Sequence-specific interactions of UvrABC endonuclease with psoralen interstrand cross-links.

The nature of the Uvr protein-DNA complexes formed on psoralen-DNA interstrand cross-links was analyzed by DNase I footprinting and correlated with the incision efficiency of the UvrABC endonuclease on the cross-links of different DNA sequences. Our results indicate that the repair specificity is dependent on the DNA sequence and the psoralen orientation in the cross-link. On the strand that will be cut, a 30-nucleotide long UvrAB footprint with a DNase I hypersensitive site at the 11th nucleotide 5' to the lesion was observed and subsequently rearranged to a 22-nucleotide long UvrB-lesion footprint. On the strand that will not be cut, the UvrAB-lesion footprint had no 5' DNase I hypersensitive site and did not form the UvrB-lesion footprint. Although UvrABC incision requires the formation of UvrB-lesion complex on the strand which will be cut, the affinities of these complexes do not correlate with the incision efficiencies, suggesting that the overall reaction can be driven forward by a favorable next step such as UvrC incision. A study of the time-dependent interconversion of UvrAB-lesion complex to UvrB-lesion complex on a cross-link revealed a secondary recognition of the UvrB-lesion complex by UvrA2(B) proteins in vitro.

The nature of the Uvr protein-DNA complexes formed on psoralen-DNA interstrand cross-links was analyzed by DNase I footprinting and correlated with the incision efficiency of the UvrABC endonuclease on the cross-links of different DNA sequences. Our results indicate that the repair specificity is dependent on the DNA sequence and the psoralen orientation in the cross-link. On the strand that will be cut, a 30-nucleotide long UvrAB footprint with a DNase I hypersensitive site at the 11th nucleotide 5' to the lesion was observed and subsequently rearranged to a 22-nucleotide long UvrB-lesion footprint. On the strand that will not be cut, the UvrAB-lesion footprint had no 5' DNase I hypersensitive site and did not form the UvrB-lesion footprint. Although UvrABC incision requires the formation of UvrB-lesion complex on the strand which will be cut, the affinities of these complexes do not correlate with the incision efficiencies, suggesting that the overall reaction can be driven forward by a favorable next step such as UvrC incision. A

study of the time-dependent interconversion of UvrAB-lesion complex to UvrB-lesion complex on a cross-link revealed a secondary recognition of the UvrB-lesion complex by UvrA2(B) proteins in vitro.
Nucleotide excision repair is a major pathway in the removal of bulky DNA lesions. In Escherichia coli, this pathway begins with the UvrABC endonuclease which is a multienzyme complex containing UvrA, UvrB, and UvrC proteins as subunits (for reviews, see Grossman and Yeung (1990), Van Houten (1990), and Sancar and Sancar (1988). This endonuclease has a broad range of lesion specificity, including covalent adducts of mitomycin A, cisplatin, psoralen, pyrimidine dimers, 6,4-photoproducts, N-acetoxy-2-acetylaminofluorene, and noncovalent DNA intercalators like ditercalinium (for review, see Van Houten (1990)). The enzyme makes two nicks in the strand containing a lesion: one at the eighth or ninth phosphodiester moiety 5' to the modified nucleotide (nt); the other at the third or fourth phosphodiester moiety 3' to the modified nt (Sancar and Rupp, 1983;Yeung et al., 1983;Van Houten et al., 1986;Jones and Yeung, 1988). * This work was supported, in part, by National Science Foundation Grant DMB88-02091 (to A. T. Y.), National Science Foundation Institutional Instrumentation Grant DIR-8812108, National Institutes of Health Institutional Grants CA06927 and RR05539 (to Fox Chase Cancer Center), an appropriation from the Commonwealth of Pennsylvania, and a grant from Glenmede Trust to the Fox Chase Cancer Center. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 2488, Fax: 215-728-3574. $To whom correspondence should be addressed. Tel.: 215-728-A brief summary of the mechanism of the UvrABC endonuclease is as follows. UvrA is a lesion-specific DNA binding protein. It has a greater affinity for damaged DNA than for undamaged DNA. UvrA dimerizes (Oh and Grossman, 1989) in solution, to form a UvrAz complex. This complex associates with UvrB to form a UvrAnB complex (Orren and Sancar, 1989). Oh et al. proposed that this UvrApB complex binds to the DNA duplex and translocates along the DNA to the lesion by using the 5' to 3' helicase DNA unwinding activity of the UvrAzB complex Grossman, 1987, 1989). Upon encountering a lesion, this forms a UvrAzB-lesion complex. According to one model, UvrAz dissociates from the UvrAzBlesion DNA complex to form a UvrB-lesion complex (Orren and Sancar, 1990). The DNA in this complex is kinked by 127" (Shi et al., 1992). The UvrC protein may bind to the UvrAzB-lesion complex or to the UvrB-lesion complex to result in one 5' and one 3' incision on the damaged strand (Visse et al., 1992). From site-directed mutagenesis studies on UvrB and UvrC, Lin et al. (1992) suggested that UvrC is responsible for the 5' incision, but UvrB is responsible for the 3' incision. This difference in protein requirement for the two incision events may further increase the complexity of the sequence specificity of the UvrABC endonuclease.
