 Good morning, everyone. We're going to go ahead and get started. Our speaker this morning is going to be Dr. DeAngeles. She's been here for approximately six months, and she's coming from Boston. And she's going to talk with us about age-related macular degeneration today. Thanks for having me here at your Grand Rounds. And I'm going to cover a lot of information today. So if you have any questions, I might gloss over some things you want to know more about, you just stop me, ask me, or we can follow up with discussion at a later time. So I'm going to talk about some of the work our laboratory is doing in conjunction with Gregory Hageman's laboratory to get to disease causality by understanding the mechanisms and pathways underlying age-related macular degeneration. Now, I'm sure, as many of you know, you know more than I do clinically about this disease, but it is the leading cause of blindness in the United States and Westernized countries. But it's increasingly becoming a problem in low-resource countries as people live longer. Also, as the population ages, age-related macular degeneration will surpass glaucoma and diabetic retinopathy combined. So although we've made these great discoveries from a genetic perspective and an epidemiological perspective, people are still going blind from this disease. This is just a review of the different subtypes. Here, Fundus photo showing the various forms of age-related macular degeneration. And we use the category called the age-related eye disease scale. There are several categories that are available out there, but more and more people use what's known as the arid study. And this is just to tell you, it's very important to get robust phenotypic ascertainment with any study of a complex disease because you can really skew your results and come up with false positives. So this is just showing the early dry. This is a progressive disease with geographic atrophy, accounting for about 5% of the advanced form and 10% of all AMD going to neovascular AMD, which you know is the growth of abnormal blood vessels beneath the retina. As you know, most therapies right now, like Avastin and Lucentis, are directed against the neovascular form, but they require invasive delivery methods. And they don't reverse or prevent vision loss over the long term. So there's really a lot of room for improvement to stop this disease before it starts or intervene possibly here before you advance to the blinding forms of age-related macular degeneration. But this also illustrates the phenotypic heterogeneity that you see that can cause problems when you start to look, how am I going to analyze this disease? How am I going to find causes for this disease? Now one of the ways we've done it is a lot of groups have come. How do we even know there's a genetic component to this disease? Well, it's true with any complex disease like cardiovascular disease, diabetes, certain forms of cancer, you have what's called familial aggregation. And what that means here, if you look at this pedigree of the different generations in this family where blue is with the disease and orange is without the disease, you have several people getting the disease. But if you have a parent or a sibling, a full degree sibling, you're at three to six times risk for getting age-related macular degeneration than someone from the general population. Also, studies of twins show that concordance for age-related macular degeneration is greater for identical twins than fraternal twins. But these types of studies just tell you there's a genetic component. It doesn't tell you what genes are involved. Additionally, like diseases such as cardiovascular disease and diabetic retinopathy, you're going to have a number of epidemiological factors, social and environmental factors that contribute to disease pathogenesis. As you see here, almost all of these have been implicated at one time or another in age-related macular degeneration. But you notice many of these factors are risks for diabetes, for cardiovascular disease, and for age-related macular degeneration, cigarette smoking is the most consistent risk factor from study to study. But we can't think about looking at one factor at a time. We have to think about gene-gene interactions, how the genes interact with the environment to contribute to a complex disease, a disease with high prevalence and more than one factor at the root of its causality. And why do we even care about this? How can knowing what risk factors you have from a genetic and epidemiological perspective help you? Well, the idea is to develop therapies that you have tangible targets, that you can either target from a preventative aspect or once the disease starts progressing to try to stop it. Because as you know right now, people already have the disease by the time they appear in your clinic. And hopefully, with all these new genomic technologies, hopefully the idea is to design custom therapeutics. And we have, as you probably have heard from Greg Hageman, one of the biggest genetic risk factors is genes involved in the complement system. And so I'm just going to briefly review this, because the work I'm going to show is with chromosome 10, is that what's interesting about the complement system is that the genes involved in here, and Greg Hageman did a lot of this work before the genetics even came to light, the complement factor HG, he approached it