mitogen
Bluelighter
- Joined
- Nov 8, 2004
- Messages
- 127
"GPCRs: part of a network of signalling machinery" [long, advanced(ish) molpharm]
Hey guys, here's the latest addition to my series of articles and reviews this year. I've also posted it into the Textbook thread, so if anyone is keen to rip it up, head over there. For general comments and discussion, reply to this thread.
I can't post the .doc file here because its too big, so that means, unfortunately, no figures. If anyone so desires the figures, drop me a line on [email protected] - thanks.
“G-protein coupled receptors: part of a network of signalling machinery”
The section on “G-protein coupled receptors” (GPCRs) in the popular undergraduate level text book “Pharmacology” by Rang, Dale, Ritter & Moore [1], presents a relatively one-dimensional view of GPCR signalling. A neat diagram sums up the events that occur on binding of a ligand to its cognate GPCR: ligand binding attracts a GTPase (‘g’-) protein to the receptor, and the GDP bound to the g-protein is swapped for a GTP. The now-activated subunits of this heterotrimeric g-protein then dissociate and activate another molecule in the signalling cascade, such as adenylyl cyclase, which subsequently produces ‘second messenger’ molecules. These second messengers are responsible for activation of other downstream effectors, and the signal that began with a ligand binding to the GPCR is propagated.
In contrast, even a cursory glance at any recent review [2,3,4] on the subject of GPCR signalling will reveal to the reader that the textbook model of GPCR signalling is woefully inadequate. Rather, these reviews present the concept that a whole network of interacting proteins and biomolecules nucleated around scaffolding proteins are involved in what was once thought to be a relatively simple, linear transduction process. These protein networks are involved with fine-tuning and regulation of every facet of GPCR function. The ‘receptosome’ concept, that whole networks of molecules are spatially compartmentalised into plasma membrane microdomains such as caveolae and other lipid rafts is integral to all recent models of how GPCR signalling is effected.
In this review I will discuss how each receptosome exists as a self-contained, functional signalling unit, and the importance of spatial compartmentalisation of GPCR signalling machinery. Important experimental observations which lead to the invalidation of older GPCR signalling dogma and forced reconsideration of the whole signalling paradigm will be included. Attention will be paid to the roles that particular groups of proteins play in receptosomes, and the modality of their interactions with other receptosome proteins. Finally I will discuss some of the methodology that is currently being used to determine these interactions and their importance to aspects of GPCR signalling.
The old GPCR signalling dogma:
The older “1-dimensional” (meaning that signal transduction follows a defined, stepwise path, as opposed to three dimensional networks of interactions) model of GPCR signalling has been superceded by models like the ‘receptosome’ model described in this review. Essentially, too many contradictions of and paradoxes in the old model arose from experimental observations. Some of the observations that forced the creation of a newer more comprehensive model are detailed below.
Old dogma states that a specific ligand binds to its cognate GPCR, which then undergoes a conformational change, such that its cytosolic domain develops high affinity for one particular G-protein family and subtype, which it recruits. The activated GPCR activates the recruited G-protein by functioning as a guanine nucleotide exchange factor, exchanging G-protein bound GDP for GTP. The activated G-protein then splits into its alpha and beta-gamma subunit components, which activate secondary targets. G-alpha usually modulates the activity of a second messenger producing enzyme, such as activation of phospholipase C in the case of Gq-alpha, or activation of adenylyl cyclase by Gs-alpha. The second messenger activates second messenger dependent kinases which proliferate the signalling cascade. Additionally, there is room in the one-dimensional model for GRK mediated phosphorylation and arrestin mediated desensitisation, which is why although these molecules were discovered relatively early on, they did not push for creation of a new model. [1,3,4]
Organisation of GPCR signalling machinery
One of the fundamental concepts of biochemistry is that the proteins that comprise the majority of cellular machinery interact with each other as huge networks of multiprotein complexes, with specific chemical affinities determining the strengths of these interactions [1,5]. Thus when a ligand binds to its cognate GPCR, a conformational change is induced in the GPCR which creates a chemical site for which specific G-protein(s) have high affinity. Additionally, according to the laws of mass action, the magnitude and rate of chemical interactions and reactions are heavily dependent on the concentrations of the reactants. This raises a fundamental problem with the classical one-dimensional theory of GPCR signalling as can be found in most generic pharmacology textbooks: if one looks at the average concentration of each protein component involved in GPCR signalling, it is far too low to possibly account for the rapidity of the biochemical response to receptor agonism [5]. Kinetics of the protein-protein interactions required to form a signalling cascade must therefore be simply too unfavourable for any agonist directed response to occur if these proteins were randomly or even uniformly distributed across the plasma membrane or cytosol. This appears to be a massive flaw in the old model of GPCR signalling.
Several more flaws can be identified when the old model is compared with recent observations about the characteristics of GPCR signalling.
The old dogma of “1 g-protein couples to one GPCR” has been comprehensively disproved – in fact GPCRs more often than not couple to more than one G-protein [6]. This has significant ramifications for the signalling pathways activated by a particular GPCR. One particularly illuminating example of this observation is that the thyrotropin receptor is able to couple to all four major G-protein families [6]. Other experiments have shown that the majority of GPCRs have at least some affinity for each type of G-protein: therefore the preference for activation of a particular G-protein subtype actually lies on a continuum. In light of this concept, it is possible to infer that it is possible to describe GPCR interactions with particular G-proteins statistically: for example, a particular GPCR may interact with G-protein X 90% of the time, G-protein Y 9.99% of the time and G-protein Z a biochemically negligible 0.01% of the time. These statistics would be based on the chemistry of the interaction sites on the GPCR and G-proteins X, Y & Z. Interaction of the GPCR with G-protein X is obviously the most thermodynamically favourable binding interaction in a mixture of the four proteins at equilibrium.
A further observation that can not be integrated with the old model of GPCR signalling is that agonism by different ligands induces GPCRs to have different affinities for particular G-proteins. The paper “Opioid agonists differentially regulate Mu-opioid receptors and trafficking proteins in vivo” [7] is a good example of how different agonists can induce different biochemical responses in the cell. It is likely that the mechanism for this involves the two agonists used, morphine and etorphine, inducing different receptor conformations, and therefore recruiting different groups of G-proteins to the Mu-opioid receptor.
It is also possible, however, that this effect is not actually mediated by G-proteins at all, and involves direct interaction of other non-G-protein signalling machinery at the Mu receptor. Models have been suggested in which particular receptors may have a number of different conformations which they can assume, and different agonist ligands thermodynamically stabilise particular conformations, which each have a set of G-proteins they activate to different extents. This is a discrete model – there are a defined number of conformations that a receptor can take, and the potency of the agonist to induce that conformation and therefore the overall activity of the drug at the receptor, depends on the degree of thermodynamic stabilisation of that conformation [8]. Alternately, it is possible to imagine a continuous model, where each agonist induces an individual receptor conformation, which alters the G-protein coupling of the receptor and therefore the properties of the signal induced by that agonist. It is even possible to amalgamate these two theories, and conceptualise a model where each different agonist does continuously induce a different receptor conformation, but there are ‘peaks’ in agonist affinity and efficacy which correspond to stabilisation of particular, discrete conformations [9]. Needless to say, a “one GPCR binds to one G-protein” model is completely unable to account for any of these ideas.
Another observation that has forced progression from the older GPCR signalling dogma is that G-protein coupling is not necessarily required for biochemical responses to receptor agonism [10]. Following receptor activation and subsequent G-protein activation, the GPCR is often phosphorylated by a G-protein coupled receptor kinase (GRK,) [10] (or sometimes a by second messenger dependent kinases,) [11] and it is this chemical modification that creates a binding site on the GPCR for a group of proteins called arrestins, which attache to the GPCR and blocks any further coupling to G-proteins, in effect causing the cessation of G-protein mediated signalling [10]. It has been shown, however, that Beta-arrestin may act as a scaffolding molecule and serve to recruit other non-G-protein related signalling machinery. Experiments have shown that arrestin-2 can recruit the tyrosine kinase Src by binding to its SH3 domain, and can also activate MAP kinase pathways. Other experiments showed binding of JNK3 and ASK1, which is a JNK kinase kinase. Thus GPCRs can activate MAPK and tyrosine kinase pathways via their interaction with arrestin proteins [11]. Recent studies have shown an interaction between activated beta2adrenoreceptors (B2AR) and Src which is increased by overexpression of beta-arrestin. Additionally, inhibition of beta-arrestin binding to either B2AR or Src attenuates B2AR mediated activation of MAPK. [10]
Beta-arrestin has also been implicated in regulation of receptor trafficking and endocytosis by its interaction with the heavy chain of clathrin and the clathrin adaptor protein AP2 [10]. These observations of arrestin molecules as scaffolds that nucleate assembly of non-G-protein mediated signalling processes add further detract from old GPCR signalling dogma.
