Amino Acid Residues in the GIY-YIG Endonuclease II of Phage T4 Affecting Sequence Recognition and Binding as Well as Catalysis

Abstract

Phage T4 endonuclease II (EndoII), a GIY-YIG endonuclease lacking a carboxy-terminal DNA-binding domain, was subjected to site-directed mutagenesis to investigate roles of individual amino acids in substrate recognition, binding, and catalysis. The structure of EndoII was modeled on that of UvrC. We found catalytic roles for residues in the putative catalytic surface (G49, R57, E118, and N130) similar to those described for I-TevI and UvrC; in addition, these residues were found to be important for substrate recognition and binding. The conserved glycine (G49) and arginine (R57) were essential for normal sequence recognition. Our results are in agreement with a role for these residues in forming the DNA-binding surface and exposing the substrate scissile bond at the active site. The conserved asparagine (N130) and an adjacent proline (P127) likely contribute to positioning the catalytic domain correctly. Enzymes in the EndoII subfamily of GIY-YIG endonucleases share a strongly conserved middle region (MR, residues 72 to 93, likely helical and possibly substituting for heterologous helices in I-TevI and UvrC) and a less strongly conserved N-terminal region (residues 12 to 24). Most of the conserved residues in these two regions appeared to contribute to binding strength without affecting the mode of substrate binding at the catalytic surface. EndoII K76, part of a conserved NUMOD3 DNA-binding motif of homing endonucleases found to overlap the MR, affected both sequence recognition and catalysis, suggesting a more direct involvement in positioning the substrate. Our data thus suggest roles for the MR and residues conserved in GIY-YIG enzymes in recognizing and binding the substrate.

Endonuclease II (EndoII) of coliphage T4, encoded by gene denA, catalyzes the initial step in host DNA degradation. It causes irreversible host shutoff and also initiates a nucleotide scavenge pathway that provides precursors for phage DNA synthesis (8). T4 phage DNA is protected from EndoII by the substitution of 5-hydroxymethyl deoxycytosine for dC during T4 DNA synthesis (9, 43). EndoII shares the sequence elements defining the GIY-YIG family of proteins (13, 20) that includes the repair endonuclease UvrC and many homing endonucleases, such as the intron-encoded T4 endonuclease I-TevI (20) and I-BmoI (10), as well as some type II restriction endonucleases (2, 19). Like the homing endonucleases in the GIY-YIG family, EndoII cleaves a long, asymmetric, and ambiguous DNA sequence (4-6, 21). However, in contrast to what has been found for I-TevI (24), EndoII cleaves the two strands independently of each other (7), and only a CG dinucleotide 2 bp away from the scissile bond is strongly conserved (5). Double-strand cleavage by EndoII is the result of concerted single-strand nicks (7), but many positions are only nicked, not cleaved.

In UvrC (44), as well as in the GIY-YIG homing endonucleases (10, 12), the conserved motifs and catalytic activity reside in the amino-terminal domain, but the primary binding energy is conferred by a carboxy-terminal domain, the equivalent of which is absent in EndoII. The GIY-YIG restriction endonucleases (Eco29kI, NgoMIII, and MraI) (2) lack an additional domain, instead having insertions and terminal extensions in the catalytic domain (13), but show low sequence similarity to EndoII. Binding and cleavage by Eco29kI (19) and its isoschizomer Cfr421 (17) have been described recently, but otherwise these restriction endonucleases have not been extensively characterized (11, 42, 45, 46). Structural analysis of the GIY-YIG-domains of UvrC (40) and I-TevI (41) show significant similarities in their suggested catalytic surface despite low sequence similarity (15% amino acid identity (40). EndoII shows 17% amino acid identity to UvrC from Fusobacterium nucleatum (EAA24392; 15% identity to Tma UvrC (Q9WYA3) (Fig. 1) and 12% identity to I-TevI (P13299) and can be threaded into either structure (see Fig. 1b). Since EndoII also shares all amino acids shown to be important for catalysis by UvrC (Fig. 1), we consider it likely that its catalytic surface is organized similarly.

FIG. 1.

(a) Alignment of T4 EndoII (AAD42558), RB49 EndoII (one of 13 [Oct 2007]) completely sequenced T4-like phage that carry a denA homolog (genome sequences available at http://phage.bioc.tulane.edu/; the presence of the MR was verified here in 48 additional phages) and the N terminus of Thermotoga maritima UvrC (Q9WYA3). The GenBank annotation of T4 EndoII indicates a start 8 amino acids upstream of that shown here. However, that start codon (GUG) is not conserved in T4-related phage having denA homologues and therefore not included here. Symbols (*, identity; colon, conservation of strong groups; point, conservation of weak groups) below the UvrC Tma sequence denote sequence similarity between that enzyme and T4 EndoII; symbols above the RB49 sequence denote similarities among the 13 denA homologs. Motifs A, B, D, and E, conserved in GIY-YIG endonucleases (20), are boxed in all three sequences. In Tma UvrC structural elements are located as follows: β1, amino acids 17 to 23; β2, amino acids 26 to 31; α2, amino acids 33 to 42; α3, amino acids 48 to 58; β3, amino acids 60 to 64; α4, amino acids 68 to 84 (8 Tma UvrC coordinates; PDB 1YD0). The NTR and MR regions are boxed in the two EndoII sequences. Residues in Tma UvrC shown to be important for its activity (40) are marked with filled circles (•) below the Tma UvrC sequence. T4 EndoII residues where mutations (to alanine unless otherwise noted above the symbol) were introduced and analyzed in the present study are marked above the T4 EndoII sequence. Symbols between the RB49 and T4 EndoII denote NU and BU per ng for the respective mutants (see Table 2). Nicking: ★, similar to wild type (not measured); ▪, ≥25 NU/ng; ▴ or ▵, 1 to 25 NU/ng; ▾ or ▿, ≤1 NU/ng; •, 0 NU/ng. Binding: solid symbols, >0.15 BU/ng, open symbols, ≤0.15 BU/ng. (b) Structural model of EndoII residues 34 through 136, based on the structure of Tma UvrC (1YD0). The two images to the left show EndoII at 90° angles; the image to the right shows Tma UvrC in the same orientation as the middle EndoII representation. Amino acid residues shown here to be particularly important for recognition and catalysis by EndoII are highlighted.

As a “stand-alone” GIY-YIG endonuclease lacking sequences corresponding to the C-terminal domains of the UvrC and homing endonucleases and with easily studied sequence preferences, EndoII offers a unique opportunity to investigate how the catalytic domain of a GIY-YIG enzyme can recognize, bind, and interact with its targets. We report here an analysis of residues important for these interactions. Our results, summarized in Fig. 1, point to roles in binding, as well as catalysis for residues previously implied in the catalysis by GIY-YIG enzymes and for residues in the N-terminal region (NTR) and middle region (MR) of EndoII.

