Genome-wide analysis reveals specificities of Cpf1 endonucleases in human cells

Abstract

Programmable clustered regularly interspaced short palindromic repeats (CRISPR) Cpf1 endonucleases are single-RNA-guided (crRNA) enzymes that recognize thymidine-rich protospacer-adjacent motif (PAM) sequences and produce cohesive double-stranded breaks (DSBs). Genome editing with CRISPR-Cpf1 endonucleases could provide an alternative to CRISPR-Cas9 endonucleases, but the determinants of targeting specificity are not well understood. Using mismatched crRNAs we found that Cpf1 could tolerate single or double mismatches in the 3′ PAM-distal region, but not in the 5′ PAM-proximal region. Genome-wide analysis of cleavage sites in vitro for eight Cpf1 nucleases using Digenome-seq revealed that there were 6 (LbCpf1) and 12 (AsCpf1) cleavage sites per crRNA in the human genome, fewer than are present for Cas9 nucleases (>90). Most Cpf1 off-target cleavage sites did not produce mutations in cells. We found mismatches in either the 3′ PAM-distal region or in the PAM sequence of 12 off-target sites that were validated in vivo. Off-target effects were completely abrogated by using preassembled, recombinant Cpf1 ribonucleoproteins.