Probing the indirect readout of the restriction enzyme EcoRV. Mutational analysis of contacts to the DNA backbone.

According to the crystal structure of the specific EcoRV•DNA complex, not only the functional groups of the nucleobases but also the phosphate groups of the DNA backbone are contacted by the enzyme. To examine the contribution of backbone contacts to substrate recognition and catalysis by EcoRV, we exchanged 12 amino acids residues located close to phosphate groups by site-directed mutagenesis. We purified the resulting EcoRV mutants and characterized them with respect to their DNA binding and cleavage activity. According to our steady state kinetic analysis, there are strong interactions between three basic amino acid residues (Lys-119, Arg-140, and Arg-226) and the phosphate backbone that support specific binding presumably by inducing and maintaining the kinked conformation of the DNA observed in the specific EcoRV•DNA complex. These contacts are important in both the ground state and the transition state. Other, uncharged residues (Thr-93 and Ser-112), which could be involved in hydrogen bonds to the phosphate groups, are needed primarily to stabilize the transition state. An especially important amino acid residue is Thr-37, which seems to couple recognition to catalysis by indirect readout.

The specific interaction of proteins with DNA is of fundamental importance in biology. Much effort, therefore, has been undertaken to understand the molecular basis of the underlying recognition process. It has been suggested that specificity is due to specific hydrogen bonds between the protein and the edges of the bases of the DNA (direct readout) (1). Such contacts are indeed observed in nearly every structure of specific protein-DNA complexes determined so far. In the first co-crystal structure of the trp repressor and its operator DNA, however, no direct contacts to the bases were seen that could account for the observed specificity of DNA binding; it was argued, therefore, that in this case, specific recognition of DNA is primarily due to contacts to the phosphate groups of the DNA backbone (indirect readout) (2). This means that during interaction with DNA, the protein recognizes a specific sequencedependent conformation of the phosphodiester backbone in addition to functional groups of the bases. This concept was initially met with some reservation, but it has proven to be very fruitful, although it was recently demonstrated that the specific interaction between the trp repressor and its operator is not only due to phosphate contacts but also due to water-mediated base contacts as well as DNA-induced tetramerization of the protein on the DNA (3). From comparisons of cocrystal structures of specific and nonspecific complexes of DNA binding proteins with DNA, it became evident that specific recognition of DNA by proteins is mediated by direct and indirect readout also in other systems (e.g. glucocorticoid receptor (4); EcoRV (5)); in general, in the specific complexes, much more interactions are observed between the protein and the bases but also between the protein and the phosphate backbone of the DNA than in the nonspecific complexes.
It is difficult to assess the relative contributions of both mechanisms of specific recognition because in virtually all complexes between specific DNA binding proteins and their cognate DNA, direct and indirect readout is the result of an interconnected network of hydrogen bonds to functional groups of the bases exposed in the major and minor grooves and hydrogen bonds or ionic interactions to the phosphate groups of the DNA. The crystal structure analyses of restriction enzyme⅐DNA complexes provide beautiful examples that illustrate the complexity of the network characteristic for the protein-DNA interfaces, as demonstrated for EcoRI (6,7), EcoRV (5,8), PvuII (9), and BamHI (10). In the case of EcoRV, there are various reasons to assume that the indirect readout contributes to the recognition process. In contrast with the cocrystal structures of the other restriction enzymes, where direct or water-mediated hydrogen bonds are observed to every base pair of the respective recognition sequences, in the specific EcoRV⅐DNA co-crystal structure, no hydrogen bonds to the bases of the two inner AT base pairs of the recognition sequence GAT2ATC (the arrow denotes the position of phosphodiester bond cleavage) are seen. Furthermore, among the four cocrystal structures obtained so far for restriction enzymes, the EcoRV⅐DNA structure is unique in having a highly distorted DNA with a sharp central kink of approximately 50° (5,11,12), which renders the inner two base pairs inacessible to the protein. Recognition of these two base pairs, therefore, cannot be direct. On the other hand, there are a multitude of amino acid residues contacting the phosphate groups of the DNA both within and outside of the recognition sequence (5) (see Fig. 1), which might compensate for the lack of base contacts.
