Increased Activity and Fidelity of DNA Polymerase β on Single-nucleotide Gapped DNA*

DNA polymerase β (pol β) is an error-prone polymerase that plays a central role in mammalian base excision repair. To better characterize the mechanisms governing rat pol β activity, we examined polymerization on synthetic primer-templates of different structure. Steady-state kinetic analyses revealed that the catalytic efficiency of pol β (k cat/K m,dNTP app) is strongly influenced by gap size and the presence of a phosphate group at the 5′-margin of the gap. pol β exhibited the highest catalytic efficiency on 5′-phosphorylated 1-nucleotide gapped DNA. This efficiency was ≥500 times higher than on non-phosphorylated 1-nucleotide and 6-nucleotide (with or without PO4) gapped DNAs and 2,500 times higher than on primer-template with no gaps. The nucleotide insertion fidelity of pol β, as judged by its ability to form G-N mispairs, was also higher (10–100 times) on 5′-phosphorylated single-nucleotide gapped DNA compared with the other DNA substrates studied. These data suggest that a primary function of mammalian pol β is to fill 5′-phosphorylated 1-nucleotide gaps.

The biochemical activities of purified pol ␤ are consistent with a role in gap-filling DNA synthesis. Early studies showed that pol ␤ is non-processive on single-stranded DNA templates, prefers short-gapped DNA substrates, and is capable of filling gaps to completion (11)(12)(13)(14)(15)(16). More recently, Wilson and colleagues (17) observed that pol ␤ fills short gaps (2-6 nt) by a processive mechanism that requires a PO 4 group at the 5Јmargin of the gap. Binding of pol ␤ to these short-gapped substrates is also strongly enhanced by the presence of a 5Ј-PO 4 (18). These experiments, together with recent structural data, suggest a model in which pol ␤ binding to gapped DNA is mediated by interactions between the 8-kDa domain of pol ␤ and the 5Ј-PO 4 at the downstream margin of the gap (7,18). Processive DNA synthesis on short (2-6-nt) gaps is consistent with roles for pol ␤ in long-patch BER (4) and in the completion of gap-filling synthesis initiated by other cellular DNA polymerases (9, 10, 14 -16).
Although DNA polymerization by pol ␤ on single-stranded and short-gapped DNAs is understood in some detail, much less is known about pol ␤ activity on its short-patch BER substrate, 1-nt gapped DNA. The model of pol ␤ binding through its 8-kDa domain to the 5Ј-PO 4 in short-gapped DNA does not appear to apply to 1-nt gaps; reducing the gap size from 5 to 1 nt decreases binding slightly, and the 5Ј-phosphorylation requirement is lost (18). This suggests that pol ␤ may interact with 1-nt gapped DNA by a distinct mechanism.
To better characterize the parameters governing pol ␤ activity on 1-nt gapped DNA, we examined the steady-state kinetics of DNA polymerization on synthetic primer-templates of different structure. We show that the catalytic efficiency (k cat / K m,dNTP app ) and nucleotide insertion fidelity of pol ␤ are strongly influenced by gap size and that the 5Ј-phosphorylation requirement is retained for these activities even on 1-nt gapped DNA. These data have important implications for models of pol ␤ DNA binding and provide biochemical evidence that 5Ј-phosphorylated 1-nt gapped DNA is the preferred substrate for pol ␤.

EXPERIMENTAL PROCEDURES
Materials-Recombinant rat DNA polymerase ␤ was purified as described previously (19). All oligonucleotides were synthesized and high pressure liquid chromatography-purified by Operon Technologies. 5Ј-32 P Labeling of the primers was performed with [␥-32 P]ATP (3,000 Ci/mmol; Amersham Corp.) using T4 polynucleotide kinase (U. S. Biochemical Corp.) according to the manufacturer's protocol. Labeled primers were separated from excess [␥-32 P]ATP after labeling by gel filtration through 0.5-ml Sephadex G-50 (Pharmacia Biotech Inc., DNA grade) spin columns. 2Ј-Deoxyribonucleoside 5Ј-triphosphates (dNTPs) were from Calbiochem or Pharmacia. Concentrations of individual dNTPs were determined by UV spectroscopy (Beckman DU65). Protein concentrations were determined by the method of Bradford (Bio-Rad) according to the manufacturer's protocol. All other reagents were of the highest grade available from Fisher Scientific or Sigma.
