Adenosine dysfunction in epilepsy



Extracellular levels of the brain's endogenous anticonvulsant and neuroprotectant adenosine largely depend on an astrocyte-based adenosine cycle, comprised of ATP release, rapid degradation of ATP into adenosine, and metabolic reuptake of adenosine through equilibrative nucleoside transporters and phosphorylation by adenosine kinase (ADK). Changes in ADK expression and activity therefore rapidly translate into changes of extracellular adenosine, which exerts its potent anticonvulsive and neuroprotective effects by activation of pre- and postsynaptic adenosine A1 receptors. Increases in ADK increase neuronal excitability, whereas decreases in ADK render the brain resistant to seizures and injury. Importantly, ADK was found to be overexpressed and associated with astrogliosis and spontaneous seizures in rodent models of epilepsy, as well as in human specimen resected from patients with hippocampal sclerosis and temporal lobe epilepsy. Several lines of evidence indicate that overexpression of astroglial ADK and adenosine deficiency are pathological hallmarks of the epileptic brain. Consequently, adenosine augmentation therapies constitute a powerful approach for seizure prevention, which is effective in models of epilepsy that are resistant to conventional antiepileptic drugs. The adenosine kinase hypothesis of epileptogenesis suggests that adenosine dysfunction in epilepsy undergoes a biphasic response: an acute surge of adenosine that can be triggered by any type of injury might contribute to the development of astrogliosis via adenosine receptor-dependent and -independent mechanisms. Astrogliosis in turn is associated with overexpression of ADK, which was shown to be sufficient to trigger spontaneous recurrent electrographic seizures. Thus, ADK emerges as a promising target for the prediction and prevention of epilepsy. © 2011 Wiley Periodicals, Inc.


The purine ribonucleoside adenosine is a primordial metabolite that likely played important roles in the origin of life on Earth (Oro,1961). As constituent of the energy metabolite adenosine-5′-triphosphate (ATP) and of RNA, adenosine is uniquely suited to adjust cellular activity (RNA as a primary source for new protein synthesis) to available energy (ATP). Adenosine has therefore been described as a “retaliatory metabolite” that protects cells against the effects of excessive energy deficits, an important role to maintain energy homeostasis in most organ systems including brain (Newby et al.,1985). Early work from Bertil Fredholm and Thomas Dunwiddie has first recognized adenosine as an endogenous modulator of neuronal excitability (Dunwiddie,1980; Fredholm and Hedqvist,1980).

Today the role of adenosine as endogenous anticonvulsant and neuroprotectant is well recognized (Boison et al.,2010; Fredholm et al.,2005a; Ribeiro,2005; Stone et al.,2009). Importantly, microdialysis studies performed in human patients with temporal lobe epilepsy have shown that endogenous adenosine levels rise as a consequence of seizures and consequently seizure termination is thought to be mediated by the seizure-induced release of endogenous adenosine (During and Spencer,1992). Adenosine provides potent seizure control in all experimental model systems studied including a model of pharmacoresistant epilepsy (Gouder et al.,2003). The anticonvulsant effects of adenosine are largely mediated by activation of pre- and postsynaptic Gi/o protein coupled adenosine A1 receptors (A1Rs), which induce presynaptic inhibition by reducing the inward flows of calcium, and which reduce excitability of the postsynaptic membrane by enhancing postsynaptic outflow of potassium through G-protein coupled inwardly rectifying potassium channels (Fredholm et al.,2005a). In addition to the global inhibitory tone provided by A1R activation, adenosine further fine-tunes neuromodulation through all four types of adenosine receptors (A1, A2A, A2B, A3), which interact, at least in part, by heterodimerization with other G protein coupled receptors, and thereby affect all major neurotransmitter and neurotrophin systems (Sebastiao and Ribeiro,2009). Therefore, endogenous adenosine is uniquely able to control neuronal excitability on multiple levels, and consequently any pathological disruption of adenosine homeostasis is likely to affect network excitability (Fig. 1).

Figure 1.

Disruption of network homeostasis in epilepsy. Epilepsy is a comorbid syndrome comprised not only of spontaneous recurrent seizures, but also of cognitive impairment, depression, psychiatric conditions, and changes in sleep patterns. Neuronal dysfunction is only part of the epilepsy syndrome and downstream of glial dysfunction, which includes adenosine deficiency, and which is influenced by changes in the blood brain barrier, inflammatory processes, and bioenergetic and epigenetic factors. It is therefore not surprising, that antiepileptic drugs that exclusively act on neuronal downstream targets fail to restore network homeostasis in epilepsy.

