The Role of Conserved Amino Acids in Substrate Binding and Discrimination by Photolyase"

DNA photolyases catalyze the light-dependent repair of pyrimidine dimers in DNA. We have utilized chemical modification and site-directed mutagenesis to probe the interactions involved in substrate recogni- tion by the yeast photolyase Phr 1. Lys617 was protected from reductive methylation in .the presence of sub- strate, but not in its absence, and the specific and nonspecific association constants for substrate binding by Phrl(Ly~"'~ + Ala) were decreased 10-fold. These results establish a role for Lys6I7 in substrate binding. Mutations at Ardo7, Lys'", and reduced both the overall affinity for substrate and substrate discrimination. Sites of altered interactions in ES complexes were identified by methylation and ethylation interference techniques. Interaction with the base immedi- ately 3' to the dimer was altered in the Phr1(Lys617 + Ala).DNA complex, whereas interactions with the phosphate and base immediately 5' to the dimer were reduced when Phr1(Ar&07 Multiple 5' and 3' the were perturbed in h ( r ~ In addition the quantum yield for photolysis by 3-fold. The locations of these mutations establish that a portion of the DNA binding domain is comprised of residues in the highly conserved carboxyl-terminal Quenching-Fluorescence quenching experiments were performed using a Shimadzu RF5000U spectrofluorometer at 15 "C as described by Kim et al. (1992). 1.5 X lo-' M photolyase in 100 mM NaC1,l mM EDTA, and 50 mM Tris-HC1, pH 7.5, was exposed to 295 nm radiation, and tryptophan fluorescence emission was monitored at 334 nm. Aliquots of 7 M acrylamide were added to the solution and after each addition fluorescence at 334 nm recorded. The data were analyzed using the modified Stern-Volmer equation (Lakowicz, 1983), Fo/(Fo - F) = l/f + l/fKsV[Q]. Fo and Fare, respectively, the relative fluorescence intensities without and with quencher at concentration [Q], f is the fractional accessibility of the tryptophans in photolyase to quenching agent, and Ksv is the Stern-Volmer constant which reflects the efficiency of quenching. KSV was determined from the slope of a plot of (Fo - F)/Fo uersus l/[Q].

the enzymes display 15% sequence identity; however, isolated regions display significantly greater sequence homology, most notably the carboxyl-terminal 150 amino acids where 30% of residues are identical.
All photolyases share at least three common functions: (i) binding of the ubiquitous FADHz chromophore; (ii) binding of a second chromophore which, depending upon the source of the enzyme, is either 5,lO-methenyltetrahydrofolate (E. coli, S. typhimurium, N. crassa, S. cereuisiae) or a derivative of 8-hydroxy-5-deazaflavin (A. nidulans, H. halobium, S. griseus); (iii) specific binding to pyrimidine dimers. Surprisingly, neither flavin, folate, nor DNA binding domains are apparent from the primary sequences of the enzymes. Recent studies by Malhotra et al. (1992) on the Phrl photolyase of S. cerevisiae demonstrated that the carboxyl-terminal 275 residues of the enzyme comprise the FAD binding domain, whereas residues 15-326 contain the folate binding domain. However neither of the fragments comprising the chromophore binding domains was capable of binding dimer-containing DNA with high affinity and specificity. We report here the results of chemical modification and site-directed mutagenesis studies which have identified residues involved in substrate binding and discrimination by Phrl. Our results implicate 4 conserved amino acids ( L y P , Arg507, L~S~~, and Trp387) in substrate binding, thereby establishing that at least a portion of the DNA binding site is composed of residues in the conserved carboxyl-terminal domain. Alanine substitution at each of these sites reduces the affinity of photolyase for substrate and produces specific changes in the interactions between Phrl and DNA phosphates and bases surrounding the dimer. The implications of these results for models of substrate recognition and discrimination by photolyases are discussed.

EXPERIMENTAL PROCEDURES
Materials-Sources of materials were the same as reported in our previous study (Baer and Sancar, 1989). Micrococcus luteus pyrimidine dimer glycosylase-AP endonuclease was purchased from Applied Genetics Inc., oligo(dT)1M, was from Pharmacia LKB Biotechnology Inc., and hydroxylapatite Ultrogel was from IBF Biotechnics. Acrylamide for quenching experiments was obtained from American BioNuclear and was washed twice in ice-cold 95% ethanol prior to use.
Enzyme Preparations-All photolyases were expressed from tacregulated cloned genes carried on derivatives of plasmid pCB1241 (Sancar and Smith, 1988), propagated in E. coli strain CSR603/ F ' h P (Sancar, 1985a). The enzymes were purified as described previously (Sancar et al., 1987a) with the following modifications. After elution from DNA cellulose, fractions containing photolyase were pooled, dialyzed into hydroxylapatite column buffer (20 mM potassium phosphate, pH 6.8, 1 mM EDTA, 10 mM b-mercaptoethanol, 20% (v/v) glycerol) and loaded onto a hydroxylapatite column (16 X 48 mm) equilibrated in the same buffer. Following a 50-ml wash with column buffer, a 50-ml linear gradient of 20-400 mM potassium phosphate was applied. One-ml fractions were collected and 10 pl from each fraction were analyzed by SDS-polyacrylamide 16717 This is an Open Access article under the CC BY license.