Described above is a series of experimental observations that obligatorily invalidate the one-dimensional model of GPCR signalling, while at the same time building the concept of GPCR signalling as involving a whole network of interacting proteins, with some acting as nodes and scaffolds onto which other proteins nucleate, while others are involved in fine tuning and regulation of the signalling machinery, and still others involved in the trafficking and regulation of the receptors themselves. The most important feature of a newer model of GPCR signalling would have to incorporate the principles of biochemical kinetics and concepts such as collision theory. If all the protein components required for GPCR signalling were to be randomly or even uniformly distributed throughout the cytosol and plasma membrane, the observed rapid response of GPCRs to agonism could not possibly occur. Thus, a new model must include a spatial dimension. The components must be spatially organised such that the biochemistry is actually possible. In answer to this requirement, the literature is packed with reviews and papers documenting the existence of membrane microdomains, or lipid rafts, such as caveolae, in which many of the signalling components and receptors are often congregated.
Caveolae are small (50-100nm) invaginations in the plasma membrane of cells, and are considered to be a subclass of lipid rafts. The lipid composition of caveolae includes characteristically high levels of cholesterol and sphingolipids, along with caveolins, a group of proteins which comprises three isoforms: cavelolin-1, caveolin-2 and caveolin-3. It is generally accepted that caveolae will form if a cell expresses caveolin-1, or in the case of striated muscle myocytes, caveolin-3. Thus, while the plasma membranes of most or all cells contain lipid rafts, only some cells contain caveolae. A 2003 paper in the Journal of Neurochemistry [12] gives a good example of a GPCR being localised to caveolin membrane fractions, and shows “molecular and functional association of mGluR1a receptors with caveolins.” The study demonstrates that agonistic activation of mGluR1a receptors increased ERK phosphorylation in low density caveolin enriched membrane fractions, but not in high density membrane fractions containing no caveolins. Also mentioned in the study was the observation that mGluR1 heterodimerizes with adenosine A1 receptor and calcium sensing receptor; all three of these proteins localise to caveolin rich membrane microdomains [12].
Another example of the role of localisation of receptors and signalling machinery to caveolae is the comparison of Beta1-adrenoreceptor (B1AR) and B2AR signalling in cardiac myocytes. B2ARs activate adenylyl cyclase 6 (AC6) with a lower efficacy than B1ARs, and it appears that this is due to rapid translocation of B2ARs out of caveolae and into clathrin coated pits after receptor activation. AC6 is localised strictly to caveolae, and as such when the B2AR is translocated, it can no longer physically contact AC6 to activate it [5].
A concept that is integral to the model of spatial compartmentalisation of signalling proteins into regions such as caveolae and other lipid rafts is the selective expression of certain isoforms of G-proteins and second messenger synthesising enzymes such as adenylyl cyclase (AC) to particular types of raft, or not to any raft at all. There are nine AC subtypes, but not all of them localise to lipid rafts [5]. Therefore, the fact that many different subtypes exist of G-proteins, second messenger synthesising enzymes and other signalling proteins such as Regulators of G-protein Signalling (RGS,) is a way of increasing the diversity of plasma membrane domains and microdomans.
Many proteins which associate with caveolin proteins contain a caveolin- or caveolin-like- binding motif [13]. Le Clerc et al. published a study in 2002 in the journal Endocrinology examining the effect of Angiotensin II Receptor Type 1’s (AIIR1) caveolin-like binding motif (CLBM) (_X_XXXX_XX_, where _ represents an aromatic amino acid residue) on AIIR1’s signalling and trafficking properties. They mutated this binding motif by replacing each aromatic residue with alanine, a small, sterically unintrusive molecule. The mutated receptor was shown to be four-fold less effective at activating phospholipase C, indicating that the functional CLBM is required for proper signalling. The authors proposed that the CLBM could be acting as a site for nucleation of proteins involved in the regulation of function of AIIR1 [13]. A similar study by Tomohiro Yamaguchi and colleagues [14] examined the effects of interaction of endothelin type A and B (ETaR & ETbR, respectively) receptors with caveolin-1. It was found that ETbR only interacted with caveolin-1 in the absence of an agonist, or bound to the antagonist RES-701-1. When endothelin-1 or another antagonist BQ788 were added, the complex dissociated. ETaR, however, bound to caveolin-1 irrespective of whether a ligand was bound or not. Additionally, overexpression of caveolin-1 dramatically increased the amount of ETbR localised to caveolae, while addition of endothelin-1 reduced caveolar localisation. Disruption of caveolae by filipin reduced the effect of endothelin-1 agonism on ERK1/2 phosphorylation [14].
Taken together, the concepts and experimental observations described here provide the framework for a GPCR signalling platform that is heavily based around spatial compartmentalization of a network of interacting components. This has been called a ‘receptosome’ in some publications [4], and it is quite possible that these receptosomes are the functional unit of plasma membrane receptor signalling, like a cell is the functional unit of a tissue. Agnati et al. in a review publication called “On the molecular basis of the receptor mosaic hypothesis of the engram,” suggest that signalling units such as receptosomes form mosaics on the pre- and post-synaptic membranes of synapses, and that these mosaics are the computational entity that actually decodes the neurochemical messages. They move on to suggest that the arrangement of these mosaics of receptosomes could form ‘supramolecular networks’ that store information about the previous activity patterns of the synapse. While it is important to note that not all GPCR related signalling machinery is congregated into lipid rafts, it is likely that the receptosome theory applies to the majority of GPCR signalling, principally because compartmentalisation of signalling proteins makes such good sense kinetically.
Figure 1. Shows the 5HT2c receptor and its interacting proteins forming a receptosome.
Figure 1. Proteins that interact with the 5HT2c receptor: an example of a synaptic receptosome.
Componentry and organisation of the receptosome: GPCRs & GPCR Interacting Proteins
Having identified general features of the receptosome and the logic behind organising signalling machinery this way, this section of the review will discuss the main groups of proteins that are likely to be part of the receptosome network and their functions.
I will address the questions of what these proteins are, where and how they interact with each other, and why these interactions are fundamental to GPCR / heptahelical transmembrane receptor signalling. Several protein-protein interaction domains such as PDZ, SH2 and SH3 domains are common in receptosome proteins, and the roles of these domains in protein interactions will be highlighted where appropriate.
GRKs:
It has been observed experimentally for decades that GPCRs undergo desensitisation and subsequent internalisation under repeated agonist stimulation. The first event in this process is usually phosphorylation of the receptor. There are at least two methods which the cell uses to perform this function: phosphorylation by second messenger activated kinases such as Protein Kinase A (PKA), and phosphorylation by non-second messenger dependent G-protein coupled Receptor Kinases, which are specific to activated GPCRs. The former is an example of ‘heterologous’ desensitisation, whereby agonism of one receptor can result in activation of PKA and subsequent phosphorylation and desensitisation of another receptor [1]. This effect is usually weak and short lasting, and the phospho-residue is not a target for arrestin binding. GRK mediated desensitisation is termed ‘homologous,’ since agonism of a receptor induces desensitisation of the same receptor [1]. Unlike phosphoresidues created by PKA or other second messenger activated kinases, GRK will phosphorylate different sites, and these phosphoresidues are targets for arrestin binding. Once arrestin is bound, various events occur, most importantly blockade of GPCR access to G-proteins. It is not the actual phosphorylation event that desensitises the GPCR in this case, but arrestin binding. GRK mediated phosphorylation was first discovered in the context of rhodopsin-dependent visual signalling, and later, beta2adrenergic receptor signalling. Since then, it has been established that the majority of GPCRs are desensitised in this way [15].
