Analyzing the handoff of DNA from UvrA to UvrB utilizing DNA-protein photoaffinity labeling.

To better define the molecular architecture of nucleotide excision repair intermediates it is necessary to identify the specific domains of UvrA, UvrB, and UvrC that are in close proximity to DNA damage during the repair process. One key step of nucleotide excision repair that is poorly understood is the transfer of damaged DNA from UvrA to UvrB, prior to incision by UvrC. To study this transfer, we have utilized two types of arylazido-modified photoaffinity reagents that probe residues in the Uvr proteins that are closest to either the damaged or non-damaged strands. The damaged strand probes consisted of dNTP analogs linked to a terminal arylazido moiety. These analogs were incorporated into double-stranded DNA using DNA polymerase beta and functioned as both the damage site and the cross-linking reagent. The non-damaged strand probe contained an arylazido moiety coupled to a phosphorothioate-modified backbone of an oligonucleotide opposite the damaged strand, which contained an internal fluorescein adduct. Six site-directed mutants of Bacillus caldotenax UvrB located in different domains within the protein (Y96A, E99A, R123A, R183E, F249A, and D510A), and two domain deletions (Delta2 and Deltabeta-hairpin), were assayed. Data gleaned from these mutants suggest that the handoff of damaged DNA from UvrA to UvrB proceeds in a three-step process: 1) UvrA and UvrB bind to the damaged site, with UvrA in direct contact; 2) a transfer reaction with UvrB contacting mostly the non-damaged DNA strand; 3) lesion engagement by the damage recognition pocket of UvrB with concomitant release of UvrA.

The current model for UvrABC protein function during the initial steps of bacterial NER is as follows. In solution, UvrA dimerizes to become UvrA 2 , and ATP binding by the UvrA monomers promotes dimer formation (5,7). UvrA 2 interacts with UvrB to form the UvrA 2 B complex. Although the UvrA 2 dimer possesses the ability to recognize DNA lesions by itself (6,8), it is believed that the UvrA 2 B complex is the DNA damage recognition complex of the NER machinery (9). We have proposed a padlock model for UvrB detection of DNA damage in which the ␤-hairpin of UvrB is key to the proper recognition and processing of the lesion (9 -11). In this model, the UvrA 2 B complex binds to DNA and searches locally for DNA lesions fueled by ATP-dependent motions between domains 1a/b and the ␤-hairpin and domain 3 of UvrB. If no lesion is encountered, the UvrA 2 B complex dissociates from the DNA. If the proteins encounter a lesion, conformational changes occur such that the DNA containing the lesion, which is believed to be intimately in contact with UvrA, is passed via an unknown mechanism to UvrB (12,13). Thus, major conformational changes occur within the proteins as well as to the DNA (14). During this transition stage of damage verification, UvrA is believed to hydrolyze ATP and dissociate from the UvrA 2 B⅐DNA complex, or the process of UvrB engaging the lesion may in and of itself trigger the release of UvrA (15).
The padlock model predicts that UvrB binds ATP and harnesses the energy of this molecule to impose an unfavorable conformation on the DNA; the ␤-hairpin moiety of UvrB separates the DNA strands to facilitate incision (9). DNA in complex with UvrB has also been visualized as being bent and wrapped around the protein at this stage of the reaction (5,16). In addition, atomic force microscopy has revealed that UvrB might function as a dimer (17). Once UvrA departs from the UvrB⅐DNA complex, UvrC, which has two separate catalytic sites (18,19), can bind and execute the dual incision events. The incision mediated by the N-terminal nuclease domain of UvrC occurs 4 -5 nucleotides 3Ј to the lesion and precedes the incision 8 nucleotides 5Ј to the site of the lesion, mediated by the C-terminal UvrC nuclease domain. This dual action creates a 12-nucleotide fragment containing the lesion (18 -21).
Classically, protein-DNA interactions have been studied via the electrophoretic mobility shift assay (EMSA). Several laboratories including our own have utilized this technique to determine whether the UvrA and UvrB proteins interact in a productive manner (22)(23)(24). This technique can readily distinguish between UvrA 2 ⅐DNA and UvrB⅐DNA complexes. However, heterotrimeric complexes such as UvrA 2 B⅐DNA are poorly resolved, with the exception of complexes involving the ⌬␤hairpin UvrB mutant, which lacks residues 97-112 (11,15). In this case, a stable UvrAB⅐DNA complex is created, which migrates slower than the UvrA 2 ⅐DNA complex. However, it is not known which protein is in direct contact with the damaged DNA in this UvrAB⅐DNA complex. In general with current methodologies it has been a challenge to determine where in the reaction pathway the DNA containing the lesion is transferred to UvrB.
Here we describe the successful utilization of UV-activated photoreactive DNA probes containing azido groups that allow us to capture the UvrA and UvrB proteins during the NER process while they are in close proximity to these groups in dsDNA. We have taken a two-pronged approach in investigating the molecular handoff of damaged DNA from UvrA to UvrB. First, two dNTP analogs linked to terminal arylazido groups were incorporated into dsDNA using DNA polymerase ␤ (pol ␤). In this manner, these reagents served as both the site of DNA damage, as well as the cross-linking reagent. In the second approach, an arylazido compound was coupled to a phosphorothioate-modified backbone of an oligonucleotide opposite the damaged strand containing an internal fluorescein adduct. Thus, by design, protein contact sites on both the damaged and non-damaged sides of the DNA were probed.
We chose to examine various mutations of the UvrB protein thought to be important in the handoff procedure. Due to their varied location within the UvrB protein (see Fig. 1) as well as their phenotypes observed in the incision and EMSA assays, six site-directed mutations of Bacillus caldotenax UvrB (Y96A, E99A, R123A, R183E, F249A, and D510A) as well as two domain deletions (⌬2, which lacks amino acids Val 158 -Phe 244 , and ⌬␤-hairpin, which lacks 16 amino acids from Gln 97 to Asp 112 ) (15) were assayed via UV-activated photoaffinity crosslinking. Combined with techniques such as the EMSA and incision assays, the successful DNA-protein cross-linking technique described here helps to elucidate the role of the ␤-hairpin region, domain 2, and Tyr 96 of UvrB in the handoff of DNA from UvrA to UvrB.
