Does arousal interfere with operant conditioning of spike-wave discharges in genetic epileptic rats?

Epilepsy Research (2010) 90, 75—82

journal homepage: www.elsevier.com/locate/epilepsyres

Does arousal interfere with operant conditioning of spike-wave discharges in genetic epileptic rats? Lasse Osterhagen a, Marinus Breteler b,c, Gilles van Luijtelaar a,∗ a

Donders Centre for Cognition, Donders Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen, Nijmegen, The Netherlands b Behavioural Science Institute, Radboud University Nijmegen, Nijmegen, The Netherlands c EEG Resource Institute - Neurofeedback, Nijmegen, The Netherlands Received 30 October 2009; received in revised form 1 March 2010; accepted 17 March 2010 Available online 11 April 2010

KEYWORDS Absence epilepsy; Spike-wave discharges; Arousal; WAG/Rij rats; Neurofeedback; Brain—computer interfaces

Summary One of the ways in which brain computer interfaces can be used is neurofeedback (NF). Subjects use their brain activation to control an external device, and with this technique it is also possible to learn to control aspects of the brain activity by operant conditioning. Beneficial effects of NF training on seizure occurrence have been described in epileptic patients. Little research has been done about differentiating NF effectiveness by type of epilepsy, particularly, whether idiopathic generalized seizures are susceptible to NF. In this experiment, seizures that manifest themselves as spike-wave discharges (SWDs) in the EEG were reinforced during 10 sessions in 6 rats of the WAG/Rij strain, an animal model for absence epilepsy. EEG’s were recorded before and after the training sessions. Reinforcing SWDs let to decreased SWD occurrences during training; however, the changes during training were not persistent in the post-training sessions. Because behavioural states are known to have an influence on the occurrence of SWDs, it is proposed that the reinforcement situation increased arousal which resulted in fewer SWDs. Additional tests supported this hypothesis. The outcomes have implications for the possibility to train SWDs with operant learning techniques. © 2010 Elsevier B.V. All rights reserved.

Introduction

∗ Corresponding author at: Donders Center for Cognition, Donders Institute for Brain, Cognition and Behavior, Radboud University Nijmegen, PO Box 9104, 6500 HE Nijmegen, The Netherlands. Tel.: +31 24 3615621; fax: +31 24 3616066. E-mail addresses: [email protected], [email protected] (G. van Luijtelaar).

Neurofeedback (NF) is a behavioural treatment for several mental and behavioural disorders, and considered an alternative or additional treatment to medical treatments of epilepsy. The origin of NF in the field of epilepsy dates back to the early seventies of the last century when it was demonstrated in cats that the sensorimotor rhythm (SMR) could get under operant control (Wyrwicka and Sterman, 1968) and that SMR trained cats had elevated seizure thresholds

0920-1211/$ — see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.eplepsyres.2010.03.010

76 (Sterman, 1972). Next, SMR training proved to be effective in epileptic patients with poorly controllable seizures (Cott et al., 1979; Sterman and Egner, 2006). Later, studies with patient groups (Lantz and Sterman, 1988; Andrews and Schonfeld, 1992) and meta-analyes (Sterman, 2000; Tan et al., 2009) verified the effectiveness of SMR training as a treatment for epilepsy. Others demonstrated that volunteers and patients can be trained to improve self-control over slow cortical potentials (SCP) (Elbert et al., 1979) and that SCP training might be effective as adjunctive treatment in drug-refractory epilepsy patients (Birbaumer et al., 1991; Rockstroh et al., 1993; Kotchoubey et al., 1996, 2001). Most studies about the efficacy of NF employ subject groups with mixed seizure types, with partial epilepsies being preponderant (Monderer et al., 2002). As far as we know, no study has been published yet that systematically investigated whether efficacy of NF training is dependant of seizure type, including absences. Although absence epileptic patients were included in the Kotchoubey et al. (2001) study, success of the treatment has not been differentiated by seizure type. The reason for the absence of such studies might most often be that large homogenous patient groups are not readily available. Experiments employing animal models provide the possibility to overcome this problem. Animal experiments have additional advantages: they permit control of possibly confounding factors such as genetic variability (by using inbred strains), age, age of onset, medication, and environmental conditions including upbringing and learning history. Furthermore, uncontrollable factors that might exert great influences on efficacy outcomes in humans e.g. expectancies that mediate the placebo effect (Kirsch, 1997) play a less significant role in animal experiments. The WAG/Rij strain is a valid animal model for human absence epilepsy (Coenen and van Luijtelaar, 2003; Depaulis and van Luijtelaar, 2006). Clinical features of rats from this strain resemble clinical features in humans with absence epilepsy. The EEG of a WAG/Rij rat during a seizure also has the typical Spike-and-Slow-wave pattern (Sitnikova and van Luijtelaar, 2007), though the frequency at which spikes and waves occur during a discharge train is 7—11 Hz in comparison to 2.5—4 Hz in humans; but there is no reason why the frequency should be the same across species. Spike-wave discharges (SWDs) in both species are also accompanied by a decrease of consciousness, as was revealed by outcomes of visual and auditory evoked potential studies, and in time estimation tasks. The evoked potentials made during SWDs mimicked most those of slow-wave sleep (when the level of consciousness has decreased) and the occurrence of SWDs influenced the accuracy of the estimation of time elapsed in both species (Meeren et al., 2001; van Luijtelaar et al., 1991a,b). The aim of this study is to investigate whether NF can be utilized to alter the frequency of SWD occurrences in WAG/Rij rats. It will be tried to bring SWDs under operant control by providing the rats with incentives contingently after SWD onsets, thus reinforcing SWDs. It is expected that the number of SWDs will increase with conditioning training and will be higher than the number of SWDs that occurred spontaneously during a previous (non-reinforced) baseline measurement. To test whether the change in frequency of SWD occurrences achieved by reinforcement remains sta-

