Progressive loss of a glial potassium channel (KCNJ10) in the spinal cord of the SOD1 (G93A) transgenic mouse model of amyotrophic lateral sclerosis


Address correspondence and reprint requests to Clemens Neusch, Department of Neurology, Georg-August University Göttingen, Robert-Koch-Strasse 40, 37075 Göttingen, Germany.


Transgenic mice expressing the superoxide dismutase G93A mutation (SOD1G93A) were used to investigate the role of glial inwardly rectifying K+ (Kir)4.1 channels, which buffer extracellular K+ increases in response to neuronal excitation. A progressive decrease in Kir4.1 immunoreactivity was observed predominantly in the ventral horn of SOD1G93A mutants. Immunoblotting of spinal cord extracts mirrored these changes by showing a loss of Kir4.1 channels from presymptomatic stages onwards. Kir4.1 channels were found to be expressed in the spinal cord grey matter, targetting astrocytes and clustering around capillaries, supporting their role in clearance of extracellular K+. To understand the functional implications of extracellular K+ increases, we challenged the NSC34 motor neurone cell line with increasing extracellular K+ concentrations. Exposure to high extracellular K+ induced progressive motor neurone cell death. We suggest that loss of Kir4.1 impairs perineural K+ homeostasis and may contribute to motor neurone degeneration in SOD1G93A mutants by K+ excitotoxic mechanisms.

Abbreviations used

amyotrophic lateral sclerosis












dorsal horn


Dulbecco's modified Eagle's medium


enhanced green fluorescent protein


familial ALS


fetal calf serum


glial fibrillary acidic protein


glutamate transporter


(human) superoxide dismutase


extracellular K+ concentration


inwardly rectifying K+; L IX, lamina IX




normal goat serum






phosphate-buffered saline








sodium dodecyl sulphate




tris-buffered saline tween




ventral horn


wild type

Amyotrophic lateral sclerosis (ALS) is a late-onset neurodegenerative disease characterized by upper and lower motor neurone degeneration. The majority of ALS cases are sporadic, but approximately 10% are hereditary (familial ALS; FALS). Some 15–20% of FALS cases have been associated with dominant mutations in the Cu/Zn superoxide dismutase (SOD)1 gene (Rosen et al. 1993). Transgenic mice bearing FALS-associated human SOD1 (hSOD1) mutations show lower motor neurone degeneration and clinical symptoms reminiscent of human ALS (Deng et al. 1993; Gurney et al. 1994, 1996, 1998; Wong et al. 1995; Gurney 1997). The pathogenic concept of a cell-autonomous motor neurone disorder has recently been challenged. First, the astroglial glutamate transporter EAAT2 (GLT1) is reduced in some hSOD1 transgenic mouse (Bruijn et al. 1997; Bendotti et al. 2001) and rat (Nagai et al. 2001; Howland et al. 2002) lines and also in patients with ALS (Rothstein et al. 1990, 1992, 1995). Moreover, glial glutamate transport is decreased in brain and spinal cord of patients with ALS (Rothstein et al. 1992) whereas glutamate levels are increased in the CSF (Rothstein et al. 1990), suggesting a role for glutamate excitotoxicity in the pathogenesis of the disease. Second, transgenic mouse lines expressing mutant SOD1 driven by a neurone-specific promoter (Pramatarova et al. 2001; Lino et al. 2002) or by an astrocyte-specific promoter (Gong et al. 2000) failed to provoke the disease. Third, chimeric mice that contain mixtures of normal SOD1- and SOD1 mutant-expressing cells suggest that mutant SOD1-expressing non-neuronal cells are required to induce neuronal cell death and that SOD1-negative non-neuronal cells that surround motor neurones may rescue affected motor neurones (Clement et al. 2003). Finally, diminishing the transgenic expression of mutant SOD1 in microglia sharply slowed disease progression in later phases of the disease (Boillee et al. 2006). Proposed mechanisms for glial-induced motor neurone death include glutamate excitotoxicity owing to loss of the glial-specific GLT1 and/or ionic dysbalance in the perineuronal environment as a result of insufficient glial control.

In glial cells, inwardly rectifying K+ (Kir) channels, e.g. Kir4.1 channels (KCNJ10), buffer extracellular K+, which is released during neuronal activity, and siphon it toward sinks, such as capillaries and the vitreous humor (Newman et al. 1984; Newman 1986, 1987, 1993; Newman and Reichenbach 1996; Kofuji et al. 2002). A polarized cellular distribution of Kir4.1 expression with clusters around small vessels or the vitreous bulb has been attributed to its intrinsic role of taking up K+ from the extraneuronal space and extruding it into the blood circulation. Recent literature suggests that Kir channels also serve a more general function by setting the resting membrane potential in glial cells, which in turn would govern the transmembrane gradients of many transported molecules, including neurotransmitter molecules such as glutamate. The Kir4.1 channel subunit is expressed predominantly in glial cells throughout the spinal cord and brainstem (Neusch et al. 2001; 2006; Olsen et al. 2005). High expression levels of Kir4.1 have been observed in grey matter and to a lesser extent in white matter structures in early postnatal days (Neusch et al. 2001). Genetic inactivation of the Kir4.1 channel subunit (Kofuji et al. 2000) leads to extensive developmental spinal cord defects (Neusch et al. 2001) and also to secondary motor neurone degeneration in early postnatal weeks.

