Selective refinement of glutamate and GABA synapses on dorsal raphe 5-HT neurons during postnatal life

Serotonin (5-hydroxytryptamine, 5-HT) neurons, which are topographically distributed within different brainstem raphe nuclei, have been implicated in the neural mechanisms contributing to the etiology and therapeutics of mental disorders, including stress vulnerability, anxiety and depression (Stockmeier, 1997; Ressler and Nemeroff, 2000; Southwick et al., 2005; Beyeler et al., 2021). Trauma and stressful experiences during childhood emerge as the main risk factors for the development of psychiatric disorders later in life (Heim and Nemeroff, 2001; Lupien et al., 2009; Insel, 2014; Teicher et al., 2016). Preclinical evidence from rodent models indicates the existence of critical periods of vulnerability during brain postnatal development, when maladaptive neural mechanisms can have detrimental consequences on adult mood and emotional responses (Benekareddy et al., 2010; Rebello et al., 2014; Adjimann et al., 2021). In such periods, 5-HT neurons from the dorsal raphe nucleus (DRN), the main source of forebrain 5-HT, appear to be critically involved in the maladaptive mechanisms. Thus, a reduction of the activity of DRN 5-HT neurons has been shown to rescue some of the long-lasting emotional alterations produced by the early postnatal exposure to the selective 5-HT reuptake inhibitor fluoxetine (Teissier et al., 2015). Other studies have shown that the activation of 5-HT2A/C receptors can modulate the long-term emotional alterations produced by the early postnatal stress of maternal separation (Benekareddy et al., 2011). At later stages, 5-HT1A receptor blockade during adolescence has been shown to reproduce the anxiety phenotypes observed in the 5-HT1A receptor knockout mouse (Lo Iacono and Gross, 2008). Moreover, selective knockdown of forebrain 5-HT1A receptor expression during adolescence is sufficient to produce depressive-like behaviors in adult life (Garcia-Garcia et al., 2017). These findings highlight the pivotal role for the DRN 5-HT system in the neural mechanisms that contribute to the early vulnerability to mood disorders. However, little is known about how excitatory and inhibitory synaptic inputs, and 5-HT-related inhibitory mechanisms, start shaping the early function of DRN 5-HT networks during postnatal life.

DRN 5-HT neurons are subjected to direct excitatory and inhibitory synaptic modulation arising from different cortical and sub-cortical brain regions (Soiza-Reilly and Commons, 2011a; Pollak Dorocic et al., 2014; Weissbourd et al., 2014). Additionally, feedforward inhibition of 5-HT neuron activity is also present in the DRN through local GABAergic neurons (Geddes et al., 2016). Finally, DRN 5-HT neurons are also finely modulated by 5-HT1A receptor-mediated feedback inhibitory loops, including direct autoinhibition through locally released 5-HT, as well as long feedback loops engaging forebrain 5-HT (Sharp et al., 2007; Soiza-Reilly et al., 2015). Evidence indicates that developmental maturation of forebrain-driven 5-HT1A receptor feedback loops occurs during the first postnatal weeks, with a strong impact on adult emotional control (Béïque et al., 2004; Garcia-Garcia et al., 2017).

In the present study, we determined the postnatal trajectories of glutamate excitatory and GABA inhibitory synaptic inputs to DRN 5-HT neurons, studying whether the development of these afferents may be accompanied by changes in 5-HT1A receptor-mediated inhibition over 5-HT neurons. To achieve this, we combined high-resolution quantitative analyses (array tomography) with ex vivo electrophysiology and Fos immunohistochemistry in Swiss mice. We found that both cortical glutamate and subcortical GABA synaptic inputs to DRN 5-HT neurons undergo a profound refinement process after the third postnatal week, whereas subcortical glutamate synapses do not. Evaluation of 5-HT1A receptor-mediated inhibition over 5-HT neurons indicated that this influence is similarly active during the third and fourth postnatal weeks. This study uncovers a precise developmental pattern of synaptic refinement of DRN excitatory and inhibitory inputs, when 5-HT-related inhibitory mechanisms are in place.

