 So we must be live now. Excellent. Excellent. Hello, everybody. And welcome to the second SAS exhibition city stock within the worldwide neuro initiative. I'm George, currently a master's student at the group of Thomas Euler and soon to be PhD student with Tom badman. As your host for today, I would like to first and foremost thank the organizers and more specifically team boggles for the logistics of this talk, but also and probably even more importantly for this initiative towards a greener and more accessible seminar world. The extraordinary and during situation we all witness provides an excellent opportunity for discussions and formats needed and science led changes. Getting back, however, to the reason we all gathered here today, please allow me to introduce our guest, Dr Gautam of a Tramani from University of Victoria in Canada. Upon acquiring his PhD in physiology and biophysics from State University, New York. He worked as a fellow in the labs of Lawrence Trussell team of the Murphy and bottomed Rosca, during which time he was investigating neural activity and transmission down to sign up to resolution. After a couple of years at Dalhousie University, he moved in 2012 to University of Victoria, where he is located ever since and currently hold the title of associate professor. At his lab, they use optical and electrical recordings together with pharmacological or optogenetic manipulations in order to investigate neural circuits and their physiological or neurodegenerative conditions. And today we will be hearing about the latest and I'm pretty sure exciting findings. Without any further ado, please don't welcome Dr Avatramani. Good morning everybody. Good afternoon or good evening. So, can you see my screen? Or do I have to do a share screen first screen first right. Can you see the screen? Great. Yeah. Okay. Hi. So, I'm a little bit nervous. The first online talk and I live on a beautiful island where the internet connection is sometimes spotty. So, I apologize in advance if this, I have some mal connection issues. I'm also nervous because off late my five year old daughter has been especially fond of me and might come barging into the room. And I'm putting a socially isolated and only has a dad to play with. But hopefully those aren't issues today. What I'd like to talk about today is work in our lab that we've carried out in the last few years, regarding the role of inhibition and excitation in the direction selective circuit. So, typically, can you see the pointer? Yes. Yeah. So, typically, you have in a classic model of a neuron you have inhibition and excitation that are collected by the den rights that funnel down to the soma. And then it's the net sum of inhibition or excitation that either cause the cell to fire or not after reaching a certain threshold. So it's the collective sum of inhibition and excitation at the soma. More recently, it's being recognized that some important steps of integration might happen at the deneridic sites themselves at distilled den rights and accessing these is quite a challenge. So today I'm going to tell you a work in which we make somatic and deneridic recordings, and we see how we can piece those together. So, in the first part of my talk, I'll just give you a basic introduction of the direction selective circuit and what the basic mechanisms are emphasizing the role of ACH. This work has been done a lot by Laura and Santhosh. In the second part of the talk, I will talk about the spatial temporal dynamics of ACH at single synapses, at single colonnagic connections between the starburst and the DS ganglion cell or the direction selective ganglion cell. This is going a little bit into the weeds and you'll see we go quite into detail, but then we kind of zoom out and try and understand how the spatial temporal dynamics relate to directional computations. Okay, and this work has been done by Akahiro, Maksimoto, a postdoc in Kezuki Yonahara's lab in Denmark, and a favorite postdoc or graduate student, I should say, David Berson, who did a lot of the SP serial block EM, and then a lot of deneridic imaging by Varsha Jain and Ben Murphy and some modeling by Jeff Rosenroth. And the basic conclusions I want to convince you about are that the steps of integration happen not at the whole cell and not even at single denerides, but little pieces of denerides, five to 10 microns in length. Okay, and these are the origins of the nonlinearities that create what will show you to be deneridic spikes. I'd also like to convince you that both ACH and GABA are tuned for direction at the level of single denerides. So it's not just the non-direction GABA that's tuned, but ACH and GABA are tuned for direction. And finally, we'd like to show you that the computations that happen come from a variety of mechanisms that are distributed heterogeneously along the deneridic tree of ganglion cells. Okay, so let's give you a little brief introduction. So all this started a while back when almost 50 years ago when Barlow and Levick were characterizing the receptor fields of ganglion cells. So when they used small spots and flashed them over the receptor field, what they noticed were on and off responses shown by the plus or minus, and these were evoked only when the stimulus was presented in the center of the receptor field. Now when they moved the spot, they found robust responses when the spot move in the preferred direction, but very little responses when the spot moved in the opposite or null direction. And this is pretty amazing because it's the same amount of light, but there's a certain computation that's happening that prevents and lets the cell decide which direction the stimulus is moving in. So they also had the brilliant insight to realize that since the emotion in the null direction was suppressing something in the center, it must invoke some sort of a null circuitry. Okay, and for the last 40, 50 years, what we've been trying to do is identify that inhibitory circuitry, and we've done that quite successfully and identified the starburst amicron cell as being the mediator of that null inhibition. The starburst amicron cells are a mirror symmetric population and you hear you can see the on and off starburst amicron cells. They send their den rights into the inner plexiform layer and they co stratify with DS ganglion cells, the direction select with ganglion cells. The release GABA and a CH from the distal terminals and importantly, when you destroy these cells by initially done by the cell specific genetic ablation, or more recently by chemo genetic silencing by Malafela's group. You see that you completely lose direction selectivity in downstream ganglion cells. So these cells play a really important role in mediating first computing direction and then mediating that to downstream ganglion cells where they can be output to the rest of the brain. Okay, so how did