In this report, we have examined the sequence-specific interactions of UvrAB proteins with psoralen interstrand cross-links. Psoralens are naturally occurring linear furocoumarins. 4,5',8-Trimethylpsoralen (TMP)' is a good example of this class of compounds. In the presence of UV light (360nm wave length), a double bond of either the furan ring or the pyrone ring of the TMP molecule (intercalates at a 5' TA or 5' AT site) photoreacts with the 5,g-double bond of a thymine residue to form a cyclobutane adducts (Song and Tapley, 1979). The adducts are called furan-side monoadduct or pyrone-side monoadduct, respectively. Only a furan-side monoadduct can undergo further photochemical reaction with the double bond of the thymine moiety in the opposite strand to produce an interstrand cross-link (Cimino et al., 1985). The UvrABC endonuclease incision can be on the furan-side strand or on the pyrone-side strand of the cross-link (Jones and Yeung, 1988;Van Houten et al., 1986). The choice of strand for repair depends on the composition of the bases flanking the cross-link (Jones and Yeung, 1990). In this report, we extend the analysis of this repair specificity to the level of the protein-DNA complex formed in the reaction by examining how different interstrand cross-links interact with UvrA and UvrB proteins in a sequence-dependent manner and how these interactions affect the repair efficiency.

486
UvrABC Cross-link Repair Specificity DNase I footprinting, we have adopted this method for our current study of site-specific cross-links with Uvr proteins (Van Houten et al., 1987;Bertrand-Burggraf et al., 1991;Munn and Rupp, 1991). Understanding the sequence-specific interactions of UvrABC proteins with TMP cross-links may provide some insights to the mechanism of DNA lesion recognition and repair. The results presented in this work show the relationship between the extent of protein-DNA complex formation and the sequences flanking the cross-link in the DNA adduct.

EXPERIMENTAL PROCEDURES
Preparation and Isolation of Specific Cross-links-The sequence of the 121-bp fragment is shown in Fig. 1. The preparation and isolation of a set of site-specific cross-links is the same as previously reported for 168-bp fragment. 88 pg of 5' 32P-labeled 121-bp fragment ( Fig. 1) containing lac UV5 po region was used as the starting material (Jones and Yeung, 1990). The 3' labeling of the cross-links on the 117-mer strand was done using the Klenow fragment of DNA polymerase I and [m3*P]dATP as described by Sambrook et al. (1989).
