Highly Tolerated Amino Acid Substitutions Increase the Fidelity of Escherichia coli DNA Polymerase I*

Fidelity of DNA synthesis, catalyzed by DNA polymerases, is critical for the maintenance of the integrity of the genome. Mutant polymerases with elevated accuracy (antimutators) have been observed, but these mainly involve increased exonuclease proofreading or large decreases in polymerase activity. We have determined the tolerance of DNA polymerase for amino acid substitutions in the active site and in different segments of E. coli DNA polymerase I and have determined the effects of these substitutions on the fidelity of DNA synthesis. We established a DNA polymerase I mutant library, with random substitutions throughout the polymerase domain. This random library was first selected for activity. The essentiality of DNA polymerases and their sequence and structural conservation suggests that few amino acid substitutions would be tolerated. However, we report that two-thirds of single base substitutions were tolerated without loss of activity, and plasticity often occurs at evolutionarily conserved regions. We screened 408 members of the active library for alterations in fidelity of DNA synthesis in Escherichia coli expressing the mutant polymerases and carrying a second plasmid containing a β-lactamase reporter. Mutation frequencies varied from 1/1000- to 1000-fold greater compared with wild type. Mutations that produced an antimutator phenotype were distributed throughout the polymerase domain, with 12% clustered in the M-helix. We confirmed that a single mutation in this segment results in increased base discrimination. Thus, this work identifies the M-helix as a determinant of fidelity and suggests that polymerases can tolerate many substitutions that alter fidelity without incurring major changes in activity.

Fidelity of DNA synthesis, catalyzed by DNA polymerases, is critical for the maintenance of the integrity of the genome. Mutant polymerases with elevated accuracy (antimutators) have been observed, but these mainly involve increased exonuclease proofreading or large decreases in polymerase activity. We have determined the tolerance of DNA polymerase for amino acid substitutions in the active site and in different segments of E. coli DNA polymerase I and have determined the effects of these substitutions on the fidelity of DNA synthesis. We established a DNA polymerase I mutant library, with random substitutions throughout the polymerase domain. This random library was first selected for activity. The essentiality of DNA polymerases and their sequence and structural conservation suggests that few amino acid substitutions would be tolerated. However, we report that two-thirds of single base substitutions were tolerated without loss of activity, and plasticity often occurs at evolutionarily conserved regions. We screened 408 members of the active library for alterations in fidelity of DNA synthesis in Escherichia coli expressing the mutant polymerases and carrying a second plasmid containing a ␤-lactamase reporter. Mutation frequencies varied from 1 ⁄ 1000-to 1000fold greater compared with wild type. Mutations that produced an antimutator phenotype were distributed throughout the polymerase domain, with 12% clustered in the M-helix. We confirmed that a single mutation in this segment results in increased base discrimination. Thus, this work identifies the M-helix as a determinant of fidelity and suggests that polymerases can tolerate many substitutions that alter fidelity without incurring major changes in activity.
DNA polymerases function in DNA replication, repair, and recombination and are essential for maintaining the integrity of the genome. Multiple DNA polymerases have been found in prokaryotes, eukaryotes, and viruses and may have different properties, including variations in the accuracy, or fidelity, of DNA synthesis (1)(2)(3)(4)(5). Polymerases that replicate large genomes typically have high fidelity, with error rates on the order of 10 Ϫ5 to 10 Ϫ6 (6 -8) to prevent the accumulation of deleterious mutations in the genome. Deficits in DNA polymerase fidelity have been associated with cancer, accelerated aging, and infertility (9 -11).
Mutator and antimutator DNA polymerases harbor amino acid substitutions resulting in decreased and increased fidelity, respectively. These variant enzymes provide insights into the structural basis of accurate DNA synthesis (12)(13)(14)(15)(16)(17)(18). Many mutators have been identified due to the availability of powerful selection methods (19). The majority of studied mutators harbor amino acid substitutions in well conserved polymerase motifs. Mutator polymerases have also been used extensively in biotechnological applications, such as DNA sequencing and error-prone PCR (20). Far fewer antimutator enzymes have been identified due to the lack of selection methods. The ones identified thus far either exhibit substantially diminished catalytic activity or possess increased 3Ј-5Ј exonuclease proofreading (21)(22)(23). Decreased polymerase activity allows more time for removal of misincorporations by a proofreading exonuclease. Recent efforts to identify antimutators in proofreading-deficient constructs through 96-well screening have focused only on 3 residues within Motif C and have produced antimutators with less than 10% of wildtype activity (24).
