S4E1 - Niels Niethard - Neural mechanisms underpinning sleep

Speaker 1 Hello and welcome to the Sleep Science Podcast. I'm Penny Lewis, a neuroscientist specializing in sleep and memory, and the presenter of this show. In the podcast, we talk about all things related to sleep, from dreaming and sleepwalking to what sleep does for our brain and body and how we can get more out of our sleep. Today we're going to take a much closer look at cellular level cortical circuitry during sleep. Our guest is Doctor Neils Netard, a research group leader from the University of Tubagen in Germany. Neils is an expert on cortical circuitry and sleep and has published A revolutionary theory of how cellular dynamics facilitate consolidation during specific times in non REM sleep. Neils, welcome to the show. Speaker 2 Hey Penny, thanks for having me. Speaker 1 As we've said in our prior discussions, this is a tricky topic for me. You're going to have to teach me from the very beginning. I wonder if we could start by talking about the circuits that we're dealing with. So tell us a bit about these cortical circuits. Which cell types are we talking about and how are they connected and what do they do? Speaker 2 Yeah, that's actually a great question because it is very complex when we look on a cellular level in our brains. But I think there is a very fundamental aspect that is important to keep in mind. So we have neurons that project to each other and then they start firing and that has an effect on the other cell. But what is very important, we have two major types and one of them is excitatory cells. So if they fire, they also induce higher firing rates in the other cell which they are projecting to. And then we have inhibitory cells which do exactly the opposite. So if they get more active, they reduce the activity of the cell that they are projecting to. And I think this is a very essential aspect that we have to keep in mind. And this is fundamentally the kind of the main question that I'm interested in. How is the interaction between excitation and inhibition affected by sleep, for example? And of course we have a lot of different excitatory and inhibitorial cells in our brains, but this is the most fundamental differentiation that we can look at on a neuronal level. Speaker 1 And you are specifically focused on the cortical circuitry, is that right? Is it a particular layer of cortex? Speaker 2 I mean, we typically record cells from layer 2/3 in the cortex. However, we actually also go a little deeper nowadays and we even reach down to the hypothalamus. So this is probably the deepest that you can go, but it's more because the accessibility is much easier in superficial layers than deep layers, because then you can actually image them, for example, without removing tissue, which you have to if you want to go deeper. Speaker 1 And this is all in rats? Or is some of it in mice as well? Speaker 2 In most of it, it's done in mice and that is a simple reason because we need to have access to genetic tools and there are much more available in mice than rats. So for example, if you want to investigate different interneuron types, then we have to do that in mice because there's actually no transgenic line and rats commercially available that would allow us to look at a specific subset of inhibitory cells in the cortex or any other brain area in the rat. Speaker 1 And do you also look in the hippocampus or is it, is it mainly in the cortex and the thalamus? Speaker 2 We do look also in the hippocampus. This is not published yet, but we have several projects running where we also now look into the hippocampus, in particular, of course related to all the memory process that we know heavily rely on hippocampal activity during sleep. Speaker 1 OK. And in in terms of the cortical areas, is it a particular area you tend to focus on or does that depend on the question? Speaker 2 So the things that we did so far were mainly somatosensory cortex. Again, this was also because of its accessibility and we wanted to very basically characterize cortical activity during sleep because that wasn't done before. And now we shift a bit more into the topic to relate this to behaviour. So for example, we are now focusing on the retro spinal and prefrontal cortical areas to see whether we can sort of say track memory N grams over time and how they are affected by changes in excitation and inhibition on a local but also on kind of systems level. Which also brings into play the recordings in the hippocampus, for example, to see the dialogue between hippocampus and neocortical cells during sleep, but also during wakefulness, of course. Speaker 1 OK, so there are lots of different circuits that you can look at basically. Speaker 2 Essentially with the tools that we have, we can look more or less everywhere. It's just a bit more demanding for the surgery and also the processes that you have to undertake for actually developing the different approaches. But but essentially you can, you can look everywhere with pros and cons that come along with that, yeah. Speaker 1 Yeah. OK. So let's talk about the actual cells. I have one of the diagrams, one of the most famous diagrams from your work in front of me with the the pyramidal cells and the somatostatin cells and the parvo parvo cellular neurons. So maybe you can just talk a bit about those different cell types and why they're important and the significance of them for for this work. Speaker 2 Yes. So when we look at the more fine grained scale and we look at excitatory and inhibitory cells, then it's of course very important to see how they actually affect the different cells. So when we look at inhibitory cells and how they project to excitatory neurons, then we see that there are different types of interneurons like the inhibitory cells. And on one side, like very basic differentiation is done on genetic markers like somalistatin and power mean. But also we can see that they target different compartments of the excitatory cells. And in particular, there's some other statin. Positive cells target the apical dendrite, like the area where the cell receives input, while the power mean positive cell targets the soma. So the orias action or initial segment, which is actually the area where the cell or to say sends out the signals. And by that they have different possibility to actually act on the network level. So there's some other 13 positive cells are, for example, very powerful in modulating arriving activity, while the power mean positive cell is very powerful in optimizing the timing of the output of the cell. And by that they have different roles and modulating activity levels in the network. Speaker 1 OK, that makes a lot of sense. So the pyramidal cells are the big excitatory cells and are they? And I'm going to show my ignorance here. Are they in layer 3? Speaker 2 They spread across all layers. So you have pyramidal cells in layer 23, you have them down to layer five and six. They are different in size and they are also probably different in their like overall functioning. And what they the role for memory for example, is their ideas that different aspects of memory are encoded in different cortical layers. And of course also the distribution of interneurons is not equal across the different layers. But what is very important that if you want to target the apical dendrite of the cell, most of the apical dendrites are actually in the superficial