 Good morning everybody. Thank you for coming and thank you for having me here. It's really, I really appreciate the opportunity to present some of the findings that we have from one of the many projects that we have in the pre-sci lab. Now, since I have completely no medical training, my idea of what you already know about this subject is very limited. And most of what I think about is coming from less than academic sources. So for this reason, and because I'm not a native English speaker, I'd like to ask you to stop me and ask whenever something is not clear. I've prepared a 45 minute talk with 45 slides. This is already the second. So we really have enough time for any questions and to keep everything clear. So what I plan for today is talk a little bit about Stargard 3 and a whole lot more about mouse models that have been developed in recent years, partially by us, in order to better understand the disease. Now, Stargard 3, Stargard's disease is a macular degeneration and as such it presents with a central atrophy surrounded most of the time with a subrational ring of flex. And as such it is pretty similar to the much more prevalent age-related macular degeneration. However, it is different from AMD in many aspects. One very important aspect is the early onset. It starts usually in the late teens, early 20s, so by the time an average AMD patient starts to notice something is wrong, a Stargard's patient may already be legally blind. Another very important difference is that all the known forms of Stargard 3, of Stargard disease are related to a single gene defect. Now, the one that I'm going to talk about today, Stargard 3, is that the one that we focus our research on is an autosomal dominant disease and it is related to a defect in the ELOVL4 gene. Now, what's ELOVL4? The gene codes an enzyme which is also called ELOVL4 and that stands short for elongase of very long-chain fatty acids number 4. What this enzyme does is that it takes an already long-chain fatty acid, like EPA, and attaches two-carbon group to this already long chain, making it even longer, eventually increasing the length above 30 carbon atoms, making very long-chain fatty acids. Some of these are polyunsaturated as you can see in many double bonds and if the substrate is polyunsaturated, then the product will be called very long-chain polyunsaturated fatty acids. I'm just going to use the VLC PUFA for short for that. So that's the product of the enzyme. Now, where is the ELOVL4? All the species that have been looked at contain ELOVL4 in the retina and inside the retina it's here in the photoreceptor cell layer. This is a better picture. This is in the mouse. The green fluorescence shows where ELOVL4 can be found. It's in the photoreceptor cell layer. Now, outside the retina it can be also found in the brain, the skin, the lens and the testes, but at a much lower level of expression. Now, this expression in the skin will become, however, pretty important later. Question, what is wrong with the ELOVL4 in Stargardt's three patients? All the currently known mutations affect the last exome, the last coding sequence of the gene, the sixth exome, and all of them cause the same effect. The end of the protein is snipped. It's called truncation, and the truncated part contains a small but very important sequence which is called the ER retention signal. This ER retention signal makes it possible for proteins to stay inside the endoplasmic reticulum. Now, without this signal in theory, these mutated proteins, mutated versions of ELOVL4 cannot stay in the endoplasmic reticulum. Why is that bad? It's bad because lipid sympathies, including all the steps of lipid syndrome, happen inside the ER. If the protein cannot stay in there, that is taking part in this, it will lose its function simply because of that. Now, is that really true? Because this is just having lost a retention signal. It may not affect it really, but experiments have shown that that's actually probably the case. Now, in this experiment, they took GFP tagged ELOVL4 and expressed it in a cell line. It's shown green. And they expressed it together with an endoplasmic reticulum marker that was connected to a red fluorescent protein. And as you can see, the two markers, one for the ELOVL4 and one for the ER, shows perfect localization. Now, when they did the same thing with the mutated version, the Stargardt 3-causing version of the protein, it formed clumps, and those clumps were not found anymore in the ER. So it seems that what happens here, there is a mislocalization of the protein and there's also aggregation. Both of these are not good for cells. And so this is the basis that is accepted now as what is the first problem in Stargardt 3 patients. So normally, we have ELOVL4 sitting in the membrane of the ER, making VLC-puffas for the photoreceptor cells. Now in a patient, the mutated version of ELOVL4 clumps together with the healthy allele, the healthy version of the protein, and together they leave the ER, forming these clumps which are actually called agrosomes. Clump is a very sloppy word for that. Sorry about that. So from this comes two theories of what happens