|Home | About | Journals | Submit | Contact Us | Français|
Abnormalities in astrocytes occur in the brains of patients with Tuberous Sclerosis Complex (TSC) and may contribute to the pathogenesis of neurological dysfunction in this disease. Here, we report that knock-out mice with Tsc1 gene inactivation in glia (Tsc1GFAPCKO mice) exhibit decreased expression of the astrocytic connexin protein, Cx43, and an associated impairment in gap junction coupling between astrocytes. Correspondingly, hippocampal slices from Tsc1GFAPCKO mice have increased extracellular potassium concentration in response to stimulation. This impaired potassium buffering can be attributed to abnormal gap junction coupling, as a gap junction inhibitor elicits an additional increase in potassium concentration in control, but not Tsc1GFAPCKO slices. Furthermore, treatment with a mammalian target of rapamycin inhibitor reverses the deficient Cx43 expression and impaired potassium buffering. These findings suggest that Tsc1 inactivation in astrocytes causes defects in astrocytic gap junction coupling and potassium clearance, which may contribute to epilepsy in Tsc1GFAPCKO mice.
Tuberous sclerosis complex (TSC) is a multisystem genetic disease caused by mutation of either the TSC1 or TSC2 gene and typically featuring disabling neurological symptoms, such as mental retardation, autism, and seizures (Kwiatkowski, 2003; Crino et al., 2006; Holmes et al., 2007). Epilepsy in TSC is particularly severe and refractory to available treatments (Sparagana et al., 2003; Holmes et al., 2007). The pathological hallmark of brains from TSC patients are areas of disrupted cortical lamination, termed tubers, which may represent the epileptogenic foci for seizures. Neuropathological studies demonstrate abnormalities in astrocytes in cortical tubers, such as astrocytosis and abnormally differentiated giant cells with glial, as well as neuronal, features (Crino, 2004; Sosunov et al., 2008; Talos et al., 2008), suggesting that astrocyte dysfunction may be centrally involved in epileptogenesis and other neurological deficits in TSC.
Recent studies have demonstrated novel physiological roles of glial cells in regulating neuronal excitability and preventing epilepsy (Binder and Steinhauser, 2006; Jabs et al., 2008; Wetherington et al., 2008). For example, astrocytes maintain normal homeostasis of excitatory substances, such as extracellular potassium. The immediate uptake of extracellular potassium is at least partially dependent on specific potassium channels and sodium-potassium pumps of astrocytes (Haglund and Schwartzkroin, 1990; Xiong and Stringer, 2000; D’Ambrosio et al., 2002). Furthermore, potassium homeostasis is also regulated by spatial buffering via astrocyte networks, in which potassium released at sites of high concentration is redistributed to distant sites of lower potassium concentration (Kofuji and Newman, 2004). This spatial buffering of potassium involves coupling of astrocytes through gap junctions. Studies of both animal models and human epilepsy demonstrate a variety of abnormalities in potassium buffering mechanisms of astrocytes that may predispose to neuronal hyperexcitability and seizures (Janigro et al., 1997; Bordey and Sontheimer, 1998; Gabriel et al., 1998a,b; Xiong and Stringer, 1999; Hinterkeuser et al., 2000). Changes in expression of astrocytic gap junction proteins have been reported in different epileptic animal models and human brain specimens (Naus et al., 1991; Elisevich et al., 1997; Sohl et al., 2000; Aronica et al., 2001; Fonseca et al., 2002). Furthermore, mice deficient in astrocytic gap junction proteins exhibit reduced thresholds for generating epileptiform activity (Wallraff et al., 2006), suggesting that spatial buffering mechanisms of astrocytes are important in preventing epilepsy.
Our previous work in a mouse model of TSC (Tsc1GFAPCKO mice) has demonstrated that inactivation of the Tsc1 gene in glia results in the development of seizures (Uhlmann et al., 2002). This epilepsy phenotype is associated with increased neuronal excitability in response to elevated extracellular potassium and decreased expression and function of astrocytic potassium channels (Jansen et al., 2005). In the present study, we tested the hypothesis that Tsc1-deficient astrocytes exhibit impaired gap junction coupling and spatial potassium buffering that may contribute to epileptogenesis in Tsc1GFAPCKO mice.
