END nucleases: Antiphage defense systems targeting multiple hypermodified phage genomes

Prokaryotes carry clusters of phage defense systems in "defense islands" that have been extensively exploited bioinformatically and experimentally for discovery of immune functions. However, little effort has been dedicated to determining which specific system(s) within defense islands limit lytic phage reproduction in clinical bacterial strains. Here, we employed the CRISPR-based Cascade-Cas3 system to delete defense islands in a Pseudomonas aeruginosa clinical isolate to identify mechanisms of lytic phage antagonism. Deletion of one island in a cystic fibrosis-derived clinical isolate sensitized the strain to phages from the Pbunavirus family, which are commonly used as therapeutics. The causal defense system is a Type IIS restriction endonuclease-like protein (ENDPaCF1), common in Pseudomonads, however it lacks an associated methyltransferase typical Type IIS R-M systems. ENDPaCF1 protects bacteria against phages with hypermodified DNA and is surprisingly agnostic to the specific structure of the modification, which is unlike typical type IV restriction endonucleases. In ENDPaCF1, the endonuclease domain is fused to a catalytically inactive Endonuclease III (iEndoIII), a domain that recognizes non-canonical bases to repair DNA in prokaryotes and eukaryotes. We therefore propose that nucleases containing an iEndoIII domain (END nucleases) can sense diverse DNA hypermodifications. Our findings reveal modularity of the sensing and cleavage domains, as expected of a modification-dependent endonucleases. We further show that some hypermodified phages, including Pbunavirus family members and Wrowclawvirus family (Pa5oct-like) of jumbo phages, encode END nuclease inhibitors that directly bind to the nuclease, likely via the iEndoIII domain. These inhibitors are necessary for Pbunavirus to plaque on clinical isolates and sufficient to enable other hypermodified phages to plaque in the presence of this defense system.