Astrocyte dysfunction in temporal lobe epilepsy: K+ channels and gap junction coupling



Astrocytes are endowed with the machinery to sense and respond to neuronal activity. Recent work has demonstrated that astrocytes play important physiological roles in the CNS, e.g., they synchronize action potential firing, ensure ion homeostasis, transmitter clearance and glucose metabolism, and regulate the vascular tone. Astrocytes are abundantly coupled through gap junctions, which is a prerequisite to redistribute elevated K+ from sites of excessive neuronal activity to sites of lower extracellular K+ concentration. Recent studies identified dysfunctional astrocytes as crucial players in epilepsy. Investigation of specimens from patients with pharmacoresistant temporal lobe epilepsy and epilepsy models revealed alterations in expression, localization, and function of astroglial inwardly rectifying K+ (Kir) channels, particularly Kir4.1, which is suspected to entail impaired K+ buffering. Gap junctions in astrocytes appear to play a dual role: on the one hand they counteract the generation of hyperactivity by facilitating clearance of elevated extracellular K+ levels while in contrast, they constitute a pathway for energetic substrate delivery to fuel neuronal (hyper)activity. Recent work suggests that astrocyte dysfunction is causative of the generation or spread of seizure activity. Thus, astrocytes should be considered as promising targets for alternative antiepileptic therapies. © 2012 Wiley Periodicals, Inc.


Astrocytes have now emerged as active partners in brain signaling. Electrophysiology and Ca2+ imaging techniques detected that astrocytes in acute brain slices or after fresh isolation from the tissue express a similar broad spectrum of functional ion channels and transmitter receptors as neurons (Verkhratsky and Steinhäuser,2000) which enables them to sense and respond to neuronal activity (Halassa and Haydon,2010). In addition to the recently discovered modulatory actions on brain signaling and circulation (Petzold and Murthy,2011) only, astrocytes are known for decades to serve homeostatic functions, including the clearance of neuronally released K+ and glutamate from the extracellular space. A prerequisite for this function is the abundant coupling of astrocytes through gap junctions, allowing them to redistribute elevated K+ from sites of excessive neuronal activity to sites of lower extracellular K+ concentration (Kofuji and Newman,2004).

Despite the fact that the pathways enabling the activation of glial channels and receptors under physiological conditions are still ill-determined, increasing evidence suggests a critical role of astrocyte dysfunction in the pathogenesis of neurological disorders, including epilepsy (reviewed by Seifert et al.,2006). Investigation of specimens from patients with pharmacoresistant mesial temporal lobe epilepsy (MTLE) and corresponding animal models of epilepsy revealed alterations in expression, subcellular localization, and function of astroglial K+ channels and gap junctions, resulting in impaired K+ buffering and disturbance of delivery of energetic substrates to neurons (Heinemann et al.,2000; Kivi et al.,2000). Together with the malfunction of glutamate transporters and the astrocytic glutamate-converting enzyme, glutamine synthetase (Eid et al.,2004; reviewed by Seifert and Steinhäuser,2011), observed in epileptic tissue, these findings suggested that astrocyte dysfunction is causative of hyperexcitation, neurotoxicity, and the generation or spread of seizure activity. In this article, we will summarize current knowledge of altered functioning of astroglial inwardly rectifying K+ (Kir) channels and gap junctions in TLE and discuss putative mechanisms underlying these alterations.


Neuronal activity leads to rapid fluctuations of the extracellular K+ concentration ([K+]o) because of the restricted volume of the extracellular space. If increases in [K+]o remain uncorrected, the resting potential would become more positive and affect activation of transmembrane ion channels, receptors, and transporters. During neuronal hyperactivity in vivo, [K+]o may increase from 3 mM to a ceiling level of 10–12 mM. Such high [K+]o levels are observed during epileptiform activity. Two different mechanisms are thought to balance [K+]o during neuronal activity: K+ uptake and spatial K+ buffering (for review, see Kofuji and Newman,2004). K+ uptake, mediated by the Na,K-adenosine triphosphatase (Na,K-ATPase) or Na-K-Cl cotransporters, is accompanied by cell swelling and local depolarization of astrocytes. Spatial K+ buffering is driven by the difference between the glial syncytium membrane potential and the local K+ equilibrium potential. K+ enters astrocytes at sites of enhanced [K+]o where the cells' resting potential is negative to the K+ equilibrium potential (due to the electrical coupling with neighbouring cells exposed to normal [K+]o) while efflux of the ions occurs at sites distant to the [K+]o increase. This allows transfer of K+ from regions of elevated [K+]o, through the syncytium, to regions of lower [K+]o (Fig. 1). Spatial buffering depends on proper distribution and function of astrocytic K+ channels and gap junctions.