Here, we report a progressive loss of the Kir4.1 protein in the ventral horn from presymptomatic to end-stage animals in hSOD1 (G93A) (SOD1G93A) transgenic mice. We show that in the adult spinal cord Kir4.1 channels are expressed predominantly on astroglial cells and cluster on capillaries, indicating that the channel is implicated in regulation of extracellular K+ concentration ([K+]o). Finally, we present evidence that increasing extracellular K+ levels affects motor neurone cell survival in an in vitro model.

Materials and methods

SOD1G93A transgenic mice and mice expressing enhanced green fluorescent protein (EGFP) under the glial fibrillary acidic protein (GFAP) promoter

Mice transgenic for the mutated human SOD1G93A (TgN[SOD1 G93A]1) and wild-type (wt)-SOD1-overexpressing mice were originally obtained from the Jackson Laboratory, Bar Harbor, ME, USA (strain B6.Cg-Tg(SOD1-G93A)1Gur/J) and bred at the Max-Planck Institute for Experimental Medicine, Göttingen animal facility. Mice expressing the fluorescent protein EGFP under the GFAP promoter have been generated and characterized at this institute. This mouse line (designated FVBN-TgN(GFAP-EGFP)GFEC-FKi; GFAP-EGFP in short) showed a similar expression pattern to the mouse line described earlier (Nolte et al. 2001). GFAP-EGFP mice were used to study cellular expression of Kir4.1 channels. Genotyping of the SOD1G93A animals was performed according to standard protocols and based on earlier descriptions (Rosen et al. 1993). For the Kir4.1 expression studies using quantitative immunoblotting, original SOD1 (G93A) mice were used. GFAP-EGFP mice were used to determine cell-specific and subcellular Kir4.1 expression in the spinal cord. SOD1G93A mice were classified as ‘presymptomatic’ when showing no signs of motor deficits [postnatal (P)60–P90], ‘early symptomatic’ (P90–P108) when an abnormal gait was first apparent, ‘late symptomatic’ (P100–P120) when paralysis of at least one limb was observed, and ‘end-stage’ (P120–P140) when righting reflex failure occurred owing to the complete paralysis of hindlimbs and at least one forelimb (Howland et al. 2002). Animals monitored in this study showed first signs of disease at ∼ 90 days of age and reached end-stage disease within ∼ 120–140 days. This classification reflects functional changes of behaviour rather than chronological categories leading to a time overlap between different disease stages.

Immunohistochemistry and immunofluorescence

Animals were anaesthetized and cardioperfused with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA) in PBS. Spinal cords were removed and postfixed in 4% PFA/PBS overnight. Organs were embedded in paraffin and sectioned at 8 µm. For immunohistochemistry, paraffin sections were dewaxed in xylene and rehydrated in ethanol, heated for 10 min at 55°C and then 10 min at 95°C in an antigen demasking solution (Vector Laboratories, Burlingame, CA, USA). After blocking for 20 min in 0.2% Triton X-100 and 10% normal goat serum (NGS) in PBS, samples were blocked for 30 min in 5% skim fat dry milk in PBS, and then washed and incubated with the primary antibody in 1% NGS at 4°C for 24 h. Samples were incubated with a fluorescent Cy-3 (1 : 500)- or Cy-5 (1 : 500)-conjugated secondary antibody (Jackson Immunoresearch, West Grove, PA, USA) for 1 h. Rabbit Kir4.1 polyclonal antibody (1 : 200; Alomone Laboratories, Jerusalem, Israel) was made against a synthetic peptide KLEESLREQAEKEGSALSV, corresponding to residues 356–375 of the rat Kir4.1 protein. For control experiments a different rabbit anti-Kir4.1 antibody was used; this was directed against the peptide REQAEKEGSALSVRISNV, corresponding to amino acids 362–379 in the C-terminus of mouse Kir4.1 (Kir4.1-PK). Other antibodies used in this study were goat anti-aquaporin (AQP)4, directed against amino acids 280–320 of AQP4 of human origin (1 : 250, used with donkey serum instead of NGS, no blocking with skim fat dry milk; Santa Cruz Biotechnology, Santa Cruz, CA, USA). Nissl staining was performed according to the manufacturer's protocol (Molecular Probes, Eugene, OR, USA). Images of tissue sections were obtained on a Zeiss Axiovert 200M/LSM 510 Meta confocal laser-scanning microscope (Zeiss, Jena, Germany). All immunocytochemical experiments were performed at least in quadruplicate.

