Mutations in the Primer Grip Region of HIV Reverse Transcriptase Can Increase Replication Fidelity*

Mutations in the primer grip region of human immunodeficiency virus reverse transcriptase (HIV-RT) affect its replication fidelity. The primer grip region (residues 227–235) correctly positions the 3′-ends of primers. Point mutations were created by alanine substitution at positions 224–235. Error frequencies were measured by extension of a dG:dA primer-template mismatch. Mutants E224A, P225A, P226A, L228A, and E233A were approximately equal to the wild type in their ability to extend the mismatch. Mutants F227A, W229A, M230A, G231A, and Y232A extended 40, 66, 54, 72, and 76% less efficiently past a dG:dA mismatch compared with the wild type. We also examined the misinsertion rates of dG, dC, or dA across from a DNA template dA using RT mutants F227A and W229A. Mutant W229A exhibited high fidelity and did not produce a dG:dA or dC:dA mismatch. Interestingly, mutant F227A displayed high fidelity for dG:dA and dC:dA mismatches but low fidelity for dA:dA misinsertions. This indicates that F227A discriminates against particular base substitutions. However, a primer extension assay with three dNTPs showed that F227A generally displays higher fidelity than the wild type RT. Clearly, primer grip mutations can improve or worsen either the overall or base-specific fidelity of HIV-RT. We hypothesize that wild type RT has evolved to a fidelity that allows genetic variation without compromising yield of viable viruses.

dressed the role of RT in the variability of the genome (4 -14). RT contributes to the generation of sequence diversity, partly because it produces frequent replication errors. One reason for its low fidelity is that RT lacks a 3Ј to 5Ј exonuclease activity (15). DNA polymerases in other organisms generally have this activity, which removes incorrectly added nucleotides by recognizing the mismatched base with the template sequence. RT misincorporates nucleotides at an estimated frequency of 1 per 5000 polymerized and extends mismatched termini at varying efficiencies (4, 6, 11, 14 -20). Additionally, RT alters the sequence of viral progeny by participating in recombination between the two copies of the viral genome (13,(21)(22)(23)(24)(25)(26)(27)(28). Polymerization errors and recombination produce a high frequency of frameshift, deletion, and deletion with insertion mutations observed both in vivo and in vitro (6,7,11,13,21,29). This proclivity for mutations by RT helps HIV to evade immune responses and drug treatment.
HIV-1 RT is a heterodimer composed of p66 and p51 subunits. The p66 subunit contains both the polymerase and RNase H active sites. The p51 subunit is a proteolytic product of p66 lacking the 15-kDa carboxyl-terminal (RNase H) domain (30). Crystallographic studies have shown that the p51 folds into a different conformation than the p66, which prevents the p51 from having functional polymerase activity (31)(32)(33)(34). In the native RT structure, the p66 folds into a conformation that resembles a right hand. The subdomains of this subunit are referred to as fingers, palm, thumb, connection, and RNase H. The polymerase active site resides on the palm subdomain within conserved residues Asp-185, Asp-186, and Asp-110 (31)(32)(33)35). The 3Ј-end of the primer is positioned near this active site within conserved residues 224 -235 on the ␤11b loop and ␤12 and ␤13 hairpin of the RT molecule. Residues 227-235 are referred to as the primer grip region (33). One striking feature of this region is that residues Phe-227, Trp-229, and Leu-234 line the binding pocket for non-nucleoside RT inhibitors, yet no drug-resistant mutations occur at these sites (36). This attests to the importance of these amino acids for polymerase function.
Recent analyses of amino acid substitutions have shown the primer grip region to influence many different RT functions. The W229A mutation reduces RNA and DNA-primed synthesis (37)(38)(39)(40)(41)(42). Since the binding affinity of this mutant is 27-fold lower than that of wild type RT (41), we believe that this mutation results in lower processivity of the enzyme. Also, residues 226, 227, 229, 230, 231, 233, and 235 are involved in proper positioning of RT for extension and processing of the polypurine tract primer (40). RNA-primed minus strand synthesis from the primer binding site is reduced or eliminated by mutations in residues 226, 227, 229, 231, 232, and 235 (41). From these results, several groups have proposed that alterations in the primer grip region affect RNA versus DNA-primed synthesis differently (37,38,41). Surprisingly, mutations in the primer grip residues also affect the distant RNase H domain. Residues 226 and 227 have been shown to be involved in positioning RT correctly for 5Ј-end-directed cleavage of RNA fragments remaining after minus strand synthesis (39). Mutations P226A, F227A, L228A, M230A, G231A, Y232A, E233A, and H235A had reduced RNase H activity for the specific cleavage involved in the removal of the polypurine tract (38,40). These results highlight the unique roles of the individual primer grip amino acids.