MATERIALS AND METHODS

Strains and plasmids.

Escherichia coli BL21(DE3)/pLysS (Novagen) was used for the overexpression of EndoII. Site-directed mutagenesis (using a QuikChange site-directed mutagenesis kit from Stratagene) was carried out according to the manual to generate plasmids carrying specific denA mutations and altogether six His codons between EndoII and the pelB leader and to mutagenize the DNA substrate. Plasmids are listed in Table S1 in the supplemental material. Plasmid DNA was purified by using a Qiaprep spin miniprep kit (Qiagen), and DNA concentrations were estimated by ethidium bromide fluorescence or using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies).

Oligonucleotides.

Oligonucleotides were purchased from Pharmacia, Life Technologies, or Sigma Genosys. The sequences of mutagenizing oligonucleotides are presented in Table S2 in the supplemental material.

Radioactive labeling of DNA.

Oligonucleotides were labeled with 20 pmol of [γ-³²P]ATP (Dupont/New England Nuclear) and 6 U of T4 polynucleotide kinase (USB) per 50 pmol oligonucleotide (in 1× polynucleotide kinase buffer, 50 μl). Radiolabeled oligonucleotides were passed over a G50 column to remove unincorporated isotope.

PCR amplification.

PCRs were carried out with 50 pmol of each primer, up to 1 μg of DNA, and two Ready-to-Go beads from Pharmacia in a final volume of 50 μl and with 35 amplification cycles at 95, 55, and 72°C. The different amplicons are shown in Table 1. Amplicons were purified by polyacrylamide gel electrophoresis and stored at −20°C.

TABLE 1.

PCR amplicons

Amplicona	Primersb	Substrate	Length (bp)
I	714-734, 1404-1381	pBR322	691
II	3659-3639, 3540-3562	pPL3604	148
III	3659-3639, 3540-3562	pPL3602	120

^aAmplicon II contains 24 bp from the 807C region cloned at coordinate 3611 of pBR322, while amplicon III contains pBR322 DNA from the same region lacking the embedded 807 site. Both amplicons were derived from a plasmid with six bases to the left of 807C region replaced to remove a nicking site utilized frequently by the wild-type enzyme.

^bPrimers were named for their starting and ending pBR322 coordinates, 5′ to 3′.

Preparation of EndoII.

Wheat germ extracts (Promega) were used to prepare EndoII in vitro by sequential transcription and translation.

To overexpress EndoII in vivo, BL21(DE3)/pLysS containing an EndoII-expressing plasmid derived from pACN01 (21) was grown at 30°C in TB (containing per liter: tryptone, 13.3 g; yeast extract, 26.7 g; glycerol, 4.5 ml; KH₂PO₄, 2.6 g; K₂HPO₄, 14 g [pH 7.2 to 7.4]) supplemented with carbenicillin to 50 μg/ml and chloramphenicol to 40 μg/ml. All constructs yielded EndoII with an amino-terminal PelB and an His₆ tag; constructs expressing mutants R57A and E118A were also prepared without the pelB leader. At an optical density at 600 nm of ∼0.6, enzyme expression was induced by the addition of IPTG (isopropyl-β-d-thiogalactopyranoside) to 1 mM, and the cultures were incubated for an additional 2 h. The cells were pelleted, resuspended in buffer 1 (20 mM Tris-HCl, 1% Triton X-100, 20% glycerol) containing 10 mM imidazole and 1 M NaCl, and frozen at −70°C. Upon thawing they were sonicated with an MSE 100-W ultrasonic disintegrator three times for approximately 10 s each time (amplitude, 5 μm). After centrifugation at 43,000 × g and filtration through a 0.45-μm-pore-size filter, His-tagged EndoII was obtained by affinity chromatography. The supernatant was loaded on a 1-ml HiTrap Chelating HP column charged with NiSO₄ (Amersham Pharmacia Biotech), which was then washed with buffer 1 containing 300 mM NaCl and 40 to 60 mM imidazole. Bound EndoII was eluted with buffer 1 containing 300 mM NaCl and 300 to 500 mM imidazole. The eluate was desalted on a 5-ml HiTrap desalting column (Amersham Pharmacia Biotech), using 50 mM sodium phosphate buffer (pH 7.9), 10 mM NaCl, 1 mM dithioerythritol, and 20% glycerol.

Determination of protein concentrations.

Only EndoII E118A and R57A were obtained in sufficient quantities and purity to permit concentration determination using a Bio-Rad protein assay with bovine gamma globulin as a standard. Concentrations of the other mutant enzymes were determined by comparing the staining intensities of the EndoII bands in Western blots relative to that of different amounts of EndoII R57A analyzed on the same blot, using QuantityOne 4.6.3 (Bio-Rad). Concentrations of the wild-type enzyme could not be estimated since it was not possible to obtain it in sufficient quantities to be visible on a Western blot.

Protein gels.

Proteins were analyzed on discontinuous 5% (stacking)/14% (separating) sodium dodecyl sulfate (SDS) polyacrylamide gels (30:0.8; Bio-Rad) with 0.025 M Tris-0.192 M glycine-0.1% SDS (pH 8.3) as a running buffer. Gels were run in a Mini-Protean II cell apparatus (Bio-Rad) at 170 V for 65 min. After electrophoresis, the gels were fixed and silver stained essentially as described by Oakley et al. (25) and finally dried between cellophane sheets.

Western blotting.

After electrophoresis, the gels were blotted and probed essentially as described by Gallagher et al. (16). Proteins were transferred to Immobilon-P (Millipore) transfer membranes in the Mini-Protean II cell (Bio-Rad) blotting apparatus with transfer buffer (0.025 M Tris, 0.192 M glycine, 20% methanol [pH 8.3]) at 30 V overnight at 4°C. The primary antibody was monoclonal anti-His₆ antibody (Amersham, diluted 1:3,000 in Blotto), and the secondary antibody was horseradish peroxidase-conjugated sheep anti-mouse immunoglobulin G antibodies (Amersham, diluted 1:1,000). The membrane was then developed with enhanced chemical luminescence reagent (Amersham) and exposed to X-ray film.

EndoII activity assays.

