Distinct Roles of the Active-site Mg2+ Ligands, Asp882 and Asp705, of DNA Polymerase I (Klenow Fragment) during the Prechemistry Conformational Transitions*

DNA polymerases catalyze the incorporation of deoxynucleoside triphosphates into a growing DNA chain using a pair of Mg2+ ions, coordinated at the active site by two invariant aspartates, whose removal by mutation typically reduces the polymerase activity to barely detectable levels. Using two stopped-flow fluorescence assays that we developed previously, we have investigated the role of the carboxylate ligands, Asp705 and Asp882, of DNA polymerase I (Klenow fragment) in the early prechemistry steps that prepare the active site for catalysis. We find that neither carboxylate is required for an early conformational transition, reported by a 2-aminopurine probe, that takes place in the open ternary complex after binding of the complementary dNTP. However, the subsequent fingers-closing step requires Asp882; this step converts the open ternary complex into the closed conformation, creating the active-site geometry required for catalysis. Crystal structures indicate that the Asp882 position changes very little during fingers-closing; this side chain may therefore serve as an anchor point to receive the dNTP-associated metal ion as the nucleotide is delivered into the active site. The Asp705 carboxylate is not required until after the fingers-closing step, and we suggest that its role is to facilitate the entry of the second Mg2+ into the active site. The two early prechemistry steps that we have studied take place normally at very low Mg2+ concentrations, although higher concentrations are needed for covalent nucleotide addition, consistent with the second metal ion entering the ternary complex after fingers-closing.

tionally, both metal ions can stabilize negative charge in the pentacovalent intermediate or transition state. It is reasonable to assume that metal B enters the polymerase complex along with the dNTP; however, it is currently unclear when metal A becomes bound. Experiments with DNA polymerase ␤ and RB69 DNA polymerase, using exchange-inert dNTP complexes with Cr(III) or Rh(III), show that some early steps of the polymerase reaction can indeed take place when the only metal ion present is in the metal B position, coordinated to the dNTP phosphates (10 -13). The subsequent addition of a second metal ion to fill the metal A site is required before phosphoryl transfer can take place.
Crystal structures of a large number of nucleic acid polymerases complexed with DNA and nucleotide show the two catalytic metal ions bound at the polymerase active site by two carboxylate side chains that are invariant throughout the entire polymerase superfamily (8,14,15). In Pol I(KF), an Afamily DNA polymerase, the carboxylate ligands, located in the palm subdomain, are Asp 705 and Asp 882 (Fig. 2). Substitution of either of these side chains by alanine reduced polymerase activity to almost undetectable levels (16,17). Although the mutational studies clearly established that Asp 705 and Asp 882 are essential for nucleotide incorporation, they did not indicate which step(s) of the reaction pathway is affected. In this study, we have investigated the D705A and D882A Pol I(KF) mutants using the two fluorescence assays described above, which report directly on early conformational transitions in the polymerase reaction. The results indicated which prechemistry steps are dependent on the aspartates and, by implication, on their coordinated metal ions, and suggested a sequence of events that assembles the polymerase active site for catalysis of phosphoryl transfer.
Expression and Purification of Pol I(KF) and Mutant Derivatives-The Pol I(KF) construct used for fluorophore labeling has been described previously (3). It has an N-terminal hexahistidine tag for purification, the D424A mutation to inactivate the 3Ј-5Ј exonuclease, the C907S mutation to remove the single native cysteine, and the L744C mutation to provide a unique labeling site on the fingers subdomain. For simplicity, the N-His 6 , D424A, C907S, L744C genotype will be referred to as wild type (WT) throughout. The D705A and D882A mutations were introduced into the Pol I(KF) construct described above, using the QuikChange site-directed 2) represent the open and closed conformations of the polymerase. E C to E C ‡ (step 3) represents a further transformation, whose identity is not known, corresponding to the rate-limiting step in single nucleotide incorporation. The existence of additional, as yet undetected, steps is not ruled out.  (6) were aligned using PyMOL (DeLano Scientific LLC). Both complexes were essentially identical except for the mobile portion of the fingers subdomain (residues 680 -714, equivalent to 732-766 of Pol I(KF), shown in blue in the binary and gray in the ternary complex; the non-mobile portion of the protein (from 1L3U) is shown in cyan. The residues Asp 705 and Asp 882 of Pol I(KF), used in our study, correspond to the residues Asp 653 and Asp 830 of Bst pol, colored in yellow and green, respectively. The red spheres indicate the attachment point for the donor fluorophore on the protein (residue 692 of Bst pol, equivalent to residue 744 of Pol I(KF)). The DNA primer-template is shown in orange, with the attachment position of the dabcyl quencher on the template strand shown as a black sphere. mutagenesis kit (Stratagene), according to the manufacturer's instructions. Pol I(KF) derivatives were expressed and purified to homogeneity as described previously (3,18).
Labeling of Pol I(KF) Derivatives with IAEDANS-The proteins carrying a single cysteine at position 744 were labeled with a 2-3-fold molar excess of IAEDANS as described previously (3). Excess fluorophore was removed by gel filtration on a Bio-Spin 30 column. Labeled proteins were stored at Ϫ20°C in 50 mM Tris-HCl, pH 7.5, 1 mM DTT, 40% (v/v) glycerol. The extent of labeling, calculated from the UV spectrum, was typically Ն70%.