UvrABC Endonuclease Incision Reactions-About 15 fmol of endlabeled DNA was incised by 1 pmol of UvrA, 1 pmol of UvrB, and 1 pmol of UvrC (Yeung et al., 1986a) in a volume of 150 p1 of UvrABC reaction buffer (85 mM KC1, 50 mM Tris-HC1, pH 7.5, 5 mM dithiothreitol, 10 mM MgC12, and 2 mM ATP). After 45 min of incubation at 37 "C, the reactions were terminated with the addition of 10 pl of 0.5 M EDTA, pH 8.0. After another 10 min of incubation at 37 'C, the samples were ethanol-precipitated and resolved on a 10% polyacrylamide, 7 M urea DNA sequencing gel at 49 "C. The autoradiograms of the gel were quantified using an AMBIS two-dimensional densitometer (AMBIS Inc., San Diego, CA) from which the percentage efficiencies of the UvrABC endonuclease 5' or 3' cutting reactions were calculated.
DNase I Footprinting Reactions-About 15 fmol (75 pM) of crosslinked DNA was incubated with UvrA or UvrA + UvrB (concentrations as indicated in figure legends) in 200 p1 of UvrABC reaction buffer at 37 "C. After 15 min of incubation, 150 ng of pBR322 plasmid DNA was added. After 5 more min of incubation, the reaction mixture was cooled to 24 "C. Two p1 of CaClz (100 mM) and 2 pl of DNase I (15 ng) were added and the mixture was incubated at 24 "C for 2 min before being terminated by the addition of a 200-pl mixture of 20 mM EDTA, 70 mM sodium acetate, pH 5.2,l pg of calf thymus DNA, and 1% SDS. The DNA was ethanol-precipitated. To break the psoralen-DNA linkages, the DNA pellet was dissolved in 1 M piperidine and incubated at 90 "C for 30 min in a microcentrifuge tube tightly sealed with Teflon tape. This alkali-reversed and DNase I-hydrolyzed sample was analyzed on a 10% polyacrylamide, 7 M urea DNA sequencing gel.
Qwntitation of Footprints-The autoradiograms obtained from these experiments were quantified using the AMBIS densitometer. At least three bands that did not change during the footprinting of the UvrB protein binding were used to normalize the bands in each lane. The value obtained for the UvrAB footprints in each lane was divided by the value obtained from the corresponding control lane that did not contain UvrA or UvrAB. The fraction so obtained was subtracted from 1 and the resulting value represents the fraction of DNA bound to the UvrA or UvrAB proteins at the specified concentration of proteins ("the fraction bound value") and was plotted against the UvrA protein concentration. The value corresponding to 50% saturation represents the K d value of the protein-DNA complex under the reaction conditions used (Snowden and Van Houten, 1991;Munn and Rupp, 1991;Brenowitz et al., 1986). Since for most of the cross-links the formation of Uvr protein-DNA concentration reached a maximum at low UvrA concentration, the (&) values represent the maximum approximate value for cross-links 12, 14, 17, 18 and 30. Time-dependent UvrAB Assembly on a Cross-link-About 380 fmol of cross-link 19 (0.4 nM) was incubated at 37 "C with UvrA + UvrB (2.2 pmol each, 2.4 nM) in 900 pl of UvrABC reaction buffer. 50-pl aliquots were withdrawn after 0, 5, 15, 30, and 60 min and added to tubes containing 150 ng of pBR322 DNA, 10 ng of DNase I, and 2 pl of 100 mM CaC12. After 2 min of incubation at 20 "C, the reactions were stopped with EDTA, ethanol-precipitated, and analyzed as described above. The fraction of DNA bound to UvrB protein ("fraction bound values") were calculated as described above. The concentration of the DNA bound to UvrB complex was calculated as 0.4 nM X fraction bound value. The concentration of the free DNA (Sf) present in the reaction mixture at any given time was calculated as  (Orren and Sancar 1990). Alternatively, the slope of a second order kinetic plot of log (Sf/UvrB,) versus time can be used to derive the rate constant for UvrB complex formation.