from a brute force proteomic level, is that the complement dysfunction of a complement system is associated with many other diseases, like dense deposit disease, myocardial infarction. And that may give support to the fact that this may be a systemic disease we're looking at, rather than a localized disease. And currently, there are some therapies and clinical trials directed at this system for the geographic atrophic form of AMD. And subsequent to that work is that, you know, I'm sure you've heard of complement factor H, and some of these C3, C2, CFB, contributing more modestly to disease, is that this verified the protein work. And now there's drug targets here, and complement factor D, and complement factor I right now. But more than just CFH, this is a schematic of our entire genome, all 22 autosomes. And there's been several genes, almost the whole, each one of these is a chromosome, several genes throughout that have been at one time or another associated with disease. But for every report, you say, you see that, oh, this gene increases susceptibility. You'll see another report saying, no, it doesn't. We can't find it. One of those controversial genes is the apoE, which is involved in Alzheimer's disease. But these genes you see circled here are the most consistent from study to study. And there's still other parts of our genome where we've done studies like genome-wide association and linkage where you look at the entire genome at one time for areas harboring AMD susceptibility genes that we have no idea what the gene is. So we're still trying to tease this out and get to disease causality. We do not have disease causality. And to have the strongest loci been identified, yes, probably so on chromosome one in 10. However, that doesn't tell us about disease path of physiology. And like the complement system I just showed you, that has a defined pathway where we know the genes, we know how they interact. But for the chromosome 10, which is the arms two, H-tra one region, we don't know the function and we don't even know the pathway it functions in. And you'll see that additional genes like maybe is an example of a complement pathway may function modestly, but that doesn't always correlate with importance to disease. And as some of you may already know, I'd like to give the Alzheimer's disease the work there as an example. The pre-Cenolin genes, mutations in those genes are only responsible for a few thousand, for Alzheimer's disease in a few thousand people worldwide. But the discovery of that gene led to the identification of the beta amyloid protein, which functions in that pathway and that's what therapeutic targets are directed at. So you've got to think in terms of how genes function together, not one gene in isolation. And we know that other genes exist. So when we think about risk or effect, that doesn't always correlate with importance to pathogenesis because it's not taking into account gene-gene interaction or how genes or pathway interact with the environment. Now I talked to you about chromosome one. Chromosome 10 is the other big region you may have heard about. And it's the most strongly associated to date with the wet form of age-related macular degeneration. And this is just a schematic of some of the variations that have been found in this region and what they are. However, we don't know if this is one gene or two separate genes that are associated with the disease. Moreover, we don't have the reports that have been published saying this is a mitochondrial protein. People are backing off of that now. Actually, the author of that paper told me it's not a mitochondrial protein, it's something else. And they're hopefully gonna publish that work soon. But the thing that's interesting too, like chromosome one, variation in H-tribe one is responsible for other diseases like rheumatoid arthritis. So that gives some credibility to the hypothesis this may be, in fact, a systemic disease. The thing is, just based on genetic studies, you can't separate these two things out. And you need the functional studies to complement that. And also, you need to look at different populations that have different prevalences of AMD from different parts of the world. Because you wanna figure out if something's truly associated with disease, you shouldn't see it in a population that doesn't have age-related macular degeneration. And we do have populations between Dr. Hageman and I that we've collected throughout the world. One approach that we've taken to try to tease out that region, figure out a function and a pathway is to use this genomic convergent approach or systems biology approach. Some people may be familiar with that. But it's taking all your data, not just genetic data, but looking at RNA expression and protein data, but also very importantly, robust phenotypic data because different variants may be associated with different subtypes of the disease. And you're gonna see an example of that with some work we're doing right now. And the idea is to computationally and mathematically put this all together so you can design more appropriate in vivo and vitro models and develop appropriate therapeutics. Example of a genetic approach that we've used and I started this work as a postdoc when I was in Boston collecting these families is to look at extremely discordant sibling pairs. And what that is is a sip because clearly because the age is the