Ep16 Caitrin McDonagh

Getting Personal: Omics of the Heart

Jane Ferguson: Hi, everyone. Welcome to Getting Personal: Omics of the Heart. This is podcast episode 16 from May 2018. I'm Jane Ferguson from Vanderbilt University Medical Center and this podcast is brought to you by Circulation Genomic and Precision Medicine and the AHA Council on Genomic and Precision Medicine.

Jane Ferguson: This month we talked to Dr. Caitrin McDonough from the University of Florida. We briefly mentioned her paper in last month's episode Genetic Variants Influencing Plasma Renin Activity in Hypertensive Patients From the PEAR Study, but we wanted to go into it in more depth this month. Caitrin shared with us that this manuscript actually resulted from student course work and was a collaborative effort between students and instructors. The manuscript highlights has successful as approach can be both in increasing student engagement and as an effective way to do high quality research. You can hear her talk more about her innovative approach to student learning and the study findings later in this episode.

Jane Ferguson: Of course, we have a great lineup of papers in Circulation Genomic and Precision Medicine this month. First up, a paper entitled, "SCN5A Variant Functional Perturbation and Clinical Presentation Variants of a Certain Significance" by Brett Kroncke, Andrew Glazer, and Dan M. Roden and colleagues from Vanderbilt University Medical Center. They were interested in investigating the functional significance of variants in the cardiac sodium channel in particular to see if they could explain why some variant carriers present with cardiac arrhythmias while others remain asymptomatic. Through a comprehensive literature search, they identified 1712 SCN5A variants and characterized the carriers by disease presentation. Variants associated with disease were more likely to fall in transmembrane domains consistent with the importance of these domains for channel function.

Jane Ferguson: Using American College of Medical Genetics Criteria for variant classification, they found that variants classified as more pathogenic were also more penetrant. Penetrance was also associated with electrophysiological parameters. This approach highlights how modeling the penetrance of different variants can help define disease risk for individuals who carry potentially pathogenic variants.

Jane Ferguson: Next we have a paper from Vincenzo Macri, Jennifer Brody, Patrick Ellinor, Nona Sotoodehnia and colleagues from the University of Washington and Massachusets General Hospital. This is also related to sodium channels and the paper is entitled, "Common Coding Variants in SCN10A Are Associated With the Nav1.8 Late Current and Cardiac Conduction". They were interested in SCN10A and sequenced this gene in over 3600 individuals from the CHARGE consortium to identify variants associated with cardiac conduction. They were able to replicate associations between variants and PR and the QRS intervals in a sample of almost 21,000 individuals from the CHARGE Exome sample. They identified several missense variants have clustered into distinct haplotypes and they showed that these haplotypes were associated with late sodium current.

Jane Ferguson: Continuing the cardiac conduction theme, Honghuang Lin, Aaron Isaacs and colleagues published a manuscript entitled, "Common and Rare Coding Genetic Variation Underlying the Electrocardiographic PR Interval". They conducted a meta-analyses of PR interval in over 93000 individuals which included over 9000 individuals of African ancestry. They identified 31 loci, 11 of which have not been reported before. We see SCN5A come up again as a gene of interest in this study but their analyses also implicated a novel locus, MYH6.

Jane Ferguson: Next up moving from the heart to the vasculature, Janne Pott, Markus Scholz and colleagues from the University of Leipzig published a manuscript entitled, "Genetic Regulation of PCSK9 Plasma Levels and Its Impact on Atherosclerotic Vascular Disease Phenotypes". They were interested in whether circulating PCSK9 can be used as a diagnostic or predictive biomarker. To address this, they conducted a GWAS of plasma PCSK9 in over 3000 individuals from the LIFE-Heart study. They found that several independent variants within the PCSK9 gene were associated with plasma PCSK9 as well as some suggestive variants in another gene locus FBXL18. They used Mendelian randomization to probe causality and the data suggest that PCSK9 variants have a causal role in the presence and severity of atherosclerosis.