layers. So then the most of the cells that directly target them from the interneuron side, like the somatostatine cells, they are also mainly located in the superficial layers of the cortex. But again, this is probably also a little different between different brain areas. But I think it's essential to keep in mind that the majority of cells is excitatory. So it's roughly like 70% of the human brain. And then mice it's actually even even more. And then you have a sub group of inhibitory cells that are particularly important for modulating dendritic activity and you have subgroups of cells that are particularly important for modulating the activity in relation to what actually is the output of the excited neurons. And there are other types of inter neurons that actually mediate between them which we didn't touch upon. But there are also inhibitory neurons that project to other inhibitory neurons. And that's in particular interesting because through that inhibition you actually increase the excitability of the network. And this is, so to say kind of the very simplified picture of the cortical networks that. Speaker 1 Look at yeah, OK, So I think we need to keep it simplified for now. But of course, in the prior episode we looked at glia and you know how those support and modulate these neurons as well. So people listening might be aware of that too. OK, so we've got these big pyramidal cells that are excitatory and they're the kind of output cells. And that's really interesting that you point out that the somatosatin cells project onto their dendrites. And so they are influencing the input into those cells and particularly because they're inhibitory, they are able to inhibit input. So I guess they can prevent those cells from getting information about what's going on that they otherwise would have. So they can give them a rest essentially. Speaker 2 Yes, they can do that, but I think what is very important to keep in mind this is not an on and off scheme. So what is very important is that those cells are very locally manipulating the activity on the dendrite. So if you have an inhibitory input to the dendrite, typically it's not affecting the entire dendritic compartment. That really shuts down the arriving information to that cell. But instead you kind of apply a filter that you have local signals that can pass and others that are suppressed. And that's a very powerful tool if you want to be selective in your processing. So I think in that regard, you can think of instead of having the neuron as the computational unit, we need to look at on a finer scale, like the synapse is the computational unit, which is between like excited or not excited, like 0 or one thinking in binary terms. And the inhibitory input can manipulate which synapses are actually passing the threshold and which ones are not. Or a piece of the dendritic branch like a short segment, which is actually excited and which ones are not. I think this is very important to keep in mind. So typically those do not act on a global level. There are exceptions probably, but I think that's very important to keep in mind. Speaker 1 And would a individual somatostatin cell be selectively inhibiting a subset of the synapses onto a particular pyramidal cell? Or maybe it would be do they project to more than one pyramidal cell so it might choose to inhibit 1 pyramidal cell and not another where IT projects? Or would it just inhibit everything that IT projects to at the same to the same extent? Speaker 2 So first of all, they span across different areas, especially some other study neurons typically span between like white distances in the cortex. So they can even connect different columns and even different brain areas so to say, which puts them in a very interesting position for mediating on a systems level. So you can manipulate activity between cells that are like quite far from each other. And of course, if the cell becomes active, then it sends out its inhibitory input to everything where it's projecting to. But of course, the strength is not always the same. It's a synapse. So depending on the strength on the synapse and the amount of synapses that are projecting to the other side, you have a stronger or weaker input. And I think it's one way to think like the active inhibition that occurs in a moment. But you can also have like a kind of tonic inhibition and you selectively release inhibition in a subset that has the same effect that you have like a very dedicated modulation of the input that arise to that say, because you're kind of inhibiting most of your network. But then you release it selectively and by that you increase the excitability of that particular excitatory neuron where you release the inhibition. So I think both is possible and both is important for our computations that are ongoing in our brain. Speaker 1 OK, that makes a lot of sense. And of course, you point out that it depends on the strength of the synapse and the synapses are all different. So I think I'm forming a nice picture of the somatostatin cells. Should we talk about the parvo cells? I'm going to call them parvo cells for short. Speaker 2 OK, Yeah, sure. No, I think it's very interesting to look at both the power mean positive cells are very important for mediating the output of the cell. And why I think this is particularly important because we know that timing matters. So if we want to have plastic changes occurring in our networks, in our brains, like we want to encode information, we want to store memories, then we need to have cells that are activated simultaneously more or less or in a very, very short time window that we actually have long term prudentiation occurring. And if you want to optimize the conditions that two cells become active at the same time, then having synchronous release and inhibition is very powerful. So imagine that you have pavamine positive cells being active during a certain period and then you synchronously released inhibition. So the cells that are affected will be more likely to actually affect other cells exactly in the same time point. And by that they might form synapses together. And I think this is fundamental principle that we know is for example also important for memory encoding. So we see that the interaction between excitation inhibition and particular pavamine positive and some other 13 positive cells are important for memory encoding processes and memory concentration processes. So if you would block that, then you interfere with all those plasticity processes that we need to have a stable storage of information in the network. Speaker 1 So presumably they interact in wake as well as in sleep. So maybe we should start by talking about their interactions in wake and then we can move into the more complex interactions in sleep. Or how would you do it? Speaker 2 Yeah, sure. So I think talking about memory for example, you need to have the memory being encoded so that sleep can do anything afterwards. So looking from the perspective from wakefulness absolutely makes sense. I think what is important to keep in mind that we are like on a very simplified level here and there are a lot of different modulators that affect those cells differentially. Like for example, all the neuromodulators act on different scales and they might be different in wake and sleep and in particular when we talk REM sleep and non REM sleep. But on a general level, I think what is very important that during