Identification ofTaqI Endonuclease Active Site Residues by Fe2+-mediated Oxidative Cleavage

Metal cofactors (Mg2+ and Mn2+) modulate both specific DNA binding and strand cleavage in the TaqI endonuclease (Cao, W., Mayer, A. N., and Barany, F. (1995) Biochemistry 34, 2276-2283). This work attempts to establish the structural basis of TaqI-DNA-metal2+ interactions using an affinity cleavage technique. The protein was cleaved by localized hydroxyl radicals generated by oxidizing Fe2+ within the metal binding sites. Cleavage fragments were separated by SDS-polyacrylamide gel electrophoresis, and cleavage sites were determined using micropeptide sequencing. Eleven amino acid residues in the vicinity of cleavage sites were selected for site-directed mutagenesis. The negative charge at Asp137 is essential for DNA cleavage but not required for sequence specific binding. Mutations at Asp142 abolish both specific binding and catalysis, except for D142E, which converts TaqI into a completely Mn2+-dependent endonuclease. The positive charge at Lys158 appears to be important for both specific binding and catalysis. Mutations at other sites affect binding and/or catalysis to different degrees, except Trp113 and Glu135, which appear to be nonessential for the TaqI enzyme activity. The critical residues for TaqI function are distinct from the PDX14-20(E/D)XK catalytic motif elucidated from other endonucleases.

TaqI endonuclease recognizes and cleaves a short palindromic double-stranded DNA sequence, T2CGA, with exquisite specificity. Its recognition mechanism has been probed with base analogues (major groove modifications) and thiomethylmodified phosphates (1,2). Major groove contacts, as identified at N 7 of Gly and N 6 and N 7 of Ala, appear to contribute to transition state stabilization (1). The phosphate contacts direct the initial DNA-protein interactions at pTpCpGA for formation of the ground state complex, as well as mediating transition state interactions by providing a controlling point at GpA (2).
Roles of metal cofactors (Mg 2ϩ and Mn 2ϩ ) in modulating substrate specificity have been investigated using both steadystate and single turnover kinetics analysis (3). When Mg 2ϩ serves as the metal cofactor, TaqI exhibits high affinity for its cognate sequence but fails to form specific TaqI-DNA-Mg 2ϩ complex with DNA sequences that differ from its cognate sequence by even one base pair (star sequence), thereby precluding any further catalytic interactions. Mn 2ϩ , the alternative metal cofactor, enables TaqI to form three distinctive tertiary complexes: the highly active TaqI-TCGA-Mn 2ϩ , partially active TaqI-TCAA-Mn 2ϩ (TCAA represents a star sequence), and nonspecific inactive TaqI-TGCA-Mn 2ϩ (TGCA represents a nonspecific DNA sequence). As a result, TaqI complexed with Mn 2ϩ is able not only to cleave its cognate sequence with a high single turnover rate but also to cleave its star sequences at significantly reduced rates.
Much progress has been made in identifying DNA elements involved in substrate recognition and elucidating the roles of the metal cofactors in specific DNA-protein interactions. The structural basis of TaqI protein that governs substrate binding and catalysis has not been clearly defined. In the absence of a high resolution three-dimensional structure, alternative approaches are needed. Hydroxyl radicals generated from metal oxidation are able to attack both phosphate and peptide backbones, resulting in chain breakage (4). Fe 2ϩ -mediated cleavage has been used to determine the metal and metal-substrate sites in pig heart NADP-specific isocitrate dehydrogenase (5). Fe 2ϩ , in place of Mg 2ϩ , has been coupled with tetracycline to map the binding sites of the drug on its repressor protein (6). Cleavage of the repressor was specific and occurred in the immediately proximity of Fe 2ϩ binding, suggesting that this method may be broadly applied in identifying specific interactions between metal and its coordinating sites on the protein (7). Specific cleavage of calmodulin and staphylococcal nuclease have also been reported (4,8).
