Integrated signaling in heterodimers and receptor mosaics of different types of GPCRs of the forebrain: relevance for schizophrenia

Abstract

Receptor–receptor interactions within receptor heterodimers and receptor mosaics formed by different types of GPCRs represent an important integrative mechanism for signaling in brain networks at the level of the plasma membrane. The malfunction of special heterodimers and receptor mosaics in the ventral striatum containing D₂ receptors and 5-HT_2A receptors in cortical networks may contribute to disturbances of key pathways involving ventral striato-pallidal GABA neurons and mediodorsal thalamic prefrontal glutamate neurons that may lead to the development of schizophrenia. The ventral striatum transmits emotional information to the cerebral cortex through a D₂ regulated accumbal–ventral pallidal–mediodorsal–prefrontal circuit which is of special interest to schizophrenia in view of the reduced number of glutamate mediodorsal–prefrontal projections associated with this disease. This circuit is especially vulnerable to D₂ receptor activity in the nucleus accumbens, since it produces a reduction in the prefrontal glutamate drive from the mediodorsal nucleus. The following D₂ receptor containing heterodimers/receptor mosaics are of special interest to schizophrenia: A_2A–D₂, mGluR5–D₂, CB₁–D₂, NTS₁–D₂ and D₂–D₃ and are discussed in this review. They may have a differential distribution pattern in the local circuits of the ventral striato-pallidal GABA pathway, predominantly located extrasynaptically. Specifically, trimeric receptor mosaics consisting of A_2A–D₂–mGluR5 and CB₁–D₂–A_2A may also exist in these local circuits and are discussed. The integration of receptor signaling within assembled heterodimers/receptor mosaics is brought about by agonists and allosteric modulators. These cause the intramembrane receptor–receptor interactions, via allosteric mechanisms, to produce conformational changes that pass over the receptor interfaces. Exogenous and endogenous cooperativity is discussed as well as the role of the cortical mGluR2–5-HT_2A heterodimer/receptor mosaic in schizophrenia (Gonzalez-Maeso et al. 2008). Receptor–receptor interactions within receptor heterodimer/receptor mosaics of different receptors in the ventral striatum and cerebral cortex give novel strategies for treatment of schizophrenia involving, e.g., monotherapy with either A_2A, mGluR5, CB₁ or NTS₁ agonists or combined therapies with some of these agonists combined with D₂-like antagonists that specifically target the ventral striatum. In addition, a combined targeting of receptor mosaics in the ventral striatum and in the cerebral cortex should also be considered.

Introduction

Receptor–receptor interactions within receptor heterodimers and receptor mosaics of different types of G protein coupled receptors (GPCRs, high order heteromers) represent an important integrative mechanism for signaling that already occurs at the level of the plasma membrane (Agnati et al. 2003, 2007b; Carriba et al. 2008; Fuxe et al. 2003a, 2007a; Gozes 2005; Guo et al. 2008; Marshall 2001, 2005; Maurel et al. 2008; Milligan and Smith 2007)]. During this integrative process, transmitters act at their orthosteric binding sites and induce conformational changes in other receptors through allosteric mechanisms. This integrative process can involve cooperative allosteric interactions when receptor subtypes that recognize the same ligand are involved (Agnati et al. 2007b, 2008; Fuxe et al. 2007a, 2008). In addition, allosteric modulators that act at allosteric modulatory sites of one receptor may also participate in this integrative process of heterodimer and receptor mosaic signaling (Agnati et al. 2006, 2007a; Fuxe et al. 2007a; Milligan and Smith 2007). The latter has been elegantly illustrated through works on homocysteine (Agnati et al. 2006, 2007b). Homocysteine acts as a negative allosteric modulator of dopamine D₂ receptors (D₂) and binds noncovalently to two arginine-rich epitopes of its third intracellular loop through an eletrostatic interaction between arginine and thiol groups of homocysteine (Agnati et al. 2006, 2007b).

In the present review, we will discuss how a dysfunction in special heterodimers and receptor mosaics in the ventral striatum and cortical networks may contribute to malfunction of key pathways involving ventral striato-pallidal γ-aminobutyric acid (GABA) neurons and prefrontal cortical glutamate projections to the ventral tegmental area (VTA). In the VTA, the dopamine (DA) A10 cell groups are located and project into the ventral striatum and the prefrontal and cortical limbic regions (Dahlstroem and Fuxe 1964; Anden et al. 1966; Berger et al. 1976). Such neuronal systems appear to participate in schizophrenia in view of the meso-limbic DA hyperfunction hypothesis as well as the N-methyl-d-aspartic acid (NMDA) glutamate receptor hypofunction hypotheses of schizophrenia in which striatal D₂ receptors play a major role and are targeted in treatment therapies by antipsychotic drugs (Carlsson and Lindqvist 1963; Carlsson 1988; Svensson 2000; Seeman et al. 2006; Fuxe et al. 2007b; Stahl 2007a). D₂ containing heterodimers and receptor mosaics located within such pathways and networks may therefore become important novel targets for treatment of schizophrenia once these are fully understood (Fuxe et al. 2005, 2007a, 2008.

Ventral striatum

The ventral striatum is an important interface between the limbic and motor system (Mogenson et al. 1980; Groenewegen et al. 1996; Groenewegen and Trimble 2007) in which disturbances in this interface may contribute to the development of schizophrenia (Fuxe 1970, Grace 2000, Heimer 2000, Fuxe et al. 2007b). The ventral striatum receives inputs from regions of the prefrontal cortex, the amygdala, and the hippocampus belonging to the limbic system and the midline thalamus. It is also densely innervated by the meso-limbic DA system that originates in the VTA and is involved in the motivational aspects of behavior and in the gating of the flow of emotional information to the cerebral cortex, specifically the prefrontal cortex (Dahlstroem and Fuxe 1964; Fuxe 1965; Anden et al. 1966; Schultz 2002; Fuxe et al. 2007b). A dense DA innervation exists in the majority of the ventral striatum including the nucleus accumbens core and shell, as well as the olfactory tubercle (Fuxe 1965; Fuxe et al. 1979). All of the subregions are rich in both dopamine D₁ receptor (D₁) and D₂ receptors; however, dopamine D₃ receptors (D₃) are specifically found within the nucleus accumbens shell and the islands of Calleja (Joyce and Gurevich 1999; Schwartz et al. 2000).

The efferent projections from the nucleus accumbens core are complex. They reach the subcomissural ventral pallidum as well as the medial part of the globus pallidus interna (GPi) and the pars reticulata of the substantia nigra. The ventral pallidum connects to the subthalamic nucleus becoming part of the indirect pathway from the dorsal striatum (Groenewegen et al. 1993). It is of particular interest that the medial parts of GPi and pars reticulata project back to the medial and orbital prefrontal regions via thalamic relay stations, especially the mediodorsal thalamic nucleus (MD). This completes one accumbal–thalamocortical circuit, since these prefrontal regions project into the nucleus accumbens core (Alexander et al. 1990; Groenewegen et al. 1996).

Instead, the nucleus accumbens shell projects into the ventral and medial parts of the ventral pallidum and the lateral preoptic area. It also innervates the VTA and the dorsal tier of the zona compacta of the substantia nigra, all with high densities of DA cell bodies. In view of nigral DA projections to the dorsal striatum, it illustrates that a spiraling circuitry of ventral striato-nigrostriatal pathways can arise from the nucleus accumbens shell to the dorsolateral striatum (Haber et al. 2000). In fact, cocaine-seeking habits rely on DA-dependent serial connectivity from the ventral striatum to the dorsal striatum for their development (Belin and Everitt 2008). The nucleus accumbens takes control over the dorsal striatal processes via the spiraling striato-nigrostriatal circuitry.