wakefulness inhibitory activity is extremely important because it guides the activity and keeps it in a stable level. So you have this immediate feedback. When you have an overshooting of your excitation, then there is an inhibitory input that directly kind of keeps this in a balance. And I think this is extremely important because if this doesn't work, you will have excitation crossing thresholds which are reaching pathological conditions. So this is a very, very important mechanism in our brains. And at the same time, as I said, the rule for example, for forming plastic changes like encoding information, they are essentially independent of whether you want to use this for a consolidation process during sleep or whether you want to use that during the encoding. Also during encoding, you need to make sure that you have a selective pattern that's activated. And if you want to have selectivity, as I told you before, local inhibition is key for this mechanism because you prevent the excitation from like just spreading randomly in your network. But you have a guidance which is actually based on the interaction between excitation and inhibition. Speaker 1 And do these parvo cells also project to many different pyramidal cells across many layers in the way this hematostatin cells do? Or are they more local? Speaker 2 Typically they are more local. They typically don't have this long range projections between different brain areas. Of course they also project to a lot of different cells, but much more locally and the overall activity patterns also quite different. So the parvamin positive cells are a majority of them are so-called basket cells and they are like fast biking cells. So they are highly active most of the time compared to the rest of the network. While the somata satin are much more sparse in the activity. So again, it's also very interesting that beside the aspect where they target in urine, they also have different aspects in their properties that relate to the spatial extent where they project to, but also in their overall firing properties like the amount of action potentials that they fire. And I think that's also and very important because again you want to ensure that your network stays in stable excitation levels and by that you need to prevent mainly the output. So if the output is in a balance, then you will not have an overshooting excitation because they are not receiving that much input. And therefore you need to have a like relatively high tonic activity level in the sense that are mediating the output. So you keep this in a balance and that's tightly connected to the local circuitry and that's why they are also not so or it's not needed that they actually be projecting that far, Yeah. Speaker 1 It will be more about which cells fire rather than the strength of their individual synapses because they're they're acting more locally. Speaker 2 Yeah, I think it's always the combination of both. And in the end, for the Somata 13, it's essentially the same because you always have this combination of the strength of the synapses, but you also have the amount of activity. And it's not only this one time point when they're active that they have an impact, but every single action potential. So to say, every synapse that has a neurotransmitter release is kind of integrated in the ongoing processing within that cell. And even if you have on one side new synapse being formed, then this has an effect on all the synapses within that cell. So that's why I think it's, it's very important to keep both of them together. You can't really disentangle them. So like the overall activity level and the strength of the synapses, they are like tightly connected and very much depending on each other. Speaker 1 OK. I mean fair. I shouldn't try to simplify it too much. So because these power of a main positive cells are firing a lot of action potentials and they're inhibiting the pyramidal cells. They're, as you said, they're kind of keeping the pyramidal cells in check so that they don't over excite and and probably that way they keep them so that they're actually ready to fire when they need to and they're not just in refractory period all the time. That makes sense to me. So shall we take the next step and think about what happens? I don't know. Should we talk about what happens during learning with these cells or should we go directly to slow oscillations and how they behave with slow oscillations? Speaker 2 I mean, we can first start with what they actually do during learning and I touched upon that a little bit before. But I think what is important to keep in mind is that you need to have local mediation and inhibition on your excitatory cells during learning because otherwise it's very unlikely that you have a systematic encoding of your information. So you need to have the selective inhibitory input for example on the apical to Android coming from the some other statin positive cells or the selective release inhibition of those cells to allow plasticity to actually form use synapses which is the basis of encoding information. And I think this is a very fundamental process which is well studied, you know comparison to a lot of the process that we encounter during sleep. However, there is also a lot of unstudied phenomena there because we don't exactly understand the cellular or the internal processes within one cell which tries to be stable in the amount of synapses and overall synaptic strength and how this is actually depending on inhibitorial input. And what is the driving force behind those plastic changes in where they occurred. Like people argue in this tagging mechanisms, but typically this is actually not encountering the interaction between excitation inhibition and whether they're independent or depending on each other. That's something which is very difficult to answer currently. Speaker 1 So this balance between the excitation and the inhibition is already important at the learning phase is what you're saying? Essentially, yes. So OK, I think we have to assume the animal has learned something and this balance helped. And the synapse. The plasticity has happened at the synapse, and the silences have been strengthened appropriately. Now the animal falls asleep. What happens with the circuitry that is relevant to our story? Speaker 2 First of all, I think what is very important, what happens during sleep is that you have an overall reduction in activity. And this is essentially because you don't need to process all the incoming information. And then you also reduce the overall network activity because you don't need to act on such an immediate level that you respond and your brain is aware of all the things that happen around you. At the same time, we know that we have this reactivation processes and all that going on during sleep. So by that when you reduce overall signal levels, like the overall activity is reduced during sleep. This is very powerful in regard of improving the signal to noise because to reduce noise like the overall activity is reduced. And then with the selective reactivation process, you can actually have a stronger impact in relation to the overall network activity by just having kind of normal activity levels as you had them previously during wake. And beside that, we. We know that we have those processes occurring where we reactivate the cells that were active during previous encoding during sleep. We also know that this is