Differential allosteric modulation within dopamine D2R - neurotensin NTS1R and D2R - serotonin 5-HT2AR receptor complexes gives bias to intracellular calcium signalling

Proceeding investigations of G protein-coupled receptor (GPCR) heterocomplexes have demonstrated that the dopamine D2 receptor (D2R), one of the hub receptors in the physiology of schizophrenia, interacts with both the neurotensin NTS1 (NTS1R) and the serotonin 5-HT2A receptor (5-HT2AR) in cell lines and rodent brain tissue. In situ proximity ligation assay and BRET-based saturation experiments confirmed interacting receptor assemblies in HEK293T and neuronal HT22 cells. The NTS1R agonist NT(8-13) reduces the Gαq-mediated calcium signal in the NTS1R-D2R complex compared to the NTS1R monomer which could be reversed by D2R antagonists. The bivalent ligand CS148 (NTS1R-agonistic, D2R-antagonistic) increased the calcium response addressing the dimer, consistent with the effect of the monovalent ligands suggesting an allosteric D2R-mediated modulation. In contrast, the 5-HT2AR-D2R heteromer did not show a calcium-altering receptor-receptor interaction. Despite their common coupling-preference for Gαq, 5-HT2AR and NTS1R supposedly interact with D2R each in a unique mode. This remarkably diverse ligand-mediated signalling in two different D2R heteroreceptor complexes illustrates the complexity of receptor-receptor interactions and their potential of modifying cell responses to external stimuli. Therefore, GPCR heteromers may provide a very promising novel target for the therapy of neuropsychiatric disorders.


Results
Neurotensin receptor 1 and dopamine receptor D2 form heteromers in HT22 cells and show characteristic signalling. As an initial step, we aimed to strengthen the hypothesis that neurotensin NTS1 and dopamine D 2 receptors are able to form heteromers in a neuronal cell line. We applied the antibody based in situ PLA to visualize the heteromerization of NTS1R and D 2 R receptors in the plasma membrane of transiently transfected, fixed HT22 cells, an immortalized cell line of murine hippocampal origin 34 . Subsequent to a rolling circle amplification, the red fluorescent hybridization product that can only appear when the two relevant GPCRs are within a mutual distance of 10-20 nm was detected by confocal microscopy 35 . NTS1R-D 2 R co-transfected cells showed significantly high numbers of PLA positive clusters (Fig. 1a,b) indicating the formation of receptor heteromers. To show specificity, we added both primary NTS1R and D 2 R antibodies and the PLA required secondary antibodies to non-transfected cells. There, no statistically significant amount of PLA positive signals could be observed proving the selectivity of the method (Fig. 1a,c) and validating the appearance of heteromers in our cellular model.
Effects of D 2 R ligands on NTS1R-D 2 R mediated calcium signalling. Next, we investigated if D 2 R agonists can regulate the observed negative modulation within the NTS1R-D 2 R heterodimer. Therefore, the cells were incubated with dopamine and quinpirole, respectively, for 5 min before stimulation with 15 nM NT(8-13) (Fig. 2a), a concentration sufficient to evoke a maximum calcium response (Fig. 1d). Neither dopamine nor quinpirole were able to restore or intensify the calcium release provoked by NT (8)(9)(10)(11)(12)(13) in NTS1R-D 2 R expressing cells (Fig. 2b). Co-transfected HT22 cells still showed approximately 35% lower intracellular calcium levels after activation compared to NTS1R mono-expressing cells, indicating that the intrinsic dopaminergic modulation is already at its maximum efficacy in the basal receptor state.
Subsequently, the effect of D 2 R antagonists on the antagonistic allosteric NTS1R-D 2 R interaction was analyzed. NTS1R mono-transfected cells were not affected by incubation with the D 2 R antagonist haloperidol, leading to a plateau calcium concentration of 624 ± 30.2 nM, mean ± SEM (Fig. 2c). In co-expressing cells, subnanomolar and low nanomolar concentrations of haloperidol did not change the allosteric receptor-receptor interaction exerted by the D 2 R protomer on the partner NTS1R protomer signalling. However, increasing concentrations of haloperidol (10-50 nM) resulted in intracellular calcium levels approaching the level of NTS1R mono-transfected cells. The weak partial D 2 R-like agonist aripiprazole showed a similar effect on the calcium signalling in NTS1R-D 2 R expressing cells (Fig. 2d). Since haloperidol and aripiprazole were able to block the constitutive D 2 R-mediated allosteric modulation within the NTS1R-D 2 R complex, we conclude that by antagonizing the D 2 R protomer the heterodimeric complex might change insofar as the inhibitory allosteric receptor-receptor interaction can no longer occur, thus leading to the same "full" calcium response in NTS1R-D 2 R heteromers as in NTS1R monomers.
