Dynamics of the Interaction of Human Apurinic Endonuclease (Ape1) with Its Substrate and Product*

We investigated the interaction dynamics of human abasic endonuclease, the Ape1 protein (also called Ref1, Hap1, or Apex), with its DNA substrate and incised product using electrophoretic assays and site-specific amino acid substitutions. Changing aspartate 283 to alanine (D283A) left 10% residual activity, contrary to a previous report, but complementation of repair-deficient bacteria by the D283A Ape1 protein was consistent with its activity in vitro. The D308A, D283/D308A double mutant, and histidine 309 to asparagine proteins had 22, 1, and ∼0.02% of wild-type Ape1 activity, respectively. Despite this range of enzymatic activities, all the mutant proteins had near-wild-type binding affinity specific for DNA containing a synthetic abasic site. Thus, substrate recognition and cleavage are genetically separable steps. Both the wild-type and mutant Ape1 proteins bound strongly to the enzyme incision product, an incised abasic site, which suggested that Ape1 might exhibit product inhibition. The use of human DNA polymerase β to increase Ape1 activity by eliminating the incision product supports this conclusion. Notably, the complexes of the D283A, D308A, and D283A/D308A double mutant proteins with both intact and incised abasic DNA were significantly more stable than complexes containing wild-type Ape1, which may contribute to the lower turnover numbers of the mutant enzymes. Wild-type Ape1 protein bound tightly to DNA containing a one-nucleotide gap but not to DNA with a nick, consistent with the proposal that substrate recognition by Ape1 involves a space bracketed by duplex DNA, rather than mere flexibility of the DNA.

Base excision repair (1, 2) is a multistep process that corrects a diverse array of spontaneous and mutagen-induced DNA lesions. The process can be initiated by DNA glycosylases, which excise damaged bases to create apurinic/apyrimidinic (AP) 1 sites. Modified abasic sites can arise directly from oxidative DNA-damaging agents, such as ionizing radiation (3). The AP and modified abasic sites are substrates for class II AP endonucleases, which cleave the 5Ј phosphodiester of both regular (hydrolytic) AP sites (4) and oxidized abasic residues (5). Excision of the lesion, DNA repair synthesis, and ligation complete the process (1,2).
Ape1 protein is quantitatively the main AP endonuclease of human cells (6,7). This DNA repair protein was independently named Hap1 (8) and Apex (9), and as an in vitro activator of DNA binding activity for some oxidized transcription factors, it was also named Ref1 (10 -12). Ape1 protein can be separated into two functionally distinct regions, with the N-terminal domain possessing the Ref1 redox activity (10,12) and the Cterminal domain contains the AP endonuclease activity within a 280-residue polypeptide sequence (10,12,13) homologous to exonuclease III of Escherichia coli (27% identity), the ExoA protein of Streptococcus pneumoniae (39% identity), the Rrp1 protein of Drosophila melanogaster (50% identity), and the Arp protein of Arabidopsis thaliana (57% identity) (4,14).
The proteins of the Ape1/exonuclease III family exhibit a range of enzymatic activities on duplex DNA. Along with the class II AP endonuclease activity, exonuclease III possesses duplex-specific 3Ј-5Ј exonuclease activity, 3Ј-repair phosphodiesterase activity, 3Ј-phosphatase activity, and RNase H activity (15)(16)(17)(18). Ape1 protein also has these multiple activities, but with 3Ј-repair diesterase, 3Ј-phosphatase, RNase H, and exonuclease activities 100 -10,000-fold lower than its AP endonuclease activity (4, 19 -21). The presence of multiple catalytic functions for exonuclease III, a relatively small (M r ϳ30,000), monomeric protein suggested that a single active site catalyzes all the enzymatic reactions (22). In that model, distinct recognition elements in the protein were proposed to interact with duplex DNA 5Ј to the target site, and a "space" in the DNA duplex caused by a missing base or frayed 3Ј-end.
