Three Metal Ions Participate in the Reaction Catalyzed by T5 Flap Endonuclease*

Protein nucleases and RNA enzymes depend on divalent metal ions to catalyze the rapid hydrolysis of phosphate diester linkages of nucleic acids during DNA replication, DNA repair, RNA processing, and RNA degradation. These enzymes are widely proposed to catalyze phosphate diester hydrolysis using a “two-metal-ion mechanism.” Yet, analyses of flap endonuclease (FEN) family members, which occur in all domains of life and act in DNA replication and repair, exemplify controversies regarding the classical two-metal-ion mechanism for phosphate diester hydrolysis. Whereas substrate-free structures of FENs identify two active site metal ions, their typical separation of >4 Å appears incompatible with this mechanism. To clarify the roles played by FEN metal ions, we report here a detailed evaluation of the magnesium ion response of T5FEN. Kinetic investigations reveal that overall the T5FEN-catalyzed reaction requires at least three magnesium ions, implying that an additional metal ion is bound. The presence of at least two ions bound with differing affinity is required to catalyze phosphate diester hydrolysis. Analysis of the inhibition of reactions by calcium ions is consistent with a requirement for two viable cofactors (Mg2+ or Mn2+). The apparent substrate association constant is maximized by binding two magnesium ions. This may reflect a metal-dependent unpairing of duplex substrate required to position the scissile phosphate in contact with metal ion(s). The combined results suggest that T5FEN primarily uses a two-metal-ion mechanism for chemical catalysis, but that its overall metallobiochemistry is more complex and requires three ions.

hydrolysis of the phosphate diester linkages of nucleic acids. The uncatalyzed hydrolysis of phosphate diesters under biological conditions is an extremely slow process with an estimated half-life of 30 million years at 25°C (1). Protein nucleases and RNA enzymes produce rate enhancements of 10 15 -10 17 to allow this reaction to proceed on a biologically useful time scale. Most enzymes catalyzing phosphate diester bond hydrolysis have a requirement for divalent metal ions. Based largely upon crystallographic observations, most metallonucleases are proposed to catalyze reactions using a two-metal-ion mechanism (Fig. 1a) analogous to that suggested for the phosphate monoesterase alkaline phosphatase (2,3), although this view is not universally accepted. Three recent reviews present contrasting views on the roles of metal ions in protein nuclease and RNA enzyme reactions and illustrate this controversy (4 -6).
One family of metallonucleases over which there has been considerable mechanistic debate are the flap endonucleases (FENs) 3 (7)(8)(9)(10)(11)(12), which are present in all domains of life and play a key role in DNA replication and repair. Unlike most metallonucleases, which typically possess a cluster of three or four active site carboxylates, the FEN active site is constructed from seven or eight acidic residues located in similar positions in FENs from a range of organisms (Fig. 1b, see also supplemental Fig. S1) (7,9,10,(13)(14)(15)(16). Several FEN x-ray structures also contain two active site carboxylate-liganded divalent metal ions, designated as metals 1 and 2 (9,(13)(14)(15). The position of metal 1 is similar in all cases, but the metal 2 location varies. In all but one structure of human FEN (hFEN), the spacing of these metal ions exceeds that demanded by a two-metal-ion mechanism, which needs to be Յ4 Å for the two ions to bind the same oxygen atom (Fig. 1a). Only two structures of enzyme-DNA complexes exist, but neither contain bound metal ions (8,17).
To investigate the roles played by metal ions in the FEN reaction, we have carried out a detailed evaluation of the magnesium ion response of T5FEN, and of the inhibition of magnesium-and manganese ion-supported reactions by calcium ions. Together the data suggest that the major acceleration of phosphate diester hydrolysis requires at least two viable cofactor ions, and so could utilize a two-metal-ion mechanism, but that the metallobiochemistry of T5FEN is more complex and overall involves at least three metal ions.