Isolation and Characterization of TMP Interstrand Cross-
links-Psoralen (TMP) can react with the DNA duplex shown in Fig. 1 to form 18 possible cross-links. The positions of the psoralen adduction in the cross-links were identified (data not shown) by the sizes of the fragments generated by the digestion of the 5"labeled cross-links with the 3' to 5' exonuclease activity of the T4 DNA polymerase (Sage and Moustacchi, 1987), followed by alkali reversal of the cross-link (Yeung et al., 1988b;Jones and Yeung, 1990). The mobility of cross-linked DNA on a sequencing gel depends on the position of the psoralen adduct in the DNA duplex ( Fig. 2 A , lanes 1-6) and therefore the cross-links can be separated on a sequencing gel. The duplexes with the crosslinks in the middle of the restriction fragment showed the least mobility and the cross-links in the end showed the greatest mobility. The even numbered cross-links were prepared from the DNA duplex in which the 117-mer contained the monoadducts, the odd numbered cross-links were prepared from the DNA duplex in which the 121-mer contained the monoadducts.

Incision of TMP Cross-links by UurABC Endonuclease-
The UvrABC 5' cuts for each cross-link were identified by 5' labeling one of the strands of each cross-link. For example, in Fig. 2 A , lanes [7][8][9][10][11][12] show positions of the UvrABC cuts for cross-links 20,30,18,14,12, and 10, respectively. The pyroneside cuts for cross-link 20 and 30 are shown in Fig. 2B (lanes 1 and 2). Weak incision on the pyrone-side of cross-link 30 is not seen in Fig 2B ( l a n e 2 ) but visible in Fig. 2A ( l a n e 8).
The incised sites on the cross-links were identified on a 10% acrylamide DNA sequencing gel by comparing the size of the labeled fragments generated by UvrABC reaction with the bands in the reference DNA sequencing ladder.
The cross-links are cut at the ninth phosphodiester moiety 5' to the lesion and at the second or third phosphodiester A UvrABC Cross-link Repair Specificity UvrABC cutting of the 117-bp cross-links. Panel A, the autoradiogram of a 10% acrylamide DNA sequencing gel reeolving the UvrABC cut fragments containing 5' 32P-label on the bottom strand of cross-linked DNA. Lanes 1-6 are the control lanes for cross-links 20,30,18,14,12, and 10, respectively, without UvrABC. G, G + A, T, and C chemical sequencing lanes are as indicated (Maxam and Gilbert, 1980;Rubin and Schmidt 1980;Yeung et al. , 1988~). Lanes 7-12, cross-link samples (0.1 nM) in lunes 1-6 treated with 6 nM of UvrABC at 37 "C for 45 min as described under "Experimental Procedures." The pyrone cuts produced bands of sizes greater than 117 nt in the gel. The position of the modified thymines are circled and UvrABC furan-side cuts are shown as breaks (//) between nt corresponding to each crosslink. For cross-link 10, the position of the cut site cannot be assigned, since the bands move close to the top of the gel. Panel B, DNA sequencing gel resolving the UvrABC cut fragments from cross-linked DNA containing 5' 3ZP-label on the top strand. Lanes 1 and 2 are pyrone-side strands of cross-links 20 and 30, respectively, treated with 6 nM of UvrABC as described. Weak pyrone side incision (<4%) is not visible in this figure. moiety 3' to the lesion (data not shown). When UvrABC makes two cuts on the labeled furan-side strand, it produces a single-stranded fragment, smaller than 117 nt as shown in Fig. 2A. When the two cuts are on the pyrone-side strand, it produces a fragment of size larger than 117 nt (117 +12 nt) as shown in Panel A (pyrone cuts). When only one of the two UvrABC cuts is made on the furan-side or pyrone-side, it is called an uncoupled cut and is seen below the uncut crosslink DNA (Fig. 2 A ) . A UvrABC cut fragment contains a 3'-OH group but a chemical sequencing ladder fragment contains a 3' PO:group. Therefore, the 5' labeled UvrABC cut fragment moves 0.5 nt more slowly than the corresponding size sequencing fragment.
