Crystal Structure of Bacteriophage T4 5′ Nuclease in Complex with a Branched DNA Reveals How Flap Endonuclease-1 Family Nucleases Bind Their Substrates*

Bacteriophage T4 RNase H, a flap endonuclease-1 family nuclease, removes RNA primers from lagging strand fragments. It has both 5′ nuclease and flap endonuclease activities. Our previous structure of native T4 RNase H (PDB code 1TFR) revealed an active site composed of highly conserved Asp residues and two bound hydrated magnesium ions. Here, we report the crystal structure of T4 RNase H in complex with a fork DNA substrate bound in its active site. This is the first structure of a flap endonuclease-1 family protein with its complete branched substrate. The fork duplex interacts with an extended loop of the helix-hairpin-helix motif class 2. The 5′ arm crosses over the active site, extending below the bridge (helical arch) region. Cleavage assays of this DNA substrate identify a primary cut site 7-bases in from the 5′ arm. The scissile phosphate, the first bond in the duplex DNA adjacent to the 5′ arm, lies above a magnesium binding site. The less ordered 3′ arm reaches toward the C and N termini of the enzyme, which are binding sites for T4 32 protein and T4 45 clamp, respectively. In the crystal structure, the scissile bond is located within the double-stranded DNA, between the first two duplex nucleotides next to the 5′ arm, and lies above a magnesium binding site. This complex provides important insight into substrate recognition and specificity of the flap endonuclease-1 enzymes.

The flap endonuclease-1 (FEN-1) 3 nuclease family is conserved in sequence and structure from bacteriophage to humans. These nucleases play essential roles in DNA replication by removing the RNA primers from lagging strand fragments. In addition, FEN-1-related nucleases are important in long-patch base excision repair and in maintenance of genomic stability. Homozygous knockouts of FEN-1 in mice are lethal to embryos (for review, see Refs. [1][2][3]. Like other family members, bacteriophage T4-encoded RNase H shows 5Ј to 3Ј exonuclease activity on either RNA/ DNA or DNA/DNA duplexes and endonuclease activity on either flap or fork DNA structures (4,5). T4 rnh deletion mutants give no phage production and accumulate unligated, lagging strand fragments in Escherichia coli hosts with defective polymerase I 5Ј nuclease (6). In addition, the mutants are hypersensitive to UV irradiation and anti-tumor agents (7).
T4 RNase H 5Ј nuclease removes a short oligonucleotide (1-4 bases) each time it binds its substrate. T4 32 protein, binding on single-stranded DNA behind the nuclease, increases its processivity so that about 10 -50 short oligonucleotides are removed in a single binding event (5). On nicked substrates, T4 RNase H is stimulated by the T4 replication clamp but not by 32 protein (9). The interaction of T4 RNase H with 32 protein is largely responsible for the maturation of lagging strand fragments in the T4 replication system in vitro and T4 phage production in vivo (9). The primers are normally removed, whereas the single-stranded DNA covered by 32 protein is still between the nuclease and the polymerase that extends the adjacent fragment. The clamp likely serves a backup function, allowing the nuclease to be loaded on nicked molecules that are formed when polymerase completes synthesis of the next fragment but before RNase H removes the primer. The T4 mechanism is different from current models for the maturation of lagging strand fragments in eukaryotes. In the latter, strand displacement synthesis by polymerase ␦ creates flaps that are removed mainly by the FEN-1 flap endonuclease, a reaction stimulated by the proliferating cell nuclear antigen clamp (17)(18)(19)(20).
Our structure of native T4 RNase H (PDB code 1TFR) revealed an active site composed of six highly conserved Asp residues plus two hydrated Mg 2ϩ ions (21). This structure shows strong similarity to those in other members of the FEN-1 family (22)(23)(24)(25)(26)(27)(28). In the present report, we describe the 3.0-Å x-ray analysis of an active site mutant of T4 RNase H co-crystallized with a fork DNA ligand. This is the first structure of a FEN-1 enzyme containing a branched DNA substrate in its active site. No major conformational changes were detected in the protein structure upon binding of the DNA molecule. The crystal structure of the enzyme-substrate complex clearly delineates the arrangement of T4 RNase H, T4 32 protein, and the T4 45 clamp on the lagging strand and provides insight into the range of specificities displayed by individual members of the FEN-1 family of proteins.
