Erik Ingelsson; Advisory on EHR data; Precision Medicine Update

Getting Personal: Omics of the Heart

Jane Ferguson: Hello, and welcome to episode two of "Getting Personal: Omics of the Heart". I'm Jane Ferguson, an Assistant Professor of Medicine at Vanderbilt University Medical Center. This Podcast is brought to you by the Functional Genomics and Translational Biology Council of the American Heart Association.

If you're a current or prospective member of the American Heart Association but not yet affiliated with our council, I do encourage you to join us. FGTB is a vibrant council with a diverse membership spanning disciplines from basic research to clinical practice, with shared interests in genomics, precision medicine and translational research.

You can find out more by going to the AHA professional website at professional.heart.org and selecting FGTB from the list of scientific councils. If you're listening to this, you've obviously already figured out a way to access this Podcast. We do have several convenient options to make sure you never miss a new episode. You can stream each episode and find additional information on links to articles on the Podcast website fgtbcouncil.wordpress.com. You can also subscribe to the Podcast on iTunes or if you are an Android user, you can subscribe via Google Play. Just search for "Getting Personal: Omics of the Heart" and click, Subscribe.

In this episode, Kiran Musunuru talks to Erik Ingelsson about research from his group on epigenetic patterns in blood and how these relate to coronary heart disease, which was published in the February 2017 issue of "Circulation: Cardiovascular Genetics".

We highlight a recent AHA Science Advisory on merging electronic health record data and genomics, and Naveen Pereira and I discuss precision medicine and whether it can live up to the hype.

Kiran Musunuru: Hello. This is Kiran Musunuru. I'm on the faculty at University of Pennsylvania and it's my pleasure to represent the Functional Genomics and Translational Biology Council of the American Heart Association. Today I have the privilege of interviewing Dr. Erik Ingelsson who is Professor of Medicine in the Division of Cardiology at Stanford University School of Medicine. We're going to be discussing a very nice paper on which he is senior author that was published last month in "Circulation: Cardiovascular Genetics" titled "Epigenetic Patterns in Blood Associated With Lipid Traits Predict Incident Coronary Heart Disease Events and Are Enriched for Results From Genome-Wide Association Studies". It's all right there in the title. Erik, welcome.

Erik Ingelsson: Thanks.

Kiran Musunuru: It's a pleasure to have you. Maybe you can say a word or two to introduce yourself and your research interests.

Erik Ingelsson: Yeah, it's a pleasure to be on. Yes, as you said I'm a professor of Medicine at Stanford, an MD, PhD trained really in epidemiology but started to do genetics about 10 years ago. I've been most of my career in Sweden but moved to Stanford now about one and a half year ago. I'm doing broadly studies within omics and molecular epidemiology but also have a translational part where I do [inaudible 00:03:29] and model systems.

Kiran Musunuru: That's great. To the subject at hand, so I think we all appreciate how, with the completion of the Human Genome Project about 15 years ago now, genetics has really taken off. What's interesting is, over the last few years, there's been a bit of a shift in focus from genetics to the layer of regulation that lies right above genetics and that's epigenetics, so modifications of DNA and the proteins that are bound to DNA and how this interacts with genetic expression and then has consequences in terms of clinical traits and diseases.

What caught my eye about your study is that you're actually looking at epigenetic regulation of gene expression but not in a very traditional, one locus at a time or one gene at a time fashion, but really in a genome-wide fashion. Whereas, starting in 2005, we started to see genome-wide association studies. Now we're starting to see, just over the last few years, epigenome-wide association studies.

Personally speaking, one of my research interests is lipid traits. I thought it was very nice how you were able to apply an epigenome-wide association study to lipid traits and actually find some very interesting things. Why don't I start by asking you simply describe the main goals of your study.

Erik Ingelsson: As you've already referred to, we wanted to look at variation in DNA methylation, which is one of the ways to look at epigenetics. I think either it's the most common way to look at epigenetics, at least if you want to do it genome-wide. We looked at variation in DNA methylation in relation to circulating lipid levels, and we did this through this epigenome-wide study and in whole blood derived DNA.

We did it with about 2,300 individuals from the Framingham Heart Study and from the PIVUS cohort, and then we had an independent external replication in about 2,000 additional individuals.

