Long-patch Base Excision DNA Repair of 2-Deoxyribonolactone Prevents the Formation of DNA-Protein Cross-links with DNA Polymerase β*

Oxidized abasic sites are a major form of DNA damage induced by free radical attack and deoxyribose oxidation. 2-Deoxyribonolactone (dL) is a C1′-oxidized abasic site implicated in DNA strand breakage, mutagenesis, and formation of covalent DNA-protein cross-links (DPCs) with repair enzymes such as DNA polymerase β (polβ). We show here that mammalian cell-free extracts incubated with Ape1-incised dL substrates under non-repair conditions give rise to DPCs, with a major species dependent on the presence of polβ. DPC formation was much less under repair than non-repair conditions, with extracts of either polβ-proficient or -deficient cells. Partial base excision DNA repair (BER) reconstituted with purified enzymes demonstrated that Flap endonuclease 1 (FEN1) efficiently excises a displaced oligonucleotide containing a 5′-terminal dL residue, as would be produced during long-patch (multinucleotide) BER. Simultaneous monitoring of dL repair and dL-mediated DPC formation demonstrated that removal of the dL residue through the combined action of strand-displacement DNA synthesis by polβ and excision by FEN1 markedly diminished DPC formation with the polymerase. Analysis of the patch size distribution associated with DNA repair synthesis in cell-free extracts showed that the processing of dL residues is associated with the synthesis of ≥2 nucleotides, compared with predominantly single nucleotide replacement for regular abasic sites. Our observations reveal a cellular repair process for dL lesions that avoids formation of DPCs that would threaten the integrity of DNA and perhaps cell viability.

Cellular DNA is under continuous assault by DNA-damaging agents of both endogenous and exogenous sources (1,2). Loss of DNA bases generates abasic (AP) 3 sites, perhaps the most common DNA lesions, which can be mutagenic and cytotoxic if not repaired appropriately (3,4). AP sites are formed by spontaneous hydrolysis of the N-glycosylic bonds or through the removal of damaged or mismatched bases by various DNA glycosylases (5,6). In either case, the resulting AP sites are repaired by the base excision DNA repair (BER) pathway. In mammalian cells, the major AP endonuclease, Ape1 (also called Apex, HAP1, or Ref-1), incises the 5Ј-phosphodiester bond of the AP site to generate a BER intermediate that contains a single-strand break bracketed by 3Ј-hydroxyl and 5Ј-deoxyribose-5-phosphate (5Ј-dRP) termini.
Subsequent steps may follow either of two distinct BER sub-pathways that replace either a single nucleotide (short-patch BER) or multiple nucleotides (long-patch BER). In short-patch BER, most 5Ј-dRP excision is attributable to DNA polymerase ␤ (pol␤), specifically the dRP lyase activity of its amino-terminal 8-kDa domain (7,8). The DNA polymerase activity of pol␤ is also the major enzyme for DNA synthesis during short-patch BER, as demonstrated in vitro by using purified enzymes and cell extracts from wild-type and pol␤-null mouse embryonic fibroblasts (MEFs) (9,10). The long-patch BER pathway involves strand displacement repair synthesis of at least two nucleotides and the excision of the 5Ј-dRP residue as part of a flap oligonucleotide released by the FEN1 nuclease (11)(12)(13).
pol␤ may initiate strand displacement DNA synthesis, but involvement in long-patch BER of other DNA polymerases, such as pol␦ and pol⑀, has been suggested (14 -16). A reconstituted enzyme system was developed for long-patch BER of a reduced AP site utilizing purified Ape1, pol␤, pol␦, proliferating cell nuclear antigen (PCNA), FEN1, and DNA ligase I, where pol␦ substituted for pol␤ when PCNA was present in the reaction (12). PCNA-dependent long-patch BER has also been demonstrated in extracts of pol␤-deficient MEFs, but it was shown to be dependent on using a circular DNA substrate (9,17). An additional variation of BER has been suggested, because some bifunctional DNA glycosylases associated with AP lyase activity can carry out incision of AP sites by ␤-elimination, resulting in 3Ј-blocked ends that must be removed by enzymes such as Ape1 and polynucleotide kinase prior to repair DNA synthesis (18,19).
