CRISPR-Cas9 genome editing has potential to cure diseases without current treatments, but therapies must be safe. Here we show that CRISPR-Cas9 editing can introduce unintended mutations in vivo, which are passed on to the next generation. By editing fertilized zebrafish eggs using four guide RNAs selected for off-target activity in vitro, followed by long-read sequencing of DNA from >1100 larvae, juvenile and adult fish across two generations, we find that structural variants (SVs), i.e., insertions and deletions ≥50 bp, represent 6% of editing outcomes in founder larvae. These SVs occur both at on-target and off-target sites. Our results also illustrate that adult founder zebrafish are mosaic in their germ cells, and that 26% of their offspring carries an off-target mutation and 9% an SV. Hence, pre-testing for off-target activity and SVs using patient material is advisable in clinical applications, to reduce the risk of unanticipated effects with potentially large implications.

To investigate the extent and distribution of unintended mutations induced by CRISPR-Cas9 in vivo, we edited the genome of fertilized zebrafish eggs and investigated DNA from >1100 larvae, juvenile and adult fish in the F0 and F1 generations. Four guide RNAs (gRNAs) were used, selected from 23 gRNAs with high on-target efficiency in vivo in previous functional experiments. CRISPR-Cas9 outcomes were analyzed by long-read sequencing of on-target sites and off-target sites detected in vitro. In founder larvae, on-target editing of the four gRNAs was 93-97% efficient, and three sites across two gRNAs were identified with in vivo off-target editing. Seven percent of the CRISPR-Cas9 editing outcomes correspond to structural variants (SVs), i.e., insertions and deletions ≥50 bp. The adult founder fish displayed a mosaic pattern of editing events in somatic and germ cells. The F1 generation contained high levels of genome editing, with all alleles of 46 examined F1 juvenile fish affected by on-target mutations, including four cases of SVs. In addition, 26% of the juvenile F1 fish (n=12) carried off-target mutations. These CRISPR-induced off-target mutations in F1 fish were successfully validated in pooled larvae from the same founder parents. In conclusion, we demonstrate that large SVs and off-target mutations can be introduced in vivo and passed through the germline to the F1 generation. The results have important consequences for the use of CRISPR-Cas9 in clinical applications, where pre-testing for off-target activity and SVs on patient material is advisable to reduce the risk of unanticipated effects with potentially large implications.