Novel Allosteric Modulators of G Protein-coupled Receptors*

G protein-coupled receptors (GPCRs) are allosteric proteins, because their signal transduction relies on interactions between topographically distinct, yet conformationally linked, domains. Much of the focus on GPCR allostery in the new millennium, however, has been on modes of targeting GPCR allosteric sites with chemical probes due to the potential for novel therapeutics. It is now apparent that some GPCRs possess more than one targetable allosteric site, in addition to a growing list of putative endogenous modulators. Advances in structural biology are also shedding new insights into mechanisms of allostery, although the complexities of candidate allosteric drugs necessitate rigorous biological characterization.

G protein-coupled receptors (GPCRs) are allosteric proteins, because their signal transduction relies on interactions between topographically distinct, yet conformationally linked, domains. Much of the focus on GPCR allostery in the new millennium, however, has been on modes of targeting GPCR allosteric sites with chemical probes due to the potential for novel therapeutics. It is now apparent that some GPCRs possess more than one targetable allosteric site, in addition to a growing list of putative endogenous modulators. Advances in structural biology are also shedding new insights into mechanisms of allostery, although the complexities of candidate allosteric drugs necessitate rigorous biological characterization.
G protein-coupled receptors (GPCRs), 4 also known as 7-transmembrane (7TM) receptors, are the largest superfamily of cell surface receptor proteins encoded by the human genome (1). These integral membrane proteins are highly dynamic and exist in an equilibrium between various functionally distinct conformational states (2). GPCRs fulfil the vital biological function of transducing a wide range of extracellular signals (e.g. photons, lipids, neurotransmitters, hormones, peptides, enzymes, ions, odorants) across the cell membrane into the cytosolic space. Physiologically, the process begins when an endogenous extracellular signal interacts with the primary ("orthosteric") binding site of GPCR, resulting in a conformational rearrangement that conveys the signal through the plasma membrane-spanning 7TM region and subsequently triggering intracellular signaling cascades via heterotrimeric G proteins and other accessory proteins (3). Because GPCR-mediated signaling systems are involved in regulating a multitude of physiological and pathophysiological processes, it is not surprising that the GPCR superfamily encompasses the targets of more actual and potential drugs than any other family of proteins (4,5).
To date, the majority of probe compounds and marketed drugs that target GPCRs are small molecules, but it is noteworthy that there is a growing interest in utilizing biologics and antibodies to target these receptors as well (6). In addition, although the mode of action of the bulk of GPCR-targeting agents remains orthosteric, the turn of the millennium has witnessed substantial efforts in alternative methods of modulating GPCR activity, specifically by targeting topographically distinct allosteric sites. This minireview discusses some of the key characteristics associated with GPCR allostery and ongoing challenges and opportunities in understanding and exploiting the phenomenon.

Characteristics of GPCR Allostery
Allostery is a widespread biological phenomenon that describes the ability of interactions occurring at one site of a macromolecule to modulate interactions at a spatially distinct binding site on the same macromolecule in a reciprocal manner. Since allosteric effects were first described in archetypal examples, such as the heme-heme interactions of hemoglobin, allostery has been acknowledged as a means by which proteins and other molecules (e.g. DNA) may amplify, attenuate, bias, and otherwise fine-tune their physiological functions (2,(7)(8)(9). Initial observations of allosteric phenomena in enzymes were mechanistically summarized first in the Monod-Wyman-Changeux (MWC) and subsequently in the Koshland-Nemethy-Filmer (KNF) models (10,11). Although the MWC model depicts allostery as a concerted process (i.e. conformational selection), and the KNF model describes it as a sequential process (i.e. conformational induction), each model reflects valid key aspects of the nature of allostery, which involves ligandmediated shifts in the population of pre-existing macromolecular conformational ensembles and resulting changes in the interactive properties of the new ensembles (12). In addition to enzymes, it became apparent that other protein classes, including GPCRs, possess many of the characteristics associated with allosteric proteins (13). GPCRs are conformationally dynamic proteins that act as conduits for the transfer of energy over a distance. Indeed, GPCR signal transduction is intrinsically allosteric as it involves the binding of an extracellular stimulus and subsequent propagation of the signal through the protein to a topographically distinct (e.g. ϳ50 Å) intracellular site recognized by G proteins, ␤-arrestins, and others. Moreover, because of the broad diversity of endogenous activators of GPCRs, an orthosteric region on one type of receptor (e.g. class A biogenic amine receptor) may represent an allosteric domain in another type of receptor (e.g. class B secretin family or class C glutamate family receptors) (14) (Fig. 1).
