INTRODUCTION

The structures of the five type II endonucleases determined by x-ray diffraction, EcoRI (1), BamHI (2, 3), EcoRV (4, 5), PvuII (6, 7) and Cfr10I (8), share substantial elements of similarity. Analysis of the enzymes complexed to DNA, where known, suggests a preliminary classification into two different groups (9). The endonucleases that produce 5-overhanging ends, EcoRI and BamHI, approach the DNA from the major groove, where most of their base-specific contacts are made. The endonucleases that produce blunt-ended DNA products, EcoRV and PvuII, approach the DNA from the minor groove side and establish contacts in the major groove by wrapping around the DNA. Structure-function studies have been limited to a small number of type II enzymes. Years of studies have produced an extensive amount of information about EcoRI (10-18), EcoRV (18-24), and NaeI (25) and, more recently, BamHI (26, 27). These studies have focused on the identification and analysis of residues involved in catalysis, testing alternative mechanisms of catalysis, and understanding how a specific DNA sequence is recognized. Attempts to alter the sequence specificity have proven to be difficult, with successful examples limited to mutants showing relaxed specificity (10, 11, 28) or displaying activity toward DNA substituted with unnatural bases (15, 29). One explanation for this difficulty is that a change in specificity not only requires recognition of a different DNA sequence, but also requires that the linkage between recognition and catalysis be retained. A class of mutations that might illuminate this linkage is the catalytic mutations, where specific DNA binding is retained but catalysis is reduced (26, 30). In particular, mutants in this class that are outside the catalytic center might be deficient in the linkage between recognition and catalysis (28).

PvuII endonuclease (hereafter referred to as PvuII) from Proteus vulgaris is a type II restriction endonuclease that cleaves DNA between the central GC base pair of its recognition sequence (5-CAGCTG-3) in a Mg²⁺-dependent reaction, generating blunt-ended products (31). PvuII approaches DNA from the minor groove side, with the DNA-binding domain reaching around the sides of the DNA and contacting base pairs in the major groove. These contacts are made by residues present in a pair of antiparallel -strands (see Fig. 1) (7). The recognition region includes a His triplet (His⁸³-His⁸⁴-His⁸⁵), an Asn doublet (Asn¹⁴⁰-Asn¹⁴¹), and Ser⁸¹. Another major region of DNA contact is the loop between the first two -helices. Residues in this loop (Gln³³, Asp³⁴, and Asn³⁵), which constitute the floor of the DNA-binding cleft, face the minor groove and make direct contacts with backbone phosphates and the central GC base pair. Two negatively charged residues and one positively charged residue on a short -sheet near the scissile phosphate are predicted to form the catalytic center. PvuII is a good candidate for structure-function studies and protein engineering since it is one of the smallest restriction enzymes (32) and has a limited number of protein-DNA interactions predicted from its crystal structure (7). The work reported here was designed to test the predictions of the crystal structure and to begin to probe the mechanism of catalysis and its linkage to DNA binding, with the ultimate goal of changing the specificity of the endonuclease.

EXPERIMENTAL PROCEDURES

LB medium and LB agar were prepared as described previously (33). When cells contained plasmids coding for ampicillin resistance, the media were supplemented with 100 µg/ml ampicillin. 5-Bromo-4-chloro-3-idolyl -D-galactopyranoside (X-gal)¹ was added to the media at a final concentration of 200 µg/ml. Plates used for testing in vivo DNA binding were supplemented with 125 µg/ml spectinomycin and 15 µM isopropyl--D-thiogalactopyranoside.