DNA Substrates-All DNA substrates were synthesized by Sigma-Genosys (Woodlands, TX), except for the S-1 substrate, which was synthesized by Midland Certified Reagent Co. (Midland, TX). Upon receipt, oligonucleotides were resuspended in 1 mM Tris-HCl (pH 7.8) and 0.1 mM EDTA and stored at Ϫ20°C.
Radiolabeled DNA Duplex Notation-Throughout this paper we discuss a variety of duplex DNA substrates. For clarification purposes, we wish to point out in advance that the strand listed first in the naming of a duplexed oligonucleotide is 5Ј-32 P-labeled (e.g. for substrate F26/ ND, the F26 strand is 5Ј-labeled versus S-1 AZ /F26 where the S-1 AZ strand is 5Ј-labeled).
Fluorescein Substrate, F26/ND-The DNA sequence of the 50-mer double-stranded substrate containing a single internal fluorescein (FldT; Fig. 3A, position in sequence indicated by red T 26 ) adduct was: F26, (5Ј-GACTACGTACTGTTACGGCTCCATC[FldT] 26 CTACCGCAA-TCAGGCCAGATCTGC-3Ј), while the complementary non-damaged strand (ND) was 5Ј-GCAGATCTGGCCTGATTGCGGTAGCGATGGAG-CCGTAACAGTACGTAGTC-3Ј. The F26 strand was 5Ј-end-labeled using OptiKinase (United States Biochemical Corp.) and [␥-32 P]ATP (3000 Ci/mmol, Amersham Biosciences) according to the manufacturer's instructions. The reaction was terminated by the addition of 20 mM EDTA, and the enzyme was heat denatured by incubation for 10 min at 65°C. After labeling, the free nucleotides were removed by gel filtration chromatography (Micro Biospin-6, Bio-Rad). The labeled oligonucleotide was annealed with the complementary strand (ND) using equimolar amounts. The double-stranded character of the 50-mer duplex was analyzed on a native 12% polyacrylamide gel.
AZ Cross-linker Substrates, F26/S-1 AZ  phosphorothioate modification at the phosphodiester linkage between the 25th and 26th nucleotide from the 5Ј-end (indicated by blue S-1, Fig.  3A). The 32 P-labeled substrate was then modified to contain the azidophenacyl cross-linking moiety as described by Yang and Nash (26) with minor modifications. A stock solution of 10 mM p-azidophenacyl bromide was prepared in fresh Me 2 SO under reduced light conditions. The S-1 50-mer (25 l, 36 pmol), dissolved in 40 mM sodium bicarbonate (pH 9), was mixed with 10 mM p-azidophenacyl bromide (25 l) and incubated for 1 h at room temperature. Butanol (500 l) and dH 2 O (100 l) were added to the reaction mixtures. The butanol layer was removed, and the aqueous layer was extracted with an additional 2 ϫ 500 l of butanol. The aqueous layer was then mixed with glycogen (10 l; 2 mg/ml), 3 M sodium acetate (pH 5.5) (10% of total reaction volume), and 3 additional reaction volumes of ethanol and allowed to precipitate overnight at Ϫ20°C. The samples were then centrifuged for 60 min at 4°C. The pellets were washed with ice-cold 70% ethanol (1 ml) and then dried under vacuum. The azidophenacyl-modified oligonucleotide (S-1 AZ ) was resuspended in TE-100 (10 mM Tris (pH 7.5), 1 mM EDTA, 100 mM KCl) and annealed to the appropriate complementary oligonucleotide as described above. The ND oligonucleotide was subjected to the same procedures described above for S-1 to gauge nonspecific incorporation of the AZ modification.
Nicked Heteroduplex Substrates, dT 26 , dC 25 , XL1 26 , and XL2 25 -Nicked heteroduplex substrates were prepared by incubating the gapped heteroduplex substrates Gap 26 and Gap 25 (ϳ22 pmol) with 100 pmol of dTTP and XL1, or 400 pmol dCTP and XL2, respectively, in the presence of human pol ␤ (45 pmol, a generous gift from Drs. Rajendra Prasad and Samuel Wilson, NIEHS, National Institutes of Health) in 13 l of 1ϫ pol ␤ buffer (50 mM Tris-HCl (pH 7.5), 50 mM KCl, 20% glycerol) for 1 h at 37°C. Incorporation of natural or base-substituted photoreactive dNMP (XL1, XL2) residues into the 3Ј-end of the DNA was analyzed on a 10% denaturing polyacrylamide gel and electrophoresis was performed at 350 V in 1ϫ Tris-borate-EDTA buffer (TBE: 89 mM Tris, 89 mM boric acid and 2 mM EDTA). The gels were dried and exposed to a PhosphorImager screen (Amersham Biosciences) overnight. Incorporation efficiency was calculated using the Molecular Dynamics software ImageQuant.
UvrABC Incision Assay-The 5Ј-end-labeled duplex DNA (2 nM) was incised by the thermophillic UvrABC enzymes (20 nM B. caldotenax UvrA, 100 nM B. caldotenax UvrB, 50 nM T. maritima UvrC) in 20 l of UvrABC incision buffer (50 mM Tris-HCl (pH 7.5), 50 mM KCl, 10 mM MgCl 2 , 1 mM ATP, and 5 mM dithiothreitol) at 55°C for 30 min. The reaction was terminated by addition of 20 mM EDTA or via the direct addition of 10% of the reaction volume to 5 l of formamide, followed by heating to 85°C for 15 min and then cooling on ice. The incision products were resolved on a 10% denaturing polyacrylamide gel, and electrophoresis was performed at 350 V for 45 min in 1ϫ TBE. The gels were processed and developed as described above. The incision efficiency was calculated using the Amersham Biosciences software ImageQuant.
