Smoothened receptor signaling regulates the developmental shift of GABA polarity in rat somatosensory cortex

Sonic hedgehog (Shh) is an activity-dependent secreted synaptic molecule (Beug et al., 2011). In the presence of Shh ligand, the receptor patched-1 (Ptch1) relieves its constitutive inhibition of the G-protein-coupled transducer smoothened (Smo), leading to activation of downstream signaling factors (Belgacem and Borodinsky, 2015; Briscoe and Thérond, 2013; Riobo et al., 2006). Shh signaling plays a wide range of functions during embryonic development from neuronal cell proliferation to differentiation (Belgacem et al., 2016; Briscoe and Thérond, 2013; Ruat et al., 2012; Traiffort et al., 2010). Furthermore, in the developing mammalian central nervous system, Shh signaling is involved in axonal elongation (Hammond et al., 2009; Parra and Zou, 2010; Yao et al., 2015) and formation of the cortical connectivity (Harwell et al., 2012; Memi et al., 2018). After birth, Shh signaling components are found in several brain regions, including the cerebral cortex and hippocampus, from an early stage of postnatal development to adulthood (Charytoniuk et al., 2002; Petralia et al., 2011, 2012; Rivell et al., 2019; Traiffort et al., 2010). In the adult hippocampus, the Shh pathway plays an essential role in neurogenesis in the dentate gyrus (Antonelli et al., 2019; Breunig et al., 2008). In addition, Shh receptor Ptch1 and its signal transducer Smo are also expressed at the synaptic junctions of the immature and adult hippocampus (Charytoniuk et al., 2002; Mitchell et al., 2012; Petralia et al., 2011), suggesting roles other than control of neurogenesis (Zuñiga and Stoeckli, 2017). For instance, Shh signaling was reported to exert a modulatory action on neuronal electrical activity in the adult brain (Bezard et al., 2003; Pascual et al., 2005) and, more recently, Shh–Smo signaling has been shown to regulate the formation of glutamatergic and GABAergic terminals in hippocampal neurons (Mitchell et al., 2012). These morphological changes are accompanied by an increase in the frequency of excitatory postsynaptic currents (Feng et al., 2016; Mitchell et al., 2012). Taken together, these findings support the view that the Shh signaling pathway plays a crucial role during early postnatal neuronal circuit construction and synaptic plasticity in the hippocampus and cortex. Thus, its impairment may affect neuronal network formation and lead to brain dysfunction. Likewise, both in human and animal models, accumulating evidence indicates that impairment of the Shh–Smo pathway at postnatal stages may contribute to the emergence of neurodevelopmental disorders, including autism spectrum disorders (ASD) (Al-Ayadhi, 2012; Halepoto et al., 2015) and seizures (Feng et al., 2016; Su et al., 2017).

In the early postnatal life, GABA_A receptor function changes from depolarizing to hyperpolarizing in a KCC2 (also known as SLC12A5)-dependent manner (Ben-Ari et al., 1989; Rivera et al., 1999). Alteration in the GABA developmental sequence leads to a compromised balance between excitatory and inhibitory transmission that is associated with severe changes in the neuronal network circuitry (Ben-Ari, 2002; Ben-Ari et al., 2007; Kilb et al., 2013; Sernagor et al., 2010; Wang and Kriegstein, 2009) and, subsequently, with the onset of neurodevelopmental brain disorders (Ben-Ari and Holmes, 2005; Kuzirian and Paradis, 2011; Mueller et al., 2015).

Although both Shh–Smo and GABA signaling are critically involved in similar processes of neuronal network formation and are implicated in etiology of the same neurodevelopmental disorders, their putative interplay remains unclear. Here we postulated that Shh components may contribute to the onset of the GABA polarity shift and suggested that deregulation of the Shh–Smo signaling pathway might then lead to an impaired neuronal network and to the emergence of brain disorders in adulthood. Because loss of Smo function results in an embryonic lethal phenotype (Zhang et al., 2001) and Smo activation in Wnt1-Cre;R26SmoM2 mice induces a hyperplasia of the facial processes at embryonic stages (Jeong et al., 2004), we used in utero electroporation (IUE) to target the somatosensory cortex with constructs encoding either negative-phenotype (Smo Δ570–581, referred to as Smo-ΔN) (Kim et al., 2009) or constitutively active (Smo A1, referred to as Smo-CA) (Chen et al., 2011) Smo variants. Our results reveal an unexpected function for the Smo signaling pathway in regulation of GABAergic developmental timing in the rat cerebral cortex, notably through the modulation of cell-surface expression and stability of KCC2. We further show that impairment of Smo signaling alters the phosphorylation state of KCC2, thus maintaining a depolarizing action of GABA, which leads to an alteration of GABAergic inhibitory transmission and might contribute to the emergence of brain disorders.

Previous studies have shown that Shh signaling pathway components, including Ptch1 and Smo, are present in the postnatal mammalian brain (Harwell et al., 2012; Mitchell et al., 2012; Petralia et al., 2011; Rivell et al., 2019). In the mouse neocortex, Shh signaling regulates the formation of neuronal networks in immature brain, suggesting a potential function in construction and maturation of neural circuits during the postnatal periods. In order to evaluate the role of Shh signaling in the developing brain, we used IUE of plasmids to express GFP alone, as a control, or in combination with constitutively active (Smo-CA) or negative-phenotype (Smo-ΔN) forms of Smo, an essential component for Shh signaling. IUE of rat brains was performed at embryonic day 15 (E15) in order to target mainly pyramidal neurons of layers V and VI (Kriegstein and Noctor, 2004). Because Shh signaling is involved in cell division and growth of cortical progenitors (Araújo et al., 2014; Radonjić et al., 2016), we investigated whether Smo-related constructs might produce alterations in the localization of electroporated neuronal cells in embryonic day 20 (E20) rat embryos. We found no significant difference between the distribution of cells electroporated with Smo-CA or Smo-ΔN and the distribution of GFP-only control cells (Fig. 1A,B) within the cortical plate (CP) (mean of 92.28%, 90.79% and 93.34%, respectively; P=0.37 for Smo-CA and P=0.26 for Smo-ΔN, Mann–Whitney test), the intermediate zone (IZ) (6.10%, 7.71% and 5.62%, respectively; P=0.74 for Smo-CA and P=0.21 for Smo-ΔN, Mann–Whitney test) and the subventricular/ventricular zones (SVZ/VZ) (1.61%, 1.49% and 1.04%, respectively; P=0.58 for Smo-CA and P=0.49 for Smo-ΔN, Mann–Whitney test). These findings suggest that electroporation of Smo-related constructs has no gross impact at the cellular level on maturation and localization of electroporated cells in developing rat somatosensory cortex.

We next performed immunostaining of electroporated tissues at postnatal day 15 (P15) using the neuronal markers NeuN and FoxP2 (deeper layer neuron markers) to assist in the histological identification of cortical layers (Fig. 1C). The fluorescence intensity profile of both Smo-CA- and Smo-ΔN-expressing neurons did not show any obvious differences in the cortical layering when compared to that of control neurons. Thus, Smo-CA- or Smo-ΔN-expressing neurons electroporated at E15 most likely represent layers V and VI of the somatosensory cortex. The cortical thickness of Smo-CA- and Smo-ΔN-expressing rats measured at P15 was not different from that of control GFP-expressing rats (median of 1772 µm, 1779 µm and 1772 µm, respectively; P=0.07 for Smo-CA and P=0.17 for Smo-ΔN, Mann–Whitney test; Fig. 1C,D). The density of NeuN immuno-positive cells in Smo-CA and Smo-ΔN electroporated somatosensory cortices, measured in layer V, revealed no difference when compared to that in control tissues [46 neurons per observation field for control versus 41 neurons for Smo-CA (P=0.72) and 41 neurons for Smo-ΔN (P=0.44), Mann–Whitney test; Fig. 1E]. Similarly, the average soma neuronal size remained unchanged (210.3 µm² for control versus 229.2 μm² for Smo-CA and 213.8 µm² for Smo-ΔN; P=0.26 and P=0.48 when compared to control, respectively; Mann–Whitney test; Fig. 1F).

