The 10-year-old gene-editing tool known as CRISPR is indispensable for engineering plants, tailoring lab animals, and probing basic biology. But there’s a caveat when it is used to tweak human genes: Unlike lab mice, which are usually inbred and genetically identical, people’s genomes differ individually and by ancestry.
These ancestry differences mean CRISPR does not always edit some genomes as intended, particularly in people of African descent, whose genomes are most likely to differ from those used to steer CRISPR to a specific gene. A new analysis finds that failing to account for ancestry slightly skewed a massive sweep for cancer genes, causing it to miss genes important as drug targets in those of recent African descent.
The study, posted online, “shows a clear example of this [ancestry] problem,” says computational biologist Luca Pinello of the Harvard University–affiliated Massachusetts General Hospital, who has also studied the issue. He and the scientists behind the new work propose tools to avoid it.
To edit a human gene, scientists first design a short strand of RNA to match part of the gene’s DNA sequence. This guide RNA then leads CRISPR’s DNA-snipping enzyme to the right spot. If the guide RNA does not closely match the genome being edited, CRISPR may not make the desired cut.
The guide RNAs are usually based on a reference genome compiled from just a few people’s DNA that does not fully capture human diversity. And people with African ancestry are more genetically diverse than Europeans or Asians, whose ancestors left Africa and spread across the globe only relatively recently.
Ancestry issues with CRISPR were first reported 5 years ago, but Sean Misek, a postdoc at the Broad Institute of MIT and Harvard and lead author of the new preprint, wanted to explore them in cancer biology. His team turned to the Cancer Dependency Map, a collaboration that used CRISPR to systematically knock out roughly 18,000 genes in 1000 lines of cancer cells grown from individual human tumors. The project looked for genes involved in cancer growth or survival that could be targeted with drugs.
The team, which included senior authors Jesse Boehm at the Massachusetts Institute of Technology and Broad and Rameen Beroukhim of Broad and the Dana Farber Cancer Institute, found CRISPR failed to knock out 2% to 5% of the 18,000 genes in an individual cell line. These errors were about 20% more common in the 41 cell lines from people with recent African ancestry than in other groups, the team reports. As a result, “We’re missing [cancer drug] targets in individuals of African descent,” Misek says.
Similarly, Pinello’s lab and others have shown how ancestry mismatches can cause CRISPR to cut the genome in the wrong spot when it is used to treat diseases such as sickle cell disorder, which mainly affects people of African ancestry. Such off-target cuts could lead to cancer. The CRISPR ancestry problem is an example of how excluding diverse populations in genomics studies “may inevitably contribute to cancer health inequity,” says translational cancer biologist Olorunseun Ogunwobi of Hunter College of the City University of New York.
Both teams have built free web tools that compare a proposed guide RNA with tens of thousands of genomes procured from diverse populations. Ancestrygarden.org from the Boehm and Beroukhim labs lets users assess the impact of ancestral diversity on standard CRISPR guide RNAs. And CRISPRme from Pinello’s team checks for off-target matches. “The hope here is that we push people to rethink how they’re using CRISPR in the laboratory,” Misek says.
“This is an important issue,” and the web tools should help, says Melissa Davis, a genomicist who studies disparities in breast cancer at Weill Cornell Medicine. Adam Phillippy of the National Human Genome Research Institute agrees. “It is exactly tools like this—tools that make human genomic variation easy to explore and understand—that can help alleviate the bias in the first place.”