Introduction

Serotonin (5-hydroxytryptamine, 5-HT) is involved in many psychiatric diseases including depression, addiction, impulsivity, or psychosis. The 5-HT neurons that innervate forebrain originate predominantly from the rostral cell group of neurons in the dorsal raphe nucleus (DRN) [1, 2]. These neurons express the serotonergic markers tryptophan hydroxylase (TPH2) and 5-HT transporter (SERT), and also the negative autoreceptors, 5-HT1A and 5-HT1B receptors, whose expression is restricted to somatodendritic compartments of 5-HT neurons, and to axonal terminals, respectively [3]. The 5-HT1A autoreceptor activation elicits an outward current carried through G protein-coupled inwardly rectifying potassium channels (GIRK) of the Kir3 family leading to membrane hyperpolarization and inhibition of 5-HT neuron firing [4]. The presence of synaptic vesicles in dendrites of 5-HT neurons led to the suggestion that autoinhibition is mediated via dendritic release of 5-HT, for a review see Andrade et al. [5]. However, activity of 5-HT DRN neurons can also be positively modulated by 5-HT2A/2B/2C receptors triggering directly or indirectly inward currents [6,7,8,9,10]. Upon electrical stimulation of leech 5-HT neurons, transmembrane Ca2+ entry through L-type channels first evokes an early dendritic exocytosis; subsequently, the released 5-HT activates dendritic 5-HT2 autoreceptors coupled to Gq and phospholipase C, resulting in a positive feedforward loop that maintains sustained exocytosis [11]. It has thus been proposed that DRN neurons can display responses ranging from inhibition to excitation depending on a balance of functional 5-HT1A and 5-HT2 receptors [12]. However, a direct action of 5-HT2 receptor subtypes at 5-HT neurons has not yet been clearly established.

In humans, a loss-of-function polymorphism of 5-HT2B receptors is associated with 5-HT-dependent phenotypes, including impulsivity and suicidality [13]. Pharmacological experiments indicated that the 5-HT2B receptor preferential agonist BW723C86 [14] mimicked serotonin-specific reuptake inhibitor (SSRI) action in the forced swimming test (FST), which was abolished by injection of RS127445, a highly selective 5-HT2B receptor antagonist, or in mice totally knocked out (KO) for 5-HT2B receptor gene (Htr2b−/−) [15]. Long-term behavioral and neurogenic SSRI effects were also abolished after either genetic ablation of 5-HT2B receptors or chronic exposure to RS127445 [16]. Genetic (KO) or pharmacologic (antagonist) manipulation of 5-HT2B receptors interferes similarly with effects of other molecules that target 5-HT neurons including amphetamine derivatives, 5-HT releasers MDMA (3,4-methylenedioxy-methamphetamine) and dexfenfluramine [14, 17, 18]. For example, the locomotor response to the psychostimulant MDMA was abolished in Htr2b−/− or RS127445-treated wild-type mice [18]. Furthermore, local infusion of BW723C86 in DRN by microdialysis increased extracellular 5-HT that was blocked by RS127445, supporting a functional role of this receptor within the raphe and the selectivity of BW723C86 for this effect [18].

Together, these studies suggested that 5-HT2B receptors could be implicated in 5-HT-dependent behavior by acting directly and positively onto the 5-HT neurons. Nevertheless, the precise localization and way of action of 5-HT2B receptors are still poorly investigated. Main difficulties include a lack of specific antibody as well as low level of expression of 5-HT2B receptors in mouse [19] or human brains [20, 21]. Nevertheless, expression of 5-HT2B receptor mRNA was confirmed in several brain nuclei including DRN [13, 22]. Besides, previous study using single-cell reverse transcription PCR [16] permitted to establish a 5-HT2B receptor expression not only in raphe but more specifically in 5-HT neurons. Nevertheless, precise role of 5-HT2B receptors in 5-HT neurons remains elusive since pharmacologic manipulations or full gene KO mice do not distinguish direct effects on 5-HT neuron from circuit effects.

In this work, we tested the hypothesis of a direct positive regulation of 5-HT neurons activity by 5-HT2B receptors. First, we established that stimulation of 5-HT2Breceptors can increase the firing frequency of 5-HT neurons, and counteract 5-HT1A autoreceptor-dependent reduction in firing rate and hypothermic response in wild-type mice. Then, a conditional genetic ablation that eliminates 5-HT2B-receptor expression specifically and exclusively from 5-HT neurons (Htr2b5-HTKO mice) demonstrated a lack of effects of drugs, including SSRIs and MDMA, targeting 5-HT neurons, which present a lower firing rate and a stronger 8-OHDPAT-induced hypothermic response revealing its importance to maintain serotonergic tone. These observations established the direct modulation of 5-HT neurons by 5-HT2B receptors acting in an opposite way as to 5-HT1A-negative autoreceptors.

