 So, I believe we are officially live, hello everybody and welcome to another session of our Sussex Vision Seminar Series, Always Within the Worldwide Neuroinitiative. I'm George Caffetzis, a former master's student in Thomas Euler's lab and currently a PhD student with Tom Badden and as your host for today, I would like to once again begin by thanking Tim Vogels and Panos Bozellos for putting forward this very initiative towards a greener and much more accessible seminar world. And having said that, please allow me of course to get back to the reason we all gathered here for today and introduce our guest from UCLA, Professor AP Sampath or Sam as people know him. Following undergraduate studies in Berkeley, he went on and obtained his PhD in Physiology from UCLA by studying light dependent changes in salamander cone photoreceptors. He then, but only briefly, left California and joined Stanford as a postdoc and University of Washington as a senior fellow before actually returning in 2005 as an assistant professor in USC at the School of Medicine. In 2013, he joined UCLA where he has been located ever since and nowadays holds the titles of professor of ophthalmology and neurobiology, associate director of the Stein Eye Institute and chief of the vision science division. Although he's talked today, as the title suggests, will be focused on single photon responses throughout his career. Sam has been successfully investigating the physiological function of photoreceptor cells over broad range of light intensities, how this function is altered or compromised in the generation and how it is constrained in energetic or evolutionary terms and the latter is achieved by studying organisms from transgenic mice to ancient agnothan vertebrates, the lamp rays. Therefore today, we have the pleasure of hearing about their latest and I'm sure exciting findings in his talk entitled mechanisms, underlying detection and temporal sensitivity of single photon responses in the mammalian retina. So without any further ado from my side, please all welcome Professor Sampath, Sam, the stage solios. Great, thank you, George. And thank you for that wonderful introduction. I want to give a special thanks to George and Tom for the invitation to the seminar series. As we've all experienced over the last year, there's not been a ton great that's happened in the world, but I think the one thing that I've appreciated is being able to return to the seminar series on a weekly basis and see just some great people. It's really my pleasure to be here today. Thank you for having me. Let me screen share at this point. Can we all see this? Yes, everything looks proper. I'd also like to thank, you know, in addition to George and Tom, also the Sussex vision group, I think, you know, hosting this has been a pretty big deal for the vision community. And I just want to acknowledge that right here. As I was thinking about what to talk about today, we've in the laboratory over the last few years, have been trying very hard to get out all of the work that was in some partial state of completion. And so I wanted to talk about something that hadn't been discussed previously or something that, you know, wasn't published at all. And as I was scanning through the various projects that the laboratory is working on, I came across these two. And what makes these two very interesting is that in some sense, they're very old. They were initiated many, many years ago by my very first PhD student. This is Haru Okawa right here. When I started my laboratory at USC in 2005, he joined pretty shortly thereafter. And for many years, I should say for many months at that point, it was he and I in a pile of boxes. And he assembled the first setups and he did a remarkable body of work that included, you know, work that set up our very fruitful collaboration with Kirill Martimionov's group, recordings that were used by Gordon Faden and Simon Lachlan to calculate ATP energy consumption in mammalian rods. And what I also believe to be one of Jeff Diamond's favorite papers. And despite this large set of work, he left me with two other pieces of work, one more nascent than the other. And we can see that he left the laboratory in 2009. And I won't do the math for you, but it's been a very long time since these pieces of work were initiated. And so what I'd like to do today is I'd like to share with you these two pieces of work to sort of as a tribute to Haru and his contribution to my laboratory. And so what I'll do for today's talk is the following. I'm going to give a brief introduction into photon detection. And then I'm going to talk about these two projects. One of them, as I mentioned, is more nascent than the other. The mechanisms underlying dark noise and rod photo transduction that influence the detection threshold is a project that he initiated mostly through the breeding of mice. And they were gradually handed off to another postdoctoral fellow in the laboratory at the time, Johann Paulberg. And since that time, it's become a collaborative effort between my group and Johann's group. Johann's now at the NIH, along with his postdoc, Elise Spokero. And on my end, my graduate student, Chris Griffith, has contributed to a lot of these recordings. Following the discussion of the detection sensitivity for single photon responses, I'll then move to the temporal filtering aspect of this. And this was a manuscript that Haru more or less left me reasonably complete when he left in 2009. This is the piece of work I'm actually a little bit more embarrassed about and quite frankly, a little bit guilty. But I hope that, you know, both of these manuscripts, which are in process right now, will be out for your consumption sometime later this year, or perhaps as early as the summer. So starting with the introduction to photon detection, we appreciate now, based on the seminal work of Hexlar and Perrin, you know, some 80 years ago, that the human visual system in the dark adapted state is remarkably sensitive. It's sensitive down as we understand now to the level of individual photon absorptions and the measurements that they made at the time allowed them using this very simple optical system to deliver flashes of light to the three observers, you know, Hexlar and Perrin, and for them to vary the strength of the flash that was delivered and then to report whether or not they had seen the flash after they had been dark adapting for some time. And they assumed a very simple model. And the simple model was given the Poisson nature of light absorption that they postulated that some threshold or greater number of photons were required for seeing. And if you were to take a cumulative Poisson distribution, what you would have is you varied the threshold from one to a larger number is that this transition becomes deeper and steeper. And so what they did with their data was they essentially fit the data to the steepness of the transition. And what that allowed them to get around is that allowed them to get around errors in the estimation of light at the level of the photoreceptor cells that were scattered by ocular structures. And what emerged from their studies is that a spot of light that subtends about 500 rods in the periphery of the retina that a threshold of somewhere between five and seven photon absorbed were required to provide the perception of light. And indeed what that means is that the single photon response in individual rods must be relevant for behavioral sensitivity because the likelihood that any one