All experiments were conducted according to an animal protocol approved by the Washington University Animal Studies Committee. Tsc1flox/flox-GFAP-Cre knock-out (Tsc1GFAPCKO) mice with conditional inactivation of the Tsc1 gene in GFAP-positive cells starting around embryonic day 14.5 were generated as described previously (Uhlmann et al., 2002). Tsc1flox/+-GFAP-Cre and Tsc1flox/flox littermates, which have been shown to have normal phenotypes, were used as control animals in these experiments.
Western blotting was performed on 4–5 week old Tsc1GFAPCKO mice and control mice to detect expression of astrocyte specific gap junction proteins, Connexin 43, 30, and 26, as well as the neuronal specific Connexin 36, using standard methods as described previously (Zeng et al., 2008). In additional experiments, 2–3 week old mice were used to assay Connexin 43 expression at a younger age that should precede seizure onset in these mice (Erbayat-Altay et al., 2007). In brief, neocortex and hippocampus were dissected, sonicated, and centrifuged. Equal amounts of total protein extract were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes. After incubating with primary antibodies specific to Cx43 (1:1000; ZYMED laboratories, San Francisco, CA), Cx26 (1:250; ZYMED laboratories), Cx30 (1: 250; ZYMED laboratories), Cx36 (1: 250; ZYMED laboratories), and actin (1:5,000; Sigma), the membranes were reacted with a peroxidase-conjugated secondary antibody. Signals were detected by using enzyme chemiluminescence reagent (Pierce, Rockford, IL) and quantitatively analyzed with ImageJ software. In a separate set of experiments, rapamycin (3mg/kg/d, i.p), or vehicle, was given to 2 week old Tsc1GFAPCKO mice for 3 weeks to inhibit mammalian target of rapamycin (mTOR) pathway activation (Zeng et al., 2008), and Western blotting was then performed to assess the expression of Cx43, as described above. In other experiments, immunohistochemistry for connexin 43 labeling was performed on hippocampal sections of 4–5 week old Tsc1GFAPCKO and control mice, using connexin 43 antibody (1:100, Zymed); general methods for processing tissue are described below.
Horizontal entorhinal cortex–hippocampal slices were prepared from 2 to 3 week old control or Tsc1GFAPCKO mice, using standard methods as described previously (Wong and Yamada, 2000). In brief, horizontal slices (400 μm) were cut with a vibratome, while submerged in ice-cold, oxygenated (95% O2/5% CO2) solution containing (in mM): 87 NaCl, 2.5 KCl, 1.25 NaH2PO4, 7 MgCl2, 0.5 CaCl2, 25 NaHCO3, 25 glucose, and 75 sucrose. Slices were maintained at room temperature for at least 1 hour before use in an oxygenated (95% O2/5% CO2) artificial cerebrospinal fluid (aCSF) containing (in mM): 125 NaCl, 2.5 KCl, 2CaCl2, 1 MgCl2, 1.25 NaH2PO4, 26 NaHCO3, 10 glucose, pH 7.4. Slices were then transferred to a submerged recording chamber and perfused with oxygenated aCSF at ~32°C. Astrocytes in the stratum radiatum of the CA1 region were directly visualized with infrared differential interference contrast (IR-DIC) optics on a Nikon upright microscope. Astrocytes were identified by their characteristic morphology, location within stratum radiatum, negative resting membrane potentials (less than −70 mV), and lack of voltage-activated sodium currents. Furthermore, only cells with time- and voltage-independent currents responses were included for further study (GluT cells from Matthias et al., 2003; Wallraff et al., 2004; linear cells from D’Ambrosio et al., 1998), as these represent the subtype of astrocytes that exhibit gap junction coupling (Wallraff et al., 2004). Under whole-cell voltage clamp mode, electrical signals were acquired through a patch pipette, processed with an Axopatch-200B amplifier, digitized at 10 kHz, and stored and analyzed with pCLAMP software (Molecular Devices, Sunnyvale, CA).
