High seizure frequency prior to antiepileptic treatment is a predictor of pharmacoresistant epilepsy in a rat model of temporal lobe epilepsy


  • Wolfgang Löscher,

    1. Department of Pharmacology, Toxicology, and Pharmacy, University of Veterinary Medicine Hannover, and Center for Systems Neuroscience, Hannover, Germany
    Search for more papers by this author
  • Claudia Brandt

    1. Department of Pharmacology, Toxicology, and Pharmacy, University of Veterinary Medicine Hannover, and Center for Systems Neuroscience, Hannover, Germany
    Search for more papers by this author

Address correspondence to Dr. Wolfgang Löscher, Department of Pharmacology, Toxicology and Pharmacy, University of Veterinary Medicine, Bünteweg 17, D-30559 Hannover, Germany. E-mail: wolfgang.loescher@tiho-hannover.de


Purpose: Progress in the management of patients with medically intractable epilepsy is impeded because we do not fully understand why pharmacoresistance happens and how it can be predicted. The presence of multiple seizures prior to medical treatment has been suggested as a potential predictor of poor outcome. In the present study, we used an animal model of temporal lobe epilepsy to investigate whether pharmacoresistant rats differ in seizure frequency from pharmacoresponsive animals.

Methods: Epilepsy with spontaneous recurrent seizures (SRS) was induced by status epilepticus. Frequency of SRS was determined by video/EEG (electroencephalography) monitoring in a total of 33 epileptic rats before onset of treatment with phenobarbital (PB).

Results: Thirteen (39%) rats did not respond to treatment with PB. Before treatment with PB, average seizure frequency in PB nonresponders was significantly higher than seizure frequency in responders, which, however, was due to six nonresponders that exhibited > 3 seizures per day. Such high seizure frequency was not observed in responders, demonstrating that high seizure frequency predicts pharmacoresistance in this model, but does not occur in all nonresponders.

Discussion: The data from this study are in line with clinical experience that the frequency of seizures in the early phase of epilepsy is a dominant risk factor that predicts refractoriness. However, resistance to treatment also occurred in rats that did not differ in seizure frequency from responders, indicating that disease severity alone is not sufficient to explain antiepileptic drug (AED) resistance. These data provide further evidence that epilepsy models are useful in the search for predictors and mechanisms of pharmacoresistance.

Despite the availability of numerous effective antiepileptic drugs (AEDs), about 20–40% of patients with epilepsy continue to have frequent seizures during treatment (French, 2007). Such pharmacoresistant epilepsy (also known as medically intractable or refractory epilepsy) is often a chronic, lifelong problem, associated with increased psychosocial and physical morbidity and increased mortality (Devinsky, 1999). Identifying clinical predictors for pharmacoresistant epilepsy early in the course of the disorder may be important for directing epilepsy patients to more aggressive pharmacotherapy or an effective nonpharmacologic treatment, such as surgery or vagus nerve stimulation (Arroyo et al., 2002; French, 2002; Hitiris et al., 2007; Rogawski & Johnson, 2008). Factors that have been repeatedly identified as potential predictors of refractory epilepsy include high frequency, density, or clustering of seizures before onset of treatment; a history of psychiatric comorbidity; brain lesions such as hippocampal sclerosis; and failure of response to the first AED (Camfield & Camfield, 1996; Kwan & Brodie, 2000; French, 2002; Kwan & Brodie, 2002, 2004; French, 2007; Hitiris et al., 2007). However, in the absence of understanding the mechanisms of resistance to treatment, more work is needed to determine the causes and predictors of pharmacoresistance. Animal models of epilepsy that allow direct comparison of pharmacoresistant and pharmacoresponsive animals in the absence of the various confounding factors that complicate epidemiologic studies may be valuable tools for understanding the biologic basis of drug resistance and whether or how it can be predicted (Löscher, 2006).

