G protein coupled receptors (GPCRs) constitute one of the largest gene families in the human genome. Due to their critical role in cell-cell communication and as causal agents in disease, they are a target for the development of numerous drugs, many of which have received regulatory approval. Emerging knowledge of GPCR physiology has begun to alter our approach to GPCR drug discovery and is providing insight into how to discover new generations of compounds as human therapeutics [1,2].

GPCRs are integral cell membrane proteins and their association with membrane lipids is critical for conferring, and maintaining, receptor structures required for function. The phospholipid environment controls the inherent flexibility of the receptor mainly due to the structural disorder of the third intracellular loop, a domain known to undergo major conformational changes following agonist binding and to be critical for G protein coupling [3,4]. The pre-requisite for lipid association and structural flexibility hitherto hindered the growth of GPCR crystals needed for X-ray diffraction analysis. This hurdle has recently been overcome with resolution of crystal structures for several GPCRs [4-11], which in turn, allows rational drug design to be employed in GPCR drug discovery and thus novel GPCR drug structures to be identified [12-14].

Unlike most enzymes that are energetically unidirectional (e.g. kinases phosphorylate proteins selectively; proteases hydrolyze selective substrates), GPCR activation can produce a greater diversity of cellular functions as they interact with multiple Guanine nucleotide binding proteins (G proteins) to regulate discrete signaling pathways [15-21]. G proteins bind to allosteric sites on the receptor that are distal to the orthosteric, ligand binding pocket [22]. Furthermore, binding sites exist on the extracellular face of the receptor, distinct from the orthosteric binding domain, representing sites that can be targeted for allosteric drug development. Consequently, a unique opportunity now exists to develop drugs selectively acting at those allosteric sites that modulate GPCRs in a very different manner from archetypal orthosteric ligands.

Some drugs of this type are generically referred to as allosteric modulators, whilst others are referred to as ‘dualsteric’, being molecules that bind both to orthosteric and allosteric sites [23-28]. Recent research has also led to the development of pepducins [29]; cell penetrating peptides that mimic or abrogate GPCR interaction with specific G proteins to induce a discrete cellular response. This new class of GPCR ligands provides several favorable properties over typical orthosteric GPCR agonists, inverse agonists and antagonists.

GPCRs are also known to produce constitutive activity, that is activity in the absence of endogenous ligand or agonist [1,2,29-34]. Constitutive activity may be considered to be an equilibrium state of a GPCR between its ground state of no activity and its fully activated state. Agonists can still bind to constitutively active GPCRs to produce further agonism. However, because of the high basal activity, some compounds can bind to constitutively active receptors and produce opposite responses to agonists. For example, if an agonist at a particular GPCR stimulates an increase in Ca⁺⁺ mobilization and the constitutively active receptor has high basal Ca⁺⁺ mobilization, some ligands are able to bind to this receptor to reduce basal Ca⁺⁺ mobilization down towards the ground state. These ligands have been referred to as ‘inverse agonists’. In many cases, these ligands were originally characterized as antagonists on non-constitutively active GPCRs and their inverse agonism is primarily seen when they interact with constitutively active GPCRs. The existence of constitutive activity has required a reconsideration of how we think about GPCR antagonists. Antagonists are classically considered compounds that bind to GPCRs to block binding and activation by agonists, but produce no activity of their own. In fact, compounds that block both agonist and inverse agonist binding but produce no response on GPCRs are now referred to as ‘neutral antagonists’.

Constitutive activity of GPCRs is believed to be due to their formation of oligomers in addition to single unit assemblies [29-34]. In many immortalized cell systems GPCRs are over-expressed, and thus form homodimers. They may also form heterodimers in which two different GPCRs associate. In some cases dimers exhibit different pharmacological properties than seen with monomers. Notably they may differ in agonist binding due to the formation of a unique orthosteric binding domain. Previously, dimers were considered to only exist in artificial cell systems classically used in GPCR drug discovery. However, emerging evidence suggests that GPCR oligomerization also occurs in endogenous tissues [31]. Consequently, discovery programs aimed at targeting GPCR oligomerization may result in drugs selective for discrete GPCR oligomers, providing unique, and potentially desirable, pharmacological properties in contrast to compounds targeting receptor monomers only.

This review will focus on these new avenues of GPCR drug discovery. We also describe briefly the scientific basis of discoveries leading to these new fields of drug discovery. Finally, we provide information on the potential of novel drugs, as well as new technologies used to screen for this new generation of GPCR compounds.

