When one thinks of the “connections” between epilepsy and gap junctions (GJs), the usual interpretation is that the GJs form direct intercellular cytoplasmic connections between neurons, promoting the hypersynchronous neuronal activity associated with seizures. However, as this review discusses, that concept is rather naïve and greatly complicated by much new data concerning the potential roles of gap junctional communication (GJC) and membrane Cx hemichannels and pannexin channels. (Giaume et al., 2013). Although there are many other relevant reviews of GJs and epilepsy (Dudek et al., 1998; Carlen et al., 2000; Perez Velazquez and Carlen, 2000; Traub et al., 2004; Nemani and Binder, 2005; Salameh and Dhein, 2005; Jin and Chen, 2011; Carlen, 2012; Steinhauser et al., 2012), the roles of pannexins and Cx hemichannels are mainly ignored.

It is commonly assumed that the key location for GJC and seizure generation is between neurons, usually between dendrites (Vazquez et al., 2009). For example our group demonstrated GJC between stratum oriens interneurons in the more electrotonically remote distal dendrites based on the relatively low coupling coefficients and the available anatomical evidence (Zhang et al., 2004). However, other locations for GJC critical for seizure generation have now been proposed. Using a double knockout (dKO) of the glial connexins 30 and 43 (Wallraff et al., 2006; Rouach et al., 2008) demonstrated that GJC between astrocytes for the delivery of glucose or lactate to astrocytes was necessary to sustain excitatory synaptic transmission and epileptiform activity. GJC also can spread apoptotic signals (Lin et al., 1998; Andrade-Rozental et al., 2000), a process which could be quite relevant in severe seizure activity (Belousov and Fontes, 2013).

An in vivo and ex-vivo study of adult rats treated with 4-aminopyridine (4-AP), a K⁺ channel blocker which induces seizures, have shown that dephosphorylation of connexin 43 associated with astrocytic swelling, resulted in reduction of astrocytic gap junction permeability (Zador et al., 2008). These results suggest that, during acute seizures, a prolonged inhibition of intercellular coupling develops in the astrocytic network. Long-lasting (weeks) astrocyte swelling was observed following seizures in the kindling seizure model (Khurgel and Ivy, 1996). It is well-known that glial swelling decreases the cerebral extracellular space (Dietzel and Heinemann, 1986; Sykova, 2004; Badaut et al., 2011). This enhances ephaptic transmission which promotes seizure activity (Jefferys, 1995; Dudek et al., 1998; Shahar et al., 2009). Since astrocytes play an important functional role in extracellular K⁺ and pH homeostasis, pathological brain states that result in K⁺ and pH dysregulation may also cause astrocyte swelling (Florence et al., 2012). The presence of gliotic scars in chronic focal epilepsy patients has led to the suggestion that glia can play an important pathophysiological role in chronic epilepsy (De Lanerolle et al., 2010). The astrocytes in sclerotic hippocampi differ from those in non-sclerotic hippocampi in their membrane physiology and related microvasculature (De Lanerolle et al., 2010). In addition, within these sclerotic hippocampal tissues, there is increased expression of many molecules normally associated with immune and inflammatory functions.

Traub and colleagues have introduced the concept of inter-axonal GJC playing a significant role in seizure-genesis and in the generation of sharp wave ripple complexes (Traub et al., 2002, 2005; Simon et al., 2013). Another underexplored feature of GJC in the CNS is the role of combined electrochemical synapses (Pereda, 2014) well-established in the invertebrate CNS, and recently demonstrated in mossy fiber terminals of the hippocampus by several groups (Hamzei-Sichani et al., 2007; Nagy, 2012; Vivar et al., 2012). What role these synapses play in epilepsy is presently unknown.

The sharp-wave ripple complex (SPW-ripple) (Figures 1B,C) is highly synchronous physiological activity that is generated in the hippocampus. Emerging evidence suggests that under pathological conditions the neuronal assemblies generating SPW-ripples may also be responsible for generating epileptiform activity (Staba et al., 2004; Khosravani et al., 2005; Behrens et al., 2007; Bragin et al., 2007; Beenhakker and Huguenard, 2009; Liotta et al., 2011; Simeone et al., 2013).