Western blot analysis and quantification

Animals were anaesthetized with 7% chloralhydrate and decapitated. Spinal cord was removed and immediately homogenized in ice-cold buffer containing 1% Triton X-100, 50 mm Tris-HCl, 150 mm NaCl, 5 mm EDTA, 6 mm NaF, 2.5 mm sodium pyrophosphate, 1 mm sodium orthovanadate, 1 mm dithiothreitol, 1 mm phenylmethylsulphonyl fluoride and 1 × Boehringer complete (Boehringer, Manheim, Germany). Extracts were lysed on ice for 30 min and then clarified by centrifugation at 4°C and 13 000 g. The supernatant was removed and protein concentration was determined using the bicinchoninic acid assay (Pierce, Rockford, IL, USA). Samples were prepared containing 1 × sodium dodecyl sulphate (SDS) sample buffer and 30 µg protein, separated on a SDS–polyacrylamide gel (10%) and immunoblotted on to a nitrocellulose membrane (0.45 µm; Schleicher and Schüll, Dassel, Germany). Membranes were blocked with 5% dry milk in tris-buffered saline tween (TBST) for 30 min at room temperature (21°C) and incubated with the primary antibody at 4°C overnight followed by horseradish peroxidase-conjugated anti-mouse (1 : 2500), anti-rabbit (1 : 2500) or anti-rat IgG (1 : 2500), and visualized by enhanced chemoluminescence. The following antibodies were used: rabbit Kir4.1 (1 : 1000; Alomone Laboratories), mouse β-tubulin (1 : 5000; Sigma, Seelze, Germany), rabbit AQP4 (1 : 5000; Santa Cruz Biotechnology) and rat anti-GFAP (1 : 1000; Zymed, San Francisco, CA, USA). The specificity of the Kir4.1 antibody was assessed by co-incubation with the immunizing peptide (Alomone Laboratories). For peptide blocking, 5 µL of the Kir4.1 antibody alone or 5 µL of the Kir4.1 antibody (0.3 mg/mL) and a five-fold weight excess of blocking peptide (0.4 mg/mL) were preincubated in TBST on a rotator for 1 h at room temperature. Both preparations were centrifuged for 5 min at 13 000 g to remove possible aggregates. The supernatant was mixed with 4.5 mL 5% non-fat milk in TBST to get a 1 : 1000 dilution of the Kir4.1 antibody. Membranes were then probed and developed as described above. For quantification, membranes were digitally scanned and analysed. Densitometric analysis was performed using Quantity One® software (Bio-Rad, Hercules, CA, USA). Samples from SOD1G93A mice were compared with respective controls blotted on the same membrane and the relative density of bands was calculated as a percentage of the control value, which was set at 100%. Calculations were made only from blots in which the exposed bands of the standard lanes were below the saturation of the film.

Cell culture and cell viability testing

NSC34, a mouse embryonic spinal cord–neuroblastoma cell line with a motor neurone phenotype (Cashman et al. 1992; Durham et al. 1993) was used as an in vitro model system to investigate K+ toxicity in the spinal cord. The cell line was generously provided by Dr J. Weishaupt (University of Göttingen). Cells were cultured at 37°C under 5% CO2/humidified air in Dulbecco's modified Eagle's medium (DMEM) (PAA, Coelbe, Germany) supplemented with 10% heat-inactivated fetal calf serum (FCS) (PAA), 100 µg/mL streptomycin sulphate and 100 U/mL penicillin (Cashman et al. 1992; Durham et al. 1993). The medium was replaced every second day and cells were passaged once a week. To differentiate cells into a motor neurone-like shape, the medium was replaced by DMEM/Ham's F12 supplemented with 1% FCS, 100 μg/mL strentomycin sulphate and 100 U/mL pencillin and 1% non-essential amino acids (NEM). Eighty per cent confluent differentiated NSC34 cells were cultured in culture dishes on glass coverslips (15 mm diameter) at a density of 10 000 cells/cm3 and subsequently exposed to different concentrations of [K+]o. High [K+]o (7.5, 10 and 25 mm) was added to differentiated motor neurones after 3 days in culture and cells were exposed to the indicated K+ concentrations for either 24, 48, 72 or 120 h. In blocking experiments, cells were co-incubated with high K+ solution and the respective blocker. Neuronal injury was assessed quantitatively by staining the nuclei with 4′,6′-diamidino-2-phenylindole (DAPI). At the desired time points, cells were washed with PBS once for 5 min at room temperature (21°C). After washing with PBS, cells were fixed with 4% PFA/PBS at 24, 48, 72 and 120 hours after K+ supplementation. Cells were stained with DAPI using a ready-to-use mounting medium containing DAPI (Vectashield; Vector). The nuclei of dying cells were highly fluorescent and condensed compared with those of live cells. Neuronal cell death was scored by counting in three randomly chosen subfields of two coverslips with a 40 × objective for each sample. Results are expressed as mean ± SEM of at least three independent experiments. Analysis was verified by a second observer who was unaware of the culture conditions.