We believe this pathological neuronal mechanism may also exist in schizophrenia where obsessive-compulsive symptoms may develop (Sturm et al. 2003; van Kuyck et al. 2007). In fact, an increase in meso-limbic DA activity is postulated to exist in schizophrenia that subsequently leads to the increased activation of DA receptor subtypes in the nucleus accumbens (Fuxe 1970). Hence, D₂ receptors are the major target for antipsychotic drugs (Seeman et al. 2006). Furthermore, the NMDA receptor hypofunction hypothesis of schizophrenia also exists and is based on the fact that the NMDA receptor antagonist phencylidine can mimic the positive as well as the negative symptoms of schizophrenia (Stahl 2007a). This indicates that many glutamate synapses are dysfunctional with reduced informational processing including those of the prefrontal–nucleus accumbens glutamate system. Thus, the prefrontal cortex can no longer exert its proper executive function over the nucleus accumbens and this deficit will allow the ventral striatum, with the nucleus accumbens being its major structure, to take over control of behavior via its integrative role in emotions and motivation (mainly nucleus accumbens shell) and strong link to the motor systems (mainly nucleus accumbens core). The NMDA receptor hypofunction hypothesis is also supported by observations that the genetics of schizophrenia appear to converge on the NMDA glutamate receptor (Stahl 2007b).

Another major nucleus when discussing the neuroanatomy of schizophrenia is the MD nucleus since a loss of neurons in this nucleus has been reported in those who suffer from schizophrenia (Pakkenberg 1990; Popken et al. 2000; Young et al. 2000). In the study of (Popken et al. (2000), it is of interest to note their discovery of a specific subnucleus loss of nerve cells in MD nucleus with significant reductions in the total neuron number only in the parvocellular (pc) and densocellular (dc) subnuclei. The pc receives inputs especially from the ventral pallidum and sends its glutamate outputs to the dorsal and lateral areas of the prefrontal cortex, which are affected in schizophrenia, showing hypoactivity. These findings are not in agreement with the work of (Danos et al. (2005), showing a reduction in the volume of the MD thalamic nucleus, associated with a lack of reduction in the total neuron number and neuron density. Volume reductions were also observed in another association thalamic nucleus, the medial pulvinar nucleus and in the whole thalamus in schizophrenia (Danos et al. 2003; Danos 2004). Deficiency in other limbic circuits most likely exist in schizophrenia since there is a reduction in the paralbumin positive glutamate/aspartate containing neurons of the mammillary bodies projecting into the anterior thalamus in schizophrenia (Bernstein et al. 2007). It should be noticed, however, that other groups have failed to observe any reductions in the number of MD neurons or in the volume of this nucleus (Cullen et al. 2003; Dorph-Petersen et al. 2004).

Thus, the strength of the MD glutamate projections into the prefrontal cortex may be reduced in schizophrenia. This will bring down the glutamate drive in this region and add to the postulated NMDA receptor hypofunction hypothesis. The D₂ regulated accumbal–ventral pallidal–MD–prefrontal circuit therefore becomes a circuit of special interest in schizophrenia in view of the reduced number and/or activity of glutamate MD–prefrontal projections making this circuit especially vulnerable to D₂ receptor activity in the nucleus accumbens. Taking this into consideration, D₂ receptor activity within the nucleus accumbens will result in a further reduction in the activity of the MD–prefrontal glutamate projections and exacerbate schizophrenic symptoms. However, there is the possibility that certain subpopulations with schizophrenia may not show this neuropathology.

D₂ receptor containing heterodimers/receptor mosaics of different types of receptors in the ventral striatum

A large number of different types of D₂ receptor heterodimers/receptor mosaics appear to exist in the ventral striato-pallidal GABA neurons at the dendritic, terminal and collateral levels (Fuxe et al. 2007a, 2008). Within their local circuits, such heteromers can be found inter alia in the dendritic spines of the striato-pallidal GABA neurons, their GABA collateral terminals and in the cortical glutamate terminals. In the present review, we will deal with the D₂ receptor heterodimers/receptor mosaics that are of special interest in schizophrenia: adenosine A_2A receptor (A_2A)–D₂, metabotropic glutamate receptor 5 (mGluR5)-D₂, cannabinoid receptor 1 (CB₁)–D₂, neurotensin receptor 1 (NTS₁)–D₂ and D₂–D_3,, which may have a differential distribution pattern in the local circuits of the ventral striato-pallidal GABA pathway mainly located extrasynaptically (Agnati et al. 2003; Maggio et al. 2005; Ferre et al. 2007; Fuxe et al. 2007a, 2008; Ferraro et al. 2008). The existence of trimeric receptor mosaics of A_2A–D₂–mGluR5 mainly located extrasynaptically in the dendritic spines of these local circuits has also been proposed (Ferre et al. 2002; Fuxe et al. 2003a, 2007a, 2008; Schwarzschild et al. 2006).

Furthermore, novel evidence inter alia with the recently developed sequential bioluminescence resonance energy transfer (BRET) and fluorescence resonance energy transfer (FRET) techniques indicate the existence of trimeric A_2A–D₂–CB₁ receptor mosaics (Carriba et al. 2008; Marcellino et al. 2008), which may be located mainly extracellularly in the dendritic spines of the ventral striato-pallidal GABA neurons, cortico-accumbal glutamate terminals and the collateral GABA terminals of the local circuits of these neurons as well as in the GABA terminals of the external globus pallidus belonging to the ventral striato-pallidal GABA pathway (Andersson et al. 2005a; Pickel et al. 2006).

The D₂-long isoform of the D₂ receptor is mainly located in the dendrites of the striato-pallidal GABA neurons, while the D₂-short isoform represents the D₂ autoreceptor on DA terminals inhibiting DA release (Khan et al. 1998). Also, the D₂-mediated presynaptic inhibition of striatal glutamate transmission mainly involves the D₂-short receptor, while the presynaptic inhibition of the striatal GABA transmission involves both D₂ isoforms (Centonze et al. 2004). Such differences in distribution of these two D₂ isoforms may contribute to differences in the receptor heterodimer and receptor mosaic assembly and their receptor–receptor interactions depending on their location in the elements of the local circuits and thus in their physiology and pharmacology.

It should be considered that endogenous cooperativity exists in the heterodimer/receptor mosaic assembly process based on the allosteric interactions between the receptor protein monomers as they bind to each other via the receptor interfaces (Agnati et al. 2007b; Fuxe et al. 2008). Cooperativity in macromolecular assembly has recently been discussed and represents a thermodynamic process to drive the macromolecular complex formation and make it more stable than the sum of its components (Williamson 2008). An endogenous cooperativity can explain the constitutive changes in receptor recognition, G protein coupling and selection, and receptor trafficking of the receptors participating in the heterodimer/receptor mosaics in the absence of transmitter ligands and allosteric modulators (Fuxe et al. 2008).

