GPCR systems pharmacology: a different perspective on the development of biased therapeutics

INTRODUCTION

G protein-coupled receptors (GPCRs) represent the largest family of transmembrane receptors and currently serve as the target for ∼30% of all Food and Drug Administration-approved small molecule drugs (1). GPCRs can signal through many effector proteins; however, the best characterized are the G proteins and β-arrestins. It was recently discovered that some ligands can preferentially activate specific signaling pathways over others, a phenomenon termed “biased agonism” (2). This discovery has spurred many efforts to develop G protein- or β-arrestin-biased drugs that demonstrate heightened therapeutic efficacy with improved side effect profile. However, these efforts have not yet demonstrated enhanced clinical utility (3). With the hindsight of nearly 50 years of scientific advancement in the GPCR field, the binary of G protein versus β-arrestin model of biased receptor signaling is too reductive when considering the pluridimensional signaling observed at GPCRs. Rather than associating a particular signaling output with a single proximal effector protein (G protein or β-arrestin), it may be better to adopt a “systems pharmacology” approach, where the mechanisms underlying a desired or pathological phenotype are explored at multiple levels beyond the receptor. Here, we provide a historical context of GPCR signaling and the discovery of biased agonism. We also discuss the current methods used to study biased agonism at GPCRs and highlight the potential utility of adopting a “systems pharmacology” approach to understand and exploit biased GPCR signaling.

A BRIEF HISTORY OF GPCR SIGNALING

Although the idea of a cell receptor was first conceptualized as early as 1905, the study of receptor regulation and signaling mechanisms was made possible in the early 1970s through the emergence of new tools including photoaffinity labeling, radioligand binding, and affinity chromatography (4, 5). These techniques drastically improved our ability to provide specific and quantitative information about the regulation of receptors and the binding affinity for a wide variety of ligands. Receptor biology research no longer had to rely on inferences made solely through observations of downstream signaling effects. Rather, it allowed for the interpretation of the now commonly used agonist and antagonist binding curves and their incorporation into receptor theory. Eventually, these efforts led to the development of the widely used ternary complex model where receptor binding to an intracellular effector allosterically enhances the binding affinity of the receptor for the ligand, and vice versa (6).

The first proposal of the ternary complex model of G proteins was suggested in 1980 after the successful purification of the heterotrimeric protein in charge of regulating adenylyl cyclase activity, with the classification as a “G protein” reflecting its ability to bind guanine nucleotides (7). In later years, the Lefkowitz laboratory would use chimeric receptors of the α₂ and β₂ adrenergic receptors to demonstrate that the α₂ receptor inhibits adenylyl cyclase through the inhibitory G protein Gα_i whereas the β₂ adrenergic receptor (β2AR) stimulates adenylyl cyclase through the stimulatory G protein Gα_s (8). Subsequently, the purification of the β2AR additionally allowed for the design of oligonucleotide probes used to clone the cDNA and gene of the β2AR, leading to the discovery of canonical features associated with the GPCR family, such as the seven transmembrane domains, sites for regulatory phosphorylation, and sites of consensus sequences for glycosylation at the amino terminus (9).

With the sequencing of the β2AR came the realization that it shared significant homology with rhodopsin, a receptor which has contributed much to our understanding of canonical GPCR signaling (10). Rhodopsin was identified as a protein in the 1930s that underwent signal transduction after exposure to light through cis to trans isomerization of conjugated retinal. During the 1970s into the 1980s, it was found that rhodopsin activated a cyclic nucleotide phosphodiesterase through a GTPase now known as the G protein transducin (10). Around the same time, researchers discovered the existence of rhodopsin kinase, now known as GRK1, which was able to phosphorylate light-activated rhodopsin (11). In addition, reports identified a protein, formerly known as 48 kDa protein or retinal S antigen and now known as visual arrestin, that was involved in the deactivation of rhodopsin (12). As more GPCRs were identified using sequence homology, the focus shifted to understanding the phenomenon of receptor desensitization, which eventually led to the identification of the β-adrenergic receptor kinase (BARK), now known as GPCR kinase 2 (GRK2), one of the seven classified GRKs primarily responsible for phosphorylation of GPCRs (13). β-arrestin, closely related to visual arrestin, was discovered shortly after in 1987 for its ability to “arrest” signaling across the GPCR superfamily (14).

The current paradigm of GPCR signaling involves interaction of the receptor with the heterotrimeric G proteins, the GRKs, and the β-arrestins. Following stimulation, the activated receptor interacts with the G proteins, resulting in a cascade of signaling followed by phosphorylation of the receptor by the GRKs. β-arrestin then binds to the phosphorylated receptor and sterically inhibits further G protein activation, thus resulting in receptor desensitization and a diminished physiological response. It was subsequently shown that β-arrestin can also serve as its own signal transducing system by acting as a scaffold that links the receptor to other intracellular molecules (15). Of these pathways, the most rigorously studied is the ability of β-arrestin to scaffold effectors in the mitogen-activated protein kinase/extracellular signal-related kinase (MAPK/ERK) pathway, and also mediate clathrin-dependent endocytosis, thus proving capable of facilitating receptor desensitization, receptor internalization, and signaling (5).