Figure 1.

Space-dependent spatial K+ buffering. Activity in a group of neurons has produced local increase in [K+]o to 12 mM (shaded area, left). This increase produces a depolarization of the membrane potential (Vm) that passively spreads through the coupled astrocytes. The positive shift of the K+ equilibrium potential (EK) is stronger than that of Vm of the cell exposed to high [K+]o because the latter is “clamped” by the neighbouring, more negative astrocytes exposed to lower (normal) [K+]o (4 mM). The difference between EK and Vm drives K+ inward at the region where it is raised and outward at distant regions. The result is a net flux of K+ away from the region where it has accumulated extracellularly. The average [K+]i is not affected. The graph shows the distribution of EK and Vm as a function of distance along the astroglial syncytium. Modified from Orkand, Ann N Y Acad Sci,1986, 481, 269-272, reproduced with permission.

In astrocytes, the inwardly rectifying K+ channel, Kir4.1, which is regulated by intracellular adenosine triphosphate (ATP), is thought to allow for K+ influx at negative membrane potentials (for review see Butt and Kalsi,2006; Hibino et al.,2010; Olsen and Sontheimer,2008). Genetic inactivation of the Kir4.1 encoding gene, KCNJ10, in animal models substantiated an outstanding role for Kir4.1 in glial K+ buffering (Djukic et al.,2007; Kofuji etal.,2000). Mice with global and conditional knockout of Kir4.1 showed a severe phenotype, including a smaller size, reduced body weight, and deficits in controlling posture and movement. They developed seizures and died about 3 weeks after birth (Djukic et al.,2007; Kofuji et al.,2000). In the CNS, Kir4.1 channels are specifically localized to macroglial cells. Accordingly, conditional knockout under the control of the human GFAP promoter entailed loss of Kir4.1 in NG2 glial cells, oligodendrocytes, and astrocytes. Ablation of Kir4.1 led to spongiform vacuolization, axon swelling, and hypomyelination in the spinal cord, with strongly depolarized oligodendrocytes (Neusch et al.,2001). It has been suggested that Kir4.1 in concert with Cx32/Cx43 activity-dependently contribute to K+ redistribution in the white matter of spinal cord and optic nerve (Menichella et al.,2006). Gap junction coupling between astrocytes and oligodendrocytes has been considered a prerequisite not only for proper K+ buffering, but also for myelin formation and axonal conductivity (Maglione et al.,2010; Magnotti et al.,2011b). In the retina, Kir4.1 knockout leads to depolarization of Müller cell processes contacting blood vessels, and disappearance of the PIII response of the electroretinogram, which is generated by K+ flux (Kofuji et al.,2000). Gray matter astrocytes are also depolarized after Kir4.1 deletion, which not only impaired K+ buffering but glutamate uptake as well (Djukic et al.,2007; Kucheryavykh et al.,2007). These changes led to decreased neuronal activity but enhanced synaptic potentiation in the hippocampus. The key role of Kir4.1 in K+ uptake and spatial buffering has also been confirmed by showing that its deletion causes an epileptic phenotype (Chever et al.,2010; Haj-Yasein et al.,2011).

Ultrastructural analyses in rat demonstrated spatial overlap of Kir4.1 and the water channel aquaporin 4 (AQP4) in astroglial endfeet contacting capillaries (Higashi et al.,2001; Nielsen etal.,1997). This finding gave rise to the hypothesis that K+ clearance through Kir channels might depend on concomitant transmembrane flux of water. Subsequent functional work corroborated this idea by showing that in mice the clearance of extracellular K+ is compromised if the number of perivascular AQP4 channels is decreased (Amiry-Moghaddam et al.,2003). In line with this assumption was also the observation of impaired K+ buffering and prolonged seizure duration in AQP4 knockout mice (Binder et al.,2006) although spatial K+ redistribution was more efficient in these mice, probably due to enhanced astrocyte gap junction coupling and volume regulation (Benfenati et al.,2011; Strohschein et al.,2011). However, more recent data from astrocytes and retinal Müller cells suggested that Kir4.1 channel function is independent of AQP4 expression (Ruiz-Ederra et al.,2007; Zhang and Verkman,2008).


