Summary

Anti-Epileptic Drugs as Prophylactics?

Because patients often experience immediate (early) seizures directly following an initial brain insult, a logical early effort to prevent the emergence of long-term seizure disorders after such insult was to use a known AED prophylactically. This assumption that early seizure prevention will preclude later seizures is fundamentally rooted in the idea that ‘seizures beget seizures,’ such that blocking early seizures should necessarily slow or stop disease progression (Gowers, 1881, Ben-Ari & Holmes, 2008). However, this notion is not without challenge (Marson et al., 2005, McIntosh & Berkovic, 2005, Sills, 2007), and it is still unclear whether the seizures themselves or some inherent predisposition toward progressive epilepsy causes further seizures. Still, despite several clinical trials using a number of the known AEDs (which themselves have considerable side-effects) to prevent late seizures following a number of brain insults, none have provided any protection against the emergence of epilepsy (Nissinen et al., 2004; reviewed inTemkin, 2001, Dudek et al., 2008).

To make this distinction clearer, allow us to liken this situation to another well-known clinical paradigm; imagine trying to develop only antibiotics as prophylaxis against infection. While most antibiotics are designed for effectively treating pre-existing infections (i.e. symptomatic treatment), they can be rather ineffective, and indeed can on occasion cause more harm than good as prophylactics (although the prophylactic use of certain antibiotics can also be clinically warranted and extremely effective). The development of vaccines to protect against various infections requires a dramatically different approach than simply changing the timing of antibiotic administration. Vaccines are designed to incite an early inflammatory process to prevent a later infection from taking hold, while antibiotics are designed to target elements of an already infected system. As such, antibiotics work on entirely different substrates than vaccines. These substrates are unique to the infected system and are not present prior to infection, or even necessarily during a latent period.

In much the same way, successful prophylaxis for acquired epilepsy will likely require employing very different paradigms than those used to suppress the seizure symptoms. And expecting that those drugs specifically designed to alleviate symptoms in an already hyperexcitable system should necessarily prevent the initial development of such a state may be just as futile as using most antibiotics as prophylactics—the substrate for the drug may simply not be present early in the progression of the disorder. But what strategies are rational approaches to antiepileptogenesis if anticonvulsants are a dead end?

Cannabinoid Pharmaceuticals

In this review, we will focus on the role of the endocannabinoid system in the epileptogenic process, as it provides an excellent example of the important distinction between symptomatic antiepileptic strategies and prophylactic antiepileptogenic strategies. Thus, we will begin by briefly highlighting some of the history and the nature of cannabinoid drugs and their endogenous analogues.

Endocannabinoids are a particular type of lipid messenger, so named because their original discovery was prompted by research into the mechanism of action of the psychoactive component of the plant, Cannabis sativa, or marijuana. The pharmacological potential of this class has been known since at least 3000 BC, when Chinese texts describe the use of marijuana for pain relief. The active ingredient in marijuana was identified for the first time as tetrahydrocannabinol (THC), a lipid molecule (Fig 1A), in 1964 (Gaoni & Mechoulam, 1964). An excellent review of the history of cannabis use can be found in (Mechoulam, 1986,Pacher et al., 2006, Murray et al., 2007) and others.

Figure 1

Cannabinoid-targeting molecules

The cannabinoid type 1 receptor (CB1) is the main type of cannabinoid receptor in the CNS (Herkenham et al., 1990, Herkenham et al., 1991). Other cannabinoid receptors and endocannabinoid molecules have also been described since that time, including the CB2 receptor, GPR55, and others (Freund et al., 2003,Pertwee, 2005, Gong et al., 2006, Pertwee, 2007, Ryberg et al., 2007). Anandamide and 2 arachidonyl glycerol (2-AG) (Fig 1B) are the main endogenous ligands for the CB1 receptor (Devane et al., 1992,Mechoulam et al., 1995, Sugiura et al., 1995), In this review, we will focus primarily on the role of 2-AG as a retrograde messenger (detailed below). These endogenous cannabinoids are known as ‘endocannabinoids.’

