DNA structure and aspartate 276 influence nucleotide binding to human DNA polymerase beta. Implication for the identity of the rate-limiting conformational change.

Structures of DNA polymerase (pol) beta bound to single-nucleotide gapped DNA had revealed that the lyase and pol domains form a "doughnut-shaped" structure altering the dNTP binding pocket in a fashion that is not observed when bound to non-gapped DNA. We have investigated dNTP binding to pol beta-DNA complexes employing steady-state and pre-steady-state kinetics. Although pol beta has a kinetic scheme similar to other DNA polymerases, polymerization by pol beta is limited by at least two partially rate-limiting steps: a conformational change after dNTP ground-state binding and product release. The equilibrium binding constant, K(d)((dNTP)), decreased and the insertion efficiency increased with a one-nucleotide gapped DNA substrate, as compared with non-gapped DNA. Valine substitution for Asp(276), which interacts with the base of the incoming nucleotide, increased the binding affinity for the incoming nucleotide indicating that the negative charge contributed by Asp(276) weakens binding and that an interaction between residue 276 with the incoming nucleotide occurs during ground-state binding. Since the interaction between Asp(276) and the nascent base pair is observed only in the "closed" conformation of pol beta, the increased free energy in ground-state binding for the mutant suggests that the subsequent rate-limiting conformational change is not the "open" to "closed" structural transition, but instead is triggered in the closed pol conformation.

Proficient DNA repair systems are critical to maintaining the stability of the human genome. Many forms of endogenous base damage, including alkylation, oxidation, and deamination, result in the formation of abasic sites. In particular, normal and damaged DNA bases are lost spontaneously or removed by lesion-specific DNA glycosylases. The resulting abasic sites are repaired by the base excision repair (BER) 1 pathway to prevent the accumulation of this miscoding lesion. Following base removal, apurinic/apyrimidinic (AP) endonuclease incises the damaged strand 5Ј to the abasic site, leaving a single-nucleotide gap with a 5Ј-deoxyribose phosphate (dRP) flap. The gap is then processed by DNA polymerase (pol) ␤ to create nicked DNA that will be ligated by DNA ligase I or III.
DNA polymerase ␤ possesses both the nucleotidyltransferase activity to fill the one-nucleotide gap and lyase activity required to remove the dRP flap. The central role of pol ␤ in BER has been established (1,2), and, in the absence of pol ␤, alternate BER pathways have been reported (3,4).
DNA polymerase ␤ is an attractive model to study polymerase mechanisms employed to assure efficient and faithful DNA synthesis. Its small size and lack of accessory proteins has facilitated its biochemical, kinetic, and structural characterization. Although pol ␤ appears to have evolved separately from other classes of DNA/RNA polymerases of known structure (5), it shares many general structural and mechanistic features. Each of the polymerases possesses a groove along a broad face of the enzyme where nucleic acid binds. The polymerase domain of these enzymes has been likened to a right hand and is composed of finger, palm, and thumb subdomains (6). Polymerases have at least two acidic residues in the palm subdomain that bind catalytically essential metals. The crystal structures of the substrate complexes of pol ␤ (7,8), T7 DNA polymerase (9), Klentaq DNA polymerase (10), and HIV-1 reverse transcriptase (11) indicate that the reactive groups (i.e. metals, dNTP, and template-primer) have a similar three-dimensional arrangement. These structures are consistent with a "two metal ion" mechanism for nucleotidyl transfer (12).
In general, pol ␤ also utilizes a similar kinetic mechanism as most other DNA polymerases. Steady-state kinetic analyses indicate that pol ␤ follows an ordered addition of substrates (13). Employing pre-steady-state kinetics, it has been shown that Escherichia coli Klenow fragment (14,15), T4 DNA polymerase (15,16), T7 DNA polymerase (17), HIV-1 reverse transcriptase (18), and pol ␤ (19) utilize a two-step nucleotide binding mechanism. Initial dNTP binding to a pol-DNA complex places the nucleotide within the active site. A subsequent conformational change results in the alignment of the catalytic atoms and rapid chemistry. Binding of the correct nucleotide facilitates this conformational change, whereas binding of the incorrect nucleotide does not. This "induced-fit" mechanism is consistent with the numerous structural differences that are observed upon binding of a correct nucleotide (8). In particular, the carboxyl-terminal subdomain (residues 262-335) is observed to reposition itself after binding a correct dNTP by rotation about an axis, ␣-helix M, that positions ␣-helix N so that several side chains can interact with the nascent base pair in this "closed" conformation.
