Both ATPase sites of Escherichia coli UvrA have functional roles in nucleotide excision repair.

The roles of the two tandemly arranged putative ATP binding sites of Escherichia coli UvrA in UvrABC endonuclease-mediated excision repair were analyzed by site-directed mutagenesis and biochemical characterization of the representative mutant proteins. Evidence is presented that UvrA has two functional ATPase sites which coincide with the putative ATP binding motifs predicted from its amino acid sequence. The individual ATPase sites can independently hydrolyze ATP. The C-terminal ATPase site has a higher affinity for ATP than the N-terminal site. The invariable lysine residues at the ends of the glycine-rich loops of the consensus Walker type "A" motifs are indispensable for ATP hydrolysis. However, the mutations at these lysine residues do not significantly affect ATP binding. UvrA, with bound ATP, forms the most favored conformation for DNA binding. The initial binding of UvrA to DNA is chiefly at the undamaged sites. In contrast to the wild type UvrA, the ATPase site mutants bind equally to damaged and undamaged sites. Dissociation of tightly bound nucleoprotein complexes from the undamaged sites requires hydrolysis of ATP by the C-terminal ATPase site of UvrA. Thus, both ATP binding and hydrolysis are required for the damage recognition step enabling UvrA to discriminate between damaged and undamaged sites on DNA.

The surveillance of the integrity of genetic material by organisms is monitored by DNA repair enzymes. One of the best characterized repair processes in Escherichia coli is nucleotide excision repair. The nucleotide excision repair enzyme UvrABC endonuclease is unique in its dual incision pattern that is seven nucleotides 5' and three to four nucleotides 3' to a UV-damaged site (Yeung et al., 1983;Sancar and Rupp, 1983). This enzyme complex exhibits broad specificity by recognizing lesions that may cause a common secondary structure, formed in the major or minor groove such as bulky nucleotide adducts, UV-mimetic lesions, or intraor interstrand cross-links Pu et al., 1989;Van Houten, 1990).
The UvrA protein is suggested to be the damage recognition subunit of the endonuclease. UvrA possesses a DNA-independent ATPase/GTPase (Seeberg and Steinum, 1982;. Its binding affinity to DNA is enhanced by ATP and DNA damage (Seeberg and Steinum, 1982;Yeung et al., 1986a). A monomer-dimer equilibrium of UvrA protein is established in the presence of ATP Orren andSancar, 1989, Mazur and, and it is shifted toward dimer formation in the presence of ATPyS' or ADP . The UvrAB protein complex manifests an enhanced ATPase activity dependent on DNA , which drives a helicase activity Grossman, 1987, 1989). The helicase activity of the UvrAB complex is limited to a short stretch of about a 22-base-paired region that requires either ATP or dATP hydrolysis for its activity Grossman, 1987,1989). The endonuclease activity of the UvrABC complex stably bound damaged DNA is driven by ATP binding and not its hydrolysis .
The predicted amino acid sequence of UvrA and UvrB proteins, two of the subunits of the UvrABC endonuclease, have sequence motifs common to many ATP binding proteins and ATPases Doolittle et al., 1986;Arikan et al., 1986;Backendorf et al., 1986;Walker et al., 1982;Fry et al., 1986;Higgins et al., 1988). Most of these ATPases contain two distinct motifs, an "A" type consensus of a hydrophobic stretch of /3-strand-GXXGXG (a flexible glycinerich loop)-KS/T (a-helix) (Fig. 1) and a "B" type consensus of hydrophobic stretch of (3-strand-DE/D (Gorbalenya and Koonin, 1990). It is suggested that the central flexible loop of the A motif binds the phosphoryl moiety of ATP possibly by forming a giant anion "hole" (Schulz, 1987). It is thought that the conserved aspartate residue of the B motif chelates Mg2+ or MgNTP (Fry et al., 1986). The UvrA protein is one of the few proteins to have two ATP binding motifs tandemly arranged in a linear sequence (Fig. 1). It is generally believed that internal duplication or exon shuffling during evolution may generate such tandem motifs (Doolittle et al., 1986;Dorit et al., 1990).
In this study, we have investigated the role(s) of the two nucleoside triphosphate binding motifs of UvrA in UvrABCmediated excision repair. It is our goal to understand whether the two motifs are functional and what roles they play in excision repair. This would serve as a useful prototype model to study proteins with two tandemly arranged ATP binding motifs. This question is approached by generating site-directed mutations at each motif as well as both motifs. The invariable lysine residue at the end of the glycine-rich loop in the A motif was mutated to an alanine, a polar glutamine, or to a conservative arginine residue (Fig. 2). Mutagenesis of the invariable lysine residue in proteins containing a single ATP binding motif in two of the repair proteins, namely UvrB and ' The abbreviations used are: ATP-yS, adenosine 5'-O-(thiotriphosphate); MOPS, 4-morpholinepropanesulfonic acid; DTT, dithiothreitol; SDS, sodium dodecyl sulfate; PAGE, polyacrylamide gel electrophoresis; kb, kilobase(s).