of those 500 rods were absorbing two photons is remarkably small. We know that based on this work and a tremendous amount of work after this by Barlow, by Sackett, by Hallett, by Teich, they were able to add to the nuance of this type of frequency of seeing experiment to add components like retinal noise to look at the role of the criteria of the observer. And what comes out of this is this notion that our visual system is sensitive, perhaps to as little as one photon absorbed per 10,000 rods. And so this is actually quite remarkable. Now the demonstration that individual rods are sensitive to single photons came, you know, some almost 40 years later. And it started with work by, you know, Baylor Lamb and Yau, which you can see here. What they did was they recorded from toad rod photoreceptors. This is a suction electrode and they're drawing in and measuring the outer segment current, which they're plotting here as a function of time. In this experiment, what they've done is they've repeated a flash over and over again that on average generates roughly one or slightly less than one photoisomerization per trial given the Poisson nature of light absorption. And what you can see is that in any given line here that you have some mixture of responses, you have some flashes that don't yield a response. Some that look like an elementary response, some that look like perhaps they're the sum of two responses. And indeed, these single photon responses are not unique to toad. In fact, what we understand now at this point, based on recordings from a tremendous number of species, is that the single photon response in a rod is one of the remarkable conserved features of rod photoreceptors all the way back to the lamprey, as we've shown previously in the laboratory. And in fact, if you were to take into account the temperature and volume of the rod, as has been done by Christian Donner and other laboratories, basically a rod is a rod. You can account for the differences in the speed and the sensitivity of the response to a great degree. So rod photoreceptors can reach the limit imposed by the physics of the stimulus, meaning that the division of light into individual quantum. And this is not unusual for sensory receptors, is that we understand at this point that many sensory receptors have evolved detection limits that coincide with the nature of the physics of the stimulus, including the ability of hair cells to sense motion on the order of the brownian motion dependent on temperature under certain conditions. Also, factory receptors can detect reliably the bindings of single odorant molecules. So this is not unusual. When we consider the entirety of the visual system, the whole range of light intensities over which we can see, if we start with the photon absorbed per 10,000 rods near the threshold, the photon absorbed per 1,000, photon absorbed per 100, per 10, photon absorbed per rod, these lowest four orders of magnitude of light intensity are dependent on the single photon response in individual rods. And that has important health implications as well. That means that the degradation of the ability to resolve the single photon response above the noise will necessarily result in the loss of the lowest four orders of magnitude of light intensity, making you effectively night blind. Now, I should note here is that the ability to reach the highest signal to noise ratio during dark adaptations also affected, for instances, when you age, then the rate of dark adaptations slower. So understanding how single photon responses are generated and how they're transmitted across the visual system is actually quite important in terms of understanding deficits and processing. Now, detection by itself has been considered sort of with the following scheme, right? So this is a picture of rod phototransduction right here, just a schematic. And it's initiated by rhodopsin and rod outer segments, which when activated by light catalyzes the exchange of GTP for GDP on the alpha subunit of transducent, which in turn stimulates cyclic GMP phosphodiesterase to break down cyclic GMP to GMP and in turn closing cyclic GMP gated channels in the plasma membrane. This process has been studied quite extensively and we know a lot about it, including where various forms of noise arise. And so what do I mean by dark noise? This is a recording that was actually from Fred Rieke that we were able to use in this review article that Greg Field and I wrote. And what it shows is it shows a suction electrode recording, I believe from primate rod over time in darkness. And what you can see here is that when you look at the current, there are a couple forms of noise that are present. The first are these, what I'm going to call discrete noise events. These discrete noise events are identical, if you will, to the single photon response. So here I've taken a discrete noise event and I've super imposed it on top of the average single photon response, which is the thin trace. And the presumption is that they're produced by the thermal activation of the visual pigment rhodopsin. Because the size and shape of the discrete noise event is identical to the single photon response. It must be creating an intermediate or meta 2 rhodopsin that has approximately the same lifetime to drive the same amount of activity through the signaling cascade. The identity of the single photon response with the discrete noise event mitigating this form of noise impossible. Because there's no filtering mechanism or non-linearity you can apply that would selectively remove these. And so what the visual system has appeared to do is it's appeared to stabilize this form of the visual pigment to reduce the probability that this occurs. And in fact, if you look over the 60 second window, you see on average perhaps one of these events on every trial. So the other form of noise here, as the name implies, is continuous. It's happening all the time. And it's the result of fluctuations in the concentration of cyclic GMP. It's called continuous noise here. So cyclic GMP, we can think of it like this. The cyclic GMP concentration at any time is dependent on its rate of synthesis from GTP by a guanyl cyclase, which is calcium dependent. And it's degradation by the cyclic GMP phosphodiesterase. And so both the synthesis and the degradation are constantly have an interplay between them. What happens during a light response is that the degradation is accelerated following followed by an acceleration in the synthesis. But what this creates is it creates fluctuations in the cyclic GMP concentration. And these fluctuations can be eliminated if you applied a light that's bright enough to cleave all the cyclic GMP and close all the cyclic GMP gated channels. And so this rod saturation shows you what the residual level of instrumenting noise is. So this has been how most people in the field have thought about sort of important sources of noise that may impact detection threshold. But detection threshold, I've discovered is actually a bit more nuanced than this. And this was based on sort of my time in Fred Rieke's lab as a postdoc and conversations that I've had largely with Greg Field is that when you consider the detection of single photon responses, there's an element of time that comes into this as well. So this this has been published, you know, in 2019 by Greg Field, Fred Rieke, and also by E.J. Chichlnisky, and he's a graduate student at the time of Valerie Yuzelle. I should note that what I'm about to show you was part of