets in regulating blood glucose homeostasis and lipid metabolism; 74 , 75 the B2 subfamily is critical in modulating sensory, endocrine, and gastrointestinal systems. 76 Class C GPCRs, the glutamate receptors, are unique in their large ECDs, conserved venus fly traps (VFTs), cysteine-rich domains (CRDs) on the ligand binding sites, and constitutive dimers for receptor activation. 77 With mGluRs (metabotropic glutamate receptors) taking the lead in clinical transformation, the physiological functions of class C GPCRs are implicated in cancer, migraine, schizophrenia, and movement disorders. 77 Class F GPCRs, comprising 10 frizzled receptors (FZDs) and one smoothened receptor (SMO), are distinctive in their conserved CRD regions and involvement in Hedgehog and Wnt signaling pathways. Therefore, they are mainly associated with cancer, fibrosis, and embryonic development. 78 The current drug discovery is only focused on SMO, 79 leaving broad exploration space for the therapeutic potential of FZDs. Particularly, although taste 2 receptors (TAS2Rs), the receptors modulating taste perception of humans, show structure similarity with class A GPCRs, their low sequence homology (<20%) with the existing types of GPCRs isolates them to a novel category of class T GPCRs, 76 deepening our understanding of the entire GPCR family. Involvement of GPCRs in sensory perception, neurotransmission, and endocrine regulation Rhodopsin, TAARs, and TASRs in sensory perception One of the most significant physiological functions GPCRs exercise is mediating sensory information such as light perception, taste, olfaction, and pheromone sensation. Rhodopsin, which contributes to the first stage of visual activation in vertebrates, exhibits the typical and representative features of class A GPCRs. Upon absorbing photons, the orthosteric ligand of rhodopsin, retinal, experiences conformational flipping within picoseconds, thus rapidly triggering signal propagation from the receptor to G proteins, cGMP phosphodiesterase, or cGMP-gated ion channel. 80 The covalent linkage of retinal with the receptor, and the instantaneous overturning and signaling serve as a paradigm for elucidating the efficiency of GPCRs in sensory perception. Olfactory sensory receptors, which can be categorized into odorant receptors (ORs) and trace amine-associated receptors (TAARs), are a valuable medium for researchers to understand olfactory information encoding. Guo et al. 81 has recently revealed the universal mechanism of TAARs recognition of amine odor molecules and the structural basis of “combinatorial coding” of the olfactory receptor in ligand recognition. Notably, the selective coupling of mTAAR9 with Gs and Golf is also delineated, which serves as a pioneer in the field of mammalian olfactory recognition. Apart from selective G-proteins, the downstream transduction mechanism of olfactory receptors is also associated with adenylyl cyclase and cAMP-gated ion channel, 82 leaving favorable exploration opportunities. To regulate the sensory function of taste, which is one of the most important sensations in human life, taste receptors (TASRs) are extensively studied from physiological and pharmacological perspectives. Among them, type I taste GPCRs function by forming heterodimeric complexes to stimulate sweet (TAS1R2/TAS1R3) and umami (TAS1R1/TAS1R3) sensation, whereas Type II are monomeric TAS2Rs that regulate bitter flavor. 83 Tastant binding to the receptor activates downstream secondary messengers, resulting in depolarization and sensitizing the transient receptor potential (TRP) channel, which in turn innervates the gustatory cortex in the brain. 84 Given the inapplicability of the previous GPCR expression techniques in TAS2Rs, 85 overcoming difficulties in the structural determination of taste receptors will further facilitate their physiological research. μOR and CBR in neurotransmission Currently, neurological therapeutic demands mainly revolve around neuropathic pain alleviation, treatment of depression, psychiatric disorders, and Parkinson’s diseases. μ-Opioid receptors (μORs), possessing a research history of over 50 years, have been extensively researched about their mechanism of analgesic action in the peripheral nervous system (PNS) and the central nervous system (CNS). For instance, μORs reduce the release of nociceptive substances and decrease Ca 2+ production following nerve injury by interacting with TRPV1, H1R, and NK1R in nociceptive receptors, 86 whereas in spinal dorsal horn neurons, μORs modulate 5-HT receptors, glycine receptors, and norepinephrine receptors to activate pain inhibitory pathway. 87 Orthosteric biased modulators, allosteric modulators, and bitopic modulators have been successively developed to exert analgesic effects while alleviating side effects like respiratory depression and addiction. 88 Cannabinoid receptors (CBRs) are also representative targets involved in neurotransmission and neuropathic pain pathophysiology. The subtype CB1R is primarily found in presynaptic terminals of neurons in CNS, the activation of which inhibits neurotransmitter release and algesthesia transmission, 89 while CB2R is highly expressed in immune cells, the activation of which can inhibit inflammatory factors that promote pain sensitization. 90 No-selective orthosteric CB1R and CB2R activators can produce an antinociceptive effect and improve sleep in several animal models, while selective positive allosteric modulators (PAMs) like ZCZ011 ( 40 ) are rising as more promising ligands without inducing cannabis-like side effects. 91 GLP-1R and GPR120 in endocrine regulation Endocrine syndrome has been rising as one of the most critical health issues in the 21 st century. Numerous metabolism-related GPCRs, which are usually activated by energy metabolites or substrates, are pivotal sensors of endocrine dysregulation. GLP-1R and GPR120 (also known as free fatty acid receptor 4), for example, are both promising therapeutic targets for the treatment of type 2 diabetes and obesity. 74 , 92 Mechanistically, the endogenous ligand of GLP-1R, GLP-1, can reduce the secretion of glucagon in pancreatic α cells and promote insulin secretion in pancreatic β cells. For GPR120, however, the binding of omega-3 polyunsaturated fatty acids (ω3-FAs) and receptor activation can reduce inflammation of adipose tissue and protect against insulin resistance. 93 The receptor’s coupling with G q/11 subsequently stimulates the PI3K/Akt pathway, resulting in the uptake of glucose in adipocytes. 94 As GLP-1R agonist liraglutide takes the lead in FDA-approved drugs treating type 2 diabetes and obesity, 95 drug development of more endocrine-related targets such as GPR35, GPR40, GPR41, GPR43, GPR81, and GPR119 are supposed to come into our view. Receptor promiscuity and cross-talk between different signaling pathways GPCR receptors convert the extracellular stimuli to intracellular signals to control cellular function and phenotype. GPCR receptors convert the extracellular stimuli to intracellular signals to control cellular phenotype and function. These intracellular signaling pathways intersect with each other to enhance or downgrade relevant responses in a phenomenon known as “cross-talk.” The promiscuity of the GPCR signaling network is consequently outlined, resulting in more extensive regulation, low selectivity, and possible adverse effects. Promiscuity and cross-talk can occur at three levels, including the GPCR receptors, G-proteins/β-arrestins, and the downstream effectors. The receptor promiscuity lies in the formation of heterodimers, which can either be constituted of subtypes of the same receptor family or those of different families. A compelling case is the heterodimerization of GABA b(1) and GABA b(2) which leads to the functionality of modulating GIRK (G-protein gate inward rectifying channel) potassium channels, whilst neither of them is functional when expressed as a monomer. 96 Another well-established example is the plentiful interrelationship of adenosine receptors and dopamine receptors, where the activation of A 1A and A 2A adenosine receptors decreases dopamine binding to D1 and D2 dopamine receptors. 97 The bivalent ligands that bind adenosine receptors and dopamine receptors at each end further demonstrate the occurrence and functionality of heterodimerization. 98 The participants of heterodimerization are assumed to share a common G-protein pool, thus contributing to the redistribution of their interaction of G-proteins and reshaping the signaling landscape. 99 Given this, by direct cross-talk between two GPCR receptors, ligands can be designed towards one receptor to modulate the affinity and efficacy of the other target, although certain pharmacological profiles remains unclear. At the second stage of the hierarchical signaling of GPCRs, namely the recruitment of G s , G i , G q , G 12 , β-arrestin 1, and β-arrestin 2, a spectrum of coupling strengths ranging from highly selective coupling to promiscuous coupling is exhibited. MD simulations performed by Sandhu et al. revealed that engineered mutant GPCRs can alter the coupling of non-cognate G-proteins by reshaping the intracellular interface, 99 demonstrating that “dynamic structural plasticity” of the GPCR cytosolic pockets is the foundation of G-protein promiscuity. Mutants, orthosteric and allosteric modulators that exert long-range and delicate effects towards the cytosolic binding interface are therefore principal strategies to achieve selectivity of G-protein signaling. The promiscuity of distinct downstream effectors, known as the third stage of signaling, is highly correlated with the cross-talk of G-proteins. Normally, stimulation of G s , G i, and G q results in the activation of AC, the inhibition of AC, and the stimulation of PLC, respectively. 100 However, once distinct G-proteins are recruited near the membrane at a similar time, βγ subunits released from respective G-protein activation are “exchangeable” between diverse signaling pathways and can potentiate responses mediated by other G-proteins. 101 The second messengers then phosphorylate, activate, or deactivate each other to construct a fine-tuning network (Fig. 3b ). Albeit conducting a great deal of research, the precise control of GPCR promiscuity remains obscure. Impact of GPCR mutations on human diseases and therapeutic implications Besides being involved in numerous physiological processes, mutations in GPCRs can be linked to manifold human diseases, underlying the necessity of GPCR genomics, and imposing therapeutic implications. Till date, over 2350 mutations in GPCR genes have been identified as the major causes of more than 60 inherited monogenic diseases in humans (Fig. 4a ), with missense mutations harboring the maximum proportion (>60%) and small inserts/deletions ranking the second (>15%). 16 Fig. 4 a Categories of Representative human diseases caused by GPCR dysfunctions. b Classification of the effects of mutations on GPCR dysfunctions Classification of the effects of mutations on GPCR dysfunctions The effects of mutations in GPCRs can be categorized into gain-of-function (GoF) and loss-of-function (LoF), corresponding to physiological hyperfunction and hypofunction, respectively. Recent studies have provided a more detailed explanation of the diverse underlying mechanisms of GoF and LoF mutations. Compared with the wild-type (WT) GPCR activation, the common pharmacological mechanisms of activating and inactivating mutations lie in three aspects: (1) Mutations transform micro-switch cascades within the receptors and induce active/inactive conformations, thus altering the constitutive activity of GPCRs and affecting the recruitment of downstream effectors. (2) Mutations influence receptor expression directly or indirectly increasing/decreasing receptors’ intracellular transport, degradation, and recycling. (3) Some mutations affect ligand potency, specificity, or promiscuous recognition, thereby exerting regulatory functions by shifting the conformational population, redistributing the downstream couplings, or altering receptor dimerization. Furthermore, all variants are not pathogenic. This provides robust evidence for another classification of “driver” and “passenger” mutations (Fig. 4b ). 102 Computational approaches are recently emerging to predict the driver ability of mutations in GPCR-related diseases, based on abundant clinical data of mutations and relevant GPCR dysfunctions. Correlation of GPCR mutations and human diseases The genomic alterations induced by GPCR mutations serve as the major driver of various monogenic diseases. Some well-established examples include missense mutations in SMO receptor causing basal cell carcinoma, 103 missense and nonsense mutations in MC4R causing obesity 104 , and missense mutations in FSHR inducing ovarian hyperstimulation syndrome. The majority of mutations are highly conserved and thus in an advantageous position during evolution. 105 Therefore, the pathological relevance between GPCR mutations and human diseases may be more effectively predicted taking the evolutionary conservation of a certain residue into consideration. Therapeutic implications and approaches of GPCR pathologies Terapeutic approaches of GPCR pathologies mainly include symptomatic and etiological treatment. As many GPCR dysfunctions ultimately result in endocrine diseases with end-organ resistance or cancer, the administration of hormones or chemotherapeutics may be considered to reduce pathologic phenotypes. 106 , 107 More state-of-the-art therapeutic implications, however, are oriented towards etiological treatment. Missense mutations in GPCRs can mislead protein folding and post-translational modifications to cause trafficking alterations, in which pharmacological chaperones are applicative therapeutic regimens. 108 For receptor truncation resulting from nonsense mutations or frame-shifting mutations, RNA interference, gene replacement approaches, and the genome editing approach CRISPR/Cas9 may rescue the receptor integrality, provided that at least the first three transmembrane helices remain in the mutant receptor. 