Oxidative damage to DNA, mediated by free-radicals and reactive oxygen species, produces structurally distinct AP sites that are handled differently by BER enzymes (18). Such lesions include 2-deoxyribonolactone (dL), a C1Ј-oxidized AP site, which has been reported to be introduced into DNA by numerous genotoxic agents including longwave UV and ionizing radiation, organometallic oxidants, copper-phenanthroline chemical nuclease, and the chromophore of the antitumor agent neocarzinostatin (20 -22). Little was known about repair of dL until the recent development of synthetic oligonucleotides that yield a site-specific dL through a photosensitive nucleotide analog (23)(24)(25)(26). Initial investigation of the reaction of dL with Escherichia coli endonuclease III, a DNA glycosylase with an associated AP lyase activity, revealed formation of a stable DNA-protein cross-link (DPC) between the dL lesion and the catalytic lysine of the enzyme (27). However, the AP endonuclease activity of E. coli exonuclease III (Exo III) and endonuclease IV (Endo IV) process dL lesions efficiently (28), and such 5Ј-incision would prevent AP lyase enzymes from acting on dL. Consistent with these observations, the dL-induced mutation frequency was highly elevated in E. coli-deficient in both major AP endonucleases (29). The major human AP endonuclease Ape1 also incises dL residues rather efficiently, leaving a 5Ј-terminal oxidized AP residue (30,31). However, an Ape1-incised dL residue forms stable DPC with pol␤, the next BER enzyme, dependent on its catalytic lysine (residue 72) for its dRP lyase activity (31). This DPC linkage is evidently a stable amide bond between the lysine-72 ⑀-nitrogen and the dL C1Ј-carbonyl (18,31).
Although in vitro studies with purified E. coli proteins and mammalian pol␤ have provided useful information about chemical mechanisms involved in dL-mediated DPC (31,32), the biological significance of such DPC formation and the cellular repair of dL lesions remains unclear. In the present study, we utilized mammalian cell-free extracts to determine whether different cellular proteins are prone to dL-mediated DPC formation. To explore possible repair of dL lesions, and its correlation with DPC formation, the complete repair of closed circular DNA substrate with a defined dL residue was analyzed for evidence of both sub-pathways of BER. We found that pol␤ is indeed the major cellular protein associated with formation of dL-specific DPC, but that DPC formation can be largely prevented by rapid removal of dL through mechanisms of long-patch BER.

EXPERIMENTAL PROCEDURES
Materials-Oligonucleotides containing a dL precursor residue (1Јt-butylcarbonyluridylate; tBU) (31) were provided by Dr. M. Greenberg (Johns Hopkins University, MD). Other oligonucleotides were synthesized and high-performance liquid chromatography purified by Operon Technologies. pGEM-3Zf (ϩ) plasmid and helper phage (R408) were obtained from Promega. Radionucleotides were from PerkinElmer Life Sciences. HeLa and pol␤-proficient (MB16tsA, clone 1B5) or -deficient (MB19tsA, clone 2B2) SV-40 immortalized MEF cell lines were obtained from American Type Culture Collection. T4 polynucleotide kinase, T4 DNA polymerase, T4 DNA ligase, Klenow fragment DNA polymerase, E. coli Exo III, and all restriction enzymes were from New England Biolabs. Proteinase K was from Sigma/Aldrich. Recombinant human Ape1 and E. coli Endo IV were purified as described previously (33). Human pol␤ was kindly provided by Drs. R. Prasad and S. H. Wilson (NIEHS, National Institutes of Health). Goat anti-Po␤ polyclonal antibody was from Santa Cruz Biotechnology. Recombinant human FEN1 was expressed and purified to apparent homogeneity as previously described (34) P]dCTP using the exonuclease-free Klenow fragment of DNA polymerase I. Closed circular pGEM-3Zf (ϩ) plasmid DNA containing a site-specific lesion was constructed as previously described (36), except that different primers were employed. Briefly, single-stranded pGEM plasmid (ϩ) DNA was produced with the aid of helper phage (R408) and purified by the CTAB DNA precipitation method (36). Each upstream primer (Up-T, Up-U, and Up-tBU) was 5Ј-phosphorylated and annealed to single-stranded pGEM DNA. During the preparation of 32 P-labeled pGEM DNA substrates that were employed in patch size distribution analysis, each upstream primer was 5Ј-end-labeled using T4 polynucleotide kinase and a molar excess of [␣-32 P]ATP. To construct 32 P-labeled pGEM DNA substrates used for DPC detection, 5Ј-32 P-labeled downstream primer (Down), in addition to the upstream primer Up-tBU, was also hybridized to single-stranded pGEM DNA. Each primed template was subjected to a primer extension reaction followed by ligation with T4 DNA ligase. Covalently closed circular duplex DNA for either unlabeled or 32 P-labeled pGEM (T/A), (U/A), and (tBU/A) were isolated by cesium chloride gradient centrifugation as described previously (36). Typically, isolated plasmid DNA was found to contain Ͼ98% form I molecules as determined by 1% agarose gel.
To generate a site-specific dL residue in DNA substrates, 2-5 pmol of DNA duplex containing a tBU residue was subjected to the photolysis reaction as previously described (31). The efficiency of the photo-conversion was typically Ͼ90% when monitored by dL-specific DNA cleavage using hot-alkali treatment and subsequent analysis of DNA by denaturing polyacrylamide gel electrophoresis. Where indicated, DNA containing a dL lesion was subjected to the treatment with catalytic amounts of either Endo IV or Ape1 to generate a site-specific dL-5phosphate (5Ј-dL) residue. Oligonucleotide and plasmid DNA containing a uracil residue was treated with a catalytic amount (1-10 units) of a The underlined nucleotides U and tBU denote uridine and 1Ј-t-butylcarbonyluridylated residues, the precursor for dL, respectively.