The crystal structure of exonuclease III has been solved, and a catalytic mechanism was proposed involving a single active site (23). The recently proposed crystal structure of Ape1/Hap1 protein (24) prompted an essentially identical proposal that involves a single active site for this enzyme. Both models involve amino acid residues that are conserved in the Ape1/ exonuclease III family, and the properties of some site-specific mutant Ape1 proteins are consistent with the proposal (25,26). In a reverse approach, Gu et al. (27) used an in vivo complementation assay to isolate mutants of the Drosophila Rrp1 protein with altered repair capacity or specificity. Some of the alterations in these proteins were mapped to conserved residues that do not correspond to functional residues proposed for exonuclease III or Ape1. One of the Rrp1 mutant proteins had diminished 3Ј-repair diesterase activity but normal AP endonuclease activity. Thus, either these proteins have more than one active site, or substrate specificity can be modulated independently of the catalytic activity.
We have previously addressed the mechanism of substrate recognition by Ape1 protein using electrophoretic mobility shift assay (EMSA) and footprinting methods (28). Those studies showed that Ape1 binds specifically around the abasic site in DNA and causes a pronounced distortion at the abasic site in a preincision complex (19). However, the EMSA conditions and those of others (26,29) were not suitable for biochemical characterization, because the protein-DNA complexes were relatively unstable. Here we report improved conditions for EMSA and use the assay to monitor Ape1 binding and dissociation from an abasic substrate and the incised product. We have examined the interactions for several site-specific mutant forms of Ape1. We show here that Ape1 has a high affinity to the incised product DNA and to DNA with a one-nucleotide gap, and the several mutations that affect the catalytic activity to different degrees do not significantly affect abasic site recognition but do affect the dynamics of the protein-DNA interaction.

Site-directed Mutagenesis and Plasmid Construction
Construction of Mutant Ape1 Proteins-"DeepVent" DNA polymerase (New England Biolabs, Beverly, MA) was used in the polymerase chain reaction (PCR) to generate site-specific mutations in APE1 cDNA. In the final expression constructs, any DNA products that resulted from PCRs were sequenced to verify that only the desired mutation had been introduced.
Mutant D308A (Aspartate to Alanine Substitution at Residue 308)-Using plasmid pCW26 (7) as the template DNA, the 5Ј-portion of the APE1 gene was amplified with T3 (ATTAACCCTCACTAAAG) and D308A(-) primers (TAGGACAGTGAGCACTCGCGA). Likewise, the 3Јportion of the gene was amplified with T7 (AATACGACTCACTATAG) and D308A primer (TCGCGAGTGCTCACTGTCCTA) in a separate reaction. The reaction products were then isolated, mixed together, and PCR-amplified with the T7 and T3 primers to generate a complete APE1 cDNA with the desired site-specific mutation. The final PCR product was cut with EcoRI and subcloned into Bluescript KS (Stratagene, La Jolla, CA) at the EcoRI site. Note that the sequence of D308A primer had been designed to change two amino acid residues, although the final plasmid had in fact only one mutation, D308A.
Mutants H309N (Histidine to Asparagine Substitution at Residue 309) and D283A (Aspartate to Alanine Substitution at Residue 283)-Mutagenesis was exactly as for D308A, except that the H309N (CG-GCAGTGATAACTGTCCTAT) or D283A (GTTGGTTGGCGCCTTGCT-TAC) primer and the H309N(-) (ATAGGACAGTTATCACTGCCG) or D283A(-) (GTAAGCAAGGCGCCAACCAAC) primer were used to mutate the gene. The intermediate PCR products were amplified with hAPE15NcoI (CAGCTGCCATGGGGTTCG) and hAPE13HindIII (TCTCTGAAGCTTGTTTAAAG) primers to generate the entire gene with the desired site-specific mutation. The final PCR product was cut with NcoI and HindIII and subcloned into NcoI/HindIII-digested pSE380 (Invitrogen, Carlsbad, CA).
Mutant D283A/D308A (Aspartate to Alanine Substitutions at Residues 283 and 308)-The technique described for generating the D308A substitution was used to construct this double mutant, except that the template DNA for the PCR was the single mutant D283A in pSE380. The primers D308A and D308A(-) were used to create the second site-specific mutation.
To construct expression plasmids for the mutant portion, the 238base pair PstI-HindIII fragment encompassing the 3Ј-portion of APE1 cDNA from pXC53 was replaced with the corresponding PstI-HindIII fragment from the mutagenized plasmids. For in vivo complementation tests, a XbaI-HindIII fragment containing the entire APE1 DNA from pXC53 or its derivatives was inserted downstream of the arabinoseregulated promoter of pBAD22A (obtained from The Cloning Vector Collection of the National Institute of Genetics, Shizuoka, Japan).

