Impaired Mismatch Extension by a Herpes Simplex DNA Polymerase Mutant with an Editing Nuclease Defect*

The D368A mutation within the 3′-5′-exonuclease domain of the herpes simplex type 1 DNA polymerase inactivates this nuclease and severely interferes with virus viability. Compared with the wild type enzyme, the D368A mutant exhibits substantially elevated rates of incorrect nucleotide incorporation, as measured in a LacZ reversion assay. This high rate occurs in the presence of high levels of dNTPs, a condition that forces the enzyme to extend mismatched primers. Hence, the mutant fails to correct many misincorporations that are removed in the wild type. In addition, the mutant shows a much reduced ability to replicate DNA templates primed with a 3′-mismatch as compared with wild type. This extension defect also appears more severe than observed for replicases which naturally lack editing nucleases. Based on these findings, we suggest that the inability of the D368A herpes simplex mutant polymerase to replicate beyond a mismatched base pair severely inhibits viral replication.

A major mechanism for controlling the accuracy of chromosomal replication involves 3Ј-5Ј-exonucleases associated with DNA polymerases. Although DNA polymerases actively control fidelity by selecting correctly base-paired dNTPs during the insertion step of polymerization (1,2), misincorporation does occur. In polymerases with editing functions, the polymerase responds to a mismatched primer by pausing, thus facilitating removal of the misincorporated nucleotide (1). However, conditions, such as high concentrations of dNTPs, may favor extension of the mismatched primer rather than editing (1). Misincorporations that escape editing can later be repaired by mismatch repair mechanisms, which act on DNA containing mispaired bases (3).
This study examines the 3Ј-5Ј-exonuclease activity associated with the herpes simplex virus type 1(HSV-1) 1 DNA polymerase (6). The exonuclease domain within this enzyme is indicated by the presence of three highly conserved Exo motifs (Exo I, II, and III) (Fig. 1) and the retention of nuclease activity by an N-terminal proteolytic fragment (7,8). Mutations within the motifs inactivate the exonuclease activity (9 -11). Based on the structural model provided by DNA polymerase I (Escherichia coli), these motifs encompass sites that bind two divalent Mg 2ϩ ions and coordinate the binding of reaction components. Aspartate residues within these motifs are critical for ion binding and catalytic activity (4,5). The HSV-1 D368A mutation described here inactivates the aspartate residue in the Exo I motif.
Mutations that inactivate DNA polymerase-associated 3Ј-5Јexonucleases frequently result in high mutation rates (1,(11)(12)(13)(14)(15)(16)(17). These effects presumably result from fixation of replication errors by extension of mismatched primers after the editing function fails to act. In some systems, exonuclease deficiencies impair viability. This response has often been attributed to production of a high level of deleterious mutations. For example, certain mutations in the exonuclease subunit of DNA polymerase III (E. coli) are lethal, unless suppressed by a second mutation that lowers the high mutation rate by increasing insertion accuracy or overproducing mismatch repair enzymes (13). Reduced viability occurs in yeast (Saccharomyces cerevisiae) carrying exonuclease deficiencies in the mitochondrial polymerase or in both the ␦ and ⑀ nuclear DNA polymerases (17,18). The nuclear mutations become lethal when post-replication mismatch repair is also inactivated (19).
However, another factor that may contribute to low viability in editing-deficient mutants is the pausing of the polymerase after misincorporation. Failure to edit might prevent extension of a mismatched primer, thus impairing replication. In support of this possibility, exonuclease deficiencies in both T4 bacteriophage and HSV-1 DNA polymerases show reduced abilities to extend mismatched primers (10,23). In addition, several exonuclease-deficient DNA polymerases exhibit lower processivity that might be attributable to a mismatch extension deficiency. For example, mutants from bacteriophages ø29 and prd1 fail to replicate duplex templates that require strand displacement for fork movement (20,21). Mutants from yeast mitochondria (S. cerevisiae), prd1, and E. coli (DNA polymerase III) produce short replication products during DNA synthesis on primed single-strand templates (17,21,22). Finally, mutants from T4 exhibit low polymerization activity when mis-incorporation can occur, as compared with activity on homopolymer templates (23).
Certain mutations within the Exo I in motif 3Ј-5Ј-exonuclease domains produce more severe phenotypes than other alleles. For example, the lethal exonuclease alleles associated with DNA polymerase III (E. coli) are within Exo I. In addition, HSV-1 viruses carrying mutations in this region cannot be isolated and appear to be inviable (9). In contrast, HSV-1 mutants with changes in the Exo III motif are viable (11). Since the Exo I motif affects binding of both the metal ions required for catalysis (4,5), mutations within this motif may be especially inhibitory to catalysis.