Because of their presumed role in K+ homeostasis, the properties of astroglial Kir channels have been investigated in experimental and human epilepsy. Measurements of [K+]o with ion-sensitive microelectrodes and patch-clamp studies suggested that impaired K+ buffering in the sclerotic human hippocampus resulted from altered Kir channel expression. Differences were observed in the effect of Ba2+ on stimulus-induced changes in [K+]o in the CA1 region of hippocampal brain slices obtained from epilepsy patients with hippocampal sclerosis (HS) or without sclerosis (lesion-associated epilepsy). In tissue from patients with lesion-associated epilepsy, Ba2+ application significantly enhanced [K+]o, while this effect was not observed in HS specimens. Since Ba2+ at low concentrations (≤ 100 μM) is a blocker of Kir channels in astrocytes of the hippocampus (Seifert et al.,2009), this finding suggested impaired function of these channels in the sclerotic tissue (Jauch et al.,2002; Kivi et al.,2000). The hypothesis could be confirmed with patch-clamp analyses demonstrating downregulation of Kir currents in the CA1 region of HS patients (Bordey and Sontheimer,1998; Hinterkeuser et al.,2000). Transcript analysis confirmed expression of Kir4.1 in human glial cells (Schröder et al.,2000). Accordingly, in HS, impaired K+ buffering and enhanced seizure susceptibility result from reduced expression of Kir channels. However, it is still unclear whether these changes are the cause or the consequence of the condition.


Fine mapping of a locus on mouse chromosome 1 identified KCNJ10, the gene encoding Kir4.1, as a candidate gene exhibiting a polymorphism potentially important with regard to seizure susceptibility (Ferraro et al.,2004). Similarly, variations in KCNJ10 in the human genome associate with multiple seizure phenotypes. Missense mutations of KCNJ10 influence the risk of acquiring forms of human epilepsy (Buono et al.,2004; Lenzen et al.,2005). Mutations in the KCNJ10 gene cause a multiorgan disorder in patients with clinical features of epilepsy, sensorineural deafness, ataxia, and electrolyte imbalance (EAST/SeSAME syndrome; Bockenhauer et al.,2009; Scholl et al.,2009). These patients suffer from generalized tonic-clonic seizures and focal seizures since childhood. Single nucleotide mutations in the KCNJ10 gene cause missense or nonsense mutations on the protein level in the pore region, transmembrane helices, or the C terminus, the latter resulting in deletion of a PDZ binding domain and improper membrane localization of Kir4.1. These loss-of-function mutations of KCNJ10 are homozygous or compound heterozygous. Heterologous expression demonstrated that the mutations indeed affected Kir4.1 function and produced depolarization and reduced transmembrane currents (Bockenhauer et al.,2009; Reichold et al.,2010; Williams et al.,2010). The EAST/SeSAME syndrome-related decrease in K+ conductance could not be rescued by co-expression with Kir5.1, and mutated Kir4.1 had no dominant-negative effect when co-expressing wildtype Kir4.1 (Tang et al.,2010). The mutations increased pHi-sensitivity of the channel and impeded its surface expression (Sala-Rabanal et al.,2010; Williams etal.,2010). Other patients carried gain-of-function mutations of KCNJ10. The heterozygous mutations were localized to the N-terminus and the first transmembrane region and the afflicted patients presented with seizures, autism spectrum disorders, and intellectual disability (Sicca et al.,2011).

Single nucleotide variations in the Aqp4 and KCNJ10 genes were identified in DNA from TLE patients presenting with HS or TLE with antecedent febrile seizure (TLE-FS). Subsequent multivariate analysis identified a correlation of three single nucleotide polymorphisms (SNPs) in the Aqp4 gene together with two SNPs in the KCNJ10 gene (Heuser et al.,2010). Further association analysis identified eight SNPs in the KCNJ10 gene and one between KCNJ10 and KCNJ9. Multivariate analysis showed that a combination of SNPs from KCNJ10, Aqp4, and the area between KCNJ10 and KCNJ9 was significantly associated with TLE-FS (Heuser et al.,2010).


Traumatic brain injury harbours a risk factor for developing seizures. In a post-traumatic epilepsy animal model (fluid percussion injury, FPI), a decrease of Cs+-sensitive Kir currents and elevated [K+]o upon antidromic stimulation of Schaffer collaterals were observed, indicating impaired glial K+ homeostasis (D'Ambrosio et al.,1999). Moreover, chronic spontaneous partial seizures were generated in the neocortex and hippocampus, with the rate of hippocampal seizures greatly increasing over time post-injury (D'Ambrosio etal.,2005). Immunohistochemical and electrophysiological analyses showed that Kir4.1 channels, particularly in astrocyte processes, were lost after FPI. This injury is accompanied by serum extravasation which might account for the loss of Kir currents and seizure generation (Stewart et al.,2010). Dysfunction of the blood-brain barrier (BBB) seems also to be involved in seizure generation. In rat, transient opening of the BBB is sufficient for induction of focal epileptogenesis (Seiffert et al.,2004). BBB lesion is also considered a primary event in human TLE, leading to the extravasation of serum albumin which is taken up by neurons, astrocytes, and microglia (Seiffert et al.,2004; Van Vliet et al.,2007). In contrast, no changes in astrocyte Kir currents were observed in a systemic kainate rat model, 7–16 days following status epilepticus (SE) (Takahashi et al.,2010).