The active site organization of a number of restriction enzymes has been elucidated by high resolution crystallographic analysis or genetic means (9 -19). One of the common features among these enzymes is that two or more carboxylates of Asp or Glu are involved in coordinating the metal cofactor in the active site. TaqI enzyme requires Mg 2ϩ or Mn 2ϩ as the metal cofactor for its activity (20). Therefore, replacement of the metal cofactor with a redox-active metal ion could lead to specific cleavage at the proximity of the active site. This study reports Fe 2ϩ -mediated cleavage of TaqI protein and site-directed mutagenesis analysis of 11 amino acid residues. The resulting TaqI mutants were grouped into four categories based on their binding and catalysis properties. The active site of TaqI was compared with that of other known type II endonucleases. purchased from New England Biolabs (Beverly, MA). MgCl 2 ⅐6H 2 O and CaCl 2 ⅐2H 2 O were purchased from Fisher Scientific; MnCl 2 ⅐4H 2 O was from Alfa Products (Danvers, MA); methanol was from J. T. Baker (Philipsburg, NJ); and FeCl 2 ⅐4H 2 O, sodium ascorbate, CAPS, HEPES, and other chemicals were from Sigma. [␥-32 P]ATP was purchased from NEN Life Science Products. PVDF membrane was purchased from Bio-Rad. TaqI endonuclease was purified as described previously (21). Kinase-ligase buffer consisted of 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 1 mM EDTA, and 6 mM ␤-mercaptoethanol. Oligonucleotides were synthesized using an Applied Biosystems 384 DNA synthesizer and purified on denaturing sequencing gel (7 M urea/10% polyacrylamide) (22). SDS-PAGE sample buffer (2ϫ) consisted of 0.12 M Tris-HCl, pH 6.8; 20% glycerol; 2% SDS; 0.01% bromphenol blue; 2% ␤-mercaptoethanol. SDS-PAGE running buffer (1ϫ) consisted of 0.025 M Tris-HCl, 0.192 M glycine, 0.1% SDS. Gel shift electrophoresis running buffer consisted of 90 mM Tris, 90 mM boric acid, 10 mM CaCl 2 . MS buffer (1ϫ) consisted of 10 mM Tris-HCl, pH 8.0, at 60°C; 100 g/ml bovine serum albumin; 50 mM NaCl; and 10 mM dithiothreitol. Glycinate agarose gel running buffer (1ϫ) consisted of 200 mM glycine, 12.5 mM NaOH, 0.1 mM EDTA.
Cleavage of TaqI Protein-TaqI protein (6 -150 g) was incubated at room temperature with 2 mM FeCl 2 , 25 mM sodium ascorbate, 10% glycerol, 20 mM HEPES (pH 7.6) in a final volume of 50 l for 2 h. The caps of the 0.5-ml reaction tubes were pierced by a 25 1 ⁄2 gauge needle to allow oxygen flow. The reaction tubes were placed in a low speed vortex to ensure proper mixing. In the case in which a DNA substrate was included in the cleavage reaction, 5 M of annealed duplex oligonucleotide substrate (AMC201/202, see below for sequence) containing a single TaqI recognition site was incorporated into the reaction mixture along with other reagents. The cleavage reaction was terminated by addition of 40 l of 2ϫ SDS-PAGE sample buffer supplemented with 20 mM EDTA.
SDS-PAGE-A 0.1% SDS-15% polyacrylamide gel was prerun at 25 V for 1 h in a Bio-Rad Protean II apparatus filled with 1ϫ SDS-PAGE running buffer. Sodium thioglycolate (approximately 4 mg) was incorporated in the upper gel buffer solution to quench reactive compounds in the gel. The cleavage samples were heated at 37°C in a water bath for 15 min prior to loading to the gel. Electrophoresis was conducted at 50 V overnight.
Electroblotting-The PVDF membrane and the polyacrylamide gel were pretreated and assembled in a Bio-Rad Trans-Blot Electrophoretic Transfer Cell apparatus according to the manufacturer's instruction manual. The transfer buffer contained 10 mM CAPS (pH 11), 10% methanol chilled in a cold room prior to use. After transferring at 30 V for 8 h on ice, the PVDF membrane was rinsed with deionized water, stained in 0.025% Coomassie Blue R-250 solution containing 40% methanol for 5 min, destained with 50% methanol solution for 15 min, and air dried until peptide bands became distinct. Each individual peptide band was excised from the membrane with a clean razor blade and stored at Ϫ70°C until micropeptide sequencing was performed.
Micropeptide Sequencing-N-terminal sequencing was performed on cleavage fragments blotted to the PVDF membrane using a Applied Biosystems protein sequencer (23). In the cases of faint bands, two slices of identical cleavage fragments were subjected to the sequencing run.
