Reaction Mechanism of (6-4) Photolyase*

The (6-4) photolyase catalyzes the photoreversal of the (6-4) dipyrimidine photoproducts induced in DNA by ultraviolet light. Using the cloned Drosophila melanogaster (6-4) photolyase gene, we overproduced and purified the recombinant enzyme. The binding and catalytic properties of the enzyme were investigated using natural substrates, T[6-4]T and T[6-4]C, and the Dewar isomer of (6-4) photoproduct and substrate analogs s5T[6-4]T/thietane, mes5T[6-4]T, and theN-methyl-3′T thietane analog of the oxetane intermediate. The enzyme binds to the natural substrates and to mes5T[6-4]T with high affinity (K D ∼10−9-10−10 m) and produces a DNase I footprint of about 20 base pairs around the photolesion. Several lines of evidence suggest that upon binding by the enzyme, the photoproduct flips out of the duplex. Of the four substrates that bind with high affinity to the enzyme, T[6-4]T and T[6-4]C are repaired with relatively high quantum yields compared with the Dewar isomer and the mes5T[6-4]T which are repaired with 300–400-fold lower quantum yield than the former two photoproducts. Reduction of the FAD cofactor with dithionite increases the quantum yield of repair. Taken together, the data are consistent with photoinduced electron transfer from reduced FAD to substrate, in a manner analogous to the cyclobutane pyrimidine dimer photolyase.

genes for the  photolyase from Drosophila melanogaster (11) and from Xenopus laevis (12) has provided the opportunity for in-depth study of this enzyme.
Surprisingly, the genes for the (6-4) photolyase apoenzyme exhibit a high degree of sequence identity to those of PyrϽϾPyr photolyases, in particular to the type I sequence photolyases which include most of the microbial enzymes (9,11). This finding suggested that  photolyases may have structure and reaction mechanisms similar to those of PyrϽϾPyr photolyases. Indeed, a reaction mechanism encompassing the main features of photoreversal by PyrϽϾPyr photolyase was advanced for the  photolyase. It was proposed that  photolyase repairs the (6-4) photoproduct by first converting it to the oxetane intermediate (13). The cloning of the D. melanogaster  photolyase gene has provided us the opportunity to produce recombinant protein in sufficient quantities for biochemical characterization. In this study we report the overproduction and purification of Drosophila  photolyase and the probing of the action mechanism of the enzyme with regard to binding specificity and catalytic mechanism by using conventional substrates, substrate analogs, and presumed transition state analogs. Remarkably, despite the lack of obvious structural similarities between PyrϽϾPyr and Pyr  Pyr, this study and the accompanying paper by Todo and colleagues (14) on X. laevis photolyase show that the (6-4) photolyase binds to its substrate and repairs it by a mechanism quite similar to the recognition and repair of PyrϽϾPyr by its cognate photolyase.
Construction of MBP-PL  Expression Vector-The plasmid pDM64PR containing the cDNA of Drosophila (6-4) photolyase (11) was a kind gift of Dr. T. Todo (Kyoto University). The cDNA within this plasmid was amplified by polymerase chain reaction using primers designed for in-frame fusion of the malE gene of pMal-c2 and the (6-4) photolyase gene. The sequences of the primers were as follows: forward 21-mer, TCC GGA ATT CCC AGG GTC GAC; reverse 18-mer, GGA TCC TCT AGA GCG GCC. Following polymerase chain reaction amplification, the amplified 2-kilobase pair fragment was digested with EcoRI and XbaI endonucleases and inserted into the pMal-c2 vector to obtain malE-phr  in-frame fusion. This plasmid was named pXZ1997. In this plasmid the fusion joint sequence is as follows: GGA AGG ATT TCA GAA TTC CCA GGG TCG ACC CAC GCG TCC GAA ATG (initiation codon of (6-4) photolyase).
Purification of MBP-PL )-E. coli UNC523/pXZ1997 was grown in 2 liters of Luria broth to A 600 ϭ 0.6 to 0.8 at which point isopropyl-␤-D-thiogalactoside was added to 0.3 mM, and incubation was continued for an additional 6 h. The cells were harvested by centrifugation; cellfree extract was prepared, and the MBP-PL-(6-4) fusion protein was purified through a 20-ml amylose column as described by the manufac-turer (New England Biolabs). The material purified through the affinity resin was loaded onto a 10-ml heparin-agarose column equilibrated with 100 ml of buffer B (50 mM Tris⅐HCl, pH 7.5, 1 mM EDTA, 10 mM ␤-mercaptoethanol, and 10% glycerol) ϩ 0.1 M KCl. The column was washed with 100 ml of the same buffer, and then the  photolyase was eluted with a gradient of 0.1-0.5 M KCl in buffer B. Fractions containing the enzyme were identified by SDS-polyacrylamide gel electrophoresis, combined, and concentrated with Centricon-30 (Amicon) filter, and dialyzed against storage buffer which contained 50 mM Tris⅐HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, 5 mM dithiothreitol, and 50% (v/v) glycerol.