One strategy applied to examine the indirect readout of restriction enzymes has been to use modified oligonucleotides as substrates, e.g. phosphorothioates (EcoRI (13,14), EcoRV 1 ) or S-methylphosphorothioates (TaqI (16)). Here we present a complementary approach focussing on the contribution of the protein to the indirect readout by a systematic mutational analysis of amino acid residues, which in the crystal structure are located sufficiently close to the phosphodiester backbone to * This work was supported by Deutsche Forschungsgemeinschaft Grant Pi 122/12-1 and by the Fonds der Chemischen Industrie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. Tel.: 641-702-5824; Fax: 641-702-5821. be considered candidates for indirect readout. We will present evidence that the catalytic efficiency of the restriction enzyme EcoRV depends on both modes of specific recognition.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis and Purification of EcoRV Variants-Sitedirected mutagenesis of the EcoRV gene was carried out following the method described by Ito et al. (17). This technique is based on the incorporation of a desired mutation with a PCR 2 primer by two successive PCR reactions. The presence of the mutations as well as the absence of unwanted mutations elsewhere in the gene were confirmed by sequencing both strands of the entire gene of each mutant. All mutant proteins had an affinity tag of six His residues on their N terminus to facilitate purification (18). Protein expression and purification was carried out essentially as described by Wenz et al. (18).
Cleavage Assays with Plasmid DNA-For plasmid cleavage assays, 21 nM pATRV (a derivative of pAT153 containing an additional EcoRV cleavage site) linearized with AseI (U. S. Biochemical Corp.) was incubated at 37°C in cleavage buffer (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 10 mM MgCl 2 ) or star buffer (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM MnCl 2 ) with varying amounts of wild-type or mutant EcoRV (diluted in 10 mM Tris-HCl, pH 7.5). After defined time intervals, aliquots were withdrawn, and the reaction was stopped by adding one-fifth volume agarose gel loading buffer containing 250 mM EDTA. The reaction products were separated by electrophoresis on 1.2% agarose gels, which were subsequently stained with ethidium bromide and photographed using a video system (Intas, Göttingen, Germany). EcoRV activity is given in relative units defined by the amount of enzyme needed to completely digest the DNA under standard conditions.
Oligodeoxynucleotide Substrate-The oligodeoxynucleotide employed in the cleavage assay was synthesized on solid support with a Milligen Cyclone DNA synthesizer and purified by denaturing polyacrylamide gel electrophoresis. It was labeled at the 5Ј-end with [␥-32 P]ATP (Amersham Corp.) and T4 polynucleotide kinase (MBI Fermentas) according to the protocol of the suppliers. Throughout the text, oligodeoxynucleotides that carry a 5Ј-phosphate group are abbreviated by d(p . . . ) to indicate this fact.
Steady State Cleavage Experiments-The rate of the EcoRV catalyzed DNA cleavage reaction was investigated as a function of oligodeoxynucleotide concentration to determine K m and k cat values. The reaction mixtures contained 100 pM to 10 M radioactively labeled oligodeoxynucleotide in cleavage buffer supplemented with 100 g/ml bovine serum albumin in a total volume of 5-40 l; the concentration of EcoRV dimer was usually at least 10 times lower than the substrate concentration, only at the lowest substrate concentration it was 5 times lower. The reaction was started by the addition of 20 pM to 0.1 M EcoRV (diluted in 10 mM Tris-HCl, pH 7.5, 100 g/ml bovine serum albumin). The reaction mixture was incubated at 25°C. After defined time intervals, 0.5-4-l aliquots were withdrawn from the reaction mixture, spotted onto a DEAE cellulose plate (Macherey-Nagel) and subjected to homochromatography (19). The detection and quantitation of the separated substrates and products was carried out using an Instant Imager system (Canberra Packard).
