Human endonucleolytic incision of DNA 3' and 5' to a site-directed psoralen monoadduct and interstrand cross-link.

Human chromatin-associated protein extracts were examined for endonucleolytic activity on a defined 132-base pair DNA substrate containing a single, site-specific 4,5'-8-trimethylpsoralen plus long wavelength ultraviolet light-induced furan side or pyrone side monoadduct or interstrand cross-link. These extracts produced incisions on both the 3' and 5' sides of each of these lesions. The distance between the 3' and 5' incisions at sites of a furan side monoadduct or cross-link was 9 nucleotides, and at sites of a pyrone side monoadduct or cross-link it was 17 nucleotides. Incisions on the 3' side of both types of furan side and pyrone side adducts were similar and were either at the fourth or fifth phosphodiester bond from the adducted thymine, depending upon the adduct. However, greater differences were observed between sites of 5' incision. This incision occurred at the fifth and sixth phosphodiester bonds from the adducted thymine at sites of furan side monoadducts and cross-links, respectively, and at the 13th and 14th phosphodiester bonds at sites of pyrone side monoadducts and cross-links, respectively. Thus, direct analysis of sites of endonucleolytic incision reveals that the location of sites of incision on TMP-adducted substrates depends upon the type of adduct present.

Repair of DNA interstrand cross-links is critical for a number of cellular processes such as transcription and DNA replication and therefore is particularly important for the maintenance of genetic integrity and cellular survival. A nucleotide excision repair mechanism is responsible in both prokaryotes and eukaryotes for the removal of this type of lesion (1)(2)(3). Although the molecular basis of repair of DNA interstrand cross-links has been extensively studied in bacteria, the proteins involved in the removal of these lesions in mammalian cells and their mechanism of action is largely unknown. A critical, rate-limiting step in this repair process is the initial damage recognition and incision step in which a protein specifically locates or recognizes a site of damage and an endonuclease makes an incision on the DNA strand at or near this site. We have recently identified a damage recognition protein in normal human chromatin, which binds to DNA containing interstrand cross-links produced by 4,5Ј,8-trimethylpsoralen (TMP) 1 plus long wavelength ultraviolet (UVA) light, and our studies suggest that it plays a role in the repair of interstrand cross-links (4). However, an endonuclease that specifically incises DNA at sites of these interstrand cross-links has heretofore not been isolated from mammalian cells.
A number of different agents have been shown to produce interstrand cross-links in DNA. One of the most definitive of these, whose reaction with DNA has been well characterized, is psoralen plus UVA light. Psoralens are a group of three-ring heterocyclic furocoumarins that contain two reactive double bonds, a 4Ј,5Ј-double bond in the furan ring and a 3,4-double bond in the pyrone ring (5)(6)(7)(8). Psoralen-DNA adducts are formed in three stages; psoralen first intercalates into the DNA duplex in a noncovalent manner and then, upon photoreaction with UVA light, forms either a furan side or a pyrone side monoadduct with the 5,6-double bond of a pyrimidine, which the majority of the time is a thymine. If the furan side of the molecule is linked to DNA, then further exposure to UVA light leads to production of an interstrand cross-link (5)(6)(7)(8)(9)(10).
In Escherichia coli, the UvrABC nuclease is responsible for repair of psoralen-photoinduced monoadducts and interstrand cross-links. The UvrA, UvrB, and UvrC proteins act in concert to excise these lesions by making sequential incisions on both the 3Ј and 5Ј sides of the modified nucleotide (11)(12)(13)(14)(15)(16)(17). Although repair of psoralen monoadducts and interstrand cross-links has been shown to occur in mammalian cells (18 -26), the precise nature of the initial events and proteins involved in removal of these lesions are less clear. We have isolated from the nuclei of normal human cells a chromatin-associated DNA endonuclease complex, pI 4.6, which recognizes and incises DNA containing TMP or 8-methoxypsoralen plus UVA interstrand cross-links and another, pI 7.6, which recognizes and incises DNA containing psoralen monoadducts (22,23,25,27). Each complex contains proteins involved in damage recognition, chromatin interaction, and endonucleolytic incision, and each has been demonstrated to be involved in the repair process by its ability to correct the repair defect in repair-deficient cells when introduced into them via electroporation (28,29).