Two CB1 agonists are currently used clinically to treat both the nausea & vomiting associated with cancer chemotherapy agents and the cachexia (anorexia-induced weight loss) of AIDS patients (Consroe, 1998). CB1 agonists that do not induce the psychotropic effects of marijuana are being investigated clinically to treat osteoporosis, chronic pain, and anxiety, and even for tumor suppression (Consroe, 1998). One of the first CB1 antagonist drugs to be used clinically was SR141716A (SR, Fig 1C), under the trade name of Rimonabant (Rinaldi-Carmona et al., 1994, Compton et al., 1996). Its effects are generally opposite to those observed after marijuana use: locomotor stimulation (Compton et al., 1996), anxiety-like behavior (Navarro et al., 1997), appetite & addiction suppression (Arnone et al., 1997, Lange & Kruse, 2008,Rasmussen & Huskinson, 2008), hyperalgesia (Richardson et al., 1997, Rice et al., 2002), and enhanced memory (Terranova et al., 1996, Lichtman, 2000). When given to obese patients, Rimonabant causes significant weight loss and has positive effects on cardiovascular health (Despres et al., 2005, Pi-Sunyer et al., 2006, Lange & Kruse, 2008). However, its side-effect of depressed mood has precluded its approval for anti-obesity in the United States and clouded its future as a therapeutic strategy for chronic use.

Despite the evidence of the medical potential of specific cannabinoid receptor-targeting drugs, pharmaceuticals within this system have just begun to be developed for clinical use because the advanced study of lipids has only recently become feasible. Advances in technology such as those in mass spectrometry and atomic force microscopy have permitted many advances, leading to the formation of the field of brain lipidomics (Piomelli et al., 2007). It is now clear that lipids such as cannabinoids are complex, dynamic molecules with the capacity to serve information-carrying roles. In fact, the dry weight of the brain is composed of more than half lipid material (Herkenham et al., 1991, Piomelli et al., 2007). Lipid molecules have unique properties, containing multiple domains whose modification can lead to functional changes in molecular action. Lipids can target surface, cytoplasmic, or even nuclear receptors, a versatility that is not possible for most amino acids, nucleic acids, and carbohydrates. Additionally, lipids have unique pharmacokinetics and pharmacodynamics, and the extent of their capabilities as pharmaceuticals has yet to be fully explored. The ongoing development of drugs that utilize the unique properties of lipid molecules promises to yield many novel therapeutic tools.

Endocannabinoids, lipids themselves, have putative roles in a number of neurological disorders. In fact, the CB1 receptor is the most abundant G-protein coupled receptor (GPCR) in the CNS, underscoring its broad importance in numerous brain regions (Herkenham et al., 1991, Piomelli et al., 2007). Because of their unique effects on both excitatory and inhibitory signaling (described below), endocannabinoids are well-suited to mediate long-term synaptic plasticity (Freund et al., 2003, Mato et al., 2004; reviewed in Alger, 2004). In addition, cannabinoids modulate immune responses and inflammation (Klein, 2005; reviewed inFriedman et al., 1995), metabolic rate and satiety (Leker et al., 2003), and protect against excitotoxicity (Hansen et al., 2001a, Hansen et al., 2001b, Shouman et al., 2006). Thus, drugs which target the cannabinoid system may have potential far beyond what has already been realized.

A well-developed repertoire of CB1-targeting lipid molecules may allow the rapid translation of research findings into novel therapeutic uses. We will describe in this review how cannabinoid compounds have important roles in modulating both epilepsy and epileptogenesis and thus may prove to be good targets for future symptomatic or prophylactic pharmaceuticals for acquired epilepsies.

Essentials of Endocannabinoid Signaling

In order to understand the way in which endocannabinoids may be involved in epileptogenesis, we will first outline a few features of the mechanics of the system. There are several more detailed reviews of endocannabinoids’ actions (Freund et al., 2003, Piomelli, 2003, Kano et al., 2009), so we will highlight only the most salient points for our discussion of epilepsy and epileptogenesis.

