DDB1-DDB2 (Xeroderma Pigmentosum Group E) Protein Complex Recognizes a Cyclobutane Pyrimidine Dimer, Mismatches, Apurinic/Apyrimidinic Sites, and Compound Lesions in DNA*

The DDB protein complex, comprising the subunits DDB1 and DDB2, binds tightly to UV light-irradiated DNA. Mutations in DDB2 are responsible for xeroderma pigmentosum group E, a disorder with defects in nucleotide excision repair of DNA. Both subunits are also components of a complex involved in ubiquitin-mediated proteolysis. Cellular defects in DDB2 disable repair of the major UV radiation photoproduct in DNA, a cyclobutane pyrimidine dimer, but no significant direct binding of DDB to this photoproduct in DNA has ever been demonstrated. Thus, it has been uncertain how DDB could play a specific role in DNA repair of such damage. We investigated DDB function using highly purified proteins. Co-purified DDB1-DDB2 or DDB reconstituted with individual DDB1 and DDB2 subunits binds to damaged DNA as a ternary complex. We found that DDB can indeed recognize a cyclobutane pyrimidine dimer in DNA with an affinity (Kappa) 6-fold higher than that of nondamaged DNA. The DDB1-DDB2 complex also bound with high specificity to a UV radiation-induced (6-4) photoproduct and to an apurinic site in DNA. Unexpectedly, DDB also bound avidly to DNA containing a 2- or 3-bp mismatch (and does not bind well to DNA containing larger mismatches). These data indicate that DDB does not detect lesions per se. It instead recognizes other structural features of damaged DNA, acting as a sensor that probes DNA for a subset of conformational changes. Lesions recognized may include those arising when translesion polymerases such as POLH incorporate bases across from DNA lesions caused by UV radiation.

The human disorder xeroderma pigmentosum (XP) 2 has been intensely studied because of the striking phenotypes of sunlight sensitivity, predisposition to skin cancer, and an association with defects in DNA repair. Seven of the eight complementation groups of the disorder (designated XP-A through XP-G) have defects in nucleotide excision repair (NER) of damaged DNA (1). An eighth group, XP-V, is defective in DNA POLH (pol ) (2), which can bypass the major photoproduct caused by UV radiation in DNA, the cis-syn cyclobutane pyrimidine dimer (CPD). The clinical and cellular characteristics of XP group E are similar to those of the other XP complementation groups but are generally milder. The overall level of NER as measured by unscheduled DNA synthesis in XP-E cells is 50 -80% of normal cells. The gene defective in the XP-E group is DDB2. The 48-kDa DDB2 protein, together with the 127-kDa DDB1 protein (3)(4)(5), constitute the DNA damage binding factor known as DDB or UV-DDB (6). The DDB subunits are also found in a protein complex that includes components of a cullinbased ubiquitin ligase as well as the COP9 signalosome (7). The function and physiological substrates of this complex are presently unclear (8).
Despite the ability of the DDB protein to bind UV light-irradiated DNA, the specific role of DDB in NER is uncertain. One difficulty in determining the function is that NER can be reconstituted with purified components and damaged naked DNA in the absence of DDB (9 -11). Varying results have been reported regarding the effect of adding DDB to a purified NER system using a DNA lesion with a single CPD, ranging from no stimulation (12) to a fewfold stimulation (13) to 17-fold stimulation (14). There is a clear phenotype of DDB deficiency, however. Most dramatically, the ability to remove the most abundant UV radiation-induced photoproduct (the CPD) from non-transcribed DNA is severely impaired in XP-E cells (15,16). This finding is puzzling because significant binding of DDB to a CPD in DNA has not been detected (12,(17)(18)(19). In fact, some recent models for NER are based on the conclusion that no single protein factor can distinguish between a CPD and undamaged DNA (12,43). By contrast, the less abundant UV lightinduced photoproduct, the (6-4) photoproduct, is well bound by DDB, yet an absence of DDB in cells has little or no effect on the repair of (6-4) photoproducts (6).
