Functional up‐regulation of the M‐current by retigabine contrasts hyperexcitability and excitotoxicity on rat hypoglossal motoneurons

Abstract

Key points

• Excessive neuronal excitability characterizes several neuropathological conditions, including neurodegenerative diseases such as amyotrophic lateral sclerosis.

• Hypoglossal motoneurons (HMs), which control tongue muscles, are extremely vulnerable to this disease and undergo damage and death when exposed to an excessive glutamate extracellular concentration that causes excitotoxicity.

• Our laboratory devised an in vitro model of excitotoxicity obtained by pharmacological blockade of glutamate transporters. In this paradigm, HMs display hyperexcitability, collective bursting and eventually cell death.

• The results of the present study show that pharmacological up‐regulation of a K⁺ current (M‐current), via application of the anti‐convulsant retigabine, prevented all hallmarks of HM excitotoxicity, comprising bursting, generation of reactive oxygen species, expression of toxic markers and cell death.

• ○Our data may have translational value to develop new treatments against neurological diseases by using positive pharmacological modulators of the M‐current.

Abstract

Neuronal hyperexcitability is a symptom characterizing several neurodegenerative disorders, including amyotrophic lateral sclerosis (ALS). In the ALS bulbar form, hypoglossal motoneurons (HMs) are an early target for neurodegeneration because of their high vulnerability to metabolic insults. In recent years, our laboratory has developed an in vitro model of a brainstem slice comprising the hypoglossal nucleus in which HM neurodegeneration is achieved by blocking glutamate clearance with dl‐threo‐β‐benzyloxyaspartate (TBOA), thus leading to delayed excitotoxicity. During this process, HMs display a set of hallmarks such as hyperexcitability (and network bursting), reactive oxygen species (ROS) generation and, finally, cell death. The present study aimed to investigate whether blocking early hyperexcitability and bursting with the anti‐convulsant drug retigabine was sufficient to achieve neuroprotection against excitotoxicity. Retigabine is a selective positive allosteric modulator of the M‐current (I _M), an endogenous mechanism that neurons (comprising HMs) express to dampen excitability. Retigabine (10 μm; co‐applied with TBOA) contrasted ROS generation, release of endogenous toxic factors into the HM cytoplasm and excitotoxicity‐induced HM death. Electrophysiological experiments showed that retigabine readily contrasted and arrested bursting evoked by TBOA administration. Because neuronal I _M subunits (Kv7.2, Kv7.3 and Kv7.5) were expressed in the hypoglossal nucleus and in functionally connected medullary nuclei, we suggest that they were responsible for the strong reduction in network excitability, a potent phenomenon for achieving neuroprotection against TBOA‐induced excitotoxicity. The results of the present study may have translational value for testing novel positive pharmacological modulators of the I _M under pathological conditions (including neurodegenerative disorders) characterized by excessive neuronal excitability.

Key points

• Excessive neuronal excitability characterizes several neuropathological conditions, including neurodegenerative diseases such as amyotrophic lateral sclerosis.

• Hypoglossal motoneurons (HMs), which control tongue muscles, are extremely vulnerable to this disease and undergo damage and death when exposed to an excessive glutamate extracellular concentration that causes excitotoxicity.

• Our laboratory devised an in vitro model of excitotoxicity obtained by pharmacological blockade of glutamate transporters. In this paradigm, HMs display hyperexcitability, collective bursting and eventually cell death.

• The results of the present study show that pharmacological up‐regulation of a K⁺ current (M‐current), via application of the anti‐convulsant retigabine, prevented all hallmarks of HM excitotoxicity, comprising bursting, generation of reactive oxygen species, expression of toxic markers and cell death.

• ○Our data may have translational value to develop new treatments against neurological diseases by using positive pharmacological modulators of the M‐current.

Introduction

Neuronal hyperexcitability is a common symptom in many pathological conditions affecting the CNS, comprising several neurodegenerative diseases, such as amyotrophic lateral sclerosis (ALS) (Geevasinga et al. 2016; Park et al. 2017). In the latter case, hyperexcitability of motoneurons in the brainstem and spinal cord often precedes their neurodegeneration and results in classical symptoms such as muscle fasciculation and atrophy (Kiernan et al. 2011). In the brainstem, hypoglossal motoneurons (HMs) are an early target for neurodegeneration, particularly in the bulbar form of ALS, in which patients show severe dysarthria and dysphagia (Cifra et al. 2011a). HMs are the main cell type in the hypoglossal nucleus (Viana et al. 1990; Cifra et al. 2012) and exclusively innervate tongue muscles to carry out inspiration, vocalization, chewing and swallowing (Lowe, 1980). This class of motoneurons is extremely vulnerable to a variety of insults (von Lewinski & Keller, 2005) because of their intrinsic properties, such as low Ca²⁺ buffering ability (Ladewig et al. 2003), expression of Ca²⁺‐permeable AMPA receptors (Laslo et al. 2001) and low ability with respect to glutamate clearance (Rothstein et al. 1992). In ALS, enhanced excitability of HMs reflects an excitotoxic state developed within the nucleus hypoglossus as a result of impaired glutamate uptake (Van Den Bosch et al. 2006), which causes over‐activation of glutamate receptors. In agreement with the excitotoxic theory of ALS, increased glutamate levels in the cerebrospinal fluid of patients (Spreux‐Varoquaux et al. 2002) and impaired glutamate transport (Rothstein et al. 1992) have been reported. Excitotoxicity triggers a variety of pathological mechanisms eventually leading to cell death (Ngo & Steyn, 2015). Drugs that are able to contrast hyperexcitability may therefore be important for protecting HMs during the early stages of ALS.

To mimic the impact of ALS on the hypoglossal nucleus, our laboratory has developed an in vitro model of excitotoxicity using rat brainstem slices (Sharifullina & Nistri, 2006). The model exploits the endogenous glutamate release that leads to an excitotoxic state after the administration of dl‐threo‐β‐benzyloxyaspartate (TBOA), a competitive, non‐transportable blocker of excitatory amino acid transporters (EAATs) (Shigeri et al. 2004). The increased glutamate level in the synaptic cleft triggers a well‐known set of pathophysiological mechanisms leading to HM death, such as electrophysiological bursting activity (Sharifullina & Nistri, 2006; Cifra et al. 2011b; Corsini et al. 2016; Ghezzi et al. 2017b), Ca²⁺ overload (Sharifullina & Nistri, 2006; Corsini et al. 2017), oxidative stress, mitochondrial damage (Tortora et al. 2017) and an unfolded protein response (Corsini et al. 2016), which have been previously characterized by our laboratory. These processes are effectively prevented by a few pharmacological agents such as riluzole (Cifra et al. 2011b), nicotine (Corsini et al. 2016, 2017) and propofol (Ghezzi et al. 2017b), which, although acting on different molecular targets, share the ability to reduce HM hyperexcitability and inhibit bursting.

