Although now appreciated to represent a substantial over-simplification (Milligan and IJzerman, 2000; Strange, 2002), the two-state model of GPCR function (Samama et al., 1993) remains an extremely useful concept. Agonist ligands stabilize or increase the fraction of the active state of a GPCR such that it can interact with and activate a G protein. As such, basic thermodynamics define that there must be a finite probability that this active state also occurs in the absence of the agonist. Equally, if agonists enrich such active states, then it should be possible to identify ligands (inverse agonists) that stabilize or enrich the inactive state. With hindsight, it is easy to argue that compounds (neutral antagonists) that bind to GPCRs without altering the equilibrium between active and inactive states of the receptor are likely to be rather rare. However, early cartoons that specifically illustrated this point (Milligan et al., 1995) were contentious because they did not reflect pharmacological experience in which `antagonists' were common reagents. However, within such models, the scale of ligand efficacy ranged from 1 (full agonist) to -1 (full inverse agonist) and neutral antagonists were defined very precisely as possessing 0 efficacy. Heterologous expression of many GPCRs resulted in the detection of ligand-independent signal transduction that increased in an essentially linear fashion with increasing levels of GPCR expression (Tiberi and Caron, 1994). After clear demonstration that the constitutive activity of the GPCR was not associated with the presence of low concentrations of endogenous agonists, the ability of many traditional `antagonists' to block the constitutive activity of expressed GPCRs rapidly saw these compounds reclassified as inverse agonists (Milligan et al., 1995). This process continues to the present with many studies still reporting the inverse agonist activity of ligands at a wide range of both native and mutated GPCRs (Daeffler and Landry, 2000). Compounds with close to zero efficacy are vital tools in ligand classification because any compound with this characteristic that binds in a competitive manner will act as a functional antagonist to compounds with agonist or inverse agonist properties. They can thus be used to exclude apparent inverse agonism that derives from competition with an endogenous ligand rather than via suppression of receptor constitutive activity.

GPCRs Display Varying Levels of Constitutive Activity

A key question for the potential importance of inverse agonists and whether they may have inherent benefits as therapeutic agents relates to the level of constitutive activity of individual GPCRs. Many clinically important medicines have been demonstrated to behave as inverse agonists when tested against either wild-type or mutated GPCRs (Table 1). In most cases, these studies have used GPCRs expressed recombinantly in cell lines, but in certain cases, significant overexpression of GPCRs in tissues of transgenic animals has been employed to produce measurable levels of constitutive activity. Although virtually all GPCRs can be modified to display constitutive activity (Chalmers and Behan, 2002), many GPCRs display rather low levels of constitutive activity when expressed in recombinant systems and display greatly increased activity in the presence of an agonist. A classic example is the hamster α_1b-adrenoceptor. In this GPCR, mutation of Ala²⁹³ to any other amino acid increased constitutive activity (Kjelsberg et al., 1992). This indicates a strong evolutionary pressure to maintain the wild-type form of the receptor in a nearly silent state. Indeed, for many GPCRs, mutational changes in the region near the the interface at the end of the third intracellular loop and the start of transmembrane helix VI result in elevated levels of constitutive activity (Pauwels and Wurch, 1998; Seifert and Wenzel-Seifert, 2002). This may relate to the well-appreciated movement of the bottom of transmembrane helix VI in response to the binding of agonist ligands (Hubbell et al., 2003). It is also noteworthy that mutations in this region of a number of GPCRs with endocrine ligands have been associated with relatively rare but extremely well studied, hormone-independent, human diseases (Parnot et al., 2002). These diseases would certainly be treated more effectively with inverse agonists compared with neutral antagonists. Enhanced levels of constitutive activity can also be produced by judicious mutation in many other locations in GPCRs (Pauwels and Wurch, 1998; Seifert and Wenzel-Seifert, 2002). Indeed, a recent random mutagenesis study employing the DOP opioid receptor used the development of constitutive activity to trace pathways of GPCR activation (Decaillot et al., 2003). These studies further indicate significant evolutionary pressure to maintain low levels of agonist-independent activity. In the case of rhodopsin, it is vital that levels of constitutive activity are virtually nil, otherwise the sensitivity and amplification of the visual pathway would result in permanent photobleaching. This is achieved by the covalent association of the ligand retinal in the binding pocket of the GPCR. In the 11-cis configuration, retinal holds rhodopsin in an inactive state, and this has resulted in 11-cis retinal being described as an endogenous inverse agonist (see below). Spontaneous isomerization of 11-cis retinal to the agonist all-trans retinal is an extremely rare event. However, the rapidity and efficiency of the photon-induced switch of 11-cis retinal to the all-trans configuration results in highly effective signal transduction.