Replicative DNA polymerases and reverse transcriptases often have an intrinsically associated exonuclease or endonuclease activity that complements their fundamental nucleotidyltransferase function. DNA polymerase ␤ also has an associated accessory activity that complements its DNA synthesis step in BER. Mild proteolysis of pol ␤ separates a 31-kDa fragment, which possesses the nucleotidyltransferase activity, from an amino-terminal 8-kDa fragment (20). This smaller domain binds strongly to 5Ј-phosphate groups in gapped DNA (21) and possesses the dRP lyase activity that is needed for removal of the dRP flap remaining after AP endonuclease cleavage of the abasic site intermediate in BER (22,23). In the absence of a downstream DNA strand (i.e. no DNA gap), a crystal structure of a pol ␤-DNA-ddNTP complex revealed that the 8-kDa domain does not interact with the DNA or the 31-kDa domain, but instead is positioned some distance from the 31-kDa domain (Fig. 1A). The crystal structure of pol ␤ bound to onenucleotide gapped DNA demonstrates that the 8-kDa domain binds to the 5Ј-phosphate in the DNA gap and interacts with the carboxyl terminus of the 31-kDa domain (Fig. 1A) (8). This results in a more compact, doughnut-shaped structure forming a channel with dimensions appropriate for dNTP diffusion (Fig.  1B) (24). Thus, this conformation of the 8-kDa domain is potentially significant due to its proximity to the nucleotide binding pocket of pol ␤. The experiments described here were performed to examine nucleotide binding to pol ␤ and to ascertain the influence of a gapped DNA structure. Additionally, the role of Asp 276 and ␣-helix N in dNTP binding was investigated by comparing the effect of valine substitution for Asp 276 on nucleotide binding with alternate DNA substrates. Asp 276 is observed to form van der Waals contact with the base of the incoming nucleotide triphosphate in the closed pol ␤ ternary substrate complex. Employing steady-state and pre-steadystate kinetics, we examine nucleotide binding to pol ␤. The results indicate that residue 276 interactions with the incoming nucleotide influence ground-state nucleotide binding during template base recognition. The implication is that ␣-helix N and Asp 276 influence ground-state binding in the closed conformation that occurs prior to the rate-limiting conformational change and that subdomain closure is very rapid and not rate-limiting. A 34-mer oligonucleotide substrate containing a single nucleotide gap was prepared by annealing three gel-purified oligonucleotides (Oligos Etc., Wilsonville, OR) to create a gap at position 16. Each oligonucleotide was resuspended in 10 mM Tris-HCl, pH 7.4, and 1 mM EDTA, and the concentration determined from their UV absorbance at 260 nm. The annealing reactions were carried out by incubating a solution of 10 M primer with 11 M each of downstream and template oligonucleotides at 90°C for 2 min followed by slow cooling to room temperature. The sequence of the gapped DNA substrate was as follows: primer, 5Ј-CTGCAGCTGATGCGC-3Ј; downstream oligonucleotide, 5Ј-GTACG-GATCCCCGGGTAC-3Ј; template, 3Ј-GACGTCGACTACGCGGCATGC-CTAGGGGCCCATG-5Ј.

Materials
The upstream primer in each case was 5Ј-labeled with [␥-32 P]ATP (specific activity ϭ 6.6 ϫ 10 6 dpm/pmol) using T4 polynucleotide kinase (New England Biolabs) and contaminating radioactive ATP was removed with a MicroSpin G-25 column. The downstream primer was synthesized with a 5Ј-phosphate. A non-gapped DNA substrate was prepared by omitting the downstream oligonucleotide. A nicked substrate was prepared by annealing a 16-mer primer, where the 15-mer primer had a dCMP added to the 3Ј-end, and downstream oligonucleotide with template as described above.
Human recombinant pol ␤ (wild type and D276V) was overexpressed in Escherichia coli cells and purified as described (25). The enzyme concentration was determined from the absorbance at 280 nm using an extinction coefficient of 21,200 cm Ϫ1 M Ϫ1 (26).
Kinetic Assays-Rapid-quench assays were performed at 37°C using a KinTek Model RQF-3 rapid-quench-flow apparatus (KinTek Corp., Austin, TX). Unless noted otherwise, all concentrations refer to the final concentration after mixing. The final reaction mixture typically consisted of 50 mM Tris-HCl, pH 7.2, 100 mM KCl, 100 g/ml bovine serum albumin, 10% glycerol, and 5 mM MgCl 2 . The final concentration of enzyme and substrates are given in the text or figure legends.
Enzyme and substrate DNA were preincubated for 1 min and rapidly mixed with dCTP and MgCl 2 to initiate each reaction. After various time periods, reactions were stopped by the addition of 0.25 M EDTA. The quenched samples were mixed with an equal volume of formamide dye and the products separated on 12% denaturing polyacrylamide gels. bound to non-gapped DNA (data not shown) (7) is superimposed with the palm of pol ␤ (blue-green 31-kDa pol domain and white 8-kDa lyase domain) bound to a one-nucleotide gapped DNA (8). The template strand of the one-nucleotide gapped DNA is red, while the primer and downstream strands are pink. The pink incoming ddCTP is illustrated in a Corey-Pauling-Koltun representation and identifies the 3Ј-primer terminus. The 8-kDa domain of pol ␤ bound to non-gapped DNA does not interact with DNA, but instead is positioned some distance from the 31-kDa domain. A transparent molecular surface of pol ␤ that is bound to the one-nucleotide gapped DNA substrate is shown and illustrates the doughnut-shaped structure of pol ␤ in this conformation. B, a view of the channel that is formed when the amino-terminal 8-kDa lyase domain (white) interacts with the carboxyl terminus of the 31-kDa pol domain (blue-green). This channel has dimensions appropriate for dNTP diffusion (24). This figure was made with GRASP (60) and Molscript (61) and rendered with Raster3D (62).
