Excision of C-4′-oxidized Deoxyribose Lesions from Double-stranded DNA by Human Apurinic/Apyrimidinic Endonuclease (Ape1 Protein) and DNA Polymerase β*

Oxidative damage to DNA deoxyribose generates oxidized abasic sites (OAS) that may constitute one-third of ionizing radiation damage. The antitumor drug bleomycin produces exclusively OAS in the form of C-4-keto-C-1-aldehydes in unbroken DNA strands and 3′-phosphoglycolate esters terminating strand breaks. We investigated whether two human DNA repair enzymes can mediate OAS excision in vitro: Ape1 protein (the main human abasic endonuclease (also called Hap1, Apex, or Ref1)) and DNA polymerase β, which carries out both the abasic excision and the resynthesis steps. We used a duplex oligonucleotide substrate with one main target for bleomycin-induced damage. Ape1 catalyzed effective incision at the C-4-keto-C-1-aldehyde sites at a rate that may be only a few-fold lower than incision of hydrolytic abasic sites at the same location. Consistent with several previous studies, Ape1 hydrolyzed 3′-phosphoglycolates 25-fold more slowly than C-4-keto-C-1-aldehydes. DNA polymerase β excised the 5′-terminal OAS formed by Ape1 incision at a rate similar to its removal of unmodified abasic residues. Polymerase β-mediated excision of 5′-terminal OAS was stimulated by Ape1 as it is for unmodified abasic sites. Escherichia coli Fpg (MutM) protein also excised 5′-terminal OAS, but in our hands, the RecJ protein did not. These observations help define mammalian pathways of OAS repair, point to interactions that might coordinate functional steps, and suggest that still unknown factors may contribute to removal of 3′-phosphoglycolate esters.

Under physiological conditions, the human genome is continuously damaged by hydrolysis; reactive metabolic by-products, such as oxygen radicals; exogenous agents that penetrate cells; and radiation (1,2). The tremendous sizes of chromosomal DNA molecules render them susceptible to even infrequent events that covalently alter their structure. For example, all of the hydrogen atoms of DNA deoxyribose can potentially be abstracted by free radicals. Four of the seven sugar hydrogens in B-form duplex DNA, i.e. C-1Ј-H, C-2Ј-H(R), C-4Ј-H, and C-5Ј-H(S) are exposed to free radicals through the minor groove and the others through the major groove (3,4). ␥-Radiation initiates oxidative DNA damage on sugar moieties by hydrogen abstraction from the C-1Ј, C-2Ј, and C-4Ј positions (5). Hydro-gen abstractions from minor groove-accessible C-1Ј, C-4Ј, or C-5Ј positions by some antitumor drugs, such as the enediyne agents and bleomycins (6 -9), result in modified abasic residues and in strand breaks terminated with different damaged sugar residues at either the 3Ј-or the 5Ј-end of the breaks. Such oxidized abasic sites (OAS) 1 may comprise up to one-third of total oxidative damage (2,10). It is expected that the abasic sites and DNA fragmentation resulting from oxidative agents would block DNA replication and be cytotoxic or mutagenic (1,2). However, little is known about the biochemistry of repair of these naturally occurring and environmentally induced lesions that may threaten genetic stability.
Genetic instability can be counteracted by various pathways of DNA repair (1,2,11,12). Nucleotide excision repair, mediated by ϳ30 proteins in mammalian cells, acts on a wide variety of DNA lesions (11,12). Base excision repair (2) handles altered bases through the action of DNA glycosylases that produce apurinic/apyrimidinic (AP) sites. The AP sites generated by glycosylases can be incised at the 5Ј-side by a number of AP endonucleases, such as the mammalian Ape1 protein (13). Ape1 protein 2 (also called Ref1 (14), Hap1 (15), or Apex (16)) is the major AP endonuclease found in human cells and belongs to a large family of nucleases homologous to exonuclease III of Escherichia coli (2,17). Ape1 has a broad specificity for AP sites and incises them to generate a 3Ј-OH and a 5Јterminal deoxyribose-5-phosphate (5Ј-dRp) residue (17)(18)(19). The resulting 5Ј-dRp can be removed by an intrinsic activity of human DNA polymerase ␤ (Pol␤) (20,21), followed by DNA repair synthesis to fill the gap and ligation to complete the repair process (2). Recently, human Ape1 protein and Pol␤ were found to act in an orchestrated manner mediated by protein-protein interactions (22). An alternative base excision repair pathway is used when strand displacement repair synthesis generates a single-stranded 5Ј-end that must be removed by a "FLAP" endonuclease (23).