The main mechanism of action of endocannabinoids in the CNS is through retrograde signaling. Briefly, depolarization of a postsynaptic cell causes the synthesis of the endocannabinoid 2-AG by the enzyme diacylglycerol lipase (DGL), which is present in postsynaptic cells such as pyramidal cells of the hippocampus. 2-AG then mediates the activation of presynaptic CB1 receptors through retrograde synaptic transmission. The mechanisms of release and transport of lipophilic molecules such as 2-AG in the aqueous synaptic cleft are the subjects of active investigation, and while several theories have been proposed (Lovinger, 2007, Lovinger, 2008, Kano et al., 2009), definitive answers have yet to be determined. Activation of presynaptic CB1 receptors then suppresses the release of neurotransmitter from specific CB1-containing axon terminals (Bodor et al., 2005, Glickfeld & Scanziani, 2006), which can be either inhibitory or excitatory. This immediate suppression of transmission is called the depolarization-induced suppression of inhibition (DSI) (see Fig 2) or excitation (DSE), respectively. Importantly, cannabinoids can be synthesized and released not only after depolarization, but also following metabotropic glutamate and muscarinic receptor activation (Pitler & Alger, 1994, Maejima et al., 2001,Varma et al., 2001, Ohno-Shosaku et al., 2003, Straiker & Mackie, 2007).

Figure 2

Basic Schematic of Depolarization-Induced Suppression of Inhibitory Transmission by Cannabinoids

Because of the ‘on-demand’ nature of endocannabinoid synthesis, DSI and DSE play a role in silencing the incoming inhibitory and excitatory synaptic inputs to very active cells. Additionally, because CB1 receptors are G-protein coupled, their activation may initiate an extensive downstream molecular cascade which can have both immediate and long-term consequences (see, for example, Chevaleyre & Castillo, 2003). Thus, endocannabinoids are unique retrograde messengers whose actions include both the provision of a rapid activity-dependent modulation of both excitatory and inhibitory neurotransmission and longer-term effects.

In addition to the phasic control of transmitter release, tonic release of endocannabinoids from postsynaptic pyramidal cells has also been demonstrated (Losonczy et al., 2003, Neu et al., 2007), revealing an ongoing role of endocannabinoids in synaptic modulation even in the absence of activity-dependent factors. Interestingly, the tonic inhibition of GABA release can be modulated in an activity-dependent manner (Foldy et al., 2006), indicating that CB1-dependent tonic control of release probability may play even more significant functional roles than the short term plasticity effects reflected in DSI and DSE.

Cannabinoids and Epilepsy

Cannabinoid agonists are thought to hold potential as novel neuroprotectants and anticonvulsants (Shohami et al., 1993, Panikashvili et al., 2001, Mechoulam et al., 2002, Mechoulam & Lichtman, 2003, Wallace et al., 2003, Deshpande et al., 2007a). The details of the neuroprotective effects of cannabinoids are beyond the scope of this discussion, but may be mediated through a number of cannabinoid actions including activity-dependent control of neurotransmission leading to decreased excitotoxicity in the affected area, decreased inflammation due to effects of cannabinoids on leukocytes, or hypothermia induced by the cannabinoid, all of which are neuroprotective (Leker et al., 2003, Fernandez-Ruiz et al., 2007). As we saw in the last section, the mechanics of the endocannabinoid system strongly suggest that in addition to its role in other neurological processes, it should have a direct modulatory effect on seizures and neuronal excitability. Both agonists and antagonists have in fact been studied for their modulatory properties on neuronal excitability. We will begin by going over clinical evidence, and then will outline some experimental evidence of the acute effects of cannabinoid-targeting drugs on seizures.

Epidemiological studies of human patients have shown that, unlike other drugs of abuse, chronic marijuana use is actually protective against first onset seizures (Ng et al., 1990), suggesting the clinical anti-epileptic potential of agonist drugs. However, the clinical efficacy of cannabinoid drugs as AEDs for existing human epilepsy syndromes has yet to be conclusively shown. Several case reports and even a few small-scale drug trials have been performed in humans using cannabis as an anticonvulsant (Keeler & Reifler, 1967,Consroe et al., 1975, Ellison et al., 1990); however, no large-scale clinical trial has been performed, largely due to concerns over the psychoactive effects of marijuana. Additionally, other studies have revealed no effect (Gordon & Devinsky, 2001) or even increased seizure susceptibility (Keeler & Reifler, 1967) with marijuana use. And while THC seems to exhibit some anticonvulsant properties, toxicity or induced withdrawal effects can be characterized by CNS excitation, ataxia, or even convulsions (Gordon & Devinsky, 2001, Lutz, 2004), further complicating the clinical picture of the therapeutic utility of this drug class against epilepsy.

