Molecular pharmacology of metabotropic receptors targeted by neuropsychiatric drugs

Bryan L. Roth

Metabotropic receptors are responsible for so-called ‘slow synaptic transmission’ and mediate the effects of hundreds of pep- tide and non-peptide neurotransmitters and neuromodulators. Over the past decade or so, a revolution in membrane-protein structural determination has clarified the molecular determinants responsible for the actions of these receptors. This Review focuses on the G protein–coupled receptors (GPCRs) that are targets of neuropsychiatric drugs and shows how insights into the structure and function of these important synaptic proteins are accelerating understanding of their actions. Notably, elucidat- ing the structure and function of GPCRs should enhance the structure-guided discovery of novel chemical tools with which to manipulate and understand these synaptic proteins.

eurotransmitter receptors are essential for mediating the GPCRs, which have complex actions mediated by second messen- effects of neurotransmitters in the brain and peripheral gers and protein kinases21,22 (Fig. 1). Nnervous system. There are generally considered to be two This Review will discuss how structural insights into neurotrans- types of neurotransmitter receptors: ionotropic and metabotropic. mitter GPCRs have transformed the understanding of these recep- While ionotropic receptors are typically ligand-gated ion chan- tors, focusing on example GPCRs that are targets of drugs approved nels, through which ions pass in response to a neurotransmitter, for use in the treatment of neuropsychiatric disorders. The past metabotropic receptors need G proteins and second messengers to decade has witnessed remarkable progress in elucidating the struc- indirectly modulate ionic activity in neurons. G protein–coupled ture and function of GPCR family, with perhaps 60 or so GPCRs receptors (GPCRs) represent the largest family of metabotropic having their structures determined by X-ray crystallography7 or receptors, although receptor tyrosine kinases1 and guanylate cyclase cryo-EM7. Additionally, various intermediate signaling states, receptors2,3 can also be considered metabotropic receptors4. GPCRs ranging from inactive to active, have been elucidated for exem- also constitute the largest family of druggable targets5,6 encoded by plar GPCRs6. Finally, progress has been made on the use of such the human genome, and 34% of medications approved by the US structural information for structure-guided and structure-inspired Food and Drug Administration have GPCRs as their main thera- neuropsychiatric drug discovery8. This Review will provide an over- peutic target7,8, especially those used for the treatment of neuropsy- view of the understanding of how structure informs the function chiatric disorders6,9,10 (Table 1). of representative neurotransmitter GPCRs. It will also show how There are many distinct chemical classes of neurotransmitters, an understanding of structure elucidates neuropharmacology and including the following: (1) small molecules (e.g., glutamate, nor- will highlight therapeutic challenges and opportunities for neuro­ epinephrine and serotonin); (2) neuropeptides (e.g., enkephalins, transmitter-targeted GPCRs. endorphins and neurokinins); and (3) others, including metabolites (e.g., endocannabinoids, ATP, ADP and adenosine) and gases (e.g., Structural genomics of neurotransmitter-targeted GPCRs nitric oxide). Although the precise number of brain neurotransmit- Structural genomics has been defined as the approach that proposes ters is not known with certainty, it is likely that hundreds of distinct ‘determining protein structures on a genome-wide scale’23, which neurotransmitters exist, including many ‘orphan’ neuropeptides11. boils down to ultimately determining the three-dimensional struc- Each neurotransmitter system has distinct cellular and region-spe- ture of every protein encoded by the human genome. Obtaining cific distribution patterns in the brain; these can be visualized at the GPCR structures has been historically challenging. Thus, although mRNA level with the Allen Brain Atlas12 or via genetically encoded low-resolution structures of rhodopsin were available as early as markers with the GENSAT (Gene Expression Nervous System 1993 (ref. 24), and useful models of helical arrangements were gen- Atlas) resource13. Almost all neurotransmitters transduce their sig- erated through the use of that information25, the first bona fide nals at least in part by activating GPCRs, and the regional and cellu- high-resolution GPCR structure was not achieved until 2000, with lar distributions of brain GPCR mRNAs are well described12,14. As is rhodopsin26. The first non-opsin GPCR structures were obtained the case with neurotransmitters, each of the metabotropic receptors in 2007 (refs. 27–29), and since then, structures have been obtained, for neurotransmitters has a distinct regional and cellular distribu- mainly via X-ray crystallography, for around 60 distinct human tion in the brain12,14 and elsewhere14. GPCRs7 (Fig. 2a). To generate useful structures, GPCR crystallog- GPCRs modulate synaptic transmission via so-called ‘slow raphy typically requires high-affinity ligands, which serve to stabi- synaptic transmission’15, which occurs in a time frame of seconds lize the receptor in a distinct state suitable for crystallography. Such to minutes in the central nervous system. Metabotropic receptors compounds not only need to be of high affinity (e.g., in the low such as GPCRs may be found pre- and post-synaptically. Pre-synaptic nanomolar to picomolar range) but also need to have slow dissocia- 30 Gi-coupled GPCRs, such as the 5-HT1A and 5-HT1B serotonin tion rates . Ideally, then, to obtain complete structural coverage of receptors16,17, inhibit neurotransmitter release via Gβ/γ-mediated all GPCRs encoded by the human genome, one would need high- activation of inhibitory channels, such as G protein–coupled affinity, slowly dissociating ligands for each of them. Such ligands inwardly rectifying potassium channels18, and via the inhibition of do not necessarily need to be specific, as, for example, the non- vesicle-docking SNARE-like proteins19,20 (Fig. 1). Post-synaptically, selective drug ergotamine has been used to obtain the structures of 31 GPCRs can enhance neuronal excitability via Gs- and Gq-coupled three different serotonin receptor subtypes: 5-HT1B (ref. , 5-HT2B

Department of Pharmacology, University of North Carolina Chapel Hill, School of Medicine, Chapel Hill 27514 NC, USA. e-mail: [email protected] Table 1 | Representative metabotropic neurotransmitter receptors as targets for neuropsychiatric disorders and drugs of abuse T arget class Neurotransmitter Receptor Representative Therapeutic indications or Pharmacological Structure medication abuse action determined (ref.) GPCR Dopamine D2 dopamine Ropinirole Parkinson’s Disease Agonist 75 GPCR Dopamine D2 dopamine Nemonapride Schizophrenia Antagonist/ 75 inverse agonist

GPCR Serotonin 5-HT2C serotonin Lorcaserin Obesity Agonist 33

GPCR Serotonin 5-HT1A serotonin Buspirone Anxiety and depression Partial agonist None reported

GPCR Serotonin 5-HT1B Ergotamine Migraine headaches Antagonist 31,32

GPCR Adenosine A2A adenosine Caffeine Alertness Antagonist 131

GPCR Norepinephrine β1 and β2 Propranolol Post-traumatic stress Inverse agonist 132,133 adrenergic disorder; anxiety disorders

GPCR Serotonin 5-HT2A serotonin Pimavanserin Psychosis related to Inverse agonist 134 Parkinson’s Disease GPCR Serotonin Many 5-HT LSD Hallucinogen use and Agonist 70 receptors abuse GPCR Endocannabinoids CB1 cannabinoid Tetrahydrocannibinol Anorexia related to cancer Agonist 135 chemotherapy; cannabis abuse GPCR GABA GABA-B Baclofen Spasticity-related Agonist None reported movement disorders GPCR Histamine H1-histamine Diphenhydramine Sedative Antagonist 41 GPCR Endorphins; μ-opioid Morphine Pain; abused opioid Agonist 59,136 endomorphans GPCR Dynorphin κ-opioid Nalfurafine Itch Agonist 52,137 GPCR Dynorphin κ-opioid Salvinorin A Hallucinogen use and Agonist 52,137 abuse GPCR Orexin OX1 orexin Suvorexant Insomnia Antagonist 138 GPCR Norepinephrine α2 adrenergic Clonidine Anxiety; opioid withdrawal Agonist None reported GPCR S1P1 sphingosine Fingolimod Multiple sclerosis Downregulates 139 antagonist receptor Receptor Nerve growth TRK Entrectinib Cancers including Antagonist No full-length tyrosine factors astrocytoma structure kinase reported Guanylate Atrial natriuretic NPRA atrial Atrial natriuretic None (heart failure) Agonist 140 cyclases peptide natriuretic receptor peptide

GABA, γ-amino butyric acid.

