Allosteric modulation of GPCRs is a topic of great interest, and has been the subject of a substantial body of literature; unfortunately, due to space limitations, we are unable to reference every paper and, instead, will focus only on seminal studies that are relevant to the concept of modulation of one receptor by another. Early observations made from agonist binding analysis of many GPCRs including the beta-2 adrenoceptor (β₂AR), muscarinic receptors, dopamine receptors and glucagon receptors all demonstrated the existence of both high and low affinity states of a GPCR. Thus, the simplistic binary model of inactive receptor R → active receptor R* would not work as a sufficient model for describing GPCR signaling. Keeping in mind this notion of a receptor existing in more than one conformation, the shift from high to low affinity states is likely related to increases in second messenger activity leading to feedback onto the receptor itself [3]. The Lefkowitz group first suggested that the signaling cascade of the β₂AR leading to increases in adenylyl cyclase (AC) activity is the result of a ternary complex that reflects the state of the receptor, R, the ligand or hormone, H, and an additional membrane component (in this instance, the G protein), g. Ligand H can bind saturably and reversibly to the receptor R, and the rate of the reaction is driven by the concentration of the ligand, the equilibrium dissociation constants, and the additional component g (Figure 1A) [3].

Just as the canonical example of hemoglobin illustrated that proteins often have more than one binding site, and that these sites can work cooperatively (that binding at one site has influence on the binding at other sites), most GPCRs also have more than one ligand binding site. Most GPCRs possess an orthosteric site where the endogenous ligand binds, and an allosteric site, where another ligand or molecule can bind to modulate receptor activity. Binding of an allosteric modulator can influence receptor signaling and behavior in either a positive or negative direction. All GPCRs have at least one allosteric site; binding of G protein incurs allostery in the most general sense of the term, and can shift the affinity for the orthosteric ligand. Recent studies have shown that many GPCRs bear one or more allosteric sites in addition to the G protein binding site. Allosteric modulators of GPCRs are defined as ligands that bind to an allosteric site on the GPCR to modulate the binding and/or signaling properties of the orthosteric site [4]. Allosteric modulators can increase or decrease the propensity of the receptor to bind the orthosteric ligand by stabilizing the active or the inactive state of the receptor, or by directly altering ligand selectivity. In addition to engineered ligands that serve as allosteric modulators of GPCRs, naturally occurring molecules at physiological concentrations such as ions (as in the case of zinc modulation of β₂AR) or small proteins (heparin binding to the neurokinin NK1 receptor) also serve as allosteric modulators of GPCRs (Figures 2A, 2B) [5,6]. GPCRs can also be modulated by other GPCRs (Figure 2C); we will discuss this in depth in a later section.

Examination of GPCR ligand binding and signaling within the framework of allostery requires further definition of different types of ligands that bind a GPCR. An agonist is a ligand that binds to the orthosteric site to shift the equilibrium to the receptor being in an active conformation (Figure 2A), inverse agonists shift the equilibrium towards the receptor's inactive state, decreasing the basal/constitutive receptor activity, and an antagonist prevents other ligands from binding without a shift in equilibrium. For example, β₂AR binding of isoproterenol (ISO, in Figure 2A, or H, in Figure 1) would lead to one stimulus (S). Figure 2A is a representative recording of currents elicited in a Xenopus oocyte expressing β₂AR and the Cystic Fibrosis Transmembrane Conductance Regulator (CFTR). When isoproterenol, a non-selective β₂AR agonist, signals through its cognate G protein leading to subsequent increases in cAMP, PKA phosphorylates CFTR, and opens the channel. Hence, CFTR can be used to monitor the activity of the β₂AR.

A molecule that binds at the allosteric site does not necessarily require any agonist or inverse agonist functionality of its own, only that it be able to shift the equilibrium of the receptor towards a more active or inactive state by stabilizing or destabilizing a certain receptor conformation. A positive allosteric modulator is a molecule that increases the receptor's affinity or efficacy at binding its ligand at the orthosteric site, or the receptor's ability to enter the active conformation, and a negative allosteric modulator would do the opposite. For example, cinacalcet, a positive allosteric modulator of the Class C Calcium Sensing Receptor, changes the receptor's sensitivity to circulating calcium, thereby being beneficial in disease states such as hyperparathyroidism in which there is a calcium sensing receptor deficiency [8]. Maraviroc, a negative allosteric modulator of the Class A CCR5 chemokine receptor, has been shown to block CCR5-mediated entry of HIV into cells [8]. Additionally, allosteric modulators can possess “pharmacological silence”, as in the case of M-5MPEP, an allosteric ligand for the MGluR5 (Metabotropic Glutamate receptor 5) receptor. M-5MPEP does not inactivate or activate MGluR5, but it partially blocks the response upon agonist binding, leading to partial mGluR5 inhibition [9].