The C-4Ј oxidized deoxyribose damage initiated by bleomycin is a mixture of oxidized abasic lesions and strand breaks (6,7,9,24). It is proposed (6,7) that the activated bleomycin, after binding into the DNA minor groove, position-specifically abstracts a hydrogen atom from the C-4Ј position of the deoxyribose of the pyrimidine (Py) in a d(GpPy) sequence to initiate the degradation process. One pathway (Fig. 1, lower branch) leads to the formation of a C-4Ј-hydroxylated abasic site (2deoxypentose-4-ulose), which can be chemically cleaved by hydrazine treatment to form a strand break with a 3Ј-pyridazine (25). Strand breaks with 3Ј-phosphoglycolate (3Ј-PG) can be generated directly (Fig. 1, upper branch) with the concomitant release of base-propenal. Although the mechanisms are different, C-4Ј hydrogen abstraction by enediyne antitumor antibiotics, such as neocarzinostatin (8,26) and C1027 (27), results in the same DNA damage products in either the presence or the absence of thiols. Since C-4Ј-oxidized abasic sites constitute as much as 10% of the damage formed by ionizing radiation (10), they are also likely to be significant products of the chemical oxidation that occurs in vivo as a by-product of aerobic metabolism (2).
Here we address whether these C-4Ј-oxidized damage products can be enzymatically excised in vitro and investigate the efficiency with which these products are processed by known enzymes. We have designed a fragment of duplex DNA (Fig. 2, duplex D1⅐D2) that contains one predominant attack site for bleomycin, which exhibits a limited sequence dependence for its reactions (6,7). The results show that the C-4Ј-oxidized abasic residues are readily excised by the human enzymes and confirm that 3Ј-PG lesions are poor substrates for excision by Ape1, consistent with previous studies of randomly introduced 3Ј-PG lesions (see "Discussion"). As they do acting on hydrolytic AP sites, Ape1 and Pol␤ can work cooperatively in the excision of OAS.

EXPERIMENTAL PROCEDURES
Materials-High pressure liquid chromatography-purified DNA oligonucleotides were commercial products from Operon Technologies (Alameda, CA). Radiolabeled nucleotides and the enzymes used for labeling were from NEN Life Science Products and New England Biolabs, Inc. (Beverly, MA), respectively. Bleomycin (1.2-1.7 units/mg) was purchased from Sigma. Recombinant human Ape1 protein expressed in E. coli was purified (Ͼ95% homogeneity as judged by silver staining of polyacrylamide-SDS gels) as described by Wilson et al. (19). Endonuclease IV (Endo IV) was purified (Ͼ98% purity) by D. M. Wilson III using a method described previously (28). Recombinant human Pol␤ protein was purified by R. A. O. Bennett as described previously (22) using an expression vector generously provided by Drs. J. Carney and S. Linn (University of California, Berkeley). E. coli uracil-DNA glycosylase and Fpg protein were kindly provided by Dr. D. Mosbaugh (Oregon State University) and Dr. A. Grollman (State University of New York, Stony Brook), respectively.
Terminal Labeling of Oligonucleotides with 32 P-Single-stranded DNA oligomers were end-labeled with 32 P at the 5Ј termini with [␥-32 P]ATP and T4 polynucleotide kinase (29) and annealed to the complementary strands. For 3Ј-end labeling, the oligonucleotides were first annealed and then incubated with [␣-32 P]dATP (for D1 and D3) or [␣-32 P]dCTP (for D2) and the Klenow fragment of E. coli DNA polymerase I (29).
Internal Labeling of Oligonucleotides with 3 H or 32 P-For 32 P internal labeling, oligomer D5 (which corresponds to the 3Ј half of D1; Fig. 2) was first 5Ј-32 P-end labeled as described above and then gel-purified. Oligomer D4 (the 5Ј-half of D1; Fig. 2) was placed adjacent to the 5Ј-labeled D5 by annealing both oligonucleotides to a third oligomer, D8 (Fig. 2). The nick between D4 and D5 was sealed by T4 DNA ligase to produce an internally 32 P-labeled D1. 3 H internal labeling at the bleomycin target site was done following the method of T. A. Winters et al. (30) with some modifications. Briefly, the deoxyribose of the target T nucleotide in D1 was labeled with 3 H at the C-1Ј-and C-2Ј-positions by incubation of [methyl-1Ј,2Ј-3 H]dTTP and duplex D4⅐D6 with the Klenow fragment of DNA polymerase I in the presence of dGTP and dATP. Similarly, the C-5Ј-position was labeled with 3 H by incubation of [5,5Ј-3 H]dCTP with Klenow fragment in the presence of duplex D4⅐D7, dGTP, and dATP. The internally 3 H-labeled D4 oligonucleotide was then gelpurified and annealed to oligonucleotide D8 (Fig. 2). The 3Ј portion of D1 was then synthesized by Klenow fragment in the presence of all four dNTPs. When [5,5Ј-3 H]dCTP was incorporated, which substitutes the target T nucleotide on D1, a complementary strand containing a G on the opposite position was annealed. The DNA duplex with C replacing T is still expected to provide a good target for bleomycin reaction, but more double-stranded lesions are expected to be generated (24). However, these double-stranded lesions should not interfere with the analysis conducted here. All labeled oligonucleotides were purified using electrophoresis in denaturing polyacrylamide gels (National Diagnostics) and then annealed to their complementary strands at a molar ratio of 1:1.
