Protein-Protein Interactions at the Adrenergic Receptors

The adrenergic receptors are among the best characterized G protein-coupled receptors (GPCRs) and knowledge on this receptor family has provided several important paradigms about GPCR function and regulation. One of the most recent paradigms initially supported by studies on adrenergic receptors is that both βarrestins and G protein-coupled receptors themselves can act as scaffolds binding a variety of proteins and this can result in growing complexity of the receptor-mediated cellular effects. In this review we will briefly summarize the main features of βarrestin binding to the adrenergic receptor subtypes and we will review more in detail the main proteins found to selectively interact with distinct AR subtype. At the end, we will review the main findings on oligomerization of the AR subtypes.

The adrenergic receptors, and in particular the β 2 AR, are among the best characterized G protein-coupled receptors (GPCRs) and knowledge on this receptor family has provided several important paradigms about GPCR function and regulation.
One of the most recent paradigms initially supported by studies on adrenergic receptors is that G protein-coupled receptors can act as scaffolds binding a variety of proteins and this can promote multiple signaling events which results in growing complexity of the receptor-mediated cellular effects (reviewed in ref. [1]).
In the late 1980s, the first protein found to interact with the β 2 AR, beyond G proteins, was βARK1 (βAR kinase 1) [2], discovered as the first member of the G protein-coupled receptor kinase family (GRK) [3]. The ability of GRK2 and 3 to selectively interact with the agonist-bound form of the β 2 AR was a crucial observation to identify their role in homologous desensitization.
Soon after the discovery of GRKs, it became apparent that β 2 AR desensitization required also the interaction of an arrestin protein with the phoshorylated receptor [2,4]. This interaction resulted in both receptor-G protein uncoupling and receptor endocytosis into clathrin-coated pits [5].
In the last ten years, an important function of βarrestins has come to light as scaffolds for a growing number of *Address correspondence to this author at the Départment de Pharmacologie et de Toxicologie, Université de Lausanne, Switzerland; Tel: +41-21-6925350; Fax: +41-21-6925355; E-mail: susanna.cotecchia@unil.ch signaling proteins thus coordinating complex signal transduction events [6]. In particular, it is well established that βarrestins are scaffolds for components of the mitogen activated protein kinase (MAPK) cascade thus mediating MAPK activation induced by various GPCRs.
This seminal work encouraged the search for novel protein-protein interactions at several GPCRs with the aim of identifying previously unappreciated signaling mechanisms that might represent new targets of pharmacological intervention. A number of approaches have been followed to identify novel proteins interacting with the receptors, including yeast two-hybrid screen using cytosolic portions of the receptors as bait, pull-down or in vitro overlay assays using purified proteins, co-immunoprecipitation of receptorprotein complexes from recombinant or native cells, FRET (fluorescence resonance energy transfer) or BRET (bioluminescence resonance energy transfer) technology in cells. These studies resulted in the identification of a variety of proteins interacting with the adrenergic receptors, several of them in a receptor subtype selective pattern. A major challenge faced by these studies has been to identify the functional implications of these interactions. Some interacting proteins have been found to either promote or impair receptor-mediated signaling whereas others are involved in receptor trafficking or endocytosis.
Among the protein-protein interactions found to regulate GPCR function, receptor oligomerization has been extensively investigated in recent years [7]. Both homo-and hetero-oligomerization have been reported for different adrenergic receptor subtypes using different experimental approaches. This phenomenon seems to have implications in different aspects of receptor function, including its pharmacological profile, signaling, trafficking or endocytosis.
The canonical interactions of the adrenergic receptors with G proteins, GRKs and βarrestins have been extensively studied and exhaustively reviewed elsewhere [8,9]. In this review we will briefly summarize the main features of βarrestin binding to the adrenergic receptor subtypes and we will review, more in detail, a number of proteins found to selectively interact with distinct AR subtype. At the end, we will review the main findings on oligomerization of the AR subtypes. Considering the large number of studies on protein-protein interactions at GPCRs, our review might not systematically include all published data.
The direct interaction of GPCRs with selected partners has recently emerged as a new mechanism of receptor signaling and regulation. Since these mechanisms might be specific for distinct receptors or cell types, the study of these interactions has interesting implications in pharmacology and drug development.

βARRESTIN INTERACTION WITH THE AR SUBTYPES
The well established crucial role played by βarrestin1 and 2 in coordinating a variety of signaling networks might imply that these proteins can form macromolecular complexes with virtually any GPCR.
The interaction of βarrestins with the β 2 AR has been extensively characterized both at functional and molecular level using different approaches including in vitro binding of purified proteins, co-immunoprecipitation, BRET or FRET, βarrestin translocation as well as confocal microscopy to assess colocalization of the proteins [6,8,9]. The β 2 AR displays a pattern of interaction with βarrestins defined as "Class A" characterized by greater affinity for βarrestin 2 than 1 and a short-lived receptor/βarrestin complex leading to rapid receptor recycling after endocytosis. The interaction with βarrestins is important in mediating β 2 AR-induced activation of ERK1/2 (see references in [6]).
In contrast to the significant amount of data on the β 2 AR, much less is known about the interaction of βarrestins with other AR subtypes. The interaction of the β 1 AR with βarrestin is much weaker than that displayed by the β 2 AR subtype and this seems to correlate with the resistance of the β 1 AR to undergo agonist-induced endocytosis [10]. However, the β 1 AR can transactivate the epithelial growth factor (EGF) receptor in a βarrestin-dependent mechanism and this effect might have implications in cardioprotection [11]. This is suggested by a recent study reporting that recruitment of βarrestin to the C-tail of the β 1 AR is required for maintaining the β 1 AR/EGF receptor complex.
Also the β 3 AR does not seem to interact with βarrestins as suggested by two lines of evidence. First, the β 3 AR is resistant to agonist-induced endocytosis [12]. Second, the ability of the β 3 AR to activate MAPK is independent from βarrestin binding since its activation does not result in βarrestin recruitment to the plasma membrane [13].