Escherichia coli ER1991 (F (argF-lac)U169 supE44 e14 endA1 thi-1 (mcrC-mmr)114::IS10) (34) was the parental strain for the construction of reporter and lysogenic strains. ER2170, a dinD::lacZts derivative of ER1991, was used as a reporter strain to estimate the endonuclease activity of PvuII mutants in vivo (see below). E. coli ER1755 ( F proA⁺B⁺ lacI^q L8/kdgK51 xyl-5 mtl-1 argE3 thi-1 mutD5 thr-1 ara-14 leuB6 lacY1 tsx-3 supE44 galK2 hisG4 rfb-1 mgl-51 rpsL31) was used for random mutagenesis. Expression of PvuII requires the presence of the PvuII methylase to protect cellular DNA from digestion (35). Lysogenic strains PE1991 and PE1755 were made with the phage CAC1.3A, a  phage carrying the PvuII methylase and kanamycin resistance, recovered from a lysogenic derivative of K802 (36) (K802(CAC1.3A), kindly provided by J. Menin and W. Jack). A single plaque from a spontaneous induction was amplified by plate lysate. ER1991 or ER1755 cells were infected at a multiplicity of infection of ~2 and grown to lysis in liquid culture, and the surviving cells were pelleted, resuspended, and plated on LB agar plates containing 15 µg/ml kanamycin. Isolated kanamycin-resistant lysogens were streak-purified twice and cross-streaked with imm21, imm434, and vir to verify lysogeny.

The mutD strain PE1755 was transformed with pBBE3, a plasmid carrying the PvuII gene (37). Individual colonies were picked and grown, and then mutagenized plasmid was isolated, purified, and retransformed into E. coli PE1991 to allow the recovery and expression of mutant proteins. Purified DNA was then transformed into E. coli ER2170 (methylase-deficient dinD::lacZ) and plated on LB agar plates containing ampicillin and X-gal. Only cells containing plasmids expressing mutant endonucleases survive this selection. The DinD::LacZ phenotype of individual colonies was scored (see below); plasmid DNA was prepared, and mutations in PvuII were identified by sequencing the entire gene.

Site-directed mutants were obtained by oligonucleotide mutagenesis using the method of Kunkel et al. (38). To generate the single-stranded DNA needed for mutagenesis, a BglII fragment of the PvuII gene (Asp⁵-Leu¹²⁰) was subcloned into LITMUS 28 (New England Biolabs Inc.). After introducing the desired changes in the LITMUS background, the BglII fragments were excised and reinserted in the original plasmid. The correct orientation of insertion was determined by restriction mapping.

Plasmid DNA was prepared by the alkaline lysis method (33). To identify or confirm mutations, the complete PvuII gene was sequenced using the CircumVent DNA sequencing kit (New England Biolabs Inc.) according to the manufacturer's instructions.

In vivo activity of mutant endonucleases was estimated by their ability to induce the SOS response. The reporter strain ER2170 (methylase-deficient dinD::lacZ) was transformed with plasmids carrying mutant PvuII genes. Cells were plated on LB/ampicillin/X-gal medium, incubated overnight at 30 °C, and then kept at 4 °C until color developed. The intensity of colony color was compared against a negative control and taken as an estimation of DNA damage caused by the mutant endonuclease.

The binding affinity of PvuII mutants for specific sites in vivo was estimated by a modification of the in vivo transcriptional interference assay (39) as described by Dorner and Schildkraut (26). The single copy reporter plasmid (pPVUII388) compatible with the PvuII expression vector was a derivative of pNN388 (39). This plasmid carried one copy of the aadA gene (spectinomycin resistance) with a strong antisense promoter containing a unique PvuII recognition site as an operator. The binding of catalytically inactive PvuII mutants to the operator allows the expression of resistance directly related to the strength of binding, measured as the plating efficiency on spectinomycin plates. Log-phase cultures of ER2170 cells containing pPVUII388 and a mutant PvuII gene were serially diluted in LB medium. A given volume (0.1 ml) of different dilutions was added to 2.5 ml of top agar, and the mixture was poured onto LB/ampicillin agar plates supplemented with spectinomycin (125 µg/ml) and isopropyl--D-thiogalactopyranoside (15 µM). After 24 and 48 h, resistant colonies were counted and compared with a control plating of cells carrying the vector with no PvuII insert.

A series of 65-mer oligonucleotides was used to test the binding of PvuII proteins to the canonical sequence. The sequence of the basic wild-type oligonucleotide containing one PvuII recognition site was 5-GAATCGCTTCATGACTACGCTGATGAGCTTTACCGCAGCTGCCTAGCACGTTTCAGTGATGACAG-3.

