Neutralizing Mutations of Carboxylates That Bind Metal 2 in T5 Flap Endonuclease Result in an Enzyme That Still Requires Two Metal Ions*

Flap endonucleases (FENs) are divalent metal ion-dependent phosphodiesterases. Metallonucleases are often assigned a “two-metal ion mechanism” where both metals contact the scissile phosphate diester. The spacing of the two metal ions observed in T5FEN structures appears to preclude this mechanism. However, the overall reaction catalyzed by wild type (WT) T5FEN requires three Mg2+ ions, implying that a third ion is needed during catalysis, and so a two-metal ion mechanism remains possible. To investigate the positions of the ions required for chemistry, a mutant T5FEN was studied where metal 2 (M2) ligands are altered to eliminate this binding site. In contrast to WT T5FEN, the overall reaction catalyzed by D201I/D204S required two ions, but over the concentration range of Mg2+ tested, maximal rate data were fitted to a single binding isotherm. Calcium ions do not support FEN catalysis and inhibit the reactions supported by viable metal cofactors. To establish participation of ions in stabilization of enzyme-substrate complexes, dissociation constants of WT and D201I/D204S-substrate complexes were studied as a function of [Ca2+]. At pH 9.3 (maximal rate conditions), Ca2+ substantially stabilized both complexes. Inhibition of viable cofactor supported reactions of WT, and D201I/D204S T5FENs was biphasic with respect to Ca2+ and ultimately dependent on 1/[Ca2+]2. By varying the concentration of viable metal cofactor, Ca2+ ions were shown to inhibit competitively displacing two catalytic ions. Combined analyses imply that M2 is not involved in chemical catalysis but plays a role in substrate binding, and thus a two-metal ion mechanism is plausible.

Flap endonucleases (FENs) 4 , essential in all life forms, are members of the 5Ј-nuclease superfamily of structure-specific phosphate diesterases that carry out essential divalent metal ion-dependent nucleic acid hydrolysis (1). FENs act upon 5Јbifurcated structures known as flaps formed during lagging strand DNA replication and long patch base excision repair. Hydrolysis predominantly occurs one nucleotide into the 5Ј-duplex adjoining the site of bifurcation. An unusual feature of these enzymes is their high density of active site carboxylates that coordinate essential divalent metal ion cofactors. Bacterial and bacteriophage FENs possess eight active site acidic residues, seven of which are conserved in FENs from higher organisms and also in other members of the 5Ј-nuclease superfamily (1)(2)(3)(4). In contrast, a typical metallonuclease active site consists of three or four carboxylates (5). However, a subclass of FEN paralogues, typified by Escherichia coli ExoIX, exists in eubacteria and appear to lack three of the metal-binding carboxylates (6).
Although the most common mechanism suggested for metal-dependent phosphoryl transferases involves two metal ions in contact with the scissile phosphate diester, this is not universally accepted (5,(7)(8)(9)(10). Debate about the validity or otherwise of two-metal ion catalysis has focused on several enzymes including flap endonucleases (11)(12)(13)(14)(15)(16). Crystallographic studies of substrate-free FENs have demonstrated two active site-bound metal ions (14,(17)(18)(19). Metal ion 1 (M1) occupies a similar position in all structures. However, there is some variation in the position of the site coordinating a second metal ion (M2), often observed in FEN crystal structures. The distance between the M1 and M2 sites varies, and the separations observed are generally much greater than the 4 Å required for both metals to contact the same phosphate diester. For example, in bacteriophage T5FEN, 5 the M1-M2 separation of two Mn 2ϩ ions is 8 Å (Fig. 1A) (17). However, in a substrate-free structure of hFEN1, two magnesium ions are bound with a separation of 3.4 Å in one of three proteins bound to proliferating cell nuclear antigen (Fig. 1B) (18). A very recent structure of hFEN1 bound to product DNA has two Sm 3ϩ ions bound in equivalent positions, and these contact the cleaved phosphate monoester (4).