Astrocyte-Based Adenosine Cycle

Astrocytes play a key role in regulating adenosine homeostasis by releasing ATP as the major physiological precursor of adenosine. Elegant studies from Phil Haydon's lab have employed transgenic mice with an inducible astrocyte-selective dominant negative mutation of soluble N-ethylmaleimide-sensitive factor attachment protein receptor (dnSNARE-mice) to study the contribution of astrocytic ATP release to the regulation of synaptic transmission (Pascual et al.,2005). Importantly, in the mutant animals the adenosine-dependent tonic suppression of synaptic transmission was removed, providing the first direct evidence that astrocytes are the source of synaptic adenosine that regulates the tonic suppression of neuronal networks (Pascual et al.,2005). Apart from the vesicular release of ATP, astrocytes can also release ATP more directly via hemichannels (Halassa and Haydon,2010; Kang et al.,2008). Once in the extracellular space, ATP is rapidly dephosphorylated into adenosine via a series of ectonucleotidases. In contrast to classical neurotransmitters, the physiological activity of adenosine is not terminated by energy-driven transport systems. Instead, astrocytes express two types of equilibrative nucleoside transporters that permit rapid equilibration of intra- and extracellular levels of adenosine (Baldwin et al.,2004). Influx of adenosine into the astrocyte is driven by metabolic clearance through the intracellular enzyme adenosine kinase (ADK; EC, a phosphotransferase that converts adenosine into adenosine-5′-monophosphate (AMP). Importantly, the intracellular levels of adenine nucleotides are about 100,000-fold higher than levels of the nucleoside (i.e. adenosine) (Fredholm et al.,2005a); thereby, changes in ADK activity will affect preferentially the levels of ambient adenosine without affecting the pool of the energy metabolites AMP, ADP, and ATP. Due to the existence of a substrate cycle between adenosine and AMP, even minor changes in ADK activity rapidly translate into major changes in the tone of ambient adenosine. A recent study using online quantification of adenosine in hippocampal slice preparations with adenosine microelectrode biosensors documented that basal synaptic adenosine levels are largely regulated by astrocytic ADK (Etherington et al.,2009). Although astrocytic ATP (independent of ADK) represents the major source for synaptic adenosine, its reuptake is controlled by intracellular astrocytic ADK; thus the tone of ambient adenosine is controlled by an astrocyte-based adenosine-cycle.

Adenosine Kinase (ADK)—An Astrocyte-Based Enzyme

In the adult brain ADK is predominantly expressed in astrocytes (Studer et al.,2006). Two alternatively spliced isoforms of ADK exist, ADK-long and ADK-short, which are expressed within the nucleus and cytoplasm, respectively (Cui et al.,2009). Interestingly, during early brain development in rodents (Studer et al.,2006) and humans (unpublished data) ADK expression undergoes a coordinated switch from primarily neuronal expression (nuclear isoform) to primarily astrocytic expression (nuclear and cytoplasmic isoform), with the exception of neurons from the olfactory bulb, which maintain cell division and high levels of ADK expression into adulthood (Gouder et al.,2004). The early developmental expression of ADK within the nucleus, but not the cytoplasm, of immature neurons and maintenance of nuclear ADK expression in adult astrocytes (Studer et al.,2006) suggests a specific function of nuclear ADK related to gene regulation, development, or cellular plasticity. It is striking to note that only those cells that are newly born or still capable to divide, in particular immature neurons, adult olfactory neurons, and astrocytes, express the nuclear isoform of ADK. In contrast, the cytoplasmic isoform of ADK appears to be involved primarily in the regulation of the adenosine tone (see below for details), which in adult brain is maintained within the 25 to 250 nM range (Fredholm et al.,2005a). Thus, ADK appears to be an enzyme with multiple functions (Studer et al.,2006). In particular, putative functions of ADK related to gene regulation and cellular plasticity might play crucial roles in pathological conditions, such as epilepsy, that involve long-lasting changes in network plasticity. ADK is essential because deletion of the gene leads to early postnatal mortality due to reduction of body temperature, prolonged periods of apnea, and fatal hepatic steatosis (Boison et al.,2002).