Among the endogenous mechanisms that neurons naturally express to prevent hyperexcitability, the M‐current (I _M) is widespread in the CNS (Delmas & Brown, 2005; Greene & Hoshi, 2016). I _M is a voltage‐dependent, non‐inactivating K⁺ current activating in the subthreshold voltage range where it provides a persistent outward current (Adams et al. 1982; Delmas & Brown, 2005) that contrasts spiking activity (Zaika et al. 2006). Homo‐ or hetero‐tetrameric assemblies of Kv7.2, Kv7.3 and Kv7.5 K⁺ channel subunits (encoded by the KCNQ2, KCNQ3 and KCNQ5 genes) underlie the neuronal I _M (Wang et al. 1998; Schroeder et al. 2000): these channels are localized in various cell compartments, such as the axon initial segment (AIS), nodes of Ranvier (Devaux et al. 2004) and somatodendritic compartment (Shah et al. 2002), as well as in presynaptic (Martire et al. 2004) and postsynaptic terminals (Fidzinski et al. 2015). I _M has been demonstrated to be an important regulator of excitability in a variety of neurons (Brown & Passmore, 2009), comprising both hypoglossal (Ghezzi et al. 2017a) and spinal motoneurons (Lombardo & Harrington, 2016). Furthermore, given their widespread distribution, Kv7 channels additionally control network excitability; for example, by regulating the spike‐dependent release of neurotransmitters (Martire et al. 2004; Vervaeke et al. 2006). From a clinical perspective, boosting the I _M through selective drugs, such as the anti‐convulsant retigabine, has been demonstrated to be an adjuvant treatment for drug‐resistant partial seizures (Porter et al. 2007). Retigabine acts as an M‐channel opener because it stabilizes its open conformation, thus shifting the activation curve toward more hyperpolarized potential (Tatulian & Brown, 2003).

We have previously shown that HMs express a functional I _M that is an important modulator of their excitability (Ghezzi et al. 2017a). Because retigabine opens M‐channels that are usually closed at resting state, it therefore induces a potent constraint to HM excitability (Ghezzi et al. 2017a). Thus, we considered whether positive modulation of I _M may have beneficial effects on the in vitro model of excitotoxicity through its ability to dampen excitability. A corollary of the present study aims to explore whether enhancing inhibitory conductances such as I _M can comprise a braking system as effective as blocking hyperactivation of glutamate receptors (Cifra et al. 2011b; Corsini et al. 2016). Although sales of retigabine (Trobalt or Ezogabine; GlaxoSmithKline, London, UK) for clinical use have recently been discontinued because of side effects with respect to the eye and the skin after prolonged treatment, the results of the present study might have translational value given the promising role of the I _M as a possible therapeutic target for a variety of hyperexcitability‐related pathological conditions, comprising neurodegenerative diseases (Barrese et al. 2018).

Methods

Ethical approval

All experiments were performed in accordance with the European Union guidelines (2010/63/UE) and Italian law (decree 26/14) and were approved by the International School for Advanced Studies (SISSA) ethical committee (prot. 3599, 28th May 2012). All efforts were made to reduce the number of animals used for these experiments and their suffering.

Animals and slice preparation

All experimental procedures were performed with an in vitro preparation of brainstem medullar slice containing the hypoglossal nucleus, dissected out from neonatal Wistar rats (postnatal days 2–5), decapitated under i.p. urethane anaesthesia (10% solution, 0.1 mL injection). Slices were prepared in accordance with a protocol described previously (Ghezzi et al. 2017a). Briefly, the whole brainstem was isolated in carbogenated (95% O₂, 5% CO₂) dissection Krebs solution (in mm: 130 NaCl, 3 KCl, 1.5 NaH₂PO₄, 1 CaCl₂, 5 MgCl₂, 25 NaHCO₃ and 11 glucose; pH 7.4; 310–330 mOsm) and then cut into either 270 or 450 μm slices. An average of two slices per animal was first incubated in the same solution as above for 20 min at 32°C and then maintained at room temperature for at least 10 min.

Immunohistochemistry

Immunohistochemistry was performed on 450 μm thick slices as described previously (Corsini et al. 2017). After stabilization, slices were either immediately fixed in 4% paraformaldehyde for 4 h or were incubated (4 h) in either recording Krebs solution (in mm: 130 NaCl, 3 KCl, 1.5 NaH₂PO₄, 1.5 CaCl₂, 1 MgCl₂, 25 NaHCO₃ and 12 glucose; pH 7.4; 300–320 mOsm) alone as the sham condition or supplemented with selected drugs and fixed afterwards. Subsequently, slices were cryoprotected in 30% sucrose for at least 24 h at 4°C. After mounting in killik medium (Bio Optica, Milan, Italy), slices were frozen at –20°C for at least 12 h before being sectioned with a cryostat into 30 μm thick sections and then placed on histology slides. After 3 h of blocking, slides were incubated overnight at 4°C with primary antibodies and then with corresponding secondary antibodies for 2 h at room temperature; all antibodies were diluted in antibody solution (2% normal goat serum, 10% BSA and 1% triton X‐100 in PBS). As previously validated primary antibodies, we employed SMI32 (mouse monoclonal, dilution 1:200; Millipore, Billerica, MA, USA; Campbell & Morrison, 1989) to identify HMs (Cifra et al. 2012) and anti‐NeuN (mouse monoclonal, dilution 1:50; Millipore) to label the nucleus of differentiated neurons (Cifra et al. 2012). The anti‐apoptosis‐inducing factor (AIF) antibody (rabbit monoclonal, dilution 1:200; Millipore) (Kanungo et al. 2008) was used as biomarker of neurotoxicity (Monni et al. 2017). For the detection of Kv7 channel subunits, the primary antibodies employed were: anti‐Kv7.2 (KCNQ2, rabbit polyclonal, dilution 1:400; Alomone, Jerusalem, Israel) (Devaux, 2010), anti‐Kv7.3 (KCNQ3, rabbit polyclonal, dilution 1:100; Alomone) and anti‐Kv7.5 (KCNQ5, rabbit polyclonal, dilution 1:400; Alomone). Primary antibodies were visualized with AlexaFluor 488 and 594 (dilution 1:500; Molecular Probes, Invitrogen, Carlsbad, CA, USA) secondary antibodies. Finally, slides were stained with the DNA dye 4′,6‐diamidino‐2‐phenylindole (DAPI; Sigma‐Aldrich, St Louis, MO, USA; dilution 1:1000 in PBS) and mounted with a fluorescence mounting medium (Fluoromount; Sigma‐Aldrich) to reduce fading.

For the correct detection of anti‐Kv7 channel subunit antibodies, a heat‐induced epitope retrieval method was necessary before the blocking phase. In particular, 30 μm thick slices were treated with glycine 0.3 m in PBS for 30 min and then incubated with 10 mm citrate buffer (pH 6) at 80°C for 30 min. For each Kv7 channel subunit antibody, the specificity control experiment was performed by pre‐absorbing the antibody with the specific control antigen (1 μg peptide per 1 μg antibody) at room temperature for 2 h, in acordance with the manufacturer's instructions. Subsequently, slices were stained in accordance with the canonical immunostaining protocol.

Immunohistochemical images were collected through a C2 confocal microscope (Nikon, Tokyo, Japan), using a 40× (1.30 NA) oil‐objective (oil mounting medium, 1.515 refractive index) and then analysed with a high‐performance 3‐D imaging software (Volocity 6.3; PerkinElmer, Waltham, MA, USA). To evaluate cell death and toxicity after TBOA treatment, SMI32‐immunopositive HMs were counted within a 554.86 mm³ region of interest (ROI) (5 μm Z‐stack, 1 μm step) for each section, whereas the AIF intensity signal was measured within each HM and averaged. For the experiment with anti‐Kv7 channel subunit antibody, SMI32‐positive, NeuN‐positive and Kv7‐positive neurons were counted within the whole imaged field.

A small subset of slices was stained with Toluidine blue (dilution 1:100) immediately after cryostat sectioning and viewed with an upright microscope (DM600; Leica Microsystems, Wetzlar, Germany) with a 5× objective.

