Excitatory action of low frequency depolarizing GABA/glycine synaptic inputs is prevalent in prenatal spinal SOD1^G93A motoneurons

Abstract

Amyotrophic lateral sclerosis (ALS) is a fatal adult-onset neurodegenerative disease characterized by progressive motor neuron degeneration and muscle paralysis. Recent evidence suggests the dysfunction of inhibitory signalling in ALS motor neurons. We have shown that embryonic day (E)17.5 spinal motoneurons (MNs) of the SOD1^G93A mouse model of ALS exhibit an altered chloride homeostasis. At this prenatal stage, inhibition of spinal motoneurons (MNs) is mediated by depolarizing GABAergic/glycinergic postsynaptic potentials (dGPSPs). Here, using an ex vivo preparation and patch clamp recording from MNs with a chloride equilibrium set below spike threshold, we report that low input resistance (R_in) E17.5 MNs from the SOD1^G93A ALS mouse model do not correctly integrate dGPSPs evoked by electrical stimulations of GABA/glycine inputs at different frequencies. Indeed, firing activity of most wild-type (WT) MNs with low R_in was inhibited by incoming dGPSPs, whereas low R_in SOD1^G93A MNs were excited or exhibited a dual response (excited by low frequency dGPSPs and inhibited by high frequency dGPSPs). Simulation highlighted the importance of the GABA/glycine input density and showed that pure excitation could be obtained in SOD-like MNs by moving GABA/glycine input away from the cell body to dendrites. This was in agreement with confocal imaging showing a lack of peri-somatic inhibitory terminals in SOD1^G93A MNs compared to WT littermates. Putative fast ALS-vulnerable MNs with low R_in are therefore lacking functional inhibition at the near-term prenatal stage.

Key points

• We analysed the integration of GABAergic/glycinergic synaptic events by embryonic spinal motoneurons (MNs) in a mouse model of the amyotrophic lateral sclerosis (ALS) neurodegenerative disease.
• We found that GABAergic/glycinergic synaptic events do not properly inhibit ALS MNs with low input resistance, most probably corresponding to future vulnerable MNs.
• We used a neuron model to highlight the importance of the GABA/glycine terminal location and density in the integration of the GABAergic/glycinergic synaptic events.
• Confocal imaging showed a lack of GABA/glycine terminals on the cell body of ALS MNs.
• The present study suggests that putative ALS vulnerable MNs with low R_in lack functional inhibition at the near-term stage.

Introduction

Amyotrophic lateral sclerosis (ALS) is an uncurable, fatal neurodegenerative disease characterized by a gradual degeneration of motoneurons (MNs) located in the brainstem and spinal cord, as well as in pyramidal motor neurons located in the primary motor cortex (Kiernan et al., 2011). Despite decades of effort, the disease mechanisms are still largely unknown, resulting in a conspicuous absence of any treatment. Most of the ALS cases are sporadic but ∼10% of the cases are familial, where the disease is transmitted in families, usually with dominant inheritance traits (Taylor et al., 2016). Using the SOD1^G93A transgenic mouse carrying the mutated human SOD1 gene (Gly93→Ala substitution) (Gurney et al., 1994) and expressing pathological and clinical symptoms similar to those seen in patients with ALS (Tu et al., 1996), we have found that SOD1^G93A MNs are hyperexcitable when maintained in primary culture from embryonic day (E)12–14 (Kuo et al., 2004) and, at E17.5, have an elevated input resistance (R_in) resulting from a shorter dendritic tree (Martin et al., 2013). In SOD1^G93A mice, E17.5 is the day before birth. At this terminal prenatal stage, the spinal locomotor networks become functional and it is possible to evoke left–right alternating locomotor-like activities from ex vivo brainstem-spinal cord preparations (Branchereau et al., 2019), indicating that spinal inhibitory commissural connections are probably functional (Branchereau et al., 2000). However, at E17.5, the effects of GABA and glycine on lumbar spinal MNs are not hyperpolarizing but rather evoke depolarizing GABAergic/glycinergic post-synaptic potentials (dGPSPs) (Delpy et al., 2008), which have mixed inhibitory (shunting) and excitatory (depolarizing) components (Branchereau et al., 2016).

Converging lines of evidence highlight alterations in neuronal inhibitory processes (Clark et al., 2015; Hossaini et al., 2011; Ramírez-Jarquín et al., 2014; Van den Bos et al., 2018), with a perinatal time window of vulnerability (Cavarsan et al., 2022; Eisen et al., 2014; Kiernan et al., 2019), among numerous potential causal links in ALS-associated dysfunction and degeneration. With this line of evidence, we found previously that chloride homeostasis is altered in embryonic E17.5 SOD1^G93A MNs (Branchereau et al., 2019). However, we did not investigate the effect of dGPSPs on the repetitive firing ability of E17.5 lumbar MNs that drive the contraction of muscle fibres, constituting the final common pathway (Manuel & Zytnicki, 2011). At high frequency, dGPSPs are inhibitory because of chloride conductance (gCl) summation (Branchereau et al., 2016) but modelling performed on E17.5 MN models has indicated that low frequency (<50Hz) dGPSPs could be excitatory in SOD-like MNs, but not in wild-type (WT)-like MNs with a similar chloride equilibrium potential (ECl) (Branchereau et al., 2019).

Here, we performed ex vivo electrophysiological experiments using whole-cell patch clamp recordings from WT and SOD1^G93A fetal E17.5 mouse lumbar spinal MNs. ECl was imposed to assess specifically the effect of dGPSPs on additional parameters including MN morphology. Physiological dGPSPs were evoked by activating local pharmacologically isolated GABA/glycine fibres with ventro-lateral funiculus (VLF) electrical stimulations. The effect of dGPSPs was assessed on the firing of MNs elicited by intracellular injection of current. Our results demonstrate dual effects of dGPSPs on MN firing frequency, namely excitatory at low dGPSP frequency and inhibitory at high frequency (dual MNs) and also show that dGPSPs are able to exert purely inhibitory effects at low frequency (inhibited MNs). The majority of SOD1 MNs were dual regardless of their R_in, whereas the response of WT MNs was largely dependent on their input resistance: for low R_in WT MNs, the majority were inhibited MNs, whereas most high R_in WT MNs were dual MNs. Simulation demonstrated that reducing the density of inhibitory synapses on MNs favours the excitatory effect of repetitive dGPSPs on MN firing. Interestingly, confocal imaging performed from recorded MNs revealed that the density of GABA and glycine releasing terminals was lower in SOD MNs compared to WT MNs. Taken together, our data show that a low density of GABA/glycine terminals on the peri-somatic motoneuronal membrane prevents low R_in SOD E17.5 MN from being inhibited by GABAergic/glycinergic synaptic events that are depolarizing at late prenatal stages, highlighting the early dysfunction of the GABA/glycine inhibitory system in the SOD1^G93A strain.

Methods

Ethical approval

All procedures were carried out in accordance with the local ethics committee of the University of Bordeaux and the UE Directive 2010/63. The project was approved by the French Ministry of Higher Education, Research and Innovation (APAFIS#5022). To apply the 3Rs rule, all efforts were made to curtail animal suffering and reduce the number of animals used.

Experimental models

B6SJL-Tg(SOD1*G93A)1Gur/J mice expressing the human G93A Cu/Zn superoxide dismutase (SOD1) mutation (Gly93→ Ala substitution) (JAKSIMA-2726; Jackson Laboratory; https://www.jax.org/strain/002726) were obtained from Charles River Laboratories (Saint Germain Nuelles, France). Heterozygous B6SJL-Tg(SOD1*G93A)1Gur/J (named SOD in the present study) were maintained by crossing heterozygous transgene-positive male mice with B6SJL F1 hybrid females (Janvier Labs, Saint Berthevin, France). Gestation in B6SJL mice lasts between 18 and 20 days (see reproduction data) but most of time it lasted 18.5 days, with E0.5 being defined as the day after the mating night. Experiments were performed on E17.5 fetuses of either sex, collected 1 day before their birth.