Despite the 8 Å distance between M1 and M2 in substratefree T5FEN structures, magnesium ion stimulation data are consistent with a mechanism where the major rate acceleration is brought about by two metal ions but where overall (k cat /K m ) has a three-divalent ion requirement (3). Thus an additional ion (M3) is bound with higher affinity to the enzyme-substrate complex. If two metal ions that participate in chemical catalysis are coordinated to the central carboxylate of the T5FEN active site (Asp-130), positioned as those observed in hFEN1 structures ( Fig. 1) (2, 4), two-metal ion catalysis mechanisms where both metals coordinate the scissile phosphate diester could take place. As only M1 is observed in this position in substrate-free structures of bacteriophage FENs, in this scenario, the third ion implicated by functional studies would have to bind in this location, and the catalytic ions would therefore be M1 and M3 (Fig.  1A). However, alternative explanations of a two-ion requirement for the major rate acceleration are possible. For example, M2, as positioned in T5FEN substrate-free structures, would not be able to directly interact with the scissile phosphate diester but instead could catalyze chemical catalysis by more remote interactions. On the other hand, if catalytic M1 and M3 provide the main rate acceleration, M2 could play a role in enzyme-substrate complex stability.
Here we test earlier proposals for the roles of metal ions in reactions catalyzed by T5FEN. To investigate whether divalent metal ions alter the stability of enzyme-substrate complexes, dissociation constants were determined in the absence of catalysis at varying concentration of Ca 2ϩ ions and pH. Previous studies of the T5FEN-catalyzed reactions demonstrated that catalytically inert Ca 2ϩ ions inhibited reaction with a 1/[Ca 2ϩ ] 2 dependence, which is consistent with an absolute requirement for two viable metal cofactor ions that the Ca 2ϩ ions displace. Potentially, this could be used to test for a requirement for two metal ions in mutated FENs and other metallonucleases. However, it is possible that decreases in the rate of reaction may involve Ca 2ϩ binding non-competitively to the enzyme, substrate, or enzyme-substrate complex. To establish the competitive inhibition of FEN-catalyzed reactions by Ca 2ϩ ions, the inhibitory characteristics were studied with different supporting metal cofactor ions at varying concentration. To probe the role of T5FEN M2, both the stimulation by viable metal cofactor ions and the inhibition by Ca 2ϩ ions were evaluated for a mutated T5FEN where two of the carboxylates that bind M2 are altered. The combined analyses support a mechanism for T5FEN where the main rate acceleration is afforded by two ions, but not by M2, and imply a role for M2 in stabilization of enzyme-substrate interactions.
Steady State Kinetic Analyses-Steady state kinetic parameters of WT and D201I/D204S T5FEN (E) were evaluated at 37°C using HP5F substrate (S) in 25 mM CHES or potassium glycinate, pH 9.3, 0.1 mg/ml BSA, and 1 mM DTT as described (3). MgCl 2 was added to the desired concentration, and the ionic strength (MgCl 2 ϩ KCl) was adjusted to 80 mM using KCl. Substrate concentrations were varied from 0.12 to 8 ϫ K m , and reactions were sampled and quenched by the addition of an equal volume of 25 mM EDTA at the appropriate time intervals. Reactions were analyzed by denaturing HPLC equipped with a fluorescence detector, and initial rates (v) were calculated as described (26  mixtures containing 0.5 or 2 mM MgCl 2 , 0.5 M HP5F, 25 mM CHES, pH 9.3, 0.1 mg/ml BSA, CaCl 2 with the appropriate amount of KCl were initiated by the addition of the appropriate amount of D201I/D204S T5FEN (final concentration 5-40 nM). All reactions were sampled and quenched as above and used to determine k obs (v/[E]) as a function of calcium ion concentration.
Fluorescence Anisotropy-T5FEN DNA interactions were evaluated by fluorescence anisotropy using a HORIBA Jobin Yvon FluoroMax-3 fluorometer with automatic polarizers. The excitation wavelength was 490 nm (excitation slit width 8 nm) with emission detected at 509 -511 nm through an 8-nmwide slit. Experiments contained 2 mM EDTA or varying concentrations of CaCl 2 (0.5-10 mM), with the appropriate amount of KCl as for kinetic analyses, 100 nM HP5F (for WT T5FEN) or 10 nM HP5F (for D201I/D204S), 25 mM HEPES, pH 7.5, or 25 mM potassium glycinate, pH 9.3, 0.1 mg/ml BSA, and 1 mM DTT. All experiments were carried out at 37°C. Anisotropy (r) was determined as described (27) initially prior to the addition of protein and then on the cumulative addition of enzyme.
Data were fitted, using KaleidaGraph, to the following equation with corrections made for dilution, where r is the measured anisotropy, [E] is the total protein concentration and [S] is the total HP5F concentration, r min is the anisotropy of free DNA, r max is the anisotropy of the DNAprotein complex, and K D is the dissociation constant (27).
Data Analyses-Curve fitting of Mg 2ϩ -dependent behavior and Ca 2ϩ -dependent inhibitions were carried out by non-linear regression fitting using KaleidaGraph software weighted according to individual error values when required, using Equations 2-7 (see "Results").