Preparation of Oligonucleotides-Oligonucleotides containing 18 and 24 bases to act as fork DNA substrate were designed according to previously reported sequences (30). High performance liquid chromatography-purified oligonucleotides (IDT, Coralville, IA) were dissolved in buffer (10 mM Tris, pH 8.0, 1 mM EDTA, 100 mM NaCl) to a final concentration of 5 mM. Annealed fork substrates were prepared by first mixing 1:1 molar ratios of two partially complementary DNA strands in 0.5-ml Eppendorf tubes. These tubes were placed in hot water (95°C), slow-cooled to room temperature overnight, moved to 4°C, and then stored at Ϫ20°C. Samples of fork DNA substrates used in the crystallization experiments were a 12-base complementary duplex, a 6-base 5Ј arm, and a 12-base 3Ј arm (Scheme 1).
Mobility Shift Assays-Oligonucleotides were 5Ј end-labeled with [␥-32 P]ATP, and electrophoretic mobility shift assays were carried out as described by Jones et al. (31). Reaction mixtures were prepared containing 25 mM Tris acetate, pH 7.5, 60 mM potassium acetate, 20 mM dithiothreitol, 1 mM EDTA, and 200 g/ml bovine serum albumin. Aliquots of 5 l of the reaction mixture were made containing 2 nM 5Ј 32 P linear and fork DNAs, as shown in Fig. 1A. The aliquots were incubated at 30°C for 1 min, and enzyme D132N T4 RNase H was added to give concentrations of 0, 10, 30, or 90 nM. The solutions were kept at 30°C for an additional 5 min. The protein-DNA complexes were separated on 6% DNA retardation gels (Invitrogen).
Nuclease Assays-For nuclease assays (described in Gangisetty et al. (9)), the reaction mixtures of 5 l contained 2 nM 5Ј 32 P DNAs and varying concentrations of wild type T4 RNase H, as indicated in Fig. 1, B and C. Reaction mixtures were incubated at 30°C for 1 min before the addition of wild type RNase H for 30 s. Products were separated on a 15% polyacrylamide (19:1), 7 M urea denaturing gel. Gels were exposed to Kodak Biomax MR film. DNA sizing ladders were produced by digesting the 32 P-labeled 18-mer in the RNase H nuclease reaction buffer with snake venom phosphodiesterase (Worthington Biochemical Corp.) for 2 min at 30°C or DNA polymerase I, large Klenow fragment (New England Biolabs) for 10 min at the same temperature.
Crystallization of D132N T4 RNase H with Fork DNA Substrate-Purified D132N RNase H was dialyzed at 4°C against 25 mM Bis-Tris, pH 6.5, 2 mM EDTA, 150 mM NH 4 Cl, concentrated to 20 mg/ml (Millipore Amicon Ultra, 10,000 molecular weight cutoff), and mixed in an equimolar ratio with fork DNA substrate (10 mg/ml protein). Sparse matrix screening at 23°C (Emerald BioStructures, Inc.) yielded crystals that were optimized by hanging drop vapor diffusion of droplets composed initially of 2 l of protein, 4 l of water, and 2 l of the well solution. The latter consisted of 0.1 M Tris-HCl, pH 7.5, 18% polyethylene glycol 2000 monomethyl ether (MME), and 20% glucose.
Data Collection and Processing-Crystals for data collection were flash-cooled in liquid nitrogen, and diffraction data were collected at 100 K on a Rigaku FR-E High Brilliance x-ray generator (copper anode) interfaced with a Saturn 92 CCD detector (The Ohio Macromolecular Crystallography Consortium, The University of Toledo, Toledo, OH). Analysis of the x-ray data indicated that the enzyme-substrate complex crystallized in the orthorhombic space group P2 1 2 1 2 1 , with cell dimensions a ϭ 68.16 Å, b ϭ 84.99 Å, and c ϭ 89.29 Å (Table 1). Data, which extended to 3.0 Å resolution, were processed with CrystalClear (Version 1.3.5, Rigaku MSC). Crystal statistics are presented in Table 1. Attempts to collect higher resolution data were unsuccessful. The Matthews coefficient of 2.7 and the solvent content of 53.4% indicated that the crystallographic asymmetric unit SCHEME 1 consisted of one D132N RNase H molecule and one bound 18/24 fork DNA ligand (32).