For K m and k cat determination, the initial velocities were calculated from the linear part of the progress curves obtained with six or more different substrate concentrations. The K m and k cat values were obtained by a best fit to the cleavage data using the computer program ENZFITTER (Biosoft, Cambridge, UK). Substrate concentrations were chosen to cover at least a range from 0.2 to 5 K m .
Single Turnover Cleavage Experiments-Single turnover experiments were performed by incubating identical concentrations (0.5 M) of radioactively labeled oligodeoxynucleotide and EcoRV in cleavage buffer at 25°C. The processing of the samples was identical to the procedure described for the steady state experiments. The progress curves were fitted to a function with a single exponential using the computer program ENZFITTER (Biosoft, Cambridge, UK).
Gel Electrophoretic Mobility Shift Experiments-A 32 P-labeled 382-bp PCR product with one centrally located EcoRV cleavage site was generated as described by Jeltsch et al. (20). The equilibrium binding experiments were performed with Ca 2ϩ as cofactor that inhibits cleavage but supports specific binding (12). Radioactively labeled 382-mer (20 -1000 pM) was incubated with 0 -0.1 M wild-type or mutant EcoRV in binding buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM CaCl 2 , 10 mM 2-mercaptoethanol, 2 mM spermine, 100 g/ml bovine serum albumin) for 30 min at room temperature. To 10 l of this mixture, 3 l of gel loading buffer (50% (v/v) glycerol, 0.25% (w/v) xylene xyanol, 0.15% (w/v) chromotrope FB in binding buffer) were added. Electrophoresis was carried out in 10 ϫ 10-cm 6% polyacrylamide gels at room temperature in 0.5 ϫ TT (100 mM Tris, 29 mM taurine) supplemented with 10 mM CaCl 2 . The detection and quantitation of the radioactive bands was carried out using an Instant Imager. The dissociation constants K d were determined from the resulting titration curves, which were obtained under conditions where the concentration of the DNA was at least 10 times lower than the estimated K d value. Under these conditions, the K d values are given by the enzyme concentration where 50% of the DNA is bound.

Selection, Generation, and Purification of EcoRV Mutants-
According to the available EcoRV⅐DNA co-crystal structures (Brookhaven databank entries 2RVE, 4RVE, 1RVA, 1RVB, 1RVC) (5,8), there are 12 amino acid side chains with hydrogen bond donors in close vicinity to the nonbridging oxygen atoms of the phosphate groups of the DNA substrate (heteroatom distance Ͻ 0.4 nm, Fig. 1, Table I). Residues of the catalytic center, the R-and the Q-loops, which were the target of previous mutational studies (21,22) were not included in the work described here. Amino acid residues, which due to their proximity to the DNA backbone could be responsible for the indirect readout of EcoRV, were exchanged by PCR mutagenesis to Ala (in the case of Arg, Lys, Ser, Thr) or to Phe (Tyr). Wild-type EcoRV and the 12 mutants were produced as N-terminally His 6 -tagged proteins and purified to near homogeneity (better than 95%) by chromatography on nickel chelate columns. The specific activity of the wild-type EcoRV preparation determined with -cleavage assays was, within the limits of error, the same as measured previously in our experiments (18) or in other groups (23,24).