next and what actually kills the photoreceptor cells in the process. So one that I've already mentioned and that is the more prevalent in the literature is that VLC-puffa loss is behind the death of photoreceptor cells. Now this happens because there is a loss of ELOVL4 activity because of its mislocalization and degradation. And VLC-puffas, these very long-changing fatty acids, are considered very important for maintaining the cell structure, the cell membrane structure, whenever the membrane makes sharp turns and falls. And that's what happens, actually, in the outer segments of the photoreceptor cells where you have the discs and where the membrane makes sudden turns. Now if this is really what's happening, then of course a therapy for CYR3 would be supplementation of VLC-puffas. We would just have to put this in the food and give it to patients and that would be it. Now the alternative theory that has also appeared in the literature and is somewhat also accepted is that these pumps, what agrosomes, are putting the cells under stress, a constant stress because they're trying to clear them away and while they're trying to do so, and so-called unfolded protein response, or UPR, is initiated. Now UPR is a regulatory mechanism which if you activate it for too long, it will eventually lead to cell death through apoptosis. So if this is the real theory, if this theory tells us what's happening in the photoreceptor cells, then giving VLC-puffas to patients will do nothing because the clumps will still be there and the cells will still die. In this case, the only good way to treat these people would be to knock down this bad protein so that it doesn't take out the good and doesn't form these, yes. Couldn't it not be a little bold? Absolutely. I mean, our own Paul Bernstein has done a study with supplementation seeing a mild but significant impact on disease, so there's no reason why both of them are not in addition. You're totally right, and this is totally hushed up in the literature. So what I see in the literature is that you're expected to put your weight into either this or this theory, attacking the other, and of course that doesn't make too much sense because both can happen at the same time, but the relative importance of it is also not pretty important to know. So many, most research groups have tried to tackle this problem of what's the mechanism by using models, biological models. Why do we use biological models? Well, first of all, researchers can't touch patients, of course. But mostly because the methods that we use, how many times destructive, I cannot ask for a blood sample, but I cannot ask a patient, a stag-out patient, to please give me one of their retinas. I can do that with a mouse. I can take out the retina of the mouse. I cannot do that with a patient, and I can do experiments on that retina if I want to. I cannot do that with a patient. And of course, repeat and control is the basics of a good experiment, and we can do that as many times as we want if we work with mice or cells with patients that have not seen the case. So in order to better understand what's going on at the molecular and cellular level in stag-out patients, we try to find good biological models that recapitulate the processes, the processes that happen in patients. And using those models, they also give us a possibility, a test bet, for trying whether a possible future therapy might work or not. So while I use mouse models, there can be other mouse models, mice have a couple of advantages. One is that they are mammals, and so at the biochemical level, they are recognized pretty similar to humans. And because they are not humans and they are not even apes, the ethical issues are much simpler to tackle. So I mentioned I can take out retina, nobody will stop me in that if I have an Iococ protocol for that. Also, because the ethical issues do not stop us, we can manipulate the genome of the mouse, which is actually already sequenced. So we have the ethical allowance, or it is allowed ethically, we can do it technically, and it's not so expensive that we cannot do it economically, so it's feasible to use mice as models. And they are used for many diseases. Now, there is one important additional asset in using mice, and that's their short lifespan. And that means that if we have a Stargardt 3 model, for example, in our hand, and we don't have to wait 15 to 25 years for the disease to present itself as in humans, we'll only need to wait a couple of months until the mouse matures. Of course, on the other side of the same coin is because they are short-lived, any process that is a short cumulative process that takes years or dozens of years to happen will never happen in a mouse, unless we somehow artificially explode the process to be really unnaturally quick. And the other major problem with every mouse model is that they do not have a macula or any other structure that would resemble a macula. And that's a major issue in all the mouse models that are related to AMD and the other macula degeneration. Now, what researchers contend with and say usually at this time is that, well, yeah, no macula, but they have