Identification of Residues of T4 RNase H Required for Catalysis and DNA Binding*

Bacteriophage T4 RNase H, which removes the RNA primers that initiate lagging strand fragments, has a 5′- to 3′-exonuclease activity on DNA·DNA and RNA·DNA duplexes and an endonuclease activity on flap or forked DNA structures (Bhagwat, M., Hobbs, L. J., and Nossal, N. J. (1997) J. Biol. Chem. 272, 28523–28530). It is a member of the RAD2 family of prokaryotic and eukaryotic replication and repair nucleases. The crystal structure of T4 RNase H, in the absence of DNA, shows two Mg2+ ions coordinated to the amino acids highly conserved in this family. It also shows a disordered region proposed to be involved in DNA binding (Mueser, T. C., Nossal, N. G., and Hyde, C. C. Cell (1996) 85, 1101–1112). To identify the amino acids essential for catalysis and DNA binding, we have constructed and characterized three kinds of T4 RNase H mutant proteins based on the possible roles of the amino acid residues: mutants of acidic residues coordinated to each of the two Mg2+ ions (Mg2+-1: D19N, D71N, D132N, and D155N; and Mg2+-2: D157N and D200N); mutants of conserved basic residues in or near the disordered region (K87A and R90A); and mutants of residues with hydroxyl side chains involved in the hydrogen bonding network (Y86F and S153A). Our studies show that Mg2+-1 and the residues surrounding it are important for catalysis and that Lys87 is necessary for DNA binding.

; and Mg 2؉ -2: D157N and D200N); mutants of conserved basic residues in or near the disordered region (K87A and R90A); and mutants of residues with hydroxyl side chains involved in the hydrogen bonding network (Y86F and S153A). Our studies show that Mg 2؉ -1 and the residues surrounding it are important for catalysis and that Lys 87 is necessary for DNA binding.
Bacteriophage T4 encodes a 5Ј-to 3Ј-nuclease with T4 RNase H activity which removes the pentamer RNA primers synthesized during lagging strand DNA replication (1). This T4-encoded nuclease is sufficient for phage DNA synthesis because wild type T4 does not require host enzymes for processing of RNA primers (2). Escherichia coli polymerase I and RNase HI can substitute to some extent if T4 RNase H is deleted, but replication is slower and less accurate in the absence of the T4 enzyme (2).
The 5Ј-to 3Ј-nuclease of T4 RNase H degrades both RNA⅐DNA and DNA⅐DNA duplexes, releasing short oligonucleotide products from the 5Ј-end. In the accompanying paper (3) we have shown that T4 RNase H continues to degrade double-stranded DNA until it reaches 8 -11 nucleotides from the 3Ј-end. Although T4 RNase H by itself is a nonprocessive exonuclease, the rate and processivity of the nuclease reaction are increased by the T4 gene 32 single-stranded DNA-binding protein. At the T4 DNA replication fork, single-stranded DNA is covered with the gene 32 protein; hence, T4 RNase H must be acting as a processive exonuclease, removing the RNA primer and some adjacent DNA. Our recent results indicate that this nuclease performs only one round of processive degradation during each lagging strand cycle, removing 10 -50 nucleotides before polymerase elongates the next Okazaki fragment creating a nick sealed by ligase. 1 T4 RNase H also has a flap endonuclease activity, which cuts preferentially near the junction of the single-and doublestranded DNA of the flap or fork structures removing the 5Ј-single-stranded tail (3). This endonuclease activity of T4 RNase H is inhibited by the gene 32 protein. The rates of the 5Јto 3Ј-nuclease and flap endonuclease reactions are similar. However, in the absence of Mg 2ϩ ions, T4 RNase H binds 100 times better to flap and fork structures than to double-stranded or nicked DNA (3).