The interaction of βarrestin with the α 2 AR was initially investigated measuring the effect of overexpressed βarrestin on receptor endocytosis [14]. Overxpression of βarrestin significantly increased the endocytosis of the α 2B and α 2C AR, but had no effect on the α 2A AR suggesting poor interaction of this receptor subtype with βarrestin. The lack of βarrestin interaction with the α 2A AR was confirmed by in vitro studies measuring the binding of purified βarrestin to peptides derived from the 3i loop of the three α 2 AR subtypes [15].
The interaction of the α 1a and α 1b AR subtypes with βarrestin has been investigated by a recent study using different approaches including co-immunoprecipitation, endo-cytosis and confocal microscopy [16]. Whereas the α 1b AR displayed robust agonist-induced endocytosis, the α 1a AR did not. The results from both co-immunoprecipitation experiments and βarrestin translocation assays indicated that the agonist-induced interaction of the α 1a AR with βarrestin was much weaker than that of the α 1b AR. The interaction of βarrestin with the α 1d AR has not been directly explored so far.
Altogether these findings indicate that the interaction pattern of βarrestin with distinct AR subtypes is divergent and correlates with differences in the internalization properties of the receptors. Overall, despite the large number of studies on the β 2 AR, a lot remains to be explored concerning the implications of βarrestin in adrenergic receptor function and regulation.

PROTEINS INTERACTING AT THE β 1 AR
The first cytoplasmic proteins found to interact selectively with the β 1 AR subtype were endophilins 1/2/3, also called SH3p4/p8/p13 [17]. This SH3 domain-containing protein family bound the proline-rich third intracellular (i3) loop of the β 1 AR in both pull-down assays and yeast two hybrid screens. The primary effect of this interaction was to promote agonist-induced internalization of the receptor and to modestly decrease its coupling to Gs. Intriguingly, a similar proline-rich sequence is found in the i3 loop of the β 3 or α 2A AR, but these receptors do not bind endophilins. The mechanism through which endophilin might regulate receptor coupling and internalization is not known, but it could include steric hindrance or allosteric modulation of G protein and arrestin binding.
A second class of proteins selectively interacting with the β 1 AR are proteins containing the PDZ (PSD-95/Discslarge/ZO-1 homology) domain which recognizes the extreme C-terminus of the target protein. The β 1 AR possesses a type I PDZ binding sequence, E-S-K-V, at the end of its C-tail and it has been shown to bind the postsynaptic protein PSD-95 [18]. This protein is abundant in brain where it co-localizes with the β 1 AR in postsynaptic densities. This interaction was found in a yeast two-hybrid screening using the β 1 AR Cterminus as bait and it markedly attenuated β 1 internalization, while having no impact on the receptor desensitization or signaling to adenylate cyclase. Association with PSD-95 could also facilitate the linkage of the β 1 to NMDA glutamate receptors, known to be regulated by adrenergic signaling in the brain, and, more in general, to facilitate targeting of the β 1 AR subtype at synapses. Other studies demonstrated that the interaction with PSD-95 is negatively regulated by agonist treatment, through phosphorylation of the receptor by GRK5, thus allowing receptor internalization [19].
A second PDZ domain containing protein interacting with the β 1 AR is membrane-associated guanylate kinase inverted-2 (MAGI-2), a multidomain scaffold protein, also known as S-SCAM (synaptic scaffolding molecule). The first PDZ domain of MAGI-2 binds with high affinity to the β 1 AR C-tail, as demonstrated by overlay and pull-down techniques [20]. The β 1 AR/MAGI-2 interaction occurred constitutively in cells, but it was further enhanced by isoproterenol treatment. It favored agonist-induced internalization of the receptor, an effect opposite to the one observed for PSD-95, another PDZ-domain protein [18]. MAGI-2 also promoted β 1 AR association to βcatenin, a known MAGI-2 partner.
A third PDZ domain containing protein binding the β 1 AR is GIPC (Gα-interacting protein-interacting protein, C terminus) [21]. This interaction decreased ERK1/2 activity stimulated by the β 1 AR, through a Gi-dependent process, with no observable effects on the Gs-mediated cAMP accumulation or on receptor internalization. One last example of a PDZ domain-mediated interaction with the β 1 AR is CAL (Cystic Fibrosis Transmembrane Conductance Regulator-Associated Ligand), found by pull-down techniques to interact with the ESKV sequence of the β 1 AR [22]. Overexpression of CAL, a protein localized in the Golgi apparatus, reduced surface expression of the receptor, a process competitively reversed by PSD-95.
GPCRs can indirectly activate Ras through the Gβγ subunits of the G protein that can recruit c-Src, Grb-Sos and PI3 kinase. The β 1 AR was the first GPCR for which a direct interaction with a Ras-GEF (GTP exchange factor), CNrasGEF, had been described [23]. CNrasGEF bound the ESKV sequence of the β 1 AR through its PDZ domain and mediated isoproterenol-induced Ras activation.
The importance of the PDZ binding sequence in the Ctail of the β 1 AR was highlighted by findings obtained expressing the receptor in mouse cardiac myocytes [24]. In these cells, stimulation of the β 1 AR increases the contraction rate, whereas the β 2 AR has a biphasic effect with an initial increase followed by a decrease mediated by receptor coupling to Gi. In addition, whereas the β 2 AR undergoes endocytosis, the β 1 AR does not. Interestingly, the mutation of the PDZ binding sequence of the β 1 AR enabled both receptor internalization and coupling to Gi thus providing evidence that differences in the interaction with PDZ containing proteins might dictate the distinct physiological effects induced by the two βAR subtypes.