Fine mapping of putative seizure susceptibility loci on mouse chromosome 1 identified KCNJ10 as a critical candidate (Ferraro et al.,2004,2007). Comparative electrophysiological analysis of seizure-resistant (B6) and seizure-susceptible (D2) mice, which differ at position 262 of KCNJ10, revealed a profound reduction of Ba2+-sensitive Kir currents in the latter (Inyushin et al.,2010). Astrocytes from D2 mice also displayed impaired glutamate clearance. Interestingly, spinal cord injury-induced downregulation of Kir4.1 could be partially rescued by administration of β-estradiol. Inhibition of estrogen receptors also reduced Kir4.1 channel mediated currents, suggesting that the regulation of these channels depends on nuclear estrogen receptor signaling (Olsen et al.,2010).

In addition to spatial buffering, net K+ uptake through Na,K-ATPase and the Na-K-Cl cotransporter NKCC1 may counterbalance transient K+ accumulation, at the cost of cell swelling due to concomitant water influx (Kofuji and Newman,2004). In rodent hippocampus, the Na,K-ATPase might have a role in maintaining low [K+]o levels and clearing elevations in [K+]o after epileptiform activity (D'Ambrosio et al.,2002; Xiong and Stringer,2000). Whether net K+ uptake mechanisms are altered in epileptic tissue and contribute to the increased [K+]o levels has still to be determined.


The main gap junction protein isoforms expressed by astrocytes are Cx30 and Cx43 (Nagy and Rash,2000). The two isoforms display distinct unitary conductances, selectivity, sensitivity to transjunctional voltage, and gating properties. For instance, Cx30 channels possess a higher unitary conductance than Cx43 channels (179 vs. 100 pS) and are more sensitive to transjunctional voltage (Valiunas et al.,1999,2000). Tracer spread studies performed on Cx30-deficient mice indicated that coupling of astrocytes in the hippocampus predominately depends on Cx43 and only to a minor extent (22%) on Cx30 channels (Gosejacob et al.,2011). Astrocytic gap junction coupling has been supposed to have several functions, including spatial buffering of K+ ions (Ransom,1996), trafficking and delivery of energetic metabolites to neurons (Giaume et al.,1997), intercellular propagation of Ca2+ waves (Scemes and Giaume,2006), volume regulation (Scemes and Spray,1998), and adult neurogenesis (Kunze et al.,2009).

Role of Gap Junction Coupling in K+ and Neurotransmitter Homeostasis and Nutrient Supply

The role of interastrocytic gap junction communication in the pathophysiology of epilepsy is still elusive. A potential anti-epileptic function of the astroglial network might be explained by the K+ spatial buffering concept (see above). According to this concept, impaired gap junction coupling could result in insufficient extracellular K+ clearance and, consequently, to epileptic seizures. To test this hypothesis, Wallraff et al. (2006) assessed changes in K+ buffering after antidromic stimulation in hippocampal slices obtained from mice with coupling-deficient astrocytes (Cx30−/− Cx43flox/flox hGFAP-Cre mice; dko mice) using K+-selective microelectrodes. Unexpectedly, in these mice they found preserved K+ buffering in the stratum radiatum (but not in the stratum lacunosum-moleculare) (Fig. 2), suggesting that gap junction independent mechanism may be involved (“indirect coupling,” Wallraff et al.,2006). However, the dko mice showed spontaneous epileptiform events and a reduced threshold for the generation of epileptic activity. In a recent paper, Pannasch et al. (2011), using the same mice, also reported disturbed glutamate clearance during synaptic activity. They attributed the extracellular accumulation of glutamate and K+ not only to an insufficient uptake/buffering by disconnected astrocytes but also to the reduction of the extracellular space volume due to astrocytic swelling, aggravating the accumulation of neurotransmitters and K+ in the extracellular space. In accordance with the findings by Wallraff et al. (2006), they detected enhanced excitability of CA1 pyramidal neurons in dko mice (Pannasch et al.,2011).