they actually provide knowledge in addition. This relies on two important factors. Okay, so first of all, we have to go people have found that the starburst amicron cell or sack only connects to the ganglion cell from the null side. Okay, so this is the DS ganglion cell is going to prefer direction and stimulus coming in from the null side. Activate starburst first. Okay, so if you, although the starburst that present right through the genetic field of the ganglion cell, if you stimulate starburst over here, they do not provide the aburgic inhibition. And this was confirmed both functionally using paid recordings and also anatomically using zero block EM by Brigham at all. Now, the second important feature is that the sack den rise themselves a direction selective. Okay, and they'd like motion from so much to the genetic tip. So when the stimulus comes in from the soma to the genetic tip, it evokes a large GABA logic output, which suppresses the glutamate from the bipolar cell. And conversely, when the stimulus moves in the ganglion cells preferred direction, it does not activate the genetic tip there's very little GABA, and the glutamate signal goes through and evokes robust spiking in the preferred direction. So this is the basic fundamental model that we've kind of reached about 510 years ago. Now, the other aspect of this, where people put a lot of effort is to realize the bipolar cell output is not directional. And this they did by very elegant G camp imaging of the terminal or measuring a measuring glutamate with I glue sniffers. And they found that, you know, the output of the bipolar cell is not directional, although, of late, there might be some inkling that few bipolar cells might be directional, but by and large, let's just say the bipolar cell input is not directional. So it's mainly the GABA logic inhibition that shaping the directional response. Now, I mentioned these are GABA logic and colonogic neurons. So the question is, what is a seed of pulling doing and right from the beginning, when they realized this neuron was releasing both this led to the, you know, imagination of a mechanism that would invoke that that would allow the starburst to provide both excitation and inhibition for creating direction selectivity. And this is a picture on the left from David Vainie, who's showing excitation coming from the left from the preferred side, and inhibition coming from the rest of the receptor field. And he imagined that all then right to the starburst with either release a C H or some would release GABA and then therefore you could get this mechanism. Now he didn't. He wasn't quite right in the the mechanistic details. And this wasn't true. But the ideas were correct. Okay, and later on, 20 years later, work from Jimmy zoo's lab, made beautiful bed recordings to test this model. And what they found was when you made bed recordings from the starburst on the null side, you could invoke large GABA GIG inhibition, as well as colonogic excitation which you measure at with the holding potential of zero millivolts for inhibition, or minus 60 or minus 70 for excitation. And there you could get beautiful release of GABA and a CH. Okay, so it kind of doesn't support this model. However, from the preferred side, you don't get any GABA GIG connection, but you get a robust excitatory current mediated by a serial colleague. Okay, so that's a differential transmission of a CH and GABA to the ganglion cell. And this was quite exciting. At the time. The issue was with these recordings was Jimmy found that in the natural stimulus a CH didn't do much so when they blocked in a big a CH, the response wasn't changed very much. In contrast, what we found in the mouse retina was blocking a CH had a huge effect on the Yang mean cell response under a variety of conditions. So on the left, you see the response to high contrast stimuli. This is almost 200 or 300%. It's getting reduced by 50%. If you go to normal stimuli, which are kind of lower contrast, the response is completely squashed. Okay, it goes to zero for a video of a natural movie taken on from a camera on a mouse's head. And you see the responses again very strongly suppressed. Okay, so regardless of whether a CH is directional or not. It's really a driving force for excitation in the DS ganglion cell. So this is also, you know, this indicates that a CH is providing excitation and inhibition to the ganglion cell. Similarly, it should be able to drive direction selectivity. And this turns out to be true and can be tested using an optogenetic approach. So this experiment done by Santhosh again shows a recording from a DS ganglion cell under control conditions. And then the light intensities increase the luck to activate the channel adoption that is expressed only in starburst amicron cells. And then you can see the same kind of tuning properties. And I guess you can see this best over here. You can see that the control tuning matches that of the channel adoption tuning. Okay, and this is for one cell and this is for many cells. Okay, so you can see the starburst network itself has all the elements to create direction selectivity. Okay, so just summarize then we kind of modified the basic model in two to two distinct ways. First of all, we suggest that it's not the bipolar cell that drives excitation, but the preferred starburst amicron cells. Okay, they drive active they are the main source of excitation. And now, you know, you might wonder what happens to the bipolar cell input. After all, if you look anatomically, it's the same set of bipolar cells that drive the starburst and DS ganglion cells. And this is the starburst to release acetylcholine why they're not driving the DS ganglion cells. And it turns out they are releasing glutamate onto ganglion cells, but this glutamate is mediated by NMDA receptors. And this is a kind of neat system where NMDA acts on the ACH and GABA and it acts in a multiplicative fashion. And this is work done from Jeff Diamond and, and even previously, they found that when you block NMDA receptors, you don't change the tuning of the ganglion cell by a similar factor. So it would multiply a null response by two, it would multiply the preferred response by two. Okay, so this is a multiplicative aspect of NMDA. In this sense, the ACH is the, and the GABA, the primary drivers, and the NMDA is the neuromodulator. So another last aspect about the circuit is it would allow the, to generate certain EI offsets between ACH and GABA, okay, which this circuit does not do as well. So, and that's kind of schematized over here. So when you come in the null direction, you activate the null starburst that release acetylcholine and GABA. So that happens at the same time, basically. Okay, so you have in GABA and ACH coming in by these null starbursts, and then a small, a smaller delayed excitation from the bipolar cells. When you come in