Site-directed Mutagenesis-An overlap extension PCR procedure was used for site-directed mutagenesis (24). Partially degenerate internal primers were designed to generated multiple mutations at a particular site. PCR products digested with a pair of restriction enzymes, either HindIII and MluI or MluI and PstI, were ligated to cloning vector pFBT69 treated with the same pair of restriction endonucleases (see Ref. 21). The ligated vectors were transformed into Escherichia coli strain AK53 (mrrB Ϫ , MM294). Plasmids containing inserts were reisolated and sequenced on a Applied Biosystems sequencer using dyedideoxy terminator chemistry to identify the mutated sequence and ensure that the constructs were free of PCR or ligation error. Mutant proteins were expressed and partially purified as described previously (25).
Gel Shift Binding Assay-Oligonucleotide AMC 201, with the single TaqI recognition site underlined, was 5Ј-end labeled as described previously (3) and annealed with a 5-fold excess of AMC202 to form a duplex DNA substrate for the binding experiment as follows. The excess amount of AMC202 did not interfere with our assay (26). Wild-type TaqI and mutant proteins (0.3 nM) were incubated with above duplex DNA substrate (10 pM) in 20 l of binding buffer containing 10 mM Tris-HCl (pH 8.0 at 60°C), 10 mM dithiothreitol, 50 mM NaCl, 0.1 mg/ml bovine serum albumin, 10 mM CaCl 2 , 0.02% bromphenol blue at 60°C for 15 min. The free and bound duplex oligonucleotides were separated on a 6% polyacrylamide gel (30:0.8) containing 89 mM Trisborate and 10 mM CaCl 2 by electrophoresis in a Bio-Rad Protean II apparatus at 100 V for 1.5 h (27). The intensity of 32 P-labeled DNA bands was quantified using a PhosphorImager (Molecular Dynamics). Results were reported as the averages of three independent experiments.
TaqI Activity Analysis by Cleavage Assay-A plasmid substrate containing a single TaqI recognition site (pFB76, 20 nM) was digested with 2 nM of partially purified wild-type or mutant enzymes in the 1ϫ MS buffer at 60°C. The reaction mixture (10 l) was incubated for 5 min with 10 M MgCl 2 as the metal cofactor or 30 min with 2 mM MnCl 2 as the metal cofactor. After quenching the reaction with 5 l of 4ϫ agarose gel loading buffer, samples were electrophoresed on a 0.8% agarose gel. The substrate and product bands were visualized by incorporating ethidium bromide (0.5 g/ml) into the agarose gel and glycinate running buffer. Quantification was performed using ImageQuant software (Molecular Dynamics) after scanning the gel photographs. Results were reported as the averages of three independent experiments.

RESULTS
Cleavage of TaqI Protein-Successful application of Fe 2ϩmediated cleavage depends on specific interactions between the metal ion and TaqI protein. Previously, we have established by kinetic analysis that Mn 2ϩ metal ion may interact with TaqI protein specifically in a DNA-independent fashion (3). To determine whether a redox-active metal such as Fe 2ϩ can serve as a metal cofactor, we tested TaqI activity with Fe 2ϩ along with other divalent metal ions. As shown in Fig. 1A, the only alkaline metal that supported TaqI cleavage was Mg 2ϩ . Mar-  (3), with modifications. The final concentration of purified TaqI enzyme was 0.25 M; 32 P-labeled oligonucleotide substrate, 5 nM; M 2ϩ Cl 2 , 10 mM. The reaction mixtures were incubated at 60°C for 60 min. Cleavage activity was quantified on a PhosphorImager. Arrows indicate that the cleavage reactions had completed long before the incubation at 60°C (1 h) was stopped.
ginal activity was observed with Ca 2ϩ after extended incubation at 60°C for 1 h. For the fourth period transition metals, Fe 2ϩ showed significant activity, suggesting that it could interact with TaqI protein specifically to form both ground state and transition state complexes (Fig. 1B). From Mn 2ϩ to Zn 2ϩ , the cleavage activity generally followed a descending order.
We first determined the optimal time required for the cleavage of TaqI protein. A time course of TaqI protein cleavage was conducted with 2 mM Fe 2ϩ and 25 mM sodium ascorbate ( Fig.  2A). Primary cleavage fragments appeared at the earliest time point (30 min) and accumulated to its maximum after about 2 h of incubation. At later time points, the reaction kinetics is dominated by concurrent cleavage of both intact protein and secondary degradation of primary cleavage products.
