Scientist Stories: Kari North, Genetic Underpinnings of Obesity

Good afternoon, everyone. Thank you for joining us. I'm Ludmila Pawlikowska. I am a program director in the Division of Digestive Diseases and Nutrition at the National Institute of Diabetes and Digestive and Kidney Diseases. I'd like to welcome you to the 26th session of the NIH Genomics and Health Disparities lecture series. The first lecture was held in May 2015. The series aims to highlight the opportunities of genomics research to address health disparities. In addition to the National Institute of Diabetes and Digestive and Kidney Diseases, the seminar series is co-sponsored by the National Institute on Minority Health and Health Disparities, by the National Heart, Lung, and Blood Institute, the National Human Genome Research Institute, and the Office of Minority Health and Health Equity at the Food and Drug Administration. The speakers have been chosen by the series co-sponsors to present their research on the opportunities for genomics to improve health for all populations and to discuss the challenges of making sure these improvements are accessible and applicable to all populations. The speakers in this series approach this problem from different research perspectives, including basic science, population genomics, translational, clinical, and social science research. We are very pleased to have Dr. Carrie North joining us as our speaker for this afternoon. My colleague, Dr. Griffin Rodgers, who is the director of the National Institute of Diabetes and Digestive and Kidney Diseases, will introduce our speaker. Dr. Rodgers.

Thanks, and it's certainly a pleasure to introduce Dr. Carrie North. Dr. North is a genetic epidemiologist and a director of the Cardiovascular Disease Genetic Epidemiology Computational Laboratory in the Gillings School of Global Public Health at the University of North Carolina. Dr. North is recognized leader in research in obesity and health disparities in underrepresented populations, including those of American Indian, Hispanic, and African American ancestry. She has more than 450 research publications and has been continually funded by the National Institutes of Health since 2003. In 2018, Dr. North received the Shariqi Koumanicki Diversity and Disparity Leadership Award from the Obesity Society for her significant impact in the area of obesity-related disparities research and her leadership and mentoring of new investigators. Dr. North's research team focuses on identifying the genetic underpinning of a wide range of disease outcomes, including cardiovascular, obesity, and diabetes trait with the goal to better understand causes of chronic common diseases and to enable disease prevention. Just a housekeeping note, after Dr. North's presentation, my colleague, Dr. Ludmila Pawlowska, who you just heard from, is the NIDDK program director will facilitate the discussion. Please submit your questions at any time via the question and answer box. Now, please join me in welcoming Dr. North.

