[MUSIC] [BLANK_AUDIO] I would now like to talk about the the genetics behind the common forms of type 2 diabetes. Again, I showed you the atlas of non autoimmune diabetes. In the upper left corner, you will have the monogenic forms of diabetes and we just learned that generic diagnosis enable us with a treatment. You can predict the development of diabetes and we have it soon for genetic counseling. In the lower right corner, we have gained a lot of insight into common variants associating with type 2 diabetes. And you can see all these variants they're very prevalent in the population, the frequency of about 5%, but the risk increase is modest. Until 2006 we didn't know much about the genetics causing common form of diabetes. Using the candidate gene approach, the variation or the PPARG gene and the KCNJ11 gene was described in 1998 and 1999. In 2006, the TCF7L2 gene was identified when they decoded in Icelandic was studying familial form of diabetes. But then in 2007 there came an explosion in the technology and in our ability to study common forms of the disease. This was the Genome-Wide Association Study. This method is a hypotheses generating method. It needs large sample sizes and as you will see, very large sample sizes will improve your power to identify genetic variance associating with disease. It's a two stage design with a discovery sample and a replication sample. It's also technology driven, so it's the progress in genotyping technology which has enabled us to examine millions of variants in individuals, and these variants we know from the human genome project and the HapMap project so that we can choose variant and put in chips and when we do not have the chips we will have a very high throughput method to do the genotyping. I said that it was a two stage design with a discovery stage and a replication or follow-up stage. Typically in the discovery stage in the early days, we genotype maybe 3 to 500, 1000 variants. Now we genotype up to 5 or 10 million variants. Either they are directly genotyped or we're using a technology where we recall imputing so we get information also on variations which you actually have not genotyped. We can do it in a case control setting. We have part of our study sample are those with the disease, and the remaining are those, the controls. We could also use it for cohort studies, and we are using a lot of tests because we have these millions of genotypes we're testing, so we need a very, very significant a threshold in order to say we have identified a novel gene variant associated with disease. So we, we have a significance level, which should be below 5 times 10 to the minus 8. So this is an example of a genomite association study. On the x axis you have the chromosomes. On the y-axis, you have the significance level and each dot here represent a statistical test. And you can see for example, on chromosome 10, there are several dots which are very significant of 2 10 times to the minus 15th. And they are also regions in the genomes other regions which is quite significant. But we should ensure that this is not just chance finding. So repeat the most significant regions, and then we do replication genotype in a new case-control data set. This could sometimes be just a few SNPs, sometimes it could be many SNPs we are genotyping. And then we do meta-analysis on stage 1 and stage 2, and those who get below the significance threshold of 5 times 10 to the minus 8, we will consider regions in the genome with significantly associated with disease. Normally data presented with a Q-Q plot of quantile quantile plot, this is the upper part of this figure on the x-axis. You have the expected significance, a level on the lockes scale, on the y-axis you have the observed significant level on the y axis, and if you have a excess of very significant observed findings, it's an indication that you have some true significant association to disease. On the lower part, you will see a Manhattan plot. Again, you have all the chromosomes on the x-axis and on the Y axis you will have the significance level. So this is the way data they are presented. Genomite association studies has provided an explosion in the number of identified risk variants for complex traits, and that's all different complex traits. Until 2007, we knew very few gene variants associating with any complex trait. But since 2007, this figure illustrates that we, the blue columns have a huge number of publications describing genomite association studies. The red column, how many new regions in the genome associating to any traits we identify each half year, and the green column, or the green line, illustrates that we now know about more than 2,000 SNPs in the genome associating with a complex trait. And if we only look at type 2 diabetes, then you can also see that that's all the, the green columns that genomite association studies has provided us with a lot of genetic insight since 2007. Here we have the years on the x axis. And you can see the first genomite association of studies was provided in 2007, and the height of each of these bars illustrates the effect size of each of these variants. You can see that many of the green bars in 2007 and 8, they had a risk increase per SNP about 10% to 20%, but with the increasing number of variants we identify, the risk increase is less and now many of the variants we discover, the risk increase is below 10% for each of the variants. Can we use all this new information actually to predict who will develop type 2 diabetes? So in this study, 37 of the validated type 2 diabetes SNPs has been examined in what we call a receiver operating curve in order to see whether we can discriminate between type 2 diabetes, and those who don't have diabetes. In such a curve, we would really like to have high sensitivity and high specificity, so we would like to be at the upper left corner, that is the perfect test with an area under the curve of 1. If we have a test which does not work which is similar to flipping a coin, we will lie on the diagonal in this figure and we'll have an area under the curve of 0.5. If we have a useful clinical test we will lie on the blue line or above the blue line with an area under the curve of .8 to .9 and what about if we apply genetics? We only have an area under the curve of .64. So we are not there yet where we can use genetics to predict who will develop type 2 diabetes, but it gives us a lot of biological insight. And how can we use this new genetic knowledge to gain biological insight? We can do association tests with intermediary phenotypes. We can see whether the variants affect beta cell function or insulin sensitivity, or for example, body mass index. We can look at expression profiles of different genes in different tissues, and we can do in vitro or in vivo functional studies. And this has revealed that many of these variants, they actually influence the function of the beta cell. So decreased beta cell function, decreased insulin secretion, will lead to development of Type 2 diabetes. A few of the variants are known to influence insulin resistance. Increased insulin resistance will lead to Type 2 diabetes. And one of the variants, this primarily affecting obesity, which then causes insulin resistance and type 2 diabetes. But even though we now have a tool to study the underlying pathophysiology there are still many many of these gene variants where we don't know how they actually is causing type 2 diabetes. Also we can look at the pre-diabetics rates. And for