32 33 40 (ref. ) and 5-HT2C (ref. ). Once structures are obtained, even with ligand-free state has been solved —those GPCRs for which there non-selective ligands, the structural features for distinct pharmaco- are reported structures typically have hundreds to thousands of logical properties can be elucidated (as in refs. 32,33). Furthermore, annotated ligands (Fig. 2a). No clear relationship exists between obtaining structures of related GPCRs frequently facilitates the the number of annotated ligands in Chembl and the number structure-guided discovery and design of subtype-selective drugs34. of distinct structures (Fig. 2b), although, as mentioned above, An analysis of available GPCR structures and cognate ligands, high-affinity, slowly dissociating ligands are typically key for done with open-source databases of small molecules that bind obtaining structures. Thus, comprehensive elucidation of the known protein targets, along with databases of GPCR structural structural genomics of GPCRs will require that suitable ligands information, can be the first step toward determining the feasibil- be obtained for a large number of these receptors. Efforts are now ity of a comprehensive structural elucidation of the ‘GPCR-ome’. ongoing to obtain these ligands, including the ‘Illuminating the Chembl35, the KiDatabase36 and PubChem37 are databases that Druggable Genome’ initiative (https://commonfund.nih.gov/idg), match compounds and their targets. Figure 2a shows the number which has designated those GPCRs for which there is little chemical of ligands with Chembl-annotated activity, plotted against a total matter a top priority. of 393 non-olfactory GPCRs and, simultaneously, the number Taking into account the wealth of data noted above, and with of structures available for each of these receptors (as of February new GPCR structures now appearing every week or so, what are 2019). Similar to previously published analyses6–8,38,39, in this analy- the opportunities and challenges for the completion of a structural sis, around half of all human non-olfactory GPCRs are well anno- genomics survey for neurotransmitter GPCRs? Even for the 240 tated with respect to chemical matter, while the rest have few to no class A (Rhodopsin-like) GPCRs, which comprise most neurotrans- compounds associated with them in openly available databases35. mitter receptors and for which there is the most structural cover- With one exception—the receptor FZD4 (Frizzled 4), for which age, fewer than 50 distinct members have elucidated structures. the structure of the seven-transmembrane domain (TM7) in the Several classes — B2 (Adhesion), C (Glutamate), F (Frizzled) and T (Tastant 2), for example — have two reported structures (Glutamate and Frizzled) to no reported structures (Adhesion and Tastant). Even among the thoroughly structurally annotated families of neurotransmitter receptors, such as the biogenic amine receptor

+ family, structural annotation is still very sparse. Thus, only one (the K 41 Diminished excitability H1 histamine receptor) of four histamine receptors has a reported Inhibition of structure, and there are no publicly available structures for α1 or neurotransmitter release γ + 17 β K α2 adrenergic receptors. Finally, among 14 serotonin receptors , 31–33 only 3 (5-HT1B, 5-HT2B and 5-HT2C) have published structures.

Gi Thus, considerable opportunities exist for obtaining structures of γ β representative members of each neurotransmitter GPCR subfam- SNAREs ily. Additionally, even for neurotransmitter receptor families such as the muscarinic receptor family, which has acetylcholine as its neu- rotransmitter and for which four (M1, M2, M3 and M4)42–44 of five Na+ members have structures reported, the structural determinants for subtype selectivity are still to be elucidated and fully exploited for the creation of selective muscarinic agonists and antagonists45. For P Gq G s Na+ muscarinic receptors, in particular, it is clear that drugs targeting PKA M1 and M4 muscarinic receptors may have special utility for the cAMP Ca++ enhancement of cognition in Alzheimer’s disease and schizophre- nia, while interactions with M3 muscarinic receptors are associated Enhanced excitability 46,47 Increased neurotransmitter release with debilitating side effects . In terms of creating subtype-selec- tive drugs, a variety of structure-guided approaches can be used, Fig. 1 | Metabotropic receptors such as GPCRs modulate synaptic including targeting allosteric and orthosteric sites (as discussed transmission. This diagram shows several of the mechanisms by which in refs. 6,8,48). As the foregoing makes clear, although considerable metabotropic receptors such as GPCRs can modify synaptic transmission. progress has been made toward a comprehensive understanding

Gi-coupled GPCRs can attenuate pre-synaptic release by activating various of the structural genomics of human neurotransmitter GPCRs, a channels, including inhibitory GIRKs (G protein–coupled inwardly rectifying vast amount of work remains, which will require the integrated and potassium channels)18,128, and by inhibiting vesicle-release machinery, coordinated efforts of structural biologists, pharmacologists, chem- 19 including SNARE proteins . Post-synaptic Gq- and Gs-coupled GPCRs ists and computational biologists. can induce or potentiate neuronal firing via varous intracellular second messengers129 and secondary modulation of ion-channel activity21. PKA, Structural insights into activation and biased signaling protein kinase A. As with GPCR structural genomics, considerable insight into the structural features of neurotransmitter-targeted GPCR activation and biased signaling has been obtained. As for GPCR activation, a 393 Human GPCRs the first structure of a GPCR in complex with hetereotrimeric 49 150,000 Structures 50 G proteins was obtained in 2011 (ref. ), and since then, several A2A adenosine ChemBl compounds structures of nanobody-stabilized active states (e.g., adrenergic

GPCR structures ( ) GPCR structures β2 (8,384 ligands; 44 structures) 40 100,000 receptors50, M2 muscarinic receptors51 and κ- and μ-opioid recep- 52,53 30 tors ) and heterotrimeric G protein–stabilized active states (e.g., 50,000 54 β1 adrenergic 20 calcitonin receptors , CGRP (calcitonin gene–related peptide) (2,919 ligands, 20 structures) 55,56 57,58 59 receptors , the 5-HT1B serotonin receptor , μ-opioid receptors 450 FZD4 10 60 300 and the GLP-1 (glucagon-like peptide-1) receptor ) have appeared, ChemBl compounds ( ) 150 (0 ligands, 1 structure) 0 0 essentially all with neurotransmitter-targeted GPCRs. Furthermore, structures of arrestin-biased agonists at the receptor 5-HT2B

BAI2 32,61,62 63 ET-A HRH4 C3aR GLP1R FFAR1 AVPR2 SSTR1 LPAR1 GPR34 FZD10 GPR32GPR82 LPAR6 CHRM1 MCHR1CXCR6 GPR115GPR157 PAQR6 (refs. ), arrestin-bound rhodopsin and G protein–biased agonists at the GLP-1 receptor64 have been reported. Finally, there 50 b are inactive-state structures stabilized by sodium (e.g., A2A adenos- A2A adenosine 65 66 34 (8,384 ligands; 44 structures) ine receptors , δ-opioid receptors and D4 dopamine receptors ), 40 negatively modulating allosteric nanobodies at the β2 adrener- gic receptor67 and the ligand-free basal structure of the receptor 30 NTSR1 neurotensin FZD440. Figure 3 depicts representative structures of these various

ctures (522 ligands; 8 structures) 20 states, along with a graphical representation of an extended ter- St ru D2 dopamine nary complex model of receptor activation that includes complexes (9,304 ligands; 1 structure) 68 69 10 stabilized by both G protein and β-arrestin . The outward movement of TM6 is the hallmark for GPCR activation, as are 49,53 0 several other canonical rearrangements . These include rear- 0 2,000 4,000 50,000 100,000 150,000 32 ChemBL rangements of several microswitches such as the ‘P-I-F’ motif, formed by residues P5.50, I3.40 and F6.44 (Fig. 4). Here there is an Fig. 2 | The availability of chemical matter is useful for obtaining GPCR inward shift of P5.50, along with a rotamer switch of I3.40 and a large structures. a, Compounds annotated in Chembl (blue circles) plotted inward movement of F6.44. Additional microswitches that undergo against a particular human GPCR (horizontal axis) versus the number rearrangement include the D(E)/RY motif in TM3 and the NPxxY of X-ray or cryo-EM structures deposited for each particular GPCR (red motif in TM7. The salt bridge between D3.49 and R3.50 is typically squares) (effective date, 1 February 2019). b, Comparison of the number broken in active-state structures49,53. At the NPxxY motif, activation of compounds annotated and the number of structures currently available reveals a rotation of Y7.53. Additionally the sodium site collapses due shows that there is no direct linear relationship between these. to rearrangements of key residues that coordinate sodium53 (Fig. 4). Ligand-free FZD4 Sodium-bound -opioid -adrenergic-Nb60-carazolol complex δ β2 a (6BD4) (4N6H) (5JQH) Inactive sodium-bound D4 dopamine (5WIV) Ser3.39 LSD-bound arrestin-biased 5-HT2B serotonin (5TVN) Inactive Na+ conformations Ser7.45