Structure Determination and Refinement-The structure was solved by molecular replacement with the program AMoRe (33) using native T4 RNase H (PDB code 1TFR) as the starting model (21). Nucleic Acid Builder was used to generate the DNA coordinate file with the appropriate sequence. A 6-bp DNA duplex was first manually positioned in the initial electron density map using the program O (34). Several rounds of model building and refinements allowed us to fit the nucleotides of the fork DNA substrate, typically adding a few nucleotides after each refinement. Refinements were carried out with CNS (35) followed by Refmac5 (36). The / angles in the Ramachandran plot of the final protein model show the following distribution (using PROCHECK (37)): 81.5% most favored, 16.3% allowed, 1.8% generously allowed, and 0.4% disallowed (1 residue, Lys-128, see "Results"). The complete 12-bp duplex, five bases of the 5Ј arm (PDB chain C) and five bases of the 3Ј arm (PDB chain D) were visible in the final model. Disorder prevented modeling of the remaining bases. Molecular figures of the enzyme and its DNA ligand were displayed and rendered with the program PyMOL. The enzyme-substrate complex was superimposed on the native T4 RNase H and on the Afu FEN-1 structures (PDB  code 1RXW) (23) using the secondary-structure matching superposition program inside COOT (38). Atomic coordinates and structure factors have been deposited. Fig. 1A illustrates the analyses of linear and fork DNA substrates by electrophoretic mobility shift assay to identify the smallest DNA fragment that could form a tight complex with T4 RNase H. Successful candidates contained a 12-base duplex with varying lengths of non-complementary 5Ј and 3Ј arms. Gel shift assays indicated that a 6-base 5Ј arm and a 12-base 3Ј arm (the 18/24 fork substrate) are required for formation of stable binary complexes. Binding is not improved by the introduction of a 3Ј duplex arm (data not shown) or the lengthening of the 3Ј arm to 18 bases or an extension of the 5Ј arm to 12 bases. Binding is abrogated in the absence of a 5Ј arm.

T4 RNase H DNA Complexes and Nuclease Activity-
Hydrolysis of the 18/24 fork DNA primarily released a 7-mer single-stranded DNA fragment from the 5Ј arm, with the cut site located one base into the duplex region near the branch point ( Fig. 1B). Lower amounts of the 5-and 8-mers and traces of the 6-mer were also found. In contrast to the FEN-1 enzymes, which exhibit a preference for a double-flap DNA substrate with a short 5Ј arm and a single 3Ј overhanging base at the branch point (39), T4 RNase H normally shows preference for single-flap and pseudo-Y (fork) substrates (5). Although catalytic activity seems opposite to what is usually observed for FEN-1 enzymes, the cleaving specificity of T4 RNase H on these three substrates is the same as that displayed by FEN-1 enzymes (30,40). In our assays (Fig. 1C), T4 RNase H released mostly 19and 21-base nucleotides from the single flap substrates, cleaving the flap strand on either side of the branch point but not at the branch point (20-base nucleotide product). Activity on double-flap substrate released 21 bases (one base into the duplex region), creating a nicked product identical to those generated by FEN-1 enzymes (Fig. 1C). However, the amount of hydrolysis was much lower for this substrate compared with the two other substrates (fork and single flap).
Structure of T4 RNase H Bound to a Fork DNA Substrate-The two hydrated magnesium ions of the native enzyme, Mg1 and Mg2, are coordinated through water to the cluster of conserved acidic residues, with Mg1 having inner-sphere coordination to Asp-132 (21). Any Asp to Asn mutation of residues coordinated to Mg1 (Asp-19, Asp-71, Asp-132, and Asp-155) eliminates nuclease activity. In contrast, similar substitutions of residues coordinated to Mg2 (Glu-130, Asp-157, and Asp-200) only reduce the nuclease activity of T4 RNase H (29).
Early attempts to co-crystallize chelated native enzyme with DNA substrates failed to produce any crystals. However, using the D132N mutation to eliminate catalytic activity, crystals were obtained of the enzyme in complex with 18/24 branched DNA substrate. Data collection and refinement statistics are shown in Table 1. These statistics are comparable with other crystal structures of protein-DNA complexes that have been solved at similar resolution. Initial phases were obtained using molecular replacement of the protein model alone. A stereo image of the initial F o Ϫ F c electron density map is presented in Fig. 2A showing the density in the region of the DNA using phases obtained from the protein alone. Initially, a doublestranded 6-mer was manually positioned in this electron density with additional nucleotides added as phases improved. D132N T4 RNase H consists of residues 1-305, of which residues 91-96 are missing in the bridge region of the refined structure. The 18/24 fork DNA is modeled as a 17/17 fork DNA, with a double-stranded 12-mer and two 5-mer single-stranded arms. The final 2F o Ϫ F c electron density map is presented in Fig. 2B. The first base of the 5Ј arm and the last seven bases of the 3Ј arm are disordered. The entire 12-mer duplex belongs to the B-form of DNA with all Watson-Crick base pairs intact, including the base pair at the branch point. The 5Ј arm traverses the active site cleft and continues in a direction perpendicular to the 12-mer duplex helical axis (Fig. 2C), whereas the 3Ј arm follows a path parallel to the duplex region toward the C terminus of the protein (Fig. 2D and supplemental movie 2D associated with Fig. 2D).
