Characterization of DNA Recognition by the Human UV-damaged DNA-binding Protein*

The UV-damaged DNA-binding (UV-DDB) protein is the major factor that binds DNA containing damage caused by UV radiation in mammalian cells. We have investigated the DNA recognition by this protein in vitro , using synthetic oligonucleotide duplexes and the protein purified from a HeLa cell extract. When a 32 P-labeled 30-mer duplex containing the (6-4) photoproduct at a single site was used as a probe, only a single complex was detected in an electrophoretic mobility shift assay. It was demonstrated by Western blotting that both of the subunits (p48 and p127) were present in this complex. Electrophoretic mobility shift assays using various duplexes showed that the UV-DDB protein formed a specific, high affinity complex with the duplex containing an abasic site analog, in addition to the (6-4) photoproduct. By circular permutation analyses, these DNA duplexes were found to be bent at angles of 54° and 57° in the complexes with this protein. From the previously reported NMR studies and the fluorescence resonance energy transfer experiments in the present study, it can be concluded that the UV-DDB protein binds DNA that can be bent easily at the above angle.

The UV-damaged DNA-binding (UV-DDB) protein is the major factor that binds DNA containing damage caused by UV radiation in mammalian cells. We have investigated the DNA recognition by this protein in vitro, using synthetic oligonucleotide duplexes and the protein purified from a HeLa cell extract. When a 32 Plabeled 30-mer duplex containing the (6-4) photoproduct at a single site was used as a probe, only a single complex was detected in an electrophoretic mobility shift assay. It was demonstrated by Western blotting that both of the subunits (p48 and p127) were present in this complex. Electrophoretic mobility shift assays using various duplexes showed that the UV-DDB protein formed a specific, high affinity complex with the duplex containing an abasic site analog, in addition to the (6-4) photoproduct. By circular permutation analyses, these DNA duplexes were found to be bent at angles of 54°and 57°in the complexes with this protein. From the previously reported NMR studies and the fluorescence resonance energy transfer experiments in the present study, it can be concluded that the UV-DDB protein binds DNA that can be bent easily at the above angle.
Irradiation of DNA with ultraviolet (UV) 1 light produces various damaged bases (1,2), which induce mutations that can cause cellular transformation and cell death. Among the photolesions, cyclobutane pyrimidine dimers (CPDs) and pyrimidine pyrimidone photoproducts ((6-4) photoproducts), formed at dipyrimidine sites in DNA, have been studied most intensively (1). In eukaryotic cells, damaged DNA is repaired by multiple mechanisms, and the photoproducts described above, especially the (6-4) photoproducts, are repaired by the nucleotide excision repair (NER) pathway (3,4). The mechanism of human NER has been investigated in relation with the inherited syndrome, xeroderma pigmentosum (XP). XP is caused by a defect in NER and is categorized into seven complementation groups, XPA through XPG, and a variant. In the proposed model, the XPA protein, which is absent in XPA patients, forms a complex with replication protein A and rec-ognizes the DNA lesion at the first step of NER (5,6). It was shown recently that the XPC⅐HR23B complex recognizes the damage prior to the XPA step in the global genome NER pathway (7). Other DNA-binding proteins that recognize damaged DNA include replication protein A (8) and the UVdamaged DNA-binding (UV-DDB) protein, which is related to XPE (9,10).
In terms of DNA recognition by proteins, we have focused on the UV-DDB protein, which is the major factor that binds UV-damaged DNA in human cell extracts. The human UV-DDB activity was originally isolated from placenta, as a protein binding to UV-irradiated DNA (11,12). Recently, the human UV-DDB protein was purified from HeLa cells (13,14), and it was found that this protein consisted of two subunits, with molecular masses of 124 and 41 kDa (14). Subsequent sequence determination of the isolated cDNAs revealed that the actual molecular weights were 1.27 ϫ 10 5 and 4.8 ϫ 10 4 (15). The  photoproduct was regarded as the preferential binding site for the UV-DDB protein (10,16). A kinetic analysis using DNA duplexes, each containing a photoproduct at a single site, revealed that this protein bound to the (6-4) photoproduct with an association constant of 1.6 ϫ 10 10 M -1 , whereas its affinity for the cis-syn CPD was reported to be equal to or marginally above that for the undamaged thymines (17).