The Effect of Divalent Metal Ions on WT FEN-Substrate
Interactions-Studies on the magnesium ion dependence of the WT T5FEN-catalyzed reaction were carried out previously using the 5Ј-fluorescent overhanging hairpin substrate HP5F under pH-independent maximal rate conditions (pH ϭ 9.3) (3). These earlier kinetic studies revealed a requirement for two Mg 2ϩ ions to maximize 1/K m . However, the interpretation of this behavior was complicated by the nature of this parameter, which in FEN-catalyzed reactions that are partially rate-limited by the release of products reflects the stability of all forms of the enzyme-bound substrate and product. To examine binding of HP5F to T5FEN under similar conditions to catalytic experiments but in the absence of catalysis, fluorescence anisotropy was employed. In the presence of EDTA at pH 9.3, anisotropy (r) increased with the addition of protein up to 10 M, but no maximal value was reached, implying K D Ͼ 5 M ( Fig. 2A). In contrast, in the presence of non-catalytically competent Ca 2ϩ ions (1 mM), r reached a maximum value and could be fitted to a single binding isotherm (Equation 1) to yield a K D ϭ 202 Ϯ 27 nM ( Fig. 2A). Thus divalent metal ions substantially stabilize the T5FEN-HP5F complex at pH 9.3.
However, at pH 7.5 in the absence of divalent metal ions, K D ϭ 231 Ϯ 13 nM (Fig. 2B). At this pH, the presence of Ca 2ϩ ions (1 mM) has a modest effect, with K D ϭ 120 Ϯ 13 nM. Thus either lowering the pH or adding divalent metal ions has a large effect on T5FEN-substrate interactions, but these effects are not additive. Similar stimulation of binding of DNA substrates at lowered pH or by Ca 2ϩ ions has been observed with other metallonucleases and been assigned to protonation or binding of Ca 2ϩ ions at the active site carboxylates (28). Here the protonation state of DNA-binding residues could also be a factor (29,30).
The variation in K D as a function of [Ca 2ϩ ] was determined at both pH 9.3 and pH 7.5. Calcium ion concentration was varied from 0.5 to 10 mM, maintaining the ionic strength by adjusting the concentration of KCl. Under these conditions, the slopes of log K D versus log [Ca 2ϩ ] were 0.15 Ϯ 0.10 (pH 9.3) and Ϫ0.06 Ϯ 0.06 (pH 7.5), indicating that the changes in enzyme-substrate affinity were negligible (Fig. 2D). The relative insensitivity of K D to concentrations of divalent ions Ͻ20 mM has been noted in other systems where divalent ions are cofactors (31). Importantly, at both pH values, the changes in enzyme-substrate affinity on altering Ca 2ϩ ion concentration are relatively small and so do not account for the inhibition characteristics of the enzyme-catalyzed reaction by calcium ions described below.
Inhibition of WT T5FEN-catalyzed Reactions with Ca 2ϩ Ions-Earlier studies have revealed that Ca 2ϩ ions inhibit the T5FENcatalyzed reaction, and under some conditions, this is dependent on 1/[Ca 2ϩ ] 2 (3). Assuming competitive inhibition with viable metal cofactor ions, a two-Ca 2ϩ ion inhibitory response could be a diagnostic test for a requirement for two viable metal ion cofactors. Previously, it was demonstrated that the concentration of Ca 2ϩ ions required to reduce the rate of reaction by 50%, K I in Equation 2, was a function of both the concentration and the identity of the viable metal cofactor (3). This analysis assumes that replacement of a single viable cofactor ion with Ca 2ϩ ion prevents reaction, where k obs is the observed normalized initial rate of reaction (v/[E]) and k 0 is v/ [E] in the absence of inhibitory ion. Thus at least one Ca 2ϩ ion displaces one viable cofactor ion to inhibit reaction. However, the origins of the most informative feature of the inhibition, a two-Ca 2ϩ ion inhibitory form, required further investigation.