L. Osterhagen et al. ble in non-reinforced sessions, the number of SWDs during a post-measurement will be recorded as well. Although the increase of the number of SWDs occurrences has no clinical application, we chose for increasing rather than decreasing SWDs, because the primary aim is to demonstrate if direct operant control of SWDs is possible at all. We did not choose to try to decrease the number of SWDs for three reasons. First, using punishment as a mean to decrease SWDs is no option, because it is ethically questionable and has unwanted side-effects including emotional responses like fear and aggression (Azrin and Holz, 1966). Second, the idea of reinforcing SWD-free periods is not feasible, because the total time that the rats are having SWDs is small compared to the total seizure-free time, which would require rewarding them almost continuously. Third, our study design is very similar to the design employed by Wyrwicka and Sterman (1968), which has proven to be effective. In both cases, a distinctive and easily identifiable EEG pattern with moderate frequency of occurrence was chosen for contingent reinforcement.

Methods Animals and surgery procedure Six male WAG/Rij rats, age 9 months, body weight between 302 and 341 g, bred and raised at the Biological Psychology Department of the Radboud University Nijmegen, were used as experimental subjects. Before surgery, rats were housed in pairs in standard cages with cage enrichment (Enviro Dry® ) and a light—dark cycle of 12/12 h with white light on at 8 am. Rats had unrestricted access to food (standard rodent chow) and water before and during the experiment. Rats were handled before surgery and before the first EEG recording. The experiment and its protocol were approved by the Animal Ethics Committee of the Radboud University Nijmegen. Rats were implanted with a tripolar (Plastics One, Roanoke, VA, USA; type: MS333/2a) and a bipolar (type: MS303/2-A) stainless steel electrode set. Bare electrode wires’ diameter was 0.2 mm, surrounded by 0.03 mm polyimide insulation with only the tip of the wires dismantled. Electrodes were implanted epidurally through circular holes in the skull (0.8 mm diameter for single wires of the tripolar electrode set; wires of the bipolar set shared one hole of 1.2 mm). The location of the bipolar electrode wires were 0.5 mm apart from each other at AP −1.8, L −3.2 (all coordinates in mm relative to bregma. AP = anterior/posterior; positive values: more anterior. L = lateral; positive values more left). The bipolar set was not used in this experiment. The location of the tripolar electrode wires were AP +4.7, L −1.7, used as ground; AP −1.8, L +3.2 (sensorimotor cortex); and above the cerebellum, not further specified. Differential EEG was measured between the last two electrodes. Anaesthesia was induced by Isoflurane (Nicholas Piramal (I) Limited, London, UK). Before surgery, rats were medicated with 0.1 ml atropine sulphate and 0.12 ml Rimadyl® (diluted; Vericore Ltd, Dundee, UK). Lidocaine was used as local analgesic. Body temperature was monitored and kept constant by a heating pad. Four stainless steel screws were attached to the skull to hold the electrode sets fixed to the skull by dentist’s cement (Simplex Rapid; Association Dental Product Ltd, Purton, Swindon, Wiltshire, UK). 24 and 48 h after surgery rats were medicated again with 0.12 ml Rimadyl® . After surgery, rats were housed individually in standard cages and a light—dark cycle of 12/12 h with white light on at 9 pm. Rats had unrestricted access to food and water. Rats had a 2-week recovery period before the first recording session. Cage enrich-

Does arousal interfere with operant conditioning of spike-wave discharges in genetic epileptic rats? ment (Enviro Dry® ) was added to the home cages 1 week after surgery.

