Heterodimerization of serotonin receptors 5-HT_1A and 5-HT₇ differentially regulates receptor signalling and trafficking

G-protein-coupled receptors (GPCRs) belong to a large and diverse family of integral membrane proteins that participate in the regulation of many cellular processes and, therefore, represent key targets for pharmacological treatment. Until recently, GPCRs were assumed to exist and function as monomeric units that interact with corresponding G proteins in a 1:1 stoichiometry. However, biochemical, structural and functional evidence collected during the last decade indicates that GPCRs can form oligomers (Devi, 2001; Bulenger et al., 2005).

Oligomerization can occur between identical receptor types (homomerization) or between different receptors of the same or different GPCR families (heteromerization). Heteromerization is of particular interest because it can specifically modulate receptor properties. It can lead to significant changes in receptor pharmacology, either by affecting the ligand binding on individual protomers or by the formation of new binding sites (Franco, 2009; Rozenfeld and Devi, 2011). Accumulating evidence also indicates that heteromerization might affect signalling pathways regulated by a given protomer. For example, a synergistic increase of the receptor-mediated signalling was observed for the adrenergic α_1AAR–α_1BAR and the muscarinic M₂R–M₃R heterodimers (Hornigold et al., 2003; Israilova et al., 2004; Fuxe et al., 2005). On the other hand, G protein signalling might be attenuated upon heteromerization, as has been reported for the opiate MOR–DOR, the α_2AAR–MOR and the adenosine–dopamine A_2AR–D₂R heterodimers (Gomes et al., 2000; Jordan et al., 2003). Moreover, heteromerization can lead to a switch in G protein coupling, as previously shown for dopamine D₁R–D₂R heteromers (Lee et al., 2004). Thus, heteromerization might provide an additional level of control for the regulation of cellular processes by fine tuning of receptor-mediated signalling.

In the present study, we examined the heteromerization of two members of the serotonin receptor family, 5-HT_1A and 5-HT₇ receptors. The 5-HT_1A receptor is coupled to members of the G_i/o protein family, which induce inhibition of adenylyl cyclase and subsequent decrease of intracellular cAMP levels (Barnes and Sharp, 1999; Raymond et al., 1999; Pucadyil et al., 2005). In addition, stimulation of 5-HT_1A receptors leads to a Gβγ-mediated activation of K⁺ channels as well as to activation of the mitogen-activated protein (MAP) kinase Erk2 (Fargin et al., 1989; Garnovskaya et al., 1996). With respect to physiological functions, considerable interest in the 5-HT_1A receptor has been raised due to its involvement in depression and anxiety states (Parks et al., 1998; Gordon and Hen, 2004).

The 5-HT₇ receptor is one of the most recently described members of the 5-HT receptor family (Barnes and Sharp, 1999; Hedlund and Sutcliffe, 2004). The 5-HT₇ receptor stimulates cAMP formation by activating adenylyl cyclases via the G_s proteins (Norum et al., 2003). This receptor is associated with a number of physiological and pathophysiological responses, including serotonin-induced phase shifting of the circadian rhythm (Lovenberg et al., 1993) and age-dependent changes in the circadian timing (Duncan et al., 1999). In addition, a large body of evidence indicates an involvement of the 5-HT₇ receptor in the development of anxiety and depression, and recent studies have shown that the 5-HT₇ receptor is most probably clinically relevant for the treatment of depression (Hedlund, 2009).

Recently, we have demonstrated that 5-HT_1A receptors form homodimers at the plasma membrane (Kobe et al., 2008; Woehler et al., 2009). Formation of 5-HT_1A homomers (including the higher-order oligomers) was further confirmed by several more recent publications (Ganguly et al., 2011; Paila et al., 2011). Here, we report that 5-HT_1A receptors can form heterodimers with 5-HT₇ receptors both in vitro as well as in vivo. In addition, we propose a dynamic dimerization model that allows calculation of relative concentrations of monomers, homo- and heterodimers as a function of receptor expression level. We also demonstrate that heterodimerization decreases G_i protein coupling of the 5-HT_1A receptor and attenuates receptor-mediated activation of potassium channels without substantial changes in the coupling of the 5-HT₇ receptor to the G_s protein. Moreover, heterodimerization significantly facilitates internalization of the 5-HT_1A receptor as well as its ability to activate MAP kinase.

Specific interaction between 5-HT_1A and 5-HT₇ receptors was analysed by co-immunoprecipitation experiments in N1E-115 cells coexpressing haemagglutinin (HA)- and YFP-tagged receptors. Fig. 1A shows that after immunoprecipitation with an antibody against the HA-tag, YFP-tagged receptors were identified only in samples derived from cells coexpressing both HA- and YFP-tagged receptors. To assay the extent of artificial protein aggregation, cells expressing only one type of receptor (HA- or YFP-tagged) were mixed prior to lysis and analysed in parallel. As shown in Fig. 1A, individual receptors can be detected by the same antibody, but co-immunoprecipitation did not occur. This further verifies the specificity of 5-HT_1A–5-HT₇ hetero-oligomerization.

We then examined Förster resonance energy transfer (FRET) occurrence between fluorophore-labelled 5-HT_1A and 5-HT₇ receptors in living neuroblastoma cells. To avoid artefacts resulting from overexpression, we adjusted the receptor expression to 1.000–1.200 fmol/mg protein, which is similar to the endogenous expression level in vivo (Pazos and Palacios, 1985; Hoyer et al., 1986; Kobe et al., 2008). Fig. 1B shows the typical fluorescence emission spectra at 420 nm excitation obtained in suspensions of cells expressing 5-HT_1A–CFP, 5-HT₇–YFP or coexpressing 5-HT_1A–CFP and 5-HT₇–YFP as a FRET pair. When cells were transfected with only CFP-fused receptor, the typical emission spectrum of CFP was obtained with emission peaks at 475 nm and 500 nm (Fig. 1B). The emission spectrum obtained from cells expressing only the YFP-fused receptor showed a very weak peak at 525 nm. By contrast, cells coexpressing 5-HT_1A–CFP and 5-HT₇–YFP receptors demonstrated a significantly larger emission peak at 525 nm concomitant with a smaller CFP emission, which demonstrates the energy transfer from CFP to YFP (Fig. 1B) and confirms the 5-HT_1A–5-HT₇ heterodimerization in living cells. To analyse the effect of agonist stimulation on receptor oligomerization, we measured the FRET efficiency in suspensions of cells co-transfected with donor (5-HT_1A–CFP) and acceptor (5-HT₇–YFP) proteins at a 1:1 ratio during receptor stimulation with serotonin. Fig. 1C demonstrates that the time course of FRET obtained upon 5-HT treatment was indistinguishable from that in cells treated with phosphate-buffered saline (PBS), demonstrating that the oligomerization state of 5-HT_1A–5-HT₇ complexes is not modulated by the agonist.

