Molecular Structure and Novel DNA Binding Sites Located in Loops of Flap Endonuclease-1 from Pyrococcus horikoshii *

The crystal structure of flap endonuclease-1 fromPyrococcus horikoshii (phFEN-1) was determined to a resolution of 3.1 Å. The active cleft of the phFEN-1 molecule is formed with one large loop and four small loops. We examined the function of the conserved residues and positively charged clusters on these loops by kinetic analysis with 45 different mutants. Arg40 and Arg42 on small loop 1, a cluster Lys193–Lys195 on small loop 2, and two sites, Arg94 and Arg118-Lys119, on the large loop were identified as binding sites. Lys87 on the large loop may play significant roles in catalytic reaction. Furthermore, we successfully elucidated the function of the four DNA binding sites that form productive ES complexes specific for each endo- or exo-type hydrolysis, probably by bending the substrates. For the endo-activity, Arg94 and Lys193–Lys195 located at the top and bottom of the molecule were key determinants. For the exo-activity, all four sites were needed, but Arg118-Lys119 was dominant. The major binding sites for both the nick substrate and double-stranded DNA might be the same.

replication, repair, and recombination. FEN-1 has 5Ј-flap endonuclease and 5Ј-3Ј-exonuclease activities. In DNA replication, FEN-1 removes the RNA primers during the maturation of the Okazaki fragment (6 -9). For DNA repair, FEN-1 removes damaged nucleotides after apurinic/apyrimidinic endonuclease has incised the 5Ј side of the apurinic/apyrimidinic site in long patch base excision repair (10 -12). FEN-1 is also required for nonhomologous DNA end joining of double strand DNA breaks (13). The FEN-1 sequence is conserved among eukaryotes and archaea (14 -17). Two crystal structures of FEN-1 have been reported exclusively in thermophilic archaea (18,19). The molecular structures of the members of the FEN-1 family, T5 exonuclease, T4 RNase H, and the exonuclease domain of Taq polymerase, were also reported (20 -22). They have in common a large helical arch mounted upon a globular domain containing the active site. It was postulated that the flap strand of the substrate DNA threads through this arch (22). Several studies have provided evidence of a role for the arch in tracking the flap strand (23)(24)(25). However, it remains unclear whether the flap strand of the substrate DNA threads through this arch because FEN-1 could cleave the flap strand with a secondary structure even at a reduced rate and could efficiently hydrolyze the branched structure (26,27).
The FEN-1 homologue (PH1415) was identified in the P. horikoshii genome (2) and then overexpressed in Escherichia coli, and the recombinant protein was characterized in detail using 35 different substrates (17). FEN-1 possesses 5Ј-flap endonuclease and 5Ј-3Ј-exonuclease activities. The flap endonuclease cleaves the flap strand at the junction, and the activity is independent of the length of the 5Ј flap strand, cleaving both the 1-and 5-mer flap strands efficiently as well as the 19-mer flap strand (28). The secondary structure of flap strand inhibited the endonuclease activity of eukaryotic FEN-1 (25)(26)(27). Archaeal FEN-1 could cleave the double flap strand (15,17). The flap endonuclease activity has been shown to require the upstream primer, which fills up until the junction portion, and expanding the 3Ј of the upstream primer elevates the activity (16,17,23). The 5Ј-3Ј-exonuclease activity digested the double strand DNA containing the nick, gap, and 5Ј recess-end, and the activity was elevated by expanding the 3Ј end of the upstream primer in the nick portion (28,29). There have been a few studies concerning the dual function of acidic residues located at the active center, which binds to both the substrate and active Mg 2ϩ ions (30 -32). Site-directed mutagenesis of T5 5Ј-3Ј-exonuclease showed that a conserved lysine residue (Lys 83 ) located near the active site was required for exonuclease activity but not for endonuclease activity (33). This finding indicates a difference in substrate recognition for exo-and endonuclease activities.
The substrate specificity of FEN-1 was investigated in detail as described above, whereas the DNA-binding mechanism, including the DNA-binding sites and the function of the large loop, is still unclear. To investigate the substrate-binding mechanism of phFEN-1, we determined the molecular structure of phFEN-1 as reported here. On the basis of this structure, we analyzed the function of one large and four small loops by site-directed mutagenesis, and identified several key amino acid residues involved in endo-and exo-type DNA binding and catalysis.