Dissection and isolation of the embryonic brainstem-spinal cord preparation

Pregnant mice were killed by cervical dislocation. The fetuses were removed from the mother using a laparotomy surgical procedure and then transferred into cold artificial cerebrospinal fluid (aCSF) (6–8°C), oxygenated with a 95% O₂/5% CO₂ mixture. The aCSF comprised (in mm): 114.5 NaCl, 3 KCl, 2 CaCl₂.2H₂O, 1 MgCl₂.6H₂O, 25 NaHCO₃, 1 NaH₂PO_4c·H₂O and 11 d-glucose, with a pH of 7.4 and osmolality is mosmol kg^–1 H₂O. Fetuses were chosen randomly, decapitated and the brainstem–spinal cord preparation was dissected out under oxygenated cold aCSF. The suprasegmental transection was performed under a binocular microscope at the level of the r1 alar plate (https://developingmouse.brain-map.org/static/atlas). The brainstem was not detached to keep intact descending inputs that already reach the lumbar spinal cord at E17.5, such as serotonergic (Ballion et al., 2002; Rajaofetra et al., 1989) and noradrenergic (Rajaofetra et al., 1992) fibres. Meninges were removed all along the brainstem–spinal cord after opening the spinal cord dorsally (open-book preparation). This procedure allowed easy access to lumbar MNs for the patch clamp electrode. The preparation was then positioned in a recording chamber with the ventral side up, maintained open under a nylon mesh and superfused (∼1.5 mL min⁻¹) with oxygenated aCSF. All experiments were carried out at constant temperature (30°C) (Temperature CXontroller V; Luigs & Neumann, Ratingen, Germany). Experiments were performed blindly without knowing the genotype of the animals. We recorded only one or two MNs from each brainstem–spinal cord preparation. Fetuses were genotyped at a genotyping platform (Magendie Neurocentre, Bordeaux) and was performed by standard PCR from mice tail samples using the primers stated in a protocol from the Jackson Laboratory (https://www.jax.org/Protocol?stockNumber=002726&protocolID=29082).

Tissue processing

Immunohistochemistry was then conducted Some MNs were recorded with glass pipettes filled with 0.4% (wt/vol) neurobiotin (CliniSciences, Montrouge, France) diluted in the intracellular medium (see below). After recording sessions, the whole brainstem-spinal cord (SC) preparations were fixed in 4% paraformaldehyde prepared with 0.1 m phosphate buffer (w/v) for 2 h at room temperature. They were rinsed three times with 0.1 m phosphate-buffered saline (PBS), blocked with a blocking buffer containing 2 % (w/v) bovine serum albumin and 0.5% PBST [0.5% Triton X-100 in 0.1 m PBS (v/v], followed by incubation with primary antibody and streptavidin-Cy3 (dilution 1:400; Invitrogen, Thermo Fisher Scientific, Life Technologies, Villebon-sur-Yvette, France) diluted in incubating solution made with 0.1 m PBS containing 0.1% Triton X-100 (Sigma-Aldrich Chimie SARL, Saint Quentin Fallavier, France) and 0.2% bovine serum albumin (Sigma-Aldrich Chimie SARL) for 48 h at 4°C. We processed brainstem–SC preparations with a rabbit antibody directed against the vesicular inhibitory amino acid transporter (VIAAT) (dilution 1:1000; antibody Provided by B. Gasnier, SPPIN – Saints-Pères Paris Institute for the Neurosciences, Paris, France). VIAAT reflects the synaptic release of the GABA and glycine neurotransmitters (Dumoulin et al., 1999). The brainstem–SC preparations were thereafter washed four times for 20 min, incubated with secondary antibody conjugated to Alexa Fluor488 goat anti-rabbit IgG(H + L) (dilution 1:500; Invitrogen) for 2 h, at room temperature, rinsed abundantly in 0.1 m PBS, and finally mounted with an anti-fade reagent (Fluoromount, Electron Microscopy Sciences, CliniSciences, France) and stored at 4°C in obscurity until confocal observation.

Confocal imaging

Fluorescence images were acquired with a LSM 900 confocal microscope (Carl Zeiss, Marly le Roi, France), with a 63× oil-immersion objective (NA 1.40) and optical section thicknesses set at 300 nm. Imaging parameters were kept constant across preparations and genotypes. Confocal stacks were acquired with 10–15 optical sections for each MN. Lasers were selected according to the wavelength required for visualization. VIAAT immunostaining was visualized as spot aggregates, which were detected using an ImageJ macro (https://imagej.nih.gov/ij/developer/macro/macros.html) allowing delineation of the periphery of MNs and quantification of the VIAAT staining density. The macro was developed by Gilles Courtand (Research Engineer at the INCIA Lab). MNs were identified as neurobiotin-positive staining. VIAAT terminals located on the cytoplasmic membrane and its close proximity (a 20 pixel wide band image) were considered in our analysis. Based on the Nyquist criterion and our confocal acquisition parameters (1 pixel = 0.099 μm), the area of the smallest VIAAT-positive spot was 0.0518 μm² [0.099 μm × 2.3 in x and y direction = (0.099 μm × 2.3)² = 0.0518 μm²]. The ImageJ macro therefore defined VIAAT-positive terminals as areas of immune-positive spots greater than 0.0518 μm². The global density of VIAAT was then assessed by analysing five consecutive optical sections. The VIAAT density was calculated by dividing the total number of VIAAT spots by the total length of their selected band image. Three to six proximal dendrites (dendritic length <50 μm) per MN and two sides of each dendrite were quantified.

Electrophysiological procedures and data analysis

Patch clamp electrodes were constructed from thin-walled single filamented borosilicate glass (1.5 mm outer diameter; Harvard Apparatus, Les Ulis, France) using a two-stage vertical microelectrode puller (PP-830; Narishige, Tokyo, Japan). Patch electrode resistances ranged from 4 to 6 MΩ. All recordings were made with an Axon Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA, USA). Data were low-pass filtered (2 kHz) and acquired at 20 kHz on a computer via an analog-to-digital converter (Digidata 1322A; Molecular Devices) and a data acquisition software (Clampex 10.7; Molecular Devices). An Axioskop 2 FS Plus microscope (Carl Zeiss) was used to visualize the MNs. A patch clamp electrode was positioned on a visually identified MN using a CCD video camera (Axiocam MRm; Carl Zeiss) using motorized micromanipulators (Luigs & Neumann). All recorded MNs were located in the lateral column as previously described (Branchereau et al., 2019) and were identified by their pear-shaped cell bodies. Values of R_in and membrane capacitance (C_m) also further confirmed the motoneuron identity of recorded neurons (Martin et al., 2013).

Evoked dGPSPs in MNs were elicited using a concentric bipolar electrode (CBBPE75; FHC, Bowdoin, ME, USA) positioned on the VLF, <500 μm rostral to the recording electrode, aiming to activate local inhibitory fibres that were pharmacologically isolated (see below) (Fig. S1). Monophasic single stimulations delivered to the VLF (duration 1 ms, intensity ranging from 3 to 30 μA) were driven by a programmable Master 8 Stimulator/Pulse generator (Master-8; A.M.P.I. Jerusalem, Israel). The intensity of the VLF stimulation was adjusted to obtain evoked dGPSPs having an amplitude in the range of the maximum spontaneous dGPSPs.

E_GABAA/glyR, considered as ECl because of the low HCO3⁻ conductance of GABA_AR/glycineR in embryonic spinal MNs (Bormann et al., 1987), was set to −45 mV (−57.5 mV after correction of the junction potential, see below), which was in the range of the recorded physiological E_GABAAR in E17.5 MNs (Branchereau et al., 2019). Intracellular medium composition (in mm) comprised: 115 K gluconate, 5 NaCl, 15 KCl, 1 CaCl₂, 2H₂O, 10 HEPES, 10 EGTA and 2 MgATP (297 mosmol kg^–1 H₂O) adjusted to pH 7.4 using 1 m KOH. Measurements were corrected for liquid junction potentials (−12.5 mV) calculated with the Clampex junction potential calculator and confirmed by experimental measurements. Actual ECl was measured by assessing the reversal membrane potential of evoked IPSCs elicited by a single VLF stimulation (resulting in local inhibitory fibre activation) at different membrane voltages (Fig. 1Fa; for the specificity of GABA_AR/glyR activation, see also Fig. S2). The reversal potential was calculated by a linear fit (Prism 9; GraphPad Software Inc., San Diego, CA, USA) (Fig. 1Fb).