Bleomycin Reactions with DNA-Bleomycin reactions with DNA were carried out according to published protocols (6,7,24). In brief, bleomycin was first mixed with an equal amount of ferrous ammonium sulfate at 4°C for exactly 1 min to make "activated bleomycin." The activated bleomycin was then added to the annealed DNA oligonucleotides in Hepes-KOH buffer (50 mM, pH 7.8) containing 50 mM NaCl to initiate the drug-DNA reactions (6,7). Alternatively, the drug-DNA reactions were started by adding ferrous ammonium sulfate to a mixture of DNA and bleomycin (24). The reactions were allowed to proceed for 1-2 h at room temperature and then stopped by the addition of sodium acetate to 0.3 M (from a pH 7.0 stock) and ethanol precipitation or by washing with 20 volumes of 5 mM Hepes-KOH buffer (pH 7.6) in an ultrafiltration apparatus (Centricon 10, Amicon, Beverly, MA). These procedures remove most of the free bleomycin.
Repair Enzyme Reactions-The freshly prepared, bleomycin-treated DNA substrates (0.25 M final concentration) with position-specific, C-4Ј-oxidized deoxyribose lesions were incubated with the indicated amounts of Ape1, Endo IV, Fpg, or Pol␤ protein in 50 mM Hepes-KOH buffer (pH 7.6) containing different enzyme-specific components. For Ape1 reactions, 50 mM KCl, 100 g/ml BSA, 0.05% Triton X-100, and 10 mM MgCl 2 were included (18). The Endo IV reactions contained 50 mM KCl, 50 g/ml BSA, and 1 mM EDTA (28). The Ape1 or Endo IV reactions were stopped by flash freezing and drying in a Speed-Vac concentrator or further incubated with Fpg or Pol␤ in Hepes-KOH buffer containing 2 mM dithiothreitol, 50 mM NaCl, 2.5% glycerol, and 100 g/ml BSA (22). The repair enzyme reactions were assembled at 0 -4°C and incubated at room temperature. In order to stabilize the C-4Ј-oxidized abasic sites and unexcised 5Ј-dRp, Pol␤ reaction mixtures were terminated by treatment with 300 mM NaBH 4 for 1 h at room FIG. 1. Proposed mechanism for iron-bleomycin-mediated C-4 DNA damage at a target T nucleotide. These reactions take place within the context of a DNA duplex, which for simplicity is not shown. See Refs. 6 and 7 for details.
temperature. The DNA samples were then precipitated in ice-cold 75% ethanol containing 0.3 M sodium acetate and pelleted by centrifugation.
High Resolution Sequencing Gel Analysis-After the DNA-bleomycin or DNA-enzyme reactions, the samples were either directly frozen and dried in a Speed-Vac concentrator or precipitated with ethanol and pelleted by centrifugation. The DNA samples were then taken up in a loading buffer containing 80% formamide and dyes, and then separated on 15 or 20% polyacrylamide sequencing gels as described previously (27). The gels were visualized by autoradiography, and the intensities of the bands were quantitated using a phosphor imager (Bio-Rad). When needed, the bands were excised, the materials inside the gel slices were eluted with distilled H 2 O, and the amount of radioactivity was determined by liquid scintillation using an LS 1801 counter (Beckman Instruments, Fullerton, CA).