EXPERIMENTAL PROCEDURES
Chemicals-The vector pET-11b and E. coli strain BL21 (DE3) were purchased from Novagen (Madison, WI). The vector pGEMEX was obtained from Promega (Madison, WI). The restriction enzymes were purchased from Promega. Isopropyl-␤-D-thiogalactopyranoside was also obtained from Takara Shuzo. The oligonucleotides labeled with fluorescent dye, fluorescein (FAM), were purchased from Sawady Technology, Ltd. (Tokyo, Japan).
Expression, Purification of Recombinant phFEN-1s, and Measurement of CD-Each mutant gene was constructed on the basis of the vector pGEMEX/FEN-1 using the site-directed mutagenesis procedure and the PCR method. The wild-type and mutated genes were inserted into the vector pET-11a using NdeI and XhoI sites, resulting in the expression vectors pET-11a/phFEN-1s.
The E. coli strain BL21 (DE3) was transformed with the expression vectors. The transformant was propagated in 200 ml of 2ϫ yeasttryptone medium containing ampicillin at 37°C. The recombinant protein was induced at A 600 ϭ 1.0 -1.5 with 1 mM of isopropyl-1-thio-␤-Dgalactopyranoside for 12 h. The cells were collected by centrifugation and stored at Ϫ20°C. The frozen cells were then thawed at room temperature and mixed with 1 ml of 50 mM Tris-HCl buffer (pH 8.0). The sample was heated at 83°C for 30 min and centrifuged at 2300 ϫ g for 30 min, and then the supernatant was again heated at 83°C for 15 min to remove endogenous proteases completely before being centrifuged. The supernatant was loaded on a HiTrap SP column (Amersham Biosciences) and eluted with a linear gradient of 1 M NaCl in a 50-mM Tris-HCl (pH 8.0) buffer. The recombinant phFEN-1s were eluted with buffer containing 200 -300 mM NaCl.
The secondary structure of the purified proteins was measured using a circular dichroism (CD) spectrometer, AVIV model 62DS (Shimazu, Kyoto, Japan) and a quartz cuvette having a path length of 1.0 or 2.0 mm. The protein concentration was ϳ0.1 mg/ml of 50 mM Tris-HCl (pH 8.0).
Structural Determination-Crystals were obtained by hanging-drop vapor diffusion using the R42E mutant protein. To a 2-l solution containing 2% protein and 200 mM NaCl, pH 8.0 (50 mM Tris-HCl buffer), an equal amount of reservoir solution containing 20% PEG 6000 at pH 6.0 (100 mM MES buffer) was added, and a droplet of the solution was equilibrated over a 1-ml reservoir solution. A rod-shaped crystal was obtained belonging to the trigonal space group P31, and the cell dimensions were a ϭ b ϭ 62.67 and c ϭ 180.69 Å. The asymmetric unit contained two independent molecules (V m ϭ 2.56 Å 3 /Da). The intensity data were measured to a resolution of 3.0 Å using synchrotron radiation ( ϭ 1.0 Å) on an ADSC Quantum 4R charge-coupled devices detector at the BL-6A beam line station of the Photon Factory. A total of 135,478 observed reflections were merged to a set of 15,903 (100% completeness) with an R merge value of 0.104.
The structure was solved by the molecular replacement method using the structure of the FEN-1 from Methanococcus jannaschii. A self-rotation search using the program X-POLR indicated that the two independent molecules are related by a local 2-fold axis. The structure was refined by X-PLOR for 13,576 reflections in the resolution range from 8.0 to 3.1 Å. The refinement converged at R and R free values of 0.190 and 0.279, respectively. The root mean square deviations of bond distances and angles from their ideal values were 0.020 Å and 3.1°, respectively. The coordinates were deposited in the Protein Data Bank (ID code 1MCB).
Oligonucleotides Used and Preparation of Substrates-The following oligonucleotides were designed to construct various DNA substrates: Oligonucleotide A, 54-mer, 5Ј-GAGCTAGATGTCGGACTCTGCCTCAA-GACGGTAGTCAACGTGCACTCGAGGTCA-3Ј; Oligonucleotide B, 27mer, 5Ј-TGACCTCGAGTGCACGTTGACTACCGT-3Ј; Oligonucleotide C, 57-mer, 5Ј-GCATCTGACGGATGTCAAGCAGTCCTAACTCTTGAG-GCAGAGTCCGACATCTAGCTC-3Ј; Oligonucleotide D, 27-mer, 5Ј-CT-TGAGGCAGAGTCCGACATCTAGCTC-3Ј; and Oligonucleotide E, 54mer, 5Ј-TGACCTCGAGTGCACGTTGACTACCGTCTTGAGGCAGAG-TCCGACATCTAGCTC-3Ј. Oligonucleotides C and D were labeled at the 5Ј terminus with a fluorescence group (FAM) and then purified using high pressure liquid chromatography. Annealing reactions to construct various DNA substrates were performed as described previously (17). The flap substrate was constructed with oligonucleotides A, B, and C labeled with FAM. The nick substrate was constructed with oligonucleotides A, B, and D labeled with FAM. The dsDNA was constructed with oligonucleotides A and E.