Figure 1. Differential effects of dGPSPs on E17.5 MN spiking activity

Aa–Ad, example of an excited MN. Ba–Bd, example of a dual MN. Ca–Cd, example of an inhibited MN. Data in Aa–Ad are from the same MN. Same for Ba–Bd and Ca–Cd. Aa, representative current clamp traces of a lumbar E17.5 spinal MN exhibiting constant firing in response to a square depolarizing current pulse (150 pA). Upper: showing the instantaneous spiking frequency, whereas the upper traces (right corner) depict evoked dGPSPs in the range of amplitude of maximum spontaneous dGPSPs. Ab, in the same MN, a 30 Hz VLF stimulation that specifically activates local GABA/glycine fibres (see Methods), increases the MN's firing frequency. Ac, a 60 Hz VLF stimulation induces a further increase in MN firing frequency. Ad, quantification of the change in MN firing frequency (spike number change, % of control) for VLF stimulations between 7.5 Hz to 200 Hz. Note that VLF stimulation evokes an increase in firing frequency whatever the tested frequency. Thus the MN is classified as ‘excited.’ Ba, similar traces for a different representative E17.5 MN showing an excitation (increase in spike frequency) with a 30 Hz VLF stimulation frequency (Bb), but switching to an inhibitory response at 60 Hz (Bc). The graph of the spike number change (Bd) shows that the switch occurs between 50  and 60 Hz for this MN, which is therefore classified as ‘dual.’ Ca, a third representative fetal MN showing partial inhibition of its firing activity when receiving dGPSPs at a VLF stimulus frequency of 30 Hz (Cb), then transitioning to a full inhibition at a frequency of 50 Hz (Cc). This type of fetal MN is classified ‘inhibited.’ Cd, quantification of the change in MN firing frequency for VLF stimulations between 7.5 and 200 Hz. D, distribution of excited, dual and inhibited MNs in WT and SOD E17.5 MNs from the same littermates. Note that SOD MNs are predominantly excited or dual, whereas WT MNs are mainly inhibited. ****P < 0.0001, chi-squared test. Ea, control firing frequency of inhibited (blue bar) and dual (green bar) WT, and inhibited (blue bar), dual (green bar) and excited (red bar) SOD MNs. Note the similar firing frequencies (no statistical differences, Kruskal–Wallis test followed by Dunn's multiple comparisons test). Eb, VLF stimulation intensity (μA) used to elicit similar dGPSPs in all MNs. Note the lower mean VLF intensity values required in WT MNs compared to SOD MNs. **P = 0.0092, Kruskal–Wallis and Dunn's tests. Ec, amplitudes of VLF-evoked dGPSPs are comparable between groups, except in excited SOD MNs that exhibit smaller evoked dGPSPs. Kruskal–Wallis and Dunn's tests. Fa, voltage clamp traces showing the VLF-induced (single stimulation, at arrow) current that is upward for more positive holding potentials and downward for negative ones. Fb, I_GABA/gly currents expressed as a function of the holding potential allowing evaluation of the actual ECl value (−48.1 mV). Actual ECl (Fc) and gIPSC/C_m (Fd) values are similar in WT and SOD MNs groups. Kruskal–Wallis and Dunn's test. [Colour figure can be viewed at wileyonlinelibrary.com]

The Clampex 10.7 (Molecular Devices) membrane test was used to monitor the passive properties of a given neuron, namely membrane input resistance (R_in), access resistance (R_a) and membrane capacitance (C_m). Only neurons with a resting membrane potential more hyperpolarized than −50 mV (value not corrected for the junction potential) were used for recording. Rheobase was recorded as the minimum injected current pulse (duration 60 ms) required to elicit an action potential. Intracellular stimulation intensity was adjusted for each recording to evoke a train of action potentials at ∼20 Hz (Fig. 1Ea), the duration of which was set to 5 s. The intensity of VLF stimulation was adjusted to elicit synaptic events equivalent to spontaneous ones observed in MNs (Fig. 1A–C, upper traces). Nine VLF frequencies were selected from previous simulation data: 7.5, 10, 20, 30, 40, 50, 10, 150 and 200 Hz. The duration of VLF stimulation was maintained at 2.5 s for all experiments, occurring 2.5 s after MN spiking activity (Fig. 1A–C). The recording protocol consisted of one control burst (by intracellular stimulation of cell to produce a train of spikes), followed by one test burst (intracellular stimulation accompanied by a VLF stimulation after 2.5 s). The spike number change (% of control) was assessed during 1 s by comparing spike number during VLF stimulation with the spike number in the preceding control, during the same second. The interburst duration was set at ∼40 s. This frequency discharge protocol was recorded in current clamp mode and analysed using a Spike2 V7 (Cambridge Electronic Design Ltd, Cambridge, UK) script developed by one of the authors (DC), whereas the actual ECl for each MN was assessed from voltage clamp recordings analysed using Clampfit 10.7 (Axon Instruments, Molecular Devices).

Pharmacology

In whole-cell patch clamp experiments, GABA_AR/GlyR responses were pharmacologically isolated by using a cocktail solution containing 4 mm kynurenic acid (Tocris Bio-Techne, Noyal-Châtillon-sur-Seiche, France) and 5 μm dihydro-beta erythroidine hydrobromide, which blocked glutamate and cholinergic input, respectively, to MNs. A further mixture of methysergide maleate (10 μm) (Tocris Bio-Techne ) plus ketanserine tartrate (10 μm) (Tocris Bio-Techne ) was used to ensure the blockade of 5-HT_2R, which are the preferential targets of these drugs. Because 5-HT2A receptor activation is able to modulate chloride homeostasis by acting on the K⁺-Cl⁻ cotransporter KCC2 (Bos et al., 2013; Martin et al., 2020; Medina et al., 2014), this cocktail was used in all experiments to saturate 5-HT2A receptors and prevent a putative modulation of chloride homeostasis by descending 5-HT (Ballion et al., 2002) during the VLF stimulation. Gabazine 3 μm (SR 95531 hydrobromide; Tocris Bio-Techne ) and strychnine hemisulfate 3 μm (Sigma-Aldrich Chimie SARL) were used to verify that our VLF stimulation specifically elicited GABA_AR- and GlyR-related synaptic events (Fig. S2).

Membrane properties and data analysis

Data analyses were performed using Clampfit 10.7 (Axon Instruments) and Spike2 V7 (Cambridge Electronic Design Ltd). For synaptic event analysis, representative 1–2 min recording samples were selected from each MN. The MN capacitance (C_m), membrane input resistance (R_in) and access resistance (R_a) were determined immediately after establishing whole-cell patch clamp.

Computer simulations

GABA/glycine synaptic events in E17.5 MNs were simulated using a multicompartment neuron model that was elaborated using NEURON 7.3 (Hines & Carnevale, 1997). Two simulated neurons were constructed: an E17.5 WT-like MN and an E17.5 SOD-like MN. These canonical MNs, designed from the average topology of real E17.5 MNs, were built on models used in a previous study. Both MNs were virtually identical, with similar channels and morphologies, except for the terminal dendritic segments of the SOD-like MN that were 40% less extensive than the WT terminal segments (Martin et al., 2013). In each section (dendrites and axon), the number of segments was prescribed an odd number that was calculated according to the d-lambda rule (Carnevale & Hines, 2006; Hines & Carnevale, 2001). The conductances used were: a passive leakage current (I_leak), a transient K channel (I_A), an SK calcium-dependent potassium channel, also called I_K(Ca) (Gao & Ziskind-Conhaim, 1998; McLarnon et al., 1995), and a high-threshold calcium current (I_L) (Walton & Fulton, 1986). I_leak was simulated in each MN section [I_leak = (E_leak – E) × G_leak, with E_leak = −73 mV and G_leak = 1/R_m]. R_m, the specific membrane resistance, was set to 21,200 Ω.cm² to obtain an input resistance R_in of 120 MΩ in the WT MN. I_A was present in the axon initial segment (AIS), with G_IA = 0.0005 S·cm⁻²; SK was present in soma, with G_SK = 0.0035 S·cm⁻², and was activated via the calcium current I_L also present in the soma, with G_IL = 9.10^{− 5} S·cm⁻². Each of the axon segments (n = 750) was equipped with Na and K Hodgkin–Huxley channels (G_Na = 0.012 S·cm⁻² and G_K = 0.0036 S·cm⁻², respectively) used to generate spikes. The densities of Na and K channels in the initial segment (G_Na = 0.5 S·cm⁻² and G_K = 0.15 S·cm⁻²) were adjusted to obtain a spike threshold of −48 mV.