The present study examines several biochemical properties of an HSV-1 DNA polymerase carrying a mutation in the Exo I motif (D368A). Since this mutation appears to be lethal (9), we were interested in uncovering mechanisms to explain the se-verity of this defect. We find that the D368A mutation dramatically elevates the mutation frequency during replication but only under conditions that force the extension of mismatched primers. In addition, this mutation causes an exceptional deficiency in extension of mismatched primers. These abnormalities may drastically impair viral replication, thus resulting in the lethality of this mutant.

EXPERIMENTAL PROCEDURES
Polymerase Purification-Sf9 cells were infected with recombinant baculovirus carrying the appropriate HSV-1 polymerase gene, and the cell lysates were processed as described (9). (NH 4 ) 2 SO 4 was added to the clarified cell lysates to a final concentration of 20%. Lysates were subjected to low pressure liquid chromatography using phenyl-Sepharose columns (Amersham Pharmacia Biotech) equilibrated in Buffer C (9) supplemented with 20% glycerol, 50 mM NaCl, and 30% (NH 4 ) 2 SO 4 . Fractions were eluted with a decreasing (NH 4 ) 2 SO 4 gradient from 20 to 0% and then with a step gradient of Buffer C supplemented with 20% glycerol, 50 mM NaCl, and either 0.1, 0.5, 1, or 2% Triton X-100. Both wild type and D368A mutant polymerase peaks eluted between 1 and 2% Triton X-100. Fractions containing peak activity were pooled, dialyzed against Buffer C containing 20% glycerol and 50 mM NaCl, and further purified by ion exchange chromatography (phosphocellulose), using a linear NaCl gradient (50 -600 mM), in Buffer C containing 20% glycerol. Both wild type and D368A mutant polymerase peaks eluted at about 300 mM NaCl. Peak polymerase fractions were identified, dialyzed, and further purified by affinity chromatography on either singlestranded DNA agarose (wild type) or heparin-Sepharose (D368A mutant) (Amersham Pharmacia Biotech). Columns were eluted with a linear gradient (50 -400 mM NaCl) in Buffer C containing 20% glycerol. The wild type polymerase peak eluted from the single-stranded DNA column at about 300 mM NaCl, and the D368A mutant polymerase peak eluted from the heparin-Sepharose column at about 350 mM NaCl. Peak polymerase fractions were pooled, dialyzed against Buffer C containing 50 mM NaCl and 60% glycerol, and stored at Ϫ20°C. Throughout purification, polymerase activity was identified by assays with salmon sperm DNA templates (9), and protein concentration was determined by the method of Bradford (39). The presence of the HSV-1 polymerase was assessed by electrophoresis on sodium dodecyl sulfate-polyacrylamide gels and by Western blotting using anti-HSV DNA polymerase antibodies (not shown).
Reversion Assays-Assays measured the reversion of an opal codon in a lacZ␣ gene (E. coli) cloned within a circular gapped DNA duplex prepared from M13mp2A89 bacteriophage (a generous gift of Dr. T. A. Kunkel). Reactions for the HSV-1 polymerases were performed using the buffer conditions described for standard polymerase assays (9) and contained 20 mM (NH 4 ) 2 SO 4 , 0.5 g of gapped DNA (0.1 pmol), and dNTPs at the indicated concentrations. Reverse transcriptases were obtained from Life Technologies, Inc., and were assayed under conditions recommended by the manufacturer. Reactions containing either 3 units (HSV-1 DNA polymerases) or 5 units (reverse transcriptases) of enzyme were incubated at 37°C for 1 h and stopped by addition of EDTA to 30 mM. Aliquots were subjected to electrophoresis on 0.8% agarose gels to verify that complete replication of the gapped DNA had occurred. Reactions were then diluted in water 10-fold, and a series of 5-l aliquots was used to transform 200 l of chemically competent JM103 bacteria (40). Cells were added to YT soft agar and plated on YT plates containing 0.001% isopropyl-␤-D-thiogalactopyranoside and 0.004% 5-bromo-4-chloro-3-indoyl-␤-D-galactopyranoside. Transformations gave approximately 1 ϫ 10 4 plaques per plate. Both light and dark blue revertant plaques were identified, and the mutant phenotypes were verified as described (28), following resuspension in 0.9% NaCl and re-plating in the presence of the parental M13mp2A89 to provide a standard for assessing the blue color.
Primer Extension Assays-Assays were performed essentially as described (36) with modifications for different substrates and buffer conditions. Oligonucleotides were manufactured by the Laboratory of Molecular Systematics and Evolution (University of Arizona), except for the C-G oligo, which is the universal (Ϫ40) M13 sequencing primer (New England Biolabs). Oligos were 5Ј-labeled with 32 P by polynucleotide kinase reactions (41). Primed M13mp18 DNA was prepared by annealing 10 pmol of 5Ј-32 P-labeled oligomer to 20 pmol of singlestranded phage DNA. The oligos contained a terminal 3Ј-base that was either matched (C-G) or mismatched (T-G or A-G), as follows.

