Genome-embedded ribonucleotides arrest replicative DNA polymerases (Pols) and cause DNA breaks. Whether mammalian DNA repair Pols efficiently use template ribonucleotides and promote RNA-templated DNA repair synthesis remains unknown. We find that human Polθ reverse transcribes RNA, similar to retroviral reverse transcriptases (RTs). Polθ exhibits a significantly higher velocity and fidelity of deoxyribonucleotide incorporation on RNA versus DNA. The 3.2-Å crystal structure of Polθ on a DNA/RNA primer-template with bound deoxyribonucleotide reveals that the enzyme undergoes a major structural transformation within the thumb subdomain to accommodate A-form DNA/RNA and forms multiple hydrogen bonds with template ribose 2′-hydroxyl groups like retroviral RTs. Last, we find that Polθ promotes RNA-templated DNA repair in mammalian cells. These findings suggest that Polθ was selected to accommodate template ribonucleotides during DNA repair.

Disrupting either the DNA annealing factor RAD52 or the A-family DNA polymerase POLQ can cause synthetic lethality with defects in BRCA1 and BRCA2 , which are tumor suppressors important for homology-directed repair of DNA double-strand breaks (DSBs), and protection of stalled replication forks. A likely mechanism of this synthetic lethality is that RAD52 and/or POLQ are important for backup pathways for DSB repair and/or replication stress responses. The features of DSB repair events that require RAD52 vs. POLQ, and whether combined disruption of these factors causes distinct effects on genome maintenance, have been unclear. Using human U2OS cells, we generated a cell line with POLQ mutations upstream of the polymerase domain, a RAD52 knockout cell line, and a line with combined disruption of both genes. We also examined RAD52 and POLQ using RNA-interference. We find that combined disruption of RAD52 and POLQ causes at least additive hypersensitivity to cisplatin, and a synthetic reduction in replication fork restart velocity. We also examined the influence of RAD52 and POLQ on several DSB repair events. We find that RAD52 is particularly important for repair using ≥ 50 nt repeat sequences that flank the DSB, and that also involve removal of non-homologous sequences flanking the repeats. In contrast, POLQ is important for repair events using 6 nt (but not ≥ 18 nt) of flanking repeats that are at the edge of the break, as well as oligonucleotide microhomology-templated (i.e., 12–20 nt) repair events requiring nascent DNA synthesis. Finally, these factors show key distinctions with BRCA2, regarding effects on DSB repair events and response to stalled replication forks. These findings indicate that RAD52 and POLQ have distinct roles in genome maintenance, including for specific features of DSB repair events, such that combined disruption of these factors may be effective for genotoxin sensitization and/or synthetic lethal strategies.

No abstract