Structure-based, rational drug discovery is widely employed by the pharmaceutical industry to design compounds with optimal specificity and potency to molecular targets, often at soluble enzymes such as protein kinases. The structural information provides information on the physical properties of the ligand binding site(s) and helps to design pharmacophore models of the ligand when docked. This provides the foundation for screening structural families of molecules with predictable chemical characteristics as well as optimal functional properties to modulate the molecular target [35].

This approach has not been used widely in the GPCR field, because GPCRs are not easily crystalized for structural evaluation. However, within the last decade, six different GPCRs have been crystalized in their inactive form, including the avian β₁- adrenoceptor [8], human β₂-adrenoceptor [4-6], human adenosine A_2A receptor [7], human dopamine D3 receptor [9], human CXCR4 receptor [10], as well as rhodopsin [11]. The structural data from these studies not only provides critical information on the molecular nature of drug interaction with GPCRs, but also provides approaches that allow for crystallization of other GPCRs, revealing a more rational method for drug development.

Rhodopsin was the first full receptor to be subjected to X-ray analysis [11,36-40]. However, the analysis of more ‘druggable’ GPCRs only occurred with the development of approaches to grow the entire β-adrenoceptor as crystals [3,4,41]. GPCRs are difficult to grow as crystals because they are structurally unstable in purified form. Kobilka and Deupi [3] attributed this instability to the unordered structure of the large third intracellular loops of the receptors. To overcome this problem, Rosenbaum et al. [4] replaced the large third intracellular loop of the β₂-adrenoceptor with T4 lysozyme, a sequence that is more structurally rigid and polar than the native intracellular loop of the receptor. This conferred greater stability on the receptor as a whole and had the advantage that the mutant receptor could be expressed at high levels in a crystal form for X-ray spectroscopy. T4 lysozyme was likewise inserted into the CXCR4 and dopamine D3 receptors to generate crystals [9,10]. For studies on the β₁-adrenoceptor, Warne et al. [8] truncated the termini of the receptor to reduce proteolytic cleavage and made mutations of the receptor that increased thermal stability to allow for crystallization as was also done on the D3 receptor [9]. In all of these cases, to generate GPCR crystals, the functional moieties of the receptor needed for G protein coupling were altered to form a protein sufficiently stable for the rigorous physical procedures needed for crystallization. These mutant receptors could be analyzed by radioligand binding and possessed expected affinities for antagonists and inverse agonists. As the receptors were unable to couple to G proteins, to establish functionality Rosenbaum et al. [4] inserted flurophores in cysteine residues in transmembrane 6 of the receptor. These allowed detection of conformational changes induced by ligand binding as a measure of activity of the mutant receptor. They found that the full agonist, isoproterenol, produced a greater change in fluorescent intensity than the partial agonist, salbutamol while inverse agonists induced no or little conformational shift on the β₂-adrenoceptor.

These authors then conducted structural analysis of the mutant receptor bound to the high affinity inverse agonist, carazolol which further stabilized the receptor. This allowed for analysis of the physical sites and amino acids critical for ligand binding particularly in the hydrophobic pore of the receptor. Similar studies have now been done on other druggable GPCRs. These studies revealed common structural requirements in ligand binding to the orthosteric domain in the transmembrane core. Such information, along with the extensive physical data on these and other GPCRs is now being used to develop computational models of the nature of small molecules binding to GPCRs, thus facilitating medicinal chemistry efforts to discover novel drugs.

To this point, Kolb et al. [42] used crystal structure information of the β-adrenoceptor as a basis for virtual screening of a chemical library of a million compounds. They employed a molecular docking program (DOCK3.5.54) for their studies, which assesses compounds for interaction at a binding site on the receptor. From the virtual screen, 25 compounds were selected and then tested functionally on the receptor. Some bound with nanomolar affinities and six were identified as potent inverse agonists. Since the crystal structure analysis is biased towards evaluation of carazolol binding, this finding was not unexpected. Similarly, Michino et al. [14] have recently described a docking model based on physical analysis of the adenosine A2 receptor referred to as “GPCR DOCK 2008” which can be employed for similar virtual screening. Presumably, as more extensive structural information arises on the GPCRs, these analyses will be further refined and possibly broadened to facilitate binding of other ligands such as agonists.

