Enzyme-DNA interactions required for efficient nucleotide incorporation and discrimination in human DNA polymerase beta.

In the crystal structure of a substrate complex, the side chains of residues Asn279, Tyr271, and Arg283 of DNA polymerase β are within hydrogen bonding distance to the bases of the incoming deoxynucleoside 5′-triphosphate (dNTP), the terminal primer nucleotide, and the templating nucleotide, respectively (Pelletier, H., Sawaya, M. R., Kumar, A., Wilson, S. H., and Kraut, J.(1994) Science 264, 1891-1903). We have altered these side chains through individual site-directed mutagenesis. Each mutant protein was expressed in Escherichia coli and was soluble. The mutant enzymes were purified and characterized to probe their role in nucleotide discrimination and catalysis. A reversion assay was developed on a short (5 nucleotide) gapped DNA substrate containing an opal codon to assess the effect of the amino acid substitutions on fidelity. Substitution of the tyrosine at position 271 with phenylalanine or histidine did not influence catalytic efficiency (kcat/Km) or fidelity. The hydrogen bonding potential between the side chain of Asn279 and the incoming nucleotide was removed by replacing this residue with alanine or leucine. Although catalytic efficiency was reduced as much as 17-fold for these mutants, fidelity was not. In contrast, both catalytic efficiency and fidelity decreased dramatically for all mutants of Arg283 (Ala > Leu > Lys). The fidelity and catalytic efficiency of the alanine mutant of Arg283 decreased 160- and 5000-fold, respectively, relative to wild-type enzyme. Sequence analyses of the mutant DNA resulting from short gap-filling synthesis indicated that the types of base substitution errors produced by the wild-type and R283A mutant were similar and indicated misincorporations resulting in frequent T•dGTP and A•dGTP mispairing. With R283A, a dGMP was incorporated opposite a template thymidine as often as the correct nucleotide. The x-ray crystallographic structure of the alanine mutant of Arg283 verified the loss of the mutated side chain. Our results indicate that specific interactions between DNA polymerase β and the template base, but not hydrogen bonding to the incoming dNTP or terminal primer nucleotide, are required for both high catalytic efficiency and nucleotide discrimination.

In the crystal structure of a substrate complex, the side chains of residues Asn 279 , Tyr 271 , and Arg 283 of DNA polymerase ␤ are within hydrogen bonding distance to the bases of the incoming deoxynucleoside 5-triphosphate (dNTP), the terminal primer nucleotide, and the templating nucleotide, respectively (Pelletier, H., Sawaya, M. R., Kumar, A., Wilson, S. H., and Kraut, J. (1994) Science 264, 1891-1903). We have altered these side chains through individual site-directed mutagenesis. Each mutant protein was expressed in Escherichia coli and was soluble. The mutant enzymes were purified and characterized to probe their role in nucleotide discrimination and catalysis. A reversion assay was developed on a short (5 nucleotide) gapped DNA substrate containing an opal codon to assess the effect of the amino acid substitutions on fidelity. Substitution of the tyrosine at position 271 with phenylalanine or histidine did not influence catalytic efficiency (k cat /K m ) or fidelity. The hydrogen bonding potential between the side chain of Asn 279 and the incoming nucleotide was removed by replacing this residue with alanine or leucine. Although catalytic efficiency was reduced as much as 17-fold for these mutants, fidelity was not. In contrast, both catalytic efficiency and fidelity decreased dramatically for all mutants of Arg 283 (Ala > Leu > Lys). The fidelity and catalytic efficiency of the alanine mutant of Arg 283 decreased 160-and 5000-fold, respectively, relative to wild-type enzyme. Sequence analyses of the mutant DNA resulting from short gap-filling synthesis indicated that the types of base substitution errors produced by the wild-type and R283A mutant were similar and indicated misincorporations resulting in frequent T⅐dGTP and A⅐dGTP mispairing. With R283A, a dGMP was incorporated opposite a template thymidine as often as the correct nucleotide. The x-ray crystallographic structure of the alanine mutant of Arg 283 verified the loss of the mutated side chain. Our results indicate that specific interactions between DNA polymerase ␤ and the template base, but not hydrogen bonding to the incoming dNTP or terminal primer nucleotide, are required for both high catalytic efficiency and nucleotide discrimination.
