Ligand-directed signalling at β-adrenoceptors

Abstract

β-Adrenoceptors (ARs) classically mediate responses to the endogenous ligands adrenaline and noradrenaline by coupling to Gsα and stimulating cAMP production; however, drugs designed as β-AR agonists or antagonists can activate alternative cell signalling pathways, with the potential to influence clinical efficacy. Furthermore, drugs acting at β-ARs have differential capacity for pathway activation, described as stimulus trafficking, biased agonism, functional selectivity or ligand-directed signalling. These terms refer to responses where drug A has higher efficacy than drug B for one signalling pathway, but a lower efficacy than drug B for a second pathway. The accepted explanation for such responses is that drugs A and B have the capacity to induce or stabilize distinct active conformations of the receptor that in turn display altered coupling efficiency to different effectors. This is consistent with biophysical studies showing that drugs can indeed promote distinct conformational states. Agonists acting at β-ARs display ligand-directed signalling, but many drugs acting as cAMP antagonists are also able to activate signalling pathways central to cell survival and proliferation or cell death. The observed complexity of drug activity at β-ARs, prototypical G protein-coupled receptors, necessitates rethinking of the approaches used for screening and characterization of novel therapeutic agents. Most studies of ligand-directed signalling employ recombinant cell systems with high receptor abundance. While such systems are valid for examining upstream signalling events, such as receptor conformational changes and G protein activation, they are less robust when comparing downstream signalling outputs as these are likely to be affected by complex pathway interactions.

This article is part of a themed section on Molecular Pharmacology of GPCR. To view the editorial for this themed section visit http://dx.doi.org/10.1111/j.1476-5381.2010.00695.x

Introduction

G protein-coupled receptors (GPCRs) represent the largest family of cell surface proteins that are targeted therapeutically. More than 30% of drugs on the market target GPCRs, and these have been derived from knowledge of about 200 receptors. We now know that there are more than 800 GPCRs (Kobilka, 2007), with further functional variation provided by interaction of receptors with accessory proteins, the formation of oligomers or the adoption of multiple receptor conformations. Classically, GPCRs change the activity of cells following binding of an agonist by interacting with heterotrimeric GTP-binding proteins (G proteins) leading to the release of Gα subunits that normally activate signalling enzymes to produce second messengers, and Gβγ subunits that activate additional pathways. Advances over the last decade in our understanding of how GPCRs work have challenged or added to many of the original concepts of signalling. For example, although activation of many GPCRs drives signalling in a preferred or canonical pathway, it is now increasingly recognized that receptor stimulation often results in a multitude of signalling outputs. This may involve coupling to multiple G proteins, Gα or Gβγ signalling, and pathway activation that is independent of G proteins. In addition, drugs can no longer be classified simply as agonists, partial agonists or antagonists, as it is now recognized that antagonists that block agonist-stimulated receptor activation may also act as inverse agonists to suppress basal receptor activation, or as protean agonists that block one effector pathway but stimulate one or more alternative pathways (Kenakin, 2001). The latter observation of differential pathway activation also applies to agonists. Indeed, it has been proposed that ‘ligands induce unique, ligand-specific receptor conformations that frequently can result in differential activation of signal transduction pathways associated with that particular receptor’ (Urban et al., 2007). Many of these concepts have been explored in detail utilizing adrenoceptors, a therapeutically important group of GPCRs that respond to the catecholamines, adrenaline and noradrenaline (NA).

Differential pathway activation by β-adrenoceptor (β-AR) ligands has been the subject of excellent recent reviews (Perez and Karnik, 2005; Galandrin et al., 2007; Kenakin, 2007; Audet and Bouvier, 2008; Hoffmann et al., 2008; Seifert and Dove, 2009). The capacity of receptor ligands to exhibit this behaviour has been described by numerous terms including stimulus trafficking, ligand bias or biased agonism and functional selectivity. We have chosen to use ‘ligand-directed signalling’ as a somewhat neutral term that does not have associations with other biological processes; however, we suggest that a consensus term that includes the key words ‘bias’ and ‘signalling’ may be adopted in the future. In this paper, we describe the evidence for ligand-directed signalling among the myriad pathways associated with β-AR activation.

G protein coupling and desensitization

Classically, β-ARs couple through Gαs to activate AC and increase intracellular levels of cAMP. Recent comparison of the structure of inactive rhodopsin with active opsin crystallized in the presence of a synthetic peptide derived from the C-terminus of Gαt has indicated that upon receptor activation, the C-terminus of the Gα subunit binds within a cavity created by outward tilting of transmembrane helix 6 (TM6), altered positioning of TM5 and restructuring of the link between TM7 and helix 8 (Scheerer et al., 2008). The Gα subunit undergoes a conformational change leading to the release of GDP, binding of GTP, an altered interaction with Gβγ subunits and binding to effector proteins. Given the conservation of key amino acids and overall structural similarity of family A GPCRs, the mode of Gs activation by the three β-ARs may be essentially equivalent to that seen for rhodopsin (Swaminath et al., 2004; 2005; Yao et al., 2006; 2009; Warne et al., 2008). Although GPCRs have been traditionally categorized by their coupling to one type of Gα, it is now well established that receptors are not simple binary switches, but instead show promiscuous coupling and activate multiple signalling pathways.

Factors that have emerged as important determinants of G protein coupling and the spectrum of signalling associated with each β-AR include interaction with different signalling proteins or multiprotein complexes utilizing motifs present in the intracellular loops and the C-terminal tail. Indeed, these interactions govern some of the functional differences between the three β-ARs. Numerous signalling proteins include domains or conserved modules that mediate protein–protein interactions. For example, proteins containing PDZ domains [post-synaptic density protein (PSD95)/Drosophila disc large tumor suppressor (DlgA)/zonula occludens-1 protein (zo-1)] bind to the C-terminal tails of target proteins including the β₂-AR (Hall et al., 1998a,b; Cao et al., 1999; Xiang and Kobilka, 2003) and β₁-AR (Xiang et al., 2002; He et al., 2006) to influence receptor trafficking and signalling. The Na⁺/H⁺ exchange regulatory factor binds to the C-terminal tail of the β₂-AR via a PDZ binding motif, D-S/T-X-L (Hall et al., 1998a,b;), where it mediates adrenergic regulation of Na⁺/H⁺ exchange (Hall et al., 1998b) and controls endocytic sorting of the β₂-AR in HEK293 cells (Cao et al., 1999). The β₂-AR PDZ motif also controls receptor recycling and coupling to Gi in cardiac myocytes derived from neonatal mice (Xiang and Kobilka, 2003).

In contrast, the β₁-AR PDZ domain E-S-K-V (He et al., 2006) interacts with PSD-95, preventing receptor internalization and interaction of the receptor with Gαi in cardiac myocytes (Hu et al., 2000; Xu et al., 2001; Xiang et al., 2002). A mutant β₁-AR, with the PDZ motif mutated (E-A-A-A), internalizes and couples to Gi like the β₂-AR. Binding of MAGI-2 to the PDZ domain of the β₁-AR promotes receptor internalization (Xu et al., 2001), and binding of MAGI-3 modulates β₁-AR-mediated Erk1/2 activation via Gαi/o without influencing β₁-AR-mediated cAMP generation (He et al., 2006). The β₁-AR can also associate with another PDZ protein, GIPC, which limits β₁-AR-mediated Gi/o-mediated Erk1/2 activation (Hu et al., 2003). Hence, interaction of β-ARs with different PDZ scaffolds plays a role in G protein coupling, receptor internalization and cell signalling pathways, although to date studies have been conducted with β-AR agonists such as isoprenaline (ISO), but few other β-AR ligands.

Many GPCRs display desensitization in response to continuous exposure to agonists (Krupnick and Benovic 1998; review by Hanyaloglu and von Zastrow, 2008, includes excellent schematic diagrams; Ferguson, 2001). The desensitization of responses involves three distinct stages: receptor phosphorylation; interaction with scaffolding proteins, such as arrestins; and internalization. It has been the accepted paradigm that β-ARs are phosphorylated at consensus sites within the third intracellular loop and C-terminal tail by PKA (PKA–cAMP-dependent protein kinase), PKC and G protein receptor kinases (GRKs), although the exact pattern of phosphorylation depends on multiple factors including cell type, levels of receptor expression and also receptor occupancy. Recently, an elegant study used the ICUE2 sensor to measure dynamic changes in cAMP levels in HEK293 cells expressing low levels of endogenous β2-ARs (Violin et al., 2008). In the absence of a phosphodiesterase (PDE) inhibitor, cAMP peaks within 90 s of agonist stimulation, then decays to basal levels within 5–6 min. Much of the decay is due to activation of PDE4 by PKA, and the remainder can be attributed to phosphorylation of the β2-AR by GRK6 and binding of arrestin-2/3.