DNA-UvrAB Cross-linking Assay-The 5Ј-end-labeled duplex DNA containing the cross-linker moiety (2 nM, XL1 26 , XL2 25 , or S-1 AZ ) was incubated with the UvrAB proteins (200 nM B. caldotenax UvrA, 1000 nM B. caldotenax wild-type (WT) UvrB or mutants) in 20 l of reaction volume containing UvrAB cross-linking buffer (50 mM Tris-HCl (pH 7.5), 50 mM KCl, 10 mM MgCl 2 , 1 mM ATP) at 55°C for up to 30 min in the dark. Reaction vials were immediately transferred to a platform 5 cm below a 365 nm UV lamp (UVP, Blak-Ray Longwave UV Mercury lamp, 100 watts) for indicated time points, typically 5 min). A sample of the reaction (8 l) was then removed and incubated with 10 l of NuPage 4ϫ LDS sample buffer (Invitrogen) containing dithiothreitol and heated for 15 min at 85°C. The cross-linked products were resolved on a 10% NuPage Bis-Tris gel (Invitrogen) in 1ϫ NuPage MOPS SDS running buffer (Invitrogen). The gels were processed and developed as described above. The percent of DNA cross-linked to protein was calculated using the Amersham Biosciences software ImageQuant.
Electrophoretic Mobility Shift Assay-Binding reactions were performed with 2 nM duplexed DNA substrate (5Ј-32 P-labeled F26/NDB), 25 nM B. caldotenax UvrA and 125 nM B. caldotenax WT UvrB or mutant in 20 l of UvrABC buffer for 20 min at 55°C. The products were resolved on a 4% native polyacrylamide gel (19:1). The gel and the buffer contained 0.5ϫ TBE (44.5 mM Tris-HCl (pH 7.5), 44.5 mM boric acid, 1.25 mM EDTA) with 1 mM ATP and 10 mM MgCl 2 . Electrophoresis was carried out at 4°C for ϳ2 h (ϳ40 mA per gel). The gels were developed and processed as described above. The percent of DNA shifted was determined using the Amersham Biosciences software ImageQuant. Previous studies have shown that a centrally located lesion such as fluorescein (F26/NDB) is well recognized by the UvrABC NER system (11,15). By placing cross-linking reagents on either the damaged or non-damaged strands of the DNA, this study aims to capture the regions of the proteins that are in closest proximity to the lesion during protein assembly for NER and assess the handoff of damaged DNA from UvrA to WT UvrB and UvrB mutants from the perspective of both the damaged site and the non-damaged strand.

Experimental Strategy-Oligonucleotides for this study (see
Photoaffinity labeling has been used to study protein-DNA interactions within a variety of biochemical systems (27)(28)(29). Orren and Sancar (30) showed, using a psoralen monoadduct, that UvrB forms intimate contacts with the damage site and that UvrD (helicase II) is necessary to dissociate the excised oligonucleotide from the post-incision protein-DNA complex. It was suggested that within this complex, UvrB remains bound to the non-damaged strand and was released by the action of DNA polymerase I. In a similar manner, Reardon and Sancar (31) probed the molecular anatomy of the preincision complexes of human nucleotide excision repair proteins.
Base-substituted analogs of dNTP carrying photoreactive arylazido groups (32) have been utilized previously to design photoreactive DNAs to study the molecular interactions of select base excision repair proteins (33,34) and to study other proteins such as human replication protein A (35,36). Similarly, a variety of reactive dNTPs with photoreactive azide functional groups have been surveyed for their ability to crosslink DNA to yeast RNA polymerase in their transcription complexes (37). Phosphorothioester backbone modifications have been utilized to successfully couple an azidophenacyl moiety to oligonucleotides and cross-link proteins such as integration host factor and transcription factor IIIB to DNA (26,38). By incorporating both photoreactive azide-containing dNTPs and the azidophenacyl cross-linking moieties into oligonucleotides in this study, we sought to gain new insight into the interaction of UvrA and UvrB with damaged DNA. We have included a graphic representation of our modified DNA substrates superimposed on the crystal structure of B. caldotenax UvrB Y96A, Fig. 3E, (10, 15). As stated above, we have strategically designed our substrates to contain a cross-linking probe on either the damaged or non-damaged strand (XL1 26 C, a photoreactive AZ functional group was coupled to the backbone of a phosphorothioate-containing (at the phosphodiester linkage, S-1) oligonucleotide. The terminal azido group (N 3 ) in each reagent is activated by exposure to UV light (365 nm) allowing DNA-protein cross-linking to occur via a highly reactive nitrene intermediate.  26 ϩ Nick, not ligated, small gray box) and with cross-linking functionality (XL1 26 and XL2 25 ϩ Nick, not ligated) were prepared via incorporation of dCTP or XL2 at position 25 and dTTP or XL1 at position 26 into the appropriate Gap 25/26 heteroduplex substrates using DNA pol ␤. C, the F26/ND substrate, our standard UvrABC NER damaged oligonucleotide, contains a centrally located fluorescein adducted thymine (FldT, yellow oval). D, the S-1 AZ /F26 substrate also contains the FldT lesion as well as the AZ (blue box) moiety in the non-damaged strand at the S-1 position, which is 5Ј to the FldT lesion on the opposite strand. ND AZ /F26 was treated with AZ in the same manner as the S-1 AZ /F26 substrate to assess nonspecific incorporation of the AZ group in the absence of a specific phosphorothioate modification. E, putative DNA binding model based on the crystal structure of B. caldotenax UvrB Y96A (9, 10, 15) with a graphic representation of the potential points of contact for our cross-linking reagents. The fragment of dsDNA (red) is shown separated by the ␤-hairpin region of UvrB (blue C␣ trace) with the non-damaged strand of DNA clasped between the ␤-hairpin and domain 1b. Undisturbed base pairs are represented with yellow and green spokes within the DNA. The points of 3Ј-and 5Ј-incision are indicated with black arrows. (25) followed by incorporation of a photoreactive dNMP moiety into a gapped DNA substrate using DNA pol ␤ (34, 39). The backbone modification required synthesis of a 50-mer containing a specific backbone phosphodiester modification followed by reaction with AZ. Prior to assaying our UvrB point mutations for their ability to interact with UvrA and damaged DNA, a series of optimization experiments were carried out as described below.