Chemical Quench Experiments-Single-turnover measurements of nucleotide incorporation were carried out at room temperature (22°C) in a rapid quench-flow instrument (Kin-Tek Corp., model RQF-3) for fast reactions or with manual quenching when the reaction was slow. The DNA substrate was the linear duplex L:unmod:3ЈOH (Fig. 3A), consisting of a 13-mer primer, 5Ј-labeled with 32 P, annealed to a 1.5-fold molar excess of the complementary 19-mer, with A as the templating base. The enzyme-DNA solution contained 0.5 M polymerase and 0.1 or 0.05 M primer-template duplex. The reaction was initiated by mixing the enzyme-DNA solution with an equal volume of a dNTP solution. The standard reaction buffer was 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, and 10 mM MgCl 2 . In some experiments, the concentration of MgCl 2 or MnCl 2 was varied, as indicated. Reactions were quenched at appropriate time intervals using excess EDTA and were fractionated on denaturating polyacrylamide-urea gels and quantitated on a Fuji FLA 5100 scanner.
Fluorescence Emission Spectra-Steady-state fluorescence spectra were recorded at 22°C using a Photon Technology International scanning spectrofluorometer. The solutions for 2-AP fluorescence measurements contained 1 M L:T(ϩ1)2-AP:3ЈH duplex DNA ( Fig. 3B; formed by annealing the T(ϩ1)2-AP template strand with a 1.5-fold molar excess of the dideoxy-terminated complementary primer strand) and 2 M wild-type or 4 M mutant Pol I(KF) in 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 10 mM MgCl 2 , and 0.1 mg/ml BSA. The polymerase/DNA ratio was selected as that which gave the maximum increase in the fluorescence of the 2-AP DNA duplex. The complementary nucleotide, dGTP, was added to a final concentration of 200 M dGTP for wild-type Pol I(KF) or 1 mM for the mutants. Samples were excited at 310 nm, and emission spectra were scanned from 330 to 460 nm. Spectra were corrected by subtraction of the fluorescence emission contributed by the identical concentration of Pol I(KF) in buffer and were adjusted to account for the dilution due to the addition of the dNTP solution.
Emission spectra of AEDANS-labeled Pol I(KF) complexes with the dabcyl-modified H:T(Ϫ8)D:3ЈH DNA (Fig. 3C) were recorded using solutions containing 1 M labeled polymerase and 2 M DNA (for wild type) or 4 M DNA (for the mutants) in 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 10 mM MgCl 2 , and 0.1 mg/ml BSA. The polymerase/DNA ratio was chosen so as to obtain the maximum change in the fluorescence of the AEDANS-labeled protein. Ternary complexes were formed by adding the complementary nucleotide, dTTP, to a final concentration of 200 M for wild-type Pol I(KF) or 1 mM for the mutants. AEDANS-labeled samples were excited at 350 nm, and emission spectra were collected from 360 to 650 nm. All spectra were corrected by subtraction of the buffer background and for the dilution factor.
Stopped-flow Fluorescence-Stopped-flow experiments were performed at 22°C using an Applied Photophysics SX.18MV spectrofluorometer. For measurements using the T(ϩ1) 2-AP reporter, one drive syringe contained a solution of L:T(ϩ1)2-AP:3ЈH (Fig. 3B) Fig. 7), giving 0.25-5.8 M free Mg 2ϩ and 3.1-67 M Mg-dGTP, as described by Bakhtina et al. (12). The excitation wavelength for the 2-AP probe was 313 nm, and fluorescence emission was detected using a 345-nm long pass filter. . Oligonucleotides used in this study. A, DNA substrate used in chemical quench experiments. The 13-mer primer, 5Ј-labeled with 32 P (indicated as an asterisk), was annealed to the 19-mer template. B, DNA duplex oligonucleotide used for 2-AP fluorescence measurements. The 2-AP fluorophore (X) was placed 5Ј to the templating base, designated as the T(ϩ1) position (relative to the templating position as 0 in our numbering system). C, hairpin DNA substrate used in FRET-based fluorescence measurements of fingers-closing. The dabcyl-dT quencher was placed at the T(Ϫ8) position. D, as in C but without the dabcyl quencher, used as a control for the FRETbased measurements. Except for A, the primer strand was dideoxy-terminated (3ЈH) so as to prevent the reaction from proceeding beyond the ternary complex. In some experiments, extendable (3ЈOH) versions of B-D were also used. All DNA substrates employed essentially the same duplex DNA sequence. The observed rates of fluorescence changes or of dNTP incorporation in our assays are altered very little by changing the templating base (2, 3). . The buffer in both syringes was 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, and 5 mM MgCl 2 (for wild type) or 10 mM MgCl 2 (for the mutants). In some experiments, Mn 2ϩ replaced Mg 2ϩ , at 2 mM MnCl 2 (for wild type) or 10 mM MnCl 2 (for the mutants). Experiments at low MgCl 2 concentrations were carried out as described above, using the extendable H:T(Ϫ8)D:3ЈOH DNA and adding MgCl 2 only to the drive syringe containing dTTP (at 2 mM).
To measure DNA dissociation from a binary complex with D882A Pol I(KF), one drive syringe contained 0.1 M AEDANSlabeled polymerase with 0.2 M of dabcyl-modified H:T(Ϫ8)D:3ЈOH DNA (Fig. 3C), and the other contained a 2 M concentration of an unmodified DNA duplex. The buffer in both syringes was 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, and 10 mM MgCl 2 . Upon mixing the two solutions, the excess unmodified DNA acts as a trap preventing the free labeled polymerase from rebinding to dabcyl-modified DNA molecules. The resulting increase in protein fluorescence reports DNA dissociation. To measure DNA dissociation from the Pol-DNA-dNTP ternary complex, the DNA trap solution contained 2 mM dNTP.