Substrate DNA was incubated with EndoII (diluted in 1× reaction buffer) in 1× reaction buffer (10 mM Tris-HCl [pH 8], 10 mM MgCl₂, 1 mM NaCl, 0.1 mM Na₂EGTA) for 1 h at 30°C. After incubation, the samples were treated with EDTA (to 50 mM), Triton X-100 (to 1%), and proteinase K (to 0.1 mg/ml) for 10 min at 37°C. Native gels contained 8% polyacrylamide (19:1) in 1× Tris-borate-EDTA (TEB), whereas denaturing gels contained 8% polyacrylamide (19:1) in 1× TEB-7 M urea. For loading on native gels 5-μl sample was mixed with 5 μl of 20% (vol/vol) glycerin, 20 mM Na₂EDTA, 20 mM Tris-HCl (pH 7.5), 20 mM NaCl, 0.3% (vol/vol) Triton X-100, 0.05% (wt/vol) bromophenol blue, and 0.05% (wt/vol) xylene cyanol. Before the cleavage products were analyzed on denaturing gels, samples were denatured at 90 to 95°C for 5 min (5-μl sample plus 5 μl of 95% [vol/vol] formamide, 20 mM Na₂EDTA [pH 8], 0.05% [wt/vol] bromophenol blue, 0.05% [wt/vol] xylene cyanol). One cleavage unit of enzyme activity was defined as previously (21) as that amount cleaving 50% of all substrate molecules using 1 ng of intact linear pBR322 DNA for 60 min at 37°C. Since some of the mutants did not cleave the substrate, one nicking unit (NU) of enzyme activity was defined as the amount nicking the bottom strand in 25% of all substrate molecules at least once, using 1.25 ng of amplicon II for 60 min at 30°C. For the wild-type enzyme, these two units roughly agree. In the present study “nicking” refers to single-strand cutting, while “cleavage” refers to double-strand cutting. “Top” and “bottom” strands refer to the conventional way of writing the pBR322 sequence (GenBank JO1749, as shown in Fig. 4b). The sizes of fragments, and thereby nicking positions, were determined from their relative mobility on denaturing gels using sequence ladders and restriction fragments as size standards.

FIG. 4.

(a) Nicking of amplicon II labeled in the bottom (lanes 1 to 10) or top (lanes 12 to 18) strand. Samples are from the same experiment as that shown in Fig. 3b, separated by 8% denaturing PAGE. Numbers to the left and right identify sizes of standards in nucleotides (nt) (intact and PstI-cleaved amplicon III labeled in the bottom or top strand [120, 72, and 52 nt, respectively, lanes 3, 11, and 19]) and intact amplicon II (148 nt); large boldface numbers 4, 5, and 10 identify fragments resulting from nicking at these positions within the 807C region (67, 72, and 85 nucleotides), respectively (see panel b). For each mutant, the relative preference for different nick sites did not change much between 0.4 and 2 NU per assay (data not shown). (b) Map of the central part of amplicon II. Arrows point to the strand and position nicked by EndoII. Sites 4, 5, and 10 are the preferred nick positions A, B, and C previously described within cleavage region 807C (7) (indicated by an oval). The amplicon contains 51 bp to the left of the sequence shown and 53 bp to the right.

EndoII binding assays.

The gel shift substrate was the same PCR amplicon II (148 bp) used for cleavage assays (see Table 3). Various amounts of EndoII were mixed with substrate (1.25 ng) on ice in 10 μl of binding buffer (final concentrations 10 mM Tris-HCl [pH 8.3 at room temperature], 5 mM Na₂EDTA, 30 mM NaCl [varying the salt concentration between 10 and 40 mM slightly affected the extent of binding but not the shift patterns], 10% glycerol, 0.3 mg of bovine serum albumin/ml). Samples were incubated for 15 min at 30°C (incubations for shorter times or at lower temperatures gave essentially similar results). Complexes were loaded directly (no sample buffer) on precooled (+4°C) 5% (37.5:1; Bio-Rad) nondenaturing polyacrylamide gels in 1× TEB (pH 8.3) and resolved at 8 W for 3 h at +4°C. One binding unit (BU) was defined as the amount of protein that shifted 25% of 1.25 ng of this substrate under these conditions as quantified by densitometry.

TABLE 3.

Relative nicking efficiency of wild-type and mutant EndoII in the presence of different divalent cations

EndoII mutant	Relative nickinga
	Top strand			Bottom strand
	Mn	Ni	Ca	Mn	Ni	Ca
Wild type	0.4	0.1	ND	0.8	0.1	ND
G49A	2	0.2	0.3	2	0.3	0
R57A	0.5	0.1	0.2	0.5	0.1	0
S72A	0.2	0.2	0.06	0.08	0.7	0.05
K76A	0.7	0.4	0.1	0.2	0.8	0.03
L84P	0.1	0.2	0.2	0.2	0.1	0
P127L	0.6	0.5	0.5	0.4	0.4	0.6
N130A	0.5	0.5	0.2	0.5	0.8	0.2

^aSummed nicking frequencies were calculated for all mutants with all cations, from data shown in Fig. 7 and similar gels. The numbers show the summed nicking frequency at the same strand by the same enzyme in the presence of Mn²⁺, Ca²⁺, or Ni²⁺ relative to that in the presence of Mg²⁺. ND, not determined.

Other reagents and chemicals.

Restriction endonucleases, Klenow DNA polymerase, and T4 DNA ligase were from Pharmacia or Amersham and used as recommended by the manufacturer. Sequenase 2.0 was from U.S. Biochemicals, and GeneClean was from Bio 101, Inc. Agarose, ethidium bromide, and the other chemicals used for electrophoresis were purchased from Sigma. All other chemicals were reagent grade.

Data analyses.

Homologs to DNA and protein sequences were searched by using BLAST (1) (http://www.ncbi.nlm.nih.gov/BLAST). Protein sequence and structure alignments were carried out using CLUSTAL W at the Biology Workbench (http://seqtool.sdsc.edu/CGI/BW.cgi) and the SwissModel Alignment Interface (http://swissmodel.expasy.org/), respectively. Protein structures were viewed by using the Swiss PDB viewer (http://www.expasy.org/spdbv/text/download.htm), and secondary structure predictions were carried out by using PSIPred (http://bioinf.cs.ucl.ac.uk/psipred/) (22) and PredictProtein (http://www.predictprotein.org/) (32); both programs predict the positions of α-helices and β-sheets in Tma-UvrC and I-TevI that are correct within a couple of amino acid residues.

Homology modeling of EndoII.