cones and they have rods, and the ratio is approximately the same as in humans, so the makeup of the retina is mostly the same if you forget about the macula. And so just remember this because this is a serious limitation to all mouse models and people tend to forget about that. And so what we try to achieve, basically, is that we say that macula degeneration is a corn rod degeneration. So what we want to achieve, basically, is a mouse that shows a progressive corn rod dystrophy that is early on set, and we do that by manipulating the ILVR4 gene. And that would be the mouse model. And I think that by the end of my talk you will find out that we do not have such a model, although I'm going to speak about them. So this is just a very simple schematic of what's going on at the level of ILVR4 gene because it might get complicated and I found out that people get mixed up in what is a knockout and what are the different genetic manipulations doing with ILVR4 gene. So in a human patient we have two copies of ILVR4 and one of them contains the mutations and the other is the Z-caustaryl free, right? So here's a header cycle. In a wild type mouse we have also two copies of mouse ILVR4 genes. Now fortunately for us there is a 93% homology with the human gene, which means that it is extremely conserved. It's an important protein and there have been almost no changes. Above 70-75% homology so they are homologous, they are the same. So this is very close. Now the first model that has been tried was, which is the simplest thing, is to just take out ILVR4. And when people took out ILVR4, that was about eight years ago, and they had, so then they produced a knockout where one of the copy was there, one of the copy was not there and the result was that nothing happened. The mouse was totally fine. When they took out both of these, that's a homozygote knockout, the mouse died within an hour after birth. Usually within five minutes. Now the reason why this happened was that ILVR4 is also expressed in the skin if you remember that little graph at a lower concentration, but still. And it turns out that in the skin, ILVR4 produces a certain type of lipid called ceramides, which are essential for the water barrier function of the mouse skin. Without that, the mouse skin stays permeable to water and basically as soon as it comes out of the mother, it suddenly dries out and there is nothing you can do about that. So this model was not very helpful because the heterozygote didn't tell us anything, nothing really happened. The homozygote died. So the next model for Saga 3 was knock-in model. In this model, they took the mouse and put the human mutation into the mouse gene. So they basically changed only a short sequence here so that it resembled now what's in the human mutant version. Now, when they made a homozygote, so when both of these carry this mutation, it died the same way as a knock-out mouse. So that told us that this little mutation here definitely takes out the function of the protein. Now, what's interesting is that nothing else happened until about 8 or 9 months of age, which is relatively old in a mouse. At that time, what researchers have found was a slight increase in the photopic ERG. And this led some people to say that there is a whole degeneration in this mouse model. Now, I think that's totally stupid. If you have a degeneration of the whole system, the photopic ERG goes away. It decreases. There could be a cone dysfunction which could generate increased ERG, but that's all. So although this model resembles best what's going on in the humans in terms of genetics, because there is just one good copy and one bad copy and nothing else, the phenotype is really, really weak and it's coming up only at a later age. The mouse model that is considered the best today and has been used by more than one group was developed here in Utah by Feng Zhang's group. And what they did is that they introduced the bad human gene, the limited human allele into the mouse in addition, giving this in addition to the mouse its own complement of ELVF4 genes. And they expressed this at different increasing levels. So this is the expression level of this transgene compared to the mouse on ELVF4. And by expressing this, they found a degeneration that was early onset. It started already at one month because if you look at the number of photoreceptive cell rolls in O and L, the upper nuclear layer, you see a decrease already at one month. And then it sharply drops later. So they found a progressive disease and they did some ERG measurements as well to back up the findings. So that's the currently accepted best model for a starboard screen at this point. Now they used, as a control, another mouse train where they put the normal human gene into the mouse. And the comparison is usually against this mouse train. What we decided to do first was to look at how these different mouse mice see. Because that's pretty important and nobody has looked at it actually. And now you can't just ask a mouse to please sit down and read the silent chart for you. But you can fortunately use a mouse reflex. It's called the optomotor head turning reflex to