In fact, Ivetac and McCammon [13] have recently mapped potential allosteric binding sites on the β₁- and β₂-adrenoceptors using the structural information from the crystallographic studies and the FTMAP algorithm employing molecular dynamic modeling. Essentially, the computational analysis identifies binding pockets on the receptor surface that could constitute allosteric sites for ligand interaction. Similarly, using the crystal structure analysis of rhodopsin, Taylor et al. [45] employed virtual screening using a peptidomimetic structure which experimentally had been shown to stabilize the rhodopsin-G protein state. A pharmacophore model was developed from the peptidomimetic to screen scaffolds of commercial and publicly available small molecule libraries. Three of the compounds found to bind to rhodopsin in experimental studies interacted with allosteric sites, suggesting that these computational approaches can be employed in a rational approach to identify allosteric compounds binding outside the orthosteric core of GPCRs.

The studies on the crystal structure analysis of the mutant GPCRs provides important information on the static structure of ligand binding to GPCRs. However, the information gained has its limits, in particular with regards to how agonists cause activation. While Rosenbaum et al. [4] actually reported that agonists bound to the mutant β₂-adrenoceptor with high affinity, they attributed this to the formation of a constitutively active nature of the receptor conferred by the mutations. The absence of the G protein binding domains of the receptor, a critical allosteric modulator of agonist binding, puts limits on the utility of the structural analysis of agonist binding for rational drug design. However, crystal structure analysis is not the only physical method used to study GPCRs. In fact, to better investigate the conformational changes induced in the wild-type β₂-adrenoceptor by agonist binding, Bokoch et al. [43] for the first time employed NMR spectroscopy on a full length druggable GPCR. At around the same time, nuclear magnetic resonance (NMR) analysis of rhodopsin was also described by Gautier et al. [44]. Unlike crystal structure analysis, NMR can provide greater insight into protein dynamics of ligand binding and is extensively used to provide critical information for rational drug design against other molecular targets such as kinases.

Bokoch et al. [43] focused their studies on investigating dynamic changes occurring around a natural salt bridge formed between extracellular loops 2 and 3 of the β-adrenoceptor that make up the extracellular limits of the transmembrane core of the orthosteric binding domain. The authors reasoned that agonist induced conformational changes would be expected in this region based on the crystal structure analysis of carazolol binding site as well as from fluorescent spectroscopy studies done in the past by this group on this receptor in studies on agonist binding. Importantly, they speculated that agonists would likely either break the salt bridge or alter the conformation around this region of the receptor. Furthermore, they suggested that studies on the surface dynamics of the receptor using NMR would be important in understanding the role of extracellular loops of the receptor in mediating conformational changes needed for receptor activation, and that those regions might be fertile sites for allosteric ligand binding and regulation.

In fact, they noted that extracellular loops may be important sites for discovery of better and more efficient receptor subtype selective drugs. For example, the orthosteric binding domains of the β₁- and β₂-adrenoceptors are almost identical with 15 or 16 amino acids being the same. In contrast, the extracellular loops of the receptor differ greatly. If the loops are critical in transducing orthosteric agonist induced conformational changes, then small molecules binding to those allosteric sites might be expected to either serve to block or enhance agonism without directly affecting agonist binding and since the loops differ substantially in sequence, then drugs targeting those sites may have greatly enhanced subtype selectivity.

Bokoch et al. [43] did NMR spectrum analysis of the β₂-adrenoceptor by labeling a lysine group near the salt bridge with ¹³C-methyl group which served as a probe for the NMR analysis. This modification did not alter the crystal structure of the receptor nor did it affect ligand binding or G protein coupling of the wild-type receptor. Crystal structure analysis showed that the inverse agonist carazolol interacted with a phenylalanine (Phe) at position 193 in the receptor, adjacent to the start of the salt bridge of the receptor and carazolol produced a shift in conformation of the second and third extracellular loops. Other inverse agonists produced similar shifts and the authors speculated the conformational shifts were due to the tricyclic or bicyclic aromatic rings of the inverse agonists interacting with Phe¹⁹³. In contrast, the neutral antagonist alprenolol, which has only a single aromatic ring, did not cause a conformational shift because this molecule is unable to effectively interact with Phe¹⁹³ in the carazolol binding pocket. This information provides structural clarity on the nature of differences in neutral antagonist and inverse agonist binding to this receptor.