The dried gels were analyzed using a PhosphorImager (Molecular Dynamics) to quantify product formation. Steady-state time courses were measured by manual mixing and quenching at 37°C to determine steady-state constants (i.e. apparent k cat and K m ). Following the appropriate incubation period, individual reactions were quenched with EDTA and the quenched reaction mixtures were applied to DEAE cellulose filters (DE-81). The unincorporated [␣-32 P]dCTP was removed, and radioactive nucleotide incorporation was quantified as described previously (27).
Pyrophosphorolysis-A nicked DNA substrate (300 nM) was preincubated with 1 nM pol ␤ in reaction buffer. The reaction was initiated by the addition of 3 mM pyrophosphate and 5 mM MgCl 2 . The reaction was quenched at various time periods with EDTA.
Data Analysis-Time courses were fitted to appropriate equations by nonlinear least squares methods. These equations are described under "Results." Progress curves were also fitted to kinetic models with Kin-TekSim, which is based on the programs KINSIM (28) and FITSIM (29).

RESULTS
Influence of DNA Structure on Steady-state Efficiency of dCTP Incorporation-The efficiency of correct nucleotide (i.e. dCTP) incorporation by pol ␤ was dependent on the nature of the template-primer (Table I). DNA polymerase ␤ catalyzed single-nucleotide gap filling at a steady-state rate of 0.8 s Ϫ1 . A similar rate of polymerization was observed when pol ␤ was bound to non-gapped DNA (i.e. the downstream oligonucleotide was omitted from the annealing reaction). In contrast, the dCTP concentration dependence of DNA synthesis indicated that K m(dCTP) with a single nucleotide gap DNA substrate was 12-fold lower than for a substrate without a gap ( Table I), suggesting that the enzyme might possibly have a higher affinity for nucleotide when bound to gapped DNA. As a result of the lower K m , catalytic efficiency (k cat /K m(dCTP) ) was over an order of magnitude greater for the gapped DNA substrate assayed under steady-state conditions than for the non-gapped DNA substrate.
Equilibrium Nucleotide Binding-Since steady-state kinetic parameters are often the composite of several rate constants, we sought to determine the equilibrium constant (K d ) for nucleotide binding to pol ␤-DNA. Using rapid-mixing and quenching techniques, we measured pre-steady-state product formation. When a high concentration of pol ␤ is preincubated with excess single-nucleotide gapped DNA (DNA/enzyme ϭ 3) and rapidly mixed with a high concentration of the complementary nucleoside triphosphate (i.e. dCTP) and Mg 2ϩ , product formation is biphasic. A rapid burst of product formation precedes a slower apparent linear rate (Fig. 2, dashed line). The biphasic nature of the time course indicates that a step following chemistry is rate-limiting during steady-state catalysis. The slow step has been attributed to dissociation of the extended pol-DNA complex (19). The rates of product formation observed in Fig. 2 indicate that in contrast to processive replicative polymerases, k pol is not considerably greater than k off(DNAϩ1) for pol ␤. Scheme I describes the basic kinetic mechanism for singlenucleotide incorporation utilized by all DNA polymerases that have been examined. Following an ordered binding of substrate DNA and dNTP, a conformational change (k pol ) precedes rapid chemistry (see below). Pyrophosphate release is rapid, but dis-sociation of the extended product DNA is usually slow and at least partially rate-limiting.
Since dNTP binding is in rapid equilibrium, k pol is dependent on dNTP concentration and is given by Equation 1.
Fitting the time course in Fig. 2 to an equation with rising exponential and linear terms, as shown by Equation 2, results in a burst rate constant (k obs ) of 12.3 s Ϫ1 and an apparent linear rate (v ss ) of 120 nM/s (dashed line). With this analysis, the burst amplitude (A) represents the "apparent" active concentration of enzyme (A ϭ 40 nM) so that k ss (v ss /pol active ) would be 3.0 s Ϫ1 . Yet, since k obs is similar to k ss for pol ␤, this analysis is not satisfactory and inappropriate. According to Scheme I, pol ␤ will cycle quickly (k pol is similar to k off(DNAϩ1) ), so that the burst amplitude is an underestimate of the active enzyme concentration. The amplitude is given by Equation 3 (30).