Photolyase Binding Reaction-All binding reactions were performed at 23 "C in a 25-p1 mixture containing 50 mM Tris-HC1, pH 7.5, 1 mM EDTA, 100 mM NaC1, 1 mM 8-mercaptoethanol, 6% (v/v) glycerol, and 5 X lo-' M substrate DNA. The substrate was a 43-bp' oligonucleotide (shown below), with a centrally located thymine dimer (indicated by <>), and was constructed as described previously (Husain et al., 1987;Baer and Sancar, 1989) with 32P label at the 5' end of the top strand. After a 20-min incubation at room temperature, free DNA and enzyme-bound DNA were separated on the basis of electrophoretic mobility (EMSA; Fried and Crothers, 1981;Garner and Revzin, 1981). Reaction mixtures were loaded onto a 6% polyacrylamide gel in 90 mM Tris base, 90 mM boric acid, 2.5 mM EDTA and were electrophoresed at 35 mA for 90 min. Binding reactions and electrophoresis were performed under gold fluorescent lamps to prevent photoreactivation. Radioactivity in all bands was quantified using an Ambis Radioanalytic Imaging System. The active molecule concentration for each photolyase preparation was determined by titrating enzyme with substrate followed by separation and quantitation of free and bound substrate as described above. Since photolyase binds substrate as a monomer, the x intercept of a Scatchard plot of the data yields the fraction of molecules active in substrate binding. The equilibrium binding constant (KA) was determined from Scatchard plots of results obtained by titrating 5 X lo-' M substrate with enzyme, correcting for the fraction of active Phrl. Nonspecific binding constants (KNs) were determined by binding competition as follows. Photolyase was bound to labeled substrate such that 70% of substrate was in ES complexes, then increasing concentrations of oligo(dT)20 were added to compete with labeled substrate for binding. KNS was determined from the x intercept of a Dixon plot of competitor concentration versus l/(fraction of ES complexes remaining) (Segel, 1975). The intrinsic specific binding constant (Ks) was then determined from the relationship KO&& = &(l+[D]&S) (Terry et al., 1983), in which [Dl is the molar concentration of nondimer nucleotides in the 43-bp substrate.
Kinetics of Inactivation of P h r l by Reductive Methylation-The method used for reductive methylation of lysines in Phrl was based on that of Cabacungan et al. (1982). Photolyase (9.4 X lo-' M) was incubated at room temperature in 250 p1 of 100 mM formaldehyde in modification buffer (50 mM Hepes buffer, pH 7.4, 100 mM KCl, 6% (v/v) glycerol, and 20 mM NaCNBH,). At various times 20-4 samples were removed and 2.5 pl of 1.0 M lysine were added to quench the reaction. Labeled substrate (2.5 pl of 10" M) was then added to the samples and incubated for 10 min prior to EMSA. In a similar experiment, photolyase was incubated with labeled substrate (lo-' M) for 10 min at room temperature prior to treatment with formaldehyde, and 20-ml aliquots were removed at various times. Five pl of 500 mM lysine were added to quench the reaction. Bound and free substrate were separated by EMSA and quantitated.
Preparative Scale Reductive Methylation of Phrl-Oligo(dT)l~ containing five dimers/molecule was prepared by irradiating a solution of oligonucleotide (2 X 10" M in 10 mM NaCl) with 254-nm light. Total dimer content was determined by the decrease in absorbance at 280 nm. Three nmol of Phrl in 100 pl were incubated with 100 mM [3H]HCH0 (80 mCi/mmol) as described previously by Takeda et al. (1986). In a similar reaction 100 mM ["CIHCHO (10 mCi/ mmol) was used to modify Phrl in the presence of 1 X lo-' M dimercontaining oligo(dT)leo ([dimer] = 5 X lo-'). Greater than 95% of EMSA, electrophoretic mobility shift assay; DMS, dimethyl sulfate; The abbreviations used are: bp, base pair(s); kbp, kilobase pair; HPLC, high performance liquid chromatography. enzyme was bound to substrate under these conditions. One minute after addition of labeled HCHO, 5 volumes of nonlabeled 100 mM formaldehyde were added to both reactions. In a third reaction 900 nmol of Phrl in 300 p1 were reductively methylated using 100 mM unlabeled formaldehyde. Ten min after addition of unlabeled formaldehyde, all reactions were quenched with lysine as described above, the mixtures were combined, and photolyase was precipitated at room temperature by addition of trifluoroacetic acid to a final concentration of 1% (v/v). The precipitate was recovered by centrifugation at 16,000 X g for 10 min, dissolved in 6 M guanidine hydrochloride, and dialyzed into 10 mM NaCl at 4 "C. During dialysis, Phrl precipitated. Enzyme was recovered by centrifugation as described above, dissolved in 200 pl of 70% formic acid (v/v), and treated with 200 mM CNBr under argon for 24 h at room temperature. The digest was lyophilized then dissolved in 250 pl of 0.1% trifluoroacetic acid (v/v).