T4 RNase H is a member of the RAD2 family of prokaryotic and eukaryotic replication and repair nucleases (Fig. 1A). Like T4 RNase H, many of these enzymes are 5Ј-nucleases that remove primers initiating lagging strand fragments. These include the bacteriophage T5 D15 and T7 gene 6 exonucleases, the 5Ј-to 3Ј-exonuclease domains of DNA polymerases from bacteria such as E. coli, Mycobacterium tuberculosis (Mtb), and Thermus aquatics (Taq), and the eukaryotic nucleases such as murine FEN 2 -1 and human RAD2 analog (also called MF-1 or FEN-1) (for review, see Ref. 4). All of these 5Ј-to 3Ј-nucleases degrade DNA⅐DNA and RNA⅐DNA duplexes to short oligonucleotide products. Many of these proteins have also been shown to possess a flap endonuclease activity (5-7). These 5Ј-nucleases have sequence similarity to the larger repair proteins such as human XPG (bottom sequence, Fig. 1A). In this figure, the highly conserved residues are marked in red, and the moderately conserved residues are marked in green. Sequence alignments for additional members of this family can be found in Ref. 4.
The crystal structure of T4 RNase H, in the absence of DNA, has been solved (4) (Fig. 1B, left panel). The proposed active site (Fig. 1B, right panel) contains 2 Mg 2ϩ ions separated by approximately 7 Å and a number of conserved amino acids. The first Mg 2ϩ ion (Mg 2ϩ -1) is coordinated directly to Asp 132 and through water molecules to Asp 19 , Asp 71 , and Asp 155 . The second Mg 2ϩ ion (Mg 2ϩ -2) is coordinated through water molecules to Asp 157 , Asp 200 , and Tyr 86 . The hydroxyl side chain of Tyr 86 as well as that of Ser 153 , which is near Asp 19 , are involved in the hydrogen bonding network surrounding the Mg 2ϩ ions. The crystal structure shows a disordered region from residues 89 to 97, containing a number of positively charged residues. It has been proposed that this region is involved in DNA binding (4).

Two different mechanisms have been proposed for E. coli
RNase H, which also acts on RNA⅐DNA duplexes: a one-metal mechanism based on that for DNase I (8,9) and a two-metal mechanism based on that for the 3Ј-to 5Ј-exonuclease of polymerase I (10). In the one-metal mechanism, an active site acidic amino acid acts as a nucleophile and abstracts a proton from a water molecule to create a hydroxyl ion that attacks the phosphodiester bond. The pentavalent species formed during the transition state is stabilized by the Mg 2ϩ ion. In the twometal mechanism, one of the bound Mg 2ϩ ions promotes the formation of the hydroxyl ion that attacks the scissile phosphodiester bond. The second Mg 2ϩ ion serves to stabilize the oxyanion leaving group and also to stabilize the pentavalent species formed during the transition state. In 3Ј-to 5Ј-exonuclease of polymerase I, the divalent metal ions are 3.9 Å apart (11). T4 RNase H does contain two Mg 2ϩ ions, but they are separated by 7 Å, which is far apart for the two-metal mechanism. However, the crystal structure was solved in the absence of DNA, and it is possible that when DNA is bound, there may be a substantial rearrangement bringing the ions closer.
What is the mechanism of the nuclease activity of T4 RNase H? To address this question, we have made three kinds of mutants of T4 RNase H based on the possible roles of the amino acid residues (Fig. 1B): mutants of the acidic residues surrounding the Mg 2ϩ ions, mutants of the residues with hydroxyl side chains that are involved in the hydrogen bonding network, and mutants of the residues that may be involved in DNA binding. In this paper, we report the effects of these mutations on the exonuclease activity, the flap endonuclease activity, and on binding to DNA. The implications of these results on the possible roles of the two Mg 2ϩ ions and the active site residues are discussed.

EXPERIMENTAL PROCEDURES
Unless otherwise indicated, the materials and methods are those described in the accompanying paper (3).
Each amplification reaction was carried out using 30 cycles of 95°C for 1 min, 50°C for 1 min, and 72°C for 2 min, followed by a final elongation phase at 72°C for 7 min. The amplified products were purified from an agarose gel using the Qiaquick kit (QIAGEN). The final amplified product was cut with PshAI and NsiI and then ligated into pNN2202 that had been cut with the same enzymes. The DNA was transformed into E. coli MV1190 (13), and the entire amplified DNA was sequenced using Sequenase 2.0 to confirm the mutation and to make sure that no other mistakes were introduced by the pfu DNA polymerase.