Arrestins:
Arrestins have been known to interact with GPCRs for a relatively long time, and their function was not particularly difficult to fit into the classical GPCR signalling dogma. Arrestins bind, as described previously, to GRK-phosphorylated GPCRs, and for a long time it was thought that arrestins were only involved in desensitisation and internalisation [4]. While some GPCRs internalise independent of arrestins, the usual scenario involves the bound arrestin attaching to clathrin – one of the major components of endocytotic machinery. Follow clathrin binding, arrestin acts as a scaffold protein and nucleates several other proteins to form the multiprotein complex that will effect receptor endocytosis. Other proteins identified in this complex include: AP2 (assembly particle-2,) a large (340kDa) protein that binds to the globular domain at the end of each clathrin heavy chain and function to promote clathrin triskelion formation and oligomerization into the cage that coats membrane invaginations to form clathrin coated pits [16], NSF (n-ethylamide sensitive factor), an intracellular trafficking protein, ARF6, an ADP-ribosylation factor and its exchange factor ARNO, which together regulate vesicle budding. Additionally, arrestin-2 can act as signalling intermediates, and attaches to multiple of the tyrosine kinase c-Src, including its SH3 and SH1 domains to activate MAPK pathways [11]. Arrestin-2 also has an ERK1/2 phosphorylation dependent regulation site at residue Ser-412 which modulates c-SRC and GPCR binding [11]. There are three beta-arrestin subtypes: arrestin-1, 2 & 3, each with different binding specificities and signalling functions. Arrestin-1 is specific to the visual GPCR rhodopsin, while arrestin-2 has a much wider GPCR specificity, and while arrestin-1 is dimeric, arrestin-2 exists as a monomer in solution [11]. These varying characteristics of arrestins add to the overall specificity and complexity of GPCR signalling.
Figure 2. shows a schematic of GPCR activation, arrestin mediated desensitisation, internalisation, and degradation or resensitisation. Some arrestin interacting proteins are shown.
Figure 2. The roles of arrestins in GPCR desensitisation, internalisation, degradation and resensitisation [4]
RGS’s:
Regulators of G-protein Signalling, or ‘RGS’ proteins play a crucial role in regulating the function of G-proteins, and therefore in the signalling efficacy of the receptor system. There are, like many other GIPs examined in this review, a number of members of the RGS family, each with their own G-protein subunit specificity. The mammalian RGS family comprises several subfamilies, termed: Rz, R4, RA, R12 and R7, which are classified on the basis of structural and sequence homology. RGS proteins contain an RGS box which allows them to interact with activated G-alpha subunits and increase the rate that the G-alpha subunit hydrolyses GTP to GDP. The net effect of this interaction is to reduce the time that the G-alpha subunit actively signals to other proteins. As well as their characteristic RGS box domains, RGS proteins often contain other protein-protein interaction domains such as PDZ domains on RGS-R12 members. These protein-protein interaction domains make RGS proteins the target of considerable research efforts because of the implication that RGS proteins can, like arrestins, act as signalling intermediates as well as their role in regulating G-alpha signalling. For example, RGS proteins containing the RBD domain have been shown to initiate MAPK signalling [17]
The roles of RGS proteins in mu-opioid receptor signalling have been quite extensively studied, and examples of these studies are demonstrative of general RGS function. RGS2 and RGS3, for example, increase opioid agonist potency, while RGS4 and RGS16 reduce the potency of agonists. It is not known whether RGS2 and RGS3 actively reduce the rate of G-alpha GTP hydrolysis, or whether their effect is mediated by one of their other protein-protein interactions [18]. Experiments in which RGS9-2 is knocked out show increased response to Mu-opioid agonists and impaired desensitisation [18]. Garzon et al. in a 2004 paper [17] demonstrated that morphine “alters the selective association between mu-opioid receptors and specific RGS proteins in mouse periaqueductal gray matter,” and in pull-down assays, they noted that certain proteins increased or decreased in their association with mu opioid receptors. It is possible that this may be something to do with morphine altering the receptor conformation and subsequently the network of proteins, particularly G-proteins, which interact with it. RGS proteins have selectivity for specific G-proteins, and if the group of G-proteins present in the network changes, then the group of RGS proteins present would also be likely to change.
Homer:
Several metabotropic glutamate receptors, such as mGluR1a and mGluR5a & b, along with Ca++ permeable IP3 receptors, ryanodine receptors, TRP channels, dynamin II and shank proteins contain the sequence (-PPxxFR-) which is a binding sequence for ‘Homer’ proteins. These Homer proteins, contain an enabled “VASP homology-like” domain which binds to the Homer binding sequence, and a C-terminal coiled coil domain which allows them to homo- and heteromultimerise. It is Homer’s coiled coil interactions that allow the above proteins to form large complexes. A complex containing mGluR’s, Homer proteins, TRP channels, ryanodine receptors and P/Q Ca++ channels, according to Bockaert et al. [4] would “constitute an ideal machinery for intracellular Ca++ release.” Homer proteins act primarily as scaffolding for protein complex formation, but experiments inhibiting Homer activity by using Homer1a, which lacks the coiled coil domain and acts as a dominant negative form of Homer, have shown that Homer has regulatory effects on mGluR signalling and ryanodine channel function [4].
GPCR-GPCR Interactions:
GIPs are essential to the function of a receptosome, but it is important to note that GPCRs do not just interact with non-GPCR proteins: the recent literature [19] documents many experiments exploring GPCR-GPCR interactions, including homo- and hetero-oligomerisation. Oligomerisation of GPCRs can affect many properties of GPCR function or sometimes only one or none, depending on the particular oligomer. For example, heteromeric complexes of B2ARs and delta or kappa opioid receptors doesn’t affect the pharmacology of either the adrenergic or opioidergic units, but profoundly alters the trafficking properties of the heteromer [20]. Again a familiar concept can be found in the nature of GPCR oligomerisation: signalling specificity and complexity are increased by a further level.
RAMPs:
The discovery of Receptor Activity Modifying Proteins or RAMPs revolutionised the field of GPCR signalling, because it demonstrated that not only could GIPs fine tune GPCR signalling, modulate trafficking and activate secondary signalling pathways, they could also turn a receptor into a completely different entity, with a totally different cognate ligand. There are three members of the RAMP family that have been identified so far, designated RAMP1, 2 & 3. As an example of RAMP function, RAMP1 can bind to the calcitonin receptor-like receptor (CL,) and convert it into a ‘high affinity calcitonin gene-related peptide receptor.” Alternatively, interaction of CL with RAMP2 or 3 produces an adrenomedullin receptor. RAMPs are now known to interact with the majority of GPCR Class II receptors, and are regulated heavily by physiological and pathophysiological processes. For example, RAMP2 and adrenomedullin mRNA are elevated in models of cardiac hypotrophy, and during pregnancy, progesterone causes upregulation of all three RAMPs. It is also thought that many of the orphan ligands which have been found (i.e. no receptor has been identified,) are ligands to GPCR-RAMP complexes, when the GPCR already has a cognate ligand in its non-RAMP complexed state [20].
‘Magic tail’ interacting proteins & PDZ Domains:
The C-terminal tail of many GPCRs contains a PDZ ligand, to which proteins with the PDZ protein-protein interaction domain common to many proteins involved in receptor signalling bind [21]. The protein PICK1 (Protein Interacting with C-kinase 1) is one such protein: by binding to the PDZ ligand motif of mGluR7a, PICK1 induces clustering of these receptors at presynaptic terminals. It is also proposed that PICK1 interaction with mGluR7a receptors mediates coupling to Ca++ channels [22].
The protein NHERF (Na+/H+ exchange regulatory factor) also contains PDZ domains, and controls the signalling properties of parathyroid hormone receptor, which binds to NHERF by its PDZ ligand. PDZ-ligand mediated coupling of NHERF to B2AR’s and kappa opioid receptors is likely the way that NHERF controls the Na+/H+ exchanger protein [21,23].
PDZ-ligand interactions between the PDZ domain of cyclic nucleotide Ras guanine exchange factor and the PDZ ligand on B2A enables B2AR to activate Ras and the associated MAPK pathway [21].
PDZ-ligand interactions also play an important role in receptosome scaffolding. The protein Shank spatially organises receptors and ion channels and provides interaction between receptors and the cytoskeleton [21]
Methods: Proteomics and experimental determination of protein-protein interactions
The dawn of the new millennium has seen the development of high throughput methods which generate vast amounts of novel data on protein-protein interactions. A number of different methods have been used to generate this data, all with their respective advantages and limitations. Use of different methods, or even variations of conditions within methods, can produce conflicting data sets. Appropriate synthesis of data sets produced by different methods is required to produce a coherent ‘map’ of interactions.