the preferred direction, what you see is this excitation is drawn out to the left, and it's become really offset from the inhibition. Okay, and this is very well tuned as well. So in this model, you don't need the direction selectivity of the starburst denrites. In a nice set of experiments done by Laura Hansen, she found a way to manipulate the DS and the denrites of the starburst amyprin cells such that they produce non directional output of GABA. And under these conditions, she could still generate a robust direction selectivity in the ganglion cells. Okay, so you can see there's a multiple mechanisms to generate DS in the ganglion cells at a macroscopic scale. Okay, so all this would be well and good and you'd probably stop here if we it wasn't for these beautiful recordings from Ben Sevier. So, if again, if the excitation and the inhibition were summed at the soma, then a work would be done. But early work from Christoph Koch, just based on the passive biophysical properties of the denrites and the natural branching patterns. He suggested that the inhibition and excitation will interact non linearly in these dendritic subunits. Okay, more recently, Ben Sevier managed to get an electrode and then measure activity in the on denrites of a rabbit on DS ganglion cell. Okay, and he found very robust dendritic spikes. Okay, and that's kind of shown over here. So as the stimulus moves in, you see these dendritic spikes measured in red precede the somatic action potential. The activity was initiated right over here. And as as the stimulus moves along you see the latency changes, and when the stimulus is back here, then it's the somatic action potential that's posing this identity action potential to back propagate. The difference is, it's pretty exciting because it says that even when you have a computation at a distal site, that information can be reliably transferred to the soma. Okay, in a passive model that the inputs closer to the soma would win out. If you put a little action potential over here, you can very efficiently transfer information to the soma. The downside of this is, if you have misinformation, so if inhibition and excitation are not balanced correctly over here, and the response is not directional, then you degrade the output of the cell. Okay, so having dendritic nonlinearity is like this requires the inhibition and excitation to be really tightly coupled, and you have to have this kind of configuration right through the dendritic field. That's the prediction. Now, measuring this is quite difficult, but you kind of get a sense of how well inhibition might be balanced through the dendritic field, and just by looking at the anatomy. Okay, so this is a diagram from Kevin Brigman's paper, and it shows the receptive field of a ganglion cell in gray, and what these little vectors are synapses and which point to the soma of the starburst. Okay, so here's the preferred side, and you can see all the vectors are pointing towards the null side. So if you do the quantification of this, you see the vectors of the starburst dendrites are oriented in the opposite direction of the ganglion cells physiological response. Okay, and this is consistent with the model I showed you so the stimulus comes in this way you would activate inhibition and suppressed suppressed the ganglion cell. You have inhibition right through the dendritic field, although if you look a little closer you might see there's a systematic shift in the in the tuning, but let's not worry about that right now, but the important thing is you have inhibition distributed through the dendritic tree. For excitation, especially acetylcholine excitation. The issue is a little more tricky. Okay, and this is because, you know, we think acetylcholine might be mediated through volume transmission. So in the retina, the starburst American cell of the main source of acetylcholine, and I showed you this they're by stratified and the receptors are all through the inner retina. And that means the ACH has to travel in great distances to reach that reset those receptors. Now for the DS ganglion cells of course they co stratified so the distance doesn't have to be that great. However, if you look at the differences between the functional wiring and the physiology, you can envision that the acetylcholine is not synapse specific. And Kevin, Kevin, these little circles depict the wrap around synapses Kevin found, and you see they're more on the null side very few on the preferred side. Yet when you measure excitation, you have an equal almost equal amount of excitation from preferred and null sacks. And if they're not synapses one is, you know, one imagines that this connection would be through volume and therefore doesn't need the wrap around connection. Now, if you imagine volume transmission, and you realize that there's lots of varicostates within this deneridate plexus. So the starburst coverage is huge. At each point, they're almost anywhere from 10 to 30 den rights get each point in this honeycomb. So you imagine that to be a lot of starburst varicostates each coding a different direction. And this is shown in different colors. And now if AC eight spread a big long distance, then you imagine the signal at the single den right, but you know integrate from multiple varicostates and therefore the signal at the den right should be non directional and relatively slow. However, if somehow the synapses are more synaptic like and there were some synapses that we missed over here, you'd get much more transient input to each point in the den right. And since there wouldn't be the spread the signal at the den right would be directional, just because the starburst den right is is releasing in a directional manner. Okay, so you can so the direction selectivity of the deneridic signal largely depends on the spatial dynamics of a CH. And this is what we sought to determine. Now the first clue that it wasn't paracrine, or at least the whole mechanism wasn't paracrine came from a simple recording of the spontaneous activity in voltage clam DS ganglion cells. And over here you see just holding the cell at minus 16 in a in a cocktail of glutamate and Gabba antagonists, you can still see very fast events. And these are quantal mini like events. They have a fast rice time and a fast decay time and a very characteristic shape. Okay, the, the, the peaks are slightly larger than the ampereceptor ones we measure and smaller than the gabbergic ones, but importantly, they rise exactly at the same time as gabbergic minis. Okay, so they seem as fast as gabberge minis they might be a bit slower than the ampere minis. Now if you put on alpha six receptor antagonist you see all of them are gone, suggesting that they're mediated by this alpha six subtype of nicotinic acetylcholine receptors. Okay, so this is telling us, okay, it can be all paracrine that must be some synaptic