Assuming two inhibitory binding sites and competitive inhibition, where calcium ions compete for the same sites as viable metal cofactors (M 2ϩ ), potentially eight enzyme mono or -dimetal species can exist when Ca 2ϩ and/or M 2ϩ are present, and so speciation could be very complex. This is particularly so when the inhibitory ions are bound with higher affinity by the protein than M 2ϩ and especially in the context of reactions where M 2ϩ is not supplied at saturating concentrations. The simplest responses are predicted to arise when M 2ϩ is bound with higher affinity than the inhibitory ion, a criteria fulfilled by the Mn 2ϩ /Ca 2ϩ pair. Hence to reveal whether Mn 2ϩ -sup-ported FEN-catalyzed reactions are inhibited by displacement of two viable metal cofactor ions by calcium ions, reactions were studied at three concentrations of viable cofactor ions (0.1, 0.2, and 0.4 mM) with saturating HP5F substrate and varying concentrations of Ca 2ϩ . The reactions were studied at pH 7.5, where the Mn 2ϩ -supported FEN reaction is optimal but where formation of manganese hydroxides is avoided. These three concentrations of Mn 2ϩ span subsaturating (0.1 mM) and maximal rate concentrations (0.2 and 0.4 mM) of viable metal cofactor ion. As the concentrations of Mn 2ϩ and Ca 2ϩ were varied, the ionic strength of reaction mixtures was kept constant by adjusting the amount of potassium chloride present. To avoid the concentration of ions being perturbed by binding to DNA, the substrate was supplied so that the total concentration of phosphate diesters in HP5F was equal to or less than 5% of the concentration of Mn 2ϩ .
As observed previously (3), but shown here with a more extensive data set, at lower concentrations of Ca 2ϩ (up to 80% inhibition, k obs /k 0 ϭ 0.2), the data were fitted acceptably to the simple competitive inhibition scheme (Equation 2). As the concentration of Mn 2ϩ was increased, K I was also increased, indicating that inhibition is competitive so that K I ϭ 1.1 Ϯ 0.1 mM at 0.1 mM Mn 2ϩ , K I ϭ 2.2 Ϯ 0.3 mM at 0.2 mM Mn 2ϩ , and K I ϭ 3.8 Ϯ 0.4 mM at 0.4 mM Mn 2ϩ (Fig. 3A). The magnitude of K I confirms that in analogy to dissociation constants determined for the substrate-free enzyme, Mn 2ϩ ions must bind to the protein with higher affinity than Ca 2ϩ ions.
In the case of the highest concentration of Mn 2ϩ , most of the data range fits acceptably to the simple inhibition model, although deviation is evident at the very highest concentrations of Ca 2ϩ . At lower concentrations of Mn 2ϩ , the data deviate from this simple inhibition scheme when inhibition is greater than 80% (k obs /k 0 Ͻ 0.2). The slopes of log-log plots of normalized rates versus calcium ion concentration where k obs /k 0 Ͻ 0.2 are Ϫ2.0 Ϯ 0.1 (0.1 mM Mn 2ϩ ) and Ϫ1.9 Ϯ 0.1 (0.2 mM Mn 2ϩ ) rather than the slope of Ϫ1 predicted by the simple inhibition scheme. This indicates that a two-Ca 2ϩ form of the enzyme is also catalytically inactive. Although a complete data fit to the complex overall inhibition scenario requires a knowledge of all eight interrelated dissociation constants, it is possible to generate a relatively simple description that assumes two Ca 2ϩ ions bind competitively and independently and that all forms of the enzyme that contain Ca 2ϩ are inactive. Equation 3 models the expected variation of rate k obs /k 0 with [Ca 2ϩ ], defining two apparent K I values for Ca 2ϩ ions, where K I1 is an apparent dissociation of dicalcium forms of the enzyme and K I2 is an apparent dissociation constant of monocalcium forms.
Although good non-linear regression curve fits can be obtained using Equation 3, the similar magnitudes of the apparent inhibition constants result in the same value for each with large errors. Similar non-linear regression curve fitting problems have been noted previously (32). When K I2 is fixed according to the single-site inhibition model, resulting values of K I1 ϭ 5.3 mM (0.1 mM Mn 2ϩ ) and K I1 ϭ 8.4 mM (0.2 mM Mn 2ϩ ) are observed. At higher Mn 2ϩ , the data fit to a simple one-site inhibition model, implying that the value of K I1 is outside the data range (Fig. 3B). Hence the magnitude of this parameter is dependent on the concentration of viable metal cofactor ions present, implying that two Ca 2ϩ ions compete for the same sites as the two Mn 2ϩ ions.
Further evidence for a requirement for two catalytic metals can be derived by an alternative analysis of the data. Assuming that the only catalytically competent species has two viable metal ion cofactors bound, it follows that when the dominant species contains two inhibitory ions, the dependence of the normalized rate of reaction on viable cofactor metal ions will approach the square of its concentration. This will occur in situations where the viable metal cofactor ion has higher affinity than the inhibitory ion. Analyzing the data in the region k obs /k 0 Ͻ 0.2, a plot of log k obs /k 0 versus log [Mn 2ϩ ] at fixed Ca 2ϩ concentration does indeed display the expected slope close to 2 (Fig. 3C).