www.nature.com/scientificreports www.nature.com/scientificreports/ To further support the D 2 R-mediated allosteric effect on the NTS1R dependent release of intracellular calcium, two recently described heterobivalent ligands, CS142 and CS148 37 , were applied to co-expressing HT22 cells. In contrast to the prior used substances that solely address the orthosteric binding site of either of the two receptors, these bivalent ligands comprise two independent pharmacologically active head groups that are connected by an appropriate spacer. Therefore, such compounds serve as perfect tools to co-stimulate closely adjacent receptors stabilizing physically interacting protomers. Both ligands contain the NTS1R agonistic pharmacophore NT (8)(9)(10)(11)(12)(13). CS142 features a D 2 R agonistic aminoindane partial structure, whereas CS148 incorporates a 1,4-disubstituted-piperazine head group exhibiting antagonistic properties at the D 2 R protomer (Fig. 3a) 37 . We expected the simultaneous binding to provide further evidence of calcium signalling of NTS1R-D 2 R heteromeric origin. Consistent with the results obtained with monovalent dopamine receptor agonists, the inhibitory effect of the D 2 R protomer in the co-expressing cell system remained untouched by the D 2 R-agonistic bivalent ligand CS142. Consequently, the calcium level is significantly lower in comparison to HT22 cells that express NTS1R only (Fig. 3b). Stimulation with CS148, consisting of NT(8-13) and a D 2 R antagonist, resulted in calcium concentrations that are highly comparable to NTS1R mono-expressing cells (Fig. 3c). These observations support the conclusion that supposedly an allosteric inhibitory mechanism within NTS1R-D 2 R heteromers regarding G protein-mediated downstream signalling is present. Again, dopaminergic agonism does not affect this modulation, yet D 2 R-antagonism seems to rescue decreased intracellular calcium levels.
Unchanged calcium signal in co-expressed 5-HT 2A and D 2 receptors. As a next step, we investigated whether the D 2 R-based effects on the ligand-triggered calcium release are specific for the NTS1R-D 2 R www.nature.com/scientificreports www.nature.com/scientificreports/ heterodimer. Therefore, we changed the Gα q -coupled protomer and co-expressed the D 2 R together with the serotonin 5-HT 2A R. Co-localization as well as direct interaction of D 2 R and 5-HT 2A R have previously been described in HEK293 cells and in rat brain tissue 32,33 . Especially regions in the dorsal striatum and the nucleus accumbens are rich in 5-HT 2A R-D 2 R complexes 33 . Again, HT22 cells served as the experimental model and expressed the 5-HT 2A R either alone or together with the D 2 R after transfection. Interestingly, activation of the 5-HT 2A R by its selective agonist DOI (100 nM) or its endogenous ligand 5-HT did not result in altered calcium levels independent of the presence of the D 2 R (Fig. 4a, Supplementary Figs S2 and S3). Varying the ratio of co-transfected cDNA did not significantly change the quality of the signal ( Supplementary Fig. S4). Neither agonism by high concentrations of dopamine (2 µM) nor antagonism at the D 2 R binding site by haloperidol (40 nM) before DOI stimulation were able to affect the amount of calcium released intracellularly. These findings clearly differ from the allosteric interaction detectable in the NTS1R-D 2 R heterodimer. To exclude that these observations are cell line specific, calcium imaging was additionally performed in HEK293T cells. Again, we did not detect quantitative changes in the concentration of released calcium in mono-or co-expressing cells, with or without preincubation with dopaminergic ligands (Fig. 4b).