component to it. So the most parsimonious explanation for those is okay, of course, you know, null starburst make bona fide wraparound connections. That must be the source of the minis we see. So we said, okay, let's go and test that by making paid recordings. Okay, you know, then the null input should be much faster. However, when we made paid recordings, when stimulating the starburst we found the response in ganglion cells was quite slow. Okay, so here pay attention on the rice time. You see that this is in black is the rice time of the minis, but the rice time of the evoked are much slower. And this is true, even when we dropped the voltage that we used to stimulate the starburst and evoked very small responses in the ganglion cell. The responses still remain slow, presumably because when we stimulate the starburst, you have vesicle release all through the arbors in an asynchronous manner. So this made it very difficult to ascertain that the properties are single starburst connections. However, at that point, a paper from my fellow's lab caught attention, where they actually replace the calcium in the back with strontium, and which led to a desynchronization of GABA jake vesicles. Okay, and this is a trick actually used by Wade Regear and Chuck Stevens to study delayed release at cerebellum synapses as well. Okay, so it's a standard method to can desynchronize release from a neuron. So you replace calcium with strontium and you give a brief depolarization of the starburst what you see is a brief response, followed by this barrage of asynchronous activity. Okay, so riding on this plateau at these really really fast events. And this is done from a preferred starburst that doesn't make many wraparound synapses with the DS ganglion cell. So if you compare the response to the spontaneous activity, you see the asynchronous responses are equivalent to those evoked to those observed spontaneously. They're exactly the same waveform. Okay, they have exactly the same amplitude distributions, and exactly the same rice times. Okay, so it's saying that the evoked asynchronous EPSC is likely mediated by a single vesicle. You can see also the important fact that when you measure make these measurements from a preferred starburst, and you know it's preferred because it has a serial calling and not gather. You can see, again, it's has the same distribution of peak amplitudes and rice time compared to the now starburst American cells, which released both a CH and gather. You can see the green because it's just hidden over there, or maybe in this slide of sorry, it's in it's sorry that the knowledge and blue blue versus red and you can see they're completely overlay. So this is pretty surprising because you have in our situation where you have wraparounds and non wraparounds, and they're still giving you a CH with the same connections with the same kinetics and the same properties. So this was really surprising. Okay. You know, if if you're having really fast transmission from preferred sacks. We wonder then, you know, are there synapses that we missed out on the side. If when Kevin originally did these studies, he used data sets in which the ultra structure was missing. These were the first data sets used and to allow for a rapid tracing of the neurons use fixation methods where you couldn't see vesicles, and you could just see these clear cylinders which allowed for rapid racing, and he he was able to see the divergic synapses by their wraparound morphologies. Okay. So we wondered if there could be a second synapse on the periphery of these wraparounds, which weren't observed in this data set. Okay, so a topic. This is called a topic release when you have release away from the pre synaptic zone. Hypothesized to mediate past colonnagic transmission in the autonomic service system, and also reckoned bipolar cells and maybe other parts of the brain as well. So topic releases is an important mechanism. And we wondered if it had any role here. However, in this, if you go to a topic release, you imagine that the release would be independent of the main pre synaptic release at the main pre synaptic specialization. Hypothesis Santosh measured activity from neighboring pairs of DS ganglion cells. So we're here you're looking at the pair of DS ganglion cells you see they're quite the somers are quite close and the genetic fields are quite overlapping. And then he asked, you know, are these, and in this case, you can see the DS, the coding opposite directions. And what do you see over here the minis, and they're again measured in a cocktail blocking GABA receptors and rudiment receptors. And you see every once in a while, a mini that's quite synchronous. Okay, so this is kind of really weird. Right. I mean, normally a mini is a kind of hallmark of a synaptic localized response. And here we're looking at minis that are synchronized across to DS ganglion cells. And if you kind of zoom into that, these synchronized events, you can see them at a higher time resolution, and you can see their eyes is is perfect together and they really go, you know, at the same time. So this give us the indication that a single vesicle releases a sealed Colleen, but that a seed of Colleen can spread to two DS ganglion cells and activate nicotinic receptors rapidly within within a millisecond, you've got your response. So this this is is really neat. And it's kind of a new form of transmission, so to speak. And we kind of refer to it as multi directed transmission, because the ACH is being directed to two, at least two dendrites as possible. It's even three, but as right now we can only measure from two DS ganglion cells. So if you look at the amplitude and decay kinetics, these are quite correlated, which is another surprising aspect. So usually the decay kinetics of a mini is shaped by the deactivation of the receptor. So when receptors bind transmitters, the transmitter, it holds on to the transmitter for a while and then the, the unbinding of the deactivation of the receptors is what determines the decay of miniature events. In this case, you see that there's a very strong positive correlation in the decay kinetics. So you have a slow vent in one cell you have a slow vent on the other. You have a fast event on the cell we have a fast event in the other. So okay, so there's a very strong correlation in the decay kinetics, suggesting maybe they were shaped by a common factor. Now what this common factor is we don't know, but maybe it's the concentration of a CH in the cleft or in that in that local vicinity. That's the hypothesis. Okay, so the next question we addressed is how far does a CH spread. Okay, so we did to address this question we looked at the relationship between the physiological and the dendritic overlap. And this is a method designed initially by Trong and Ricky. And in this method what you do is you measure the physiological