actually accompanied by specific oscillatory phenomena that are sleep specific. In particular, we know that slow oscillations and sleep spinners are very important beside the hippocampus ripple, which I don't emphasize that much here now because they are very fast. And typically when we do the imaging experiments, it's difficult to have such a high temporal resolution that we can actually really look into those phenomena. This is about to change, but we don't know much about it. So forgive me that I will focus on the slower events like the slow oscillation and sleep spinners. And what we found there, for example, is that we have on one side for the slow oscillation, which is a very prominent feature where we have this pattern of traveling waves from frontal to occipital areas with this off state where you have more or less cortical silence for, I don't know, 100 or 200 milliseconds. And then the activity returns back. And I think what is essentially or very important to understand is that this is not really an oscillation, which is, I don't know like feet activity where we have this ongoing more or less sign of the curve. But the slow oscillation is a single event type of oscillation which is characterized really by this period of cortical silence. And what we found is interestingly is that right before the initiation of that downstate, we have an increase of some other statine positive cell activity. And if you remember the some other statine positive cells are the ones that actually target the apical dendrite. And as you mentioned the very beginning, they're like very powerful if you really have a massive recruitment of them to kind of keep the activity down. So they suppress all the driving input. So the network shuts down and probably the some other statine cells are highly involved in the initiation of this downstate. So they suppress the input by that the activity really goes down after a few 100 of milliseconds stay, recover. And then also there's some other 13 positive cells are not active. So they are briefly active. The entire network shuts down also to some other 13 positive cells actually reduce the activity. And that might be one of the mechanisms that kind of guides activity patterns to be synchronized. Because by that you have this kind of 1 moment where you shut down the network and then there is a more synchronous recovery across the network based on when the inhibition is not sufficient anymore to suppress the overall activity levels. Speaker 1 I think that that's really clear. But just one question. So is it then the case that the somatostatin cells go into a refractory period? I mean, how do they provide the appropriate timing? Is it because of the refractory period in the somatostatin cells perhaps? Or is it the refractory period in the pyramidal cells? Speaker 2 I think it's probably on one side an internal process and some other 13 positive cells. So they have a certain refractory period probably we don't know. But what we know is actually there is 1/3 cell type that I touched in the very beginning and these are the inhibitory cells that actually project inhibitory cells. In particular there are so-called VIP interneurons, vasointestinal peptide positive interneurons which is 1/3 group and they directly project to some other 13 and I didn't touch upon that. But what we see, those are the soul cells that are active during the downstate. So it seems like the samadasatine become active, then the VIP sites inhibit the samadasatine and by that they kind of stop this downstate to exceed certain durations. Whether there are interactions on a level that the excitatory neurons also affect that circuits, it's very likely, but I don't know. We don't know yet whether there's actually a direct effect of the excitatory activity on, for example, the VIP interns which then go to the samadasatine. We don't know. Speaker 1 OK. Oh, that's really interesting. I mean, that's already a kind of finer grained understanding than I've ever heard before. So that's really useful. So the somatostatin cells are involved and so are the VIP neurons in those slow oscillations. It's also interesting that you say they're active right before the downstate. So that's the timing of it when they occur. So what are the PARVA cells doing at that point? Speaker 2 The power while being positive cells actually decreasing because also some other statin also inhibit them and vice versa. So we see this kind of differential pattern that the some other statin go up, the PV cells go down and then the downstate occurs and then all of them recover more or less. Whether it is important that they are reducing their activity to kind of provide the environment for the kind of period afterwards where you have this upstate where you want to have the synchronous activity occurring, we don't know because this is a very short time window that you have right before. Then everything reduces its activity during the downstate and then also the PV cells have to recover. So also the pavilion have to recover, but of course with the reduced inhibition at that time point on the soma and the output of the cell from the PV cell cells, then you might actually facilitate a moment where you have this window that it is more likely for cells to just show action potentials. Which also is essentially probably necessary to for the network to recover. Because if you would keep the output low, then you could not recover from the downstate. You need to have the cells becoming active again, sending signals to other cells so that the actual network activity can go back to to previous levels. Speaker 1 So the somatostatin cells have this period of not firing and therefore the excitatory information can come into the pyramidal cells, they can get excited and the parvamin positive cells also stop firing briefly at some point there. And so this kind of loss of inhibition on the cell body also allows. So basically both of the inhibitory inputs are switched off briefly and this is probably what allows the firing of those cells, which is the the upgoing phase. The peak of the slow isolation is that. Speaker 2 Yes, essentially, that's right. Even though I need to admit that we don't have the temporal resolution to really disentangle whether this is a synchronous mechanism or whether there's a slight time shift. So when we look in those modulations, we have temporal resolutions with 100 milliseconds, for example. So this is very slow compared to the actual activity pattern that you see in networks. What I want to emphasize is on that we see this modulation and we see when we look at behavioral data for example, that those oscillations are predictive for memory performance. And when we when we now try to combine those two phenomena, then we see in particular that the slow oscillations that are coupled with a spindle are predictive. And there we have a different modulation of excitation inhibition in the Upstate compared to the solitary slow oscillations. That's very important. Speaker 1 So tell me about what happens in the system during a sleep spindle. Speaker 2 So when we look at sleep spindles, then we have those two options, so to say. Either they occur solitarily or they occur in the upstate of the slow oscillation. Let's focus first on the spindle that occurs without the slow oscillation preceding it. When we look at that, the interesting thing is we see exactly the opposite pattern as as I described it before the