In the present study, we describe the construction of a 132base pair oligonucleotide substrate that contains a centrally placed single, site-specific TMP monoadduct or interstrand cross-link. These substrates are so constructed that endonucleolytic activity on either the furan side or the pyrone side of each type of adduct may be selectively studied. We have used these four substrates to directly examine the pattern of incision produced by human chromatin-associated protein extracts that contain both the endonuclease complexes, pI 4.6 and pI 7.6. Our results show that these extracts make incisions in DNA on both the 3Ј and 5Ј sides of each of the two types of furan side and pyrone side adducts (i.e. monoadducts and cross-links). The incision patterns on each of these four substrates are different. The distance between the 3Ј and 5Ј incisions made at sites of furan side monoadducts and cross-links is 9 nucleotides, whereas the distance between the 3Ј and 5Ј incisions made at sites of pyrone side monoadducts and cross-links is 17 nucleotides.

MATERIALS AND METHODS
Chromatin-associated Protein Extracts-Normal human (GM 1989 and GM 3299) lymphoblastoid cell lines (transformed with Epstein-Barr virus) were obtained from the Coriell Institute for Medical Research (Camden, NJ). The cells were grown in suspension culture in RPMI 1640 medium, supplemented with 12.5% fetal calf serum (Grand Island Biological Co.), and harvested under conditions of maximal proliferation (30). Cell cultures were routinely tested for mycoplasma (30).
Cell nuclei were isolated, and the chromatin-associated proteins were separated from the nucleoplasmic proteins in a series of steps as described previously (22,31). Chromatin-associated proteins were dialyzed into 50 mM potassium phosphate (pH 7.1), 1 mM ␤-mercaptoethanol, 1 mM EDTA, 0.25 mM phenylmethylsulfonyl fluoride, and 40% ethylene glycol and then passed through a CM Sephadex column and stored at Ϫ20°C (4,22,31). Protein concentrations were determined by the Bio-Rad protein assay (Bio-Rad).
DNA Oligonucleotide Synthesis and Purification-The DNA substrate used was the region from position 61 to 200 of the nucleotide sequence of the 5 S rRNA gene from Lytechinus variegatus (32). This region was synthesized as seven separate oligonucleotides (ML1-ML7) ( Fig. 1) (33). The full-length products were purified by electrophoresis on 10 or 20% polyacrylamide 7 M urea gels and by subsequent isolation of the bands identified by autoradiography. The labeled phosphates were then removed by calf intestinal alkaline phosphatase (Boehringer Mannheim) and replaced by unlabeled phosphates. These purified oligonucleotides were then used in ligation reactions to produce the 132-bp substrates described below (Fig. 2).
Preparation of the 18-Mer Oligonucleotide (ML3) Containing the Furan Side TMP Monoadduct-The region from position 124 to 141 of the nucleotide sequence of the top strand of the 5 S rRNA gene from L. variegatus (designated ML3) ( Fig. 1), which is approximately at the center of the DNA fragment, was used for construction of the furan side TMP monoadduct. ML3 (18-mer) and ML3A (22-mer) were designed to have a single 5Ј-TpA-3Ј dinucleotide sequence at positions 132 and 133, in place of the 5Ј-ApT-3Ј original dinucleotide sequence, since this sequence has been shown to be a hot spot for photochemical reaction with TMP, which photoreacts with the thymine in this sequence (9,13). ML3A was a part of the sequence of the ML6 oligonucleotide (Fig. 1). The TMP monoadduct was formed as follows. 180 g of 5Ј-phosphorylated ML3 was mixed with 200 g of 5Ј-phosphorylated ML3A, along with ␥-32 P-5Ј-end-labeled ML3 and ML3A as tracers, in 1000 l of 10 mM Tris-HCl, pH 7.6, 0.4 mM EDTA, 100 mM NaCl and 5 g/ml TMP (Sigma). The solution was irradiated in 100-l aliquots with long wavelength ultraviolet (365 nm) light (13 milliwatts/cm2), from a super-high intensity black light lamp (Spectroline model SB-100, Spectronics Corp.) for 5 min. at 20°C. The DNA was extracted with chloroform/ isoamyl alcohol (24:1) to remove unbound TMP, ethanol-precipitated, and then electrophoresed on a denaturing 7 M urea 20% polyacrylamide gel so as to separate the strands and isolate the monoadducted ML3 oligonucleotide. Also formed in this reaction was the monoadducted ML3A oligonucleotide and two types of cross-linked species, since TMP can photoreact with the ML3 and ML3A oligonucleotides in two different orientations. The band containing the ML3 furan side monoadduct (18-mer), which migrated more slowly than the unmodified ML3, was visualized by autoradiography, extracted, and purified as described (33). Labeled phosphates were then removed using calf intestinal alkaline phosphatase and replaced by unlabeled phosphates. DNA was quantitated by measuring absorbance at 260 nm. The yield was 22 g of the ML3 furan side monoadduct. This was stored in aliquots at Ϫ20°C in 10 mM Tris-HCl, pH 7.6, and 0.4 mM EDTA.