ATPase
Sites of UvrA RAD3, and other proteins such as adenylate kinase, led to the loss of their functionality Grossman, 1989, 1990;Sung et al., 1988;Reinstein et al., 1988). The present study suggests that both of the ATP binding motifs of UvrA are indispensable for its functioning and localizes the roles of ATP binding and hydrolysis in the pre-incision partial repair reaction.
Molecular Cloning-The uurA gene was excised from pLAC1, a derivative of pATY1103 . The largest BamHIIHindIII fragment of pLACl carrying the uurA gene was cloned into pTZ18R (Pharmacia LKB Biotechnology Inc.) at the HindIIIIBamHI site in the multiple cloning region t o generate pSST1. The FlIG region of pSSTl derived from the pTZ18R enabled this plasmid to be used in generating single-stranded uurA gene containing template for mutagenesis experiments. pSSTlO is a derivative of pSSTl with the uurA gene under control of the PL promoter, which is in turn regulated by the cI857 repressor on the same construct. pSSTl was digested with SacI for linearization. pHE6 (Milman, 1987) was digested with HaeIIIXmnI, and the largest fragment carrying the PL promoter and cI857 region was isolated. The isolated fragment of pSST1 linearized by SacI and the HaeII/XmnI fragment of pHE6 were first treated with mung bean nuclease (Pharmacia). The fragments were isolated again and digested with BamHI. The BamHIlmung bean nuclease-treated fragments derived from pSSTl and pHE6 were ligated to generate pSST10. Mutant uurA genes generated using the pSST1 construct were recloned into pSSTlO by digesting pSST1 mutant plasmids with BamHI/XmnI to generate the mutant uurA gene containing fragments and ligating it into the BamHIIXmnI fragment of pSSTlO deleted of uurA wild type gene but containing the PL promoter and cI857 repressor genes.
Preparation of Single-stranded DNA-The CJ236 strain of E. coli transformed with pSSTl was used to generate a uracil-incorporated single-stranded DNA(ssDNA) template. Cells from a minimal plate were grown a t 30 "C until midlog phase (A59s of 0.5 to 0.8) in 2 X YT media supplemented with 0.25 pg/ml uridine. The midlog phase cells were infected with M13K07 a t a multiplicity on infection of about 10, allowed to stand for 15 min at room temperature, and continued incubation at 30 "C for 1 h. Kanamycin was added to a final concentration of 75 pg/ml, and the culture continued shaking at 300 rpm for 6 to 12 h. The supernatant fraction of the culture was used to prepare the ssDNA in a standard method for phage DNA preparation (Sambrook et al., 1989). Site-directed Mutagenesis-Oligonucleotide primers 23 nucleotides long were synthesized on an Applied Biosystems 380A automated synthesizer by Scott Morrow (Dept. of Biochemistry, The Johns Hopkins Oligonucleotide Facility). The nucleotide changes are in bold letters: CGGGTTCTGGCGCATCCTCGCTC, K37A; CGGGT-TCGCTC, K37R;GTTCCGGTGCATCGACGCTGATT, K646A; CTCGACGCTGATT, K646R.
We have employed a modification of the procedure described by Kunkel et al. (1987) for oligonucleotide-directed mutagenesis.
The primer was first phosphorylated at the 5'-end by polynucleotide kinase and annealed with the uracil containing ssDNA template in an annealing buffer of 40 mM Tris-HC1, p H 7.5, 15 mM MgCl,, and 50 mM NaCl, in a volume of 14 pl. The primer-template mixture at a ratio of 1O:l in the annealing buffer was incubated at 60 "C for 10 min and allowed to cool to the room temperature. Second strand synthesis was continued in a 50-p1 reaction mixture of 20 mM Tris-HC1, pH 7.5, 10 mM MgCl,, 2 mM DTT, 4% (v/v) glycerol, 600 p M dNTPs (150 p~ each), 500 p~ ATP, 1.5 pg/ml SSB, and 2 pg of DNA pol 111 holoenzyme (a gift from Dr. K. Stephens and Dr. R. Mc-Macken, The Johns Hopkins University) or 3 units of T4 DNA polymerase (Pharmacia). The reaction was carried out at 30 "C for 30-60 min and heat-inactivated at 70 "C for 10 min. The ligation TCTGGCCAGTCCTCGCTC, K37Q; CGGGTTCTGGCCGCTCC-GTTCCGGTCAGTCGACGCTGATT, K646Q; GTTCCGGTCG-reaction was continued overnight a t 15 "C by the addition of 2 units of T4 DNA ligase (New England Biolabs). The ligation mixture was transformed into competent E. coli JM109 cells, and plasmids isolated from randomly picked colonies were sequenced using T 7 DNA polymerase (Pharmacia) by the Sanger dideoxy method (Sanger et al., 1977;Tabor and Richardson, 1987).