Many studies have aimed toward the elucidation of the biological function of the UV-DDB protein. It was shown that the UV-DDB activity was absent in XPE cells (9), although further analyses revealed that a defect in this activity was not a general feature of all XPE patients (18,19). Therefore, the relationship of the UV-DDB protein to NER has been investigated intensively. While it was reported that an injection of the purified UV-DDB protein into XPE cells lacking the UV-DDB activity stimulated DNA repair to normal levels (20), this protein was not essential for the reconstitution of NER with purified proteins (21), and another group described that the in vitro excision activity of cell-free extracts from XPE cells was not restored by the addition of the purified UV-DDB protein (22). It was suggested recently that this protein has a role in the repair of DNA in chromatin (23). A more recent model suggests that the UV-DDB protein recognizes the CPD at the initial step of global genomic repair, and then is replaced by the XPC⅐HR23B complex (24). Besides its relationship to NER, several proteins with the ability to interact with the UV-DDB protein have been identified (25,26). The function of each subunit has also been investigated. The 127-kDa subunit was reported to bind the UV-damaged DNA specifically (27), and it was proposed that p48, which is mutated in XPE cells (28), activates p127 by a hit-and-run mechanism (29). Since the expression of the p48 gene was found to be dependent on the p53 tumor suppressor, it is proposed that this subunit acts as a link between p53 and NER (24).
We focused our study on the DNA recognition by the UV-DDB protein. To simplify the interpretation of the experimental results, synthetic oligonucleotide duplexes were used in this study, with the protein purified from a HeLa cell extract, and a single complex was obtained in each assay. We report here that a ternary complex between p127, p48, and DNA was detected in our in vitro assays, and propose that a kink or a kinkable nature of the damaged DNA is recognized by this protein.
Preparation of a Column for Affinity Chromatography-A (6-4) 49mer, d(AGCTACCATGCCTGCACGAAT(6-4)TAAGCAATTCGTAATC-ATGGTCATAGCT), and its complementary oligonucleotide with biotin at the 5Ј-end, biotin-d(TTTTTTTTTTTAGCTATGACCATGATTACGA-ATTGCTTAATTCGTGCAGGCATGGTAGCT), were synthesized on a 1.0-mol scale, and purified by reversed-phase HPLC. The (6-4) 49-mer (about 15 nmol) and the biotinylated 60-mer (about 10 nmol) were mixed in water (90 l), and the solution was heated at 95°C for 5 min and cooled to room temperature. This solution was diluted with a buffer (410 l) containing 20 mM potassium phosphate (pH 7.5) and 0.5 M NaCl, and was added to UltraAvidin-agarose (1.8 ml; Leinco Technologies), which was prewashed and suspended in the same buffer, and the suspension was kept on ice. Binding of the biotinylated duplex to the avidin-agarose was monitored by measuring the absorbance of the supernatant on a spectrophotometer. After 90 min, the damaged DNAcontaining agarose was collected by centrifugation and packed into an Econo-Column (0.5 ϫ 10 cm; Bio-Rad). This column was equilibrated with buffer containing 20 mM potassium phosphate (pH 7.5), 0.1 M KCl, 1 mM EDTA, 1 mM DTT, 0.25 mM phenylmethylsulfonyl fluoride, 0.01% Triton X-100, and 10% glycerol before sample loading.