To determine the postnatal ontogenetic profile of glutamate excitatory and GABA inhibitory synaptic inputs influencing the early activity of the DRN, we performed a quantitative analysis of synaptic afferents using the high-resolution microscopy technique array tomography (Micheva and Smith, 2007; Soiza-Reilly and Commons, 2011b). This method enables 3D quantitative analysis of multiple neurochemically identified populations of synapses within the same tissue volume. Glutamate axon terminals were labeled for vesicular glutamate transporter types 1 and 2 (VGLUT1 and VGLUT2, respectively) (Fig. 1A-C). VGLUT1 is predominantly expressed in cortical axons, whereas VGLUT2 is almost exclusively present in axon projections arising from subcortical structures (Soiza-Reilly and Commons, 2011a). GABAergic synaptic boutons were labeled for the enzyme glutamate decarboxylase 65 (GAD2) (Fig. 1A,D). Synaptic boutons were defined by co-labeling for synapsin 1, a general marker for synaptic vesicles (Soiza-Reilly et al., 2013) (Fig. 1A-D).

The ontogenetic analysis of global glutamate and GABA synaptic innervations present in the DRN showed that the density of VGLUT1⁺ synapses steadily increases from P4 until P21 (F_5,25=10.05, P<0.0001; P4 versus P7 and P14, P=0.0178 and P=0.0031, respectively), followed by a reduction that became more marked by P60 (F_5,25=10.05, P<0.0001; P21 versus P60, P=0.0308) (Fig. 1E). VGLUT2⁺ synapses presented a highly stable density in the first four postnatal weeks, with a marked increase towards P60 (F_5,25=12.46, P<0.0001; P28 versus P60, P=0,004) (Fig. 1F). The analysis of global DRN GABAergic synapses indicated that the pattern of growth of the first three postnatal weeks (F_5,25=23.67, P<0.0001; P7 versus P14, P=0.0023; P14 versus P21, P=0.0413) is followed by a pronounced reduction by the fourth postnatal week (F_5,25=23.67, P<0.0001; P21 versus P28, P=0.0281) that remains unchanged until adulthood (Fig. 1G). Examination of possible ontogenetic variations in the volume of glutamate and GABA synaptic boutons showed that all studied synapses steadily increase their size throughout postnatal life (Fig. S1), further supporting a process of synaptic pruning in the DRN. Additionally, VGLUT1⁺ and VGLUT2⁺ synapses would have similar ultrastructural features in the adult DRN (Commons et al., 2005).

Next, we examined whether there was a selective refinement of the glutamate and GABA synaptic populations associated with 5-HT neurons. To do this, using array tomography we evaluated associations between VGLUT1⁺, VGLUT2⁺ and GAD2⁺ synapses and 5-HT neurons with immunolabeling against the 5-HT biosynthetic enzyme tryptophan hydroxylase (TPH) (Soiza-Reilly et al., 2013) (Figs 1A, 2A-C). These results showed that VGLUT1⁺ synapses on 5-HT cells steadily increase their abundance between P4 and P21 (F_5,25=8.624, P<0.0001; P4 and P7 versus P21, P<0.001) (Fig. 2A,D). Subsequently, VGLUT1⁺ boutons undergo a marked reduction by the fourth week that continues until adulthood (F_5,25=8.624, P<0.0001; P21 versus P28 and P60, P=0.0396 and P=0.0002, respectively) (Fig. 2A,D). On the contrary, the predominantly subcortical VGLUT2⁺ synapses associated with 5-HT cells present a sustained growth towards P60, without any apparent refinement process (F_5,25=5.826, P<0.0011; P4, P7 and P14 versus P60, P=0.0005, P=0.0201 and P=0.0071, respectively) (Fig. 2B,E). These findings are consistent with a selective refinement of VGLUT1⁺ synaptic inputs innervating 5-HT neurons after the third postnatal week. GABAergic synapses associated with 5-HT neurons showed an early increase of their density until the third postnatal week (F_5,25=29.23, P<0.0001; P7 versus P14 and P21, P<0.0001) followed by marked reductions by P28 and P60 (F_5,25=29.23, P<0.0001; P21 versus P28, P<0.0236; P28 versus P60, P<0.0221) (Fig. 2C,F), in agreement with their observed global reduction (Fig. 1G).