In conclusion, we assume that independent of the in vitro test system, 5-HT 2A R and D 2 R do not show a significant positive or negative receptor-receptor modulation, at least at the level of downstream calcium signalling. Hence, the NTS1R-D 2 R heterodimer allosteric interaction is a highly specific means of regulation. ] stimulated 188 ± 49.9 nM) did neither compensate nor enhance reduced Ca 2+ levels in NTS1R-D 2 R co-expressing cells still showing significant differences towards the mono-transfected system (**p < 0.01). (c) In contrast, increasing concentrations of the dopaminergic antagonist haloperidol led to similarly high intracellular Ca 2+ compared to NTS1R-only HT22 cells. (d) Incubation with 40 nM haloperidol or 40 nM of the weak D 2 R partial agonist aripiprazole led to a full recovery of the intracellular Ca 2+ response in co-expressing cells after stimulation with 15 nM NT (8)(9)(10)(11)(12)(13) no longer presenting significant differences between both expressing systems (*p < 0.05, ns = non-significant). Data were analyzed with one-way ANOVA and Tukey's multiple comparisons test presented as mean ± SEM, n = 6. www.nature.com/scientificreports www.nature.com/scientificreports/ To exclude that the absence of allosteric modulation does not rely on the absence of 5-HT 2A R-D 2 R heteromers, we analyzed whether 5-HT 2A R and D 2 R are expressed in immediate proximity, thus theoretically enabling a mutual physical interaction. To this end, we made use of a BRET saturation assay. Either of the receptors was C-terminally tagged with a coelenterazine-converting enzyme Renilla reniformis luciferase variant (Rluc8), whereas the other protomer served as the acceptor bearing the yellow fluorescent protein mVenus at its intracellular C-terminus. Keeping the concentration of the Rluc8 donor construct constant, we transfected HEK293T cells with increasing amounts of mVenus-fused receptor plasmids. Irrespective of the orientation of the BRET pair, characteristic saturable BRET titration curves could be obtained (Fig. 4c,d) indicating a specific protein-protein interaction 32,38 . This supports the assumption that the absence of any allosteric modulation regarding the calcium signalling does not depend on the absence of heterodimers.

Discussion
There is evidence that heteromeric GPCR complexes differ in their way of transferring external stimuli to the inside of a cell compared to their monomeric counterparts. For instance, the activation of the dopamine receptor heterodimer D 1 R-D 2 R results in an unexpected Gα q -mediated calcium release, whereas the individual monomers modulate the effector adenylyl cyclase in response to ligand binding 39 . This 'non-rigidity' of interacting GPCRs can multiply the possibilities of signal transduction opening up a wide field for GPCR-targeting novel therapeutics.
In the present study, we demonstrate that two different D 2 R-containing GPCR heterocomplexes, each with a Gα q -coupling partner (NTS1R and 5-HT 2A R, respectively), do not show identical behavior determining the intracellular calcium release after ligand activation. Immortalized neuronal HT22 cells served as a relevant model to investigate schizophrenia-related receptor interactions 34 . Interestingly, according to our data the mere co-expression of NTS1R and D 2 R is sufficient to reduce the level of intracellularly released calcium after NTS1R activation with NT(8-13) compared to singly expressed NTS1R. Being aware that the reduced signal may be influenced by the degree of receptor expression in the cells, we transfected different cDNA ratios. In each proportion, a significantly lowered dimer-specific signal could be obtained (Supplementary Fig. S1). The application of dopaminergic ligands provided more information about this negative allosteric modulation. We presume that blocking the D 2 R leads to a conformational change in the transmembrane helices involved in the NTS1R-D 2 R interface. This altered inter-receptor contact may change the geometrical properties of the NTS1R, which could result in an improved recruitment of Gα q and therefore in a calcium signal identical to the monomeric response. In contrast, co-activation of the heterodimer with NT(8-13) and DA or quinpirole, respectively, does not seem to contribute to a relevant change in the interacting receptor domains indicated by the reduced intracellular calcium response. The application of bivalent ligands, which prevented the protomers from dissociation by their high-affinity The applied bivalent ligands consist of three distinct structural elements: NTS1R-agonistic pharmacophore NT(8-13) (red), spacer (blue) and D 2 R-agonistic (CS142, black-orange) or D 2 R-antagonistic (CS148, black-cyan) pharmacophore, respectively. (b) Treatment