 Hello everyone and welcome back to another Sussex Vision Talk, which forms part of the very successful Worldwide Neuro Talks. My name is Tessa Herzog and I'm a PhD student in the labs of Tom Barden and Leon Lanyardo here at the University of Sussex, which is currently on the frozen south coast of England. So today I'm very happy to welcome Teresa Fattaseri, who is an assistant professor at UC Berkeley, and we'll be talking to us today about direction selective ganglion cells in the primate retina and their role in reflexive gaze stabilisation. So those who don't know her, Teresa completed her PhD with Dr Erica Fletcher in Melbourne and went on to postdoc in Melbourne, Portland and Berkeley, then becoming an assistant professor at the University of California. So her research focuses on how visual signals are encoded and transmitted in the healthy retina, specifically looking at how neurons extract motion and spatial detail from the environment and the contribution of receptors and ion channels into shaping these response properties. Teresa also looks at what happens to retinal signalling during retinal degeneration, looking at the knock on effects of photoreceptive degradation in the structure and function of the inner retina. So I'm very much looking forward to hearing more about the direction selectivity story and gaze stabilisation. So without much further ado, I'll hand over to Teresa now. Thanks very much, Tessa. Let me just get sharing here. Okay, is that looking okay for you? Yeah, looks great for us. Excellent. Well, thanks again, Tessa for that introduction. And thank you to George and Tom for the opportunity to share some of the recent work from my lab with you today. So as Tessa mentioned, I've had a longstanding interest in the cell types and circuits of the primate retina and how they encode different features in the visual environment. And so much of our current focus has been on understanding the ganglion cell types, the output neurons of the retina specifically in the primate. And so we know that in primates, as in other mammals, there are at least 20 or more different ganglion cell types. And as in other species, we expect that each of these will uniformly tile the retina and encode a specific set of features in the visual scene. And so different ganglion cell types we know, optimised to extract different spatiotemporal features or chromatic features from the environment. So each of these primate ganglion cell types is then sending projections to different areas within higher visual areas of the brain. We know that much of the output from the primate retina is targeted to the lateral geniculate nucleus. Some cell types are going to the superior colliculus and other cells are going to other interesting brain regions. And I'll talk about one such cell type today. And so as in other mammals, we know that light enters the retina from the ganglion cell side, activates the rodent cone photoreceptors, and these signals are then passed on to around 10 to 12 different bipolar cell types. Bipolar cells in turn relay this information into different types of ganglion cells. And we know that these signals, these so-called three-week signatory signals are modulated by lateral inhibition from horizontal and amachrine cells. And so if we look at the classic primate, the classic textbook view of the primate retina, we often see something that looks like this. Really emphasising the major ganglion cell types that are most abundant and most well characterised. So these are the midget, parasol and small by stratified cells, which together make up around 70 to 80% of all of the output from the primate optic nerve. And the main functions of these cells I've just highlighted here. But we know that there's much greater diversity and that picture I just showed you is really an oversimplification and that there's much greater diversity of ganglion cell types that is recognised at the level of molecules, morphology and function. And so if we start at the molecular level, we know from the recent beautiful work from Yirong Peng and her colleagues that there are around 18 transcriptomically distinct primate retinal ganglion cell types in the peripheral retina. And so this T-SNE plot, each of these coloured clusters represents a cell type. Here I've highlighted the midget and parasol type cells and this really gives you a sense of their relative abundance. And whilst the remaining cell clusters are comparatively small. And so this teeny tiny cluster here is the smallest of all of them. These are the intrinsically photosensitive retinal ganglion cells. And despite the fact that they are the most sparse cell type, they are serving a very important function in circadian rhythms and pupillary reflexes. So this is just to make the point that even though a cell type might be sparse, certainly doesn't mean that it's unimportant. So this level of molecular diversity is very much married at the morphological level. And so surveys over many years now have converged on around 17 or so morphologically distinct ganglion cell types. This is a recent summary from Dennis Dacey's work from last year. And we see each of these cell types differs with respect