To obtain  photolyase without the MBP fusion, 50 g of MBP-PL-  was mixed with factor Xa protease (5 g) in 100 l of reaction mixture (20 mM Tris⅐HCl, pH 8.0, 100 mM NaCl, 2 mM CaCl 2 ) and was incubated for 4 h at 22°C. The reaction mixture was then loaded onto a 2-ml heparin-agarose column equilibrated with buffer B ϩ 0.1 M KCl and washed with 5 column volumes of buffer B ϩ 0.1 M KCl;  photolyase was then eluted with buffer B ϩ 0.5 M KCl. The fractions containing the enzyme were identified by SDS-polyacrylamide gel electrophoresis, combined, concentrated, dialyzed in storage buffer, and stored at Ϫ80°C. The concentration of (6-4) photolyase was determined from the absorbance of FAD upon release of the cofactor at low pH (16,17).
Spectroscopy-The absorption spectrum was obtained with a Hewlett-Packard 8451A spectrophotometer, and the fluorescence spectra were recorded using a Shimadzu RF5000 V spectrofluorimeter. The spectrum of the native protein was measured in storage buffer. To obtain the spectra of the chromophores, the enzyme was first denatured in 0.8% SDS and 0.1 M HCl by heating at 95°C for 10 min. Fluorescence spectra were recorded, and then 5 M NaOH was added to the mixture to adjust the pH to 10 -11. Under this condition, the FAD fluorescence is quenched (17), making it possible to detect the fluorescence of other species such as pterins which are known to be present in some photolyases and blue-light photoreceptors (16).

Synthesis of d(GAGTAmes 5 T[6-4]TATGAG)-
The thietane substrate analog, d(GAGTAs 5 T[thietane]TATGAG), was synthesized according to the procedure described for the synthesis of d(GTAs 5 T[thietane]TATG) (18) and was methylated with 13 CH 3 I (Sigma) by adapting a procedure of methylating oligonucleotides containing 4-thiothymidine (19). 250 l of 334.4 M thietane was dried in a Speedvac concentrator and then dissolved in 2 ml of potassium phosphate buffer, pH 8, together with 200 l of N,N-dimethylformamide. 10 l of 13 CH 3 I was added to the mixture and stirred at 25°C for 3 h. The mixture was extracted with ethyl ether (3 ϫ 2 ml), and the trace of ethyl ether was removed under a gentle stream of nitrogen gas. The aqueous phase was dried in a Speedvac concentrator and was dissolved in 0.4 ml of doubly distilled water and filtered with a 0.45-m Nylon-66 syringe filter. The filtered solution was then purified by reverse phase high pressure liquid chromatography with 75 mM potassium phosphate, pH 6.6, and 40% methanol in 75 mM potassium phosphate, pH 6.6. The major fraction was collected, concentrated to dryness, and desalted by dissolving in doubly distilled water. The sample was then injected into a reverse phase column and eluted with 50% aqueous acetonitrile. The purified product was characterized by its inability to be photoreversed by 254 nm irradiation, and the fact that raising the pH to 10 did not lead to any UV absorption increase at 326 nm as was observed for the thietane. 1  Substrates-Oligonucleotides used in this work were constructed by ligating photoproduct-containing oligomers, 6 -12 nucleotides in length, with a three-oligomer scaffold to obtain duplexes of 49, 55, or 56 bp. 10 pmol of photoproduct-containing oligomer was phosphorylated and mixed with 100 pmol each of the three other oligomers which were phosphorylated separately. The mixture (50 l) in annealing buffer (50 mM Tris⅐HCl, pH 7.4, 100 mM NaCl) was heated to 65°C for 10 min, cooled to 16°C, and then adjusted to 10 mM MgCl 2 plus 2 mM ATP, and 50 units of T4 DNA ligase was added. Ligation was for 10 h at 16°C. Then the duplexes were purified by gel electrophoresis on both denaturing and native polyacrylamide gels and stored in annealing buffer at Ϫ20°C. The oligomer building blocks were as follows: The preparation and characterization of the latter two oligomers were carried out as described previously for the corresponding products of d(GTATTATG) (18). The substrates were internally or 5Ј terminally labeled by phosphorylating the appropriate oligomer with [␥-32 P]ATP and T4 polynucleotide kinase. The concentrations of the internally labeled substrates were determined from their specific activity, and those of the terminally labeled substrates were determined spectrophotometrically.
Binding Assays-The apparent equilibrium binding constant and the dissociation rate constant were determined by gel-electrophoretic mobility-shift assay as described previously (22,23). The bound and unbound fractions of DNA were quantified by scanning the autoradiograms on a Computing Densitometer (Molecular Dynamics). The apparent equilibrium dissociation constant (K D ) was obtained by using the SigmaPlot program based on 1:1 stoichiometry of substrate to enzyme. The dissociation rate constant was determined from the slope of the decay curve of the first-order rate plot of enzyme-substrate complexes following the addition of competitive DNA to 25-fold molar excess over the radiolabeled substrate. All binding assays were conducted under yellow light from General Electric "Gold" fluorescent lights.
Enzymatic and Chemical Probing of DNA-Protein Complexes-DNase I footprinting assays with 120-bp duplex containing a centrally located T[6-4]C were conducted under yellow light as described previously (23). To detect base flipping upon binding of (6-4) photolyase to substrate, we performed thymine permanganate cleavage assays. Permanganate selectively modifies unpaired thymines, making them susceptible to cleavage by piperidine. The 120-bp T[6-4]C substrate (0.5 nM) was incubated with 2 nM (6-4) photolyase in 30 l of binding buffer at 23°C for 10 min, after which KMnO 4 (1 mM) was added and incubation was continued for 1 min. The reaction was then quenched by the addition of ␤-mercaptoethanol to 5 mM. The DNA was precipitated with ethanol, dissolved in 10 l of 1 M piperidine, and heated at 90°C for 30 min. The DNA was then lyophilized, resuspended in 10 l of formamide/ dye mixture, and resolved on an 8% denaturing polyacrylamide gel.