Except for the very specialized example of rhodopsin noted above, measurable constitutive activity can be recorded, without mutation, for virtually all GPCRs. However, in assays for constitutive activity that monitor GPCR-mediated GTPase activity in membrane fractions, the contribution of the GPCR of interest to the total high-affinity GTPase can be relatively poor. This can be improved by addition of regulator of G protein signaling (RGS) proteins to the assay (Welsby et al., 2002). These proteins function to accelerate the intrinsic GTPase activity of many heterotrimeric G proteins (Neubig and Siderovski, 2002). It is of little significance to the RGS whether GTP was loaded onto the G protein by an agonist-mediated or agonist-independent event. Thus, the presence of sufficient RGS increases the fraction of basal GTPase activity contributed by the constitutive activity of a GPCR and provides a greater level of activity that can be inhibited by inverse agonists and hence measured (Welsby et al., 2002). In intact cells, RGS proteins and other modulators of G protein function will regulate the constitutive activity of GPCRs and thus the likely therapeutic effectiveness of inverse agonists.

Closely related GPCRs frequently display significantly different levels of constitutive activity when expressed at equal levels. For example, a number of studies have noted higher levels of constitutive activity of the β₂-adrenoceptor compared with the β₁-adrenoceptor (Engelhardt et al., 2001; Zhou et al., 2000) and of the dopamine D5 receptor compared with the dopamine D1 receptor (Tiberi and Caron, 1994). Possible reasons for such differences are discussed later. Tissue-targeted transgenic expression of wild-type GPCRs also results in constitutive signaling, which can result in physiological function largely independent of requirement for agonist. One of the earliest examples was the generation of transgenic mice overexpressing the β₂-adrenoceptor to a high degree via a heart-specific promoter. Isolated atria from the transgenic mice displayed isometric tension as high in the absence of agonist as that produced by a maximal concentration of isoprenaline in atria of control mice. Basal heart rate was also much higher in the transgenic animals and not further increased by infusion of isoprenaline (Milano et al., 1994). Interestingly, this phenotype was not observed by transgenic expression of a constitutively active mutant of the β₂-adrenoceptor (Samama et al., 1997). This reflected the now well appreciated instability of such mutants that frequently results in low steady-state levels of expression. Treatment of the animals with β-blockers that are inverse agonists up-regulated the mutant protein and uncovered the constitutively active phenotype (Samama et al., 1997). In contrast to examples noted earlier, a number of GPCRs do seem to have significant levels of constitutive activity when expressed in cell lines; in some cases, ligand-induced stimulation of activity is relatively small compared with the signal in the absence of ligand. Three families of GPCRs that cluster closely together in sequence similarity plots are the sphingolipid receptors, the melanocortin receptors, and the cannabinoid receptors. These tend to display high levels of agonist-independent activity, and it is noteworthy that unlike the vast majority of the rhodopsin-like class A GPCRs, they do not have the possibility to form a disulfide bond between cysteine residues positioned adjacently in extracellular loops II and III. However, it is not a simple issue to test whether lack of this disulfide bond enhances the constitutive activity of a range of GPCRs, because mutational alteration of these residues frequently prevents transport of the modified GPCR to the cell surface (Ai and Liao, 2002; Kuwasako et al., 2003).

Virally Encoded GPCRs

A number of viruses encode combinations of chemokine-like GPCRs and GPCR ligands in their genome (Rosenkilde et al., 2001). These include ORF74 of human herpes virus-8 (also called Kaposi sarcoma herpes virus) and US28 and US33, both encoded by human cytomegalovirus. Presumably, these were pirated from mammalian cells infected by the ancestors of these viruses. ORF74 displays homology to the CXCR2 receptor and was shown to possess constitutive activity when expressed in heterologous systems. Furthermore, ORF74 transfected NIH-3T3 cells can induced tumor formation in nude mice (Bais et al., 1998) and mice transgenic for ORF74 expression develop Kaposi sarcoma-like symptoms (Yang et al., 2000; Guo et al., 2003). The majority of chemokine receptors can be regulated by a range of chemokine ligands; indeed, in certain cases, chemokines seem to act as endogenous inverse agonists of these virally encoded GPCRs. Both human interferon-γ-inducible protein 10 and stromal cell-derived factor-1α have been shown to inhibit the constitutive signaling of ORF74 (Rosenkilde et al., 1999). The US28 gene product is distantly related to the human CCR5 and CXCR4 chemokine receptors. Like these, US28 allows infection of CD4-positive human cell lines by primary isolates of HIV-1 and HIV-2 (Pleskoff et al., 1997). Both the US28 and US33 GPCRs also display constitutive activity when expressed in heterologous systems, and a nonpeptidergic ligand, VUF2274, has been shown to act as a relatively low-affinity inverse agonist of US28 without affecting the constitutive activity of either US33 or ORF74 (Casarosa et al., 2003). This is of considerable interest because VUF2274 was also shown to be able to partially inhibit HIV-1 entry into US28-expressing, CD4-positive cells (Casarosa et al., 2003). This viral GPCR demonstrates high levels of constitutive endocytosis (Waldhoer et al., 2003) but studies with variants of US28 lacking elements of the C-terminal tail have dissociated constitutive internalization and constitutive signaling (Waldhoer et al., 2003), implying that high levels of constitutive activity does not directly induce endocytosis. This might have been expected to be the case if the constitutive activity were associated with ligand-independent-phosphorylation and translocation of β-arrestins. However, there are potentially conflicting recent data on the likely importance of interactions with β-arrestins in the constitutive internalization of this GPCR (Fraile-Ramos et al., 2003; Miller et al., 2003).