Over the last 20 years we have developed and characterized two rat models of pharmacoresistant temporal lobe epilepsy (TLE): phenytoin-resistant kindled rats and phenobarbital (PB)–resistant epileptic rats (Löscher, 2006). In the latter model, spontaneous recurrent seizures (SRS) occur after a latency period following induction of status epilepticus (SE) via electrical stimulation of the basolateral amygdala (BLA) (Brandt et al., 2003). By prolonged treatment of such epileptic rats with PB, two subgroups of rats can be differentiated: PB responders and nonresponders (Brandt et al., 2004). Greater than 80% of the PB nonresponders are also resistant to treatment with a second AED (Bethmann et al., 2007), so that—in parallel to the clinical situation—inadequate seizure control after initial treatment is a predictor for subsequent failure to other AEDs in this model. In the present study, we investigated whether PB-resistant rats differ from PB-responsive rats in seizure frequency before treatment. In previous studies on this model, we had repeatedly observed a trend for higher pretreatment seizure frequency in PB-nonresponders, but this difference from responders was not statistically significant because of the small sample size (Brandt et al., 2004; Bethmann et al., 2007). Therefore, we combined rats from three separate experiments for the present study, thereby increasing the statistical power for such a comparison. This allowed us to evaluate whether intrinsic disease severity is a determinant of AED refractoriness, as recently suggested from clinical data (Rogawski & Johnson, 2008).

Materials and Methods


As in our previous experiments in rats with SRS developing after SE induced by prolonged electrical stimulation of the BLA (Brandt et al., 2003; Volk & Löscher, 2005; Volk et al., 2006; Bethmann et al., 2007), adult female Sprague-Dawley rats (Harlan-Winkelmann, Borchen, Germany) were used for this study. All rats were purchased at the same age and body weight (200–230 g). Following arrival, the rats were kept under controlled environmental conditions (24–25°C; 50–60% humidity; 12 h light/dark cycle; lights on at 6:00 a.m.) with free access to standard laboratory chow (Altromin 1324 standard diet) and tap water. Before being used in the experiments, the rats were allowed to adapt to the new conditions for at least 1 week. All animal experiments were carried out in accordance with the European Communities Council Directive of 24. November 1986 (86/609/EEC) and were formally approved by the animal subjects review board of our institution. All efforts were made to minimize the number of animals used and their suffering.

Electrode implantation and SE induction

Electrodes were stereotactically implanted into the right anterior BLA under anesthesia as described in detail recently (Brandt et al., 2003) and served for electrical stimulation and electroencephalography (EEG). About 4 weeks after electrode implantation, rats were electrically stimulated via the BLA electrode for induction of a self-sustained SE as described previously (Brandt et al., 2003, 2004; Volk & Löscher, 2005). The following stimulus parameters were chosen: stimulus duration 25 min; stimulus consisting of 100 msec trains of 1 msec alternating positive and negative square wave pulses. The trains were given at a frequency of 2/s and the intratrain pulse frequency was 50/s. Peak pulse intensity was 700 μA. For this pulsed-train stimulation, an Accupulser A310C stimulator connected with a Stimulus Isolator A365 (World Precision Instruments, Berlin, Germany) was used. In all rats, EEG was recorded via the BLA electrode during SE and up to 20 h after termination of SE by diazepam (see subsequent text in this article). Only rats that developed a self-sustained SE with generalized convulsive seizures were used for further experiments. SE was interrupted after 4 h by diazepam (10 mg/kg, i.p.), which is highly effective for termination of SE in this model (Brandt et al., 2003; Bankstahl & Löscher, 2008). In 16 of the 33 rats used in this study, it was necessary to repeat the application of diazepam for complete suppression of clinical and EEG seizures. Rats that were subsequently categorized as PB-nonresponders (see subsequent text) did not require more diazepam to terminate SE than PB-responders (p = 0.2960). As reported previously, diazepam completely stopped motor and EEG seizure activity after either one or two injections in all rats, as shown by continuous video and EEG recording for up to 20 h after injection of diazepam (Brandt et al., 2003, 2004; Bethmann et al., 2007). Starting 6 weeks later, the rats were monitored by video-EEG recordings for about 2 weeks until the first spontaneous seizures were detected as described recently (Brandt et al., 2003, 2004). Rats with SRS were then used for treatment with PB. In the 33 rats used for the present study, baseline seizure recording (predrug control) for the PB trial (see subsequent text) started after an average interval of 14 weeks following the SE, at which time seizure frequencies have become relatively stable in most rats in this model (Brandt et al., 2003); this corresponds to the data reported by Gorter et al. (2001) on rats in which SE was induced by electrical tetanic stimulation of the angular bundle.

