Ancestral photoreceptor diversity as the basis of visual behaviour

uh in the space of visual system development and function um he started as a as an undergrad all the way through to a PhD student at at Cambridge uh telling me earlier it was in Zoology so that's sort of fit uh he then did his first postdoc with Le Lano before going onto the lab of Thomas uler in Germany um before moving to Sussex in uh 2016 uh so p has again made a lot of seminal discoveries in the field of visual system development it's probably easier for me to list the things he doesn't work on when it comes to color vision than it does uh to to name them all but essentially he's he's been one of the Pioneers in using zebra fish to study high high Acuity vision for example uh cone adaptability prey capture so on and so forth and he's going to tell us a little bit about his his recent work on color vision uh today uh Tom has won a number of awards uh the idorf prize the lisst prize uh amongst a number of others um so he's very well decorated in that space but also I think what should be mentioned is p is a big proponent of open science um and he was a very early adapter of bioarchive for example everything goes on there You' started open software uh platforms um and you you basically share everything you do including probably this seminar uh soon online right so I think that that's a really commendable thing for science nowadays is really sharing everything and and Tom's at the Forefront of that uh the other thing that Tom does which I think is particularly impressive is he started Trend uh for Africa or Trend in Africa um which is a nonprofit uh organization where he basically uh science training and communication in Africa I think there's over 40 academics who contribute to this it involves uh training in the field but also acquiring equipment that will allow them to you know start to work towards Cutting Edge uh Science Education in Africa and that's something that Tom started uh several years ago so that's also a very impressive uh feat I think um and something that Tom should be certainly recognized for so um with no further Ado um I will turn the the seminar over to Tom and we'll we'll hear all about his uh really amazing work so I will start this and I will give you a pointer it's all yours all right well thank you very much for the uh invitation and the introduction um I won't be talking about Africa or open signs and I will barely talk about cision because having worked on cision for so many years kind of gone off it because I think television is an ny phenomenon let me let me explain what I mean by that um hey clicking doesn't work this does so so this might as well be on the cover of a textbook on color vision right so that's that's kind of how we think of it so we've got our rods there one type um doesn't really matter what spectral wavelengths they're sensitive to because there's only one which means you can't do spectral comparisons um you can't tell intensity independent from wavelength that's the definition of color vision in most people's books um and therefore everything that we see at night is gray and you can test that go to the forest at night it's gray it's just correct um however when you switch on the light or you wait to the daytime then we get to use our cones which are less sensitive than the rods the cones as it so happens in the human ey come in three spectral flavors um that means that if we were to take the signals from these cones and compare them rather than add them together uh we can make a spectral uh comparison that is different from just an intensity measurement um and based on that um we can see color or we can tell we can tell wavelengths that depend from from from intensity and our brains interpret that as color um so that's great um So based on this what you might conclude then is therefore the presence of multiple cone types enables color vision and I think that's true if you have multiple cones you can do this that's fine but I think what we've been doing and I think it puzzles me how how this this hasn't been challenged before it kind of what we always say that therefore the purpose of having more than one cone type is to have color vision but that makes no sense um because they could have evolved for something else and then we just use them for color vision right so we've kind of flipped our causalities and it doesn't get challenged um and what I'm hopefully trying to convince you of today is that at least the original purpose of having multiple cont types is to enable something as fundamental as Vision so aspects of vision that are not about color um and they are specifically important in the water where Vision First evolved right so All Eyes almost all eyes across animals have evolved in the water first and one day they came out of the water in a species such as ourselves and then you just work with what you inherit from your Aquatic ancestors and I think that's also the case for our vision so um if we take our little diagram here of our rods and Cone rod and cones and we take our combination contrast we feed them everything into the vision box and we say this is vision and then once on multiple times probably in the history of evolution of vision this system has then become co-opted to also allow color vision but this is like an optional Branch you don't need it you humans use it but some species might not but quite fundamentally this just happens because the spectral information that comes in through these different channels is not identical it's actually quite overlapping but it's not identical therefore if for any reason other than color vision you take two of these cones and you contrast the signals you're going to end up in aely with a spectrally complicated signal some sort of opponency will happen and if it's here used for something useful you could also use it then to in addition get the sense of color vision out of it so I think this is sort of the a more plausible history of color vision um and now I need to convince you of that so here's water uh and the rainbow and hopefully what you can see is that this rainbow is very rainbowy in the top and not very rainbowy in the right there's nothing magic about this this is just physics this is the interaction of light with water depending on the wavelengths I.E energy of the photon that you have you interact with water stronger or weaker there are some wavelengths that penetrate water better than others that means that well color or the wavelength composition is informative about all kinds of stuff in the water and this would have been the evolutionary pressures driving everything to do with how how we set up our detectors and um now what we've done here um is a very simple thought experiment so here paint in the background what you see is the average spectral content um of light and shallow fresh water so this it's not muddy it's just like basically res fish it's like half a meter deep it's clear water it's lovely stick spectrometer in there and just see what's the wav length composition and hopefully it's and not is that flipped here so normally what we plot is we plot red over here and UV over here but here this time I red here and U be over here because this is an energy axis which is invert length just to make my argument here and what I ploted on top of it is the invo spectral sensitivity functions of two of the photo receptors of the Ze fish The ancestral red cone and the ancestral cone and hopefully and we've measured these using to phon with all kinds of tricks in that paper so um hopefully what you can appreciate is the red cone is basically a copy of all the light right the red cone is a light detector it just sees the light that's there the UV cone is doing something completely different the UV con is very specifically trying to get the base that is actually UV okay so you don't really have a red and a UV system you have a gray and a UV system if anything um now starting on that premise we might be able to then just plot these two counts in in this arbitary space