ADK—A Key Regulator of Synaptic Adenosine

An alternative pathway for metabolic adenosine clearance is based on deamination of adenosine into inosine via the metabolic enzyme adenosine deaminase (ADA). However, Km values for adenosine in rat brain are much lower for ADK (∼2 μM) than those for ADA (∼17 μM) (Phillips and Newsholme,1979) suggesting that ADK is the key enzyme for metabolic adenosine clearance under physiological baseline conditions, whereas the contribution of ADA might become relevant only under conditions of stress, which lead to increased extracellular adenosine concentrations (Latini and Pedata,2001). In contrast to the hypothalamus where ADA is highly expressed and thought to be implicated in sleep regulation, the hippocampus is the brain region with the lowest expression and activity levels of ADA (Geiger and Nagy,1986). Several lines of direct experimental evidence now demonstrate that ADK is the key regulator of hippocampal adenosine. Data derived from hippocampal slices demonstrated that pharmacological inhibition of ADK but not of ADA increased endogenous adenosine and depressed neuronal activity (Pak et al.,1994). Using a combination of electrophysiology and adenosine biosensor technology in an in vitro hippocampal slice model of electrically-evoked epileptiform activity, it was shown that pharmacological inhibition of ADK led to an increase in synaptic adenosine, which in turn suppressed glutamatergic excitatory synaptic transmission in an A1R-dependent manner under both baseline and Mg2+-free (ictogenic) conditions (Etherington et al.,2009). Under conditions of ADK inhibition, seizure activity induced by high-frequency stimulation was significantly reduced (Etherington et al.,2009). These findings indicate that ADK, expressed only in astrocytes of the adult hippocampus, controls synaptic adenosine homeostasis under both baseline and pathological conditions. Proof, that changes in ADK expression effect basal adenosine levels likewise in vivo was recently achieved using adenosine biosensors in mice with genetically engineered changes in ADK expression in brain. Triple-mutant fb-Adk-def mice underexpress ADK in the entire telencephalon, but to overexpress ADK in all other brain areas (Li et al.,2008). Adenosine microelectrode biosensors implanted into cerebral cortex (65% of wild-type ADK) or striatum (164% of wild-type ADK) of fb-Adk-def mice revealed significant increases (163% of wild-type cortex) and decreases (50% of wild-type striatum) of adenosine levels (Shen et al.,2011). These data demonstrate an inverse relationship between ADK expression and adenosine levels in vivo and imply that pathological increases in ADK results in adenosine deficiency.

Overexpression of ADK and Adenosine Deficiency in Epilepsy

Astrogliosis is a pathological hallmark of the epileptic brain. Due to the astrocyte-specific expression profile of ADK we hypothesized that astrogliosis might disrupt homeostatic functions of the ADK-adenosine system, which might have direct implications on the epileptic phenotypic spectrum. Overexpression of ADK in hypertrophied astrocytes was first identified in a mouse model of mesial temporal lobe epilepsy (MTLE) in which status epilepticus (SE) triggered by a unilateral intrahippocampal injection of the excitotoxin kainic acid (KA) resulted over a time span of 4 weeks in ipsilateral neuronal cell loss in area CA1, mossy fiber sprouting, granule cell dispersion, hippocampal sclerosis, and spontaneous recurrent seizures (Gouder et al.,2004) (Fig. 2). Interestingly, ADK expression followed a biphasic kinetic following the KA injection: acute downregulation of ADK during the first 24 h following the injury was followed by gradual overexpression of the enzyme during the progression of epileptogenesis (Gouder et al.,2004). Eight weeks following the KA-injection, the epileptogenic hippocampus was characterized by a 177% increase in the enzymatic activity of ADK suggesting a relationship among increased ADK immunoreactivity, enhanced metabolic adenosine clearance, and seizures (Gouder et al.,2004). Importantly, seizures that are resistant to conventional antiepileptic drugs in this model, were completely suppressed after treating the animals with the ADK inhibitor 5-iodotubercidin (ITU; 3.1 mg/kg, i.p.), a therapeutic effect that was prevented when ITU was combined with the A1R antagonist 8-cyclopentyl-1,3-dipropylxanthine (DPCPX; 1 mg/kg i.p.) (Gouder et al.,2004). Although astrogliosis and overexpression of ADK correlated with spontaneous seizures in this model of MTLE, a causal relationship between astrogliosis and ictogenesis could not be established in this complex model of epilepsy, which, like human MTLE, is characterized by a multitude of histopathological alterations.

Figure 2.

Overexpression of ADK in the KA-treated hippocampus of a mouse. Representative images of the hippocampal formation (A, B) and dentate gyrus (C, D, areas magnified from insets shown in A and B) from coronal brain sections of a control mouse (A, C, n = 4) and an epileptic mouse (B, D, n = 4) 4 weeks after the intrahippocampal injection of KA after staining for ADK immunoreactivity (immunoperoxidase method) show profound overexpression of ADK in astrocytes of the epileptic hippocampus. Note the lack of ADK expression in dispersed granular neurons (located between arrow heads in B) and prominent expression in hypertrophied astrocytes (see arrows in D) of the epileptic dentate gyrus. Specific areas of the Ammon's horn are indicated as CA1, CA2, and CA3. DG identifies the dentate gyrus. Scale bars: A, B = 1 mm; C, D = 250 μm.