Greg Field's dissertation talk in 2004. And he was working on this in 2003 and 2004. That's a long time. You know, what's 12 years really when this is more like 15, right? So I don't feel that terrible about what I'm about to tell you later. But in any case, what Greg and Fred, along with E.J., were working on is this notion that the detection sensitivity has to be considered also in terms of the timing of the stimulus. So let's consider these two photon responses right here. When if you consider an ideal observer analysis, and you have two separate stimuli and you ask the observer to tell you which interval the flash occurred in, when the stimuli are fairly far apart, it's easy to tell which interval occurred. This flash, of course, occurred during stimulus A, and not likely during stimulus B. However, when the time shift between the two stimuli becomes reasonably close, it becomes hard to distinguish between these two. And this is another factor that impacts the detection of photon responses is this timing element. They formalized this a little bit more. It went by essentially creating a plot of a surface along which the flash strength sits along one axis, the time shift another, and the probability correct here. And the surface essentially could be broken up into two components, right? If we were to take a slice along this axis right here at a fixed time shift, you would then have a detection threshold that corresponded to, you know, some probability of events being correct, right? But that's dependent on a particular time shift. Similarly, if we were to take a slice along this axis right here at a particular flash strength, you would have the same, you have the reverse is that there would be a temporal threshold for the ability to detect the difference between two flash responses. And so it's a bit more nuanced, but it's a much more fair representation of the detection problem to consider this as a surface that takes into account both the detection sensitivity as well as the temporal sensitivity of photon responses. So what I'd like to do today, as I described for you earlier, is I'd like to share with you some work that we've done on both sides of this, starting with the mechanisms underlying the detection threshold and the forms of dark noise in the log photoreceptors. And then the second piece of work that I was describing for you that Haru left me with when he left the laboratory in 09 is the temporal filtering of the single photon response and how it affects signal transmission through the retina circuitry. So I'm going to start first with the mechanisms underlying dark noise. So what our experimental design for this part of the project would be is that we're going to be measuring the properties of the rod photocurrent. We do this in one of two ways, either with suction electrodes from dark adapted mouse tissue or with patch electrodes in voltage clamp from dark adapted mouse retina slices. What we're going to do is we're going to study the variations in the amount of noise dependent on the differential expression of proteins involved. And we're going to be using mouse lines that have variations in the rhodopsin, the rod transducin, or the cyclic GMP phosphodiesterase. The idea is that we're going to evaluate the magnitude of the discrete and continuous noise to try to establish their relationship between the level of protein expression, level of noise, and infer what the functional consequences are of these changes. So going back to the phototransduction cascade, I'm going to start with the discrete noise events. Now, interestingly enough, I mean, the identity of the single photon response with the discrete noise events would suggest that the species that's creating the discrete noise event is rhodopsin. And it's the thermal activation of rhodopsin either by heat that's causing the conversion to R star, which is metroidopsin two, which then drives the rest of the photoconstruction cascade. I may be wrong about this, and I would be happy if anybody would like to comment about this at the end of the talk, but I'm not aware of a really direct manipulation that shows this to be correct. Even though a wealth of circumstantial information indicates that this is so. What we did in order to test the role of rhodopsin and creating the thermal noise events is we got rhodopsin knockout mice from Janislem, and we bred these rhodopsin knockout mice with the GCAP knockout background. So the GCAP knockout, which I'm going to all talk about in the second half of the talk as well, is for the guanyl cyclase activating protein. And this is the key calcium sensor that controls the activity of the guanyl cyclase. And so when you eliminate GCAPs, there's no calcium dependent feedback to phototransduction. And it makes phototransduction really slow and photon response is really big. And this was an insight from Marie Burns from her original paper with Jeannie Chen on the GCAP knockout. But what you can do under these conditions is you can simply count the number of events. So what Johann did here, these are recordings by Johann Kahlberg, is that what Johann did here was he made suction electrode recordings from this rhodopsin plus plus or the wild type GCAP knockout and a rhodopsin heterozygid GCAP knockout. The rhodopsin heterozygid expresses half the rhodopsin of the wild type animal. And we know this a couple of different ways. We know this by micro spectrophotometry. This is a spectra from rods that were collected by Johann Kahlberg with Richard Fredrickson from my laboratory. You can see that the rhodopsin het has about half the optical density. And Elise Poquero in Johann's lab has also done western blotting from our segments. And he can show that in the het that it has about roughly half the expression of normal. So the experiment that Johann did here was he first delivered a really bright flash of light that closed all the cyclic jampi gated channels and then he waited for a minute afterwards to count the number of discrete noise events as I showed you previously in the work that Fred Rieke had done. And so you can see that the single photon response in each of these cases is easily detectable above the intrinsic noise and they can simply be counted. And what he showed over time is when he recorded for a very long period of time sitting in the dark all by himself is that he was able to record a number of events and when divided by the amount of time he could estimate the event rate. And what you see is in a rhodopsin het g-cap knockout that the event rate is almost exactly half. This gives more credence to the notion that it's the level of rhodopsin itself that's controlling these discrete noise events. And this is another interpretation has been in recent years that perhaps the discrete noise events are coming from the free APO ops in itself. But I think that to be unlikely, I think the fact that you can change the total level of rhodopsin in half the events and that the events have exactly the size and shape of the single photon response I think speaks to the fact that it's coming from the thermal activation of rhodopsin directly. So we evaluated the continuous noise by doing one of two things, by either manipulating the level of transducin or the level of cyclic GMP phosphodiesterase. Now, previous work from Fred Rieke and Dennis Bailer suggested that transducin doesn't participate in the continuous noise. And they did recordings from truncated toed outer segments where they put in recombinant transducin and didn't show a change