In all stopped-flow experiments using the AEDANS reporter, the excitation wavelength was 350 nm, and fluorescence emission was detected with a 400-nm long pass filter. Stopped-flow data were collected for 10 s using a logarithmic time base, and averages were typically taken from four or more traces. Reaction rates were derived from curve fitting to exponential equations using Kaleidagraph (Synergy Software, Reading, PA).

RESULTS
Consistent with earlier data on the D705A and D882A Pol I(KF) mutants (16,17), single-turnover rate measurements of dNTP incorporation (Table 1) demonstrate the importance of the invariant carboxylates, Asp 705 and Asp 882 , in the polymer-ase reaction. Ala substitution of either side chain caused a decrease of at least 6 orders of magnitude in the rate of dNTP addition. Because Asp 705 and Asp 882 are ligands to divalent metal ions at the polymerase active site, we examined the effect of metal ion concentration on the polymerase activity of these mutants (supplemental Fig. S1). Wild-type Pol I(KF) had a broad Mg 2ϩ optimum from 5 to 10 mM. The very low activity of the D705A and D882A Pol I(KF) mutants, measured at 10 mM Mg 2ϩ , was not improved by further increasing the metal ion concentration. In wild-type Pol I(KF) and the D882A mutant, substitution of Mn 2ϩ for Mg 2ϩ had little effect, causing only a small decrease in the reaction rate. By contrast, the use of Mn 2ϩ instead of Mg 2ϩ increased the activity of the D705A mutant by ϳ10-fold at 20 mM metal ion and by up to 20-fold at metal ion concentrations above 50 mM.
In light of the crucial role of Asp 705 and Asp 882 in the overall process of dNTP incorporation, further experiments in this study investigated their participation in the early steps of the reaction. We used two fluorescence assays that we had previously developed to study prechemistry conformational transitions (2, 3). A 2-AP probe on the DNA substrate reports step 2.1, and a FRET-based assay reports step 2.2, the fingers-closing transition (Fig. 1).
Asp 705 and Asp 882 Are Not Required for the Prechemistry Step 2.1-The DNA duplex, L:T(ϩ1)2-AP:3ЈH (Fig. 3B), which has a 2-AP reporter as the 5Ј neighbor of the templating base, provides a sensitive fluorescence assay for step 2.1. The primer strand was dideoxy-terminated so as to block covalent addition of a dNTP while allowing observation of the prechemistry steps. The fluorescence emission spectra of complexes of the 2-AP DNA with the D705A and D882A mutants were similar to that previously reported for wild-type Pol I(KF) (2); relative to unbound 2-AP DNA, there was a ϳ1.5-fold increase in 2-AP fluorescence upon forming the Pol-DNA binary complex and a ϳ2.5-fold increase upon forming the Pol-DNA-dNTP ternary complex with the complementary dGTP (supplemental Fig. S2). We assume that the increase in 2-AP fluorescence represents a conformational rearrangement that decreases the stacking of the T(ϩ1) base with adjacent bases and may resemble the dislocation of the T(ϩ1) base seen in ternary complex cocrystal structures (4 -6).
We used the stopped-flow instrument to monitor the changes in 2-AP fluorescence, corresponding to step 2.1, as a function of time. The results for wild-type Pol I(KF) and both mutants were similar (Fig. 4). Upon the addition of dGTP to the Pol-DNA binary complex, there was an initial very rapid fluorescence increase within the instrument dead time (ϳ3 ms), indicated by the increasing start points of the traces, followed by a slower second phase. Fitting the observable slow fluorescence increase to a single exponential gave an approximate rate for the second phase of the reaction. To refine this determination and to estimate the rate of the initial fast phase, we fitted each trace to a double exponential, taking account of the amplitude missing due to the rapid fluorescence changes in the first 3 ms (for details of the fitting procedure, see supplemental Fig. S3). The results of this analysis (Table 2) indicate a rate of at least 400 s Ϫ1 for the initial phase of the fluorescence change. The differences between the be-

Single-turnover rates of dNTP incorporation by wild-type Pol I(KF) and the D705A and D882A mutants
Values reported as mean Ϯ S.D. were from two independent experiments; the others were from single determinations. The reactions were carried out under single-turnover conditions (see "Experimental Procedures") in the presence of a 10 mM concentration of the indicated divalent metal ion for WT and D882A and 20 mM for D705A. The DNA substrate, L:unmod:3ЈOH (Fig. 3A), has A at the templating position.
a In addition to the listed mutations, all proteins in this study were N-His 6 , D424A, L744C, and C907S. b The reaction with WT Pol I(KF) contained 40 M dTTP. c Reactions with the mutant Pol I(KF) derivatives contained 500 M dTTP. Because the reactions were extremely slow, the rates were calculated from the yield of product DNA after 4 h.
havior of the two carboxylate mutants and wild-type Pol I(KF) are relatively subtle: a faster Rate 1 for D705A, a slower Rate 2 for both mutants, and differences in the relative amplitudes of the two phases. We interpret the stopped-flow kinetics within the framework of Scheme 1.
In this scheme, K S , K 1 , and K 2 are the equilibrium constants for the initial binding of nucleotide and the two subsequent transitions. For simplicity, we assume that the 2-AP fluorescence increase results from the Rate 1 conformational transition, and the slower (Rate 2 ) transition shifts the equilibrium in favor of products without further changes to the 2-AP fluorescence. Because of the substantial difference in Rate 1 and Rate 2 , we can make the following inferences from our analysis ( Table 2). The ratio of the amplitudes of the two phases should approximate the equilibrium across the first conformational transition; relative to wild type, this transition appears more favorable in D705A and less favorable in D882A. 3 The overall K d for dNTP binding was determined from the hyperbolic dependence on dNTP concentration of the end point of the fluorescence traces (supplemental Fig. S3) and is equal to 1/(K S K 1 K 2 ).