It is therefore clear that D₂ receptors in different constitutive D₂ receptor heterodimers/receptor mosaics can have different functional and pharmacological properties even if they contain the same D₂ isoform. It should be considered that the D₂ partial agonist aripiprazole may be an antipsychotic drug (Tamminga and Carlsson 2002; Novi et al. 2007; Carlsson and Carlsson 2008; Stahl 2008), since it mainly targets the D₂ autoreceptor heterodimer/receptor mosaic on the DA nerve terminals as an agonist. The postjunctional extrasynaptic D₂ heterodimer/receptor mosaics on the terminals and dendrites of the local circuits may be less affected. At these locations, this compound may exert antagonistic actions by its partial agonist activity on these receptor mosaics. Based on this theory, the development of novel antipsychotic D₂ antagonists could also be based on targeting the different postjunctional extrasynaptic D₂ containing heterodimers/receptor mosaics in the local circuits of the ventral striato-pallidal GABA neurons. This may reduce the positive symptoms of schizophrenia with reduced extrapyramidal side effects and cognitive deficits related to blockade of dorsal striatal and cortical D₂ receptors, respectively (Tamminga and Carlsson 2002; Del Arco et al. 2007; Stahl 2007a). The endogenous cooperativity in the receptor mosaic should however not only be discussed in terms of allosteric cooperativity, but also in terms of configurational cooperativity (Whitty 2008). The receptor proteins may be regarded as polyvalent ligands, which contain multiple binding epitopes, similar or different, binding the other receptor. As discussed by (Whitty (2008) the first binding interaction makes the subsequent binding site interactions more easily performed if the receptor geometries are suitable for an interaction. If so, an enhanced binding is observed since the initial binding interaction pre-organizes the other binding sites for receptor protein interactions by reducing the number of inefficient configurations, making possible binding interactions at reduced entropic cost. The multiple interactions between the two receptor proteins may also cooperatively increase the enthalpic stability of all the binding interactions between the receptor proteins (Williams et al. 2004).

The integration of receptor signaling in the assembled heterodimer and receptor mosaic of different types of GPCRs is instead brought about by exogenous agonist ligands and allosteric modulators not belonging to the receptor protein. This causes the intramembrane receptor–receptor interactions via allosteric mechanisms, with conformational changes passing over the receptor interfaces. Exogenous cooperativity develops when similar subtypes of receptors having the same transmitter directly interact (Agnati et al. 2007b; Fuxe et al. 2007a, 2008; Agnati et al. 2008). The functional outcome of the integrative process with emergent function is determined by the stoichiometry, topology, agonist concentrations and receptor recognition properties determining the rank order of receptor activation and how this process becomes altered by allosteric modulators and the microenvironment (Agnati et al. 2007b, 2008, Fuxe et al. 2007a, 2008).

The D₂–D₃ heterodimer/receptor mosaic

D₂–D₃ heterodimers have been discovered in cellular models (Scarselli et al. 2001; Maggio et al. 2003) and are of particular interest for schizophrenia since they are colocated in ventral striato-pallidal GABA neurons, especially in the nucleus accumbens shell and in limbic DA terminals (Surmeier et al. 1996; Gurevich and Joyce 1999; Joyce and Gurevich 1999; Diaz et al. 2000; Schwartz et al. 2000). It is likely that in DA terminals, the D₃ and the D₂-short autoreceptor form a heterodimer/receptor mosaic and together modulate DA release (Diaz et al. 2000; Millan et al. 2000; Sokoloff et al. 2006). Instead, in ventral striatal nerve cells, D_2L–D₃ heterodimers/receptor mosaics may be formed. Specific D₃ antagonists have also been developed and proposed to be antipsychotic drugs (Millan et al. 2000; Schwartz et al. 2000; Joyce and Millan 2005).

The actions of the partial D₂ agonist antipsychotic aripiprazole as well as the partial D₂ agonist candidate antipschotics including S33592, bifeprunox, N-desmethylclozapine, and preclamol have recently been characterized with regard to potential D_2L–D₃ heterodimer/receptor mosaics in cellular models (Novi et al. 2007). An interesting finding was obtained that the partial D₂ agonist actions of these compounds on D_2L receptors were counteracted in the D_2L–D₃ heterodimer/receptor mosaics due to a reduced coupling efficiency in the D_2L receptor brought about by the allosteric interaction in the D₂–D₃ interface. Instead, they behaved as D₂ antagonists at the D_2L–D₃ heterodimer/receptor mosaic. This analysis was made possible by the use of COS-7 cells co-transfected with a D₃ receptor-insensitive, chimeric adenylate cyclase (AC)-V/VI (Novi et al. 2007). These results indicate that one reason for the putative antipsychotic effects of the partial D₂ agonists may be that in the ventral striatum, they exert mainly antagonist action at the D_2L receptor due to the substantial existence of postjunctional D_2L–D₃ heterodimers/receptor mosaics canceling the partial D₂ agonist actions of these candidate antipsychotics. Instead in the dorsal striatum, such heterodimers/receptor mosaics hardly exist, and therefore the partial D₂ agonist effects can be exerted at the postjunctional D₂ receptors, leading to reduced extrapyramidal side effects due to improved motor function via increases in the D₂ receptor activity.

The A_2A–D₂ heterodimer/receptor mosaic

The A_2A–D₂ heterodimer/receptor mosaic was first indicated from the demonstration of antagonistic intramembrane receptor–receptor interactions through biochemical binding techniques in striatal membranes (Ferre et al. 1991; Fuxe et al. 1993). Further evidence for its existence were demonstrated using resonance energy transfer techniques (BRET and FRET) and co-immunoprecipitation experiments (Hillion et al. 2002; Canals et al. 2003; Kamiya et al. 2003). The zones of receptor interactions were mapped and modeled and important epitope–epitope electrostatic interactions between negatively charged epitopes of the A_2A (specifically involving a phosphorylated serine in its C terminal tail) and positively charged epitopes (rich in arginine residues) in the intracellular loop (IC) 3 of the D₂ receptor were discovered (Ciruela et al. 2004). For reviews, see Fuxe et al. (2005, 2007a, 2008).

The recent discovery of higher order oligomers of D₂ receptors in living mammalian cells at physiological expression levels is highly interesting. These were demonstrated through the use of bioluminescence/fluorescence complementation together with resonance energy transfer (Guo et al. 2008). Likewise, higher order A_2A receptor oligomers have been observed using combined BRET and bimolecular fluorescence complementation techniques (Gandia et al. 2008). Therefore, A_2A–D₂ heterodimers that display antagonistic interactions should be considered with D₂ coupled to G_i or G_q/11 to regulate AC and phospholipase C (PLC), linked to inhibition of L-type voltage-dependent calcium channels (VDCC) (Ferre et al. 2007, Fuxe et al. 2007a, Fuxe et al. 2007d, Fuxe et al. 2008). Receptor mosaics of A_2A and D₂ receptors with an unknown stoichiometry and topology, where A_2A may antagonize D₂ signaling to G_i and G_q/11, should also be considered. In such a receptor mosaic the D₂ receptors via G_i in turn inhibit the A_2A-activated AC keeping the A_2A signaling at a low basal level.