Internalization of receptors occurs within minutes of agonist exposure, and involves the interaction between GRK-phosphorylated receptors and arrestin-2 and/or arrestin-3. The receptor/arrestin complexes accumulate in clathrin-coated pits that are pinched off by the protein dynamin to form endosomal vesicles. The internalized receptors do not activate G proteins and may be recycled to the cell membrane or undergo degradation (Hanyaloglu and von Zastrow, 2008). β₂-ARs are the most susceptible to this process, whereas β₁-ARs are more resistant. Activation of the human β₃-AR increases cAMP accumulation, but this subtype also couples to Gi to modulate AC activation. Coupling of the β₃-AR to Gi cannot involve receptor phosphorylation or internalization because the β₃-AR is not phosphorylated (Liggett et al., 1993) and does not bind arrestins or internalize (Cao et al., 2000; Breit et al., 2004).

Activation of non-canonical signalling pathways

A major breakthrough in the GPCR field was the recognition that the primary wave of signalling is accompanied by other receptor interactions, leading to desensitization of the initial response and activation of secondary signalling pathways. In recent years, the adrenoceptor field has developed many new layers of complexity, and the receptors and associated proteins have been shown to influence an almost bewildering array of signalling mechanisms (Lefkowitz et al., 2002). To date, much of the work on ligand-directed signalling at β-ARs has focused on differential activation of cAMP versus mitogen-activated protein kinase (MAPK) signalling pathways. In this section, we highlight a broader range of effector pathways that have been associated with β-AR activation (Figure 1), in the hope that future studies of ligand-directed signalling will encompass additional instances of differential pathway activation that may have important clinical benefits.

Figure 1.

Effector pathways stimulated by β-ARs. Each panel shows pathways that have been demonstrated experimentally across different cell types. Some effector mechanisms are observed in multiple systems, whereas others are more restricted or opposite between cell types. For example, β₃-AR-stimulated activation of p38 MAPK is mediated by a cAMP–PKA pathway in adipocytes, whereas in CHO-K1 cells expressing the β₃-AR, increasing levels of cAMP cause inhibition of p38 MAPK phosphorylation (Sato et al., 2007). Note that for the β₂-AR, arrestin-mediated Erk1/2 phosphorylation can be dependent on EGFR transactivation (e.g. Maudsley et al., 2000a), but in many studies the involvement of EGFR has not been investigated. G* denotes activated forms of Gs or Gi/o, and R*, R‡, R^# and R′ represent active receptor conformations. In light of recent evidence, G*αsG*βγ is shown as an intact heterotrimer, while G*αi/o and G*βγ are shown as dissociated subunits (Digby et al., 2006). References: ¹ Leblais et al., 2004; ² Morisco et al., 2005; ³ Galandrin et al., 2008; Kim et al., 2008; ⁴ Kim et al., 2008; ⁵ Schmitt and Stork, 2002b; ⁶ Zheng et al., 2000; ⁷ Gong et al., 2008; ⁸ Ma et al., 2000; ⁹ Ciccarelli et al., 2007; ¹⁰ Yamauchi et al., 1997; ¹¹ Yano et al., 2007; ¹² Shenoy et al., 2006; ¹³ Maudsley et al., 2000a; ¹⁴ Luttrell et al., 1999; ¹⁵ Sun et al., 2007; ¹⁶ Lindquist et al., 2000; ¹⁷ Cao et al., 2001; ¹⁸ Westphal et al., 2008; ¹⁹ Gerhardt et al., 1999; Soeder et al., 1999; ²¹ Sato et al., 2008; ²² Cao et al., 2000; Hutchinson et al., 2002.

c-Src-dependent and c-Src-independent activation of Erk1/2 by β-ARs

Much work has focused on phosphorylation and activation of the closely related MAPKs Erk1 (p44 MAPK) and Erk2 (p42 MAPK). These kinases are ubiquitously expressed at relatively high abundance, and play key roles in determining cell proliferation, differentiation or apoptosis. Many studies have addressed the mechanism whereby agonist-occupied β₂-ARs stimulate Erk1/2 phosphorylation (Ahn et al., 2003; Tohgo et al., 2003; Kobayashi et al., 2005; Lynch et al., 2005; Shenoy et al., 2006). The recruitment of arrestins that initiate β₂-AR internalization and recycling or degradation also drives the activation of MAPK pathways. In fact, arrestins function as scaffolds or adaptor proteins for the activation of many signalling networks, including phosphorylation of Erk1/2, c-Jun N-terminal kinase (JNK) or p38 MAPK, and other kinases including phosphoinositide 3-kinase (PI-3-kinase) and Akt (Luttrell et al., 1999).

β-AR-activated signalling networks also involve c-Src, a non-receptor tyrosine kinase that is important in determining cell fate. Although the exact mechanism varies, c-Src mediates Erk1/2 phosphorylation in various cell types in response to activation of the β₁-AR (Galandrin et al., 2008; Kim et al., 2008), β₂-AR (Daaka et al., 1997; Luttrell et al., 1999; Schmitt and Stork, 2000; 2002a; Friedman et al., 2002; Klinger et al., 2002; Sun et al., 2007) and β₃-AR (Cao et al., 2000; Hutchinson et al., 2002). In HEK293 cells expressing the β₂-AR, c-Src recruitment and subsequent Erk1/2 activation are secondary to binding of arrestin to the receptor (Luttrell et al., 1999). Alternatively, a study in HEK293 cells suggested that phosphorylation of the β₂-AR by PKA promotes switching of β₂-AR coupling from Gs to Gi, with c-Src activation and Erk1/2 phosphorylation dependent on Gβγ released from activated Gi proteins (Daaka et al., 1997). However, later work showed equivalent Erk1/2 responses in HEK293 cells expressing either the wild type or a mutant β₂-AR with all PKA phosphorylation sites removed (Friedman et al., 2002), and in the study of Violin et al. (2008) PKA was not involved in receptor phosphorylation. In mouse embryonic fibroblast cells, β₂-AR activation increases Erk1/2 phosphorylation in a biphasic manner. At low agonist concentrations, the response is due to Gαs activation, whereas at higher concentrations c-Src is involved independently of both G proteins and arrestins (Sun et al., 2007).

Although the β₃-AR does not bind arrestins, it also activates Erk1/2 (Gerhardt et al., 1999; Soeder et al., 1999; Hutchinson et al., 2002). This activation is cell dependent and may involve cAMP (Lindquist et al., 2000), Gi-derived βγ subunits (Gerhardt et al., 1999; Soeder et al., 1999) or c-Src activation (Cao et al., 2000; Hutchinson et al., 2002). Following agonist stimulation, it has been shown that c-Src co-precipitates with the β₃-AR utilizing a Gαi, but not arrestin-mediated pathway to activate Erk1/2 (Cao et al., 2000). In addition, the SH3 domain of c-Src is known to interact with other proteins via proline-rich motifs. Interestingly, mutation of the Pro-X-X-Pro motifs in the β₃-AR third intracellular loop or C-terminal tail completely blocks Erk1/2 phosphorylation in response to the agonist CL316243, without affecting cAMP accumulation (Cao et al., 2000), suggesting that direct interaction between the β₃-AR and c-Src is involved in Erk1/2 activation.

p38 MAPK

p38 MAPK is activated by a wide variety of stimuli including inflammatory cytokines, stress (osmotic and mechanical), growth factors, UV light and heat shock, and by a wide variety of GPCRs including β-ARs. p38 MAPK has been widely investigated in cardiac tissues as it is involved in cell death caused by ischaemia–reperfusion (Kaiser et al., 2004), acute myocardial infarction (Tenhunen et al., 2006) and cardiac failure (Braz et al., 2003; Nishida et al., 2004).

β₂-ARs have been shown to activate p38 MAPK by a cAMP–PKA-dependent mechanism in mouse cardiac myocytes (Zheng et al., 2000) and B lymphocytes (McAlees and Sanders, 2009). However, as with Erk1/2 signalling, many other mechanisms may be involved. For instance, β₂-ARs activate p38 MAPK utilizing Gβγ in HEK293 cells (Yamauchi et al., 1997), and β₁- and/or β₂-ARs activate p38 MAPK through a mechanism involving Gi in rat cardiac myocytes (Communal et al., 2000). A recent study (Gong et al., 2008) in HEK293 cells expressing the β₂-AR, showed biphasic activation of p38 MAPK to ISO: initial (within minutes) activation of p38 MAPK by an arrestin-2/Rac1/NADPH oxidase pathway and delayed activation (at least 90 min) due to a cAMP–PKA mediated mechanism.