Materials and methods

Generation and use of mice

KO mice, 129S2.Cg-Htr2btm1Lum/Htr2btm1Lum (Htr2b−/−), were generated on the 129S2/SvPas strain (see extended methods in Supplementary materials). Floxed mice, Htr2btm2Lum/Htr2btm2Lum, (Htr2blox/lox), were generated on a mixed B6;129S2 background and backcrossed >10 times onto the 129S2 strain. Htr2blox/lox mice were inactivated for Htr2b in 5-HT neurons by crossing with 129S2.Cg-Tg(Fev-cre)1Esd/0 (ePet1-Cre BAC transgenic mice or Pet1-Cre+/0) [23] generating 129S2.Cg-Pet1-Cre+/0; Htr2blox/Htr2blox conditional KO mice (Htr2b5-HTKO) and littermate controls (Htr2blox/lox). Then, 129S2.Cg-Tg(ROSA)26Sortm1(CAG−EGFP)Fsh (Rosa26;CAG-loxP-STOP-loxP-EGFP or RCE) were crossed with Pet1-Cre+/0 generating (Pet1-GFP) that express GFP in Pet1-positive 5-HT neurons only after Cre recombination. All experiments involving mice were approved by the local ethical committee (No. 1170.02).

Ex vivo electrophysiology: cell attached

Electrophysiological recordings and molecular characterization of 5-HT neurons was conducted in acute brain slices from male Pet1-GFP mice aged 3–4 weeks old according to published procedures [24, 25].

Viral constructs and stereotaxic injection

To express HA-tagged 5-HT2B receptor specifically in 5-HT neurons, we use a double floxed inverse orientation (DIO) adeno-associated virus (AAV) construct that allows Cre-mediated expression of the transgene (pAAV-EF1A-DIO-WPRE-pA vector; Addgene). The viruses packaged into AAV2.9 serotype with titers of 1012–1013 viral particles/ml were obtained (UNC Vector Core, Chapel Hill, NC, USA). AAV2.9 DIO-TdTomato was used as a control. AAVs were injected in the B7 raphe nuclei of Pet1-GFP following described procedure [26].

Ex vivo electrophysiology: current clamp

Mice of either sex aged ~9 weeks were deeply anesthetized with ketamine and xylazine (150 and 10 mg/kg, respectively; intraperitoneally; Sigma-Aldrich, France) and immediately sacrificed. Coronal brain slices containing the raphe nucleus were prepared as previously described [27].

8-OHDPAT-induced hypothermia

Body temperature was measured intrarectally in mice every 10 min after the agonist injection during 40 additional minutes [28].

In vivo electrophysiology of DRN 5-HT neurons

The extracellular recordings of 5-HT neurons in DRN were performed in anesthetized mice (chloral hydrate, 400 mg/kg; intraperitoneally) according to Rainer et al. [29]. The DRN 5-HT neurons were identified according to the criteria of Aghajanian and Vandermaelen [30].

Locomotor response to novelty and MDMA-induced hyperlocomotion and sensitization

Locomotor activity was measured as previously described [18], in a circular corridor with four infrared beams placed at every 90° (Imetronic, Passac, France). MDMA-induced locomotor sensitization was performed in a two-injection protocol as previously described [17].

Prepulse inhibition of acoustic startle

Sensorimotor gating was assessed as previously described [31].

Forced swimming test

Mice FST was conducted as described [15].

Fluoxetine treatments, proliferation assay

Fluoxetine dose selection and FST were performed as described [15]. Neurogenesis study was performed as described [16].

Synaptosome study

Crude synaptosomes were prepared as previously described [32]. Synaptosomes 5-HT uptake and citalopram competition experiments were performed as described [18].

DOI-induced head-twitch

DOI-induced head-twitch experiment was performed in clear plastic cages lined with bedding. Head-twitches (lateral movements of the head from side to side) were induced by ±DOI (1, 3, and 5 mg/kg, intraperitoneally), and after 10 min, were scored for 10 min, by an experimenter unaware of mice genotype.

Statistical analysis

To determine differences between experimental groups, responses were analyzed by two-tailed unpaired Student's t test or Mann–Whitney test when distribution was not normal. Also, a one- or two-way analysis of variance (ANOVA) repeated measure (RM) when appropriate, with Bonferroni’s test was used for post hoc multiple comparisons. In all cases, P < 0.05 was considered statistically significant.

Results

Ex vivo electrophysiological responses to 5-HT2B-receptor stimulation or overexpression

To examine the ability of 5-HT2B receptors to regulate 5-HT neuron activity, we first performed cell-attached electrophysiological recordings of identified DRN 5-HT neurons in coronal slices of wild-type mice expressing GFP in Pet1-positive neurons (Pet1-GFP mice). In vivo, DRN neurons fire with a regular pattern varying from 0.5 to 5 Hz. 5-HT neurons in ex vivo coronal brain slices are electrically quiescent, but an excitatory noradrenergic tone facilitates firing [33]. In order to reproduce noradrenergic input, subsaturating concentrations of phenylephrine (Phe, 100–300 nM) were added to the bath [25]. Low concentration (100 nM) of Phe was unable to initiate regular firing nor was the application of a preferential 5-HT2B receptor agonist BW723C86 (1 µM). Higher concentration of Phe (300 nM) was able to initiate and maintain a regular firing. Subsequent bath application of BW723C86 (1 µM) induced a significant increase in firing frequency (1.73 ± 0.23-fold) compared to Phe alone (Fig. 1a) (one-way ANOVA RM, effect of treatment, F1.6, 11.0 = 98; P < 0.0001, n = 8 cells). Bonferroni’s post hoc analysis showed a significant increase in firing frequency between each group (Fig. 1a). These data suggested that 5-HT2B-receptor stimulation could increase Pet1-positive raphe 5-HT neuron firing activity.