correlations and then look at the dendritic overlap. Imagine if there's a huge, a perfect dendritic overlap then all the events would be correlated. And of course, if you have only a little overlap then you have a few chances for overlap events. We computed the distribution of nearest neighbors and that's what's kind of you're seeing in this in this figure. So in the heart regions that there's a very strong overlap. Okay, and then this is the cumulative distribution function for those nearest neighbors, and you can see this ramps up very quickly. Okay, but 10 microns is a very high chance that you'd have to den rights within that 10 micron vicinity. So 10, if AC 8 spread 10 microns, then you would expect, you know, 75 or 80% of the events to be correlated. So then what we did was we plotted the strength of the correlation against the fraction of dendritic overlap in these different bins. Okay, so if you looked at the first bin, and he said look, you need to have a perfect overlap, then you would see this relationship. Okay, and you see, it's it's underestimating the amount of correlation. However, one micron, when you can consider this bin over here, then you see a good mapping of the functional correlation. Okay, if you go to larger distances, then again, they're not matching. So it seems from this analysis that AC 8 spreads about one micron from the from from the release site. So dendrites need to be within kind of one micron region, and then you have correlated activity. So this is pretty exciting because it said, look, you've got a condition where you're getting synaptic and para crying forms of transmission at the same time and it kind of requires a new definition for cool magic transmission basically. So I went to David Burson and told him our results and he was quite excited, and he dug deep into his data sets and examined these barricades again with a lot of detail. So he contacted Kevin and he bought a new data set, which is published by ding at all in nature with where they had ultra structural details. And he again did the analysis where he mapped out the wrap arounds and followed them to different kinds of DS gangling cells. And basically what he found was when he looked at the barricade in detail and green you see it's making a wrap around synapse that kind of where the Gabba transmission happens. Quite often you had a second dendrite in the periphery peripheral region. There was no synaptic specialization over here. Okay, and over here you might see some vesicles but normally they weren't even that many vesicles over here. Okay, so there was no sign of ectopic release per se. But quite often you had, and at least 50% had off the barricades had a second dendrite from a DS gangling cell coding in opposite or another direction passing by within one micron. You looked at the angle of that second DS gangling cell, you see it tends not to be of the same one that's making the wraparands and apps. Okay, but it could be any of the other three. Okay, so you have this kind of novel tripartite complex we have a single. Ecologic axon terminal that's kind of has two post synaptic sites, you know, you know, in a very small region. Okay, just to summarize this part of the talk, you have a novel form of multi directed transmission, where a single vesicle release from the starburst terminal. It doesn't create many so it's not para crying. Okay, it's not spill over because in a spill over means it would go and create a large response in the now and then into a preferred but you see nothing of the sort. You don't you see the responses in both so it's not the typical spill over transmission that we refer to, and it's not a topic release because we see again both. simultaneous responses into ganglion cells simultaneously. Okay, so it's similar to para crying in a sense because the transmitter is spreading a large distance, but the main difference is in para crying you don't normally have fast responses. And to understand why we kind of probed a little further into this, and I won't go into the details of this, but I'll just give you the highlight from Jeff's modeling work, but he took he reproduced the concentrations of a CH that might be experienced in the in the synaptic left in the simulations from Barbara and how so, where you see, when you have release at the site, you have a very large concentration of a CH but that comes crashing down within a few microseconds. Within a millisecond, the, the concentration of equilibrated between a proximal and a distal side maybe one micron away. You can simulate this and you ask how receptors would behave. And to do that, you kind of use these reaction mechanisms with great constants. And again, all this from a published literature and we just kind of for an initial step we took the values from this paper modeling to the genetic responses in the autonomic nervous system. And what we found is if you looked at Alpha three receptors, then they behave very differently at the proximal and distal site. So this could not mediate multi multi directed transmission. Alpha the Alpha seven receptor, and then we modify the desensitization connects because really we have Alpha six but we don't have these parameters available for Alpha six, we slowed down the desensitization of Alpha seven. And what you see then is the proximal and distal responses are much more similar. They both are rising within the millisecond, and they have comparable this one is a bit smaller but at least they're comparable. So in black you see the waveform of the miniature spontaneous EPSC that we measure. Okay, so again, this might be wrong in the precise detailed parameters and we kind of trying to optimize that. The main point is this fast paracrine or multi directed transmission really depends on the biophysical properties of the receptor. Finally, we're trying to understand what all this means. Why are we going so much and trying to understand why ACH needs to be fast and so local. And to directly get an idea of what the light evoked responses will look like, we then collaborated with Kezuki, Yonahara and Akira, and also Yulong in China who developed these ACS sensors. Okay, and here, Santosh made a recording from a starburst amocrine cell and he's expressed the sensor in the whole network of starbursts. Okay, but he's loaded only one starburst with a red indicator. When he delivers a pulse, you see a very rapid ACS signal, and you see that those signals all over the dendritic tree, except they always colloquialize or very close to the dendritic barycosides that you see in red. Okay, so here you're looking at these red little spots are the barycosides of a single starburst. And in each one, you can see these really fast transients. Okay, and these are showing that starbursts release acetylcholine from barycosides. You don't get release at every barycoside, I would say about 50%. These are