The mechanisms responsible for ripple generation are not fully understood, however a prominent theory suggests that gap junction proteins play a critical role (Figures 1B,C). Traub and colleagues modeled hippocampal ripple activity through axo-axonal electronic coupling of pyramidal neurons (Traub et al., 2002). According to this model, spontaneously generated action potentials in CA1 axons depolarize electronically coupled neurons, resulting in a wave of spike generation that travels along the axonal plexus, and is detected in the LFP as a high frequency oscillation (Traub et al., 1999; Simon et al., 2014). The first experimental evidence implicating a role for GJs in high-frequency ripple oscillations in the hippocampus, came from utilizing multisite single unit recordings in the awake rat (Ylinen et al., 1995). They inhibited ripple oscillations in the CA1 by application of the anesthetic drug halothane, which, along with other effects, also blocks GJs. Subsequently, ripple oscillations were observed in extracellular field recordings in rat hippocampal brain slice preparations (Draguhn et al., 1998; Traub et al., 2002). The ripple oscillations were enhanced following bath application of a calcium-free solution, which enhances GJC and membrane excitability (Perez-Velazquez et al., 1994). Similar to ripples recorded in vivo, their occurrence was completely inhibited by pharmacological agents known to block GJs, such as octanol, carbenoxolone, halothane, and low pH.

However, as noted below, gap junction modulators are non-specific, and may mediate their effects through gap junctional-independent mechanisms (Nakahiro et al., 1991; Rouach et al., 2003). Recently, (Behrens et al., 2011) utilized a more specific gap junction blocker, mefloquine, which blocks Cx36, Cx43, and Cx50, and observed no effect on SPW or associated ripples. In Cx36 knockout mice, in vivo ripple oscillations were unaltered in terms of power, frequency, or probability of occurrence (Buhl et al., 2003). In contrast, in vitro studies on Cx36 knockout mice, demonstrated a reduced frequency of SPW and associated ripples (Pais et al., 2003). However, other in vitro studies in Cx36 knockout mice have reported no significant differences in ripple oscillations when chemical synaptic transmission was absent (Hormuzdi et al., 2001) or during application of the glutamate receptor agonist, kainate (Pais et al., 2003). It has to be recognized also that upregulation of other Cxs could significantly alter/blunt the effects of eliminating Cx36. Future studies should investigate the possibility that alternative connexins, connexin hemichannels, or pannexins are involved in the generation of SPW and ripple rythmogenesis.

GJ channels provide the basis for direct cell-to-cell communication, whereas Cx hemichannels and Panx channels allow the exchange of ions and signaling molecules such as ATP, NAD+, glutamate and other molecules less than 1000 Daltons between the cytoplasm and the extracellular milieu (Willecke et al., 2002). Recent papers have discussed the different properties of Cx hemichannels and Panx channels from the perspective of ATP release (Lohman and Isakson, 2014) and electrophysiological characteristics (Patel et al., 2014). These hemichannels and channels support autocrine and paracrine communication through a process called “gliotransmission,” which involves the uptake and release of metabolites such as glucose and glutathione and also the release of autocrine/paracrine molecules such as ATP, glutamate, NADP+ and adenosine into the extracellular medium (Sohl and Willecke, 2004; Bruzzone et al., 2005; Cherian et al., 2005; Retamal et al., 2007; Lin et al., 2008; Stridh et al., 2008). Studies have shown that the opening of Cx43 hemichannels is promoted by positive transmembrane voltages and reduced concentrations of extracellular divalent cations such as Ca²⁺. High intracellular H⁺, and phosphorylation of Cx43 promotes the closure of the hemichannels (Chang et al., 1996; Bennett et al., 2003; Saez et al., 2005; Verselis and Srinivas, 2008). The release of ATP and glutamate from astrocytic Cx hemichannels induces neuronal death through activation of Panx1 hemichannels (Orellana et al., 2011), a process which may contribute to the increased apoptosis observed in epileptic tissue. Furthermore, astrocytes release ATP in response to raised intracellular Ca²⁺, which subsequently is broken down to adenosine. Adenosine, a neuromodulator, can inhibit presynaptic neurotransmitter release via the activation of P2X receptors (Pascual et al., 2005; Montana et al., 2006; Zhang et al., 2007; Pankratov et al., 2009).