Site-specific Bleomycin Damage in Oligonucleotides-Based
on the reported sequence specificity of bleomycin (6, 7, 24), a 36-mer DNA oligonucleotide was designed and 5Ј-32 P-end-labeled for in vitro studies. When this duplex DNA fragment (duplex of oligomers D1 and D2; Fig. 2) was reacted with different amounts of bleomycin, one major damage site was found on strand D1 (Fig. 3, lanes 4 -9) and two minor sites on D2 (lanes 13-18), respectively. The band in Fig. 3, lane 4 (indicated by an arrow) that moves slightly faster than the Maxam-Gilbert "T" sequencing marker corresponds to a DNA fragment bearing a 3Ј-PG terminus, which is diagnostic for C-4Ј chemistry (27). To reveal additional OAS, samples were treated with hydrazine, which specifically cleaves C-4Ј-oxidized abasic sites to generate strand breaks terminated with 3Ј-pyridazine ( Fig. 1). As shown in Fig. 3, lane 5, the band (marked by an asterisk) that was generated by hydrazine treatment and moves more slowly than the "T" sequencing marker corresponds to a DNA cleavage product containing a 3Ј-pyridazine terminus (27). Such a band was not observed without bleomycin treatment (Fig. 3, lanes 2 and 11). Quantitation by phosphor imaging of the bleomycin damage products showed that, under our conditions, C-4Ј-oxidized abasic sites and direct strand breaks (3Ј-PG) were generated at a ratio of ϳ1.5:1. Comparison of lanes 5 and 14 of Fig. 3 shows that, when almost all of the D1 strand had been modified, 35% of oligomer D2 remained without detectable bleomycin damage. Thus, the central T nucleotide within oligonucleotide D1 was indeed the major site for bleomycin attack in this DNA fragment. However, we also note that, at the highest level of bleomycin treatment, considerable damage at secondary sites was observed, which was accompanied by some loss of material from the gel analysis. Therefore, to ensure that the bleomycin-treated DNA fragments prepared for later enzymatic treatment would contain only one damaged nucleotide, the extent of damage in the D1 oligomer was limited to 5-7% of the total D1 oligomer by using bleomycin concentrations of 3-5 M.
Excision of 3Ј-PG Residues and 5Ј Incision of the C-4Ј-oxidized Abasic Sites by Human AP Endonuclease Ape1-The abundance of Ape1 protein in human cells (13,18) and its ability to cleave DNA at diverse types of abasic sites (19) prompted us to test the enzyme's ability to act on OAS. Incubation of the bleomycin-treated, 5Ј-labeled D1⅐D2 duplex with Ape1 initially generated a single cleavage product (band labeled b in Fig. 4), which remained as the main product at the highest level of enzyme treatment (Fig. 4, lane 6). For comparison, the 5Ј-labeled D2⅐D3 DNA duplex, treated with uracil-DNA glycosylase to generate a single AP site, was cleaved by Ape1 protein to generate a single product of the same mobility (Fig. 4, lanes 7-10). As expected, the mobility of this product (faster than the G (not shown) and slower than the T Maxam-Gilbert markers with 3Ј-phosphate ends generated by piperidine treatment) was consistent with cleavage by Ape1 at the single AP site to form a DNA fragment containing a 3Ј-OH end. Thus, Ape1 cleaves C-4Ј-oxidized abasic lesions in the same way. Because only some of the DNA in the bleomycin-treated FIG. 3. Denaturing gel analysis of DNA damage products formed by bleomycin. The 5Ј-32 P-end-labeled oligomer D1 (lanes 1-9) and D3 (lanes 10 -18) were separately annealed to the D2 complementary strand. The DNA substrates (0.5 M) were incubated for 1 h with bleomycin at different drug:DNA molar ratios (numbers shown at the top of the lanes) as described under "Experimental Procedures." ϩ, samples treated with hydrazine; this treatment converts C-4Ј-oxidized abasic sites to strand breaks with 3Ј-pyridazine ends (marked by an asterisk in lane 5). Samples were analyzed on a 15% sequencing gel. C ϩ T, Maxam-Gilbert sequence markers. The arrows indicate the 3Ј-PG formed at the target T site in D1 and two minor sites in D2.

FIG. 2. Sequences of oligodeoxyribonucleotides.
The predominant target T nucleotide for bleomycin attack is printed in boldface in oligomer D1. The sequence of oligomer D3 is same as that of D1 except the target T is substituted by a U, which can be removed by uracil-DNA glycosylase to generate an AP site. D2 is the complementary strand for both D1 and D3. Oligomers D4 to D8 were used for internal labeling (see "Experimental Procedures").
samples contained OAS at the target site, and allowing for possible inhibitory effects of other damage in these samples, we cannot make an accurate quantitative comparison of Ape1 activity on C-4Ј-oxidized versus hydrolytic abasic sites. However, we note that 54 nM Ape1 acting on C-4Ј-oxidized abasic sites generated product at Ն10% of the amount generated from hydrolytic abasic sites (Fig. 4, lane 5 versus lane 10). Since other lesions in the bleomycin-treated DNA probably compete for Ape1 with the C-4Ј-oxidized abasic residues at the target site, the relative efficiency of the enzyme on this type of OAS may be within a few-fold of the activity on regular AP sites.