of how many photons you have the red gets a lot more Photon than UV just because there's a lot more light when you measure all the light um but the UV uh cone gets the good photons the high energy photons um UV is the highest energy light we can see if you go higher energy that didn't when it doesn't get through skin doesn't get through anything right so this is this sort of biological limit of how far you can push this this is the best of the light but this is the most of the light um and you can use those for different purposes so for example you could say if I can count a lot of photons that means that I can shorten my integration time therefore my photo receptor becomes fast therefore if I want to do fast Vision You Really Want A system that catches all the light like the red cone the UV cone is terrible at speed but with the UV cone is really good at its game the reason it's good at game is because the energy is so high that you don't run into thermal noise right so this is one of the key things there's a reason that we don't see infrared infrared is really low energy light and it's so low that the energy coming in to trying to photocon convert your your opsin well your your rmal is so low that it gets matched or and and swamped by the thermal energy of the system right so if we operate something like 38 degrees in your eye or whatever um you just can't push the Ops further into an infrared region you will just start seeing nothing you run into theral noise um that problem exists on this end it doesn't exist on this end on this end the problem is cancer right so as long as you can keep your DNA Dage at Bay um you want to go that way and that means that a UV cone really hasn't got much thermal noise at all right you could use it as a photon counter it would be fairly clean and that means though that even though there's not a lot of phons in the UV you can take your gain and you crank it up all the way as much as you want you're not going to run into Noise Okay so you've got a completely different detector system you've got a fast low gain system and you've got a slow High Gain system and this is not the properties of the photo setor yet this is the property of this is just what inevitably happens if you build a detector that works in a certain wavelength range um and it's not just invertebrates it's true pretty much ACR any species you look that has multiple Spectrum receptors they always like that now um what then becomes quite interesting I think in verteb at least is that the vertebrate eye of course doesn't have the photo receptors in isolation has them in the circuit and the circuit does things and the first thing that the circuit does is it takes these photo receptor signals which are already different and it put them apart and makes them more different so for example here the photo receptors in ey as most ver eyes are contacted by horizontal cells horizontal cells do interesting things if you take your horizontal cell uh well if you if you record a red cone this is a red cone glutamate response to the stimulus you can see it follows the low frequencies and eventually gives up and then you block the horizontal cells nothing else then you clip The High Frequency response right so what that means is that the the presence of the horizontal cell through whatever mechanis I don't want to get into that makes the cone faster right that's the purpose of this system um if you do the similar experiment for the UV cone we didn't use a CH here we just used the flash of light it goes the other way around the cone is normally quite slow um is is slow in the presence of horizontal cells and when you block them it gets quicker okay so the horizontal cells make the red cones faster and the UV cones slower right that take the intrinsic differences and make them bigger suggesting that this is a difference that you want to work with Okay so we've moved the dots a bit then let's talk about space so um if you open my textbook and you look at how a cone is supposed to respond to light in space it's supposed to have centers suround organization you hit the exact the outer segment of the cone it responds negatively because these are off cells so the decrease activity but if you hit it near the center but not on the center it should go actually the other way around because there's an opponent circuit an inhibitory circuit that goes the other way around um so and in redes we can repr produ this right so here's a very simple experiment you've got little red bars we recording red cones and it keeps missing and here hits it and then misses again right so this is just a central receptive field it's quite small it's couple of degrees makes sense in the baby have a fish because the ey is small um but then what you can do is you can contrast invert that stimulus and play the same game and you will find a place where the black bar sits exactly on the cone and the Red Bar sit exactly not on the cone that causes the like that hits the opponency and then get you flip the signal you turn it into an onell you know so this this this is textbook stuff this is what should happen with the con it's a center surround unit what Center surround does is it makes when it decorrelates in space it basically it makes it makes your ability to encode edges and space better yeah and the entire visual system is kind of based on that principle it's just it starts in the cones you try that in in a in a UV code it just doesn't work right so here this and this is equivalent except that we're recording from a different photo receptor and hopefully you can appreciate that the receptor feels huge or something like 10° this is crazy big for a photo receptor bear in mind that the outer segment of the photo receptor is not 10 degrees it's maybe one degree okay so something's going on there I think it's chromatic aberation so basically you can't focus all W flinks the same and the eye just lives with that and then that becomes basically a massive detector um but also we we just can't oops I haven't got this here we can't get a surr out of this so we've tried this experiment doesn't work at all we've tried all kinds of clever other experiments doesn't work we've never managed to invert the polarity of a UV con based on a cleverly designed spatial stimulus implying that either there is no surround or it's so weak that we can't detect it okay so what that really implies then is that that the way there's a tiny little thing the center and the surround of the uh of the red one yeah they don't seem to be in the same place or near it's a different cone sorry oh it's a different one okay go no yeah good point though um yeah so they just do completely different things then a different position speed in a different position game and they're in a different position space they're basically two completely different ways of sampling the world yeah and if you sort of mentally translate into what that would do red is basically your detail channel right it sees everything as well as it possibly can be seen pretty much like if you take a a grayscale camera and you just point it on under the water and you have enough pixels that's roughly what it'll tell you plus some processing you get a nice black and white image U image by itself will also give you black and white image right it's no color computation it's just it looks at the UV bit but it'll it'll it'll have a really high gain it'll be quite slow um which also drives the gain up again um uh and it'll have terrible spal solution right it's just it's trying to look at a completely different thing it's a very sensitive Channel but it's not not a detail Channel okay so um it seems then that this this is sort of the principle of feature selectivity in the visual system right like the idea usually is that you have a where the photo receptors detect the light and then it goes to the bipolar the gang cells goes into the brain does kind of things and then you get these features and these features might be you've got a big receptive field you got a small receptive field you've