DNA polymerase I (Pol I) 4 is a high fidelity polymerase belonging to Family A, which includes Taq Pol I, T7 DNA polymerase, and human DNA polymerase ␥, , and (6,25,26). In this work, we identified determinants of Pol I fidelity that do not involve alterations in proofreading by the 3Ј-5Ј exonuclease and that do not cripple the polymerase activity. This was attained by conducting a comprehensive analysis of amino acid substitutions in Pol I that result in antimutator or mutator activity. We focused on mutations that directly affect nucleotide selection at the incorporation step. We report on the spectrum of amino acid substitutions that both maintain activity and confer either antimutator or mutator phenotypes. Our results indicate that although the catalytic site is optimized for base selection without loss of activity, it is still feasible to create more accurate DNA polymerases without greatly reducing catalytic activity. Furthermore, many of the mutations that affect fidelity are distant from the active site.
Creation of a Pol I Mutant Library-The E. coli polA gene encoded in plasmid pECpolI-3Јexo Ϫ was mutated to introduce a PstI restriction site at nucleotide 1493. The polymerase domain was amplified from this plasmid using the primers GGACGT-TACGCTGCAGAAG and CGACGGCCAGTGAATTC-TTAG according to the error-prone PCR GeneMorph II kit (Stratagene, La Jolla, CA), which includes error-prone Taq Pol I mutant 53 (29). The PCR product was gel-purified (Qiagen, Valencia, CA), cut using PstI and EcoRI (New England Biolabs, Ipswich, MA), and ligated into a pECpolI-3Јexo Ϫ dummy vector. The dummy vector had Pol I nucleotides 2219 -2646 deleted, resulting in an inactive polymerase. The ligated plasmids were transformed into TOP10 cells (Invitrogen). Library size was calculated by counting a fraction of the surviving transformants.
Selection of Active Pol I Mutants-The library was transformed into JS200 cells, which were then plated on 2XYT agar and grown overnight at 37°C. Surviving colonies were inoculated into LB and grown at 30°C. Individual plasmids from both the unselected and selected library were recovered and sequenced. Analysis of the library mutation spectrum was performed on Phred-Phrap software (30,31). The index of substitutability at each position was calculated as the percentage of clones containing a mutation at that residue divided by the total number of clones sequenced.
Comparison of Evolutionary Conservation Index with Substitutability Index-An index of evolutionary conservation (Fig. 2B) (32,33). The DNA polymerases analyzed were DNA polymerase I (bacteria) or DNA polymerase ␥ (eukaryotes) unless otherwise noted. DNA polymerase ␥ sequences were included to further differentiate residues that are absolutely required for catalytic activity. Sequences were obtained from Pubmed (available on the World Wide Web at www.ncbi.nlm.nih.gov/) and aligned with ClustalW version 1.83 (available on the World Wide Web at www.ebi.ac.uk/clustalw/).
Screen for Mutations in Pol I That Alter Fidelity of DNA Synthesis-Plasmid pECpolI was recovered from active clones and transformed into fresh JS200. After recovery, cells were transformed with the second plasmid pLA230 and grown in LB to an A 600 of 2.0 at 30°C. The culture was then inoculated to a final dilution of 5 ϫ 10 Ϫ6 into 1 ml of 2XYT containing chloramphenicol (30 g/ml) and kanamycin (50 g/ml). Cultures were then grown with aeration for 20 h at 37°C and plated on LB agar containing either chloramphenicol and carbenicillin (100 g/ml) or chloramphenicol and kanamycin at a dilution sufficient for 100 -500 colonies per plate. Several controls were included: wild type sequences, D424A,I709N, and D424A, I709N,A759R (34). Assays were performed in duplicate.