early recordings, but at least half the barycosides, we can detect sizable cholinergic responses. Now, the average, these are very localized and if you make an average of these, this is the plot you get, and this has a half width, a full max of about one and a half microns. So, so when this is again saying that the acetylcholine released from starbursts is really spatially compartmentalized. Importantly, Akira then did the express the sensor in DS ganglion cells using a certain Cree line initially called Cree and then OTX Cree which labels only ganglion cells, and what he saw was something quite striking. So again, all these colors represent the spatial compartmentalization of the cholinergic signal. In this case, when you come with light and stimulate, you get a response throughout the dendritic tree. But if you go and do this or multiple trials, you'll see the variance of the response is not correlated across small sites. So over here in black, you're looking at how the correlation of the noise falls with distance. And again, you see that's neighboring distances do not share the same variance, they don't share the same noise. Okay, so they're independent units. Again, and that number that he gets matches up with this spread that we see over here. So what this is saying in a sense is even when you come with light that the responses are quite compartmentalized in these small sections of den right. And importantly, if you look at the response to multiple directions and that's shown in ROI one and two, you can see the responses beautifully directionally tuned. So here and and and neighboring sites can have different tuning. This one, the optimal responses over here. And for this ROI, it's over here and and this is actually it's now response. Okay, and if you quantify this and ask what is the directional tuning of all these responses, you see it's basically around the clock. Okay, and this is actually expected, given that you can go anywhere around the DS ganglion cell in the book a colonnagic response. But it's saying the colonnagic response is localized and it's not being averaged over multiple viacosities across around the den right. Okay, at least as reported by the sensor. It's also important to note that the sensor has a relative an affinity that's similar to acetylcholine receptors about 23 micromolar. Now, this suggested another colonnagic inputs are coming the tuned in all directions. Okay, so there's no bias towards the preferred direction so it's saying again, locally it's providing a response, but globally, really it's it's it's not. So if you imagine the den rights integrated information from large sections of of the dendritic tree, then the, you know, having these little local tuned inputs would not matter at all. However, sorry, just just before I get into that I just wanted to show you that when you block acetylcholine esterase you lose the tuning over here, suggesting it's really the acetylcholine esterase is what's keeping the response very localized. So sorry coming back to the integration. We predict that if it integrates over a large area, you get a non directional response but if it integrates over a small area, we get a directional effect. And previous work using calcium imaging actually suggested that the response would be integrated over a small section of the den right. Okay, so we're here, what we've done is these representations of the directional tuning, showing these different colors suggesting that the independent units. So over here, this is the section of den right, and you can see if you look at the calcium response. Again, I guess the side three and four, you can see the tuning is very different so this one has a response over here this one's gone here the peak is kind of going to 30 degrees and here it's going up to 90 degrees. They have different tuning saying that they independent. The signal seems to be integrated over small areas. I must say that these responses are not physiological in the sense that measured in TTX, because of the denitic spiking. If you didn't have TTX you would have a denitic spike, and you would get a calcium response through the denitic tree, and you wouldn't see the impact of local inputs, but TTX allows us to do is to monitor the local inputs. Okay. Now, you can argue about whether TTX, the sodium spike would allow activity to spread. We don't think so, but it probably does to a small extent. So, you're probably not convinced by this that activity is really local, and we did a few more steps to show that what we're looking at these different colors really mean local inputs. This Herculean experiment, Varsha, recorded from a DS ganglion cell, recorded with preferred and null responses, and then in a mouse line where the starburst somas were labeled. And these little dots, I don't know if you see them on your screen, but you can actually see GFP in the nucleus of the starburst American cells. And then she went with an electrode and electroporated them and basically destroyed a handful of starburst shown by these white arrowheads. And she destroyed the ones on the null side that provide inhibition. When she did that, you can see an increase in null spiking. When she did that, she then went and imaged the dendritic tree, and most of the tree was still normal because there's a lot of starbursts over here and destroying a handful doesn't really destroy inhibition. And you can see in red, most of the dendrite is intact. However, a few dendrites lose their direction selectively. And if you look at this dendrite in particular, you can see on the same dendrite, one response is strongly tuned shown in red, and the other one is in blue showing its weekly tune. So the inhibition and excitation are very experienced by these two sections of the same dendrite can be very different, even though they're barely 10 microns away from each other. This is just another example of a before and after. So before the ablation, you have a very strong ds tuning around 0.5 in many parts of the dendrite tree. And then when you ablate, in this case we've lost a ds in most of the other of this small section that we've zoomed into, except we seem to have a strong ds at one point. And again, this is an average of four or five trials, so it's not just some sort of noise, but it's saying that it almost seems like there was one input from starbursts remaining over here and that was able to create direction selectivity within this five micron pocket of the dendrite. Okay, so inhibitions is very, very local. This is the final slide in case I'm probably running over time. But over here, what Santhosh did was he did the converse experiment where he loaded starbursts, amulet cells, patched onto starbursts, amulet cells and stimulated them and loaded the calcium dye into a gangrene cell. And that's kind of shown in green over here. When you stimulate the starburst what you see