Fig. 1
figure 1

Ex vivo electrophysiological recordings of wild-type Pet1-GFP mice. a Recordings of 5-HT neuron in ex vivo wild-type slices. Cell-attached recordings made in identified Pet1-GFP neuron did not detect firing in the presence of 100 nM phenylephrine (Phe) or of 1 µM of the 5-HT2B-receptor agonist BW723C86 (top line—5 cells). In the presence of higher concentration of phenylephrine (300 nM), Pet1-positive neuron firing was observed, and BW723C86 (1 µM) was able to increase this firing rate (8 cells). Representative traces (top) and quantification (bottom left) reveal a significant BW723C86-induced increase in firing frequency (one-way repeated measures (RM) ANOVA; bar graph and scatter plots and mean ± SEM; Bonferroni post test, *P < 0.05). b Current-clamp recordings of ex vivo raphe slices from AAVs injected Pet1-GFP mice, overexpressing 5-HT2B-HA receptor in Pet1-positive 5-HT neuron. (Left) Sample traces of action potential for a +200 pA current step in two experimental groups. (Right) Quantification of the number of action potentials obtained in function of injected currents showed a significant increase in action potential number in Pet1-positive 5-HT neurons of mice overexpressing 5-HT2B-HA receptor, compared to controls (n = 20 cells from three control mice and 21 cells from four 5-HT2B-HA mice; two-way ANOVA RM, Bonferroni post test, *P < 0.05; data are mean ± SEM). c Input resistance values. Bar graph and scatter plot for input resistance values showed that input resistance was also increased in mice overexpressing HA-5-HT2B receptor in Pet1-positive 5-HT neuron (unpaired ttest, *P < 0.05; data are mean ± SEM)

Full size image

To further establish that 5-HT2B receptors can affect 5-HT neuron activity, we overexpressed a HA-tagged 5-HT2B receptor specifically in Pet1-positive neurons, using a DIO AAV construct that allows Cre-mediated expression of the tagged receptor. Viruses packaged into AAV2.9 serotype, AAV-DIO-5-HT2B-HA or control AAV-DIO-TdTomato, were unilaterally injected into B7 raphe nucleus of Pet1-GFP mice (Fig. S1A). We confirmed the proper injection site by colocalization of either TdTomato expression or HA immunofluorescence with Pet1-GFP-positive neurons (Fig. S1B). Coronal raphe-containing brain slices were used to record Pet1-GFP-positive neuron excitability performed in current clamp mode. Recordings at various current steps showed that the number of action potentials obtained in function of the injected current was significantly increased in Pet1-positive neurons overexpressing 5-HT2B receptors (two-way ANOVA RM, effect of overexpression, F1, 39 = 4.94, P = 0.032, n = 20 and 21 cells from 3 and 4 mice) (Fig. 1b). Input resistance value from recorded neurons was also significantly increased in 5-HT2B-HA-overexpressing neurons (unpaired ttest, t39 = 2.05, P = 0.047) (Fig. 1c). Ex vivo recordings indicated that the 5-HT2B-receptor overexpression increased Pet1-positive neuron excitability.

In vivo responses to 5-HT1A and 5-HT2B receptor agonists

Since our initial ex vivo results indicated that 5-HT2B receptors could increase 5-HT neuron activity, we hypothesized that it could work in an opposite way as to 5-HT1A-negative autoreceptor. We thus tested the effect of BW723C86 on 5-HT1A receptor agonist 8-OHDPAT-induced inhibition of neuronal firing frequency in vivo. Interestingly, 8-OHDPAT was significantly less potent in suppressing putative 5-HT neuron firing activity after BW723C86 injection as shown by two-way ANOVA RM (main effect of BW723C86 treatment, F1, 9 = 7.34; P = 0.024, n = 6–5 mice). Bonferroni’s post hoc analysis revealed a reduced effect of 8-OHDPAT at 50 and 100 µg/kg on putative 5-HT neuron firing activity (Fig. 2a). 8-OHDPAT ED50 was shifted about 3.3-fold from 45 to 148 µg/kg; BW723C86 alone did not modify firing (not illustrated).

Fig. 2
figure 2

In vivo response of wild-type mice to combined 5-HT2B and 5-HT1A receptor agonists. a In vivo extracellular electrophysiological recordings of putative raphe neurons in anesthetized wild-type mice. (Left) BW723C86 (5 mg/kg subcutaneously (s.c.)) injected 20 min before test counteracted the 5-HT1A agonist 8-OHDPAT (0.05 mg/kg s.c.) inhibitory cumulative effects to putative 5-HT neuron firing (n = 6–5 mice per group; two-way ANOVA RM, followed by Bonferroni’s multiple comparisons test, *P < 0.05; data are mean ± SEM). (Right) Examples of typical recordings of putative DRN 5-HT neurons obtained in each experimental group. Each arrow represents an injection of 8-OHDPAT (0.05 mg/kg s.c.). The injection of the 5-HT1A receptor antagonist WAY100635 (0.3 mg/kg s.c.) completely reversed the inhibitory effect of 8-OHDPAT. b In vivo hypothermic effects. The 5-HT2B-receptor agonist BW723C86 (5 mg/kg s.c.) injected 20 min before the test was able to counteract 5-HT1A agonist 8-OHDPAT (0.3 mg/kg s.c.) hypothermic effects on wild-type mice (scattered plot, n = 4–4 mice two-way ANOVA RM, followed by Bonferroni’s multiple comparisons test, *P < 0.05)