Thank you so much for that wonderful introduction, and I'm very pleased to be here today. I, um, As you'll see from my talk today, I'm quite hopeful that innovations in genomics research will assist us in more holistic interrogation of the causes and the consequences of obesity, both in and of themselves, but also critically, an improved understanding of the social and environmental effects on obesity. So, I'll begin by saying, you know, quickly centering us all and saying, what are health disparities? We know that they are, you know, notable differences in disease risk by self-reported measures of race and ethnicity. And so, we need to ask what causes these disparities. And I want to start by making a disclosure that I'm not a card-carrying health equity or health disparities researcher. Rather, my expertise is in genetic epidemiology, and in particular, in the genetic causes and consequences of obesity. So what factors influence health disparities? Back to that point. Number 1, humans live in diverse environments. As you can see on the slide here, we live in the Arctic, we live in the desert. In those different environments, we have different exposures to sunlight, to vitamins, to different bacteria and viruses that are endemic in those regions. And this can have a large influence on health disparities. We also live in diverse social environments. Social environments are extremely influential on disparities of disease risk. They can, they can influence how much food you eat, how little food you have available to you that is fresh fruits and vegetables, the stress that you have, how you handle said stress, your access to healthcare, your trust in healthcare, how much walkable your neighborhoods are, and how much physical activity you get. In fact, I would argue that especially for diseases like obesity, this is probably the greatest source of variability, and our rapid increase in obesity over the last several years speaks to this as well. But I'm here to you today to talk to you about how genetics and how humans vary genetically and how this might be important for obesity pathogenesis. I'll begin by asking, how much do we really know about human genetic variation and its influence on health and disease? Well, in the last 15 years, there's been a radical change in our understanding of the human genome. The Human Genome Project began in 1990, and I'm showing you over here with these slides. The first map of the genome was completed in 2003. This project was extremely costly and controversial at the time, given its extreme cost, But I would argue that it's been absolutely transformative in terms of what we know and what we understand about human disease. It's not only transformative in what we know, it's also transformed the way that we can study here the genes, the chromosomes, and their functions. And it's facilitated our ability to query the genome and to ask questions about health and disease that we never before could. Humans are a highly variable species, and that variability has been shaped by hundreds of thousands of years of evolution, with the ultimate source of all variation in mutation events, and then the spread of populations that came out of Africa. Although here, it's shown here in this map, and although we may argue about how many migrations came out of Africa, whether it was multiple migrations, how large those migrations were, I think we all agree that we did come out of Africa. And these migrations coupled with the forces of genetic drift, which are shown over here, this is genetic drift, the influence of allele frequencies in a very small population where you have genetic drift across generations. And down here is showing you in a large population across a number of generations how little the variation is. Of course, most of these migrations were small bands of individuals that were not representative of the whole population from which it came. This could have had a huge effect on variability in the gene allele frequencies that we see today, coupled with other forces of evolution such as natural selection. Today when I'm talking about racial disparities in health and the importance of genetics on the influence of health and disease, I want to be sure to establish some ground rules of what I mean. First, all humans have the same genes. They're human genes. Some variants in genes have higher frequencies in some populations due to their ancestral background and due to the forces that shape them, just like I showed you. Because there are more individuals in a given population that have ancestors that migrated out of Africa at some recent time in history, you might see variability in gene frequencies across populations. The data that I'll be describing today and speaking from has collected information on race and ethnicity as self-report and as a social factor. When I'm talking about ancestral or allele frequency differences in populations, I'm talking about ancestry in terms of continental origin of the variants. I'll be describing— what I'm describing is the variants that I'll be describing. We know about them and we make hypotheses about their origins based on comparison of participants' genotypes to continental and/or global reference populations. But I want to note, I make note at the bottom here that we have an incomplete availability and the small sample sizes of these reference populations. So it's an important concern, and discrete labeling of ancestral population completely oversimplifies genetic variation. Lastly, the differences across populations that I'll be describing today reflect differences by lifestyle, environment, admixture, ancestry, race, and ethnicity. So by all means, I'm not— I do not mean in any way to equate genetic ancestry to race/ethnicity. Given these acknowledgments and the transformation in genomics that I described when I described the Human Genome Project, we have begun in earnest to explore the role of genetics in health and disease. This slide shows the workhorse, so to speak, of these genetic studies, the genome-wide association study. Genome-wide association studies query the genome without specific hypotheses, for particular genes and biological pathways. These studies began in earnest in about 2007. Well, they began before that, but they began to be published in about 2007. And they asked questions about intervals across the genome and case-control status or variation in a phenotype, continuous variation in a phenotype. After a GWAS, we often produce these so-called Manhattan plots, which are shown here. And this is basically querying across the genome, across your autosomes and sex chromosomes. The evidence for association with lots and lots, millions of variants across the genome. And the Manhattan skyscrapers, so to speak, are those regions where we see the best evidence for association and where they likely harbor genes of interest that might be influential in susceptibility to our trait. Genetic discoveries rely on genetic variation that is present in the populations that we study. And unfortunately, overwhelmingly, they've been performed in populations with high levels of European ancestry. In fact, as you can see from this plot here showing the beginning of the GWAS era and here up through 2018, and this, this was a review paper by Alicia Martin. According to that GWAS catalog, about 79% of all GWAS participants are of European descent, despite making up here— this is the global population in billions— only 16% of the global population. So here's the population, here's the participants in GWAS studies. Anyone can look at this and say we've got a really big mismatch here. So because of that, our knowledge on the roles of genes as an influence on susceptibility to disease is limited and biased. Thus, I would say we have a major research disparity. That needs to be fixed. And that's what I have spent a lot of time in my career trying to do. So for the rest of my talk, what I want to go over today is my outline of 3 points. I want to try and convince you of the enormity of responsibility and the enormous need to make sure that diverse participants