β₃-ARs activate p38 MAPK in 3T3-L1 cells and brown adipocytes utilizing a Gs-cAMP mediated pathway (Moule and Denton, 1998; Cao et al., 2001; 2004; Mizuno et al., 2002). In contrast, the cAMP pathway can also inhibit p38 MAPK signalling in some cell types. For example, in CHO-K1 cells expressing the mouse β₃-AR at low levels (B_max 115 fmol·mg⁻¹ protein), the β₃-AR agonist CL316243 activates p38 MAPK, whereas in high-expressing cells (B_max 1150 fmol·mg⁻¹ protein) it has no effect. This is most likely due to the high levels of cAMP generated by the agonist in high-expressing cells, as stable cell-permeable analogues of cAMP also inhibit phosphorylation of p38 MAPK (Sato et al., 2007; 2008;). In some studies, β-ARs have no capacity to stimulate phosphorylation, instead only inactivating or dephosphorylating p38 MAPK. In accord with our findings, NA acting at β₁- and/or β₂-ARs in rat spinal microglia decreases ATP-mediated p38 MAPK activation by a cAMP–PKA mechanism (Morioka et al., 2009), and in chick cardiomyocytes ISO causes dephosphorylation of p38 MAPK (Tsang and Rabkin, 2009).

Transactivation of tyrosine kinase receptors

Other recurring players in the activation of MAPK pathways by GPCRs are the receptor tyrosine kinases (RTKs) including the EGF, PDGF and IGF receptors. For example, GPCR activation increases the activity of membrane-spanning matrix metalloproteases (MMPs) that cleave heparin-binding EGF, a single transmembrane-spanning protein, leading to shedding of a soluble EGFR ligand that is capable of activating the EGFR (Prenzel et al., 1999). GPCRs can stimulate MMPs by activation of Gα, release of Gβγ subunits following activation of Gαi or Gαq, and activation of Src or PKC (Pierce et al., 2001; Wetzel et al., 2001; Gschwind et al., 2002). GPCR activation induces a rapid increase in tyrosine phosphorylation of RTKs (Linseman et al., 1995; Rao et al., 1995; Daub et al., 1996). In rat-1 fibroblasts and COS-7 cells, activation of lysophosphatidic acid (LPA), endothelin-1, thrombin receptors or the β₂-AR results in tyrosine phosphorylation of the EGFR, and both GPCR-induced tyrosine phosphorylation and Erk1/2 activation can be blocked by dominant-negative mutants of the EGFR or inhibition by AG1478 (Daub et al., 1996; 1997; Maudsley et al., 2000a). In cells that lack endogenous EGFRs, such as L cells or CHO-K1 cells, LPA or β₂-AR stimulation can transactivate the PDGFR to increase Erk1/2 phosphorylation (Herrlich et al., 1998; Maudsley et al., 2000b).

PI-3-kinase

There is extensive evidence showing that β-ARs activate pathways involving PI-3-kinase and Akt. These pathways are involved in insulin-independent stimulation of glucose uptake mediated by GPCRs including β-ARs (Nevzorova et al., 2002; 2006; Hutchinson et al., 2007; 2008;). PI-3-kinases are classified into three classes according to structure and substrate specificity (Vanhaesebroeck et al., 2001). Class I_A PI-3-kinases are activated by RTKs. There are three isoforms of the 110 kDa catalytic subunit (p110α, β and γ, encoded by separate genes), seven isoforms of 85 kDa regulatory subunits that are produced by alternative splicing of three different genes (p85α and β, and p55γ), as well as a separately encoded p101 regulatory subunit. The only known member of class I_B PI-3-kinase is PI-3-kinaseγ, which consists of the p110γ catalytic subunit and the p101 subunit. The PI-3-kinaseγ isoform can be activated by Gβγ release following GPCR stimulation (Kurosu et al., 1997; Maier et al., 1999; Brock et al., 2003; Czupalla et al., 2003). Additionally, GRK2 can directly interact with PI-3-kinase, thereby directing recruitment to the β₂-AR (Naga Prasad et al., 2002). The function of class I PI-3-kinases is to convert PI to PI 3-phosphate [PI(3)P], PI 4,5-bisphosphate [PI(4,5)P₂] to PI 3,4,5-trisphosphate [PI(3,4,5)P₃], and PI 4-phosphate [PI(4)P] to PI 3,4-bisphosphate [PI(3,4)P₂]. PI(3,4)P₂ and PI(3,4,5)P₃ are able to bind to pleckstrin homology domains in other signalling molecules such as PDK1 and Akt (PKB) (Vanhaesebroeck et al., 2001).

The evidence for β-AR signalling to PI-3-kinase/Akt comes mainly from studies performed with the commercially available PI-3-kinase inhibitors, wortmannin and LY294002. β₂-ARs stimulate PI-3-kinase activity in H9c2 cells through a non-Gs-mediated pathway involving Gαi, Src and transactivation of the PDGF receptor (Yano et al., 2007). In rat cardiomyocytes, β₂-AR stimulation increases PI-3-kinase activity, possibly by a Gαi–Gβγ pathway, to restrict β₂-AR-mediated cAMP/PKA signalling (Jo et al., 2002). In the same cells, β₁-AR stimulation also increases PI-3-kinase activity by a PKA-mediated mechanism that limits β₁-AR cAMP production and subsequent positive inotropic responses (Leblais et al., 2004). In rat cultured aortic endothelial cells, activation of β₂-ARs phosphorylates Akt by a pathway involving Gαi–Src that is dependent upon Gαi rather than the dissociated Gβγ subunits (Ciccarelli et al., 2007). In H9c2 cells, β₂-AR stimulation causes Akt phosphorylation at both Ser473 and Thr308 (Yano et al., 2007). While all these studies involve classical agonist stimulation, there are others that report that the β₁-AR antagonist metoprolol causes Akt phosphorylation in the heart in both normoxic and ischaemia–reperfusion models (Sharma et al., 2008; Kovacs et al., 2009).

There are also reports of cross-talk between β-ARs and insulin receptor signalling. In cardiomyocytes, short-term β₁-AR stimulation phosphorylates Akt on Thr308 and Ser473 involving a PKA–CaMK calcium-dependent, but PI-3-kinase-independent pathway to increase cardiac glucose uptake (Morisco et al., 2005). However, longer-term β₁-AR stimulation causes impaired insulin-stimulated glucose uptake through a PKA–CaMK–PI-3-kinase-dependent pathway and reduced tyrosine phosphorylation of the insulin receptor. Similar results were obtained in brown adipocytes where β₃-AR stimulation caused decreased insulin-stimulated Akt and PI-3-kinase activities involving PKA–PKC pathway (Klein et al., 1999). In contrast, β₃-AR stimulation activates Akt/PI-3-kinase in rat endothelial cells (Figueroa et al., 2009) and adipocytes (Moule and Denton, 1997; Zmuda-Trzebiatowska et al., 2007). It is interesting to note that the β₃-AR-mediated increases in glucose uptake are sensitive to PI-3-kinase inhibitors, but there is no corresponding phosphorylation of Akt (Chernogubova et al., 2004).

Janus kinase/signal transducers and activators of transcription (JAK/STAT)

The JAK/STAT pathways are mainly associated with cytokine and growth factor receptor signalling. Binding of ligands to cytokine/growth factor receptors leads to tyrosine phosphorylation of JAK proteins, which then recruit and activate STATs. The phosphorylated STAT proteins dimerize and translocate to the nucleus where they bind to specific promoter elements of genes that regulate cellular proliferation, differentiation and apoptosis. Some GPCRs such as the angiotensin AT₁ (Marrero et al., 1995; Ali et al., 1997; Pelletier et al., 2003), thrombin (Rodriguez-Linares and Watson, 1994; Pelletier et al., 2003) and bradykinin (Ju et al., 2000) receptors are known to signal via JAK/STAT, although the mechanisms remain poorly understood. In the case of the AT₁ receptor, Jak2 appears to bind to a YIPP motif in the C-terminal tail of the receptor (Ali et al., 1997). A recent study (Pelletier et al., 2003) showed that angiotensin and thrombin effects on JAK/STAT are dependent upon Rho GTPases and NADPH oxidase activation. β₃-AR stimulation of adipocytes and β-AR stimulation of murine heart both activate JAK/STAT signalling through a cAMP–PKA-mediated pathway (Yin et al., 2003; Westphal et al., 2008), which may be involved in β₃-AR regulation of angiotensin II signalling in adipocytes and β-AR-mediated effects on cardiac remodelling respectively.

NOS

Adrenaline acts at β₁-/β₂-ARs and at β₃-ARs to increase nitric oxide and cGMP levels in rat mesenteric arteries (Figueroa et al., 2009). The involvement of β₃-ARs was demonstrated by blockade of responses in the presence of the β₃-AR-selective antagonist SR59230A. Other studies used the β₃-AR agonist BRL37344 to demonstrate increased soluble guanylate cyclase and nitric oxide levels in human endomyocardial and left ventricular tissues; however, this drug may also have agonist actions at β₁-/β₂-ARs (Gauthier et al., 1998; Brixius et al., 2004; Figueroa et al., 2009). Another study done using rat heart provides more convincing evidence for β₃-AR mediation of both negative inotropic and lusitropic responses by a pertussis toxin (PTX)-sensitive, nitric oxide–cGMP–PKG pathway (Angelone et al., 2008). Here, significant responses to 10 nM BRL37344 were completely blocked by 100 nM SR59230A or 100 nM L748337 (both selective β₃-AR antagonists), but not by 100 nM nadolol.

