Modulation of the 5'-deoxyribose-5-phosphate lyase and DNA synthesis activities of mammalian DNA polymerase beta by apurinic/apyrimidinic endonuclease 1.

The Ape1 protein initiates the repair of apurinic/apyrimidinic sites during mammalian base excision repair (BER) of DNA. Ape1 catalyzes hydrolysis of the 5'-phosphodiester bond of abasic DNA to create nicks flanked by 3'-hydroxyl and 5'-deoxyribose 5-phosphate (dRP) termini. DNA polymerase (pol) beta catalyzes both DNA synthesis at the 3'-hydroxyl terminus and excision of the 5'-dRP moiety prior to completion of BER by DNA ligase. During BER, Ape1 recruits pol beta to the incised apurinic/apyrimidinic site and stimulates 5'-dRP excision by pol beta. The activities of these two enzymes are thus coordinated during BER. To examine further the coordination of BER, we investigated the ability of Ape1 to modulate the deoxynucleotidyltransferase and 5'-dRP lyase activities of pol beta. We report here that Ape1 stimulates 5'-dRP excision by a mechanism independent of its apurinic/apyrimidinic endonuclease activity. We also demonstrate a second mechanism, independent of Ape1, in which conditions that support DNA synthesis by pol beta also enhance 5'-dRP excision. Ape1 modulates the gap-filling activity of pol beta by specifically inhibiting synthesis on an incised abasic substrate but not on single-nucleotide gapped DNA. In contrast to the wild-type Ape1 protein, a catalytically impaired mutant form of Ape1 did not affect DNA synthesis by pol beta. However, this mutant protein retained the ability to stimulate 5'-dRP excision by pol beta. Simultaneous monitoring of 5'-dRP excision and DNA synthesis by pol beta demonstrated that the 5'-dRP lyase activity lags behind the polymerase activity despite the coordination of these two steps by Ape1 during BER.

The repair of damaged DNA is a critical function in all organisms and is required to maintain the integrity of genetic information. DNA can be damaged by exposure of cells to ionizing radiation, ultraviolet light, and xenobiotic agents as well as by exposure to endogenously generated alkylating agents and reactive oxygen species (1,2). DNA base lesions are a common result of these genotoxic agents, which contribute to the formation of mutations that lead to cancer (3). In mamma-lian cells, these DNA lesions are repaired by enzymes of the base excision repair (BER) 1 pathway.
BER is initiated by a family of DNA glycosylase enzymes. These proteins recognize and excise modified bases from the sugar-phosphate backbone of DNA (4,5). The resultant apurinic/apyrimidinic (AP) site is both a repair intermediate and also a highly prevalent primary DNA lesion that is also created via acid-catalyzed hydrolysis of DNA bases (6,7). An estimated 10,000 spontaneously generated AP sites are formed daily in each human cell (8), in addition to the contributions made by the action of DNA glycosylases. Indeed, the steadystate level of AP sites has been estimated to be quite high in some tissue types, especially following oxidative stress (9 -11).
AP sites are processed by AP endonucleases, which are highly conserved in both prokaryotic and eukaryotic organisms. Mammalian Ape1 hydrolyzes the phosphodiester backbone of DNA 5Ј of AP sites to create nicks with 3Ј-hydroxyl and 5Ј-deoxyribose 5-phosphate (dRP) termini. DNA polymerase (pol) ␤ follows Ape1 and has two important functions during BER. In addition to its gap-filling DNA polymerase activity, pol ␤ also has an intrinsic 5Ј-dRP lyase activity that resides in an 8-kDa domain at its amino terminus (12). Repair of AP sites is completed by either DNA ligase I or III (13).
Ape1 (also known as Apex, HapI, or Ref1) is a 35-kDa multifunctional protein that possesses 3Ј-phosphodiesterase, 3Јphosphatase, and 3Ј 3 5Ј-exonuclease activities in addition to its 5Ј-AP endonuclease function (14,15). These 3Ј-repair activities are required for the repair of oxidatively damaged DNA (6), mismatches arising from polymerase misincorporation (16,17), and 3Ј-blocking groups produced as repair intermediates by the AP lyase activity of bifunctional DNA glycosylases (18). Ape1 potentiates the DNA binding activities of several transcription factors via a redox mechanism (reviewed in Ref. 19) and also forms transcriptional repressor complexes for genes containing nCaRE sequences (20 -22). Recently, Ape1 was also found to be an important proteolytic target of the cytotoxic T lymphocyte granzyme A-mediated cell death pathway (23).