Kinetic Analysis of Mutants Using Substrates with Fluorescence-The levels of activity of phFEN-1s were measured using substrates with a highly fluorescent dye, fluorescein (FAM), at the 5Ј end of the flap strand or 5Ј end of the upstream primer of the nick substrate. The pattern observed for the digestion of fluorescent substrate was the same as that observed for the radiolabeled substrate. The product amount was evaluated with a fluorescence scanner after electrophoresis. This method has almost the same sensitivity as the radioactive-labeling methods employed for the phFEN-1 assay. Hence, we used the fluorescent substrate throughout the study.
Measurement of Kinetic Parameters-The kinetic parameters were determined using the fluorescence substrate. The reaction mixture was 50 mM Tris-HCl (pH 7.4) containing the labeled substrate, 1.5 mM MgCl 2 , and 1.5 g of bovine serum albumin. The enzyme and substrate concentrations were varied from 0.025 to 200 ng and from 0.1 to 75 pmol, respectively. A 10-fold volume (150 l) of the reaction mixture was preheated at 60°C for 5 min and 14 l was added to 10 l of stop solution as a control for a zero time reaction. Nine volumes of enzyme were added, and 15 l of the reaction mixture was added to 10 l of the stop solution every 1 min for 7 min. The samples were denatured at 95°C for 10 min. Five l of sample was loaded onto a 12% polyacrylamide gel (10 ϫ 10 cm) with 7 M urea and 1ϫ TBE, and electrophoresed with the same buffer for 1.5 h at 200 V. The reaction products were visualized and quantified using a FluorImager 585 (Molecular Dynamics, Inc., CA). The initial velocities were obtained directly from the initial slopes of the time course plots. The K m and k cat values were calculated using the Michaelis-Menten equation and the least squares method (34).
Surface Plasmon Resonance Experiments-The interaction between phFEN-1 and dsDNA were quantitatively analyzed on a BIAcoreX apparatus (Biacore, Uppsala, Sweden) at 25°C. The flow cell was routinely equilibrated with 50 mM Tris-HCl buffer (pH 8.0), 10 mM NaCl, 5 mM EDTA, and 0.005% Tween 20. The biotinylated dsDNA, which was constructed with 5Ј-biotinylated oligonucleotide A and oligonucleotide B, was coupled to a streptavidin-dextran layer on the surface of Sensor Chip SA (Biacore). The 5Ј-biotinylated dsDNA were injected for 120 s, resulting in an immobilization corresponding to ϳ600 resonance units. The protein were diluted in the same buffer and injected for 120 s. In all cases, at least nine different concentrations ranging from 10 nM to 10 M protein were injected. The obtained sensograms were analyzed by the evaluation software (Biacore) to calculate the association constants on the basis of the simple 1:1 binding model.

RESULTS
Three-dimensional Structure of phFEN-1-The crystallization was investigated with the wild type and some mutant enzymes, and a rod-shaped crystal was obtained with the R42E mutant. The crystal structure was solved by molecular replacement using the structure of FEN-1 from M. jannaschii and was refined at a resolution of 3.1 Å. The two independent molecules are related by the local 2-fold axis as shown in  (17), the folding structure of small loop 2 differs markedly between P. horikoshii and P. furiosus as shown in Fig. 1B. The structural flexibility of these two loop regions of phFEN-1 is also indicated by the superimposition of two independent subunit molecules from the dimer structure (data not shown).
The two independent subunits form a dimer structure having a pseudo 2-fold symmetry as shown in Since several basic amino acids and their clusters were observed on these loops that were expected to be important for binding DNA, 45 mutant genes were constructed by site-directed mutagenesis to analyze the function of these positively charged clusters. The mutation points are summarized in Fig. 2. The expression level of each mutant in E. coli was ϳ3-fold lower than that of the wild type. These mutants were purified using the same procedure as was used for the wild type. The molecular weight of each mutant was the same as for the wild type except R40E, which had a molecular weight half of that of the wild type, presumably due to proteolytic digestion. The R40E protein was not used in the subsequent experiment.