In its native form, T4 RNase H can be divided into a large ␣/␤ subdomain and a small ␣-helical subdomain. A short disordered bridge region extends above the cleft that contains the active site and two bound Mg 2ϩ ions (21). The new model of the T4 RNase H in a complex with fork DNA provides critical information for understanding the structure-specific DNA recognition by the FEN-1 family of enzymes (Fig. 2, C and E). In Fig. 2F, the amino acid sequence of T4 RNase H is aligned with the secondary structure to highlight the interactions with the DNA substrate. These interactions are examined in more detail in Fig. 3A.
A helix-hairpin-helix (HhH) motif was previously described for T4 RNase H (24). HhH is a non-sequence-specific DNA The density for the 3Ј arm, refined at half occupancy, does not appear at this contour level (see "Results"). C, stereo view of a ribbon diagram of T4 RNase H (light gray) is shown with the modeled portion of the 18/24 fork DNA substrate (5Ј arm, cyan; 3Ј arm, magenta). The 5Ј arm lies across the active site below the disordered bridge (gray spheres). Helix H1, the residues shown in red (part of helix H4), and the S5-H7 loop make specific interactions with the 5Ј arm and the branch point. Active site residues associated with magnesium binding are dark blue. The 3Ј DNA arm (magenta) extends past the C terminus of the protein. The double-stranded 12-mer DNA is bound to the HhH2 motif (orange), which includes helices H7 and H8 and the extended loop between them. D, surface rendering of the T4 RNase H enzyme bound to the 18/24 fork DNA substrate. The colors are the same as in panels C, E, and F. This figure has an associated video (movie 2D, available online in the supplemental material). E, ribbon diagram of T4 RNase H with the DNA substrate removed. Note the positions of specific residues involved in substrate recognition. F, secondary structure aligned with the primary sequence of T4 RNase H indicates the highly conserved residues (white on blue, defined in Fig. 4A of Ref. 21), the D132N mutation used in the structure (red asterisk), residues interacting with the 3Ј arm DNA strand (magenta triangles), and the 5Ј arm DNA strand (cyan triangles). The only residue (His-174) interacting with both DNA strands is indicated in both colors (magenta triangle underlined in cyan). As in panels C and D, the HhH2 motif is orange, whereas helices H1 and H4 and the part of loop S5-H7 interacting with the branch point and the 5Ј arm are in red. As in panel E, the N-terminal and the C-terminal regions of the enzyme are shown in yellow and green, respectively. These areas are further discussed in Fig. 6. OCTOBER 26, 2007 • VOLUME 282 • NUMBER 43

JOURNAL OF BIOLOGICAL CHEMISTRY 31717
binding motif consisting of two antiparallel ␣ helices connected by a four-residue hairpin (type II ␤ turn) (41). The 5Ј to 3Ј nucleases have extended insertions in the hairpin region (42), modifying the motif name to H3TH (24). In T4 RNase H, the corresponding structure is designated the HhH motif class 2 (HhH2) (European Bioinformatics Institute, InterPro accession number IPR008918). Sites in the fork DNA recognized by the enzyme include the 12-mer duplex region bound to the HhH2 motif (Fig. 3B), the branch point, the 5Ј arm traversing the active site under the bridge (Fig. 3C), and the 3Ј arm extending from the branch point to the C terminus of the enzyme. Interactions of the enzyme with the duplex region and the 3Ј arm segment of DNA involve mainly polar residues. In contrast, the branch point and the 5Ј arm near the active site tend to make more hydrophobic contacts with the protein.
Interactions of the HhH2 Motif with the Duplex Segment of DNA-T4 RNase H interacts with double-stranded DNA at two specific sites. The HhH2 region, helices H7 through H8, impinges on the duplex DNA section of the 3Ј arm (chain D) (Fig. 3, A and B), and His-174 closely approaches the branch point (Fig. 3C). As in the hallmark interactions of an HhH fold (42), the peptide amide groups of residues Thr-221, Met-224, and Thr-226 form hydrogen bonds with the phosphate oxygen atoms of D6Gua, D5Thy, and D4Cyt, respectively. A side view (Fig. 3B) of the DNA duplex shows nucleotides (D4Cyt through D7Cyt) from chain D interacting with the loop of the HhH2 unit. Salt bridges link the side chain of Lys-195 and the phosphate group of D5Thy and of Lys-198 and Arg-215 with D7Cyt (Fig. 3, A and B). Within the HhH2 loop, Arg-220 binds in the minor groove of the duplex DNA and interacts with the ribose ring oxygen atoms of both D5Thy and D6Gua and with the thymidine ring carbonyl of D5Thy (Fig. 3, A and B). Interaction of the HhH2 motif and the duplex DNA probably control the spacing and alignment of the DNA substrate for enzymatic cleavage.