got one that is very sensitive one that's not very sensitive well it looks like the cones are kind of doing that for free right like the visual system starts with these features but this is just two of them okay so you might call both of them by itself they brightness channels are black and white Channels with different properties now what about the other two channels that fish like many surface dwelling tell or basically anything that swims in the water that evolutionarily never left the water has four cones that's State there the green and the blue and a while ago we did similar experiments for the spectral sensitivity function that show you in the beginning um for these and turns out that these are color opponent at the firsts of vision right so basically these guys through the horizontal c network um they don't tell you greenness or blueless they're just a color opponent signal they respond positively one W lengths and negative Le other wavelengths now if you've got a color opponent system um the mind immediately goes to color vision right we know that the currency of color vision is color opponency and that's true in the humans as well um so that's how wrote up the paper and published it and thought that was great but actually coming to think of it I think that was shortsighted I mean yes it will give you color vision if you if you choose to use it that way but it's so much more useful for much more basic things so if you think about about this this gradient again the gradient doesn't just work down it also works along the water column right if you stick your head in the water look along the cor Reef you're going to see this you see color where you are and you see bless where you're not right and it works for the fishes well that fish is near that fish is far okay so the spectral composition of an underwater image tells you about distance now I would POS it to a animal that swims around the water unless it's a very specific purpose distance is more important than color right because distance allows you to interact with the world long before you understood what you're interacting with stops you from running into things it stops pred getting you stuff like that so working on that assumption let's think about what these wavelength channels would do with color well first of all red sees all the light therefore it sees the Fest right there's just no way around that UV actually doesn't travel very well in water at all therefore if you only have a UV Channel you don't see very far so if you only have a red in UV Channel you will get two distance shelves um just for free but then these color opponent signals they will actually Travel Intermediate es and because the color opponent system don't have an identical zeroc Crossing fish two different zero Crossings it will give you two different distance shells so kind of just for free by The Way That these channels are set up you will get four distance shells if they're space like this I don't know but something like this and it would be crazy for fish to not have evolve to use this it seems like the cheapest way of doing distance um which incidentally should work without even having any decent spal resolution like this you could do with the single Pixel detector that just measures different wavelengths you could do this in syation like that so if we then think conceptually as the input Channel and Z fish not just as four color channels but basically as two completely different detail uh two different gray scale channels on the extremes plus a couple of different distance channels here um I think it makes more sense and I just want to point out so what I've done here this is a GoPro image so red green blue human optimized right you stick it in the z natural habitat this is actually taken in India um and this is what it looks like when you look at the when you just pull the channels this is what what it looks like when you take a red Channel and your green Channel and you flip the green Channel and put them together right I didn't do any magic here I just made a color opponent a fake human Centric color opponent signal and already you can see the foreground and the the background just disappears right so these color opponent computations they're really powerful tools to get distance information um they're basically for ground enhancers now working on that so if this is our basic model it's not very satisfying though is it it's just a bunch of boxes so how can we unpack this and really understand what's under the hood so um my proposal is this we have in the Rea of fish but also in human still two main systems we've got the ancest red system which goes together with rods you can ask me about why I think it goes to R later just take my word for it for now and on the other hand you've got the U system which goes by itself and they give you the high resolution versus high sensitivity systems and they can individually Drive certain behaviors so for example opot the kinetic response type things are generally associated with long W link vision and I think it's actually just the red cone circuit um pre capture in zra fish for example is a UV circuit right so there are examples where really the cone seems to be the key thing um and I would argue that this is all of vision um why would I say that um I just a detour just to make the point that ancal red cones in Z fish are both red and green cones in humans and the blue cones in humans is the ancestral system right so humans don't have the green and the blue The ancestral ones they've got the ancest red UV ones um and this is important because it's true for pretty much all vertebrates here's a little F tree of most major branches of the vertebrates up here you've got squidy little fish more interesting looking fish then we started to get into lizards amphibians and whatnot we put the MS down here and hopefully what you can appreciate is that if you look at the rod always there right there are reports of rodless species like a chameleon was T to be rodless turns out it's wrong they just don't have very many so they haven't they been miss so I'm not aware of any vertebrate that doesn't have rods not even hafish which barely have an eye they still have rods um then red cones is the next best one almost all vertebras have red cones the only ones where it's unclear is lon fish live very very very deep silicons very very weird and very distantly related to anything that's life today and hack fish barely ey okay so basically red cones are always there and red cones will forever shift the spectral sensitivity to match whatever makes sense so for example in the human case there's the red and the Green version but it doesn't stop them from being ancestor red con they're just different versions of it and that happen actually a bunch of times in evolutionary story of animal okay so the red cone is almost always there next best one is the one it's like the red cone except you've got some special cases like seals whales basically large aquatic secondarily aquatic species um that have losted and again these these wees and the Sharks and race which are a little bit well they're basically the size of a dolphin similar problem um so almost all vertebras have this system and that is entirely untrue for this system the blue and green cones are missing and pretty much half of the lineages none of the mammals have them there's a report of platypus I think this is probably wrong um and you can ask me in a year or two after we've checked it um if the Platypus is wrong all of the mammals are done there's no survival of this system and there might be a new occurrence in the in the stunard which is very unclear basically mammals don't have this system can you clarify which system we don't have with mammals because I mean we do have green and blue c and justari so that that's um that's what I meant here that our green and blue what we call to be the green and uh what we call the green cone is an ancestral red cone with a green shifted dos and what we call the blue cone is an ancestral UV cone with