Agonist binding also induced conformational changes in the receptor that differed considerably from inverse agonist binding. The agonist binding causes a shift in the structure of the salt bridge connecting extracellular loops 2 and 3, focused on lysine (Lys) 305, the other end of the bridge from the Phe¹⁹³. Agonists appear to cause movement of the two extracellular loops. In contrast, inverse agonists appear to stabilize the salt bridge at Phe¹⁹³ to reduce movement of the extracellular loops. Thus, in the model developed, the unbound GPCR in a basal state exerts flexibility in its third intracellular loop and extracellular loops, and can couple to G protein to produce basal GTPase activity. Inverse agonists stabilize the salt bridge to prevent basal conformational shifting of the receptor to uncouple from G protein to inhibit basal GTPase activity. Agonists cause profound movement of the extracellular loops to orientate the receptor to a most favorable conformation to couple to G protein to increase G protein coupling.

Most importantly, the authors [43] speculated that ligands binding to the extracellular surface of the receptor, distinct from the orthosteric binding pore but within regions of extracellular loops 2 and 3 could modulate the conformational shifts induced by agonists, either potentiating (allosteric activator) or by blocking (allosteric inhibitor). This suggestion is consistent with data from modeling studies by Ivetac and McCammon [13] who showed potential binding pockets at the mouth of the orthosteric binding pore of the β-adrenoceptor. They indicated that similar allosteric sites have been identified in muscarinic receptors, and ligand binding to these pockets could act by blocking ability of orthosteric ligands from gaining entrance to the orthosteric pore, acting as a sort of gate keeper. In effect, allosteric inhibitors could stabilize a “closed” form of the GPCR preventing access of orthosteric agonists.

The model however does not explain how allosteric activators would work. Consistent with the NMR analysis of Bokoch et al. [43], such activators could bind to regions near extracellular loops 2 and 3 to facilitate conformational shifts induced by agonists to enhance G protein coupling and could also explain the actions of dualisteric ligands, which are able to bind within the orthosteric pore to activate receptor and also interact with surface regions of the receptor to tweak conformational changes needed for receptor stimulation. While the authors acknowledge that they have limited information on the generality of their data to other GPCRs, the data provides a useful model to begin to design allosteric modulators of GPCRs and implies that NMR could be used to directly test efficacy of such compounds.

The focus of most of the crystal structure and NMR analysis has been on receptors which bind endogenous ligands which are small molecules, such as dopamine, epinephrine, norepinephrine and adenosine. The models and information gained from these studies show how ligands interact with a central hydrophobic core of the GPCRs. However, a large number of GPCRs have large endogenous ligands and those molecules do not easily fit into a hydrophobic binding pocket first described for β-adrenoceptor. In fact, it has been shown that for peptide receptors, such as the opioid receptors, the endogenous ligands, including enkephalins, endorphins and dynorphins appear to primarily bind to orthosteric domains in the extracellular loops of the receptors [46-49]. For example, dynorphin A, the endogenous ligand for the kappa opioid receptor has been shown by mutagenesis studies, NMR analysis and molecular modeling to bind to the second extracellular loop of the kappa receptor [49-54]. In contrast, small molecule selective antagonists bind to different regions [46,57]. For other opiate receptors, the N-terminus and extracellular loops 1 and 3 appear to be critical regions involved in selective peptide agonist [55,56,58,59-61].

Importantly, the recent crystal structure analysis of the CXCR4 receptor bound to a small molecule antagonist and a cyclic peptide show the ligands bind to pockets more closely associated with the extracellular domains of this GPCR [10]. The endogenous ligands for CXCR4 receptors are chemokines which are small proteins which would not be expected to easily fit into a deep hydrophobic pocket as found with the adrenoceptors. The structural studies support the notion that differences exist in the manner by which peptide or protein endogenous ligands bind to GPCRs compared to how endogenous small molecule ligands bind to their receptors.

In fact, NMR has been employed to study the binding of a number of different endogenous peptides and their agonists to fragments of GPCRs (not the entire receptor) [53,54,62]. For example, CCK8 was shown to bind to the N-terminus and extracellular loop 3 of the CCK1 and CCK2 receptors by NMR analysis [63,64] which provided information on the specific amino acids in these extracellular domains that the ligand makes contact with and the helical nature and structures conferred on the receptor upon binding. For these receptors it was also found that non-peptide agonists had overlapping binding domains in extracellular loop 3 with CCK8 [65].