Protein Purification-A HeLa cell extract from a 30-liter culture, prepared as described previously (32), was loaded on a column (2.5 ϫ 20 cm) of cellulose phosphate (Whatman P11) equilibrated with an elution buffer (20 mM potassium phosphate (pH 7.5), 1 mM EDTA, 1 mM DTT, 0.25 mM phenylmethylsulfonyl fluoride, 0.01% Triton X-100, and 10% glycerol) containing 0.1 M KCl. Proteins were eluted by stepwise increases of the KCl concentrations (0.1, 0.45, and 1.0 M) in the above elution buffer. More than 3 column volumes of the eluant were used at each KCl concentration, and protein binding to the (6-4) photoproductcontaining probe was analyzed by the electrophoretic mobility shift assay (EMSA) described below. The fractions at 0.45 M KCl, which contained the target protein, were combined and dialyzed against the elution buffer containing 0.1 M KCl. The protein solution was divided into halves at this step. One-half was loaded on a column (0.7 ϫ 10 cm) of double-stranded DNA-cellulose (Sigma) equilibrated with the elution buffer containing 0.1 M KCl, which was eluted in steps (0.3, 1.0, and 2.0 M KCl). The fractions at 1.0 M KCl, containing the target protein, were combined, and after dialysis against the elution buffer containing 0.1 M KCl, the solution was loaded on the affinity chromatography column described above. Elution was performed with 0.3, 1.0, and 2.0 M KCl in the elution buffer, and the target protein was found in the fractions at 2.0 M KCl. The remaining half was purified in the same way.
Protein Sequencing-After desalting by dialysis, the purified protein sample was treated with lysyl endopeptidase (0.5 g; Wako Pure Chemical Industries) at 37°C for 15 h in a 50 mM Tris-HCl buffer (pH 8.5) containing 4 M urea. The resultant peptides were separated by HPLC, using a Bakerbond Octyl column (4.6 ϫ 250 mm; J. T. Baker) with a linear gradient of acetonitrile (0 to 60%) in 0.1% trifluoroacetic acid. After lyophilization, the sequence of each peptide was determined on a Shimadzu PPSQ-10 protein sequencer.
Electrophoretic Mobility Shift Assays-The modified strands (20 pmol) were mixed with [␥-32 P]ATP (560 kBq, about 3 pmol; Amersham Pharmacia Biotech) and T4 polynucleotide kinase (6 units; Takara Shuzo) in a buffer (25 l) containing 50 mM Tris-HCl (pH 8.0), 10 mM MgCl 2 , and 5 mM DTT, and the mixtures were incubated at 37°C for 80 min. After heating at 70°C for 10 min, the oligonucleotides were precipitated with ethanol. Double-stranded DNA probes were prepared by mixing aliquots of the 32 P-labeled strands with a small excess of each complementary strand, followed by heating the mixtures at 95°C for 2 min and cooling them to room temperature. The mixtures for binding (10 l) included 10 nM 32 P-labeled probe, 0.1 mg/ml poly(dI)⅐poly(dC), and the purified UV-DDB protein in 13.6 mM Hepes-KOH (pH 7.9), 0.14 mM EDTA, 3.4 mM MgCl 2 , 0.14 mg/ml bovine serum albumin, 10 mM DTT, and 3% glycerol. After an incubation at 37°C for 20 min, a dye solution (1 l) containing xylene cyanol and bromphenol blue in 50% glycerol was added, and the mixtures were loaded onto a nondenaturing 5% polyacrylamide gel (acrylamide:N,NЈ-methylenebisacrylamide ϭ 19: 1). The gel size was 15 ϫ 15 cm, and a buffer containing 50 mM Trizma (Tris base), 0.4 M glycine, and 2 mM EDTA was used. Electrophoresis was performed at 20 mA, and the bands were detected on either a BAS-2500 or a BAS-5000 bioimaging analyzer (Fuji Photo Film) and by autoradiography.
Western Blotting-Antibodies against p127 and p48 were prepared using peptides, LDITPLGDSNGLSPLC (positions 540 -555) and KRSR-SPLELEPEAKKLC (positions 22-38), respectively. After separation by SDS-PAGE or in the same way as the EMSA, the subunits and the protein-DNA complex were transferred to a polyvinylidene difluoride membrane, and probed with each antibody.
