 in three, two, one, two, three. Hey, hello everyone, and welcome for another Vision Web Seminar. So summer is finally upon us. So we have only two online talks left, and then we will leave the stage for the retinal secret symposium that will take place at the end of the month. And then we will finally take some months after this off and take some vacations. So we will be back on mid-September, so I can only encourage you to subscribe to this channel if you do not want to miss any future talk. As usual, I'd like to remind you that these online events are part of the worldwide neuroinitiative. So while you enjoy your vacation, maybe try to watch some of the podcasts. You will find all the relevant links in the description. Today, we are very happy to receive Professor Timothy Brown from the University of Manchester. Based within the Center for Biological Timing at the University of Manchester, Timothy established his own grouping in 2012 and has since focused on the control of circadian rhythm and the effects of visual signals on physiology and behavior. Timothy and going research activities combine large-scale multi-electrode recording approaches alongside intersectional genetics tool for neuro-secret mapping and whole-animal physiological measurement. These activities have contributed to significant agencies in our understanding of the sensory control of the circadian rhythm, such as the influences of daily changes in the color of daylight around dawn and dusk, as well as revealing an aesthetic aspect of more conversional vision, like the contribution of melanopsin photoreception and binocular processing individual types. So those are all very interesting topics, but I think the best will still to hear about them directly from Tim. So hello to Timothy. Thank you for accepting the invitation. How are you doing today? Good, thank you very much. So thank you so much for the invitation. It's a real pleasure to be able to talk to you today about some of our work. So I'll just bring my presentation up. Thank you. Yeah, so thanks for the introduction. So what I really wanted to talk to you today is about some of the work we've been doing over the past few years relating to processing of color signals in the mouse brain. And particularly I want to talk about some of our recent stuff to do with color and conventional aspects of vision in mice. But first of all, because I think it helps to provide some sort of useful context for understanding why we've done what we've done and also hopefully because it's interesting in its own right, I'm going to say a bit about beyond in this title here by which I mean subconscious, aspects of how retinal output influences physiology and behavior. So in the introduction you just heard, obviously a major interest of mine for a long time has been the circadian system. And really that's how I became interested in vision and the visual system in mammals. So around about that time when I first got interested in this area, which is I guess 15 years ago now maybe, this idea had emerged of kind of two parallel visual systems in the mammalian brain. So obviously what I might call here conventional visual system, so output from the retina going to the lateral geniculate nucleus and supporting thermocortical visual processing and also obviously associated accessory projections to superior colliculus, et cetera. But obviously at the same around that time when I became interested in this a really important discovery for us in the circadian field was the identification of these intrinsically photosensitive melanopsin expressing retinal ganglion cells. And at that time, it seemed that this specialized class of retinal neuron specifically projected to parts of the brain involved in subconscious effects of light on physiology and behavior. So in particular projections to the suprachiasmatic nucleus, which is the site of the central clock regulating circadian rhythms, projections to the pre-tectal or livery nucleus, which obviously controls the pupil and also some projections to parts of the thalamus, the intergeniculate leaflet and ventral LGN which are also involved in circadian and other processes. So because these brain regions at the time seemed to be the main targets of the IPRGCs, are generally considered to be mediating these subconscious responses to ambient light intensity but more sophisticated aspects of visual processing, the ability to distinguish form and motion, et cetera, weren't considered to be so relevant to these. This kind of gave rise to the idea of this non-image forming visual system which involved the IPRGCs. And essentially the IPRGCs are a specialized class of retinal ganglion cells that provide information about ambient light levels and regulate these subconscious responses. So that's kind of where I became interested in this area. And I guess it's fair to say that now that this distinction between image forming and non-image forming visual systems is a bit less clear cut than it once was. But nonetheless, even back in those days when I first started working on this, it was certainly apparent that melanopsin didn't account for all aspects of non-image forming vision. So just like any conventional retinal ganglion cell, the IPRGCs and the retina, yeah, they've got their own photoreceptor melanopsin that they can also receive signals from the rods and cones. And so in the circadian field, there was a lot of interest in understanding, what, you know, why is this? What, how does signals from the rods, cones and melanopsin combine to dictate the sensory properties of the circadian system? And obviously this is important in practical terms because it's well established that kind of the daily patterns of light exposure that we tend to experience nowadays, potentially disrupting our clock. So if we could understand the sensory control over circadian rhythms and maybe we could devise lighting systems that were better for our overall health and wellbeing, et cetera, and obviously to do that, you need to know the sensory properties of the photoreceptive input. So that, so we wanted to tackle that problem initially. And obviously doing that is actually kind of challenging. And one of the reasons why it's so hard is because if you look at the spectral sensitivity of the different opsins in the retina of a mammal, they overlap quite significantly, particularly in the nocturnal rodents that people were mainly using for studying the circadian system. So