Similarly, for Class B1 GPCRs which include the peptide receptors for corticotropin releasing factor (CRF), glucagon-like peptide-1 (GLP-1), gastric inhibitory peptide (GIP), pituitary adenylate cyclase activating peptide (PACAP), parathyroid hormone (PTH) and calcitonin gene-related peptide (CGRP), both crystal structure analysis and NMR studies on fragments of the receptors have shown that both endogenous ligands and peptide antagonists bind to the N-termini of the receptors [66-71]. In the case of CRF, binding to the N-terminus of the CRF1 receptor induces helical formation of the peptide ligand and binding to the N-terminus of the receptor occurs along the hydrophobic face of the ligand [67,68]. The binding of CRF to the N-terminus induces a conformational change in the receptor to orientate the N-terminal segment of the ligand to make contact with the extracellular loops of the receptor needed for receptor activation. Thus, these physical studies clearly show that binding of endogenous ligands to peptide receptors have different orthosteric binding domains than small molecule hormones binding to GPCRs such as the β-adrenoceptor.

This however, does not negate the important nature of the crystal structure and NMR information on the β-adrenoceptor nor its application in understanding how GPCRs work in general. In fact, the studies by Bokoch et al. [43] and Ivetac and McCammon [13], suggest that multiple binding sites, both in the hydrophobic core and the extracellular loops can influence both activation and inhibition of GPCRs. Thus, while sites in the extracellular loops of the β-adrenoceptor may provide targets for allosteric modulation and activation, sites in the extracellular loops of peptide receptors may serve as primary sites of orthosteric ligand binding. These extracellular sites can provide a rich source of targets for new small molecule drug discovery at either set of receptors. Thus, physical analysis and modeling based on the β-adrenoceptor is likely to be extremely useful in development of novel ligands and drugs acting upon peptide receptors and GPCRs in general.

Most drugs developed against GPCRs bind to the orthosteric ligand domain. However, as described above, GPCRs have many other extracellular facets besides their orthosteric binding domains that can affect receptor activity and function that can serve as drug binding targets [22-28]. Furthermore, intracellular domains of GPCRs couple to G proteins and interact with a variety of protein kinases and the β-arrestin family [72]. These intracellular sites are physically distinct from the orthosteric binding domain and can bind allosteric modulators. Small molecules that bind to allosteric regions of GPCRs have properties distinct from classical orthosteric ligands and can provide potential therapeutic advantages over classical GPCR drugs [1,2,22-24]. There are at least five unique properties of allosteric modulators. First, they can be highly selective for individual GPCRs because the structures of their binding pockets are unique for each receptor. For example, orthosteric binding domains of subtypes of receptors for some neurotransmitters, such as serotonin, dopamine and acetylcholine, can have similar affinities for the endogenous ligand. For this reason, it can be difficult to develop subtype selective drugs targeting the receptor subtypes because the subtypes have similar ligand binding proteins. However, allosteric sites on receptor subtypes do not generally have endogenous ligands. As a consequence, the determinants for their binding sites may be different amongst receptor subtypes providing approaches to more easily develop subtype selective ligands.

Similarly, some GPCR subfamilies have unique properties that make the discovery of orthostatic agonists or antagonists difficult to accomplish. An example is the protease activated receptor (PAR) subfamily which mediates the biological functions of thrombin. Thrombin is a protease and a major physiological factor in blood coagulation and wound healing. Thrombin activates PAR through a unique mechanism in which the protease cleaves the N-terminus of the receptor and the resulting truncated receptor bends back upon the receptor to serve as a tethered ligand to cause continuous activation. Efforts to develop anti-thrombin drugs have primarily focused on inhibiting thrombin activity or in some cases generating PAR receptor orthostatic binding antagonists. Because of the unique manner in which thrombin binds to PAR, development of small molecule antagonists and agonists has proven difficult. However, allosteric enhancers or inhibitors need not bind to PAR in the same manner as thrombin and therefore are not constrained in their receptor interaction as orthostatic agonists or antagonists. As described below, allosteric PAR modulators are beginning to be developed that provide a unique family of drugs targeting this clinically important GPCR family.