To determine whether our morphological findings could relate to functional changes in glutamate and GABA synaptic inputs to DRN 5-HT neurons, we performed ex vivo patch-clamp recordings of 5-HT neurons identified as previously described (Soiza-Reilly et al., 2013). Glutamate and GABA postsynaptic currents were recorded in the presence of tetrodotoxin (TTX, 1 µM) and strychnine (3 µM) as previously described (Soiza-Reilly et al., 2013) (Fig. 3A). We compared mice of three and four postnatal weeks of age because they represented key transitional stages in our morphological analyses (Figs 1E-G, 2D-F, 3B). Spontaneous miniature excitatory postsynaptic current (mEPSC) recordings in the presence of bicuculline (20 µM) showed a lack of change in the frequency or amplitude of glutamate synaptic inputs between P21 and P28 (frequency: t₁₄=0.693, P=0.500; amplitude: t₁₄=−1.846, P=0.086) (Fig. 3B-D). However, in our electrophysiological recordings, mEPSCs are examined without identifying specific populations of synapses that may arise from different sources (i.e. VGLUT1⁺ versus VGLUT2⁺ synaptic inputs). VGLUT1⁺ synapses preferentially associate with small-caliber and more distal 5-HT dendritic processes and spines, whereas VGLUT2⁺ synapses are often found in larger dendritic shafts closer to 5-HT somas (Commons et al., 2005). In an attempt to functionally identify these two populations of glutamate inputs in our recordings, we assessed the half-width duration of mEPSCs to study their kinetic properties. A double-peak Gaussian function provided the best fit to histogram distributions of mEPSC half-widths (Fig. S2). This indicated the presence of two populations of miniatures with different kinetics that could correspond to VGLUT1⁺ and VGLUT2⁺ synapses (Fig. S2). Importantly, slower mEPSCs were more frequent at P21, whereas at P28, faster mEPSCs became more predominant (Fig. S2). Consistently, glutamate currents displayed shorter decay times at P28 (Mann–Whitney U=11, P=0.028 for P21 versus P28) (Fig. 3E). Thus, kinetic differences detected may be reflecting specific age-related changes in VGLUT1⁺ synapses that are expected to produce slower mEPSCs in comparison with VGLUT2⁺ synapses, which are located closer to 5-HT somas (Commons et al., 2005).

A marked age-dependent effect on GABAergic miniature inhibitory postsynaptic currents (mIPSCs) was evidenced in the presence of 6-cyano-7-nitroquinoxaline-2,3-dione disodium (CNQX, 20 µM). We found a marked reduction in the frequency of GABA currents between P21 and P28 (frequency: Mann–Whitney U=7, P=0.004) with larger amplitudes (amplitude: t₁₇=−2.252, P=0.038) (Fig. 3B-D), although no changes in mIPSC kinetics were detected (t₁₇=−1.500, P=0.152 for P21 versus P28) (Fig. 3E). Interestingly, mIPSCs but not mEPSCs often presented burst-like events that became less frequent with age (mIPSCs: Mann–Whitney U=12.5, P=0.014 for P21 versus P28; mEPSCs: Mann–Whitney U=28, P=0.721 for P21 versus P28) (Fig. 3D, Table S1). Amplitude analyses of intra- and extra-burst synaptic currents suggested that both events would arise from the same release sites (P21: t₁₂=0.494, P=0.630; P28: Mann–Whitney U=11, P=0.841) (Table S1) (Popescu et al., 2010).

Our findings are in agreement with previous reports suggesting the occurrence of maturational changes on the synaptic inputs of 5-HT neurons during postnatal life (Rood et al., 2014; Morton et al., 2015). While DRN 5-HT neurons undergo dendritic remodeling after the third postnatal week (Rood et al., 2014), here we describe, at the same age, a synaptic refinement of cortical glutamate inputs. DRN cortical afferents predominantly arise from the prefrontal cortex (Jankowski and Sesack, 2004; Pollak Dorocic et al., 2014; Weissbourd et al., 2014), and usually associate with small-caliber and more-distal dendritic processes and spines (Commons et al., 2005). Recent evidence indicates that 5-HT-related maladaptive changes upon the synaptogenic potential and connectivity of cortico-raphe synaptic circuit, during a critical postnatal period, contribute to adult vulnerability to psychiatric disorders (Soiza-Reilly et al., 2019; Olusakin et al., 2020). Here, we demonstrate the presence of a selective synaptic refinement of the cortico-raphe circuit after the third postnatal week. However, future experiments involving selective manipulation of this circuit, e.g. by using optogenetics, will give us further mechanistic insight into the synaptic properties of this circuit during the refinement period. Last, postnatal modifications were not observed in glutamate synaptic inputs that preferentially originate from subcortical regions, including the hypothalamus, lateral habenula, laterodorsal tegmental nucleus and parabrachial nucleus (Soiza-Reilly and Commons, 2011a). However, whether VGLUT2⁺ synaptic afferents to DRN neurons could modify their functional properties to counteract the reduction of VGLUT1⁺ inputs remains an unresolved issue.