Peptides were separated by HPLC using a W-Porex (Phenomenex) reverse phase C18 column (250 X 4.6 mm). Samples were loaded onto a column pre-equilibrated with 0.1% (v/v) trifluoroacetic acid, then peptides were eluted with a 60-ml linear gradient of 0-70% (v/v) methanol. Peaks were collected manually. Multiple runs at various flow rates were used to separate closely eluting peptides which were detected by absorbance at 220 nm using an in-line ultraviolet detector. I4C and 3H disintegrations/min in each fraction were determined by scintillation counting in an LKB RackBeta instrument employing a predetermined correction for spectral overlap. Isolated peptides were sequenced by Dr. David Clapper (University of North Carolina at Chapel Hill) using an Applied Biosystem model 4700 automated peptide sequencer with an in-line Waters HPLC. Each amino acid peak was collected and analyzed for 3H and "C content.
M13 mpl8:PHRlBK was also used to reconstruct a PHRl overexpression vector lacking the BglII site at bp 1710 and containing the KpnI site at bp 1365. Replicative form I DNA was digested with BgAI and PstI to release a 2.1-kbp fragment containing the PHRl coding sequence 3' to bp 400 (Sancar, 1985b). This fragment was ligated into BglII-PstI-digested pCB1241 (Sancar and Smith, 1988) yielding plasmid pCB1241Mut, in which expression of PHRl is driven from the tac promoter.
Amino acid substitution mutants. were generated by oligonucleotide-directed mutagenesis usmg single-stranded M13 mpl8:PHRlKpn as template (Kunkel, 1985). KpnI-PstI restriction fragments containing the desired mutations were used to replace the KpnI-PstI fragment of pCB1241Mut, thereby reconstructing an intact mutated PHRl gene. The nucleotide sequence for each reconstructed gene was confirmed by DNA sequencing (Sanger et al., 1977).
Probing Phosphate and Major Groove Contacts-Methylation and ethylation interference experiments were performed as described previously (Baer and Sancar, 1989) except that photolyase concentrations were adjusted to achieve 50% substrate bound. (Higher concentrations of most of the mutant enzymes interfered with separation of free and bound substrate by EMSA in these experiments. This difficulty was not encountered in other experiments reported here and is related to the large amount of DNA and protein required in the interference experiments.) Strand cleavage products and the products of chemical sequencing reactions were visualized by electrophoresis through 12% DNA sequencing gels and autoradiography as described (Siebenlist and Gilbert, 1980). The relative amount of radioactivity in each band was quantitated using either an Ambis Radioanalytic Imaging System or by densitometry using a Molecular Dynamics densitometer. Counts/min or total area in a single band located outside of the binding site was used to correct for lane-tolane loading variation.
Relative Quantum Yield-Monochromatic photoreactivation experiments were performed at 15 "C using a Quanticount Monochrometer/Actinometer from Photon Technology International. Photo- lyase was incubated with substrate such that 80% of the substrate was bound. The reaction mixture was exposed to successive 500 erg/ mm2 fluences of 380 nm radiation. 2 0 4 aliquots were removed after each irradiation and photolyase-bound substrate and free substrate were separated by EMSA and quantitated. Quantum yields were determined by plotting total fluence received uersus the fraction of ES complexes remaining (Harm and Rupert, 1968). An alternative method for determining repair was used in some cases. 10-pl aliquots were removed after each irradiation, extracted with phenol, dissolved in buffer (50 mM Tris-HC1, pH 8.0, 40 mM NaCI, 1 mM EDTA), and treated for 90 min at 37 "C with 320 units of M. luteus pyrimidine dimer glycosylase-AP endonuclease, which cleaves the intradimer phosphodiester bond. Control experiments indicated that all nonrepaired substrate was cleaved under these conditions. Cleaved and full-length (repaired) substrates were separated on a 12% sequencing gel and quantified.