It has also been suggested that β₃-ARs produce a negative inotropic effect in human cardiac ventricle via the NO pathway. For example, NA produces a robust increase in peak tension in the presence of 1 µM prazosin (to block α₁-ARs), but this is converted to a weak negative inotropic response when 10 µM nadolol (a β₁-/β₂-AR antagonist) is also added (Gauthier et al., 1998). Given that β₁-/β₂-ARs increase peak tension, the observation that BRL37344 has negative inotropic effects with an EC₅₀ in the 10 nM range would suggest the involvement of β₃-ARs rather than β₁- or β₂-ARs. In contrast, human atrium displays positive inotropic responses to SR58611A, CGP12177A and BRL37344 that are blocked by bupranolol, but not nadolol (Skeberdis et al., 2008). These responses are mediated by a cAMP–PKA–L-type Ca²⁺ channel pathway, and importantly, stimulation of L-type Ca²⁺ channels by BRL37344 or CGP12177A is blocked completely by 1 µM L748337, indicating that this response is mediated by β₃-ARs. The authors suggest that the opposing actions of β₃-ARs in atrial and ventricular myocytes may reflect differing abundance of the receptor, or distinct compartmentation of signalling proteins between the two cell types. It should be noted, however, that studies on the role of β₃-ARs in human heart tissues must be viewed with caution. For example, Kaumann and Molenaar (2008) found no negative or positive inotropic effects of BRL37344 or other β₃-AR agonists in human right ventricular trabeculae from heart failure patients undergoing transplantation.

The β₂-AR inverse agonist ICI118551 reduces NA-induced vasoconstriction by an eNOS–nitric oxide mechanism in murine pulmonary arteries, although the only ICI118551 concentration reported was 10 µM (Wenzel et al., 2009). This response was diminished in the presence of butoxamine (a β₂-AR antagonist) and lost in β₁-/β₂-/β₃-AR triple knockout mice, but was not affected by CGP20712A (β₁-AR antagonist) or SR59230A (β₃-AR antagonist), indicating that ICI118551 produces vasodilation by acting at the β₂-AR. This was not a class action of β-AR antagonists because propranolol and butoxamine failed to affect vasodilation in the absence of ICI118551. Nebivolol is a selective β₁-AR antagonist for cAMP responses, but has been shown to induce vasodilation by an eNOS–nitric oxide mechanism in rodent and human coronary arteries, again only at high concentrations (Dessy et al., 2005; Rozec et al., 2006; Gupta and Wright 2008). Nebivolol induces NO release in human umbilical vein endothelial cells (HUVECs) via β₁-/β₂ as well as β₃-ARs (Evangelista et al., 2007). ln mice, on the other hand, the relaxation of coronary arteries and also angiogenesis induced by nebivolol in wild-type mice is largely lost in β₃-AR knockout mice, indicating only minor involvement of β₁- or β₂-ARs (Dessy et al., 2005). Compared to its vasodilatory effects, nebivolol has more potent negative inotropic activity in human endomyocardial biopsies, with an EC₅₀ in the nM range (Rozec et al., 2009). The nebivolol-induced reduction in peak tension is largely blocked by inhibitors of NOS and by the β₃-AR antagonist L748337, but not by nadolol, providing further evidence that β₃-ARs can signal by nitric oxide–cGMP–PKG pathways. The therapeutic actions of nebivolol in chronic heart failure may involve both antagonism at β₁-ARs and agonism at β₃-ARs. There are little data on β-AR-mediated regulation of iNOS, but a recent study (Pekarova et al., 2009) showed that the agonists adrenaline and NA, as well as the β-AR antagonist carvedilol, inhibit LPA-induced nitric oxide production in macrophages by a mechanism involving iNOS.

Ligand-directed signalling at β-ARs

The multiplicity of signalling pathways activated by β-ARs raises the intriguing possibility that it may be possible to activate these selectively by drugs. This can and has been done simply by exploiting the often different coupling efficiency displayed by particular signalling pathways in different cells. Thus, in a well-coupled system, a drug classified as a partial agonist may have high efficacy, whereas the same drug in a less well-coupled system may act as an antagonist. In this situation, selectivity is based on ‘strength of coupling’ that can vary with cell type and could be termed cell-based functional selectivity (Urban et al., 2007). However, the more interesting possibility from a therapeutic standpoint would be drugs that selectively activate beneficial pathways, while in the same cell type either block or have reduced activity at clinically detrimental pathways.

This is illustrated in Figure 2A, where signalling pathways 1–3 show progressively lower receptor/effector coupling. Drug A is a full agonist in pathway 1, a strong partial agonist in pathway 2 and a weak partial agonist in pathway 3. Drugs B and C display the same pattern, but have lower efficacy, and act as partial agonists even in the strongly coupled pathway 1, and have weak or no agonist activity for pathways 2 and 3 (and are highly likely to be antagonists of these pathways). Thus, these three drugs have the same rank order of efficacy for separate effector pathways, and the differences in responses can be explained purely by the coupling efficiency of the receptor to each effector. In Figure 2B, this is also the case for pathways 1 and 3, but drug B now has a much higher efficacy than drug A in pathway 2. The observation that drug A has higher efficacy than drug B for pathway 1, but a lower efficacy than drug B for pathway 2 is generally referred to as a ‘reversal of efficacy’ (Urban et al., 2007). This term can be used only when comparing two different drugs across two pathways, and does not refer to the ability of a particular drug to act as an antagonist for one pathway, but an agonist for another pathway. The most likely interpretation of an observed reversal of efficacy is that the two drugs are inducing or stabilizing different receptor conformations that in turn couple preferentially to different signalling pathways (Kenakin, 2007). Thus, reversal of efficacy provides a relatively unambiguous demonstration of ligand-directed signalling, that is, it represents a sufficient condition to state that this is occurring. However, we discuss below the evidence that reversal of efficacy is not always a necessary condition to be able to identify ligand-directed signalling.

Figure 2.

Two approaches to achieving selectivity of action at G protein-coupled receptors. In (A), three drugs, A, B and C, interact with three signalling pathways, E1–E3, that display different efficiencies of coupling to their functional response. For effector pathway E1, drug A is a full agonist, but B and C also have high efficacy; in E2 where the coupling efficiency is lower, only A is a full agonist, B is a low-efficacy partial agonist and C has no agonist properties; in E3 which is poorly coupled, only A has any agonist properties. Although some selectivity has been achieved as B and C will be agonists in some tissues and likely antagonists in others, this does not amount to ligand-directed signalling. However, in (B), the two drugs, A and B, display reversal of efficacy. For effectors E1 and E3, drug A acts as an agonist, whereas B is a low-efficacy agonist or has no effect. However, for effector E2, drug B is a full agonist, whereas A has no effect. The reversal of efficacy seen for E1 versus E2, and E2 versus E3 strongly suggests ligand-directed signalling.

Agonists and partial agonists promote distinct β-AR conformations

As shown in Figure 2, the capacity of two or more drugs to display ligand-directed signalling can be convincingly demonstrated by the measurement of multiple downstream signalling outputs and demonstration of reversal of efficacy (Urban et al., 2007). However, it is now also possible to examine changes in receptor conformation directly. This approach has been used to show that the endogenous β-AR agonist NA, catechol itself and the partial agonist salbutamol produce distinct modes of β2-AR activation (Swaminath et al., 2005). Each element within the structure of catecholamines has been shown to interact with amino acid side chains in the ligand-binding pocket of the β2-AR (Liapakis et al., 2004; Swaminath et al., 2004). The hydroxyl groups on the catechol ring undergo hydrogen bonding with Ser203 (5.42), 204 (5.43) and 207 (5.46; Strader et al., 1989a; Liapakis et al., 2000), the chiral β-hydroxyl interacts with Asn293 (6.55; Wieland et al., 1996; Bhattacharya et al., 2008; Reynolds et al., 2009), the aromatic ring undergoes hydrophobic interaction with Phe290 (6.52; Strader et al., 1989b) and the bioamine–NH₃⁺ group interacts with Asp113 (3.32; Strader et al., 1988). In addition, the amine substituent group present in full agonists such as adrenaline and ISO may interact with as yet unidentified residues in TM6 and TM7 (Liapakis et al., 2004; Swaminath et al., 2004). While optimal agonist efficacy is associated with the presence of all the elements present in adrenaline and ISO, compounds missing only one element, including NA, N-methyldopamine and salbutamol, are still strong partial agonists, and compounds missing two or more elements can still be weak partial agonists (Liapakis et al., 2004). It is interesting to note that the catechol analog U-0521 (3′,4′-dihydroxy-2-methylpropio-phenone), which lacks an amine group, demonstrates positive chronotropic effects in rat atria, albeit with about 10⁵-fold lower potency than ISO (Kaumann et al., 1977). In fact, U-0521 also increases the spontaneous beating rate of isolated ventricular myocytes despite its inability to promote cAMP accumulation, or to act as an antagonist for ISO-stimulated responses. These early findings perhaps indicate that U-0521 promotes an active receptor conformation distinct from that of ISO.