Lane 3 of Fig. 4 shows the direct cleavage product of bleomycin acting on the 5Ј-labeled D1⅐D2 duplex to generate a strand break with 3Ј-PG (indicated by an arrow). Incubation with increasing amounts of Ape1 protein led to the disappearance of the 3Ј-PG band, with nearly complete excision at the highest Ape1 level (Fig. 4, lane 6). Band b contains the common product of Ape1 3Ј-PG excision and abasic site incision activities. However, at an intermediate Ape1 concentration, most of the 3Ј-PG substrate remained intact in the same sample in which there was extensive incision at the C-4Ј-oxidized abasic site (Fig. 4, lane 5). This result suggested that Ape1 acts more readily on the intact abasic residues formed by bleomycin than it does on the 3Ј-PG sites, and this conclusion was verified by a kinetic analysis (see below).
The same products were formed by Endo IV (Fig. 4, lanes [11][12][13][14], in keeping with earlier reports of this enzyme's activity upon bleomycin cleavage (28,31). However, Endo IV excised 3Ј-PG as efficiently as it incised C-4Ј-oxidized abasic sites (compare lanes 12 and 13 of Fig. 4). In addition, the action of the bacterial enzyme on the OAS was nearly as robust as its activity on hydrolytic AP sites (see lanes 15-18 of Fig. 4). Band c in Fig. 4 is probably generated by a trace amount of contaminating exonuclease in this particular Endo IV preparation (data not shown).
To verify the incision sites of Ape1 and Endo IV at bleomycin damages, the D1 oligomer was annealed with D2 and 32 Plabeled at its 3Ј-end before treatment with bleomycin. Ape1 or Endo IV treatment of the 3Ј-labeled substrate generated a single cleavage product (Fig. 5, lanes 2 and 7, labeled with an asterisk) of slower mobility than the 5Ј-phosphate produced by the direct action of bleomycin (arrow in Fig. 5). The slower mobility of this product is consistent with the presence of an abasic residue at the 5Ј terminus. Thus, Ape1 protein recognizes and incises the C-4Ј-oxidized abasic residues on their immediate 5Ј side.
Kinetics of the 3Ј-Diesterase and 5Ј-Incision Activities of Ape1 on C-4Ј-oxidized Deoxyribose Lesions-Ape1 protein can remove 3Ј-PG residues from DNA single-stranded breaks, onebase gaps, and from double-stranded breaks with either blunt or two-base recessed 3Ј-termini (Refs. 30 and 32 and this work). Since we found that Ape1 can also incise on the 5Ј side of C-4Ј-oxidized abasic sites, it was of interest to compare the excision and incision activities of Ape1 in the same reaction system. For this purpose, oligomers were 32 P-labeled at either the 5Ј-end or internally in the phosphodiester at the immediate 5Ј side of the target T nucleotide in sequence D1 (Fig. 2). With the internal-labeled oligomers, the DNA fragments with 3Ј-PG residues (direct damage products) and the fragments with 5Јterminal oxidized abasic residues (Ape1 dependent products) can be well separated as two bands on 15% sequencing gels (see Fig. 7). Quantitation by phosphor imaging showed that the incision activity of Ape1 on C-4Ј-oxidized AP sites was much more active (Ն25-fold) than its 3Ј-diesterase activity (Fig. 6). After a 2-h incubation with Ape1 protein at a Ape1:DNA molar ratio of 1.1:1, there was almost no diesterase activity detected, while Ͼ50% of the C-4Ј-oxidized abasic sites had been incised  Fig. 4. The strand breaks with 5Ј-PO 4 ends directly generated by bleomycin are indicated by an arrow. The DNA fragments with 5Ј-terminal C-4Јoxidized abasic residues generated by the 5Ј-incision activities of Ape1 (labeled Ape) or Endo IV are marked by an asterisk (see lane 2). The incubation with Pol␤ was for 60 min at room temperature. The final concentrations of the enzymes are shown above the lanes. (Fig. 6A). Doubling the relative amount of Ape1 protein allowed some 3Ј-PG excision to be detected (Fig. 6B) at a rate Ͼ15-fold lower than the enzyme's activity on the oxidized abasic residue. This result is consistent with the relative 3Ј-diesterase and AP incision activities of Ape1 determined using a synthetic substrate (18). The bacterial protein Endo IV, however, showed a similar efficiency for both activities (Fig. 4), consistent with published data (33).