Two oligonucleotides of 40 and 41 bases were prepared, corresponding to the first 40 bases and the complement of the last 41 bases of the sequence above. These oligonucleotides shared a complementary region of 16 bases at their 3-ends. Full-length double-stranded oligonucleotides were produced by annealing the two shorter oligonucleotides and extending them with the Klenow fragment of DNA polymerase I in the presence of the four dNTPs (33). Synthesis of radiolabeled DNA was performed in the presence of both [-³³P]dATP and [-³³P]dCTP (NEN Life Science Products). The efficiency of the labeling reaction was monitored by adsorption to DE81 filters (Whatman) (33). Extended DNA was ethanol-precipitated and resuspended in 10 mM Tris (pH 8.0) and 1 mM EDTA (pH 8.0). ³³P-Labeled DNA was separated from the free nucleotides using NucTrap Probe columns (Stratagene) as recommended by the manufacturer.

A PvuII "star" site oligonucleotide was used as a negative control. It contained a single change at the PvuII site (CATCTG) and was resistant to PvuII cleavage (data not shown). Mutant oligonucleotides containing symmetrical changes in the central bases of the PvuII recognition site are indicated by the first base substituted (e.g. G to A identifies the oligonucleotide containing the sequence CAATTG).

The interaction between PvuII and DNA was analyzed by gel electrophoresis mobility shift assays. Binding was assayed using 65-mer oligonucleotides containing either one PvuII site (5-CAGCTG-3) or one modified site as is described for each experiment. Binding reactions (15 µl) contained purified protein (ranging from 10 nM to 0.625 pM final concentration of dimer) and ³³P-labeled DNA (5-10 pM) in 10 mM Tris-Cl (pH 8.0), 50 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 10 mM CaCl₂, 30 µg/ml bovine serum albumin, and 7.5% glycerol. The binding reaction was incubated for 10 min at room temperature. When crude protein extracts were used (PvuII at 50-100 pM), reactions were supplemented with 0.2 µg of poly(dI-dC). Samples were loaded, without the addition of loading buffer or dye, on 15% polyacrylamide gels that were prerun for 2 h at 120 V/cm in 45 mM Tris-Cl, 45 mM boric acid, 1 mM EDTA, and 10 mM CaCl₂ (pH 8.0). Gels were run for 2.5 h, fixed in 7% acetic acid for 10 min, dehydrated with 100% methanol for 30 min, dried, and exposed to x-ray film (40). The resulting autoradiographs were digitized with a Microtek Scanmaker III, and the intensity of the bands was analyzed with NIH Image (Version 1.59).

Enzyme activity was tested at 37 °C in 25 µl of NEBuffer 2 (50 mM NaCl, 10 mM Tris-Cl (pH 7.9), 10 mM MgCl₂, and 1 mM dithiothreitol) containing 0.5 µg of DNA. 1 unit of PvuII will completely digest 1 µg of  DNA in a 50-µl reaction in 60 min as defined by New England Biolabs, Inc.  DNA, LITMUS 28 DNA, and synthetic oligonucleotides utilized to test activity and specificity of the enzymes were from New England Biolabs Inc.

Overnight cultures of cells expressing mutant PvuII proteins were diluted 1:100 into 25 ml of LB/ampicillin medium and incubated at 37 °C until the cultures reached 80-130 Klett units. Protein expression was induced by the addition of isopropyl--D-thiogalactopyranoside to 0.3 mM, and the cultures were incubated at 37 °C for 2 h. Cells were pelleted by centrifugation and stored at 70 °C. Cell pellets were resuspended in 1 ml of 100 mM KCl, 20 mM Tris-Cl (pH 8.0), 2 mM EDTA, and 20 mM -mercaptoethanol and treated with 0.4 mg/ml lysozyme for 10 min at 0 °C. The cell suspension was sonicated and then centrifuged for 30 min at 15,000 × g. Supernatants were diluted with 1 volume of 100% glycerol and stored at 20 °C.

The concentration of PvuII protein in crude extracts was estimated by gel electrophoresis. A calibration curve was made by dilution of purified PvuII protein with 2.5 µl of protein extract from the E. coli parental strain. Samples to be estimated (2.5 µl of the crude extracts) were loaded on a 10-18% polyacrylamide gel side by side with the standards. After staining with Coomassie Brilliant Blue, gels were scanned, and the intensity of the individual protein bands was analyzed with NIH Image. The PvuII concentration of individual samples was extrapolated from the calibration curve.