The strongest evidence linking GJs and seizures is the seizure-blocking actions of agents that disrupt connexon-based GJ intercellular communication in in vitro and in vivo epilepsy models (e.g., Figure 1A). Although there are several GJC blockers available which possess anticonvulsive actions, they are not specific and do not usually discriminate between Cx isoforms or cell types (Perez-Velazquez et al., 1994; Ross et al., 2000; Kohling et al., 2001; Rozental et al., 2001; Jahromi et al., 2002; Steinhauser and Seifert, 2002; Szente et al., 2002; Gajda et al., 2003; Samoilova et al., 2003; Bostanci et al., 2006; Gigout et al., 2006; Nilsen et al., 2006; Bostanci and Bagirici, 2007; Medina-Ceja et al., 2008; Giaume and Theis, 2010). Carbenoxolone is one of the drugs widely used a GJ blocker, although one report showed a proconvulsive effect (Voss et al., 2009). Intracellular alkalinization, such as caused by trimethylamine and ammonium chloride, increases GJC, and also enhances epileptiform activity (Perez-Velazquez et al., 1994; Kohling et al., 2001). An antimalarial drug, quinine strongly blocks neuronal GJCs formed by Cx36 and to a lesser degree, Cx45 (Srinivas et al., 2001), but paradoxically enhances in vitro neocortical seizure activity caused by low magnesium perfusion, possible by blocking inhibitory interneuronal synchrony (Voss et al., 2009). Compounds of the fenamates family have also been reported to inhibit GJs composed of Cx43 with the following order of efficacy; meclofenamic acid > niflumic acid > flufenamic acid (Harks et al., 2001). Flufenamic acid was also shown to inhibit GJs composed of Cx26 and Cx32, but the selectivity was reported to be low (Srinivas and Spray, 2003). Most of the gap junctional blockers were also shown to affect Cx hemichannels and Panx channels (Giaume and Theis, 2010). Gap junctional blocking agents also affect the conductance of Panx membrane channels (Thompson and Macvicar, 2008; Giaume et al., 2013). Compared to Cx channels, Panx1 channels and Cx hemichannels are even more sensitive to most GJ channel blockers including carbenoxolone (CBX), flufenamic acid, mefloquin (MFQ) (Bruzzone et al., 2005; Iglesias et al., 2008) and Cx mimetic peptides (Dahl, 2007).

Several of the compounds from the fenemate family of blockers block Panx1 hemichannels in the following order: mefloquine > carbenoxolone > flufenamic acid (Iglesias et al., 2009). Studies have also shown that Panx1 hemichannels are 1000- to 10,000-fold more sensitive to mefloquine than Cx43 GJCs (Cruikshank et al., 2004; Iglesias et al., 2008). Although substances such as gadolinium (Gd³⁺) and lanthanum (La³⁺) block hemichannels without inhibiting gap junction or Panx1 channels, these ions are not specific to GJ proteins since they also block other channels such as Ca²⁺ channels (Mlinar and Enyeart, 1993; Liu et al., 2008). Long-chain alcohols are known to block Cx hemichannels but have only a very small effect on Panx channels. In contrast, low concentrations (5–10 μ M) of carbenoxolone block Panx1 channels and only have a minor inhibitory effect on Cx hemichannels (Schalper et al., 2008).

Connexin mimetic peptides, which correspond to the extracellular loops of connexins, are a class of specific and reversible inhibitors of GJC (Evans and Boitano, 2001; Herve and Dhein, 2010). In rat organotypic cultures, which show spontaneous epileptiform activity, prolonged (>10 h) application of Cx43 mimetic peptides, reduced the spontaneous seizure activity (Samoilova et al., 2008). This long period of incubation was required presumably to block the docking of homologous Cx43 hemichannels, a process which forms GJs, thereby diminishing inter-astrocytic GJC. These data suggest that enhanced inter-astrocytic gap junctional based coupling promotes seizure activity. Under normal physiological conditions, the probability of hemichannels opening is low because of the blocking effects of divalent extracellular ions such as magnesium and calcium (Ebihara, 2003; Ebihara et al., 2003). But when extracellular magnesium and/or calcium levels are reduced, for example, in the case of low Mg²⁺ or low Ca²⁺ seizure models, or the reduced extracellular Ca²⁺, which is seen during seizures in several seizure models (Lux and Heinemann, 1978; Somjen, 2002) the role of hemichannels and the effects of mimetic peptides must also be considered. In this case, higher concentrations of mimetic peptides are presumed to be required in order to block both GJ channels as well as hemichannels (Ebihara, 2003; Ebihara et al., 2003). In a slice culture model wherein added bicuculline caused epileptiform activity, low concentrations of a Cx43 mimetic peptide (amino acid sequence VDCFLSRPTEKT) targeting the extracellular loop two of Cx43, mainly blocked Cx43 hemichannels and prevented seizure-induced neuronal death (Yoon et al., 2010). Higher doses of the peptide, which inhibited GJs in addition to the hemichannels, increased the severity of the seizure-induced lesion. These data suggest that in this model, the neuronal damage from epileptiform neuronal hyperactivity is prevented by Cx43 gap junctional blockade and exacerbated by Cx43 hemichannel opening.

