Action Mechanism of Escherichia coli DNA Photolyase I. FORMATION OF THE ENZYME-SUBSTRATE COMPLEX*

Escherichia coli DNA photolyase (photoreactivating enzyme) is a flavoprotein. The enzyme binds to DNA containing pyrimidine dimers in a light-independent step and, upon illumination with 300-600 nm radia-tion, catalyzes the photosensitized cleavage of the cy- clobutane ring thus restoring the integrity of the DNA. We have studied the binding reaction using the tech- niques of nitrocellulose filter binding and flash photolysis. The enzyme binds to dimer-containing DNA with an association rate constant kl estimated by two dif-ferent methods to be 1.4 % 10' to 4.2 x 10' M-' s-'. The dissociation of the enzyme from dimer-containing DNA displays biphasic kinetics; for the rapidly dissociating class of complexes s-l, while for the more slowly dissociating class k2 = 1.3 X to 6 X s-'. The equilibrium association constant Ka, as determined by the nitrocellulose filter binding assay and the flash photolysis assay, was 4.7 % to 6 X M", in reasonable agreement with the values predicted from kl and kz. From the dependence of the association constant on ionic strength we conclude that the enzyme contacts no more than two phosphodiester bonds upon binding; this strongly suggests that the pyrimidine di- mer is the main structural determinant of specific pho-tolyase-DNA interaction and that nonspecific ionic in- teractions do not contribute significantly to substrate

Irradiation of DNA with 254 nm of light induces the formation of dimers between adjacent pyrimidine bases in the same DNA strand; left unrepaired such lesions result in mutation or cell death. DNA photolyases (EC 4.1.99.3) are a group of enzymes which catalyze the light-dependent reversal of pyrimidine dimers in DNA thus repairing these lesions in situ. The reaction proceeds by a two-step mechanism (Rupert, 1962): (i) the substrate binding step, which can occur in the dark, and (ii) dimer reversal, which occurs only in the presence of light, the precise wavelength being dependent upon the source of the enzyme (see Harm, 1976, for a review). Thus the overall reaction scheme can be written as the following. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Escherichia coli (Rupert et al., 1958) and the action of the enzyme has been studied extensively in vivo Harm, 1970aHarm, , 1970b, until recently the paucity of the enzyme (about 20 molecules/cell in wild type cells) had made in vitro studies of the enzyme in a fully defined system impossible. Several years ago we began a concentrated effort to obtain sufficient protein to undertake studies on the molecular mechanism of action of photolyase; to that end we cloned the phr gene (Sancar and Rupert, 1978b), constructed overproducing E. coli strains in which >15% of total cellular protein is photolyase, and purified the enzyme to homogeneity (A. Sancar et al., 1984b). Pure photolyase consists of an M,-53,994 polypeptide (G. Sancar et al., 1984) and two intrinsic chromophores, a blue neutral FAD radical and a second chromophore the chemical identity of which is not yet known Jorns et al., 1984); in vitro the enzyme is capable of repairing pyrimidine dimers in the absence of any additional proteins or cofactors and at a rate similar to that observed in vivo. Using this fully defined system we report in this and the following two papers the results of studies on the thermodynamic, kinetic, and photochemical processes involved in photoreactivation by the E. coli enzyme. In this article we deal with the first half of the reaction sequence, formation of the enzyme-substrate complex.
We have previously reported that purified photolyase binds specifically to UV-irradiated DNA and that this binding is relatively insensitive to changes in salt concentration or pH (G. Sancar et al., 1985). In this report we extend those studies by quantitating the specific association constant K A using both the nitrocellulose filter binding assay and the flash photolysis-transformation technique; in addition we have quantitated the kinetic constants for complex formation, kl and k2. Both equilibrium and kinetic methods yield estimates for KA of 4.7 X lo7 to 1.4 X 10' M-', values which are unusually low for a site-specific DNA-binding protein. In addition, quantitation of the effect of NaCl concentration on KA indicates that the enzyme contacts only 1-2 phosphates on the DNA backbone when bound to the pyrimidine dimer; taken together with the relatively low k1 (4.2 X lo6 M" s-') exhibited by the enzyme, these results suggest that in contrast to other site-specific DNA-binding proteins, nonspecific ionic interactions contribute very little to the equilibrium or kinetics of photolyase binding.

