Abstract
The molecular basis for the known intramembrane receptor/receptor interactions among G protein-coupled receptors was postulated to be heteromerization based on receptor subtype-specific interactions between different types of receptor homomers. The discovery of GABAB heterodimers started this field rapidly followed by the discovery of heteromerization among isoreceptors of several G protein-coupled receptors such as δ/κ opioid receptors. Heteromerization was also discovered among distinct types of G protein-coupled receptors with the initial demonstration of somatostatin SSTR5/dopamine D2 and adenosine A1/dopamine D1 heteromeric receptor complexes. The functional meaning of these heteromeric complexes is to achieve direct or indirect (via adapter proteins) intramembrane receptor/receptor interactions in the complex. G protein-coupled receptors also form heteromeric complexes involving direct interactions with ion channel receptors, the best example being the GABAA/dopamine D5 receptor heteromerization, as well as with receptor tyrosine kinases and with receptor activity modulating proteins. As an example, adenosine, dopamine, and glutamate metabotropic receptor/receptor interactions in the striatopallidal GABA neurons are discussed as well as their relevance for Parkinson's disease, schizophrenia, and drug dependence. The heterodimer is only one type of heteromeric complex, and the evidence is equally compatible with the existence of higher order heteromeric complexes, where also adapter proteins such as homer proteins and scaffolding proteins can exist. These complexes may assist in the process of linking G protein-coupled receptors and ion channel receptors together in a receptor mosaic that may have special integrative value and may constitute the molecular basis for some forms of learning and memory.
I. Experimental Evidence on Protein/Protein Interactions Involving G Protein-Coupled Receptors in the Central Nervous System
A. Early Indications for Intramembrane Receptor/Receptor Interactions Involving G Protein-Coupled Receptors
Emerging evidence shows that G protein-coupled receptors (GPCR2) can form homo- and heteromers (Bouvier, 2001; Marshall, 2001). It all began in 1979-1980 in search of an explanation of where all the recently discovered neuropeptides in the brain could integrate their messages with those of classical transmitters such as the monoamines. Luigi F. Agnati and Kjell Fuxe postulated that an intramembrane interaction between neuropeptide and monoamine receptors could be involved. The first observations were published in 1980 (Agnati et al., 1980), showing that substance P could modulate the high-affinity serotonin binding sites in spinal cord membrane preparations using biochemical binding techniques (also, see Agnati et al., 1983b). The same year, an interesting paper was published by Maggi et al. (1980) showing that the β adrenergic receptor agonist isoproterenol could increase α2 adrenergic receptor binding in cortical slices, supporting the concept of intramembrane receptor/receptor interactions of GPCR, in this case among isoreceptors. Subsequently, in 1981 the existence of cholecystokinin (CCK) receptor/dopamine D2 receptor interactions using biochemical binding techniques was indicated since CCK-8 could modulate the dopamine D2 receptor antagonist and agonist binding sites in striatal membrane preparations (Fuxe et al., 1981, 1983b; Agnati et al., 1983a,b, 1985). Further evidence for receptor/receptor interactions came in 1982 from Lundberg, Bartfai, and colleagues (Lundberg et al., 1982) and from Zarbin and colleagues (Zarbin et al., 1982). Using the same type of approach, a large number of papers were published in 1983 that suggested the existence of intramembrane receptor/receptor interactions between different GPCR (Fuxe et al., 1983b; Agnati et al., 1984; Fuxe and Agnati, 1985). Those included neurotensin (NT receptor)/D2 (Agnati et al., 1983c; Nemeroff, 1986; Von Euler and Fuxe, 1987; Von Euler, 1991), CCKB/serotonin 5-HT2 (Agnati et al., 1983a, 1985), vasoactive intestinal peptide (VIP) receptor/serotonin 5-HT1 (Rostene et al., 1983a,b), neuropeptide Y (NPY) receptor/α2 adrenergic (Agnati et al., 1983d) and neurokinin NK1/5-HT1 receptor/receptor interactions (Agnati et al., 1983e). Subsequently in the 1980s, indications for glutamate receptor/D2 receptor interactions (Fuxe et al., 1984) were obtained in striatal membrane preparations after earlier observations had indicated the existence of interactions at the membrane level among glutamate receptor subtypes (Fuxe et al., 1983c). This early research led to the following postulation in the opening address of Fuxe and Agnati at the International Wenner-Gren Symposium on receptor/receptor interactions in 1986 (Fuxe and Agnati, 1987) “... we will find out that some sophisticated elaborations are performed at the membrane level, via interactions within and among different classes of macromolecules (such as receptors, ion pumps, ion channels)” (also, see Agnati et al., 1988). In 1988, evidence for galanin (Gal) receptor/serotonin 5-HT1A receptor interactions in limbic cortical membranes (Fuxe et al., 1988a), as well as for angiotensin II receptor (AT1)/α2 adrenergic receptor interactions (Fuxe et al., 1988b) in the medulla oblongata membrane preparations were obtained. In the early 1990s, adenosine A2A receptor/D2 receptor interactions (Ferré et al., 1991d, 1993; Ferré and Fuxe, 1992) were demonstrated in striatal membrane preparations.
Thus, not only neuropeptide and monoamine receptors were involved in intramembrane receptor/receptor interactions but also certain types of glutamate and adenosine receptors (Agnati et al., 1986, 1990, 1993; Härfstrand et al., 1988; Tanganelli et al., 1989, 1990, 1993; Von Euler et al., 1989; Fuxe et al., 1990a-c, 1991, 1992a-c; Ferré et al., 1992, 1993b; Fior et al., 1993; Yang et al., 1994b). These results were all obtained at the recognition site of the receptors, using saturation and competition binding experiments. The modulation of binding could be shown as changes in KD and Bmax values (saturation analysis) and as KL, KH, and RH values (competition analysis) allowing a determination of modulation of the high- versus the low-affinity states of the receptor. An indication of an effect on the G protein-coupling and thus on the efficacy of the modulated receptor could be obtained by studying how, e.g., the modulator could control the GTP-induced disappearance of the high-affinity state of the receptor (reduction of the RH values). This would imply a G protein activation with formation of Gα-GTP and βγ dimers associated with a cross-regulation of the GPCR with a disappearance of the high-affinity state of the receptor.
In this period, the above work was extended to show multiple receptor/receptor interactions. Thus, evidence was obtained for a dopamine D1 receptor involvement in the CCKB receptor/D2 receptor interaction. Coactivation of D1 and D2 receptors led to an enhancement of the affinity of D2 receptor agonist sites by CCK-8, instead of a reduced affinity of D2 receptor agonist sites observed without D1 receptor stimulation (Li et al., 1994a). These results are in line with the findings of Seeman et al. (1989) suggesting reciprocal interactions between D1 and D2 receptors in striatal homogenates. Thus, there may exist striatal nerve cell populations where intramembrane multiple CCKB receptor/D1 receptor/D2 receptor interactions can take place (Agnati et al., 1982).
In the early 1990s, evidence was also obtained that striatal NT receptors involved in the G protein-independent antagonistic regulation of striatal D2 receptors (Von Euler et al., 1991) may represent a novel type of a high-affinity NT receptor. This was suggested in view of the rank order of potency found among COOH-terminal NT fragments, neuromedin N, and NT in this response versus the rank order of potency found at the cloned high-affinity NT receptors (NT1 receptors) (Li et al., 1993a,b). These effects were stronger in striatal sections (Li et al., 1994b), and recently, the NT-induced reduction of D2 receptor affinity in striatal sections has been found to be blocked by a NT1-like antagonist (Diaz-Cabiale et al., 2002a).
In 1993 (Zoli et al., 1993), the hypothesis was introduced that the molecular mechanism for these large numbers of intramembrane receptor/receptor interactions among GPCR could be the formation of heteromeric complexes, the simplest being a heterodimer. This concept was based on the indication at the time that dimerization upon agonist activation may be a general phenomenon essential for receptor activation (Hollenberg, 1991), the best example being the dimerization of tyrosine kinase receptors (Schlessinger 1988, 2000; Helldin 1995). Thus, it was assumed that GPCR exist mainly as homodimers that interact with other types of homodimers to form heterodimers. The relative proportions of homo- and heterodimers would be determined by the concentrations of the two transmitters, the density of the two receptors and their distribution patterns, and the unique features of each receptor/receptor interaction (Zoli et al., 1993). In fact, early evidence obtained on purified GPCR by, e.g., Venter and Fraser (1983) and Conn et al. (1982) indicated that the functional GPCR were in a dimeric state. The same year that our review article appeared, the first evidence was published that GPCR can exist as dimers (Ng et al., 1993). Thus, the 5-HT1B receptor in Sf9 cells was in the immunoblot analysis shown to exist as dimers and monomers. Finally, it should be mentioned that in 1982 we had introduced the receptor mosaic hypothesis of learning and memory based on the formation of membrane receptor clusters and thus of high order oligomeric receptor complexes (Agnati et al., 2002). It was postulated (Agnati et al., 1982; Zoli et al., 1993) that the formation and/or stabilization of the heteromeric complexes of GPCR could be enhanced by associated (adapter) proteins especially in the synaptic membranes. It must be noted that most of the GPCR are located in extrasynaptic membranes and therefore the potential heteromeric complexes discussed above may also be reached by volume transmission (VT) signals (Agnati and Fuxe, 2000).
Recent experimental data exist confirming this hypothesis, and they are described in this review (see Table 1). Despite all this novel experimental evidence, there are many questions regarding the molecular mechanism of receptor heteromerization and among them the mapping of the residues involved in the interaction in the case of direct interactions and the identification of scaffolding and adapter proteins in the case of indirect interactions. It should be noted that both types of interaction, direct and indirect, are likely to occur.
B. G Protein-Coupled Receptors Homo- and Heteromerization
1. Homomerization of G Protein-Coupled Receptors.
A number of new approaches made it possible to obtain convincing evidence for the existence of homomers of many types of GPCR. Those include complementary chimeras, coimmunoprecipitation with differentially epitope-tagged receptors, the use of sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), often in combination with covalent cross-linking, and finally biophysical methods, namely bioluminescence resonance energy transfer (BRET) and fluorescence resonance energy transfer (FRET).
In the mid 1980s, evidence for dimerization was obtained with experiments with photo-affinity labeling, radiation inactivation, cross-linking, and hydrodynamic analysis (Fraser and Venter, 1982; Avissar et al., 1983; Herberg et al., 1984; Peterson et al., 1986). However, it was the work of Maggio, Wess, and colleagues (Maggio et al., 1993) that offered strong indications for the existence of dimers, with the transcomplementation results obtained using cholinergic M3 muscarinic receptor/α2 adrenergic receptor chimeras. In fact, when a chimera consisting of 5TM regions of one receptor and two of the other was expressed, there was a lack of ligand binding and of function that was recovered after coexpression of the two types of chimeras. This finding had a major impact, and the interpretation that the functional transcomplementation was caused by intermolecular interactions leading to the formation of a dimeric complex was early on accepted (Monnot et al., 1996). The early work in the mid-1970s demonstrating negative cooperativity in β-adrenergic receptors, opening up the possibility of dimer formation should also be mentioned (Limbird et al., 1975; Limbird and Lefkowitz, 1976). More recently, structural implications for V2 vasopressin receptor oligomerization have been given by Schultz et al. (2000) from functional reconstitution studies.
The SDS-PAGE approach provided one of the first demonstrations of GPCR dimers, namely of 5-HT1B, D1, and D2 receptor homomers (Ng et al., 1993, 1994a,b, 1996; George et al., 1998; Zawarynski et al., 1998). These observations are based on the fact that several homomers are resistant to the denaturation properties of SDS. Thus, it was possible early on for George, O'Dowd, and colleagues to observe upon SDS-PAGE that the 5-HT1B, D1, and D2 receptors expressed in cell lines formed molecular species not only corresponding to monomers but also to dimers (Ng et al., 1993, 1994a,b; Lee et al., 2000). These dimeric and even higher order oligomeric complexes were not caused by glycosylation of monomers nor to the presence of G proteins (Lee et al., 2000). Treatment with covalent cross-linkers before solubilization increased the proportion of dimeric complexes and facilitated the demonstration of dimers in immunoblots (Hebert et al., 1996; Romano et al., 1996). It should be considered that monomers at least in some cases could represent the disruption of dimers or higher oligomeric complexes (Lee et al., 2000). Also by using the SDS-PAGE strategy, the existence of D2 and A1 receptors homomers was demonstrated, for the first time, in brain tissue, showing their existence in situ and not only in cell lines where artificially high levels of receptors are expressed (Ciruela et al., 1995; Ng et al., 1996).
The coimmunoprecipitation approach was first described in the article by Bouvier and colleagues in 1996 (Hebert et al., 1996) on β2 adrenergic receptors with coexpression of differentially tagged β2 adrenergic receptors. This coimmunoprecipitation was taken as evidence for the existence of a β2 adrenergic receptor dimer. Peptides corresponding to TM6 were found to disrupt the dimerization as well as receptor activation indicating a participation of the hydrophobic forces in the TM6 region in the dimerization interface (Hebert et al., 1996; Ng et al., 1996). The focus on the TM6 came from the work of Engelman et al. (Lemmon et al., 1992; Lemmon and Engelman, 1994), showing that dimerization is driven by specific interactions among TM α-helices. A number of GPCR homomers have been demonstrated with this approach, e.g., the metabotropic glutamate receptor (mGlu5) homomer, where the disulfide bridges between the large extracellular NH2-terminal domains play a role in the formation of the homomer (Romano et al., 1996; Bouvier, 2001).
The techniques of BRET and FRET imply as close distances as 5 to 10 nM between donor and acceptor for energy transfer. In FRET both the donor and acceptor are fluorescent molecules, whereas for BRET the donor is bioluminescent and the acceptors fluorescent. These techniques have been very valuable in detecting dimers in living cells without the risk of solubilization artifacts (for details on methodology, see Bouvier, 2001). In 2000, the BRET approach could for the first time demonstrate homomerization of β2 adrenergic receptors in living cells (Angers et al., 2000) independently, and at the same time, the FRET procedure revealed the homomerization of the yeast α mating factor in living cells (Overton and Blumer, 2000). In 2001, using the FRET and BRET technologies, constitutive homo-oligomerization could be demonstrated for δ opioid receptors in intact cells (McVey et al., 2001). The same was also found to be true for the thyrotropin-releasing hormone (TRH) receptors using the BRET technique (Kroeger et al., 2001). With FRET technique, the somatostatin receptor subtypes have been shown to assemble, e.g., as homomers (Rocheville et al., 2000b) using differential epitope tagging and fluorescently labeled antibodies against the epitopes.
Together with results from coimmunoprecipitation experiments these results indicate that several, if not many, GPCR undergo constitutive homomerization, i.e., the basal state of the GPCR may be the dimer. It remains to be shown if the existence of constitutive dimers can help explain the constitutive, agonist independent, activity of several GPCR. We have, e.g., recently observed that D2 receptor antagonists do not affect or even reduce the D2 receptor clustering (increased by D2 receptor agonists) in the basal state of CHO cell lines expressing human D2L receptors on their surface. Such an effect of the D2 receptor clustering by a D2 receptor antagonist in the absence of a dopamine receptor agonist could at least in part explain an inverse D2 agonist activity of the D2 antagonist (L. F. Agnati, S. Ferré, R. Franco, and K. Fuxe, unpublished data). Therefore, the action of the agonist at GPCR may sometimes be to produce a conformational change in the basal homomer leading to the development of the active state. Again it should be emphasized that the basal state instead sometimes may be represented by a monomeric or higher oligomeric form of the GPCR.
2. Heteromeric Complexes Involving G Protein-Coupled Receptors.
The existence of heteromers of GPCR was postulated in the 1993 Molecular Neurobiology review article (Zoli et al., 1993) to give a molecular basis for the large amount of evidence we had obtained on the existence of receptor/receptor interactions among GPCR (see above). It was therefore inspiring when the first evidence of the existence of a GPCR heteromer, namely the GABAB heteromer, came in 1998/99 (Jones et al., 1998; Kaupman et al., 1998; White et al., 1998; Kuner et al., 1999; Marshall et al., 1999; Ng et al., 1999). The field of heteromerization has been excellently summarized by Marshall (2001) and by Bouvier (2001). Here some key examples will be discussed together with the evidence suggesting that the heterodimerizaton and hetero-oligomerization may be the main molecular basis for the previously observed receptor/receptor interactions among GPCR (Fuxe and Agnati, 1985, 1987; Agnati et al., 1993; Zoli et al., 1993; Fuxe et al., 1995, 1996, 1998; Ferré et al., 1997). The functional relevance, the pathological implications and the relevance for new drug development (see Agnati et al., 1986, 1990; Fuxe et al., 1989, 1998) of certain selected receptor heteromers are also discussed.
One question that arises out of the recent reports is how to understand the architecture of cross-talk among heteromeric complexes. It is interesting to ask, for instance, how the strongly antagonistic A2A/D2 intramembrane receptor/receptor interaction through its heteromeric complex (Hillion et al., 2002) becomes integrated with the strongly facilitatory SSTR5/D2 intramembrane receptor/receptor interaction through its heteromeric complex (Rocheville et al., 2000a). Do these heteromeric complexes directly interact? Are they part of the same molecular circuits of the same striatal nerve cell or are they independently located in different membrane domains of the same cell or even in different striatal nerve cell populations? These will be important studies to perform in order to understand the integration of transmitter signals in the striatum and in general in the brain. How do the various heteromeric complexes of GPCR interact at membrane level and downstream at intracytoplasmatic level? How are these two levels of interactions integrated? These are fundamental questions to be answered and a possible heuristic frame to tackle these questions is discussed below (Section II.C.).
a. The GABAB Receptor Heterodimer.
The first evidence that two subtypes of GABAB receptors, GABABR1 and GABABR2, undergo heterodimerization and that this process is essential for the cell surface expression of the functional receptor was given in three studies published simultaneously in December 1998 (Kaupmann et al., 1998; Jones et al., 1998; White et al., 1998). This was also demonstrated by Kuner et al. and Ng et al. in January 1999 (see also Gordon et al., 1999; Mitrovic et al., 2000). The physiological relevance of these findings is supported by the demonstration of coimmunoprecipitation in cerebral cortex membranes of GABABR1 (a or b) with GABABR2 proteins, of their colocalization in dendritic spines (Kaupmann et al., 1998) and of substantial degree of coexpression of GABABR1 and GABABR2 mRNA levels in many nerve cell populations (Kuner et al., 1999). Changes also occurred at the recognition site level as a result of the heterodimerization, since the potency of agonists and partial agonists became increased (Kaupmann et al., 1988).
The most dramatic change is, however, that the formed GABAB receptor heteromer, unlike its monomeric components, can become functional and couple to the G protein leading to regulation of the inwardly rectifying K+ channels, the Ca2+ channels, and adenylyl cyclase (Alger and Nicoll, 1979; Bettler et al., 1998). Thus, it seems as if the predominant native GABAB receptor is the GABAB receptor heteromer. Evidence has been presented that the COOH-terminal domain is involved in the formation of this heterodimer by a coiled-coil interaction (Kammerer et al., 1999; Kuner et al., 1999).
In 2000, the important observation was made that the coiled-coil interaction at the COOH-terminal domain blocks a retention motif for the endoplasmic reticulum of the GABABR1 receptor (Margeta-Mitrovic et al., 2000). The masking of this motif allows the heterodimers to travel to the cell surface. However, even if in the mutant GABABR1 the retention motif had been removed and this mutant receptor could be expressed on the cell surface, it still remained functionally inactive underlining a probable crucial role of GABAB receptor heterodimerization in signaling.
These important findings give a clear example of the functional relevance of intramembrane receptor/receptor interactions through heteromerization, namely in receptor trafficking, including receptor maturation and receptor cell surface expression, and in receptor signaling, i.e., in G protein coupling and in increased binding potency of agonists and partial agonists, in line with previous work on receptor/receptor interactions (see Fuxe and Agnati, 1985, 1987; Zoli et al., 1993).