Both Ape1 and pol ␤ are critical for development, as mouse knockouts for either gene exhibit an embryonic lethal phenotype (24 -27). pol ␤ is a bifunctional protein with deoxynucleotidyltransferase and 5Ј-dRP lyase activities (12). Cells lacking pol ␤ are hypersensitive to alkylating agents and other DNA base-damaging agents. However, this phenotype can be complemented by expression of a mutant form of pol ␤ that is active only as a 5Ј-dRP lyase and not as a DNA polymerase, which shows that the critical role of pol ␤ lies in its dRP lyase function (28).
The 5Ј-dRP lyase activity of pol ␤ proceeds via formation of a Schiff base intermediate between amino acid lysine 72 and the Ape1-incised 5Ј-abasic residue (29,30). This intermediate can be captured by reduction using sodium borohydride (31). The analogous reaction can also result in the formation of a stable pol ␤-DNA cross-link during attempted repair of 5Ј-incised deoxyribonolactone lesions (32). The 5Ј-dRP excision step has been proposed as the rate-limiting step of BER based on a comparison of kinetic constants from several sources (33). We demonstrated previously that Ape1 interacts physically with pol ␤ to recruit the polymerase to the 5Ј-incised AP site and to stimulate excision of the 5Ј-dRP moiety (34). However, the nature and mechanism of this stimulation remained unclear. In this work, we further investigated the ability of Ape1 to modulate two key functions of pol ␤ during BER.

EXPERIMENTAL PROCEDURES
Reagents and Enzymes-Urea was obtained from American Bioanalytical (Natick, MA). A 40% acrylamide-bisacrylamide solution (29:1 ratio) was purchased from Bio-Rad. Radionuclides [␣-32 P]dCTP and [␥-32 P]ATP were obtained from PerkinElmer Life Sciences. Uracil DNA glycosylase, T4 polynucleotide kinase, and the Klenow fragment (3Ј 3 5Ј-exonuclease-free) of Escherichia coli DNA polymerase I were purchased from New England Biolabs (Beverly, MA). The wild-type pol ␤ protein was kindly provided by Drs. Rajendra Prasad and Sam Wilson, NIEHS, National Institutes of Health (Research Triangle Park, NC). The wild-type and D283A/D308A mutant forms of Ape1 were purified as described previously (35). The R177A mutant form of Ape1 was provided by Drs. Tadahide Izumi and Sankar Mitra, University of Texas Medical Branch (Galveston, TX). E. coli Fpg DNA glycosylase was a gift from Dr. Arthur Grollman, State University of New York (Stony Brook, NY). E. coli endonuclease IV was prepared as described previously (36).
DNA Substrates-Oligonucleotides were synthesized, and purified by high pressure liquid chromatography was purified at Operon Technologies (Alameda, CA). DNA substrates used in this study are shown in Table I. Substrates were 3Ј-end-labeled by incorporation of [␣-32 P]dCTP using the exonuclease-free Klenow fragment of DNA polymerase I. For polymerase primer extension assays, oligonucleotides were 5Ј-end-labeled using T4 polynucleotide kinase and a molar excess of [␥-32 P]ATP. Unincorporated label was removed using Micro Biospin P-30 columns (Bio-Rad), following the manufacturer's protocols. Double-stranded DNA substrates containing 5Ј-incised AP sites or tetrahydrofuran residues were prepared just prior to use by treatment with a catalytic amount of uracil-DNA glycosylase and either Ape1 or endonuclease IV. For Ape1 pretreatment, 25 fmol of Ape1 protein was incubated with 1000 fmol of substrate DNA; thus, with the typical assay containing 10 nM substrate, the residual amount of Ape1 in the pol ␤ reactions was 0.25 nM.
Enzyme Assays-The pol ␤ primer extension and 5Ј-dRP lyase reactions were performed at 30°C in standard reaction buffers containing 50 mM HEPES-KOH (pH 7.5), 5% (v/v) glycerol, 0.5 mM dithiothreitol, 0.1 mg/ml bovine serum albumin, 10 nM DNA substrate, either 0 or 8 mM MgCl 2 , and either 0 or 2 mM EDTA (37). Nucleotide concentrations, enzyme concentrations, and incubation times are indicated in the figure legends. Unless otherwise noted, reactions were initiated by addition of pol ␤. The 5Ј-dRP lyase reactions were stopped at the indicated times by addition of 300 mM NaBH 4 and then desalted by ethanol precipitation with the addition of glycogen as carrier. After recovery, samples were resuspended in formamide loading buffer (90% formamide, 10 mM EDTA, bromphenol blue, and xylene cyanol), boiled for 2 min, and then resolved by electrophoresis on 20% denaturing polyacrylamide gels containing 7 M urea. Polymerase chain reactions were stopped by addition of formamide loading buffer, boiled for 2 min, and then resolved by electrophoresis on 14% denaturing polyacrylamide gels. Gels were dried, analyzed by a Molecular Imager System (model GS-525, Bio-Rad), and quantified using Molecular Analyst software (Bio-Rad). The percentage of total dRP excised was calculated by dividing the amount of the dRP lyase product formed in each reaction by the sum of this product and the amount of the substrate DNA containing intact 5Ј-dRP. The percentage of total substrate extended in the primer extension assays by the nucleotidyltransferase activity of pol ␤ was calculated in a similar fashion by dividing the amount of the 22-mer primer extension product by the sum of the 21-mer substrate and the 22-mer products.