In this study, we examined the effects that alterations in the primer grip region have on the error rate of RT. Recent work has shown that mutations in amino acids involved in positioning of the primer-template can alter fidelity (10,35,43,44). Specifically, Beard et al. (43,44) have shown that mutations in residues Gly-262 and Try-266 on the ␣H-helix contacting the sugar phosphate backbone of the primer lowered the dissociation rate and the fidelity of RT. Also, Kim et al. (45,46) have made amino acid substitutions in the finger regions involved in template positioning that alter fidelity. This group found that mutation D76R increases overall fidelity. Also, some nucleoside analogue-resistant mutations of RT, such as E89G, L74V, and K65R, which affect primer-template interactions and dNTP binding, display higher fidelity than wild type enzyme (47)(48)(49)(50)(51)(52)(53). Based on these observations, we examined the role of the primer grip region in determining the fidelity of RT by using alanine-scanning mutagenesis of each residue (42). We measured mismatch extension capacity for each mutation and base substitution efficiencies for mutations F227A and W229A. Our results showed that a number of residues in this region affect fidelity, and each residue alters error rates in a different manner. We conclude that alterations in the ability of the RT to position a primer influence fidelity. Since single amino acid changes can readily improve or worsen fidelity, we suggest that the intermediate fidelity of the wild type RT provides an evolutionary advantage.

Materials
T4 polynucleotide kinase, dNTPs, alkaline phosphatase, and spin gel filtration columns were purchased form Roche Molecular Biochemicals. Radiolabeled nucleotides were from NEN Life Science Products, and oligonucleotides were purchased from Genosys Biotechnologies, Inc.

Construction and Purification of the Primer Grip Mutant RTs-The
individual substitution of p66 residues Glu-224 to His-235 by alanine was performed by using BcgI cassette mutagenesis as described previously (42). The mutated p66 subunit was expressed separately from the p51 subunit with a six-residue histidine tag (His 6 -p51). Extracts containing the primer grip mutated p66 and the His 6 -p51 were mixed to produce a heterodimer. The reconstituted RT was purified on a nickelnitrilotriacetic acid-Sepharose column and by ion exchange chromatography (S-Sepharose). The purified protein was examined by SDS-polyacrylamide gel electrophoresis (42).
Labeling of DNA Primers-DNA primers were synthetically generated by Genosys Biotechnologies, Inc. and were 5Ј-end-labeled using [␥-32 P]ATP (3000 Ci/mmol) and 10 units of T4 polynucleotide kinase. This reaction was incubated at 37°C for 1 h and stopped by heating at 65°C for 5 min. Excess radionucleotides were removed by spin gel filtration columns, and the primers were purified by 15% polyacrylamide gel electrophoresis. The DNA primers were removed from the gel using an elution buffer (0.1% SDS, 1 mM EDTA, and 0.5 M ammonium acetate) and resuspended in water treated with diethyl pyrocarbonate to ensure that they did not have RNase H contamination when hybridized to an RNA template.
DNA-DNA or RNA-DNA Hybridization-The RNA or DNA templates were annealed with the DNA primers in 10 mM Tris-HCl (pH 8.0), 1 mM EDTA, and 80 mM KCl for 10 min at 65°C. The substrate was allowed to slow cool over 90 min.
Mismatch Extension Assay-The reaction contained 2 nM substrate, 50 mM Tris-HCl (pH 8.0), 1 mM dithiothreitol, 1.0 mM EDTA, 34 mM KCl, and 2 units of mutant or wild type HIV-1 RT (1 unit is equal to 4000 nmol of dTMP incorporated in poly(rA:dT) 20 /min/mg of protein). This was preincubated at 37°C for 5 min to allow prebinding of the protein. The reaction was started by the addition of MgCl 2 (6 mM final concentration) and 50 M of each dNTP. Reactions were stopped after 20 min with a 2ϫ termination dye (90% formamide, 10 mM EDTA (pH 8.0), and 0.1% each of xylene cyanole and bromphenol blue). Products were fractionated by 15% denaturing polyacrylamide gel electrophoresis, and the data were quantified by PhosphorImager analysis (Molecular Dynamics, Inc., Sunnyvale, CA). Relative synthesis by the mutant versus wild type RTs was determined in the following way. The amount of synthesis past the mismatch was divided by the amount of synthesis on a correctly matched substrate. This worked out to 33% for the wild type RT when comparing substrates A and B. The percentage of synthesis past the mismatch was determined in the same way for the mutants. This standardizes the activities of the enzymes. The relative ability of wild type and mutant RTs to synthesize past the mismatch was calculated by dividing the percentages.
Running Start Fidelity Assay-We performed a running start experiment as described by Goodman and co-workers (54 -57). The reactions contained 20 nM substrate, 50 mM Tris-HCl (pH 8.0), 1 mM dithiothreitol, 1.0 mM EDTA, 34 mM KCl, and 50 ng of mutant or wild type HIV-1 RT. The substrate was incubated for 5 min to allow the enzyme to prebind. The reactions were started with 6 mM MgCl 2 and varying concentrations of dATP, as indicated in the figure legends. The reaction was stopped after 2 min with 2ϫ termination dye.