Reactive oxygen species (ROS) measurement

ROS production was monitored using a method described previously (Tortora et al. 2017). In brief, 270 μm thick slices were first incubated for 2 h in either sham condition or with selected drugs and then stained with the membrane permeable dye dihydrorhodamine 123 (DHR 123, 5 μM; Molecular Probes) and with the nuclear dye Hoechst 33342 (dilution 1:1000; Molecular Probes) for 20 min at room temperature. Free ROS oxidizes DHR 123 into the fluorescence probe rhodamine 123 (Rho 123) (Nani et al. 2010). Thus, after a rapid wash, slices were transferred into a Petri dish containing Krebs solution and visualized with a C2 confocal microscope (Nikon) using a 20× (0.50 NA) objective. Fluorescence signals were acquired in a 40 μm Z‐stack (5 μm sectioning) and images were further analysed with Volocity, version 6.3 (PerkinElmer): the Rho 123 signal was averaged within a 406.4 mm³ ROI, which more or less corresponds to one hemi‐nucleus.

Electrophysiology

Electrophysiological recordings were performed on 270 μm slices as described previously (Corsini et al. 2016). Slices were placed into a recording chamber and continuously superfused through a gravity‐driven system (2‐3 mL min^–1) with carbogenated recording Krebs solution at room temperature. The whole‐cell patch‐clamp technique was employed to record from HMs, visually identified through infrared microscopy according to their large soma diameter (∼25 μm). Patch pipettes, pulled from borosilicate glass capillaries, were filled with a CsCl‐based intracellular solution (in mm: 130 CsCl, 5 NaCl, 2 MgCl₂, 1 CaCl₂, 10 Hepes, 10 EGTA and 2 ATP‐Mg, pH 7.2 with CsOH; 290–310 mOsm; 3.5‐4.5 MΩ resistance), which has been shown to inhibit the I _M in HMs (Ghezzi et al. 2017a). Moreover, this solution did not affect the occurrence of bursting in HMs during TBOA application (Sharifullina & Nistri, 2006). HMs were voltage clamped at –70 mV holding potential (V _h, corrected online for liquid junction potential of 3 mV) and their series resistance (R _s) (5–20 MΩ) was monitored throughout the experiment without applying compensation; recordings were discarded whenever R _s exceeded 20% of the initial value. All recorded data were filtered at 3 kHz, sampled at 5 or 10 kHz and acquired using Clampex, version 9.2 (Molecular Devices, Sunnyvale, CA, USA).

Current traces were analysed using Clampfit, version 10.0 (Molecular Devices): cell input resistance (R _in) was calculated by measuring current responses to small 10 mV hyperpolarizing voltage steps from V _h. During experiments with TBOA, HMs were classified as either burster or non‐burster cells according to their ability to perform bursting during the first 15 min of TBOA administration. Bursting episode duration was measured as the time from the first to the last burst. Spontaneous postsynaptic currents (sPSCs), either mixed or pharmacologically isolated, were detected through the template search event detection method implemented in Clampfit, version 10.0: the amplitude of single events was analysed as well as their overall frequency.

Drugs

The sources of the drugs used were: TBOA, (2R)‐amino‐5‐phosphonovaleric acid (APV), strychnine hydrochloride and bicuculline methiodide (Tocris, Bristol, UK); 6,7‐dinitroquinoxaline‐2,3‐dione (DNQX) (Abcam, Cambridge, UK); XE‐991 (Sigma‐Aldrich); and retigabine (Alomone).

Statistical analysis

All results are expressed as the mean ± SEM unless otherwise indicated, where n corresponds to the number of cells recorded or slices for each independent experiment. Statistical analysis was performed using SigmaPlot, version 9.0 (Systat Software, Chicago, IL, USA). Normality and equal variance tests were first run to discriminate between parametric and non‐parametric data sets and to direct the correct choice of the statistical tests for comparison. Multiple groups were compared through one‐way ANOVA or Kruskal–Wallis ANOVA for parametric or non‐parametric data, respectively. The correction for multiple comparisons was performed with the Student–Newman–Keuls or Dunn's method, respectively. Two independent parametric data sets were compared with the Student's unpaired t test whereas two dependent sets were compared with Student's paired t test; corresponding non‐parametric data sets were compared with the Mann–Whitney rank sum or the Wilcoxon signed rank tests, respectively. P = 0.05 was considered statistically significant.

Results

Retigabine prevents TBOA‐induced HM death and toxicity

Because the I _M is a potent constraint to HM excitability (Ghezzi et al. 2017a), we first investigated whether its positive modulation by retigabine may prevent HM death evoked by TBOA (4 h of continuous application). HMs were identified as large soma diameter (>25 μm), SMI32‐immunopositive cells (Cifra et al. 2012), whereas AIF immunostaining was employed as a biomarker of cell damage (Monni et al. 2017). Representative images of brainstem slices are shown in Figs 1 and 2, whereas HM number and AIF intensity signal quantification are shown in Fig. 3. In accordance with our former studies (Corsini et al. 2017; Ghezzi et al. 2017b), treatment with TBOA (50 μm) led to a significant reduction in HM number (Fig. 1, second row, green pseudocolor, and Fig. 3 A) and an increase in AIF expression (Fig. 1, second row, red pseudocolour, and Fig. 3 B) compared to sham slices (Fig. 1, upper row; nuclear DAPI staining in blue). Furthermore, although retigabine (10 μm) per se did not affect either HM number (Fig. 1, third row, and Fig. 3 A) or AIF expression (Fig. 1, third row, and Fig. 3 B), it did prevent TBOA‐induced cell death and toxicity (Fig. 1, lower row, Fig. 3) when co‐applied with TBOA.

Figure 1. Histological neuroprotection by retigabine (10 μm) on HMs challenged by TBOA (50 μm)‐induced excitotoxicity.

Representative confocal images of a portion of the hypoglossal nucleus in which HMs are labelled with SMI32 (green, left column) or an anti‐AIF antibody (red, middle column); merged images are shown in the right column where cell nuclei are stained with DAPI (blue). Images were obtained from slices previously treated for 4 h in either sham condition (upper row) or with the following drugs: TBOA (second row), retigabine (third row), or TBOA and retigabine (lower row). [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 2. XE‐991 (10 μm) inhibits retigabine‐induced neuroprotection.

Representative confocal images of a region of the hypoglossal nucleus in which HMs are labelled with SMI32 (green, left column) or an anti‐AIF antibody (red, middle column); merged images are shown in the right column where cell nuclei are stained with DAPI (blue). Images were obtained from slices previously treated for 4 h in either XE‐991 alone (upper row), TBOA and XE‐991 (middle row), or TBOA, retigabine and XE‐991 (lower row). [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 3. Quantification of HM viability and AIF toxicity from immunohistochemical data.

A, HM number counted within a ROI for the treatments: sham (S; n = 14 slices), TBOA (T; n = 20), retigabine (R; n = 16), TBOA and retigabine (T+R; n = 17), XE‐991 (XE; n = 18), TBOA and XE‐991 (T+XE; n = 17), and TBOA, retigabine and XE‐991 (T+R+XE; n = 11). There are statistical significant differences between treatments (Kruskal–Wallis ANOVA: P ≤ 0.001; Dunn's method: *P ≤ 0.05 vs. T unless otherwise indicated by horizontal bar between two treatments). B, average AIF intensity signal within single HMs after treatments shown in (A). There are statistical significant differences between treatments (Kruskal–Wallis ANOVA: P ≤ 0.001; Dunn's method: *P ≤ 0.05 vs. T).