Wild-type PvuII and D34G mutant proteins were overexpressed and purified as described previously (37). Purified enzymes were stored at 20 °C in 10 mM Tris-Cl, 50 mM NaCl, 0.1 mM EDTA, 1 mM dithiothreitol, and 50% glycerol (pH 7.4).

RESULTS

The specific interaction between PvuII and its DNA recognition sequence (5-CAGCTG-3) (Fig. 1) was analyzed using a gel electrophoresis mobility shift assay. Similar to EcoRV, a stable complex was formed in the presence of 10 mM Ca²⁺, although the activity of PvuII using Ca²⁺ as a cofactor was negligible (Fig. 2 and data not shown) (41). Several attempts to visualize specific PvuII-DNA complexes in the absence of divalent ions were attempted unsuccessfully (data not shown). PvuII showed highly specific binding to its canonical sequence. Incubation of a 65-mer oligonucleotide containing a single PvuII site in the presence of enzyme produced one band with reduced electrophoretic mobility (Fig. 2A). The mobility of the complex was independent of protein concentration in a range from 1.2 to 625 pM (Fig. 3). The shifted band could not be detected using a mutant oligonucleotide differing in only 1 base in the PvuII site (Fig. 2B). Competition experiments showed that the mutant oligonucleotide was 100 times less effective than the specific oligonucleotide in displacing ³³P-labeled DNA bound to the complex (Fig. 2A). A K_d(app) of 110 ± 50 pM (n = 8) was estimated as the enzyme concentration needed to bind half of the labeled DNA (Fig. 3). The affinity of binding was dependent on the reaction conditions. An increase in affinity was observed with increased concentrations of glycerol and bovine serum albumin (data not shown). The binding was performed in 7.5% glycerol and 30 µg/ml bovine serum albumin, conditions in which the enzyme still displayed its maximal activity and specificity for cleavage (data not shown).

The participation of different residues of the PvuII restriction endonuclease in activity and DNA binding was analyzed using a collection of random and site-directed mutants. Cells expressing an active restriction endonuclease need the presence of a modification methylase to protect their DNA. Random mutants with reduced activity (Fig. 1A, left monomer) were selected in a PvuII methylase-deficient dinD::lac fusion strain of E. coli. The expression of the dinD::lac fusion, scored by the intensity of blue color of individual colonies in the presence of X-gal, was presumably a consequence of the induction of the SOS system by the residual endonuclease activity in vivo (30). To decrease the frequency of isolates containing deletions or nonsense mutations and to increase the chances of obtaining a variety of different mutants, colonies with a wide range of blue color were selected (Table I). After random mutants were sequenced and characterized, the collection was enlarged by oligonucleotide-directed mutagenesis, targeting residues implicated in binding and catalysis by the crystal structure (Fig. 1A, right monomer) (7).

Table I. Properties of PvuII mutants PvuII mutant Predicted interactiona Survival in PvuII methylase-deficient cells SOS responseb Cleavage of  DNAc Spectinomycin resistanced Gel shifte % Random mutants   D34G C3 minor groove + ++ 104 70-80 +   S81L G1 + +/ NDf 0     N140D C1, A2 + +++ ND 0     N141K G1 +   ND 3   Site-directed mutants   E55A Binding of Mg2+   NA 1-101 NA +   D58A Binding of Mg2+ + +/ ND 2     E68A Binding of Mg2+ + +/ ND 5     K70A Close to scissile P +   ND NA +   K70R +   ND 0     H83A P4   NA 102 NA +   H84A G3 major groove   NA 102 NA +   H85A Closes off the + + ND 50g +   H85L binding cleft + +/ <104 1.5 + a The predicted function or interaction according to Cheng et al. (7) is indicated. The positions at the PvuII site (5-CAGCTG-3/ 3-GTCGAC-5) were numbered from 1 to 6 starting at the left-hand CG base pair and ending at the right-hand GC base pair. Phosphate groups are given the number of the preceding nucleotide. b Methylase-deficient cells carrying a dinD::lac fusion were transformed to express mutant proteins and plated on LB/X-gal (200 µg/ml) agar plates. The intensity of the blue color, indicated by the  to +++ scale, gave an estimation of the DNA damage cause by the mutant protein. c The specific activity of mutant proteins was compared with the activity displayed by the wild-type enzyme. d PvuII mutant plasmids were transformed into PvuII methylase-deficient cells carrying the plasmid pPvuII388. This plasmid contains a PvuII site positioned as an operator sequence in an antisense promoter opposed to the aadA gene (spectinomycin resistance). Repression of this promoter allows the expression of the resistance gene. The survival is dependent on the strength of the binding of the catalytically deficient enzyme to its recognition site. The data were reported as survival percent of total viable cells. e DNA binding was estimated by gel shift assay. f ND, not detectable; NA, not assayed (mutants that required the expression of PvuII methylase and mutants that had low viability without spectinomycin were not assayed). g Colonies appeared after 48 h rather than overnight.