Repair Assays-The repair of DNA by  photolyase was tested by either coupled enzyme assay or band shift assay as described previously (24). In coupled enzyme assay, a restriction enzyme site such as TTAA (MseI), which has become refractory to cleavage because of the photoproduct at the TT site, is rendered restriction enzyme-sensitive by the repair reaction. In the band shift assay, samples from an enzyme/ substrate mixture with saturating amount of enzyme are exposed to photoreactivating light, and the samples are then subjected to gel electrophoresis on a non-denaturing polyacrylamide gel. Under these experimental conditions, initially all of the substrate is bound to the enzyme specifically and hence the entire substrate is in the slower migrating band in the band shift assay. Upon exposure to light the repaired substrate (product) dissociates from the enzyme and migrates as "free" DNA. The photoreactivation experiments were done under aerobic conditions unless stated otherwise. Photoreactivation with dithionite-reduced enzyme was conducted anaerobically as described previously (24). The amount of free DNA in the band shift assay is a measure of (6-4) photolyase repair activity.
Primer extension assay (13) was used to test the reaction product resulting from photoreactivation of T[6-4]C. A 17-mer primer with the sequence 5Ј-ATT ACG AAT TGC CTC AT (2.5 nM) was labeled with [␥-32 P]ATP and annealed to a photoreactivated 55-mer substrate containing T[6-4]C (0.5 nM) in 10 l of DNA polymerase buffer (10 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 7.5 mM dithiothreitol, 2 M of each individual dNTPs). The primer extension was initiated by adding 0.5 unit of Klenow fragment of DNA polymerase I. After incubation at 23°C for 2 min, the reaction was quenched by adding 25 mM EDTA. The DNA was then lyophilized, resuspended in 10 l of formamide/dye mixture, and resolved on a 12% denaturing polyacrylamide gel.

RESULTS
Purification of (6-4) Photolyase-Using the cloned cDNA for Drosophila (6-4) photolyase, we constructed vectors for expressing the (6-4) photolyase without peptide fusion in Sf 21 insect cells and in E. coli. The protein made in insect cells was soluble, but the expression level was too low for purifying sufficient enzyme for biochemical characterization. The (6-4) photolyase without a peptide fusion was greatly overproduced in E. coli, but more than 99% of the protein was insoluble, and attempts to reconstitute the enzyme from inclusion bodies were unsuccessful. Hence, we decided to express the (6-4) photolyase as a fusion protein with the E. coli maltose binding protein (MBP). It is known that MBP confers solubility on heterologous proteins otherwise resistant to be expressed in soluble form in E. coli (15). Fig. 1A shows the physical map of the plasmid construct used for expressing the (6-4) photolyase in the form of MBP-PL-(6-4) fusion protein. The fusion protein exhibits the same binding properties of the natural protein (see below), and hence, because of its ready availability, it was used in most of our experiments for characterizing the (6-4) photolyase.
As shown in Fig. 1B, cells containing the expression plasmid, upon induction, synthesized the MBP-PL-(6-4) fusion protein to a level of about 5% of total cellular proteins. The fusion protein was purified by affinity chromatography on amylose resin, and after this step the protein was Ͼ90% pure. We conducted most of our experiments with this fusion protein. For certain experiments the (6-4) photolyase was separated from MBP by digesting the fusion protein with factor Xa which cleaves at the linker region between MBP and (6-4) photolyase [PL ]. Following proteolytic cleavage, PL(6-4) was purified away from MBP by chromatography on heparin-agarose. After this purification step, although the yield was low, the (6-4) photolyase was highly pure (Fig. 1B, lane 4).
Spectroscopic Properties of (6-4) Photolyase-All of the photolyase/blue-light photoreceptor family proteins investigated to date contain flavin adenine dinucleotide as a cofactor (12,15,16,25,26). In addition, these proteins contain a second chromophore which may be a pterin (folate) or a deazaflavin depending on the source of the enzyme (27,28). Since neither Drosophila nor E. coli can synthesize deazaflavin (29), this cofactor was not considered as a potential candidate for a chromophore in  photolyase. Instead, we conducted assays to detect flavin and pterin in the recombinant enzyme preparation. Fig. 2A shows the absorption spectrum of the purified enzyme. The spectrum is quite similar to those of Drosophila PyrϽϾPyr photolyase and to plant and human blue-light photoreceptors (15,16,30) which were shown to contain FAD and a pterin. To find out if the (6-4) photolyase contained flavin, the enzyme was denatured at pH 1 and analyzed by fluorescence spectroscopy. Fig. 2B shows that under these conditions, a species with fluorescence excitation and emission spectra typical of flavin is released. Upon raising the pH, the flavin fluorescence is quenched indicating that the cofactor is FAD and not FMN (17). Furthermore, high pH revealed the presence of another fluorescent species with an excitation maximum at 360 nm and emission maximum at 440 nm (Fig. 2C). These properties are consistent with a pterin species as the second chromophore (27). Indeed, this blue fluorescent species was separated from FAD and other contaminants by high pressure liquid chromatography (data not shown). However, the amounts obtained after the purification steps were not sufficient for unambiguously identifying the second chromophore by other analytical methods as a pterin derivative.