Taken together, these observations indicate that expression of Smo-CA or Smo-ΔN does not alter the mean neuronal density or the positioning of pyramidal neurons in the cerebral cortex.

In hippocampal neurons, exogenous treatment with Shh or Smo agonist (SAG) upregulates presynaptic nerve terminals (Mitchell et al., 2012). These observations led us to investigate whether Smo-CA or Smo-ΔN could modulate the density of presynaptic terminals in electroporated cortices. To assess the possible changes at the presynaptic level, we performed quantitative analysis of the area fraction of synaptophysin on MAP2-positive neurons at postnatal day 30 (P30). Compared with control tissues, Smo-ΔN-electroporated tissues showed a significant decrease in the area fraction of synaptophysin immunolabeling (median of 16.05% for Smo-ΔN versus 19.67% for control; P=0.01, Mann–Whitney test; Fig. 2A,B), whereas Smo-CA electroporation increased the area fraction of synaptophysin-positive terminals (24.77% for Smo-CA; P=0.01 when compared to control and P<0.0001 when compared to Smo-ΔN, Mann–Whitney test; Fig. 2B).

Recent evidence indicates that Smo signaling inhibits neuronal apoptosis (Wang et al., 2018), whereas loss of Smo activity leads to increased neuronal apoptosis (Qin et al., 2019). In order to assess whether Smo-CA or Smo-ΔN induces apoptosis in electroporated neurons, we performed immunohistochemical analysis to detect cleaved caspase-3, a marker for cell apoptosis (Logue and Martin, 2008) on electroporated rats at P15. Quantification of cleaved caspase-3-positive neurons showed no significant difference between the constitutively active (mean±s.d. of 0.24±0.13% and 0.34±0.23% in Smo-CA and control, respectively; P=0.70, Mann–Whitney test; Fig. 2C,D) or negative-phenotype (0.23±0.15; P=0.74, Mann–Whitney test; Fig. 2C,D) forms of Smo when compared to control tissues.

These results suggest that IUE of Smo-related constructs does not affect apoptosis mechanisms in the postnatal cerebral cortex. By contrast, Smo-CA and Smo-ΔN significantly change the formation of synaptic terminals. These observations are in accordance with previous findings showing Smo-dependent enhancement of synapse formation in hippocampal neurons (Mitchell et al., 2012).

To investigate the functional expression of Smo mutants, we performed immunofluorescence labeling in somatosensory cortical tissues overexpressing Smo-related constructs at P15 using an antibody raised against Smo protein. As illustrated in Fig. 3A, the anti-Smo antibody detected the expression of Smo-CA and Smo-ΔN in the electroporated somatosensory cortex. Smo activation mobilizes activator forms of Gli transcription factors, which in turn lead to an increase in Ptch1 and Gli1 expression (Jacob and Lum, 2007). To validate functional expression of the Smo-related constructs, we quantified Gli1 and Ptch1 mRNA in cortices overexpressing Smo-CA or Smo-ΔN mutants using reverse transcriptase quantitative PCR (RT-qPCR). As expected, Gli1 mRNA expression was increased in Smo-CA-expressing animals when compared with levels in GFP-expressing control animals [median of 1.39 arbitrary units (a.u.) versus 0.76 a.u. in Smo-CA and control, respectively; P=0.041, Mann–Whitney test; Fig. 3B] and, conversely, was decreased in Smo-ΔN-expressing animals (0.48 a.u.; P=0.039 when compared to control, Mann–Whitney test; Fig. 3B). When compared with levels in control or Smo-CA-expressing rats, Ptch1 mRNA transcripts were downregulated in Smo-ΔN-expressing rats (3.61 a.u. in Smo-ΔN versus 6.97 a.u. in control and 6.21 a.u. in Smo-CA; P=0.026 and P=0.008, respectively; Mann–Whitney test; Fig. 3C). No significant difference was observed between Ptch1 mRNA levels in control and Smo-CA-expressing rats (6.21 a.u. for Smo-CA; P=0.66, Mann–Whitney test; Fig. 3C). We next measured Shh protein expression levels in the somatosensory cortex of GFP-, Smo-CA- and Smo-ΔN-expressing animals at P15 and P30 by ELISA. Our results indicated that Shh protein was expressed at a constant and similar level in all samples analyzed (2.58 ng/ml for control versus 2.42 ng/ml for Smo-CA and 1.93 ng/ml for Smo-ΔN at P15, P>0.05; 2.27 ng/ml for control versus 2.11 ng/ml for Smo-CA and 2.57 ng/ml for Smo-ΔN at P30, P>0.05; Mann–Whitney test; Fig. 3D).

We show here that, in electroporated somatosensory cortices, the constitutively active form of Smo significantly promotes the expression of the target gene Gli1, whereas the negative-phenotype form of Smo acts in the opposite way by reducing Gli1 expression.

Cortical neuronal network construction during the postnatal period requires the depolarizing-to-hyperpolarizing shift of GABA to occur with precise timing during maturation (Ben-Ari and Holmes, 2005; Wang and Kriegstein, 2009; Wu and Sun, 2015). Because Shh signaling plays a nodal role in brain development (Álvarez-Buylla and Ihrie, 2014; Delmotte et al., 2020; Yao et al., 2016), we investigated whether the manipulation of Smo activity affects the onset of the GABA excitatory-to-inhibitory postnatal shift at the network level. We performed field recordings of multiunit activity (MUA) in acute cortical slices from control (GFP), Smo-CA- and Smo-ΔN-expressing rats and measured the effect of bath application of the GABA_A receptor (GABA_AR) agonist isoguvacine (10 µM) on their spiking activity at P14, P20 and P30. In accordance with the known depolarizing action of GABA in the immature neocortex (Kirmse et al., 2015; Riffault et al., 2018), isoguvacine induced an increase of spiking activity at P14 in cortical slices from control and Smo-ΔN-expressing rodents [median of +33.87% (P=0.019) and +86.12% (P=0.019), respectively; Wilcoxon matched-pairs signed test; Fig. 4A,B]. In contrast, isoguvacine decreased the spiking activity in Smo-CA-expressing animals (−12%; P=0.009, Wilcoxon matched-pairs signed test; Fig. 4A,B). At P20, isoguvacine induced either a decrease or an increase of spiking activity in control (−11.37%; P=0.519, Wilcoxon matched-pairs signed test; Fig. 4B). In slices obtained from Smo-CA-expressing rats we observed a decrease in spike frequency (−47.03%; P=0.011, Wilcoxon matched-pairs signed test), whereas in Smo-ΔN-expressing rats, isoguvacine treatment increased the spike frequency (+60.21%; P=0.042, Wilcoxon matched-pairs signed test). At P30 (Fig. 4A–C), isoguvacine induced a decrease in spike frequency in control and Smo-CA-expressing cortical slices [−22.86% (P=0.008) and −42.15% (P=0.015), respectively; Wilcoxon matched-pairs signed test], whereas an increase in spiking activity was observed in Smo-ΔN-expressing rats (+12.89%; P=0.037, Wilcoxon matched-pairs signed test). Thus, the overall effects of isoguvacine treatment on spiking activity suggest that the constitutively active and the negative-phenotype forms of Smo respectively accelerate and delay the developmental hyperpolarizing GABA shift (P<0.0001, Chi-squared test; Fig. 4C). In developing neocortex, the depolarizing-to-hyperpolarizing shift of GABA depends primarily on enhancement of the functional expression of KCC2 (Ben-Ari et al., 1989; Rivera et al., 1999). To investigate whether the hyperpolarizing shift of GABA observed in Smo-CA-expressing rats might involve activation of KCC2 we applied the KCC2-selective blocker VU0463271 (VU) on cortical slices obtained from Smo-CA-expressing rats at P14. VU application shifted the isoguvacine response from inhibitory to excitatory when compared with the response of non-treated Smo-CA-expressing slices (−12% without VU versus +57.1% with VU; P=0.0009, Mann–Whitney test; Fig. 4D).