To better understand the mechanism of DDB action, we produced highly purified recombinant DDB protein complex to investigate and quantify the preference of DDB for binding to different types of DNA damage. We found that DDB is indeed able to efficiently discriminate a single CPD from nondamaged DNA. Further, we found that DDB can bind avidly to abasic (AP) sites in DNA, to DNA containing mismatches without a covalent lesion, and to "compound" lesions containing both mismatches and lesions. A DDB complex containing amino acid substitutions in the DDB2 subunit, as found in XP-E, patients is greatly deficient in binding to damaged DNA.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-To overproduce tagged DDB1 and DDB2, 5Ј primers containing the tags and an SpeI restriction site were used in PCR reactions together with a 3Ј primer containing a SpeI site. The fragments (NЈ-His 10 -FLAG-TEV-DDB2, NЈ-His 10 -FLAG-TEV-DDB1, and NЈ-His 10 -DDB1) were then cloned into pBlueBac4.5/V5-His (according to the Invitrogen instructions for Topo cloning). All con-structs were sequenced to be sure that no mutations were introduced by PCR. FLAG-tagged DDB2 virus was made as described (20). Single amino acid substitutions K244E and R273H in DDB2 were introduced by the two-stage, PCR-based overlap extension method (21). To facilitate purification of the recombinant mutant DDB complexes, the wild type and mutant DDB2 genes were fused to DNA encoding a His 10 -FLAG-TEV tag at the N terminus. Plasmids for generating the viral stocks were sequenced to verify the mutation before transfection. For R273H, DNA from the large scale virus stock used for protein expression was extracted, and the sequence was re-verified.
Protein Expression and Purification-Recombinant baculoviruses encoding DDB1 and DDB2 were prepared according to the manufacturer's protocol (Invitrogen Bac-N-Blue). Sf9 insect cells were infected with virus encoding each subunit alone or co-infected with two baculoviruses to obtain a DDB complex and incubated for 48 h at 27°C. After centrifugation, the cell pellet was lysed by freezing in liquid nitrogen. Buffer A (300 mM KCl, 50 mM potassium phosphate pH 8.0, 5% glycerol, and 0.1% Nonidet P-40) was added, the pellet was thawed on ice, and extracts were prepared in a Dounce homogenizer with an A-type pestle. After centrifugation for 45 min at 41,000 rpm in a Sorvall T.647.5 rotor, the cleared lysate was adjusted to 10 mM imidazole and applied to a nickel-agarose column (Qiagen) equilibrated with buffer A containing 10 mM imidazole. After washing extensively with buffer A containing 10 mM imidazole, the protein was then eluted with buffer A containing 200 mM imidazole, and peak fractions were applied to a column containing anti-FLAG M2 affinity gel (Sigma) pre-equilibrated with buffer A. The column was washed with 20 volumes of buffer A and subsequently eluted with buffer A containing 100 g/ml FLAG peptide (Sigma). The yield of pure DDB complex was ϳ200 g/liter Sf9 culture. The fractions collected were stored at Ϫ80°C. An EDTA-free protease inhibitor mix was added to all buffers used in the protein purification (Roche Applied Science).
Immunoblotting analysis of the complex was performed on 4 -15% gradient gels (Bio-Rad). Gels were electrophoresed at 40 mA for 1.5 h and transferred to polyvinylidene difluoride membrane (PVDF) overnight at 30 V. The membrane was blocked with 20% dry milk, 1ϫ phosphate-buffered saline, and 0.1% Tween 20 and incubated with primary antibody for 2 h at room temperature in 5% milk, 1ϫ PBS, and 0.1% Tween. Primary antibodies were kindly provided by Dr. Vesna Rapic-Otrin and used as described (22). After washing in 1ϫ phosphate-buffered saline and 0.1% Tween 20, membranes were incubated for 30 min with a secondary horseradish peroxidase antibody, washed, and visualized by using a Femto signal kit (Pierce).
DNA Substrates-TABLE ONE shows a schematic representation of the 30-mer duplex substrates utilized. The sequence of the top strand was 5Ј-CTCGTCAGCATCTTCATCATACAGTCAGG-3Ј (position of UV radiation photoproducts shown in boldface). The sequence of one strand of the 49-mer was 5Ј-AGCTACCATGCCTGCCTGCAC-GAATTAAGCAATTCGTAATCATGGTCATAGCT-3Ј (position of UV radiation photoproducts shown in boldface). Single-stranded DNA and duplex DNA substrates used for binding assays and two-dimensional gel analysis were all 32 P-5Ј-end-labeled with T4 polynucleotide kinase (Invitrogen). Substrates, except for those containing the CPD and the (6-4) photoproduct, were purchased high performance liquid chromatography-purified from Sigma-Genosys and were Ͼ95% pure. Substrates containing CPD and (6-4) photoproduct were made as described (23,24). Annealing of duplexes was done by adding a 2-fold molar excess of the unlabeled strand, heating to 95°C for 5 min, and then cooling down to 20°C. Single-stranded DNA was irradiated (where indicated) with a 254-nm peak, germicidal UV-C light at a dose of 18 kJ/m 2 .
