Two pathways for base excision repair in mammalian cells.

Abasic sites (apurinic/apyrimidinic, AP sites) are the most common DNA lesions generated by both spontaneous and induced base loss. In a previous study we have shown that circular plasmid molecules containing multiple AP sites are efficiently repaired by Chinese hamster extracts in an in vitro repair assay. An average patch size of 6.6 nucleotides for a single AP site was calculated. To define the exact repair patch, a circular DNA duplex with a single AP site was constructed. The repair synthesis carried out by hamster and human cell extracts was characterized by restriction endonuclease analysis of the area containing the lesion. The results indicate that, besides the repair events involving the incorporation of a single nucleotide at the lesion site, repair synthesis occurred also 3' to the AP site and involved a repair patch of approximately 7 nucleotides. This alternative repair pathway was completely inhibited by the presence in the repair reaction of a polyclonal antibody raised against human proliferating cell nuclear antigen. These data give the first evidence that mammalian cell extracts repair natural AP sites by two distinct pathways: a single nucleotide gap filling reaction targeted at the AP site and a proliferating cell nuclear antigen-dependent pathway that removes a short oligonucleotide containing the abasic site and 3'-flanking nucleotides.

Abasic sites (apurinic/apyrimidinic, AP sites) are the most common DNA lesions generated by both spontaneous and induced base loss. In a previous study we have shown that circular plasmid molecules containing multiple AP sites are efficiently repaired by Chinese hamster extracts in an in vitro repair assay. An average patch size of 6.6 nucleotides for a single AP site was calculated. To define the exact repair patch, a circular DNA duplex with a single AP site was constructed. The repair synthesis carried out by hamster and human cell extracts was characterized by restriction endonuclease analysis of the area containing the lesion. The results indicate that, besides the repair events involving the incorporation of a single nucleotide at the lesion site, repair synthesis occurred also 3 to the AP site and involved a repair patch of approximately 7 nucleotides. This alternative repair pathway was completely inhibited by the presence in the repair reaction of a polyclonal antibody raised against human proliferating cell nuclear antigen. These data give the first evidence that mammalian cell extracts repair natural AP sites by two distinct pathways: a single nucleotide gap filling reaction targeted at the AP site and a proliferating cell nuclear antigen-dependent pathway that removes a short oligonucleotide containing the abasic site and 3-flanking nucleotides.
The base excision repair (BER) 1 pathway involves the action of specific DNA glycosylases that catalyze excision of the base, leaving noncoding apurinic/apyrimidinic (AP) sites in DNA. These lesions are further processed mainly by the abundant 5Ј-acting AP endonucleases, which cleave the DNA backbone generating a 5Ј-terminal deoxyribose-phosphate moiety. In Escherichia coli the 5Ј-nucleotidyl hydrolases, exonuclease III and endonuclease IV, constitute more than 90% of the total cellular AP endonuclease activity (1). In human cells the major 5Ј AP endonuclease is termed HAP1 or APE protein (2,3) and is the human homologue of the E. coli repair protein, exonuclease III. In E. coli as well as in mammalian cells, the BER excision step seems to involve only the deoxyribose-phosphate residue with generation of a single nucleotide gap (4). DNA polymerase I and DNA polymerase ␤ have been suggested as the major activities responsible for gap filling in E. coli and mammalian cells, respectively (4,18).
Recent data obtained both in prokaryotic and eukaryotic cells raise the question of the existence of an alternative BER pathway. Two activities have been detected in E. coli extracts that are able to excise free deoxyribose-phosphate from DNA: RecJ and MutM gene products. Nevertheless, E. coli fpg/recJ double mutants retain the ability to repair AP sites, suggesting the existence of backup systems (5). In a previous study (6) we have shown that circular plasmid molecules containing multiple AP sites are efficiently repaired by Chinese hamster cell extracts in an in vitro repair assay with an estimated repair patch of 6.6 nucleotides for a single AP site, suggesting that tracts longer than one nucleotide are synthesized at the lesion site. The analysis of repair of AP sites in a reconstituted repair system of Xenopus laevis oocytes has provided very convincing evidence that AP sites are repaired not only via a DNA polymerase ␤-dependent system but also by a PCNA-dependent system (7). Therefore, DNA polymerase ␦ or ⑀ seems to function in repair of this ubiquitous form of DNA damage. Accordingly, evidence has been recently presented that DNA polymerase ␦ is required for base excision repair of DNA methylation damage in Saccharomyces cerevisiae (8).