To assay astrocyte coupling, single astrocytes were filled with biocytin through the patch pipette. Patch pipettes (4–7 MΩ) were filled with an internal solution containing the following (in mM): 140 K-gluconate, 4 NaCl, 1 CaCl2, 10 HEPES, 10 EGTA, and 2 Mg-ATP. In addition, biocytin (Nε-biotinyl-L-lysine; Sigma, St. Louis, MO) was dissolved at 0.5–0.6% in the internal solution. After obtaining whole-cell mode, the biocytin was allowed to diffuse into the cell for exactly 20 minutes. Current-voltage plots were obtained and input resistance was monitored – any cells showing a greater than 20% change in resistance during the 20 minute recording period were discarded. Only one cell per slice was injected with biocytin. After the 20 minute labeling period, slices were removed from the recording chamber and immersion-fixed in a solution of 4% paraformaldehyde and 0.1% glutaraldehyde in 0.1 M sodium phosphate buffer (PB), pH 7.4, for 4–12 hr at 4°C. The slices were rinsed in 0.1 M PB and then infiltrated with 10% sucrose in 0.1 M PB for 1hr, followed by 30% sucrose for 8–12 hr. For further analysis, the 400 μm slices were then sectioned into 40μm-thick sections using a microtome. Sections were rinsed in 0.1 M PB and then in 0.1 M Tris-HCl buffer (TB), pH 7.4. The sections were then incubated with 1% H2O2 in 0.1 M TB for 2hr to suppress endogenous peroxidase. Sections were then treated with 2% bovine serum albumin (BSA; Boehringer Mannheim, Indianapolis, IN), 0.25% dimethylsulfoxide (DMSO; Sigma, St. Louis, MO), and 0.05 M Tris-buffered saline (TBS), pH 7.4, for 1hr to reduce nonspecific background staining and to permeabilize membranes. Sections were rinsed in 0.1 M TBS for 30min and then incubated with an Elite ABC kit (Vector Laboratories, Burlingame, CA), diluted 1:500 in 0.5% BSA, 0.25% DMSO, and 0.05 M TBS for 36–48 hr at 4°C. Sections were then rinsed thoroughly in 0.1M TBS followed by 0.1 M TB, pH 7.6, and incubated in SIGMAFAST™ DAB with metal enhancer tablet set (D-0426, Sigma, St. Louis, MO) for 10–15 min. The reaction was stopped by rinses in 0.1 M TB, and the sections were mounted on gelatin-subbed slides and imaged with a Zeiss Axioskop microscope and AxioCamHR digital camera (Carl Zeiss MicroImaging Inc., Thornwood, NY). All DAB-positive cells were counted in all sections.
To determine astrocyte density, brains from 2- to 3-week old control and Tsc1GFAPCKO mice were perfusion-fixed with 4% paraformaldehyde and cut into 50μm sections with a vibratome. Sections were labeled with GFAP antibody (rabbit anti-mouse; 1:500; Sigma, St. Louis, MO) and then anti-rabbit Alexa Fluor 488 (1:800; Invitrogen, Carlsbad, CA, and mounted in Vectashield (Vector Laboratories). Stacks of optical sections (2 μm intervals) were acquired through the entire depth of a slice with a Zeiss LSM PASCAL confocal microscope (Zeiss, Thornwood, NY). From merged images from the stack, GFAP-immunoreactive cells in a 200 × 200 × 50 μm region from the stratum radiatum of hippocampus were counted from four sections per mouse from a total of four mice per group, to calculate astrocyte density.