Electrophoretic Mobility Shift Assay-1 ng (2.6 nM concentration) of 32 P-labeled substrate was incubated at 20°C with the indicated amount of recombinant protein for 30 min in a 20-l reaction volume (0.5 mM dithiothreitol, 150 mM NaCl, 20 mM Tris-HCl, pH 7.5, 1 mM MgCl 2 , 0.2 mM EDTA, and 5% glycerol). Time courses showed that all binding reactions were complete within 15 min. The protein-DNA complexes were separated by electrophoresis on a non-denaturing 5% polyacrylamide gel at 175 V for 2 h and visualized by autoradiography. The dried gels were exposed to an imaging screen, and quantification was carried out using a Fuji FLA 3000 Bioimager with ImageQuant software. The graphs shown are each from a single experiment, and all results were verified by repeat experiments.
Two-dimensional Gel Electrophoresis-The DDB binding reactions were performed as described above except that 12 ng (18.6 nM) of probe was incubated with 100 ng (28.6 nM) of DDB in each reaction. The first dimension of the gel was run according to the description above for an electrophoretic mobility shift assay (EMSA). Two identical lanes were run, one for exposure and one for processing in the second dimension. The lane was excised and cast into the stacking region of an SDS-polyacrylamide gel (4% stacking and 8% resolving). A modified gel comb allowed a size marker and a control sample of the DDB complex to be included in the second dimension. The gel was then transferred to a PVDF membrane and analyzed by immunoblotting as described above.

Expression and Purification of the DDB Heterodimer and Single
Subunits-To obtain highly purified DDB protein, the recombinant DDB1-DDB2 heterodimer and the single DDB1 and DDB2 subunits were expressed in insect cells using a baculovirus system. For purification of the single DDB1 and DDB2 subunits, each was fused to a FLAG-His 10 -TEV tag at the N terminus. To facilitate purification of the heterodimeric complex, the N terminus of the DDB2 subunit was fused to a nine-amino acid FLAG tag, and the N terminus of the DDB1 subunit was fused to a 10-amino acid His tag. This ensured that after passing extracts containing the co-expressed heterodimer over both anti-FLAG-M2 affinity gel and nickel-NTA columns, only the DDB1-DDB2 complex was purified. The DDB1-DDB2 complex was purified 1550fold from whole cell lysates by this two-step scheme (Fig. 1A). In addition to the DDB1-DDB2 complex, this procedure yielded highly purified individual DDB1 and DDB2 proteins (Fig. 1B) with a size and subunit composition confirmed by immunoblot analysis (Fig. 1C). In preparations of the DDB2 subunit alone, a substantial amount of DDB2 was insoluble, and a polypeptide of ϳ 60 kDa was reproducibly present that did not react with DDB2 antibody. This polypeptide could not be separated from DDB2 even in buffers containing 1 M NaCl or 40% ethylene glycol (data not shown). This 60-kDa peptide derived from insect cells is likely a tightly bound chaperone that solubilizes DDB2 in the absence of DDB1.
DDB1 and DDB2 Associate to Form DDB Unaided by Other Protein Factors and Remain in the Complex following Binding-A simple mechanistic question that has not been investigated directly by utilizing pure protein preparations is whether both the DDB1 and DDB2 subunits are required for DNA binding. To investigate this, a 30-bp duplex containing a centrally placed (6-4) photoproduct (TABLE ONE) was used in an EMSA. Binding activity was reconstituted only when the two subunits were mixed together (Fig. 1D, lane 4) and not with either of the single subunits. The position of the mobility shift was identical to that of DDB complex where the two subunits were co-purified (Fig. 1D, lane 5) and identical to that obtained with HeLa cell extract (not shown). Both subunits are therefore necessary and sufficient for binding to the DNA substrate.
Two or more bands of distinct mobilities appeared in the EMSA, as noted in other studies (18,25). To determine whether all of these bands contained both subunits of DDB protein or if only one subunit was present, two-dimensional gel electrophoresis was used to separate the components of the different DNA-bound complexes (Fig. 1E). A 49-bp DNA duplex with a centrally located (6-4) photoproduct was incubated with DDB1-DDB2 complex and electrophoresed on a 5% non-denaturing gel. Two identical lanes were run. One lane was excised and prepared for autoradiography to record migration in the first dimension and the other was excised and cast into the stacking section of an SDSpolyacrylamide gel. The polypeptides were separated in this second dimension together with a control sample containing the DDB1-DDB2 complex. Proteins were then visualized by immunoblotting and aligned with an autoradiograph of the first dimension gel (Fig. 1E). Each of the shifted complexes (labeled C1 and C2) detected in the first dimension EMSA contained both DDB1 and DDB2 subunits in the same ratio and, consequently, both DDB subunits remained in these complexes after DNA was bound.
