I Want To...
I Want To...
Find Research Faculty
Enter the last name, specialty or keyword for your search below.
School of Medicine
I Want to...
Andrew Feinberg MD, MPH
For more than a quarter of a century, you’ve been trying to understand how epigenetics is related to human disease: You did the first experiments on the epigenetics of human cancer in the early 1980s with your mentor, Bert Vogelstein, and for last seven years or so, you’ve been investigating the epigenetics of other common and complex diseases. What’s new?
FEINBERG: A couple months ago, I started working in the lab again, doing experiments; that’s how I got started and it’s what I like to do the most. I have a good administrator doing the business part of the job, and that frees up time for me.
What is it that you like best about working in the lab?
FEINBERG: It’s much more cheerful in the lab than in my office — there are neat people and great equipment. The most fun in science is generating data and getting to see for the first time how it’s looking and what it shows. It’s an incredible thrill — to get a look at things nobody has ever seen before. With post-docs and students, you have this thrill vicariously, and you can do so much more working with a team of people. But it’s sure a lot of fun to do it yourself.
What specifically has occupied your time in the lab?
FEINBERG: Just recently, I was in CHARM school. Statistician Rafael Irizarry and I invented an analysis method about two years ago called CHARM, an acronym alluding to Baltimore as “Charm City” and standing for Comprehensive High-Throughput Relative Methylation. It’s odd to have invented a method and never have done it. It bothered me a lot because there’s so much work we do that depends on this procedure. I understood the methods, of course, but that’s different from having done them myself.
What does CHARM do? What do you use it for?
FEINBERG: It’s a relatively inexpensive tool using molecular and statistical procedures, to do high-throughput array-based DNA methylation analysis. It is suitable for genome-wide analysis, allowing individual samples to be assayed reliably at very high array (sequence) density, allowing locus-level genome-wide epigenetic discrimination of individuals, not just groups of samples. Unlike some other approaches, CHARM is highly quantitative, a substantial advantage in application to the study of human disease.
Why is doing it yourself so important?
FEINBERG: Such a large part of what we do in research now is statistical analysis. While some other molecular biology labs might view statisticians as servants of the data, I see them as full partners in what we discover. When I started using the software and looking at the different algorithms that statisticians have written for normalizing data and clustering data, I could see limitations in how that’s done; I saw relative pitfalls that wouldn’t have been obvious had I not been doing it myself.
You’ve weathered criticism over the years from naysayers who contend that it’s too difficult to study the epigenetics of disease in human populations because there’s too much heterogeneity and so many epigenomes. How do you deal with that?
FEINBERG: One thing that’s changed in the last half-decade came as a result of our Center of Excellence in Genome Science grant from the NIH. CEGS has brought together world-class epidemiologists like Dani Fallin of the School for Public Health, disease-based biologists like Jimmy Potash in the Department of Psychiatry, statisticians like Raphael Irizarry, and molecular biologists like me. Together we’re coming up with new ways of comprehensively looking at how to collect data, interpret it, and relate it to human populations and phenotype, as well as how to integrate that into existing samples. This partnership effort has broadened my view. I came from a cancer focus and now it has broadened to include neuropsychiatric and other diseases. For instance, we have a big project starting that looks at the epigenetics of schizophrenia. The questions, methods and approaches we’re developing can be applied generally, however. The CEGS has led to a broader way of thinking about epigenetics and human development.
You published a paper in PNAS amending Darwin’s view of evolution, suggesting that genes that don’t themselves directly affect the inherited characteristics of an organism but leave them increasingly open to variation may be a significant driving force of evolution. What’s the origin of this interest?
FEINBERG: I was asked to speak at a meeting of the National Academy of Sciences about evolution and medicine, so I was thinking hard about how epigenetics fit into evolutionary biology and how that is relevant to disease. I realized you can’t really understand human development or disease unless you look at it from an evolutionary perspective, because what we have is what’s been selected for over many generations, and you have to explain things that way.
