 So, Terry and I thought that maybe it would be worthwhile for me to discuss means by which we've identified causal variation in HIV disease, which is what my lab works on primarily. But I have to say that just from what I've heard, I think you have all identified, you've really honed in on the allele in each case almost. And not only that, there's some really nice work to show why that specific allele is having the effect. So, I feel like I'm going to be preaching to the choir and that probably you've moved ahead of many of the areas in which I work, so I just wanted to make that clear. Now, the human major histocompatibility complex is the most rich 4-megabase region of the human genome in terms of association with different types of diseases. And some of those are shown here. There are a lot of different types, autoimmunity, cancer, viral disease, bacteria, and even some oddballs like schizophrenia and Parkinson's, which we didn't expect to be involved with HLA class one or class two. Now, I don't have drug hypersensitivity up here, but I really need to go back and do that because in most of the, in the vast majority of these cases, there is nowhere near the odds ratios that you are seeing in association with any of these diseases. So, in HIV disease, we feel excited about an odds ratio that goes above two. So, anyway, I will definitely get drug hypersensitivity up here and it's really, I think, an important point. Now, HLA variation, we think of it as being really important in antigen presentation. So this is just the peptide binding groove of a class one molecule with a peptide sitting in the groove. And we know that a lot of the effects that we see when we look at disease actually are due to the types of peptides that sit in the groove and the ability of individuals with specific alleles to respond appropriately to a given pathogen, for example. But there are other characteristics of the class one molecules, and I'm sure this is true for class two, though we haven't studied that as closely, that really should be considered when you think about the role of variation at class one in human disease. And one of those is the ability of the peptide binding groove to bind receptors on innate immune cells. So these lillers and the cure, I won't have time to talk about those there encoded on chromosome 19, the MHC is on chromosome six. These molecules that are expressed on innate immune cells such as natural killer cells and certain monocytes and dendritic cells, they actually bind, their ligands are HLA class one. It's this region of the peptide binding groove that the cure receptors recognize. So they see this region and they can send activating or inhibitory signals to NK cells, for example, and affect the innate immune response. So it's not just the acquired immune response, it's also the innate immune response. More recently, and what I will be talking about, the example I'll give today is their differential expression levels, which we think is having an effect, modifying the effect of alleles on different disease outcomes. And I won't have time to talk about this, but there's been a good bit of data recently to show that certain HLA class one and class two alleles actually serve as trans expression quantitative trait loci. For some reason, they appear to be associating with the expression of unlinked genes, other genes in the genome. And all of these things I think should be considered and probably a lot more than these in terms of when we start thinking about what is the effect of HLA on disease. So the example I'd like to discuss is HIV control, and we know there are HLAB alleles that have been shown consistently to associate with HIV control, and I've just put three of those up here. The HLAB locus is really key in terms of allelic effects on HIV disease. But recently, well, not so recently, but Jacques Faelet in David Goldstein's lab identified through a GWAS a variant between HLAC and HLAB that associated with viral load set point. And we followed up to find out whether in our cohorts we also saw an effect of that SNP, which is 35KB upstream of HLAC, but between HLAC and HLAB, those two are only 150KB apart. And indeed, we saw that there was an effect on viral load of this minus 35 SNP, what we call minus 35, where the TT associated with higher levels of viral load and CC with lower viral load. And this agreed nicely with the data from David's lab. We also showed in normal donors that TT associated the same minus 35 SNP associated with low expressions of HLAC and the CC with higher expression. So was there a connection between these two? Higher expression levels associating with CC also associated with lower viral load. What did the effect of minus 35 have to do with expression levels? And I won't, I'm going to give you a summary of what we did. Now I'll say minus 35 is upstream, 35KB upstream of HLAC. It's not in the promoter region proper. And the effect was in whites, and the effect was absent in blacks in GWAS. These things made it very unlikely that minus 35, which was identified through GWAS, to be the causal SNP. Now we started looking at promoter variants and 3-prime UTR variants, and I'm just going to give you a