Each of the four PDZ domain containing proteins above described, although binding to the same sequence of the β 1 AR, exerted different effects both on receptor trafficking and signaling. It is noteworthy that these four proteins do not share the same tissue distribution, PSD-95 and MAGI-2 being almost exclusively found in the brain, whereas GIPC and CNasGEF being predominantly expressed in the heart. A recent proteomic screen aimed at providing an exhaustive view of PDZ-mediated interactions at the β 1 AR, confirmed the association with PSD-95, MAGI-2, GIPC and CAL, and identified two novel ones, SAP97 and MAGI-3 [25]. MAGI-3 co-expression profoundly impaired β 1 AR-mediated ERK1/2 activation.
To assess the GPCR selectivity of these interacting proteins a recent study used a library of 59 GPCR C-tails, among which those of the β 1 and β 2 AR, in a pull-down assay [26]. This approach identified the lysosomal targeting protein GASP1 (G protein-coupled receptor-Associated Protein) as a potential interacting partner of the β 1 AR, but this finding was not explored further. GASP1 had been previously identified in a yeast two-hybrid screen as a protein interacting with the C-tail of the δ opioid receptor and shown to interfere with the post-endocytic sorting of the receptor. It was later discovered that GASP1 is member of a family of ten proteins displaying some sequence similarities whose functional implications are still largely unknown (reviewed in [27]). Whereas GASP1 can interact with several GPCRs other proteins of this family have been shown to modulate transcription. Considering these two important features, the GASP proteins might represent a promising area of investigation in the GPCR field.
Another protein recently shown to interact with the β 1 AR and influence it's trafficking is golgin-160 which is a ubiquitously expressed Golgi membrane protein [28]. Golgin-160, whose function in Golgi structure is unknown, can interact with the third intracellular loop of the β 1 AR and its depletion in cells leads to a significantly reduced cell surface expression of the receptor.
An interesting interaction was recently found between the β 1 AR and the adaptor protein 14-3-3epsilon both in recombinant cells and heart [29]. The receptor/14-3-3 complex can compete with the interaction occurring between 14-3-3epsilon and the voltage-gated potassium channel, Kv11.1, and this might represent a mechanism involved in adrenergic regulation of cardiac repolarization. Whereas the wild type β 1 AR inhibits potassium current, a receptor mutant lacking its interaction with 14-3-3 can activate it. This is a fascinating finding highlighting the role of multiple protein interactions in fine tuning of the physiological effects mediated by the adrenergic receptors on cardiac rhythms.
Another finding highlighting the potential functional impact of different signaling complexes in heart cells concerns the interaction of the β 1 and β 2 AR subtypes with phosphodiesterases (PDE) [30]. Whereas the β 2 AR forms a complex with βarrestin and the PDE4D5 isoform, the β 1 AR can directly interact with the PDE4D8 isoform and the agonist can dissociate this complex. Both the receptor and PDE4 are regulated in a complex manner by the cAMP/PKA cascade and binding of PDE4D to the β 1 AR might allow for the control of cAMP levels in proximity of the receptor.

PROTEINS INTERACTING AT THE β 2 AR
PDZ domain-mediated interactions are also known to occur at the C-tail of the β 2 AR, which includes the type I PDZ-binding C-terminal sequence D-S-L-L. The first protein found to interact specifically with this sequence of the β 2 AR, but not with β 1 AR, was NHERF1 (Na+/H+ Exchanger Regulatory Factor 1), also named EBP50 (ERM-Binding Protein 50) [31]. This protein contains two PDZ domains and regulates the activity of the Na + /H + exchanger type 3 (NHE3). In addition, NHERF factors represent a well established link between ERM (Ezrin-Radixin-Moesin) proteins and specific polytopic membrane proteins [32]. The β 2 AR/NHERF interaction might play a role in the receptormediated regulation of cellular pH. Typically, an increase of intracellular cAMP levels inhibits NHE3 activity (via PKAmediated phosphorylation of NHERF), but the stimulation of the β 2 AR potentiates it. This is probably due to the fact that binding of the β 2 AR to NHERF relieves the NHERFmediated inhibition of NHE3.
NHERF can also be a link between the β 2 AR and other proteins which interact with NHERF, such as the platelet-derived growth factor (PDGF) receptor [33] and cystic fibrosis transmembrane conductance regulator (CFTR) [34]. Whereas PDGF receptor-mediated signaling is potentiated by its interaction with NHERF, the formation of the β 2 AR/ NHERF complex might regulate PDGF receptor function. A macromolecular complex composed of the β 2 AR, CFTR and NHERF was shown to form in human airway epithelia and this complex might represent an important mechanism underlying the β 2 AR-stimulated increase of CFTR activity.
In addition to mediating these functional effects of the β 2 AR, the interaction with NHERF seems to play a direct role in receptor trafficking. The β 2 AR is a member of class A GPCRs, being characterized by rapid agonist-induced endocytosis and recycling back to the plasma membrane. It was demonstrated that disruption of either NHERF binding to the β 2 AR or NHERF binding to the actin cytoskeleton (mediated by NHERF/ERM interaction) caused missorting of endocytosed β 2 AR to the degradation pathway and prevented its recycling [35]. Actin depolymerization had a similar effect, implying that anchoring of the receptor to the actin cytoskeleton through NHERF and ERM proteins ensures its proper trafficking following endocytosis. A similar role for NHERF in directing the recycling of several other GPCRs was subsequently documented [32].
Another protein that associates with the distal part of the β 2 AR C-tail is NSF (N-ethylmaleimide-Sensitive Factor) [36]. Despite lacking PDZ domains, NSF also recognizes the extreme C-terminal sequence D-S-L-L and its association with the receptor is increased upon agonist treatment. NSF binding to the β 2 AR was shown to be required for both receptor internalization and recycling, but it is unclear whether a competition between NSF and NHERF binding to the receptor takes place and how this process is regulated.