Fluorescence Resonance Energy Transfer (FRET)-The duplexes shown in Fig. 5A were designed using the published structure (33) and a B-form DNA generated with Insight II version 97.0 (Molecular Simulations) on a Silicon Graphics O 2 workstation. These duplexes (0.22 nmol) were dissolved in a buffer (2.2 ml) containing 100 mM NaCl, 10 mM sodium phosphate (pH 7.5), and 0.1 mM EDTA, and their steadystate fluorescence emission spectra, excited at 490 nm, were measured at 10°C on a Hitachi F-4500 fluorescence spectrophotometer. The distance between the two dyes (R) was calculated using the equation, E ϭ where E is the energy transfer efficiency, and I DA and I D are the emission intensity in the presence and the absence of the acceptor, respectively (34). The critical Förster distance (R 0 ) was calculated to be 53.3 Å for this duplex (34).
Circular Permutation Analysis-A 277-bp DNA was prepared by ligating 14 oligonucleotides with chain lengths between 30 and 47. Only the 30-mers containing a damaged site were phosphorylated using [␥-32 P]ATP. Typically, four 5Ј-phosphorylated oligonucleotides were mixed in a buffer (400 l) containing 50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 10 mM DTT, 1 mM ATP, and 25 g/ml bovine serum albumin, and treated with T4 DNA ligase (2,000 units) at 16°C for 16 h. After an aliquot was analyzed by nondenaturing 5% PAGE, two other oligonucleotides and the ligase were added, and the mixture was incubated again. The DNA fragments were purified by nondenaturing PAGE, eluted out of the gel, and precipitated with ethanol. The 277-bp DNA was cleaved with restriction endonucleases, as shown in Fig. 6, and the resultant 157-bp fragments were used for EMSA at a concentration of 25 nM. The bend angle (␣) was calculated using the equation, M / E ϭ cos(␣/2), where M and E represent the mobility of the protein-DNA complex with the damage at the middle and at the end of the DNA fragment, respectively (35).

Purification and Identification of UV-DDB Protein-
In another study, 2 we assayed HeLa cell extracts to find proteins that bind DNA containing the (6-4) photoproduct, using synthetic oligonucleotide duplexes, and when (6-4) 30 (Table I) was used, a single specific complex was detected. The protein forming this complex was purified first in this study. A whole cell extract was prepared as described (32), and the first purification was performed by cation-exchange chromatography using a column of cellulose phosphate. The proteins were eluted by stepwise increases of the KCl concentration in the buffer, and the fractions containing the target protein were determined by the EMSA. A band was detected in the fractions at 0.45 M KCl. The second purification was performed on a column of DNAcellulose. After stepwise elution, the target protein was detected by the EMSA in the fractions at 1.0 M KCl. For further purification, we tried to exploit the high affinity of the target protein for the (6-4) photoproduct, and a synthetic oligonucleotide duplex containing this photoproduct was linked to a matrix for chromatography, using the avidin-biotin interaction. By stepwise elutions of this column with increasing KCl concentrations, the target protein was detected by EMSA in the fractions at 2.0 M KCl. Analysis by SDS-PAGE revealed that the purified sample contained three polypeptides, one of which gave only a weak band (Fig. 1A). The amount of the protein obtained from the HeLa cell extract was quantified as 0.1 mg by the dye-binding method using Coomassie Brilliant Blue G-250.
This polypeptide mixture was degraded with lysyl endopeptidase, and the resultant peptides were separated by reversedphase HPLC. The sequences of 13 peptides of up to 19 amino acids were determined, and it was found that all of them were identical with the partial sequences of two subunits (127 and 48 kDa) of the human UV-DDB protein, which was reported previously (15). Since no other sequence was obtained, the third polypeptide, with an estimated molecular mass of 85,000 by SDS-PAGE, was regarded as a degradation product from the 127-kDa subunit. This 85-kDa impurity could not be removed by either gel filtration or glycerol gradient centrifugation. Antibodies against p127 and p48 were prepared using peptides derived from each sequence, and as shown in Fig. 1B, the two subunits were detected specifically with these antibodies.