The CD profiles of the mutant proteins were measured to confirm that no significant changes had occurred by sitedirected mutagenesis. As shown in Fig. 3, the CD spectra of the mutants were the same as the spectrum of the wild type. Hence, the following kinetic analysis was performed using these mutant proteins.
Kinetic Parameters of Mutants on Four Small Loops and One Large Loop Using Flap Substrate-As shown in Fig. 2, two pairs of basic amino acids (Arg 40 and Arg 42 , and Lys 51 and Arg 53 ) were observed typically on both edges of small loop 1 (39 -55) with highly conserved amino acids (Leu 47 , Gly 44 , and Gly 52 ) between the eukaryotes and archaea. Various mutations were introduced into this typical region to investigate the function of these residues. As shown in Fig. 4, the K m of the single mutants R40G and R42G both increased 7-fold, compared with that of the wild-type enzymes (WT). The K m value of the single mutant R40Q increased 10-fold, whereas that of R42Q was almost the same as the value for of WT. The K m and k cat /K m values of R42E were 19-fold higher and 25-fold lower than those of WT, respectively. These results indicate that the mutational effect was increased in the following order: no side chain Ͻ neutral side chain Ͻ negative side chain. The tendency was magnified for the double mutants. The K m value of R40G/ R42G was elevated 26-fold compared with that of the wild type, as well as R40Q/R42Q. The K m of R40E/R42E was increased 105-fold compared with WT, suggesting strong repulsion among these residues and DNA. The k cat and k cat /K m values were also decreased 4-and 680-fold, respectively. On the other hand, the mutations of Lys 51 and Arg 53 showed little difference in the K m , k cat , and k cat /K m values compared with those of WT. Even the double mutant K51E/R53E retained 30% of the WT k cat /K m value.
The K m value of L47G was increased 20-fold, whereas the values of a conservative mutant L47F was similar to that of WT. The k cat values of both L47G and L47F were almost the same as that of WT. These findings indicated important hydrophobic interactions between Leu 47 and other hydrophobic side chains to maintain the typical loop structure. The mutations at Gly 44 and Gly 52 had no significant effect on the K m , k cat , and k cat /K m values.
Concerning small loop 2 (187-206), the basic cluster, Lys 193 , Arg 194 , and Lys 195 , is present on one side of the loop in which the conserved Lys 199 is located in the middle as shown in Fig.  2. The K m value of each of the single mutants K193A, R194A, and K195A was 4-, 5-, and 8-fold of WT, respectively, whereas the k cat values of these mutants were almost the same as for WT as shown in Fig. 4. For the triple mutant K193A/R194A/ K195A, the K m value increased markedly and the k cat value decreased moderately. The negatively charged triple mutant K193E/R194E/K195E showed similar but more magnified effects on both parameters compared with the alanine triple mutant K193A/R194A/K195A, indicating ionic interactions of the positive cluster with DNA phosphate groups in an analogous manner to Arg 40 and Arg 42 on small loop 1. The values of K199A differed little from those of WT.
One DNA binding motif was putatively identified at small loop 3 (234 -249) in the molecular structure of FEN-1 from P. furiosus (19). Small loop 4 (257-263) is also close to small loop 3. The positive residues, Lys 243 , Lys 248 , Lys 249 , and Lys 263 , and the conserved residue Tyr 237 are located inside or close to the motif as shown in Fig. 2. Several mutants were made to investigate the function of these positive or conserved residues. However, the K m and k cat /K m values of the negative mutants (K243E and K263E) and the alanine mutants (K243A, K248A, K249A, K263A, and Y237A) did not change markedly compared with those of WT, indicating less significant roles for these positive residues compared with those on small loops 1 and 2.