are included in genetic discovery. I'll begin this starting with the first point, where I will talk about obesity as a massive public health problem, with major disparities. So what is obesity? At the simplest level, it's an excess of fat. Body mass index, shown here, is typically used to measure obesity according to clinical criteria, which are shown here and defined on the left. It is basically a measure of adult's weight in relation to his or her height. It's a very crude measure, but it— and it has a lot of problems in terms of, is it really correctly classifying obese individuals? But it is widely available in many populations across the world and is so widely used as a study of obesity across many of the world's populations. There's been some discussion in terms of self-reported race, ethnicity criteria for our definitions of obese. Here, I'm showing you the European criteria, which is 30 or greater. There are some suggestions of a different criteria for African Americans. And for Asian American populations, for example. This slide shows the prevalence of obesity by self-report data on height and weight from the Behavioral Risk Factor Surveillance System from the CDC. Obesity is defined as a body mass index, um, as I said earlier, of 30 or higher, which is plotted on this graph. And you can see there's substantial variability across the states, in the prevalence of obesity. This is from 2019. I did look and there is a newer map, so I should have gotten— I should have grabbed that one, but it's, it's really alarmingly just the same. And so there's certain populations like here, Colorado, that has a very low prevalence of obesity. So I don't know what they're doing, but they're doing something very right. And then over here, there's a lot of red and orange where we have a high prevalence of obesity in these populations. Obesity has a tremendous impact on health. It has comorbidities affecting every organ system and every medical specialty. It has huge economic costs ranging from $150 to $200 billion per year. It has a huge cause of loss of life with about 3 million people dying each year from complications of overweight and obesity. It's really a huge public health crisis. Yet obesity is viewed often as a one-size-fits-all simple problem of energy balance. Yet most of our research today shows us that obesity results actually from a complex individual susceptibility and environmental factors. It's a complex multifactorial deal disease. The energy balance equation is affected by a host of individual predispositions that interact in very complex ways with environment, with lifestyle factors to influence behaviors, which in turn influence energy balance. I showed you before the map of the United States, but obesity is also a serious global health problem. According to the Noncommunicable Disease Risk Factor Collaboration in 2017, 2 billion adults are overweight, and that's a BMI of 25 or greater. 671 million adults have obesity, and by 2025, we expect that 1 billion adults, or 20% of the global population, will have obesity. Obesity prevalence is— also varies. Substantially across the United States by self-identified race and ethnicity. These plots show the obesity prevalence in the U.S. according to self-report of race and ethnicity averaged over 3 years. These maps help demonstrate the geographic and race and ethnic disparities in obesity burden. The combined data from 2015 to 2017 allowed for assessment by major racial and ethnic categories and found that non-Hispanic Black adults had the highest prevalence of obesity. They're here shown in the middle, and followed by Hispanic adults, which is 32.6% shown here, and with the lowest prevalence among non-Hispanic white adults of 28.96% shown in this map. There was a notable variability in obesity. One state, In— if we looked at a criteria of rates of 35% or higher and you looked in white individuals, only one state had a value where at least 35% of the population met that criteria. The most in the non-Hispanic Black adults and somewhere in the middle for the Hispanic adults. We also know that downstream disease risk, like for type 2 diabetes, for example, which is largely influenced by obesity, varies by self-reported race ethnic groups as well. These plots show females and males in the UK Biobank, which is a large electronic health record biobank study of 500,000 individuals. These plots show the prevalence of diabetes plotted against the level of adiposity by race and ethnic group. Stratified by sex. So, on the left here, we have women; on the right, we have men. As you can see, the prevalence of diabetes among nonwhite groups is fairly equivalent to the white groups at a low level of adiposity. So, here we have very lean individuals, and you can see that, at least especially in women, but also somewhat in men, that you have a coming together in the very— at the very lean levels of prevalence. In contrast, if I just highlight women on the left, this example compared with the women with BMI of 30, diabetes prevalence was equivalent in South Asian women with a BMI of 22.0, in Black women with a BMI of 26.0, and in Chinese women with a BMI of 24. Similar trends were seen in men to the right side of the slide. There's huge variability by self-reported race of when obesity or what level of BMI you start to develop type 2 diabetes. What are the major drivers of obesity disparities? Obviously, and as I've said before, lifestyle, cultural norms, healthcare access, psychosocial, socioeconomic disadvantage, environmental toxins, they all play a role in obesity susceptibility. And in fact, it's probably a combination of these factors. And yet our methods of interrogation have largely been limited to single exposures because of the complexity of trying to look at integrative analysis across all of this complexity. So I think we need to try and do better. We need to try and consider social and environmental factors in the context, in particular, of genetic susceptibility, which is one of the things that I will talk about today. So, in fact, I argue that some of the susceptibility to obesity may be influenced by genetic factors, which might be ancestry-specific or have complex interactions with environmental factors that are patterned across populations. And in fact, I like to think of our genomes as history books that have unbiasedly captured human experiences over thousands and millions of years. And I, and I ask you, and I ask myself this regularly, what is that book teaching us about obesity? So, now I'm going to turn— switch gears and talk a little bit about what we know about the genetics of obesity. As you all know, genes encode proteins, which represent enzymes, receptors, and hormones. This particular slide shows some of the main organs, so your brain, your pancreas, your liver, the adipose tissue that are involved in obesity. And they play a huge role in hunger and satiety and energy expenditure, and many of the main hormones shown that we know are highly influential in obesity pathogenesis. And these all have known structural genes that influence their expression. In fact, these hormones play a major role in monogenic, or so-called Mendelian obesity, which is typically seen as early onset and severe, so shown over here. These were among the first On the left— on the right-hand side of the slide are among the first obesity genes that were mapped, and they are mostly mapped for monogenic obesity. Up until very recent studies are starting to identify them— different variants in those genes as highly influential in polygenic obesity. Indeed, our conceptions of biological underpinnings of polygenic obesity, you know, that is the common forms of obesity that plays the largest burden on obesity worldwide are evolving. Polygenic obesity and rare severe early-onset obesity have been often polarized prior as distinct diseases, but now we know gene discovery has shown us that they have shared genetic and biological underpinnings, pointing to a key role of the brain in the control of body weight. Much of what we've learned about the genetic underpinnings of