 Can you guys see the Neuromatch conference? I saw a bit of it. I saw your talk, actually, that was good. I thought it worked really well. I was impressed also by the platform and the organization and the software, yeah. It's just 200,000 people came. This sort of seems extraordinary to me. Yeah, it's crazy. They're doing a couple of them. Yeah. And I also saw that CNS is now going fully online. Oh, really? The Nuclear Science Society, yeah. Oh, yeah, yeah, yeah, yeah. Oh, okay. Yeah, that's cool. I feel like it's gonna have a big effect in the long term. There's people who will realize it's possible and will realize that they don't need to travel hover around the world. Yeah. I don't know whether it'll totally kill all, like, will conference as well. I think it'll certainly make an impact. Yeah. I think it seems like there've been more and more conferences popping up. It might just force people to refocus a little bit in which ones are actually worth traveling for. Yeah, exactly. Tim, we have 75 attendees online and we're live on YouTube. So over to you. Ah. I'm gonna start right now. Is that right? Please do, yeah. Hi. So welcome to our open research webinar, Eli's open research webinar. So what we're trying to do here is give early career researchers a platform to distribute their work during the time when a lot of conferences have been canceled and when it's difficult to get jobs, et cetera. And so this is an E-Life program and there are a number of E-Life programs trying to help people in these situations. So I'm Tim Barron. So I'm a Deputy Editor at E-Life. So I'm also a neuroscientist. And today we're gonna hear from, come over to the next slide, yeah, from Catherine Voet, post-doc at Harvard, talking about internal state configures or factory behavior and early sensory processing and Drosophila larvae. And Megan Creed, who's an assistant professor at WashU, who's gonna talk about ventral palatal neurons, modulate a cumble activity to amplify reward value. And Ilke Spak, who's a Vaini research fellow at Donders in Holland, who's gonna talk about the hippocampal and prefrontal theta band mechanisms underpin implicit spatial context learning. Great, so after each 10 minute talk, we have five minutes for questions for the speaker and then we'll move on to the next talk. And so if you wanna ask a question, you can type it into the chat on Zoom or directly into the Google document linked here. You can see that and the link is also in the chat. And we're also joined by Miranda, Anya, and Naomi from Elife, who are working in the background to support you. And so they'll help you, they'll help line up your questions. So if you're able to, I'll invite you by name to ask your question and the team will then unmute you so you can do that. Otherwise, I can read your question out loud and include your name where possible. So the open notes document is also a place for you to contribute shared public notes. I mean, welcome to that. And so if you've got chats or discussion about something that's unclear in the talk, you think the other people can help with. And you can list yourself as a contributor on the list above for today's webinar. Okay, finally, I'd just like to let you know we're recording this webinar and also live streaming it on YouTube. And so I'm just gonna ask you to be respectful, honest, inclusive, accommodating, appreciative, and open to learning. Don't attack, demean, disrupt, harass, or threaten others or encourage such behavior. And if you feel uncomfortable or unwelcome on any of these webinars, please contact by Elife by email to eventsatelifesciences.org. And so this inbox is being watched by Miranda and Anya Elife. And lastly, the Elife organizers reserve the right to ask anyone to leave or deny access to a subsequent webinar on Zoom. Cool, thanks very much indeed. So first up, we'll hear from Catherine. Over to you, Catherine. Thanks. I'll share my screen. Thank you all for coming and thank you for the opportunity to present my work here. Start this. Okay, does it work? Yeah, I think that's great. So yeah, thank you. And I'm a postdoc in Harvard and I'm working with Christopher LaLarva and how internal state effects are factory processing. And as a neuroscientist, I'm interested in how the brain works, how the brain integrates information and comes up with the best behavior in the end. So you might imagine a situation where there's a sensory queue where it's really good to have a fast and a flexible behavior. But you also might think that you don't even really need a brain for such a stereotype behavior. However, there's other situations, for example, when you go shopping in the supermarket, where there's a sensory queue that can elicit several behaviors. And it really depends on context, on your experience and also on your internal state, how you respond to the sensory queues, for example, what to buy in the supermarket. And I think you really need your brain for that. And especially internal state is really important and I think all of you know this example where you normally wanna do and go shopping something like this, but in the end, you end up with a cart like this because you're hungry. So I think this is really interesting because we're exposed to the same stimuli, but our behavior changes due to some change in internal processing of the sensory queue. And I'm interested in how this works in the brain and what is the underlying neural circuit. And so this is really hard to look at this in humans. So I turn to the fly larvae and we don't send the fly larvae to the supermarket, but we can ask them about their olfactory preference. So we put them in this little petri dish and on the right side, you see a little order well where we have an order and we ask them if they approach or avoid the order. And we can use fed larvae fresh from the food or we can food deprived them in border for a few hours and then test them in the setup. And we found this order journal acetate where in the fed state, the larvae do not approach the order well. You can see that they accumulated the opposite side of the petri dish. Whereas in the food deprived state, they actually go and approach the order and check it out. So there's a drastic change in behavior due to internal state. They go from avoidance to the order to attraction to the order. So indeed they showed us, I think sophisticated behavior. And now in the drosophila, we can go and we can use the broad range of genetic tools to manipulate single cell types and cell components, especially in the larvae, we can do functional imaging and intact animal because they are transparent. So we can look at the activity of the neurons and we already know very well the anatomy of the neural circuits because there is EM reconstruction of almost the whole brain available. Okay, so now if we test larvae that have non-functional sensory or factory neurons, they do not respond to the order. So we think it's something is happening in the processing of the olfactory information. And the information flows, the brain is following. So the order is received by olfactory receptor neurons, the orange. And then projected to the brain, to the antenna lobe, which in mammals would be the olfactory bulb, where it's sent to glomeruli. And then from these glomeruli projection neurons bring the information to the higher brain in the fly, it's to the lateral horn for innate behavior to the much from body for learning in memory. And so I'm asking, so where in this processing system, where is