MATERIALS AND METHODS
Enzyme Preparation!-E. coli DNA photolyase was purified to homogeneity as described previously (A. Sancar et al., 1984b) with the modification that at Step VI, Blue Sepharose (Sigma) was substituted for DNA cellulose; this change enabled us to obtain pure enzyme at higher yield (about 10% versus 5.4%) than previously. The enzyme had a blue-purple color when freshly prepared due to its neutral FAD radical cofactor (Jorns et al., 1984) and concentrated 478 This is an Open Access article under the CC BY license.
solutions (5-10 mg/ml) retained this color for over a year when stored at -80 "C. We will refer to such enzyme as "blue enzyme." In contrast to the stability of the enzyme under the conditions described above, prolonged purification procedures and/or storage a t -20 "C in 40% ethylene glycol containing 10 mM 0-mercaptoethanol for more than 6 months results in the appearance of "yellow enzyme." Spectrophotometric analysis (absorption at 450 and 580 nm) indicated that 70% of the FAD in yellow enzyme was in the fully oxidized form (Jorns et al., 1987). The oxidized FAD remained firmly associated with the enzyme as overnight dialysis of the yellow enzyme did not result in a detectable decrease in the 450-nm absorption.
Visualization of ES Complexes by Electron Microscopy-DNA molecules that contained pyrimidine dimers in a slightly off-center location were prepared as follows. Plasmid pBR322 DNA was digested with EcoRI and BanHI restriction endonucleases and the (EcoRI-B U~H I )~,~ and ( B~~H I -E C O R I )~~, B fragments were separated on agarose gels and purified as described by Vogelstein and Gillespie (1979). The small fragment was irradiated with 1.5 kJ/m2 of 254 nm of light to produce an average of 3 pyrimidine dimers/molecule. The irradiated DNA was mixed with the nonirradiated ( B~~H I -E C O R I )~~~ fragment at a concentration of 4 pg/ml each and covalently linked using T4 DNA ligase. After overnight incubation at 14 "C, ligase was inactivated by heating at 65 "C for 10 min. The DNA was then digested with PuuII (which cuts a t base pair 2067) and separated on a 1 % agarose gel; a full-length fragment (4363 base pairs) was purified from this gel and used for electron microscopic studies. Photolyase-DNA complexes were fixed with 1% formaldehyde, 0.6% glutaraldehyde, mounted onto thin carbon films, and rotary shadowed with tungsten (Griffith and Christiansen, 1978); micrographs were taken using a Phillips EM400 TLG electron microscope.
Preparation of Substrate for Filter Binding and Flush Photolysis Studies-Radiolabeled pBR322 DNA was obtained from E. coli AB2463 (recA13)/pBR322 as previously described (G. Sancar et al., 1985). The specific activity was 1.3 X lo5 cpm/pg. The DNA was diluted to a concentration of 20 pg/ml and irradiated with 254 nm of light. The average number of UV-induced lethal lesions/molecule was determined by transformation selecting for tetracycline resistance (Sancar and Rupert, 1978a;A. Sancar et al., 1984aA. Sancar et al., , 1984b; G. Sancar et al., 1985); incubation of UV-irradiated DNA with a molar excess of photolyase for 1 h under photoreactivating light (supplied by a Sylvania F15TF8BLB bulb at a fluence rate of approximately 10 J/m2), followed by transformation was used to quantitate the average number of pyrimidine dimers/molecule. When performed as described, the transformation assay is a sensitive and accurate method for quantitating dimer repair. Typically transformation with 0.1 pg of plasmid DNA containing an average of 6 pyrimidine dimers/ molecule results in 500 (&lo%) transformants/ml of culture, while cultures transformed with plasmid molecules containing fewer dimers contain proportionately larger numbers of transformants according to the relationship S/S, = e-A, where X is the number of pyrimidine dimers/molecule. For experiments to determine k2, competing DNA was prepared by irradiating plasmid pBM150 (tetn amp'; Johnston and Davis, 1984) at a concentration of 20 pg/ml with sufficient 254-nm light to produce 100 dimers/plasmid molecule. Plasmid DNA was concentrated by ethanol precipitation and the concentration was determined by absorbance at 260 nm.