Asp 705 Is Not Required for the Fingers-closing Transition (Step 2.2)-Our assay for the fingers-closing transition (3) uses FRET between an AEDANS fluorophore at position 744, on the mobile portion of the fingers subdomain, and a dabcyl quencher at position T(Ϫ8) on the DNA template strand (Fig.  3C). Fig. 2 illustrates the decrease in donor-to-quencher distance, resulting in a decrease in AEDANS emission, associated with fingers-closing and also shows the positions of the Asp 705 and Asp 882 side chains in the present study. As in the 2-AP experiments, the DNA primer terminus was dideoxyterminated to limit observations to the prechemistry steps. Emission spectra of 744-AEDANS Pol I(KF) were consistent with our previous study (3); formation of the binary Pol-DNA complex with the dabcyl-DNA decreased the AEDANS fluorescence by ϳ30%, and the addition of the complementary dTTP caused a further ϳ20% decrease. The D705A and D882A mutations did not change the response upon forming the binary complex; however, the fluorescence decrease upon the addition of dTTP was smaller for D705A (ϳ10% de-  Table S2).  S3 and Table S2). Rate 2 showed very little variation with dNTP concentration, and therefore we report the average of Rate 2 values at all nucleotide concentrations. b The ratio of the maximum values of the amplitudes of the fast and slow phases, each determined by plotting amplitude against dGTP concentration and fitting the data to a hyperbolic equation. c K d overall was determined by plotting the end point of each fluorescence trace against dGTP concentration and fitting to a hyperbolic equation (supplemental Fig. S3).  . S4).

Active-site Carboxylates in DNA Polymerase Prechemistry Steps
Using the stopped-flow instrument, we analyzed the kinetics of the fluorescence decrease associated with closed complex formation. As described previously (3), mixing of 744-AEDANS Pol I(KF), in complex with the dabcyl-DNA duplex, with the complementary dTTP resulted in a biphasic fluorescence decrease ( Fig. 5A and Table 3). The rates obtained from the data are consistent with our previous study. Our interpretation of the biphasic kinetics is analogous to the scheme described above for the 2-AP stoppedflow fluorescence experiments, namely that the fingers-closing transition (causing the fluorescence decrease) corresponds to the faster rate, and the second phase is a subsequent non-fluorescent transition that displaces the overall equilibrium toward products.
The D705A mutant also showed biphasic kinetics, but both rates were substantially faster than those of wild-type Pol I(KF) (Fig. 5B and Table 3). The first phase of the reaction was essentially complete within 3 ms (corresponding to a rate of Ն500 s Ϫ1 ) and therefore was observed as a decrease in the start points of the fluorescence traces. No fluorescence change was observed when using a control DNA duplex that lacked the dabcyl quencher, ruling out the possibility that the  Table S3 for curve fitting and residuals), yielding the reaction rates and amplitudes reported in Table 3. In the absence of the dabcyl quencher on DNA, essentially no change in AEDANS fluorescence was observed (supplemental Fig. S6A).  Fig. S5. b The ratio of the maximum values of the amplitudes of the fast and slow phases, determined by plotting the individual amplitudes against dTTP concentration and fitting the data to a hyperbolic equation. c The parameters k max and K d(app) were derived from rate measurements at a series of dTTP concentrations by plotting Rate 1 against dTTP concentration and fitting the data to a hyperbolic equation. d K d overall was determined from the hyperbolic dependence on dTTP concentration of either the total amplitude or the end point of the fluorescence traces. e The rate of this very fast process was estimated assuming that 80% of the process was accomplished within 3 ms. start point decrease in Fig. 5B might be an instrument artifact (supplemental Fig. S6A). The D705A mutation also caused a decrease in the dNTP binding affinity, as shown by the 15fold increase, relative to wild type, in the overall K d(dTTP) for the fingers-closing process (Table 3). Essentially identical fluorescence changes were observed when a deoxy-terminated primer was used instead of the dideoxy-terminated DNA (supplemental Fig. S6B and Table S1).
Asp 882 Is Required for the Fingers-closing Transition-No fluorescence decrease corresponding to fingers-closing was observed with the D882A mutant, regardless of whether the DNA substrate had a 3Ј-H (Fig. 6A) or a 3Ј-OH (supplemental Fig. S6D). Instead, we observed a dNTP-dependent fluorescence increase with a rate of 3.1 s Ϫ1 . Because dissociation of the dabcyl-DNA from the ternary complex could account for this fluorescence increase, we measured the DNA dissociation rate directly using a DNA trap. The complex of labeled D882A Pol I(KF) with dabcyl-DNA was mixed in the stoppedflow instrument with a 10-fold excess of an unmodified DNA duplex to serve as a trap and prevent reassociation of free labeled protein with the dabcyl-DNA (Fig. 6B). The rate of DNA dissociation measured in the presence of dTTP, so as to form a correctly paired ternary complex, was similar to the rate of the fluorescence increase observed for D882A in the fingers-closing assay ( Table 4), suggesting that the latter is indeed caused by DNA dissociation. Moreover, the DNA dissociation rate was faster in the presence of nucleotide (both the complementary dTTP and the mispaired dGTP) than in its absence, consistent with the traces in Fig. 6A showing the fluorescence increase to be dNTP-dependent.