polygenic obesity have come from studies in the context of the Genetic Investigation of Anthropometric Traits, or the GIANT Consortium. It's an international effort between many individual studies and smaller consortia interested in the genetic underpinnings of anthropometric traits. I have had the pleasure of working in the context of this consortia for the last 20 years, where we've been assembling large populations with genome-wide association studies and combining them together into very large meta-analyses. As of 2020, the GIANT Consortium had identified 1,000 loci associated with anthropometric traits, including for body size and shape and measures of obesity. This slide summarized the findings from GIANT from over the last 15 years in the discovery of those 1,000 loci. Again, beginning in 2007, when we first started these types of studies, all the way up to 2020. And it demonstrates— obviously, one of the things you can see very clearly is a rapid expansion of loci discovery over the short term, but also a key problem, which is with much of the work of this GWAS. And I showed you this before in a plot that was looking across all traits. This is specific to obesity GWAS, and you can see blue is showing you the European ancestry populations, And this pink and purple or mauve colors are showing you African and Asian populations, and this little black bar up here, other. But as you can see, there's a massive overrepresentation of European ancestry participants in these obesity GWAS studies. Lastly, I would like to report that in our most recent effort, we have been able to assemble a sample size of 5.6 million study participants. We're working on those papers right now, and we have doubled the number of loci that are discovered by using such a large sample size, with over 2,000 signals for BMI alone. That being said, we still suffer from a lack of ancestral diversity. The large majority of that sample of 5.6 million individuals has primarily European ancestry. So what have we learned about the GWAS findings that we've matched? So these 1,000 loci for BMI, several hundred for central adiposity, etc. We do have a lot of challenge. We've identified all of these loci, but we still have much to learn from our GWAS studies. And still, indeed, as with many other common complex traits or common complex diseases, We're still majorly behind in translating GWAS signals to biology, but we have learned some key points, one of which I've already indicated in the uniting of monogenic and polygenic forms. And that is a huge overrepresentation of CNS regulation in, in overall obesity. So this is showing you some of the pathway analyses that we put together when we were doing our GWAS. Looking at overall— the overall GWAS studies and looking at— and both sets of these analyses here shown in B and A show a major overrepresentation of neuroendocrine pathways and tissues. These early loci, in fact, showed us the importance of the hypothalamus and the pituitary gland, which are both known appetite regulation sites, the hippocampus and the limbic system, which are involved in learning, cognition, and emotion, And our newest results also indicate other brain areas, including the hippocampus and the insula and substantia nigra, which are related to addiction and reward. And I find these loci extremely interesting, as many of them overlap with things like alcohol addiction and smoking addiction. Interestingly, the enrichment of immune-related cells, such as lymphocytes, that predominate in our central adiposity GWAS are not found— are found to be much weaker in our overall adiposity GWAS. Although enrichment analyses have provided preliminary insights into the broad biology represented by genes in these GWAS loci, determining which genes, variants, and/or underlying mechanisms are causal has proved an arduous task. On the right, I show exceptions to this rule where we've been able to go from locus to biology, and these have been extremely rewarding and wonderful. But on the left side, I show you one of our biggest loci that everyone probably— oops, sorry— has seen on there somewhere or another, the FTO locus. And this locus has been extensively studied. And the more we study it, the more we realize that it is a very complex story. It probably has a role of all three genes shown here, one long distantly regulated, the IRX3. I think probably some of you have heard about that story, with really interesting and intricate patterns of differential expression in adipose tissue, the brain, and more broadly. Most of what I've described today in terms of the genetic underpinnings of obesity has been very focused on European ancestry populations. I would argue, and the work that I've worked on for a long time is arguing that we really need to study diverse populations to truly understand the pathogenesis of any disease. As an example here today in obesity pathogenesis. I want to show you two cherry-picked examples from work that I've been working on to make some points about why it's so important to study ancestrally diverse populations and to try and convince you that if we study them, we really are going to advance science. So the first study, the example that I'll talk to you about is TOPMed. In this example, the, the Transomics for Precision Medicine Program, or TOPMed, is part of a broader precision medicine initiative, which aims to provide disease treatments tailored to an individual's unique genes and environment and social factor. TopMed contributes to this initiative through integration of whole genome sequencing and other omics, so metabolomic profiles, epigenomics, protein, proteomics, and RNA expression patterns. And it couples that data with molecular, behavioral, imaging, environmental, and clinical data. And if you think about one of the earlier slides that I showed you with the Human Genome Project, where it took us several years and many billions of dollars to sequence basically an amalgam of a couple of genomes, but one sequence. It should be quite revolutionary to think that at this point we have now sequenced hundreds of thousands of human genomes, and largely TOPMed has a lot to do with the success of that across in the United States, also the Genome Sequencing Project, and then also lots of other consortia that are worldwide. But it's a really exciting time to be doing this kind of science. This project, this example that I want to show you, is a project that I always call team science. If people ask what I do, I say I'm a team scientist, because when you have large populations and lots of people and lots of important questions to address in genetic epidemiology, you really need a big team to try and accomplish these goals. So I wanted to point out to the two junior leads of this project and to give them credit. Alice is my student, Jen Brody is at UW, and these two have done the lion's share of the work for this project. Another senior lead on the project is Anne Justice at Geisinger. TopMed in 2022, this is showing you the Phase 1 through 7 of TopMed where there's 180,000 individuals that have been sequenced, which is kind of mind-boggling. They're selected for different phenotypes. This plot over here is showing you the different phenotypes that have been chosen. Some proposals went after heart, some after lung, multi-phenotype, etc. There's a small component of studies that have been focused on sleep, but they're all aggregated together in one large database, and you can ask permission to use all of these samples which is what we've done for our obesity work, because most studies do collect, luckily for us, information on BMI. So that opens up a lot of studies for us to collaborate with to try and answer our questions. What I do want to draw your attention to here on the right is an exciting factor, and that is our little pie graph here, which is not way out of whack like a lot of the other plots that I've been showing you. We do in TopMed a much better job of trying to look and be inclusive of populations with ancestral diversity from Africa, ancestral diversity from American Indians, Asian, South Asian, etc. Real