a blue shift like for example L they all have this but also you know cats dos horses basically frogs crocodiles it's uh it happens quite often that this UV Ops Shift Into the Blue range but the photo receptor itself is the ancestral U V1 yeah so um and hopefully organize them a little bit from live in the water live sometimes in the water don't live so much in the water and hopefully you see there's a bit of a pattern there it's like it's a water system it survives in some animal that don't live in the water like some birds some lizards I'll get to that if I have time um but basically it's an ancestral water system and what I would posit is that this is not used in how we think cones are used I'm saying that this is used as a regulatory system we know in the Rea we've got horizontal cells as lateral connections we've got Ames as lateral connection I posit that these middle cones are part of that system they're the regulatory system they're not the drive system and I know this is heretical so I will justify it so how do you test this let's go back to the distance so there's a lot of aspects that you could test but we thought the distance ones maybe the sexest to look at so um how do we think about distance well actually a color disappears with distance that's quite clear but color is the EP phenomenon of a much more simple phenomenon in physics and that is the the narrowing of the wavelength with this right stuff that's near you have a lot of wavelength available to make spectral judgments that's I think do look colorful but really it's the whiteness that disappears in the front stuff can be white but if you take a white T-shirt and you start swimming and the distance it would be a blue t-shirt right so the spectral width is the is the easiest to measure factor that changes um so just to illustrate how that it's really powerful right so here's an RGB movie these are artifacts these are not real right so you know that so basically you've got a little um uh underwater movie again where LIF what we can do is we can just extract the brightness component by just adding all the channels and yeah us being clever humans we can tell where's the foreground and where's the background right but it's it's it's cognitive you need to think this through right if you just look at the whiteness so if you just look at how road is the Spectrum based on these three pixel like the RGB values that are available in the video I think it's really obvious that the forground here is super high weightless and it just very obviously decreases and um I like to point at this particular home of grass or sewe or whatever it is right it sort of comes out of the nothing so the whiteness really is a very powerful and easy way to signify distance or lack of distance so how would you measure this so if you are a fish um you could use different photo receptors to to pick it up so let's let's start with the ex Spectrum so here's the hypothetical Spectrum it's not what actually looks like this the um so if if this is super near you as you as you go further away there the sh the Spectrum will shake towards a much more narrow one typically bias towards the long wave length right so this is near this is far this is roughly what would happen um if you have a spectrally broad non opponent system like one down here it'll just copy the Spectrum more or less right so you lose exitation there's less light to integrate Therefore your signal goes down but it goes down a bit right as you make this narrower you just lose a little bit of signal but the opponent systems they're more interesting because in the beginning when you've got a broad spectral input the negative and the positive will cancel and you get nothing right so opponent systems are silent under white light but as you get rid of some of the wavelengths but not others that's when the opponent systems kick in right so the opponent system actually start to um be more and more active up to a point then they go down again right so if you have an opponent system in your eye and a non opponent system in your eye and you just contrast them right you basically say I'm going to take my non opponent system to drive everything and I'm going to use my opponent system to regulate the system then all of the entire output of the visual system will become white Biers like stuff that you do will work really well if it's white and stuff that you do uh if it's not white your responses are going to go down your retinal output is going to go down right so you don't need to do clever nitive computations or anything you just need to take an oppon a non opponent system contrast them and make that the output of the eye yes is there an assumption here that the reflectance of the objects that you see uh is going to be very broad for example if all objects were blue this will not work you're completely right yes okay so what is your assumption there because I wonder whether actually you're assuming that it's flat that there are objects of all colors using the words that we would use for color in the in in the sea in the ocean in I don't know this is fresh water this is fresh water um so you assuming that yes but actually if you so um if you walk around nature with spectrometer and you pointed at things almost everything gives you some sort of broad spectrum with a little bit of a kink this way or this way right so if you point it at a leaf it goes this way if you point it at the sky it goes this way but it never goes like this right whereas if you pointed at a children's toy or something perfect red then it goes like this okay right so it's it's it's man-made objects that are like this and the occasional occurrence where animals or plants have evolved to try to be narrow like some plants for example some feathers of birds they are narrow but they come much later in evolution than any of this would have been important and actually there's very little evidence for the existence of structural color on pigmentation of skin or anything in in in old things right so the the first clear evidence for the existence of should we call a purposeful color um is um something like 300 million years old uh but the I would have evolved 550 right so this it's uh it's quite a difference but you're completely right yes it doesn't work if everything is monochromatic then this fails um okay so how do you test this um but we have tested this and I'm going to give you the punch line before I show you um first of all using Ze fish we see that fish Vision Ze fish vision is white biased and it's not a little bit white biased it's crazy white bias if you give a zeber fish white light everything works better two I'm going to show you that the red and the UV system is necessary and sufficient to drive all of vision so when you take it out vision goes away as long as it's there vision's fine the green and blue system completely different you can Che out the green and blue system vision is fine and actually some aspects of vision improve um or the gain goes up improves maybe the wrong word so the green blue system is not necessary nor sufficient for vision it um and it's actually packed a mutually competitive way to inhibit Vision um four I'm going to show you what that means in behavior and five if you've got time I'm going to talk a little bit about Evolution so here is finally your standard Ze fish experiment here Ze fish that's the brain lots of neurons they're all expressing gamp and we have a two photo microscope to record the activity uh we scan this bit of the brain this is the tectum this is the other tectum and fish all the information goes across right so if you stimulate from here all the activity will be here specifically in this Red Zone which would be the retinotopically aligned bit of the brain from where our stimulus is then we've got a custom stimulator that can do it can give it fish white so real white or all kinds of combinations thereof um here's the actual stimulus that we use so here is the spectrum of light in nature that's the average one and here's how we've placed our