The adenosine kinase hypothesis of epileptogenesis

2.1. The acute brain injury

Most experimental data regarding the pathophysiology of epileptogenesis have been obtained from studies, in which epileptogenesis was induced by status epilepticus (SE), which will be discussed in the following. It remains to be determined whether these mechanisms can be extrapolated to other triggers of epileptogenesis, such as stroke or traumatic brain injury. According to several studies, SE, induced by kainic acid (KA), pilocarpine, or electrical stimulation, is associated with hippocampal cell loss, particularly in the dentate hilus (Dudek et al., 2002; Pitkanen et al., 2002; Sutula, 2004). In a rat model of SE, evoked by electrical stimulation of the amygdala, hilar cell loss correlated positively with duration between SE and time point of analysis, but not with the number of spontaneous seizures occurring after SE (Pitkanen et al., 2002). Similar results were obtained in a related study on parahippocampal neuronal damage (Gorter et al., 2002). Most importantly, however, mossy fiber spouting does not appear to be necessary for spontaneous seizures to occur (Longo and Mello, 1998) and no correlation was found between mossy fiber sprouting and the quantity of seizures in a chronic seizure model of pilocarpine-induced SE (Longo and Mello, 1999). These experimental data have been confirmed in an analysis of 572 human resected MTLE specimens (Mathern et al., 2002). According to these studies an initial epileptogenesis precipitating injury was principally responsible for hippocampal neuronal loss, while late occurring recurrent limbic seizures may further damage the hippocampus. In addition, hippocampal sclerosis, the neuropathological hallmark of MTLE, was commonly (87%) preceded by an initial precipitating injury (Mathern et al., 2002). In summary, these studies suggest that an initial precipitating injury (e.g. SE) causes both acute neuronal cell loss and hippocampal sclerosis (i.e. astrogliosis).

2.2. Astrogliosis

Astrogliosis is a hallmark of the epileptic brain. Autopsy and surgical resection specimens have suggested that chronic temporal lobe epilepsy and post-traumatic seizures may originate from gliotic scars (Duffy and MacVicar, 1999; Rothstein et al., 1996; Tashiro et al., 2002). The involvement of astrocytes in the pathogenesis of epileptogenesis can no longer be ignored since recent findings that astrocytes control synaptic transmission by the vesicular release of glutamate, ATP, and by the release of the NMDA receptor co-agonist D-serine (Haydon and Carmignoto, 2006). The concept of the tripartite synapse (Araque et al., 1999; Halassa et al., 2007a) is based on findings showing that glia respond to neuronal activity with an elevation of their internal Ca²⁺ concentration, which in turn triggers the release of chemical transmitters from glia and thus causes feedback regulation of neuronal activity and synaptic strength. This is even more important, since one cortical astrocyte contacts approximately 300 to 600 neuronal dendrites, leading to “functional islands of synapses” in which a given set of synapses can be controlled by gliotransmitter release from one single astrocyte (Halassa et al., 2007b). These “local modules” appear to be integrated by neurotransmitter receptor crosstalk including receptor heteromerization (Ferre et al., 2007). Thus, enhanced or reduced gliotransmission might directly translate into neurological disease. Recently, it was demonstrated that prolonged episodes of neuronal depolarization evoked by astrocytic release of glutamate contributed to epileptiform discharges (Tian et al., 2005). This study provided support for the concept that astrocytes are involved in the initiation, maintenance and spread of seizures.

2.3. Chronic epilepsy

Among other histopathological changes, chronic epilepsy is largely characterized by selective neuronal cell loss, mossy fiber sprouting, and astrogliotic scar formation. This constellation obviously raises the chicken-and-egg-question, i.e. what is cause or consequence of chronic seizures, and challenges the traditional oversimplified concept, that chronic epilepsy is merely due to an imbalance of excitation and inhibition. Rather, the recruitment and synchronization of oscillatory networks is considered to be an emergent property of epileptic activity (Javitt and Coyle, 2004), which leaves open the question whether there might be a key regulator or endogenous inhibitor for the synchronization of neuronal networks. The concept of heterosynaptic depression was established 30 years ago by findings that high frequency stimulation of a particular excitatory pathway caused a depression of nearby pathways (Lynch et al., 1977). This could indeed be an endogenous mechanism to prevent synchronization of neuronal networks. Today it is well recognized that the purine ribonucleoside adenosine is the main messenger mediating heterosynaptic depression (Manzoni et al., 1994).