Perhaps not surprisingly, therefore, the attractiveness and tractability of GPCRs as drug targets, coupled with advances in drug screening at these receptors, have uncovered different types of allosteric GPCR modulators ( Table 1) that are generally classified operationally based on their modes of pharmacology. An allosteric ligand that potentiates an agonist-mediated receptor response is referred to as a "positive allosteric modulator" (PAM), whereas one that attenuates activity is known as a "negative allosteric modulator" (NAM) (15). Mechanistically, these effects can be achieved through modulation of the binding affinity of orthosteric ligands and/or through changes in the ability of the orthosteric ligand-occupied receptor complex to interact with intracellular transducer proteins. In contrast a ligand that binds at the allosteric site without affecting receptor or orthosteric ligand activity (at equilibrium) is classed as "neutral allosteric ligand" (NAL). "Allosteric agonists" are ligands that are capable of directly activating the receptor from an allosteric site even in the absence of an orthosteric agonist (15), whereas "PAM agonists" and "NAM agonists" display mixed modes of modulation and direct GPCR activation depending on the cellular context (15). More recently, "bitopic ligands" have also been described, which are defined as hybrid molecules possessing separate orthosteric and allosteric pharmacophores that concomitantly engage with their respective sites on a single GPCR to mediate novel pharmacology; several molecules initially classified as allosteric agonists in the literature have since been reclassified as bitopic ligands (15)(16)(17)(18).
There are a number of general characteristics associated with allosteric GPCR modulators that present both unique advantages over orthosteric ligands as well as challenges to successful detection or validation of allosteric compounds. The first characteristic is the potential for allosteric ligands to exhibit greater receptor subtype selectivity. This property has two potential origins: i) a decreased evolutionary pressure for sequence conservation within allosteric sites relative to the orthosteric site between GPCR subtypes (assuming there is no endogenous allosteric ligand for such a site, but see the next section) and/or ii) selective cooperativity with an orthosteric site at one receptor subtype while exhibiting neutral cooperativity at other subtypes of that receptor family (19). The nature of the cooperativity between orthosteric and allosteric sites on a GPCR also represents a second important characteristic of allostery that has practical and therapeutic implications. If the modulator displays minimal direct allosteric agonism in its own right, then it will act as a PAM or NAM only when and where the endogenous ligand is released, thus maintaining the natural spatiotemporal "rhythms" of the endogenous orthosteric ligand. Furthermore, very subtle degrees of positive or negative cooperativity (which may be all that is necessary for certain GPCRs and disease states) result in an allosteric "effect ceiling" that increases the likelihood of on-target safety in overdose situations, although this also poses a challenge for the screening of modulators with low degrees of cooperativity (20). The ability to achieve unprecedented modes of on-target selectivity and/or fine-tune endogenous responses as a consequence of pure PAM or NAM activity may prove particularly important in diseases where tight physiological regulation is vital, such as neurodegeneration, schizophrenia, diabetes, and endocrine disorders.
A particularly interesting phenomenon associated with allostery that has been most noted at GPCRs is the property of "probe dependence," wherein the magnitude and direction of an allosteric effect can change depending on the nature of the interacting ligands (15,21). Probe dependence has substantial implications for GPCR drug discovery and GPCR biology. For instance, many GPCRs have more than one endogenous orthosteric agonist, and totally different effects can be observed  (or missed altogether) depending on which agonist is used to activate the receptor in the presence of a given allosteric modulator (22). In addition, the allosteric nature of GPCR signal transduction means that the ability of a ligand-bound receptor to recognize cellular effector molecules is also subject to probe dependence, this time directed intracellularly rather than between two ligands. Specifically, the ability of orthosteric or allosteric GPCR ligands to stabilize different functionally relevant conformations of the same GPCR can give rise to the phenomenon of "biased agonism," whereby only a subset of the possible signaling repertoire of the receptor is recruited at the relative expense of other pathways (21,23,24). Although beyond the scope of the current minireview, the allosteric pharmacology of bitopic ligands also presents a unique set of characteristics. Because of their dual pharmacophore nature, bitopic ligands can provide both greater selectivity through interaction with an allosteric site and higher affinity through concomitant engagement of the orthosteric site. A judicious choice of orthosteric and allosteric building blocks can also yield bitopic ligands that display novel biased agonism (25). Although the spatiotemporal control of endogenous signaling afforded by pure allosteric modulator ligands is lost with bitopic ligands, the latter may prove particularly useful in situations where endogenous agonist tone is progressively lost (e.g. neurodegenerative disorders) (18).