PvuII mutants were initially characterized by their ability to be transformed into PvuII methylase-deficient cells, induce the SOS response, cleave  DNA in vitro, and bind to PvuII sites in vivo (Table I). Cleavage of  and plasmid DNAs in vitro was measured in crude extracts of cells expressing the mutant enzymes. The interaction with DNA in vitro was also characterized in detail for the wild-type and all mutant proteins using the gel shift assay described above.

Table I summarizes the properties of the mutant endonucleases. The assays of in vivo cleavage activity consisted of the ability of cells bearing the mutants to survive in the absence of the methylase and the degree to which they induced the SOS response as indicated by expression of the dinD::lacZ fusion. In vitro activity was measured by cleavage of  DNA as well as plasmid DNA in some cases (data not shown). The random mutants D34G, S81L, N140D, and N141K, which are all predicted to be involved in DNA recognition, were of course selected for survival in the methylase-deficient strain. Most of the site-directed mutations were also able to transform the methylase-deficient strain, with the exception of E55A, H83A, and H84A. Measurement of in vitro cleavage activity correlated well with the ability to transform; mutants that had 1% of the wild-type activity or more could not be established in the methylase-deficient strain.

The in vivo assay of SOS induction gave a range of levels resulting in colony colors from white to medium blue. However, the level of SOS induction did not correlate well with in vitro activity. This may have been because the mutant endonucleases were associated with some other activity (for example, DNA nicking) that does not parallel their double-strand cleavage activity. A test of this hypothesis could not be accomplished with crude extracts because of a high background of nonspecific nicking activity.

E55A retained approximately the wild-type level of activity. This mutation is near the scissile phosphate, but contrary to earlier predictions (7), these results suggest that it is not involved in catalysis. His⁸³ and His⁸⁴ make a phosphate contact at ApG and a hydrogen bond with the O-6 atom of the third guanine on the complementary strand, respectively. The relatively high level of activity in the H84A mutant contrasts with the lower level found in the other mutants predicted to be involved in nucleoside recognition.

Structural and sequence data suggest that charged residues make up the active site of type II endonucleases (9). The common active-site motif is (E/D)X_n(E/D)ZK, where n varies from 9 to 18 and Z is a hydrophobic residue (42). This domain has been described in EcoRI, EcoRV, and PvuII. Data from this study, when combined with the crystal structure, indicate that Asp⁵⁸, Glu⁶⁸, and Lys⁷⁰ are the active-site residues in PvuII. When alanine is substituted for any of these residues, endonuclease activity is undetectable in crude extracts.

The difference in cleavage activity between H85A and H85L deserves comment. His⁸⁵ forms a hydrogen bond with its counterpoint on the other subunit, closing off the DNA-binding cleft (top of Fig. 1A). H85A destroys this hydrogen bond, and the mutant has an undetectable level of cleavage activity in crude extracts. Replacing the histidine with a leucine rather than an alanine leads to a low but detectable level of activity, suggesting that a hydrophobic contact between the two leucine side chains can partially compensate for the lost hydrogen bond.