Full size image

Although the mechanism of 5-HT1A receptor agonist-induced hypothermia is incompletely understood [34], the 8-OHDPAT-induced hypothermia in mice is known to be mediated by 5-HT1A autoreceptor [35]. We thus used the 8-OHDPAT-induced hypothermia as another in vivo readout of the functional status of 5-HT1A autoreceptor and tested putative effects of stimulating 5-HT2B receptors. In agreement with the above-mentioned effect on firing, a pretreatment of wild-type mice with BW723C86 (5 mg/kg s.c,) was able to significantly reduce the ability of 8-OHDPAT (0.3 mg/kg s.c.), to induce hypothermia in wild-type mice (Fig. 2b), as shown by two-way ANOVA RM analysis (interaction between the time and treatment, F9, 54 = 10.34; P < 0.0001, n = 4 mice per group). Bonferroni’s post hoc analysis showed a significant reduction by BW723C86 at 30 and 40 min post-injection of the 8-OHDPAT-induced hypothermia (Fig. 2b). These data indicated that 5-HT1A autoreceptor inhibitory activities can be attenuated by concomitant activation of 5-HT2B receptors.

Absence of response to ecstasy, MDMA, in Htr2b 5-HTKO mice

To establish putative direct actions of 5-HT2B receptors on 5-HT neurons, we developed mice with conditional ablation of this receptor gene specifically in Pet1-positive neurons. We inserted recombination sites (loxP) flanking the second exon of Htr2b gene (Htr2blox/lox) and crossed these mice with mice expressing the Cre recombinase under Pet1 gene promoter (BAC transgenic). Pet1 gene expression in the brain is restricted to most of differentiated 5-HT neurons and their postmitotic precursors [23]. We thus generated Pet1-Cre+/0; Htr2blox/lox mice (Htr2b5-HTKO) (Fig. S2A). Restricted recombination in raphe 5-HT neurons was revealed by analysis of raphe DNA and by colocalization of a Cre-dependent GFP reporter with 5-HT staining (Fig. S2B–C) [23]. We quantified the efficiency of recombination and found that 83.6 ± 6.1% of the Pet1-Cre-dependent GFP was colocalized with 5-HT staining, 11.6 ± 3.9% was only 5-HT positive and <5% (4.8 ± 2.9%) was only GFP positive (n = 12 from 3 different mice), supporting an efficient and specific recombination in 5-HT neurons (Fig. S2C).

MDMA, the active compound of ecstasy, is a substrate of SERT leading to massive 5-HT release from synaptic vesicle stores, which is partially calcium-dependent [36]. The Htr2b5-HTKO mice injected with MDMA (20 mg/kg) did not display increased locomotor responses in contrast to control Htr2blox/lox littermate mice (Fig. 3a, b). Two-way ANOVA RM analysis showed a main effect of genotype (F1, 30 = 4.22; P = 0.049, n = 16 mice per group). Bonferroni’s post hoc analysis showed a significant difference between genotypes from t = 25 to t = 45 min post-injection (Fig. 3a). The analysis of the total locomotor activity over the first 60 min after MDMA injection confirmed the significant effect of genotypes (two-way ANOVA RM, F1, 60 = 4.61; P = 0.036, n = 16 mice per group). Bonferroni’s post hoc analysis showed a significant increase in locomotion in control Htr2blox/lox mice, but not in Htr2b5-HTKO mice (Fig. 3b). We further assessed the contribution in Pet1-positive neurons of 5-HT2B receptors to MDMA-induced locomotor sensitization using a two-injection protocol [37]. A significant increase in locomotor activity was observed in Htr2blox/lox control mice as shown by two-way ANOVA RM analysis (main effect of genotype, F1, 14 = 5.12; P = 0.04). Bonferroni’s post hoc analysis showed a significant increase in locomotion in control Htr2blox/lox mice at the second injection of MDMA (20 mg/kg) (first injection 639 ± 222 vs. second injection 1,291 ± 295 1/4 of turns, n = 8), but not in Htr2b5-HTKO (first injection 254 ± 132 vs. second injection 353 ± 174, n = 8) (Fig. 3c). These results indicated a lack of behavioral and sensitizing effects of MDMA in Htr2b5-HTKO mice as found in full KO, Htr2b−/− mice, confirming the need for this receptor in Pet1-positive 5-HT neurons.