spectral channels um um in power and wavelength hopefully you can see we're trying to approximate this white light right so the idea is that if we pop on all four spectral TRS at the same time we're approximating the natural Spectrum I'm going to cool this white um if you take if if you do the exact same thing but you take out the UV I'm going to call it visible this what we would use maybe and but this for example would be you be only stimulus and the experiment now is very simple having set this up we're just playing a stimulus like this right we're showing some great things that start to move then we're switching off the light we're showing some dots this would be like a pre capture type stimulus and then we're doing a contrast inverted thing of the same thing on with the dark dots and we're just seeing what can we get okay so this is now in white and here just a bunch of example cells so for example this is a cell that responds to grings every time you show GR thingss it responds it doesn't respond to anything else it's a gr cell simple this is a DOT cell responds to light dots this non cell if not cell you find what kinds of things most cells are quite selective for one of them sometimes they get two but that's actually an edge Edge case most are pretty selective now where in the brain are they okay the grating cells are in the retinotopically aligned area of the brain where they're supposed to be the dott cells are also in that part of the brain the on and off cells are all over the place it looks like they um either they don't come from the eye all of them because the fish has other photosensitive bits or if they come from the eye they get distributed in such a way they reach the whole brain which what the two it is I don't know but hopefully just the location of the cells shows you very clearly that this is not the same system it's something else is is the full retinotopic within the area yeah so the screen is roughly where the yeah okay but I took away my punch line but here's my punch line we take the white we delete the U and leave everything else intact so we just taking away this tiny little bit of light here and you see how this breaks right the great thing in your dot response yes there are still cells but the proportion of cells compared to what you get with white is pitiful yeah whereas this on and off system isn't I mean it's affected a bit but not so much much okay so when people study zbra Vision do they usually use a UV monitor uh some do it's rare okay but if they don't they're missing a lot it there's a criticism that would be valid yeah you said it not me okay so but then you might say well maybe you Fe amazing maybe you Fe the remaining 85% here and then we solve the problem um but that's not true because if you only show you we you also get P for resp it light dots go up a little bit and I think that's because PR capture and Z fish is a very UV heavy Behavior but even that is less than half compared to White right so even bra capture isn't UV only it's just UV helps kind of system so what's going on obviously we tried more colors more color combinations and I just want to make the point just how strong is this for grating is extremely strong for for for Dots Whenever there UV you get some activation but basically you get this is white works really well anything else works very badly the best thing after white you can lose this which is white minus the blue they not the blue doesn't seem to hurt as much but even that hurts right and it's not a lot of light so clearly this is a white biased output to the brain at least the brain representation of the stimulus is white biased um dots is completely lost I mean this this is a few percent and you can work with this right but this is not the native response this is the resp um so what's going on um well this being fish can do things that you can't do in other animals and that is you can break things without getting into trouble so here is a wild Ty fish what do you think would happen if you were to genetically upate the red cones acutely so there's no developmental anything you just take a fish take out the red cones what should happen this okay almost complet lose the grating response um and actually this is so the significance is always versus zero not against here right so this is significantly more than zero this is not sign significantly more than zero um the dots is still significant and there are still a few dot responses but hopefully you can see that this is much worse than this okay so the red cone is basically most of it the green cone is not most of it you take out the green cone not only are responses still there whether they go up or down you can debate but actually if you just these maps are not just the number but the number multiplied by the amplitude of responses and if you take that as a factor then actually at least the light dot responses go up when you kill the green cone right so the green count isn't there to drive these responses if anything it's there to stop them a bit can you remind me the ratio FR plus um in the baby ze fish they're almost one to one to one to one not quite but roughly there right so the number of cones are not crazily different um here's what happens when you take out the blue cone it's a bit like the green except that now the gratings are a little bit more potentiated than the dots but again it's still there this fish can still see they got the UV cone it's almost like the red cone it's just the other way around a bit now the great things are a little bit alive and the dots are less alive even though they come non significant okay so red and UV matches green and blue do weird stuff okay so but we do have some sh left both in the right and the new V so what happens if you take a both then it's completely toast okay so the red and Cones are necessary for this certainly and I'll show you later it's it's the whole visual system um if you take out the middle tube you clean up the texto right get potentiation and both and everything looks pretty much intact so green and blue cones not necessary nor sufficient for vision they must be doing something else and actually what's kind of cool is there's there's an additional phenotype that I didn't talk about so when you play in a control fish dark dots they don't love it there are responses is not many um but on a map like this they come out as basically nothing if you take out this middle system they come back right so the Ze fish has some sort of circuitry by which it takes its middle cones and suppresses the dark dot response Maybe maybe different species of fish has the light response whatever the point is that you take out cones and stuff comes back stuff stuff gets more not less okay so remember I talked about the white effect in the beginning so let's look at this at a bigger picture so this is now many colors we're looking at gring responses and we've always got the white uh the controls and the double the green blue double upated and hopefully can appreciate that you starting to potentiate especially in the long wavelength range um the responses to gratings so for example if you've got a control fish and you give it red gratings which is sort of the standard stimulus used in the field then you get this little hump here and then if you take out the green blue cone you rescue it almost to control levels okay so the presence of this green blue system causes the white effect yeah you take it out the Ze fish vision is no longer white biased um so we wondering how can we map this onto the different C the nice thing is that we have special sity functions of the cones and we can therefore take each of these stim convol them with our special sensitivity functions and make predictions of how the cones would respond to the different colors and then once we have that we can use those cones and fitter model and basically predict which cones probably underly which kind of things so here is for example a model trying to explain in a green blue up lated fish the grating response which is this line um