for example, here you can see the spectral sensitivity of the opsins in mice. So like most mammals, they've got two cone opsins, an S cone opsin and an M cone opsin. And those are relatively well separated spectrally, but you can see that all of the other opsins, melanopsin, rods, M cones, they all overlap very strongly. And so for that reason, if you're using the sort of conventional approaches that people tended to use, so you're comparing sensitivity to monochromatic light of different wavelengths, makes it very difficult to separate how each of these different opsins is contributing to the response. And so obviously one thing you can do if you're using mice is use transgenics. And so obviously a lot of our insights have come from studies where people have knocked out either opsins or aspects of the phototransduction cascade of rods or cones. And that's definitely produced some useful insights. But really what we wanted to be able to understand is how signals from these different opsins were interacting in the intact retina. And that's quite difficult to do if you're knocking things out. But fortunately some time ago now, Jeremy Nathan's lab developed this, this really useful mouse model where they took the native M cone opsin and replaced it with the human L cone opsin. So I'll refer to these as red cone mice. So actually, by generating these mice, Nathan's and Co-Ray were able to show some really interesting things about how color processing might work and evolve in animals. But for our purposes, they were just really useful because it gave us this really nice spectral separation between the M or now L cones and rods and melanopsin. And so what we could do then is by using multi-primary visual stimuli. So in this case, we've got three LEDs with peaks in the UV blue and orange part of the spectrum by adjusting the relative intensity of those three different primaries. We're able to produce pairs of stimuli that differed in brightness for one or more options of our choice, but appeared of identical brightness to the other, what we would call silenced opsin. So this is the silent substitution approach that's been quite popular in human psychophysics. So we applied this to these red cone mice and that allowed us to generate stimuli where we could selectively modulate, for example, just the L cones or just the S cones, or we could modulate them in unison, for example, increase brightness for both L and S cones simultaneously or change brightness for the two opsins in anti-base to produce changes in color, all while keeping signals from melanopsin and rods silent. So this is really important then allow this to get at this question of what cones might be contributing to the responses of the circadian system in mice initially. So to look at that, then we performed multi-electro recordings from the SCN in mice and these ties mice, this is. So for these experiments, we tend to use these Buzaki style silicon probes. So we have four shanks each with eight closely spaced recording sites that we can insert into the SCN. And then based on those kind of signature wave for electrophysiological waveform, we see on adjacent recording sites that allows us to isolate individual neurons with pretty good reliability. So in conjunction with those stimuli, I just showed you, which allows us to modulate the activity of specific cone classes. And we could also do it for melanopsin. It wasn't shown on the slide, but obviously we could also look at what happens if we modulate melanopsin signals to these cells. So we could get the back question of how the different sources of photos of perceptive input were combining to dictate the response of SCN neurons. And obviously at the time, our understanding was that essentially all of the retinal input to the SCN was coming from the IPI GCs. And sure enough, we found that all of those cells pretty much showed evidence of melanopsin-based responses. But what about cone-based responses? So certainly most, if not all of them, responded to our cone-isolating stimuli, but we found different types of responses. So quite a lot of cells responded, as you can see in this example at the bottom. So they would show excitatory responses to steps in brightness selectively targeting either the L or the SCN-opsin and those signals combined. So if you stimulated both opsin types in unison, you got a bigger response. If you stimulated the manti phase, you got a slightly smaller response. But more interestingly, we also found other cells that behaved like this. So they showed this kind of classic opponent behavior, for example, showing an inhibitory response to selective stimulation of the L-opsin and excitatory responses to selective stimulation of the S-opsin or in some cases vice versa. So as evidence that this is kind of your prototypical response to color, what we found is that if we modulated activity of the two cone types in anti-phase, so this would produce a large change in apparent color, that is the ratio of L to S-opsin activation, without changing the net brightness for the two cone types together, those cells would show very, very large modulations in firing, whereas if we modulate the two cone types in unison to produce an achromatic change in what you might call illuminance, then the responses are very small. So we found robust color responsive cells in the central clock, and there were quite a lot of them. So all in all, somewhere in the region of a third of the cells we recorded had these color-sensitive responses in the central clock. So that led us to ask, well, obviously, there's information about color here in the clock, why is it helpful? And the obvious reason why it might be helpful, given the idea, the goal of the clock is to align rhythms in the animals, physiology and behavior to changes in the outside world. If you look at what happens to the spectral content of ambient illumination around dawn and dusk, it's not just the amount of light that changes when the sun goes below the horizon, but also its spectral content. For example, as you can see here, when the sun's below the horizon, there's a strong loss of these middle wavelengths, which is caused by that light having to pass through more ozone before it reaches us when the sun's below the horizon, and that selectively filters out that band. So to us as humans, this appears as a kind of blue shift in the color of twilight, and that's true also for most mammals that have dichromatic visual