UV Suruiual-UV survivability of the UvrA mutant and wild type strains was determined by streak tests and the sensitivity range obtained by this procedure was used to perform UV survival experiments. The mutant and wild type uurA gene carrying pSSTl or pSSTlO and the control pTZ18R plasmids in an E.
coli MHlAA background were used in these experiments.
The streak tests were performed as follows. The overnight cultures were streaked across the 50 pg/ml carbenicillin containing 2 X YT plates. Shortwave (254 nm) UV lamp in the dark room was set at 12 microwatts cm-' to generate an exposure of 7.5 J m-' min-' at a distance of 41 cm. The UV exposure was increased from left to right perpendicular to the bacterial streaks by moving an opaque shield across the plate from right to left. The exposures were as indicated in Fig. 3A. At the end, the plates were immediately transferred to an incubator set at 30 "C and protected from laboratory light exposure in order to avoid photoreactivation.
The UV survival curves were executed as follows. The bacterial strains were grown up to midlog phase (A6g5 of 0.7-0.8); the cells were centrifuged, resuspended in M9 salts lacking a carbon source, and permitted to grow with shaking for another 1.5 h. The cells were diluted in M9 salts lacking a carbon source to yield approximately 3000 cells on a 2 X YT plate containing 50 pg/ml carbenicillin and further diluted for mock irradiation. The UV exposed cells at the exposures indicated in Fig. 3B were incubated in the dark at 30 "C. The colonies were counted after an overnight incubation.
Protein Purification-The mutant proteins K37A, K646A, and K37A K646A as well as wild type protein were purified from pSSTlO constructs transformed into MH1 AA strains by a modified procedure of Yeung et al. (1986b). Cells carrying the appropriate plasmid were grown at 30 "C in 2 X YT media containing 50 pg/ml carbenicillin until it reached an A595 of 0.7-0.8. The temperature was shifted to 42 "C to start induction of protein expression, and the cells were grown for another 1.5 to 2 h. The cells were immediately chilled and pelleted by centrifugation and resuspended in 0.1 M Tris-HC1, pH 7.5, and stored a t -80 "C. Twenty grams of cell pellet was used in each preparation. Cells were thawed in a buffer of 0.1 M Tris-HC1, p H 7.5, 12 mM EDTA, 2 mM DTT, 1 mM phenylmethylsulfonyl fluoride, and 1.5 mg/ml lysozyme at 4 "C. After 1.5-2 h of incubation, 5 M NaCl was added to a final concentration of 0.3 M and sonicated to further disrupt the cells. The crude extract was obtained after centrifugation and used in the subsequent steps of purification.
The chromatography steps essentially consisted of Affi-Gel Blue (Bio-Rad), phosphocellulose P-11 (Whatman), and single-stranded DNA (Bethesda Research Laboratories) columns as described previously by Yeung et al. (198613). However, single-stranded DNA column chromatography was modified as follows. The sample was applied onto the column and washed with a buffer of 0.1 M KP04, pH 7.5, 1 mM EDTA, 2 mM DTT, 15% (v/v) glycerol, and 0.1 M KC1. The column was washed again with a buffer of 50 mM K+-MOPS, pH 7.5, 1 mM EDTA, and 2 mM DTT (buffer C) and 0.1 M KC1 to remove the phosphate. Buffer C, containing 0.1 M KC1 and 5 mM ATP, was which bind ATP but not DNA. Washing with buffer C and 0.1 M KC1 used to wash the column again to remove any contaminating proteins was followed to remove excess ATP. The wild type UvrA protein was eluted with buffer C and 0.3 M KC1. The column material with bound mutant UvrA proteins was first washed with buffer C and 0.3 M KC1 and eluted with a linear gradient of 0.3 M to 0.6 M KC1 in buffer C. The salt concentration was lowered, and proteins were concentrated simultaneously by a two-step centrifugation in Centriprep 30 and Amicon 30 microconcentrator cartridges, respectively. The purified mutant and wild type proteins were judged to be greater than 90% pure by SDS-PAGE (Laemmli, 1970) on 10% polyacrylamide gels after staining with Coomassie Blue (Diezel et al., 1972).