so so the way to read this is here would be red cone only here would be UV cone only here would be both and these would be opponent and these would be both negative right and hopefully you can see very clearly that all of the good model fits end up on this side which is basically red only and maybe a little bit tilted upwards which tells us that the system is mostly red with a little bit of UV mixed in for fun okay and then we can see given that we've now looked at this we can see what's the remainder what's the difference that we're getting here because that must come from these two cones if you do that and you fit the model then it folds into the bottom quadrant which is green minus blue right so it's not like the green and blue system are working together they're working against each other to regulate the red UV so like the difference this slide and the one before um in so in this slide we're fitting the green blue upated right so basically what's left up they take out the system and that you can explain with a red cone fairly well whereas this you can't explain with a red cone you need the green blue and you need the green blue in this sort of configuration green positive but blue negative so it kind of suggests that we've got here again this system positive and then here a regulation and actually these two are fighting with each other for the right to regulate so it's green minus Blue System okay so far so good this is great things what about Dots Dots is actually even clearer the DOT tuning functions if you look at the updated version you get a very clear R plus UV a little bit wise towards UV but basically both cones are important and if you look at the remaining response in the in the control it's very clearly green minus blue so again same architecture gives you a slightly different tuning function because the weights are different but the architecture is the same now what about Behavior so we thought well the the easiest Behavior we can do is nothing we just put the fish in the dish and we watch them let's see what they do so here's some control fish this is probably the most boring bit they swim around and they do this with the bouts and then do a little turn and then they do some bouts you can count the bouts you can count the turns you can see how many they do you can do this in the light you can do it in the dark and this is the non phenotype of it's not phenotype this is what fish do when you put them in the light they don't move so much you put them in the dark they move a bit more the same so now we take out the cones and what I think is very cute is that we basically get a light phenotype and dark phenotype right so the red cones no matter what you do if you show them light or if you don't show them show them the light they behave like a control fish that is in the dark so it's almost like they don't see right they always think it's that the UV con are the other way around if you take out the UV con they behave no matter what you do as if they are in the light so it almost suggests that the red UV system has a has a push pull on on the on the behavioral decision as do I respond as if I'm in the light or in the dark so maybe it tells the fish it sort of an underlying behavioral stage type thing something we're looking into um if you're killing the Blue by itself it's a little bit like the controls they're all right in the light and then they s of almost reach the dark state but not quite the grain are completely broken they go the wrong way around it's not significant but it's certainly not doing the right thing but what I think is my favorite is that if you UPL both you kind of rescue them right so the control to remember goes from here to here and the green blue from here to here so they're almost like the controls and I think what the key thing to notice here is that the double upated fish are better than the individual upated fish you take out only the green system they B you take out only the blue system they B you take up both they're kind of okay yeah so it's like a misregulation when you have only one out but when you take out those right again there is a very clear phenotype of the double lated ones though and that's this one so instead of counting the number of turns you can just see how much they turn like do that turn like this or do that turn like this like a good control fish and the double up later ones the terrible they do this yeah it's a very striking phenotype it's a lot of fun to watch so that can do fishy things but they're sort of a bit overe excitable right they it's it's almost as if they're lacking the regulation to to do the right thing or to to do a modest thing um so what happens you start stimulating them optim motor is fairly easy Behavior test so we just adopted this setup published from Flor lab um and basically you don't opot response you watch the fish you give it some Stripes turny quantify controls are very good at it red up L fish are very bad at it so far so good you V blad the fish are very bad at it makes sense the green blue um they also kind of bad but notice that they're better than these guys um but where it becomes really interesting is uh you take out the green by itself absolutely terrible take out the blue by itself not so bad and notice that both of these are significantly different from the double ablated phenotype right so again it's a sort of rescue of the green effect the blue isn't so involved doesn't matter so much but the green really breaks it and if you want the Green from not breaking it you also need to take out the blue so it's the green blue system they're talking to each other and if you take out one it goes wrong okay so this is the phenotype I just talked about where where they turn too much this is spontaneous well it turns out that if you do stimulate them it overrides it I mean some of the mutation neotypes do do wrong things a little bit but but the uh scale is completely different R so so they're like 10% turn too much it's just the numbers are so high that it becomes out significant right but hopefully you can appreciate that the the basic phenotype is essentially gone right so these green blue uped fish in the absence of systematic stimulus they get ere excited then you give them something easy to do like a phot sub Stripes they okay they can do it yeah okay so that's op theot respon I just want to point out so I haven't got the data here because it's literally coming right now but this was done in a setup that doesn't do UV this is a red bean blue setup it's a standard setup with a standard screen and everything and we thought what's what's up with that and we thought why are these so bad they're meant to be better so what we have done now is we have set up an aquarium outside and we've put it in the sun which sometimes comes out in this country as you might have witness um and we put a big conveyor Bel type thing under the tank and a camera above the tank and like try basically a natural type thing and if you do that not only are the green blue are these dou BL fish um completely the same as These Guys these two are still terrible but what's actually really cool is if you then increase the water depth the controls stop responding and the green blue up BL it still respond so basically this distance idea that I was pointing out it looks like it actually plays out even in something as city as an opom mot response right it's not it's not just a break capture thing over way you might expect that this is important like something as fundamental op the motor it seems to work we just need the numbers to make a nice plot and see if it if it comes out a significant okay so this is optimo um but we wanted to test if it generalizes across other behaviors so here is perhaps even more uncensor Behavior as photot taxis right photot taxis is you switch on the light you check out if the fish go to WS or away from the light and turns out that in Ze fish um it's a special behavior so if you give them your V light they go towards it if you give them blue light they don't care