this food deprivation come in and change the processing? So first we wanted to look at the activity of these neurons and we are using this functional imaging setup where we build a microfluidics chip where we can immobilize the fly larva and we can present olfactory stimuli. So on the bottom left, you see the tubes that come into the chip that bring the olfactory sensory stimuli. On the right, you can see the little larva squeezed in this channel with its nose stuck to out of the channel to proceed olfactory stimulus. And in the larva, we can see, for example, something like this. So here you can see the olfactory sensory neurons and one of them highlighted in green. And you can see that they're sit at the nose of the larva on the top, where there's the dendrites and the cell bodies and then further down in the body of the larva as the brain and we can see the antenna lobe. And all the imaging I'm doing is done in this, in this antenna lobe in the first olfactory processing center. So if we now look at the response of the orange to the order, here you see orange or AB2A. You can see that they strongly respond to the order but there is really no difference whether they're fed or food deprived. So we don't see any modulation at this stage. However, if we look a step further down in the olfactory projection neurons, we can actually see that one of them, the uniglomerular projection neurons shown increase in activity upon order. And then based on the M data, we also know that there is a second type of projection neurons, the so-called multi-glomerular projection neurons that project to a different part of the brain. And if we look in one of these neurons, we can actually see that we see a decrease in activity upon food. And then we were also interested if these neurons actually play a role for behavior. So we silenced these neurons. And if we do this for the uniglomerular projections, we can see that they do not switch anymore from avoidance to attraction, but they always show avoidance to the order. So the food deprived behavior is impaired. However, if we look at the multi-glomerular projection neurons, actually the avoidance behavior is impaired and they already show attraction in the fat state. So the modulation has to come in in the antenna lobe. And what's happening upon food deprivation is that the avoidance pathway is inhibited and the attraction pathway is excited. But we were wondering, how does it actually work in the antenna lobe? So first we let's look at the inhibition of the multi-glomerular projection neurons. So we noted glutamate is actually inhibitory in the fly antenna lobe because it activates the glutamate-gated chloride channel. So we were wondering what happens if we take out this chloride channel using RNAi, is there still this inhibition going on? And actually we do not see this inhibition anymore. So we think this is mediated via glutamate and we also see the phenotype in the behavior. So if we take out this glutamate-gated chloride channel from this multi-glomerular projection neurons, they do not change their behavior to attraction anymore. So this was really interesting. But we also know, I don't have time to show the data, but we noted this glutamate is coming from local inhibitory neurons. So we don't think that this is actually the source of the food deprivation signal. So we actually also looked at this uniglomerular projection neurons which show the increase in activity. If you look at those, it has been shown that they express excitatory serotonergic receptor, the 5-HT7 receptor. So we used a novel cell-specific CRISPR-Cas9 knockout tool to actually get rid of this receptor in these uniglomerular projection neurons. And if we do this, yeah, we don't see the increase anymore. So we think this is actually mediated via serotonergic signaling. And the same in behavior. If we block this receptor and use neurons and we test the larvae for behavior, they do not switch to the attraction behavior. And so the source of serotonin we know is the serotonergic CSD neuron which is very prominent in the olfactory system of the fly larvae. On the bottom, you see the antenna lobes. On the top, you see that it has a lot of projections in the higher brain. It's actually input regions. So it receives a lot of input in the higher brain and projects down to the antenna. If you look at the activity of this neuron, you can also see that it has increased response into food deprived state. If we silence this neuron, you can see that it's also required for the attraction. So what we see is happening. And then we were wondering if there's actually connections between the CSD neuron and the uniclomerular projection neuron because they seem to have a big influence on each other. But actually we do not see any synapses in the wiring diagram in the EM data. So what we think is happening that upon increased activation of the serotonergic neuron, we get a synaptic overspell or non-synaptic release of serotonin that only then the serotonin is able to modulate the uniclomerular projection neurons. So since the CSD is a feedback neuron from the higher brain, we were wondering if this serotonergic signaling is actually sufficient to induce the food deprived behavior and also modulates the avoidance path failure at the MPNs. So what we did is we actually artificially activated the CSD neuron to see if we can do a normally fat larvae would avoid the order. And we want to see if we activate this neuron in a fat larva, if we can mimic the food deprived behavior which would be attraction to the order. And indeed that's what we see. If we activate the CSD neuron, the larvae are already attracted to the order and mimic the food deprived behavior. So we think that CSD is actually the food deprivation signal that's feeding back from the higher brain to the internal. And to bring all this data together, we also built a very simple computational dynamical systems model where we put in information from the connectome about connection connectivity and the connection weights based on number of synapses. And we also integrated information about the neural activity from experiments about the non-synaptic neural modulation but also about the signs of connections which is the either excitatory inhibitory. And this model actually works and we can also perform manipulations in this model and we see also the same results as we see in the experiments. Okay, so to sum this up, I hope I could show you that actually the early, the first early olfactory processing centers much more complicated thought. If you wanna read more about this, there is a preprint available in bioarchive about this study. And shortly I just wanna highlight that we found a new projection neuron pathway that's going to a different region in the higher brain which is also required for innate avoidance behavior. So it's not just these uniglomerular projection neurons as we thought before. And also I could show that there is certain energetic feedback from the higher brain which brings information into the antenna lobe about internal state and which severely modulates the processing in the antenna lobe. And I think this could be a general mechanism how internal state could affect different sensory processing regions, also in other animals. I hope I could convince you that drosophila is a really great model organism. We can all use all these tools, genetic tool here, cost imaging, EM reconstruction, modeling to actually really learn a lot about the brain. And if you wanna read