to its dendritic morphology as well as to the level of stratification of the dendrites in the inner plexiform layer of the retina. And again, if we highlight the midget, parasol and small by stratified cells, we see these are relatively small field neurons whilst the remaining cell types are much more wide field. And then finally we come to function. And this is where I think we have much more of a gap in knowledge. And so less than 50% of the primate retinal ganglion cell types have been functionally characterized. And so far only five or so cells have been unequivocally linked to a molecular identity. And those are the cells I already mentioned, the midget parasols and IPRGCs. I'm just highlighting this recent work from Ejectionist group who've been using high throughput recordings using multi-electrode array to try and classify some of the different cell types. This work's gone a long way to understanding the diversity of primate RGC function, but we still have some way to go. And so really I think an opportunity and a challenge for our field now is to try and connect the dots on these different modalities. So to try and link a specific molecular cell type with its morphology and to determine its function and also ultimately to understand where these different cell types project to in the brain. And so in this talk today, I really want to highlight an approach we've taken to try and do just that to resolve what I think has been a very longstanding question in our field as to whether on type direction selective ganglion cells are present in primate retina. I'll show you evidence that these cells belong to this little cyan cluster here called RGC-10. I'll describe the morphology and connectivity of these cells and finally show you evidence for direction selectivity in this cell type. So firstly, what are direction selective ganglion cells? We know these are cells that respond preferentially to motion in their so-called preferred direction while showing little or no response to motion in the opposite or null direction. And these cells come in two major varieties, at least from work in lower order mammals. We know that there is an on-off type DSGC that responds to both the on and off set of a light stimulus. These cells receive input from both on and off type bipolar cells. And an on-off DSGC type was just very recently described in work from Dennis Dasey's group last year. The other cell type is the focus of my talk today and these are on-type DSGCs, up until this point they haven't been described in primate retina. And these receive input only from the on pathway. So why are we interested in the on-type DSGCs? Well, these cells we know serve a really important role in reflexive gaze stabilization. And so if we take this schematic from the mouse retina, we know that on DSGCs project to a collection of mid-brain nuclei that are associated with the accessory optic system. And that on DS cells that encode horizontal gaze directions, horizontal motion rather, project to the NOT, DT and nuclei. And those that encode upward and downward directions of motion project to the MTN nuclei. So these accessory optic system nuclei are important for driving the optokinetic reflex. And so this is the reflexive eye movements that compensate for retinal sleep to stabilize slowly moving images on the retina. And so the classic example is when we look out a train window, for example, we see this type of optokinetic nystagmus. And in this gif that I'm showing here, you see this subject fixating this global motion stimulus that's in the top left-hand corner of the movie. You can see that the subject's eyes are reflexively following the direction of this global motion and that's the slow phase of the eye movement, followed by a fast phase corrective saccade back to fixation. And the bottom trace here is just showing the fast and slow phases of this eye movement. So this is all happening reflexively. And if we look at this movie in the top right-hand corner, this is showing the calculated retinal position taking into account the gaze stabilization. And so what you can hopefully appreciate is this is allowing for increased periods of foveation and this reduces image blur and improve security. And so despite the importance of the optokinetic reflex for gaze stabilization, when we know this reflex works in concert with the vestibular ocular reflex to stabilize our gaze, there's been no evidence for on-type DSGCs to this point in primates. And so how then might this work? Well, a prevailing view in the field, and this is a quote that I've taken from one of the seminal textbooks on eye movements. The quote here is that unlike other species, primates like directionally sensitive retinal ganglion cells and processing of moving visual stimuli occurs within the cerebral cortex. And so we know that the accessory optic system nuclei do receive inputs from multiple different brain regions. And whilst they receive direct inputs from retinal ganglion cells in all lower-order mammals, this has really been the question mark for primates. But one view has been that perhaps other cortical areas are areas of striate cortex and also extra striate regions like MT and MST