Selection of responders and nonresponders by prolonged treatment with phenobarbital

For selection of pharmacosensitive and pharmacoresistant rats by prolonged treatment with PB, we used a dosing protocol with an intraperitoneal bolus dose of 25 mg/kg on the morning of the first treatment day, followed 10 h later by an administration of 15 mg/kg, i.p., and then twice daily 15 mg/kg, i.p. for the 13 subsequent days (Brandt et al., 2004). We have shown previously that this dosage regimen leads to maintenance of therapeutic plasma levels of PB (10–40 μg/ml; Baulac, 2002) in rats (Brandt et al., 2004). Before onset of drug treatment, baseline seizure frequency was determined over 2 weeks (predrug control period); then PB was administered over 2 weeks, followed by a postdrug control period of 2 weeks. In this way, each animal served as its own control, accounting for differences between animals, for example, variability in baseline seizure frequency. During drug treatment, rats were closely observed for adverse effects (ataxia, sedation, muscle relaxation) as described previously (Brandt et al., 2006). Ataxia was repeatedly scored by a rating system over the 2 weeks of treatment with PB. Blood was sampled by retroorbital puncture (after local anesthesia with tetracaine) 10 h after the first drug injection and 10–12 h after the last drug injection (in 22 of the 33 rats also 10 h after drug injection on day 7) for PB analysis in plasma by high-performance liquid chromatography (HPLC) with ultraviolet detection (Potschka et al., 2002). In all rats, seizures were continuously (24 h/day, 7 days/week) monitored by video-EEG recording over the 6 weeks of the experiment as described later. For the present study, data from two previous experiments with selection of PB-responders and PB-nonresponders (Brandt et al., 2004; Bethmann et al., 2007) were combined with data from an unpublished third selection trial to obtain enough rats for the purpose of this study. In this way, a total of 33 rats with SRS that underwent selection into responders and nonresponders could be used. All of these rats had the same age and body weight at onset of the experiments, and all experiments were performed with the same protocol, thereby allowing combined evaluation of these experiments. Furthermore, body weight was tracked over the entire duration of the experiments in all rats.

Monitoring and analysis of spontaneous recurrent seizures

For continuous EEG-monitoring, an 8-channel amplifier (CyberAmp 380; Axon Instruments Inc., Foster City, CA, U.S.A), eight 1-channel bioamplifiers (ADInstruments Ltd., Hastings, East Sussex, U.K.), and two analog-digital converters (PowerLab/800s; ADInstruments Ltd.) were used. This system allowed simultaneous recording of EEG from up to 16 rats over the experimental periods. The data were recorded and analyzed with the CHART4 for windows software (ADInstruments Ltd). The sampling rate for the EEG recording was 200 Hz. A high pass filter for 0.1 Hz and a low pass filter for 60 Hz were used.

Simultaneously to the EEG recording, all rats used in this study were video-monitored continuously during the experimental periods, using four light-sensitive black-white cameras (CCD-Kamera-Modul; Conrad Electronic, Hannover, Germany), which allowed video-recording of up to four rats per camera. The cameras were connected to a multiplexer (TVMP-400; Monacor, Bremen, Germany), which converted the signals from the four cameras to a video recorder (Time Lapse recorder; Sanyo TLS-9024P, Monacor, Bremen, Germany). To allow video-recording of seizures during the night, infrared light was used during the dark phase. Rats were housed in clear glass cages (one per cage) to allow optimal video observation.

For detection of spontaneous seizures, the EEG recordings were visually analyzed for characteristic ictal events. The typical ictal discharges occurring during spontaneous seizures in this model have been described in detail previously (Brandt et al., 2003). To evaluate the severity of motor seizure activity during a paroxysmal alteration in the EEG, the corresponding video-recording was viewed. In addition to seizures observed by video and EEG recordings, seizures observed during handling or other manipulations of the animals were noted. For rating of seizure severity of spontaneous seizures, Racine’s scale (1972) was used. In addition to the five seizure stages rated by this grading system, a stage 6 was used to characterize running-bouncing seizures, which were occasionally observed before or after a generalized convulsive seizure. Based on this scale, seizures were subdivided into nonconvulsive (stages 1 and 2) and convulsive (stages 3–6). Some rats displayed an additional convulsive seizure type not covered by the Racine scale, which was characterized by jerky backward movements of the head sometimes associated with a short, clonic movement of one or both forelimbs.