Assay for Pol I Polymerase Activity-JS200 cells containing pECpolI mutants were grown in LB at 30°C with 1 mM isopropyl 1-thio-␤-D-galactopyranoside to induce protein expression. Cells were grown to an A 600 of 2.0, at which time a 1.5-ml aliquot was pelleted, washed with 1 ml of 20 mM sodium phosphate (pH 7.2), and resuspended in 0.1 ml of the same buffer. Cells were lysed with 5 l of lysozyme (10 mg/ml) and disrupted by freezing (Ϫ80°C) and thawing. The cell extract was collected by centrifugation at 15,000 rpm for 15 min. The polymerase activity of the supernatant was assayed for ability to incorporate radioactive dTMP into activated calf thymus DNA at 37°C as previously described (35). Assays were performed in triplicate. The background polymerase activity of the cell, measured from cells grown with an empty plasmid vector, was subtracted from all measurements.
Protein Purification-Mutant and wild-type Klenow fragments, encompassing the 3Ј-5Ј exonuclease domain and the polymerase domain, were subcloned into the pLEX vector (Invitrogen), expressed, and purified as previously described (27). This Pol I construct included a six-histidine N terminus tag. In brief, the pLEX expression plasmid was introduced into E. coli GI724 (F2, l2, lacIq, lacPL8, ampC::Ptrp cI, mcrA, mcrB, INV(rnnD-rnnE)), and cells were grown at 30°C in 500 ml of induction medium composed of M9 salts, 0.2% casamino acids, 0.5% glucose, 1 mM MgCl 2 , and 100 mg/ml carbenicillin. When the cultures reached an A 600 of 0.5, tryptophan was added to a final concentration of 100 g/ml, and the culture was incubated at 37°C for 4 h. The cells were collected by centrifugation, washed with 40 ml of binding buffer (5 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9), and suspended in 4 ml of the same buffer containing 200 g/ml lysozyme and 0.04 ml Protease Inhibitor Mixture III (Calbiochem). Extracts were prepared by freezing/thawing on ice for 2 h, followed by centrifugation at 15,000 rpm for 15 min, and loaded onto a 4-ml nickel-resin column (Novagen, San Diego, CA). The column was washed with 80 ml of wash buffer (15 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9) and serially with 8 ml of high imidazole wash buffer (30, 60, 120, and 200 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9) and eluted with 8 ml of 400 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9. The eluted sample was mixed with 2.5 ml of 80% glycerol and 12.5 l of 1 M dithiothreitol and stored at 80°C. Protein concentrations were determined by the method of Bradford.
To assay the polymerase and 3Ј-5Ј exonuclease activity, a DNA duplex with a 5Ј overhang (template strand) was prepared using a template oligonucleotide, CCCGGGAAATTTCCGG-AATTCGATATTGCTAGCGGGAATTCGGCGCG, and the primer oligonucleotide CGCGCCGAATTCCCGCTAGCAA-TAG, the latter of which was 32 P-radiolabeled at the 5Ј termi-nus. The purified enzyme (40 fmol) was incubated with the DNA duplex (20 fmol) in 10 l of 50 mM Tris-HCl, pH 7.5, 10 mM KCl, 2.5 mM MgCl 2 , 1 mM dithiothreitol at 37°C for 30 min. Klenow 3Ј-5Ј exonuclease-proficient enzyme (4 fmol) was used as a control. Reactions were terminated with 20 l of 98% formamide, 10 mM EDTA; 5 l of each product was analyzed by electrophoresis through 14% denaturing acrylamide electrophoresis and quantified by phosphorimaging analysis.
M13 Gapped Assay-We characterized the DNA synthesis fidelity of purified mutant and wild-type polymerase in vitro on the M13 gapped forward mutation assay, as previously described (36). Gap fill-in reactions occurred in 10 l of 50 mM Tris-HCl, pH 7.5, 10 mM KCl, 2.5 mM MgCl 2 , 1 mM dithiothreitol at 37°C with all four dNTPs at 50 M each. The complete fill-in of the M13 gapped substrate was monitored by gel (data not shown) and by the consistency of mutation rates after incubation with increased amounts of the DNA polymerase.

