Newswise — Bidirectional links between epilepsy and sleep have been known for thousands of years. Despite nearly a century of research using EEG investigations, the relationships are still not well understood.

Dr. Laurent Sheybani interviewed Dr. Birgit Frauscher, an epileptologist and sleep specialist at the Montreal Neurological Institute. They discussed what is known about sleep patterns, how sleep may influence epileptic activity and vice versa, and how sleep research may lead to more effective epilepsy treatments.

Listen below or download the episode.

Find Sharp Waves episodes on SpotifyApple PodcastsGoogle PodcastsAmazon MusiciHeart Radio, and Stitcher.


Sharp Waves episodes are meant for informational purposes only, and not as clinical or medical advice.

Podcast Transcript

Dr. Birgit Frauscher: I’m Birgit Frauscher. I am an epileptologist and I’m also board certified in sleep. I’m a professor of neurology at the Montreal Neurological Institute (MNI) and Hospital, and group leader of epilepsy and director of the neurophysiology lab at the MNI.

Dr. Laurent Sheybani: Very nice to meet you, thank you very much. So you are a clinician and also a scientist with a dual interest in epilepsy and in sleep. Why did you choose these two fields of interest? Why did you think they were important?

Frauscher: Well, they have something in common and that is the EEG. That’s basically the reason why I became interested first in sleep, and then later on in epilepsy, because I really think the most interesting EEG patterns are found in the context of the disease epilepsy.

Sheybani: So you started actually by being interested in sleep and not epilepsy; I thought it was the other way around.

Frauscher: No. During medical school, during one of the internships actually I chose neurology, and I was at that time with a neurologist who was actually the head of the sleep lab, so then I saw all the tracings. And what I find fascinating in EEG in general is the objectivity of this measure and also that you can directly assess function of the human brain. I started out with sleep and during residency I became very interested in EEG and epilepsy, and so I then later on pursued a career in both.

Sheybani: So besides the fact that sleep and epilepsy share the same investigation method, EEG, why do you think, if you think so, why do you think epileptologists should be interested in sleep and should take care of investigating sleep for their patients?

Frauscher: The reason is that many of our patients with epilepsy have also sleep disorders, right. And if we look at patients with drug-resistant epilepsy and that’s most of the patients I deal with, it’s up to 50%. And we know that there are these reciprocal interactions with sleep disorders, such as, for example, sleep apnea aggravating epilepsy and also epilepsy can aggravate sleep disorders, such as aggravating insomnia, et cetera. And then we have this overlapping window between sleep-related hypermotor epilepsy and parasomnia, where we have this fascinating differential diagnostic spectrum. So I’m not saying that every epileptologist needs to be an expert in sleep medicine, but I still think one should have a good knowledge, and then liaise with colleagues from sleep medicine if one can’t do everything by oneself.

Sheybani: So sleep is obviously very difficult to define, but there are markers in the EEG that can help to know whether someone is sleeping or not. Can you tell us a bit more about these EEG markers?

Frauscher: It’s all defined with the criteria of the American Academy of Sleep Medicine. Sleep is coded in 30-second epochs, so basically we differentiate grossly between wake and non-REM sleep, subdivided into N1, N2 and N3 sleep, and N3 is the previous deep sleep, and then REM sleep. And so basically all of these stages are defined so that we can like, with boxes, dedicate one of the boxes depending on what we see. So for example, classical hallmarks of N1 sleep would be the so-called vertex waves. For sleep stage N2 it is K complexes or sleep spindles; I can talk about the definitions later if of interest. For the N3 sleep it is defined as there is a minimum of 20% delta activity present, and then we have REM sleep, where we have low-amplitude, desynchronized EEG patterns associated with rapid eye movements and muscle atonia. 

Again, what is important, maybe if you come from EEG, is that for the scoring of sleep, we need additional channels. Usually what we place is EOG (electro-oculograms). That being said, it can also be very well estimated from some of our EEG channels like F7, F8, but then we also add EMG (electromyography) leads on the chin to assess muscle tone, which is one of the hallmarks when we are in rapid eye movement.

Sheybani: REM sleep, non-REM sleep, non-REM sleep stages 1, 2, 3 – do they occur randomly during sleep or is there a certain periodicity, a certain rhythm?

Frauscher: Yes, it’s a good question, thank you for asking. Basically, what everybody does is undergo several of the so-called sleep cycles per night. Usually, it’s four or five. We have more non-REM sleep in the first part of the night and then more REM sleep in the second part of the night. That is also explanatory for some of the diseases occurring in sleep, for example sleep walking, which usually occurs mostly during non-REM sleep, N2 N3, it’s something you more frequently see in the first half of the night, whereas REM sleep behavior disorder, REM sleep parasomnia, occurs more during the second half of the night. So that’s why it’s also important to ask for the timing.