Similar analyses have also been conducted for Mg 2ϩ -supported T5FEN reactions, which also display a biphasic response to Ca 2ϩ inhibition. These studies were carried out at pH 9.3, which is optimal for the Mg 2ϩ -supported reaction, and at saturating substrate. Two subsaturating concentrations of Mg 2ϩ were studied (0.5 and 2 mM), a criteria imposed to keep total divalent ion concentration below a value that could potentially alter enzyme-substrate affinity. Analysis according to the onesite inhibition model (Equation 2) reveals K I ϭ 0.11 Ϯ 0.01 mM at 0.5 mM Mg 2ϩ and K I ϭ 0.22 Ϯ 0.01 mM at 2 mM Mg 2ϩ (Fig.  3D). The apparent K I values are smaller than the concentration of Mg 2ϩ , indicating that Mg 2ϩ ions are bound more weakly than Ca 2ϩ . Applying the two-inhibitory ion model (Equation 3) yields apparent inhibition constants that are similar, K I1 ϭ 8.7 Ϯ 0.1 mM (0.5 mM Mg 2ϩ ) and K I1 ϭ 9.0 Ϯ 1.3 mM (2 mM Mg 2ϩ ), presumably reflecting the relative affinity of Ca 2ϩ versus Mg 2ϩ and the subsaturating conditions with respect to viable metal cofactor (Fig. 3E).
The Magnesium Ion Requirements of a Metal Site 2 Mutant (D201I/D204S)-Previous analyses, including the observation that D201I/D204S mutations do not drastically impair reaction, suggest that the T5FEN M2 may not play a direct role in catalysis (12,24). In the designed D201I/D204S mutant, two of the carboxylates that bind M2 are altered (Fig. 1A) to abolish the binding site for this ion as observed in E. coli ExoIX. With HP5F as substrate, the D201I/D204S mutation slowed the maximal single turnover rate of the reaction by a factor of 40 at saturating Mg 2ϩ , a modest factor in the context of the entire T5FEN rate acceleration of 10 16-10 17 (33). To elucidate the role and minimal number of metal ions involved in the D201I/D204S mutant catalyzed reaction, kinetic parameters were determined as a function of magnesium ion concentration in an analogous fashion to experiments previously conducted on the WT enzyme using HP5F substrate (pH 9.3) and at constant ionic strength. Like WT T5FEN, the major site of reaction of D201I/D204S with HP5F is one nucleotide into the double-stranded region of the 5Ј-overhanging hairpin substrate, although slightly increased amounts of reaction elsewhere are observed with the mutant protein. Total concentrations of products were evaluated during all kinetic analyses (25).
The variations of the individual kinetic parameters of D201I/ D204S mutant with magnesium ion concentration are shown in Fig. 4, A-C. The turnover number (k cat ) is dependent upon increasing magnesium ion concentration until it reaches 10 mM Mg 2ϩ . Data could not be collected below 0.05 mM Mg 2ϩ as the end points of the slowed reactions were Ͻ100% at these concentrations. A similar effect was previously noted for the WT enzyme, albeit at lower [Mg 2ϩ ] (3).
A previous study of the k cat or k ST response of WT T5FEN displayed a clear biphasic response to Mg 2ϩ ions, where a second order dependence on Mg 2ϩ at low concentrations of ions becomes first order at higher concentrations before approaching saturation (3). This situation can only be explained by the participation of two ions in chemical catalysis and was best fit to a model where reaction required the presence of two ions that were bound independently and with differing affinity (Equation 4).
However, the k cat data for D201I/D204S fit acceptably to a single binding isotherm (Equation 5), where (k cat ) obs is the observed turnover number at a given concentration of magnesium ions, K D Mg 2ϩ ES is the magnesium ion dissociation constant from ES complex, and (k cat ) max is the maximal turnover rate at infinite magnesium ions, giving K DMgES ϭ 2.5 Ϯ 0.3 mM. This value is similar to that obtained for the lower affinity magnesium ion associated with the k cat response of the WT protein (K D2MgWTES ϭ 2.1 mM (3)).