Binding of (6-4) Photolyase to Substrates and Substrate Analogs-We investigated the binding of (6-4) photolyase to various substrates by gel mobility shift assay and by DNase I footprinting and chemical probing. We synthesized six different substrates or substrate analogs designed to probe the repair mechanism ( Fig. 3A) and incorporated them into 49 -56-mers to conduct binding and catalysis studies (Fig. 3B). The substrate analogs were synthesized to test the model that catalysis by (6-4) photolyase proceeded through an oxetane intermediate (13,26). However, it was necessary to demonstrate high affinity binding before these compounds could be used as mechanistic probes. Because of its ready availability, MBP-PL-(6-4) was used in majority of our experiments. In order for the results obtained with the fusion protein to be a true representation of the natural protein, we wished to establish that the two forms behaved similarly in a binding assay.
MBP-PL-(6-4) or (6-4) photolyase cleaved from MBP by factor Xa and purified free of MBP was mixed with various substrates. The protein-DNA complexes were then separated on a 5% non-denaturing polyacrylamide gel. Fig. 4 shows the results of these binding experiments. Neither the fusion protein nor the full-length (6-4) photolyase bound to undamaged DNA under these experimental conditions, which were designed to bind about 50% of the substrate (lanes 1 and 2). Both forms bound to T T and T C with about equal affinity (lanes 3-6), with the complexes formed with the full-length protein migrating faster than those formed with fusion protein (lanes 4 and 6 versus lanes 3 and 5). As is evident from this figure, both forms bound very weakly to the thietane analog of (6-4) photoproduct (lanes 7 and 8) which was designed to test certain predictions regarding the reaction mechanism. The important conclusion, however, that emerges from this figure is that the fusion and non-fusion forms of the (6-4) photolyase have very similar binding properties and hence the binding characteristics deter- mined for the fusion protein should be applicable to the nonfusion form as well.
To determine the equilibrium binding constant, we gener-ated binding isotherms under conditions of constant substrate and variable enzyme concentrations. The results of a typical assay are shown in Fig. 5. The apparent K d values calculated from these experiments as well as data from similar experiments with other substrates are summarized in Table I. As apparent, the enzyme binds with high affinity and specificity to its natural substrates, with subnanomolar dissociation constants for both T T and T C. The binding to the Dewar isomer is 3-4-fold weaker than the precursor photoproducts. Of the substrate analogs, mes 5 T T bound with the same affinity as the natural photoproducts, but the other two analogs did not bind at a level measurable above background.
To study the kinetics of the binding reaction, we determined the dissociation rate constant (k off ) by band shift assay using the competing substrate method (22,23). In this method, to a pre-formed enzyme-substrate complex, 25-fold excess non-radiolabeled substrate was added, and at various time points following the addition of the competing DNA, samples were loaded onto a native polyacrylamide gel under constant voltage. Autoradiograms of such gels revealed a "Canadian geese" pattern where the distance between the free and bound DNA and the amount of bound DNA decrease with increasing time points. An example of such gel for T C is shown in Fig. 6A, and quantitative analysis of the data is in Fig. 6B. As apparent from the figure, the complex formed with the T[6-4]C substrate dissociates rather slowly with k off ϭ 1.4 ϫ 10 Ϫ3 s Ϫ1 (t1 ⁄2 ϭ 8.3 min). The off rates for the other substrates, in general, reflected the magnitude of the equilibrium binding constants as shown in Table I. It is thus concluded that the formation of the enzyme-substrate complex is diffusion controlled and that the main source of difference in the binding constants is the difference in the off rates of complexes formed with different substrates.
Base Flipping by  Photolyase-The (6-4) photolyase, either in the form of fusion protein or in isolation, makes an approximately 20-bp DNase I footprint which is nearly symmetrical around the photoproduct and which in many ways resembles the DNase I footprint of PyrϽϾPyr photolyase (23) on its cognate substrate (Fig. 7A). However, in addition to the protection footprint, the (6-4) photolyase also induces DNase I-hypersensitive sites (Fig. 7A marked by filled circles) that are ascribed to opening of the minor groove by kinking (23). Thus, it appears that binding of (6-4) photolyase induces significant conformational change in the substrate.
One important conformational change induced by a number of DNA modifying or repair enzymes is base flipping. In this reaction, the base(s) is extruded from within the duplex to a cavity in the enzyme, and depending on the enzyme involved, the base is either methylated, cleaved at the glycosylic bond, or repaired before dissociation of the enzyme-substrate complex (31)(32)(33). Although base flipping was originally discovered with M. HhaI (cytosine-5) DNA-methyltransferase (34 -35), it was later found in DNA glycosylases as well (see Ref. 31). Of special relevance to our topic, both the crystal structure of E. coli DNA photolyase (36) and biochemical analyses of its reaction mechanism (see Ref. 6) provided strong evidence that photolyase also operated by a base flipping mechanism (36). It was proposed that photolyase flips out the thymine dimer from within the duplex to a cavity within the enzyme and following photorepair releases the repaired dinucleotide moiety (36). Considering the structural, and perhaps catalytic mechanism, similarities between the classical photolyases and the (6-4) photolyase, we wished to find out if the (6-4) photolyase employed base flipping as part of its reaction mechanism.