A Single Metal Ion Is Sufficient for the Prechemistry Steps 2.1 and 2.2-To investigate the requirements for the two Mg 2ϩ ions during the prechemistry steps, we limited the amount of free Mg 2ϩ in the reaction buffer using the approach developed by Bakhtina et al. (12). The rationale of this experiment is that Mg 2ϩ will bind to dNTP and thus to the polymerase at the metal B site at micromolar concentrations, whereas the metal A site has a much lower affinity for Mg 2ϩ , indicated by a Mg 2ϩ optimum for dNTP incorporation in the millimolar range. Maintaining the concentration of free Mg 2ϩ in the micromolar range therefore has an effect on the polymerase reaction similar to that of an exchange-inert metal-dNTP complex, without introducing an unnatural metal ion. We carried out stopped-flow fluorescence assays for steps 2.1 and 2.2 in the presence of very low concentrations of free Mg 2ϩ and Mg-dNTP, controlled by the inclusion of EDTA in the nucleotide solution, as described (12). Fig. 7A shows the 2-AP fluorescence assay for step 2.1 using the extendable DNA duplex, L:T(ϩ1)2-AP:3ЈOH DNA. In the presence of 3 mM Mg 2ϩ , the initial rapid fluorescence increase was followed by a fluorescence decrease corresponding to the incorporation of dGTP, as reported previously (2). However, at the lower Mg 2ϩ concentrations indicated in Fig.  Table S3.

TABLE 4 Rates of dissociation of D882A Pol I(KF) complex with DNA
The DNA oligonucleotide was H:T(Ϫ8)D:3ЈOH (Fig. 3C). Values reported as mean Ϯ S.D were from two independent experiments; the others were from single determinations. 7A, the fluorescence signal was almost indistinguishable from that observed with a nonextendable DNA (compare Fig. 4A), with only a slight fluorescence decrease seen at long time points with 0.7 mM total Mg 2ϩ .
Analogous results were obtained in the fingers-closing FRET assay using 744-AEDANS Pol I(KF) and the extendable dabcyl substrate, H:T(Ϫ8)D:3ЈOH (Fig. 7B). In 3 mM Mg 2ϩ , the addition of the complementary dTTP resulted in the expected fingers-closing fluorescence decrease followed by a fluorescence increase due to the covalent addition of the nucleotide (3). As in the 2-AP experiment, the lower Mg 2ϩ concentrations gave traces very similar to those obtained with the corresponding nonextendable DNA (Fig. 5A), with the fluorescence increase detectable only at long times with 0.4 and 0.7 mM total Mg 2ϩ . Together, these experiments show that binding of dNTP-Mg 2ϩ alone is sufficient for the polymerase reaction to progress as far as formation of the closed ternary complex, but further addition of Mg 2ϩ is required to fill the second metal ion site and support dNTP incorporation.
Effects of Mn 2ϩ on the Early Stages of the Reaction-When the stopped-flow assays for steps 2.1 and 2.2 were carried out with Mn 2ϩ as the metal cofactor, the results were significantly different from those obtained with Mg 2ϩ . With the T(ϩ1)2-AP probe, reporting step 2.1, the characteristic fluorescence increase was much slower, ϳ12 s Ϫ1 for wild-type Pol I(KF) and 1 s Ϫ1 for D705A (supplemental Fig. S7), compared with Ն400 s Ϫ1 in the Mg 2ϩ reaction.
In the fingers-closing assay, wild-type Pol I(KF) gave similar reaction rates in Mn 2ϩ and Mg 2ϩ (Fig. 5C and Table 3). However, K d overall , obtained from the dTTP concentration dependence of the amplitudes or the end points of the fluorescence traces, was much lower in Mn 2ϩ than in Mg 2ϩ . As a result, the fingers-closing transition in Mn 2ϩ was essentially irreversible, with all of the fluorescence traces reaching the identical end point, even at dTTP concentrations as low as 1 M. With the D705A mutant, the reaction was much slower in Mn 2ϩ compared with Mg 2ϩ and was no longer detectably biphasic ( Fig. 5D and Table 3). The dTTP affinity of D705A, measured by K d overall , was almost 20-fold greater in Mn 2ϩ . Because the T(ϩ1)2-AP fluorescence increase in Mn 2ϩ is slower than the fingers-closing transition, we infer that the DNA rearrangement reported by the 2-AP probe must take place after fingers-closing in Mn 2ϩ and is not necessarily identical to the equivalent step observed in Mg 2ϩ . Intriguingly, the 2-AP fluorescence increase seen with the D705A mutant in Mn 2ϩ is preceded by a fluorescence decrease at ϳ60 s Ϫ1 , similar to the fingers-closing rate (supplemental Fig.  S7 and Table 3).
In the case of the D882A mutant, a low amplitude fluorescence decrease was observed when Mn 2ϩ was used as the metal cofactor (Fig. 6C), suggesting that Mn 2ϩ , unlike Mg 2ϩ , allows the fingers-closing transition to proceed to a small extent. As with Mg 2ϩ , a fluorescence increase, most probably due to DNA dissociation, was also observed. Because of the low amplitudes, it was difficult to get accurate determinations of the rates of these two fluorescence changes; we estimate the fluorescence decrease to be ϳ60 s Ϫ1 , similar to the wildtype fingers-closing rate, and the increase to be ϳ2 s Ϫ1 . The behavior of the D705A and D882A mutants in the presence of Mn 2ϩ was unchanged when an extendable (deoxy-terminated) DNA substrate was used instead of the nonextendable DNA in the experiments described above (supplemental Fig.  S6 and Table S1).