To investigate whether mammalian cells repair abasic sites via alternative repair pathways a circular DNA duplex molecule was constructed with a single AP site in a defined sequence and the repair patch at the lesion was measured in both hamster and human cell-free repair assays.
We provide the first evidence that, in addition to the single nucleotide insertion pathway, a PCNA-dependent longer patch repair system is active in mammalian cells.

Construction of DNA Molecules Containing a Single AP Site-An
oligonucleotide that contains a single uracil, 5Ј-GATCCTCTAGAGUC-GACCTGCA-3Ј, and a control oligonucleotide, 5Ј-GATCCTCTAGAGTC-GACCTGCA-3Ј, were prepared by automated DNA synthesis. Closed circular DNA was produced by priming single-stranded (ϩ) pGEM-3Zf DNA (Promega) with a 30-fold molar excess of uracil-containing or control oligonucleotide and incubating with T4 DNA polymerase holoenzyme, single-stranded DNA-binding protein and T4 DNA ligase (Boehringer Mannheim). Closed circular DNA duplex molecules were purified by cesium chloride equilibrium centrifugation. The single uracil-containing plasmid, pGEM-U, and the control oligonucleotide-containing plasmid, pGEM-T, were stored at Ϫ80°C. The characterization of the substrate was performed on duplex DNA molecules synthesized by using a 5Ј-end 32 P-labeled oligonucleotide. 30 ng of pGEM-U or pGEM-T DNA were then incubated with E. coli uracil-DNA glycosylase (1 ng, specific activity: 3.1 ϫ 10 5 units/mg) at 37°C for 30 min, followed by digestion with E. coli endonuclease III (0.45 g, specific activity: 3.8 ϫ 10 6 units/mg) at 37°C for 30 min when required. The incubation buffer for both enzymes was 7 mM Hepes-KOH (pH 7.8), 5 mM 2-mercaptoethanol, 2 mM EDTA disodium salt, and 35 mM NaCl. The uracil-DNA glycosylase and the endonuclease III, purified to apparent homogeneity as described in (9,10), were kindly provided by S. Boiteux, Institute Gustave Roussy, Villejuif, France. For the repair experiments, pGEM-U DNA was incubated with uracil-DNA glycosylase immediately before the repair reactions to generate the plasmid pGEM-X, which contains a single abasic site at a known position of its genome. The plasmid pGEM-T was also treated with uracil-DNA glycosylase before incubation with cell extracts.
In Vitro Repair Synthesis-Extracts from CHO-9 and HeLa cells were prepared as described (6). Repair reactions were carried out with 150 g of cell extract protein in reaction buffer containing 20 M dGTP, dCTP, dATP, 8 M dTTP, and 2 Ci of [ 32 P]dTTP (3000 Ci/mmol). When [ 32 P]dATP (3000 Ci/mmol) was added to the repair reactions, the concentration of dATP was lowered to 8 M. In the experiments carried out with human cell extracts the repair reaction buffer contained 10 M dGTP, dCTP, dTTP, and dATP and 5 Ci of [ 32 P]dTTP (3000 Ci/mmol) and/or [ 32 P]dCTP (3000 Ci/mmol). After 3 h at 30°C the DNA was recovered and digested with the appropriate restriction enzymes (40 units for 5 h at 37°C). The digestion products were resolved on 7% top/20% bottom polyacrylamide gels. Gels were exposed to Kodak XAR film with intensifying screens at Ϫ70°C.