To further probe the role of the I _M in the context of HM excitotoxicity, we performed immunohistochemical experiments after treatment with the selective M‐channel inhibitor XE‐991 (10 μm) (Wang et al. 1998). Figure 2 (upper row) shows that XE‐991 alone did not affect HMs or AIF expression (see also quantification in Fig. 3). Nevertheless, after co‐treatment with XE‐991 and TBOA, cell death was consistently observed (Fig. 2, middle row, and Fig. 3 A), whereas AIF immunopositivity was similar to sham slices (Fig. 2, middle row, and Fig. 3 B). XE‐991 completely abolished the neuroprotective ability of retigabine because slices treated with TBOA, retigabine and the I _M inhibitor displayed cell loss and AIF toxicity (Fig. 2 lower row, Fig. 3) similar to slices treated with TBOA alone.

Our results suggest that the I _M positive modulator retigabine provided neuroprotection against TBOA‐induced excitotoxicity by preventing early AIF expression and cell loss.

Retigabine prevents TBOA‐induced ROS production

An important outcome of HM hyperexcitability developed during TBOA‐induced excitotoxicity is the massive influx of Ca²⁺ into the cytoplasm (Sharifullina & Nistri, 2006; Corsini et al. 2017), which, in turn, leads to ROS production and mitochondrial damage (Tortora et al. 2017), two phenomena prodromal to cell death. In particular, it is clear that ROS production precedes mitochondrial damage with maximal expression after 2 h of treatment with TBOA (Tortora et al. 2017). Thus, we studied ROS production with the rhodamine fluorescence method previously used in our laboratory (Nani et al. 2010; Tortora et al. 2017). Representative confocal images of brainstem slices stained with Rho 123 (probe for ROS generation) and Hoechst 33342 (for nuclear staining) are shown in Fig. 4 A: from these images and the data quantification shown in Fig. 4 B, it was apparent that TBOA induced a large increase in ROS production within the hypoglossal nucleus (Fig. 4 A, second from left) compared to sham slices (Fig. 4 A, far left). Although retigabine per se did not affect this parameter (Fig. 4 A, middle), it prevented TBOA‐induced ROS production when co‐applied with this toxic agent (Fig. 4 A, second from right). Prevention of ROS generation by retigabine was abolished after co‐treatment with TBOA, retigabine and the I _M inhibitor XE‐991 (Fig. 4 A, far right). Thus, inhibition of ROS production by retigabine may be one important process to induce neuroprotection against excitotoxicity.

Figure 4. Effect of retigabine on ROS generation in the hypoglossal nucleus during TBOA‐induced excitotoxicity.

A, representative confocal images of a portion of the hypoglossal nucleus stained with Rho 123 (red; to label free ROS) and Hoechst 33342 (blue; to label cell nuclei) obtained after 2 h of treatment in either sham condition (left) or the following treatments: TBOA (second from left) or retigabine alone (middle), TBOA and retigabine (second from right), or TBOA, retigabine and XE‐991 (right). B, average Rho 123 fluorescence intensity signal within a pre‐selected ROI for the following treatments: sham (S; n = 7), TBOA (T; n = 8), retigabine (R; n = 8), TBOA and retigabine (T+R; n = 7), and TBOA, retigabine and XE‐991 (T+R+XE; n = 8). There are statistical significant differences between treatments (one‐way ANOVA: P ≤ 0.001; Student–Newman–Keuls method: ^# P ≤ 0.05 vs. T). [Color figure can be viewed at http://wileyonlinelibrary.com]

Retigabine suppresses HM bursting and reduces hyperexcitability

The results obtained from immunohistochemical and ROS experiments suggested that retigabine has a substantial neuroprotective action against TBOA. Because an early hallmark of TBOA‐induced excitotoxicity on HMs is the emergence of strong bursting (Sharifullina & Nistri, 2006), it was of interest to investigate whether retigabine contrasted this phenomenon. Accordingly, electrophysiological experiments were performed on HMs voltage‐clamped at –70 mV V _h. In view of the network origin of TBOA‐induced bursting (Sharifullina & Nistri, 2006), we used, as a readout of this activity, single HM recordings with a CsCl‐based intracellular solution to isolate the I _M network effects of retigabine (Ghezzi et al. 2017a). The database of the present electrophysiological experiments comprises 49 HMs characterized by a mean R _in of 164 ± 14 MΩ.

Immediately after bath administration of TBOA to brainstem slices, a typical pattern of strong bursting appeared on ∼50% of HMs (Sharifullina & Nistri, 2006; Cifra et al. 2011b; Corsini et al. 2016; Ghezzi et al. 2017b). A representative current trace (V _h = –70 mV) of a bursting HM is shown in Fig. 5 A: in this example, TBOA (50 μm) triggered 13 bursts that spontaneously stopped after 33 min. Overall, bursting appeared in 17 out of 29 HMs (58%), with the average burst episode duration and burst number shown in Fig. 5 C and D (filled bars). TBOA also induced an inward current of –45 ± 13 pA (from –200 ± 17 pA to 245 ± 17 pA; paired t test: P ≤ 0.01) without affecting cell R _in (from 168 ± 20 to 174 ± 19; Wilcoxon signed rank test: P ≥ 0.05; n = 29). Furthermore, Fig. 5 E shows that, regardless of bursting ability, TBOA significantly increased both sPSC frequency (Fig. 5 E, top) and amplitude (Fig. 5 E, bottom), in accordance with previously published data (Sharifullina & Nistri, 2006; Corsini et al. 2016), validating the strong increase in network excitability.

Figure 5. Effect of retigabine (10 μm) on TBOA (50 μm)‐induced hyperexcitability and bursting.

A, B and F, representative current traces from three different HMs (V _h = –70 mV) recorded in the presence of TBOA alone (A), TBOA before retigabine (B) and TBOA after retigabine (F). The protocol of drug administration is exemplified by lines above each trace; interruptions in the traces are because of cell R _in and R _s monitoring. C and D, average bursting episode duration (C) and burst number (D) from recordings obtained with TBOA alone (filled bars, n = 7) or TBOA and retigabine together (open bars, n = 10; ***t test: P ≤ 0.001 vs. TBOA). E, plots of average sPSC frequency (top) or amplitude (bottom) vs. time obtained from recordings during sequential administration of TBOA and retigabine (drug administration protocol is shown over the time axis). Statistical significance was evaluated comparing control (5 min), TBOA (20 min) and TBOA + retigabine (35 min) and indicated by horizontal bars (paired t test: **P ≤ 0.01 and ***P ≤ 0.001).

It has been shown previously that riluzole (Cifra et al. 2011b) and nicotine (Corsini et al. 2016, 2017) stop bursting when they are applied immediately after the first burst had occurred. By employing a similar protocol in 10 out of 17 bursting HMs, retigabine (10 μm) was applied after bursting emerged, as shown in the representative current trace of Fig. 5 B in which bursting only lasted 10 min with a total number of five episodes. Average data are shown in Fig. 5 C and D (open bars), indicating that retigabine drastically reduced bursting episode duration (Fig. 5 C) and burst number (Fig. 5 D) compared to TBOA alone (filled bars). Furthermore, retigabine also potently decreased sPSC frequency (Fig. 5 E, top) and amplitude (Fig. 5 E, bottom) despite the continuous presence of TBOA. Finally, retigabine contrasted the TBOA‐induced inward current by inducing a net outward current of 43 ± 15 pA (from –248 ± 21 to –205 ± 17 pA; Wilcoxon signed rank test: P ≤ 0.01) without affecting cell R _in (from 168 ± 14 MΩ to 177 ± 14 MΩ; paired t test: P ≥ 0.05; n = 16).

In four out of 12 non‐bursting HMs, XE‐991 (10 μm) was applied after 15 min of TBOA application (not shown). The I _M antagonist did not unmask bursting in any of these cells, suggesting that basal, network‐located I _M was not the main determinant for preventing bursting in those non‐bursting cells.