Construction of DNA Substrates Containing Site-specific TMP Monoadducts and Interstrand Cross-links-As can be seen in Fig. 1A, the top strand of the DNA substrate (ML8), which is a 136-mer fragment, was assembled from four oligonucleotides: ML1 (20-mer), ML2 (43-mer), ML3 (18-mer), and ML4 (55-mer). Similarly, the bottom strand (ML9), which is also a 136-mer fragment, was assembled from 3 oligonucleotides: ML5 (45-mer), ML6 (50-mer), and ML7 (41-mer). The preparation of DNA substrates containing site-specific TMP monoadducts and crosslinks is outlined in Fig. 2. The top strand (ML8) of the substrate was internally labeled at the 5Ј-end of the ML2 oligonucleotide (thymine at position 81) ( Fig. 1) to visualize the furan side incision events. Similarly, the bottom strand (ML9) of the substrate was internally labeled at the 5Ј-end of ML6 (guanine at position 159) (Fig. 1) to visualize the pyrone side incision events. The full-length labeled top strand (ML8) and bottom strand (ML9) were purified by electrophoresis on a 5% denaturing polyacrylamide gel in order to remove incomplete ligation products. Both ML8 and ML9 had the same mobility on these gels and had migrating along with them the unlabeled complementary ML9 or ML8 strands. The bands corresponding to the full-length ligated products were extracted from the denaturing gel, and the DNA was purified as described (33). These bands represent both the top and the bottom strands in their unhybridized forms. Therefore, it was necessary to reisolate the hybridized duplex full-length DNA substrates on a nondenaturing 5% polyacrylamide gel. These substrates were 132 bp in length with a 4-base overhang on the 5Ј-ends. In these substrates, TMP was adducted to a thymine at position 132 in ML8 and at position 133 in ML9 (Fig. 1A). Preparation of the specific ligated, psoralen-monoadducted, and cross-linked substrates is described below.
For construction of the furan side monoadduct, ML2 (180 pmol) was dephosphorylated with calf intestinal alkaline phosphatase and 5Ј-endlabeled with 1 mCi of [␥-32P] ATP (6000 Ci/mM; DuPont NEN) by using T4 polynucleotide kinase (Pharmacia). ML2 was freed from protein, salts, and unincorporated label by passing through Nensorb 20 nucleic acid purification cartridges according to the manufacturer's directions (DuPont NEN). ML2 was then eluted with 50% ethanol and dried in a Speed Vac in a siliconized eppendorf tube. To the dried pellet, 180 pmol of 5Ј-phosphorylated ML1, ML4, ML5, ML6, and ML7 and 540 pmol of ML3 oligonucleotide containing the furan side TMP monoadduct were added in a total volume of 150 l containing 10 mM Tris-HCl, pH 7.6, 0.4 mM EDTA, and 25 mM NaCl (Fig. 2). The amount of the ML3 furan side monoadduct intermediate added was three times that of the other oligomer components, since the efficiency of ligation was dependent upon the amount of ML3 monoadduct in the reaction mixture. The mixture was heated at 95°C for 3 min, and then the DNA strands were hybridized at 60°C for 30 min followed by slow cooling to 25°C. To this mixture 100 units of DNA ligase (Boehringer Mannheim) was added, and ligation was carried out at 12°C for 90 min, followed by 4°C overnight. Ligation was found to be equally efficient when carried out at 22°C for 1 or 2 h. The ligated full-length product was isolated by electrophoresis of the sample through a denaturing 5% polyacrylamide gel. As mentioned above this product also contained the complementary unlabeled full-length bottom strand (ML9). The purity of the duplex DNA was ensured by repurifying this product on a nondenaturing 5% polyacrylamide gel (data not shown). This product represented the furan side monoadducted substrate in which TMP was adducted to a thymine at position 132 on ML8 (Fig. 1). The preparation of undamaged DNA substrate was similar to this except that an undamaged ML3 sequence was used for this purpose.