The role of the PDZ binding sequence in the β 2 AR was validated in cardiac myocytes by the same group which investigated the β 1 AR-mediated effects on contraction rate [37]. Deletion of the PDZ binding motif in the β 2 AR abolished its coupling to Gi resulting in higher contraction rate, in contrast to the effect observed for the β 1 AR, where a similar mutation promoted Gi coupling and decreased contraction [24]. In agreement with previous findings, the mutated β 2 AR lacking interaction with NHERF was unable to recycle back to the plasma membrane. The experimental design did not allow to discriminate between NSF-and NHERF1-mediated effects.
Stimulation of the β 2 AR raises intracellular cAMP levels leading to activation of protein kinase-A (PKA). The functional connection between the β 2 AR and PKA prompted the investigation on potential interactions between this receptor and PKA-anchoring proteins, or AKAPs. The first AKAP found to bind the β 2 AR was AKAP250, also known as gravin. Association of gravin to the receptor was increased upon agonist treatment [38] and involved the Cterminal portion of the receptor [39]. Suppression of gravin expression using an antisense strategy disrupted receptor resensitization and impaired its association to GRK2, βarrestin and clathrin, suggesting that gravin might serve as a scaffold bringing together kinases, phosphatases and proteins of the endocytic machinery.
Another AKAP, AKAP79, directly and constitutively interacts with the β 2 AR and promotes its phosphorylation by PKA, which is obligatory for MAPK activation by the receptor [40]. The anchored PKA was further shown to phosphorylate GRK2 enabling its translocation to the membrane and subsequent phosphorylation of the β 2 AR [41]. Thus, AKAP79 is involved in the process of β 2 AR desensitization and internalization through the clathrindependent pathway. The molecular determinants of the receptor interaction with AKAPs were unfortunately not finely mapped and the question of the specificity of this interaction among adrenergic receptor subtypes has not been addressed yet.
An interaction of the β 2 AR with the adaptor protein Grb2 has also been reported [42]. This interaction was not identified through unbiased screening, but specifically tested based on the observation that insulin abolishes catecholamine response by stimulating tyrosine phosphorylation of the β 2 AR. This phosphorylation creates an SH2 site on the β 2 AR which induces Grb2 binding to the receptor at Tyr250/254 thus leading to receptor-G protein uncoupling. Grb2 binding to the receptor was increased upon cell treatment with insulin and promoted also receptor internalization [43].
Interesting interactions have been found between the β 2 AR and ion channels involved in the regulation of membrane excitability. The β 2 AR was found to directly interact with the voltage-gated calcium channel Cav1.2 in hippocampal neurons [44]. A complex containing the β 2 AR, the Cav1.2, G protein, an adenylate cyclase, PKA and the PP2A phosphatase might represent a mechanism ensuring a specific and highly localized signaling process.
It is well known that β 2 AR and PKA activation can increase BKCa activity in neurons, smooth muscle cells and other excitable cells. Recent findings indicate that the β 2 AR can interact with calcium sensitive K + channels (BKCa) in a complex containing the receptor, BKCa and AKAP79 [45]. Thus, the β 2 AR might interact with both the Cav1.2 calcium channel and the BKCa calcium sensitive K + channel, and these interactions might enable a highly localized control of membrane excitability.
A recent study reports a fascinating finding concerning the interaction between the β 2 AR and two proteins, the von Hippel-Lindau tumor suppressor protein (pVHL)-E3 ligase complex and the dioxygenase EGLN3 [46]. It is known that in response to hypoxia there is a decreased expression of β 1 AR in heart attributed to high levels of catecholamines whereas the β 2 AR abundance increases suggesting a direct regulation of this receptor by oxygen. The metallo-sensory enzyme dioxygenase EGLN3 transduces O 2 -responsive signals through modifications of target proteins. Thus, the interaction of EGLN3 with the β 2 AR results in hydroxylation of Pro382 and Pro395 of the receptor. After hydroxylation of the β 2 AR, pVHL-E3 ligase is recruited triggering ubiquitinylation of the β 2 AR and its degradation. This finding highlights the broadness and complexity of protein-protein interactions that can be explored to understand receptor function and regulation.

PROTEINS INTERACTING AT THE β 3 AR
The β 3 AR subtype was cloned several years after the other βAR subtypes and its distribution seems to be restricted to adipose tissue, skeletal and smooth muscles. For these reasons little is known to date about the specific interactions of this AR subtype. One cytoplasmic partner has been identified, namely the tyrosine kinase Src, that can bind through its SH3 domain to the polyproline region in the third intracellular loop and the C-tail of the receptor [13]. The direct binding of Src of the β 3 AR seems required for ERK1/2 activation by the receptor. Previous reports from the same group had demonstrated an involvement of Src in MAPK activation by the β 2 AR, but this effect was mediated by the interaction of Src with βarrestin. As mentioned above, direct Src binding to the β 3 AR occurred through its SH2 domain on the phosphotyrosine Tyr350. The paucity of data regarding β 3 AR interactions leaves open an entire field of investigation in the signaling of this receptor.

PROTEINS INTERACTING AT THE α 1A AR
Few interactions have been shown to occur selectively at the α 1a AR subtype. Yet, this receptor contains a PDZ binding sequence G-E-E-V at its C-terminus that can be expected to give rise to PDZ-domain mediated interactions. An early report, at the issue of a yeast two-hybrid screen, identified the type III PDZ domain of nNOS (neuronal nitric oxide synthase) as a potential α 1a AR interacting protein [47]. However, co-immunoprecipitation studies, while confirming this interaction, failed to highlight selectivity for the α 1a AR subtype since all three α 1 AR subtypes could be coimmunoprecipitated with nNOS and this even when they were lacking their C-terminus. This interaction appeared to be without apparent physiological implications in spite of the known role of NO in the regulation of blood pressure and of nNOS as local metabolic inhibitor of α 1 AR-mediated vasoconstriction.