Thus, even when k pol is 1 order of magnitude greater than k off(DNAϩ1) , the burst amplitude is still only 83% of the true TABLE I Steady-state kinetic constants for wild-type pol ␤-DNA complexes Gapped and non-gapped DNAs (300 nM) were preincubated with 1 nM pol ␤ prior to addition of dCTP. Initial velocity data were fitted to the Michaelis equation by nonlinear least squares methods.
a The gapped DNA substrate contains a one-nucleotide gap and was prepared as described under "Experimental Procedures." FIG. 2. Pre-steady-state incorporation of dCTP by DNA polymerase ␤. DNA polymerase ␤ (100 nM) was pre-incubated with 300 nM gapped DNA substrate prior to mixing with an equal volume of 50 M dCTP and 5 mM MgCl 2 to initiate the reaction. Reactions were stopped at time intervals and the products isolated and quantified as described under "Experimental Procedures." The time course of product formation appears to be biphasic with an initial rapid exponential phase followed by a linear phase. The dashed line represents the best fit to an equation (Equation 2 under "Results") with rising exponential and linear terms. The observed rate constant of the burst phase was 12.3 s Ϫ1 followed by an apparent linear rate (v ss ) of 120 nM/s. With this analysis, the burst amplitude represents the apparent active concentration of enzyme (40 nM) so that k ss (v ss /pol active ) would be 3.0 s Ϫ1 . However, since k pol / k off(DNAϩ1) is not considerably greater than 1, this analysis is not appropriate (see "Results" for details). Assuming rapid ligand binding and an "active" enzyme fraction of 70%, the data can be fitted to the model outlined in Scheme I to derive k pol and k off(DNAϩ1) of 7.7 and 2.8 s Ϫ1 , respectively (solid line). The simulated time course also indicates that the steady-state phase is short-lived (i.e. time course is non-linear).
active fraction of enzyme. The apparently linear portion of the time course also suffers from the high concentration of free product DNA (DNA ϩ1 ) that is accumulating after the first turnover, therefore competing with extendable substrate DNA (i.e. product inhibition). Correct analysis of the time course in Fig. 2 requires a model that takes into account product competition. Assuming rapid ligand binding and an "active" enzyme concentration of 70%, the data in Fig. 2 were fitted to the model outlined in Scheme I to estimate k pol and k off(DNAϩ1) (solid line).
The fitted values for k pol (7.7 Ϯ 0.5 s Ϫ1 ) and k off(DNAϩ1) (2.8 Ϯ 0.2 s Ϫ1 ) are similar to the values determined independently below.
To eliminate interference of enzyme cycling in determining the equilibrium nucleotide dissociation constant, an excess of enzyme was preincubated with DNA (enzyme/DNA ϭ 5). Under this condition, nearly all of the substrate DNA is bound to enzyme so that, upon addition of dCTP/Mg 2ϩ , dNTP binding and incorporation limit catalysis. Under these single-turnover conditions, the first-order time courses were dependent on the concentration of dCTP (Fig. 3). A secondary plot indicated that the apparent first-order rate constants can be fitted to Equation 1 with a K d and k pol of 5.6 M and 10.0 s Ϫ1 , respectively ( Fig. 3B and Table II). Interestingly, however, the amplitude of the fast exponential phase was only 50% (i.e. 50 nM) of that expected. Extended incubation of the pol with substrates demonstrated that greater than 90% of the primer could be ex-tended (data not shown). This result suggested that pol ␤ can bind to the template-primer in a catalytically nonproductive mode. This nonproductive binding form competes with the excess free enzyme for productive DNA binding resulting in a biphasic time course. To determine whether enzyme that bound DNA nonproductively can isomerize to form productive enzyme-bound DNA or must dissociate from DNA to allow enzyme the opportunity to bind productively, we added excess unlabeled DNA (5 M) with 50 M dCTP to pol ␤ (300 nM) preincubated with 100 nM 5Ј-labeled DNA. The unlabeled DNA binds free enzyme and enzyme that dissociates from the radioactively labeled DNA to eliminate enzyme cycling. The excess non-radioactively labeled DNA trap had no effect on the amplitude or rate of the rapid or slow phases (data not shown), indicating that for the wild-type enzyme that the nonproductive DNA binding form can isomerize to the productive form without dissociating from the DNA.
To ascertain whether dNTP incorporation during these single-turnover conditions is limited by chemistry, the rate constant for the incorporation of the ␣-thio-substituted analogue of dCTP was determined. Due to the lower electronegativity of sulfur relative to oxygen, a significant decrease in rate upon sulfur substitution would suggest that chemistry is rate-limiting. Model studies with phosphate triesters indicate a large elemental effect upon substitution of sulfur at a nonesterified position, whereas studies with phosphate diesters indicate smaller decreases in rate upon sulfur substitution (see Ref. 31 for discussion). From the dCTP␣S concentration dependence of the rate of incorporation, the affinity (K d ) and rate constant for incorporation of the analogue were 9.8 M and 4.7 s Ϫ1 , respectively (data not shown). Thus, the thioelemental effect observed for pol ␤ is only 2.1, suggesting that a step other than chemistry is rate-limiting.