DDB Can Recognize a Single CPD or AP Site in DNA-The major phenotype of cells lacking DDB activity is their extremely low capacity to remove CPD from their genomes by NER, with a milder defect in removal of (6-4) photoproducts (13,15,16,26). Paradoxically, experiments have suggested that DDB is unable to discriminate a cis-syn CPD from undamaged DNA (17,18,27,28). The source of DDB in previous studies was either cell extracts (25,28,29) or purified protein preparations (18,27,30). However, competitor DNA such as poly(dI-dC) is in most cases included in the binding reaction mixtures to lower the background arising from the binding of damaged DNA by contaminating proteins. The addition of competitor to binding reactions can potentially mask binding of DDB to some substrates. Few studies have employed single defined lesions, complicating the assessment of affinity of DNA for various types of DNA damage.
To better understand the features in damaged DNA sensed by DDB, we used single lesions and highly purified recombinant DDB complex so that nonspecific competitor DNA could be omitted. Four different 30-mer DNA duplex substrates with centrally located single alterations (TABLE ONE) were tested ( Fig. 2A). Measurements were initially carried out with equimolar amounts of DDB and DNA substrate. Under conditions where 87% of the (6-4) photoproduct substrate was bound, there was 29% binding to the CPD substrate, 89% binding to DNA containing an AP site, and 15% binding to undamaged duplex ( Fig. 2A, lanes  2, 5, 8, and 11). For comparison with results obtained in previous studies, binding to the substrates was also tested in the presence of 0.5 , and 1 g of the FLAG column eluate (FLAGpool) were analyzed on a 4 -15% gradient polyacrylamide-SDS gel followed by silver staining. Sizes of markers are shown (kDa). B, purified recombinant DDB heterodimer (DDB-co), DDB2, and DDB1 were analyzed by electrophoresis on a 4 -15% gradient SDS-polyacrylamide gel followed by silver staining. Positions of the DDB1 and DDB2 are indicated, and the insect chaperone protein co-purifying with DDB2 migrates at ϳ60 kDa. C, immunoblotting analysis of the samples. After separation on a 4 -15% gradient polyacrylamide gel, the polypeptides were transferred to a PVDF membrane that was cut lengthwise at the 63-kDa marker and probed with DDB2 or DDB1 antibody. D, DNA binding activity of DDB subunits. The DNA binding ability was tested on a 32 P-labeled 30-mer oligonucleotide duplex containing a single  photoproduct paired to the AA-containing strand (see . The reaction mixtures were separated on a non-denaturing 5% polyacrylamide gel and analyzed by autoradiography. E, analysis of the DDB-DNA complex. A non-denaturing 5% gel was used in the first dimension to separate bound from unbound complex in a reaction containing 12 ng (18.6 nM) of a 32 P-labeled 49-mer containing a single (6-4) photoproduct (see "Experimental Procedures") and 100 ng (28.6 nM) of purified DDB1-DDB2 complex in the co-expressed form (the portion of the gel containing the unbound probe was removed). The slice excised from this native gel was cast into the stacking of the SDS gel used for the second dimension. A control sample of co-expressed DDB (DDB-co) was loaded before running the second dimension. The resulting gel was transferred to PVDF and probed with DDB2 and DDB1 antibodies. g/reaction poly(dI-dC) (a molar excess of 500-fold over 30-mer). This procedure virtually abolished binding to undamaged DNA while retaining 3% binding to the CPD, 36% to the AP site, and 67% to the (6-4) photoproduct. These data show that DDB can efficiently discriminate between a CPD and non-damaged DNA, but that this discrimination can be obscured by competition with an excess of poly(dI-dC).
Binding of DDB to Base-Base Mismatches-UV radiation photoproducts and AP sites cause loss of base pairing potential, with accompanying distortions in double-stranded DNA. We therefore asked whether DDB might bind to DNA containing only mispaired, undamaged bases. A set of 30-bp DNA duplex substrates was created with 1-5 T residues mismatched to C residues on the opposite strand (TABLE ONE). The experiments were done with a 1:2 ratio of substrate to DDB in the absence or presence of poly(dI-dC).
These mismatched substrates were indeed better bound by DDB than by paired DNA (Fig. 2B). The substrates with two or three base mismatches were bound most avidly (88% for two mismatches and 75% for three mismatches in the absence of competitor), and substrates containing one, four, or five mispairs were bound less well (17-26% in the absence of competitor). Substrates containing a centrally located T loop of varying length creating a bulge were not bound by DDB better than by nondamaged DNA (data not shown). We also tested several different single-stranded substrates in the EMSA assay using a 10-fold molar excess of protein. A poly(dT) 50-mer with minimal self-pairing potential was not bound by DDB, whether unirradiated or irradiated. A 49-mer substrate with a mixed nucleotide content containing either no lesion or a single (6-4) photoproduct or CPD was bound marginally (1-3%), and irradiation of the 49-nucleotide substrate increased binding by 3-fold (to 8%). It is possible that photoproduct formation promotes limited internal pairing in the substrate that is bound by DDB. We conclude that the binding of DDB to substrates with reduced base pairing potential such as mismatches, AP sites, and (6-4) photoproducts cannot be explained by recognition of single-stranded DNA character.
