In search of models of pharmacoresistant epilepsy


Address correspondence to Uwe Heinemann, Institute of Neurophysiology and NeuroCure Research Centre, Charité Universitätsmedizin Berlin, Humboldt and Freie Universität Berlin, Tucholskystr. 2, D 10117 Berlin, Germany. E-mail:


Clinically available anticonvulsant drugs fail to control seizures in approximately 30% of epileptic patients. If hippocampal sclerosis is combined with focal dysplasia or similar developmental alterations, the likelihood of incomplete seizure control may reach >90%. Because only a minority of epilepsy patients benefit from epilepsy surgery, we need more research into the mechanisms of drug refractoriness. In this review we analyze different approaches to study pharmacoresistance and underlying mechanisms using in vitro models. Epileptiform discharges after prolonged application of low Mg2+ artificial cerebrospinal fluid (ACSF) or 4-aminopyridine (4-AP), or combined application of these convulsants with bicuculline in acute hippocampal–entorhinal cortex slices reveal pharmacoresistance and point to loss of γ-aminobutyric acid (GABA)ergic function, in part due to reduced delivery of GABA from presynaptic terminals. Interestingly, epileptiform activity in immature tissue (organotypic hippocampal slice cultures and acute intact hippocampus) is immediately resistant to available antiepileptic drugs, and preliminary evidence points to a role of alterations in Cl homeostasis. Seizure-like events can also be induced in dissected tissues from human epileptic patients. Future research on human tissue may provide useful information for understanding the mechanisms underlying pharmacoresistance.

Clinically available anticonvulsant drugs fail to control seizures in approximately 30% of epileptic patients. Drug refractoriness is particularly frequent with focal epilepsies. These concern epilepsies due to abnormalities in brain development and conditions in which abnormal connectivity has developed, as is the case with hippocampal sclerosis. If hippocampal sclerosis is combined with focal dysplasia or similar developmental alterations, the likelihood of incomplete seizure control may reach >90% (Schmidt & Löscher, 2005). This is not to say that the antiepileptic drugs have no effect at all, but for maintaining professional activities and permission to drive vehicles, complete seizure control is usually required. Because only a minority of patients with epilepsy benefit from epilepsy surgery we need more research into the mechanisms of drug refractoriness.

A first idea that comes to mind is that regional variations in circuitry and cellular properties not only determine seizure susceptibility but also determine drug refractoriness. Based on this idea, a number of acute seizure models have been studied. In entorhinal–hippocampal slices prepared from adult animals, tonic–clonic seizure-like events (SLEs) can be induced by elevation of K+, lowering of Ca2+, or lowering of Mg2+ in artificial cerebrospinal fluid (ACSF) (for review see Heinemann et al., 1994). In addition, interference with ion channels can induce such kind of activities. Therefore, application of 4-aminopyridine (4-AP)—which is well known to interfere with different types of K+ channels that include D type and A type K+ currents and a subportion of delayed rectifier currents—induces SLEs (Fueta & Avoli, 1992). SLEs induced in acute slices are characterized by a negative potential shift superimposed by high-frequency field potential transients (tonic period), followed by a period in which clonic-like afterdischarges occur. It has generally been found that seizure-like activity induced in temporal neocortex, hippocampus, and entorhinal cortex had similar sensitivity to antiepileptic drugs. Therefore low Ca2+ induced SLEs or high K+ induced SLEs in the hippocampus responded well to classical anticonvulsants, which act either by augmenting γ-aminobutyric acid (GABA)ergic synaptic transmission such as phenobarbital, valproic acid, benzodiazepines, or which dampen neuronal excitability such as phenytoin, carbamazepine, or valproic acid. In addition, some agents influence glutamatergic synaptic transmission (felbamate, lamotrigine). These studies suggested that pharmacoresistance is not intrinsic to some regions of the brain.

In vivo Models of Pharmacoresistant Epilepsy

Although acute SLEs in hippocampus and other temporal lobe structures respond well to antiepileptic drugs, seizures occurring in animals in which epilepsy had been induced are often pharmacoresistant. For example, different types of spontaneous seizures in epileptic dogs, spontaneous seizures in amygdala-kindled rats, and post status epileptic seizures after prolonged electrical stimulation of the basolateral amygdala have all been shown to be pharmacoresistant in subgroups of the experimental animals (Brandt et al., 2004). The idea that this is due to genetic variance to date has not been confirmed, since breeding paradigms did not result in a strain that developed pharmacoresistance in a predictable manner. This does not exclude genetic influence in general, however.