with 0.5-1 µM CS142 resulted in significantly reduced intracellular Ca 2+ fluxes in co-expressing compared to NTS1R singly expressing cells (*p < 0.05). (c) The application of CS148 neutralized the diminished Ca 2+ release in the NTS1R-D 2 R co-expressing system. The observed calcium recovery is displayed by non-significant differences between mono-and co-transfected cells for all tested concentrations. Data were analyzed with one-way ANOVA and Tukey's multiple comparisons test presented as mean ± SEM, n = 6, performed in hexaplicates. *p < 0.05, **p < 0.01, ***p < 0.001, ns = non-significant. (2019) 9:16312 | https://doi.org/10.1038/s41598-019-52540-8 www.nature.com/scientificreports www.nature.com/scientificreports/ bivalent binding mode 37 , supported that the detectable effect is dimer-specific and not provoked by disruption of the heterodimer. We suggest the notion of allosteric interaction via immediate conformational changes within the GPCRs, since the phospholipase C-β mediated release of calcium from cytosolic stores upon ligand binding is a short-term effect 40 . Whereas we here describe a negative receptor modulation arising from the D 2 R protomer towards the NTS1R, Borroto-Escuela et al. revealed, on the contrary, a direct receptor-receptor interaction within the NTS1R-D 2 R heterodimer 27 . Stimulation of the NTS1R led to an antagonistic modulation of the D 2 R-mediated adenylyl cyclase inhibition. On the other hand, they postulate that co-stimulation of both protomers accounts for a synergistic activation of the protein kinase C, likely involving agonist-mediated β-arrestin recruitment, resulting in an enhanced D 2 R signalling. Moreover, bivalent ligands with high affinity to NTS1R-D 2 R heteromers show a characteristic signalling behavior regarding cAMP accumulation in co-expressing compared to D 2 R singly expressing cells, potentially implying allosteric modulatory mechanisms 37 . Consequently, it becomes evident that allosteric receptor-receptor interactions by no means are clearly directional but are able to appear within one GPCR complex both positively and negatively in either direction.
Surprisingly, we did not see any influence on the intracellular calcium response in 5-HT 2A R-D 2 R complexes compared to the mono-expressed 5-HT 2A R, irrespective of the 5-HT 2A R-D 2 R ratio ( Supplementary Fig. S4). Neither any constitutive, nor any D 2 R agonist-or antagonist-mediated allosteric interaction was detected. This, however, does not automatically demonstrate the absence of any direct crosstalk, but at least at the level of downstream calcium signalling there appears to be a lack of allosteric modulation. Nevertheless, 5-HT 2A R-D 2 R www.nature.com/scientificreports www.nature.com/scientificreports/ interactions have been described, mainly investigating D 2 R-promoted signalling 32,33 . Depending on the ability of the 5-HT 2A R agonist to provoke hallucinations or not, the D 2 R signalling was influenced differently. Hallucinogenic DOI enhanced, but non-hallucinogenic 5-HT 2A R ligands on the contrary rather diminished the Gα i -mediated actions of the D 2 R protomer. The authors concluded to have witnessed a dynamic set of allosteric receptor-receptor interactions that develop through conformational changes at the receptor interface upon ligand binding 33 . A similar bias was apparent at the level of inositol phosphate production, a process upstream of Ca 2+ -release from intracellular stores, after 5-HT 2A R activation with and without D 2 R co-expressed or activated, respectively 17 . In the latter study, dimeric modulation was ligand and readout dependent. Contrarily, there is evidence for increased calcium signalling after 5-HT 2A R-D 2 R co-stimulation 32 . There, the ensemble cell assay depicting combined signals of mono-and co-expressing cells may account for the signal enhancement in contrast to our single cell imaging technique where exclusively 5-HT 2A R-D 2 R cells were analyzed. By trend, this finding might still be reflected in our HEK293T cell system showing a slightly higher, but non-significant signal for the DOI/DA co-activation (Fig. 4b). Yet, in hippocampus derived HT22 cells, an altered calcium signal could not be detected. Occurrence of a different dimer-specific behaviour in a more neuron-like system possibly gives a hint that modulatory receptor-receptor interactions may be dynamic in the environment they appear. Thus, the mechanism of atypical antipsychotic drugs may also proceed via heteromeric complexes in various areas of the brain. This fact makes GPCR heterodimers to very exciting target structures for drug development that need to be studied further to gain more knowledge about their pharmacology.
