LETTERS Identification of a serotonin/glutamate receptor complex implicated in psychosis Javier Gonzalez-Maeso1,2, �� Rosalind L. Ang1, Tony Yuen1, Pokman Chan1, Noelia V. Weisstaub5,6, Juan F. Lopez-Gimenez8, �� �� Mingming Zhou5, Yuuya Okawa1, Luis F. Callado9,10, Graeme Milligan8, Jay A. Gingrich5,6,7, Marta Filizola3, J. Javier Meana9,10 & Stuart C. Sealfon1,4 The psychosis associated with schizophrenia is characterized by alterations in sensory processing and perception1,2. Some anti- psychotic drugs were identified by their high affinity for serotonin 5-HT2A receptors (2AR)3,4. Drugs that interact with metabotropic glutamate receptors (mGluR) also have potential for the treatment of schizophrenia5���7. The effects of hallucinogenic drugs, such as psilocybin and lysergic acid diethylamide, require the 2AR8���10 and resemble some of the core symptoms of schizophrenia10���12. Here we show that the mGluR2 interacts through specific transmem- brane helix domains with the 2AR, a member of an unrelated G-protein-coupled receptor family, to form functional complexes in brain cortex. The 2AR���mGluR2 complex triggers unique cellular responses when targeted by hallucinogenic drugs, and activation of mGluR2 abolishes hallucinogen-specific signalling and behavioural responses. In post-mortem human brain from untreated schizophrenic subjects, the 2AR is upregulated and the mGluR2 is downregulated, a pattern that could predispose to psychosis. These regulatory changes indicate that the 2AR��� mGluR2 complex may be involved in the altered cortical processes of schizophrenia, and this complex is therefore a promising new target for the treatment of psychosis. The 2AR and mGluR2/3 show an overlapping distribution in brain cortex in autoradiography studies13. The mGluR2 and mGluR3 are not distinguished by autoradiographic ligands. We used fluorescent in situ hybridization (FISH) to determine whether either of these receptor subtypes is co-expressed by the same neurons. In layer V mouse soma- tosensory cortex (SCx), cells positive for 2AR mRNA were mostly positive for mGluR2 mRNA. The level of expression in SCx was much lower for mGluR3 mRNA, which rarely co-localized with 2AR mRNA (Fig. 1a). Control studies validated the sensitivity and specificity of the assay, and similar 2AR���mGluR2 mRNA co-localization was found in cortical primary cultures (Fig. 1a���c and Supplementary Fig. 1). Translation of 2AR protein in cortical pyramidal neurons was found to be necessary for normal mGluR2 expression. Mice with globally disrupted 2AR expression (htr2A2/2 mice) showed reduced cortical mGluR2 binding and expression, whereas mice in which 2AR expres- sion had been selectively restored in cortical pyramidal neurons8,14 showed expression levels equal to those in the control (Supplemen- tary Table 1 and Supplementary Fig. 2). The effects of mGluR2/3 activation on 2AR responses have generally been attributed to synaptic mechanisms5,6,13,15. However, the co-localization of 2AR and mGluR2 and the decrease in mGluR2 expression levels in htr2A2/2 mice led us to examine whether a direct mechanism contributed to cortical cross- talk between these two receptor systems. Recent studies have demonstrated that some G-protein-coupled receptors (GPCRs) belonging to the same sequence classes can form dimers16 or, potentially, higher-order oligomers17. Although the 2AR and mGluR2 belong to different GPCR classes, we established the existence of 2AR���mGluR2 heterocomplexes by several methods: co- immunoprecipitation of human brain cortex samples (Fig. 1d) and of HEK-293 cells transfected with epitope-tagged receptors (Fig. 2b), bioluminescence resonance energy transfer (BRET) (Fig. 1e and Supplementary Fig. 3), and fluorescence resonance energy transfer (FRET) (Fig. 2d) studies in transfected cells. To determine whether the formation of the 2AR���mGluR2 com- plex has functional consequences, we first examined, in mouse SCx membranes, the effects of an mGluR2/3 agonist on the competition binding of several hallucinogenic 2AR agonists (Fig. 1f, top) and of a 2AR agonist on the competition binding of several mGluR2/3 agonists (Fig. 1f, bottom). The agonist affinities for the 2AR and mGluR2/3 were decreased when receptor���G-protein complexes were uncoupled by GTP-cS (Supplementary Fig. 4 and Supple- mentary Tables 2 and 3). The glutamate agonist LY379268 increased the affinity of all three hallucinogens studied for the 2AR-binding site. Furthermore, the 2AR agonist DOI (1-(2,5-dimethoxy-4- iodophenyl)-2-aminopropane) decreased the affinity of the three mGluR2/3 agonists for the glutamate-receptor-binding site. The allosteric interactions observed were eliminated by antagonist for each modulator (see Supplementary Tables 2 and 3 and Supplemen- tary Fig. 4 for additional concentrations of DOI and LY379268 and for the elimination of the allosteric effects by antagonists). Although the glutamate agonists studied do not distinguish between the mGluR2 and mGluR3 subtypes18, the rarity of mGluR3 and 2AR mRNA co-expression in cortex, the absence of evidence for 2AR��� mGluR3 complex formation by co-immunoprecipitation, BRET and FRET, and the detection of 2AR���mGluR2 complexes by these same assays all indicate that the cross-talk identified results from 2AR��� mGluR2 complexes. The differences in the capacity of the mGluR2 and mGluR3 to interact with the 2AR and their close sequence similarity provided the basis for identifying the specific mGluR2 domains responsible for heterocomplex formation. A study of a series of molecular chimaeras of the mGluR2 and mGluR3 (see Fig. 2a) demonstrated that the segment containing transmembrane (TM) helices 4 and 5 of the mGluR2 receptor was both necessary and sufficient for the formation of a complex with the 2AR. The mGluR3 receptor chimaera contain- ing only this segment from the mGluR2 (mGluR3DTM4,5) was cap- able of co-immunoprecipitating with the 2AR (Fig. 2b), mediating 1Department of Neurology, 2Department of Psychiatry, 3Department of Structural and Chemical Biology and 4Center for Translational Systems Biology, Mount Sinai School of Medicine, New York, New York 10029, USA. 5Department of Psychiatry, 6Sackler Institute Laboratories and 7Lieber Center for Schizophrenia Research, Columbia University, and the New York State Psychiatric Institute, New York, New York 10032, USA. 8Division of Biochemistry and Molecular Biology, Institute of Biomedical and Life Sciences, University of Glasgow, Glasgow G12 8QQ, UK. 9CIBER of Mental Health, and 10Department of Pharmacology, University of the Basque Country, E-48940 Leioa, Bizkaia, Spain. Vol 452|6 March 2008|doi:10.1038/nature06612 93 NaturePublishing Group ��2008

allosteric cross-talk (Fig. 2c) and maintaining close proximity with the 2AR as indicated by FRET (Fig. 2d). In contrast, mGluR2DTM4,5 did not show evidence of complex formation with the 2AR (Fig. 2, Supplementary Figs 5 and 6, and Supplementary Tables 4 and 5 show complete curves, analysis and evidence of membrane expres- sion of all chimaeras). The absolute and relative levels of expression of heterologous constructs were comparable to the physiological levels found in mouse SCx, and in cortical primary cultures (Supple- mentary Fig. 5 and Supplementary Table 4). Our data do not exclude the possibility that the predicted 2AR���mGluR2 heterodimer, a model of which is