Standing Start Fidelity Assay-Reactions were performed as in the running start experiment with some exceptions. The reactions were started with 6 mM of MgCl 2 and varying concentrations of dTTP, dATP, dGTP, or dCTP (4, 3, 2, 1, 0.5, 0.1, and 0.05 mM). Mutant W229A required a 15-min reaction in order to obtain a 20% extension from the DNA primer. The products were fractionated by 20% denaturing polyacrylamide gel electrophoresis, and the data were quantified by Phos-phorImager analysis. Equations used to determine the fidelity of the mutants and wild type RT are based on Goodman's fidelity assays and obey Michaelis-Menten kinetics (54 -57) as follows.
In the Michaelis-Menton equation, V pol is defined as the rate of addition of nucleotides to the primer. This is an important factor to measure in a fidelity determination, because it allows a comparison of rates of polymerization for incorrect versus correct nucleotides. To determine V pol in these experiments, we used a method developed by Goodman and colleagues (54 -57) that requires the intensities of the extended product (I e ) and the unextended primer (I u ) as follows, where t represents the reaction time. A Lineweaver-Burk graph of 1/V pol versus 1/[dNTP] was plotted to obtain the V max and K m values for the insertion of the incorrect or correct nucleotide used in this experiment. These values were used to calculate the frequency of insertion, which is a ratio of incorrect to correct insertions as follows.
The fidelity of these enzymes was obtained as the reciprocal of the frequency of insertion (31). Primer Extension in the Presence of Biased dNTP Pools-Procedures were identical to those described by Kim et al. (45). These reactions contained a 63-nt DNA template annealed to a 32 P-labeled 14-nt DNA primer at a ratio of 2.5:1 template to primer. The 20 l assay mixture contained 100 nM template-primer, three or four dNTPs at 250 M each, 25 mM Tris-HCl (pH 8.0), 40 mM KCl, 2 mM dithiothreitol, 5 mM MgCl 2 , and 0.1 mg/ml bovine serum albumin. Wild type RT, F227A, and W229A were added in four different quantities (wild type: 0.05, 0.5, 5, and 50 ng; F227A and W229A: 25, 50, 100, and 200 ng). The reactions were incubated at 37°C for 5 min, and the reactions were terminated with 2ϫ termination dye. The reactions were analyzed on denaturing 15% polyacrylamide gels.

RESULTS
The primer grip region of RT correctly positions the polymerase active site to the 3Ј-end of the primer for synthesis (32)(33)(34)(35). We anticipated that mutations in this region would distort primer positioning in a way that could alter the choice of nucleotides added to the primer terminus. If the active site of the wild type RT were optimized for fidelity, we would expect all mutations to be neutral or decrease fidelity.
Mismatched Nucleotide Extension by the Primer Grip Mutants and Wild Type RT on a DNA Primer-Template-The primer grip mutant RTs were generated by site-directed mutagenesis and purified essentially to homogeneity as indicated under "Methods." Substrates A and B were used to compare DNA synthesis by the wild type and mutant RTs (Fig. 1A). Substrate A consists of a DNA primer correctly matched to a DNA template, whereas substrate B has a DNA primer hybridized to the same DNA template except with a dG:dA mismatch at the 3Ј terminus of the primer. The DNA primers were radiolabeled at the 5Ј-end, and extension of these primers produced a 40-nucleotide full-length product (Fig. 2). As seen in this figure, each primer grip mutant was able to extend substrate A, the matched primer-template, with an efficiency similar to that of the wild type RT. Extension past the dG:dA mismatch was evaluated by PhosphorImager analysis, allowing comparison of each mutant to the wild type RT. The wild type extends the mismatch with 30% of the efficiency that it extends the correctly matched substrate (24). Mutations E224A, P225A, P226A, L228A, and E233A were approximately equal to wild type in ability to extend a dG:dA mismatch (compare lanes 4, 6, 8, 14, 26, and 28). F227A, W229A, M230A, G231A, and Y232A had lower levels of extension on the mismatched substrate. F227A, W229A, and M230A had 40, 66, and 54% less mispair extension past a dG:dA mismatch compared with wild type (compare lanes 12, 16, 20, and 28). Mutations G231A and Y232A had 72 and 76% less extension past this dG:dA mismatch (see lanes 22, 24, and 28). L234A was not examined, since this mutation affected dimerization (42).