In Vivo Complementation Test
Strain BW528 was transformed by pBAD22A or the derivatives carrying the wild-type or mutant APE1 gene. The resulting strains were tested for their ability to form single colonies by streak-dilution on LB agar plates (33) containing 0.025% methyl methanesulfonate (MMS) and 0.2% L-arabinose. The colony-forming ability was checked after incubation for 24 h at 37°C.

DNA Substrates
Poly[d(A-T)] containing [␣-32 P,uracil-3 H]dUMP at a frequency of one per 500 nucleotides was synthesized as described by Levin and Demple (34), except that the [␣-32 P]dUTP was prepared from labeled dCTP (Andotek, Irvine, CA) by deamination with dCTP deaminase (35) (kindly provided by Dr. Bernard Weiss, University of Michigan Medical School). After the reaction, more than 95% of dCTP had been converted to dUTP, as determined using thin layer chromatography on polyethyleneimine cellulose (36).
A 51-mer synthetic oligonucleotide (37) containing a tetrahydrofuran residue (51F), a uracil (51U), or a cytosine at position 22 (51C) and the related oligonucleotides (see Fig. 6A) were purchased from Operon Technologies Inc. (Alameda, CA). The concentration of high pressure liquid chromatography-purified oligonucleotides was determined by measuring absorbance at 260 nm (38). To generate a double-stranded substrate, 100 pmol of oligonucleotide was 5Ј-end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase (New England Biolabs, Beverly, MA) and then annealed to the complementary oligonucleotide (see Fig. 6A). After precipitation with ethanol, the amount of DNA recovered was determined by monitoring the recovered radioactivity.
To prepare incised 51F, the double-stranded oligonucleotide (100 pmol) was treated with 12.5 pmol of Ape1 in a 100-l reaction (50 mM Tris-HCl (pH 8.4), 10 mM MgCl 2 , and 0.2 mg/ml BSA) at 37°C for 10 min. The reaction was stopped by adding 5 l of 0.5 M EDTA. After the reaction, 98% of substrate had been converted to the incised form (data not shown). The DNA was sequentially extracted with phenol-chloroform and chloroform and then precipitated with ethanol.
A substrate with a single AP site was prepared according to Bennett et al. (39), except that the Ape1 reaction buffer was used (50 mM HEPES-KOH (pH 7.5), 50 mM KCl, 0.1 mg/ml BSA, 10 mM MgCl 2 , and 0.05% Triton X-100). After treatment with recombinant E. coli uracil-DNA glycosylase (a generous gift of Dr. Dale Mosbaugh, Oregon State University), the reaction mixture was used directly for incision assays.

Protein Purification
To purify the recombinant Ape1 proteins, expression was accomplished by addition of isopropyl ␤-D-thiogalactopyranoside when the bacterial cultures had reached an absorbance at 600 nm to 0.5, and the incubation was continued for 90 min according to Marians (40). Cell lysates (fraction I) were prepared as described by Hupp and Kaguni (41). The cell lysate was fractionated at a 55% saturation of ammonium sulfate, and the remaining soluble proteins were then precipitated at 80% saturation of ammonium sulfate. The proteins were resuspended in Buffer A (50 mM HEPES-KOH (pH 7.5), 1 mM EDTA, 0.1 mM dithiothreitol, 10% (v/v) glycerol) containing 100 mM KCl and dialyzed against the same buffer (fraction II). Fraction II was applied to a 20-ml phosphocellulose (Whatman P11) column and then eluted with a linear gradient of 100 -750 mM KCl in Buffer A. The peak fractions containing Ape1 were pooled and dialyzed against Buffer A containing 100 mM KCl, (fraction III). Fraction III was applied to a MonoS 5/5 column (Amersham Pharmacia Biotech) and then eluted with a linear gradient of 100 -250 mM KCl in Buffer A. The peak fractions containing Ape1 were pooled and concentrated by dialysis against Buffer A containing 200 mM KCl and 20% polyethylene glycol (average molecular weight, 8000) (fraction IV). Fraction IV was applied to a Superose-12 column (Amersham Pharmacia Biotech), which yielded a first, symmetrical peak corresponding to Ape1 protein and the AP endonuclease activity, and a second peak corresponding to residual lysozyme from the cell lysis procedure (41).