Cleavage of a Plasmid Substrate by EcoRV Mutants-The mutant EcoRV proteins were characterized first by cleavage assays with pATRV. This plasmid harbors two EcoRV sites with very different flanking sequences ( . . . GCGG-GATATCGTCC . . . versus . . . GAAAGATATCTTTT . . . ). To allow for the identification of the products of cleavage of pATRV, the plasmid was linearized with AseI prior to cleavage with EcoRV. All mutants display specific EcoRV activity in normal cleavage buffer (containing MgCl 2 ) and star buffer (containing MnCl 2 ); the mutants T111A, S112A, and, to a smaller extent, S41A, show an enhanced star activity compared with the wild-type enzyme, i.e. cleavage at sites that differ in one base pair from the canonical site ( Fig. 2) (25). At the substrate concentration used in this assay (21 nM), most of the mutants display moderately altered relative activities in cleavage buffer, ranging from a slightly enhanced to by a factor of 50 decreased activity in comparison with the wild-type enzyme (Table II). Only one EcoRV variant, T37A, shows a 1000-fold diminished relative activity. This is comparable in magnitude with the activity of various mutants with substitutions in the recognition loop, for example N185A, N188A, and S183A, which have residual activities of 1/5000, 1/600 -1/5000, and 1/1000, respectively (26). These data indicate that Thr-37 is as important for the catalytic efficiency of EcoRV as amino acid residues of the recognition loop, which are engaged in specific hydrogen bonds to the bases of the recognition sequence.
It is known that the DNA cleavage rate of EcoRV is influenced by sequences flanking the recognition sequence (27). The use of a plasmid substrate with two EcoRV sites embedded in very different sequence surroundings offers the possibility to screen the mutants for an enhanced sensitivity toward flanking sequences. Whereas in cleavage buffer the wild-type enzyme cleaves both sites with similar rates, three EcoRV variants, namely R226A, R140A, and T93A, show a preference for one site (Table II, Fig. 3), all three mutants prefer the EcoRV site with AT rich flanking sequences.
Cleavage of Oligodeoxynucleotide Substrates by EcoRV Mutants-As the crystal structure analysis of the EcoRV ϫ d(AAA-GATATCTT) 2 complex has produced evidence for an interaction between EcoRV and position Ϫ5 of the substrate DNA (cf. Fig. 1a), a 20-mer oligonucleotide substrate was chosen for a detailed kinetic analysis of the mutant proteins under steadystate conditions. This substrate can be expected to mimic a high molecular weight DNA as it is longer than the entire binding pocket of EcoRV, which covers 14 -16 bp (5,20). By variation of the substrate concentration over 5 orders of magnitude, it was possible to determine the kinetic parameters K m and k cat for all mutants except the T37A mutant (Table III). Fig. 4 shows as an example the Michaelis-Menten diagrams for the wild-type EcoRV and the mutant R226A. The steady-state kinetic parameters determined for the cleavage of the 20 bp substrate by the wild-type enzyme are in the same order of magnitude as corresponding values reported previously for plasmid substrates, e.g. for pAT153, the reported K m , k cat , and k cat /K m values are 0.5 nM, 0.9 min Ϫ1 , and 3.0 ϫ 10 7 M Ϫ1 s Ϫ1 , respectively (25).
Depending on their K m and k cat values, it is possible to divide the EcoRV mutants into three groups (Table III).
Group I consists of mutants with a similar k cat /K m value as the wild-type enzyme: S41A, T94A, Y95F, T111A, R221A, S223A.
Group II consists of mutants with lower k cat /K m values due to an increase in K m : K119A, R140A, R226A. The k cat /K m values of these mutants are reduced by 2-3 orders of magnitude compared with the wild-type enzyme, and this group, therefore, comprises the EcoRV variants with the lowest activities.
Group III consists of mutants with lower k cat /K m values due to a decrease in k cat : T93A and S112A. The k cat /K m values of these mutants are by more than 1 order of magnitude lower than that of the wild-type enzyme.
The only EcoRV variant that could not be classified according to the above scheme is the T37A mutant. Due to its low activity, DNA cleavage by T37A is only measurable at equimolar concentrations of enzyme and substrate, i.e. under single turnover conditions. In good agreement with the results of the plasmid DNA cleavage experiments (Table II), the single turnover cleavage rate constant k for oligodeoxynucleotide DNA cleavage by this mutant is 700 times lower than the k cat of the wild-type enzyme (Table III). As the K d value measured for T37A (see below) is well below the substrate concentration that was employed in the single turnover experiments (0.5 M), the enzyme is likely to be saturated with substrate. Hence, it is reasonable to compare the k value of the mutant with the k cat value of the wild-type enzyme. Thr-37 is the only amino acid residue involved in a phosphate contact that has an impact on both specific binding (K d ) and catalysis (k), although the effect on catalysis is much more pronounced.