systems. So for example, you can see here, the native mouse M and S cone options well positioned to detect that spectral change. So obviously, in that sense, there's information here in the spectral content that could tell the mouse what time of day it is in the outside world, but really the value potentially of color in the obviously measuring the amount of light can also tell you what time of day it is, is what happens if you look at what happens on clear and cloudy days. So we took data recorded in Manchester between August and October for ambient illumination, we broke it down into clear and cloudy days, and what you can see here is that obviously, if it's cloudy, that can have a big impact on the overall level of irradiance that you might experience, like a 10-fold reduction easily. But the relative spectral content is quite similar. So this difference between twilight and daytime is retained whether it's a clear or a cloudy day. So that sort of tells you something about why color might be useful then. If you were just relying on measuring brightness to figure out the position of the sun, if it was a cloudy day, your estimate could be out by, your estimate of sunrise or sunset might be out by 45 minutes or so compared to a clear day, whereas that doesn't happen if you're measuring color. So our thoughts then were that probably by combining information about brightness and color that enhances the ability precision with which the clock can infer the position of the sun. And in later studies, we went on to provide evidence for this. So I'm not gonna go through all the data from this relatively recent paper in detail, but just to give you an idea of the kind of things we did, we generated a kind of housing environment for mice where we could change the spectral content of ambient illumination presented to animals in their home cage. And then for example, we could do things like this. So we could change between pairs of stimuli where the brightness for metanopsin and rods was identical, but the relative activation of SNL cones was skewed one way or the other to make the illumination appear relatively yellow or blue. And what we found was that bluer colors evoked weaker responses, weaker circadian responses in the mice. So an example of that is shown here. In this particular example, we kind of simulated jet lag by maintaining mice under a light dark cycle and then shifting the time of the day. And we found that if we made the day's color blue, resembling twilight, it took them longer to re-align their activity than if we made the days appear yellow more better resembling day. And we confirmed that effect was due to cones because if we do the same thing in mice that lack the cone specific cyclic nucleotide gated channel, that effect disappears. And in this study, we're also able to actually recreate naturalistic daily patterns of luminance, including or lacking clouds and including or lacking color. And using those kind of approaches, we're able to actually provide some strong evidence in support of the hypothesis I just outlined. Essentially what's happening here is that by combining signals about color and brightness, it helps to buffer the clock against this kind of effect of stochastic variations in brightness due to clouds, which render light intensity and ambiguous time of day. So color enhances the precision with which the mouse clock can tell what time of day it is. So given that data showing this regulation of the circadian system, which obviously entirely dependent on signals from IPRGCs has a major contribution of color, we were interested in determining whether color signals might be also be important for other IPRGC driven responses in mice. And obviously the kind of prototypical response that a lot of people have studied is control of the pupil. So pupil responses are known to be dependent on IPRGC inputs. So for example, in this nice study from Gula Ratau, in Nature 2008, where they lesioned the IPRGCs genetically, they found major different deficits in light dependent pupil constriction. And I guess more to the point, there's a number of studies now, here's one very nice example from Manuel Spitzchan, have shown that pupil responses in humans involve an S cone opponent effect. So there's a color influence on human pupil responses. So we're interested in determining is that true also in mice? So as a starting point to look at that, we performed electrophysiological recordings similar to what I've just told you, but now rather than recording from the SCN, we're recording from the livery pre-tectal nucleus, which is obviously the main central relay regulating pupil responses. Now, for these studies, obviously, we know that IPRGC projections to the pre-tectum are important for regulating the pupil. However, unlike the SCN, where all of the retinal input is coming from IPRGCs, as you can see here in the top panel, IPRGC inputs labeled in green and conventional IPRGC inputs labeled in magenta, in the pre-tectum, there's a lot more input from non-IPRGCs as well. So this is a study from Bayer et al. 2020. So in that sense, really what we wanted to do when we were recording from the pre-tectum is find a way of specifically identifying those neurons that were getting input from IPRGCs rather than conventional RGCs. And so to do that, we designed a pair of stimuli that had identical brightness for the S and L cones in these red cone animals, but differed very greatly in their ability to activate melanopsin and rods, so somewhere in the region of a 500-fold difference in brightness. So we really wanted to produce the biggest difference we could with these stimuli because as some of you will probably be aware, our understanding now of the IPRGCs is there's lots of different types, some of which don't have very much melanopsin, so we really wanted to give ourselves the best chances of possible of being able to distinguish cells that even had small amounts of IPRGC input. That being said, then the challenge with the approach we've used here is, because we're also presenting stimuli with a big difference in brightness for rods, if we see differences in response to these two stimuli, how do we know whether they're coming from melanopsin or rods? And so the approach we took was to present these as discrete light steps from darkness, the idea being that provided they're sufficiently bright, even the melanopsin-low stimulus should drive transient rod saturation. And basically that