Jane Ferguson: Moving on to another biomarker, Lisanne Blauw, Ko Willems van Dijk and colleagues from the Einthoven Laboratory for Experimental Vascular Medicine report on CETP in their manuscript Cholesteryl Ester Transfer Protein Concentration A Genome-Wide Association Study Followed by Mendelian Randomization on Coronary Artery Disease. They aimed to assess potential causal effects of circulating CDP on cardiovascular disease through GWAS and Mendelian randomization.

Jane Ferguson: In a study of over 4000 individuals from the Netherlands Epidemiology of Obesity Study, they identified three variants in CTP that associated with plasma levels of CETP and explained over 16% in the total variation in CDP levels. Genetically predicted in CETP was associated with reduced HDL and LDL cholesterol suggesting that CETP may be causally associated with coronary disease.

Jane Ferguson: Rounding out the table of contents we also have a clinical case perspective from Nosheen Reza, Anjali The Importance of Timely Genetic Evaluation in family members in cases of cardiac disfunction and cardiomyopathy. We have a report from Adrianna Vlachos, Jeffrey Lipton and colleagues on the Diamond Blackfan Anemia Registry and we have a clinical case from Yukihiro Saito, Hiroshi Ito and colleagues on TRP and poor mutations in patients with ventricular non-compaction and cardiac conduction disease.

Jane Ferguson: To read all of these papers and the accompanying commentaries, log on to circgenetics.aha.journals.org and if you're a visual learner or you need a work related excuse to spend time on YouTube, you can also access video summaries of all our articles from the CircGen website or directly from our YouTube channel Circulation Journal. Lastly, follow us on Twitter at circ_gen or on Facebook to get new content directly in your feed.

Jane Ferguson: I'm joined today by Caitrin McDonough from the University of Florida and Caitrin is an Assistant Professor in the Department of Pharmacotherapy and Translational research in the College of Pharmacy and she's the first author on a recently published manuscript entitled, "Genetic Variants Influencing Plasma Renin Activity in Hypertensive Patients From the PEAR Study". This was published in the April 2018 issue of Circulation Genomic and Precision Medicine. Welcome, Caitrin.

Caitrin M.: Thank you.

Jane Ferguson: For listeners who haven't had a chance to read the paper yet, I wonder could you give us a brief overview of what prompted you to do this study?

Caitrin M.: Sure so this looks at plasma renin activity and just initially a GWAS but it was done in a hypertensive population from the pharmogenomic evaluation of antihypertensive responses study. Particularly, since our group here at the University of Florida is more interested in pharmacogenomics we wanted to address plasma renin since it can influence blood pressure response to antihypertensive medication particularly if you use it as something to predict but also to correlate it with that as there have been also prior data from our group that shows if you have different levels of plasma renin that would predict if you would respond better to certain types of antihypertensive medications such as a beta-blocker or a diuretic.

Caitrin M.: We used both a GWAS approach as well as a prioritization through blood pressure response to focus in on signals and then furthered by using prioritization using data from RNA seq and looking at eQTLs and then finally looking at more of just a traditional net replication of the original plasma renin activity signal.

Caitrin M.: Overall, one of the interesting things and why we were initially doing this study was really in connection with a graduate course that myself and another faculty member here who's also an author on the paper, Yan Gong [inaudible 00:09:12]. We often have the students analyze data from the PEAR study as we have a lot of data from that study and it helped us analyze additional papers but we didn't necessarily know if this was going to be an interesting phenotype but through that course work which turned out that it really did have some interesting signals so we wanted to follow up more on.

Jane Ferguson: Yeah, I love that approach so I think that's a really smart way to do it. To actually get your students to analyze your data and get them really involved in the process. How much then did the students ... how much were then they able to get involved when it started transpiring that their results would actually be something that could be put together for a manuscript?