To demonstrate the specificity of Fe 2ϩ -mediated cleavage of TaqI protein, we denatured the protein by incubating with 0.1% SDS for 20 min or heating at 95°C for 15 min (Fig. 2B). Upon denaturation with SDS, discrete cleavage bands were no longer formed, although cleavage of intact TaqI protein appeared to be significant. This result suggests that when TaqI protein is denatured and linearized by SDS treatment, the cleavage reaction becomes essentially nonspecific. In the case of thermal denaturation, TaqI protein remained largely uncleaved. This result indicates that the protein may have undergone unfolding and refolded into a random coil-like state that disorganizes the metal binding site(s) and becomes more resistant to cleavage. By and large, the data demonstrate that the cleavage pattern shown in Fig. 2A is dependent on the proper tertiary or even quarternary structure of TaqI native protein.
Peptide Sequencing of Cleavage Fragments-To obtain enough fragmented peptides for N-terminal sequencing, the cleavage reaction was scaled up by using 75 g of pure TaqI protein. As shown in Fig. 3, eight cleavage fragments were generated in the absence of DNA substrate. A few low yield fragments (F2, F4, F5, and F8) were accumulated in sufficient quantity for N-terminal sequencing. Two new minor bands (F4a and F5a) were generated when 5 nM oligonucleotide substrate containing a single TaqI recognition site was incorporated into the Fe 2ϩ cleavage mixture; these could result from conformational changes induced by the substrate. In addition to smaller molecular mass cleavage products, peptide bands with molecular masses greater than that of the intact TaqI monomer (33 kDa) were generated as well (Fig. 3). A similar observation was reported in the course of cleaving tetracycline repressor protein and was taken as an indication of cross-linking of hydroxyl radical generated cleavage peptides to the uncleaved TetR (7). An identical cleavage mechanism may have occurred with TaqI protein, which results in accumulation of high molecular mass cross-linked products after primary cleavage events.
These fragments were electroblotted onto PVDF membrane, and each individual band was excised out for N-terminal sequencing. For some faint bands, two identical fragments were loaded to a protein sequencer. Sequence data were obtained from at least five cycles, and in some cases eight cycles. Three fragments (F2, F3, and F8) corresponded to the original N terminus of TaqI protein, which had a seven-amino acid PhoA signal sequence (MKQSGGN) fused onto the gene (Table I). F4a and F5a uniquely generated from substrate-induced cleavage also corresponded to the original N-terminal sequences, from which little information about cleavage sites can be extracted. Internal cleavage sequences were obtained from minor amino acid residues in fragments F1, F4, F5, and F7 and from major amino acid residues in fragment F6 (Table I). Some fragments with single internal sequence data available (F1, F4, and F6) were readily mapped to the correct TaqI sequence. For fragments with several internal amino acid residues identified in each cycle, all the possible combinations with a match of sequence more than four amino acids were selected (Fig. 4) Ile 141 , Ile 156 2Lys 157 , and Arg 173 2Lys 174 . These cleavage sites, from Asp 47 to Lys 174 , span a 127-amino acid stretch that is highly conserved among TaqI and its three isoschizomers (TthHB8I, TfiTok6A1I, and Tsp32I R ), except that Lys 174 in some cases is replaced by arginine (28,29).
Site-directed Mutagenesis-The cleavage sites originated from Fe 2ϩ -mediated cleavage could point to regions that play an important role in coordinating metal binding and organizing the active site in TaqI protein. We selected 11 amino acid residues (Asp 48 , Trp 113 , Glu 135 , Asp 137 , Gln 139 , Gly 140 , Asp 142 , Lys 157 , Lys 158 , Arg 173 , and Lys 174 ) for site-directed mutagenesis study based on the following two guidelines. First, they are located at or nearby the cleavage sites. Second, they are likely to play a functional role in TaqI enzyme activity, for example, negatively charged amino acids (Asp and Glu) were chosen for their potential roles in coordinating metal cofactors, positively charged amino acids (Lys and Arg) were chosen for their potential roles in interacting with phosphate backbone or making specific base contacts.
Site-directed mutagenesis was performed using a PCR-based overlap extension procedure (24). Three to five mutations were isolated at each individual site by using partially degenerate internal mutagenesis primers. These mutations included both highly conservative substitutions (i.e. K158R) and more general alterations (i.e. K158S, K158C, and K158G).