Stimulation of the pol ␤ 5Ј-dRP Lyase Activity by Two
Distinct Mechanisms-Work from this laboratory demonstrated previously (34) that Ape1 interacts with pol ␤ to stimulate excision of 5Ј-dRP. We explored the mechanism of this stimulatory effect further by investigating the ability of Ape1 to modulate the 5Ј-dRP lyase activity of pol ␤ in the presence of magnesium chloride and nucleoside triphosphates, which support DNA synthesis by pol ␤. The 3Ј-endlabeled DNA substrate containing a 5Ј-dRP migrates more slowly than the product of dRP excision and can be resolved by 20% denaturing PAGE following reduction with NaBH 4 to stabilize uncleaved 5Ј-dRP. Kinetic studies indicated that 5Ј-dRP excision by pol ␤ is not significantly affected by the presence of 8 mM MgCl 2 (Fig. 1A, panels a and b, left side). Consistent with previously published data (34), addition of 50 nM Ape1 stimulated 5Ј-dRP excision but only in buffer containing EDTA and not in buffer containing MgCl 2 (Fig. 1A, a The 32 P-labeled nucleotide is shown in boldface. b F and U denote tetrahydrofuranyl and uridine residues, respectively. c Preincised AP DNA substrates were created following treatment of oligonucleotides with catalytic amounts of uracil DNA glycosylase and Ape1. d Gap substrates were created by annealing together three oligonucleotides. The 5Ј terminus of the downstream oligonucleotide on the top strand was chemically phosphorylated. panels a and b, right side). When 20 M dCTP was included in the reactions, we observed an opposite effect. The dRP lyase activity was enhanced by dCTP in buffer containing MgCl 2 but not in buffer containing EDTA (Fig. 1A, panels c and d, left side). This increase in dRP excision by pol ␤ occurred independently of Ape1, which stimulated 5Ј-dRP excision in buffer containing EDTA and dCTP (Fig. 1A, panel c, right side) but did not further stimulate dRP excision in buffer containing both MgCl 2 and dCTP (Fig. 1A, panel d, right side). A slight stimulation of 5Ј-dRP excision by pol ␤ under conditions that support DNA synthesis was noted by another group but was not analyzed in depth (33). We also detected a modest increase (typically 10 -15%) in the spontaneous loss of the labile 5Ј-dRP moiety from DNA in reactions containing both EDTA and Ape1 (Fig. 1A, panels a and c). This increase occurred before reactions were initiated by addition of pol ␤ and was not observed in reactions containing MgCl 2 . The cause for this increase in background dRP loss is currently under investigation.
We next sought to quantify the ability of Ape1 to stimulate dRP excision by pol ␤ in magnesium-free buffer (Fig. 1B). Whether or not magnesium was included in the reaction, 5Ј-dRP excision proceeded at a slow rate when only pol ␤ was present. A comparison of estimated initial rates from these reactions indicated that addition of Ape1 stimulated 5Ј-dRP excision by pol ␤ at least 9-fold. As observed previously, the stimulation occurred only in buffer containing EDTA, and the effect was abrogated by addition of MgCl 2 (Fig. 1B). Because the release of Ape1 from its 5Ј-AP site incision product is induced by magnesium (38), these data suggest that the mechanism employed by Ape1 to stimulate 5Ј-dRP excision by pol ␤ likewise involves the tight binding of Ape1 to its 5Ј-incised abasic DNA product.
We subsequently investigated the stimulatory effect of dCTP on 5Ј-dRP excision by pol ␤ in buffer containing MgCl 2 . The dRP excision reactions were performed in the absence of dCTP or in the presence of either 1 or 20 M dCTP. The 5Ј-dRP lyase activity was enhanced at least 2-fold at all pol ␤ concentrations when dCTP was present in the reaction, and the enhancement was maximal at 20 M dCTP (Fig. 1C). Similar results were obtained when all four deoxynucleoside triphosphates were included in the reaction (data not shown). The apparent K m of pol ␤ for dCTP on single-nucleotide gapped DNA has been calculated to be 0.2 M but can vary depending on the structure of the DNA substrate (39,40). Although the K m value of pol ␤ for dCTP on 5Ј-incised abasic DNA has not been reported, it is likely that the 20 M concentration used here is saturating.