In addition to looking at these DNA methylation associations with lipids, we also wanted to look at these DNA methylation patterns in relation to incident coronary heart disease. We also wanted to integrate all of this with genetic variation, gene expression and also actually with metabolites through metabolomics. The whole idea here is trying to understand genomic regulatory mechanisms that link lipid measures to coronary heart disease risk.

Kiran Musunuru: That's one thing I really liked about this paper, how you really took it all on. It wasn't just one particular type of omics analysis. It started with epigenomics but then you really went the extra mile, I thought, to connect it to genetic variation, and then to disease, and to metabolomics and so it was very comprehensive that way. Why don't we discuss the actual findings. You actually found quite a bit in your analysis, didn't you?

Erik Ingelsson: Yeah. I think some of it were already actually in the title. We did, as I said, several different layers of things. The first thing was really to look at methylation patterns. We looked at CpG sites across the whole genome, and we identified almost 200 such sites that were different lipid levels in the discovery but then going to the replication stage, we had a little bit more than 30 of them being replicated and 25 of them had never been reported in relation to lipids before. That's one layer, so it is new associations. A lot of the genes that were then enriched they were involved in lipids and amino acid metabolism so it makes a lot of sense biologically.

There is the one example of an interesting finding there with ABCG1 that we perhaps can discuss a little bit later. Other larger things that we found was that there was a lot of cis-methylation quantitative triglycerides so that means that there were a lot of genetic variants that were associated with these methylation levels. In fact, actually, 64% of all of the CpG sites that we found, they also had genetic variance determining the level of the methylation. So quite large fraction being genetically determined. We also-

Kiran Musunuru: That's actually quite interesting because typically when you hear it in the lay press or what not about epigenetics, they tend to equate epigenetics with more environmental influences. It's a simple dichotomy or simplistic dichotomy of your genes are what you're born with but then epigenetics is the way that environment actually modifies your genetics in ways. But what you're suggesting from your findings is that it's actually genetic variation itself that could be directly responsible for epigenomic variation, which then would have effects on gene expression.

Erik Ingelsson: I agree. I think we're seeing a shift a little bit in this field. Again, my background is not really within that genetics field so I'm a little bit on the side here but what I see is that it's come more from an approach or focus really on inherited epigenetic changes so studies in animals, primarily, I guess a lot, but also in some human studies so more on that level to something that had been, as you mentioned, a lot of focus on environment causing methylation changes and now almost more into a focus of gene regulation and then gene expression and that focus.

Perhaps the ENCODE project and the Epigenome Roadmap and those projects have moved this field a little bit towards more focus on gene regulation and gene expression and that's kind of a part, a linking variation to gene expression. I think we're seeing a shift a little bit in that field.

Kiran Musunuru: That's very interesting. Can you give an example of a particular locus or particular gene where epigenetic regulation really seems to be playing an important role, not just with respect to lipids but even, perhaps, connecting to disease. I think you'd mentioned ABCG1 very briefly.

Erik Ingelsson: That's actually a pretty interesting locus. It's been recorded in the past, as well, in relation to methylation but we linked it all together. Basically, we see this intronic variant here where the minor allele is associated with increased methylation at the CpG site in that 5 prime UTR region of this gene of ABCG1 and then so that minor allele leads to increased methylation. It also leads to decreased expression of ABCG1 in blood. I think that makes sense. Quite often in the past, people have recorded that increased methylation should decrease expression.

As we see that, we also see an effect on triglyceride levels and HDL levels as well and, interestingly, also, on the risk of coronary heart disease. In addition, also, associations with several of the metabolites, so single myelins and[karomites 00:11:40] which have also been implicated in coronary heart disease in some prior studies. It all comes together quite nicely at this locus where you have a minor allele increasing methylation, decreasing expression, increasing triglyceride levels and increasing the risk of coronary heart disease along with increases in some of the metabolites that also have been linked to coronary heart disease.

Kiran Musunuru: Wow. Fascinating.

Erik Ingelsson: Yeah, I think it's pretty interesting, actually. We could link it all together in the study.

