Thermodynamic parameters governing interaction of EcoRI endonuclease with specific and nonspecific DNA sequences.

Equilibrium binding of EcoRI endonuclease to DNA has been analyzed by nitrocellulose filter and preferential DNA cleavage methods. Association constants for pBR322 and a 34-base pair molecule containing the EcoRI site of this plasmid in a central position were determined to be 1.9 X 10(11) M-1 and 1.0 X 10(11) M-1 at 37 degrees C, respectively, with the stoichiometry of binding being 0.8 +/- 0.1 mol of endonuclease dimer per mol of DNA. In contrast, the affinity of the enzyme for a pBR322 derivative from which the EcoRI site has been deleted is 3.2 X 10(9) M-1 as judged by competitive binding experiments. If it is assumed that each base pair can define the beginning of a nonspecific binding site, this value corresponds to an affinity for nonspecific sites of 7.4 X 10(5) M-1. Furthermore, the affinity of the endonuclease for the EcoRI-methylated sequence is at least three orders of magnitude less than that for the unmodified recognition site. The dependence on temperature and ionic strength of the equilibrium constant governing specific interactions has also been examined. The temperature dependence of the reaction indicates that entropy increase accounts for 70% of the free energy of specific binding at 37 degrees C. Affinity of the endonuclease for the EcoRI site is highly dependent on NaCl concentration. Analysis of this dependence according to the theory of Record and colleagues (Record, T. M., Jr., Lohman, T. M., and deHaseth, P. (1976) J. Mol. Biol. 107, 145-158) has implicated 8 ion pairs in the stability of specific complexes, a value identical with the number of phosphate contacts determined by ethylation interference analysis (Lu, A. L., Jack, W. E., and Modrich, P. (1981) J. Biol. Chem. 256, 13200-13206). Extrapolation to 1 M NaCl suggests that nonelectrostatic interactions account for 40% of the free energy change associated with specific complex formation.

The most widely studied of the Type I1 restriction endonucleases has been that of EcoRI specificity (1). The recognition sequence 5 cleavage being symmetrically disposed about the dyad axis (2,3). This endonuclease has been purified to homogeneity ( 4 5 ) and the primary sequence has been determined (6,7). The protein exists in solution as an equilibrium mixture of dimers and tetramers (5, 8). It has been argued that the dimer represents the functional unit of the enzyme and that specific endonuclease-DNA complexes may possess elements of 2-fold symmetry (1,9). This idea has been supported by results of alkylation protection and interference experiments (10).

"G-A-A-T-T-C 3°C-T-T-A-
Although EcoRI sites are typically embedded within a large background of nonspecific DNA, the endonuclease is capable of efficient location of its recognition sequence. Several models have been proposed to explain facile site location by sequence-specific proteins (11)(12)(13). Such models invoke nonspecific binding of protein to DNA and then translocation in a dimension-limited diffusion process until the recognition site is located. As a prerequisite, such proteins must exhibit an affinity for nonspecific sequences sufficient to permit a significant level of interaction but not so high as to represent a kinetic barrier to translocation.
Several studies have examined interaction of EcoRI endonuclease with specific and nonspecific DNA sequences (9, [14][15][16][17][18]. However, interpretation of these experiments has been complicated by substantial variation in reported binding constants. The majority of this work has relied on the nitrocellulose assay for trapping of protein-DNA complexes, but questions of membrane retention efficiency of specific and nonspecific complexes have not been addressed. Furthermore, these studies have been performed under a variety of conditions of temperature and ionic strength rendering ccmparison difficult, and no attempts have been made to correct apparent specific binding constants for enzyme bound to nonspecific DNA sequences (1,13). For these reasons we have performed a systematic analysis of interaction between EcoRI endonuclease and specific and nonspecific sequences of plasmid pBR322 using the intact plasmid, a derivative from which the EcoRI site has been deleted, and a 34-base pair DNA which contains the EcoRI site of the molecule. In these studies we have employed two assays for specific binding, and further, have examined the dependence of specific interactions on temperature and ionic strength.