summary of what Smitho Cochrane identified. In the 3-prime UTR of HLAC, Sumoleos of HLAC, there is a binding site for a microRNA. MicroRNAs are known to down-regulate expression of certain messenger RNAs, either by degradation or decreasing their translation. What Smitho found was that Sumoleos have a mutation in that mere binding site, whereas other ones have the appropriate binding site for this microRNA. This variant in the 3-prime UTR in whites is in good linkage disequilibrium with minus 35. So now as I mentioned, HLAC is very close to HLAB, and one very large concern was whether the effect that we see with minus 35 or even with this variant, and I'll get to that, could be due to HLAB. So I'll come back to that in a second. So individuals who have HLAC alleles with the microRNA binding site have lower levels in general on the cell surface than those individuals who fully escape the microRNA regulation. They have higher levels. Now the next part I'm going to tell you really eliminated the possibility that this was due to HLAB, linkage disequilibrium with HLAB. The microRNA gene is located on chromosome 7, unlinked to the MHC. Having an insertion at this position in the microRNA associates with low levels of microRNA, Smitho showed. Having the deletion mutation, the deletion variant in this gene, associates with high levels of the microRNA. This variant associates with HIV outcome, but only among individuals who have HLAC alleles that are inhibited by the microRNA. In these individuals, this variant has no effect. So this eliminated the possibility that this was due to HLAB because HLAB is not regulated by this microRNA. Now, so the minus 35 SNP is associated with HIV control in GWAS in whites, but not blacks. That SNP associates with C expression in whites, but not blacks, the minus 35. And we think it's unlikely to be the causal SNP. Variation in the 3-prime UTR of HLAC that's in LD with minus 35 in whites that determines microRNA regulation affects HLAC levels and associates with HIV control in whites. Among those who have C alleles that are regulated by the microRNA. A variant in the microRNA gene interacts with the 3-prime UTR of HLAC. This strengthens the causal effect of HLAC 3-prime UTR and rules out the possibility that simply marking the HLAB locus. Now, and I've got to zip through this because I know I'm running, I'm going to be running late. Richard Aps in the lab went back and looked at HLAC expression levels across individuals, each dot is a different individual dividing them by the HLAC types. And there was a significant association of expression across HLAC types where C1402, for example, is expressed significantly higher on the cell surface on average than is, for example, C0302. If we use C expression as a continuous variable in our analyses, okay, so now we know the genotype of HLAC, we can assign expression levels by imputation, and we can ask, using HLAC expression in our cohorts as a continuous variable, we can ask whether it has a significant effect on HIV control. And this is a much more direct way of doing them, for example, with the minus 35 SNP. And indeed, higher HLAC expression associated significantly with better control of HIV. And it did so in various disease outcomes that are shown here. Some of these are completely independent of one another. And as you have seen as well, you're seeing effects of specific HLA alleles across distinct cohorts, and that I think is really an important point. We're seeing this also in African Americans, so across different populations, HLA, higher HLAC expression confers protection. We see it across different outcomes of HIV infection. We see it across populations. And finally, we think it's quite important to look at the functional significance of any genetic association, and I think in general this group has done that and is beginning to do that for more of the drug hypersensitivity studies. Here we looked at the frequency of HLAC restricted responses to HIV peptides and found that it correlates with HLAC expression levels. This is a complicated slide, actually. Philip Goulder at Oxford University had data from over 1,000 individuals from South Africa where they looked at overlapping peptides across the HIV proteome, and they knew exactly what CTO responses were present in each of those 1,000 individuals. We asked then, so for example, this dot represents individuals who have C14, C1402, remember, I told you that's a high-expression allele and that's what's on the x-axis. On the y-axis is basically the odds ratio of a CTO response against C1402 restricted epitopes in this case. In this case, these are the C0102 individuals, C01, and they would restrict different epitopes because this is C01. Basically what this shows is that the higher the level of expression of the C allele, the greater the likelihood that individuals with that C allele would respond to their specific HLAC restricted epitopes. Higher expression results in a greater CTO response. It would be actually interesting to look at expression levels in the context of some of the cohorts that you've mentioned today and yesterday. What I've shown you is that measurement of HLAC expression levels across