Fig. 3
figure 3

Behavioral response to the 5-HT releaser MDMA in Htr2b5-HTKO mice. a MDMA-induced locomotion. Mice were injected with MDMA (20 mg/kg i.p.) (arrow) after 30 min habituation. A lack of MDMA-induced locomotion was observed in Htr2b5-HTKO mice, while control Htr2blox/lox mice showed a clear increase in locomotion. Data between −30 to +60 min were analyzed using two-way ANOVA RM (means ± SEM, n = 16 mice per group) and a Bonferroni post test was applied on each graph, *P < 0.05. b Cumulative MDMA-induced locomotion. Cumulative locomotion during the first hour following MDMA injection showed a significant difference between the two genotypes. Data were analyzed using two-way ANOVA (n = 16–16 mice, scattered plot, mean ± SEM). A Bonferroni post test was also applied on each graph (*P < 0.05 Htr2b5-HTKO vs. Htr2blox/lox; #P < 0.05 MDMA vs. Veh). c Locomotor sensitization by two MDMA injection protocol. The stimulant effect of a challenge dose of MDMA (20 mg/kg i.p.) 7 days after the first (2nd) was significantly enhanced compared to the first injection in control Htr2blox/lox mice, while it had no effect in Htr2b5-HTKO mice. Data were analyzed using two-way ANOVA RM (n = 8–8 mice, scattered plot, mean ± SEM). Bonferroni post test was also applied on each graph (*P < 0.05 Htr2b5-HTKO vs. Htr2blox/lox; #P < 0.05 1st vs. 2nd injection)

Full size image

Absence of response to the SSRI fluoxetine in Htr2b 5-HTKO mice

SSRIs are known to block 5-HT reuptake by SERT, leading to extracellular 5-HT accumulation following vesicular 5-HT release. A classical outcome for acute response to SSRIs is the reduced immobility time observed in FST. We tested the effect of acute fluoxetine injection (3 mg/kg intraperitoneally (i.p.)), the optimal dose determined for 129S2 strain in FST (Diaz et al., 2011). Fluoxetine injection did not affect immobility time in Htr2b5-HTKO mice (Fig. 4a) as shown by two-way ANOVA analysis (time and genotype interactions, F1, 24 = 10.67; P < 0.003, n = 7–7). However, a significant reduction of immobility time was observed in control Htr2blox/lox littermates as Bonferroni’s post hoc analysis showed the only significant difference between vehicle and fluoxetine-treated Htr2blox/lox control mice (Fig. 4a).

Fig. 4
figure 4

Antidepressant action in Htr2b5-HTKO mice. a Forced swimming test (FST). The time spent immobile in the FST was significantly reduced in control Htr2blox/lox mine but not in Htr2b5-HTKO mice 30 min after SSRI antidepressant fluoxetine (Flx 3 mg/kg i.p.) injection (two-way ANOVA, Bonferroni post tests, *P < 0.05; n = 7–7 mice, scattered plots with mean ± SEM). b Neurogenesis in subgranular zone (SGZ) of the hippocampus. Fluoxetine (3 mg/kg/day i.p.), daily injected for 4 weeks, induced a significant increase in BrdU incorporation in neuron of the SGZ of control Htr2blox/lox mice, but had no effect in Htr2b5-HTKO mice (two-way ANOVA, Bonferroni post test; *P < 0.05; n = 7–8 mice, scattered plots with mean ± SEM). c SERT expression and function. (Left) SERT expression in conditional Htr2b5-HTKO and Htr2blox/lox control mice was evaluated using heterologous competition binding assays of [3H]citalopram on synaptosome membranes prepared from whole brain. No differences in the affinity (Ki) or expression (Bmax) between Htr2b5-HTKO and Htr2blox/lox genotypes were observed. (Right) Saturation isotherms for [3H]5-HT uptake of this synaptosomal preparation were similar in conditional Htr2b5-HTKO and Htr2blox/lox control mice. Nonlinear regression analysis did not reveal differences in the Km or Vmax. Shown are representative curves of at least two independent experiments performed in duplicates. Individual values are presented

Full size image

Long-term effects of SSRI are known to be associated with hippocampus subgranular zone (SGZ) neuronal proliferation. We performed a daily i.p. injection for 4 weeks of fluoxetine (3 mg/kg), bromodeoxyuridine (BrdU) injections the 27th day, and quantified BrdU incorporation in SGZ neurons of the hippocampus at the 28th day. Chronic injection of fluoxetine did not produce changes in BrdU incorporation in Htr2b5-HTKO mice (Fig. 4b) (two-way ANOVA analysis showed a trend for treatment and genotype interactions, F1, 26 = 3.88; P = 0.059, n = 7–8). An increase was detected in control Htr2blox/lox littermates as Bonferroni’s post hoc analysis showed the only significant difference between vehicle and fluoxetine-treated Htr2blox/lox controls (Fig. 4b). These results indicated that the lack of both acute behavioral and chronic neurogenic effects of SSRIs found in Htr2b−/− mice was due to 5-HT2B-receptor elimination from Pet1-positive neurons.

We recently reported that Htr2b−/− mice displayed novelty-induced hyperlocomotion and a global deficit in sensorimotor gating [31]. However, locomotor activity in a new environment was not different between Htr2b5-HTKO and Htr2blox/lox littermate control mice over the first 60 min (Fig. S3A). Similarly, prepulse inhibition (PPI) of startle reflex and startle amplitude were not different between Htr2b5-HTKO and Htr2blox/lox littermate control mice (Fig. S3B). Although present in Htr2b−/− mice, these deficits were not found in mice lacking the 5-HT2B receptor selectively in 5-HT neurons, supporting the specificity of these conditional mice.