The structure of EndoII residues 34 through 136 was modeled by using MODELLER (33; http://www.salilab.org/modeler/) with the crystal structure of the Thermotoga maritima UvrC N-terminal domain (PDB code 1YD0) as a template. Four out of seven secondary structure elements present in UvrC have homologous sequences in EndoII. These were fixed in the alignment; β-strands 1 and 2 (motif A) and α-helices 2 (motif B) and 4 (motif D) (secondary structure elements counted from the N-terminal end of UvrC). EndoII residues 72 through 84 (EndoII numbering) that partially overlap nuclease-associated DNA-binding domain 3 (NUMOD3) but lack homologies in the Tma UvrC N-terminal domain were modeled as an α-helix (MODELLER keyword “rsr.add”) based on homologous sequences in the C-terminal of I-TevI (PDB code 1I3J). EndoII has no sequence homology to the UvrC β-strand 3, but since the region was predicted to contain β-strands, a β-stand was modeled here. The N-terminal domain in EndoII was left out in the final model since it could not be modeled reliably. Images were prepared in the PyMOL Molecular Graphics System (http://pymol.sourceforge.net/).

RESULTS

Structural model of EndoII.

As a guide in the analysis of importance of different amino acids in catalysis by EndoII, we modeled EndoII using the structure of the homologous sequences of Tma UvrC (PDB code 1YD0). The NTR and MR of EndoII (see Fig. 1) lack homologies to other GIY-YIG enzymes, and since it has not been possible to obtain crystals of EndoII, their structures are unknown. However, structure prediction programs suggest that the NTR is mainly unstructured or β-sheets, while the MR is helical.

I-TevI and UvrC both have heterologous α-helices between GIY-YIG motifs B and D, where the MR is located. Partially overlapping the MR, EndoII positions 67 to 80 show similarities to the NUMOD3. This motif is present in the C-terminal DNA-binding domains of several homing endonucleases (39) and has been shown to be helical in I-TevI (38). Figure 1b shows a model of EndoII based on these homologies and similarities; the N-terminal 33 amino acids of EndoII were left out since this part could not be modeled reliably. Based on EndoII structure predictions and the NUMOD3 structure, the N-terminal part of MR (amino acids 72 to 84) was modeled crudely as a helix. Because of the large number of charged and polar residues, the MR may form an extended structure that does not necessarily pack tightly to the main body of the enzyme; the crystal structure of the NUMOD3 sequence in the C-terminal domain of I-TevI (PDB code 1I3J) shows that it binds to DNA in a very extended conformation. Four out of ten residues in the N-terminal end of the NUMOD3 element in EndoII are basic (Lys and Arg) and could interact with the phosphate backbone of DNA. Alternatively, this helix might be located in the same position as helix 3 of Tma UvrC (see Fig. 1b); this would position the catalytically important residue K76 close to the active surface.

Inspecting the final model visually for the presence of structural features not consistent with globular proteins revealed a large cluster of relatively exposed hydrophobic residues at the surface opposite the catalytic surface. However, it is not unlikely that the N-terminal residues of EndoII packs close to this patch in the true structure.

Isolation of mutant EndoII.

Amino acids in EndoII (see Fig. 1) were selected for mutagenesis based on three criteria: (i) their importance for catalysis and their position on the putative catalytic surface (G49, R57, E118, and N130) (20, 40, 44); (ii) the loss of in vivo activity of EndoII (E91G, L84P and P127L) (3, 18, 43; K. Carlson and A. C. Nyström, unpublished data); and (iii) conservation of residues in regions unique to EndoII—the MR that is conserved in EndoII homologues in T4-like phage (http://phage.bioc.tulane.edu/; S72, K76, I80, A83, E86, and V90) and the less-conserved NTR (K12, L16, and I24). In addition, a glutamic acid next to E118 (E117) and a proline next to P127 (P128) were selected to test whether the two residues in these two neighboring sets had distinct roles. All mutants were alanine substitutions except as noted in Fig. 1 and created by using site-directed mutagenesis.

Active EndoII very rapidly degrades the template from which it is being produced, and it has not been possible to protect the template from degradation. Therefore, only mutant enzymes with low activity can be obtained by in vivo overexpression in sufficient quantities for good affinity purification (Fig. 2). Ion exchange or size-exclusion chromatography of the R57A and E118A enzymes did not yield purer enzymes. The predominant contaminant (arrow in Fig. 2) was identified by mass spectrometry as an FKBP-type peptidyl-prolyl cis-trans isomerase (31). Mass spectrometry of the purified R57A enzyme identified it unambiguously as EndoII. Wild-type EndoII, as well as mutant enzymes I80A and E86A, could not be detected by Western blotting upon overexpression in vivo, indicating high catalytic activities causing degradation of the template. Wild-type EndoII and some of the most active mutant enzymes were therefore produced in vitro by consecutive transcription and translation. All mutants were expressed with an N-terminal pelB-His₆ tag fusion. The presence or absence of the PelB leader in EndoII mutants R57A and E118A did not affect their catalytic or binding properties (see Fig. 6), as previously found for the wild-type enzyme (4).

FIG. 2.

SDS-polyacrylamide gel electrophoresis analysis of EndoII. Proteins were visualized by silver staining. Lanes: 1, crude extract from cells expressing R57A before induction; 2, same crude extract after 2 h of IPTG induction; 3, affinity-purified EndoII R57A; 4 and 6, molecular weight standards (low molecular weight standards; Bio-Rad); 5, affinity-purified EndoII E118A. Lanes 1 to 4 were from different portions of a single gel, lanes 5 to 6 were from different portions of another, less-exposed, gel. The bands marked with asterisks are EndoII. The arrows point at FKBP-type cis-trans-isomerase, the major contaminant in our preparations. Both bands at the asterisk in lane 3 were identified as EndoII by mass spectrometry; the faster-migrating band is likely a degradation product from EndoII that could not be removed by additional purification steps. The purities of the R57A and E118A enzymes were similar; the remaining enzyme preparations were of somewhat lower purity.

FIG. 6.