in the power of the noise. We wanted to investigate this using transgenic method. And so we're going to look at reductions in the expression of transducin as well as reductions in the expression of the phosphodiesterase to see, you know, how those influence the magnitude of the continuous noise. So starting with transducin, when you look at the transducin heterozygid, these are recordings that were done by Johann. And so what he's done here is he's recorded in the dark and then at this time, at five seconds right here, he's delivered a step of light that essentially saturates the photocurrent. And from these, he can calculate the power of the noise in the dark adapted state and the power of the noise in the light saturated state is essentially subtracted to find the extent of the cellular noise. And he's done that for the wild type rod here, as well as the genat one or the rod transducin heterozygid. It should be noted that the rod transducin heterozygid doesn't have that much of a decline in the total transducin concentration. It's about 80% of normal. But what he sees under these conditions when he plots the power spectrum, the power as a function of the frequency is that these relations essentially sit on top of one another. And so a modest reduction in the transducin concentration doesn't appear to appreciably affect the continuous noise. We decided to go even further in this direction using what I'm going to be calling here a genat one under-expressor. These were mice that were generated by my colleague, Genie Chen. And they effectively on a genat one knockout express a trans gene that puts the transducin concentration about 10% of normal. And what we see under these conditions is that the power spectrum of the noise is also very similar. At these intermediate frequencies here between one and four hertz, it appears to be a little bit larger in the control mice. But it takes a pretty large reduction in the transducin concentration to get these. The zero frequency asymptote is also very, very similar. So what we would conclude from this is that the continuous noise isn't heavily affected by the transducin concentration until perhaps you get to a very low transducin concentration. We're exploring this a little bit further. The bigger changes in continuous noise come from when we start to reduce the PDE concentration. So these are recordings also from Chris Griffiths. I should say Chris in my lab did the recordings on the right and previous panel as well, where he looked at control rods. He looked at RD1. This is the retinal degeneration mice. He looked at RD1 heterozygous, which are known to express about half the concentration of PD6B, beta B, beta B. And we also looked at one other line of PDE under expression mice. These are the PDAD167A mice. They were produced by Mingya Liu from East China Normal University with an assist from Jason Chen at Baylor. These essentially don't express any PDE A, but they express residual PDB, which gives them about 10% of normal activity. So in recordings from these types of rods, what Chris found is that the amount of noise gradually increases. So you can see this here. These are dark noise. You're all scaled to the same axis here. And this is what happens when you apply a saturating light. So what happens, what you can see essentially from the response itself is that when we look at the size of the response as a function of the flash strength used to evoke these response families, that the reduction in the phosphodiesterase expression shifts this curve to lower light intensities. The responses become more sensitive. They become, you know, they also start to become marginally slower. You can see a modest slowing, which we're working on in the PDE, in the RD1 heterozygote. But the responses become very slow when you look at this. I'm going to call this the NCBS mouse. This mutation is in the non-catalytic binding site of the cyclic gene PFOSVADIS trace. But in these mice, it becomes tremendously slower. You can see the difference in the time scale here. But these are both the company. These slowing of the responses is accompanied by a sensitization of the response and more noise. The noise can be viewed here on the bottom in this power spectrum, applying the power as a bunch of frequency, is that at low frequencies, particularly, you start to have an elevated amount of noise that becomes quite substantial by the time you consider the NCBS knockout. So there appears, even within the phosphodiesterase expression level, this trade-off between the detection and timing that I was describing to you before. As you decrease the expression of the cyclic gene PFOSVADIS trace, you make the responses slower and slower. And at the same time, you increase the noise, right? So the temporal characteristics of the photon response are degrading. But again, this is done under conditions where the detection of the photons is greater. But it's also done under this consideration of greater noise. So what this is suggesting to us is the fact that the level of cyclic gene PFOSVADIS trace in rods is set probably by selective pressures that are trying to minimize the amount of continuous noise, while at the same time trying to make the photon responses fast as possible. One of the factors that we appreciate is important in setting the time scale of the dimplash response or single photon response is the cyclic gene P turnover rate. So by increasing the amount of PD expression, we can increase the rate in the dark at which cyclic gene P is turned over, making responses faster. So we believe that the set point of the phosphodiesterase activity is an important factor in controlling the detection of photon responses and in mitigating noise. This also holds up, I just wanted to include this data here. This was collected by Ulysse who looked at the same phenomenology in the RD10 retina. This is another PD6B mutation. And in the RD10 heterozygote, he did essentially the same experiment where he calculated the noise before and after saturating background light. And what he can see is that the power of the noise in this frequency range under a hertz is a little bit larger. Note the log scale on the left. And when he calculates the integral of the noise, you can see that in a population of the rods, it's a little bit higher. So we're in the process of investigating this further. But the concentration dependence of the noise and the characteristics of the response of PD, I think, are intertwined with one another. So in this part of the talk, what I've shown is that the thermal activation of redopsin, I believe, causes the discrete noise events in rods. And when you have the redopsin in a redopsin heterozygote, the thermal event rate or discrete noise event rate is exactly half. The continuous noise appears to be relatively insensitive to alterations in transducent levels until the transducent levels become very low. We're going to be looking now at what that means in terms of the loss of power in the frequency range of one to four hertz. We're unclear about the mechanism surrounding that and why that might be relevant. But that's currently an area of investigation. We also showed that reducing phosphodiesterase levels also increases the continuous noise. But it also slows the time course of the response and it increases sensitivity. And what we infer from this is that the phosphodiesterase level appears to be critical for, as I mentioned, for the minimization of noise and to ensure a temporarily quicker single photon