For Pol␤ purification, expression of the protein and preparation of lysate (fraction I) were performed basically the same as for Ape1 purification. After dialysis against Buffer B (50 mM Tris-HCl (pH 8.0), 1 mM EDTA, 0.1 mM dithiothreitol, and 10% (v/v) glycerol) containing 100 mM KCl, fraction I was applied to a 60-ml phosphocellulose (Whatman P11) column and then eluted with a linear gradient of 100 -800 mM KCl in Buffer B. The peak fractions containing active Pol␤ were pooled and dialyzed against Buffer B containing 100 mM KCl, (fraction II). Fraction II was applied to a 1-ml heparin column (Amersham Pharmacia Biotech) and eluted with a linear gradient of 50 -750 mM KCl in Buffer B.
Fractions containing active Pol␤ were pooled. Analysis of these fractions by SDS-PAGE indicated that they contained only a single protein detectable by silver staining (data not shown).
Protein concentrations were determined by the Bio-Rad protein assay (Bio-Rad), based on the method of Bradford (42), using BSA (Amersham Pharmacia Biotech) as the standard.

DNA Binding Assays by EMSA
The following binding methods were adapted from the protocols of Wilson et al. (28), Ng and Marians (43), and Hendrickson and Schleif (44). Modification of the binding conditions of Wilson et al. (28) involved exploring the pH range 7.5-8.8 with different buffers and a range of KCl concentrations up to 50 mM NaCl (data not shown); however, the most important factors in the improved electrophoretic resolution were the reduction of the concentration of EDTA and omission of glycerol from the buffer.
For binding measurements, reaction mixtures (10 l) in 50 mM Tris-HCl (pH 8.4), 1 mM EDTA (pH 8.0), and 0.2 mg/ml BSA containing 51F DNA (1 nM) were mixed with proteins as indicated in the figure legends. The mixtures were incubated on ice for 10 min and loaded on a prerunning 8% polyacrylamide gel (80:1 acrylamide:bisacrylamide). The electrophoresis buffer contained 6 mM Tris-HCl (pH 7.8), 5 mM sodium acetate, and 0.1 mM EDTA (43), and the gels were subjected to a constant voltage of 8 V/cm applied for 2 h at 4°C. Following gel electrophoresis, the gels were dried and autoradiographed at -80°C. The amount of DNA present in each band was quantified using a Molecular Imager system (Bio-Rad).
To follow dissociation, binding reaction mixtures (100 l) containing 1 nM 51F DNA and 7.8 nM Ape1 protein were incubated on ice for 10 min. The subsequent steps were carried out at 0 -4°C using equipment prechilled to 4°C. At time 0, 10-l samples were removed, mixed with 1 l of a 1-M unlabeled 51F stock, and incubated for various times on ice before being loaded on a prerunning gel to separate free DNA from the Ape1-DNA complexes.

Product Depletion Assay
To deplete the Ape1 reaction product from the reaction mixture, we used human Pol␤ and a 51-mer oligonucleotide (100 nM) containing a natural AP site as the substrate. The concentration of proteins used is indicated in the figure legends. After the incision reactions, the AP site was treated with NaBH 4 (140 mM) to reduce and stabilize it against spontaneous ␤-elimination (45). After inactivation of the enzymes by heating at 80°C for 1 min, the DNA was analyzed on a 20% polyacrylamide gel containing 8 M urea. The amount of DNA present in each band was quantified using a Molecular Imager system (Bio-Rad).

Other Enzyme Assays
To monitor AP endonuclease activity during purification, poly[d(A-T)] containing [␣-32 P,uracil-3 H]dUMP was used as a substrate. Just before use, the polymer was treated with uracil-DNA glycosylase (kindly provided by Dr. Dale Mosbaugh, Oregon State University) at 37°C for 1.5 h under standard assay conditions. The AP endonuclease assays contained DNA with 1.8 pmol of AP sites in 25 l, and the reactions were initiated by addition of an enzyme sample (Յ2 l), followed by incubation at 37°C for 5 min. The reactions were stopped by addition of 25 l of stop solution (0.45 N NaOH, 25 mM EDTA) and heating for 45 min at 65°C. The amount of incised AP sites was determined as the acid-soluble, Norit-nonadsorbed radioactivity following treatment at alkaline pH; ␤-elimination of nonincised AP sites does not liberate the 32 P label in Norit-nonadsorbable form (34). One unit of AP endonuclease was defined as the amount of activity that cleaves 1 pmol of AP sites per min.