Endogenous Allosteric Modulators
Although most studies of GPCR allostery have traditionally focused on the actions of exogenous allosteric modulators because of the implications for novel drug discovery (see below), these receptors can also be modulated by a variety of endogenous substances (26). As mentioned above, the best characterized allosteric interaction at GPCRs is the positive cooperativity exhibited between the intracellular G proteinbinding site with the orthosteric site (3). The GPCR intracellular face can also interact with ␤-arrestins: endogenous GPCR accessory proteins that were originally characterized as scaffolding proteins involved in the termination of GPCR signaling, internalization, and recycling of receptors. It is now recognized that ␤-arrestins can be involved in G protein-independent signal transduction (27)(28)(29)(30). These unique signaling properties emerge as a result of specific receptor conformations that have been shown to interact more readily with ␤-arrestins, leading to increased ␤-arrestin binding and biased signaling through noncanonical signaling pathways (31)(32)(33). Although beyond the scope of the current review, it is also widely acknowledged that different types of GPCRs have the potential to associate with other proteins, such as receptor activity-modifying proteins (RAMPs) or melanocortin receptor accessory proteins (MRAPs), or with each other in the form of homo-or heterodimers (or higher order oligomers) with novel pharmacological properties (26). In a number of such instances, cooperative interactions have been noted between orthosteric ligands within such complexes (34).
In addition to the ubiquitous allosteric sites utilized by G proteins and ␤-arrestins, individual GPCR classes or subclasses possess more specific allosteric sites that may be targeted by endogenous allosteric modulators. For example, pharmacological and crystallographic evidence has shown that sodium is vital for stabilizing the inactive state of many class A GPCRs via an allosteric site centered on a highly conserved aspartate residue, Asp 2.50 (26). Additionally, aromatic amino acids (e.g. L-phenylalanine, L-tryptophan, and L-tyrosine) bind at a site near to, but spatially distinct from, the orthosteric site within the "Venus flytrap" (VFT) N-terminal domain of the calciumsensing class C GPCR to potentiate the actions of extracellular calcium at a number of intracellular signaling pathways (35)(36)(37). A similar allosteric site has been shown to be located in the Venus flytrap of the metabotropic glutamate (mGlu) class C GPCRs. This site was initially reported as an allosteric chloridebinding site in the mGlu 1 subtype, but some synthetic small molecule agonists have since been reported to bind this site in mGlu 4 (38 -40).
As summarized in Table 1, a growing number of substances encompassing not only amino acids and ions, but also lipids, peptides, and proteins, have been proposed to act as putative endogenous allosteric modulators of different types of GPCRs  (26). It should be noted that in many of these instances, conclusive validation of an allosteric mechanism remains to be established, but the study of endogenous GPCR modulators may prove to be a fertile ground for uncovering novel biology in health and disease. For instance, the nature and composition of endogenous lipidic substances and numerous types of peptides and proteins can vary dramatically in inflammation; if it can be shown that some of these substances are bona fide allosteric modulators of specific GPCRs, then this may represent a novel avenue for understanding and targeting GPCR functionality in inflammatory disease (26). In addition, GPCR-directed autoantibodies have been identified as potential endogenous allosteric ligands related to a number of chronic disease states. In particular, autoantibodies have been identified in patients with a variety of cardiovascular diseases, including autoantibodies against the AT 1 receptor, the ␤ 1 -adrenergic receptor, the M 2 muscarinic acetylcholine receptor (mAChR), the ␣ 1 -adrenergic receptor, the ET A receptor, and the 5-HT 4 receptor (26). Additionally, the central nervous system can be affected by autoantibody activity. For instance, autoantibodies against GABA B have been found in the cerebrospinal fluid of limbic encephalitis patients, whereas patients with basal ganglia encephalitis with dominant movement and psychiatric disease present with autoantibodies against the dopamine D 2 receptor (26).