Status Epilepticus and Pharmacoresistance

In the kindling and pilocarpine model of epilepsy, frequent seizures or status epilepticus has preceded the appearance of spontaneous seizures. This raises the possibility that pharmacoresistance is acquired. Indeed, status epilepticus with complex partial seizures can relatively readily develop into drug refractory status epilepticus, which no longer is controlled by the usual medication for status epilepticus such as benzodiazepines, phenytoin, or barbiturates (Holtkamp et al., 2005). We found it worthwhile to test this aspect by using complex entorhinal cortex hippocampal slices to study pharmacologic properties of recurrent seizures as a model of status epilepticus. Low-Mg2+ treatment led relatively rapidly to transition of seizure-like events into late recurrent discharges, which were resistant to clinically employed antiepileptic drugs such as phenytoin, carbamazepine, phenobarbital, and midazolam (Dreier et al., 1998). When low-Mg2+ treatment was combined with application of the GABA receptor antagonist bicuculline, the emerging epileptiform activity was immediately not only resistant to valproic acid and to phenobarbital but also to phenytoin or carbamazepine, which exert their actions by prolonging recovery of Na+ currents from inactivation. Furthermore, prolonged exposure to 4-AP or combining 4-AP with bicuculline resulted in pharmacoresistant epileptiform activity (Brückner & Heinemann, 2000).

The mechanisms underlying these effects may be related to loss of inhibitory function. The mechanisms leading to loss of GABAergic function are presently not understood. Prolonged status epilepticus is associated with increased production of reactive oxygen species (ROS), including nitric oxide and derivatives. This might affect mitochondrial function and the capability of neurons to adapt their metabolism to neuronal function (Kann & Kovács, 2007 for review). As a result GABA may be consumed for energy production in the GABA shunt of the tricarboxylic acid cycle. Alternatively, altered properties of astrocytes may lead to reduced supply of GABA when the glutamine-glutamate-glutamine shuffle between astrocytes and neurons is affected, with glutamine being a precursor required for GABA synthesis from glutamine and glutamate (Eid et al., 2008; Zahn et al., 2008). In line with this finding is the observation that GABA and GABAA receptor agonists block late recurrent discharges in these models of status epilepticus. Increased production of ROS may also modify properties of ion channels and thereby interfere with pharmacosensitivity. Interestingly, production of free radicals accumulates during status epilepticus, whereas functionality of mitochondria decreases. Functional alterations of mitochondrial properties are also seen in chronic epileptic tissue and may contribute, therefore, to pharmacoresistance (Kann & Kovács, 2007).

Developmental Aspects of Pharmacoresistance

Quilichini has reported that SLEs induced by low Mg2+ in the acutely prepared intact hippocampus (Quilichini et al., 2003) are pharmacoresistant. However, this preparation is possible only until postnatal day 9 in rats. The finding that drug refractoriness is associated with induction of seizures in tissue similar to nervous tissue from preterm or new born babies may also indicate that one property resulting in pharmacoresistance is the general developmental state of the tissue. Indeed, preliminary data from our lab suggest that in acute hippocampal–entorhinal cortex slices prepared from postnatal 3–5-day-old rats, seizure-like events can be readily induced in the hippocampus, but that these events are pharmacoresistant to clinically employed drug concentrations of phenytoin, carbamazepine, valproic acid, or phenobarbital. In this age group these agents may even prolong seizure duration. This might relate to incomplete myelination, open blood–brain barrier (BBB), incomplete development of astrocytic properties, and delayed expression of ion channels and receptors for neurotransmitters and neuromodulators, but also to immaturity of Cl homeostasis, which matures around postnatal day 9.