EXPERIMENTAL PROCEDURES
Materials-T5FEN was purified to homogeneity as described (18). HP5F substrate (5Ј-FAM-pd(CGCTGTCGAACA-CACGCTTGCGTGTGTTC)) was prepared as described (19) and after purification, divalent metal ion contaminants were removed by treatment with Chelex resin. ICP mass spectrometry was used to confirm the absence of significant divalent metal ion contamination in all materials.
Steady State Kinetic Analyses-Steady state kinetic parameters of T5FEN were evaluated at 37°C using HP5F substrate in 25 mM CHES or potassium glycinate, pH 9.3, 0.1 mg/ml bovine serum albumin, and 1 mM dithiothreitol as described (19). MgCl 2 was added to the desired concentration and the ionic strength (MgCl 2 ϩ KCl) adjusted to 80 mM using KCl. Substrate concentrations were varied around the K m , and reactions were sampled and quenched by addition of an equal volume of 25 mM EDTA at appropriate time intervals. Reactions were analyzed by dHPLC equipped with a fluorescence detector and initial rates (v) were calculated as described (19

Stimulation of FEN Catalysis by Magnesium
Ions-To elucidate the role and minimal number of metal ions involved in the T5FEN-catalyzed reaction, kinetic parameters were monitored as a function of magnesium ion concentration. As the concentration of Mg 2ϩ was varied, the ionic strength of reaction mixtures was kept constant by adjusting the amount of potassium chloride present. A fluorescent 29 nucleotide 5Ј-overhanging DNA hairpin substrate (HP5F) was employed. The characteristics of the T5FEN-catalyzed reaction of HP5F have been reported previously (19 -21). As both maximal steady state and single turnover rates of Mg 2ϩ -supported reactions reach a pH-independent plateau above pH 8.5, pH 9.3 was selected as the pH optimum for the experiments. The variations of the individual kinetic parameters with magnesium ion concentration are shown in Fig. 2, a-c, and sample individual experiments are shown as supplemental Fig. S2. The turnover number (k cat ) is magnesium ion-dependent and increases until a Mg 2ϩ -independent plateau is observed above 10 mM. Below 10 M Mg 2ϩ , reactions no longer proceed to completion, and therefore data were not collected below this concentration. At low magnesium ion concentration (10 -  100 M), k cat has a second order dependence on divalent metal ion concentration (Fig. 2a, green slope of 2 in log-log plot) that becomes first order (Fig. 2a, blue slope of 1) at cofactor concentrations greater than 100 M. This change in reaction order cannot be accounted for by the participation of a single Mg 2ϩ in the FEN-catalyzed reaction (see Equation 1). A reaction supported by a single ion would display first order dependence (Fig.  2a, slope of 1 in log-log plot) at all concentrations of magnesium ions Յ K DMgES . Instead the minimal number of metal ions that could give rise to the behavior observed in Fig. 2a is 2. Moreover the change in reaction order requires that these two metal ions are bound independently and have differing affinities, but that for a reaction to proceed, both ions must be present. A model of this situation leads to Equation 2, where (k cat ) obs is the observed turnover number at a given concentration of magnesium ions, K DMgES1 and K DMgES2 are the respective apparent magnesium ion dissociation constants from the ES complex, and k cat(max) is the maximal turnover rate at infinite magnesium ion concentration. At low concentrations of metal ions ([Mg] Ͻ K DMgES1 and K DMgES2 ), two metal ion binding sites of differing affinity need to be occupied to produce a catalytically active form of the enzyme giving rise to the slope of 2. At higher concentrations of , the higher affinity site is saturated, and the reaction becomes first order as only the low affinity site is titrated and the occupancy of both sites is necessary for reaction to proceed. Curve fitting of the data to Equation 2 yields metal ion dissociation constants of K DMgES1 ϭ 0.04 Ϯ 0.003 mM and K DMgES2 ϭ 2.25 Ϯ 0.06 mM.