A genetic system established in E. coli allows for TaqI endonuclease expression even in the absence of the corresponding methylase, which protects the endogenous Thermus host genome by methylating the TaqI recognition sites (30). Under our growth conditions (32-37°C), the thermophilic TaqI may generate a single-strand nick that could be repaired by E. coli DNA ligase. Thus, an E. coli cell hosting a TaqI endonuclease gene becomes viable at permissive temperature. We employed this system in our site-directed mutagenesis study so that mutants without enzyme activity or with partial activity would have a higher expression level, as they are less damaging to the E. coli cells. Three expression levels were observed among the fortythree mutants. Level 1, wild-type level: mutants had a similar low level of expression as the wild-type, indicating that these mutants were as active as the wt TaqI enzyme. This group includes three Gln 139 mutants, as shown in Fig. 5A, mutants at Asp 48 , Trp 113 , Glu 135 , and Arg 173 , and some mutants at Lys 174 . Level 2, partially overexpressed: mutants had significantly higher level of expression than the wild-type, indicating moderate effects on binding or catalysis. This group includes mutants at Gly 140 (Fig. 5A), Lys 157 , Arg 173 , and some mutants at Lys 174 (K174I and K174M). Level 3, completely overexpressed: mutants had a severalfold higher level of expression than the partially overexpressed group, indicating drastic effects on binding or catalysis. This group includes all the mutants at Asp 137 , Asp 142 , and Lys 158 positions. It is noteworthy that three of the mutants (D142A, D142V, and D142G) migrated slightly faster than D142E, which migrated at a position identical to that of the wt TaqI in 12.5% SDS-PAGE (data not shown). This result suggested that elimination of a negative charge at Asp 142 may be responsible for a more complete unfolding of the TaqI protein, which was found to migrate with a higher apparent molecular weight than expected in the presence of 0.1% SDS in 12.5% polyacrylamide gel (21). In general, the level of protein expression is inversely proportional to catalytic activity.
Characterization of TaqI Mutants-To characterize the binding affinity of the forty-three mutants to the TaqI recognition site (TCGA), we adopted a Ca 2ϩ gel shift binding assay procedure previously described for EcoRV (31) and NaeI (32). Ca 2ϩ , in place of the normal metal cofactor Mg 2ϩ , is able to support specific binding but not catalysis of EcoRV to its cognate site,  Minor product(s). Fig. 1A demonstrated that Ca 2ϩ is also unable to support significant catalytic activity (Ͼ10%) with TaqI. This technique is particularly useful in characterizing wild-type and mutant binding because both EcoRV and TaqI require Mg 2ϩ for binding and catalysis. To evaluate the binding property of these mutants, we used a binding condition in which wt enzyme achieved less than 100% binding to the oligonucleotide. Relative binding affinity was quantified by taking the intensity ratio of each bound mutant enzyme over bound wt enzyme. The mutants can be classified into three categories (Table II). Category 1: wild-type binding (Fig. 5B) Cleavage assay conditions were designed such that limited cleavage occurs on a plasmid substrate (pFB76) containing a single TaqI recognition site. The ratio of linearized product formed by a particular mutant over wt TaqI was used to determine relative cleavage activity. The majority of mutations (Asp 48 , Trp 113 , Glu 135 , Gln 139 , Gly 140 , Lys 157 , Arg 173 , and Lys 174 ) retained wt or reduced activity (Table II). Most mutations at Asp 137 , Glu 142 , and Lys 158 completely lost DNA cleavage activity even after incubation with 10-fold more enzymes (data not shown). The effects of individual mutations on binding and cleavage are summarized in Table II and described as follows.

GATATC (31). Our results shown in
Asp 48 -DNA binding was lowered by 20% by changing Asp to Glu, albeit D48E retains a negative charge at this position.
Elimination of the negative charge (D48A and D48G) further lowered the binding to 50%, whereas catalysis was comparable with the wt enzyme, suggesting that the influence of mutations at Asp 48 mainly takes effect on the formation of ground state ES complex.
Trp 113 and Glu 135 -Mutations at these two positions did not seem to affect binding or catalysis, although they are located at or near the Fe 2ϩ -mediated cleavage sites and are conserved among four homology groups of TaqI isoschizomers (29). These results suggest that Trp 113 and Glu 135 may not be involved in direct DNA binding and catalysis. However, some of the mutants (W113C and E135V) may disturb catalysis when Mn 2ϩ is used as the metal cofactor.
Asp 137 -Conservative change from Asp to Glu led to about 25% loss of binding affinity but 80% loss of catalytic activity (Fig. 5). The other three mutants (D137A, D137V, and D137G), which eliminated the negative charge, led to complete loss of catalysis. Thus, mutations at Asp 137 in general do not affect substrate binding, suggesting that Asp 137 is a critical amino acid residue for catalysis, possibly by binding to a catalytic metal ion in the formation of transition state complex.