Mismatch Extension on a DNA-primed RNA Template-Substrates C and D have the same sequences as A and B but with an RNA template. These substrates were used to determine possible influences that an RNA template may have on the fidelity of reverse transcription compared with a DNA template. When the 5Ј-labeled DNA primer was extended on the RNA template, a 40-nucleotide-long product was again expected. However, after gel electrophoresis, an additional longer product was observed (Fig. 3). This product is apparently the result of fold-back synthesis. As observed previously with RNA templates (58), fold-back synthesis occurs because the RNase H activity of the RT degrades the RNA template after synthesis converts it to an RNA-DNA hybrid. The elongated DNA primer can then fold back to a favored site and then continue extension until reaching the 5Ј-end of the DNA. Each of the examined RTs made the 40-nucleotide product with substrate C. The fold-back product was only observed with wild-type RT and the mutants E224A, P225A, P226A, F227A, L228A, E233A, and H235A (lanes 3, 5, 7, 9, 11, 21, 23, and 25). As discussed above, each of these mutants extended past a G:A mismatch on a DNA template with efficiency ranging from 40% less extension to equal levels of mismatch extension to wild type RT. The mutant RTs W229A, M230A, G231A, and Y232A, which were worse at extending a DNA primer-template dG:dA mismatch, failed to produce the fold-back product (lanes 13, 15, 17, and 19). A likely explanation for this phenomenon is that production of the fold-back product involves extension past a mismatch. Mutants that are not able to extend a mismatch efficiently would then be unable to further elongate the 40-nucleotide intermediate.
Mismatch extension on substrate D followed the same trends as with the DNA primer-template. E224A, P225A, L228A, E233A, and H235A performed DNA synthesis on substrate D with similar efficiency as the wild type (lanes 4, 6,12,22,24,26). F227A, W229A, M230A, G231A, and Y232A were ineffective at extending the mismatch (lanes 10, 14, 16, 18, and 20). Unexpectedly, P226A did not extend the dG:A mismatch on the RNA template, although it could extend this same mismatch on a DNA template (lane 8). Possibly, this mutation had affected the ability of the RT to extend a mismatch in the context of the A-form helix of the RNA/DNA hybrid versus the B-form DNA-DNA primer-template.
Analysis of extension efficiencies of the primer grip mutants past a mismatched nucleotide suggested that several of the amino acids influence viral mutagenesis. For analysis of nucleotide misinsertion fidelity, we decided to narrow our focus to the mutations F227A and W229A for the following reasons. As discussed previously, although these amino acids are in the nonnucleoside inhibitor binding pocket, they do not mutate in response to drug treatment, attesting to a significant role of these residues in RT function (36). Also, we previously found that Phe-227 is important for correct positioning of RT to the 5Ј-end of RNA fragments that remain after synthesis of the minus (39). Since the Phe-227 mutant RT has a positioning defect that alters RNase H function, we reasoned that it might also have a defect in positioning for polymerization, resulting in a change in fidelity. Furthermore, both the F227A and W229A RTs were unable to extend efficiently past a mismatched primer on an RNA or DNA template, suggesting that Substrate B consisted of a 30-nt DNA primer with a 3Ј-end dG:dA mismatch when annealed to the 40-nt DNA template. B shows the substrates for the mismatch extension experiments on an RNA template. Substrates C and D are homologous to the substrates in A but contain a 40-nt RNA template. C displays the substrates used in the Goodman fidelity assays. Substrate E was used for the running start experiments. The template contains two T residues after the annealed primer in order to allow synthesis to occur in the presence of dATP before the enzyme has to misinsert an A across from a G. Substrates F and G were used for the standing start experiments and consist of a 45-nt DNA template that is homologous to the nef gene. Substrate F has a 26-nt DNA primer fully annealed to this template such that the next correct nucleotide addition is a dTTP. This substrate was used to examine G, C, and A misinsertions across from a template A. Substrate G is a 23-nt DNA primer fully annealed to the template such that the next correct nucleotide addition is an A. This substrate was used to examine T, G, and C misinsertions across from a template T. the error rate was improved. It seemed unusual and significant that mutations able to decrease viral mutation rates would be so readily created. The result suggested that this region of the primer grip is particularly important for controlling fidelity. In addition to mismatch extension, nucleotide misinsertions also contribute to the ultimate introduction of errors into the DNA product. Consequently, we expanded our analysis of these mutants by measuring nucleotide misinsertion rates using the fidelity measurements devised by Goodman and colleagues (54 -57).