Acrylumide Quenching-Fluorescence quenching experiments were performed using a Shimadzu RF5000U spectrofluorometer at 15 "C as described by Kim et al. (1992). 1.5 X lo-' M photolyase in 100 mM NaC1,l mM EDTA, and 50 mM Tris-HC1, pH 7.5, was exposed to 295 nm radiation, and tryptophan fluorescence emission was monitored at 334 nm. Aliquots of 7 M acrylamide were added to the solution and after each addition fluorescence at 334 nm recorded. The data were analyzed using the modified Stern-Volmer equation (Lakowicz, 1983), [Q]. Fo and Fare, respectively, the relative fluorescence intensities without and with quencher at concentration [Q], f is the fractional accessibility of the tryptophans in photolyase to quenching agent, and Ksv is the Stern-Volmer constant which reflects the efficiency of quenching. KSV was determined from the slope of a plot of (Fo -F)/Fo uersus l/[Q].

RESULTS
Reductive Methylation of P h r l " P h r l interacts extensively with DNA phosphates on the dimer-containing strand (Baer and Sancar, 1989) and displaces 3-4 Na' upon binding.2 Because interactions between proteins and DNA phosphates frequently involve salt bridge formation (Siebenlist and Gilbert, 1980), we decided to examine the effect of lysine modification on substrate binding. Treatment of proteins with formaldehyde in the presence of sodium cyanoborohydride results in reductive methylation of solvent exposed lysines (Cabacungan et al., 1982). P h r l was treated for various times, then challenged with 32P-labeled dimer-containing 43-bp substrate, and bound and free substrate were separated by EMSA.
As can be seen in Fig. 1, DNA binding activity was rapidly G . B. Sancar and F. W. Smith, manuscript in preparation.
lost; the initial slope yielded a t1/2 of 1.8 min. In contrast, when P h r l was incubated with substrate such that 80% of substrate and enzyme were bound, then treated with formaldehyde, tlI2 increased to 5.0 min. Protection was not observed when P h r l was treated in the presence of an identical concentration of 43-bp substrate lacking pyrimidine dimer. Thus substrate binding protects one or more lysines in P h r l from methylation.
To identify the specific lysines protected, P h r l was treated with [3H]formaldehyde in the absence of DNA and ["C] formaldehyde in the presence of dimer-containing oligo-(dT),,. The enzymes were then mixed, cleaved with CNBr, and the resulting peptides were separated by HPLC. In the initial separation 57 peaks were detected (Fig. 2, panel A); this is approximately three times the number of expected peaks (Phrl contains 18 methionines) and was due to incomplete cleavage by CNBr. Both the intact protein (data not shown), and the majority of peaks displayed approximately a 1:l ratio of 3H:'4C ( Fig. 2, panel B). To ensure that peaks with elevated 3H:'4C ratios were not missed due t o peak overlap, all peaks with ratios >0.9 were repurified. Upon repurification peak 38 displayed a 3H:'4C ratio of 2.0 ( Fig. 2, panels C and D), indicating the presence of a lysine protected from modification in the ES complex; all other peaks displayed ratios 51.3 and were not examined further. Sequence analysis of the peak 38 peptide and scintillation counting of the cleaved phenylthiohydantoin derivatives revealed that the protected lysine was at position 517 in the Phrl sequence (Fig. 3).
To confirm that Lys6I7 plays a role in DNA binding, this residue was changed to an alanine by site-directed mutagenesis, and the mutant enzyme was purified to apparent homogeneity. As can be seen in Fig. 1, substrate DNA did not protect Phrl(Lys517 + Ala) from inactivation by reductive methylation. The tl12 for inactivation in both the presence and absence of substrate was 1.0 min. Therefore, we conclude that methylation of Lys5I7 in the wild type enzyme is responsible for the initial loss of binding activity. In all sequenced microbial photolyases the residue at the equivalent position is either Lys or Arg (Fig. 3), suggesting a strict functional requirement for a positive charge at this position.