Design and Optimization of Damage-dependent Cross-linking Reagents (XL1 26 and XL2 25 )-The photoreactive dTTP or dCTP analogs (XL1 or XL2) contain an arylazido group projecting into the major groove from either the C5 or N4 positions, respectively. The 5Ј-32 P-labeled, heteroduplexed, 50-base pair dsDNA substrates with a gap at position 25 or 26 (Fig. 3B, Gap 25/26 ) were incubated with dCTP or XL2 and dTTP or XL1, respectively, and human pol ␤ in reaction buffer for 1 h at 37°C. A dNTP analog (XL1 or XL2) was added to the respective reaction mixtures containing gapped DNA substrates in the absence of DNA polymerase ␤ as a negative control for incorporation (Fig. 4A, lanes 1 and 4). Incorporation of dNTPs or the dNTP photoreactive analogs went essentially to completion as indicated by the more slowly migrating species shown in Fig.  4A (lanes 2, 3, 5, and 6). The DNA was not ligated but left as a nick to minimize the size of DNA that remained covalently bound to the protein. After successful incorporation of the photoreactive dNMP moieties into DNA, various cross-linking conditions were tested and optimized for the XL modified substrates. With exposure to UV (365 nm) held constant at 10 min, it was determined that optimal cross-linking of XL1 26 (2 nM) to UvrA (ϳ15%) was consistently obtained at 200 -250 nM UvrA (Fig. 4B). Utilization of the longer, lower energy 365 nm wavelength not only facilitated DNA-protein cross-linking but also avoided the structural damage to both proteins and DNA imposed by shorter UV wavelengths (e.g. 254 nm). Subsequently, keeping the concentration of UvrA constant at 200 nM, the length of UV (365 nm) exposure was varied from 0 to 15 min (Fig. 4C). The gel and graph clearly indicated that optimal cross-linking was achieved for both XL1 26 and XL2 25 substrates (2 nM) with a 5-min exposure. Last, we examined the ability of substrate XL1 26 (2 nM) to cross-link to UvrB in the presence of UvrA by holding the concentration of UvrA constant at 200 nM and titrating the concentration of UvrB from 200 to 2000 nM (Fig. 4D). It was determined that optimal cross-linking conditions for XL-containing DNAs were 200 nM UvrA, 1000 nM UvrB, and 5 min UV exposure. It is also observed from Fig. 4D that as the concentration of UvrB is in- creased, the cross-linking to UvrA is decreased consistent with the handoff of damaged DNA from UvrA to UvrB. It is important to note that no DNA-protein cross-linking occurs with either XL reagent in the absence of UV exposure (Fig. 4B, lane  2, and Fig. 4C, lane 1) or in the absence of either XL reagent (Fig. 4B, lane 1, and 3D, lane 2), and no DNA cross-linking to UvrB occurs in the absence of UvrA (Fig. 4D, lane 5). The full gel shown in Fig. 4D, representative of all cross-linking experiments described in this paper, clearly shows cross-linking bands specific to UvrA and UvrB and that as cross-linking bands appear, the amount of free DNA decreases accordingly. Bovine serum albumin (BSA) was used as a control to check for nonspecific cross-linking to proteins with no known DNA damage recognition functions.
The cross-linking bands that are shown and quantitated in all figures of this study are attributed to DNA substrate cross-linked to B. caldotenax UvrA (ϳ105 kDa) or B. caldotenax UvrB (ϳ76 kDa) protein monomers. While it is conceivable that the slower migrating band observed in the full-length gels shown in Figs. 4D and 5, A and B, could correlate with UvrA still bound to the intein-chitin binding domain tag (total MW ϳ161 kDa), this is unlikely based on the purity of UvrA used in these studies (Fig. 7A). Since the exact nature of the faint, slower migrating cross-linking band is currently unknown, the counts associated with this band are not included in the analysis.
Design and Optimization of Non-damaged Strand Crosslinking Probe (ND AZ /F26 Versus S-1 AZ /F26)-Cross-linking conditions for the AZ-treated DNA substrates were also optimized prior to assaying the UvrB mutants for their ability to interact with UvrA and DNA. The concentration of UvrA (Fig.  5A) and length of UV (365 nm) exposure (Fig. 5B) were varied FIG. 5. Characterization of non-damaged strand cross-linking reagent (AZ). Cross-linking conditions for the AZ-treated substrates were optimized for UvrA concentration (n ϭ 2, mean) (A) and length of UV (365 nm) exposure (n ϭ 2, mean) (B). The DNA concentration in each reaction was 2 nM. Both S-1 AZ /F26 and ND AZ /F26 substrates were treated identically with the AZ reagent. Only the S-1 AZ /F26 substrate contained a specific phosphorothioate modification. Cross-linking was observed with the ND AZ /F26 substrate indicating nonspecific incorporation of the AZ reagent. The percentage of nonspecific cross-linking observed to each protein was subtracted as background for each reaction utilizing the AZ-treated substrates. It was determined that the optimal cross-linking conditions were 200 nM UvrA and 5-min UV exposure at a 5-cm distance from the UV source. A molecular weight protein marker is shown to the right of the full gel in each panel. Cross-linking bands outlined in the full gels are expanded to the right (A, panels i and ii; B, panels iii and iv). Percent cross-linking is shown for each substrate in the graphs just below the gel expansions (ND AZ /F26, gray diamonds; S-1 AZ /F26, black boxes).