In animal models, cannabinoid agonists are acutely effective at stopping generalized tonic-clonic and both simple and complex partial seizures, though not generalized absence seizures (Consroe, 1998). Exogenously applied endocannabinoids also robustly inhibit electroshock-induced seizures, and reduce seizure frequency better than current clinically used AEDs in chemically kindled animals (Fig 3A) (Wallace et al., 2003, Lutz & Monory, 2008). These beneficial effects can be blocked by the CB1 antagonist, SR141716A, suggesting that cannabinoids’ acute anticonvulsant properties are mediated through CB1 activation.

Figure 3

Acute effects of cannabinoids on seizures

Although CB1 receptors are indeed more abundant at presynaptic inhibitory than excitatory terminals (Katona et al., 1999, Hajos et al., 2000, Katona et al., 2000), CB1 agonists’ anticonvulsant properties indicate that silencing of incoming excitation, rather than inhibition, has the dominant effect on acute seizures. To further examine this, Monory et al (2006) used the selective deletion of CB1 receptors from either excitatory or inhibitory terminals to elucidate the differential role of the receptor at these locations. CB1 receptor deletion from excitatory terminals—a phenomenon which is also seen in human epileptic brain tissue (Ludanyi et al., 2008)—increased seizure susceptibility to kainate, while deletion from inhibitory terminals did not have a significant effect on seizure susceptibility to kainate (Monory et al., 2006). While the precise mechanism of the latter finding is not yet understood, these experiments indicate that the CB1 receptors on excitatory terminals are those which may be primarily responsible for the acute anticonvulsant properties of cannabinoid agonists.

Predictably, considering the results of experiments with CB1 agonists, CB1 receptor antagonists increase seizure frequency and decrease seizure threshold during hyperexcitable states (such as in chemically or electrically kindled animals, see Fig 3C), but not in control animals (Wallace et al., 2002, Marsicano et al., 2003, Bernard et al., 2005, Chen et al., 2007, Deshpande et al., 2007b). The acute effects of CB1-targeting drugs on seizures are therefore largely explained by their direct effects at presynaptic excitatory terminals.

The endogenous production of endocannabinoids such as 2-AG and anandamide is also increased in the CNS following injury and seizures (Fig 3B) (Hansen et al., 2001b, Panikashvili et al., 2001, Marsicano et al., 2003, Wallace et al., 2003), perhaps serving to directly quell seizure activity, to aid in neuroprotection, or to play another as yet undetermined role. However, this rise in endocannabinoid levels following seizures indicates that CB1 receptors on both excitatory and inhibitory terminals are being broadly activated after brain insult and urges us to consider the long-term consequences of this CB1 activation—and its antagonism—following initial brain insult on epileptogenic processes.

Long-term effects of cannabinoids on epilepsy and epileptogenesis

As might be expected based on the G-protein coupled nature of CB1 receptors, the acute modulation of seizure activity we have discussed is not the only role that endocannabinoids play in epilepsy. Having explored the acute role of cannabinoids in seizures, which we have seen makes them interesting as potential future symptomatic treatments (Katona & Freund, 2008), we will discuss the long-term modulatory effects of cannabinoid agonists and antagonists on the epileptogenic process (Lutz & Monory, 2008).

Unlike the acute effects we discussed above, long-term effects of CB1 receptor activation seem to be more evident at inhibitory than excitatory terminals. Long-term CB1 receptor expression changes have been documented following seizures of many etiologies. After seizures in animal models, CB1 receptor expression is increased overall (Chen et al., 2003, Wallace et al., 2003, Schuchmann et al., 2006, Chen et al., 2007), but is particularly increased at inhibitory terminals after experimental prolonged febrile seizures (Chen et al., 2003). In human epileptic tissue, there is an overall decrease in CB1 expression when compared to normal tissue (Ludanyi et al., 2008). However, the same study showed that CB1 receptor expression at inhibitory terminals remains relatively constant, indicating that this decrease in expression is primarily at excitatory terminals. Thus despite different overall CB1 receptor levels, in both animal models and in human epileptic tissue there is a selective enhancement of the ratio of CB1 receptor expression on inhibitory versus excitatory terminals. The net effect of this change in ratio would be expected to makeinhibitory synapses more prone than usual to silencing by endocannabinoids through DSI (as demonstrated in Chen et al., 2003) and excitatory synapses less prone than usual to silencing through DSE. When spread across an entire network under pathological circumstances, this change could actually contribute to the underlying hyperexcitability seen in epilepsy. However, there is probably much more to this story, as for example, new research has indicated that after seizures, mossy fibers—the axons of hippocampal dentate gyrus granule cells that form recurrent excitatory connections uniquely in epilepsy—actually begin to express CB1 receptors, a property which normal mossy fibers do not possess (Bhaskaran & Smith, 2007). This is contrary to the relatively decreased CB1 receptor expression on excitatory terminals normally expressing CB1 receptors. Additionally, in the pilocarpine model of epilepsy, CB1 receptors are dramatically redistributed to dendritic regions (Falenski et al., 2007). Thus, the overall changes in CB1 receptor expression likely have complexities relating to specific cell types and brain areas which remain to be explored.