Construction of Phrl Mutants and Equilibrium Binding
Analy~is-Lys'~~ lies in the conserved carboxyl-terminal region of P h r l , suggesting that this region comprises at least a portion of the DNA binding domain of the enzyme. We therefore introduced mutations at additional conserved residues within this region and determined the effect on specific and nonspecific binding to substrate. Argo7 is conserved in all reported microbial photolyase sequences and is located near Lys517, whereas Lys463 is the only lysine that is conserved in all microbial photolyase sequences. Trp3s7 was also targeted because previous work on E. coli photolyase (Li and Sancar, 1990) has implicated the equivalent Trp residue in DNA binding. In all cases the original amino acid was replaced with an alanine residue, which is equally accommodated in hydrophobic and hydrophilic environments (Schultz and Schrimer, 1979). Each mutant photolyase was overproduced at near wild type levels and purified to apparent homogeneity. Wild type P h r l (Sancar et al., 1987a) and the mutant photolyases displayed similar absorbance spectra with peaks at 277 and 377 nm (data not shown). In addition each mutant enzyme contained folate, flavin, and apoenzyme in 1:1:1 stoichiometry, and in no case were there detectable quantities of blue neutral flavin radical or FAD,, (data not shown). We conclude that the amino acid substitutions do not introduce large structural perturbations in the enzyme which might be expected to Following removal of unincorporated counts and treatment with CNBr, peptides were applied in 0.1% (v/v) trifluoroacetic acid to a reverse phase C18 column and were eluted with a linear gradient of 0-70% (v/v) methanol in 0.1% (v/v) trifluoroacetic acid. Panels A and B show, respectively, the absorbance at 220 nm and 3H:"C ratios obtained during the initial fractionation of CNBr-treated Phrl. The first 17 fractions were collected across the break-through peak and appeared to contain primarily large partial digestion products (data not shown). Panels C and D show the absorbance and 3H:14C ratios obtained upon repurification of material from peak 38 of the initial separation. The arrows in panels A and C indicate the locations of peak 38. Purified material from the separation shown in panel C was subjected to amino acid sequence analysis.  whereas Ks for Phrl(Arg607 + Ala), P h r l ( L y~~~~ + Ala) and P h r l ( T~p~'~ + Ala) decreased by 2 orders of magnitude. Thus each of the mutant photolyases has reduced affinity for dimercontaining DNA. The discrimination ratio (KS/KNs) is a measure of the ability of a DNA binding protein to selectively bind to its target. The discrimination ratio for wild type P h r l is lo4. Phrl(Lys517 + Ala) displayed a similar discrimination ratio while Phrl(Argo7 + Ala), P h r l ( L y~~~~ + Ala), and P h r l ( T~p~'~ + Ala) had discrimination ratios of approximately lo3. This suggests that interactions involving Lys517 contribute equally to specific and nonspecific binding, whereas Lys463, and Trp387 participate in interactions which contribute to a greater extent to substrate discrimination.

Phosphate Contacts of Mutant Photolyases-Ethylnitrosou-
rea ethylates phosphate oxygens on the DNA backbone and interferes with photolyase binding to dimer-containing DNA either by eliminating ionic interactions between Phrl and DNA or by steric hindrance (Baer and Sancar, 1989). T o map the sites of altered interactions between the mutant photolyases and DNA phosphates, we compared the ethylation interference patterns of the mutant and wild type enzymes.
The results are shown in Fig. 4. As in our previous work (Baer and Sancar, 1989) we found that, for wild type P h r l , ethylation of the phosphate immediately 5' to the dimer weakly inhibited binding, whereas ethylation of any of the three phosphates immediately 3' to the dimer strongly inhibited binding. Ethylation at the fourth phosphate 3' to the dimer also inhibited binding by Phrl but to a lesser extent. (Interference at the 5' site was substantially weaker than in our previous study; this probably reflects the fact that the current experiments were performed with subsaturating enzyme concentrations (see "Experimental Procedures")). The interference pattern obtained with Phrl(LysS17 + Ala) was identical Each photolyase was tested at least twice, and scanning densitometry was used to confirm interference in questionable cases. The sequence of the substrate is shown on the left with the bracket indicating the position of the dimer. Numbering is identical to that shown under "Experimental Procedures." to that obtained with wild type enzyme. In contrast, the patterns obtained with P h r l ( A r p + Ala), Phrl(Lysm + Ala), and Phrl(TrpS7 + Ala) revealed altered interactions between the enzymes and DNA phosphates. For Phrl(ArC7 + Ala), interference at the first phosphate 5' to the dimer was reduced compared with wild type Phrl, implicating A r p ' in a contact with this phosphate. The patterns obtained with Phrl(Lysm + Ala) and Phrl(Trpm7 + Ala) showed extensive differences from wild type. In both cases ethylation of the first phosphate 5' to the dimer as well as the first two phosphates 3' to the dimer no longer interfered with binding. These results suggest that either the Lys463 + Ala and TrpS7 + Ala mutations substantially perturb the structure of the DNA binding site or that loss of the contacts normally made by L Y S '~ and TrpS7 permits realignment of the enzyme on its substrate.