most of the pre-diabetics rates now we can explain for example for insulin secretion up to 10% of the variation and its rate. Here, I show you some data on body mass index and this is a study in international collaboration, where we look at 250,000 individuals and you can see a lot of loci pops up in this Manhattan plot. The most significant is the FTO gene. And now, using this common variants, we can then explain up to 10% of the variation in BMI. Also very interestingly, if you look where these PMI SNPs are located in the body. Many of them seems to be expressed in the brain. We can also look at distribution of the fat. So here we have looked at waist hip ratio corrected for BMI. And then another set of genes pops up. If we look at the expression of these genes, then they are not expressed in the brain but more in the peripheral tissues, like the fat tissue, the bone tissue, the muscle tissue. So this indiciates that BMI, total BMI is determined by expression in the brain, but the distribution of fat and muscle that is determined by another set of genes. So how much of the variation in BMI is explained? To the left here you have a slide where we count off the number of risk allele increasing your BMI, on the y-axis you have the number of individuals and you can see to the far left that those with few risk alleles, they will have an average BMI of 25.5. And then with increasing numbers of risk allele then there are those with more than 38 risk alleles, then on average, they will have a BMI of 28.5, so a BMI increase of 3 units. So each allele will increase your body weight with 4 to 500 gram. That's highly significant. And in the population, those with few risk allele will have a body weight which is approximately 8 kilogram lighter than those with many risk alleles. So we really start to get an impression of some of the factors determining body weight. But can we use it to predict those who will be obese? This is again a receiver operating curve, and the answer is no, the area under the curve is below .6. We also, in one of the most important genes you now have the opportunity to look into gene environment or gene lifestyle interaction. This is a Danish study. The Inter99 study. It's about 6,000 individuals and the low risk genotype TT is found in nearly 2,000 individuals and you can see their BMI on average is 25.9, and if you compare the TT carriers to those with the high risk genotype the AA carriers then the AA carriers have a BMI which is more than one BMI unit higher than the TT. They have a body weight which is 3.5 kilograms higher. They have a waist circumference which is 2.5 centimeters higher. So, just one gene seems to influence the body weight, to height degree in the population. We then ask the question, those who carry this variant, does it interact with physical activity? And here we have compared high risk AA carriers to low risk TT carriers. And if you have low physical activity, it's not one BMI unit, it's two BMI units you are gaining in weight. But if you're physically active, the influence of this risk variant is much less. So to sum up for the, what we know about common variants, we have now identified more than 65 regions in the genome causing Type 2 diabetes. We have identified regions affecting plasma glucose, circulating lipid levels, BMI and obesity, hypertension and cardiovascular disease. In general, these variants, they have low impact at the individual level, but explain a lot at the population level. Many of the variants, we don't know, in which gene they are influencing their effect. We don't even know whether the SNP is the causative SNP or whether it's inherited together with another SNP which is actually the causative. And for many of these traits we can now explain up to 10% of the genetic contribution to the trait variation for lipids, we can explain much more. So back to our atlas of the genetics of non-auto immune diabetes. To the upper left corner, as you have seen, we have very good information on a lot of genes which we can use in the clinic to improve treatment, to say something about prognosis and genetic counseling. Now to the lower right corner, we have gained a lot of new knowledge of common SNPs which increases risk of diabetes but they have low penetrance. The effect is too small to use it at individual level, but all this information is very important tool to gain novel, biological insights. But we can explain now, for many traits about 10% of the heritability of complex traits. So where should we look for the missing heritability? Should we look more into gene-gene interaction, or gene environment interaction, GxG, GxE? Could the heritability be over estimated? Is this structural variation? Are there more common variants? Are there epigenetic effects? Or could it be rare and low frequency SNP's? A lot of the focus these years are on many of these areas, but most research is actually focusing on trying to describe the rare and low frequency SNPs. And there the rare variant hypothesis states that a significant proportion of the inheritance of complex disorders is due to the cumulative load of rare variants. And in order to understand this, we need to have international collaboration and have listed some of the most important ones. These are collaborations using excellent chip data, excellent chips are chips with information on the variation of the populations which are in the coding regions of proteins. They are using other chip data, and they're using whole x-nome and g-nome sequence data. There are go Type 2 diabetes consortium focused on looking at Type 2 diabetes gene variants and related phenotypes. The Magic contortion glucose, homeostasis traits and insulin secretion and action. There's a Giant consortium anthropometric traits. the Global Lipid Consortium looking at lipid levels and the Charge consortium, looking at traits related to, for example, pulmonary function, inflammation, and other traits. But we have already got some insights into the low frequency and the rare variants. We can see that both the low frequency and the rare variance they also have an impact of the risk of Type 2 diabetes, but the effect size is maybe not as big as we have hoped, so, so we need much more insight before we can use this as a clinical setting. So, where should we look, and where are the the research hitting in the coming years? That is to identify more common variants. We will use exome an whole genome sequencing in large population to identify new variation and their variants, we have to work with very large collaborations to identify the low frequency variants and to understand what they mean. We're using exome and whole genome sequencing in isolated populations to look for variations which are specific to disease in various population, and then, as I said, we will have structural variation, epigenetic modifications, gene x gene interactions and gene x environmental interactions, which will also explain part of the genetic risk. So to sum up what we know about genetics of diabetes, the identification of monogenic forms of diabetes is important, it makes the right diagnosis, it helps counseling and it can guide choice of treatment. We also have experienced many new genetic variants increasing the risk of common Type 2 diabetes. They cannot yet be used to predict Type 2 diabetes. They cannot yet be used to guide choice of treatment, but these years we get a lot of novel, new biological insights. Genetics are here to stay. Thank you for your attention. [MUSIC] [BLANK_AUDIO]