approach works, so evidence of that. Here you can see a bunch of the neurons from the pre-tactum that we recorded that we designated as melanopsin responsive, and you can see something kind of very consistent in their response. So this is the same population of neurons tested across three different log-space light intensities. And essentially what you can see is the initial component of their response to the light step is pretty similar for the melanopsin-low and the melanopsin-high, but then a difference in the response appears over later components of the stimuli. And this of course matches our understanding of how melanopsin works in quite a sluggish manner. So it takes 500 milliseconds or so for the responses really to become apparent. And so we have a sort of nice intrinsic control here where we can look for cells where the initial part of the response doesn't differ significantly, but the later components do. And just evidence that what we're seeing here really is a melanopsin effect and not a rod effect. We can do the same thing in melanopsin-knockout red-cone animals and we found that that kind of sustained component was massively reduced when we did the experiment in those animals. Okay, so these cells that we identify that seem to get input from IPIGCs or show evidence of melanopsin-dependent responses, we call melanopsin responsive. And so we could then look at that population of neurons and say how the cone signals influence their activity. And so here's the average responses of those group of neurons. And once again, just as in the SCN, we found that you could split them down broadly into three types of cells. So down at the bottom here, there are some that show excitatory responses to both L and S-optin stimuli or in some cases just responses to one and no response to the other. But importantly, we can confirm these have sort of conventional non-opponent type responses by virtue of the fact that when we stimulate both ops in types together, they show a large response and when we stimulate the two ops in types in anti-phase, they show a smaller response. And then of course, by distinction to that, we can also clearly identify opponent type cells and I think I forgot to mention earlier, but we're referring to these cells that show inhibitory responses to L-opsin and excitatory responses to S-opsin. So S-on, L-off, we refer to them as blue-on by analogy to the kind of the color channel in humans that this replicates. So we can also find blue-on cells and yellow-on cells here in the pre-tectum among cells that seem to be getting input from IPRG seeds. And similar to what I just told you for the SCN, among the cells that respond to these stimuli and we find a small subset of neurons that don't reliably respond to these cone-isolation stimuli. But among those that do, again, about a third, in fact, in this case, maybe even slightly more than the third, are color opponent, most commonly blue-on, but also with some yellow-on cells in there. And then this little panel underneath is just to show you that these color opponent cells were all found within the pre-tectal livery nucleus PON and intermingled with the non-opponent cells. So they're certainly found within the part of the brain where we might expect them to influence or contribute to pupil responses. Before I leave this, also just to make the point that in these experiments, we specifically confirmed that our kind of silent substitution approach was not misleading us and resulting in off-target responses, for example, mediated by rods. So we wouldn't expect rods responses per se because we present, first of all, because normally there's almost zero rods contrast and secondly, because we present these at high brightness level where we would expect rod responses to be suppressed. But nonetheless to confirm that we aren't getting rod responses, we also applied the same stimuli in coneless animals. And as you can see here, where as we get robust responses in both the normal red cones and the red cone melanoxin knockout mice, there's no response to these stimuli at all in the coneless animals. So they're definitely cone mediated. Okay, so anyway, moving on, there's abundant evidence of color opponents in the PON. How does it influence pupil responses? So that's what we looked at next. So we use this identical stimuli, although we presented them in a slightly different paradigm. So for the things I've shown you previously, we're presenting kind of square wave modulations at 0.25 Hertz. Here to look at the pupil because the response is a bit more sluggish, we present these contrast steps where we give a step up five seconds and then step down for five seconds or vice versa. And what you can hopefully see here is that if we modulate brightness for just the L or the S-conoxin, we see very similar responses in the increases in brightness are associated with pupil constrictions and decreases in brightness associated with pupil dilations. So much as you would expect for the pupil, but no evidence of any opponents in here. And indeed, you know, further indication that that's the case is that if we modulate brightness for the two conoxins in anti-phase, so this S minus L condition, we didn't actually find any significant pupil modulation. Whereas if we modulate brightness for the two octins in unison, we get big, robust pupil responses. All together then what we found was that essentially the L and S cone signals are combining in an additive rather than subtractive manner to regulate pupil responses in mice. And actually that seems to combine with meta-oxin signals. So in this further panel here, what we're doing is we're providing a spectrally neutral change in brightness of the same contrast as this L plus S-oxin stimuli. And what you can see is that that actually results in bigger responses. And if you look at the time course of how this changes and compare the L plus S condition to the all-oxin condition, you can see that for the initial couple of seconds post-change in light step, responses are the same, but then the all-oxin condition gradually deviates, suggesting this kind of slowly building up meta-oxin contribution that I kind of alluded to in the electrophysiological data. So to summarize that first bit then non-image-forming responses, I guess what I've hopefully convinced you is that color-oponency is very widespread across brain regions that are targeted by IPI-GCs and play key roles in non-image-forming