Reconstituted phospholipid vesicles containing β2-ARs labelled at Cys265 (6.27) with the environmentally sensitive fluorophore tetramethylrhodamine maleimide were used to show that the receptor undergoes two kinetically distinct conformational changes upon agonist activation (Swaminath et al., 2004). It was then demonstrated that whereas NA induces a biphasic conformational change composed of both rapid and slow phases, catechol induces only the rapid phase, and salbutamol induces only the slow phase. Interestingly, catechol was still able to induce the rapid conformational change following pre-incubation of the receptor with the partial agonist salbutamol, with the antagonists timolol and alprenolol or the inverse agonist ICI118551, but not after pre-incubation with NA or ISO (Swaminath et al., 2005). All β-AR ligands with a charged amine group interact with Asp113 (3.32). Molecular modelling indicated that catechol, NA and ISO also share the TM5 binding sites in the lower part of the binding pocket, but the aromatic ring of salbutamol appears to be orientated towards the upper part of the binding pocket and extracellular loop (ECL) 2 (Swaminath et al., 2005). This orientation of salbutamol prevents interaction between the chiral β-hydroxyl and Asn293 (6.55), consistent with earlier findings that the affinity and efficacy of partial agonists that lack a catechol group are not affected by mutation of Asn293 (Wieland et al., 1996). By inference, β-AR antagonists interact with Asp113 (3.32), but other than this, each compound would be expected to produce a unique network of interactions with amino acid side chains in the binding pocket.

These β₂-AR conformational studies indicate that there are not only quantitative differences in the activity of agonists and partial agonists, but also qualitative differences in their capacity to induce particular receptor conformations. Further work using live cells has confirmed that β-AR agonists display ligand-directed signalling. An elegant study has recently shown that agonist stereoisomers promote differential Gs/Gi coupling of the β₂-AR in adult rat cardiomyocytes (Woo et al., 2009). Fenoterol is a β₂-AR-selective partial agonist with two chiral centres, and can be synthesized as S,R; R,R; S,S; or R,S isomers. The S,R and R,R isomers of both fenoterol and the related compound methoxyfenoterol all stimulated cardiomyocyte contractility that was blocked in the presence of the β₂-AR-selective inverse agonist ICI118551. Importantly, the R,R isomers had higher potency than the S,R isomers, and the S,R but not the R,R responses showed a clear leftward shift in the presence of PTX. The PTX sensitivity was even more apparent in Erk1/2 phosphorylation assays, with the (S,R)-fenoterol response largely blocked, but the (R,R)-fenoterol response unaffected. Differential Gα coupling was verified by subtype-specific immunoprecipitation of activated Gα subunits labelled with [γ-³²P]GTP-azidoanilide. For example, the S,R isomer of fenoterol stimulated substantially higher activation of Gαi2 than the R,R isomer, whereas (R,R)-fenoterol produced a threefold higher activation of Gαs than (S,R)-fenoterol. These results are compatible only with the conclusion that the two stereoisomers stabilize distinct conformations of the β₂-AR. This work provides a clear demonstration of ligand-directed signalling in a physiological system, and may have important clinical implications. The authors suggest that agonists that selectively stimulate β₂-AR Gs coupling, without stimulating β₂-AR Gi coupling or β₁-AR activation, may have considerable therapeutic benefit (Woo et al., 2009).

Further demonstrations of ligand-directed signalling by β-AR agonists have used recombinant systems. Recently, bioluminescence resonance energy transfer (BRET) or fluorescence resonance energy transfer (FRET) has been used to directly measure changes in the association or relative conformation of receptors and interacting proteins. The data obtained can then be compared with associated signalling outputs, namely changes in cAMP or Erk1/2 phosphorylation (e.g. Drake et al., 2008). One such study has demonstrated that, relative to ISO, drugs acting as partial agonists for cAMP production can nonetheless act as full agonists for arrestin-3 recruitment (Drake et al., 2008). In fact, cyclopentylbutanephrine (CPB) and ISO demonstrate the reversal of efficacy that verifies the presence of ligand-directed signalling, while α-ethylnoradrenaline and isoetharine are both partial agonists for cAMP, but full agonists for arrestin-3 recruitment. Unlike these three drugs, the remaining full or partial agonists (adrenaline, NA, methylnoradrenaline, protokylol, deoxyadrenaline, zinterol, metaproterenol, terbutaline, fenoterol, procaterol, formoterol, albuterol, salbutamol and salmeterol) all have equivalent efficacy in the two assays. It was pointed out that drugs, such as CPB, α-ethylnoradrenaline and isoetharine, which display ‘biased agonism’ towards arrestin recruitment, share an ethyl group on the αC atom (Drake et al., 2008). As this is in close proximity to the NH₃⁺ group that interacts with Asp113 (3.32), there may be a steric effect of this α-ethyl group that compromises receptor conformational changes linked to G protein activation, without affecting those that promote phosphorylation or arrestin binding. In fact, addition of the α-ethyl group in α-ethylnoradrenaline substantially improves the rate of arrestin-3 recruitment without changing cAMP production relative to NA. It should be noted that these studies were done using HEK293 cells expressing the β₂-AR at 1 pmol·mg⁻¹ protein, and that the agonists were used at a concentration 100 times their K_D, in some cases up to 300 µM (including NA and ethylnoradrenaline). While these conditions are non-physiological, the experiments have clearly been designed to ensure maximal receptor occupancy, such that the observed responses solely reflect efficacy, or maximum agonist effect values. The main drawback of using such high concentrations would be the possibility of off-target effects. This is unlikely for the arrestin recruitment studies, however, as the FRET response can only involve the expressed β₂-ARs that are tagged with cyan fluorescent protein.

Another study that compared cAMP accumulation with activation of cAMP response element (CRE)-mediated reporter gene transcription provides a third demonstration that agonists and partial agonists promote qualitatively different conformations of the β₂-AR (Baker et al., 2003b). While both of these responses are downstream of the receptor activation process, they differ substantially in their degree of signal amplification and assay timing (10 min for cAMP accumulation and 5 h for the reporter gene assays). Most notably, the longer time period required for CRE reporter gene transcription favours receptor desensitization more than the short-term cAMP assay. The partial agonists salbutamol and terbutaline display efficacies for CRE activation that are comparable with those of the full agonists ISO and adrenaline. However, salbutamol and terbutaline are, respectively, 16 and 19 times more potent in the reporter gene assay than for short-term cAMP accumulation, whereas ISO and adrenaline are four- and sixfold less potent in the reporter gene assay compared to cAMP. The antagonists ICI118551 and propranolol show a two- to sixfold lower affinity for the β₂-AR based on their capacity to block ISO or adrenaline-stimulated CRE activation compared to their antagonism of cAMP responses. Furthermore, based on the CRE responses, ICI118551 and propranolol show a 10-fold lower affinity at the β2-AR in the presence of ISO or adrenaline compared to salbutamol or terbutaline. This discrepancy in antagonist pK_B values was not seen in the cAMP assays. The reporter gene assay data are not consistent with the generally held view that antagonist affinity is constant for a particular receptor, irrespective of the agonist or the bioassay used. Indeed, they suggest that the properties of the receptor are altered depending on whether it is activated by a full or a partial agonist. This alteration can be attributed in part to previous findings that the β2-AR is phosphorylated to a greater extent in response to full compared to partial agonists (January et al., 1997). When the reporter gene assays were carried out in cells expressing a mutant β2-AR lacking all of the possible PKA and GRK phosphorylation sites, there were two important differences in the data obtained. Firstly, ISO was substantially more potent at the mutant β2-AR (pEC₅₀ 9.45 vs. 8.11 for the wild-type receptor), and secondly, there was only a twofold difference in the pK_B values obtained for ICI118551 between ISO and salbutamol, compared to the 10-fold difference seen with the wild-type β2-AR (Baker et al., 2003b). This study brings together various previous findings and new ideas, namely that the receptor conformations induced by the full agonist ISO and the partial agonist salbutamol differ in their capacity for phosphorylation and desensitization of responses, and that the conformation of the β₂-AR that is phosphorylated and possibly interacting with additional proteins has a lower affinity for antagonists than the non-phosphorylated state.