109 Designing peptides or small molecule modulators is the most straightforward means for the restoration of receptor pharmacology, though high expenditure in multiple mutations remain an intractable issue. Advances in GPCR drug discovery Overview of traditional and emerging approaches for GPCR drug discovery Since enkephalin was first recognized as the endogenous ligand of opioid receptors, 110 the discovery of modulators with diverse regulatory effects is constantly endowing meaning in the research of GPCRs. Several decades have witnessed the transformation from serendipity to rational design in the field of GPCR drug discovery, and the ligands have been expanded from natural products to synthesized compounds and engineered antibodies. Currently, apart from the traditional molecular docking and SBDD, more screening methodologies of wet experiments have been established to facilitate the selection of high-quality hits, including FRET/ BRET (Bioluminescence Resonance Energy Transfer) assay, NanoBiT (NanoLuc Binary Interaction Technology) assay, Tango assay, and 19 F NMR. 111 Once the hits were obtained, structure-activity relationship (SAR) optimization in synergistic application of computational methodologies such as fragment-growing, property prediction, and MD simulations, was conducted to initiate the hit-to-lead and lead-to-drug campaign. 99 Herein, we specially emphasize on the interaction and signaling mechanism of synthetic small-molecule modulators bound to GPCRs, with the aim of enlightening the discovery of more ingenious molecules with high potency, selectivity, and potential biased effects. Structure-based drug design targeting the orthosteric sites of GPCRs Orthosteric small molecule modulators are the most universal non-peptide regulators of GPCRs. By competing with endogenous ligands, they interact with the orthosteric binding pocket (OBP) and exert a full agonistic 112 , 113 /partial agonistic 114 /antagonistic function 115 , 116 by triggering the conformational displacement of GPCR internal structures. 117 , 118 Despite their relatively mature development, side effects derived from low subtype selectivity and promiscuous signaling remain the major hurdle. 49 , 119 Over the past 30 years, the widespread use of X-ray and Cryo-EM has facilitated the characterization of GPCR-orthosteric ligand complexes, with 657 class A, 16 class B1, 6 class B2, 19 class C, 18 class F, and 1 class T structures solved (supplementary Table 1 – 5 ). 120 Here, we meticulously selected five representative complexes in which ligands have been launched recently to elucidate the mechanisms of ligand recognition, specificity, and elaborate signaling transduction. Furthermore, we exemplified two cases to demonstrate the beneficial engagement of structural information in exploiting not only SAR but also the structure-functional selectivity relationship (SFSR). Considering these seven cases as a paradigm, we aimed to condense valuable hints based on a detailed analysis of approved drugs or selective compounds and provide a constructive outlook for the high-quality discovery of GPCR orthosteric modulators that may overcome the current dilemma. μOR in complex with oliceridine With morphine and fentanyl ( 1 ) the most effective drugs treating acute or chronic pain, 121 , 122 their common receptor μOR was revealed to be responsible for both analgesic and adverse effects. 123 – 125 To attenuate side effects and broaden the therapeutic window, modulators that can abolish β-arrestin activity while maintaining relatively intact G-protein signaling are of intense pharmaceutical interest. 126 – 128 Oliceridine ( 2 ), a partial agonist binding at the orthosteric site of μOR, was approved by the FDA in 2020 for its ability to biased signaling via the G protein pathway and thus alleviating side effects (Fig. 5a, b ). 129 Therefore, casting light on the oliceridine-μOR complex structure and the underlying mechanism of biased signaling will provide insight in developing a novel generation of analgesic drugs. Fig. 5 a A bridged general view of fentanyl and oliceridine inducing distinct pharmacological profiles. b 2D structure of fentanyl and oliceridine shown for clarity. c Superimposed views of μOR–fentanyl (gray cartoon, gray sticks; PDB: 8EF5) and μOR–oliceridine (light green cartoon, light green sticks; PDB: 8EFB) complex structure, together with the comparison of ligand binding modes and arrestin coupling interfaces, are presented. 2D structures of two designed biased modulators are also presented By aligning the complex structures of μOR–oliceridine and μOR–fentanyl, a well superimposed binding mode in OBP above Trp295 6.48 was found. The only exception was that the pyridine ring of oliceridine tilts 35° toward TM2 relative to the n-aniline group of fentanyl, resulting in weaker hydrophobic interactions with TM6/7 than that with fentanyl. Based on the results of MD performed by Zhang et al., 130 extended interactions with TM6/7 can be inferred to have elicited inward movement of TM6 and TM7-H8 toward the TM core, shaping adaptive intracellular pocket conformation for both G-protein and β-arrestin coupling and thus leading to neutral signaling, whereas reduced interactions may have kept the intracellular end of TM6/7 relatively away from the TM core and therefore stabilize an intracellular pocket preferential for G protein binding and signaling. Two fentanyl-derived μOR agonists ( 3 - 4 ), which substituted the aniline group on fentanyl with n-propyl or isopropyl to reduce hydrophobicity with TM6/7, were thereupon designed as “proof-of-concept” to successfully achieve biased signaling via the G protein pathway (Fig. 5c ). Different from the “trial-and-error” mode when developing biased ligand oliceridine, 131 the comprehensive study by Zhang et al. serves as a paradigm for dissecting co-crystallized complexes to understand the molecular basis of preferential signaling mechanisms initiated from the orthosteric pocket and broadens the avenue for designing biased modulators of ORs through SBDD strategies. S1PR in complex with siponimod Sphingosine-1-phosphate receptor (S1PR), a family of class A GPCR consisting of five subtypes, S1PR1-S1PR5, modulates diverse physiological functions, including lymphocyte trafficking, vascular development, endothelial integrity, and heart rate. 132 – 136 Although Fingolimod received regulatory approval from the FDA in 2010 as a first-in-class S1PR agonist, 137 its low subtype selectivity has led to several “off-target” effects, including bradycardia and atrioventricular blockade. 138 Therefore, a second-generation, highly subtype-selective S1PR modulator is crucially needed. Siponimod ( 5 ) was globally approved in 2019 for the treatment of adults with relapsing MS by selectively targeting S1PR1 and S1PR5. 139 Insights into the mechanisms of drug recognition and receptor activation will provide a framework for understanding ligand selectivity and signal transduction in GPCRs. 140 , 141 Yuan et al. presented the cryo-EM structures of siponimod–S1PR1–G i and siponimod-S1PR5 complexes, in which the ligands exhibited an identical linear conformation across a polar module and the deep hydrophobic cavity of the orthosteric pocket (Fig. 6a ). 142 Given that members of the S1PR family display different extracellular vestibules, distinct extracellular leaflets have been reported to have contributed to diverse access channels for ligand entry and thus relate to specificity among subtypes (Fig. 6b ). 143 Moreover, further careful comparison of the siponimod-S1PR1-G i complex with antagonist ML056-bound S1PR1 structure underlines the “twin toggle mechanism” during receptor activation. 144 Upon ligand binding, Leu128 3.36 rotates 130° away from TM5 to form a direct interaction with the hydrophobic portion of siponimod, disrupting its previous interaction with Trp269 6.48 and triggering a synergistic downward movement of Trp269 6.48 . The dramatic displacement of the two residues can therefore loosen the interaction between TM3 and TM6, inducing a consequent outward movement of TM6 that can accommodate G protein binding (Fig. 6c, d ). Similar activation mechanism involving corresponding mechanical switches can also be found in CB1 and MC4R, 145 , 146 which provides valuable hints that designing ligands forming elaborate hydrophobic interaction with residue 3.36 or directly inducing reconfiguration of 3.36-6.48 may contribute to enhanced activation efficacy. Fig. 6 a Detailed binding modes of S1PR5 in complex with siponimod. Labels of the residues engaged in polar contacts with siponimod are colored in blue, with hydrogen bonds presented by orange dashes. The residues of the hydrophobic pocket that stabilizes ligand binding are marked with green labels, while residues that are critical for signal transduction are labeled in red. b Superimposed views of S1PR1 (orange cartoon, PDB: 7T6B), S1PR2 (light green cartoon, PDB: 7C4S), S1PR3 (light purple cartoon, PDB: 7YXA), and S1PR5 (yellow cartoon, PDB: 7TD4) GPCR structures, where TM1 and TM7 of S1PR5 are highlighted for clarity. c Superimposed views of active S1PR1-siponimod complex (cyan cartoon, cyan stick, PDB: 7TD4) and inactive S1PR1 structure (gray cartoon, gray stick, PDB: 3V2Y) to illustrate the “toggle switch” activation mechanism. d 2D structure of siponimod is shown for clarity OX2R in complex with lemborexant Orexin receptors are expressed throughout the central nervous system and demonstrate therapeutic potential for insomnia by regulating the sleep-wake cycle. 147 – 149 The two subtypes, OX1R and OX2R, dominate the respective regulatory behaviors, with OX1R involved in gating rapid eye movement (REM) sleep and OX2R involved in gating non-REM and REM sleep. 150 Lemborexant ( 6 ), an orthosteric competitive antagonist approved by the FDA in 2019, exhibits outstanding inhibitory activity against OXRs. 151 , 152 However, the most important features of lemborexants lie in two aspects: (1) Why lemborexants show moderate selectivity toward OX2R over OX1R, 152 which will facilitate the design of OX1R/OX2R-selective modulators that can be applied to REM and non-REM functionality studies? 2) What is the basis of the dynamic parameters of lemborexant that may explain the relationship between drug-induced improvement of sleep onset and a decrease in wake time after sleep? To elucidate the mechanism of lemborexant subtype selectivity and provide guidance for anti-insomnia drug development, Asada et al. presented the crystal structure of the OX2R–lemborexant complex and compared its ligand-binding mode with that of the previously solved OX1R–lemborexant complex structure. 153 Despite the ligand’s shared hydrogen bonds with Gln126 3.32 of OX1R and Gln134 3.32 of OX2R, lemborexant binds OX1R as a mixture of two orientations owing to the small side chain of Ala127 3.33 , whereas lemborexant binds OX2R in only one configuration because of the steric hindrance of Thr135 3.33 , which is inferred to be the primary cause of the difference in its affinity for OX1R and OX2R (Fig. 7a, b ). In contrast, by simulating lemborexant in solution, the intramolecular stacking of two aromatic rings was observed to play a vital role in shaping the conformation of lemborexant close to the bound state before receptor binding, which explains the high k on value of the ligand. In addition, the higher binding free energy of lemborexant compared to other OXR modulators may contribute to a higher k off value. Collectively, these observations highlight the possibility of obtaining a high k on by optimizing the conformation of free molecules via intramolecular interactions (Fig. 7c, d ). By extension, separately modulating the enthalpy of molecular binding to the receptor and entropy derived from the intramolecular structure may be important strategies for designing drugs with enhanced kinetics and dynamics. Fig. 7 a Detailed binding mode of lemborexant in complex with OX2R (receptor: light orange, ligand: cyan, PDB: 7XRR), where steric hindrance of T135 3.33 only allows one orientation of the ligand. b Detailed binding mode of lemborexant in complex with OX1R (receptor: light pink, ligand: yellow, PDB: 6TOT), where small side chain of A127 3.33 accounts for two orientations of the ligand. c Abridged general view of employing MD simulation to predict the conformation of the ligand before receptor binding, to improve K on values. d 2D structure of lemborexant is shown for clarity 5-HT 1F in complex with lasmiditan The 5-HT1 receptor subtypes, including 5-HT 1A , 5-HT 1B , 5-HT 1D , 5-HT 1E , and 5-HT 1F , are well-known class A GPCRs that respond to the endogenous neurotransmitter serotonin and have been proven to be promising targets for the treatment of migraine, depression, and schizophrenia. 154 – 156 Although traditional targeted agonists have been clinically used as anti-migraine drugs for decades, side effects such as therapeutic vasoconstrictive actions owing to the non-selective activation of 5-HT 1B and 5-HT 1D remain a major hindrance. 157 Lasmiditan ( 7 ), a potent and highly selective drug toward 5-HT 1F was approved by the FDA in 2019 because of its vasoconstrictive side effects and high-penetration properties. 158 Elucidation of the scaffold features of lasmiditan and the mechanism of 5-HT 1F -selective activation will provide a template for the rational design of safer anti-migraine drugs. Through the 5-HT 1F -lasmiditan-G i1 complex solved by Huang et al., an overview of the lasmiditan-binding mode was presented. 