might provide the input to the AOS to drive the OKR in primates. Okay, so why should we care about whether this direct pathway is present? Well, aside from understanding the basic mechanisms here, we are also learning that there are certain human gaze stabilization disorders that may have a retinal origin. And so the example I'm using here is that of infantile mastagma syndrome. This is a condition that leads to these uncontrolled repetitive horizontal eye movements that we're seeing in the subject here and a loss of horizontal optic kinetic reflex. Now interestingly, these patients often have mutations in a gene that's called FRMD7. And the FRMD7 gene has been linked to the loss of horizontal OKR, not only in humans, but has been shown in a mouse model of FRMD7 mutation to lead to the loss of horizontal OKR. So this beautiful work from Yonahara and colleagues showed that in the mouse not only was there a loss of the horizontal OKR, but there is a loss of horizontal direction selectivity. And that the FRMD7 gene is expressed in the starbursts, Democrat cells, a critical inhibitory into neuron for the computation of DS. So when this study came out, this was really, I found this study to be extremely exciting. And I think this was a real clue that the direction selective circuit may be playing a role in primates as in other mammals. More recent work from Martin Kammemann's group has also shown a possible link between aberrant activity or oscillatory activity in on DSGCs and a form of congenital nystagmus that occurs in congenital stationary night blindness. So this again was a piece of evidence that suggests these cells might be present. And then more recently work from Sarah Patterson and her co-workers had identified a candidate morphological cell type that might be the on DSGC. And this was based on serial block face EM reconstruction work that showed a ganglion cell type that connected specifically with starbursts, Democrat cells. So at this point, we thought, right, these cells must be there. How can we find them? And so our search really began at the molecular level. And our approach was really based on the idea that specific mechanisms, specific inhibitory mechanisms are critical for the directional and velocity tuning of on DSGCs. So firstly, we knew that gabbargic input from starbursts, Democrat cells is critical for the directional tuning of DSGCs, specifically the asymmetric inhibitory input from these cells. But importantly, we also knew from work in multiple species that the alpha two GABA receptor subunit is expressed at these synapses. And I'll tell you in a moment why that's important. Now more recent work has shown that a glycinergic amicron cell provides the key input to enable the slow velocity tuning of the on DSGCs. And again, this had been shown in multiple species. And so we reasoned that if there is a primate on DSGC, then it would likely express this alpha two GABA receptor, assuming that there was conservation across species, and that they would also likely express glycine receptors. So whilst this might not seem like much of a clue. So indeed, if we look at the peripheral retinal ganglion cell types, we see that most of these cell types express the alpha two GABA receptor, as well as this specific glycine receptor subunit, the alpha two subunit. So these are the midget and parasol cells here, and the other peripheral retinal ganglion cell types from macaque retina. This is the data set from Peng and colleagues. And we see the expressions higher across the board. So if we now look at these same cell types in the foveal retina, we see something interesting. And that is that most of these retinal ganglion cell types show a markedly lower level of expression of these receptors. Now, when we quantified this, we found that there were two cell types that seem to be an exception to this rule. And they were showing comparable levels of expression of these receptors in both their foveal and peripheral cell types. And so we pursued this cell type here, RGC 10 as a primary candidate for the on DSG C's. We could mine the transcriptome to look for candidate markers that would help us label these cells. And so we identified high level of expression of the transcription factor BNC two in the cell type of interest. And we could distinguish these cells from other cell types, since the only other cell type that expresses BNC to also expresses the transcription factor Fox P2. So I'll get into this in a moment. Okay, so we went after this cell type here RGC 10. This is where it is in that original clustering plot that I showed you. And what we did was to look then at the comparative transcriptomics to see whether we might be on the right track. And so we turned again to the data set from Pang et al. And what we found was that RGC 10 was one of the few macaque retinal ganglion cell types that had a clear molecular homologue in the mouse retina. And so these are the four or five types that show a high level of molecular correspondence. These are different mouse cell types identified by in a single cell aren't RNA sequencing study from room et al. And so what we tried to do was connect the dots here across multiple studies. This