Asp2.50

Ser7.45

R R1LR2L

LSD-bound 5-HT b UK-432097 (agonist)-bound A2A 2B (3QAK) (5TVN)

Active-like Ile3.40 conformations

Phe6.44 Inactive sodium-bound D4 dopamine (5WIV) R* + L R*L R**L LSD-bound arrestin-biased

5-HT2B serotonin (5TVN) Pro5.50 G protein active 5-HT1B serotonin (5V54) β2-adrenergic-Gs complex Rhodopsin-arrestin complex (3SN6) (5DGY)

Fig. 4 | Structural rearrangements associated with distinct GPCR states. a, Rearrangements within the sodium pocket associated with stabilization Signaling complexes of an arrestin-biased state of the 5-HT2B serotonin receptor: key residues involved in stabilizing sodium (Ser3.39 and Asp2.59), along with Ser7.45, collapse to occlude the sodium site. Green indicates the inactive sodium- stabilized state of the D4 dopamine receptor; cyan shows the LSD-bound R*G R*GL R**LβArr 5-HT2B receptor (key). b, Rearrangements in the P-I-F motif associated with inactive (green), G-protein (magenta) and arrestin-biased (cyan) Fig. 3 | Structural validation of the extended ternary complex model conformations (key). of GPCR action. The extended ternary complex model6,68,69 predicts the existence of multiple interconvertible GPCR states stabilized by ligands (agonists, neutral antagonists and inverse agonists) and transducers (G proteins, arrestins and other transducers). The various high-resolution Notably, Ser7.45 rotates in to occlude the sodium pocket (Fig. 4a). structures here have provided validation for this schema. These states Additionally, a key residue (Leu209) in extracellular loop 2 (EL2) include the following: inactive states stabilized by allosteric nanobodies seems to form a ‘lid’ over LSD, retarding its dissociation and being 67 70 66 (RL2) ; active sodium-bound (R*L) and inactive sodium-bound states essential for the recruitment of arrestin. A similar role for the same stabilized by agonists and inverse agonists, respectively; and coupled states residue in EL2 was verified in studies of the D2 dopamine receptor (R*GL and R**LβArr). The model also provides for the possibility of biased and several other biogenic amine receptors61,62,72. Notably, the P-I-F signaling due to agonist-induced stabilization of distinct active and coupled and NPXXY motifs, along with conformational rearrangements in 31,32 states (e.g., R*L and R**L). Finally, spontaneously active states (R*) and TM6 and TM7, are consistent with an intermediate activated state61 68 coupled states (R*G), which have been demonstrated in many systems , (Fig. 4b). Collectively, these rearrangements suggest that concerted are predicted to exist, although structures of these are not yet available. conformational changes involving residues throughout the recep- βArr, β-arrestin. Each state for which a structure is known is presented with tor, although they do not fully mimic the changes in the rhodop- the Protein Data Bank accession code in parentheses directly above and a sin–arrestin complex63, are responsible for the arrestin-biased description of the structure at top. Here, * and ** represent distinct active conformation that favors binding of arrestin (Fig. 4b). In contrast, and coupled states. the GLP-1 complex with the G protein–biased agonist exendin-P5 shows the most pronounced difference in TM1, along with extra­ cellular regions of TM6, TM7 and extracellular loop 3 (EL3), for Finally, there is typically a large outward movement (of 10–15 Å) which there is support by extensive mutagenesis (details in ref. 64). of the cytoplasmic end of TM6, which facilitates interactions with The inactive states typically show a conserved sodium site, with G proteins49 and other transducers53. In general, the ‘active-like’ sodium visualized in the structures of highest resolution34,65,66. receptor states in Fig. 3 show the various transitions mentioned Additionally, a conserved ‘ionic lock’ between D/E3.49 (of the D(E)/ above without the large outward movement of TM6. RY motif) and N6.30, which stabilizes the ground state of several Currently, there are two published structures of GPCRs in GPCRs73,74, is occasionally seen. Remarkably, in an inactive state complex with β-arrestin-biased agonists, including LSD70 and stabilized by Nb60 (nanobody 60) and carazolol67, this ionic lock is 32 ergotamine with the 5-HT2B serotonin receptor, and one struc- recapitulated by the nanobody. Additionally, in many inactive-state ture of a G protein–biased ligand in complex with hetereotrimeric structures, the P-I-F, NPxxY and D(E)/RY motifs are all in a gener- G protein for the GLP-1 receptor64. The β-arrestin-biased agonist ally conserved ‘inactive-like’ state. structures (Fig. 3) seem to represent an ‘intermediate’ state between The studies discussed above demonstrate clear progress in the signaling complex states and the inactive states. Notably, the the understanding of the structural features required for differ- sodium site is collapsed, with rearrangements of key residues that ent modalities of receptor activation and signaling. Clearly, much stabilize the bound sodium ion, including (with the Ballesteros– remains to be done in this direction, and the discovery of many Weinstein numbering convention71) Ser3.39 and Asp2.50 (Fig. 4a). novel permutations of these features is eagerly anticipated. abD2 dopamine D3 dopamine c D4 dopamine d 6CM4 3PBL 5WIL D4 L3.28

D4 F2.61

fgh e D2 F3.28 Docking 138 million compounds Eticlopride Structure-inspired 67 pM D2 design 160 pM D3 27 nM D4 D2 V2.61

Compound 19 ZINC621433144 6/8 nM D3 160 pM D4 >10,000 nM D2 >10,000 nM D2 >10,000 nM D3 >10,000 nM D3

Fig. 5 | Structure-guided design of selective GPCR ligands for dopaminergic modulating neurotransmission. a–c, Solved structures of the three members of the D2 family of dopamine receptors: D2 (ref. 75), in magenta (a); D3 (ref. 76), in cyan (b); and D4 (ref. 34), in green (c). d, Location of the D4 selectivity filter, with nemonapride bound in green. e, The selectivity filter in d is occluded in the D2 receptor (D2 residues in magenta). f–h, As described in the text, the D3 receptor structure was used to modify the non-selective D3 inverse agonist eticlopride (f) to compound 19 (g), which has high affinity and selectivity. Alternatively, ultra-large scale docking led directly to ZINC621433144 (h) which emerged as a picomolar potent and selective D4 agonist79. Affinities for eticlopride are from ref. 130.