is very local responses. So again, it's saying, you know, ACH can only spread a little bit so if your varicosity is far away, you don't see anything. And also that the response is really compartmentalized. Okay, so now I must emphasize these are calcium recordings. So you shouldn't get confused with calcium and voltage. Okay, the voltage spread is bound to be much more not so localized, but calcium gives you is just the peak of the voltage so to speak. Okay, so it really thresholds the signal and all you can see are the peaks of the voltage. But I would argue that, you know, this is what determines whether a dendritic spike occurs or not and these are the deciding points, the information is kind of nonlinearly transformed. So this in fact is the important part and the calcium, the thresholding by the calcium allows you to view these important points along the dendritic tree. Okay, to summarize, we have a conventional mechanisms for generating DS where you have non directional excitation and a directional inhibition that some that the soma to give you your somatic spiking response. However, what I've shown you just now is that the important steps of dendritic integration is that acting integration happen in the dendrites in these small regions. Okay, our first, first thought was that this design must be repeated right through the dendritic tree. So you have this directional inhibition right through the dendritic tree, and you have non directional excitation right through the dendritic tree. This was our initial inkling. However, what we found was the ACH was tuned as well. Okay, so it wasn't the simple pattern repeated through, but different patterns of inhibition and excitation will appear to be occurring in different parts of the dendritic tree. So in some parts, in some sites you could have purely tuned excitation and other sites you could have inhibition and excitation canceling and canceling each other out as previously thought. Okay, these are distinct mechanisms right this is an and not you need inhibition and excitation. This is an and computation way just need the excitatory mechanism. Now, there's a huge advantage to this because in a half year tuning is already accomplished by excitation. The amount of inhibition you need is now much small we don't need inhibition at each point. And because a CH is rapid, you know, the path that you need where you need most inhibition in the null direction. Okay, this is going to happen naturally. Okay, in the null direction you're going to get the activation of these null connections or wraparounds, which might co release a seal Collina Gabba, and then that would happen in a very synchronous temporally precise manner, and therefore you can cancel them really easily. Okay, if your a CH is spreading around a lot, then it wouldn't be coming from the same varicosity and you would have orthogonals also being activated and contributing the response, and you could not have this precise timing between inhibition and excitation. Okay, so this provides a very method to to synchronize your inhibition and excitation for cancellation. Finally, you ask why do you need all these subunits and and and if you have these small subunits what we speculate is that you can have a continuous readout of the direction as the stimulus passes through. So if you had a stimulus coming in and have a quick turn, the DS Canyon cell would respond to that. Okay, but this we still again need to verify. The final point of speculation, looking at these beautiful super resolution images of the the inhibitory sites. We can speculate where you have lots of clusters of inhibitory synapses. That's where the conventional inhibitory mechanisms would be implemented. In many parts of the den right where you don't have clusters don't have inhibition. Maybe that's where you get these excitatory mechanisms and tuned excitation play an important role. And these you see in many parts of the cell, you don't have you have long regions where you don't have inhibition. But again, this is just speculation. Okay, so finally, sorry if I ran over time. I just want to, again, I've told you who's done the work. This is a beautiful campus in the University of Victoria and that's a lab over here. And I just want to acknowledge the funding sources that allowed us to do the work. And again, the collaboration of Kizuke and you long and the person allowing us to put the story together. So again, thanks for your attention, and be happy to take any questions. Thank you very much. Got on for this fascinating talk. It's really nice to see that you attempted to explicitly figure out what the role of a set the whole line is in the direction selectivity circuit. We have indeed a couple of questions and while I go through them and ask you this. So the audience could still use the chat to inquire about the topic. So the first one is from Brent Young. Are there any post synaptic densities that can be seen in the EM data set in the ectopic RGC, or you don't think that ectopic cells form densities. So this is a, it's a good question and it's in fact, a question that's led the whole field into a state of disarray in a sense, because in the CNS. Cologic synapses do not often have post synaptic densities. And if you see a varicosity and you see no post synaptic density, you assume that the cologic signals are being transmitted by volume transmission. So this is all in many parts of the brain. For us, we did not see a post synaptic specialization either. If you go to the preferred side, the peripheral synapses that we see with the varicosity, we don't see any post synaptic specialization. And it's only because we had the physiological recordings, we can say that there must be transmission happening over there. Without the physiological recordings, we would think nothing was happening over there. That's kind of why Kevin also probably came to that same conclusion. There's only a combination of physiology and the SPM, you can tell that cologic events must be happening at these sites without post synaptic densities. But finally, we are trying to localize the receptors to realize, you know, that the receptor must be there even without the post synaptic density. The second question is from Marla Feller. Is it known whether acetylcholine and GABA are packaged in different vesicles, so different calcium dependence of release indicating different release mechanisms. Yes, Marla, good morning. I'm glad you woke up. Yes, in rabbit, GABA and acetylcholine seem to depend on different channels, the N type and the P types, and they might be differentially coupled. In mouse, they don't seem to, they seem to, we haven't really checked the calcium channel dependence, but they don't seem to depend on different biophysical mechanisms. And this I say because what we did was we measured noise correlations in the ganglion cell. Okay, so we measured using