The EMSA was carried out using the purified UV-DDB pro-tein, and a single complex was detected with the (6-4) 30 probe (Fig. 1C, lane 1). To determine the subunit composition of this complex, a Western blotting analysis was carried out using the antibodies described above, after the separation of the protein-DNA complex on a gel in the same way as the EMSA. It was demonstrated, as shown in Fig. 1C, that the complex obtained in the EMSA contained both subunits of the UV-DDB protein.
Specificity of UV-DDB Protein-damaged DNA Interaction-In order to gain insight into the interactions between the UV-DDB protein and the damaged DNA, the substrate specificity of this protein was determined using various damaged or modified duplexes. As shown in Table I, the probes used in this study included oligonucleotide duplexes containing a cyclobutane pyrimidine dimer (CPD 30), a cis-diamminedichloroplatinum(II) adduct (cisplatin 30), 3-hydroxy-2-(hydroxymethyl)tetrahydrofuran as a stable abasic site analog (abasic 30), 7,8- The sequence was d(CTCGTCACCTCTGGTCTCCTACAGT-CAGTG).
b The damaged strand was separated between X 1 and X 2 . The 5Ј-end of the left 13-mer was linked to the 3Ј-end of the complementary strand with a TTTT loop, and the right 17-mer was phosphorylated with [␥-32 P]ATP.

FIG. 1. Characterization of the purified UV-DDB protein. A,
SDS-PAGE analysis of the purified protein on a 10% polyacrylamide gel. B, Western blotting analysis of the subunits. After separation by 7.5% SDS-PAGE, the polypeptides were transferred to a polyvinylidene difluoride membrane, and were probed with the antibodies prepared with the peptides derived from the sequences of p127 and p48. C, electrophoretic mobility shift assay and Western blotting analysis. A 32 P-labeled 30-mer oligonucleotide duplex containing the (6-4) photoproduct at a single site was mixed with the purified UV-DDB protein, and the protein-DNA complex was separated on a nondenaturing 5% polyacrylamide gel. Lane 1, the complex was detected by autoradiography, and lanes 2 and 5, the subunits in the complex were analyzed using the antibodies specific to p127 and p48, respectively. Both subunits were detected in the complex. dihydro-8-oxoguanine (8-oxo-G 30), O 6 -methylguanine (O 6 -Me-G 30), a T⅐C or T⅐T mismatch (T⅐C-, T⅐T-mismatch 30), and a single-strand break (nick 30), as well as (6-4) 30. All of the oligonucleotides were synthesized chemically, except that DNA ligase was used to prepare cisplatin 30. The EMSA was carried out using these probes, and the results are shown in Fig. 2. The probes containing the (6-4) photoproduct and the abasic site analog gave a thick retarded band, showing that these duplexes formed a stable complex with the UV-DDB protein. Bands were also found in the lanes with the CPD and cisplatin probes, but they were relatively weak. The other duplexes did not exhibit obvious complex formation with this protein. EMSA experiments, using single-stranded oligonucleotides as probes, confirmed that the complexes were formed only with doublestranded DNA (data not shown). The specificity was tested by competition experiments, as shown in Fig. 3. In these EMSA experiments, unlabeled duplexes containing the (6-4) photoproduct, the CPD, and the abasic site analog were used as competitors against each 32 P-labeled probe, and the concentration-dependent decrease in the radioactivity of the bound probe was compared with those using unmodified 30-bp duplexes. In all cases, competition was observed only for the damage-containing duplexes. These results demonstrate that the complex formation between these duplexes and the UV-DDB protein is specific to the damage in the DNA. Finally, the affinity of the duplex containing the abasic site analog was compared with those of the (6-4) photoproduct-and CPD-containing duplexes, whose K a values were reported previously (17). Competition EMSA experiments were carried out against the 32 P-labeled (6-4) 30 probe, and as shown in Fig. 4, the abasic 30-mer duplex competed with (6-4) 30 less and more efficiently than the  and CPD duplexes, respectively. The results revealed that the affinity for the UV-DDB protein is in the order of (6-4) Ͼ abasic Ͼ CPD. This order can also be estimated from the thickness of the bands in Fig. 2.