DNA Binding Experiments with EcoRV Mutants-In addition to the kinetic analyses, equilibrium binding studies with a 381-bp DNA fragment harboring a single EcoRV site were performed and analyzed by gel electrophoretic mobility shift assays. These experiments were carried out in the presence of 10 mM CaCl 2 (instead of MgCl 2 ), which inhibits cleavage but supports specific DNA binding (12). The equilibrium DNA binding constant measured for binding of the wild-type enzyme to the 381-bp DNA fragment is the same as that determined previously under similar reaction conditions (12). By variation of the protein concentration over 4 orders of magnitude, it was possible to detect a specific band shift for all mutants (Fig. 5). It was not possible, however, to monitor complete binding curves for the mutants K119A and R226A because of their low affinity to DNA. The retardation of the specific complex in comparison with the free DNA was identical for all EcoRV variants, implying that the degree of DNA bending is within Ϯ5°(the estimated accuracy of the determination), the same for the wild-type enzyme and all EcoRV mutants. For the wild type enzyme and most of the mutants, the K d values measured in the presence of Ca 2ϩ (which permits cleavage, but supports binding) are 1 order of magnitude lower than the K m values measured in the presence of Mg 2ϩ . In a double logarithmic plot of K d versus K m , most mutants are well represented by the regression line (Fig. 6). Exceptions are the mutants T93A and S112A; their K d values are more than 1 order of magnitude higher than the corresponding K m values. As in the two sets of experiments (binding versus cleavage) only the divalent cation used was different; this finding may suggest that in the presence of specific DNA, these two mutants have a lower affinity for Ca 2ϩ than for Mg 2ϩ compared with the wildtype enzyme and the other mutants. DISCUSSION Restriction endonucleases recognize their cleavage sites on double-stranded DNA with remarkable high specificity by forming contacts both to the bases and to the backbone of the DNA (28,29). As the essential role of the base contacts had been established before (21,18), it has been the aim of the work presented here to define the importance of contacts between amino acid residues and phosphate groups for the mechanism of DNA recognition by the restriction endonuclease EcoRV. For this purpose, we have exchanged all amino acid residues that are located in proximity to the phosphodiester backbone of the DNA in the co-crystal structures of EcoRV⅐DNA complexes (5,8). The resulting EcoRV variants were analyzed in terms of DNA binding and cleavage activity. According to our experimental results, the amino acid residues Thr-37, Arg-226, Lys-119, Arg-140, Thr-93, and Ser-112 (arranged in the order of decreasing importance for catalysis) are of major importance for EcoRV activity. They are located close to the phosphates ϩ2, Ϫ5, Ϫ2, ϩ3, ϩ1, and Ϫ1, respectively (Fig. 1a). This finding is in agreement with the results of a complementary analysis in which phosphate contacts of EcoRV were probed with phosphorothioate containing oligonucleotides; it was shown that all phosphates demonstrated by our study to be subject to indirect readout show effects upon modification (the position ϩ5 was not subject of this study). 1 The various mutants can be classified according to whether they are impaired in K m and/or k cat . In the subsequent discussion this classification will be used. Table III)-Six out of 11 EcoRV mutants (S41A, T94A, Y95F, T111A, R221A, S223A) display a similar catalytic efficiency (k cat /K m ) as the wild-type enzyme when analyzed in steady state cleavage experiments. The energy penalty ⌬⌬G app associated with the deletion of the respective functional group in one enzyme monomer is less than 1 kJ/mol or 0.24 kcal/mol (⌬⌬G app ϭ Ϫ0.5 RT ln ((k cat /K m ) mutant /(k cat / K m ) wild type ), Table III), which is well below the value of 5.5 kJ/mol or 1.3 kcal/mol estimated for an average protein phosphate contact (30,31). Consequently, during the enzymatic turnover, the amino acid residues Ser-41, Thr-94, Tyr-95, Thr-111, Arg-221, and Ser-223 most likely interact only weakly with the phosphates of the DNA substrate and are of minor importance for substrate binding and cleavage.