The efficiencies of repair (3' or 5' side cut) were determined by quantifying the cut bands with respect to the total counts in each lane (Table I). The UvrABC incision experiments were done at 6 nM of UvrABC proteins, which gave the optimum cutting value for most of the cross-links except for cross-links 29 and 19, which showed improved cutting effi-ciencies when UvrABC concentration was lowered from 6 nM to 2.4 nM. It appears that cross-links which are better UvrABC substrates are more easily inhibited by as little as 6 nM of UvrABC, while others are not. The cross-links studied in this work contain the psoralen adduct in different positions within the duplex, therefore the main difference among them is in their flanking sequences. The repair efficiencies among these cross-links varied from 0 to 34%. The cross-links 14, 18, and 30 showed a preference for the furan-linked strand (117-mer strand). The converse is true for cross-link 20 which preferred repair on the pyrone-linked strand (121-mer strand). Some of the cross-links studied in this work have identical sequences flanking the cross-link but with the adducted psoralen molecule in opposite orientations at the 5' TA or 5' AT site. These pairs of cross-links which differ only in the psoralen orientation are known as orientation isomers. The pairs compared are cross-links 11 and 12,17 and 18,19 and 20, and 29 and 30. The UvrABC cutting efficiencies on the same DNA strand for these isomers may be the same or differ by as much 488 UurABC Cross-link Repair Specificity UurABC endonuclease repair efficiencies of cross-links and the UurAB apparent equilibrium dissociation constants for binding to different cross-links 5'4abeled site-specific cross-links were treated with UvrABC proteins at 37 "C for 45 min as described under "Experimental Procedures." The UvrABC treated samples were analyzed in a 6% polyacrylamide-7 M urea denaturing gel. The autoradiograms of the gel were quantified using the AMBIS two-dimensional densitometer. From these quantitations the percent efficiencies of the UvrABC 3' or 5' cut at cross-links were calculated. The percent uncoupled cut values were determined by quantitating the band migrating close to the uncut DNA in Fig. 3A (lanes 8 and 11) and expressing as a fraction of the total DNA. as 34% (Table I). These results suggest that repair specificity depends both on the sequence flanking the cross-link and the psoralen orientation within a cross-link.

Cross-link
UvrAB Footprint Intensity on Cross-links Does Not Correlate with Cutting Efficiency-UvrABC endonuclease mechanism consists of several reaction steps involving different protein-DNA interactions. Previous work by others have indicated that the nature of the Uvr protein-DNA complexes so formed as UvrA2-DNA, UvrAZB-DNA, and UvrB-DNA complexes each with a distinct footprint (Visse et al. 1992;Van Houten et al., 1987;Bertrand-Burggraf et al., 1991;Orren and Sancar, 1989;Mazur and Grossman, 1991;Munn and Rupp, 1991). Because the UvrAB protein-DNA complexes are not completely stable under gel-shift assay conditions (Visse et al., 1992); we have used the more reliable DNase I footprinting technique as a tool to quantity these Uvr protein interactions on different psoralen cross-links.
Cross-links can be bound with equal affinity to UvrAB proteins and yet have different repair efficiencies. Fig. 3 shows the furan-side and the pyrone-side footprint on two orientation isomers, 18 and 17, respectively (Fig. 3, A and B). Incubation of cross-links 18 and 17 with increasing amounts of UvrA resulted in gradual UvrA dimer binding on these cross-links as shown by DNase I protection on the strand which will be cut by UvrABC. A clear DNase I protection resulted at 54 and 18 nM of UvrA for cross-links 18 and 17, respectively (Fig. 3, A , lane 5, and B, lane 4 ) . This 32-nt long * M. Ramaswamy and A. T. Yeung, unpublished results. is DNase I hydrolysis of control sequence without cross-link. The 5' and 3' hypersensitive sites of the UvrB-cross-link footprint are indicated as H3 and H5, respectively. The exposed DNase I-sensitive phosphodiester bond in the middle of the footprint is shown as H4.
The position of the cross-link is one base below the hypersensitive site H4. The X and Y regions indicated are the 5' half and 3' half of the footprint with respect to the exposed phosphodiester moiety 5' to the psoralen modified thymine residue. The region Z shows the footprint of the UvrAAB) recognition of the UvrB-lesion complex. UvrA dimer footprint is similar to the one formed on a HMT monoadduct (Van Houten et al., 1987). We have assigned this footprint as corresponding to UvrA2-DNA complex. This UvrA2-DNA complex footprint also shows a DNase I hypersensitive site 18 nt 5' to the modified thymine.