Accurate DNA synthesis during replication and DNA repair is crucial in maintaining genomic integrity. Although DNA polymerases play a central role in these essential processes, the fundamental mechanism by which they select the correct deoxynucleoside 5Ј-triphosphate (dNTP) 1 from a pool of structurally similar compounds and substrates to accomplish rapid and efficient polymerization is poorly understood. Vertebrate DNA polymerase ␤ (␤-pol) has been suggested to play a role in both DNA repair (1)(2)(3)(4)(5) and replication (6 -8). The x-ray crystal structures of rat and human ␤-pol in complex with substrates have suggested a detailed model of the chemical mechanism for the nucleotidyl transfer reaction and also have suggested several protein/substrate interactions that may play a role in nucleotide discrimination (9 -12). Additionally, these structures allow us to experimentally test model-derived predictions about the role(s) of individual amino acids.
DNA and RNA polymerases, for which the structure has been determined, have been described by analogy to the anatomical features of a hand as consisting of fingers, palm, and thumb subdomains (13). Conserved carboxylates, which bind catalytically essential divalent metal ions, are found in the palm subdomains of these polymerases. The dNTP binding site of ␤-pol is formed by the DNA template base, the 3Ј-terminal nucleotide of the primer strand, and the palm and thumb subdomains of the polymerase (10). Only three amino acid residues of the thumb subdomain have side chains that are within hydrogen bonding distance to the nucleotide bases within this binding pocket. These hydrogen bond donors are indiscriminate in that they bond to the O2 of pyrimidines or the N3 of purines in the DNA minor groove (14). The structure of the ␤-pol ternary complex reveals a single hydrogen bond between the base of the incoming ddCTP and Asn 279 ; Tyr 271 and Arg 283 are also within hydrogen bonding distance to the O2 and N3 atoms of the terminal primer and templating base, respectively ( Fig. 1). To assess the role of these interactions in nucleotide selection and incorporation, we replaced Tyr 271 , Asn 279 , and Arg 283 with alternate residues by site-directed mutagenesis to remove and/or alter each interaction.
Mutagenesis of the Human ␤-Pol Gene-Oligonucleotide site-directed mutagenesis was performed using a procedure described previously (15). M13 phage containing the human ␤-pol target DNA was propagated using the bacterial host CJ236 (dut Ϫ ung Ϫ ) and phage DNA purified for use as template. Synthetic oligonucleotide primers containing the desired codon change were annealed to the template DNA and the The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ To whom correspondence should be addressed.
primers extended with Sequenase Version 2.0 (U. S. Biochemical Corp.). The following mutations were introduced into the M13 ␤-pol vector, 5Ј to 3Ј: Y271F (TAT to TTC), Y271H (TAT to CAC), N279A (AAT to GCG), N279L (AAT to CTG), R283A (AGG to GCG), R283K (AGG to AAA), and R283L (AGG to CTG). To ensure that the resulting ␤-pol genes contained the desired change, the entire coding sequence of each mutant was confirmed by DNA sequence analysis. The mutated ␤-pol gene was inserted into the ClaI and HindIII sites of the P L promotorbased expression system pWL-11 provided by T. A. Patterson (Ares, Inc.) and overexpressed in Escherichia coli TAP56. Protein Purification-Wild-type and mutant enzymes were purified as described previously (16). All enzyme preparations were assayed for contaminating 3Ј 3 5Ј exonuclease activity on a mismatched primer and had at least 10-fold lower exonuclease activity relative to Klenow fragment (17).
␤-Pol Polymerization Assays-Enzyme activities were determined using a standard reaction mixture (50 l) containing 50 mM Tris-HCl, pH 7.4 (22°C), 5 mM MnCl 2 , and 100 mM KCl. Other reaction conditions are described in the figure legends. Reactions were initiated by addition of enzyme, incubated at 22°C, and stopped by the addition of 20 l of 0.5 M EDTA, pH 8. Quenched reaction mixtures were spotted onto Whatman DE-81 filter disks and dried. Unincorporated [␣-32 P]dTTP was removed, and filters were counted as described (18).