response. So let me go forward now and just talk a bit about temporal filtering and then I can field questions at the end of the talk. As I mentioned earlier in my introduction, that the detection of photons has to be considered in addition to the temporal sensitivity and that these two parameters are essentially intertwined. They have to be represented together in some way. So we investigated how the single photon response's time course is controlled as it processes through the retinal circuitry. And so the experiments that Haru did was we studied mice with variations in the time course of the single photon response. And these include, you know, wild type mice. They include redox and heterozygots. And they include G-cap knockout mice. As I showed you before, the G-cap knockouts have a slowed response because of the lack of calcium feedback to cyclic GMP synthesis. The redox and heterozygots are an interesting case, and I'm going to show you something in just a second here. We received these mice as brown mice as they came from Janus Lem, you know, roughly in 2007 or 2008. And they display a faster single photon response. And I'm going to get back to this in a second. So we have mice that are, you know, have a faster single photon response and a slower single photon response. And the idea is that we're going to measure the properties of light dependent signal as they traverse the retinal circuitry from rods, from rod bipolar cells and from alpha ganglion cells to determine how the temporal characteristics of the response are processing as the signal flows to the retinal output. So this is what the responses look like. So these are suction electrode recordings now from these three genotypes, plotting the current as a function of time for a flash of light delivered at time zero. And so the flash is either just perceptible by the rod, and it increases in strength until it saturates the photocurrent. And what we can see here in these mice is, as I mentioned previously, the redox and heterozygot responses are faster than the wild type responses and the GCAP knockout responses are slower. And from these, if we derive the single photon responses, they're shown here in this top right panel, plotting the percent suppression of the dark current as a function of time. The wild type responses suppress about 5% of the dark current. This is fairly characteristic for rods across all species. And they decay over, of course, of a second in our suction electrode recordings. The redox and heterozygots have a faster temporal characteristic, and the GCAP knockouts have a much slower temporal characteristic and suppress perhaps as much as 20% of the dark current for photon because of the lack of calcium feedback. When we normalize these, you can see these temporal differences more readily. Now, this is where I wanted to put in the comment about the redox and heterozygot. When we receive the redox and heterozygot from Janislem 2008, 2007, 2008, the mouse was brown and it displayed a faster single photon response. And we had become concerned at that time about the differences in the line. Because the different mouse strains or lines all have different features. And if the ultimate goal in these experiments was to compare these different mice behaviorally, that it was important for us to cross them all into a common background. And we didn't know any better. And so we chose to breathe them into C57 black six line. And so we did this for five generations. And lo and behold, what happened over time is that the speeding phenotype went away. So these are now recordings from C57 black six mouse and from a redox and heterozygot, which had been bred for five generations on the black six background. And now if you were to take the single photon response and you were to normalize it, you can see that the time force is identical. But indeed it's a knockout because it has half the sensitivity. It has half the visual pigment, which means half the collecting area, which means you need twice as much light and low to evoke the same size response. So we worked with Clint McKino on this for a little while. As it turns out, the phenotype in the redox and heterozygous for the brown mouse was because of, as the Palschewski lab had shown, is a reduction in the outer segment volume. They showed this by EM. But what happened over time when we bred this into the black line, I sent retinas to Clint McKino, he did a similar analysis and found that the outer segment volume was identical. So there was some strain dependent effect that was causing the heterozygot in the brown line to have smaller outer segments, but in the black line not to have outer segments. Now I'm happy to talk about this at the end. I don't know that we can ever replicate these starting conditions, but it's fascinating nonetheless. The recordings I'm going to show you going forward are from the brown mice though, okay? So what Haru did was he recorded essentially across the retinal circuitry recording from rods, user section like for recordings, and then recording from rod bipolar cells or retinal ganglion cells and voltage clamp. And what he, these are, you can see these various responses here, response families here. And one of the things that you can appear that you can appreciate qualitatively at this level is that there appears to be a speeding of the responses as they process from rod to rod bipolar cell to ganglion cell. They appear to become briefer and briefer, you know, or they're high pass filtered. To look at this in more detail, what Haru did was he estimated the elementary response per photon at each of these levels, right? So these are for the rods. This is data that I've shown you already in the previous slide, but he's also done this for the rod bipolar cells and for the retinal ganglion cells plotted as a fraction of the maximum current. So what you can see here is if I start in, in the, with the rows is that the rod bipolar responses are a bit briefer in the Rodopsin heterozygote and a bit slower in the GCAP knockout. And this relative difference, you know, is still there, but becomes proportionally smaller if you look at the retinal ganglion cells. When we look in any individual genotype, what you see is that the rod response is very slow and that the retinal ganglion cell response, which is in the dark line here, is pretty well, the time to peak is, you know, fairly close to the rod bipolar response, which is the dotted line. And this appears to be true across all of the genotypes. So we wanted to look carefully at what the temporal transformations were. And so we used a linear filter approach to this. What Haru did was these are now the responses, we're going to break this down into two steps. One is looking at the conversion from the rod to the rod bipolar response. And the second is looking in the rod bipolar to the retinal ganglion cell response. What Haru did was he modeled each of these responses in the wild type as a cascade of Poisson processes that they produced effectively the dashed lines that you see on top of the actual responses. He did the fast Fourier transform, he divided them, he did the reverse transform, generated the linear filter, then asked to what extent does the linear filter, when applied to the rod responses of the redox and heterozygote or the GCAP knockout, to what degree do they recapitulate the characteristics of the bipolar response, right? So you can see this is done here, right? So when you make, when you create a response for the rods, you know, based on