Collectively, our data suggest that increasing Smo activity prematurely shifts GABA polarity from excitatory to inhibitory in a KCC2-dependent manner, whereas inhibiting Smo signaling delays the switch in GABA polarity.

To further link Smo signaling with the shift in GABA polarity and neuronal chloride homeostasis, we used the gramicidin-perforated patch-clamp technique to measure GABA_A reversal potential (E_GABA) in dissociated hippocampal cultures of immature neurons [9 days in vitro (DIV)]. To ensure that the overexpression of Smo-related constructs did not modify E_GABA by itself, we compared the values of E_GABA for neurons transfected with Smo SA0-5, a constitutively inactive form of Smo (Chen et al., 2011), and control neurons transfected with mCherry, and found that they were identical (median of −56.68 mV for Smo SA0-5 versus −53.16 mV for mCherry; P=0.50, Mann–Whitney test; Fig. 5B). Consistent with our results obtained in somatosensory cortical slices, the measurements of E_GABA in Smo-CA-transfected neurons showed more hyperpolarized values when compared to control mCherry-transfected neurons and Smo-SA0-5-transfected neurons [−61.9 mV in Smo-CA versus −53.16 mV in control (P=0.0004) and −56.68 mV in Smo SA0-5 (P=0.02); Mann–Whitney test; Fig. 5A,B]. Contrary to action of Smo-CA, overexpression of Smo-ΔN did not produce a statistically significant change of E_GABA, as compared to that of control mCherry-transfected neurons (–57.13 mV in Smo-ΔN; P=0.45, Mann–Whitney test; Fig. 5B). To investigate further whether Smo acts through the canonical Gli-dependent pathway, we used the Gli antagonist GANT61, a downstream inhibitor of the canonical Shh–Smo signaling pathway (Lauth et al., 2007). We found that application of GANT61 abolished the hyperpolarized E_GABA values observed in Smo-CA-expressing neurons, restoring it to values similar to those observed in control neurons (−54.92 mV in Smo-CA with GANT61 versus −61.9 mV for Smo-CA and −53.16 mV for the control; P=0.04 and P=0.36, respectively; Mann–Whitney test; Fig. 5A,B).

Next, we used single cell RT-qPCR to quantify Gli1 mRNA levels in Smo-transfected neurons, with or without the application of 10 µM GANT61. In Smo-CA-transfected cells, Gli1 mRNA levels were significantly increased compared to levels in control and Smo SA0-5-transfected cells (0.01 a.u. in Smo-CA versus 0.0029 a.u. in control and 0.0036 a.u. in Smo SA0-5; P=0.008 and P=0.008, respectively; Mann–Whitney test; Fig. 5C). Conversely, Gli1 mRNA levels were downregulated in Smo-ΔN-transfected neurons (0.0004 a.u. in Smo-ΔN versus 0.0029 a.u. in control and 0.0036 a.u. in Smo SA0-5; P=0.015 and P=0.016, respectively; Mann–Whitney test; Fig. 5C). Moreover, the effect of Smo-CA on the target gene Gli1 was blocked in the presence of GANT61 (0.0024 a.u. in Smo CA with GANT61; P=0.016 when compared to Smo-CA alone; P=0.41 and P>0.99 when compared with control and Smo SA0-5, respectively; Mann–Whitney test; Fig. 5C). These results, therefore, suggest that the constitutively active form of Smo prematurely shifts GABA polarity through the activation of the canonical Gli-dependent transcription pathway.

To corroborate the results obtained in neuronal cultures, we studied whether Smo signaling modifies intracellular chloride concentration ([Cl⁻]_i) in acute slices of somatosensory cortex isolated from Smo-CA- and Smo-ΔN-expressing rats. For this purpose, mCherry (control), Smo-CA or Smo-ΔN were co-electroporated in utero at E15 with constructs encoding a ratiometric chloride sensor (Cl-Sensor; Friedel et al., 2013, 2015) (Fig. 5D). Electroporated slices prepared at P10 and P30 harbored hundreds of neurons expressing Cl-Sensor in cortical layers V and VI. Bath application of isoguvacine (10 µM, 3 min) to P10 Smo-ΔN-expressing and control slices produced a uniform decrease of the fluorescence intensity ratio R_430/500 (indicated by an arrowhead in Fig. 5E and by the negative value of ΔR/R in Fig. 5G), reflecting a chloride ion extrusion characteristic of depolarizing action of GABA (median of −0.04 a.u. for Smo-ΔN versus −0.09 a.u. for control; P=0.17, Mann–Whitney test; Fig. 5E,G), whereas in Smo-CA-expressing slices, isoguvacine triggered either an increase, no change or slight decrease of R_430/500. The median of these changes in Smo-CA slices was 0.00 a.u., and statistical analysis revealed significant difference when compared to control (P=0.005, Mann–Whitney test; Fig. 5E,G). This difference indicates that in Smo-CA-expressing slices GABA is switching from depolarizing to hyperpolarizing actions. In control and Smo-CA-expressing slices at P30, isoguvacine produced a chloride influx (indicated by an arrow in Fig. 5F) as shown by the positive value of ΔR/R (+0.22 a.u. for control versus +0.26 a.u. for Smo-CA; P=0.63, Mann–Whitney test; Fig. 5F,G), whereas Smo-ΔN-expressing slices showed a slight increase of R_430/500 following isoguvacine treatment (+0.06 a.u.; P=0.002 and P=0.01 when compared to control and Smo-CA, respectively; Fig. 5F,G). The latter finding suggests that a Smo-dependent pathway is active at rest in mature P30 neurons to maintain the low [Cl⁻]_i and hyperpolarizing inhibitory action of GABA. The inhibition of Smo action using Smo-ΔN led to an increase in [Cl⁻]_i that reduces the inhibitory strength of GABA.

Thus, in both immature acute slices and primary cultures, activation of the Smo signaling pathway using Smo-CA led to a hyperpolarizing shift of Cl⁻, whereas Smo-ΔN did not affect the [Cl⁻]_i at this stage.

Regulation of the KCC2-dependent developmental shift of GABA relies on a complex mechanism involving a progressive increase in the amount of KCC2 protein and its posttranslational modifications via at least two distinct phosphorylation sites, serine 940 (Ser⁹⁴⁰) and threonine 1007 (Thr¹⁰⁰⁷) (Côme et al., 2019; Medina et al., 2014; Moore et al., 2017). We measured the protein and mRNA expression of KCC2 from the somatosensory cortices of Smo-CA- and Smo-ΔN-expressing rats at P10. We observed similar levels in control rats and rats expressing the Smo-related constructs for KCC2 mRNA (median of 5.02 a.u. for control tissues versus 3.3 a.u. for Smo-CA and 2.92 a.u. for Smo-ΔN; P=0.84 and P=0.42, respectively, compared to control; Mann–Whitney test; Fig. 6A) and protein (1.23 a.u. for control tissues when normalized to β3 tubulin versus 1.26 a.u. for Smo-CA and 1.25 a.u. for Smo-ΔN; P=0.48 and P=0.48, respectively; Mann–Whitney test; Fig. 6B,C). The membrane stability and transporter activity of KCC2 are dependent on the phosphorylation state of intracellular C-terminal domains. For instance, phosphorylation of the Ser⁹⁴⁰ residue increases stability and functionality of KCC2, whereas phosphorylation on the Thr¹⁰⁰⁷ residue enhances KCC2 endocytosis (Lee et al., 2007; Medina et al., 2014). To determine which steps in KCC2 trafficking may be regulated by Smo-related constructs, we carried out a quantitative immunoblotting assay using antibodies recognizing the Ser⁹⁴⁰ (Lee et al., 2011) and Thr¹⁰⁰⁷ (de Los Heros et al., 2014) phosphorylated forms of KCC2. We found that the ratio of phospho-Ser⁹⁴⁰ (pKCC2 S⁹⁴⁰) to total KCC2 protein (pKCC2 S⁹⁴⁰/KCC2) was upregulated in Smo-CA-electroporated cortices when compared to the ratio in control and Smo-ΔN conditions (0.56 a.u. for Smo-CA versus 0.39 a.u. for control and 0.27 a.u. for Smo-ΔN; P=0.02 and P=0.02, respectively, when compared to Smo-CA; Mann–Whitney test; Fig. 6B,D). The ratio of phospho-Thr¹⁰⁰⁷ (pKCC2 Thr¹⁰⁰⁷) to total KCC2 protein (pKCC2 T¹⁰⁰⁷/KCC2) was not modified in cortices electroporated with Smo-related constructs (0.90 a.u. for control tissues versus 0.89 a.u. for Smo-CA and 0.78 a.u. for Smo-ΔN; P=0.88 and P=0.17, respectively, when compared to control; Mann–Whitney test; Fig. 6B,E).