The large loop (80 -128) is 49 amino acids in length, and   Fig. 2. For the first cluster, the K m and k cat values of R88A and K89A did not change markedly, compared with those of WT. Interestingly, the k cat value of K87A was decreased 400-fold, whereas the K m value was almost the same as that of WT, suggesting a significant role for Lys 87 in the catalytic mechanism. Additive effects were observed for the triple mutant K87A/R88A/K89A in which the K m was 5-fold higher, although the k cat was 184-fold lower than that of WT due to the large contribution of K87A to the k cat value. In the second cluster, the R94A mutation showed a 12-fold increase in the K m value and a 15-fold decrease in the k cat value, whereas the effects of the mutations K93A and R95A on both parameters were minor. The K m and k cat values of the triple mutant K93A/R94A/R95A was increased by 17-fold and decreased by 96-fold, respectively, compared with those of WT, indicating the central role of Arg 94 in binding the substrate in the middle of the positive cluster. In the third cluster, the K m value of the double mutant R118A/K119A was increased 10-fold, while the k cat value was almost the same as for WT.
Kinetic Parameter of Mutants on Four Small Loops and One Large Loop Using Nick Substrate-FEN-1 possesses 5Ј-flap endonuclease and 5Ј-3Ј-exonuclease activities, and little is known about how the substrate is recognized by FEN-1 endo/ exonucleolytically. To clarify differences between the endo-and exo-type binding, a kinetic analysis of 5Ј-exonuclease was performed using nick substrate. The kinetics of exo-activity were analyzed using single, double and triple mutants (Arg 40 , Arg 42 , Lys 87 , Arg 94 , Arg 118 , Lys 119 , and Lys 193 -Lys 195 ) with significant differences in endo-activity compared with WT.
The K m values of two double mutants, R40G/R42G and R118A/K119A were magnified by 6-and 17-fold, respectively, compared with the K m of WT, suggesting a large contribution by these two regions to the exo-type binding as shown in Fig. 5. The k cat value of the single mutant R94A was decreased 200fold compared with that of WT. The k cat value of the double mutant R118A/K119A was decreased 111-fold. The k cat /K m values of R40G/R42G, R118A/K119A, and K193A/R194A/K195A decreased 222-, 1851-, and 76-fold, respectively compared with the value of WT, indicating that small loops 1 and 2 and the two regions of the large loop play important roles in recognizing the nick substrate exonucleolytically.
The single mutant K87A showed weak activity on the nick substrate, too weak to determine the kinetic parameters, suggesting the functional importance of the residue in exonucleolysis as well as endonucleolysis.
Binding Analysis by Surface Plasmon Resonance Experiments-The band shift assay was not successful because phFEN-1 could not enter the native gel. The substrate binding affinity was therefore analyzed by surface plasmon resonance measurements. A 54-mer dsDNA was immobilized on the Sensor Chip. The affinity constants (K a ) of WT and mutant protein are shown in Table I. The K a value of R40G/R42G and R118A/ K119A dropped 6-and 4-fold, respectively, and the values for each single mutant decreased moderately compared with that of WT, whereas the K a value of the other mutants showed little difference to that of WT. The results indicated that both these sites, Arg 40 and Arg 42 and Arg 118 -Lys 119 , contribute to the binding of dsDNA. Furthermore, we tried to analyze the binding affinity with single-stranded DNA. The K a values of all these mutants for single-stranded DNA did not decrease significantly compared with the value for WT (data not shown).  archaea (14,18) for which differences in substrate binding were proposed. According to the two-modeled structures of the protein-substrate complex, the 5Ј end of the single-stranded flap strand threads through the hole formed by the L1 loop (corresponding to the large loop in our work). However, the orientation of the main double-helical strands shown in the proposed complex crosses at a 90°angle. Despite the abundance of structural information on thermostable FEN-1s from archaea, the substrate binding models were too different to comprehend the general type of binding, and little was known about how the substrates were recognized by FEN-1s endo/exonucleolytically. To understand the substrate recognition mechanism of phFEN-1 at the molecular level, the crystal structure was determined to a resolution of 3.1 Å. phFEN-1 has a novel dimer structure in which one large loop and four small loops form a large cylindrical active cleft as shown in Fig. 1A. On the basis of the structural information for phFEN-1, 24 amino acid residues in several loops were mutated systematically to search for recognition sites in the phFEN-1 molecule (Fig. 2). The mutant proteins were purified completely, and the structural integrity was checked by a CD measurement as shown in Fig.  3. The nuclease activities were measured with 32 P-labeled substrates, but the reproducibility of the reactions was too poor to allow determination of the kinetic parameters because the labeling efficiency with T4 kinase and unstable 32 P was variable. Finally, a new assay method using fluorescence substrates and a FluorImager was developed, and the kinetic parameters were determined for 45 mutants, including single, double, and triple mutants at different loop regions as shown in Figs. 4 and 5.