ATPase Assays-The purified proteins were assayed in the absence of any DNA or in the presence of undamaged DNA (DNA) or UVirradiated DNA (uvDNA). Assay mixtures (20 pl) contain 50 mM K+-MOPS, pH 7.6, 100 mM KC1, 15 mM MgCl', 50 pg/ml bovine serum albumin, 2 mM DTT, 24 nM UvrA proteins, and 25 to 350/650 p M ATP (labeled with ["HIATP (ICN or Amersham, [3H]ATP at 1 Ci/ mmol ATP). pPYC3 plasmid DNA (Yeung et al., 1983) is used at 30 p~ to check the effect of DNA or uvDNA on the ATPase activity. UV-irradiated DNA was obtained by exposure to 720 J m-' generating approximately six pyrimidine dimers per kilobase pair of random DNA sequences. Reactions were initiated by adding ATP. At various time points, 05/11 aliquots of the reaction mixture were spotted onto polyethyleneimine-cellulose TLC plates (Brinkmann) prespotted with 0.5 p1 ATP/ADP (100 mM) markers. The TLC plates were developed with 1 M formic acid, 0.5 M LiC1. The dried spots corresponding to ATP and ADP were visualized using shortwave UV light (254 nm) and excised to determine the level of radioactivity in scintillation fluid (Bio-Safe NA, Research Products International). The initial rates were determined by linear regression analyses, and the kinetic parameters were calculated from Lineweaver-Burk plots.
ATP Binding Assay-The binding of ATP to the UvrA proteins was assayed by retention of ATPy"S employing a modification of the procedure described by Gonsky et al. (1990). The binding mixture (75 p l ) contained 10 mM K+-MOPS, pH 7.6, 100 mM KC1, 15 mM MgCI,, 50 pg/ml bovine serum albumin, 2 mM DTT, 1 mM EDTA, 5% (v/v) glycerol, 100 pg/ml pPYC3 plasmid DNA, and 3 p~ ATPyS (labeled with ATPy'"S a t 70 Ci/mmol), and 25 to 500 nM wild type and mutant UvrA proteins were added to initiate the reaction. The control reactions were assembled in a similar manner except for the inclusion of UvrA protein. The reactions were incubated a t 37 "C for 30 min. 20-4 portions of the reaction mixture were filtered through 0.45-pm nitrocellulose filters (HAWP025, Millipore) pre-equilibrated in the filter buffer (10 mM Kf-MOPS, pH 7.6, 100 mM KCI, 15 mM MgC1, and 1 mM EDTA). The filter was washed once with 100 p1 of filter buffer and dried for 10 min, and the level of ATPy'"S adsorbed onto the filters was determined by counting in scintillation fluid. The level of radioactivity retained by the UvrA proteins was corrected by subtraction of the background radioactivity without protein.
DNA Binding Assay-Nucleoprotein complexes were assembled in a 1OO-pl reaction volume consisting of 50 mM K+-MOPS, pH 7.6, 100 mM KCI, 15 mM MgCI,, 1 mM DTT, 50 pg/ml bovine serum albumin, 10% (v/v) glycerol, 150 ng ["HIpHEG plasmid DNA (Milman, 1987;Yeung et al., 1986a) equivalent to 57-fmol circles (specific activity = 8.0 X lo4 cpm/pg) and 2 mM ATP, ADP, or ATPyS. Whenever the reaction mixture contained ATP, an ATP regenerating system (4 mM phosphoenolpyruvate and 500 units/ml pyruvate kinase (Boehringer Mannheim)) was also included in the reaction. UV-irradiated ['HI DNA was prepared by exposure of DNA to 254 nm UV light to 680 J m-?, which was estimated to generate approximately 18 cyclobutane pyrimidine dimers per molecule of pHE6. The reaction was initiated by adding 1 to 25 nM UvrA protein and incubated to equilibrium a t :17 "C for 30 min. 3O-pl aliquots of the reaction mixture were filtered through 0.45-fim nitrocellulose filters (HAWP025, Millipore) preequilibrated in the filter buffer (50 mM K+-MOPS, p H 7.6, 100 mM KCI, and 15 mM MgCI2). Then t,he filter was washed once with 300 p1 of filter buffer and dried for 10 min, and the level of [,"H]DNA adsorbed onto the filters was determined by counting in scintillation fluid. The radioactivity bound to the filters was calculated from the average of three readings from the same experiment and in which a t least two independent experiments were performed. The average number of complexes formed was calculated by Poisson analysis (Yeung et al., 1986a;Grossman, 1989, 1990;Mazur and Grossman, 1991). The total number of complexes was determined by multiplying the average number of complexes with the amount of substrate (57-fmol circles).
Half-life Measurement of 2 X SSC-resistant Complexes-DNAprotein complexes were assembled under conditions described for filter binding assays using 2.5 nM wild type and mutant UvrA proteins and ATP as the nucleoside triphosphate in the presence of an ATP regenerating system. 33O-pl reactions were incubated a t 37 "C for 30 min. 30-pl aliquots of the reaction mixture were filtered and washed with 300 p1 of filter buffer (same as in the filter binding assay) to p l ) was diluted into 3 ml of cold (4 "C) 2 X SSC (300 mM NaCI, 30 determine the zero time point. The rest of the reaction mixture (300 InM sodium citrate (pH 7.0)). At various time intervals, 300.~1 aliquots were filtered through 0.45-pm nitrocellulose filters (HAWP025, Millipore) pre-equilibrated in 2 X SSC and washed twice with 300.~1 aliquots of cold 2 X SSC. Filters were dried, and DNA adsorbed to the filters was determined by counting in scintillation fluid. The halflives of the complexes were determined as described previously by Yeung et al. (1986a).