while cross-linking to azidophenacyl-treated substrates S-1 AZ / F26 and ND AZ /F26. The concentration of DNA in each reaction was 2 nM for all experiments. Again, optimal cross-linking conditions were determined to be 200 nM UvrA and 5-min UV exposure (365 nm) at 5 cm below the UV source. Both S-1 AZ / F26 and ND AZ /F26 substrates were treated identically with the AZ reagent but only the S-1 AZ /F26 substrate contains a specific phosphorothioate modification. Surprisingly, the ND AZ /F26 substrate exhibited cross-linking activity indicating that the azidophenacyl bromide reacts with DNA, regardless of specific phosphorothioate modifications, at some background level. This background reactivity of azidophenacyl bromide has not apparently been reported previously. We speculate that the nonspecific modification observed for the ND AZ /F26 substrate may be due to mild nucleophilic reactivity of purines with azidophenacyl bromide. Therefore, for each cross-linking reaction involving AZ, we probably overestimate the contribution of cross-linking due to AZ incorporation at nonspecific (i.e. nonphosphorothioester) sites within the DNA substrate. The percentage of nonspecific cross-linking observed to each protein was subtracted as "background" for each reaction utilizing the AZ-treated substrates.
Recognition of Substrates via UvrABC Incision Assay-It was necessary to ensure that the substrates prepared for this study, despite their initial ability to cross-link to the UvrA and UvrB proteins, were well recognized and incised as a DNA adduct by the UvrABC NER machinery. Therefore, each substrate was subjected to our standard UvrABC incision assay (11). Each 5Ј-32 P-end-labeled substrate (2 nM, see Fig. 3, A-D) was incubated with 20 nM B. caldotenax UvrA, 100 nM B. caldotenax WT UvrB, and 50 nM T. maritima UvrC for 30 min at 55°C in reaction buffer (Fig. 6, A and B). The reactions were terminated with EDTA, and the incision products were analyzed on a 10% denaturing polyacrylamide gel. Fig. 6C is a graphic comparison of the incision activity of UvrABC on each substrate. We typically observe ϳ70% incision of our laboratory's standard substrate F26/ND. Surprisingly, we also observed that substrates containing a single nucleotide gap (Gap 25/26 ) or nick (dT 26 ) were all incised with nearly the same efficiency as F26/ND (65, 69, and 64% respectively, versus 71%). The nicked dC 25 substrate yielded a lower average incision of 40%. We observed that both cross-linking substrates XL1 26 and XL2 25 were also well recognized by the UvrABC repair system (82 and 79% incision, respectively).
As expected, the S-1 AZ /F26 substrate was incised only 15% (Fig. 6B, lane 6). A 15% incision result indicates a 4 -5-fold preference of the Uvr proteins for the damaged (F26) versus the non-damaged (S-1 AZ ) side of the DNA. In this manner, we show that UvrABC is oriented upon the F26 strand, leaving the S-1 AZ moiety free to interact with regions of the proteins positioned on the "non-damaged" side of the DNA. Incision of F26 when duplexed to the AZ-treated S-1 oligonucleotide yielded a reduced overall incision efficiency relative to F26/ND (55% versus 70%, respectively; Fig. 6B, lanes 2 and 4 versus Fig. 6A,  lane 2). Thus, at least 80% of the UvrABC system complexes assembled on the F26 strand of the DNA substrate.
Electrophoretic Mobility of UvrAB Complexes-Stable protein-DNA interactions can be visualized by employing the EMSA technique; however, transient or labile protein-DNA interactions are not easily resolved by this technique. Prior to EMSA, each protein utilized in this study was examined on a SDS-PAGE gel to ensure that stock solutions of all proteins were essentially pure and of the same concentration (1.25 M, Fig. 7A). WT or mutant UvrB proteins were then subjected to EMSA analysis in the presence of B. caldotenax UvrA (Fig. 7B). The UvrA 2 ⅐DNA complex can be readily identified in most lanes. However, in some lanes there is a significant upward smearing pattern indicative of dissociation of labile larger complexes, probably the UvrA 2 B⅐DNA complex; compare the compact band of the lane containing only UvrA 2 with the lanes of UvrA 2 plus WT UvrB or Y96A (Fig. 7B, lane 2 versus lanes 3 or  4). The addition of a UvrB protein results in a band that migrates slightly slower compared with the pattern of UvrA 2 alone. More readily distinguishable from the UvrA 2 ⅐DNA complex is the stable, a slower migrating UvrAB⅐DNA complex formed in the presence of the UvrB mutant ⌬␤-hairpin (Fig. 7B,  lane 11).
Last, we can visualize the UvrB⅐DNA complex as the faster migrating species in our EMSAs. WT UvrB in the presence of UvrA and damaged DNA produces a UvrB⅐DNA complex (26%). The F249A mutant forms a UvrB⅐DNA (30.5%) complex comparable with WT UvrB. Weaker complexes (ϳ50% or less of WT complex formation but still detectable) are observed for R123A (14.6%) and R183E (7.8%). A very weak complex is observed for E99A (1.4%). The remaining mutants examined here, Y96A, D510A, ⌬2, and ⌬␤h, are completely defective in producing a UvrB⅐DNA complex suggesting that these regions of UvrB are essential for the transfer of damaged DNA from UvrA to UvrB. Fig. 7C summarizes the percent DNA shifted for all protein-DNA complexes. Within the UvrAB⅐DNA complexes we cannot distinguish on which protein the lesion containing DNA resides; therefore we chose to evaluate DNA-protein cross-linking with each UvrB mutant in the presence of UvrA to determine whether the DNA was ever transferred from UvrA to the various mutant proteins.