Hippocampal slices (400μm) were prepared with a vibratome from control or Tsc1GFAPCKO mice, placed in a submerged recording chamber, and perfused with oxygenated (95% O2/5% CO2) aCSF at a rate of ~ 2 ml/min at ~32°C, as described above. To block synaptic transmission, 50 μM D,L-APV, 10 μM CNQX, and 100 μM bicuculline were also added to the aCSF, and slices were equilibrated for 1 h. Double-barreled glass microelectrodes were constructed to allow simultaneous recording of potassium concentration and extracellular field potentials, as previously described (Meeks and Mennerick, 2007). In brief, theta glass capillaries (1.5mm, Sutter Instrument, Novato, CA) were pulled to a tip diameter of 1–5 μm with a horizontal puller. One barrel was silanized with Sigmacote (Sigma, St.Louis, MO) vapor. Immediately after silanization, the capillaries were baked at 120°C for 15–30 minutes. To create a potassium ion-sensitive barrel, the silanized barrel was tip filled with the K+-selective ionophore IE-190 (World Precision Instruments, Sarasota, FL) and then back-filled with 500 mM KCl. The non-silanized reference barrel was back-filled with 140 mM NaCl. The recording electrode was placed between the stratum pyramidale and stratum oriens in the in the CA1 region at a depth of approximately 100 μm. Voltage signals from both barrels were recorded with an Axoclamp 2B amplifier in bridge mode and pClamp software. The signal from the reference barrel was subtracted from the signal from the potassium-selective barrel, and the resulting signal was correlated with the subsequent calibration curves to calculate potassium concentration (see below). The signal from the reference barrel was also used to measure extracellular field potentials in response to stimulation. A bipolar stimulation electrode was placed in the alveus of the CA1 region near the subiculum to trigger potassium release from antidromic action potentials by electrical stimulation, using trains of stimuli (0.1 ms pulses, 20 Hz for 10 s). After obtaining an input-output response to a range of stimulation intensities, the amplitude of stimulation was adjusted to evoke population spike amplitudes at ~50% and 100% of maximal response to produce comparable stimulation intensities between different experiments. After recording, the electrode was calibrated in the standard aCSF containing varying K+ concentrations (2.5, 4, 10, 25, and 50 mM). A log-linear fit to the [K+]0. responses was used to calibrate the electrodes and calculate [K+]0. Data from an experiment were not used for this study unless the electrodes displayed voltage changes of at least 40 mV per ten-fold increase in [K+]0. The maximal increase in [K+]0 following stimulation and the subsequent decay time (t1/2 − time for [K+]0 to decay to half the maximum value; τ − a weighted time constant from a two exponential fit) was calculated for each experiment, using pCLAMP software. To control for variability in the stimulation and neuronal activation of potassium release, the absolute rise in potassium concentration was normalized to the population spike amplitude (measured from base to peak) at each level of stimulation. In addition to comparing these parameters between control and Tsc1GFAPCKO mice, to investigate if spatial buffering through gap junctions contributed to extracellular potassium concentration, we also compared the [K+]0 changes in the hippocampal slices before and after perfusion with the gap junction inhibitor carbenoxolone (100μM).
Data are expressed as mean values ± SEM. Student’s two-tailed t test was used for quantitative comparisons between two groups and ANOVA for comparisons for more than two groups, with Tukey multiple comparisons post-tests. Statistical significance was defined as p<0.05.
Connexin 43 (Cx43) is the predominant gap junction protein expressed in astrocytes and involved in astrocytic coupling, although other connexins (Cx26, Cx30) in astrocytes have also been reported (Nagy and Rash, 2000). By using western blotting, we found that Tsc1GFAPCKO mice have decreased Cx43 expression in both neocortex and hippocampus compared with control mice (Figs. 1A and C). In contrast, expression of Cx36, a major neuronal-specific gap junction protein, appeared to be slightly increased in Tsc1GFAPCKO mice, although this was not statistically significant (Figs. 1B and D). Furthermore, there were no significant differences in the astrocytic gap junction proteins,Cx30 and Cx26, between Tsc1GFAPCKO and control mice (data not shown). Immunohistochemical studies also confirmed a decrease in Connexin 43 expression in Tsc1GFAPCKO mice (Supplemental Fig. 1A).
Although video-EEG was not directly performed in these studies, the abnormal Cx43 expression was correlated to the onset and evolution of epilepsy in Tsc1GFAPCKO mice as documented in detail in previous video-EEG studies (Erbayat-Altay et al., 2007). The reduction in Cx43 expression was observed in Tsc1GFAPCKO mice not only at 4–5 weeks old (Fig. 1), which is around the age seizures start in these mice (Erbayat-Altay et al., 2007), but also at 2–3 weeks of age (Supplemental Fig. 1B), which should precede onset of seizures.