For construction of the furan side cross-linked DNA, half of the above furan side monoadducted substrate in 100 l of 10 mM Tris-HCl, pH 7.6, and 0.4 mM EDTA was irradiated with the UVA light (20 milliwatts/ cm 2 ) for 25 min at 20°C (Fig. 2). The furan side cross-linked substrate was purified by electrophoresis on a denaturing 5% polyacrylamide gel.
The strategy for formation of the pyrone side interstrand cross-link was similar to that used for the furan side cross-link. However, in this instance in the ligation reactions the ML6 rather than the ML2 oligonucleotide was 5Ј-end labeled (Fig. 2). This resulted in the bottom strand (ML9) being internally labeled rather than the top strand (ML8) (Fig. 1). In this substrate the pyrone ring of TMP was adducted to a thymine at position 133 on ML9 (Fig. 1A).
The pyrone side monoadducted substrate was prepared according to a protocol described by Yeung, Dinehart, and Jones (34). The psoralen cross-link was reversed by treatment with hot alkali, which has been shown to specifically cleave the psoralen cross-link at the furan side, whereas the pyrone side monoadduct is relatively resistant to this treatment (34). Half of the above pyrone side cross-linked substrate in 150 l of 10 mM Tris-HCl, pH 7.6, 0.4 mM EDTA, and 100 mM KOH was heated at 90°C for 30 min, and the product was purified on a denaturing 5% polyacrylamide gel (Fig. 2). The preparation of the undamaged DNA substrate was similar to this except that it contained an undamaged ML3 sequence on the top strand.
All of the above substrates were cleaned up with the Nensorb 20 nucleic acid purification cartridges according to the manufacturer's protocol (DuPont NEN) and were stored in aliquots at Ϫ20°C in 10 mM Tris-HCl (pH 7.6) and 0.4 mM EDTA.
Assay for Endonucleolytic Incision-For the incision reactions approximately 100 fmol (concentration determined by In Vitrogen DNA Dip Stick kit, version 3.2) of the 132-bp labeled DNA substrate, either TMP-modified or unmodified, was treated with 10 g of the human chromatin-associated protein extract in a volume of 40 l containing 50 mM Tris-HCl, pH 7.6, 10 mM MgCl 2 and 100 ng of salmon sperm DNA (Strategene) for 3 h at 37°C. The reactions were terminated by the addition of 15 mM EDTA and samples deproteinized using Nensorb 20 nucleic acid purification cartridges. The incision products were eluted with 50% ethanol and dried in a speed vac. Before sequence analysis, all samples containing TMP cross-linked DNA substrates were photoreversed by irradiation with 254-nm UV light at a dose of 1500 microwatts/cm 2 for 5 min at room temperature using a germicidal UV lamp (G15T8, American Ultra Violet Co.). The pellet was dissolved in 90% formamide, 1 ϫ TBE containing 0.25% bromphenol blue and xylene cyanol. The DNA samples were denatured by heating at 90°C for 1 min and quickly cooled down in ice. Incision sites were identified by running the normal endonuclease reactions along with the sequencing ladders of undamaged substrates in a denaturing 6% polyacrylamide gel using the method of Maxam and Gilbert (33). Sequencing gels were exposed to DuPont Reflection autoradiography film (NEF-496), with a reflection intensifying screen, overnight at Ϫ80°C. The percentage of substrate converted to incision products was quantitated from the x-ray films from each experiment using a Leica Quantimet 500 Image Analysis System with Sony video camera. Means of two to six determinations for each set of experiments were obtained.

