ABSTRACT

Pif-1 proteins are 5′→3′ superfamily 1 (SF1) helicases that in yeast have roles in the maintenance of mitochondrial and nuclear genome stability. The functions and activities of the human enzyme (hPif1) are unclear, but here we describe its DNA binding and DNA remodeling activities. We demonstrate that hPif1 specifically recognizes and unwinds DNA structures resembling putative stalled replication forks. Notably, the enzyme requires both arms of the replication fork-like structure to initiate efficient unwinding of the putative leading replication strand of such substrates. This DNA structure-specific mode of initiation of unwinding is intrinsic to the conserved core helicase domain (hPifHD) that also possesses a strand annealing activity as has been demonstrated for the RecQ family of helicases. The result of hPif1 helicase action at stalled DNA replication forks would generate free 3′ ends and ssDNA that could potentially be used to assist replication restart in conjunction with its strand annealing activity.

A double-strand DNA break was introduced at a specific site within the lacZ gene of plasmid pUC18 using one of several restriction enzymes, and the plasmid exposed to nuclear extracts from human cell lines. Physical rejoining of DNA was monitored by Southern analysis after gel separation, and the fidelity of rejoining by expression of the lacZ gene after bacterial transformation with the treated plasmid. Breaks at the SalI and EcoRI sites were rejoined by extracts to form circular monomers, but the efficiency of rejoining was much higher at the SalI site. Measurement of rejoining at several adjacent sites having different types of termini, consistently showed a range of efficiencies with 5' 4-base greater than 3' 4-base overhangs and 4-base greater than 2-base greater than no overhang. Similar efficiencies were found for nuclear extracts from transformed cell lines, both from a 'normal' individual and an ataxia-telangiectasia (A-T) patient, and from a non-transformed normal cell culture. In contrast at some sites, especially those with a low rejoin efficiency, the fidelity of rejoining was very much lower for the A-T extracts than for normal cell extracts. Mis-rejoining was, however, unrelated to rejoin efficiency at other sites, suggesting that factors such as the exact sequence at the break site on the molecule may also influence the fidelity of rejoining.

DNA molecules carrying a site-specific double-strand break were exposed to nuclear extracts from human cell lines. It was shown previously that breaks could be rejoined correctly by human extracts, but that a proportion of the rejoined molecules had suffered deletions and insertions. The 'mis-rejoined' proportion was higher with cell extracts from an individual with the disorder ataxia-telangiectasia than with normal cell extracts. We now show by sequence analysis that deletions in extract-treated molecules occur exclusively between short direct repeats (2-6 base pairs). A mis-rejoined molecule containing an insertion of 300 bp also had a repeat-based deletion at the same site. A number of different direct repeats are involved; however, some clustering of these occurs especially on the upstream side of the initial breakpoint. These data are most simply interpreted in terms of a model of deletion formation involving single-strand exposure and repair, perhaps with the action of other DNA-metabolising enzymes influencing the frequency with which some repeats are involved.

The nature of mutations occurring in two colorectal carcinoma cell lines deficient in mismatch repair and displaying mutator phenotypes was determined. One of the lines (HCT116) exhibited a higher level of microsatellite instability than the second (DLD-1), although the rate of mutation at the selectable locus encoding the purine salvage enzyme hypoxanthine guanine phosphoribosyl transferase (HPRT) was equally elevated (about 350-450-fold relative to mismatch repair proficient cell lines). Transitions were the major class of mutations in the two mutator lines. In DLD-1 these mutations recurred at several sites that appeared to be hotspots. Frameshifts at a run of six guanine residues in the coding sequence for HPRT constituted 35% of mutations in HCT116. These frameshifts were highly unstable and reverted to wild type at high frequency. Larger deletions were also detected in HCT116. Although these deletions constituted a small proportion of mutations compared with the other types, our data suggest that the rate of deletion is elevated relative to mismatch repair proficient (hMLH1+) cell lines. These observations suggest that the gene(s) altered in DLD-1 may preferentially affect the repair of base mismatches while the alteration(s) in HCT116 may affect the repair of both mismatches and frameshifts.