A marked process of synaptic refinement was also evidenced in GABA synaptic inputs to DRN neurons, typically originating from the hypothalamus, substantia nigra, ventral tegmental area, periaqueductal gray and DRN (Gervasoni et al., 2000; Kirouac et al., 2004). Thus, whether this refinement process could selectively modify synaptic inputs arising from extrinsic versus local GABA sources remains to be determined, and will help us to understand how feedforward inhibition (Geddes et al., 2016) and presynaptic control of glutamate release in the DRN (Soiza-Reilly et al., 2013) are ontogenetically established. Finally, the precise cellular mechanisms involved in the synaptic refinement remain to be elucidated, determining, for example, whether microglial activity (Paolicelli et al., 2011) or specific plasticity events such as long-term depression (Piochon et al., 2016) could be engaged in this process.

Glutamate and GABA synaptic inputs to DRN neurons are finely modulated by 5-HT1A receptor-mediated feedback inhibition involving both prefrontal afferents and local 5-HT (Hajós et al., 1999; Celada et al., 2001). We explored the possibility that glutamate and GABA synaptic refinement could be accompanied by adjustments in the activity of this inhibitory feedback loop. To this purpose, we compared the capacity of the 5-HT1A receptor-mediated inhibition to modulate the activation of DRN 5-HT networks in response to an acute swim stress in the third and fourth postnatal weeks. Previous studies have established the relationship between behavioral responses to this type of stress and the activity of the forebrain 5-HT1A receptor-dependent feedback loop (Soiza-Reilly et al., 2015). Mice received a single injection of the 5-HT1A receptor selective antagonist (WAY-100635, 0.5 mg/kg i.p.) 5 min before their exposure to a swim stress (Fig. 4A). This approach allows the indirect evaluation of the 5-HT1A receptor-mediated inhibition by measuring the activation of pharmacologically disinhibited DRN 5-HT neurons with Fos immunohistochemistry (Fig. 4B,C). Our three-way ANOVA of double-labeled Fos⁺/5-HT⁺ neurons showed that WAY-100635 treatment enhanced the activation of 5-HT neurons in the DRN in response to stress (treatment×age×sex: F_1,48=2.701, P=0.1068; treatment: F_1,48=38.05, P<0.0001) as previously shown (Soiza-Reilly et al., 2015). This effect was similarly present at both P21 and P28 (treatment×age: F_1,48=0.3256, P=0.5709) (Fig. 4D). Examination of possible sex differences revealed that males and females had similar Fos responses to stress within DRN 5-HT neurons (treatment×sex: F_1,48=0.08492, P=0.7720). Finally, no interaction between sex and age factors was observed (sex×age: F_1,48=0.8521, P=0.3606). These findings indicate that 5-HT1A receptor-mediated inhibition is largely active and not modified between P21 and P28. Consistently, previous studies have shown that the hyperpolarizing effects of 5-HT1A receptors on prefrontal glutamate neurons seen in adults are not detected before the third postnatal week (Béïque et al., 2004). Additionally, selective genetic ablation of prefrontal 5-HT1A receptors, starting from the fourth postnatal week, produced long-lasting adult depressive-like symptoms (Garcia-Garcia et al., 2017). However, whether the forebrain 5-HT1A receptor signaling has a direct role in the molecular mechanisms involved in the synaptic refinement remains to be determined. In fact, previous data show that dendritic growth can be modulated by activating 5-HT1A receptors during the third to fifth postnatal weeks (Ferreira et al., 2010).