Running Start Misinsertion Assay for Mutant and Wild Type RTs-The running start assay examines the capacity of RT to add several correct nucleotides and then a mismatched nucleotide. The experiment was designed to measure the ability of mutants F227A and W229A to misinsert an A during replication of substrate E (Fig. 1C). Only dATP was added in increasing concentrations to a DNA synthesis reaction. The template contained two T residues after the annealed primer, allowing RT to add two correct nucleotides. In order for DNA synthesis to continue, the RT had to misinsert an A across from a tem-plate G. The primer was 5Ј-end-labeled, and the products from this reaction were analyzed on a 20% polyacrylamide gel (Fig.  4). The wild type RT was able to produce this mismatch. Also, this enzyme was able to catalyze the next misinsertion, producing a dA:dC mismatch. The majority of synthesis ceased after this mismatch was created. There is a faint extension product further than the dA:dC mismatch, but this does not correspond to full-length product. W229A and F227A were not able to catalyze a dA:dG misinsertion such that DNA synthesis ceased after the addition of the first two correctly inserted A residues (Fig. 4). These two mutants had higher fidelity for the misinsertion of an A across from a G compared with the wild type. Extension to a full-length product was not observed for either F227A, W229A, or wild type in the presence of only dATP, because none of the RTs could add the required mismatches.
Running Start Misinsertion Assay on a Template That Allows Misalignment Synthesis-We also performed a running start experiment using substrate A. In this experiment the template has two T residues and one G after the 3Ј end of the annealed primer, which is similar to substrate E. After the G, the substrate A template has two additional T residues that possibly promote a misalignment as described by Kunkel (8,10,59,60). By this mechanism, once the first two T residues in the template are extended by RT, the last A of the primer can bind to the next set of T residues in the template, causing the G to loop out. The third or fourth A could then be added. The enzyme can continue synthesis with the G looped out to create a deletion. Alternatively, the third A added to the primer could mispair with the G and the fourth could correctly pair with a T, allowing further extension. In Fig. 5, we observed F227A and wild type RTs synthesizing a misinsertion of an A across from a G on this template. Mutant F227A did not extend past the misinsertion as efficiently as the wild type RT, because the mutant required a severalfold higher concentration of dATP in order to sustain comparable synthesis past the mismatch (lanes 1-16). Furthermore, mutant F227A was not able to make a detectable misincorporation product with the same mismatch on substrate E, which did not contain the two T residues after the G on the template. This suggests that the two T residues in substrate A promote extension past the G in the template and that the F227A mutant uses the misalignment mechanism for this process. It is also possible that F227A misinserted an A across from a G. Misinsertion might have been favored by the particular sequence context of substrate A (61). W229A was not able to catalyze an A misinsertion across from G, similar to its behavior on substrate E, and this mutant was not able to  catalyze extension of a misalignment like F227A or wild type RT (Fig. 5B).
Notably, the F227A mutant catalyzed the dA:dG misincorporation and mismatch extension with a lower efficiency than the wild type RT (Fig. 5A). Extension past the 32-nucleotidelong product by F227A to create longer products having a mismatch required a 40-fold higher concentration of dATP in order to generate an equal amount of extension products as compared with wild type RT. In contrast, F227A extended the DNA primer to make a 38-nucleotide product at 0.5 mM dATP. This product is barely visible with the wild type RT at all concentrations of dATP added. To create this extension product, the mutant had to misinsert three A nucleotides across from three A residues in the template. Clearly the F227A mutant was more efficient at creating and extending dA:dA mismatches than the wild type RT. We conclude that the F227A mutant displays higher fidelity for dA:dG mismatches and lower fidelity for dA:dA mismatches than wild type. The result shows that a single amino acid substitution can increase or decrease fidelity, depending on the base pair being formed in the reaction.
Standing Start Misinsertion Assay for Mutant and Wild Type RTs-In a standing start measurement, the RT must immediately add a mismatched nucleotide. Again, the mutants F227A and W229A were compared with the wild type RT. Reactions were performed using substrate F, which contained sequence homology to the nef gene of HIV with some nucleotides changed to prevent the formation of secondary structure. The template nucleotide immediately adjacent to the primer terminus is A. We measured and compared the rates of incorporation of the correctly complementary T with the mismatched G, C, or A. Four separate reactions were performed containing only one of the four deoxynucleoside triphosphates. A large excess of substrate to enzyme was used to ensure that extension from the primer resulted from only one interaction of RT with the primer terminus. The primer was labeled at the 5Ј-end, and samples were fractionated by denaturing 20% polyacrylamide gel electrophoresis. Results were quantified by Phosphor-Imager analysis.