DISCUSSION
Residues Asp 705 and Asp 882 of Pol I(KF) each make bifurcated interactions with the two catalytically essential metal ions (A and B) at the polymerase active site in the closed ternary complex. Therefore, it is no surprise that mutation of either carboxylate results in an enzyme that is essentially inactive in terms of incorporation of nucleotides into DNA (16,17). The current study examines the role of Asp 705 and Asp 882 and their associated metal ions in the prechemistry steps of the DNA polymerase reaction that assemble the active site in readiness for the phosphoryl transfer reaction. Despite the dramatic effect of the D705A and D882A mutations on the chemical incorporation rate, the carboxylate side chains have much less influence on some of the early steps of the reaction. The behavior of the D705A and D882A mutants in fluores- cence-based assays that probe prechemistry conformational transitions sheds light on the sequence of events in the early steps of the DNA polymerase reaction pathway. Interpreting our results in the context of available crystal structures of Afamily DNA polymerases in complexes with DNA and nucleotide substrates, we present a structural description of the assembly of the catalytically competent polymerase active site (Fig. 8), with the caveat that such a structural description is limited by the lack of cocrystal data for the early (and probably unstable) intermediates that we are most interested in.
Nucleotide Binding-The sequence of prechemistry steps in the DNA polymerase reaction is initiated by the binding of nucleotide to a Pol-DNA binary complex, whose structure we have taken as the starting point in our mechanism. We propose that the initial contacts made by an incoming dNTP resemble those seen in the DNA-free polymerase-dNTP binary complex cocrystals of Pol I(KF) and Klentaq (19,20), specifically interactions between the dNTP phosphates and conserved positively charged side chains on or close to the Ohelix of the fingers subdomain (Fig. 8A). Kinetic studies have demonstrated the importance for dNTP binding of residues such as Arg 754 of Pol I(KF), which is predicted to interact with the ␥-phosphate (21). Moreover, the interactions between Klentaq and dCTP, observed in the Pol-dNTP binary complex, are preserved in the closed Pol-DNA-dNTP ternary complex, implying that movement of the O-helix simply delivers the bound dNTP into the active site (5).
Our experiments using very low concentrations of Mg 2ϩ support a model in which the dNTP binds to the Pol-DNA complex as dNTP-Mg 2ϩ , whereas the second Mg 2ϩ , required for the completion of catalysis, does not bind until later. The fluorescence changes associated with steps 2.1 and 2.2 were not observed in the absence of divalent metal ions; however, these conformational transitions, but not the subsequent phosphoryl transfer, could take place with Mg 2ϩ at only micromolar levels (Fig. 7). These observations are consistent with studies of Pol ␤ and RB69 DNA polymerase showing that exchange-inert metal-dNTP complexes, containing Cr(III) or Rh(III), promote early conformational change steps in the absence of exogenous divalent metal ions, the latter being required for covalent incorporation of the nucleotide into DNA (10 -13). The idea that metal B enters the active site as dNTP-Mg 2ϩ is supported by NMR studies of Pol I(KF) showing that binding of dNTP to the polymerase domain creates a high affinity binding site for a divalent metal ion (22). Moreover, cocrystals of A-family Pol-DNA binary complexes do not have a bound metal ion in a position equivalent to the metal A site (6,23). (Some binary complex cocrystals have a divalent metal ion coordinated to the two active-site carboxylates, in a location close to the metal B site, but its mechanistic significance is unclear.) DNA Rearrangement, Step 2.1-The fluorescence change of the T(ϩ1)2-AP reporter is affected very little by the D705A or D882A mutations, indicating that the active-site carboxylates are not important either for the initial binding of dNTP and its bound Mg 2ϩ or for this subsequent conformational transition. If indeed the incoming dNTP is bound as we have suggested in Fig. 8A, the Asp 705 and Asp 882 side chains would be too distant to interact at this early stage with the chelated metal B, although both aspartates are ligands to metal B in the closed ternary complex. We know very little about the structural change associated with step 2.1, except that the polymerase must remain in an open conformation because fin-  5KTQ coordinates). B, the ternary complex (taken from Protein Data Bank entry 1LV5) (6) immediately after the fingers have closed but before metal A binds, showing the change in position of the Asp 705 and Asp 882 side chains, upon going from the open (yellow, as in A) to closed (cyan) conformation. Arg 754 and Lys 758 remain in contact with the incoming nucleotide in the closed conformation but are omitted for clarity. Since the primer is dideoxy-terminated in this structure, the interaction between the 3Ј-OH group and the carboxylate of Asp 882 is absent. C, the closed complex is illustrated (as in B), showing the entry of metal A (red, from 1LV5). This structure represents the enzyme poised for catalysis.
gers-closing is a subsequent step. The fluorescence increase associated with step 2.1 implies that the T(ϩ1) base becomes less stacked; however, the binary complex crystal structures do not provide much information on the interactions responsible for the observed change. In the Pol-DNA binary complex, the templating base is bound in the preinsertion binding pocket formed by the O and O 1 helices and consequently is not able to stack with the T(ϩ1) base, its 5Ј neighbor. Therefore, the relevant interactions with the T(ϩ1) base probably involve either the protein or the T(ϩ2) base (absent in the majority of binary complex structures), but our earlier attempts to identify the contacts proved inconclusive (2). Currently, we visualize step 2.1 as a local rearrangement of the unpaired template DNA that decreases the stacking interactions made by the T(ϩ1) base. An intriguing possibility is that a change in the environment of the T(ϩ1) base could be associated with movement of the templating base into a position suitable for testing the complementarity of the incoming dNTP.