It is well known that GABAB receptors play a distinct role in modulating the neuronal networks, and agonist drugs acting on these receptors appear to have inter alia anticonvulsive and anxiolytic properties. It is therefore of substantial interest that the anticonvulsive compound gabapentin is a selective agonist at the GABABR1a/GABABR2 heteromer (Ng et al., 2001). This is an example of how the molecular composition of the heteromer determines its pharmacological profile and gives rise to a novel GABAB receptor agonist selective for a certain type of GABAB receptor heteromer dependent on the splice variant involved. In this study, these results were correlated with a selective ability to increase postsynaptic GABAB receptor signaling (opening of inwardly rectifying K+ channels) without altering GABA transmission at the presynaptic level (Ng et al., 2001). This novel type of pharmacological selectivity based on unique heteromers may therefore have considerable potential for drug development. It serves to show the pharmacological relevance of intramembrane receptor/receptor interactions that may give rise to novel receptor subtypes with a unique pharmacology based on the composition of the heteromer formed changing the biochemical characteristics of the binding pocket of the receptor.
b. Heteromerization of δ and κ Opioid Receptors and of μ and δ Opioid Receptors.
After the discovery of the GABAB receptor heterodimers the discovery of the δ/κ opioid receptor dimer came next (Jordan and Devi 1999; Jordan et al., 2000) followed by another interesting paper on μ/δ opioid receptor heteromerization by George et al., (2000). In this case, however, the two receptors of the heterodimer were functional on their own and could reach the cell surface without heteromerization with another opioid receptor subtype.
These discoveries were in a way expected, since early work had given indications for the existence of a μ/δ opioid receptor complex (Rothmann et al., 1988; Schoffelmeer et al., 1990). Furthermore, the opiate receptor field had for some time discussed the possibility that heterodimerization among the cloned μ, δ, and κ opioid receptors could explain the existence of more than three opioid receptor subtypes as characterized pharmacologically (Kieffer, 1999). The δ/κ opioid heteromer (Jordan and Devi, 1999; Jordan et al., 2000) was shown to have a unique pharmacology with high affinity for rather unselective ligands but very little affinity for δ- and κ-selective compounds. Nevertheless, the selective δ and κ agonists, when given at the same time, bound synergistically to the heterodimers associated with a synergistic activation of the mitogen-activated protein kinase (MAPK). Thus, a novel subtype of opioid receptor binding pocket may have appeared through this receptor/receptor interaction via δ/κ opioid receptor heterodimerization. Also, the δ and κ heteromer had consequences for the agonist-induced internalization of the δ opioid receptor, which became reduced. Thus, another functional role of this intramembrane receptor/receptor interaction through heteromerization may be the control of receptor internalization.
The μ/δ opioid receptor heteromer, demonstrated by George et al. (2000), also showed changes in the pharmacological properties at the recognition site with reduced affinity for selective agonists and increased affinity for certain enkephalin peptides. Can in fact the μ/δ heteromer be the target of distinct enkephalin peptides? Of special interest was the demonstration that the μ/δ heterodimer, unlike the μ and δ receptors when expressed alone, could become coupled to G proteins resistent to pertussis toxin, like GZ. Thus, the G protein coupling has become markedly altered in the heteromer. In fact, the major function of this intramembrane receptor/receptor interaction based on μ/δ heteromerization may be a change in the selection of G protein coupling involving a conformational change in the G protein interface of the μ/δ receptor heteromer. The other functional change is altered binding properties of the recognition site seen as a novel pattern of ligand binding based on affinity changes; a novel binding pocket seems to have appeared (see Levac et al., 2002).
c. The Serotonin 5-HT1D/5-HT1B Receptor Heteromer.
This is an interesting demonstration of how two receptor subtypes of the type A receptor family (rhodopsin-like GPCR) when coexpressed preferentially form heteromers (Xie et al., 1999) without homomers. In contrast, when the two receptor subtypes were expressed alone homomers were formed. It is of substantial interest that the two receptor subtypes when coexpressed prefer the heteromer, since it indicates that at least in some cases there is a markedly displaced equilibrium between homomers and heteromers, since the heteromer is so clearly preferred. In this example the formation of the heteromer was not associated with a change in the binding pocket, and the functional relevance of this receptor/receptor interaction still remains to be determined. Based on their expression patterns in the brain it seems as if they form different types of complexes in the brain, namely 5-HT1B receptor and 5-HT1D receptor homomers and 5-HT1B/5-HT1D receptor heteromers.
d. The Dopamine D2/D3 Receptor Heteromer.
The D2 and D3 receptors are known to exist as monomers and homomers (Nimchinsky et al., 1997; see Lee et al., 2000). In 2001, the evidence also came that D2 and D3 receptors can form heteromers with unique functional properties (Scarselli et al., 2001). Coimmunoprecipitation experiments using differentially tagged D2 and D3 receptors showed that D2 and D3 receptors in HEK-293 cells can form a heteromeric complex. Furthermore, they were able to demonstrate that heterologous cotransfected dopamine receptor fragments [D2 trunc (TM1-5)/D3 tail (TM6-7); D3 trunc (TM1-5)/D2 tail (TM6-7)] could form functional dopamine receptors that bound dopamine agonists and antagonists with a different pharmacological profile compared with native D2 and D3 receptors, with the highest affinity of all being found with the D3 trunc/D2 tail fragment combination. Thus, split D2/D3 heteromers may be formed through the domain-swapping mechanism as proposed by Gouldson, Reynolds, and colleagues (Gouldson et al., 1998, 2000) based inter alia on demonstrations of functional complementation between chimeras of α2 adrenergic and M3 muscarinic receptors (Maggio et al., 1993).
In agreement with the formation of D2/D3 heteromers in cells, these types of D2 and D3 receptor fragments, when coexpressed with native D2 and D3 receptors, reduced the expression of native dopamine receptors indicating fragment/native receptor complex formation. However, not only are D2 trunc/D3 tail and D3 trunc/D2 tail receptors able to bind ligands but they can also couple in an inhibitory way to adenylyl cyclase and to the same extent as the native D2 receptor. It is also of substantial interest that the D3 receptor under conditions in which it cannot inhibit adenylate cyclase VI (Robinsson and Caron, 1997) can develop such a coupling by cotransfection with D2 receptors. It is therefore possible that in the D2/D3 heteromeric receptor complex formed, the D2 receptor can make possible the G protein coupling of the D3 receptor to adenylyl cyclase VI. Alternatively, the D3 receptor binding pocket upon activation by D3 agonists can, through conformational changes, transfer the D2 pocket of the heteromer into an activated state, leading to Gi activation and adenylyl cyclase VI inhibition. Finally, in the case of adenylyl cyclase V activity, the coexpression of D2 and D3 receptors even resulted in an increased potency of the D3 receptor agonist to inhibit this adenylyl cyclase compared with wild-type D2 receptors when activated by D2 agonists. One function of the D2/D3 heteromeric receptor complex may therefore be to allow a stronger inhibitory coupling of the D3 receptors to adenylyl cyclase.
Colocalization of D2 and D3 receptors has been demonstrated in nerve cells of the basal ganglia (Le Moine and Bloch, 1996; Gurevich and Joyce, 1999) showing that there is the potential to form functional D2/D3 heteromers also in vivo.
e. The Somatostatin SSTR5/SSTR1 Receptor Heteromer.
The study by Rocheville et al. (2000b) gives a fine illustration of intramembrane receptor/receptor interactions and their functional relevance and of the relationship of monomers, homomers, and heteromers among five somatostatin receptor subtypes. Using FRET analysis, the human somatostatin receptor subtype SSTR5 was shown to exist as a monomer in the basal state, which upon agonist activation was converted into a homomer. The data suggested that the agonist-induced dimerization of SSTR5 receptors was essential for signaling. Agonist-induced heteromerization of SSTR5 and SSTR1 receptors could also be demonstrated, which appeared to be subtype specific.
The suggestion was made that the reported high level of basal homomer expression of GPCR could be due to receptor overexpression. The intramembrane SSTR5 and SSTR1 receptor/receptor interactions through heteromerization was shown to have important functional consequences for the participating receptors. Besides the changes in agonist affinity that usually develop upon changes in the oligomeric state, marked alterations in agonist-dependent internalization and an up-regulation of SSTR1 receptors occurred through formation of a heteromer with SSTR5 receptors. Thus, the SSTR1 receptor only underwent agonist-induced internalization as a heterodimer with the SSTR5 receptor. Furthermore, the heteromerization allowed a somatostatin receptor agonist (not binding to SSTR1) to induce up-regulation of agonist binding at the SSTR1 receptor. SSTR5 receptor signaling via its G protein is probably not involved in this response of the heteromer, since the COOH-terminal tail of the SSTR5 receptor had been deleted abolishing adenylyl cyclase regulation. These results are of substantial interest since in this way the desensitization of activated somatostatin receptor subtypes can be compensated for by an up-regulation of the nonactivated somatostatin receptor subtypes, such as SSTR1, made possible through the heteromerization. Thus, another functional meaning of intramembrane receptor/receptor interactions via heteromerization may be the sensitization of one isoreceptor as the other isoreceptor of the heteromer undergoes desensitization.
f. The Somatostatin SSTR5 and Dopamine D2 Heteromeric Receptor Complex.
The discovery of this intramembrane receptor/receptor interaction through hetero-oligomerization gave a novel way to understand the well known somatostatin/dopamine interactions in the brain involved, e.g., in the control of motor activity (Cohn and Cohn, 1975; Havlicek et al., 1976; Kastin et al., 1978; Chneiweiss et al., 1985; Glowinski and Premont, 1985; Martin-Iverson et al., 1986; Leblanc et al., 1988; Izquierdo-Claros et al., 1997; Rodriguez-Sanchez et al., 1997). In this case the hetero-oligomerization involved distinct GPCR and not isoreceptors having the same or similar endogenous ligands. It was by means of photobleaching FRET microscopy that the direct SSTR5/D2 receptor/receptor interaction could be determined, and oligomerization was hardly observed in the basal state but only after treatment with either agonist. Simultaneous treatment with the two types of agonists together had no further action (Rocheville et al., 2000a). It is known that D2 receptor homomers exist in the basal state (Lee et al., 2000), and somatostatin homomers are induced by somatostatin receptor agonists (see above). It remains to be determined whether heterodimers are formed or larger oligomeric complexes in the case of the SSTR5/D2 heteromerization.
The functional meaning of this direct intramembrane receptor/receptor interaction appeared to be severalfold (Rocheville et al., 2000a). First, the binding pocket of SSTR5 receptor was markedly altered, since a 30-fold increase in affinity was found upon D2 receptor agonist activation, whereas D2 receptor antagonists reduced the affinity of SSTR5 receptors for the somatostatin agonist SST-14. Thus, different conformational states (agonist-antagonist states) of the D2 receptor have a substantial modulatory action on the binding pocket of the SSTR5. The interaction at the recognition site level was also reciprocal since the somatostatin agonist enhanced the affinity of the D2 receptor for antagonists. Second, the G protein coupling of the SSTR5 receptor was enhanced by the D2 receptor activation, since the reduction of SSTR5 receptor agonist binding by GTPγS was enhanced by the D2 agonist. Furthermore, the inhibitory responses on cAMP accumulation were significantly enhanced by simultaneous agonist treatments, emphasizing the enhancement of the functional activity through the hetero-oligomer formed and the associated conformational changes induced by agonists in this complex. In fact, these functional changes in the intramembrane receptor/receptor interaction induced by SSTR5 and D2 receptor agonists may explain the increased somatostatin- and D2 receptor-mediated neurotransmission found in vivo after somatostatin or dopamine agonist treatments.
In this study, Rocheville et al. (2000a) also used the mutant Δ 318-SSTR5 receptor, with a COOH-terminal tail deletion. This mutant SSTR5 receptor had been previously shown to bind somatostatin agonists with unchanged affinity, but different than the wild-type SSTR5 receptor, it is not able to produce inhibition of forskolin-induced cAMP accumulation (Hukovic et al., 1998). The interesting finding was that cotransfection with D2 receptors and mutant Δ 318-SSTR5 receptors could restore the somatostatin agonist signaling to the adenylate cyclase provided the D2 recognition site was not blocked by a D2 antagonist. These observations can be explained by the formation of a hetero-oligomer in the CHO-k cells used, in which the SSTR5 binding pocket when activated by agonists can signal via a conformational change in the dopamine D2 binding pocket. This would lead to a coupling of the D2 recognition site to the Gi protein followed by its activation and subsequent inhibition of adenylate cyclase. This conformational change cannot occur when the D2 binding pocket is in an antagonistic binding state. It indicates in fact that the SSTR5 receptor can signal via a conformational change in the D2 receptor similar to that produced by the D2 agonist. In other words, a cross-activation of the D2 receptor can occur in the absence of dopamine by a direct receptor/receptor interaction in the receptor interfaces of the hetero-oligomer. Thus, the activated SST receptor cannot only modulate the activated D2 receptor/Gi protein coupling to adenylate cyclase but also produce a constitutive activity of the D2 receptor when it is not locked into an antagonistic state. From another perspective, it represents an example of how a mutant GPCR can rescue its signaling by activating another receptor coupled to the same type of G protein.
g. The Adenosine A1 and Dopamine D1 Heteromeric Receptor Complex.
The article on A1/D1 heteromers came out a couple of months after the appearance of the SST5/D2 receptor oligomer article and gives another example of heteromerization between distinct GPCR (Gines et al., 2000). A number of morphological and neurochemical observations indicate that adenosine A1 and dopamine D1 receptor/receptor interactions exist in the basal ganglia (Ferré et al., 1994b, 1996a,b, 1997; Fuxe et al., 1998, 2002; Franco et al., 2001) and colocalization of A1 and D1 receptors exists in primary cortical cultures (Gines et al., 2000). The article by Ginés et al. (2000) gives the first evidence that this receptor/receptor interaction can involve A1/D1 heteromeric receptor complexes since such complexes could be demonstrated in cotransfected A1/D1 fibroblast Ltk- cells by means of coimmunoprecipitation. Thus, the previously found A1 receptor-induced uncoupling of the D1 receptor, demonstrated as the A1 receptor-induced disappearance of the high-affinity D1 receptor agonist binding sites in membrane preparations (Ferré et al., 1994b, 1998; Fuxe et al., 1998), could be the result of a physical interaction of the A1 receptor with the D1 receptor in this heteromeric complex, leading to an uncoupling of the D1 receptor to its Gs-like protein in this functionally interacting heteromeric complex. The coimmunoprecipitation analysis demonstrates its existence already in the basal state and the specificity by the failure to show A1/D2 receptor heteromerization in A1/D2 receptor cotransfected fibroblast cells.
However, A1/D1 receptor heteromerization in the cotransfected fibroblast cells was strongly reduced by the D1 receptor agonist treatment, showing an agonist dependence, and simultaneous D1 and A1 receptor agonist treatment blocked this disruption of the heteromeric complex. Thus, like the Rocheville et al. (2000a) study, this study shows how agonists alone or simultaneous treatment lead to conformational changes in their respective binding pockets that are transmitted to the heteromeric interface and results in strengthening or disruption of the complex. In this case, the physical interaction is maintained when the A1 and D1 receptor binding pockets are simultaneously activated by agonists allowing the antagonistic intramembrane receptor/receptor interaction to take place, namely the G protein uncoupling with the disappearance of the high-affinity state of the D1 receptor for agonists. One functional meaning of this intramembrane receptor/receptor interaction is therefore uncoupling of the D1 receptor from Gs protein. This is in sharp contrast to the enhanced functional activity of the SSTR5/D2 receptor oligomers, especially after combined agonist treatment (Rocheville et al., 2000a).
The A1/D1 heteromeric receptor complex may therefore give the molecular basis for the well documented antagonistic A1/D1 receptor/receptor interactions found in the neuronal networks of the brain (Ferré et al., 1997; Fuxe et al., 1998, 2002; Franco et al., 2000, 2001). The A1/D1 receptor heteromerization also appears to have an impact on receptor trafficking (Ginés et al., 2000). Thus, an A1 receptor agonist, after 3 h of exposure, produced a coaggregation of A1 and D1 receptors. On the other hand, aD1 receptor agonist after 3 h of exposure only produced an aggregation of D1 receptor immunoreactivity with a lack of coaggregation in agreement with the ability of the D1 receptor agonist to disrupt A1/D1 receptor heteromerization (see above). The D1 receptor signaling remained unaffected by the formation of D1 receptor or A1/D1 receptor clusters, as seen in terms of an unchanged D1 receptor-stimulated cAMP accumulation and thus with no signs of D1 receptor desensitization. In contrast, combined A1 and D1 receptor agonist treatments under the same conditions did not result in the formation of A1/D1 receptor clusters, but the diffuse A1/D1 receptor colocalization was maintained. Furthermore, now signs of D1 receptor desensitization developed as seen from reductions in D1 receptor-induced increases of cAMP levels. Thus, essential features of D1 receptor desensitization may be a maintained heteromerization with no A1/D1 receptor coaggregates formed after prolonged combined exposure to A1 and D1 receptor agonists with no indications of receptor internalization. It seems possible that the D1 receptor desensitization may be mainly caused by a prolonged allosteric change in the D1 receptor brought about by the A1/D1 receptor/receptor interaction within the heteromeric complex, which could be related to subsequent phosphorylation changes and/or association with β-arrestin-like molecules (Lefkowitz, 2000; McDonald and Lefkowitz, 2001), leading overall to a reduced D1 receptor/Gs coupling. Thus, it may be suggested that the intramembrane A1/D1 receptor/receptor interaction in this heteromeric complex is relevant not only for acute antagonism of D1 receptor signaling but also for a persistent long-term antagonism of D1 signaling to the Gs protein. The details of the composition and stoichiometry of the A1/D1 heteromeric receptor complex is unknown, and A1 and D1 receptors are known to exist as monomers and homomers (Ciruela et al., 1995; Franco et al., 2000; Lee et al., 2000). It is unknown if heteromers are preferred when A1 and D1 receptors are coexpressed in the same cells.
h. The Metabotropic Glutamate mGluR1α and Adenosine A1 Heteromeric Receptor Complex.
There is evidence that the group I metabotropic glutamate receptor mGluR1α and adenosine A1 receptors colocalize in certain types of cerebellar neurons and that functional interactions occur between adenosine and glutamate receptors in the brain (Ferré et al., 1999a; Ciruela et al., 2001a,b). Coimmunoprecipitation experiments on soluble extracts from the rat cerebellum synaptosomes have shown that mGluR1α receptors can coimmunoprecipitate with anti-A1 receptor antibodies. Thus, mGluR1α and A1 receptors may exist as heteromers in certain cell populations of the cerebellum (Ciruela et al., 2001a). Subsequent coimmunoprecipitation studies on transiently cotransfected HEK-293 cells showed that mGluR1α/A1 heteromeric receptor complexes exist also in these cells. The receptor subtype specificity was shown by the failure of the COOH-terminal splice variant, the mGluR1β to immunoprecipitate with the A1 receptor, indicating the involvement of the COOH-terminal tail in the formation of this heteromeric complex. It is unknown how this heteromeric complex relates to mGluR1 receptor dimer, where the interface modulates the glutamate binding site of the extracellular region of the receptor (Kunishima et al., 2000).