Secondly, allosteric ligands do not compete with endogenous transmitters or hormones like orthosteric ligands. As a consequence, allosteric modulators generally can be employed at lower dosing than orthosteric ligands to be effective. Thirdly, allosteric modulators are less likely to desensitize like orthosteric agonists. For allosteric activators, this property means they might be less prone to tolerance development, which plagues many orthosteric agonists, such as in the case of opiate analgesics. Fourthly, allosteric modulators tend to only effect the agonist occupied GPCR without causing actions of their own, reducing toxicity potential. As a consequence, allosteric modulators are usually found to have a ceiling effect, since when the receptor is fully occupied by an orthosteric agonist, the allosteric modulator can produce no further effect, no matter what concentration is employed. This allows allosteric modulators to temper activity of receptors rather than simply turning them off or on, providing a means to fine-tune GPCR activity.

Finally, allosteric modulators have the capacity to affect some functions of an individual GPCR and not others. In fact, as described by Rajagopal et al. [72] there are already examples of orthosteric agonists producing biased agonism by directing the receptor to selective G protein activation or distinct signaling pathways such as the β-arrestin pathway. For example, most muscarinic agonists can activate both G_s and G_q signaling pathways. However, some muscarinic orthosteric agonists only stimulate G_q signaling [72,73]. Similar differentiation of signaling pathway activation is found with angiotensin receptors where some ligands activate G proteins while others, acting through the same receptor, only stimulate β-arrestin signaling pathways involving mitogen-activated protein kinases (MAPKs), SRC, nuclear factor-κB (NF-κB) and phosphoinositide 3-kinase (PI3K). Thus, bias agonism can come about when the receptor is directed to couple with some but not all of their signaling pathways through different conformational changes.

In a similar manner, allosteric modulators can block or enhance GPCR coupling to one G protein and not affect coupling to other G proteins or signaling pathways. Furthermore, allosteric modulators that can prevent β-adrenoceptor kinase (βARK) and β-arrestin interaction with GPCRs may prevent desensitization, thereby providing an approach to prolong the activity of orthosteric agonists.

One of the most obvious examples of allosterism that focuses on G protein coupling domains of GPCRs is found with pepducins. These are ligands corresponding to specific sequences of the intracellular domain of a particular receptor designed to either block receptor coupling or mimic receptor coupling to a particular G protein to produce desired cellular effects. The compounds are synthesized with membrane anchoring segments which allow the compound to enter cells and localize to the cell membrane to be in close proximity to the GPCR target. Because the intracellular sequences of the GPCRs can differ, the pepducins can have high GPCR selectivity as well as target selective functions of a receptor.

Thus, allosteric ligands provide greater diversity in selecting desired functional and potentially clinical outcomes than orthostatic agonists and antagonists. Specially, with respect to orthostatic ligands, allosteric modulators provide greater specificity in targeting individual GPCRs, and can temper receptor activation in a manner not found with agonists. Furthermore, they can direct the activation of individual GPCRs to mediate some but not all of the functions of endogenous ligands and agonists at a given receptor by causing preferred GPCR-G protein interactions. Importantly, structural information now exists for some allosteric sites to allow for rational drug design and high throughput screening (HTS) assays have been developed to identify allosteric modulators. Consequently, the tools needed to begin to develop families of allosteric modulators for targeted GPCRs are at hand to generate new allosteric therapeutics.

Classically, GPCRs have been viewed as monomeric proteins. However, evidence has accumulated that GPCRs may physically interact with each other and that oligomeric forms of the same receptor (homodimers) or different receptors (heterodimers) may be functionally active [29-34,106,107]. The evidence for oligomerization has been based on biochemical co-immunoprecipitation studies using antibodies directed at one or the other dimer component [107]. This has involved studies using epitope tagged receptors transfected into cells and using antibodies against the tag on one receptor to co-immunoprecipitate the other receptor. Antibodies against the receptors themselves have also been employed to co-immunoprecipitate receptor heterodimers. Furthermore, biophysical approaches have also been used to measure dimer formation. These studies have employed fluorescence resonance energy transfer (FRET), time resolved FRET (TR-FRET), fluorescence recovery after photobleaching (FRAP), internal reflection fluorescence microscopy (TIRFM) and bioluminescence resonance energy transfer (BRET) based assays to measure either homodimers or heterodimers association [32,33,108-112]. Furthermore, complementation assays have been employed to measure oligomerization in which two receptors are tagged on their N- or C-termini with different constructs, each by themselves unable to give off a fluorescent response but when in close proximity to each other, as in a dimer, form a complement that gives off a luminescent response [113-115]. Finally, studies by Wu et al. [10] showed in crystal structure analysis that homodimers of the CXCR4 receptor form and these authors described the physical nature of this interaction and provided information not only on the sites of contact of the receptor monomers but an explanation of how the nature of the contact could explain the promiscuity of this receptor to form dimers with other GPCRs both within its receptor subfamily and with other receptor types.