Figure 2.

Impaired spatial buffering in the stratum lacunosum-moleculare of mice with genetic deletion of Cx43 and Cx30 in astrocytes (dko mice). A: Experimental setup for the analysis of laminar profiles of changes in [K+]o. Stimulation was performed in the alveus (20 Hz, 100%), while the recording electrode was stepped from the stratum pyramidale to the hippocampal fissure (100 μm step size). B: Mean rises in [K+]o (normalized to rise at the stratum pyramidale) plotted against the distance from stratum pyramidale [thin line, wild type (wt): n = 8 animals, 13 slices; thick line, dko, n= 8 animals, 15 slices]. Normalized [K+]o in dko mice reached lower levels at 400, 500, and 600 μm from the stratum pyramidale when compared with wt. See inset for relative changes. C: This demonstrates the difference in astrocyte morphology and orientation in the stratum radiatum vs. stratum lacunosum-moleculare, as obtained after biocytin injection into a stratum radiatum astrocyte proximal to the stratum lacunosum-moleculare. Note the small size and random orientation of cells in the stratum lacunosum-moleculare. Scale bar: 50 μm. alv., alveus; fis., fissure; stim., stimulation electrode; rec., recording electrode; s.r., stratum radiatum; s.l.m., stratum lacunosum-moleculare. From Wallraff et al., J Neurosci,2006, 26, 5438-5447, reproduced with permission.

A potential pro-epileptic function of astroglial networks was proposed by Rouach et al., (2008). These authors demonstrated that gap junctions mediate trafficking of energetic metabolites from blood vessels to the site of high energy demand in an activity-dependent manner and suggested that this process might be important to sustain glutamatergic synaptic activity under pathological conditions (Fig. 3). Moreover, they showed that spontaneous epileptiform activity increased glucose diffusion through the astrocytic syncytium, a process important to sustain epileptiform activity. A second possible seizure-promoting function of the astroglial syncytium may arise from its contribution to the propagation of intercellular Ca2+ waves (Scemes and Giaume,2006) which have been shown to mediate non-synaptic neuronal synchronization via glutamate release (Fellin et al.,2004; Tian et al.,2005). Increased astrocytic coupling may facilitate the propagation of calcium waves, leading to hypersynchronization and spread of ictal activity (Gomez-Gonzalo et al.,2010).

Figure 3.

Glucose supply to astrocytic networks partially sustains neuronal epileptiform activity. A: Sample traces showing that epileptiform activity, induced by 0 Mg2+-picrotoxine (100 μm, 1–4 h) and recorded by extracellular fEPSPs, is reversibly abolished by exogenous glucose deprivation during 30 min in control conditions (n = 6) while (B) it is partially maintained when glucose (20 mM) is delivered to astrocytic networks through a patch pipette (+ glucose astrocytes, n = 6). Scale bar, 0.25 mV, 10 s. C: Summary bar graph (n = 6). From Rouach et al., Science,2008, 322, 1551-1555, reproduced with permission.

Taken together, these studies lead to conflicting conclusions as to the role of gap junctions in epileptogenesis. Reduction of astrocytic coupling, for instance, would on the one hand provoke neuronal hyperexcitability, due to impaired K+ and glutamate clearance but, in contrast, hyperactivity could not be fuelled, due to insufficient supply of energetic metabolites. In addition, decreased coupling would hamper the spread of Ca2+ waves and therefore attenuate synchronization of neuronal activity. Enhanced astrocytic coupling, on the contrary, would not only improve clearance of K+ and transmitters and dampen excitability, but also facilitate intercellular propagation of Ca2+ waves and delivery of energetic metabolites, supporting neuronal hyperactivity.

Gap Junction Communication Between Astrocytes and Oligodendrocytes

Gap junction coupling occurs not only between astrocytes but also between astrocytes and oligodendrocytes. Oligodendrocytes express the gap junction proteins Cx32 and Cx47 which may form heterotypic channels with astrocytic gap junctions (Magnotti et al.,2011a). There are, however, conflicting reports regarding the relevance and functionality of the different heterotypic channel combinations which can form between the two cell types. In HeLa cells, the pairings Cx30/Cx32, Cx30/Cx47, and Cx43/Cx47 gave rise to functional channels (Magnotti et al.,2011a) whereas in N2A cells only the pairings Cx30/Cx32 and Cx43/Cx47 but not Cx30/Cx47 were functional (Orthmann-Murphy et al.,2007). Freeze-fracture immuno-labeling microscopy indicated that gap junctions between oligodendrocytes and astrocytes contain mainly Cx47 while autologous (reflexive) oligodendrocytic gap junctions were mainly composed of Cx32 (Kamasawa et al.,2005). Coupling analyses performed on acute slices from Cx43- and Cx47-deficient mice suggested that the communication between astrocytes and oligodendrocytes relies mainly on Cx30/Cx47 channels (Maglione et al.,2010). In contrast, immunolocalization studies in Cx32 and Cx47 KO mice rather indicated that heterotypic Cx32/Cx30 and Cx43/Cx47 are involved in heterocellular coupling (Altevogt and Paul,2004; Li et al.,2004; Nagy et al.,2003). Hence, additional work is needed to clarify the contributions of the different connexin isoforms to the astrocyte–oligodendrocyte communication.

