Whole-genome sequencing characterizes multiple mutational mechanisms resulting from off-target effects of CRISPR-Cas9 and TALEN treatments in human embryonic stem cells. R. L. Collins1, A. Veres8,9,7, H. Brand1,4, A. Ragavendran1, S. Erdin1, Q. Ding8,9, B. S. Gosis8,9, K. Musunuru8,9,7,4,5, M. E. Talkowski1,2,3,4,6 1) Center for Human Genetic Research, Massachussets General Hospital, Boston, MA; 2) Molecular Neurogenetics Unit, Massachussets General Hospital, Boston, MA; 3) Psychiatric and Neurodevelopmental Genetics Unit, Massachussets General Hospital, Boston, MA; 4) Broad Institute, Cambridge, MA; 5) Division of Cardiovascular Medicine, Brigham and Women's Hospital, Boston, MA; 6) Department of Neurology, Harvard Medical School, Boston, MA; 7) Department of Medicine, Harvard Medical School, Boston, MA; 8) Department of Stem Cell and Regenerative Biology, Harvard University, Cambridge, MA; 9) Harvard Stem Cell Institute, Cambridge, MA.

   Recent innovations in genome editing technologies have opened access to revolutionary new approaches in genetic research. The use of targeted nucleases, such as CRISPR-Cas9 and TALENs, have become widespread in human biology. The specificity of these tools to the targeted editing site, and the confounds presented off-target effects, have been widely debated, with most studies to assess off-target effects focusing on single nucleotide variations (SNVs) or indels at predicted sites based on sequence specificity. Here, we performed direct comparison of off-target effects from two genome-editing technologies (CRISPR-Cas9 and TALENs) in HUES9 embryonic stem cell clones using unbiased whole-genome deep sequencing (~60X coverage). We sequenced a parental, untreated line, and nine treated clones. Each treatment group contained two clones with successful on-target mutations and one clone without targeted edits, serving as a treatment control. We selected three clones each with TALEN and CRISPR/Cas9 editing of the SORT1 locus, as well as three CRISPR-Cas9 clones from an independent locus for comparison (LINC00116). We investigated the full mutational spectrum of genomic variation accessible to short read sequencing in these clones. Of the clonal SNVs that differed from the parental line, none were within 100bp of a predicted off-target site. Of 28 small indels identified across all clones, only one (from a TALEN treatment) was in proximity to a predicted off-target site, and none were attributable to CRISPR-Cas treatment. We also surveyed all balanced and unbalanced structural variation (SV) using the convergence of three different paired-end and split read mapping methods, as well as focal read depth analysis. Two clonal SVs were detected: a 5.5kb deletion with no discernible homology to the forecasted on- or off-target sites, and an intriguing 564bp interchromsomal insertion at the on-target site of LINC00116 that exhibited five bases of microhomology with the target locus, suggesting homology-directed repair of the double-stranded DNA break. Finally, we also find support for integration of the Cas9 vector into the host genome. These results suggest that off-target mutations related to TALEN/CRISPR genome editing technologies are relatively infrequent, though such effects could present confounds when modeling disease and warrant consideration.

You may contact the first author (during and after the meeting) at