In conclusion, our study demonstrates that cortical glutamate and subcortical GABA synapses driving the activity of DRN 5-HT neurons undergo a marked refinement process after the third postnatal week. At the same time, this synaptic refinement process is not accompanied by changes in 5-HT1A receptor-mediated inhibition controlling the activity of DRN 5-HT networks. Finally, whether neurodevelopmental alteration of excitatory/inhibitory synaptic drives upon DRN neurons could be part of maladaptive mechanisms increasing the risk of developing mental disorders later in life remains to be elucidated.

Swiss-Webster outbred mice of various ages (ranging from P4 to P60, both sexes) used in the study were acquired from Janvier Labs (France) and from Facultad de Farmacia y Bioquímica, Universidad de Buenos Aires (Argentina), originally obtained from Charles River (USA). Dams were individually housed in separate cages with litters until postnatal day 21 (weaning). After that, mice were sexed and grouped in cages with four or five individuals per cage. All animals were bred and maintained in the animal facility on a 12 h light-dark cycle (7 AM lights on; 7 PM lights off) with ad libitum food and water. All procedures related to care and treatments of mice were approved by the local institutional animal care and use committees (CICUAL 122b, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Argentina; Comité Darwin, Region Ile de France, Inserm UMR-S 1270, France), in compliance with the standard European Union and United States National Institutes of Health ethical guidelines for Care and Use of Laboratory Animals. All efforts were made to minimize any possible distress experienced by mice, as well as to reduce the number of animals required to generate reliable scientific data.

P21 and P28 mice (both sexes) were injected with WAY-100635 (0.5 mg/kg i.p.; Tocris) using a solution of 0.05 mg/ml in saline. Mice were returned to their home cage for 5 min before subjecting them to an acute swim stress that is known to engage DRN 5-HT networks (Roche et al., 2003; Soiza-Reilly et al., 2015). For the swim stress experiments, animals were placed in a cylindrical glass tank (21 cm high×15.5 cm diameter) filled with water (22-24°C) for 10 min. After swimming, mice were removed from the tank, dried with paper towels and returned to their home cage with the littermates. Perfusions with fixative solution started 120 min after the end of the swim session to allow the nuclear accumulation of Fos protein (Soiza-Reilly et al., 2019).

At the specified ages, mice of both sexes were anesthetized with pentobarbital (200mg/kg) (in array tomography experiments) or tribromoethanol (150 mg/kg; Sigma-Aldrich) (in Fos immunohistochemistry experiments), and then killed by transcardiac perfusion through the ascending aorta with 4% paraformaldehyde in phosphate-buffered saline (PBS; pH 7.5). Brains were removed and post-fixed in 4% paraformaldehyde overnight (at 4°C). On the next day, brains were equilibrated in 30% sucrose in PBS for at least 2 days. Brain tissue containing the mid part of the DRN was sectioned at 300 µm in a vibratome (Microm HM650, ThermoFisher Scientific) for array tomography. In Fos immunohistochemistry experiments, brains were sectioned at 40 µm in a cryostat (Roundfin Model 2230) for confocal microscopy analyses.

Coronal cryostat sections (40 µm) containing the mid part of the DRN were collected in a series of three and then used for immunohistochemistry immediately or after storage at −20°C in cryoprotective solution (30% ethylene glycol and 30% sucrose in 0.1 M phosphate buffer, pH 7.4). Antibody incubations were carried out in blocking solution containing 0.04% bovine serum albumin (Sigma-Aldrich), 0.25% Triton X-100 (Merck) and 0.01% sodium azide (Sigma-Aldrich) in PBS (pH 7.5). Primary antisera were applied for 48 h at 4°C, and secondary antibodies for 2 h at room temperature. Antisera anti-5-HT (goat from Abcam, Ab66047, 1/1000) and anti-Fos (rabbit from Abcam, Ab190289, 1/1000) were used. After three washes in PBS at room temperature, primary immunolabeling was revealed using appropriate fluorescent-conjugated secondary antisera made in donkey (Alexa 488, 711-545-152; Alexa 647, 705-605-147; Jackson ImmunoResearch, 1/500) without cross-reactivity. After three washes, sections were mounted onto glass slides using fluorescence mounting medium (S3023, DAKO, USA). Immunolabeled sections were imaged using a confocal microscope (Olympus FV1000). The total number of double-labeled cells (5-HT neurons with Fos) were manually counted by an experienced researcher blind to the pharmacological treatment, age or sex of the animals. Counting was carried out across the entire rostrocaudal axis of the midline DRN (B7 and B6, without considering the lateral wings), using Fiji software (NIH) as previously described (Soiza-Reilly et al., 2015).