Another study reported that mammalian tolloid (mTLD) could interact with the α 1a AR [48]. The CUB5 domain of mTLD, a zinc-finger matrix metalloprotease of the astacin family, interacted with α 1a AR C-tail in a yeast two hybrid screen. The interaction was specific for the α 1a AR and the binding region on the receptor could be narrowed down to a sequence of 37 aminoacids. Overexpression of mTLD reduced the number of cell surface receptors without affecting total receptor level or affinity when transiently expressed in HEK293 cells. mTLD also appeared to facilitate calcium signalling evoked by phenylephrine. No mechanism was proposed to account for the observed phenomena.
Interesting prospects were opened by the report of the direct interaction between RGS2 (Regulator of G protein Signaling 2) and the third intracellular loop of the α 1a AR [49]. RGS proteins are well characterized inhibitors of heterotrimeric G protein function, acting as GAPs (GTPase activating proteins) to increase the rate of GTP hydrolysis at Gα subunits and thus terminate signaling. More than 30 RGS proteins have been identified so far, but many RGS proteins can non-selectively bind to and inhibit Gαi/o and Gαq11 in reconstituted systems, suggesting that other factors may regulate their specificity for a particular signaling pathway. RGS2 was found to interact with the α 1a AR third intra-cellular loop confirming what previously shown for other Gq-coupled receptors, namely the cholinergic muscarinic M1, M3 and M5 receptors [50] and it inhibited agonistinduced inositol phosphate responses without affecting ligand binding.

PROTEINS INTERACTING AT THE α 1B AR
Two main interacting partners were pulled out of a yeast two-hybrid screen for the α 1b AR: the µ2 (or AP50) subunit of the clathrin adaptor complex AP2 [51] and ezrin, a member of the ERM protein family [52]. The AP2 complex is part of the endocytic machinery mediating clathrin-dependent endocytosis of membrane proteins and it is recruited to agonist-activated GPCRs through the intermission of βarrestins. Interactions of the AP50 subunit are dependent upon a YxxF motif present in the cargo protein. However, in the case of the α 1b AR, binding of AP50 relied on a basic stretch of eight arginines in the proximal C-tail of the recaptor. Direct association of the α 1b AR to AP50 contributed to the agonist-induced internalization of the receptor as demonstrated by the fact that a receptor mutant lacking the AP50 binding motif was delayed in internalization. The presence of the eight arginine motif in the C-tail of a GPCR is not common, which rules out the hypothesis that direct AP50 interaction is a common mechanism for clathrinmediated endocytosis. Interestingly, this feature is shared by the α 1d AR, which contains a stretch of seven positive charges in its C-tail, but no studies were undertaken using this receptor subytpe.
In addition to AP50, the same yeast two-hybrid screen identified ezrin as a potentially direct binding partner of the α 1b AR [52]. Ezrin belongs to the ERM family of proteins, primarily described as linkers between membrane proteins and cortical actin. Ezrin was also shown to be involved in the remodelling of the actin cytoskeleton, in the modulation of Rho signaling (by binding to Rho GTP dissociation inhibitor (GDI) and through direct association to several Rho GTP/GDP exchange factors (GEFs) as well as in anchoring of PKA. Ezrin interactions with polytopic membrane proteins generally occur through the adaptor proteins EBP50 (NHERF1) and E3KARP (NHERF2), but direct contacts were also described between ezrin and proteins such as the Na+/H+ exchanger type 1 (NHE1) and podocalyxin. So far, a role for the ERM proteins in GPCR trafficking was inferred from the finding that NHERF1 binding to some GPCRs promoted their recycling, depending on its binding to ERM proteins. The α 1b AR is the first GPCR for which a direct interaction with ezrin has been found. Disruption of this interaction by overexpression of a dominant negative mutant of ezrin inhibited receptor recycling after internalization, as did actin depolymerization. Thus, the involvement of ERM proteins in GPCR recycling, through either a direct or indirect interaction with the receptor, might represent a general paradigm for the regulation of GPCR trafficking.
Intriguingly, ezrin and AP50 shared the same binding site on the α 1b AR C-tail, consisting in the eight arginines stretch positioned after the putative palmitoylation site. In pulldown experiments the binding domain of ezrin and that of AP50 competes with each other for the same binding site of the receptor C-tail (unpublished data). How these events are regulated within the cell is a matter that awaits further inquiry.
Another protein, the receptor for globular "Heads" of c1q (gC1qR), was reported to interact with the same argininerich sequence in the α 1b and the α 1d AR [53]. gC1qR is a glycoprotein mainly displaying intracellular localization, but also present on the surface of macrophages and T cells through anchoring to β-integrin, where it is part of a complement receptor. No functional relevance was demonstrated for its interaction with the α 1b or α 1d AR.

PROTEINS INTERACTING AT THE α 1D AR
The α 1d AR was for a long time a "poor relative" to the other α 1 AR subtypes, the α 1a and α 1b because poorly expressed at the cell surface in heterologous systems, probably because of its long N-terminus. This peculiarity hampered the investigation of its potential interactions with other proteins. Apart from the above mentioned interaction with gC1qR, whose functional implications are unknown [53], another interacting partner of the α 1d AR was α-syntrophin [54]. α-syntrophin, a protein containing one PDZ domain and two PH (pleckstrin homology) domains, specifically recognized the C-tail of the α 1d AR, but not that of the α 1a or α 1b , in the yeast two-hybrid assay. The PDZ domains of syntrophin isoforms α, β1 and β2, but not γ1 or γ2, could interact with the α 1d AR C-tail. The α 1d AR possesses the Cterminal sequence E-T-D-I, whose mutation impaired syntrophin binding to the receptor and markedly decreased norepinephrine-induced inositol phosphate accumulation. This mutation also dramatically decreased receptor expression levels. Interestingly, syntrophins seemed to interact equally well with intracellular or surface-expressed α 1d AR receptors. Taken altogether these results suggested that syntrophins act to maintain the stability of the α 1d AR through a PDZ-mediated interaction.