Although crystallography is the most definitive methodology for proving base flipping, there are certain biochemical meth- ods that can provide valuable indirect evidence on whether or not base evisceration occurs in an enzyme-substrate complex. Three of these methods were employed with (6-4) photolyase.
First, if base flipping is part of the binding/recognition mechanism, then if the target base(s) are within a mismatch sequence (bases complementary to photoproduct), and they are expected to bind with higher affinity compared with normal duplex because of the lower free energy of flipping out of a mismatched base (34). The second test that is related to the first one is based on the observation that nearly all enzymes that employ base flipping as part of their recognition mechanism bind about equally well to single-and double-stranded DNA substrates, presumably because the flexibility of the target base(s) in single strand compensate for any loss of binding free energy provided by contacts made between the comple-mentary strand and the enzyme with duplex substrate (23,36,37). Fig. 7B shows that (6-4) photolyase binds to singlestranded substrate and to double-stranded substrate with two mismatches, with higher affinity (K D ϭ 2.5 ϫ 10 Ϫ10 M) than a duplex with a (6-4) photoproduct (K D ϭ 5 ϫ 10 Ϫ10 M). These data are consistent with the notion that binding of (6-4) pho- A third biochemical test for base eversion is to probe for unpaired bases at the target site upon enzyme binding (38,39). When we probed the complementary strand of a T[6-4]C substrate with permanganate which specifically reacts with unpaired thymines (40), it was found that the two residues on either side of the AG across from T[6-4]C became hypersensitive to oxidation by permanganate (Fig. 7C). Thus, binding of (6-4) photolyase confers single strandedness to about a 4-bp region at the site of lesion. Although these individual biochemical experiments in isolation do not prove base flipping, taken together, our data are consistent with the flipping out of the (6-4) photoproduct during binding and catalysis by (6-4) photolyase.
Assays for DNA Repair by (6-4) Photolyase-The (6-4) photolyase binds its substrate independent of light and initiates  catalysis only upon absorbing a near UV-visible photon, as does PyrϽϾPyr photolyase (10,41). To study the catalysis by the purified enzyme, we used three assays. One is a coupled enzyme assay and is based on restoration of restriction enzyme sensitivity of a recognition sequence that has become resistant to a particular restriction enzyme because of the photoproduct (13,24,41). An example of this assay is shown in Fig. 8A. As apparent, the T[6-4]T photoproduct is repaired, but the two substrate analogs are not. The repair of s 5 T[thietane]TAA would generate s 5 TTAA which is cleavable by MseI, but the N 3 -methylthymidine reversion product of s 5 T[thietane]m 3 T may not be susceptible to cleavage by MseI. The lack of repair of this photoproduct was firmly established by the second assay (binding assay) which measures the loss of binding site as a result of photorepair (data not shown). Fig. 8B shows an example of the binding assay with a repairable photoproduct where binding is measured by the gel electrophoretic mobility shift assay.
Finally, restoration of normal bases by photoreactivation was tested by the primer extension assay. Previously, we showed that photoreactivation treatment restored T T photoproduct to normal bases as evidenced by the primer extension assay which measures the base incorporated across from the photoproduct (13). Fig. 8C shows that the repair of T C photoproduct also restores cytosine to its normal coding properties, and hence reversal of this lesion represents a nonmutagenic repair as in the case of T T photoproduct. The binding assay is technically simpler and is less subject to experimental variability; hence, it was the assay of choice for determining the action spectrum.
Action Spectrum-An action spectrum may be defined as the rate of a photochemical reaction as a function of wavelength. It yields information about the nature of the chromophore involved in the photochemical reaction. In contrast, an absolute action spectrum is a plot of ⑀ (extinction coefficient ϫ quantum yield) as a function of wavelength; it provides a quantitative value of the efficiency of the photochemical reaction, and provided that the extinction coefficient at each wavelength is known, it enables the calculation of the quantum yield of repair, that is the probability that an absorbed photon would catalyze a reaction.
