Early steps of excision repair of cyclobutane pyrimidine dimers by the Micrococcus luteus endonuclease. A three-step incision model.

The early steps of excision repair of cyclobutane pyrimidine dimers are investigated. It is demonstrated that the apurinic/apyrimidinic endonuclease associated with the Micrococcus luteus uv-specific endonuclease cleaves the phosphodiester bond on the 3' side of the deoxyribose leaving a 3' hydroxy terminus and a 5' phosphoryl terminus. This nick is not a substrate for T4 polynucleotide ligase. The 3' base-free deoxyribose terminus is not a substrate for either the polymerase or the 3' to 5' exonuclease activities of Escherichia coli DNA polymerase I. However, the 3' terminus of the nick is converted to a substrate for DNA polymerization by the action of a 5' apurinic/apyrimidinic endonuclease. A three-step model for the incision step of excision repair of cyclobutane pyrimidine dimers is presented.

T4 polynucleotide ligase. The 3' base-free deoxyribose terminus is not a substrate for either the polymerase o r the 3' t o 5' exonuclease activities of Escherichia coli DNA polymerase I.
However, the 3' terminus of the nick is converted to a substrate for DNA polymerization by the action of a 5' apurinic/apyrimidinic endonuclease. A three-step model for the incision step of excision repair of cyclobutane pyrimidine dimers is presented.
Ultraviolet light induces the formation of stable cyclobutane dimers between adjacent pyrimidines in DNA (1). In repairdeficient strains of bacteria, the existence of one genomic cyclobutane dimer may constitute a lethal event; consequently, pyrimidine dimers must be repaired. Two general repair pathways exist for the repair of pyrimidine dimers: direct photoreversal of the cyclobutane dimer by a photoreactivating enzyme and long wavelength ultraviolet light (2) and a dark repair pathway that involves enzymatic excision of the dimer from the DNA followed by resynthesis of the excised region (3,4). The initial step in the dark repair pathway occurs by the action of endonucleases that recognize cyclobutane dimers. Incision at the site of cyclobutane dimers for the uvspecific endonucleases purified from either Micrococcus luteus or phage T4-infected Escherichia coli (5)(6)(7)(8)(9)(10) was recently shown to result from the action of two successive enzymatic steps. The fist step of the incision pathway is the action of a pyrimidine dimer DNA glycosylase that cleaves the glycosylic bond between the 5' pyrimidine of the dimer and its corresponding sugar. Following rupture of the glycosylic bond, an apurinic-apyrimidinic (AP)' endonuclease activity cleaves the phosphodiester bond on the 3' side of the base-free sugar. The action of these enzymatic activities leaves a novel structure at * This work was supported by National Cancer Institute Grant CA26716. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ' The abbreviation used is AP, apurinic-apyrimidinic. the site of cleavage which has a 3' apyrimidinic deoxyribose and a 5' pyrimidine cyclobutane dimer (Fig. 1).
The existence of this type of nicked structure raises several questions regarding subsequent enzymatic steps in the dark repair pathway. A major question regarding the excision repair of ultraviolet light-irradiated, M. luteus uv-specific endonuclease-treated DNA is the ability of the nicked DNA to act as a substrate for DNA polymerase. If the structure is not a substrate, what further enzymatic steps are required to convert the nick into a substrate for DNA polymerase?
To investigate these questions, we examined the ability of ultraviolet light-irradiated DNA that had been treated with the M. luteus uv-specific endonuclease to serve as a substrate for subsequent enzymatic treatments. For these studies, DNA fragments of defined sequence labeled at the termini were used. The use of such DNA fragments permits detailed analysis of the reactions that occur at individual dimer sites within a complex sequence.
Here we show that the product of strand scission with M. luteus uv-specific endonuclease is not a substrate for either E. coli polymerase (Pol I) or the Klenow fragment of E. coli polymerase I. However, human placental AP endonuclease, which is shown to remove the 3' A P deoxyribose moiety, converts the DNA to a substrate for DNA polymerase I. In related experiments with the human placental AP endonuclease and bacterial alkaline phosphatase, we further characterize the site of action of the M. luteus uv-specific endonuclease.
