CRISPR–Cas immunity protects prokaryotes against invading genetic elements¹. It uses the highly conserved Cas1–Cas2 complex to establish inheritable memory (spacers)^2–5. How Cas1–Cas2 acquires spacers from foreign DNA fragments (prespacers) and integrates them into the CRISPR locus in the correct orientation is unclear^6,7. Here, using the high spatiotemporal resolution of single-molecule fluorescence, we show that Cas1–Cas2 selects precursors of prespacers from DNA in various forms—including single-stranded DNA and partial duplexes—in a manner that depends on both the length of the DNA strand and the presence of a protospacer adjacent motif (PAM) sequence. We also identify DnaQ exonucleases as enzymes that process the Cas1–Cas2-loaded prespacer precursors into mature prespacers of a suitable size for integration. Cas1–Cas2 protects the PAM sequence from maturation, which results in the production of asymmetrically trimmed prespacers and the subsequent integration of spacers in the correct orientation. Our results demonstrate the kinetic coordination of prespacer precursor selection and PAM trimming, providing insight into the mechanisms that underlie the integration of functional spacers in the CRISPR loci.

The bidirectional replication of a circular chromosome by many bacteria necessitates proper termination to avoid the head-on collision of the opposing replisomes. In Escherichia coli, replisome progression beyond the termination site is prevented by Tus proteins bound to asymmetric Ter sites. Structural evidence indicates that strand separation on the blocking (nonpermissive) side of Tus-Ter triggers roadblock formation, but biochemical evidence also suggests roles for protein-protein interactions. Here DNA unzipping experiments demonstrate that nonpermissively oriented Tus-Ter forms a tight lock in the absence of replicative proteins, whereas permissively oriented Tus-Ter allows nearly unhindered strand separation. Quantifying the lock strength reveals the existence of several intermediate lock states that are impacted by mutations in the lock domain but not by mutations in the DNA-binding domain. Lock formation is highly specific and exceeds reported in vivo efficiencies. We postulate that protein-protein interactions may actually hinder, rather than promote, proper lock formation.