Major Groove Contacts-DMS methylates N7 of guanine in the major groove of DNA and interferes with DNA-protein interactions by disrupting hydrogen bonds at this site or by steric hindrance. The effect of DMS modification on substrate binding by wild type and mutant photolyases was determined by methylation interference analysis. The results are shown in Fig. 5. In agreement with our previous report (Baer and Sancar, 1989), methylation of G215' to the dimer and of G25 and of G27 3' to the dimer interfered with binding of wild type Phr, whereas methylation of G24 enhanced the ability of Phrl to bind substrate. The interference patterns for binding by each of the four mutants differed from wild type. Methylation of G25 failed to interfere with binding by Phrl(Lys"' + Ala), and a new interference band was apparent at G24. Thus the effect of the Lys517 + Ala mutation is localized to the 2 bases immediately 3' to the dimer. Methylation at G21 not only failed to inhibit binding by Phrl(ArgW' + Ala) but actually appeared to enhance binding slightly; additionally, methylation interference at G27 was also absent for this mutant. Because interference at the latter site is normally weak for wild type Phrl, we believe that the primary site at which interaction is perturbed is at G21 and the adjacent phosphate (see above). As was evident in the ethylation interference studies, multiple interactions are affected in the P h r l ( T~p~'~+ A l a ) and P h r l ( L y~~~ + Ala) complexes. Methylation at G21 did not interfere with binding by either mutant, whereas, in contrast to wild type, methylation of G24 interfered with binding by Phrl(TrpS7 + Ala). Thus interactions with the bases immediately flanking the dimer are altered in the Phrl(Trp3" + Ala) ES complex. Substrate binding by P h r l ( L y~'~~ + Ala) appeared to be insensitive to methylation as indicated by the absence of interference at any site. Because photolyase binds pyrimidine dimers regardless of the surrounding sequence context (Myles et al., 1987;Rupert, 1962aRupert, , 1962b, methylation interference probably reflects steric constraints rather than the loss of hydrogen bonds between the enzyme and DNA bases. The results indicate that interactions with major groove residues are altered in each mutant Phrl ES complex. Relative Quantum Yield-The quantum yield for photolysis of dimers by Phrl at 384 nm is 0.5 and proceeds via energy transfer from the folate chromophore to FADHz followed by electron transfer from flavin to the dimer (Sancar, 1990). This process is potentially sensitive to changes in the alignment of and distance between enzyme-bound FADHz and the dimer. Additionally, changes in enzyme structure which alter interaction between the folate chromophore and FADH, may affect the quantum yield of photolysis. To further characterize the substitution mutants, we determined the quantum yield for each mutant enzyme relative to wild type. As can be seen in Table 11, the quantum yields displayed by two of the mutants, Phrl(TrpS7 + Ala) and P h r l ( L y~~~~ + Ala), were substantially different from wild type. To confirm these differences, we used a more direct method for quantitating repair. After the reaction mixtures were exposed to photoreactivating light, each sample was treated with M. luteus pyrimidine dimer glycosylase-AP endonuclease which cleaves the glycosylic bond linking the 5' dimer base and sugar as well as the intradimer phosphodiester bond (Haseltine et al., 1980). Fulllength and cleaved substrate were separated on a 12% sequencing gel. When the quantum yield was calculated by this method, the value obtained for Phrl(TrpS7 + Ala) was onethird that of wild type Phrl. This is in contrast to results reported by Li and Sancar (1990) which indicate that (different) mutations at the equivalent site in E. coli photolyase do not affect the quantum yield. In contrast the quantum yield displayed by P h r l ( L y~~~ + Ala) was similar to wild type. We believe that the latter value for this mutant is the correct one and that the greater value obtained in the EMSA analysis reflects denaturation of the enzyme during irradiation and/or electrophoresis; the released substrate would be scored as repaired in the EMSA assay. This is consistent with the fact that preparations of this enzyme consistently displayed the lowest number of active molecules in the binding assay (40%). In addition the known reaction mechanism for photolysis precludes a quantum yield greater than 1.0, which is implied by the results of the EMSA experiment.
Conformational Analysis of Phrl and Photolyase Mutants by Acrylamide Quenching-The intrinsic fluorescence of solvent-exposed tryptophan residues is quenched by acrylamide. Thus structural perturbations which expose one or more tryptophans to solvent can be detected by both increased tryptophan fluorescence and increased quenching by acrylamide. To test whether any of the mutations introduced into Phrl had greatly perturbed the structure of the enzyme, acrylamide quenching experiments were performed. This method is particularly suited for analysis of the mutants described here as the carboxyl-terminal region of Phrl, in which the mutations are located, contains 9 tryptophan residues distributed throughout the region. The enzymes were exposed to 295 nm radiation, and fluorescence emission at 334 nm was measured. Addition of small aliquots of 7 M acrylamide quenched the fluorescent emission of the tryptophans and permitted us to determine KSV, which reflects the efficiency of quenching. KSV for all mutant photolyases was similar to wild type Phrl, and quenching was insignificant compared with that seen with an equimolar concentration of free tryptophan (Table 11).