Similarly, analysis of the footprint of 2-54 nM of UvrA + 20 nM of UvrB on cross-links 18 and 17 (Fig. 3, A, lanes 6-9, and B, lanes 5 and 6) showed that UvrB-lesion complexes were formed at 2 nM of UvrA + 20 nM of UvrB (Fig. 3A, lane   6, and B, lane 5 ) . These two 20-nt long footprints showed the key feature of the UvrB-lesion footprint as reported by others for HMT monoadduct and cross-link (Bertrand-Burggraf et 489 al., 1991;Munn and Rupp, 1991): a strong DNase I hypersensitive site was observed at the 11th and 12th nt 5' to the lesion. The other 5' DNase I hypersensitive site at 18 nt from the lesion, characteristic of UvrAz-DNA complex, disappeared in the UvrB-lesion complex. In addition, the footprints showed that the nucleotide next to the modified thymine was not protected from DNase I in this complex. This DNase I sensitivity at the base next to a modified thymine was also seen in other cross-link footprints and suggested that the adjacent bases next to the lesion were not protected by UvrB protein in the UvrB-lesion complex and were accessible from the minor groove. A weakly protected region, 2, in these footprints is due to nonspecific UvrAz(B) binding to UvrBlesion complex and it will be discussed below. Although we have not assigned the nature of the complex independently by other techniques such as gel-shift and supershift assays, we have tentatively assigned this complex as UvrB complex based on the footprint characteristics.
The DNase I protected or hypersensitive band intensities in the autoradiograms of the gels were quantified as described under "Experimental Procedures." From these values, UvrA lesion binding curves were constructed (data not shown). From the UvrA lesion binding curves for cross-link 17 and 18, we have calculated the apparent dissociation constants of the UvrAz-DNA complex as the concentration of UvrA corresponding to 50% saturation of DNA lesion sites: maximum Kd = 8-12 nM for cross-link 18; maximum Kd = 6-7 nM for cross-link 17 ( Table I).
The UvrA-dependent UvrB binding curve for cross-link 18 (data not shown) was constructed from the DNase I hypersensitive sites of cross-link 18 footprint autoradiogram. From this, we have estimated the apparent equilibrium dissociation constant for UvrA-dependent UvrB binding on DNA as maximum Kd = 1-1.6 nM (for cross-link 18) and an approximation of the Kd of cross-link 17 as 0.5 nM (Table I). Therefore the affinities of UvrAzB or UvrB complexes for cross-link are 5-8 times greater than that of the UvrAz complex. This analysis also showed that for a pair of orientation isomers (17 and 18), the affinities of the UvrB complexes formed on the cut strand are similar (KO = 1-2 X lo9) irrespective of whether the UvrABC incision will be on the furan-or pyrone-linked strand. Although the binding of UvrA and UvrB proteins to these two cross-links on the same strand were similar, the UvrABC incision efficiency for these cross-links were different. It appears that the differences in the subsequent binding and cutting by UvrC after UvrAB-lesion complex formation must be an additional parameter controlling the overall incision efficiency of a cross-link. Cut-Fig. 4, A and B, shows the difference in footprints of the UvrAB-DNA complex formed on the two strands of cross-link 20 in which only the top strand will be cut by UvrABC. Incubation of this cross-link with increasing concentrations of UvrA (0-13.5 nM, Fig. 4A, lanes 2-6) did not show a UvrA-specific DNase I footprint on the bottom strand (117-mer strand, furan-linked strand). This absence of UvrA footprint is expected because at least 27 nM of UvrA was required to produce the UvrA footprint for cross-link 18. Under the same conditions but in the presence of only 3 nM UvrB, a 30-nt footprint emerged on this uncut strand even at a low concentration of UvrA (0.8 nM). However, this footprint did not contain a DNase I hypersensitive site at the 11th phosphodiester moiety 5' to the lesion and the footprint did not change in size with increasing concentrations of UvrA or UvrB. This footprint also contained a DNase I-sensitive site at the 28th phosphodiester moiety 5' to the lesion. Because the appearance of this footprint on this strand required both UvrA and UvrB proteins, this footprint corresponds to a UvrApB-lesion complex. The absence of repair on this strand may be related to failure to form the lesion-specific UvrB complex.