Yet, there is still limited understanding of the structural determinants of the different signalling behaviour of these two pairs of receptors. To elucidate NTS1R-D 2 R interaction, Hübner et al. combined X-ray crystal structures and homology modelling to design a heterodimer model 37 . They suggested transmembrane helices (TM) 1, 2 and helix 8 to be involved in the formation of the interface, supported by binding data of heterobivalent ligands. Further attempts to analyze the interface between D 2 R and NTS1R or 5-HT 2A R, respectively, have been made by bioinformatical approaches, where short amino acid sequences that are predominantly found in heteromeric complexes were identified 27,32 . According to these findings, NTS1R and D 2 R are likely to have triplets in common that interact via TMs 2, 4 and 6 as well as intracellular loop 3 27 . 5-HT 2A R and D 2 R, contrarily, are described to share homologies in TM1 and TM3 32 . The motifs within NTS1R-D 2 R and 5-HT 2A R-D 2 R interfaces are non-identical so that dimer-specific interactions might be possible. For the adenosine A2A receptor (A2AR)-D 2 R dimer, however, interdisciplinary approaches have unravelled TM4 and TM5 of D 2 R to form the potential interface 41 , strengthening the evidence that D 2 R is able to develop different interaction sites in different receptor complexes. Since NTS1R homodimerization has been described in a "rolling interface" model 25 , according to which the interacting TMs are interconverting, presence of NTS1R within a heterocomplex might result in dynamic changes at the interface in contrast to other possibly more rigid dimers. Postulating that allosteric modulation may be transduced via conformational changes, different adjacent structural elements at dimer interfaces would therefore contribute to the differences in ligand mediated calcium response for NTS1R-D 2 R compared to 5-HT 2A R-D 2 R.
In the plasma membrane of neurons, 5-HT 2A R and NTS1R are by far not the only receptors to interact with D 2 R. For instance, A2AR has been described to form heteromers with D 2 R showing a strong antagonistic relation. Activation of the A2AR decreases Gα i -mediated signalling of the D 2 R while promoting the recruitment of β-arrestin2 to the dopamine D 2 protomer at the same time 42,43 . Interestingly, three different D 2 R-heterocomplexes show clearly individual and unique allosteric modulation illustrating the importance of potential receptor-receptor interactions for the neuronal system to adapt its signal transduction to physiologically altered environments.
We are aware that the neuronal cell line used does not perfectly represent the molecular environment of brain tissue. Within a synapse, the release, the local distribution and concentration as well as the reuptake of neurotransmitters are tightly controlled by regulatory mechanisms 44 . Nevertheless, the hippocampus-derived cells are an adequate model to study receptor-receptor interactions relevant to neuropsychiatric disorders.
In summary, we here demonstrate the existence of two different pairs of GPCR heteromers, namely the NTS1R-D 2 R and the 5-HT 2A R-D 2 R using in situ PLA and BRET. There is evidence that the allosteric interactions between the protomers may be due to conformational anomalies or changes at facing transmembraneous domains. In addition, the extent of the inter-receptor modulation is highly dimer and ligand specific as we could show the altered modulation of intracellular calcium responses by live cell calcium imaging (Fig. 5). Having in mind that these heteromers are inter alia present in schizophrenia associated limbic areas 14 , it is of particular interest to continue the studies on dimer expression, distribution, monomer-dimer equilibrium and signalling. Further work is inevitably necessary to better understand these receptor formations in order to use them as potential targets for specific and effective therapy of neuronal diseases.