Materials
EcoRI endonuclease and methylase were homogeneous preparations isolated as previously described (5,19). Plasmid pBR322 DNA (20) and pBR322ARI DNA (17) were isolated according to Hardies and Wells (21). T4 polynucleotide kinase was a gift from Dr. R. Kolodner (Harvard University). Escherichia coli alkaline phosphatase and PvuII endonuclease were purchased from commercial sources.

Methods
DNA Substrates"PlasmidpBR322 (4362 bp') was hydrolyzed with PuuII endonuclease to yield a molecule with the EcoRI endonuclease site near the middle (2067 bp from the nearest end). A 34-bp fragment containing the EcoRI site of this plasmid in a central position (16 bp from the nearest end) was prepared as previously described (10). DNA fragments were 5'-end labeled with T4 polynucleotide kinase (23) and purified by reverse phase system 5 chromatography (24).
Plasmid pBR322ARI is a derivative of pBR322 that has been cleaved with EcoRI endonuclease, DNA termini repaired with T4 DNA polymerase, and then circularized using T4 DNA ligase (17). Linear pBR322ARI was obtained by PuuII endonuclease hydrolysis.
Plasmid pBR322 and 34-bp fragment were methylated with EcoRI methylase in vitro as previously described (19). The extents of methylation were 1.8 and 1.9 methyl groups incorporated per DNA molecule, respectively.
Assays of DNA Binding-Specific endonuclease. DNA complexes formed in the absence of M e are efficiently retained on nitrocellulose filters (5,14,17,16). Unless specified otherwise, reaction mixtures contained 0.1 M Tris. HCI, pH 7.6, 1 mM EDTA, 0.05 mg/ml of bovine serum albumin, [5'-32P]DNA, and EcoRI endonuclease as indicated. Incubation was at 37 "C until well after equilibrium was attained, usually 1-2 h. Duplicate reaction samples were filtered through nitrocellulose membranes (presoaked in 0.1 M Tris.HC1, pH 7.6, 1 mM EDTA) and the DNA retained was quantitated by liquid scintillation counting. Analysis of the temperature dependence of DNA binding was performed in a similar manner. However, the pH was adjusted to 7.6 at the incubation temperature in order to compensate for the temperature dependence of the Tris buffer.
An alternate method was also employed for assay of specific EcoRI endonuclease-DNA complexes (25). In this procedure enzyme and [5'-32P]DNA are allowed to reach equilibrium with respect to specific complex formation at 37 "C in the absence of M e as described above. M e and a large excess of the same DNA in nonradioactive form is then added, incubation continued for 20 s, and the reaction quenched. Under these conditions the fraction of labeled DNA cleaved is directly proportional to the amount of enzyme.DNA complex formed at equilibrium with the yield being 0.8-1.0 mol of double strand breaks per mol of endonuclease dimer at saturating DNA concentrations?
Analysis of ionic strength dependence of specific binding of EcoRI endonuclease to pBR322 was determined in reactions containing 0.02 M Tris. HCl (pH 7.61, 1 mM EDTA, 0.05 mg/ml of bovine serum albumin, 0.05-0.25 M NaCl, and DNA and endonuclease as indicated. Reactions were incubated at 37 "C until equilibrium was attained (empirically determined) before analysis by nitrocellulose membrane filtration.
DNA concentrations were determined by two methods. The concentration of stock solutions of pBR322 and derivatives of pBR322 were determined spectroscopically using an extinction coefficient of 20 liters/g cm at 260 nm and a molecular weight of 2.88 X lo6 g/mol (26). The concentration of 34-bp [5'-32P]DNA (above) was determined by isotope dilution using a preparation of the unlabeled molecule whose concentration had been accurately determined spectroscopically. DNA concentrations are expressed in terms of molecules unless otherwise stated. EcoRI endonuclease concentrations were determined using an extinction coefficient of 0.830 liter/g cm at 278 nm and are expressed in terms of dimer equivalents (5).