Apparatus Operant conditioning and pre- and post-EEG measurements took place in Skinner boxes of size 25 cm length × 25 cm width × 40 cm height. Boxes are open at the top. Front and back side were made of clear Plexiglas. The right side and the floor of the boxes were made of 3 mm stainless steel rods spaced 1.4 cm. The left side was made of aluminium. A food magazine in the form of a triangular prism with opening size 4 cm width × 5 cm height and depth of 5 cm was build into the left wall. The food magazine could be illuminated by a dimmed green light. In order to reduce artefacts, cable connection between rat electrodes and recording cable were equipped with impedance transformers which convert high impedance input to low impedance output before the signal travels through the cable to a swivel contact that was connected to the amplifier. The swivel, which was installed above the box, allows for free moving of the animal and minimal movement artefact during EEG recording. Boxes were located in individual soundproofed and air ventilated chambers. Boxes were cleaned with 70% ethanol before each usage. The analogue EEG signal was band-pass filtered between 1 and 100 Hz. First, the signal was digitized at 256 samples/s with 10-bit resolution (DATAQ Instruments, Akron, OH, USA) and recorded to hard disc for offline analysis. Second, the signal was sent to a QPET EEG device (Bio-Medical Instruments Inc., Warren, MI, USA) which was connected via Bluetooth to an IBM-compatible desktop PC running Bioexplorer software (Cyberevolution Inc, Seattle, WA, USA), working at a sample rate of 200 Hz, for online detection of SWDs.

Procedure For a complete schema of the various recording and training sessions see the flow chart in Fig. 1. All recording and training sessions were done in the Skinnerboxes during the rats’ active (dark) period of the day. At several days before the first training session, a few sucrose pellets (45 mg, Campden Instruments, Loughborough, Leicestershire, UK) were used as reward were administered to the rat in their home cage to familiarize the rats to the sucrose pellets. With this protocol it is not necessary to deprive rats from food. First, rats were attached to recording cables and put into the Skinner boxes for 2 h to acquaint them to the experimental setting.

Figure 1 Flow chart of the various sessions in the experiment. Two baseline sessions (BL1 and BL2) were followed by 10 neurofeedback training sessions (T1—T10), and two after training sessions (PM1 and PM2) in order to establish whether the decrease in SWD could be generalized to outside the training conditions. In PM3 and PMEnv the accessibility of the food magazine was manipulated, the relation between the presentation of food pellets and the occurrence of SWDs was investigated in RandPel.

77

The following day, a 2-h baseline EEG recording was done (BL1), 2 weeks later this was followed by another 2-h session (BL2). The food magazine was made inaccessible during these non-training sessions by means of an aluminium division wall that was installed in front of the intelligence panel that contained the food magazine; these recording sessions were done in complete darkness. After these two baseline sessions, rats got next a 2-h magazine training session. Standard sucrose pellets were used as reward. Pellets were given randomly at intervals between 90 and 210 s. When a pellet had been dispensed, the food magazine became illuminated until the rat visited the magazine as detected by a photo sensor. Magazine training seemed to be effective: in the course of the training, the rats took a typical waiting posture in front of the food magazine. Every rat ate all its received pellets. On SWD training days, rats were attached to recording cables and placed into the Skinner boxes. The training session began 5 min later, which was indicated by switching on a dimmed light at the ceiling of the individual chambers. Online detection of SWDs was achieved by a customized algorithm implemented in the Bioexplorer environment. An illustrative example of a detected SWD is shown in Fig. 2, in this case the SWD is aborted by the presentation of a food pellet. The reliability of the SWD online detection was verified by visual inspection of EEG data from all six rats during training session two and the first hour of session eight (altogether 12 h of EEG). 80 SWDs with minimum duration of 2 s were identified from which 79 had been correctly classified by the online detection algorithm. 3 non-SWD-like events (intermediate state, Gottesmann et al., 1998) were misclassified as SWDs. A SWD with minimal duration of 2 s serves as criterion for receiving a reward. When this criterion had been reached, a sucrose pellet was released into the food magazine and the magazine became illuminated until it was visited by the rat. Rats had ten training sessions (T1—T10), the first four sessions lasted 60 min. Because the occurrence of SWDs during sessions became less and less frequent, session duration was increased by 30 min for the fifth session and another 30 min for the remaining five sessions. Two days after the last training session two 2-h postmeasurements EEG (PM1 and PM2) were recorded in the same Skinner boxes, now again with inaccessible food magazines, similar to the two baseline sessions. Because the results of the training (a decrease in SWDs) were against our expectation, the experiment was extended by two additional tests in three different sessions. First, the effect of the training environment on the occurrence of SWD was assessed: session 1, 1 week later, post-measurement (PM3). Regular EEG recordings with inaccessible food magazines, in the same way as the other post-measurements. Immediately thereafter (session 2), access to food magazines was re-established and the light that indicated the beginning of training was switched on to emulate the training session environment (PMEnv), however, food pellets were not delivered. Second, the effect of administering food pellets in itself on the frequency of occurrence of SWDs was assessed: a last test (session 3, RandPel) was done 2 days after PM3/PMEnv. In this session the training environment was emulated as mentioned above, additionally, sucrose pellets were released into the food magazines at intervals ranging randomly between 619 and 873 s. Non-constant intervals were chosen so that the rats did not built up expectancy with respect to when pellets were dispensed. The length of the interval was based on the number of pellets per hour that the rats received during the last NF training session averaged across all rats: 10 pellets were given to each rat in a 2-h session.

