The action of DNA ligase at abasic sites in DNA.

Apurinic/apyrimidinic (AP) sites occur frequently in DNA as a result of spontaneous base loss or following removal of a damaged base by a DNA glycosylase. The action of many AP endonuclease enzymes at abasic sites in DNA leaves a 5'-deoxyribose phosphate (dRP) residue that must be removed during the base excision repair process. This 5'-dRP group may be removed by AP lyase enzymes that employ a beta-elimination mechanism. This beta-elimination reaction typically involves a transient Schiff base intermediate that can react with sodium borohydride to trap the DNA-enzyme complex. With the use of this assay as well as direct 5'-dRP group release assays, we show that T4 DNA ligase, a representative ATP-dependent DNA ligase, contains AP lyase activity. The AP lyase activity of T4 DNA ligase is inhibited in the presence of ATP, suggesting that the adenylated lysine residue is part of the active site for both the ligase and lyase activities. A model is proposed whereby the AP lyase activity of DNA ligase may contribute to the repair of abasic sites in DNA.

DNA repair pathways have evolved to process a wide range of chemically distinct lesions in DNA (1). One of the most common types of damage is spontaneous or enzymatic hydrolysis of the N-glycosidic bond between a DNA base and the sugar phosphate backbone generating an abasic site (AP site). 1 AP sites are highly mutagenic and require rapid and efficient repair. AP sites are processed by a base excision repair pathway that is frequently initiated by the action of a class II AP endonuclease that cleaves the DNA backbone adjacent to the lesion to produce 3Ј-OH and 2Ј-deoxyribose 5Ј-phosphate termini (2). The latter residue, referred to as a 5Ј-dRP moiety, is relatively alkali labile and can be removed by AP lyases that facilitate ␤-elimination. Most enzymes demonstrated to have dRPase activity operate through this lyase mechanism, although some, such as the Escherichia coli recJ protein, catalyze hydrolysis (3). The two classes of dRPase enzymes can be distinguished by the fact that the lyase mechanism frequently involves formation of a transient Schiff base intermediate in which an amino group on the enzyme is covalently bound to the DNA. This lyase mechanism, first proposed for E. coli endonuclease III (4), provides a simple method to identify a polypeptide with AP lyase activity because the Schiff base intermediate can be trapped in a stable form by reaction with a strong reducing agent, such as NaBH 4 or NaBH 3 CN. Thus, with an appropriate radioactively labeled substrate, the label can be transferred to the AP lyase enzyme. This borohydride trapping has been documented for several repair enzymes (5)(6)(7)(8)(9) and, more recently, for DNA pol ␤ (10,11).
The combined action of AP endonuclease and AP lyase leaves a one-nucleotide gap that is filled by a DNA polymerase. The final step in repair involves DNA strand sealing by DNA ligase. The mechanism of DNA ligase involves covalent modification of the enzyme by adenylation, transfer of the AMP residue in a phosphoanhydride linkage to the 5Ј-phosphate of nicked DNA, followed by resealing of the DNA strand driven by the energy of AMP hydrolysis.
We recently characterized a mtDNA ligase as part of an effort to reconstitute repair of AP sites using mitochondrial enzymes (12). The size, template specificity, and immunological properties of the mtDNA ligase suggested that this was a form of DNA ligase III. In the course of this work we found that mtDNA ligase is active on a DNA substrate containing an AP site incised on the 5Ј side by a class II AP endonuclease. Our observations prompted a detailed investigation of the action of T4 DNA ligase as a prototype for ATP-dependent DNA ligases at AP sites in DNA. In this paper we show that in the presence of ATP, T4 DNA ligase is able to reseal an incised AP site. In the absence of ATP, T4 DNA ligase acts as an AP lyase to facilitate a ␤-elimination reaction that leads to removal of the 5Ј-dRP residue. A model is presented whereby an intrinsic AP lyase activity in DNA ligase may facilitate repair of AP sites.

