The research, published in June 2019 in Nature Medicine1, had suggested that people with two copies of a natural genetic mutation that confers HIV resistance are at an increased risk of dying earlier than other people. It was conducted in the wake of controversial experiments by the Chinese scientist He Jiankui, who had attempted to recreate the effects of this mutation in the gene CCR5 by using the CRISPR gene-editing tool in human embryos. The twin girls born last year as a result of the work did not end up carrying this exact mutation, but the research attracted attention because of its potential relevance to such experiments.
But a flurry of studies2,3,4 that looked anew at the Nature Medicine research — some of which analysed new data from genome databases comprising sequences from hundreds of thousands of people — have rejected the results and find no evidence that people with the mutation die early. The erroneous conclusion about CCR5 was caused by technical errors in how the mutation was identified in a population-health database.
“I feel I have a responsibility to put the record straight for the public,” says Rasmus Nielsen, a population geneticist at the University of California, Berkeley, who led study, which the authors retracted on 8 October. Nielsen also co-authored one of the papers rebutting its findings.
Some researchers stress that because the twins did not receive exactly the same mutation that occurs naturally, the original research and its retraction would not necessarily offer insights into their health anyway. But the episode raises questions about how best to assess the safety of future attempts to edit genes in human embryos.
He Jiankui shocked the scientific world when he announced, in November 2018, that his team had used CRISPR to disable the CCR5 gene in two babies born that month. He, who was at the time a biophysicist at the Southern University of China in Shenzhen, said he chose to target CCR5 because people with a 32-DNA letter deletion known as delta-32 in the gene are resistant to HIV but seem not to experience significant related health problems.
He has not published data supporting his work, but his announcement — presented at a scientific meeting — indicated that, for one of the twins, both copies of CCR5 were altered, whereas the other twin carried edits in just one of her two copies. None of the changes exactly matched the delta-32 variation.
Research has hinted that the delta-32 mutation, which is relatively common in people of European ancestry, might carry downsides — one small study5 found that carriers were more likely than other people to die from influenza. To tackle the question in larger data sets, Nielsen and his Berkeley colleague Xinzhu Wei looked at the UK Biobank, a database containing genome and health data from 500,000 British people.
Their Nature Medicine paper1 reported that people with two copies of delta-32 — whom they estimated to make up about 1% of biobank participants — were slightly more likely to die by the age of 76 than were those with one or no copies. They also found that the database harboured fewer people with two copies of delta-32 than evolutionary theory predicted it should — a sign that individuals with two copies were dying earlier, on average, than the population at large, Wei and Nielsen argued.
Results not replicated
Questions over the conclusion emerged as soon as the paper was published. Sean Harrison, an epidemiologist at the University of Bristol, UK, attempted to replicate the findings that night. He did not have UK Biobank data on the gene variant that Wei and Nielsen used to identify carriers of delta-32, so he analysed genetic variants near it on the genome that should have given the same result (adjacent parts of the genome tend to be inherited together, allowing scientists to infer the presence or absence of a DNA sequence by analysing neighbouring variants). When they didn’t, he described his findings in a series of tweets and later a blogpost.
The discrepancy Harrison identified piqued the interest of David Reich, a population geneticist at Harvard Medical School in Boston, Massachusetts, whose lab is studying CCR5. Working with Nielsen, his team discovered2 that Neilsen and Wei’s method had caused them to undercount the number of people in the UK Biobank with two copies of the delta-32 mutation, because the probe that measured the variant they were tracking did not always identify its target sequence. This — and not the supposed harmful effects of the mutation — explained the apparent absence of carriers from the UK Biobank database, says Nielsen.
He stresses that the undercounting problem is unique to the gene variant his team looked at, and not a general issue with UK Biobank data. “There were checks we could have done and should have done that we didn’t do. We missed the fact that there was a genotyping error,” Nielsen says.
Another follow-up study3, posted on a preprint server last week and based on genome databases that together include nearly 300,000 people from Iceland and Finland, also found no evidence that people with two copies of delta-32 die earlier than others.
No green light
Researchers stress that the unravelling of Wei and Nielsen’s results does not mean that it is a sound idea to target CCR5 for gene editing. “It’s very reasonable to expect that it might have a valuable function that we just don’t know how to measure. It seems very unwise to edit it out,” says Reich.
Gaétan Burgio, a geneticist at the Australian National University in Canberra, says the original Nature Medicine paper offered no insights into the health of the gene-edited twins. “Therefore, the retraction and these additional studies on European populations will still have no relevance to CRISPR babies, in my view,” he says. Population-based studies are very unlikely to give insights on these two babies, who don’t carry the CCR5-delta-32 mutation, he says.
Nielsen hopes that his team’s error does not dissuade others from using databases such as the UK Biobank to understand the effects of editing the human germ line, the DNA that can be passed on to future generations.
Kári Stefánson, head of the company deCODE genetics in Reykjavik and a co-author of one of the papers that found no evidence that delta-32 is harmful, says that Nielsen’s original study was not a valuable contribution to debates over germline gene editing. But he agrees that resources such as the UK Biobank and his company’s data on Iceland’s population can inform future efforts. “I think databases like this provide a fairly good way of assessing the probable effect of altering bases, no question about it,” he says.
Nature 574, 307 (2019)