Short Gap Fidelity Assay-A gapped DNA substrate was constructed in which the single-stranded gap contains the lacZ ␣-complementation sequence of M13mp2, which has been modified by the introduction of an in frame opal codon (see Fig. 3A). The construction of the gapped substrate will be described in detail elsewhere. Polymerase reactions (20 l) contained: 20 mM Tris-HCl, pH 8.0, 2 mM dithiothreitol, 25 mM NaCl, 10 mM MgCl 2 , 5% glycerol, 1 mM ATP, 150 ng (32 fmol) of gapped DNA, 500 M each dATP, dCTP, dGTP, and dTTP, 400 units of T4 DNA ligase and ␤-pol. An enzyme titration was performed to determine the amount of each ␤-pol variant required to completely fill the 5-nucleotide gap (see below). Following incubation at 37°C for 60 min, reactions were stopped by adding EDTA to 37.5 mM and the products separated on an agarose gel. Gel slices containing covalently closed circular product DNAs were isolated. The DNAs were electroeluted from gel slices and concentrated. DNA samples were introduced into competent E. coli cells by electroporation and plated on Petri dishes containing the chromogenic indicator 5-bromo-4-chloro-3-indolyl-␤-D-galactoside and a lawn of E. coli ␣-complementation host cells (CSH50) as described (19). Molar ratios of enzyme to DNA used in these assays are as follows: 25:1 for wild-type, N279A, N279L; 100:1 for Y271H, Y271F; 200:1 for R283K, R283L; 250:1 for R283A.
Structure Determination-The human wild-type and mutant enzymes were complexed with DNA by mixing 54 l of enzyme (20 mg/ml in 0.1 MES, pH 6.5, 10 mM (NH 4 ) 2 SO 4 ) with 43 l template⅐primer (T⅐P) (5.5 mg/ml in 20 mM MgSO 4 ) at room temperature. ddTTP was added in 40-fold molar excess to the enzyme-DNA complex. The template and primer sequences are 5Ј-CAAACTCACAT-3Ј and 5Ј-TGATGTGAG-3Ј, respectively. Crystals of the complex were grown at room temperature from sitting drops prepared by mixing 5 l of protein-DNA complex with 5 l of reservoir containing 13% polyethylene glycol 3350, 180 mM NaCl, 50 mM cacodylate, pH 6.5. Crystals appeared within a week. Macroseeding was employed to enlarge crystal size. Crystals belong to space group P2 1 2 1 2 (a ϭ 180, b ϭ 57.5, c ϭ 47.9 Å) and are isomorphous with crystals obtained previously (11). The crystals grew to 0.6 ϫ 0.3 ϫ 0.2 mm and diffracted to 3.3 Å. Using a rotating anode x-ray generator equipped with Xuong-Hamlin multiwire area detectors (20), 43,435 observations of 8411 unique reflections were collected with 97% completeness to 3.3 Å.

RESULTS AND DISCUSSION
Expression constructs of human ␤-pol were prepared: Tyr 271 was replaced with either histidine or phenylalanine; Asn 279 was replaced with either alanine or leucine; and Arg 283 was replaced with either alanine, leucine, or lysine. Each altered human ␤-pol gene was expressed in E. coli and the recombinant enzymes were soluble in the crude cell extracts. Following purification, SDS-PAGE analysis indicated that the mutant ␤-pol polypeptides had the same apparent molecular weight as the wild-type enzyme and were greater than 99% homogeneous (data not shown).
