Integrated events in central dopamine transmission as analyzed at multiple levels. Evidence for intramembrane adenosine A_2A/dopamine D₂ and adenosine A₁/dopamine D₁ receptor interactions in the basal ganglia¹

Abstract

An analysis at the network and membrane level has provided evidence that antagonistic interactions between adenosine A_2A/dopamine D₂ and adenosine A₁/dopamine D₁ receptors in the ventral and dorsal striatum are at least in part responsible for the motor stimulant effects of adenosine receptor antagonists like caffeine and for the motor depressant actions of adenosine receptor agonists. The results obtained in stably cotransfected cells also underline the hypothesis that the intramembrane A_2A/D₂ and A₁/D₁ receptor interactions represent functionally important mechanisms that may be the major mechanism for the demonstrated antagonistic A_2A/D₂ and A₁/D₁ receptor interactions found in vivo in behavioural studies and in studies on in vivo microdialysis of the striopallidal and strioentopeduncular GABAergic pathways. A major mechanism for the direct intramembrane A_2A/D₂ and A₁/D₁ receptor interactions may involve formation of A_2A/D₂ and A₁/D₁ heterodimers leading to allosteric changes that will alter the affinity as well as the G protein coupling and thus the efficacy to control the target proteins in the membranes. This is the first molecular network to cellular integration in the nerve cell membrane and may be well suited for a number of integrated tasks and can be performed in a short-time scale, in comparison with the very long-time scale observed when receptor heteroregulation involves phosphorylation or receptor resynthesis. Multiple receptor–receptor interactions within the membranes through formation of receptor clusters may lead to the storage of information within the membranes. Such molecular circuits can represent hidden layers within the membranes that substantially increase the computational potential of neuronal networks. These molecular circuits are biased and may therefore represent part of the molecular mechanism for the storage of memory traces (engrams) in the membranes.

Introduction

In the eighties, the concept was introduced of multiple intrasynaptic transmission lines, which interact at presynaptic and/or postsynaptic levels. These interactions could also take place at plasma membrane level. In this way reciprocal modulatory influences can occur in neurotransmitter release and recognition/decoding processes [6]. Several types of membrane integral proteins (ion channels, receptors, etc.) as well as membrane-associated proteins (enzymes, cytoskeleton proteins, etc.) may be involved. In this context, we have studied one type of these intramembrane (and/or membrane associated) molecular interactions, namely the receptor–receptor interactions 1, 2, 4, 7, 26, 27, 28, 29, 30, 31, 32, 33, 37, 38, 39, 40, 73. By receptor–receptor interactions is meant that a transmitter by binding to its receptor can modify the binding characteristics of the receptor for another transmitter or modulator. It should be noticed that, in principle, it is also possible that a transmitter by binding to one subtype of its receptors can modulate the recognition/decoding characteristics of another subtype of its receptors. Receptor–receptor interactions have so far only been studied on G protein coupled receptors and can be either G protein dependent or G protein independent 4, 22, 29, 73. In the latter case, a direct interaction between two receptor proteins may occur. The receptor cross-talk mechanisms may constitute the molecular network for a cellular integration characterized by an extraordinary miniaturization of computational elements. Through the membrane receptor–receptor interactions the intracellular effects of an active receptor will become potentiated, inhibited or even directed towards a different intracellular effector cascade in relation to the state of activity of other receptors which interact with that receptor 2, 38, 39, 40, 73.

In principle, the demonstration of the existence and functional significance of the interaction between two receptors must comprise an analysis of their reciprocal influence from the behavioural level down to the neurochemical and molecular levels. In order to prove its functional relevance, the molecular phenomenon must be shown to be necessary for the occurrence of the interaction at the higher levels. In addition, the precise mechanism of the receptor–receptor interaction at molecular level must be characterised.

Therefore, several experimental steps must be performed to assess the functional significance of the receptor X–receptor Y interaction:

• The (e.g., pharmacological) activation of the transmitter-identified system X influences a transmitter Y-dependent neurochemical effect which is relevant for the elicitation of a specific behavior (transmitter Y-dependent behavior).

• The modulatory action of transmitter X is exerted on the same cellular target on which transmitter Y acts to elicit the effect of point 1.

• In the membrane of the cellular target of point 2, (a) the activation of the receptor for the transmitter X changes the binding characteristics of the receptor for the transmitter Y and (b) this interaction is necessary for the elicitation of the effect of point 1.