Plasmids. For calcium live imaging, wildtype human NTS1R, D 2L R and 5-HT 2A R in pcDNA3.1 (+) (cDNA Resource Center, www.cdna.org) and plasmids containing the cDNA for the fluorescently labeled receptor fusion proteins NTS1R-eCFP and D 2L R-eYFP in pcDNA3.1 (+) 30 were used. The cDNA encoding human 3xHA 5-HT 2A R was designed in the lab of Kjell Fuxe, using standard molecular cloning. The BRET 1 biosensors 3xHA 5-HT 2A R-Rluc8, 3xHA 5-HT 2A R-mVenus, D 2L R-Rluc8 and D 2L R-mVenus in pcDNA3.1 (+) were cloned applying (Invitrogen). 48 h after mono-transfection with NTS1R or co-transfection with NTS1R and D 2L R-YFP, HT22 or HEK293T cells were washed twice with HBSS (138 mM NaCl, 6 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 5.5 mM glucose, 10 mM HEPES, pH 7.4, 37 °C). Experiments with 5-HT 2A R alone or in combination with D 2L R-YFP were performed 24 h after transfection, as low cell viability and frequent detachment were observed at later time points. Loading with Fura-2-AM, (0.35 µM in HBSS with 2% FBS, 0.01% Pluronic F-127 (Invitrogen)) was performed at 37 °C under light exclusion for 30 minutes. After two washing steps, cells were kept dark at room temperature for 30 minutes. Cells were incubated with either dopamine (20 nM, 2 µM), haloperidol (40 nM), quinpirole (10 nM), aripiprazole (40 nM) or HBSS (control) for 5 minutes at room temperature and transferred to the imaging chamber. Cells were alternately excited with 340 nm and 380 nm for 50 ms each. A 510 nm mirror unit was used to collect Fura-2 fluorescence. Following stimulation with NT(8-13), CS142 or CS148 as NTS1R agonists or 5-HT and DOI as 5-HT 2A R ligands, the calculated 340/380 absorption ratio provided information about changes in intracellular calcium concentrations. Transformation of ratio values into absolute calcium concentrations was carried out according to literature 46 . Δ[Ca 2+ ] i values were determined calculating the difference between basal calcium before and maximum response after ligand application. D 2L R transfected cells were identified using their eYFP fluorescent tag and only cells that showed eYFP fluorescence and concomitant release of calcium upon NTS1R or 5-HT 2A R agonist application were analyzed for co-transfections. For the investigation of NTS1R or 5-HT 2A R mono-transfected cells, only cells providing a calcium response were included in data analysis. BRet 1 saturation assay. BRET saturation experiments have been described previously 47 . Briefly, HEK293T cells were transfected in suspension at a density of 1.25 × 10 5 cells/mL in complete media using polyethylenimine (PEI, linear, 25 kDa, Polysciences, Heidelberg, Germany) at a 3:1 PEI to cDNA ratio. Increasing amounts of mVenus-labeled BRET acceptor cDNA (0-700 ng) were mixed with a constant amount of Rluc8-tagged donor cDNA (100 ng for 3xHA 5-HT 2A R-Rluc8 and 50 ng for D 2L R-Rluc8, respectively) and complemented with ssDNA (Sigma Aldrich) to a total amount of 1 µg cDNA per 1.2 mL cell suspension. The cDNA/PEI complexes were incubated at room temperature for 30 minutes, before the transfection complex was added to the cells and cells were transferred to white 96-well plates (100 µL/well, Greiner Frickenhausen, Germany). After 48 h at 37 °C and 5% CO 2 , the media was removed and substituted by 90 µL of PBS supplemented with CaCl 2 and MgCl 2 (Invitrogen) for 1 h. A CLARIOstar Plate Reader (BMG Labtech, Ortenberg, Germany) was used to determine mVenus fluorescence (excitation 497 nm, emission 535 nm). BRET 1 measurements were performed 20 minutes after addition of 10 µL of coelenterazine-h (5 µM final concentration, Promega, Mannheim, Germany) using a BRET 1 filter set (475 − 30/535 − 30 nm). The BRET 1 signals were calculated as the ratio between the intensity of light emitted by the acceptor (mVenus, 535 nm) divided by the donor emission (Rluc8, 475 nm). Specific BRET (net BRET) was determined by subtraction of the signal obtained in the absence of any fluorescent acceptor protein and plotted against the ratio of fluorescence (mVenus) to luminescence (Rluc8) to receive saturation curves. Statistical analysis. Experimental data are presented as mean ± SEM. Graphical and statistical analysis was performed using GraphPad Prism 6.0 (GraphPad Software, Inc.; San Diego, CA, USA). One-way ANOVA or Student's t-test together with post-hoc Bonferroni's or Tukey's test provide information about statistical significance. p values less than 0.05 were considered statistically significant (*p < 0.05; **p < 0.01; ***p < 0.001), above 0.05 as non-significant (ns). Number of single experiments ranges between n = 3-6.

Data availability
All data generated and analyzed during this study may be requested from the corresponding author (K.Fr.).