In our attempt to understand how neurons process information in the neuronal network, the observation that astrocytes listen and talk to synapses by exerting both excitatory and inhibitory actions on neurons (Araque et al., 1999; Brockhaus and Deitmer, 2002; Di Castro et al., 2011; Fellin et al., 2004; Jourdain et al., 2007; Kang et al., 1998; Panatier etal., 2006, 2011; Parpura et al., 1994; Pascual etal., 2005; Pasti et al., 1997, 2001; Perea and Araque, 2007; Serrano et al., 2006; Shigetomi et al., 2011; Zhang et al., 2003) represents one of the most relevant finding in brain research over the last decades. This discovery literally revolutionized our view of a brain function based “only” on billions of neurons interacting dynamically in the neuronal network. Against this background, it is not surprising that an integrated view of the role of astrocytes not only in the processing of sensory information, but also in the genesis of brain disorders is now gradually emerging (Blackburn et al., 2009; Maragakis and Rothstein, 2006; Seifert et al., 2006). A full understanding of the astrocyte role in brain pathologies represents, indeed, a formidable challenge in neurobiological research.
An increasing body of evidence has documented a dysregulation of astrocyte-specific functions in human and experimental epilepsy. We will focus this article on the role of gliotransmission in the generation of epileptiform activity. We will first provide some background information and a brief overview of the “historical” findings that suggested an active role of astrocytes in the epileptic brain network. We will then discuss the most recent studies that unravelled an intriguing complexity in the astrocyte action in this brain disorder.
Epilepsy Is a Complex Brain Disorder
Epilepsies comprise a family of multifactorial neurological disorders that affects at least 50 million people worldwide (Ngugi et al., 2010; Thurman et al., 2011). Seizures represent the clinical manifestation of epilepsies caused by both genetic and acquired factors, such as trauma, perinatal injury, postinfection lesions, and tumors. The seizure reflects a highly synchronous neuronal discharge that arises at restricted brain sites, the epileptogenic foci, and then secondarily spreads to large portions of the brain (Avoli et al., 2002; Jefferys, 1990; Pinto et al., 2005; Traub and Wong, 1982; Trevelyan et al., 2006). Epilepsy can thus be considered a disorder of excess synchronisation of neurons, fundamentally linked to an imbalance between excitatory and inhibitory activities that produces hyperexcitability. In support of this view, in experimental animal models recurrent seizures are evoked by enhancing excitatory and/or impairing inhibitory activity.
In the last decade, significant advances have been made in epilepsy research and we have knowledge of the experimental conditions that may lead to neuronal hyperexcitability. For example, mutations in genes encoding voltage-gated Na+, Ca2+, and K+ channels, as well as those involved in inhibitory synaptic transmission (Cl- channel) have been identified to be associated with the so-called idiopathic epilepsies (McNamara et al., 2006). This observation is not surprising given that a dysfunction of ion channels, which directly regulate membrane excitability, can lead to uncontrolled neuronal hyperexcitability.
Our knowledge of the cellular events that cause a brain tissue to become epileptic is, however, largely unsatisfactory. For example, how and why specific subsets of neurons from a specific region are suddenly activated and display the widespread synchronous activity that characterizes epileptic discharges? The ictal event, that represents the cellular correlate of the seizure, occurs sporadically and seems not to be predictable. It can be preceded by a series of interictal discharges, but it can also occur without any previous neuronal hyperactivities. These observations lead to the following fundamental question: what is the nature of the signal that triggers the ictal transition, thus leading to seizure onset? In spite of an intense experimental research in this field, the mechanism at the basis of seizure generation, propagation, and cessation remains poorly defined.
The extracellular concentration of K+ ions ([K+]o) is an important factor that regulates the neuronal excitability in the brain network. An increased [K+]o deriving from intense neuronal firing discharges tends, indeed, to depolarize neuronal cells and facilitate the development of epileptiform discharges. Peaks of 10–12 mM in the [K+]o are reached during the hypersynchronous neuronal activities that characterized epileptic disorders and are proposed to play an important role in ictogenesis (Heinemann et al., 1977; Pedley et al., 1976).