The functional role of this heteromerization was especially studied in the HEK-293 cells. After cotransfection of the mGluR1α and A1 receptors, it was found that quisqualic acid substantially enhanced the increase in Ca2+ signaling produced by A1 receptor activation, and the same was true when A1 receptor modulation of mGluR1α receptor function was studied. Thus, it seems as if heteromerization led to the development of synergistic responses in Ca2+ signaling upon simultaneous activation of the receptors within the mGluR1α/A1 receptor heteromer. Also in primary cortical cultures, where a high colocalization was observed at dendritic locations, a synergistic interaction was found in terms of a reduction of NMDA-induced neurotoxicity (Ciruela et al., 2001a,b). A reduced hypoxic neuroprotection has been observed in mice lacking A1 receptor (Johansson et al., 2001). It seems possible that the mGluR1α/A1 receptor heteromerization can take place indirectly, since the COOH terminus interacts with specific targeting proteins. Thus, the protein Homer-1a with an enabled VASP homology 1-like domain binds to the COOH-terminal of mGluR1α receptor. Homer-1c can link together proteins with a proline-rich motif (PPSPF), since it binds to this motif (Tu et al., 1999). It is of substantial interest that this or a similar motif is found in both the COOH-terminal part of mGluR1α and in the COOH-terminal part of the A1 receptor. Thus, the Homer can be an important part of this heteromeric complex as an adapter protein (Xiao et al., 2000; Ciruela et al., 2000). It should also be noticed that the Homer can also link mGluR1α to Shank, since it also contains a similar motif (PPEEF). Shank is a scaffolding multimeric postsynaptic protein that may bring the mGluR1α to the appropriate location on the cell surface and is part of the NMDA receptor-associated PSD-95 complex (Naisbitt et al., 1999; Tu et al., 1999).
i. The Purinergic P2Y1 and Adenosine A1 Heteromeric Receptor Complex.
In 2001, the fascinating finding was made in coimmunoprecipitation experiments on cotransfected HEK-293 T cells that Gi/o-coupled P1 purinoceptor adenosine A1 receptors can form a heteromeric complex with Gq-coupled P2 purinoceptor ATP P2Y1 receptors (Yoshioka et al., 2001), showing less than a 5% homology with each other (Fredholm et al., 1994). The COOH-terminal part of A1 receptors was shown not to be involved in this type of heteromerization. In agreement immunofluorescence studies with confocal imaging showed a marked colocalization of the A1 and P2Y1 receptors. Furthermore, A1 receptors coimmunoprecipitated with P2Y1 receptors indicating that heteromerization between P1 and P2 receptor subtypes could be a rather widespread mechanism for the immediate cross-talk between, e.g., inhibitory A1 receptors and excitatory P2Y1 receptors.
A marked change in the signaling was found in the heteromeric complex. Thus, it became possible for the P2Y1 receptor agonist ADPβS to induce signaling via the Gi/o protein coupled to the A1 receptor, an action blocked by pertussis toxin and the A1 receptor antagonist but not by the P2Y1 receptor antagonist.
These interesting results seemed to be explained by the development of an ability of the ADPβS to reduce 3H-labeled (-)-N6-phenylisopropyladenosine binding of the heteromeric complex in the high-affinity range. It was therefore suggested that the A1 receptor ligand pocket of the heteromer had markedly changed so as to bind the P2Y1R agonist associated with activation of the Gi/o protein. Thus, it was correctly suggested that the heteromeric association produces a P2Y1-like A1 receptor. However, it is also possible to propose another molecular mechanism. Thus, there could exist an A1/P2Y1 receptor/receptor interaction at the recognition site level that changes the pharmacology of both the A1 and P2Y1 receptor binding pockets. The P2Y1 agonist-induced conformational change in the P2Y1 receptor binding pocket now not only leads to an activation of Gq proteins but also to a change in the conformation of the A1 receptor, converting it into an agonist state capable of turning on the Gi/o protein. Such a conformational change may no longer occur when the A1 receptor binding pocket is occupied by the A1 receptor antagonist locking it into an antagonist state. The P2Y1 receptor antagonist may not block the action of the P2Y1 receptor agonist since it seems possible that in the heteromeric complex the antagonist used may not have sufficient affinity for the P2Y1 receptor binding pocket. The possible existence of the P2Y1/A1 heteromeric receptor complex in the brain could explain the demonstration of theophylline-sensitive P2Y receptors (Mendoza-Fernandez et al., 2000). This P2Y1/A1 heteromeric receptor complex is of substantial interest, since it allows the excitatory ATP receptor P2Y1 upon activation to immediately activate in parallel the inhibitory A1R mechanism. In this way, the excitation and increased energy expenditure brought about by the ATP P2Y1 receptor activation begins to be counteracted even at a moment when the extracellular ATP has not been broken down to adenosine, the major ligand for the A1 receptor (Fredholm, 1995a,b; Ferré and Fuxe, 2000; Fredholm et al., 2001).
j. The Adenosine A2A and Dopamine D2 Heteromeric Receptor Complex.
In 1991, the antagonistic A2A/D2 receptor/receptor interaction was demonstrated in striatal membrane preparations with A2A receptors reducing the affinity of D2 receptors, especially in the high-affinity state, for agonists (Ferré et al., 1991d). This offered a novel mechanism for the reported antagonistic adenosine/dopamine interactions found in the brain (Ferré, 1992, 1997; Fuxe et al., 1993, 1998; Lepiku et al., 1997). The molecular mechanism was proposed to be one of heteromerization of A2A/D2 receptors (Zoli et al., 1993). The same antagonistic intramembrane modulation of D2 receptor recognition mechanisms by A2A receptor activation was observed in different cell lines stably cotransfected with different species and isoforms of A2A and D2 receptors. These were a native A2A receptor/human D2L receptor neuroblastoma cell line (Salim et al., 2000), a dog A2A receptor/human D2L receptor Ltk- fibroblast cell line (Snaprud et al., 1994; Yang et al., 1995; Dasgupta et al., 1996a), and a human A2A receptor/rat D2S receptor CHO cell line (Kull et al., 1999). This indicated that the same type of intramembrane A2A/D2 receptor/receptor interaction occurs in all cell types and that both D2L and D2S receptors could undergo the same modulation by A2A receptor activation, at least at the recognition site level. The specificity is demonstrated by the failure of A1 receptor agonists to alter the affinity of the D2 receptors (Ferré et al., 1991d). Hillion et al. (2002) have recently reported, based on coimmunoprecipitation experiments, that heteromerization of human A2A and human D2L receptors exists in the basal state in neuroblastoma SH-SY5Y cells stably transfected with D2L receptors and containing native A2A receptors and in fibroblast Ltk- cells stably transfected with human D2L receptors and transiently transfected with tagged dog A2A receptors. There also exists a high degree of colocalization of D2 and A2A receptors in these cotransfected cells and in primary cultures of rat striatal neurons. The existence of monomers and homomers versus the heteromeric complexes in these cotransfected cells remains to be determined as well as the existence of the simplest heteromeric complex, the A2A/D2 heterodimer. Again, it should be emphasized that this heteromeric complex exists in the absence of exogenous agonists and the specificity of the A2A/D2 receptor heteromerization is shown by the absence of A2A/D1 receptor coimmunoprecipitation in cells expressing D1 receptors and tagged A2A receptors. One functional meaning of this intramembrane receptor/receptor interaction through heteromerization is then to reduce the affinity of the high-affinity agonist state of D2 receptors. Another meaning is to counteract D2 receptor G protein coupling, since the A2A agonist counteracts the GTP analog-induced disappearance of D2 receptors in the high-affinity state (RH) through a site of action independent of the GTP binding site (Ferré et al., 1993b). Thus, the essence of this A2A/D2 receptor heteromerization may be to convert the D2 receptor into a state of strongly reduced functional activity. In line with this view A2A receptor activation counteracts D2 receptor-induced intracellular Ca2+ responses (Salim et al., 2000) and D2 receptor-mediated inhibition of cAMP formation (Kull et al., 1999; Hillion et al., 2002). Based on studies of D1/D2 receptor chimeras (Kozell et al., 1994; Kozell and Neve, 1997; Torvinen et al., 2001) where the 5th and 6th TM domain plus the IC loop 3 of the D2 receptor has been replaced by the corresponding domain of the D1 receptor, it is likely that these D2 receptor domains are part of the A2A/D2 interface, since the affinity of this D1/D2 receptor chimera for dopamine can no longer be modulated by the A2A receptor agonist activation (Torvinen et al., 2001). So far it has not been possible to show a reciprocal A2A/D2 receptor affinity regulation by which the D2 receptor upon activation controls the agonist affinity of the A2A receptor.
The A2A/D2 receptor intramembrane receptor/receptor interaction through heteromerization also has an impact on receptor trafficking (Hillion et al., 2002). Thus, coaggregation of D2 and A2A receptors in the cell membrane of neuroblastoma cells could be demonstrated after A2A or D2 receptor agonist treatment for 3 h by means of immunocytochemistry in combination with confocal image analysis of nonpermeabilized cells. The D2 receptor agonist-induced aggregation of A2A receptors was absent in parental neuroblastoma cells, with a very reduced expression of D2 receptors. The increased development of the A2A/D2 receptor coaggregates on the cell membrane after prolonged A2A or D2 agonist treatment was associated with a failure of the A2A receptor agonist to increase cAMP levels. Thus, the A2A/D2 receptor coaggregates that developed were associated with the appearance of both homologous and D2 receptor-mediated heterologous desensitization of A2A receptors.
In contrast, the D2 receptor did not desensitize under these conditions in terms of inhibition of forskolin-induced cAMP accumulation, possibly related to the substantially higher density of D2 receptors, several of which could represent spare receptors. A high degree of colocalization of A2A and D2 receptors was also found in cultured striatal neurons and also here the A2A agonist or the D2 agonist after a prolonged exposure could induce coaggregates of A2A/D2 receptors.
Evidence for coaggregation followed by cointernalization of A2A/D2 receptors was observed after prolonged cotreatment of the neuroblastoma cells with A2A and D2 receptor agonists. Thus, under these conditions an increase in the uneven distribution of the A2A/D2 receptor immunoreactivity on the membrane was found associated with a marked reduction of the intensity of the immunoreactivity over the A2A/D2 receptor coaggregates. The cointernalization of A2A/D2 receptors could also be directly demonstrated by incubating fluorescent-labeled D2 and A2A receptor antibodies together with A2A and D2 receptor agonists at 4°C for 2 h followed by incubation for 3 h at 37°C, allowing the labeled A2A/D2 receptors to internalize under the influence of the two agonists. Such a synergism with regard to coaggregation and cointernalization of A2A/D2 receptors could not be demonstrated in primary striatal neurons.
It is of substantial interest that in the cAMP accumulation experiments on the neuroblastoma cells, combined agonist treatment was associated with the development of a D2 receptor desensitization as seen from the reduced inhibition by D2 receptor activation of the forskolin-induced cAMP accumulation (Hillion et al., 2002).
Thus, the A2A/D2 receptor heteromerization appears to be involved in the coaggregation, cointernalization, and codesensitization of the A2A and D2 receptors (Hillion et al., 2002). Finally, this intramembrane A2A/D2 receptor/receptor interaction through heteromerization may help understand the cross-tolerance and cross-sensitization found in vivo between dopamine agonists and drugs acting at A2A receptors (Garrett and Holtzman, 1994; Fenu et al., 2000) and also the reduced antiparkinsonian activity and the dyskinesias found after chronic intermittent L-DOPA treatment (Zeng et al., 2000).
k. The Metabotropic Glutamate mGlu5 and Adenosine A2A Heteromeric Receptor Complex.
In 2002, Ferré et al. were able to demonstrate in cotransfected HEK-293 cells a substantial overlap in the distribution of differentially tagged A2A and the group I metabotropic glutamate receptor mGluR5 receptors. Furthermore, in these transiently cotransfected cells (cDNAs for Flag A2A receptor and hemagglutinin-mGluR5 receptor) coimmunoprecipitation experiments showed that the mGluR5 and A2A receptors formed heteromeric complexes that appeared to be selective since such complexes were not formed between mGluR5 and mGluR1β. Importantly, A2A/mGluR5 heteromeric complexes were also demonstrated in rat striatal membrane preparations with coimmunoprecipitation experiments (Ferré et al., 2002).
These findings are of special interest, since in the striatum the A2A and mGluR5 receptors seem to have a similar distribution pattern in the striatopallidal GABA neurons with a perisynaptic localization to asymmetric postsynaptic, putative glutamatergic synapses (see Section II.D.). Furthermore, in behavioral studies A2A and mGluR5 receptor agonists synergize in counteracting D2 receptor-induced turning behavior at supersensitive dopamine receptors (Popoli et al., 2001) and, in biochemical studies, both A2A and mGluR5 receptor agonists reduced the affinity of D2 receptor high-affinity agonist binding sites (Ferré et al., 1999a; Rimondini et al., 1999; Popoli et al., 2001). At the moment, it is unknown whether the A2A receptor and the mGluR5 are linked together via direct heteromerization or whether, e.g., the cytosolic Homer proteins are involved that can bind to the COOH-terminal part of mGluR5 and produce their clustering. The Shank proteins having a scaffolding role with multiple protein/protein interaction motives such as proline-rich regions and PDZ domains could also be involved (Milligan and White, 2001), especially since they participate in linking together the mGluR5 with the NMDA receptors (Sheng and Kim, 2000).
The A2A/mGluR5 heteromeric receptor complex in the cotransfected HEK-293 cells failed to show synergism in Ca2+ mobilization and cAMP accumulation. Nevertheless, a substantial synergism was found after coagonist treatments in terms of MAPK [extracellular signal-regulated kinase 1/2 (ERK 1/2)] and c-fos expression in the cotransfected HEK-293 cells (Ferré et al., 2002). It is presently unknown how signals from the heteromer can bring about this strong synergistic functional interaction that was also observed in the striatum in vivo after combined A2A and mGluR5 agonist treatments (see Section II.D.). Thus, mGluR5 and A2A receptor may mediate glutamate adenosine synergism in case of c-fos expression in the striatum involving the A2A/mGluR5 heteromeric receptor complex. There is a distinct possibility that the combined activation of the two receptors of the A2A/mGluR5 heteromeric complex may lead to reduced desensitization of the mGluR5 by allowing an increased dephosphorylation to develop thanks to increased activation and/or availability of protein phosphatase 2B at the mGluR5 (Cho and Bashir, 2002; Dale et al., 2002).
It seems likely that the demonstrated synergism in rat striatal expression of c-fos has important functional consequences, since it was matched by a synergism of the mGluR5 receptor agonist CHPG and of the A2A receptor agonist CGS 21680 to counteract phencyclidine-induced motor activity in rats, which is a behavioral response known to be highly dependent on D2 receptor function. It seems possible that the combined activation of the A2A and mGluR5 receptors in the striatum may counteract the well known strong tonic D2 receptor-mediated inhibition of adenylyl cyclase and expression of immediate-early genes in the striatopallidal GABAergic neurons (see Section II.D.). Since immediate-early genes are involved in the connection between short- to long-term adaptive neuronal responses, the A2A/mGluR5 heteromeric receptor complex may have a role in striatal plasticity inter alia long-term depression and potentiation as well as in the sensitization to psychostimulants linked to dopamine-independent c-fos expression (see Section II.D.). Finally, chronic but not acute treatment with a mGluR5 antagonist can reverse a kinetic deficit in a 6-OH-dopamine model of Parkinson's disease (Breysse et al., 2002). This may be related to an altered trafficing of the A2A/mGluR5 heteromer, leading to its internalization and/or redistribution allowing a dominance of D2 signaling.
l. The Bradykinin B2 and Angiotensin AT1 Heteromeric Receptor Complex.
The first indications of the possible existence of a bradykinin/angiotensin II receptor/receptor interaction was obtained by quantitative receptor autoradiography in the nucleus tractus solitarius of the rat brain, a central cardiovascular center (Fior et al., 1993). The findings suggested that in the nucleus tractus solitarius bradykinin B2 receptors were involved in modulating in a differential way the affinity of the high and low affinity binding sites of the angiotensin II (AT1) receptors without effects on the Bmax values of the AT1 agonist binding sites. Thus, the affinity of the high-affinity agonist state of the AT1 receptors was reduced by bradykinin while bradykinin increased the affinity of the low affinity agonist state using agonist and antagonist radioligands for the AT1 receptor. It was suggested that this receptor/receptor interaction can contribute to the central vasopressor activity of bradykinin by reducing and increasing AT1-mediated transmission at high and low affinity agonist states, considered to be involved in vasodepressor and vasopressor activity, respectively (Fior et al., 1993). However, another interpretation of the results from the competition experiments with an iodinated AT1 receptor antagonist versus angiotensin II (revealing mainly the low affinity agonist component) is that bradykinin reduces the affinity of the AT1 receptor antagonist binding sites, allowing an improved competition by angiotensin II seen as a reduction in the IC50 values. Overall it may be considered that the antagonist state of the AT1 receptor can be differentially regulated by B2 receptor activation versus the agonist state. The modulation of the AT1 receptor antagonist binding sites by bradykinin, however, still remains to be determined.
Recently the discovery was made that angiotensin AT1 and bradykinin B2 receptors form heteromers in smooth muscle cells and HEK-293 cells, coexpressing AT1 and B2 receptors (AbdAlla et al., 2000) indicating that this may be the molecular basis for the intramembrane receptor/receptor interactions previously observed between these two receptors.
Immuno-affinity chromatography was performed on proteins from smooth muscle cells and AT1 receptor dimers were coenriched with the anti-B2 receptor antibodies. Since bradykinin and angiotensin II had been cross-linked to the B2 and AT1 receptor antibodies before immuno-affinity chromatography, the results suggested that high-affinity AT1 and B2 receptors form heteromeric complexes on smooth muscle cells. The HEK-293 cells, when expressing only one of the two receptors showed only a monomeric form, but when coexpressing the AT1 and B2 receptors a heteromer was demonstrated, consisting of the AT1 and B2 receptors. The stable AT1/B2 receptor heteromer could be demonstrated by SDS-PAGE (nonreducing conditions) and was not dependent on agonists but on the density of the two receptors. Thus, it seems likely that intramembrane receptor/receptor interactions reported earlier (Fior et al., 1993) reflect agonist-induced conformational changes in the binding pockets of preformed heteromers leading to alterations in ligand affinity of the other binding pocket. The most impressive finding in the article from AbdAlla et al. (2001) was the increase in the AT1 receptor/G protein coupling in the AT1/B2 receptor heteromer. This was seen, e.g., by the increased degree of AT1-stimulated redistribution of Gα protein into the cytosol, by the marked increase of angiotensin II-stimulated GTPγS binding and the substantial increases in inositol phosphates. An elegant analysis with B2 receptor mutants demonstrated that the AT1 signal increase in the heteromer was dependent on the G protein interface of the B2 receptors but not on the binding of bradykinin to the B2 receptors. The heteromer was, however, formed independently of interference with G protein coupling and with bradykinin binding. Thus, an important functional meaning of the heteromer in this case is the enhancement of AT1 receptor/G protein coupling and thus of AT1 receptor signaling. It has also been shown that the increased presence of the AT1/B2 receptor heteromer may contribute to the development of angiotensin II hypersensitivity in preeclampsia (AbdAlla et al., 2001).
Still another functional meaning may be a change in receptor trafficking, since the AT1/B2 receptor heteromer becomes internalized through a dynamin-dependent pathway in contrast to the case when they are expressed alone (dynamin- and clathrin-independent pathway).
m. Other Heteromeric G Protein-Coupled Receptor Complexes.
It has been reported that also β2 adrenergic receptors and δ opioid or κ opioid receptors can undergo heteromerization using coimmunoprecipitation technology (Jordan et al., 2001). Furthermore, protease-activated receptors PAR3 and PAR4 can also form heteromers, where PAR3 is a cofactor for PAR4 activation by thrombin in platelets (Nakanishi-Matsui et al., 2000).
A large number of functional receptor/receptor interactions exists among GPCR in the brain for which evidence of heteromerization is still lacking. This is because there has been no time so far to perform such studies but many are on the way. Based on the evidence for intramembrane receptor/receptor interactions at the recognition level, the following GPCR heteromerizations are postulated: NK1/5-HT1, NT receptor/D2, CCKB/D2, A2A/D3, Gal receptor/5-HT1A, Gal receptor/α2, NPY receptor/α2, AT1/α2, NPY receptor/AT1, B2/α2, and oxytocin receptor/α2 receptor/receptor heteromerizations (see previous citations in the Introduction, and Härfstrand et al., 1988; Aguirre et al., 1991; Hedlund et al., 1991, 1994; Fior et al., 1994; Yang et al., 1994a, 1996; Fior and Fuxe, 1995; Li et al., 1995a,b; Dasgupta et al., 1996b; Hedlund and Fuxe, 1996; Ferraro et al., 1997; Diaz-Cabiale et al., 2000a-2000e, 2001; Tanganelli et al., 2001). Finally, muscarinic acetylcholine receptor heterodimerization may also exist (Chiacchio et al., 2000).