To monitor the Pol␤ activity during purification, a hairpin-structured oligonucleotide DNA (46)   washed three times with 0.5 M Na 2 HPO 4 . The amount of incorporated dTMP was determined as the radioactivity retained on the paper (38).

Purification of Mutant Ape1
Proteins-The site-specifically mutated Ape1 proteins were purified following overexpression in E. coli strain BL21 (DE3). After induction of the D283A and D308A mutant proteins, the crude extracts (fraction I) containing these proteins displayed significantly higher AP endonuclease activity than extracts of the control strain (data not shown), which indicated that both the D283A and D308A proteins retain substantial enzymatic activity. This result seemed to contrast with the data of Barzilay et al. (25), who reported a D283A mutant protein that had ϳ2000-fold reduced activity. Therefore, we purified the mutant proteins very carefully, and both the AP endonuclease activity and the amount of the induced M r 37,000 protein were monitored during all purification steps.
In contrast to the above, for the D283A/D308A double-mutant and H309N proteins, no increase was detected in the total AP endonuclease activity in crude extracts of cells after protein induction, even though substantial amounts of the induced recombinant proteins were detected by SDS-PAGE (data not shown). Thus, only the induced M r 37,000 protein could be monitored during the purification of the D283A/D308A and H309N proteins. In the final step of the H309N purification, the leading edge of a bacterial AP endonuclease peak introduced Ͻ5% contaminating activity (data not shown).
Analysis of these preparations by SDS-PAGE and staining with Coomassie Brilliant Blue (Fig. 1) or silver (data not shown) indicated that they contained only a single predominant polypeptide (Ն95%) with M r ϳ37,000 as expected for this protein of 35,500 Da (7). Interestingly, the two proteins containing the D283A mutation migrated slightly faster in the gel. Table I summarizes the enzymatic activity of the purified proteins.  The enzymatic activity of the mutant proteins was also addressed by complementation of the nfoxth -E. coli strain BW528 (31). This strain has Ͻ5% of the normal AP endonuclease activity and is highly sensitive to MMS, but the wild-type APE1 gene can restore MMS resistance to it (7). Consistent with the activities measured in vitro, the D283A and D308A mutant APE1 genes under inducing conditions restored MMS resistance in BW528, whereas the genes containing the D283A/ D308A double mutation, or the H309N single mutation, did not. 2 Binding of Ape1 to 51F DNA-We previously (28) showed that an EMSA allows the identification of specific complexes between wild-type Ape1 and DNA containing an abasic site. However, these complexes are evidently unstable, as indicated by the "smear" of DNA migrating in gels between the protein-DNA complex and the free DNA (28). This material likely represents protein-DNA complexes that dissociated during electrophoresis. Thus, the assay probably underestimates the amount of DNA bound by protein. The binding and electrophoresis conditions were therefore modified (see under "Experimental Procedures") to allow the free DNA and the DNA-Ape1 complexes to be separated cleanly by electrophoresis ( Fig. 2A,  panel a). Nonspecific DNA binding by the Ape1 proteins was estimated by using 51C DNA as the substrate, which revealed only small amounts of protein-DNA complexes with any of the proteins (Fig. 2B). Small amounts of these nonspecific complexes were also observed for 51F DNA incubated with the highest levels of Ape1 protein (see, for example, Fig. 2A, panel  d). Thus, the modified EMSA detects predominantly damagespecific binding by Ape1.