159 In the orthosteric binding pocket, the primary amine on the methylpiperidine group largely contributes to the stability of lasmiditan by forming a canonical charge interaction with Asp103 3.32 of the receptor while simultaneously forming a hydrogen bond with Tyr337 7.42 . Notably, in the extended binding pocket (EBP), the trifluorobenzene group of lasmiditan forms additional hydrophobic interactions with Ile174 ECL2 and Pro158 4.60 and forms hydrogen bonds with residue Glu313 6.55 , Asn317 6.59 , Thr182 5.40 , and His176 ECL2 . Structural alignment of 5-HT 1F with other 5-HT1 receptor subtypes revealed that the TM4-TM5-ECL2 region, which is highly conserved in the other four subtypes, underwent a notable conformational change, thereby disrupting the interaction between lasmiditan and 5-HT 1A , 5-HT 1B , 5-HT 1D , and 5-HT 1E . Thus, designing ligands that accommodate EBP and form specific interactions with the TM4-TM5-ECL2 region may enable high 5-HT 1F selectivity (Fig. 8a, b ). Activation mechanical analysis by Huang et al. revealed that lasmiditan triggers the downward movement of the toggle switch residue Trp 6.48 and then induces conformational rearrangement of the PIF, DRY, and NPxxY motifs. Particularly, structural comparison of 5-HT 1F -G i complex and other 5-HT 1 -G i/o showed that the αN of 5-HT 1F -bound G i shifts away from other 5-HT 1 receptor-bound G i/o , suggesting unique G i coupling and corresponding specific downstream effects (Fig. 8c ). Therefore, designing modulators that interact with the toggle switch residue and optimize their blood-brain-barrier (BBB) penetration properties may yield effective and safer 5-HT 1F agonists. Fig. 8 a Detailed binding mode of 5-HT 1F in complex with lasmiditan. Hydrogen bonds are presented by orange dashes, while halogen bonds are presented by green dashes. b Superimposed views of 5-HT 1A (light green cartoon, PDB: 7E2X), 5-HT 1B (light orange cartoon, PDB: 5V54), 5-HT 1D (light gray cartoon, PDB: 7E32), 5-HT 1E (light pink cartoon, PDB: 7E33), and 5-HT 1F (light purple cartoon, PDB: 7EXD). The TM4-ECL2-TM5 region of the 5-HT 1F receptor is highlighted for clarity. c The structure alignment comparison of αN helices of G protein coupling with their corresponding 5-HT receptors. αN helix of G i protein coupled with 5-HT 1F is highlighted for clarity. d 2D structure of lasmiditan GnRH1 in complex with elagolix The representative class A GPCR, gonadotropin-releasing hormone 1 receptor (GnRH1R), once activated by its endogenous peptide activator, gonadotropin-releasing hormone (GnRH), can initiate the reproductive hormone cascade and release gonadotropins through the activation of the G q protein pathway. 160 – 162 With the first availability of GnRH1R non-peptidic antagonist elagolix ( 8 ) on the market in 2018, 163 structural insights into the GnRH1R-elagolix complex have gained pharmaceutical interest. 164 Additionally, unlike other class A GPCRs, GnRH1R lacks a C-terminal helix (helix 8) in the cytoplasmic region and harbors Asn 2.50 instead of the highly conserved Asp 2.50 present in other receptors, 165 leaving a wide space for different microswitches along the signaling cascade within 7TMD. The crystal structure of the GnRH1R-elagolix complex studied by Yan et al. revealed that polar network residues composed of Lys121 3.32 and Asp98 2.61 play critical roles in forming polar interactions with the ligand, whereas Tyr283 6.51 and Tyr290 6.58 are engaged in ligand recognition by contributing to hydrophobic interactions (Fig. 9a ). Notably, Elagolix is located closer to TM7, resulting in an enlarged orthosteric pocket that allows N-terminal entry and co-occupation of the site. Structural alignment and IP accumulation assays showed that, unlike some GPCRs in which ligands can contact residue Trp 6.48 directly and trigger the toggle switch, the special motif Tyr283 6.51 -Tyr284 6.52 -Trp280 6.48 -Phe276 6.44 in TM6 was suggested to be a critical structural motif involved in mediating the propagation of signal transmission (Fig. 9b ). Moreover, only 4% of class A GPCRs, including GnRH1R, have asparagine at the 5.58 position, which is implicated in a polar interaction with Ser136 3.47 GnRH1R, thus leading to TM6 packing tightly with TM3 and TM5 in GnRH1R and exercising an antagonistic function. Collectively, these analyses highlight the distinctive features of GnRH1R in the binding of a representative antagonist and provide insights for structural biologists. Fig. 9 a Detailed binding mode of GnRH1 in complex with elagolix (receptor: light gray, ligand: light pink, PDB: 7BR3), where N-termini of GnRH1R is highlighted in a light purple to present its co-occupation with elagolix in the orthosteric pocket. b Overview of the special signal transduction mechanism in GnRH1R. c 2D structure of elagolix for clarity SFSR study utilizing structural information While the development of crystallography over the last decade has revealed an attractive possibility of SBDD, the mainstream strategy of GPCR drug discovery remains extensive SAR study and fragment-based drug design (FBDD). 166 – 168 This is partially due to the “activity-cliff” phenomenon which to some extent, undermines the profits from structural information. Nevertheless, the promising prospect still deserves expecting. Two examples are analyzed here to arouse future interest in SFSR studies utilizing structural information. The first example is the efficient discovery and optimization of A 2 AR selective antagonist 1,2,4-triazine derivative 4d ( 9 ) via SBDD strategy. With Biophysical Mapping (BPM) approach and crystal structure analysis, compound 4d was revealed to be primarily stabilized by two hydrogen bonds between the triazine core and N253 6.55 , with ring A oriented towards TM2 and TM7 (Fig. 10a ). Hence, the presence of a hydrogen bond acceptor at the para position of ring A to interact with His278 7.43 , as well as the introduction of one or more flanking lipophilic substituents on the same ring to interact with Ile66 2.64 was suggested as the focus of the SAR program. Introducing either a phenolic hydroxyl or 4-pyridyl nitrogen at the para position of ring A, and fine-tuning affinity by various combinations of small lipophilic substituents efficiently yielded compound 4k , which proves the best balance of potency and efficacy (Fig. 10b ). 169 Further research compared the binding pockets of A 2 AR in complex with adenosine (agonist), ZM241385 (antagonist), and compound 4e (antagonist). The hydrophobic sub-pocket in the lower chamber was observed to be occupied by the ribose ring system of adenosine analogs in agonist complexes, though was typically unoccupied when antagonists bound. The same region also allowed optimization of selectivity for A 2 AR over A 1 AR (Fig. 10c ). Therefore, expanding chemotypes into this region may harvest a more efficient chemical series when designing selective and diverse functional modulators. 170 Fig. 10 a Detailed binding mode of A 2 AR in complex with compound 4d (receptor: light gray, ligand: orange, PDB: 3UZA). b SAR study of A 2 AR antagonist. c Comparison of the orthosteric binding site of A 2 AR–Adenosine complex (light blue, PDB: 2YDO), A 2 AR–ZM241385 complex (light yellow, PDB: 4EIY), A 2 AR–Compound 4e complex (light green, PDB: 3UZC), the difference in cavity occupation is highlighted by red circles and arrows The second paradigm entails the structure-based drug design of novel β-arrestin-biased D2R agonists commencing with aripiprazole, so as to alleviate the movement disorders associated with the adverse effects of antipsychotics. 171 The dichlorophenylpiperazine portion of aripiprazole was first replaced with an indolepiperazine, leading to 12 that displayed comparable activity in both G i/o -mediated cAMP inhibition and β-arrestin2 recruitment assays. Molecular docking with a D 2 R homology model revealed that the indole NH of 12 formed a hydrogen bond with Ser5.42, which has been shown to mediate G-protein-dependent signaling in highly homologous β2 adrenergic receptors. A methyl group was thus attached to the NH of indole to fine-tune the binding conformation of 12 and thereby preclude TM5 engagement (Fig. 11a ). Inspired by structural information from homologous 5-HT 2B receptor, where ligand interactions with hydrophobic residues on ECL2 appear to promote β-arrestin recruitment (Fig. 11a ), a second methyl was introduced to position 2 of the indole ring, yielding 13 with a β-arrestin bias factor of 20 and potentially reduced side effects (Fig. 11b ). 172 To our knowledge, this is the first successful attempt at using structural information for the rational design of GPCR-biased ligands, underlining the necessity of interactive structural comparison in SFSR study. Fig. 11 a Detailed binding mode of D 2 R in complex with compound 1 ( 12 ) (receptor: light gray, ligand: salmon, the receptor is modeled from PDB: 3PBL). TM5 of the receptor is colored in pink and ECL2 is colored in blue for clarity. b SAR study of β-arrestin biased agonists of D 2 R Delineation of GPCR structures complexed with small-molecule allosteric modulators and allosteric signaling Over the past 10 years, allosteric drug discovery targeting GPCRs has witnessed significant progress in structural understanding, with the advance in knowledge of GPCR allostery. 173 , 174 Till February 2024, the crystal structures of 59 allosteric small-molecule modulators bound to GPCRs have been solved, including 33 class A, 7 class B, 18 class C, and 1 class F modulators. These structures reveal that despite the intrinsic dynamic nature of GPCRs and the structural diversity among different GPCRs, only limited locations function as allosteric pockets, and the same pockets are present in GPCRs with different homologies. 175 Even within a single receptor, more than one allosteric site has been identified. In addition, druggable allosteric hotspots spread throughout the receptor and can be divided into the following sections: extracellular vestibule, transmembrane domain, intracellular surface, outside 7TMD, and inside 7TMD domains. 174 , 175 As will be discussed, allosteric binding sites in all GPCRs are currently known to be located at 11 distinct locations with some consensus, 176 depicted in Fig. 12 . In this figure, the locations of all pockets identified in different GPCRs are mapped onto the structure of an example GPCR to facilitate the comparison of these sites. Fig. 12 11 allosteric binding sites reported across GPCRs mapped onto representative class A GPCR CB1R. Gray pockets represent binding pockets within 7TMD, and white pockets represent binding pockets outside 7TMD. For each pocket, the number of unique ligands is indicated using boldface type, and the number of GPCRs containing the pocket is provided in parentheses. The boundary of the lipid bilayer is indicated by gray dashes Based on the compounds’ ability to affect the stimulatory activity of orthosteric ligands, allosteric ligands can be classified into several categories, including positive allosteric modulators (PAMs), negative allosteric modulators (NAMs), allosteric modulators, and allosteric inverse agonists. 22 , 177 A PAM, such as cinacalcet, targets the calcium-sensing receptor (CaSR) and potentiates the response of the receptor to its orthosteric agonist. Conversely, NAM attenuates the response of the receptor to its orthosteric agonist, mavoglurant, which targets the metabotropic glutamate receptor 5 (mGluR5). 178 Ago allosteric modulators can activate or inhibit a receptor without an orthosteric agonist such as compound 2, which targets the glucagon-like peptide-1 receptor (GLP-1R). Targeting of GPCR extracellular vestibule (outside and inside 7TMD) After the first FDA approval of cinacalcet (a PAM of CaSR) in 2004 as a treatment for hyperparathyroidism, 179 small-molecule allosteric modulators bound to the extracellular vestibule have developed rapidly. Till date, four of these modulators have been approved by the FDA, and one has entered clinical trials, as summarized in Table 1 . Resolved crystal structures have revealed three extracellular binding sites: the pocket outside helices I and II, the pocket outside helices II and III, and the pocket inside 7TMD (Fig. 13 ). 180 , 181 Due to their proximity to the traditional active sites of class A and B GPCRs, such allosteric modulators may exert their effects by directly altering the binding of orthosteric ligands to the receptor. As GPCRs evolved from a common ancestor, this allosteric site, found on receptors, may represent the ancestral orthosteric site. 176 , 182 Table 1 Solved GPCR structures complexed with synthetic allosteric modulators bound to the extracellular vestibule Structure Type GPCR Type GPCR Modulator Highest Phase Modulator type Number PDB code Allosteric site Refs Cryo-EM Class B GLP-1R LSN3160440 Pre-clinical PAM (14) 6VCB outside 7TMD (I-II) 180 Cryo-EM class A GPR101 AA-14 Pre-clinical Allosteric agonist (15) 8W8S outside 7TMD (II-III) 201 X-ray diffraction Class A CCR5 maraviroc Approved Allosteric inverse agonist (16) 4MBS inside 7TMD 181 X-ray diffraction Class A PAR2 AZ8838 Pre-clinical Allosteric antagonist (17) 5NDD inside 7TMD 451 X-ray diffraction Class A GPR52 c17 Pre-clinical Allosteric agonist (18) 6LI0 inside 7TMD 452 Cryo-EM Class A LHCGR Org43553 Pre-clinical Allosteric agonist (19) 7FIH inside 7TMD 453 X-ray diffraction Class A M2R LY2119620 Pre-clinical PAM (20) 4MQT inside 7TMD 216 Cryo-EM Class A M4R LY2119620 Pre-clinical PAM (20) 7V68 inside 7TMD 217 Cryo-EM Class A M4R compound-110 Pre-clinical Allosteric agonist (21) 7V6A inside 7TMD 217 Cryo-EM Class A M4R LY2033298 Pre-clinical PAM (22) 7TRP inside 7TMD 454 Cryo-EM Class A M4R VU0467154 Pre-clinical PAM (23) 7TRQ inside 7TMD 454 Cryo-EM Class A TSHR ML109 Pre-clinical Allosteric agonist (24) 7XW6 inside 7TMD 455 Cryo-EM Class A MRGPRX1 ML382 Pre-clinical PAM (25) 8DWG inside 7TMD 456 X-ray diffraction Class C mGluR1 FITM Pre-clinical NAM (26) 4OR2 inside 7TMD 457 Cryo-EM Class C CaSR cinacalcet Approved PAM (27) 7M3F inside 7TMD 230 Cryo-EM Class C CaSR evocalcet Approved PAM (28) 7M3G inside 7TMD 230 Cryo-EM Class C CaSR NPS-2143 Pre-clinical NAM (29) 7DD5 inside 7TMD 458 Cryo-EM Class C CaSR R-568 Pre-clinical PAM (30) 7SIL inside 7TMD 459 Cryo-EM Class C mGluR2 JNJ-40411813 Phase 2 PAM (31) 7E9G inside 7TMD 460 X-ray diffraction Class C mGluR2 NAM563 Pre-clinical NAM (32) 7EPE inside 7TMD 461 X-ray diffraction Class C mGluR2 NAM597 Pre-clinical NAM (33) 7EPF inside 7TMD 461 X-ray diffraction Class F SMO vismodegib Approved Allosteric antagonist (34) 5L7I inside 7TMD 462 Fig. 13 Three extracellular allosteric binding sites in GPCRs and the corresponding small-molecule allosteric modulators. Stick models of small-molecule ligands are mapped to representative members of outside 7TMD (I and II) (GLP-1R, PDB: 6VCB), outside 7TMD (II and III) (GPR101, PDB: 8W8S), and within 7TMD (M4R, PDB: 7V68) GPCRs. The position of an orthosteric ligand of M4R (shown in gray and sphere-and-stick representation) is mapped onto the overview of allosteric modulators for comparison. For each pocket, the number of unique modulators is indicated in boldface type, and the number of GPCRs containing the pocket is indicated in parentheses