Further scrutiny of GPCR signaling uncovered the discovery of biased ligands, which refers to the ability of an agonist to selectively activate some pathways, such as those mediated by G proteins or β-arrestins, compared with a reference agonist. Classically, agonists are capable of eliciting maximal signaling responses, partial agonists generate submaximal responses, antagonists do not generate any change in receptor signaling but compete for the receptor orthosteric site, and inverse agonists suppress basal receptor signaling. Importantly, biased agonists do not fit this classical paradigm as they may demonstrate agonist behavior at certain signaling pathways but may be partial agonists or even antagonists at other signaling pathways, i.e., they signal with different efficacies through signaling pathways downstream of a receptor. Recent studies have provided structural evidence that biased agonists stabilize the receptor in particular conformation that differ from the classical “active” and “inactive” states of the receptor, thereby allowing the activation of selective responses (16).

Biased agonists have served as tool compounds to understand specific GPCR signaling pathways that are associated with either β-arrestin or G protein signaling (2). For example, certain GPCRs demonstrate β-arrestin-dependent activation of receptor internalization and activation of MAPK/ERK signaling (15). Conversely, there is evidence for G protein-dependent signaling including the regulation of cyclic adenosine monophosphate (cAMP) levels through adenylyl cyclase, calcium levels through phospholipase C, and the activity of Rho guanine nucleotide exchange factor (17). However, not all β-arrestin- and G protein-mediated signaling pathways are independent of each other. For example, several studies have demonstrated that the MAPK/ERK signaling pathways can be activated by both β-arrestins and G proteins, albeit with differential kinetic and spatial patterns (15). In addition, recent work has observed an agonist-induced physical association between Gα_i and β-arrestin, even at GPCRs that do not canonically couple to Gα_i (18).

CURRENT APPROACHES TO STUDY GPCR-BIASED AGONISM

There are many mechanisms that result in a functionally selective GPCR response. The most commonly studied is “ligand bias,” where a ligand demonstrates biased signaling relative to a reference agonist, commonly the endogenous ligand, at the same receptor (Fig. 1A). For example, G protein-biased agonists of the µ-opioid receptor have distinct physiological effects compared with reference agonists (19–21). In “receptor bias,” a GPCR demonstrates preferential signaling through one transducer over another, largely irrespective of the ligand used to activate the receptor (Fig. 1B), a phenomenon that has been described at atypical chemokine receptors that recruit β-arrestins but do not activate G proteins (22). Methods for quantifying bias based on cellular analyses provide data for one defined set of experimental conditions, whereas in vivo, ligands encounter a range of tissues with different receptor densities, effector expression levels, and agonist sensitivities (23). In “systems bias,” an unbiased ligand and receptor may demonstrate biased signaling between cell types due to differences in the stoichiometry of downstream effectors (24) (Fig. 1C). For example, a β-arrestin-biased D2 dopamine receptor agonist had distinct effects in the cortex and striatum, which was associated with differences in GRK2 expression (25). In “location bias,” the specific subcellular location of a GPCR may demonstrate a signaling profile that is different than that observed at the plasma membrane (Fig. 1D) (26). For example, endosomal signaling by GPCRs has been shown to have distinct effects on transcription (27) and therapeutically targeting neurokinin endosomal signaling has distinct effects on the pain response (28).

In addition, granularity of biased agonism exists that allows for even greater signaling specificity beyond the canonical G protein versus β-arrestin paradigm. We refer to this granularity in biased agonism as “subdomains” of bias. For example, β-arrestin is known to adopt many different conformations (tail, core, and megaplex) which are directly associated with specific functions, such as receptor internalization or inhibition of G protein signaling (29) (Fig. 1E). GPCRs are now appreciated to hetero- or homodimerize, and even oligomerize, with other receptors or with accessory proteins, which impact their signaling (30) (Fig. 1F). The biased agonism observed by a GPCR may also vary based on the specific kinetic profile of the observed signaling (31) (Fig. 1G). In addition, a GPCR can simultaneously couple to many different isoforms of both G protein and β-arrestins, and these isoforms can demonstrate divergent signaling (32) (Fig. 1H). Finally, there are many identified signaling effectors that couple to GPCRs beyond G proteins and β-arrestins, which also may demonstrate biased signaling (Fig. 1I).