The rate of nucleotide insertion was determined as a function of the nucleoside triphosphate concentration, and this rate obeys Michaelis-Menten kinetic equations (see "Experimental Procedures"). To determine the rate of polymerization (V pol ), the intensity of the extension products (I e ) and the primer (I u ) were quantitated and calculated by the equation V pol ϭ I e /t⅐I u , where t is the reaction time. A Lineweaver-Burk analysis was plotted as 1/V pol versus 1/[dNTP] values to determine the K m and V max . The insertion efficiency was defined as (V max /K m ). The frequency of insertion (f ins ) of nucleotides was described as the ratio of incorrect to correct insertion efficiencies by the equation f ins ϭ (V max /K m ) W /(V max /K m ) R . Table I shows the values calculated for wild type, F227A, and W229A RTs. The characteristics of the F227A mutant followed the same trends observed in the running start experiment. It had higher fidelity for the insertion of a G or C across from a template A when compared with wild type. F227A produced a dC:dA and a dG:dA misinsertion at frequencies of 1.0 ϫ 10 Ϫ3 and 2.2 ϫ 10 Ϫ3 , and wild type RT produced these misinsertions at frequencies of 3.3 ϫ 10 Ϫ3 and 3.1 ϫ 10 Ϫ3 . This means that F227A misinserts these nucleotides at 30 and 70% of the error rate of the wild type RT. On the other hand, the F227A mutant had a frequency of 3.9 ϫ 10 Ϫ3 for dA:dA misinsertions, and wild type had a corresponding frequency of 7.1 ϫ 10 Ϫ4 . This indicates that F227A produces a dA:dA misinsertion 500% more often than wild type. These measurements are consistent with the results of the running start assay, which showed that the F227A mutation altered fidelity differently for the formation of particular mispairs.
W229A produced dA:dA misinsertions at a frequency of 4.8 ϫ 10 Ϫ4 , which corresponds to 67% of the error rate of the wild type RT (Table I). This mutant did not misinsert G or C across from a template A within the detectable range of the Phospho-rImager even at high concentrations of the incorrect deoxynucleoside triphosphate. Although this is indicative of high fidelity synthesis, we could not use the undetectable values to calculate the kinetic parameters for these misinsertions. To determine whether W229A is generally a high fidelity DNA polymerase, we also examined it for misinsertion of a T, C, or G across from a template T (Table II). W229A RT produced dT:dT, dT:dG, and dT:dC mispairs at frequencies of 1.5 ϫ 10 Ϫ3 , 1.4 ϫ 10 Ϫ3 , and 1.6 ϫ 10 Ϫ3 , respectively, which corresponds to 32, 36, and 38% of the error rate of the wild type RT. While every possible mispair remains to be tested, the results suggest that the W229A mutation has generally higher misinsertion fidelity.
Primer Extension Assay with a DNA Template and Biased dNTP Pools-To determine whether mutants F227A and W229A display generally higher fidelity than wild type RT, we also performed an assay that examines both misinsertion and mismatch extension. In this assay, only three of the four dNTPs were added to the reaction, causing RT to either terminate synthesis or misinsert a wrong nucleotide at specific sites on the template. Fig. 6 shows synthesis by the mutants compared with the wild type enzyme on substrate H. Overall, this assay demonstrates that the two mutants replicate with higher fidelity and terminate synthesis earlier than the wild type enzyme. Fig. 6A shows the results of the reaction with all four dNTPs. All enzymes extended the primer to a full-length product. The mutants only show differences in synthesis levels and a pause site when compared with the wild type. F227A had slightly higher levels of synthesis, and W229A had lower levels of full-length product. Both mutants paused at a site 29 nt from the 3Ј-end of the template. Since the wild type RT did not pause at this site, the structure of this site specifically affected the mutants. Fig. 6B shows synthesis performed in the absence of dGTP. Wild type RT and W229A produced a full-length product to a lesser extent than in the presence of all dNTPs. F227A terminated at a 32-nt product that corresponds to the fifth C site in the template. Furthermore, all of the enzymes paused at several sites corresponding to a misinsertion across from a C on the template. The first prevalent pause site is the first nucleotide added to the primer. This is the first site where the enzymes have to misinsert a wrong nucleotide. Wild type, F227A, and W229A RTs terminated 64, 65, and 98% of their synthesis products, respectively, at this site (Fig. 6B, lanes 5, 9,  and 13). All enzymes appeared to readily misinsert and extend a mismatch at the second C site in the template, since no pause site occurs at this position. The second major pause site occurs at the third C in the template. Wild type, F227A, and W229A RTs terminate 20, 26, and 4% of their synthesis at the 29th position on the template. The third important pause site creates a 32-nt product. F227A terminates synthesis at this site and does not produce a full-length product. Wild type and W229A also pause at this site, terminating 5 and 0.4% of synthesis products. Both wild type and W229A could generate a full-length product. Since F227A terminated synthesis at a 32-nt product and W229A terminates 98% of its total synthesis at the first C in the template, these results show that both of these mutants display higher fidelity than the wild type enzyme. When synthesis was performed in the absence of dATP, all enzymes terminated synthesis at the first T in the template (Fig. 6C). These results do not reveal a difference in fidelity for the wild type and mutant enzymes. Fig. 6D shows the results for synthesis in the absence of dCTP. Wild type RT extended 1.4% of the primer 17 nt and stopped synthesis at a G in the template (Fig. 6D, lanes 1-4). F227A extended 1% of the primer to make a 15-nt-long product and terminated synthesis at a G (lanes 5-8). W229A is unable to misinsert a wrong nucleotide across from a G in the template and terminated synthesis after the addition of the first nucleotide to the primer (lanes 9 -12). These results show that both mutants are higher fidelity and terminate synthesis at earlier sites on the template compared the wild type RT.