Taken together, these data suggest that enhanced Smo activity led to KCC2 posttranslational modifications, with increased Ser⁹⁴⁰ phosphorylation in Smo-CA-electroporated somatosensory cortex.

The phosphorylation of Ser⁹⁴⁰ described above is well known for its ability to stabilize KCC2 in the plasma membrane, whereas dephosphorylation of Ser⁹⁴⁰ promotes KCC2 endocytosis (Kahle et al., 2013; Lee et al., 2007). We therefore assessed whether Smo-related constructs are able to regulate KCC2 trafficking to the cell surface. As a tool, we used a KCC2 construct tagged in an external loop with the fluorescent protein pHluorin (KCC2-pH_ext) (Kahle et al., 2014). This construct allows both measurement of the KCC2-pH_ext-dependent shift of [Cl⁻]_i and visualization of surface expression and/or internalization of KCC2-pH_ext. (Friedel et al., 2015, 2017).

As expected, overexpression of KCC2-pH_ext in neurons at 9 DIV resulted in a negative shift of E_GABA (median of −53.16 mV in control neurons and −86.25 mV in neurons overexpressing KCC2-pH_ext; P<0.001, Mann–Whitney test; Fig. 7A,B), indicating also that at least some portion of KCC2-pH_ext molecules were delivered to the plasma membrane. The overexpression of Smo-CA did not produce additional hyperpolarizing shift of E_GABA in KCC2-pH_ext-positive neurons (-83.82 mV for Smo-CA and KCC2-pH_ext versus −86.25 mV for mCherry and KCC2-pH_ext; P=0.69, Mann–Whitney test; Fig. 7A,B), whereas overexpression of Smo-ΔN resulted in a significant depolarizing shift of this parameter (−53.1 mV in Smo-ΔN and KCC2-pH_ext versus −86.25 mV in mCherry and KCC2-pH_ext; P=0.006, Mann–Whitney test; Fig. 7A,B) indicating either a Smo-ΔN-dependent inactivation of the membrane pools of KCC2-pH_ext or reduction of KCC2-pH_ext expression at the plasma membrane.

To corroborate these observations, we performed a live-staining analysis of surface-expressed and internalized KCC2-pH_ext. Three different pools of KCC2-pH_ext were revealed using a multistep immunolabeling protocol: the total expressed amount of protein (F_t), the amount of KCC2-pH_ext cell surface expression (F_m) and the amount of internalized KCC2-pH_ext (F_i) (Fig. 7C). As positive and negative controls we used overexpression of KCC2-pH_ext mutants T906A/T1007A (A/A-KCC2-pH_ext) and ΔNTD-KCC2-pH_ext, which are known for their increased and perturbed surface expression abilities, respectively (Friedel et al., 2017). Live-cell immunolabeling of the control non-mutated KCC2-pH_ext revealed a well detectable F_m signal in the form of clusters, whereas the post-hoc multistep immunolabelings revealed the F_i pool of molecules and the total amount of expressed KCC2-pH_ext (Fig. 7C). The five tested constructs showed similar intensities of F_t (median of 0.95 a.u. for KCC2-pH_ext, 1.25 a.u. for A/A-KCC2-pH_ext, 0.85 a.u. for ΔNTD-KCC2-pH_ext, 1.062 a.u. for Smo-CA with KCC2-pH_ext, and 1.067 a.u. for Smo-ΔN with KCC2-pH_ext; Fig. 7D). The relative amounts of F_m and the membrane-labeled cluster size in KCC2-pH_ext-expressing neurons were lower than those in A/A-KCC2-pH_ext-expressing neurons and significantly higher than the values revealed in ΔNTD-KCC2-pH_ext-transfected cells (for F_m: 0.94 a.u. for KCC2-pH_ext versus 1.49 a.u. for A/A-KCC2-pH_ext and 0.008 a.u. for ΔNTD-KCC2-pH_ext; P=0.002 and P<0.0001, respectively; Fig. 7E; for single cluster membrane size: 1.02 a.u. for KCC2-pH_ext versus 1.09 a.u. for A/A-KCC2-pH_ext and 0.55 a.u. for ΔNTD-KCC2-pH_ext; P<0.0001 and P<0.0001, respectively; Mann–Whitney test; Fig. 7G). Co-expression of Smo-CA with KCC2-pH_ext did not produce a statistically significant change in F_i, F_m or single surface-labeled cluster size (Fig. 7E–G). By contrast, neurons co-expressing KCC2-pH_ext and Smo-ΔN showed a significant twofold higher internalization rate of labeled KCC2-pH_ext molecules (0.84 a.u. for KCC2-pH_ext versus 2.51 a.u. for Smo-ΔN with KCC2-pH_ext; P<0.0001; Mann–Whitney test; Fig. 7C,F), a significant twofold lower amount of F_m (0.94 a.u. for KCC2-pH_ext versus 0.57 a.u. for Smo-ΔN with KCC2-pH_ext; P=0.009; Mann–Whitney test; Fig. 7C,E) and significantly smaller size of surface-located KCC2-pH_ext clusters (1.02 a.u. for KCC2-pH_ext versus 0.92 a.u. for Smo-ΔN with KCC2-pH_ext; P<0.0001, Mann–Whitney test; Fig. 7G).

Thus, consistent with data illustrated in Figs 4–7, both [Cl⁻]_i measurement and live-cell immunolabeling revealed a striking Smo-dependent change in function and surface expression of KCC2-pH_ext.

Large number of studies have illustrated that KCC2 dysfunction facilitates initiation of epileptic seizures (Chen et al., 2017; Moore et al., 2018). Changes in KCC2 Ser⁹⁴⁰ phosphorylation that modify neuronal chloride homeostasis and depolarizing strength of GABA strongly affect seizure susceptibility in KCC2 transgenic mice (Silayeva et al., 2015). Finally, we investigated whether overexpression of Smo mutants may influence the susceptibility to seizures. Rats expressing Smo-ΔN or Smo-CA and age-matched control rats (transfected with GFP alone) at P30 were intraperitoneally injected with subconvulsive doses of pentylenetetrazol (PTZ; 25 mg/kg; Fig. 8A), an inhibitor of the GABA_A receptors (Klioueva et al., 2001). Previous studies have shown that the lowest threshold susceptibility to PTZ can be observed during the first 2 postnatal weeks, and the highest at around P30 (Klioueva et al., 2001; Vernadakis and Woodbury, 1969). These results suggest that the vulnerability to PTZ is correlated to the maturation of the GABAergic system during postnatal development. Injections of PTZ were continued every 10 min until each rat had generated a generalized tonic–clonic seizure. Seizure susceptibility and severity were quantified using the canonical Racine's scale of seizure severity (Lüttjohann et al., 2009), including dose and latency to onset of the first generalized tonic–clonic seizure. Generalized seizures in Smo-ΔN-expressing rats occurred at lower doses of PTZ (median of 87.5 mg/kg compared with 100 mg/kg for both control and Smo-CA-expressing rats; P=0.005 and P=0.003, respectively; Mann–Whitney test; Fig. 8C) and with shorter latencies than in control and Smo-CA-expressing rats (30.83 min compared with 32.71 min in control and 32.69 min in Smo-CA; P=0.041 and P=0.024, respectively; Mann–Whitney test; Fig. 8B). No significant difference was observed between control and Smo-CA-expressing rats in the latency (32.69 min compared with 32.71 min in control; P=0.5158; Fig. 8B) or cumulative dose of PTZ to induce tonic–clonic seizures (P=0.501; Fig. 8C).