The D132N and S153A mutants were prepared by cutting the plasmid pNN2202 with HindIII and BsaBI or with BsaBI and NsiI, respectively, and inserting the oligonucleotides containing the respective mutation. The mutant sequences were confirmed by sequencing the DNA between the appropriate sites.
The mutants D19N, D155N, D157N, D200A, and D200N were made from pVC1109 (wild type T4 RNase H (1)) single-stranded DNA by the Kunkel method (14), modified by using the T4 44/62 and 45 polymerase accessory proteins in addition to T4 DNA polymerase to copy the singlestranded DNA template. Mutants were identified by sequencing using Sequenase 2.0.
Purification of the Mutant Proteins-All of the plasmids, except the one with D71N mutation, were transformed into E. coli BL21(DE3)-pLysS (15) for expression of the mutant T4 RNase H proteins, as described for the wild type protein (16). Because the plasmid with D71N mutation could not be transformed into E. coli BL21(DE3)pLysS, the D71N mutant protein was purified from E. coli MV1190 carrying the appropriate plasmid by infecting with M13mGP1-2, encoding T7 RNA polymerase, as described (1). All of the mutant proteins except D155N and D200A were expressed well and were soluble. There was enough soluble protein to purify the D155N but not the D200A mutant protein.
The mutant proteins were partially purified using Whatman P-11 phosphocellulose resin. All purification steps were performed at 4 -8°C. The cell paste from 100 ml of culture was suspended in 5 or 10 ml of sonication buffer (50 mM Tris-Cl, 7.5, 0.2 M KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 10% glycerol, and 0.1 mM AEBSF). The cell extract was ultracentrifuged at 100,000 ϫ g for 2 h. The supernatant was then diluted with 2 volumes of PC buffer (50 mM Tris-Cl, 7.5, 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, and 0.1 mM AEBSF). Phosphocellulose resin (2 ml, 1:1 suspension), equilibrated in PC buffer, was added, and the mixture was rotated slowly for 1 h to allow binding of T4 RNase H to the resin. The resin was collected by centrifugation, and loosely bound proteins were removed by washing the resin with PC buffer containing 50 mM KCl. The resin was then transferred to a small column and washed successively with PC buffer (1 ml) containing 0.1 M KCl, 0.3 M KCl, and 0.5 M KCl. T4 RNase H was eluted in the 0.3 M KCl and 0.5 M KCl fractions. The purity of the proteins was checked using a 12% SDS-polyacrylamide gel (Novex), stained with Coomassie Blue. The concentrations of the proteins for the assay were estimated by comparison of the band intensities of mutant proteins with those of the known concentrations of highly purified wild type T4 RNase H protein, visually or using the dv1 program and a Desktop Plus scanner from PDI, Inc.
Nuclease Activity-The 5Ј-to 3Ј-nuclease activity was assayed at 30°C using a DNA 34-mer annealed to M13 single-stranded DNA, and the flap endonuclease activity was assayed using a synthetic oligonucleotide flap DNA structure, as described in the accompanying paper (3). To study the effect of temperature on the activity, degradation of the 34-mer⅐M13 was measured at room temperature and at 30, 35, 40, and 45°C for 1 min. To study the stability of the wild type and the mutant proteins, the proteins were diluted as described (3) and incubated at 45°C for 0, 1, 5, 10, and 20 min. The nuclease reaction mixture containing the 34-mer⅐M13 substrate was preincubated at 45°C for 2 min. The reaction of the preincubated protein with the preincubated substrate was performed at 45°C for 1 min. The products were separated on 20% polyacrylamide and 7 M urea (19:1) gels and exposed to XAR or BMR film. The films were scanned and the products quantitated using the dv1 program from PDI, Inc.
DNA Binding-Binding of the flap DNA structure by the wild type and mutant proteins was measured with a gel mobility shift assay, as described (3).