the model and then convolve it with the linear filter, you don't quite capture the time to peak of the bipolar response which is shown right there. And the same is true for the GCAP knockout, it doesn't quite capture the time to peak. A more complicated filter for the bipolar to ganglion cell response, you know, that does, this is a slightly better job. But, you know, this, it's obvious by this, at this point that a majority of the speeding hasn't happened and the majority of the speeding of the response can already be seen at the level of the rod bipolar cell, as I described previously. So when we look at the rod photo current and compare it against the rod bipolar cell current, this is really effectively two steps. The first step is the conversion of the rod photo current into a photo voltage. And the second is the rod photo voltage in turn changing the trans, the rate of transmitter release to create the new rod bipolar current. So Haru looked more carefully at the rod photo voltage in current clamp and these are what photo voltage families look like, right? Plotting the membrane potential as a function of time with the flash at time zero. And from these, he estimated the single photon flow voltage. And you can see that here. So this is for a wild type, for redox and heterozygine for the cheek cap knockout. What you can see at the bottom here is that the photo voltage already accounts for most of the speeding in the rod bipolar response. If we took the slow trace here as the suction electrode recording of the single photon response from the rod outer segment, the dashed line is the photo voltage shown here. And the dark line here as the rod bipolar response, we can see that the time to peak is already fairly well represented. Whether you look at wild type, whether you look at the redox and heterozygine, mostly the case for the cheek cap knockout. The cheek cap knockout is a little bit more complicated in some ways because the response is so slow and so much larger as a fraction of the total current. So what we did was we took the same approach to look at these two steps. We looked at the photo voltage to photo current to photo voltage conversion and then the photo voltage to the bipolar conversion using the same methodology. And what we find is when we take the linear filter that converts the photo current to the photo voltage for normal rods and applied it to the redox and heterozygine, you get a fairly good recapitulation of the time to peak. And it does a much better job even on the cheek cap knockout. And similarly, if you look at the photo voltage to the rod bipolar current, there's ways where the time to peak is not fully captured. And so it looks just on the face of this that the conversion of the photo current to the photo voltage that a single linear filter accounts pretty well for the time to peak, but a linear process doesn't fully account for the rod photo voltage to rod bipolar response. And we have some sense of that already from recordings from rod bipolar cells, which are markedly nonlinear work that Greg Field and Fred Ricci did as long as well as work that Fred and I did when I was post-doc in this lab, among others. So there appears to be that the major speeding of the photon response in the retina appears in the photo current to photo voltage conversion. And you can already see that the faster photo current in these models or the slower photo current in these models is also display a faster or slower photo voltage, you know, compared to wild type. And what Haru was curious about is he was curious about to the extent to which these changes in the photo voltage that are present in the rod layer are contributing to the retinal output. So what he did was he recorded from on alpha cells on alpha cells are reasonably easy to identify in a home out. They have really large cell bodies. And when you record from them and deliver a step of light, they deliver sustained action potentials. So what Haru did was he recorded from these cells in voltage clamp, either near the reversal potential for inhibitory input or for excitatory input, and he looked essentially at the two. You can look at the excitation here. You can see that they have very large excitatory currents that scale with the strength of the flash. And similarly that there's inhibition as well, that's a little bit smaller. And what he did was he picked a flash strength that yielded a small excitatory current, but negligible inhibition. What he would always do in these recordings is he would go down on cell and record the spikes while in on cell mode. And you can see that many, many trials delivering this weak flash strength, he collected a bunch of brassters, which he can use in spike histograms. And then he would break into the cell and then make the voltage clamp recordings afterwards. And so if you were to look at the excitatory input on a number of trials, the average to a very small excitatory input. And then the question is, to see in each of these mouse models, when you change the rod photo voltage time course, how does that translate into changes in the excitatory input as well as the spike generation? And so what Haru found was this, is that if you looked at the excitatory synaptic input, isolated here, the elementary input per photon, you can see the very same relationship is what we see at the level of the rod photo voltage. Is that the redox and heterozygote, it's a little bit faster in the GCAP knockout, it's a little bit slower. These are the spike histograms that he collected. And when you overlay them, you see the same relationship is that the spike histogram and the redox and heterozygote gets a little briefer and has a shorter latency than the GCAP knockout, which is more prolonged in time. And indeed, if you were to overlay the time course of the excitatory input to the retinal ganglion cells on top of the spike histograms, they match pretty well. And so what we take from this is the notion that this initial conversion from the photo current to the photo voltage in the rod itself is responsible for most of the temporal filtering for what you see at the level of the retinal output. Indeed, the differential control of the excitatory synaptic input to the retinal ganglion cell is essentially defining the envelope over which action potentials are generated. And the presumption is that along with this will come changes in the temporal sensitivity of these cells to stimuli in the single photon regime. So the conclusions to the second part of this talk is that the single photon responses are sped as they flow through the retinal circuitry. And the high pass filtering is critical essentially to establish the timing of arrival of single photons to individual rods, as has been suggested by Bialik and Owen, among others. A majority of the speeding of the single photon response appears due to the linear conversion of the rod photo current to the rod photo voltage. And there are non-linear components downstream that sharpen the waveform even further. But interestingly, the spike responses of the on-alpha retinal ganglion cells are confined to the temporal window that's set by the excitatory input, which is largely defined by the speeding that happens at the level of the rod photo voltage. So what I hope I've shared with you today are two sides of the single photon detection and temporal sensitivity issue, looking first at the mechanisms that underlie the rod noise itself, and then looking to see what the transformations in the rod signal are as it