C. Direct Protein/Protein Interactions between G Protein-Coupled Receptors and Multisubunit Ligand-Gated Ion Channels
1. The GABAAand Dopamine D5Heteromeric Receptor Complex. The direct protein/protein interaction between GABAA and dopamine D5 receptors was reported in an impressive paper by Liu et al. (2000). A colocalization of GABAA and D5 receptors was demonstrated in cultured hippocampal neurons by means of immunofluorescence studies in combination with confocal laser microscopy. This was in line with previous work indicating that D5 receptors in hippocampal neurons exist on dendritic shafts and in the axon hillock, regions rich in GABA synapses (Bergson et al., 1995; Nusser et al., 1995).
In Western blots, hippocampal GABAA receptors could be demonstrated after affinity precipitation with D5 but not D1 receptor COOH-terminal GST fusion proteins. Furthermore, GST fusion proteins with the GABAA receptor γ2 (short) second intracellular loop precipitated the hippocampal D5 but not D1 receptors. These results indicated a physical interaction between the COOH-terminal part of the D5 receptor and the GABAA receptor γ2 subunit, more precisely the second intracellular loop. GABAA/D5 receptor heteromerization was further demonstrated in coimmunoprecipitation experiments. In blot overlay assays, it was shown that the SDS-PAGE separated GST fusion protein of the γ2 subunit but no other subunit could directly bind the in vitro translated [35S]methionine-labeled D5 COOH-terminal peptide. Likewise, the [35S]methionine-labeled second intracellular loop of the γ2 subunit could directly bind to the GST-D5 receptor-COOH-terminal fusion protein. It should also be considered that the GST fusion protein of the second intracellular loop of the β2 subunit could not bind to the D5 receptor COOH-terminal part in the blot overlay assay in spite of the fact that this part can immunoprecipitate D5 receptors. This interaction may therefore be indirect via associated proteins or involve other parts than the COOH terminus of the D5 receptor. However, this part of the GABAA receptor may also have a role in the formation of the GABAA/D5 heteromeric receptor complex. It is of interest that there exists a GABAA receptor-associated protein that can interfere with the ability of the D5 receptor COOH-terminal tail to interact with the γ2 intracellular loop (Wang, 2002).
Studies in HEK-293 cells demonstrated that agonist coactivation of D5 and GABAA receptors (transient coexpression of GABAα1, β2, and γ2 receptor subunit) was necessary for the coimmunoprecipitation to take place. Thus, agonist-induced changes in the second intracellular loop of the γ subunit and in the COOH-terminal part of D5 receptor are essential for the formation of this heteromeric complex. One functional meaning of this heteromerization appears to be to allow mutually inhibitory cross-talk to take place. Thus, in cotransfected HEK-293 cells the D5 but not D1 receptor activation reduced by 30% the GABAA currents by decreasing the slope of the current-voltage curve. These changes were brought about by a cAMP-independent mechanism and were by means of D1/D5 receptor chimeras shown to be entirely dependent on the D5 receptor COOH-terminal/γ2 interaction in the heteromeric complex formed. Thus, these results indicate that D5 can reduce the synaptic strength over GABAA receptors via this complex. In agreement with this view, electrophysiological studies in hippocampal slices showed that a D1/D5 agonist could reduce the amplitude of the GABAA receptor-mediated miniature inhibitory postsynaptic currents independent of protein kinase C (PKC) and PKA. The injection of a GST-encoded D5 COOH-terminal peptide into the recorded neuron prevented this action. This heteromeric complex may therefore be of relevance in vivo, where it may control the synaptic strength of the GABAA receptor.
Signaling over the D5 receptor was in turn modulated by the GABAA receptor in an inhibitory way. Thus, in cells coexpressing the two receptors the GABAA receptor reduced the maximal dopamine activation of the adenylate cyclase by 45%, which was selective for the D5 versus the D1 receptor. This action did not involve changes in the D5 recognition site as shown by absence of changes in dopamine receptor agonist and antagonist affinity nor were any changes in Bmax values of D5 antagonist binding sites observed. Instead, it seems as if the GABAA activation of this heteromeric complex reduces the D5 receptor/Gs protein coupling through the γ2 intracellular loop 2/D5 receptor COOH-terminal physical interaction. Thus, expression of minigenes encoding the γ2 intracellular loop 2 sequences or the D5 receptor COOH-terminal sequences blocked the GABAA modulation of D5-induced cAMP accumulation. The use of D1/D5 chimeras gave further evidence for the crucial involvement of the agonist-induced D5 receptor COOH-terminal/γ2 complex in the regulation of the G protein coupling of the D5 receptor.
Another functional meaning of this dynamic heteromeric complex may be a role in receptor trafficking (Wang, 2002). Thus, there exist indications that agonist activation of either the GABAA or the D5 receptor makes possible endocytosis of both receptors and thus cotrafficking (Wang, 2002). It should also be considered that the γ2 subunit may be essential for the clustering of the postsynaptic GABAA receptors. A potential role of this GABAA receptor/D5 receptor complex in the pathophysiology of schizophrenia was also postulated in view of the fact that alterations in D1/D5 and GABAA γ2 containing receptors and their functions may exist in the schizophrenic brain (Goldman-Rakic and Selemon, 1997; Okubo et al., 1997; Huntsman et al., 1998; Keverne, 1999).
Very recently Liu and collaborators (personal communication) have obtained evidence that D1 receptors can directly interact with NMDA receptors via protein/protein interaction. Thus, NMDA/D1 heteromerization may have a role in the regulation of glutamate transmission. The demonstration of the GABAA/D5 and possibly NMDA/D1 receptor complexes are very exciting, since they open up a new way to understand how volume transmission (VT) and wiring transmission (WT) signals can become integrated (Agnati et al., 1987; Zoli et al., 1993; Agnati and Fuxe, 2000). Thus, the GABAA and NMDA receptors are classical fast synaptic receptors operating via regulation of their ion channels, whereas the GPCR are slow and mainly located extrasynaptically and reached by VT signals in the extracellular space. Heteromerization of ion channel receptors and GPCR offers a new mechanism for the integration of WT and VT and how to control synaptic strength of crucial importance for learning and memory (Abel and Kandel, 1998; Agnati et al., 2002a,b). It may be mentioned that several years ago we obtained indications that GABAA receptor agonists could modulate the binding characteristics of D2-like receptors in striatal membrane preparations (Pérez de la Mora et al., 1997). Thus, the GABAA agonist muscimol reduced the affinity of the high-affinity D2 receptor agonist sites as shown in competition experiments with [3H]raclopride versus dopamine. Thus, GABAA/D2 heteromeric receptor complexes may also exist, since such interactions at the recognition level have been regarded (Zoli et al., 1993) as biochemical indicators of the existence of a heteromeric complex, in this case between GABAA and D2 receptors. It is also of interest that early on we could report the ability of agonist-activated α2 adrenergic and D1 receptors to substantially modulate [3H]nicotine binding in membrane preparations from the tel-diencephalic regions (Fuxe et al., 1988c, 1989). In view of the above, it seems relevant to test the existence also of heteromeric complexes between nicotinic and α2 adrenergic receptors and between nicotinic and D1 receptors involved in the control of allosteric mechanisms at nicotinic acetylcholine receptors (Changeux and Edelstein, 2001). It is also of substantial interest that cross-inhibition between certain transmitter-gated cation channels (ATP-gated P2X2 and α3β4 nicotinic channels) has been shown to exist upon their coactivation (Khakh et al., 2000) probably reflecting heteromerization between these two ion channel receptors when coactivated.
D. Oligomeric Complexes Containing G Protein-Coupled Receptors and Receptor Tyrosine Kinases
Recently, it has been reported that the epidermal growth factor (EGF) receptors can become associated with growth hormone (GH) receptors and with β2 adrenergic receptors upon their stimulation by GH and β2 receptor agonists, respectively (Yamauchi et al., 1997; Maudsley et al., 2000). Evidence was provided that their physical association resulted in a transactivation of the EGF receptor. As an example, we will focus on how the β2 adrenergic receptor can produce EGF receptor dimerization, tyrosine auto-phosphorylation, and EGF receptor internalization (Maudsley et al., 2000) leading to MAPK activation. A prerequisite for such a RTK transactivation appears to be the β2 agonist-induced formation of a multiprotein complex containing not only the β2 adrenergic receptor and the EGF receptor but also β-arrestin and c-Src, a nonreceptor tyrosine kinase. The coimmunoprecipitation experiments demonstrated the β2 adrenergic/EGF heteromeric receptor complex that rapidly formed with a peak within minutes after β2 agonist treatment correlated with the β2 agonist-induced MAPK activation. This complex could be detected also under basal conditions probably related to a constitutive activity of the β adrenergic receptor, since this basal β2 adrenergic/EGF receptor complex was markedly reduced by a β2 adrenergic receptor inverse agonist. It was shown that Src kinase inhibitors blocked the formation of the β2 adrenergic/EGF receptor complex and MAPK activation, indicating a critical role of Src catalytic activity. Furthermore, β-arrestin recruits c-Src protein to the β2 adrenergic receptor (Luttrell et al., 1999) after β-arrestin has become linked to the β2 adrenergic receptor through its agonist-induced phosphorylation via G protein receptor kinase. It was noticed that the transactivation of the EGFR by β2 adrenergic receptor was blocked by inhibition of clathrin-mediated endocytosis. It may therefore be that the clathrin-coated pits can represent microdomains for optimizing the signaling among the assembled proteins, leading to the RAS-dependent activation of Raf after the transactivation of the EGF receptor has occurred. As pointed out by Maudsley et al. (2000), these results open up an important mechanism for how GPCRs and also cytokine receptors (Phonphok and Rosenthal, 1991; Quijano et al., 1998) may control trophic signaling, namely through agonist-induced heteromeric complexes with EGF receptors leading to regulation of its transactivation followed by MAPK activation. These studies are excellent examples of how multiprotein complexes form a crucial role also in trophic signaling.
E. Oligomeric Complexes Containing G Protein-Coupled Receptors and Receptor Activity-Modifying Proteins
1. Receptor Activity-Modifying Transmembrane Proteins.
a. The Calcitonin Receptor Family/RAMP1-3 Heteromeric Complexes.
The calcitonin family peptides such as calcitonin, calcitonin gene-related peptides (CGRP), adrenomedullin, and amylin act via GPCR. Among others, the calcitonin-receptor-like (CRL) receptor was cloned but could not bind CGRP peptides. An attempt was therefore made to clone the gene for the CGRP receptor by expression cloning (McLatchie et al., 1998). In this study, a single cDNA was finally shown to markedly enhance the responses at an endogenous CGRP receptor in Xenopus oocytes. The cDNA was shown to encode a protein with a single TM domain and an extracellular NH2 terminus (RAMP1). Expression of this protein selectively enhanced CGRP-induced actions in the oocytes. This was the way the RAMP1 was discovered (McLatchie et al., 1998).
Subsequent experiments on cell lines (HEK-293 T cells) demonstrated that coexpression of RAMP1 was necessary for CRL receptor ligand binding (increases in binding of 125I-CGRP1) and function in terms of cAMP accumulation. The question was how RAMP1 brought about the development of a functional CGRP receptor based on coexpression of CRL receptor and RAMP1. By means of epitope-tagged CRL receptor and RAMP1 in combination with fluorescence-activated cell sorting, it was shown that their coexpression made possible the cell surface expression of both receptors. It was of substantial interest that SDS-PAGE could show cross-linking of 125I-CGRP1 to two proteins (66,000 and 17,000 bands) from the surface of cotransfected HEK-293 T cells. The two bands seemed to correspond to the native CGRP receptor and RAMP1, respectively, which may form an easily disrupted heteromeric complex according to these findings. Further experiments demonstrated that the coexpression of epitope-tagged RAMP1 and CRL receptor led to the disappearance of a 58-kDa band found with the expression of CRL receptor alone with the appearance of a diffuse 66-kDa band correlating in size with the band cross-linked to
125I-CGRP1. This additional increase in size by 8 kDa could not be due to an association with the epitope-tagged RAMP1 having a size of Mr 14 kDa. Instead, experiments with endoglycosidases F and H indicated that it was related to a terminal glycosylation of CRL receptor, not found when it was expressed alone and subject only to core glycosylation. The terminal glycosylation of CRL receptor found in the presence of RAMP1 indicates that CRL receptor now can be expressed on the cell surface as a mature glycoprotein, capable of being a CGRP receptor. Nevertheless, it seems likely that also the RAMP1 coexpressed with the terminally glycosylated CRL receptor on the cell surface and capable of becoming cross-linked with 125I-CGRP1 can contribute to regulation of CRL receptor ligand selectivity and function by physical and/or indirect interactions. It has also recently been observed that multiple amylin receptors can be formed by RAMP interactions with the calcitonin receptor gene product (Christopoulos et al., 1999).
Two other RAMP proteins called RAMP2 and RAMP3 were also discovered (McLatchie et al., 1998), and it was found that RAMP2 and CRL receptor could generate a receptor for adrenomedullin (ADM). It was found that RAMP2 allowed the transport to and expression of a core-glycosylated CRL receptor (the 58-kDa protein) on the cell surface. This was shown not to be a regular CGRP receptor. It was elegantly demonstrated that oocytes expressing RAMP2 and CRL receptors substantially respond to low concentrations of ADM fragments but only weakly to CGRP. In HEK-293 T cells coexpressing these proteins, specific ADM binding and ADM-mediated increases of cAMP accumulation could be demonstrated. Thus, RAMP2-transported CRL receptor becomes ADM receptors on the cell surface. Based on this work it seems as if the functional meaning of the various RAMPs is severalfold, namely to transport the CRL receptor to the cell surface, to differentially glycosylate the CRL receptor and to interact differentially with the CRL receptor on the cell surface, which may involve heteromeric complexes. All these processes may lead to the development of receptor diversity with markedly different ligand specificities, and CGRP, ADM, and amylin receptor subtypes (see Chen et al., 1997; Perry et al., 1997) can be formed based on the CRL receptor and calcitonin receptor interactions with different types of RAMPs. Thus, CRL receptor and possibly other GPCR can markedly alter its binding pocket by interactions with single TM domain proteins. The role of core and terminal glycosylation versus receptor/RAMP protein interactions resulting in putative heteromeric complexes remain to be determined. When reading about RAMPs one becomes aware of the possibility that many receptors may not function in their own right but mainly as partners in heteromeric complexes and in association with other membrane-associated proteins.
b. The Dopamine D1 Receptor/Calcyon Heteromeric Complex.
In an attempt to understand how D1 receptors can couple to multiple G proteins, the groups around Goldman-Rakic and Bergson began searching for D1 receptor-interacting proteins with a yeast two-hybrid screen, using the COOH-terminal part of the human D1 receptor as bait (Lezcano et al., 2000). They found a 24-kDa single transmembrane protein, named calcyon, that could interact with the D1 receptor and produce enhancement of D1 receptor-induced Ca2+ signaling. The calcyon may therefore be regarded as a RAMP where the interaction is focused on the G protein coupling and not on the binding pocket selectivity as described above for RAMP1-2. The calcyon appears to have an NH2 terminus located extracellularly and a COOH terminus intracellularly located domain like the RAMP1-3 and also contains N-linked oligosaccharides. Immunocytochemistry demonstrated that the D1 receptor and calcyon colocated in the same population of pyramidal cells of the cerebral cortex and in a subpopulation of D1 receptor-containing striatal neurons. It is of interest that both calcyon and D1 receptors were located perisynaptically in dendritic spines at a postsynaptic location.
Coimmunoprecipitation experiments indicated that D1 receptor and calcyon formed a heteromeric complex in HEK-293 cells. Furthermore, using a GST fusion protein with the D1 receptor bait sequence (GSTD1) and a bacterial fusion protein with the COOH-terminal part of calcyon (S-calcyon), it could be shown that GSTD1 became bound to S-calcyon. Thus, a direct COOH-terminal interaction may be involved in the formation of this calcyon/D1 receptor heteromeric complex, where the D1 receptor COOH-terminal sequences 421-435 appear crucial since this peptide prevented the binding.
The formation of the calcyon/D1 receptor heteromeric complex resulted in a marked change in D1 receptor signaling in HEK-293 cells. After priming by activation of ATP P2Y receptors but not otherwise, the D1 receptor agonist SKF 81297 produced a rapid increase in Ca2+ signaling dependent on release from intracellular Ca2+ stores provided that transfection with calcyon had been performed. This Ca2+ response was similar in size to that produced by the P2Y receptor linked to Gq/11.
These results can be explained by assuming that the P2Y activation can increase the coupling of the D1 receptor to the Gq/11 protein, an increase that may be further strengthened by the formation of the D1 receptor/calcyon complex. In contrast, the D1 receptor signaling over the Gs to adenylyl cyclase leading to increases of cAMP accumulation was unaltered by the formation of this complex. Thus, the dual signaling of the D1 receptor with involvement also of Ca2+ signaling via Gq/11 becomes more pronounced by this heteromeric complex with calcyon.
Formation of this complex appeared to be inhibited by expression of D1421-435 in the cells, which reduced the D1 receptor-stimulated Ca2+ responses probably by competing with the D1 receptor for calcyon. The heteromeric complex formation is increased by the D1 receptor activation and reduced by a PKC inhibitor. In fact, the COOH-terminal part of the calcyon can become phosphorylated by PKC and may bind phosphoinositol-4,5-biphosphate. These observations give an increased understanding to the importance of the direct interaction between the COOH-terminal parts of D1 receptor and calcyon for the increased Gq/11 protein coupling of the D1 receptor with activation of phospholipase C (PLC) and formation of inositol 1,4,5-trisphosphate (IP3) (see Section II.D. for details of a Gq-mediated signaling).
It is of substantial interest that stimulation of endogenous muscarinic Gq/11-coupled M1 receptors, like ATP P2Y activation, prior to D1 receptor activation also made possible a strong D1 receptor-induced increase in Ca2+ signaling by the heteromeric D1 receptor/calcyon complex. It seems possible that the mechanism underlying the primary actions of ATP and M1 receptor activation can be severalfold with the end result being an increased Gq/11 coupling of the D1 receptor. Thus, the indication of multiple receptor interactions with formation of high order heteromeric complexes containing ATP and D1 receptors and calcyon or M1 and D1 receptors and calcyon may be considered leading to the increase of D1 receptor coupling to the Gq/11 protein. The activation of intracellular phosphorylation cascades involving, e.g., PKC can also be involved with phosphorylation of the COOH-terminal calcyon, especially since the priming action have so far only been observed with GPCR coupled to Gq/11.
The pioneering work of Goldman-Rakic has demonstrated the prominent role of the D1 receptor in the cerebral cortex in the modulation of glutamate receptor signaling during working memory operation (Williams and Goldman-Rakic, 1995). This work will therefore have relevance for understanding cortical plasticity. It is also inspiring that M1 receptors, like D1 receptors, are located on dendritic spines of pyramidal nerve cells (Mrzljak et al., 1993). Thus, M1/D1 receptor/receptor interaction via possible heteromeric complexes may therefore exist in dendritic spines and have a functional relevance (Wang and McGinty, 1997).
2. Receptor Activity-Modifying Cytosolic Proteins.
a. The Adenosine A1 Receptor and Adenosine Deaminase Heteromeric Complex.