assess how well a mouse sees. Now you all know about the optokinetic that was the reflex. When the visual fear moves, we follow it with our eyes. The same thing happens in mice. But in addition, they also moved their whole head. And projecting, putting the mouse in a box covered with LCD displays and projecting an image of black and white and moving the whole stripe around it. The mouse will follow it as long as it can see the stripes. Then what we do is that we increase the spatial frequency of this stimulus until the mouse can no longer tell apart black and white, it just see grays. And that doesn't move. So if it doesn't move, it will not move its head. So this way we can find the visual threshold for a certain spatial frequency. And that spatial frequency tells us what the visual acuity of the mouse is. So just to show you how this reflex looks like, there's a mouse there looking at it and the reflex is moving up right now. Can you see it's moving its head this way? So I measure the visual acuity in mice in a lot of different age categories from early teens to old mice and senescent mice. And what I found was basically nothing for the mouse would move. They have perfectly fine vision for a mouse. Now this is really bad vision for a human that they are legally blind. I mean normal mice have like 20 to 1500 vision. That's about this level. That's normal for a mouse. And well that didn't change in the mouse game. So there was nothing outside there. Now looking at the transgenic mouse, I did find a nice progressive decrease in the visual acuity. It started around this time and became very apparent. That was the D3 variant. That was the D3 variant of the older mouse. That was the highest of the two. No, this was the two. TG2, so-called TG2. We used the TG2 variant in all our research and that's what I'm going to call later as a transgenic animal. Thank you. So when I looked further into this transgenic mouse I found something weird though. I checked the ERG and what I've seen is that the broad prevalence of the ERG was going down really fast already at 1.4 months. Definitely at 2 months. It was going down very fast. All the while the ERG did not decrease at all in the first 4 months. Now this is just one stimulus intensity of 1 mouse. I did a couple of mice and a couple of different stimulus intensities and I plotted here the B-Babe amplitudes. This is the spotopic. This is the spotopic B-Babe amplitudes. They increase as you increase the stimulus intensity. Now what I would like you to see on this is that compared to this control, to the control values, we had a decrease of the spotopic B-Babe amplitudes already at 1.4 months and then this follows the progressive decrease. Whereas in this spotopic B-Babe which shows us what's happening with the control system, nothing bad is happening in the first 4 months, seemingly. And then it suddenly drops at 7 months. So why is this weird? It's weird because server 3, the TG, the transgenic mouse model is supposed to be a control mode dystrophy model. So if the controls were dying first, this would have to be reversed. We would have to see this picture here and this picture here. Now this was weird and it was showing definitely a rod corn dystrophy, more reminiscent of retinitis pigmentosa but at an early age. So I did the optometry measurements, the visual acuity measurement in darkness to assess the rod function, the rod acuity. Now that showed the same thing that I found with the ERGs that the rod system was already sub-normal, already at 1 month to start with and the visual behavior was gone at 4 months. This visual behavior that is driven by the rod system. So the co-system started to go down sharply only after the rod system was totally knocked out. So this definitely looks like a rod corn dystrophy. Now we've done a couple of a whole string of measurements together with the Bersin lab where I, who are Leo, at that time measured BRC-cufalavos in the retinitis. So because I was talking about ERGs and optometry but we haven't talked about anything about what's happening with the lipids in these mice. Now I was measured and found that the total BRC-cufalavos content at 1 month, this is shown by the magenta line, at 1 month the BRC-cufalavos content was already down by 75%. And this decreased further here. This is still the transgenic mouse. Now this contains all the information that was up here before plus the lipid information so it's a little hard to just say, I'm guessing. But if you can look at only this, what's happening is around 3 to 4 months. That's when the mouse is considered early at all. At this point, this mouse model shows about 5% of the BRC-cufalavos remaining. These broad-given visual activity is not measurable at this time already whereas the cone-driven visual activity is slightly increased at that point. So what does that tell us? It tells us that at this level around 5% of the BRC-cufalavos in the retinal the cone function was mostly retained and rod function seemed to have been lost. That would say that rods seem to be, at least mouse rods, seem to be more sensitive to VS-cufalavos in cones. Now, is that really the case? To