For humans, conception involves a fateful collision of two cells, one egg from the mother and one sperm from the father. After ﬁve days, the two cells create a primordial package called a blastocyst. This wall-like structure, as wide as a hair, protects a moist ball of a few dozen cells. One day later, the blastocyst, ponderous with potential, attaches to the uterine wall.
This journey of primary cells in the uterine environment and the complex ecology of fetal development captivate Karin B. Michels, an associate professor of obstetrics, gynecology, and reproductive biology at Harvard Medical School. She also studies how environmental factors, from maternal smoking and folic acid supplementation and fortiﬁcation to plastic bottles and soup can liners, affect human health, in utero and otherwise.
In June 2012, Michels directed a Radcliffe Exploratory Seminar on ways to analyze a new kind of data piling up in genetics labs all over the world. The result was a paper in the prestigious journal Nature Methods, published in December 2013 as “Recommendations for the Design and Analysis of Epigenome-Wide Association Studies.”
Don’t know much about biology or the epigenome? To explain, Michels goes back to the blastocyst. At ﬁrst, fast-forming fetal cells are undifferentiated, she says. Then differences emerge. What biologists call the genome—the unchanging inherited genes in every cell—becomes the phenome: the differentiated cells that early on decide to become skin, hair, blood, or organ tissues.
How they decide to be different is what Michels and her team study. “Every tissue has different function—kidney, liver, brain, skin,” she says, “yet all have the same genetic information. So there is something between the genome and the phenome. That’s epigenetics.”
“Epi” means “over,” she points out. The epigenome is atop the DNA—“over” the double helix where genes reside. It tells each cell what its function is by chemically directing the cell to switch a gene on or off in order to create a certain tissue type. Michels calls the epigenome “a complex switching mechanism.”
If genes are an array of switches ready to be coded for functionality, the epigenome is the blueprint, covering each strand of DNA like a template. But that blueprint requires an epigenetic agent—a ﬁnger, as it were, that can ﬂip the genes like switches, setting the codes for hair, blood, and other functional differentiations.
There are three of these switch-tripping “epigenetic processes,” says Michels. In the most thoroughly studied, DNA methylation, CH3 groups (composed of one carbon atom bonded to three hydrogen atoms) attach to a strand of DNA and guide protein formation by shutting off genes. Methyl groups are attractive to researchers because they remain stable in tissue samples stored for a long time in freezers. (Epidemiology labs everywhere have frozen samples, some going back decades.)
Epigenetics is an important key to turning genes off and on. But until a decade or so ago it was the province of only plant and animal studies. Today, human epigenetics is a hot ﬁeld—growing exponentially, says Michels—because it holds preventive and therapeutic promise for the future.
The genome, in humans and others, is a relatively stable archive of genes that “essentially never changes,” she says. “But the epigenome is a dynamic signature. ” It can change as the result of the environment, ﬂexing and mutating—a “plasticity” that, Michels says, may be a buffer against environmental stressors. However, too many dynamic changes in the epigenome may make you more vulnerable to disease.
Scientists—in big teams, and with big money behind them—are pondering a future of prevention and therapies. But those efforts involve identifying sites in the epigenome that are associated with disease. That relates to the puzzle Michels and others are busy trying to solve, using—in large part—just partial scans of methylation sites along the epigenome. Which have functional relevance, they ask, and which are just statistical noise?
Telling the difference means being able to sort and analyze all the epigenomic data that have piled up. The Radcliffe seminar was an attempt to set standards of epigenetic analysis and study design. “We were at the point where we had all these data and everybody took a different approach to tease out the important signals,” says Michels.
In 2012 there was no uniﬁed approach. Michels once sent data to three experts and got back three different interpretations of the important epigenetic sites. “I thought: This can’t be,” she says. “We need more standardization.”
Michels was just ﬁnishing a book on how to marry two disciplines—that of epigeneticists, who might study 5 or 10 people at a time, and that of epidemiologists (like herself), who study whole populations. (“I always joke about my epi-epi interests,” she says.) The marriage involves mixing the challenges of scale (large populations) with the challenges of depth: epigenomes that are far more complex and dynamic than genomes.
The 2012 seminar brought together some of the world’s leading experts in analyzing epigenomic data: statisticians, epidemiologists, biologists, and physicians at the top of their game, as well as sharp young researchers with something to add (another Radcliffe seminar marker). “They did all the organizing,” Michels says of Radcliffe. “I just had to write my wish list. How much better does it get?”
On her seminar wish list was the prominent Johns Hopkins University biostatistician Rafael A. Irizarry. At Harvard now, he credits the Radcliffe experience with changing the way he analyzed DNA methylation data, “which in turn inﬂuenced the way my collaborators at Hopkins analyzed data,” he says.
The two intensive days of discussion followed the Radcliffe seminar model: bring together many disciplines to puzzle over one problem. In this case, the problem was “epigenome-wide association studies” (EWAS)—broad investigations into how the epigenome is affected by the environment and other factors, and how this can result in disease. They are modeled on GWAS—similarly broad studies in the world of genomics.
In 2009 there were fewer than 20 EWAS; by 2012 there were more than 80; today there are hundreds. The seminar was “a step towards unifying methods to identify important sites along the epigenome,” Michels says.
To continue the momentum of the seminar, Michels set up a Google document so that the article could be written by the 14 participants simultaneously. She assigned each a writing task that matched his or her expertise. After six months, it was ﬁnished, and in another six, the Nature Methods study was revised and published. “Given how fast this ﬁeld moves,” says Michels, “it was important to get our recommendations out.”
By early December, the paper had taken on a life of its own as a landmark that is changing the way EWAS are conducted. Michels was nearing the end of a round of fall conferences in London, Milan, and Stockholm, where the paper was widely circulated and praised. In London the Gates Foundation asked her to advise on epigenetic research. “I am glad we were able to provide recommendations that the epigenetic community is ﬁnding helpful,” she says.
All the attention was “more mileage from the Radcliffe seminar,” says Michels—without which “I never would have written this paper or met these fantastic people with whom I now collaborate.”
Corydon Ireland is a staff writer for the Harvard Gazette.