These results indicate that the reduced seizure threshold in Smo-ΔN-expressing rats could be due to a downregulation of KCC2 and deregulation of chloride homeostasis induced by the reduction of Smo activity.

In this study, we investigated the actions of the Smo signaling pathway on GABAergic transmission in vivo and ex vivo in the rat somatosensory cortex and in vitro in primary cultures of hippocampal neurons using negative-phenotype (Smo-ΔN) and constitutively active (Smo-CA) forms of Smo. We show that activation of Smo signaling in immature neurons at early postnatal stages accelerated the GABAergic switch from depolarizing to hyperpolarizing activity in a KCC2-dependent manner, leading to a low intracellular chloride concentration. Conversely, Smo signaling blockade altered the phosphorylation state and membrane stability of KCC2, weakening the GABAergic inhibition in later stages of development and resulting in an increase of seizure susceptibility. These findings reveal an unexpected function of the G-protein-coupled receptor smoothened on chloride homeostasis regulation and, in particular, on the timing of the GABA shift from depolarizing to hyperpolarizing in the postnatal developing brain.

The postnatal developmental switch in GABA polarity is tightly regulated by the actions of several peptides, including neurotrophic factors (Aguado et al., 2003; Kelsch et al., 2001; Ludwig et al., 2011; Riffault et al., 2018), hypothalamic neurohormones (Leonzino et al., 2016; Spoljaric et al., 2017; Tyzio et al., 2006), central and peripheral hormones (Dumon et al., 2018; Sawano et al., 2013). Hence, the precise timing of the GABAergic polarity shift depends on the balance between factors promoting or delaying the onset of GABAergic inhibitory transmission. It is noteworthy that these factors are developmentally regulated and/or operate within specific time windows to control the functional expression of KCC2 and associated chloride homeostasis and, consequently, the postnatal maturation of GABAergic transmission in the central nervous system. Interestingly, the excitatory action of GABA at early stages of development coincide with the predominance of trophic factors responsible for the inhibition of KCC2 expression and function, such as the immature form of BDNF (proBDNF) (Riffault et al., 2018) and the adipocyte hormone leptin (Dumon et al., 2018). These two factors are expressed shortly after birth and surge during the first postnatal week in rodents (Dumon et al., 2018; Menshanov et al., 2015). They have been shown to promote low KCC2 protein and mRNA expression levels and, consequently, a high intracellular chloride concentration, thus maintaining a depolarizing action of GABA in neonates. The developmental decreases of leptin and proBDNF observed at the end of the first postnatal week in rodents are concomitant with an increase in oxytocin (Leonzino et al., 2016; Tyzio et al., 2006), mature BDNF (mBDNF) (Maisonpierre et al., 1990) and thyroid hormones (Sawano et al., 2013), which in turn upregulate KCC2, leading to a low intracellular chloride concentration and a hyperpolarizing action of GABA. Although we did not observe any difference in the level of Shh proteins between P15 and P30, a previous study demonstrated that Shh expression paralleled the temporal expression of mBDNF in the cortex of rodents, with a nearly undetectable level at birth and a continuous increased expression at both mRNA and protein levels during the two first postnatal weeks (Rivell et al., 2019). This temporal expression of Shh and its signaling partners during postnatal brain development has been shown to play a role in neuronal network construction through axonal growth and synaptic maturation (Yao et al., 2015). Here, we uncover a new critical function of the Shh transducer Smo in regulation of the GABA polarity switch. Indeed, we showed that activation of the GABA_A receptor by isoguvacine in immature (P14) Smo-CA-expressing cortical neurons resulted in an influx of chloride and a decrease in spiking activity, whereas mature (P30) negative-phenotype Smo cortical neurons (Smo-ΔN), maintained a depolarizing and excitatory action of GABA. Thus, in the absence of Smo signaling, the GABA shift is delayed, thereby leading to an altered chloride homeostasis and higher susceptibility to seizures, presumably reflecting an unbalanced excitation to inhibition (E/I) ratio. E/I imbalance has been linked to cognitive disorders such as schizophrenia, ASD, depression and epilepsy (Ben-Ari and Holmes, 2005; Brambilla et al., 2003; Kuzirian and Paradis, 2011; Mueller et al., 2015). Abnormal levels of Shh have also been observed in ASD patients, with increased serum level observed in children with ASD, and decreased mRNA level measured in post-mortem adult human brain tissue from autism patients (Choi et al., 2014; Halepoto et al., 2015). In addition, mutations in PTCHD1, a gene encoding a protein that displays secondary structures similar to Ptch1, are found in patients with intellectual disability and ASD (Ung et al., 2018). Consistent with a possible role of Shh components in the regulation of the E/I balance, these observations also suggest that an alteration in the Shh–Ptch–Smo pathway might have an effect on the pathophysiological mechanisms involved in ASD. In line with these observations, the role of Shh signaling in neuronal activity has been recently addressed in an elegant study (Hill et al., 2019) showing that the selective disruption of Shh signaling using a conditional knockout of Smo leads to an increase in neuronal excitability of cortical neurons, thus highlighting the importance of Shh–Smo signaling for the regulation of the E/I equilibrium and the construction of functional cortical networks. The results from this study are consistent with our findings showing that Smo signaling blockade increased neuronal network excitability in rodents expressing the negative-phenotype form of Smo. Furthermore, we found that blockade of Smo activity in Smo-ΔN-expressing rodents led to a decrease in KCC2 function and to a reduced threshold for PTZ-induced tonic–clonic seizure. In agreement with a previous study by Chen et al. (2017), these results suggest that downregulation of KCC2 function may contribute to epileptic seizures. The implication of Shh–Smo signaling in epilepsy and seizures has been indirectly addressed in previous studies showing that somatic mutations in Shh pathway genes in humans are associated with hypothalamic hamartoma and drug-resistant epilepsy (Hildebrand et al., 2016; Saitsu et al., 2016). Likewise, Shh expression is increased in hippocampus and neocortex from both human and animal models of temporal lobe epilepsy (TLE) (Fang et al., 2011), thus suggesting a potential relationship between epilepsy and Shh activity. In line with the above studies, Feng et al. (2016) reported that during TLE development, Shh contributes to epileptogenesis through the inhibition of glutamate transporter function leading to abnormal extracellular glutamate levels, whereas the blockade of Shh–Smo signaling reduced the severity of seizures. Taken together, these observations can lead to opposing interpretations regarding the possible interplay between Shh–Smo pathway and epilepsy. These apparent discrepancies could be explained by the treatments utilized (i.e. acute seizure models with PTZ versus spontaneous seizures in the pilocarpine model of chronic TLE), which are able to activate different downstream signaling cascades (Choudhry et al., 2014). Moreover, other studies have found that activation of the Shh–Smo signaling pathway increases the synthesis and secretion of neurotrophic factors like nerve growth factor (NGF) and BDNF (Bond et al., 2013; Chen et al., 2018; Delmotte et al., 2020; Radzikinas et al., 2011), which may have neuroprotective effects in TLE (Bovolenta et al., 2010; Falcicchia et al., 2018; Paradiso et al., 2009). Clearly, additional studies are needed to obtain a deeper knowledge of the cellular and molecular mechanisms linking epilepsy to Shh pathway components.