green light they go towards it again and red they love it right so that's the spectal tuning function of your um of your controls you take out the red cone this time it doesn't matter so this is one Behavior Where The Red cone doesn't seem to be so important but you take out the UV cone then it breaks it seems to be a UV cone Behavior but again where I think it becomes really interesting oh yeah and I should point out that photo toxis isn't part driven by the cones but part driven by melanopsin system so that's intrinsically photosensitive gang yeah um but those get input from cones still so it's still the cones are involved um but where become really interesting if you take out the green cones they they love it they love a bit of attractive even at the wrong colors they just go they always go for the light you take out the blue they hate it they want to don't want to do it at all you take out both they're fine yeah so again even in something as ancestral as phot taxis which doesn't even need an eye you don't need an eye PHX you just need to be able to measure how much light there is and start moving yeah even there it still seems to be the case um so it looks like you've got this regulatory system in place again so what I um would then propose is that this is probably not right at least for fish um and you've got some sort of core architecture where you've got this the core system sometimes the red is important sometimes the UV is important you can recruit Metals in circuits but there's a core drive and there's a regulatory drive and the regulation always comes from the GRE blue system and it seems that it's always in competition with each other such that when you damage it half it's bad if you damage both it's fine again is now um in Fairly rapid closing hopefully let me give you some perspective on what I think this might mean for animals that are not zra fish so ancestral State four cones quite clear I think they're Inc composite so this is this is the model if you look at basically all of the fish they more or less old habit sharks and rays are weird and many of them live at depth and have they've gone through weird bottlenecks so if we ignore the Sharks and Rays you can probably say that all the primitively Aquatic lineages have the system um then you've got this Branch this is Branch is everybody that ever went walked onto land that isn't a mamal okay what happens here is they almost all have it except for weird secondarily aquatic species like Ro um but what happened if you the Tetra pods come here tet pods are things with fors um they already have lungs they already big eyes stuff is changing but they still live in the water for quite a long time then they start becoming amphibus one branch of that is amphibians and the other big branch of that is basically everything that would become reptiles birds that sort of thing and they also stayed aquatic for ages so between tetrapod and this line we've got 100 million years of buffer window to adapt to the idea of I don't live in the water anymore let's see if I can rearrange my visual circuits to use this ancestral system for something new and they seem to have all managed to keep it and the branch that's broken off here is the mammals and it looks like the mammals didn't have this buffer zone the mammals very quickly went from completely aquatic rot here to completely terrestrial um so I would posit that it was probably related to the fact that mamals went very rapidly onto land it wasn't time to evolve new strategies to use the system because it is a subtractive negative inhibitory system is probably bad for signals and noise so if it doesn't give you benefits you should get rid of it and I would posit that these early synapses which eventually would become mammals L and just to make the point I've got this B notal bot on in here you might have heard of the nocturnal bottleneck the is the idea that mammals lost their cones those cones during the age of the dinosaurs because they were nocturnal because the dinosaurs were dial they would eat the mammals that would come out during the day so everything was nocturnal at night you don't see color but there's not enough light to see color there you lose the ground that's the ancient Theory but there's many it's just wrong but let me just give you two reasons for it one the actual window for con loss where we've got where we can got any inside of because we know the cones are present here because all of these branches have it we don't know it's if the pla was wrong thing is wrong then none of this Branch like it's possible they got lost here rather than here which is what's always proposed okay so the window for cone L is very broad um but at the very latest even if PL was has blue the blue cone must have been lost when PL was split actually be here um so that means that the nocturnal bter neck does need align with a possible window of no loss so it's not very plausible there and um what was my getting confused now um it come back to me there's more um anyway I would posit that this was probably a key reason so let me just close in making this little argument here so you've got your Ze fish and I think Z fish is a fairly typical surface to a Lim fish um of course it's all Al 550 million years removed from the ancestor but at least it still has a lifestyle that's a little bit similar to what we think our ancestors will like um we've got four counts and Rod and that fans out into a massive array of first order inter neurons you've got loads of bipolar cells something like 23 types you got four types of horizontal cells so there's a lot there like a five b Banning out home network perspective and then you've got all of these second order neurons and you go into your third order neurons which is your gangling cells in yat and that actually that Divergence isn't all that big by the Numbers it's about two two two and a half okay so what then happens if you take your fish and you try to turn it into Mouse so you just take out the system what happens to the top at the top what happens is you get a proportional reduction yeah so we still got a time five panning out and the way it's achieved is by losing some horizontal cells and some bipolar cells yeah so the numbers the ratios are exactly the same it's just you've lost some cones and therefore you lost some cell some horizontal cell but then what happens down here then be massively increase not only do mice have as many G cells cells as fish they have more that's way more right so this Factor becomes seven rather than two and a half so I would posit that mammals lost this system very early in the evolutionary history coming out of the water not having use for it um but they're still needed to do the things that the system does which is like get distance right and compute interesting things that I haven't thought of which I'm sure these counts will do still so they needed to evolve compensatory strategies and one of the ways of doing that is to put different circuits in and what better circuit to use than an AM cre cell which is highly diverse it contacts all the right bits in the Resa to do similar kinds of computations um and it seems a very I mean I've used the fish versus mouse example here but it's it's it's a mammal example mammals have crazy numbers down here and if you look at something like a bird or reptile they still have the entire complement they actually have grown the complement there's new count types this is still sort of factor five sometimes factor four up here um and then they s intermediate down here so the birds and the reptiles they seem to sort of stad Two Worlds of the of of of the sort of early Divergence in a fish and the late Divergence in a mamal so there's something to be thought about from a network P Evolution perspective here so um this is actually my in case there's time which there isn't so I will skip it and I will thank everybody apologize we're going slightly over and yeah thanks thanks Tom for a great talk there any questions yeah Mato I have two questions one is uh when