Caitrin M.: Overall, they are fairly involved. During the course work, what we usually do is give them just directly types data since a lot of them have not done this type of genetic analysis before and we split it up where each student gets about four to five chromosomes of data and then different phenotypes in the different race groups as we have both whites and African Americans. They get a certain race group, certain number of chromosomes and so they're able to conduct the analysis just using the Uplink software which is fairly user-friendly and straightforward. Then they get experience making Manhattan plots and using LocusZoom.

Caitrin M.: After they have the basic techniques, then we teach them how to start following up top signals and determine what is a good signal. They're looking at the LD or SNP function or possibly gene function or looking at their genotype, phenotype relationships and making sure that it's not just one person who's driving the whole signal. Then selecting what top reasons and top SNPs may need a follow up. That part they all do there in the class and learn more of the basics.

Caitrin M.: After the class, the students who want to continue to participate we get together and redistribute data where they would then move on to working on the imputed data sets and we teach them how to do that. Then we give them ... operate it somewhat similar to a consortia level meta-analysis type thing. I write up an analysis plan, each student does some part of the analysis. They have to bring it all back to me. I sort through it. We meet and go through it. Then we set our next steps to follow up. Then different students get different SNPs to investigate the function of or different subanalyses to do.

Caitrin M.: One of our graduate students who is on this particular project, her dissertation project was very focused on our RNA seq data so that was how we were able to bring in the eQTL analysis using the RNA seq data as she had done a lot of the groundwork with that already. In one of our discussions that was one of the ways that we were able to incorporate the prioritization since she was intimately familiar with that data set.

Jane Ferguson: Yeah and I think that's great. I can imagine that, that's a much more compelling way for students to learn about how to analyze data when they see the natural follow through. Do you find that some of the students maybe get more excited about research or are more likely to pursue future research opportunities by having had this hands on experience with the publication process and completing a project really did to this very end?

Caitrin M.: They do, yeah. I see some of the students that end up sticking with it more are the students who I work more closely with and see more closely some of the students who are from other departments still stay involved but sometimes don't stay quite as involved. But, all of them really do continue to follow up and ask if they can still help or if there's anything they need to do until we get it to publication which is really nice.

Jane Ferguson: Yeah, right. I think that's fantastic and I'm sure every study has its challenges. I'm interested what were the challenges you encountered in doing this study and which one of them may be unique to the way you have a lot of different people analyzing different aspects of the data versus the regular challenges that would come up in a study like this.

Caitrin M.: Yeah so some of it I think is just keeping everyone on track and keeping it organized, making sure I think some of our challenges with this study was just me making I think on a lot of other studies, while I had certainly hands on the data it was more of an oversight rule for some pieces of it and just making sure everything looked the way that I thought it did, double checking. Some of it I think the teaching aspect of it just making sure everything was also done correctly and then keeping everything organized made the study a little bit more challenging.

Caitrin M.: I think part of it too was with the PEAR study, it is a very rich data set. Determining what we wanted for our prioritization scheme and how to work through the different types of data sets that we had and put it all together as initially we just assign each student a different piece and we had a vague plan but it was a little bit more tricky as to work through how it was all coming together then when everyone came back together since a lot of people were doing as opposed to just one person doing it.

Jane Ferguson: Right, so yeah and I think you're touching on the part that all of us have when we're writing papers that you sometimes end up with a lot of data at the beginning, you're trying to sift through it and then sometimes at a certain point you see something and you're like, "Okay, yes. This is interesting." Then you start following it up.

Jane Ferguson: I wonder at what point did that happen? I suppose you probably ... You ran the GWAS for plasma renin activity and then find a number of suggested SNPs that were significant you associated but then ... Describe your strategy and you did so the second screening stuff to look at the pharmacological aspect defining [crosstalk 00:15:12]?