Double knockout mice lacking the oligodendrocytic Cx32 and the astrocytic Cx43 (Cx32/Cx43 dKO mice) showed selective loss of astrocytes, seizure activity, and early mortality at around 16 weeks of age (Magnotti et al.,2011b). Since the corresponding single knockouts do not exhibit astrocytic loss or seizure activity (Sargiannidou et al.,2009; Theis et al.,2003), it is reasonable to assume that this phenotype is attributable to the reduction/disruption (depending on whether Cx30 and Cx47 can form functional channels) of astrocyte–oligodendrocyte communication. However, the molecular mechanisms underlying the pathological consequences remain unclear. It has been suggested that oligodendrocyte connexins contribute to K+ buffering (Kamasawa et al.,2005; Menichella et al.,2006). In this model, K+ is taken up by oligodendrocytes and transferred through reflexive junctions to the cell body and from there through heterocellular junctions into the astrocytic network. While impairment of this mechanism may offer an explanation for the observed seizure activity, it is less plausible why astrocytic cell death occurs in Cx32/Cx43 dKO mice. Magnotti et al., (2011b) speculated that Cx32 and Cx43-mediated signaling promotes astrocyte survival. However, it is not obvious why Cx30/Cx43 dKO mice, in which not only the inter-astrocytic but also the astrocyte–oligodendrocyte communication is interrupted (Maglione et al.,2010), do not show the severe phenotype of Cx32/Cx43 dKO mice. Further work is needed to elucidate the potential role of gap junction communication between astrocytes and oligodendrocytes in seizure generation.


Several studies have investigated changes in connexin expression in animal models of epilepsy. The results of these studies are inconsistent, with some providing support for increased mRNA and/or protein levels of Cx43 (Gajda etal.,2003; Samoilova et al.,2003; Szente et al.,2002; Takahashi et al.,2010) and Cx30 (Condorelli et al.,2002), while others reporting no changes of Cx43 (Khurgel and Ivy,1996; Li et al.,2001; Söhl et al.,2000) and Cx30 (Söhl et al.,2000; Xu et al.,2009), or decreased Cx43 levels (Elisevich et al.,1997a,1998; Xu et al.,2009) (see also reviews by Giaume et al.,2010; Steinhäuser and Seifert,2002). This inconsistency may result from the different animal models used in these studies. Furthermore, differences in seizure duration and investigated brain regions could have caused conflicting findings. In addition, species differences as well as differences in the time course of the disease between animal models and the human situation limit the conclusions which can be drawn from such studies (Avoli et al.,2005).

In human epileptic tissue, unchanged (Elisevich et al.,1997b) or elevated (Aronica et al.,2001; Collignon et al.,2006; Fonseca etal.,2002; Naus et al.,1991) Cx43 mRNA and/or protein levels have been reported. For instance, Collignon et al. (2006) used immunohistochemical analysis to assess Cx43 protein levels in hippocampal specimens from patients after amygdalohippocampectomy and found a significant increase of Cx43 immunoreactivity when compared with non-epileptic postmortem controls. In contrast, Northern and Western blot analyses performed by Elisevich et al. (1997b) on hippocampal tissue from patients presenting with medically intractable epilepsy revealed no changes in Cx43 mRNA or protein levels.