Product release is partially rate-limiting for T5FEN-catalyzed hydrolysis of HP5F (22). Thus, the rate constants under single turnover conditions (k ST ) were also monitored as a function of magnesium concentration, and are 2-3-fold greater than k cat in all cases (supplemental Fig. S3). The response to magnesium ion concentration is the same as for the k cat data and could be fitted to Equation 2 to give K DMgES1 and K DMgES2 values of 0.09 Ϯ 0.01 mM and 2.8 Ϯ 0.35 mM, respectively. These values are in good agreement with those obtained from the k cat data, and so the parameters obtained from Fig. 2a apply to the magnesium-dependent events prior to product release.
HP5F binds to T5FEN weakly in the absence of divalent metal ions (K D Ͼ 15 M). Substrate/product binding equilibria, as reflected by 1/K m , are magnesium ion-dependent and K m reaches a minimum (0.04 Ϯ 0.01 M) above 0.25 mM Mg 2ϩ (Fig.  2b). In the Mg 2ϩ -dependent range, the slope of the log-log plot is 1.8 Ϯ 0.2, suggesting a second order dependence. This magnesium-dependent behavior can only be explained by the involvement of at least two metal ions, but could potentially arise from two different models. In one model, two metal ions bind independently, and the presence of both is required to stabilize one or more enzyme-substrate state(s). However, to give rise to the slope close to 2 over the whole Mg 2ϩ -dependent range, both metal ions must bind with virtually identical binding constants. Applying this model to the data (Equation 3, where (1/K m ) obs is the observed parameter at a given magnesium ion concentration, (1/K m ) max is the maximal reciprocal of the concentration of substrate required to convert half the enzyme into the form(s) preceeding the rate-limiting step(s) at infinite magnesium ion concentration, and K DMgE1 and K DMgE2 are the respective apparent magnesium ion dissociation constants from free enzyme, with K DMgE1 and K DMgE2 kept identical) yields values of K DMgE1 (ϭ K DMgE2 ) ϭ 0.21 Ϯ 0.02 mM (Fig. 2b).
In an alternative model, the metal ions bind cooperatively to the enzyme. A curve fit to this scheme (Equation 4, n ϭ 2, where K DMgE 2 is the product of 2 apparent magnesium ion dissociation constants from free enzyme) yields K DMgE 2 ϭ 0.07 Ϯ 0.01 mM 2 . The difference between independent and cooperative behavior is only evident as the curves approach saturation, where an independent binding model produces a more gradual change to metal ion-independent behavior (see Fig. 2b). Distinguishing between these two models is generally difficult and especially so for this data set because K m measurements for nucleases derived from discontinuous assays are typically associated with errors of Ϯ30%.
The magnesium ion dependence of the overall FEN reaction (k cat /K m ) is shown in Fig. 2c. The apparent second order rate constant can be obtained in two ways, either from the individual catalytic parameters reported above, or as the gradient of a plot of v/[E] versus [S] at concentrations of [S] Ͻ Ͻ K m . Both of these approaches produce identical data within experimental error (Fig. 2c). In the latter case, the total concentration of phosphate diesters in HP5F was equal to or lower than 10% of the concentration of Mg 2ϩ , ruling out the possibility that Mg 2ϩ binding to HP5F is affecting the cofactor dependence. The apparent second order rate constant increases with magnesium concentration, reaching a Mg 2ϩ -independent plateau above 2 mM. Importantly, the slope of the log-log plot at low Mg 2ϩ concentrations is 2.9 Ϯ 0.2, requiring the participation of at least three metal ions. As with the 1/K m data, either independent binding or cooperative models could explain the behavior. Furthermore, as this is at least a three-ion process, combinations of independent and cooperative models are plausible. All of these models fit the data reasonably. To maintain consistency with the data derived from the individual catalytic parameters, a combination of cooperative and independent binding models (Equation 5) with K DMgE2 2 ϭ 0.07 mM 2 produces a convincing data fit and a value for the dissociation constant of the with K DMgE1 ϭ K DMgE2 ϭ 0.21 mM), K DMgE3 ϭ 6.1 Ϯ 1.8 mM.
One caveat on the interpretation of this and any metal ion titration data is that the experiments only report on the presence and affinity of magnesium ions in the tested data range. If one or more metal ions are tightly bound over the entire range allowing data collection, then the total number of metal ions required to support the overall catalytic process is 4 or more.