Pharmacological approaches have also been employed to establish dimer formation. Studies by Zheng et al. [116] studied the interaction of mu opioid receptors with CCK2 receptors. These receptors are co-expressed in brain regions involved in pain modulation and a number of studies had suggested that the mu opioid and CCK systems may interact and that CCK antagonist may potentiate morphine induced analgesia and diminish mu receptor mediated tolerance development.

These authors investigated whether oligomer formation may be a basis of the interaction of the mu opioid and CCK system. They used a BRET assay to measure interaction of the cloned mu opioid and CCK2 receptors. Under basal conditions the receptors did not appear to associate and mu agonist and CCK ligands did not induce oligomerization as measured in the BRET assay. However, bivalent ligands containing a mu agonist and a CCK2 antagonist did cause a BRET response, suggesting that they bound simultaneously to each component of the dimer to promote oligomerization. Similar results were reported by Harikumar et al. [117] in which a fluorescent complement assay was employed to measure dimer formation and in these studies, the authors were able to fine tune the linker size between the opioid and CCK pharmacophore to measures distances of the ligand parts needed to promote dimer formation. Thus, these studies indicate that dimer formation can be induced by single ligands designed to bind simultaneously to two different receptors and suggest an approach to designing novel drugs based on their selective interaction with oligomers.

Some studies indicate that dimers may have pharmacological properties that differ from the monomers. Most studies on the pharmacological properties of dimers have shown that the pharmacological profile of ligands binding to monomers and oligomers are similar but that the affinities of the ligands may differ [34,118]. However, there are examples of compounds being identified that either bind to the oligomer differently or produce agonism or antagonism at the oligomer but not at the individual monomeric receptors [34,119,120]. These findings suggest that it is possible to identify small molecules that may selectively target dimers. However, as discussed by Birdsall [118], caution is needed in interpreting the pharmacological studies that suggest dimers exhibit differences in ligand binding properties from monomers. As suggested by Birdsall [118], in many cases, ligand binding characteristics “…do not agree with the thermodynamic constraints…” of the available models describing dimer interaction.

In fact, determining whether dimers, especially heterodimers, have pharmacologies that can be distinguished from monomers of each receptor could be more clearly established if compounds, such as the two examples indicated above [119,120] are identified that bind to or affect the activity of the heterodimer but not the homodimer or monomer. Identifying such compounds would support that notion that the ligand binding pocket of the heterodimer is different from the homodimer or monomer and the physical determinants of the heterodimer are unique.

This may be possible in the future since assays exist that can be employed to select for drugs targeting dimers. Most of the luminescent assays to measure dimer formation, in which acceptor and donors are attached to different components of the oligomer, can be employed to identify drugs that block dimer formation, in essence to identify allosteric inhibitors of the receptor complex formation. However, Milligan and associates [121-122] developed an assay to identify compounds that promote dimer formation and that can detect agonists at the dimer. For their approach, they developed GPCR-G protein fusion proteins to set up a defined receptor-G protein stoichometry. They introduced mutations in one monomeric GPCR-G protein that maintained ligand binding properties but not G protein activation. In the other GPCR-G protein pair, they introduced mutations that prevented GTP exchange following agonist activation. Each monomer alone was inactive. However, when the dimer formed, a fully functional receptor complex was formed.

This assay provides an approach to identify compounds selective for activating heterodimers. Compounds effective on the dimer but ineffective on the monomers might be considered dimer selective. Of course, the next problem would be to determine whether such compounds are effective on the native receptors expressed in native tissue.

In fact, while GPCR oligomerization has drawn much interest, a major question remains as to whether GPCR dimers form in native tissue. Studying oligomerization in native tissues has been difficult because of the low expression levels of receptors and dimers. However, recent studies by Albizu et al. [31] have attempted to directly test whether GPCR oligomers actually exist in real life (native tissue).

For their studies, they detected dimers using a FRET assay. The assay was dependent on the use of fluorescent ligands binding to the receptors. Specifically, agonists and antagonists were used with tags that contained either FRET donor or acceptors. The tags did not affect ligand binding appreciably. The binding of the ligands to monomeric receptors produced no FRET response. However, if dimers were present and the ligands with acceptor and donors were in close proximity a FRET response could be detected.