initiation of the downstate for the slow oscillation. So instead of having high somatostatin activity and low PV or power mean positive activity, now during the solitary spindle we have increased in power mean positive activity and a decrease in some other 13 positive activity. To simplify that, you need to imagine during the spindle you release the inhibition and the apical dendrite and you increase the inhibition on the output of the cell. And this is an extremely powerful configuration because you open up the window for plastic changes to occur in the dendrite, while you prevent the network from overshooting the activity by suppressing output patterns. Speaker 1 I see. So you're basically taking in all the new information that's coming from the dendrites, but you're not allowed to respond to it. Speaker 2 Yes, or at least not randomly, but only at certain time points when the PV cell releases its inhibition and it does that synchronously between other cell or other neurons. And then you get actually the possibility that those neurons fire together and following having rules, what fires together wires together. You can actually form new synapses or strengthen existing synapses by this pattern. And I think this is super interesting that we have this very unique configuration that you get this change in local inhibition on the cell, which allows actually plastic changes to occur exactly in the time port that we know. It's the strongest predictor from mammary conjugation processes during sleep, which is typically density actually. Speaker 1 Okay, super fascinating. I think I'm finally starting to understand this. But then it begs the question of whether this configuration occurs throughout the spindle, or is it just in the troughs of the spindles? Or do we not know? Speaker 2 It lasts throughout the spindle. And actually when we detect spindles, we see that the PV increase is already occurring roughly a second before the extra spindle onset, which probably is also just because the way we detect spindles in the, EG, for example, is like threshold base typically. And then the spindle actually is probably starting a little earlier, but we're not getting it. What we don't know is whether there is a real modulation within the face of the spinners because we don't have the temporary resolution with the imaging techniques. However, what we know from electrophysiological recording is that the fast spiking cells, so the ones that are typically the majority of the PV, the power B positive cells, are highly modulated by the face in the spindle, for example in prefrontal cortex. And therefore 1 needs to assume that the modulation of those cells is not like tonically hydrating the spindle, but it is face dependent, which again is a very powerful tool for modulating the temporal aspects for long term pronunciation to occur for example. Speaker 1 OK. So that makes a lot of sense. If we think about this idea that the reactivations supposedly occur more in the troughs of the spindles and the reactivations are, we think, the information that we want to encode. So we want those reactivations to be able to modify the synopsis on the dendrites and that would be facilitated by this configuration of low somatostatin and high parvo inhibition. Yeah, makes sense. Now this is all. So we're talking about what during sleep spindles, but without coupling it to slow oscillations. So I think the next step is to think about what happens when you have a sleep spindle that is riding on the peak of a slow oscillation, because I know that this is an important part of your model. Do you want to talk about that next? Speaker 2 Yes, absolutely, because I emphasized a lot now on the inhibitory input of those cells. So we talked about the somatostatin and pavamine positive cells and that we have this release and inhibition on the dendrite and the increase on the soma and the accidental additional segment, so the output of the cell. But we didn't talk about the exadatory network, which is essentially the network that we assume is important for maintaining and encoding the information itself. What is the interesting aspect now is when we look at the sleep spindle, the solitary one, when we look at the chromatic level, like where you, so to say you monitor the output of the cell, there's no increase during the sleep spinner compared to the period before. What we see is there's an increase on the dendrite, but it doesn't really necessarily lead to an increase in the output of that cell. However, when we look at the spindle that occurs in the upstate of the slow oscillation, we do see an increase. We have this unique configuration of release inhibition on the dendrite and at the same time we indeed have an increase in the network's activity, which is most likely specific to the memory content. And by that actually have exactly what you need to facilitate existing synapses, probably even form new ones. And if you provide a window where you have local inhibition and local excitation levels, this essentially what you need to be specific for memory consultation process as we need it to explain the process that we observe when we look on a behavioral level. Speaker 1 Let me just check if I understood that. So we're talking about a spindle that's riding on the peak of the slow isolation. So at the dendritic level, you can see there is an increase in excitation. But in terms of the outputs of that parameter cell, is there an increase or not? Speaker 2 When we look at the spindles that occur in the app state, there is an increase in the output and of course there's also an increase in the activity on the dendrite. When we look at the solitary spindles, we only see the increase in the dendrite, but we don't see the increase on the somatic level, which is essentially monitoring the output of the cell. Speaker 1 Yeah, OK. And you have just explained that because this is the time when we think the reactivation is happening, the input should be associated with the memory that we're trying to consolidate. So we should be able to strengthen those synapses on the dendrites. But also because we have the heavy and plasticity of that pyramidal cell now also firing and the feedback from that, we'll have neurons firing together that we need to strengthen the synapses between and so the whole circuit should be strengthened. This makes sense. OK. So let's then move on to thinking about how you know other topics that relate to this. So, for instance, could we talk about what's going on in Remsley? Speaker 2 Yeah, sure. Getting back to the circuit level, I think what's most important to note is that whatever we do, all this is done in rodents. So we don't know exactly whether this is essentially the same in in humans, but we know that we have the same cell types in humans. What we know is that when we poke an electrode into the brain of a mouse or rat or even intracranial recordings in humans, then what we typically see is we have an increase in action potential rate during REM sleep compared to slow way sleep. And that is also observable if you look for example at fMRI signals. However, when we looked at the calcium activity like the imaging of the cells where we can differentiate the activity of different cell populations, for example exhalatory inhibitory neurons, what we found is that we have a very similar configuration that I mentioned before that we see during spindles. We have it during Ramsey, you have a released