The fluorescence increase of step 2.1 is not observed if the incoming dNTP is not complementary to the templating base (2, 3), 4 consistent with the idea of an early fidelity checkpoint that discriminates against mispairs. It is possible that the discriminator step precedes step 2.1 so that only correctly paired ternary complexes initiate step 2.1. Alternatively, step 2.1 itself could be the discriminator, with the equilibrium favoring the forward reaction for a correctly paired dNTP and favoring the reverse reaction for a mispair.
Fingers-closing and the Role of Asp 882 -Step 2.2 (fingersclosing) takes place essentially normally with the D705A mutant but is strongly impaired in D882A Pol I(KF). No fingersclosing fluorescence signal was observed with D882A in the presence of Mg 2ϩ ; in Mn 2ϩ , there was a limited extent of reaction at a rate similar to the wild-type fingers-closing rate. This suggests that D882A has not lost the ability to form the closed complex, but instead, the mutation has changed the open-closed equilibrium of the ternary complex so that it is biased in favor of the open complex (more so in Mg 2ϩ than in Mn 2ϩ ). Single-molecule FRET studies of Pol I(KF) demonstrate a finely balanced equilibrium between open and closed conformations, with the closed conformation favored by ϳ1 kcal/mol in a correctly paired ternary complex (24), so it is not unreasonable to imagine that the loss of a key side chain could shift the equilibrium so that the ternary complex is predominantly in the open conformation. In this scenario, we assume that the low level of catalysis observed with D882A occurs via the sparsely populated closed ternary complex.
What is the role of Asp 882 in stabilizing the closed conformation? Alignment of open binary and closed ternary complexes of Bst pol and Klentaq shows that the position of the side chain equivalent to Asp 882 changes very little. Thus, Asp 882 may serve as an anchor point during the fingers-closing transition, making an interaction with metal B as the movement of the O-helix delivers the dNTP-Mg 2ϩ complex into the active site (Fig. 8B). The Bst pol binary complex illus-trated here (as well as other binary complexes in which the primer is deoxy-terminated) has a hydrogen bond between the homologue of Asp 882 and the primer 3Ј-OH (Fig. 8A) (6,23). In the closed ternary complex crystal structure, this hydrogen bond is absent because the primer is dideoxy-terminated; however, the similar positioning of Asp 882 in open and closed conformations suggests that, with the natural substrates, Asp 882 bridges the active site, linking the 3Ј-OH and the dNTP-Mg 2ϩ , and this could be the basis for its role as an anchor point during fingers-closing.
There is an interesting contrast between wild-type Pol I(KF), where formation of the closed ternary complex stabilizes DNA binding (3), and D882A, where the ternary complex is predominantly in the open conformation and the addition of nucleotide, even when it is complementary to the templating base, destabilizes DNA binding relative to the binary complex ( Fig. 6 and Table 4; compare with wild-type data listed in Table 4 Footnote b). We suggest that weaker DNA binding may be a property inherent to the open ternary complex rather than being a direct consequence of the absent Asp 882 side chain. The behavior of D882A resembles the destabilization of DNA binding observed for mismatched ternary complexes of wild-type Pol I(KF), which also remain in an open conformation (3). As we have suggested previously (3,24), the open ternary complex may represent an important checkpoint in which the incoming dNTP previews the template base, and the eventual outcome is determined by whether the nascent base pair is correct or mismatched. The D882A mutant could therefore be a valuable tool, allowing us to arrest the reaction at the open ternary complex and to study this intermediate using a variety of physical techniques.
The Role of Asp 705 -In our stopped-flow experiments, the D705A mutant is fully capable of carrying out the fingersclosing transition and actually does so at a faster rate than wild-type Pol I(KF). Therefore, Asp 705 does not assume its important role in catalysis until after the closed complex has formed. The Klentaq and Bst pol binary (Pol-DNA) complex structures show that, at the start of the reaction, Asp 705 , unlike Asp 882 , is not positioned appropriately to be a metal ligand; instead, it faces away from the active site region (Fig.  8A). In the closed ternary complex structures (with both metals A and B) the side chain of Asp 705 has rotated so that one of the carboxylate oxygens is a ligand to metal B (Fig. 8B). Two possible scenarios explain the faster rate of fingers-closing observed with the D705A mutant: 1) the repositioning of the Asp 705 side chain could take place, at least in part, during the fingers-closing conformational change and might involve a significant energy barrier; alternatively, 2) the position of the Asp 705 side chain itself might impede the fingers-closing process. In either case, the presence of Asp 705 (in the wildtype enzyme) requires the input of energy, relative to D705A, giving a higher energy transition state, which could enhance the catalysis of subsequent steps. Structural studies of RB69 DNA polymerase, HIV-1 reverse transcriptase, and Pol ␤ also show alternative conformations of active-site side chains in the prechemistry steps of the reaction pathway, suggesting that the mobility of key side chains may be a common feature of the polymerase mechanism (11,(25)(26)(27). In Pol ␤, both di-rect measurements and computational studies indicate that some side chain motions are energy-requiring and can potentially be rate-limiting for the reaction kinetics (28).