Ions can also modulate GPCRs. For example, zinc has been shown to alter the ligand binding properties of a few different GPCRs at physiologically relevant concentrations (Figure 2B) [10]. Binding of either an agonist or an antagonist to 5-hydroxytryptamine/serotonin receptor 1A (5HT1A-R) is completely prevented by zinc [11]. In the continued presence of zinc, D1 and D2 dopamine receptors do not lose all ligand binding properties like the 5HT1A-R does, but they do have decreased affinity for ligand [12]. Referring back to our example of the β₂AR in Figures 1B and 2B, addition of zinc (X in Figure 1B) would shift the equilibrium constant such that binding of ISO would lead to a modified stimulus (S*β) compared to the stimulus S in the absence of the allosteric modulator.

We propose that a second GPCR can also act as an allosteric modulator of a GPCR. In the case of the β₂AR, our previous work has shown that both in vitro and in vivo, binding of bradykinin (BK) to B2R transactivates β₂AR [7]. Control experiments show that this transactivation is second messenger independent, and is a result of conformational changes in both receptors [7]. Figure 2C shows that upon coexpression of B2R and β₂AR, stimulation with BK leads to activation of both a calcium activated chloride channel (as indicated by the sharp current spike) and CFTR (as indicated by the slow current wave); the former is via B2R-mediated activation of Gαq signaling, and the latter is via heterodimer-mediated activation of Gαs signaling [7]. Stimulation with BK a second time also leads to β₂AR-mediated CFTR opening. These results show that B2R is able to transactivate β₂AR. Other results suggest that B2R is a negative allosteric modulator of β₂AR, as shown in the concentration response curve in the bottom panel of Figure 2C. In this case, the concentration response curve generated in Xenopus oocytes in response to increasing concentrations of terbutaline, a selective β₂AR agonist, exhibited a rightward shift in cells also expressing B2R, indicating that B2R dampens the ability of β₂AR to signal through the Gαs pathway. Because we hypothesize that one GPCR can modulate another, it is important to review the consequences of GPCR/GPCR interaction.

GPCR heterodimerization can lead to changes in a cognate receptor's ability to signal upon ligand binding. Given our previous definition of an allosteric modulator (an entity that binds a receptor at a site other than the orthosteric site thereby changing the properties of that receptor), then GPCRs can also serve as allosteric modulators of one another. In the specific examples detailed below, coexpression of a second receptor, R₂, changes the properties of the first receptor, R1, thereby altering the response to R1's cognate ligand. Thus, R₂ is an allosteric modulator of R1 (Figure 3).

The interaction of the family of Gαi-coupled OPRs is a classical example of GPCR heterodimerization leading to changes in pharmacology. It has been shown that κOPR heterodimerizes with δOPR but not μOPR [32]. Although the κOPR/ δOPR heterodimer does not exhibit altered selectivity for either receptor's specific agonist or antagonist, the heterodimer can bind partial agonists for either receptor with increased affinity (Figure 3A). Heterodimerization of these two OPRs also leads to enhanced agonism in the presence of both ligands; addition of the δOPR agonist DPDPE in the continued presence of the κOPR agonist U69563 increases binding affinity and receptor activation (Figure 3C). This synergy is also seen in the presence of both receptor's antagonists, suggesting the presentation of a novel ligand binding pocket in the heterodimer [39]. This κOPR/ δOPR heterodimer was later found to exist in vivo because the ligand 6′GNTI, an agonist that has tissue-specific analgesic effects, selectively targets the heterodimer; in contrast, this ligand is able to target the κOPR expressed singly at a very low efficacy and targets δOPR expressed singly not at all [27].

One GPCR in a heterodimer also can have an effect on the other receptor's cognate signaling pathways. The α_2A-AR and μOPR, both Gαi coupled receptors, when coexpressed in HEK cells were shown to heterodimerize by FRET measurements. Formation of the heterodimer enhances μOPR signaling when stimulated with morphine only, but simultaneous addition of an α_2A-AR agonist greatly decreased the μOPR signaling response, thus representing both a change in pharmacology and in signaling (Figure 3B). In the inverse experiments, there was also a dampening of norepinephrine-mediated G protein activation in the continuing presence of morphine, but not to the same extent [40].