processes through the retinal circuitry. Let me end here by making the following acknowledgments. As I said, I'm really deeply indebted to Haru Okawa for essentially helping set the lab up on a really tremendous trajectory. I also appreciate all the other lab members. Please don't think that I don't. In particular, Chris Griffiths provided a lot of the recordings that you saw in the first half of the talk along in collaboration with Johan Paulberg and Lee Spokaro. Please, you know, expect to manuscript somewhat soon on the detection work. This work also is, this has been, we're always pleased to have, you know, Gordon Fain, you know, as part of the laboratory. It's just been a tremendous opportunity to leverage his wisdom and his knowledge in this field. I'd like to really give special thanks also to Greg Field. I think the discussions with him over the years, you know, have really helped me help frame how I think about, you know, single photon responses more generally. And also, I'd like to acknowledge Jurgen Reingo over here, who's actually been involved in a lot of modeling on our behalf. And the insights that come out of this are actually quite remarkable. This is some of the funding that supports the laboratory. And I'd like to thank you all for, in some cases, either viewing this very early in the morning or very late at night. And I can take some questions. Thank you very much, Sam, for this fascinating talk. There is a lot of hype in the chat. I mean, I'm pretty sure you will be checking the conversation after your talk. And before I start moderating the post-talk discussion, I would like to remind to our audience or introduce in case we have a new audience joining in for today's talk, that we will have like a 15 minute session that I will be the moderator. And then you are more than invited to join us in this very Zoom room that we are currently seated in, where you can discuss in a more informal tone. So I would like to start with a question from Jeff. It was already like five minutes in your talk, saying, how can you talk about single photon responses in LA? Like Finland, he does understand, but in LA, how are you talking about single photon responses? Jeff. It gets dark here in the middle of the night too. I'll just leave it at that. Oh, I should also mention before anything else, this talk is supported by Thor Labs. And I request them to send the snacks directly to me because they're being consumed in the lab before I get them. But yeah, so yes. Okay, do we have more questions, George? Yeah, sure. Absolutely. So the first one is from Petri. How does halfing the expression level of rhodopsin leading to halfing the dark noise allow to address the apo-opsin hypothesis? Isn't noise going to be one half? The answer to that is yes, Petri. I think the argument for me is less about that and it's more about the fact that to generate an event that has the size and shape of the single photon response, you really need to generate the profile of metatouridopsin. And I think that requires there to be chromophore in the pocket, in the binding pocket, and the cis-trans confirmation change. I think that the problem that you face if you want to make the apo-opsin argument is that you're arguing that a pigment molecule devoid of chromophore in its pocket is capable of transiently either taking up a chromophore or generating an event that looks like metatouridopsin without the chromophore. I think I find that to be more unlikely, but your point is well taken. I don't think it necessarily distinguishes that from the other. So before I continue with the questions, Sam I would like to ask you to stop screen sharing so we appear simultaneously and it's easier for the audience to follow. Great. So next one up is Phyllis Robinson. Are there any PD6 mutations that affect the kinetics of the enzyme? I think this was at the end of your first part. That's a good question Phyllis. I think this is a question really for Rick Cote. I might need to ask. I am not aware, but that's a very good question. I think yeah I'm not really sure Phyllis, but that is a good question. Next one is from Brian Jones and there has already been some discussion in the chat, but I will make sure you also get the question. Was the rhodopsin content halved in the normal size outer segments? I think this is for the strains. Yes and so yes and so indeed when the rhodopsin content in the normal size outer segment when we bred it into the black background, the content was halved and you can see that as a combination of both micro spectrophotometry as well as the shift in the response intensity relationship by twofold. So yes the answer is yes. The content looks like it's halved, but it doesn't change the time course of the response, which was the original argument made by Clint McKino's lab that the speeding of the photon response in the brown mouse was caused by a reduction in the crowding of the disk membrane. So that appears largely now to have been more an effective outer segment volume and not necessarily crowding the disk membrane. I see. Xiaoyu Li, Duzi, Rd1, Rd10 heterozygotes saw the same response kinetics across all ages. We've not looked systematically across all ages. The mice that we're using are typically around eight weeks of age, six to eight weeks of age, so somewhat young. I'm assuming the same is true for Ulisse on his side, but no, we haven't systematically looked at that. One thing notably is that the Rd1 heterozygotes doesn't degenerate as aggressively or at all compared to the regular Rd1 mice, where the degeneration is pretty aggressive. The same is true for Rd10. By about two months, the Rd10 mouse has very few rod photos left. So it does appear that at least half the expression keeps the rods alive and functioning well for a much longer period of time. At this moment, I want to apologize for not conveying it sooner, but there are people congratulating you for your talk in the chat. I'm pretty sure you are not following the conversation. Next one is from Henrik von Gerdstorf. What is the membrane time constant of the own alpha RGC? Does this dictate the time course of the decay? Hi, Henrik, how are you doing? Yes, so I should say the recordings, this site, I couldn't tell you for sure, but I can say that the one thing that Haru did in his recordings is that he recorded with a cesium internal in order to tighten the space constant. So the answer to your question is, I don't think we have a precise estimate of what the membrane time constant is, but the fact that the synaptic input, that the time course of the excitatory input and the spikes line up so well and with such correlation between the two, I think would suggest that the membrane time constant is not the limiting factor here. And I'm taking the questions chronologically as they appear. Next one is from Tom. What might a single photons glutamate release response from a rod look like on average? Would that not give you a similar speeding up effect simply because of how the ribbon works? Okay, so let me just give a plug here to Cassandra Hayes and to Wally Thorsen and Greg Field. I don't know if you saw this recent paper in E-Life, but they've looked at the role of rod release using glutamate uptake as a marker of rod release. I think that the issue is this, is that the transmitter is being released continuously by the ribbon. And so I don't think this is in darkness. And so I don't think that the depletion is an issue the way it is at the rod bipolar A2 synapse in that regard. I'll have to go