Preparation of Anti-PCNA Antibody-A polyclonal monospecific anti-PCNA antibody, termed 3009, was raised by immunization of a rabbit with keyhole limpet hemeocyanin, as described elsewhere. 2 Antibody was purified from whole serum essentially as described by (11). Briefly, 1 ml of immune serum was incubated overnight at 4°C with 1 ml of protein G-Sepharose beads (Pharmacia Biotech Inc.) in phosphatebuffered saline (142 mM NaCl, 7.2 mM Na 2 HPO 4 , 3.26 mM NaH 2 PO 4 ). Beads were loaded onto a 1-ml column that was washed with 10 column volumes of phosphate-buffered saline, then antibody was eluted in 100 mM glycine HCl (pH 2.5) and 0.5-ml fractions neutralized with 50 l of 2 M Tris-HCl (pH 8.0). Peak antibody fraction was determined by Bradford assays and purity verified on Coomassie-stained polyacrylamide gels. Purified antibody retained the ability to recognize PCNA as determined by Western blotting.
Inhibition of Repair Synthesis by Anti-PCNA Antibodies-Protein extracted from CHO-9 cells were preincubated for 20 min at 30°C with the indicated amounts (from 0.15 to 4.6 g) of anti-PCNA polyclonal antibody. Plasmid DNA was then added in the repair reaction, which was allowed to proceed for 3 h at 30°C.

Synthesis of Duplex DNA Containing a Single AP Site-
Single-stranded (ϩ) pGEM-3zf DNA was primed with 5Ј-end 32 P-labeled uracil-containing oligonucleotide or control oligonucleotide and DNA was synthesized in vitro on this primed template ( Fig. 1). As shown in Fig. 2, in both cases closed circular molecules were obtained with an efficiency close to 100% (Fig. 2, A and B, lane 1). When the duplex DNA molecules containing the single uracil residue, pGEM-U, were digested with uracil-DNA glycosylase followed by endonuclease III, a complete conversion to Form II was observed ( Fig. 2A, lane 4), confirming that these duplex molecules are homogeneous closed circular molecules all containing a single uracil residue. Both pGEM-U and pGEM-T DNA were not significantly cleaved (conversion Form I to Form II Ͻ5%) by incubation with uracil-DNA glycosylase ( Fig. 2A, lane 2 and Fig. 2B, lane 4) or endonuclease III (Fig. 2A, lane 3 and Fig. 2B, lane 5) alone. pGEM-T and pGEM-X duplex molecules were further characterized by restriction endonuclease analysis (Fig. 2, B and C). As expected from the correct insertion of the primer oligonucleotide, the duplex molecules contained a single BamHI (Fig.  2B, lane 2, and Fig. 2C, lane 7) and a single AccI site (Fig. 2B, Fig. 2C, lane 8). In the case of pGEM-X (Fig. 2C), the presence of the AP site within the recognition sequence of AccI caused a partial resistance to cleavage by this restriction enzyme (Fig. 2C, lane 8). Interestingly, the presence of a methoxyamine-modified AP site in the same sequence leads to a complete inhibition of the cleavage of the lesion-containing strand (data not shown).