Mutants with Similar Catalytic Efficiency as the Wild-type Enzyme (group I in
Interestingly, some of the mutants (T94A, Y95F, and R221A), which have a similar k cat /K m value as the wild-type enzyme, have a 2-fold higher k cat and K m value (group I* in Table III). Moreover, their K d is also higher than that of the wild-type enzyme. These results may be interpreted in terms of small differential effects on substrate binding and transition state stabilization; it appears that these residues are involved in contacts that stabilize the ground state of the enzyme-substrate complex, not, however, the transition state (Fig. 7).
Mutants with a Reduced DNA Affinity (Group II in Table  III)-During the recognition process, EcoRV distorts the cognate sequence, which leads to the unstacking of the two central base pairs and a bending of approximately 50°as seen in the crystal structure of the specific EcoRV⅐DNA complexes (5,8) and confirmed by gel shift experiments with catalytically inactive mutants (11) or with Ca 2ϩ as a substitute for Mg 2ϩ (12). If the DNA distortion were of similar importance for the transition state as it is for the ground state, one would expect that mutants defective in contacts responsible for bending will have a higher K m or K d and a more or less unaltered k cat , as the energy difference between ground and transition state is the same (Fig. 7). There are three amino acid residues contacting the phosphate backbone which fit into this scheme, namely the basic amino acid residues Arg-226, Lys-119, and Arg-140; the corresponding alanine mutants show a drastic increase in K m and K d but an only slightly changed k cat (Table III). Consequently, the results of both steady state cleavage and equilibrium binding experiments demonstrate that basic residues that are not in direct contact with functional groups of the bases of the recognition sequence but close to the phosphate groups of the DNA backbone, are important for specific binding. The only exception to this rule is the amino acid Arg-221, as the R221A mutant shows only moderately enhanced K m and K d values.
We suggest that the main function of the residues Arg-226, Lys-119, and Arg-140 is to stabilize the distorted conformation of the DNA in the specific EcoRV⅐DNA complex. The results obtained for the K119A and R140A mutants can be rationalized by a comparison of the co-crystal structures of EcoRV with specific and nonspecific DNA. Upon formation of the specific complex, there is a significant movement of the residues Lys-119 and Arg-140 with respect to the DNA that shortens the   distance between the interacting partners (Fig. 8, Table I). Thus, weak ionic interactions with nonspecific DNA become considerably stronger in the complex with the specific substrate leading to a gain in binding energy at the transition from the nonspecific to the specific binding mode. The distances between Lys-119, Arg-140, and the phosphate groups Ϫ2 and ϩ3, respectively, are not the only ones that are altered in the co-crystal structures with nonspecific and specific DNA (Table  1), but we suppose that the changes in positions of Lys-119 and Arg-140 are the most important ones because the interacting partners are charged.
Whereas the decrease in apparent binding energy that accompanies the substitution of Lys-119 (6.7 kJ/mol or 1.6 kcal/ mol) and Arg-140 to alanine (4.0 kJ/mol or 0.95 kcal/mol), respectively, are roughly compatible with the deletion of a single phosphate contact, the energy penalty associated with the exchange at position 226 (8.2 kJ/mol or 2.0 kcal/mol) is significantly higher. The large effect observed with the R226A mutant is particularly noteworthy as Arg-226, which is in a disordered region in the co-crystal structure of the unspecific EcoRV⅐DNA complex, contacts the DNA backbone two residues outside of the recognition sequence. This implies that Arg-226 must have an additional role. It has been suggested that this amino acid residue apart from promoting specific binding is responsible for the stabilization of a loop between ␣ helices D and E encompassing residues 221-228 (8) (cf. Fig. 1b). The importance of the C terminus of EcoRV for catalysis was dem-

for the cleavage of d(pGATCGACGATATCGTCGATC) and dissociation constants (K d ) for the binding of a 382-bp DNA fragment with one EcoRV site for wild-type EcoRV and EcoRV mutants
Apparent binding energies corresponding to one contact formed by the homodimeric enzyme to its palindromic substrate were determined from ⌬⌬G app ϭ Ϫ0.5 RT ln ((k cat /K m ) mutant /(k cat /K m ) wild-type ) (15). For K m and k cat , standard errors calculated by the program ENZFITTER are given; the binding experiments were carried out at least in duplicate, errors of the dissociation constants are estimated to be Ϯ30%. For the classification of the mutants, see Fig. 7.