kudos to TOPMed for working to make sure that diverse populations are not left behind. We did an analysis of about 88,000 of those individuals. That had their whole genome sequence available and BMI and were willing to collaborate with us. This included about half of which were non-European ancestry. We performed a whole genome association, sequence association analysis in these 88,000 individuals who represent 15 self-reported race and ethnic groups from 36 different cohorts. We modeled BMI as a function of the single nucleotide polymorphisms across the genome here in the Manhattan plot, age, sex, study, self-reported race and ethnicity, principal components of ancestry to help us account for population structure, and then sequencing center and phase of the project to help us clean up some of the QC issues that might be present. As you can see from our result, we find— like everyone finds— huge evidence in support of FTO. It's in most studies, it's the biggest effect variant that's common across multiple populations. We also have found evidence for a lot of other well-known obesity susceptibility genes. But I'd like to draw your attention here to chromosome 22. Chromosome 22, we find an association that's genome-wide significant with rs111490516. This is in the MTMR3 gene, and I draw your attention to it because it has this— this variant in this gene is very diverse across ancestral groups. As you can see in our population, it is monomorphic in European, in individuals who are primarily European ancestry, so almost not found at all. And in African individuals who have African ancestry, it's found at a relatively common frequency, 13%. So in essence, by doing, by being inclusive of populations with ancestral diversity, we were able to map this susceptibility variant whereas previous studies had not been able to map it. MTMR3 encodes a member of the myotubulin dual-specificity protein phosphatase gene family. There's multiple transcript variants encoding different isoforms that have been found for this gene. This is borrowed from GTEx, which is an online resource of bulk tissue gene expression that is primarily from European ancestry individuals, But we can see it's rather ubiquitously expressed, and it's expressed in our tissues of interest. So adipose tissue, brain— sorry, adipose tissue, brain, and whole blood. If we look at our particular variant, which again, remember, is almost monomorphic in European ancestry, you see that it is not an eQTL for this gene, for gene expression in any of these tissues that we looked at. And that's largely because they, in this resource, We do not have diverse individuals. We do not have ancestrally diverse individuals. So one of the things that we're really interested in doing is looking in individuals that we have as a follow-up that have a lot of high-level African ancestry, and in TOPMed that also have information on BMI, where we can look and further probe into gene expression for— as an influence of obesity susceptibility. We're very excited about this finding and we've gone on to replicate it. We've been very lucky to have great response from our collaborators. We were able to engage investigators from the multiethnic cohort, the Millions Veterans Program, BioMe, which is a large biobank from New York City, UK Biobank that does have a sample of about 50,000 African ancestry individuals, and then the REGARD study. And so in our meta-analysis, the allele frequency was about 11%. So similar to what we saw, and we do see significant evidence for replication across these populations. So we're writing these, these findings up right now. We're excited about it. And we also would like to draw your attention to the fact that individual independent variants in this gene have also recently been associated with fat mass and fat mass measures, circumferences in the UK Biobank study, which is largely European ancestry. So, our variant is not in LD with any of these variants. So, this suggests that this gene is influential and that across different populations, different variants might be important for influencing susceptibility. So, what that means is that this gene is important in humans, There's different variants that might vary across different populations in their frequency that could be influential in pathogenesis of obesity. The second example that I wanted to talk about is a study of epigenetics. As with this last project, this is a large team science effort, and I want to acknowledge the person who's done the lion's share of this work, And that is my graduate student, Tay Glover, who I'm very proud of and will be graduating this year. And the other senior leads of the project, Mario Sims at UAB, who's a wonderful socioepidemiologist who's helped a lot with interpretation of this study because it's outside of my realm, and Lindsay Fernandez-Ralph as well. And then Anne Justice, who's the other major genetic epidemiologist that has been influential in this work. As I stated before, the largest factors I think influencing obesity disparities are likely social factors. So that's why I put this nature versus nurture street sign on the side, and I will spend a lot of time in this project describing what I think is an ideal project where we try and bring together social factors and genetic factors to try and understand the pathogenesis of obesity. And in fact, I tell all of my students that we need to throw nature versus nurture debate out out, and that we all need to agree that both factors are influential in health and disease. And honestly, you cannot separate the human genotype from their environment. They're inherently together. So the overarching aim of the work is to explore if epigenetic mechanisms are involved in the association of socioeconomic adversity or disparities with obesity. We want to ask the question, how is socioeconomic adversity embedded in biology to influence health and disease, and in this case, for obesity. As you all know, socioeconomic status is a broad concept that refers to the placement of persons, families, household, and census tracts or other aggregates with respect to the capacity to create or consume goods that are valued in our society. The 3 most common objective measures of SES that have been studied and collected in lots of large studies are education, occupation, and household family income. Now, these are not the best measures. In some ways, they're quite crude, but if they're available in a large number of studies and they're available historically so that we can look at the effect as it changes across time in a lot of our longitudinal studies. There are other studies that examine more exquisite measures of SES, but we will not be considering them here today because they were not available. So in terms of socioeconomic status, there's disparities in socioeconomic status. This is showing you the annual income of U.S. households by race and ethnicity. Here you have African Americans at the top, Hispanic Latinos, and then you have white and Asians down here. So you can see there's notable disparities in income, in household— household home ownership in the United States.. And this study that I'm going to be describing today is looking at African Americans, and African American families and communities often have lower SES when compared to white counterparts. According to the U.S. Census Bureau, Blacks have the highest poverty rate in 2020 of 19.5%. Racial residential segregation, physical separation of African American families to poor residential areas, is the foundation of socioeconomic difference between African American and white adults in the United States. And socioeconomic status has an impact on obesity in general, but in particularly on African ancestry— African American populations. Low SES is inversely associated with obesity in the United States and also globally. And here are the associations for men. Here are the associations for women. And in general, you do see stronger associations in women than you do for men. And this is a plot I should have explained of poverty income rate. The blue is showing you those individuals