Based on individual responses of rats to treatment, they were considered either responders or nonresponders. Responders were defined by complete seizure suppression during treatment or a seizure suppression of at least 75% compared to seizure frequency in the control periods. In some of our previous studies, we had used > 50% seizure reduction as criterion for response to PB (Brandt et al., 2004; Bethmann et al., 2007), but all responders that we selected by this criterion had at least 75% seizure reduction, so that we decided recently to use this stricter criterion (Bethmann et al., 2008; Gastens et al., 2008). This did not change the results of selection, because none of the nonresponders exhibited > 50% seizure reduction (see Results).


The significance of differences between the predrug, drug, and postdrug recordings within the same group of rats was calculated by analysis of variance (ANOVA) for paired nonparametric data (Friedman test), followed by Dunn’s multiple comparison test. Significance of differences between different groups of rats was calculated by either Student’s t-test or Mann-Whitney U test, depending on whether data were normally distributed or not. Frequencies were analyzed by Fisher’s exact test. All tests were used two-sided; p < 0.05 was considered significant.


Selection of PB-responders and PB-nonresponders

With the dosing protocol used for selection of responders and nonresponders, PB induced marked sedation and ataxia, indicating that maximum tolerated doses were used. In most rats, ataxia lasted over the 2 weeks of treatment, but in some of the animals the severity of ataxia was decreased in the second week. Analysis of plasma drug concentrations showed that drug concentrations within or above the therapeutic range (10–40 μg/ml) known from patients with epilepsy (Baulac, 2002) were maintained throughout the period of treatment except for two rats (NIH 62 and NIH 95), in which, therefore, the dose of PB was increased to 20 mg/kg twice daily.

In the 33 rats used for testing anticonvulsant efficacy of PB, average seizure frequency in the 2 weeks before onset of treatment with PB was 2.0 seizures per day (Fig. 1A), with a large variation of individual data (range 0.07–16.4 seizures per day). As described previously (Brandt et al., 2003, 2004), most spontaneous seizures that were recorded were generalized convulsive seizures, resembling stage 4 or 5 seizures on the Racine scale (1972). Compared to predrug seizure frequencies, seizure frequency increased significantly during treatment with PB (Fig. 1A). In the postdrug control period, seizure frequency increased in several rats compared to that determined before onset of treatment (Fig. 1A), but the difference compared to predrug control was not statistically significant.

Figure 1.

Effect of phenobarbital (PB) on spontaneous recurrent seizures (SRS) in rats. SRS were recorded over a period of 2 weeks before onset of PB treatment (predrug control), followed by drug treatment for 2 weeks, and then a 2-week postdrug control period. All data are shown as mean ± SEM (standard error of the mean). (A), (B) and (C) show the average number of seizures per day recorded over the three 2-week periods, whereas (D) illustrates the average plasma concentration of PB from the blood samples taken during treatment. In (A), average seizure data from 33 epileptic rats recorded in three separate experiments are given, whereas (B) shows respective data from the 20 responders and (C) data from the 13 nonresponders of these experiments (see text for definitions). Analysis of data by nonparametric repeated measures ANOVA (Friedman test) indicated significant differences for the data shown in (A) (p = 0.0007) and (B) (p < 0.0001) but not (C) (p = 0.1180). Post hoc analysis indicated that seizure frequency during PB treatment was significantly higher versus predrug and postdrug control in (A) (p < 0.05; indicated by asterisk), whereas in (B) PB significantly suppressed SRS compared to the predrug and postdrug periods (p < 0.001; indicated by asterisk). In the nonresponder group shown in (C), there was a tendency for increased seizure frequency in the postdrug versus predrug control period, but the difference was not statistically significant. PB plasma levels of responders and nonresponders did not differ significantly (ANOVA, p = 0.152). The shaded area in (D) indicates the therapeutic plasma concentration range of PB.