5Ј-GTTTTCCCAGTCACGAC
...TCCCAAAAGGGTCAGTGCTGCAACATTTT...   Reactions (10 l) were conducted as described for the mutagenesis assays (above) for 15 min at 37°C, using 0.05 units of wild type or 0.06 units of D368A mutant polymerase, 0.1 pmol of primer/template, and one or more dNTPs. Reactions were stopped by rapid cooling on ice and addition of EDTA (pH 8.0) to 30 mM followed by ethanol precipitation. Pellets were resuspended in 20 l of H 2 O and 10 l of loading buffer (95% formamide, 0.25% bromphenol blue, and 0.25% xylene cyanol FF). Samples were run on denaturing 20% polyacrylamide gels containing 8 M urea. Gels were scanned on a Molecular Dynamics PhosphorImager, bands quantified using the program ImageQuant (Molecular Dynamics), and K m and V max values calculated by fitting the data to rectangular hyperbola by nonlinear regression. Values for V max corresponded to the extrapolated velocity at infinite substrate concentration, and values for K m corresponded to the substrate concentration at 1 ⁄2V max .

RESULTS
To gain insight into the mechanism of the 3Ј-5Ј-exonuclease associated with the HSV-1 DNA polymerase, we have compared the biochemical characteristics of the wild type polymerase and a mutant carrying the D368A mutation. As shown in Fig. 1, this mutation lies within the Exo I motif that is highly conserved between the HSV-1 polymerase and DNA polymerase I (E. coli). Structural studies with DNA polymerase I indicate that the corresponding residue, Asp-355, is essential for exonuclease function because it binds two Mg 2ϩ ions required for catalysis (4,42,43). We previously reported that the D368A mutation in HSV-1 DNA polymerase reduces exonuclease activity to undetectable levels (9, 24), demonstrating that this residue is also required for enzyme function.
Purification of the HSV-1 DNA Polymerase-To obtain enzymes for these experiments, wild type and D368A mutant polymerases were overexpressed using baculovirus vectors (9) and purified as described under "Experimental Procedures" and Table I. Although we had previously purified the wild type HSV-1 DNA polymerase to near-homogeneity (24), the D368A mutant polymerase failed to survive the glycerol gradient step in this original procedure (not shown). Hence, we devised a new purification scheme using a three-step procedure (Table I) consisting of hydrophobic interaction chromatography on phenyl-Sepharose, anion exchange chromatography on phosphocellulose, and affinity chromatography. In the affinity chromatography step, the wild type polymerase was purified using single-stranded DNA agarose. The D368A mutant enzyme failed to bind to this matrix (not shown) and was consequently purified on heparin-Sepharose. Gel analyses (not shown) reveal no large differences between these two preparations in the levels of minor contaminants. Hence, the difference in specific activities between these polymerases is most likely due to differences in the relative fractions of active enzyme molecules.
The identity of the purified enzymes was confirmed by two biochemical tests. First, the enzymes were assayed for polymerase activity in the presence of phosphonoacetic acid, a specific inhibitor of the HSV-1 polymerase. As shown in Fig. 2A, both the wild type and D368A mutant enzymes were severely inhibited by phosphonoacetic acid at concentrations at or above 10 g/ml. In contrast, DNA polymerase I (Klenow fragment) from E. coli remained relatively active at concentrations up to 100 g/ml. Second, both enzymes were subjected to exonuclease measurements using a single-stranded DNA substrate. As shown in Fig. 2B, the wild type enzyme contains exonuclease activity, and the D368A mutant is completely deficient. This observation confirms our earlier finding that the D368A mutant lacks detectable exonuclease activity (9) and indicates a lack of contaminating exonucleases in the mutant preparation.
Reversion Frequencies-Studies from several systems, including HSV-1 (11), have revealed that DNA polymerases with inactive editing nucleases show increased mutation frequencies compared with wild type (1). The D368A mutation in HSV-1 inactivates the polymerase-associated exonuclease but also appears to be lethal to the virus (9). Hence, we suspected that an editing deficiency might increase the viral mutation frequency, producing intolerable levels of deleterious mutations.
To test this idea, we compared the reversion frequencies of purified wild type and D368A mutant HSV-1 polymerases in vitro. A partially duplex, gapped M13 bacteriophage DNA molecule was used as a primer/template for these polymerase reactions. The gap contained a mutant lacZ␣ gene (from E. coli) carrying a single base change that produces a nonsense codon. The polymerase reaction allowed filling of the gap by DNA synthesis and production of LacZ revertants by replication errors at the nonsense codon site. These revertants were de-tected by transfecting the reaction products into E. coli and scoring for blue M13 plaques in agar containing 5-bromo-4chloro-3-indoyl-␤-D-galactopyranoside.