shown in Fig. 2f, assembles into tetramers or larger recep- tor oligomers19,20. The changes in high-affinity binding caused by 2AR���mGluR2 cross-talk suggested that this complex may serve to integrate sero- tonin and glutamate signalling and modulate G-protein coupling21,22. This hypothesis was tested by measuring the regulation of Gaq/11 and Gai proteins by 2AR. High-affinity activation of Gaq/11 by the 2AR was decreased by co-expression of mGluR2 (Fig. 2e and Supple- mentary Table 6). The activation of Gai by the 2AR was markedly enhanced by mGluR2 co-expression (Fig. 2e and Supplementary Table 7). The mGluR2-dependent effects on both Gaq/11 and Gai regulation by the 2AR were reversed in the presence of mGluR2 agonist (Fig. 2e and Supplementary Tables 6 and 7). Consonant with the results from co-immunoprecipitation, allosteric modulation and FRET, the functional assays of G-protein activity also show that the TM4���5 segment of the mGluR2, when substituted into the mGluR3, was sufficient for signalling cross-talk to occur (Fig. 2e). These data support the presence of functional and physiological 2AR���mGluR2 complexes that integrate serotonin and glutamate neurotransmission to specify the pattern of G-protein regulation. Similar evidence for specification of G-protein subtype regulation was also observed for the endogenous brain 2AR���mGluR2 complex Number of copies per �� g RNA 40,000 30,000 20,000 10,000 Somatosensory cortex Thalamus 0.6 mGluR2-Rluc + 5HT2A���GFP2 mGluR2-Rluc + pGFP2 mGluR2-Rluc + 5HT2C���GFP2 mGluR3-Rluc + 5HT2A���GFP2 BRET 2 0.4 0.2 0 0 100 75 50 0 ���10 ���9 ���8 ���7 0 ���10 ���9 ���8 ���7 0 ���10 ���9 ���8 ���7 0 ���10 ���9 ���8 ���7 0 ���10 ���9 ���8 ���6 ���7 0 ���7 ���6 ���5 ���4 100 75 50 100 75 50 100 75 50 100 100 75 50 25 0 75 50 0.5 1.0 Fluorescence/luminescence log([LY379268] (M)) log([DOI] (M)) log([DCG-IV] (M)) log([DOM] (M)) log([L-CCG-I] (M)) log([DOB] (M)) [ 3 H]LY341495 bound (%) [ 3 H]Ketanserin bound (%) 1.5 2.0 0 mGluR2 mGluR3 118 47 118 47 kDa kDa 118 118 5-HT2AR mGluR2 5-HT2AR mGluR3 mGluR3 mGluR3 Overlay Overlay 5-HT2AR mGluR2 5-HT2AR mGluR3 Overlay Overlay a d e f IP 2AR WB mGluR2 IP 2AR WB mGluR3 WB mGluR2 WB mGluR3 47 WB 2AR V kDa b c VI V VI Figure 1 | 2AR and mGluR2 co-localize and interact. a, 2AR and mGluR2, but not mGluR3, are co-expressed in cortical neurons. Top, mouse somatosensory cortex bottom, mouse cortical primary culture. Scale bars, 50 mm (top) and 10 mm (bottom). Nuclei are blue. Inset: co-expressing neuron. b, FISH for mGluR3 in thalamus. Top, mouse thalamus bottom, thalamic primary culture. Scale bars, 25 mm (top) and 10 mm (bottom). c, mRNA levels measured by real-time PCR (Error bars show s.e.m. n 5 6 per group). d, Specific co-immunoprecipitation of 2AR and mGluR2 in duplicate human frontal cortex samples (arrows). WB, western blot IP, immunoprecipitation. e, BRET2 shows specific 2AR and mGluR2 interaction in HEK-293 cells. Data are means 6 s.e.m. (n 5 3). The mGluR2/ 2AR curve is fitted better by a saturation curve than by a linear regression, F test (P , 0.001). The other co-transfection data sets show linear regressions. f, Top, [3H]ketanserin displacement curves in mouse SCx membranes. 2AR agonist affinities were higher in the presence of the mGluR2/3 agonist LY379268 at 10 mM (red) than in vehicle alone (black). [3H]LY341495 displacement curves (bottom panels). mGluR2/3 agonist affinities were lower in the presence of the 2AR agonist DOI at 10 mM (red) than in vehicle alone (black). DCG-IV, (2S,29R,39R)-2-(29,39-dicarboxycyclopropyl)- glycine L-CCG-I, (2S,19S,29S)-2-(carboxycyclopropyl)-glycine. LETTERS NATURE|Vol 452|6 March 2008 94 NaturePublishing Group ��2008