Since SERT-targeting drug (MDMA and fluoxetine) action is affected by the lack of 5-HT2B receptors in Pet1-positive neurons, we next determined possible alterations of SERT expression or function in brain synaptosome preparations from Htr2b5-HTKO mice. Heterologous competition binding experiments showed no difference in the density of citalopram binding sites (Bmax 305 ± 18 vs. 389 ± 26 fmol/mg of protein) or affinity (Ki 0.4 ± 0.1 vs. 0.17 ± 0.06 nM) between Htr2b5-HTKO mice and their Htr2blox/lox littermate control mice (Fig. 4c). Additionally, 5-HT uptake experiments on brain synaptosomes showed no difference in 5-HT transport maximum velocity (Vmax; 5.57 ± 0.32 vs. 6.49 ± 0.38 fmol/sample/min) or apparent affinity (Km 40.7 ± 2.6 vs. 27.7 ± 2.5 nM) between Htr2b5-HTKO mice and their Htr2blox/lox littermate controls (Fig. 4c). Together, these results demonstrated that selective ablation of 5-HT2B receptors in raphe Pet1-positive neurons eliminates MDMA and fluoxetine actions but does not affect SERT expression and activity.

Htr2b 5-HTKO mice display a hyposerotonergic phenotype

To determine whether the 5-HT2B receptor had an overall effect on 5-HT neurons, we measured firing rates of putative 5-HT DRN neurons in vivo in Htr2b5-HTKO mice. Neurons were included in this analysis based on characteristics and averaged traces of their action potentials [25]. Htr2b5-HTKO mice display a higher percentage of putative 5-HT neurons discharging with a low firing mode (16.1% vs. 9.3% < 1 Hz) and a lower percentage of neurons with a high firing mode (5.9% vs. 13.9% > 4 Hz) relative to Htr2blox/lox mice. This observation was reflected by a significant shift in cumulative distribution of neurons with lower firing rate in Htr2b5−HTKO mice compared to control Htr2blox/lox mice (n = 108 and 118 neurons, respectively; Kolmogorov–Smirnov test; P = 0.0008) (Fig. 5a). These results revealed that ablation of 5-HT2B receptors in raphe Pet1-positive neurons is sufficient to modify 5-HT neuron firing rate.

Fig. 5
figure 5

Hyposerotonergic phenotype of Htr2b5−HTKO mice. a In vivo extracellular recordings of putative DRN 5-HT neurons in anesthetized mice. (Left) The firing frequency of individual putative DRN 5-HT neuron was shifted from high to low firing rates in Htr2b5-HTKO mice. (Right) A shift in cumulative distribution was observed with significantly lower firing rate in Htr2b5−HTKO mice than in control Htr2blox/lox mice (n = 108 and 118 neurons, respectively; Kolmogorov–Smirnov test; *P < 0.05). b Head-twitch dose–response to DOI. Control Htr2blox/lox and Htr2b5−HTKO mice were i.p. injected with ±DOI (1, 3, and 5 mg/kg i.p.). Ten minutes later, the head-twitch response was scored for 10 min. DOI-induced head-twitch response was significantly increased in conditional Htr2b5-HTKO (*P < 0.05; multiple t test) compared to littermate control mice at 5 mg/kg of DOI (data are presented as scattered plot, n = 3–5 mice per group, and means ± SEM). c In vivo hypothermic effects of 8-OHDPAT. The hypothermic response to 8-OHDPAT (0.1 mg/kg s.c.) was significantly stronger in Htr2b5-HTKO mice compared to control Htr2blox/lox mice (n = 3–8 mice, scattered plot, two-way ANOVA RM, followed by Bonferroni’s multiple comparisons test, *P < 0.05)

Full size image

Head-twitch response is a rhythmic paroxysmal rotational head movement that occurs in mice and rats treated by a variety of serotonergic hallucinogens, including LSD and DOI [38]. This behavior is specifically linked to 5-HT2A receptor activation, since selective 5-HT2A receptor antagonists block head-twitch response induced by DOI and other hallucinogens, and it is absent in Htr2a−/− mice [38]. Here, we tested ±DOI at 1, 3, and 5 mg/kg i.p. We found that a dose of 5 mg/kg induced a larger increase in head-twitch in Htr2b5-HTKO mice, as compared to Htr2blox/lox littermate control (+109%, t13 = 2.68, n = 3–5, P = 0.016, multiple t test) (Fig. 5b). Similarly, a greater head-twitch response was observed in Htr2b−/− compared with control mice at the same dose (Fig. S4A). Nevertheless, no significant change in 5-HT2A or 5-HT1A receptor expression or in 5-HT content and turnover was found in PFC from Htr2b−/− compared to control mice (Fig. S4B–D). Since a drastic increase in the number of DOI-induced head-twitch was observed in the Tph2-R439H knock-in mouse [34], a mouse model with 60–80% reduction in TPH2 activity and thus with low serotonergic tone, the increased behavioral response to DOI confirmed a lower serotonergic tone in Htr2b5-HTKO mice.

Finally, we used the 8-OHDPAT-induced hypothermia as an in vivo readout of the functional status of 5-HT1A autoreceptor in Htr2b5-HTKO mice. The hypothermic response to 8-OHDPAT (0.1 mg/kg s.c.) was significantly increased in Htr2b5-HTKO mice compared to control Htr2blox/lox mice as shown by two-way ANOVA RM analysis (interaction between time and treatment, F4, 36 = 5.30; P = 0.0018, n = 3–8 mice). Bonferroni’s post hoc analysis showed a significantly stronger hypothermic effect at 20 min post-injection of the 8-OHDPAT (Fig. 5c). These data supported that 5-HT1A autoreceptor inhibitory activity is increased upon inactivation of 5-HT2B receptors in Pet1-positive neurons.