Not only is CB1 receptor expression altered following seizures, but DSI is also enhanced in animals that have experienced prolonged febrile seizures compared to littermates not exposed to hyperthermia (Chen et al., 2003). Additionally, tetanic activity in acute slices from control animals leads to an enhancement of DSI, while exposure to prolonged febrile seizures in vivo one week prior to electrophysiology precludes the further enhancement of DSI with tetanic stimulation, suggesting that tetanic stimulation and prolonged febrile seizures may enhance DSI through a common mechanism. Interestingly, we found that the induction of both long-term increases in CB1 receptor expression and enhanced DSI in this model (Chen et al., 2003) seem to be mediated primarily by the activation of CB1 receptors themselves (Chen et al., 2007). Specifically, increases in both DSI and CB1 receptor expression following prolonged febrile seizures could be blocked by the in vivo application of the CB1 antagonist SR141716A (SR) at the time of the febrile seizures (Chen et al., 2007). Thus, we hypothesized (Chen et al., 2007) that long-term hyperexcitability in vivo might also be prevented by early CB1 antagonism during prolonged febrile seizures. Indeed, animals given a single application of SR and then subjected to hyperthermia still had early seizures (as expected, based on the aforementioned pro-convulsant effects of the CB1 antagonist SR in seizure-prone animals), but without further intervention were as resistant to kainate challenge six weeks after experimental febrile seizures as littermates not exposed to hyperthermia (Fig 4). In contrast, animals that had experienced prolonged febrile seizures without being given SR showed significantly increased kainate sensitivity at this six week timepoint (Fig 4). This single dose of SR could be given either before or 2 minutes after the initiation of febrile seizures with the same outcome on long-term seizure susceptibility. Thus, not only did CB1 blockade, either before or during the time of brain insult preclude the upregulation of CB1 and enhancement of DSI, but it also decreased the animals’ long-term seizure susceptibility (Fig 4) (Chen et al., 2007). This suggests that while using a CB1 agonist—an acute anticonvulsant—at the time of initial brain insult may stop seizures, the activation of CB1 receptors at this time has the potential to contribute to later changes in CB1 receptor expression and DSI which may in turn actually contribute to later overall hyperexcitability in vivo.

Figure 4

SR141716A (SR), an acute proconvulsant in already hyperexcitable animals, blocks the emergence of increased seizure susceptibility in the hyperthermia-induced prolonged febrile seizure model of epileptogenesis

In order to determine whether SR treatment might affect common underlying mechanisms of hyperexcitability in more than one animal model, we employed a similar strategy, using a single dose of SR following lateral fluid percussion injury (FPI), a model of TBI, to prevent long-term seizure susceptibility changes (Echegoyen et al., 2009). This treatment was effective at preventing late increases in seizure susceptibility in this model as well.

Experiments done in these two models have allowed some insight into the time window within which single SR application is effective at preventing long-term seizure susceptibility changes. As mentioned above, in the prolonged febrile seizure model, application of SR before hyperthermia was effective in this regard. However, as we will discuss further later, an ideal agent would be administered after a problem has been detected, rather than in an anticipatory fashion. Thus, it is significant that in both the traumatic brain injury and prolonged febrile seizure models, SR application within 2 minutes following brain insult prevented long-term seizure susceptibility changes. However, in both the FPI and the prolonged febrile seizure models, SR application 20 minutes after head injury or 30 minutes after the start of the induction of febrile seizures failed to have any long-term beneficial effect on seizure susceptibility (Echegoyen et al., 2009; K. Chen & I. Soltesz, unpublished observations). These experiments have begun to outline the brief time window during which early epileptogenic processes might begin, and therefore how quickly prophylactic interventions may need to be administered (Santhakumar et al., 2001, Chen et al., 2007,Echegoyen et al., 2009).