you were doing the calcium Imaging the gratings and the dots were moving stimul right yeah the drifting gratings as well so could you be that this applies to rting images but not still images that the fish sees um so there are very few so what you might what you might have noticed in the stimulus is that there's a static rating and a moving rating back to static the responses are almost all to the moving so I think the transition from like a black screen to a stripy screen that doesn't move is I mean yes the fish sees it but it was represents it with a handful of neurons whereas if you start moving it then it goes so um yes what I've said applies to moving stimul but I think and part of that is because non-moving stimul are well a rare um and maybe less is less difficult to represent them so you don't need many neurons and the other thing I was wondering so I always thought on you talk I the way I estimate distance mainly is because things got more blurry and uh I mean it's interesting to you suggest fishal this might use changing color as well um color appearance do you have any way to estimate which one the cas is faster like to say to say that the change in how white or how colorful something appears changes faster than how they blur yeah I mean I don't know if it's efficient to jump back to that slide but um actually I can probably just scroll it up I mean if you look at my silly little movie uh where is it no it's not this one it playing the GI yeah so I mean you can see the liness the bright right but I would argue that this effect is stronger so I think the whiteness effect is the over shorter distances I mean this is anecdotal I haven't Quantified this but this is my hunch any other Nicole so you mentioned a couple times that you think that the monitoring findings they don't have U cones anymore think sort of a weird Mion thing maybe not being right so do you think it's that they have the ancestral UV cone but then turned on the sws2 Ops and lost the sws1 yeah that's what I think yeah it's just so evidence for the Platypus blue cone rather than U cone is the presence of the gene for the Ops in the Gen that's it right there's no no one's looked even in the Rea if it's so yeah the evidence is weak okay any other yeah one one last question then let's go with that no you're you're good you're good a question about the mouse because the mouse seems odd in in that it's got this UV uh for receptors that are looking mostly at the sky I think and the and the greenish ones which are the gray for you are looking mostly at the Earth or down so what's what's your view with that well this is uh one of the things that are on my optional slides so thank you for that I will take my payment uh okay so here's my mouse bit I'll try to zip through it very quickly so this is basically the way that's described so first of all what's happening is that the real UV cones still exist and the the red cones have a green or UV sensitive option depending on location so there's actually two versions of the UV sensitive cone the original red one and the original UV one so that's the first critical distinction then um they are positioned in different places the top of the retina is standard bottom of the retina is the weird one because that has the real UV countes the fake UV countes and the occasional still green cone which is red just to confuse you so what happens when you the video no it's not okay so this is this is data from Thomas lab basically what they did is they build a camera that sees in Mouse wavelength screen and UV and just took videos of the same scene here's the green view here's the UV View and basically this is what it would do right so you got a green I just greened it out here right so this is what it would see and this is what would see of as glue together picture and hopefully what you can appreciate is that UV picture is better than the green picture it's sharper right so above the air UV gives beautiful C very nice um where but below the below is terrible because nothing UV doesn't Bounce U V light Silhouettes right so anything on the ground you don't want to use UV for it so it makes sense from ecological view to do that but okay so this is the world that the mouse fa faces itself right like okay I need to see this how do I do it do I take my UV system completely rewire how my red might work and point them at the sky in order to use this niceness or they just take my red system and put a UV hat so that the UV system sees in so that the red system sees in UV but codes like a red system that seems like a much simple strategy to do this yeah um and then actually quite recently they discovered that the true U cones the ancestral ones have a massive splodge in the bottom of the right now looking up that's different from that um distribution that you just saw this is the real UV cones they exist everywhere but there's a huge hot spot that hot spot looks at the sky and I think it looks for birds and here's a very nice video that again from from to paper so they've got the green andv and up here you see a drone and it's really hard to see it in the green Channel but in the U sample it's really easy to see right so if you're using u channel to look at the sky not only to get the nice cting from the trees and the grass and everything but also see your predatory Birds better right but it's nice to do it better or you can use your mega sensitivity trick of the UV system to see it even better I mean this obviously just take enhanced but the point is that if you take your real UV system and you you really crank up the game um and look and point it at the sky you will see your birds so I think that's in part what it's doing and here's some um data from how they respond so here um at the time we didn't we couldn't distinguish the real and fake UV cones in the mouse but basically what we found is we found some photo receptors that are very linear and respond to Green some photo receptors that are very linear and respond to UV but not green and then we found some photo receptors that were extremely nonlinear and biased to dark contrasts they were always UV right so we at the time couldn't tell if these are the real UV cones and the other ones were the fake UV cones but I suspect they are so we're testing that now um but that's basically what I think is going on bu I picked up a lot of rods because they also co-opted some like UBC based system have any thoughts on that or whether that's real you think need spend them it's the so an non laab did this uh a while ago this is a few years ago now where they basically found that some early Rod precursors started or they determined to be Rod precursors for trying to turn on UV opim but then they wouldn't continue into a cone fate they would switch them to A rod fate so this idea that they were taking cones that were destined to be a cone making them into rods instead as part of the nocturnal Bal neck solution so I I can't give you anything concrete about that but I think there is a it comes back to the question of where the Ros and Cones come from I think and I think if you do we've done this with if you do transcri toic across all the rods and all the counts and all the species you can get your hands on you do clever clustering and you sort of compare what is most like the next you'll find that the rods are by far the most molecularly distant for receptor like you've got all the cones and you've got the rods so that suggests that if they come from a single precursor Once Upon a Time the rods are the first thing to split off and then the cones follow spectral order the red ones are like the green ones the green ones are like the blue ones the blue ones are like the UV ones so probably either the UV ones or the red ones are the original one if it's the UV one it gives you rod and towards the other cones that might explain it um I do have reasons to suspect though that the red one is the original one so I think there's something to be cool great thank you very much Tom again start okay thank you very much there any questions