Steady-state Kinetics of Single-nucleotide Insertion-The kinetics of dNMP incorporation opposite template positions G 21 (see Fig. 1) were determined in polymerization reactions (10 or 20 l) containing 0.1-20 nM pol ␤, 20 nM primer-template, and 0 -5,000 M of a single dNTP in 50 mM Tris-HCl, pH 8.0 (22°C), 10 mM MgCl 2 , 2 mM dithiothreitol, 20 mM NaCl, 20 mM KCl, 2.5% glycerol, 0.2 mg/ml bovine serum albumin. Primer-templates were first incubated with pol ␤ for 5 min at 37°C in the absence of dNTPs, and then polymerizations were initiated by the addition of a single dNTP. After continued incubation for 3-15 min at 37°C, reactions were terminated by adding 0.1 volume of 0.5 M EDTA. 1-2-l aliquots were removed and mixed with 5 l of formamide loading dye (20), boiled for 5 min, and immediately transferred into an ice slurry for 5 min. Products were resolved by PAGE (7 M urea, 16% acrylamide) and then visualized and quantified using a Phosphor-Imager and Imagequant software (Molecular Dynamics). Reaction times and enzyme concentrations were adjusted for each substrate to optimize product detection while ensuring that all reactions were conducted in the steady state. Only those reactions that fell within the linear range of substrate utilization (Յ20% primer extension) were used for kinetic analyses.
Steady-state kinetic analyses were based on the Michaelis-Menten equation. For correct dCMP incorporation, k cat and K m,dNTP app values were determined using a non-linear curve fitting program (SigmaPlot). For mispairs, k cat /K m,dNTP app values were determined from the initial slopes of Michaelis-Menten plots (20,21), and the frequencies of misincorporation were calculated as described (22). To detect extension products resulting from dNMP misincorporation, it was often necessary to increase pol ␤ concentrations and/or incubation times. As expected in steady state (22), V max values were directly proportional to enzyme concentration (data not shown).
Processivity-The processivity of pol ␤ was determined on the same substrates used in the kinetic assays (see Fig. 1) and under similar conditions, except the reactions were started by adding all four dNTPs at saturating concentrations (1.25 mM each). Linear regions of product yield versus pol ␤ concentration curves were used to quantify average (statistically weighted) processivities. Statistical weighting was performed by multiplying the lengths of the products by the relative intensities of corresponding bands on the gel.

RESULTS
Design of Recessed and Gapped DNA Substrates-To examine pol ␤ activity on DNA substrates of different structure, a series of primer-templates was constructed (Fig. 1). These DNAs all contained the same 46-mer template sequence based on a region of bacteriophage X174 DNA used in previous fidelity studies (Refs. 20 and 23, and references therein). All of the substrates also contained the same [5Ј-32 P]20-mer primer hybridized to template residues 22-41. This places the primer 3Ј-OH terminus such that polymerization of the first dNTP  (Fig. 1). pol ␤ concentrations and incubation times were adjusted to optimize detection of primer extension products; only data obtained from reactions conducted in the steady state (i.e. Յ20% primer extension) were used in the kinetic analyses in Tables I and II an asterisk in B). Each 21-mer mobilized at a characteristic rate, with the correct C-containing 21-mer running 1.5-3 mm ahead of the incorrect G-, A-and T-containing 21-mers. The 22-mer products formed on gap-1 (A) presumably result from partial displacement of the downstream oligonucleotide and incorporation of the next correct nucleotide dATP (see Fig. 1). occurs opposite template G 21 (which corresponds to residue 587 in X174 DNA). The simplest DNA substrate, comprised of [5Ј-32 P]20-mer primer hybridized to 46-mer template, had a "recessed" primer with 21 nt of downstream single-stranded DNA template (Fig. 1A). Two gapped substrates were also constructed ( Fig. 1, B and C). The substrate designated gap-6 contained a second oligonucleotide (15-mer) hybridized to template residues 1-15, thereby creating a primer-template with a 6-nt gap immediately downstream from the 5Ј-32 P-primer 3Ј-OH (Fig. 1B). A similar substrate with a 1-nt gap (designated gap-1) was constructed by hybridizing a 20-mer oligonucleotide to template residues 1-20 (Fig. 1C). Variants of gap-6 and gap-1 containing PO 4 moieties at the 5Ј margins of the gaps (designated P-gap-6 and P-gap-1, respectively) were also made by starting with 5Ј-phosphorylated oligonucleotides.