Long-patch BER of dL Prevents Cross-link Formation
uracil-DNA glycosylase to prepare DNA substrates containing a sitespecific AP site. Preparation of Cell-free Extracts-HeLa and MEF cell-free extracts were prepared from confluent cells as previously described (17), and dialyzed extensively against 20 mM Hepes-KOH (PH 7.6), 100 mM NaCl, 1 mM dithiothreitol, 0.1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10% (v/v) glycerol, and protease inhibitor mixture (Sigma-Aldrich). Protein concentrations of the cell-free extracts were determined using a Bio-Rad Protein Assay reagent. The levels of pol␤ in cell-free extracts were monitored by Western blot analysis as described previously (37).
Analysis of Cross-linking Reactions-Standard cross-linking reactions contained 50 mM Hepes-KOH (pH 7.5), 20 mM NaCl, 0.5 mM dithiothreitol, 2 mM EDTA, 5% (v/v) glycerol, 0.1 mg/ml bovine serum albumin, 10 nM 3Ј-32 P-labeled 31-mer DNA substrate, and protein concentrations as indicated in the figure legends. Following incubation at 30°C for the specified times, reactions were terminated by the addition of SDS-PAGE loading buffer (31) and heating at 100°C for 5 min. DPCs and free DNA were resolved by 8% SDS-PAGE, and 32 P radioactivity associated with DNA-protein cross-links was quantified using a Phos-phorImager and the ImageQuant program (Amersham Biosciences).
In Vitro DNA Repair Assay with Purified Enzymes-A partial reconstitution of long-patch BER was performed in reactions containing 50 mM Hepes-KOH (pH 7.5), 50 mM NaCl, 0.5 mM dithiothreitol, 8 mM MgCl 2 , 5% (v/v) glycerol, 0.1 mg/ml bovine serum albumin, 10 nM 3Ј-end 32 P-labeled 31-mer DNA substrate, and various BER enzymes as indicated. Where indicated, 50 M each of dATP, dCTP, and dGTP were included to permit only limited repair related strand displacement DNA synthesis. For the FEN1 assay, reactions also contained the indicated concentrations of FEN1 enzyme and duplex DNA substrates bearing unannealed 5Ј-flap structures. After incubation at 30°C for the specified times, reactions were terminated by addition of formamide loading buffer (31) and heating at 100°C for 3 min. 32 P-DNA products were resolved by 15% polyacrylamide, 7 M urea gel electrophoresis.
Analysis of BER Reaction Products-BER reactions were performed at 30°C using a standard reaction buffer containing 100 mM Hepes-KOH (pH 7.5), 50 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 0.1 mM EDTA, 2 mM ATP, 0.5 mM ␤-NAD, 20 M each of dNTPs, 5 mM phosphocreatine, 200 units/ml phosphocreatine kinase, 5-10 g/ml of the appropriate pGEM DNA substrate, and cell-free extract as indicated in the figure legends. To monitor DPC formation with unrepaired dL residues, a 32 P-labeled pGEM (dL/A) substrate was employed in the reaction. Following incubation for the specified times, reaction products were digested with 10 units of BamHI and HindIII for 1 h at 37°C, and then analyzed by 8% SDS-PAGE. To distinguish between either the short-or long-patch BER pathways, a modified restriction analysis method was used that differed slightly from that previously described (38). Briefly, unlabeled pGEM DNA substrates were incubated with cell-free extract under standard BER conditions, except reactions contained, as indicated, 20 M of either 32 P-dTTP or a combined mixture of dTTP and 32 P-dCTP. After incubation for 30 min at 30°C, reactions were terminated, and DNA products were isolated as previously described (36). Samples (100 ng) were removed for digestion with various restriction endonucleases, as indicated, for 1 h at 37°C. The resulting 32 P-DNA fragments were resolved by electrophoresis on 15% polyacrylamide, 7 M urea gel.
Determination of Repair Patch Size Distribution-Standard BER reactions (100 l) were prepared as described above except that 2Ј-deoxyribonucleoside ␣-thiotriphosphate (dNTP[␣S]) were substituted for normal dNTPs. pGEM DNA substrates contained a 32 P-labeled dAMP residue located 11 nucleotides upstream of the target residue. After BER reactions were performed, DNA products were isolated as described above, and 100 ng of each DNA sample was digested with 10 units of HindIII for 30 min at 37°C. The reactions were terminated at 70°C for 10 min, and samples were incubated with 100 units of Exo III for 1 h at 37°C. Exonuclease III was inactivated at 70°C for 10 min, and the 32 P-DNA was then cleaved with BamHI for 30 min at 37°C. The resulting 32 P-DNA fragments were resolved by 15% polyacrylamide, 7 M urea gel electrophoresis, and quantified using a PhosphorImager and Image-Quant software (Amersham Biosciences).