Developmental age plays a role in gap junctional expression and seizure susceptibility. It is well-known that immature humans and animal models are more susceptible to undergoing seizures than mature adults (Rakhade and Jensen, 2009). Also expression of glial and neuronal connexins is developmentally regulated (Rozental et al., 2000; Montoro and Yuste, 2004). Within the developing neocortex, significant GJC was measured in four compartments (Sutor and Hagerty, 2005); (1) gap junction-coupled neuroblasts of the ventricular zone and GJs in migrating cells and radial glia, a compartment which disappears with maturity; (2) gap junction-coupled glial cells (astrocytes and oligodendrocytes); (3) gap junction-coupled pyramidal cells (which exists primarily during the first two post-natal weeks in rodents); and (4) gap junction-coupled inhibitory interneurons. It is hypothesized that the increased GJC between pyramidal cells plays a major role in lowering the seizure threshold in immature animals, but the roles played by the age-dependent differential expression of the various gap junctional and pannexin proteins remain unclear. Among the 21 known members of Cxs and three Panxs, it has been determined that there are at least 10 Cxs and 2 Panxs expressed in the brain (Willecke et al., 2002; Baranova et al., 2004). Their expression patterns are not uniform and the distribution of different gap junction subtypes is dependent upon location and developmental maturity (Willecke et al., 2002; Baranova et al., 2004).

Analysis of the mRNA expressions of different Cxs and Panxs in the hippocampal tissues from 15 day mice, showed the highest expression of astrocytic Cxs 43 and 30, followed by Panxs, and the lowest expression was observed in the neuronal Cxs 36, 40, and 45, a few orders of magnitude lower than the pannexins and astrocytic connexons (Mylvaganam et al., 2010). Since the half-lives of most Cxs are short, between 1 and 5 h (Herve et al., 2007), changes in Cx expression can rapidly alter Cx levels in cells (Laird, 2010). Recent studies have demonstrated that several mechanisms are involved in the control of Cx expression. In addition, the GJ proteome which includes complex translational and post-translational mechanisms (e.g., phosphorylation) that can introduce changes in protein conformation, activity, charge, stability and localization, can alter downstream signaling pathways that may contribute to the pathophysiology of several neurological diseases (Laird, 2010; Gehi et al., 2011; Chen et al., 2013). Recently, Cx43 phosphorylation and Panx1 glycosylation were shown to occur following in vitro Co2+-induced seizures (Mylvaganam et al., 2010). Although Panx regulation has not been explored to the level studied for Cxs, from the functional point of view, Panxs might also show in seizure models, compensatory, overlapping, or unique physiological roles compared to those of Cxs.

In brief, the roles of GJs, connexins, and pannexins in epilepsy remain unclear. The lack of specificity of pharmacological agents now used to block or enhance GJC is a challenge if one wants to understand the role of a particular type of connexin-based GJ, and if wants to develop a specific therapeutic tool. The fact that most of these agents affect the Cx hemichannels and the Panx membrane channels further confuses the interpretation of experimental data. The ongoing development of peptides targeting specific Cxs and Panxs will permit a more precise dissection of the functions of the different Cx and Panx species. Although the gap junctional blocking drugs are almost always anticonvulsant, there is not as of yet an anticonvulsant drug on the market that is reputed to have gap junctional blocking properties, although both acetazolamide and topiramate inhibit carbonic anhydrase activity which should cause an intracellular acidosis thereby blocking GJC. Another problem is the overwhelming evidence that tissue from animal models and human epileptics show increases in Cxs expression in glia but not in neurons. This begs the question as to what are the roles of glia in seizure generation. If applying a gap junctional blocking agent to epileptic tissue is blocking interglial GJC, then what is the underlying anticonvulsant mechanism? Or could these agents have their major anticonvulsant action by blocking the conductance of Cx hemichannels and/or Panx channels? The scientific community is just now starting to consider the possible role of these membrane channels in the pathophysiology of epilepsy. In summary, although there is no doubt that GJs, connexins and pannexins are intimately related to epilepsy and seizure generation, the specific details of exactly how they are involved and how we can modulate their function for therapeutic purposes remain to be elucidated.