Researchers studying protein-protein interactions have a large toolbox of methodologies at their disposal. These include complementation assays, mass spectrometric approaches, chip based methods and bioinformatic analysis. The nature of the data produced by various methods differs: data can be qualitative or quantitative, and can describe pairwise interactions between two interaction partners, or can describe grouped interactions within a complex. The inability of most methods used to investigate large scale interactomes to measure quantitative information about interactions such as kinetics raises an important question: what exactly constitutes an interaction? Some biologically relevant interactions may occur on short timescales with very low affinity, but might be considered irrelevant by, or be below the sensitivity of such methods [25].
One issue that is particularly applicable to the study of protein-protein interactions occurring in receptosomes, and particularly interactions with membrane bound proteins, is the difficulty of resolving hydrophobic proteins in 2D gels [25,26]. Modern two-dimensional liquid chromatographic techniques have been able to provide improved resolution of hydrophobic proteins but preparation of pull-down assay experiments still proves difficult with membrane proteins [26]. One of the other important problems in GPCR and GIP interaction analysis is the low cellular concentrations of these proteins. If the experimenter chooses to overexpress a particular GPCR or GIP, they run the risk of ruining the stoichiometry of the interaction network [26].
One particularly effective method of analysing protein complexes is called SEAM, which stands for Sequential Epitope tagging, Affinity tagging and Mass spectrometry. In this process, a protein is selected and epitope tags such as Myc are fused to one of its termini. It is then overexpressed in a cellular system of choice and the cell lysate is run through an affinity column where anti-Myc antibodies are attached to the beads. A second mixture of proteins is then run through the column, and those proteins that are bound to the epitope tagged proteins are resolved by 2D liquid chromatography and fed into a mass spectrometer for identification. Subsequently, one of the MS identified proteins is then Myc tagged and the procedure run again. In this way, it is possible to build up complexes of proteins [27].
Obviously a vast amount of information has to be gathered regarding protein-protein interactions between components of signalling machinery before any kind of mathematical modelling process can be applied to these networks. First, it is necessary to determine the stoichiometry of each complex, and the precise interactions of each protein with each other. It would also be exceptionally useful to have the crystal or NMR structures of each protein involved. Additionally, having identified the qualitative aspects of the system, quantitative biophysical data would be needed concerning the strength and kinetics of interactions. This task will be a massive undertaking, but eventually researchers will be able to build these mathematical models of GPCR signalling and incorporate them into pre-existing models of human brain function, such as the Blue Brain project that is being run on IBM’s Blue Gene supercomputer (http://bluebrainproject.epfl.ch/). Once GPCR and GIP interactions can be comprehensively modelled, the potential for drug design targeted to, and therapeutic intervention of these systems will be unprescendented.
REFERENCES:
1.) Rang, Dale, Ritter, Moore “Pharmacology,” 5th Edition, Elsevier Sciences Ltd. 2003.
2.) Ostrom RS. “New determinants of receptor-effector coupling: trafficking and compartmentation in membrane microdomains.” Mol Pharmacol. 2002 Mar;61(3):473-6.
3.) Bockaert J, Roussignol G, Becamel C, Gavarini S, Joubert L, Dumuis A, Fagni L, Marin P. “GPCR-interacting proteins (GIPs): nature and functions.”
Biochem Soc Trans. 2004 Nov;32(Pt 5):851-5.
4.) Bockaert J, Fagni L, Dumuis A, Marin P. “GPCR interacting proteins (GIP).
Pharmacol Ther. 2004 Sep;103(3):203-21.”
5.) Ostrom RS, Insel PA. “The evolving role of lipid rafts and caveolae in G protein-coupled receptor signaling: implications for molecular pharmacology.”
Br J Pharmacol. 2004 Sep;143(2):235-45. Epub 2004 Aug 2.
6.) Kukkonen JP. “Regulation of receptor-coupling to (multiple) G proteins. A challenge for basic research and drug discovery.” Receptors Channels. 2004;10(5-6):167-83.
7.) Patel MB, Patel CN, Rajashekara V, Yoburn BC “Opioid agonists differentially regulate mu opioid receptors and trafficking proteins in vivo.” Mol Pharmacol. 2002 Dec;62(6):1464-70
8.) Hermans E. “Biochemical and pharmacological control of the multiplicity of coupling at G-protein-coupled receptors.” Pharmacol Ther. 2003 Jul;99(1):25-44.
9.) No reference: personal theory
10.) Hall RA, Premont RT, Lefkowitz RJ. “Heptahelical receptor signalling: beyond the G protein paradigm” J Cell Biol. 1999 May 31;145(5):927-32.
11.) Milano SK, Pace HC, Kim YM, Brenner C, Benovic JL. “Scaffolding functions of arrestin-2 revealed by crystal structure and mutagenesis.” Biochemistry. 2002 Mar 12;41(10):3321-8.
12.) Burgueno J, Enrich C, Canela EI, Mallol K, Lluis C, Franco R, Ciruela F “Metabotropic glutamate type 1 alpha receptor localizes in low-density caveolin-rich plasma membrane fractions” J Neurochem. 2003 Aug;86(4):785-91.
13.) Leclerc PC, Auger-Messier M, Lanctot PM, Escher E, Leduc R, Guillemette G. “A polyaromatic caveolin-binding-like motif in the cytoplasmic tail of the type 1 receptor for angiotensin II plays an important role in receptor trafficking and signaling.” Endocrinology. 2002 Dec;143(12):4702-10.
14.) Yamaguchi T, Murata Y, Fujiyoshi Y, Doi T. “Regulated interaction of endothelin B receptor with caveolin-1.” Eur J Biochem. 2003 Apr;270(8):1816-27.
15.) Inglese J, Freedman NJ, Koch WJ, Lefkowitz RJ “Structure and mechanism of G protein-coupled receptor kinases” J Biol Chem. 1993 Nov 15;268(32):23735-8.
16.) Lodish, Berk, Zipursky, Matsudaira, Baltimore, Darnell “Molecular Cell Biology” 4th Edition, p734-5 W.H. Freedman & Company 2000
17.) Garzon J, Rodriguez-Munoz M, Sanchez-Blazquez P. “Morphine alters the selective association between mu-opioid receptors and specific RGS proteins in mouse periaqueductal gray matter.” Neuropharmacology. 2005 May;48(6):853-68.
18.) Garzo n, J., Rodrıguez-Diaz, M., Lopez-Fando, A., Sanchez-
Blazquez, P. “RGS9 proteins facilitate acute tolerance
to mu-opioid effects.” European Journal of Neuroscience 13,
801-811. 2001
19.) Maggio R, Novi F, Scarselli M, Corsini GU “The impact of G-protein-coupled receptor hetero-oligomerization on function and pharmacology” FEBS J. 2005 Jun;272(12):2939-46.
20.) Jordan BA, Trapaidze N, Gomes I, Nivarthi R, Devi LA “Oligomerization of opioid receptors with B2-adrenergic receptors: a role in trafficking and mitogen-activated protein kinase activation” PNAS Jan 2, 2002 vol.98(1)343-348
21.) Morfis M, Christopoulos A, Sexton PM. “RAMPs: 5 years on, where to now?” Trends Pharmacol Sci. 2003 Nov;24(11):596-601.
22.) Boudin H, Doan A, Yia I, Shigemoto R, Huganir RL, Worley P, Craig AM. “Presynaptic clustering of mGluR7a requires the PICK1 PDZ domain binding site.” Neuron. 2000 Nov;28(2):485-97.
23.) Hall RA, Premont RT, Chow CW, Blitzer JT, Pitcher JA, Claing A, Stoffel RH, Barak LS, Shenolikar S, Weinman EJ, Grinstein S, Lefkowitz RJ. “The beta2-adrenergic receptor interacts with the Na+/H+-exchanger regulatory factor to control Na+/H+ exchange.” Nature. 1998 Apr 9;392(6676):626-30.
24.) Doronin S, Malbon CC. “Functional proteomics of G-protein-coupled receptors: analysis of large scale signalling devices” Pharmaceutical News (2002) 9, 347-355.
25.) von Mering C, Krause R, Snel B, Cornell M, Oliver SG, Fields S, Bork P. “Comparative assessment of large-scale data sets of protein-protein interactions.”
Nature. 2002 May 23;417(6887):399-403. Epub 2002 May 8.
26.) Wu CC, Yates Jr. III “The application of mass spectrometry to membrane proteomics” Nat Biotechnol. 2003 Mar;21(3):262-7.
27.) Deshaies RJ, Seol JH, McDonald WH, Cope G, Lyapina S, Shevchenko A, Shevchenko A, Verma R, Yates JR 3rd. “Charting the protein complexome in yeast by mass spectrometry.” Mol Cell Proteomics. 2002 Jan;1(1):3-10.
