Covalent Trapping of Human DNA Polymerase (cid:1) by the Oxidative DNA Lesion 2-Deoxyribonolactone*

, Oxidized abasic residues in DNA constitute a major class of radiation and oxidative damage. Free radical attack on the nucleotidyl C-1 (cid:1) carbon yields 2-deoxyri-bonolactone (dL) as a significant lesion. Although dL residues are efficiently incised by the main human abasic endonuclease enzyme Ape1, we show here that subsequent excision by human DNA polymerase (cid:1) is im-paired at dL compared with unmodified abasic sites. This inhibition is accompanied by accumulation of a protein-DNA cross-link not observed in reactions of polymerase (cid:1) with unmodified abasic sites, although a similar form can be trapped by reduction with sodium boro-hydride. The formation of the stably cross-linked species with dL depends on the polymerase lysine 72 residue, which forms a Schiff base with the C-1 aldehyde during excision of an unmodified abasic site. In the case of a dL residue, attack on the lactone C-1 by lysine 72 proceeds more slowly and evidently produces an amide linkage, which resists further processing. Consequently dL residues may not be readily repaired by “short-patch” base excision repair but instead function as suicide substrates in the formation of protein-DNA

Mutagenesis and disruption of the cell cycle caused by DNA damage is counteracted by DNA repair systems. In the base excision repair pathway (1)(2)(3), DNA glycosylases eliminate damaged bases to generate abasic (AP) 1 sites, which are also formed in large numbers by spontaneous depurination (2). In either case, AP sites are incised by an AP endonuclease to allow subsequent DNA repair synthesis and excision of the abasic residue. In mammalian cells, incision is carried out by the major AP endonuclease Ape1 protein (also called Apex, Hap1, or Ref1), while the excision step for regular abasic residues is thought to be mainly carried out by DNA polymerase ␤ (Pol␤) using a ␤-elimination mechanism. A distinct branch of the base excision pathway involves strand displacement repair synthesis and excision of the displaced, damaged strand by the FEN1 nuclease (4 -6). Still another variation is potentiated by the initial DNA glycosylase (7) because some of these enzymes carry out a second reaction to cleave at the abasic site by ␤-elimination (1,3). The resulting 3Ј-blocked products must then be removed by an enzyme such as Ape1 before repair synthesis can proceed (1).
Base excision repair acts on a wide variety of deaminated, alkylated, or oxidized bases (2,3). However, oxidative damage to DNA also produces various modified abasic residues that may complicate the repair scenario (1). For example, free radical attack forms strand breaks with fragmentary or oxidized products of deoxyribose; when these are present at the 3Ј terminus, removal by Ape1 may be the rate-limiting repair step (8,9). Oxidized abasic residues without direct strand breakage (10) include 2-deoxypentos-4-ulose residues (a major lesion produced by the antitumor drug bleomycin) and 2-deoxyribonolactone (dL) residues (formed by diverse oxidative agents). 2-Deoxypentos-4-ulose residues are processed rather efficiently in vitro by the central base excision enzymes Ape1 and Pol␤ (11). However, the effectiveness of repair on dL was unknown, and the irreversible cross-linking of dL observed for Escherichia coli endonuclease III (12) suggested that Pol␤ might encounter the same fate. We have shown that Ape1 acts effectively on dL residues in DNA. 2 We show here that excision of Ape1-incised dL residues by Pol␤ is hampered by the formation of a stable covalent cross-link between the Pol␤ and the dL site in the DNA.
Enzymes-Polynucleotide kinase, Klenow fragment DNA polymerase, and uracil-DNA glycosylase were from New England BioLabs, Inc. (Beverly, MA). DNase I was from Ambion, Inc. (Austin, TX). Proteinase K was from Sigma/Aldrich. Recombinant human Ape1 was purified as described previously (13). Hexahistidine-tagged human DNA Pol␤ 3 was Ͼ95% pure as determined following SDS-PAGE and silver staining. For the experiment depicted in Fig. 4B, both wild-type and lysine 72 to alanine (K72A) mutant Pol␤ (14) were provided by Drs. R. Prasad and S. H. Wilson (NIEHS, National Institutes of Health, Research Triangle Park, NC).