Repair of a Single AP Site by Hamster Cell Extracts-The ability of CHO-9 cell extracts to repair the single AP site was tested in vitro (Fig. 3). pGEM-T (lanes 1, 2, 5, and 6) and pGEM-X (lanes 3, 4, 7, and 8) plasmids, which are identical except for the presence of the AP site, were used as substrates to determine the localization and extent of repair synthesis stimulated by the excision of a single abasic site. The restriction fragments XbaI-HincII, containing the AP site and 6 nucleotides 5Ј to it (A), and SalI-PstI, containing the AP site and 9 nucleotides 3Ј to it (B), were selected for this analysis. The incorporation of radiolabeled dTMP and dAMP residues was determined in these two fragments, which contain the same number of Ade and Thy residues. Repair synthesis was exclusively AP site-dependent as shown by the lack of incorporation in the undamaged A and B fragments (lanes 1-2 and 5-6, respectively). In the case of the single AP site containing substrate (lanes 3, 4, 7, and 8), incorporation of radioactively labeled dTMP residues was observed in both fragments A (lane 4) and B (lane 8) with a higher incorporation in fragment B. These data indicate that: (i) the replacement of the AP site was predominantly with a dTMP residue, (ii) repair synthesis extended also 3Ј to the lesion, and (iii) no significant repair replication occurred 5Ј to the abasic site. The extension of the repair patch exclusively 3Ј to the abasic site was confirmed by the incorporation of labeled dAMP residues in fragment B (lane 7) but not in fragment A (lane 3). The background incorporation present over the top half of the gel is likely due to unspecific incorporation at nicks which occur randomly along the plasmid genome (12).
The extent of DNA repair replication around the AP site was further characterized by a panel of restriction enzyme diges- tions as shown in Fig. 4. In all the following experiments only the incorporation of labeled dTMP residues was evaluated. No incorporation was detected in the XbaI-SalI (A) and SmaI-XbaI (D) fragments extending 5Ј to the lesion, but excluding the AP site. These data together with the lack of dAMP incorporation observed in the XbaI-HincII fragment (see Fig. 3, lane 3), indicate that the incorporation found in the XbaI-AccI fragment (B), originally containing the AP site and the residues 5Ј to it, is exclusively due to the replacement of the AP site by a single nucleotide. Incorporation was also detected in the HincII-PstI fragment (C), which contains residues 3Ј to the AP site but not the AP site residue itself. No incorporation was found in the more distal 3Ј PstI-HindIII fragment (E).
These data confirm the occurrence of a repair patch extending 3Ј to the AP site. Since the incorporation of radiolabeled dTMP was detected at the Thy closest to the AP site (7 nucleotides 3Ј to the lesion), but not at the Thy located 15 nucleotides 3Ј to the AP site, the patch of resynthesis could be mapped to a tract of 7-14 nucleotides.
Identification of the DNA Polymerase Involved in the Long Patch Repair Synthesis at AP Sites-Gap filling of the single nucleotide patch at the AP site seems to be catalyzed by DNA polymerase ␤ (4, 18). In a search for the DNA polymerase involved in the longer patch repair synthesis, a polyclonal antibody directed against the DNA polymerase ␦/⑀ auxiliary factor PCNA was used. This antibody was raised against the C-terminal 15 amino acids of human PCNA and was able to inhibit completely SV40 replication in an in vitro assay. 2 The ability of this antibody to efficiently inhibit PCNA-dependent repair synthesis was tested in UV-damaged plasmid DNA (Fig.  5). CHO cell extracts were preincubated with 0.15, 1.5, 3.7, or 4.6 g of purified anti-PCNA antibody for 30 min at 30°C prior to repair reactions. As shown in Fig. 5 (right panel), the antibody was able to inhibit UV-induced repair synthesis mediated by hamster extracts (compare lanes 1 and 2 with lanes 3-6), causing a reduction of incorporation almost to background levels.
The role of PCNA in the AP site-induced repair synthesis was investigated. CHO cell extracts were preincubated with 3009 anti-PCNA antibody prior to repair reactions on AP sitecontaining plasmid DNA. As shown in Fig. 6, the repair synthesis 3Ј to the AP site (fragment B) was strongly inhibited by anti-PCNA antibody (compare lane 5 with lanes 6 and 7). In contrast, the 1-nucleotide insertion pathway (fragment A) was not inhibited under these conditions (lanes 1-3). The slight increase in repair synthesis observed in the presence of the antibody (compare lane 1 with lanes 2 and 3) indicates that some competition might exist between the two pathways.
These data demonstrate that two repair pathways are involved in AP site repair in mammalian cells. The single nucleotide gap filling is PCNA-independent, while the resynthesis of longer patches extending 3Ј to the AP site is PCNA-dependent. Therefore besides DNA polymerase ␤ (4, 18), PCNA-dependent polymerase(s) (DNA polymerase ␦ and/or ⑀) might be involved in AP site repair.