Sequence preference for nicking by EndoII. Twenty nick sites were aligned according to the position of the nicks, which are in the opposite strand between bp 8 and 9 of the sequence shown here. The frequency of nicking at a particular site was calculated for each mutant as the ratio between the amount of label in that fragment to the amount of label in DNA of this size and larger (21); at the low enzyme concentration used here most molecules are nicked only once. These frequencies were used to weight the respective sequences (6) to calculate the information content (36) at each of the 16 nucleotide positions for each enzyme, subtracting the maximal baseline variation due to small number of sites [E(n) = 0.114 for n = 20] and taking into account the average base composition of the substrate (maximal information content, 1.98 bits). The probabilities of base occurrences are calculated as logarithms to make them additive, and the use of log₂ sets the range of information content between 0 (log₂ 1 = 0, all bases appearing with equal frequency) and 2 (log₂ 4 = 2, only one base appearing) (36). Numbers after the respective strain designations in the panels show the total information content for the 16 positions (Rseq) in bits (standard deviation = 0.4). (a) Information content at wild-type, K12A, S72A, P127L, and N130A sites. The consensus sequence of wild-type sites is shown below the abscissa. The consensus sequences for mutants not represented here were essentially the same as that for the wild type. (b) Information content at the R57A, R57A without the pelB leader, G49A, K76A, and L84P sites. The consensus sequences at R57A and G49A nick sites are shown below the abscissa. Lowercase letters in consensus sequences denote information content of 0.2 to 0.4; uppercase letters denote information content of >0.4. Abbreviations: H = A, C, or T; K = G or T; M = A or C; R = A or G; S = C or G; V = A, G, or C; and Y = C or T.

Cleavage.

The cleavage activity of different EndoII mutants and wild-type EndoII was initially screened by using in vitro-translated enzymes and a long amplicon (amplicon I, 691 bp, Table 1) derived from pBR322 containing the in vitro-favored EndoII cleavage regions 807N and 827N, in addition to several minor nicking and cleavage sites (7). In vitro-translated EndoII mutants P128A, L84A, E91G, and E117A (Fig. 3a and data not shown) were catalytically active, with cleavage preferences similar to those of the wild-type enzyme and therefore not further investigated; the E91G enzyme was somewhat less active than the others. No activity could be discerned in in vitro-translated preparations of G49A, R57A, L84P, E118A, P127L, and N130A (Fig. 3a and data not shown). Since it was not possible to purify the in vitro-translated protein for more detailed studies, we switched to in vivo overexpression of these seemingly inactive enzymes. A shorter amplicon (amplicon II, 148 bp, Table 1) containing the in vitro favored cleavage region 807C (16 bp from region 807N cloned into the PstI site of pBR322) and a number of nicking sites (7) (see Fig. 4b) was used to test cleavage by in vivo-overexpressed mutant enzymes. All but E118A were now found to cleave the substrate (Fig. 3b). For G49A, R57A, and L84P the cleavage patterns differed from that of the wild-type enzyme (little cleavage or different cleavage pattern), indicating perturbed substrate interaction.

FIG. 3.

Cleavage by wild-type and mutant EndoII. Samples were separated by 8% nondenaturing PAGE. (a) Cleavage by in vitro-prepared enzymes. Amplicon I was digested by two cleavage units of the wild-type enzyme or corresponding amounts of mutant enzymes for 60 min at 30°C. Lane 1, 100-bp ladder (Amersham Pharmacia Biotech); sizes are shown to the left. To the right of the panel the positions of the uncleaved amplicon (691 bp), as well as of the fragments resulting from cleavage at regions 807N and 827N, are shown. (b) Cleavage by in vivo-overexpressed mutants and in vitro-expressed wild-type EndoII. Amplicon II was digested with approximately 1 NU of each enzyme (≤24 ng; 65 ng of E118A). The size markers are a mixture of uncleaved amplicon III (120 bp) and amplicon III labeled in the top strand (68 bp) or bottom strand (48 bp), cleaved with PstI; the sizes are shown to the left and right. The boldface 807C identifies fragments cleaved in the 807C region (see Fig. 4b).

Nicking.

Since cleavage by EndoII is the result of two consecutive nicks (7), perturbed substrate interaction can be studied more precisely as nicking, not cleavage. To enable analysis of the two strands separately, the short substrate was labeled in either end, and the products were analyzed on denaturing gels. Representative results are shown in Fig. 4. No enzymatic activity was found for the E118A mutant (however, faint bands can be seen in Fig. 7). For all active mutants except for R57A, the cleavage activities were 20 to 70% of the nicking activities (compare Fig. 3b with Fig. 4a, showing analyses of the same samples), as previously found for the wild-type enzyme (7). For the R57A mutant this number was 5 to 7%, and we were never able to obtain more than 25% total nicking with the R57A enzyme over a 40-fold variation in enzyme added (data not shown). These data suggest that a significant fraction of this mutant enzyme forms enzyme-substrate complexes impaired in catalytic activity.

FIG. 7.

Nicking by wild-type and mutant EndoII in the presence of Mn²⁺, Ni²⁺, or Ca²⁺. Top-strand-labeled amplicon II was digested with the same enzyme preparations as used for Fig. 3b, 4a, and 5, using approximately 1 NU of each enzyme (≤24 ng; 65 ng of E118A) in the presence of 10 mM Mg²⁺ (data not shown), Mn²⁺ (lanes 2 to 8), Ni²⁺ (lanes 10 to 16), or Ca²⁺ (lanes 18 to 23). Samples were analyzed by 8% denaturing PAGE, together with PstI-cleaved amplicon II labeled in the top or bottom strand (lanes 1, 9, 17, and 24; sizes are shown to the left, together with the size of the fragment resulting from nicking at position 10 [see Fig. 4b]). Samples analyzed in lanes 1 to 16 were analyzed on one gel (together with samples cleaved in the presence of Mg²⁺), while lanes 17 to 24 are from a separate gel.

All favored nick sites and most unfavored ones were the same for all EndoII variants investigated, although utilized with different preferences (Fig. 4a and data not shown). This was most notable at nick site 10. Nicking here comprised ca. 50% of the total nicking by the wild-type enzyme and most of the NTR and MR mutants; other mutants nicked this site to a lower extent (ca. 30% for S72A and N130A, ca. 20% for K76A, L84P, and P127L, and ca. 5 to 10% for G49A and R57A). The nicking patterns of mutant enzymes G49A, R57A, K76A, and L84P deviated the most from those of the wild-type enzyme, suggesting the importance of these residues in the interaction between enzyme and substrate.

The catalytic activities of the mutants are summarized in Table 2. Mutation of the GIY-YIG conserved residues and MR residues K76 and L84P resulted in much reduced catalytic activities, as did as the P127L mutation. Other MR mutants and the NTR mutants had higher catalytic activities, the most active one (K12A) being 2 orders of magnitude more active than the least active mutant (L84P) (Table 2). Since the concentration of the wild-type enzyme was too low for determination by the methods used here, mutant activities cannot be compared to the activity of the wild-type enzyme. We have previously estimated the EndoII concentration in in vitro-translated preparations to be around 10 nM (2 ng/ml) by [¹⁴C]leucine incorporation (4). With such a concentration, the in vitro-translated wt enzyme would contain about 8,000 NU per ng. Although we do not routinely monitor these EndoII preparations through ¹⁴C labeling, the variations in enzymatic activity are small, less than 1 order of magnitude from one preparation to another.

