Since most dominant human mutations are single nucleotide substitutions1,2, we explored gene editing strategies to disrupt dominant mutations efficiently and selectively without affecting wild-type alleles. However, single nucleotide discrimination can be difficult to achieve3 because commonly used endonucleases, such as Streptococcus pyogenes Cas9 (SpCas9), can tolerate up to seven mismatches between guide RNA (gRNA) and target DNA. Furthermore, the protospacer-adjacent motif (PAM) in some Cas9 enzymes can tolerate mismatches with the target DNA3,4. To circumvent these limitations, we screened 14 Cas9/gRNA combinations for specific and efficient disruption of a nucleotide substitution that causes the dominant progressive hearing loss, DFNA36. As a model for DFNA36, we used Beethoven mice5, which harbor a point mutation in Tmc1, a gene required for hearing that encodes a pore-forming subunit of mechanosensory transduction channels in inner-ear hair cells6. We identified a PAM variant of Staphylococcus aureus Cas9 (SaCas9-KKH) that selectively and efficiently disrupted the mutant allele, but not the wild-type Tmc1/TMC1 allele, in Beethoven mice and in a DFNA36 human cell line. Adeno-associated virus (AAV)-mediated SaCas9-KKH delivery prevented deafness in Beethoven mice up to one year post injection. Analysis of current ClinVar entries revealed that ~21% of dominant human mutations could be targeted using a similar approach.