Statistical analyses Because one rat bit through its recordings cable during T6 and T7, the rat will be excluded from analysis for these sessions. Due to a

78

L. Osterhagen et al.

Figure 2 EEG recordings of absence seizures in a rat. The upper signal shows the EEG; the lower signal indicates when the feeder motor is switched on to release a sucrose pellet as a reward (upward signal = feeder on). (A) Recording of a single absence seizure. (B) Recording of three consecutive absence seizures. Rewarding absence seizures results in seizure abortion. Note the different time scales in A and B. dysfunctional recording connection, there is no data for another rat on PM2. For this rat, the post-measurement value is based on PM1 alone. To ascertain that lengthening the duration of training sessions had no effect on the mean number of SWDs per hour (SWDs/h, this is equivalent to pellets/h during training sessions), first a two-way (session and hour) repeated-measures ANOVA was conducted on the data of 2-h sessions to test whether there was a difference in number of SWDs/pellets between the first and the second hour. The main effect of hour and the interaction with session were non-significant. Hence, lengthening of session durations caused no problem for the mean SWDs/h as dependent variable and will be used throughout the following analyses. Five out of six rats’ number of SWDs/h increased to some extend between BL1 and BL2, but the mean increase was non-significant. Therefore both pre-measurements were combined into one single baseline (BL). The difference between PM1 and PM2 was also not significant. Both follow-up measurements were combined into one post-measurement (PM). A repeated-measures analyses of variance with sessions as within-subject factor was used as omnibus test in order to establish whether the number of SWDs/h change in the course of the experiment, while post hoc tests (Student t-tests for dependent variables) were be used to further differentiate where the differences could be found. Orthogonal trend analyses were used to describe the changes over T1—T10. The rat for which measurements during T6 and T7 are missing, was excluded from this analysis. Mean duration of SWDs did not differ between BL1 and BL2 and did not differ between PM1 and PM2 (t(5) = 0.54, p = .610). A repeated-measures ANOVA revealed no significant differences in SWD mean durations between training sessions. Therefore, combined measurements were computed by taking the mean of the baseline measurements (BL), the post-measurements (PM), and the training measurements (T) respectively. A Z-test (Pitman, 1993) was used on the RandPel data to test the hypothesis whether times of occurrences of SWDs were independent from the times pellets were delivered. For this purpose, all time segments bounded by dispensing a pellet and 619 s later, which was the shortest interval between two pellet releases, were considered. Under the null hypothesis, the probability for a SWD onset

is independent from pellet dispensing and thus the same throughout the whole interval. The expected value for the mean interval between SWD onset time and the time when the last pellet had been dispensed is half the segments length (309.5 s). Three rats had no SWDs during RandPel. Therefore, Z statistics were computed for the remaining three rats. Because longer mean distances were expected between SWD onsets and the last pellet, one-sided tests were used.

Results SWDs/pellets per hour Mean number of SWDs/h for the various sessions are shown in Fig. 3. The repeated-measures ANOVA showed that the

Figure 3 Mean number of SWDs/h across sessions. Error bars indicate SEM. Black bars show baseline measurements (BL); grey bars show trainings measurements (T); white bars show post-measurements (PM). The number of SWDs/h decreased significantly with training session which is indicated by the dotted trend line.

Does arousal interfere with operant conditioning of spike-wave discharges in genetic epileptic rats?

79

Table 1 Z statistics for the average distances between SWD onset time and the time when the last pellet had been dispensed for those rats that had SWDs during RandPel. xexp is the expected distance under H0 . xobs is the observed distance for the respective rat. The difference between xobs and xexp is given in s. Rat no.

Number of SWDs

Standard error

xobs − xexp

Z

p

#1 #3 #4

68 59 26

21.68 23.28 35.06

65.4 68.9 61.3

3.01 2.95 1.74

.001 .002 .041

Training environment and SWDs

Figure 4 Box plots of the mean duration of SWDs during baseline sessions (BL), training sessions (T), and post-measurement sessions (PM).

number of SWDs/h changed in the course of the experiment (F(11, 44) = 6.75, p = .001). The following post hoc tests were done: BL vs. T10 (t(5) = 7.37, p = .001), the mean number of SWDs/h was higher during baseline compared to the last training session. T10 vs. PM, t(5) = 4.92, p = .004. Mean number of SWDs/h was lower during the last training session compared to the post-measurement. BL vs. PM: (t(5) = 1.51, p = .190). The number of SWDs/h did not differ between baseline and post-measurement. T1 vs. T10 (t(5) = 4.40, p = .007): the mean number of SWDs/h was lower during the last in comparison to the first training session. The orthogonal trend analyses over all ten training sessions indicated a significant linear trend (F(1, 4) = 19.90, p = .011), this is interpreted as that the mean SWDs/h decreased with training session.