From a mechanistic standpoint, we can distinguish true membrane receptor–receptor interactions, when all the molecular elements necessary for the interaction are integral membrane (type B receptor–receptor interaction, [73]) and/or membrane-associated molecules (type C receptor–receptor interaction, [73]) from receptor–receptor interactions involving intracellular molecules such as second messenger cascades (type D receptor–receptor interaction, [73]).

In the last 20 years, many studies have been successfully devoted to the demonstration of points 1, 2 and 3a in several neural systems, although the demonstration of point 3b often remains circumstantial. In addition, molecular tools have started to make it possible to address the issue of the precise molecular mechanisms involved in the receptor–receptor interactions.

The interest of our group has mainly been to characterize the heteroregulation of dopamine (DA) D₂ and DA D₁ receptors in the basal ganglia. Evidence has been obtained for the existence of intramembrane cholecystokinin-B (CCKB)/D₂, neurotensin (NT)/D₂, adenosine A_2A/D₂ and adenosine A₁/D₁ receptor interactions 3, 5, 8, 10, 15, 16, 17, 18, 19, 21, 22, 34, 35, 36, 39, 51, 63, 71within the dorsal and ventral striatum.

In the present review article we will discuss how a combined approach using behavioural, neurochemical and molecular biological techniques has provided evidence that

• A_2A/D₂ and A₁/D₁ transmissions, respectively, antagonistically interact in the ventral and dorsal striatum and are to a large extent responsible for the motor stimulant effects of adenosine receptor antagonists like caffeine and for the motor depressant actions of adenosine receptor agonists 15, 22, 25.

• A_2A/D₂ and A₁/D₁ receptors are present in the same target cells and

• true membrane receptor receptor–receptor interactions as defined above take place in these target cells

Subsequently, evidence will be presented that these receptor–receptor interactions can occur without involving G proteins and are therefore an example of type B receptor–receptor interactions. It will be proposed that receptor heterodimerization is one possible molecular mechanism for type B receptor–receptor interactions.

Finally, indications will be presented for multiple receptor–receptor interactions, supporting the view of the existence of membrane molecular circuits, which have as crucial elements receptors for neurotransmitters.

Motor initiation can be brought about by activation of the D₁-regulated direct efferent pathway of the basal ganglia made up of strionigral and strioentopeduncular GABA neurons. Loss of motor inhibition is instead produced by inhibition of the D₂-regulated striopallidal pathway, which is the first link of the indirect efferent pathway of the basal ganglia. The D₂ receptors are inhibitory and the D₁ receptors are excitatory with regard to the indirect and direct pathways, respectively. The A_2A receptors in the basal ganglia are concentrated in the striopallidal GABA neurons (dorsal and ventral parts), while the A₁ receptors have a more wide-spread distribution, but are found in the D₁ rich strionigral and strioentopeduncular GABA pathways (for review, see Ferré et al. [22]).

As illustrated in Fig. 1, low doses of the A_2A receptor agonist CGS 21680 can effectively diminish the increases of motor activity produced by D₂ agonists, but cannot counteract the increases of motor activity produced by D₁ receptor agonists 13, 14, 22. In contrast, A_2A receptor antagonists can enhance the motor actions of D₂ as well as D₁ receptor agonists [59]. Since A_2A/D₁ receptors do not coexist within the same striatal efferent neurons (see below), the enhancement by A_2A receptor antagonists of the motor actions on D₁ receptor agonists probably reflect an interaction at the network level, due to synergistic interactions of the indirect and direct pathways within the globus pallidus interna and within the zona reticulata of the substantia nigra [59]. Therefore, the behavioural analysis alone leaves open the question whether the demonstrated interactions are due to receptor–receptor interactions or interactions between different neuronal systems.

As illustrated in Fig. 2, in in vivo dual probe microdialysis experiments, perfusion of the somatodendritic region of the striopallidal GABA neurons with the A_2A agonist CGS 21680, fully counteracts the ability of the D₂ agonist pergolide to reduce pallidal GABA release in the awake freely moving rat [16]. In line with these results using the same methodology, combined perfusion with low doses of the adenosine receptor antagonist theophylline and of the D₂ receptor agonist pergolide, led to a considerable enhancement of the ability of the D₂ receptor agonist to reduce the extracellular levels of pallidal GABA [16]. The existence of an antagonistic A_2A/D₂ transmission interaction in the striopallidal GABA neurons can also explain the ability of CGS 21680 to reduce the ability of a D₂ receptor agonist to increase the c-fos expression in the globus pallidus [56].