These results indicate that functional up‐regulation of the I _M through retigabine readily contrasted a pathophysiological state of hyperexcitability as a result of TBOA application. We also tested whether retigabine could prevent burst onset by applying this drug before TBOA, as shown in Fig. 5 F. Retigabine did not modify either the baseline holding current or the cell R _in (Table 1), an observation consistent with blockade of the I _M in the recorded HM by intracellular Cs⁺ (Ghezzi et al. 2017a). Nevertheless, retigabine did induce a significant decrease of both baseline sPSC amplitude and frequency, as shown in Table 1. Figure 5 F shows that, after retigabine was pre‐applied, TBOA did not evoke bursting, as observed in all of the cells tested (n = 5; Fisher's exact test: P ≤ 0.05 vs. TBOA alone or pre‐applied before retigabine, n = 29). Thus, the up‐regulation of network‐located I _M could depress burst onset, as well as arrest established bursting.

Table 1.

Effect of retigabine on basal HM electrophysiological properties and mixed sPSC characteristics

	Control	Retigabine (10 μm)
Cell R in	148 ± 27 MΩ	145 ± 27 MΩ
Holding current	–180 ± 35 pA	–202 ± 35 pA
sPSC amplitude	–126 ± 13 pA	–72 ± 5 pAa
sPSC frequency	5.2 ± 0.5 Hz	2.9 ± 0.3 Hza

^aPaired t test: P ≤ 0.01 vs. control (n = 5).

Retigabine depresses network excitability by acting on major neurotransmitter systems

The ability of retigabine to restrain hyperexcitability and to suppress bursting may depend on M‐channels widely located in the network within the brainstem slice, in addition to its direct effect on HM excitability (Ghezzi et al. 2017a). To investigate the contribution of various neurotransmitters to the effect of retigabine, we performed electrophysiological experiments during pharmacological isolation with selective antagonists in the absence of TBOA. Thus, we employed DNQX (10 μm), APV (50 μm), strychnine (0.4 μm) and bicuculline (10 μm), at validated concentrations, to block AMPA, NMDA, glycine and GABA receptors, respectively (Donato & Nistri, 2000; Essin et al. 2002).

Isolated GABAergic, glycinergic and glutamatergic sPSCs are shown in Fig. 6 A–C (upper traces), respectively. Figure 6 A (lower trace) shows that retigabine decreased pharmacologically isolated GABAergic sPSC frequency and amplitude (see quantification in Fig. 6 D and E). Similar results were obtained with isolated glycinergic sPSCs (Fig. 6 B, lower trace) with a lower frequency and amplitude (Fig. 6 D and E). Glutamatergic sPSCs were modulated by retigabine as shown in Fig. 6 C (lower trace) with a decreased frequency (Fig. 6 D) but unchanged amplitude (Fig. 6 E). Because retigabine did not affect basal HM properties such as cell R _in (Fig. 6 F) and baseline current (Fig. 6 G) as a result of the CsCl intracellular solution, these results point to a widespread fall in network excitability by retigabine, which targeted GABAergic, glycinergic and glutamatergic neurons.

Figure 6. Effect of retigabine on network neurotransmitter systems.

A–C, representative current traces of three different HMs (V _h = –70 mV) in the presence of neurotransmitter receptor blockers to isolate GABAergic (A), glycinergic (B) or glutamatergic (C) sPSCs, recorded before (upper traces) and during retigabine (10 μm, lower traces) application. D and E, average GABAergic (filled bars), glycinergic (grey bars) and glutamatergic (open bars) sPSC frequency (D) and amplitude (E) recorded under control conditions (left bars) or during application of retigabine (right bars), showing significant differences (paired t test: *P ≤ 0.05 and **P ≤ 0.01 vs. correspondent control values). F and G, cell R _in (F) and holding current (G) obtained under control conditions (left) or during retigabine application (right), after pharmacological isolation of GABAergic (filled bars), glycinergic (grey bars) or glutamatergic (open bars) synaptic events, showing no significant differences.

Kv7.2 and Kv7.5 M‐channel subunits are expressed in medullary nuclei

The ability of retigabine to modulate sPSCs suggests the presence of functional M‐channel subunits in pre‐motoneurons within the rat brainstem slice, in addition to HMs. Nevertheless, to the best of our knowledge, there are no reports of Kv7 channel subunit expression in the hypoglossal nucleus and in brainstem nuclei known to provide functional inputs to HMs (Cunningham & Sawchenko, 2000; Peever et al. 2002; Chamberlin et al. 2007; van Brederode et al. 2011). Previous immunohistochemical studies have shown that Kv7.2, Kv7.3 (Wang et al. 1998) and Kv7.5 (Schroeder et al. 2000) are the main channel subunits expressed in central neurons, with Kv7.1 present in the cardiac tissue (Sanguinetti et al. 1996) and Kv7.4 being restricted to auditory and vestibular pathways (Kharkovets et al. 2000). Thus, our investigation was limited to the former three subunits.

Representative confocal images of Fig. 7 A show that HMs (labelled by SMI32 antibody) strongly express all Kv7 channel subunits tested (Fig. 7 A, right). By quantifying the ratio of Kv7‐ over SMI32‐positive neurons within the whole imaged field (Fig. 7 B), it was clear that each Kv7 channel subunit was expressed by the majority of HMs in the nucleus (>85%), which is in agreement with our previous electrophysiological study (Ghezzi et al. 2017a).

Figure 7. Expression of Kv7 channel subunits in the hypoglossal nucleus.

A, representative confocal images of a portion of the hypoglossal nucleus in which HMs are labelled with SMI32 (green, left column) or anti‐Kv7.2 (red, top row), anti‐Kv7.3 (red, middle row) or anti‐Kv7.5 (red, bottom row) channel subunit antibodies. B, percentages of Kv7‐positive (Kv7⁺) over SMI32‐positive (SMI32⁺) neurons, counted in the whole imaged field (exemplified in A), for Kv7.2 (open bar; n = 3), Kv7.3 (grey bar; n = 3) and Kv7.5 (filled bar; n = 3). Error bars indicate the standard deviation of the mean. [Color figure can be viewed at http://wileyonlinelibrary.com]

We next considered whether Kv7 channel subunits are also expressed in other medullary nuclei known to provide functional inputs to HMs. Accordingly, we performed immunohistochemical experiments by employing the specific neuronal marker NeuN together with each anti‐Kv7 channel subunit antibody tested. Among all of the medullary nuclei, our analysis was focused on the portion of lateral reticular formation present in the slice (dorsal medullary reticular column; DMRC), known to provide excitatory and inhibitory inputs to HMs (Cunningham & Sawchenko, 2000; Peever et al. 2002; Chamberlin et al. 2007), and the nucleus of Roller, known to provide exclusively GABAergic inputs to HMs (van Brederode et al. 2011). The location of these nuclei within the brainstem slice is shown in Fig. 8 B. Confocal images of Fig. 8 A demonstrated that Kv7.2 (Fig. 8 A, second row) and Kv7.5 (Fig. 8 A, bottom row) subunits are abundantly expressed in both the dorsal and ventral portion of the DMRC (DMRCd, Fig. 8 A, left column, and DMRCv, Fig. 8 A, middle column, respectively), as well as in the nucleus of Roller (Fig. 8 A, right column), whereas Kv7.3 subunit (Fig. 8 A, fourth row) was less expressed. Quantification of the ratio between Kv7‐ and NeuN‐positive neurons (Fig. 8 C) confirmed a substantial expression of Kv7.2 (Fig. 8 C, open bars) and Kv7.5 (Fig. 8 C, filled bars) subunits in all nuclei tested (>80% and >90%, respectively) and a lower expression of Kv7.3 subunit (Fig. 8 C, grey bars) (>40%).