Mean duration of SWDs Mean duration of SWDs over sessions is shown in Fig. 4. A repeated-measures ANOVA revealed significant differences between baseline (BL), training (T) and post-measurements (PM) (F(2, 10) = 47.99, p = .0001). While mean SWD duration did not differ between BL and PM, (t(5) = 0.28, p = .795), it was significant lower during T than during BL or PM, t(5) = 18.58, p = .0001, t(5) = 6.40, p = .001, respectively. These differences are due to the fact that during training sessions the majority of SWDs were aborted almost immediately after a pellet had been dispensed.

The number of SWDs/h between PM3 and PMEnv decreased slightly, but this was not significant (t(5) = 0.31, p = .773), and also the number between PMEnv and RandPel decreased non-significantly (t(5) = .67, p = .534). During RandPel, three rats showed no SWDs, for one rat the number of SWDs/h decreased and for two rats it increased. Mean SWDs/h decreased from PM3 to PMEnv (t(5) = 2.82, p = .037) and PMEnv to Randpel, though the latter non-significantly (t(5) = 0.67, p = .534). Table 1 shows the number of SWDs, standard error, and mean distances for the three rats with SWDs during RandPel. The mean distance between SWDs and the last pellet dispensed was significantly higher (all p’s <.05, the Z statistics can be found in Table 1) for all three rats than the expected value under the null hypothesis. Therefore it can be concluded that dispensing sucrose pellets postponed the onset of SWDs.

Discussion The overall number of SWDs/h decreased in the course of the training, after ten training sessions it was lower than during the first training session and lower than baseline. Therefore it seems that the effect of rewarding SWDs was in the opposite direction than expected. However, the decrease of SWDs was not permanent: it returned to the baseline level at the post-measurements. The first conclusion that can be drawn is that increasing the incidence of SWDs in WAG/Rij rats by means of operant conditioning was not successful. The question towards the cause of this decrease remains. Which factors of the training procedure could be accountable for the decrease of SWDs? It could be speculated that the rats had difficulties to associate their seizures with the reward, because absence seizures are accompanied by impaired consciousness (Oller, 2000; van Luijtelaar et al., 2007) and have a negative effect on information processing. This may mean that they did not notice the release of the reward timely. Patients with absence epilepsy make more omission errors on the visual (Mirsky et al., 1960) as well as on an auditory Continuous Performance Test (Mirsky and van Buren, 1965; Mirsky et al., 1986). WAG/Rij rats respond later in an operant (fixed interval) task when a SWD occurred during the trial in comparison to SWD-free trials (van Luijtelaar et al., 1991a). On

80 the other hand, others found hardly any errors in a continuous performance task if the duration of SWDs was less than 3 s in comparison to SWDs with longer duration (Goode et al., 1970) and in our experiment pellets were released 2 s after SWD onset. Drinkenburg et al. (2003) exposed WAG/Rij rats ictally to auditory stimuli that were either reinforced or non-reinforced previously. Reinforced stimuli led to a much higher percentage of SWD abortions than nonreinforced stimuli. We found the same in our experiment: releasing a reward from the food pellet dispenser led almost always to an immediate abortion of SWDs. This result verifies the existent information processing capability during SWDs. Therefore it can be concluded that the rats took notice of the reward, a prerequisite for learning the association between SWD and reward. Other factors might also account for the decrease of SWDs across training sessions. The presentation of a reward shortly after SWD onset let consistently to SWD abortion, meaning that the rats did not get full-sized seizures and the total duration of SWD activity was decreased during training. One could suspect this of having also an effect on the number of seizures. However, there exists no literature on such an effect. Moreover, the number of SWDs returned to initial levels after training, suggesting at the utmost a short-term effect. An important factor that influences seizure occurrence likelihood is the behavioural or vigilance state of the subject. Bureau et al. (1968) noted that children with absence epilepsy have fewer seizures while being stimulated or being mentally or physically active. Coenen et al. (1991) described that SWDs in rat models are more likely to occur during low vigilance states, such as passive wakefulness and drowsiness and that active behaviour decreases the probability for the occurrence of SWDs. van Luijtelaar et al. (1991a) showed that while WAG/Rij rats were involved in a task at which they could acquire food rewards, they had significantly fewer seizures than during baseline measurements. This experiment shares some properties to our experiment: both experiments took place in Skinner boxes different from the rats’ home cages and both tasks involve administering food incentives to the rats. Killeen et al. (1978) demonstrated with pigeons that the presentation of rewards generated arousal that decayed over time since the last presentation. Furthermore, they observed that the pigeons’ starting level of activity increased with every subsequent session that involved presentation of food rewards. They attributed this effect to the possibility that the environment where feeding took place, became increasingly associated to food and thereby became a conditioned elicitor for arousal. This theory can explain why high arousal levels can be expected in our rats during training sessions which ultimately led to the decreased number of SWDs. They can also explain why the number of SWDs progressively decreased with training sessions. The idea that conditioned responses are not only associated to conditioned stimuli, but also to the context in which conditioning takes place is described by the Rescorla—Wagner model (Rescorla and Wagner, 1972). The model has been repeatedly confirmed in conditioning paradigms in which emotional (fear) responses became associated to the experimental environment (Bouton and Bolles, 1979; Bouton, 1984). However, there exists no literature yet about arousal conditioned to the environment and its