Mutant
Group  onstrated previously with a mutant whose C-terminal end, starting from position 216, had been deleted. This mutant showed no activity in cleavage assays with -DNA (21). Furthermore, the W219C mutant was shown to be very defective in DNA binding in the presence of Mg 2ϩ (20).
In comparison with the wild-type enzyme, two mutants of group II (R140A, R226A) show an enhanced selectivity toward different flanking sequences (Table II, Fig. 3). Flanking sequences are likely to influence the structure and dynamics of the DNA, both of which are important in the recognition process, especially for the propensity of the recognition sequence to be bent in a unique way. The EcoRV restriction endonuclease distorts the recognition sequence GATATC during the recognition process by introducing a sharp central kink (5). It appears, as if reducing the ability of EcoRV to deform the DNA, for example through mutation of an amino acid residue that stabilizes the kinked conformation, leads to an increased sensitivity of EcoRV to flanking sequence effects. This consideration may explain the preferences of some phosphate contact mutants for EcoRV sites with AT-rich flanking sequences because these sequences may facilitate the distortion of the cognate DNA during the recognition process.
Mutants That Selectively Affect the k cat of the Reaction (Group III in Table III)-Compared with the basic amino acid residues discussed above, Thr-93 and Ser-112 interact with the phosphate backbone in a fundamentally different way. They are responsible for backbone contacts required for the catalytic efficiency of the enzyme. Exchanging these residues leads to EcoRV variants with a reduced k cat value, whereas their ability to bind the substrate is not or only marginally decreased (Table  III), which means that the energy of the ground state complex is unaltered but that the energy of the transition state complex is higher than for the wild-type enzyme (Fig. 7). The way in which the residues Thr-93 and Ser-112 influence the catalytic machinery of the enzyme is obvious, because they contact the phosphate groups ϩ1 and Ϫ1 (Fig. 1a), respectively, which are in close vicinity to the catalytic center and participate in the hydrolysis of the phosphodiester bond (32,33). Thus, hydrogen bonds established by these residues may help directly to stabilize the precise assembly of the catalytic center.
A remarkable property of these mutants is a pronounced metal ion effect on the binding of DNA. In the presence of Ca 2ϩ , T93A and S112A bind to the DNA by more than 1 order of magnitude more weakly than in the presence of Mg 2ϩ , whereas the wild-type enzyme and all other mutants bind to DNA by 1 order of magnitude more strongly in the presence Ca 2ϩ than in FIG. 6. Double logarithmic plot of K d versus K m for wild-type EcoRV and all EcoRV mutants. The data were taken from Table III. The slope of 1 of the regression line suggests that wild-type EcoRV and the various mutants, with exception of S112A and T93A, differ by a constant energy increment (⌬⌬G ϭ RT ln (K m /K d )) of 5.7 kJ or 1.4 kcal in their DNA binding affinity in the presence of Mg 2ϩ and Ca 2ϩ , respectively.