Organotypic Hippocampal Slice Cultures as a Model of Pharmacoresistance

The susceptibility to induction of SLEs may be altered by lesions within the central nervous system. Many abnormalities found in human tissue with hippocampal sclerosis are also found in tissue that has been explanted early after birth in rodents. Therefore, in organotypic hippocampal cultures, most cell types and their connections are preserved, but some reorganization also occurs. Consequently, projections from area CA1 to the dentate gyrus and back-projections from area CA1 to CA3 have been described. Moreover, mossy fibers from granule cells may display axonal sprouting and recurrent connectivity (Gutierrez & Heinemann, 1999) comparable to that observed in patients (Franck et al., 1995) and in animal models of temporal lobe epilepsy (Mello et al., 1993). In general, slice cultures respond readily to convulsants with recurrent seizures (Figs 1 and 2) (Gutierrez et al., 1999; Albus et al., 2008; Wahab et al., 2009). We, therefore, began to use such preparations for the analysis of drug refractoriness. To this end we prepared slice cultures from 2–12-day-old rats and exposed them to either low Mg2+ or 4-AP at different times after slice culture preparation. More than 90% of slice cultures showed robust SLEs when treated with either low Mg2+ or 4-AP. These activity patterns could be elicited as late as 8 weeks after preparation (Albus et al., 2008). However, there is some variance in evoked activity. In 52% of slices, recurrent SLEs can be induced (Fig. 1A and B). In a subportion of 25% the SLEs transit immediately into late recurrent discharges (Fig. 2B), and in 23% we found both SLEs and late recurrent short discharges (Fig. 2A) (Albus et al., 2008).

Figure 1.

Antiepileptic drugs failed to block seizure-like events in organotypic hippocampal slice cultures. (A) (a–c) Diazepam failed to suppress seizure-like events induced by low Mg2+, even at a very high concentration. In this sample recording, the slice culture was explanted from 6-day-old rats and kept for 17 days in the incubator before experiment. (B) (a–c) Phenobarbital (100 μm) failed to suppress seizure-like events induced by low Mg2+. In this sample recording, the slice culture was explanted from a 6-day-old rat and kept for 8 days in the incubator before experiment. [K+]o (extracellular potassium concentration) and FPs (field potentials) were recorded in the pyramidal cell layer of CA3.

Figure 2.

GABA mimetic substances suppress seizure-like events in organotypic slice cultures. (A) (a–c) Muscimol, a GABAA-receptor agonist, at a concentration of 10 μm reversibly blocked seizure-like events induced by low Mg2+ in organotypic hippocampal slice culture. In this sample recording, the slice culture was explanted from an 8-day-old rat and kept for 26 days in incubator before experiment. (B) (a–c) Nipecotic acid, a GABA-uptake inhibitor, at 1 mm concentration reversibly blocked late recurrent short discharges induced by low Mg2+. In this sample recording, the slice culture was explanted from an 8-day-old rat and kept for 5 days in incubator before experiment. [K+]o (extracellular potassium concentration) and FPs (field potentials) were recorded in the pyramidal cell layer of CA3.

When testing for effects of anticonvulsant drugs one can follow two different regimens. The slice cultures can be treated long term or acutely with anticonvulsant drugs, and then SLEs can be induced by application of convulsants. Alternatively SLEs can be induced first and then drug response is monitored (Fig. 1). With both protocols we noted that SLEs could not be prevented by phenytoin, carbamazepine, valproic acid, phenobarbital, diazepam, and clonazepam. This was independent of the testing protocols. It also applied to drugs when applied in concentration clearly above the maximal doses tolerated by patients (Fig. 1). However, SLEs were suppressed in 50% of slice cultures when an anesthetic concentration of phenobarbital was used (Albus et al., 2008). We again addressed the issue of whether loss of GABAergic function was involved and tested for the effects of the GABAA agonists muscimol and isoguvacine, the GABAB agonist baclofen, the GABA-uptake blocker nipecotic acid, and the neurosteroid alfaxalone on SLEs in slice cultures at about 1 week after preparation using 6–12-day-old rats (Wahab et al., 2009). Low Mg2+–induced SLEs were dose dependently suppressed by the GABAA receptor agonists muscimol (Fig. 2A), isoguvacine, and alfaxalone and by the GABA uptake inhibitor nipecotic acid (Fig. 2B). The GABAB receptor agonist baclofen attenuated but did not suppress SLEs. Therefore, our findings do not necessarily point to a loss of inhibitory function due to major alterations in GABA receptors. The loss of effect by barbiturates, which augment GABA dependent Cl channel opening, or of benzodiazepines, which increase affinity for GABA at GABAA receptors, suggests that the loss of efficacy of these drugs is due to reduced availability of GABA. This reduced GABA availability may be due to increased consumption of GABA in the GABA shunt of the tricarboxylic acid cycle, to reduced GABA uptake into presynaptic terminals, to changes in efficacy of vesicular transporters, or to reduced delivery of the GABA precursor glutamine from astrocytes, which is converted into glutamate and subsequently into GABA in GABAergic neurons.