RESULTS
To determine the amino acids important for the cleavage of the phosphodiester bond by T4 RNase H, we selected residues for mutation based on the crystal structure of T4 RNase H (4) and its sequence similarity to other prokaryotic and eukaryotic nucleases (Fig. 1A). As indicated in the Introduction, the proposed active site contains two Mg 2ϩ ions surrounded by a number of conserved acidic amino acids (Mg 2ϩ -1: Asp 19 , Asp 71 , Asp 132 , and Asp 155 ; Mg 2ϩ -2: Asp 157 and Asp 200 ). We mutated each one of these aspartates individually to aspargine and in addition, Asp 200 to alanine. We also mutated two residues with hydroxyl side chains, Tyr 86 and Ser 153 , to phenylalanine and alanine, respectively. In addition, we made mutants of residues close to (K87A) or within (R90A) the disordered region to study the possible roles of these amino acids in binding to DNA.
Purification of Mutant T4 RNase H-We partially purified all of the mutant proteins and characterized the effects of each mutation on the nuclease activities of T4 RNase H and on its binding to DNA. The mutants were constructed by site-directed mutagenesis and the proteins expressed, as described under "Experimental Procedures." They were partially purified by chromatography on phosphocellulose. The 0.5 M KCl step elution fraction, which contained fewer contaminating proteins than the 0.3 M KCl fraction, was used for the nuclease and DNA binding assays. Fig. 2 is an SDS-polyacrylamide gel of fractions of the Y86F and K87A mutant proteins. Similar purification was achieved with all other mutant proteins (data not shown). The activities of the partially purified wild type and mutant T4 RNase H were compared with corresponding fractions from cells with the vector plasmid and with highly purified wild type protein. The partially purified wild type enzyme showed activity similar to that of homogeneous T4 RNase H, except for some contaminating 3Ј-to 5Ј-exonuclease activity (see Fig. 4). Exonuclease Activity of the Mutant Proteins-T4 RNase H cuts double-stranded DNA from the 5Ј-end and releases short oligonucleotides until it reaches 8 -11 nucleotides from the 3Ј-end (3). Fig. 3 shows the products of the reactions of the wild type and mutant T4 RNase H proteins with the 5Ј-end-labeled 34-mer annealed to M13 circular DNA. This assay shows the size of the short oligonucleotide products released at the first cut from the 5Ј-end. The mutants D19N, D155N, D71N, K87A, and D132N lost their nuclease activity; D200N and Y86F had activity equal to or close to that of the wild type; and D157N, R90A, and S153A had reduced activity. Because all of these mutants (except K87A) retained flap DNA binding activity similar to that of the wild type enzyme (see Fig. 5), it is likely that they are properly folded.
Although the D200N mutant protein maintained the exonuclease activity, the size distribution of its products differed from that of the wild type enzyme. Wild type T4 RNase H released dinucleotides and trinucleotides as the major products in about equal amounts at 30°C, whereas the D157N and D200N mutant proteins released more dinucleotides than trinucleotides under similar reaction conditions (Fig. 3).
Flap Endonuclease Activity of the Mutant Proteins- Fig. 4 shows the products of the reactions of the wild type and the mutants of T4 RNase H with the flap substrate. The sequences of the oligonucleotides used to construct this flap substrate are described in the accompanying paper (3). Highly purified wild type T4 RNase H cuts the flap substrate near the junction of the single-and double-stranded DNA, releasing the 19-mer and 21-mer oligonucleotides as the major products (3). Because of contaminating 3Ј-to 5Ј-exonuclease activity in the partially purified proteins, the 19-mer and 21-mer oligonucleotide products were partially degraded at higher protein concentrations and at longer time periods. The effects of most of the mutations on the flap endonuclease activity were similar to those on the exonuclease activity. The mutants D19N, D155N, K87A, D132N, and D71N lost their flap endonuclease activity; D200N and Y86F had activity close to or equal to that of the wild type; and D157N, R90A, and S153A have reduced activity (data for S153A not shown). However, relative to the wild type protein, D157N mutant protein retained more flap endonuclease than exonuclease activity.