The A_2A–D₂ heterodimer/receptor mosaic is mainly located extrasynaptically on the dendritic spines of the local circuits of the ventral and dorsal striato-pallidal GABA neurons. These receptor mosaics may also exist on either glutamate terminals and/or collateral GABA terminals within these local circuits [for reviews, see Ferre et al. (2007), Fuxe et al. 2007a, 2007d, 2008]. Recently, further studies have been performed to obtain a physiological correlate to the antagonistic A_2A–D₂ receptor interactions in this receptor mosaic by pairing afferent stimulation with postsynaptic spikes in a burst mode (Shen et al. 2008). In D₂ positive striato-pallidal neurons, long-term potentiation (LTP) was induced when spiking was induced 5 ms after the synaptic stimulation, while long-term depression (LTD) developed when postsynaptic spiking was induced prior to synaptic stimulation. The D₂ antagonist sulpiride blocked LTD underlining the importance of D₂ receptors for LTD. In contrast, the A_2A antagonist SCH58261 counteracted the introduction of LTP in the D₂ positive striato-pallidal nerve cells (Shen et al. 2008). This may be at least in part explained by the removal of the tonic A_2A reduction of postjunctional D₂ signaling through the A_2A/D₂ receptor mosaic (Fuxe et al. 2007d). The increased D₂ activity will not allow the increased excitability to be maintained via its gating effects on currents over the voltage-dependent calcium channels (Cav), inward rectifier potassium channels (Kir), a glutamate-gated ion channel that is permeable to Na⁺ and K⁺ (AMPA) and voltage-dependent sodium channels (Nav), favoring the silencing of the D₂ striato-pallidal neurons (Surmeier et al. 2007).

In line with these results, the D₂ agonist quinpirole transformed the normal LTP development into LTD, and when the A_2A agonist CGS 221680 and quipirole were co-administered LTP was restored (Shen et al. 2008). These results can also be explained by the antagonistic A_2A–D₂ receptor interaction in the A_2A–D₂ receptor mosaic, markedly reducing D₂ signaling and setting free the A_2A signaling by removal of the inhibitory tone of D₂ via G_i on the A_2A activated AC mediated via G_olf (Fuxe et al. 2007c). It seems likely that the resulting A_2A-mediated activation of protein kinase A (PKA) contributes to the excitability and upstate (depolarized plateau potential favoring the elicitation of action potentials) of this neuron system and restoration of LTP by phosphorylation of L-type voltage-dependent calcium channels, dopamine and cyclic AMP-regulated phosphoprotein with molecular weight 32 kDa (DARPP-32) at threonine 34 and of GluR1–AMPA receptors playing an important role in synaptic plasticity (Hernandez-Lopez et al. 2000, Hakansson et al. 2006, Bateup et al. 2008). The strength of the A_2A agonist action in inducing LTP is evident by its ability to do so in the presence of a D₂ receptor antagonist using a protocol that should not result in LTP (postsynaptic spiking prior to synaptic activation) (Shen et al. 2008). These results by (Shen et al. 2008) therefore indicate that the A_2A–D₂ receptor–receptor interactions in receptor mosaics play an important role in striatal synaptic plasticity, contributing to making it bidirectional and Hebbian in the striato-pallidal GABA neurons.

In animal models of Parkinson’s disease (PD) with severe degeneration of DA terminals, abnormal striatal plasticity was observed, since with the protocol normally inducing LTD, LTP was observed (spiking prior to synaptic activation) (Shen et al. 2008). This LTP was blocked by A_2A antagonists underlining the view of a dominance of A_2A over D₂ signaling in the A_2A–D₂ receptor mosaic in PD (Fuxe et al. 2007c). The impact of D₂ for LTD was again demonstrated in the PD model by recruiting LTD with a D₂ agonist. The results underline the value of introducing A_2A antagonists as antiparkinsonian drugs (Schwarzschild et al. 2006; Fuxe et al. 2007d, 2008).

On the basis of the existence of the antagonistic A_2A–D₂ receptor interactions in the ventral striato-pallidal GABA pathway, A_2A agonists were early on proposed to be atypical antipsychotic drugs (Ferre et al. 1994, Fuxe et al. 2007a, 2008). In fact, A_2A agonists show an atypical antipsychotic profile in the rat amphetamine and phencyclidine model of schizophrenia (Rimondini et al. 1997) and in the Cebus apella monkey model of schizophrenia (Andersen et al. 2002). A glutamate releasing action in the nucleus accumbens may also contribute to its atypical antipsychotic profile (Fuxe et al. 2008). Recently, evidence has been obtained that the antagonistic A_2A–D₂ receptor interactions importantly modulate the information flow in the neuronal circuit from the nucleus accumbens to the prefrontal cortex via the ventral pallidum and the mediodorsal thalamic nucleus. The antipsychotic activity of A_2A agonists appears to mainly rely on the ability of these compounds to counteract the D₂ receptor-induced reduction of the glutamate drive from the mediodorsal thalamic nucleus to the prefrontal cortex via their reduction of D₂ signaling in the nucleus accumbens (Fuxe et al. 2008). It brings the A_2A agonists targeting the A_2A–D₂ receptor mosaic into the center of the D₂ and NMDA hypothesis of schizophrenia where combined treatment with low doses of A_2A and D₂ antagonists may be an optimal treatment to avoid collateral effects (Fuxe et al. 2008).

A_2A–D₃ heterodimers/receptor mosaics and putative A_2A–D₂–D₃ receptor mosaics

A_2A–D₃ heterodimers/receptor mosaics have been demonstrated in cellular models and exhibit an antagonistic receptor interaction where A_2A agonists reduce both D₃ receptor agonist binding and D₃ receptor signaling (Torvinen et al. 2005). It is therefore possible that such receptor complexes exist in the ventral striatum where A_2A and D₃ receptors are co-distributed, especially in the nucleus accumbens. A_2A agonists may therefore exert their antipsychotic effects in animal models of schizophrenia by targeting these receptor mosaics and producing a reduction in D₃ receptor signaling (Fuxe et al. 2008), since D₃ antagonists may have antipsychotic actions (Schwartz et al. 2000; Joyce and Millan 2005). D₃ tetramers have been found in the brain (Nimchinsky et al. 1997), opening up the possibility for the existence of A_2A/D₃ receptor mosaics in the ventral striatum.

In view of the likely existence of D₂–D₃ heterodimer/receptor mosaic in the ventral striatum, it seems possible that A_2A–D₂–D₃ receptor mosaics exist in the ventral striatum, especially in the nucleus accumbens, modulating the local circuits of the ventral striato-pallidal GABA pathway, mainly originating in the nucleus accumbens shell and core (Fuxe et al. 2008). It will be of interest to evaluate if in fact such complex receptor mosaics may be a target of antischizophrenic drugs such as D₂ antagonists and of antipsychotic candidates such as D₃ antagonists and A_2A agonists.

Putative A_2A–D₂–mGluR5 receptor mosaics

The first indications for the existence of glutamate receptor–D₂ receptor complexes was obtained in 1984 through the demonstration that L-glutamate could reduce the affinity of the D₂ agonist binding sites in striatal membrane preparations (Fuxe et al. 1984). Later on, mGluR5 was demonstrated as the glutamate receptor involved in mediating this antagonistic receptor–receptor interaction in putative mGluR5–D₂ receptor heterodimer/receptor mosaics (Ferre et al. 1999; Popoli et al. 2001). Furthermore, combined activation of A_2A and mGluR5 receptors enhanced the reduction of the affinity of the D₂ agonist binding sites (Popoli et al. 2001) and coimmunoprecipitation experiments showed the existence of A_2A–mGluR5 heteromeric complexes (Ferre et al. 2002; Fuxe et al. 2003a). Together with the chemical anatomical analysis (Fuxe et al. 2003a, Ferre et al. 2004, Conn et al. 2005), these results indicated the existence of extrasynaptic A_2A–D₂–mGluR5 receptor mosaics located mainly not only on the dendritic spines of the local circuits of the ventral and dorsal striato-pallidal GABA neurons, but also on the cortico-striatal glutamate terminals (Fig. 1).