Steady-state kinetic characterization of mouse pol ␤ failed to demonstrate a reversal of the polymerization reaction (i.e. pyrophosphorolysis) in the presence of PP i with activated DNA. Nonetheless, PP i was inhibitory for DNA synthesis (13). To ascertain the possible significance of pyrophosphorolysis on single-nucleotide gap filling DNA synthesis, nicked DNA was used as a substrate to measure the magnitude of the reverse reaction. The substrate was identical to that used in the onenucleotide gap filling reaction except that the primer was one nucleotide longer (i.e. ϩdCMP). Pyrophosphorolysis was initiated by the addition of excess PP i (3 mM) and MgCl 2 to pol ␤ preincubated with DNA. The steady-state removal of dCMP to create a one-nucleotide gap occurred with an apparent rate constant of 0.01 s Ϫ1 (data not shown).
Affinity of DNA Polymerase ␤ for Gapped DNA-The biphasic time course observed when DNA/pol Ͼ 1 (Fig. 2) indicates formation of a "stable" pol-DNA complex. When k pol Ͼ Ͼ k off(DNAϩ1) , the amplitude of the rapid (burst) phase is equal to the active-enzyme concentration, but in the case of pol ␤ is only  (Table II). proportional to the concentration of pol-DNA complex, as dis- The K d and maximum burst amplitude were 22 and 70 nM, respectively. From Equation 3 and k pol (Fig. 3B and Table II), the maximum burst amplitude can be used to calculate k off(DNAϩ1) . This results in a dissociation rate constant of 1.9 s Ϫ1 and in turn can be used to estimate an association rate constant of 8.6 ϫ 10 7 M Ϫ1 s Ϫ1 (K d ϭ k off /k on ).
Influence of DNA Gap on Nucleotide Affinity-As shown in Table I, pol ␤ fills single-nucleotide gapped DNA substrates nearly 15 times more efficiently than non-gapped substrates under steady-state conditions. To determine if the binding affinity of the incoming nucleotide (i.e. K d(dCTP) ) is influenced by the single-nucleotide gap, single-turnover analysis (pol/DNA ϭ 5) was employed to quantify nucleotide binding to a non-gapped DNA substrate (the downstream primer was omitted). Using this non-gapped DNA substrate, pol ␤ was found to bind and incorporate dCTP with K d of 22 M and k pol of 5 s Ϫ1 (Table II). This represents a 4-fold decrease in binding affinity as compared with a gapped DNA complex and an 8-fold decrease in insertion efficiency.
Role of Residue 276 in the Nucleotide Binding Pocket-The structure of the pol ␤-DNA-ddCTP complex reveals that there are van der Waals interactions between the base moiety of the incoming nucleotide and the C␤ of Asp 276 , Fig. 5 (7, 8). Steadystate kinetic analysis and cross-linking studies previously suggested that removal of the electronegative side chain by valine or glycine substitution resulted in an apparent increase in binding affinity (32). To determine whether the binding affinity for the incoming nucleotide is altered by removing the negative charge, but retaining van der Waals interactions, the K d(dCTP) was determined for the D276V mutant. Single-turnover analysis employing both non-gapped and gapped DNA substrates indicated that the K d for the incoming nucleoside triphosphate decreased 3.6-and 9-fold, respectively, as compared with wildtype enzyme (Table II). Thus, the binding affinity for the incoming nucleotide is increased nearly 40-fold for the mutant enzyme when utilizing a gapped DNA substrate as compared with wild-type enzyme on a non-gapped DNA substrate.
Incorrect Nucleotide Binding-To investigate whether Asp 276 participates in the binding of an "incorrect" nucleotide (e.g. dTTP), the binding affinity of thymidine triphosphate with a gapped DNA substrate was determined for wild-type pol ␤ and the D276V mutant. For the kinetic mechanism described above (Scheme I), when k pol becomes very slow (i.e. rate-limiting), then K m(dNTP) is equivalent to K d(dNTP) (33)(34)(35). DNA polymerases generally discriminate against incorrect nucleotides by binding them weakly and incorporating them slowly. Thus, incorporation of an incorrect nucleotide is always ratelimiting for polymerases that exhibit low processivity, where processivity is defined as a competition between nucleotide incorporation and polymerase dissociation from the templateprimer (i.e. k pol /k off(DNA) ). This analysis does not require that the active fraction of enzyme be known. When the dTTP concentration dependence of incorporation opposite the template deoxyguanine in the single-nucleotide gapped DNA substrate was examined, the K m(dTTP) was 330 Ϯ 60 and 130 Ϯ 20 M (five independent determinations) for the wild-type and the D276V enzymes, respectively. Therefore, valine substitution at Asp 276 modestly increased (2.5 Ϯ 0.6-fold) the binding affinity of dTTP opposite a templating deoxyguanine. DISCUSSION Nucleic acid polymerases must select (bind and incorporate) the correct nucleotide from a pool of structurally similar molecules to ensure accurate and efficient nucleic acid synthesis. DNA polymerase binding of the incoming nucleoside triphosphate has been mechanistically and kinetically described as occurring in at least two steps. After initial complex formation, an isomerization of the polymerase ternary complex leads to a productive catalytic complex where chemistry occurs rapidly (Scheme II). Each of the steps illustrated in Scheme II (groundstate binding, isomerization, chemistry) offers an opportunity for selection of the correct nucleotide.