Last, Fig. 6E shows the results for replication in the absence of TTP. In this assay, wild type RT shows low fidelity replication compared with the mutants. Wild type RT was able to produce a full-length product with several pause sites corresponding to positions where a misinsertion occurred. F227A did not produce a full-length product and terminated synthesis at the 56-nt position on the template. Also, F227A terminated 71% of synthesis at nt positions 23-28 because of several A residues in the template. W229A terminated synthesis after the addition of 9 nt to the primer. This site corresponds to the second A residue of the template. Again, this assay showed that the two mutants displayed higher fidelity than the wild type RT, because they did not generate a full-length product.

Effects of Other Amino Acid Substitutions at Position 229 on Mismatch Extension-Residue
Trp-229 is invariant among retroviruses and appears to be a pivotal residue involved in positioning of RT for several different activities. In the previous experiments, the bulky aromatic tryptophan residue 229 was substituted with a smaller aliphatic alanine. In this experiment, we examined mismatch extension of residue Trp-229 replaced with other bulky aromatic residues (phenylalanine or tyrosine). Substrates A and B were used as in Fig. 2 to measure the ability of these mutants to extend a dG:dA mismatch. The W229Y RT was unable to extend this mismatch (Fig. 7, lane 2). Mutant W229F showed 76% less extension past the mismatch as compared with wild type (Fig. 7, lane 4). These results show that replacing residue 229 with a chemically similar amino acid  still produces an RT with higher mismatch extension fidelity than the wild type.

DISCUSSION
The region of HIV-RT known as the primer grip is highly conserved among retrovirus RTs (42). Crystallographic studies suggest that the primer grip amino acids are involved in the binding and proper positioning of RT to the 3Ј-end of the primer for DNA synthesis. Other studies show that making structural changes in regions of RT involved in primer and template positioning will alter fidelity of nucleotide addition (35,(43)(44)(45)(46)(47). In this study, we found that alanine substitutions for a number of the residues at positions 224 -235 of the p66 subunit do indeed influence replication fidelity. Interestingly, some substitutions increased fidelity. Specifically, the mutants F227A, W229A, M230A, G231A, and Y232A have 40, 66, 54, 72, and 76%, respectively, less synthesis past a dG:dA primer-template mismatched terminus than the wild type RT. These mutants displayed similar high fidelity for synthesis on an RNA template. One additional mutant, P226A, displayed similar mismatch extension on the DNA primer-template as the wild type, but this mutant did not extend the dG:A primer-template mismatch on the RNA template.
We also examined the effects of alanine substitutions of amino acids 224 -235 on the p51 subunit of HIV-RT. RTs were created in which the p66 subunit had the wild type amino acid, and the substitution was solely on p51. All of these p51 mutants had wild type levels of mismatch extension (data not shown). This was expected, since the primer grip region in the p51 subunit is not folded into the same structure as it is in the p66 subunit, and this region has not been shown to be involved in primer binding.
Overall synthesis error rate is controlled by the rate of misinsertion of nucleotides and the ability of those mismatched nucleotides to be extended. A mutant RT that resists both misinserting nucleotides and extending mismatched termini would be expected to replicate the virus with high fidelity. We examined the nucleotide misinsertion fidelity of two mutants F227A and W229A. Since these amino acids are in the nonnucleoside inhibitor-binding pocket but do not mutate in response to drug treatment, we anticipated that it is important to the virus that they remain unchanged. We measured nucleo-tide misinsertion rates using the fidelity measurement techniques of Goodman and colleagues (54 -57). Our results showed that the W229A mutant had overall high fidelity for the misinsertion of A, G, and C across from a template A and misinsertion of T, G, and C across from a template T. In fact, the W229A mutant would not add a G or C across from a template A on substrate G. We also confirmed that W229A had higher fidelity than wild type RT by performing primer extension assays in the presence of only three dNTPs. This assay showed W229A to misinsert a wrong nucleotide and to perform mismatch extension less frequently than the wild type enzyme.
Is Trp-229 maintained in the viral RT because a higher fidelity is a selective disadvantage? It has been proposed that high fidelity replication of retroviruses would prevent them from altering structure so as to evade immune response and antiviral drugs (50,52,53). Certain drug treatments select for drug-resistant mutations of HIV-RT that are coincidentally higher fidelity. Specifically, treatment of infected cells with 2Ј,3Ј-dideoxy-3-thiacytidine selects for drug-resistant RT mutations M184V and M184I (47)(48)(49)(50)(51)(52)(53). These mutations result in an increase in fidelity compared with wild type RT. Making the virus in an infected person into a higher fidelity form might also allow more effective subsequent drug treatments and improve immune sensitivity to the virus. The accuracy of this concept is controversial, awaiting a better understanding of the role of fidelity in viral therapy (16,(63)(64)(65)(66)(67)(68).