RESULTS
The presence of two tandemly arranged Walker type (1982) ATP binding motifs in the UvrA protein lends itself to analysis of the roles of each of the repeated functional sites (Fig.   1). Site-directed mutagenesis was focused on the invariable lysine residue at the end of the glycine-rich loop of the consensus "A" type motif of a single putative ATP binding site. Substitution of this lysine leads to effects such as loss of the ability of the protein to hydrolyze ATP, reduced affinity for the NTP (Reinstein et Parsonage et al., 1988, Seeley and Grossman, 1989, 1990, change in phenotype resulting in transformation (Sigal et al., 1986;Barbacid, 1987), or loss of the ability to induce transformation and tumorigenicity (Snyder et al., 1985). However, it is conceivable that, when a protein has two ATP binding sites, its duplication may be an artifact of evolution in which one is redundant in the function of the protein. On the other hand, both of these sites may have been selected during evolution to be retained as a minimal cooperative requirement for the protein's macromolecular phenotypes. The site-directed mutations of UvrA, shown in Fig. 2, were used to analyze these two possibilities.
The described mutagenesis technique provided a 10 to 80% yield of mutants in a plasmid construct pSSTl which was easily manipulated for further cloning and eventual expression of the respective mutant proteins (Fig. 4 ) . Further, there are no other mutations in the mutant uvrA gene(s) as the mutant(s) reverted back to wild type when the mutant segment of the DNA was replaced with the corresponding wild type segment of DNA in the plasmid constructs (data not shown). The yield of ssDNA template for the mutagenesis experiments using the M13K07 superinfection method was very poor when pSST10, an 8.1-kb plasmid with the phage origin was employed. However, this was overcome by using the 6.6-kb plasmid pSST1. It is possible that the phage capsid   Doolittle et al., 1986); MlUurA is the UvrA analog of Micrococcus luteus (Shiota and Nakayama, 1989); Factor Y(n') is a component of the replication-priming apparatus of E. coli and 6x174 (Nurse et al., 1990;Lee et al., 1990); RbsA is a ribose transport protein in E. coli (Bell et al., 1986); Pfmdr is the multiple drug resistance gene in Plasmodium falciparum (Foote et al., 1989); STE6 is the gene for the 0-factor pheromone export system of Saccharomyces cerevisiae (McGrath and Varshavsky, 1989); mmdrl and mmdrll are members of a mammalian multiple drug resistance gene family (Gros et al., 1986;; hmdrl is a human multiple drug resistance gene (Chen et al., 1986); CFTR is the human cystic fibrosis gene (Riordan et a/., 1989). imposes a limitation on the size of the ssDNA which can be tion of the host defect by the ATPase site mutations were as packaged.
follows: K37A, 2.7 J m-*; K37Q, 4.6 J m-'; K37R, 16.3 J m-*; Effect of the Mutations on UV Suruiual-Wild type and K646A, 1.4 J m-2; K646Q, K646R, K37A K646A, K37Q mutant uurA gene containing plasmid (pSST1 or pSST10) in K646Q, and K37R K646R, 1.3 J m-'. Both types of plasmids the AuurA background was studied by streak tests and UV (pSST1 and pSST10) carrying the mutant and wild type genes survival curves (Fig. 3, A and B ) . K37R was the only mutant of uurA yielded similar results despite the fact that the gene able to complement the chromosomal defect in uurA to gen-is expressed a t slightly higher levels in pSSTlO due to "leakerate a UV-resistant phenotype. The levels of resistance con-iness" of the temperature-sensitive repressor (cI857) regulated ferred by the other mutants were as follows; K37Q > K37A > PL promoter. The levels of UvrA protein in all of the strains K646A > K646Q, K646R, K37A K646A, K37Q, K646Q, and were found to be comparable (data not shown). From such K37R K646R. The estimated D:j7 values (the UV dose suffi-observations it is suggested that the presence of both invaricient to produce a single inactivating hit, In N/No of 1 = 0.37) able lysine residues at the ends of the glycine-rich loops of for wild type uurA gene complementation was 16 J m-2 while the A consensus Walker type motif of N T P binding proteins it was 1.3 J m-' for the host strain MHlAA with a chromo-present in the UvrA protein enables this subunit to confer soma1 deletion of the gene.

Sites
of UvrA 11399 clease. Furthermore, the ability of the K37R mutant to exhibit essentially similar levels of survival as the wild type uurA gene indicates that the two ATPase sites could also have different roles in the overall excision repair process.
