Our combined EEG-behavioral study with Wistar rats chronically treated with the CB1 antagonist SLV326 showed two major findings. Firstly, the behavioral spasms, which were observed during daily clinical observations, were identified as generalized convulsive seizures. The aberrant activity in the EEG showing generalized, high amplitude sharp spiky activity, and in some cases post-ictal depression, was concomitantly accompanied by severe clinical symptoms. Therefore it can be concluded that these muscle spasms are indeed epileptic phenomena.

The second major finding was that spectral changes were found, specifically during active behavior, suggesting that locomotor activity or sensory motor processing may be affected.

The convulsive seizures were present in 6 out of 24 treated animals (25%). All seizures were scored 4 or 5 on Racine’s scale. The violent movements (or muscle spasms as the biotechnicians really described it like that) were rhythmic movements which developed in time and represented well-ordered consequence of motor manifestations. The complex behavioral pattern is known to be specific for epileptic seizures generated by limbic structures in rodents.

Additionally, interictal spikes were found in more SLV treated animals (37.5%). Interictal spikes can serve as a sensitive marker for epilepsy [22, 30] or epileptogenesis [31, 32]. Therefore it is likely that the number of animals with seizures is actually larger than 25%. The spikes found in the treated animals were all generalized, whereas the spikes in the single control animal were localized. Epileptic spikes are expected to be found in close proximity to the epileptic focus. This might indicate the spikes in the control animal to be local cortical changes due to electrode. In view of the limbic nature of the generalized SLV326 seizures, a cortical focus is not likely, but cortical electrodes can register high limbic activity. However, much is still unclear about the relationship between seizures and spikes [33], and conflicting results have been described in literature [34].

The presumption that more animals than the 25% we observed might have seizures is also dictated by behavioral observations made during the regulatory repeated dose study that preceded the present EEG study. The animals’ behavior was monitored during daily cage observations; aberrant behavior, described as muscle spasms, was observed in 70% of the animals of the treatment group (totally 39 animals out of 60) during the course of the treatment period (personal communication). Given the short period of EEG recordings (24 hours) in the present study, it is likely that seizures would be observed in more animals over a longer time frame. Moreover, the animals with EEG-behavioral seizures consisted of an equal amount of animals with and without previously reported “spasms”. We expect that the actual number of animals with seizures might be higher than 25%, or even 70%.

The ‘hands-off’ approach during the EEG recordings excludes the possibility that external influences, such as handling stress, were culprit, as was suggested for rimonabant by the European Medicines Agency; the seizures that we noticed did occur in their home cages and occurred spontaneously. The animals were freely moving and were habituated and adapted to the EEG leads and EEG recording home cages. Moreover, the seizures were evenly distributed over the 24 hours of recording, and no relation with the time of gavage was found. Therefore, handling can be excluded as a factor.

This study confirms the findings of our previous study, using the CB1 antagonist rimonabant [6]; this antagonist induced de novo limbic seizures in healthy Wistar rats as well. These findings suggest that the EEG changes that were observed in both studies are related to CB1 antagonism, rather than being compound-specific. Data from Vinogradova, Shatskova [35], showed that CB1 antagonists augment the spreading of brainstem seizures, induced by audiogenic stimulation, to limbic areas, signifying that a proper functioning of the endocannabinoid system in limbic areas is of importance in the protection against seizures.

Results from animal models for brain injuries suggest a seizure protective role of the endocannabinoid system as well. In a model of chronic brain injury, viral encephalopathy, rimonabant induced spontaneous seizures [36]. Also, rimonabant increases the severity of pilocarpine-induced seizures [16, 37]. Moreover, impairment of endocannabinoid synthesis increased the seizure susceptibility [38]. G protein Gq/G11 knockout mice, which have impaired endocannabinoid synthesis, showed spontaneous epileptic seizures. It is notable that both the frequency of seizures in each animal and the number of affected animals increased with age [38].

In all, the present study strengthens our previous proposal that the endogenous cannabinoid system protects against seizures. Indeed, cocaine-induced seizures in mice were inhibited by enhancing the endocannabinoid system [39]. These inhibitory effects were not observed in the presence of a CB1 antagonist.

Since the limbic system is especially vulnerable to injuries [40], it is not surprising that limbic seizures were observed in our animals, including hippocampal seizure activity. In humans, age-dependent seizures are often of a limbic nature [41, 42]. Apart from the epileptic activity induced by SLV326, changes in spectral content of background EEG were observed. When active, SLV326 treated animals display a lowering of the theta peak frequency and a decrease in relative gamma power. Because the spectra were normalized, a decrease in gamma power coincides with an increase of the power in other (low) frequency bands.

It is interesting to speculate what this means in terms of changes in the absolute spectrum. Our results suggest that CB1 antagonists induce a reduction of gamma power rather than an increase in the lower end of the frequency spectrum. Indeed, the absolute power spectra of our data (not shown) seem to confirm this, although no significant differences were found due to the high variance between animals.

So far only reports on the effects of agonists on the spectral power have been published. In humans a CB1 agonist causes a decrease in the EEG power spectrum between 7 and 27 Hz [23]. Effects of cannabinoids on the gamma band have been described in animals [24] as well as in humans. Nottage, Stone [43] report a THC-induced increase in the magnitude of gamma activity. The decrease in gamma activity we found, after exposure to an antagonist, is in line with this. However, CB1 agonists have also been found to reduce gamma power [44], although in this case repeated exposure to cannabis blunted the response. When taking into consideration that the gamma band is typically associated with higher cognitive functions, it is not surprising that the complete picture is more complex. Indeed, effects of cannabis on gamma activity have been found to be dependent on dose [45] and time after administration [43]. Interestingly, the effects on the EEG spectrum were state-dependent, as they were only present when the animals were highly active. Indeed, theta rhythms are associated with locomotor activity [46]. Nevertheless, we did not find any changes in the quantity of locomotor behavior (S1 Fig; p = 0.5). Theta rhythm is shown to be a marker of complex behavior [47]. It is likely that the behavioral effects caused by SLV326 are of a more complex nature than, and therefore do not show up in the PIR, as CB1 receptors are known to be involved in, among others, compulsion and food intake behavior, all of which can be observed predominantly during active wakefulness [48]. Sprague-Dawley rats treated with CB1 antagonist AM-251 during adolescence showed more active stress coping behavior and greater risk assessment during adulthood, but no changes in locomotor activity [49]. Moreover, in several studies using animal models for cognitive impairment CB1 antagonists have been shown to improve cognition [50–54].

Whether and how the epilepsy and the spectral changes are connected remains to be investigated. In humans with epilepsy both differences in absolute and relative power of the frequency spectrum have been found [55]. In children, epilepsy was found to be highly correlated with changes in the alpha band [56]. A study from Maheshwari, Marks [57] suggests that relative gamma power may serve as a biomarker of AED efficacy in absence epilepsy; drugs that reduce interictal gamma power increase the incidence of absence seizures. This might imply that the observed deficit in gamma power in SLV-treated rats can be ascribed to enhanced susceptibility to seizures.

In summary, blocking of the endogenous cannabinoid system makes the healthy rat brain more vulnerable to epilepsy. Additionally, it alters the frequency spectrum of the EEG when rats are highly active, suggesting effects on complex sensory motor functions.

The role of the endocannabinoid system is still far from clear, but its properties make it very suited for protection from excessive oscillatory activity. A better understanding of the epileptogenesis induced by removing or blocking this protection system may contribute to a better understanding of the endogenous cannabinoid system altogether.