And seizures usually occur out of the not-consolidated sleep stages, N1, N2, frequently.

Sheybani: Spindles, and slow oscillations, those are two main, say, EEG markers of sleep? Can you tell us a bit more about these?

Frauscher: Yes. So let’s start with sleep spindles. Sleep spindles, there is an electrophysiologic definition but more important there is a function attributed to them. Usually, they are more than 0.5 seconds in duration and have a frequency between 10 and 16 Hertz (Hz). Then basically we have slow sleep spindles and fast sleep spindles, and they differ also in how they are distributed across the brain. For example, the slow spindles are more seen in the frontal area and the fast spindles in the central parietal area. Sleep spindles are super important for our functioning; they have a role in brain maturation but also for memory consolidation and various other cognitive function. There is quite some active research going on in the sleep community, where it was shown that these spindles are really important for sleep-related memory consolidation; there is most data on that. 

Slow oscillations are also a very important rhythm, because basically they decide these up and down states, which occur periodically and modulate physiological rhythms such as sleep spindles, for example, also higher frequencies such as ripples above 80 Hz, gamma oscillations, they usually occur in the up state and are regulated like this. And then also if we think of pathology, different patterns are attributed to that as well.

Sheybani: Among the many functions of spindles, they are also involved in brain maturation. Does that mean that babies do not have spindles and they start to express them later, or are they born with spindles? How does it work?

Frauscher: I think they start a couple of months after being born, and what I remember very strikingly, if you look at EEGs of 6-, 8-, 9-month-old children, they are really of very long duration. It’s not unusual that they can be up to 3 seconds in duration. Evidently when you become an adult, they become shorter in duration but they’re still very important and I think there’s really good evidence on studies where they do memory testing before the night and then after the night and they could see that more spindles were associated with better memory consolation than lower rates.  That brings us to diseases where we don’t see them, so for example if you think of patients with ESES (electrical status epilepticus in sleep) for example, honestly at least I can’t differentiate which stage of non-REM sleep they are on, right, because it’s full of epileptic activity, and by definition that’s more than 85%. Very often you don’t see classical sleep features. And that might among other factors be related to the cognitive dysfunction they have, so this is something which is very important and a widespread phenomenon.

Sheybani: Do spindles or slow oscillations or both interact with epileptic activities?

Frauscher: They do. There is much work out – I want to apologize if I now focus a lot on what our group did, because evidently a lot of other people did the same, which is reassuring that there is a trend behind for example, regarding sleep spindles, I think what is very interesting is for example, and that is work in children with what is now called self-limited childhood epilepsy, the centrotemporal spikes, what was previously called Rolandic epilepsy. It was very nicely shown there was an inverse relationship between sleep spindles and the occurrence of spikes, and this was particularly seen in this centrotemporal region, and they could find various dysfunctions on memory tasks. That is, for example, one work. What we did and that’s a study on intracranial EEG a couple of years ago, published in Sleep, is that we found an inverse relationship between sleep spindles in the hippocampus and spiking. So in other words, the more spikes there were, the lower the spindle rates were. 

Recently, one of my now post docs, Katharina Schiller, she did a study with high-density EEG in adults with various epilepsy types. At the sub-lobar location of the epileptic focus there were significantly reduced spindle rates compared with the rest of the brain, and she could also find some correlation with memory function.

So with the second half of the question with the slow waves, that is also something I think first notion for that is from Terzano and Parrino, from a cyclic alternating pattern where they found some associations, but then what we did in a study in intracranial EEG published in Brain in 2015, is we looked at the coupling of epileptic activity to these slow oscillations. And when we went into the study, we knew already very well from animal work that in this so-called “up” stage of the slow waves, in the active stage, right, we have physiological rhythms. So we hypothesized that is the same, that also the spikes are coupled there. But then in the end, that’s not what we found – it’s at the transition, and the epileptic activity, if you think of ripples which can be both physiological and pathological, that they are at the opposite transition then, that indeed, it really could even help us to disentangle between both. But what we also found, which was interesting, is that if we looked at the highest amplitude slow waves we found they went along with highest spiking waves. 

So that is all work in sleep that was then replicated with slightly differing methods by our group but also others, for example the group of Eishi Asano and then, and that is work from yours also, that you found this coupling to slow waves outside of the sleep context if I remember. So I think there is now also a vast literature of decoupling so that we not only speak about, and I think this is an interesting change, that we come away from these big boxes that we use for classification but now really look at microstructure, associations such as sleep spindles, slow waves, if we want to go for REM we could also look at the basic REM where we have the lowest rates of epilepsy activity as opposed to the tonic REM, et cetera et cetera. So it’s really a fascinating topic.