back and look at that paper, but I'm not sure that you would expect a temporal speeding just because of the transmitter release because the signal is reversed. You're working under conditions where you're continuously releasing transmitter, and then it's stopping briefly. I think when it's the other way around, the depletion does create effects. I have already posted the Zoom room link of the room we are currently sitting in, so please feel free to start joining us here. Henrique again, what controls the GTP concentration within the rod? Is this rate limited by the levels of ATP to GTP conversion? Is this free GTP about four or five millimolar? And the answer to those questions is that we don't assume that ATP or GTP to be limiting in any way, and we know that there is a robust mechanism for the interconversion between ATP and GTP under conditions where one is depleted. And so I think, I think if I understand the question correctly, I think that that's the consideration about whether the ATP or GTP, and so the concentration of cyclic GMP is being held fixed through the feedback by the rate of synthesis that's calcium dependent from the cyclase, and the dark rate of consumption of cyclic GMP by the cyclic GMP phosphodiesterase. And so those are kept in the micro molar level, if I'm not in the high micro molar level. So I don't think that that's, I don't think they're influenced necessarily by the ATP and GTP concentrations. We assume those to be in excess. I have another couple, if I'm not mistaken in the chat, and people have already started joining us here. Next one is from Anguillera. To modify PDE kinetics, could you overexpress the gamma subunit of the PDE? Presumably it should shorten PDE activation plus decrease continuous noise. That's a good question. And I have a feeling that the gamma overexpression has been done by Steve Sang. I'll have to look at that one. I think that's one way to look at it. Another mouse that we've been thinking about in terms of changing the temporal characteristics, and perhaps even speeding up the PDE activity, is this R9-AP95 mouse, which overexpresses the entire GAP complex, the RGS9 GAP complex in photoreceptors. And that speeds the temporal characteristics of the response by accelerating the PDE decay. That's interesting. We'll look into the gamma overexpressor. And the last one that appears on the chat is from Wei Li. Is it possible to manipulate guanil cyclase activity in the PDE mutant mice to rebalance the CGMP dynamics and restore the dark noise? That's a really good question. The one thing I should note is that in the NCBS mouse in particular, it displays a higher dark current because you've perturbed the balance between the two. The dark current is a little bit higher, like maybe 20% higher. That's going to be one part of the contributor to the noise. And the answer to this is, yeah, I have to think about how you would do that. The cyclase activity is actually pretty well regulated by the calcium dependence through GCAP. I don't know how you would tweak that too much. And the effect of GCAP is actually dual. On one side, as the calcium level falls, it accelerates the production of cyclic GNP. But as it rises, I think there's also an inhibitory effect on the cyclic GNP production. I think you would have to fundamentally reset that relationship. And I believe with that, if I'm not mistaken, because multitasking is quite tricky sometimes. We conclude the formal segment of today's talk. I would like to thank you very much, Sam, for honoring us and presenting so much unpublished data. And I really hope sooner than later you can wrap it up. And we can consume it in the printed version as well. I would like to thank everyone out there for joining us for another Sussex Visions seminar series talk. I would like to remind you that there is another one tomorrow at two o'clock UK time, if I'm not mistaken, about a primate color vision from Bevil Gunway from NIH. And I see people are already here and more are joining. Please, I would like to remind to the audience that once I stop live transmission in five minutes from now, that the link will stop appearing for some time. So make sure you do join this Zoom room if you wish to continue and participate in this informal chit chat. So, obviously, I wave my moderator rights. Rich, I see you have raised your hand. Please go ahead and mute yourself and ask the question. Hey, Sam. Hey, Rich. Nice talk. I was glad I got up at 8 a.m. in time for it. Okay, good. I just had a question in these heterosegous knockout rhodopsin mice. So if the outer segment doesn't change, I mean, you mentioned this very briefly about the molecular crowding issue and whether that is affecting the kinetics and the response. But there's something else I kind of wondered about, which is that normally this wouldn't be an issue, but the density of rhodopsin is so high in discs and in outer segments, but I guess it really pertains only to the outer segment. I'm just wondering if the membrane capacitance, I mean, I guess the plasma membrane capacitance wouldn't be affected by this. But has anyone ever measured the specific capacitance of maybe in cones, this is relevant, where the proteins are on the cell surface of cells that are heterosegous for rhodopsin. There's so much less protein there in total. As I was asking the question, I realized it's not relevant for rods because of discs. You know, if you're curious about the capacitance measurements for mouse rods and cones, they're published in a paper from Norian Ingram, who was a graduate student in the laboratory. It's in the Journal of General Physiology in 2019. So we've made, for a number of lines, we've made capacitance measurements and they're there, can be found there. Okay, and are they, you know, is there any different, you know, I guess I just wonder yeah, we've not looked at the rhodopsin heterozygot in this way, but like I said, I think that there's a number of pieces of evidence that suggests that the outer segment structure is what you would expect. I think, like I said, Clint McKinoh had done, you know, this EM work to show that the rod outer segment diameter is unchanged in the rhodopsin heterozygot on the black background. That's one way. The second is that the response sensitivity is shifted in the expected direction by expected amount based on the loss of the visual pigment. And the MSP also says that, you know, and Ricard, I can see Ricard is here too. So if there's half as much rhodopsin, what is, and they're sort of in this semi, this quasi-crystalline array, what is taking the place of that missing rhodopsin? Is it just empty space? Is it more membrane, you know? I have no idea actually, Rich. Thanks. And I hope that was worth waking up at eight o'clock. Yeah, no, absolutely. Very nice talk. Great. Thank you. I, you know, just now that I have all of you here, I just, I want to say is that I am deeply, deeply embarrassed that some of this work has been lingering for so long, particularly the work on the temporal filtering, which has been sitting on, you know, my computer for the better part of 10 years. So, you know, I'm really happy about the opportunity to at least talk about it, because it's going to push me to at least get that paper out hopefully soon. We all have our skeletons in the closet. Except Jeff. Jeff doesn't have any skeletons in the closet. I can't see his closet, but I wish my unpublished data looked that good. You know, there's usually a reason that it's unpublished. That was great, Sam.