Binding of the Mutant Proteins to the Flap Substrate-We used a gel mobility shift assay to study the effects of the mutations on binding to the flap substrate. Because T4 RNase H requires Mg 2ϩ for its activity, the binding studies were performed in the absence of Mg 2ϩ to prevent degradation of the DNA substrate. Fig. 5 shows the mobility shift caused by the binding of wild type, Y86F, R90A, and K87A to the flap substrate. The binding of K87A mutant protein was reduced by approximately 50% compared with that of wild type, and the binding of R90A was reduced slightly. All other mutant proteins had DNA binding affinity similar to that of wild type T4 RNase H (data for remaining mutant proteins not shown).
Effect of Temperature on the Exonuclease Activities of the Mutant Proteins-The mutants of the residues coordinated directly or through water molecules to Mg 2ϩ -1 completely lost their nuclease activities (D19N, D71N, D132N, and D155N) but still bound to flap DNA. Hence, Mg 2ϩ -1 has a catalytic role. In contrast, the mutants of the residues surrounding Mg 2ϩ -2 either maintained their activity (D200N and Y86F) or had reduced activity (D157N). To determine whether mutations in residues surrounding Mg 2ϩ -2 changed the stability of the enzyme, we first compared the exonuclease activity of the active mutant protein D200N with that of the wild type at various temperatures (Fig. 6). The activity of the pure wild type enzyme increased 10 times, and that of D200N increased about 2 times as the reaction temperature was increased from room temperature to 45°C. Partially purified wild type enzyme showed a similar increase in activity (data not shown). For the wild type and the D200N mutant proteins, the ratio of the trinucleotides to dinucleotides released increased with increasing temperature.
Further, we incubated both the wild type and the mutant proteins at 45°C in the absence of the DNA substrate for various time periods, followed by a 1-min reaction at 45°C. The results indicate that the D200N and D157N mutant proteins were not more thermolabile than the pure wild type T4 RNase H protein (Fig. 7) or the partially purified wild type T4 RNase H protein (data not shown).
Effects of Mg 2ϩ and Mn 2ϩ Ion Concentrations on Exonuclease Activity-To determine whether Mg 2ϩ -2 is loosely bound in the mutant proteins, we studied the effect of altering the Mg 2ϩ ion concentration on the activities of the D157N and D200N mu-

FIG. 3. Exonuclease activity of the wild type and mutant T4 RNase H.
The substrate (5Ј-end-labeled 34-mer DNA annealed to M13, 1 nM) was treated with wild type or mutant T4 RNase H proteins at 30°C for the indicated time periods. The approximate concentrations of the proteins, represented by large and small rectangles, respectively, were 35 and 7 nM in the left and the middle panels. In the right panel, the large and small rectangles represent protein concentrations of approximately 35 and 3.5 nM, respectively. The products were analyzed on 20% polyacrylamide, 7 M urea gels. tant proteins. Similar to the pure wild type, D200N mutant protein showed maximal activity in 10 mM magnesium acetate (Fig. 8). The partially purified wild type T4 RNase H also showed maximal activity in 10 mM magnesium acetate (data not shown). The low activity of D157N did not increase with the elevated magnesium acetate concentration. However, the D157N mutant had higher activity when magnesium chloride was replaced by manganese chloride (Table I). In addition, the ratio of trinucleotide to dinucleotide products released by both D157N and D200N increased when magnesium chloride was replaced by manganese chloride (Table I).    6. The size distribution of the oligonucleotide products is different for wild type and D200N mutant T4 RNase H. The substrate, 5Ј-end-labeled 34-mer DNA annealed to M13 (1 nM) was reacted with pure wild type T4 RNase H and partially purified D200N mutant protein at various temperatures for 1 min. The products were analyzed on 20% polyacrylamide, 7 M urea gels. The ratio of trinucleotide to dinucleotide products and the ratio of total products (trinucleotides plus dinucleotides) at the indicated temperature to that at 25°C are listed below each lane. Similar results were obtained for the partially purified wild type protein (not shown).