RESULTS
To explore the protein sequence space of Pol I, we generated a random mutation library of the polA gene using errorprone PCR. Because we wanted to focus on mutations in the polymerase domain, nucleotides at positions 1493-2784 that encode the polymerase catalytic domain were randomly substituted, and the 3Ј-5Ј exonuclease proofreading activity was inactivated by a D424A substitution (27,37). For simplicity, mutants henceforth will only be named by their polymerase domain genotype. For example, the D424A mutant is referred to as "wild type," whereas the double mutant D424A,I709N is referred to as only I709N. A library of greater than 1.0 ϫ 10 4 mutants was established, with a mean amino acid substitution of 4.2 per gene (polymerase domain only). To select for active mutants, the library was transfected into JS200 E. coli cells, which contain a temperature-sensitive allele of Pol I; growth at the restrictive temperature requires complementation with a plasmid-borne active Pol I (27,38). Approximately 10% of the original library formed colonies at the restrictive temperature. From the surviving population, 592 individual clones were sequenced to reveal a spectrum of tolerable substitutions across the polymerase domain ( Fig. 1). A mean of 2.8 amino acid mutations were present in the active clones. Across the entire polymerase domain, the probability (30) that a single random amino acid substitution will inactivate the enzyme is 34 Ϯ 2% (mean Ϯ S.E.).
To compare the tolerance to substitutions at different amino acid positions, an index of substitutability was calculated based on the percentage of active clones that harbored a mutation at that residue (30). These indices were consistent with the prediction that the active site of the polymerase, located in the interior-facing surfaces of the enzyme, harbors the most highly conserved residues, whereas residues located on the outer surfaces and lacking direct contact with the reaction substrates more readily tolerate substitutions ( Fig. 2A and Table 1). To quantify how well these results match the amino acid conservation among Family A DNA polymerases, we compiled a protein sequence alignment of 30 Family A DNA polymerase species (17 prokaryotic, 10 eukaryotic, and 3 viral) and calculated indices for evolutionary conservation at each position (Fig. 2B).
By comparing the substitutability of different amino acid residues in Pol I with the evolutionary conservation among Family A polymerases, we were able to identify regions of discrepancy between the functional and evolutionary conservation (Fig. 2B). Residues where the functional tolerance is at least 35% greater than the evolutionary conservation include 516, 656, 720, 723, 726, 738, 757, and 802.
To assess the accuracy of the mutant polymerase enzymes, we screened 408 clones from our active library for their polymerase fidelity using a second plasmid pLA230 as a reporter (28). This reporter contains an opal (TAA) codon at the N terminus of the ␤-lactamase protein, which terminates its translation. Errors that are produced in this codon during the Pol I-catalyzed replication of pLA230 render the bacteria resistant to the antibiotic carbenicillin. The reversion frequency has been shown to be inversely related to the in vitro fidelity of the Pol I enzyme (28). The reversion frequency for cells expressing the wild-type Pol I was 8.9 Ϯ 1.9 ϫ 10 Ϫ7 (mean Ϯ S.E.). Clones with a reversion frequency at least 10-fold below or above wildtype were classified as antimutators and mutators, respectively (Fig. 3). The lowest and highest reversion frequency were ϳ1/1000and 1000-fold that of wild type. Approximately 12% (51) of the clones were antimutators, whereas only 2.4% (10) were mutators.
To determine the catalytic activity of the same 408 clones, we measured their capacity to incorporate radiolabeled dTMP into activated calf thymus DNA in vitro (35). The activity of the polymerases varied greatly (Fig. 3), ranging from 15 to 125% of that of clones expressing similar amounts of wild-type enzyme as determined by Western blots. The only clones with activity greater than wild-type were mutators. Both mutators and antimutators were obtained without significant diminution in catalytic activity.
To verify that the results of our reversion and polymerase activity assays were internally valid, we included three wildtype controls that had nucleotide substitutions that did not change the encoded amino acid sequence. In addition, two previously characterized extreme mutators, I709N and the double mutant I709N,A759R, were included (28,34). Their reversion frequencies are ϳ200and 1700-fold, respectively, above wildtype, consistent with prior studies.