Interactions of the T4 RNase H with the DNA Branch Point and the 5Ј Arm-Three ␣-helices, H1, H2, and H4, separate the two DNA strands; H1 interacts with the branch point, H2 with the 3Ј arm, and H4 with the 5Ј arm of the fork DNA substrate (Fig. 2C). The scissile bond is located between C7Thy and C8Thy within the DNA duplex (Fig. 1B). At the C terminus of H1, D12Ade and C7Thy make hydrophobic contacts with Val-30 and Leu-29, respectively, and the branch point base pair tilts slightly (Fig. 3C). Abutment of Leu-29 against C7Thy and the interaction of His-174 with the base pair of C8Thy:D11Ade in the major groove help position the scissile bond over Asp-155, a conserved active site residue (Fig. 3C). A stereo image of the electron density of the crucial interactions at the branch point is presented in Fig. 3D. Helix H4 is located immediately after the bridge region, a disordered segment (residues 91-96) above the active site (Fig.  2C). Residues Trp-101 and Phe-105 of H4 form a -stack with the first and second unpaired bases of the 5Ј arm, C6Thy and C5Cyt (Fig. 3, C and D). Base-stacking occurs within the singlestranded 5Ј arm, between C4Ade and C5Cyt (Fig. 3C) and between C3Ade and C2Thy (not shown). His-109 contacts C5Cyt perpendicular to the ring (cation-) and makes an H-bond with the ring nitrogen atom of C4Ade. C2Thy is in contact with the blunt end of the DNA duplex of an adjacent symmetry molecule, and C1Cyt is disordered.
Interactions of the T4 RNase H with the DNA 3Ј Arm-The electron density for the duplex and 5Ј arm of the DNA was well defined. For the 3Ј arm the electron density was not continuous, but it was possible to place 5 of 12 nucleotides (D13Gua to D17Gua), which could then be refined at half-occupancy. Defensible interactions between the 3Ј arm and the enzyme are listed in Fig. 3A. A large, flat region of electron density stacked against Phe-53 was interpreted as the guanidine moiety of D13Gua, which was the first unpaired base in the 3Ј arm. Electron density was also observed for phosphate groups and bases of nucleotides D15Cyt, D16Gua and D17Gua. D15Cyt hydrogen bonds to Lys-52, D16Gua interacts with Asn-300 and Ser-301, and D17Gua interacts with Asn-303 (Figs. 2C and 3A). D14Ade, visible in a 0.5map, was built into a hydrophobic pocket near Lys-56.
Changes in the Structure upon Substrate Binding-The structure described in the present report is a complex of a fork DNA with the T4 RNase H enzyme that has a D132N substitution in the active site. Both an inactive mutant (with a D132N mutation) and the use of the chelating agent EDTA were essential to grow these crystals. Presumably, both were required to eliminate any trace of residual activity and hydrolysis of the DNA substrate. In the protein-DNA crystal structure described here, divalent cations required for exonuclease activity were not present in the active site of the enzyme. A previous report detailed our structural work on the wild type form of the enzyme with two magnesium ions bound in the active site (PDB code 1TFR) (21). After superposition of these two structures, the root mean square deviation was found to be 1.05 Å over 283 residues: 12-88, 98 -180, and 183-305 (Fig. 4A). This comparison facilitated plausible positioning of the magnesium ions adjacent to the scissile bond in the DNA-protein complex (Fig.  4B) and placement of the DNA substrate in the active site of native metal-bound RNase H (Fig. 4C). For subsequent analysis of the active site, it was assumed that the magnesium ions are in the same position in the presence of a DNA substrate as they are in the native structure. Analyses of other nucleases suggest that active site rearrangements can occur on substrate binding (43). We are currently attempting structural studies of the T4 RNase H enzyme (native and D132N mutant) in the presence of both the fork DNA and the metal ions in the active site.