DISCUSSION
Bacteriophage T4 RNase H is a 5Ј-nuclease that is required to remove the RNA primers from lagging strand fragments during DNA replication and has significant amino acid sequence similarity to other prokaryotic and eukaryotic nucleases with the same function (for review, see the Introduction and Fig. 1A). In the recent crystal structure of the enzyme, solved in the absence of a substrate, the highly conserved acidic residues are clustered together surrounding two Mg 2ϩ in a cleft that appears to be wide enough for single-but not doublestranded DNA (4) (see Fig. 1B). At the top of the cleft there is a short basic disordered region that has been proposed to play a role in binding the substrate. As a step toward understanding the mechanism of this important enzyme, we have constructed and characterized T4 RNase H mutants altered in active site residues that are coordinated to the two Mg 2ϩ , are in the disordered region, or possess hydroxyl side chains.
Residues Required for Catalysis-Our studies on the mutagenesis of the active site residues of T4 RNase H show that mutation of any of the residues coordinated directly or through water molecules to Mg 2ϩ -1 (D19N, D71N, D155N, or D132N) causes the complete loss of both the exonuclease and flap endonuclease activities but does not affect binding to DNA. These results indicate that residues Asp 19 , Asp 155 , Asp 132 , Asp 71 , and Mg 2ϩ -1 have a role in catalysis. At present it is not clear whether any of these aspartates has a specific function beyond being an essential part of the binding site for Mg 2ϩ . A divalent cation may increase the rate of the hydrolysis of a phosphodiester bond in a number of ways. It may neutralize the substrate, the pentavalent transition state, or the product (17), or it may be involved in the formation of the hydroxyl ion that attacks the scissile phosphodiester bond. The inactive mutant proteins of T4 RNase H are currently being used for co-crystallization with DNA to define the DNA binding site and the actual role of Mg 2ϩ -1 and the aspartates coordinated to it. 3 Our finding that T4 RNase H proteins with mutations in the residues coordinated through water to Mg 2ϩ -2 either retain their activity (D200N and Y86F) or have reduced activity (D157N) (Figs. 3 and 4) suggests either that Mg 2ϩ -2 can remain bound in the absence of any one of these residues or that this metal ion is not required for catalysis. We were unable to test the possibility that the carbonyl oxygen of D200N is important for activity because the D200A protein was expressed poorly (not shown).
Mg 2ϩ -2 could play a less direct role by stabilizing the protein or by forming a part of the binding site for the substrate, as initially proposed by Mueser et al. (4). However, our studies indicate that the D157N and D200N mutant proteins are not less thermostable than the wild type (Fig. 7). In addition, Mg 2ϩ -2 is not loosely bound in the D157N mutant protein, as increasing the Mg 2ϩ ion concentration did not increase the activity of D157N mutant protein markedly (Fig. 8). However, the exonuclease activity of the D157N mutant is increased when magnesium chloride is replaced by manganese chloride. A similar increase in activity when Mn 2ϩ replaced Mg 2ϩ has been reported for the active site mutants of EcoRV (18) and BamHI (19) restriction endonucleases. For both the D200N and D157N mutant proteins, the ratio of trinucleotide to dinucleotide products released is lower than that of the wild type protein in Mg 2ϩ (Fig. 6 and Table I) and increases in Mn 2ϩ (Table I). It is possible that there is a distortion in the metal or DNA binding site on mutating Asp 157 or Asp 200 to asparagine, which is partially corrected by the Mn 2ϩ substitution. Thus, at this time the role of Mg 2ϩ -2 is still ambiguous, but it should be clarified when a structure of the enzyme with its substrate is available. As already noted in the Introduction, in the absence of substrate, the two Mg 2ϩ in the T4 RNase H cleft are more widely separated (7 Å) (4) than those in the 3Ј-to 5Ј-exonuclease domain of E. coli polymerase I (3.9 Å), for which a twometal hydrolytic mechanism has been proposed (11).  7. The D157N and D200N  Our studies suggest that the tyrosine closest to Mg 2ϩ -2 (Tyr 86 ) and the serine closest to Mg 2ϩ -1 (Ser 153 ) in the T4 RNase H cleft are not essential for the stabilization of the transition state or the product during the exonuclease reaction. The mutant Y86F maintains its activity, whereas that of the S153A protein is moderately reduced (Figs. 3 and 4).