To compare the catalytic efficiency of the wild-type and mutant enzymes, the steady-state kinetics on a simple T⅐P system, poly(dA)-p(dT) 10 with dTTP as the incoming nucleotide, were analyzed (Fig. 2). Whereas the catalytic activity of the Tyr 271 and Asn 279 mutants were not significantly influenced (i.e. Ͻ10fold), k cat of the lysine, alanine, and leucine mutants of Arg 283 were decreased greater than 20-, 150-, and 600-fold, respectively ( Fig. 2A). In contrast, the K m for T⅐P was increased with the R283L mutant (17-fold), and the K m for dTTP was increased to the greatest extent with the R283A mutant (29-fold) ( Fig. 2A). Catalytic efficiency, as measured by the ratio of k cat and K m,dNTP , was not influenced by the histidine substitution at Tyr 271 , while the phenylalanine mutant displayed a modest (2-fold) decrease (Fig. 2B). Since the phenylalanine substitution had only a small effect on catalytic efficiency, substrate interactions with Tyr 271 appears to offer very little transition state stabilization. Elimination of the hydrogen bond between the incoming dNTP and the Asn 279 side chain with an alanine or leucine substitution decreased catalytic efficiency further, but again only modestly (Ϸ10-fold). In this case, catalytic efficiency was dependent solely on the apparent dNTP binding affinity, 2 since k cat of each mutant was similar to wild-type enzyme. The most dramatic decrease in catalytic efficiency was observed for the mutants of Arg 283 (Ala Ͼ Leu Ͼ Lys). A 5000-fold decrease in efficiency was observed for the alanine mutant, whereas catalytic efficiency of the lysine mutant, which could potentially hydrogen bond to the template base, was decreased over 100-fold.
In vivo, ␤-pol is involved in short gap DNA repair (1)(2)(3)5). DNA polymerase ␤ is an ideal polymerase to examine "intrinsic" base substitution fidelity, because it lacks an associated 3Ј 3 5Ј proofreading exonuclease. In vitro, pol-␤ fills these short gaps (Յ6 nucleotides) processively, whereas longer gaps are filled distributively (21). The fidelity of ␤-pol-dependent long gap DNA synthesis (i.e. Ͼ100 nucleotides) had previously been examined on undamaged (22,23) and damaged DNA templates (24,25). To determine the fidelity of wild-type ␤-pol on a physiologically relevant DNA substrate and to assess the effect of the amino acid substitutions on fidelity, a reversion assay was developed on a short (5 nucleotide) gapped DNA substrate containing an opal codon (Fig. 3A). This codon is within the non-essential lacZ␣ gene of bacteriophage M13mp2. Polymerase errors that restore ␣-complementation activity yield a blue or light blue plaque phenotype. This assay can detect eight different base substitution errors.
The result of in vitro gap filling synthesis by wild-type ␤-pol and the mutants described above on the reversion of the opal codon is shown in Fig. 3B. Wild-type ␤-pol produced one revertant per 370 filled gaps (reversion frequency of 27 ϫ 10 Ϫ4 ). Whereas deletion of the hydrogen bond donor at Tyr 271 did not alter the reversion frequency, alanine substitution at Asn 279 significantly reduced it signifying an apparent increase in fidelity. This apparent increase in fidelity could reflect a reduced misinsertion rate or a reduced ability to extend mispairs, since both must occur to score a mutant. In contrast, alteration of the Arg 283 side chain, which interacts with the templating base, dramatically lowered fidelity, as demonstrated by the strong increases in reversion frequency (Fig. 3B).
Sequence analyses of the DNA of lacZ␣ mutants resulting from short gap filling synthesis indicated that the types of base substitution errors produced by the wild-type and R283A mutant were similar (Table I). However, the frequency of each type of error was much greater for the R283A mutant. The base substitution errors observed in the polymerization products of both enzymes reflected misincorporations resulting in relatively frequent T⅐dGTP and A⅐dGTP mispairs. Seven of the eight mispairs detected by this reversion assay were observed in the products of wild-type enzyme and the strong mutator mutant R283A. For the mutant polymerase, a dGMP was incorporated opposite a template thymidine nearly 46% of the time, whereas the correct nucleotide was incorporated only 48% of the time. Additionally, sequence analysis often detected two misincorporations by both wild-type and R283A polymerases within the 5-nucleotide gap. These misincorporations were, in many instances, consecutive, and in one case, three consecutive misincorporations were observed. Consecutive misincorporations had not been observed previously in the forward mutation assay employing a long single-stranded template (22,23). This suggests that a difference may exist between the fidelity of ␤-pol during short processive gap filling as compared FIG. 3. Short gap fidelity assay. A, experimental outline for the short gap fidelity assay as described under "Experimental Procedures." B, mutation frequencies for the products synthesized by the wild-type and mutant ␤-polymerases. The background reversion frequency for the assay was Յ0.001%. Frequencies are shown as the mean and standard deviation of at least two independent determinations.