The contribution of each of these steps to the accuracy of nucleotide selection for DNA synthesis depends on the specific DNA polymerase. Whereas polymerases generally bind the incorrect nucleotide weakly and incorporate (k pol ) them slowly (e.g. T7 DNA polymerase and HIV-1 reverse transcriptase), the Klenow fragment of E. coli pol I does not appear to utilize ground-state binding to increase selectivity (K d(correct) Ϸ K d(incorrect) ) (36). The lack of a significant elemental effect when the non-bridging oxygen on the ␣-phosphorus is substituted with sulfur is consistent with a rate-limiting conformational change in a two-step dNTP binding mechanism for human DNA polymerase ␤. This interpretation is supported by fluorescence changes of the template base analogue 2-aminopurine that occur upon nucleotide binding (37). The environment of the templating base, as monitored by fluorescence, changes at a rate that is similar to nucleotide incorporation even when incorporation is prevented (i.e. no chemistry) by using a dideoxynucleotide-terminated primer. A similar approach had been used to separate chemical and non-chemical steps of nucleotide binding with Klenow fragment and T4 DNA polymerase (15).
DNA polymerase ␤ utilizes ground-state binding to selectively incorporate dCTP opposite a templating deoxyguanine. The selectivity over thymidine triphosphate (K d(dTTP) / K d(dCTP) ) is approximately 60. This is similar to the selectivity of 25 (38) and 120 (39) reported for HIV-1 reverse transcriptase, but less than the selectivity of 410 for T7 DNA polymerase (40), when considering formation of the same mispair. The selectivity of the D276V mutant was moderately increased to 220 by binding the correct deoxynucleoside triphosphate more tightly than the incorrect thymidine triphosphate relative to wild-type enzyme.
Steady-state and pre-steady-state kinetics indicate that DNA polymerases bind substrates in an ordered fashion; DNA binding precedes nucleotide binding. However, this is clearly not obligatory. For example, utilizing photoreactive nucleotide analogues, it is possible to covalently cross-link these analogues to pol ␤ and other DNA polymerases in the absence of DNA. Upon the addition of DNA, only analogues cross-linked at the polymerase active site are incorporated, resulting in DNA that becomes cross-linked to the polymerase (32). Additionally, crystal structures of binary pol-dNTP complexes reveal that the dNTP is bound in the active site through its triphosphate moiety (41)(42)(43). The triphosphate moiety, therefore, offers minimal specificity toward accuracy for correct nucleotide binding. The sugar also affords some specificity, since 2Ј-deoxyribose is preferred over ribose in the presence of DNA. However, discrimination during dNTP binding originates mainly from the identity of the base and its hydrogen bonding capacity and steric complementarity with the templating base. Ground-state binding of a nucleotide offers the opportunity to check proper Watson-Crick hydrogen bonding and/or steric complementarity. Formation of a base pair with good geometry induces a conformational change that results in rapid incorporation of the nucleotide. Thus, the rate of insertion is limited by a conformational change that triggers chemistry. The identity in structural terms of this rate-limiting conformational change will be discussed here.
Several significant structural changes occur upon substrate (DNA and dNTP) binding to pol ␤ and other DNA polymerases as inferred from comparison of structures of apoenzyme with substrate bound complexes. These conformational changes occur with both the polymerase and the substrate. The ratelimiting conformational change observed kinetically has been postulated to be the large subdomain movement that is observed by comparing the crystal structures of open pol and pol-DNA complexes with those of the closed ternary pol-DNA-dNTP complexes (see Ref. 44 for a review). For pol ␤, in addition to significant movement of the 8-kDa lyase domain upon binding gapped DNA, the carboxyl-terminal subdomain is observed to "close" upon the exposed face of the nascent base pair. Rotation around an axis, ␣-helix M, positions ␣-helix N to probe correct Watson-Crick geometry (7,8,24). Superimposing the polymerase catalytic subdomains of the binary one-nucleotide gapped DNA substrate with that including the incoming ddCTP (ternary complex) illustrates this subdomain movement (Fig. 6A) and reveals that the 8-kDa domain moves slightly to form an even more compact structure (not shown) (24). Numerous active site side chain and substrate motions are associated with these conformational differences. Assuming that valine substitution for Asp 276 does not produce additional active site structural changes, the finding that interactions between Asp 276 and the incoming dNTP influence ground-state dNTP binding implies that the conformational change limiting insertion during the first turnover may not involve movement of ␣-helix N. This follows because C␤ of Asp 276 must be in close contact with the base of the incoming dNTP during template base annealing. The van der Waals interaction between C␤ of Asp 276 and the correct dNTP is observed in the closed conformation indicating that the closed conformation itself may trigger the rate-limiting conformational change. A "structural signal" to indicate that positioning of ␣-helix N in the closed position has occurred could be transmitted to the catalytic metal-binding site through the altered positioning observed with Arg 283 , Glu 295 , Arg 258 , and Asp 192 in the open and closed forms of pol ␤ (Fig. 6B). From a comparison of open and closed forms of pol ␤ ternary complexes, it has also been suggested that thumb closure repositions the mononucleotide binding motif (residues 179 -189) closer to the primer 3Ј-hydroxyl (8).