Gln 139 -A conservative change (Q139N) seemed to affect neither binding nor catalysis. Failure to retain the hydrogen bond donor and acceptor property (Q139D and Q139Y) reduced binding or both binding and catalysis. Some Asn or Gln residues in restriction endonucleases and other proteins are involved in DNA recognition. Mutations at Gln 139 caused only limited reduction in substrate binding, suggesting that it may not play a direct role in DNA recognition. However, Gln 139 may be part of a hydrogen bond network in the vicinity of the active site.
Gly 140 -All mutations at Gly 140 led to loss of binding affinity, which roughly corresponded with the bulkiness of the amino acid side chain. As to DNA cleavage, whereas a methyl side chain exhibited little effect, an isopropyl or a hydroxylethyl side chain reduced cleavage by 60%. These results indicate that Gly 140 may influence both binding and catalysis by providing flexibility to the protein confirmation in the vicinity of the active site.
Asp 142 -Unlike Asp 137 , which retained binding affinity and some catalytic activity with a conservative change (D137E), both binding with Ca 2ϩ and catalysis with Mg 2ϩ are completely abolished by any replacement at this position. We considered two scenarios to explain the role of Asp 142 . First, the carboxylate of Asp 142 may be involved in forming a metal binding site to coordinate Mg 2ϩ , which is required for both ground state ES and transition state interactions. The inability to bind a Mg 2ϩ metal cofactor results in loss of binding and catalysis. An alternative explanation is that Asp 142 is a critical determinant for the TaqI three-dimensional structure, such that mutations at the Asp 142 position would drastically distort the protein conformation, resulting in a total loss of enzyme activity. However, the high cleavage rate of D142E mutant with Mn 2ϩ suggests that the protein conformational alteration is unlikely to occur globally, at least not by the Asp to Glu substitution, for a global structural change would likely impede cleavage with Mn 2ϩ as metal cofactor as well. On the other hand, small conformational changes could have occurred with D142E, which altered the metal binding site in such a way that Ca 2ϩ or Mg 2ϩ may no longer be able to coordinate, but still could accommodate Mn 2ϩ binding.
Lys 157 -Mutations at Lys 157 in general led to better than wt binding with the exception of K157G, which may cause some conformational disturbance due to the complete elimination of the side chain. Catalysis in general was diminished, particularly when the positive charge was not maintained (K157S, K157C, and K157G). In light of its vicinity to the catalytic center residue Lys 158 described below, the positively charged Lys 157 may play a supportive role in catalysis.
Lys 158 -Mutations at Lys 158 reduced both substrate binding and DNA cleavage drastically. Even the conservative change from Lys to Arg resulted in complete loss of catalysis using either Mg 2ϩ or Mn 2ϩ as a metal cofactor, suggesting that the precise configuration of Lys could not be achieved by Arg at Lys 158 . The weak binding affinity observed in K158R indicates that Arg may slightly compensate the role of Lys in ES complex formation, but it failed to interact with the DNA sequence to form an active transition state complex. Therefore, the positive charge at Lys 158 is critical for the formation of ES complex as well as transition state interactions. The weak binding and  cleavage observed in K158G are seemingly contradictory at first glance; however, the replacement of the Lys at 158 by a Gly may confer enough flexibility such that the adjacent Lys at 157 could fulfill the role of Lys 158 for both binding and catalysis to a certain extent. As a result, limited DNA cleavage was observed.
Arg 173 and Lys 174 -The positive charges at positions 173 and 174 are important for binding and catalysis. Although the role of Arg 173 in substrate binding could not be entirely replaced by Lys, Lys and Arg are interchangeable at 174 without affecting either binding and catalysis. Both Arg and Lys residues at 174 have been observed in several TaqI isoschizomers (29). Whether the positive charges interact with DNA substrate directly or are required for structural integrity remains to be investigated.