Kir4.1 channel is reduced in the ventral horn of the spinal cord and this loss precedes clinical symptoms in SOD1G93A animals

Earlier immunohistochemical data showed that Kir4.1 is widely expressed throughout the mouse spinal cord, predominantly in grey matter areas (Figs 1a and d inset). Given its abundant expression in the grey matter, we investigated whether Kir4.1 expression is affected in the SOD1G93A mouse model of ALS. In spinal cord sections of asymptomatic mice (60 days of age), Kir4.1 labelling was not different when SOD1G93A mutants were compared with non-transgenic (ntg) animals (not shown for 60 days of age). However, in spinal cord sections of early symptomatic mice (∼ 90 days of age), an apparent decrease in Kir4.1 immunofluorescence was detected in the ventral horn of the lumbar spinal cord, but the dorsal horn was not affected (Figs 1b and e). In end-stage mice, an almost complete loss of Kir4.1 immunolabelling was observed in the ventral horn of the spinal cord in cervical (not shown) as well as lumbar sections (Figs 1c and f). To determine whether Kir4.1 reduction is a consequence of post-transcriptional protein modifications, we used a Kir4.1 antibody raised against a slightly different peptide sequence (Kir4.1-PK). We observed a similar Kir4.1 reduction in SOD1G93A mutants with end-stage disease (Fig. 1g) as observed with the commercially available Kir4.1 antibody. Furthermore, we examined the cerebellum, a CNS area that is not typically involved in the pathogenic process in SOD1G93A mutants. Immunolabelling of cross-sections of the cerebellum of ntg and end-stage mice revealed a comparable Kir4.1 expression pattern in deep cerebellar nuclei (not shown) and the cerebellar cortex (Figs 1h and i).

Figure 1.

 Kir4.1 immunoreactivity is reduced in the spinal cord of SOD1G93A animals. Kir4.1 immunofluorescence labelling of cross-sections prepared from lumbar control spinal cord at 4 months of age showed abundant expression within spinal cord grey matter (a); dh, dorsal horn; vh, ventral horn. Kir4.1 immunoreactivity was reduced in the ventral horn of SOD1G93A transgenic mice in early symptomatic stages (b).  (c) In end-stage animals, Kir4.1 labelling was highly reduced in the ventral horn of the lumbar spinal cord, with the remainder showing a patchy staining pattern. Note the focal Kir4.1 loss in the ventral horn of end-stage SOD1G93A transgenic mice while the dorsal horn was only weakly affected. (d–f) Higher-magnification images of the ventral horn including lamina IX (L IX). Kir4.1 labelling of a control cervical spinal cord (d, inset of a). In early symptomatic animals an apparent loss of Kir4.1 immunoreactivity was detected in ventral horn grey matter (e, inset of b). (f) Inset of (c) of an end-stage animal. (g) Labelling of end-stage spinal cord sections using an antibody directed against a different amino acid sequence of the Kir4.1 protein showed a comparable loss of Kir4.1 channels in the ventral horn (inset represents control staining of a ntg spinal cord section). Kir4.1 expression in cerebellar sections was not different in ntg (h) and end-stage (i) mice. Scales bars 200 µm in (a–c, g) and 100 µm in (d–i).

We next examined Kir4.1 protein levels using an immunoblot assay of spinal cord lysates. Kir4.1 immunoblotting revealed multiple bands at high molecular weights, supposedly representing a complex consisting of the tetrameric channel and putative binding partners of the Kir4.1 channel subunit (∼200 kDa), such as aquaporin water channels, other Kir subunits or PDC domain-containing syntrophins (Hibino et al. 2004). A band at ∼160 kDa is supposed to represent the tetrameric form. A single band at ∼85 kDa is generally regarded as dimeric form of Kir4.1, whereas the monomer gives a band at ∼ 42 kDa (Fig. 2a). A profound reduction in Kir4.1 monomeric levels in spinal cord lysates (Fig. 2b) from separate lumbar and cervical sections was apparent in SOD1G93A end-stage mice (Fig. 2b). In some end-stage animals an almost complete loss of Kir4.1 in lumbar spinal cord lysates was observed (not shown). Kir4.1 channel protein levels were not decreased in 60-day-old asymptomatic animals; earliest protein reduction was observed at presymptomatic stages (∼ 80 days old) and levels decreased further as the disease progressed (Fig. 3a). Densitometric quantification of immunoblots revealed a progressive loss of Kir4.1 monomers that reached a maximum of ∼ 80% loss in end-stage transgenic mice compared with control values in ntg or wt-SOD1-overexpressing mice (Figs 2d and 3).

Figure 2.