Applying a model where the reaction requires the presence of two ions that are bound independently, as used to fit WT data previously (Equation 4), places a higher affinity dissociation constant outside the data range, although this number can only be regarded as a potential upper limit (Fig. 4A). Earlier studies of the WT protein revealed that both 1/K m and k cat /K m are Mg 2ϩ -dependent. Below 0.25 mM Mg 2ϩ , 1/K m has a [Mg 2ϩ ] 2 dependence (slope of 2 log-log plot). This required a two-ion model for the WT enzyme, where two ions were bound independently with similar affinity or cooperatively. The magnesium ion response of 1/K m for D201I/D204S differs from that of the WT protein and fits a single binding isotherm (Equation 6), yielding K DMgE ϭ 0.5 Ϯ 0.2 mM (Fig. 4B).
Below 0.25 mM, k cat /K m for the WT enzyme was dependent on [Mg 2ϩ ] 3 , requiring a three-ion model. In contrast, a plot of log (k cat /K m ) versus log [Mg 2ϩ ] for D201I/D204S (Fig. 4C) reveals a slope of 1.8 Ϯ 0.2, implying the overall involvement of two ions in the reaction catalyzed by the mutated protein, rather than the three-ion requirement of the WT enzyme. Fitting the data to Equation 7, assuming that the mutated FEN is only capable of reaction with two ions present that are bound independently, with K DMgE1 ϭ 0.5 mM (as defined by the 1/K m data), yields K DMgE2 ϭ 3.3 Ϯ 0.2 mM.
Thus combined magnesium stimulation data for D201I/D204S provide evidence that the overall reaction of the mutated protein requires at least two Mg 2ϩ ions and differs from the WT protein, but the restricted data range makes it impossible to ascertain whether chemical catalysis (k cat ) has a one-or two-ion requirement.
Substrate Binding and Calcium Ion Inhibition of D201I/D204S-Binding of HP5F and D201I/D204S was studied in an analogous fashion to the WT protein at pH 9.3 to ascertain whether divalent metal ions were required to bind substrate and whether binding of Ca 2ϩ ions by DNA substrate could potentially be inhibitory. In contrast to the WT protein, in the absence of divalent metal ions, a dissociation constant could be determined (K D ϭ 2300 Ϯ 200 nM), presumably due to the substitution of two active site carboxylates with neutral residues (Fig. 2C). This was decreased by 2 orders of magnitude upon the addition of 0.5 mM Ca 2ϩ ions (K D ϭ 28 Ϯ 3 nM). The effect of varying [Ca 2ϩ ] on K D was modest; the slope of a log-log plot of K D versus [Ca 2ϩ ] is 0.09 Ϯ 0.04.
To determine the number of viable metal cofactor ions required by D201I/D204S, Ca 2ϩ inhibition of the Mg 2ϩ -supported reactions was investigated in an identical manner to the WT protein. Applying the single-site inhibition model (Equation 2) yielded K I ϭ 2.1 Ϯ 0.2 mM (2 mM Mg 2ϩ ) and K I ϭ 1.1 Ϯ 0.1 mM (0.5 mM Mg 2ϩ ) (Fig. 4D). Importantly, data obtained at 0.5 mM Mg 2ϩ clearly display 1/[Ca] 2 dependence at Ca 2ϩ ion concentrations above 6 mM, whereas increasing the concentration of Mg 2ϩ leads to deviation from the single-site inhibition model only at the very highest [Ca 2ϩ ] employed (Ͼ15 mM) (Fig.  4E). Respective values of K I1 ϭ 7.7 Ϯ 0.9 mM (0.5 mM Mg 2ϩ ) and 40 Ϯ 18 mM (2 mM Mg 2ϩ ) were obtained by fitting the data to Equation 3. As the addition of Ca 2ϩ ions at low concentration stimulates substrate binding and the stability of complexes is similar at inhibitory concentrations, the data imply that when a substantive part of the T5FEN M2-binding site is abolished, the FEN-catalyzed reaction still requires at least two catalytically viable ions.

DISCUSSION
The work described here supports the hypothesis that the major rate acceleration in the reactions catalyzed by flap endonucleases requires the presence of two metal ions. Stimulation of chemical catalysis (k cat or k ST ) by WT T5FEN has revealed previously that the reaction has a requirement for two magnesium ions that are bound independently with differing affinity (3). Confirmation of this requirement is demonstrated by the characteristics of inhibition of the WT reaction by catalytically inert Ca 2ϩ ions, shown here to be competitive with viable metal cofactor ions. The response of the FEN reaction to added Ca 2ϩ is biphasic and ultimately dependent on 1/[Ca] 2 , indicative of a requirement for two viable metal cofactors. The possibility that higher order inhibition is the result of altered T5FEN-substrate interactions as a result of added Ca 2ϩ is ruled out. Moreover, stimulation of reactions under higher Ca 2ϩ inhibitory conditions by Mn 2ϩ is a second order process with respect to stimulatory metal cofactor ion. Thus the conclusion of more than one experimental approach is that the chemical process catalyzed by T5FEN involves at least two catalytic metal ions.