is a single cell RNA sequencing data set from Tron et al. We found that 10 novel was the corresponding cell type here. These cells also express a high level of BNC to. And then it became clear in from multiple studies from independent labs that this cell cluster corresponds to a functional on type DSG C. And so the key point here is that the cell cluster we were pursuing seems to be the molecular homologue of the mouse on DSG C's. So at this point we were encouraged and we felt that we had a good lead. So the first thing that we did was to try to label these cells using an antibody for BNC to. And here we're looking at the macaque ganglion cell layer on the left labelled for BNC to the same retina labelled on the right for the pan retinal ganglion cell marker RB PMS. And if we overlay these two channels, we can see that the BNC to expression is in the ganglion cells here. And this is clearer if we zoom into these cells. Now we wanted to make sure that we could distinguish these two different ganglion cell types. And so we are labelled for both BNC to and Fox P2. And indeed we could find cells that express both of these markers and we classified those as RGC 16. And then we also found cells that expressed only BNC to and these we classified as RGC 10. So in this way, we could then classify these two different cell types across multiple retinas. And based on the expression of these two markers, we could reliably classify these two cell types. So now that we could identify the RGC 10 cells, the cell type of interest, we could then determine their density. And we found these cells were present at a density of around 30 cells per millimetre square. This is in the peripheral nasal macaque retina and that they made up around two and a half percent or so of all of the retinal ganglion cells in this region. One thing we had noted is that the distribution of these cells seem to be somewhat non-uniform. And so here we've taken a nasal hemi retina from the macaque and I'm showing at the position of all of the cell bodies of the BNC to positive cells in this sample. And I think what you'll appreciate is there is a higher density of these cells concentrated on the horizontal midline of the retina. And in this largest sample, we found these cells made up around one percent of the total RGC population out of around close to half a million total RGCs in this piece. So this distribution was interesting to us, especially in light of some prior studies that were conducted over 20 years ago where TELCA-ZL had injected into the N-O-T-D-T-N, which was one of those accessory optic system nuclei that I mentioned earlier. And what they found was that there were retrogradally labeled ganglion cells in the retina. And interestingly, and the nasal retina similar to what we see here, there was a higher density of these cells on the midline. So with this we thought was sort of exciting and might be suggestive that the cells we'd labeled are projecting to the nuclei of interest. So we just quantified the distribution of the cells in that piece. Here I'm showing a heat map of the total RGC density for that hemi retina. And we see that the highest density of ganglion cells is close to the optic nerve here. If we plot the ganglion cell density as a function of the distance from the optic nerve, which would be located right here, we see that there's a steady decline in total RGC density. And this is in line with prior studies. If we now look at the density of the BNC2 positive RGCs along this midline, we see a relatively flat density profile similar to what I just showed you. So this is just quantification of that data. And interestingly, when we looked at the relative percentage of these cells as a fraction of the total RGC population, again, if we concentrate on the midline region here, we see that the percentage of these cells increases as you move out towards the retinal periphery. But more interesting, I think, is that the percentage of these cells is highest in the superior retina, which corresponds to the inferior visual field. This I thought was interesting because we know that in our natural environment, the objects in the inferior part of our visual field tend to be closer or closer to fixation than those in the superior visual field. And so we felt this might possibly be a mechanism by which to increase the angular velocity of image motion on the retina should be a nice stimulus for these cells. Okay, so that we know in other species that on DSGCs come in different subtypes and that those subtypes encode different preferred directions. And so we know that there are at least three, possibly four different cell types that encode a forward direction or nasal direction, upward and downward motion. And there's some evidence also for a cell type that encodes temporal motion. So we expect that each of these subtypes will uniformly tile the retina to form a mosaic like I'm showing here for the forward cells, the upward cells and the downward cells. And if we combine these different mosaics, what we expect is that the regularity of the retinal mosaic will be lower as we add in these different cell types. And so what we then did was to quantify the mosaic