DNA binding was tested with a variation of the transcriptional interference assay (26, 39); binding of a catalytically deficient mutant to a PvuII site on a reporter plasmid allowed the expression of spectinomycin resistance (see "Experimental Procedures"). Two mutants gave a significant indication of in vivo binding: D34G and H85A (Table I). In the case of H85A, there was a defect in cell growth on the spectinomycin plates, with no colonies appearing until day 2. It is possible that this assay is under-reporting the number of mutants that bind DNA in vivo. H85A and H85L behave identically in gel shift assays (see below), but only H85A indicates binding in the spectinomycin assay. The other difference between these two mutants is that H85L has a low but detectable level of cleavage activity. The low level of activity may interfere with the assay either by a generalized defect in growth due to DNA damage or by specifically cleaving the spectinomycin reporter plasmid at the PvuII site. In addition, K70A gave no indication of binding in vivo, although it did bind in the gel shift assay (see below); cells bearing K70A had an undetermined defect in growth that was unlikely to be related to activity since this mutant is catalytically inactive.

Fig. 4 shows the results of the gel shift assay with the mutants and a DNA fragment containing the wild-type PvuII site. Alanine substitutions of the two residues that are predicted to bind the divalent metal ion, Asp⁵⁸ and Glu⁶⁸, showed no binding. This is consistent with these mutants being defective in metal binding since, under the assay conditions, Ca²⁺ is required for binding. A substitution of alanine for Lys⁷⁰ bound specifically, approximately as well as the wild-type enzyme, consistent with its predicted role as part of the catalytic site. When arginine was substituted for Lys⁷⁰, binding was abolished, presumably because the larger arginine side chain conflicts sterically with the DNA.

A number of mutants that are not part of the catalytic site in the crystal structure also showed specific binding to the DNA. The His⁸⁵ mutants bound the specific DNA fragment, a little less tightly than the wild-type protein. This histidine residue is not predicted to contact the DNA, but rather to form a hydrogen bond between the two subunits. Substituting either alanine or leucine for this histidine has a profound effect on cleavage activity (see above), but only a small effect on binding. This suggests that this interaction is necessary for forming the correct conformation of the protein-DNA complex to get cleavage. The H83A and H84A mutants also bound reasonably well, indicating that the specific DNA contacts that His⁸³ and His⁸⁴ make are not necessary for binding under these conditions. This is consistent with the relatively high level of residual activity in the alanine substitutions. His⁸⁴ is predicted to be the only major groove contact with the central GC base pair in the PvuII site. These results suggest the possibility that these residues are involved more in coupling cleavage to recognition than in providing binding energy, although investigation of this hypothesis would require purification of the mutant proteins and examination of their binding and cleavage in detail. The D34G mutant bound the specific DNA fragment as tightly as the wild-type protein. This residue contacts the same base pair as His⁸⁴, the central GC, through the minor groove. Thus, neither of the residues that are predicted to contact the central base is individually necessary for specific binding under these conditions. This indicates either that one or the other of these contacts is sufficient for specific binding or that recognition of the specific sequence takes place by indirect readout (28, 43).

The remaining mutants, S81L, T82A, N140D, and N141D, affect residues that are predicted to be involved in contacting the outer 2 base pairs of the PvuII site. None of these mutants showed DNA binding under any of the conditions tested, indicating that these mutations either directly or indirectly interfere with binding.

To see if any of the mutants were altered in the specificity of DNA binding, all of the mutants were tested in the gel shift assay against oligonucleotides in which two positions of the PvuII site were altered symmetrically. None of the mutants bound sites where the outer 2 bases were changed (data not shown). However, when the mutants were tested against sites altered in the central 2 base pairs, the D34G mutant showed a surprising result (Fig. 5). This mutant bound the modified sites CAATTG and CATATG with high affinity and bound the site CACGTG with reduced affinity. Thus, the in vitro DNA binding showed that this mutant had lost the ability to discriminate the base present in the central position.

The D34G mutant protein was purified to near homogeneity, and its cleavage activity was studied in more detail. Overdigestion of  DNA with the D34G mutant protein still gave the wild-type banding pattern (data not shown). In addition, the level of "star activity" associated with D34G was measured in comparison with wild-type PvuII. Contrary to the expected result, the presence of 7.5% dimethyl sulfoxide induced the same level or a higher level of star activity for wild-type PvuII than for the D34G mutant, indicating that, although the mutant is promiscuous for binding, it still cleaves with normal specificity (data not shown).

