Scientists say they can read nearly the whole genome of an IVF-created embryo


A California company says it can decipher almost all the DNA code of a days-old embryo created through in vitro fertilization (IVF)—a challenging feat because of the tiny volume of genetic material available for analysis. The advance depends on fully sequencing both parents’ DNA and “reconstructing” an embryo’s genome with the help of those data. And the company suggests it could make it possible to forecast risk for common diseases that develop decades down the line. Currently, such genetic risk prediction is being tested in adults, and sometimes offered clinically. The idea of applying it to IVF embryos has generated intense scientific and ethical controversy. But that hasn’t stopped the technology from galloping ahead.

Heart conditions, autoimmune diseases, cancer, and many other adult ailments have complex and often mysterious origins, fueled by a mix of genetic and environmental influences. Hundreds of variations in the human genome can collectively raise or lower risk of a particular disease, sometimes by a lot. Predicting a person’s chance of a specific illness by blending this genetic variability into what’s called a “polygenic risk score” remains under study in adults, in part because our understanding of how gene variants come together to drive or protect against disease remains a work in progress. In embryos it’s even harder to prove a risk score’s accuracy, researchers say. “Ultimately, how are we going to validate this in embryos?” says Norbert Gleicher, an infertility specialist at the Center for Human Reproduction in New York City who was not involved in the research. “We’ll have to wait for 40 or 50 years” to find out whether a person develops the diseases they were screened for as an embryo.

With current technologies, it’s very difficult to accurately sequence a whole genome from just a few cells, though some have tried with different methods. The new work on polygenic risk scores for IVF embryos is “exploratory research,” says Premal Shah, CEO of MyOme, the company reporting the results. Today in Nature Medicine, the MyOme team, led by company co-founders and scientists Matthew Rabinowitz and Akash Kumar, along with colleagues elsewhere, describe creating such scores by first sequencing the genomes of 10 pairs of parents who had already undergone IVF and had babies. The researchers then used data collected during the IVF process: The couples’ embryos, 110 in all, had undergone limited genetic testing at that time, a sort of spot sequencing of cells, called microarray measurements. Such analysis can test for an abnormal number of chromosomes, certain genetic diseases, and rearrangements of large chunks of DNA, and it has become an increasingly common part of IVF treatment in the United States. By combining these patchy embryo data with the more complete parental genome sequences, and applying statistical and population genomics techniques, the researchers could account for the gene shuffling that occurs during reproduction and calculate which chromosomes each parent had passed down to each embryo. In this way, they could predict much of that embryo’s DNA.

The researchers had a handy way to see whether their reconstruction was accurate: Check the couples’ babies. They collected cheek swab samples from the babies and sequenced their full genome, just as they’d done with the parents. They then compared that “true sequence” with the reconstructed genome for the embryo from which the child originated. The comparison revealed, essentially, a match: For a 3-day-old embryo, at least 96% of the reconstructed genome aligned with the inherited gene variants in the corresponding baby; for a 5-day-old embryo, it was at least 98%. (Because much of the human genome is the same across all people, the researchers focused on the DNA variability that made the parents, and their babies, unique.)

“What they presented is a nice method to sequence the genomes of all embryos,” says Shai Carmi, a statistical geneticist at the Hebrew University of Jerusalem. Such an accomplishment “is not trivial.” Kumar hopes being able to reconstruct most of an embryo’s genome will provide information well beyond what’s now available to people undergoing IVF, to determine an offspring’s chances of staying healthy. “It’s not enough to focus on the single gene anymore,” he says.

Once they had reconstructed embryo genomes in hand, the researchers turned to published data from large genomic studies of adults with or without common chronic diseases and the polygenic risk score models that were derived from that information. Then, MyOme applied those models to the embryos, crunching polygenic risk scores for 12 diseases, including breast cancer, coronary artery disease, and type 2 diabetes. The team also experimented with combining the reconstructed embryo sequence of single genes, such as BRCA1 and BRCA2, that are known to dramatically raise risk of certain diseases, with an embryo’s polygenic risk scores for that condition—in this case, breast cancer.

“We’re talking about providing information on risks that people care about—heart disease, cancer, autoimmune disease,” says Kumar, who is also a pediatric medical geneticist. He still sees patients and sometimes encounters frustration from parents wanting to avoid conferring a high risk of ailments that run in their families to their offspring. At the same time, Kumar stresses, “This is a new technology. It’s going to have controversies and challenges.”

In fact, many researchers say it’s premature to use polygenic risk scores to select which embryos are transferred. Such risk scores are “primarily still a research tool, even in adults,” says Barbara Koenig, a medical anthropologist who works on bioethics at the University of California, San Francisco. She’s involved in a large study called Women Informed to Screen Depending On Measures of risk that offers some women polygenic risk scores for breast cancer along with screening recommendations. “The scores are constantly being refined, every week they change,” Koenig says. “It’s like a constantly moving target.”

Kumar and his co-authors acknowledge the scores’ limitations, including that they are based on DNA from populations of overwhelmingly European ancestry and may be less accurate in other groups. Because of that, the MyOme team did not create disease risk assessments for embryos whose genome reflected at least 20% Asian or African ancestry. Even the DNA array technologies used to reconstruct the embryonic genomes have a European bias, says Genevieve Wojcik, a genetic epidemiologist at Johns Hopkins University, and may be less reliable for those with non-European ancestry. “You have a tool that cannot be used for a large proportion of the population,” she says. Kumar says the company is working to make the technology more broadly applicable.

There are other concerns, too. Although Carmi says the accuracy of polygenic risk scores in adults has improved, it’s unknown whether scores based on adult DNA and health data translate to embryos, in part because the environment can play a major role in shaping outcomes. “It’s difficult to say whether this will be meaningful,” Carmi says. He and his colleagues have seen this limitation up close: They’ve used computer modeling to assess whether height and IQ can be boosted by selecting embryos using polygenic risk scores for either trait, and found that generally, it doesn’t work. “We’re still missing a lot” when it comes to understanding genetics, even for highly heritable traits such as height, he says. In another computer modeling paper, however, Carmi found certain disease polygenic risk scores in embryos may prove useful. That’s because unlike height, which runs across a spectrum, heart attacks, say, either happen or they don’t. And pulling down genetic risk somewhat by implanting a different embryo, he says, may be enough to avoid that outcome.

But like a painting with only one corner visible, much of the human genome remains shrouded, including how genes interact with each other and the multiple effects one gene may have. Gleicher worries about the unintended consequences of applying polygenic risk scores to embryos. “You can achieve omission of one disease but at the same time, by doing that, induce another disease.” For example, modeling suggests selecting an embryo with a high polygenic risk score for educational attainment could also increase its risk for bipolar disorder. In December 2021, the European Society of Human Genetics urged against using polygenic risk scores for embryo selection—a position firmly endorsed by Gleicher, who calls such practice “unethical.”

Still, some companies and fertility clinics already claim they can help parents select embryos for IQ and risk of various diseases. MyOme, meanwhile, is applying the methods from this latest study to another that’s ongoing, working with IVF clinics and couples who want to learn polygenic risk scores for their frozen embryos. Couples may opt to decide which embryos to implant based on that information. “When you have a lot of information presented in this context, is it going to provide empowerment, or is it just going to confuse the parents?” Kumar asks. That’s one question he hopes this ongoing study can answer.

Kumar says he’s well aware of the criticisms, including that polygenic risk scores may not even be accurate for embryos. “That point is heard,” Kumar says. “Our focus is doing this research because we see promise.”

Science, 21 March 2022