L. Osterhagen et al. influence on SWDs in WAG/Rij rats or in comparable animal models. To test the environmental factor of the arousal hypothesis, we extended the experiment by recording the number of SWDs in an environment that indicates by the presence or absence of discriminative stimuli that rewards can be obtained, while the incentives were not presented. It was expected that the number of SWDs is higher in the environment without the discriminative stimuli because the rats should be less aroused. The results were non-significant, only a decrease in mean duration of SWDs was found between accessible and non-accessible food dispenser. Next we tested whether administering incentives arouse rats and thereby decrease the likelihood of SWDs, sucrose pellets were given to the rats at random times. Because this effect on arousal decays over time as mentioned above, it was expected that the likelihood of SWDs is low directly after presentation of the reward and increases with time. Randomly administering rewards delayed the occurrence of SWDs in three subjects or inhibited them completely, as was shown by three other WAG/Rij rats. Both findings are in full accordance with the arousal hypothesis. In the first case, arousal was high directly after releasing a reward, thereby preventing the occurrence of SWDs. With the passage of time, arousal decreased to a level that made the occurrence of SWDs possible again. In the second case, this lower arousal level was not reached before the next reward was delivered. In summary, staying with the arousal theory as best supported by literature and most plausible, it can be concluded that the effect of arousal outperformed the effect of operant conditioning of SWDs. Two directions of further research might be worthwhile to pursue. The first is to further explore the arousal hypothesis. Then, it would be interesting to ascertain whether it is possible to increase the arousal level of subjects in such a manner, that longer lasting effects on SWDs are observable. The results of Coenen et al. (1991) suggest this possibility. They showed that WAG/Rij rats had decreased SWD occurrences during REM-sleep deprivation days and that the return of the number of SWDs to baseline levels was slow after discontinuation of REM-sleep deprivation. The second research direction is teaching subjects by NF to change other EEG patterns and observe their influence on SWDs. One could e.g. look for an EEG pattern that is similar to SMR in human. SMR is rarely described in rats, if such a rhythm exists in rats is a hot discussed topic with one group of researchers arguing for its existence (Wiest and Nicolelis, 2003; Fontanini and Katz, 2005; Marini et al., 2007), other groups interpret the rhythm as absence-related activity since it is not occurring during attentive wakefulness but during a drowsy state (Robinson and Gilmore, 1980; Vergnes et al., 1982; Shaw, 2004, 2007; Sitnikova and van Luijtelaar, 2007). The alternative, applying SCP training to animals might be difficult, because non-neural generators for DC-shifts, e.g. eye movements (Strehl, 2009) or changes in brain CO2 -levels by means of hyperventilation (Voipio et al., 2003) are hardly controllable in animals. An interesting combination of an NF and an arousal approach might be to provide subjects with arousing feedback on SWD precursor activity (Sitnikova and van Luijtelaar, 2009), so that the arousal generated in the subject eventually will suppress the otherwise following SWD.

Does arousal interfere with operant conditioning of spike-wave discharges in genetic epileptic rats?

Acknowledgements We thank Jos Wittebrood for realising the real-time EEGSkinnerbox facilities and interface, Hubert Voogd for programming Skinnerbox control, and Gerard van Oijen for maintenance of recording devices. We would also like to thank Saskia Hermeling and Hans Krijnen for animal care and surgery assistance. Gilles van Luijtelaar was supported by the BrainGain project.