The ability of antipsychotic drugs to counteract the increases of locomotion produced by d-amphetamine and other DAergic drugs is probably related to the blockade of DA receptors in the nucleus accumbens and the olfactory tubercle and appears to predict antipsychotic potency [75]. In contrast, the production of catalepsy and counteraction of stereotypes produced by DAergic agents, such as d-amphetamine, instead involves blockade of DA receptors in the dorsal striatum. A detailed analysis has therefore been performed in order to characterize the actions of the A_2A agonist CGS 21680 on d-amphetamine and phencyclidine (PCP) induced locomotion and on apomorphine-induced stereotypes (Fig. 3A, B) as well as its potency to produce catalepsy. In this analysis Rimondini et al. [62]found that CGS 21680 had a substantially stronger efficacy to antagonize PCP and d-amphetamine induced locomotion and motility than to induce catalepsy in the rat. Higher doses of CGS 21680 were also needed in order to block the stereotypes produced by the non-selective DA receptor agonist apomorphine.

These results demonstrate that the A_2A agonist CGS 21680 displays an atypical antipsychotic profile in animal models. There exist several hypothetical advantages with such an atypical antipsychotic drug operating through antagonistic A_2A/D₂ transmission, and, possibly, receptor interaction (see below). Since the drug would not directly act on the D₂ receptor, a development of supersensitivity lower than that usually seen with direct DA receptor blocking agents is to be expected. As a consequence, a lower incidence of tardive dyskinesia should occur. Autoradiographic studies [18]indicate a more powerful regulation of the D₂ receptors by A_2A agonists within the nucleus accumbens than in the caudate putamen, as also indicated by the present behavioural analysis. Thus, less parkinsonian side-effects should develop.

The fact that CGS 21680 was even more potent in antagonizing PCP induced motor activation than d-amphetamine induced motor activation is unexpected in view of the major role of the A_2A/D₂ transmission interaction. Nevertheless, it is known that the PCP induced motor activation is a DA dependent phenomenon. Therefore, a markedly enhanced DA tone, like in the case with d-amphetamine, can be less vulnerable to an A_2A induced inhibition of the D₂ receptor signal.

To test the hypothesis of the existence of an inhibitory A_2A/D₂ transmission interaction in the somatodendritic region of the ventral striopallidal GABA pathway a dual probe in vivo microdialysis study has been performed [18]. The D₂ receptor antagonist raclopride perfused into the nucleus accumbens was found to increase pallidal GABA release, an action that was mimicked by the A_2A agonist CGS 21680 when also perfused into the nucleus accumbens. Furthermore, as seen in Fig. 4, when low doses of raclopride and CGS 21680 were locally perfused into the rat nucleus accumbens they synergized and produced significant increases of extracellular GABA levels in the ventral pallidum, each drug alone in this case having no action. These results provide biochemical support for the existence of an inhibitory A_2A/D₂ transmission interaction within the ventral striopallidal GABA pathway in line with the behavioural analysis.

A prerequisite for the existence of receptor–receptor interactions is the colocalization of the two types of receptors in the same cell. Both in situ hybridization and immunocytochemical experiments have shown that DA D₂ receptor distribution is restricted to around half the medium spiny striatal neurons, namely, those projecting to the globus pallidus and to some small populations of striatal interneurons, such as the cholinergic neurons (reviewed in Ferré et al. [22]). A_2A are concentrated in DA D₂-containing GABA neurons projecting to the globus pallidus [65]. A_2A receptors may also exist in the cholinergic striatal interneurons, where A_2A receptor activation appears to lead to enhancement of acetylcholine release, in part involving a reduction of the ability of the D₂ receptors to inhibit the release of acetylcholine [47]. Thus, an enhancement of cholinergic tone in the neostriatum by A_2A receptor activation of cholinergic interneurons may contribute to the findings obtained on pallidal GABA release. Such a mechanism may also contribute to the above mentioned behavioural findings that CGS 21680 can inhibit the D₂ agonist induced motor behaviours [69]. The biochemical studies described above show that a main site for the interaction between A_2A and D₂ transmissions is the soma and dendrites of the striopallidal GABA neurons. A similar interaction may also occur at the nerve terminal level of these striopallidal GABA neurons [55].

In conclusion, the functional and anatomical evidence reviewed above points to the fact that the interaction between A_2A and D₂ transmissions in the striatum is functionally relevant and occurs in the same target neurons. Therefore, the prerequisites for the existence and functional significance of A_2A/D₂ receptor–receptor interactions stated at points 1 and 2 (see Section 1) are met.