5Ј-Excision Activity of Pol␤ for C-4Ј-oxidized Abasic Residues-
The foregoing experiments show that the C-4Ј-oxidized abasic sites are effectively incised on the 5Ј side by Ape1 or Endo IV to generate strand breaks with 3Ј-OH and 5Ј-abasic phosphate residues. For repair to proceed, the 5Ј-terminal abasic residues need to be removed. Mammalian DNA polymerase ␤ has this ability for the 5Ј-dRp generated by AP endonucleases at a glycosylase-generated AP site (11,20,22). Since the excision occurs by ␤-elimination (20,21), chemically reduced AP sites and some OAS are resistant to Pol␤ (23). To determine whether C-4Ј-oxidized residues specifically are substrates for excision by Pol␤, the 3Ј-32 P-labeled duplex was treated with bleomycin and then with an excess of Ape1 protein or Endo IV to generate an incised substrate. This material contained direct strand breaks with 5Ј-phosphate ends (Fig. 5, lanes 1 and 6 indicated by an arrow) accompanying 3Ј-PG formation, and 5Ј-terminal C-4Ј-oxidized abasic residues (Fig. 5, lanes 2 and 7, denoted by asterisk) generated by the endonucleases. After additional incubation with different amounts of Pol␤, the intensities of the bands corresponding to the endonuclease prod-ucts were decreased, while the amount of the DNA fragment with a 5Ј-phosphate end was correspondingly increased (Fig. 5,  lanes 3-5 and 8 -10). Thus, C-4Ј-oxidized abasic residues can be excised by human Pol␤ to generate a one-nucleotide gap with normal 3Ј-OH and 5Ј-phosphate ends. Chemical reduction of the 5Ј-terminal oxidized abasic residues with NaBH 4 before incubation with Pol␤ prevented the excision (results not shown), consistent with their removal by ␤-elimination. As for Ape1 incision at C-4Ј-oxidized abasic sites, it is difficult to make an accurate quantitative comparison of Pol␤ excision activity on the hydrolytic versus oxidized abasic residues using the bleomycin-treated substrate.
A careful analysis of the Pol␤ excision reactions shows that the polymerase was more efficient in the presence of Ape1 protein than when Endo IV was present. Approximately 3-fold more Pol␤ was required for extensive removal of the abasic residue when only the bacterial enzyme was present than when the cognate enzyme Ape1 was available (Fig. 5, compare lanes  3-5 and 8 -10). This activation by Ape1 protein is consistent with its stimulation of 5Ј-dRp excision by Pol␤ (22). A quantitative analysis showed that Ͼ95% of the 5Ј-terminal C-4Јoxidized abasic residues were removed by Pol␤ in a reaction containing Ape1 protein in 30 min at a DNA:Pol␤ molar ratio of 5.3:1 (data not shown).
The Fpg protein of E. coli can excise 5Ј-dRp by a ␤-elimination reaction analogous to that of Pol␤ (34). In a separate experiment with internally 32 P-labeled D1 oligomer in a bleomycin-treated D1⅐D2 duplex, Fpg protein was tested for exci-

FIG. 6. Kinetics of 3-PG excision and C-4-oxidized abasic site incision by Ape1 protein.
The DNA (D1⅐D2) substrates treated with bleomycin to contain 3Ј-PG and abasic sites at the target T were incubated with Ape1 protein at room temperature at Ape1:DNA molar ratios of 1.1:1 (A) or 2.2:1 (B). Samples were removed at the indicated times, mixed with an equal volume of 10% SDS, and flash-frozen in liquid nitrogen. Samples were then thawed and treated with NaBH 4 to stabilize the unincised abasic lesions. The bands containing DNA fragments with 3Ј-PG or 3Ј-OH ends were quantified by phosphor imaging. The 5Ј-incision activity of Ape1 was calculated by subtracting the amount of 3Ј-PG excised from the amount of 3Ј-OH generated; 3Ј-OH is the common product of the two enzymatic activities of Ape1 protein acting on 3Ј-PG and abasic sites. q--q, abasic site incision activity; E --E, 3Ј-diesterase activity. There was no detectable spontaneous decomposition (Ͻ5%) of the C-4Ј-oxidized abasic lesions under the reaction and analysis conditions used here. sion of the 5Ј-terminal C-4Ј-oxidized abasic residues (Fig. 7). Consistent with the results of Fig. 4 involving 5Ј-end-labeled D1, Endo IV removed 3Ј-PG (disappearance of the 3Ј-PG band in lane 2) and incised on the 5Ј-side of OAS to generate 5Јoxidized abasic residues (the new band denoted by a in lane 2). The bands migrating between the intact DNA and band a were probably due to other minor bleomycin-damaged sites; these secondary sites appear to be processed by the enzymes in the same way as the primary site (compare Fig. 7, lanes 1 and 2). Incubation with Fpg protein excised the 5Ј-terminal oxidized abasic residues produced by Endo IV (disappearance of band a). E. coli RecJ protein was also reported to have dRpase activity (35); however, in our hands, RecJ even at a molar ratio of 5:1 (enzyme:DNA) was unable to excise either a C-4Ј-oxidized or a hydrolytic abasic residue from our DNA substrates (data not shown).