To isolate the contribution of astrogliosis and associated dysfunction of adenosine homeostasis to seizure generation in epilepsy we developed and characterized a minimalistic mouse model of focal epileptogenesis that is comprised of focal astrogliosis, associated overexpression of ADK, and spontaneous electrographic seizures, but without any confounding histopathological alterations commonly seen in MTLE such as neuronal cell loss, mossy fiber sprouting, or granule cell dispersion. In this model a single unilateral injection of KA into the basolateral amygdala triggered focal onset SE that was terminated after 30 min with lorazepam. As a result of this manipulation acute neuronal cell loss was restricted to the ipsilateral amygdala and CA3 (Li et al.,2008,2012). Importantly, within 3 weeks astrogliosis, overexpression of ADK, and spontaneous electrographic seizures at a rate of around four seizures per hour resulted at the sites of the original injuries, i.e. restricted to the ipsilateral amygdala and CA3 (Li et al.,2008,2012). These data demonstrated a tight spatial association of astrogliosis, overexpression of ADK, and spontaneous recurrent seizures. In a time-course experiment, onset of seizures at around 12 days following the KA-injection was also associated with the onset of astrogliotic overexpression of ADK suggesting not only a spatial but also a tight temporal association of astrogliosis, overexpression of ADK, and spontaneous recurrent seizures (Li et al.,2007a). To identify the focal origin of seizures and to study mechanisms of seizure spread in this model, we recently performed synchronous electroencephalographic (EEG)—recordings from multiple intracranial leads. Specifically, we recorded simultaneously from the ipsilateral amygdala, CA3, CA1, dentate gyrus, and the cerebral cortex (Li et al.,2012). Spontaneous seizures were restricted to the ipsilateral amygdala and CA3. Importantly, seizures between these two foci occurred independently and were not synchronized based on the focal nature of adenosine deficiency (Li et al.,2012). These focal seizures appeared stable over time; however, they rapidly synchronized and generalized once adenosine signaling was globally impaired in brain. Thus, astrogliosis, associated dysfunction in adenosine signaling, and focal onset of subclinical seizures might constitute an early event in the epileptogenic cascade (Li et al.,2012).

Overexpression of ADK appears to be a general pathological hallmark of the epileptic brain. It seems to be independent of the model used and was recently described following the electrical induction of SE in rats (Aronica et al.,2011). Most importantly, robust overexpression of ADK and association with astrogliosis was found in epileptogenic tissue from human patients with hippocampal sclerosis that underwent surgical resection of their hippocampus as a treatment of their epilepsy (Aronica et al.,2011; Masino et al.,2011b). These findings suggest that overexpression of ADK is also a prominent feature of human epilepsy and further imply that consequential adenosine deficiency is at least a contributing factor to seizure generation in human epilepsy.

Overexpression of ADK is Sufficient to Trigger Seizures

The evidence outlined above demonstrates a tight association of astrogliosis, overexpression of ADK, and ictogenesis. However, to ascribe a specific role of increased ADK (i.e. decreased adenosine) for ictogenesis, it is necessary to molecularly dissect overexpression of ADK from the underlying astrogliosis. Several lines of evidence now suggest that overexpression of ADK per se and in the absence of astrogliosis or any other epileptogenic event sufficiently triggers electrographic seizures. Thus, transgenic Adk-tg mice with a brain-wide 1.4-fold overexpression of the cytoplasmic isoform of ADK on top of an ADK-deficient background (to eliminate the endogenous regulatory mechanisms of ADK expression) experience spontaneous recurrent hippocampal seizures at a rate, frequency, and duration comparable to spontaneous seizures in the mouse model of intra-amygdala KA-induced focal epilepsy (Li et al.,2007a,2008). More direct evidence for the crucial role of ADK for seizure generation was recently obtained in a viral gene expression approach. Adeno associated virus serotype 8 (AAV8) was used to overexpress a cDNA of the cytoplasmic isoform of ADK under the control of a truncated GFAP promoter as a molecular strategy to selectively overexpress ADK in astrocytes (Theofilas et al.,2011). Four weeks following the intrahippocampal injection of the ADK-overexpressing AAV8, robust electrographic seizure activity was recorded from the AAV8-injected hippocampus (Theofilas et al.,2011). Importantly, overexpression of the cytoplasmic isoform of ADK in astrocytes was sufficient to trigger the seizures. Conversely, the knockdown of ADK in Adk-tg mice using an AAV8-based antisense approach robustly abrogated the spontaneous seizures within the injected hippocampus (Theofilas et al.,2011). Together, the transgenic and viral approaches described above demonstrate that overexpression of ADK as such and in the absence of astrogliosis is sufficient to trigger spontaneous electrographic seizures. However, what are the causal relationships between astrogliosis, overexpression of ADK, and ictogenesis?

Astrogliosis and ADK

Overexpression of ADK always seems to be associated with astrogliosis, even in nonrelated pathologies such as Alzheimer's disease and amyotrophic lateral sclerosis (unpublished data). It is unlikely that overexpression of ADK is the cause for astrogliosis, since the AAV8-based overexpression of ADK in astrocytes did not trigger astrogliosis (Theofilas et al.,2011). However, the temporal coincidence of astrogliosis and overexpression of ADK (Li et al.,2007a) suggests that the two might be triggered by common mechanisms.