The role of astrocytes as modulators of epileptogenesis was initially proposed over 20 years ago and for a long period this role focused on the ability of astrocytes to buffer extracellular K+ or neurotransmitters (released in excess during epileptic discharges). The K+ channels mainly responsible for the astrocyte K+ buffering capacity are a class of inward rectifying potassium (Kir) channels. Sixteen different Kir channel subunits have been identified and the Kir4.1 and, to a lesser extent, the Kir2.3 are the most intensively studied channels (Neusch etal., 2006; Olsen and Sontheimer, 2008). Unlike other K+ channels that allow large K+ efflux from depolarized cells, Kir channels have the peculiarity to mediate small K+ efflux from depolarized cells and large influx at hyperpolarized potentials. Importantly, Kir4.1 channels have a high open probability at resting potential (Ransom and Sontheimer, 1995) and a channel conductance that is proportional to [K+]o (Newman, 1993; Sakmann and Trube, 1984). These features allow astrocytes to remove large amounts of K+ from the extracellular space. Notably, Kir4.1 channels are expressed on astrocytes in different brain regions (Higashi et al., 2001; Ishii et al., 1997; Li etal., 2001; Poopalasundaram et al., 2000; Takumi et al., 1995) and mediate the majority of K+ currents in these cells (Seifert et al., 2009).
As crucial regulators of [K+]o and network excitability astrocytic Kir channels might be targets of a novel therapeutic approach in epilepsies. However, while studies performed over the last decade, in both animal models and human epilepsy, demonstrated that a dysregulation of K+ buffering in the astrocyte network predisposes to neuronal hyperexcitability and seizures (Wallraff et al., 2006), additional studies appear necessary to fully understand whether or not a defective removal of extracellular K+ by astrocytes is a major factor in the excessive K+ accumulation that characterizes the epileptic tissue.
Gliotransmission in the Generation of Epileptiform Activities
The discovery that through a Ca2+-dependent glutamate release astrocytes can directly excite groups of neighbouring neurons (Parpura et al., 1994) and favour synchronised activities mediated by extrasynaptic N-methyl-D-aspartate (NMDA) receptor activation (Fellin et al., 2004) were the initial observations that hinted at a more direct role of Ca2+-dependent gliotransmission in the generation of epileptiform activities. Indirect support for this hypothesis were results from subsequent studies in brain slice and in vivo preparations that described a significant increase in the frequency of Ca2+ oscillations in astrocytes during epileptiform activity (Fellin et al., 2006; Tian et al., 2005), and its reduction in the presence of anticonvulsant drugs (Tian et al., 2005). In animal models of temporal lobe epilepsy, the astrocytic expression of mGluRs that mediates Ca2+ oscillations was also found to be increased (Aronica et al., 2000; Ulas et al., 2000). These data suggest that the excessive synchronization of neuronal activity that characterizes the epileptic discharge might derive, at least in part, from a hyperactivity of astrocytes. In support of an astrocytic role in epileptogenesis, it has been reported that the paroxysmal depolarising shifts, i.e., the cellular correlate of interictal events recorded between seizures, are tetrodotoxin-resistant and mediated by glutamate released from astrocytes (Tian etal., 2005). This conclusion was, however, disputed by others (Fellin et al., 2006; Fellin and Haydon, 2005) who demonstrated that astrocytic activation of neuronal NMDA receptors was not necessary for the generation of inter-ictal-like or ictal-like events and fuelled a controversial debate on the role of these glial cells in focal epileptogenesis (D'Ambrosio, 2006; Seifert et al., 2006; Wetherington et al., 2008). More recent results obtained in a slice model of focal epilepsy hint at a contribution of astrocytes to the generation of focal seizures-like ictal discharges rather than of interictal discharges (Gomez-Gonzalo et al., 2010). In this model - that we recently developed in slice preparations obtained from the rat and mouse entorhinal cortex (EntCx) (Gomez-Gonzalo et al., 2010; Losi et al., 2010)—an episode of intense activity was triggered in a small group of layer V-VI neurons by two subsequent NMDA stimulations obtained by applying brief pressure pulses to an NMDA-containing glass pipette. In the presence of 4-amino pyridine (4-AP, 50-100 μM) (Perreault and Avoli, 1989; Perreault and Avoli, 1991; Rutecki et al., 1987) and 0.5 mM Mg2+, double NMDA stimulations reliably evoked focal ictal discharges. Notably, a single NMDA stimulation was regularly observed to be ineffective. Because in this model we know in advance when and where a focal seizure will initiate (Gomez-Gonzalo et al., 2010; Losi et al., 2010), we have the unique opportunity to study the cellular events that develop early at the focal site and predispose neurons to generate a seizure discharge there. We found that after a double, but not a single NMDA stimulation, a high number of astrocytes exhibited a large, tetrodotoxin-sensitive Ca2+ elevation just prior to the onset of the ictal discharge. This