TABLE I
Base substitution specificity of wild-type and R283A ␤-pol DNA from 47 wild-type and 92 R283A ␤-pol revertants was sequenced yielding a total of 49 and 98 base substitution mutations, respectively. Error rates per detectable nucleotide polymerized were calculated as described (19). Multiple base substitution mutations in a revertant were treated as independent events in the calculation. with distributive DNA synthesis on large gaps. Processive short gap filling synthesis is modulated by the binding of the amino-terminal 8-kDa domain to the downstream 5Ј-phosphate group in gapped DNA (26).
To understand the structural basis for the lower catalytic efficiency and fidelity of the alanine mutant of Arg 283 , we determined the x-ray crystal structure of this mutant in complex with substrates (Fig. 4). In contrast to the rat ␤-pol ternary complex, where the thumb subdomain is in a closed conformation (10), the human mutant ␤-pol ternary complex crystallized in a different crystal packing form with the thumb in the open conformation (space group P2 1 2 1 2). Hence, Ala 283 is moved over 12 Å away from the template. This open conformer had been observed previously when the wild-type human enzyme is bound to the blunt end of duplex DNA (11). We have been unable to crystallize the mutant with the thumb in the closed conformation. There is electron density corresponding to the incoming thymidine moiety of ddTTP at the primer 3Ј terminus; however, the ddTTP had not reacted with the primer and the 5Ј-triphosphate is disordered and not visible. A F o(R283A) Ϫ F o(native) difference Fourier map reveals a negative peak enveloping the Arg 283 side chain consistent with mutation to alanine (Fig. 4). No other significant changes were detected between wild-type and mutant enzymes. The fingers and thumb subdomains are structurally diverse among the different classes of polymerases, and except for ␤-pol, the dNTP binding site is not clearly defined. Therefore, the functional role of each subdomain may be unique to each class of polymerase, and care must be taken in extrapolating the present results to the thumb subdomain of other DNA polymerases (27).
In summary, fidelity assays coupled with kinetic and structural evaluation of the alanine mutant of Arg 283 indicate that this residue plays a central role in nucleotide discrimination by correctly positioning and stabilizing the templating base for efficient nucleotide incorporation. Although the guanidinium group of Arg 283 is within hydrogen bonding distance to N3 of the template guanine, the hydrogen bond geometry is unfavorable. Therefore, correct van der Waal's interactions may also be important at this site. This is consistent with the low catalytic efficiency and reduced fidelity exhibited by the lysine mutant of Arg 283 , which would be expected to preserve hydrogen bonding to the templating base. Our results support the hypothesis that discrimination and catalytic efficiency are modulated by polymerase interactions near the templating base and are sensitive to precise Watson-Crick base pairing by possibly "sensing" C1Ј distances and bond angle geometry (28,29). In contrast, alteration of direct interactions with the incoming dNTP decreased dNTP binding affinity but not fidelity. Thus, the coupling between catalytic efficiency and discrimination is residue-specific. Our results indicate that we can modulate discrimination and catalytic efficiency based upon ternary complex crystal structures, and site-directed mutagenesis will be a productive avenue for future analysis of polymerase structure-function relationships. Although the thymidine moiety of ddTTP is observed, the 5Ј-triphosphate is disordered. In contrast to the rat ␤-pol ternary complex where the thumb subdomain is in a closed conformation ( Fig. 1) (10), the human enzyme (wild-type and mutant) crystallized with the thumb in the open conformation. Therefore, Ala 283 is moved over 12 Å away from where it is observed in the closed ternary complex.