statistics of the cell type of interest RGC 10. And so here I'm showing a retinal area with all of the cell bodies of these cells indicated by the blue dots. And then for comparison, I'm showing the position of the somers of the RGC 16 cells with these red dots. And what we've shown here are the Voronoi domain areas for these two different cell types. Now on the right here, what we're doing is quantifying the Voronoi domain regularity index, which is a measure of the mean Voronoi domain area over the standard deviation. And the higher this index, the higher VDRI would indicate a more regular mosaic, whereas a lower VDRI would indicate a more random mosaic. And the dotted line on this plot indicates the cutoff under which we expect this mosaic to be random. And so what we found is that for RGC 10, the real mosaic was only slightly above that which we'd expect for a random distribution and was not significantly different from a simulated random distribution of these cells. In contrast, RGC 16, that other cell type, showed a much higher regularity index that was significantly higher than a simulated random array. And so these data sort of suggest that we may have this overlapping, these overlapping mosaics as I showed you on the previous slide. Now, this is one piece of evidence that suggests this, but another way we might think about this is at the molecular level. And so we turn back to the transcriptomic data, we look closely at RGC 10. And one thing that was interesting here is that there was this appearance of perhaps subsets of clusters within this group. And so what we did was to analyze these separate subclusters, firstly for expression of BNC 2, which we found to be similar between the three groups. But we then looked for a gene that we expected might be expressed only in one of the subclusters. And that gene is FSTL4, it encodes in protein called SPIG1. And prior work from Yonahara and colleagues has shown that SPIG1 is only expressed in on DSGCs that prefer upward motion in the visual field. And when we looked at expression of SPIG1, we found that only one of these three subclusters expressed this gene. So this to us is exciting and I think suggestive that there may be multiple cell types within this cluster and within the RGC 10 cluster. And we're working to validate this now using RNA scope. Okay, so we have a good molecular candidate, but what about the human retina? And is this relevant to humans? So what we did was to look at the comparative transcriptome, this work from Yarnetel. And we found that the group RGC 10 corresponds to a cell type in humans called RGC 11. We then mined that data and we found that this cell type also expresses high levels of alpha 2 GABA and alpha 2 glycine receptor and also expresses the transcription factor BNC 2. We could then try to label these cells in the human retina. This is what we're looking at here. So this is the BNC 2 expression here labeled also with a retinal ganglion cell marker. And indeed we find a sparse subset of retinal ganglion cells that express BNC 2 in humans. And these cells made up just under 1% of the total RGC population. Okay, so at this point, we had a molecular candidate and we wanted to know, well, do these cells have the morphology and the connectivity we might expect for an on type DSGC. And so to do this, what we did was to label the fixed macaque retina with an antibody for BNC 2. We labeled these same retinas with an antibody for chat. And then we targeted the BNC 2 positive cells with sharp electrodes full of dye. So we could trace their complete dendritic morphology. And so when we did this, we found cells that had this wide field morphology. Here are some examples. These cells have relatively wide field morphology. They are stratifying in the onset lamina of the retina. And here I'm showing an example of one of these filled cells in comparison to an on parasol retinal ganglion cell at the same eccentricity. So this really gives you a sense of the size and scale of these cells. Now on average at the eccentricity shown here, these cells have a dendritic field diameter of approximately or close to 400 microns. And now that we had a dendritic diameter for these cells, we also know the density of these cells at the same eccentricity. We can calculate the estimated coverage factor, which we found to be a bit higher than three. So this is again consistent with the idea that there may be these overlapping populations of cells, different subtypes encoding different preferred directions. Now the morphology of these cells we identified appeared to be very similar to the recursive monostratified cells described previously by Dennis Stacy's group. And also similar to the cell type that Sarah Patterson had identified as a candidate on DSGC. And so morphologically these cells we think look very similar. This cell I've scaled up for comparison, it was reconstructed at one millimeter eccentricity, but I think the morphology looks very similar here. Okay, so this is just coming back to the coverage factor result here. So this high coverage factor of 3.45 we think might be consistent with this idea that we have multiple subtypes overlapping. Okay, so if these cells