Challenges and opportunities for drug design and discovery from the initial docking screen was further optimized by dock- Selectivity. Given the large number of GPCR structures and reason- ing and subsequent identification of analogs; that ultimately led to able coverage of some GPCR families, these structures should, at the highly selective D4 partial agonist compound 9-6-2434, which least in theory, be useful for the de novo design of selective drugs. was >1,000-fold selective for the D4 receptor and lacked activity at This is particularly true when new ‘pockets’ are discovered, as was 320 other GPCRs. the case for the D275, D376 and D434 dopamine receptors. As shown Subsequent to that, the computational approach has been in Fig. 5a–c, these structures revealed potentially unique binding enhanced by automated docking-based screening of ultra-large surfaces that could be exploited for the design of selective ligands. libraries in a screen with 138 million drug-like compounds. Figure 5d shows a more detailed view of the D4 selectivity filter, This in silico screening, wherein each compound was docked in while Fig. 5e shows how this is occluded in the D2 receptor34. Given >100,000 conformations, required the analysis and scoring of 70 that all three structures were obtained with chemically distinct trillion docking events79 (Fig. 5h). In this study79, the previously and non-selective ligands, however, it is conceivable that the dif- identified D4 selectivity filter34 composed of the pocket formed by ferent binding pockets might simply reflect the fact that different residues Leu3.28 and Phe2.61 (Fig. 5c–e) was again computationally ligands with distinct chemical scaffolds (e.g., different chemotypes) targeted. That docking screen led to the discovery of a large num- engage different residues in the receptors. One way to test the ber of chemically novel and highly selective D4 ligands. Of these, hypothesis that these different binding pockets are pharmacologi- ZINC621433144 (Fig. 5h) was the most potent of the agonists cally relevant is to create new ligands that are designed to engage tested, with a median effective concentration (EC50) of 180 pM and these ‘selectivity filters’. For the D3 receptor, remarkable success was selectivity for D4 over D2 and D3 (Fig. 5h). On the basis of the rate achieved77 with a compound with 1,700-fold selectivity for the D3 of true positive results among the predicted active compounds, the receptor and minimal off-target activity (Fig. 5). Starting with the authors estimate that as many as 400,000 distinct D4-active com- seed compound eticlopride (Fig. 5f), which is a potent and non- pounds with more than 70,000 diverse chemotypes79 could exist selective D2 and D3 antagonist with weaker activity at D4, that in the library of 138 million compounds. Similar, albeit less com- group of researchers77 used a combination of docking and medici- putationally intensive, docking-based approaches have yielded 80 nal chemistry to identify a potential modified scaffold predicted to selective ligands for the 5-HT1B serotonin receptor , κ-opioid target a putative selectivity region within the D3 receptor previously receptor81, D2 dopamine receptor81 and other receptors (reviews identified by mutagenesis and molecular modeling76 (Fig. 5g). Via available in refs. 8,82). this structure-inspired design approach, compound 19 (Fig. 5g) was eventually synthesized and was found to be a potent and selective Structure-guided design of biased ligands D3 antagonist. Functional selectivity, also known as ‘biased signaling’83, has been An alternative approach is to target such ‘selectivity filters’ via defined as the process by which “...ligands induce (or stabilize) automated docking (Fig. 5c–e,h). In an initial proof-of-concept unique, ligand-specific receptor conformations…result(ing) in dif- study, the D4 dopamine receptor selectivity filter was targeted by ferential activation of signal transduction pathways associated with a docking screen of 600,000 commercially available compounds34 that particular receptor”83. The phenomenon of functional selec- from the ZINC database78. A set of ‘seed’ compounds discovered tivity occurs when different ligands at the same receptor lead to https://doi.org/10.1038/s41594-019-0252-8 preferential activation of different G proteins (e.g., G protein–sub- a type bias), a preference for G-protein signaling over arrestin (e.g., G-protein bias) or a preference for arrestin signaling over G pro- teins (e.g., arrestin bias). Reports of functional selectivity at GPCRs have appeared for many decades (an example of this is in ref. 84), CH CH3 3

H3C and such bias is now recognized as a nearly universal phenomenon N HN for GPCRs6. Indeed, there are now multiple reports of the discov- N N N N ery and optimization of arrestin-biased ligands85,86 and G protein– O O biased ligands87–89 for many GPCRs (reviews available in refs. 6,90,91). As mentioned above, so far there is only one structure of a GPCR HN HN (a GLP-1 receptor) in complex with a G protein–biased ligand O O 64 (peptide exendin-P5) and hetereotrimeric G proteins , a handful Compound 6 Compound 7 of structures of arrestin-biased ligands in complex with the 5-HT2B 100 G protein 100 G protein

32,62,70 max max serotonin receptor , and no structures of arrestin-biased ligands 75 β-arrestin 75 β-arrestin in a GPCR–arrestin complex. Thus, although there is a paucity of 50 50 Bias = 20 structural information on GPCR functional selectivity, there have 25 25 0 0 been several reports in which structure-informed and structure- % quinpirole E % quinpirole E –12 –11 –10 –9 –8 –7 –6 –5 –12 –11 –10 –9 –8 –7 –6 –5 inspired design of neurotransmitter-targeted G protein and arres- Log (compound 6) M Log (compound 7) M tin-biased ligands have been achieved.

O For example, the discovery of G protein–biased agonists for the 3.1 million compounds b NH 92 HN μ-opioid receptor has been reported , although this discovery was docked to GPR68 model N O

N+ O accomplished without knowledge of any of the structural features -O responsible for biased signaling. In that study, molecular docking of 3 million compounds was performed against the inactive conforma- H tion of the -opioid receptor targeting a highly conserved aspartic N N NH2 μ N N Iterative docking and 3.32 acid (Asp ) and engaging a conserved water network and a tyrosine analog identification 3.33 residue (Tyr ) putatively involved in selectivity and agonist activity. F F Compounds with predicted activity were tested for functional activ-