DISCUSSION
Fe 2ϩ -mediated Cleavage-This study attempts to elucidate the structural basis of TaqI endonuclease catalytic center by Fe 2ϩ -mediated oxidative cleavage. The observation that Fe 2ϩ could replace Mg 2ϩ as the metal cofactor validates the use of this technique for TaqI. Comparison of Fe 2ϩ -mediated cleavage with native and denatured protein further ensures that the cleavage is specific to the three-dimensional structure of the protein. Although the cleavage may progress in many buffer conditions, such as TES, Tris-HCl, and HEPES, phosphate buffer was found to be inhibitory to the cleavage reaction (data not shown). Hydrogen peroxide, which is used for the generation of hydroxyl radicals in some studies, was excluded from the final protocol because it diminished the accumulation of cleavage products. All of the cleavage fragments analyzed were products of a single break in the TaqI protein by hydroxyl radicals, as indicated by their amino acid sequences and molecular weights. In some cases, the amino acid sequences of both fragments from a single cleavage were identified, where the molecular weights of the two fragments add up to that of intact TaqI protein. In other cases, only one sequence was found, suggesting that some fragments may not be able to maintain their native conformation. The unfolded peptide fragments became vulnerable to nonspecific cleavage. There was only one minor fragment identified as a multiple cleavage product (data not shown).
Two additional cleavage fragments (F4a and F5a) were identified when the oligonucleotide substrate containing a single TaqI recognition site was incorporated into the cleavage reaction. These two fragments, when sequenced, corresponded to the original N terminus of TaqI. We have not found the Cterminal fragments from these substrate-dependent cleavage events. More experimentation is needed to purify these two fragments and identify their C-terminal cleavage sites by other means.
Active Site of Other Endonucleases-The active site amino acid residues of several restriction endonucleases have been identified by high resolution crystallographic analysis, as well as by genetic and biochemical approaches. On the basis of the structural and biochemical studies of EcoRI and EcoRV, a PDX 14 -20 (E/D)XK motif (where X means any amino acid residue) was proposed as the Mg 2ϩ binding site and catalytic center (Fig. 6). This motif is also found in PvuII (16,17), FokI (33), NaeI (32), and MunI (18). Although some restriction enzymes indeed contain such a sequence element, many other restriction endonucleases do not posses this motif (34). The catalytic center of BamHI endonuclease was initially identified by genetic selection (13), later confirmed by structural studies (14,15). The organization of the BamHI catalytic center can be considered as a degenerate form of the PDX 14 -20 (E/D)XK motif, where an amino acid is inserted into the PD doublet and the positively charged Lys is replaced by a negatively charged Glu in the triplet (E/D)XK (Fig. 6). In PvuII, the ELK triplet (amino acids 68 -70) was proposed as part of the catalytic center (Fig.  6). However, a PD doublet was not found within the upstream sequence (16,17). Glu 55 and Asp 58 are implicated as playing a role similar to the PD doublet in metal cofactor binding. Glu 55 is followed by a Gly, which bears some resemblance to the PDG triplet (amino acids 90 -92) in EcoRI. It was suggested that Pro before and Gly after the Asp may properly position the Asp for metal binding (17).
These examples strongly suggest that the catalytic centers of restriction endonucleases are more degenerate than originally thought. In TaqI, a complete PDX 14 -20 (E/D)XK motif does not exist. A PD doublet (amino acids 47-48) is conserved among TaqI isoschizomers (29). Fe 2ϩ -mediated cleavage also occurred right after the doublet, PD2L. Nevertheless, mutations at Asp 48 appeared to affect mainly DNA binding. These results suggest that although Asp 48 may play some roles in DNA binding or regulating ES complex formation, it does not seem to play a critical role in coordinating the metal cofactor required for catalysis.
Role of Asp 137 and Asp 142 -Based on the binding and catalysis studies of the mutants, we propose that Asp 137 , Asp 142 , and Lys 158 are the essential active site amino acid residues in TaqI endonuclease (Fig. 6). Mutations at Asp 137 generate a binding-proficient, catalysis-deficient mutant, which is equivalent to Asp 74 in EcoRV (35), Asp 91 in EcoRI (35), and Asp 74 in BamHI (13). These negatively charged amino acid residues in EcoRV, EcoRI, and BamHI are involved in coordinating a catalytic metal cofactor at the catalytic center; a similar role is expected for Asp 137 in TaqI. The failure to cleave DNA substrate by these mutants could be attributed to the loss of binding affinity to the catalytic metal cofactor, which may serve as a Lewis acid to polarize the nonbridging POO bonds in a previously proposed catalytic mechanism of EcoRI and EcoRV (36). In the other endonucleases, at least two carboxylates are identified in metal binding, one of which is a Glu or Asp in the (E/D)XK half of the catalytic motif. In TaqI, the only triplet signature found is EAK (residues 164 -166), but the Lys is substituted by Arg in TaqI isoschizomers, and K166Q did not disrupt for TaqI function (25). Additionally, there is no evidence that Fe 2ϩ -mediated cleavage has occurred around the EAK triplet. Thus, some other amino acid residue(s) of TaqI, such as Asp 142 , may be involved in coordinating metal cofactor. Mutations at this position eliminated both binding with Ca 2ϩ as cofactor and catalysis with Mg 2ϩ as cofactor. D142E in fact turned TaqI into a Mn 2ϩ -dependent endonuclease. A similar mutant was described in EcoRV when a nearby active site isoleucine was replaced by leucine (37). Asp 142 in TaqI is situated in a sequence context similar to that of Asp 58 in PvuII, both of which are the second amino acid residue downstream of a Gly. However, Asp 142 mutants bear no resemblance to Asp 90 mutants in EcoRV, the E111A mutant in EcoRI, or Glu 111 mutants in BamHI because the later still bind to their cognate sequence. More studies are needed to better define the role of Asp 142 in TaqI.