Requirement of the Correct Incoming Nucleotide to Stimulate 5Ј-dRP Excision by pol ␤-Our results showing that dCTP can stimulate 5Ј-dRP excision by pol ␤ suggested that the mechanism of stimulation may involve either general pol ␤ protein conformational changes upon nucleoside triphosphate binding or altered protein-DNA interactions accompanying dCMP incorporation. Because of the location of the 32 P label on the DNA substrate, we could not ascertain from the experiments of Fig.  1 which of these possibilities led to the enhanced dRP lyase activity. We therefore sought to analyze further the 5Ј-dRP lyase activity under conditions that support DNA synthesis. The DNA substrate we utilized contained a template G opposite a preincised abasic site, which allowed for the incorporation of dCMP by pol ␤. We observed enhanced 5Ј-dRP excision in reactions containing dCTP, but not in reactions containing dTTP, which cannot be efficiently incorporated by pol ␤ opposite the template G ( Fig. 2A). We observed a slight reduction in the stimulatory effect of dCTP when Ape1 was included in the reaction, but Ape1 otherwise had no measurable effect in reactions containing dTTP or lacking dNTPs altogether. In contrast, neither dCTP nor dTTP had any effect on 5Ј-dRP excision by pol ␤ in reactions lacking magnesium, when DNA synthesis was not possible (Fig. 2B). However, Ape1 did stimulate 5Ј-dRP excision under these conditions.
The Effect of Nuclease-defective Ape1 Proteins on dRP Excision by pol ␤-The ability of Ape1 to enhance dRP excision only in the absence of magnesium indicated that this stimulatory mechanism likely involves binding by Ape1 to its incised AP-DNA product. To investigate this mechanism further, we assessed the ability of two mutant forms of Ape1 to stimulate 5Ј-dRP excision by pol ␤. The R177A mutant of Ape1 was shown to have a 2-fold reduced affinity for intact abasic sites relative to wild-type Ape1 but was 25% more active overall as an AP endonuclease because of a complementary increase in k cat (14). Crystal structures of Ape1 in complex with duplex DNA containing an abasic residue indicate that Arg-177 makes contact  Table I  with the nonterminal phosphate adjacent to the incised abasic residue. This phosphate becomes the 5Ј-terminal phosphate moiety following 5Ј-dRP excision by pol ␤. The aspartic acid residues at positions 283 and 308 of Ape1 are conserved in all members of the Ape1/E. coli exonuclease III family of class II AP endonucleases (15). Substitution of these aspartic acid residues with alanine results in a mutant form of Ape1 that possesses only ϳ1% of the AP endonuclease activity of wildtype Ape1 (35) but with no measurable 3Ј 3 5Ј-exonuclease or 3Ј-phosphodiesterase activities (17). This mutant protein also exhibits a 2-fold increased half-life for binding to its 5Ј-incised abasic DNA product (35).
Stimulation of the pol ␤ dRP lyase by the wild-type and mutant Ape1 proteins occurred in reactions containing EDTA, but not with magnesium (Fig. 3A). Quantification of these results demonstrated that the D283A/D308A protein enhanced 5Ј-dRP lyase activity of pol ␤ nearly 2-fold more strongly than did wild-type Ape1 (Fig. 3B). We observed no significant difference between the R177A and wild-type forms of Ape1 in their ability to stimulate 5Ј-dRP excision. The stimulatory effect of all three Ape1 proteins was abrogated by addition of magnesium to the reactions (Fig. 3C), which has been shown to destabilize the Ape1-5Ј-incised DNA complex (35).
Modulation of the Polymerase Activity of pol ␤ at 5Ј-Incised Abasic DNA-We next investigated the effect of Ape1 on the DNA synthesis activity of pol ␤ on a duplex oligonucleotide substrate containing a 5Ј-incised tetrahydrofuran residue. Tetrahydrofuran is a chemical analog of an abasic site that can be recognized and hydrolytically cleaved by Ape1 and other AP endonucleases with kinetics similar to the incision of unmodified AP sites (41)(42)(43). The resultant 5Ј-incised abasic residue cannot undergo ␤-elimination and therefore cannot be excised by pol ␤ (29,44). However, this incised intermediate is able to form a stable enzyme-product complex with Ape1 in the presence of EDTA (38).