Sheybani: So maybe to follow up on the last point you raised, the point of that REM sleep is kind of resistant to epileptic activity. Do we know why?

Frauscher: Well, I think there is good evidence and it might be really at the end the importance of synchronization versus desynchronization, right. Because we know that particularly in phasic REM sleep, our group and others showed that, and before that in animal models it was nicely shown that phasic REM sleep is the most desynchronized phase, even more so than in tonic REM sleep. So this is one mechanism – the desynchronization seems to be protective, and what are the underlying molecular mechanisms? We have to think of neurotransmitters.

Sheybani: You talked about these two papers showing that the more spikes we have, the less spindles we have. Does that fit with a theory that has led to the publication of several papers regarding the hypothesis that epileptic discharges might kind of hijack the brain circuits necessary for the expression of sleep patterns?

Frauscher: I think it does, and actually I think there was some fundamentally important work done here at MNI by Pierre Gloor and others. We know, right, for spindles the thalamocortical circuit is very important and we know this is the main driver, and then depending on the state of the cortex we have a variable expression of spindles as we could show in a paper in Neuroimage in 2015. And so basically similar circuits are also responsible, if we think of the classical 3-per-second spike and wave discharges, and so that theory is that this physiological activity is displaced by these pathological mechanisms and that is one way to explain it. And it’s a fascinating model which could indeed be the case and be suggested from this research.

Sheybani: So if you talk about the 3 Hz spike and wave you’re mainly speaking about generalized epilepsy, right, and absence epilepsy in particular. But is it also true for focal epilepsies, this hijack of the brain circuits by epileptic discharges?

Frauscher: It’s a theory. It’s a theory, right, which is basically adopted from this classical 3 Hz spike and wave generalized model where Pierre Gloor and others worked on it. But it’s still a kind of a theory. Now evidently another explanation might also be, you know, if there is this pathological activity or the cortex is not functioning correctly then it’s also expected to have less physiologic function. That is something I think that we should look more into, because what I see in epileptology, what happens is, and that I see with all the trainees always, you are very much focused on where are the spikes, where are the seizures. But maybe we should broaden a little bit our view of things and also pay attention to okay, what is not normal in the background, what does that tell us in addition? So I think we should not only hop on the epileptic but also look at the reduction of physiologic phenomena.

Sheybani: Agreed. I think that’s also an important point. When we think about sleep, we generally think that this is a general phenomenon, right, we are either sleeping or we are awake. Is it the way that the brain sleeps? Does it sleep homogeneously, or is there another mechanism?

Frauscher: Yeah that’s a great question. And really I mean if you would have asked people 20, 30 years ago everybody would have agreed and said, “Well, sleep is a global phenomenon.” We know that some species, like the dolphin, they sleep with one hemisphere and are awake on the other, that’s a different phenomenon but we would have assumed for humans that it’s global. But now with all the possibilities that we also have, and the unique opportunity offered by invasive intracranial EEG, there is more evidence that this is indeed not the case. That there is asynchrony, and I remember beautiful work from Lino Nobili, where he could show that, for example, we are really deeply asleep, right, and then we see this awake-like activity in the motor cortex. Or for example, the scalp EEG shows wakefulness minutes before we see spindles in the hippocampus. Our group started in the last 10 years to work on that topic. Yes, there is a global mechanism, then depending on which state the various cortical region is in, you have a variable expression of spindles. Slow waves, there is also excellent experimental work done which showed that it’s a use-dependent principle, right? If a region was very much used during the day, you have more increase in slow waves in that region. There is fascinating work out going in that direction and that is also why I think that yes, we do invasive intracranial EEG to find a surgical solution for our drug-resistant epilepsy patients, but as a byproduct, these recordings can be used to answer interesting sleep questions, because it’s the only method where basically we can have chronic recordings of various regions over several days.

Sheybani: You have recently published a paper using intracranial EEG where you managed to score sleep, indicating in which stage of sleep we are, REM sleep, non-REM sleep, et cetera, and you did that with intracranial EEG. Given that sleep is not a homogenous system, did you find differences between the intracranial electrodes? Meaning did you find that some were in REM sleep and others were in non-REM sleep, for example?

Frauscher: That’s a very interesting topic. Let me talk a little bit about this work and basically it’s attributed to Nicolas von Ellenrieder, who is the engineer behind Sleep-SEEG, which is by the way publicly open, available, and I saw already first groups using it and a first paper coming out actually using it, so I’m very happy that people find that it’s useful. 