However, by averaging data from all 33 rats as shown in Fig. 1A, striking differences in individual response to PB were masked. In 14 of these rats, complete control of seizures was achieved, and 6 other rats exhibited at least 75% reduction in seizure frequency (on average 84%, range 75–94%). These 20 rats were considered responders. Data from these rats are shown in Fig. 1B, demonstrating a significant anticonvulsant effect of PB in this subgroup when compared to either pre- or postdrug seizure frequency. In 13 other rats, that is, 39% of the 33 rats treated with PB, this AED did not exert an anticonvulsant effect, so these rats were considered nonresponders. Data from these rats are shown in Fig. 1C. Analysis of seizure frequency in nonresponders by ANOVA did not indicate any significant difference between the predrug, drug, and postdrug data. Compared to predrug seizure frequency, seizure frequency increased in 8 of the 13 nonresponders during treatment with PB, whereas the remaining 5 nonresponders showed a moderate decrease of seizures by 35%, on average (range 16–50%). As previously reported (Brandt et al., 2004, 2006; Bethmann et al., 2007), the severity or duration of the initial brain insult (the SE) did not differ between responders and nonresponders. Furthermore, responders and nonresponders did not differ significantly in the efficacy of diazepam to interrupt SE. Body weight of PB-nonresponders (316 ± 5.8 g) was slightly (6.4%) but significantly higher than that of PB-responders (297 ± 5.1 g; p = 0.0227) at the end of the postdrug vehicle phase.

In the nonresponder group, there was a tendency for increased seizure frequency in the postdrug versus predrug control period (Fig. 1C), but the difference was not statistically significant. At least in part, this trend for increased seizure frequency in the postdrug control was due to some rats, in which seizure frequency increased shortly after termination of treatment with PB, indicating withdrawal hyperexcitability, as previously described for prolonged treatment with PB in rats (Löscher & Hönack, 1989).

Plasma levels of PB did not differ significantly between responders and nonresponders at any of the 2–3 determinations (i.e., on days 1, 7, and 14) during the treatment period; plasma concentrations averaged from the 2–3 determinations in all 33 rats and in the responder and nonresponder subgroups are shown in Fig. 1D. No significant difference was found between these values. Furthermore, the severity of adverse effects of PB did not differ significantly between the responder and nonresponder subgroups (not illustrated), thus excluding that resistance was due to too low doses (or plasma concentrations) of PB in the nonresponders.

Differences in seizure frequency between PB responders and nonresponders before onset of treatment

As shown in Fig. 2, PB responders exhibited a relatively low variation in seizure frequency before onset of treatment with PB. Average seizure frequency was 0.54 (range 0.07–1.29) seizures per day. In contrast, a large variation in frequency of seizures was determined in PB-nonresponders (Fig. 2). Average seizure frequency was 4.31 (range 0.14–16.4) seizures per day. Average seizure frequency in PB-nonresponders differed significantly from seizure frequency in PB-responders (p = 0.0050). However, as shown in Fig. 2, the significantly higher seizure frequency of nonresponders was due to six rats that exhibited > 3 seizures per day. Such high seizure frequency was observed only in the nonresponder group (6 of 13) and not in the responder group (0 of 20), which was significantly different by Fisher’s exact test (p = 0.0015). Therefore, high seizure frequency (> 3 seizures per day) predicted resistance to PB. The data in Fig. 2 indicated that there may be two subgroups of PB-resistant rats, a subgroup with high frequency of SRS and a subgroup with low SRS frequency. Therefore, these subgroups were further compared.

Figure 2.

Frequency of spontaneous recurrent seizures (SRS) in 33 rats developing epilepsy after status epilepticus induced by sustained stimulation of the basolateral amygdala. Thirteen (39%) of the 33 rats were resistant to prolonged treatment with phenobarbital (PB) (nonresponders), whereas the other 20 rats responded to treatment with a reduction of seizure frequency of at least 75% (see Fig. 1). The seizure frequency shown was determined by continuous video-EEG (electroencephalography) recording over 2 weeks prior to onset of treatment with PB. The average frequency of nonresponders (indicated by horizontal line) was significantly higher than average seizure frequency of responders (p = 0.0050; indicated by asterisk).

Differences between PB-resistant rats with low and high seizure frequency

Average predrug seizure frequency in the seven PB-nonresponders with low seizure frequency was 0.34 (range 0.14–0.86) seizures per day (Fig. 3A), which was not significantly different from predrug seizure frequency in PB responders (p = 0.2403). In contrast, seizure frequency in the six PB-nonresponders with high seizure frequency (> 3 seizures/day) was 8.94 (4.1–16.4) seizures per day (Fig. 3B), which was significantly higher than seizure frequency in the other subgroup of nonresponders (p = 0.0012) and seizure frequency in responders (p < 0.0001), respectively.

Figure 3.