The quantitative binding curves for 51F and the various Ape1 proteins are shown in Fig. 2C. For all five proteins, the amount of protein-DNA complex formed was proportional to the protein concentration up to 4 nM (Fig. 2C). This result allowed us to estimate the relative binding affinities of wildtype and mutant Ape1 proteins (Table II). Surprisingly, all the mutant proteins showed high affinity for 51F in this assay, with quantitative values very close to that for wild-type Ape1 protein ( Fig. 2C; Table II). The D308A and D283A/D308A proteins showed slightly higher affinity for the abasic DNA than did wild-type Ape1, and the H309N mutant had only ϳ2-fold reduced affinity (Table II), even though its AP endonuclease activity was markedly reduced (Table I). Thus, the defects in the D283A/D308A and H309N proteins do not seem to influence substrate recognition, in contrast to an inactive Ape1 mutant protein that was reported to lose both activity and damage-specific DNA binding (26). This result shows that APspecific binding and enzymatic activity can be separated by certain amino acid substitutions.
Dissociation of Ape1 from 51F DNA-Dissociation of an enzyme from its DNA substrate prior to cleavage can be an important factor in limiting its catalytic activity. We studied the dissociation of preformed complexes of the Ape1 proteins and labeled 51F DNA by adding a large excess of unlabeled 51F DNA (see under "Experimental Procedures"). The vast majority of any dissociated Ape1 protein will bind to this unlabeled DNA rather than rebind to the small portion of labeled DNA. Samples were taken at various times after adding the competitor DNA and analyzed by EMSA. Fig. 3A shows that at 0°C, Ape1-DNA complexes were converted to free DNA during incubation with the unlabeled 51F substrate in a time-dependent manner. The quantitative dissociation curves allowed us to determine the half-lives of the protein-DNA complexes in the EMSA procedure (Fig. 3B). For wild-type Ape1 protein, this value was 50 s. Previously, we developed a filter binding assay to measure the half-life of Ape1 complexes with DNA containing an abasic site, and the new value agrees quite well with the previous result (28). Surprisingly, the half-lives of the F-DNA complexes with the D283A, D308A, and D283A/D308A proteins were considerably longer than for wild-type Ape1 (5, 9, and 7 min, respectively). In contrast, the half-life of the complex with H309N protein was too short to measure in this assay, because the majority of the protein had dissociated within 10 s (Fig. 3B; Table II).
Binding of Ape1 Proteins to Incised 51F DNA-The catalytic activity of Ape1 may be limited by interaction with its reaction product, DNA containing an incised abasic site, but this question had not been explicitly studied. We approached this issue by using incised 51F DNA as a binding substrate. EMSA using the incised 51F DNA showed that both the wild-type and mutant Ape1 proteins had high affinity for incised 51F DNA (Fig.  4A, panels a-d). For the wild-type, D283A, D308A, and D283A/ D308A proteins, we estimate 50% binding of incised 51F DNA at 4 nM Ape1 protein (Table II). The dissociation of Ape1 proteins from complexes with 51F DNA was also measured by EMSA, again using unlabeled, nonincised 51F DNA as a competitor (Fig. 5) For the wild-type, D283A, and D283A/D308A proteins, the half-lives were 35, 50, and 65 s, respectively (Table II); for the D308A protein, the half-life was 130 s (Table  II). Thus, under our conditions, wild-type Ape1 and all but one of the mutant proteins bound the endonuclease cleavage product with affinities very close to those for the nonincised abasic DNA substrate.
Incised 51F DNA also bound Pol␤ tightly (Fig. 4, panel f), even though this material is not a substrate for deoxyribosephosphate excision by the ␤-elimination activity of the polymerase (45). The affinity of Pol␤ for incised 51F DNA was such that a higher protein concentration (estimated ϳ12 nM; see Fig.  4B) was required for 50% binding by Pol␤ than by Ape1 (compare with Table II).
High-affinity Binding of Ape1 Protein to DNA Containing a One-nucleotide Gap-The high affinity of Ape1 for the incised F site suggested that the enzyme might bind generally to nicked or gapped structures in DNA. To test this idea, we generated duplex oligonucleotides with a one-nucleotide gap at the same position as the F residue in 51F, a nick at this position, or protruding 3Ј-or 5Ј-single-stranded regions; the structures are shown in Fig. 6A. In EMSA experiments, Ape1 bound the structure of the duplex oligonucleotides with a one-nucleotide gap with affinity similar to that for incised or intact F sites (Fig. 6B). In contrast, Ape1 did not have high affinity for the structure with a nick or the 5Ј-single-stranded or 3Ј-singlestranded structures (Fig. 6B). We have not explored the gap size-dependence of Ape1 binding affinity.