In order to confirm that the enhanced fidelity of DNA synthesis measured in vivo reflects a greater accuracy by the DNA polymerase, we purified the polymerase from a mutant, K601I,A726V, that exhibited a 10-fold enhanced accuracy without a significant reduction in activity. First, both the mutant and a wild-type control were cloned into the pLEX vector and expressed with a six-histidine N terminus tag. After chromatography on a nickel column, both the purified mutant polymerase and the wild-type control polymerase were tested for 3Ј-5Ј exonuclease activity to ensure that the mutant did not exhibit any proofreading (Fig. 4). The polymerases were incubated with double-stranded DNA with a mismatched G:A (primer-template) at the 3Ј-primer terminus. Less than 0.001% of the terminal nucleotides were hydrolyzed in 30 min at 37°C in a reaction containing a 4 nM concentration of the mutant or wild-type DNA polymerase and 2 nM DNA. 3Ј-5Ј exonuclease activity was also absent when the polymerases were incubated with double-stranded DNA with a matched A:T at the 3Ј-primer terminus (data not shown). The fidelity of the mutant protein and of the wild type were then measured in vitro using an M13 gapped forward mutation assay (36), in which the enzymes catalyze the incorporation of four dNTPs into gapped M13 circular DNA. The complete fill-in of the M13 gapped substrate was monitored by gel (data not shown) and by the consistency of mutation rates after incubation with increased amounts of the DNA polymerase. A 5-fold increase in the amount of enzyme used for fill-in did not alter the mutation frequency significantly.
In two separate experiments, the frequency of mutations in the gapped segment of M13 DNA was 3.1-and 2.6-fold less in reactions catalyzed by K601I,A726V than in reactions catalyzed by the wild-type polymerase ( Table 2). 68 and 82% of the mutations were single base substitutions with the wild-type and mutant polymerase, respectively. The largest reduction in error FIGURE 1. Tolerated amino acid changes in the Pol I polymerase domain. The primary sequence of Pol I, numbered as residues 514 -928, is depicted in five rows, with the wild-type amino acid identity placed below each number and the tolerated mutations placed above each number. This spectrum of mutations was compiled from 592 sequenced active mutants. Amino acids are colored by type. Boxes below the wild-type sequence identify absolutely conserved residues and highly conserved motifs of Family A DNA polymerases (53,54). Identification of ␣-helices and ␤-strands, also denoted below the wild-type sequence, was obtained from Swiss-PdbViewer version 3.7 (59) using Protein Data Bank file 1KLN (60).

FIGURE 2. Functional and evolutionary conservation of the Pol I polymerase domain.
A, after active mutants of Pol I were identified, the number of tolerable mutations at each residue was determined and used to calculate an index of substitutability at that position. Values for these indices are represented as a color along a gradient, from blue to red, representing the gradient from fewest substitutions to the most substitutions. The crystal structure of Pol I Klenow fragment (Protein Data Bank file 1KLN) was then colored accordingly at each residue. The two structures depict opposite faces of Pol I, with the duplex DNA (magenta) positioned between the 3Ј-5Ј exonuclease domain (gray) and the polymerase domain (colored). The active site of the polymerase domain is denoted with an asterisk. B, the functional conservation of Pol I, obtained from genetic complementation, is compared with its evolutionary conservation, obtained from sequence alignment. Amino acid sequences of 30 members of Family A DNA polymerases were aligned to determine conservation at each position. The same color gradient used in A was used to code the conservation of each residue along the primary sequence of Pol I, represented as colored boxes, with numbers along the bottom denoting the amino acid position. rate, 10-fold, occurred with A to C transversions ( Table 3). The mutant exhibited reduced mutation rates for 10 of the possible 12 misinserted bases. Mutation rates for the remaining two, A to G transitions and C to A transversions, were increased less than 2-fold. Many of the error rates for the wild-type Klenow control were comparable with those that have been previously published (13).
The mapping of mutations present in each antimutator and mutator revealed regions where mutations were more frequent. One region, encompassing residues 720 -728 and corresponding to the M-helix in the Pol I protein, clearly stood out. Approximately 12% of the mutations among the antimutators occurred in this segment. This high mutation frequency in the M-helix was not due to a bias in the input population, since only 2.6% of the random substitutions in mutants in our screen were located in this region.