Specijic
Binding-EcoRI endonuclease. DNA complexes are efficiently retained on nitrocellulose membranes in the absence of Mg2+ (5,14,17,18). Incubation of a fixed concentration of pBR322 or a 3Cbp DNA containing the EcoRI site of this plasmid with increasing endonuclease results in hyperbolic binding isotherms that appear to represent a true equilibrium process (Fig. 1). EcoRI endonuclease titration of pBR322 occurred with an apparent stoichiometry of 0.78 & 0.08 mol of endonuclease (dimer) per mol of pBR322 and an association constant of 1.9 t 0.6 X "1 at 37 OC. The   DNA yielded specific association constant KS = 1.7 X 10" M" and a stoichiometry of 0.93 mol of pBR322 bound per mol of endonuclease (Fig. 2). A number of experiments in which endonuclease was incubated with near saturating (>95%) DNA concentrations yielded a stoichiometry of 0.8 k 0.1 mol of endonuclease dimer per mol of EcoRI site. This value indicates that at least 80% of the enzyme was active in specific binding and that efficiency of retention of specific complexes on nitrocellulose membranes is near unity. In the experiments reported here we have assumed that this stoichiometry represents the fraction of active enzyme, rather than a filter retention effect, and have incorporated this correction by guest on March 24, 2020 http://www.jbc.org/ Downloaded from into analysis of binding curves. During the course of analysis of specific interaction between the endonuclease and DNA, all binding curves were found to be hyperbolic indicating the absence of cooperative effects in specific binding.
Since EcoRI endonuclease also interacts with nonspecific DNA sequences (Refs. 14, 15, and below) the observed association constant is expected to be a function of the DNA concentration (13), where KS is the intrinsic specific association constant, KN is the nonspecific association constant, and D is the free DNA concentration (in base pair equivalents), which in the case of experiments presented here can be approximated by the total DNA concentration. Using the nonspecific association constants described below, it can be shown that under the exper- Hence, under the dilute conditions of these experiments the difference between observed and intrinsic association constants is well within experimental uncertainty. Furthermore, no variation in the experimentally determined association constant was observed over a 100-fold range of endonuclease and 6-fold range of DNA concentration. Given the similar specific association constants for pBR322 and the 34-bp molecule derived from this plasmid, these experiments also illustrate that the affinity of the endonuclease for the EcoRI site of the plasmid is much greater than that for the remainder of the 4362-bp molecule.
One possible reservation concerning the results cited above is the potential for perturbation of DNA .protein complexes during collection on nitrocellulose membranes. Therefore, formation of specific complexes was monitored by an alternate method in which pre-existing DNA. endonuclease complexes were scored as those susceptible to preferential cleavage upon addition of Mg2' (25).' Using this assay, we have demonstrated that binding isotherms are independent of method of measurement. Fig. 3 presents results of an experiment in which specific binding was scored by both methods. As can be seen, the two procedures yield comparable results. In the case of the preferential cleavage assay, the binding curve shown yields an association constant for binding to pBR322 of 1.8 X 10" a value identical with that cited above.

FIG.
3. Comparison of nitrocellulose filter and DNA cleavage assays for quantitation of specific endonuclease-DNA complexes. 13'P]pBR322 (12.7 PM) and EcoRI endonuclease (dimer) as indicated were incubated at 37 "C for 1 h. Samples (0.5 ml) were removed and analyzed by the nitrocellulose filter binding assay (0). The remainder was supplemented with unlabeled pBR322 and M&12 to yield final concentrations of 5.4 nM and 9 mM, respectively. After a 20-9 incubation at 37 "C, the reactions were quenched and subjected to agarose electrophoresis (25). Radioactive bands were excised and quantitated by liquid scintillation counting (A).