Nitrocellulose Filter Binding Assay-The details of this assay have been published previously (Riggs et al., 1968;Madden et al., 1973;Seawell et al., 1980; G. Sancar et al., 1985). Briefly, the reaction mixture (110 pl) consisted of 50 mM Tris, pH 7.5, 1 mM EDTA, 10 mM 8-mercaptoethanol, 100 pg/ml bovine serum albumin, DNA at the concentrations indicated in the figure legends, and 100 mM NaCl unless stated otherwise. After addition of photolyase to the desired concentration, the reaction mixture was incubated at 22 "C for 1 h before filtering through Schleicher and Schull BA85 filters (24-mm diameter). All experiments were performed under illumination from G.E. gold fluorescent bulbs.
Flush Photolysis-This method is a modification of Porter's (1960) classic flash photolysis technique adapted by  to the study of the action mechanism of photolyase. DNA-photolyase complexes present in a reaction mixture are subjected to a high intensity flash of photoreactivating light of approximately I-ms duration, resulting in repair of pyrimidine dimers bound by the enzyme at the instant of the flash. Since the turnover rate of the enzyme is considerably longer than the duration of the flash, the number of dimers repaired (as determined by biological or chemical methods) is a measure of the number of ES complexes present at the moment of the flash. (See also Jorns et al., 1986, in this series.) Originally Rupert and colleagues used Yashika PRO 50 photographic flash units to provide the flash. We have found that any of several commercially available photographic flash units are equally applicable provided that the sample is screened with Pyrex glass and/or the plastic lid of a Petri dish to eliminate short wavelength radiation. In the studies described in this paper we used Vivitar 2500 photographic flash units for flash photoreactivation.
Substrate for the flash photolysis assay was UV-irradiated pBR322 DNA, labeled and irradiated as described above. Following flash photoreactivation, photolyase was removed by phenol extraction and the DNA was removed from photolyase assay buffer and concentrated by ethanol precipitation. An aliquot of the DNA was then used in the transformation assay to determine the number of UV-induced lesions and pyrimidine dimers remaining per molecule. Losses during extraction and precipitation were monitored by scintillation counting.

Visualization of Photolyase-DNA Complexes
We have previously reported (G. Sancar et al., 1985) that photolyase binds specifically to UV-irradiated DNA and has little affinity for nonirradiated DNA. This is also apparent from analysis of DNA/photolyase mixtures by electron microscopy. Plasmid pBR322 DNA containing dimers only in a FIG. 1. Binding of photolyase to linearized pBR322 DNA containing pyrimidine dimers in a slightly off-center location. The pBR322 substrate, prepared as described under "Materials and Methods," was incubated with photolyase at a concentration of 1.8 nM DNA and 9.3 nM enzyme under standard binding conditions for 1 h; photolyase-DNA complexes were then fixed and processed for transmission electron microscopy as previously described (Griffith and Christiansen, 1978). Bar indicates 100 nm. Upper panel shows one photolyase-DNA complex while the DNA molecule in the lower panel appears to bear two complexes; it should be noted that less than 10% of the molecules analyzed contained two ES complexes.  Sancar et al., 1985). The data points shown in Fig. 2 were then plotted on these curves to obtain the K A values. Symbols are the same as described in the legend to Fig. 2. 374-base pair region in a slightly off-center location relative to the ends of the linearized molecule was mixed with photolyase in the dark and examined by electron microscopy. Fig.   1 shows two representatives of the majority of ES complexes we observed. The enzyme binds with measurable frequency only to the irradiated region and binding to this region is not seen if the mixture is exposed to photoreactivating light prior to fixing the complexes with glutaraldehyde (not shown). From the micrographs shown in Fig. 1 it is appafent that photolyase is a spherical molecule of about 50 A diameter; furthermore, under the same conditions and magnification, DNA polymerase I (M, = 103,000) and UvrA protein (M, = 104,000) appear to be about twice the size of photolyase (Mr = 53,994; G. Sancar et al., 1984) consistent with photo-otolyase to DNA lyase binding to DNA as a monomer (data not shown). This observation is in agreement with our previous results obtained from gel filtration studies which showed that photolyase elutes as a M, = 49,000 protein (A. Sancar et aL, 1984b). Because of difficulty in visualizing proteins as small as photolyase by electron microscopy we have not obtained a sufficient number of photographs to quantitate the specific and nonspecific association constants by this method; however, we may draw the qualitative conclusion from this and other micrographs that photolyase does not bind to nonirradiated DNA to any significant extent.