In order to examine the biochemical mechanism of the repair deficiency of these mutations, representative mutant UvrA proteins were purified for further characterization (Fig.  5). The levels of UV resistance, sites of the mutations, and the amino acid change were considered in choosing the proper mutant protein. The K37A and K646A single mutants and the K37A K646A double mutant proteins were chosen to study since they represent an increasing level of UV sensitivity. Further, mutations were localized at each of the single site and both sites, and all had a similar lysine to alanine change in the ATP binding motif.
Expression of the Wild Type and Mutant UvrA Proteim-The analysis of the soluble fractions of the mutant and wild type uurA gene carrying pSSTlO/MHlAA strains suggest that the majority of the UvrA protein in each of the fractions does not form inclusion bodies due to overexpression (Fig. 4). The presence of higher levels of UvrA protein in pSSTlO/MHlAA than pSSTl/MHlAA and MHlAA (data not shown) even at 30 "C (before induction) suggests that the pL promoter is not tightly repressed ("leaky") by cI857.
Purification of Proteins-The mutant proteins were purified by essentially the same protocol as the wild type UvrA (Fig. 5). However, unlike the wild type protein, proteins were eluted from the ssDNA column using a higher salt gradient. This suggested that the mutant proteins are stably bound to ssDNA. Similarly, the mutant proteins were surprisingly more stably bound to double-stranded DNA (Fig. 7, Table 11). Western blotting of the purified wild type and mutant UvrA proteins using anti-UvrA and anti-UvrB antibodies confirmed that the mutant proteins could cross-react with the anti-UvrA raised against the wild type protein, and the protein preparations have no contaminating UvrB (data not shown). ATPase Activity-Both of the single site mutants retained ATPase activity while the double mutant was unable to hydrolyze ATP (Table I). These observations suggest a crucial role of the lysine residue in the mechanism of ATP hydrolysis. Further, it is assumed that the ATPase activities exhibited by the single mutants are derived from the nonmutated ATPase site(s). This assumption relies on the absence of any gross conformational changes in the mutant UvrA protein(s) which could affect the ATPase activity derived from the nonmutated site(s). The ability of the mutant proteins to bind ATP at levels similar to the wild type protein indicate that the conformation of the ATP binding pocket is not significantly altered by these ATPase site mutations (Fig. 6). Furthermore, while the double mutant completely lost its ability to hydrolyze ATP, the single mutants retained ATPase activities with different kinetic parameters ( Table I). The ATPase activity of the wild type protein has an apparent K , of 149 FM, intermediate between the apparent K, values of the C-terminal site (60 PM) and the N-terminal site (312 p~) . The apparent second order rate constant (Kca,/K,) for the association of the enzyme and substrate of the C-terminal site (derived from K37A) is only slightly decreased while that of the N-terminal site (derived from K646A) is about 10-fold lower than the wild type ATPase. Thus, the C-terminal site apparently has a higher affinity for ATP binding than the Nterminal site. In addition, a substantial decrease in the Kc,, values of the individual sites compared to the wild type protein suggests cooperativity between the sites in the catalysis of ATP hydrolysis.

TABLE I
Kinetic parameters of wild-type and mutant ATPase activity The initial rates of ATPase activity were measured using polyethyleneimine-cellulose TLC based assay as described under "Experimental Procedures." The assays contained 24 nM UvrA. The DNA used in these assays was pPYC3 plasmid DNA. UV exposure was at 720 J m-?. The KG46A ATPase activity in the presence of DNA was measured using 25 to 650 PM ATP (labeled with ["HIATP) while 25 to 350 PM ATP was used in all other measurements. WT, wild-type; NT, N-terminal ATPase; CT, C-terminal ATPase; ND, not detectable (the initial rates of ATP hydrolysis a t all the concentrations tested were less than 0.1 nmol/min).  UV-damaged as well as native DNA modulate wild type UvrA-associated ATPase activities exhibited by both of the individual sites in the mutant proteins as well as the wild type protein. The apparent K,,, values of the individual ATPase sites are 3-to 23-fold higher and the apparent second order rate constants are 3-to 17-fold lower than the wild type protein with both functional ATPase sites operating. These results further suggest allosteric interactions between the sites in both ATP binding and hydrolysis when in the presence of DNA. The significant levels of inhibition of ATPase activity of UvrA protein in the presence of either UV-damaged or native DNA is in agreement with the notion that ATP-bound UvrA protein has an enhanced affinity for DNA binding while hydrolysis of the bound ATP favors its dissociation from DNA.

Km
Effect of ATPase Site Mutations on Binding to ATP-Mutagenesis of the lysine at the end of the glycine-rich loop of the consensus NTP binding motifs did not significantly affect these mutant UvrA proteins to bind ATP (Fig. 6). The wild type and the mutant UvrA proteins showed essentially similar levels of ATP binding. These findings further suggest that the inability of the double mutant UvrA protein to hydrolyze ATP is not due to a defect in ATP binding ( The retention of ATP binding ability by these ATPase site mutants provides a system with great potential for an examination of the role(s) of binding and hydrolysis of ATP by the UvrA subunit of the UvrABC endonuclease.