that have a poverty income rate greater than the 350th percentile. The green would be the average, 130th to the 350th. And this would be less than 130th percentile. And these are data according to the CDC. Epigenetics, or DNA methylation, directly affects the DNA in a cell and thus highly dynamic. In the process of DNA methylation, proteins attach chemical tags called methyl groups, which are shown here, at the bases of DNA molecules at CpG sites, which are shown here. The methyl groups turn genes on and off by affecting interactions between the DNA and other proteins. So it's a dynamic genomic factor that we can study, something that we can— is very much cell-specific. CpG sites cluster in the genome, and in particular, many are found around the transcription start site, which makes sense if you think of it as a DNA— as a gene silencing mechanism. We are able to study methylation quite efficiently these days with arrays. The Illumina array is a very popular array to study methylation across the genome. And on an array, we measure methylation in a population of cells. So here's your sample. Sorry. Here's your sample. Here's your cells. And you can— at these CpG sites, you can query whether or not they have— whether it's methylated or not. So in this cell here, you have 2 that are methylated. In this cell, you have 1 that is methylated. In this cell, you have 0 that are methylated. So an individual cell can be either 0, 0.5, or 1 at 1 CpG, and across the population we can get a continuous measure between 0 and 1. So socioeconomic— so socioepigenomics is a field that's very interested in trying to look at epigenomics as a mechanism of embodying socioeconomic adversity into influences on poor health or health disparities. There's mounting evidence of biological embedding of early life socioeconomic exposures. We also— several studies have additionally found evidence for adults' socioeconomic exposures. And for example, looking at low adult education and occupation has been associated with differential methylation of stress reactivity and inflammation genes.. And methylation loci may be sensitive to individual and community-level exposures or stressors across the life course. So at various levels of socioeconomic adversity, we're trying to study how it might influence disease. So our study took place in ARIC, which is a multi-study longitudinal cohort study of men and women from multiple ethnic— race-ethnic groups. 45 to 64 years. The focus of this analysis on African American participants. We've constructed a socioeconomic adversity score, and we've looked at Illumina methylation arrays across the entire genome, and we've modeled BMI as a continuous measure. The cumulative socioeconomic adversity score was developed where we looked at each SES indicator having 3 categories. Each of these categories was assigned a 0, a 0.5, or 1 based on socioeconomic index developed by James et al. And we looked at education attainment, household income, and occupation. We summed the values, and the total score is a continuous standardized— continuous standardized measure, which I show here. And we categorized it. We created a categorical measure based on the median split of— this would be high socioeconomic adversity and this would be lower socioeconomic adversity. The characteristics of these 2,600 study participants are shown here. On average, they are middle-aged, more predominantly female, have moderate levels of smoking and alcohol use, and overall on average are obese. So our strategy was to not skip through all my slides like I'm doing. We ran two separate models finding sites significantly associated with socioeconomic adversity and BMI, adjusting for confounders of smoking, alcohol take, age, and sex. We used R and EasyStrata to identify significant CpG sites that were associated with methylation on the one hand, and BMI associated with methylation on the other side— other hand. We took the sites that were common for these domains and carried them forward for formal mediation analysis. When we did this, we found that there are 26 sites that were significantly associated with both cumulative SES and BMI. In this slide, I'm displaying the 17 sites that were significantly associated and the results from the mediation analysis. This is the indirect effect or the mediated effect of these various CPG sites. So these are all the various 17 CPG sites. And as you can see, as you notice from the plot, some of the sites had very small mediated effects, but others were quite large. And in fact, we're very excited about this finding for the SOX3 gene, as this gene is associated with leptin sensitivity, obesity, overnutrition, and cumulative stress. So, this makes a great candidate gene story for mediation of socioeconomic adversity and its influence on obesity. We took forward this data for replication in the Jackson Heart Study, which is a single-site longitudinal cohort that is very— it's collecting its data from Jackson, Mississippi, from the same area where Eric's study was collected. So we've excluded any individuals that overlapped from these studies, and this is truly an independent sample of African Americans from that Jackson, Mississippi site. We extracted the sites of interest from the Eric analyses, the 26 sites, and we had a small, slightly smaller sample size in Jackson of 1,700. We took those forward and we were very excited that 3 of the sites significantly replicated in our mediation analysis. So they demonstrated directionally consistent, statistically significant evidence of mediation, just like we did with the largest effect in this SOX3 gene. And I'll just go really quickly back to the slide because I wanted to show you these are the 3 sites that mediated, but for the majority of these sites, even though they didn't have a statistically significant p-value, they were directionally consistent, and in some cases associated with BMI and SES. They just didn't demonstrate significant mediation. We're really jazzed about these findings, and we are moving forward to vet these findings more with trying to first look at the components of the SEA score and decide which ones are the most influential. Next, we're going to move to replication and generalization in other populations. Indicated over here on the left are the populations that have already agreed to take part in our study. We really want to dig into what is the biological basis of socioeconomic adversity effects on CpG and look into the ultimate mechanism. CpG methylation, we say, is a gene silencing or gene— it's going to silence this gene expression. We actually have gene expression data, and so we're going to look, does this CpG methylation actually influence expression in our one tissue of interest that we happen to have, which is whole blood. But we'd also like to go beyond whole blood and look in adipose tissue, et cetera. So in summary, I hope that I have convinced you that ancestrally diverse populations will advance science. The differing burden of obesity, both globally and within populations, suggests an imperative for diversification of biomedical studies. I first described a novel loci for a novel variant for obesity susceptibility that was common in African ancestry haplotypes. I next talked to you about how socioeconomic adversity effects may be operationalized through changes in CpG sites in DNA methylation and their effects on obesity. I would just end by saying, you know, our findings demand a reevaluation of how future genetic studies are designed and implemented. And we strongly advocate for continued large genome-wide efforts in diverse populations to maximize genetic discovery and make sure that all populations benefit from precision medicine advances. And I'll just end by thanking the NIH, who's been extremely supportive of our work, and all the study participants and study personnel and co-authors of all the papers that we've worked on, and the wonderful team that we have in our computational lab. Including many faculty affiliates, postdoctoral scientists, predoctoral fellows, and programmers. So thank you for your attention.