In an attempt to identify a mechanistic link between Tsc1 gene inactivation and connexin expression in astrocytes, the effect of the mTOR inhibitor, rapamycin, on Cx43 expression in Tsc1GFAPCKO mice was tested, as the mTOR pathway can regulate protein synthesis (Sandsmark et al., 2007). Rapamycin has previously been shown to reverse the abnormal hyperactivation of the mTOR pathway in Tsc1GFAPCKO mice (Zeng et al., 2008). Pretreatment with rapamycin prevented the decreased Cx43 expression in Tsc1GFAPCKO mice, but had no effect on control mice (Fig. 2), indicating that the mTOR pathway does mediate the effects of Tsc1 gene inactivation on Cx43 expression.
Since Tsc1GFAPCKO mice exhibited decreased levels of the astrocytic gap junction Cx43 expression, we next examined astrocyte gap junction coupling by assessing the extent of transcellular diffusion of biocytin from single, patch-clamped astrocytes in stratum radiatum of the CA1 of hippocampal slices. With whole-cell recording, only cells with time- and voltage-independent currents responses in response to voltage steps were included for further study (see Fig. 3E), as these represent the subtype of astrocytes that exhibit gap junction coupling (Wallraff et al., 2004). There was no significant difference between Tsc1GFAPCKO and control mice with respect to resting membrane potential and input resistance of astrocytes (Vm = −73.2 ± 1.5 mV, Rm = 2.4 ± 0.6 MΩ, n = 20 KO astrocytes; Vm = −69.4 ± 1.6 mV, Rm = 2.2 ± 0.2 MΩ, n = 16 control astrocytes). However, the number of coupled astrocytes as detected by biocytin staining was significantly less in hippocampal slices from Tsc1GFAPCKO mice compared to control mice (Figs. 3A, B, and F). In Tsc1GFAPCKO mice, biocytin spread from single injected astrocytes to 118 ± 57 encircling cells (n=20 injections) after 20 minutes of whole-cell recording, as opposed to 243 ± 87 cells in control mice (n=16 injections), indicating a significant impairment of astrocytic gap junction coupling in Tsc1GFAPCKO mice (Fig. 3F, p<0.001 by t-test). This decrease in astrocyte coupling in Tsc1GFAPCKO mice was observed, despite a very similar density of GFAP-positive astrocytes between Tsc1GFAPCKO and control mice (Fig 3C and D; 14,550 ± 1,220 cells/mm3, n=4 Tsc1GFAPCKO mice; 14,650 ± 1,250 cells/mm3, n=4 control mice), indicating that the changes in astrocyte coupling and Cx43 expression are not related to changes in astrocyte number that develop later in Tsc1GFAPCKO mice (Uhlmann et al., 2002).
The decrease in astrocyte coupling in the hippocampal slices from Tsc1GFAPCKO mice did not have a completely uniform spatial laminar distribution. When comparing parallel ~40 μm sub-sections of the original 400 μm hippocampal slice, the number of biocytin-positive cells decreased in sections more distant from the injected astrocyte in both Tsc1GFAPCKO and control mice and was consistently lower in Tsc1GFAPCKO in all sections (Fig. 3G). However, when comparing different strata within the hippocampus, control mice exhibited an expected decrease in coupled cells in the more distal stratum lacunosum-moleculare compared to the stratum radiatum, whereas Tsc1GFAPCKO mice had equal numbers of coupled cells in these two strata (Fig. 3H). Thus, compared to control mice, the number of coupled cells in Tsc1GFAPCKO mice was only decreased in the stratum radiatum, not the stratum lacunosum-moleculare. Although the cause or functional consequences of these laminar-specific differences were not specifically investigated in this study, laminar differences in spatial potassium buffering have previously been reported in connexin knock-out mice (Wallraff et al., 2006) and helped target our electrode placement in the subsequent potassium concentration studies.