dL-mediated DPC Formation in Cell-free Extracts-
We showed previously that Ape1 efficiently incises DNA at dL lesions to yield 5Ј-terminal dL-5-phosphate (5Ј-dL) residues (30,31). During attempted excision of a 5Ј-dL residue, pol␤ becomes covalently cross-linked to the oxidized lesion (31). To understand the biological significance of such DPC formation, we have now conducted in vitro assays with cell-free extracts and determined major cellular proteins involved in dL-mediated DPCs. A 32 P-labeled oligonucleotide DNA substrate bearing a sitespecific dL residue (formed from the precursor by UVA exposure) was incubated with a HeLa cell-free extract, and the profile of DNA and DPC products was examined by SDS-PAGE (Fig. 1). Initial reactions were conducted in the presence of EDTA to suppress nuclease activity; AP lyases and dRP lyases are usually active under these conditions and might form DPC (27,31). To allow incision of dL residues under these conditions, we added catalytic amounts of the EDTA-resistant enzyme E. coli Endo IV (39). Incubation of the incised dL substrate with purified pol␤ produced the expected DPC product (Fig. 1, lane 3). Incubation with the cell-free extract revealed several new labeled bands migrating more slowly than the DNA substrate (Fig. 1, lane 4). These new species were not produced with DNA containing the unconverted precursor residue (Fig. 1, lane 2), and they were completely eliminated upon posttreatment with proteinase K (Fig. 1, lane 6). These observations indicate the formation of dL-specific DPCs.
Heat treatment of the extracts was used to test for those proteins that required a folded structure for the formation of cross-links. At least one of the DPCs with M r ϳ20,000 was heat-insensitive, but the major DPC species with M r ϳ45,000 was lost when the cell extract was pretreated by heating (Fig. 1, lane 7). A similar profile of DPCs was also observed when the dL-DNA was pre-incised by a catalytic amount of Ape1 and incubated with cell-free extract in the presence of EDTA (data not shown). Thus, the formation of the major DPC appeared to be mediated by a Mg 2ϩ -independent, heat-sensitive HeLa protein.
The M r of the major HeLa DPC product formed with AP endonuclease-incised dL was very close to that formed with purified pol␤ (Fig. 1. compare lanes 3 and 4; see also supplemental Fig. S1). To address directly whether the major protein for dL-mediated DPC in mammalian cells was indeed pol␤, experiments were performed with cell-free extracts from POLB ϩ/ϩ or POLB Ϫ/Ϫ MEFs. The M r 45,000 DPC product initially appeared after a 10-min incubation with a POLB ϩ/ϩ cell extract and increased in a time-dependent manner ( Fig. 2A, lanes 2-6). Extracts of POLB Ϫ/Ϫ MEFs did not produce this DPC product after incubations as long as 120 min (Fig. 2A, lane 8), which indicates that pol␤ was the major protein forming DPCs with 5Ј-dL residues. Western blot analysis confirmed the POLB genotypes and indicated that the extract of wild-type cells contained ϳ0.2 pmol of pol␤ per cross-linking reaction (Fig. 2B). Thus, DPC formation with endogenous pol␤ extracts was 8-fold less efficient than the reaction with purified pol␤ (Fig. 2A,  lane 1 versus 6). However, both the POLB ϩ/ϩ and POLB Ϫ/Ϫ cell-free extracts gave a time-dependent increase of minor DPC species, which appeared as additional faint bands on the gel ( Fig. 2A, lanes 2-8; indicated by arrows). These results imply that DPCs may be formed with various mammalian proteins, although pol␤ is the main activity responsible for DPC formation with AP endonuclease-incised dL.

Reduced dL-mediated DPC Formation under DNA Repair Conditions-
Although the removal of 5Ј-dRP residues by pol␤ is a key step in accomplishing short-patch BER (17,40), the formation of DPC in reactions with 5Ј-dL is problematic. In principle, long-patch BER (17,40) would avoid this problem. We therefore examined reactions of dL with cellfree extracts under the conditions that would allow long-patch BER.
To protect the substrate from nonspecific degradation and to promote the assembly of long-patch BER proteins (38,41), a circular plasmid DNA substrate was constructed containing a site-specific 5Ј-dL residue (Fig. 3). This DNA substrate was first incubated with MEF cellfree extracts under the DNA repair conditions in the presence of dNTP, MgCl 2 , and an ATP regenerating system (17). After various incubation times, the DNA was subjected to the cross-linking reaction with an excess amount of purified pol␤ to detect the level of 5Ј-dL residues that remained available to form DPC. The analysis showed that DPC with pol␤ decreased as a function of the preincubation time ( Fig. 3B; see also supplemental Fig. S2). Quantification showed that DPC formation was reduced to 50% after a 10-min preincubation, to 25% after 20 min, and effectively eliminated after a 60-min preincubation with an MEF cellfree extract (Fig. 3C).