 Kir4.1 protein levels in SOD1G93A transgenic mice are diminished compared with levels in ntg and wt-SOD1-overexpressing controls. (a) In adult tissue, the Kir4.1 antibody recognized several bands around ∼ 160–200 kDa, corresponding to its tetrameric form and putative higher-molecular weight complex partners (multimeric form), a band around 85 kDa that represents the dimeric form of Kir4.1, and a band at ∼ 42 kDa corresponding to its monomeric form. The specificity of the Kir4.1 antibody was tested by preabsorption with the Kir4.1 peptide (pept). All described bands disappeared upon peptide blockade. Note that in some cases a non-specific band appeared around 100 kDa that did not interfere with the bands of interest. (b) Loss of Kir4.1 immunoreactivity was paralleled by a reduction in Kir4.1 protein levels in immunoblots of SOD1G93A transgenic mice. In end-stage SOD1G93A transgenic mice a slight reduction in the multimeric form was observed in some cases, whereas the monomeric protein was consistently decreased. Cervical (cerv) spinal cord was compared with lumbar (lumb) spinal cord. Both spinal cord levels were affected in symptomatic (not shown) and end-stage animals (b) when tested for Kir4.1 channel proteins. Immunoblots of end-stage mice showed a lumbar to cervical gradient of Kir4.1 protein reduction. Data are representative of five independent experiments for the spinal cord of cervical versus lumbar levels. Tubulin was used as loading control. (c) Kir4.1 protein levels were no different in adult (100–120 days old) wt-SOD-overexpressing mice and ntg control mice. (d) Quantitative analysis of wt-SOD1-overexpressing mice, ntg control mice and end-stage G93A transgenic mice showed a significant reduction in Kir4.1 protein in G93A mice compared with the controls (n = 3 each). Values are mean ± SEM. **p < 0.01 (anova with Newman–Keuls post-test).

Figure 3.

 Time course of Kir4.1, AQP4 and GFAP expression in SOD1G93A mutants during disease progression. (a) No obvious differences were observed when Kir4.1 protein levels were investigated in asymptomatic (asympt.) SOD1G93A transgenic animals (∼ 60 days of age) compared with ntg control mice, whereas in presymptomatic (presympt.) mice (∼ 80 days of age) a reduction in the Kir4.1 immunoblot signal was observed. In early and late symptomatic (sympt.) as well as in end-stage mice, Kir4.1 monomeric protein levels were highly reduced. Data are representative of three independent experiments for asymptomatic and presymptomatic animals, and nine independent experiments for early symptomatic to end-stage animals. (b) Densitometry was used to quantify the relative expression levels of the Kir4.1 subunit. Density values are expressed as percentage of Kir4.1 monomer control levels from ntg littermates and results are combined from at least three independent experiments blotted on one membrane. A significant reduction of the Kir4.1 monomeric form was observed in presymptomatic SOD1G93A transgenic mice. Values are mean ± SEM. *p < 0.05, **p < 0.01 (anova with Newman–Keuls post-test). (c) Expression of AQP4 and GFAP proteins in SOD1G93A transgenic animals. The APQ4 antibody recognized a predominant band at ∼ 30 kDa. In SOD1G93A transgenic mice, AQP4 levels slightly increased with disease progression. Note that levels of immunoreactivity for the GFAP protein were increased compared with respective control levels, indicating reactive astrocytosis. Tubulin was used as loading control.

We next determined whether other major astroglial membrane proteins are affected in SOD1G93A mutants during the disease course. Reactive astrocytosis, as has been observed in SOD1G93A mutants, is accompanied by an increase in the number and change in phenotype of astrocytes. As a consequence, activated astrocytes may have distinct membrane properties and may express different membrane proteins. We chose to test SOD1G93A mutants for the expression of GFAP and AQP4 water channels. Owing to putative formation of high molecular weight membrane complexes, which include AQP4 channels and Kir4.1 in the retina and the hippocampus (Guadagno and Moukhles 2004; Pannicke et al. 2004), we carried out an expression analysis of AQP4 in SOD1G93A mice. According to the present view, AQP4 channels regulate K+-driven water flux (Amiry-Moghaddam et al. 2003; Pannicke et al. 2004) and may play a role in cytotoxic oedema (Saadoun et al. 2003). AQP4 expression in SOD1G93A mutants was slightly increased compared with that in ntg animals (Fig. 3c), regardless of the observed Kir4.1 protein reduction (Fig. 3a). As shown in earlier studies (Hall et al. 1998; Howland et al. 2002), we detected an increase in the astrocytic cytoskeletal protein GFAP in the course of the disease, indicating reactive astrocytosis (Fig. 3c).