Human Tissue as a Tool to Investigate Pharmacoresistance

Because no reliable and easily usable in vivo models for pharmacoresistance exist, we also were interested in studies on human tissue. Initially the question was whether one could induce SLEs in such tissue. We used a number of protocols such as lowering Mg2+ or Ca2+ and application of bicuculline or 4-AP, which works in normal tissue but which failed in chronic human tissue (Gabriel et al., 2004). This is of interest as it might suggest that anticonvulsant mechanisms are upregulated during the development of epilepsy as well. However, we found that with elevation of extracellular potassium concentration, SLEs can be relatively reliably induced in human tissue including the dentate gyrus, the subiculum, and in temporal neocortex where elevated K+ had to be combined with the GABAA receptor blocker bicuculline. We, therefore, studied initially the effects of carbamazepine at therapeutically relevant concentrations on epileptiform activity electrophysiologically recorded in acute hippocampal slices of patients with mesial temporal lobe epilepsy (MTLE) or extrahippocampal tumors (Jandová et al., 2006). Quantitative analysis of data revealed that epileptiform activity induced by elevation of K+ in tissue of tumor was regularly suppressed by carbamazepine, indicating that the “epilepsy model” used is carbamazepine sensitive. In contrast, epileptiform activity in tissue patients with drug-resistant MTLE was resistant to carbamazepine in 82% of patients, partially suppressed in 11%, and completely suppressed in 7%. We expanded this study to other anticonvulsants with similar results (Lehmann T, Gabriel S, Sandow N, Kim S, Heinemann U, unpublished data).


Overexpression of drug efflux pumps at the BBB has been suggested to contribute to pharmacoresistance (Löscher & Potschka, 2005). However, in such slices BBB is not maintained, suggesting that overexpression of drug efflux transporters at the BBB is not a major cause of pharmacoresistance. This does not exclude that drug transporters affect tissue levels by expression in neurons and astrocytes within the tissue. Interestingly, already in the study of Jandová et al. (2006) it was noted that in a subset of slices coming from pharmacoresistant patients, some tissue was pharmacoresistant whereas some was not. This suggests that in many cases of epilepsy surgery pharmacoresistance may be more restricted than the epileptogenic zone.

At present, therefore, we have no good understanding of mechanisms involved in pharmacoresistance. This requires the search for alternative mechanisms. For example alterations in network properties may contribute to pharmacoresistance. This would be the case if a major part of information transfer is provided for by disinhibition chains such as those that transmit information through the basal ganglia. In such structures GABAergic cells innervate other GABAergic cells, causing disinhibition in the next cell within a neuronal transmission chain. In such a network both sodium channel blockers may lose efficacy if they prevent high-frequency discharges of some types of GABAergic cells, and barbiturates and benzodiazepines may lose efficacy if they augment disinhibition. In chronic epilepsies development of tolerance may also contribute to reduced efficacy of GABAergic drugs.

Finally the mechanisms underlying pharmacoresistance may be due to mechanism of seizure generation, which exploits mechanisms not targeted by presently available drugs. Research into seizure onset in such patients has, for example, suggested that alteration in Cl homeostasis is an important factor in generation of epileptiform discharges, and often seizure-like events seem to be preceded by abnormal GABA responses (Uva et al., 2009). Therefore the possibility must be left open that the mechanisms of seizure generation in pharmacoresistant patients are different from those of patients whose seizures can be controlled by presently available drugs.


In conclusion, it is important to understand the mechanisms of pharmacoresistance to develop new drugs for the treatment of patients with pharmacoresistent epilepsy. In vitro models are useful tools for investigating the mechanisms of drug resistance. In the future the research on human tissue may provide us useful information for understanding the mechanisms and developing new therapies for pharmacoresistant patients. Because frequently antiepileptic drugs that augment GABAergic function do not work in all these models, but competitive GABA receptor agonists do, the latter might be an alternative in the search for novel anticonvulsants suitable for the treatment of patients with pharmacoresistant epilepsy.


This research was supported by EpiCure and the SFB TR3 as well as by a grant of the Hertie Foundation.


None of the authors have any conflict of interest to disclose, as all authors are employed by a research institution and the work was not funded by industry.