H ity at G protein and arrestin signaling and, after several rounds of H2N N N medicinal chemistry optimization, PZM21 was identified as a N N potent and efficacious G protein–biased agonist92. Similar G pro- OH 93,94 tein–biased ligands have been discovered for μ-opioid receptors Ogerin and κ-opioid receptors89,95, although these were not informed by GPR68 positive allosteric modulator structural determinants. On the other hand, via a combination of molecular dynamics and synthetic studies, a derivative of salvinorin Fig. 6 | Structure-inspired design of ligands for GPCRs to modulate 96 97 synaptic transmission. a, Strategy for the design of ligands with arrestin A was discovered for μ-opioid receptors that is G protein biased 72 and thus probably has less potential for abuse. Similarly, in a study bias for biogenic amine neurotransmitter receptors (details in text). Here, through the use of a combination of molecular dynamics simulations, of κ-opioid receptors, residues identified from structural studies as being essential for biased signaling ultimately led to the creation of docking and medicinal chemistry, compounds were created that were novel structure-guided G protein–biased opioid agonists52. predicted to interact with a conserved EL2 residue that can impart arrestin 72 bias. b, Computational strategy for the discovery of small-molecule A structure-inspired approach for the creation of biased ligands 102 at aminergic GPCRs, all of which are essential targets for the neu- probes with which to modulate oGPCRs such as GPR68 Here, 3.1 million rotransmitters serotonin, norepinephrine, dopamine and histamine, compounds were docked to a model of GPR68 and then, via iterative proved successful as well. For this study72, the authors took advan- docking and testing of analogs, ogerin was identified as a selective GPR68 tage of prior studies indicating that interactions with TM5 serine positive allosteric modulator. and other residues are essential for G-protein signaling62 at amin- ergic GPCRs, while interactions with EL2 residues can be essential for arrestin signaling61. Via a combination of molecular modeling, challenging. Thus, for example, there are no truly selective agonists molecular dynamics simulations, automated docking and synthesis, for D5 dopamine or M5 muscarinic receptors, although a modestly the arrestin-biased compound 7 was discovered for the D2 dopa- selective D5 antagonist100 and a fairly selective M5 antagonist101 mine receptor72 (Fig. 6a). As previous studies have shown that such have been reported. Unfortunately, off-target pharmacology has not arrestin-biased compounds may have efficacy in the treatment been reported for these compounds, so their selectivity over other of schizophrenia and related neuropsychiatric disorders85,98, this GPCRs is unknown. could represent a useful approach for optimizing arrestin-biased Although there are no published structures of orphan or under- medications for the treatment of neuropsychiatric disorders. studied brain GPCRs (oGPCRs38), the use of homology models Finally, it is also worth mentioning that biased signaling might be and automated docking has provided a structure-inspired approach influenced at the level of recruitment of and receptor phosphoryla- for ligand discovery. Thus, for example, an integrated approach tion by GPCR kinases99, and that this potentially can be influenced using parallel screening, homology modeling, docking and by GPCR ligands. analog synthesis led to ogerin — a selective positive allosteric modu­ lator for GPR68 — and selective negative allosteric modulators Orphan and understudied GPCRs. As previously mentioned, for GPR65102 (Fig. 6b). The allosteric modulator ogerin was nearly 50% of GPCRs, most of which are highly expressed in the demonstrated to affect learning and behavior in mice, which brain14, have a paucity of known ligands. Even for those neurotrans- indicates that GPR68 is a potentially important GPCR in the mitter receptors that have been the subject of intensive investiga- brain102. A similar approach led to the discovery of selective ago- tion, such as muscarinic and dopamine receptors, the identification nists for the oGPCR MRGPRX2, which is involved in pain and of suitably selective ligands for the various subtypes continues to be itch103 and binds many important neurotransmitters, including substance P and other peptides103. These approaches are facilitated eventually lead to neuropsychiatric medications with greater effi- by parallel screening approaches in which hundreds of oGPCRs cacy and fewer side effects. As on-target and off-target side effects, can be screened simultaneously102,104; active compounds are used to along with clinical effectiveness and safety, continue to be major inform homology models for docking and subsequent discovery of drivers of drug failures in clinical trials123, insights into the struc- new ligands102,103,105–107. tural basis of ligand engagement could provide new opportunities Genetic and model organism studies are also facilitating for GPCR-based neuropsychiatric drug discovery. Thus, where a understanding of the potential therapeutic importance of oGP- G protein–biased ligand might show greater efficacy and fewer side CRs for neuropsychiatric disease. Thus, a published study has effects, insights into the structural basis of such bias could acceler- indicates that the oGPCR MRGPRX4 is involved in the pain ate the discovery of novel chemotypes with biased signaling pat- and itch sensations mediated by bile acids108 Another study has terns. Indeed, this approach has been successful in the generation shown that MRGPRX4 is also essential for the preference for of biased tool compounds for many neurotransmitter receptors, menthol cigarettes in certain ethnic populations109. Given the including D4 dopamine receptors34,124, μ-opioid receptors92 and distribution of MRGPRX4 in peripheral nerves, it is likely that other receptors62,72. Although there are currently no structure- MRGPRX4 and related receptors are neurotransmitter recep- guided biased drugs in clinical trials, encouraging results have tors involved in peripheral sensations such as pain and itch110. appeared for TRV130 as a G protein–biased opioid receptor ago- Databases devoted to the brain distribution of oGPCRs39,111,112 nist for pain125, albeit with some abuse liabilities, as manifested in should facilitate discovery of their endogenous neurotransmit- preclinical studies126,127. ters, their function and neuropsychiatric implications for patho- As is clear from the foregoing, the past decade has witnessed genesis and treatment. astounding progress in understanding of the structure and func- tion of metabotropic receptors for neurotransmitters. This Polypharmacology. Historically, the most effective drugs for many increasing wealth of structural information, as well as advances complex neuropsychiatric diseases have targeted multiple GPCRs in structure-guided and structure-inspired drug discovery, prom- and other molecular targets9,113. It is now understood that this is ise to accelerate the discovery of drugs that target metabotropic probably because of the exceedingly complex genetic landscape of receptors for neurotransmitters with improved efficacies and common diseases in which hundreds to thousands of genes might fewer side effects. exert small effects114. This has led to the understanding that com- 114 plex diseases are omnigenic rather than polygenic , meaning Received: 13 March 2019; Accepted: 15 May 2019; that nearly every gene may ultimately exert a small effect on core Published online: 3 July 2019 disease pathways. Perhaps not surprisingly, the desire to discover increasingly selective drugs has resulted in lower overall success rates of drug-discovery screens115. Such a lack of success has led References to the hypothesis that generating polypharmacological drugs 1. Ullrich, A. & Schlessinger, J. Signal transduction by receptors with tyrosine with designated multiple targets represents a useful approach kinase activity. Cell 61, 203–212 (1990). for the treatment of complex diseases113,116. For aminergic 2. Schulz, S., Chinkers, M. & Garbers, D. L. Te guanylate cyclase/receptor family of proteins. FASEB J. 3, 2026–2035 (1989). GPCRs, there has been some success in discovering the 3. Chinkers, M. et al. A membrane form of guanylate cyclase is an atrial structural features responsible for polypharmacological activi- natriuretic peptide receptor. Nature 338, 78–83 (1989). ties33, which comprise a series of nine semi-conserved residues 4. Purves, D. et al. Two families of postsynaptic receptors. in Neuroscience (Asp3.32, Ile/Val3.33, Cys3.36, Thr3.37, Ala/Ser/Thr5.46, Trp 6.48, Phe6.51, 2nd edn. (Sinauer Associates, 2001). Phe6.52 and EL2 Val/Ile/Leu). Although in theory, structure- 5. Overington, J. P., Al-Lazikani, B. & Hopkins, A. L. How many drug targets are there? Nat. Rev. Drug Discov. 5, 993–996 (2006). guided approaches should be helpful for the discovery and Ref. 5 is classic paper that provides estimates for the size of the design of polypharmacological drugs, so far, this has not been ‘druggable’ genome. entirely successful117,118 6. Wacker, D., Stevens, R. C. & Roth, B. L. How ligands illuminate GPCR molecular pharmacology. Cell 170, 414–427 (2017). Therapeutic and commercial opportunities for 7. Hauser, A. S., Attwood, M. M., Rask-Andersen, M., Schiöth, H. B. & Gloriam, D. E. Trends in GPCR drug discovery: new agents, targets and structure-guided GPCR drug discovery indications. Nat. Rev. Drug Discov. 16, 829–842 (2017). Heptares (acquired by Sosei in 2016)119, and Receptos (acquired by 8. Roth, B. L., Irwin, J. J. & Shoichet, B. K. Discovery of new GPCR ligands to Celgene in 2015)120, along with Conformetrix121, have invested con- illuminate new biology. Nat. Chem. Biol. 13, 1143–1151 (2017). siderable resources in bringing GPCR structure-guided drug dis- 9. Roth, B. L., Shefer, D. J. & Kroeze, W. K. Magic shotguns versus magic bullets: selectively non-selective drugs for mood disorders and covery to fruition. So far, publicly available information indicates schizophrenia. Nat. Rev. Drug Discov. 3, 353–359 (2004). that the exemplar compounds are targeted mainly to well-studied 10. Marder, S. R. et al. Advancing drug discovery for schizophrenia. GPCRs for which selective ligands are difficult to obtain. Thus, for Ann. NY Acad. Sci. 1236, 30–43 (2011). example, Heptares has advanced an M4-selective agonist to phase I 11. Fricker, L. D. & Devi, L. A. Orphan neuropeptides and receptors: Novel trials, presumably for cognition enhancement (https://clinicaltrials. therapeutic targets. Pharmacol. Ter. 185, 26–33 (2018). 12. Lein, E. S. et al. Genome-wide atlas of gene expression in the adult mouse gov/ct2/show/NCT03244228?term=heptares&rank=4). Heptares brain. Nature 445, 168–176 (2007). has also brought a structure-guided mGlu5 negative allosteric 13. Gerfen, C. R., Paletzki, R. & Heintz, N. GENSAT BAC cre-recombinase modulator122 for metabotropic glutamate receptors to initial phase driver lines to study the functional organization of cerebral cortical and I trials, presumably for application to the central nervous system basal ganglia circuits. Neuron 80, 1368–1383 (2013). (https://clinicaltrials.gov/ct2/show/NCT03785054?term heptare 14. Regard, J. B., Sato, I. T. & Coughlin, S. R. Anatomical profling of G protein- = coupled receptor expression. Cell 135, 561–571 (2008). s&rank=2). Receptos, in partnership with Celgene, has advanced 15. Greengard, P. Te neurobiology of slow synaptic transmission. Science 294, RPC1063, an S1P receptor agonist for the treatment of multiple 1024–1030 (2001). sclerosis (https://www.accessdata.fda.gov/scripts/opdlisting/oopd/ 16. Berger, M., Gray, J. A. & Roth, B. L. Te expanded biology of serotonin. detailedIndex.cfm?cfgridkey=611517), to phase III clinical trials. Annu. Rev. Med. 60, 355–366 (2009). 17. McCorvy, J. D. & Roth, B. L. Structure and function of serotonin G protein- coupled receptors. Pharmacol. Ter. 150, 129–142 (2015). Future directions 18. Andrade, R., Malenka, R. C. & Nicoll, R. A. A. G protein couples serotonin Given the wealth of structural information now available for many and GABAB receptors to the same channels in hippocampus. Science 234, GPCRs, it is anticipated that structure-guided approaches will 1261–1265 (1986). 19. Blackmer, T. et al. G protein βγ directly regulates SNARE protein fusion 50. Rasmussen, S. G. et al. Structure of a nanobody-stabilized active state of the machinery for secretory granule exocytosis.Nat. Neurosci. 8, 421–425 (2005). β2 adrenoceptor. Nature 469, 175–180 (2011). Ref. 19 is one of the frst demonstrations that G-protein β- and 51. Kruse, A. C. et al. Activation and allosteric modulation of a muscarinic γ-subunits regulate vesicle-fusion machinery to inhibit synaptic acetylcholine receptor. Nature 504, 101–106 (2013). transmission. 52. Che, T. et al. Structure of the nanobody-stabilized active state of the κ 20. Gerachshenko, T. et al. Gβγ acts at the C terminus of SNAP-25 to mediate opioid receptor. Cell 172, 55–67.e15 (2018). presynaptic inhibition. Nat. Neurosci. 8, 597–605 (2005). 53. Huang, W. et al. Structural insights into µ-opioid receptor activation. Nature 21. Skeberdis, V. A. et al. Protein kinase A regulates calcium permeability of 524, 315–321 (2015). NMDA receptors. Nat. Neurosci. 9, 501–510 (2006). 54. Liang, Y. L. et al. Phase-plate cryo-EM structure of a class B GPCR-G-protein 22. Carver, C. M. & Shapiro, M. S. Gq-coupled muscarinic receptor complex. Nature 546, 118–123 (2017). enhancement of KCNQ2/3 channels and activation of TRPC channels in Ref. 54 provided the frst cryo-EM structure of an active GPCR in multimodal control of excitability in dentate gyrus granule cells. J. Neurosci. complex with G protein.