All of the antimutators having a mutation in the M-helix also harbored additional mutations elsewhere. To determine whether a single substitution in the M-helix was capable of producing an antimutator, we genetically dissected the dou-ble mutant, K601I,A726V, into individual mutations and assayed the reversion ability of each (Fig. 5). The double mutant K601I,A726V exhibited a 20-fold decrease in reversion frequency. The K601I mutation and the A726V mutation separately reduced the reversion frequency by 7-and 8-fold, respectively, relative to wild type. This result suggested that each of the substitutions increased base selection and that there was an additive interaction between the two mutations.

DISCUSSION
The centrality of DNA polymerases to the DNA synthetic process, the high degree of evolutionary conservation, and the stringent requirements for selection of four different nucleotide substrates would suggest that these enzymes should have limited tolerance to substitutions within their catalytic domains. Previous studies on random substitutions in Motif A of E. coli Pol I and Taq Pol I indicated that this region tolerates many substitutions (27,39). The results presented in this paper identify many other regions of the polymerase domain of E. coli Pol I that were highly tolerant of amino acid replacements, some of which altered the fidelity of DNA synthesis.
Properties of Antimutators and Mutators in the Polymerase Domain-Mutations in Pol I that alter the reversion frequency of E. coli by as much as 1000-fold were found to be widely distributed across the polymerase domain and not restricted to specific motifs or domains. Many of the antimutator polymerases exhibit polymerase activity near wild type (Fig. 3). This stands in contrast to all of the characterized antimutators, which manifested more extensive decreases in polymerase activity relative to exonuclease proofreading activity. Mutators, in some cases, had a level of activity that was greater than wild type, which was also unpredicted.
Three mechanisms have generally been proposed to account for polymerase antimutators (22,40). First, the antimutator possesses greater selectivity during nucleotide incorporation.
Second, the antimutator phenotype results from a decrease in the ratio of polymerase to exonuclease proofreading activity. Third, the ratio of polymerase to exonuclease activity remains constant, but the nascent DNA is somehow able to translocate more efficiently between the two domains. Since our antimutators lacked an essential residue in the exonuclease domain and did not demonstrate exonucleolytic activity, the most direct explanation is that these antimutators exhibit greater selectivity during nucleotide incorporation. Based on studies with nucleotide analogs of different sizes, Kool and co-workers (41) have suggested that the active site of wild-type Pol I is not optimized for maximum fidelity. Increases in the  (1, 1).

TABLE 1 Mean substitutability indices of Pol I regions
The tolerance to mutation of each amino acid was derived by dividing the number of observed substitutions at that residue by the number of sequence reads obtained and expressed as a percentage (ϫ100). A mean index for each region was calculated by taking the mean of all indices for residues belonging to that region. The t test was performed on the mean indices for region and nonregion residues to evaluate the significance of the difference, which is indicative of relative importance to protein function. tightness of the active site may serve as one mechanism by which the fidelity of the enzyme increases (42). A Novel Determinant of Fidelity, the M-helix-The M-helix consists of residues 720 -728 located at the juncture of the fingers and palm subdomains (43). It is located on the outer surface of the protein and lacks any direct contacts with substrates. Many antimutator mutations (a higher number than expected from random distribution) mapped to the M-helix. Our genetic analysis of a single point mutation in this region, A726V, revealed that a mutation here is sufficient to produce an antimutator phenotype (Fig. 5). Family A DNA polymerases undergo a global conformational change, from an "open" to a "closed" state, during each nucleotide addition reaction (44 -46). It has been argued that kinetic proofreading of the incoming deoxynucleotide may occur at this step (47). Recent computational analyses of DNA polymerases have identified amino acid residues, including several in the M-helix, that function as "hinges" during this transformation (48). An amino acid substi-tution here may disrupt the function of the hinge and allow more time for kinetic proofreading that need not involve the exonuclease (49).