DNA polymerase ␤ is an integral component of the BER pathway. It is responsible for removing the 5Ј-dRP flap and filling the one-nucleotide gap generated during BER (for a review, see Refs. 45 and 46). DNA polymerase ␤ processively fills short DNA gaps (ϭ 5 nucleotides) where the downstream DNA is 5Ј-phosphorylated (23,47). In contrast, DNA synthesis occurs in a distributive fashion with non-gapped DNA. This is attributed to the 5Ј-phosphate binding activity of the aminoterminal 8-kDa domain (23). The 8-kDa domain interacts with the carboxyl-terminal domain when bound to a one-nucleotide DNA gap (Fig. 1A). This interaction buries approximately 130 Å 2 of surface area. In the closed ternary complex of the onenucleotide gap DNA structure, this solvent-inaccessible interface increases to 300 Å 2 .
We have examined the influence of the template-primer structure on nucleotide binding by pol ␤. Previous studies that examined the influence of template-primer structure on the catalytic efficiency of pol ␤ have reported dissimilar effects. In one study, the catalytic efficiency was increased over 2500-fold when using a single-nucleotide gap, as compared with a nongapped DNA substrate (48). In contrast, only a modest influence (10-fold) of gap structure on catalytic efficiency has also been reported (49). In the current study, the catalytic efficiency increased about 10-fold as determined from steady-state (Table  I) or pre-steady-state kinetics (Table II) with a one-nucleotide DNA gap, as compared with non-gapped DNA. With a template-primer system that is typically employed for kinetic study of DNA polymerases (non-gapped DNA), pol ␤ binds the correct nucleotide with a lower affinity than when bound to gapped DNA (Table II). This could be due to the role that the aminoterminal 8-kDa domain may play in forming the dNTP binding pocket. In the absence of downstream DNA (i.e. no gap), the dNTP binding site is "exposed" relative to that observed in the ternary complex with a one-nucleotide gapped DNA (Fig. 1). As illustrated in Fig. 1, this results in a doughnut-like structure that apparently gates dNTP binding through a cavity leading to the polymerase active site. This type of polymerase architec- ture is similar to that observed with B-type replicative polymerases (50,51), as well as with the RNA-dependent RNA polymerase from hepatitis C (52,53).
The modestly higher dNTP binding affinity when pol ␤ utilizes a one-nucleotide gapped DNA substrate indicates a role for the 8-kDa domain in dNTP binding with the one-nucleotide gapped DNA substrate. Additionally, both wild-type enzyme and the D276V mutant exhibit a 2-fold increase in the rate of nucleotide insertion (k pol ) with a one-nucleotide gapped DNA substrate (Table II). A strong hydrogen bond between Asp 276 of ␣-helix N and Arg 40 (␣ -helix B) of the 8-kDa lyase domain in the closed ternary complex is the only electrostatic interaction between the 8-kDa domain and ␣-helix N. The increased binding affinity with gapped DNA suggests that the negative charge contributed by Asp 276 restricts binding (reduced association and/or increased dissociation) and that Arg 40 may increase binding affinity in the gapped structure by charge compensation. Except for that contributed by Asp 276 , the electrostatic surface potential around the sugar and base of the incoming nucleotide is positive (Fig. 7). It is noteworthy that most other X family polymerases have an arginine or lysine at this position (54). Additionally, Arg 72 of HIV-1 reverse transcriptase is observed to stack with the incoming dNTP and form hydrogen bonds with the ␣-phosphate. It is highly conserved among retroviral reverse transcriptases and telomerase. In contrast to the results reported here for Asp 276 , modification of the Arg 72 side chain results in mutant enzymes that have dramatically lower catalytic efficiency (55,56). In the case of pol ␤, it appears that the surrounding positive potential (i.e. blue) can facilitate dNTP binding with a one-nucleotide gapped DNA substrate. This interpretation is consistent with the observation that removal of the acidic side chain by valine substitution (D276V) results in ground-state binding affinity on non-gapped DNA similar to that of wild-type enzyme on gapped DNA (Table II). In this situation, the positive potential is removed by side chain replacement, whereas in the latter situation, the highly basic 8-kDa domain facilitates binding through charge neutralization.
When a one-nucleotide gapped DNA substrate is utilized by the D276V mutant polymerase, the binding affinity for the incoming nucleotide is increased even further (Table II). This result indicates that additional interactions and conformational changes give rise to an increased free energy of binding. The additional methylene-group in the valine side chain, as compared with aspartate, may increase the van der Waals contacts with the base of the incoming nucleotide.