more about functional connectomics in drosophila, feel free to look at a little perspective that I wrote which is on Journal of Neurogenetics. And yeah, then I end and thanks to the lab and all the collaborators and the funding and thank you all for listening. Cool, okay. So hang on a second. So, sorry, I guess there are loads of questions and I'm new to this too. So I'm just figuring out how I get the question. So superb talk by the way, lots of round of applause. So Naomi, should I invite Peter Ape up to give this first question here? Now we've posted you a first question in the chat, I'll post it again. How do I look at the chat? Do I have to stop? So Naomi's posting to me in the chat and I'm supposed to be the clever one. Oh, so this question here that I'm supposed to ask is apparently from, okay, so could you refer the circuit function just based on the conectomics data? I guess the answer to that is no, but you should answer that question. Yeah, no, I mean, our initial hypothesis was based on the conectomics data because we saw that there might stay dependent activity because we found a cascade of inhibitory neurons. So there might be two states in one state, one of the inhibitory neuron is active in another state. The downstream inhibitory neuron would be active. So the conectomic data already suggested that there's like a two-state system, but we really know what's going on and how it works and how it switches. And yeah, and as you could see, we also found some connection which was not synaptic and which was not predicted by the EM. So yeah, I think it's really hard. I think we can get a lot of hypotheses and ideas from the conectome, but in the end, we really have to look at details and neurotransmitters and receptors to really see what's going on. Cool, excellent. Good, Miranda, who shall I call up now? Who, what shall I be doing now? Sorry. Hi, Tim, don't worry. There's Dana Galleli on the line who's got a good question here in the chat. Okay. If you're ready, we can unmute you. I'm just coming to you now. It's a very long list of attendees. It's amazing. All right, Dana, you're allowed to talk if you'd like to share your question. Hello, my question is, do you think other PN channels are also affected by feeding state or is this channel unique? So we already found two channels that are affected. It's this type of uniglomerular projection neurons where there's 21 and they're all very similar. And then we found this multi-glomerular projection neuron, which is one out of 14. So there's 13 other multi-glomerular projection neurons that all have different projection patterns and input patterns. And I guess they might also be affected, but we don't really know what's their function. And for most of them, we don't have driver lines so we can't really look at them right now. So I'm happy we found this one line for this MPM that I showed you. But yeah, I think there is much more stuff happening, yeah. And next we have Karen Chang. You've also got an interesting question. I've allowed you to talk if you'd like to unmute yourself. Hi, great talk. I was just wondering what is known about CSD neurons modulate? What do you think internal state modulation in adults also might be mediated through them? But there is a lot of studies in the adult fly and also it's in moth and in other insects. But so far it's not really known what's the function of this neuron. We know that it affects the antenna lobe in similar ways. Also in the adult, there's different types of serotonin receptors expressed in different neurons in the antenna lobe and they're affected by the CSD at least most of them. But what's the actual function? We don't know yet, yeah. Okay, so it looks like that was a fantastic talk. So we sure have another, I don't know how rounder applause works on Zoom but we should all have another flap. And then we're gonna move on to Megan now talking about ventral palatal neurons and I'll leave it up to you. Fantastic, thanks Catherine. Okay, so thank you very much for the opportunity to speak to you today. I want to tell you about new work from my lab looking at different populations of ventral palatal neurons and how they modulate reward seeking and processing. So for those of you who don't know, the Vettral balladum is a critical node in the basal ganglia. According to classical models of the basal ganglia, you have the nucleus accumbens which integrates excitatory inputs from a variety of brain structures and sends inhibitory projections downstream to the ventral paladum. And one of the more consistent findings in behavioral reward neuroscience is that during reward consumption, you have an inhibition of the nucleus accumbens which disinhibits or releases these downstream structures which is involved in the seeking or processing of reward. And I say this is a consistent finding in neuroscience. This is a recent meta analysis looking across a variety of behavioral paradigms and looking at the proportion of inhibitory responses recorded in the nucleus accumbens during reward consumption. However, there's a lot of heterogeneity in these tasks so different reinforcers are used. There's different operant contingencies or metabolic states. So the first thing I wanna show you is that we can replicate this in our own hands. And the way we did this was to do in vivo recordings of the nucleus accumbens shell in freely moving mice. These were sated mice so there was no metabolic need and there was no operant contingency. So these are just mice that are freely able to like a highly palatable reward in a free access manner during a one hour session. And this is the licking behavior shown here just aligned to a optical locometer. And even these conditions who are able to replicate this result, here's a single unit example in the nucleus accumbens showing a pronounced decrease of accumbal firing. The average of all are significantly modulated units. And consistent with prior literature we did find that even in the absence of any instrumental responding or contingencies we still saw this robust inhibition of the nucleus accumbens shell during reward consumption. We did a further experiment to show that this was in fact causal to reward consumption by using an inhibitory opsin injected into the nucleus accumbens shell to specifically inhibit these neurons in a closed loop way during reward consumption. So that's what you're seeing here. So the closed loop stimulation is being driven by the reward consumption. So we're further inhibiting these neurons during this reward consumption state trying to mimic this endogenous pattern or amplify this endogenous pattern. And we see that in our GFP control mice we don't see any effective stimulation but in our channel rhodopsin mice, sorry, our archaeodopsin mice when we inhibit the neurons we're able to further drive reward consumption by amplifying this inhibition. So nothing too controversial here. We've repeated this finding from the literature that nucleus accumbens shell inhibition is causally related to reward consumption. But what we don't know is what the neural substrate of this nucleus accumbens inhibition is. And turning back to our very classic simplified model of the ventral basal ganglia where we have the nucleus accumbens shell inhibiting the ventral pallidum. We know that the ventral pallidum also plays a really active role in reward processing. So firing rates within the VP encode the expected and actual value of reward. They also encode the bigger or motivation to go seek