3.1. Adenosine A₁ receptors

Inhibitory neuromodulation by adenosine is largely mediated by activation of A₁ receptors that are coupled to inhibitory G_i or G_o containing G-proteins. These inhibit adenylyl cyclase, activate inwardly rectifying K⁺ channels, inhibit Ca²⁺ channels and activate phospholipase C. As a net result, the release of various neurotransmitters, in particular glutamate is inhibited by presynaptic A₁ receptors. Consequently, adenosine and A₁-selective receptor agonists are effective in seizure suppression (Boison, 2007a) and neuroprotection (Cunha, 2005). In addition to these synaptic effects, A₁ receptors are believed to provide beneficial extra synaptic effects, which are based on a decrease in brain metabolism (Haberg et al., 2000) and the control of astrocyte function (van Calker and Biber, 2005). Tonic activation of A₁ receptors by an endogenous tone of adenosine is believed to lead to long-distance A₁ receptor mediated inhibition and thus to mediate heterosynaptic depression (Manzoni et al., 1994). In regard to epilepsy, A₁ receptors, which have high expression levels in epilepsy-prone regions such as hippocampus (Fredholm et al., 2001), would thus be uniquely positioned to convey paracrine anticonvulsant functions of adenosine (Güttinger et al., 2005a; Güttinger et al., 2005b) and to prevent the spread of epileptogenic activity (Fedele et al., 2006; Kochanek et al., 2006).

3.2. Adenosine A_2A receptors

In contrast to A₁ receptors, A_2A receptors are coupled to stimulatory G_s or G_olf proteins (Dunwiddie and Masino, 2001) and both inhibitory as well as excitatory responses mediated via A_2A receptors have been described in discrete brain areas (Fredholm et al., 2001). Inhibitors of the A_2AR have profound neuroprotective functions (Cunha, 2005) and the capability to prevent apoptosis (Silva et al., 2007). In addition to regulating neuronal vulnerability, these receptors play a key role in modulating the action of other neurotransmitters and neuromodulators (Sebastiao and Ribeiro, 2000). In contrast to the global action of inhibitory A₁ receptors the activity of A_2A receptors appears to be locally restricted to active synapses (Ferre et al., 2005). Thus, stimulated nerve terminals can synaptically release ATP. This occurs preferentially at high frequency stimulation (Cunha et al., 1996b), and extracellular degradation of synaptically released ATP to adenosine is not associated with the activation of inhibitory A₁, but with activation of facilitatory A_2A receptors (Cunha et al., 1996a). Synaptic activation of A_2A receptors can subsequently downregulate A₁ receptors or its responses (Ciruela et al., 2006; Lopes et al., 1999). Thus, synaptic stimulation of A_2A receptors under high frequency conditions in epileptic circuits could lead to downregulation of A₁ receptors, a finding confirmed in chronic epilepsy (Ekonomou et al., 2000; Glass et al., 1996; Rebola et al., 2003). An additional layer of complexity in adenosine receptor function is created by heteromerization with each other (e.g. A₁/A_2A heteromers) and with other neurotransmitter receptors (e.g. A_2A/D₂ heteromers) (Fuxe et al., 2007). In summary, a shift in the A₁ / A_2A receptor ratio in chronic epilepsy, may reinforce the epileptic state and limit the therapeutic efficacy of adenosine-augmenting therapies. Nevertheless, augmentation of adenosine with an ADK inhibitor is very effective in preventing pharmacoresistant seizures in a mouse model of chronic MTLE (Gouder et al., 2004). These powerful therapeutic effects may be due to a global increase in adenosine, which is not restricted to the synapse, but able to exert predominantly inhibitory effects by activation of extra-synaptic A₁R_s. Eventually, the combination of adenosine-augmentation with A_2AR antagonism might be a preferable strategy for therapeutic intervention. In this regard, it needs to be mentioned that a safe A_2AR antagonist KW-6002 is available and already in phase III clinical trials in patients with advanced Parkinson’s disease (Bara-Jimenez et al., 2003; Hauser et al., 2003).