ADK and seizures

The evidence described above clearly demonstrates that overexpression of ADK can directly cause electrographic seizures. However, the reverse assumption that seizures might cause overexpression of ADK requires some thought: It is known that seizures trigger the release of endogenous adenosine (During and Spencer,1992), which is expected to require higher demands on metabolic clearance. Therefore it is tempting to speculate that the seizure-associated release of adenosine might also contribute to trigger astrogliosis and overexpression of ADK as a means to enhance metabolic clearance of adenosine. This could in fact lead to a self-propagating vicious cycle in which seizures induce overexpression of ADK, which in turn promotes further seizure activity. Based on current data however, this scenario seems to be unlikely. In the rat kindling model electrically induced seizures do not seem to trigger astrogliosis or overexpression of ADK (unpublished data). Moreover, in the mouse model of focal epileptogenesis, the degree of ADK overexpression appears to be stable over time at least for several weeks (Li et al.,2012). In addition, A1R knockout mice, which experience spontaneous recurrent electrographic seizures that are quantitatively and qualitatively equivalent to seizures recorded from Adk-tg mice or from epileptic mice following the intraamygdaloid injection of KA, do not display any signs of overexpression of ADK (Li et al.,2007a; Masino et al.,2011b).

Overexpression of ADK in Epilepsy: A Likely Explanation for Comorbidities

Clinically, comorbid cognitive impairments are among the most debilitating and persistent concerns of chronic epilepsy, with a comorbidity rate reaching up to 64% compared with a maximal of 27% among nonepilepsy controls (Hermann et al.,2008). In particular, episodic memory impairment is now recognized as a key feature of MTLE and cognitive impairment has been described as “the most problematic of the comorbidities of epilepsy” (Bell et al.,2011).

As outlined above, overexpression of ADK—and resulting adenosine deficiency—in sclerotic tissue of the epileptic brain can be a direct cause for seizures. Studies performed in ADK-overexpressing Adk-tg mice not only revealed spontaneous electrographic seizures, but also comorbidities commonly associated with epilepsy (Yee et al.,2007). In comparison with nontransgenic control mice, Adk-tg mice were characterized by devastating deficits on both the working and reference memory versions of the Morris water maze spatial memory tests, and by pronounced and selective deficits in the acquisition and expression of a conditioned response to a discrete conditioning stimulus as assessed in the conditioned freezing paradigm (Yee et al.,2007). Thus, adenosine deficiency in Adk-tg mice was not only associated with spontaneous seizures but also with significant comorbid impairment. Adenosine deficiency in epilepsy is also likely to affect psychomotor control by insufficient activation of A2A receptors (Shen et al.,2008), sleep regulation (Palchykova et al.,2010), and might contribute to depression (Hanson,2009). Therefore it is tempting to conclude that glial disruption of adenosine homeostasis in epilepsy is not only associated with epileptic seizures but also with several comorbidities that are part of the epileptic syndrome. This conclusion might provide a more holistic view of the complexity of the disease.

Seizure Suppression by Therapeutic Adenosine Augmentation

If adenosine deficiency is a pathological hallmark of epilepsy, then therapeutic adenosine augmentation should be an effective tool to control epileptic seizures. Indeed, systemic augmentation of adenosine signaling by either A1 receptor agonists or by ADK-inhibitors effectively suppress seizures in a wide range of epilepsy models including those that are resistant to conventional AEDs (Benarroch,2008; Gouder et al.,2003; McGaraughty et al.,2005). Unfortunately, systemic augmentation of adenosine signaling is not a therapeutic option due to widespread, mainly cardiovascular, side effects, and due to liver toxicity of ADK disruption (Boison et al.,2002; Fredholm et al.,2005a). Therefore, focal adenosine augmentation approaches, with the aim to reconstruct adenosine homeostasis within an epileptogenic brain area, become a therapeutic necessity. Focal adenosine augmentation to suppress epileptic seizures was first successfully performed in the rat kindling model of epilepsy using intraventricular implants of baby hamster kidney fibroblasts that were engineered to release adenosine and that, before transplantation, were encapsulated into a semipermeable polymer membrane to prevent immune rejection (Huber et al.,2001). In this approach recipients of the adenosine releasing cells were completely protected from generalized kindled seizures, whereas recipients of control cells continued to display their preimplantation seizure profile (Huber et al.,2001). Importantly, focal adenosine augmentation, in contrast to systemic adenosine augmentation, was not associated with any sedative side effects (Güttinger et al.,2005). Unfortunately, duration of seizure control (∼12 days) was limited by a reduced life expectancy of the encapsulated cells (Huber et al.,2001). Focal adenosine augmentation as a strategy for seizure control was subsequently validated in an independent laboratory using a strategy of intracranial injection of adenosine to prevent epileptiform events in a rat model (Anschel et al.,2004). Currently, four different approaches are being pursued to make therapeutic use of adenosine augmentation and are as follows.