To determine whether the decreased astrocyte gap junction coupling in Tsc1GFAPCKO mice might affect potassium buffering properties, we measured stimulus-evoked extracellular potassium concentrations ([K+]0) in hippocampal slices from Tsc1GFAPCKO and control mice using potassium selective electrodes. Increases in [K+]0 were elicited by alvear stimulation of CA1 pyramidal neurons with a train of stimuli (10s, 20Hz) at stimulus intensities to elicit 50% and 100% of maximal population spike amplitude. Maximal field potential amplitudes did not differ significantly between control and Tsc1GFAPCKO mice (5.5 ± 0.5mV, n=14 slices from 8 Tsc1GFAPCKO mice; 7.0 ± 1.0 mV, n=12 slices from 8 control mice). In response to stimulation, an increase in [K+]0 reached a peak immediately following the end of the stimulus train and decayed over seconds (Fig. 4A). Higher stimulation intensities (to elicit 100% of the maximal population spike) induced larger increases in [K+]0 compared to lower stimulation (to elicit 50% of the maximal population spike) in both control and Tsc1GFAPCKO mice (Fig. 4B). However, with both lower and higher stimulation levels, the peak increase in [K+]0 was significantly great in Tsc1GFAPCKO mice compared to controls (1.05 ± 0.12 mM in 14 Tsc1GFAPCKO slices versus 0.64 ± 0.12 mM in 12 control slices for 50% stimulation, p<0.05; 2.57 ± 0.17 mM in 14 Tsc1GFAPCKO slices versus 1.25± 0.12 mM, in 12 control slices for 100% stimulation, p<0.001) (Fig. 4B). Since higher stimulation intensities cause greater activation of neurons and more potassium release, the increases in [K+]0 was normalized to field potential amplitude to evaluate effects on [K+]0 independent of the degree of neuronal activation. After normalization, there were still significant differences between Tsc1GFAPCKO and control mice in normalized [K+] rises at both 50% (0.46 ± 0.06 mM/mV, n=14 Tsc1GFAPCKO slices; 0.23 ± 0.05mM/mV, n=12 control slices, p<0.01) and 100% (0.49 ± 0.04 mM/mV, n=14 Tsc1GFAPCKO slices; 0.21 ± 0.03 mM/mV, n=12 control slices, p<0.001) stimulation intensity (Fig. 4C), indicating a deficiency in the [K+]0 clearance in the Tsc1GFAPCKO mice. While the peak increase in [K+]0 was larger in hippocampal slices from Tsc1GFAPCKO mice, there was no significant differences in decay time in recovery of [K+]0 back to baseline following stimulation (50% stimulation: t1/2= 4.9 ± 2.6s, n=14 Tsc1GFAPCKO slices; t1/2= 3.5± 1.6s, n=12 control slices; 100% stimulation:, t1/2= 8.1 ± 4.6s, n=14 Tsc1GFAPCKO slices; t1/2= 7.6 ± 3.7s, n=12 control slices).
Although the deficiency in astrocyte gap coupling could be responsible for the increased [K+]0 following stimulation in the Tsc1GFAPCKO mice, there are other possible mechanisms contributing to potassium buffering that could also account for this difference. Thus, we used a pharmacological gap junction inhibitor, carbenoxolone, to determine the contribution of gap junctions to K+ buffering in control and Tsc1GFAPCKO mice. Peak increases in [K+]0 at 50% and 100% intensity stimuli were compared before and after application of 100μM carbenoxolone for 20 minutes. In control mice, carbenoxolone treatment resulted in a significant, additional increase in peak [K+]0 in response to both 50% and 100% stimulation in control mice that was similar to peak [K+]0 in Tsc1GFAPCKO mice without carbenoxolone (Fig. 4D), indicating that potassium buffering through gap junctions was active in control mice. In contrast, there was no significant difference in [K+]0 rises for 50% and 100% stimulus intensities comparing before and after carbenoxolone treated in Tsc1GFAPCKO mice (Fig. 4D), suggesting that the impaired potassium buffering in Tsc1GFAPCKO mice is due, at least in part, to the deficient astrocyte gap junction coupling.
To provide evidence for a mechanistic link between this impaired gap junction potassium buffering, deficient Cx43 expression, and Tsc1 gene inactivation, we again tested the effect of mTOR inhibition. Similar to the effect on Cx43 expression, rapamycin treatment (3 mg/kg/d for 3 weeks) reversed the impairment of potassium buffering in Tsc1GFAPCKO mice (Fig. 5). With both lower and higher stimulation levels, the peak increase in [K+]0 was significantly decreased in rapamycin-treated Tsc1GFAPCKO mice compared to untreated Tsc1GFAPCKO mice (Fig. 5A). After normalization to population spike amplitude, there were still significant differences between untreated and rapamycin-treated Tsc1GFAPCKO mice in normalized [K+] rises at both 50% and 100% stimulation intensity (Fig. 5B), indicating that the mTOR pathway does mediate the effects of Tsc1 gene inactivation on Cx43 expression and potassium buffering.