Kiran Musunuru: That's very nice. Another aspect of this study that caught my attention is that you really did it in a fairly rigorous way. You had your discovery cohorts in which you did the initial screen or the initial association study, but then you also had replication cohorts where you were then able to go independently test your findings and then accrue more evidence or lack of evidence for replication in the ones for which there was evidence of replication, those are, obviously, much more stronger results.

I expect that we have among our listeners trainees who might be interested in hearing more about how you were able to assemble so many different cohorts to be able to get this study done.

Erik Ingelsson: I think that's an important question. I would say that it goes back a little bit to the development that we've seen in genomics in the past 10 years. People coming in from gene studies to GWAS realizing that you really need to work together both because the science is better but also just if you want to establish any robust findings that can be replicated, you need to combine the data.

I think we've seen that for GWAS clearly, but I think we're starting to see that also for other [inaudible 00:13:28] approaches as we move forward. Because all of these approaches are prone to false-positives so if you just do your analysis in your own data, then you're more likely to report false-positives and you need replication.

I think we're lagging behind a little bit for epigenomics and other omics methods, but we're truly starting to see this happening also in other omics fields. I think, in a sense, the field is prime for collaboration and then I'm talking about the broad, molecular epidemiologist field or the people having cohorts and this kind of data, they're all used to working together from the GWAS era and also realize the need for it. I think for that reason it's usually not that difficult to get people together.

Then how do you do it practically? It's easier if you know people, of course, since before and that's probably more common nowadays than it would have been 15, 20 years ago because you always used to work with people in the GWAS era and you can even add a junior level set up these collaborations because you might have been involved in some other collaboration before and know some postdocs in some other labs, etc.

That might be one way to go about but the other thing is also that you have an interest in a certain phenotype and then you reached out to people that you think have the data. You can know about either from other publications and other phenotypes or on the same phenotype or just by word of mouth you know it since you've met people at conferences, you've seen some poster on the same phenotype, etc.

I would say that people, in general, are very open to collaborations, and I think we've seen that change and shift of the past 10 years. I think we see it now also for other omics methods, and I definitely do think that's the way forward. To report more robust findings, in general.

Kiran Musunuru: In closing, I'd say that seeing your study and seeing the very nice results, it seemed very promising with respect to what we're going to find going forward and doing epigenetic studies. Do you see more of this happening in the near future? Maybe even what happened with GWAS where it just got increasingly larger and larger studies and finding more and more results as these studies became increasingly powered.

Erik Ingelsson: Yeah, I think so. I think for epigenomics, as with some other omics, I think we will see the same development that we saw with GWAS, which is the people start to publish in relatively small settings with perhaps a few discovery cohorts, a few replication cohorts, and that parts happen kind of independently of each other. Then the next stage is you're grouping together and you're starting to involve other people as well and these consorts get larger and larger.

I think the value of this data can be exponentially increased if you can actually combine it with other data sets. We've seen that in genomics. There's a large return on your investments by collaborating with other people. I definitely do see the same kind of development happening here, as well.

Kiran Musunuru: Well, Erik, thank you so much. That's all the time we have for today but we greatly appreciate your taking the time out of your busy schedule to discuss with us this really nice paper that you and your colleagues published very recently. I would encourage all of our listeners to go take a look at the paper themselves. As I recall, this particular paper is open access so it should be freely available to anyone who is interested. Is that correct?

Erik Ingelsson: Yes, it's an open access. And thanks, Kiran. It was a pleasure.

Kiran Musunuru: Thank you very much.

Jane Ferguson: An AHA Science Advisory from the FGTB Council published in 2016 focused on the challenges and the potentials in merging electronic health data with genomics data to advance cardiovascular research. Jennifer Hall, John Ryan and colleagues published this on behalf of the Functional Genomics and Translational Biology Council as well as the councils on clinical cardiology, epidemiology and prevention, quality of care and outcomes research and the stroke council.

As electronic health records have become ubiquitous in medical practice, there is an opportunity to utilize existing stored data and add new types of data to the EHR to facilitate research through EHR-coupled biobanks and to improve patient care through the use of precision medicine approaches based on genomic and clinical data stored in a patient's record.