A microscope-based acceptor photobleaching FRET assay (Kobe et al., 2008) was applied to study 5-HT_1A–5-HT₇ heterodimerization at the subcellular level. CFP- and YFP-tagged receptors were expressed in N1E-115 cells, and the plasma membrane localized receptors were targeted for acceptor photobleaching analysis (Fig. 2A). Fig. 2B,C illustrates changes in emission intensities of donor and acceptor fluorescence in the bleached and non-bleached regions of interest, demonstrating that a loss of acceptor fluorescence was accompanied by an increase of donor emission intensity, which is characteristic for FRET. For cells expressing fluorescence-tagged 5-HT_1A receptors with similar donor (CFP) to acceptor (YFP) ratios, a mean apparent FRET efficiency of 20±1% (n=24) was measured (Fig. 2D), which is in accordance with our previous results (Kobe et al., 2008). Similar FRET values were obtained for 5-HT₇ homodimers, with a mean apparent FRET efficiency of 21±2% (n=20) (Fig. 2D). In the case of coexpression of 5-HT₇–CFP (donor) and 5-HT_1A–YFP (acceptor) as a FRET pair, the apparent FRET efficiency was 20±2% (n=21) (Fig. 2D), and this value was not significantly different when 5-HT_1A–CFP was used as a donor and 5-HT₇–YFP as an acceptor (Ef_D=23±2%, n=22). As a negative control, we used co-transfection of 5-HT₇–CFP receptor and non-relevant transmembrane protein CD86–YFP. This protein is known to be a monomer and, therefore, it is often used as a negative control in methods that study protein–protein interaction by resonance energy transfer (James et al., 2006; Dorsch et al., 2009). In accordance with published data, in such negative control experiments we found significantly reduced, but still not zero, apparent FRET values (11±1% n=22; Fig. 2D). The main reason for such observation is an enriched local concentration of CD86–YFP and 5-HT7–CFP (which both are transmembrane proteins) at the plasma membrane after co-transfection, which results in nonspecific donor–acceptor interactions. Significantly lower FRET efficiency was also obtained after the co-transfection of CD86–YFP with 5-HT_1A–CFP and of CD86–YFP with CD86–CFP (data not shown). These results indicate that 5-HT_1A and 5-HT₇ receptors can form both homo- and heterodimers at the cell surface.

Results of the acceptor photobleaching FRET experiments demonstrated the existence of three principal kinds of oligomers after receptor coexpression, which includes two types of homodimer (5-HT_1A–5-HT_1A and 5-HT₇–5-HT₇) as well as heterodimer (5-HT_1A–5-HT₇). In addition, a certain number of receptors are expected to be expressed as monomers. To study the oligomerization behaviour of 5-HT_1A and 5-HT₇ receptors in more detail, we used the quantitative lux-FRET method (Wlodarczyk et al., 2008) to calculate and visualize (Fig. 3B) the apparent FRET efficiencies for donors, Ef_D, and acceptors, Ef_A, over a wide range of donor molar fractions, x_D, where x_D =[D^t]/([D^t]+[A^t]), [D^t] is the total donor concentration and [A^t] is the total acceptor concentration. To be able to compare FRET values obtained at different donor to acceptor ratios, the total concentration of plasmids encoding for donor and acceptor was held constant in all experiments. Based on the dependence of both Ef_D and Ef_A on x_D, we first estimated the number of units (n) participating in complex formation (Veatch and Stryer, 1977; Meyer et al., 2006) and obtained a best fit for the value of n=2 (R²=0.94 and 0.89 for 5-HT_1A and 5-HT₇ receptors, respectively), demonstrating the preferential formation of dimers (supplementary material Fig. S1). This is also in accordance with our previous results on the 5-HT_1A receptor (Kobe et al., 2008; Woehler et al., 2009). Further analysis revealed a linear dependence and symmetry of the apparent FRET efficiencies Ef_D and Ef_A over x_D in the case of 5-HT_1A and 5-HT₇ homodimers (Fig. 3A). By contrast, coexpression of 5-HT_1A and 5-HT₇ receptors resulted in highly non-symmetrical distribution of the Ef_D and Ef_A values (Fig. 3A), which cannot be sufficiently fitted (R²=0.76) by the model suggested by Veatch and Stryer (Veatch and Stryer, 1977). To explain such asymmetry, we developed a general dimerization model describing Ef_D and Ef_A as a function of the total donor and acceptor concentrations (see Materials and Methods for details). The model also considered possible differences in the interaction efficiencies between monomers for the formation of homo- and heterodimers as well as different characteristic FRET efficiencies for the dimer compositions (Fig. 3C; supplementary material Fig. S2). By fitting the model to experimental data we obtained relative dissociation constants in the order of K1A–1A=1.05>K_1A–7=0.27>K_7–7=0.016 (R²=0.93; Fig. 3C), where lower K values correspond to a higher tendency to form the particular dimer. For this calculation, the unknown total concentration of receptors (sum of donor and acceptor fluorophores) was assumed to be constant and used as the ‘unit’ concentration. Note that at the relative low, physiologically relevant total receptor concentrations used in the present study, the variability of Ef_D and Ef_A at high x_D values (i.e. high amount of donor and low amount of acceptor) is relative high, because at such conditions the specific YFP fluorescence can hardly be distinguished from cell background. According to these results, the 5-HT₇ receptor possesses the highest affinity to form homodimers, followed by the 5-HT₇–5-HT_1A heterodimers and 5-HT_1A–5-HT_1A homodimers. Using the reaction scheme shown in Fig. 3C, we were also able to predict the relative concentration of monomers and dimers at any given defined expression ratio between 5-HT_1A and 5-HT₇ receptors. As shown in Fig. 3D, differences in the affinity for forming homo- and heterodimers led to asymmetric distributions of relative concentrations for 5-HT_1A–5-HT_1A, 5-HT₇–5-HT₇ and 5-HT_1A–5-HT₇ dimers as well as the corresponding monomers, when plotted against the expression ratios (see also supplementary material Fig. S3). For instance, an equal amount of homo- and heterodimers can be obtained only at a 5-HT_1A to 5-HT₇ ratio of 2:1 (x_D value of 0.65), whereas at equal expression levels (1:1 ratio, x_D value of 0.5), the relative amount of 5-HT₇–5-HT_1A heterodimers will be higher than for the 5-HT_1A homodimers (Fig. 3D).