EXPERIMENTAL PROCEDURES
Materials-mtDNA ligase, mitochondrial AP endonuclease, and DNA ligase I were purified from Xenopus ovary tissue as described (12). T4 DNA ligase was obtained from Boehringer Mannheim. T7 DNA ligase was a gift from Dr. J. Dunn (Brookhaven National Laboratory). Variable quantities of both preparations of bacteriophage DNA ligase were subjected to SDS-PAGE analysis (13) in parallel with standard proteins of known concentration. The gels were stained with Coomassie Blue to confirm that both preparations were essentially homogeneous and to permit estimation of protein concentrations using densitometry of the stained gels. Radiochemicals were purchased from ICN Radiochemicals. Uracil DNA glycosylase (UDG) was obtained from Epicentre Technologies (HK-UNG). FPG protein was a gift from J. Tchou and A. P. Grollman (SUNY-Stony Brook). Sodium borohydride and sodium thioglycolate were obtained from Sigma-Aldrich. Other reagent grade chemicals were obtained from Sigma-Aldrich or Fisher. The Poros Q 4.6 ϫ 50-mm column used for anion exchange HPLC was obtained from Perceptive Biosystems. Oligonucleotides were either synthesized by the phosphoramidite method at the SUNY-Stony Brook Oligonucleotide Synthesis Facility or were obtained from Operon. A continuous duplex oligonucleotide was prepared by annealing a 5Ј-32 P kinase-labeled 32mer (5Ј-CATGGGCCGACATGAUCAAGCTTGAGGCCAAG) to a complementary oligonucleotide (5Ј-TCTTGGCCTCAAGCTTGATCAT-GTCGGCCCATG). Two nicked duplex oligonucleotides were prepared by annealing a 5Ј-32 P kinase-labeled 17-mer (5Ј-UCAAGCTTGAGGCC-AAG; referred to as U17) and either a nonradioactive 15-mer (5Ј-CAT-GGGCCGACATGA) or 12-mer (5Ј-GGGCCGACATGA) to the same complementary strand described above. The 12-mer was used in the experiment in Fig. 7C to permit better resolution of the reaction product from the initial labeled substrate.
Methods-Oligonucleotides were phosphorylated using standard procedures (14) and annealed by heating to 80°C in 0.1 M NaCl, 10 mM Tris, pH 8, 1 mM EDTA and slow cooling to 4°C. Oligonucleotides were treated immediately before use with UDG in 20 mM Hepes, pH 7.5, and diluted into an assay mixture with DNA ligase in 10 mM Hepes, pH 7.5, 1 mM MgCl 2 , unless otherwise indicated. Borohydride trapping was performed by a modification of published procedures (9, 11) by including 20 or 50 mM NaBH 4 in the binding reaction. After 30 min at 25°C, the solution was adjusted to contain 6 mM CaCl 2 and 10 g/ml micrococcal nuclease. After 20 min of incubation at 37°C, proteins were precipitated with trichloroacetic acid, analyzed by SDS-polyacrylamide gel electrophoresis (13), and detected by autoradiography or PhosphorImager analysis (Molecular Dynamics). dRPase activity was measured as described (3), and HPLC analysis of products generated in the presence of sodium thioglycolate was performed as described (5,15), except that a Poros Q anion exchange column was used.

RESULTS AND DISCUSSION
DNA Ligases Are Able to Reseal DNA Strands Nicked on the 5Ј Side of an AP Site-Our laboratory has studied the base excision repair pathway with nuclear and mitochondrial protein fractions using templates containing precisely positioned single AP sites embedded in covalently closed circular DNA (12,15). These sites are readily cleaved on the 5Ј side by class II AP endonuclease to yield a 3Ј-OH terminus and a 5Ј-deoxyribose phosphate (dRP) residue. When templates bearing incised AP sites are incubated with DNA ligase in the presence of ATP, the DNA ligases are able to reseal the nicked strand ( Fig. 1). This reaction has been reported for T4 DNA ligase (16) but not for eukaryotic DNA ligases. Because ligation of a 5Ј-dRP moiety directly reverses the action of AP endonuclease, it is counterproductive for repair. It is also a potential confounding factor in efforts to reconstitute repair reactions in vitro, because religation of an abasic site might not be differentiated from actual repair. In the experiment in Fig. 1, we used a substrate with a synthetic analogue of an abasic site, a tetrahydrofuran analogue that has been used extensively in repair studies (15,17). This is an analogue of a reduced deoxyribose moiety and is not subject to ␤-elimination. Experiments presented below show that T4 DNA ligase can also reseal an authentic (nonreduced) AP site.