Discussion

Our results show that 5-HT2B receptors can positively modulate 5-HT neuron activity in brain slices of wild-type mice, and counteract 5-HT1A-negative autoreceptor actions. Furthermore, mice lacking 5-HT2B receptors exclusively in Pet1-positive 5-HT neurons, Htr2b5-HTKO mice, display an excess of inhibitory control exerted by 5-HT1A autoreceptors supporting an opposite control of 5-HT2B receptors on 5-HT neurons.

5-HT2B receptors positively regulate 5-HT neurons

There is a growing consensus that 5-HT neurons are non-homogeneous as supported by anatomical, biochemical, and electrophysiological studies [12, 24, 39, 40]. Sub-populations of 5-HT neurons, either within the DRN or between various raphe nuclei, are interconnected, and form complex circuits [41,42,43]. The activity of 5-HT neurons can be modulated by both 5-HT1A and 5-HT2A/2B/2C receptors [7, 9]. Identified 5-HT neurons are known to respond to 5-HT1A receptor agonists by a 5-HT-induced outward current [7,8,9, 44]. A significant proportion of TPH2-positive neurons (about 50%) also respond to 5-HT2-receptor activation by an inward current [12]. Tonic spiking of 5-HT neurons establishes synaptic 5-HT levels. We show here, by cell-attached recordings of identified wild-type raphe Pet1-positive neurons, that stimulation of 5-HT2B receptors by BW723C86 can increase their firing frequency. The effect on Pet1-positive neuron firing frequency of the 5-HT2B receptor agonist can only be detected in the presence of Phe (300 nM), suggesting that although the two receptors share common Gq signal transduction, they are likely acting at different effectors that remains to be identified. Independent electrophysiological current-clamp recordings showed that overexpression of 5-HT2B receptors in Pet1-positive 5-HT neurons was sufficient to increase their excitability. These results indicate that 5-HT2B receptors can positively control the firing of 5-HT neurons. This was confirmed in vivo by extracellular recordings in Htr2b5-HTKO mice of putative 5-HT neurons that showed a significant shift to low firing rate. Together, the present results revealed a need for 5-HT2B receptors in 5-HT neurons to positively regulate their activity.

Serotonergic tone results from an opposite control exerted by 5-HT1A and 5-HT2B receptors

The lack of effects of MDMA and fluoxetine in the absence of 5-HT2B receptors in Pet1-positive 5-HT neurons, previously observed in full KO mice Htr2b−/− associated to reduced extracellular 5-HT accumulation as assessed by microdialysis [16, 18], raised the possibility of an interaction of 5-HT2B receptors with SERT. However, the absence of modification in SERT uptake and expression lowers this possibility. The unique control of dendritic 5-HT release has important implications for DRN physiology and actions of SERT-targeting drugs, SSRIs and MDMA. Packaging by the vesicular monoamine transporter 2 is essential for 5-HT transmission; glutamate receptor activation in dorsal raphe brain slice can evoke somatodendritic release by vesicle exocytosis [45]. SSRI antidepressants, which have a more limited effect on 5-HT release from dendrites than from the soma and terminals, markedly increase extracellular 5-HT in DRN that involves both somatic and dendritic release [45]. The 5-HT released within DRN induces feedback inhibition of 5-HT neuron firing activity by stimulation of somatodendritic 5-HT1A-negative autoreceptors, which results from local release rather than extended diffusion of 5-HT throughout the extracellular space [46]. We found that the hypothermic response to 8-OHDPAT, known to be mediated by 5-HT1A autoreceptors but not heteroreceptors in mice [35], is attenuated by pretreatment with the 5-HT2B receptor agonist BW723C86. However, we never observed an effect of the 5-HT2B-receptor agonist in the absence of 5-HT1A-receptor agonist. These observations support either a dominant effect of 5-HT1A over 5-HT2B receptors or combination of 5-HT2B autoreceptors and heteroreceptors having opposite actions. Richardson-Jones et al. [35] generated a mouse strain differing in 5-HT1A autoreceptor expression by approximately 30–40% below the wild-type level (1A-Low). These 1A-low mice showed reduced 8-OHDPAT-induced hypothermia and their neurons exhibit a shift toward higher firing rates. To the opposite, in the absence of 5-HT2B receptors in Pet1 neurons (in Htr2b5-HTKO mice), we found a significant increase in hypothermic response to 8-OHDPAT and a significant shift toward lower frequency firing neurons. These findings support that the lack of 5-HT2B receptor in Pet1-positive 5-HT neurons is associated with a higher 5-HT1A-autoreceptor reactivity and thus a lower activity of 5-HT neurons. The lower serotonergic tone observed in the absence of 5-HT2B receptors in Pet1-positive neurons would thus result from the opposite control exerted by 5-HT1A and 5-HT2B receptors on DRN neurons. This may explain the lack of actions of SERT-targeting drugs, SSRIs and MDMA, although the detailed mechanism remains to be identified. An interaction between 5-HT1A and 5-HT2B receptors directly or via trans-regulation could be involved since we previously reported such cross-talks between 5-HT1B and 5-HT2B receptors [47].