Fig. 1

Illustration of the possible existence of CB₁–D₂–A_2A and A_2A–D₂–mGluR5 receptor mosaics in different local circuits of the accumbal-pallidal GABA neurons projecting into the ventral pallidum reached by volume transmission signals. The extrasynaptic CB₁–D₂–A_2A receptor mosaic is inter alia located on the dendritic spines of these neurons and on the incoming synaptic glutamate terminals to these spines. Activation of the CB₁ receptors with CB₁ agonists in this receptor mosaic leads to inhibition D₂ recognition and signaling and removal of D₂–G_i-mediated inhibition of A_2A activated adenylate cyclase with a release of its brake on A_2A signaling as shown in the upper left (1). The postulated extrasynaptic A_2A–D₂–mGluR5 receptor mosaic is illustrated in another local circuit of the same neurons and mainly in the dendritic spines, but may also be located in the same local circuit as you find the CB₁–D₂–A_2A receptor mosaic. Combined activation of the A_2A and mGluR5 receptors in this receptor mosaic leads to synergistic inhibition of D₂ recognition and signaling as illustrated in the upper right (2)

In this receptor mosaic, A_2A and mGluR5 co-activation appears to synergistically increase c-fos expression as well as extracellular signal-regulated kinase (ERK) and DARPP-32 phosphorylation (Ferre et al. 2002; Nishi et al. 2003). A_2A and mGluR5 co-activation counteracts D₂ recognition and signaling, probably resulting in an activation of the ventral and dorsal striato-pallidal GABA pathways. This has in fact been shown with dual-probe microdialysis showing that coactivation of mGluR5 and A_2A receptors in the nucleus accumbens leads to a synergistic rise of GABA release in the ventral pallidum (Diaz-Cabiale et al. 2002b). Behavioral evidence also exists that the A_2A agonist CGS21680 and the mGluR5 agonist (RS)-2-chloro-5-hydroxy-phenylglycine (CHPG) synergize in counteracting phencylidine-induced motor activity (Ferre et al. 2002). Based on this work it was proposed that combined treatment with low doses of A_2A and mGluR5 agonists with or without treatment with low doses of D₂ antagonists may be an interesting new strategy for treatment of schizophrenia (Fuxe et al. 2008). The finding that LTD in striato-pallidal GABA nerve cells highly dependent on D₂ receptors is antagonized by mGluR5 antagonists (Shen et al. 2008) can, however, not be explained on the basis of this hypothesis, but probably is caused by a reduction of retrograde endocannabinoid signaling being enhanced by mGluR5 coupled to G_q/11 (Uchigashima et al. 2007).

Parallel to this work, it has been shown that mGluR5 activation enhances NMDA currents (Conn et al. 2005). In view of the NMDA hypofunction hypothesis of schizophrenia, it has therefore been suggested that mGluR5 agonists and positive allosteric modulators produce antipsychotic actions by enhancing synaptic NMDA signaling (Kinney et al. 2003, 2005; Conn et al. 2005). It has been shown that mGluR5 agonist counteracts the amphetamine-induced reduction in prepulse inhibition and the mGluR5 antagonist 2-methyl-6-(phenylethynyl)-pyridine (MPEP) enhances the phencyclidine (PCP)-induced disruption of prepulse inhibition (Kinney et al. 2003).

Taken together, it seems likely that the mechanism underlying the antipsychotic-like actions of mGluR5 agonists and modulators involves in part an inhibition of D₂ signaling in the extrasynaptic A_2A–D₂–mGluR5 receptor mosaics located in the ventral striato-pallidal GABA neurons. However, it also involves an enhancement of synaptic NMDA signaling via indirect mGluR5–NMDA receptor interactions in the postsynaptic density in several brain regions, including the ventral striatum and the prefrontal cortex.

Receptor mosaic of D₂ and NMDA receptors

A direct interaction between the N-terminal portion of the third intracellular loop of the D₂ receptor and a C-terminal region of the NR2B subunit occurs and takes place within the microdomains of the postsynaptic densities of striatal glutamate synapses (Liu et al. 2006). This receptor interaction appears to be increased through the activation of the D₂ receptors and subsequently leads to reduced NMDA receptor signaling by a reduction in the phosphorylation of NR2B through the counteraction of Ca²⁺/calmodulin-dependent protein kinase II binding to the NR2B receptor. Thus, D₂ receptors at glutamate synapses on ventral striato-pallidal GABA neurons may assist in silencing these neurons through direct interactions with NR2B containing NMDA receptors. This receptor mosaic represents an interesting novel target for antischizophrenic drugs including D₂ antagonists and/or partial D₂ agonists. It is possible that synaptic mGluR5 receptors may, in part, increase NR2B–NMDA signaling through its antagonistic interaction with D₂ receptors [see above and Fuxe et al. (2008)]. It should also be noted that timing-dependent LTP in the striato-pallidal GABA neurons is blocked by NMDA antagonists and the antagonistic D₂–NR2B–NMDA interaction helps understand the crucial role of D₂ receptors for the development of LTD in these neurons (Shen et al. 2008).

CB₁–D₂ heterodimer/receptor mosaics and CB₁–D₂–A_2A receptor mosaics

Indications for the existence of CB₁–D₂ heterodimers/receptor mosaics were first obtained in 2003 (Fuxe et al. 2003b). In 2005, coimmunoprecipitation of CB₁ and D₂ receptor was found in cellular models (Kearn et al. 2005) and resonance energy transfer (FRET) tecniques gave further evidence for the existence of CB₁–D₂ receptor heterodimers/receptor mosaics in cell lines in 2008 (Marcellino et al. 2008). Antagonistic CB₁–D₂ receptor interactions were demonstrated at the level of D₂ recognition, where reduced D₂ agonist affinity in ventral and dorsal striatum was found upon CB₁ stimulation (Marcellino et al. 2008). It is of special interest that CB₁ signaling appears to switch from a G_i-mediated inhibition of AC to a G_s/G_olf-mediated activation of AC upon concurrent activation of both CB₁ and D₂ receptors in cellular models (Glass and Felder 1997; Kearn et al. 2005). In the ventral striatum, CB₁ and D₂ receptors are colocated in dendrites, terminals and local collaterals of the ventral striato-pallidal GABA neurons as well as in the cortico-striatal glutamate terminals (Pickel et al. 2006; Uchigashima et al. 2007). In the dorsal striatum, their colocation is mainly found in the terminals and local collaterals of dorsal striato-pallidal GABA neurons.

The antagonistic CB₁–D₂ receptor interaction may exist in CB₁–D₂ heterodimer/receptor mosaics located in different regions of the striato-pallidal GABA neurons (Maneuf et al. 1997). This antagonistic interaction may be part of an inhibitory feedback mechanism aimed to counteract strong D₂ signaling. Evidence for this includes a marked increase in extracellular levels of anandamide following D₂ agonist administration as well as D₂ regulation of 2-arachidonoyl-glycerol (2-AG) signaling (Giuffrida et al. 1999; Piomelli 2003; Uchigashima et al. 2007). Therefore, the two major endocannabinoids are increased by D₂ receptor activation. Via volume transmission and the antagonistic CB₁–D₂ receptor interaction, these increases in endocannabinoids makes possible the counteraction of excessive or aberrant D₂ signaling, thus restoring the activity of striato-pallidal GABA neurons and reducing motor activation (Rodriguez de Fonseca et al. 1998; Fuxe et al. 2008; Marcellino et al. 2008).