Figure 8. Expression of Kv7 channel subunits in medullary nuclei functionally linked to HMs.

A, representative confocal images of the dorsal and ventral portions of the dorsal medullary reticular column (DMRCd, left column and DMRCv, middle column) and the nucleus of Roller (right column) stained with NeuN (green, top, third and fifth rows) and anti‐Kv7.2 (red, second row), anti‐Kv7.3 (red, fourth row) or anti‐Kv7.5 (red, bottom row) channel subunit antibodies. B, schematic representation of half brainstem slice (left) and toluidine blue‐stained half brainstem slice (right) highlighting the location of the hypoglossal nucleus and surrounding medullary nuclei. The scheme on the left is a modified version of that from Ashwell & Paxinos, 2015. C, percentages of Kv7‐positive (Kv7⁺) over NeuN‐positive (NeuN⁺) neurons, counted in the whole imaged field (exemplified in A) of DMRCd (left bars), DMRCv (middle bars) and nucleus of Roller (right bars), for Kv7.2 (open bar; n = 3), Kv7.3 (grey bar; n = 3) and Kv7.5 (filled bar; n = 3). Error bars indicate the standard deviation of the mean. [Color figure can be viewed at http://wileyonlinelibrary.com]

Discussion

The main finding of the present study is that retigabine confers neuroprotection to HMs in an in vitro model of excitotoxicity established during glutamate uptake block. Functional up‐regulation of the I _M through retigabine (Ghezzi et al. 2017a) probably led to a strong reduction in HM and network excitability, which is effective in contrasting TBOA‐evoked bursting. Furthermore, retigabine prevented the large ROS generation after 2 h of TBOA application and, after 4 h of treatment, this drug prevented both HM death and AIF overexpression. The widespread expression of Kv7 K⁺ channel subunits underlying neuronal I _M throughout the medullary slice preparation highlights an important role for network I _M in this neuroprotection mechanism.

Expression of Kv7 K⁺ channel subunits in the brainstem medullary slice preparation

Although the presence of a functional I _M on HMs has been demonstrated previously (Ghezzi et al. 2017a), when studying the effect of selective drugs on network phenomena such as TBOA‐induced excitotoxicity, it is fundamental to understand whether and where Kv7 channel subunits may be located in the brainstem slice preparation. Thus, to confirm the expression of Kv7 K⁺ channel subunits within the hypoglossal nucleus and functionally related medullary nuclei containing pre‐motoneurons, we performed immunohistochemical experiments to fill the lack of current data. Our results show that all main Kv7 channel subunits underlying neuronal I _M (Kv7.2, Kv7.3 and Kv7.5) (Wang et al. 1998; Schroeder et al. 2000) were highly expressed in the soma of HMs, as indicated by their co‐localization with the motoneuron marker SMI32 (Cifra et al. 2012). This result provides further validation concerning the presence of an I _M in HMs, as previously demonstrated exclusively through electrophysiological methods (Ghezzi et al. 2017a).

The hypoglossal nucleus receives functional projections from a variety of other brainstem nuclei. Excitatory and inhibitory inputs to HMs have been shown to originate mainly from an extended region of the caudal reticular formation, termed the DMRC (Cunningham & Sawchenko, 2000; Peever et al. 2002; Chamberlin et al. 2007), located immediately lateral to the hypoglossal nucleus itself (Fig. 8 B). Another important source of inputs is the pre‐Bötzinger complex, providing respiratory rhythmic drive (Feldman et al. 2013), whereas GABAergic innervation comes from interneurons located within the hypoglossal nucleus (Peever et al. 2002) or in the nearby ventral nucleus of Roller (van Brederode et al. 2011). Our investigation was focused on the DMRC and the nucleus of Roller given the probable absence of a functional pre‐Bötzinger complex in our thin slice preparation (Feldman et al. 2013); we also did not study hypoglossal interneurons given their very low prevalence in the hypoglossal nucleus (Viana et al. 1990; Cifra et al. 2012). In all areas under investigation, a similar expression of Kv7 channel subunits was found: Kv7.2 and Kv7.5 subunits were expressed in the majority of neurons in the DMRCd, DMRCv and nucleus of Roller, whereas Kv7.3 subunits were expressed only in less than half of the imaged neurons. Furthermore, expression of Kv7 channel subunits was confirmed to be somatic by co‐staining with NeuN, the selective marker for the neuronal nucleus. Although more detailed immunohistochemical studies will be necessary in the future, our results suggest that the molecular correlates of the I _M in HMs and neighbouring nuclei may constitute homomeric assemblies of either Kv7.2 or Kv7.5 subunits or heteromeric assemblies of Kv7.3 with one of the other subunits (Wang et al. 1998; Greene & Hoshi, 2016). Their somatic expression is in accordance with a direct modulatory effect of I _M on neuronal excitability, whereas the widespread expression of such subunits throughout the slice preparation alludes to the possibility of an ample downregulation of excitability and neurotransmission in the pre‐motoneuron network. Indeed, because pharmacological blockers of glutamate receptors or tetrodotoxin suppress TBOA‐evoked pathological bursting (Sharifullina & Nistri, 2006), it appears that the development of excitotoxicity requires the major role of pre‐motoneurons with respect to generating a phenomenon eventually manifested by a loss of HMs.

Broad depression of HM and network excitability by retigabine

Among voltage‐gated K⁺ conductances, the I _M has the unique characteristic of providing a non‐inactivating outward current at a depolarized membrane potential from rest. Thus, neurons expressing M‐channels possess an intrinsic mechanism for dampening their excitability (Adams et al. 1982; Delmas & Brown, 2005). Although there are few endogenous compounds that are able to functionally up‐regulate the I _M, this braking system can be effectively boosted by drugs such as the anti‐convulsant retigabine. This drug shifts the activation curve of single M‐channels toward a more hyperpolarized membrane potential (Wickenden et al. 2000; Tatulian et al. 2001; Tatulian & Brown, 2003), with the consequence that a greater number of those channels would be open at (or near) resting potential. This mechanism was demonstrated to be effective in constraining neuronal excitability in a number of neurons (Brown & Passmore, 2009) comprising motoneurons (Lombardo & Harrington, 2016; Ghezzi et al. 2017a). Moreover, retigabine has been clinically employed as an adjuvant treatment against drug‐resistant seizures, where hyperexcitability is the main symptomatic event (Porter et al. 2007), even though clinical use of the drug has been discontinued because of its side effects.

In the context of our model of brainstem slice comprising the hypoglossal nucleus, the presence of the I _M in HMs has been demonstrated previously, it is targeted by retigabine (Ghezzi et al. 2017a), and remains functional even when synaptic transmission is inhibited (Ghezzi et al. 2017a). In our previous study, we also demonstrated that retigabine opened M‐channels usually closed at resting membrane potential, hence shifting the spike threshold toward a more depolarized potential and significantly reducing HM intrinsic excitability (Ghezzi et al. 2017a). However, in a relatively complex context such as the brainstem slice, in which part of the original neuronal network is still present, the contribution of pre‐motoneurons becomes crucial for determining the final motoneuron output (Rekling et al. 2000).