Recent clinical and basic science studies are beginning to provide insights into the mechanisms of neurological deficits and epileptogenesis in TSC (Holmes et al., 2007; Wong, 2008). Consistent with modern trends recognizing the importance of astrocytes in brain function and neurological disease, pathological studies of brains from TSC patients document a number of histological abnormalities in astrocytes (Crino, 2004; Sosunov et al., 2008; Talos et al., 2008), suggesting that astrocyte dysfunction may be central to the pathophysiology of neurological deficits and epilepsy in TSC. We have explored the role of astrocytes in a mouse model of TSC involving Tsc1 gene inactivation primarily in astrocytes (GFAP+ cells) (Uhlmann et al., 2002). In the present study, we demonstrate that Tsc1GFAPCKO mice are deficient in the astrocyte-specific gap junction protein, connexin 43, and have an associated impairment in gap junction coupling between astrocytes. Furthermore, we have provided evidence that this deficiency in gap junction coupling may contribute to altered potassium buffering and increased extracellular potassium levels in Tsc1GFAPCKO mice. The abnormal connexin 43 expression and impaired potassium buffering in Tsc1GFAPCKO mice were reversed by rapamycin, indicating that the mTOR pathway mediates these effects. Thus, given the known effects of extracellular potassium on neuronal excitability, we propose that abnormal astrocyte-controlled buffering of potassium may promote epileptogenesis and other neurological deficits in TSC.
The role of astrocytic gap junction coupling in the spatial buffering of potassium and other neuroactive compounds has long been suspected, but has only recently been directly tested. Although pharmacological inhibitors of gap junctions have lacked specificity to differentiate the contribution of different cell types, cell-specific differences in the molecular components of gap junctions, the connexin (Cx) proteins, have allowed the contribution of astrocytes to homeostatic mechanisms to be isolated. Astrocytic gap junctions consist primarily of Cx43, as well as Cx30 and Cx26 (Nagy and Rash, 2000). Genetically-engineered mice deficient in Cx43 and Cx30 completely eliminates astrocyte coupling within hippocampus (Theis et al., 2003; Wallraff et al., 2006). Furthermore, as evidence of the physiological importance of astrocyte coupling, these mice exhibit reduced potassium buffering capabilities and increased suspectibility to epileptiform activity (Wallraff et al., 2006).
The significance and relationship of astrocytic gap junction coupling to epilepsy per se is more complex and controversial. While astrocytic gap junctions may inhibit neuronal excitability and seizures by buffering extracellular potassium, it has also been proposed that the coupling of astrocyte networks might promote seizures, such as by non-synaptic synchronization of neuronal activity via glial calcium waves or other intracellular signals propagated between astrocytes (Steinhauser and Seifert, 2002; Samoilova et al. 2008). Variable changes (primarily increases, but also decreases) in astrocytic connexin proteins have been documented in human epilepsy brain specimens and animal models, and it is difficult to determine whether these changes represent causative or compensatory mechanisms for epileptogenesis (Naus et al., 1991; Elisevich et al., 1997; Sohl et al., 2000; Aronica et al., 2001; Fonseca et al., 2002). Finally, the distinction between astrocytic and neuronal gap junctions is also important, as they may have different physiological roles related to epilepsy. While spatial buffering by astrocytes may reduce neuronal excitability, it is likely that neuronal gap junctions contribute directly to the spread of seizures among neuronal networks and thus that gap junction inhibitors may be effective anti-convulsant drugs (Jahromi et al., 2002; Gajda et al., 2003; Gigout et al., 2006; Medina-Ceja et al., 2008).
