Doubts raised on key points of Nature paper on CRISPR gene editing of human embryos

Egli Mitalipov preprint
Egli, et al. preprint Fig 1e-f

Is it possible that CRISPR gene editing actually didn’t happen in many of the human embryos in that big Nature paper that made such news a couple weeks back?

Some doubts have emerged that call the main conclusions of the paper into question and argue that more definitive studies are needed to be sure.

Egli Mitalipov preprint
Egli, et al. preprint Fig 1e-f

An international team of top scientists led by first author Dieter Egli has responded via a preprint on Biorxiv to that Mitalipov team high-profile Nature paper on CRISPR gene editing of human embryos. Egli, et al. raise the possibility that the CRISPR gene editing as reported in the Nature study may actually not have happened, at least not in every case and perhaps not the way the Ma, et al. paper argued it did (via homology directed repair (HDR)-based CRISPR-Cas9 action specifically depending on interaction between normal maternal and mutant paternal chromosomes).

On one level it isn’t so unusual to see a scientific critique of and technical questions raised about a published paper that made splashy news. However, I see this particular case as a striking turn of events because although the new Egli, et al. piece is very collegial and diplomatic, they convincingly lay out a number of rather compelling reasons why the main conclusions of the Ma paper might be incorrect and the reasons why there may not have been CRISPR gene editing in many of the embryos. To be clear, Egli and colleagues don’t seem to be saying the Ma, et al. paper is definitely wrong, but they describe some quite reasonable ways in which the Ma paper could hypothetically have inadvertently reached incorrect central conclusions. To me these possible alternative explanations just simply make a lot of sense and are things that should have been ruled out as alternative explanations.

The preprint author group includes highly respected scientists (Dieter Egli, Michael Zuccaro, Michal Kosicki, George Church, Allan Bradley, and Maria Jasin) and the questions they raise should be taken seriously. On my first reading of their well-written piece, I was already convinced that their concerns may be valid. Early on in the piece they write this:

“Considering the data presented in Ma et al., alternatives to recombination between homologues are possible and would seem more likely, as the cell biology of fertilized eggs would appear to preclude the direct interaction between the maternal and paternal genomes required for inter-homologue HDR. Therefore, clear evidence for a novel linkage of maternal and paternal alleles is an imperative for any embryo that would be considered for future implantation.”

What are the main issues articulated in the Egli paper that raise at least some doubts about the main conclusions of Ma, et al. paper?

First, the Ma paper makes the unusual argument that HDR-driven gene editing occurred after CRISPR-Cas9-induced DNA breaks in the mutant paternal allele essentially exclusively using the normal maternal chromosome as a template within the same 1-cell embryo rather than via an introduced synthetic template. In fact, in some of the Ma paper’s studies no template was included so that CRISPR-Cas9 gene editing had to rely on endogenous DNA in the embryo for HDR. However, Egli, et al. point out that this is exceedingly unlikely because the male and female pronuclei are entirely physically separated in the 1-cell embryo (Figure 1e-f above). How could the maternal and paternal chromosomes have physically come together to mediate this HDR during meiosis? Hypothetically possible? I suppose, but it’s hard to imagine a likely mechanism.

What about HDR later during mitosis? This is theoretically possible as the preprint notes the maternal and paternal genomes come together at that later time: “Merging of maternal and paternal chromosomes does not occur until microtubule action assembles both genomes on a common metaphase plate at the first mitosis…direct interactions between maternal and paternal genomes required for inter-homologue repair do not seemingly occur until embryos enter the 2-cell stage when the two genomes are packaged within the same nucleus.” If gene editing via HDR did occur much later during mitosis (but note that such recombination is thought to occur much less frequently during mitosis than meiosis), the Mitalipov team should have seen dramatically more mosaicism. Importantly, in addition if the gene editing only took place that late, why would they have observed such an apparently large difference in outcomes between MII and zygote injections?

Second, Egli, et al. point out the potential risks of relying as the Ma team did for some assays just on the apparent absence of a detectable mutant allele. Concerningly, there are a number of reasons other than CRISPR gene editing of a mutant allele back to wildtype that could more simply explain why no mutant alleles were detected. The Egli piece argues that one such possibility that wasn’t ruled out is the presence of moderate-to-large deletions resulting from CRISPR activity and eliminating primer binding sites, leading to no amplification of mutant alleles. This may in theory result in only WT alleles being detected leading to the potential incorrect conclusion of gene editing reversion of mutant alleles, when in the fact in that scenario the altered (not repaired) mutant alleles are there but just not detectable.