Next, we examined the consequences of 5-HT_1A–5-HT₇ heterodimerization for the agonist-induced receptor endocytosis by using quantitative analysis of surface-expressed receptors labelled with quantum dots (QDs). Equal labelling of each receptor subtype was assessed by fluorescence-activated cell sorting (FACS) analysis. Serotonin-mediated receptor internalization was then analysed in N1E-115 neuroblastoma cells using total internal reflection fluorescence (TIRF) microscopy by counting the number of QD-labelled puncta visible on the surface during the incubation time (Fig. 4). Analysis of non-stimulated cells revealed that the QDs were stably associated with the plasma membrane and did not change during the incubation period (data not shown). Prolonged stimulation of cells expressing HA-tagged 5-HT_1A receptors with serotonin showed no significant receptor internalization, even after 30 minutes of observation (n=4) (Fig. 4A,F,G; supplementary material Movie 1). By contrast, serotonin treatment of N1E cells expressing myc-tagged 5-HT₇ receptors resulted in profound internalization of receptors after 11±2 minutes (n=4) (Fig. 4B,F,G; supplementary material Movie 2). To analyse whether heterodimerization can influence receptor internalization, neuroblastoma cells were co-transfected with HA-tagged 5-HT_1A and myc-tagged 5-HT₇ receptors, and serotonin-mediated re-distribution of QDs bound to the 5-HT_1A receptors was studied. As shown in Fig. 4C, 5-HT_1A–5-HT₇ heterodimerization led to pronounced agonist-mediated co-internalization of 5-HT_1A receptor (n=4) (Fig. 4C,F,G; supplementary material Movie 3). It is noteworthy that treatment of cells coexpressing 5-HT_1A and 5-HT₇ receptors with specific 5-HT_1A receptor antagonist WAY100635 did not affect serotonin-mediated 5-HT_1A receptor internalization (Fig. 4D,F,G). By contrast, pharmacological blockade of 5-HT₇ receptor with SB269970 completely abolished agonist-induced 5-HT_1A receptor internalization (Fig. 4E,F,G). These results suggest that 5-HT₇-receptor-mediated signalling is necessary for initiation of the co-internalization of 5-HT_1A receptors.

Finally, we verified that the decreased intensity obtained by the TIRF analysis is indeed caused by the internalization of receptor-bound QDs. After TIRF measurements, all samples were imaged using confocal microscopy followed by 3D-image reconstruction. Such analysis revealed that the loss of QD fluorescence at the cell surface was accompanied by accumulation of fluorescence signal within intracellular compartments (Fig. 5).

The 5-HT_1A and 5-HT₇ receptors differ in their intracellular signalling in that 5-HT_1A is coupled to pertussis-toxin-sensitive members of the G_i/o families, whereas the 5-HT₇ receptor stimulates adenylyl cyclases via the G_s protein. To determine whether 5-HT_1A–5-HT₇ hetero-dimerization leads to changes in receptor-mediated signalling, we first examined receptor-mediated activation of heteromeric G proteins through a GTPγS coupling assay (Kvachnina et al., 2005). As expected, significant increase in [³⁵S]GTPγS binding to stimulatory Gα_s-subunit was measured upon serotonin treatment of cells expressing only the 5-HT₇ receptor (Fig. 6A). More importantly, 5-HT₇-receptor-mediated activation of G_s protein was not affected by the coexpression of 5-HT_1A receptors. By contrast, 5-HT_1A-receptor-mediated activation of inhibitory G_i protein obtained in cells expressing the 5-HT_1A receptors was decreased after coexpression of the 5-HT₇ receptor (Fig. 6A). Thus, 5-HT_1A–5-HT₇ heterodimerization specifically attenuates the ability of 5-HT_1A receptor to activate G_i protein.

The 5-HT_1A receptor can also activate the MAP kinases Erk1 and Erk2 (Della Rocca et al., 1999; Papoucheva et al., 2004). We, therefore, next examined whether heterodimerization can influence the Erk phosphorylation in cells expressing a constant amount of 5-HT_1A receptors, either alone or together with increased concentrations of YFP-tagged 5-HT₇ receptors (Fig. 6B). As shown in Fig. 6C,D, serotonin treatment resulted in a robust increase of Erk phosphorylation in cells expressing 5-HT_1A alone, and this response was continuously enhanced after coexpression of increasing amounts of the 5-HT₇ receptors. It is noteworthy that increased Erk phosphorylation was not mediated by the co-activation of Erk via the 5-HT₇ receptor, because (similarly to the untransfected cells) we did not detect any serotonin-mediated Erk activation in cells expressing the 5-HT₇ receptor alone (Fig. 6D). Taken together, these results demonstrate that the degree of heterodimerization specifically regulates 5-HT_1A-receptor-mediated Erk signalling.