5-HT2B receptors contribute to SSRI therapeutic effects

The excess of inhibitory control exerted by 5-HT1A receptors in the absence of 5-HT2B receptors in Pet1-positive 5-HT neurons may also explain the lack of response to chronic SERT blockers (fluoxetine) in Htr2b5-HTKO mice. Chronic SSRI antidepressant responses are at least partially ascribed to desensitization of somatodendritic 5-HT1A receptors [35]. Recent works using chemogenetic approaches (i.e., Designer 
Receptors Exclusively Activated by Designer Drugs (DREADDs)) showed that CNO activation of SERT-positive or Pet1-positive neurons expressing the Gq-coupled M3Gq DREADD induced an increase in 5-HT neuron firing rate and a reduction in immobility in FST [48, 49]. We showed previously that activation of 5-HT2B Gq-coupled receptors with BW723C86 mimicked both acute and chronic behavioral and neurogenic effects of SSRI antidepressants and led to extracellular 5-HT accumulation, which were eliminated in Htr2b−/− mice or by RS127445 [16, 18]. We found here that knocking out the 5-HT2B receptors exclusively from Pet1-positive neurons (Htr2b5-HTKO) mice is sufficient to eliminate behavioral effects in FST and neurogenic effects of fluoxetine and that 5-HT2B-receptor overexpression increases Pet1-positive neuron excitability. It appears thus that 5-HT2B receptors contribute to SSRI therapeutic effects by their positive Gq-dependent signaling on adult raphe 5-HT neurons, which may be revealed upon somatodendritic 5-HT1A-receptor desensitization.

The 5-HT neuron firing relies on a balance of functional 5-HT1A and 5-HT2B receptors

The reason why positive 5-HT2B receptors acting in an opposite manner to negative 5-HT1A autoreceptors has not been previously identified could have several explanations. Recently, the role for 5-HT1A receptor-mediated autoinhibition of the DRN in homeostatic control of firing rate has been questioned. As discussed by Andrade et al. [5], 5-HT1A autoinhibition may participate in regulating glutamate signaling to 5-HT neurons [50] or in mediating inputs from distal serotonergic cell groups [43]. De Kock et al. [51] first showed that following calcium influx through N-methyl-d-aspartate receptor (NMDA) receptors, 5-HT could be released from DRN neuron dendrites in the absence of postsynaptic firing. Colgan et al. [50] reported that 5-HT release from dendrites is secondary to calcium influx through L-type calcium channels that open in response to the local dendritic depolarization elicited by synaptically released glutamate. A contribution of dendritic 5-HT release to 5-HT1A-autoreceptor activation would thus result from excitatory glutamatergic inputs to DRN via locally triggered calcium influx rather than by neuronal firing. Independently, it has been reported that upon electrical stimulation of leech 5-HT neurons, transmembrane Ca2+ entry through L-type channels can first evoke an early dendritic exocytosis; subsequently, the released 5-HT activates 5-HT2 autoreceptors coupled to Gq and phospholipase C, resulting in a positive feedforward loop that maintains sustained exocytosis [11]. In frog motoneurons, a potentiation of NMDA-induced depolarization has been shown to depend on the activation of 5-HT2B receptors causing an influx of extracellular Ca2+ through L-type Ca2+ channels and a reduction of the open-channel block of NMDA receptors [52]. Since serotonergic DRN neurons can respond to 5-HT with responses ranging from inhibition to excitation with the net effect of 5-HT relying on a balance of functional 5-HT1A and 5-HT2A/2B/2C receptors [12], combined expression levels of these receptors in various 5-HT sub-populations may set-up DRN firing levels. Since DRN receives serotonergic inputs from caudal raphe nuclei [43], 5-HT released in DRN may also originate from extrinsic serotonergic afferents. Our results show that 5-HT2B receptors can positively modulate 5-HT neuron activity, and counteract 5-HT1A-negative autoreceptor actions. Furthermore, our recent unpublished data revealed a somatodendritic expression of 5-HT2B receptors as 5-HT1A receptors. Whether 5-HT2B receptors act at different 5-HT neurons subtypes, from different raphe nuclei, or by direct interaction with 5-HT1A receptors, remain to be determined.

Concluding remarks

A remaining question is why both positive and negative autoreceptors are needed to regulate 5-HT neuron activity. In locus coeruleus (LC), the major noradrenergic nucleus of the brain, resting membrane potential, and pattern of spontaneous firing neurons in newborns are affected by both α1-adrenergic and α2-adrenergic receptors activation [53]. Application of α2-adrenergic receptor agonists activates GIRK channels, resulting in membrane hyperpolarization and inhibition of spontaneous firing of action potentials [54]. By contrast, stimulating α1-adrenergic receptors, known to couple to a Gq protein, induces membrane depolarization and accelerates spontaneous firing rates in neonatal rat LC neurons [55], whose alterations may trigger pathological conditions [53]. Mixtures of positive and negative feedback has been modeled and appears to be necessary to create oscillatory signal outputs [56], which are important for neuronal network functions such as DRN or LC. Our findings established that Gq-coupled 5-HT2B receptors expressed by Pet1-positive 5-HT neurons act in an opposite manner as to 5-HT1A autoreceptors. The 5-HT2B receptor can thus be considered as a positive modulator of serotonergic tone that acts at 5-HT neuron excitability. This positive modulation has to be taken into account in the studies of the regulatory mechanisms of 5-HT neurons including those of antidepressants.