Our results demonstrate that pol ␤ can catalyze dCMP insertion at double-stranded DNA containing 5Ј-incised tetrahydrofuran residues (Fig. 4A). However, addition of Ape1 inhib-ited DNA synthesis by pol ␤ about 3-fold (estimated from initial rates shown in Fig. 4A). This inhibition occurred even though the reactions contained magnesium, which promotes the dissociation of Ape1 from its incision product (38). The polymerase inhibition was likely not due to substrate degradation by the 3Ј 3 5Ј-exonuclease activity of Ape1, which is active at singlenucleotide gaps but is poorly efficient on substrates containing 5Ј-incised tetrahydrofuran (17). Furthermore, the inhibitory effect required fully active Ape1 protein, because the D283A/ D308A mutant of Ape1 did not affect dCMP incorporation by pol ␤ (Fig. 4A). The D283A/D308A mutant of Ape1 was not merely inactivated protein, as it retained the ability to stimulate 5Ј-dRP excision by pol ␤, possessed measurable AP endonuclease activity, and bound AP sites with the same affinity as wild-type Ape1 (data not shown).
The inhibition of the pol ␤ polymerase activity by wild-type Ape1 was specific for DNA containing 5Ј-incised abasic residues because no inhibition was observed on substrates containing a single-nucleotide gap lacking an abasic residue (Fig. 4B). To verify that the observed inhibition of the pol ␤ polymerase activity by Ape1 was not a feature specific to DNA containing tetrahydrofuran residues, we also performed the polymerase primer extension assays on substrates containing uracil-de-  Table I Table I  rived AP sites. DNA containing a 5Ј-incised AP site was generated by treating a substrate containing a central uracil residue with catalytic amounts of E. coli uracil-DNA glycosylase and Ape1 protein. Consistent with results in Fig. 4A, we observed that pol ␤ could insert dCMP at 5Ј-incised abasic sites (Fig. 4C, filled circles) but that addition of Ape1 inhibited DNA synthesis by about 3-fold (Fig. 4C, open circles). However, the inhibitory effect was not as great as that observed in Fig. 4A, perhaps due to spontaneous loss of some fraction of the labile 5Ј-dRP moiety. Ape1 did not inhibit pol ␤ when the 5Ј-dRP was first excised by pretreatment of the Ape1 incision product with E. coli Fpg protein (Fig. 4C, open triangles), thereby corroborating our results showing that Ape1 can inhibit pol ␤ specifically at 5Ј-incised abasic DNA but not at single-nucleotide gaps. These two substrates mimic DNA repair intermediates just prior to and following 5Ј-dRP excision, which suggests that the polymerase activity of pol ␤ can be coordinated by Ape1 in the presence of the 5Ј-dRP repair intermediate.
Concentration-dependent Inhibition of pol ␤ by Ape1 and E. coli Endonuclease IV-We next investigated the effect of Ape1 concentration on the inhibition of dCMP incorporation by pol ␤. Wild-type Ape1 inhibited the polymerase activity of pol ␤ on DNA containing 5Ј-incised tetrahydrofuran by up to 60% at the highest concentration of Ape1 assayed (Fig. 5B, gray bars). No significant inhibition of pol ␤ was observed on the single-nucleotide gap substrate (Fig. 5B, black bars). These data confirmed that inhibition of pol ␤ by Ape1 occurs specifically at 5Ј-incised abasic sites. The polymerase activity was not inhibited when boiled Ape1 was added to the reactions (data not shown), which indicates that the inhibitory effect was not due to a buffer contaminant. Furthermore, the D283A/D308A mutant form of Ape1 did not inhibit dCMP incorporation by pol ␤ on either substrate (Fig. 5C). Most interesting, E. coli endonuclease IV, which is structurally unrelated to Ape1 but shares the same enzymatic functions, also inhibited pol ␤ at incised abasic DNA. However, endonuclease IV inhibited pol ␤ on single-nucleotide gapped DNA as well, albeit to a lesser extent (Fig. 5D). Following incision, Ape1 and endonuclease IV both remain tightly bound to their abasic DNA products (35,45), so these two proteins may inhibit DNA synthesis by pol ␤ using a similar mechanism. However, the substrate-specific inhibition of pol ␤ by Ape1, and the apparent lack of specificity of endonuclease IV, may reflect different modes of binding employed by these two enzymes (14). Both proteins induce DNA bends, but Ape1 induces a relatively mild 35°bend compared with the 90°kink caused by endonuclease IV (45,46).
Simultaneous  Table I (n ϭ 3). B, primer extension reactions containing 51Gap DNA were performed and analyzed as described for A. C, pol ␤ primer extension assays on preincised 51AP DNA. Reactions were performed and analyzed as described for A, except reactions contained 0.5 nM pol ␤, 10 nM 5Ј-incised 51AP DNA (circles), or preincised 51AP DNA pretreated with 10 nM (Fpg) for 30 min to excise 5Ј-dRP (triangles). Reactions included either no added Ape1 (filled circles and triangles) or 50 nM Ape1 (open circles and triangles).