Constitutive Activity of GPCRs in Native Systems

As noted above, in many situations, the inability to clearly exclude a contribution from endogenous ligands to basal activity has made unambiguous scoring of GPCR constitutive activity in native systems difficult. As such, a thorny issue is whether significant levels of constitutive activity are seen only in recombinant systems in which GPCR expression levels can be manipulated and are often relatively high. Early data to suggest that this is not the explanation in all cases has been reviewed elegantly and extensively by de Ligt et al. (2000).

The availability of ligands with close to zero efficacy is required to allow detailed study in native tissues. After expression of splice variants of the rat histamine H3 receptor in Chinese hamster ovary cells, both histamine and imetit were able to elevate [³H]arachidonic acid release, whereas thioperamide and a range of other histamine H3 blockers were able to inhibit basal release. By contrast, proxyfan was without effect. Importantly, the effects of both agonists and inverse agonists were blocked by proxyfan in a concentration-dependent fashion (Morisset et al., 2000). Based on this definition of proxyfan as a neutral antagonist for the rat histamine H3 receptor, this compound was used to examine the relevance of constitutive activity of this receptor on either autoreceptor-mediated [³H]histamine release or basal and ligand-regulated guanosine 5′-O-(3-[³⁵S]thio)triphosphate binding in rat brain membranes. In both assays, proxyfan reversed both the agonist effects of imetit and the inverse agonist effects of thioperamide without producing significant effects itself (Morisset et al., 2000). This remains one of the most impressive examples of constitutive activity in a native setting and is consistent with significant tonic regulation via ligand-independent activity of a GPCR. In a clinical content, these results have led to the suggestion that inverse agonists might be preferred to neutral antagonists as `cognitive enhancers' (Schwartz et al., 2003).

Most studies on constitutive activity and inverse agonism continue to rely on heterologous expression of GPCRs. A further complication for the use of transfected cell systems as a means to gauge the potential for constitutive activity of GPCRs in native tissues is that it is now clear that GPCRs generally do not exist in isolation but may have (many) potential binding partners. Interactions between intracellular regions of GPCRs, particularly the C-terminal tail (Bockaert et al., 2003), and other polypeptides has become an intensely studied topic (Kreienkamp, 2002; Premont and Hall, 2002). In many cases, such interactions serve to regulate the cellular location, scaffolding and/or trafficking of a GPCR. Such interactions can, however, also regulate the degree of constitutive activity observed. If the heterologous cell system selected for expression of a GPCR lacks expression of a specific interacting protein, this can modulate the observed degree of constitutive activity. For example, when expressing either the mGluR 1a or mGluR 5 receptor in HEK293 cells, Fagni and colleagues observed constitutive inositol phosphate generation (Ango et al., 2001). This was not observed, however, in neurons. Although differences in expression levels in the two host systems might have provided a trivial explanation for these findings, a more interesting one emerged. The C-terminal tail of the mGluR1a and mGluR 5 but not other members of this GPCR subfamily family contain consensus regions for the binding of homer proteins. Homer proteins have been particularly well studied owing to their capacity to provide a bridge between cell surface receptors and the cortical actin system of the cytoskeleton and thus anchor receptors in specific locations (e.g., the synapse) (Thomas, 2002). Cultured cerebellar granule cells were shown to express homer 3 but not other members of the homer family (Ango et al., 2001). After treatment of such cells with antisense oligonucleotides targeting homer 3 expression, basal inositol phosphate production increased. This was unaffected by addition of the mGluR1 neutral antagonist CPCCOEt but reversed by addition of the inverse agonist BAY36. Such constitutive activity of the mGluR1a receptor was not uncovered when cells were treated with a scrambled oligonucleotide sequence based on the effective antisense (Agno et al., 2001).

Conclusions

Partially because of the relative simplicity of the assays, `antagonist' ligands are now routinely assessed for inverse agonism. Many clinically effective medicines can be shown to be inverse agonists. However, the bulk of such studies still rely on heterologous expression of GPCRs in cell lines, and it remains unclear whether inverse agonists offer inherent advantages over neutral antagonists in the vast majority of clinical scenarios, or indeed whether the inverse agonism of some medicines contributes significantly to their effectiveness. Unraveling this issue remains a key challenge. The complexity of signal transduction scaffolds and the varying distribution of GPCR interacting proteins implies that simple heterologous expression systems may not provide a useful assessment of the significance of inverse agonism in a clinical setting.