Probing the Damage Recognition Domains of UvrA and UvrB with XL1 26 and XL2 25 -DNA substrates XL1 26 and XL2 25 (2 nM) were incubated with UvrA (200 nM) and WT UvrB or point mutants (1000 nM) for 30 min at 55°C in the dark followed by UV irradiation (365 nm, 5 min). BSA, which does not bind specifically to damaged DNA, was used as a negative control. Fig. 8, A and B, display the DNA-protein cross-linking products observed for the XL1 26 and XL2 25 photoreactive DNA substrates to UvrA and WT and mutant UvrBs. No DNA-protein cross-linking is observed to BSA (lane 1), no DNA-protein crosslinking occurs in the absence of UV irradiation (lane 2), and no DNA-protein cross-linking is observed to WT UvrB in the absence of UvrA (lane 4). The percentage of DNA cross-linked to protein is summarized in Fig. 8, C and D, for the XL1 26 and XL2 25 substrates, respectively. In general, the percent of DNA cross-linked to protein was nearly 2-fold higher for the XL2 25 substrate compared with the XL1 26 substrate (average crosslinking to UvrA was 14.8% for XL2 25 versus 8.3% for XL1 26 ). The presence of WT UvrB caused a decrease in the crosslinking to UvrA of ϳ50% for the XL1 26 substrate (Fig. 8, A and  C, lanes 3 and 5). The XL2 25 substrate cross-linked to UvrA with the same efficiency in both the presence and absence of WT UvrB (Fig. 8, B and D, lanes 3 and 5). All UvrB proteins, with the exception of the ⌬2 mutant, were captured to some extent (1.3-6.7%) by the XL2 25 substrate (Fig. 8D). Only WT and F249A UvrB were captured with XL1 26 (Fig. 8C). Crosslinking to UvrA is enhanced by the presence of the ⌬2 and ⌬␤h mutants (40% increase for XL1 26 and 70% increase for XL2 25 ). Neither substrate was able to capture the Y96A UvrB mutant.
Probing the Non-damaged Strand with S-1 AZ /F26 -DNA substrates ND AZ /F26 and S-1 AZ /F26 (2 nM) were incubated with UvrA (200 nM) and WT UvrB or point mutants (1000 nM) for 30 min at 55°C in the dark followed by UV irradiation (365 nm, 5 min). BSA was also used as a negative control. Fig. 9, A and B, are representative of the DNA-protein cross-linking bands observed for the ND AZ /F26 and S-1 AZ /F26 substrates, respectively. The amount of non-target (ND AZ ) cross-linking observed in Fig. 9A was subtracted from the specific (S-1 AZ ) cross-linking observed in Fig. 9B to yield the percentage of DNA cross-linked to each protein reported in Fig. 9C. As for the XL reagents, no DNA-protein cross-linking is observed to BSA. Due to the proposed orientation of the AZ reagent in these two substrates (see Fig. 3E), we believe that most (80% or greater) protein "captured" would be closest to the non-damaged strand of the DNA. Cross-linking of AZ containing substrates to UvrA in the presence and absence of WT UvrB or mutants is 1.5 to 3 times lower compared with the XL1 26 and XL2 25 substrates, respectively (average cross-linking to UvrA for S-1 AZ /F26 was 5.4%). Approximately 2-4% cross-linking to UvrB mutants (except ⌬2) is achieved with the S-1 AZ /F26 substrate (Fig. 9C). Again, cross-linking to UvrA is enhanced in the presence of ⌬2 UvrB, while no cross-linking to ⌬2 is achieved. In contrast to both the XL-containing substrates, the Y96A mutant is captured (ϳ2%) by the AZ-containing substrates. Finally, the DNA backbone modified substrate S-1 AZ /F26 consistently captures (ϳ2-4%) a majority (eight out of nine) of the UvrB proteins examined here, whereas the damaged strand cross-linking reagents, XL1 26 and XL2 25 , capture two and seven out of nine UvrB proteins, respectively.

DISCUSSION
To probe the molecular handoff of damaged DNA from UvrA to UvrB during NER, we have developed two types of photoaffinity labeled DNA substrates that probe the regions of proteins making contact along the damaged or non-damaged strands. Two of these DNA-protein cross-linkers are arylazidomodified dNTPs, FABC-dUTP (XL1) and FAB-dCTP (XL2), reagents that function as both the lesion and the cross-linker within the DNA substrate (Fig. 2, A and B). The third DNAprotein cross-linking substrate consisted of a phosphorothioate (S-1) DNA backbone modified with AZ (Fig. 2C) on the strand of DNA opposite a fluorescein adduct at position 26. Sitespecific mutants and domain deletion mutants of UvrB were used to systematically address how the DNA travels from UvrA to UvrB during the process of nucleotide excision repair. In the course of this study we found that the XL substrates (XL1 26 and XL2 25 , Fig. 6C) were incised 82 and 79%, respectively, compared with a single nucleotide gapped DNA substrate (Gap 26 and Gap 25 , 69 and 65% incised, respectively) and fluorescein-containing DNA substrate (F26/ND; 71% incised) indicating that these DNA-XL substrates were well recognized by the NER proteins. Surprisingly, we also discovered that substrates containing a single nick were good substrates for the thermophilic UvrABC NER machinery (Fig. 6C, dT 26 and dC 25 , 64 and 40% incised, respectively). The S-1 AZ /F26 substrate was poorly incised (ϳ15%) indicating that the Uvr proteins targeted the fluorescein adduct (F26) as bona fide DNA damage, leaving the AZ modification available to probe the amino acid residues that are in proximity to the non-damaged strand of DNA (Fig.  6C). Goosen and co-workers (40) have shown previously that the UvrBC nuclease efficiently incised DNA gaps and nicks in a UvrA-independent manner. In contrast, we find UvrA to be required for efficient incision of nicked or gapped substrates. Thus we believe our thermophilic UvrABC proteins are sensing a nick or a gap, which at 55°C might be sufficiently dynamic and partially denatured, to be recognized as a DNA lesion.