Our understanding of biased agonism has been aided by technological breakthroughs, such as those in receptor pharmacology, structural biology, and biotechnology, enabling more rapid, thorough, and targeted detailing of GPCR structural properties and signaling activity (Table 1). In particular is the adoption of resonance energy transfer (RET)-based assays with fluorescence (FRET) and bioluminescence (BRET) biosensors. Using these technologies, multiple aspects of GPCR signaling including heterotrimeric G protein dissociation (35), G protein nucleotide status (36), β-arrestin recruitment and trafficking (33), β-arrestin conformation (40), second messenger accumulation (54), and receptor dimerization (39) can be interrogated. We do not discuss the methods used to investigate the structural mechanisms underlying biased agonism as they are beyond the scope of this review.

Many assessments of biased agonism relied on the detection of an amplified signal, where the efficacy of GPCR signaling through a particular pathway can be easily misinterpreted (59). Many of these RET-based assays are stoichiometric in nature, allowing for accurate assessments and direct comparison of relative coupling with G proteins and β-arrestins. Because RET-based technologies can interrogate many of the aforementioned subdomains of biased agonism, they have become widely adapted as tools to assess GPCR functional selectivity. Using these technologies, one can mathematically model, either quantitatively or qualitatively, the extent of biased agonism using bias plots or by calculating bias factors (59). A limitation to these RET-based assays is that many of them require fusion to a fluorescent or bioluminescent probe that may impact the normal function of the protein being studied. However, novel methods have been developed to use RET methods in a manner that allows for interrogation of GPCR transducers, such as G protein nucleotide status, without the need to modify the protein of interest (36).

It is next important to determine whether the observed biased pharmacological response corresponds to a functionally selective cellular response. Quantitative imaging technologies, such as high content imaging (HCS) (60), label-free optical tools including dynamic mass redistribution (DMR) (61) and surface plasmon resonance (SPR) (62), and electric biosensor technologies, such as cellular impedance spectroscopy (63) have provided a more comprehensive tool set for determining phenotypes associated with GPCR activity. Such methods can report on cellular processes such as cell proliferation, migration, cell death, cell-cell communication, and adhesion (64). These whole cell phenotypical profiling approaches can delineate collective cellular responses and ultimate biological outputs of GPCR signaling activity. Importantly, these assays can be used in conjunction with tools that manipulate GPCR signaling to determine the contribution of G proteins and β-arrestins to the cellular response. For example, one can study the function of a biased agonist in the context of pharmacological inhibitors of specific G protein isoforms, or by using cell lines devoid of G proteins or β-arrestins (15).

Once a biased ligand is associated with a desirable cellular or in vivo response, one can identify GPCR therapeutics using screening-based approaches on a desired set of pharmacological properties. There are also efforts to use in silico methods to develop GPCR therapeutics based on existing ligand-bound GPCR structures (65). For example, the development of the FDA-approved antagonist Maraviroc, used in the treatment of human immunodeficiency virus (HIV), began with high throughput screening (HTS) using a CC-chemokine receptor 5 radioligand-binding assay, which identified a weak agonist that had no cellular antiviral activity, but served as a starting point for compound modification and synthesis based on structure-activity relationship studies (66). However, attempts to translate such pharmacological findings for biased agonists to therapeutics is difficult, particularly due to problems with drug efficacy, toxicity, or unexpected biology (67). Cell culture conditions, in which many signaling studies are conducted, are often not analogous to biological conditions. For example, studies in Chinese hamster ovary (CHO) cells have shown that when the mouse β₃-adrenoceptor is expressed at high concentrations, cAMP-mediated signaling predominates, but when the same receptor is expressed at low concentrations, MAPK/ERK pathways become the most significant signaling pathway (68). Due to the context-based specificity of biased signaling, addressing such discrepancies is critical for successful biased drug development.

UTILIZING SYSTEMS PHARMACOLOGY TO UNDERSTAND AND EXPLOIT GPCR-BIASED AGONISM

The discovery of biased agonism at GPCRs raised the possibility of developing functionally selective pharmaceutical drugs which activate therapeutic signaling pathways while simultaneously avoiding deleterious effects (2). Currently, the general framework in developing or discovering biased GPCR therapeutics is to associate biased signaling of compounds at the level of the GPCR with a desired or pathological physiological output in vivo or in cell culture. Frequently, the observed phenotype is associated with signaling through the proximal transducers with little insight into the signaling pathways that are directly responsible for the observed phenotype (Fig. 2A). This approach is best highlighted at the µ-opioid receptor, where previous work suggested that signaling via β-arrestins leads to constipation, respiratory, and tolerance whereas signaling through G protein leads to antinociception (69, 70). However, more recent studies have challenged those findings, and it is unclear if this paradigm is as simple as previously described (71). This general approach relies on associating differential activation of a few signaling events proximal to the receptor, but the overall effect of an agonist in vivo is driven by the activity of multiple pathways whose activity is collectively modulated by the ligand.