Construction of a 132-bp DNA Substrate Containing
Sitespecific TMP Monoadducts or Cross-links-A 132-bp DNA substrate, with a 4-base overhang on the 5Ј-end of each strand, was constructed by ligating together seven component oligomers (Fig. 1). This DNA fragment contained a TMP-modified thymine, at a centrally defined position, which formed either a furan side or pyrone side monoadduct or interstrand cross-link.
An 18-mer oligonucleotide (ML3), containing the furan side monoadduct of TMP, was used as an intermediate in the construction of all of these unique TMP-modified 132-bp DNA substrates ( Figs. 1 and 2). The efficiency of formation of TMP monoadducts on this intermediate was determined by examining the ability of TMP-modified ML3 to form interstrand crosslinks when hybridized to complementary ML3A and then irradiated with UVA light. It was found that under these conditions all of the furan side monoadducts were able to form interstrand cross-links (data not shown).
To construct the 132-bp furan side monoadducted DNA, TMP-modified ML3 was ligated together with the other six component oligomers. The upper strand was internally labeled on the 5Ј-end of oligonucleotide ML2 (Figs. 1 and 2). The ligation reaction occurred with over 50% efficiency. The purity of the duplex DNA was ensured by repurifying this product on a nondenaturing gel (data not shown). To form the furan side cross-linked DNA, the above monoadducted substrate was photoreacted with 365-nm UV light (Fig. 2). The formation of the furan side cross-link was also very efficient. Greater than 90% of the monoadducted substrate was converted to the crosslinked form (Fig. 3A, upper band, lane 1).
For construction of the pyrone side adducted substrates, TMP-monoadducted ML3, was used in the ligation reactions in which the bottom strand was internally labeled on the 5Ј-end of ML6 (Fig. 2). As was the case for the top strand, the ligation reaction for the bottom strand occurred with over 50% efficiency. The purity of the duplex DNA was ensured by repurifying this product on a nondenaturing gel (data not shown). The monoadducted 132-bp substrate was then driven to the cross-linked form by exposure to 365-nM UV light (Fig. 2). Greater than 90% of the monoadducted substrate was converted to the cross-linked from (Fig 3A, upper band, lane 2). The lower bands in Fig. 3A, lanes 1 and 2,  unreacted full-length monoadducted substrates. To form the pyrone side monoadduct, the pyrone side cross-link was reversed by treatment with hot alkali (Fig. 2). The alkali reversal was also very efficient since the cross-linked substrate was completely reversed to the monoadducted form (Fig. 3B).
Identification of Sites of Endonucleolytic Incision at TMP Furan Side Adducts-The human chromatin-associated protein extracts produced incisions on the 132-bp substrate on both the 3Ј and 5Ј sides of TMP furan side monoadduct. Two bands were produced that represent incisions made at the fifth phosphodiester bond 3Ј to the adducted thymine and at the fifth phosphodiester bond 5Ј to this same modified base (Fig. 4, lane 1; bands representing 3Ј and 5Ј incisions comprise 4.5 and 2.0% of the substrate, respectively). These incisions were not seen on undamaged DNA (Fig. 4, lane 3). In these reactions, salmon sperm DNA was used as a nonspecific competitor DNA. The same results were obtained when two other nonspecific competitor DNAs were used: poly(dI-dC)⅐(dI-dC) and a 260-bp fragment from the 5 S rRNA gene from L. variegatus, which included the sequence for the 132-bp substrate (data not shown). These dual incisions were also observed in the absence of competitors (data not shown). Minor bands were observed that were also present on the undamaged DNA and represent incisions by nonspecific nucleases.