DISCUSSION
DNA photolyases repair pyrimidine dimers in a variety of sequence contexts yet discriminate as efficiently between target and nontarget sequences as do sequence-specific DNA binding proteins (Rupert, 1962a(Rupert, , 1962bSancar et al., 198%;Sancar, 1990). Unlike the latter proteins, which utilize primarily sequence-specific base contacts to identify target sequences, photolyases must recognize an altered DNA structure introduced by and characteristic of a pyrimidine dimer. This structure includes the two pyrimidines in the dimer, as shown by the observation that dimer-containing UpU, TpT, and the TT base dimer are repaired by E. coli photolyase (Witmer et at., 1989;Kim and Sancar, 1991). However additional interactions with flanking residues are at least as important in stabilizing the ES complex; comparison of the KA for binding to dimer-containing poly(dT) uersus TpT indicates that 50% of the binding free energy comes from such flanking interactions (Kim and Sancar, 1991). More importantly, the specific binding equilibrium constant ( Ks) decreases with substrate length, indicating that at least some of these flanking interactions are utilized in substrate discrimination.* Mapping of the contacts made by Phrl on the DNA surrounding the dimer has established that residues on yeast photolyase lie in close proximity to or interact with four to five phosphates on the dimer-containing strand as well as with major groove residues 5' and 3' to the dimer (Baer and Sancar, 1989). The same contacts are utilized by photolyases from a prokaryote (E. coli; Husain et al., 1987) and an archaebacterium (Methanobacterium thermoautotropicum; Kiener et ai., 1989), suggesting that all photolyases recognize the same structural determinants which uniquely specify the site of a dimer. However previous studies have not revealed which interactions play a role in binding specificity nor have the amino acids involved in specific substrate recognition by Phrl been identified. In the present study we have addressed these questions by constructing and characterizing a set of mutants which are defective in substrate binding and discrimination and which have lost specific contacts to residues on the DNA surrounding the dimer.
Alanine substitution mutagenesis was utilized in this study in an attempt to minimize the potential for structural perturbation of the enzyme. Consistent with this, several lines of evidence indicate that the Phrl mutants described here have not suffered extensive structural alterations. (i) Each protein contains a complete complement of the flavin and folate chromophores, and (ii) the redox states of the chromophores are unchanged. Thus the chromophore binding sites as well as the interactions which maintain the chromophores in the reduced state are not altered. (iii) With the exception of Phrl (Trp387 "-* Ala) the quantum yield for photolysis for each mutant is similar to that of wild type photolyase, indicating that the interactions between the two chromophores and between the flavin chromophore and the pyrimidines in the dimer are not significantly changed. (iv) With the exception of P h r l ( L y~'~ + Ala), the fraction of photolyase molecules active in DNA binding exceeded 70% in each preparation (data not shown). (v) All mutants are able to discriminate to a significant extent between dimer and nondimer DNA. (vi) All mutants display a response to acrylamide quenching indistinguishable from wild type, suggesting that at most only minor local structural changes have resulted from the substitutions.
Each of the mutations reported here produces a characteristic decrease in the free energy of binding, with AGOb ranging from +1.4 to +3.1 kcal/mol (Table I). Fersht (1989) has estimated that a single hydrogen bond between uncharged groups contributes 0.5-1.5 kcal/mol to binding energy, whereas a hydrogen bond to a charged group contributes 3-4 kcal/mol. Similarly in 0.1 M NaCl a single electrostatic interaction between a DNA phosphate and the €-amino group of lysine contributes 1.2 kcal/mol (Record et al., 1976). Thus the changes in binding energy that we observe are consistent with loss of only one or a few contacts. In addition the mutants displaying the greatest losses in binding energy display the greatest number of altered interactions in methylation and ethylation interference assays. This is strong evidence that the changes in KS reflect primarily altered interactions at the DNA-photolyase interface. The sites of altered interactions that we have detected are summarized in Fig. 6.
Lys517 is protected from methylation when Phrl is bound to substrate but not when the enzyme is free in solution, implying that it is buried between Phrl and DNA in the ES complex. Furthermore the fact that, in contrast to wild type Phrl, the initial kinetics of inactivation of Phr1(Lys5l7 + Ala) are unaffected by substrate binding indicates that the rapid loss of binding activity for wild type Phrl is the consequence of methylation of L y P . Further evidence for the involvement of Lys517 in binding comes from thermodynamic and binding interference analysis. Both the specific and nonspecific equilibrium binding constants are reduced 10-fold for Phrl(Lys517 + Ala); thus interactions involving Lys517 contribute approximately equally to the free energy of specific and nonspecific binding. Loss of methylation interference at G25 suggests that Lys517 interacts with the second base 3' to the dimer. It is unlikely that the decrease in binding energy exhibited by Phrl(Lys517 + Ala) reflects loss of a hydrogen bond between N7 of G25 and Phrl; the lack of sequence specificity of photolyase binding implies the absence of sequence-specific hydrogen bonds with the bases surrounding the dimer. Rather the methyl groups of the Lys517 side chain may participate in van der Waals interactions with G25 and methylation at N7 may sterically interfere with binding of Phrl. This does not rule out additional interactions