Another interesting result of the study using SR in the FPI model was that the positive effect of SR on long-term hyperexcitability was blocked with concurrent administration of pentobarbital to block early post-traumatic seizures (Echegoyen et al., 2009). It is still unclear precisely what mediates this interaction. Whether the underlying mechanism has to do with blocking the activity-dependent release of endocannabinoids, some target-specific effect of SR or pentobarbital, or even the pro-convulsant nature of SR remains to be resolved with further experiments. Nevertheless, it is important to note for future studies of this drug class that AEDs themselves may interfere with potential antiepileptogenic agents.

The ability of SR to prevent hyperexcitability in these two very different seizure models demonstrates that the underlying molecular mechanisms leading to long-term hyperexcitability in both models share sensitivity to a single agent. This implies that there are indeed unifying themes in epileptogenesis between models which can be targeted for future therapeutics. It is important to note that the previously-described studies used convulsant challenge to assess seizure threshold as a major mechanistic component and indicator of post-traumatic epilepsy. As we will discuss below in “Challenges to the Development and Clinical Application of Anti-Epileptogenic drugs,” until animal models can be developed which are better suited for studying drug effects on spontaneous seizures, or until a reliable surrogate marker for epileptogenesis is found, hyperexcitability has been and will continue to be the most experimentally feasible end point for identifying potential antiepileptogenic agents.

Preclinical Implications

In the previous section, we saw that the long-term effects of CB1 receptor activation and antagonism are very different than the acute effects we had discussed prior. Indeed, a CB1 antagonist, despite being an acute pro-convulsant (Fig 3C) (Marsicano et al., 2003, Wallace et al., 2003, Bernard et al., 2005, Chen et al., 2007, Deshpande et al., 2007b), prevented long-term hyperexcitability following two different types of brain insults. This result was highly sensitive to both temporal factors and the concomitant administration of pentobarbital, an AED.

What, then, are the clinical implications of these findings? First, we have seen that different cannabinoid targeting drugs have the potential to subserve either symptomatic or prophylactic strategies for treating acquired epilepsies, but that the necessary type of drug depends critically on the timing of administration. While a cannabinoid antagonist may be acutely pro-convulsant, it could have significant value as a prophylactic against epilepsy within an appropriate time window. On the other hand, a cannabinoid agonist may be acutely effective at seizure suppression, but have no effect on long-term epileptogenesis. To revisit our previous analogy, a CB1 agonist, then, is more comparable to the antibiotic—potentially useful in managing the symptoms of an existing infection, or in this case, pre-existing epilepsy—whereas a CB1 antagonist, useless or even deleterious against existing epilepsy, can be likened to the vaccine—when given within a critical period it can participate in a sequence of events which prevents the emergence of late hyperexcitability.

This somewhat counterintuitive concept may prove to be a sound conceptual base upon which to develop further attempts at pharmaceutical prophylaxis for acquired epilepsies. Because CB1 antagonists are known pro-convulsants in animal models of hyperexcitability (Wallace et al., 2002, Marsicano et al., 2003,Bernard et al., 2005, Chen et al., 2007), the fact that their application immediately after brain insult in more than one clinically relevant animal model prevented the development of hyperexcitability (Chen et al., 2007, Echegoyen et al., 2009) implies the possibility that the pro-convulsant effects of this drug may contribute to its anti-epileptogenicity. This is not a conceptually new idea in medicine—the necessarily pro-inflammatory nature of vaccines in preventing later infection, for example, has been well-understood for some time. Certainly, the fact that interfering with seizure activity using pentobarbital blocks the beneficial effect of SR supports an essential role of the pro-convulsant effect (Chen et al., 2007, Echegoyen et al., 2009). However, further study of SR in both genetically normal and transgenic animals will be required to determine whether its beneficial properties arise from its mild proconvulsant effects, or its direct or secondary effects. The idea of using a proconvulsant as an antiepileptogenic agent has also been visited by the Pitkänen group, who found that atipamezole, an α2-adrenoceptor antagonist which is also a proconvulsant, has beneficial effects against epileptogenesis after status-epilepticus induction (Pitkänen et al., 2004). In this study, atipamezole, a mild proconvulsant, was administered continuously for 9 weeks and beneficial effects on seizure frequency & severity were observed, along with decreased markers of epileptogenesis such as hilar cell loss and mossy fiber sprouting. However, the drug did not prevent the development of epilepsy (Pitkänen et al., 2004). More studies will be required to determine what properties of these agents—their proconvulsant nature or their mechanisms of action—are indeed the ones which mediate their antiepileptogenic properties.