TABLE 2.

Activity of EndoII mutants

EndoII mutant	NU/nga	BU/ngb	NU/BUc	Symbold
K12A	25	0.25	100	▪
L16A	3.0	0.21	14	▴
I24A	5.8	0.21	28	▴
G49A	0.11	0.13	0.85	▿
R57A	0.048	0.079	0.61	▿
S72A	3.2	0.15	21	Δ
K76A	0.57	0.34	1.7	▾
A83L	6.8	0.22	31	▴
L84P	0.042	0		▿
V90A	11.8	0.24	49	▴
E118A	0	0.31	0	•
P127L	0.78	0.09	8.7	▿
N130A	1.0	0.23	4.3	▾

^aNU per ng of EndoII as estimated from Western blots from activity measurements using different ratios of EndoII to DNA. Estimates of nicking units per ng of enzyme varied by ca. 5 to 25% with different enzyme preparations, suggesting that variations of >0.02 NU/ng among the least active mutants and >2 NU/ng among the more active mutants were significant.

^bBU per ng of EndoII as estimated from Western blots from EMSA titrations using the same enzyme preparations as for the activity determinations. Estimates of binding units per ng of enzyme varied by 10 to 40% with different enzyme preparations, indicating that the highest values (K76A and E118A) were significantly different from the lowest (R57A and P127L).

^cThat is, the ratio of the numbers in columns 2 and 3.

^dThe symbols, used in Fig. 1, denote NU/ng and BU/ng levels. Nicking: ▪, ≥25 NU/ng; ▴ and Δ, 1 to 25 NU/ng; ▾ and ▿, ≤1 NU/ng; •, 0 NU/ng. Binding: solid symbols, >0.15 BU/ng, open symbols, ≤0.15 BU/ng.

Binding.

The ability of the mutant enzymes to bind to the substrate was tested in the presence of EDTA to prevent catalysis, using an electrophoretic mobility shift assay (EMSA). It was not possible to test the binding abilities of the wild-type enzyme in this assay due to the high background of the wheat germ extract. Typical results are shown in Fig. 5. All mutant enzymes except L84P (up to 140 ng added per 1.25 ng of DNA) bound to the substrate, producing distinct shifted bands (Fig. 5 and data not shown). The rates of migration of the shifted bands were similar for all binding mutants except E118A, which produced several faster-migrating complexes not seen with the other mutant enzymes at any enzyme/substrate ratio (Fig. 5 and data not shown). These faster-moving E118A complexes could be chased into the more slowly moving ones with increasing enzyme/substrate ratios (data not shown). The nature of these complexes will be addressed elsewhere (unpublished data). Our preliminary experiments suggest that native EndoII forms dimers and higher multimers. Possibly, the E118A enzyme is defective in multimerization and therefore forms fewer higher-order multimers than other mutants.

FIG. 5.

Binding of EndoII mutants to DNA. EndoII mutants (0.7 to 2.5 BU of enzyme; [7 ng of L84P] and 1.25 ng of radiolabeled amplicon II) were tested in an EMSA. Lane C, control (no enzyme added). The asterisks mark the position of the unshifted substrate.

Binding propensities of the different mutant enzymes are also summarized in Table 2. Differences in binding (among those enzymes that did bind) were not as variable as the differences in catalysis, but the E118A and K76A enzymes bound relatively more efficiently to DNA and mutants P127L and R57A least efficiently to DNA. The NU/BU ratios in Table 2 illustrate the relative importance of the different residues for binding and catalysis. The low NU/BU ratios for mutants affecting MR residue K76 and the universally conserved GIY-YIG residues compared to those for the NTR mutants and most other MR mutants suggests that the former mutations affect enzyme-substrate interactions required for catalysis to a higher extent than the latter.

Sequence recognition.

Analysis of the DNA sequence preference of different mutants provides information on how the enzyme-substrate interaction is perturbed by the different amino acid substitutions and thereby sheds some light on the roles of individual amino acids in these interactions. Sequence conservation implies direct or indirect recognition of the respective base. To obtain easily comparable numeric values for sequence recognition, the information content was calculated for the sites shown in Fig. 4a and six additional sites where nicking was less frequent and therefore not visible in this figure. The information content, a measure of the extent of base conservation at each position of a sequence, is measured in bits, ranging from 0 if all four bases appear with equal frequency at the position, to 2 if only one base appears (36). For each mutant all sites were weighted by the frequencies with which they were nicked (6); the frequencies of the different bases in the substrate and the small sample size were also taken into account.

The total information content (Rseq) at the nick sites was highest for the wild-type enzyme (9 bits) and lower for all other mutant enzymes (3 to 8 bits), confirming their reduced precision (Fig. 6). Only the CG base pair in position 5 was strongly conserved for all mutants. The relative information content at each position and also the bases preferred by most NTR and MR mutants, as well as P127L and N130A, was similar to that for the wild-type enzyme (Fig. 6a and data not shown), suggesting that these mutants recognized the DNA approximately the same way, although with varying precision. For these strains, three of the four two-base ambiguities resulted in retained hydrogen-bonding propensities (occurrence of amino and keto groups, respectively, e.g., M [A or C] or K [G or T]) in the major groove, suggesting the possibility of involvement of such hydrogen bonds in binding the protein.

For mutants K76A, L84P, G49A, and R57A (Fig. 6b), on the other hand, both the relative information content at each position and the preferred bases were different from those for the wild type, especially in the middle (positions 7 to 9) and on the right side (positions 11 to 15) and also different from each other. For these mutants, sequence conservation was less strong on the left side (positions 1 to 4) of the consensus sequence. The G49A enzyme showed an increased preference for A and G in positions 7 and 8, respectively. The preference for G in position 8 was shared by the R57A enzyme, which in addition showed base preferences in positions 4 and 13 where no other enzyme showed any preference. Thus, nicking by these mutants required somewhat different interactions between the enzyme and substrate than was the case for the enzymes analyzed in Fig. 6a.

Catalytic activity with different divalent cations.

Since EndoII does not nick DNA in the absence of divalent cations (7), we tested Mn²⁺, Zn²⁺, Co²⁺, Cd²⁺, Sr²⁺, Cu²⁺, and Ni²⁺, in addition to Mg²⁺, for their ability to support nicking by EndoII (data not shown). In the presence of Mn²⁺ the wild-type enzyme showed approximately half the activity seen in the presence of Mg²⁺, and in the presence of Ni²⁺ it showed ca. 10% of the activity. No activity could be discerned in the presence of any of the other cations tested.