Systemic administration of the adenosine A₁ receptor agonist cyclopentyladenosine selectively counteracted a series of behavioural effects produced by the DA D₁ agonist SKF 38393, namely motor activation in reserpinized mice, oral dyskinesias in rabbits and EEG arousal in rats 19, 60. Recently, Popoli et al. [61]showed that the adenosine A₁ receptor antagonist 8-cyclopentyl-1,3-dimethylxanthine significantly potentiated the motor effects induced by the DA D₁ agonist SKF 38393 in reserpinized mice and in rats with unilateral 6-hydroxydopamine (6-OHDA) induced lesions of the substantia nigra (Fig. 5). These results strongly support the existence of an antagonistic adenosine A₁/DA D₁ transmission interaction in the basal ganglia and possibly also in the cerebral cortex (EEG arousal). As discussed by Popoli et al. [61], the inability of the adenosine A₁ antagonist to modulate the motor actions of the DA D₂ agonist quinpirole can be explained on the basis of the wide-spread distribution of brain adenosine A₁ receptors. Thus, simple synergistic interactions of the direct and indirect pathways in striatal output structures may not be possible.

By means of dual probe in vivo microdialysis it has been possible to demonstrate that the striatal DA D₁ receptor can enhance GABAergic neurotransmission in the rat strioentopeduncular pathway, while A₁ receptor activation alone exerted no effects on this pathway [21]. However, the striatal adenosine A₁ receptors were found to strongly antagonize the D₁-mediated facilitation of GABA release in the entopeduncular nucleus, suggesting the existence of an antagonistic A₁/D₁ interaction within the strioentopeduncular GABA pathway. The fact, however, that the noncompetitive NMDA antagonist MK 801 could partially antagonize the action of SKF 38393 suggests that an A₁ receptor mediated inhibition of glutamate release also can contribute to these results [21]. Nevertheless, this strioentopeduncular GABA neuron appears to be one important locus for A₁/D₁ receptor interactions in the brain and may be a major basis for the demonstrated A₁/D₁ receptor interactions at the behavioural level.

Contrary to A_2A receptors which are concentrated in the D₂-containing striopallidal GABA neurons, A₁ receptor distribution is more wide-spread. A₁ receptors are colocalised with D₁ receptors in the strionigral and strioentopeduncular GABA neurons. At these levels, A₁/D₁ receptor–receptor interactions may take place. However, A₁ receptors are also present, for instance, in the striopallidal GABA neurons as well as in glutamatergic neurons projecting to striatum from a number of cortical areas [22]. The biochemical experiments reported above indicate that the soma and dendrites of the strioentopeduncular neurons is a main site for the interaction between A₁ and D₁ transmissions.

In conclusion, the functional and anatomical evidence reviewed above is compatible with the existence of A₁/D₁ receptor–receptor interactions in the basal ganglia and prefrontal cortex, although the interaction between A₁ and D₁ transmissions may take place, in a more indirect way, at a number of other sites.

In 1991 Ferré and colleagues demonstrated that activation of the adenosine A_2A receptor via CGS 21680 reduces the affinity of DA D₂ receptors for agonists but not for antagonists in membrane preparations of the rat caudate putamen (Fig. 6) 11, 12. The analysis of competitive inhibition curves with DA versus the D₂ antagonist [³H]raclopride demonstrated that the A_2A agonist CGS 21680 increased the equilibrium dissociation constant for both low (K_L) and high affinity (K_H) states of the D₂ receptors, as well as produced an increase in the proportion of DA D₂ receptors in the high affinity state (R_H) (Fig. 7, see also Table 1). These results indicated the existence of an intramembrane antagonistic A_2A/D₂ receptor interaction within the caudate putamen. Under pathophysiological conditions this intramembrane interaction might be even more strongly developed, in view of the demonstration that striatal DA denervation or chronic treatment with the DA D₂ antagonist haloperidol resulted in an increased efficacy of the A_2A agonist to increase the K_H, K_L and R_H values of the D₂ receptors of the rat neostriatum 14, 20. It has also been demonstrated that chronic treatment with haloperidol results in upregulation of both A_2A and D₂ receptors in the rat caudate putamen [58]. In line with these findings an enhancement of the motor activating effect of adenosine receptor antagonists has been found after DA denervation or chronic treatment with haloperidol. These results indicate that A_2A agonists may be effective in the treatment of tardive dyskinesias and A_2A antagonists in the treatment of Parkinson's disease.