Helices H7, H8, H9, and H10 comprise the small ␣-helical subdomain and contain the HhH2 motif. Two loops, H10-H11 and S5-H7, tether the small subdomain to the large subdomain; S5-H7 is linked to the HhH2 motif (Fig. 4A). In the native structure, the N-terminal segment of S5-H7 (residues 172-183) forms a short ␤-strand (S6), whereas the C-terminal dipeptide segment (Lys-181 and Ser-182) is disordered. When DNA is present, Lys-176, Lys-179, and Lys-181 interact with phos-phates from the duplex, disrupting strand S6, and ordering residues Lys-181 and Ser-182. Concomitantly, the small domain rotates inward toward the duplex, and this motion brings the N-terminal portion of the HhH2 loop about 4 Å closer to the S5-H7 loop. As part of this rotation, loop S2-H3 (residues 73-78), which forms a ridge on the back side of the cleft, moves slightly, thus opening the active site (Fig. 4A).
Helices H1-H6 and H11-H13 plus five parallel ␤-strands (S1-S5) make up the large subdomain of the enzyme. The segment between helices H3 and H4 (residues 83-101) forms an open bridge above the active site. This region appears as a hel- ical arch in the comparable structure of the 5Ј to 3Ј exonuclease from the bacteriophage T5 (22). In the structure of the native T4 RNase H, a segment of the bridge (residues 89 -96) was too disordered to be modeled. It was initially speculated that the bridge segment would become ordered when substrate was bound, considering that the large number of basic residues in this region (Lys-87, Lys-88, Arg-90, Lys-92, and Arg-94) would be available for binding to the negatively charged DNA. With DNA in the active site, however, none of the basic residues seemed to interact with the 5Ј arm of the substrate. Moreover, bridge residues 91-96 remained disordered. Polypeptide backbone segments of Met-173 and His-174 moved ϳ2 Å, enabling the His-174 side chain to interact with the penultimate base pair (Figs. 4D and 3C). More dramatically, a displacement of nearly 5 Å of the side chain of Gln-22 led to the formation of a hydrogen bond with Ser-153 and Ser-154, both situated next to the active site (Fig. 4D).
Residues Asn-89 and Thr-98, located immediately before and after the disordered bridge, residues Asp-8 and Ser-182, located in disordered regions of the native structure (PDB code 1TFR) and residue Lys-35 located in the bend of a loop are in the generously allowed region of the Ramachandran plot (Table  1). Other structural differences can be attributed to lattice contacts. The N-terminal 11 residues, completely disordered in the native structure (PDB code 1TFR), are modeled in this new structure adjacent to a neighboring molecule. The 3Ј arm of the fork DNA is positioned nearby but does not appear to interact with the N-terminal segment. Also, the conformation of residue Lys-128 is distorted by the lattice contact between protein and the blunt end of a symmetry-related DNA molecule. This is the residue in the disallowed region of the Ramachandran plot (Table 1).

DISCUSSION
Most structures of FEN-1 enzymes have two divalent metal ions positioned in highly conserved active sites (21,22,24,25,28). A two-metal mechanism has been proposed for the FEN-1 family, analogous to that proposed for the 3Ј to 5Ј exonuclease domain of DNA polymerase I from E. coli (2,44). In this mechanism, the metal ion in site A facilitates the formation of a hydroxyl anion that carries out a nucleophilic attack on the scissile phosphate. A metal ion in site B stabilizes both the oxyanion of the leaving group and the phosphorane intermediate.
With the fork DNA as substrate, the assay of nuclease activity of T4 RNase H indicated a primary hydrolysis site at C8P between bases C7Thy and C8Thy, thereby releasing a 7-mer oligonucleotide (Fig. 1B). Secondary cleavages were detected at C6P (releasing a 5-mer) and at C9P (releasing an 8-mer), but hydrolysis at C7P was limited. With respect to the scissile phosphate, the location of Mg2 in the active site of native T4 RNase H does not favor a two-metal mechanism. Both metals in the composite structure (Fig. 4C) appear to be too far away for inner sphere coordination to the phosphates. Moreover, the twometal mechanism requires the distance between the metal ions to be less than 4 Å. Interestingly, the crystal structures of the native T4 RNase H and the archaeal Mj and Pfu FEN-l enzymes show that the two bound metal ions in the active site are ϳ7, 5, and 5 Å apart, respectively (21,24,25). These observations sug-gest that either the two magnesium ions move closer together upon DNA binding (24) or, alternatively, that FEN-1 enzymes carry out hydrolysis by a different two-metal mechanism.