Thank you very much, Dr. North, for this informative and thought-provoking presentation. I'd like to remind attendees that questions can still be placed in the Q&A, and I will moderate the discussion by conveying some of the questions that have been asked. So, first of all, going back to the genomic studies, with such large sample sizes in current studies, what do you think are some reasons for the continued disproportionate representation of European participants? And do you feel that there is enough of an effort being done to change this?

Well, I think it's the historical legacy and The bad things that have been— there's some distrust, I think, in some populations in participating in studies, and rightly so, because of the Havasupai and various problems that have occurred in the past. And I think that you need to really work on building trust in these communities and implementing studies where individuals who are recruiting and studying individuals are those very same individuals from the population. So engaging them in research. And when you do this, it can work quite effectively. And there's several population-based studies. I work in a population on the border in Texas where we do that. Everyone is from that community. There's a great trust in the medical community, and we really engage and give back to the community. So I think that that's really important. And I think that there are a lot of efforts, and I'm really grateful that NIH is actually starting to make huge efforts. I'm hearing that pharmaceutical companies are trying to diversify their, their clinical trials. So I think there's a lot of movements forward to make things better, but it's just going to take a little bit of time, I think, to catch up and be able to recruit the sizable numbers of populations that we need for these very large genomic studies.

So the next question relates to what happens when we actually get there. So once the genomic data available has more diversity from non-European ancestral populations, Do you feel that some of the mapped loci in the currently top loci will wash out or become less relevant?