Results from this assay (Table II) showed that the D368A mutation caused significantly elevated reversion frequencies compared with wild type but that the reversion rate was dependent on dNTP concentrations. We conducted additional analyses to understand the substantial variation observed between reactions carried out under different conditions and to identify conditions most appropriate for comparing the activities of the wild type and D368A mutant polymerases. Based on these analyses, we conclude that the wild type enzyme edits many of its replication errors at low dNTP concentrations but favors mismatch extension at higher concentrations, thus producing higher reversion rates. In contrast, the D368A mutant polymerase fails to replicate the template efficiently at low dNTP concentrations, possibly due to stalling at mismatches, but is able to extend mismatches at higher concentrations. Under conditions where the wild type exonuclease is fully functional and both enzymes completely replicate the gapped template (i.e. 10 3 M dNTPs), the mutant showed an 18-fold higher reversion rate than wild type.
We first examined the effect of dNTP concentrations on the wild type polymerase. In this case, the reversion frequency remained low at 10 to 10 3 M but increased 6.7-fold at 10 4 M. Experiments with DNA polymerase II from E. coli (25), human, and calf DNA polymerase ␦ (26, 27) and DNA polymerase I (Klenow fragment) from E. coli (27) have shown similar increases in reversion rates but at dNTP concentrations of 10 3 M Numbers indicate the N-terminal residue in each Exo sequence. The aspartic acid residue in bold from DNA polymerase I binds two divalent metal ions that are required for exonuclease function (42). A mutation at residue Asp-368 in the HSV-1 polymerase was used for experiments described in the text. Polymerases were purified from recombinant baculovirus-infected Sf9 cells as described under "Experimental Procedures." Sequential purifications were performed as indicated. After each purification step, the peak fractions were pooled and analyzed for protein concentration and enzyme activity. Activity measurements were determined using enzyme dilutions that exhibited linear incorporation with incubation time. One unit is defined as the amount of enzyme necessary to convert 10 nmol of nucleotides to acid-insoluble material in 30 min at 37°C. ssDNA, single-stranded DNA. or less. It was concluded from these previous studies that high levels of dNTPs shift the balance between editing and extension of the mismatch in favor of extension. Mismatch mutations become fixed, and the reversion frequency rises. It seems likely that the elevated reversion frequency seen here for the HSV-1 DNA polymerase at high dNTP levels is also caused by enhanced mismatch extension. Our observation that this effect occurs only at very high dNTP concentrations (10 4 M) suggests that the exonuclease of this enzyme is exceptionally active. Hence, to establish a reversion level for the wild type polymerase that occurs in the presence of a fully active editing exonuclease, dNTP concentrations of 10 3 M or lower are required.
For the D368A exonuclease-deficient polymerase, we observed an 8-fold increase in the apparent reversion frequency when the dNTP concentration was raised from 10 to 10 3 M. In contrast, the MLV reverse transcriptase, which also lacks an exonuclease activity, showed nearly the same frequency at both these dNTP concentrations. Hence, we suspected that the change in reversion frequency seen for the D368A mutant might be explained by some effect of dNTPs on the polymerase reaction. For example, the polymerase might stall at misinserted bases in the presence of low dNTPs but be able to extend past the misinsertions at higher concentrations. Partially replicated molecules produced by stalling might be repaired after transfection into E. coli leading to a lower observed reversion frequency. Although the majority of our polymerase reactions appeared by electrophoretic analysis to have completely filled the gap (not shown), a small fraction of partially replicated molecules would not be detectable by this test and might still be present.
To test the possibility that a stalling mechanism produced the low apparent reversion rate of the D368A mutant enzyme at 10 M dNTPs, we used a two-step protocol. We first allowed the polymerase to replicate in the presence of 10 M dNTPs and then increased the dNTP concentration to either 10 3 or 10 4 M. If stalling had occurred at 10 M, then the subsequent increase in dNTP concentration should force the polymerase past the mismatch, allowing complete replication. As shown in Table II, when the D368A mutant polymerase was subjected to the twostep protocol, the reversion rate was comparable to that seen when high dNTPs were present throughout the reaction and was substantially higher than when the reaction was conducted entirely at 10 M dNTPs. In contrast, when the wild type HSV-1 polymerase was subjected to the two-step protocol (i.e. a shift from 10 to 10 4 M), the reversion frequency was about the same as when 10 M dNTPs were present throughout the reaction. These results suggest that whereas the wild type polymerase completely replicates the gap at 10 M dNTPs without stalling, the D368A mutant stalls at low dNTP levels, leading to a lower observed reversion rate, but completes gap filling at higher dNTP levels (Ͼ10 M). Hence, the true reversion frequency of the D368A mutant polymerase is represented by the values observed at high dNTP concentrations.