Substitutability
Error Spectrum of K601I,A726V Mutant-In order to verify that the enhanced fidelity measured in vivo reflects an intrinsic increase in the accuracy of the DNA polymerase, we compared the in vitro fidelity of mutant K601I,A726V and the wild-type polymerase. The error rate of the mutant was 3.1-and 2.6-fold lower than wild type as measured on the M13 gapped forward mutation assay in two independent experiments. The error spectrum of the mutant in this assay appears to be different from other Family A antimutator DNA polymerases reported (50). It is characterized by predominantly C to T transitions and G to C transversions. The K601I mutation is located near Motif 1, which has been postulated to be involved in the binding to and positioning of the DNA duplex (51), whereas the A726V mutation is located on the fingers subdomain, which interacts with the incoming nucleotide and template base as well as the catalytic magnesium ions.
We also compared the shift in mutation spectrum of the K601I,A726V double mutant with the several previously reported TaqDNA Pol I antimutators (50) and found the shifts to be dissimilar. The triple mutant A661E,I665T,F667L and the single mutant F667L (50) harbor mutations exclusively in Motif B of the fingers subdomain and demonstrate 2.8-and 2.7-fold increases in fidelity, respectively, and 3-fold decreases in activity. Their error spectrums have reduced A to T and G to T    (30), we calculate that the overall probability for any single amino acid substitution to inactivate Pol I is 34 Ϯ 2%. This probability is strikingly similar to results obtained for other proteins, including human 3-methyladenine DNA glycosylase (AAG) and the E. coli lac repressor (52). The probabilities reported for AAG and the lac repressor are 30 -39 and 34%, respectively. Given that these three proteins are vastly dissimilar in tertiary structure and in function, this similarity of inactivation probabilities may predominantly reflect overall requirements for global processes, such as protein folding at the secondary structure level (30).
The mutational tolerance of the Pol I polymerase domain parallels the extent of its evolutionary conservation ( Fig. 2B and Table 1). The highly conserved Motifs A, B, and C, which harbor catalytically important residues, are the least tolerant of substitutions. The second most conserved regions are Motifs 1, 2, and 6, which interact with the polymerase primer or template and are found across all Family A DNA polymerases (53,54). Hinge region residues (48) also exhibit an intermediate level of tolerance for substitutions compared with Motifs A, B, and C ( Table 1). Several of these residues, 794, 822, 826, 845, 906, and 912, are well conserved evolutionarily but mutable nevertheless (Fig. 2B). This quality may make them excellent targets for altering the properties of DNA polymerases without abolishing catalytic activity (55).
One residue that stood out in our analysis of evolutionary conservation is Asp 854 (Fig. 2B). This residue is highly conserved among bacterial Pol I proteins and interacts with Arg 821 (56). Our alignment of eukaryotic and viral Family A DNA polymerases reveals that this conservation extends across even more diverse phyla. Asp 854 has 90% identity and 97% similarity across prokaryotic, eukaryotic, and viral evolution, which is the highest for any non-substrate-binding residue. Functionally, it is also well conserved (Fig. 2B). Asp 854 is positioned on the T-helix, which spans both the fingers and palm subdomains of the polymerase. The residue appears to anchor this helix to the rest of the fingers through its hydrogen-bonding partners, Arg 821 (56) and the previously unreported Thr 817 , which are located in the loop between ␤-strands 10 and 11. This stabilizing interaction is likely to be critical for polymerase function, because the T-helix contains the evolutionarily immutable residue Gln 849 , which binds the base that is 3Ј to the template base (57).
In summary, we have identified novel antimutator DNA polymerases with high catalytic activity, described the activity of polymerases across a broad range of fidelity mutants, pinpointed the M-helix as a novel determinant of fidelity, and presented the most comprehensive analysis of Pol I protein conservation in E. coli and across diverse evolutionary species. Viewed together, our results indicate that the catalytic site of E. coli Pol I is highly conserved and yet is able to accept multiple substitutions. Both mutator and antimutator mutations can be accommodated. In vivo, many of these mutations can alter base selection at this site without severely compromising catalysis. Considering the structural and functional conservation among different DNA polymerases, it is likely that other DNA polymerases are similarly positioned. Error-prone mutations may have a role in generation of random mutations in malignancies, whereas enhanced accuracy may be of importance in delaying the onset of diseases that are associated with mutation accumulation (58).