NEW YORK (GenomeWeb) – As they've begun to dig into the available data on He Jiankui's germline genome editing experiment, CRISPR researchers have continued to condemn the work. But now, these investigators are not only appalled at the seemingly unethical nature of He's approach and lack of transparency, they're also disturbed by the science itself.
During his presentation at the Second International Human Genome Editing Summit in Hong Kong on Nov. 28, the Southern University of Science and Technology's He presented parts of what he said was a larger paper that was still being edited and peer-reviewed. He outlined the impetus for his work, noting that the naturally occurring CCR5Δ32 deletion mutation confers profound resistance to HIV1, and walked the assembled researchers through his experiments doing CCR5 knockouts in mice, designing guide RNAs for CCR5 in human cell lines, conducting experiments in monkeys, and then moving on to the human embryos. Once edited, these embryos were implanted and resulted in the birth of twin girls referred to as Lulu and Nana.
In part of his presentation, He paid particular attention to the work he and his team did to search for off-target effects in the human embryonic stem cells. The researchers had detected one off-target effect in an intergenic region, 279 kilobases away from any known genes and not near any known non-coding RNA or transcription factor binding sites, He said. In one embryo, the team identified a 6-kilobase deletion at the target site, which did not affect any gene other than CCR5.
"Pre-implantation genetic diagnosis showed two sites were edited. One was a frameshift lockout, which should shorten the CCR5 protein similar to the natural protective variation," He said during his presentation. "Another had a deletion, which was expected to maybe destabilize local protein structure around the nearby HIV1 binding site."
He also noted that deep sequencing of the cord blood after the birth of the twins confirmed the editing pattern the team saw in PGD testing, but neither deep sequencing nor Sanger sequencing detected the intergenic off-target effect that was observed during PGD testing. "This suggests that this was an artifact of a single cell amplification, or a mosaic off-target that happened to occur in the few blastocyst cells assembled for PGD," he added.
Digging into the science
At first, the reaction to He's work centered on the ethics and lack of transparency. Arthur Caplan, a bioethics professor at NYU School of Medicine, told GenomeWeb that in his opinion, what He did was "stupendously unethical" and even "eugenic" in nature. Martin Markowitz, clinical director of the Aaron Diamond AIDS Research Center at Rockefeller University, concurred, noting problems with He's informed consent procedures, and even calling his stated motivation for the experiment into question.
"So, the idea is that he was protecting these babies from HIV infection from their father? Well we know very clearly now that the chances of this HIV-infected father infecting the mother is probably zero if he's got a viral load that's undetectable for six months," Markowitz said. "So, I don't even understand what this is all about. I'm a little bit confused as to how this is going to lead us to an HIV vaccine. He called it HIV vaccine research, but I don't think that's the case. I think that what he did is eugenics research. This whole thing stinks."
Indeed, the live reaction to He's presentation seemed to center on the ethical questions as well, as the assembled audience questioned him on whether his human subjects had truly understood what they were consenting to and whether he understood what his responsibility would be to Lulu and Nana going forward.
But as the initial shock started to wear off, researchers began to look at the technical aspects of the experiment itself, referring to the slides He had presented in anticipation of the full data set. What they found disturbed them even further. Sean Ryder, a principal investigator and professor in the department of biochemistry and molecular pharmacology at the University of Massachusetts Medical School, summed up the experiment in a single graphic that he posted on Twitter — Ryder showed that none of the mutations generated by He's CRISPR work had actually resulted in a CCR5Δ32 deletion in either twin.
"I calculated the conceptual translation products of the mutant alleles in the two babies," Ryder noted in a statement. "One mutation ('-15') produces an in-frame deletion of five amino acids. A second, '-4,' produces an out-of-frame mutation and a truncated protein product. The third, '+1', makes a different truncated protein product that is similar but not identical to the Δ32 allele."
(Graphic courtesy of Sean Ryder)
Ryder, who uses CRISPR in his lab as a tool to make mutations in Caenorhabditis elegans, told GenomeWeb that initial news reports about the experiment had left him feeling a little incredulous and morbidly curious. But when he saw He's slides, he found that what the researcher kept referring to as CCR5Δ32 deletion mutations in the twins were actually three new mutations that, as far as Ryder could tell, had never been tested in an animal model.
"In addition to the fact that this actual thing had been done, I found it shocking that he would proceed to implantation of these embryos and making babies without knowing for sure that the mutations that he put in would have a benefit and would do no harm," Ryder said.
