Skip to Main Content

Editors’ note (Sept. 27, 2019): The senior author of the study on which this story is based has cited likely errors in the research. Read more here

Last January, the Chinese scientist who created the first “CRISPR babies” — twin girls born from gene-edited embryos — sent an American bioethicist friend an email suggesting he was having second thoughts about what he’d done.

“I have been thinking,” He Jiankui wrote Stanford University bioethicist Dr. William Hurlbut. “I recognize I pushed too quickly into a first-of-kind clinical study without the necessary open dialog with regulators, the scientific community, and the public.” Hurlbut shared the email with STAT — the first communication to be made public in which He is known to have expressed any regrets.


The premature birth last October of the girls, whom He calls Nana and Lulu, created an uproar because he rushed to alter the DNA of human embryos in ways whose consequences are largely unknown — changes that would be inherited by the girls’ offspring. A new paper provides the latest evidence of why critics worldwide attacked He, formerly of Southern University of Science and Technology in Shenzhen, for a deplorable ethical lapse.

People with the rare genetic variants that He tried to engineer into the embryos, using CRISPR genome-editing technology, have a 21% higher mortality rate than people with the more common form of the gene, scientists reported on Monday.


“We keep learning things we didn’t know [about the CCR5 gene that He tried to alter], and that should give us pause,” said science historian J. Benjamin Hurlbut of Arizona State University, the son of William Hurlbut. “Heritable genetic modifications will intrinsically be an experiment — an experiment conducted on children. We better be damn sure what we’re doing before we embark on that.”

The CCR5 variant, called Δ32, is best known for protecting against infection with HIV, the virus that causes AIDS. The gene codes for a receptor on the surface of immune cells. The Δ32 variant, which has 32 fewer DNA bases, produces a receptor that is too short to poke through the cell surface, so HIV cannot use it the way it normally does to enter cells. Δ32 is most common in northern Europeans (it’s thought to have come from the Vikings) but still rare, found in 16% of that population compared to 6% of Italians, 4% of Greeks, and even lower percentages of other ethnic groups.

In explaining his choice of CCR5, He has said he wanted to protect children against HIV infection, which is extremely stigmatized in China, but acknowledged that Δ32 had been associated with a higher risk of serious complications and death from influenza as well as infection with West Nile virus.

To get a better sense of the health effects of Δ32, Xinzhu Wei and Rasmus Nielsen of the University of California, Berkeley, analyzed data from 409,693 people of British ancestry in the United Kingdom Biobank, an ambitious effort to study the effects of genes on health and other characteristics. They examined three Δ32 possibilities: two copies (one from mom and one from dad, which 4,349 people had), one copy (83,038 people), or zero copies (meaning both of a person’s CCR5 genes are full-length, not lacking the 32 letters missing in Δ32, as was the case with 308,317 people).

People with two copies of Δ32 are about 20% less likely to make it to age 76 than people with one copy or none, the Berkeley researchers reported in Nature Medicine. And given the frequency of the Δ32 mutation among Britons (11.59%), fewer people with two copies had volunteered for the U.K. Biobank than expected. That suggests they might have become too ill to participate; such studies, in general, have more healthy participants than very sick ones.

“∆32/∆32 individuals have a significantly higher death rate than the other two genotypes,” the researchers wrote.

Nielsen, a professor of integrative biology at Berkeley, called ∆32 “probably not a mutation that most people would want to have. You are actually, on average, worse off having it,” including because of the influenza and West Nile consequences. The higher risk of early death, he told STAT, “likely has to do with reduced efficacy of the immune response to specific diseases. The effects depend on the environmental context, specifically the pathogenic environment.”

The death rate of people with a single copy of ∆32, however, is no higher than people with none, possibly because “immune function is largely preserved as long as you have one functional copy of the gene,” Nielsen said.

On the other hand, the centuries-long persistence of ∆32 suggest it confers benefits beyond HIV protection (HIV appeared too recently to explain why a Viking-era mutation has hung on). The best guess comes from small studies suggesting that ∆32 protects against flaviviruses (such as yellow fever) and smallpox, and might help people recover from stroke.

The largest legitimate study to try to change CCR5 genes into the HIV-protecting form was conducted by Sangamo Therapeutics. It did not report any excess mortality in its adult participants — all adults — but that’s probably because they were relatively young (30 to 55) and have been followed for barely 10 years, said Fyodor Urnov of Berkeley’s Innovative Genetics Institute, who helped lead the Sangamo study.

“The subjects have not been tracked long enough,” he said.

“If there ever was a poster child [for using], at present, genome editing solely to treat existing disease in consenting adults, and pediatric subjects where the parents can consent, this is it,” Urnov said of the Nature Medicine paper.

As it happens, He’s chaotic experiment missed its stated goal. Although he intended to edit embryos’ CCR5 gene to produce the ∆32 variant, he instead created chaotic edits. Lulu has one normal CCR5 gene (that is, CRISPR missed) and one with a 15-base-pair deletion, not the 32-base-pair deletion. Nana has one CCR5 gene missing four base pairs and one with an extra base pair.

No such CCR5 mutations are known to exist in nature, so their health consequences are unknown.

  • «Lulu has one normal CCR5 gene (that is, CRISPR missed) and one with a 15-base-pair deletion, not the 32-base-pair deletion. Nana has one CCR5 gene missing four base pairs and one with an extra base pair.

    No such CCR5 mutations are known to exist in nature, so their health consequences are unknown..»

    Like the ∆32 variant, all three of these are frame shifts around the cut site. While we can’t know their effects for certain until we study them, we can make a pretty good guess that they should be similar to said ∆32 variant (Lulu will be fine & will likely receive neither benefit nor harm whereas Nana’s phenotype will be like ∆32).

Comments are closed.