G-protein-gated inwardly rectifying potassium (GIRK or Kir3) channels constitute an important physiological downstream target of the 5-HT_1A receptor, and channels of this type are activated by direct binding of βγ-subunits of inhibitory G proteins (Huang et al., 1995). Therefore, we next analysed whether the heterodimerization can alter the 5-HT_1A-receptor-mediated activation of Kir3.1/3.2 concatemers after their coexpression in Xenopus oocytes. When 5-HT_1A receptors were expressed together with Kir3.1/3.2, basal inward currents of 1.45±0.14 μA (n=23) were elicited upon elevation of the extracellular potassium concentration. Application of 500 nM serotonin further increased current amplitudes to 2.81±0.21 μA (n=23) (Fig. 7A,B). Notably, 5-HT₇ receptor expressed together with Kir3.1/3.2 (but without 5-HT_1A receptor) did not influence Kir3.1/3.2 channel-mediated currents under basal conditions nor after treatment with serotonin (Fig. 7A, lower trace). However, when 5-HT₇ receptors were expressed in addition to 5-HT_1A receptors and Kir3.1/3.2, both basal and agonist-induced currents were significantly reduced (basal Kir3.1/3.2 currents decreased to 0.53±0.08 μA and serotonin-mediated current to 0.81±0.11 μA, n=29, P<0.01; Fig. 7A,B). These effects were not mediated by the decreased amount of the 5-HT_1A receptor at the cell surface, because receptor density was not altered in oocytes coexpressing 5-HT_1A and 5-HT₇ receptors (supplementary material Fig. S4A,B). Note that the relative increase in concentrations of injected 5-HT₇ receptor RNA led to an increased inhibition of Kir3.1/3.2 currents. Although at a 5-HT_1A to 5-HT₇ ratio of 1:1 the current amplitude was reduced by 49%, a relative increase in the amount of 5-HT₇ cRNA resulting in a ratio of 1:5 led to an augmented reduction of current by 71% (Fig. 7C). The inhibitory effect of 5-HT₇ receptor was not affected after pre-incubation of oocytes with the selective 5-HT₇ antagonist SB-269970 (10 μM; Fig. 7D), suggesting the importance of receptor–receptor interaction rather than 5-HT₇-receptor-mediated signalling for the Kir3.1/3.2 inhibition.

Finally, we tested whether the inhibitory effect on potassium currents is selectively caused by the coexpressed 5-HT₇ receptors or it can be adjusted by other GPCRs. Agonist-induced currents were not significantly changed in comparison with values obtained in oocytes expressing only 5-HT_1A receptors when the serotonin receptor 5-HT_2C was coexpressed with 5-HT_1A(relative current 0.92±0.05, n=5). Also, coexpression of G_s-coupled β1-adrenergic receptors (relative current 0.78±0.15, n=7) as well as G_q-coupled histamine H1 (relative current 0.88±0.11, n=4) or bradykinin B1 receptors (relative current 1.07±0.08, n=7), did not result in any significant change in potassium current (supplementary material Fig. S4C,D). These experiments demonstrate that 5-HT₇ receptors can selectively inhibit activation of Kir3.1/3.2 currents via interaction with 5-HT_1A receptors.

Having demonstrated the inhibitory role of 5-HT_1A–5-HT₇ receptor heterodimerization on 5-HT_1A-receptor-mediated potassium currents in a recombinant system, we next analysed whether this effect also takes place in neurons. As a model system we used mouse hippocampal neurons, which have been shown to produce robust 5-HT_1A-receptor-mediated K⁺ current via GIRK channels (Delling et al., 2002). As illustrated in Fig. 8A, hippocampal neurons express both 5-HT_1A and 5-HT₇ receptors and these receptors are highly co-localized at the plasma membrane. Co-immunoprecipitation assays performed with brain samples prepared from mice at postnatal day 6 (P6) also demonstrated that these receptor can form heterodimers in vivo (Fig. 8B). To study the functional implication of 5-HT_1A–5-HT₇ heterodimerization, we developed short interfering RNAs (siRNAs) to specifically knockdown the endogenously expressed 5-HT₇ receptor (Kobe et al., 2012) (supplementary material Fig. S5A). The expression vectors encoding the specific siRNA also contained GFP, allowing for simple identification of transfected neurons by green fluorescence (Fig. 8C). Transfection of hippocampal neurons with a mixture of 5-HT₇ receptor silencing vectors resulted in an increase in basal GIRK currents, as compared with control neurons transfected with the scrambled siRNA (P<0.05, U-test; Fig. 8D,E). Application of 5-HT_1A receptor agonist 8-OH-DPAT significantly potentiated GIRK currents in both transfected groups (P<0.05, Wilcoxon signed rank test), and the amplitude of currents remained significantly larger in 5-HT₇ receptor silenced neurons (Fig. 8D,E). Because all experiments were performed in the presence of 5-HT₇ receptor antagonist SB-269970, these data strongly suggest that direct 5-HT_1A–5-HT₇ receptor interaction rather than 5-HT₇-receptor-mediated signalling is responsible for the smaller currents in non-silenced cells.

Finally, we analysed whether the relative concentration of heterodimers, which crucially depends on the expression ratio of both receptors (Fig. 3), undergoes developmental changes. For that, we determined the expression profiles for both 5-HT_1A and 5-HT₇ receptors in the mouse hippocampus at different stages of postnatal development using real-time PCR. This approach demonstrated that 5-HT₇ receptor transcripts were strongly expressed during early postnatal stages (P2 and P6) and downregulated during later developmental stages (supplementary material Fig. S5B). By contrast, expression levels of the 5-HT_1A receptor mRNA transcripts were not significantly modulated during development (supplementary material Fig. S5C). Because the protein expression level is assumed to roughly correlate with the level of mRNA transcripts, the above data suggest that receptor expression also undergoes developmental regulation. Such differences in the expression levels result in drastic changes of the 5-HT_1A to 5-HT₇ ratio from 3:1 at P2, to 6:1 at P6, 12:1 at P12 and 35:1 at P90 (Fig. 8F). According to our dimerization model (Fig. 3D), this means that at the early postnatal stage (P2) hippocampal neurons express similar amounts of homo- and heterodimers ([5-HT_1A–5-HT_1A]=13% and [5-HT_1A–5-HT₇]=9%). During development, the relative concentrations of 5-HT₇ receptors continuously decreased, resulting in a decrease in the amount of 5-HT_1A–5-HT₇ heterodimers (e.g. at P90, [5-HT_1A–5-HT_1A]=23% and [5-HT_1A–5-HT₇]=2%). These combined results demonstrate that the relative amount of 5-HT_1A–5-HT₇ receptor heterodimers and, consequently, their functional role in inhibition of GIRK currents is progressively decreased during brain development.

The existence of GPCR homo- and heterodimers has become generally accepted, and a growing body of evidence points to the functional importance of oligomeric complexes for the receptor trafficking, receptor activation and G protein coupling in native tissues (Bouvier, 2001; Rivero-Müller et al., 2010). The clinical significance of GPCR oligomerization has also become more evident in recent years, leading to identification of receptor oligomers as a novel important therapeutic target (Waldhoer et al., 2005; González-Maeso et al., 2008).