The elimination of dL residues cross-linkable to pol␤ suggested that dL modification, removal, or competition with other proteins to form DPC was taking place in the extracts. Examination of reactions of the dL plasmid substrate with increasing amounts of either POLB ϩ/ϩ or POLB Ϫ/Ϫ extracts revealed that no detectable DPC were being formed (Fig. 4A, lanes 2-5 and 12-15, respectively; see also supplemental Fig.  1 and 2, respectively) are shown, along with 0, 0.13, 0.5, 2, and 8 pmol of purified pol␤ (lanes 3-7). S3). Minor bands with slower mobility were also observed, but these occurred with both POLB ϩ/ϩ or POLB Ϫ/Ϫ extracts (Fig. 4A), and so were not related to the dL-mediated DPC with pol␤. However, supplementation of either extract with a large amount of purified pol␤ showed the expected DPC with the polymerase (Fig. 4A, lanes 6 and 16 versus 7  and 17), the formation of which diminished with an increasing proportion of added extract (Fig. 4A, lanes 6 -10 and 16 -20; quantification in Fig. 4B). Considering that the level of DPC with pol␤ was reduced without a compensating increase in DPC with other proteins (Figs. 3C and 4A), the observed decrease in pol␤-DPC formation most likely resulted from repair (or modification) of dL residues, rather than a simple competition between pol␤ and other proteins to form DPC. Collectively, the results imply the existence of dL-repair mechanisms that prevent the cross-linking of pol␤ with dL.

Inhibition of dL-mediated DPC by in Vitro Reconstitution of Longpatch BER-
To determine whether the processing of dL residue can occur by long-patch BER to prevent DPC formation, we examined the profiles of both DNA processing and DPC formation in the reactions reconstituted with various repair components involved in long-patch BER (Fig. 5). Untreated 3Ј-32 P-labeled dL DNA substrate displayed heterogeneous mobility on a sequencing gel, probably due to cleavage at the heat-labile dL site during gel electrophoresis (Fig. 5A, lane 1). Treatment of the DNA substrate with Ape1 converted the majority of the DNA substrate to the 18-mer DNA cleavage product, consistent with incision at the 5Ј site of the dL residue by Ape1 (30,31). Additional treatments with pol␤ and FEN1, or both, did not mediate further processing of DNA in the absence of dNTPs (Fig. 5A, lanes 3-5). However, the inclusion of dATP, dCTP, and dGTP to permit limited DNA repair synthesis (only 7 nucleotides, specified by the substrate DNA sequence) in a reaction containing Ape1, pol␤, and FEN1 produced a distinct DNA product of 11 nucleotides (Fig. 5A, lane 10). The generation of the 11-mer is consistent with strand displacement DNA synthesis of 7 nucleotides by the polymerase, followed by removal of the displaced DNA flap by FEN1. Parallel analysis of the same reaction mixtures by SDS-PAGE displayed the profile of DPC formation, which was dependent on both Ape1 and pol␤ (Fig. 5B, lanes 4 and 5). The generation of DPC was markedly reduced when the reaction allowed the combined action of repair synthesis by pol␤ and flap excision by FEN1 (Fig. 5, A  and B, lane 10). pol␤-mediated repair synthesis alone did not block dL-mediated DPC formation (Fig. 5B, lane 9), which indicates the importance of FEN1-mediated removal of the dL-containing fragment in preventing cross-linking.
To test directly whether FEN1 can indeed excise a 5Ј-dL flap, the endonuclease activity of FEN1 was assayed with a pre-assembled flap DNA structure. Processing of flaps by FEN1 occurred about equally efficiently for 5Ј-dL, 5Ј-dRP, or the dL-precursor residue (Fig. 6), consistent with published observations that FEN1 tolerates a variety of flap modifications (42). Overall, these results demonstrated that repair of dL by long-patch BER averts the formation of dL-mediated DPC.
Altered Patch Size Distribution Associated with dL-mediated DNA Repair-To examine how dL lesions affect the mode of DNA repair synthesis, we analyzed the repair patch size distribution for dL lesions compared with regular AP sites. Briefly, the approach (36) relies on the incorporation of 2-deoxyribonucleoside-␣-phosphorothioates during DNA synthesis, which renders the repaired DNA region resistant to subsequent digestion by E. coli Exo III. As illustrated in Fig. 7A, the pGEM-derived plasmid DNA substrates utilized in this experiment contained the target dL or AP site on the (Ϫ)-strand 17 nucleotides 5Ј and 12 nucleotides 3Ј, respectively, to unique HindIII and BamHI sites. A 32 P-dAMP label was introduced 5Ј to the target, and the repair patch size distribution was evaluated by determining the length of 32 P-DNA   1-5) or the presence (lanes 6 -10) of a dNTP mix excluding dTTP. The asterisk indicates the position of the radiolabel, X denotes the dL residue, and the underlined nucleotide sequence represents the DNA segment that would be displaced by the incorporation of 7 nucleotides with dCMP, dAMP, and dGMP. After the incubation, one-half of each reaction mixture was analyzed on a 15% polyacrylamide/urea DNA sequencing gel. 3Ј-32 P-labeled 31-mer DNA substrate containing U instead of a dL residue was either untreated (lane U) or incubated with Ung and Ape1 (lane T) to generate reference standards. B, the remainder of each reaction mixture was analyzed by SDS-PAGE, and the 32 P-labeled DNA bands were visualized using a Phos-phorImager. The band positions of the M r markers are indicated at the right. NOVEMBER 25, 2005 • VOLUME 280 • NUMBER 47 fragments produced after sequential treatment of the reaction products with HindIII, Exo III, and BamHI. Because 3Ј-to-5Ј digestion by Exo III terminates at the first phosphorothioate linkage (43), the size of the remaining 32 P-DNA fragments indicates the 3Ј-boundary of DNA repair synthesis tracts.