39, 1566–1587 (2019). 55. Liang, Y. L. et al. Cryo-EM structure of the active, Gs-protein complexed, 23. Stevens, R. C. & Wilson, I. A. Tech.Sight. Industrializing structural biology. human CGRP receptor. Nature 561, 492–497 (2018). Science 293, 519–520 (2001). 56. Draper-Joyce, C. J. et al. Structure of the adenosine-bound human

24. Schertler, G. F., Villa, C. & Henderson, R. Projection structure of adenosine A1 receptor-Gi complex. Nature 558, 559–563 (2018). rhodopsin. Nature 362, 770–772 (1993). 57. García-Nafría, J., Nehmé, R., Edwards, P. C. & Tate, C. G. Cryo-EM

Ref. 24 provided the frst cryo-EM structure of a membrane protein. structure of the serotonin 5-HT1B receptor coupled to heterotrimeric Go. 25. Baldwin, J. M. Te probable arrangement of the helices in G protein- Nature 558, 620–623 (2018). coupled receptors. EMBO J. 12, 1693–1703 (1993). 58. García-Nafría, J., Lee, Y., Bai, X., Carpenter, B. & Tate, C. G. Cryo-EM

26. Palczewski, K. et al. Crystal structure of rhodopsin: A G protein-coupled structure of the adenosine A2A receptor coupled to an engineered receptor. Science 289, 739–745 (2000). heterotrimeric G protein. eLife 7, e35946 (2018). Ref. 26 provided the frst crystal structure of a GPCR. 59. Koehl, A. et al. Structure of the µ-opioid receptor-Gi protein complex. 27. Rosenbaum, D. M. et al. GPCR engineering yields high-resolution Nature 558, 547–552 (2018). structural insights into β2-adrenergic receptor function. Science 318, 60. Zhang, Y. et al. Cryo-EM structure of the activated GLP-1 receptor in 1266–1273 (2007). complex with a G protein. Nature 546, 248–253 (2017). 28. Rasmussen, S. G. F. et al. Crystal structure of the human β2 adrenergic 61. Wacker, D. et al. Crystal structure of an LSD-bound human serotonin G-protein-coupled receptor. Nature 450, 383–387 (2007). receptor. Cell 168, 377–389.e12 (2017).

29. Cherezov, V. et al. High-resolution crystal structure of an engineered 62. McCorvy, J. D. et al. Structural determinants of 5-HT2B receptor activation human β2-adrenergic G protein-coupled receptor. Science 318, and biased agonism. Nat. Struct. Mol. Biol. 25, 787–796 (2018). 1258–1265 (2007). 63. Kang, Y. et al. Crystal structure of rhodopsin bound to arrestin by Refs. 27–29 provided the frst crystal structures of non-opsin GPCRs. femtosecond X-ray laser. Nature 523, 561–567 (2015). 30. Katritch, V., Cherezov, V. & Stevens, R. C. Structure-function of the G Ref. 63 provided the frst crystal structure of a GPCR with arrestin. protein-coupled receptor superfamily. Annu. Rev. Pharmacol. Toxicol. 53, 64. Liang, Y. L. et al. Phase-plate cryo-EM structure of a biased agonist-bound 531–556 (2013). human GLP-1 receptor-Gs complex. Nature 555, 121–125 (2018). 31. Wang, C. et al. Structural basis for molecular recognition at serotonin 65. Liu, W. et al. Structural basis for allosteric regulation of GPCRs by sodium receptors. Science 340, 610–614 (2013). ions. Science 337, 232–236 (2012). 32. Wacker, D. et al. Structural features for functional selectivity at serotonin 66. Fenalti, G. et al. Molecular control of δ-opioid receptor signalling. Nature receptors. Science 340, 615–619 (2013). 506, 191–196 (2014). 33. Peng, Y. et al. 5–HT2C receptor structures reveal the structural basis of 67. Staus, D. P. et al. Allosteric nanobodies reveal the dynamic range and GPCR polypharmacology. Cell 172, 719–730.e714 (2018). diverse mechanisms of G-protein-coupled receptor activation. Nature 535,

34. Wang, S. et al. D4 dopamine receptor high-resolution structures enable the 448–452 (2016). discovery of selective agonists. Science 358, 381–386 (2017). 68. Samama, P., Cotecchia, S., Costa, T. & Lefowitz, R. J. A mutation-induced 35. Gaulton, A. et al. ChEMBL: a large-scale bioactivity database for drug activated state of the β2-adrenergic receptor. Extending the ternary complex discovery. Nucleic Acids Res. 40, D1100–D1107 (2012). model. J. Biol. Chem. 268, 4625–4636 (1993). 36. Roth, B. et al. Multiplicity of serotonin receptors: useless diverse molecules 69. Roth, B. L. DREADDs for neuroscientists. Neuron 89, 683–694 (2016). or an embarrassment of riches? Neuroscientist 6, 252–262 (2000). 70. Wacker, D. et al. Crystal structure of an LSD-bound human serotonin 37. Wang, Y. et al. PubChem: a public information system for analyzing receptor. Cell 168, 377–389.e312 (2017). bioactivities of small molecules. Nucleic Acids Res. 37, W623–33 (2009). 71. Ballesteros, J. A. & Weinstein, H. Integrated methods for the construction 38. Roth, B. L. & Kroeze, W. K. Integrated approaches for genome-wide of three-dimensional models and computational probing of structure- interrogation of the druggable non-olfactory G protein-coupled receptor function relations in G protein-coupled receptors. Meth. Neurosci. 25, superfamily. J. Biol. Chem. 290, 19471–19477 (2015). 366 (1995). 39. Oprea, T. I. et al. Unexplored therapeutic opportunities in the human 72. McCorvy, J. D. et al. Structure-inspired design of β-arrestin-biased ligands genome. Nat. Rev. Drug Discov. 17, 377 (2018). for aminergic GPCRs. Nat. Chem. Biol. 14, 126–134 (2018).