In addition, calcium dynamics (Blaustein, 1988) were added in the soma to reproduce calcium accumulation, diffusion and pumping (Destexhe et al., 1993). The parameters used in the intracellular calcium dynamics were: depth = 0.17 mm, tau_r = 10¹⁰ ms, Ca_inF = 0.0002 mm, Kt = 0.00025 mm·ms⁻¹, Kd = 0.0001 mm, Ca_i = 2.4 × 10^−4 mm and Ca_o = 3 mm.

Inhibitory synaptic inputs were inserted on the somatic MN compartment using two-exponentials equations (one for the rising phase and the other for the decay phase). Depolarizing GABAergic/glycinergic post-synaptic current kinetics was set at tau_rise = 0.3 ms and tau_decay = 20 ms. More details about equations for channel properties and synaptic activation are provided in Branchereau et al. (2019) and Martin et al. (2013).

Statistical analysis

Prism 9  (GraphPad Software Inc.) was used to analyse all data. The results are presented as the mean ± SD unless otherwise specified; n is the number of MNs used in the analysis and N is the number of fetuses. P < 0.05 was considered statistically significant. The statistical differences between two data sets were assessed with the Mann–Whitney test for non-parametric data. Values for the Kruskal–Wallis test followed by a Dunn's comparisons test are also given in Table 1.

Table 1. Physiological values measured in E17.5 lumbar spinal MNs

MN subgroup Number of MNs	WT inhibited (n = 19)	WT dual (n = 13)	SOD inhibited (n = 11)	SOD dual (n = 21)	SOD excited (n = 7)
Female MNs	n = 11	n = 7	n = 7	n = 8	n = 3
Male MNs	n = 8	n = 6	n = 4	n = 13	n = 4
Number of litters	N = 11	N = 13	N = 10	N = 16	N = 7
ERest (mV)	−74.2 ± 5.0	−74.9 ± 5.3	−75.5 ± 4.9	−76.8 ± 3.9	−77.8 ± 3.8
Rin	82.0 ± 25.1	128.7 ± 56.6	143.3 ± 70.2	119.8 ± 45.5	153.1 ± 70.6
Cm (MΩ)	171.9 ± 57.7*	122.3 ± 37.4	130.5 ± 36.2	129.4 ± 40.9	124.1 ± 26.2
Rheobase (pA)	246.8 ± 129.0	199.5 ± 162.8	173.8 ± 94.5	201.2 ± 121.9	202.0 ± 209.2
SpNbCh30 (%)	34.9 ± 41.6**	101.6 ± 29.8	65.7 ± 31.5**	97.6 ± 28.3***	121.3 ± 9.4****
Spike amp. (mV)	50.2 ± 6.2	49.0 ± 7.3	48.1 ± 5.4	52.8 ± 6.4	54.6 ± 7.1
Spike width (ms)	1.5 ± 0.2	1.4 ± 0.1	1.7 ± 0.3	1.4 ± 0.2	1.4 ± 0.2
Vthr – Vr (mV)	28.5 ± 6.5	31.2 ± 7.9	30.5 ± 6.3	31.6 ± 6.6	31.8 ± 2.8
Firing freq (Hz)	18.8 ± 5.6	18.0 ± 5.3	20.1 ± 4.5	18.3 ± 4.8	16.0 ± 4.8
VLF intensity (μA)	8.3 ± 6.6**	8.4 ± 4.8	15.0 ± 7.2	15.1 ± 6.6	13.6 ± 10.1
Evoked dGPSP amp. (mV)	3.9 ± 1.5 (18)	4.1 ± 1.0	3.9 ± 1.8	4.4 ± 1.8	3.0 ± 1.0
ECl (mV)	−52.1 ± 6.9 (18)	−47.4 ± 5.6	−51.6 ± 9.8	−50.6 ± 9.2 (20)	−43.6 ± 10.0 (6)
gIPSC (nS)	4.0 ± 2.5 (18)	3.8 ± 2.3	4.6 ± 2.9	4.6 ± 2.8 (20)	3.2 ± 2.5 (6)
gIPSC/Cm (nS/pF)	25.3 ± 18.1 (18)	34.1 ± 19.9	39.3 ± 28.0	39.6 ± 24.9 (20)	27.6 ± 19.3 (6)

• Values in parenthesis are numbers of MNs where corresponding values are not mentioned above. E_Rest: ns. R_in: ns. C_m: *Between WT inhibited and WT dual (P = 0.0422). Rheobase: ns. SpNbCh30: **Between WT Inh and WT dual (P = 0.0014). ***Between SOD dual and WT inhibited (P = 0.0002). ****Between SOD excited and WT inhibited (P < 0.0001). **Between SOD inhibited and SOD excited (P = 0.0021). Spike amp.: ns. Spike width: ns. V_thr – V_r: ns. Firing freq.: ns. VLF intensity: **Between WT inhibited and SOD dual (P = 0.0092). Evoked dGPSP amp.: ns. ECl: ns. gIPSC: ns. gIPSC/C_m: ns. ns, not significant; *P < 0.05; **P < 0.01; ****P < 0.0001. Kruskal–Wallis test followed by Dunn's multiple comparisons test.

Principal component analysis (PCA)

Principal component analysis was performed using R software (R Foundation, Vienna, Austria) (Ade4 package) based on the behavioural variables described in Table 2. In a first step, the contribution of each variable to the variance in the first and second components of the PCA was calculated (Tables 3 and 4). Only those variables for which the contribution to the variance of a given component was larger than the mean contribution of all variables to that component (i.e. #2000) were considered (Asterisks in Tables 3 and 4).

Table 2. Variables used for PCA

Variables	Definition	Measurement	Range
SpNbCh30	% Change in number of spikes during 30 Hz VLF stimulation	%	0–150
Rin	Input resistance	MΩ	113–264
Rheobase	Threshold current	pA	60–510
Vthr – Vr	Threshold Mb potential –- resting Mb potential	pA	21–43
Spike width	Mid-amplitude spike width	ms	1–2.25

Results

SOD lumbar MNs are more excited by low frequency dGPSPs

Using simulations made on E17.5 WT-like MNs, we previously found that repetitive dGPSPs, delivered to the MN soma, were able to totally block ongoing MN discharge when reaching cut-off frequency (Branchereau et al., 2019). Interestingly, similar simulations made on E17.5 SOD-like MNs exhibiting shorter dendritic trees (Martin et al., 2013) revealed that such repetitive dGPSPs were also able to exert an excitatory effect on ongoing MN discharge when occurring at low frequency before switching to a full inhibitory effect at higher frequency (cut-off frequency) (Branchereau et al., 2019). Here, we initially aimed to verify whether this excitatory effect was found in real prenatal SOD1^G93A MNs. To address this, we used whole-cell patch clamp recordings from lumbar E17.5 MNs, with a total of 71 MNs (N = 39 litters) (Table 1) located in the lateral column (Branchereau et al., 2019), in which ECl was set to −57.5 mV (i.e. in the range of the recorded physiological E_GABAAR in biological E17.5 spinal MNs) (Branchereau et al., 2019; Delpy et al., 2008). SOD or WT genotype was unknown at the time of recording. MNs were depolarized by injecting through the recording patch clamp pipette in current clamp mode, using a positive current to elicit tonic firing (∼20 Hz) (Fig. 1A–C) during 5 s, whereas dGPSPs were evoked at different frequencies by electrical stimulations of the VLF in the presence of glutamate, cholinergic and serotonergic blockers. As shown in Fig. 1, three types of dGPSP-mediated effects were found: dGPSPs having an excitatory effect at both low and high frequency (up to 200 Hz) (termed excited MNs) (Fig. 1A), dGPSPs having an excitatory effect at low frequency, but switching to an inhibitory effect above ∼40 Hz (termed dual MNs) (Fig. 1B) and dGPSPs exerting inhibitory effects at low and high frequencies (between 10 and 200 Hz) (termed inhibited MNs) (Fig. 1C).