The low processivity of pol ␤ makes it difficult to accurately analyze the biphasic time courses of product formation when DNA is in excess by pre-steady-state methods (enzyme/DNA Ͻ 1; Fig. 2). This is because k pol and k off are both partially ratelimiting, and product inhibition makes it difficult to estimate the linear steady-state rate of product formation. Processivity is kinetically defined as the probability of inserting a nucleotide (i.e. k pol ) to that of dissociating from the DNA substrate (i.e. k off ). Thus, with a one-nucleotide gapped DNA substrate the processivity of DNA polymerase ␤ is modest (k pol /k off ϭ 10 s Ϫ1 /1.9 s Ϫ1 ϭ 5.3). This is consistent with the qualitative gel assay that had indicated pol ␤ can processively fill short gaps of up to 6 nucleotides (47). Single-turnover analysis (enzyme/ DNA Ͼ Ͼ 1) of dNTP binding to DNA polymerase ␤ eliminates enzyme cycling. From the nucleotide-dependent exponential time courses, the ground-state binding affinity for dCTP and k pol for the wild-type and D276V polymerases were determined ( Fig. 3; Table II). The amplitude of the exponential time courses should be equivalent to the concentration of DNA. Under the conditions of the assay, 95% of the DNA should be enzyme-bound. For several determinations, the amplitude of the rapidexponential phase was 63 Ϯ 8% and 37 Ϯ 6% of the extendable DNA for the wild-type and D276V polymerases, respectively. This indicates that a population of these polymerases can bind in a nonproductive conformation that is sensitive to the valine substitution at Asp 276 . To determine whether this nonproductive DNA binding form could isomerize to a productive complex without dissociating from DNA, unlabeled DNA was added to trap enzyme that may dissociate during the course of the reaction. The excess non-radioactively labeled DNA trap had no effect on the amplitudes or rates of the rapid or slow phases, indicating that for the wild-type enzyme that the nonproductive DNA binding form can isomerize to the productive form without dissociating from the DNA. Similar results were observed with the mutant enzyme except that the slow phase was partially sensitive to the added trap. A thorough kinetic and thermodynamic analysis of these alternate binding modes is currently under study.
The alternate binding modes result in a decreased population of polymerase poised for nucleotide insertion. Suo and Johnson (57,58) observed similar kinetic behavior when HIV-1 reverse transcriptase binds templates with secondary structure and suggested that the template nucleic acid is blocking access to the pol dNTP binding pocket. A similar proposal for pol ␤ suggests that the DNA substrate may bind at the polymerase active site such that the terminal template-primer base pair is sitting where the nascent base pair would form. In this conformation the template-primer is out of register for DNA synthesis and would block an incoming nucleotide from entering the polymerase active site. This conformation is observed in the binary complex of pol ␤ bound to nicked DNA. In addition, the ability of DNA polymerases to catalyze pyrophosphorolysis requires that the terminal primer nucleotide move into the polymerase active site to achieve PP i addition. As the valine substitution at Asp 276 stabilizes the incoming nucleotide sitting in the active site, it also may stabilize the terminal primer nucleotide in the polymerase active site. This is consistent with the lower apparent active concentration of the mutant polymerase.
In conclusion, the dNTP binding affinity is increased for pol ␤ with a one-nucleotide gapped DNA substrate, suggesting a role for the amino-terminal lyase domain in nucleotide binding. Interestingly, altering the interactions with the incoming nucleotide and/or lyase domain by site-directed mutagenesis of Asp 276 also increased the binding affinity of the correct incoming nucleotide. This is the first example that we are aware of where strategic modification of the dNTP binding pocket has resulted in an increased nucleotide binding affinity. It will be interesting to determine if the observed changes in binding affinity are due to "ionic tethering" (59) where Asp 276 and Arg 40 modulate domain interactions to influence ligand binding. Finally, since this interaction between Asp 276 and the incoming dNTP is observed only in the closed conformation of pol ␤, the increase in the free energy in ground-state binding for the D276V mutant suggests that the rate-limiting conformational change is not the open to closed structural transition, but instead is triggered in the closed polymerase conformation. Deoxynucleoside triphosphate binding, therefore, occurs in at least three steps. First, there is an initial nonspecific binding that occurs primarily through polymerase interactions with the triphosphate/metal and sugar moieties of the nucleotide. This step is weak so as to facilitate rapid sampling of the nucleoside triphosphate pools and is kinetically silent. This sampling is probably associated with the subdomain movements inferred from the crystal structure differences between the binary pol-DNA and ternary substrate complexes. The second step, ground-state binding, is associated with template base recognition through hydrogen bonding and/or steric complementarity. In turn, the degree of complementarity influences the probability of triggering a rate-limiting conformational change (i.e. step three) that leads to a catalytically competent complex that induces chemistry.