B. caldotenax UvrA and B. caldotenax UvrB were efficiently cross-linked (3.7-20.9% and 1.3-6.7%, respectively) to the two types of photoaffinity probes. Experiments with UvrA and WT UvrB indicated that UvrA was cross-linked about 2-fold more efficiently than WT UvrB. With respect to the XL1 and XL2 containing substrates, results indicate that UvrA makes more intimate contacts with the major groove azido moiety and that the azido-DNA adduct is further removed from UvrB, despite its tight binding to DNA. Furthermore, cross-linking DNA to UvrB via FAB-dCTP (XL2) was two to four times more efficient than FABC-dUTP (XL1) indicating that the cross-link arm of XL1, which projects from the five position of uridine into the FIG. 9. Cross-linking data for non-damaged strand cross-linker (AZ). DNA substrates ND AZ /F26 and S-1 AZ /F26 (2 nM) were incubated with UvrA (200 nM) and WT or mutant UvrB (1000 nM) for 30 min at 55°C in the dark. Samples were exposed to 5 min of UV and processed as described under "Experimental Procedures." A and B are representative gels that display the DNA-protein cross-linking bands observed for the ND AZ /F26 and S-1 AZ /F26 substrates, respectively. For each protein, the amount of nonspecific (ND AZ ) cross-linking observed in A was subtracted from the specific (S-1 AZ ) cross-linking observed in B to yield the percentage of DNA cross-linked to protein reported in C (n ϭ 3 for S-1 AZ /F26 experiments, n ϭ 4 for ND AZ /F26 experiments; percent cross-linked is the mean Ϯ S.D.). Gray bars indicate percent cross-linked to UvrA; black bars indicate percent cross-linked to WT or mutant UvrB. major groove, may not be in optimal juxtaposition in the UvrB⅐DNA complex for efficient cross-linking.
Analysis of photoaffinity cross-linking of several UvrB point mutations and domain deletions that arrest repair at specific points has given molecular insight into the transition of a UvrAB⅐DNA complex to a UvrB⅐DNA complex (Fig. 10). In this model, UvrA and UvrB are shown to bind to the adducted DNA in a series of discrete steps. After forming a UvrA 2 B complex (step 1), which is dependent upon domain 2 of UvrB, the initial contact with DNA damage occurs through UvrA (step 2).
Step 3 involves Uvr isomerization where the DNA is physically moved from UvrA to UvrB. Step 4 is characterized by the release of UvrA and engagement of the DNA damage by UvrB to form the UvrB⅐DNA propreincision complex (41) followed by the formation of the UvrB⅐DNA* complex, which is incisioncompetent (42). UvrC is then recruited to the preincision complex near the site of DNA damage to perform both 3Ј and 5Ј incisions (step 5). The positions of the major defects for each of the UvrB mutants are indicated by a blocked arrow in each step of Fig. 10 and described below.
Based on our previous work (15), domain 2 and Arg 183 of UvrB are essential for forming a productive UvrA 2 B complex (Fig. 10,  step 1). However, based on EMSA (Fig. 7B) (15) and cross-linking data (Figs. 8 and 9), it would appear that another region of UvrB can interact with UvrA to shift the equilibrium of UvrA from monomer to dimer, facilitating DNA binding. It is possible that the C-terminal coiled-coiled region of UvrB, while helping to promote interaction with UvrA, helps to stabilize UvrA dimer formation but is not sufficient for interaction with UvrA, in the absence of DNA (15). Our data also suggest that conversion of Arg 183 to glutamic acid (R183E) is a "leaky" mutation in that it greatly reduces, but does not completely block, UvrA-UvrB inter-action (Fig. 10, step 1). Thus, a small but detectable amount (ϳ2%) of UvrB was cross-linked to the photoaffinity substrates, XL2 25 and S-1 AZ /F26 (Figs. 8B and 9B, respectively).
Asp 510 is located in domain 3 and forms part of the ATP binding pocket of B. caldotenax UvrB. We have shown previously that the D510A UvrB mutant is defective in UvrA⅐DNA damagedependent ATPase activity leading to a loss of incision efficiency (43). This mutant does not form a UvrB⅐DNA complex as determined by EMSA (Fig. 7B). Results presented here would suggest that ATP binding by UvrB is sufficient for UvrA 2 B complex formation, and since we see little enhanced UvrA binding by either EMSA or photoaffinity cross-linking of UvrA, we believe the primary defect of the D510A mutant of UvrB is an inability to promote efficient UvrA detection of damage (Fig. 10, step 2). It is interesting to note that a small but detectable amount (2%) of UvrB cross-linking is observed with this mutant to the XL2 25 and S-1 AZ /F26 substrates (Figs. 8 and 9), suggesting that while ATP hydrolysis by UvrB greatly stimulates the subsequent isomerization process, it might not be absolutely required (Fig. 10, step 3). ATP hydrolysis is absolutely required for UvrB loading (Fig. 10, step 4) as evidenced by the lack of a UvrB⅐DNA complex by EMSA for D510A UvrB (Fig. 7B, lane 9). This idea is consistent with the work of Goosen and co-workers (41) who suggest that ATP binding allows the formation of the propreincision UvrB(UvrA)⅐DNA complex, but hydrolysis is required for the formation of the UvrB⅐DNA preincision complex.