Repair of a Single AP Site by Human Cell Extracts-To investigate whether the size heterogeneity of repair patches at AP sites was a peculiar feature of hamster cell extracts, the repair of the single abasic site was analyzed by using HeLa cell extracts. In this case, prior to restriction digestions close to the AP site, the repaired pGEM-X duplex molecules were digested with FokI that presents multiple recognition sites on the plasmid genome. In particular, FokI cuts a fragment 1388 bp long, including the AP site where the abasic site is located 122 bp from the 5Ј terminus. The extent of both dTMP and dCMP incorporation was analyzed. As shown in Fig. 7, repair synthesis by HeLa cell extracts was observed in the SalI-PstI fragment (A, lanes 2, 4, and 6) as well as in the HincII-PstI fragment (B, lanes 3, 5, and 7) independently of the labeled nucleotide used. The identical fragments from the control plasmid DNA incorporated no label (data not shown). In the presence of [ 32 P]dTTP (lanes 4 and 5), the higher level of incorporation detected in fragment A as compared with fragment B indicates that the substitution of a single nucleotide at the AP site is the favored pathway in our cell extracts. The detection of a labeled repair patch in the presence of [ 32 P]dCTP (lanes 6 and 7), which is not incorporated at the lesions site, is the formal proof that resynthesis 3Ј to the AP site is indeed occurring also in human cells. The level of incorporation of dCMP in the two fragments was proportional to the number of Cyt residues present in the A or B fragments that might be substituted, indicating that the patch of resynthesis 3Ј to the lesion is at  1 and 6); XbaI and AccI to release the 7-bp fragment (B) containing the AP site and nucleotide residues 5Ј to it (lanes 2 and 7); HincII and PstI to release the 8-bp fragment (C) containing nucleotide residues flanking 3Ј to the AP site (lanes 3 and 8); SmaI and XbaI to release the 9-bp fragment (D) containing nucleotide residues 5Ј to the AP site (lanes 4 and 9); PstI and HindIII to release the 8-bp fragment (E) containing nucleotide residues 3Ј to the AP site (lanes 5 and 10). least 8 bp long. The same results were obtained with CHO cell extracts (data not shown). The most likely explanation for the different levels of incorporation at the lesion and 3Ј to it observed in the previously described experiments is the efficiency of the restriction cleavage in the vicinity of the lesion. The restriction with a multiply cutting enzyme (i.e. FokI) prior to the digestions in the proximity of the lesion allows to control the efficiency of restriction digestions by monitoring the appearance of a new fragment (in this case 124 or 122 bp long) easily recognizable on the gel (lanes 2-7).
The production of repair patches longer than one nucleotide was therefore confirmed in human cell extracts. The higher level of incorporation detected in fragment A, as compared with fragment B, is in agreement with the hypothesis that in mammalian cell extracts two alternative repair pathways for AP sites repair are present. DISCUSSION The BER pathway is a major cellular defense against both spontaneous and chemical/radiation-induced DNA damage. AP sites are generated as a central intermediate in this repair process.
The biological consequences of unrepaired AP sites are numerous. AP sites can block progression of RNA and DNA polymerases, resulting in impairment of gene transcription and DNA replication. If polymerases succeed in bypassing these lesions during transcription or replication, the effects may be even more deleterious for genome stability. Because AP sites are non-instructional lesions to polymerases, there is a high probability that a "wrong" base will be inserted opposite an AP site leading to gene mutations (13,14) or to mutated transcripts (15). Studies on yeast mutants defective in the major AP endonuclease, APN1, have shown that a failure to rapidly repair AP sites leads to a strong increase in spontaneous mutation frequency (16). This is presumably due to AP sites arising from the action of endogenous DNA damaging agents such as oxygen radicals or alkylating agents. Similarly, HeLa cell transfectants expressing HAP1 antisense RNA and unable to efficiently repair AP sites are hypersensitive to killing by a range of DNA-damaging chemicals and radiation, including oxygen radical-generating agents and alkylating agents (17). Thus, the repair of AP sites is essential for suppression of mutations and, therefore, vital for genome stability. It is not surprising if backup systems are employed for the repair of these very common and potentially harmful DNA lesions.