Quantitation of the Association Constant KA of Blue
Photolyase Nitrocellulose Filter Binding-The nitrocellulose filter binding technique first used to study Lac repressor-operator interactions has become a standard technique for quantitating enzyme-substrate complex formation for DNA-binding proteins (Riggs et al., 1968(Riggs et al., ,1970. We have utilized this technique to study the equilibrium binding of photolyase to UV-irradiated DNA. As can be seen in Fig. 2, incubation of radiolabeled pBR322 DNA containing an average of 5.5 pyrimidine dimers/ molecule with increasing concentrations of photolyase results in increasing retention of the DNA by the nitrocellulose filters up to the point at which saturation is apparently obtained. However, the relationship between the fraction of total DNA retained on the filter and the total number of ES complexes present in solution is nonlinear in this case because 1) the efficiency of retention of molecules containing a single ES complex is less than 1.0 for photolyase-pyrimidine dimer complexes (G. Sancar et al., 1985) and 2) the substrate (pyrimidine dimers) is not distributed uniformly among the DNA molecules, but rather follows a Poisson distribution. As is described in the Appendix,' beginning with the relationship derived by Woodbury and von Hippel (1983) for a substrate with multiple independent binding sites, we have derived the following equations which express the quantitative relationship between the fraction of total DNA retained by the filter (3) KA was determined from the filter binding data shown in Fig.   2 as follows: the known values for P,, D, X, and t (=0.34; quantitated by titrating enzyme with dimer-containing DNA, G . Sancar et al., 1985) were used to generate theoretical curves of predicted retention ( R ) versus log KA for each protein concentration (Fig. 3). The experimentally observed retention at each protein concentration was then plotted on the appro-'The "Appendix" are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are available from the Journal of Biological Chemistry, 9650 Rockville Pike, Bethesda, MD 20814. Request Document No. 86M-1951, cite the authors, and include a check or money order for $1.60 per set of photocopies. Full size photocopies are also included in the microfilm edition of the Journal that is available from Waverly Press. priate curve and the K A values thus obtained (Fig. 3) when averaged gave a value for K A = 6.0 f 2.1 X lo7 M-' for the binding of photolyase to pyrimidine dimers in DNA. The good agreement between the observed retention of dimer-containing DNA and that predicted for this K A is shown by the fact that the solid curoe drawn through the data points in Fig. 2 is the theoretical curve calculated from Equations 1-3 using K A = 6.0 X lo7 M-', while the dashed lines indicate the retention expected for K A values of 1 standard deviation from the mean.