The 3Ј and 5Ј incisions were reduced (67 and 64%, respectively) upon the addition to the reaction mixture of 100 ng of unlabeled competitor DNA, which contained TMP furan side monoadducts (Fig. 5, lane 1; 3Ј and 5Ј incisions represent 4.9 and 2.8% of substrate, respectively, compared with lane 2, 3Ј and 5Ј incisions representing 1.6 and 1.0% of substrate, respectively). Nonspecific cuts were reduced by only 38% (Fig. 5, lane 1 compared with lane 2). Some reduction in nonspecific cuts would be expected since nonspecific nucleases would be acting on the competitor as well as on the substrate DNA. The dual incisions were totally abolished when 500 ng of the competitor was added (lane 4). In several of the figures, the DNA sequence around the damaged sites, when reacted with the protein extracts, appeared to be relatively protected. Since this pattern is seen on both damaged as well as on undamaged DNA (Fig. 4), it may represent sequence-specific protein binding rather than binding by a damage-specific binding protein, particularly since the 3Ј and 5Ј endonucleolytic incisions are seen only on the damaged and not on the undamaged DNA.
For those substrates in which incisions at sites of interstrand cross-links were examined, the cross-linked DNA was photoreversed by UVC light before gel electrophoresis. This was done in order to break the interstrand cross-link and allow us to detect potential incisions on both sides of the lesion on the same strand in a single assay (35,36). Experiments were carried out to make sure that irradiation of the cross-linked DNA with UVC light was not affecting the electrophoretic migration pattern of the incised DNA. Monoadducted DNA, reacted with the human extract and irradiated with UVC light prior to electro-  phoresis, produced patterns of migration that were the same as when the DNA was not UVC-irradiated (data not shown). Therefore, irradiation with UVC light did not produce a shift in migration pattern of the DNA bands.
Examination of the ability of the normal extracts to incise DNA containing a furan side interstrand cross-link indicated that dual endonucleolytic incisions were also produced on this substrate (Fig. 4). Incisions occurred at the fourth phosphodiester bond 3Ј to the adducted thymine and at the sixth phosphodiester bond 5Ј to this modified base (Fig. 4, lane 2; 3Ј and 5Ј incisions, 4.4 and 2.8% of substrate, respectively). For both the furan side monoadduct and cross-link, the distance between sites of incision was 9 nucleotides. These same studies were carried out with similar extracts from HeLa cells, and the same results were obtained (data not shown).
Sites of Endonucleolytic Incision at TMP Pyrone Side Adducts-Incision signals by the protein extracts were not as strong on DNA containing pyrone side adducts. As is shown in Fig. 6, two bands were produced when the extract was reacted with the pyrone side monoadducted DNA substrate. An incision was produced at the fifth phosphodiester bond 3Ј to the adducted thymine and at the 13th phosphodiester bond 5Ј to this modified base (Fig. 6, lane 1, 3Ј and 5Ј incisions, 2.7 and 1.9% of substrate, respectively; lane 2, 3Ј and 5Ј incisions, 3.2 and 2.3%, of substrate, respectively). The extracts did not incise the undamaged DNA substrate (Fig. 6, lanes 3 and 4).
Dual incisions were also produced by the normal extracts on pyrone side cross-linked DNA. However, as noted above, these incisions were produced less efficiently than for the other TMPadducted substrates. The 3Ј incision was at the fourth phosphodiester bond from the adducted thymine, and the 5Ј incision was at the 14th phosphodiester bond from this modified thymine (Fig. 7; lane 1, 3Ј and 5Ј incisions, 2.5 and 1.7% of substrate, respectively; lane 2, 3Ј and 5Ј incisions, 2.7 and 2.1% of substrate, respectively). The distance between sites of incision on both the pyrone side monoadducted and cross-linked substrates was 17 nucleotides, a greater distance than that between sites of incision on furan side adducted DNA.
All of the incision experiments described above were carried out using two different normal human cell lines. Each experiment was repeated 10 -20 times using two to four different extractions from each cell line and three to four different substrate preparations. In all instances the results obtained were always the same. The addition of ATP was not required for any of these incision events.