Hey guys, here's the latest addition to my series of articles and reviews this year. I've also posted it into the Textbook thread, so if anyone is keen to rip it up, head over there. For general comments and discussion, reply to this thread.
I can't post the .doc file here because its too big, so that means, unfortunately, no figures. If anyone so desires the figures, drop me a line on [email protected] - thanks.
“G-protein coupled receptors: part of a network of signalling machinery”
The section on “G-protein coupled receptors” (GPCRs) in the popular undergraduate level text book “Pharmacology” by Rang, Dale, Ritter & Moore [1], presents a relatively one-dimensional view of GPCR signalling. A neat diagram sums up the events that occur on binding of a ligand to its cognate GPCR: ligand binding attracts a GTPase (‘g’-) protein to the receptor, and the GDP bound to the g-protein is swapped for a GTP. The now-activated subunits of this heterotrimeric g-protein then dissociate and activate another molecule in the signalling cascade, such as adenylyl cyclase, which subsequently produces ‘second messenger’ molecules. These second messengers are responsible for activation of other downstream effectors, and the signal that began with a ligand binding to the GPCR is propagated.
In contrast, even a cursory glance at any recent review [2,3,4] on the subject of GPCR signalling will reveal to the reader that the textbook model of GPCR signalling is woefully inadequate. Rather, these reviews present the concept that a whole network of interacting proteins and biomolecules nucleated around scaffolding proteins are involved in what was once thought to be a relatively simple, linear transduction process. These protein networks are involved with fine-tuning and regulation of every facet of GPCR function. The ‘receptosome’ concept, that whole networks of molecules are spatially compartmentalised into plasma membrane microdomains such as caveolae and other lipid rafts is integral to all recent models of how GPCR signalling is effected.
In this review I will discuss how each receptosome exists as a self-contained, functional signalling unit, and the importance of spatial compartmentalisation of GPCR signalling machinery. Important experimental observations which lead to the invalidation of older GPCR signalling dogma and forced reconsideration of the whole signalling paradigm will be included. Attention will be paid to the roles that particular groups of proteins play in receptosomes, and the modality of their interactions with other receptosome proteins. Finally I will discuss some of the methodology that is currently being used to determine these interactions and their importance to aspects of GPCR signalling.
The old GPCR signalling dogma:
The older “1-dimensional” (meaning that signal transduction follows a defined, stepwise path, as opposed to three dimensional networks of interactions) model of GPCR signalling has been superceded by models like the ‘receptosome’ model described in this review. Essentially, too many contradictions of and paradoxes in the old model arose from experimental observations. Some of the observations that forced the creation of a newer more comprehensive model are detailed below.
Old dogma states that a specific ligand binds to its cognate GPCR, which then undergoes a conformational change, such that its cytosolic domain develops high affinity for one particular G-protein family and subtype, which it recruits. The activated GPCR activates the recruited G-protein by functioning as a guanine nucleotide exchange factor, exchanging G-protein bound GDP for GTP. The activated G-protein then splits into its alpha and beta-gamma subunit components, which activate secondary targets. G-alpha usually modulates the activity of a second messenger producing enzyme, such as activation of phospholipase C in the case of Gq-alpha, or activation of adenylyl cyclase by Gs-alpha. The second messenger activates second messenger dependent kinases which proliferate the signalling cascade. Additionally, there is room in the one-dimensional model for GRK mediated phosphorylation and arrestin mediated desensitisation, which is why although these molecules were discovered relatively early on, they did not push for creation of a new model. [1,3,4]
Organisation of GPCR signalling machinery
One of the fundamental concepts of biochemistry is that the proteins that comprise the majority of cellular machinery interact with each other as huge networks of multiprotein complexes, with specific chemical affinities determining the strengths of these interactions [1,5]. Thus when a ligand binds to its cognate GPCR, a conformational change is induced in the GPCR which creates a chemical site for which specific G-protein(s) have high affinity. Additionally, according to the laws of mass action, the magnitude and rate of chemical interactions and reactions are heavily dependent on the concentrations of the reactants. This raises a fundamental problem with the classical one-dimensional theory of GPCR signalling as can be found in most generic pharmacology textbooks: if one looks at the average concentration of each protein component involved in GPCR signalling, it is far too low to possibly account for the rapidity of the biochemical response to receptor agonism [5]. Kinetics of the protein-protein interactions required to form a signalling cascade must therefore be simply too unfavourable for any agonist directed response to occur if these proteins were randomly or even uniformly distributed across the plasma membrane or cytosol. This appears to be a massive flaw in the old model of GPCR signalling.
Several more flaws can be identified when the old model is compared with recent observations about the characteristics of GPCR signalling.
The old dogma of “1 g-protein couples to one GPCR” has been comprehensively disproved – in fact GPCRs more often than not couple to more than one G-protein [6]. This has significant ramifications for the signalling pathways activated by a particular GPCR. One particularly illuminating example of this observation is that the thyrotropin receptor is able to couple to all four major G-protein families [6]. Other experiments have shown that the majority of GPCRs have at least some affinity for each type of G-protein: therefore the preference for activation of a particular G-protein subtype actually lies on a continuum. In light of this concept, it is possible to infer that it is possible to describe GPCR interactions with particular G-proteins statistically: for example, a particular GPCR may interact with G-protein X 90% of the time, G-protein Y 9.99% of the time and G-protein Z a biochemically negligible 0.01% of the time. These statistics would be based on the chemistry of the interaction sites on the GPCR and G-proteins X, Y & Z. Interaction of the GPCR with G-protein X is obviously the most thermodynamically favourable binding interaction in a mixture of the four proteins at equilibrium.
A further observation that can not be integrated with the old model of GPCR signalling is that agonism by different ligands induces GPCRs to have different affinities for particular G-proteins. The paper “Opioid agonists differentially regulate Mu-opioid receptors and trafficking proteins in vivo” [7] is a good example of how different agonists can induce different biochemical responses in the cell. It is likely that the mechanism for this involves the two agonists used, morphine and etorphine, inducing different receptor conformations, and therefore recruiting different groups of G-proteins to the Mu-opioid receptor.
It is also possible, however, that this effect is not actually mediated by G-proteins at all, and involves direct interaction of other non-G-protein signalling machinery at the Mu receptor. Models have been suggested in which particular receptors may have a number of different conformations which they can assume, and different agonist ligands thermodynamically stabilise particular conformations, which each have a set of G-proteins they activate to different extents. This is a discrete model – there are a defined number of conformations that a receptor can take, and the potency of the agonist to induce that conformation and therefore the overall activity of the drug at the receptor, depends on the degree of thermodynamic stabilisation of that conformation [8]. Alternately, it is possible to imagine a continuous model, where each agonist induces an individual receptor conformation, which alters the G-protein coupling of the receptor and therefore the properties of the signal induced by that agonist. It is even possible to amalgamate these two theories, and conceptualise a model where each different agonist does continuously induce a different receptor conformation, but there are ‘peaks’ in agonist affinity and efficacy which correspond to stabilisation of particular, discrete conformations [9]. Needless to say, a “one GPCR binds to one G-protein” model is completely unable to account for any of these ideas.
Another observation that has forced progression from the older GPCR signalling dogma is that G-protein coupling is not necessarily required for biochemical responses to receptor agonism [10]. Following receptor activation and subsequent G-protein activation, the GPCR is often phosphorylated by a G-protein coupled receptor kinase (GRK,) [10] (or sometimes a by second messenger dependent kinases,) [11] and it is this chemical modification that creates a binding site on the GPCR for a group of proteins called arrestins, which attache to the GPCR and blocks any further coupling to G-proteins, in effect causing the cessation of G-protein mediated signalling [10]. It has been shown, however, that Beta-arrestin may act as a scaffolding molecule and serve to recruit other non-G-protein related signalling machinery. Experiments have shown that arrestin-2 can recruit the tyrosine kinase Src by binding to its SH3 domain, and can also activate MAP kinase pathways. Other experiments showed binding of JNK3 and ASK1, which is a JNK kinase kinase. Thus GPCRs can activate MAPK and tyrosine kinase pathways via their interaction with arrestin proteins [11]. Recent studies have shown an interaction between activated beta2adrenoreceptors (B2AR) and Src which is increased by overexpression of beta-arrestin. Additionally, inhibition of beta-arrestin binding to either B2AR or Src attenuates B2AR mediated activation of MAPK. [10]
Beta-arrestin has also been implicated in regulation of receptor trafficking and endocytosis by its interaction with the heavy chain of clathrin and the clathrin adaptor protein AP2 [10]. These observations of arrestin molecules as scaffolds that nucleate assembly of non-G-protein mediated signalling processes add further detract from old GPCR signalling dogma.