Comparison and interpretation of the human studies should, however, be handled with care because: (1) epilepsy is not a homogeneous condition but comprises a large number of syndromes which vary greatly with respect to their clinical features, treatment, and prognosis; (2) tissue specimens are only available from the chronic state of the disorder; and (3) tissue is usually obtained from patients with pharmacoresistant epilepsy, raising the possibility that treatment with various antiepileptic drugs differently affected the tissue samples. In addition, in most cases, tumor or autopsy specimens were used as controls and it cannot be ruled out that the apparent increases in connexin levels were actually caused by decreased connexin expression in such “control” tissue (Nemani and Binder,2005). In accordance with this concern, downregulation of Cx43 expression in tumor tissue has been shown in several studies (Huang et al.,1999; Laird etal.,1999; Soroceanu et al.,2001; but see Aronica et al.,2001). It also has to be emphasized that alterations in mRNA or protein levels do not necessarily correlate with changes in functional coupling. Regulatory modifications, for example phosphorylation, may profoundly influence gating properties, assembly, or subcellular localization of the junctional channels, limiting predictions from expression studies. Functional assays are therefore inevitable to reliably investigate the role of astrocyte gap junctions in epilepsy.


Seizure-induced changes in astrocytic gap junction coupling have been investigated by Xu et al. (2009) in a genetic mouse model of tuberous sclerosis complex, a human disease which is commonly associated with medically intractable seizures. Using biocytin diffusion they detected a significant reduction of inter-astrocytic coupling in the hippocampal CA1 region. Opposite results have been reported by Takahashi et al. (2010) who used the same method in a post-SE rat model of epilepsy (systemic kainate injection) and observed increased coupling of hippocampal astrocytes during the latent period (7–16 days post-SE). Increased coupling has also been found with fluorescence recovery after photobleaching (FRAP) in hippocampal slice cultures chronically exposed to bicuculline (Samoilova et al.,2003). So far, there is only one culture study investigating coupling in astrocytes from human epilepsy. In this study, Lee et al. (1995) used FRAP to demonstrate stronger coupling between astrocytes cultured from epileptic specimens. Of course, culture condition may have influenced the functional phenotype of these human astrocytes.

Many studies have explored the effect of pharmacological disruption of gap junction communication on seizure activity in a variety of in vitro and in vivo models of epilepsy. Most of these studies reported that gap junction inhibitors, such as carbenoxolone, halothane, octanol, heptanol, oleamide, and taurine as well as substances producing intracellular acidification (sodium propionate, carbon dioxide) possess anticonvulsive effects (Bostanci and Bagirici,2006,2007; Gajda et al.,2003; Gigout et al.,2006; Jahromi et al.,2002; Köhling et al.,2001; Medina-Ceja et al.,2008; Perez-Velazquez et al.,1994; Ross etal.,2000; Samoilova et al.,2003; Szente et al.,2002). Consistently, substances promoting gap junction communication by intracellular alkalinization (trimethylamine, ammonium chloride) have been found to increase epileptiform activity (Köhling et al.,2001; Perez-Velazquez et al.,1994). It should be noted, however, that others have reported proconvulsive effects of gap junction blockade through carbenoxolone (Voss et al.,2009). In this context one has to consider that most gap junction blockers are not specific, i.e., they have many side effects (Rozental et al.,2001) and barely discriminate between connexin isoforms and thus cell types (Giaume et al.,2010; Steinhäuser and Seifert,2002). Hence, it remains mostly unclear whether the effects seen after application of the agents were caused by inhibition of electrical synapses between neurons or by blockage of the glial syncytium. Specific inhibition of individual connexin isoforms can hitherto only be achieved with mimetic peptides. Samoilova et al. (2008) treated rat organotypic hippocampal slice cultures with Cx43 mimetic peptides and found reduced spontaneous (but not evoked) epileptiform activity, suggesting that enhanced interastrocytic coupling promotes seizures. This result is, however, not in line with the reported hippocampal hyperexcitability found in acute slices obtained from mice with coupling-deficient astrocytes (Wallraff et al.,2006). The situation is further complicated by the fact that both pharmacological blockade and genetic ablation may not only inhibit intercellular gap junctions but also connexin hemichannels which might differently affect excitability.