The mutants whose sequence preference was analyzed in Fig. 6 (except for K12A) were tested with the three ion cofactors that supported wild-type activity and also with Ca²⁺ (which cannot be tested with the wheat germ-translated wild-type enzyme). The results are shown in Fig. 7 and Table 3. The mutant E118A in this set of experiments showed very low activity, amounting to nicking of ≤1% of all substrate molecules and only at position 10. Thus, the residual activity of this enzyme appears to be similar in sequence preference to that of the wild-type enzyme. Ni²⁺, and for most mutants also Mn²⁺, supported less enzymatic activity than Mg²⁺ (10 to 70%). The G49A mutant, however, was more active with Mn²⁺ than with Mg²⁺. Using Ca²⁺, all mutants nicked the top strand with efficiencies similar to those with Mn²⁺ or Ni²⁺, but only P127L and N130A nicked the bottom strand noticeably. The sequence preference of all enzymes was somewhat reduced when a cation other than Mg²⁺ was used, as evidenced by lower total Rseq values (data not shown). The plots of Rseq(L) versus position had the same shape as the Mg²⁺ plots shown in Fig. 6 (data not shown), showing that the overall recognition was not altered.

DISCUSSION

The GIY-YIG enzymes form a very diverse group of nucleolytic enzymes (13) catalyzing single-strand breaks or double-strand cleavages at sequences recognized precisely (restriction endonucleases), ambiguously (EndoII and homing endonucleases), or because of damages (UvrC endonucleases). Similarities between EndoII and the N-terminal domains of I-TevI and UvrC render it likely that its catalytic surface has a similar organization. The well-characterized UvrC and I-TevI enzymes contain carboxy-terminal sequences beyond the last conserved amino acid (N130 in EndoII) that provide the major part of the binding energy (10, 12, 19, 34, 38). However, about one-quarter of altogether 784 bioinformatically identified GIY-YIG enzymes (13) are 150 amino acids or shorter, like EndoII, suggesting that they are single-domain enzymes. EndoII therefore provides a unique opportunity to address the role of individual amino acids in the catalytic domain of a GIY-YIG endonuclease in sequence recognition and binding. We have identified altogether 13 amino acid alterations that reduce the activity of the enzyme significantly, and also affect the mode or extent of sequence recognition and binding.

Our data, discussed in the following, are consistent with residues in the putative catalytic surface (G49, R57, E118, and N130) having catalytic roles similar to those described for I-TevI and UvrC. In addition, they are important for recognition and binding to the substrate. We suggest that the conserved glycine and arginine are essential for deformation of the substrate involved in forming catalytically competent enzyme-DNA complexes and that the conserved asparagine and an adjacent proline residue contribute to positioning the catalytic domain correctly. The NTR and MR distinguish EndoII from other characterized GIY-YIG enzymes. The NTR appears to contribute to binding strength without affecting the mode of substrate binding at the catalytic surface. The MR, on the other hand, appears to contribute to both binding and organization of the catalytic surface.

Residues in the GIY-YIG motives. (i) The conserved glutamate is essential for activity and affects the mode of binding.

The E118A EndoII mutant bound efficiently to DNA (Table 2) but produced catalytically inactive DNA-enzyme complexes that were less retarded than those formed by all other mutant enzymes (Fig. 5 and data not shown). The residue that corresponds to E118 in both UvrC (40) and I-TevI (41) coordinates the divalent metal ion necessary for activity. Its importance for catalytic activity in EndoII suggests a similar role here.

Two previous reports have dealt with effects on binding by mutations of this conserved GIY-YIG residue. In R.Eco29kI (19), the corresponding glutamate mutant produces EMSA shifts similar to those of both the wild-type and other mutant enzymes. In I-BmoI the corresponding glutamate mutant produces a slightly variant footprint, suggesting this residue affects enzyme-substrate interactions in this enzyme. However, since stable binding by I-BmoI requires the presence of its C-terminal domain, it is difficult to ascertain the precise role of its glutamate residue in binding. The EMSA results in Fig. 5 provide solid evidence for altered substrate binding by an EndoII mutant defective in the catalytic glutamic acid residue, possibly caused by reduced multimerization capacity by the mutant enzyme.

(ii) The conserved glycine and arginine are important for binding, catalysis, and specificity.

The G49A and the R57A enzymes showed the lowest NU/BU ratios (Table 2), indicating that their nicking was relatively more perturbed than their binding compared to the other mutant enzymes. This demonstrates the importance of these residues for catalysis, as has been found for other GIY-YIG endonucleases (40, 41). Both EndoII mutants recognized their substrate differently, both compared to wild type and compared to each other (Fig. 6). They showed a stronger preference for G at position 8, just to the left of the scissile bond than did the wild type. They also showed reduced base preference in the leftmost and rightmost parts of the sequence, suggesting reduced base-specified contacts to these distal parts. Strong synergistic effects on cleavage by base replacements in positions 8 and 12 in vivo, despite low sequence conservation in both positions (5), suggest an indirect readout by the wild-type enzyme in this part of the sequence, i.e., that a particular DNA structure or deformability is needed for efficient incision to the right of position 8. The structure at steps involving G:s and C:s is context dependent (26), and most flexibility measures suggest that steps involving G:s and C:s are more flexible than steps involving A:s and T:s (26, 28). We suggest that the sequence preference to the left of the nick site for G49 and R57 mutants reflects the need for the DNA to adopt a structure where these mutant enzymes can nick. If the G49A and R57A enzymes have lower ability to distort DNA upon binding, increased flexibility at the scissile bond may facilitate their incision.

The role(s) of the conserved arginine in GIY-YIG endonucleases has been enigmatic. In both UvrC (40) and I-TevI (41) the arginines are positioned in α-helices on the catalytic surface, though located rather far from the catalytic glutamates (8 to 10Å). The I-TevI R27A substitution did not result in perturbance of the overall enzyme structure. R27 of I-TevI could be positioned correctly to interact with the DNA molecule, but there is no direct evidence for a role in DNA recognition (41). The I-BmoI R27A mutant showed a reduced footprint in the region of the scissile bond (10), suggesting this mutation affected binding. This arginine residue has also been suggested to stabilize an intermediate or stabilize the negative charge of the leaving 5′-phosphate after DNA incision in I-TevI (41) and UvrC (40). Few of the EndoII R57A complexes formed appeared to be catalytically proficient, since even a 20-fold excess of the amount required to shift all substrate in the EMSA nicked it to only 25%. This suggests that this mutant EndoII is unable to drive the incision to completion. Moreover, few of these complexes that formed appeared sufficiently stable to yield double-stranded cleavage, since the ratio of cleavages to nicks was much lower for this mutant than for any other. If the arginine of EndoII stabilizes an intermediate, as suggested for UvrC and I-TevI, the indirect readout and the altered sequence preference at the scissile bond suggest that this stabilization involves indirect recognition of residues at the scissile bond.