The data in Table II also show reversion assays using reverse transcriptases from Moloney murine leukemia and avian myoblastosis viruses (MLV-RT, AMV-RT). These enzymes lack any editing function and are known to exhibit high mutation rates (20). These assays reveal that, in the presence of high dNTP levels (Ն10 3 M), the D368A HSV-1 mutant reversion frequencies are significantly higher than those of the reverse transcriptases (7.2-fold for MLV-RT and 3.5-fold for AMV-RT). This comparison suggests that the D368A mutant HSV-1 polymerase exhibits an exceptionally high mutation rate, which is substantially greater than those of the strongly mutagenic reverse transcriptases.
From these reversion data, we can calculate a corresponding base substitution rate for the D368A mutant polymerase, using the method of Bebenek and Kunkel (28). This calculation corrects for spontaneous background revertants, for the extent of revertant expression during scoring, and for the fact that revertants can arise by any of several changes within the target nonsense codon. By using this method, we find that a reversion rate of 2.11 ϫ 10 Ϫ4 (D368A at 10 3 M dNTPs, Table II) corresponds to 1.2 ϫ 10 Ϫ4 errors per replicated base.
Reversion frequencies from Table II were used to generate bar graphs in Fig. 3 to evaluate the statistical significance of these data. Non-overlapping vertical error bars indicate statistically significant differences between values. All values for polymerase reactions are significantly greater than that of the control without enzyme (p Յ 0.005). In reactions with the wild type HSV-1 polymerase, the value at 10 4 M is significantly higher than those obtained under other conditions (p Յ 0.001), whereas these other values do not differ significantly from each other. This analysis supports our hypothesis that the wild type favors mismatch primer extension, rather than editing, at 10 4 M dNTPs. In reactions with the D368A mutant polymerase, the value at 10 M dNTP concentration is significantly different from the other four values (p Յ 0.001), and these other values do not differ significantly. This result supports our contention that the reactions at 10 M dNTPs do not represent true reversion frequencies but rather result from incomplete gap filling after misinsertions. Finally, the D368A mutant values at dNTP levels that allow complete gap filling were significantly higher than either the wild type values obtained at dNTP levels that allow editing (p Յ 0.001) or values for the MLV (p Յ 0.001) and AMV (p Յ 0.01) reverse transcriptases. These comparisons confirm our conclusion that the D368A mutant DNA polymerase exhibits exceptionally high mutation rates in vitro, significantly higher than those of the wild type HSV-1 polymerase

Reversion of a LacZ␣ mutation by HSV-1 DNA polymerases and reverse transcriptases in the presence of varying dNTP concentrations
Assays were conducted with purified polymerases using a gapped M13mp2A89 primer/template, as described under "Experimental Procedures." Reversion of the LacZ␣ mutation within the template was scored by the appearance of blue plaques. Each reaction contained equimolar quantities of dNTPs at the indicated concentrations. Arrows indicate two-step reactions in which a given reaction mix, containing 10 M dNTPs, was first incubated for 60 min, and then additional dNTPs were added, to a final concentration of 10 3 or 10 4 M (as indicated), and incubation was continued for 60 min. Reverse transcriptases were from MLV-RT and AMV-RT. Means Ϯ S.E. are calculated and displayed in Fig. 3. Values in parentheses are the ratio of the mutant reversion frequency compared with that of wild type. ND, not determined. Mismatch Extension-The reversion experiments described above suggest that the D368A mutant polymerase stalls during replication at low dNTP levels, probably at sites of misinsertion. Since the D368A mutation also appears to be lethal (9), we suspected that inefficient extension of mismatched primers might severely impair viral replication in vivo. To obtain evidence for a stalling mechanism, we conducted experiments in vitro to compare the efficiency of primer extension by the wild type and the D368A mutant polymerases.
This analysis was conducted using primer/templates consisting of 5Ј-labeled oligonucleotides annealed to single-stranded M13 DNA. Primers contained a 3Ј-terminal nucleotide that was either matched or mismatched to the template. The HSV-1 polymerases were allowed to extend the primers by five nucleotides in the presence of one or two dNTPs. Because the D368A mutant polymerase cannot excise mismatches, full-length products with a mismatched primer should only be produced by this polymerase by extension of the primer. However, the wild type polymerase should be able to excise a mismatch and, depending on the dNTPs present, should be able to insert the correct nucleotide or should be forced to re-insert the mismatch prior to extension. Products from reactions at varying concentrations of dNTPs were analyzed by gel electrophoresis, and the percentages of fully extended primers were measured (Fig. 4). These data were used to calculate the kinetic values K m and V max to determine extension efficiencies (V max /K m ) (Table III).