UvrB-Lesion-specific Footprint Zs Formed Only on the Strand That Zs
On the top strand of cross-link 20, the strand which will be cut by UvrABC endonuclease in the cross-link (121-mer, pyrone-linked strand), concentrations up to 13 nM of UvrA (Fig. 4B, lanes 1-6) did not result in a characteristic UvrA-DNA footprint, However, at 3 nM of UvrB and 0.75 nM of UvrA, a weak UvrAzB footprint with the characteristic 5' DNase I hypersensitive site was formed (Fig. 4B, lane 7). As the concentration of UvrA was increased from 0.75 to 13.5 nM, the footprint shrank from 34 to 23 nt long (Fig. 4B, lane  12). Thus the lesion-specific UvrB footprint with 5'-hypersensitive site at the 11th phosphodiester moiety is formed only on the strand which will be cut by the UvrABC endonuclease. These footprinting experiments on two strands of the same cross-link illustrate the difference in the nature of the protein-DNA interactions for the strand to be cut and the strand not to be cut.
From the binding curve constructed from the densitometer scans of the autoradiogram of this experiment, we have estimated the equilibrium dissociation constant of UvrB-lesion complex (&) as 0.75 nM ( Table I). The KO values of crosslinks 20, 18, and 17 are comparable (KO = 1-2 x lo9).
To verify whether the formation of a UvrB footprint, characterized by a hypersensitive site at the eleventh phosphodiester moiety 5' to the lesion, is a prerequisite for UvrABC cutting, we have footprinted the furan-side of the cross-links 12, 14, and 30 with UvrA and UvrB proteins. The footprint on cross-link 12 is an example of a lesion positioned only 10 nt away from the 3' termini. The UvrB complex formed on this lesion protected only 6 nt on the 3' side and 11 nt on the 5' side (data provided to the reviewers but not shown). We estimated the maximum Kd of UvrB-lesion complex for crosslink 12 to be <2 nM. Similarly the cross-links 14 and 30 produced UvrB footprints with the 5' hypersensitive site on the UvrABC cut strand (furan-linked, 117-mer strand, data not shown). Therefore, it seems likely that most of the crosslinks which will be incised by UvrABC endonuclease will form UvrB footprints with a 5' DNase I hypersensitive site on the strand to be cut.

Kinetics of UvrB-Lesion Complex Formation, and Detection of Secondary Recognition of UvrB-Lesion Complex by UvrAZB
Complex in Vitro-The kinetics of UvrB-lesion complex formation were investigated by DNase I footprinting using crosslink 19. The densitometer tracings of the footprints (autoradiograms not shown) obtained for various aliquots withdrawn after incubation times of 0, 5, 10, 15, 25, 35, and 60 min are shown in Fig. 5. The tracings are divided into two segments (I and ZI) to aid analysis. The footprint obtained at 5 min showed some protection of segment I. By 9 min, a portion of the bands in this protected region (43 nt long) reappears on the 3' side (Fig. 5, segment I, peaks 2-8, 10-min time point) with the simultaneous appearance of the 5' DNase I hypersensitive sites at the 11th and 12th phosphodiester moieties 5' to the lesion (Fig. 5, segment IZ, peaks 12 and 13, 10- used was low. No additional UvrA was added after the formation of the UvrB-lesion complex, therefore the natural progression of the protein complexes was not disturbed. The amount of UvrB-lesion complexes formed was plotted against the time of incubation (Fig. 6A). From the initial slope, the rate constant for the loading of UvrB was calculated using UvrA and DNA concentrations as the limiting factors. This rate constant also agreed with the second order rate constant derived from the slope of the plot of log( [free DNA] It is assumed that the fraction of UvrA used in substrate recognition is catalytically regenerated as shown in the literature (Orren and Sancar, 1989). The rate constant ($, = 7.6 X lo4 M-' s-' for UvrB loading onto cross-links) obtained by both of these methods, agreed with the $" values reported for UvrB loading on pyrimidine dimers (Orren and Sancar, 1989).