Preparation of DNA Substrates-Oligonucleotides containing a 1Ј-tbutylcarbonyl-uridylate residue (indicated by X in the 30-mer 5Ј-GTC-ACGTGCTGCAXACGACGTGCTGAGCCT or the 17-mer 5Ј-XACGAC-GTGCTGAGCCT) were prepared as described previously (15). Other oligonucleotides were purchased from Operon Technologies, Inc. (Alameda, CA). Using standard methods (16), the DNA substrates were ** To whom correspondence should be addressed. Tel.: 617-432-3462; Fax: 617-432-0377; E-mail: bdemple@hsph.harvard.edu. 1 The abbreviations used are: AP, abasic; Pol␤, DNA polymerase ␤; dL, 2-deoxyribonolactone. labeled at the 5Ј-end and hybridized to a complementary strand or hybridized first and labeled at the 3Ј-end. To generate a site-specific dL lesion, 2-10 pmol of radiolabeled duplex DNA containing the modified uridylate were diluted with water to a volume of 30 l, transferred to a glass tube, and subjected to photolysis in a Photochemical Reactor (RPR-100 from Rayonet Corp., Branford, CT) at 350 nm, 9200 microwatts/cm 2 for 150 min (15). This material was then used immediately in enzyme reactions. To generate unmodified abasic sites, radiolabeled duplex DNA containing a uracil residue was treated with 2-3 units of uracil-DNA glycosylase for 60 min and used at once.
Assays-Reactions (12.5 l in 50 mM HEPES-KOH, pH 7.5, 5% glycerol, 8 mM MgCl 2 , 0.5 mM dithiothreitol, 0.1 mg/ml bovine serum albumin) contained 40 -50 nM radiolabeled DNA substrate and the enzyme concentrations indicated in the figure legends. After incubation at 25°C for the specified times, samples were either directly analyzed by electrophoresis (see below) or subjected to further chemical or enzymatic treatment. Subsequent incubations with DNase I or proteinase K were performed at 37°C prior to SDS-PAGE. Treatment with 48 -300 mM NaBH 4 was carried out for Ͼ30 min at 25°C followed by desalting where indicated using Microcon filters and analysis by SDS-PAGE.
Analysis-For analysis on 19% polyacrylamide, 8 M urea gels (Fig. 1), samples were mixed 1:1 with a 2-fold concentrated formamide loading buffer (16). Samples for analysis by SDS-PAGE on 8% polyacrylamide gels (Figs. 2-4) were mixed 5:1 with 6-fold concentrated loading buffer containing SDS. Samples were heated for 4 min at 90°C prior to electrophoresis. Following electrophoresis, SDS-PAGE gels were silverstained and dried, while urea-containing gels were dried immediately. The dried gels were analyzed using a Molecular Imager System (model GS-525, Bio-Rad), and images were obtained by autoradiography.

RESULTS
The selective generation of dL via a photosensitive precursor enables analysis of the repair enzymology of this key oxidative lesion. Using this approach, a recent study showed that bacterial endonuclease III becomes cross-linked to dL during its ␤-elimination cleavage reaction (12). We have shown that the hydrolytic endonuclease Ape1 cleaves dL in DNA effectively. 2 Thus, although human cells contain enzymes similar to endonuclease III (2), the abundant Ape1 protein would likely convert much of dL in vivo to a cleaved form available to the downstream repair enzyme Pol␤. Since excision of abasic sites by Pol␤ occurs through a ␤-elimination mechanism, the process might lead to cross-linking of the polymerase to dL during attempted repair.