In order to further understand the mechanism involved in A_2A/D₂ receptor interactions, kinetic studies have been performed in striatal membranes. The D₂/D₃ receptor agonist [³H]quinpirole was used. CGS 21680 increased the K_A value (initial formation of receptor ligand complex) and the K_I value (isomerization rate) of the complex but did not affect the dissociation (deisomerization). Thus, A_2A activation can produce changes in several steps of the D₂ agonist binding to D₂ receptors in striatal membrane preparations [50].

In looking for simpler models to analyze the mechanism of the A_2A/D₂ receptor antagonistic interaction COS-7 cells were used transiently transfected with DA D₂ receptors and adenosine A₂ receptor cDNAs [67]. However, in spite of high densities of adenosine A_2A and DA D₂ receptors with high affinity binding characteristics it was not possible to demonstrate the A_2A/D₂ antagonistic interaction in these cells, possibly related to the use of transient transfection. Therefore, stable transfections with the dog adenosine A_2A receptor cDNA [52]was performed in human DA D₂ receptor (long form) containing fibroblast cells (Ltk⁻). In the stably dog A_2A/human D_2L cotransfected fibroblast cells CGS 21680 (100 nM) increased the K_L, K_H, and R_H values in membrane preparations in the same way as in striatal membrane preparations. Furthermore, the effects were blocked by the adenosine receptor antagonist 8-phenylteophylline [10]. In view of these results it seems likely that in the transiently transfected cells only a low number of cells contained the two receptors.

It is possible that adenylyl cyclase (AC) activation is involved in the A_2A/D₂ receptor interaction, since the A_2A receptors via a Gs protein stimulates AC. However, in the three clones analyzed, CGS 21680 only produced weak (clone 7) or no (clones 8, 9) cAMP accumulation in the A_2A/D₂ cotransfected cells (Fig. 8). Indeed, D₂ receptors (coupled to AC in an inhibitory fashion) only weakly counteracted forskolin induced cAMP accumulation. These results suggest that both A_2A and D₂ receptors are only weakly coupled to G protein in the transfected cells and also, more importantly, that AC is not involved in this antagonistic A_2A/D₂ receptor interaction. In further support of this view NECA, a non-selective adenosine agonist, and forskolin produced a marked cAMP accumulation in the control cells, having only D₂ receptors, but were not able to modulate the binding characteristics of the D₂ receptors. The NECA-induced cAMP accumulation is produced via activation of low affinity A_2B receptors present in both control and A_2A/D₂ cells [10]. Based on this evidence it is not likely that a protein phosphorylation process is involved. Furthermore, ATP was not added in the binding assays, a condition which is necessary for protein kinase activation.

Besides inducing modifications in ligand binding characteristics, A_2A agonists are able to influence D₂ receptor function in the A_2A/D₂ Ltk⁻ cells although not at the AC level. Thus, CGS 21680 concentration dependently reduced the influx of Ca²⁺ ions upon activation of the D₂ receptors with quinpirole (Fig. 9) [72], an effect that may be related to the demonstrated alteration of the D₂ binding by the A_2A agonist. Thus, the change in the D₂ binding by A_2A may reflect a change in efficacy caused by altered isomerization favouring a non-activated state and/or by a reduced G protein coupling.

Studies have also been initiated in CHO cells cotransfected with human A_2A and rat D₂ short cDNAs. In this cell line, the D₂ and A_2A receptors show a stronger coupling to G protein and effectively inhibit and stimulate cAMP accumulation, respectively (Kull et al. in preparation). Again, CGS 21680 can potently increase the K_L, K_H, and R_H values of D₂ receptors in the CHO membranes as previously described in striatal membranes and Ltk⁻ membranes. Thus, D_2L and D_2S can both become modulated by A_2A receptor activation. In contrast, previous work [12]has demonstrated that the A_2A receptor cannot modulate the binding characteristics of D₁ receptors in striatal membranes, demonstrating DA subtype specific changes upon A_2A activation.

Although we have focused our review on the intramembrane interactions between A_2A/D₂ receptors (as well as A₁/D₁ receptors, see below) (Fig. 10) it must be pointed out that classical interactions at the adenylate cyclase level may occur as well and contribute to A_2A/D₂ transmission interaction 8, 25, 43. Stimulation of A₁ and D₂ receptors inhibits AC activity through Gi proteins, and A_2A or D₁ receptors activate AC through G_s or possibly G_olf proteins [42]. Thus, interactions take place also at the second messenger level and beyond [22]. For instance, there may exist an antagonistic D₂ control of A_2A receptors (D₂/A_2A) that may operate to a large degree through a direct inhibition of AC (Fig. 10), contributing to the ability of a D₂ agonist to antagonize the CGS 21680 induced c-fos expression in the DA denervated striatum [56].