Substrate Binding-Because T4 RNase H cuts RNA⅐DNA and DNA⅐DNA duplexes and fork and flap structures at similar rates, it must have a substrate binding site that can accommodate all of these substrates. Our mutagenesis studies show that these substrates must be positioned on the enzyme with the scissile phosphodiester bond (phosphodiester bond between the first few nucleotides from the 5Ј-end for RNA⅐DNA and DNA⅐DNA duplexes and the phosphodiester bonds near the junction between the single-stranded DNA and the duplex in the flap and fork substrates) near Mg 2ϩ -1. Because there was no DNA in the crystal structure of T4 RNase H (4) or of the T5 D15 exonuclease (7) or the 5Ј-to 3Ј-nuclease domain of Taq DNA polymerase I (20), the position and the orientation of the substrates on this family of enzymes are still speculative.
There is a short sequence between the conserved N and I regions (Fig. 1A) that contains several basic residues in most members of this RAD2 nuclease family, particularly the short nucleases that are not attached to a polymerase. The exceptions are the larger repair nucleases, like human XPG, in which the N and I regions are separated by more than 600 residues (Fig. 1A). In the T4 RNase H structure, the region between N and I is located above the cleft with the two Mg 2ϩ ions, and nine of its residues (89 -97) are disordered (Fig. 1B). Mueser et al. (4) speculated that this disordered region might be involved in substrate binding. The crystal structure of the 5Ј-to 3Ј-nuclease domain of Taq polymerase also shows a disordered region, a small part of which overlaps the region disordered in the T4 enzyme (20). In the T5 D15 nuclease structure, there is a helical arch large enough to accommodate single-stranded but not duplex DNA at the back of the concave surface containing the two Mg 2ϩ ions (7). Part of this arch is composed of residues corresponding to all of those disordered in T4 RNase H and some of those disordered in the Taq nuclease. Ceska et al. (7) proposed that the single-stranded region of the flap threaded through this arch and that the duplex was bound on the enzyme surface outside the concave surface. A similar orientation for the duplex part of the flap substrate might be possible for the T4 enzyme, with the single strand threading under the disordered region. An alternate possibility is that the cleft of T4 RNase H, or the concave surface of T5 exonuclease, which in the absence of substrate appears to be too narrow to accommodate an RNA⅐DNA or DNA⅐DNA duplex, might open to bind a duplex in an orientation such that the scissile bond is above Mg 2ϩ -1. Even with the duplex in this position, the single strand of the flap or the released oligonucleotide products might move under the arch or disordered region.
The notion that the single strand threads through the nucleases of this family is supported by evidence that the flap endonuclease of the calf thymus FEN-1 protein is inhibited by either binding a protein on the single strand or annealing a complementary strand (21) and that the T4 gene 32 single-stranded DNA-binding protein inhibits the flap endonuclease of T4 RNase H (3). Although simple duplex and flap substrates are hydrolyzed at similar rates by the T4 enzyme, only the flap and fork substrates bind tightly enough to be retarded in mobility shift assays (3). We mutagenized the two most conserved basic residues in and adjoining the disordered region (Lys 87 and Arg 90 ), with the expectation that these changes might have more effect on the binding and degradation of flap than duplex substrates. Although the K87A mutant protein did show re-duced binding to flap DNA, about 50% of the wild type, it had also lost the ability to cut both flap and duplex substrates. The nuclease activity of the R90A protein was reduced to a similar extent on flap and duplex substrates, but its binding to flap DNA was only slightly impaired relative to the wild type protein.