We cannot conclude that Trp-229 is conserved in the virus solely because altering it would improve fidelity. Trp-229 has previously been found to be important for maintaining the processivity and nucleotide incorporation rate of the RT (37,38,40). Mutations in this residue cause the RT to synthesize in a highly distributive manner, dissociating frequently from the primer terminus. They also have a reduced pre-steady-state rate of nucleotide addition. The high fidelity of mutant W229A could be a consequence of a lower affinity of the RT for the primer terminus, allowing it to dissociate rather than add an incorrect nucleotide or extend a mismatched nucleotide. Wohrl et al. (41) have reported that W229A has 27-fold lower binding affinity for a DNA/RNA primer-template than the wild type RT. Clearly, W229A might put the virus at a replication disadvantage for several reasons.
An indirect but compelling argument that high fidelity DNA synthesis is a disadvantage to the virus comes from the ease of generating mutants that are higher fidelity than the wild type RT. Five of the alanine substitutions in positions 224 -235 show characteristics of higher fidelity DNA synthesis. This would suggest that the RT is not constrained to its natural level of fidelity and could readily change to a higher fidelity structure. This assertion is also supported by the observations of others regarding drug-resistant mutants. M184V has higher fidelity for dT:dT and dC:dT misinsertions (53), and E89G has higher fidelity for dT:dT, dT:dG, and dG:dA (48). Since changes to a higher fidelity require selective pressure, these observations suggest that the natural fidelity of wild type RT provides the maximum potential for survival.
We have found that mutant F227A RT favors particular base substitutions. Specifically, F227A makes dC:dA and dG:dA misinsertions 3.2 and 1.4 times less frequently but makes dA:dA misinsertions 5.5 times more frequently than the wild type enzyme. Potentially, this mutation affects the positioning of RT to the 3Ј-end of the primer for nucleotide insertion. This alteration in the positioning of RT onto the substrate could cause the RT to favor certain misinsertions over others. Also, a change in positioning could influence the ability of RT to bind to mismatched primers leading to lower mispair extension. Another consideration is that residue Phe-227 could interact with the adjacent ␤10 loop that binds dNTPs. Changing the structure of this dNTP binding pocket could cause the RT to favor the binding of particular nucleotides versus others, improving the likelihood of specific base substitutions.
The decrease in dA:dG and dA:dC misincorporation by F227A may be significant for another reason. Nevirapine, a nonnucleoside inhibitor, creates drug-resistant mutations by changing residues 181 and 188 in RT (69 -71). This mutation is usually a Tyr to Cys that causes an A to G base substitution to occur. Mutations at position Phe-227 may not be favored during nevirapine treatment, because this residue reduces the frequency of the misinsertion needed to evade this drug. This reasoning further supports the concept that RT is designed by evolution to have a fidelity that is low enough to generate an advantageous number of mutations.
Although the F227A mutation increases the probability of a dA:dA misinsertion, the overall fidelity of this mutant is greater than that of the wild type RT as measured in a primer extension assay containing three dNTPs. This type of assay allows synthesis over a large region of template. During the course of synthesis, there are numerous opportunities for misinsertion and mismatch primer extension in different sequence contexts. This produces an estimate of the average error rate for polymerization. The generally higher fidelity of the F227A is an additional indicator of how readily mutations in the primer grip region can improve fidelity.
The primer grip region of HIV-RT is highly conserved in the virus, and mutations in the 224 -235 residues are not observed. Changes within the primer grip produce many defects in RNAprimed DNA synthesis and RNase H activities. We have shown here that this region is also important for maintaining replication fidelity at the level measured in the wild type RT. Amino acid substitutions in this region alter base substitution frequencies and mismatch extension efficiencies. Interestingly, some amino acid substitutions lower fidelity, while many raise it. Since the region is highly conserved, this observation suggests that the level of fidelity conferred by the natural amino acid sequence is advantageous for the survival of the virus. HIV-RT has a replication fidelity lower than that of avian myeloblastosis virus RT and murine leukemia virus RT (72)(73)(74)(75)(76), and this fidelity is much lower than cellular polymerases such as Escherichia coli DNA polymerase I and mammalian DNA polymerases ␣, ␦, and ⑀ (9, 62, 77-88). This higher error frequency allows the RT to introduce frequent changes into the genome. The error frequency of the RT is thought to be a major determinant of the diversity of sequence generated during growth of HIV. Our results suggest that naturally occurring amino acid substitutions in regions like the primer grip could readily alter the fidelity of the RT positively or negatively. Presumably, the wild type fidelity represents a desirable balance between survival advantages of diversity and lowered viral titers. The importance of fidelity in viral evolution suggests that therapeutic approaches aimed at altering fidelity deserve continued consideration.