DNA
Experiments were performed at 37 "C as described under "Methods." The equilibrium constants are expressed in terms of DNA molecules.
Nonspecific Binding of EcoRI Endonuclease t o DNA-Nonspecific DNA. EcoRI endonuclease complexes are poorly retained on nitrocellulose membranes (17), with retention efficiencies for such complexes varying between 10 and 30% depending on the batch of nitrocellulose membranes used. 3 We have, therefore, used competition methods to analyze nonspecific binding. In order to assess the affinity of endonuclease for pBR322 sequences external to the EcoRI site, competition studies were performed using pBR322AR1, a plasmid in which the EcoRI site has been destroyed by a 4-bp insertion (17). As shown in Fig. 4, specific binding of EcoRI endonuclease to pBR322 DNA was inhibited by pBR322AR1, and double reciprocal plots indicated this inhibition to be competitive in nature (not shown). In addition, covalently closed circular and linear pBR322ARI were equally effective in this respect indicating that DNA topology does not play a significant role in equilibria governing nonspecific interactions. The competition curves shown in Fig. 4 indicate that interaction of the endonuclease with sequences external to the EcoRI site of pBR322 are governed by an association constant of 3.2 X lo9 M-', a value 65 times less than that for the EcoRI site. If it is assumed that each nucleotide pair can initiate a potential nonspecific binding site, this value corresponds to a nonspecific association constant on a per site basis of 7.4 X lo5 M" under standard assay conditions (see under "Methods"). Therefore, this analysis, like that described above, indicates that affinity of the endonuclease for the EcoRI site of pBR322 is much greater than that for the remainder of the plasmid DNA molecule. Effects of the EcoRI Modification on Binding of the Endonuclease and Its Recognition Site-Modification of the EcoRI site involves methylation of the 6-NH2 of adenine residues adjacent to the recognition site dyad, rendering the sequence resistant to cleavage by the restriction endonuclease (3). The interaction between endonuclease and EcoRI-methylated DNAs was, therefore, examined. Since we have previously shown that complexes formed between the endonuclease and modified DNA are retained with low efficiency on nitrocellulose membranes (17) This value is comparable to that observed for binding of the endonuclease to pBR322AR1, indicating that the endonuclease has a markedly reduced affinity for the modified EcoRI sequence.
The association constant determined in this experiment may reflect interactions between endonuclease and the methylated EcoRI site, between the enzyme and sequences external to the recognition site, or a combination of both. In order to more carefully assess the affinity of the endonuclease for the methylated EcoRI site, competition experiments were performed in which binding of 34-bp [5'-32P]DNA was examined in the presence of the unlabeled methylated 34-bp molecule as competitor. In these experiments (not shown) competition by the methylated short DNA was governed by an association constant of 1.9 x IO8 M-', although inhibition deviated from linear competitive above 7 nM, indicative of more than one DNA molecule binding to endonuclease (28) or the presence of trace amounts of substrate in the inhibitor (29). The possibility that the preparation of methylated DNA contained trace amounts of unmethylated molecules cannot be excluded. Nevertheless, this association constant demonstrates that affinity of the endonuclease for the unmodified EcoRI is at least three orders of magnitude greater than that for the methylated sequence. If the methylated form of the 34-bp DNA contains multiple binding sites for the enzyme, then the difference in affinity will be even larger.
Enthalpy and Entropy Changes Associated with Specific Complex Formation-The binding of EcoRI endonuclease to pBR322 was investigated as a function of temperature. From the thermodynamic relationships at equilibrium it can be shown that d(ln K)/d(l/T) = -AH"/R (30). Construction of a van't Hoff plot of In K uersus 1/T enables calculation of the enthalpy of binding ( A H " ) from the slope (Fig. 5 ) . The values obtained are: A H o = AH (37 "C) = -4.7 kcal/mol, AG (37 "C) = -15.9 kcal/mol, and A S = 36.2 cal/mol deg (TAS = -11.2 kcal/mol at 37 "C). It is apparent that the primary driving force for specific EcoRI endonuclease binding is a large increase in entropy.
Dependence of Specific Complex Formation on NaCl Concentration-If a protein binds electrostatically to DNA, counterions are displaced from the phosphate backbone increasing the entropy of the system. Consequently, the position of equilibrium with respect to polyelectrolyte interactions can be altered by variation of counterion concentration. Record FIG. 5. Dependence of the equilibrium association constant on temperature. The association constant for EcoRI endonuclease binding to [32P]pBR322 was measured as a function of temperature using the nitrocellulose filter binding assay. Conditions were as described in the legend to Fig. 1 except that pH was adjusted to pH 7.6 at each temperature. The line is a linear least squares regression fit, and error bars represent one standard deviation of quadruplicate determinations.