Comparative Affinity of DDB for Different Lesions-To obtain more detailed information on the affinity of DDB for various substrates, titration experiments were performed in the absence of competitor. Maximum binding close to 100% could be obtained for all of the substrates (Fig. 2C and data not shown). Apparent equilibrium affinity constants (K a app ) for those substrates were estimated by determining the DDB concentration at which 50% of the substrate was bound. The affinity for the CPD photolesion was 1 ϫ 10 8 M Ϫ1 , 6-fold higher than the affinity for undamaged DNA (1.8 ϫ 10 7 M Ϫ1 ). The estimates for binding affinity to the (6-4) photoproduct (1.5 ϫ 10 9 M Ϫ1 ), the AP site (8.3 ϫ 10 8 M Ϫ1 ), and the two mismatch (9.1 ϫ 10 8 M Ϫ1 ) constitute minimal affinities, as binding was so avid that there is not a vast molar excess of protein at the DDB concentration at which 50% of the substrate was bound. Competition experiments were also performed by incubating the different substrates with a 10-fold molar excess of DDB and the indicated amounts of poly(dI-dC) (Fig. 2D). The results confirm the order of affinities for the different substrates established above and show that addition of poly(dI-dC) eliminates the binding of DDB to both undamaged DNA and to DNA containing a CPD.
DDB Does Not Detect DNA lesions per se but Recognizes Distortions in Duplex DNA-Given the above results, we next asked whether DDB has an increased affinity for DNA containing both a photoproduct and a mismatch (Fig. 3). Several substrates with such "compound lesions" were tested (TABLE ONE). The potential for base pairing across from the CPD was eliminated by placing an AC or CC sequence across from the lesion instead of the normal AA. This increased the binding of DDB from 26% for the "paired" CPD:AA to 88% for the CPD:CC (Fig. 3). The latter value was similar to that obtained for the paired (6-4) photoproduct (6-4:AA); introducing mispairs opposite the (6-4) photoproduct did not give significantly higher binding. As these values were obtained at near saturation, titration experiments were performed for the double mismatched substrates (data not shown; as for other substrates, maximum binding was very close to 100%). A minimum estimate of the apparent equilibrium affinity constant K a app for the CPD:CC substrate was 1.1 ϫ 10 9 M Ϫ1 , a 10-fold higher value than for the paired CPD:AA substrate (1.0 ϫ 10 8 M Ϫ1 ), demonstrating a marked impact of abrogating the base pairing across from the CPD. In contrast the K a app for the 6-4:CC substrate (2.5 ϫ 10 9 M Ϫ1 ) was only slightly increased over the paired 6-4:AA substrate (1.5 ϫ 10 9 M Ϫ1 ). Experiments were also performed to compare the relative binding of DDB to all substrates by competition with a substrate containing two mismatches and no photolesion. The relative affinity of binding of DDB to substrates was in good correspondence with the order obtained from the estimated values for K a app (data not shown). These data show that the CPD structure itself presents no obstacle to high affinity binding of DDB. Further, the results indicate that the affinity of DDB for DNA is not determined by the structure of a particular lesion but more likely involves recognition of distortion in the DNA helix caused by some lesions.