Flash Photolysis-As an increasing proportion of the substrate-binding sites on each DNA molecule are bound by photolyase, the nitrocellulose filter binding assay becomes less sensitive to increases in ES complexes/molecule; in addition, as can be seen in Fig. 3 and Equations 1-3, as saturation is approached small differences in the apparent fraction of DNA molecules retained make large differences in the apparent K A . Therefore, we chose to use a second method to confirm our estimate of K A , namely the flash photolysis technique originally used by Rupert (1968, 1970) to study binding of yeast photolyase in vitro. Various concentrations of photolyase were mixed with UV-irradiated DNA and after 1 h of incubation in the dark, the mixture was exposed to a single intense flash of photoreactivating light sufficient to repair all dimers complexed with active enzyme; determination by transformation assay of the number of dimers/DNA molecule before and after the flash provides a direct measure of the number of ES complexes present at the time of the flash. As can be seen in Fig. 4, a typical enzyme saturation plot is thus obtained. Analysis of the data by an Eadie-Scatchard plot (Segel, 1975) is shown in Fig. 4; photolyase binds to dimers with an apparent stoichiometry of 0.75 enzyme molecules/dimer and with an association constant K A = 4.7 X lo7 M-' in good agreement with the value for K A obtained by the nitrocellulose filter binding technique. It should be noted that at great (200-2000-fold) enzyme excess, only about 75% of dimers are repaired by a single flash suggesting that some enzyme molecules are able to bind pyrimidine dimers but are deficient in photolysis (see below and G.

in this series).
Binding of Yellow Photolyase to Pyrimidine Dimers E. coli photolyase contains a FAD neutral blue radical (Jorns et al., 1984) and has a blue-purple color when freshly prepared. Upon storage at -20 "C the radical is oxidized to FAD,, and the enzyme turns yellow. Since at present we are unable to oxidize the blue radical in an experimentally controlled manner, we do not have fully oxidized enzyme. However, we have obtained an enzyme preparation in which 70% of the blue neutral radical has been converted to FAD,, (Jorns et al., 1987) after 6 months of storage at -20 "C. When we conducted the filter binding experiment with this enzyme preparation the results shown in Fig. 5 were obtained. The yellow enzyme preparation was still capable of binding specifically to UV-irradiated DNA, displaying an apparent association constant K A = 4.8 X lo7 M-'. Clearly the binding seen is in excess of that expected if only the remaining enzyme containing neutral blue radical were capable of binding DNA (Fig. 5 , broken line).

Dependence of KA on [NaCl]
When proteins bind to DNA by ionic interactions between positively charged groups on the protein and negative charges on the DNA phosphate backbone, counter ions are displaced from the phosphates as the result of binding. Therefore, if an important fraction of the total binding energy is this ionic interaction, KA shows a strong dependence on the cation concentration in the reaction mixture. Record et al. (1976) have derived a quantitative relationship between K A , ion pairs formed upon binding, and the counter ion concentration:   Lohman et al., 1980) and m' is the number of ion pairs formed between the DNA and protein. Table I shows K A for photolyase binding as a function of [NaCl] as determined by the filter binding assay. From the slope of the Record-Lohman plot (Fig. 6) we obtain m' = 1.6 f 0.3. These results indicate that photolyase makes ionic contact with only 1-2 phosphates in the phosphodiester backbone when bound to the dimer.
Extrapolation of Koba to 1~ NaCl concentration can be used to estimate the nonelectrostatic contribution (AG,,) to the total free energy of photolyase-DNA binding (Record et al., 1976). From the data shown in Table I and Fig. 6 we obtain AGIM = -8.9 kcal/mol a t 20 "C. Since: (where AGlYs = 0.2 kcal/mol, the standard free energy of a lysine-phosphate ionic interaction at 1 M NaCI), we obtain AG,, = -9.2 kcal/mol. Under our standard binding conditions AG = -10.5 kcal/mol and thus only about 10% of the total free energy of binding of photolyase to a dimer-containing region is due to ionic interactions.