are in fact the on DSGC's, we would expect that they should cofasciculate with the starbursts, Democrat cells, which are the critical inhibitory into neurons, providing that asymmetric inhibition for the computation of DS. And so what we did was to look at the relationship between the DI field cells and the starburst cells. So we're going to take this area of the dendritic arbor of this cell. And I'm going to play a video here. We're starting at the level of the ganglion cell layer and we're going to focus down towards the inner plexiform layer as I play this. So these are the starburst cell bodies. And as we focus down, we start to see the dendrites of our field cells come into focus around the same time that the starburst dendrites also become visible. Let's do that again here. Okay, and then if we look at this in the static view, we can appreciate that the starburst dendrites shown in green are showing really nice cofasciculation with the field cells. We can now look at the cross-sectional view of this. Here we're seeing the oncolinergic chat band shown here in green. And if we overlay the DI field cells, we get a sense of this level of co-stratification in the IPL. So we could quantify this across a number of cells. And what we found is that the peak intensities of the chat band and the DI field cells were coincident in the inner plexiform layer. And these peaks were at a depth of around 65% and in the on sublamina here. So this suggests that these cells may be interacting with each other. We could also quantify the extent of this interaction by looking at the contact area between these two channels with the chat signal in its normal orientation and after rotating the chat signal by 90 degrees. And when we quantify this contact area in these two orientations, we see that there's a significantly higher contact area for the normal orientation, suggesting again that these two cells might be interacting with one another. And then from prior work in rabbit and also from Sarah Patterson's work in primate, there's evidence that the starburst cells tend to form these so-called wraparound synapses around the dendrites of the on DSGC's. And so I'm going to play this video, which is a 3D reconstruction of one of these filled cells. We're going to just focus on this dendrite here and the green chat processes surrounding this dendrite. And as I play this, hopefully what you can appreciate is that the chat dendrites are really very much wrapping around the dendrites of these cells in a way that is similar to what has been described previously. So at this point, we had a good molecular candidate. We showed that these cells have the morphology and the apparent connectivity that we'd expect for a non DSGC. And so really the next step was to look at these cells and determine whether they were functionally direction selective. And so given how sparse these cells were, what we did was to use two photon calcium imaging so we can make population recordings from the retinal ganglion cell layer. And so this is just showing our two photon setup here on the left. And on the right, I'm playing one of the movies from the macaque ganglion cell layer. This is sped up to 10 times actual speed. And here we've exogenously loaded. We've loaded the exogenous calcium indicator Cal 520 AM. This is just bolus loaded into the ganglion cell layer. And what we're watching here is the activity of the cells in the GCL in response to a bar stimulus that is moving in eight different directions as indicated by the arrow here. And this bar is drifting at around 500 microns per second or around two and a half degrees per second. So here I'm showing one of these two photon scan fields and the calcium signal is shown here. This is a projected image stack shown here in gray in red. I'm showing a vascular marker sulfurotamine 101, which we use here for landmark detection. And in the top right hand corner, I'm showing the same scan field. But this time what we're plotting is a direction selectivity index pixel based direction selectivity index, which is showing us hotspots in this scan field where there are areas that are responding asymmetrically to the bars drifting in these directions. So these different directions. So this is really highlighting possible areas where we might have direction selective responses. So if we now take out these ROIs and look at the Delta F over F in these regions, we see that these cells are showing asymmetries in their response to these different different drifting bar directions. And here on the right, we see the polar plot summarizing these data. Similarly, we see some tuning in this cell type in this ROI here. Now, if we compare these ROIs to other cells in the scan field, we see that other cells are showing much more symmetric responses to this stimulus. And so this is the example here from region three and region four. So this was nice. We were seeing some evidence for directional signals, but are these cells the cell types that we, the cell type that we had identified earlier. And so to address this, what we did was to fix the samples after the recordings and to label the same sample with our molecular markers that I described earlier. And