The assay system was developed and optimized using tumor cell lines expressing the cloned GPCRs. The authors conducted studies on the cloned vasopressin, oxytocin and dopamine D2 receptors. Interestingly, antagonists could produce strong fluorescent responses while agonists produced much weaker FRET responses. The diminished FRET responses were not due to the agonist causing dimer uncoupling. There may be a number of reasons or interpretations for the weak FRET response to agonist binding. However, the authors suggested that the weak FRET responses following agonist binding were due to negative cooperativity and that essentially one agonist molecule binds to a dimer at a time. In contrast, the strong FRET response found with antagonists is likely to reflect two antagonist molecules binding to a single dimer at a time.

Ultimately, the purpose of these studies was to test whether the FRET based assay could be employed to detect dimers in native tissue. For this, the investigators studied oxytocin receptors in mammary glands from lactating rats. They employed the same ligands as used for the studies on the cloned receptors and showed that antagonist binding yielded FRET responses and in contrast agonist binding yielded much weaker responses. These studies were interpreted as supporting the notion that oxytocin receptor dimers can be detected in native tissues.

However, it should be noted that the studies used tissues that express extremely high levels of oxytocin receptor (1–3 pmol/mg protein), similar to the levels that may be normally expressed in tumor cell lines transfected with recombinant receptors. The investigators could not detect the dimers in tissues with lower receptor expression. In fact, using the same approach, they were not able to detect oxytocin or vasopressin receptor dimers in brain, a tissue with relatively robust receptor expression although at much lower levels than in mammary gland of lactating animals. Thus, the question remains whether dimers are naturally found in tissues in which receptor expression is not abnormally high. This issue is critical in determining the value of dimer selective drugs discovered in immortalized cell lines in which receptor expression is unnaturally high.

Classical approaches to discover drugs targeting GPCRs have either employed radioligand binding assays, assays to measure GTPase activity or second messenger responses including changes in cAMP and Ca⁺⁺ signaling [1,2]. Within the last decade, newer technologies have been developed and adapted for HTS that can allow for identification of novel types of GPCR drugs. These technologies include luminescent assays as well as label free technologies, some of which can now be employed to study GPCRs in native tissues.

Basic research is driving the development of entirely new fields of GPCR drug discovery and development. It is providing the basis for discovery of compounds with new mechanisms of action, greater GPCR selectivity and unique biological actions. The hope is that these new generations of compounds can be translated into drugs with distinct and desirable therapeutic properties.

First and foremost is the development of allosteric ligands as designer drugs. Physical analysis of GPCRs by crystallography and NMR not only provides insight into the mechanics of drug-receptor interaction but yields critical information on how to rationally design allosteric modulators at specific receptors. We can begin to pick out specific sites on a receptor to affect receptor activation to produce designed responses, such as affecting one signaling pathway over another or to produce activation without desensitization or to facilitate activation only when the endogenous transmitter is present but not when it is absence to produce discrete signaling through the GPCR. The physical and computational analysis allows one to select compounds with particular chemistries to interact with selected sites on the receptor. The new screening technologies provide approaches to select for allosteric modulators using pharmaceutically user friendly HTS assays. Some of these technologies allow for phenotypic responses in native tissues to identify compounds more likely to produce desired therapeutic effects than assay systems dependent on immortalized cell lines over-expressing the receptors. Thus, one would predict a shift in direction of GPCR drug discovery to identification of new allosteric modulators discovered against endogenous GPCRs which may provide more interesting and potentially useful drugs than found with more classical orthosteric ligands.

Drug discovery against dimers also holds promise for the development of unique pharmacological agents and drugs. This may be particularly the case if the dimers have ligand binding properties distinct from the monomers. In this regard, heterodimer drug discovery may provide an approach to develop drugs that are region or tissue selective, which is presently not easily accomplished with classical drug discovery approaches. Furthermore, because dimers are predicted to be in equilibrium with monomers, drugs targeting dimers may shift the equilibrium thereby increasing the biological impact of dimer selective drugs. However, before the full potential of dimer pharmaceutics is realized, much more basic research will be necessary to establish the physiological role of these complexes in vivo. A major hurdle in this effort has been the problem of sensitivity, that is, the assays to detect dimers in physiological systems is still not adequate. This is due to the generally low level of endogenous expression of dimers in non-artificial systems. As the technologies improve, it is likely the evidence for unique dimers in brain and other organs will increase as will the interest and necessity to focus on these proteins for drug discovery.