Exonucleolytic proofreading and DNA mismatch repair (MMR) act in series to maintain high-fidelity DNA replication and avoid mutagenesis. MMR defects elevate the overall mutation rate and are associated with increased cancer incidence. Hypermutable colorectal and endometrial tumors with functional MMR were recently reported to carry amino acid substitutions in the exonuclease domain of DNA polymerase ε (Polε). This created a notion that loss of the proofreading activity of Polε is an initiating cause of some sporadic human cancers. In this study, we identified a somatic P286R substitution in the conserved ExoI motif of Polε in a collection of 52 sporadic colorectal tumor specimens. This change has been repeatedly observed in colorectal and endometrial tumors in previous studies despite many possible ways to inactivate Polε proofreading. To understand the reasons for the recurrent appearance of the P286R variant, we characterized its functional consequences using the yeast model system. An analogous substitution in the yeast Polε produced an unusually strong mutator phenotype exceeding that of proofreading-deficient mutants by two orders of magnitude. This argues that the P286R mutation acts at some level other than loss of exonuclease to elevate cancer risk. Heterozygosity for the variant allele caused a strong mutator effect comparable to that of complete MMR deficiency, providing an explanation for why loss of heterozygosity is not required for the development of Polε-mutant human tumors.

Alterations in the exonuclease domain of DNA polymerase ε (Polε) cause ultramutated tumors. Severe mutator effects of the most common variant, Polε-P286R, modeled in yeast suggested that its pathogenicity involves yet unknown mechanisms beyond simple proofreading deficiency. We show that, despite producing a catastrophic amount of replication errors in vivo, the yeast Polε-P286R analog retains partial exonuclease activity and is more accurate than exonuclease-dead Polε. The major consequence of the arginine substitution is a dramatically increased DNA polymerase activity. This is manifested as a superior ability to copy synthetic and natural templates, extend mismatched primer termini, and bypass secondary DNA structures. We discuss a model wherein the cancer-associated substitution limits access of the 3’-terminus to the exonuclease site and promotes binding at the polymerase site, thus stimulating polymerization. We propose that the ultramutator effect results from increased polymerase activity amplifying the contribution of Polε errors to the genomic mutation rate.

In metazoans, the mechanism by which DNA is synthesized during homologous recombination repair of double-strand breaks is poorly understood. Specifically, the identities of the polymerase(s) that carry out repair synthesis and how they are recruited to repair sites are unclear. Here, we have investigated the roles of several different polymerases during homologous recombination repair in Drosophila melanogaster. Using a gap repair assay, we found that homologous recombination is impaired in Drosophila lacking DNA polymerase zeta and, to a lesser extent, polymerase eta. In addition, the Pol32 protein, part of the polymerase delta complex, is needed for repair requiring extensive synthesis. Loss of Rev1, which interacts with multiple translesion polymerases, results in increased synthesis during gap repair. Together, our findings support a model in which translesion polymerases and the polymerase delta complex compete during homologous recombination repair. In addition, they establish Rev1 as a crucial factor that regulates the extent of repair synthesis.

DNA replication fidelity relies on base selectivity of the replicative DNA polymerases, exonucleolytic proofreading, and postreplicative DNA mismatch repair (MMR). Ultramutated human cancers without MMR defects carry alterations in the exonuclease domain of DNA polymerase ε (Polε). They have been hypothesized to result from defective proofreading. However, modeling of the most common variant, Polε-P286R, in yeast produced an unexpectedly strong mutator effect that exceeded the effect of proofreading deficiency by two orders of magnitude and indicated the involvement of other infidelity factors. The in vivo consequences of many additional Polε mutations reported in cancers remain poorly understood. Here, we genetically characterized 13 cancer-associated Polε variants in the yeast system. Only variants directly altering the DNA binding cleft in the exonuclease domain elevated the mutation rate. Among these, frequently recurring variants were stronger mutators than rare variants, in agreement with the idea that mutator phenotype has a causative role in tumorigenesis. In nearly all cases, the mutator effects exceeded those of an exonuclease-null allele, suggesting that mechanisms distinct from loss of proofreading may drive the genome instability in most ultramutated tumors. All mutator alleles were semidominant, supporting the view that heterozygosity for the polymerase mutations is sufficient for tumor development. In contrast to the DNA binding cleft alterations, peripherally located variants, including a highly recurrent V411L, did not significantly elevate mutagenesis. Finally, the analysis of Polε variants found in MMR-deficient tumors suggested that the majority cause no mutator phenotype alone but some can synergize with MMR deficiency to increase the mutation rate.