A designed metal site 2 mutant catalyzes phosphate diester hydrolysis with a decrease in the rate of reaction of only 40-fold, implying that the T5FEN M2 is not responsible for major rate accelerations. Previous experiments have indicated that the overall reaction of the WT enzyme (k cat /K m ) requires at least three divalent ions. In contrast, magnesium ion stimulation data demonstrate that the overall reaction catalyzed by D201I/ D204S requires at least two ions. However, within the experimental data range tested, chemical catalysis requirements for the mutated protein (k cat ) can be assigned to a single-ion dependence. In theory, these data could indicate that a single metal ion is required for catalysis by the mutant. However, inhibition of the mutated protein by Ca 2ϩ ions clearly demonstrates a requirement for two viable metal cofactor ions, in an analogous fashion to the WT protein. The combined data therefore imply that another ion (referred to here as M3), not present in substrate-free T5FEN structures, is bound by both the WT and the mutated enzymes. Together with M1, this provides the major rate enhancement of phosphate diester hydrolysis.
Although the data presented here implicate two ions in catalysis, they do not provide information on the location of the T5FEN M3. For a two-metal ion mechanism where both ions contact the scissile phosphate diester, M3 would have to be located within Ͻ4 Å of M1, where both could be bound by the central carboxylate of the T5FEN active site (Asp-130). As this is the arrangement of two Mg 2ϩ ions shown in a substrate-free structure of hFEN ( Fig. 1B) (18) and the position of two Sm 3ϩ ions in an hFEN1 product complex (4), this positioning of ions for chemical catalysis appears highly plausible. Thus we suggest that the T5FEN M3, not visible in substrate-free structures of bacteriophage enzymes, is equivalent to the ion designated M2 in hFEN1 structures (Fig. 1, A and B).
However, despite the necessity of two ions for chemical catalysis, the WT T5FEN protein does have a three-metal ion requirement to maximize the rate of its overall (k cat /K m ) reaction (3), which is reduced to two ions for D201I/D204S. In the absence of catalysis at pH 9.3, the dissociation constant of substrate HP5F is Ͼ5 M, but the addition of Ca 2ϩ ions substantially lowers this value. Enhancement of substrate affinity is also observed with the T5FEN M2 site mutant upon the addition of Ca 2ϩ , although overall dissociation constants in the presence and absence of divalent ions are lowered as the result of the neutralizing mutations. This, together with the 1/K m divalent metal dependences of both the WT and the D201I/D204S mutant proteins, implies a role for metal ions in FEN-substrate complex stability.
Unlike the WT protein, which requires two ions to maximize 1/K m , data for the mutated FEN fit to a single binding isotherm. Therefore, the combined data suggest that the T5FEN M2 stabilizes protein-substrate interaction in WT T5FEN but is not an absolute requirement when neutralizing mutations are present in D201I/D204S.
Earlier consideration of functional and structural analyses led us to suggest that for FENs and related superfamily members to achieve a state where the scissile phosphate diester contacted metal ions involved in chemical catalysis, the terminus of the duplex where the reaction occurs would have to be unpaired (3). Structural analyses of mutant T4FEN bound to a pseudo-Y substrate in the absence of divalent ions (15) demonstrate that the scissile phosphate diester is located within a duplex bound in front of and parallel to the carboxylate-rich active site (Fig. 5,  A and B). Modeling indicated that the unpairing of two nucleotides would be required to place the scissile phosphate in the vicinity of M1 (and presumably T5FEN M3) (Fig. 5C). Recent structural analyses of DNA complexes of hFEN1 and hEXO1 lend support to this unpairing mechanism (2,4). When the 5Ј-phosphate monoester of products contacts active site metal ions, one nucleotide of the duplex is unpaired, implying that for the scissile bond to occupy the same site, two nucleotides must unpair.
To achieve contact between the scissile phosphate and the metal ions, the unpaired DNA has to traverse the active site. It is interesting to note that the phosphate diester adjacent to the scissile bond (Ϫ1) would also have to move farther toward the carboxylate-rich region, closer to the T5FEN M2, which would presumably act to stabilize this unpaired state. We suggest that in accord with the functional analyses of WT T5FEN, two metal ions are required to position the substrate in the reactive conformation: M1, which later plays a direct role in chemical catalysis, and "non-catalytic" M2. It is tempting to speculate that an M2 that plays a substrate-positioning role can be replaced with Ca 2ϩ , explaining why WT T5FEN has an overall three-metal ion requirement but is inhibited by replacement of only two viable metal cofactor ions.
