Mitalipov rebuts CRISPR pub critiques: strengthens case, but puzzles remain

Remember that Shoukhrat Mitalipov lab paper on the use of CRISPR in human embryos? It’s back in the news.

One of the biggest stories of 2017 centered on a Nature paper (Ma, et al., see my quick, initial review shortly after it was published here) from Mitalipov’s lab claiming both efficient repair of a disease-causing mutation in human embryos via CRISPR-Cas9 gene editing and that the correction happened at the MYBPC3 gene via an unexpected mechanism using the normal maternal chromosome as a template (the mutation was paternal) via so-called “interhomolog repair.”

Then another group’s preprint (Egli, et al.) upped the intensity by challenging the Ma paper’s central conclusions and proposing alternative explanations that some of us (see my posts here and here) found quite plausible.

Now Mitalipov’s team has published a rebuttal of a sort to the critiques in Nature along with the finalized paper from Egli, et al. and a comment from another group led by Paul Thomas.

Now everything’s crystal clear, right?

Not really, but a quick take-home is that Mitalipov’s group has convinced me that it is more likely than not that some interhomolog repair happened in some embryos at least, but things are still somewhat up in the air.

I won’t rehash all the main issues that led to questioning of the Ma paper, but just try to be concise about it and how Mitalipov’s group responded.

Chopped to be undetectable rather than fixed? First, there was some concern that rather than really repairing the mutant allele via CRISPR-Cas9 and the maternal WT chromosome as template, CRISPR-Cas9 had further damaged the allele rendering it undetectable by PCR/sequencing and hence giving the incorrect appearance of an entirely WT embryo. Mitalipov’s team responded to this mainly with PCR data in Figure 1 of their rebuttal. They used a series of increasingly widely spaced primers to generate an expanding set of amplicons using genomic DNA from embryos/blastomeres or ES cells as templates. If there were alleles with large deletions, they reasoned this approach would detect them. On the 1% agarose gels shown, there are no intense small product bands that one might expect with certain primer sets if deletions were present. However, there are some important caveats here. The data in Figure 1 are frankly not so clear nor precise, and pretty messy in some cases (see above). Also, there are in fact, as the authors acknowledge, extra bands here and there. While they dismiss these extra bands without much of a look, they could have meaning. Surprisingly, they also did not sequence their dominant amplicons in Figure 1, which could have probably made things clearer.

Lost rather than fixed? Second, it was also pointed out by Egli, et al. that it was possible that the mutant paternal allele wasn’t fixed, but rather lost either via chromosome loss of parthenogenic embryo development. Such loss would mimic the appearance of a “fixed”, only WT embryo. Mitalipov’s team aims to rebut this with various SNP data in their reply now suggesting that there isn’t parthenogenesis and that various embryos and blastomeres did in fact arise from both maternal and paternal contributions. I’m still trying to wrap my head entirely around the data here, but so far I’m thinking it’s generally supporting their claim. There could be other explanations, although I’m not sure how likely they are.

A new, unusual mitotic model of human embryo gene editing raises more puzzles. One of the take-homes from the Mitalipov reply piece is a new model that differs somewhat from the original Ma paper in that they are now arguing for much later gene editing occurring probably at the two-cell stage and via mitosis.

While an intriguing model, it has some issues and puzzling aspects to it.

Mitotic interhomolog repair happens such as in ES cells, but is very rare, while the Ma paper suggests it happens at least in embryos quite often.

If it does happen during mitosis, wouldn’t this sometimes lead to LOH and reduced genetic diversity? I’m not an evolutionary biologist, but it seems like mitotic interhomolog repair in early embryos could cause issues and be inconsistent with our current understanding of how things work.

Perhaps one of the biggest issue is that late CRISPR-Cas9 triggered allelic repair really should lead to a high degree of mosaicism, while Ma, et al. reported little mosaicism.

Also, if CRISPR-Cas9 hangs around dormant but stable in embryos until this late point, why would Ma see such differences in outcomes with the timing of injection of CRISPR-Cas9 machinery either at MII or when they did it much later just before the 2-cell stage? In addition, embryos just tolerate a DSB for all that time prior to mitosis and don’t fix it with NHEJ that would then probably interfere with HDR later?

So overall, Mitalipov’s group has made a more convincing case for interhomolog repair with CRISPR in human embryos, but with so many questions, until we all see more data from various groups, we can’t be sure. One practical challenge is that almost no labs can do this kind of research. Still I’d imagine a year or two from now we’ll have learned more that hopefully will clarify things further.

In the bigger picture I also remain concerned about using CRISPR in human embryos for heritable genetic modification, a notion that was one potential motivation for this work originally.