Adenosine releasing silk

Silk is an FDA-approved biocompatible biopolymer that can be engineered for controlled focal drug release. In two recent studies, infrahippocampal implants of silk that were engineered to release defined daily doses of up to 1 mg adenosine provided complete control of seizures in the kindled rat for the duration of the adenosine release (1 mg adenosine per day for 10 days) (Szybala et al.,2009), but also dose-dependently suppressed kindling epileptogenesis when transplanted before the onset of kindling (Szybala et al.,2009; Wilz et al.,2008).

Adenosine releasing stem cells

Adenosine releasing mouse embryonic stem cells and human mesenchymal stem cells were engineered to release adenosine based on genetic disruption of the endogenous Adk gene or by expressing a miRNA directed against Adk via a lentiviral vector (Fedele et al.,2004; Ren et al.,2007). Progeny of adenosine releasing stem cells potently suppressed kindling epileptogenesis after transplantation into the infrahippocampal fissure (Li et al.,2007b). More importantly, when adenosine releasing stem cells were transplanted into the infrahippocampal fissure of mice 24 h after the intra-amygdaloid injection of KA, 3 weeks later astrogliosis was significantly attenuated, ADK expression levels were normalized, and not a single seizure was recorded in >100 h of EEG recordings. In contrast, recipients of control cells, or sham control animals displayed profound astrogliosis, overexpression of ADK, and approximately four seizures per hour in area CA3 ipsilateral to the injection of KA (Li et al.,2008). These findings support a novel antiepileptogenic effect of focal adenosine delivery.

Gene therapy

As outlined above, glial expression of an Adk-antisense construct in the hippocampus of Adk-tg mice via an AAV8-based gene expression system completely abrogated seizures in this model system (Theofilas et al.,2011). Thus ADK is a therapeutic target for antisense-based gene therapy approaches.

Ketogenic diet

Ketogenic diets have successfully been used for >80 years to suppress seizures in patients with epilepsy including those with refractory seizures, although the underlying anticonvulsive mechanisms were incompletely understood (Kossoff and Rho,2009). We recently demonstrated that a ketogenic diet reduces the expression of ADK in mice (Masino et al.,2011b). While the diet effectively suppressed seizures in adenosine deficient Adk-tg mice, it failed to suppress seizures in A1R knockout mice (Masino et al.,2011b). These findings demonstrate that dietary manipulations can be used to enhance adenosine signaling and A1R activation in the brain.

The ADK Hypothesis of Epileptogenesis

The data discussed above suggest that astrogliosis and changes in adenosine homeostasis are intricately linked to both epileptogenesis and subsequent ictogenesis. The ADK hypothesis of epileptogenesis is based on biphasic changes in adenosine homeostasis during disease progression (Boison,2008) and the following sequence of events is proposed and supported by experimental data as indicated: Acute injury to the brain, e.g. brain trauma (Clark et al.,1997), stroke (Pignataro et al.,2008), or SE (Gouder et al.,2004) leads to an immediate and acute surge in adenosine at micromolar levels that is further potentiated by acute downregulation of ADK. In models of stroke and SE, ADK levels drop down to minimum levels at 2 to 3 h following the injury and gradually recover to normal within 24 h (Gouder et al.,2004; Pignataro et al.,2008). The acute adenosine response makes physiological sense, as it significantly enhances the neuroprotective capacity of the brain. Under conditions of impaired adenosine signaling the brain is indeed more susceptible to aggravated neuronal injury and lethal outcome following traumatic brain injury, stroke, or SE (Fedele et al.,2006; Kochanek et al.,2006; Li et al.,2008; Pignataro et al.,2007). While beneficial acutely following an initial epileptogenesis precipitating injury, the same surge of adenosine might also be a trigger for subsequent epileptogenesis. Several adenosine receptor mediated downstream effects might contribute to epileptogenesis, among which increased signaling via the A2A receptor appears to be crucial.

Brain insults lead to an increase in A2AR signaling

The expression of A2ARs in glial cells appears generally to be induced following brain insults. In one study, LPS treatment led to a robust induction of A2AR mRNA and protein in primary cultures of glial cells with a peak at 48 h after the treatment (Saura et al.,2005). Likewise, A2AR expression was induced in microglial cells and astrocytes of mouse substantia nigra 24 h after intoxication with 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP).

Stimulation of astrogliosis

Astrocyte proliferation is in part regulated by adenosine receptors expressed on the astrocyte membrane. In particular, activation of A2ARs by extracellular adenosine increased astrocyte proliferation and activation, whereas inhibition of the A2AR prevented brain derived neurotrophic factor (BDNF)-induced astrogliosis (Brambilla et al.,2003; Hindley et al.,1994).