T4 DNA Ligase Has AP Lyase Activity-We previously showed Xenopus laevis mtDNA ligase can be labeled using a borohydride trapping procedure that is specific for AP lyase activities (12). To determine whether this is a general property of ATP-dependent DNA ligases, we tested T4 and T7 DNA ligases for the ability to react with AP sites in a borohydride trapping assay. This assay employed an oligonucleotide substrate designed to contain a specific U residue adjacent to a nick, referred to as 15-*U17:33 to denote a 15-mer annealed adjacent to a kinase-labeled (*) 17-mer to a 33-mer complementary strand. Treatment of the duplex oligonucleotide with UDG results in a 5Ј-phosphoryl abasic site. This is an exact model of the substrate that would be generated by action of a class II AP endonuclease. Fig. 2 shows that both T4 DNA ligase and T7 DNA ligase react in this assay. Controls in lanes 3 and 4 of Fig.  2 demonstrate that cross-linking was dependent on the presence of NaBH 4 and on the AP site, because the reaction was not observed with a control oligonucleotide containing thymine in place of uracil.
We performed a variety of control experiments to characterize the putative AP lyase activity in T4 DNA ligase. The extent of cross-linking with a nicked substrate varies with solution conditions. Cross-linking is reduced at 5 mM MgCl 2 in comparison with the standard reaction conducted at 1 mM MgCl 2 and is inhibited more than 90% by 1 mM ATP (Fig. 3). The standard reaction includes post-treatment with micrococcal nuclease to degrade the oligonucleotide cross-linked to protein by NaBH 4 . When this nuclease treatment was omitted, the electrophoretic mobility of the major cross-linked protein species was reduced, as expected for a DNA-protein complex (Fig. 3B). In the absence of micrococcal nuclease treatment a minor cross-linked labeled species with essentially the same gel mobility as unmodified T4 DNA ligase was also observed. This faster migrating species is expected to be formed as an intermediate in the action of an AP lyase, as shown in Fig. 4. The initial product of an attack by AP lyase on an AP site is a Schiff base intermediate in which the enzyme is joined to DNA. The AP lyase reaction proceeds with elimination of the DNA from the C3Ј position of deoxyribose, producing an enzyme-dRP intermediate with a Schiff base linkage. Both the enzyme-DNA and enzyme-dRP species can be reduced by NaBH 4 to generate the doublet of cross-linked species in Fig. 3B. These borohydride FIG. 1. DNA ligases can reseal a nick adjacent to a 5-phosphorylated abasic site. A 5Ј-end labeled oligonucleotide containing a single synthetic AP site (3-hydroxy-2-hydroxymethyltetrahydrofuran, designated F for furan) was ligated into a gapped heteroduplex, and the covalently closed circular DNA product was purified as described (15). PAGE-urea gel analysis of a HinfI digest of this substrate confirmed that all substrate molecules were ligated with the tetrahydrofuran residue embedded in a 46-mer fragment, denoted as 46(F) (lane 1). Treatment with mitochondrial AP endonuclease led to cleavage on the 5Ј side of the tetrahydrofuran residue, providing a 26-mer fragment following diagnostic HinfI cleavage (lane 2). This fragment with a 5Ј-tetrahydrofuran residue is identified as F26. Samples of this mitochondrial AP endonuclease-incised substrate were incubated with X. laevis DNA ligase I, mtDNA ligase, T4 DNA ligase, or T7 DNA ligase (lanes 3-6). The products were deproteinized by organic extraction and cleaved with HinfI endonuclease prior to electrophoresis. The radioactive species moving slightly slower than the F26 fragment, which is most apparent in lane 6, is an intermediate in the ligase reaction produced by transfer of AMP to the 5Ј terminus at the nick.
cross-linking results are consistent with the hypothesis that T4 DNA ligase contains AP lyase activity.
In other experiments, we found that it was not necessary to present the AP site in the context of a nick in DNA, although this is the preferred substrate. Borohydride trapping was observed when the 15-mer oligonucleotide was omitted from the standard nicked substrate, leaving a 5Ј-AP site adjacent to single-stranded DNA. We also observed cross-linking to a free oligonucleotide with a 5Ј-dRP site generated by the action of UDG on the 5ЈU-17-mer oligonucleotide (5Ј-32 P-UCAAGCTT-GAGGCCAAG). The efficiency of borohydride trapping with a single-stranded 5Ј-AP oligonucleotide was about 50% of that observed with the nicked oligonucleotide substrate. The singlestranded oligonucleotide substrate was used in some experi-ments described below to simplify preparation of large quantities of substrate for AP lyase assays.