Inhibition of FEN Catalysis by Calcium Ions-When supplied as the sole cofactor, calcium ions do not support FEN catalysis. However, as is typical for other divalent metal ion-dependent nucleases, Ca 2ϩ ions have been demonstrated to stabilize protein-DNA interactions in the absence of catalytically viable cofactors (12). To investigate the effect of added calcium ions on the Mg 2ϩ -supported FEN reaction, the rate of reaction was measured as a function of Ca 2ϩ concentration at pH 9.3 in the presence of subsaturating Mg 2ϩ , but with saturating HP5F. The effects of addition of calcium ions on the Mn 2ϩ -supported FEN reaction were also investigated. For Mn 2ϩ -supported reactions, the pH was reduced to 7.5 to avoid the problems associated with the formation of manganese hydroxides at higher pH values. This pH value is still above the measured pK a for the Mn 2ϩsupported FEN-catalyzed reaction of HP5F (21). All reactions were inhibited by addition of Ca 2ϩ (Fig. 3). To evaluate the apparent K i at a given concentration of cofactor, data were fitted to a simple competitive inhibition model (Equation 7) where k obs is the normalized initial rate (v/[E]) at a given concentration of calcium ions, k 0 is v/[E] in the absence of calcium ions, and K i is equal to the concentration of calcium ions that produces 50% inhibition at the respective concentration of viable divalent metal ion cofactor. In all cases, the data fit acceptably to a simple competitive inhibition model up to 80% inhibition (k obs /k 0 ϭ 0.2) (Fig. 3a). At 2 mM Mg 2ϩ K i ϭ 0.22 Ϯ 0.02 mM whereas decreasing the concentration of Mg 2ϩ to 0.1 mM decreased K i ϭ 0.037 Ϯ 0.002 mM, indicating that Ca 2ϩ inhibition is competitive with Mg 2ϩ . At 0.1 mM Mn 2ϩ the K i ϭ 1.2 Ϯ 0.1 mM. The 30-fold higher concentration of Ca 2ϩ ions required to inhibit the Mn 2ϩ reaction by a factor of 2 may in part be a consequence of the greater affinity of the FEN protein for Mn 2ϩ versus Mg 2ϩ (12,21). However, by necessity Mn 2ϩ and Mg 2ϩ experiments had to be carried out at different pH values, and pH-dependent effects cannot be ruled out. When Mn 2ϩ was supplied as the cofactor for the EcoRVcatalyzed reaction of plasmid substrates, addition of calcium ions stimulated or first stimulated then inhibited reactions (23). This was explained by the formation of a mixed enzyme Mn 2ϩ Ca 2ϩ species that was more reactive than an Mn 2ϩ Mn 2ϩ form. However, a detailed study at low concentrations of Ca 2ϩ revealed that no stimulation of activity was observed in the Mn 2ϩ -supported FEN reaction (Fig. 2b).
The observation that the overall FEN reaction requires the presence of at least three magnesium ions (Fig. 2c) suggests that a competitive inhibition scheme based on binding one Ca 2ϩ ion (Equation 7) could be too simplistic as the calcium ions could replace more than one of the viable cofactor ions. Evidence that this is the case is observed at higher concentrations of inhibitory ions, where the Ca 2ϩ dependence deviates from that predicted from a simple competitive inhibition model (Equation 7). In these regions, the slopes of log-log plots are Ϫ2.0 Ϯ 0.1 (0.1 mM Mn 2ϩ ) and Ϫ2.1 Ϯ 0.2 (2 mM Mg 2ϩ ) rather than the slope of Ϫ1 predicted by the simple competitive inhibition model (Fig. 3, b and c). Thus the simultaneous binding of two calcium ions inhibits the FEN reaction. However, a model where the presence of two calcium ions is the sole inhibitory form of the enzyme-substrate complex is also inadequate as this would give rise to a slope of Ϫ2 over the entire Ca 2ϩ -dependent data range. Assuming that Ca 2ϩ ions inhibit the reaction by displacing catalytically competent cofactors, the FEN protein must possess at least two viable cofactor ions at specific sites, and replacement of either or both of these by Ca 2ϩ significantly reduces activity. However, it is possible that one of the three FEN metal ions can be replaced by Ca 2ϩ without altering the ability to catalyze the reaction. A similar biphasic response to inhibitory Ca 2ϩ ions has been noted for the magnesium-supported reactions of the Tetrahymena ribozyme (24).