Effect of phenobarbital (PB) on spontaneous recurrent seizures (SRS) in two subgroups of nonresponders with either low (A) or high (B) seizure frequency. The allocation of rats to the two subgroups was based on the SRS data in Fig. 2, indicating that there are two subpopulations of nonresponders: one subpopulation with predrug seizure frequency < 1 seizure per day and a second subpopulation with > 3 seizures per day. The predrug seizure frequency of these two subgroups differed significantly (p = 0.0012). The two subgroups also differed in their response to PB. As shown in (A), PB did not exert any effect on seizure frequency in nonresponders with low seizure frequency (ANOVA; p = 0.2359). However, in nonresponders with high seizure frequency, PB significantly increased seizure frequency (ANOVA; p = 0.0012); post hoc testing indicated a significant increase in seizure frequency compared to predrug control (p < 0.05; indicated by asterisk). The seizures illustrated in this figure include both convulsive and nonconvulsive seizures (see text for further details).

When the response to PB was separately analyzed for PB-nonresponders with low and high seizure frequency, PB significantly increased seizure frequency in rats with high predrug seizure frequency (Fig. 3B), whereas this was not observed in PB nonresponders with low seizure frequency (Fig. 3A). Therefore, there seemed to be a qualitative difference in the effects of PB in these two subgroups of PB-nonresponders. Interestingly, in the six rats with high predrug seizure frequency, the type of seizures partly (four rats) or completely (two rats) changed during treatment with PB from secondary generalized convulsive (stage 4–6) to nonconvulsive or focal (stage 1–2), whereas most seizures recorded in the other rats during the control or treatment periods were secondary generalized convulsive seizures. The frequency of nonconvulsive (or focal) seizures in the six rats with high seizure frequency was 0.23 ± 0.12 seizures per day in the predrug period, 11.1 ± 6.3 during treatment with PB (p < 0.05 vs. predrug), and 3.0 ± 1.4 in the postdrug period. Plasma levels of PB were not significantly different in the two subgroups of nonresponders.

The six PB-nonresponders with high seizure frequency differed also behaviorally from all other PB-nonresponders and responders in that they were difficult to handle, hyperexcitable, and aggressive. There was no indication of any differences in the duration or severity of the initial SE in PB-nonresponders with frequent or rare seizures. Furthermore, the efficacy of diazepam in terminating SE was comparable in the two subgroups.

Three of the 13 nonresponders died 7–8.5 months after the SE, which was not observed in any of the 20 responders (p = 0.0524 by Fisher’s exact test). Two of the three nonresponders that died belonged to the group with high seizure frequency. General appearance or body weight of the three rats that died did not differ from those of other epileptic rats, excluding that death was due to a steady deterioration in the health of the animals that died. One of these rats was video-monitored at the time of death, showing that this animal did not experience seizures immediately prior to death. Postmortem examination (necropsy) of this rat after did not indicate any obvious pathologic cause of death.


A history of high seizure frequency, density, or clustering prior to onset of treatment is widely accepted as a negative predictor of treatment outcome (French, 2007; Rogawski & Johnson, 2008). However, such multiple seizures often occur in difficult-to-treat localization-related epilepsies, so that it is not clear whether it is the high seizure frequency or the type of epilepsy, or both, that predicts pharmacoresistance in such patients. In the present study in a rat model of TLE, all rats with multiple seizures prior to treatment were resistant to PB, so that high seizure frequency was a powerful predictor of pharmacoresistance. However, such high seizure frequency was observed in only 46% of nonresponders, whereas the other 54% (7 of 13 nonresponders) did not differ in predrug seizure frequency from PB-responders. Therefore, factors other than high seizure frequency were obviously involved in drug resistance in this model.

We have previously reported that the majority (90%) of PB-nonresponders exhibit hippocampal damage, whereas such damage is determined in only 7% of PB-responders, so that neuron loss in the hippocampus, particularly in the dentate hilus, is a characteristic feature of PB-resistant rats (Volk et al., 2006; Bethmann et al., 2008). Again, this observation in our rat model is in line with clinical experience, in that hippocampal sclerosis in patients with TLE carries a poor prognosis (French, 2007). In addition to hippocampal damage, PB nonresponders differ from responders in increased brain expression of the efflux transporter P-glycoprotein (Pgp) (Volk & Löscher, 2005) and subunit expression and binding characteristics of γ-aminobutyric acid (GABA)A receptors (Volk et al., 2006; Bethmann et al., 2008), which may be critically involved in the lack of antiepileptic efficacy of PB in nonresponders. These clear differences between PB-responders and PB-nonresponders also indicate that the strict definition of response that we chose for selection was suitable for differentiating between pharmacoresistant and pharmacoresponsive rats, despite the low seizure frequency of many epileptic rats in the model used for our experiments. This conclusion is substantiated by the recent finding that most of the PB-nonresponders were also resistant to another AED, that is, phenytoin (Bethmann et al., 2007), which is in line with the clinical experience that failure to respond to initial treatment predicts failure of response to other AEDs (Kwan & Brodie, 2000; French, 2002).