1) Outside 7TMD (TM I-II):

GLP-1R–LSN3160440 structure The glucagon-like peptide-1 receptor (GLP-1R) is a peptide hormone class B GPCR whose activation stimulates the glucose-dependent stimulation of insulin and decreases glucagon secretion. 183 – 185 For such peptide receptors, allosteric pockets on GPCRs may be easier to target for small-molecule drugs than orthosteric drugs. 186 Therefore, highly potent agonists and PAMs of GLP-1R must be developed to treat type 2 diabetes. 187 – 191 LSN3160440 ( 14 ) (Fig. 14 ) is a small-molecule PAM targeted GLP-1R with an EC 50 of 1 μM to enhance the potency and efficacy of GLP-1(9-36) becoming a full agonist. 180 The cryo-EM structure of GLP-1R in complex with LSN3160440, the orthosteric ligand GLP-1, and the Gs protein revealed a clear depiction of the U-shaped binding mode of LSN3160440. The allosteric site is formed by residues on helices I and II in the extracellular vestibule (Fig. 15a ). 180 Within the binding pocket, the benzimidazole moiety of LSN3160440 (Fig. 15c ) formed hydrophobic contacts with Leu142 1.37 (the superscript represents the generic residue numbers of GPCRs) and engaged in aromatic interactions with Tyr145 1.40 (Fig. 15b ). Mutation and molecular dynamics (MD) simulation results also suggest that water-mediated hydrogen bonds may form between N3 of benzimidazole and Lys202 2.72 . 192 , 193 Notably, LSN3160440 interacts with GLP-1 and acts as a molecular glue. 194 The 2,6-dichloro-3-methoxyl phenyl moiety of LSN3160440 forms van der Waals interactions with Phe12 GLP-1 , Val16 GLP-1 and Leu20 GLP-1 simultaneously. Fig. 14 Two-dimensional (2D) chemical structures of synthetic allosteric ligands targeting the GPCR extracellular vestibule Fig. 15 a Schematic representation of PAM LSN3160440 and orthosteric GLP-1 bound to GLP-1R (PDB: 6VCB). GLP-1 is indicated in pink. b Detailed binding modes of GLP-1R bound to LSN3160440; π–π stacking is indicated in gray dashes. c 2D structure of small-molecule allosteric ligand LSN3160440 presented for clarity. d Superposition of orthosteric small-molecule agonists Boc5 (displayed with purple sticks), TT-OAD2 (displayed with salmon sticks), LY3502970 (displayed with yellow sticks), and CHU-128 (displayed with blue sticks) to LSN3160440–GLP-1–GLP-1R structure reveals a partial overlap in the TM1-TM2 cleft. The conserved residue Tyr145 1.40 is highlighted Several structures of orthosteric small-molecule agonists complexed with GLP-1R were resolved (Fig. 15d ). 195 – 198 Structural comparisons of these ligands with LSN3160440 revealed a shared region situated at the extracellular termini of the TM1-TM2 cleft, further suggesting that this is a promising area for lead optimization for both orthosteric and allosteric agonists. Within the binding site, the aromatic interactions with Tyr145 1.40 are conserved.