Our histological data concerning the expression of Kv7 channel subunits in the brainstem slice suggest that pre‐motoneurons may be endowed with this K⁺ conductance and that modulation by retigabine may affect the pattern of synaptic inputs onto HMs. Network I _M function could be inferred by studying sPSCs as a result of ongoing neurotransmitter release by pre‐motoneurons on HMs and isolating the presynaptic effect with a Cs⁺‐containing intracellular solution that has been shown to inhibit the I _M of motoneurons (Ghezzi et al. 2017a). Thus, the reduction in the frequency of mixed and pharmacologically isolated sPSCs during retigabine administration is an index of a widespread reduction in network excitability (Fidzinski et al. 2015) that affects all main neurotransmitter systems (glutamatergic, GABAergic and glycinergic). Conversely, the decrease in inhibitory sPSC amplitude may have a variety of explanations. One possibility is the involvement of network located M‐channels in the spike‐dependent release of GABA and glycine neurotransmitters (Shah et al. 2008; Sun & Kapur, 2012), which is conceivable given their localization at the AIS and nodes of Ranvier (Devaux et al. 2004). Another possibility would require the presence of M‐channels at the synapse: indeed, the effect of presynaptic I _M may be to regulate resting potential and neurotransmitter release (Huang & Trussell, 2011), whereas postsynaptic Kv7 channels in dendrites (i.e. that might be spared by Cs⁺ inhibition) may shunt synaptic events (Fidzinski et al. 2015). In our slice preparation, the expression of Kv7.2 and Kv7.3, which were found to localize preferentially at the AIS (Devaux et al. 2004; Tzingounis et al. 2010), and Kv7.5 channels, mainly localized at synapses (Caminos et al. 2007; Huang & Trussell, 2011; Fidzinski et al. 2015), is compatible with either possibility.

Pathophysiological model of HMs challenged by glutamate uptake block‐induced excitotoxicity

In recent years, our in vitro model of excitotoxicity has proved useful and reliable for investigating electrophysiological activities and the molecular events characterizing the progressive deterioration of a motor network. By employing TBOA as a pharmacological agent to inhibit EAATs and therefore glutamate uptake (Shigeri et al. 2004), an excitotoxic state is rapidly (2–3 min) obtained within the hypoglossal nucleus, with half of the HMs displaying bursting activity characterized by relatively large inward currents or depolarizing waves (Sharifullina & Nistri, 2006; Cifra et al. 2011b; Corsini et al. 2016). Previous studies in our laboratory have shown that these inward currents reflect hyperexcitability of motoneurons and are correlated with a large influx of Ca²⁺ into their cytoplasm (Sharifullina & Nistri, 2006; Corsini et al. 2017). In turn, the disruption of Ca²⁺ homeostasis in HMs is the most probable trigger for a cascade of pathophysiological processes and molecular alterations, eventually leading to cell damage and/or death (Jaiswal, 2014). Indeed, after 2 h of continuous glutamate uptake block, there is a significant rise in ROS production within the nucleus hypoglossus, which is followed by mitochondrial damage (Tortora et al. 2017). Subsequently, after 4 h of treatment with TBOA, several toxic compounds are released from impaired mitochondria into the cell cytoplasm (Corsini et al. 2016, 2017). One of them is AIF, which can migrate into the nucleus and act as the ultimate trigger for caspase independent apoptosis (Joza et al. 2001) of a significant proportion of HMs, whereas surviving ones display an increased expression of this factor throughout the subcellular compartments (Corsini et al. 2017; Monni et al. 2017). It is not possible to understand the fate of surviving HMs because of the time‐limited survival of the in vitro preparation (Corsini et al. 2016). Nevertheless, an excessive expression of AIF most probably correlates with imminent cell death because this factor triggers irreversible chromatin condensation and DNA damage (Joza et al. 2001).

Although HMs comprise a heterogeneous population of motoneurons (namely, protruders and retractors located in the ventral and dorsal part of the hypoglossal nucleus, respectively) (Lewis et al. 1971; Krammer et al. 1979), previous studies have suggested that antagonist muscles may be equally recruited during complex physiological behaviours of the tongue (such as licking, swallowing and rhythmic activity involved in respiration) to carry out synchronous oscillatory discharges (Wiesenfeld et al. 1977; Travers et al. 2000; Feldman et al. 2013). Interestingly, antagonist tongue muscles are also coherently activated during the inspiratory phase of respiration (Rice et al. 2011) via a distributed pre‐motoneuron innervation to stabilize the tongue during complex behaviours (Dobbins & Feldman, 1995). Another determinant of coherent activity of HMs may be the expression of electrical synapses formed by gap junctions between neighbouring cells, which are a necessary requirement for TBOA‐induced bursting (Sharifullina & Nistri, 2006; Corsini et al. 2017). The reason why only half of the HMs display bursting behaviour remains incompletely understood. Nevertheless, previous Ca²⁺ imaging studies did not identify any differential susceptibility to TBOA‐induced excitotoxicity between different classes of motoneurons because the presence of bursting HMs was scattered across the whole hypoglossal nucleus (Sharifullina & Nistri, 2006; Corsini et al. 2017).

Model limitations

A clear limitation to the translational value of the present research is the use of an in vitro preparation obtained from neonatal animals to explore the basics of neurodegeneration typical of adult life. This is a result of the well‐known difficulties encountered when performing similar experiments in preparations of motoneurons from adult animals. Nevertheless, it is noteworthy that TBOA remains strongly neurotoxic after in vivo microinjection into the adult rat brain (Selkirk et al. 2005). Furthermore, one theory suggests that adult‐onset ALS is the consequence of processes initiated during early development (van Zundert et al. 2012). Indeed, a widely employed genetic mouse model of ALS (overexpressing the hSOD1^G93A gene) shows early pathophysiological features of human ALS even during the first postnatal days (Kanjhan et al. 2016). Clinical observations indicate that the neonatal brain is highly sensitive to excitotoxicity (Johnston, 2001), against which neuroprotection remains an important goal (Gonzalez & Ferriero, 2009). Hence, in terms of motoneuron toxicity, neonatal brain tissue appears to offer a model with basic pathophysiological properties not very dissimilar to those of the adult one.

Our in vitro model focuses on one of the possible pathogenetic mechanisms for ALS, namely excitotoxicity. Indeed, sporadic and familial ALS patients may show a decreased function of EAAT in the cortex and spinal cord with an abnormal increase in extracellular glutamate (Rothstein et al. 1992; Spreux‐Varoquaux et al. 2002). Beside the ‘glutamate hypothesis’, other mechanisms for the onset of this complex disease have been proposed recently in alternative animal models: these include SOD1 mutations (Philips & Rothstein, 2015), as well as aberrant changes in the neuronal protein TDP‐43 (Gao et al. 2018) or FUS (Guerrero et al. 2016). Thus, a unique model that represent the whole complexity of ALS phenotypes and causes is still missing, given the multifactorial nature of this disease and its possible subdivision into various clinical subtypes (Al‐Chalabi et al. 2012).

Retigabine confers neuroprotection against excitotoxicity by contrasting hyperexcitability

Because the first phenomenon occurring immediately after TBOA application is hyperexcitability of HM and bursting (Sharifullina & Nistri, 2006), drugs that are able to contrast this pathophysiological state may be important for providing protection against cell death. Indeed, previous studies by our laboratory demonstrated that several drugs such as riluzole (Cifra et al. 2011b), nicotine (Corsini et al. 2016, 2017) and propofol (Ghezzi et al. 2017b), acting on different molecular targets, share the ability to contrast bursting and thus prevent cell death. The retigabine‐dependent reduction in network excitability affecting both HMs (Ghezzi et al. 2017a) and pre‐motoneurons is sufficient to readily arrest bursting activity or prevent its appearance, in accordance with the experimental protocol employed. Although investigating the differential contribution by inhibitory or excitatory pre‐motoneurons expressing I _M requires additional studies, I _M‐mediated dampening of the excitability of GABAergic and glycinergic neurons is expected to enhance bursting and toxicity because, at this developmental age, GABA and glycine are already inhibitory (Marchetti et al. 2002). Conversely, decreasing (via I _M activation) excitatory synaptic transmission is probably an important factor for contrasting bursting and excitotoxicity and is compatible with the role of this conductance in modulating excitatory synaptic transmission (Fidzinski et al. 2015). The present data are therefore consistent with respect to proposing retigabine as a prototypic agent against excitotoxicity acting via I _M at the network and HM levels.