Long-patch BER of dL Prevents Cross-link Formation
Patch size analysis was conducted for both POLB ϩ/ϩ and POLB Ϫ/Ϫ extracts. Reactions with dL and AP sites yielded discrete DNA products (Fig. 7B), which were used to define the repair patch. These results were quantified, with the repair patch sizes expressed as one nucleotide (short-patch) or longer than one nucleotide (long-patch) BER (Fig. 7C). However, the same amount of Exo III completely digested mock treated DNA (Fig. 7B, lane T), which indicates that the incorporation of 2-deoxyribonucleoside-␣-phosphorothioates depends upon repair DNA synthesis by the cell-free extracts. Reactions with DNA containing a dL-precursor residue did not produce any detectable labeled product in both types of cell extract (Fig. 7B, lanes 5 and 10), which shows that the detected repair patches are dependent on BER.
As reported previously (12,16), insertion of a single nucleotide was the predominant product during repair of AP sites in a pol␤-proficient extract (Fig. 7B, lane 1), corresponding to 35% of the total (Fig. 7C). In contrast, dL-mediated repair in the POLB ϩ/ϩ extract produced patch sizes of Ն2 nucleotides and very little single-nucleotide product (Fig. 7B,  lane 3), Ͻ10% of the total (Fig. 7C). In the pol␤-deficient cell extract, the repair patch size distribution for AP sites was shifted to longer products (Fig. 7B, lane 6), such that single-nucleotide BER constituted Ͻ10% of the total (Fig. 7C). For dL repair, pol␤-deficiency also shifted repair toward products that were shorter than those found when pol␤ was present (Fig. 7B, lanes 3 and 8), but the proportion of single-nucleotide products remained low (Fig. 7C). Because pol␤ is the major activity for removing 5Ј-dRP moieties and filling in the resulting single-nucleotide gaps, our observations suggest that the inability of the polymerase to remove 5Ј-dL lesions alters the repair mode during BER of the oxidized lesion compared with a regular AP site.
The addition of a p21 peptide to block interactions of PCNA with other BER proteins and suppress long-patch BER (44,45) diminished the formation of multinucleotide repair patches for the pol␤-proficient extract acting on AP sites (Fig. 7B, lanes 1 and 2). A smaller but analogous effect was observed for AP repair in the pol␤-deficient extract (Fig.  7B, lanes 6 and 7). The p21 peptide also affected the products seen with dL: in the presence of pol␤, the peptide increased the amount of material scored as single-nucleotide (Fig. 7C), but there was almost no effect in the pol␤-deficient extract (Fig. 7B, lanes 8 and 9; Fig. 7C). The products of the dL reactions scored as single-nucleotide BER may include incomplete repair intermediates in which a single nucleotide has been inserted but no further reaction has yet occurred. The effect of the p21 peptide in the POLB ϩ/ϩ extract compared with the POLB Ϫ/Ϫ extract was consistent with the generation of these products by pol␤. Activation of the dL precursor also generates a proportion of strand breaks with 3Ј-phosphate residues (30,31), which could also generate some singlenucleotide repair product following the action of polynucleotide kinase (18,19).