3.3. Adenosine A_2B and A₃ receptors

Less is known about the contribution of the low-affinity A_2B and the low-density A₃ receptors to the regulation of epileptic circuits. While A₁ and A_2A receptors have important functions under physiological conditions, A_2B and A₃ receptors may be of relevance under pathological conditions. As an example, traumatic brain injury or cerebral ischemia lead to a massive surge in adenosine, elevating ambient levels of adenosine more than tenfold to levels sufficient to activate A_2B and A₃ receptors (Clark et al., 1997; Pearson et al., 2006). Recent findings suggest that the physiological consequences of A₃ receptor activation show dose-dependent effects: While CA1 hippocampal A₃ receptors stimulated by adenosine released during brief ischemia might exert A₁ receptor-like protective effects on neurotransmission, severe ischemia (i.e. higher dose of adenosine) would transform the A₃ receptor-mediated effects from protective to injurious (Maria Pugliese et al., 2007). In line with this notion, activation of A₃ receptors led to heterologous desensitization of A₁ receptors (Dunwiddie et al., 1997), and activation of A_2A and A₃ receptors by endogenous adenosine aggravated epileptic activity in slice cultures (Etherington and Frenguelli, 2004).

3.4. Pharmacological approaches for seizure suppression

Due to the widespread systemic distribution of adenosine receptors pharmacological approaches to antiepileptic therapy are limited by widespread systemic side effects including suppression of cardiovascular functions (Dunwiddie and Masino, 2001; Monopoli et al., 1994). While A₁ receptor-selective antagonists exacerbate convulsions indicating that endogenous adenosine modulates epileptiform activity via the A₁ receptor (Etherington and Frenguelli, 2004), A₁ selective agents such as 2-chloro-N⁶-cyclopentyladenosine (CCPA) have previously been used to suppress seizures in various models of epilepsy (Huber et al., 2002; Monopoli et al., 1994). In addition, A₁R agonists were shown to be effective in a mouse model of pharmacoresistant epilepsy, in which seizures could not be suppressed by conventional antiepileptic drugs (Gouder et al., 2003). The latter finding implies that augmentation of the adenosine system might be a general strategy for the treatment of otherwise refractory epilepsy. The potential of A_2A receptor activation for seizure suppression has been investigated in only few studies and the results are partly controversial and seem to depend on the animal models and drugs used (De Sarro et al., 1999; Huber et al., 2002; Malhotra and Gupta, 1997; Zhang et al., 1994).

In addition, approaches involving the chronic activation of adenosine receptors by selective ligands may lead to rapid adaptive changes, thus limiting their clinical potential (Jacobson et al., 1996). The low blood brain barrier permeability and pronounced peripheral side effects of adenosine receptor agents (Malhotra and Gupta, 1997) constitute a significant limitation to their therapeutic use. Even the combination of the non-brain permeable non-selective adenosine receptor antagonist 8-sulfophenytheophylline with brain permeable selective adenosine receptor agonists failed to abolish centrally mediated sedative side effects in a rat model of audiogenic brain stem epilepsy (Huber et al., 2002). Therapeutically, it thus appears to be more reasonable to increase the extracellular level of adenosine by controlling its clearance or metabolism.

4.1. Presynaptic vesicular release of ATP

Most cells in the brain, including neurons and astrocytes, can release ATP via regulated vesicular transport (Fields and Burnstock, 2006). Extracellular ATP is then cleaved into adenosine by a cascade of ectonucleotidases (Cunha, 2001; Zimmermann, 2000). The presynaptic release of ATP and its extracellular conversion to adenosine has important consequences for the modulation of synaptic transmission. Thus, ATP was preferentially released (and subsequently degraded to adenosine) upon high-frequency stimulation of hippocampal slices (Cunha et al., 1996b) leading to the preferential activation of excitatory A_2A receptors (Cunha et al., 1996a); this mechanism allows the selective potentiation of high frequency bursts within a tonically inhibited network.