Substrates Primers
UD a 5Ј-CTCGTCAGCATCTTCATCATACAGTCAGTG-3Ј UD a 3Ј-GAGCAGTCGTAGAAGTAGTATGTCAGTCAC-5Ј AP site 5Ј-CTCGTCAGCATC-TCATCATACAGTCAGTG-3Ј AP site 3Ј-GAGCAGTCGTAGAAGTAGTATGTCAGTCAC-5Ј CPD:AA 5Ј-CTCGTCAGCATCTϽϾTCATCATACAGTCAGTG-3Ј CPD:AA 3Ј-GAGCAGTCGTAG AA GTAGTATGTCAGTCAC-5Ј CPD:AC 5Ј-CTCGTCAGCATCTϽϾTCATCATACAGTCAGTG-3Ј CPD:AC 3Ј-GAGCAGTCGTAG AC GTAGTATGTCAGTCAC-5Ј CPD:CC 5Ј-CTCGTCAGCATCTϽϾTCATCATACAGTCAGTG-3Ј CPD:CC 3Ј-GAGCAGTCGTAG CC GTAGTATGTCAGTCAC-5Ј 6-4: highly conserved residues within WD40 repeats in DDB2, K244E, and R273H (32) (Fig. 4A). To gain insight into the role of protein domains in DDB1-DDB2 complex formation and DNA binding, we studied DDB2 mutant proteins with these single amino acid substitutions. Like all other XP-E cell lines, extracts from the corresponding cell lines do not show DDB activity as measured by EMSA with UV light-irradiated DNA (25). However, immunoprecipitation of endogenous DDB from the cell lines indicates a difference between the two mutant proteins (20). DDB2 K244E is present in normal amounts in XP82TO cells and able to interact with DDB1. By contrast, DDB2 R273H is detected in only trace amounts in XP2RO and XP3RO cells and does not co-immunoprecipitate with DDB1 (20). No previous studies reported purification of mutant DDB2-DDB1 complexes, and so we decided to investigate these mutant complexes in recombinant form. The mutant DDB2 proteins were co-expressed with DDB1 in the baculovirus/insect cell system and purified by a two-step procedure on nickel-nitrilotriacetic acid and anti-FLAG-M2 affinity gels (Fig. 4, B and  C). The additional chaperone polypeptide of ϳ 60 kDa was present in all three recombinant protein preparations. Because in this case the DDB complex was purified only by virtue of tags on the DDB2 protein, DDB2 was in excess over DDB, and the excess DDB2 was likely associated with the chaperone. When an untagged version of DDB1 was taken through the same purification steps, no DDB1 remained as determined by immunoblotting, showing that DDB1 alone did not bind to the purification resins utilized (data not shown).
Binding of normal DDB and the mutant DDB2-K244E/DDB1 and DDB2-R273H/DDB1 complexes to 30-bp DNA duplexes with different centrally located lesions was tested by EMSA using equimolar amounts of protein and DNA (Fig. 4D). Both mutant complexes had only residual binding activity as compared with the wild-type protein, and both had the highest affinity (10% for the K244E complex and 3% for R273H complex) for the compound 6-4:CC substrate.

DDB Recognizes Helical Distortion rather than Specific DNA Lesions-
XP-E cells have a prominent defect in removal of CPDs from DNA (13,15,16,26). However, a conundrum has arisen because in vitro data have indicated that DDB binds a CPD with an affinity only marginally or no higher than that of undamaged DNA (18,27,28). We found here that a CPD is in fact recognized by DDB with a K a app 6-fold higher than that for undamaged DNA. It is worth noting that the absolute affinity of DDB for CPD lesions is not particularly low on a nanomolar basis; it is just lower than that found for a (6-4) photoproduct. DDB has a high affinity for DNA. Indeed, its affinity  . DDB containing mutant DDB2. A, protein sequence alignment of two adjacent WD40 repeats from DDB2 in vertebrates. Mutations leading to K244E and R273H changes, found in XP-E, are in highly conserved residues of these repeats. The corresponding normal residues are shown in boldface with asterisks. Other residues very highly conserved between both repeats are indicated in boldface only. Sequences were from Homo sapiens (human), Mus musculus (mouse), Rattus norwegicus (rat), Bos taurus (cow), Canis domesticus (dog), Xenopus laevis (frog), Gallus gallus (chicken), and Tetraodon nigroviridis (puffer fish). B, purified recombinant wild-type (DDB2-His-FLAG)-DDB1 (WT), K244E (DDB2 His-FLAG)-DDB1, R273H (DDB2-His-FLAG)-DDB1, and wild type (DDB2-FLAG)-(DDB1-His) (DDB-co) were analyzed by electrophoresis on a 4 -15% gradient SDS-polyacrylamide gel followed by silver staining. Sizes of markers are shown (kDa). C, immunoblotting analysis of the samples. After separation on a 4 -15% gradient polyacrylamide gel, the polypeptides were transferred to a PVDF membrane and probed with antibodies (see "Experimental Procedures"). The amount of DDB complex was quantified by immunoblotting of DDB1 in each protein preparation. Sizes of markers are shown (kDa). WT, wild-type; DDB-co, coimmunoprecipitated DDB. D, DNA binding activity of wild-type (WT) and mutant DDB complexes. Mutant and wild-type DDBs were incubated with a 32 P-labeled 30-mer oligonucleotide duplex containing different lesions at a single position (the type of lesion is indicated above each lane; see also TABLE ONE). The reaction mixtures were separate on a non-denaturing 5% polyacrylamide gel and analyzed by autoradiography. Percentage of bound substrate is indicated underneath the gel. UD, undamaged DNA, (6-4)PP, (6-4) photoproduct; 2MM, two mismatched substrates.
for undamaged DNA is close to that estimated for the distortion recognition factor XPC-RAD23B on UV light-irradiated DNA (17). An affinity of DDB for AP sites has been suggested previously (28,30) but never quantified, and studies have always included excess poly(dI-dC) competitor DNA. We find that substrates containing an AP site were bound with ϳ80% of the affinity of DNA containing a (6-4) photoproduct.