PROTEINS INTERACTING AT THE α 2 AR SUB-TYPES
The three α 2 AR receptor subtypes (α 2 A α 2 B and α 2 C) are coupled to the Gi/o family of heterotrimeric G proteins, and hence to the inhibition of adenylyl cyclase and voltage sensitive calcium channels, and to the activation of potassium channels. They are differently expressed in various tissues including the basolateral membrane of polarized renal epithelial cells where differences in the targeting and trafficking of the three subtypes have been found. The proper basolateral retention of the α 2 AR receptor subtypes is dependent upon the integrity of their third intracellular loop, a finding which prompted the search for partners interacting with this region of the receptors. A gel overlay strategy using in vitro translated i3 loops of the α 2 AR receptors highlighted the interaction of the zeta isoform of 14-3-3-proteins [55]. 14-3-3 proteins are ubiquitous and involved in the regulation of a number of signaling pathways, among which the Ras/ MAPK cascade. 14-3-3 interacted preferentially with the α 2A and α 2B than with the α 2C . A detailed study of the sequence requirements for this interaction at the α 2A AR did not identify a single linear motif suggesting that a three dimensional structure in the i3 loop is needed for 14-3-3 binding [56].
An important binding partner of all three α 2 AR subtypes is spinophilin [57,58]. This interaction was specifically tested with the rationale that spinophilin is enriched beneath the basolateral membrane of MDCK cells. Spinophilin binding to α 2A AR was enhanced by agonist treatment and the region responsible for binding was loosely mapped to the Nand C-terminal ends of the i3 loop [56]. Interaction with spinophilin contributed to the cell surface stabilization of the α 2B AR subtype, a receptor that displays the unique property of being first randomly delivered to both apical and basolateral side of the cell, with a much faster apical versus basolateral turnover which ends up in its accumulation on the basolateral side. Apical delivery of a spinophilin subdomain extended the half-life of the α 2B AR in this region; furthermore, agonist-stimulated internalization of the receptor was accelerated in fibroblats derived from spinophilin knock-out mice [59]. Presumably, similar effects could occur at the other two α 2 AR subtypes. An explanation of this effect came from the finding that spinophilin blocks GRK2 association to the α 2B ARs thus inhibiting receptor endocytosis [60].
Spinophilin was also found to interact with other GPCRs, including the α 1b AR, as well as with the N-terminal domain of RGS proteins (RGS1, 2, 4 and 16) which participates in GPCR recognition [60,61]. Thus spinophilin might represent an interesting functional bridge between RGS and α 1 AR subtypes that don't bind RGS, like the α 1b AR. In fact, it has been found that spinophilin increases the RGS2induced inhibition of the α 1b AR calcium response. In support of this finding, a constitutively active α 1b AR mutant did not bind spinophilin and was resistant to inhibition by RGS2. Similar resistance to RGS2 inhibition was found to occur in spinophilin knock-out cells. These data offer a glimpse into a potentially more general regulatory mechanisms of GPCR function by spinophilin.
The most recent protein found to interact with α 2 ARs is ubiquitin carboxyl-terminal hydrolase-L1 (Uch-L1), a protein associated with Parkinson disease [62]. Uch-L1 binds preferentially to the α 2A AR subtype and its overexpression inhibits the receptor-induced activation of MAPK. This interaction might have implications in the neuroprotective effect of α 2 ARs which needs to be further investigated.

OLIGOMERIZATION OF THE ADRENERGIC RECEPTORS
Findings in the last decade challenged the widely held view of GPCRs functioning as monomeric units [7]. Early biochemical studies showing higher molecular weight recaptor bands migrating in SDS-PAGE gels, stabilized by crosslinking, suggested that adrenergic receptors might form oligomers. Afterwards, co-immunoprecipitation of differentially tagged GPCRs or functional complementation of pairs of co-expressed inactive receptor mutants provided stronger evidence in favor of GPCR oligomers. The widespread use of biophysical techniques such as FRET or BRET between GPCRs carrying the appropriate pair of fluorescent/ bioluminescent labels suggested oligomerization of a variety of GPCRs. Each technique employed has its own shortcomings: whereas co-immunoprecipitation cannot rule out indirect interactions, energy transfer techniques can only certify that the two partners are in close proximity, not necessarily in immediate contact. Therefore, whether receptor oligomerization involves direct interactions among receptor monomers versus their increased proximity in micro-domains of the cell membrane cannot be unequivocally demonstrated. Despite the fact that the precise molecular events are not fully understood, the term of GPCR oligomerization is largely accepted to indicate the existence within the cell membrane of macromolecular complexes formed by two or more receptor monomers. It is important to highlight that most studies have been performed in recombinant cells overexpressing the receptors with very few examples of oligomerization occurring in physiological systems.
Whereas the stoichiometry of these complexes is hard to assess with either biochemical or biophysical methods used so far, the structural architecture of receptor oligomers has been addressed by some studies using computational, pharmacological and biochemical approaches. Two types of receptor-receptor interaction have been proposed, involving either lateral contact or domain swapping. A model was proposed for a β 2 AR dimer where the two receptor monomers swapped helices V and VI [70]. Other studies pointed at helices I and VII [66] or I/II and III/IV [71] as the probable interface for receptor-receptor contact in the case of the α 1b AR. Neither the C-terminal tail of the α 1b AR nor its glycosylation state or the presence or absence of a glycolphorin dimerization motif GxxxG in the transmembrane domains affected its oligomerization [66]. This was in contrast with the previous report of Hébert et al., showing that helix VI containing the GxxxGxxxG motif was responsible for oligomerization of the β 2 AR [63].
The functional implications of AR homo-ligomerization have been explored by several studies. Preventing receptor association has been technically difficult with few exceptions. In the case of the β 2 AR, a peptide derived from helix VI was shown to disrupt dimerization and its use was demonstrated to impair isoproterenol-stimulated cAMP production by the β 2 AR [63] implying that β 2 oligomers would be the functional form of the receptor.