In this study we have provided clear evidence that natural AP sites are repaired by two distinct pathways by mammalian cell extracts. Besides the repair route which involves a single nucleotide gap filling reaction at the abasic site, a new pathway was identified that is PCNA-dependent and involves the replacement of a short oligonucleotide containing the AP site and 6 -13 nucleotides 3Ј to it. Similarly, a PCNA-dependent system for repair of AP sites has been previously identified in X. laevis oocytes (7).
How much is the relative contribution of the two pathways to the repair of abasic sites still remains an open question. The generation of a single nucleotide repair patch seems to be the favored route for AP site repair in mammalian cell extracts, although the PCNA-dependent pathway is undoubtedly active on natural AP sites. The short patch BER pathway has been shown to be specific for AP site repair, while the long patch BER is also able to repair AP site analogues like the tetrahydrofuran residues (7). It might well be that in vivo after cell damage the type and rate of AP site production determines whether the long patch BER pathway enters into action as backup system.
In a recent study aimed at identifying the DNA polymerase(s) requirement for BER in mammalian nuclear extract (18), only DNA polymerase ␤ was shown to be required for the gap filling step following uracil excision. However, it is important to notice that in this study a 51-bp synthetic oligonucleotide, e.g. a linear double-stranded molecule, containing a single uracil was used as substrate. Clear evidence has been provided recently that a stable RF-C⅐PCNA complex can be assembled on circular but not on linear DNA (19). It has been proposed that RF-C, after binding to DNA, might load PCNA onto circular duplex DNA and then act as a clamp to ensure the correct loading of the replicative polymerase at the synthesis site. The mechanism of assembly of polymerase ␦ or ⑀ auxiliary proteins on DNA would then be compromised on a linear substrate containing the AP site, thus favoring the PCNA-independent, polymerase ␤-mediated, repair pathway.
The repair of an AP site via 1-nucleotide replacement involves the generation of a single nucleotide gap as a reaction intermediate. DNA polymerase ␤ is able to catalyze both the excision of the 5Ј-terminal deoxyribose-phosphate at the incised AP site (20) and DNA synthesis to fill the gap. The repair of an AP site via the PCNA-dependent pathway might imply the formation of a flap structure intermediate. A DNA flap is a bifurcated structure composed of a double-stranded DNA and a displaced single-strand. It will be of interest to investigate whether, in the PCNA-dependent pathway, polymerase ␦/⑀ and 5Ј-3Ј-exonuclease of FEN-1/DNase IV (21,22) may function together to replace the damaged bases by nick translation.
In conclusion, these observations lead to a redefinition of the model of BER in mammalian cells. As proposed by Dianov and Lindahl (23) for E. coli cells, this excision process is likely to be a branched pathway in which, besides the very short patch pathway likely mediated by DNA polymerase ␤, another sys-FIG. 7. Repair of a single AP site by HeLa cell extracts. Autoradiograph of a nondenaturing polyacrylamide gel is shown. Repair replication was performed in the presence of [ 32 P]dTTP and/or [ 32 P]dCTP as indicated. pGEM-X DNA was digested with FokI (lane 1) followed by digestion with: SalI and PstI to release the 10-bp fragment (A) containing the AP site and nucleotide residues 3Ј to it (lanes 2, 4, and 6) or HincII and PstI to releases the 8-bp fragment (B) containing nucleotide residues 3Ј to the AP site (lanes 3, 5, and 7). tem enters into the scenario. This repair system results in the replacement of several nucleotides by a PCNA-dependent polymerase, DNA polymerase ␦ or ⑀.