The inhibitory effects observed during GABA/Gly synapse activation could result from an inactivation of sodium channels at potentials around −45 mV, or to a shunting effect induced by GABA/Gly synapse activation.  We therefore performed a simulation to explore the effect of activating GABA/Gly synapses on the MN discharge and spike amplitude (Fig. S3). We used the WT MN model subjected to a depolarizing current of 280 pA throughout the whole recording time. A barrage of stimulation (83.33 Hz) of the GABA/Gly synapse was started at t = 900 ms (ECl = −50 mV). With gGABA/Gly set at 16 pS, GABA/Gly stimulation resulted in an excitatory effect (Fig. S3A). During this response, there was no inactivation before each spike onset (Fig. S3B). The spike recorded in the soma was smaller than that detected in the initial axon segment or in the axon itself (Fig. S3C) because there was no voltage-dependent sodium and potassium channels present in the soma, and also because of the shunting effect of the GABA/Gly synapse. However, when gGABA/Gly was increased to 30 pS, the shunting effect became correspondingly larger, leading to a net inhibitory effect on MN firing (Fig. S3D) and a decreased in spike amplitude (Fig. S3D). Note that the amplitudes and kinetics of sodium channel activation/inactivation (m, h) and potassium channel activation (n) (Fig. S3E) were similar to those obtained with gGABA/Gly = 16 pS (Fig. S3B). Moreover, there was no inactivation before onset of each spike (Fig. S3E). Nevertheless, the shunting effect of GABA/Gly synapse was responsible for slowing down MN discharge and for the amplitude decrease of spikes recorded in soma (Fig. S3F). Our simulation therefore indicates that an inactivation of sodium channels at potentials around −45 mV is not required to produce inhibitory effects during the dGPSP barrage. Moreover, an inactivation is mostly absent, since in other recordings at a similar membrane potential (−45 mV), the dGPSP barrage resulted in an increase in MN discharge.

Subsequent genotyping revealed that ∼60% of the WT MNs (19/32; seven female and eight male MNs; N = 11 litters) were inhibited, whereas the remaining ∼40% (13/32; seven female and six male MNs; N = 13) were dual MNs. Interestingly, the genotyping also revealed that excited MNs were exclusively SOD MNs (n = 7; three female and four male MNs; N = 7), corresponding to ∼18% of the total SOD MNs (7/39), with most of the remainder being dual (∼54%, n = 21; eight female and 13 male MNs; N = 16) or inhibited MNs (∼28%, n = 11; seven female and four male MNs; N = 10) (Fig. 1D) (P < 0.0001, chi-squared test). Therefore, although our electrophysiological data did not confirm the exclusivity of a dual dGPSP effect in SOD MNs because some WT MNs could also be a dual phenotype, they did show that fetal E17.5 SOD MNs were more excited by low frequency repetitive dGPSPs than WT MNs from the same littermates, with some of them being unable to switch to the inhibited MN phenotype when dGPSP frequency attained >40 Hz. This propensity to be excited by low frequency repetitive dGPSPs was not a result of the MN resting membrane potential (E_Rest), spike amplitude (Spike Amp.), spike width or the difference between spike threshold and membrane potential (V_thr – V_r ), which were all of similar values between WT dual, WT inhibited, SOD dual, SOD inhibited and SOD excited MNs (Table 1). It was also not a result of the initial firing frequency of MNs because this was not significantly different between the five groups (Fig. 1Ea and Table 1). In each experiment, VLF intensity was adjusted to evoke dGPSPs with amplitudes corresponding to those of spontaneous dGPSPs (Fig. 1A–C, upper traces). We found that the VLF stimulus threshold intensity was significantly higher (P < 0.0001, Mann–Whitney test) in SOD MNs (14.8 ± 7.3 μA, n = 39) compared to WT MNs (8.3 ± 5.8 μA, n = 32) (Fig. 1Eb; for individual comparisons, see Table 1). Even though the theoretical ECl was set to −57.5 mV, the actual ECl was measured for each MN by determining, in voltage clamp mode, the reversal potential of single VLF-evoked IPSCs (Fig. 1Fa and Fb). We did not find any significant difference between the five groups (Fig. 1Fc and Table 1). The slope of the current–voltage curves deriving from actual ECl assessment provides the IPSC conductance (gIPSC) that was divided by the membrane capacitance (gIPSC/C_m). No statistical differences were found between WT and SOD MN groups (Fig. 1Fd) (Table 1), indicating an equal contribution of GABA_A/glycine receptors when the VLF intensity was adjusted to obtain comparable dGPSPs (or IPSCs). In conclusion, assessing ex vivo the effect of repetitive dGPSPs on the firing ability of prenatal E17.5 MNs revealed that SOD1^G93A MNs were lacking an inhibited response compared to WT littermate MNs. We next aimed to assess, using non-correlated parameters, whether the two groups identified amongst MNs from WT embryos and the three MN groups identified in SOD embryos expressed distinguishing properties.

PCA of the different subgroups of SOD and WT MN responses

As described above, depending on their response to VLF stimulation, E17.5 MNs were classified as excited (MNs excited at all tested VLF stimulation frequencies), dual (MNs excited at low VLF frequencies and then inhibited with higher VLF stimulation frequencies), and inhibited (MNs that were exclusively inhibited by VLF discharges). Five non-correlated parameters were used to characterize these responses: percentage of variation of the number of spikes produced by the MN during the one second VLF stimulation at 30 Hz, compared to control spiking discharge at the same time (SpNbCH30); MN input resistance (R_in); Rheobase (lowest value of injected current able to elicit an action potential; voltage difference between threshold potential for spikes and membrane potential (V_thr – V_r); and spike width (SpkWdth). The first parameter corresponded to the cut-off VLF frequency at which dSPSPs were either excitatory or inhibitory, whereas the remaining four parameters were used to discriminate mouse postnatal day 6–10 delayed MNs from immediate MNs, most probably motoneurons innervating future fast-twitch muscle-fibers (F-type MNs) and future slow-twitch fibres (S-type MNs), respectively (Leroy et al., 2014; Sharples & Miles, 2021).

The data from of the PCA are shown in Fig. 2A for WT MNs and Fig. 2B for SOD MNs. The contribution of variables to the first and second components (Tables 3 and 4) indicates that the first component of the PCA (horizontal axis) mostly represented R_in and Rheobase (in red) in both SOD and WT MNs. Spike width (SpkWdth) was also a determinant in the first component for SOD MNs but not for WT MNs. Using a similar method, the second component was interpreted as representing the change in MN discharge (number of spikes during VLF stimulation at 30 Hz expressed as % of control: SpNbCh30) and the spike width, in both SOD and WT MNs (in green).

Figure 2. PCA of the subgroups of MNs made on the basis of their d-responses to VLF inhibitory stimulation

A, in SOD MNs, three groups of MNs were tested: excited, dual and inhibited (for explanations, see text). PCA is presented in (Aa); results of Montecarlo tests to simulate 1000 inertia (see histograms) (Ab). Real inertia value is presented as a red vertical line. The P value was then calculated as the ratio of the number of simulations in which the simulated inertia was larger than the real inertia to the total number of runs. The top histogram represents the inertia calculated for all groups. The three smaller histograms represent inertias calculated for pairs of groups. B, in WT MNs, PCA was made for the two dual and inhibited groups (Ba); the Montecarlo test indicates that these two groups are significantly different (Bb). BCA, between class analysis. [Colour figure can be viewed at wileyonlinelibrary.com]

Table 3. SOD MNs: contributions to PCA first and second component variance

	Component 1	Component 2
SpNbCh30	1421	4177*
Rin	2797*	1294
Rheobase	3190*	1044
Vthr – Vr	504	1343
SpkWdth	2088*	2142*
Mean	2000	2000

Table 4. WT MNs: contributions to PCA first and second component variance

	Component 1	Component 2
SpNbCh30	403	4494*
Rin	4142*	402
Rheobase	4480*	40
Vthr – Vr	274	1816
SpkWdth	700	3248*
Mean	1999.8	2000

To estimate whether the probability groups were distinct, the separation between pairs of groups was evaluated by calculating the inertia, which was defined as the ratio of the between-group variance to the global variance (e.g. between class analysis, BCA). The statistical significance of inertia for group separation was estimated using a Monte Carlo permutation test (1000 runs) (see Methods). For WT MNs, the dual group was significantly different from the inhibited group (P = 0.001, Fig. 2Ab and Table 5). For SOD MNs, the excited and dual groups were not found to be statistically different (P = 0.213), whereas the inhibited group was significantly different from the dual (P = 0.023) and excited (P = 0.013) groups (Fig. 2Bb and Table 5). On this basis, therefore, the dual and excited groups were combined in subsequent analyses. Altogether, our PCA/BCA analysis confirmed that WT E17.5 MNs could be classified into two groups according to their dGPSP responses, whereas the three groups of SOD MNs, as in WT, could be reduced to two groups. In the subsequent part of the study, only dual and inhibited groups were considered for both WT and SOD MNs.