Effect of the Product on the Nicking Reaction in Solution-The high affinity of Ape1 protein for the incised product and for one-nucleotide gapped DNA suggested that the enzyme might exhibit product inhibition, although such an effect has not been reported. One possibility is that Ape1 has high affinity for the incised tetrahydrofuran (F) abasic site, but not for an incised hydrolytic AP site. Because the latter moieties can undergo degradation during electrophoresis (45), we performed steadystate kinetic studies using a substrate with a glycosylase-generated AP site.
In Ape1 incision reactions, the kinetics exhibited a short linear range restricted to Ͻ30% incision (Fig. 7A, open circles). We obtained essentially the identical result when we used the 51F substrate (data not shown). The nonlinearity observed as the incision reactions proceeded could result from competitive inhibition by the incised product. If the incision product is a competitive inhibitor, then the depletion of the product by the deoxyribose-phosphatase and DNA polymerase activities of Pol␤ (45) might enhance the reaction by Ape1. Indeed, in the presence of Pol␤ and dCTP, Ape1 incision kinetics showed an expanded linear range (Fig. 7A, closed circles), although Pol␤ cannot incise abasic sites (Fig. 7A, closed triangle) (45). The same effect of Pol␤ was seen for the D283A and D308A proteins (Fig. 7, B and C). This result is consistent with the results from EMSA using 51F, which show that the Ape1 proteins have affinity for incised 51F DNA.
To confirm that Pol␤ actually depletes the Ape1 reaction products, the amount of the substrate, the product, and the DNA fragment extended one nucleotide by Pol␤ were measured formamide and 20 mM EDTA. The products were analyzed on a 20% polyacrylamide gel containing 8 M urea. The intensities of intact and incised DNA bands were measured by a Molecular Imager system, and the amount of incision was quantitated. The first time point was at 15 s after addition of the Ape1 proteins. Essentially the same result was obtained in two independent experiments, one of which is shown; we estimate the measurement errors in these determinations to be Յ5%. wt, wild-type. at various concentrations of Pol␤ (Fig. 8A). In the presence of dCTP, the incised product (21-mer) was converted to a 22-mer by the gap-filling activity of Pol␤ (Fig. 8, A and B). The total amount of incised DNA (21-mer ϩ 22-mer) increased in a manner dependent on the Pol␤ concentration up to ϳ15 nM and then continued to increase with a lower Pol␤ dependence up to a plateau. The beginning of this plateau (at 50 -100 nM Pol␤) was identical to concentration at which almost all the incised product (21-mer) had been converted to the 22-mer form (Fig.  8B). The total amount of incision (21-mer ϩ 22-mer) in this single time point experiment was inversely correlated with the amount of the Ape1-incised compound (21-mer) present (Fig.  8B). On the other hand, in the absence of dCTP, the smaller enhancing effect of Pol␤ on the Ape1 reaction might not be enzymatic (Fig. 8C), but rather the Ape1-incised DNA molecules may be sequestered through binding by high concentrations of Pol␤. These results are consistent with the result from EMSA that Ape1 has high affinity for its incised product. DISCUSSION Our studies have addressed the substrate and product binding dynamics of Ape1, the predominant AP endonuclease of human cells. The results show that the incised product of the AP endonuclease reaction is a competitive inhibitor for Ape1 and that the next enzyme of the base excision repair pathway, DNA polymerase ␤, can capture the Ape1 product to overcome this limitation. Furthermore, analysis of catalytically impaired mutant Ape1 proteins demonstrated that substrate binding and the chemical steps of phosphodiester cleavage are separable by certain mutations.
We created four different mutant proteins by site-directed mutagenesis. These altered enzymes had reduced AP endonuclease activity between 22 and ϳ0.02% that of wild-type Ape1. Unexpectedly, the D283A protein was quite active (10% of wild-type), although Barzilay et al. (25) reported that this mu-tation markedly reduced the activity (ϳ2000-fold) and concluded that aspartate 283 was critical for catalysis. Repeated DNA sequencing and independent subcloning efforts showed that only the D283A change was introduced by our procedure. The activity of the D283A protein was not derived from contaminating bacterial AP endonuclease: first, the AP endonuclease activity was dependent on expression of the human protein; second, during purification, the activity copurified with the induced M r 37,000 protein; and third, the D283A mutant cDNA restored resistance to MMS in an xthnfostrain of E. coli. We found that the D283A protein could aggregate under low-salt conditions (Buffer A containing 25 mM KCl), which is one possible explanation for the dramatically divergent results. Alternatively, the D283A derivative generated by Barzilay et al. (25) may contain an unintended second mutation. Regardless, aspartate 283 contributes to activity in Ape1 but is not essential.