Although most nucleases utilize a two-metal mechanism (43), it seems possible for T4 RNase H to manifest activity through a metalactivating process where active site residues act as a base and/or acid. In this type of nuclease activity, the role of the metal is to coordinate the substrate and shield the negative charges in the active site. In the composite structure, Asp-155 or Asp-19 could act as the base activating a water molecule for nucleophilic attack on the substrate, and Lys-199 or Lys-87 could then act as the acid, donating a proton for product release (Fig. 4B). Lys-199 interacts with Mg2 through an intervening water molecule in the native T4 RNase H (21). In mutational studies (29), Lys-87 was identified as an important contact residue for DNA binding, as evidenced by a 50% decrease in DNA binding affinity when alanine was substituted for Lys-87 in T4 RNase H. In the absence of an active site rearrangement, the general reaction mechanism for the FEN-1 family probably differs from that proposed for the Klenow fragment. Further studies will be required for a more comprehensive understanding of the mechanisms of action of this family of enzymes (24,(45)(46)(47).
Substrate Recognition by the FEN-1 Family of Enzymes-Several models for a 5Ј nuclease-DNA complex of this family of enzyme have previously been proposed (54,55). In Fig. 5 the structure of T4 RNase H (gray) with its branched substrate is superimposed on that of Afu FEN-1 (bronze) (PDB code 1RXW), the only other available structure of a FEN-1 nuclease complexed with DNA (23). The ligand in the latter complex was a DNA duplex containing a single 3Ј base overhang (23). In the superposition, the 3Ј fork arm from T4 structure (magenta) overlaps the corresponding strand of the duplex from the Afu structure (green). The 3Ј flap of the Afu duplex (red) resides in a pocket above the C-terminal helices. Comparison of duplex arms of the two superimposed DNA substrates indicates that a flap substrate was oriented at an angle of 110 -120°between the duplex axes, approximating the angle 90 -100°suggested by fluorescence resonance energy transfer experiments (23) (Fig.  5A). The 3Ј overhang binding site and the 5Ј cleavage site are separated by ϳ26 Å, positioned on opposite sides of helices of the large domain of T4 RNase H (Fig. 5A).
Sequence alignment of the FEN-1 family revealed a variable region (30) appearing as a helical arch in the T5 exonuclease and as "helical clamp" structures in the FEN-1 enzymes (22,24,25). Presumably, the helical clamps located above the active site are involved in recognition of the 5Ј arm, provided that a conformational change occurs to assist in substrate binding. Superposition of the T4 complex and Afu FEN-1 indicate that the 5Ј arm in the T4 structure threads under the helical clamp of Afu FEN-1 ( Fig. 5A and supplementary movie 5A associated with Fig. 5A) as predicted (48). However, the disordered "bridge" in T4 RNase H does not adopt a discernable specific structure in the presence of DNA.
Eukaryotic and archaeal FEN-1 enzymes exhibit their highest activities on double-flap DNA substrates, which have a single 3Ј base overhang (39). T4 RNase H displays similar cleavage specificity but with significantly reduced activity on double-flap substrates (Fig. 1C). In the structure of Afu FEN-1, the sugar moiety on the 3Ј overhang fits into a pocket formed by loops ␣2-␣3 and ␣14-␣15. The 3Ј-hydroxyl group is within hydrogen-bonding distance of three residues in these loops (Lys-48, Thr-55, and His-308, Fig. 5B). Superposition of the T4 and Afu nucleases shows that T4 RNase H has an additional turn of helix H2 (superimposes on Afu helix ␣3), a shorter loop H12-H13 (Afu loop ␣14-␣15), and that the T4 H1-H2 loop differs in length and position from the Afu ␣2-␣3 loop. Thus, the 3Ј flap pocket does not seem to be present in the T4 RNase H structure, as no residue is available for hydrogen bonding to the substrate. Moreover, T4 Ser-41 resides where the 3Ј-OH of the double-flap would bind, effectively eliminating selective binding of a 3Ј overhang (Fig. 5B). These differences in the structures of the T4 and Afu enzymes might explain why the doubleflap substrate is not preferred by T4 RNase H.
Interactions between T4 RNase H, T4 45 Clamp, and T4 32 Protein on the Lagging Strand-The structure of T4 RNase H complexed with fork DNA provides important insight into how the nuclease integrates into the rest of the T4 replication system (Fig. 6). In Fig. 6A, chain D with the 3Ј arm (magenta) of the fork and chain C with the 5Ј arm (cyan) correspond to the lagging strand template and fragment, respectively. Before the polymerase has finished elongating the upstream fragment in normal replication, T4 RNase H removes the RNA primers and some adjacent DNA in a 32 protein-dependent reaction (Fig.  6B). T4 RNase H stimulation by the 45 clamp likely serves a back-up function by loading the nuclease on nicks formed when polymerase finishes the upstream fragment before the primers have been removed by the nuclease (Fig. 6C) (9,49).