4.2. Astrocytic vesicular release of ATP

The elegant work of Phil Haydon’s group has recently identified vesicular release of ATP from astrocytes as a major source of extracellular adenosine under physiological conditions (Pascual et al., 2005). Transgenic overexpression of a dominant-negative form of synaptobrevin-2 selectively in astrocytes disrupted the vesicular release of ATP from astrocytes (Zhang et al., 2004). This manipulation led to the loss of the A₁R mediated tonic inhibition in hippocampal slices (Pascual et al., 2005) indicating that under physiological conditions astrocytic release of ATP (and subsequent cleavage by ectonucleotidases) is a major source of extracellular adenosine. Since each astrocyte contacts thousands of synapses (Ventura and Harris, 1999), it is conceivable that astrocytic release of ATP has the function to set a global adenosine-mediated inhibitory tone within a neuronal network, while presynaptic release of ATP from neurons (see above) and its conversion to adenosine may have the function to selectively potentiate individual synaptic transmission within a globally inhibited network.

4.3. Nucleoside transporters

Since most nucleosides and their analogues are relatively hydrophilic and thus cannot readily pass the plasma membrane, nucleoside transporters are essential in adjusting extra- and intracellular adenosine levels (Cass et al., 1999). In addition, nucleoside transporters play an important role in the transport of antiviral nucleoside derivatives (Yamamoto et al., 2007). Human nucleoside transporters are classified into three known isoforms of concentrative nucleoside transporters (CNT) that are coupled to a Na⁺ gradient (Gray et al., 2004; Smith et al., 2007) and into four known isoforms of equilibrative nucleoside transporters (ENT), which rapidly equilibrate pools of extra- and intracellular adenosine (Baldwin et al., 2004). Under physiological conditions ENTs appear to play the predominant role in adjusting extra- and intracellular pools of adenosine in the brain, since the extracellular concentration of adenosine can be enhanced by inhibition of ENTs with dipyridamole, dilazep, nitrobenzylthioinosine, or cannabidiol (Carrier et al., 2006; Kong et al., 2004).

4.4. Control of extracellular adenosine by intracellular metabolism via adenosine kinase

As outlined above, under physiological conditions, astrocytic release of ATP and its extracellular degradation to adenosine appears to be the major source of extracellular synaptic adenosine, which is then rapidly equilibrated with the adenosine pool of intracellular compartments due to the ubiquitous expression of nucleoside transporters. Thus, intracellular metabolism of adenosine is uniquely suited to control the influx of adenosine into the cell and thus the levels of extracellular adenosine. Adenosine kinase (ADK, EC 2.7.1.20), which phosphorylates adenosine to 5’-AMP is considered to be the key metabolic enzyme for the regulation of extracellular adenosine levels in brain (Boison, 2006). Thus, inhibition, or genetic disruption, of ADK leads to rapid increases in extracellular adenosine (Fedele et al., 2004; Güttinger et al., 2005b; Huber et al., 2002), while overexpression of ADK leads to a reduction of the synaptic adenosine tone (Fedele et al., 2005). Alternative adenosine metabolizing enzymes such as adenosine deaminase or S-adenosyl-homocysteine hydrolase appear to play minor roles in the regulation of brain adenosine (Boison, 2006). Given the existence of an equilibrative transport system for adenosine, ADK fulfills the role of a metabolic reuptake system for adenosine, being the motor driving the influx of adenosine into the cell. The existence of a highly active substrate cycle between adenosine and AMP involving ADK and 5’-nucleotidase in brain allows minor changes in the activity of ADK to be translated rapidly into major changes in ambient adenosine (Arch and Newsholme, 1978; Bontemps et al., 1983). Thus, hypoxia-induced inhibition of ADK potentiated cardiac adenosine release (Decking et al., 1997), and pharmacological inhibition of ADK increased endogenous adenosine and depressed neuronal activity in hippocampal slices (Pak et al., 1994). Since intracellular ATP levels are about 100,000 times higher than respective adenosine levels (Fredholm et al., 2005a), modulation of ADK activity will translate rapidly into changes in adenosine, but will not influence levels of ATP. Likewise, ADK will not influence the de novo synthesis of purines, which is dependent on the formation of IMP as a common precursor of AMP and GMP.