In the present study, we provide biochemical and biophysical evidence for the heteromerization of two serotonin receptors, 5-HT_1A and 5-HT₇. Although our experimental results suggest preferential formation of heterodimers, we still cannot exclude the possibility that these receptors can form higher-order oligomers. Indeed, the models that have been previously developed for estimating the number of units interacting in an oligomeric complex can identify a case of dimerization, although they cannot accurately quantify the number of units reacting if this number is above two (Veatch and Stryer, 1977; Meyer et al., 2006). Moreover, homo-FRET analysis of 5-HT_1A receptors stably expressed in CHO cells provided first experimental evidence for the existence of higher-order 5-HT_1A homo-oligomers (Ganguly et al., 2011). Therefore, future investigations involving homo-FRET experiments in combination with extended oligomerization models will be needed for a better understanding of 5-HT_1A and 5-HT₇ oligomerization behaviour.

The results of co-immunoprecipitation experiments in mouse brain provided direct evidence that these receptors can form heteromers in vivo. Utilizing FRET techniques, we demonstrated that 5-HT_1A and 5-HT₇ form constitutive and agonist-independent heterodimers at the plasma membrane of living cells. We also found that both 5-HT_1A and 5-HT₇ receptors can form homodimers when expressed alone (Kobe et al., 2008; Woehler et al., 2009). This observation suggests that, in addition to 5-HT_1A–5-HT₇ heterodimers, two types of homodimers composed either of 5-HT_1A or 5-HT₇ receptors together with the corresponding monomers can simultaneously exist in cells coexpressing both types of receptor (which is often the case in native tissues). This should also be true for other oligomerizing receptors, and the coexistence of the corresponding homomers was experimentally confirmed when heterodimerization of AT₁–B₂ and δOR–β₂AR were analysed (AbdAlla et al., 2000; Jordan et al., 2001). However, the issue of the relative concentration of monomers, homo- and heteromers still remains open, not least because of the absence of suitable methodology. Such knowledge, however, is of particular importance because heterodimers often possess distinct pharmacological or functional properties in comparison with monomers and homodimers (Rozenfeld and Devi, 2011).

So far, various FRET strategies have been used to prove the existence of well-defined complexes only (either homo- or heterodimers). Thus, in the case of heterodimers the possible coexistence and/or quantitative analysis of corresponding homodimer fractions has not been taken into consideration. In the present study, we were able to calculate the relative dissociation constants for hetero- and homodimers by a combination of lux-FRET with an appropriate dimerization model. This model allowed us for the first time to compare the relative concentrations of homo- and heterodimers as well as the corresponding monomers under physiological conditions. A detailed analysis of oligomerization behaviour revealed that the 5-HT₇ receptor possesses a higher binding affinity for formation of homodimers than for 5-HT₇–5-HT_1A receptor heterodimer and 5-HT_1A receptor homodimers. One functional consequence of different affinities for homo- and heterodimers is that, even at a relatively low expression level of 5-HT₇ receptors, the amount of 5-HT₇–5-HT_1A receptor heterodimers and, consequently, their functional implication will be relatively high.

Analysis of the functional consequences of dimerization between 5-HT_1A and 5-HT₇ receptors revealed that heterodimerization decreases the 5-HT_1A-receptor-mediated activation of G_i protein without affecting 5-HT₇-receptor-mediated G_s protein activation. Because G protein activation is mainly mediated through the stabilization of receptor in the active conformation (Gether et al., 2002; Wess et al., 2008), decreased activation of G_i protein in the case of 5-HT_1A–5-HT₇ heterodimer might be explained by the partial destabilization of the 5-HT_1A receptor conformation induced by the direct interaction with the 5-HT₇ protomer. This might result in formation of a modified binding surface that provides increased specificity for the G_s protein. Based on the atomic model of rhodopsin, it has been proposed that one GPCR dimer possesses the optimal docking interface for only one G protein heterotrimer (Fotiadis et al., 2006; Palczewski, 2010). Thus, activation of 5-HT₇ protomer in the dimer might induce preferential association of G_s protein with the complex, leading to diminished G_i-protein-mediated signalling. Such mechanism of allosteric modulation between two protomers is further confirmed by the fact that heteromerization often results in an increased G protein activation of the one associated receptor within a heteromer (Rocheville et al., 2000; González-Maeso et al., 2008).

Another important finding in this study is that heterodimerization markedly alters the internalization profile of 5-HT_1A receptors. Whereas 5-HT_1A receptors expressed alone are resistant to the agonist-mediated internalization, 5-HT_1A receptors participating in 5-HT_1A–5-HT₇ heterodimers undergo efficient internalization upon serotonin treatment. The fact that the pharmacological blockade of 5-HT₇ receptors, but not of 5-HT_1A receptors, abolishes internalization of both 5-HT₇ homo- and heterodimers suggests that 5-HT₇-receptor-mediated signalling represents an initial step responsible for 5-HT_1A co-internalization. Generally, internalization of GPCRs is initiated by the agonist-mediated receptor phosphorylation by GPCR kinases followed by the recruitment of β-arrestin and the assembly of clathrin-coated pits, leading to removal of receptor from the cell surface (Ferguson, 2001; Drake et al., 2006). Once internalized, receptors can initiate additional, G-protein-independent signalling pathways such as a β-arrestin-mediated coupling to MAP kinase (Kovacs et al., 2009). The best-studied example of such signalling is the angiotensin AT₁ receptor, which activates MAP kinase Erk in two different ways: first, by a G-protein-dependent pathway that results in transient Erk phosphorylation and targets Erk into the nucleus or, second, by a β-arrestin-dependent pathway that leads to sustained ERK phosphorylation, which directs Erk to the cytosol (Ahn et al., 2004). Such spatio-temporal segregation of ERK signalling has been shown to result in activation of distinct downstream signalling cascades (Luttrell et al., 2001; Wei et al., 2004). Our experimental data suggest that a similar scenario is also relevant for the 5-HT_1A receptors residing within 5-HT_1A–5-HT₇ heterodimers. When 5-HT_1A receptor monomers and/or homodimers build a dominant population, receptor-mediated G_i protein activation represents the main factor responsible for Erk phosphorylation (Fig. 6) (Papoucheva et al., 2004). In the case of heterodimers, Erk phosphorylation significantly increases (despite the fact that the coupling of 5-HT_1A receptor to G_i protein is reduced under these conditions), suggesting that serotonin-mediated co-internalization of 5-HT_1A receptor can initiate β-arrestin-mediated Erk phosphorylation. Thus, dependent on the relative amount of heterodimers, this mechanism can allow the same ligand (serotonin) to activate distinct Erk-mediated pathways (i.e. G-protein-dependent or β-arrestin-dependent). This also raises the possibility that conditions that selectively promote or inhibit heterodimerization could have a significant physiological relevance.