between the €-amino group of Lys517 and DNA at sites not currently amenable to interference analysis, for example the intradimer phosphate, adjacent ribose moieties, or O4 of the thymines in the dimer. Phrl(ArSo7 -+ Ala) is the only other mutant in this study that exhibits a small number of changes in the interference pattern. Ethylation interference at the first phosphate 5' to the dimer is absent from this mutant, as is methylation interference at G21, the base immediately 5' to the dimer. Loss of these interactions is accompanied by a 100-fold decrease in KS and a 20-fold decrease in KNs, indicating that ArgSo7 contributes to substrate discrimination. Although interactions between positively charged amino acid side chains and DNA phosphates have usually been considered "nonspecific," studies on substrate binding by EcoRI endonuclease have demonstrated that such interactions can serve to "anchor and orient protein recognition elements" and thus make a significant energetic contribution to substrate discrimination (Lesser et al., 1990). Our results suggest that interaction between Arg607 and the phosphate 5' to the dimer plays such a role in photolyase binding. Discrimination probably entails recognition of the position of phosphoryl oxygens, which is influenced by local DNA conformation. This is consistent with the currently accepted model of dimer-containing DNA in which the conformation of the sugar-phosphate backbone is altered from the first phosphate 5' to the dimer to at least the third phosphate 3' to the dimer (Pearlman et al., 1985;Taylor et. al., 1990). In addition to the altered interference pattern at residues 5' to the dimer, the normally weak methylation interference at the fourth base 3' to the dimer is missing in the Arg60' + Ala mutant. This may reflect a subtle change in the alignment of the enzyme on DNA which is the indirect consequence of loss of the 5' contacts. Similar distal effects have been reported for the binding of EcoRI to variant sites (Lesser et al., 1990). Clearly the alignment is not substantially altered since the quantum yield for dimer repair is similar to that of wild type Phrl ( Table 11). Replacement of either Trp387 or Lys463 disrupts multiple DNA contacts and many of the affected sites are identical in the two mutants. For both, interference is lost at the first base and phosphate 5' to the dimer as well as at the two phosphates immediately 3' to the dimer. In addition methylation of the second and fourth G's 3' to the dimer (G25 and G27) fails to inhibit binding of Phrl(Lys4= + Ala). The large number of changes resulting from these single amino acid substitutions suggests that TrpS7 and play a crucial role (direct or indirect) in orienting many of the Phrl specificity determinants at the binding site. In the case of Lys463, current evidence does not permit us to discriminate between a role in maintaining the local secondary structure of the binding site in the absence of DNA uersus participation in a network of interdependent interactions between Phrl and DNA. However studies on E. coli photolyase suggest a direct role in DNA binding for Trp3". Based on alignment of the amino acid sequences, Trp"' in the bacterial enzyme is the homologue of of Phrl. Kim et al. (1992) have shown that TrpZ7' of E. coli photolyase photosensitizes photoreactivation at 280 nm with high quantum yield, implying that Trp277 lies in close proximity to the dimer. In addition a Trp277 --* Arg mutation decreases KS 200-fold and increases K N~ 5fold (Li and Sancar, 1990). Helene and Maurizot (1981) have demonstrated that the tripeptide Lys-Trp-Lys can bind to DNA via nonspecific interactions between the 2 Lys residues and DNA phosphates and intercalation of the central Trp residue. Because intercalation is only partial, the bases flanking the insertion site are forced open at an angle and the helix is kinked. This is similar to the altered structure predicted at the two pyrimidines in the dimer (Pearlman et al., 1985) and suggests that intercalation of Trp at the dimer may be energetically favored and thus contribute to substrate discrimination. The large number of altered contacts surrounding the dimer in the Phrl(TrpS7 + Ala) ES complex may reflect the importance of intercalation in correctly orienting additional recognition elements.
Phrl(TrpS7 + Ala) also displays a %fold reduction in the quantum yield for photolysis at 380 nm. Given the disruption of normal interactions at the DNA-protein interface, we believe that this is the indirect result of misalignment of the enzyme-bound FADHz and the dimer in an abnormal "adaptive" ES complex. In contrast, Li and Sancar (1990) have reported that substitution of TrpZ7' of E. coli photolyase with Arg, Glu, Phe, or His does not significantly affect the quantum yield, despite the fact that both the Trp2" + Arg and Trp"" + Glu mutants are clearly defective in substrate binding and discrimination. At present the reason for the different effects of mutations in the two enzymes is not clear.
Each of the four sites probed by site-directed mutagenesis in this study is conserved in all photolyases characterized to date, with the exception of photolyase from the goldfish Carassius auratus (Yasuhira and Yasui, 1992). (The predicted amino acid sequence for C. auratus photolyase is unlike that of the microbial enzymes.) Thus it is likely that these residues are part of a dimer recognition motif which is common to most or all DNA photolyases. Solution of a photolyase-DNA cocrystal structure will be necessary to establish whether these residues directly contact the DNA or anchor other side chains which make these contacts. In either case, the results reported here establish functional roles for Trp3", Lys463, Argo7, and LysS1' in substrate binding and discrimination and suggest some of the sites on dimer-containing DNA that are involved in these processes. Any proposed photolyase structure must be consistent with these observations.