Exogenous Allosteric Modulators
Numerous synthetic small molecules with an allosteric mode of action have been reported for GPCRs ( Table 1). The mAChR family is arguably the most well studied class A GPCR system in this regard (41). Indeed, the first examples of GPCR allosteric modulators, the alkane bis-ammonium family of ligands, were identified at the mAChRs (42). Since that time, allosteric modulators of nearly every mode of action have been found to target the mAChRs, including PAMs, NAMs, PAM and NAM agonists, and bitopic ligands (17,(43)(44)(45)(46)(47)(48). Interestingly, most of these ligands bind to the mAChRs at a shared site (albeit with different affinities depending on the subtype), often referred to as the "common" allosteric site and highlighting how the targeting of a common allosteric domain can yield markedly different biological behaviors (49,50). However, evidence has been provided for the existence of a second allosteric site on the mAChRs, recognized by indolocarbazole analogues of staurosporine (e.g. KT5720 and KT5823) and the benzimidazole derivatives WIN63577 and WIN51708. As noted with the "common site" modulators, the second-site compounds also demonstrated positive, negative, or neutral interactions with acetylcholine depending on the mAChR subtype, but showed largely neutral interactions with the binding of common site modulators, further supporting the presence of at least two allosteric sites on a single GPCR (51)(52)(53)(54).
The notion that a single GPCR may possess more than one allosteric site for exogenous small molecules is likely more widespread. For instance, a number of allosteric modulators have been described for the gonadotropin-releasing hormone receptor, including Furan-1 and its derivative, FD-1 (55), as well as a series of amiloride derivatives (56). Interestingly, interaction studies between the two different classes of allosteric ligand revealed neutral cooperativity, despite each class indi-vidually modulating orthosteric ligands (56), again suggestive of more than one allosteric site on this receptor family. The intracellular face of GPCRs may also harbor allosteric sites for selective small molecules. For instance, binding sites for allosteric antagonists of the CXCR2, CCR4, and CCR5 chemokine receptors have been identified among the intracellular loops (57)(58)(59). Intracellular allosteric sites are also utilized by pepducins: cellpenetrating, lipidated peptides that target the intracellular loops. By tailoring the peptide's design to the intracellular domain of a particular GPCR, researchers have been successful in producing pepducins selectively targeting a number of GPCRs involved in inflammatory diseases, including proteaseactivated receptors (PAR1, -2, -4) and chemokine receptors (CXCR1, -2, -4) (60,61).
Class B GPCRs, such as the secretin, glucagon, and glucagon-like peptide 1 receptors, have proven notoriously intractable to small molecule discovery and have remained an area of intense research with regard to allosteric drugs. One common observation is the discovery of direct-acting allosteric agonists, in addition to PAM agonists or NAMs (62). Pure class B GPCR PAMs have thus far remained relatively more elusive, but whether this reflects a fundamental property of this class of GPCR or the relative immaturity of detailed studies of Class B allostery remains to be determined. In common with the preceding examples, however, there is clear evidence suggesting the presence of more than one allosteric site on Class B GPCRs (63,64).
Within the class C GPCRs, the mGlu receptors have a very rich allosteric pharmacology. In particular, allosteric modulation of the mGlu 5 receptor has been of interest as a target for the treatment of schizophrenia and Alzheimer disease (65,66). As such, multiple small molecule PAMs have been developed for mGlu 5 (67)(68)(69)(70)(71)(72), with the bulk acting at a common site in the upper region of the 7TM domain often termed the "MPEP-binding site" after the prototypical mGlu 5 NAM, MPEP (65,66). However, several ligands have been identified that are non-competitive with the MPEP site, including the PAMs VU357121 and CPPHA, indicating the presence of at least two distinct allosteric sites (67,69,73,74). Further investigations have shown that PAMs acting on each of these sites may have varied effects on different signaling pathways, indicating that allosteric modulation of discrete allosteric sites may have significant effects on the response of the GPCR (75,76).