DISCUSSION
The data presented here demonstrate that the overall T5FEN-catalyzed reaction requires at least three magnesium ions. However, most FEN x-ray structures (in the absence of substrate) show only two divalent metal ions bound (Fig. 1b), consistent with ITC studies of divalent metal ion binding to T5FEN (12). This implies that a third metal ion binds in the presence of substrate and has a lower association constant for the free enzyme. There are several structural precedents for the binding of additional active site metal ions by nucleases in the presence of substrate or products. A recent example is RNase HI where x-ray crystallography revealed the presence of a single magnesium ion in the absence of substrate (25), but an inhibited enzyme-nucleic complex contained two active site magnesium ions (26). However, debate has focused on the functional relevance of the metal ions within co-crystal structures and whether these additional metal ions are required for catalysis of phosphate diester hydrolysis.
For the T5FEN reaction, the magnesium ion response of the maximal FEN reaction rate (k cat or k ST ) requires at least two independently bound magnesium ions. The calcium ion inhibition data also provide evidence for two inhibitory metal ion binding sites. This is most simply explained by a competitive inhibition scheme where replacement of either or both of the catalytically essential viable cofactors leads to inactivation. The finding that at least two viable cofactor ions are required to render the FEN protein catalytically competent makes it plausible that FEN derives its major rate acceleration using a twometal-ion mechanism. This mechanism involves two metal ions interacting with the same oxygen atom of the scissile phosphate diester and requires they must have a separation of Յ4 Å (Fig. 1a). The separation of the two metal ions (designated metal 1 and 2) observed in the T5FEN structure (8 Å) is much greater than this (Fig. 1b). However, in one hFEN structure two metal ions are liganded with a separation of 3.4 Å (Fig. 1b and  supplemental Fig. S1) (15) to the same central carboxylate of the FEN active site, in an analogous fashion to the metal ion coordination observed in other nucleases (3,26). The third T5FEN metal ion (metal 3), implicated by this study, could bind to Asp-130 in close proximity to metal 1 in an analogous fashion to the metal coordination observed in hFEN, and support a twometal-ion mechanism for chemical catalysis. Mutation of Asp-130, or its equivalent in other FENs, leads to undetectable activity, supporting a key role for this Asp-130 site in metal ion coordination (12,27,28). In contrast mutation of metal 2 coordinating carboxylates in T5FEN produces readily detectable activity (12), suggestive metal 2 does not play a vital role in catalysis. Superposition of hFEN and T5FEN structures demonstrate that the third metal ion implicated by this study could be accommodated within the T5FEN active site bound to Asp-130 to support a two metal ion mechanism (Fig. 1b). However, the possibility remains that metal ions not involved in direct scissile bond interactions could contribute substantive rate acceleration by transition state stabilization (29) and the location of the third metal ion required by the T5FEN reaction merits further investigation.
Metal ions involved in the overall FEN catalytic process could potentially play roles in substrate binding equilibria and/or chemical catalysis. Therefore, this study provides evidence that between one and three ions could participate in FEN substrate equilibria as the overall FEN reaction has a three metal ion requirement. The magnesium ion dependence of 1/K m can only be explained by the participation of at least two ions. For enzyme reactions rate limited by steps other than the chemical reaction under steady state conditions, K m values define the concentration of substrate required to convert half the enzyme into the form(s) proceeding the rate-limiting step(s) (30). As product release is partially rate-limiting for the WT T5FEN-catalyzed reaction of HP5F (k ST ϭ 2.3 k cat ), the stability of enzyme-product species is measured by K m . Furthermore, the stability of any intermediates formed after initial interaction of enzyme and substrate would also be reflected in this parameter. In the absence of direct interrogation of all of these enzyme-substrate and enzyme-product equilibria, any interpretation of the metal ion dependence of 1/K m is tentative, but the data suggest a role for metal ions in stabilizing T5FEN-substrate equilibria. Support for a requirement for metal ion(s) for productive substrate binding is provided by a recent bacteriophage T4FEN structure solved in complex with pseudo-Y DNA (8). This structure lacks divalent metal ions and has a mutation of metal 1-liganding central carboxylate (D132N). The DNA substrate makes no contacts with the FEN active site lacking metal ions, presumably due to the electrostatic repulsion by the carboxylates. Thus, bound metal ion(s) are probably required to stabilize a FEN-DNA complex where substrate is positioned for reaction to occur.