As previously reported (Brandt et al., 2004; Volk et al., 2006; Bethmann et al., 2007), differences in the frequency of SRS or the pharmacologic response of SRS to PB was not a consequence of differences in the severity of the initial brain insult, that is, the SE. Detailed analyses of the video and EEG recordings performed during SE demonstrated that the clinical seizure activity and the number and type of paroxysmal EEG alterations did not differ significantly between responders and nonresponders (Volk et al., 2006). The duration of SE was 4 h in all rats, because it was terminated by diazepam (10–20 mg/kg) after this period of convulsive activity. Furthermore, nonresponders did not require more diazepam than responders to terminate the SE. Recurrence of SE activity was excluded by EEG recordings following SE termination. Therefore, the finding that a comparable or equivalent brain insult results in major interindividual differences in the severity and pharmacologic responsiveness of epilepsy developing after the insult can be explained only by genetic differences between the age-matched rats of the outbred strain (Sprague-Dawley) used for our experiments.

The subgroup of six epileptic rats with extremely high seizure frequency (> 3 seizures per day) in the population of 33 epileptic animals examined in the present study differed in several aspects from all other rats. First, whereas rats of the present model are normally easy to handle without any obvious behavioral abnormalities, these six rats were hyperexcitable and aggressive, resembling epileptic rats of chemical models of TLE, such as the pilocarpine or kainate models. Second, these rats differed in their response to PB from all other rats in that PB significantly increased seizure frequency, that is, exerted a proconvulsant effect in such animals. This increase in seizure frequency was particularly due to the occurrence of nonconvulsive seizures, which were observed only rarely before treatment with PB. These observations indicate that epileptic rats with a high frequency of PB-resistant seizures represent a condition that is distinct from resistant rats with lower frequency of seizures.

Two of the six rats with frequent pharmacoresistant seizures died suddenly and unexpectedly after 6.5 and 8.5 months following the SE, which was also observed in one rat with less frequent pharmacoresistant seizures, but none of the rats with pharmacoresponsive seizures. In patients with epilepsy, uncontrolled partial or generalized convulsive seizures are a risk factor for sudden unexpected death in epilepsy (SUDEP), which is the most important direct epilepsy-related cause of death (Tomson et al., 2005; Nashef et al., 2007). The mechanisms of SUDEP are unclear but most probably involve seizure-induced changes in respiratory and cardiac function (Leung et al., 2006; Nashef et al., 2007). To our knowledge, the present data are the first evidence that sudden death also occurs in rats with pharmacoresistant spontaneous seizures.

Interestingly, in PB-resistant rats with high seizure frequency, the type of seizures changed during treatment from secondary generalized convulsive (stages 4–6) to focal (stages 1–2), which did not occur in the rest of the group. In rat models of TLE, sustained or frequent seizures have been shown to induce a locally restricted increase in the expression of the drug efflux transporter Pgp in epileptogenic brain regions, which locally decreases concentrations of AEDs that are substrates of Pgp (Löscher & Potschka, 2005; van Vliet et al., 2007). Therefore, one explanation for the present findings could be that levels of PB, which is a Pgp substrate (Löscher & Potschka, 2005), are decreased in such brain regions, so that PB does not affect focal seizures originating in these regions but still inhibits seizure generalization in brain regions adjacent to the focal areas. This, however, does not explain the proconvulsant effects of PB observed in several PB-nonresponders.