Finally, the Class F GPCR family, which includes the Smoothened and Frizzled receptors, has also been considered difficult to target with small molecules, but recent breakthroughs have revealed that these receptors possess allosteric binding sites for exogenous ligands. For example, a series of small molecules that were originally designed to act as pharmacological chaperones for a misfolded mutant of Frizzled4 were subsequently identified as novel allosteric modulators of the wild-type form of Frizzled4; the binding site for these compounds was proposed to be located in the vicinity of intracellular loop 3 of the receptor (77). The recent crystal structure of the Smoothened receptor bound to the allosteric modulator, Sant1, also revealed an allosteric pocket that is located deep within the transmembrane-spanning cavity of the receptor, toward the cytosolic end; this is in contrast to the binding site that is closer to the extracellular entrance and utilized by canonical ligands of this receptor (78).
It is important to note that although small molecules have made an undeniable impact on the study of allostery, there is growing interest in utilizing biologics to target allosteric sites. Thus far, allosteric antibodies have shown the most promise in targeting receptor tyrosine kinases, such as the insulin receptor, where the monoclonal antibody XMetA allosterically binds and activates the receptor (79,80). A subcategory of allosteric antibodies known as "allosteric ligand-modifying antibodies" (ALMA) acts by binding the endogenous ligand prior to interacting with the receptor (15). One example is gevokizumab, an anti-interleukin-1␤ antibody that binds interleukin and alters the conformation of the endogenous ligand. Subsequently, the gevokizumab-interleukin-1␤ complex binds in a characteristic ternary complex with the receptor (81,82). Although examples of rationally designed allosteric antibodies targeting GPCRs have yet to be described, the novel mechanisms of allosteric antibodies and allosteric ligand-modifying antibodies offer a valuable new strategy for the study of GPCR allostery in the future.

Chemical Biology Challenges in Designing Allosteric Modulators of GPCRs
The production of highly site-specific allosteric probe molecules is a crucial factor in the study of allosteric interactions. Structure-based drug design is still a relatively nascent field with respect to GPCRs, although substantial progress has occurred in recent years (see below). Within both industry and academia, a large number of resources have been dedicated to the generation, collection, and curation of large libraries of natural and synthetic small molecules from which to screen for and optimize novel allosteric ligands, and this remains the major source of such compounds (83,84). Given this significant investment, the methodologies with which these novel ligands are interrogated are central to any research campaign. The field of chemical biology has provided a number of tools to probe allosteric mechanisms of GPCR structure and function. These techniques may provide the biochemical and biophysical information necessary to garner a clear understanding of the structural dynamics and signaling behavior of allosteric interactions on a chemical level. Techniques such as bioluminescent and Förster resonance energy transfer as well as single-molecule detection fluorescence have gained popularity for the study of GPCRs and validated binding partners. However, there are limitations for using these techniques in the process of screening for novel ligands, not least of which is the challenge of applying these methods to a high throughput screening format (85,86). Measurement of second messengers of G protein signaling (e.g. calcium and cAMP) or ␤-arrestin recruitment has proven more successful in this respect; nevertheless, it remains challenging to develop selective small molecule allosteric modulators.
For example, allosteric ligands often possess delicate structure-activity relationships. That is, in a given chemical series, a seemingly minor modification to a molecule's steric or electronic properties often leads to the complete ablation of its activity. Similarly, certain chemical series demonstrate "mode switching" whereby small modifications to the structure can result in dramatically changed pharmacological profiles (87). These concerns are particularly relevant when further modifications are applied to the probes, for instance in the generation of irreversible or photoactivatable allosteric molecules (88,89), or through the use of these chemical probes in vivo, where biochemical transformations undertaken by metabolic processes may alter a modulator's potency, cooperativity, receptor selectivity, or mode of action.
Novel chemotypes of allosteric modulators also pose challenges to pharmacological characterization. For instance, many allosteric modulators display phenomena such as probe dependence, and can impose biased signaling on the actions of orthosteric ligands ( Table 2). These may go unnoticed with the use of a single screening methodology. Only through the measurement of an allosteric ligand's effect on multiple downstream pathways in the presence of all relevant endogenous orthosteric ligands can a compound's properties be fully elucidated, and this must be factored into all allosteric discovery programs (90). Further complexities in characterizing a novel allosteric ligand may be caused by differences between species' receptor isoforms, differential effects on orthosteric ligand affinity and efficacy, varied kinetics between different signaling pathways, and the possibility of undetected endogenous allosteric interactions. Therefore efforts must be made to fully characterize and validate novel chemical probes in as many experimental paradigms as possible to compose the clearest picture of the compound's properties.