For small loop 1 (39 -55), the kinetic parameters summarized in Fig. 4 clearly indicated that the former pair, Arg 40 and Arg 42 , bound equally to the substrate, and the latter pair, Lys 51 and Arg 53 , played a less significant role for the substrate binding and the catalytic reaction. In the molecular structure, Leu 47 was located in the middle of the two positive pairs on small loop 1 and showed tight hydrophobic interactions with the methyl and methylene groups of Ile 39 , Gln 41 in the same loop, and Ser 59 located near the loop. The kinetic parameters of Leu 47 mutants, as shown in Fig. 4, strongly suggested that their hydrophobic interactions might be critical in maintaining the typical loop structure and, consequently, in adjusting the orientation of the positive side chains of Arg 40 and Arg 42 to the substrate. The mutations for removing the hydrophobicity of Leu 47 might increase the K m value by eliminating the function of Arg 40 and Arg 42 .
For small loop 2 (187-206), the three positive residues Lys 193 , Arg 194 , and Lys 195 bind cooperatively to the DNA by ionic interaction in the same manner as ARG 40 and Arg 42 on loop 1 as shown in Fig. 4. However, Lys 199 has a less significant function in binding the substrate than does the cluster. In the sequence alignment of small loop 2 among six FEN-1s from eukaryotes and archaea, the small loop 2 of phFEN-1 is longest, having nine extra residues, compared with those of the eukaryote FEN-1s, as previously reported (17). As shown in Fig.  1B, phFEN-1 and pfFEN-1 show a large difference in small loop 2 folding despite their high sequence identity (95%), suggesting a marked flexibility of the loop. The conformational change of the small loop might be induced to fit the positive cluster (Lys 193 , Arg 194 , and Lys 195 ) to bind DNA strands because it was reported that the helicity of the human FEN-1 structure was increased up to 4% by addition of DNA substrates (35).
The positive residues Lys 243 , Lys 248 , Lys 249 , and Lys 263 are also located in small loop 3 (234 -249) and small loop 4 (257-263) as shown in Fig. 2. The kinetic parameters of various mutants for these residues, even the negatively charged mu-tants, as shown in Fig. 4, were not changed markedly suggesting less significant roles for these positive residues compared with those in small loops 1 and 2.
The three positive clusters (87-89, 93-95, and 118 -119) were located at the large loop (80 -128). For the first cluster (87-89), only residue Lys 87 showed drastic changes in the k cat value for endo-type hydrolysis on replacement of alanine compared with WT, although the K m value of the mutants did not change markedly as shown in Fig. 4. Our findings indicated a significant contribution of Lys 87 to the catalytic mechanism. Furthermore, the functional importance of Lys 87 was noteworthy, regarding the molecular structure. The Lys 87 residue forms a hydrogen bond with Pro 84 with a distance of 2.9 Å. Interestingly, two successive proline residues (corresponding to Pro 84 and the neighboring Pro 83 in the phFEN-1 molecule) are conserved completely in the initial part of the large loop (80 -128) for all FEN-1s reported from eukaryotes and archaea, suggesting their structural importance (17). Because the proline residue has a fixed peptide bond angle, the two successive Pro residues, Pro 83 and Pro 84 , might be essential in maintaining the specific structure of the initial part containing the catalytic Asp 80 . The mutation K87A that loses the hydrogen bond with Pro 84 might cause a positional shift of the catalytic Asp 80 from the active center due to a slight modification in the loop structure leading to a decreased k cat value.
In the second positive cluster (93-95), only Arg 94 showed marked changes in the kinetic parameters on replacement of alanine as shown in Fig. 4. The K m and k cat values of R94A were greatly increased and decreased, respectively, clearly suggesting a markedly decreased ratio of the productive complex per total ES complex according to the subsite theory (36). Arg 94 is essential to bind the substrate, whereas Lys 93 and Arg 95 play supplemental roles. The positive charges of Lys 93 and Arg 95 might be important in adjusting environmental conditions for Arg 94 .
In the third cluster (118 and 119), the K m value of the double mutant R118A/K119A was increased 10-fold as shown in Fig. 4, although the single mutant, R118A or K119A, showed a moderate increase in the value probably due to the functional complementarity between the two residues. These findings clearly indicate that the substrate DNA is recognized by ionic interactions with Arg 94 , Arg 118 , and Lys 119 located at both typical sites of the large loop as shown in Fig. 2, whereas Lys 87 located in the early part of the loop is deeply involved in the catalytic mechanism.