In the present study, the functional and pathophysiological significance of impaired astrocyte gap junction coupling in relation to epilepsy and other neurological deficits in Tsc1GFAPCKO mice was not directly addressed. However, several lines of evidence from this and previous studies suggest that the deficient astrocyte coupling is detrimental and promotes epileptogenesis in this mouse model of TSC. First of all, the initial genetic defect in Tsc1GFAPCKO mice primarily affects astrocytes, and progressive astrogliosis slightly precedes the onset of epilepsy in these mice (Uhlmann et al., 2002; Erbayat-Altay et al., 2007), suggesting that astrocyte dysfunction is responsible for epileptogenesis in this model. Furthermore, previous studies have found that hippocampal slices from Tsc1GFAPCKO mice are more sensitive to elevated potassium-induced epileptiform activity (Jansen et al., 2005), which would be consistent with the impaired potassium buffering in hippocampal slices of these mice in the present study. Finally, the decreased expression of Cx43 and impaired potassium buffering could be reversed by rapamycin, a treatment that has also been shown to prevent epilepsy in Tsc1GFAPCKO mice (Zeng et al. 2008), suggesting that the impaired astrocyte gap junction coupling may participate in epileptogenesis due to mTOR hyperactivation and mechanistically linking these astrocyte defects to Tsc1 gene inactivation in these mice.
As mTOR can affect a variety of other downstream effectors through gene regulation, the abnormalities in Cx43 expression could represent just one component of a spectrum of brain abnormalities that promote epileptogenesis and other neurological deficits in Tsc1GFAPCKO mice. This might include more global effects on brain metabolism, homeostasis, or blood brain maintenance (Marchi et al., 2007) or other specific molecular defects in astrocytes or neurons. For example, although the decreased connexin expression and gap junction coupling of astrocytes could account for the excessive potassium concentrations following stimulation in the Tsc1GFAPCKO mice, other mechanisms may also contribute to differences in potassium buffering. Both inward-rectifying potassium channels and sodium-potassium pumps of astrocytes and neurons have been implicated in absorbing extracellular potassium (Haglund and Schwartzkroin, 1990; Xiong and Stringer, 2000; D’Ambrosio et al., 2002), and defects in these molecules may promote seizure generation (Janigro et al., 1997; Bordey and Sontheimer, 1998; Gabriel et al., 1998; Xiong and Stringer, 1999; Hinterkeuser et al., 2000). In fact, decreased expression of specific inward-rectifying potassium channels have been previously documented and proposed to contribute to epileptogenesis in Tsc1GFAPCKO mice (Jansen et al., 2005). However, the effects of carbenoxolone in the present study in causing a further increase in [K+]0 in control mice, but not in Tsc1GFAPCKO mice, suggest that gap junction mechanisms account, at least in part, for the observed differences in potassium buffering between control and Tsc1GFAPCKO mice.
Assuming that defects in astrocyte gap junction coupling and potassium buffering do contribute to epileptogenesis in Tsc1GFAPCKO mice, these findings have potentially significant clinical and therapeutic implications for epilepsy in human TSC. Although to our knowledge, abnormalities in astrocytic connexin expression or gap junction function have not yet been reported in human TSC, there have been several recent examples where initial findings in animal models of TSC have been subsequently confirmed in human TSC (Wong, 2007). As the role of astrocytes in epilepsy in general has been receiving more attention recently (Binder and Steinhauser, 2006; Jabs et al., 2008; Wetherington et al., 2008), future anti-epileptic therapies may be targeted specifically for astrocytes. As exemplified by the potentially opposite effects of neuronal versus astrocytic gap junctions on brain excitability, designing therapies that modulate cell-specific gap junctions will be important. While there is already much excitement and promise for developing novel anti-epileptogenic treatments for epilepsy in TSC (Zeng et al., 2008), further understanding of astrocytic mechanisms of epileptogenesis should lead to safer, more effective treatments.
This work was supported by the National Institutes of Health (NIH) grants K02 NS045583 and R01 NS056872 (to M.W.), the Tuberous Sclerosis Alliance (to M.W.), and the NIH Neuroscience Blueprint Center Core Grant P30 NS057105 (to Washington University). The authors thank Julian Meeks and Steven Mennerick for guidance on potassium-selective electrode methods and David Gutmann for ongoing advice and support related to the mouse model.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.