Vibratome tissue sections (300 µm) were dehydrated and flat embedded in LR White resin (Medium grade; Electron Microscopy Sciences) for array tomography (Micheva and Smith, 2007; Soiza-Reilly et al., 2013). Briefly, after tissue dehydration in a graded alcohol series (50%, 70% and 70%) and then in a 1:3 mixture of 70% and LR White, for 5 min each, tissue sections were flat embedded between a glass slide and a sheet of ACLAR plastic (Electron Microcopy Sciences; #50425-10) with pure LR White resin, and then curated for 24 h at 55°C in the oven. Tissue was then glued on top of resin blocks for subsequent ultrasectioning at 70 nm with a Histo Jumbo diamond knife (Diatome) and an ultramicrotome (Leica). Serial-ultrathin sections were collected on gelatin coated glass coverslips, dried and heated for 30 min at 55°C on a hot plate to maximize adherence of the sections. Samples were stored at room temperature until use.

Immunofluorescent labeling was performed using the following antibodies previously validated for array tomography (Micheva and Smith, 2007; Soiza-Reilly et al., 2013, 2019): anti-VGLUT1 (guinea pig, 1:1000; Millipore; AB5905), anti-VGLUT2 (guinea pig, 1:1000; Millipore; AB2251), anti-glutamate decarboxylase 65 (GAD2) (rabbit, 1:200; Cell Signaling Technology; 5843), anti-synapsin 1a (rabbit, 1:400; Cell Signaling Technology; 5297) and the antiserum anti-tryptophan hydroxylase (TPH) (sheep, 1:200; Millipore; AB1541) to identify 5-HT neurons. Different combinations of antibodies were applied in a random order and subsequently imaged. After imaging, the antibodies were eluted from the sections with a solution of 0.02% SDS and 0.2M NaOH in distilled water for 20 min. Afterwards, sections were re-immunolabeled with a different combination of antibodies and re-imaged. The TPH labeling was used in each round of immunolabeling as a reference channel for section alignment. Primary antisera prepared in blocking solution (0.05% Tween; 0.1% bovine serum albumin in Tris buffer saline; pH 7.6) were applied for 2 h to sections encircled with a hydrophobic pen (ImmEdge; Vector Labs; H-4000). After three washes with PBS, sections were incubated with fluorescence-conjugated secondary antisera raised in donkey (Alexa 488, 705-545-147; CY3, 711-165-152; Alexa 647, 706-605-148; Jackson ImmunoResearch) for 24 min and rinsed with PBS as before. The coverslips with the sections were mounted onto slides using SlowFade Gold Antifade mountant with DAPI (Thermo Fisher Scientific) and imaged within 3 h. Negative controls without primary antisera were run to corroborate the complete elution of primary antibodies. Immunolabeled sections were imaged on a Leica DM6000 epifluorescence microscope with a 63× NA 1.4 Plan Apochromat oil objective, a CoolSNAP EZ camera and Metamorph software (Molecular Devices). The multiple color channels of serial images were aligned and converted into stacks in Fiji software (NIH) with StackReg and MultiStackReg plug-ins. Two stacks of at least 20 images were analyzed per mouse to obtain a mean value of the number of synaptic boutons per µm³ of tissue. The quantitative analysis was carried out using a sampling mask of at least 90 μm×90 μm. Images were converted to binary images with the threshold function, and, if needed, manually adjusted to avoid non-specific spurious background. Axon boutons were identified by labeled voxels positive for VGLUT1, VGLUT2 or GAD2, and co-labeled with synapsin. This resulted in density values of different axon boutons per tissue volume (puncta per μm³). The dilate function was applied when analyzing synaptic associations to tryptophan hydroxylase-positive profiles. This procedure introduced a mask expansion of one of the objects smaller than 0.2 μm, and helped to avoid possible underdetection of nearby labeled objects, located in close proximity but within different synaptic compartments (Soiza-Reilly et al., 2013). For the statistical analysis, array tomography data obtained from two stacks per mouse were averaged to generate a mean density value per mouse.