Strand scission at cyclobutane dimers results in a 3' AP deoxyribose with a 3' hydroxyl group and a 5' phosphate. In contrast to earlier studies, we also show that this nick cannot be sealed by the action of the T4 DNA ligase.

EXPERIMENTAL PROCEDURES
DNA-Fragments of defined sequence, labeled at either the 5' or 3' terminus, were prepared from the pLJ3 plasmid. The isolation and labeling of these fragments were described previously (11). The sequences of the DNA substrates are given below. Unprimed nunbers refer to DNA labeled at the 5' terminus and primed numbers refer to DNA labeled at the 3' terminus.  Ultraviolet light-irradiated DNA was prepared by exposure to 254nm light from a germicidal lamp (GE G15T8) at a dose rate of 7.5 J/ m2/s and a total dose of 5000-7500 J/mz. The DNA was treated with enzymes as described below. Alkaline hydrolysis was perfomed by incubating the DNA in 0.01 M Tris-HC1, pH 7.4, 0.001 M Na2EDTA, and 0.1 M NaOH for 30 min at 90 "C. The DNA was precipitated and the fragments resolved on 12% polyacrylamide gels that contained 7 M urea (11). Products of the DNA fragments generated by the dimethylsufate (12), hydrazine (12), and neocarzinostatin (13) DNAsequencing reactions were layered on adjacent lanes of the gels to permit determination of the length of each scission product. The labeled DNA fragments were visualized by autoradiography.
M. luteus uu-specific Endonuclease-M. luteus uv-specific endonuclease was provided by R. Grafstrom and L. Grossman, Johns Hopkins University. The C-75 fraction (14) was used in these experiments and contained lo5 units/ml (1 unit of activity produces one break per nanomole of uv 4x174 DNA in 20 min at 37 "C). DNA was incubated in a reaction volume of 50 pl of 0.01 M Tris HCl, pH 7.4, 0.05 M NaC1, and 0.001 M NazEDTA with an excess of the uv-specific endonuclease. Reactions were incubated for 60 min at 37 "C and were terminated by precipitation with 0.3 M Na acetate and 3 volumes of ethanol at -70 "C, by heating at 60 "C for 5 min before precipitation, or by sequential extraction with phenol, chloroform, and ether before precipitation. Human Placental AP Endonuclease-Human placental AP endonuclease was provided by N. Shaper and L. Grossman Quantitation-The amount of each enzymatic cleavage product was determined by measurement of the amount of Cerenkov radiation in each band relative to the total Cerenkov radiation in the lane.

RESULTS
Site of Incision of the J A P Endonuclease Associated with the M. Luteus uu-specific EndonucZease-DNA fragments of known sequence were used as substrates for these experiments. Resolution of the cleavage products produced by enzymatic treatment of these molecules on high resolution polyacrylamide gels permitted analysis of the events that occurred at individual cleavage sites.
Previous studies demonstrated that strand scission by the M. luteus uv-specific endonuclease involves sequential action of a pyrimidine dimer DNA-glycosylase and an AP endonuclease that cleaves the phosphodiester bond 3' to the base-free sugar (5, 6). These experiments did not distinguish between two alternative sites of scission: cleavage of the bond between the base-free sugar and the phosphate or cleavage between the phosphate and the 3' nucleotide (see Fig. 1). These two possibilities can be distinguished by analysis of the 5' termini of the nicks, as one route would leave a 5' phosphoryl group, whereas the other would create a 5' hydroxyl terminus. The presence or absence of a 5' phosphoryl group was tested by examination of the sensitivity of the cleavage product to bacterial alkaline phosphatase, an enzyme that removes both 3' and 5' phosphoryl groups from polynucleotides ( Fig. 2) (16). Removal of a 5' phosphoryl group should decrease the charge to mass ratio of a DNA fragment and thereby retard the electrophoretic mobility of the unphosphorylated fragment relative to the phosphorylated fragment on a high resolution polyacrylamide gel.