DISCUSSION

This work describes the isolation of random and site-directed mutants in the PvuII endonuclease and their analysis using a number of in vivo and in vitro tests. In principle, the combination of site-directed and random mutagenesis allows both a direct test of residues that are indicated as central by the PvuII crystal structure and an unbiased probe for additional important residues. In this case, the random mutagenesis was too limited to be comprehensive, and most of the residues that were identified from the random mutagenesis were coincidentally implicated by the crystal structure. Additional mutations were recovered that produced unstable proteins, occurred as double mutants along with an implicated residue, or produced bulky substitutions of residues adjacent to identified motifs. The use of the dinD::lac fusion helped to facilitate isolation of a range of different classes of mutations, but the level of SOS induction has not yet been correlated with an in vitro activity of the mutant proteins.

To characterize the DNA binding of the mutants, in vivo and in vitro assays of DNA binding were employed. A variation of the transcriptional interference assay of Elledge and Davis (39), where repression of an antisense transcript by a DNA-binding protein leads to the expression of spectinomycin resistance, was used to assess in vivo binding. This assay has the advantage of probing for DNA binding under physiological conditions, but comparing the results with the in vitro data indicates that binding to the antisense promoter may not be sufficient to give spectinomycin resistance. In particular, mutants that produced a growth defect or retained low levels of endonuclease activity failed to exhibit spectinomycin resistance, even when in vitro data indicated binding. In these cases, the growth defect may have had a pleiotropic effect on the expression of spectinomycin resistance. On the other hand, this assay showed the binding of two mutants, D34G and H85A, in the presence of in vivo levels of Mg²⁺. This increases the confidence in the correlation of the in vitro binding (see below).

The in vitro DNA binding was assayed by devising a gel shift assay in the presence of Ca²⁺. This assay is nearly identical to the one used for the structurally related enzyme EcoRV (41), except that a 15% acrylamide gel was necessary to trap the protein-DNA complex under the conditions used here. A recent report indicates that EcoRV can bind to DNA in the absence of Ca²⁺ or Mg²⁺ under different conditions (44); attempts to visualize a gel shift for PvuII in the absence of metal ions or in the presence of Mg²⁺ for the catalytically inactive mutants have thus far been unsuccessful.

Similar mutagenesis studies have been done on EcoRI (10-13, 15, 45), EcoRV (18, 20, 23, 24, 28), and BamHI (26, 27, 30). These studies have anticipated or confirmed the predicted roles of residues identified as central from the crystal structures of the respective enzymes. The crystal structures have been used to identify three domains: the subunit interface domain, the DNA recognition domain, and the catalytic domain. The previous work has allowed the identification of the charged residues in the active sites of these nucleases, while the crystal structures show how the corresponding residues line up in similar positions relative to the scissile phosphate in all three enzymes: for EcoRI, Asp⁹¹, Glu¹¹¹, and Lys¹¹³; for BamHI, Asp⁹⁴, Glu¹¹¹, and Glu¹¹³; and for EcoRV, Glu⁷³, Asp⁹⁰, and Lys⁹². The identity of the residues that compose the PvuII active site has been proposed based on sequence and structural homologies with EcoRV and EcoRI (7). The data presented in this work support the assignment of Asp⁵⁸, Glu⁶⁸, and Lys⁷⁰ as the key residues of the PvuII active site. Alanine substitutions by site-directed mutagenesis have shown that these three residues are essential for endonuclease activity. The wild-type characteristics of the E55A mutant also indicate that glutamate 55 is not part of the catalytic center and is not involved in the coordination of metal ions (7). The K70A mutant binds DNA but fails to cleave, similar to mutations at EcoRV Lys⁹² and EcoRI Lys¹¹³. The fact that, unlike their counterparts in EcoRV, EcoRI, and BamHI, the D58A and E68A mutants do not bind DNA in the gel shift assay used probably reflects the fact that the assay requires Ca²⁺ since these residues are thought to bind the metal ion.