Such an anti-epileptogenic treatment could eventually be made available in a variety of situations where risk factors for acquired epilepsy are likely to be identified—in hospitals, at sporting events, and in combat situations, for example—and a single treatment close to the time of insult might then drastically improve the outlook for these patients, for whom there is currently nothing to do but to wait and hope that epilepsy does not emerge months to years later. Still, there are several unanswered issues which must be addressed before we consider any agent a true antiepileptogenic drug.

Challenges to the Development and Clinical Application of Anti-Epileptogenic drugs

Any truly clinically useful antiepileptogenic strategy will necessarily address several criteria. It should be practically useful, clinically testable, and also implementable. In order to be practically useful, for example, the agent should be effective even when given only after a risk factor has been identified. Just as we don’t routinely vaccinate individuals against every disease, but rather only against those diseases to which an individual is likely to be exposed, with few exceptions, an agent that can only be given prior to a brain insult is of little clinical utility. Further, the simplicity of a prophylactic strategy is essential for its implementation—if the prevention of a disease is as invasive, dangerous, or arduous as dealing with the disease itself, or has more serious adverse effects, then the treatment is contraindicated, as prophylaxis is necessarily provided to not only individuals who would develop the disease without it but also to those who would never develop the disease at all.

Because no antiepileptogenic agent currently exists, the major criterion which must be met first is that of testability. From a research perspective, the single major question regarding the current evidence for any antiepileptogenic strategy involves evaluating the realistic nature and predictive utility of our current animal models. Finding models which are both amenable to addressing the types of clinically-oriented criteria described above and which also produce enough seizures to statistically identify promising treatment effects is a difficult endeavor.

Some models, such as the chemical and electrical kindling models, produce robust spontaneous seizure phenotypes, but have no comparable etiology in human acquired epilepsy. Additionally, there are problems of interpretation in utilizing such models to evaluate potential prophylactic treatments. Electrical kindling models generally require multiple sequential rounds of induction, making it difficult to identify the precise time at which epileptogenic events occur and thus difficult to define the window(s) in which to test potential antiepileptogenic agents, particularly considering that an ideal agent would be given after the etiologic event has occurred. Models that involve injecting proconvulsants are also difficult to use for testing antiepileptogenic agents because 1) there is a delicate balance between inciting epileptogenesis and an unacceptably high mortality rate, which may be disrupted by an agent that has direct effects on acute seizures and 2) the interactions of the injected agents, both direct and through physiological effects, must be taken into account. If the agent stops the induced seizures, can later beneficial effects be considered antiepileptogenic, or could they simply indicate a failed induction of epilepsy due to disruption of the etiologic event? If the agent is proconvulsant, how does one normalize the degree to which animals experience the etiologic event itself?

Models with more clinically relevant etiologies include, among others, those which mimic traumatic brain injuries or prolonged febrile seizures mentioned above (Baram et al., 1997, Chen et al., 1999, Dube et al., 2000, Santhakumar et al., 2001, D’Ambrosio et al., 2004, Dube et al., 2006, Kharatishvili et al., 2006,Schuchmann et al., 2006, Pitkänen et al., 2007). In these models, the etiologies do more closely mimic those seen in human disease, but decreased seizure threshold or cellular hyperexcitability are typically taken as measures of epileptogenesis for the purposes of drug testing, simply because the incidence of true epilepsy—late unprovoked seizures—and the frequency of the resulting seizures are rather low (Dube et al., 2000, Kharatishvili et al., 2006, Chen et al., 2007, Howard et al., 2007, Kharatishvili et al., 2007,Pitkänen et al., 2007; but see D’Ambrosio et al., 2004).

In fact, using data from the clinically relevant lateral fluid percussion injury (FPI) model of post-traumatic epilepsy as a template (Kharatishvili et al., 2006), a power analysis of the number of animals and monitoring time per group needed to reveal a statistically significant drug effect (given that the drug produced a 70% reduction in the frequency of late seizures) in animal models of spontaneous seizures suggests that at a minimum, over 6,000 days worth of video EEG data would need to be analyzed (seeSupplemental Material). And despite this amount of effort, such a study could still overlook a less robust, but still very clinically desirable drug effect, not to mention the increase in time and numbers that would be required to compare more than two groups to look at interactions between or differences among pharmaceutical agents.