Unlike wild-type EndoII and all other mutants, the G49A enzyme also showed a higher relative nicking efficiency with Mn²⁺ compared to Mg²⁺ (Table 3) but recognized the same sequence with both ions. In I-TevI the universally conserved glycine residue is part of a β-sheet and positioned next to the catalytic glutamate residue, leaving room between them for the divalent cation required for catalysis (41). In UvrC it is positioned right behind the metal ion (40). If the glycine is similarly positioned in EndoII, it is possible that the slightly smaller Mn²⁺ ion (0.67Å) (14) fits more easily than an Mg²⁺ ion (0.72Å) (14) when G49 is replaced by a bulkier residue. The fairly large effect of the metal species on the catalytic activity, positively on G49A activity and negatively on that of all other mutants (Table 3), may suggest that a metal ion is used in positioning and activating a nucleophile rather than just stabilizing a charge (15).

(iii) Structural roles for the GIY-YIG-conserved N130 residue and the EndoII-conserved P127 residue?

Mutants N130A and P127L showed similar low nicking activity with the same sequence recognition as the wild-type enzyme (Fig. 6a). Alanine replacement of the conserved asparagine residue in I-TevI results in very low activity (20), and it was suggested that the residue is of structural importance (41). In UvrC, the conserved asparagine is connected to the metal ion via two steps of hydrogen bonds (to a second amino acid bonded to one of the water molecules coordinated by the metal ion) (40). An alanine replacement mutant is catalytically inactive, but a structural role for this residue was ruled out by the lack of perturbation by the mutation of the active-site architecture (40). Instead, it was suggested to have a role in positioning the catalytic domain relative to other domains of the enzyme. EndoII N130 is not likely to coordinate any other domains since the enzyme ends only six residues beyond N130 and most likely consists of a single domain, excluding the possibility that it has a role like that assigned to this residue in UvrC.

A proline in a position corresponding to EndoII P127 is conserved among UvrC and a few other endonucleases as part of the conserved motif E (13), though in I-TevI this position is occupied by a serine (20). The structural proximity of the proline in UvrC and the serine in I-TevI to the invariant asparagine residue suggests that they may be involved in positioning that residue correctly. If these residues are located at the EndoII catalytic surface, their replacements may impair proper contacts and bonding between this surface and the substrate, leading to the defects observed. P127L and N130A were the only mutants tested that showed significant activity with Ca²⁺, an ion slightly larger than Mg²⁺. Perhaps the P127L and N130A mutations create more space on the catalytic surface permitting accommodation of a larger cation cofactor.

Residues in the EndoII-conserved NTR and MR regions.

The NTR and MR of EndoII (Fig. 1) show little similarity to the catalytic domains of I-TevI and UvrC enzymes, or to other GIY-YIG endonucleases (10, 11, 19, 23, 30, 35, 37, 42, 45, 46); I-TevI lacks sequences N-terminal to motif A. In I-TevI and UvrC the largely helical heterologous regions between motives B and C are believed to be structurally important, stabilizing the hydrophobic core of the domain (40, 41). The MR sequence (amino acids 72 to 92 in T4 EndoII) and predicted structure (mostly helical) is well conserved in EndoII homologs in T4-like phages, while the NTR is less strongly conserved.

The high NU/BU ratios (Table 2) of mutants with substitutions in the NTR, and most MR mutants, indicate that they are relatively more proficient in nicking than other mutants. Most showed the same sequence preference as the wild-type enzyme, although with somewhat more variability (Fig. 6), resulting in lower Rseq values. We suggest that these regions primarily contribute to binding.

Two MR mutants (K76A and L84P) deviated from this pattern. The L84P mutant was the only mutant tested that failed to bind stably enough to produce a band shift in the EMSA (Fig. 5, Table 2). It also showed perturbed sequence recognition, with a total information content of only 3 bits for the 16 positions analyzed (compared to 9 bits for the wild-type enzyme, Fig. 6) and, as a consequence, little and promiscuous catalytic activity. Since replacement of L84 by alanine did not affect EndoII activity very much (Fig. 3a), it is likely that the proline substitution perturbs the overall protein structure so that it no longer can contact the DNA efficiently. If the MR region in EndoII is located as helix 3 of UvrC (see Fig. 1b), L84 could contribute to stabilizing the hydrophobic core of the protein. Such roles have been demonstrated for the heterologous leucines or isoleucine of I-TevI and UvrC (I-TevI α2 L45, Bca-UvrC α3 L56, Tma-UvrC α3 I54) (40, 41). Although the MR was predicted helical also for the L84P mutant, the confidence of the prediction was lower than for the wild-type sequence, suggesting this substitution indeed may perturb protein structure.

The presence of a NUMOD3 element (39) partially overlapping the MR supports a role for this region in DNA binding. Residues corresponding to S72 and K76 in EndoII are conserved in NUMOD3 elements (39), suggesting their importance for its function. Mutations of both residues in EndoII reduced catalysis (Table 2); despite overall good DNA binding, the K76A mutant showed reduced recognition especially of distal parts of the consensus sequence (Fig. 6a), suggesting that this lysine is important for positioning the enzyme correctly on the DNA and forming catalytically competent complexes. If the MR is positioned as helix 3 of Tma UvrC, this residue (and S72) would indeed be positioned close to the catalytic surface.

EndoII in vivo.

Four of five sequenced in vivo-isolated T4 denA missense mutants carry amino acid substitutions that map within the MR (3), while none perturb any of the important GIY-YIG residues. Although this may be fortuitous, it may also reflect some unknown aspect of the biology of this enzyme. The in vivo-isolated EndoII⁻ mutants L84P, E91G, and P127L all retain some EndoII activity; the severely defective in vivo-isolated mutants L84P and P127L are defective primarily in DNA binding. Possibly, there is some advantage to the phage in retaining at least partially active EndoII and some disadvantage in having an inactive enzyme that is capable of binding to DNA. Among genes encoding proteins involved in host DNA breakdown and phage genome modification, denA is the most ubiquitous, found in all but 1 of 11 T4-related phage (29). Thus, we consider it likely that this enzyme has a second biological role more important to the phage than degradation of host DNA, where its activity appears dispensable (27).