Described above is a series of experimental observations that obligatorily invalidate the one-dimensional model of GPCR signalling, while at the same time building the concept of GPCR signalling as involving a whole network of interacting proteins, with some acting as nodes and scaffolds onto which other proteins nucleate, while others are involved in fine tuning and regulation of the signalling machinery, and still others involved in the trafficking and regulation of the receptors themselves. The most important feature of a newer model of GPCR signalling would have to incorporate the principles of biochemical kinetics and concepts such as collision theory. If all the protein components required for GPCR signalling were to be randomly or even uniformly distributed throughout the cytosol and plasma membrane, the observed rapid response of GPCRs to agonism could not possibly occur. Thus, a new model must include a spatial dimension. The components must be spatially organised such that the biochemistry is actually possible. In answer to this requirement, the literature is packed with reviews and papers documenting the existence of membrane microdomains, or lipid rafts, such as caveolae, in which many of the signalling components and receptors are often congregated.
Caveolae are small (50-100nm) invaginations in the plasma membrane of cells, and are considered to be a subclass of lipid rafts. The lipid composition of caveolae includes characteristically high levels of cholesterol and sphingolipids, along with caveolins, a group of proteins which comprises three isoforms: cavelolin-1, caveolin-2 and caveolin-3. It is generally accepted that caveolae will form if a cell expresses caveolin-1, or in the case of striated muscle myocytes, caveolin-3. Thus, while the plasma membranes of most or all cells contain lipid rafts, only some cells contain caveolae. A 2003 paper in the Journal of Neurochemistry [12] gives a good example of a GPCR being localised to caveolin membrane fractions, and shows “molecular and functional association of mGluR1a receptors with caveolins.” The study demonstrates that agonistic activation of mGluR1a receptors increased ERK phosphorylation in low density caveolin enriched membrane fractions, but not in high density membrane fractions containing no caveolins. Also mentioned in the study was the observation that mGluR1 heterodimerizes with adenosine A1 receptor and calcium sensing receptor; all three of these proteins localise to caveolin rich membrane microdomains [12].
Another example of the role of localisation of receptors and signalling machinery to caveolae is the comparison of Beta1-adrenoreceptor (B1AR) and B2AR signalling in cardiac myocytes. B2ARs activate adenylyl cyclase 6 (AC6) with a lower efficacy than B1ARs, and it appears that this is due to rapid translocation of B2ARs out of caveolae and into clathrin coated pits after receptor activation. AC6 is localised strictly to caveolae, and as such when the B2AR is translocated, it can no longer physically contact AC6 to activate it [5].
A concept that is integral to the model of spatial compartmentalisation of signalling proteins into regions such as caveolae and other lipid rafts is the selective expression of certain isoforms of G-proteins and second messenger synthesising enzymes such as adenylyl cyclase (AC) to particular types of raft, or not to any raft at all. There are nine AC subtypes, but not all of them localise to lipid rafts [5]. Therefore, the fact that many different subtypes exist of G-proteins, second messenger synthesising enzymes and other signalling proteins such as Regulators of G-protein Signalling (RGS,) is a way of increasing the diversity of plasma membrane domains and microdomans.
Many proteins which associate with caveolin proteins contain a caveolin- or caveolin-like- binding motif [13]. Le Clerc et al. published a study in 2002 in the journal Endocrinology examining the effect of Angiotensin II Receptor Type 1’s (AIIR1) caveolin-like binding motif (CLBM) (_X_XXXX_XX_, where _ represents an aromatic amino acid residue) on AIIR1’s signalling and trafficking properties. They mutated this binding motif by replacing each aromatic residue with alanine, a small, sterically unintrusive molecule. The mutated receptor was shown to be four-fold less effective at activating phospholipase C, indicating that the functional CLBM is required for proper signalling. The authors proposed that the CLBM could be acting as a site for nucleation of proteins involved in the regulation of function of AIIR1 [13]. A similar study by Tomohiro Yamaguchi and colleagues [14] examined the effects of interaction of endothelin type A and B (ETaR & ETbR, respectively) receptors with caveolin-1. It was found that ETbR only interacted with caveolin-1 in the absence of an agonist, or bound to the antagonist RES-701-1. When endothelin-1 or another antagonist BQ788 were added, the complex dissociated. ETaR, however, bound to caveolin-1 irrespective of whether a ligand was bound or not. Additionally, overexpression of caveolin-1 dramatically increased the amount of ETbR localised to caveolae, while addition of endothelin-1 reduced caveolar localisation. Disruption of caveolae by filipin reduced the effect of endothelin-1 agonism on ERK1/2 phosphorylation [14].
Taken together, the concepts and experimental observations described here provide the framework for a GPCR signalling platform that is heavily based around spatial compartmentalization of a network of interacting components. This has been called a ‘receptosome’ in some publications [4], and it is quite possible that these receptosomes are the functional unit of plasma membrane receptor signalling, like a cell is the functional unit of a tissue. Agnati et al. in a review publication called “On the molecular basis of the receptor mosaic hypothesis of the engram,” suggest that signalling units such as receptosomes form mosaics on the pre- and post-synaptic membranes of synapses, and that these mosaics are the computational entity that actually decodes the neurochemical messages. They move on to suggest that the arrangement of these mosaics of receptosomes could form ‘supramolecular networks’ that store information about the previous activity patterns of the synapse. While it is important to note that not all GPCR related signalling machinery is congregated into lipid rafts, it is likely that the receptosome theory applies to the majority of GPCR signalling, principally because compartmentalisation of signalling proteins makes such good sense kinetically.
Figure 1. Shows the 5HT2c receptor and its interacting proteins forming a receptosome.
Figure 1. Proteins that interact with the 5HT2c receptor: an example of a synaptic receptosome.
Componentry and organisation of the receptosome: GPCRs & GPCR Interacting Proteins
Having identified general features of the receptosome and the logic behind organising signalling machinery this way, this section of the review will discuss the main groups of proteins that are likely to be part of the receptosome network and their functions.
I will address the questions of what these proteins are, where and how they interact with each other, and why these interactions are fundamental to GPCR / heptahelical transmembrane receptor signalling. Several protein-protein interaction domains such as PDZ, SH2 and SH3 domains are common in receptosome proteins, and the roles of these domains in protein interactions will be highlighted where appropriate.
GRKs:
It has been observed experimentally for decades that GPCRs undergo desensitisation and subsequent internalisation under repeated agonist stimulation. The first event in this process is usually phosphorylation of the receptor. There are at least two methods which the cell uses to perform this function: phosphorylation by second messenger activated kinases such as Protein Kinase A (PKA), and phosphorylation by non-second messenger dependent G-protein coupled Receptor Kinases, which are specific to activated GPCRs. The former is an example of ‘heterologous’ desensitisation, whereby agonism of one receptor can result in activation of PKA and subsequent phosphorylation and desensitisation of another receptor [1]. This effect is usually weak and short lasting, and the phospho-residue is not a target for arrestin binding. GRK mediated desensitisation is termed ‘homologous,’ since agonism of a receptor induces desensitisation of the same receptor [1]. Unlike phosphoresidues created by PKA or other second messenger activated kinases, GRK will phosphorylate different sites, and these phosphoresidues are targets for arrestin binding. Once arrestin is bound, various events occur, most importantly blockade of GPCR access to G-proteins. It is not the actual phosphorylation event that desensitises the GPCR in this case, but arrestin binding. GRK mediated phosphorylation was first discovered in the context of rhodopsin-dependent visual signalling, and later, beta2adrenergic receptor signalling. Since then, it has been established that the majority of GPCRs are desensitised in this way [15].