We found that Ape1 protein has high affinity for incised abasic sites, which was not detected in previous studies (28,39). The previous EMSA conditions may have been too stringent for this detection, because the complex of Ape1 and the incised abasic site is less stable than the complex of Ape1 and an intact abasic site (compare Fig. 2 with Fig. 4). However, the binding determinants for incised abasic sites might overlap with those for uncleaved abasic residues, because the wildtype, D283A, D308A, and D283A/D308A proteins had identical affinity. The much lower affinity of the H309N protein for incised abasic sites compared with the other Ape1 derivatives could reflect an important role of histidine-309 in product binding specifically, as this protein has near-wild-type affinity for the substrate (an intact abasic site).
Interestingly, the stability of the Ape1 DNA complexes varied among the wild-type and the various mutant proteins. The slower dissociation of the D283A, D308A, and D283A/D308A proteins from the intact abasic site than from the incised lesion indicates that the preincision complex is stabilized by the continuous DNA structure and perhaps by substrate contacts within the active site. In a current model (23,24), the aspartate residue that was replaced in the D283A and D283A/D308A proteins is postulated to act as general base. Without this aspartate residue, the histidine residue might stabilize the preincision complex. In fact, the unincised 51F DNA complex with the H309N protein had a much lower stability, consistent with loss of an important contact. The model of Gorman et al. (24) suggests that histidine 309 interacts with the 5Ј-phosphate bond of the abasic site through a water molecule. Structural studies of these mutant proteins, particularly in complexes with DNA, would be useful.
What does Ape1 protein recognize specifically in DNA? We showed previously that neither the deoxyribose of the AP site itself nor the base opposite the AP site is necessary for incision (19). Indeed, strong binding by Ape1 can occur when no residue is present at all, as shown with the 51-Gap substrate (Fig. 6). A mere nick was insufficient to allow Ape1 to bind stably, but Strauss et al. (29) showed that Ape1 protein has high affinity for a 3Ј-incised abasic site resulting from ␤-elimination. We found that Ape1 binds tightly to a 5Ј-incised AP site, which together with the foregoing observations indicates a strong binding determinant for the gap or "space" itself, as originally suggested by Weiss (22) for exonuclease III. Beyond this conclusion, the 3Ј-OH and 5Ј-phosphate moieties of the gapped substrate are evidently not important binding determinants. One observation that seems at first inconsistent with this model is the report that Ape1 can incise a duplex oligonucleotide 5Ј to a 3,N 4 -benzetheno-2Ј-deoxycytidine residue (47). However, molecular modeling suggests that the local structure at the adduct site is similar to that of an AP site (48), with the adducted base shifted away from the base in the opposite strand.
Although the EMSA approach is not suitable for determination of rate constants, we were able to show in steady-state reactions that Pol␤ enhances Ape1 activity. This observation supports the suggestion from EMSA experiments that the incised DNA product would be a competitive inhibitor by strongly binding to Ape1 protein. The activating effect of Pol␤ on Ape1 is most likely due to its gap-filling activity, as the activation was most effective when a nucleoside triphosphate was present to allow DNA synthesis. These experiments also extended the range of experiments in which the F analog and a hydrolytic AP site are handled identically by Ape1 (19,28,39).
We previously proposed a model that Ape1 protein loads Pol␤ onto the incised abasic site by specific protein-protein interactions and in the process enhances the deoxyribose-phosphate excision reaction of Pol␤ (5,39). We demonstrated here that Ape1 protein has high affinity for the product of the DNA Pol␤ excision reaction, a one-nucleotide-gapped DNA. Although it is unknown how often one-nucleotide gaps arise in DNA in vivo, this binding activity of Ape1 protein for one-nucleotide gaps might promote repair of those gaps by allowing Ape1 to act as a molecular matchmaker (49) for Pol␤ and gapped DNA.