From these analyses, we observed that both the wild type and D368A polymerases extend a C-G matched terminus with approximately equal efficiency. As calculated in Table III, the extension efficiency (V max /K m ) was 375 (%⅐min Ϫ1 ⅐M Ϫ1 ) for the wild type polymerase and 509 (%⅐min Ϫ1 ⅐M Ϫ1 ) for the D368A mutant. Hence, the D368A mutation does not have a substantial effect on the efficiency of polymerization.
In reactions with a mismatched primer, both the wild type and D368A mutant polymerases exhibit impaired extension efficiencies compared with the matched primer. In general, this response results from large increases in the K m with relatively small changes in V max . These results suggest that a correctly matched primer is necessary for stable binding of the next nucleotide.
The efficiency of the extension of the mismatched primer depended dramatically on the potential for editing of the mismatch. In reactions with a T-G mismatched primer, the wild type polymerase showed an efficiency of 0.0088 (%⅐min Ϫ1 ⅐M Ϫ1 ) when only dTTP was present, a condition that disallows productive editing. When the correct nucleotide (dCTP) was also added, the extension efficiency increased (180-fold) to 1.6. Thus, when productive editing is prevented, the wild type polymerase stalls for a substantially longer time than when editing is allowed. Similar but more dramatic results were obtained with an A-G mismatch, and a 4800-fold increase in extension efficiency was observed when productive editing was allowed. We believe that the larger increase in extension efficiency of the A-G mismatch as compared with the T-G mismatch results from the decreased stability of the A-G base pair, causing increased editing efficiency and/or decreased extension efficiency. In contrast to the wild type, the D368A mutant polymerase failed to show substantial increases in extension efficiencies of either the T-G or A-G mismatch when two dNTPs were present. This result is consistent with the editing deficiency of the D368A mutant polymerase.
To compare extension efficiencies of the wild type and D368A mutant polymerases, we calculated standard extension efficiencies for each enzyme by dividing the mismatched extension efficiencies by the matched efficiency, as shown in Table III. This comparison revealed a defect in the D368A mutant in mismatch extension. For the T-G mismatch, when two dNTPs are present, allowing productive editing by the wild type polymerase, the wild type enzyme extended much more efficiently (380-fold) than the D368A mutant. In contrast, when only one dNTP was present, the wild type polymerase extended slightly less efficiently (0.27-fold) then the D368A mutant. Similar results were obtained with the A-G mismatch. In the presence of both dNTPs, the wild type extended 590-fold more efficiently than the mutant, whereas with only one dNTP it extended less efficiently (0.22-fold). These results suggest that the D368A mutant polymerase stalls substantially at mismatches compared with a productively editing wild type polymerase.
Evidence for Polymerase Cycling-In the primer extension experiments above, the wild type polymerase is less efficient at extending mismatched primers when editing is prevented. We  Table II and are shown here in bar graph format with error bars indicating the S.E. S.E.s were calculated by dividing the standard deviation from the mean revertant frequency by the square root of the total number of plaques scored. Arrows indicate a two-step protocol in which the dNTP concentration is increased after an initial incubation at 10 M. The Student's t test for statistical significance was used for comparing wild type and mutant frequencies (p Ͼ 0.99).
suspect that under these conditions, the wild type enzyme becomes trapped in a cycle of removal and reinsertion of the mismatch. To test this possibility, we conducted dNTP turnover assays. The polymerase was incubated with an unlabeled primer/template carrying a T-G mismatch at the primer 3Ј-end in the presence of [␣-32 P]dTTP. Reaction products were analyzed by thin layer chromatography and autoradiography. As shown in Table IV, radioactive dTMP was generated in the course of these experiments, indicating incorporation and then hydrolysis of the dTTP by the polymerase. Based on the amount of primer/template present, we calculated that approximately 50 dTMP bases were generated for every primer/template during the 30-min reaction time. This result is consistent with the possibility that the low extension efficiency of the wild type polymerase in the absence of productive editing results from cycling of the wild type polymerase at the mismatch.
Effects of Internal Mismatches on Extension Efficiency-Because our primer/templates allowed incorporation of multiple nucleotides after a mismatch, we were able to evaluate the effects of the mismatch on addition of both the next nucleotide and other downstream nucleotides. This analysis revealed that internal mismatches also promote stalling by the D368A mutant HSV-1 polymerase.