Table 5. P values from Montecarlo test after PCA

SOD MNs			WT MNs
Groups	Excited	Dual	Dual
Dual	0.213
Inhibited	0.013	0.023	0.001

Low input resistance SOD MNs are not specifically inhibited

We previously showed that SOD E17.5 spinal lumbar MNs have a higher R_in than WT MNs from the same littermates (Branchereau et al., 2019; Martin et al., 2013). Interestingly, our present data revealed a similar finding: R_in was 101.0 ± 46.4 MΩ (n = 32; 18 female and 14 male MNs; N = 24) for WT MNs and 132.4 ± 58.1 MΩ (n = 39; 18 female and 21 male MNs; N = 33) for SOD MNs (P = 0.0194, Mann–Whitney test) (Fig. 3Aa). Rheobase, the minimum current injection to elicit spiking activity, was lower, even though not significant (P = 0.2552), in SOD MNs (193.9 ± 131.3 pA, n = 39) compared to WT MNs (227.6 ± 143.1 pA, n = 32), in agreement with the former's higher R_in values (Fig. 3Ab). The between group analysis (Kruskal–Wallis test followed by Dunn's comparisons test) showed that the R_in of WT inhibited MNs was significantly reduced compared to the R_in of dual (P = 0.0098) and inhibited (P = 0.0306) SOD MNs (Fig. 3B), and a similar trend, although not significant (P = 0.0534), was found when compared to the R_in of dual WT MNs. Plots of rheobase with R_in values revealed an expected relationship for WT and SOD MNs (i.e. MNs with low R_in express a high rheobase and vice versa) (Fig. 3Ca and Cb). However, it also showed that most R_in values were below 100 MΩ for inhibited MNs (Fig. 3Ca), which was not the case for SOD MNs that were dual or inhibited, regardless of their R_in values (Fig. 3Cb). The proportion of inhibited MNs with R_in below 100 MΩ was ∼90% (2/19) in WT MNs and only 45% in SOD MNs (5/11) (P < 0.0001, chi-squared test) (Fig. 3Cc). In conclusion, rheobase = f(R_in) plots revealed that low R_in WT MNs (i.e. putative future F-MNs) were preferentially inhibited by repetitive dGPSPs, in contrast to low R_in SOD MNs that could be excited.

Figure 3. WT inhibited MNs have a low R_in, but not SOD MNs

Aa, SOD MNs (18 female, 21 male, N = 33 litters) have lower R_in values compared to WT MNs (18 female, 14 male, N = 24 litters). *P = 0.0194, Mann–Whitney test. A2, the trend, albeit not statistically significant, is also toward a lower rheobase in SOD MNs. B, RIN is low exclusively in WT inhibited MNs. *P = 0.0306, **P = 0.0098, Kruskal–Wallis test followed by Dunn's comparisons test. Ca and Cb, rheobase values are inversely proportional to R_in values in WT and SOD MNs. Unlike SOD MNs (Cb), most inhibited WT MNs (Ca) have values <100 MΩ (Ca). Cc, percentage of inhibited MNs <100 MΩ (bright blue) in both genotypes. ****P < 0.0001, chi-squared test. [Colour figure can be viewed at wileyonlinelibrary.com]

Impact of ECl, distal dendritic length and synapse position on dGPSP bursts: a simulation analysis

In a previous study, we have used computer simulations to show how ECl interacts with V_rest and other MN parameters (morphology, synaptic conductance, etc.) to produce mixed inhibitory and excitatory responses to single PSPs (Branchereau et al., 2016). Here, we addressed this question more precisely using the same protocol as that used in physiological experiments. Basically, the model MN received a continuous amount of constant depolarizing current to produce a stable 12 Hz discharge (Fig. 4Ab). Superimposed on this basal activation, a train of dGPSPs (corresponding to GABA/glycine inputs) was elicited at a range of frequencies from 0.2 to 300 Hz. When ECl was around −50 mV, a small variation in ECl could produce opposite effects. For example, in a WT MN, when its ECl was set to −47.4 mV (as in physiological experiments) a 20 Hz GABA/glycine fibre stimulation evoked an excitatory response that increased with frequency up to 166 Hz and then decreased and became inhibitory for GABA/glycine stimulation frequency >300 Hz (Fig. 4Aa, red symbols, illustrated up to 200 Hz). A modest decrease of ECl (to −52.1 mV) led to a total absence of an excitatory effect from 20 Hz and above. The MN discharge was even totally blocked with a GABA/glycine stimulation frequency of 50 Hz (and above) (Fig. 4Aa, blue symbols, and Fig. 4Ab).

Figure 4. Simulation of the effects of ECl, distal dendritic length and synapse position on DGPSP burst

Aa and Ab, comparison of the effects of two ECl values: −47.4 mV (red circles) and −52.1 mV (light blue circles) (Aa) for a WT MN [for characteristics of WT and ‘cut ends’ (CE) MNs, see Methods]. Raw data are presented to illustrate the procedure (A2). A constant current is injected into the MN (200 pA) so that its discharge is ∼12 Hz. GABA synapse stimulation starts at t = 900 ms. B, comparison of the responses presented in (A) for WT MNs, with CE MNs (40% decrease in distal dendrite length, see Methods). Ba, simulation results. Bb, shape of the dendritic trees of WT and CE simulated MNs. C, comparison of the effect of GABA synapse stimulation with two different GABA synapse (syn) locations: on the soma (Ca) and at 100 μm from soma on a dendrite (Cb). Cc, illustration of the geometric disposition. The green pipette symbolizes a recording made at the soma level of the simulated MNs. The default MN soma diameter was 11.17 μm. [Colour figure can be viewed at wileyonlinelibrary.com]

By comparison, the effect of dendrite length was less prominent (Fig. 4Ba and Bb). In this new simulation, we used WT-like MNs and SOD-like MNs. SOD-like MNs had the same characteristics as WT-like MNs, except for terminal dendrites that were 40% shorter than those of the WT model, as observed in SOD MNs (Martin et al., 2013). These SOD model MNs were therefore called ‘cut ends’ or CE MNs. This reduced length of their terminal dendrites increased their input resistance. Interestingly, this morphological change increased the excitability of these CE MNs in such a way that, at low GABA/glycine stimulation frequencies (<50 Hz), CE MNs expressed an excitatory response for ECl values of −47.4 mV and −52.1 mV (Fig. 4Ba), transforming the previously inhibited response (Fig. 4Aa, light blue trace) into a dual response (Fig. 4Ba, dark blue trace). For CE MNs, with ECl = −52.1 mV, this response switched to an inhibition for GABA/glycine stimulation frequency >70 Hz (Fig. 4Ba, dark blue trace), whereas with ECl = −47.4 mV, the excitatory response to GABA/glycine stimulation was larger than for WT MNs.

In the last series of simulations, we aimed assess the effect of dGPSP synapse position on the response of SOD MNs to GABA/glycine stimulation (Fig. 4Ca–Cc), estimated for variations in the dGPSP synapse conductance. The dGPSP synapse was moved from the soma (x = 0 μm) to a proximal dendrite at x = 100 μm from the soma (Fig. 4Cc). This modest displacement had dramatic consequences on the effect of dGPSP trains. A dGPSP synapse in soma position x = 0 μm induced a mixed excitatory (at low GABA/glycine stimulation frequencies) and inhibitory response (at high frequencies) (Fig. 4Ca), regardless of the dGPSP conductance (12–20 nS), whereas solely excitatory responses were elicited by dGPSP trains when the dGPSP synapse was positioned on the dendrite (x = 100 μm) (Fig. 4Cb).

In conclusion, our simulation data highlighted the major role of ECl value in the polarity of dGPSPs effects, as well as the excitatory effect arising from a morphological alteration (i.e. shorter dendritic tree) found in most SOD MNs. Modelling revealed that reducing the chloride conductance of dGPSPs occurring on the soma strongly favours excitation and mimics the overexcitation found in some SOD MNs by moving dGPSP input away from the soma. Because, altogether, our simulation data suggested that SOD MNs are lacking peri-somatic inhibitory inputs, we therefore conducted an immunohistochemical study aiming to quantify the actual amount of peri-somatic GABA/glycine terminals in SOD vs. WT MNs.