β-AR ligands that are antagonists for cAMP accumulation are able to activate MAPK phosphorylation

The concept of ligand-directed signalling is a topic of immense interest and has recently been extended to drugs that act as antagonists for the cAMP pathway. As stated by Urban et al. (2007), ‘At the extreme, functionally selective ligands may be both agonists and antagonists at different functions mediated by the same receptor’. Several studies describe the activation of Erk1/2 phosphorylation by drugs classified as β-AR antagonists in cells expressing β₁- or β₂-AR (Azzi et al., 2003; Baker et al., 2003a). It has now been demonstrated that a wide range of β-AR ligands have complex efficacy profiles for cAMP generation and Erk1/2 activation at both β₁- and β₂-ARs (Galandrin and Bouvier, 2006). In addition, recent studies on mouse (Sato et al., 2007) and human (Sato et al., 2008) β₃-ARs showed that the antagonists SR59230A and L748337 act as classical competitive antagonists for cAMP accumulation, but are powerful agonists for both Erk1/2 and p38 MAPK activation. These effects again suggest that many compounds previously thought to interact with receptors to block the actions of agonists (as antagonists or inverse agonists) may in fact have the ability to selectively activate discrete pathways by inducing or interacting with particular conformations of the receptor. The idea that compounds acting as antagonists can in fact induce an active receptor conformation is not novel, as it has been known for some time that non-conventional partial agonists at high concentrations have cardiostimulant effects and produce cAMP accumulation via the β₁-AR (reviewed by Kaumann & Molenaar, 2008). For example, the compound CGP12177A blocks agonist-stimulated cAMP accumulation at low concentrations, but also binds to a low-affinity ‘agonist site’ utilizing interactions with residues that are distinct from the high-affinity ‘catecholamine site’ of the β₁-AR (Joseph et al., 2004; Baker et al., 2008).

The fact that many compounds previously regarded as ‘blockers’ express their own spectrum of pharmacological properties has potentially far-reaching consequences for the use of these drugs therapeutically. To date, there is not extensive literature that directly relates clinical efficacy to the ability of β-AR antagonists to act as agonists for MAPK or other signalling pathways; however, this area is likely to expand greatly in the near future. It has been suggested that the therapeutic benefits of carvedilol in heart failure patients are related to its unique capacity to activate Erk1/2 signalling by a G protein-independent mechanism (Wisler et al., 2007), but it is difficult to draw conclusions on the basis of one compound given that other β-blockers have similar clinical efficacy. To highlight an approach based on a series of different compounds, a key example of functional ligand selectivity is the finding that antipsychotic drugs acting at the dopamine D2 receptor can have opposite effects on Gi/o-mediated decreases in cyclic AMP compared to receptor recruitment of arrestin-3 (Masri et al., 2008). All clinically effective antipsychotics block arrestin-3 recruitment, despite having effects on Gi/o coupling that vary widely, ranging from partial agonists to neutral antagonists and inverse agonists. Here again, further work is needed to determine the downstream signalling pathways that are inhibited by antipsychotics, although the authors suggest that Akt and GSK-3 are important effectors.

Recent studies provide insights into how antagonists activate MAPK signalling. An array of β-AR blocking agents have been tested for their capacity to stimulate cAMP production (using the ICUE2 sensor) or Erk1/2 phosphorylation in HEK293 cells stably expressing the β₂-AR (Wisler et al., 2007). β-AR blocking agents that are partial agonists for cAMP accumulation, namely acebutolol, alprenolol, atenolol, labetalol, oxprenolol, pindolol and practolol, also stimulated Erk1/2 phosphorylation. All of the other agents tested (betaxolol, bisoprolol, carvedilol, ICI 118551, metoprolol, nadolol, propranolol, sotalol and timolol) are inverse agonists for cAMP in cells pretreated with 250 µM IBMX. Of these, only carvedilol and propranolol stimulated Erk1/2 phosphorylation. To determine the mode by which these two drugs promote Erk1/2 signalling, a mutant β2-AR was used that is deficient in G protein activation (T68F,Y132G,Y219A) (Wisler et al., 2007). Relative to ISO, carvedilol still acted as a partial agonist for Erk1/2 phosphorylation, whereas propranolol produced no response. Previous experiments had shown that ISO produces its response partly by coupling to Gi, and partly by receptor phosphorylation and recruitment of arrestin (Shenoy et al., 2006). Not surprisingly, the carvedilol Erk1/2 response was sensitive to depletion of arrestin-3 by siRNA, but was not sensitive to PTX. Thus, carvedilol, but not propranolol, causes receptor phosphorylation, recruitment of arrestin3–GFP and receptor internalization without changes in cAMP levels (Wisler et al., 2007). This study indicates that carvedilol induces or stabilizes a β2-AR conformation that does not activate G proteins, but can facilitate activation of arrestin-dependent signalling.

A similar study carried out in HEK293 cells expressing the human β₁-AR showed that ISO stimulates Erk1/2 phosphorylation by both Gi-dependent and G protein-independent pathways (Galandrin et al., 2008). Relative to ISO, bucindolol was a partial agonist and propranolol an inverse agonist for cAMP, whereas both bucindolol and propranolol stimulated Erk1/2 phosphorylation (E_max ∼ 30% of ISO response). The ISO-stimulated Erk1/2 response was partially blocked by PTX (30% of control) and by βARK-CT (a C-terminal peptide derived from GRK2 that sequesters Gβγ subunits), but the bucindolol and propranolol responses were unchanged (Galandrin et al., 2008). In cells co-expressing Gαi1 tagged with Renilla luciferase (Gαi1-91hRluc), Gγ2 tagged with green fluorescent protein (GFP10-Gγ2) and untagged β₁-AR, only ISO caused a reduced BRET signal due to dissociation of Gα and Gγ subunits. BRET was also measured in cells co-expressing β₁-AR-hRluc and GFP10-Gγ2 in the presence of untagged Gαi1, or alternatively β₁-AR–GFP10 and Gαi1-91hRluc. In both cases, the conformational change induced by ISO (10 µM) caused an increase, whereas bucindolol and propranolol decreased the BRET signal. These experiments show that ISO and two prototypical antagonists bucindolol and propranolol promote distinct conformations of the β₁-AR.

Whereas it is clear that activation of Erk1/2 phosphorylation by carvedilol at the β₂-AR involves arrestin recruitment (Wisler et al., 2007), the β₁-AR responses to ISO, bucindolol or propranolol were not sensitive to co-expression of a dominant negative arrestin-2 or siRNAs that knock down arrestin-2/3 (Galandrin et al., 2008). Instead, the c-Src inhibitor PP2 caused almost complete blockade of Erk1/2 phosphorylation for all three ligands, suggesting that the upstream signalling pathways activated by these ligands converge at or above the level of c-Src tyrosine kinases. This raises an interesting point, as like the β₃-AR, there are Pro-X-X-Pro motifs in the third intracellular loop and the C-terminal tail of the β₁-AR that could conceivably mediate interaction with c-Src or other SH3 domain proteins. It has been shown in vitro that the β₁-AR third intracellular loop does not bind c-Src directly. The entire loop containing the proline-rich motif binds specifically to endophilins (SH3p4/p8/p13). but not to other SH3 proteins including the adapter protein Grb2, c-Src or the synaptic vesicle trafficking protein amphiphysin 2 (Tang et al., 1999). It is possible that the C-terminal tail of the β₁-AR does bind c-Src, or alternatively, that interaction between this receptor and c-Src is mediated by adapter proteins other than arrestins. We suggest that binding of bucindolol or propranolol to the β₁-AR promotes a conformation that is able to activate c-Src without G protein involvement, while ISO promotes a conformation that can activate both Gi and c-Src, or alternatively, conformations that can activate each of these pathways independently.

The β₂-AR lacks any Pro-X-X-Pro motifs, but there is evidence that c-Src can bind directly to helix 8 (Sun et al., 2007). Despite this, the β₂-AR is more commonly found to activate c-Src by direct Gαs interactions (Ma et al., 2000), direct Gαi interactions (Ciccarelli et al., 2007), PKA phosphorylation (Schmitt and Stork, 2002b) or arrestin recruitment (Luttrell et al., 1999). The β₁-AR findings described earlier indicate that Erk1/2 phosphorylation in response to ISO, bucindolol or propranolol does not involve arrestins (Galandrin et al., 2008). Another recent study suggests, however, that carvedilol and alprenolol act at the β₁-AR to promote arrestin-2/3 recruitment and consequent transactivation of the EGF receptor and Erk1/2 phosphorylation (Kim et al., 2008). In agreement with this, propranolol was unable to stimulate arrestin recruitment, and bucindolol was not tested (Galandrin et al., 2008). These studies provide strong evidence that different drugs may have distinct modes of action not only with respect to cAMP and Erk1/2 signalling, but also in terms of the upstream signalling effectors that they activate. There is one caveat, as the study describing the actions of carvedilol and alprenolol used the mouse β₁-AR, while the effects of propranolol and bucindolol were demonstrated in cells expressing human β₁-ARs (Galandrin et al., 2008; Kim et al., 2008). Although both the human and mouse β₁-AR have multiple Pro-X-X-Pro motifs in the third intracellular loop and the C-terminal tail, there may be other amino acid differences that differentially affect phosphorylation or arrestin recruitment.