In the mismatch extension reactions shown in Fig. 4B (T-G), a band corresponding to the primer extended by one nucleotide (P ϩ 1) was typically observed, indicating a kinetically slow step for the addition of a second nucleotide onto the mismatched primer. Although not as prominent, a similar band was seen in reactions involving the extension of A-G mismatches. Kü hn and Knopf (10) have observed a similar phenomenon. Subsequent addition, up to the five allowed in these reactions, occurred very rapidly, and no further intermediate bands were observed. Hence, these later additions contribute negligibly to the extension rates.
The effect of internal mismatches is quantified in Table V for the T-G mismatched primer. K m and V max values were calculated using graphs of percent substrate extended versus dNTP concentration. Primer extension efficiencies for the addition of both the first (indicated as P 3 P ϩ 1) and second (indicated as P ϩ 1 3 P ϩ 5) nucleotides reveal that the second addition is at least as slow as the first. This effect may act as an additional pause in the elongation of a primer, amplifying the effectiveness of proofreading by increasing the chance that a mismatch will be detected. A similar effect has been noted for the T4 bacteriophage DNA polymerase (23). DISCUSSION This study examines the effect of an impaired editing nuclease during DNA replication by the HSV-1 DNA polymerase. We showed previously that the D368A mutation inactivates the exonuclease and severely interferes with virus viability (9,24). In the present experiments, we find that this defect increases the production of replication errors but only in the presence of high levels of dNTPs to favor the extension of mismatched primers. In addition, this mutation reduces the efficiency of extension of 3Ј-mismatched primers. These abnormalities appear to be exceptionally severe when compared with other replicases. Hence, loss of editing activity in the HSV-1 DNA polymerase may interfere with viral growth by enhancing the production of deleterious mutations and/or by prematurely terminating replication forks at sites of misinsertions.
High Mutation Rates in the D368A Mutant-Inactivation of editing exonucleases increases mutation rates in many systems (1). Elevated mutation frequencies have been reported for HSV-1 derivatives carrying mutations in the highly conserved Exo III motif within the 3Ј-5Ј-exonuclease domain of the DNA polymerase (11). However, viruses with mutations in the conserved Exo I motif appear to be inviable (9). Therefore, we have analyzed the mutation frequency of the Exo I D368A mutant polymerase in vitro by measuring reversion of a nonsense codon. This analysis reveals up to an 18-fold increase in the reversion frequency of the D368A mutant polymerase compared with wild type. Although the reversion levels vary with the reaction conditions, we argue that this maximum increase best reflects the intrinsic difference between these enzymes. This difference was obtained using a dNTP concentration of 10 3 M, a condition that allows maximum activity of the wild type editing exonuclease and efficient replication of the template/ primer by both the wild type and mutant enzymes.
A comparison of our findings with other studies reveals that reversion frequencies for the wild type HSV-1 DNA polymerase resemble those of other mammalian polymerases with editing functions. In contrast, the mutant HSV-1 polymerase exhibits an exceptionally high reversion frequency, even higher than enzymes that naturally lack an editing function.  Table III. ( Table II).
The large increases in mutation frequencies exhibited by our HSV-1 D368A mutant and by other HSV-1 exonuclease mutants described by Hwang et al. (11) suggest that the wild type HSV-1 polymerase readily incorporates mis-paired nucleotides during replication and relies substantially on its editing function to correct these errors. Poor insertion fidelity is also suggested by reports that the HSV-1 DNA polymerase readily misincorporates nucleotide analogs (29 -31) and by analyses of HSV-1 anti-mutators which show decreased viral mutation rates due to enhanced polymerase insertion specificity (32,33). The 3Ј-5Ј-exonuclease of the HSV-1 DNA polymerase appears to be exceptionally active. We find that, although high levels of dNTPs shift the balance between editing and mismatch extension to favor extension, these levels are much higher than levels that suppress exonucleases in other systems (27,34,35). Hence, loss of the HSV-1 exonuclease may have catastrophic consequences, leading to high levels of deleterious mutations.
Stalling of the D368A Mutant at Mismatches-Primers with mispaired 3Ј termini slow primer extension by DNA polymerases (1,36), providing an opportunity for editing. Processivity defects have been reported for exonuclease-deficient mutants in several organisms (17,21,22,37), suggesting that misinsertion events might block replication or promote polymerase disassociation when the editing exonuclease is inactivated. Evidence supporting these stalling mechanisms occurs in T4 bacteriophage, where exonuclease-deficient polymerases are defective in extension of mismatched primers and exhibit poor replication of templates which allow misincorporation (23). By using a more quantitative approach, our studies indicate that the loss of exonuclease in the HSV-1 DNA polymerase results in an exceptionally severe defect in mismatch extension. This defect may contribute to the inviability of the D368A mutant.