References Andrews, D.J., Schonfeld, W.H., 1992. Predictive factors for controlling seizures using a behavioral approach. Seizure 1, 111—116. Azrin, N.H., Holz, W.C., 1966. Punishment. In: Honig, W.K. (Ed.), Operant Behavior: Areas of Research and Application. AppletonCentury-Crofts, New York, pp. 390—477. Birbaumer, N., Elbert, T., Rockstroh, B., Daum, I., Wolf, P., Canavan, A., 1991. Clinical-psychological treatment of epileptic seizures: a controlled study. In: Ehlers, A. (Ed.), Perspectives and Promises in Clinical Psychology. Plenum Press, New York, pp. 81—96. Bouton, M.E., 1984. Differential control by context in the inflation and reinstatement paradigms. J. Exp. Psychol. [Anim. Behav.] 10, 56—74. Bouton, M.E., Bolles, R.C., 1979. Role of conditioned contextual stimuli in reinstatement of extinguished fear. J. Exp. Psychol. [Anim. Behav.] 5, 368—378. Bureau, M., Guey, J., Dravet, C., Roger, J., 1968. Frequency of petit mal absences in children as a function of their activity. Rev. Neurol. 118, 493—494. Coenen, A.M.L., Drinkenburg, W.H.I.M., Peeters, B.W.M.M., Vossen, J.M.H., van Luijtelaar, E.L.J.M., 1991. Absence epilepsy and the level of vigilance in rats of the WAG/Rij strain. Neurosci. Biobehav. Rev. 15, 259—263. Coenen, A.M., van Luijtelaar, E.L., 2003. Genetic animal models for absence epilepsy: a review of the WAG/Rij strain of rats. Behav. Genet. 33, 635—655. Cott, A., Pavloski, R.P., Black, A.H., 1979. Reducing epileptic seizures through operant conditioning of central nervous system activity: procedural variables. Science 203, 73—75. Drinkenburg, W.H.I.M., Schuurmans, M.L.E.J., Coenen, A.M.L., Vossen, J.M.H., van Luijtelaar, G., 2003. Ictal stimulus processing during spike-wave discharges in genetic epileptic rats. Behav. Brain. Res. 143, 141—146. Depaulis, A., van Luijtelaar, E.L.J.M., 2006. Genetic models of absence epilepsy in the rat. In: Pitkänen, A., Schwartzkroin, P.A., Moshe, S.L. (Eds.), Models of Seizures and Epilepsy. Elsevier Academic Press, Amsterdam, pp. 233—248. Elbert, T., Birbaumer, N., Lutzenberger, W., Rockstroh, B., 1979. Biofeedback of slow cortical potentials: self-regulation of central-autonomic patterns. In: Birbaumer, N., Kimmel, H. (Eds.), Biofeedback and Self-regulation. L. Erlbaum Associates, Hillsdale, NJ, pp. 321—337. Fontanini, A., Katz, D.B., 2005. 7 to 12 Hz activity in rat gustatory cortex reflects disengagement from a fluid self-administration task. J. Neurophysiol. 93, 2832—2840. Goode, D.J., Penry, J.K., Dreifuss, F.E., 1970. Effects of paroxysmal spike-wave on continuous visual-motor performance. Epilepsia 11, 241—254. Gottesmann, C., Gandolfo, G., Arnaud, C., Gauthier, P., 1998. The intermediate stage and paradoxical sleep in the rat: influence of three generations of hypnotics. Eur. J. Neurosci. 10, 409—414. Killeen, P.R., Hanson, S.J., Osborne, S.R., 1978. Arousal: its genesis and manifestation as response rate. Psychol. Rev. 85, 571—581.