Caitrin M.: Yeah, our initial plan going in was the first two steps, to do the GWAS for plasma renin activity and then to do the prioritization through blood pressure responses. I was very familiar with what our lab was familiar with but then after we got there, I think we were then troubled with what we did next and where to go. When we decided to bring in the RNA seq data, I think that was when it really started coming together as our top signal, the SNN-TXNDC11 gene region really stuck out then and it showed up. That seemed like a much stronger signal and it gave us a little bit more focus and also brought it much more of a functional aspect where we would maybe start to believe that signal more. That I think was really when we did that more of a turning point for the study and helped us focus more on where then to go with the results.

Jane Ferguson: As far as the data you had I think over 700 people for your GWAS. Then you had a pretty large number of ... Was it the same subjects or different subjects where you also had the RNA seq data to do the QTL analysis?

Caitrin M.: The same subject so not everyone has RNA seq. We have RNA seq data on 50 individuals and they were selected from whites and at the extremes of the blood pressure response so that it has a slightly interesting selection process. It's the main data analysis there was a best responder, worse responder to thiazide diuretics.

Caitrin M.: When we do the eQTL analysis, we aren't always sure what we're going to get since we're missing the middle of blood pressure response. But, when we're just looking strictly at eQTL analysis sometimes we get lucky and sometimes it looks weird.

Jane Ferguson: In your case as well you had the added issue of subjects were randomized to drug treatment so it was some where responders were ... I guess some people got the drug that worked for them and some people did not get the drug that worked for them.

Caitrin M.: Yeah.

Jane Ferguson: Did you I guess were incorporating both groups so good responders to either and some of that was because of their gene. They got the right drug for their genotype.

Caitrin M.: Exactly, yeah.

Jane Ferguson: It's good and then you were able to replicate this. After you were able to prioritize your gene region based on the GWAS and the drug response and the eQTL data, you actually ended up being able to go to a second sample to replicate the association right?

Caitrin M.: Yes, it was in a lot of the same investigators, we have a second study PEAR-2, which has a very similar design to the PEAR study but used different drugs but also collected baseline plasma renin activity. We were able to use that phenotype again. We did have slight differences in GWAS to imputation panels at that point in time for when we were conducting this study so we ended up using a proxy to replicate but we did see the same signal in the second population which was very nice to see.

Jane Ferguson: What is known about this gene region or either of these two genes?

Caitrin M.: Overall, that was I think one of our harder points when we started trying to make the connections back to our phenotype. This was one of the areas where we did also have help from our students and the students as that was part of their initial training where they really looked to see what function was of various different genes and how to follow them up. That was one area where they came in, was to help look up some of the function of these ... there have been some connections with the various genes, the other phenotypes and with SNN and to atherosclerosis and other inflammatory cytokines such as TNF-alpha. Then there have been data also from [inaudible 00:19:37] that really show that there is an eQTL in this region that which supports what we saw in our own data. However, there really wasn't any direct connections with renin and the renin angiotensin aldosterone system and blood pressure regulation that we could find in the literature.

Caitrin M.: We're not exactly sure how it connects but based off of our functional data and levels of evidence and then we saw some of that in publicly available data, we're still very interested in the region.

Jane Ferguson: I think the data is compelling enough that it looks like you've identified the new region that probably is the mechanistically related that will require a whole bunch of basic mechanistic research to figure out what exactly the genes in this region are doing and how this ultimately connect back to blood pressure and response to drug therapy.

Caitrin M.: Exactly.

Jane Ferguson: I could see this ... I mean obviously there's a whole lot of potential functional work there and then probably also the clinical work, I wonder what you think about how this would affect any pharmacogenetic therapeutic ... You know at present I think you can look at plasma renin activity and use that as a predictor of drug response to help guide therapies. Would you think that a genotype guided therapy may end up being more effective than the plasma renin activity measurement in this case?