Among the functional implications of AR homooligomers, different studies investigated their localization within the cell. BRET studies suggested that β 2 AR oligomers are present at the plasma membrane [64]. Biotinylation with a membrane-impermeable showed expression of α 1a , α 1b and α 1d homodimers at the cell surface [67]. Moreover, homooligomerization of the β 2 AR seems to be a prerequisite for its plasma membrane targeting [72]. In fact, the peptide derived from helix VI of the β 2 AR, which blocks receptor oligomerization, prevented normal targeting of the receptor to the plasma membrane. These results imply that receptor oligomers form as early as the stage of their synthesis in the endoplsmic reticulum and that oligomerization influences their maturation and trafficking ability from then onwards. Indeed, co-expression with a recycling-defective mutant diverted wild type β 2 AR from normal recycling to the plasma membrane to the proteolytic degradative pathway [73]. Similar data were obtained for the α 2B AR subtype; when co-expressed with a mutant deficient in cell surface targeting, the wild type α 2B AR remained trapped in the endoplsmic reticulum [69].
A matter of debate has been the regulation of the oligomeric status of GPCRs by ligands. One among the first reports indicated that agonist binding could increase the amount of β 2 AR dimers [63], corroborated by BRET data [64]. In contrast, agonist treatment had no apparent effect on the oligomerization of α 1 AR subtypes [66,67]. Furthermore, constitutively active or non-functional α 1b AR mutants displayed the same propensity to oligomerize as the wild type receptor [66], indicating that the activation state of the receptor is irrelevant for this process.

Hetero-Oligomerization
Different AR subtypes are often co-expressed in the same cells and cross-talk among them can occur. Various studies addressed the hypothesis that hetero-oligomerization could account for cross-talk effects. A number of interactions between different AR subtypes or between ARs and more distantly related GPCRs have been reported ( Table 2).
Hetero-oligomerization was found to have various functional effects upon the behavior of individual receptors, ranging from regulated targeting to modified pharmacological, signaling or trafficking profile in recombinant systems. Co-expression of the α 1d AR with the α 1b AR [76] or the β 2 AR [77] was able to rescue surface expression of the α 1d AR, the majority of which is intracellular when expressed alone in various cell lines. A similar phenomenon was already observed for the GABA-B receptor and for the calcitonin/adrenomedullin receptor, both receptors needing hetero-dimerization in order to be properly targeted to the plasma membrane. Interestingly, the interaction with the α 1b AR modified the pharmacological profile of the α 1d AR which looses its affinity for its selective ligand BMY7378 when it is co-expressed with the α 1b AR. The α 1b /α 1d dimer behaves as a single functional entity with increased response to norepinephrine relative to either monomer alone. The α 1d AR receptor was long supposed to be little expressed in the heart, as its selective ligand BMY7378 could detect only minimal levels of the receptor. However, these findings should be considered in a new light, given that the α 1b and α 1d AR subtypes co-exist in this tissue and the pharmacological profile of the α 1d AR might be different than expected because of oligomerization.
Hetero-dimerization of the β 2 AR with the α 2C AR, which is normally poorly expressed in recombinant cells, increased the expression at the cell surface and ERK1/2 signaling of the α 2C AR [79]. Coexpression of the α 2C AR with more than twenty-five GPCRs revealed that only the β 2 AR induced this effect. Coexpression of the β 2 AR with the olfactory receptor (M71) also promoted the cell surface localization of this receptor which is often co-expressed with the β 2 AR in olfactory neurons [80]. Again, only the β 2 AR among the AR subtypes had this effect.
Beyond its effect on receptor targeting to the cell surface, hetero-oligomerization seems to have an impact on various aspects of receptor trafficking and signaling. For example, several studies have shown that receptor monomers can mutually influence their endocytosis pattern. Whereas the α 1b AR undergoes agonist-induced internalization, the α 1a AR does not. However, when the two AR subtypes were coexpressed forming heterodimers, the endocytosis of each monomer could be triggered by stimulation of the other [66]. Colocalization of the two monomers could be seen in endocytic vesicles suggesting that the α 1a /α 1b dimers remained stable throughout the endocytosis process.
Strikingly, in β 1 /β 2 heterodimers, the internalization of the β 2 AR within the complex was inhibited. Furthermore, β 2 AR-induced ERK activation was equally blocked by coexpression of β 1 AR [78]. Similar results were also found for the β 2 /β 3 heterodimer [79] in which β 2 AR internalization was impaired. The β 3 AR is resistant to agonist-promoted endocytosis and, like the β 1 AR, acted as a dominant negative on β 2 AR internalization.
Functional cross-talk was also described between the β 2 AR and the opioid receptors (OR) that are coupled to stimulatory and inhibitory G proteins, respectively. Whereas in the β 2 /δ opioid dimer the β 2 AR and δOR could facilitate the endocytosis of each other, in the β 2 /κ opioid dimer the internalization of the β 2 AR was inhibited. Moreover, isoproterenol-induced MAPK activation was diminished in presence of κOR, showing that the κOR acts as a dominant negative modulator of the β 2 AR [81].
The same group explored the possible interaction between α 2 AR and opioid receptors that colocalize in neurons and affect the nociceptive response. The α 2A /µ opioid could be isolated from recombinant cells as well as from primary neurons. In the α 2A /µ opioid dimer the activation of each monomer increases G protein and MAPK signaling whereas the activation of both monomers decreases it [86]. In addition, when the receptors were expressed in a neuronal cell line the α 2A AR increased the δOR-mediated neurite outgrowth [87]. These results support the notion that α 2 AR and opioid receptors are synergic in spinal analgesia as also demonstrated by the decrease in morphine-induced analgesia in α 2A -/-knock out mice. Thus, the physical association between the α 2 AR and opioid receptors might explain their functional interaction.