Importantly, our in vitro experiments revealed that in immature neurons of both primary rat hippocampal cultures and rat neocortical acute slices the E_GABA shifted towards a more hyperpolarized level in Smo-CA-expressing neurons, indicating a Smo-CA-initiated decrease of neuronal [Cl⁻]_i. This effect of Smo was blocked by preincubation with the Gli transcription factor inhibitor GANT61, suggesting that in the current study, Smo acts through Gli-dependent canonical signaling (Belgacem et al., 2016) to modulate the reversal potential of GABA. Our previous research demonstrated that Smo also participates in the maturation of GABAergic networks in the postnatal rat hippocampus through a non-canonical pathway (Delmotte et al., 2020). Non-canonical Shh signaling has been shown to acutely modulate Ca²⁺-mediated spontaneous electrical activity in the embryonic spinal cord through a G_αi protein (Belgacem and Borodinsky, 2011). In this context, the G_αi-dependent non-canonical signaling acts as a negative regulator of canonical Shh signaling. The activation of Smo leads to an increase of Ca²⁺ spike frequency, leading to protein kinase A activation and subsequent inhibition of the activator forms of Gli transcription factors (Gli1 and Gli2^A) concomitant with an enhancement of the repressor forms of Gli (Gli3^R) in embryonic spinal neurons (Belgacem and Borodinsky, 2015). Here, we show an enhancement of Gli activity upon Smo activation, indicating that the non-canonical Smo signaling pathway is not involved (Belgacem and Borodinsky, 2011, 2015). This highlights the diversity of signaling pathways recruited by the Shh signal transducer Smo among cell types, central nervous structures and developmental stages.

In immature neurons the increased [Cl⁻]_i and corresponding depolarizing values of E_GABA are determined primarily by effective intrusion of Cl⁻ by sodium-potassium-chloride cotransporter type 1 (NKCC1, also known as SLC12A2) and absence of effective Cl⁻ extrusion by KCC2. The blockage of NKCC1 (Yamada et al., 2004) or activation of KCC2 (Dumon et al., 2018; Inoue et al., 2012; Khirug et al., 2005) both lead to a hyperpolarizing shift of E_GABA. In this study, we revealed a novel Smo-dependent pathway of KCC2 upregulation during neuronal development. Whether the Smo pathway also regulates NKCC1 activity, as well as other transporters and channels controlling Cl⁻ homeostasis, in immature neurons should be a subject of future studies.

The Cl⁻ extrusion activity of KCC2 in developing neurons depends on protein expression level (Rivera et al., 1999) and posttranslational modifications that regulate the surface targeting and clustering of KCC2 (Côme et al., 2019; Medina et al., 2014). Thus, KCC2 cycles between synaptic sites and endocytic compartments, depending on the phosphorylation state of different amino acid residues (Côme et al., 2019; Kahle and Delpire, 2016; Kahle et al., 2014). However, this membrane turnover also requires regulation by the cytosolic N- and C-termini of KCC2 (Friedel et al., 2017). For instance, N-terminal truncation of KCC2 prevents its export to the cell surface, whereas C-terminal truncation leads to a decreased surface expression of KCC2 (Friedel et al., 2017). The observed loss of KCC2 surface expression in the C-terminal mutants is explained by an increase in the internalization rate without affecting cell-surface delivery. Furthermore, KCC2 function can also be regulated through lateral diffusion at the plasma membrane (Chamma et al., 2013). This suggests that KCC2 membrane dispersion may, in turn, mediate membrane destabilization and endocytosis, resulting in dysregulation of intracellular chloride homeostasis (Côme et al., 2019). In agreement with this, we found that inhibition of the Smo signaling pathway in mature neurons induces a decrease in KCC2 cluster size and enhances the intracellular concentration of chloride, which sets E_GABA at more positive values than the resting potential and reduces cell-surface expression levels of KCC2. We further show that these effects occur through an increased rate of KCC2 turnover.

In conclusion, these data uncover an unexpected role for the Shh transducer Smo during early postnatal development through the control of chloride homeostasis. Thus, Smo signaling can tune GABA inhibitory transmission by controlling phosphorylation of KCC2 that affects KCC2 stability in the cell membrane and modulates neuronal chloride homeostasis. Finally, these results suggest that Smo signaling is able to set the cursor of the GABAergic inhibitory switch during the critical maturational period where developmental pathogenesis takes place.

Animal procedures

All animal procedures were carried out according to guidelines set by the INSERM animal welfare committee through the local committee (CEEA n°14) and the European Communities Council Directives (2010/63/UE). Male and female Wistar rats were purchased from Janvier Labs (https://www.janvier-labs.com/fiche_produit/rat_wistar/). Animals were raised and mated at INMED A2 animal facility and were housed under a 12 h light/dark cycle at 22–24°C with access to food and water ad libitum.

The following reagents were purchased from the indicated sources: 1,2,3,4-Tetrahydro-6-nitro-2,3-dioxo-benzo[f]quinoxaline-7-sulfonamide (NBQX) was obtained from the Molecular, Cellular and Genomic Neuroscience Research Branch (MCGNRB) of the National Institute of Mental Health (NIMH, Bethesda, MD, USA). Tetrodotoxin (TTX) was purchased from Abcam (Cambridge, UK). Isoguvacine and VU0463271 were purchased from Tocris Cookson (Bristol, UK). Strychnine, GANT61 and bumetanide were from Sigma (St Louis, Missouri, USA).

In utero injections and electroporations were performed in embryos from timed pregnant rats (embryonic day 15) that received buprenorphine (Buprecare at 0.03 mg/kg) and were anaesthetized with sevoflurane (4.5%) 30 min later. Briefly, the uterine horns were exposed, and a lateral ventricle of each embryo was injected using pulled glass capillaries and a microinjector (PV 820 Pneumatic PicoPump; World Precision Instruments, Sarasota, FL) with Fast Green (2 mg/ml; Sigma, St Louis, MO, USA) combined with the DNA constructs encoding GFP or mCherry and/or Smo-ΔN and/or Smo-CA and/or Cl-Sensor (molar ratio 1:2 between fluorescent protein DNA and other DNA constructs). Cl-sensor constructs was cloned into gw vector (Friedel et al., 2015). Smo constructs were provided by Jin Jiang and Philip Beachy (Chen et al., 2011; Kim et al., 2009). Plasmids were further electroporated by delivering 40 V voltage pulses with a BTX ECM 830 electroporator (BTX Harvard Apparatus, Holliston, MA, USA). The voltage was discharged in five electrical pulses at 950-ms intervals via tweezer-type electrodes (Nepa Gene Co, Chiba, Japan) placed on the head of the embryo across the uterine wall. We performed in utero electroporation in embryonic rats at E15, corresponding to an active period of both radial and tangential migration of newborn neurons in the cortex (Kriegstein and Noctor, 2004). At birth, successfully electroporated pups were selected after transcranial visualization of the GFP reporter fluorescent protein only. Following experimentation, morphological analysis of electroporated tissues, selected by GFP or mCherry expression under a fluorescence stereomicroscope (Olympus SZX 16), was performed. The criteria commonly used for selection of animals were localization and cell density of transfected cortices and absence of abnormal cortex morphology (i.e. cortical disruption, thickening and abnormal cortical organization). Our analyses revealed no alterations in cortical layer position or morphology in GFP or mCherry, Smo-CA and Smo-ΔN conditions, although we cannot exclude the possibility that Smo constructs impact the neuronal cells at the cellular and subcellular levels. Nevertheless, ∼20% of electroporations failed due to the absence of transfected cells, as revealed by the lack of fluorescent cells (in GFP or mCherry and Smo conditions). These animals were excluded from the study.