81

Kirsch, I., 1997. Specifying nonspecifics: psychological mechanisms of placebo effects. In: Harrington, A. (Ed.), The Placebo Effect: An Interdisciplinary Exploration. Harvard University Press, Cambridge, MA, pp. 166—186. Kotchoubey, B., Schneider, D., Schleichert, H., Strehl, U., Uhlmann, C., Blankenhorn, V., Froscher, W., Birbaumer, N., 1996. Selfregulation of slow cortical potentials in epilepsy: a retrial with analysis of influencing factors. Epilepsy Res. 25, 269—276. Kotchoubey, B., Strehl, U., Uhlmann, C., Holzapfel, S., Konig, M., Froscher, W., Blankenhorn, V., Birbaumer, N., 2001. Modification of slow cortical potentials in patients with refractory epilepsy: a controlled outcome study. Epilepsia 42, 406—416. Lantz, D., Sterman, M.B., 1988. Neuropsychological assessment of subjects with uncontrolled epilepsy: effects of EEG biofeedback training. Epilepsia 29, 163—171. Marini, G., Ceccarelli, P., Manica, M., 2007. Characterization of the 7—12 Hz EEG oscillations during immobile waking and REM sleep in behaving rats. Clin. Neurophysiol. 119, 315—320. Meeren, H.K.M., Capellen van Walsum, A.M., van Luijtelaar, G., Coenen, A.M.L., 2001. Auditory evoked potentials from auditory cortex, medial geniculate nucleus, and inferior colliculus during sleep-wake states and spike-wave discharges in the WAG/Rij rat. Brain. Res. 898, 321—331. Mirsky, A.F., Duncan, C.C., Myslobodsky, M.S., 1986. Petit-mal epilepsy: a review and integration of recent information. J. Clin. Neurophysiol. 3, 179—208. Mirsky, A.F., Primac, D.W., Marsan, C.A., Rosvold, H.E., Stevens, J.R., 1960. A comparison of the psychological test performance of patients with focal and nonfocal epilepsy. Exp. Neurol. 2, 75—89. Mirsky, A.F., van Buren, J.M., 1965. On nature of absence in centrencephalic epilepsy: a study of some behavioral electroencephalographic and autonomic factors. Electroencephalogr. Clin. Neurophysiol. 18, 334—348. Monderer, R.S., Harrison, D.M., Haut, S.R., 2002. Neurofeedback and epilepsy. Epilepsy Behav. 3, 214—218. Oller, L.F.-V., 2000. Childhood absence epilepsy and juvenile absence epilepsy. In: Meinardi, H. (Ed.), The Epilepsies. Elsevier Science, Amsterdam, pp. 143—155. Pitman, J., 1993. Probability. Springer, New York. Rescorla, R.A., Wagner, A.R., 1972. A theory of Pavlovian conditioning: variations in the effectiveness of reinforcement and nonreinforcement. In: Black, A.H., Prokasy, W.F. (Eds.), Classical Conditioning II. Appleton-Century-Crofts, New York, pp. 64—99. Robinson, P.F., Gilmore, S.A., 1980. Spontaneous generalized spikewave discharges in the electrocorticograms of albino rats. Brain. Res. 201, 452—458. Rockstroh, B., Elbert, T., Birbaumer, N., Wolf, P., Duchtingroth, A., Reker, M., Daum, I., Lutzenberger, W., Dichgans, J., 1993. Cortical self-regulation in patients with epilepsies. Epilepsy Res. 14, 63—72. Shaw, F.Z., 2004. Is spontaneous high-voltage rhythmic spike discharge in Long Evans rats an absence-like seizure activity? J. Neurophysiol. 91, 63—77. Shaw, F.Z., 2007. 7—12 Hz high-voltage rhythmic spike discharges in rats evaluated by antiepileptic drugs and flicker stimulation. J. Neurophysiol. 97, 238—247. Sitnikova, E., van Luijtelaar, G., 2007. Electroencephalographic characterization of spike-wave discharges in cortex and thalamus in WAG/Rij rats. Epilepsia 48, 2296—2311. Sitnikova, E., van Luijtelaar, G., 2009. Electroencephalographic precursors of spike-wave discharges in a genetic rat model of absence epilepsy: power spectrum and coherence EEG analyses. Epilepsy Res. 84, 159—171. Sterman, M.B., 1972. Studies of EEG biofeedback training in man and cats. In: Hightlights of 17th Annual Conference: VA Cooperative Studies in Mental Health and Behavioral Sciences, Central NP Research Laboratory, Veterans Administration Hospital.

82 Sterman, M.B., 2000. Basic concepts and clinical findings in the treatment of seizure disorders with EEG operant conditioning. Clin. Electroencephalogr. 31, 45—55. Sterman, M.B., Egner, T., 2006. Foundation and practice of neurofeedback for the treatment of epilepsy. Appl. Psychophys. Biof. 31, 21—35. Strehl, U., 2009. Slow cortical potentials neurofeedback. J. Neurotherapy 13, 117—126. Tan, G., Thornby, J., Hammond, D.C., Strehl, U., Canady, B., Arnemann, K., Kaiser, D.A., 2009. Meta-analysis of EEG biofeedback in treating epilepsy. Clin. EEG Neurosci. 40, 173—179. van Luijtelaar, G., Sarkisova, K., Midzyanovskaya, I., Tolmacheva, E.A., 2007. Stress vulnerability and depressive symptoms in genetic absence epileptic rats. In: Hollaway, K.J. (Ed.), New Research on Epilepsy and Behavior. Nova Science Publishers, New York, pp. 211—278. van Luijtelaar, E.L.J.M., van der Werf, S.J., Vossen, J.M.H., Coenen, A.M.L., 1991a. Arousal, performance and absence seizures in rats. Electroencephalogr. Clin. Neurophysiol. 79, 430—434.

L. Osterhagen et al. van Luijtelaar, E.L.J.M., de Bruijn, S.F.T.M., Declerck, A.C., Renier, W.O., Vossen, J.M.H., Coenen, A.M.L., 1991b. Disturbances in time estimation during absence seizures in children. Epilepsy Res. 9, 148—153. Vergnes, M., Marescaux, C., Micheletti, G., Reis, J., Depaulis, A., Rumbach, L., Warter, J.M., 1982. Spontaneous paroxysmal electroclinical patterns in rat: a model of generalized non-convulsive epilepsy. Neurosci. Lett. 33, 97—101. Voipio, J., Tallgren, P., Heinonen, E., Vanhatalo, S., Kaila, K., 2003. Millivolt-scale DC shifts in the human scalp EEG: evidence for a nonneuronal generator. J. Neurophysiol. 89, 2208—2214. Wiest, M.C., Nicolelis, M.A., 2003. Behavioral detection of tactile stimuli during 7—12 Hz cortical oscillations in awake rats. Nat. Neurosci. 6, 913—914. Wyrwicka, W., Sterman, M.B., 1968. Instrumental conditioning of sensorimotor cortex EEG spindles in waking cat. Physiol. Behav. 3, 703—707.