The presence of functional connexin hemichannels in astrocytes has been demonstrated in several studies (for review, see Giaume et al.,2010). Under basal conditions hemichannels possess a very low open probability, which however increases upon depolarization, enhanced intracellular and decreased extracellular Ca2+ concentrations, metabolic inhibition, increased levels of proinflammatory molecules and, in case of Cx43, phosphorylation (Giaume et al.,2010; Saez et al.,2005). Many of these changes occur in epileptic tissue, raising the possibility of enhanced hemichannel opening in this condition. Astrocyte hemichannels have been reported to be permeable for ATP, glutamate, and glucose (Cotrina et al.,2000; Retamal et al.,2007; Ye etal.,2003). Increased release of substances through astrocytic connexin hemichannels might promote seizures generation: (1) Elevated extracellular ATP levels have been found in epileptic tissue and implicated in the pathophysiology of epilepsy, although the underlying mechanism is still unclear (for review, see Kumaria et al.,2008). It is conceivable that enhanced release of ATP through hemichannels facilitates the propagation of intercellular glial Ca2+ waves, thereby supporting hypersynchronization and spread of neuronal discharges during seizure activity. (2) Glutamate release from astrocytes may trigger transient depolarizations and focal ictal discharges (Gomez-Gonzalo et al.,2010; Kang etal.,2005). Under pathological condition, hemichannels might contribute to excessive glutamate release. (3) Open connexin hemichannels may provide an alternative mechanism for astrocytic glucose uptake under pathological conditions. Transient increases in extracellular glucose levels have been demonstrated after pilocarpine-induced SE in rat (Slais et al.,2008). Enhanced uptake of glucose by astrocytes through hemichannels (Retamal et al.,2007) and the resulting increased formation of lactate may be required to cover the high neuronal energy demand during epileptiform activity (Rouach et al.,2008).

Accordingly, hemichannels might have differential roles in epilepsy. Yoon et al. (2010) used a bicuculline slice culture model of epilepsy to show that low concentrations of a connexin mimetic peptide, which mainly blocked hemichannels rather than gap junctions, had protective effects on seizure spread while inhibition of both hemichannels and gap junctions by high doses of the peptide exacerbated the lesions. Distinct functions of gap junctions vs. hemichannels might be particularly relevant in the context of epilepsy because it has been shown that they are oppositely regulated during inflammation (Retamal et al.,2007). The cytokines interleukin-1β and tumor necrosis factor-α as well as conditioned medium from activated microglia were found to reduce gap junction coupling but to increase the open probability of hemichannels in cultured astrocytes (Meme et al.,2006; Retamal et al.,2007). Proinflammatory cytokines are overproduced in human and rodent epileptic tissue (Avignone et al.,2008; Vezzani et al.,2008 ). Hence, it is conceivable that cytokine-mediated decrease of coupling causes hyperexcitability due to impaired clearance of K+ and glutamate, while at the same time the increased hemichannel activity facilitates hypersynchronization and seizure spread through release of ATP and glutamate, and glucose uptake.

It has to be noted that the functional relevance of connexin hemichannels is still disputed. Recent evidence suggests that pannexins, rather than connexins, form functional hemichannels under physiological conditions (reviewed by Scemes et al.,2009). These proteins, which show substantial homology with invertebrate gap junction proteins (innexins), are expressed by neurons and glial cells and are probably unable to form functional gap junctions. Pannexin hemichannels are activated by elevated [K+]o, depolarization, mechanical stress, and P2 receptor activation (Scemes et al.,2009). In neurons, NMDA receptor stimulation triggers pannexin hemichannel opening, which in turn enhances the postsynaptic responses and promotes epileptiform activity (Thompson et al.,2008). Several studies have shown that pannexins mediate ATP release from astrocytes and neurons (Bao etal.,2004; Iglesias et al.,2009; Scemes et al.,2007), leading to the notion that these channels may contribute to seizures by increasing the extracellular concentration of ATP. To test this hypothesis, Santiago et al. (2011) investigated consequences of pannexin blockade or deletion on seizure activity in a SE mouse model. The results from this study indicate that pannexins may indeed contribute to SE in vivo, presumably due to [K+]o-dependent opening. However, a differentiation between the relative contributions of neuronal and glial pannexins on this effect could not be elucidated. The findings by Santiago et al. (2011) are in line with an earlier report demonstrating enhanced pannexin mRNA levels in experimental epilepsy (Mylvaganam etal.,2010).


Astrocytes are increasingly accepted as communication partners of neurons. This new insight also sheds new light on a potential role of these cells in epilepsy. Recent work from several laboratories reported dysfunction and dysregulation of astroglial Kir channels and gap junction proteins in human and experimental epilepsy. This particularly concerns impaired removal and redistribution of K+ as observed in MTLE-HS. However, several important questions still need to be addressed. Importantly, it is still unclear whether the reported astroglial alterations are causative or a consequence of the condition. Moreover, evidence is emerging demonstrating functional heterogeneity among astrocytes (Matyash and Kettenmann,2010; Zhang and Barres,2010) which complicates interpretation and comparison of individual studies. The combination of molecular genetics with high resolution imaging and electrophysiology will help clarifying whether MTLE is a glial rather than a neuronal disorder and to develop more specific therapeutic approaches to treat this condition.


The authors thank Dr. I. Nauroth for comments on the manuscript, and apologize to all those, whose work could not be discussed due to space constraints.