vision. So about a third of the cells in the SCN and about a third of the meta-oxin-responsive cells in the PON have color-oponency. Most commonly this is blue on, but there are also some yellow on responses in there. What I didn't kind of talk about in much detail is the potential origin of these responses. It's definitely known that there are some subtypes, at least one subtype of IPI-GC, the M5s that have blue on responses. And there's also some evidence that some M4 IPI-GCs, which seem to be synonymous with on-alpha-RGCs could exhibit color-oponent responses if their receptive fields are located in the middle part of the retina. But there's also the possibility that some central processing is underlying what I've just told you and maybe if people are interested, we can come back to that at the end. In terms of the function of color-oponency, that's become a bit... Well, in the case of the SCN, we have a good idea. So I showed you that color signals adjusts the KDM responses to brightness to provide a mechanism for compensating for the effects of clouds. In the PON, we don't see an effect of color, which is surprising given, obviously, the best known role of the PON is in regulating the pupil. We don't see an effect of color on that. So we assume, then, that these color signals relate to other things that the PON does, which are not that well-defined, but include effects on eye movements and potentially also contributions to circadian control. But anyway, moving on from the non-image-forming stuff, I guess we're still interested in these questions and pursuing them in the lab, but for the rest of the day, I want to cover another question we asked, which is, is this widespread appearance of color in non-image-forming processing just indicative of a general widespread role of color in all aspects of mouse vision, or is it something selective to the non-image-forming responses? So in terms of color and mouse vision, it's been known for some time, mice can discriminate colors. For example, this is a kind of forced choice, a wavelength discrimination task reported back in 2004 that showed that mice can distinguish ultraviolet from longer wavelength light. And more recently, a very nice study from Denman and Co. In a kind of immersive virtual dome setup showed that mice can perform color discrimination and particularly they can do it relatively well for stimuli appearing at high elevations. So above their head, but not so much for things appearing below the head. So there seems to be some sort of spatial inhomogeneity in terms of where mice can discriminate color. Well, I guess it's been a bit controversial is the mechanisms underlying this more wavelength for color discrimination in mice. So many of you will probably know that the mouse retinas kind of unusual in that there's this very strong gradient of cone-opsin expression such that in dorsal parts of the retina, most of the cones just express the M-opsin, whereas in ventral parts of the retina, you have some cones that just express the S-opsin and other cones that co-express both types. So there's not much overlap in terms of the regions where you get pure S and pure M cones. And so that's sort of thought to be a limitation on mouse color vision, although it's worth saying that plenty of studies have found evidence for spectral opponents at the level of the retina. I briefly mentioned a couple of them. Here's another really nice extensive recent study from Katrin Frankes group where they use kind of imaging techniques to look at color processing at all stages of the retina. And this is for retinal ganglion cells. They found lots of cells that showed evidence of the ponency to full field violet and green spectral modulations, particularly in the ventral parts of the retina. So these red colors here. So in particular, what they did was they found that four stimuli presented just over the cells receptive field and adjacent regions. They didn't find many cells that were opponent to that, but they have a chromatic bias between center and surround. So for example, in this neuron, you can see center stimuli, it's violet biased and surround stimuli, it's green bias, which gives your eyes to this kind of full field opponent type response. And that's also been common of other studies that've looked at opponents in the retina. It seems to involve generally surround mechanisms. So for example, here's another nice study, which looked at a specific RGC type, the JRGCs. And they found that in central and ventral parts of the retina, many of these have UV off centers and green on surrounds. And of course, because there seems to be a lot of this ponency in ventral parts of the retina, where there are very few pure M cones, it's, I guess an idea is emerged that maybe rods are playing a particularly important role in color discrimination in mice. So that's really, we wanted to look at the extent to which cones might also be playing a major role in driving color responses in the mouse visual system. And to get to that, we use the same kind of stimuli I've just described, but now we're recording in the mouse LGN. So just to remind you, the DLGN, which is where we mainly targeted with these recordings is the main thalamine cortical relay, but we also did some recordings from the IGL and ventral LGN parts, which are involved in this cadence system. So we applied our stimuli and in keeping with everything I've told you so far, we found quite a diversity of different types of responses. These are just for example, neurons. But basically what we found was you can find plenty of neurons that aren't opponent. So they either respond to just S or just L or sometimes a combination of both. But we also found a good number of cells that showed opponent responses. For example, yellow on illustrated here on the top row and blue on the bottom row. So here's data from a whole bunch of neurons we were recording that in that kind of approach. So in particularly we found over 400 that responded to our cone isolating stimuli. And you could see based on the response to L and S only stimuli and L plus S versus L minus S, we're able to cluster these into yellow on, blue on and one opponent on and off types. And this is just sort of population data from those groups of cells to illustrate the point that for the non-opponent cells, they typically very strongly bias towards one ops in class or another, most commonly S ops in, but obviously they're also L bias cells. So on