2) Outside 7TMD (TM I-II)

GPR101–AA-14 structure GPR101 is an orphan class A GPCR that is highly expressed in the nucleus accumbens and the hypothalamus and has constitutive Gs and Gq activity. 199 GPR101 gene duplication or mutation modulates its constitutive activity, rendering GPR101 a promising target for metabolic diseases. 200 Recent studies have identified AA-14 ( 15 ) (Fig. 14 ) as an allosteric agonist of GPR101, demonstrating robust Gs activation activity and high subtype selectivity. 201 In vivo studies have shown that AA-14 exerts rejuvenating effects by activating GPR101 in the pituitary. The cryo-EM structure of the AA-14–GPR101–Gs complex unveils two distinct binding sites for AA-14 (Fig. 16a ): one located outside 7TMD, surrounded by helices I, VI, and VII, while the other is positioned outside TM2–TM3 and ECL1. 201 Within the extracellular allosteric site, the 3-(trifluoromethyl) phenyl group (Fig. 16c ) establishes polar interactions with Asn100 ECL1 and hydrophobic interactions with Phe103 3.24 and Trp87 2.60 (Fig. 16b ). The 4-methyl-2-pyridinyl group packs against Phe96 ECL1 and Leu99 ECL1 . Fig. 16 a Schematic representation of allosteric agonist AA-14 bound to GPR101 (PDB: 6VCB). b Detailed binding modes of GPR101 bound to AA-14. c 2D structure of small-molecule allosteric ligand AA-14 presented for clarity

3) Inside 7TMD:

GPR52–c17, MRGPRX1–ML382, PAR2–AZ8838, LHCGR–Org43553, M2R–LY2119620, M4R–LY2119620, M4R–compound-110, TSHR–ML109, CaSR–cinacalcet, CaSR–evocalcet, CaSR–NPS-2143, CaSR–R-568, mGluR1–FITM, SMO–vismodegib, mGluR2–JNJ-40411813, mGluR2–NAM563, mGluR2–NAM597, and CCR5–maraviroc structures