early Ca2+ elevation in astrocytes was not a mere consequence of neuronal activity and it rather had a causative role in the generation of focal ictal discharges. Indeed, after Ca2+ elevations in astrocytes from the focal site of ictal discharge generation were inhibited by the Ca2+ chelator 1,2-bis(o-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid (BAPTA), introduced in the astrocyte syncytium by patching an individual astrocyte with a BAPTA-containing pipette, the episode of neuronal hyperactivity induced by double NMDA stimulations failed to generate an ictal discharge In the presence of BAPTA, the number of recruited neurons upon the double NMDA pulse was significantly lower than in controls (Gomez-Gonzalo et al., 2010) suggesting that when an episode of hyperactivity in a group of neurons, such as that stimulated by a double NMDA pulse application, consistently engages nearby astrocytes, a feed back signal, i.e., Ca2+-dependent release of glutamate and/or D-serine, is generated causing a larger population of neurons to be recruited into a coherent synchronous activity. If this feedback signal operates on a brain network prone to seizures—a condition that in our slice models was established by lowering the extracellular Mg2+ and applying either 4-AP or picrotoxin—it contributes to drive neurons towards the ictal discharge threshold (Gomez-Gonzalo et al., 2010; Losi et al., 2010). The initiation site is thus represented not only by the neurons activated by NMDA, but also by those that are secondarily activated in a recruitment process that involves astrocytes. Consistent with this view, it was also found that a single NMDA pulse—that repetitively failed to activate an ictal discharge—became effective when it was coapplied with a selective stimulation of astrocytes by TFLLR (Gomez-Gonzalo et al., 2010). This peptide has been reported to trigger in astrocytes both Ca2+ elevations and glutamate release by activating the thrombin protease activated receptor-1 (PAR-1) (Gomez-Gonzalo et al., 2010; Lee et al., 2007; Shigetomi et al., 2008). Notably, in both rodent and human brain, the PAR-1 receptor is present in some specific neuronal populations, such as the hippocampal granule cells, but it is prominently expressed in astrocytes (Junge et al., 2004). In rodent, CA1 hippocampal region and EntCx no functional evidence, however, has been obtained for the presence of PAR-1 receptors in neurons (Gomez-Gonzalo etal., 2010; Lee et al., 2007), and results from immunocytochemical experiments revealed that in the rat EntCx PAR-1 receptor expression is restricted to astrocytes (Gomez-Gonzalo et al., 2010). Consequently, the PAR-1 specific activator TFLLR can be reasonable used in these regions to stimulate Ca2+ elevations in astrocytes selectively.
In agreement with the view that a critical number of neurons from a restricted brain site need to be intensively activated in order to generate an ictal discharge, we found that the number of neurons activated upon NMDA/TFLLR coapplications was significantly higher than that activated by NMDA alone. Most interestingly, we also found that when the astrocyte contribution was reduced by inhibiting Ca2+ signals in these cells, as in the BAPTA experiments, the ictal discharge could be recovered by applying three, instead of two, NMDA pulse applications (Gomez-Gonzalo et al., 2010). A plausible interpretation of this latter observation is that the three successive NMDA stimulations directly activated a large number of neurons and evoked a level of correlated neuronal activity that were sufficient for seizure-like discharge generation, bypassing the astrocyte contribution in the recruitment process. If this single result is interpreted in the context of the results obtained in the other experiments, the plausible conclusion is that the contribution of astrocytes is not an absolute requirement to, but it can be important for ictal discharge generation. If instead the result from the experiments with the three NMDA pulse were interpreted in isolation from all the other results, the only plausible, but ultimately erroneous, conclusion would be that astrocytes had no role in seizure generation. These observations underline the importance of exhaustively evaluating neuron-astrocyte reciprocal signaling before concluding how astrocytes contribute to the remarkable complexity of the brain network.
An additional collective property of astrocytes that may contribute to the control of seizure propagation is represented by ATP-mediated Ca2+ waves propagating across the astrocyte network. Consistent with this hypothesis, a recent study provided evidence that a local Ca2+ decrease in the extracellular space initiates a Ca2+ wave in astrocytes mediated by ATP release through connexin 43 hemichannels (Torres et al., 2012). Most interestingly, this astrocytic ATP enhances inhibitory transmission by acting on the P2Y1 receptors expressed in a subset of hippocampal interneurons (Torres et al., 2012). Given that during the epileptic discharges the extracellular Ca2+ is markedly reduced, the consequent release of astrocytic ATP may potentiate inhibitory transmission thereby working as an anticonvulsant feedback mechanism that opposes seizure propagation.