Photoconversion of the precursor nucleotide to dL generated a site that was sensitive to Ape1 protein. Ape1 treatment yielded an oligonucleotide bearing a dL-5-phosphate residue at its 5Ј terminus, which thus had slower mobility (Fig. 1, lane 2) than the 5Ј-phosphate product (Fig. 1, lane 1). The latter product accompanies the generation of 3Ј-phosphates by the same C-1Ј radical that produces dL (15) and may also reflect instability of dL during electrophoresis (e.g. chemically induced ␤-elimination (15)). Treatment with piperidine to cleave at dL generated an adduct of slower mobility (Fig. 1, lane 3) due to addition of the amine to the dL aldehyde as previously noted (17). This procedure also demonstrated that the photoconversion to dL and 3Ј-phosphates was only partial (the top band in Fig. 1 represents noncleaved oligonucleotide).
A 25-min incubation of the Ape1-cleaved dL product with increasing amounts of Pol␤ diminished the amount of the 5Јterminal dL substrate only partially (Fig. 1, lanes 4 -6). In contrast, a 10-min incubation with Pol␤ sufficed to completely remove the unmodified 5Ј-terminal deoxyribose residues produced by Ape1 acting on a glycosylase-generated abasic site (Fig. 1, lanes 9 and 10). Thus, Pol␤ was clearly less effective against dL than at conventional abasic sites, although the nonenzymatic cleavage products noted above interfered with accurate quantitation using PAGE.
To test the possibility that the poor activity of Pol␤ in dL excision might be related to the cross-linking reaction proposed above, we turned to SDS-PAGE. This analysis revealed other products with Pol␤ that did not correspond to DNA trimmed of the dL residue. These experiments (Fig. 2) demonstrated the time-dependent formation of a major radiolabeled species with electrophoretic mobility (M r ϳ45,000) greatly shifted from the DNA substrate (bottom of the gel in the figure) and significantly slower than free Pol␤ (indicated just below the M r 40,000 marker; silver-stained gel not shown). The formation of this new species depended on the prior reaction with Ape1 (Fig. 2). The amount of the M r 45,000 product also increased as a function of the Pol␤ concentration (data not shown). These results are consistent with the formation of the hypothesized Pol␤-dL cross-link.
Smaller amounts of a product of M r Ͼ50,000 were sometimes observed (Fig. 2). Results to be detailed elsewhere 3 indicate that this secondary species may represent the same product as the M r 45,000 species but retaining the complementary DNA strand.
The nature of the hypothetical Pol␤-dL cross-link was addressed by selective digestion of either the protein or the DNA component. A 32 P-labeled 5Ј-terminal dL in a 17-mer oligonucleotide was annealed to the complementary and upstream strands, and reaction of this substrate with Pol␤ generated a substantial amount of the M r 45,000 complex (Fig. 3). Treatment with DNase I did not eliminate the extra band but shifted its mobility by ϳ5 kDa, consistent with the removal of ϳ15 nucleotides. In the same sample, the residual unlinked DNA (Fig. 3, bottom of gel) was completely destroyed. Treatment with proteinase K eliminated the M r 45,000 band (and the Pol␤ and bovine serum albumin bands observed on silver-stained gels) but did not remove the residual DNA. These features are consistent with the formation of a cross-link in which labeled dL was directly bonded to the Pol␤ protein and resistant to nuclease digestion.
Pol␤ excises unmodified, 5Ј-terminal abasic residues by the SCHEME 1   FIG. 1. Pol␤ poorly excises 5-terminal dL. Duplex 3Ј-end-labeled DNA substrates (50 nM) containing either a dL site (lanes 1-7) or an unmodified abasic residue (lanes 8 -11) were incubated without or with 5 nM Ape1 for 10 min as indicated. Pol␤ was then added in the amounts shown in the figure, and the incubation was continued for 25 min (lanes 1-7) or 10 min (lanes 8 -11). For the sample in lane 3, after Ape1 incision piperidine (100 mM) was added followed by a ϳ25-min incubation at 90°C. All reactions were terminated with 300 mM NaBH 4 and then desalted before analysis on a 19% polyacrylamide/urea DNA sequencing gel. To the left of the autoradiogram are diagrams depicting the fulllength oligonucleotide containing the abasic residue (top), the Ape1incised product (middle), and the fragment lacking the abasic residue (bottom); the asterisks indicate the 3Ј-label. Note that the portion of the autoradiogram between the full-length oligonucleotides and the cleaved products has been eliminated to save space. pdRp, 2-deoxyribonolactone-5-phosphate or 2-deoxyribose-5-phosphate; p, phosphate; Reg., regular. transient formation of a Schiff base, which can be trapped by reduction to a secondary amine (14,18). As expected, addition of fresh NaBH 4 to a Pol␤ reaction with unmodified, Ape1cleaved abasic sites produced bands of M r ϳ45,000 and (more weakly) 52,000 (Fig. 4A, lane 3). These bands are of the same mobility and relative intensity as those formed by Pol␤ acting on an Ape1-cleaved dL substrate in the absence of NaBH 4 (Fig.  4A, lane 1).