greatest risk factor, you don't have parents available to study the diseases you traditionally would for Mendelian disorder and the children are usually too young to have manifested the disease. Robert Alston came up with the theory of using sibling pairs to study diseases of complex etiology back in the 70s and Neil Rich took it and applied it to looking at extreme discordant. So the idea is you have an index patient or a pro band who has a sibling who has complete absence of the disease who's older than the index patient because age is your greatest risk factor. So if they pass that age and they still have clean retina then their chances of getting it may be small. So we collected such sibling pairs and to date we have over 500 sibling pairs and Julie Sylvester from the UK has also helped send Sib pairs over to supplement our cohort. And the way we initially approached this is we looked at the activity of genes like how genes are expressed between these sibling pairs and simultaneously we looked at gene variants between the Sib pairs. And we think this is a good way to do discovery because you have to do discovery, validation and replication to reduce your number of false positives. And the idea is that if this is your affected, your pro band, these two gentlemen are gonna share 50% of the time they're gonna have their, they're gonna share their alleles in common. 25% of the time they'll have no alleles in common for a given locus and 25% of the time they'll have both alleles in common. 50% of the time they'll only have one alleles in common. And the idea is to reduce the noise with the sibling pairs. This is an example of one of our sibling pairs and this is actually a set of fraternal twins. And both these, as you can see the index patient here has very end stage neovascular AMD with the disciform scarring in both eyes. And his brother has very normal healthy retina and normal macula and not even drusen that we can visualize with a fundus photo. So it's important to have good phenotypic documentation not just on your index patients but on your unaffected as well. So what we first did is we said, hey let's try to just look at gene activity as a way to look at the entire genome from an unbiased perspective. And to identify potential biomarkers with genes that might be made, there's too much being made or not enough being made. This is an example very close up of what a chip would look like. And each one of these little dots represents a gene in our genome. And the fluorescence that you see would be okay, that gene lit up. But the thing is each one of these has to be used per patient. You have to do a lot of statistical work and computation in mathematics to get some meaning from this data. And like I said, the idea is that maybe in a diseased patient, not enough of the gene is being made. And the examples I'm gonna show you and we've published on some of this is what happened. And for this work, we only used 18 subjects. As you can imagine at this time years ago this was very expensive. And also we wanted to make sure our Sib pairs were matched for things that can alter gene activity like smoking, hypertension, cholesterol. And so we only used nine Sib pairs. And just to give you an idea of the chip, it's very small. You saw the blown up version. Now to get some meaning from that data and all those circles you saw in that chip, we use a lot of computational modeling and pathway analysis. And we can start to infer function of these genes. So out of a number of probes that we cleaned up, we looked at, there were 18 probes or genes here that were at least two-fold either upregulated between an affected patient or a patient with neovascular AMD compared to their unaffected sibling. And in line with the genomic convergence approach in systems biology, we also had other evidence from our laboratory saying that this gene might be in AMD susceptibility locus in our genome either through genome-wide association or linkage. So that was nice. And the two genes I'm gonna focus on right here are robo and rora. Another thing we can do is to try to statistically look at these groups of genes and their functions and try to rank them. And here you see along the y-axis the functions of these genes and here the inverse log of the p-value to tell you the statistical significance. The take-home message from this slide is many of these genes, as you can see here, have more than one function and many of these genes share the same function. And not surprisingly, many of these functions have been implicated in an AMD pathophysiology. And that gets us to the shifting paradigm that you need to target more than one gene at a time for a therapeutic to be effective and reduce toxic side effects. From this data I've shown you, we can start thinking about how these genes interact. This is an example of several pathways meshed together with our data here circled in boxes. And red means the gene is up-regulated in an affected patient compared to an unaffected patient. But the idea is that the connectivity of these genes, and this is all mathematical modeling, this is hypothetical, this could be one pathway or more than one pathway. The idea is that the way these genes interact with each other is gonna be different between someone with disease versus someone without disease. And I like to give the example in the breast cancer field of a similar study that was done on a