The complex role of astrocytes in seizure generation is schematically illustrated in Fig. 1 The figure is composed with bars of different height which represent the mass of neurons activated upon different NMDA stimulation intensities, in the absence (left, grey bars) or presence (right, black-white bars) of an astrocyte contribution. The threshold for ictal discharge generation in EntCx slices in two different experimental models, i.e., the picrotoxin model and the 4-AP model (both used a low extracellular Mg2+), is also reported (dashed lines). The picrotoxin model can be considered a model with a low-threshold of epileptic discharges, as suggested by the recurrent epileptic discharges that occur spontaneously in this model. The 4-AP model can be considered a high-threshold model and, indeed, it lacks spontaneous epileptic activities. In the BAPTA experiments, in which astrocyte Ca2+ elevations were selectively blocked, a single NMDA pulse triggered an ictal discharge in the picrotoxin, but not in the 4-AP model. In this latter, high threshold model an ictal discharge could be evoked by three, but not by two NMDA pulses (bar a,b,c). Bars from the right side of the figure illustrate the contribution of astrocytes to ictal discharge generation, as evaluated in the experiments without BAPTA. A single NMDA pulse failed to activate a significant astrocyte response. Accordingly, its effect was unchanged with respect to that observed in the BAPTA experiments, and it resulted in an ictal discharge in the low, but not in the high threshold model (bar d). In contrast, upon a double NMDA pulse, astrocytes were massively activated and their signaling back to neurons (white bar and curved arrow in bar e) critically increased the overall activation of neurons. Under these conditions, the ictal discharge threshold was reached also in the high threshold model. An ictal discharge in this model could be also evoked when a single NMDA application (ineffective per se) was coupled with a TFLLR-mediated astrocyte activation (bar f), while astrocyte stimulation alone was effective only to evoke an ictal discharge in the low threshold model (bar g).
In summary, the threshold for a focal ictal discharge generation is reached when a critical mass of neurons enters into a phase of very intense activity. By amplifying the recruitment of neurons into a critical mass of synchronously active neurons astrocytes can be crucial for setting the threshold of ictal discharge generation. Their contribution, however, can vary depending on the general level of excitability in neurons and it can be bypassed, at least experimentally, by using protocols that comprise an overstimulation of neurons. A paradigmatic example of this is the recovery of an ictal discharge by a very intense stimulation of neurons (i.e., a three NMDA pulse stimulation, Fig. 1, bar c) in the experiments in which a double NMDA pulse failed to evoke an ictal discharge because astrocyte activation was blocked by BAPTA (Gomez-Gonzalo et al., 2010). Apparently, to unravel the contribution of astrocytes to epileptiform activities experimental conditions need to be accurately established. The same consideration holds when the astrocytic contribution to plastic changes of the synapse, such as long-term potentiation (LTP), is evaluated. Our results imply that the contribution of astrocytes to the generation of a variety of different LTP forms represents an additional level of complexity that can only be appreciated through carefully designed experiments. Indeed, similarly to the ictal discharge experimentally generated by three NMDA pulse (in which the astrocyte contribution is potentially irrelevant), if the system is “saturated” by an overstimulation of synaptic transmission, all the fine influences that modulate synaptic strength become irrelevant. Consistent with this conclusion, in aquoporin-4 knock-out mice the LTP evoked by theta-burst stimulation was found to be significantly reduced, while that evoked by high frequency stimulation was unaffected (Skucas et al., 2011). Given that aquoporin-4 channels are predominantly expressed in astrocytes, where they are functionally coupled with Kir 4.1 and Kir 5.1 channels, these results provide further support to the conclusion reported above and strengthen the hypothesis of a modulatory role of astrocytes in synaptic transmission.