The repositioning of Asp 705 during or subsequent to fingers-closing contributes a ligand to the metal A binding site, and we suggest that the important function of Asp 705 is to ensure the proper coordination of metal A in preparation for phosphoryl transfer. The participation of Asp 705 in a step after fingers-closing should further stabilize the closed complex, accounting for the tighter binding of dNTP to wild-type Pol I(KF), compared with D705A. Our earlier data on the fingers-closing process in the presence of Ca 2ϩ had suggested a metal-dependent step after fingers-closing but before chemistry (3), and the current study provides additional lines of evidence that coordination of metal A occurs after fingersclosing. First, both of the early fluorescently detectable steps take place normally in the presence of very low Mg 2ϩ concentrations, but higher metal ion concentrations are required to complete phosphoryl transfer, implying that an additional metal site must be filled after fingers-closing. Second, the behavior of the D705A and D882A mutant proteins in the fingers-closing assay is not appreciably influenced by the presence of the primer 3Ј-OH, one of the ligands to metal A, suggesting that the metal A site is irrelevant until after the fingers-closing step. The participation of Asp 705 in a late step involving metal coordination is also consistent with the 10fold increase in the rate of dNTP incorporation by D705A when Mg 2ϩ is replaced by Mn 2ϩ , a metal ion that is better able to bind to sites with imperfect geometry or missing ligands (29,30). Mn 2ϩ had the opposite effect, a slight decrease in the rate of dNTP incorporation, on wild-type Pol I(KF), which should have a normal metal A site, and on D882A, which does not proceed efficiently beyond the open ternary complex.
Steps following Fingers-closing-As described above, our data suggest that Asp 705 plays a role in the binding of metal A after the fingers-closing conformational change, a process that may correspond to the unidentified rate-limiting prechemistry step 3. Plausible candidates for this rate-limiting step are the movement of Asp 705 into its position as a metal A ligand or the entry of metal A into its coordination site. Studies of other DNA polymerases using exchange-inert dNTPmetal complexes indicate that one or more early conformational transitions can take place in the absence of the second active-site metal ion, just as we have demonstrated here, although there is a lack of consensus on when the second metal ion enters the ternary complex. In the case of the X-family DNA polymerase, Pol ␤, structural studies have shown that the fingers close in the absence of the second metal ion (10,11), and stopped-flow fluorescence indicated that binding of the second metal and catalysis occurs without kinetic consequences (12); however, computational studies suggested that fingers-closing in Pol ␤ requires both active-site metal ions (28,31). With the B-family DNA polymerase from phage RB69, a conformational transition (detected by a template 2-AP probe but assumed to be fingers-closing) is induced by the exchange-inert dNTP-Rh(III), but subsequent binding of the second (catalytically essential) metal ion is extremely slow, suggesting that it normally binds in advance of the nucleotide and cannot gain entry to the ternary complex without reversing the conformational transition (13,32).
Our data do not rule out an additional role for Asp 705 in the chemical step itself; nor do they rule out significant participation by Asp 882 in the steps that follow fingers-closing. (The latter question cannot be addressed because D882A does not proceed beyond the open ternary complex.) Indeed, a computational study of T7 DNA polymerase, a homologue of Pol I(KF), has suggested that the residue equivalent to Asp 882 may act as a general base to receive the proton from the primer 3Ј-OH, so as to generate the attacking nucleophile for the phosphoryl transfer reaction (33). This idea fits well with the consistent observation of a hydrogen bond between the Asp 882 homologue and the primer 3Ј-OH in A-family Pol-DNA binary complex cocrystals in which the primer strand has a 3Ј-OH (6,23).
Effects of Mn 2ϩ -Our preliminary observations when replacing the natural cofactor, Mg 2ϩ , with Mn 2ϩ are intriguing, and further studies are ongoing in order to understand their full implications. The largest effects were the decrease in rate, by ϳ2 orders of magnitude, of the fluorescence increase reported by the T(ϩ1)2-AP probe and the increase in overall dNTP binding affinity measured by the end point of the fingers-closing fluorescence change. An obvious inference from the slower rate seen with the 2-AP probe is that, in Mn 2ϩ , the putative DNA rearrangement must take place after fingersclosing and may not necessarily be identical to the step 2.1 process seen in the Mg 2ϩ reaction. This makes it difficult to interpret the metal-dependent differences in the fingers-closing assay because the open ternary complex structure at the start of the fingers-closing is different in Mg 2ϩ and Mn 2ϩ .
There is little definitive information that establishes why Mn 2ϩ should have such a profound effect on the early stages of the DNA polymerase reaction. In a crystallographic study of Pol ␤, Pelletier et al. (34) reported that Mn 2ϩ is coordinated to the dNTP at the active site as an ␣-␥ bidentate, instead of the ␤-␥ bidentate observed with Mg 2ϩ . They also reported tighter nucleotide binding in the presence of Mn 2ϩ , which is similar to our observations. These results raise the possibility that dNTP-Mn 2ϩ may find an aberrant, perhaps lower energy, binding site on the DNA polymerase and that this may alter the early steps of the reaction pathway. Changes to the pathway that alter or eliminate an early fidelity checkpoint might explain the ability of Mn 2ϩ to decrease polymerase fidelity (30,(35)(36)(37).
Conclusions-From our studies of the effects of active-site carboxylate mutations on the prechemistry steps in the Pol I(KF) reaction pathway, a picture is starting to emerge of the sequence of events in the pathway and the stages at which the carboxylates and their associated metal ions are involved. Our results suggest an exquisitely choreographed series of conformational transitions that sequentially assemble the active site into a reaction-competent configuration. Similar conclusions were reached in computational studies of Pol ␤ (28,31), an X-family polymerase unrelated to Pol I, suggesting that the stepwise assembly of the correct active-site geometry may be a common feature of DNA polymerases, providing multiple FEBRUARY 4, 2011 • VOLUME 286 • NUMBER 5 fidelity checkpoints at which inappropriate substrates can be diverted from the productive reaction pathway.