Franco and collaborators (Franco et al., 1997, 1999, 2000) have obtained evidence that adenosine deaminase (ADA) can form a heteromeric complex with adenosine A1 receptors. ADA is an enzyme capable of converting adenosine in inosine. ADA is also a multifunctional protein appearing on the cell surface anchored to different proteins (Lluis et al., 1998; Mirabet et al., 1999, Herrera et al., 2001). It can therefore act enzymatically but also extraenzymatically (Franco et al., 1997, 1998) as in the case of the formation of ADA/A1 receptor complexes. Formation of ADA/A1 receptor complexes was demonstrated in experiments involving confocal laser microscopy, coimmunoprecipitation, and affinity chromatography. Thus, ADA and A1 receptors coimmunoprecipitated and A1 receptors were retained in a matrix of ADA-Sepharose. Finally A1 receptors colocalized with ADA on cell membranes, including cell surface cortical neurons (primary culture; Ruiz et al., 2000). The binding of ADA to the A1 receptors appears to be essential for the high-affinity agonist binding of A1 receptors, giving a functional role of this physical interaction in A1 receptor-mediated transmission (Ciruela et al., 1996; Saura et al., 1996, 1998). Thus, ADA has a role not only as a degradative ectoenzyme but also as an A1 receptor activity-modulating protein. It therefore became of interest to study a possible role of ADA in the A1/D1 heteromeric receptor complex. In a recent study (Torvinen et al., 2002), using the same A1/D1 receptor-cotransfected fibroblast cell line as described above, evidence was obtained that in nonpermeabilized A1/D1 receptor-cotransfected cells, but not in cells only transfected with D1 receptor, ADA exists on the plasma membrane. These results indicated that ADA can be targeted to the membrane by A1 receptors but not by D1 receptors. Futhermore, A1 receptor agonist (R-PIA, 100 μM, 3 h) preincubation resulted in coaggregations of both A1 receptors and ADA and D1 receptors and ADA in the A1/D1 receptor-cotransfected fibroblast cells. These results suggested that after A1 receptor agonist treatment with maintained A1/D1 heteromerization coaggregates are formed that contain high-order molecular structures (Torvinen et al., 2002) involving ADA/A1 receptor/D1 receptor heteromeric complexes and other interacting proteins that have a special functional role, especially in receptor trafficking. However, ADA does not seem to be linked directly to D1 receptors. In line with this view, ADA/D1 receptor aggregates are no longer present after D1 receptor agonist pretreatment (SKF 38593, 10 μM, 3 h), disrupting the A1/D1 receptor heteromerization leading to aggregated D1 receptor alone, while ADA/A1 receptor immunoreactivity remain diffusely colocalized (Torvinen et al., 2002). The impact of the ADA/A1 receptor complex on D1 receptor signaling was demonstrated by the blockade of the A1 agonist-induced uncoupling of the D1 receptor by the irreversible blockade of ADA function using deoxycoformycin. This counteraction was unrelated to the rise of endogenous adenosine levels (Torvinen et al., 2002). Thus, ADA is part of the A1/D1 heteromeric receptor complex but directly linked only to the A1 receptor, where it makes possible the high-affinity state of the A1 receptor for agonists, allowing it to modulate the operation of the D1 receptor.
b. The Adenosine A1 Receptor and hsc73 Heteromeric Complex.
Sarrió et al. (2000) demonstrated that adenosine A1 receptors interact with a protein of the family of heat shock proteins. By affinity chromatography the heat shock cognate protein hsc73 was identified as a cytosolic component able to interact with the third intracellular loop of the receptor. As demonstrated by surface plasmon resonance, purified A1 receptors interact specifically with hsc73 with a dissociation constant in the nanomolar range (0.5 ± 0.1 nM). The hsc73/A1 receptor interaction leads to a marked reduction in the affinity of A1 receptor agonist ligands, a reduction of A1 receptor antagonist binding, and prevents activation of G proteins, as deduced from [35S]GTPγS binding assays. Interestingly, this effect on A1 receptor agonist binding was stronger than that exerted by guanine nucleotide analogs, which uncouple receptors from G proteins, and was completely prevented by ADA, which interacts with the extracellular domains of A1 receptors (see above). As assessed by immunoprecipitation, a high percentage of A1 receptors in cell lysates are coupled to hsc73.
Members of the hsp70 family interact with a number of cellular proteins. Due to the molecular chaperone function of hsp70 proteins, they appear capable of recognizing “non-native” forms of proteins. This is not the case for A1 receptors for some reasons. One is due to the fact that solubilized A1 receptors are very sensitive to the composition of the medium, which affects ligand binding to the soluble molecule. Thus, a precise combination of detergent and salts is required to achieve a high recovery of binding sites in solubilized preparations of A1 receptors. The strong effect of nanomolar concentrations of hsc73 upon ligand binding to purified soluble A1 receptors is evidence for a specific interaction between hsc73 and functional A1 receptors. The specificity of the interaction has also been demonstrated in primary cultures of neurons where other GPCR (A2A receptors, A2B receptors, or metabotropic glutamate mGluR4 receptors) do not colocalize with hsc73.
On the other hand, colocalization between A1 receptors and hsc73 is not restricted to a specific zone of the cell, even in cells naturally expressing the proteins, and this is evidence that the interaction occurs with completely folded functional receptors. Apart from the regulatory role of the interaction in ligand binding, there are data supporting the idea that hsc73 is relevant for the trafficking of the A1 receptors. As a matter of fact, colocalization between hsc73 and A1 receptors was detected in specific regions of rat cerebellum and in the nerve cell bodies of cortical neurons but not in dendrites or synapses. Moreover, it seems that agonist-induced receptor internalization leads to the endocytosis of A1 receptors by two qualitatively different vesicle types, one in which A1 receptors and hsc73 colocalize and another in which hsc73 is absent. These results open the interesting possiblity that the signaling and trafficking of GPCR may be regulated by heat shock proteins.
The novel findings presented by Sarrió et al. (2000) suggest a specific role for hsp70 proteins in regulating the activation and operation of A1 receptors. Although a relevant role for chaperones in signaling by steroid hormone receptors has already been demonstrated (Bohen et al., 1995; Caplan et al., 1995), our results are the first evidence suggesting a control by hsc73 of signaling via plasma membrane GPCR. It should also be noted that this member of the GPCR can be regulated differently by a protein interacting with extracellular domains of the receptor (ADA, see above) and by a cytosolic protein interacting with the third intracellular loop of the receptor (hsc73).
II. On the Functional Implications of Receptor/Receptor Interactions
A. The Context of the Present Discussion
The existence of homodimers, heterodimers, homo-oligomers, and hetero-oligomers of GPCR provides the structural framework to explain the function of GPCR in a variety of systems. From a biochemical point of view, the formation of GPCR homo- and heteromers explains some of the data on ligand binding and of cross-talk that have been reported for many years in the literature and that were interpreted in an erroneous or incomplete way. The knowledge of the molecular mechanisms underlying receptor function introduces, however, complexities derived from the fact that different conformations of a single receptor may arise and that receptor molecules with different conformation and from different receptors can interact to give rise to multiple oligomeric structures in specific membrane microdomains. The role of structural diversity in GPCR function will be the topic covered first in this section (Gutkind et al., 1998; Bockaert and Pin, 1999; Gether, 2000; Heuss and Gerber, 2000; Lefkowitz, 2000).
On the other hand, it is important to give an example of how heteromerization of the receptors can improve our understanding of how signals are integrated in a given system. Since substantial indirect evidence for receptor/receptor interactions has been provided, studying receptors for neurotransmitters and recent evidence for heteromeric receptor/receptor interactions has been given in the central nervous system, a complete and comprehensive account of functional implications of certain receptor/receptor interactions occurring in the basal ganglia is provided. Thus, the interactions between adenosine, dopamine, and glutamate metabotropic receptors in the GABAergic striatopallidal neurons will be covered in full. It should be noted that the heteromerization concept gives new therapeutic directions for treatment of diseases involving inter alia this brain region, such as Parkinson's disease, schizophrenia, and drug addiction.
B. Structural Basis of Receptor Function
1. Conformational Diversity.
It seems likely that the receptor can assume not only two conformations related to two functional states: active (R*) versus inactive (R) receptor, but rather several slightly different conformations. In fact, proteins can assume a large number of slightly different structures each of which with potentially different biochemical characteristics.
The pharmacology of GPCR has led to the frequent finding of negative cooperativity in agonist binding. In fact, the work of Limbird et al. (1975) and Limbird and Lefkowitz (1976) gave clear evidence for negative cooperativity for binding of adrenergic agonists to β adrenergic receptors, which was further demonstrated for a variety of GPCR. In the case of catalytic proteins a negative cooperativity could be explained only in multimeric enzymes and assuming that there was intercommunication between the enzyme subunits. In a scenario where GPCRs were assumed to be present as monomers in the membranes, negative cooperativity was difficult to explain. One possible explanation for this apparent negative cooperativity was the assumption that a given receptor can exist in two conformational states with different affinity for the agonist. It is frequently assumed that the high-affinity form is due to the conformation of the receptor/G protein complex whereas the low affinity form is due to the conformation of the receptor uncoupled from the G protein. These coupled and uncoupled receptor states would correspond to the functionally active (R*) and inactive forms (R). The view that some of the GPCRs displaying negative cooperativity existed in membranes as two totally independent conformations had to change by the finding that the affinity constant deduced from equilibrium binding or kinetic binding experiments were different (Casadó et al., 1991). This should not be the case for two independent affinity states.
The existence of receptor dimers would lead to an explanation of the kinetic data on agonist binding assuming that there is negative cooperativity among the two interacting receptor monomers. The truth may be not as simple as that. The occurrence of clustering clearly suggests that G protein-coupled receptors form high order molecular structures, in which multimers of the receptors and probably other interacting proteins form functional complexes (Fuxe and Agnati, 1987). Therefore, negative cooperativity has to be explained taking into account the variety of conformations in which a given GPCR can exist in the cell membrane. There is a model for GPCR operation that takes into account that different GPCR molecules can display different conformations at a given time and that agonist binding can change this conformational pattern. The cluster-arranged cooperative model, which accounts for the kinetics of ligand binding to adenosine A1 receptors (Franco et al., 1996), shows that high- and low-affinity sites are a consequence of the negative cooperativity of agonist binding and may not be related to the content of free and G protein-coupled species. Conceptually this model takes advantage of the fact that GPCRs are not isolated proteins but linked with other components of the membrane. This intramembrane intercommunication is the basis of the observed negative cooperativity and can even participate in the multiple processes involved in ligand-induced desensitization. Assuming that the GPCR communicate with each other in the membrane, the idea of the model is that agonist binding decreases the affinity of receptors that are not yet interacting with the ligand. The molecular mechanisms of this effect are multiple and probably vary from GPCR to GPCR. The validity of the model was proven with A1 adenosine receptors, which cluster after agonist binding (Franco et al., 1996). For this receptor, it is likely that clustering is one of the factors affecting the conformation of the receptor in such a way that it decreases the affinity for the agonist. This is also the reason that the model was named as cluster-arranged cooperative model. The model considers that the cluster may be formed by molecules other than the A1 receptor itself, and therefore, it is open to heteromeric and multimeric interaction involving two or more proteins and even membrane lipids (Franco et al., 1996). This broad conformational spectrum that is possible for GPCR confers functional plasticity of GPCR since the conformational pattern and thus the energy landscape in response to a variety of effectors, chemical and also physical stimuli will indeed vary. These physicochemical-induced influences on the GPCR energy landscape is homologous to what happens in computational processes (Frauenfelder et al., 1991). Therefore, GPCRs are capable of detecting the specific features of the cellular microenvironments in which they are embedded to assume a certain preferential conformation, which in turn affects function (Fig. 1). Thus, the following can be proposed (Agnati et al., 2002a,b):
-
GCPR conformation at a given circumstance is the transient result of the chemical-physical factors acting upon the receptor and coming simultaneously from the extracellular space (Ciruela et al., 1996; Saura et al., 1996, 1998), the membrane (Casadó et al., 1990, 1991, Ginés et al., 2001), and the intracellular space (Sarrió et al., 2000).
-
Ligand induces conformational changes and conformational changes in the receptor would affect ligand binding and, therefore, signaling (Franco et al., 1996).
-
The chemical-physical inputs can lead to a conformation that weakens the intramolecular stabilizing interactions favoring constitutive ligand-independent activation of the receptor [see the so-called “protonation hypothesis” for GPCR conformation (Gouldson et al., 2000)].
-
The GPCR conformation may affect signaling; that is, one GPCR conformation may favor the activation by the receptor of certain particular decoding pathways (Kenakin, 1995, 1997). It should be noted that signaling is also dependent on the G protein interacting with the GPCR, but again, it is reasonable to assume that a given G protein interacts preferentially with a given conformation of the receptor (Cordeaux et al., 2000).
-
The GPCR conformation may affect receptor turnover, and thus signaling and receptor trafficking (Ginés et al., 2000; Hillion et al., 2002).
-
As a result of the above, the conformational pattern and thus the energy landscape is the key for the output in terms of function.
2. Oligomeric Diversity: The Receptor Mosaic
After the reports indicating the existence of heterodimers and heteromers of GPCR, the present data on receptor/receptor interactions cannot distinguish between domain-swapped dimers/heteromers and contact dimers/heteromers (Gouldson et al., 2000; Schulz et al., 2000). It is our opinion that very likely both types of interactions (domain-swapping and domain contact) can occur allowing the formation of high order hetero-oligomers. The prevalence of one of the two ways may depend on the receptor type and on the chemical-physical environments in which the interacting receptors are embedded that affect GPCR conformations (see above).
Recent papers have carefully discussed the mechanistic aspects of receptor/receptor interactions (Gouldson et al., 2000). Based on this work, it is possible to distinguish two pseudoindependent units in the GPCR; thus the NH2-terminal and helices 1 through 5 constitute the A-GPCR domain, whereas helices 6 and 7 through to the COOH terminus constitute the B-GPCR domain. The two A and B domains are connected by a hinge loop (ICL3) that is, frequently, the longest loop in GPCR and therefore very well suited to allow reciprocal movements of the two A and B-GPCR domains. It has been suggested that helices 5 and 6 form the dimerization interface and the 5 and 6 domain-swapped dimer may be the active (R*) form of the receptor that interacts with the G protein. Examination of various structures of adrenergic receptors by means of molecular dynamics suggests that the role of the agonist may be that of stabilizing the 5 and 6 dimer through conformational changes in helices 5 and 6. One problem is the dilemma whether agonist promotes dimerization or dimers are already preformed. It has been suggested that dimers are preformed and merely rearranged in the presence of the agonist (Gouldson et al., 2000). Although this can happen for some receptors (Ginés et al., 2000; Hillion et al., 2002), current evidence indicates that agonists affect markedly the oligomerization state of some GPCR (see Table 1 and Franco et al., 1996). It is important to underline that functional sites have been identified on the external face of helix 2, which could be involved in the formation of heterodimers and thus in the formation of high order hetero-oligomers as well as in heterodimerization processes with other proteins.
This view is in agreement with the receptor mosaic hypothesis suggested in 1982 (Agnati et al., 1982, 1990, 2002, 2003; Fuxe et al., 1983a; Zoli et al., 1996), with the cluster arranged cooperative model and with the fact that quite often agonists lead to clustering of GPCR cell surface receptors (Franco et al., 1996). Taking into account that oligomeric complexes are likely composed of ordered conglomerates of a number of different membrane, extracellular and intracellular proteins, such clusters were viewed as computational units having an important role in the information handling by the cell and were defined by Agnati et al. (1982) as a “receptor mosaic” or, more generally, a “protein mosaic” that may contain also other plasma membrane-associated proteins (Fuxe and Agnati, 1985; Agnati et al., 2002, 2003).
It is reasonable to think that the composition, geometry, and characteristics of this protein mosaic will depend on the above-mentioned physical-chemical factors, and in turn, the characteristics of the mosaic will target specific patterns of signaling pathways. Therefore, activation of one receptor would lead to different cell responses depending upon the nature of the mosaic where the receptor is located. It is evident that different cells, or the same cell in different conditions (activation, mitosis, etc.) could lead to different signaling scenarios for a given receptor since the composition and function of the mosaic will depend on the pattern of trafficking and energy landscape (conformational state) of the proteins forming the mosaic (Agnati et al., 2003).
Some basic features of a receptor mosaic with functional consequences are described as follows (Agnati et al., 2002):
-
A receptor cluster works as a receptor mosaic if and only if the cluster of proteins (heteromeric receptor complex involving GPCR and/or other proteins as ion channel receptors, tyrosine kinase receptors or receptor activity-modulating proteins) is such that activation of a receptor modulates the biochemical/functional features of at least another receptor of the cluster, i.e., if and only if receptor/receptor interactions are in operation in the cluster of molecules, where adapter protein may play an important role in stabilizing the receptor mosaics.
-
The fluctuation of each receptor (of the receptor mosaic) among its possible conformational states is conditioned by the conformations of the other receptors in the receptor mosaic (receptor/receptor interactions). Hence, each receptor will respond to its ligand in a way that depends on its conformation and thus on its interactions with the other receptors of the macromolecular complex.
-
The receptor mosaic is connected with effector proteins (e.g., G proteins, ion channels) and works as an integrative unit.
It should be noted that proteins can work as modules capable of building up supra molecular complexes (Pawson, 1995; Pawson and Scott, 1997; Pawson et al., 2001). This can be used for a variety of molecular events, for instance, to “hide” functional groups of a protein or to have emergent complex functions performed by protein aggregates. On the other hand, protein/protein association is, inter alia, strongly regulated by protein phosphorylation (Huganir and Greengard, 1987; Greengard et al., 1998, 1999; Hunter, 2000). Examples of protein/protein interactions are the receptor/receptor interactions, the assembly of αβγ subunits of the G proteins, and those interactions occurring along the MAP kinase pathways (Impey et al., 1999). Some of the possible interacting proteins can work indirectly as “scaffolds” (Pawson, 1995; Luttrell et al., 1999, 2001; Kohout et al., 2001; Miller and Lefkowitz, 2001). These proteins do not have the capability to transfer or elaborate information, but provide a scaffold along which other proteins transfer and/or elaborate information in an ordered and efficient fashion. For example, they provide a scaffold along which a series of enzymes process their substrates in a well defined sequence and with an efficiency, specificity, and rate that is clearly not possible in a freely diffusing system. These scaffolding proteins have indeed a relevant role that must be taken into account to understand how recognition occurs and signaling is transmitted. They may even be involved in processes such as those related to memory in the central nervous system.
Aspects on receptor activation should be revised in the light of the receptor plasticity and especially in light of receptor/receptor interactions. The different models of receptor activation that have been proposed (Kenakin, 1995, 1997) such as the two-state model versus the multistate model and the conformation selection versus conformation induction of the receptor binding site are still useful. The cluster arranged cooperative model (see above) has given a new perspective on the ligand binding and activation of GPCR (Franco et al., 1996). However, on the basis of the recent evidence on GPCR plasticity and on the suspected complexity of receptor mosaics, a new more global model should be devised. This model should take into account that the receptor recognizes and decodes the neurotransmitters or other physical-chemical signals according to the inputs that it receives from the other proteins of the membrane (in particular the other receptors of the cluster) and from the extracellular and intracellular signals impinging on it. On the other hand, the receptor mosaic concept can help in understanding better at a molecular level how complex integrative tasks performed by neurons (such as memory, see below) can take advantage of molecular circuits located at the plasma membrane level.
3. The Role of the Receptor Mosaic in Learning and Memory.
Grant and Husi (2001) have elegantly described the synaptic multiprotein complexes associated with the NMDA receptors and the PSD-95 (postsynaptic density protein 95) and how they can process information and encode memories (Migaud et al., 1998). In recent articles (Agnati et al., 2002a,b) building on earlier articles (Agnati et al., 1982; Fuxe et al., 1983a), we have postulated the existence of three-dimensional molecular circuits built up of intramembrane receptor mosaics, ion channels, and G proteins linked to cytosolic protein networks of protein kinases, protein phosphatases, scaffold, and anchoring proteins (Weng et al., 1999). These three-dimensional molecular circuits can store information in nerve cells, where engram formation may depend on the resetting of receptor mosaics (higher order hetero-oligomeric complexes). Transient stabilization of the receptor mosaic is postulated to result in short-term memory associated with an appropriate change of synaptic weight. Long-term memory, i.e., engram consolidation, may according to our hypothesis be caused by the ability of the molecular circuits involved to form intracellular signals translocating to the nucleus and activating a large number of transcriptional factors leading also to induction of various immediate early genes. In this way inter alia a number of postulated unique adapter proteins can be formed that will stabilize the receptor mosaic and lead to the consolidation of the receptor mosaic so that it becomes a long-lived heteromeric receptor complex (Agnati et al., 2002a,b). It will then, when again activated, induce ion channel activity and protein kinase and phosphatase cascades in its molecular circuit that will develop the synaptic weight at the time of the perception of the event to be remembered, and the retrieval of the engram can take place, since the appropriate pattern of firing rate in the nerve cell can occur. It must be underlined that this theory on the molecular basis of learning and memory is in agreement with the Hebbian rule stating that learning is associated with simultaneous changes in pre- and postsynaptic activity. According to our theory, the change of presynaptic activity will cause the specific change of postsynaptic activity by reorganizing the receptor mosaics of the postsynaptic membrane, leading to novel or altered molecular circuits that upon integration with other available three-dimensional molecular circuits in the nerve cell will produce a change of the firing rate pattern that is linked to the change in presynaptic activity.