reward. And critically, a recent work out of Patricia Genek's lab has shown that if you record in both the accumbens and the ventral pallidum simultaneously that these value signals are more widely represented in the VP and they actually occur earlier in the ventral pallidum relative to the nucleus accumbens. So we started looking at old anatomical papers where they've reported this thin palatal fugal projections back to the nucleus accumbens. And we're referring to these for the purposes of this talk as the ventral archipaladol neurons after the pathway that's been described in the dorsal basal ganglia from the globus pallidus back to stritum. And our hypothesis was that this thin projection of neurons from the ventral pallidum back to the accumbens is actually an important source of inhibition that drives this pause in accumbal firing. So rather than just being a passive output of the nucleus accumbens this VP is actively orchestrating accumbal responses as well. And to test this whether the VP is a source of inhibition we first need to tell you that the nucleus accumbens a heterogeneous. So it contains different classes of projection neurons that come in two types based on their expression of the dopamine receptor either D1 spiny projection neurons or D2 spiny projection neurons. It also has different classes of interneurons. Here we're focusing on the cholinergic interneurons or chats and the part-validine neurons or the PV cells. So we first tried to answer this question by injecting a channel radopsin into the ventral paladum. And then we took slices of the accumbens and we used different transgenic lines that expressed a fluorescent reporter in each of these cell types to record from identified nucleus accumbens neurons. This is showing you what that looks like under the patch microscope. So we're going in and targeting identified cells and this is a labeled fibers from the ventral paladum expressing channel radopsin in close opposition. And we patched over a hundred cells of each cell type. So about 420 cells total. And this is showing you the location of those cells within the accumbens here. This is the commissure and the locations of recorded neurons around the accumbens shell. And what's readily apparent is that there's really high rates of connectivity. So this is quantified here looking at the spiny projection neurons where we see over 75% of the neurons that we recorded had this input from the ventral paladum. And a smaller but still significant proportion of interneurons were also innervated. So about 28 to 30% of the cholinergic and parvalbumin interneurons. And I don't have time to show you any of the data, but pharmacologically and biophysically we determined that these connections are all inhibitory. So they release GABA and they're monosynaptic. So we say they're monosynaptic, but we know that the nucleus accumbens has a lot of recurrent local connectivity. And so when we take slices we might be missing a lot of the important network dynamics. And so stimulation of the ventral paladum might have complex effects on the structure. So we've repeated this experiment but instead of cutting slices we performed in vivo recordings. So we have the same channeled-opsin injection into the ventral paladum, but instead of taking slices we're now in planting an electrode array into the nucleus accumbenshell that's coupled to an optic fiber to stimulate in vivo. And we recorded from the accumbenshell and we used electrophysiological properties to isolate putative medium spiny neurons. These are the projection neurons and putative interneurons. We also had a fair number of multiunits where we couldn't distinguish what cell type were determined that they only came from one unit. And consistent with our sliced data what we saw is when we illuminated the terminals in the ventral paladum those cells in the nucleus accumbens showed this robust inhibition in vivo. So these are all single unit examples that you're seeing here. This is the firing of a putative medium spiny neuron when we shine the light we see this robust and immediate inhibition that lasts for the duration of the light pulse. And this is the average data shown here. So very robust consistent response regardless of the post-synaptic cell type. So what we've confirmed is that in vivo and ex vivo the ventral paladum does send a projection back to the accumbenshell and can inhibit those neurons. This is the summary data shown here showing the inhibition. But we haven't shown. So if we know that is that the, or we haven't asked rather is are these ventral paladal neurons active during reward consumption? So if we know the nucleus accumbens is inhibited during reward consumption that these cells inhibit the nucleus accumbens in vivo we predict that they would in fact be active during reward consumption. And we answered this question using fiber photometry in vivo to measure bulk calcium signal. We injected first a retrograde Cree virus into the nucleus accumbenshell and then a Cree dependent calcium indicator G-Camp6 into the ventral paladum. This is going to allow us to express that option only in these NAC projecting VP neurons. So selectively targeting the subpopulation. And we just measure bulk calcium signal in vivo as the mouse is doing the same pre access reward consumption task. If we look at the calcium signal aligned to the onset of these lickbouts we see that yes, consistent with our hypothesis we do see an increase in the G-Camp signal immediately after reward consumption onset. We don't see this in our GFP controls which is reassuring. And we took this analysis one step further so for each subject we took the top quartile trials so the quartile where the mice showed the highest G-Camp response and the quartile where they showed the lowest G-Camp response at the onset of lickbout. So all the mice are included in this analysis and then we asked how does the behavior or does the behavior differ between these two classes of trials? We found that in the mice, in the trials that were preceded by the highest G-Camp signal had significantly longer lickbouts. So even 10 seconds after that G-Camp peak we see that mice are, there's a 50% chance that the mice are being, are so engaging in this lickbout whereas if we look at the behavior on those low G-Camp trials immediately after the mice have initiated the lickbout they've dropped to a 50% probability of continuing. So we do see a really robust change in behavior that's predicted by this initial difference in the G-Camp signal immediately after reward consumption onset. And we wanted to test this relationship in a more causal way. So this is correlating the endogenous activity with the behavior. But our next question was whether this ventral palatal arc whether this archipalatal activity can actually promote reward consumption. And we did this by turning back to our optogenetic stimulation. So here we're injecting we're still sticking with that same free access task. We're injecting channel redoxin into the ventral palatum and we're stimulating fibers over the nucleus accumbens to selectively target those ventral palatal projections innervating the nucleus accumbens shell. And we stimulated in two ways. The first was using that same closed loop stimulation. So whenever the mice approach the rewards this triggers the onset of the light stimulation in a closed loop way. And then it almost as a control experiment we did an open loop stimulation. So we took back the same pattern of light but played it in a