Ryder's graphic showed that Nana's CCR5 genes were edited with a 4-base pair deletion and a 1-base pair insertion, and that one of Lulu's CCR5 genes was edited with a 15-base pair deletion. It seemed that the second copy of CCR5 in Lulu was left unedited, which may result in her being heterozygous for the mutation and possibly lacking the protection against HIV1 that He's experiment intended in the first place.
But this is still unclear, according to Ryder. "What we know from the deep sequencing data post-birth of the umbilical cord, the placenta, and the cord blood is that there are two genotypes [in Lulu]," he said. "There's the -15 mutation and there's the wildtype or the unedited gene. It could be that in other tissues there will be other genotypes that reveal themselves. Having looked more closely at the chromatograms of the sequencing from before implantation, it suggests that the embryos were in fact mosaic. I would say that it's probably the case that she's not going to be protected from HIV, and she could be heterozygous. It's at least possible that half of the cells are completely edited, and half of the cells are unedited. And whether or not she's protected depends upon what genotypes are present in her CD4+ T cells which are the cells that HIV infects."
It is this uncertainty which has angered CRISPR experts.
"I have both serious ethical and scientific concerns with Dr. He's work. The lack of an unmet medical need, the failure to fully engage the communities of stakeholders, and numerous breaches of community and institutional policies and laws represent major ethics violations," Harvard University and Broad Institute researcher David Liu said in an email interview. "The disturbing lack of correspondence between the DNA sequencing data and the claimed genotypes of the babies, the somewhat superficial off-target analysis, and the lack of characterization of the consequences of the edited forms of CCR5 in these babies raise additional scientific questions that are deeply troubling."
Liu, who pioneered the CRISPR base-editing technique, also noted that Lulu's -15 mutation would result in an edited CCR5 gene that produces a protein missing five amino acids, rather than an edited CCR5 gene with an early stop codon. "I was shocked when Dr. He noted that he decided to proceed with initiating a pregnancy with this embryo after asking the parents. Placing the burden of such a decision on the parents, who presumably are not experts in the potential consequences for their daughter of this protein lacking these five amino acids, is unconscionable," he added.
In fact, according to Liu, the variability of the resulting mutations was almost predictable. Non-homologous end-joining, a pathway that repairs double-strand breaks in DNA, normally creates a mixture of products that are difficult to accurately predict, he said, adding, "The kind of outcomes Dr. He reported (-4, +1, -15 bp) are very typical of how mammalian cells resolve double-stranded DNA breaks such as those created by CRISPR-Cas9 nuclease."
Although some have argued that the specific mutation that was generated might not matter as long as CCR5 was deactivated, Liu sees plenty of potential for harm to Lulu and Nana. "The two frame-shifted variants will encounter early stop codons during translation that will lead to truncated proteins. But even these truncated variants are not necessarily biologically inert," he said. "And the variant lacking five amino acids should really be characterized before creating a baby who has to live with this variant (and whose descendants would as well)."
Ryder has also been trying to figure out if the specific mutations matter beyond simply inactivating CCR5. Previous studies on the gene — such as a 2014 New England Journal of Medicine paper on the deactivation of CCR5 using zinc-finger nucleases — have shown that there are various mutations produced through gene editing that could inactivate the gene, he said, but when one looks at the mutations that actually exist in humans, the list shrinks.
Ryder searched through the 1000 Genomes Project's database, looking for all insertions or deletions that occur within CCR5, and for how frequent or common these variations are. "Most of them were detected once or twice in 100,000 to 200,000 people, so [CCR5 mutations are] very, very, very uncommon with the sole exception of the famous Δ32 mutation, which is more common in northern Europeans," Ryder said. "I don't quite know what that means other than it implies that there's something special about the Δ32 mutation as opposed to any old deactivating mutation conferring the same benefit. And, so, it's really unknown — in a human population sense — whether or not [other mutations] have any benefit against HIV or any cost in terms of other diseases."
The bottom line for both Ryder and Liu — and indeed much of the research community — is that although it's entirely possible that Lulu and Nana will lead normal, healthy lives, the uncertainty surrounding the effects these edits will have means these embryos should never have been implanted and allowed to grow into babies. At the very least, they both noted, He should have conducted animal studies on these specific mutations before implanting the embryos.
"We should never need to have conversations about what these edited gene variants might do to a patient after the decision to proceed with a pregnancy containing these variants was made," Liu said. "These scientific uncertainties can be illuminated, if not fully resolved, using known methods, and we owe it both to patients and to other scientists in the field to do much, much better than was done in Dr. He's study."