A new, brief study in mice from the Vinit Mahajan lab argues CRISPR can have very large numbers of off-target sites. Their paper is entitled “Unexpected mutations after CRISPR–Cas9 editing in vivo” and was published in Nature Methods.This work has garnered a lot of attention in the media. Let’s take a journal club review kind of approach to this pub.
In this paper there were two major reported findings. First, the number of reported CRISPR-associated mutations in the two mice studied (versus one control) were high, over a thousand overall and a small number were predicted to be functionally important in the sense of being disruptive to gene expression. Second, many of the off-target sites were not ones that would readily be anticipated in advance based on widely used computational predictive approaches and were not particularly similar to gRNA sequences. In the two mice the off-target sites strongly overlapped suggesting a non-random phenomenon.
The big question from this paper is how often this kind of large-scale off-target activity would be expected to occur throughout the community of scientists using CRISPR-Cas9 and is there something unusual about the methods in this paper that led to this activity? From the methods, “Briefly, an sgRNA-expressing plasmid had been coinjected, into FVB/NJ zygotes, with the single-stranded oligodeoxynucleotide (ssODN) donor template and Cas9 protein to generate mosaic F0 founders.”
I’m not clear on the rationale for use of both plasmid and Cas9 protein. For more detail, see their separate original paper on generation of the mice. At the moment I cannot seem to access that full original paper, but on Twitter Lluis Montoliu indicated that the plasmid pX335 had been used, which could lead to persistent activity and ultimately aberrant genetic modifications, particularly if it integrated. Comparing the data here to defined off-target activity in mice in which gene editing was mediated transiently only by RNPs would be quite interesting.
In the journal club spirit of this review, I’ve asked for a couple of expert opinions on the paper. Gene editing guru Gaetan Burgio, who is a Group Leader and Head of the Transgenesis Core at The Australian National University, told me this about the paper:
“The claims over this paper are unsurprising as Cas9 enzyme could remain in the cells for days and create random indels in the genome. However, the main issues for me resides in the overestimation of the number of off target effects due to the lack of rigor in the experimental design to detect these unexpected mutations. In short my main point is these unintended mutation are likely to have preexisted prior to the injection of CRISPR system.”
I also asked Stephen Floor, a top CRISPR researcher who is an Assistant Professor at UCSF, for his take on this paper:
“I was excited to read the report of unexpected genomic alterations following zygotic Cas9 editing in mouse by Schaefer et al. This study highlights the importance of examining regions near to and far from the targeted site for genomic alterations. It will be interesting to repeat this study using other guide RNAs, repair templates, and editing protocols to determine if something about the conditions used in this work induces promiscuous editing. The ongoing use of CRISPR-Cas9 editing in many organisms suggests off-target editing is generally lower than is reported in this work. Determining the origin of this difference will inform in vivo editing protocols and associated analysis.”
The small N in this study of just 2 mice and 1 guide combined with use of plasmid DNA injection (along with protein) means that the experiment both has very limited power and has the potential for very long periods of persistent nuclease activity that together are likely to make the reported magnitude of off-target effects unusually high. CRISPR-Cas9 off-target activity should be studied carefully in both a direct way as research unto itself and considered for individual gene editing studies. Overall, this study adds a bit to the knowledge base, but it has been overinterpreted in the media. Most likely in general this magnitude of off-target activity is not occurring in most gene editing research, but more studies with larger N’s and different approaches will help to clarify the situation.