mouse model of breast cancer that showed that you needed to intervene on four genes to stop metastases from the breast to the lung. They used a combination of three genes or whatever. They could not block that metastases. And the idea is that genes that function on the periphery might be represent good therapeutic targets as opposed to hub genes. And more importantly, what was interesting is the HTRA1 gene came up. So we're like maybe we have a pathway for HTRA1 because what this program does is it adds additional genes that may be interacting with your genes. But it's not enough just to look at gene expression. You've got to verify this at another level. And the level we chose to do was at the DNA level. So how do you look at things at the DNA level? I'm sure this is a review slide for many of you, but we use single nucleotide polymorphisms because they're the most abundant variation in our genome. And you have to appreciate that we're only looking at this 0.1% that differs between each one of us. And that's what's responsible for health and disease in our systems. So we harness those. And to that end, we were able to use our extremely discordant SID pairs in the discovery cohort to show that this gene, the retinoic acid receptor, it's a mouthful, was indeed associated with age-related macular degeneration. But more importantly, it's not enough just to show in one cohort. Your discovery cohort helps you to direct you to where to look in that gene because many of these genes are huge. And in fact, we had several validation cohorts recruited in many different ways, both from Boston, cohort we have from central Greece, and a prospective cohort. So prospective, as you know, is followed from baseline with no disease over a number of years. And using all these cohorts for validation, we found indeed that this gene was, in fact, associated with neo-vascular age-related macular degeneration in three very different populations. Moreover, we found that this gene statistically interacted with H-tron arms 2. The gene I told you about earlier, that's the most significantly associated with neo-vascular AMD. So this makes us start thinking about a pathway to look at these genes more fully and completely in all of our cohorts. And to that end, we published the RORA. We started to look at the other gene, Robo1. And this is data we're about to submit, probably, I hope, in the next 10 days, showing that this gene 2 not only is significant in all three of these cohorts, similar to RORA, but more importantly, interacts with RORA. And that starts to tell us, hey, these genes are likely functioning in the same pathway. And the example I'm just showing you here is not to confuse you with these statistics, the odds ratio, if it's above one, it increases one's risk. If it's below one, it decreases one's risk. But depending where the variation is in the gene, you get increased risk or decreased risk for wet or dry age-related macular degeneration. So that tells you there may be different pathophysiological processes involved of who is converting to the more advanced forms. So this is data that we hope gets accepted soon. And the thing is, moreover, it's not just those three genes. We've actually shown statistical interactions between several of those genes in that pathway. And this arms 2, H-tri1. Again, red means it's up-regulated and affected patients, and green means it's down-regulated and affected patients compared to the unaffected. So this is really good. But it's one thing to show something statistically and mathematically that they interact. But what does that mean biologically? And to that end, not only do you have to look at several different populations on a large scale, because if you identify, if these are potential therapeutic targets, you want to make sure they have some global applicability. And it's very important to look at several populations. And you can look at it at a large scale with these chip-based methods. And to that end, we have many replication cohorts between Greg and myself with several people and some of our own people here, Amy Hartnett and Paul Bernstein, helping us with these populations. But also, population and ethnic specific populations. For example, the population from T. Moore has no retinal pathophysiology. They don't get age-related macular degeneration or diabetic retinopathy. We want to know why. We also have a study, a prospective study, on healthy aging in New Zealand in Christchurch. And we're trying to follow those patients along with aging of both the brain and the cardiovascular system and getting defined quantitative endpoints. So is this a potential drug-able pathway for age-related macular degeneration? Well, the idea is we hope so. Something else I didn't show you because it was getting a little off track was that we've also done work to show that these genes actually bind in the mouse macula. However, the mouse isn't, as you know, the best model for age-related macular degeneration, but it's good to know these genes are actually binding. So maybe we have a drug-able pathway here with targets of these various genes, and they are involved, many of them involved, in the neural retina, in development. For example, Robo-1 persists all the way through development, whereas other isoforms of Robo turn off in before adulthood. And what are we doing to get to disease causality? Well, one thing we're doing is we're doing