What Can Structural Biology Reveal About Allosteric Mechanisms?
Although the use of selective allosteric probe molecules and functional assays can reveal much about allosteric pharmacology, the information provided by crystal structures and other high resolution approaches is invaluable for more direct, molecular level insights into GPCR allostery. Since the turn of the millennium, when the first high resolution GPCR crystal structure was solved for bovine rhodopsin (91), there has been a seemingly exponential growth in the number of crystal structures solved for a range of class A, B, and C GPCRs (Fig. 2), although the majority of these rely on some form of protein engineering to improve stability and crystal formation. This has allowed the observation of conformational snapshots adopted by structures co-bound to ligands and/or interacting proteins, offering an unprecedented view into the fundamental structural basis for receptor function.
Of note, there have been a number of inactive state crystal structures of GPCRs in binary complexes with small molecule NAMs. These include the chemokine receptor CCR5 bound to maraviroc, the corticotropin-releasing factor receptor CRF 1 bound to CP-376395, the mGlu 1 bound to FITM, the mGlu 5 bound to mavoglurant, and the aforementioned Smoothened receptor bound to Sant1 (78,(92)(93)(94)(95). Despite providing new insights into the binding behavior of these NAMs, the inactive and binary complex natures of these structures do not capture the complete structural mechanisms that underlie cooperativity between the allos-teric and orthosteric sites. Indeed, the CRF 1 and mGlu structures lack the key N-terminal domains that constitute much (or all) of the orthosteric binding site.
A major breakthrough in recent years has been the crystal structure depicting the classic ternary complex of an orthosteric agonist binding the ␤ 2 -adrenergic receptor coupled to a G s protein (96). This structure offered the first observation of how a GPCR orthosteric site is allosterically coupled to G protein activation. In terms of small molecule allostery, the recent structure of an active M 2 mAChR in complex with the high efficacy orthosteric agonist, iperoxo, and the PAM, LY02119620 (97), also represented a major structural biology advance in understanding allostery at a GPCR. For example, the structure is consistent with the predictions of the MWC model of allostery in that the PAM preferentially recognizes and stabilizes a preformed active state of the receptor. Nonetheless, these structures still represent first steps in our molecular level understanding of mechanisms underlying GPCR allostery. Phenomena such as biased allosteric modulation, probe dependence, and the actual mechanisms underlying transmission of cooperativity remain challenging as they require the ability to capture multiple states in the absence and presence of multiple ligands. In the meantime, additional insights into the structural and dynamic mechanisms of allostery are being obtained via other methods. For example, the inactive M 2 mAChR structure has been subjected to long timescale molecular dynamics simulations with a diverse range of small molecule allosteric mod-ulators, in the absence or presence of an orthosteric ligand, to identify binding poses and mechanisms underlying cooperativity for a broad set of NAMs (98). The use of NMR has also allowed researchers to study the conformational flexibility of the receptor as a whole, providing more information on receptor dynamics than can be obtained through static crystal structures, although no study has directly applied this approach to GPCR allosteric modulators to date (99,100).

Conclusions
Allosteric modulation of GPCRs is now a widely accepted phenomenon with substantial implications for novel drug discovery, yet many fundamental issues remain to be addressed. For example, it is now clear that a single GPCR can possess more than one allosteric site, but whether the targeting of such sites can lead to differential behaviors remains unknown. The prevalence of endogenous allosteric modulators remains to be determined, but they may prove to be a mechanism of tissuespecific regulation in normal physiology or disease that may be amenable to chemical manipulation. The ascendance of biologics as therapeutics opens new vistas for targeting of GPCRs with a greater degree of specificity than previously possible, yet the extent with which such substances can interact allosterically with GPCRs is largely unexplored. As more detailed GPCR structural information becomes available, it too will profoundly affect our understanding of allostery in GPCRs as well as the manner by which allosteric molecules are designed. In prepar-

Ligand
Receptor Bias profile (and orthosteric agonist probe) ing for such an eventuality, the rigorous biological characterization of each new allosteric site and allosteric ligand discovered remains paramount.
Author Contributions-All authors wrote the manuscript.