Complete Repair of dL Is Mediated Exclusively by Long-patch BER-To verify that the complete repair of dL residues indeed occurred through long-patch BER, we analyzed the repaired DNA products by the restriction digestion approach adapted from a published procedure (38). As shown in Fig. 8A, pGEM plasmid DNA substrates that con-  1-3), the tBU precursor lesion (lanes 4 -6), or a dL residue (lanes 7-9) were incubated with 0, 0.1, and 1 nM FEN1 for 30 min at 30°C. The reaction products were analyzed on a 15% polyacrylamide/urea DNA sequencing gel and visualized using a PhosphorImager. Schematic representations of the substrates are depicted for each set of lanes; the asterisks indicate the positions of the 32 P-label in each substrate.  1, 2, 6, and 7), pGEM (dL/A) (lanes 3, 4, 8, and 9), or pGEM (tBU/A) (lanes 5 and 10) DNA, together with 50 g of POLB ϩ/ϩ (lanes 1-5) or POLB Ϫ/Ϫ (lanes 6 -10) MEF extract were incubated for 60 min at 30°C in the absence or presence of 50 M of p21 peptide (lanes 2, 4, 7, and 9). Following the incubation, the DNA products were isolated, and samples (100 ng) were digested with BamHI, subsequently incubated with 20 units of Exo III, and then digested with HindIII. As controls, 200 ng of 32 P-labeled pGEM (T/A) DNA was also similarly treated, but without (lane U) or with (lane T) the same amount of Exo III. The DNA fragments were analyzed on a 15% polyacrylamide/urea DNA sequencing gel. The DNA size markers (a 30-mer generated by digestion of pGEM (U/A) DNA with BamHI/HindIII, and a 12-mer produced by partial treatment of the 30-mer with Ung/Endo IV) are shown in lane M. C, the 32 P-radioactivity for each band was quantified using a PhosphorImager. The relative intensity for each band was determined by dividing the amount of 32 P radioactivity detected in that band by the total 32 P signal detected for all bands in the same lane and multiplying by 100%. The proportion of repair products with a single-nucleotide patch (black bars) and the sum of those with multinucleotide patches (white bars) are expressed as percent distributions. This quantitative analysis represents the relative amount of the single short-patch (black bars) versus long-patch (white bars) BER products, and may not reflect the visually perceived intensities of the image in panel B. The mean Ϯ S.D. for three experiments are shown. tained a site-specific AP or dL lesion, located within two unique restriction sites, were incubated with MEF cell-free extracts. We initially forced short-patch BER by supplying 32 P-dTTP as the only nucleotide, which would allow DNA repair synthesis of just one residue (Fig. 8A). The ligated products of such single-nucleotide BER are expected to generate a 32 P-labeled 30-mer DNA fragment from BamHI and HindIII digestion, with the unligated products yielding a 13-mer fragment. Under these conditions, only the completely repaired DNA product would contain the recognition sequence for AccI to generate a 13-mer 32 P-DNA fragment by co-digestion with BamHI. For long-patch BER, the reactions were conducted in the presence of 32 P-labeled dCTP and unlabeled dTTP. Under these conditions, incorporation of 32 P-dCMP would require repair DNA synthesis of two nucleotides (Fig. 8A). Digestion of the repaired DNA with BamHI/HindIII or HincII/BamHI was expected to produce 30-mer or 14-mer 32 P-DNA fragments, respectively, indicating complete repair by the long-patch BER pathway.
For AP sites, discrete 32 P-DNA restriction fragments were observed under both short-and long-patch conditions (Fig. 8B, lanes 9 -10 and  11-12, respectively). DNA repair synthesis was specific for damaged DNA, as reactions with control pGEM (A/T) DNA did not produce any detectable 32 P-DNA restriction fragments (Fig. 8B, lanes 1-4). For dL residues, in contrast, under the short-patch BER conditions complete repair was not observed (Fig. 8B, lane 5), and only a small amount of 32 P-dTMP incorporation into dL-DNA was registered as the 32 P-13mer (Fig. 8B, lane 5 and 6). The small amount of ligated product observed with the dL substrate likely corresponds to repair of the minor proportion of 3Ј-phosphate residues present in dL substrates prepared in this way (30,31). Incubation of the dL substrate under long-patch BER conditions, however, produced the 32 P-DNA products expected for digestion by the corresponding restriction enzymes (Fig. 8B, lanes 7  and 8; compare with lanes 11 and 12 for AP site repair), which indicated complete repair. We note that the amount of DNA repair synthesis associated with complete long-patch BER for dL appeared to equal to that for the long-patch component of AP repair (Fig. 8B, compare lanes  7 and 11). This observation suggests that long-patch BER of dL may occur as efficiently as it does for an AP site, or the same factor is limiting for both. The repair of a site-specific dL lesion has not been reported previously, but our results clearly demonstrated that the long-patch BER pathway is capable of handling dL lesions, which are resistant to processing by short-patch BER mechanisms.

DISCUSSION
Cellular DNA exposed to free radical attack and oxidation due to metabolic by-products or agents such as ionizing radiation contains various DNA modifications, including several types of oxidized abasic sites (18,46). Assessing the biological consequences of individual lesions, such as dL residues, has been difficult because of the complex nature of free radical DNA damage (18) and, until recently, a lack of methods to generate the lesions site-specifically. Our recent work had already shown that dL residues have the potential to generate DPC with pol␤ (31), which makes the analysis of this lesion especially important. In this study we have used DNA substrates containing a site-specific dL residue to examine repair by either purified BER enzymes or cell-free extracts, and we have demonstrated the complete repair of dL residues by long-patch BER.