The N-terminal sequence responsible for clamp binding matches a widely conserved group adopting a common clamp interaction motif (50). This motif is found in the C-terminal region of the DNA polymerase of phage T4 and its close relative RB69, as well as near the C-terminal segment of FEN-1 nucleases. A peptide with the sequence of the clamp binding motif of the RB69 phage DNA polymerase has been shown to bind in a hydrophobic pocket on the interdomain loop of the RB69 clamp (51). The latter is analogous to the position of the clamp binding sites for the Afu and human FEN-1 proteins (23,28). In the T4 RNase H structure, the N-terminal clamp binding motif is close to the C-terminal binding site for the 32 protein (Fig.  6D). It is also located in a position relative to the active site that is similar to the position of the C-terminal clamp binding regions in other FEN-1 family nucleases (23,28).
A model of the ternary complex of the nuclease, DNA, and 32 protein was constructed by positioning the core of 32 protein (PDB code 1GPC) (52) adjacent to the end of the 3Ј fork arm in the T4 RNase H-DNA complex (Fig. 6D). The selected core structure of the 32 protein contains residues 22-239 of the 301 in the full-length protein. Residues 213-239 (blue in Fig. 6D) are part of a region (residues 213-255) thought to interact with T4 DNA polymerase (53). With the orientation of the 32 core used in Fig. 6D, the polymerase binding site is located on the opposite side from the nuclease. Polarity of the single-stranded DNA within the 32 protein remains uncertain 4 and, therefore, was not stipulated in the model of the ternary complex. In studies of T4 RNase H mutants, deletion of C-terminal residues 286 -305 (green in Fig. 6D) abolished stimulation of the nuclease by the 32 protein. Mutation of L3A or M6A or deletion of the N-terminal residues 2-10 (yellow in Fig. 6D) abolished its stimulation by the 45 clamp (9). The juxtaposition of the N and 4 Y. Shamoo, personal communication.  (23). Note the overlap of the 3Ј arm (magenta) with the corresponding strand of the Afu duplex (green). The 3Ј flap of the Afu duplex (red) lies in a pocket above the C-terminal helices. The 5Ј arm of the fork DNA from the T4 nuclease (cyan) threads beneath the helical clamp of Afu FEN-1. This figure has a video associated with it (movie 5A, available online in the supplemental material). B, a close-up view of the 3Ј overhang base together with the sugar moiety hydrogen bonding with loops ␣2-␣3 and ␣14-a15 of the Afu FEN-1. In T4 RNase H, the positions of the corresponding H1-H2 and H12-H13 loops are different. The comparable region in T4 RNase H is ␣-helical, and the main chain atoms are unavailable for hydrogen bonding. The absence of these interactions is consistent with diminished activity of the T4 nuclease on the double-flap substrate. See "Discussion" for a further comparison of the structures.
C termini strongly suggests that the binding of the 32 protein would interfere with the binding of the 45 clamp and vice versa.
In our structure, three bases of the 5Ј arm of fork DNA participate in specific interactions with the nuclease that are important for substrate recognition, substrate binding, and proper substrate cleavage (Fig. 3C). However, these interactions could not take place on the gapped and nicked DNA substrates produced during lagging strand replication (Fig. 6, B and  C). Therefore, it seems likely that the unwinding of a small number of base pairs in the gapped and nicked DNA substrates may be necessary for proper activity of the enzyme. Consistent with this suggestion, the predominant products of hydrolysis of gapped substrates by T4 RNase H alone or in a complex with the 32 protein are dimers and trimers rather than monomers (5,9,49). This generalization also holds for the hydrolysis of nicked substrates by the nuclease affixed to the T4 45 clamp.
The structure of T4 RNase H in the presence of a fork DNA substrate has direct implications for the understanding of DNA recognition by the FEN-1 family of enzymes. The HhH2 motif is crucial for substrate binding of the DNA duplex, while the 5Ј arm threads below the bridge/helical arch region. Three bases closest to the branch point interact with aromatic residues of the enzyme. Comparison with the FEN-1 enzyme reveals that the 3Ј flap binding pocket is not present in T4 RNase H, an observation that explains its substrate preference for single rather than double flaps. This novel structure also provides important insights into the processes of lagging strand synthesis, since the 3Ј arm of the DNA is situated in an optimal position for binding both the T4 32 protein and the T4 45 clamp.