alleles allowed a direct test for association between C expression levels and HIV control where we take into account individual allelic effects rather than involving the minus 35 SNP proxy or a variant that only accounts for part of the differential expression, which is true of the microRNA. It does not account totally for differential expression of HLAC. The effect of C expression levels is consistent across black and white cohorts and across different HIV outcomes. Really important in validating any effect, any genetic effect. And then functional data explain or support the effect of HLAC expression on HIV control through enhanced CTL activity. It's possible that it's also having an effect on NK natural killer cell activity as well. So I'll stop there and what I've tried to do is just walk through an example where we've tried to really pin down what is the variant that's causing the effect. Which locus is it? What variant is it? And can we generate functional data to support that? Now be happy to take questions. So we have time for a couple of questions. Just one question. When you were doing your comparison with blacks, was that African-Americans? So I can speak louder? I'll speak louder. People don't tell me that a lot. Is that with African-Americans or was that also people from the continent? So the data that I showed you from our cohorts is with African-Americans, but the functional data was with South African blacks. And they are infected with two different clades of HIV, so the South African are primarily infected with clade C and Americans are primarily infected with clade B. So we think that this is going to be across the board. We think that if we looked in a Chinese cohort, we would see the exact same thing. Yeah. Thank you. Matt Nelson at GSK. Really interesting results. Is the effect of the different HLA alleles, B or C, is that simply, are those just a proxy for HLA C expression? Do those different allele types have an effect even after accounting for the differences in expression or are they just simply along for the ride? So it's a good question. And even given everything that I've shown you, it's not always easy to parse that out completely, but we do multivariate analyses where we include individual alleles, all individual alleles of A, B, and C in the analysis along with expression as a continuous variable. And out of those models, you get specific alleles like, for example, B57 still shows up, B81 still shows up. Alleles are well known to be protective alleles, and there's a lot of data because of the type of epitope they present. So we were happy to continue to see those, and B3501 shows up. So in this case, C1601 shows up even though we're using C as an expression of C as an allele, so we assume that 1601 has an independent effect, independent of its expression. I don't think it's so easy to totally tease all of those apart. Yeah. Thank you. Mary, you're... Thank you very much. I wanted to ask you. We talked about biomarkers for risk, and clearly the quantitation, the set of expression of the particular HLA that might be the putative HLA for, let's say, a skin reaction, the level becomes an important measure itself. So the question is, as a translational measure of risk, given that these levels are fluctuating over time in the context of infection and other things that are going on, is there a... How do you think about the idea of quantitating HLA levels? If you knew, let's say 1502 was a problem of carbamazepine, to determine whether there is a set point for those individuals who will then go on to really have a problem versus most people with the same marker, who will not if the quantitation becomes a key question and whether that can be measured as a translational measure? So when we do these measurements, it's actually... If you measure self-surface expression levels, and we did these from healthy donors, it's tricky because you need to measure them on a fax machine. Everybody that you're measuring, if you're going to compare them with one another, you need to measure them at the same time without recalibrating the fax machine. So the way we did this with 200 people is at Duke University, they brought the people in every day for a week and a post-doc just churned through them for that week so that we could compare those. Now you can look at CDNA and that way you can freeze, but for example, if you had, it would be interesting to look at people with 1502 and see whether those people who are tolerant of the drug actually express lower levels of 1502, and you can do that by CDNA. Once an individual has a reaction, there is not, I mean, that's probably massive immune activation in those individuals, so that is probably not going to measure their steady state levels of HLAB. You'd have to catch them before you have this massive disease going on, I think, yeah. Thank you. Because of time, we have to move to the next talk, sorry, 5th during lunchtime. Thank you. Thank you. So the next talk is on drug-induced liver injury and across reactivity, across drugs and organs, given by Jay Huffnagle, director of the Liver Disease Research Branch at the National Institute of Diabetes and Justice and Kidney Diseases at NIH.