Effects of Ionic Conditions on Endonucleolytic Incision-The effects of ionic conditions on endonucleolytic incision were examined on the furan side monoadducted 132-bp DNA substrate. Mg 2ϩ (10 mM) was needed for both the 3Ј and 5Ј incisions (Fig. 8, lane 1; 3Ј and 5Ј incisions represent 3.1 and 1.6% of the substrate, respectively). Incision signals were much weaker when the Mg 2ϩ was reduced to 2 mM (Fig. 8, lane 2; 3Ј and 5Ј incisions, 1.1 and 0.9% of the substrate, respectively) or eliminated (Fig. 8, lane 3; 3Ј and 5Ј incisions, 0.4 and 0.5%, respectively). The presence of EDTA inhibited both of the incisions (Fig. 8, lane 4; 3Ј and 5Ј incisions, 0.6 and 0.7%, respectively). The addition of Zn 2ϩ to the reaction had minimal influence on the incisions (Fig. 8, lane 5; 3Ј and 5Ј incisions, 3.5 and 1.9%, respectively). Mn 2ϩ (10 mM), in the presence of 10 mm Mg 2ϩ , slightly enhanced both the 3Ј and 5Ј incisions (Fig. 8,  lane 6; 3Ј and 5Ј incisions, 4.0 and 2.0%, respectively). Incision activity was decreased when 2 mM manganese alone was used (Fig. 8, lane 7; 3Ј and 5Ј incisions, 1.9 and 1.2%, respectively). DISCUSSION An important step in elucidating the mechanisms of action of human DNA repair proteins is to determine the events involved in the critical initial damage recognition and incision step in the repair process. Engineering and construction of DNA substrates of defined sequence and length containing a single site-directed lesion provide an excellent tool for investigation of the exact sites and precise nature of the incision events that can occur at such a lesion. Utilizing such a uniquely modified 132-bp substrate, which contained a single site-directed TMP adduct placed on a central thymine, we have directly examined the incision events involved in repair of psoralen plus UVA light-induced DNA monoadducts and interstrand cross-links by human chromatin-associated proteins. The results show that the human extracts produced incisions in DNA on both the 3Ј and 5Ј sides of each of the four types of TMP-thymine adducts. The distance between the 3Ј and 5Ј incisions for furan side monoadducts and cross-links was always 9 nucleotides, whereas the distance between these incisions for the pyrone side monoadducts and cross-links was always 17 nucleotides. This increased distance between sites of incision for the pyrone versus the furan side adducts could be due to the fact that, since TMP is an asymmetric molecule, more distortion may occur in the DNA in the vicinity of the thymine adducted to the pyrone side of TMP as compared with that of the thymine bound to the furan side (7,9,10,37). The sites of incision on the 3Ј side of both types of furan side and pyrone side adducts were similar; they were either at the fourth or fifth phosphodiester bond from the adducted thymine, depending upon the adduct (Fig. 9). Greater differences were observed, however, in the sites of incision on the 5Ј side. These incisions occurred at the fifth and sixth phosphodiester bonds from the adducted thymine at sites of furan side adducts and at the 13th and 14th phosphodiester bonds at sites of pyrone side adducts (Fig. 9). Thus, production of a potentially greater distortion in the DNA by the pyrone side adducts appears to have little effect on the site of the 3Ј incision but to increase markedly the distance of the 5Ј incision from the adducted thymine, compared with the distance of this incision from the site of furan side adducts.