In the glutamate terminals, our hypothesis is that CB₁ receptor homodimers (Mackie 2005) may dominate over CB₁–D₂ receptor heteromers and lead to CB₁-mediated inhibition of glutamate release (Piomelli 2003). In fact, in D₂ containing striato-pallidal nerve cells, timing-dependent LTD is not able to develop after the blockade of CB₁ receptors on the glutamate terminals, demonstrating that these receptors are required for the reduction of glutamate release (Shen et al. 2008). These CB₁ receptors mediate the retrograde endocannabinoid signaling through a postsynaptic release step (Adermark and Lovinger 2007). In PD models, it has recently been shown that endocannabinoids can rescue striatal LTD in the indirect pathway demonstrated through the use of inhibitors of endocannabinoid degradation (Kreitzer and Malenka 2007). It was also found that combined treatment with D₂ agonists and such inhibitors rescued motor deficits in the PD models used.

One explanation for these findings is that CB₁ receptors mainly exist as homomers on the glutamate terminals and act to reduce glutamate release, thereby enhancing motor function when activated. Thus, CB₁ may preferentially heteromerize with D₂-long since D₂-short are found on glutamate terminals (Centonze et al. 2004). Alternatively, the CB₁ receptor in the heterodimer/receptor mosaic may have a lower affinity than CB₁ homodimers. These are possible explanations for the indicated dominance of the CB₁ homomer signaling on the glutamate terminals with inhibition of glutamate release and induction of LTD in the striato-pallidal GABA neurons.

The antagonistic CB₁–D₂ receptor interaction may also occur at the postjunctional level in the striato-pallidal GABA neurons where the CB₁–D₂ heterodimer/receptor mosaic may exist especially at the terminal level. This inhibitory feedback may only take place in the presence of a highly exaggerated D₂ signaling resulting in a higher release of endocannabinoids. Such events may take place due to the postulated reduced affinity of the CB₁ receptor in the heteromer versus the homomer. This inhibitory feedback may become effective not only by reducing D₂ recognition and signaling, but also because the CB₁ receptor in the CB₁–D₂ heterodimer/receptor mosaic may be coupled to G_olf. This will further help switch the indirect pathway towards excitation, which can also involve increased glutamate release (Fuxe et al. 2008).

In recent years, indications have also been obtained for the existence of possible CB₁–A_2A–D₂ receptor mosaics in the terminal parts of the dorsal striato-pallidal GABA neurons and in the dendritic regions of the ventral striato-pallidal GABA neurons and in cortico-striatal glutamate terminals (Rosin et al. 1998; Shindou et al. 2002; Pickel et al. 2006; Simola et al. 2006; Fuxe et al. 2008). In striatal slices, CB₁ agonists can produce an increase in PKA-mediated DARPP-32 phosphorylation through activation of A_2A receptors (Andersson et al. 2005b; Borgkvist et al. 2007). A_2A–CB₁ heterodimer/receptor mosaics have been demonstrated in cell lines in which CB₁ signaling depends on A_2A signaling (Carriba et al. 2007) and A_2A antagonists can counteract the ability of CB₁ agonists to antagonize the D₂ agonist-induced motor activation (Marcellino et al. 2008). Furthermore, in the brain there exists a high molecular weight form of CB₁ devoid of G proteins, which may represent a trimeric CB₁–A_2A–D₂ receptor mosaic (Mukhopadhyay et al. 2000). Using sequential BRET–FRET, Franco et al. (Carriba et al. 2008) have this year in fact identified trimeric receptor mosaics of cannabinoid CB₁, dopamine D₂ and adenosine A_2A receptors in living cells (Fig. 1).

Based on these findings, the hypothesis has been introduced that the molecular basis for the motor inhibition produced by CB₁ agonists is through the specific target of these trimeric receptor mosaics. Within these receptor mosaics, you can set free the A_2A signaling to AC from D₂ inhibition via the antagonistic CB₁–D₂ receptor interactions reducing D₂ signaling over G_i (Marcellino et al. 2008). Also in this trimeric receptor mosaic, the CB₁ receptor may be coupled to G_olf and not to G_i. These receptor mosaics probably represent important integrators of adenosine, DA and endocannabinoid signals and probably exist also in the cortico-striatal glutamate terminals.

The ventral striato-pallidal GABA pathway is of special interest, since here the CB₁ receptors have been demonstrated also in the dendrites including the dendritic spines where they are colocated with A_2A and D₂ receptors (Pickel et al. 2006). Thus, CB agonists may more effectively inhibit D₂ signaling in the ventral striato-pallidal GABA pathway than in the dorsal striato-pallidal GABA pathway, giving them a potential atypical antipsychotic profile. However, the CB₁ agonists should preferentially target the CB₁/D₂ heterodimer/receptor mosaic and the trimeric A_2A–D₂–CB₁ receptor mosaic in the ventral striato-pallidal GABA pathway and the cortico-striatal glutamate terminals, which probably exhibit a special CB₁ pharmacology versus the CB₁ homodimers/receptor mosaics. The CB₁ agonist delta-9-tetrahydrocannabinol (Δ⁹-THC) may exacerbate psychotic symptoms in schizophrenia (D’Souza et al. 2005) because of a high affinity for CB₁ homodimers/receptor mosaics. This would result in a reduction of glutamate release inter alia in the prefrontal cortex and the nucleus accumbens and reduced activity in the ventral striato-pallidal GABA pathway. Instead, cannabidiol, a weak CB₁ antagonist appears to be an antipsychotic drug (Zuardi et al. 2006a, 2006b. Based on the present findings, it is postulated that this action of cannabidiol is produced by its ability to act as a preferential CB₁ agonist at CB₁/D₂ heterodimers/receptor mosaics and at trimeric A_2A–D₂–CB₁ receptor mosaics. In line with this hypothesis, it has been reported that in schizophrenic patients the anandamide cerebrospinal fluid (CSF) levels are inversely correlated to psychotic symptoms, probably due to the activation of the CB₁ receptors in the mentioned D₂ containing receptor mosaics (Leweke et al. 2007). For further discussion, see Fuxe et al. (2008).

Based on the likely existence of the CB₁–D₂–A_2A receptor mosaic in the ventral striatum, it may be suggested that combined treatment with A_2A agonists and CB₁ agonists preferentially acting on this receptor mosaic may be an interesting novel strategy for treatment of schizophrenia, especially in combination with low doses of typical and atypical D₂-like antagonists.

Putative NTS₁–D₂ heterodimers/receptor mosaics

In 1980, it was proposed that Neurotensin (NT) may be an endogenous antipsychotic based on antatomical and neurochemical evidence of NT/DA interactions (Nemeroff 1980). In agreement, low CSF NT concentrations have been found in a subset of drug-free schizophrenic patients (Breslin et al. 1994; Sharma et al. 1997). Furthermore, clinical improvement was associated with normalization of CSF NT, and low CSF NT concentrations were positively correlated with severity of thought disorder, deficit symptoms, etc. (Caceda et al. 2006). Also NT receptor agonists, similar to known antipsychotic drugs, prevent amphetamine and dizocilpine-induced disruption of prepulse inhibition of the startle response in the rat (Feifel et al. 1999).