The borohydride trapping reaction is a very sensitive probe of AP lyase activity but is not always efficient because it acts on a transient intermediate in the overall AP lyase reaction. For DNA glycosylases that remove a base and attack the AP site in a sequential dual mechanism, such as FPG protein, this borohydride trapping procedure is relatively easy to control. It is more difficult to trap a large fraction of a protein that does not contain an intrinsic glycosylase activity because the NaBH 4 required to cross-link the enzyme-DNA Schiff base intermediate can also react with the ring open form of the sugar residue to inactivate the substrate. As an independent assay for AP lyase activity, we monitored the release of free 5Ј-dRP from DNA as an acid or ethanol-soluble species, as shown in Fig. 5. When reactions were performed with a high concentration of the single stranded 5Ј-dRP-17-mer oligonucleotide, we found that dRP release was linear with time, showed a clear temperature optimum, and was inhibited by ATP as observed for the borohydride trapping assay. The ethanol-soluble species released by T4 DNA ligase in the presence of thioglycolic acid was observed to have the same chromatographic properties as the product released by FPG protein, which has a well characterized AP lyase activity ( Fig. 6 and Refs. 5, 7, and 18).
The borohydride trapping and dRP release assays show that T4 DNA ligase is an authentic AP lyase by the same criteria used to show that DNA pol ␤ possesses AP lyase activity (10,11). Although T4 DNA ligase is capable of processing multiple AP site substrates (Fig. 5A), the turnover number is less than 10% of the vigorous rate observed for FPG protein (5). It should be noted that the glycosylase activity of FPG protein is reported to be significantly slower than the AP lyase activity (5). A lower turnover number for the AP lyase activity in T4 DNA ligase may be expected because the enzyme is likely to bind persistently to the gapped substrate generated by AP lyase action on a site previously cleaved by AP endonuclease.
We have not yet identified the active site for the DNA ligaseassociated AP lyase. The borohydride trapping reaction requires attack on the C1Ј residue of deoxyribose by an N-terminal amino group or by an internal lysine (19). The fact that the AP lyase activity is suppressed by ATP suggests that the active site lysine residue that is adenylated in DNA ligase (20) may be involved directly in the nucleophilic attack that promotes ␤-elimination. This residue is normally in close proximity to the nick in a DNA substrate in the course of a DNA ligation reaction. However, we have not ruled out the possibility that another lysine residue may be involved in this attack, because the deadenylated enzyme may have an altered conformation that interacts differently with DNA substrates containing 5Ј-dRP residues. It will be particularly interesting to map the active site residue in T7 DNA ligase that reacts in the borohydride trapping reaction because the structure of this enzyme has been determined (21). This enzyme has the added advantage that it is relatively small, with only 359 amino acid residues.
A Model for the Role of AP Lyase Associated with DNA Ligase-We have observed AP lyase activity using the borohydride trapping assay for the following four different ATP-dependent DNA ligases: T4 and T7 DNA ligase (Fig. 2), mtDNA ligase (12), and DNA ligase I (data not shown). Because ATPdependent DNA ligases as a class share structural and functional features (22), it is likely that the presence of AP lyase activity will be conserved in this family. To date we have not been able to document AP lyase activity in bacterial DNA ligases from either E. coli or T. aquaticus either in the presence or the absence of their cofactor, NAD (data not shown). The critical question raised by our observations is whether the AP lyase activity associated with T4 DNA ligase plays a significant physiological role. A model for the action of T4 DNA ligase at AP sites is shown in Fig. 7. In living cells, AP sites are very rapidly incised by AP endonuclease to generate the sort of nicked AP substrate we have used in our reactions. The experiments reported here show that T4 DNA ligase can act as an AP lyase at these sites in the absence of ATP. Under these conditions, the deadenylated enzyme cannot seal the nick and instead facilitates ␤-elimination, leading to loss of the 5Ј-dRP residue. This produces the single nucleotide gap structure diagramed as species 5 in Fig. 7. This single base gap may be repaired by the action of DNA polymerase and the conventional strand sealing action of DNA ligase.