Structure of Abasic Probe-From the results of the substrate specificity, it was assumed that the common feature of the DNA recognized by the UV-DDB protein was a kink, i.e. a sharp bend usually accompanied by a gap between the helix axes, induced by the damage. The solution structure of a DNA duplex con-taining the (6-4) photoproduct was determined by NMR, and an overall helix bending of 44°was reported previously (36). This conformation is supported by the structure of the dinucleoside monophosphate of the (6-4) photoproduct, which is distorted from that of thymidylyl (3Ј-5Ј)thymidine (37). In the abasic case, the NMR structures of duplexes containing the same analog (33) and the true abasic site (38,39) were reported. The former duplex was kinked by an angle of about 30°, but a hydrogen bonding interaction between the unpaired thymine and the cytosine residue 5Ј adjacent to the abasic site was found in a residue-specific manner. Since this interaction might have caused a kink, the structure of our abasic probe was analyzed by FRET, which is widely used to obtain structural information of nucleic acids (34,40,41).
The duplexes for the FRET measurements are shown in Fig.  5A. To avoid any hypothesis on the location of the dyes, fluo-  Table I, and after incubation, the mixtures were subjected to nondenaturing 5% PAGE.

FIG. 3. Specificity of the UV-DDB protein for damaged DNA.
Competition experiments were carried out against the 32 P-labeled (6-4) 30, CPD 30, and abasic 30 (lanes 1-7, 8 -14, and 15-21, respectively) using damaged (lanes 2-4, 9 -11, and 16 -18) and undamaged (lanes 5-7, 12-14, and 19 -21) 30-mer duplexes as competitors. The concentration of each 32 P-labeled probe was 9 nM, and those of the competitors were 180, 450, and 900 nM (from left to right). After an incubation with the purified UV-DDB protein, the complex was separated from the free probe by nondenaturing 5% PAGE.  8 -10). The concentration of the 32 P-labeled probe was 10 nM, and those of the competitors were 200, 500, and 1000 nM (from left to right). rescein and tetramethylrhodamine were attached to the C-5 position of thymine through a linker with 12 atoms. The sequence of abasic 30 was slightly changed to incorporate the thymidylate with the amino linker, and two extra base pairs were added to restrict the flexibility of the linker in the major groove of the duplex. In these duplexes, the distance between the two dyes can be calculated by fixing the linkers on a line perpendicular to the helix axis. Using the structures of the abasic site analog-containing duplex (33) and a B-form DNA, the sites for the attachment of the dyes were chosen so that the two dyes were arranged on the same side of the helix. The distances between the tetramethylrhodamine and the fluorescein in FlRh-A 32 and FlRh-abasic 32 would be 78 and 56 Å, respectively, if the abasic 32-mer duplex is kinked as reported.
Steady-state fluorescence emission spectra of these duplexes, excited at 490 nm, are shown in Fig. 5, B and C. No difference was observed between FlRh-abasic 32 and FlRh-A 32, and in both cases, the distance between the two dyes was calculated to be 73 Å from the FRET efficiency of 13%, using the equation and the R 0 value described under "Experimental Procedures." The obtained distance is close to that measured for a canonical B-form DNA, as described above, and this result indicates that the abasic duplex we used was not kinked itself.