I don't think that they'll become less relevant at all. There, you know, all of these traits, and especially like polygenic obesity, it's highly polygenic. There's lots of different genes and there's lots of different variants that are influential and important. There's a lot of movement towards trying to use polygenic risk prediction. And in order for polygenic risk prediction, we really need to have trans-ancestry populations that inform those predictions. What we found in work that's supported by NHLBI, an R01 that we have, when we use trans-ancestry large populations as training sets to, to start our risk scores and then to predict them into populations, they have much better performance than if you use a PRS, the parent GWAS population that is exclusively European ancestry, for example. And I imagine as samples expand larger and larger, and as our methods improve for polygenic risk score estimation by accounting for local ancestry differences and things like that, I think we'll do a much better job. So I don't— I think that it's in certain populations, you, any population that you look at, there may be a variant that's more or less important in that population., but they're all gonna have, you know, all human genes. There's gonna be a representation of all of these genes that play some role. There's gonna be different things that rise to the top in different populations, but they're all going to be important and influential.

And regarding methodology, we have a question on the usefulness of local ancestry admixture mapping. Is that, does that still have a role?

I think it absolutely does. So we have— I work on a study called the Hispanic Community Health Study, Study of Latinos, and we have been doing— using ancestry admixture mapping. I think local ancestry does have a great role to play, especially in populations that are admixed, like African Americans and Hispanic Latinos. You couple that finding the locus, trying to understand the local ancestry pattern there, and then that can help you inform haplotypes what can be happening on that haplotype, what the actual causal variant is. So, as a benchmark of jumping off to functional studies, I think it's a really great idea still.

So then, regarding sort of deepening the phenotypic work, do you have— in these cohorts, are you analyzing data on, for example, visceral versus subcutaneous fat, and of course, diet, microbiome? I mean, there are so many things linked here. So, what are the main sort of directions you're going in with these types of multifaceted analyses?

Oh, man, I could have planted that question. So, that's an excellent question. And of course, I emphasized in my talk quite a bit the crudeness of BMI. And so, this is a problem. And I think that more exquisite measures like DEXA, and et cetera, are really exciting avenues. One of the things that we're trying to do is instead of looking— you know, obesity heterogeneity is important. And if you're looking just at one BMI measure at one point in time, you're missing a lot of the picture. Severity matters, duration matters. So we've been doing some measures, trying to use multiple measures in a cohort and trying to look at, like, trajectories across time, weight gain, weight change. We're also using more exquisite measures for studies that are looking at gene expression, for example. So in the Cameron County, that study on the border population that I mentioned, We're looking at severe obesity and differential expression in whole blood using severe obesity as a measure. Then we took those findings and then working with collaborators that have adipose tissue, trying to replicate those findings. And so, trying to understand what that signaling cascade is. We're looking in the brain, we're looking in adipose tissue, we're looking in whole blood, but looking at the protein products, the proteomics. So, more and more our work is becoming more multi-omic. Starting with the germline variant, but then going broader and seeing how the translation of those effects occur in the body.

So we've pivoted to the multi-omics and the epigenomics, which is a very interesting part of your talk. A question here is how you take into account the genetics that affect the methylation levels.

That's such a great point. Yeah, so that's a real problem in this work, right? So one of the things that we do is we exclude those methylation sites that are very clearly controlled by nearby gene, nearby SNPs. So that's one way. The other way is to just try and test. You can look at the distribution of the methylation and the peaks of it. Are you seeing evidence that it's being driven by a SNP? So there's lots of different ways of QCing, so to speak, that you can try and our knowledge of the genome where that's relevant. So that's what we're trying to do for right now. But it is an issue. And I think that there are some interesting methods development in that arena to try and solve or help us with that problem.

And also about epigenomics, epigenomics is cell and tissue specific. And so how— what do you feel will be— what will be the utility? What will be the right samples to look at that? And also using epigenomics as a predictor, given that you need to be looking in the right tissue.

Any thoughts on that? That is such an important question. And I get it on all of my grant submissions. So some things are ubiquitously expressed. Especially when you have a disease like obesity, where you have CNS regulation and signaling and, and the such. So, there's going to be some things that you're going to pick up. You're going to miss some things. There's no doubt about it. But we are working right now to try and get those tissues. So, one of the things, luckily for our field, obesity, a lot of people are willing to part with some of their obesity tissue so that we can study things like gene expression and methylation. And a lot of the work that I'm focusing on now actually is doing that to try and see, to answer questions, you know, how useful is the whole blood samples that we have? I think that there is, there's, we know from GTEx that there is a lot of overlap, but there's also going to be some unique things that we can only know by looking at the specific tissues that are playing a role.

Thank you for those answers. We are unfortunately out of time. So thank you very much again, Dr. North. For a very inspiring presentation. And thank you to all our attendees. And thanks and goodbye, and come back for another talk in a couple months.

Thank you so much, everybody. Bye-bye. Bye.