Most importantly, in adult brain, ADK is almost exclusively expressed in astrocytes (Studer et al., 2006). Given the astrocytic synaptic release of ATP as a major source of adenosine, an astrocyte-based adenosine-cycle becomes plausible (Fig. 1). Thus synaptic concentrations of adenosine are dependent on astrocytic release of ATP, extracellular cleavage by ectonucleotidases into adenosine, and ADK-driven reuptake via nucleoside transporters. This implies that astrocytes play a prominent role in setting the inhibitory adenosine tone, which is of importance in setting the thresholds for seizure susceptibility.

5.1. Biphasic regulation of adenosine kinase expression

As an acute response to stress (e.g. SE, ischemia, brain injury) ADK is rapidly downregulated leading to elevated levels of adenosine, thus enhancing the protective functions of the adenosine system via increased activation of A₁Rs. Thus, the induction of ischemia in hippocampal slices led to the rapid increase in extracellular adenosine within 10 min to levels >20 µM above baseline (Pearson et al., 2006). This rapid rise of adenosine was demonstrated to be due to the release of adenosine as such and was not dependent on extracellular cleavage of ATP (Frenguelli et al., 2007). In line with these findings, oxygen-glucose deprivation in cultured cortical neurons led to downregulation of ADK and rapid rise in extracellular adenosine, which was maintained for at least one hour (Lynch et al., 1998). In vivo, ADK was downregulated within 2 hours after SE in mice (Gouder et al., 2004) and within 3 hours after middle cerebral artery occlusion in a mouse model of stroke (Pignataro et al., 2007a). In these in vivo models downregulation of ADK was transient, reaching back to normal levels within 24 hours (Gouder et al., 2004; Pignataro et al., 2007a). Downregulation of ADK was also demonstrated as a consequence of seizure preconditioning in a recent gene expression study (Borges et al., 2007). Thus, acute downregulation of ADK and the resulting increase in extracellular adenosine might be a general astrocyte-based protective mechanism of the brain.

Conversely, under conditions of chronic epilepsy, ADK was found to be upregulated. Thus, kainic acid-induced SE led to a 1.8 fold increase in hippocampal ADK activity and spontaneous seizures within a time frame of four weeks (Gouder et al., 2004). Otherwise pharmacoresistant seizures in this mouse model (Gouder et al., 2003) could be suppressed by application of the ADK inhibitor 5-iodotubercidin (3.1 mg/kg, i.p.); 5-iodotubercidin-mediated seizure suppression was abolished by co-application of the A₁R antagonist DPCPX. These findings demonstrate that upregulation of ADK in chronic epilepsy is implicated in epileptogenesis by reducing the adenosine tone at the A₁R. Upregulation of ADK was also found in chronically kindled rats and in the rat pilocarpine model of epilepsy (D. Boison, unpublished observations). Upregulation of ADK in chronic epilepsy is paralleled by changes in G-protein coupled receptor expression such as downregulation of A₁Rs (Ekonomou et al., 2000; Rebola et al., 2003; Rebola et al., 2005), reorganization of cannabinoid type 1 receptors (Falenski et al., 2007), or increases in GABA_B receptor expression (Kokaia et al., 2001; Morimoto et al., 2004).

In summary, ADK expression undergoes a biphasic regulation during epileptogenesis, with acute downregulation after injury (e.g. after SE) but upregulation in chronic epilepsy. Since in adult brain ADK is primarily an astrocytic enzyme, astrogliosis is likely the culprit for enhanced ADK expression in chronic epilepsy. How does the biphasic regulation of ADK expression relate to astrogliosis during epileptogenesis?