Caitrin M.: In this case since this is so connected with a phenotype that you could use with plasma renin, I think if you're able to draw a plasma renin you may just want to do that. I think our overall goal would be if someone had preexisting genetic data and you weren't wanting to do an additional test or if you're contemplating response to a lot of different drugs that perhaps you could use a genetic data. One of the issues that was brought up on review and that are a lot of group considers quite a lot is that we have a lot of signals and that our group has certainly published a lot in this area and there's a lot of signals that we have to a lot of different drugs and how do you incorporate all of them together, is there overlap between them or where do they all fall?

Caitrin M.: That is certainly something that we're still working on as more I think ultimate goal would be more to delve more of a SNP score or gene score and some type of risk score that would help you determine what drug you would best respond to. We've done that a little bit in some of our prior publications but we haven't yet taken all of our data together and help to build something that would if you had a lot of data on an individual and various different alleles at various different genes, how that would respond.

Caitrin M.: Overall, when we look at blood pressure response as a pharmacogenetic signal, certainly we see larger affect sizes than you would in disease genetics but we're not seeing affect sizes like you would with more of an adverse drug event. We're in between there and we're often times it's not necessarily just going to be one SNP or one gene that would tell your whole story but a combination of quite a few of them.

Jane Ferguson: I wonder are there more similar stories like this from the same data set? You know you've been through this process from start to finish and building in the functional work and do you think that next year's class will be able to do this again with the same data set? That maybe pick one of the next priority candidate down the list and maybe find another interesting story like this?

Caitrin M.: Yeah, so we actually just finished our class this year and they looked at potassium. We just got done grading final papers and submitting grades so we will over the summer be working with them a little bit more. I think some of our new graduate students too are starting to work on trying to make more connections between a lot of our different phenotypes and as you start to layer those together what it exactly means for a patient or implementation perspective.

Jane Ferguson: Yeah, interesting. We'll have to look for that story whenever you guys get done with it. Otherwise, are you planning on following up this specific SNP region in any other way or any other studies? Where's next for you guys overall?

Caitrin M.: I think one of the things we would like to do is look at this more in PEAR-2. We really just brought the PEAR-2 data set in here as replication of the top region in that last stage but we have that data set and we can certainly look at that data set.

Caitrin M.: The other thing that I would like to do is as we started this project in conjunction with the class that was a couple of years ago at this point in time, we used [TAP/MAP 3 00:24:35] imputed data since that was what we had in the lab and what we were using at that point in time. At this point in time, we have now imputed both PEAR and PEAR-2 2000 genomes phase three data. It'd be interesting to see if we are able to see any additional signals or if these regions become stronger or exactly what would happen using a more imputation panel that has more coverage and where we would have the same panel between both PEAR and PEAR-2.

Jane Ferguson: Right because you may or may not have identified the causal SNPs in the previous access but-

Caitrin M.: Yeah.

Jane Ferguson: -yeah so it'd be nice to see if you can actually get that out. That potentially could end being a drugable target maybe suitable for something more specific but who knows. Is there anything else that we haven't covered yet that you'd like to mention?

Caitrin M.: Overall, I think that just this type of model of utilizing more of a real world analysis and data in a class project really certainly engages our students a lot and I think they all enjoy actually being able to work with data that came out of this study and have a lot more hands-on experience and really project-based analysis experience. We've been very happy with this model and have used it multiple times. We have an HDL paper, the renin paper, our glucose response paper and now we're working on the potassium project. It's been a good model for us here with our pharmacogenomics class.

Jane Ferguson: Yeah, I mean I think it's a really smart and intuitive way to think about education. It's mutually beneficial it sounds like, so it's helping you guys get your data analyzed. It's really helping the students learn so I think it's a win-win situation. I think it's a model that a lot of other people would really be interested in adopting.

Caitrin M.: Yeah.

Jane Ferguson: Okay well thanks so much for talking to me and talking about your model and your research. It's been great.

Caitrin M.: Yes, thank you very much for having me.

Jane Ferguson: That's it from us for May. Thank you for listening and come back for more next month.