figure that out, we really had to take out ELVF4 and we cannot do that in the whole mouse but we can do that, and we did that by creating our own new mouse models where we took out ELVF4 specifically in rods or specifically in cones using the pre-ELOVS genetic system. I'll tell about you this if there will be time. Basically, what we do here is that we take out both of the copies in rods whereas we do not touch the copies in cones and anywhere else. So these mice are perfectly fine in terms of barrier function and all that. Then we characterize these mice with that rod initial knockout and compared to the normal mouse where you see ELVF4 is present in the polysector cell layer and in the polysector cell layer of the knockout mouse you can see still some columns of rods that still show ELVF4 presence. We estimated that the knockdown was about 65% efficient. And when I, who are at the university, let measured via Sikufa content of these mice she also found that the approximate the same 65% loss of via Sikufa's from these red mice. So this was not bad. So we tested these mice and checked whether the anatomy changed and we found no sign of degeneration. We measured, we counted the number of cell rows in the ELVF4 now and there was no difference between the controls and the knockouts. Overall, retina integrity looked fine. We asked the Mark Lab, specifically Brian Jones to perform an electron microscopic analysis of the outer segments because that's where we were thinking ELVF4 and via Sikufa's are especially required. But he found absolutely no change in the outer structure of the other segments. So we thought that maybe the problem that we do not see anything here is that we only knocked out ELVF4 to 65%. So what if the rest of the cells, the 35% of the cells are resupplying the others through the RPV with via Sikufa's. So the next step, what we did was made another better conditional knockout for us and this time knockout was looked to be perfect. All the rods seem to have been labeled with CRE that takes out the gene and I have determined that there was almost no via Sikufa left in these retinas. The total via Sikufa level of the retinas went down by 98%. So only 2% left which could have been in the cones. Probably wasn't in the cones. So we were really happy to see this and so I measured the visual acuity of these mice in darkness and what I found was no statistically significant difference between the new knockout and its controls. Same thing with the old conditional knockout it wasn't a big surprise. There was no difference there and also we didn't see an effect on acuity in the cone condition knockouts in darkness. That was 3-4 months but what happens at 6-7 months? Very good question. I did not know yet. Sorry, I actually know that. I did not know what's happening at 10 months. 6-7, still nothing. Yeah, same thing with the cone system. The photopic visual behavior was not different in any of the three knockout strains. Then I measured some ERGs as well in darkness. It has the broth system and the photopic ERGs and the BV8 amplitudes were not different when I compared the congenic control to the knockout the congenic control to the convolutional knockout and we respected what level of the visual stimulus was. So to summarize, we've seen that 98% loss of TSE2 plus did not cause any problems in young mice. At a similar level, at 95% decrease of TSE2 plus, we saw a huge reduction in the rod capacity or in the rod system in the transgenic mice. So that tells us that at least in the transgenic mice it is not the loss of TSE2 plus that caused the degeneration of the rods. Yeah, that's basically would be that. That would tell us that at least in the TG mouse it's probably the UPR the other theory is more prevalent. Now, we think it's probably not the UPR because the UPR, the unfolded protein response requires certain key elements to be expressed at the higher level and these elements here we've tested with artificial and we did not see a significant increase in many of them. So we have to see at least a big increase and or a child increase to believe that there is a UPR and unfolded protein response going on in these mice. But it probably isn't. We checked it ourselves whether it was really good that RGDCI by checking the GFAP which is a CLEOSIS signal and it was up a lot and that's simply because GFAP is expressing more CLEOSIS in response to the degeneration and that happens inside in the transgenic mice. So RVAC proof is required. Well, we can be sure that they're not required for many months at their normal level. 