While logistical and ethical considerations remain, this is an area with great promise. You can read more in the Science Advisory published in the March 2016 issue of "Circulation: Cardiovascular Genetics", which along with all the papers mentioned in this episode, are linked on the Podcast website at fgtbcouncil.wordpress.com

This Podcast has the focus of precision medicine, and I saw an interesting back and forth in the JAMA comments section about the hype of precision medicine. I think even those of us who are fond of precision medicine would agree that there's probably a certain amount of hype surrounding it.

There was this interesting opinion published in JAMA last October addressing the question of, will precision medicines really have an impact on population health? I think there is some important points that really to improve population health, there may be other options rather than precision medicine, which may be more focused on the individual or on certain subgroups, which may not actually raise the broad population's health.

But then there was response to that published in JAMA in January, which was arguing against it. I thought it would be some interesting thing for us to talk about a little to see do we agree? Is this over-hyped? Or is precision medicine really something that could fundamentally change population and individual level of health in the future?

Naveen Pereira: I agree. There seems to be a tension between precision medicine that stresses on the individual and using omic technology and molecular markers to determine individualistic response or characteristics and population health in general, which looks at population trends. Both of them in principle and philosophy appear to be deferring fields. I guess the question is how do we integrate both of them to improve overall, not only individual but large population health?

Jane Ferguson: I think there's probably some disconnect maybe between what people think of as precision medicine and what sort of things it includes because I think our first thought could be that precision medicine is very much based in genetics and genetic risk scores, using genotype as a way to predict an individual's response to a drug or their risk of disease.

I think maybe one of the things we have to think about with precision medicine is to encompass all of these additional omic technology. So, yes, genotype alone is unlikely to really affect population health on a broad scale, but when you add in gene expression and proteomic biomarkers, metabolomics and microbiomes, I think then we do start to get to a point where it's mathematically complex but it would theoretically be possible to predict risk and implement precision medicine approaches, even on a large-population scale.

Naveen Pereira: Right. One of the things I've always wondered is should we move away from our traditional classification of disease? For example, hypertension. Is all hypertension the same? We know it's not, it's such a heterogeneous disease process. Are we still stuck in the 19th century where we think of hypertension as blood pressure? Should we move away from that? Should we integrate all this great input from omics technology and phenotype hypertension is a better disease process, which would, perhaps, improve outcomes.

Jane Ferguson: I think that's a great point. Honestly, probably a lot of the challenge in this is just us in thinking about things differently. You're right. We're very used to thinking of hypertension and we recognize it, we treat it. But it really is just ... The underlying causes of hypertension in the individual may be very different and it may need very different treatments.

I think a paradigm shift is probably needed in thinking about a lot of these complex diseases. Diabetes is another one where really that's the causes and then the way it progresses in different individuals is probably really distinct subtypes of disease rather than being one broad disease that we can classify as such.

Naveen Pereira: Exactly. And that would enable, perhaps, more dramatic treatment effects, too. I keep thinking of the example in cystic fibrosis where the genetic mutation in the cystic fibrosis gene actually proved that a certain therapy for cystic fibrosis in those patients who carry that gene mutation had a dramatic response. It didn't take tens of thousands of patients to demonstrate that effect but it took several hundred patients.

Jane Ferguson: That's a great point. I think if we're accurately substratifying individuals so that we really are looking at people who really do have the same underlying causes of disease, then I think we will have a lot more power to see effects in smaller numbers of people and we can move away from these huge GWAS of hundreds of thousands of people as being necessary to find effect.

Naveen Pereira: In fact, what we could do is take some of the knowledge from precision medicine and apply it at a population level and, hence, perhaps what we need to do is integrate the two disciplines better and people need to speak to each other more often. What do you think, Jane?

Jane Ferguson: Absolutely. I think that is key. We're used to thinking about our own little narrow field and focusing on that but I think integration and finding good ways for it. The humans to integrate and also to integrate the data mathematically, I think that will be key. I think that certainly caveats, I mean, these approaches may not find everything but I think there's definitely a lot of promise that has not yet been fully exploited.

Naveen Pereira: Absolutely.

Jane Ferguson: Last time we talked, we were talking about a paper that used gene expression profiling in CAD. I think you found a really interesting paper for us to talk about this month looking at gene expression profiling but in the setting of heart transplant and heart transplant rejection.