To investigate the structure of the 5' terminus of the nick, a DNA fragment labeled at the 3' terminus was used. The DNA was labeled by incorporation of a-"P-labeled precursors at the 3' end of a restriction fragment that contained a 5' overlapping terminus (6). This labeling strategy resulted in a fragment that contained a 3' hydroxyl end. Thus, the labeled terminus of this fragment was resistant to the action of the bacterial alkaline phosphatase. The DNA was irradiated, treated with the M . luteus uv-specific endonuclease, denatured, and layered on a high resolution polyacrylamide gel before or after treatment with bacterial alkaline phosphatase. In a parallel series of reactions, the same DNA fragment was subjected to chemical degradation with dimethyl sulfate or hydrazine according to the DNA-sequencing protocol of Maxam and Gilbert (12). These reactions result in DNA fragments that have 5' phosphorylated termini (12). The products of these reactions were also layered on the gel before and after bacterial alkaline phosphatase treatment to serve as positive controls for the activity of the phosphatase. The result of such an experiment is shown in Fig. 3. Treatment of the products of the M. luteus reaction with bacterial alkaline phosphatase resulted in a decrease in the electrophoretic mobility of each scission product. A similar decrease in mobility after phosphatase treatment was evident for each of the products of the DNA-sequencing reactions. The change in electrophoretic mobility was that predicted for loss of a single phosphate group. Therefore, we conclude that the 5' terminus of the nick is phosphorylated and that the AP endonuclease activity associated with the M. luteus uv-specific endonuclease cleaves the phosphodiester bond between the base-free sugar and the phosphate (alternative 1, Fig. 1). The result i s consistent with an earlier report (18)    erase, the elongated molecules should have migrated more slowly than the scission products, and new bands should have appeared in the higher molecular weight region of the gel. Fig.  4A shows that the electrophoretic mobility of the M . luteus scission products was not altered in rections that included either DNA polymerase I or the Klenow fragment of the polymerase I in the presence or absence of nucleotide precursors.
As a positive control for this type of analysis of the polymerase function of Pol I or the Klenow fragment, similar experiments were performed with a known substrate for polymerization. For these experiments, 5' end-labeled DNA was first nicked with DNase I and then treated with Pol I or the Klenow fragment of Pol I in the presence or absence of deoxynucleotide triphosphate precursors. DNase I leaves a 3' hydroxyl terminus that serves as a site for initiation of DNA polymerase (20). When the DNase I-nicked DNA was incubated in the absence of deoxynucleotide triphosphate precursors with either Pol I or the Klenow fragment, there was no significant change in the electrophoretic mobility of the DNA fragment (Fig. 5). When thymidine triphosphate was included in the reaction mix for either enzyme, some of the cleavage products migrated more slowly on the gel. These products correspond to positions of the sequence at which thymidine should have been added as the next nucleotide. When all four deoxynucleotide triphosphate precursors were included in the reaction mixture, there was substantial elongation of the fragments generated by the action of DNase I. Thus, the electrophoretic mobility of DNA fragments on high resolution polyacrylamide gels can be used to examine the incorporation of deoxynucleotide triphosphate precursors into an end-labeled substrate at each nick.