Sequence-dependent Variation in DNA Interstrand Crosslink Structure Can Affect
Repair-The structure of psoralen interstrand cross-link has very interesting structural features. NMR spectroscopy experiments combined with molecular modeling studies revealed that the psoralen interstrand crosslink produces a bend toward the major groove (Tomic et al., 1987). Gel electrophoretic analysis of psoralen cross-linked duplexes containing mismatched bases revealed that the two sides of the helix flanking the cross-link contain domains of different stabilities (Kumaresan et aL, 1992). It is likely that these features of psoralen cross-links flanked by different sequences can adopt different DNA conformations in a sequence-dependent manner. Similar observations for the sequence-dependent variations in the conformations of other DNA-bulky adducts are known. Examples include DNA adducts of (+)-anti-benzo[a]pyrene diol epoxide (Roche et al., 1991;Rodriguez and Loechler, 1993) and C-%guanine adducts of 2-acetyl-aminofluorene (Veaute and Fuchs, 1991;Belguise-Valladier and Fuchs, 1991). The sequence-specific variations in psoralen cross-link conformation may influence the efficiencies of formation of various UvrABC protein-DNA complexes, like UvrA, UvrAl3, UvrB, UvrB(AB) and Uvr(A)BC complexes. The sequence-dependent variations in DNA conformation and their role in DNA-protein interactions has been recently reviewed (Travers, 1989).
Three major points can be inferred from our work discussed in this paper. (i) We have observed that UvrB-lesion footprint is formed only on the strand which is to be cut, but not on the strand to be cut weakly or not cut. The strand-specificity in the cross-link repair may partly stem from the competition between the lesions on the two strands of a cross-linked DNA duplex for Uvr protein-DNA complex formation. (ii) Although we have observed the formation of a footprint corresponding to UvrB complex in all the cross-links on the strand that is to be cut, the strength of UvrB complex binding did not correlate with the UvrABC incision efficiency. This is in accord with the observation that repair efficiency of N-2acetylaminofluorene modified DNA substrates at different sites of a DNA duplex could not be accounted for by the binding of the UvrAB complex alone (Seeberg and Fuchs, 1990). But this is in contrast with the observation made by Visse et al. (1992) who showed, for a cisplatin substrate present in a defined sequence, that cutting and UvrAB binding can be correlated. (iii) The time-dependent secondary recognition of UvrB-DNA complex by UvrA or UvrAB can lead to UvrB(AB) DNA complex. This may interfere with UvrC binding and cutting. Similarly with increasing concentration of UvrA protein, in the presence of fixed concentration of UvrB protein, Visse et al. (1992) observed that, in addition to the formation of UvrB-DNA complex, another nonspecific UvrAB-DNA complex was formed in their reaction. More From the initial slope of the plot (between time 5 and 15 min), an apparent rate constant for UvrB loading was calculated. Using this value, a true second order reaction rate for the loading of UvrB onto a cross-link was calculated by assuming UvrA and DNA as limiting factors as described by Orren and Sancar (1990). Panel B, a second order kinetic plot of log (Sf/ UvrBJ as a function of time. From the slope, a rate constant k for UvrB loading was derived. indirectly, others have explained the UvrA concentration dependent change in the footprint of UvrB protein-DNA complex as due to a reversal of UvrB-DNA complex to a UvrAB complex (Bertrand-Burggraf et al., 1991); alternatively UvrA can compete with UvrAB for binding to DNA, which results in UvrA-DNA complex (Snowden and Van Houten, 1991). Our approach of using low concentrations of UvrAB proteins in our experiment has allowed us to observe the sequential formation of UvrB complex footprint followed by the appearance of a secondary recognition of the UvrB-lesion complex by another UvrAB complex.