To study mismatch extension by the HSV-1 DNA polymerase, we measured extension efficiencies of primers annealed to M13 bacteriophage DNA. By controlling the types of dNTPs, we could determine whether the polymerase had the potential to correct the mismatch. The wild type polymerase rapidly extended the primer when editing was allowed. When editing was inhibited, extension was reduced, and the polymerase entered into repeated cycles of removing and re-inserting the mismatch, thus interfering with extension. In contrast, primer extension by the D368A mutant DNA polymerase was several hundred-fold less efficient than the wild type under conditions where the wild type could edit.
Comparison of our primer extension data with those of other studies reveals that the D368A mutant polymerase stalls at mismatches considerably longer than at least two other polymerases that naturally lack an editing nuclease. This comparison is possible since in each case the extension efficiency for mismatched primers is standardized to the efficiency for matched primers, thus eliminating intrinsic differences in efficiency of matched primer extension for different enzymes. This comparison reveals that the D368A mutant polymerase extends a T/G mismatch 40-fold less efficiently than Drosophila melanogaster DNA polymerase ␣ and 110-fold less efficiently than AMV reverse transcriptase (36). For an A/G mismatch, although both polymerase ␣ and reverse transcriptase extend this mismatch much less efficiently than a T/G mismatch, the HSV-1 D368A mutant is again more inefficient (30-and 2-fold less, respectively). Hence, the D368A mutant enzyme appears to be severely inhibited in its ability to extend mismatched primers.
Editing Defects and Loss of Viability-Editing defects have

TABLE V Kinetics of addition of individual nucleotides after a T-G mismatch by
D368A mutant HSV-1 DNA polymerase DNA polymerase reactions were conducted as shown in Fig. 4, using conditions that allowed extension of the primer, of length P, by up to five nucleotides. The kinetically slow steps of addition of the first base past the mismatch (P 3 P ϩ 1) and addition of the second and subsequent bases (P ϩ 1 3 P ϩ 5) were measured independently from the same autoradiograms. 3Ј-5Ј-Exonuclease of HSV-1 DNA Polymerase been linked to lower viability in several systems (9,11,(17)(18)(19). A model frequently proposed to explain this phenomenon is error catastrophe due to an overproduction of deleterious mutations. In support of this model, certain lethal mutations in DNA polymerase III (E. coli) become viable if coupled with a genetic change that lowers the high mutation rate by enhancing the insertion specificity of the polymerase or by raising the mismatch repair level (13). In T4, those exonuclease-deficient phages with the highest mutation rates show the strongest effects on viability (23). Our studies of the D368A mutant HSV-1 DNA polymerase suggest that high mutation rates might contribute to the low viability of this mutant. From our reversion data, we calculate that this mutant polymerase makes errors in vitro at 1.2 ϫ 10 Ϫ4 per replicated base. Although in vivo mutation rates should be influenced by additional replication enzymes, if the local dNTP concentration at the replication fork is sufficiently high, it seems likely that a D368A virus would make multiple misinsertions during replication of its 1.5 ϫ 10 5 base pair genome. However, our data on primer extension suggest that stalling at mismatches may also contribute to the impaired viability of exonuclease mutants. Our experiments show that, depending on the type of mismatch, the D368A mutant HSV-1 DNA polymerase stalls 300 -600-fold longer than the wild type. At this level, assuming multiple misinsertions within the HSV-1 genome, we predict that replication of such mutant viruses might be reduced by several thousand-fold. This effect, coupled with the accumulation of deleterious mutations, might explain the severity of the D368A phenotype. However, other possibilities such as poor interactions of the polymerase with other replication proteins have not been ruled out.
Since other exonuclease mutants in HSV-1 are viable (11), the defect in our D368A mutant appears to be exceptionally severe. This mutant carries an altered aspartic acid residue within the Exo I motif of the exonuclease site. The Exo motifs are highly conserved in 3Ј-5Ј-exonucleases (7) and bind two divalent metal ions, based on the structural model of E. coli DNA polymerase I (4,38). This model suggests that mutations in the Exo I motif might lead to loss of both ions, whereas Exo II or Exo III mutations might lead to loss of only one (5). Hence, the Exo I defects may be especially debilitating. The poor mismatch primer extension efficiency of the HSV-1 D368A mutant suggests that this mutant may trap mismatched primers in non-productive complexes within the altered exonuclease site, thus blocking further replication.
Summary-We have characterized the D368A mutation, which inactivates the editing nuclease of the HSV-1 DNA polymerase. We show that the D368A mutation causes the polymerase to exhibit an exceptionally high mutation frequency and an exceptionally low ability to extend mismatched primers. These phenotypes appear more severe than those reported for other polymerases that naturally lack editing functions. The high mutation rate and low mismatch extension ability of the D368A mutant may substantially reduce the extent of viral replication and may explain the apparent lethality of this mutation.