random way scrambled for the trial to ask if the stimulation on its own would drive reward approach. And this is showing the histology here showing our infection site in the ventral palatum and the terminal fields in the accumbens shell with the optic fiber implanted over the accumbens shell to activate those terminals. So what we found is that relative to the unstimulated condition when we applied stimulation in the open loop way so randomly throughout the trial we didn't see any effect on behavior. But if we stimulated in a closed loop way we saw this twofold increase in the amount of time that the mice spent drinking or consuming this highly palatable reward. And this was driven by an increase in the mean bout length. So the mice are spending longer time at each reward bout. And it wasn't driven by the number of reward approaches or the number of like about. So we're not changing how many times the mouse is approaching the reward we're just extending those ongoing like abouts. And this is significant because this shift in the microstructure of the looking towards longer lookouts is one index of palatability or hedonic value of reward. And we hypothesize that the ventral pallidum is encoding a value signal. And we tested this a little bit more explicitly in two different ways. The first was by filming mice at 140 frames per second and scoring these oral facial reactivity reactions to the consumed reward. Something that has been done for decades in the behavioral pharmacology literature. And we did in practice see that relative to unstimulated trials when the mouse were consuming rewards they did show more of these hedonic facial reactions which represent the effective response independent of motivation. And we also ran the mice in a two bottle choice task. So this was a choice where the mouse had the choice between two different bottles of reward that were exactly identical. Except one of the bottles was paired with reward with the stimulation, sorry. And the other bottle was unstimulated. And despite the fact that these two bottles were identical the majority of mice developed a preference for the bottle that was paired with the stimulation of this ventral archipelago pathway which was an effect that was not evident in the control mice that just received light stimulation without the active oxen being expressed. And as a final control, we did an operant task where the mouse were allowed to just nose poke and receive stimulation directly independently of reward. And in fact, we didn't see any preference so the mice were not willing to work or we're not willing to work to receive stimulation of the ventral archipelago pathway in the absence of the reward. So suggesting that stimulation of this pathway itself is not reinforcing, but when paired with another reward it promotes or enhances the value of that reward. So in summary, what I've shown you is that the Nucleus Cummins inhibition is in fact necessary for reward consumption that these ventral palatal neurons inhibit the nackshell both in vivo and ex vivo. That the endogenous activity of the ventral archipelago pathway predicts subsequent reward consumption and that these ventral palatal neurons amplify reward value to drive further consumption. With that, I'll thank lab members, specifically Jessica Tuley and Ivan Baches who did a lot of the work on this project and take any questions. Hey, so thanks for really very much indeed, Megan. I think I'm a bit more in control this time so I know what I'm doing, but unfortunately the chat's been closed so we have to wait for questions by email. So in the meantime, yeah, so I'm sorry about that. In the meantime, so it's really, really interesting stuff and so it seemed to me that at the end of the day you think that these palatal neurons are encoding value and feeding that value back to the accumbens and I feel like that would do two things. One, it would make them drink longer and the second thing it would do would make them find the stimulus more rewarding next time they were going to, so they would learn something from that value, right? In a sort of classic reward prediction or reinforcement learning style way. And I wonder if you could dissociate that, those two effects, is it just like a physical mechanism for making them continue to drink or is it something about the value? By making them, by figuring out do they actually return more often to that thing or is it just that they stay there longer when they're there? Yeah, so that's a good question whether it's specifically for meal termination or whether it is actually the value. So I think that the two things, the first is that this oral facial reactivity for what it's worth, that's all post-ingestive behavior in response to the reward. So that's one sort of piece of evidence that it's not just a meal termination response. But the other thing is when we look at the two-bottle choice task where we had the two identical rewards, this was one in an overnight session, so they had two identical rewards, one shared with the stimulation, the other wasn't. If you look at the time course of that behavior, it does emerge as a learning signal. So we didn't do a progress afterwards to see if we condition like left, but it's right-bottle doesn't. I noticed that graph was a percentage of total time or something at the reward. It wasn't like how often they would choose that one versus the other one, is that correct? Right, okay, so that was the volume, the total volume that they drank out of the assumed rewards. But if we look at the learning throughout the nighttime session, if we take the last half of the night, then we see that the mice are returning to that bottle more and they're spending more time with that bottle. So we don't see that at the beginning of the session, although whether that's really learning because they're still getting the stimulation even though they're going back there more, it's harder. But there is a learning component to it if it's changing. Yes. It is, but we haven't done the experiment where we put them back in the next day and ask, now that we've removed the stimulation, we haven't done a probe test to see if we really have conditioned that response over the course of the night, which would be interesting to do. And do you have a feeling for where the VP is getting these signals from or how it's computing these signals itself? That's a pretty, that's a very loaded question. We know that the VP gets input from, well, it gets a lot of the input from the accumbens. We also know that it, I mean, I know that there are lots of people working actively on this problem, whether they're our brainstem or early signals that, or polemic signals that are coming in that are maybe a little bit more immediate than ones that have gone through this accumbul loop and come back to the VP. It can't be accumbul stuff because it's because accumbul stuff is being, it's silent, it's that input in order to get an input from somewhere else. Right. So now we have a question from, so can we unmute, Marie? Is it Labus, Labus? I'll try, Marie, I'm gonna try and find you right now. Or I could just ask this question from Marie. Marie is able to unmute. Yes, can you hear me? Yes, we can. Well, actually my question was the question that Tim just, what the input might be. I was wondering whether dopamine neurons arrive in the VP and might promote activity of archipelago neurons. I don't know that's, obviously that's not what you've been studying, but I was just curious. Yeah, so we