The experiment pictured in Fig. 4A shows that DNA nicked a t dimer sites with the M. luteus enzyme was not a substrate for the polymerase activities of either Pol I or the Klenow fragment of Pol I. Furthermore, the experiment performed in the absence of deoxynucleotide triphosphate precursors shows that the nicked DNA is also not a substrate for the 3' to 5' exonuclease activities of the enzymes as no digestion of the cleavage products was noted. However, the conditions for the 3' to 5' exonuclease were not optimal, as the 3' to 5' activity requires single-stranded DNA. In a separate experiment, DNA was heat denatured after treatment with the M . luteus enzyme and prior to incubation with the polymerases. Incubation of denatured DNA with the Klenow fragment resulted in extensive degradation of the full length DNA species (Fig. 6). The full length molecules contain a 3' hydroxyl group and, therefore, can serve as substrates for the exonuclease activity of the enzyme. These species completely disappear from the high molecular weight region of the gel. Short, incomplete digestion products were evident in the low molecular weight region of the gel after digestion. These probably represent the for Incision Repair limit digestion products of the full length DNA species. However, the DNA scission products produced by cleavage at the dimer sites by the M. luteus enzyme were resistant to digestion by the 3' to 5' exonuclease activity of the polymerases. Furthermore, following treatment with the Klenow fragment, these DNA scission products underwent a change in electrophoretic mobility upon alkaline hydrolysis characteristic of the 3' base-free sugar. Identical results were obtained with Pol I (not shown). These experiments demonstrated that the 3' deoxyribose terminus of the nick generated by the M . luteus uv-specific endonuclease at dimer sites is not a substrate of DNA polymerase I of E. coli. The enzyme cannot initiate DNA synthesis at the terminus, nor can it initiate degradation in the 3' to 5' direction. The observation that the electrophoretic mobility of the scission products is not altered by incubation with the polymerases implies that DNA polymerase I does not remove the 3' deoxyribose. sible for the inhibition of Pol I activity. T o investigate this possibility, we attempted to remove this group enzymatically.

Conversion of the Incision Site to a Substrate for DNA
A 5' end-labeled DNA fragment was first treated with the M. luteus uv-specific endonuclease and then treated with human AP endonuclease (15). The human AP endonuclease was selected since it cleaves the phosphodiester bond 5' to the AP deoxyribose and leaves a 3' hydroxyl terminus. If the AP endonuclease removed both the 3' deoxyribose and the intervening phosphate from the terminus of the M . luteus cleavage site, the result would be a 3' terminus that should serve as a primer for DNA polymerase I (Fig. 2).
The possibility that the human AP endonuclease was active at the dimer cleavage site was first investigated by examination of the electrophoretic mobility of the DNA products on polyacrylamide gels before and after treatment with the AP endonuclease. Removal of both the phosphate and deoxyribose from the 3' terminus of the fragment should alter the electrophoretic mobility of the product. Moreover, after AP endonuclease treatment, no further change in electrophoretic mobility of the product should occur after alkaline hydrolysis. In contrast, the electrophoretic mobility of the untreated scission product would be altered upon alkaline hydrolysis since the p-elimination reaction that occurs at high pH will result in removal of the deoxyribose leaving a 3' phosphorylated terminus (6) (Fig. 2).
A series of such reactions is pictured in Fig. 7. Alkaline hydrolysis of the M. luteus uv-specific endonuclease-treated DNA resulted in an increase in the electrophoretic mobility of cleavage products, the expected consequence of the loss of the deoxyribose (5-7). Treatment of the same substrate with the human AP endonuclease also resulted in an increase in the electrophoretic mobility of the cleavage products. However, in the latter case, the electrophoretic mobility of the cleavage products was intermediate between the mobility of the untreated cleavage products and the mobility of the fragments produced by alkaline hydrolysis. This was the anticipated electrophoretic mobility for a product that differed from the rapidly migrating species by loss of a terminal phosphate. Exposure of the AP endonuclease-treated DNA products to alkaline hydrolysis did not result in further changes in the electrophoretic mobility of the cleavage products. We conclude that the human AP endonuclease does recognize and act to remove the terminal 3' deoxyribose. The scheme pictured in Fig. 2 summarizes the reactions described above.
We examined the ability of DNA that had been treated sequentially with both the M. luteus uv-specific endonuclease and the human AP endonuclease to serve as a substrate for DNA polymerase I of E. coli. No change in the electrophoretic mobilities of the cleavage products was observed upon incubation of the DNA with E. coli DNA polymerase I or the Klenow fragment of Pol I in the absence of nucleotide precursors (Fig. 4B). Limited elongation of the cleavage fragments is evident in the reactions that contained only thymidine triphosphate. The products observed are consistent with the addition of thymidine residues to the 3' end of the cleavage products, as is expected from the these reactions did not require the 5' to 3' exonucleolytic activity of Pol I as the Klenow fragment of Pol I is deficient in this activity. It is likely that a displacement of the strand ahead of the growing chain occurred in these reactions. This is reasonable in view of the relatively short length of the DNA duplex fragments used as substrates in these experiments.