Given the aforementioned example of α_2A-AR and μOPR heterodimers, it is now clear that heterodimerization also can lead to transinhibition of either R₁ or R₂ upon addition of antagonist against one of the two receptors (Figure 3E). The Gi-coupled chemokine receptors CCR2 and CXCR4 heterodimerize, and are well studied because of their role in Human Immunodeficiency Virus (HIV) entry into cells and in other inflammatory diseases (CCR2 is expressed on the surface of T lymphocytes and is involved in the recruitment of monocytes to atherosclerotic lesions). Antagonists against either CCR2 or CXCR4 transinhibit the other receptor, preventing binding of chemokines to the opposite receptor in both leukocytes and heterologous cell lines. The CCR2/CCR5 heterodimer also exhibits the same transinhibition [41].

Heterodimerization also can alter trafficking of either receptor compared to when they are expressed singly. For example, when coexpressed in HEK293 cells, V1a and V2 vasopressin receptors can heterodimerize. Although both receptors associate with β-arrestin, V1aR when singly expressed rapidly dissociates from β-arrestin in order to be rapidly recycled to the plasma membrane whereas V2R singly expressed remains stably associated with β-arrestin, thereby accumulating intracellularly. The V1aR/V2R heterodimer followed the V2R endocytic/recycling pattern when stimulated with a nonselective agonist. Selective agonism of the V1aR led to the co-endocytosis of the V2R, suggesting that the heterodimer is influenced by the activated receptor (Figure 3D) [28]. Similarly, cotransfection of β₂AR, a Gαs coupled receptor, and δOPR, a Gαi coupled receptor, in CHO and HEK cells led to heterodimerization. The heterodimer underwent both isoproterenol-mediated and opioid-mediated endocytosis (Figure 3D). Interestingly, the δOPR/ β₂AR heterodimer did not exhibit any changes in ligand selectivity compared to each cognate receptor [29]. As mentioned previously, α_1B-AR and α_1D-AR also heterodimerize; not only does this interaction change ligand binding characteristics as discussed above, but co-expression of α_1D-AR greatly increases the trafficking of α_1B-AR to the cell surface. Without α_1D-AR present, only negligible amounts of α_1B-AR are expressed at the cell surface [31].

Despite the increasing amount of knowledge about GPCRs, and the likelihood that they often exist as homo- or heterodimers, there is still a gap in the number of dimer pairs that have been shown to interact in both heterologous and endogenous systems (Table 1). Because transfection into a cultured cell line typically creates overexpression of a given protein, it is often suspected that the apparent dimerization and functional consequences are artifacts of non-physiological conditions. Furthermore, methods such as BRET and FRET, often used to study heterodimerization, assume that the introduction of the fluorescent tags attached to the proteins of interest does not induce any false positive interaction. A recent paper from the Sheikh group shows in three separate commonly used cell lines that previously published data on the Angiotensin type II receptor (ATR1)/B2R heterodimer are the result of artifact [42]. Thus, when examining potential interactions, both physical and functional, between two GPCRs, it is of utmost importance to be thorough in technique and cell type using both heterologous and endogenous expression systems to lend credence to statements regarding a putative heterodimer pair.

Two sets of studies have shown that interactions between two GPCRs also can lead to transactivation, where the ligand-induced conformational change in the R1 receptor leads to activation of the R2 receptor (Figure 3F). The first was the recognition that GABA_BR1 and GABA_BR2 must heterodimerize to form a functional receptor. In this case, GABA_BR1 can bind ligand but cannot activate signaling, while GABA_BR2 cannot bind ligand, but can activate Gαi-mediated signaling. Hence, binding of ligand to GABA_BR1 is thought to induce the adoption of an activated conformation by GABA_BR2. Our own work demonstrated that upon binding BK, B2R is able to transactivate β₂AR (Figure 2C, top). Additionally, coexpression of B2R is sufficient to hinder β₂AR's ability to respond to its selective ligand (Figure 2C, bottom). Thus, in this case the B2R is a negative allosteric modulator of β₂AR [7]. We know that B2R and β₂AR are physically associated in Xenopus oocytes, and in these cells, B2R coexpression alters β₂AR pharmacology as well as activating this receptor directly in the absence of a β₂AR-selective ligand (Figure 2C, top). Schwartz and Holst define an allosteric modulator as: “a ligand that functions as both an agonist on its own and as an allosteric modulator of the efficacy and/or potency of the orthosteric ligand” [43]. Although B2R is not a ligand, it fits the definition of an allosteric modulator: BK-bound B2R acts as an agonist at β₂AR, yet B2R negatively affects the sensitivity of β₂AR to activation by its cognate ligand.