TABLE I1
NaCl dependence of the specific association constant for EcoRI endonuclease binding to pBR322 Reactions were performed at 37 "C in 0.02 M Tris.HC1 (pH 7.6), 1 mM EDTA, 0.05 mg/ml of bovine serum albumin, and the indicated NaCl concentration. Data shown is presented graphically in Fig. 6.
The uncertainties are the standard deviation in the slope of the Eadie-Scatchard plot used to determine the association constant. As shown in Table 11, the equilibrium constant for EcoRI

EcoRI Endonuclease
DNA Binding endonuclease binding to pBR322 is quite dependent on NaCl concentration. Furthermore, the salt dependence of Ks obeys the relationship predicted by polyelectrolyte theory over the range of 0.088-0.17 M NaCl (Fig. 6). Assuming $ = 0.88 for double helical DNA (33), the slope of this plot yields a value for m' of 8.1 f 0.6 ion pairs formed during specific binding by EcoRI endonuclease. Subsequent extension of binding to 0.24 M NaCl does not alter the value of m'. Although the reason for deviation from theory of the dependence of Ks on NaCl concentration at low counterion concentrations is not understood, a similar effect has been observed in the case of the lac repressor-operator system (34).

DISCUSSION
In order to assess interactions between EcoRI endonuclease and specific and nonspecific DNA sequences we have examined binding of the enzyme to plasmid pBR322, a derivative in which the EcoRI site has been destroyed, and a 34-bp DNA which contains the EcoRI site of the plasmid but lacks >99% of the nonspecific sequences. Results obtained with these three DNAs are internally consistent. Thus, the affinity of the enzyme for pBR322 is about 60 times that for the plasmid from which the recognition site has been deleted, a finding consistent with the observation that specific affinity for a 34bp molecule containing the EcoRI site of pBR322 is essentially identical with that for the intact plasmid. These results are also in agreement with previous findings, based on equilibrium competition, that the intrinsic affinity of the endonuclease for the EcoRI site of pBR322 is independent of DNA chain length (25). Furthermore, if it is assumed that each base pair can represent the beginning of a nonspecific binding site, it is evident that the affinity of the endonuclease for the EcoRI site of pBR322 is on the order of lo5 times greater than that for nonspecific sites within this DNA under the binding conditions routinely employed (see under method^").^ We have previously concluded that the functional form of EcoRI endonuclease is the dimer based on simple steady state kinetic behavior of the enzyme under dilute conditions where the dimer is the thermodynamically stable species (Kd for the tetramer to dimer transition is M) (1,5,8). The binding isotherms reported here are also consistent with this idea. In particular, specific complex formation upon titration of DNA with the endonuclease under dilute conditions is governed by a simple hyperbolic function. This indicates that association of dimers to tetramers is not necessary for specific binding. The stoichiometry of binding (0.8 f 0.1 mol of endonuclease dimer per mol of EcoRI sites) is also in accord with this view, and furthermore, indicates that efficiency of retention of specific complexes on nitrocellulose membranes is near unity at moderate salt concentrations.
The finding that the affinity of the endonuclease for EcoRImodified pBR322 is comparable to that for a molecule in which the EcoRI sequence had been destroyed indicates that specific methylation markedly reduces the affinity of the endonuclease for its recognition site. In order to more carefully evaluate the affinity of the enzyme for the methylated sequence, the EcoRI-modified 34-bp DNA was employed. The affinity of the enzyme for this molecule was found to be three orders of magnitude less than that for an unmodified recognition site. This provides a minimum estimate of relative 4 Since nonspecific binding constants were obtained by equilibrium competition methods in which competitor DNA was in large excess over endonuclease, values cited are expected to be biased toward nonspecific sequences which have a higher affinity for the enzyme than the typical nonspecific site. Hence, affinity for the EcoRI site relative to the average nonspecific sequence may be greater than that cited in the text. affinity of the endonuclease for methylated and unmethylated states of the EcoRI site. If the modified form of the short DNA contains multiple binding sites for the protein, then the difference could be substantially larger. However, this limit indicates that it is possible that residual affinity of the enzyme for the methylated sequence may be as much as 10 to 100 times that for a nonspecific site.
Temperature dependence of the specific association constant indicates that binding of EcoRI endonuclease to its recognition site is primarily entropy driven. At 37 "C, increased entropy accounts for 70% of the binding free energy, although this should be viewed as an approximate value in light of the limited temperature range studied and the relatively small variation in Ks with temperature.