Kir4.1 channels are expressed in glial cells in adult spinal cord grey matter and co-localize with AQP4 on astroglial endfeet

The Kir4.1 channel subunit was abundantly expressed in the grey matter of the adult mouse spinal cord (Fig. 4b). In contrast to Kir4.1 expression in early postnatal animals, no obvious Kir4.1 labelling was observed in white matter of adult control mice, indicating that channel expression is developmentally regulated and shifts from the oligodendroglial to astroglial lineage during postnatal development (Neusch et al. 2001; Kalsi et al. 2004). GFAP-EGFP mice that expressed EGFP under control of the GFAP promoter were used as a tool to visualize astrocytes and their extensive process network, including glial endfeet structures (Fig. 4a). Kir4.1 immunoreactivity was localized predominantly to the processes (Fig. 4d and insets) in adult mice whereas motor neurone labelling was not apparent. Kir4.1 expression was most abundant in astroglial branches (Figs 5a–d, arrowheads) and surrounding capillaries (Figs 5e–g), but only weak labelling was detected on astroglial cell bodies (arrow in Figs 5a–d). In adult spinal cord grey matter, AQP4 expression was found predominantly on astroglial endfeet surrounding the microvasculature (Fig. 5c) as has also been observed in recent studies by other groups (Rash et al. 1998; Manley et al. 2004; Oshio et al. 2004; Vitellaro-Zuccarello et al. 2005). Double-immunolabelling of Kir4.1 and AQP4 on GFAP-EGFP transgenic mice revealed co-localization of both membrane proteins on astrocytic endfeet that are in close contact with the microvasculature (Fig. 5d).

Figure 4.

 Predominant expression of the Kir4.1 channel subunit on astroglial cells in the adult mouse spinal cord. (a–d) Cross-sections of the ventral horn of the cervical spinal cord from adult control mice. Motor neurones were labelled with the neuronal marker Nissl (c, red) and astrocytes were visualized using transgenic mice expressing EGFP under control of the GFAP promoter (a, green). The Kir4.1 channel subunit (magenta) was expressed predominantly in the grey matter, including the ventral horn (b).  (d) Merged images of (a–c). The insets in (d) represent higher magnification of the indicated ventral horn area. Kir4.1 immunolabelling (magenta) was diffusely distributed in grey matter areas adjacent to, but sparing large motor neuronal cell bodies (red). The overlap of all three high-magnification images indicated that Kir4.1 overlaps with EGFP fluorescence of astrocytes (green). Note that Nissl stain also labelled glial nuclei. Scale bars 50 µm in (a–c) and 40 µm in (d).

Figure 5.

 Subcellular distribution of Kir4.1 in adult mouse spinal cord and co-localization with aquaporin water channels to form putative multimeric complexes. (a–d) Kir4.1 immunolabelling (b, red) showed clustering on astrocytic endfeet (a, green) contacting small vessels in the ventral horn spinal cord. (c) AQP4 (magenta) immunoreactivity on EGFP/GFAP-positive astrocytes predominantly labelling astrocytic endfeet. (d) Merged images of (a–c). Arrowheads point to an astrocytic endfoot co-labelled by Kir4.1 and AQP4. Note that the astrocytic cell body (arrow) was only weakly labelled by Kir4.1 and AQP4, whereas the expression of both membrane proteins was predominantly targeted to the astrocytic network and astrocytic endfeet. (e–g) Magnified view of an EGFP-positive astrocytic endfoot (e, green) in the cervical spinal cord of a 120-day-old control mouse surrounding a vessel (v) in cross-section. Kir4.1 (f, magenta) was diffusely expressed on astrocytic branches and clusters on capillaries overlapping with the EGFP signal. (g) Merged images (e) and (f). Scale bars 10 µm in (a–d) and 2 µm in (e–g).

Increasing extracellular K+ levels induce motor neurone death in the NSC34 cell line

Earlier work and the above data suggest that Kir4.1 channels are implicated in extracellular K+ regulation. To understand putative toxic effects of extracellular K+ increases, we performed cell viability tests by challenging a classical motor neurone cell line (NSC34) with increasing extracellular K+ concentrations. Differentiated cells were cultured in control medium, and then shifted to a medium containing 7.5, 10 or 25 mm KCl. Cell death was investigated at 24, 48, 72 and 120 h following the shift to high K+ by determining the number of pyknotic or fragmented nuclei visualized by DAPI staining. NSC34 cells chronically exposed for 120 h showed significantly increased cell death when exposed to high K+ conditions compared with NSC34 cells grown under physiological conditions (Fig. 6a). At 24, 48 and 72 h increasing [K+]o did not significantly alter the extent of motor neurone cell death. However, exposure to increasing [K+]o for 120 h induced cell death in a concentration-dependent manner. Maximal sensitivity to external K+ was reached at 120 h using 10 and 25 mm KCl (Fig. 6b). Thus increasing [K+]o altered motor neurone survival in vitro in a time- and concentration-dependent manner. In another set of experiments we tested the effect of a Kir channel blocker on K+-induced motoneurone death. Co-incubation with low Ba2+ (1 µm) significantly blocked the K+-induced cell death, indicating that cell death mechanisms are partly a consequence of K+ influx into NSC34 cells (Fig. 6c).

Figure 6.