Proper assembly of these oligonucleotides into the desired structures was confirmed in two ways. First, native PAGE showed that hybridization efficiencies were Ͼ95% as evidenced by different mobilities of the [5Ј-32 P]20-mer primer before and after hybridization to the 46-mer template alone and in combination with the downstream oligonucleotides (data not shown). As a second indirect way of confirming structure, we examined the processivity of pol ␤ on these DNA substrates. This approach is based on the observation of Singhal and Wilson (17) that pol ␤ is distributive on recessed primer-templates and short non-phosphorylated gapped DNAs but processive on short gaps containing 5Ј-phosphates. We observed the same trend on our substrates (Fig. 2). Average processivities on the recessed, gap-6, and P-gap-6 DNAs were 1.3, 1.1, and 3.5, respectively. Thus, our data confirm the results of Singhal and Wilson (17), although synthesis on our phosphorylated 6-nt gapped substrate was not strictly processive (Fig. 2, right).
Effect of DNA Substrate Structure on pol ␤ Catalytic Efficiency-We performed steady-state kinetic analyses of singlenucleotide addition (dCMP) opposite template G 21 on the different DNA substrates (Fig. 3, dCTP reactions, and Table I) (Table I; compare gap-6 with gap-1). Most striking, however, was the dramatic effect of 5Ј-phosphorylation on the 1-nt gapped substrate, where addition of a 5Ј-PO 4 resulted in a 500-fold increase in catalytic efficiency (compare gap-1 to P-gap-1). Thus, the relative catalytic efficiencies of pol ␤ on the different DNA substrates were P-gap-1 Ͼ Ͼ gap-1 Ϸ P-gap-6 Ͼ gap-6 Ϸ recessed. pol ␤ was some 10,000 and 2,500 times more efficient on P-gap-1 than on the gap-6 and recessed DNA substrates, respectively. As noted above, this increase in catalytic efficiency resulted primarily from a decrease in K m,dCTP app , although k cat values were also slightly higher on Pgap-1. The k cat value of 0.6 s Ϫ1 observed on the recessed DNA substrate is very similar to the value of 0.3 s Ϫ1 reported for a different pol ␤ preparation on a different recessed primertemplate (24).
Effect of DNA Substrate Structure on pol ␤ Fidelity-The nucleotide insertion fidelity of pol ␤ was determined on the same series of DNA substrates using a "standing start" (22) kinetic fidelity assay (Fig. 3 and Table II). The frequencies of nucleotide misinsertions opposite the template G 21 residue were similar for all substrates except P-gap-1. The fidelity of pol ␤ on P-gap-1 was 100, 50, and 30 times higher for G-T, G-G, 2 B. D. Preston, unpublished data.  Fig. 3 shows representative gels used for analysis. a Sequences and structures of the DNA substrates are shown in Fig. 1.
b Calculated using total pol ␤ protein concentration. c Values in parentheses indicate the number of independent experiments used for each analysis.  Fig. 3 shows representative gels used for analysis. Mispair formation frequencies were calculated from the initial slopes of Michaelis-Menten curves using the formula: f ins ϭ (k cat /K m,dNTP app ) incorrect / (k cat /K m,dNTP app ) correct (21,22), where "correct" corresponds to extension in the presence of dCTP to form the G-C base pair (Table I).
a Sequences and structures of the DNA substrates are shown in Fig.  1.
b Values in parentheses indicate the number of independent experiments used for each analysis.