the technique by Catherine Ricci, we measured EPSC and IPSC's near simultaneously, and looked at the noise fluctuations. And we saw, when we measured both simultaneously, we saw that every trial, if the excitation increased above the mean, so did the inhibition. Okay, so it seemed like at every instance, we had extra excitation, we had extra inhibition, and on instances where we had less excitation, we had less inhibition. So we, we seem like we had core release or core transmission of acetylcholine and GABA at the same varicose. Now, whether they come from a different vesicle, definitely not. Okay, they're co-packaged into different vesicles. So if you look at the spontaneous activity, you kind of get a sense of that as well. You don't see any biphasic events, and then if the other experiment we did was we measured correlations, and maybe I can show you this. Okay, thanks for asking this because yeah, we measured correlations between inhibition and excitation. Okay, so previously I showed you the EE correlations, and I showed you they were correlated. But if you measure EI correlations, you measure excitation at minus 60, and that's shown at the bottom over here. And this is between two cells, and you're measuring inhibition in one and excitation in the other, you see they're not correlated. Okay, if it was happening from the same vesicle, then you would expect these to be correlated. Okay, so, and, and again, the machinery required to package acetylcholine and GABA are different, and so we don't expect them to be packaged in the same vesicle either. I don't know if I have much more to add to that. Okay, before I proceed to the next question because they started really flooding in. I would like to let you know that in case you don't have open your YouTube live transmission channel that a lot of people are congratulating you and thanking you for the talk. And now back to the discussion on the question section. Mike Manukin asks, first he says beautiful talk and asks at the level of the RGC cell, the driving force on the acetylcholine input would be two to four times larger than the GABA input near spike threshold. How does this relate to response and magnitude? Yes. So, so the, the driving force for whether acetylcholine or ampers always going to be larger near spike threshold, right. I mean, the way inhibition works is is not so much by hyperpore rising the cell but through a shunting mechanism. Okay, so, so this nonlinear mechanism, it, even though it's not creating a voltage, and it's near the reversal for inhibition, it really clamps the voltage at the inhibitory reversal without producing hyperporeization. So, and so it's just a matter of relative conductance that's important rather than the driving force. And this is again clear for it's not specific for acetylcholine or ampers, but this is just the way the inhibition would work and that's why it's so effective. You can have a very large inhibitory conductance in the null direction that allows you to clamp and prevent spiking, even though you don't have the same amount of driving force for both excitation and inhibition. I have a certain questions also regarding the locality of like the dendritic segment. Anna Vlasic really amazing story I'm still a bit confused about how having many different combinations of acetylcholine and GABA interactions at local sites of dendrites of ganglion cell dendrites allows for less inhibition being necessary. Are you saying that in some areas tuning is purely being driven by acetylcholine. So sorry I went over that a little bit quickly and if you're confused about it, I probably confused everybody. So, the easiest way to imagine this is by just comparing the scenario on the left versus the right. And on the left you have non directional excitation, which means to create direction selectivity you would need inhibition at all sites. On the right you have a heterogeneous distribution of excitation. Okay, so only some of them require attention from GABA. Right, the main ones that need to be suppressed are the ones pointing in the null direction. These are these are key to suppress. And this is where inhibition would be directed. These sites do not need inhibition and yes it would come from pure excitation. Right, so if you just look at the color it kind of gives you a sense that you need much less inhibition. Right, the other aspect of it is like I mentioned is the timing of inhibition and excitation that's not shown in these polar plots. Okay, so even if you had this scenario on the left and the E and the I were not well timed, then the poorer the timing the more the inhibition you need. Okay, if excitation comes before or after inhibition, it'll be very difficult to suppress and you need larger and larger inhibition to suppress it if the timing is poor. Okay, what acetylcholine allows if it's if it's synaptic and temporally precise and occurs from the same best varicostates, then the GABA and acetylcholine can be released from the same varicostate and have the exact same timing. Okay, and the minis have exactly the same rice time so you'll have a perfect cancellation of the signal. So in that sense, it minimizes the amount of inhibition required if you have perfect timing. Okay, so these are the two important aspects of why we think inhibition you need less inhibition in this scenario of tuned excitation. Great. Given that we are running a little bit out of time I think the next question will be the last one I will ask you to address. And then like, maybe we can take advantage of the online seminars and transfer this into the zoom room that we are already sitting in. So, like for a casual discussion and whether people want to ask you in person their questions about the topic you just presented. I will choose the question from Sandra Fendt because it's something I also wanted to ask you. Hey Gautam, great talk. Do you plan to look at acetylcholine receptor distribution in the direction of selective ganglion cells. Yeah, absolutely. In fact, it's too difficult for me so I passed this on to Kezuke, who's the expert and he's using CRISPR-Cas to put a GFP tag on the acetylcholine alpha 6 receptor and then use super resolution to actually see that distribution close to the varicosity. It's an important aspect that needs to be confirmed. So it's a good question. Yes, given that a lot depends on the special temporal dynamics, not only of the release but also the reception, it's something that is definitely worth looking into. I would like to thank you Gautam for this talk today. I would like once again to thank the organizers and all of you actually being in front of your computers at different time zones and making all this happen. And I would like to invite you, as I already said, in the Zoom room that we are sitting, we will post shortly the link for this in the comments section. Thank you very much. See you.