The ⌬␤-hairpin UvrB mutant is clearly defective in the UvrA to UvrB handoff process as almost no cross-linking (Ͻ 2%) is observed for the ⌬␤-hairpin mutant in the presence of UvrA and damaged DNA (Fig. 8, A and B, lane 13, and Fig. 9B, lane  11). These data are consistent with work by Goosen and coworkers (40) who showed that the Y101A/F108A double mutant was defective in incision, suggesting that the tip of the ␤-hairpin is necessary for strand opening and efficient damage detection. UvrA is unable to transfer the DNA efficiently to the UvrB ⌬␤-hairpin mutant. Thus, UvrA retains the DNA longer causing a greater percentage of the DNA to be cross-linked to it. A faint cross-link band observed in the lane (Fig. 8B, lane 13, and Fig. 9B, lane 11) is evidence that some transfer from UvrA may occur. But, as evidenced by EMSA, no stable UvrB⅐DNA complex ever forms with the ⌬␤-hairpin UvrB mutant (Fig. 7B, lane 11). It is interesting to note that cross-linking to the phosphorothioate (S-1) DNA backbone modified with AZ for the ⌬␤-hairpin mutant is not completely inhibited. This suggests that along the pathway to forming a stable UvrB⅐DNA complex, the non-damaged strand is probably clasped between the ␤-hairpin and the wall of domain 1b (Fig. 3E) prior to the engagement of the damaged strand. This model is consistent with data generated with the Y96A mutation. Tyr 96 is located at the base of ␤-hairpin region of WT UvrB where it is proposed to be intimately in contact with the damaged strand of DNA (15,40,44). No UvrB⅐DNA complex is observed via EMSA (Fig. 7B, lane  4), no cross-linking is observed with XL1 26 , and very little crosslinking is observed with XL2 25 (lane 6 of Fig. 8, A and B, respectively). Weak cross-linking (2%) is observed with the nondamaged strand cross-linker, S-1 AZ /F26 (Fig. 9B, lane 4). These data suggest that the damaged DNA is never efficiently transferred to Y96A. It is possible that this residue is required to hold the damaged DNA in place. Thus, our cross-linking results are in agreement with previously published data (40,44) regarding the importance of Tyr 96 in damage recognition.
During the process of damage engagement (Fig. 10, step 4), UvrB is believed to engage the adduct during damage verification such that a stable UvrB⅐DNA complex is formed. It would appear that the E99A mutant of UvrB is defective in this step. When Glu 99 is mutated to alanine, incision is reduced to less indicates an altered conformation of DNA in which the DNA is bent and unwound and that the B⅐DNA* complex is capable of supporting incision (42). In step 5, DNA 3Ј and DNA 5Ј indicate DNA containing 3Ј and both 3Ј and 5Ј incisions, respectively. than 5% of WT. 2 In addition, this mutant fails to form a stable UvrB⅐DNA complex (Fig. 7B, lane 5). We have proposed that Glu 99 is involved in a salt-bridge contact or charge repulsion at the tip of the ␤-hairpin. When mutated, the ␤-hairpin may not be in the proper orientation to facilitate the opening of DNA (9 -11). It is interesting to note that E99A UvrB did not form a cross-link with XL1 26 , but cross-linking was achieved to this mutant with XL2 25 and S-1 AZ /F26, albeit 40 and 28% less than WT UvrB, respectively. This indicates that each of the three photoaffinity substrates appear to be probing a different region of UvrB and that Glu 99 is essential for strong engagement of the adduct but not as essential for the non-damaged DNA contact.
Arg 123 is located in domain 1a, close to domain 2. Very poor incision 2 and very weak cross-linking with XL2 26 and S-1 AZ / F26 (Fig. 8B, lane 8, and Fig. 9B, lane 6) to the R123A mutant suggest that this residue plays a key role in the handoff of damaged DNA from UvrA to UvrB. Arg 123 may provide an attractive force with the negatively charged phosphate on the non-damaged strand. In this manner, the mutant is trapped between the transition from binding the non-damaged strand and forming a tight UvrB⅐DNA complex. This residue may also be important in communications with domain 2, which interacts with UvrA (15).
The final residue examined in this study was Phe 249 , which is located in the beginning of domain 1b. Mutation of Phe 249 to alanine enhanced both incision and UvrB⅐DNA complex formation 2 (Fig. 7B, lane 8). This mutant was also cross-linked up to 2-fold better to the photoaffinity substrates than WT UvrB. The highest amount of cross-linking to this mutant was achieved with XL2 25 (12.9% UvrA and 6.7% F249A UvrB; Fig. 8B, lane  10). These results indicate that Phe 249 may play a key role in DNA binding. Phe 249 is located near the DNA binding pocket of UvrB. Replacement of the larger phenylalanine residue with a less sterically interfering alanine may cause the binding pocket to collapse slightly; damaged DNA can still fit in the pocket but is held more tightly once there. In this manner, the DNA is more readily transferred to UvrB resulting in a less transient, more stable UvrB⅐DNA preincision complex (Fig. 10, step 4, white arrow), which is consistent with the results achieved in this study. CONCLUSIONS A new assay to study the interactions of NER proteins with damaged DNA during the DNA repair process has been successfully implemented and described. DNA-protein photoaffinity cross-linking has allowed a visualization of the architecture of the DNA when in complex with UvrA and UvrB proteins such that we can now dissect the molecular handoff of DNA from UvrA to UvrB into discrete steps, the most important step being Uvr isomerization in which UvrB comes into close contact with the adduct within the UvrA 2 B complex, prior to UvrA dissociation. We believe that the ␤-hairpin deletion and Y96A mutants are arrested during this normally transient step that proceeds by movement of the non-damaged strand into the pocket between the ␤-hairpin and domain 1b. This is followed by engagement of the damaged strand at the base of the ␤-hairpin through hydrophobic residues, primarily Tyr 96 . The DNA probes containing photoreactive groups on both the damaged and non-damaged strands and assays described here provide a novel approach to identify and dissect specific regions of the UvrA and UvrB proteins that are in closest proximity to the site of DNA damage during NER. Experiments utilizing tandem liquid chromatography-mass spectrometry are currently in progress to further characterize the specific peptide regions of the proteins that are cross-linked.