On flap and forked substrates, wild type T4 RNase H and the active mutant proteins cut on each side of the junction between the single-and double-stranded DNA, whereas on duplex DNA they removed a mixture of short oligonucleotides (1-5 nucleotides) from the 5Ј-end. The average size of the products from duplex substrates was smaller with the D157N and D200N mutant proteins and increased with increasing reaction temperature with both the mutant and wild type enzymes. These observations are consistent with a need for the 5Ј-end to fray to reach the catalytic site, in a fashion similar to the fraying of the 3Ј-end to reach the active site of the proofreading 3Ј-to 5Јnuclease domains of polymerases (22,23). However this fraying does not correlate simply with the melting temperature of the duplex. There was a similar distribution of product sizes with increasing temperature on the 34-mer annealed to M13 DNA (Fig. 6) and on (dT) 36 ⅐poly(dA) (not shown).
Comparison with Related Nucleases-Mutagenesis studies have been performed for three proteins homologous to T4 RNase H: the 5Ј-nuclease domains of E. coli (24) and Mtb (25) polymerase I and human FEN-1 (26). The mutants of amino acids corresponding to T4 RNase H Asp 19 , Asp 71 , Asp 132 , and Asp 155 were inactive in E. coli polymerase I (D13N, D63A, D115A, and D138N) and in Mtb polymerase I (D21N, D73N,  D125N, and D148N). For human FEN-1, the mutants of the residues corresponding to Asp 19 and Asp 71 (D34A and D86A) lost activity, but that analogous to Asp 155 (D179A) was active. (The mutant corresponding to D132N was not reported.) Thus, with the exception of the mutant of human FEN-1 corresponding to D155N, the mutants of the residues surrounding the Mg 2ϩ -1 were inactive in all of these homologous proteins.
T4 RNase H differs from the other proteins in the importance of residues surrounding Mg 2ϩ -2. In each of the three other homologous proteins where mutagenesis studies have been reported, the mutants corresponding to D157N and D200N lost activity. In contrast, for T4 RNase H, D157N had activity, albeit reduced, and D200N had activity comparable to the wild type. However, as indicated above, the size distribution of the products released by these T4 mutant proteins was different from that of the wild type. One possible reason for the activity of T4 D200N is that Asp 200 may not have an important role in the active site of T4 RNase H because it is neutralized by the adjacent Lys 199 . None of these other homologous proteins has lysine or arginine in the corresponding position. Finally, although the T4 mutant K87A was inactive and had reduced binding to DNA, the analogous mutant in human FEN-1 (R103A) was active.
The comparison of the mutant activities of T4 RNase H and these three homologous proteins indicates that the role of Mg 2ϩ -1 is catalytic in all of these proteins, but the role of Mg 2ϩ -2 varies. Differences in the structure surrounding Mg 2ϩ -2 may contribute to the different biochemical characteristics of the 5Ј-to 3Ј-exonuclease and flap endonuclease activities of these proteins. E. coli polymerase I releases more monomers than oligomers as the major product of the exonuclease activity (27), whereas T4 RNase H releases dimers and trimers (3). In addition, E. coli polymerase I makes only one cut on flap structures, between the first two paired nucleotides (24), whereas yeast, mouse, and calf thymus FEN-1, and T4 RNase H make two cuts, one on each side of the junction between single-and double-stranded DNA (3, 21, 28). Our analysis suggests that despite their sequence homology, these four proteins (T4 RNase H, Mtb nuclease, E. coli polymerase I, and human FEN-1) have subtle differences in their active site structures. Two mutants of the more distantly related eukaryotic nuclease, human XPG protein (Fig. 1A), have recently been reported (29). In contrast to T4 RNase H and related 5Ј-exonucleases and flap endonucleases, XPG protein, which is required for nucleotide excision repair, has 3Ј-junction cutting activity on bubble substrates. The alanine mutants of Asp 77 and Asp 812 (corresponding to Asp 71 and Asp 157 of T4 RNase H) had, respectively, reduced and no residual 3Ј-junction cutting activity.
In summary, our mutagenesis studies of T4 RNase H reported here identify the importance of Lys 87 in binding to DNA, and Mg 2ϩ -1 and its coordinating residues (Asp 19 , Asp71, Asp 155 , and Asp 132 ) in catalyzing the exonuclease and flap endonuclease activities of T4 RNase H.