Q&A With Art Petronis. Questions by Hakon Heimer and Victoria Wilcox.
Q: To start off, could you summarize the most important findings and your conclusions from them?
A: The most important finding is that, although we still don’t have the final answer, we cannot exclude a possibility that epigenetics may be a secondary molecular substrate of heritability. Most, if not all, human diseases, including all psychiatric diseases, exhibit evidence for inherited predisposition. The most elegant way to estimate heritability is by comparison of concordance of monozygotic (MZ) and dizygotic (DZ) twins. In schizophrenia, the numbers are as follows: MZ twins exhibit 50 percent concordance, while DZ twins exhibit 15 percent concordance for schizophrenia, and the heritability is 70 percent. The traditional understanding is that DNA sequence is the molecular basis of heritability.
In our recent twin study, we compared the degree of epigenetic differences between MZ twins to the degree of epigenetic differences between DZ twins. We detected that MZ twins, despite their numerous epigenetic differences, are still more similar to each other in comparison to DZ twins. There could be several explanations for this. The null hypothesis is that the larger epigenetic differences in DZ twins are secondary; i.e., they are induced by DNA sequence differences in DZ twins. The alternative hypothesis is that those large epigenetic differences that we see in DZ twins represent vestiges of the inherited epigenetic profiles. MZ twins originate from a single zygote and, therefore, their epigenetic origin is the same (or very similar). DZ twins originate from two separate zygotes, the epigenetic profiles of which can be very different. Before continuing, it is necessary to explain the concept of epigenetic variation of the zygotes. In our earlier studies, we investigated the epigenetics of the germline, more specifically, sperm, and detected significant epigenetic differences across sperm cells from the same individual. Even short (about 500 nucleotide) stretches of DNA exhibited enough epigenetic variation to make each germ cell unique from the epigenetic point of view. We don’t know if oocytes exhibit such a significant degree of epigenetic variation, but it is likely to be similar to what we detected in the sperm. If germ cells are different epigenetically, zygotes will be different as well.
We performed a series of additional experiments looking at the possible effects of DNA sequence on DNA methylation and did not detect much evidence that the larger epigenetic differences in DZ twins are secondary. These findings favor—and I want to emphasize that they do not prove, but only favor—the alternative hypothesis that some of the epigenetic peculiarities in the zygotes are not completely erased; they survive all those numerous mitotic divisions experienced by an organism during development. If some of the inherited epigenetic factors play a role in phenotypic outcomes, we have a case of epigenetic heritability.
The ramifications of this finding may be quite interesting. Traditional genetic studies in complex diseases thus far could explain only a small fraction of heritability. Maybe a part of that heritability that we detect in phenomenological twin studies is actually due to epigenetic mechanisms.
Q: What questions remain about the methods and data that you want to follow up on?
Overall, this was a significant effort that required over 700 microarray experiments to test nearly 100 sets of MZ and DZ twins, but it was only a pilot study. First, the twin sample—when it is split into tissue categories (white blood cells, buccal epithelial cells, and gut cells)—brings only ~20 sets per MZ or DZ twin group. Second, we used the so-called CpG island microarrays, and we interrogated only 12,000 loci in the genome. So if we add up these 12,000 loci, it would only equal about 1 percent of the genome, which is merely a tiny fraction to interrogate. The next step should be a dedicated replication effort that would investigate a much larger twin sample and use more comprehensive microarrays; ideally, we would utilize whole genome tiling microarrays. In addition, it would be particularly interesting to investigate some complex phenotypes and try to identify links between phenotypic differences in MZ and DZ co-twins vs. epigenetic differences in such co-twins.
Q: Is anything known about the difference between methylation or epigenetics in general in the nervous system versus the tissue types that you sampled? How does that play into your thinking about this? And do you have any ideas for how to surmount the fact that you’re not looking directly at the nervous system if you start thinking about this in terms of your schizophrenia or bipolar disorder research?
A: You’re absolutely right; this is a limitation. However, we have to separate basic human epigenetic studies from those in human diseases. If we want to learn about the principles of epigenetic metastability, blood, gut, or buccal cells are sufficient. In human disease epigenetic studies, unfortunately, we often cannot acquire tissue from the disease site. This directly applies to psychiatric diseases where brain samples are available only postmortem. Still, it is important to note that the non-brain tissues may contain vestiges of epigenetic changes that were inherited or occurred before tissue differentiation in embryogenesis. In our earlier studies, we detected that, despite significant epigenetic differences across tissues, epigenetic profiles from different tissues of the same individuals exhibit some degrees of similarity. Another important aspect is that the non-brain tissues may help to address the cause-effect conundrum between epigenetic changes and disease. If epigenetic abnormalities are detected in the brains of schizophrenia patients, we cannot immediately know which of these changes caused or predisposed to schizophrenia and which ones were induced by the disease or disease-associated events (e.g., compensatory mechanisms of the brain, treatment with neuroleptics). Compared to the brain, the non-brain tissues should contain fewer secondary epigenetic changes.
Q: Is there any chance that variation in DNA sequence could contribute to the differences between the monozygotic co-twins?
A: Yes, the idea that monozygotic twins are not necessarily identical in their DNA sequences is not new. The most recent observation is actually that monozygotic twins may have some copy-number variation, and there was a paper published from Jan Dumanski’s lab in the American Journal of Human Genetics last year (Bruder et al., 2008) showing that there are some loci in monozygotic twins that exhibit copy-number variation. There are some other studies over the last 15 years demonstrating that there are some somatic mutations—trinucleotide repeats and some single-nucleotide mutations—that can potentially make monozygotic twins different. All these observations are very interesting, and I believe many more insights are still to come when sequencing of the entire human genome will become less expensive. It is early to prognosticate what proportion of phenotypic differences will be explained by DNA sequence variation or epigenetic differences. At this point all we can say is that epigenetic differences in MZ twins by far exceed the DNA sequence ones.