Taken together, the findings reported in this chapter show that

• when the two receptor types are present in the same cell membrane, A_2A receptors can inhibit D₂ receptor binding and decoding characteristics;

• this interaction does not require AC activation;

• this interaction is qualitatively similar to that observed in membranes from caudate-putamen cells, as far as binding inhibition and lack of AC activation are concerned.

The hypothesis is therefore advanced that membrane antagonistic A_2A/D₂ receptor–receptor interactions occurring in striopallidal GABA neurons are a main mechanism for the antagonistic A_2A/D₂ transmission interactions demonstrated in biochemical and behavioural experiments (see Section 1.1.1Section 1.1.2Section 1.1.3).

We have recently shown that interactions between A₁ and D₁ receptors take place also at the membrane level as evaluated in membrane preparations from the rat neostriatum [19]. In competitive inhibition experiments with DA versus the D₁ antagonist [³H]SCH23190 the adenosine A₁ receptor agonist CPA significantly reduced the R_H values (proportion of binding sites in the high affinity state) without influencing the K_L and K_H, values (Fig. 11). The adenosine antagonist 8-phenyltheophylline could antagonize the action of CPA on the D₁ receptor binding characteristics. These results were interpreted to indicate an uncoupling of the D₁ receptors to the G proteins, since the GTP analogue Gpp(NH)p has been found to exert the same actions on the R_H values. Thus, an antagonistic A₁/D₁ receptor intramembrane interaction appears to exist in the striatal membranes and may represent a major mechanism for the antagonistic A₁/D₁ receptor interactions found in the behavioural studies and in the studies with the dual probe in vivo microdialysis.

These effects have been recently reproduced in a mouse fibroblast Ltk⁻ cell line, stably cotransfected with human A₁ and human D₁ receptor cDNAs [24]. The clone selected was the one with similar amounts of both types of receptors, both being functional in terms of cAMP responses. The adenosine A₁ agonist CPA can reduce the R_H values of DA for the D₁ antagonist binding sites in the A₁/D₁ cells but not in the control cells lacking A₁ receptors, but containing D₁ receptors. In contrast, the GTP analogue Gpp(NH)p was effective both in the D₁ alone cells and in the A₁/D₁ cells in reducing the R_H values for the D₁ receptors.

Taken together, the findings reported in this chapter show that

• when the two receptor types are present in the same cell membrane, A₁ receptors can reduce the proportion of D₁ receptors in the high affinity state;

• this interaction may be due to uncoupling of D₁ receptors to the G-protein;

• this interaction is qualitatively similar to that observed in membranes from caudate-putamen cells, as far as the modulations of binding characteristics are concerned.

The hypothesis is therefore advanced that membrane antagonistic A₁/D₁ receptor–receptor interactions occurring in strionigral and strioentopeduncular GABA neurons are a main mechanism for the antagonistic A₁/D₁ transmission interactions demonstrated in biochemical and behavioural experiments (see Section 1.1.4Section 1.1.5).

In principle, an interaction between Gs and Gi proteins [42]may underlie the A_2A receptor-induced increase in the R_H values of the D₂ receptors in the striatal membrane preparations (type C receptor–receptor interaction, [74]. A_2A receptor activation would be expected to increase the formation of β- and γ-subunits from the A_2A linked Gs that could bind to the α-subunit of the D₂ linked Gi. In this way, an inhibition of Gi protein dissociation from the D₂ receptors could take place, resulting in the observed increases of the R_H values. However, this mechanism does not appear to be of substantial importance, since it was found that the A_2A mediated increase in R_H values of the D₂ receptors was independent of the effect of GTP as well as GDP agonists [17]. These results, taken together with the results obtained in the A_2A/D₂ fibroblast cells ruling out the involvement of AC in the A_2A/D₂ receptor interaction, strongly support the hypothesis that the interaction between the A_2A/D₂ receptors is a direct one (type B receptor–receptor interaction, [73]), possibly related to the formation of heterodimers (see below and [73]) leading to allosteric changes in one or both types of receptors involved. Nevertheless, this does not exclude that the interactions at the level of the Gs/Gi and adenylate cyclase represent additional important mechanisms for the integration of the A_2A/D₂ receptor-dependent signaling.