do know that the VP expresses D1, D2 and D3 dopamine receptors. And recording studies from about 20 years ago showed that if you do recordings of these cells while infusing dopamine agonists, the cells are responsive. They show very mixed patterns of activity in response to dopamine agonist infusion. So they do get dopamine innovation. It's functional in the VP, but we haven't started to tease out that heterogeneity or where these signals are coming from. And then we've got one more question, I think a clarification question that maybe, A McDonald, I'm not sure what it is. I think that's Alison. So Alison, if you're able to unmute, you can ask that question now. Go ahead, Tim, go and ask it. Thank you. So Alison's just asking a clarification question. I may have missed it, but do ventral archipelodal neurons make up a large percentage of VP neurons? No, so they make about 20%. When we've done, we had a lot of data that we obviously didn't show in the 10 minutes, but we've done retrobe tracing and dual retrobe tracing, looking at classical projection targets. So seeding in the ventral tegmental area, which is one of the classical output targets of VP and looking at the proportion of overlap. We don't see much overlap. So it seems like they're a distinct population in these classical projection ventral patterns, but they make up slightly under 20% by our rough calculation, given the imperfect nature of these retrograde dual tracing techniques. What a fabulous talk and a great set of questions. Thanks very much indeed, everybody. Fantastic. Yeah, thank you. I'm just going to hop in here. Hi, I'm Naomi, and I'm running in the background. I just want to thank everybody for their patience while we deal with some trolls and especially Megan and Ketchum who had to do their talks with some interference on the chat while they were talking. You've done an excellent job, and those questions that just came in were really good. We think we've probably managed to kick off the one or two people that were coming in with different usernames. So we're just going to proceed as normal and just thank you. Unfortunately, these things happened, but it's really great to see how many genuine attendees we have here. And these are great talks. Thanks for sharing as well, Tim. No problem at all. And welcome Ilka to give our next talk. Who's going to be talking about Hippocampal Theta? I don't have the title in front of me now, but I probably could, here we go. Yeah, Hippocampal and Refrontal Theta. So kick over to you, Ilka. Thank you very much. I'll just share my screen here. There we go. All right, yeah, so a bit of a gear change from Drosophila to rodents to humans and more of a cognitive study here. So yeah, Hippocampal, Refrontal Theta mechanisms underpin spatial context learning. But before I go into the details of this particular study, just a little bit of background of the research I'm interested in general. So I'm interested in the relationship between predictive processing in the brain and conscious visual perception. And of course, I'm not the first one to come up with the idea that there might be a relationship between the two. It's actually quite a popular topic. These last a few decades, perception as a result of an inference process with 21st century authors associated with this idea, like the ones I'm listing here. And usually when you read papers about this idea, perception as a result of an inference process, they trace the idea back to Helmholtz in the 19th century. But I just thought I'd take the opportunity in a talk like this is a point out that actually you can trace this back at least to the 11th century Arab scholar, Al-Hazan, which I thought was quite interesting. It's not my scholarly discovery, I have to say. It's something that I write in actual scholarly papers and it might, with a bit of goodwill, you can even trace it back to Ptolemy's optics from the second century. So this is a very old idea that there is this relationship between perception as a result of an inference process and conscious visual perception. So that's kind of the background of my research interest. But of course, this is still too vague for an actual experiment. So for the study I'm going to present to you today, I've narrowed down to these research questions. So how does the brain learn and exploit spatial contextual information or spatial contextual priors, if you will? And does this process require conscious awareness of what is being learned? So a little bit of a motivating example if I present to you this image of a messy desk which apparently according to science is a sign of genius. And I ask you to locate the computer mouse. You're able to do this quite well. That's where it is. And that's because the mouse is usually located next to the computer keyboard and below the computer monitor. So there are these spatial contextual associations that humans are very good at exploiting in order to navigate through a visual scene, if you will, in order to locate relevant objects. Now in the experiment I did, I did not use a messy desk but I used a very classical visual search task where human observers were presented with this search array and where they had to search for this oriented T amongst L-shaped distractors. And then once they found it, they needed to push a button left or right to indicate the orientation of the T. Now what the participants did not know is that some of these search arrays, in fact, half of them were repeated over and over again throughout the course of this experiment. So in a schema ties, you can see it like this. So if this is the first trial, then the next trial would be a new trial because it's completely different from the first one. But then later on there will be also an old trial because the configuration of the displays identical to the first one. No, they were identical. I'm talking about the location and the orientation of all the distractors and the location of the target but not the orientation of the target. So participants could not learn the correct response given the spatial context where they could learn associations between the spatial context and the location of the target. This is quite a classical study. It's known as the contextual queuing task. And what I recorded while people were doing this task is a magnetoencephalography. So MEG data to look at the brain responses while people were doing this. But first the behavioral data. So I was able to reproduce this very classical contextual queuing effect where people get faster for all trials compared to new ones and also more accurate. And this was a very clear effect during the second half of the experiment. But like I said, this is not new. This has been shown before. But what I noticed when I looked at these reaction time curves over the course of the experiment is that it appears like there's no effect in the beginning of the experiment. And then all of a sudden there is an effect that appears to remain constant throughout the rest of the experiment. And I wanted to quantify whether that is indeed the case or whether that is an artifact. So what I did was I fitted several models to this effect curve. So first of all, I computed the effect curve which was just all the minus new, basically the reaction time benefit that participants are having. And I compared that to different models. So one was a no effect model, a constant effect model or indeed a switch point or what you might intuitively expect linear effect models. And what I found was that