The preprint mentions unpublished data that such large deletions occur with CRISPR gene targeting in around 20% of gene edited cells. Sequencing to conclusively rule out Indels as a cause of failure to detect mutant alleles in various embryos or cells could be very difficult for a variety of technical reasons, but it could be done. I’m not clear as to whether the extent of genome sequencing in the Ma, et al. paper was enough to be sure.

Third, another alternative possibility discussed is that there sometimes was no contribution of the paternal genome to the zygote and hence no paternal genome present in some later embryos or ES cell derivatives. The Egli, et al. piece describes one possible way that could happen: “Zygotes with a single pronucleus are not uncommon after intracytoplasmic sperm injection, occurring in ~10% of fertilization attempts, and are mostly of parthenogenetic origin, containing only the maternal genome.” That makes sense to me as a possible alternative explanation. In addition, it is also possible that “a fraction of embryos derived from successful fertilization with mutant sperm are at more risk of paternal chromosome loss due to the occurrence of the Cas9-induced DSB.” In either case, lack of a paternal genome being present would give the incorrect appearance that mutant paternal alleles had been gene edited back to a WT state. There is some data in the Ma, et al. paper showing paternal contribution to certain embryos/ES cells, but not from enough samples to be sure overall.

I asked Gaetan Burgio, Group Leader in Genetics of Host-pathogens interactions and Genome editing, and Head of the Transgenesis Facility at The Australian National University, for his reaction to the Egli, et al. preprint:

“This preprint manuscript raises serious concerns about Ma et al. paper published in Nature over the finding that the deleterious mutation was corrected by “self repair” from the non-disease copy of the genome. They also argue that many underlying and important changes (large deletions) were undetected after editing in these human embryos.”

While the potential remarkable claim of inter-homologous repair in the Ma, et al. paper could still turn out to be correct, my sense is that it seems more likely that some of the time other events occurred such as those possibilities outlined in the Egli, et al. piece. It concludes:

“In summary, the conclusion of gene correction in human embryos requires further investigation, including direct verification. Efficient inter-homologue recombination in embryos in which the maternal and paternal genomes are undergoing distinct biological programs and in distinct nuclei would be a stunning biological finding.”

It will be important to see how the Mitalipov team responds (e.g. with additional data?) to the Egli, et al. piece. Perhaps there are good reasons why they think they are still correct.


    • Hey Leonid, Other groups have gotten human SCNT to work to produce hESCs. I wouldn’t say that paper was “never reviewed”. The review occurred during a period of only a few days. That’s possibly concerning too but not the same as there being no review.

  1. This is fascinating. The only way to resolve this is through examination of the paternal and maternal genomes separately. I think I have an idea. I’ve worked with a company called BioNano Genomics, that creates de novo maps by marking single DNA strands at known intervals and imagining long strands of DNA, using the markers to measure the physical distances. We used this method to detect some big single strand deletions in iPSCs (

    What do you think?

  2. Even given that the Mitalipov data was generated from whole genome amplification, it would be fairly easy to determine two alleles present versus one by PCR (qPCR would be undoubtedly more sensitive in this regard). I think Mitalipov’s group went to far in their claim that the repair is happening so early. With the amounts of Cas9 injected and given how long it hangs around, there’s a reasonable chance that gene conversion (I refuse to use novel terms for previously described mechanisms) happened once the genomes were together in one nucleus (either 2- cell or 4-cell stage seems possible)

    That said, I doubt we’ll see new data on this from Mitalipov’s group.

  3. Egli et al. stress the fact that pronuclei do not interact which is correct however there is clear interaction of maternal and paternal chromosomes on the first mitotic metaphase plate before division to the two cell stage. It it possible that HDR could occur at that point and Egli et al. could have discussed this in their otherwise well formed critique.

    • The chromosomes align, but my understanding was that there is probably little recombination at the first mitotic metaphase. Plus, if CRISPR gene editing via HDR can only occur then and only sometimes, wouldn’t there be expected a lot more mosaicism and/or lower efficiency than what Ma, et al. reported? For instance, if the gene editing is mitosis driven, sometimes it might not happen until the 2nd mitosis, leading to mosaicism. What do you think?

      • Perhaps there is some novel biology here that has not been demonstrated before (referring to HDR/recombination during metaphase of first mitotic division) but this is certainly more plausible than the idea put forward by the illustrations of the original authors that HDR is occurring in separate pronuclei (which is rightly criticized by Egli) Either way more detailed genetic analysis is needed to conclude anything.

Leave a Reply