inhibition of the apical dendrite, so reduced some other 13 positive cell activity. You have an increase, quite a massive increase in pavamine positive cell activity and you don't have high activity in pyramidal cells in the excitatory cells, but those are also reduced. And of course with the imaging techniques that we're using, one can argue we are not measuring action potentials, it's only a proxy. But it shows that you have this very unique configuration, again an excitation inhibition putting R.E.M. into a inhibitory state. So you have an increase in inhibition coming from the pavomy positive threats during R.E.M. Compared to slowly sleep or non REM sleep and and wakefulness. Speaker 1 Is that throughout R.E.M. or is it particularly phasic R.E.M. Or fader bursts, or we don't know. Speaker 2 What we know is that it's not stable across there. So typically it's increasing and then it's decreasing. So you have this kind of inverted U shape over the time of REM sleep epoch. Whether this is actually modulated by other phenomena, we don't know yet. So we are probably going to investigate in the near future, but I can't tell you anything in relation to REM sleep specific phenomena on a on a finer time scale. Speaker 1 I mean, fair enough. I guess it's a relatively new area of research for you guys. It's interesting though that it's, it's more similar to the solitary spindle than to the slow oscillation spindle complex, because it suggests that you're not, maybe you're not getting such massive plasticity at this time, although you are. I mean, then the neuron is listening and and getting the inputs, but it's not necessarily firing and strengthening up the whole local circuit in the same way. Speaker 2 I think what is important to keep in mind is that what we always have is that R.E.M. follows non REM sleep. And I think the entire mechanism that you can see during non REM sleep is really also solving the the purpose, serving the purpose to strengthen certain synapses to actively manipulate the synaptic strength in a way that you're even facilitating connectivities. Why when you then have REM sleep afterwards? We don't see this kind of massive recruitment of sales during in periods. We also don't see that much of reactivation even though it still occurs. But I think keeping that in mind, it makes sense if you have first a state where you kind of ensure that a certain amount of activity is really engraved into the network that you don't lose it. And then afterwards you can have this kind of environment where there is a certain reactivation possibility and you also have this kind of release inhibition on the dendrite. But it is less likely that the noise levels that you had from previous wake have a strong impact on that because you reduce them already sufficiently. You have you increase your signal to noise and then during REM sleep you can really increase signal to noise levels. You suppress the overall activity. So there's a reduced excitatory activity with a specific configuration of disinhibition on the dendrite where you then can have plastic changes occurring, which probably also serve a quite different function compared to those during non REM sleep. Because you then can can connect things which are probably not so much related as they had been before, which you can't do from the beginning because you would just randomly connect things that don't belong to each other. But during REM sleep you had this selective facilitation process going on during non REM sleep, particularly during sleep spindles for example. Or I become the rippers. Then you can rely on those processes during REM sleep without sacrificing the precision in such a dramatic way as you would do if you do that right away. Speaker 1 I see would it be reasonable to kind of say that during non REM sleep, you know you're strengthening up all the closely associated things and so by having these massive pyramidal outputs during the slow oscillation spindle complexes, you know everything that's in that local circuit, it's getting strengthened together. Whereas in REM sleep, because that's already happened, you can do something more subtle. You talked about longer distance less associated. You know, you don't want to just strengthen up everything that's close by. You've already done that. You want to be more refined and kind of careful and selective. And we don't maybe know the mechanisms for that selectivity yet. But because you've got this groundwork there, you can then work on it in it in that kind of way. Is that basically what you said? Speaker 2 I think first keep in mind that it's not a massive recruitment in the sense of that a massive amount of cells is recruited during non run. This is still a highly specific process like on a cellular. But even on a synaptic level all what we see there is only affecting a very small set of cells at a time. Because otherwise if you recruit everything this doesn't contain any information. This is a highly selective process. Actually we once looked at the amount of cells that are for example recruited during sleep spindles. It's a very small fraction. Most of the cells are silent, which makes sense also during wake even actually most of the time most of the cells are silent because otherwise you don't encode any selective information. But the interesting thing is when you then have those selective pattern engraved into your system, then afterwards during REM sleep, you can actually you don't need to rely that much on this active process of maintaining this engraving the same pattern because you can sort of say rely on what is already on your network. There might be some other things that were not originally related to that and also during non REM sleep they were not directly covered to this, but because they share certain features then it's more likely that they also reactivated during REM sleep without sacrificing that. It would just be reactivated everything. That would be very bad because reactivating unspecifically a massive amount of sales will lead to memory loss to be honest. Yeah. Speaker 1 Yeah. I think that's a really elegant way of thinking about what's going on at the cellular level and and how that relates to what we see at the behavioral level in terms of these consolidation. So, so thank you for explaining that. I'm just thinking about some of the other work that you've done that relates to this. So I know that you've done a lot of work on cells that are active during sleep spindles and how they're different and also how sleep spindles are important for downscaling in the hippocampus. Maybe you could tell us a bit about those? Speaker 2 First, an obvious question for us was whether a sleep spindle has a long lasting effect on subset of cells. Because assuming that you have this process being selective as we said, selective reactivation of a subset of cells during the sleep spindle which facilitates their activity levels for a certain period and by that strengthened the memory. So if this is actually really predicted for memory, this needs to lead to a long term change in their activity patterns. So what we did is we looked at cells that are particularly active during sleep spindles and compared their activity levels over time with spindle inactive cells. So cells that were particularly inactive during sleep spindles. And when we looked at that, what we saw is that sleep spindle active cells were the only excitatory cell cluster that we could