The development of long-lived heteromeric receptor complexes of high order may, in the nerve cell membrane, therefore be the molecular basis for learning, memory, and retrieval (Hobb, 1949). Here the important work on cooperativity in biological membranes and of clustered receptors must also be mentioned (Changeux et al., 1967; Duke and Bray, 1999) and is in line with this hypothesis.
C. Communication Processes in the Cell
The concept to be introduced here is derived in part from studies on the communication processes at the level of neural networks (Agnati and Fuxe, 2000). Thus, the communication within a compartment and between compartments in the cell may occur via WT as well as via VT. In the first instance, there is the channeling of the information along some sort of physical delimited pathway (the information channel), e.g., a set of proteins interconnected or working along scaffold proteins (Weng et al., 1999). In the second instance, there is the diffusion of signals in the cytoplasm (possibly along preferential diffusion pathways) to reach the proper target, where the signal is recognized (bound) by the target molecule, and the message is decoded (Fig. 2). Although VT is very likely important both for communication between different compartments and within compartments, WT is mainly used for communication inside a compartment. It should also be noticed that to get WT, besides molecular scaffolds, the cell can use other specific nonleaking routes such as the directional transport of signaling proteins via filament-bound motor proteins or via the vesicular transport.
The modular nature of the signaling pathways and the process of protein/protein interactions (Edwards and Scott 2000; Fraser et al., 2000, Grant and Blackstock, 2001; Marinissen and Gurkind, 2001; Pawson et al., 2001) lead to “wired molecular circuits” that can be traced from the plasma membrane compartment, through the cytoplasm, toward the nuclear compartment. Receptor mosaics represent not simple inputs to these “vertical molecular circuits” but rather “horizontal molecular circuits” capable of integrating different signals already at the plasma membrane level (Fig. 3). In fact, due to the spatial/temporal contiguity of the transductional processes not a single (elementary) response but a complex activation of several intracellular signaling pathways occurs. These pathways cross-talk, in a controlled fashion, at several levels. This leads to biochemical/ionic responses that are the result of integrations at several crucial levels (“nodal points” of the signaling pathways) and thus to a “syndromic response” by the cell (Figs. 2 and 4). It should be noticed that, according to this view, the receptor mosaic has to be considered as a nodal point of signaling pathways at the membrane level.
As pointed out above, intracellular signaling pathways can use both WT and VT. In addition, cross-talks among signaling pathways are made possible by both WT and VT. In fact, there are nodal points where VT messages can enter a “wired signaling pathway” as well as there are nodal points where several wired signaling pathways converge. Via this cross-talk among intracellular signaling pathways, integrative units can be formed that very easily can be assembled and disassembled, according to the needs of the cell.
Thus, both WT and VT are used to create intracellular molecular circuits. For example, the MAPK cascade is mainly WT. However, the terminal MAPK, once activated, migrates (VT) into the nucleus where it phosphorylates and in this way activates transcription factors.
There are some potential arrangements of high functional significance to be considered. In fact, it can be hypothesized that the following instances are all theoretically possible:
-
The WT signaling pathways are completely precabled and they need only to be activated at some nodal point (e.g., at the plasma membrane receptor site) to transmit (and/or elaborate) information.
-
The WT signaling pathways are partially precabled and assembled when needed.
-
The WT signaling pathways are to be entirely cabled.
The same logical scheme may hold for receptor mosaics. In fact, as discussed above, the following can be surmised.
-
The receptor mosaic is already present at the plasma membrane level.
-
The receptor mosaic is assembled from homo-heterodimers or simple oligomers, when needed.
These concepts may allow tackling, in a new way, some problems concerning the signaling processes within a cell and the generation of the cell response. In fact, by applying the concept of WT and VT and taking into proper consideration the hypothesis on the informational role of the receptor mosaic, it is possible to explain how the signaling pathways can operate with a limited number of components as follows:
-
with selectivity of the signaling pathways, notwithstanding the fact that common signals and common protein modules are used in the various signaling pathways; and
-
with a complexity of the biological response in spite of the activation of a single signaling pathway. In fact, as pointed out for neural networks, it is our opinion that it would be better to speak of “syndromic responses” rather than of a single response (elementary responses).
-
In this way we can also understand the extraordinary capability of the cell to adapt to the highly variable challenging conditions in which the cell must live and operate in a finalistic fashion.
As shown in Fig. 1, the adaptation of the cell to challenges starts at the level of the membrane where the receptors can modify their conformation in response to the chemical-physical environments with which they are in contact. The binding of the neurotransmitter to its receptor (or of the neurotransmitters to the various types of receptors) will trigger the activation or modulation not only along one single signaling pathway but along several pathways. In this way the intracellular machinery is tuned in a form that can be described not with a “branch” (i.e., a single signaling pathway) but with a “tree” (see Fig. 2) where some main branches can be recognized but where the entire tree is important for the biological response of the cell (syndromic response). Selectivity is obtained since WT allows the direct and “private” link to the input site (e.g., the receptor) with the target (e.g., an enzyme).
It should be noted that in a nodal point (Fig. 4), it is possible to have summations as well subtractions of WT and VT signals. Thus, the output of the nodal point is the result of such an integration. Furthermore, it should be noted that the integration in the nodal point follows a temporal code that depends on the timing of the summations/subtractions among the different signals, since for each of these signals the progression toward the nodal point is different depending on the length and speed of its transmission along the respective signaling pathway.
A peculiar aspect of this process of integration at the nodal points is its dependence on the assembly/disassembly of the single modules (proteins) of the wire that forms the WT signaling pathway in question or on the characteristic of the diffusion of signals in the case of VT signaling pathways. In the case of WT, sometimes pieces of the wire can continue to operate: see, e.g., Raf-MEKMAPK that goes on to operate also after Raf dissociates from Ras-GDP (Egan and Weinberg, 1993).
This possibility of assembling prewired segments of the various signaling pathways could, at least in part, explain the extraordinary capability of the cell to promptly adapt its responses to highly variable challenging conditions.
D. Receptor/Receptor Interactions in the Striatopallidal GABA Neurons: Implications for Parkinson's Disease, Schizophrenia, and Drug Addiction
1. Localization of Adenosine A2A, Dopamine D2, and Glutamate Metabotropic mGluR5 Receptors in the GABA Striatopallidal Neuron.
The striatum is the main input structure of the basal ganglia and a key component of the motor and limbic systems. On the basis of its afferent and efferent connections the striatum is currently subdivided into two parts, the dorsal and the ventral striatum. Accordingly, the dorsal striatum is mainly represented by the dorsolateral part of the nucleus caudate putamen and the ventral striatum is made of the ventromedial caudate putamen, the nucleus accumbens (with its two compartments, shell and core) and the olfactory tubercle (Heimer et al., 1995; Gerfen and Wilson, 1996). The basic intrastriatal circuitry is quite simple, only one type of neuron, the projecting GABAergic neuron, constitutes more that 90% of the striatal neuronal population. The GABAergic efferent neurons mainly convey information carried by dopaminergic mesencephalic cells, which are located in the substantia nigra and ventral tegmental area, and by glutamatergic cells located in cortical and limbic areas, like the amygdala and hippocampus (Heimer et al., 1995; Gerfen and Wilson, 1996) (Fig. 5).
There are two subtypes of striatal GABAergic efferent neurons that give rise to the two dorsal striatal efferent systems, which connect the dorsal striatum with the output structures of the basal ganglia, the substantia nigra pars reticulata, and the entopeduncular nucleus (Alexander and Crutcher, 1990; Heimer et al., 1995; Gerfen and Wilson, 1996). These are called direct and indirect pathways. The direct pathway is made of striatonigral and striatoentopeduncular neurons. The indirect pathway consists of the striatopallidal GABAergic neurons, pallido-subthalamic GABAergic neurons, and glutamatergic neurons, which connect the subthalamic nucleus with the output structures. Pallidal GABAergic neurons also project directly to the output structures without using the subthalamic nucleus relay (Fig. 5). The striatopallidal GABAergic neurons contain the peptide enkephalin and dopamine receptors predominantly of the D2 subtype. On the other hand, the striatonigral and striatoentopeduncular GABAergic neurons contain the peptides dynorphin and substance P and dopamine receptors predominantly of the D1 subtype. Stimulation of the direct pathway results in motor activation and stimulation of the indirect pathway produces motor inhibition. Dopamine, or dopamine agonists, will induce motor activation by activating the direct pathway (acting on stimulatory D1 receptors) and by depressing the indirect pathway (acting on inhibitory D2 receptors) (Alexander and Crutcher, 1990; Gerfen and Wilson, 1996). In Parkinson's disease a preferential degeneration of the nigrostriatal dopaminergic system produces striatal dopamine depletion with the consequent impairment of the functioning of these circuits, which is associated with hypokinesia. Hyperactivity of the GABAergic striatopallidal neurons due to the release from the strong D2 receptor-mediated tonic inhibitory effects of endogenous dopamine is probably the main pathophysiological mechanism responsible for this hypokinesia (Obeso et al., 2000). The consequent increased neuronal activity of the subthalamic nucleus and the output structures of the basal ganglia (in particular the internal segment of the globus pallidus, which is the counterpart of the rat entopeduncular nucleus in primates) seems to be the functional hallmark of the parkinsonian state, that gives the basis of the surgical treatment in this disease (lesion of the subthalamic nucleus or pallidotomy) (Obeso et al., 2000).
The same two subtypes of GABAergic neurons, with the same segregation of dopamine receptor subtypes are also found in the ventral striatum (Le Moine and Bloch, 1995). However, the organization of the targets of the ventral striatum is different from its dorsal counterpart (Heimer et al., 1995). Although the ventral striatum sends projections to the pallidal complex (ventral pallidum), the entopeduncular nucleus and the substantia nigra-ventral tegmental area, it also sends projections to the extended amygdala, the lateral hypothalamus and the lateral mesopontine tegmental nucleus. The ventral GABAergic striatopallidal neuron plays a key role in the conversion of motivation into action via transfer of information from the limbic to the motor system (for review, see Ferré, 1997). In fact, it is a final common pathway for opiate- and psychostimulant-mediated reward (Koob, 1999) and a common target for antipsychotic drugs (Ferré, 1997). Thus, blockade of D2 receptors in the ventral GABAergic striatopallidal neuron seems to be involved with the antipsychotic effects of neuroleptics or, at least, with their therapeutic effect on the positive symptoms of schizophrenia. On the other hand, blockade of D2 receptors in the dorsal GABAergic striatopallidal neuron is associated with the extrapyramidal side effects of neuroleptics, such as parkinsonism. One of the possible properties of atypical neuroleptics, which have a decreased liability of those side effects, is a preferential D2 receptor blockade in the ventral striatum (reviewed in Ferré, 1997).
In addition to dopamine and glutamate, which are believed to have opposite actions on the striatal GABAergic efferent neurons (Carlsson and Carlsson, 1990), adenosine is a very important modulator (Snyder 1985) that plays an important integrative role in the function of these neurons (Ferré et al., 1997; Fuxe et al., 1998) (Fig. 6). Extracellular adenosine depends mostly on the intracellular concentration of adenosine and nucleotides, such as ATP, AMP, and cAMP (Latini and Pedata, 2001). In some brain areas, like the hippocampus, most of the extracellular adenosine seems to depend mostly on intracellular adenosine, the concentration of which depends on the rate of breakdown and synthesis of ATP (Latini and Pedata, 2001). Thus, in this case, different from what happens with classical neurotransmitters, which are mostly released by the nerve terminals in response to the arrival of the impulse flow, adenosine is released as a neuromodulator (Snyder, 1985) by the effector cells in response to an increased metabolic demand (Ferré and Fuxe, 2000). However, in the striatum, it has recently been suggested that the main source of extracellular adenosine is extracellular cAMP (Manzoni et al., 1998), which is metabolized to AMP by means of phosphodiesterases and then to adenosine by the ectoenzyme 5′-nucleotidase. Since cAMP can only be generated intracellularly by the action of the enzyme adenylyl cyclase, striatal extracellular adenosine would mostly reflect an increased activation of receptors positively linked to adenylyl cyclase. The actions of Adenosine are mediated by specific G protein-coupled receptors. From the four subtypes of adenosine receptors so far identified (A1, A2A, A2B, and A3 receptors) A1 receptors and, particularly, A2A receptors are specially concentrated in the striatum (Fredholm, 1995a,1995b; Svenningsson et al., 1999b; Fredholm et al., 2001). A2A receptors are mostly localized in both dorsal and ventral GABAergic striatopallidal neurons, colocalized with dopamine D2 receptors (Schiffmann et al., 1991, 1993; Fink et al., 1992; Svenningsson et al., 1997, 1999b). On the other hand, adenosine A1 receptors are localized in all the striatal neuronal elements and dopamine D1 receptors in the GABAergic striatonigral-striatoentopeduncular neurons are colocalized with adenosine receptors of the A1 subtype (Ferré et al., 1996).
An important amount of evidence exists for the existence of specific antagonistic A2A/D2 receptor interactions that modulate the function of GABAergic striatopallidal neurons (Ferré et al., 1993a, 1997; Fuxe et al., 1998). Furthermore, specific antagonistic A1/D1 receptor interactions modulate the function of GABAergic striatonigral-striatoentopeduncular neurons (Ferré et al., 1997; Fuxe et al., 1998). It has been suggested that these receptor/receptor interactions are involved in the motor effects of adenosine agonists and antagonists (like the nonselective adenosine antagonist caffeine) and that they can provide a new therapeutic approach for the treatment of basal ganglia disorders and schizophrenia (Ferré et al., 1992, 1994a,b, 1997, 2001; Ferré, 1997; Lillrank et al., 1999; Fuxe et al., 2001). Here we will review the more relevant findings on the A2A/D2 receptor interactions as well as the recent data on the multiple interactions between these receptors and the group I metabotropic glutamate receptor mGluR5.
Glutamate acts on both ionotropic and metabotropic G protein-coupled receptors. Molecular and pharmacological characterization studies have currently divided the metabotropic glutamate receptor family into three groups (I to III) (Pin and Duvoisin, 1995; Bockhaert and Pin, 1999; Hermans and Challiss, 2001). Group I includes mGluR1 and mGluR5 receptors, the latter of which is particularly expressed in the striatum, especially in the striatal GABAergic efferent neurons (Shigemoto et al., 1993; Testa et al., 1995; Tallaksen-Green et al., 1998). The ultrastructural analysis of the localization of mGluR5 receptors in the striatopallidal complex in primates demonstrated that mGluR5 receptor immunoreactivity is commonly found perisynaptically to asymmetric (glutamatergic) postsynaptic synapses (Fig. 6) but also at symmetric synapses formed by tyrosine hydroxylase-immunoreactive (dopaminergic) terminals (Smith et al., 2000). The recent ultrastructural analysis of the localization of A2A receptors in the rat striatum showed a very similar localization to that described for mGluR5 receptors in primates. Thus, A2A receptors are mostly localized postsynaptically in the dendrites and dendritic spines of the striatal GABAergic neurons. Importantly, A2A receptor-immunoreactivity was observed primarily at asymmetric synapses (Rosin et al., 1998; Hettinger et al., 2001). Although less frequently A2A receptor immunoreactivity was also found presynaptically at asymmetric synapses and postsynaptically at symmetric synapses (Hettinger et al., 2001). Finally, in the rat striatum D2 receptor immunoreactivity has been found both presynaptically (in both dopaminergic and glutamatergic terminals) and postsynaptically. At the postsynaptic level, D2 receptor immunoreactivity has not only been observed at symmetric synapses but also perisynaptically at asymmetric synapses (Yung et al., 1995; Delle Donne et al., 1997). In summary, these morphological findings strongly suggest that the three receptors, A2A, mGluR5, and D2, are colocalized in the striatopallidal GABAergic efferent neuron, postsynaptically and perisynaptically at dopaminergic and glutamatergic synapses (Fig. 6). Dopamine and glutamate may be able to reach this localization, not only by spillover from the respective nerve terminals, but also by longer distance VT. Finally, the colocalization gives a morphological frame for the existence of multiple A2A/D2/mGluR5 receptor interactions.
2. Interactions between Adenosine A2A, Dopamine D2, and Glutamate Metabotropic mGluR5 Receptors in the GABA Striatopallidal Neuron: Biochemical-Cellular Level.
As previously discussed, evidence for the existence of physical interactions between A2A and D2 receptors and between A2A and mGluR5 receptors has been obtained with coimmunprecipitation experiments in cell lines that express, constitutively or after stable or transient cotransfection, the corresponding receptors (Ferré et al., 2002; Hillion et al., 2002). A2A/mGluR5 heteromeric complexes have also been demonstrated in rat striatal membrane preparations with coimmunoprecipitation experiments (Ferré et al., 2002). Different behavioral and biochemical models have demonstrated the existence of functionally significant antagonistic A2A/D2 receptor and mGluR5/D2 receptor interactions and synergistic A2A/mGluR5 receptor interactions. At the biochemical level, by using membrane preparations of rat striatum or of cell lines expressing the corresponding receptors and by using receptor autoradiography in rat and human striatal sections, it has been shown that stimulation of A2A receptors decreases the affinity of D2 receptors for dopamine or dopamine agonists. These results were shown, e.g., as an increase in KH or IC50 in competition experiments of dopamine versus a radioactively labeled D2 receptor antagonist or as an increase in KD in saturation experiments with tritiated dopamine or quinpirole (Ferré et al., 1991d, 1994a,b, 1999a; Dasgupta et al., 1996a; Dixon et al., 1997; Lepiku et al., 1997; Kull et al., 1999; Franco et al., 2000; Salim et al., 2000; Díaz-Cabiale et al., 2001). In receptor autoradiography experiments, the A2A receptor-mediated modulation of D2 receptor binding properties was found to be stronger in the ventral compared with the dorsal striatum (Ferré et al., 1994b; Díaz-Cabiale et al., 2001). Dopamine receptors of the D3 subtype are structurally and pharmacologically very similar to D2R and they are specially concentrated in the ventral striatum (Missale et al., 1998). Therefore, it was suggested that the stronger A2A/D2 receptor interaction in the ventral striatum might be related to the action of A2A receptor agonists on D3 receptors, in addition to D2 receptors, as suggested by the demonstration of the A2A receptor agonist-mediated modulation of D3 receptor agonist binding in vivo (Hillefors et al., 1999). Also, in membrane preparations from rat striatum, nonselective stimulation of group I glutamate metabotropic receptors or selective stimulation of mGluR5 receptors decreased the affinity of D2 receptors for dopamine (Ferré et al., 1999a; Rimondini et al., 1999, Popoli et al., 2001). Finally, costimulation of A2A receptors and group I glutamate metabotropic receptors or mGluR5 receptors exerted a synergistic effect on the modulation of the binding characteristics of D2 receptors (Ferré et al., 1999a; Rimondini et al., 1999; Popoli et al., 2001).