This allosteric site is located in an extracellular pocket surrounded by a 7TM helical bundle, directly above the traditional orthosteric site of family A and B GPCRs and the cholesterol-binding site of the SMO receptor (Fig. 13 ). 202 , 203 Until now, this allosteric site has been the most frequently targeted binding site for drug-like allosteric modulators, mainly because allosteric modulators can enter from the extracellular region, allowing ligand binding without the need to penetrate the membrane. 204 For these receptors, the pocket in the extracellular vestibule can be partitioned into two subpockets, namely the orthosteric and allosteric pockets. The N-terminal group and ECL2 regulate the sizes of the two sub-pockets by pushing the ligand to one side, 205 – 207 thereby contributing to the creation of a new ligand pocket. As prototypical class A GPCRs, muscarinic M1–M5 acetylcholine receptors (mAChRs) are responsible for the release of acetylcholine into the brain and play fundamental roles in the central and peripheral nervous system. 208 – 210 Muscarinic receptors have garnered attention as potential drug targets to treat several pathophysiological disorders including Alzheimer’s disease, schizophrenia, and drug addiction. 211 – 214 LY2119620 ( 20 ) acts as a PAM that has activity at both the M2 and M4 receptors (Fig. 14 ) but is inappropriate for treatment, probably because of cross-reactivity and cardiovascular liability. 215 The structures of the M2 and M4 receptors bound to PAM LY2119620 have been solved. 216 – 218 LY2119620 demonstrated a similar binding pattern to both the M2 and M4 receptors; nonetheless, subtle differences were noted (Fig. 17a ). LY2119620 binds to a spacious extracellular vestibule just above the orthosteric pocket and is segregated from the orthosteric pocket via three tyrosine residues: Tyr 3.33 , Tyr 6.51 , and Tyr 7.39 . The thienopyridine ring of LY2119620 is sandwiched by π–π stacking between Tyr177 ECL2 and Trp422 7.35 in M2 receptor (Fig. 17b ), Phe186 ECL2 and Trp435 7.35 in M4 receptor (Fig. 17c ). Particularly, in the M2 receptor, the residues Tyr80 2.61 , Asn410 6.58 , and Asn419 ECL3 formed hydrogen bonds with the modulator, and Glu172 ECL2 participated in ionic interactions with piperidine. Contrarily, in the M4 receptor, only Gln427 ECL3 formed a hydrogen bond with the modulator. Fig. 17 a Superposition of PAM LY2119620 bound to M2 receptor (pink cartoon, pink sticks; PDB: 4MQT) and M4 receptor (yellow cartoon, yellow sticks; PDB: 7V68). b Detailed binding modes of M2 receptor bound to LY2119620. c Detailed binding modes of M4 receptor bound to LY2119620. Hydrogen bonds are presented as orange dashes and π–π stacking is presented as gray dashes. d Superimposed views of highlighted residue Trp422 7.36 on M2 receptor–iperoxo–LY2119620 (pink cartoon, pink sticks; PDB: 6U1N), M2 receptor–LY2119620 (purple cartoon, purple sticks; PDB: 4MQT), and M2 receptor–iperoxo (blue cartoon, blue sticks; PDB: 4MQS) structures. The orthosteric agonist iperoxo is presented in orange In addition, the structures of M2 receptor–iperoxo–LY2119620 (PDB: 4MQT) and M2 receptor–LY2119620 (PDB: 6U1N) are highly similar, with Trp422 7.35 perpendicular to the horizontal plane and forming a π–π stacking with LY2119620 (Fig. 17d ). In contrast, Trp422 7.35 of the M2 receptor–iperoxo (PDB:4MQS), exhibits a parallel conformation, suggesting that the allosteric binding site is formed predominantly in the presence of an allosteric modulator. MD simulations have revealed that LY2119620 modulates the conformation of Trp422 7.35 , causing reorientation of Tyr426 7.39 within the orthosteric site. 219 This reorientation may explain the observed increase in affinity for iperoxo, thereby providing insight into the underlying allosteric mechanism. 220 For GPCRs that use other sites to bind endogenous ligands, the traditional orthosteric pocket is potentially druggable for allosteric modulators. 221 Calcium-sensing receptors (CaSR), members of the family C GPCR, are found primarily in the parathyroid glands and kidneys to ensure strict control of calcium homeostasis. 222 , 223 Elevated Ca 2+ levels trigger the activation of CaSR, leading to the inhibition of parathyroid hormone (PTH) secretion. Thus, CaSR has become a potential target for calcimimetic drugs to treat parathyroid disorders. 224 – 226 Cinacalcet ( 27 ) (Fig. 14 ), an orally active allosteric agonist of CaSR, has been used for the treatment of secondary hyperparathyroidism 227 , 228 whereas calcilytic NPS-2143 ( 29 ) (Fig. 14 ) is a potent NAM-targeting CaSR that exhibits favorable in vitro and in vivo activity. 229 When bound to CaSR, PAM cinacalcet adopted extended and bent poses between CaSR homodimers (Fig. 18a ). The naphthylethylamine moiety was bound to highly similar poses in both the extended and bent conformations (Fig. 18b, c ). The naphthyl group engaged in hydrophobic interactions with Ile777 5.44 on one side and formed edge-to-face π–π interactions with Phe684 3.36 and Trp818 6.50 on the other, thereby effectively securing the side chain of Trp818 6.50 inside 7TM helical bundle. The NH group formed a hydrogen bond with Gln681 3.33 . In the extended conformation (Fig. 18b ), the linker and phenyl group were parallel to TM VI, extending upward and driving Tyr825 6.57 to orient downward. In the bent conformation (Fig. 18c ), the phenyl group folded between TM V and TM VI to form a parallel-displaced π–π stacking with the naphthyl group, whereas Tyr825 6.57 assumed a conformation perpendicular to TM VI and stabilized the ligand through a σ–π interaction. Fig. 18 a Schematic representation of PAM cinacalcet bound to CaSR (PDB: 7M3F). b Detailed binding modes of CaSR bound to cinacalcet in extended conformations. c Detailed binding modes of CaSR bound to cinacalcet in bent conformations. Hydrogen bonds are presented as orange dashes and π–π stackings are presented as gray dashes In the CaSR–NPS-2143 complex, NPS-2143 exhibited the same crescent conformation as the homodimers (Fig. 19a ). The naphthyl group at one end of NPS-2143 (Fig. 19c ) was lined by residues Phe684 3.36 , Leu776 5.43 , Ile777 5.44 , Trp818 6.50 , and Ile841 7.37 in the interior of the pocket (Fig. 19b ). Conversely, the 3-chloro-2-cyano-phenyl ring of NPS-2143 protrudes out toward the lateral opening and forms hydrophobic contacts with Leu773 5.40 and π–π stacking interactions with Tyr825 6.57 . A single hydrogen bond was established between the hydroxyl group and Tyr825 6.57 . Moreover, the conformation of the NAM-bound CaSR agrees well under both active (in the presence of Ca 2+ and L-Trp) and inactive (no Ca 2+ ) conditions. Fig. 19 a Schematic representation of NAM NPS-2143 bound to CaSR (PDB: 7M3E). b Detailed binding modes of CaSR bound to NPS-2143. Hydrogen bond is presented as orange dashes and π–π stacking is presented as gray dashes. c 2D structure of small-molecule allosteric ligand NPS-2143 provided for clarity. d Superimposed views of highlighted residues on CaSR–Cinacalcet–Ca 2+ –Trp (blue cartoon, blue sticks; PDB: 7M3F), CaSR–Ca 2+ –Trp (pink cartoon, pink sticks; PDB: 7DD6), and CaSR–NPS-2143–Ca 2+ –Trp (purple cartoon, purple sticks; PDB: 7M3E) structures. e Workflow of discovery of novel CaSR PAMs utilizing structural information Despite having highly similar binding sites, NPS-2143 and cinacalcet exhibit quite different pharmacological properties, which may be explained by the conformation of the receptor residues. The conformations of NPS-2143-bound and Ca 2+ -bound CaSR were similar. 230 Nevertheless, cinacalcet binding induced significant conformational changes in Trp818 6.50 , Phe821 6.53 , and Tyr825 6.57 within the allosteric pocket (Fig. 19d ). Trp818 6.50 rotates inwardly from a vertical to a horizontal conformation, forming extensive π–π interactions with cinacalcet. Phe821 6.53 underwent an outward shift and was inserted into a crevice between TM6 and TM7 facing the dimer interfacial area. Simultaneously, Tyr825 6.57 flips down, driven by structural conflicts in the extended conformation of cinacalcet. In summary, cinacalcet induces a bent conformation of TM6 and stabilizes the homodimer interface, thereby contributing to receptor activation. 231 Contrarily, NPS-2143 decreased agonist efficacy by enhancing TM VI helicity, which spatially hindered receptor activation. Based on the special binding conformations of cinacalcet, Liu et al. conducted a virtual screening of 1.2 billion compounds to discover novel PAMs with potentially novel pharmacology. To respectively mimic the “extended” and “bent” conformation, extensive orientations and conformations of library molecules were sampled, which gave 682 trillion configurations overall and finally achieved a 3.8% and 13.6% hit rate. The hits were then optimized to a pharmacologically potent lead ( 36 ) via synergistic application of structural information, fragment hybridization, and stereochemistry separation (Fig. 19e ). 232 Such practice serves as a paradigm for its elaborate utility of solved GPCR structures and conformation sampling strategy and is generalizable in the discovery of CaSR NAMs and other allosteric modulators. Targeting of GPCR transmembrane domain (outside 7TMD) As shown by their structures, GPCRs utilize the domain outside 7TMD at the lipid interface to bind allosteric modulators. Till date, five different binding sites outside 7TMD in the transmembrane domain have been defined by their crystal structures (Table 2 ): the pocket outside helices I–III, the pocket outside helices II–IV, the pocket outside helices III–V, the pocket outside helices V–VI, and the pocket outside helices I, VI, and VII (Fig. 20a, c ). Allosteric modulator binding to these regions targets class A GPCRs. These sites are typically shallow and not as well surrounded by the 7TM helical bundle as the traditional orthosteric sites. Polar functional groups are commonly found in allosteric modulators at these sites where they anchor themselves to the pocket. Thus, hydrogen atom donor or acceptor groups exposed between the receptor and lipid bilayer are more likely to mediate the binding of such allosteric ligands. These modulators are also required to preserve their overall hydrophobic character to enter the transmembrane domain. Allosteric modulators bound to the transmembrane domain outside 7TMD appear to regulate receptor signal transduction from outside the 7TM helices in a manner that stabilizes inactive or active interaction networks or impedes or facilitates the interhelical motions required for receptor activation. 233 – 235 Table 2 Solved GPCR structures complexed with synthetic allosteric modulators bound to the transmembrane domain outside 7TMD Structure Type GPCR type GPCR Modulator Highest Phase Modulator type Number PDB code Allosteric site Refs X-ray diffraction class A P2Y 1 BPTU Pre-clinical Allosteric antagonist (37) 4XNV outside 7TMD (I–III) 242 X-ray diffraction class A PAR2 AZ3451 Pre-clinical Allosteric antagonist (38) 5NDZ outside 7TMD (II–IV) 451 X-ray diffraction class A CB1R ORG27569 Pre-clinical NAM (39) 6KQI outside 7TMD (II–IV) 257 X-ray diffraction class A CB1R ZCZ011 Pre-clinical PAM (40) 7FEE outside 7TMD (II–IV) 261 X-ray diffraction class A GPR40 compound 1 Pre-clinical Allosteric agonist (41) 5KW2 outside 7TMD (III–V) 463 X-ray diffraction class A GPR40 AP8 Pre-clinical AgoPAM (42) 5TZY outside 7TMD (III–V) 464 X-ray diffraction class A β2AR Cmpd-6FA Pre-clinical PAM (43) 6N48 outside 7TMD (III–V) 269 X-ray diffraction class A β2AR AS408 Pre-clinical NAM (44) 6OBA outside 7TMD (III–V) 465 X-ray diffraction class A C5aR1 NDT9513727 Pre-clinical Allosteric inverse agonist (45) 5O9H outside 7TMD (III–V) 466 X-ray diffraction class A C5aR1 avacopan Approved Allosteric antagonist (46) 6C1R outside 7TMD (III–V) 467 Cryo-EM class A DRD1 LY3154207 Phase 2 PAM (47) 7CKZ outside 7TMD (III–V) 277 Cryo-EM class A CXCR3 SCH546738 Pre-clinical Allosteric antagonist (48) 8HNN outside 7TMD (V–VI) 299 Cryo-EM class A A1R MIPS521 Pre-clinical PAM (49) 7LD3 outside 7TMD (I, VI, VII) 305 Cryo-EM class A GPR101 AA-14 Pre-clinical Allosteric agonist (15) 8W8S outside 7TMD (I, VI, VII) 201 Cryo-EM class C mGlu4 VU0364770 Pre-clinical PAM (50) 8JD5 outside 7TMD (I, VI, VII) 468 Cryo-EM class C mGlu4 ADX88178 Pre-clinical PAM (51) 8JD6 outside 7TMD (I, VI, VII) 468 Fig. 20 a Five allosteric binding sites in the transmembrane domain outside 7TMD of GPCRs and the corresponding small-molecule allosteric modulators. Stick models of small-molecule ligands are mapped to representative members of outside 7TMD (I–III) (P2Y 1 , PDB: 4XNV), outside 7TMD (II–IV) (CB1R, PDB: 6KQI), outside 7TMD (III and V) (C5aR1, PDB: 6C1R), outside 7TMD (V and VI) (CXCR3, PDB: 8HNN), and outside 7TMD (I, VI, and VII) (A1R, PDB: 7LD3) GPCRs. For each pocket, the number of unique modulators is indicated in boldface type, and the number of GPCRs containing the pocket is provided in parentheses. b 2D chemical structures of synthetic small-molecule allosteric ligands targeting the transmembrane domain outside 7TMD of GPCRs. c Extracellular view of the five allosteric sites

1) TM I–III:

P2Y 1 –BPTU Structure To the best of our knowledge, one small molecule allosterically targets this relatively shallow pocket. Because of the flat TM helical bundle surface and relatively narrow cavity of the binding pocket, rational drug design in this area may be challenging. In this instance, the protein target was the P2Y 1 purinergic receptor. Agonists induce the activation of the P2Y 1 receptor, leading to the potentiation of platelet aggregation that triggers platelet secretion; 236 , 237 thus, antagonists targeting the P2Y 1 receptor offer a prospective approach to treat thrombosis. 238 , 239 BPTU ( 37 ) (Fig. 20b ), a P2Y 1 antagonist, has been gaining attention as an antithrombotic treatment and is the first allosteric GPCR modulator located outside the helical bundle. 240 BPTU blocks the P2Y 1 -induced platelet aggregation with nanomolar potency and presents good selectivity for P2Y 1 receptor and highly homologous P2Y 12 receptor (P2Y 1 K i  = 75 nM, P2Y 12 K i  > 70 μM). 241 The binary complex structure of P2Y 1  − BPTU was determined and showed that the BPTU binding pocket consists mainly of residues in helices I − III (Fig. 21a ). 242 Notably, two crucial hydrogen bonds were formed between the two NH moieties of the urea group of BPTU and the main-chain carboxyl group of Leu102 2.55 (Fig. 21b ). In terms of hydrophobic interactions, the BPTU pyridyl group makes contact with the residues Ala106 2.59 and Phe119 ECL1 . Hydrophobic interactions of the tert-butyl phenyl group occur within a distinct subpocket shaped by helices II and III, including residues Leu102 2.55 , Thr103 2.56 , Met123 3.24 , Leu126 3.27 , and Gln127 3.28 . On the opposite side of the ligand, the trifluoromethoxyphenyl group participated in hydrophobic interactions with Phe62 1.43 and Phe66 1.47 . Fig. 21 a Schematic representation of the allosteric antagonist BPTU bound to P2Y 1 receptor (PDB code 4XNV). b Detailed binding modes of P2Y 1 receptor bound to BPTU. Hydrogen bonds are presented as orange dashes. c Superimposed views of the highlighted residues on 2MeSADP − P2Y 1 R − G 11 (pink cartoon, pink sticks; PDB: 7XXH) and P2Y 1 R − BPTU structures (blue cartoon, blue sticks; PDB: 4XNV) This case suggests an effective shape-complementary mechanism for allosteric ligands: bulges on 7TM helices can be utilized as anchors for fixation. In this instance, Pro105 2.58 serves as the corresponding anchor, which is conserved in 74% of non-olfactory class A GPCRs. 243 A comparison between the P2Y 1 receptor bound to the agonist 2MeSADP and the allosteric antagonist BPTU revealed that the BPTU induces a 1.4 Å shift in Tyr100 2.53 toward TM3 (Fig. 21c ), preventing the conformational change of Phe131 3.32 . 244 Thus, Phe131 3.32 interacts with Phe276 6.51 and limits the TM6 transition, which is required for the activation of class A GPCRs. 245 In addition, MD simulations have demonstrated that the binding of BPTU stabilizes the helical bundle, leading to an increase in lipid order. 246 This, in turn, stabilizes the ionic lock formed between Lys46 1.46 and Arg195 ECL2 in the inactive receptor.

2) TM II−IV:

CB1R− ORG27569 , CB1R   −   ZCZ011, and PAR2   −   AZ3451 Structures Cannabinoid receptors, which consist of two subtypes, CB1R and CB2R, are activated by neurotransmitter endocannabinoids and play key modulatory roles in synaptic transmission. 247 , 248 Among them is the most abundant GPCR in the human brain is CB1R. 249 Because of its widespread distribution and regulatory roles in various physiological functions, CB1R is considered an important target for the treatment of various central nervous system (CNS) disorders. 250 – 253 ORG27569 ( 39 ) (Fig. 20b ) is the first and most comprehensively studied NAM of CB1R. 254 , 255 However, when tested in vivo, Org27569 was not sufficiently effective in modulating the effects of orthosteric cannabinoids. 256 The co-crystal structure of CB1R and NAM ORG27569 , together with its orthosteric CP55940 agonist, has been solved. 257 ORG27569 occupied an allosteric site outside helices II and IV in the inner lobe of the phospholipid bilayer (Fig. 22a ). 258 Specifically, the chloro-indole ring of ORG27569 establishes a key aromatic interaction with the indole group of Trp241 4.50 and buries in a hydrophobic pocket surrounded by His154 2.41 and Val161 2.48 . Cys238 4.47 , Trp241 4.50 , Thr242 4.51 , and Ile245 4.54 supply hydrophobic interactions for the amide-linked piperidinylphenyl chain (Fig. 22b ). Fig. 22 a Superposition of the cocrystal structures of NAM ORG27569 (yellow, PDB: 6KQI) and PAM ZCZ011 (pink, PDB: 7FEE) with orthosteric ligand-bound CB1R. b Detailed binding modes of CB1R binding to ORG27569 . c Detailed binding modes of CB1R binding to ZCZ011. Hydrogen bond is presented as orange dashes, and π–π stacking is presented as gray dashes. d Superimposed views of highlighted residues on CB1R–CP55940– ORG27569 (pink cartoon, pink sticks; PDB: 6KQI) and CB1R–AMG315–G i structures (blue cartoon, blue sticks; PDB: 8GHV). e Superimposed views of highlighted residues on CB1R–AM6538 (pink cartoon, pink sticks; PDB: 5TGZ) and CB1R-CP55940–ZCZ011 structures (blue cartoon, blue sticks; PDB: 7FEE) The recently developed ZCZ011 (Fig. 20b ) is also an indole derivative that exerts PAM and partial agonist effects in vivo assays. 259 , 260 ZCZ011 showed good shape complementarity with the pocket outside 7TMD comprising helices II−IV (Fig. 22a ). The indole group in ZCZ011 is anchored by Phe191 3.27 and forms a hydrogen bond with the main chain of Phe191 3.27 while forming π−π stacking interactions with the side chain. (Fig. 22c ). 261 In addition, the indole group interacted with Leu165 2.52 , Ile169 2.56 , Ile245 4.54 , and Val249 4.58 from TM II and TM IV, respectively. NAM ORG27569 and AZ3451 ( 38 ), as well as PAM ZCZ011, bind to the same TM II−IV surface, however, exert opposite allosteric effects. 262 Unlike traditional NAMs, ORG27569 enhances the affinity of the agonist though reduces the activity of G i turnover. 263 Structurally, the inactivating efficacy of ORG27569 acts by stabilizing the “activation switch” formed by Phe155 2.42 and Phe237 4.46 in CB1R (Fig. 22d ). 264 , 265 Furthermore, a hydrogen bond was formed between ORG27569 and Trp241 4.50 , which, together with the hydrogen bond formed between Trp241 4.50 , Ser158 2.45 , and Ser206 3.42 (Fig. 22b ), created a polar network that contributed to maintaining the inactive conformation. 266 However, the precise mechanism through which ORG27569 augments agonist affinity remains to be elucidated. Upon comparing the CB1 structures bound to the PAM ZCZ011 and the antagonist AM6538, a notable shift of Ile169 2.56 in TM2 towards TM3 was observed. This shift results in the contraction of the receptor“s active site (Fig. 22e ), and is believed to be associated with activation. 267 , 268 Additionally, Ser173 2.60 undergoes notable inward movement and forms a hydrogen bond with CP55940, thereby stabilizing the agonist binding.

3) TM III–V:

GPR40–compound 1, GPR40–AP8, β2AR–AS408, β2AR–Cmpd-6FA, C5aR1–NDT9513727, C5aR1–avacopan, and DRD1–LY3154207 structures