Male mice of P18-22 and P27-32 were anesthetized with tribromoethanol (250 mg/kg i.p.; Sigma-Aldrich) and then decapitated. Brain slices (300 µm) containing the DRN were obtained using a vibratome (PELCO EasiSlicer, Ted Pella), submerged in ice-cold low-sodium/high-sucrose solution (composition in mM: 250 sucrose, 2.5 KCl, 3 MgSO4, 0.1 CaCl2, 1.25 NaH₂PO₄, 0.4 ascorbic acid, 3 myo-inositol, 2 pyruvic acid, 25 D-glucose and 25 NaHCO3). Slices were cut sequentially and transferred to an incubation chamber at 35°C for 30 min containing a stimulant-free, low Ca2⁺/high Mg2⁺ normal artificial cerebrospinal fluid (ACSF) (composition in mM: 125 NaCl, 2.5 KCl, 3 MgSO₄, 0.1 CaCl2, 1.25 NaH₂PO₄, 0.4 ascorbic acid, 3 myo-inositol, 2 pyruvic acid, 25 D-glucose and 25 NaHCO₃, and aerated with 95% O₂/5% CO₂, pH 7.4). Whole-cell patch clamp recordings were performed on putative 5-HT neurons (capacitance 30-50 pF) within the midline region of the DRN as previously described (Soiza-Reilly et al., 2013, 2019). Recordings were carried out at room temperature (22-24°C) in normal ACSF with MgCl₂ (1 mM) and CaCl₂ (2 mM). Patch-clamp electrodes were made from borosilicate glass (2-3 MΩ) filled with a voltage-clamp high Cl⁻, high Cs⁺ solution (composition in mM: 110 CsCl, 40 HEPES, 10 TEA-Cl, 12 disodium phosphocreatine, 0.5 EGTA, 2 Mg-ATP, 0.5 disodium-GTP and 1 MgCl₂). pH was adjusted to 7.3 with CsOH. Signals were recorded using a MultiClamp 700 amplifier commanded by pCLAMP 10.0 software (Molecular Devices). Data were filtered at 5 kHz, digitized and stored for off-line analysis.

Spontaneous (non-electrically evoked) miniature IPSCs (mIPSCs) were recorded in the presence of tetrodotoxin (TTX, 1 µM; Alomone), strychnine (3 μM; Sigma-Aldrich) and 6-Cyano-7-nitroquinoxaline-2,3-dione disodium (CNQX, 20 µM; Sigma-Aldrich). Spontaneous miniature EPSCs (mEPSCs) were recorded in the presence of TTX (1 µM), strychnine (3 μM) and bicuculline (20 µM; Sigma-Aldrich). Electrophysiological recordings were carried out at a holding potential of −70 mV. For P21 and P28 mice, mean amplitudes, frequencies and kinetic parameters of miniature postsynaptic currents were calculated. Cumulative probability graphs for both amplitude and inter-event intervals were also calculated and statistically compared using nonlinear regression analysis (curve fit). We measured the following kinetic parameters: rise time (from 10 to 90%), decay time (from 90 to 10%) and half-width (the width at half-maximal peak amplitude).

Spontaneous burst analysis: postsynaptic currents were recorded for 3 min, and spontaneous bursts were identified using the Poisson surprise algorithm with a minimum burst surprise value of 5, which determines: burst length, event number, intra-burst frequency and mean intra-burst interval (Clampfit 10, Molecular Devices). Mean amplitudes of mIPSCs inside and outside the burst were calculated.

Array tomography and Fos immunohistochemistry data were analyzed by one- or three-way ANOVA, respectively, after verifying compliance with assumptions of normality and homogeneity of variances. Post-hoc multiple comparisons were carried out using a two-tailed Tukey's test. Changes in the size of synaptic boutons were analyzed using a Kruskal–Wallis test, followed by Dunn's multiple comparisons. Electrophysiological recordings were analyzed using an unpaired t-test or a Mann–Whitney test, when normality and homoscedasticity were not reached. Histograms of probability distribution for mEPSCs and mIPSCs inter-event intervals and amplitudes were analyzed using nonlinear regression analysis and F-test.

The authors are grateful to Dr Delfina M. Romero for her insightful comments on the manuscript. We also thank the Institut du Fer à Moulin Imaging Platform and the IFIBYNE Confocal Microscopy Facility.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201121.reviewer-comments.pdf