The absolute action spectrum of (6-4) photolyase was determined using the T T and T[6-4]C as substrates. By conducting the photoreactivation reaction under enzyme excess conditions, the reaction is rendered pseudo first-order with respect to the light fluence. From the slope of the first-order rate plot, k p , the photolysis constant is obtained; k p is related to the photolytic cross-section by the following equation: ⑀ ϭ k p (mm 2 ⅐erg Ϫ1 ) ϫ 5.2 ϫ 10 9 / (nm) (42). In practice, the extent of repair is determined by the band shift assay. An example of this test conducted with T[6-4]C substrate and 410 nm light is shown in Fig. 8B. Quantitative analyses of data from Fig. 8B and of data from similar experiments conducted with T T and T C at other wavelengths are shown in Fig. 9. From the slopes of the lines in the latter figure, the absolute action spectrum was constructed (Fig. 10). The action spectrum resembles the absorption spectrum of the enzyme in the near FIG. 8. Repair assays for (6-4) photolyase. A, restriction site restoration assay. The indicated substrates (0.1 nM) were mixed with 0.75 nM  photolyase in 50 l of binding buffer and incubated at 20°C for 30 min. The mixtures were then irradiated with 366 nm from a Sylvania Blacklight at a fluence rate of 2 milliwatts/cm 2 for 45 min. The DNA was then extracted with organic solvents and digested with MseI restriction endonuclease. All three photoproducts are in TTAA sequence context, and repair of the photolesions enables MseI to cut between the two Ts, generating the 21-mer product. In lane 2, 51% of the DNA was cleaved to 21-mer, and in lane 4, 4% was digested. In all other lanes there was no cleavage above background. B, band shift assay. DNA (0.1 nM T[6-4]C) was mixed with 0.75 nM photolyase in 40 l of binding buffer, and following incubation at 20°C for 30 min, the reaction mixtures were exposed to 410 nM light (40 microwatts/cm 2 ) for the indicated times and then separated on a 5% native polyacrylamide gel. Lane 1 contained DNA without enzyme; lane 8 contained DNA without enzyme that was irradiated for 60 min; lane 9 contained DNA mixture that was exposed to photoreactivating light for 60 min and then mixed with  photolyase. C, primer extension assay. A terminally labeled 17-mer was annealed to a 55-mer with a centrally located T C that had been exposed to photoreactivating (PR) treatment (ϩ) or not (Ϫ) and then extended by Klenow fragment in the presence of the indicated dNTPs (2 M). Following incubation at 23°C for 2 min the reaction products were analyzed on a 12% denaturing polyacrylamide gel.
UV-visible region. However, because of the "tail" absorption/ scattering of the recombinant enzyme preparation in the 300 -400 nm region, the absorption and action spectra diverge at shorter wavelengths. It is most likely that the absorption spectrum of the primary photoantenna of (6-4) photolyase is similar to the action spectrum shown in Fig. 10.
Mechanism Probes with Substrate Analogs-It was previously proposed that (6-4) photolyase repairs (6-4) photoproduct by first converting it to the oxetane intermediate which can then be readily split to constituent bases by a concerted or stepwise retro[2ϩ2] reaction (13). A key element of this model is formation of the oxetane intermediate thermally. Studies with model systems supported the feasibility of this mechanism (18,43). Hence, we reasoned that the oxetane form of the (6-4) photoproduct would be efficiently repaired by the enzyme and that substrate analogs which cannot form this intermediate would be resistant to photorepair. Toward this end we tested the following substrates (Fig. 3A): the Dewar isomer (2) which cannot form the oxetane intermediate; the thietane analog of oxetane intermediate (5) which is in equilibrium with the s 5 -(6-4) product and predicted to be repaired efficiently; the N-methyl-3Ј-T thietane analog of the oxetane intermediate (6) which is locked in the thietane form and was predicted to be split very efficiently; and the s 5 -methyl analog of (6-4) product (4) which presumably cannot be reversed to thietane and was predicted to be refractory to photorepair.
Neither the thietane analog (5) nor the N-methyl-3Ј T analog (6) which is permanently locked in the thietane ring form bound the enzyme, and hence we were unable to test the effect of these presumptive transition state analogs on catalysis. In contrast, both the Dewar form of T[6-4]T (2) and the S 5 -methyl analog of T[6-4]T (4) which presumably cannot form the oxetane intermediate bound the enzyme with high affinities (Table  I) and hence proved useful as mechanistic probes. Unexpectedly, both T[Dewar]T and the S 5 -methyl analog of T T were repaired by the enzyme. The results of photoreactivation experiments with these substrates are shown in Fig. 11A. Under saturating enzyme concentration both substrates are repaired with about equal rates. More importantly, even though these substrates are repaired by the enzyme, the rate of repair and hence the quantum yield of repair is about 0.3% that obtained with reference to (6-4) photoproducts of TT or TC. These slow repair rates determined by band shift assay were confirmed by restriction site susceptibility assay for T[Dewar]T (Fig. 11B) and by primer extension assay for the S 5 -methyl analog of T T (data not shown) and hence represent true repair and not a side reaction which causes the enzyme to dissociate from substrate. The repair of these lesions, even though not supportive of an oxetane intermediate, does not rule out such an intermediate during repair of natural substrates.