Naveen Pereira: Yes, Jane. It's interesting to see increasing number of publications now looking at gene expression arrays and profiling for various disease states. In the March 7, 2017, issue of "Circulation", there was a very interesting paper looking at gene expression profiling and complementing the diagnosis of antibody-mediated heart rejection.

Just as a background, the two types of heart rejection that heart transplant recipients can have, one, is cellular rejection which we're seeing now less often due to improvements in immunosuppression; the other type of rejection is antibody-mediated rejection most often caused by anti-HLA antibodies that are directed towards the donor or what we call as donor-specific antibodies.

This paper, the first doctor is Alexandre Loupy and he is from INSERM Institute in Paris, France and the senior author is Philip Halloran who is from Edmonton, Canada. What they essentially did was look at 617 heart transplant patients from four French transplant centers. Out of these 617 recipients, there were 55 recipients who had antibody-mediated rejection.

They did a case control study, the controls being 55 recipients who did not have antibody-mediated rejection. They analyzed 240 heart biopsies in total. Unfortunately, even in this modern era, we still perform heart biopsies traditionally through the internal jugular route and endomyocardial biopsies and these biopsies are then analyzed for features of antibody-mediated rejection.

The International Society of Heart and Lung Transplant has standard definitions by consensus as to what is antibody-mediated rejection and their various features histopathologically and by immunostaining. We also use donor-specific antibody detection in the serum to finally make a diagnosis.

What this group really did was analyze these heart biopsies by performing expression microarrays and they found a very distinctive pattern in patients who had antibody-mediated rejection by traditional criteria. The gold standard was the traditional criteria, and they used the gene expression pattern to correlate it with the gold standard.

They found certain selective gene sets that they call antibody-mediated rejection gene sets. It involved transcripts of natural killer cells, endothelial cell activation, macrophages and interferon gamma. The area under the curve that they found using these gene expression patterns for these four gene sets was greater or equal to 0.8 which is quite good. This gene expression pattern was then validated in a separate cohort of patients from Edmonton, Canada.

It's an interesting manuscript, which essentially looks at using gene expression profiling in addition to traditional histopathological determination for a relatively common type of rejection in heart transplant patients to consolidate the diagnosis and give insight into pathophysiology.

But some of the questions that arise are we still submit patients to endomyocardial biopsies so this does not supplant the need to perform endomyocardial biopsies because this was looking at expression arrays within heart tissue. We are still struggling with the gold standard, the histological diagnosis of antibody-mediated rejection as to what it really means in patients, for example, who do not have dysfunction of the graft, or a low ejection fraction. Useful in many ways. I think it adds to the overall knowledge of this phenomena, but it may not change clinical practice significantly.

Jane Ferguson: That's really interesting. It's exciting but, you're right, we are subjecting people to heart biopsies isn't necessarily going to be a good way to monitor rejection or be able to predict in advance who is going to suffer rejection versus not.

I think it's definitely a very interesting study and I think, the fact that they discovered these genes that which were then validated, may give some additional insight into the underlying biology, which may help us develop new ways to start thinking about treating this unmitigating rejection.

Naveen Pereira: Right and it would be interesting to see how this corresponds to peripheral blood gene expression and whether there's an early, noninvasive way of detecting rejection. I know the Stanford group in the past has looked at circulating DNA from the donor heart, analyzed by peripheral blood, the same thing that's done in efforts to its cancer detection to see if we can pick up rejection by just a blood draw instead of doing endomyocardial biopsies.

Jane Ferguson: Yes, definitely. I wonder if this group collected any blood or is this something they may want to do in the future because I think that would be a really interesting addition to this study.

Naveen Pereira: Absolutely.

Jane Ferguson: Well, it's been great talking to you as always, Naveen, and we want to say special thank you to Rick [Andraysen 00:31:10] for the Mayo Clinic Media Support Services for helping us with this Podcast.

Naveen Pereira: Always does a great job.

Jane Ferguson: Absolutely. We'll thank everybody for listening and we'll look forward to being back with you next month with more topics related to precision medicine and getting personal with omics of the heart.

Naveen Pereira: Lot of excitement next month, Jane. Thank you.