40. Yang, S. et al. Crystal structure of the Frizzled 4 receptor in a ligand-free 73. Ballesteros, J. A. et al. Activation of the β2-adrenergic receptor involves state. Nature 560, 666–670 (2018). disruption of an ionic lock between the cytoplasmic ends of transmembrane 41. Shimamura, T. et al. Structure of the human histamine H1 receptor segments 3 and 6. J. Biol. Chem. 276, 29171–29177 (2001). complex with doxepin. Nature 475, 65–70 (2011). 74. Shapiro, D. A. et al. Evidence for a model of agonist-induced activation of 42. Haga, K. et al. Structure of the human M2 muscarinic acetylcholine 5–HT2A serotonin receptors which involves the disruption of a strong ionic receptor bound to an antagonist. Nature 482, 547–551 (2012). interaction between helices 3 and 6. J. Biol. Chem. 18, 18 (2002). 43. Kruse, A. C. et al. Structure and dynamics of the M3 muscarinic 75. Wang, S. et al. Structure of the D2 dopamine receptor bound to the atypical acetylcholine receptor. Nature 482, 552–556 (2012). antipsychotic drug risperidone. Nature 555, 269–273 (2018). 44. Tal, D. M. et al. Crystal structures of the M1 and M4 muscarinic 76. Chien, E. Y. et al. Structure of the human dopamine D3 receptor in acetylcholine receptors. Nature 531, 335–340 (2016). complex with a D2/D3 selective antagonist. Science 330, 1091–1095 (2010). 45. Liu, H. et al. Structure-guided development of selective M3 muscarinic 77. Kumar, V. et al. Highly selective dopamine D3 receptor (D3R) antagonists acetylcholine receptor antagonists. Proc. Natl Acad. Sci. USA 115, and partial agonists based on eticlopride and the D3R crystal structure: 12046–12050 (2018). new leads for opioid dependence treatment. J. Med. Chem. 59, 46. Vardigan, J. D. et al. Improved cognition without adverse efects: novel M1 7634–7650 (2016). muscarinic potentiator compares favorably to donepezil and xanomeline in 78. Irwin, J. J. & Shoichet, B. K. ZINC—a free database of commercially rhesus monkey. Psychopharmacology (Berl) 232, 1859–1866 (2015). available compounds for virtual screening. J. Chem. Inf. Model. 45, 47. Foster, D. J., Choi, D. L., Conn, P. J. & Rook, J. M. Activation of M1 and 177–182 (2005). M4 muscarinic receptors as potential treatments for Alzheimer’s disease and 79. Lyu, J. et al. Ultra-large library docking for discovering new chemotypes. schizophrenia. Neuropsychiatr. Dis. Treat. 10, 183–191 (2014). Nature 566, 224–229 (2019). 48. Tal, D. M., Glukhova, A., Sexton, P. M. & Christopoulos, A. Structural 80. Rodríguez, D., Brea, J., Loza, M. I. & Carlsson, J. Structure-based insights into G-protein-coupled receptor allostery. Nature 559, 45–53 (2018). discovery of selective serotonin 5-HT(1B) receptor ligands. Structure 22, 49. Rasmussen, S. G. et al. Crystal structure of the β2 adrenergic receptor-Gs 1140–1151 (2014). protein complex. Nature 477, 549–555 (2011). 81. Negri, A. et al. Discovery of a novel selective κ-opioid receptor agonist Ref. 49 provided the frst structure of a GPCR in a complex in its active using crystal structure-based virtual screening. J. Chem. Inf. Model. 53, state with a G protein heterotrimer. 521–526 (2013). 82. Irwin, J. J. & Shoichet, B. K. Docking screens for novel ligands conferring 109. Kozlitina, J. et al. An African-specifc haplotype in MRGPRX4 is associated new biology. J. Med. Chem. 59, 4103–4120 (2016). with menthol cigarette smoking. PLoS Genet. 15, e1007916 (2019). 83. Urban, J. D. et al. Functional selectivity and classical concepts of 110. Dong, X., Han, S., Zylka, M. J., Simon, M. I. & Anderson, D. J. A diverse quantitative pharmacology. J. Pharmacol. Exp. Ter. 320, 1–13 (2007). family of GPCRs expressed in specifc subsets of nociceptive sensory 84. Roth, B. L. & Chuang, D.-M. Multiple mechanisms of serotonergic signal neurons. Cell 106, 619–632 (2001). transduction. Life Sci. 41, 1051–1064 (1987). 111. Ehrlich, A. T. et al. Expression map of 78 brain-expressed mouse orphan Ref. 84 provided the frst explication of the principles of functional GPCRs provides a translational resource for neuropsychiatric research. selectivity and biased agonism and antagonism. Commun Biol 1, 102 (2018). 85. Allen, J. A. et al. Discovery of β-arrestin-biased dopamine D2 ligands for 112. Oprea, T. I. et al. Far away from the lamppost. PLoS Biol. 16, probing signal transduction pathways essential for antipsychotic efcacy. e3000067 (2018). Proc. Natl Acad. Sci. USA 108, 18488–18493 (2011). 113. Dar, A. C., Das, T. K., Shokat, K. M. & Cagan, R. L. Chemical genetic 86. Boerrigter, G. et al. Cardiorenal actions of TRV120027, a novel β-arrestin- discovery of targets and anti-targets for cancer polypharmacology. Nature biased ligand at the angiotensin II type I receptor, in healthy and heart 486, 80–84 (2012). failure canines: a novel therapeutic strategy for acute heart failure. 114. Boyle, E. A., Li, Y. I. & Pritchard, J. K. An expanded view of complex traits: Circ. Heart Fail. 4, 770–778 (2011). from polygenic to omnigenic. Cell 169, 1177–1186 (2017). 87. Chen, X. et al. Discovery of G protein-biased D2 dopamine receptor partial 115. Shih, H. P., Zhang, X. & Aronov, A. M. Drug discovery efectiveness from agonists. J. Med. Chem. 59, 10601–10618 (2016). the standpoint of therapeutic mechanisms and indications. Nat. Rev. Drug 88. Chen, X. T. et al. Structure-activity relationships and discovery of a G Discov. 17, 78 (2018). protein biased μ opioid receptor ligand, [(3-methoxythiophen-2-yl)methyl] 116. Besnard, J. et al. Automated design of ligands to polypharmacological (2-[(9R)-9-(pyridin-2-yl)-6-oxaspiro-[4.5]decan-9-yl]ethyl)amine (TRV130), profles. Nature 492, 215–220 (2012). for the treatment of acute severe pain. J. Med. Chem. 56, 8019–8031 (2013). 117. Anighoro, A., Bajorath, J. & Rastelli, G. Polypharmacology: challenges and 89. Lovell, K. M. et al. Structure-activity relationship studies of functionally opportunities in drug discovery. J. Med. Chem. 57, 7874–7887 (2014). selective κ opioid receptor agonists that modulate ERK 1/2 phosphorylation 118. Weiss, D. R. et al. Selectivity challenges in docking screens for GPCR while preserving G protein over βarrestin2 signaling bias. ACS Chem. targets and antitargets. J. Med. Chem. 61, 6830–6845 (2018). Neurosci. 6, 1411–1419 (2015). 119. Micklus, A. & Muntner, S. Biopharma deal-making in 2016. Nat. Rev. Drug 90. Wootten, D., Christopoulos, A., Marti-Solano, M., Babu, M. M. & Sexton, P. Discov. 16, 161–162 (2017). M. Mechanisms of signalling and biased agonism in G protein-coupled 120. Morrison, C. & Lähteenmäki, R. Public biotech in 2015 - the numbers. receptors. Nat. Rev. Mol. Cell Biol. 19, 638–653 (2018). Nat. Biotechnol. 34, 709–715 (2016). 91. Smith, J. S., Lefowitz, R. J. & Rajagopal, S. Biased signalling: from 121. Blundell, C. D. & Almond, A. Method for determining three-dimensional simple switches to allosteric microprocessors. Nat. Rev. Drug Discov. 17, structures of dynamic molecules. UK patent 0718027.6 (2017). 243–260 (2018). 122. Christopher, J. A. et al. Fragment and structure-based drug discovery for a 92. Manglik, A. et al. Structure-based discovery of opioid analgesics with class C GPCR: discovery of the mGlu5 negative allosteric modulator reduced side efects. Nature 537, 185–190 (2016). HTL14242 (3-chloro-5-[6-(5-fuoropyridin-2-yl)pyrimidin-4-yl]benzonitrile. Ref. 92 reports structure-guided discovery of biased ligands. J. Med. Chem. 58, 6653–6664 (2015). 93. Váradi, A. et al. Mitragynine/corynantheidine pseudoindoxyls as opioid 123. Harrison, R. K. Phase II and phase III failures: 2013-2015. Nat. Rev. Drug analgesics with μ agonism and δ antagonism, which do not recruit Discov. 15, 817–818 (2016). β-arrestin-2. J. Med. Chem. 59, 8381–8397 (2016). 124. Lyu, J. et al. Ultra-large library docking for discovering new chemotypes. 94. Schmid, C. L. et al. Bias factor and therapeutic window correlate to predict Nature 566, 224–229 (2019). safer opioid analgesics. Cell 171, 1165–1175.e1113 (2017). 125. Singla, N. K. et al. APOLLO-1: a randomized, controlled, phase 3 study of 95. White, K. L. et al. Identifcation of novel functionally selective κ-opioid oliceridine (TRV130) for the treatment of moderate to severe pain following receptor scafolds. Mol. Pharmacol. 85, 83–90 (2014). bunionectomy. Spine J. 17, S2017 (2017). 96. Roth, B. L. et al. Salvinorin A: a potent naturally occurring 126. Araldi, D., Ferrari, L. F. & Levine, J. D. Mu-opioid receptor (MOR) biased nonnitrogenous κ opioid selective agonist. Proc. Natl Acad. Sci. USA 99, agonists induce biphasic dose-dependent hyperalgesia and analgesia, and 11934–11939 (2002). hyperalgesic priming in the rat. Neuroscience 394, 60–71 (2018). 97. Crowley, R. S. et al. Synthetic studies of neoclerodane diterpenes from salvia 127. Austin Zamarripa, C. et al. Te G-protein biased mu-opioid agonist, divinorum: identifcation of a potent and centrally acting μ opioid analgesic TRV130, produces reinforcing and antinociceptive efects that are comparable with reduced abuse liability. J. Med. Chem. 59, 11027–11038 (2016). to oxycodone in rats. Drug Alcohol Depend. 192, 158–162 (2018). 98. Park, S. M. et al. Efects of β-arrestin-biased dopamine D2 receptor ligands 128. Kunkel, M. T. & Peralta, E. G. Identifcation of domains conferring G on schizophrenia-like behavior in hypoglutamatergic mice. protein regulation on inward rectifer potassium channels. Cell 83, Neuropsychopharmacology 41, 704–715 (2016). 443–449 (1995). 99. Choi, M. et al. G protein–coupled receptor kinases (GRKs) orchestrate 129. Alexander, G. M. et al. Remote control of neuronal activity in transgenic

