Because twin studies are a bedrock of human genetics research, it's no surprise that they are being incorporated into whole-genome and epigenome studies, many of which attempt to pinpoint why one monozygotic twin has a disease while the other does not. However, a recent report in Twin Research and Human Genetics from the Whitehead Institute's Yaniv Erlich cautions that studies using blood-based samples might not be picking up the differences between twins. GT's Ciara Curtin recently spoke with Erlich to find out why.
Genome Technology: How did you come up with the idea of studying the effect of chimerism in twins on sequencing studies?
Yaniv Erlich: There was a very elegant paper by Baranzini [in Nature that looked at the genomic, epigenomic, and RNA sequences of twins discordant for multiple sclerosis]. And then we realized that 70 percent of the twins actually have chimerism. It's remarkable; we see these studies of twins for trisomy 21 where [researchers] checked the blood of the healthy twin and they found cells with trisomy 21 and then they checked the blood of the affected twin and found normal karyotype in the blood, but when you go to the skin, you see two distinct genotypes.
GT: In your study, you first assessed the level of chimerism found in twins?
YE: The twin community has recognized that there is a problem in chimerism, but nobody studied the profiles of chimerism and the rate of chimerism. The first step in our study was just to look at the profile, and we looked extensively in the literature to identify chimerism cases. What we found was something very interesting. If you think about each discordant twin as 7Up and Coke Zero — basically as the blood system, they are equal mixtures of Coke Zero and 7Up. The ratios can vary from twin case to twin case. Important, too, is that if you look at the two twins, they have almost the same composition of normal and affected cells.
GT: So if one twin has, say 70 percent normal cells, the other twin will too?
YE: It'll be 70 percent. No matter if you are the case or the healthy twin, you'll see the same thing. When we draw this, we get a straight line, this very nice linear relationship between them.
GT: How did you determine the effect of chimerism on SNP calls?
YE: After we had the profile, we said 'OK, let's take our data from another study that we were doing with exome sequencing,' and we took data from the mother and the son. We knew exactly the SNPs that we were looking at — [mother and son] are completely discordant for those SNPs. Then we just simulated frequencies. … And we tried different SNP callers and saw that if you have average levels, the same type of chimerism as in twins, [you] just lose the signal.
GT: About how many studies do you think are out there that might be reporting a false null result?
YE: I don't know if this is really because there is no difference between the twins at all. One would expect when you sequence two human genomes, even if they start from identical zygote, you will find some accumulation of mutations between [them]. But other studies are taking that into account, using other tissues — I'm not talking about high-throughput sequencing studies, but other studies. There was a very elegant Nature Genetics paper about four years ago and they recognized the problem and they took [samples] from the gut and they took from buccal cells to study methylation between twins. If you want to use blood, you need to know if the twins are dichorionic, so they don't share the blood system. But most of the time that's very hard to know because nobody keeps records of it.