Electroporated zones of the somatosensory cortex were homogenized in RIPA buffer [150 mM NaCl, 1% Triton X-100, 0.1% SDS, 50 mM Tris-HCl, pH 8, containing protease inhibitors (Complete Mini; Roche)]. Lysates were centrifuged (10,000 g for 10 min at 4°C) and the supernatant was heated at 90°C for 5 min with Laemmli loading buffer. Loading was 20 µg of protein, as determined using a modified Bradford reaction (Bio-Rad Laboratories). Proteins were separated by 7–15% SDS–PAGE and electrophoretically transferred to nitrocellulose membranes. Membranes were blocked with 5% bovine serum albumin (BSA) in TBS containing 0.1% Tween 20 (TBST) for 2 h at room temperature, then incubated with primary antibodies diluted in TBST containing 3% BSA overnight at 4°C or for 2 h at room temperature. Blots were probed with antibodies against KCC2 pSer⁹⁴⁰ (1 mg/ml; rabbit, Novus Biologicals), KCC2 pThr¹⁰⁰⁷ (1 mg/ml; sheep, Division of Signal Transduction Therapy Unit, University of Dundee) and KCC2 (1:2000; rabbit, US Biological). After washing with TBST, membranes were incubated with HRP-conjugated secondary antibodies diluted in TBST containing 3% BSA for 60 min, washed with TBST and then developed using a G:BOX gel imaging system (Syngene). The appropriate exposure time of the digital camera for acquisition of chemiluminescent signals from immunoblots was adapted to avoid saturated pixels, and expression levels were estimated using ImageJ software (NIH, Bethesda, MD; http://rsb.info.nih.gov/ij/).

Under deep anesthesia with isoflurane prior to chloral hydrate (7% in 1 M PBS), P10 to P30 electroporated rats were intracardially perfused with cold PBS (1 M) followed by 4% paraformaldehyde (PFA) in PBS. Brains were removed, post-fixed overnight at 4°C and rinsed in PBS. Coronal cortical sections (70 μm thick) were obtained using a vibratome (Microm HM 650 V). Sections were incubated first for 1 h in PBS with 1% BSA and 0.3% Triton X-100, then overnight at 4°C with rabbit anti-Smo (1:500; ab38686; Abcam), rabbit anti-caspase-3 cleaved (1:500; 9661S; Cell Signaling Technology), chicken anti-MAP2 (1:5000; ab5392; Abcam), mouse anti-NeuN (1:1000; MAB377; Chemicon), rabbit anti-FoxP2 (1:4000; ab16046; Abcam) or mouse anti-synaptophysin (1:1000; MAB5258; Chemicon). Sections were rinsed in PBS and incubated for 2 h with the corresponding Alexa488- (1:1000; FluoProbes), Cy5- or Cy3-conjugated (1:1000; Chemicon) secondary antibodies diluted in PBS). DAPI (Vector Laboratories H-1200) was applied to stain nuclei. Control tissues used to determine the level of nonspecific staining included tissues incubated without primary antibody. Sequential acquisition of immunoreactivity of GFP-positive pyramidal-like cells was performed using a laser scanning confocal microscope (Zeiss LSM 510 Meta) with a 40× or 63× oil-immersion objective. For each experimental condition, a semi-quantitative analysis of synaptophysin (Syn)-labeled area fraction per field was measured and reported relative to the area fraction of pixels positive for MAP2 staining. The surface areas of NeuN immuno-positive neurons expressing GFP or Smo mutants were measured manually using the Freehands selection tool in ImageJ. For neuronal migration, electroporated rat brains were removed at E20 and fixed in Antigenfix (Diapath) for 24 h before they were included in agar and sectioned (70 µm) using a vibratome (Leica VT 100 S). Analysis was performed as previously described (Elias et al., 2007), with an automated subdivision into regions of equal length, from the ventricle to the surface [i.e. ventricular/subventricular zones (VZ/SVZ), intermediate zone (IZ) and cortical plate (CP)]. In each set of images, laser light levels and detector gain and offset were adjusted to avoid any saturated levels. Optical sections were digitized (1024×1024 pixels) and processed using ImageJ software. All of the images were analyzed blind.

Expressions of Gli1 and Ptch1 mRNA in the electroporated regions of the somatosensory cortex were measured using real-time RT-qPCR. Total RNA was isolated from cerebral cortices (P15 rats) using a Mini RNeasy kit (Qiagen) then converted to cDNA using 1 µg RNA and a QuantiTect Reverse Transcription kit (Qiagen) according to the manufacturer's instructions. Single-cell gene expression was performed on primary hippocampal neurons transfected with a mixture of constructs encoding mCherry (mock), Smo SA0-5, Smo-ΔN or Smo-CA with or without GANT61 (10 µM) for 48 h. Whole-cell configuration using patch pipettes was used to harvest the cytosol from 15 transfected neurons per condition, which were converted immediately to cDNA as described above. PCR was carried out with a LightCycler 480 SYBR Green I Master (Roche Applied Science) with 1 µL cDNA using the following oligonucleotides (QuantiTect Primer Assays, Qiagen): Gli1 (Gli1; QT01290324), Ptch1 (QT01579669), KCC2 (Slc12a5; QT00145327) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; QT001199633). Relative mRNA values were calculated using LC480 software with GAPDH as the housekeeping gene. PCR was performed in triplicate.

Cortical tissues from electroporated rats at the indicated age were homogenized in RIPA buffer [150 mM NaCl, 1% Triton X-100, 0.1% SDS, 50 mM Tris-HCl, pH 8, containing protease inhibitors (Complete Mini; Roche)]. Lysates were centrifuged (5000 g for 5 min at 4°C). Loading was 200 µg of protein, as determined using a modified Bradford reaction (Bio-Rad Laboratories). Quantification of Shh was performed using a Rat Shh ELISA Kit (FineTest, Wuhan Fine Biotech Co. Ltd., China) in the concentrated solutions following the manufacturer's protocol. Experiments and analysis were performed blind.

Electroporated cortical regions of P14 rats were identified using an atlas of the developing rat brain (Khazipov et al., 2015). Electroporated cortical regions of P20 and P30 rats were identified using the Paxinos and Watson rat brain atlas (Paxinos and Watson, 2005). Based on these two atlases, the somatosensory regions were prepared for acute slice experiments from P14, P20 and P30 rats in ice-cold (2–4°C) oxygenated modified artificial cerebrospinal fluid (ACSF) (0.5 mM CaCl₂ and 7 mM MgSO₄; NaCl replaced by an equimolar concentration of choline). Slices (300-µm thick) were cut with a Vibratome (VT1000E; Leica, Nussloch, Germany) and kept at room temperature (25°C) for at least one hour before recording in oxygenated normal ACSF containing (in mM): 126 NaCl, 3.5 KCl, 2 CaCl₂, 1.3 MgCl₂, 1.2 NaH₂PO₄, 25 NaHCO₃ and 11 glucose, pH 7.4 equilibrated with 95% O₂ and 5% CO₂. Slices were then transferred to a submerged recording chamber perfused with ACSF (3 ml/min) at 34°C.

Extracellular field potential recordings were performed from P14, P20 and P30 electroporated somatosensory cortical slices. Extracellular tungsten electrodes of 50 µm diameter (California Fine Wire, Grover Beach, CA, USA) were positioned in the V/VI cortical pyramidal cell layer of the transfected area to record the multiunit activity (MUA). Field potentials were recorded using a DAM80 Amplifier (World Precision Instruments, Sarasota, FL, USA) using a 1–3 Hz bandpass filter and were analyzed off-line with the Axon package MiniAnalysis program (Jaejin Software, Leonia, NJ, USA). To determine the developmental changes in GABA_A signaling, we used isoguvacine, a potent and selective GABA_A receptor agonist (Krogsgaard-Larsen and Johnston, 1978). We determined the effect of isoguvacine on MUA as a ratio of the spiking frequency at the peak of the isoguvacine response to the spiking frequency in control.

To perform non-invasive monitoring of neuronal intracellular chloride concentration ([Cl⁻]_i), we used a ratiometric genetically-encoded Cl⁻-sensitive probe called Cl-Sensor (Markova et al., 2008; Waseem et al., 2010) that was co-expressed in the cells of interest together with other constructs as described above. The acquisition of fluorescence images was performed using a customized imaging setup with consecutive cell excitation at 430 and 500 nm and emission at 480 and 540 nm. The frequency of acquisition was 0.05 Hz. The duration of excitation was selected for each cell type and was selected to avoid use-dependent bleaching of the signal (Friedel et al., 2013). Results are expressed as fluorescence ratio measured at 430 and 500 nm excitation wavelengths (R_430/500). Experiments were performed on acute cortical slices from in utero electroporated rat pups on postnatal days P10 and P30. Individual slices were transferred to a specially designed recording chamber where they were fully submerged and superfused with oxygenated ACSF complemented with 1 µM tetrodotoxin, 0.3 µM strychnine and 10 µM NBQX to prevent spontaneous neuronal activity and noncontrolled [Cl⁻]_i changes at 30–32°C at a rate of 2–3 ml/min. The applications of the ACSF solution containing isoguvacine (30 µM) or KCl (25 mM)+isoguvacine (30 µM) were performed with a perfusion system.