Substrate Binding Sites Commonly Used Exo-and Endonucleolytically-For the endo-type reaction as mentioned above, four substrate binding sites (ARG 40 and ARG 42 , Arg 94 , Arg 118 Lys 119 , and Lys 193 Arg 194 Lys 195 ) were identified and were located on small loops 1 and 2, and on the large loop, independently. FEN-1 possesses dual activities; exo-activity on dsDNA with nick and recess end, and endo-activity on the flap substrate. The same catalytic residues binding to an essential metal ion Mg 2ϩ are indispensable for both activities. Furthermore, Garforth et al. reported an interesting mutant of T5 5Ј-3Ј-exonuclease belonging to the same family as FEN-1, which lost the exo-activity but retained the endo-activity on replacement of the Lys residue close to the active center (33). However, it is still unclear how FEN-1 recognizes the different substrates exo/endonucleolytically.
To clarify the endo-and exo-type recognition mechanism of FEN-1, kinetic parameters of exo-activity against the nick substrate were determined and are shown in Fig. 5. Using the kinetic parameters for the exo/endonucleolysis summarized in Figs. 4 and 5, the DNA binding manner specific for each activity was evaluated according to the subsite theory using following equations (36 -40), where ⌺K j corresponds to the sum of the binding constants for all binding forms of the substrate. K j equals the sum of K p (the binding constant of ES complex leading to the product) and K q (the binding constant of ES complex failing in production). k int is the intrinsic rate constant of the hydrolysis. The k int value was postulated as invariant among mutant and WT enzymes, because no significant structural change had been detected by the CD measurement as shown in Fig. 3  For the exo-type hydrolysis as shown in Fig. 6A, II, the K p value was more markedly reduced by mutation in four parts of the active cleft (Arg 94 , Arg 118 -Lys 119 , Arg 40 and Arg 42 , and Lys 193 -Arg 194 -Lys 195 ). These facts indicate that all four binding sites play critical roles. The role of Arg 118 -Lys 119 is dominant in the exo-type reaction, in contrast to its less important part in the endo-type reaction. As shown in Fig. 6A, II, the K j value of the R94A was not significantly changed by the mutation. This clearly indicates that, in the exo-type reaction, Arg 94 prefers to take a productive form rather than bind to the substrate, whereas both functions of Arg 94 are needed in the endo-type reaction.
The affinity of phFEN-1 for the dsDNA was evaluated by surface plasmon resonance measurement as shown in Table I and Fig. 6A, III. The results indicate that two sites Arg 40 and   FIG. 6. A, the difference in DNA binding among four sites. I, comparison of the K j and k int K p values for endonucleolysis using the flap substrate. The K j and k int K p values were evaluated from the K m and k cat values according to the equations described under "Discussion." The insert shows the k int K p values expressed as logarithm. II, comparison of the K j and k int K p values for exonucleolysis using the nick substrate. The log (k int K p ) values are shown in the inset. III, comparison of the K a values for dsDNA. B, stereo view of the four identified DNA binding sites and Lys 87 in phFEN-1. The number indicates the position of DNA binding sites (red). Small loops 1 and 2 and the large loop were colored yellow, green, and pink, respectively. A small ball (orange) shows the active center.
Arg 42 and Arg 118 -Lys 119 mainly contribute to the binding to dsDNA. Comparing the binding affinity of the mutants between Fig. 6A, III and II in which the affinity was evaluated as the K j value, the profiles seems to be similar. The similarity suggested that the binding affinity for both the nicked substrate and dsDNA is produced by two commonly used sites (Arg 40 and Arg 42 and Arg 118 -Lys 119 ) with less of a contribution by the other two sites (Arg 94 and Lys 193 -Arg 194 -Lys 195 ). Interestingly, this indicates that the major binding form might be shared between dsDNA and the nick substrate.
Here, we successfully identified four DNA binding sites that share affinity to form productive ES complexes specific for each endo-or exo-type hydrolysis, probably by bending the substrates. The molecular structure of phFEN-1 was solved as a dimer form; however, unfortunately, we could not obtain the strong biochemical evidence that the dimer formation was necessary to fulfill the function of phFEN-1. A structural analysis of the ES complex with a substrate analogue will clarify the functional role of the dimer formation and the defined mechanism for endo/exonucleolytical substrate recognition.