DISCUSSION
This study extends the previous investigations on (6-4) photolyase which were conducted with cell-free extracts (10,13) or partially purified enzyme preparations (11,41). However, even though this study goes a long way toward characterizing the (6-4) photolyase, it has certain shortcomings. First of all, even though the majority of the enzyme made in the form of fusion protein in E. coli is soluble and thus apparently properly folded, it has substoichiometric amounts of the two cofactors. The stoichiometry of FAD to apoenzyme ranged from 0.01 to 0.05. The active enzyme concentration was based on FAD content because apoenzyme without FAD neither binds to nor can repair substrate DNA (24). The stoichiometry of the pterin cofactor was even lower. In fact the low level of the pterin cofactor precluded definitive identification of the second chromophore in (6-4) photolyase. It is known that the pterin cofactor is rather loosely bound in many photolyases and, in fact, even with E. coli photolyase purified from its natural source, quite often the purified enzyme contains very little pterin in the form of folate (27). Second, because of the low stoichiometry of FAD to apoenzyme, it was not possible to unambiguously establish the oxidation state of the flavin cofactor. However, the strong absorbance in the 420 -430 nm region and strong fluorescence at 520 nm indicate that most of the flavin in the enzyme is oxidized. Despite these shortcomings of the recombinant (6-4) photolyase, we feel that this study has uncovered an important point regarding the  photolyase and that similarities between this enzyme and the classical PyrϽϾPyr photolyase exist in almost every aspect. The accompanying paper (14) on Xenopus photolyase provides additional evidence for the similarities between the (6-4) photolyase and PyrϽϾPyr photolyase. Although some minor differences exist between the Drosophila and Xenopus (6-4) photolyases with regard to substrate specificities and catalysis, these could reflect interspe- Binding-The PyrϽϾPyr photolyase and (6-4) photolyase act on two rather different substrates. Although both PyrϽϾPyr and (6-4) photoproduct are produced by ultraviolet light, the structures of the two dipyrimidine photoproducts and their effects on DNA conformation are quite dissimilar (see Ref. 49). Yet, surprisingly, the two enzymes appear to be evolutionarily and structurally related (11). In this report we show that the similarities extend to substrate recognition and catalysis as well.
The PyrϽϾPyr photolyase contacts a region of about 20 nucleotides around the photoproduct (23,37) and binds to single-and double-stranded DNA substrates with about equal affinities (22,23). The crystal structure of E. coli photolyase strongly suggests that during catalysis the photodimer is flipped-out into the flavin binding pocket in the center of a groove of positively charged residues running the length of the molecule (36). Here we report that (6-4) photolyase makes a DNase I footprint similar to those of PyrϽϾPyr photolyases, that it binds to single-and double-stranded substrates with about equal affinities, and with binding constants comparable to those of PyrϽϾPyr photolyases. Importantly, by at least two different experimental approaches (affinity of the enzyme to a mismatched 6-4 photoproduct and permanganate hypersensitivity of neighboring thymines upon enzyme binding), we find evidence that  photolyase also flips out the photodimer during catalysis. Despite these similarities in substrate binding, there are some differences as well. Specifically, PyrϽϾPyr photolyase has a strict requirement for normal bases in the photodimer (see Ref. 4); in contrast, we find that (6-4) photolyase recognizes with high affinity the Dewar isomer which is structurally quite different than (6-4) photoproduct, and the S 5 -methyl analog of T T which has a methyl on the 5-thiol group which is transferred during catalysis.
Catalysis-Two general pathways were previously proposed for the photoreversal of T T to the normal bases that were consistent with the available data and made chemical sense (13). Both pathways involved formation of a four-membered oxetane ring before splitting the photoproduct. An oxetane intermediate appeared to be a viable intermediate considering that it is also the presumed intermediate in the photochemical formation of the T T product (see Ref. 3). In one general pathway, formation of the oxetane is catalyzed by the enzyme in a thermal step and is then split in either a photooxidative or photoreductive step. In the other general pathway, the oxetane is produced in a photooxidative step and is then split by the enzyme in a thermal step. Because (6-4) photolyase containing reduced FAD appears to be the active form of the enzyme, we consider photooxidative steps unlikely, and we will therefore focus our discussion on the photoreversal of (6-4) photoproducts by the photoreductive pathway. Support for a photoreductive mechanism comes from model studies that have shown that oxetane adducts of 1,3-dimethylthymine and either benzaldehyde or benzophenone can be efficiently reversed by photoreduction (43). Likewise, repair of the Dewar isomer is also consistent with a photoreductive pathway, in that the highly strained Dewar isomer is expected to be converted by photoreduction to the (6-4) isomer, which would then be reverted in a second photoreductive step. The Dewar isomer may not be as good an electron acceptor as the (6-4) isomer, which would explain the lower efficiency of the reaction.
A key aspect of the photoreductive pathway hinges on the ability of the enzyme to thermally catalyze the formation of an intermediate oxetane from TT, azetidine (or protonated azetidine) from TC (44), or thietanonium ion from Tsme 4 T photoproducts (45,46) (Fig. 12A). Although it has been shown that the T[thietane]T is more stable than the s 5 T[6-4]T isomer in both dinucleotides and oligonucleotides (18,47), the putative oxetane intermediates produced in the photochemical reaction of TT do not appear to be the stable form above Ϫ80°C (48). Recent theoretical calculations on the base portion of the (6-4) photoproduct of TT and its oxetane isomer estimated the gap to be about 14.5-16.5 kcal/mol (50). One indication that the calculated energy differences may be greatly overestimated comes from the fact that the same study predicted the gap between the (6-4) photoproduct and the corresponding thietane was 5 kcal/mol where in fact, in a dinucleotide, the thietane is favored 3:1 over the  isomer (47), and is the only isomer observable in an octanucleotide (18). One possible explanation for the lower energy difference between the (6-4) and thietane isomers in dinucleotides and DNA than calculated for the base portion , control substrate incubated with enzyme but not exposed to light, 0.4%. Lane 8, control reaction containing substrate which was exposed to light for 8 h in the absence of enzyme, 0.1%. only is that the (6-4) isomer may be more strained than the four-membered ring isomer when incorporated into a ring system formed by the sugar phosphate backbone. Because 3Ј-C-4 of the (6-4) isomer is sp 2 -hybridized, the bond between 5Ј-C-6 and 3Ј-C-4 would normally be coplanar with the pyrimidone ring, and energy might have to be expended to bend the pyrimidone ring to accommodate within the ring formed by the sugar phosphate backbone. Not as much energy would have to be expended for the four-membered ring isomer, however, because the bond between 5Ј-C-6 and 3Ј-C-4 is naturally bent out of the plane of the pyrimidone ring due to the sp 3 hybridization of 3Ј-C-4 (Fig. 12A). In this regard, original molecular modeling calculations on (6-4) photoproducts in DNA yielded structures with a highly bent pyrimidone ring (51). A partially bent pyrimidone ring in the (6-4) product would also explain why the corresponding Dewar product, which has a highly bent ring, binds to (6-4) photolyase with relatively high affinity (Table I).