Cross-link formation of Pol␤ with dL depended on the known mechanism of the enzyme. In deoxyribose excision by Pol␤, the ⑀-amino of lysine 72 attacks C-1 during Schiff base formation (14,18). We tested parallel preparations of wild-type Pol␤ and a K72A mutant in reactions with Ape1-cleaved dL. These experiments (Fig. 4B) showed that the M r 45,000 and 52,000 bands formed by the wild-type protein were not generated by the K72A protein. The K72A protein retained full DNA polymerase activity and therefore was not nonspecifically impaired. Conversely addition of dNTPs did not prevent formation of the dL cross-link with wild-type Pol␤ (data not shown).

DISCUSSION
The data presented here demonstrate a new complication upon processing an oxidative DNA lesion (12): formation of a covalent DNA-protein cross-link with Pol␤. Consistent with this conclusion, phosphate label associated with the abasic dL residue itself became resistant to nuclease digestion upon reaction with Pol␤. In its apparent irreversibility and dependence on the active-site nucleophile (lysine 72) of Pol␤, this crosslinking corresponds to a mechanism-based suicide inhibition of the enzyme. Given the biological role of the abasic excision activity of Pol␤ (19), this inhibition may have important cellular consequences. The imino intermediate (Schiff base) formed by Pol␤ during excision of abasic residues (20,21) is an Achilles heel in the repair of oxidative DNA damage since dL residues can covalently trap the enzyme. The nucleophilic attack by the Pol␤ lysine 72 on the lactone carbonyl of dL is typical of the reaction of alkylamines with esters and would be enhanced by the release of strain upon cleavage of the lactone. This trapping generates an amide linkage that resists further processing by Pol␤ (see Scheme 1 under "Results").
The newly discovered DNA polymerases and also have intrinsic lyase activity for 5Ј-abasic residues (22,23) and might be similarly trapped by dL. This lesion therefore constitutes a triple threat: the initial loss of genetic information and replication-blocking features of the abasic lesion; the blocking of DNA repair activities; and the formation of another type of damage, the protein-DNA cross-link.
A repair mechanism for the Pol␤-dL cross-link is unknown, but it would almost certainly be distinct from the diesterase that reverses the protein-DNA phosphodiester of aborted topoisomerase II complexes (24). Although the base excision repairassociated FEN1 endonuclease can excise displaced DNA "flaps" and generally enhances the repair of radiation-induced abasic sites in cell-free extracts (6), we have not detected cleavage of Pol␤-dL cross-linked molecules even at high levels of FEN1 protein (data not shown). This observation is consistent with the inability of FEN1 to act on flaps containing a protein bound within the displaced flap (25,26). Other enzymes must be tested for their ability to dispose of this unusual species.  1 and 2) 48 mM NaBH 4 in a 2.5-h incubation. All reactions were terminated by adding loading buffer. The samples were directly analyzed by SDS-PAGE (8%) and autoradiography. B, dependence of the cross-link with dL on lysine 72 of Pol␤. Where indicated, the dL substrate was incubated with Ape1 as described for panel A. Wild-type (lanes 3 and 4) or K72A mutant Pol␤ (lanes 5 and 6) (600 nM) was then added, and the incubation continued for 30 min followed by SDS-PAGE and autoradiography. WT, wild type; oxi., oxidized; Reg., regular.