FIG. 7. Schematic free energy profiles to illustrate the classification of EcoRV mutants in groups I*, II, and III. The relative energy levels of the various states of the wild-type enzyme and the mutants encountered during enzymatic turnover (E ϩ S 3 ES 3 ES # 3 E ϩ P) are represented by solid and dashed lines, respectively. The K m value is represented by the energy level of the ground state complex (ES), the k cat value by the energy difference of the ground state and the transition state complex (ES # ). Note that the energy levels are not drawn to scale, neither within nor among the groups.  (Table III, Fig. 6). These results suggest that the residues Ser-112 and Thr-93 participate in the correct assembly of binding sites for divalent metal ions important for specific binding. In agreement with these results, metal ion binding sites of EcoRV have been mapped to be in the catalytic center of the enzyme (5,8) and at a site remote from the cleavage site at phosphate Ϫ2 (20). Whereas Thr-93 is in close proximity to the catalytic center and hence may affect this metal binding site, the main chain carbonyl group of Ser-112 is one possible ligand for the second metal binding site. This conclusion is supported by our finding that the replacement of Ser-112 (and of Thr-111) by alanine leads to an EcoRV variant with a relaxed specificity in the presence of Mn 2ϩ .
The EcoRV Mutant T37A-The exchange of Thr-37 3 Ala has by far the most severe effects on the catalytic efficiency of EcoRV. This mutant is affected mainly in its k cat and retains the ability to bind the recognition sequence specifically and firmly (Table III). The loss of relative activity of this mutant is too severe to be explained by the deletion of a single hydrogen bond. In fact, the activity of this mutant is roughly comparable with that observed with mutants that have amino acid substitutions in the recognition loop of EcoRV (26). The importance of Thr-37 can only be rationalized in terms of a co-operative mechanism, where by the loss of this interaction causes a rearrangement leading to a loss of further protein DNA contacts. Inspection of the structure of the specific EcoRV⅐DNA complex shows that Thr-37 is at a key position of a complex network of interactions (Fig. 9). The side chain oxygen of Thr-37 is not only in hydrogen bonding distance to the phosphate ϩ2 but also to the side chain oxygen of Tyr-138, thereby connecting Thr-37 to the hydrophobic core of the protein.
Thr-37 is located at the N-terminal end of the long ␣ helix B and may be important for the positioning of this secondary structure element, which harbors the residue Glu-45 that is near the catalytic center (5,8). Moreover, there is a connection to the catalytic center of the other subunit that is mediated by a main chain contact between Thr-37 and Gln-69 of the other subunit. Gln-69 is located in the Q-loop, which is in close proximity to Asp-74, a key residue of the catalytic center of the corresponding subunit (5,22). Consequently, the residue Thr-37 is located at a position well suited to connect various functionally important parts of the enzyme. There are possible pathways that may channel information concerning specific recognition, in this case the establishment of a precise hydrogen bond to phosphate ϩ2, to both catalytic centers of the enzyme. We suggest that the important function of Thr-37 is to couple recognition to catalysis, which may explain the drastic decrease in cleavage activity associated with the mutation of this residue. Possibly, Thr-37 fulfills a similar role for the indirect readout, as Asn-188 does for the direct readout, which was proposed to establish a link between the R-loop and the catalytic center (8).
It is presumably no coincidence that Thr-37 as well as Thr-93 and Ser-112, which when substituted by alanine produce the largest k cat effects, also show the largest movements in a comparison of the enzyme-substrate and enzyme-product complexes (Tables I and III).
In summary, the restriction enzyme EcoRV uses both direct and indirect readout mechanisms to achieve its high accuracy of recognition. The primary function of basic amino acid side chains contacting the phosphate backbone of the DNA is to stabilize the kinked DNA conformation in the specific EcoRV⅐DNA complex. Other, uncharged residues help to assemble the catalytic center in the transition state. An especially important residue, Thr-37, is most probably responsible for the coupling of specific recognition and catalysis.
The effects seen with mutants of amino acid residues in close vicinity to the phosphate backbone are in general not as severe as observed with mutants of the recognition loop of EcoRV (21) but together may contribute to a considerable extent to the specificity of the recognition process and to the catalytic efficiency.