Arrestins:
Arrestins have been known to interact with GPCRs for a relatively long time, and their function was not particularly difficult to fit into the classical GPCR signalling dogma. Arrestins bind, as described previously, to GRK-phosphorylated GPCRs, and for a long time it was thought that arrestins were only involved in desensitisation and internalisation [4]. While some GPCRs internalise independent of arrestins, the usual scenario involves the bound arrestin attaching to clathrin – one of the major components of endocytotic machinery. Follow clathrin binding, arrestin acts as a scaffold protein and nucleates several other proteins to form the multiprotein complex that will effect receptor endocytosis. Other proteins identified in this complex include: AP2 (assembly particle-2,) a large (340kDa) protein that binds to the globular domain at the end of each clathrin heavy chain and function to promote clathrin triskelion formation and oligomerization into the cage that coats membrane invaginations to form clathrin coated pits [16], NSF (n-ethylamide sensitive factor), an intracellular trafficking protein, ARF6, an ADP-ribosylation factor and its exchange factor ARNO, which together regulate vesicle budding. Additionally, arrestin-2 can act as signalling intermediates, and attaches to multiple of the tyrosine kinase c-Src, including its SH3 and SH1 domains to activate MAPK pathways [11]. Arrestin-2 also has an ERK1/2 phosphorylation dependent regulation site at residue Ser-412 which modulates c-SRC and GPCR binding [11]. There are three beta-arrestin subtypes: arrestin-1, 2 & 3, each with different binding specificities and signalling functions. Arrestin-1 is specific to the visual GPCR rhodopsin, while arrestin-2 has a much wider GPCR specificity, and while arrestin-1 is dimeric, arrestin-2 exists as a monomer in solution [11]. These varying characteristics of arrestins add to the overall specificity and complexity of GPCR signalling.
Figure 2. shows a schematic of GPCR activation, arrestin mediated desensitisation, internalisation, and degradation or resensitisation. Some arrestin interacting proteins are shown.
Figure 2. The roles of arrestins in GPCR desensitisation, internalisation, degradation and resensitisation [4]
RGS’s:
Regulators of G-protein Signalling, or ‘RGS’ proteins play a crucial role in regulating the function of G-proteins, and therefore in the signalling efficacy of the receptor system. There are, like many other GIPs examined in this review, a number of members of the RGS family, each with their own G-protein subunit specificity. The mammalian RGS family comprises several subfamilies, termed: Rz, R4, RA, R12 and R7, which are classified on the basis of structural and sequence homology. RGS proteins contain an RGS box which allows them to interact with activated G-alpha subunits and increase the rate that the G-alpha subunit hydrolyses GTP to GDP. The net effect of this interaction is to reduce the time that the G-alpha subunit actively signals to other proteins. As well as their characteristic RGS box domains, RGS proteins often contain other protein-protein interaction domains such as PDZ domains on RGS-R12 members. These protein-protein interaction domains make RGS proteins the target of considerable research efforts because of the implication that RGS proteins can, like arrestins, act as signalling intermediates as well as their role in regulating G-alpha signalling. For example, RGS proteins containing the RBD domain have been shown to initiate MAPK signalling [17]
The roles of RGS proteins in mu-opioid receptor signalling have been quite extensively studied, and examples of these studies are demonstrative of general RGS function. RGS2 and RGS3, for example, increase opioid agonist potency, while RGS4 and RGS16 reduce the potency of agonists. It is not known whether RGS2 and RGS3 actively reduce the rate of G-alpha GTP hydrolysis, or whether their effect is mediated by one of their other protein-protein interactions [18]. Experiments in which RGS9-2 is knocked out show increased response to Mu-opioid agonists and impaired desensitisation [18]. Garzon et al. in a 2004 paper [17] demonstrated that morphine “alters the selective association between mu-opioid receptors and specific RGS proteins in mouse periaqueductal gray matter,” and in pull-down assays, they noted that certain proteins increased or decreased in their association with mu opioid receptors. It is possible that this may be something to do with morphine altering the receptor conformation and subsequently the network of proteins, particularly G-proteins, which interact with it. RGS proteins have selectivity for specific G-proteins, and if the group of G-proteins present in the network changes, then the group of RGS proteins present would also be likely to change.
Homer:
Several metabotropic glutamate receptors, such as mGluR1a and mGluR5a & b, along with Ca++ permeable IP3 receptors, ryanodine receptors, TRP channels, dynamin II and shank proteins contain the sequence (-PPxxFR-) which is a binding sequence for ‘Homer’ proteins. These Homer proteins, contain an enabled “VASP homology-like” domain which binds to the Homer binding sequence, and a C-terminal coiled coil domain which allows them to homo- and heteromultimerise. It is Homer’s coiled coil interactions that allow the above proteins to form large complexes. A complex containing mGluR’s, Homer proteins, TRP channels, ryanodine receptors and P/Q Ca++ channels, according to Bockaert et al. [4] would “constitute an ideal machinery for intracellular Ca++ release.” Homer proteins act primarily as scaffolding for protein complex formation, but experiments inhibiting Homer activity by using Homer1a, which lacks the coiled coil domain and acts as a dominant negative form of Homer, have shown that Homer has regulatory effects on mGluR signalling and ryanodine channel function [4].
GPCR-GPCR Interactions:
GIPs are essential to the function of a receptosome, but it is important to note that GPCRs do not just interact with non-GPCR proteins: the recent literature [19] documents many experiments exploring GPCR-GPCR interactions, including homo- and hetero-oligomerisation. Oligomerisation of GPCRs can affect many properties of GPCR function or sometimes only one or none, depending on the particular oligomer. For example, heteromeric complexes of B2ARs and delta or kappa opioid receptors doesn’t affect the pharmacology of either the adrenergic or opioidergic units, but profoundly alters the trafficking properties of the heteromer [20]. Again a familiar concept can be found in the nature of GPCR oligomerisation: signalling specificity and complexity are increased by a further level.
RAMPs:
The discovery of Receptor Activity Modifying Proteins or RAMPs revolutionised the field of GPCR signalling, because it demonstrated that not only could GIPs fine tune GPCR signalling, modulate trafficking and activate secondary signalling pathways, they could also turn a receptor into a completely different entity, with a totally different cognate ligand. There are three members of the RAMP family that have been identified so far, designated RAMP1, 2 & 3. As an example of RAMP function, RAMP1 can bind to the calcitonin receptor-like receptor (CL,) and convert it into a ‘high affinity calcitonin gene-related peptide receptor.” Alternatively, interaction of CL with RAMP2 or 3 produces an adrenomedullin receptor. RAMPs are now known to interact with the majority of GPCR Class II receptors, and are regulated heavily by physiological and pathophysiological processes. For example, RAMP2 and adrenomedullin mRNA are elevated in models of cardiac hypotrophy, and during pregnancy, progesterone causes upregulation of all three RAMPs. It is also thought that many of the orphan ligands which have been found (i.e. no receptor has been identified,) are ligands to GPCR-RAMP complexes, when the GPCR already has a cognate ligand in its non-RAMP complexed state [20].
‘Magic tail’ interacting proteins & PDZ Domains:
The C-terminal tail of many GPCRs contains a PDZ ligand, to which proteins with the PDZ protein-protein interaction domain common to many proteins involved in receptor signalling bind [21]. The protein PICK1 (Protein Interacting with C-kinase 1) is one such protein: by binding to the PDZ ligand motif of mGluR7a, PICK1 induces clustering of these receptors at presynaptic terminals. It is also proposed that PICK1 interaction with mGluR7a receptors mediates coupling to Ca++ channels [22].
The protein NHERF (Na+/H+ exchange regulatory factor) also contains PDZ domains, and controls the signalling properties of parathyroid hormone receptor, which binds to NHERF by its PDZ ligand. PDZ-ligand mediated coupling of NHERF to B2AR’s and kappa opioid receptors is likely the way that NHERF controls the Na+/H+ exchanger protein [21,23].
PDZ-ligand interactions between the PDZ domain of cyclic nucleotide Ras guanine exchange factor and the PDZ ligand on B2A enables B2AR to activate Ras and the associated MAPK pathway [21].
PDZ-ligand interactions also play an important role in receptosome scaffolding. The protein Shank spatially organises receptors and ion channels and provides interaction between receptors and the cytoskeleton [21]
Methods: Proteomics and experimental determination of protein-protein interactions
The dawn of the new millennium has seen the development of high throughput methods which generate vast amounts of novel data on protein-protein interactions. A number of different methods have been used to generate this data, all with their respective advantages and limitations. Use of different methods, or even variations of conditions within methods, can produce conflicting data sets. Appropriate synthesis of data sets produced by different methods is required to produce a coherent ‘map’ of interactions.
Researchers studying protein-protein interactions have a large toolbox of methodologies at their disposal. These include complementation assays, mass spectrometric approaches, chip based methods and bioinformatic analysis. The nature of the data produced by various methods differs: data can be qualitative or quantitative, and can describe pairwise interactions between two interaction partners, or can describe grouped interactions within a complex. The inability of most methods used to investigate large scale interactomes to measure quantitative information about interactions such as kinetics raises an important question: what exactly constitutes an interaction? Some biologically relevant interactions may occur on short timescales with very low affinity, but might be considered irrelevant by, or be below the sensitivity of such methods [25].