 Increasing concentrations of extracellular K+ induce cell death in NSC34 motor neurones. (a) Effect of 7.5, 10 and 25 mm KCl on differentiated motor neurones in vitro. After 24, 48, 72 and 120 h cells were fixed, stained with DAPI and visualized by fluorescence microscopy. (a) Example of NSC34 motor neurones exposed to the indicated KCl concentrations at 120 h. Note the increasing number of fragmented nuclei with increasing K+ concentrations (arrows) in the DAPI counterstain. The inset in (a) at 25 mm KCl represents a fragmented nucleus. Scale bar 10 µm. (b) Quantitative analysis of the effects of external K+ on motor neurone survival. Significant cell death was observed after exposure of NSC34 cells for 120 h using at least 10 mm KCl. (c) K+-induced cell death of NSC34 motoneurones was blocked by co-incubation with the Kir channel blocker Ba2+. Cell counting was performed after 120 days in culture differentiate. Cells were co-incubated with 10 mm KCl and 25 mm KCl and 1 µm Ba2+ from day 3 onwards. Values are mean ± SEM of 5–7 experiments. *p < 0.05, **p < 0.001 (anova).


The present study provides the first evidence that an astroglial K+ channel is implicated in the pathogenic process in FALS-associated SOD1G93A mutant mice. First, we showed that Kir4.1 channel expression is focally decreased in the ventral horn of the SOD1G93A spinal cord coinciding with presymptomatic stages and that this decrease progresses when animals become more seriously affected. Second, we demonstrated that glial Kir4.1 channels are expressed on astroglial cells and target to the microvasculature, but that Kir4.1 is not expressed in motor neurones of adult mice. Third, our data indicate that Kir4.1 channel expression is co-localized with AQP4 channels and clusters around the microvasculature of the spinal cord. This provides indirect evidence that a Kir4.1/AQP4 complex may perform tasks such as buffering K+ and water away from the extracellular space to extrude it into blood vessels in the spinal cord, a mechanism that has been reported in the retina and hippocampus (Newman 1986; Nagelhus et al. 1999; Kofuji et al. 2002; Amiry-Moghaddam et al. 2003; Dalloz et al. 2003; Saadoun et al. 2003; Guadagno and Moukhles 2004; Pannicke et al. 2004). Furthermore, our functional assay using the NSC34 motor neurone cell line indicated that potential extracellular K+ increases induce motor neurone death in a concentration-dependent manner and at concentrations that can be reached under physiological conditions.

Functional implications of Kir4.1 channel expression in the adult mouse spinal cord

Astroglial Kir channels are implicated in maintaining low [K+]o levels in neurally active tissue. Kir channels are regarded as one molecular substrate of the K+-buffering process in the retina and in the hippocampus, as indicated in part by Kir4.1 deletion studies (Newman et al. 1984; Newman 1986, 1987, 1993; Newman and Reichenbach 1996; Kofuji et al. 2000, 2002; Neusch et al. 2001; Amiry-Moghaddam et al. 2003; Oshio et al. 2004). The ‘polarized’ distribution of Kir4.1 in the spinal cord is appropriate for a regulatory role in extracellular K+ homeostasis: (i) Kir4.1 channel expression forms an extensive network closely surrounding motor neurones in the grey matter of the spinal cord and is localized predominantly in astrocytes in adult tissue; (ii) clustered Kir4.1 expression is detected on spinal cord capillaries, supporting the hypothesis that Kir channels serve the role of extruding K+ into the microvasculature; and (iii) Kir4.1 co-localizes with a water channel (AQP4) on capillaries to release K+ and water into the blood flow. This ‘polarized’ expression pattern is similar to that observed in the retina and hippocampus, and is suited for a role in taking up K+ from the extraneuronal space and releasing it into capillaries in the spinal cord. In the early postnatal days, Kir4.1 channels are expressed on cells of the oligodendrocytic lineage and the impact on spinal cord development has been reported in an earlier study (Neusch et al. 2001). Astroglial Kir4.1 expression is evident from the second postnatal week of development and increases in adulthood in the spinal cord and other CNS organs, such as the optic nerve (Kalsi et al. 2004). In the adult spinal cord, we did not observe significant Kir4.1 labelling in the white matter, indicating that white matter structures such as axons are not directly affected by loss of Kir4.1 channels. Whether oligodendrocytes of the grey matter express Kir4.1 has not been shown conclusively and the impact on disease progression in SOD1G93A mutant mice remains unknown.