What structural features of a DNA lesion favor recognition by the DDB1-DDB2 complex? Possibilities to consider include disruption of base pairing interactions, the extent of intrinsic DNA bending, and local flexibility of DNA containing a lesion. Regarding base pairing interactions, structural data indicate significantly reduced complementary hydrogen bonding potential across from a (6-4) photoproduct, particularly on the 3Ј side of the lesion. This reduced pairing is much less marked for a CPD (33)(34)(35). The structural effects of increasing the size of a mismatched region are not well understood. However, mismatches do confer helix flexibility, and a substrate with a 3-bp bubble is more flexible than one containing a single bp mismatch (36,37).
The degree of intrinsic DNA bending may also be relevant (30). The (6-4) photoproduct introduces a bend of 44°in the helical axis (33), whereas a CPD bends DNA at an angle between 9°(33, 34) and 30° (38). A (6-4) photoproduct results in an enlargement of the major groove and a compression of the minor grove as compared with an average B-DNA structure (39). To a lesser degree this is also the case for probes containing CPD:AG and CPD:GG (39). In contrast, the paired CPD:AA has a wider minor groove and a narrower major groove than an average B-DNA structure. Intrinsic bending cannot, however, be the only important parameter. For example, substrates containing 1,2-dGpG cisplatin lesions have a 78°bend (40) but are very poorly recognized by DDB, whereas substrates containing multiple mismatches are not bent much if at all (41) but are bound with high affinity by DDB. Substrates with mismatches forming looped-out bases are bent (41) but are not bound well by DDB. DDB may be a conformational sensor that binds DNA when bent within a defined range of angles to allow interaction with a widened major groove or flexible enough to be bent within a range of angles.
These results regarding the binding affinity of DDB have been obtained from experiments using naked DNA. Chromatinized templates might yield different results, as lesions might be displayed differently because of the DNA topology necessary for formation of nucleosomes and higher order chromatin structures.
DDB1 and DDB2 Are Both Required for Binding to Damaged DNA-The results presented here also settle several long standing uncertainties regarding DDB1 and DDB2 subunit interaction and complex stability. In this study and several previous ones, multiple bands were observed by EMSA using DDB and substrates containing single lesions. In our experiments, these bands cannot result from independent binding of DDB to multiple lesions on the same DNA molecule. Cooperative intermolecular binding of multiple DDB complexes seems most likely to explain the multiple shifted forms. In a previous study, Chu and coworkers (25) detected such differently migrating protein-DNA complexes and asked whether they contained both DDB1 and DDB2. Complexes formed on UV light-irradiated DNA with cell extracts from strains overexpressing DDB subunits were analyzed to determine whether they could be supershifted with an antibody against an epitope tag on DDB2. The epitope tag was detected in the more slowly migrating complex but not in the major shifted complex, leading to the proposal of a "hit and run" model in which the DDB2 subunit leaves the complex after facilitating the binding of DDB1 to DNA (25). Perhaps in those experiments the N-terminal FLAG epitope on DDB2 in the major DDB1-DDB2-DNA ternary complex was inaccessible to the FLAG anti-body. In another study, a complex of DDB with DNA was found by immunoblotting to contain both DDB1 and DDB2 (30). The latter result was not compatible with a hit and run model, but under the assay conditions used (a relatively small amount of protein and inclusion of the competitor poly(dI-dC)), multiple complex species were not detected. We directly investigated the composition of both the faster and the slower migrating species in band shift assays by two-dimensional gel electrophoresis and immunodetection (Fig. 1D). Both shifted species are ternary complexes containing DNA, DDB1, and DDB2. Thus, DDB2 does not function to activate DDB1 in a hit and run fashion.
Two XP-E-associated Single Amino Acid Changes Nearly Eliminate DNA Binding Activity-Two XP-E mutations are single amino acid substitutions (K244E and R273H) corresponding to XP-E patients XP82TO and the related individuals XP2RO and XP3RO, respectively (32). Extracts from cells of these lines are defective in the ability to bind UV light-irradiated DNA fragments (25). These two mutations appear to have different effects on interaction with DDB1. Immunoblotting of DDB2 in mutant cell lines shows that the K244E mutant DDB2 protein is present in normal amounts in XP82TO cells, whereas the R273H mutant DDB2 protein is present in only trace amounts in XP3RO cells (20). Mixing of cell extracts containing recombinant DDB1 and mutant DDB2 has been used previously to probe DNA binding (42). We purified mutant DDB2-DDB1 complexes and found that when overexpressed in the presence of DDB1 in a baculovirus/insect cell system, both mutant DDB2 proteins (K244E and R273H) form complexes with DDB1.