so here what we showed is that these cells here in region one and two are in fact BNC2 positive and they like expression for Fox P2. They are indeed ganglion cells. And so these cells appear to be the RGC10 cell type that we identified as our molecular candidate earlier. These are the cell types. One of them is a non-direction selective RGC. The other is an amicron cell that is not showing directional signals. So we could repeat this type of experiments across a number of different animals. And when we did that, what we found is that on average, these RGC10 cells, confirmed RGC10 cells, had a higher direction selectivity index than the other RGC group together. Now there are some cells in this group that have higher DSIs and we're interested to determine whether these may be the on off type DS cells. We're looking to identify a molecular marker that might help us label these. And on the right, what I'm showing is a polar plot of all of the RGC10 cells that we've recovered so far showing their preferred directions. And so most of the cells we've recorded from thus far have a preferred direction of inferior to superior motion on the retina. There are other cells that are encoding other directions here. But I think we need more data to really say whether there's a true bias towards this cell type or not. We tend to record from the same retinal area and so this may introduce some inherent bias there as well. So really we need more data, I think, to look at the different preferred directions here. Okay, so finally we're interested to see whether blocking gabarite dependent inhibition could suppress the direction selectivity in these cells as has been shown in other mammals. And so here what we did was again to use calcium imaging to record from the DS cells. We applied gabazine, a gabarite receptor antagonist, and in the two cells that we have so far we are able to show that these cells show more symmetric responses to the drifting bar stimulus after application of gabazine. And again we can show after these recordings that these cells are the cell type of interest. So to summarize what I've shown you, we were able to use a very powerful existing RNA, single cell RNA sequencing data set to identify a candidate molecular marker for the primate on DSGCs. These cells express a high level of the transcription factor BNC2. And they are present at relatively low density and make up around one to two and a half percent of all of the RGCs in the primate retina. I showed you evidence that these cells may be asymmetrically distributed and we're learning more and more about other retinal ganglion cells that may show some asymmetry in their distribution in the retina. The mosaic analysis of these cells suggest that there are likely to be multiple cell types within this cluster and we have some preliminary molecular evidence that suggests this as well. Using morphological analysis, we found these cells are a wide field on type cell and that the morphology seems most consistent with the recursive monostratified RGC that has been described previously. And we found that these cells cofasciculate with the on-starved estamocrine cells. And then finally, using calcium imaging, we've shown evidence that these cells show directional tuning. Our preliminary evidence suggests that this can be blocked by gabazine, but our data do suggest that there may be additional DSGC types here and we're really working on looking into that further. Our general approach here I think illustrates that we could use this type of online directions selective pixel-based mapping to identify where these cells are in our scan fields. And that should allow us to target these cells for spike recordings and for patch clamp recordings. And that to me would open up some really interesting questions. We could start looking at the velocity tuning of these cells and the mechanisms that generate that tuning, for example. So this is really where we're going in the future. And I think generally this approach could be used also to try and resolve the functions of some of these other sparse retinal ganglion cell types. So just finally, I wanted to come back to this sort of original model of the inputs to the AOS nuclei. I think we've gone from a question mark here to saying that it's most likely that these cells that we've identified are the cell type that are projecting to the AOS nuclei. And this is based on what we know from work in other mammals. And so this I think opens up a really exciting and interesting set of broader questions. So really this should force us I think to ask, how do these different pathways, this subcortical, direct subcortical, pathway and the cortical inputs to the AOS contribute to gay stabilisation during development in the mature visual system and in the diseased and disordered primary visual system? And these are questions I think for future analysis. And so finally, I just want to acknowledge all the people who put a lot of hard work and effort into this project over a number of years now. I want to shout out to Anna Wang, who is an extremely talented postdoc in the lab, who's really led the charge on this and pulled this all together. Manoj Kulkarni, who is