pol ␤ Activity and Fidelity and G-A mispair formation, respectively, compared with the recessed substrate. G-T and G-A mispairs were formed ϳ10fold more readily than G-G mispairs on all of the DNA substrates studied. DISCUSSION pol ␤ plays a central role in mammalian short-patch BER (1)(2)(3)(4)(5). This suggests that a preferred substrate for pol ␤ might be 5Ј-phosphorylated 1-nt gapped DNA. We examined the DNA substrate preferences of purified rat pol ␤ in steady-state kinetic assays using synthetic DNAs of different structure. We show that pol ␤ prefers 5Ј-phosphorylated 1-nt gapped DNA as substrate with relative catalytic efficiencies on P-gap-1 Ͼ Ͼ gap-1 Ϸ P-gap-6 Ͼ gap-6 Ϸ recessed (Table I). The efficiency of pol ␤ on P-gap-1 DNA was 500 -10,000 times higher than on the other DNA substrates examined. We also observed that the frequency of nucleotide misinsertion by pol ␤ was 10 -100-fold lower on P-gap-1 compared with the other DNA substrates (Table II).
Singhal and Wilson (17) showed that pol ␤ switches from a distributive to a processive mode of DNA polymerization on short-gapped (2-6 nt) DNA substrates but only if the 5Ј-margin of the gap is phosphorylated. The very similar effects observed in our processivity experiments using different oligonucleotides (Fig. 2) indicate that this is an intrinsic property of pol ␤ that has no obvious requirement for specific template sequences. Our steady-state kinetic analyses show that the catalytic efficiency and nucleotide insertion fidelity of pol ␤ are also influenced by gap size and 5Ј-phosphorylation. Moreover, in contrast to what is obserbed for pol ␤ binding to DNA (18), 5Јphosphorylation is required for both high catalytic efficiency and increased fidelity on 1-nt gapped DNA (Tables I and II). These data extend the model of Prasad et al. (18) by showing that 5Ј-PO 4 residues must mediate a productive catalytic interaction between pol ␤ and DNA even in 1-nt gaps.
The relative low fidelity of pol ␤ observed on the recessed primer-template (f ins ϭ 10 Ϫ3 -10 Ϫ4 ; Table II) is comparable with that reported by others on recessed DNA substrates (21,23,24). However, our observation of similar fidelities on recessed and P-gap-6 DNAs appears to conflict with recent reports suggesting that pol ␤ is less faithful during 5-and 6-nt gap-filling synthesis (17,25). This apparent discrepancy may relate to the overall higher catalytic efficiency of pol ␤ on phosphorylated short-gapped DNA (Table I), to template sequence effects, and/or to differences in the assays used to measure fidelity. Additional experiments are required to resolve this. Regardless, our data showing increased fidelity on Pgap-1 DNA indicate that pol ␤, and by inference BER, may be less error prone than once thought.
Several mechanisms may contribute to the observed effects of gap structure on catalytic efficiency and fidelity. Based on the binding studies of Prasad et al. (18), it appears that the differences in catalytic efficiency on P-gap-1 and gap-1 are not due to differences in the levels of stable DNA binding (at least for pol ␤-DNA binary complexes detected by cross-linking and competition assays). An attractive general hypothesis is that the 5Ј-PO 4 in a 1-nt gap somehow facilitates formation of a catalytically optimal pol ␤-DNA complex without affecting overall binding affinity. Amino acid changes at residues distant from the polymerase active site of pol ␤ were recently shown to affect the fidelity of DNA synthesis (19). This indicates that molecular events at the active site respond to long range changes in the pol ␤ protein. Thus, interactions between the DNA 5Ј-PO 4 and the 8-kDa domain of pol ␤, which also occur at some distance from the active site (7), may remotely alter dNTP binding and/or protein conformational changes required for chemical catalysis (24,26). Additional kinetic and structural studies will be required to delineate the contribution of these and other mechanisms to pol ␤ substrate recognition and catalytic efficiency. It is particularly germane to examine the role of the 8-kDa domain in directing the interaction of pol ␤ with P-gap-1 DNAs (7,18).
In summary, we show that purified pol ␤ exhibits relative high catalytic efficiency and fidelity on 5Ј-phosphorylated 1-nt gapped DNA in vitro. This suggests that a primary biochemical function of pol ␤ in the mammalian cell is to fill 5Ј-phosphorylated 1-nt gaps. Gaps with this structure appear to be requisite intermediates in short-patch BER (1)(2)(3)(4) and may exist in other pathways where involvement of pol ␤ is implicated (4, 8 -10).