That realization culminated in a bona fide Aha! moment. Would you please recall it for us?
FEINBERG: I was visiting London with my son, Jason, who was teaching in Spain at the time. It was February 14, 2009, and I was standing in Westminster Abbey on Darwin’s grave, looking across at Newton’s grave, and next to that is the grave of Dirac, a founder of quantum mechanics whose contribution involves the idea that things are indeterminate. I realized this is something that’s been missing from Darwinian thinking; not that Darwin’s wrong, but the 19th-century watch-glass-like regularity to things is a problem when thinking about epigenetics. The conventional wisdom in epigenetic models of evolution is you have some sort of environmental change that influences your genome and that influences your phenotype. If it’s important in human disease, it needs to be selected for in evolution. How would that happen? How would you get an environmentally induced change in your genome that would somehow change it and be inherited stably for a hundred generations, which is what it takes for natural selection to work? It’s not going to happen. It’s been a paradox.
Standing there in Westminster Abbey, it occurred to me that maybe what really happens is you have a genetically determined stochastic variation: It was like getting hit by a thunderbolt.
Could you explain the gist of the idea?
FEINBERG: There are going to be some genetic variants that will make you prone to epigenetic variation — which could be completely random or environmentally influenced, but probably a mixture — and if you have those genetic variants that make you prone to this epigenetic variation, you’re going to have a greater range of phenotype with the same genetic background. In a fixed environment, that’s going to be a disadvantage, but in the highly fluctuating environment that we in fact live in, that’s an enormous advantage. What it means is the phenotype is somewhat unpredictable, and yet, as the environment changes — say it’s been selecting for a particular given phenotype and now it’s selecting against — those individuals who have that genetically determined propensity toward greater epigenetic variation, those carrying that genetic determinant for this variation, are going to be at a selective advantage over those hundreds of generations.
What happened after the thunderbolt?
FEINBERG: My son and I walked across town and I wrote it all down at a pub near the Tower of London. When I returned to Hopkins, Rafael Irizarry and I tested the idea on data we already had generated from mice that were from exactly the same genetic background, from the same litter, with identical genotypes, and living in the same cage, eating the same food. There are certain places throughout the genome where there are huge DNA methylation variations. We call them variably methylated regions, or VMRs. The VMRs occur at really key loci for pattern formation, limb development, neural system development. It’s a counterintuitive and stunning thing because you would not expect there to be that kind of variation in these very important patterning genes. But when you think about it, it makes sense. We have to go through a developmental program and we’re exposed to lots of different potential environmental factors, and there needs to be flexibility for change at certain places in the genome. VMRs provide that flexibility by allowing for stochastic variation in gene expression. Most importantly this would lead to variation in offspring that would not otherwise occur. That very flexibility can also be a weak point, an Achilles heel for disease if we adapt to the external environment changes in an unhealthy way.
FEINBERG: Now I’m very busy testing this idea on a lot of common complex diseases. I want to emphasize in the strongest possible way that I could be completely wrong about this evolutionary model. That’s true for any idea that you have. You have to test it.
- Paternal Sperm May Hold Clues to Autism
- Researchers Identify New Genetic And Epigenetic Contributors To Diabetes
- New Tool Pinpoints Genetic Sources Of Disease
- Genes And Their Regulatory 'Tags' Conspire To Promote Rheumatoid Arthritis
- How Bees Decide What To Be
- Scientists Map Epigenetic Changes During Blood Cell Differentiation
- Scientists Suggest Framework For Epigenetics In Common Disease
- Manipulation of Epigenome Turns Off As Many Genes As It Turns On
- Epigenetics’ new findings on gene regulation spark new studies
- It's Not All Genetic: Common Epigenetic Problem Doubles Cancer Risk In Mice
- "Epigenetic" Changes Come Before Mutations
- Mendel Didn't Have The Whole Picture: Our Genome Changes Over Lifetime, Johns Hopkins Experts Say
- Deciphering an elephant