DNA Bending in Specific Complexes-The DNA structures after the complex formation with the UV-DDB protein were analyzed using the circular permutation gel retardation assays (42). The probes used for these assays are usually prepared using recombinant plasmids, but in this study, the specific damages could not be cloned in a vector. Therefore, synthetic oligonucleotides, one of which contained the (6-4) photoproduct or the abasic site analog at a single site, were ligated to make a 277-bp DNA with a sequence derived from a DNA bending vector, pBend2 (43). Circularly permuted fragments with a length of 157-bp were prepared by digesting this DNA with the MluI, NheI, SpeI, EcoRV, StuI, RsaI, and BamHI restriction endonucleases, as shown in Fig. 6.
Electrophoretic mobility shift data of the complexes with these probes are shown in Fig. 7. In both cases of the (6-4) photoproduct and the abasic site analog, the mobility of the protein-bound fractions exhibited clear anomalies. The mobility was larger when the damage was located near the end of the fragment and smaller when it was near the center. Although the mobility anomalies found in the circular permutation assays were attributed to "distortions," and were distinguished from bends in some cases (44), it is reasonable to conclude that the anomalies observed in this study were caused by the bending of DNA in the complex, because the duplex containing the (6-4) photoproduct has an intrinsic bend. The ␣ values (bend angles), calculated by the equation described by Thompson and Landy (35), were 54°and 57°for the probes containing the (6-4) photoproduct and the abasic site analog, respectively. The data in Fig. 7 also show that the center of each bend corresponds to the damage site. DISCUSSION In this study, we analyzed the DNA recognition by the UV-DDB protein from a HeLa cell extract. The two subunits, p127 and p48, were co-purified by three steps of chromatographic separation, and a single complex was obtained in the EMSA. Another group reported previously that two complexes were detected using UV-irradiated DNA and the UV-DDB protein (9,13,29), but since we used highly purified synthetic oligonucleotide duplexes, each containing defined damage at a single site, the difference in the number of complexes arose probably because their larger complex was formed by protein binding to the second site of damage in the DNA.
The first novel finding in this study is that the complex detected in our EMSA contained both of the p127 and p48 subunits, as shown by the Western blotting analysis (Fig. 1C). It was shown previously that p127 bound specifically to UVdamaged DNA (27), and this subunit was detected in a protein-DNA complex by immunoblotting (45). Recently, Chu and coworkers (29) described that p48 activates the p127 binding by a hit-and-run mechanism. There is a discrepancy between their report and our finding. In their assays using human cell extracts, the ternary complex between p127, p48, and DNA was detected as a minor band only when the p48 cDNA was expressed, whereas only this ternary complex was formed in our in vitro assays using the purified protein. A plausible explanation that links these results is the possible existence of a factor that releases p48 from the ternary complex in cells. Since p127 and p48 form a heterodimer in solution (14), it is likely that the ternary complex, detected in our assays, is formed at the initial step of damage recognition.
Although the substrate specificity of the UV-DDB protein was originally analyzed by Payne and Chu (46), we reinvestigated it using synthetic oligonucleotides containing damage at a single site, to find a clue to the structural features recognized by this protein. In addition to the (6-4) photoproduct-containing DNA, a duplex containing 3-hydroxy-2-(hydroxymethyl)tetrahydrofuran as an abasic site analog was recognized efficiently by the UV-DDB protein, and those containing a CPD and a cisplatin adduct gave a weak band in the EMSA (Fig. 2). Complex formation was not detected for duplexes containing 8-oxo-G, O 6 -Me-G, a mismatch, and a nick. The tertiary structures of these DNA duplexes have been determined, mainly by NMR spectroscopy. The solution structure of a decamer duplex containing a (6-4) photoproduct in its center was determined by NMR and relaxation matrix refinement, and a helix bending of 44°into the major groove was reported (36). A similar study of an undecamer duplex, containing the same abasic site analog as that used in our present study, revealed that this modification induced a kink of about 30°in the duplex (33). In the NMR structure of a CPD-containing duplex, a relatively small helix distortion was found on the 3Ј side of the dimer (47), and for the intrastrand cisplatin adduct at the d(GpG) sequence, several bend angles, from 35-40°to 78°, were reported in the studies of the NMR solution and x-ray crystal structures (48 -50). On the other hand, no global conformation change due to 8-oxo-G was observed in NMR and circular dichroism studies (51,52). There was no bending in the crystal structures of dodecamer duplexes containing an O 6 -ethylguanine⅐cytosine pair, although they were complexed with minor groove binding drugs (53, 54), and even when O 6 -Me-G was opposite adenine, the overall structure of this mispaired dodecamer duplex was very close to that of the parent duplex (55). NMR studies of duplexes containing a T⅐T mismatch (56) and a nick (57) revealed that these lesions had little influence on the conformation of the normal B-form DNA. These structural features strongly suggest that the UV-DDB protein recognizes a kink caused by the damage.