Effect of Nucleotide Cofactors on DNA Binding-The requirement of nucleotide cofactor binding and/or hydrolysis in support of UvrA-DNA binding was analyzed. ATP has been previously shown to be the preferred nucleotide cofactor for UvrA. DNA complex formation (Seeberg and Steinum, 1982;Yeung et al., 1986a;Seeley and Grossman, 1989). However, these studies could not distinguish between the role(s) of ATP binding and the role of ATP hydrolysis in nucleoprotein complex formation. Since the single and double mutations in the putative ATP binding sites led to the loss of ATPase activity and retention of ATP binding ability of the mutated site(s) ( Table I and Fig. 6), it provided a useful system by which to separate these roles in nucleoprotein complex formation.
The mutant proteins as well as the wild type protein bind DNA in the absence of nucleotide (Fig. 7 A ) . However, in the presence of ATP, the double mutant (K37A K646A) resulted in the highest level of DNA binding while the binding of the single mutants (K37A, K646A) were intermediate between wild type and double mutant (Fig. 7 C ) . The introduction of ADP in the binding reaction reduced the level of nucleoprotein complex formation by the mutants as well as the wild type protein (Fig. 7 B ) . These observations suggest that it is the binding of ATP which has a greater influence on nucleoprotein formation, whereas the hydrolysis of ATP seems to favor nucleoprotein complex dissociation. Interestingly, the wild type protein exhibited the lowest level of nucleoprotein complexed compared to the mutants in the presence of ATP, suggesting that the wild type UvrA is in an equilibrium between the bound form driven by nucleotide binding and the unbound form driven by the hydrolysis of ATP.
Substitution of A T P r S for A T P i n the Binding Reaction-ATPyS as a poorly hydrolyzable analog of ATP is expected to result in binding of the wild type protein to a similar level of binding exhibited by the double mutant (Fig. 70). The variability in the level of binding in the presence of ATPrS compared to ATP may be due to the difference in the inter-actions of ATP and ATP-yS to the UvrA protein.
Binding of Mutant and Wild Type Proteins to UUDNA-The binding of wild type UvrA protein to DNA was enhanced as a consequence of DNA damage (Fig. 8A). In contrast, the binding of the mutant proteins shows no differential binding between native DNA versus uvDNA. In some cases, there is a reduction in binding to the uvDNA by mutant UvrA proteins (Fig. 8, B-D). There appears to be a loss of differential recognition of damaged sites on DNA by the UvrA mutations. This is consistent with further analyses of whether hydrolysis of ATP by either one ATPase site or both sites may be important in locating a damaged site by the UvrA protein. It was approached by examining the chelator-resistant residence times of the mutant and wild type proteins on DNA (Table  11).
2 X SSC-resistant Nucleoprotein Complex Formation-The half-lives of chelator-resistant UvrA-DNA complexes were found to be less than 5 s. However, in the presence of UvrB and damaged DNA, the half-life was increased to an hour (Yeung et al., 1986a). The increase in the residence time of these nucleoprotein complexes reflects binding of the UvrAB complex to damaged sites. In the present study, we find that the residence times of the nucleoprotein complex on both undamaged and damaged DNAs are dramatically increased in the C-terminal ATPase site (ATPII) mutant protein (K646A) and in the double mutant (K37A K646A), while the Nterminal ATPase site (ATPI) mutant (K37A) protein and the wild type protein bound to the undamaged DNA formed very unstable complexes (Table 11). Further, there is a noticeable increase in the half-life of wild type protein bound to the damaged DNA compared to the undamaged DNA. The substitution of ATP-yS for ATP in the binding reaction of wild type UvrA protein to either damaged or undamaged DNA increases its half-life in 2 X SSC from <3-4.5 seconds to 30-40 seconds and 25% of the nucleoprotein complexes remain stable with a half-life of more than 3 h. These observations suggest that the initial binding of UvrA protein is largely nonspecific, and hydrolysis of ATP by the C-terminal ATPase site leads to dissociation of the protein enabling it to repeat the binding-release cycle until it finds a damaged site. This diffusion-controlled repetition of random binding and dissociation cycles may be a component of the discrimination required for damage recognition by the UvrABC system. However, the UvrA protein seems to be the damage recognition subunit irrespective of its independent action.

DISCUSSION
The roles of the two ATPase sites present in the UvrA protein have been analyzed by undertaking site-directed mutagenesis of the gene and biochemical characterization of the representative mutant proteins. The results of these experiments indicate that both of these UvrA-associated ATPase sites are functional in the overall nucleotide excision repair in E. coli mediated by the UvrABC endonuclease (Fig. 3, A  and B ) . The biochemical characterization of the respective mutants revealed that these two sites play different roles as well as function together or influence each other by allosteric interactions between the sites. The C-terminal site has a higher affinity for ATP binding than the N-terminal site. The two ATPase sites can hydrolyze ATP independently of each other. ATPase activities exhibited by both of the sites are modulated by undamaged as well as UV-damaged DNA ( Table  I). The ability to bind ATP is unaffected by these mutations at the ATPase sites (Fig. 6). The complete loss of ATPase activity in the double mutant protein with retention of its ability to bind ATP identifies that the conserved lysine residue at the end of the glycine-rich loop of the consensus A type ATP binding motif as assuming a key role in the mechanism of ATP hydrolysis.