FIG. 5. Specificity of Ape1 for inhibition of DNA synthesis by pol ␤ at 5-incised abasic DNA.
A, schematic diagram of oligonucleotide substrates (see Table I  Ape1-mediated Stimulation of Polymerase ␤ dRP Lyase that Ape1 modulates both the 5Ј-dRP lyase and DNA polymerase activities of pol ␤ under different buffer conditions. Therefore, it seemed possible that Ape1 might redirect pol ␤ activities to favor dRP excision prior to repair synthesis. To study the modulation of these activities by Ape1 together rather than in separate reactions, we used [␣-32 P]dCTP and 3Ј-32 P-labeled DNA substrates to allow for the simultaneous monitoring of both 5Ј-dRP excision and dCMP incorporation. This approach has been used to compare the relative rates of dRP excision and DNA synthesis by pol ␤ in a partial reconstitution of BER (33). However, the 1:1 ratio of Ape1 and pol ␤ employed in the referenced study did not reflect biologically relevant levels of the two enzymes, which has been estimated to range from 7:1 up to 140:1 depending on the cell type (47,48).
The results show that [␣-32 P]dCTP incorporation by pol ␤ in the absence of Ape1 occurred at comparable rates at 5Ј-incised tetrahydrofuran (Fig. 6A, lanes 1-5, lower bands), 5Ј-incised AP sites (lanes 11-15), and single-nucleotide gaps (lanes [21][22][23][24][25]. pol ␤ can still bind to both the 5Ј-tetrahydrofuran and 5Ј-dRP moieties of each substrate, although the dRP lyase would not be productive at 5Ј-incised tetrahydrofuran DNA. However, this competitive binding did not result in a significant difference in polymerase activity on any of the three substrates. As expected, addition of Ape1 inhibited dCMP incorporation on the two substrates containing 5Ј-incised abasic residues (Fig. 6A, lanes 6 -10 and 16 -20). A small reduction in dCMP incorporation by pol ␤ was observed on the single-nucleotide gapped substrate as well when Ape1 was included (Fig.  6A, lanes 26 -30). This inhibition is likely due to degradation of the internal 3Ј terminus of the single-nucleotide gap by the 3Ј 3 5Ј-exonuclease activity of Ape1, which produces a small amount of substrate bearing a template T (Fig. 6B) that cannot support efficient dCMP incorporation by pol ␤. Ape1 exonuclease might also excise labeled dCMP incorporated by the polymerase, but if this were the only Ape1 effect, it should be greater for the gapped substrate than for the incised abasic substrates (17), contrary to the observed results (Fig. 6A). The modest inhibition by Ape1 of the gap substrate in this experiment compared with the data for Fig. 4 and Fig. 5 derives from the fact that, for those experiments, the primer was 5Ј-labeled, so that N-1 products could be accounted for, whereas in the experiment of Fig. 6A, the incoming nucleotide was labeled, so that shortened primers were not observable.
As expected, no excision of the abasic residue was observed at 5Ј-incised tetrahydrofuran (Fig. 6A, lanes 1-10, upper bands) as compared with the single-nucleotide gap, which lacked the abasic residue altogether (Fig. 6A, lanes 21-30). Under the conditions employed in this assay, very little 5Ј-dRP excision was observed on the 5Ј-incised AP DNA substrate (Fig. 6A,  lanes 11-15), and addition of Ape1 did not affect dRP excision (Fig. 6A, lanes 16 -20). We repeated this experiment using a higher concentration of pol ␤ to drive the dRP lyase activity FIG. 6. Simultaneous monitoring of pol ␤ polymerase and dRP lyase activities. A, dual polymerase and dRP lyase assays. Reactions contained 1 nM pol ␤, 1.3 M [␣-32 P]dCTP, 0 or 50 nM Ape1, and 3Ј-32 P-labeled preincised 52F DNA, preincised 52AP DNA, or 52Gap DNA, as indicated above each panel. Reactions were initiated by addition of oligonucleotide substrate and incubated at 30°C. Aliquots of each reaction were removed after 0, 1.5, 3, 4.5, or 10 min (indicated by the wedges) and then stopped by addition of 300 mM NaBH 4 and formamide loading buffer. The migration of the radiolabeled 30-mer DNA substrate containing intact 5Ј-dRP, 30-mer 5Ј-dRP lyase product, and the 22-mer primer extension product are marked by the open, filled, and solid arrows, respectively. nt, nucleotide. B, Ape1 exonuclease activity under pol ␤ synthesis conditions. The substrates are indicated to the left of the panel. Every sample contained 0.5 nM pol ␤, with the addition of wild-type or the D283A/D308A mutant Ape1 protein as indicated. The reactions were conducted as described for A, except that the primer strand was labeled at the 5Ј end; the dCTP was not labeled, and time points were taken at 0, 2, 4, or 