The reason why the R273H protein level is reduced in XP3RO cells may be because it is more prone to degradation due to aberrant folding. Indeed, it is also prone to degradation in insect cells (Fig. 4B). The residual mutant protein may also bind more weakly to DDB1 so that the interaction does not survive the conditions required for detection by immunoprecipitation from cell extracts. Because a complex can be formed between DDB1 and either the K244E or R273H mutant DDB2 proteins, we conclude that neither residue is essential for the interaction between DDB1 and DDB2. Furthermore, as both purified mutant complexes display a dramatically reduced level of DNA binding, these residues are important for DNA binding activity. Functional residue mapping and a crystal structure of the DDB1-DDB2 complex will lead to a deeper understanding of the biochemistry, mechanism, and function of DDB.
Role of DDB in NER-There is good evidence that within cells DDB affects the rate of removal of UV radiation-induced CPDs from DNA. For example, in some rodent cell lines DDB2 expression is inhibited by cytosine methylation, and these cells show very slow removal of CPDs from the bulk of the genome (13). This repair defect can be restored either by reactivating DDB2 gene expression or by exogenous expression of DDB2 in the cells (15,26). DDB2-defective human XP-E (XP2RO) cells have a much reduced level of CPD repair (15). Further, DDB accumulates at sites of CPD lesions immediately after UV irradiation (14) suggesting that it plays a role in initial recognition of such damage. DDB may in fact cooperate with the XPC-RAD23B complex, which also plays a key role in damage recognition during NER. Recruitment of XPC to CPD lesions in vivo is DDB2-dependent (43), as is the tight association of XPC with chromatin following UV irradiation (44). Thus our finding that purified DDB actually recognizes a single CPD in DNA is consistent with these observations of a cellular role for DDB in repair of CPD lesions in the genome. The role of DDB may extend to other lesions as well, as at least one XP-E cell line (Ops1) has a reduced rate of removal of (6-4) photoproduct repair from the genome (16).
Because significant binding of DDB to a CPD has not been detected previously, more complex models have been proposed involving coop-erative low affinity binding of DDB, XPC-RAD23B, XPA, and RPA (replication protein A) to a CPD (12,45). Because DDB does indeed bind with significantly higher affinity to a CPD than to nondamaged DNA, such cooperative models are probably unnecessary, and it is possible that DDB is in fact the most important initial recognition factor for a CPD.
In Vivo Implications of Substrate Preferences-A role for DDB in NER is consistent with its relative abundance in cells. We determined the amount of DDB1 and DDB2 in a HeLa cell based on quantitative immunoblotting analysis of cell extracts compared with purified recombinant DDB and estimate 1.8 ϫ 10 5 DDB2 molecules and 8.4 ϫ 10 5 molecules of DDB1 per cell. These numbers are somewhat higher than previous estimates (18,29), but previous calculations were based on measurements of protein activity where reaction mixtures included a large excess of DNA competitor, resulting in an underestimate. Our estimate for DDB2 is within the range of 2.5 ϫ 10 4 to 2.5 ϫ 10 5 molecules per HeLa cell found for each of the other NER factors (46). A higher relative level of DDB1 is consistent with its participation in other cellular protein complexes not involving DDB2 (8).
The clinical features of XP-E patients are associated with sensitivity to UV radiation, underlining that recognition of DNA photolesion damage is a plausible role for DDB in NER. However, we find that DDB also has a high affinity for 2-or 3-bp mismatches, which are normally repaired by MSH2-MSH3-dependent mismatch repair (47), and for AP sites which are normally repaired by base excision repair (48). There are no reports of DDB playing a role in either of these DNA repair pathways. In view of the present results, it may now be worthwhile to investigate the functionality of mismatch repair and base excision repair in XP-E cells. It is also possible that 2-or 3-bp mismatches and AP sites might sometimes be bound by DDB in cells, either masking them from normal repair or leading to occasional processing by NER. The study of DDB may shed light on the poorly understood subject of how a particular DNA repair pathway is selected for the repair of a given lesion. Based on our findings, one may also speculate on the in vivo relevance of the high affinity of DDB for compound lesions containing both a CPD and a mismatch. CPD lesions in a temporarily mismatched environment might arise in DNA under unusually high supercoiled tension or in non-canonical DNA structures. Further, compound lesions may occasionally arise by translesion DNA synthesis (49). Conceivably, an auxiliary role for DDB would be to efficiently recognize compound lesions arising as a result of the action of specialized DNA polymerases that can bypass DNA adducts.