The equilibrium constant for interaction of EcoRI endonuclease with its recognition site is highly dependent on salt concentration. We have examined this dependence according to the theory of Record and colleagues (31,32) for interaction of polyelectrolytes. Although interpretations based on this approach can be complicated by dependence of polyelectrolyte hydration or amino acid residue pK values on salt concentration, as well as anion effects (31, 32), it is noteworthy that this analysis indicates 8.1 f 0.6 ion pairs in specific complexes. This is precisely that predicted based on the number of phosphate contacts (4 phosphates/strand) in specific EcoRI endonuclease. DNA complexes as deduced by ethylation interference experiments (10).
Extrapolation of Ks for specific complex formation to a NaCl concentration of 1 M can be employed to estimate the nonelectrostatic contribution to the free energy of binding (31). Utilizing the data of Table I1 and Fig. 6 we obtain a value of AG, = -5.31 kcal/mol at 37 "C. The nonelectrostatic contribution to the free energy of specific site binding is given by where N is the number of ion pairs and AG,, is the standard free energy of formation of a single lysine-phosphate ionic interaction. In 1 M NaC1, AQ,. = 0.2 kcal/mol (31, 35).
Therefore, AG, = -5.3 -(8)(0.2) = -6.9 kcal/mol. This indicates that under our standard binding conditions (see under "Methods," Ks = 1.9 X 10") about 40% (-6.9 kcal/ mol/-15.9 kcal/mol = 0.43) of the total specific binding free energy is due to nonelectrostatic interactions. The results cited above were obtained using a nitrocellulose membrane assay. We have also monitored specific complex formation by a preferential DNA cleavage assay, a more tedious method but one which is reliable for quantitation of specific complexes at ionic strengths up to at least 0.1 M. ' The fact that both methods yielded identical specific association constants indicates that the thermodynamic parameters governing specific interaction reported here are not assay dependent. While the nitrocellulose membrane assay can be used to accurately quantitate specific endonuclease. DNA complexes, it is inadequate for precise quantitation of nonspecific complexes between the enzyme and DNA due to the low efficiency of retention of such complexes on membranes (17). Depending on the lot of nitrocellulose membranes, we find that efficiency of retention of such complexes ranges from 10-30% (not shown). As shown here, this problem can be circumvented by monitoring nonspecific interactions by competition methods.
Previous studies using the nitrocellulose membrane assay have suggested that in the absence of M$+, EcoRI endonuclease binds preferentially to its recognition site as opposed to nonspecific DNA sequences (5,14,17,18). Interpretation of such experiments has been compromised by their failure to establish retention efficiencies for enzyme. DNA complexes, particularly in the case of those involving nonspecific interactions. Indeed, Goppelt et al. (16) have concluded that there is no preferential interaction between the endonuclease and its recognition sequence in the absence of divalent cations. This conclusion was based on measurement of complex formation between EcoRI endonuclease and several oligonucleotides by circular dichroism spectroscopy. In particular, the affinities of the enzyme for d(G-G-A-A-T-T-C-C), d(G-A-A-T-T-C), d(T-T-A-C-A-T), and d(T-A-A-A-T-G) were found to be comparable (all being in the range of 3 to 8 X lo6 M-'), and salt dependence of binding indicated involvement of two ion pairs in binding to d(G-A-A-T-T-C) (16). These conclusions are clearly at odds with the results presented here. We attribute these differences to several effects. First, measurement of binding affinity by circular dichroism required micromolar concentrations of endonuclease and oligonucleotide, values well within the range of where nonspecific interactions become significant (Ref. 15 and this paper). Secondly, it has been shown that in specific complexes the endonuclease interacts with about 10 base pairs (lo), with several of the phosphate contacts being external to the canonical recognition sequence. The 5"OH terminated octamer employed in the circular dichroism experiments lacked several of these potential phosphate contacts, and the hexamers used were single-stranded under conditions of spectroscopic analysis (16). It, therefore, seems evident that the oligonucleotides employed in the spectroscopic experiments did not occupy the entirety of the endonuclease DNA binding site, thus rendering interpretation in terms of binding specificity difficult. In contrast, the results presented here are consistent with the number of phosphate contacts as deduced by alkylation methods.