Getting old stinks; literally actually, at least in the EpiGenie office, where the pungent aroma of sports creme lingers the day after softball night. Dr. Steve Horvath (UCLA) published a doozy of a paper, DNA methylation age of human tissues and cell types, in Genome Biology recently that caught mainstream media attention from hacks like us to the Huffington Post. Dr. Horvath developed a multi-tissue predictor that can assign a DNA methylation “age” to tissues. Although we didn’t need a sophisticated predictor to verify that our connective tissue resembled that of a 90 year old, we were excited to hear more from Dr. Horvath himself on his findings.
After pouring through oodles of Illumina Infinium Array data; 82 data sets to be exact, Horvath uncovered a number of tight correlations between DNA methylation and age:
- “It is close to zero for embryonic and induced pluripotent stem cells”
- “It correlates with cell passage number “
- “It gives rise to a highly heritable measure of age acceleration”
When asked if he was surprised about his results, he shares that, “In total I analyzed over 120 data sets. Along the way I encountered many surprises. Overall, I continue to be amazed that one can construct a highly accurate epigenetic age predictor…” that allows you to forget the source of the DNA. He was also surprised that it “performs well in chimpanzees” too, since evolutionary conservation is never a bad confirmation.
“This is the first study that demonstrates that one can construct a highly accurate epigenetic predictor of age that works in most areas of the human body. It is the first biological clock that allows one to contrast the ages of different organs and cell types. Many researchers from different disciplines will be needed to carefully evaluate whether this epigenetic clock can be used to clarify why we age and what can be done against it.”
A Deeper Look into the Inner Works of the Epigenetic Clock
Horvath also shared his surprise “that age acceleration, defined as difference between DNA methylation age and chronological age, is inversely related to the number of somatic mutations in cancer tissue”, as he “would have expected to see the opposite.”
In terms of experimental design, he described three unique challenges he faced:
- “The DNA came from very different sources (adipose, bladder, blood, brain, colon, skin, etc).”
- “The data were measured on two different platforms: the Infinium 27K and the Infinium 450K array.”
- “The data were generated by many different investigators, which could lead to serious batch effects. The large sample size (several thousand arrays) allowed me to turn these challenges into strengths: the resulting age predictor is highly robust, performs well in most tissues and cell types, and is applicable to Illumina data measured on two platforms.”
“All of these results suggest to me that DNA methylation age relates to a process that maintains epigenomic stability. This is surprising given that many other articles suggested that age related changes in DNA methylation levels relate to random events, noise or “drift”. I think it is likely that the epigenetic clock relates to a biologically important process, which is why I proposed the epigenetic maintenance system model of DNA methylation age in the article.”
Given the coverage of his research by the popular press, we wanted to know if there is anything he felt was overlooked. Surprisingly he shared that, “Cancer is associated with both increased and decreased DNA methylation age. Tissue that looks much older/younger than expected may be malignant.”
An Environmental Interface?
“Since I suspect that life style factors (exercise, diet, smoking) have at best a small effect on DNA methylation levels in blood, hundreds if not thousands of samples will be needed to detect it” and pay the price of statistical power. In the spirit of Open Science, he also shared that, “Since many large cohort studies have already generated DNA methylation data or are in the process of generating these data, it will be straightforward to test these hypotheses using the software that can be found on my webpage.”
While “not sure about the role chemical exposures” have on the epigenetic clock, Horvath has begun to “suspect that age acceleration effects will mostly be tissue specific.” He concludes with “an important open research question…” about “…whether easily accessible sources of DNA (for example skin, blood, saliva) can be used as surrogate for inaccessible tissue (for example brain, heart, lung) when it comes to age acceleration effects.”
Take a deeper glance over at Genome Biology, November 2013