Analysis of the Enzymatically Released PG and C-4Ј-Oxidized Abasic Residues-In order to analyze the released products, a substrate was prepared that was internally doublelabeled with 3 H and 32 P at the bleomycin target site. After incubation with an excess of either Ape1 or Endo IV protein, 3Ј-32 PG was removed, and C-4Ј-oxidized abasic sites were incised at the 5Ј-side to produce strand breaks with 5Ј-32 P labeled abasic residues at the 5Ј-termini (band a in Fig. 8A, lane 2). Consistent with the results in Fig. 5, the 5Ј-terminal C-4Јoxidized abasic residues formed at the main target site can be released from the 5Ј terminus by further incubation with Pol␤, and excision also proceeds at other damaged sites (Fig. 8A,  lanes 3-5).
To analyze the products released by Pol␤, the samples were electrophoresed under the same conditions but for a much shorter time (45 versus ϳ200 min) after loading the samples in the same order (Fig. 8B). Because the D1 oligomer was internally labeled with 32 P in its center, only the released products and DNA fragments longer than a 17-mer were visualized on the gel under these conditions. Fig. 8B shows the results of combined Endo IV and Pol␤ treatment. There are three released products, which are denoted by PG, PO 4 , and an asterisk, respectively (lanes 2-5). The same products were released by a combined incubation with Ape1 and Pol␤ (results not shown). To identify these enzymatically released products, the bands were excised, the materials inside the gel were eluted, and the amounts of 3 H and 32 P were determined by liquid scintillation counting. The band denoted by PG in Fig. 8B was found to contain 3 H when C-5Ј-3 H-labeled D1 was used in the reactions. When C-1Ј-3 H-and C-2Ј-3 H-labeled oligomer was used, no 3 H was found in the PG band (Table I). Since C-5Ј but not C-1Ј and C-2Ј are retained in PG (Fig. 1), these results are consistent with the identification of this band as PG. Quantitation by phosphor imaging of the radioactivity in the PG bands in Fig. 8B showed an amount of 32 P almost equal to that of the 3Ј-PG band in lane 1, Fig. 8A, consistent with the origin of the excised PG from the 3Ј-PG formed by bleomycin.
The band denoted in Fig. 8 by PO 4 contained no detectable 3 H for either the C-5Ј-labeled or the C-1Ј-and C-2Ј-labeled substrate (Table I), consistent with its identification as free HPO 4 2Ϫ , which may arise as a decomposition product of incised or released lesions. The intensities of the bands marked by an asterisk, which retained the C-5Ј 3 H label in the same ratio as PG (Table I), were much less than that of band a in Fig. 8A. However, the combined total 32 P radioactivity (determined by phosphor imaging) of this band and that of the PO 4 band was almost the same as that of band a, which suggests that the released C-4Ј-oxidized abasic residue was unstable and decomposed to HPO 4 2Ϫ and an unknown product during the work-up or gel analysis. The PO 4 band and a small amount of the released abasic residue in lane 2 of Fig. 8B are probably decomposition products of the 5Ј-terminal C-4Ј-oxidized abasic residues formed by Endo IV incision. The excised products from uracil-DNA glycosylase-produced AP sites, which have a similar migration rate on the gel as that of the released C-4Јoxidized abasic residues, were found to be relatively stable, and no detectable HPO 4 2Ϫ was found in these reactions under the same analysis conditions (data not shown).

DISCUSSION
In the present study, we have shown that Ape1 protein, the major human AP endonuclease, can recognize the C-4Ј-oxidized abasic site and efficiently incise on the 5Ј side to generate a nick bounded by a 3Ј-OH end and a 5Ј-terminal oxidized abasic residue. We confirmed that the bacterial enzyme endonuclease IV incises C-4Ј-oxidized abasic sites in the same way. The 5Ј-terminal abasic residue resulting from cleavage can then be excised by human Pol␤ or bacterial Fpg protein to leave a one-nucleotide gap with a 5Ј-phosphate and a 3Ј-OH end (Fig.  9). Further, Ape1 stimulates Pol␤ excision activity on the 5Јterminal oxidized abasic residue, consistent with the recent finding that Ape1 and Pol␤ act on an AP site in an orchestrated manner (22). Completion of the repair of these lesions in vivo probably involves repair synthesis by Pol␤ and ligation by DNA ligase III complexed with the Xrcc1 scaffold protein (36) or possibly by DNA ligase I in association with Pol␤ (37).