It is also likely that (6-4) photolyase can further stabilize the oxetane intermediate relative to the (6-4) isomer through critically placed amino acid side chains. In this regard, the proposed formation and stabilization of the oxetane intermediate by addition of the 5Ј-C-5-OH to 3Ј-C-4 bears a striking similarity to the formation and stabilization of the product of the otherwise highly unfavorable (K eq ϭ 4.7 ϫ 10 Ϫ6 ) addition of water to the C-4 of zebularine that is carried out by cytidine deaminase (52)(53)(54). Zebularine is an analog of cytidine in which the cytosine ring is replaced with a pyrimidone ring that is identical in structure to the pyrimidone ring of the (6-4) product of TC, except that C4 substituent is replaced by a hydrogen.
In cytidine deaminase there is an active site glutamate (54), and in the related adenosine deaminase which catalyzes the hydrolysis of adenosine to inosine, there are two aspartates and one glutamate (55,56). The crystal structure of E. coli photolyase shows that there are three carboxylates (Glu-279, Asp-372, and Asp-374) in the substrate binding site (36). These residues are conserved in the (6-4) photolyase (Glu-304, Asp-397, and Asp-399, respectively; see Refs. 11 and 12) and hence are in a position where they could act as general acid or base to facilitate the formation of the proposed oxetane intermediate. One of these carboxylate side chains could increase the nucleophilicity of the 5Ј-C-5 oxygen of the (6-4) product by deprotonating it, whereas another carboxylic acid side chain could simultaneously activate the 3Ј-C-4 to nucleophilic attack by protonating 3Ј-N-3 to generate a highly electrophilic N-acyliminium ion (57) (Fig. 12B). The same pair of carboxylates could also be envisioned to catalyze the formation of an azetidine intermediate (or its protonated form) (Fig. 12A). Formation of the positively charged thietanonium ion intermediate could likewise be catalyzed by protonation of the 3Ј-N-3 of the pyrimidone ring and stabilized electrostatically by a negatively charged carboxylate side chain. The failure of the s 5 T[thietane]m 3 T to bind to the enzyme is also consistent with this mechanism, protonation of the 3Ј-N-3 of the pyrimidone ring is sterically blocked by a bulky methyl group.
Conclusion-In consideration of all the possible mechanisms and all the available data, enzyme-catalyzed formation of an oxetane or an azetidine intermediate followed by photoreversal catalyzed by electron transfer from a reduced flavin would appear to be the most likely mechanism for DNA repair by  photolyase. Although neither of the two thio analogs of the proposed oxetane or azetidine intermediates were substrates for the enzyme, they do not rule out these intermediates in the catalytic mechanism. The lack of binding of the s 5 T[thietane]m 3 T analog is consistent with the proposed role of 3Ј-N-3 of the pyrimidone ring in substrate binding and catalysis. The lack of binding of both thietane analogs might also be due to substantial structural differences that result from differences in the lengths of C-S and either C-O or N-O bonds (1.8 versus 1.5 Å, respectively). Overwhelming evidence indicates that catalysis in PyrϽϾPyr photolyases is initiated by The role of the second chromophore as a light harvester has been omitted for clarity. photoinduced electron transfer from FADH Ϫ to PyrϽϾPyr (4 -6). Here we show that, very likely, the same type of electron transfer reaction cycle must be involved in catalysis by  photolyase. At this point, the evidence for electron transfer from FADH Ϫ to (6-4) photoproduct is preliminary. One piece of evidence is that enzyme containing only flavin and no detectable second chromophore is capable of catalysis (Ref. 12; data not shown). Second, chemical reduction of the flavin cofactor increases the quantum yield of (6-4) photolyase (Fig. 11A). In fact, this latter type of data was the first experimental evidence that led to the model that PyrϽϾPyr photolyase splits the photodimer by electron donation and not by electron abstraction (58). Clearly, further experiments aimed at identifying the reaction intermediates are needed for formal proof of the reaction mechanism of (6-4) photolyase. Nevertheless, based on our current data, on work with model systems (18,43), and by analogy to PyrϽϾPyr photolyase and cytidine deaminase, we propose the model in Fig. 12B, which is an updated version of a model proposed earlier (13,26), as a working model for catalysis. In this model the enzyme binds to the open form of the (6-4) photoproduct of either TT or TC and catalyzes formation of the oxetane or azetidine isomer, both of which collapse to the corresponding pyrimidines upon photoinduced electron transfer from a photoexcited reduced flavin and formation of the radical anion via a stepwise or concerted retro [2ϩ2] reaction.