Determination of the Rate Constants kl and k2
From the reaction scheme for photolyase-substrate binding k1 k, E -k S c l E S , the change in concentration of ES complexes in the absence of photolysis is given by the differential equation To determine the dissociation rate constant k2 we utilized the "cold DNA chase" method of Riggs et al. (1970) and Harm and Rupert (1970). Photolyase was incubated with UV-irradiated pBR322 DNA and incubated for 1 h to allow binding to reach equilibrium; competing pBM150 DNA containing a 20-fold molar excess of pyrimidine dimers compared to the pBR322 DNA was then added and at various later times samples were removed and exposed to flash photolysis thus repairing dimers in ES complexes at the time of the flash. The binding constant at each NaCl concentration was determined as described in the legends to Figs. 2 and 3 except that the reaction mixture contained 10 mM Tris, pH 7.5, and NaCl at the appropriate concentrations. The error bars indicate the standard deviation of the mean for each K A determination. The line was obtained by linear least squares regression analysis which indicated a correlation coefficient of 0.87. m' is the slope and equals the number of protein-phosphodiester bond contacts, as described in the text. UV-irradiated pBR322 DNA containing an average of 4.2 dimers/molecule was incubated at a concentration of 3.3 nM with photolyase at 70 nM under standard reaction conditions in the dark for 1 h. Competing UV-irradiated DNA (pBM150 containing approximately 100 dimers/ genome) was added at a concentration of 2.66 nM at zero time; at various later times samples were taken and the number of ES complexes remaining were determined by flash photolysis and transformation assay. kz was determined by extrapolation of the initial slope (15-and 30-9 time points) to a value of 0.37.  Harm (1970aHarm ( , 1970b obtained similar results for the dissociation step in vivo. In addition, we have noted a marked variation in the rate of repair of specific pyrimidine dimers which appears to be dependent upon both dimer type and the surrounding sequences. ' A strategy similar to that for determining k, is frequently used to quantitate kl, i.e. reaction conditions are chosen such that the term k, [ E S ]

= k,[E][S]k [ E S ]
and by adding photolyase to a reaction mixture containing UVirradiated pBR322 DNA and subjecting the mixture to flash photolysis at various later times. As can be seen, even using the minimum concentrations of enzyme and substrate which allow us to reliably quantitate [ES] by transformation, the reaction is very rapid. Only for the 5-s point shown in Fig. 8 is kl [E], [S], less than 20% of k2[ESIt (using kp = 3 x lo-' s-'); using Equation 8 and this first point we obtain a value for k, = 4.2 X lo6 M" s-'. If, on the other hand, we use Equation 9 to determine kl, the values obtained are dependent upon the value chosen for k2 as well as the proportion of the fast and slowly dissociating complexes present at each time point.
It is of interest to compare the equilibrium association constant K A predicted from the relationship K A = kl/kp with the value obtained by the equilibrium binding experiments described above. Using values for kl of 1.4 x lo6 to 4.2 X lo6 M-' s" and for kz of 3 x lo-' s" we calculate K A = 4.7 X lo7 M" to 1.4 X lo8 M -~, in good agreement with the values of 4.7 X IO7 to 6 X lo7 obtained by equilibrium binding.

DISCUSSION
Photolyase is a spherical molecule of about 50 A diameter which binds specifically to irradiated DNA. Since the enzyme does not act on photoproducts other than pyrimidine dimers (Brash et al., 1985) we conclude that this specificity is the result of photolyase-pyrimidine dimer binding. Perhaps the most interesting and surprising result of this study is the relatively minor contribution of electrostatic interactions to the binding of photolyase to substrate. The dependence of K A on monovalent cation concentration indicates that the protein makes electrostatic contacts with only 1-2 phosphates on the DNA backbone. This conclusion is consistent with other observations: (i) the enzyme can act efficiently on substrates as small as oligo(dT), ; (ii) photolyase does not inhibit the incision by ABC excision nuclease of the fourth phosphodiester bond 3' to a pyrimidine dimer (A. Sancar et al., 1984a). The fact that only about 10% of the total free energy of photolyase-DNA binding is electrostatic at moderate salt concentrations has important implications for the specificity of binding and for the mechanism by which the enzyme "searches" for its target.