the panel on the right, I'm showing contrast response functions for single ops in stimuli and the magenta is the preferred ops in and the green is the non-preferred ops in. So you can see very little response to the non-preferred ops in strong responses to the preferred ops in. And if we stimulate both ops in classes together L plus S, we get just slightly bigger responses than the L minus S because these are essentially driven by just a single ops in. By contrast among the opponent cells, they're also often quite biased towards the single ops in, but in this case, you can see the L minus S chromatic stimuli produces way, way, way bigger responses than the L plus S acromatic cone modulation. So strong indication these are color opponent neurons. And as you can see here, we find good numbers of both yellow on and blue on cells. So there's plenty of color processing going on in the LGN. We next wanted to get at the question of whether there was some kind of anatomical order to this because obviously we were recording across all different sub regions of the LGN. We first of all looked at across all cells regardless of whether they were opponent or not, whether there was a relationship between their anatomical location and their preference to L or S cone ops in stimulation. So essentially here, a minus one means pure S ops in driven responses, plus one means pure L ops in different responses. And what you can see is that actually there's a strong segregation here. So in the DLGN, dorsal regions are very S ops in biased whereas medial and ventral regions are more kind of equal L and S. This obviously to some extent at least matches the known retinotopic arrangement of projection to the LGN that have been reported previously. But more to the point here of interest, if we break down where you find the opponent and non-opponent cells, you can see that the opponent cells are very strongly localized to this medial portion of the DLGN complex and the non-opponent cells found in surrounding regions. Now this interested us because of course, if you look at where the IPRGC projections to the DLGN go, so obviously I told you at the beginning that originally we thought that IPRGC selectively innovated non-image forming structures. We now know that some subtypes of IPRGC is also projected to the DLGN. And that seems to align quite nicely with where we find the color of opponent cells. So is melanopsin or IPRGC input providing the dominant source of color processing in the LGN? So we use those same melanopsin, melanopsin high, melanopsin low stimuli I told you about before, and that allows us to identify melanopsin responsive cells. But actually we didn't find any real difference. So essentially among the cells that responded to our stimuli, in both cases about a third of the neurons are color opponent. The big difference between the MR and the non-melanopsin responsive cells is that in the latter, many more of them just didn't respond to our cone isolating stimuli, probably because they're more tunes to other things like spatial properties, et cetera. But in any case, it certainly seems to be the case that melanopsin input is not directly predictive of the presence of color opponent cells. We next wanted to get sort of more insight or to better confirm what we were looking at really is a cone dependent response rather than off target responses, for example involving rods. So I already showed you some validation in the pretext. Here we use the stimulus called rod mel that we designed to provide 45% contrast for rods and melanopsin, but zero contrast for SNL cone options. And we then compared it to 45% contrast stimuli targeting SL, L plus SL minus, et cetera. And as you can see in all cases, the cone stimuli produced robust responses across all the different classes of cells. We never saw any response to the rods stimuli. So under these conditions, which is high background light intensity, it's definitely not rods that's driving the responses. We also looked at the stability of the cone driven responses across different background light levels. And here you can see that those responses are present across backgrounds between about 10 to the 12 and 10 to the 15 log photons, which broadly correspond to the kind of light intensities a mouse would experience between solar elevations between minus six and plus six degrees. And also obviously aligns with when you would expect cone responses to be biggest. To look at whether rods signals become more important under these lower light intensities, we'd also in these experiments compared L plus S cone modulating stimuli with the all-opsin stimuli. So this is providing contrast, not just for cones, but also for rods, some melanopsin. And in none of those cells did we find a difference in response to the two stimuli. Again, suggesting that rods contributions across this range are negligible. Further evidence of that is provided here. So for the interest of time, I'm gonna skate through this, but essentially we designed a new stimulus set, which allowed us to selectively modulate rods S and L-opsin stimuli, activation in a kind of white noise type paradigm. And basically this kind of recapitulated what I've just told you. We occasionally saw some evidence of rods responses in the spike-triggered averages at low light intensities, but we never saw any evidence of rods cone-opponency directly, whereas we could clearly see L-S cone-opponency in the cells previously categorized as yellow or blue-on. Okay, so hopefully I've convinced you with that bit we're finding widespread and robust cone-opponency in the mouse L-GN. We next wanted to see, dig into the kind of spatial properties of these responses a bit. So we started off by looking at the just the receptive field locations of the cells, estimated in a rather crude manner by presenting flashing white or black horizontal or vertical bars at various places across the visual field. And although obviously these stimuli are in principle not optimal for modulating color-opponent cells, we're actually fully able to identify receptive field locations for the yellow-on and blue-on cells using this approach. There was a tendency for yellow-on cells to have on-type receptive fields to these bar stimuli and blue-on cells to have off-type receptive fields, although that wasn't always the case. But the important point was that we could estimate at least receptive field locations for a good amount of cells using this. And