find that actually increased their activity over the time course of non remedy. This was really unique. None of the other cells increased their activity. Overall cell population decreases really related to the synaptic downscaling hypothesis. We always see there's a down regulation activity over the time course of non remedy sleep with The only exception that the cells that were active in particular during this non R.E.M. epoch during sleep spinels are up regulated. Then the next question was obviously whether it's this long lasting change. So we compared the activity of those cells during the particular non R.E.M. activity period where they were active during sleep spinels with the next one with the next non R.E.M. epoch where there was wake or R.E.M. in between. And what we saw is that if there is wakefulness in between, they stay on a high level, which is so to say the memory is preserved. That was our first idea when we saw that. However, when you look at the same phenomena, but there's REM sleep in between, they drop down to baseline levels. And this was in the very beginning was kind of disappointing. Like how does this actually explain that they carry information which is preserved on a long term? But what is the interesting fact is that they are still much less down regulated compared to the overall activity level. And what it ensures is, as I said, you have this consecutive first, you have a strengthening kind of increase the single to noise level of that particular representation, in particular during memory sleep. And then during REM sleep, you can't actually decrease their activity again because everything else is much more down regulated and by that we preserve the activity. So I think this is very important to keep in mind that of course it's a rough measure and we don't know exactly whether this is actually memory that is really encoded. But coming from the theoretical framework that we know that memories are reactivated during sleep spinners, then we know that those cells that are active during sleep spinners show this particular pattern. Speaker 1 Could it be that they decrease during R.E.M. Because you know, the consolidation that they needed to achieve during slowest sleep has the curve and the steps that needed to happen during R.E.M. to take that on to the next stage of consolidation have already happened as well. And so they can reduce or I guess in that case you wouldn't see the reduction at the beginning of the R.E.M. Epoch only later. Speaker 2 Actually, I think that's exactly what happens. So you have a certain period where some of the memories are reactivated and then once they are sufficiently consolidated, you don't need to consolidate them any further. And when we look for example, of the overlap between the cells that are considered spindle active between consecutive non R.E.M. epochs, it's relatively small. So it changes, it rapidly changes, and in particular it changes. If there is a longer period in between, then the overlap gets very, very low. It's a small fraction, like probably less than 10% of the cells are the exactly. Yeah, I think that's that's exactly the what what we see. Yeah. Speaker 1 That's fascinating. So maybe you can tell me next a bit about the synoptic downscaling in the hypothalamus. Speaker 2 Yes. So for us it was very interesting first of all to see whether what we observed on a level of activity like we looked at changes in calcium activity over time during sleep, whether this is reflected in a real structural change in brain network. And it was shown before that for example ample receptor densities and ample receptors are the main excitatory receptors in our brains that they are down regulated during non REM sleep. There was a big question whether this is also the case during REM sleep and coming from our data with the calcium activity being low in the pyramidal cell string REM sleep, we expected that REM sleep is actually the driving force for this down regulation and ample receptor densities in the cortex. So what we did is we performed an experiment where we selectively suppressed REM sleep and compared that to animals that were allowed to normally sleep or they were total sleep deprived. Then we looked at ample receptor densities afterwards and what was very interesting for us is actually that we did not see any effect of REM sleep on the downscaling the reduction ample receptor density. So this is actually independent of REM sleep, it's completely dependent on non REM sleep. And this is actually replicated by other people, which was interesting to see because that also shows that the modulations on an activity level is not necessarily one to win one couples to what we see on a structural level. So the activity and the structure, they completely depend on each other, but they don't need to always go hand in hand. Probably also because we have quite an impact coming from inhibitory activity as I mentioned before. And during REM sleep we probably have already a sufficient down regulation of the antireceptors. But what comes on top is actually the strong inhibition that I mentioned before. So I think this is important as one aspect. And then for us it was also very interesting to see whether we can extend this concept beyond cortical networks. So we wanted to look at the hypothalamus, which is a brain area which is highly involved in regulating all our homeostatic behaviors like sleep, eating and so on. We wanted to see whether actually sleep is itself regulating the synaptic connectivity in those works which are important for regulating sleep itself. And indeed what we saw is essentially the same as we found in cortex that sleep is down regulating out receptors in the hypothalamus and by that probably mediating the sleep pressure itself in the circuits that are inducing sleep, for example. Speaker 1 So if it's down regulating the amper receptors in the hypothalamus, so you think it's altering the sleep in general, but how does that make sense? Because then the next night you have to sleep again. Is it just that across the night of sleep you down regulate these receptors and and that's maybe part of the mechanism for concluding you asleep so that you would ultimately wake up and then by the next evening you have more amper receptors again? So it's just part of the homeostasis. Is that how you think of it? Speaker 2 I think that's essentially the mechanism behind. So you have this upregulation during wakefulness and then the doubt regulation during sleep and once this is sufficiently done, you can actually wake U. And if you don't really sleep long enough so that there is not a sufficient down regulation damper receptor in the hypothalamus, you are tired because the sleep promoting neurons are more excitable because they have more amper receptors. It's important to note we looked at overall amper receptor densities so we did not differentiate different sleep promoting neurons or anything like that. This would be the next step which we are working on now. I don't have any results yet. Hopefully we can show that, but this is essentially the idea that you have this synaptic scaling mechanism, the hypothalamus, which is reflecting also sleep pressure in a way, Yeah. Speaker 1 This would be amazing if you could take a measure of the sleep pressure in these mice as well and see whether those things actually correlate. I mean, So what you're suggesting here, that the empireceptors relate to the tiredness, I