A deep pocket is present outside the transmembrane helices III–V among GPCRs, which allows for the presence of a population of allosteric regulators bound to this pocket. In this case, allosteric agonists and PAMs (i.e., compound 1 ( 41 ), AP8 ( 42 ), Cmpd-6FA ( 43 ), and LY3154207 ( 47 )) localize to regions near ICL2 and stabilize the ICL2 α helix through direct interactions, facilitating the inward movement of Pro ICL2 . This movement leads to a ~3° inward displacement of TM III, which in turn determines the outward shift of TM V together with TM VI, which is a hallmark of GPCR activation. 269 , 270 Thus, the binding of allosteric modulators increases the proportion of receptors that adopt active conformations, thereby increasing their affinity for agonists. Contrarily, allosteric antagonists and NAMs (i.e., AS408 ( 44 ), NDT9513727 ( 45 ), and avacopan ( 46 )) bind to regions far from ICL2. As these ligands are bound close to the proline kink of TM V, the receptor-ligand interactions collectively inhibit the interhelical movements and rotations within TM III, TM IV, and TM V, which are required for receptor activation. 271 The dopamine D1 receptor is a prototypical example. Dopamine functions as an essential catecholamine neurotransmitter that signals via the dopamine D1 to D5 receptors. 272 , 273 The dopamine D1 receptor (DRD1) regulates neuronal growth, memory, and learning in the central nervous system. 274 , 275 LY3154207 (Fig. 20b ) is a selective PAM of the dopamine D1 receptor that improved motor symptoms associated with Lewy body dementia in a 2022 phase 2 clinical trial. 276 Two distinct binding modes of LY3154207 to DRD1 have been reported, and structural comparisons have demonstrated upright and boat conformations (Fig. 23a ). 277 – 280 In both binding modes, LY3154207 was localized at the receptor–lipid bilayer interface surrounded by TM III, TM IV, and ICL2. Fig. 23 a Superposition of the cocrystal structures of PAM LY3154207 bound to dopamine D1 receptor in upright (pink, PDB: 7CKZ) and boat (yellow, PDB: 7LJC and 7X2F) conformations. b Detailed binding modes of dopamine D1 receptor binding to LY3154207 in upright conformations. c Detailed binding modes of dopamine D1 receptor binding to LY3154207 in boat conformations. Hydrogen bonds are presented as orange dashes; π–π stacking and π–cation stacking interactions are presented as gray dashes In the upright conformation, the tertiary alcohol group of LY3154207 extends toward TM III, establishing a single hydrogen bond with Cys115 3.44 (Fig. 23b ). The central tetrahydroisoquinoline ring of LY3154207 engages in extensive hydrophobic interactions with Val119 3.48 , Trp123 3.52 , and Leu143 4.45 . Additionally, the dichlorophenyl group forms a π-cation interaction with the side chain of Arg130 ICL2 . In the boat conformation, the dichlorophenyl group participates in π–cation interactions with the side chains of Arg130 ICL2 , and interacts sandwich-like π–π stacking with Trp123 3.52 (Fig. 23c ). The tetrahydroisoquinoline ring of LY3154207 established hydrophobic interactions with neighboring residues, including Met135 ICL2 , Ala139 4.41 , Ile142 4.44 , Leu143 4.45 , and the alkyl chain of Lys134 ICL2 . In addition, two hydrogen bonds were observed between LY3154207 and the polar residues Arg130 ICL2 and Lys138 4.40 . Superposition of the D1R–LY3154207–dopamine and D1R–dopamine structures showed near-identical binding positions for dopamine and the surrounding residues. Instead, LY3154207 interacted with Arg130 ICL2 and Lys134 ICL2 and stabilized ICL2, which interacts directly with G-proteins, thereby increasing the population of D1R adopting active conformations (Fig. 23b, c ). 279 Another example of an allosteric modulator that binds outside 7TMD formed by helices III–V is the allosteric antagonist avacopan of C5a receptor 1. Human C5a receptor 1 (C5aR1), which binds to the pro-inflammatory mediator C5a, is primarily expressed on the surfaces of various immune cells, such as neutrophils, eosinophils, and dendritic cells. 281 , 282 Overactivation of the C5aR1-C5a axis will cause uncontrolled inflammation; 283 thus, C5aR1 antagonists are ideal candidates for treating various inflammatory conditions, including sepsis COVID-19, etc. 284 – 286 Avacopan (Fig. 20b ) is an orally administered allosteric antagonist of C5aR1 approved by the FDA in 2021 to treat severe autoantibody (ANCA)-ANCA-associated vasculitis (granulomatosis with polyangiitis and microscopic polyangiitis). 287 , 288 Pharmacological studies have indicated the ability of avacopan for biased inhibition of β-arrestin coupling. 289 – 291 The co-crystal structure of C5aR1 was reported with avacopan, highlighting the binding site outside 7TMD between helices III and V (Fig. 24a ). The cyclopentane group of avacopan extended into the crevice between helices III and IV and occupied the hydrophobic pocket consisting of Leu125 3.41 , Val159 4.48 , Leu163 4.52 , and Leu167 4.56 (Fig. 24b ). The o-methyltrifluoromethylbenzene group exhibited hydrophobic and aromatic interactions mediated by residues Ile124 3.40 , Leu125 3.41 , Leu209 5.45 , Trp213 5.49 , Pro214 5.50 , and Leu218 5.45 in the binding cleft between helices III and V. The m-methylfluorobenzene group lies deeper and forms non-polar interactions with residues Phe135 3.52 , Ile220 5.56 , Cys221 5.57 , and Phe224 5.60 in C5aR1. Only one hydrogen bond was observed between the carbonyl substituent of the amide bond of avacopan and Trp213 5.49 . Additionally, there is a water-mediated polar interaction between avacopan and Thr217 5.53 . Fig. 24 a Schematic representation of the allosteric antagonist avacopan bound to C5a receptor 1 (PDB: 6C1R). b Detailed binding modes of C5a receptor 1 bound to avacopan. Hydrogen bond is presented as orange dashes. c Superimposed views of highlighted residue Trp213 5.49 on C5aR1–PMX53–avacopan (pink cartoon, pink sticks; PDB: 6C1R) and C5aR1–C5a–G o (blue cartoon, blue sticks; PDB: 8IA2) structures In the inactive C5aR1 structure, Trp213 5.49 undergoes a conformational transition to accommodate avacopan binding, in contrast to the active C5aR1 structure (Fig. 24c ). Avacopan may stabilize the conformation of the residues Ile124 3.40 , Pro214 5.50 , and Phe251 6.44 in their inactive states through direct hydrophobic interactions (Fig. 24b ). This stabilization, in turn, hinders conformational changes in transmembrane helices TM5 and TM6, which are necessary for receptor activation. 292 , 293

4) TM V–VI:

CXCR3–SCH546738 structure C-X-C chemokine receptor type 3 (CXCR3), a class A GPCR, is highly expressed on effector T cells and is activated by chemokines CXCL9, CXCL10 and CXCL11. 294 Due to the critical role of CXCR3 in type 1 immunity, agonists and antagonists targeting CXCR3 have been synthesized to treat infection, autoimmune diseases, allograft rejection and cancers. 295 – 297 Among these, SCH546738 ( 48 ) (Fig. 25c ) has shown remarkable efficacy in several preclinical trials by effectively inhibiting the activation of T cell chemotaxis with an affinity of 0.4 nM. 298 Fig. 25 a Schematic representation of the allosteric antagonist SCH546738 bound to CXCR3 (PDB code 8HNN). b Detailed binding modes of CXCR3 binding to SCH546738. c 2D structure of small-molecule allosteric ligand SCH546738 provided for clarity In the CXCR3–SCH546738 structure, SCH546738 is trapped in a narrow hydrophobic pocket surrounded by TM3, TM5 and TM6 (Fig. 25a ). 299 The head of SCH546738 is surrounded by a hydrophobic pocket formed by residues Phe135 3.36 , Ala139 3.40 , Phe224 5.47 , Leu228 5.51 , Met231 5.54 , Ile261 6.41 , Ala273 6.53 (Fig. 25b ). The tail of SCH546738 stretches out to the lipid bilayer and interacts with Tyr235 5.58 , Leu239 5.62 and Leu258 6.38 . Given the unique allosteric site of SCH546738, the interposition of SCH546738 may weaken the repacking between TM5-TM6, maintaining the receptor in an inactive state.

5) TM I, VI, VII:

A1R–MIPS521, GPR101–AA-14, mGlu4–VU0364770, and mGlu4– ADX88178 Structures

The adenosine A1 receptor (A1R), a subtype of the adenosine receptor, 300 has been a highly pursued non-opioid analgesic target for the treat chronic pain. 301 – 304 Nonetheless, no selective clinically approved A1R agonists or antagonists are currently available. MIPS521 ( 49 ) (Fig. 20b ), a PAM of A1R, suppressed spinal nociceptive signaling and displayed an analgesic effect in a rat model with a pEC50 of 6.9 ± 0.4. 305 Structural analysis revealed that MIPS521 binds outside 7TMD surrounded by helices I, VI, and VII (Fig. 26a ). Residues Leu18 1.41 , Ile19 1.42 , Val22 1.45 , Leu242 6.43 , Leu245 6.46 , Ser246 6.47 , Phe275 7.40 , Leu276 7.41 , and Met283 7.48 formed shallow hydrophobic pockets at the allosteric site (Fig. 26b ). The amino group of MIPS521 (Fig. 26c ) was hydrogen-bonded to the main-chain carbonyl groups in Ser246 6.47 and Leu276 7.41 . Comparing the ADO–A1R–G i2 and MIPS521–ADO–A1R–G i2 structures showed that MIPS521 has minimal impact on the binding position of ADO and receptor conformations. Mechanistically, the binding of MIPS521 may stabilize the active conformation by interacting with the allosteric site, which, in turn, promotes the collapse of the Na + pocket (a nearby conserved class A activation motif). 245 , 306 Fig. 26 a Schematic representation of PAM MIPS521 bound to A1R (PDB code 7LD3). b Detailed binding modes of A1R binding to MIPS521. Hydrogen bonds are presented as orange dashes. c 2D structure of small-molecule allosteric ligand MIPS521 provided for clarity Targeting of GPCR transmembrane domain (inside 7TMD) An empty pocket is present in the middle of the GPCR transmembrane domain that serves as an allosteric site inside 7TMD, as shown in Fig. 27 . Allosteric modulators bound to this pocket primarily interact with helices other than TM I and TM IV. To date, a PAM of FFAR3, an allosteric antagonist of CRF1R, an allosteric agonist of PTH1R, and five NAM and one PAM of mGluR5 have been reported to bind to this region (Table 3 ). In contrast, the binding of NAMs and an allosteric antagonist blocks the outward movement of TM VI, thereby acting as an antagonist. Notably, the binding of only the allosteric agonist stabilizes the G protein through direct interactions. Fig. 27 Allosteric binding sites in the transmembrane domain within 7TMD of GPCRs and corresponding small-molecule allosteric modulators. Stick models of small-molecule ligands are mapped to representative member CRF1R, PDB: 4K5Y. The number of unique modulators is indicated in boldface type, and the number of GPCRs containing the pocket is provided in parentheses Table 3 Solved GPCR structures complexed with synthetic allosteric modulators bound to the transmembrane domain inside 7TMD Structure Type GPCR type GPCR Modulator Highest Phase Modulator type Number PDB code Allosteric site Refs Cryo-EM class A FFAR3 AR420626 Pre-clinical PAM (52) 8J20 inside 7TMD 469 X-ray diffraction class B CRF1R CP-376395 Pre-clinical Allosteric antagonist (53) 4K5Y inside 7TMD 470 Cryo-EM class B PTH1R PCO371 Pre-clinical Allosteric agonist (54) 8GW8 inside 7TMD 335 X-ray diffraction class C mGluR5 mavoglurant Phase 3 NAM (55) 4OO9 inside 7TMD 178 X-ray diffraction class C mGluR5 compound 14 Pre-clinical NAM (56) 5CGC inside 7TMD 321 X-ray diffraction class C mGluR5 HTL14242 Phase 1 NAM (57) 5CGD inside 7TMD 321 X-ray diffraction class C mGluR5 Fenobam Phase 1 NAM (58) 6FFH inside 7TMD 471 X-ray diffraction class C mGluR5 M-MPEP Pre-clinical NAM (59) 6FFI inside 7TMD 471 Cryo-EM class C mGluR5 CDPPB Pre-clinical PAM (60) 8TAO inside 7TMD 325

1) Inside 7TMD:

FFAR3– AR420626 , CRF1R–CP-376395, mGluR5–mavoglurant, mGluR5–compound 14, mGluR5–HTL14242, mGluR5–Fenobam, mGluR5–M-MPEP, and PTH1R–PCO371 structures

The metabotropic glutamate (mGlu) receptor type 5 is among the eight most widely expressed mGlu receptors in the brain. 307 Recently, mGluR5 has been shown to be involved in a growing number of cognitive and psychiatric di