involved in much of the functional work that we saw today. Jacqueline Gaye, who's a very talented technician who has done all of the immunostaining that we've seen today. Evette Yao and Max Palpstein who were undergrads in the lab who were involved in a lot of the analysis on this project. And Amanda McLaughlin who was involved in much of the early method development here. I want to acknowledge Ben Smith, who is who built the two photon systems that have allowed us to do these calcium imaging experiments. And I want to also acknowledge that really helpful input from Roland Taylor and Marla Feller on the direction of this project. And so I'd also like to thank our tissue sources, our imaging core support, and all of our funding support. And finally, thank you all for your attention. Theresa, thank you so much. That was a great talk. It was fascinating and beautifully explained as well. So we've had a couple of questions that come in. So I'll put them to you now whilst we're still live. So we have a question from Ari Aranberg who says, amazing results. The direction selectivity appears to be relatively low in many cases. Is this due to recording settings or the true biology? What was the signal to noise like in these cells? Yeah. So thanks for the question Ari. So what I've shown here is what I've plotted is a normalised vector sum. And so just to clarify the DSI metric that we're using here. So that'll be a little bit lower than a DSI that's calculated in another way. For example, using preferred minus null over preferred plus null. I think also the lower DSI that we see here may reflect the calcium imaging approach. So the sense that I have is that we may be seeing the excitatory PSPs here in the calcium response as well. And that if we were recording spikes, you know, we would see a higher tuning. And so what I'm hoping is that we can target these cells based on the calcium imaging approach that I described. And my expectation is that we might see a higher DSI in these cells. Having said that, we have done these same experiments in the mouse retina using the exact same protocols and the DSIs are higher there. And so it could be that it is true biological, you know, a true biological difference. But I think we need a bit more data to answer that. And with respect to the signal to noise, for the cells that I've shown you, the criterion for inclusion was a signal to noise of over 2.5. And only cells that pass that criterion were included in the analysis at all. So yeah, I think the cells that I've shown you here have quite high signal to noise and all of the ones that are included in this data set are exceeding that threshold. Great. Thanks. Okay, so our next question is from Enrique and he says, great talk. Have you tried to block glycine receptors to see if they affect calcium signals? Yeah. Thanks, Enrique. That's a great question. And of course, this is something we were actively working on. We've not done that experiment yet. This is what we want to do next. We expect based on the work from Roland Taylor and Vensivia and David Berson's group and Malafalla's group that the glycinergic import will be extremely important for the temporal tuning of these cells. And will be what shapes this really slow velocity tuning of these cells. But we've not yet done those experiments and I'm really excited to do that. We expect that it will be the same as in other species. Great. Okay, so I've got another question from George and he says, hi Teresa, great talk. Do the percentage of the cells resemble the ones in other animals, fervorated or not? So my understanding of the mouse retina is that at least I think from Besson study, those cells made up around the on DSGC is made up around 7% of the total RGC population. So a little higher than what we're seeing here. Great. Thanks. And yeah, I think we've got time for another question. This is from Jeremy Boll and he says, wonderful talk in the field salary construct. I must say actually there's lots of well done and perhaps which you probably can't see. But yeah, so Jeremy says wonderful talk in the field salary constructions do you see an asymmetry in the potential wraparound sinuses with starburst amicron cells. Yeah, that's a great question and we haven't looked at that we haven't quantified that. However, Sarah Patterson's work with the serial serial block face EM did look at that exact question. And I think did show that there is this asymmetry and the connections between the starburst and the DSGC's. So our work has not looked at that, but I think Sarah's beautiful study has looked at that very question. Great. Thank you so much. So we've actually hit all the questions in the chat. So I think now we might move to the more informal chat which will be hosting at the zoom. But first of all, I just once again really want to extend my thanks to you to raise a fantastic talk. It was just it was really lovely work and great to listen to. And also thank you to everyone at the other end of zoom for your time and putting questions forwards. So we'll terminate the live broadcast here, but please do click on the zoom link which is in the chat to join us. Yeah, a more informal chat. Thank you again.