then we could look at whether there was a relationship between the retinal location of the receptive field or the visual location of the receptive field and the cone-based response of the cells. So on this central panel, then opponent cells are indicated by a star and opponent cells by a diamond and the color indicates their cone preference. So what you can see here is that obviously at higher elevations, there's strong SOPs in bias, which obviously corresponds to the SOPs in dominant ventral retina. And at lower elevations, the preference becomes more towards the LOPs in, as you would expect. And indeed, if we plot the average cone preference as a function of elevation, you can see this nice gradient. We didn't see any variation on the azimuth axis, although most of our recordings come from the locations corresponding to temporal parts of the retina. So there may be some variation if we recorded more cells out here. The most important thing I want you to get from this though is that if we look at the relative proportions of color opponent and non-opponent cells across azimuth or across elevation, we didn't actually find any evidence of a significant variation. So color opponent neurons are found throughout all of the regions that we recorded from, which on the elevation axis corresponds to approximately 30 degrees below to 30, 40 degrees above the retinal midpoint on the dorsal ventral axis. So the central 60 or so degrees of the visual field, essentially. And we also wanted to look at the idea that maybe these opponent responses were originating simply by random sampling. So I guess those of you that followed the color field will know that one of the proposed mechanisms involved center surround based sampling, where the surround just happens to sample from a different population of cones to the center. So this indirectly can give you some form of color opponentcy. In the interest of time, I think I'm gonna have to skip the details of this slide, but suffice to say that in modeling to produce a situation where we got the equivalent number of opponent cells that we found in the actual data, this required that the surrounds of the cells be very strongly weighted. So essentially the surround was almost as strong as the center and it resulted in the situation where essentially all of the opponent cells would be found in the dorsal retina, which is with receptive fields in the dorsal retina, which is not exactly what we found. In fact, we found almost the opposite. So we don't think random wiring alone can fully account for the responses we've seen. There must be some kind of selective wiring going on to account for the responses. So final little bit of data then, we more directly wanted to look at the receptive field properties of the color opponent cells and the cone specific receptive fields. So to do this, we adapted a projector system originally described in this paper from Annette Allen. So it's a five primary DMD based projector system. And by using this five primary system, we're able to produce stimuli where we could present flashing squares, where we, for example, presented an L-optin increment on an L low background or vice versa or an S-optin increment on an S-optin low background or vice versa. And so using this approach, we were then able to map receptive fields of some of the cells we found. And here's just some example, one opponent cells. So for example, LS on, L off, L on, LS off. But we're also able to find a bunch of cells with clear opponent receptive fields. So for example, here are two examples of yellow on cells. You can see L on and overlapping S off receptive fields in the same retinal location. And similarly, here are two blue on cells. You can see the L off receptive fields and S on components in the same visual location. So this is a bit different to what people have seen in the retina. Seems to be an opponent center based mechanism rather than that's kind of surround based mechanism that's most commonly be seen in the retina. We did also though, in these experiments, provide full field flashes of L and S-optin isolating stimuli to ask whether there are additional color opponent cells to be found. And indeed we did see that. So here's two examples of cells where our conventional receptive field mapping didn't show us any opponents. But when we delivered the full field stimuli, we clearly see L on S off or S on L off. So all in all, we found about a third of the neurons in the LGN has cone dependent color opponent C. Among that, there was a pretty much equal mixture of yellow and blue on responses. But also there was an almost equal mixture of cells that had this kind of opponent C for small discrete spatially localized stimuli versus cells that only showed opponent C to full field stimulation. So given a lot of what we see here in the LGN is a bit different to what people have been reporting in the retina where people have rarely seen strong center based opponent C. It's interesting to speculate that maybe there are sort of central mechanisms at play giving rise to this very widespread opponent C in the LGN. But nonetheless, what we can say for certain is that color opponent cells are strongly enriched within the LGN, particularly in the medial portion. Their receptive field span at least the central plus or minus 30 degrees of retina. And collectively, this is providing a kind of diverse capacity for color discrimination across the central and upper visual field that operates across twilight to daylight light intensities and that uses cones. So I think I should end there. I should just acknowledge the people that did the work. So Josh Moulin, current postdoc in the lab, Ed Hayter, a PhD student, Abby, a PhD student and Lauren, a previous postdoc all did aspects of what I've spoken about. Thanks. Well, thank you Timothy. That was very interesting talk. Thanks a lot for that. I'd like to remind audience that these are the questions that can ask directly to the chat or they can join us directly in a room through the link I just shared on the chat. So we're waiting for your questions. I have a couple already. So I have one from Tom Badden. Yeah. Is there a light intensity dependence difference between the L and S cone responses in LGN neurons? Okay, that's a great question. So it depends on, because obviously everything we've done is modulations with respect to a background. And so where we give stimuli on a background that resembles natural daylight