SOD MNs receive less inhibitory inputs

Confocal imaging was performed on MNs that were filled with neurobiotin (Fig. 5, in blue) during the previous recording session and processed for VIAAT immunostaining to detect presynaptic GABA/glycine terminals on the soma and proximal dendrites. Inspection of the cytoplasmic membrane and its close proximity revealed a reduced VIAAT staining on SOD MNs, relative to age-matched WT MNs (Fig. 5A and B). This was confirmed by a quantitative analysis of global VIAAT (Fig. 5C) both on the soma and proximal dendrite: 97.2 ± 14.7/100 μm for WT vs. 66.0 ± 23.1/100 μm for SOD MN somata (P < 0.0001, Mann–Whitney test), 90.4 ± 25.7/100 μm for WT vs. 70.0 ± 31.4/100 μm for SOD proximal dendrites (P < 0.0001, Mann–Whitney test). This similarity in distribution on the soma membrane and proximal dendrite membrane is in agreement with the literature, where the density of GAD65/67 immunoreactive synaptic boutons on neonatal rat MNs has been found to be similarly distributed on the soma and on dendrites located up to 100 μm away (Jean-Xavier et al., 2007). Here, the proximal dendrite length analysed was 28.5 ± 11.7 μm and 24.6 ± 12.6 μm for WT and SOD MNs, respectively, with the GAD65/67 synaptic density diminishing on the most distal dendrites (200–250 μm) (Jean-Xavier et al., 2007). To verify whether the lowered VIAAT innervation of mouse SOD MNs, compared to WT littermate MNs, could impact the responsiveness to motoneuronal inputs, we analysed synaptic GABA/glycine activity in SOD vs. WT MNs. Interestingly, we found that the frequency of GABA/glycine synaptic events (for representative traces see Fig. 5Da; for related analysis, see Fig. 5Db and Dc) was significantly diminished in SOD E17.5 MNs compared to littermates WT MNs (Fig. 5E) (5.44 ± 2.16 Hz for WT MNs vs. 4.33 ± 1.69 Hz for SOD MNs, P = 0.0251, Mann–Whitney test), without any change in amplitude (Fig. 5F) (0.43 ± 0.16 Hz for WT vs. 0.45 ± 0.10 Hz for SOD, P = 0.2335, Mann–Whitney test). Therefore, our results from confocal imaging, as well as synaptic transmission analysis, provided further evidence to support the conclusion that SOD1^G93A MNs lack peri-somatic GABA/glycine inputs, as predicted from our simulation data.

Figure 5. SOD MNs lack VIAAT terminals on their soma and proximal dendrite, and exhibit a lower frequency of spontaneous dGPSPs

A and B, confocal images (63×) of a neurobiotin-injected soma (blue), VIAAT immuno-staining (green) and after merging of E17.5 WT (A) and SOD MNs (B). An example of band image analysis is presented for the WT and SOD MNs as row images (Aaa to Aca and Baa to Bca, respectively) and detected particales (Abb and Bbb, respectively). C, global VIAAT density on the soma and proximal dendrite, expressed as number per 100 μm. Da, representative raw traces from a WT and SOD MN illustrating the detection of GABA/glycine synaptic events. Db and Dc, corresponding frequency and amplitude analysis. GABA/glycine events frequency (E) and amplitude (F) in all analysed WT and SOD MNs. Quartiles and medians are indicated in the violon plots as white dotted and hatched lines, respectively. *P = 0.0251; ****P < 0.0001, Mann–Whitney test. [Colour figure can be viewed at wileyonlinelibrary.com]

Discussion

We have investigated ex vivo the effect of repetitive dGPSPs occurring at various physiologically relevant frequencies (Branchereau et al., 2019) on the repetitive firing ability of fetal E17.5 lumbar spinal MNs from the SOD1^G93A ALS mouse model in comparison with WT MNs from the same littermates. Our electrophysiological experimental approach allowed us to disclose two types of MNs: dual MNs exhibiting increased MN firing (excitation) at dGPSP frequency <50 Hz, but decreased MN firing (inhibition) at a frequency >50 Hz, and inhibited MNs showing decreased MN firing (inhibition) at all tested dGPSP frequencies. We found that 60% of WT MNs were inhibited MNs, exhibiting an R_in <100 MΩ, whereas 72% of SOD MNs were dual MNs, without any relationship with their R_in values. A reduced density of GABA/glycine terminals on the soma and proximal dendrite of SOD MNs accounted for the propensity of SOD MNs to be excited by dGPSPs occurring at a low frequency, as expected by the results of simulations.

Functional significance of the different subtypes of fetal MNs

We previously reported that prenatal E17.5 lumbar spinal MNs in the SOD1^G93A ALS mouse model are hyperexcitable because of their shorter dendrites and exhibit higher R_in values compared to littermate WT MNs (Martin et al., 2013). Our present data, obtained from similar fetal MNs located in the lateral motor column (Branchereau et al., 2019), not only confirm the high R_in of SOD MNs, but also show that some SOD MNs have a low R_in, whereas some fetal WT MNs have a high R_in. Does this correspond to two functional types of MNs, namely future large fast-twitch, fast-fatigable (FF) motor neurons that degenerate before slow (S) motor neurons appear (Nijssen et al., 2017)? Leroy et al. (2014) have shown, in P6–10-day-old SOD1^G93A mice, that only S-type (innervating the slow-contracting muscle fibres) α MNs display intrinsic hyperexcitability and a shrunken dendritic tree, whereas the excitability of F-type (innervating the fast-contracting muscle fibres) α MNs remains unchanged. Similarly, Venugopal et al. (2015) found in P8-P12 SOD1^G93A mice that predicted fast SOD trigeminal MNs showed hyperexcitable shifts marked by a reduced rheobase and an increased input resistance, whereas predicted slow SOD MNs did not display significant alterations in these properties compared to WT littermate MNs. Leroy et al. (2014) reported an immediate firing capability in S-type MNs and a delayed firing in F-type MNs. S-type were also identified as expressing the oestrogen-related receptor β (Errβ) but not matrix metalloproteinase-9 (MMP9), whereas most of the F-type were enriched with MMP9 and were Errβ negative. MMP-9 is not expressed by embryonic MNs and is first detected in the spinal cord at around P5 (Kaplan et al., 2014). We aimed to detect MMP9 in neurobiotin-injected fetal MNs but could not detect any staining. We also checked immediate vs. delayed firing ability when MNs were depolarized to spike threshold with a 5 s long depolarizing pulse, but found only immediate firing. In addition, the voltage difference between the threshold potential for spiking and resting membrane potential (V_thr – V_r), as well as spike width (SpkWdth), did not differ amongst E17.5 MNs (Table 1), unlike P6–10 MNs (Leroy et al., 2014). However, despite the difficulty in defining the future identity of our recorded fetal MNs, low R_in (<100 MΩ) WT MNs, inhibited by dGPSPs barrages, most probably correspond to future F-type MNs. Their morphology also complies with that described for WT MNs in our previous study (Martin et al., 2013).

Motoneuronal properties

Our data did not reveal any differences in MN E_Rest, spike amplitude and spike width between E17.5 SOD1^G93A and WT littermate mice. No difference in E_Rest and action potential shape was also reported between G93A and control MNs from spinal cord slices cultured over 3 weeks, starting at E12–14 (Kuo et al., 2004). P6–10 SOD1^G93A spinal MNs also did not show any change in E_Rest, spike amplitude and spike width (Leroy et al., 2014), whereas the E_Rest of P30–60 SOD1^G93A spinal MNs was found to be 10 mV more hyperpolarized than WT littermate MNs, although this difference in E_Rest becomes undetectable at P90–120 (Huh et al., 2021). The hyperpolarization of MN E_Rest is therefore not apparent at early and late stages of the disease in SOD1^G93A animals. Such a hyperpolarization was proposed to serve as a cellular adjustment that compensates for an increase in the persistent inward currents (PICs), as mediated by voltage-gated Na⁺ and Ca²⁺ channels, in postnatal SOD MNs (Huh et al., 2021), along with excessive dendritic elongation and overbranching (Amendola & Durand, 2008; Filipchuk & Durand, 2012), possibly ensuing from early morphofunctional alterations in embryonic MNs (shorter dendrites, impaired inhibition). Interestingly, an overall pattern of oscillations across the lifespan of the SOD1^G93A ALS mouse model was described, with the PIC oscillations tending to increase excitability, but with the membrane conductance and E_Rest oscillations tending to reduce it (Huh et al., 2021). An hyperexcitability of SOD1^G93A MNs mediated by excessive activity of voltage-gated Na⁺ and Ca²⁺ channels (Quinlan et al., 2011) is also counteracted by aberrant increases in cell size and conductance in juvenile P50 MNs, with this homeostatic gain being excessive in SOD1^G93A MNs (Dukkipati et al., 2018; Martin et al., 2007) and ultimately lethal (Kuo et al., 2020).