In PvuII, substitutions at residues that interact with the external bases of the canonical sequence, Ser⁸¹, Asn¹⁴⁰, or Asn¹⁴¹, abolished the enzyme activity and impaired the formation of the protein-DNA complexes. The functional relevance of these residues is brought out by the fact that these mutants were obtained by random mutagenesis, without relying on knowledge of the enzyme structure. The mutant proteins also could not bind to PvuII sites symmetrically altered at the outside bases (CAGCTG). Additional mutagenesis of these three residues needs to be done to test if substitutions of smaller neutral side chains would be able to support some level of binding or endonuclease activity at canonical or mutant DNA sequences. In particular, the profound effect of the S81L substitution is remarkable. Recognition of the outermost guanine is established by simultaneous contact with asparagine 141 and serine 81 (7). In view of the close proximity of Asn¹⁴¹, the removal of the interaction supported by Ser⁸¹ in the S81L mutant might be expected to have a smaller effect than mutations at Asn¹⁴¹. It is possible that the larger leucine side chain substituted in this mutant created steric constraints that preclude binding. In addition, the fact that the structure has been solved in the absence of Mg²⁺ means that conformational changes not yet observed might accompany binding of the metal cofactor and could alter the interactions with the DNA.

Histidine 84 and aspartate 34 establish contact with the central base pair GC (7). While His⁸⁴ contacts the guanine on the major groove side, Asp³⁴ interacts with the minor groove side. H84A had a much smaller defect in cleavage activity than D34G and retained specific binding of the DNA. Despite predictions based on the crystal structure, methylation of the central cytosine still protected DNA from cleavage by the H84A mutant. On the other hand, Asp³⁴ played a crucial role in the discrimination of the central base pair. Substitution of glycine for aspartate decreased enzyme activity by a factor of 10⁴, but allowed the enzyme to bind DNA regardless of the bases present at the central positions. More elements must be involved in the process of recognizing the central pair since the residual activity of the D34G mutant enzyme only hydrolyzed the canonical PvuII sequence. These data are consistent with several possible mechanisms of recognition of the central bases. His⁸⁴ and Asp³⁴ could be acting synergistically, with either one of them being sufficient for maintaining cleavage specificity. If this is the case, Asp³⁴, approaching the DNA through the minor groove, must be the principal residue because the removal of its side chain produces the most dramatic changes in the enzyme's binding properties. Alternatively, recognition of the central base pair could be an example of indirect readout (28), i.e. the position adopted by the scissile bond in the catalytic center and other nonspecific contacts may be the major discriminating factors, while the base-specific contacts made by His⁸⁴ and Asp³⁴ play only a secondary role. Analysis of the double mutant D34G,H84A might help to explain questions about the recognition of the central base pair.

Of the three contiguous histidine residues present in the DNA recognition region, the most sensitive to substitution was histidine 85. Mutants of His⁸⁵ bound DNA with high affinity, but their catalytic activity was seriously compromised (H85L) or totally abolished (H85A). Substitutions at this position produced another example of separation between binding to the target sequence and catalysis. The different behavior of the leucine and alanine substitutions suggested that a hydrophobic contact between leucine side chains could partially replace the His-His hydrogen bond present in the wild-type protein.

Although there is wide acceptance that DNA recognition and cleavage by restriction endonucleases are strongly linked, most of the mutagenesis studies have outlined how residues in the DNA recognition domain abolish binding and mutations in the catalytic domain abolish cleavage, usually leaving specific binding intact (11, 12, 29, 30, 46). There is a relative paucity of mutations described similar to, for example, T37A of EcoRV (28) that are located outside the catalytic domain and bind DNA specifically but are defective in cleavage. In one case where catalytic mutations of BamHI were selected directly using the transcriptional interference method, all the mutations found were in catalytic domain residues (26). This rarity may be due to how the mutations were made and analyzed rather than a reflection of their abundance in nature. At any rate, the analysis in this study yielded a relatively large number of mutations that are located at residues outside the catalytic domain and bind DNA specifically but have a substantial cleavage defect. These mutants have the potential to illuminate how residues outside the catalytic center can be involved in positioning the reactive side chains and the scissile phosphate for catalysis by changing the conformation of the protein-DNA complex.