Challenges that Must be Addressed to Understand the Role of the Astrocyte in Epilepsy
Our knowledge of the astrocyte role in epilepsy remains largely unsatisfactory and many important issues need to be clarified. Among these is the fact that astrocytes can release beside glutamate other gliotransmitters that have different impacts on seizure activity. This leads to the question of whether astrocytes have ultimately a pro- or an anti-convulsant action. An example of our difficulty in understanding the overall role of gliotransmission in epileptiform activities is ATP which is released from astrocytes through different mechanisms, including a Ca2+-dependent mechanism (Coco et al., 2003). Astrocytic ATP can exert on neurons an excitatory effect mediated by P2 receptor activation, but it can have also an inhibitory action. Under physiological conditions, adenosine derived from the degradation of astrocytic ATP by extracellular nucleotidases can, indeed, activate presynaptic, high affinity A1 receptors leading to inhibition of transmitter release and heterosynaptic depression [Pascual, 2005 #7468;Serrano, 2006 #8330;Zhang, 2003 #6691]. While these findings hint at an anticonvulsant role of astrocytes, the action of these cells in the epileptic brain may be even more complex if we consider that seizure induction in experimental epilepsy upregulates adenosine kinase (ADK) resulting in a proconvulsant decrease of extracellular adenosine concentrations. Noteworthy is that this enzyme is predominantly expressed in astrocytes (Gouder, 2004) and it constitutes an efficient metabolic reuptake system that controls ambient adenosine levels (Boison, 2005). Therefore, astrocytes are key regulators of the extracellular availability of the endogenous anticonvulsant adenosine. Notably, the ADK expression was also found to be increased in astrocytes from the hippocampus and temporal cortex of patients with temporal lobe epilepsy (Aronica et al., 2011), while genetic reduction of ADK in a kainic acid experimental model of epileptogenesis prevents seizures (Li et al., 2008). These observations suggest that ADK can be considered not only a diagnostic marker of epilepsies, but also a potential target for the development of a new therapy that could prevent epileptic seizures (Boison, 2008; Theofilas etal., 2011). All in all, it can be predicted that gliotransmission may soon be recognized as target for developing new antiepileptic therapies.
Full clarification of the functional significance of the different astrocyte-to-neuron signaling pathways represents an intriguing challenge in future studies. A great help to these studies is provided by an intense development of molecular genetic tools that allow one to affect selectively the different pathways mediating astrocyte-to-neuron signaling. However, one of the biggest problems that we still have with the study of the role of the astrocyte in the generation of epilepsy is that cell type selective manipulations have not yet been brought to this problem. Although the introduction of a Ca2+ chelator such as BAPTA is sufficient in acute brain slice studies to impair an astrocytic Ca2+ signal, it is not an approach that is appropriate for in vivo studies or for prolonged studies needed when video electroencephalography monitoring of spontaneous seizures is performed. Recent studies have shown that it is necessary to perform continuous recordings from rodents for months in order to identify the development of the seizure phenotype (Williams et al., 2009). To ask how an astrocyte contributes to this behaviour will require the introduction of molecular genetics to this problem. Recent studies have shown that perturbation of membrane fusion through the conditional expression of the SNARE domain of synaptobrevin II leads to changes in extracellular adenosine and as a consequence the trafficking of NMDA (Deng et al., 2011). Whether these changes are sufficient to alter disease progression is unclear at this time.
In addition to transgenic mice is the potential to use viral transduction approaches to study the role of the astrocyte in the development of epilepsy. However, great care needs to be taken with this approach as viral transduction per se can induce a condition known as reactive astrocytosis which changes the physiology of the brain. This unwanted side effect of the use of certain viruses was used advantageously in a recent study by Ortinsky and collaborators (Ortinski et al., 2010). In this study, adeno-associated virus serotype 5 was used to transduce astrocytes. High viral titers led to an increase in glial fibrillary acidic protein expression, the appearance of vimentin expression and a loss of glutamine synthetase expression. This is a typical hallmark of the epileptic brain. The astrocyte normally metabolizes glutamate to glutamine, via the activity of glutamine synthetase, and shuttles glutamine to neurons as a renewable source of transmitter. Gamma-aminobutyric acid (GABA)ergic interneurons rely on glutamine to synthesize GABA. In brain slices containing reactive virally transduced astrocytes, inhibitory transmission was impaired leading to hyperexcitability of the hippocampus. Simple addition of exogenous glutamine restored normal GABAergic transmission and excitability.
As we move forward in trying to understand the role of the astrocyte in epilepsy it is essential that we embrace opportunities to use molecular genetics to perturb this type of glial cells but also to integrate this approach with the laborious method of chronic video electroencephalography recordings. Only in this manner will we be able to determine whether and how the astrocytes modulate epilepsy.
It is worth underlying that some of the changes occurring in astrocyte signaling in the epileptic tissue resemble the dysregulation of molecules and pathways that occurs in other brain disorders. Indeed, astroglial glutamate receptors and transporters, water channels (aquaporins), Kir channels and connexins, that are found to be dysregulated in the epileptic brain, are also dysregulated in motor neuron disease, stroke, hepatic encephalopathy, schizophrenia, Huntington's disease, and Alzheimer's disease (Blackburn et al., 2009; Maragakis and Rothstein, 2006; Seifert et al., 2006). All in all, these findings strongly suggest that astrocytes might hold the key to understand the pathogenesis of these brain disorders.