Most acquired epilepsy models with a tested history and a strongly clinical etiological basis have rates of spontaneous epilepsy roughly similar to the lateral FPI model (but see (D’Ambrosio et al., 2004), underscoring the major problem in studying true epilepsy in clinically relevant animal models: these types of studies are very difficult to perform without using a biomarker for true epilepsy. However, a good biomarker for epilepsy currently does not exist, so that the best measure of an antiepileptogenic agent in an animal—its effect on true epilepsy—is something which is not currently logistically feasible for most laboratories to examine. Thus, most laboratories studying interventions must either resort to using chemical or electrical kindling models of epileptogenesis, in which nearly all animals develop spontaneous seizures, or try to use various biomarkers which have yet to be widely accepted as substitutes for spontaneous recurrent seizures. Clearly, a solution to this problem is desperately needed, since without an extremely convincing basis in animals, a potential anti-epileptogenic drug could hardly expect to be considered for clinical trials in humans.

Certainly, drug trials for prophylactic treatments with anti-epileptogenic agents in humans will require much larger and lengthier studies than those for symptomatic treatments with AEDs, as the incidence of epilepsy in at-risk populations is fairly low with very high variability, and the latent period prior to the emergence of seizures can be years long. Thus, before beginning such a study, we will also need to ensure that the testing of potential prophylactic treatments are implemented in a way that is compatible with the clinical management of the particular brain insult, including AEDs, which as in the example above may actually interfere with the beneficial effects of drugs such as SR on the development of hyperexcitability. These kinds of logistical issues (also see Dichter, 2009) remain major concerns which loom over the possibility of developing prophylaxis for acquired epilepsies.

Conclusions and Outlook

In summary, we have outlined the important problem of the current lack of pharmacological prophylaxis for acquired epilepsies, the history of cannabinoids as modulators of neural activity, and we have illustrated, using the endocannabinoid system, how a CB1 antagonist—a known proconvulsant under hyperexcitable conditions—may have potential as a prophylactic treatment for acquired epilepsy using more than one clinically relevant animal model. Finally, we described some of the major issues facing the translation of antiepileptogenic strategies to clinical practice. Prophylactic treatment as an alternative to symptomatic treatment of acquired epilepsy is one of the ‘holy grails’ of epilepsy research, and this discussion highlights several essential avenues of future research in this area to allow the development of rational drugs for this purpose:

1. Antiepileptogenic agents may not be anticonvulsant (i.e. AEDs) and may even have acute pro-convulsant properties (Pitkänen et al., 2004, Chen et al., 2007, Echegoyen et al., 2009). Future studies will be needed to see whether this pro-convulsant property is necessary for antiepileptogenesis.
2. Traditional AEDs, when used concomitantly, may actually interfere with the actions of potential antiepileptogenic agents (e.g. the interactions of pentobarbital and SR in Echegoyen et al. (2009)). This will be particularly important in considering the translation of research findings into a clinical setting.
3. Focusing on identifying which molecular and cellular changes are shared between different etiologies of acquired epilepsies may help identify good strategies for developing broadly useful therapeutics.
4. The time window during which potential antiepileptogenic agents are able to act is critically important and may be much smaller—on the order of minutes rather than hours or days—than has been investigated in clinical studies (Temkin, 2001).
5. Developing more robust but still clinically relevant animal models of acquired epilepsies in which spontaneous seizures arise quickly and at a frequency amenable to high-throughput screening of potential anti-epileptogenic drugs will be necessary in order to undertake a true search for pharmacologic prophylaxis of acquired epilepsy.

The development of an ideal antiepileptogenic agent has until recently seemed an elusive goal, despite the idyllic possibility of prescribing prevention rather than chasing seizure control. However, with recent advances in the field of lipidomics, the myriad new research studies outlining the molecular mechanisms of epileptogenesis, including endocannabinoid signaling, and increasing understanding of the effects of various endogenous and exogenous molecules on this process, we may indeed be closer than ever to reaching this goal.

Supplementary Material

Acknowledgments

References