Role of Lys 157 -The effect of Lys 157 is largely confined to catalysis (Table II). One possible role the Lys 157 plays is to lower the pK a of the adjacent Lys 158 , which is catalytically critical for TaqI. A typical example of pK a shift of a Lys residue by an adjacent Lys is a bacterial acetoacetate decarboxylase, in which the pK a of the active site Lys 115 is shifted to pH 6.0 by Lys 116 (38,39).
Role of Lys 158 -According to the structural motif proposed, the catalytic center of a restriction endonuclease should have a Lys about 10 -30 amino acids downstream, which is part of the (E/D)XK triplet. Although Lys 158 in TaqI does not have the triplet feature, it is critical for TaqI function, as demonstrated by the drastic effects of mutants on binding and catalysis. Similar effects were observed with K92E in EcoRV. But the role that the positively charged lysine plays in TaqI and EcoRV are still unclear. It has been proposed that Lys 92 in EcoRV may stabilize transition state complex by neutralizing the negative charges developed during transition state interactions (36). This hypothesis gains support from the observation that K92E mutant of EcoRV can recover the DNA cleavage activity to 65% of wt level when Mg 2ϩ was substituted by Mn 2ϩ (35). The carboxylate at position 92 in the K92E mutant was assumed to bind a Mn 2ϩ ion, so that the role of Lys 92 is compensated by Glu 92 /Mn 2ϩ . In our DNA cleavage analysis, the role of Lys 158 in TaqI could not be replaced by another positively charged amino acid, Arg. It will be very interesting to see whether a K158E mutant could exhibit an effect similar to that seen in K92E of EcoRV in the presence of Mn 2ϩ . Alternatively, Lys 92 of EcoRV has been speculated to act as a general base to activate the water molecule (water 18), which is in-line for attacking the scissile phosphodiester bond (12).
DXXG Motif-Although the catalytic motif PDX 14 -20 (E/ D)XK is not found in TaqI, the catalytically important acidic residue Asp 137 is situated in a DXXG motif previously found in GTPase superfamily, including p21 ras (40,41). The invariant Asp in this highly conserved motif is involved in coordinating a catalytic Mg 2ϩ through an intervening water molecule and the amide proton of the Gly forms a hydrogen bond with the ␥-phosphate of the GTP. Asp 137 in DAQG of TaqI, as discussed above, could play a similar role as Asp 57 of p21 ras , albeit it may or may not be mediated through a localized water.
In homing endonucleases, the LAGLI-DADG motif is implicated in DNA cleavage, but the roles of each individual residue have not been established (42). It remains to be seen whether the acidic residue(s) in LAGLI-DADG motif also coordinates catalytic Mg 2ϩ directly or indirectly. Recently, some catalytic residues of BsoBI have been identified by random mutagenesis (43). Interestingly, two catalytically deficient acidic residues, Asp 124 and Asp 246 , are located in a DXXG sequence context. It appears that the DXXG motif may have evolved as a Mg 2ϩ ion coordinator for DNA hydrolysis.
In summary, this study has identified a few active site residues and revealed some unique features of the TaqI catalytic center. Although the thermophilic TaqI endonuclease shares many important catalytic features with other well studied enzymes, it may also have evolved some unique means to achieve efficient substrate turnover. The hypothetical model will evolve as more experimental data accumulate. Further structural and biochemical characterization of TaqI endonuclease will shed more light on the mechanistic understanding of endonucleases in particular and nucleases in general.