8 min. The band below full-length primer (especially evident with Ape1 added to the gap substrate) corresponds to the N-1 exonuclease product; there is also a very small amount of N-2 seen with the longest incubation time. C, reactions containing 17 nM pol ␤ were performed and analyzed as described for A. toward completion. Under these new conditions, dCMP incorporation had already plateaued at the earliest time point even though 5Ј-dRP excision had not yet reached saturation (Fig. 6C,  lanes 1-5). These results therefore indicate that 5Ј-dRP excision lags behind DNA synthesis by pol ␤ in our partial reconstitution experiment. The addition of Ape1 affected neither dCMP incorporation (Fig. 6C, lanes 6 -10, lower bands) nor dRP excision (Fig. 6C, lanes 6 -10, upper bands). Thus the modulating effects of Ape1 on pol ␤ do not seem to change the order of the enzymatic steps (DNA synthesis first, followed by dRP excision) in this simplified system. DISCUSSION Base excision repair is an orchestrated process that involves a series of protein-protein and protein-DNA interactions. These interactions serve to facilitate repair and limit exposure of potentially reactive DNA repair intermediates to the cellular milieu (49). To maximize the efficiency of BER, nature has evolved these proteins to possess multiple functions encoded within a single polypeptide. Ape1 and pol ␤ are both multifunctional proteins with critical roles during BER. We demonstrated previously that Ape1 interacts with pol ␤ to stimulate dRP excision (34). In this study, we examined the functional interaction between these two enzymes by analyzing the ability of Ape1 to modulate 5Ј-dRP excision and DNA synthesis by pol ␤.
The 5Ј-dRP lyase activity of pol ␤ does not require divalent cations because it proceeds via ␤-elimination instead of hydrolysis (12,30). However, under certain conditions MgCl 2 may provide a salt effect that enhances dRP excision by enhancing protein-DNA interactions (12,50). This salt effect can also be mimicked by addition of 20 mM of the sodium salt of EDTA (30). Our results show that dRP excision proceeds slowly in the absence of MgCl 2 but is stimulated by at least 9-fold with the addition of Ape1. Furthermore, the D283A/D308A mutant form of Ape1 was more effective at stimulating 5Ј-dRP excision than the wild-type protein. The D283A/D308A protein binds incised AP DNA with the same affinity as wild-type Ape1 but forms a complex with an extended half-life (35). Consistent with the idea that formation of a long lived Ape1-DNA complex is responsible for stimulation of the relatively slow dRP excision reaction, Ape1 did not stimulate the dRP lyase activity in the presence of magnesium. Magnesium is required for Ape1 enzymatic activity but also destabilizes the complex of Ape1 with the incised DNA product to facilitate turnover (38). The E96Q mutant form of Ape1 may be an interesting enzyme for subsequent analysis, because it has been reported to bind tightly to DNA containing tetrahydrofuran residues even in the presence of magnesium (51). However, pol ␤ undergoes significant conformational changes upon binding to divalent cations and during catalysis (52), so we cannot rule out the possibility that the absence of magnesium constrains pol ␤ in a conformation favorable for stable physical interactions with Ape1.
Our results show that dRP excision by pol ␤ can also be stimulated in the presence of MgCl 2 and dNTPs, which are required to support DNA synthesis. This second mode of dRP lyase stimulation functions independently of Ape1 and requires the appropriate incoming nucleoside triphosphate to be present. By using these assays, we cannot distinguish whether binding to the correct incoming nucleotide is sufficient or whether nucleotidyltransferase activity is required to stimulate dRP excision. pol ␤ requires divalent cations to bind dNTPs and adopts multiple conformational changes upon metal binding, dNTP binding, DNA substrate binding, and during catalysis. Stabilization of any one of these conformations may result in enhanced dRP excision. Consistent with this idea, we have observed enhanced pol ␤-DNA cross-linking in sodium borohy-dride trapping experiments performed in the presence of EDTA when Ape1 or a dNTP was also present (data not shown). Interaction of pol ␤ with Ape1 or other known interacting proteins such as Xrcc1, p53, DNA ligase I, or proliferating cell nuclear antigen (53-57) may cause further conformational changes. The influence of these other proteins on the two enzymatic activities of pol ␤ remains to be determined.