An intriguing feature of the T4FEN DNA structure is that the scissile phosphate corresponding to the major site of FEN reaction, one nucleotide into the duplex region at the site of bifurcation, is located within duplex DNA bound parallel to the active site. Positioning metal 1 within the T4-DNA complex according to a T4FEN structure without substrate bound (9) placed this ion 7 Å away from this scissile phosphate (8) (supplemental Fig. S4b). Metal 2 is at a much greater distance from the major reaction site (8). Nevertheless, a model of T5FEN-DNA interaction, based upon overlay with the T4-DNA structure, is in excellent agreement with experimental data on the interactions of the protein with the duplex and the 3Ј-overhang of the pseudo-Y (supplemental Fig. S4a) (19,31). This makes it difficult to envisage how the scissile phosphate, one nucleotide into the duplex region, can bind within the active site while still maintaining a Watson-Crick terminal base pair and the functional substrate-protein contacts. However, breaking the terminal base pair and making at least the first nucleotide and reaction site extrahelical would allow active site binding while maintaining substrate-protein contacts. The metal dependence of substrate equilibria observed here may be the consequence of a metal dependent conformational change, analogous to the nucleotide flipping seen in base excision repair enzymes (32), which positions the scissile bond within the active site (supplemental Fig. S4c).
The reactions of all FENs may have an overall three metal ion requirement. However, in archaeal and hFEN structures (10, 14 -17) the protein N terminus occupies a position similar to metal ion 2 in phage enzymes (supplemental Fig. S1). In bacteriophage and bacterial FEN structures, this region is disordered (7,9,13). Bacteriophage FENs lacking the N-terminal regions possess wild type activity (12,33), whereas the functional relevance of the N-terminal region of archaeal and mammalian FENs is untested. It is possible that the N terminus might substitute for metal ion 2 in these enzymes. While the role played by metal 2 deserves further investigation it could act in stabilization of FEN-substrate equilibria and/or may play a non-essential but rate accelerating role in chemical catalysis. It is intriguing to note that the same function may be provided by the protein in higher organism FENs.
Other phosphoryl transferases have three metal ion active sites. For phosphate diesterases that contain three active site zinc ions, such as nuclease P1 (34), phospholipase C (35,36), and endonuclease IV (37,38), all three metals are suggested to act in chemical catalysis, but different mechanisms involving these three metal ions have been proposed. Several magnesiumdependent nucleases with three metal ion binding sites may also exist, such as the group I intron (39), EcoRV (40), and 3Ј-5Ј exonuclease (41). Even the prototype two-metal-ion mechanism phosphoryl transferase, the phosphate monoesterase alkaline phosphatase, has an active site composed of two zinc ions plus one magnesium ion, and magnesium ions stimulate activity (2). Thus our new results on T5FEN in the context of existing data argue that three metal ion active sites may be a more common phenomena than previously suspected. Additional metal ions could influence biological efficacy by either moderating chemical reactivity or physical steps of the catalytic cycle such as substrate association, conformational changes, and enzyme-product stability. Thus, while a mechanism can be envisaged for rate accelerations of phosphate monoesterases and diesterases involving two metal ions, the biological function of other phosphoryl transferases in DNA replication, DNA repair, RNA processing, and RNA degradation may well involve a third ion or its functional equivalent.