mean, of course, that's tricky to measure in animals, but that's a very interesting idea. Do you have plans about how you could check that? Speaker 2 I mean, what we want to do is we want to disentangle the different networks. So we want to see whether they're like sleep promoting neurons for example, that are in particularly showing those dynamics because this would be necessary to have this explanation. We are not working on that because as I said, if you imagine the hypothalamus, this is the deepest brain area that you can try to reach and you really need to pass through an entire brain. We're not working on this and we are actually aiming for looking on that on a functional level, like looking at the activity of those cells, but also looking on a structural level, seeing whether synapses are remodeled in those networks and whether this relates to sleep. But also actually as I said, we are also very much interested in the how this translates to feeding behavior, which we also know is depending on this homeostatic process. And there is a very well known phenomena for example that you eat more if you sleep less, but we don't know why this is the case. So the idea would be again that sleep is important for renormalization of the feeding promoting circuits and if you fail to down regulate the activity there, you start eating more. Speaker 1 Yeah, that's really interesting. I just have this feeling that R.E.M. is also important for the feeling of tiredness. And so I wonder whether you know, the you're measuring the amperoceptors, there may be some other cellular or synaptic level mechanism that is doing something equivalent in REM sleep. And it's just that, you know, you don't know where to look for it yet. What do you think? Speaker 2 Absolutely. So again, if we look at amper receptors, we have one major problem and that is we only look at excitatory synapses. So to be honest, this is still I feel a pain when I think about it because of course we need to include the inhibitory cells. We didn't do it yet. But for the imaging, we will actually do this differentiation and then we can disentangle the mechanisms. That's an absolute must to do. And I bet that REM sleep does something, but it's just not visible on an amper receptor level. Speaker 1 Well, that's so interesting. So you're kind of getting down to the micro level. You know, what could be mediating this feeling of tiredness and how does it change homeostatically across the cycle? That's amazing. So maybe because you mentioned the feeding behavior, maybe there's one more topic that we could touch on, because I know you've worked on starvation in mice and how that influences their sleep. Would you tell a little bit about that work? Speaker 2 Yeah, sure. So this was actually work that we were very much inspired by the work by Yogi Boujaki's group where they could show that hippocampus ripples are predictive for changes in peripheral glucose levels. And we were very much interested in whether we also see something like that, but in relation to sleep specific oscillations. And we ran experiments in humans and in rats so that we could have a kind of comparable approach with all the advantages that we have in rats that we can place invasive electrodes. So we actually implanted them with LFP and the hippocampus. And what we first found is that sleep oscillations, slow oscillation and sleep spinels are actually indeed predictive for peripheral glucose changes, which is very interesting because that has a lot of implications in the way, for example, sleep alterations in relation to diabetes and also on. I'm not going to go too much in detail about that, but I think it is very important to note that we have this interaction between peripheral signals and brain activity that is sleep specific, like sleep spinel and slow oscillations. At the same time. We wanted to see whether we can't use it the other way around. So can we manipulate sleep by manipulating the metabolic state? So we had four different groups of animals and essentially we had two things that we manipulated. Either we injected them with glucose, like an immediate increase in peripheral glucose levels, it was an IP injection, or the animals were fed at libitum or fasted and all the combinations. And what we found interestingly, first of all, they were fasted for six hours, which is not really long and they are not like really all the time running around looking for food, but they are like little hungry. What is important, during the actual recording, which was another six hours, all animals were fasted because we did not want to have an interfering feeding behavior that might affect the oscillatory activity itself. And overall sleep architecture was not affected at all, neither by fasting nor by glucose injections. Of course we controlled that with a water injection where they did not get any glucose. And then what we found is that fasting itself had a very strong effect on the occurrence of all sleep related oscillation. We see that there is an increase in sleep spinners. We see there's an increase in slow oscillations and we see there's an increase in the Co occurrence of sleep spinners and slow oscillations. And even the temporal alignment between them is affected. We see actually that the fasted animals, so to say, show all signs of better sleep. And that's very interesting. In a way it might be good to not eat too heavy before you go to bed. And we probably all know that it's not so, so nice to have a heavy meal before we sleep. Speaker 1 Do you have ideas about why? Speaker 2 I think it's two mechanisms. On one side I think that directly the circuits that are in the brain mediating feeding behavior, which are also sitting in the hypothalamus are also cells that are tightly interacting with the ones that are actually promoting sleep, for example. I think this is 1 mechanism. And on the other side, I think also if you're in a fasted state, then also your peripheral signals are affected by that, like you have different levels of peripheral glucose levels and by that you also have different levels of insulin probably. And all that we know that they are very powerful and also mediating brain activity. And by that you probably have a 2 sided interaction. 1 is like a kind of global working mechanism which entails all the hormonal changes, for example. But also on a circuit level you have this direct impact of defeating promoting neurons, for example on sleep promoting neurons. Speaker 1 Yeah, it's OK. So it could just be almost an EPI phenomena of that circuit that the neurons interact in that way and it's not necessarily a functional thing. Speaker 2 Yes, could be, yeah. But it might be good to know about it because we can make use out of it. Speaker 1 Yeah, absolutely. We can make use of it to improve our sleep and maybe also improve our feeding behaviour. I mean, this was fantastic news. I I think I finally understand something about the microcircuitry in sleep. Yeah. I really appreciate that you could have explained it in such detail and so clearly. And also explaining your theories about how things work together and what non RAM and RAM are doing. Yeah, it was great. Thank you. Great. Speaker 2 Thank you. Speaker 1 You've been listening to the Sleep Science Podcast with me, Penny Lewis and my special guest Niels Nietard from Tubigen, Germany. For me and all the researchers at Napsap Cardiff. Thanks for listening and sleep. Speaker 2 Well.