Some interesting findings have been reported concerning the potential implications of βAR hetero-oligomerization with other GPCRs in native tissues. Whether in native tissues simultaneous synthesis of hetero-dimer partners can occur or different receptors are simply clustered in the same cell membrane compartment is not known.
In freshly isolated mouse cardiomyocytes, interesting effects were reported on contractility induced by angiotensin II (AgII) acting at AT1 receptor or isoproterenol acting at βAR [82]. Whereas the beta-blocker propranolol could inhibit the effect of AgII, the AT1 blocker valsartan inhibited that of isoproterenol. This transinhibitory effect of the two antagonists seemed related to inhibition of receptor coupling to its cognate G protein. Biochemical experiments on β 2 AR and AT1 receptor expressed in HEK cells indicated that the two receptors could be co-immunoprecipitated thus suggesting that hetero-oligomerization of the two receptors could be the basis of their functional interaction in heart.
Hetero-oligomerization between the β 2 AR (coupled to Gs) and the EP1 prostaglandin receptor (coupled to Gq) has been observed in airway smooth muscle [84]. In mouse tracheal rings, activation of the β 2 AR induces muscle relaxation whereas stimulation of EP1 alone has no effect on contraction. However, stimulation of the EP1 receptor profoundly reduced the β 2 AR-induced muscle relaxation and cAMP accumulation. The modulatory effect of EP1 receptor on the β 2 AR might depend on the interaction between the two receptors which form heterodimers in airway smooth muscle cells.
Hetero-oligomerization has also been recently observed using BRET between the β 2 AR and the CB1 cannabinoid receptor expressed in recombinant cells [85]. The formation of oligomers could be increased by a CB1 inverse agonist. Interestingly, co-expression of the two receptors resulted in increased localization at the cell surface and decreased constitutive activity of the CB1 receptor. Complex functional interactions on signaling between the two receptors have been found both in the recombinant system and in primary human trabecular meshwork (HTM) cells from the eye, a tissue in which the two receptors are natively co-expressed. Beyond mutual effects on ERK activation, activation of one receptor induces cross-desensitization of the other both in recombinant and HTM cells. This might be relevant in drug therapy since CB1 agonists and β 2 AR antagonists can both reduce intraocular pressure.
Very recently, an elegant study demonstrated that the β 2 AR can form a signaling complex with the GluR1 subunit of the AMPA glutamate receptor including also the trimeric Gs protein, adenylate cyclase and protein kinase A [88]. This complex seems to be important to allow β 2 AR-induced phosphorylation of the GluR1 which results in increased GluR1 cell surface expression and current amplitudes.
Most of the interactions involving the AR subtypes have been found using various screening approaches (proteomic, yeast two hybrid or pull-down experiments) without a specific bias towards the interactions searched. In most studies, the investigation of the interactions has been performed in recombinant systems in which only a limited number of functional parameters common to all GPCRs can be explored, i.e. trafficking, signaling through known pathways, receptor pharmacology. This resulted in the identification of a number of proteins affecting receptor endocytosis (endophilins, PSD-95, MAGI-2, NSF, AKAP250, Grb2, AP50) and few others stabilizing receptor expression at the cell surface (golgin-160, spinophilin, syntrophin), favoring recycling (ezrin, NHERF) or regulating receptor coupling to G proteins (RGS2) ( Table 1). Most of these interactions have been investigated only at a specific AR subtype with few exceptions. For example, the role of spinophilin in stabilizing receptor expression at the cell surface seems a property of all three α 2 AR subtypes (α 2A , α 2B and α 2C ) [55,56]. Spinophilin has been found to interact also with the α 1b AR displaying, however, a different effect, i.e. the modulation of calcium signaling. Spinophilin certainly represents a protein of pharmacological interest which should be further investigated because of its well established role in targeting the α 2 AR subtypes in polarized cells as well as in regulating their function.
It seems quite evident that proteins containing the PDZ domain can interact with both the β 1 and β 2 AR. However, some PDZ domain containing proteins seem to display a selectivity towards either one of two receptors. For example, NHERF interacts with the C-tail of the β 2 AR, but not with the β 1 [31]. PSD-95 interacts selectively with the β 1 AR being both proteins expressed in post-synaptic densities [18]. The interaction of PDZ proteins with the β 1 and β 2 AR is also one of the few examples in which the functional implication has been investigated in a physiological system, i.e. mouse cardiac myocytes [24,37]. Deletion of the PDZ binding motif in the β 2 AR abolished its coupling to Gi resulting in higher contraction rate, in contrast to the effect observed for the β 1 AR, where a similar mutation promoted Gi coupling and decreased contraction. Because of the functional relevance of these interactions, PDZ domain containing proteins might represent interesting targets to pharmacologically interfere with trafficking and signaling of either the β 1 or β 2 AR. Recently, some success has been reported in designing small peptides blocking PDZ interactions [89]. In addition, despite the wide distribution of PDZ proteins, the overlapping expression of interacting partners might be restricted to some tissues and this could be functionally relevant. More detailed studies on this important family of proteins in the regulation of AR subtypes are required and might have important implications.
Altogether, these latter studies clearly indicate that the full elucidation of signaling events in time and space will depend on a much deeper understanding of the interactions among receptors and signaling molecules. Developing drugs acting at distinct receptor-protein interfaces might represent an approach to achieve more cell specific pharmacological effects. However, this field is still at an early stage because of the complexity of studying these events in physiological cell systems as demonstrated by the limited number of studies published so far. Therefore, the fine-tuning of GPCR activity by receptor-interacting proteins is a very promising area of investigation in which a lot remains to be explored [90]. Studies in recombinant systems can certainly provide some useful information, but the real challenge concerns the possibility of exploring the functional implications of a variety of interactions in different tissues and physiological conditions. Without these studies it will be difficult to assess which of these interactions are druggable.