Svoboda et al. (26) have examined removal of psoralen monoadducts by HeLa whole cell extracts. They utilized as substrate a plasmid DNA containing a single 4Ј-hydroxymethyl-4,5Ј8trimethylpsoralen (HMT) furan side monoadduct. Based on their results using an excision assay, which examined the sizes of DNA fragments excised from HMT-monoadducted DNA after treatment with cell extracts (26), and based on the pattern of excision of thymine dimers from DNA (26,38), they proposed that fragments of 27-32 nucleotides in length are excised and predicted that the sites of incision occur at the fourth or fifth phosphodiester bond 3Ј and the 22nd to 24th phosphodiester bond 5Ј to the adducted thymine. Their site of 3Ј incision is similar to the one we have observed, but their proposed site of 5Ј incision is significantly farther from the adducted thymine than ours. An important difference, however, between the methods utilized in the present study and those used by Svoboda et al. is that we are looking at sites of endonucleolytic incision directly and they used indirect methods of analysis. For their studies that examined the size of the DNA fragment excised, DNA was labeled at the eighth phosphodiester bond 5Ј to the HMT furan side adduct (26). However, any fragment in which the 5Ј incision occurred between the seventh phosphodiester bond and the adduct would not have been detected in their system. Therefore, using this labeling scheme, Svoboda et al. (26) would not have been able to detect the 5Ј incision site we describe at the 5th phosphodiester bond from the adduct. Svoboda et al. (26) also predicted sites of incision for HMT monoadducts based on the excision pattern of thymine dimers. In these studies the boundaries of the repair patch synthesized, after excision of the damaged fragment, were examined and used in the determination of the sites of 3Ј and 5Ј incision (26,38). By using DNA fragments containing repair patches with phosphorothioate linkages and by analyzing resistance of these linkages to exonuclease III digestion and their sensitivity to cleavage by iodine, they delineated the borders of the repair patch (26,38). Since whole cell extracts were used for these studies, it is possible that specific as well as nonspecific endonucleolytic and exonucleolytic activities present in these extracts and postincision events could contribute to the size of the DNA fragment excised. The size of this fragment in turn determines the size of the repair patch generated. Thus, these events could account for the larger distance, reported by Svoboda et al. (26), between the sites of the 3Ј and 5Ј incision. This is in contrast to results obtained when pure proteins are used in this type of excision assay, where a more direct correlation can be obtained between the size of the repair patch generated and the distance between sites of endonucleolytic incision, as has been shown for the UvrABC nuclease (39).
Comparison of the incision pattern produced on psoralenadducted substrates by the human endonucleases in the present study with that produced by the UvrABC nuclease (11)(12)(13)(14)(15)(16)(17) shows that the sites of 3Ј incision are fairly similar. Differences were observed, however, in the sites of 5Ј incision. The distance between sites of incision on all four types of psoralen-adducted substrates is 12 nucleotides for the UvrABC nuclease, compared with a distance of 9 and 17 nucleotides for the human endonucleases at sites of furan and pyrone side adducts, respectively. Also, in the present study, incisions at the pyrone side TMP cross-link were not produced as efficiently as those at the furan side cross-link. This result is analogous to those from the UvrABC system in which the pyrone side cross-linked substrate was not found to be efficiently incised in one study (14) and was not found to be incised at all in another (11). Studies indicated that UvrABC incision on the pyrone side of the cross-linked substrate may depend upon whether or not a three-stranded intermediate is present (16,17). These incisions have also been shown to be sequence-dependent (14,15). It is possible that a similar situation may play a role in the incision of a pyrone side cross-linked substrate by the human repair endonuclease(s).
The present data show that incisions both 3Ј and 5Ј to a TMP adduct are produced on the 5Ј-labeled DNA substrate. This means that the 3Ј incision occurred in the absence of or independent of the 5Ј incision; otherwise, the 3Ј incision would not have been detected. This indicates that the 3Ј and 5Ј incision events are uncoupled. The 5Ј incision, in turn, could represent just a 5Ј incision event or the generation of a 5Ј incision accompanied by a 3Ј incision. Uncoupling of dual incisions at sites of DNA adducts has also been observed for the UvrABC nuclease at sites of pyrimidine dimers (40), psoralen adducts (12,15), and CC-1065 adducts (41).
The question arises whether the same or different proteins are involved in repair of the various different types of lesions that can be produced in DNA. The chromatin-associated protein extracts utilized in the present study contain an endonuclease complex, pI 4.6, which has specificity for DNA interstrand cross-links, and another, pI 7.6, which recognizes psoralen monoadducts (22,23,25,42). A damage recognition protein with specificity for interstrand cross-links, which we have identified in these extracts, is thought to be a component of the DNA endonuclease complex, pI 4.6 (4). These results are consistent with the hypothesis that different endonuclease complexes are involved in repair of psoralen monoadducts and interstrand cross-links (23,25,42). It is, however, possible that the endonucleases involved in the production of the 3Ј and 5Ј incisions are the same in each complex but that each complex varies in the type of damage recognition protein present. On the other hand, different endonucleases could be associated with these complexes. The system described here, which specifically examines the initial incision events, combined with the use of human cell lines deficient in specific aspects of the damage recognition and incision step of the repair process, should greatly facilitate evaluation of this multifaceted and interactive process and of the proteins involved in these interactions.