98% loss is still fine for many months and that means they're not immediately required for a percent function. There's not to say that in the long term it might not be needed. And that also means on the other side that maybe 2% was perfectly enough to sustain function. Who is to say it's not? If that's the case, then a supplementation study could be helpful because we might only need to increase the VRC proof of levels just a little bit to save cells if that was the case. I've shown you also that the current models are actually not corn rod dystrophies with an early onset in mice. So at this point we do not have a model in our hand. We are working on some, but none of this more. That's what we are doing. We are trying to create 100% retinal knockouts and we are trying to do that, accomplish that in two different ways. One is to knock out ELVF in both cones and rods. That should take it out whole. And the other way that we are trying is to create a retinal knockout where we change the rods into a cone-like cell. In NRL knockout mice, they do not develop rods. These have only cone-only retinas. And through cone-conditional knockout plus NRL knockout mixing, we will, in theory, get 100% knockout in an all-corn retina. That should be the perfect model. The problem with that is that it degenerates on its own. So we will have to see whether it degenerates faster or not, but no more without VRC profiles. And we are also starting to conduct feeding studies together with the Bernstein level where we are going to try to bump up the EPA level in the retinas. And that is a substrate for the remaining ELVF fourth function to work with. Thank you very much. If you over-express human gene, then you definitely get substantial functional loss in regards to the mouse model. But if you knock out 90% of very long-chain polyunsaturated fatty acids, you don't see it get much of anything. So could there be some secondary effect that's not just knocking out the very long-chain polyunsaturated fatty acids occurring with over-expression that we haven't figured out yet? What are the levels of the very long-chain polyunsaturated fatty acids in the Zane model? How low are those? The division starts dropping at six months. Oh, I can tell you that. I want to measure that. There you go. So that's the Zane model. This is the TG2. So VLC, so it's now right there. It's starting to go lower and lower already. And then it goes down and down and down more. But you're still showing, you're still, to me, like even here, even in the four months, you're more than 2%. Yeah, it's about something else is going on besides just knocking down your very long-chain polyunsaturated fatty acids. But they would be getting the same results in four months. But you'd be getting just knocking down the... That's exactly it. So there's another effect. There's a secondary thing happening beyond just knocking down VLC proofing. Right, yeah. Definitely. Well, if... That's the system you've got to figure out. The funny thing is that one has to be very cautious with mouse models, as with any other models. Well, that's the other alternative. The mouse is just playing. Not a species that's going to be able to show this. It is possible. We might end up dumping the mouse. We might end up priming the mouse. Yeah. This is $100,000 level that would be millions or tens of million. Yeah. Yeah, so whatever you express, it doesn't have to be Stargardt III mutant in rods. If you express it at a high enough level, the rods will degenerate. They will show this degeneration. So it could be an extra-redubson molecule, which is in the right place to be. But if you push the rods to express a protein in addition to its own protein, then there is a very good chance that the rods will die. Cones are much more resistant to that. And it is possible that, actually, that's what we are seeing here. Now, if that was the case, we would probably see an unfolded protein response also. We don't see that. Now, it is possible that we just didn't look enough for that unfolded protein response. I cannot say that with 100%, that there is no UPR, absolutely. We just think that there is no UPR at the moment, because we haven't seen any evidence to the contrary. But, yeah, that's probably the case in the transgenic mouse. That's why it's not really a good model. There is probably a good reason why the CAN group did not try to express VT1, the Y-type human yellow VR4, at the same level as the Stargardt III causing mutant. And also, there was something funny, even when they expressed the Y-type. So this is the transgenic mouse that expresses the Y-type, the human yellow VR4, at a much lower level than the transgenic mutant that expresses the mutant genes. And we saw a little decrease here. And it seemed to be progressive but slow. And basically, you could get, depending on the level of expression, you could get an increasingly worsening, well, faster regeneration here. So yes, you're totally right about that other effect. That other effect is the presence of an extra gene that's not supposed to be there, and an extra load on the cells that they cannot handle. Rods in general are pushed to the limits of their ability to synthesize new proteins and lipids because they renew their own structure, about 10% of their structure is renewed every day, so that they have to remake themselves in 10 days totally. And that means a lot of fordopsin has to be produced all the time. And it's actually not a very easy molecule to make. Yeah, so that's why they're probably much more sensitive than cones. They have a little different half-segment. So it sounds like it's still back to drawing more of the protein model? Yeah, well, we are trying now with the double knockouts and the NRL knockouts. There