Adenosine modulates important functions of the brain immune system (Hasko et al.,2005). It is well known that inflammatory responses, microglial activation, and changes in the blood brain barrier contribute to epileptogenesis (Ravizza et al.,2008; Uva et al.,2008; Vezzani et al.,2008). Adenosine is known to stimulate the proliferation of naive microglial cells via simultaneous activation of A1 and A2A receptors (Gebicke-Haerter et al.,1996). Additional proinflammatory effects of adenosine are mediated via A2A receptor dependent upregulation of cyclooxygenase 2 (COX-2) and the release of prostaglandin E2 (PGE2) (Fiebich et al.,1996).

As outlined above, once astrogliosis and overexpression of ADK are established, spontaneous recurrent electrographic seizures result (i.e. contribution to ictogenesis). Indeed, overexpression of ADK as such is sufficient to trigger those seizures. It is important to note that the ADK-associated seizures are frequent (usually around four seizures per hour), relatively short in duration (around 20 s), but electrographic and subclinical in nature; ADK-associated seizures in our model of focal epileptogenesis can only be recorded with intracranial electrodes, co-localize with overexpressed ADK, and under normal conditions these seizures appear to be stable over both space and time, likely because the adenosine system in the vicinity of the epileptogenic focus is still intact (Li et al.,2007a,2008,2012). In support of this notion, we have previously demonstrated that A1R activation is necessary to keep an epileptogenic focus localized and to prevent seizure spread (Fedele et al.,2006). Consequently, injection of animals that express focal ADK-associated seizures with a nonconvulsive dose of the A1R antagonist DPCPX can turn those pre-existing electrographic seizures into clinical grade convulsive seizures (Li et al.,2012). As indicated above, multiple pathological events can trigger focal astrogliosis; such triggers can range from minor traumatic injuries or microstrokes to pathologies with known involvement of astrogliosis, such as Alzheimer's disease, autism, or amyotrophic lateral sclerosis (Rosengren et al.,1992; Schiffer and Fiano,2004; Van Eldik and Griffin1994). Could it be that “silent seizures” linked to astrogliosis and focal adenosine deficiency, are fairly widespread—if not common? Unless those seizures have a clinical manifestation, they would not necessarily be detected. However, those pre-existing subclinical seizures might still constitute a risk factor for subsequent seizure generalization and development of epilepsy. Of note, patients with Alzheimer's disease frequently suffer from episodes of sudden severe confusion that can best be explained by the occurrence of such “silent seizures” in deep brain structures. In addition, epileptic (clinical) seizures constitute a well-known comorbidity of Alzheimer's disease. Thus, focal ADK-associated seizures might well be a substrate for future seizure generalization and might constitute a first step in the epileptogenic cascade, i.e. a condition comprised of astrogliosis, overexpression of ADK, and spontaneous electrographic seizures. A secondary event or a secondary “hit” might be required leading to network rewiring, or failure of other endogenous antiepileptogenic systems might become involved. In this regard it is important to note that A1R downregulation was found both in animal models with spontaneous recurrent convulsive seizures as well as in temporal lobe resections from patients with intractable epilepsy (Ekonomou et al.,2000; Glass et al.,1996). These findings suggest that downregulation of A1Rs in chronic epilepsy might be a necessary prerequisite for seizure spread and generalization.

Adenosine Receptor-Independent Effects of Adenosine

While adenosine receptor dependent pathways of adenosine are well characterized (Fredholm et al.,2005b,2011; Stone et al.,2009) and have extensively been explored within the context of epilepsy (Benarroch,2008; Boison,2005; Jacobson and Gao,2006) it is important to point out that adenosine has additional adenosine receptor independent functions that have not yet been explored within the context of epilepsy. Of note are as follows.

Transmethylation and epigenetics

Adenosine is an obligatory biochemical endproduct of all transmethylation reactions (Mato et al.,2008). It results from the cleavage of S-adenosylhomocysteine (SAH) into adenosine and homocysteine via the bidirectional enzyme SAH-hydrolase. If adenosine is not constantly removed via ADK, the equilibrium of this reaction is shifted toward SAH formation (Boison et al.,2002). Increased SAH potently inhibits transmethylation, which also includes DNA- or histone methylation. As a primordial biochemical regulator, adenosine could thus have direct influence on epigenetic functions, which are thought to be involved in epileptogenesis, and which form the basis of the “methylation hypothesis,” suggesting that seizures by themselves can induce epigenetic chromatin modifications, thereby aggravating the epileptogenic condition (Kobow and Blumcke,2011). Interestingly, the nuclear isoform of ADK undergoes tightly controlled developmental expression switches during early postnatal brain development (Studer et al.,2006), suggesting a direct involvement of the enzyme in gene regulation.