The Nick Generated by M. luteus uv-specific Endonuclease Cannot Be Joined by T4 Polynucleotide
Ligase-Genetic evidence suggests that DNA ligase is involved in excision repair pathways (28,29). The product of cleavage of ultraviolet light-irradiated DNA with the uv-specific endonuclease has both a 3' hydroxyl group and a 5' phosphate at the internal nicks. This structure suggests that the nick may be a suitable substrate for the rejoining of nicks by T4 DNA ligase. However, unlike a typical substrate, the 3' hydroxyl is associated with an apyrimidinic deoxyribose and the 5' phosphate is associated with a pyrimidine that is covalently linked via a cyclobutane bond to another pyrimidine base (Fig. 1). T o determine whether such nicks can serve as substrates for T4 DNA ligase, DNA was irradiated with ultraviolet light, cleaved with uv-specific endonuclease, and incubated with T 4 DNA ligase. Following the ligation reaction, the DNA fragments were precipitated and analyzed by electrophoresis on high resolution polyacrylamide gels as above. In this analysis, the action of T4 DNA ligase was examined a t every nick. DNA that was nicked with DNase I, a known substrate for T4 DNA ligase, was also used in these experiments. The results are shown in Fig. 8.
The fragments produced by DNase I treatment of the 3' end-labeled DNA were sealed by the action of T4 DNA ligase as evidenced by the disappearance of such fragments (Fig. 8). The disappearance of these fragments could not be accounted for in terms of degradation of the DNA since there was no loss of total radioactivity in the sample as compared to the samples incubated in the absence of the T4 DNA ligase. In contrast, the nicks generated by the uv-specific endonuclease were not sealed by the action of T4 DNA ligase. T o determine whether the uv-irradiated, uv-specific endonuclease-treated DNA was inhibitory for the T4 DNA ligase, DNase I-treated DNA and uv-irradiated uv-specific endonuclease-treated DNA were mixed prior to incubation with the ligase. It is evident that the DNase I nicks were ligated even in the presence of endonuclease-treated DNA. The possibility that the uv-specific endonuclease remained bound to the nick, thus preventing the action of DNA ligase, was examined by extraction of the M. luteus enzyme-treated DNA with phenol and I @Or " " " _" " " " " I( " " " " " " " " _ chloroform prior to incubation with T4 DNA ligase. Such treatment did not alter the ligation of DNase I nicks; yet T 4 DNA ligase was still unable to rejoin the nicks generated by the uv-specific endonuclease (Fig. 8).
These observations were c o n f m e d by quantitative measurement of the amount of radioactivity in the different regions of the gel. Significant ligation of the DNAse I nicks was seen with 0.001 unit of T4 DNA ligase/reactions, and maximal ligation of the DNase I nicks was observed with 0.2 unit of T4 DNA ligase/reaction (Fig. 9). Ligation of nicks generated by the uv-specific endonuclease was not observed even a t 0.4 unit of T4 DNA ligase/reaction. These results indicate that the product of cleavage of uv-irradiated DNA with the uv-specific endonuclease is not a substrate for T4 DNA ligase. Previous experiments that indicated that some of the nicks introduced by the M. luteus or T4 uv-specific endonuclease can be sealed by T4 polynucleotide ligase (18,30) probably reflect the ligation of nicks introduced into the substrates by endonucleases of the DNase I type that contaminated the enzyme preparations. In any event, it is not clear how ligation of the nick would facilitate repair of the lesion, as the resultant DNA would still contain an AP site as well as the cyclobutane pyrimidine dimer.

DISCUSSION
The dark repair pathway of pyrimidine dimers involves excision of the pyrimidine cyclobutane dimer followed by resynthesis of DNA. Two enzymes involved in the fwst steps of this pathway, the M. luteus and T4 uv-specific endonuclease, carry out two functions for this pathway, recognition of the dimer and incision at the site of the dimers. Incision of the DNA by these enzymes occurs by cleavage of the Nglycosyl bond between the 5' pyrimidine of the dimer and the corresponding sugar followed by scission of the phosphodiester bond. Experiments presented here demonstrate that the second cleavage event occurs on the 3' side of the AP deoxyribose leaving a 5' phosphoryl terminus at the site of the break.