In addition, we demonstrated that 5-HT_1A–5-HT₇ heterodimerization markedly decreases the ability of 5-HT_1A receptor to activate GIRK channels, an effect mediated through the Gβγ subunits of inhibitory G proteins (Reuveny et al., 1994; Kofuji et al., 1995). The finding that pharmacological blockade of 5-HT₇ receptor does not overcome this inhibitory effect suggests that direct receptor–receptor interaction rather than 5-HT₇-receptor-mediated signalling is responsible for the reduced GIRK channel activation. The inhibitory effect of 5-HT_1A–5-HT₇ heterodimerization on GIRK channel currents was also found in hippocampal neurons, which suggests a physiological relevance of heteromerization in a neuronal context. The 5-HT_1A-receptor-mediated opening of GIRK channels, leading to membrane hyperpolarization and a decrease in neuronal input resistance, is one of the main physiological effects of serotonin in the CNS (Araneda and Andrade, 1991; Tanaka and North, 1993; Lüscher et al., 1997).

With respect to development, it has been shown that the effects of 5-HT on membrane potential undergo pronounced changes. Although at early developmental stages serotonin has only a marginal membrane hyperpolarizing effect, a stronger hyperpolarization becomes the dominant effect of serotonin in adult animals (Segal, 1990; Béïque et al., 2004). Molecular mechanisms underlying such developmental changes in 5-HT action are poorly understood. Our results suggest that differential heterodimerization rates between 5-HT_1A and 5-HT₇ receptors during development can provide an intriguing explanation for this effect. We found that the expression level of 5-HT₇ and 5-HT_1A receptors in the hippocampus varies during development. Although the amount of 5-HT₇ receptor progressively decreases, the 5-HT_1A receptor expression remains relative stable. Therefore, the relative concentration of 5-HT_1A–5-HT₇ heterodimers and, as a consequence, their functional importance also undergo pronounced developmental changes. A relative high expression level of 5-HT_1A–5-HT₇ heterodimers at the early postnatal stages will result in reduced coupling of 5-HT_1A receptor to GIRK channels and, consequently, in decreased membrane hyperpolarization due to a lower number of open channels. With increasing age, the relative amount of heterodimers gradually decreases to only 2% at P90, allowing 5-HT_1A homodimers together with monomers to become the dominant populations. Thus, the inhibitory influence of heterodimers on basal and 5-HT_1A-receptor-mediated GIRK channel activation begins to subside and is gradually replaced by a hyperpolarizing effect mediated by the 5-HT_1A homodimers and/or monomers.

In addition to the role of heterodimerization in regulation of Erk signalling and GIRK channel activation, our results demonstrate that the heterodimerization can modulate the agonist-mediated internalization of 5-HT_1A receptor. It has been shown that although the 5-HT_1A receptor is expressed both as a presynaptic autoreceptor in serotonergic neurons of raphe nuclei (Hamon et al., 1990; Riad et al., 2000) and as a postsynaptic receptor in multiple brain regions including hippocampus and cortex (Beck et al., 1992; Aznar et al., 2003), chronic receptor stimulation results in functional desensitization of only 5-HT_1A autoreceptors without affecting the postsynaptic 5-HT_1A receptors (Jolas et al., 1994; Le Poul et al., 1995). Our data suggest that the higher amount of heterodimers produced in presynaptic neurons than in postsynaptic neurons might represent a mechanism responsible for the differential desensitization obtained for the 5-HT_1A auto- and heteroreceptors. This is supported by several observations. First, analysis of the brain regional distribution of 5-HT₇ receptor has revealed that this receptor is highly enriched in serotonergic neurons of dorsal raphe nuclei in adults (Neumaier et al., 2001; Bonaventure et al., 2002; Martín-Cora and Pazos, 2004). Together with the finding that the 5-HT_1A receptor has a higher affinity for forming heterodimeric (5-HT_1A–5-HT₇) rather than homodimeric (5-HT_1A–5-HT_1A) complexes, this suggests that serotonergic neurons in raphe nuclei express a relative high level of heterodimers, i.e. [5-HT_1A–5-HT₇]>[5-HT_1A–5-HT_1A]. From a functional point of view, this will result in effective co-internalization of 5-HT_1A receptor within 5-HT_1A–5-HT₇ heterodimeric complexes upon serotonin release.

In the present study, we also found that the expression level of postsynaptic 5-HT₇ receptors in the hippocampus progressively decreased during postnatal development without changes in 5-HT_1A receptor expression. Similar results were obtained in forebrain, where expression of 5-HT₇ receptor in pyramidal neurons has been shown to diminish dramatically with increasing age (Béïque et al., 2004). These observations suggest that under physiological conditions 5-HT_1A homodimers represent a dominant receptor population in hippocampus in adulthood, i.e. [5-HT_1A–5-HT_1A]»[5-HT_1A–5-HT₇]. Because 5-HT_1A receptors expressed alone are resistant to the agonist-mediated internalization, 5-HT released in hippocampus or cortex will not reduce the amount of postsynaptic 5-HT_1A receptors at the cell surface. The mechanism proposed here not only explains the differences in desensitization between pre- and postsynaptic 5-HT_1A receptors, but also suggests that the regulated and balanced ratio of homo- and heterodimerization on pre- and postsynaptic neurons might be crucially involved in both the onset and response to treatment of psychiatric diseases such as depression and anxiety.

The construction of HA-tagged 5-HT1_A and 5-HT₇ receptors as well as receptors fused to different spectral variants of the green fluorescence proteins has been described previously (Kvachnina et al., 2005; Kobe et al., 2008). YFP-tagged CD86 was a kind gift from Moritz Bünemann, University of Würzburg, Germany (Dorsch et al., 2009). Note that the monomeric versions of CFP and YFP were used to produce all constructs used in the present study. Mouse N1E-115 neuroblastoma cells from the American Type Culture Collection (ATCC) were grown in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum (FCS) and 1% penicillin/streptomycin at 37°C under 5% CO₂. For transient transfection, cells were transfected with appropriate vectors using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. The amount of expressed receptor was measured in membrane preparations of transfected cells by using a radioactive ligand binding assay with [³H]8-OH-DPAT as a specific ligand and non-radioactive 5-HT as a competitor.