Discovered antagonistic NTS₁–D₂ receptor interactions in postulated NTS₁–D₂ heterodimers/receptor mosaics in the ventral striatum have been repeatedly proposed to be the molecular mechanism involved in the postulated antipsychotic effects of NT (Agnati et al. 1983; Tanganelli et al. 1989; Fuxe et al. 1992; Diaz-Cabiale et al. 2002a; Antonelli et al. 2007; Ferraro et al. 2007). It should be noted, however, that NT shows antipsychotic-like behavioral effects in the nucleus accumbens, but stimulant-like effects in the VTA (Kalivas et al. 1981, 1984. The mechanism for such dual actions can be explained by the probable existence of antagonistic postjunctional NTS₁–D₂ interactions in the nucleus accumbens that reduce postjunctional D₂–mediated signaling and antagonistic NTS₁–D₂ autoreceptor interactions on DA nerve cells in the VTA that increase the firing of meso-limbic DA neurons. This regional difference can also explain why systemic NTS₁ receptor antagonists demonstrate antipsychotic-like effects, namely by preferentially blocking the NTS₁–D₂ autoreceptor interaction. This will in turn lead to an increased D₂ autoreceptor function and reduced activity of meso-limbic DA neurons (Dobner et al. 2003; Ferraro et al. 2007)]. The antagonistic NTS₁–D₂ receptor interaction was first shown by the reduction of D₂ receptor agonist affinity in striatal membrane preparations (Agnati et al. 1983; von Euler 1991) and in striatal sections (Li et al. 1994).

An in vivo functional correlate to the antagonistic NT effects on D₂ recognition was obtained from microdialysis studies (Tanganelli et al. 1989, 1994; Fuxe et al. 1992; Antonelli et al. 2007, Ferraro et al. 2007). These experiments indicated for the first time the existence of a D₂ autoreceptor heteromer on DA nerve terminals in the dorsal striatum, namely the NTS₁–D₂ autoreceptor heterodimer/receptor mosaic, where NT was able to restore DA release after D₂ autoreceptor activation (Tanganelli et al. 1989; Diaz-Cabiale et al. 2002a). On the other hand, the postjunctional antagonistic NTS₁/D₂ receptor interaction in the dorsal striatum predominantly takes place in cortico-striatal glutamate terminals and results in an increase in glutamate release. The enhancement of glutamate transmission may be responsible for the subsequent activation of the dorsal striato-pallidal GABA neurons thus producing an increase in GABA outflow (Fuxe et al. 1992; Antonelli et al. 2007) and may contribute to the motor inhibition produced by NT treatment. This fact may underlie the use of NTS₁ antagonists as a novel strategy for the treatment of PD (Antonelli et al. 2007). The D₁ receptors are not modulated by the NT peptides and therefore a switch towards D₁-mediated DA transmission is obtained, which is facilitated by the antagonistic NTS₁–D₂ autoreceptor interaction favoring DA release (Fuxe et al. 1992, Antonelli et al. 2007).

In contrast to the dorsal striatum, accumbal DA terminals did not exhibit a NTS₁–D₂ autoreceptor interaction (Tanganelli et al. 1994). Rather in the nucleus accumbens, the major postjunctional antagonistic NTS₁–D₂ receptor interaction is again linked to the glutamate terminals and leads to an enhancement of glutamate release in view of the inhibitory role D₂ receptors play on glutamate release (Ferraro et al. 2007). This increase in glutamate outflow will thereby activate the ventral striato-pallidal GABA pathway and increase GABA outflow from terminals and collaterals that will reduce accumbal DA release (Tanganelli et al. 1994). In this way, NT agonists will more strongly reduce D₂-mediated DA transmission in the nucleus accumbens, which will give them an atypical antipsychotic profile. The increase in glutamate release can also explain why NT peptides can block the D₂ agonist-induced reduction of GABA release (Ferraro et al. 2007).

The mGluR2–5-HT_2A heterodimer/receptor mosaic in the cerebral cortex

In 1968–1974, we published a series of papers demonstrating that d-lysergic acid diethylamide (d-LSD), hallucinogenic drugs of the indolalkylamine type, and hallucinogenic phenylethylamines produced reductions of 5-hydroxytryptamine (5-HT) turnover. This reduction of 5-HT turnover was associated with an increase in the extensor hind limb reflex activity of the acutely spinalized rat. The reflex activity was known to be induced through a postjunctional 5-HT receptor (Anden et al. 1968; Anden et al. 1971, 1974; Fuxe et al. 1972). In the pharmacological analysis, all of these hallucinogenic drugs were determined to be postsynaptic 5-HT receptor agonists. We proposed that the hallucinogenic actions of these compounds were produced via a postsynaptic activation of a 5-HT receptor in the brain, which produced a compensatory reduction in the firing of the 5-HT raphe neuron systems that in turn led to the reduced 5-HT turnover. These effects were blocked by methergoline, a putative 5-HT receptor antagonist (Fuxe et al. 1976).

Instead, Aghajanian et al. (1968) proposed that the hallucinogenic effects of d-LSD was produced by the activation of a presynaptic 5-HT receptor on the raphe nerve cells. The receptor was very sensitive to d-LSD and other hallucinogens, and thereby caused a reduction in the firing of the 5-HT neuron systems that mediated the hallucinogenic effects. The doses of hallucinogens we used to produce postjunctional 5-HT receptor activation were considerably higher. In line with these results, we observed, with Dr. Everitt, biphasic dose response curves with d-LSD and other hallucinogens on lordosis behavior of estrogen-primed ovariectomized rats (Fuxe et al. 1976). The ascending 5-HT neuron systems play an inhibitory role in sexual behavior (Everitt et al. 1975). Therefore, as expected, a low dose (5–20 μg/kg) of d-LSD was associated with an increase in sexual behavior, indicating a presynaptic 5-HT receptor activation. Inhibition was observed with a higher dose (40 μg/kg) indicating postsynaptic 5-HT receptor activation (Fuxe et al. 1976). However, no evidence was obtained to indicate that the hallucinogenic effects were abolished with increasing doses of hallucinogens. Hence the postsynaptic 5-HT receptor hypothesis was therefore maintained, especially since antidepressants were also shown to reduce the firing of the ascending 5-HT neuron system.

Today, the postjunctional 5-HT receptor has been identified as the 5-hydroxytrypamine receptor 2A (5-HT_2A) receptor (Lieberman et al. 1998) and high densities of these binding sites have been found all over the cerebral cortex, especially in layer IV where the sensory inputs terminate (Palacios 1983; Cummins et al. 1987). Immunohistochemistry of 5-HT_2A has indicated that 5-HT_2A receptors are located extrasynaptically in large populations of pyramidal nerve cells, especially along the apical dendrites, and also in GABA interneurons distributed over the cerebral cortex. They are reached by 5-HT through volume transmission with a minimal distance of 7 μm between the 5-HT varicosities and the 5-HT_2A receptors located on the apical dendrites of the pyramidal cells (Jansson et al. 2001).

Certain antipsychotic drugs, such as risperidone, have now been identified by their high affinity for the 5-HT_2A receptor that results in a blockade of 5-HT_2A activity (Lieberman et al. 1998, Colpaert 2003, Miyamoto et al. 2005). Psilocybine induces schizophrenia-like psychosis in humans through the activation of 5-HT_2A receptors (Vollenweider et al. 1998). The effects of d-LSD and psilocybine and other hallucinogenic drugs appear to resemble core symptoms of schizophrenia (Colpaert 2003). Evidence indicates that hallucinogens act by recruiting specific cortical 5-HT_2A receptor-mediated signaling pathways (Gonzalez-Maeso et al. 2007), which non-hallucinogenic 5-HT_2A agonists cannot activate. This can also explain the observations that you can distinguish between hallucinogenic and non-hallucinogenic 5-HT_2A agonists by transcriptome fingerprints (Gonzalez-Maeso et al. 2003).