Prenatal SOD motoneurons receive less VIAAT-positive inputs

Our results show that fetal SOD MNs can be dual or inhibited MNs, regardless of their R_in, unlike WT MNs from the same littermates. A possible explanation for this initially surprising result is that inhibited low R_in WT MNs are future F-type (innervating fast-contracting muscle fibres) α MNs that start becoming mature at E17.5. Then, it is possible that the normal developmental sequence of the lumbar motoneuronal networks [decrease in MN R_in, increase in C_m (Delpy et al., 2008), increase in MN synaptic inputs (Scain et al., 2010)] is delayed in the SOD1^G93A strain compared to WT littermates. Another explanation is that the lack of GABA/glycine inputs impedes the occurrence of inhibited low R_in SOD MNs. We have shown that increasing gGABA/glycine favours dGPSP-mediated inhibition occurring on SOD MN somata (Fig. 4Da). The role of gGABA/glycine (gCl) in the dual effect of dGPSPs was also demonstrated in a previous study where increasing gCl was found to favour inhibition, both during a single dGPSP or during trains in which gCl summates (Branchereau et al., 2016). Our immunohistochemical data also provided evidence for a lower density of VIAAT staining on MN somata (and proximal dendrites) from the SOD1^G93A strain compared to WT littermate MNs. Consequently, VLF stimulation would activate less GABA/glycine inputs, thereby lowering the shunting (inhibitory effect) of dGPSPs. Our physiological experiments showed that an increase in the intensity of VLF stimulation in SOD preparations was required to evoke single dGPSPs of similar size as in WT preparations (Fig. 1Eb), in agreement with our anatomical data indicating a lower density of VIAAT positive terminals on SOD MNs. A reduction of inhibitory synapses was not found in SOD1^G93A E12.5 spinal cords maintained in culture during 14 days, but rather an increase was reported (higher ratio of inhibitory vs. excitatory synapses) (Avossa et al., 2006). However, F-type MNs from P45 SOD1^G93A mice have been recently reported to lose inhibitory glycinergic connections before S-type MNs (Allodi et al., 2021). A major reduction in inhibitory synapse densities (gephyrin puncta) in the ventral horn of the SOD1^G93A slow mutant was also described at early stages of the disease (P50) (Saxena et al., 2013), with these changes probably being linked to cholinergic dysfunction in local spinal cord circuitry (Casas et al., 2013; Wootz et al., 2013). Our previous study showed that VIAAT positive terminals are not less abundant in the marginal zone edging the soma location of E17.5 MNs (Branchereau et al., 2019). Here, we found a reduction in VIAAT density on the soma and proximal dendrites of SOD MNs, relative to age-matched WT. This indicates a selective reduction of GABA/glycine inputs to MN cell bodies that probably accounts for the lack of an inhibitory effect of dGPSP barrages in prenatal SOD^G93A MNs and for the reduction in the frequency of GABA/glycine synaptic events in SOD MNs. This selective reduction of GABA/glycine inputs may start around birth and increase over time in the adult.

Effect of recurring dGPSPs on MN firing: a complex equation

The present study highlights the effect of recurring dGPSPs on the repetitive firing capability of E17.5 lumbar MNs that drive the contraction of muscle fibres, constituting the final physiological pathway (Manuel & Zytnicki, 2011). We show that an excitatory action of low frequency dGPSPs is predominant in SOD1^G93A MNs, which can be explained by the lower density and distribution of GABA/glycine synaptic inputs on SOD MNs compared to WT MNs. Simulations also indicated that morphological changes in SOD MNs (i.e. shorter dendritic tree) favour the excitatory action (Fig. 4B). Additional parameters are undoubtedly governing the effect of repetitive dGPSPs on MN firing. Clearly, the ECl value will control this effect, and a slight hyperpolarization of ECl will enable switching the effect from pure excitation to pure inhibition (Fig. 4A). A shorter decay time of dGPSPs also enhances their excitatory effect on MN firing as we previously showed (Branchereau et al., 2019) because it leads to a lower capacity for shunting effect of dGPSP summation (Branchereau et al., 2016). The decay of dGPSPs is closely dependent on [Cl⁻]_i (Houston et al., 2009; Pitt et al., 2008) as a result of a direct effect of chloride ions acting in the pore of glycine and GABA channels (Moroni et al., 2011). Here, because [Cl⁻]_i was set at similar values in SOD and WT MNs, the decay time of dGPSPs was maintained similarly in all recorded MNs. We have previously shown that E17.5 SOD1^G93A MNs exhibit dGPSPs with an increased decay time compared to WT littermate MNs (Branchereau et al., 2019) and this change could compensate for the more depolarized ECl value found in SOD MNs. Under physiological conditions, this compensatory change may limit the overexcitatory action of repetitive dGPSPs in SOD MNs demonstrated in the present study. However, the lack of inhibitory GABA/Gly inputs to SOD MNs and their morphological alteration (Martin et al., 2013) probably renders them vulnerable.

Limitations of the study

In the present study, we did not analyse the lack of VIAAT staining on F-type vs. S-type MNs because of the lack of available specific markers at fetal stages (see Discussion). Therefore, we could not assess whether any specific early reduction in GABA/glycine terminals on F-type MNs exists at the prenatal stages. We also did not look at GABA_AR and GlyR on spinal MNs. A decrease in surface postsynaptic GlyR on SOD1^G93A MNs maintained in culture from E12.5 during 12–16 days has been previously reported (Chang & Martin, 2011). A possible reduction of GlyR on SOD MNs may also account for the higher excitability of dGPSPs in our experiments. Finally, the main limitation of our study derives from the fact that MNs were recorded in whole-cell configuration with similar ECl values. Because physiological ECl is 10 mV more depolarized in E17.5 SOD1^G93A MNs compared to WT littermate MNs (Branchereau et al., 2019), a preferable approach in future experiments will be to test the effect of dGPSP barrages on MN spiking activity using perforated patch clamp recordings that preserve physiological ECl.

Biography

• 

  Hongmei Zhu received her Master's degree in Neurobiology from joint laboratories of Monash University and Kunming Medical University, China. She was awarded a PhD scholarship from the China Scholarship Council (CSC) and joined the Institut de Neurosciences Cognitives et Intégratives d'Aquitaine (INCIA), Bordeaux University, France, in 2018, where she obtained her PhD in Neurosciences under the supervision of Professor Pascal Branchereau. She now works as a research fellow in the Shaoxing Institute, Zhejiang University, China.

Open Research

Data availability statement

The authors state that all data supporting the results are in the manuscript, figures and tables.

Competing interests

The authors declare that they have no competing interests.

Author contributions

H.Z., U.D. and P.B. performed electrophysiological experiments and analyses. P.B. conceived the experiments and wrote the manuscript. H.Z. performed immunohistochemistry, confocal imaging and amended the manuscript. D.C. performed some analyses, the computer simulations and helped with revision of the manuscript.

Funding

Funding was provided by Association pour la Recherche sur la Sclérose Latérale Amyotrophique et autres Maladies du Motoneurone (ARSLA): Pascal Branchereau, AAP2018; AFM-Telethon: Pascal Branchereau, 23185; Framework of the University of Bordeaux's IdEx ‘Investments for the Future’ program/GPR BRAIN_2030.

Acknowledgements

H.Z. was helped by the China Scholarship Council (CSC). We thank Nathalie Argenta for excellent technical assistance and animal breeding. The help of Gilles Courtand with image analyses is also acknowledged. We thank the ‘Association pour la recherche sur la Sclérose Latérale Amyotrophique et autres Maladies du Motoneurone‘ (ARSLA), as well as AFM-Téléthon for their financial support. This study also received financial support from the French government within the framework of the University of Bordeaux's IdEx ‘Investments for the Future’ program/GPR BRAIN_2030.

Additional information