In addition to the intramembrane receptor/receptor interactions, the existence of A2A/D2 and A2A/mGluR5 heteromeric receptor complexes provides the possibility for close cross-talk between A2A, D2, and mGluR5 receptors. This would synchronize the activation of these receptors to interact at the second-messenger level and beyond. The major signal transduction pathway used by A2A receptors depends on the activation of adenylyl cyclase, by means of Gs/Golf coupling (Kull et al., 1999, 2000). A2A receptor-mediated adenylyl cyclase activation generates cAMP, which activates a cAMP-dependent protein kinase (PKA), which in turn regulates the state of phosphorylation of various substrate proteins (Fig. 7). One of those proteins, DARPP-32 (dopamine and cyclic adenosine 3′,5′-monophosphate-regulated phophoprotein, 32 kDa) is expressed in very high concentration in the GABAergic efferent neurons (Greengard et al., 1998, 1999). PKA-induced phosphorylation at a single site (Thr-34 of the rat sequence) converts DARPP-32 into a potent and selective inhibitor of protein phosphatase-1 (PP-1) (Greengard et al., 1999). Furthermore, DARPP-32 is also phosphorylated to a high stoichiometry under basal conditions at another site, Thr-75, which converts it into an inhibitor of PKA (Nishi et al., 2000). PKA stimulates protein phosphatase-2A, which is the predominant phosphatase responsible for dephosphorylation of phospho-Thr-75-DARPP-32 (Nishi et al., 2000). It has been demonstrated that the removal of the inhibitory constraint on PKA by dephosphorylation of phospho-Thr-75-DARPP-32 provides a mechanism of amplification of the PKA signal transduction pathway (Nishi et al., 2000). In striatal slices, Fisone and coworkers have demonstrated that stimulation of A2A receptors produces phosphorylation of DARPP-32 at Thr-34 and dephosphorylation of DARPP-32 at Thr-75 (Svenningsson et al., 1998, 2000; Lindskog et al., 1999, 2002). Another protein phosophorylated by PKA is the constitutive transcription factor CREB (cAMP response element-binding protein). Induction of cAMP liberates the catalytic subunits of PKA, which diffuse into the nucleus and induce cellular gene expression by phosphorylating CREB at a serine residue (Ser-133) (Mayr and Montminy, 2001). CREB activity declines after a couple of hours of continuous stimulation, due to dephosphorylation at Ser-133 by PP-1 (Mayr and Montminy, 2001). Thus, through PKA activation A2AR stimulation in the GABAergic striatopallidal neurons can potentially produce a sustained increase in the transcription of some CREB-modulated genes by a mechanism involving increased CREB phosphorylation and decreased CREB dephosphorylation (through phospho-Thr-34-DARPP-32-mediated inhibition of PP-1 activity) (Fig. 7). The immediate-early gene c-fos and the preproenkephalin and neurotensin genes are very well studied target genes, whose promoters contain consensus sites for CREB-P binding (Borsook and Hyman, 1995; Evers et al., 1995; Hughes and Dragunow, 1995; Le et al., 1997; Herdegen and Leah, 1998). Recent studies have shown that A2A receptor stimulation can, under certain conditions, increase the expression of c-fos (and other immediate-early genes), preproenkephalin, and neurotensin genes (Fig. 7) (see below).
One of the main fully documented signaling effects of D2 receptors is adenylyl cyclase inhibition by means of its coupling to Gi/o proteins (for review, see Missale et al., 1998). Also dependent on Gi/o coupling, but independent of cAMP, D2 receptors have been shown to modulate the activity of K+ channels. In many preparations, including acutely dissociated striatal neurons (Freedman and Weight, 1988), D2 receptor stimulation has been shown to increase outward K+ currents, leading to cell hyperpolarization (Missale et al., 1998). Furthermore, other less well characterized cAMP-independent signaling pathways used by D2 receptors, such as activation of PLC, inhibition of inward Ca2+ currents, arachidonic acid release, inhibition of Na+-K+-ATPase and MAPK activation, have been reported (Missale et al., 1998; Yan et al., 1999). A strong antagonistic interaction between A2A receptors and D2 receptors at the adenylyl cyclase level has been demonstrated in different cell lines (Kull et al., 1999; Hillion et al., 2002) (Fig. 7). In CHO cells stably cotransfected with A2A and D2 receptor cDNAs, stimulation of A2A receptors produced a strong stimulation of cAMP, CREB phosphorylation and increased c-fos expression (Kull et al., 1999). A selective D2R agonist dose dependently counteracted these effects and a complete blockade was attained at low concentrations of the D2R agonist (Kull et al., 1999). Similarly, in striatal slices a D2 receptor agonist completely counteracted Thr-34 phosphorylation of DARPP-32 by A2A receptor stimulation (Lindskog et al., 1999).
In the striatum, D2 receptors are tonically stimulated by basal endogenous levels of dopamine (see above). Many experimental results indicate that this tonic D2 receptor stimulation strongly counteracts a tonic A2A receptor stimulation induced by the basal striatal levels of adenosine. In this way, the products of A2A receptor activation are kept at low concentration under normal basal conditions and inactivation of striatal D2 receptor-mediated neurotransmission (by administration of D2 receptor antagonists, striatal dopamine depletion or genetic D2 receptor inactivation) liberates A2A receptor-mediated function from the strong D2 receptor-mediated tonic inhibition. This results in a very significant increase in the striatal expression of c-Fos, Thr-34-phosphorylated DARPP-32, enkephalin, and neurotensin, which is partially counteracted by genetic inactivation or pharmacological blockade of A2A receptors (Schiffman and Vanderhaeghen, 1993; Morelli et al., 1995; Pollack and Fink, 1995; Boegman and Vincent, 1996; Adams et al., 1997; Pinna et al., 1997, 1999; Richardson et al., 1997; Svenningsson et al., 1999a, 2000; Ward and Dorsa, 1999; Chen et al., 2000, 2001; Zahniser et al., 2000; Dassesse et al., 2001; Ferré et al.,. 2002;, see Section I.B.). Since also at the behavioral level A2A receptor antagonists counteract the motor depressant and cataleptic effects secondary to the genetic inactivation or pharmacological interruption of D2 receptor-mediated neurotransmission (see below), these results strongly suggest that an important part of the biochemical and behavioral effects induced by interruption of D2 receptor-mediated neurotransmission are due to the liberation of A2A receptor signaling. This can have obvious implications for the possible application of A2A receptor antagonists in Parkinson's disease. On the other hand, these results also suggest that the biochemical effects related to adenylyl cyclase activation (which are only clearly apparent when the D2 receptor-mediated inhibitory tone is removed) do not play a major role in some functional and behavioral effects produced by the administration of A2A receptor agonists and antagonists (see below). For instance, the systemic administration of a low dose of a selective A2A receptor agonist produces a pronounced motor depression (already shown to be centrally mediated; see Barraco et al., 1993) and selectively counteracts D2 receptor agonist-mediated behaviors (Rimondini et al., 1997, 1998) without inducing an increased c-fos striatal expression (Morelli et al., 1995; Pinna et al., 1997; Ferré et al., 2002; see Section I.B.2.k.). Most probably those behavioral effects are related to the reciprocal antagonistic A2A/D2 receptor intramembrane interaction, which can be more effective at modulating other D2 receptor-mediated signaling pathways, such as the opening of K+ channels.
As previously mentioned, some studies suggest that a preferential blockade of D2 receptors in the ventral, compared with the dorsal striatum, is one of the main factors responsible for the atypical profile of an antipsychotic. Some of these studies are based on the differential increase in c-fos expression in the different striatal compartments. Thus, typical antipsychotics, like haloperidol, increase c-fos expression in the dorsal and ventral striatum, whereas atypical antipsychotics, like clozapine, selectively elevates c-fos expression in the ventral striatum, and especially in the shell of the nucleus accumbens (Deutch et al., 1992; Robertson and Fibiger, 1992; Merchant and Dorsa, 1993; Pinna et al., 1999). A similar differential response to haloperidol and clozapine has also been demonstrated for the striatal expression of neurotensin (Merchant and Dorsa, 1993). It has also been shown that the increased c-fos expression in the ventral striatum induced by both haloperidol and clozapine selectively takes place in the ventral GABAergic striatopallidal neuron (Robertson and Jian, 1995) and that it can be counteracted by the systemic administration of an A2A receptor antagonist (Pinna et al., 1999). Therefore, these results suggest that A2A receptors might be involved in the mediation of the antipsychotic effects of neuroleptics (see below).
It remains to be determined which are the physiological conditions (without interruption of D2 receptor-mediated neurotransmission) that would allow A2A receptor stimulation to produce a clear adenylyl cyclase activation, with the corresponding increase in Thr-34 DARPP-32 and CREB phosphorylation and increase in the expression of c-fos and the preproenkephalin and neurotensin genes. One such condition seems to be mGluR5 receptor coactivation (Ferré et al.,. 2002). The major signal transduction pathway used by mGluR5 receptors (through Gq proteins) is activation of PLC, which releases IP3 and diacylglycerol, which cause the release of intracellular Ca2+ and the activation of PKC, respectively (Pin and Duvoisin, 1995; Hermans and Challis, 2001) (Fig. 7). In HEK-293 cells transiently cotransfected with A2A and mGluR5 receptors, stimulation of A2A receptors produced an increase in cAMP levels that was not significantly modified by the concomitant stimulation of mGluR5 receptors (Ferré et al., 2002). As expected, stimulation of mGluR5 receptors induced an increase in [Ca2+]i. In addition to mGluR5 receptor stimulation, a selective A2A receptor agonist also produced a significant increase in [Ca2+]i in cotransfected cells, which would agree with recent studies in rat striatal slices which suggest that A2A receptors can also use PLC/IP3 signaling under certain conditions (Wirkner et al., 2000). However, at the level of [Ca2+]i an absence of synergistic A2A/mGluR5 receptor interaction was observed (Ferré et al.,. 2002). The lack of synergistic interaction at the second messenger level (cAMP and Ca2+) was somehow surprising in view of previous studies showing synergistic effects on cAMP accumulation of group I glutamate metabotropic receptors and receptors positively linked to adenylyl cyclase in neuronal primary cultures (Cartmell et al., 1998; Paolillo et al., 1998) and in view of recent findings on synergistic effects between A1 receptors and mGluR1α receptors on [Ca2+]i in cotransfected HEK-293 cells (Ciruela et al., 2001). Nevertheless, a strong functional synergistic A2A/mGluR5 receptor interaction on MAPK (ERK 1/2) and on c-fos expression was found in cotransfected cells (Ferré et al.,. 2002). Accordingly, a significant striatal c-fos induction was obtained after the concomitant central administration of A2A and mGluR5 receptor agonists, which were ineffective when administered alone (Ferré et al., 2002). This strongly suggests that concomitant stimulation of A2A and mGluR5 receptors is one of the mechanisms by which A2A receptor stimulation can overcome the tonic inhibitory effect of dopamine and induce striatal c-fos expression. We favor the idea that this mechanism can take place under conditions of intense glutamatergic neurotransmission, which is known to induce adenosine release, most probably, due to the neuronal metabolic demand imposed by the increased excitatory input (Ferré and Fuxe, 2000; Latini and Pedata, 2001).
The protein encoded by c-fos (c-Fos) is an inducible transcription factor, expression of which is controlled by pre-existing constitutive transcription factors, such as CREB, serum response factor, and TCF/Elk-1 proteins (Herdegen and Leah, 1998; Ches and Wang, 2001). Upon phosphorylation, CREB and TCF/Elk-1 (which requires serum response factor) activate c-fos transcription by binding to the CRE and serum response element regulatory elements of the c-fos promoter, respectively (Hughes and Dragunow, 1995; Herdegen and Leah, 1998) (Fig. 7). It is generally accepted that CREB is mostly phosphorylated by PKA (see above) whereas TCF/Elk-1 is mostly phosphorylated by MAPK (Vanhoutte et al., 1999). In the case of group I metabotropic receptors, MAPK activation seems to involve PKC (Ferraguti et al., 1999) and Ca2+/calmodulin-dependent protein kinases (CaMK) (Choe and Wang, 2001). However, recent studies indicate that both MAPK and CREB can be convergent targets for different elements of the cAMP/PKA and the PLC/PKC-CaMK signaling pathways (Fig. 7) (Impey et al., 1999; Sweatt, 2001; Wu et al., 2001). Since the A2A/mGluR5 receptor synergistic effect on c-fos expression found in cotranfected cells was completely counteracted by a ERK 1/2 kinase inhibitor (Ferré et al., 2002), MAPK seems to be a main biochemical integration element responsible for the synergistic interactions between A2A and mGluR5 receptors. Furthermore, other levels of interaction upstream to MAPK could be the nonreceptor tyrosine kinase Src or other enzymes recently demonstrated to be activated by G protein-coupled receptors through G protein-independent signaling (Heuss and Gerber, 2000). Given the key role of immediate-early genes in the coupling of early neuronal responses to long-term adaptive changes (Hughes and Dragunow, 1995; Berke and Hyman, 2000), those results suggest that A2A/mGluR5 receptor interactions might be involved in striatal neuronal plasticity. More specifically, this mechanism may underlie the recently described dopamine-independent increased c-fos expression in the striatopallidal neurons associated with sensitization to psychostimulants (Uslaner et al., 2001). In fact, both striatal A2A and mGluR5 receptors have recently been shown to be involved in certain forms of striatal plasticity, including cortico-striatal long-term potentiation and long-term depression (d'Alcantara et al., 2001; Sung et al., 2001).
3. Interactions between Adenosine A2A, Dopamine D2, and Glutamate Metabotropic mGluR5 Receptors in the GABA Striatopallidal Neuron: Physiological-Behavioral Level.
The first functional approach used to study the interactions between adenosine and dopamine receptors in the GABAergic striatopallidal neurons was the dual-probe in vivo microdialysis approach in freely moving rats. In these experiments, one microdialysis probe is implanted in the caudate-putamen or the nucleus accumbens, and the second one is implanted in the ipsilateral globus pallidus or the ipsilateral ventral pallidum, respectively. This allows determining in freely moving animals the modulation of GABAergic neurotransmission in the dorsal and ventral striatopallidal neurons (Ferré et al., 1993a, 1994a; Díaz-Cabiale et al., 2002b). With these experiments, a main difference in the effects of D2 receptor stimulation was found between the dorsal and the ventral striatopallidal pathways. The striatal perfusion of a D2 receptor agonist decreased the extracellular levels of GABA in the globus pallidus, but not in the ventral pallidum. On the other hand, D2 receptor antagonists increased GABA levels in the ventral pallidum and did not significantly modify GABA levels in the globus pallidus (Ferré et al., 1993a, 1994a; Díaz-Cabiale et al., 2002b). Also, a different effect was obtained with the striatal perfusion of an A2A receptor agonist. In this case, GABA levels were only increased in the ventral pallidum (Ferré et al., 1993a, 1994a; Díaz-Cabiale et al., 2002b). Nevertheless, recent studies by Ochi et al. (2000) have also found a significant increase in the GABA levels of the globus pallidus after the striatal infusion (through an injection needle) of an A2A receptor agonist. The reciprocal A2A/D2 receptor interactions could also be demonstrated in microdialysis experiments. Thus, the D2 receptor agonist-induced decrease and the A2A receptor agonist-induced increase in pallidal levels could be counteracted by the striatal coperfusion with A2A and D2 receptor agonists, respectively (Ferré et al., 1994a; Díaz-Cabiale et al., 2002b). Altogether, these results suggest the existence of functional regional differences in the A2A/D2 receptor interactions, with a relatively stronger A2A receptor signaling in the ventral compared with the dorsal striatum, in agreement with some biochemical (Ferré et al., 1994a; Pinna et al., 1997; Díaz-Cabiale et al., 2002b), and behavioral studies (see below). We have hypothesized that these differences in the A2A/D2 receptor interactions between both striatal compartments might explain the atypical antipsychotic profile of adenosine A2A receptor agonists in animal models (see below). As mentioned earlier, this differential effect may be related to the existence of an additional antagonistic A2A/D3 receptor interaction in the ventral striatum (Diaz-Cabiale et al., 2001). Finally, in recent microdialysis experiments we have found that the perfusion of a selective mGluR5 receptor agonist in the nucleus accumbens produces an increase in the extracellular levels of GABA in the ventral pallidum. Furthermore, coperfusion with an A2A receptor agonist produced a strong synergistic increase in pallidal GABA levels (Díaz-Cabiale et al., 2002b). From these results, we have suggested a possible use of mGluR5 receptor agonists as antipsychotic drugs, alone or in combination with A2A receptor agonists.
In vitro studies with micropunctures of rat globus pallidus have shown that an antagonistic A2A/D2 receptor interaction also modulates GABAergic neurotransmission in the terminals of GABAergic striatopallidal neurons, which also express both A2A and D2 receptors (Mayfield et al., 1996). In this experimental preparation, stimulation of pallidal A2A and D2 receptors stimulates and inhibits GABA release, respectively, and the D2 receptor-mediated inhibition of GABA release is counteracted by A2A receptor stimulation (Mayfield et al., 1996). The A2A receptor-mediated stimulation of pallidal GABA release has been recently confirmed with in vivo microdialysis experiments (Ochi et al., 2000). However, experiments using striatal and pallidal synaptosomal preparations have given contradictory results. Thus, in these preparations, A2A receptor stimulation inhibited high KCl-stimulated GABA release (Kurokawa et al., 1994). Similarly, results obtained from experiments using intracellular and whole-cell patch-clamp recording in striatal slices suggested that striatal presynaptic A2A receptors exert an inhibitory modulation of GABA release (Mori et al., 1996). Based on these findings, Richardson et al. (1997) proposed that the main mechanisms by which A2A receptors would influence the function of striatal GABAergic neurons would be by a presynaptic inhibitory modulation of GABA release from their collateral recurrent axons. At this point, this theory is incompatible with the bulk of new information about the ultrastructural localization and function of striatal A2A receptors (see above).
Many behavioral studies have shown that in animals with intact striatal dopamine innervation, A2A receptor agonists behave as D2 receptor antagonists (Heffner et al., 1989; Ferré et al., 1991a,b; Popoli et al., 1994; Kafka and Corbett, 1996; Hauber and Munkle, 1997; Rimondini et al., 1997, 1998; Wardas et al., 1999; Poleszak and Malec, 2000). Furthermore, although questioned by some studies (Kafka and Corbett, 1996; Hauber and Munkle, 1997), there are some behavioral data supporting a preferential effect of A2A receptor agonists in the ventral striatum (Barraco et al., 1993; Rimondini et al., 1997, 1998). When considering commonly used tests to screen the effects of antipsychotics, it is generally accepted that blockade of dopamine receptors in the ventral striatum is mostly responsible for the counteracting effects of the motor activation induced by novel stimuli (exploratory activity in a novel environment) and psychostimulants [such as amphetamine and phencyclidine (PCP)]. On the other hand, the counteraction of dopamine agonist-induced stereotypes and the induction of cataleptic immobility are believed to be mainly mediated by blockade of dopamine receptors in the dorsal striatum (reviewed in Ögren, 1996). The systemic administration of an A2A receptor agonist gave an atypical antipsychotic profile in these animal models, since it counteracted the motor-activating effects of psychostimulants at lower doses than those needed to counteract dopamine agonist-induced stereotypes or to induce catalepsy (Rimondini et al., 1997, 1998). In particular, very low doses of the A2A receptor agonist were necessary to counteract PCP-induced motor activity (Rimondini et al., 1997). In agreement, recent studies by Sills et al. (2001) have shown than an A2A receptor agonist can selectively reverse the reduction in prepulse inhibition of the acoustic startle response induced by PCP.