It is a major discovery that a cortical mGluR2–5-HT_2A heterodimer/receptor mosaic may have been identified and implicated in psychosis (Gonzalez-Maeso et al. 2008). In fact, metabotropic glutamate 2/3 receptors have been regarded as targets for antischizophrenic drugs (Aghajanian and Marek 2000; Marek 2004). In addition, a positive allosteric modulator of mGluR2, biphenyl-indanone A is effective in a hallucinogenic drug model of psychosis (Benneyworth et al. 2007). Furthermore, a randomized phase II clinical trial in schizophrenia using a mGluR2/3 agonist indicates their therapeutic potential in this disease (Patil et al. 2007).

Fluorescent in situ hybridization (FISH) has demonstrated colocation of 5-HT_2A/mGluR2 mRNA expression in nerve cell bodies of mouse somatosensory cortex and mouse cortical primary cultures. Also, specific 5-HT_2A and mGluR2 heteromerization has been found in HEK-293 cells using bioluminescence resonance energy transfer (BRET²) techniques (Gonzalez-Maeso et al. 2008). The interface in the mGluR2–5-HT_2A heteromer model introduced involves the TM4 and TM5 helices of the two receptors. It is of substantial interest that the hallucinogenic 2,5-dimethoxy-4-iodoamphetamine (DOI) stimulated [³⁵S]GTPγS binding in primary cortical membranes when linked to either G_αi1, G_αi2 or G_αi3 is strongly reduced by the mGluR2/3 agonist LY379268 as is the hallucinogen-specific induction of egr-2 in mouse cortex (Gonzalez-Maeso et al. 2008). In contrast, the DOI-stimulated [³⁵S]GTPγS binding when linked to G_αq/11 is only weakly affected by the mGluR2/3 agonist.

Thus, the mGluR2/3 agonist may produce their antipsychotic effects by targeting the mGluR2–5-HT_2A heterodimer/receptor mosaic probably located in distinct pyramidal and granular cells and reduce the hallucinogen-specific G_i/o signaling via allosteric interactions over the receptor interface (Gonzalez-Maeso et al. 2008). The receptor–receptor interaction is also found at the agonist recognition level in this receptor mosaic. mGluR2/3 agonists increase the agonist affinity of the 5-HT_2A receptor. This may be part of a complex allosteric mechanism within this receptor–receptor interaction, which increases 5-HT_2A recognition and reduces its coupling to G_i/o.

In this exciting paper (Gonzalez-Maeso et al. 2008), it was also found that 5-HT_2A receptor expression is increased in schizophrenia, a rise that is blocked by antipsychotic treatment. In parallel, the mGluR2 is reduced in schizophrenia and not affected by the antipsychotic treatment. These results clearly indicate that the disruption of sensory processing in schizophrenia may be related in part to pathological disturbances in the mGluR2–5-HT_2A receptor mosaic in the cerebral cortex. Alterations in its receptor–receptor interactions may lead to increases in hallucinogen-specific signaling (Fig. 2).

Fig. 2

Illustration of the possible preferential location of the mGluR2–5-HT_2A heterodimer/receptor mosaic on the shafts of the apical dendrites of pyramidal nerve cells of the cerebral cortex. This receptor assembly may mainly have an extrasynaptic location and is reached by both glutamate and 5-HT through volume transmission within the extracellular space originating from the synaptic glutamate nerve terminals and from the varicosities of the 5-HT nerve terminal plexus. No 5-HT_2A receptors exist on the dendritic spines of the pyramidal nerve cells (see Jansson et al. 2001). In the upper part, the hallucinogenic signaling of d-LSD is indicated as it targets the mGluR2–5-HT_2A heterodimer/receptor mosaic. d-LSD can activate not only the G_q/11 but also the G_i/o-mediated hallucinogenic signaling of the 5-HT_2A receptors (Gonzalez-Maeso et al. 2008). Via the inhibitory 5-HT_2A–mGluR2 receptor interaction in the receptor assembly, the signaling of the mGluR2 is also reduced by d-LSD (Gonzalez-Maeso et al. 2008). The type of G protein involved in the mGluR2 signaling in this assembly is not known. Through glutamate activation of the mGluR2 receptor within this receptor assembly, the d-LSD-mediated hallucinogenic signaling over the G_i/o in the 5-HT_2A receptor becomes blocked through an inhibitory mGluR2–5-HT_2A receptor–receptor interaction at the G protein coupling level. The receptor interaction is complex, since at the same time the affinity of the 5-HT_2A receptor becomes increased upon the activation of the mGluR2 (Gonzalez-Maeso et al. 2008) suggesting a major conformational change in the 5-HT_2A receptor that results in altered G protein selection and an enhancement of 5-HT_2A recognition

Conclusions

Receptor–receptor interactions within receptor heterodimer/receptor mosaics of different receptors give novel strategies for treatment of schizophrenia. One key region is the ventral striatum where the discovery of different types of D₂ containing heterodimers/receptor mosaics offer new therapeutic possibilities to increase activity in the ventral striato-pallidal GABA pathway. This will help and restore the glutamate drive from the mediodorsal thalamic nucleus to the prefrontal cortex and other cortical regions apart from the present use of D₂ antagonists/partial agonists. Thus, the existence of antagonistic A_2A–D₂, mGluR5–D₂, CB₁–D₂ and NTS₁–D₂ receptor interactions in the corresponding heterodimers/receptor mosaics in the local circuits of the ventral striato-pallidal GABA neurons suggests the use of A_2A, mGluR5, CB₁ and NTS₁ agonists as novel strategies for the treatment of schizophrenia. This may involve combined use of such agonists in low doses to antagonize D₂ receptor function especially in view of the likely existence also of trimeric A_2A–D₂–CB₁ and A_2A–D₂–mGluR5 receptor mosaics in these local circuits. This should reduce collateral effects, which may also be the case if also very low doses of D₂ antagonists are added to the agonist treatments.

Different primary disturbances may exist in subpopulations of schizophrenic patients in these D₂-containing receptor mosaics, modulating the activity of the ventral striato-pallidal GABA pathways. Therefore, the different agonists may be differentially active in subgroups of schizophrenic patients.

An important consequence of restoring the glutamate drive to the prefrontal cortex may be the increased activation of mGluR2 in the 5-HT_2A–mGluR2 heterodimer/receptor mosaic in cortical neurons. This will reduce the hallucinogenic-specific G_i/o signaling and behavior over the 5-HT_2A receptor with a reduction of primary disturbances in cortical sensory processing (Gonzalez-Maeso et al. 2008). In certain groups of patients, the primary disturbance may be in this receptor mosaic and therefore these schizophrenic patients may optimally respond to mGluR2 agonists and/or 5-HT_2A antagonists for antipsychotic activity. Other groups of schizophrenic patients may show optimal therapeutic responses by targeting both the malfunctioning D₂ receptor containing receptor mosaic in the ventral striatum and the 5-HT_2A–mGluR2 heterodimer/receptor mosaic in the cerebral cortex.

The field of receptor mosaics of different types of receptors and their receptor–receptor interactions in networks of the ventral striatum and the prefrontal cortex offer novel strategies for treatment of different forms of schizophrenia.