At optimum conditions of ionic strength and pH, all formerly characterized sequence-specific DNA binding proteins (e.g. lac repressor and EcoRI restriction endonuclease) have K A values 2 10" M" (see Berg and von Hippel, 1985;Terry et al., 1983), consistent with the selectivity of these proteins; in contrast we find K A = 6 X lo7 M" for E. coli photolyase, yet in vivo data indicate that the enzyme is highly specific. Harm (1970b) reported that in E. coli cells containing approximately 20 pyrimidine dimers and 20 photolyase molecules, about 10 enzyme-substrate complexes were present at equilibrium. If we assume that all of the remaining enzyme molecules bound nonspecifically and that the photolyase binding site spans 4 nucleotides , then the minimum specificity of photolyase binding (defined by the ratio of the specific to nonspecific association constants) is 8 X 106/4 X 10 = 2 X lo5 (assuming that each of the 8 X lo6 bases in the E. coli chromosome constitutes the beginning of a potential nonspecific binding site). This value is similar to those reported for sequence-specific binding proteins; thus it is the ratio of specific to nonspecific binding constants rather than the absolute values of these constants that is important in determining specificity (von Hippel and Berg, 1986). The major difference between E. coli photolyase and the lac re-pressor or EcoRI is that the latter two proteins have large nonspecific association constants mediated largely or entirely by electrostatic interactions with DNA phosphates, and therefore must also have high KA values to retain specificity de Haseth et al., 1977;Revzin and von Hippel, 1977;Berg and von Hippel, 1985;Terry et al., 1983;Jen-Jacobson et al., 1983). The S187 mutant of the EcoRI protein makes 2 phosphate contacts at pH 7.4 compared to 8 phosphate contacts for the wild type enzyme; the result is that while K A for the enzyme decreases from 10'' M-' (wild type) to 4.6 X los M" (S187), the enzyme retains its specificity for binding to the EcoRI recognition sequence (Jen-Jacobson et al., 1983), indicating that the specificity of EcoRI binding, like photolyase, derives largely from nonionic interactions.
A number of sequence-specific DNA-binding proteins utilize nonspecific electrostatic interactions to increase the kinetic efficiency of binding; the high association rate constants of proteins such as lac repressor generally exceeds the Smolouchowski limit for diffusion controlled reactions (Riggs et al., 1970;Berg et al., 1981;Berg and von Hippel, 1985) indicating that the proteins find their targets by one-dimensional or facilitated diffusion. Our results indicate that the contribution of such mechanisms of reduced dimensionality are much less important for binding of photolyase than for other DNA-binding proteins, as the bimolecular association constant kl = 4.5 X lo6 "' s-' is well within the limit of $ diffusion controlled reaction for a spherical protein of 50 A diameter. Furthermore, the minor contribution of electrostatic interactions to photolyase binding argues against the use of unidimensional diffusion as a protein must have a significant level of nonspecific, electrostatic binding to slide or track along DNA (Berg and von Hippel, 1985). The above arguments do not rule out relatively small contributions of nonspecific interactions to both the kinetics of photolyase binding and the intracellular equilibrium; indeed it is clear that some nonspecific binding of the enzyme to DNA does occur as minicells of E. coli which lack DNA are devoid of photolyase (Setlow and Cohen cited in Cohen et al., 1968) while minicells containing R factor DNA have a small but detectable amount of photoreactivating enzyme (Paterson and Roozen, 1972). However, the magnitude of these effects relative to lac repressor or EcoRI must be small.
It is of interest to compare the values which we obtain in vitro for the kinetic and equilibrium constants for photolyase binding with values obtained in vivo by Harm (1970a). Because the photolysis reaction can easily be separated from the binding reaction in vivo as well as i n vitro this enzyme offers a unique opportunity to measure reaction parameters under both conditions and to compare them. Using the flash photolysis technique i n viuo, Harm (1970a) found kl = 1.1 X lo6 " 1 s-l , kz = 1.3 X to 1.9 X s-' from which K A can be calculated to be 8.5 X lo7 to 5.6 X 10' M-'. All of these values agree reasonably with the i n vitro values reported here: kl = 1.4-4.2 X lo6 M-' s-', k2 = 2-3 X lo-' s-', and K A = 6.0 X IO7 M -~. Thus we conclude that the kinetic and equilibrium constants we obtain for photolyase binding in vitro reflect the in vivo state.