5.2. Astrogliosis

Several lines of evidence indicate that increased levels of adenosine may promote or trigger astrogliosis, an effect that appears to be largely mediated by A_2ARs. Thus, activation of the A_2AR potentiates synaptic actions of brain-derived neurotrophic factor (BDNF) in the hippocampus (Diogenes et al., 2007), stimulates glutamate outflow and leads to excessive glial activation (Popoli et al., 2007). In rat cortex, micro-infusion of the potent A_2AR agonist 5′-(N-cyclopropyl)-carboxamidoadenosine (CPCA) increased astrogliosis, an effect that was abolished by co-infusion of the adenosine A_2AR antagonist 1,3-dipropyl-7-methylxanthine (DPMX) (Hindley et al., 1994). Likewise, A_2AR blockade abolished basic fibroblast growth factor-induced reactive gliosis in rat striatal primary astrocytes (Brambilla et al., 2003). These findings nicely demonstrate that A_2AR stimulation promotes astrogliosis. Thus, it is compelling to conclude that under conditions of brain injury, hypoxia, or inflammation, which all cause both upregulation of A_2ARs (Cunha, 2005) as well as acute increase in adenosine (Dunwiddie and Masino, 2001), an adenosine-based mechanism contributes to astrogliosis. This notion is further supported by findings that stimulation of A₁Rs, which are downregulated during epileptogenesis (Ekonomou et al., 2000; Rebola et al., 2003; Rebola et al., 2005), leads to a reduction of astrocyte proliferation (Rathbone et al., 1991).

5.3. Hippocampal upregulation of adenosine kinase and hippocampal seizures

The arguments described above thus support the ADK hypothesis of epileptogenesis (Fig. 2): A variety of acute insults to the brain including brain injury, SE, and ischemic stroke lead to a homotypic brain response consisting of acute downregulation of ADK, rapid increase in extracellular adenosine, downregulation of A₁Rs and upregulation of A_2ARs, which in turn regulate astrocyte proliferation. Thus, as a consequence of acute downregulation of ADK, astrogliosis results. As described above, astrogliosis in a mouse model of kainic acid induced hippocampal epileptogenesis correlated with a 1.8-fold upregulation of ADK activity and spontaneous non-convulsive intrahippocampal partial seizures (Gouder et al., 2004). Importantly, spontaneous seizures were restricted to the kainic-acid injured astrogliotic hippocampus displaying upregulation of ADK, while generalized seizures were not detected using cortical electrodes (Gouder et al., 2004). These findings imply that spontaneous seizures “co-localized” to upregulated ADK, but also that seizures did not spread to other brain regions. It is important to note that activation of A₁Rs by ambient adenosine prevents the spread of seizures (Fedele et al., 2006); thus, normal adenosine levels in non-astrogliotic brain regions are expected to prevent the spread of focal seizures, which are possible only within the context of upregulated ADK (Fig. 3) (= reduced adenosine). The experiments described above suffered somewhat from the rather diffuse hippocampal upregulation of ADK in this model (Gouder et al., 2004).

More recently a co-localization of astrogliosis, upregulation of ADK and spontaneous electrographic seizures were described at high spatial resolution in a novel mouse model of CA3-restricted epileptogenesis (Li et al., 2007a): C57BL/6 wild-type mice received intraamygdaloid injections of kainic acid (KA), which triggered SE (terminated after 30 min with lorazepam) and subsequent neuronal cell loss restricted to the ipsilateral CA3 region. During a time span of 3 weeks astrogliosis developed together with upregulation of ADK, all restricted to the ipsilateral CA3. Most importantly, spontaneous electrographic seizures at a frequency of around 4 seizures per hour developed exclusively within the ipsilateral CA3, but were not recorded elsewhere (CA1, dentate gyrus, cortex). This experiment demonstrates high special colocalization of spontaneous electrographic seizures with astrogliosis and upregulated ADK, and suggests that astrogliosis and/or upregulation of ADK contribute to epileptogenesis.