Mitochondrial bioenergetics

As a “retaliatory metabolite” adenosine is tightly linked to its metabolic precursors AMP, ADP, and ATP. It is important to note that there is no direct biochemical biosynthetic pathway for the synthesis of adenosine. Purine biosynthesis results in 5′-inosine-monophosphate (IMP), which is then transformed into AMP. Thereby, adenosine is intricately linked to its phosphorylated precursors and the energy state of the cell. Mitochondrial bioenergetics and the mitochondrial capacity to produce ATP as the major precursor of adenosine thereby critically influence homeostatic functions of the adenosine system (Masino and Geiger,2008; Masino et al.,2011a). Conversely, adenosine homeostasis and ADK expression influence mitochondrial function, which is suggested by a severe mitochondrial pathology in Adk-knockout mice (Boison et al.,2002). Mitochondrial dysfunction in turn has been implicated in epileptogenesis (DiMauro et al.,2002; Kunz,2002). In particular, mitochondrial complex I deficiency has been identified in the epileptic focus of patients with temporal lobe epilepsy. Given the mitochondrial pathology of Adk-knockout mice it is tempting to speculate that an acute injury-induced surge of adenosine might lead to mitochondrial impairment as an additional contributing factor to epileptogenesis.

Additional biochemical pathways

Deamination of adenosine via ADA leads to the formation of inosine, which is further degraded into hypoxanthine; both compounds have moderate anticonvulsant activity (Lewin and Bleck,1985). Recent studies in a tissue culture system have shown that adenosine augmentation inhibits tissue factor induction via activation of the phosphoinositide 3-kinase/Akt signaling pathway. Those effects were maintained in cell types that did not express the A1 or the A2A receptor (Zhang et al.,2010). The authors of this study proposed an adenosine receptor independent pathway as explanation for their results, however, the underlying biochemical mechanism was not elucidated and a possible contribution of A2B or A3 receptors was not excluded.

Given the multiple activities of adenosine as an upstream regulator, controlling not only basic biochemistry but also specific receptor-mediated functions of relevance for epilepsy, adenosine is a prime candidate to modulate network activity in epilepsy in the broader sense of “homeostatic bioenergetic network regulation.” Since the epilepsy syndrome is a complex disorder of network dysfunction it is highly likely that disruption of adenosine homeostasis—linked to glial dysfunction in epilepsy—affects pathogenetic processes on multiple levels via a multitude of adenosine receptor dependent and independent effects.


Overexpression of ADK in astrocytes has been identified as a direct molecular link to neuronal hyperexcitability in epilepsy. Consequently, therapeutic adenosine augmentation might uniquely be suited to reconstruct homeostatic functions of the adenosine system. As upstream regulator—in contrast to AEDs that act on selective and neuronal downstream targets—adenosine may affect epilepsy as a syndrome on multiple different levels via adenosine receptor dependent pathways, but also through biochemical, epigenetic, and bioenergetic mechanisms. Thus, reconstruction of adenosine homeostasis might influence epilepsy in a more holistic way (Fig. 3); new data discussed here suggest disease modifying and antiepileptogenic effects of focal adenosine augmentation that go far beyond mere seizure control. Questions that need to be addressed in future work include: (i) identification of the mechanisms by which adenosine can affect network changes and disease progression in epilepsy; a focus on adenosine receptor independent effects of adenosine may yield promising results. (ii) Development of diagnostic tools to quantify ADK expression in the living brain. The development of PET or SPECT ligands directed against ADK could be used to monitor longitudinal changes in ADK expression during disease progression and could also be used as diagnostic tool to identify areas of ADK-associated neuronal hyperexcitability. (iii) Finally, clinical trials should be initiated to test safety and feasibility of focal adenosine augmentation for epilepsy. Adenosine is already FDA-approved and routinely used in clinical applications for the treatment of supraventricular tachycardia. First clinical studies could be realized by combining focal adenosine infusion with presurgical invasive diagnostics using intracranial EEG-recording electrodes coupled with cannula for the infusion of adenosine. Those studies should yield valuable safety, efficacy, and dose–response data. Given the short half-life of adenosine (∼ minutes in brain), adenosine infusion is considered to be safe.

Figure 3.

Toward a “holistic” treatment approach for epilepsy. Adenosine is can affect epilepsy and its comorbidities on different levels: Activation of A1 receptors provides potent seizure suppression and neuroprotection. In addition adenosine can modulate glial function and contribute to bioenergetic and epigenetic functions that all have antiepileptogenic potential.

Evolution might have provided us with a primordial metabolite (adenosine) that appears to be a “master switch” regulating key homeostatic functions of any given cell. Understanding those mechanisms in health and disease will eventually allow us to develop entirely new conceptual strategies to treat and cure a complex comorbid syndrome such as epilepsy in a broader sense that goes far beyond mere symptom suppression. This ambitious goal will require critical rethinking and abandonment of current dogma.