These patients have invasive intracranial EEG and very often it’s not so trivial to get the scalp EEG and then to add EOG and EMG, you know, albeit it’s noninvasive, these patients are already bothered and it might be complicated to get. And so the first question is, is it possible to avoid this and just do the sleep scoring reasonably well on what we have for intracranial EEG? And the second question is, would it be even possible if we disrespect the presence or absence of spikes, or is it needed as a first step to do a spike detection and then eliminate contacts with high-spiking waves? 

If you look at the paper, there were depending on the region always various features which performed better than others, which also points to the fact that the brain is not sleeping evenly and at the scalp, you know to me scalp EEG is a blur, it’s a mixture of everything going on. But if you’re very focally focused you see different patterns. For example I think the hippocampus is for example very fascinating. And so basically even during wakefulness you have slow wave activity, right, which is totally normal, which wouldn’t be normal if I would look at the parietal neocortex. And then if you look at sleep spindle rates, they’re significantly lower in the hippocampus than outside that structure. And there are also some regions where sleep is not very well expressed. This is often channels located in the temporal neocortex. Then if we look in the occipital lobe, it’s also differently organized. 

So in a way, this is the beauty I think of the tool that we developed, first of all we were very excited to see that really it does, you don’t have to exclude channels with spiking. It performed as well. So that already saves you quite some time and the next thing was that we were very happy with the results and it really performs I think reasonably well for N2, N3 and REM sleep, not so much for N1, but again that’s not a surprise because if we think about this it’s some kind of transition stage, and so less interesting for sleep research. 

The fundamental question, what I think in this regard, is it really correct to use the same scoring criteria we use in the scalp at the various brain regions? And I would doubt it, because what we could see when we developed Sleep-SEEG is that depending on the anatomy, there are various signatures occurring during the various stages of what we see in sleep. So this to me is one of the questions and I don’t have the final answer to it but I’m wondering if we think of that recent paper, published in PNAS, from Guthrie et al., right, where they looked at what happens in the hippocampus during sleep and the neocortex and they found that they only were in the same stage for about one-third of the night. What they did is actually they used scoring criteria that we use for our regular sleep scoring in the hippocampus, and that comes to the fundamental question – is that a good thing to do? I have not the final answer; the only thing that I can say is that different regions have different signatures in sleep. 

We just have new data from our group, it’s not published yet, it’s not submitted yet, but hopefully will happen soon, where we looked at the transitions between non-REM to REM sleep and we also found that this happened not abruptly but progressively and that there is a posterior to anterior gradient.

Sheybani: That’s interesting that means it happens progressively in spatial terms and in temporal terms.

Frauscher: Yes.

Sheybani: That’s interesting and we look forward to the papers then. Do you have anything that you would like to add, general conclusion or maybe the next three tough questions in sleep science or epileptology or anything you want to finish with?

Frauscher: Well first of all I would like to thank you for having me and giving me the opportunity to talk about what I‘m passionate about, sleep and epilepsy. And I think for the next question to be answered is really, I’m a physician and so you want to treat your patients. And so the next fascinating question could be, and that is also work we didn’t get to today, but we and others have shown that unstable sleep goes along with higher rates of epileptic activity. So one question is, could stabilization of sleep also contribute to better control of epileptic activity. So this is one fascinating avenue we could go into. 

And then now we have our automatic scoring tool to score sleep, there are many open questions we can try to answer because now it’s really easy to analyze really the whole recording durations – two weeks of patients and find out more about influences of prolonged spikes. This is also something I think could be fascinating to do with devices, with RNS (responsive neurostimulators) – unfortunately at the moment it’s not the whole recording that’s stored in the devices but for sure this is something very interesting to look into seizure prediction, because that could be very useful for our patients to know. 

Studies mentioned in the interview:

Interictal Hippocampal Spiking Influences the Occurrence of Hippocampal Sleep Spindles (Frauscher et al., 2015, Sleep)

Focal epilepsy disrupts spindle structure and function (Schiller et al., 2022, Nature Scientific Reports)

Origin and significance of the cyclic alternating pattern: Review article (Terzano & Parrino, 2000, Sleep Medicine Reviews)

SleepSEEG: automatic sleep scoring using intracranial EEG recordings only (von Ellenrieder et al., 2022, J. Neural Eng.)

Recurrent Hippocampo-neocortical sleep-state divergence in humans (Guthrie et al., 2022, PNAS)

Scalp spindles are associated with widespread intracranial activity with unexpectedly low synchrony (Frauscher et al., 2015, Neuroimage)