whole exome sequencing in our families. And that is to look at rare variants and their contribution to disease. So that's different than just the SNPs that are all over the genome. It may be that many rare variants in a gene have a large effect. And Chad Huff, who's here today, is helping us to analyze this data. Analysis of this data is not easy. There's still developing methodology for doing it. And many people who are here in my laboratory are looking to vet this in our families and make sure the variation segregates with disease. And I will tell you, we have identified some rare variants in that Robo-1 because none of those variations I showed you are causing disease. They're just associated with disease susceptibility. There's a difference. And whole genome sequencing, which we're doing on Greg Hageman's population and to get to causality in the complement factor H region. We're also doing chip assays and Greg's donor eye bank. And we also have immortalized cell lines from patients to look at how are these genes binding with each other or is a transcription factor. Again, we think these variations are changing the binding and the way these genes are interacting. And the idea is that I hope I've covered is that it's not gonna be one gene in isolation. It's gonna be variation in two, four, six genes that's gonna cause abnormal ways the genes interact with each other, but also exposure to the environment. Say somebody who smokes 140 pack years and you're gonna get disease. And this kind of makes sense with a disease that doesn't manifest really until after age 50. And the idea is to get things that really can also predict disease. Even though we have 95% confidence that a risk factor is associated with disease, we still can't separate our cases from our controls. And we hope that using the methodology we are, we can increase sensitivity and also increase the value of anything we find is a prognostic and diagnostic factor. None of this work would be possible without a team approach in collaborators. And we have some wonderful collaborators in labs we're working with here at the University of Utah, people from my lab, Margo Morrison, Denise, Sylvia, Caitlin, Nathan, and Katrina, and Mike. The Hageman Laboratory, of course. Emmy Hartnitz Laboratory is helping us with some of the cell lines and the chip assays. The Geordie Laboratory, Chad Huff, with elucidation of the rare alleles in this pathway. And of course, collaborators from all over, including where I came from, and all over because they're contributing populations, the whole genome, whole exome sequencing, and high throughput statistics as well. And Nina, who's helping us with the mouse models for this. And I'd like to acknowledge our funding agencies, including the support I get from our department. Thank you, Randy. And our NIH grants and private foundation grants, because we couldn't do any of this work without that. Thanks. Yes, Randy. Pave is finally going in this week. We found some data to suggest both at the genetic, environmental, and the protein level to suggest that vitamin D may be protective for advanced age-related macular degeneration. And the way we went about that is we found that some sunlight exposure, we tried to quantitate sunlight and our sibling pairs was protected. So then we said, hey, let's look at the entire vitamin D pathway, but we looked at genes that we had, again, linkage or GWAS data from that we knew were in regions that were in the age-related, that were in regions harboring AMD susceptibility genes. But we found that CYP24A1, I forgot where that, it binds with VDR, vitamin D receptor, was protective and increased risk for age-related macular degeneration. And Sylvia Smith here was able to work, her post-doctoral fellow was able to work on some of the stuff we brought from Boston. Yeah, and it's quiet right now because everybody's kind of dumbfounded. At least that was the feeling from Beckman. I mean, these people have published this work, I'm like, what do you think? They're like, oh, it's not mitochondrial. And some people think it's cytosolic and they're like, we don't know. But they have these papers out there with whatever technology they used and that's what they thought it was. We're not sure which one, so that's why we write it arms two, H-tri-one. Because we're not sure if it's two genes, one gene. We don't know. It looks more strongly associated from a statistical perspective, but we're gonna have to see, like with some of these functional studies, like Greg's lab's doing, what this chip seeks gonna show us. Because arms two sometimes, depending on what databases you look at, isn't even acknowledged as a real gene. It's thought of as a pseudo gene. But now, it's just a battle and it's only based on statistics and you can't separate it out based on that. And they're in high linkage disequilibrium, which means variants in them are inherited together. Yeah. Hey. Attention, is there any more questions and you can email me or stop by my office if you think of things and wanna discuss them further? Yes? Not for Robo and Rora. No. Not yet. Yes, strongly. Yeah, and I think some of the work we're doing, like in New Zealand on healthy aging, where we can measure cardiovascular dysfunction, brain dysfunction, and we have MRIs, carotid artery scans. We're gonna look at these variations in those populations. It'll be interesting to see that. Thank you.