for a wild type mouse, like the ratio of L to S ops in activation. When that matches what a wild type mouse has experienced at natural daylight, the intensity range is pretty similar. Now, one of the things I presented, I had to go through it slightly quickly because I realized I was running out of time. Apologies for that. For the white noise stimulus paradigm we tested, in order to make that work, so we could present rod selective contrast as well as cone selective contrast, we had to adjust the spectral composition of the background and make it much more skewed in favor of the L ops in. And there then we saw this shift where there were some intensities where we could see an L-driven response but not an S-driven response. So in simple terms, I don't think there is, I don't think there is a major difference in the intensity range they operate at, but you sort of have to bear in mind the background spectral composition. If you're looking at just responses to modulations on the background, it depends what background you're starting from. Fair enough. Thank you. I have a question from Christian Piller. Yeah. He says, great talk. Thank you. What is your idea of selective circuitry in particular for open-end signals coming from ventral as dominated retina? So in terms of selective circuitry, I think all that's really required is like, I think SCONE specific pathways, SCONE bipolar cells are probably enough and the surround, like a surround component can come from the mixed M plus SCONE population because of the way a ponency works as a subtractive mechanism. As long as there's a difference in where you have one pure S center, then the greater M bias of the surround would be enough to give you a surround type of ponency. Now, for the cells we find in the LGN that seem to have this kind of center of ponency, the reviewers of our paper wanted, like, were keen that we steered away from saying center of ponency because they rightly pointed out that maybe it's hard for us to directly distinguish surround and center in these experiments, which is fair enough. But my feeling, well, I have a suspicion that maybe what we're looking at here in the LGN is actually generated centrally or at least it need not come from the retina. So in that case, if you have like a retinal ganglion cell providing pure S signals or even S bias signals and another providing M bias signals, perhaps rooted via via an inhibitory intranuron or perhaps just an off versus an on ganglion cell converging onto the same LGN neuron then that could generate everything we see. I think that's an interesting possibility. We certainly know that LGN neurons receive input from more than one retinal ganglion cell. So that could definitely happen. In terms of if it was happening in the retina, I don't know, I'm not the expert here, but I don't think people have, I don't think there's like an M selective bipolar cell and obviously you've still run into this issue that I alluded to at the beginning of my talk of there are very few pure M cones in the ventral part of retina. Following up, sorry, I lost it, it's somewhere. I have another question from Tom Baddon. I mean, you kind of covered it already. It's regarding the open hand receptor fields that you showed earlier in the LGN. Are the green receptor fields bigger than UVs? I think he's referring to the third picture you showed when you had a big green. Yeah, so for all the cells where we, I didn't put this in my presentation, apologies. For all the cells where we had receptive fields for both S and L, we didn't find, there was no significant difference in receptive field size between the L component and the S component. I didn't actually specifically look across, I didn't specifically look generally, but obviously I can do quite easily. But I think it was just, yeah, just a random example. So obviously, in terms of receptive field sizes, you see a range in the LGN. Like, because of the nature of the stimuli we were delivering, I think we may have been slightly biased towards detecting cells with the smaller rather than larger receptive fields, but I think most of the ones I showed you had approximately six degree diameter receptive fields and certainly you can find cells in the LGN with like 20 degree receptive fields in some cases. Thanks for that. Sorry, I think there was a follow up. Isn't type one basic sort of M-selective? Isn't type one bipolar cell kind of L-selective? M-selective? I'll have to pass on that question. I'll take your word for it, probably, yeah. There is a bit of a delay Christian as answers the questions or the chat. So thanks for being interactive people. I'd like to remind you that I have one last question and then I will close the stream. So if you want to join us, do that now. If you won't be able to do that after I terminate the session. So we have, for you Timothy, we have one more question from Catherine Franka. Yeah. Hello, Catherine. Great talk. Can you elaborate on differences in color or potency for different stimuli, but for the similar roles? For example, full feed flashes versus full feed flicker and spatial ARF mapping. Okay. So for the full field stimuli, we, you know, all of the neurons we tested with our kind of like our standard square wave modulation and the white noise approach. And we found a very close correspondence in terms of a potency found under those different conditions in terms of the receptive field mapping and full field stimuli. We did find there were some cells, actually some of them cells where I showed you the center of potency or apparent center of potency, particularly the SON cells, cells that had an SON center and apparently an L-Off surround or sorry, an L-Off center. Often didn't respond to full field stimuli suggesting they have quite strong surround suppression. So they might even be double opponent. I don't know. It's quite hard for us in the L-GN to see evidence. We don't tend to see strong surrounds in the L-GNs. So, but they responded, yeah, they responded under the spatial stimuli not under the full field. So hopefully that answers your question sort of. Well, I guess I can only invite Katry to join us later on so we can elaborate on this. Thanks again, it was easy, it was very interesting. Very nice for you to put me up. For the other people, well, I guess I'll see you next week for our last talk is going to be hosted by Karen Carlitan from Maryland. So, thank you all, see you next week. We are offline.