The 3' terminus of the nick produced at the dimer site is a deoxyribose that has a free 3' hydroxyl group. This unusual structure raises several questions regarding subsequent steps in the enzymatic repair of the lesions. Accurate repair of a pyrimidine dimer probably involves resynthesis of DNA from the complementary strand. However, the experiments presented here as well as the recent experiments by Warner et al. (31) using DNA nicked at cyclobutane dimers by the T4 uvspecific endonuclease as a polymerase substrate demonstrate that the 3' deoxyribose cannot serve as a site for initiation of polymerization by DNA polymerase I. The failure of DNA polymerase I to initiate synthesis at this site implies that other enzymatic activities must be involved in the dark repair pathway. It is anticipated that these enzymes would convert the 3' deoxyribose terminus to a suitable primer for DNA polymerases. The conclusion that another enzyme must be involved in this process is strengthened by the observation that DNA polymerase I does not remove the 3' deoxyribose from either double-or single-stranded DNA.
The 3' deoxyribose is, however, a substrate for the human AP endonuclease. This enzyme removes the 3' deoxyribose, leaving a 3' hydroxyl terminus. Ultraviolet light-irradiated DNA exposed to both the M. luteus enzyme and the human AP endonuclease is a substrate for DNA polymerase I. In these experiments, every scission product of the M. luteus uvspecific endonuclease was converted to a primer for the polymerase as evidenced by elongation of aU of the scission products in the polymerase reactions.
A Three-step Incision Model-We speculate that excision repair in M. luteus of cyclobutane dimers may involve an AP endonuclease activity similar to that of the human AP endonuclease. In this view, incision at dimer sites would occur by the combined action of the dimer DNA glycosylase and an AP endonuclease that acts 3' to the AP site. Subsequently, the 3' deoxyribose would be removed by the action of an AP endonuclease that removes the 3' deoxyribose, leaving a 3' hydroxyl end which is suitable for initiation of DNA synthesis (Fig. 10). Several AP endonuclease activities have been detected in extracts of M. luteus and these may serve the proposed function (36, 37).
There is some evidence that a three-step incision pathway may be a general mechanism involved in DNA repair in bacteria. The uv-specific endonuclease which is encoded by bacteriophage T4 also creates nicks at cyclobutane dimer sites via the combined action of a dimer DNA glycosylase and a 3' AP endonuclease (7-10). The E. coli enzyme endonuclease 111 is both an N-glycosylase which is active at sites of dihydrothymidine and an AP endonuclease that cleaves AP sites 3' to the deoxyribose (10, 32). In both cases, the experiments of Warner et al. (31) and Dimple and Linn (10) show that the cleavage products are not substrates for DNA polymerases, but can be converted into suitable substrates by the action of AP endonucleases that act 5' to the AP sites. In E. coli, the removal of several other bases in DNA, for example, uracil (33), N7 methyl guanine (34), and N3 methyl adenine (35), may also proceed via a three-step incision mechanism as the first step in the repair of these lesions involves cleavage of the N-glycosyl bond between the modified base and deoxyribose, leaving an AP site in the DNA.
The net effect of the three-step incision mechanisms for excision of modified bases is that an entire modified nucleotide is removed from the DNA. In the case of removal of single modified bases, DNA polymerization initiated at the 3' hydroxyl terminus of the nick leads to correct repair of the damaged base. In the case of cyclobutane pyrimidine dimers, one more step in the repair process must be evoked, i.e. excision of the dimer that remains on the 5' terminus of the nick.
Two mechanisms may be envisioned for removal of the dimer linked to the DNA by a single N-glycosyl bond: excision of an oligonucleotide by cleavage of a phosphodiester bond 3' to the dimer or cleavage of the N-glycosyl bond that links the dimer to the DNA. DNA polymerase I of E. coli can carry out the former reaction in vitro (22), but at present, it is not known whether this reaction occurs in intact cells. Further experiments are required to investigate the later steps in dark repair pathways.