Co-immunoprecipitation and immunoblotting in neuroblastoma N1E-115 cells coexpressing HA-tagged 5-HT₇ and YFP-tagged 5-HT_1A receptors were performed as described (Kobe et al., 2008). The presence of YFP-tagged receptors was verified by the fluorescence scanner Typhoon 9400 (GE Healthcare).

For co-immunoprecipitation analysis from brain, samples from P6 NMRI mice were isolated and homogenized in buffer containing 10 mM HEPES (pH 7.4), 5 mM EGTA, 1 mM EDTA and 0.32 M sucrose. The membrane fraction was then isolated and dissolved in lysis buffer. The lysates were incubated with a rabbit polyclonal antibody directed against the murine 5-HT₇ receptor (1:200 dilution; AbD Serotec, Düsseldorf, Germany) followed by incubation with Protein-A Sepharose, SDS-PAGE and western blot with antibody directed against 5-HT_1A receptor (1:200 dilution; Alomone, Jerusalem, Israel). All animal experiments were performed according to the relevant regulatory standards.

Neuroblastoma N1E-115 cells expressing 5-HT_1A–CFP and/or 5-HT₇–YFP receptor were analysed using a spectrofluorometer (Fluorolog 3-22, Horiba JobinYvon, Unterhaching, Germany).

To determine the apparent FRET efficiency for 5-HT_1A homodimers, 5-HT₇ homodimers and 5-HT_1A–5-HT₇ heterodimers, we used a recently developed lux-FRET method that has been described in detail (Wlodarczyk et al., 2008). This method allows calculation of the total concentration ratio [A^t]/[D^t] of donor and acceptor, a donor molar fraction x_D=[D^t]/([D^t]+[A^t]) as well as the apparent FRET efficiencies Ef_D and Ef_A, where f_D=[DA]/[D^t] and f_A=[DA]/[A^t] are the fractions of donors and acceptors in complexes.

During the time-course experiments, the required two emission spectra for lux-FRET analysis were only obtained at the first and last time point by exciting at 440 nm and 488 nm with 2 nm spectral resolutions for emission and 0.5 second integration time. To achieve an appropriate time resolution, only the 440 nm excitation was applied at the intermediate time points, the second excitation data were approximated from the accompanying measurements at the beginning and the end. Stimulation was carried out using 5-HT (Sigma) at a final concentration of 10 μM after 4 minutes. As reference, the same volume of buffer solution was applied.

In all measurements, the spectral contributions due to light scattering and nonspecific fluorescence of the cells were taken into account by fitting reference spectra of donor and acceptor, the emission spectra of non-transfected cells (background) and the Raman scattering spectra to each spectrum.

Due to the complexity of that fourth order equation (which cab be solved only analytically), it is difficult to apply these for analysis of experimental data. However, several approximations and special cases can provide us with important information about the dimerization process.

For homodimers with K_DA = K_DD =K_AA we receive the well-known linear dependence of eq. 3 mentioned above.

In our experiments we found a linear dependence of f_D=f(x_D) for 5-HT_1A and 5-HT₇ receptor homodimers, but a significantly nonlinear dependence for the heterodimers (Fig. 3A). Moreover, in the case of heterodimer, the dependencies of f_D and f_A were not symmetric. Thus, on the basis of our model we cannot assume the high affinity case. On the other hand, experimental apparent FRET efficiencies allowed us to assume a comparably high quantity of FRET complexes, which is in conflict with the low affinity case. Thus, we analysed the measured apparent FRET dependencies using a numerical solution of eq. 5.

We found that various combinations of K_i and E can fit the individual dependencies with almost similar fit error. Thus, we analysed the dependencies of the individual K_i according to given E values. Supplementary material Fig. S2A,C shows the functional dependence of individual fits for given E from 0.2 to 1 and supplementary material Fig. S2B,D shows the fit parameters K_i and the relative error of the fit result. In the case of homodimers, the model delivers a constant fit quality for E values above a minimal threshold of E=24% and E=22%, for 5-HT_1A and 5-HT₇ homodimers, respectively, where higher E values correspond to higher K values. However, the relation between the K values remains preserved, where the K values for 5-HT_1A are slightly higher than for 5-HT₇ homodimers. The heterodimer model fit requires a significantly higher minimal FRET efficiency than for 5-HT_1A and 5-HT₇ homodimers. In contrast to the homodimer model fitting, the E value affects the fit quality in the heterodimer cases. For higher E values, the fit error significantly increases. For complexes of 5-HT_1A–CFP and 5-HT₇–YFP the best fit was obtained for E=40% and for complexes of 5-HT₇–CFP and 5-HT_1A–YFP the best fit was obtained for E=32%. Finally, we prepared a global fit that allowed individual E values for the different homo- and heterodimers (Fig. 3A). Best fitting results were found for E values slightly higher than proposed above for the individual fits. In this case, K_{5-HT1A–5-HT1A}>K_{5-HT7–5-HT1A}>K_{5-HT7–5-HT7}, which is in line with results obtained for homodimers.

With the information on the relative dissociation constants, we simulated the concentration profile of dimers and monomers according to our model for a total concentration ([Dt]+[At]) ranging from 10⁻² to 10² (supplementary material Fig. S3). The concentration pattern at log10([5-HT_1A]_tot+[5-HT₇]_tot)=0 reflects the situation shown in Fig. 3C. For higher total concentrations, log10([5-HT_1A]_tot+[5-HT₇]_tot)=2, the dimer concentration pattern becomes more symmetrical, which was already predicted from the high concentration or high affinity approximation. At this concentration range, monomer concentrations are very small and the fits do not contain any information about individual affinities (see eq. 8). However, at low total concentrations, log10([5-HT_1A]_tot+[5-HT₇]_tot)=−2, the monomeric forms are preferentially present. In this case, the distribution of dimers becomes asymmetric with respect to x_D, and the fits become sensitive to individual affinities. However, at this concentration range the dimer concentration becomes very small compared with the monomer concentration. Consequently, the amplitudes of f_D and f_A are very small and FRET signals cannot be detected. Thus, if we observe an asymmetry in the Ef_D and Ef_A functions, we can suggest that monomers and dimers are expressed at similar concentrations (which allows extraction of information about the individual affinities).