PCP is self-administered by humans and experimental animals (Carlezon and Wise, 1996; Jentsch and Roth, 1999), and its administration in humans reproduces both positive and negative symptoms of schizophrenia and exacerbates those symptoms in schizophrenic patients (Jentsch and Roth, 1999). Together with the neurochemical modifications induced by this drug, PCP administration is considered as a potential animal model of schizophrenia, and reversal of the effects of PCP is also commonly used as a model to screen for antipsychotic activity (Jentsch and Roth, 1999). The ventral striatopallidal GABA neurons are most probably the main neuronal target underlying the A2A receptor agonist-mediated counteraction of PCP-induced motor activation, since inhibition of ventral striatopallidal GABA transmission appears to be central in the mediation of some (rewarding) effects of PCP (Carlezon and Wise, 1996). Since the counteracting effect on PCP-induced motor activation is obtained with low doses of the A2A receptor agonist, which are not able to induce an increase in the striatal c-fos expression (see above), the antagonistic A2A/D2 receptor intramembrane interaction is most probably involved. In fact, the motor hyperactivity induced by PCP has been shown to be dependent on postsynaptic D2 receptors and a high degree of D2 receptor blockade is required to significantly counteract the stimulatory action of high doses of PCP (Ögren and Goldstein, 1994). Furthermore, the effect of A2A receptor stimulation was surmountable, and high doses of PCP could not be counteracted by the systemic administration of an A2A receptor agonist (Ferré et al., 2002; see Section I.B.). However, in agreement with the results obtained with striatal c-fos expression, when the A2A receptor agonist was coadministered with a mGluR5 receptor agonist, which by itself did not produce any significant effect, PCP-mediated motor activation was significantly counteracted (Ferré et al., 2002). These results also fit very nicely with those obtained in microdialysis experiments (Díaz-Cabiale et al., 2002) and suggest that costimulation of A2A and mGluR5 receptors can even counteract the effects induced by a strong stimulation of dopaminergic neurotransmission.
As previously mentioned, the systemic administration of A2A receptor antagonists counteracts most of the biochemical (see above) as well as the motor depressant and cataleptic effects secondary to the genetic inactivation or pharmacological interruption of D2 receptor-mediated neurotransmission. This has been repeatedly demonstrated in a number of experimental models involving rodents pretreated with D2 receptor antagonists, reserpine, or MPTP or after genetic inactivation of D2 receptors (Casas et al., 1988; Popoli et al., 1991; Kanda et al., 1994; Giménez-Llort et al., 1995; Pollack and Fink, 1995; Malec, 1997; Shiozaki et al., 1999; Ward and Dorsa, 1999; Wardas et al., 1999; Aoyama et al., 2000; Chen et al., 2001; Ferré et al., 2001; Hauber et al., 2001) and involving MPTP-treated monkeys (Kanda et al., 1998a,b; Grondin et al., 1999). The results of these experiments suggest that A2A receptor antagonists can be useful in the treatment of Parkinson's disease. However, although it has been claimed that A2A receptor antagonists might be devoid of the secondary dyskinetic effects associated with treatment with dopamine agonists (Kanda et al., 1998a,b; Grondin et al., 1999), it is still a matter of debate whether they can be useful as mono-therapy or whether they would be more efficacious when combined with D2 receptor agonists. In fact, in another classical experimental model of Parkinson's disease, the rat with a unilateral lesion of the nigrostriatal pathway, A2A receptor antagonists do not produce a turning behavior contralateral to the lesioned side (Fenu and Morelli, 1998; StrÖmberg et al., 2000; Ferré et al., 2001), which is a behavior that predicts antiparkinsonian activity (Ungerstedt, 1971). Nevertheless, in all the animal models of Parkinson's disease tested to date (reserpinized mice, rats with unilateral 6-OH-dopamine lesions, MPTP-treated monkeys), A2A receptor antagonists strongly potentiate the motor activation induced by D2 receptor agonists (Ferré et al., 1991b, 2001; Jiang et al., 1993; Fenu et al., 1997; Grondin et al., 1999; Kanda et al., 2000; Koga et al., 2000; Popoli et al., 2000, 2001; StrÖmberg et al., 2000). Importantly, cotreatment with A2AR antagonists and L-dopa did not increase the non-wanted dyskinetic effects in MPTP-treated monkeys (Grondin et al., 1999; Kanda et al., 2000). Finally, recent electrophysiological experiments in the rat dopamine denervated striatum showed that the infusion of an A2A receptor antagonist did not produce any effect on its own, but strongly potentiated the D2R agonist-induced inhibition of striatal neuronal activity (StrÖmberg et al., 2000). Altogether, these results suggest that some pharmacological and, maybe, therapeutical effects of A2A receptor blockade can only be observed with the concomitant stimulation of D2 receptors. This is also in agreement with some results obtained with D2 receptor knockout mice where the motor effects of A2A receptor antagonists were attenuated (Chen et al., 2001). Thus, these experiments performed under complete inactivation of D2 receptors demonstrate that some A2A receptor functions are dependent on the integrity of D2 receptors and, most probably, on the integrity of A2A/D2 heteromeric receptor complexes. This is shown even more dramatically in recent studies by Zahniser et al. (2000), where a very significant functional uncoupling of A2A receptors (lack of A2A receptor agonist-induced GABA release in striatal/pallidal slices) was found in D2 receptor knockout mice. On the other hand, a functional striatal hypodopaminergic activity (decreased striatal dopamine release and decreased psychostimulant-induced motor activation) has been found in mice lacking A2A receptors (Chen et al., 2000; Dassese et al., 2001). In this case, the possible role of a functional uncoupling of D2 receptors remains to be determined.
Finally, the role of mGluR5 and its interactions with A2A and D2 receptors in animal models of Parkinson's disease is beginning to be evaluated. In unilaterally 6-OH-dopamine-lesioned rats, the intracerebral administration of a selective mGluR5 agonist selectively inhibited the contralateral turning induced by a D2 receptor agonist (Popoli et al., 2001). The effect of the mGluR5 agonist was potentiated by an A2A receptor agonist and attenuated by an A2A receptor antagonist (Popoli et al., 2001). These results suggest that a mGluR5 antagonist, alone or in combination with A2A receptor antagonists and/or D2 receptor agonists, might provide a new therapeutic approach for basal ganglia disorders, such as Parkinson's disease. In fact, recent studies by Ossowka et al. (2001) have found antiparkinsonian-like effects of mGluR5 antagonists in rats (Chase and Oh, 2000).
4. Interactions between Adenosine, Dopamine, and Glutamate Metabotropic Receptors in the GABAergic Striatoentopeduncular and Striatonigral Neurons.
In these GABAergic efferent neurons of the striatum, which constitute the direct striatal efferent pathway (see above), evidence indicates that there might also exist multiple interactions between adenosine, dopamine, and group I metabotropic glutamate receptors. In this case, however, an antagonistic interaction between A1 and D1 receptors is involved (see above; Ferré et al., 1994b, 1996a,b, 1997, 1999b; Fuxe et al., 1998), which form heteromeric complexes (see above; Ginés et al., 2000). Although it remains to be demonstrated, most probably mGluR1, instead of mGluR5, receptors functionally interact with A1 and D1 receptors in these neurons, because of the demonstrated mGluR1α/A1 heteromeric receptor complexes (see above; Ciruela et al., 2001). Furthermore, the demonstrated synergistic interactions in mGluR1α/A1 heteromeric receptor complexes in cell lines with regard to agonist-induced increases in Ca2+ signaling (Ciruela et al., 2001) are of interest. In view of the role of D1 receptors in favoring motor initiation by excitatory effects on the direct pathway, the A1/D1/mGluR1 receptor interaction might also have implications for the treatment of Parkinson's disease. Furthermore, A1/D1 receptor interactions can also have implications for schizophrenia and drug addiction, since similar to the A2A/D2 receptor interactions (see above), A1/D1 receptor interactions are stronger in the ventral compared with the dorsal striatum (Ferré et al., 1996b, 1999b; Mayfield et al., 1999).
III. Implications of Receptor/Receptor Interactions for Drug Development
A. The Ground for Novel Therapeutical Interventions
As pointed out above, receptor/receptor interactions are one type of protein/protein interactions, and they can occur in the context of various protein/protein interactions, i.e., inside of an aggregate of several proteins forming molecular circuits. This view is in agreement with the experimental evidence that, in almost all cases, proteins do not work alone but rather as part of larger complexes. The study of protein expression and interactions is undertaken by “proteomics”. Proteomics has been defined as the large-scale study of proteins encoded by a genome (Banks et al., 2000; Grant and Blackstock, 2001; Husi and Grant, 2001). Grant and colleagues have subdivided proteomics into “expression proteomics” and “functional (or interaction) proteomics” (Grant and Blackstock, 2001). Functional proteomics should analyze how proteins interact to form cellular machines. Thus, receptor/receptor interactions as well as the concept of cellular wiring (Pawson et al., 2001) made with protein modules (Pawson and Scott, 1997) can be part of the vast field of “functional proteomics”.
This new way of looking at receptor activation and intracellular signaling pathways opens up the possibility of discovering molecular, physiological and pathological mechanisms until now unknown and also opens up an entire unexplored field for the development of drugs that should be aimed to specifically target protein/protein complex formation or to modulate the function of these protein complexes (Bond and Bouvier, 1998; Cochran, 2000). This field is just at its beginning (Tallman, 2000), and hence only the description of some of the most promising results until now obtained will be given.
It is convenient to subdivide the presentation as follows:
-
. Analysis of drug action on protein/protein interactions
-
. Possible targets for drugs acting on the heteromeric receptor complexes
B. Theoretical Strategies to Target Receptor Complexes
From a theoretical standpoint at least three strategies can be employed to pharmacologically affect protein/protein interactions:
-
To prevent the interaction by altering at least one of the two interacting protein interfaces (e.g., by a drug that localizes at the interfaces preventing the matching of the interface)
-
To address the protein/protein interaction toward the formation of a different (inactive) complex (e.g., by favoring the matching between two different interfaces of the two partners)
-
To favor the interaction of one of the two proteins toward the formation of a complex with another protein present in that compartment.
It should be underlined that to act in a predictable way on protein/protein interactions is a difficult task. Protein/protein interactions very often include discontinuous parts of the protein sequence. Thus, it is difficult to develop low molecular weight molecules capable of disrupting the protein interfaces that match together (Cochran, 2000). This difficulty seems to preclude the possibility to develop small molecules as pharmaceutical agents, but it should be kept in mind that low molecular weight is one of the features that determines better bioavailability.
The most straightforward strategy to develop a molecule that antagonizes a protein/protein interaction is that of reproducing the essential features of one of the two partner proteins in a smaller protein that, therefore, interferes with the complex formation. Some good results have been obtained in the case of interleukin-4 (Domingues et al., 1999). However, proteins even of a reduced size are still too large to be currently used as drugs. Thus, it is important to address the efforts toward using small peptides or, even better, peptidomimetic molecules as drugs (Cochran, 2000).
In a few cases it has been claimed that short peptides (10-20 amino acids), taken from a protein sequence, can disrupt the protein/protein interaction in which the parent protein is involved. However, according to Cochran these reports should be viewed with caution and carefully confirmed (Cochran, 2000).
In the field of peptides with a potential drug action an interesting case is that of an erythropoietin (EPO) agonist 20-residue peptide that has no resemblance with the natural hormone (so its affinity could not be predicted), but yet it binds to the hormone site as a dimer activating the receptor (Wrighton et al., 1996).
Another important approach in developing peptides that interfere with protein/protein interactions is that of replacing some natural amino acids with non-natural amino acids. This approach has been used to develop short peptides for the SH3 domains. These domains are very important for protein/protein interactions since they are small docking units present in many signal-transduction proteins (Pawson and Scott, 1997; Pawson et al., 2001). Thus, it has been possible to replace parts of the polyproline helix-recognition sequences with non-natural, N-substituted glycine residues (Nguyen et al., 1998). These peptide-peptoid hybrids often have higher affinity than that of the natural peptide and improved specificity for SH3 domains (Cochran, 2001) and by competing for binding to the SH3 domain they may stop the signal along the wiring pathways in the cytoplasm such as the RTK-Ras-MAP kinase pathway.
C. Possible Targets for Drugs Acting on Heteromeric Receptor Complexes
When the receptor/receptor interactions are considered it is possible to make the following considerations:
-
The receptor spans three different phases and in principle three different types of drugs could be developed according to where (extra-cellularly, intramembrane, intracellularly) the sequence of the receptor which should be targeted for the drug is located
-
The receptor component of the three-component system that forms the GPCR appears as the target of choice for drugs, namely for two reasons: stoichiometry and accessibility of the target. The greater accessibility of a drug to the receptor component is evident. As far as the stoichiometry of the three components is concerned it has been shown that the ratio of receptor/G-protein/adenylyl cyclase is in most instances equal to 1:100:3 (Ostrom et al., 2000). These data lead one to predict that the receptor or adenylyl cyclase are the best targets.
On the basis of these data and of the receptor/receptor interactions, it is possible to conceive of other pharmacological interventions besides the classical approach aimed to the activation or inhibition of the receptor due to occupation by the drug of the binding pocket for the natural ligand. Several other approaches are possible based on receptor/receptor interactions in heteromeric complexes:
-
The drug is developed for one coreceptor to modulate another coreceptor at the recognition level (binding pocket). One example is to have an adenosine A2A receptor antagonist acting on the A2A coreceptor in the antagonistic A2A/D2 heteromeric receptor complex to produce enhancement of D2 coreceptor signaling by removal of the A2A-induced reduction of affinity of the D2 coreceptor. Such a drug may become a novel antiparkinsonian drug with less side-effects as indicated from early (Fuxe and Ungerstedt, 1974; Fredholm et al., 1983; Herrera-Marschitz et al., 1988; Casas et al., 1988; Popoli et al., 1991 Jiang et al., 1993; Kanda et al., 1994) and recent work (Malec, 1997; Kanda et al., 1998a,b; Grondin et al., 1999; Shiozaki et al., 1999; Ward and Dorsa, 1999; Wardas et al., 1999; Aoyama et al., 2000; StrÖmberg et al., 2000; Chen et al., 2001; Ferré et al., 2001; Fuxe et al., 2001; Hauber et al., 2001; Pinna et al., 2001; Morelli and Wardas, 2001; T. Chase, personal communication). The best approach will be to block selectively the A2A binding pocket in the heteromeric complex and not those A2A receptors not linked to the D2 receptors. In this way a novel form of selectivity can be obtained based on the unique selectivity features of the binding pockets of the heteromeric complex. The assay systems will then be neuronal cell lines expressing A2A receptors alone and the A2A/D2 heteromeric receptor complex to discover A2A antagonistic drugs with the desired selectivity for the heteromeric A2A coreceptors. Thus, the heteromeric complex is a novel target for drug development (see also Fuxe et al., 1989). It has also been discussed that dimeric compounds can be designed for the heteromeric complex to cointeract with the two binding pockets of the heteromer (Franco et al., 2000; Marshall 2001) to obtain a signaling which better mimic that under physiological conditions. Still another target for drugs could be the interface of the heteromer where drugs can disrupt its formation (see above). Finally, it must be considered that dependent on the heteromeric complex and on the pathological conditions studied it may be beneficial to block or to enhance the intramembrane interactions in the heteromeric coreceptor complex. It must also be emphasized again that in the case of the GABAB heteromer it has been elegantly indicated that a novel anticonvulsant gabapentin is selective for the GABA R1α/GABABR2 heteromer. Thus, this heteromeric pharmacology has also had an impact on drug development in other types of heteromers (Ng et al., 2001).
-
The drug is developed for one coreceptor to address the G protein coupling and G protein selectivity of another coreceptor or the activity of an ion channel receptor existing in the same heteromeric complex. One example is the D5/GABAA heteromeric receptor complex where activation of the D5 receptor can reduce the synaptic strength of the GABAA receptor. Thus, activation of the D5 receptor of this complex offers a novel approach for selective reduction of GABAA signaling in this complex. Thus, GABAA signaling in discrete brain regions may be reduced in a selective way. It illustrates how drugs can be developed based on various types of heteromeric complexes to reduce or enhance GABA or glutamate synaptic signaling in discrete brain regions. Hence, a novel type of drugs can be developed based on heteromeric complexes containing ion channel receptors and GPCR that may be used to treat a number of neuropsychiatric disorders. It must also be underlined again that the receptor/receptor interaction is reciprocal in the D5/GABAA heteromeric receptor complex so that the GABAA receptor activation can control the D5 receptor coupling to its G protein and thus its efficacy. These types of heteromeric complexes may also allow to select drugs preferentially acting on ion channel receptors of heteromeric complexes. It should also be considered that in several examples of heteromerization such as the A2A/D2 and A2A/mGluR5 receptor heteromers the simultaneous activation of the two binding pockets may also give rise to coupling to other types of intracellular pathways such as the MAPK pathway leading to increased nuclear signaling via transcriptional factor activation with induction of marked changes in gene expression and of the phenotype. In this case development of dimeric agonists for these heteromeric complexes may have a unique and selective trophic potential and help learning and memory processes. It should be considered that the drug developed for one coreceptor may modulate both the binding and the G protein properties as well as the traffic of the other coreceptor.
-
The drug is developed for one or both coreceptors to control the receptor trafficking of the heteromeric complex. As an example the A2A/D2 heteromeric receptor complex can be mentioned, since prolonged A2A or D2 receptor agonist treatment in vitro alone produced coclustering and a certain cointernalization and homologous and heterologous down regulation of A2A receptor function in neuroblastoma cells (Hillion et al., 2002). Prolonged combined treatment with agonists for the A2A and D2 coreceptors produced a much stronger cointernalization and codesensitization involving also the D2 receptor. Thus, it seems likely that increased understanding of the joint regulation by agonists of the trafficking of the A2A/D2 heteromeric receptor complex will give us a novel understanding of the desensitization and sensitization at the D2 receptor, a key target for treatment of neuropsychiatric diseases. A2A receptor antagonists may therefore also be used in Parkinson's disease because they counteract the internalization and desensitization of D2 like receptors after prolonged L-DOPA and/or D2 receptor agonist treatment in addition to having an antiparkinsonian and neuroprotective activity (Ferré et al., 2001; Morelli and Wardas, 2001). This may be true also for other heteromeric receptor complexes and therefore offers a new way of avoiding desensitization of key receptors in heteromeric complexes after prolonged agonist treatment, namely by developing drugs that act on the coreceptors.
Acknowledgments
This work has been supported by a grant from the Swedish Research Council (14X-00715), by a grant from the European Commission (QLG3-CT-2001-01056), by grants (BIO99-0601-C02-02, SAF2001-3474) from the Interministerial Commission of Science and Technological Innovation (CICYT) and 2001/012710 from Marató de TV3 Fundation and by a Consiglio Nazionale delle Ricerche (CNR) (40%) grant.
Footnotes
-
↵1 Dedicated to Prof. G. L. Gessa.
-
↵2 Abbreviations: GPCR, G protein-coupled receptor; ADA, Adenosine deaminase; ADM, adrenomedullin; AT, angiotensin; BRET, bioluminescence resonance energy transfer; CaMK, Ca2+/calmodulin-dependent protein kinases; CCK, cholecystokinin; CGRP, calcitonin gene-related peptide; CHO, Chinese hamster ovary; CRE, cAMP response element; CREB, cAMP response element-binding protein; CRL, calcitonin receptor-like; DARPP-32, dopamine and cyclic adenosine 3′,5′-monophosphate-regulated phosphoprotein, 32 kDa; EGF, epidermal growth factor; FRET, fluorescence resonance energy transfer; Gal, galanin; GRK, GPCR kinase; GST, glutathione S-transferase; GTPγS, guanosine 5′-O-(3-thio)triphosphate; HEK, human embryonic kidney; 5-HT, 5-hydroxytryptamine (serotonin); IP3, inositol 1,4,5-trisphosphate; MAPK, mitogen-activated protein kinase; MEK, MAPK/extracellular-regulated kinase; MPTP, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine; NK, neurokinin; NMDA, N-methyl-d-aspartate; NPY, neuropeptide Y; NT, neurotensin; PAGE, polyacrylamide gel electrophoresis; PAR, protease-activated receptor; PCP, phencyclidine; PLC, phospholipase C; PKA, protein kinase A; PKC, protein kinase C; PP-1, protein phosphatase-1; PSD, postsynaptic density protein; RAMP, receptor activity-modifying protein; RTK, receptor tyrosine kinase; SST, somatostatin; TCF/Elk-1, ternary complex factor/Elk-1; TM, transmembrane; trunc, truncated; VT, volume transmission; WT, wiring transmission.
-
Article, publication date, and citation information can be found at http://pharmrev.aspetjournals.org.
-
DOI: 10.1124/pr.55.3.2.
- The American Society for Pharmacology and Experimental Therapeutics
References
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
-
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
-
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