The anticonvulsant effect of PB is thought to be related primarily to enhancement of GABA-mediated inhibitory synaptic transmission via modulation of GABAA receptors (Olsen, 2002; Rogawski & Löscher, 2004). We have reported previously that the subunit expression and binding characteristics of GABAA receptors are significantly altered in PB nonresponders (Volk et al., 2006; Bethmann et al., 2008). However, whereas such alterations in the expression and function of GABAA receptors are likely involved in resistance to PB, they cannot explain the different response to PB in nonresponders with high or low seizure frequency. When we reexamined the autoradiographic and immunohistochemical data on GABAA receptors in nonresponders, no obvious differences between animals with high or low seizure frequency were observed (unpublished data). Therefore, other explanations for the proconvulsant effect of PB in epileptic rats with frequent seizures must be sought after. A likely explanation would be a shift from inhibitory to excitatory GABAA receptors, so that GABAA receptor agonists such as PB exert proconvulsant instead of anticonvulsant activity. Such a shift in GABAergic response polarity from hyperpolarizing to depolarizing has been described both in patients with TLE and animal models of TLE (Cohen et al., 2002; Staley, 2004; Pathak et al., 2007; Li et al., 2008). We currently plan to examine whether such changes may be involved in the different response to PB in nonresponders with high or low seizure frequency.

We performed all experiments in female rats, because these animals are easier to handle during long-term pharmacologic studies, and drug elimination is typically slower than in males, which is an advantage for maintenance of effective drug levels (Löscher, 2007). However, estrous cycle variations, typically 4–5 days in rats, may affect seizure incidence and possibly also drug responses. Indeed, it is well known that seizure frequency in women largely varies with the stage of estrous cycle, and steroid hormones influence neuronal excitability (Scharfman & MacLusky, 2006). However, previous experiments of our group in female rats did not indicate any marked effects of estrous cycle on seizures or drug responses in models of TLE. First, in the amygdala kindling model of TLE, the estrous cycle did not significantly affect seizure threshold or response to phenytoin (Rundfeldt et al., 1990; Wahnschaffe & Löscher, 1992). Second, although phenytoin was less effective in male than in female kindled rats, phenytoin-resistant and phenytoin-responsive animals could be selected from both genders (Ebert et al., 1994). Third, the female epileptic rats with high seizure frequency evaluated in the present and previous studies exhibited about the same number of seizures per day without any indication of estrous cycle variations (see figures of individual rats in Brandt et al., 2004, 2006). In this respect, it is important to note that hormonal changes during the estrous cycle in rats are much less pronounced than in humans, and that the most marked changes in sex hormone levels occur during a relatively short period in the early afternoon of the 4–5 day cycle (Butcher et al., 1974). For the present study, SE induction was always performed in the morning, so that it is unlikely that the differences observed between responders and nonresponders originate in a difference in excitability related to the level of sex hormones at the time of induction of SE.

In conclusion, the present study provides further evidence that our rat model of TLE bears several interesting similarities to human TLE and can be used in the search for predictors and mechanisms of pharmacoresistance. Why the same brain insult, that is, a SE of similar severity and duration, produces epilepsy with such divergent features in age-matched rats of the same strain is not clear at present, but this is most likely due to genetic differences between animals, which are known to be present in outbred strains of rats, such as Sprague-Dawley (Festing, 1993). The clarification of these genetic differences will surely enhance our understanding of how or why pharmacoresistance develops in a particular individual. Rogawski and Johnson (2008) have recently suggested that intrinsic disease severity, characterized by high frequency of seizures in the early phase of epilepsy, is a major determinant of AED refractoriness, which is substantiated by the present experimental data. However, our data also demonstrated that there is no necessary link between seizure frequency and AED resistance, because part of the PB nonresponders did not differ in seizure frequency from PB responders, indicating that disease severity alone is not sufficient to explain AED resistance. This, again, is in line with clinical findings, in that epilepsy patients starting with few seizures do not always respond to AED treatment (Schmidt & Löscher, 2009). Therefore, the current hypotheses of drug resistance should be discussed in an integrative rather than exclusive fashion, which will be critical to the development of strategies to overcome pharmacoresistance.


We thank Prof. Dieter Schmidt (Epilepsy Research Group, Berlin, Germany) for discussions during preparation of the manuscript; and Nicole Ernst, Julia Förster, and Dr. Kerstin Bethmann for help with the video-EEG monitoring of SRS. The study was supported by a grant (Lo 274/9) from the Deutsche Forschungsgemeinschaft and a grant (R21 NS049592) from the National Institutes of Health (NIH; Bethesda, MD, U.S.A).

We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Disclosure: The authors declare that they have no conflicts of interest.