The abundance of the highly active Ape1 protein in most cell types (47,48) indicates that dL residues in DNA would likely be rapidly incised to yield a 5Ј-dL moiety at the nick (30). It is this cleaved species that forms DPC with pol␤ (31), which raised the question of whether other cellular proteins might also undergo such a cross-linking reaction. Indeed, incubation of Ape1-treated dL-DNA with cell extracts yielded several distinct DPC products, with the major species evidently due to pol␤ (Fig. 3). Because pol␤ is the major 5Ј-dRP excision activity in mammalian cells (7,8), these results indicate that efficient DPC formation with 5Ј-dL lesions depends on the dRP lyase activity of pol␤ (31). More recently pol has been ascribed with a dRP lyase activity (49), but another study suggested that pol would be trapped as a covalent Schiff base intermediate at 5Ј-dRP residues (50). Although our own studies indicate that recombinant pol can form DPC with 5Ј-dL residues, 4 there was no significant dL-mediated DPC in the cell-free extracts corresponding to the molecular mass of ϳ80 kDa (Fig. 1) expected for pol (51).
Several other dL-mediated DPCs, of lower intensity than the pol␤ product, were observed to involve cross-linking to relatively small polypeptides, which was not prevented by a heat pretreatment of the extracts (Fig. 1). These other DPC may be due to abundant DNA-binding proteins, such as histones, participating in non-enzymatic reactions with the lactone.
A recent study showed that the oxanine, a major nitric oxide-induced guanine base damage, forms stable DPC with various DNA glycosylases (human OGG1 and E. coli Fpg, AlkA, and EndoVIII) at a much higher rate than found for histone and HMG proteins (52). Oxanine and dL are base and sugar lesions, respectively, but both involve DPC formation via a lactone structure that is prone to react with nucleophilic molecules. The two major oxanine-mediated DPC species detected in human cell extracts (52) involve as-yet-unknown proteins other than hOGG1, the enzyme that plays a major role in the recognition and removal of the oxidized guanine base (53,54). For DPC formation between 5Ј-dL DNA and pol␤, the cross-linking efficiency (Fig. 2) 1, 2, 5, 6, 9, and 10), or a mixture of cold-dTTP and 32 P-dCTP (C; lanes 3, 4, 7, 8, 11, and 12) to allow DNA repair synthesis associated with either short-patch or long-patch BER, respectively. The DNA products were isolated and digested with BamHI (B) together with HindIII (H), AccI (A), or HincII (h), as indicated in the figure, followed by electrophoresis on a 15% polyacrylamide/urea DNA sequencing gel. As an internal standard (IS), a 5Ј-32 P-labeled 35-mer oligonucleotide DNA was included in each reaction and used to normalize any variation that could be associated with DNA isolation and sample loading on the gel. the intrinsic reactivity of the active-site lysine-72 and the strong affinity of the enzyme for 5Ј-dRP residues (55), but also by Ape1-mediated recruitment during BER (56). On the other hand, we found that the pol␤ dRP lyase lags behind its polymerase activity (57), which could offset the potential for DPC formation during BER.
Long-patch BER seems to provide an effective mechanism to avoid DPC formation during the excision of dL residues. A partial long-patch BER reaction reconstituted with Ape1, pol␤, and FEN1 allowed the excision of 5Ј-dL on a displaced oligonucleotide flap (Fig. 5), with the product further processed by ligation (Figs. 7 and 8). However, once DPC were formed, either FEN1 or cell-free extract appeared unable to process the DPC on a displaced DNA flap (31) (see supplemental Figs. S4 and S5).
Although short-patch BER appears to be the predominant mode for the repair of AP sites or base lesions converted to AP sites by monofunctional DNA glycosylases, several in vitro and in vivo studies also suggest a significant contribution of long-patch BER (17,40,58,59). Our data for AP site repair are consistent with these observations (Fig. 7). Repair of the dL lesion, in contrast, was accompanied exclusively by multinucleotide repair synthesis (Fig. 7). As a consequence, in cell-free extracts proficient for long-patch BER, the formation of dL-mediated DPC was dramatically diminished (Fig. 4). The cross-linking reaction between pol␤ and dL is not a fast one (31), and evidently the long-patch BER steps occur rapidly enough to prevent most of the possible DPC formation.
It has not been established what factors are involved in the selection between the short-or long-patch BER mode. For DNA base lesions, the nature of the participating DNA glycosylase is important (59): DNA glycosylases with associated AP lyase activity lead mainly to short-patch BER, whereas monofunctional DNA glycosylases lead to a mixture of the short-and long-patch pathways.
The exclusive use of the long-patch pathway for dL, a natural product, echoes a similar finding for the synthetic tetrahydrofuran analog often used as a stable substitute for AP sites. However, under conditions of oxidative stress that might accompany the formation of relatively large numbers of dL residues, BER would also be engaged with handling many other base and deoxyribose lesions. Under these circumstances, it is unclear whether the coordination that seems to operate generally in BER can be maintained. It seems possible that, in such circumstances, Ape1-incised dL residues may remain in the DNA for longer periods, increasing the opportunity for DPC formation with pol␤ and pol (enzymatically), and with other cellular proteins (non-enzymatically). The study of dL will provide opportunities for analyzing these issues and the switching mechanisms that govern the shortversus long-patch BER distribution under varying circumstances of damage load and repair enzyme availability. Developing methods to determine the extent to which dL-mediated DPC are formed in vivo is an important goal to enable such an analysis