In unpaired hFEN1 and hEXO1 DNA complexes, the Ϫ1 phosphate interacts with the protein N terminus (2, 4), which occupies a similar position to the T5FEN M2 (Fig. 1, A  and B). The N-terminal region of the protein is disordered in bacterial and bacteriophage FEN structures and is thought to be a site of protein-protein interaction (17,21,22). In T5FEN, the first 19 residues can be removed while retaining full enzymatic activity, and so the N terminus is unlikely to play a substrate-positioning role (12,23). Thus apparent discrepancies between metal ion-binding sites in FEN structures could be explained by a functional equivalence of a divalent metal ion and the N terminus.
The designed neutralization of Asp-201/Asp-204 to Ile-201/ Ser-204, based on the sequence of another 5Ј-nuclease family member (E. coli ExoIX (6)), presumably facilitates passage of the substrate across the same region of the protein without a requirement for a M2, yet still with the participation of M3. Individual mutations of the equivalent carboxylates (Asp-185 and Asp-188) in the FEN domain of E. coli DNA polymerase to alanine, which could be viewed as more conservative than the D201I/D204S double mutation, were both reported to have a much larger deleterious effect on enzyme activity than the designed double mutation introduced here (34). This is presumably in part because a greater active site charge remains in the individual Ala mutants, but stabilizing substrate interactions with the side chains of Ile and Ser are also a possibility in D201I/D204S. A mutated T5FEN where the same two carboxylates are changed to arginine (D201R/D204R) retains activity in line with a non-catalytic substrate-positioning/stabilizing role for the T5FEN M2, perhaps compensating for the lack of a bound metal ion with the cationic arginine (12). The decrease in the reaction rate observed with our designed M2 mutant may reflect a change in the partitioning of the substrate structure between a bound duplex form and an unpaired, active site occupying form. Such a situation would be consistent with the properties of exonucleolytic reactions catalyzed by the mutated protein (24). Although not a feature of the T5FEN-catalyzed reactions of 5Ј-flap substrates, 6 the exonucleolytic reactions of T5FEN are diffusion-controlled, indicating that forward reaction steps are faster than substrate dissociation. Despite more stable enzyme-DNA complexes with D201I/D204S, both in the presence and in the absence of divalent ions, reactions are no longer diffusion-controlled, implying a slowed later step in the reaction sequence. Furthermore this rate-limiting later step is insensitive to the nature of the leaving group, consistent with a physical rather than chemical process, such as substrate unpairing.
In contrast to the ExoIX subgroup, the carboxylate-rich active site is conserved in the majority of the DNA 5Ј-nuclease superfamily (1). These include enzymes that catalyze structure-specific reactions of an apparently disparate group of nucleic acid substrates including DNA bubbles (xeroderma pigmentosum complementation group G (XPG)) (35), four-way junctions (GEN-1 (xeroderma pigmentosum complementation group G-like gap endonuclease, a putative human Holliday junction resolvase)) (36), as well as flapped, nicked, and 3Ј-overhang DNAs (FEN and EXO1) (2,20,37,38). Because all of these enzymes carry out the same chemical transformation, like T5FEN, they too will require two ions to effect phosphate diester hydrolysis. Notably, the reaction specificities of the superfamily do have a common feature; the major site of reaction occurs one nucleotide into a double-stranded region of the substrate, located at a junction. It seems highly likely that the unpairing mechanism is universally conserved in the seven/eight-carboxylate enzymes to give rise to this reaction specificity. If this is the case, FEN superfamily members may also require either a further active site metal or a functionality of the protein such as the N terminus for substrate unpairing in addition to two catalytic ions. Analyses of the nuclear RNA degradative 5Ј-exoribonuclease Rat1 (nuclear 5Ј-3Ј-exoribonuclease; also known as Xrn2) (41) demonstrate similar secondary structure to FENs (5). In addition, we also note that two recent structures of the corresponding cytosolic enzyme Xrn1 (cytosolic 5Ј-3Ј-exoribonuclease) (39,40) demonstrate that these enzymes conserve the DNA 5Ј-nuclease superfamily active site and that RNA 5Ј-nucleases may in turn may also utilize a similar catalytic mechanism.