the switch point model was clearly the best one. So it appears indeed that there is no behavioral benefit in the beginning of the experiment. Then there's a sudden switch and after participants have learned that there are these repeats in the experiment, then there is a constant reaction time benefit. So far for the behavioral results, as I mentioned, I recorded MEG while people were doing this and I was mainly interested in the neural oscillatory mechanisms during this experiment and particularly how all trials might be processed differently from the new trials. So when I look at that contrast, I'm plugging that here, what I see is this clear effect in the theta band frequency range. In the first, let's say 500 milliseconds of stimulus processing with a somewhat diffuse sensitive photography which fortunately becomes quite clear after I do source reconstruction on this effect. And then I found that this effect is stemming from two clear clusters, one in the right hippocampus and one in the bilateral, mainly left superior prefrontal cortex. What I can then do is zoom into these clusters to see how this effect evolves over the time course of the experiment. And that's what I'm plotting here for the hippocampus. So blue is all trials again, red is new trials and you can see that all trials have elevated hippocampal theta power, mainly in the beginning of the experiment but no effect anymore in the later part of the experiment whereas the prefrontal cortex behaves very differently. So there's no effect in the very beginning and after which there is a sustained effect for old versus new throughout the rest of the experiment. I can take that effect and plot those on the same axis to compare them directly. Now I'm just computing again a different score old versus new. That's what I'm showing here for the hippocampus. So you can see the exact same. It ramps up and then down again to no effect levels. And now I plot the reaction time data also old versus new on the same axis. You can actually see that as this hippocampus effect ramps up and then down again only as it's ramping down. Do you see the reaction time effect ramping up? And interestingly the prefrontal cortex does something very different to the hippocampus and actually mirrors this reaction time effect quite closely. So it seems that beginning of the experiment people have not learned anything but are still trying to pick up on what's going on. That's what's involving the hippocampal theta rhythm. After hippocampal theta ramps down again you see a behavioral effect which is mirrored by a theta band effect in the prefrontal cortex. So on to the next question, what about awareness? So after people did 22 blocks of this search task I asked them a subjective question. Did you have the feeling that some displays occurred multiple times over the course of this experiment? And they gave a simple yes no answer. And after they gave an answer to this question they had to do one block of an objective recognition task. So instead of having to search for this T people were just asked, have you seen this display before? Yes or no. And that allowed me to assess the explicit memory that people might have formed about these search displays. So remember they've seen these old displays 22 times already over the course of the experiment. Now what's a classical finding in these contextual queuing tasks, this is not novel, is that people if you analyze the whole behavioral sample people are actually at chance level in recognizing displays as old. So that's usually taken as evidence that this is an implicit effect, but that's heavily debated. In my experiment ensemble I was able to replicate that. So I also see that the sample as a whole is at chance level. But I see quite a lot of variance here in this recognition accuracy. But now because I also had this subjective question, did you have the feeling that there were a PTS no? I could do a split, about half of the people answered yes, about half answered no, fortunately for me. And I see that, excuse me, the people who had no feeling that there were repeats are indeed at chance level. But the people who did have the feeling that there were repeats indeed also have objectively speaking explicit memory about the repeats in the experiment. And now the crucial question, of course, is how does this relate to the search task benefit that people form during the search task? How does this relate to the contextual queuing effect? And what I found was completely opposite to what I had expected. I see that both these groups, the people with and without feeling of recognition, have a significant contextual queuing effect. But the people who did not form any explicit knowledge have a considerably greater contextual queuing effect than the people who did form explicit knowledge. So remember, this is the split based on the subjective feeling. So I just asked them, did you have the feeling that there were repeats? Yes, no. But I also have this objective metric of accuracy. And when I then correlate that across the participants to how much contextual queuing they show during the search task, I actually see an inverse relationship here. So wrapping up, people learn spatial contextual associations, or in other words, prior expectations about spatial relationships rapidly and implicitly. And in fact, what's more is that the less conscious they are and the less explicit knowledge they have about these repeats, the greater the behavior benefit, which is quite in contrast to my expectations. The learning phase in this task transiently involves the hippocampal theta rhythm. And after that, there's not really a gradual change, but there is an abrupt switch to an exploitation phase in the behavior, which is dominated by theta rhythm in the prefrontal cortex. So taking together these findings shed light on how the brain continuously adapts its model of the world to optimally structure perception and guide behavior. And what I think is quite an interesting corollary here is that high level cortical structures like the hippocampus and the prefrontal cortex are not necessarily linked to awareness of stuff that's being learned. So I'll end by thanking you all. Also thanks, of course, to everyone in the Langer Lab. I didn't put any names, but nonetheless, thank you very much. And I'll point you to this paper in case you're interested, which is about exactly what I just presented. It was published a few months ago. So thanks, and I'll take questions. Hey, brilliant, excellent talk. Very, very clear, beautiful data. Again, we're going to wait for some questions. And I'll ask you a couple of my own in the meantime. I think now I should announce that we're going to go back to asking questions on this chat. OK, so cool. So I think that you're saying that the people who have the biggest effects are doing it subconsciously. And so does that hippocampal theta, which is like a subconscious recognition memory style effect, does that also show a cross-subject correlation? Then it should do if that's the right argument. So actually, I'm glad that you asked this, because I have a slide about this. So actually, the hippocampal theta effect does not correlate directly with the behavioral performance. But the prefrontal effect does. So this is what I'm showing here. So I'll just show them all. So I'm showing here hippocampal theta power on the x-axis for a prefrontal cortex in the right and hippocampal in the left, separately for the old trials in the top and the new trials in