biased agonism at the β2-adrenergic receptor. Sci Signal 11, mice expressing evolved G protein-coupled receptors. Neuron 63, eaar7084 (2018). 27–39 (2009). 100. Mohr, P., Decker, M., Enzensperger, C. & Lehmann, J. Dopamine/serotonin Ref. 129 was the frst paper to demonstrate that designer receptors receptor ligands. 12(1): SAR studies on hexahydro-dibenz[d,g]azecines lead exclusively activated by designer drugs can activate neuronal activity. to 4-chloro-7-methyl-5,6,7,8,9,14-hexahydrodibenz[d,g]azecin-3-ol, the frst 130. Tang, L., Todd, R. D., Heller, A. & O’Malley, K. L. Pharmacological and picomolar D5-selective dopamine-receptor antagonist. J. Med. Chem. 49, functional characterization of D2, D3 and D4 dopamine receptors in 2110–2116 (2006). fbroblast and dopaminergic cell lines. J. Pharmacol. Exp. Ter. 268, 101. Gentry, P. R. et al. Discovery, synthesis and characterization of a highly 495–502 (1994).

muscarinic acetylcholine receptor (mAChR)-selective M5-orthosteric 131. Doré, A. S. et al. Structure of the adenosine A2A receptor in complex antagonist, VU0488130 (ML381): a novel molecular probe. ChemMedChem with ZM241385 and the xanthines XAC and cafeine. Structure 19, 9, 1677–1682 (2014). 1283–1293 (2011). 102. Huang, X. P. et al. Allosteric ligands for the pharmacologically dark 132. Cherezov, V. et al. High-resolution crystal structure of an engineered human

receptors GPR68 and GPR65. Nature 527, 477–483 (2015). β2-adrenergic G protein-coupled receptor. Science 318, 1258–1265 (2007). Ref. 102 reports structure-inspired discovery of allosteric ligands for an 133. Warne, T. et al. Structure of a β1-adrenergic G-protein-coupled receptor. orphan GPCR. Nature 454, 486–491 (2008).

103. Lansu, K. et al. In silico design of novel probes for the atypical opioid 134. Kimura, K. T. et al. Structures of the 5-HT2A receptor in complex with receptor MRGPRX2. Nat. Chem. Biol. 13, 529–536 (2017). the antipsychotics risperidone and zotepine. Nat. Struct. Mol. Biol. 26, 104. Kroeze, W. K. et al. PRESTO-Tango as an open-source resource for 121–128 (2019). interrogation of the druggable human GPCRome. Nat. Struct. Mol. Biol. 22, 135. Hua, T. et al. Crystal structure of the human cannabinoid receptor CB1. 362–369 (2015). Cell 167, 750–762.e714 (2016). 105. Ngo, T. et al. Orphan receptor ligand discovery by pickpocketing 136. Manglik, A. et al. Crystal structure of the µ-opioid receptor bound to a pharmacological neighbors. Nat. Chem. Biol. 13, 235–242 (2017). morphinan antagonist. Nature 485, 321–326 (2012). 106. Ngo, T. et al. Identifying ligands at orphan GPCRs: current status using 137. Wu, H. et al. Structure of the human κ-opioid receptor in complex with structure-based approaches. Br. J. Pharmacol. 173, 2934–2951 (2016). JDTic. Nature 485, 327–332 (2012). 107. Trauelsen, M. et al. Receptor structure-based discovery of non-metabolite 138. Yin, J. et al. Structure and ligand-binding mechanism of the human OX1 agonists for the succinate receptor GPR91. Mol. Metab. 6, 1585–1596 (2017). and OX2 orexin receptors. Nat. Struct. Mol. Biol. 23, 293–299 (2016). 108. Meixiong, J. et al. Identifcation of a bilirubin receptor that may mediate a 139. Hanson, M. A. et al. Crystal structure of a lipid G protein-coupled receptor. component of cholestatic itch. eLife 8, e44116 (2019). Science 335, 851–855 (2012). 140. Misono, K. S. et al. Structure, signaling mechanism and regulation of the Additional information natriuretic peptide receptor guanylate cyclase. FEBS J. 278, 1818–1829 (2011). Reprints and permissions information is available at www.nature.com/reprints. Correspondence should be addressed to B.L.R. A cknowledgements Peer review information: Katarzyna Marcinkiewicz was the primary editor on this The author thanks W. Kroeze for editing and comments. Work described in this Review article and managed its editorial process and peer review in collaboration with the rest of was funded by grants and contracts from the US National Institute of Health as well as the editorial team. the Michael Hooker Distinguished Professorship. Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in Competing interests published maps and institutional affiliations. The author declares no competing interests. © Springer Nature America, Inc. 2019