Rather than acting discretely through G proteins or through β-arrestins, GPCRs signal through multiple transducers/effectors through a complex and interconnected signaling network. There are many potential explanations underlying this signaling including direct interactions between G proteins and β-arrestins (18), convergence of initially divergent signaling pathways, differential kinetic and spatial patterns of activation of the same pathways, multiple drug targets by a specific ligand, or the integration of both GPCR and non-GPCR signaling. Associating proximal cellular signaling assays of G protein and β-arrestin signaling with specific phenotypes may be too reductive when trying to develop biased pharmaceuticals which promote the desired physiological effects. For those reasons, we advocate for a “systems pharmacology” approach when developing biased therapeutics (Fig. 2B). Systems pharmacology has been defined as “the study of a biological process in which all of the relevant components are studied together in parallel to discover integrative mechanisms” (72). In practice, after a particular phenotype and proximal GPCR signaling profile are associated, a more rigorous approach is taken to dissect the complex signaling that underlies this association. Given the widespread availability and reasonable cost of current technologies, one can assess many aspects of cellular biology and biochemistry. Mass spectrometry can identify differential expression of proteins, posttranslational modifications such as phosphorylation, and even GPCR effectors using proximity-based assays (46,47, 73). RNA sequencing (RNA-seq), single-cell RNA-seq, chromatin immunoprecipitation sequencing (ChIP-seq), and assay for transposase-accessible chromatin using sequencing (ATAC-seq) allow for high throughput assessment of transcription, promoter function, and DNA accessibility (74). Various fluorescent and bioluminescent biosensors exist to assess kinase activity and second messengers such as cAMP and calcium (57). Use of multidimensional assays and the associated computational tools to analyze such large data sets may uncover key and novel GPCR signaling pathways that are better correlated to or even directly associated with the phenotype being analyzed. Then, drugs can be screened or developed that demonstrate a signaling profile that appropriately regulate the effector(s) identified in addition to or in place of G proteins and β-arrestins. This systems pharmacology approach has already been applied to understand the mechanisms underlying the V2R-mediated regulation of the aquaporin-2 channel in water balance disorders (74). Similarly, a systems pharmacology approach was used to understand the mechanisms underlying retinal degeneration and identify novel combination therapies that target multiple GPCRs to potentially treat this disease (75, 76). Importantly, because laboratories can now generate large -omic data sets that interrogate cellular signaling across multiple dimensions, there has been a revolution in data science to analyze “multi-omic” data sets which integrate multiple data sets to study complex biological processes (77). An ideal systems pharmacology approach will require a concerted effort that lies at the intersection of biochemistry, molecular biology, structural biology, pharmacology, and data science to completely understand how a receptor-ligand interaction is translated into a biological phenotype. Adopting a systems pharmacology approach to biased agonism at GPCRs may unveil novel effectors or complex signaling networks that are better associated with cellular and animal physiology and increase the likelihood of developing a successful biased GPCR therapeutic.

CONCLUSIONS

Biased agonism at GPCRs presents an exciting opportunity to develop targeted therapeutics with heightened clinical efficacy and an improved side effect profile. Currently, biased agonism is generally described as the relative ability of a GPCR to differentially activate proximal transducers, such as G proteins or β-arrestins. Efforts to develop biased agonists as drugs have met with limited success up to this point, possibly because attributing the complex GPCR signaling and phenotypes to proximal transducers is too reductive. Rather, we believe that a systems pharmacology approach that investigates both proximal and distal aspects of GPCR signaling may unveil mechanisms underlying various pathologies and identify effectors and networks that can be targeted directly or through a single or multiple GPCRs. Although G proteins and β-arrestins are undoubtedly critical signaling effectors that regulate both normal and abnormal physiology, they are simply the tip of the iceberg that underlie biased GPCR signaling. The complexity of GPCR signaling networks presents an exciting opportunity to redefine how we as a research community define biased GPCR signaling and how we search for and design novel GPCR therapeutics.

GRANTS

This work was supported by the National Institute of General Medical Sciences Grants T32GM00717 (to D.S.E) and R01GM122798 (to S.R); the Duke University Medical Scientist Training Program (to D.S.E); and the American Heart Association Predoctoral Fellowship No. 20PRE35120592 (to D.S.E).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

This article is part of the special collection “Advances in GPCRs: Structure, Mechanisms, Disease, and Pharmacology.” Wei Kong, MD, PhD, and Jinpeng Sun, PhD, served as Guest Editors of this collection.

AUTHOR CONTRIBUTIONS

D.S.E. prepared figures; D.S.E., U.P., J.G., and C.H. drafted manuscript; D.S.E., U.P., J.G., C.H., and S.R. edited and revised manuscript; D.S.E., U.P., J.G., C.H., and S.R. approved final version of manuscript.