the bottom. And what I see is only the theta power on the old trial in the prefrontal cortex correlates across the participants with the contextual curing effect. So like you said, I completely expected also the hippocampus to correlate, but that's not the case. Only see this effect for the prefrontal cortex, yeah. I thought you were arguing that that frontal one was like in a hormone contrast. I understand what's going on. And the, but OK, I may have missed this. No, so I think what's going on is that the, I think both these effects are reflective of the implicit, both these neural effects are reflecting the implicit behavioral benefit. And I didn't find anything related to the explicit. Sorry, I misunderstood. OK, cool. I'm going to move on to a great question. So hang on, from Daniel Whitblitt. So maybe I probably pronounced that wrong, but maybe we can invite him onto the chat to... He's got a microphone. Would you be able to read out his question, please? Sure, I can. Thanks, Naomi, for handling everything. So would you be able to explain exactly how you define theta power is the question? Yes, I can explain that. So I'll try to go back to the slide with the time frequency. So this is so I computed power in many frequency bands and across many different time points. And here theta, I take to understand oscillatory power between the three and nine hertz frequency band. Does that answer the question? Three and nine hertz. Yeah, that seems to answer the question to me. Cool. And we have a second question from Wu. I think it is... Yeah, I'm just going to see if he can talk or she can talk. They can talk. Hi. Sorry. I think I missed some information about the minutes of the figure of the blue activities and the newer sort of old contracts across the blocks, with also hippocampus fronto and also the reaction time. I think it was another strike. Yeah, not this one. This one? Yeah, I think this one. Because I think there's a lot, so many information in these figures. So I'm like, what does... Like what does this mean? I mean, for example, why the prefrontal cortex, the significant box as from like five to 10 blocks and also in the end of the experiment, I think. Mm-hmm. Yeah, so this is showing the evolution over the time course of the full experiment. So I'm not sure I'm fully understanding your question, but maybe if I take one step back, so in this slide, this is more the raw data, the raw power that I'm showing for the old trials, so the ones that were repeated over and over in blue. And then for the new trials, so the ones that were randomly generated every trial in red. And then I see here for hippocampus, you see this effect early on, where all trials have more theta power than new trials. And frontal cortex have more theta power for all trials only later on in the experiment. So then when I can just basically subtract blue minus red to just get one curve, which allows me to explore it a bit in more detail. And that's what I'm plotting here. So yeah, what I'm thinking is going on is that hippocampus is picking up the association. So hippocampus is doing the learning in the theta band. And then later on, once this knowledge needs to be exploited, which we can see in the reaction time, that's when the prefrontal theta rhythm kind of takes over after this switch. Does that answer your question? Oh, yeah, thank you. Okay. Okay, so I think I've... Naomi, can we ask some more questions here? Yeah, I feel like there's some good ones coming in. Is that okay? The round, we've got a couple of things to do before we end on the hour, but we would like to keep people on if they've got a few more minutes to stay. We'd love to support that. So let's try and do that then. So there are two or three interesting-looking questions that maybe can be asked in about two minutes' time. Perfect, thank you. So what do I... So is there something that I should be reading out now, Naomi, or is it a movie? I can do it if you like. That's okay. What we'd like to do, we'd like to see how we can make these talks more engaging. Obviously, we have to have mechanisms to deal with trolls, but that's okay, that's on us. So we would like to actually ask a few questions of the audience right now. We've actually just got the one question that I'm gonna launch right now and attendees should be seeing this question. So it's basically, if we changed anything about this format, which two things do you think we should prioritize? And it may be nothing and that's okay as well. You can't post things in the chat, I recognize, sorry. We're just gonna give them 30 seconds. And actually, so Tim, what we discussed up front, we'd like to keep this, we think we've got the trolls out now. We'd like to keep this Zoom open for 10 minutes or so just to help those questions be answered and for people to chat with the speakers. Are you comfortable with that? I can hang around for 10, 15 minutes, yeah. I have to go and sort the kids out quite soon, but I've got 10, 15 minutes. Sure, thank you. We can also wrangle that. Okay, cool, and if everyone else is up for that. Yeah, if attendees want to stay on, no one's answering this question right now. So if you can answer the question, that would be great. Can you see it? I can see a question, yes. So there were several questions. I can't see a question. All right, okay. If people can't see it, okay. So I'm just gonna end it, we'll move on. Sorry. No, here it is. I see it suddenly arrived on my screen. Oh, it's just appeared. Thank you, Tom. Oh, yes, now I can see this as well. Yes, great. We'll just give you 10 more seconds if you could click away. Oh, actually, I can't vote. Oh, yeah, sorry, it tells me that. They are not allowed to vote. You don't have the powers. Oh, yeah. And then we've just got some thank you slides, Tim, with an advert for the next one. And it's on page 11 of your doc. Good, look at me just wheeling out things. Okay. That's okay, everyone gets to see how this works. Okay, I'm gonna end the poll now. Well, I'll give you five seconds. Okay, cool. Thank you so much, everyone. So thanks very much to Catherine, Megan, and Ilka for sharing their work with us today and to you all for listening and asking great questions. So our online research talks continues on, actually, this says Thursday, but it's not Thursday. The next one is Tuesday, is it? Next week. Yeah, Tuesday, sorry. Yeah, so it should be Tuesday. And I think it's 9 a.m., not 5 p.m. as well. So my script is wrong. And that's, is that also neuroscience or is that a? No, it's not. And it's chaired by Anna, not you. Sorry. Yeah, right. And then, when's the next, let us know when the next neuroscience one is for this audience, please. So is that in order? This is the end of Neuroscience Week. There are other neuroscience talks to come along. If you carry on with the next bit, I'm gonna find out when. Okay, cool. So, yeah. So for updates about the series, if you just follow Eli Community on Twitter and you can find the schedule and registration information there, cool. So today we're gonna keep the Zoom open for another 15, 20 minutes to give you a chance to chat directly with the speakers. And the speakers are staying on with us. So if you'd like to stay, please do. And we're gonna stop the live stream, but the Zoom is gonna carry on. Yeah, for Neuroscientists, the next one's April the 14th, which is a Tuesday. So that's 9 a.m. in British summertime. 9 a.m. Tuesday 14th is the next Neuroscience one. Neuroscience talks in that one, yeah. Brilliant. Five.