At the β₃-AR, drugs that act as antagonists of cAMP responses strongly activate Erk1/2. However, the Erk1/2 responses do not involve phosphorylation of the receptor, interaction with arrestins or internalization because the β₃-AR does not undergo any of these processes. In CHO-K1 cells expressing the human β₃-AR at physiological levels, the β₃-AR ligand L748337 is a competitive antagonist for cAMP accumulation, but has high agonist potency and efficacy for Erk1/2 phosphorylation. Zinterol, on the other hand, which has agonist properties at the human β₃-AR (Hutchinson et al., 2006), had high efficacy for cAMP accumulation, but lower efficacy than L748337 for both Erk1/2 and p38MAPK phosphorylation (Sato et al., 2008). Reversal of efficacy was also demonstrated between the agonist CL316243 and the antagonist SR59230A acting at the mouse β₃-AR (Sato et al., 2007). When the functional readout is cAMP, CL316243 is a full agonist and SR59230A either a partial agonist or antagonist depending on the level of receptor expression. In the identical cells but using extracellular acidification rate (ECAR) as the functional measure, both CL316243 and SR59230A are full agonists at all levels of receptor expression. Further analysis using selective MAPK inhibitors and Western blotting confirmed that SR59230A has much higher efficacy than CL316243 for MAPK signalling. These examples of reversal of efficacy provide strong support for the concept of ligand-directed signalling.

In addition, there is evidence from the studies with β₃-AR that MAPK responses induced by agonist ligands differ from those induced by antagonist ligands in terms of the G proteins utilized. Thus, L748337 stimulation of the human β₃-AR causes an Erk1/2 response that is largely blocked by PTX, indicating that the antagonist recognizes or induces a conformation of the β₃-AR that efficiently couples to Gi/o but not to Gs. Indeed, in the presence of a classical agonist, L748337 blocks the capacity of the β₃-AR to adopt a conformation capable of Gs coupling (Sato et al., 2008; Skeberdis et al., 2008; Wuest et al., 2009). The Erk1/2 response to the agonist ligand L755507 is much less affected by PTX, suggesting coupling predominantly to Gs (Sato et al., 2008). This finding again highlights differences between the three human β-ARs, as antagonist-stimulated Erk1/2 phosphorylation at the β₁- and β₂-AR is not PTX sensitive (Wisler et al., 2007; Galandrin et al., 2008).

Studies of ligand-directed signalling in CHO-K1 cells expressing the mouse β₃-AR demonstrate additional complexity for the interpretation of efficacy data (Sato et al., 2007). In cells with high β₃-AR expression levels (950 fmol·mg⁻¹ membrane protein), SR59230A caused increases in ECAR measured in the cytosensor microphysiometer, and behaved as a weak partial agonist for cAMP accumulation. In 3T3-F442A cells that express endogenous β₃-ARs or in CHO-K1 cells expressing low levels of the β₃-AR (115 fmol·mg⁻¹ protein), SR59230A still produced full cytosensor responses, but no measurable cAMP accumulation. Changes in ECAR in the low-expressing cells were abolished by the p38MAPK inhibitor RWJ67657, adding to the evidence that SR59230A caused robust phosphorylation of p38MAPK. In fact, the efficacy of SR59230A for the p38 MAPK response in the low-expressing cells was greater than in the cells with high β₃-AR expression. A clue to the cause of this surprising finding came from the observation that the efficacy of CL316243 was only one-third of that displayed by SR59230A in low-expressing cells, and in high-expressing cells CL316243 failed to produce a measurable p38MAPK response (Sato et al., 2007). Because p38MAPK phosphorylation was inhibited by 8-Br-cAMP, we concluded that p38MAPK activation is attenuated by cAMP, a situation that occurs to a greater extent in high-expressing cells. While in many cases there is positive cross-talk between signalling pathways, the study demonstrates a negative interaction between p38MAPK and cAMP in CHO-K1 cells expressing mouse β₃-AR.

Studies done at the level of receptor conformational changes; interactions between receptors and G proteins or alternative effectors, such as arrestins; or G protein activation are likely to provide a clear indication of the capacity of ligands to stabilize or induce distinct receptor conformations. It is also worthwhile to study ligand-directed signalling based on downstream signalling events, as these are of potential importance in optimizing the clinical efficacy of new or existing drugs. Our studies on the β₃-AR indicate, however, that measuring downstream signalling events can cloud the interpretation of ligand-stimulated responses, as there may be composite effects of conformational bias at the level of the receptor, plus signalling pathway interactions that produce either synergy or essentially ‘functional antagonism’, as in the case of p38 MAPK and cAMP (Sato et al., 2007). Our findings sound a cautionary note regarding the use of recombinant cell systems with extremely high levels of receptor expression for studies of pleiotropic signalling by β-AR antagonists, as the likelihood of pathway interactions may be greatly increased.

The studies described above use cAMP and ECAR as functional readouts to identify ligand-directed signalling. ECAR is a useful screening technique that identifies changes in metabolic activity in cells, and therefore, makes no assumptions regarding the pathways being activated. We have also used reporter genes in studies of human β₃-ARs to map the efficacy of different agonists and antagonists for a range of signalling pathways. The data have been used to generate a ‘web of efficacy’ (Figure 3) that compares the efficacy of ligands in four reporter gene assays, and provides a profile for the series of β-AR ligands tested. Although at a comparatively early stage, it is interesting to note that carvedilol and nebivolol display similar profiles and quite different from that shown by the prototypical β-AR antagonist propranolol. In contrast, bupranolol behaved as a neutral antagonist in all of the reporter gene assays tested to date. The human β₃-AR-selective ligands L755507 (agonist) and L748337 (antagonist) affect a similar spectrum of reporter genes, but with different efficacy. Hopefully, this approach will provide another avenue for the identification of ligand-directed signalling.

Figure 3.

The web of efficacy. A series of seven β-AR ligands were compared using four reporter gene assays, activator protein-1 (AP-1; JNK); cAMP response element (CRE; PKA and JNK/p38MAPK); nuclear factor of κB (NF-κB); and serum response element (SRE; MAPK/JNK). Note the similar profile exhibited by carvedilol and nebivolol, and the different profile shown by the prototypical β-AR antagonist propranolol. Bupranolol was a neutral or inverse agonist in the reporter gene assays. The human selective β₃-AR ligands L755507 and L748337 had similar profiles apart from that in the AP-1 reporter gene assay.

Perspectives

Do we need to redefine ligand-directed signalling?

As discussed earlier, it has been proposed that the unequivocal demonstration of ligand-directed signalling requires a reversal of the efficacy of two drugs for two different signalling pathways (Kenakin, 2003; Urban et al., 2007). This is perfectly acceptable if only downstream signalling outputs are measured, as otherwise there is no certainty that differing efficacies reflect the capacity of the drugs to elicit distinct receptor conformations. For example, a partial agonist may fail to stimulate one pathway that is weakly coupled to the receptor, but have high efficacy in a pathway that is strongly coupled. It is salutory to revisit the studies on the capacity of β-AR antagonists to stimulate Erk1/2 phosphorylation by β₁- and β₂-ARs. If we consider only the relative efficacy of drugs for cAMP production and Erk1/2 phosphorylation, all of the data obtained are consistent with a single active receptor conformation (as described in Figure 2A). Although Gs coupling and activation of AC are considered the classical β-AR signalling pathway, the activation of Erk1/2 phosphorylation appears to be a more strongly coupled pathway. It has been noted that none of the drugs examined so far display a G protein bias relative to arrestin activity, but the suggestion has been made that there are no theoretical grounds that this could not occur (Drake et al., 2008). It may be the case, however, that there are much more rigorous structural constraints on the ability of a drug to promote a receptor conformation that can activate G proteins than to activate MAPK pathways.

The detailed experiments on activation of Erk1/2 signalling confirm ligand-directed signalling at β₁- and β₂-ARs. The study of Galandrin et al. (2008) provides an excellent example of this point. While ISO, bucindolol and propranolol all stimulate Erk1/2 phosphorylation, only the ISO response is inhibited by PTX. Only ISO induces a reduced BRET signal between Gαi and Gβγ, consistent with dissociation of these subunits upon activation (Digby et al, 2006). BRET was also used to demonstrate that ISO increases the interaction between the β₁-AR and either Gαi or Gβγ, whereas both bucindolol and propranolol reduce these interactions (Galandrin et al., 2008). Thus, these drugs fulfil the criteria of inducing distinct receptor conformations, even though at face value there is no reversal of efficacy when measuring solely downstream signalling.