In addition to its incision activity on the C-4Ј-oxidized abasic sites, we found that Ape1 protein can also catalyze the hydrolysis of 3Ј-PG residues, consistent with other reports (18,30,32,38). However, our work also demonstrates directly, by analysis of lesions at a single site, that Ape1's incision activity for C-4Ј-oxidized abasic sites is much more efficient than its 3Ј-PG excision function (Fig. 6). This conclusion is consistent with the previous studies that used randomly generated 3Ј-PG (30,38) or a structural analog of this lesion (18). In contrast, the Endo IV protein showed similar efficiencies in acting on the two lesions. The heterogeneous nature of the DNA substrate (major and minor damage sites, a mixture of strand breaks and abasic lesions) prevented assigning precise catalytic constants for the Ape1 acting on these oxidative lesions. However, our current and previous (18) results suggest that the release of 3Ј-damaged residues in human cells might involve other proteins, as noted by others (39). Indeed, earlier studies have shown that, in addition to Ape1, at least one (18) and possibly two (39) 3Ј-repair phosphodiesterase activities can be detected in mammalian cell extracts. The lability of these putative repair enzymes has hindered further characterization of their enzymatic and physical properties. However, our previous study does indicate that a second 3Ј-repair diesterase of HeLa cells has a higher ratio of 3Ј-phosphodiesterase to AP endonuclease activity than does Ape1 protein (18), consistent with a possible role in effecting excision of 3Ј-PG residues. Alternatively, other proteins could interact with Ape1 to stimulate some of its repair activities. The recent observation that Ape1 is inducible in response to bleomycin (40) underscores its likely importance in repair of oxidative DNA damage. The response to bleomycin may also involve novel inducible activities that participate in repair of OAS.
Recently, a second pathway for mammalian base excision DNA repair was found that yields repair patches of several nucleotides (23,36). This second pathway requires proliferating cell nuclear antigen and FEN1 (FLAP endonuclease), which removes a single-stranded region displaced by repair synthesis. This pathway seems to be determined in large part by modifications that prevent the excision of the abasic residue by the ␤-elimination mechanism of Pol␤. For example, chemical reduction of an AP site directs its repair into the longer patch pathway, and some OAS may also be processed in this way (23). Consistent with this report, we found that Pol␤ cannot excise a 5Ј-terminal C-4Ј-oxidized abasic residue following chemical reduction (data not shown). However, our demonstration that Pol␤ can quite efficiently excise the 5Ј-terminal C-4Ј-oxidized abasic residues indicates that Pol␤ could participate in repairing oxidative damage. Oxidation at positions other than C-4Ј (C-1Ј, C-2Ј, or C-5Ј) might generate OAS that are refractory to Pol␤, although only C-2Ј oxidation seems an obvious candidate. Whether Pol␤ can excise other 5Ј-terminal OAS certainly merits further study.
In summary, in order to study the mammalian repair mechanism of defined oxidative damage to the sugar moiety of the DNA backbone, we have constructed a 36-base pair DNA fragment containing only one predominant site for attack by bleomycin, which generates two types of C-4Ј-oxidized deoxyribose lesions, i.e. single-stranded breaks with 3Ј-PG ends, and C-4Јoxidized abasic sites. These lesions can be effectively excised by the major human endonuclease Ape1 in combination with Pol␤ to generate one-nucleotide gaps bracketed by 3Ј-OH and 5Јphosphate termini (Fig. 9). Repair of the gaps thus generated could be finished by repair synthesis of a one-nucleotide patch, followed by ligation to seal the remaining nick. The D1⅐D2 duplex, internally labeled with 32 P and with 3 H at the indicated positions on the deoxyribose of the target T nucleotide, was treated with bleomycin and incubated with Endo IV and Pol␤ as described under "Experimental Procedures." The enzymatically released products were separated by gel electrophoresis and analyzed by phosphor imaging (Fig. 8B). The product bands were excised, eluted, and quantitated by liquid scintillation. The results are expressed in cpm. a Corrected for spillover from 32 P channel. b Asterisk, the band corresponding to the band marked by an asterisk in Fig. 8B. c ND, not determined. d 3Ј-Pyridazine produced by hydrazine treatment (Fig. 1) can be excised by Ape1 or Endo IV, and the released product, which is not shown in Fig. 8B, moves between the PO 4 and the ATP bands.
FIG. 9. Repair of OAS: proposed mechanism for enzymatic removal of C-4-oxidized deoxyribose lesions. In step one, 3Ј-PG residues and C-4Ј-oxidized abasic sites are removed or incised, respectively, by the 3Ј-diesterase or 5Јincision activities of Ape1 or Endo IV. These reactions generate 3Ј-OH ends and 5Ј-terminal C-4Ј-oxidized abasic residues. The 5Ј-terminal abasic residues are then excised in step two by Pol␤ or Fpg proteins by ␤-elimination or possibly by hydrolysis with other unknown proteins.