It is also important to consider the action of T4 DNA ligase at incised AP sites in the presence of ATP, because a large fraction of DNA ligase may exist in the adenylated state in vivo. Our results suggest that the adenylated T4 DNA ligase FIG. 4. The chemistry of AP lyase action accounts for DNA-enzyme and dRP-enzyme complexes following NaBH 4 treatment. This scheme is based on those presented for other AP lyase enzymes (7,19). The Shiff base reaction scheme requires attack by a free amino group of the AP lyase on the C1Ј residue of the deoxyribose. This reactive nitrogen is referred to as N-enz. Other functional groups within the enzyme may assist in the ␤-elimination reaction as indicated. Covalent intermediates in the Schiff base reaction scheme labeled 1 and 2 may be reduced by borohydride to yield stable species with the protein cross-linked to an oligonucleotide or to a dRP moiety, respectively.
FIG. 5. Release of an acid soluble product from a 5-32 P-labeled abasic site by T4 DNA ligase. 15 pmol of 5Ј-32 P U17 oligonucleotide pretreated with UDG was incubated with 100 ng of T4 DNA ligase in 40 mM Hepes buffer, pH 7.5, for varied periods of time at 30°C (A), for 30 min at varied temperature (B), or in the presence of increasing concentrations of ATP (C). dRP release was measured as the generation of a radioactive product soluble in the presence of cold TCA and activated charcoal (3). The percentage of dRP removed was determined relative to the total alkali-labile cpm. The maximal amount of label solubilized in parallel reactions without enzyme represented 4% of the total available substrate.
FIG. 6. The putative 5-dRP product released by DNA ligase has the same chromatographic properties as that released by FPG protein, a well characterized AP lyase. The 5Ј-32 P U17 oligo pretreated with UDG was incubated with either FPG protein as a positive control (A) or with T4 DNA ligase (B) in the presence of 50 mM sodium thioglycolate. The product released by the lyase activity of FPG protein has been shown to react with thioglycolate to generate an anionic species (5,18). Reaction of the ␤-elimination product with thioglycolate leads to reduction of the aldehyde, blocking the subsequent ␦-elimination reaction for FPG protein. The ethanol-soluble reaction products were analyzed by chromatography on a Poros Q HPLC column using the indicated NaCl gradient. The intact oligonucleotide elutes from this column with the 1 M NaCl step. has a reduced ability to promote ␤-elimination. Instead, when a T4 DNA ligase molecule that is activated by adenylation binds this nicked substrate, it seals the nick to regenerate an internal AP site, as shown in Fig. 1 and diagramed as species 3 in Fig. 7A. The second product of the ligation reaction is a "disarmed" DNA ligase molecule that is no longer adenylated but is still in contact with the AP site. To test whether T4 DNA ligase is able to incise DNA on the 3Ј side of an internal AP site (i.e., without prior action of an AP endonuclease), we performed the experiment in Fig. 7B. This experiment shows that T4 DNA ligase is clearly capable of strand incision to yield a product with a slightly slower gel mobility than that produced by the well characterized AP lyase of FPG protein. This suggests that T4 DNA ligase is able to promote ␤-elimination but unlike FPG protein does not efficiently promote ␦-elimination. This sort of incision reaction was not observed in Fig. 1 because that experiment employed a reduced AP site analogue. These results suggest that when T4 DNA ligase seals a nick generated by class II AP endonuclease, it may recognize the product as a mistake and employ its lyase activity to reopen the DNA. To test this prediction, we performed the experiment shown in Fig. 7C. In this experiment, T4 DNA ligase was incubated with a 17-mer oligonucleotide containing a 5Ј-32 P-dRP residue adjacent to a nonradioactive 12-mer. DNA ligase was able to ligate the 12-mer to the 5Ј-dRP-17 mer to generate a 29-mer with an internal 32 P-dRP residue (lane 3 of Fig. 7C). A limited extent of ligation was observed without the addition of exogenous ATP (lane 2), presumably because a fraction of the T4 DNA ligase is puri-fied in an adenylated form. The more efficient ligation in the presence of ATP was followed by incision on the 3Ј side of the AP site to produce a labeled 12-mer with a 3Ј-dRP residue. Thus, the label transfer experiment in Fig. 7C confirms the model for the action of T4 DNA ligase at an AP site in the presence of ATP. The ring open 3Ј-dRP residue produced by AP lyase cannot be rejoined by DNA ligase due to the 2Ј-3Јdouble bond, but the 3Ј-dRP group would be susceptible to release by class II AP endonuclease. Taken together, these experiments suggest that the role of T4 DNA ligase in base excision repair may not be limited to the final step of strand closure.