To evaluate the susceptibility to seizures in control rats (GFP-transfected brains), and in Smo-CA- and Smo-ΔN-expressing animals at P30, pentylenetetrazol (PTZ; 25 mg/kg; Sigma) was administered via intraperitoneal injections every 10 min until generalized seizures occurred. Rats were placed in a plexiglass cage, and the time to the onset of the generalized seizure was measured by observation. There was no significant difference in weight and sex between groups of animals. Experiments were performed blind, and the electroporated region was verified after PTZ induction. Rats with a a very local electroporated cortical area of GFP fluorescence or fluorescence in other brain regions than somatosensory cortex were excluded.

Neurons from 18-day-old rat embryos were dissected and dissociated using trypsin and plated at a density of 70,000 cells cm⁻² in minimal essential medium (MEM) supplemented with 10% NU serum (BD Biosciences, Le Pont de Claix, France), 0.45% glucose, 1 mM sodium pyruvate, 2 mM glutamine and 10 U ml⁻¹ penicillin–streptomycin (Buerli et al., 2007). On days 7, 10 and 13 of culture incubation (DIV, days in vitro), half of the medium was changed to MEM with 2% B27 supplement (Invitrogen). For electrophysiology, neuronal cultures were plated on coverslips placed in 35-mm culture dishes. Twelve hours before plating, dishes with coverslips were coated with polyethylenimine (5 mg/ml). Transfection of cultured neurons was performed with 300 µl Opti-MEM medium mixed with 7 µl Lipofectamine 2000 (Invitrogen), 1 µl Magnetofection CombiMag (OZ Biosciences) per µg of DNA and 1.5 µg premixed DNA encoding the constructs of interest. The mixture was incubated for 20 min at room temperature and thereafter distributed dropwise above the neuronal culture. Culture dishes (35-mm) were placed on a magnetic plate (OZ Biosciences) and incubated for 35 min at 37°C, 5% CO₂. Transfection was terminated by the substitution of 80% of the incubation solution with fresh culture medium. Cells were used in the experiments 48–72 h after transfection. These experiments were based on co-transfection into the same cell of two different pcDNAs encoding a fluorescent marker of transfection (eGFP or mCherry, 0.3 µg), and Smo-related constructs (1.2 µg).

Gramicidin-perforated patch-clamp recordings were performed on primary hippocampal neurons, transfected with a mixture of Smo-related constructs (mCherry, Smo SA0-5, Smo-ΔN and Smo-CA) in the presence or absence of GANT61 (10 µM, 48 h) and with or without KCC2-pH_ext (Friedel et al., 2017). Measurements were performed 2 or 3 days after transfection (corresponding to 8 or 9 DIV). Coverslips with transfected neurons were placed onto the inverted microscope and perfused with an external HEPES-buffered solution (HBS: 140 mM NaCl, 2.5 mM KCl, 20 mM HEPES, 20 mM d-glucose, 2.0 mM CaCl₂, 2.0 mM MgCl₂, and 0.02 mM bumetanide, pH 7.4). For recording from neurons, external HBS contained 0.5 µM tetrodotoxin and 15 µM bumetanide. The recording micropipettes (5 MΩ) were filled with a solution containing 150 mM KCl, 10 mM HEPES and 20 µg/ml gramicidin A, pH 7.2. Isoguvacine (30 µM) was dissolved in an external solution and focally applied to recorded cells through a micropipette connected to a Picospritzer (General Valve Corporation, pressure 5 p.s.i.). Recordings were performed using an Axopatch-200A amplifier and pCLAMP acquisition software (Molecular Devices) in voltage-clamp mode. Data were low-pass filtered at 2 kHz and acquired at 10 kHz. Isoguvacine responses were recorded at voltages −110, −90, −70 and −50 mV, or at −70, −50, −30 and −10 mV depending on neuron GABA inversion potential. A linear regression was used to calculate the best-fit line of the voltage dependence of the isoguvacine responses.

For immunolabeling of KCC2-pH_ext proteins on living neurons, rabbit anti-GFP antibody was dissolved in culture medium applied to neurons for 2 h at 37°C, 5% CO2 (Friedel et al., 2015). Neurons were then rinsed three times at room temperature with HBS, labeled with anti-rabbit Cy3-conjugated antibody for 20 min at 13°C and fixed in Antigenfix (Diapath). To reveal the intracellular pool of live-labeled proteins, cells were permeabilized with 0.3% Triton X-100, blocked by 5% goat serum and incubated for 1 h at room temperature with anti-rabbit Alexa647-conjugated antibody. For visualization of the total pool of overexpressed KCC2-pH_ext, cells were labeled overnight (4°C) with mouse anti-GFP antibody and for 1 h at room temperature with anti-mouse Alexa488-conjugated antibody. For control of the cell membrane integrity during live-cell immunolabeling, one batch of cultures were routinely transfected with the KCC2-pH_ext mutant ΔNTD-KCC2-pH_ext, which does not incorporate into the plasma membrane (Friedel et al., 2017). A/A-KCC2-pH_ext, a double phosphomimetic mutant of KCC2 for T906A/T1007A that maintains KCC2 at the cell surface (Friedel et al., 2015), was used as a positive control.

Images of labeled neurons were acquired with an Olympus FluoView 500 confocal microscope [oil-immersion objective 60× (NA 1.4); zoom 1–5]. We randomly selected and focused on a transfected cell by only visualizing Alexa488 fluorescence and then acquired z-stack images of Alexa488, Cy3 and Alexa647 fluorochromes. Each z-stack included ten planes of 1 µm optical thickness, taken at 0.5 µm distance between planes. The cluster properties and fluorescence intensities of each cell were analyzed with Metamorph software. First, we used the logical ‘NOT’ conversion of pairs of Alexa647 and Cy3 images to isolate in each focal plane the Alexa647 signal that was not overlapping with Cy3 fluorescence restricted to the plasma membrane. This gave rise to additional images reflecting the fluorescence of the internalized pool of labeled clusters, called thereafter ‘NOT-conversion’. Second, the arithmetic summation for each z-stack and channel was performed to collect the whole fluorescence of the different signals (Alexa488 signal for total protein fluorescence; Cy3 signal for plasma membrane restricted fluorescence; NOT-conversion signal for internalized restricted fluorescence; and Alexa647 signal for all surface-labeled fluorescence). Third, a binary mask was created for each cell, using the Alexa488 image, to isolate the signal coming from the transfected neuron, and the fluorescence parameters (total fluorescence, single cluster fluorescence as well as density and brightness of clusters) were analyzed for each channel (Alexa488, Cy3, NOT-conversion and Alexa647) in regions overlapping with the binary mask. The analysis parameters were the same for each experiment, and all experiments were performed blind. After analysis, data were normalized to the mean value of cells transfected with KCC2-pHext+GFP.

No statistical methods were used to predetermine sample sizes. To ensure the consistency and reproducibility of our results, we conducted repeated trials in different cell cultures and acute brain slices prepared from at least three different animals for each experimental condition. If not stated otherwise, statistics are presented as the mean±s.d. for normally distributed data and as the median only for non-normally distributed data. Experiments with control and Smo-electroporated animals were processed at the same time to ensure homogeneity of experimental conditions.

We thank Drs Y. Ben Ari and J. L. Gaiarsa for critical reading of the manuscript and F. Michel at InMAGIC for technical assistance.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.247700.reviewer-comments.pdf