Impact of Kir4.1 loss on disease progression in SOD1G93A mice

The observed loss of the Kir4.1 channel subunit may have major implications for glial cell function, for the transmembrane gradient of transported molecules and for the ionic homeostasis of the perineural space. Genetic Kir4.1 inactivation induces early postnatal glial cell death and probably secondary motor neurone degeneration in the spinal cord grey matter in an in vivo paradigm (Neusch et al. 2001). Two major K+-selective membrane proteins maintain extracellular K+ balance in the CNS, the Na+, K+-ATPases and Kir channels (D'Ambrosio et al. 2002). Among glial Kir channel subunits, the Kir4.1 channel is the functionally prominent subunit in brainstem and spinal cord astrocytes. Tetrameric K+ channels formed by Kir4.1 subunits represent the principle channel mediating resting K+ conductance in vitro and in situ (Olsen et al. 2005; Neusch et al. 2006), and at least three α-subunits and three β-subunits of the Na+, K+-ATPases have been reported (Dolapchieva 1998; Ellis et al. 2003) on neuronal as well as on glial cells. The individual contribution of each of these K+-transporting molecules to extracellular K+ regulation is not clear. Specific pharmacological blockade of Kir channels in situ, for example in the hippocampus, increases [K+]o by a factor of 1–2 or even more (Gabriel et al. 1998; Jauch et al. 2002). Loss or blockade of Kir channels with subsequent extracellular K+ accumulation may therefore have an impact on normal functioning of surrounding neurones, as previously reported for epileptic hippocampal tissue (Hinterkeuser et al. 2000; Schroder et al. 2000; Huttmann et al. 2003). However, active Na+, K+-pumps are also involved in clearing local extracellular K+ increases due to neuronal activity and may therefore compensate for the loss of Kir4.1 channels. Interestingly, inactivation of Na+, K+-ATPases has been reported in the SOD1G93A mutant mouse model (Ellis et al. 2003). Therefore, both K+-selective membrane proteins that are suggested to control external K+ levels are affected in the SOD1G93A mutant mouse model. Accumulation of K+ in perineuronal compartments of the spinal cord thus seems a likely mechanism of motor neurone excitotoxicity, although there are technical obstacles to the direct detection of K+ in local perisynaptic compartments of the spinal cord. As a result of extracellular accumulation of K+, surrounding neurones and synaptic membranes will become chronically depolarized followed by uncontrolled Ca2+ influx with excitotoxic effects on neurones. To test this hypothesis and to understand the effects of high external K+ on motor neurones, we employed the NSC34 motor neurone cell line. High [K+]o led to a significant increase in motor neurone cell death in concentrations that can be reached under physiological conditions, indicating that motor neurones are vulnerable to increasing extracellular K+ levels. Impaired K+ buffering with consecutive extracellular K+ increases may thus lead to dysfunction of motor neurones or even directly induce motor neurone death in vivo.

Another general function of glial Kir channels is setting the cell membrane potential. This in turn governs the transmembrane gradients of many transported molecules, such as glutamate and other neurotransmitters. Loss of the Kir4.1 channel subunit or its pharmacological blockade depolarizes glial cell membranes in vitro, as demonstrated in a Kir4.1 knockout mouse (Neusch et al. 2001). A chronically depolarized glial membrane potential may, as a consequence, inactivate and degrade voltage-dependent membrane transporters, such as GLT1 or Na+, K+-ATPases (Bordey and Sontheimer 2003). This in turn would increase extracellular, perineural glutamate concentrations in addition to increases in [K+]o. Loss of GLT1 represents a pathogenic mechanism that has been demonstrated in several SOD1 mouse mutant lines as well as in human tissue (Rothstein et al. 1992, 1995; Trotti et al. 1999; Howland et al. 2002). The molecular mechanism of Kir4.1 channel reduction is not yet clear. Loss of GLT1 has been attributed to intracellular oxidative reactions triggered by hydrogen peroxide in some SOD1 mutant cells in vitro (Trotti et al. 1999), and to increased GLT1 internalization and degradation in SOD1G93A mutant cell lines (Vanoni et al. 2004). In our study, Kir4.1 channel reduction predominantly affected the monomeric form of Kir4.1, a phenomenon that has also been reported for the GLT1 protein reduction in ALS mouse models. We did not observe a significant reduction in multimeric protein complexes in the high molecular weight range. This is most likely due to a tight linkage of Kir4.1 channels to other complex partners, such as aquaporins, dystrophins and syntrophins, which renders the complex insoluble under SDS denaturating conditions.

Whether the observed loss of astroglial Kir4.1 channels is a specific consequence of the SOD1 mutation, an early event in reactive astrocytes or a combination of both mechanisms remains unclear. However, loss of Kir4.1 channels has not been observed previously in other neurodegenerative paradigms, suggesting that astrogliosis is not a primary cause of Kir4.1 channel reduction.

In conclusion, we suggest that loss of astroglial Kir4.1 channels represents a novel pathogenic factor that induces motor neurone dysfunction and probably motor neurone degeneration by hampering glial K+ uptake in the spinal cord of the SOD1G93A mouse model of ALS.


This work was supported by grants from the Deutsche Forschungsgemeinschaft (NE-767/3-1,3 to CN and SFB 581, TP B4 to BH). The Kir4.1 antibody (Kir4.1-PK) was generously provided by Paulo Kofuji, University of Minnesota, MN, USA).