are some drawbacks with both of those again, but still. Yeah. Yes? That's a great thing. That's a great job. I'm sorry, that was the first few minutes. Thank you. The main thing, the thing that I want to emphasize to some people, some of the new questions are, why not just give all these very long-chain fatty acids to patients? The bottom line is they're not in our diet unless you eat retinas or grains or skin as your usual diet. So that doesn't happen. And they're very hard to synthesize, at least in large amount, even to feed to a thousand gene addresses that are trying to do that. So complex amount of genes are not male. Right, so you can give them EPA as a precursor. That's a better precursor than DHA. And we know that DHA is not as good. Fish oil. Yeah, fish oil. And we know we're preparing a manuscript now where we're looking at autopsies and looking at their blood levels. Their levels in their fat and what's going on in the retina. We've clearly seen that there's a correlation to the EPA and DHA intake of the person before they die, when we see the retina. We even had the one extreme of a patient who was just a huge outlier. He had like five to ten times as much. He got a very long-chain polyunsaturated fatty acids in his retina. We asked, what was different about him? And it turned out he had been on two feeds. He was a debilitated person, but they were giving him ten grams, I think, of fish oil per day. Which is just huge. Yeah, huge amount. I mean, when we give A-rents patients that were supplemented as one gram. So it threw everything off. We could see it translate all the way into the retina. It was very easy to figure that out. The question is, in these mouse models, you have a defective yellow billboard, can you bypass that? But in normal people like us, you know, have a normal yellow billboard, there is a big influence on the diet as to what's going on in the retina. We have some evidence that with aging and with amnesia, those levels drop a lot. So there is something going on, even if you're not seeing the changes. Yeah. But didn't you have a study that was with a group of starters that by supplementing them with complex omega-3 fatty acids, you did have a statistic or something like that? So, yeah, what you're getting into is I have, and I don't know if Peter showed this in the beginning or not, but I have a cohort of family of 18 people here in Utah and California that have yellow billboard defects in us. And one of the things we found initially with this family is that some people had bad disease, some of them were all over the place, so we could correlate that actually with their fishing tape. They ate fish, they were relatively protected, and an awful lot of people absolutely ate a fish and they were going blind. And so I have an open label study now still going on where they're at least encouraging them to take the fish oil, and we're calling them by year to year-end. We're just about, it's now been about four or five years and we can actually start analyzing what's been going on. They think they're seeing better, but there's some preliminary small effects that we're seeing. It's a relatively benign intervention, so I think they're causing a big fish oil. But some of them still won't do it, so I actually have a controlled study that some of you have been very non-required. Interesting. So is it not that you would recommend at this point in time, clinically? Is it a line-up? You are recommending that. How much do you recommend? About a gram today. Fish oil. Fish oil, again, there is some difference. This translates into the AREDs study. AREDs is also using fish oil, which I think they made the right choice by having UP and DHA. There have been other studies using these mice and other where they did only DHA, pure DHA, which is a different analogy. And with that, they have a lot of failures. The explanation may be that you're not seeing the... you're not really giving them a better pre-pursuit, which is EPA. Any preliminary on the AREDs to... Even if I knew I could not tell you... It's not easy to block it. So AREDs, too, is closed. The results will be announced in May or June. I've invited them right, so I'm going to cooperate. I absolutely cannot tell you. I don't know anything now. There's a lot of people that... Very good. Yes, there is one more question there. I just have a really quick question, and I have been playing to you, connecting it up to our comas. You know, it could activate a pathway like that. I'm more important than that. We've been put in our stomach pathway just to agricultural itself. I'm losing my copy of that protein. It didn't actually have a functional effect, because you weren't generating agrosomes. Did you generate those agrosomes? So, in knockouts, you cannot expect to see any agrosomes, because there is no yellow vl4 in them at all. Now, in transgenic mice, you should see agrosomes, but interestingly, nobody has ever looked at it. We haven't seen them. The only evidence for agrosomes and the whole basis of this field is that all of it is based on cell studies, on cell culture studies. And that's, again, another interesting thing there.