Insights from the crystal structures of the β₁- and β₂-ARs

The elucidation of these structures has provided a unique opportunity to examine the determinants for β-AR function. As more insights become available, especially the structure of an active conformation of a β-AR, it may be possible to draw conclusions about the capacity of different classes of drugs to promote ligand-directed signalling. The carazolol β₂-AR structure (Cherezov et al., 2007) has already been used to model interactions of the receptor with an array of drugs (Audet & Bouvier, 2008) that had been shown previously to display complex efficacy profiles (Galandrin and Bouvier, 2006). These drugs fall into three categories, namely AC inverse agonist/MAPK agonist, AC neutral antagonist/MAPK agonist and AC inverse agonist/MAPK inverse agonist. The first two groups dock in a similar fashion, interacting for the most part with common residues in TM3, TM6, ECL2 and TM7. In contrast, the three drugs in the MAPK inverse agonist class (metoprolol, bisoprolol and atenolol) show no interaction with Asp113 (3.32) or Asn312 (7.39), instead having closer contact with ECL2 at Thr195, as well as the Phe193 and Tyr199 that are common to the other drug classes. It is suggested that because ECL2 forms a disulphide bond to the top of TM3, binding of drugs in this orientation may prevent the conformational changes that are propagated to the bottom of TM3 upon receptor activation (Audet & Bouvier, 2008). Although this docking has been done using the rigid carazolol-bound structure, it certainly provides a working hypothesis for testing the basis of ligand-directed signalling by mutating the β₂-AR at key residues.

The complexity of signal transduction networks

In some of the studies described earlier, drugs differ in their capacity to activate upstream effectors, yet produce the same overall signalling response. For example, agonists can promote Erk1/2 phosphorylation mediated by β₁- and β₂-ARs acting by both G protein- and arrestin- or c-Src-dependent pathways, whereas antagonists at cAMP signalling only activate the non-G protein-dependent effectors. This ‘convergent’ signalling has been observed as well for the angiotensin AT_1A receptor in both heterologous expression systems and native vascular smooth muscle cells (Ahn et al., 2004; Kim et al., 2009).

It is important to note that the great majority of studies describing ligand-directed signalling by β-ARs have focused on cAMP and Erk1/2 signalling. However, these studies clearly demonstrate that different ligands promote distinct receptor conformations, and thereby stimulate different upstream effectors. The response of cells to stimuli depends upon the activation of a network of signalling pathways involving protein phosphorylation or translocation, enzyme activation/inactivation and changes in gene expression due to modulation of transcriptional activators or repressors. Therefore, even though multiple upstream effectors may converge at the level of EGF receptors or Erk1/2 phosphorylation, it is highly likely that different sets of signalling pathways are activated by each effector. The work presented in this review may represent the tip of the iceberg, highlighting the value of screening drugs not only for their activity at one or two signalling outputs, but instead at multiple pathways that represent the integrated responses displayed by cells following stimulation by endogenous hormones and neurotransmitters, or by drugs aimed at disease therapy. Approaches involving high-content analysis and reporter gene technology combined with siRNA and the use of inhibitors will no doubt revolutionize our understanding of receptor signalling.

The therapeutic relevance of ligand-directed signalling at β-ARs

Most work to date on ligand-directed signalling has been done using heterologous systems, often with high levels of receptor expression. Although it is important to consider the point that ligand-directed signalling may not be a significant player in vivo where cells have lower receptor abundance and are not exposed to high agonist or antagonist concentrations, the work we have described here indicates that possible bias must always be taken into account, and may in some cases have therapeutic benefits. Firstly, the use of recombinant systems, even with high receptor abundance, represents an important first step in providing proof-of-principle that β-ARs display ligand-directed signalling. In addition, it is difficult to define a ‘physiological level’ of receptor expression, as B_max values derived from whole tissues are simply an average over the entire population of cells present. It may be that receptor abundance is unexpectedly high in particular target cells. Having said this, the use of recombinant systems with high receptor abundance does increase the likelihood that downstream pathway interactions will complicate the demonstration of ligand-directed signalling. In contrast, the work showing differential coupling of cardiomyocyte β₂-ARs to Gs and Gi in response to fenoterol stereoisomers (Woo et al., 2009) provides an elegant example of a study conducted in an authentic system of direct clinical relevance. Secondly, antagonists are developed in the pharmaceutical setting based on their binding affinity, but it is of great importance that they are screened for functional activity as well. Indeed, it will be interesting to chart the emergence of new-generation drugs that have high potency at alternative signalling pathways. Although the β₃-AR is yet to be validated as a therapeutic target, our demonstration that the cAMP antagonist L748337 is a highly potent activator of Erk1/2 phosphorylation in cells expressing the human β₃-AR provides a case in point.

Ligands at the other β-ARs are important clinically. β-AR blocking agents are routinely used in the treatment of heart failure, and so a broader understanding of whether their capacity to activate (or not activate) particular MAPK or other non-cAMP pathways correlates with their clinical efficacy will be of immense value. For example, the Erk1/2 signaling pathway is known to be cardioprotective, in part due to inhibition of the cardiomyocyte apoptosis that results from ischaemia/reperfusion injury or oxidative stress (Yue et al., 2000; Lips et al., 2004). Likewise, chronic treatment with a JNK inhibitor worsens heart failure due to increased myocyte apoptosis and interstitial fibrosis (Kyoi et al., 2006). Whereas both Erk1/2 and JNK pathways have beneficial cardiac effects, there is consensus that p38 MAPK signalling is detrimental in most models of heart failure, and that chronic systemic p38 MAPK inhibition by pharmacological agents reduces disease severity (See et al., 2004; Widder et al., 2004; Liu et al., 2005; Kyoi et al., 2006).

As the importance of ligand-directed signalling becomes more fully appreciated, we envisage that both existing and novel β-AR agonists and antagonists will be subject to screening for their interaction with multiple signalling pathways. In the case of existing drugs, it may be possible to determine activity profiles that correlate positively or negatively with clinical efficacy as has been done for the series of antipsychotic drugs acting at the dopamine D2 receptor (Masri et al., 2008). The ability to predict therapeutic benefit for newly developed drugs will depend largely on the power of this profiling, and it will be interesting to see whether profiling can be done in recombinant systems with high receptor abundance or whether it must be augmented by the use of primary human cell systems expressing endogenous receptors. Ultimately, the application of broader screening methods to drug development will need to be validated by post hoc clinical trials and long-term monitoring of clinical outcomes.

Glossary

Abbreviations:

8-Br-cAMP

8-bromoadenosine 3′,5′-cAMP

AR

adrenoceptor

AT_1AR

angiotensin II receptor type 1a

BRET

bioluminescence resonance energy transfer

CaMK

Ca²⁺/calmodulin-dependent protein kinase

catechol

1,2-benzenediol

CHO-K1

Chinese hamster ovary

CPB

N-cyclopentylbutanephrine

ECAR

extracellular acidification rate

ECL

extracellular loop

EGF

epidermal growth factor

ERK

extracellular-regulated kinase

FRET

fluorescence resonance energy transfer

GFP

green fluorescent protein

GPCR

G protein-coupled receptor

GRK

G protein-coupled receptor kinase

Gs

guanine nucleotide-binding protein that stimulates AC

Gi

guanine nucleotide-binding protein that inhibits AC

GSK-3

glycogen synthase kinase-3

GTPγS

guanosine 5′-O-(thiotriphosphate)

HEK

human embryonic kidney

IBMX

3-isobutyl-1-methylxanthine

IGF

insulin-like growth factor

ISO

isoprenaline

JAK

Janus kinase

JNK

c-Jun N-terminal kinase

LY294002

2-(4-morpholinyl)-8-phenyl-4H-1-benzopyran-4-one

MAPK

mitogen-activated protein kinase

MEF

mouse embryonic fibroblast

MMP

matrix metalloprotease

NA

noradrenaline

PDE

phosphodiesterase

PDGF

platelet-derived growth factor

PDZ

post-synaptic density protein (PSD95)/Drosophila disc large tumor suppressor (DlgA)/zonula occludens-1 protein (zo-1)

PI-3-kinase

phosphoinositide 3-kinase

PP2

4-amino-5-(4-chlorophenyl)-7-(t-butyl)pyrazolo[3,4-d]pyrimidine

PTX

pertussis toxin

Rluc

Renilla luciferase

RWJ67657

4-[4-(4-fluorophenyl)-1-(3-phenylpropyl)-5-(4-pyridinyl)-1H-imidazol-2-yl]-3-butyn-1-ol

siRNA

small interfering RNA

STAT

signal transducers and activators of transcription

TM

transmembrane segment

Conflicts of interest

N/A.