We also demonstrated in this study that the deoxynucleotidyltransferase activity of pol ␤ on 5Ј-incised abasic DNA can be suppressed by Ape1. Most interesting, this suppression occurs even in the presence of magnesium, which promotes dissociation of Ape1 from DNA and thus does not support the formation of stable protein-DNA contacts. Moreover, the D283A/D308A mutant of Ape1, which can form more stable interactions with DNA, was unable to inhibit DNA synthesis by pol ␤. E. coli endonuclease IV inhibited pol ␤ at both 5Ј-incised abasic residues and at single-nucleotide gaps. The altered specificity of endonuclease IV may reflect the 90°bend of the abasic DNA induced by endonuclease IV (45). pol ␤ itself induces a 90°bend in the DNA and may not be able to recognize DNA bound by endonuclease IV that is already severely kinked (52). By comparison, Ape1 induces a 35°bend in DNA, which may facilitate transfer of the DNA repair intermediate to pol ␤ (14). The evidence also suggests a general conformational difference in Ape1 bound to different DNA molecules; on incised abasic substrates, the enzyme is bound in such a way that both its intrinsic exonuclease activity and the polymerase activity of pol ␤ are suppressed. At the same time, this conformation would allow for an efficient handoff of the 5Ј-terminal dRp from Ape1 to pol ␤. This mode of binding by Ape1 may involve its specific contacts with the incised abasic residue.
Successful completion of BER requires both gap-filling synthesis and 5Ј-dRP excision by pol ␤. By inhibiting the polymerase activity of pol ␤ on DNA containing incised abasic sites, Ape1 may direct pol ␤ to first excise the dRP moiety. The 3Ј 3 5Ј-exonuclease activity of Ape1 has been proposed as a proofreading activity for pol ␤, which is error-prone and lacks its own intrinsic proofreading activity. Logically, this proofreading exonuclease activity should follow DNA synthesis by pol ␤. We demonstrated in another study that the exonuclease activity of Ape1 is inhibited by the presence of a 5Ј-incised abasic residue (17). Excision of the 5Ј-dRP allows the exonuclease activity to proceed at both gaps and nicks, thereby temporally placing the proofreading activity of Ape1 after both dNTP insertion and 5Ј-dRP excision by pol ␤. However, simultaneous monitoring of both the dRP lyase and the DNA synthesis steps indicated that 5Ј-dRP excision lags behind DNA synthesis (Fig. 6). Although Ape1 stimulated dRP excision in the absence of magnesium, we observed no stimulation by Ape1 under conditions that also support DNA synthesis. However, most enzymes work at substrate concentrations far below their K m values, so the real environment for Ape1 and pol ␤ likely lies somewhere in between the "all or nothing" conditions employed in our study (58).
Inhibition of the nucleotidyltransferase activity of pol ␤ may also be a mechanism whereby Ape1 functions as a molecular switch between the short and long patch BER pathways. Tetrahydrofuran is a form of reduced abasic DNA that serves as a useful model for the study of modified abasic lesions that are resistant to ␤-elimination by the 5Ј-dRP lyase activity of pol ␤ (41). The oxidative lesion 2-deoxyribonolactone is another such lesion, which forms by oxidation at the C-1Ј carbon of AP sites and cannot undergo ␤-elimination chemistry (6). These and other abasic lesions resistant to excision by pol ␤ are likely repaired by the long patch BER pathway (6). This pathway of repair involves other DNA polymerases, which may be more effective at displacing Ape1 than is pol ␤, to bypass the 5Јterminal lesion by strand displacement synthesis (59). Preliminary data indicate that dCMP insertion by the Klenow fragment of E. coli DNA polymerase I is not inhibited by Ape1 (data not shown), which may indicate that the more processive eukaryotic DNA polymerases ␦ and ⑀ can also overcome the inhibitory effects of Ape1 on repair synthesis. However, the contribution of pol ␤ accessory proteins also needs to be investigated, because pol ␤ has also been implicated in the long patch BER pathway in some studies (60 -62). These accessory proteins include proliferating cell nuclear antigen, which interacts physically with both Ape1 (63) and pol ␤ (57), and Fen1, which has been shown to interact physically with Ape1 (63) and functionally with pol ␤ to promote strand-displacement synthesis (61). It should be noted that Fen1 efficiently excises 5Ј-dRP, but only if the lesion is first displaced by at least one nucleotide during strand-displacement synthesis (64).
Work from this laboratory has shown that pol ␤ forms a stable cross-link to 2-deoxyribonolactone during an abortive attempt to excise the lesion (32). Thus, Ape1 may also help limit formation of this potentially catastrophic lesion by preventing pol ␤ from accessing the repair site. Because Ape1 can modulate both the deoxynucleotidyltransferase and 5Ј-dRP lyase activities of pol ␤, it will be interesting to determine whether Ape1 can also modulate the activities of DNA polymerases ␥, , and . Like pol ␤, these proteins also possess both dRP lyase and nucleotidyltransferase activities (65)(66)(67). A better understanding of the coordination of the different enzymatic activities at the core of this critical mammalian DNA repair pathway will shed light on the mechanisms employed by cells to maintain their genetic information.