Ferré et al. [19]showed that CPA and the GTP analogue Gpp(NH)p exerted synergistic actions on the R_H values of D₁ receptors in striatal membranes. It was, therefore, proposed that A₁ receptor modulation of the D₁ receptors involves a GTP independent uncoupling of the D₁ receptors to the G proteins. This hypothesis was further tested by using pertussis toxin in binding experiments on A₁/D₁ fibroblast cells [20]. Pertussis toxin is known to produce uncoupling of the A₁ receptor to its G protein and ADP ribosylation of the Gi proteins. This causes a reduction in the number of A₁ receptors in the high affinity state and a blockade of A₁ receptor signalling. By exposing the A₁/D₁ cell to pertussis toxin, it is therefore possible to evaluate whether G proteins are involved in the uncoupling of D₁ receptor to its G protein produced by A₁ receptor activation. It was found that pertussis toxin antagonized the intramembrane A₁/D₁ receptor interaction elicited by low (nanomolar) concentrations of CPA but not that elicited by higher (micromolar) concentrations of CPA. Some caution must be exerted in interpreting these results as the pertussis toxin induced inactivation was incomplete. However, these findings suggest that activation of the low affinity state of the A₁ receptors is capable of producing changes in the binding characteristics of the D₁ receptors via G-protein independent mechanisms, thus involving a direct interaction between A₁ and D₁ receptors (type B receptor–receptor interaction, [73].

Dimerization or oligomerization of receptor molecules may develop upon activation by an agonist and appears to be a general phenomenon [46]. Dimerization may be a necessary condition for activation or a way to produce response potentiation or metabolic stabilization of the receptors [73].

The best known cases of membrane receptor dimerization have been demonstrated within the tyrosine kinase receptor superfamily [66]. Dimerization appears to increase the affinity for the agonist. Functional G protein coupled receptors may also exist in a dimeric form and activation of LHRH receptors demonstrated that the dimerization processes are a necessary step for the receptor function [9]. Recently, indications have been obtained that both β₂-adrenergic receptors and DA D₂ receptors may exist as dimers 45, 57, 71. In the experiments with DA D₂ receptor oligomerization, coimmunoprecipitation of the D₂ receptors of the human caudate nucleus was made, demonstrating a dimeric form. D₂ receptors were found to exist either as monomers or as dimers. Peptides corresponding to the transmembrane regions VI and VII of the D₂ receptor were found to block the dimerization process. Thus, these transmembrane domains appear to be involved in the D₂ dimerization. Electrostatic intermolecular non-covalent interactions were postulated by these workers [57].

In 1993 Zoli et al. [73]postulated that heterodimerization may be the mechanism involved for the demonstrated receptor–receptor interactions between G protein coupled receptors (Fig. 12). As a matter of fact heterodimerization had early on been clearly demonstrated for the tyrosine kinase receptors [68]. There is also some evidence that the μ- and δ-opioid receptors probably are functionally and physically associated [64]. In view of the fact that the homodimers appear to be formed through interactions at transmembrane regions, especially transmembrane region VI [54](but see Maggio et al. [53]for possible involvement of other receptor domains), it seems likely that also the heterodimers postulated to be formed involve interactions in transmembrane domains, probably the same domains that are involved in the homodimerization process. It was postulated by Zoli et al. [73], that the density of interacting receptors and the number of receptors activated by their agonists in each population control the proportion of hetero- versus homodimerization and subsequently the overall action on the target function. We, therefore, presently postulate that the major mechanism for the direct intramembrane A_2A/D₂ and A₁/D₁ receptor interactions involve formation of A_2A/D₂ and A₁/D₁ heterodimers leading to allosteric changes that will alter the affinity as well as the G protein coupling and thus the efficacy to control the target proteins in the membranes. Experiments are presently ongoing to provide evidence for this phenomenon based on tagging the D₂ and A_2A receptors and their subsequent coimmunoprecipitation.

It has recently been suggested that striatal adenosine A_2A transmission interact with group I metabotropic glutamate transmission in the control of motor activity. Thus, it seems possible that multiple interactions between striatal A_2A, D₂ and type I metabotropic glutamate receptors could be involved in mediating motor actions of the group I metabotropic agonist 1S-3R-ACPD [48]. In line with this hypothesis, Ferré et al. [23]recently demonstrated that 1S-3R-ACPD substantially reduced the affinity of the high affinity state of the DA D₂ receptors, an effect similar to that previously demonstrated upon incubation with the A_2A agonist SCH 21680. This action was antagonized by the selective group I metabotropic glutamate receptor antagonist AIDA but not by the adenosine antagonist 8-phenyltheophylline. Of substantial importance was the demonstration of a strong synergistic action on the DA D₂ receptors when the striatal membranes were incubated with both the A_2A and the metabotrobic glutamate agonist in terms of K_H, K_L and R_H values. These results were furthermore matched by the ability of the 1S-3R-ACPD to potentiate the antagonistic action of the A_2A agonist on D₂ agonist induced turning behaviours in unilaterally 6-OHDA lesioned rats. Combined use of a metabotropic glutamate receptor antagonist group I and of an adenosine A_2A receptor antagonist could therefore be an important new type of treatment of Parkinson's disease and elegantly illustrates how receptor–receptor interactions can be used in drug development 22, 31, 33, 39.