Before verifying this hypothesis, we had to analyze the structure of our duplex containing the abasic site analog, because a hydrogen bonding interaction between the unpaired thymine and the cytosine residue 5Ј to the abasic site was reported in the NMR structure (33). FRET was used for this purpose, and the duplexes for the fluorescence measurement were designed very carefully, using tertiary structure models. The results (Fig. 5) indicated that our abasic duplex was indistinguishable from the unmodified DNA in terms of the global structure. Since the base 5Ј to the abasic site was thymine in our duplex, the kink observed in the NMR study must be caused by the base-specific interaction. It should be noted that such a kink was not observed for the duplexes containing a true abasic site, in which the unpaired and 5Ј adjacent bases were adenines (38,39). From these observations, it is derived that the abasic duplexes can be kinked very easily, because the formation of a single bifurcated hydrogen bond resulted in a kinked structure.
The DNA structures in the complexes were analyzed by the circular permutation assays. From the results shown in Fig. 7, the DNA fragments in the complexes with the UV-DDB protein were found to be distorted at the damage site, by an ␣ value of 54 -57°. Since the (6-4) photoproduct-containing duplex has an intrinsic bend, the bent DNA structure, but not other alterations caused by the protein binding, accounts for this distortion. From the finding that the abasic site analog-containing DNA, without an intrinsic bend, was bent by the complex formation with the UV-DDB protein, we conclude that a kinkable nature, as well as an intrinsic kink, is recognized by the UV-DDB protein. In contrast to the thick bands observed for FIG. 7. DNA bending in the UV-DDB protein-DNA complexes. A and B, circular permutation analyses on a 5% polyacrylamide gel, using the 157-bp probes containing the (6-4) photoproduct (A) or the abasic site analog (B). The probes are represented by the restriction endonucleases used for their preparation, as shown in Fig. 6. C and D, graphic presentation of the relative mobility of the complexes plotted as a function of the position of the (6-4) photoproduct (C) or the abasic site analog (D) from the 5Ј-end. Each data set was approximated by a second order polynomial equation, and the bend angles were calculated to be 54°and 57°in the (6-4) photoproduct and abasic cases, respectively. the duplexes containing the (6-4) photoproduct and the abasic site analog in our EMSA, the weak recognition of the cisplatin adduct-containing duplex by the UV-DDB protein can be attributed to the larger bend angle of this DNA. The bend angles of the cisplatin-modified DNA, in the cases of both the solution structure determined most recently (49) and the complexes with the high-mobility group domain proteins (58), were much larger than that calculated for the UV-DDB protein-DNA complexes in this study. The duplex containing the CPD was also recognized weakly by this protein. Although the CPD-containing duplex is kinked by the binding of endonuclease V (59), it is apparent that the abasic duplex can be kinked much more easily than the CPD-containing one. Since DNA structures can be distorted by interactions with proteins, the CPD-containing DNA may be kinked more easily and be recognized more efficiently by the UV-DDB protein in the nucleosome.