DNA repair outcomes at Cas9-mediated double-strand breaks are non-random and sequence-dependent

The repair of site-specific DNA double-stranded breaks (DSBs) generated by the RNA-guided DNA endonuclease Cas9 has been extensively used to alter gene function in human cells. Despite the widespread adoption of CRISPR-Cas9 technology to induce DSBs for genome engineering, the genomic scars resulting from repair of Cas9-dependent DSBs have not been examined in depth. Repair of DSBs in the absence of donor DNA has been thought to result in random, error-prone repair outcomes; here we show that the patterns of DNA repair following Cas9 cutting are non-random and are reproducible across cell lines and reagent delivery methods. Microhomology contributes reproducibly to a portion of the repair landscape at some sites, but does not fully account for the conservation of repair outcomes. Finally, we find that repair outcomes are determined by the protospacer sequence, the region of DNA targeted via Watson-Crick base pairing by the single-guide RNA (sgRNA) spacer, which opens a path to harness these mutagenic DNA-repair outcomes in a controlled manner to deliver precise editing events