Recombinant N-terminally HA-tagged 5-HT_1A and myc-tagged 5-HT₇ receptors (Santa Cruz Biotechnology) were used for labelling of receptors with QDs at the surface of living cells. Cells were incubated with 1 ng of primary antibody diluted in OptiMEM for 5 minutes and then extensively washed with OptiMEM before addition of 1 nM QD–Fab conjugates (Invitrogen) in OptiMEM for 5 minutes. QDs were removed by extensive washing over a period of 10 minutes. All staining and washing steps were performed at room temperature.

The TIRF setup was based on an IX71 microscope (Olympus) equipped with a 60× 1.45 NA Plan Apochromat Olympus objective, an Olympus TIRF condenser and a diode laser emitting at 405 nm (Toptica Photonics, Gräfelfing, Germany). Images were acquired with an Andor iXon camera controlled with Andor iQ software (Andor, Belfast, Northern Ireland). The acquisition rate was 0.25 Hz and the exposure time was 300 milliseconds. To analyse accumulation of receptors in intracellular organelles, three-dimensional reconstructions were made from confocal 3D stacks acquired with a confocal laser-scanning microscope Meta-LSM 510 (Zeiss, Germany). For data acquisition and analysis of confocal images we used the LSM 510 software. Subsequent imaging procedures were performed using NIH ImageJ (http://rsb.info.nih.gov/ij/).

Agonist-promoted binding of [³⁵S]guanosine 5-(3-O-thio)triphosphate to different G proteins caused by stimulation of 5-HT_1A and/or 5-HT₇ receptors was performed as described previously (Ponimaskin et al., 2000).

For Erk phosphorylation assay, neuroblastoma cells were transfected with 5-HT_1A–mCherry receptor (1 μg of corresponding plasmid) together with increasing amount of 5-HT₇–YFP receptor (0, 0.2, 0.6 and 1 μg of corresponding plasmid). At 24 hours after transfection, cells were stimulated for 5 minutes with 10 μM 5-HT and then lysed in the loading buffer. Equal amounts of proteins in lysates were separated by SDS-PAGE and then subjected to western blot. The membranes were probed either with antibodies raised against phosphorylated Erk1/2 (phospho-p42/44; 1:2000 dilution) or against total Erk (p42/44; 1:1000 dilution). Receptor expression in gels after SDS-PAGE was visualized by the fluorescence scanner Typhoon 9400 (GE Healthcare). The amounts of phosphorylated and total Erk1/2 were quantified by densitometric measurements using GelPro Analyser version 3.1 software.

For recombinant protein expression in Xenopus oocytes, cDNAs of 5-HT_1A receptor, 5-HT₇ receptor, 5-HT_2C receptor, β1-adrenoreceptors, H1 histamine receptors, B1 bradykinine receptors and Kir3.1/3.2 concatemers were subcloned into the polyadenylation vector pSGEM, respectively. Capped run-off poly(A)+cRNA transcripts were synthesized from linearized cDNA and subsequently injected into defolliculated oocytes. Xenopus oocytes were incubated at 20°C in ND96 solution (96 mM NaCl, 2 mM KCl, 1 mM MgCl₂, 1 mM CaCl₂ and 5 mM HEPES, pH 7.4) supplemented with 100 μg/ml gentamicin and 2.5 mM sodium pyruvate. Two-electrode voltage-clamp recordings were performed 48 hours after injection. Currents were recorded with a Turbo Tec-10 amplifier (npi electronic, Tamm, Germany) and sampled through an EPC9 (HEKA Elektronik, Lambrecht, Germany) interface using Pulse/Pulsefit software (HEKA Elektronik). For rapid exchange of external solution, oocytes were placed in a small perfusion chamber with a constant flow of ND96 or high K⁺ medium (96 mM KCl, 2 mM NaCl, 1 mM MgCl₂, 1 mM CaCl₂ and 5 mM HEPES, pH 7.4).

Primary murine hippocampal neuronal cultures from 1- to 3-day-old C57BL6/J mice pups were prepared as described previously (Dityateva et al., 2003). On day in vitro 8 (DIV 8), primary hippocampal neurons were transfected with a control pSUPER-Mamm-X/scrambled shRNA plasmid (2 μg per well) or were co-transfected with two plasmids encoding shRNA to silence the expression of 5HT₇ receptor. A modification of the calcium phosphate precipitation method was used for transfection (Jiang and Chen, 2006). Neurons were used for electrophysiological recordings 3–4 days after the transfection. Whole-cell recordings from pyramidal neurons were obtained as previously described (Moult et al., 2006). Electrodes with a resistance of 3–4 MOhm were filled with solution (130 mM potassium gluconate, 8 mM NaCl, 4 mM Mg-ATP, 0.3 mM Na-GTP, 0.5 mM EGTA and 10 mM HEPES, pH 7.25). Cells were perfused continuously with HEPES-buffered saline (HBS) of the following composition: 119 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 25 mM HEPES, 20 mM D-glucose, 0.0005 mM Na⁺ channel blocker tetrodotoxin citrate (Tocris) and 0.001 mM 5-HT₇ receptor antagonist SB269970, pH 7.3. Currents were recorded in GFP-expressing neurons with an EPC 10 USB Patch Clamp Amplifier (HEKA Elektronik). Data acquisition and command potentials were controlled by PATCHMASTER software (HEKA Elektronik) and traces were digitalized at 5 kHz and stored for off-line analysis. To activate GIRK channel currents, 200-millisecond voltage steps from −60 to −120 mV were delivered at 10-second intervals, and leak and capacitive transients were digitally subtracted (Delling et al., 2002). Transfected cultures were coded, and recordings were performed without knowing the identity of delivered plasmids. After recording basal GIRK channel currents, 1 μM of 5-HT_1A receptor agonist 8-OH-DPAT was applied, followed by 1 mM BaCl₂ to block GIRK currents. The currents recorded in the presence of BaCl₂ were digitally subtracted from basal and 8-OH-DPAT-activated currents. The mean amplitudes of currents activated 180–200 milliseconds after the beginning of the voltage step were measured and statistically evaluated using non-parametric tests.