One issue that is particularly applicable to the study of protein-protein interactions occurring in receptosomes, and particularly interactions with membrane bound proteins, is the difficulty of resolving hydrophobic proteins in 2D gels [25,26]. Modern two-dimensional liquid chromatographic techniques have been able to provide improved resolution of hydrophobic proteins but preparation of pull-down assay experiments still proves difficult with membrane proteins [26]. One of the other important problems in GPCR and GIP interaction analysis is the low cellular concentrations of these proteins. If the experimenter chooses to overexpress a particular GPCR or GIP, they run the risk of ruining the stoichiometry of the interaction network [26].
One particularly effective method of analysing protein complexes is called SEAM, which stands for Sequential Epitope tagging, Affinity tagging and Mass spectrometry. In this process, a protein is selected and epitope tags such as Myc are fused to one of its termini. It is then overexpressed in a cellular system of choice and the cell lysate is run through an affinity column where anti-Myc antibodies are attached to the beads. A second mixture of proteins is then run through the column, and those proteins that are bound to the epitope tagged proteins are resolved by 2D liquid chromatography and fed into a mass spectrometer for identification. Subsequently, one of the MS identified proteins is then Myc tagged and the procedure run again. In this way, it is possible to build up complexes of proteins [27].
Obviously a vast amount of information has to be gathered regarding protein-protein interactions between components of signalling machinery before any kind of mathematical modelling process can be applied to these networks. First, it is necessary to determine the stoichiometry of each complex, and the precise interactions of each protein with each other. It would also be exceptionally useful to have the crystal or NMR structures of each protein involved. Additionally, having identified the qualitative aspects of the system, quantitative biophysical data would be needed concerning the strength and kinetics of interactions. This task will be a massive undertaking, but eventually researchers will be able to build these mathematical models of GPCR signalling and incorporate them into pre-existing models of human brain function, such as the Blue Brain project that is being run on IBM’s Blue Gene supercomputer (http://bluebrainproject.epfl.ch/). Once GPCR and GIP interactions can be comprehensively modelled, the potential for drug design targeted to, and therapeutic intervention of these systems will be unprescendented.
REFERENCES:
1.) Rang, Dale, Ritter, Moore “Pharmacology,” 5th Edition, Elsevier Sciences Ltd. 2003.
2.) Ostrom RS. “New determinants of receptor-effector coupling: trafficking and compartmentation in membrane microdomains.” Mol Pharmacol. 2002 Mar;61(3):473-6.
3.) Bockaert J, Roussignol G, Becamel C, Gavarini S, Joubert L, Dumuis A, Fagni L, Marin P. “GPCR-interacting proteins (GIPs): nature and functions.”
Biochem Soc Trans. 2004 Nov;32(Pt 5):851-5.
4.) Bockaert J, Fagni L, Dumuis A, Marin P. “GPCR interacting proteins (GIP).
Pharmacol Ther. 2004 Sep;103(3):203-21.”
5.) Ostrom RS, Insel PA. “The evolving role of lipid rafts and caveolae in G protein-coupled receptor signaling: implications for molecular pharmacology.”
Br J Pharmacol. 2004 Sep;143(2):235-45. Epub 2004 Aug 2.
6.) Kukkonen JP. “Regulation of receptor-coupling to (multiple) G proteins. A challenge for basic research and drug discovery.” Receptors Channels. 2004;10(5-6):167-83.
7.) Patel MB, Patel CN, Rajashekara V, Yoburn BC “Opioid agonists differentially regulate mu opioid receptors and trafficking proteins in vivo.” Mol Pharmacol. 2002 Dec;62(6):1464-70
8.) Hermans E. “Biochemical and pharmacological control of the multiplicity of coupling at G-protein-coupled receptors.” Pharmacol Ther. 2003 Jul;99(1):25-44.
9.) No reference: personal theory
10.) Hall RA, Premont RT, Lefkowitz RJ. “Heptahelical receptor signalling: beyond the G protein paradigm” J Cell Biol. 1999 May 31;145(5):927-32.
11.) Milano SK, Pace HC, Kim YM, Brenner C, Benovic JL. “Scaffolding functions of arrestin-2 revealed by crystal structure and mutagenesis.” Biochemistry. 2002 Mar 12;41(10):3321-8.
12.) Burgueno J, Enrich C, Canela EI, Mallol K, Lluis C, Franco R, Ciruela F “Metabotropic glutamate type 1 alpha receptor localizes in low-density caveolin-rich plasma membrane fractions” J Neurochem. 2003 Aug;86(4):785-91.
13.) Leclerc PC, Auger-Messier M, Lanctot PM, Escher E, Leduc R, Guillemette G. “A polyaromatic caveolin-binding-like motif in the cytoplasmic tail of the type 1 receptor for angiotensin II plays an important role in receptor trafficking and signaling.” Endocrinology. 2002 Dec;143(12):4702-10.
14.) Yamaguchi T, Murata Y, Fujiyoshi Y, Doi T. “Regulated interaction of endothelin B receptor with caveolin-1.” Eur J Biochem. 2003 Apr;270(8):1816-27.
15.) Inglese J, Freedman NJ, Koch WJ, Lefkowitz RJ “Structure and mechanism of G protein-coupled receptor kinases” J Biol Chem. 1993 Nov 15;268(32):23735-8.
16.) Lodish, Berk, Zipursky, Matsudaira, Baltimore, Darnell “Molecular Cell Biology” 4th Edition, p734-5 W.H. Freedman & Company 2000
17.) Garzon J, Rodriguez-Munoz M, Sanchez-Blazquez P. “Morphine alters the selective association between mu-opioid receptors and specific RGS proteins in mouse periaqueductal gray matter.” Neuropharmacology. 2005 May;48(6):853-68.
18.) Garzo n, J., Rodrıguez-Diaz, M., Lopez-Fando, A., Sanchez-
Blazquez, P. “RGS9 proteins facilitate acute tolerance
to mu-opioid effects.” European Journal of Neuroscience 13,
801-811. 2001
19.) Maggio R, Novi F, Scarselli M, Corsini GU “The impact of G-protein-coupled receptor hetero-oligomerization on function and pharmacology” FEBS J. 2005 Jun;272(12):2939-46.
20.) Jordan BA, Trapaidze N, Gomes I, Nivarthi R, Devi LA “Oligomerization of opioid receptors with B2-adrenergic receptors: a role in trafficking and mitogen-activated protein kinase activation” PNAS Jan 2, 2002 vol.98(1)343-348
21.) Morfis M, Christopoulos A, Sexton PM. “RAMPs: 5 years on, where to now?” Trends Pharmacol Sci. 2003 Nov;24(11):596-601.
22.) Boudin H, Doan A, Yia I, Shigemoto R, Huganir RL, Worley P, Craig AM. “Presynaptic clustering of mGluR7a requires the PICK1 PDZ domain binding site.” Neuron. 2000 Nov;28(2):485-97.
23.) Hall RA, Premont RT, Chow CW, Blitzer JT, Pitcher JA, Claing A, Stoffel RH, Barak LS, Shenolikar S, Weinman EJ, Grinstein S, Lefkowitz RJ. “The beta2-adrenergic receptor interacts with the Na+/H+-exchanger regulatory factor to control Na+/H+ exchange.” Nature. 1998 Apr 9;392(6676):626-30.
24.) Doronin S, Malbon CC. “Functional proteomics of G-protein-coupled receptors: analysis of large scale signalling devices” Pharmaceutical News (2002) 9, 347-355.
25.) von Mering C, Krause R, Snel B, Cornell M, Oliver SG, Fields S, Bork P. “Comparative assessment of large-scale data sets of protein-protein interactions.”
Nature. 2002 May 23;417(6887):399-403. Epub 2002 May 8.
26.) Wu CC, Yates Jr. III “The application of mass spectrometry to membrane proteomics” Nat Biotechnol. 2003 Mar;21(3):262-7.
27.) Deshaies RJ, Seol JH, McDonald WH, Cope G, Lyapina S, Shevchenko A, Shevchenko A, Verma R, Yates JR 3rd. “Charting the protein complexome in yeast by mass spectrometry.” Mol Cell Proteomics. 2002 Jan;1(1):3-10.