As illustrated in Fig. 12, the membrane receptor–receptor interactions represent a molecular mechanism for computation occurring at membrane level. Computational mechanisms based on receptors located in the cell membrane, the main interface between the cell and its environment, are ideally suited to intervene on the selection of the extracellular events relevant for the functioning of the cell. Indeed, the 2nd messenger cascades amplify receptor modulations, so that even small changes in receptor function may result in large intracellular effects. We can consider two simple models based on excitatory and inhibitory receptor–receptor interactions, respectively. When excitatory interactions are involved we may obtain an integration of coincident signals, that is the generation of an output that is qualitatively or quantitatively different when two or more inputs are simultaneously present as compared to when they are presented alone [73]. Inhibitory interactions among receptors instead filter out incoming signals that are excessive and irrelevant for a particular cellular state. Thus, the filtration of incoming D₂ signals may be one role for the A_2A/D₂ receptor interaction and filtration of incoming D₁ signals may be one role for the A₁/D₁ receptor interactions. Taken together, the receptor–receptor interaction mechanisms appear to constitute the first molecular network to cellular integration in the nerve cell membrane and may be well suited for a number of integrated tasks. The main differential features of membrane (type B and C) receptor–receptor interactions with respect to other mechanisms of transmission interactions (including type D receptor–receptor interactions) are

• their main dependence on extracellular signals;

• their short time scale in comparison with the longer time scale of receptor heteroregulation involving phosphorylation or receptor resynthesis.

Current theories on memory trace formation are based on the modification and subsequent stabilization of the efficacy (weight) of a synapse according to certain rules relating pre- and post-synaptic activity. For instance, a synapse working according to hebbian rules is stably potentiated when both the pre- and postsynaptic sides are simultaneously active [44]. These theories consider the synapse as an unidimensional device, which varies along a single dimension (depressed-basal-potentiated). However, much evidence accumulated in the last 20 years (see [41]and this paper) favours the view that synapses are multidimensional devices. For instance,

• a synaptic transmitter usually interacts with several receptor subtypes (isoreceptors), thus triggering different intracellular effector cascades;

• in most, if not all, cases, several transmitters are released at a single synapse;

• interactions between these transmissions (different receptors and/or isoreceptors) occur at several levels of the transduction pathways, including interactions between receptors in the cell membrane.

Therefore, the synaptic signal does not simply consist of increased or decreased efficacy of the overall synaptic transmission. Rather each transmission and each transmission interaction is a different signal. As a consequence, the synapse would be able to deliver towards the cell a number of different signals. In this frame, a theory was advanced already at the beginning of the eighties 2, 74about how the existence of membrane receptor–receptor interactions may constitute a molecular mechanisms for memory trace formation. According to this theory, called the receptor mosaic hypothesis of the engram [2], the stabilization of a receptor cluster formed through multiple receptor–receptor interaction of the type discussed above would encode in the synaptic membrane a stable trace of the synaptic activity (Fig. 13) [74]. This view is based on the hypothesis that the stabilization of the receptor cluster at least outlasts the synaptic delay. The existence of these receptor clusters would then result in a biased decoding of new signals and thus a memory trace has been formed at the membrane level representing part of the synaptic weight [74]. Evidence for this phenomenon is still indirect. However, receptors (both metabotropic and ionotropic) are usually confined to specific cell membrane domains 49, 74and often organized in microaggregates 9, 74. Microaggregation is the result of extracellular, intracellular and intramembrane protein–protein interactions. Moreover, some instances of stable dimers or polymers are available in the literature 63, 70.

It is therefore concluded that multiple membrane receptor–receptor interactions through formation of receptor clusters can lead to the storage of information within the membranes. Such molecular circuits can represent hidden layers within the membranes that substantially increase the computational potential of neuronal networks. These molecular circuits are biased as discussed above and may therefore represent part of the molecular mechanism for the storage of memory traces (engrams) in the membranes.