In the presence of ATP-yS, the level of wild type UvrA nucleoprotein complex formation reaches the level observed with the double mutant protein (K37A K646A) that is completely defective in ATP hydrolysis (Fig. 7 0 ) . Similar conditions lead to a 10-fold increase in the half-life of chelatorresistant wild type UvrA-DNA complex. Furthermore, the single mutants defective in ATP hydrolysis a t only one of the ATP binding sites show an increased ability to bind DNA compared to wild type protein (Fig. 7C). These findings suggest that ATP binding, but not hydrolysis a t both of the ATP binding sites of UvrA, favors nucleoprotein complex formation. Similarly, it is ATP binding to UvrA that favours its dimerization to form (UvrA)2. In further support of this, ATPyS shifts the monomer-dimer equilibrium toward (UvrA)2 formation . (UvrA), exhibits a greater increase in binding UV-damaged DNA than undamaged DNA (Fig. 8A). The moderately increased residence time in the presence of chelators of wild type UvrA bound to UV-damaged DNA relative to undamaged DNA (Table 11) provides additional support for the notion that the UvrA protein has an intrinsic ability to discriminate between damaged and undamaged sites on DNA. Analyses of the residence times of these proteins on damaged and native DNA in the presence of  chelators suggested an important role for the C-terminal ATPase site. The K37A and K646A proteins differ only in the location of the mutations at the N-terminal site and the C-terminal site, respectively. However, the former protein exhibits a sensitivity essentially similar to chelators as the wild type protein bound to undamaged DNA whereas the latter is highly resistant to dissociation under similar conditions. Based on these observations, it is concluded that the hydrolysis of ATP at the C-terminal site of UvrA is essential for UvrA to dissociate from nonspecific or undamaged sites. The participation of UvrA in the initial steps of nucleotide excision repair in E. coli is dependent on binding and/or hydrolysis of ATP (Fig. 9). ATP binding seems to induce an allosteric effect leading to a conformational change in UvrA resulting in its self-association to form the reactive dimeric species. The initial binding reaction is a nonspecific one as the probability of the protein landing on a damaged site is very low. It is this step that the UvrA protein senses the DNA for damaged sites. The end result could be 2-fold. One, (UvrA)2 remains at the bound site if that site contains damage. Secondly, in the event of failure to find damage, hydrolysis of ATP at the C-terminal ATPase site leads to dissociation of the complex. The dissociated protein could proceed through similar steps and recycle until it finds a damaged site allowing nucleation for the subsequent steps of nucleotide excision repair. Because of the low level of discrimination between damaged and undamaged sites, it is unlikely that (UvrA)' could find a single damaged site in lo6 nucleotides in a diffusion-controlled reaction. An alternate mechanism for damage recognition by the UvrABC system is becoming increasingly evident from the recent observation that UvrAB acts as a helicase and tracks on DNA in search of damage Grossman, 1987, 1989;Koo et al., 1991). Damage recognition by the UvrABC endonuclease system, however requires the hydrolysis of ATP presumably localized at the C-terminal ATPase site of UvrA. This binding/hydrolysis step seems to provide a sensing step localizing the complex at damaged sites to initiate repair processes.
It is apparent from UV sensitivity and survival experiments of the uurA mutants that productive nucleoprotein complex formation is critical to repair (Fig. 3, A and B ) . The extreme sensitivity of the C-terminal ATP binding site mutants and the double mutants is due to essential irreversible binding of those mutant UvrA proteins to nonspecific sites, thus blocking their ability to locate UV adducts on DNA. However, reduced UV sensitivity of the N-terminal site mutants suggests that binding of ATP but not hydrolysis is required at this site by the UvrA protein during the damage recognition step. The wild type level of survival exhibited by the K37R mutant further supports this hypothesis and suggests ATP binding is not greatly impaired by the conservative change of lysine to arginine in this region of UvrA.
The biochemical characterization of the site-directed mutant proteins of UvrA has not only allowed for an understanding of the different roles of ATP in the initial steps of damage recognition by UvrA, but also provides direct evidence that UvrB exhibits ATPase activity in the presence of UvrA and DNA.' This conclusion is reached from the observation that the ability of the double mutant protein (K37A K646A) to manifest ATPase activity depends on the presence of UvrB protein and DNA. The requirement for ATP binding and hydrolysis at either ATP binding site by the UvrA subunit for the function of the UvrAB complex and the UvrABC system are under current investigation.