On 2017 May 31, Xiaolin Wu commented:

This paper raises important safety issue for gene therapy application of CRISPR-Cas9. However, there are serious doubts about the results or interpretation. First of all, the authors listed Top-10 predicted off-target sites. But all genes are wrong! looking at the sequence they listed (supp. figure 3), you will not be able to find it in the genes! After careful inspection, the first predicted off-target is actually the "on-target" sequence for pde6b gene. For such a high-profile journal, you can't be so sloppy. This is not just a typo. I inspected them and they are all assigned to wrong gene. If you can't even get your on-target correct, how do you think people can trust your data? There are some genes are assigned to even wrong chromosomes! Supp fig3 panel b, listed herc1 gene on ch11. That gene is supposed to be on chr9. After this first figure, I don't even know if any other information reported here is correct!

I then went on to inspect Supp table 1-3. The authors listed all off-targets observed from the WGS. However, Pde6b pTyr347fs/c1041_1050CGTAGCAGAA is actually the on-target indel. and the author did not even notice this is their target gene? and listed it as one of the two off-target genes with mouse phenotype? The CRISPR-cas9 system is supposed to created Indel here! You simply did not repair it. You replaced the stop codon with the indel. I downloaded the raw sequence, and found that this specific deletion (CTGAGCAGAA)can not be found. Only by reading the authors previous paper, I figured out that they mean a 10 bp deletion but they don't even have the correct deletion sequence!

After seeing all these careless mistakes, I don't even know if they mislabeled the mouse or samples! It is hard for me to imagine CRISPR-case9 causes so many homozygous deletions in two independent mice (all right, it may happen in rare case for specific sgRNA like this one). And even if some of the mutations/indels are real, they may have nothing to do with CRISPR-cas9. For example, the authors see homozygous deletion in Pde9a gene in both animals. Do the authors consider the possibility that this deletion might be created by totally unrelated mechanisms and strongly selected for in vivo? since Pde9a and pde6b are paralogues. The easiest way to test if these are real CRISPR-cas9 off-target is to check these loci in treated cells in vitro. In that setting, you can check millions of cells to see if they do occur or do not occur. Maybe none of them is created by CRISPR-cas9 off-target. But during the embryo development, these mutations are created and strongly selected to compensate for something. I admit that in vitro does not speak for in vivo. But you can't just assume these mutations are generated by CRISPR-Cas9.

On 2017 May 31, Xiaolin Wu commented:

This paper raises important safety issue for gene therapy application of CRISPR-Cas9. However, there are serious doubts about the results or interpretation. First of all, the authors listed Top-10 predicted off-target sites. But all genes are wrong! looking at the sequence they listed (supp. figure 3), you will not be able to find it in the genes! After careful inspection, the first predicted off-target is actually the "on-target" sequence for pde6b gene. For such a high-profile journal, you can't be so sloppy. This is not just a typo. I inspected them and they are all assigned to wrong gene. If you can't even get your on-target correct, how do you think people can trust your data? There are some genes are assigned to even wrong chromosomes! Supp fig3 panel b, listed herc1 gene on ch11. That gene is supposed to be on chr9. After this first figure, I don't even know if any other information reported here is correct!

I then went on to inspect Supp table 1-3. The authors listed all off-targets observed from the WGS. However, Pde6b pTyr347fs/c1041_1050CGTAGCAGAA is actually the on-target indel. and the author did not even notice this is their target gene? and listed it as one of the two off-target genes with mouse phenotype? The CRISPR-cas9 system is supposed to created Indel here! You simply did not repair it. You replaced the stop codon with the indel. I downloaded the raw sequence, and found that this specific deletion (CTGAGCAGAA)can not be found. Only by reading the authors previous paper, I figured out that they mean a 10 bp deletion but they don't even have the correct deletion sequence!

After seeing all these careless mistakes, I don't even know if they mislabeled the mouse or samples! It is hard for me to imagine CRISPR-case9 causes so many homozygous deletions in two independent mice (all right, it may happen in rare case for specific sgRNA like this one). And even if some of the mutations/indels are real, they may have nothing to do with CRISPR-cas9. For example, the authors see homozygous deletion in Pde9a gene in both animals. Do the authors consider the possibility that this deletion might be created by totally unrelated mechanisms and strongly selected for in vivo? since Pde9a and pde6b are paralogues. The easiest way to test if these are real CRISPR-cas9 off-target is to check these loci in treated cells in vitro. In that setting, you can check millions of cells to see if they do occur or do not occur. Maybe none of them is created by CRISPR-cas9 off-target. But during the embryo development, these mutations are created and strongly selected to compensate for something. I admit that in vitro does not speak for in vivo. But you can't just assume these mutations are generated by CRISPR-Cas9.