Because retigabine is assumed to have a strong selectivity for some Kv7 K⁺ channels subunits at the concentration employed in the present study (Rundfeldt & Netzer, 2000; Wickenden et al. 2000), the more parsimonious hypothesis is that the observed prevention of ROS generation, AIF expression and HM death are all direct consequences of the profound reduction in network excitability. Indeed, retigabine may strongly limit the influx of Ca²⁺ into HM cytoplasm because of arrested bursting and therefore prevent all of the pathological processes triggered by impaired Ca²⁺ homoeostasis (Jaiswal, 2014; Corsini et al. 2017). A similar scenario has been recently proposed by Wainger et al. (2014) in another in vitro model of ALS demonstrating the ability of retigabine to contrast intrinsic hyperexcitability and ameliorate cell survival of familial ALS patient‐derived motoneurons obtained from induced pluripotent stem cells (Wainger et al. 2014). Thus, in future experiments, it would be interesting to directly test the impact of retigabine and I _M modulation on HM Ca²⁺ dynamics.

The issue of hyperexcitability in ALS

The crucial role of hyperexcitability in ALS has been demonstrated by a variety of studies performed in patients (Geevasinga et al. 2016), as well as in several experimental models (Kuo et al. 2004; Wainger et al. 2014). In this context, dysfunctions of axonal Na⁺ or K⁺ channels are usually manifested in peripheral nerves (Park et al. 2017) and have been consistently observed in sporadic and familial forms of ALS (Kanai et al. 2006; Shibuya et al. 2011; Devlin et al. 2015; Geevasinga et al. 2015; Iwai et al. 2016; de Carvalho et al. 2017). Another process for eliciting motoneuron hyperexcitability might arise from a dysfunction of cortical neurons in the motor cortex (‘upper motoneurons’) that impinge upon brainstem or spinal motoneurons (‘lower motoneurons’) such that ALS may be considered as a corticofugal disease spreading across synapses (King et al. 2016; Eisen et al. 2017). Hence, a build‐up of toxic protein aggregates of the DNA‐binding protein TDP‐43 impairs the function of cortical neurons leading, in turn, to functional and molecular alterations in distant neurons receiving cortical projections (Braak et al. 2013; Hardiman et al. 2017). In our model, hyperexcitability has a network origin because it reflects an excessive extracellular glutamate concentration as a result of its inhibited uptake (Sharifullina & Nistri, 2006). Despite this network origin, the positive pharmacological modulation of a single K⁺ conductance, the I _M, is sufficient to potently contrast HM bursting and neurodegeneration.

The causative relation between hyperexcitability and degeneration of motoneurons in ALS has recently been questioned by experimental evidence obtained from the SOD1^G93A mutant mouse model of this disease (Leroy & Zytnicki, 2015). In these neonatal mice, hyperexcitability is found in spinal motoneurons innervating slow‐contracting muscle fibres, whereas no change in excitability is detected in motoneurons innervating fast‐contracting muscle fibers (Leroy et al. 2014). Later, however, in the adult mouse, fast spinal motoneurons become hypoexcitable, whereas slow motoneurons remain apparently unaffected (Delestrée et al. 2014; Martinez‐Silva et al. 2018). Because fast motoneurons are the main ones to degenerate in ALS, these studies suggest a connection between hypoexcitability (and not hyperexcitability) and spinal motoneurons degeneration (Martinez‐Silva et al. 2018). Nevertheless, it is difficult to exclude the possibility that at least some motoneurons underwent a short‐lived phase of hyperexcitability prior to becoming hypoexcitable. Furthermore, it is noteworthy that brainstem motoneurons may not share precisely the same pathophysiological characteristics as those of spinal motoneurons (Shellikeri et al. 2017): indeed, in the same SOD1^G93A mouse model, there is hyperexcitability of trigeminal fast motoneurons, as well as signs of hypoexcitability in slow ones (Venugopal et al. 2015).

In accordance with all these studies, our previous work has demonstrated that, in HMs, enhanced excitability caused by excess of glutamate was a rather transient phenomenon triggering a slowly developing cascade of metabolic and gene expression changes only later leading to cell death (Sharifullina & Nistri, 2006; Corsini et al. 2016). Although a hypothesis has been proposed suggesting that hyperexcitability may be a compensatory change (perhaps even protective) of motoneurons to the early development of the disease process in the SOD1^G93A mouse model (King et al. 2016), our model indicates that hyperexcitable motoneurons are those that subsequently die (Sharifullina & Nistri, 2006; Corsini et al. 2016). Within the framework of our theory, it is feasible that this deleterious process occurs with a distributed time‐course affecting, in particular, certain motoneurons that express vulnerable properties such as, for example, gap junctions (Corsini et al. 2017) or a differential ability to produce heat shock proteins to sequester AIF (Shabbir et al. 2015). Thus, when searching for new therapeutic approaches to this complex disease, it is fundamentally important to consider motoneuron type and disease progression (Kim et al. 2017).

Conclusions

The widespread distribution of Kv7 K⁺ channel subunits (Delmas & Brown, 2005) throughout the CNS renders the I _M as a suitable target for contrasting many hyperexcitability‐related pathologies such as epilepsy, neuropathic pain and neurodegenerative diseases (Barrese et al. 2018). The results of the present study provide further support concerning the ability of I _M positive modulation as a neuroprotective strategy against hyperexcitability and excitotoxicity in an in vitro experimental model. A direct translational value of retigabine per se is challenged by the side effects elicited by this drug after prolonged treatment (Faulkner & Burke, 2013) that recently led to discontinuation of its commercialization. Nevertheless, targeting of the Kv7 channel is still recognized as a highly promising strategy for the treatment of a variety of pathological conditions (Barrese et al. 2018) and therefore there is increasing motivation to find alternative, more selective pharmacological agents to this drug (Barro‐Soria et al. 2017; Wang et al. 2017).Thus, the results of the present study provide further evidence of possible therapeutic outcomes with respect to positive modulation of the I _M in the context of hyperexcitability and excitotoxicity.

Additional information

Competing interests

The authors declare that they have no competing interests.

Author contributions

All experiments were performed in the laboratory of Professor Andrea Nistri at the International School for Advanced Studies (SISSA). FG, LM and AN were responsible for the conception and design of the work. FG and LM were responsible for the acquisition, analysis or interpretation of data. FG, LM and AN were responsible for drafting the work or revising it critically for important intellectual content. All authors approved the final version of the manuscript submitted for publication, and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

This work was supported by an intramural SISSA grant. FG is the recipient of a Wellcome Trust studentship (102170/B/13/Z).

Biography

Filippo Ghezzi gained his BSc degree in Biology at the University of Parma and MSc degree in Neuroscience at the University of Trieste. He worked for almost 2 years in the laboratory of Professor Andrea Nistri at SISSA, Trieste, where he studied K⁺ conductances and GABAergic inhibition as neuroprotective agents in a model of neurodegeneration. Currently, he is a first‐year student at the Wellcome Trust Doctoral Training Programme in Neuroscience, University of Oxford. He has a strong interest in the development of neuronal networks and is fascinated by computational neuroscience as a tool for complementing electrophysiology and imaging in research.