INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

One of the obstacles to successful antiretroviral treatment of human immunodeficiency virus (HIV) infections is the eventual emergence of drug-resistant viruses. A primary contributor to this process is the inherently high genetic variability of HIV, commonly encountered in virus populations in the infected individuals (9, 19, 25). This high level of variability is directly related to the viral mutation rate which, in turn, is thought to be significantly influenced by the polymerase fidelity of HIV reverse transcriptase (RT).

Mutants of HIV type 1 (HIV-1) RT resistant to nucleoside analog drugs that coincidentally display higher in vitro polymerase fidelity in gel-based nucleotide insertion and primer extension assays have been identified (6, 11, 20, 21, 24, 31). These include the multi-dideoxynucleoside triphosphate-resistant E89G and the 3TC-resistant M184V RT variants. The M184V mutation arises in patients in response to antiviral therapy with ()2',3' dideoxy-3'-thiacytidine (3TC; lamivudine) (27) and confers up to a 1,000-fold increase in viral resistance to 3TC (7, 26, 30) as well as a low-level cross-resistance to ddC and ddI (8). The E89G mutation, initially isolated via a phenotypic bacterial screening assay (22), confers a broad cross-resistance to deoxynucleotide (dNTP) analogs including ddATP, ddCTP, ddGTP, ddTTP, 3TCTP, and phosphonoformic acid (foscarnet; foscavir). This mutation has also been observed in in vitro-selected, 3TC-resistant variants of HIV-1 (7). The influence of these mutations on drug sensitivity is thought to be due to changes in the geometry of the dNTP-binding pocket brought about by direct interaction of the altered residue with the incoming dNTP in the case of M184V (12, 21, 29) or via template repositioning in the case of E89G (4, 29).

Previous observations showing that E89G and M184V RTs display increased fidelity of dNTP insertion and mispair extension suggested that these enzymes may be less prone to polymerase errors that lead to base substitution mutations. However, polymerase errors that contribute to mutation rates are not limited to uncorrected polymerase misinsertions. Both base substitutions and frameshift errors can be generated through mechanisms not involving direct misinsertion but rather involving template-primer slippage or dislocation (16, 28). In fact, a large proportion of errors by HIV-1 RT have been shown to be of this type (2). Thus, fidelities of dNTP insertion and mispair extension cannot solely be used to determine an overall polymerase error rate.

In this communication, we report the overall polymerase error rates of the RT mutants E89G and M184V measured via an M13-based forward mutation assay (2, 3). The double mutant E89G/M184V was also included in the study to examine if the effects of the two mutations on the overall error rate would be cumulative. Our findings indicate that the overall levels of polymerase fidelity of wild-type, E89G, M184V, and E89G/M184V HIV-1 RTs are similar (less than twofold differences). However, the mutational spectra of the variant enzymes revealed altered error specificities. It is unclear what impact such changes will have on viral variation, and this remains to be experimentally examined.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Phage DNA and bacterial strains. Bacteriophage M13mp2, a strain that carries the lacZ gene of Escherichia coli, was used to prepare gapped duplex DNA substrate. E. coli NR9099 [(pro-lac) thi ara recA56/F' (proAB lacI^q ZM15)] served as the bacterial host for the preparation of both single-stranded and replicative-form M13 DNA. Electrocompetent E. coli MC1061 [hsdR hsdM⁺ araD (ara leu) (lacIPOZY) galU galK strA] cells were used to produce phage from the products of the fill-in reactions. E. coli CSH50 [(pro-lac) thi ara strA/F' (proAB lacI^q ZM15 traD36)] was the -complementation strain used to score mutant phage.

Enzymes. Recombinant wild-type heterodimeric RT derived from HIV-1_Hxb2 and its variants with E89G, M184V, and E89G/M184V substitutions were bacterially expressed from derivatives of plasmid pL6H-PROT (14). The construction of these derivatives is described elsewhere (10). The recombinant enzymes were purified by Ni²⁺-nitrilotriacetic acid-hexahistidine chromatography followed by DEAE- and S-Sepharose chromatography as described earlier (15). The purified wild-type, E89G, M184V, and E89G/M184V RTs had specific activities of 220, 680, 280, and 89 U/mg, respectively, and were found to be nuclease free (data not shown). One unit is defined as the amount of enzyme required to incorporate 1 nmol of dTMP into DNA on a poly(rA)-oligo(dT) template-primer at 37°C in 10 min.

Forward mutation assay. Gapped duplex M13mp2 DNA was prepared as described previously (3) as the template for DNA fill-in synthesis reactions. This template contained a single-stranded region of 361 nucleotides including the upstream regulatory sequences and the first 107 coding nucleotides of the lacZ gene. Gap-filling synthesis reactions were performed by combining purified RT (0.07 to 0.45 U) with gapped duplex DNA (75 ng) in a reaction buffer containing 75 mM Tris-Cl (pH 8.0), 80 mM KCl, 6 mM MgCl₂, 10 mM dithiothreitol, and 500 µM each dATP, dCTP, dGTP, and dTTP (Boehringer Mannheim, Indianapolis, Ind.) in a total volume of 25 µl, and incubation for 1 h at 37°C. Complete gap closure was confirmed via agarose gel electrophoresis.

Mutational specificity. Single-stranded phage DNA was prepared as the template for sequence analysis as previously described (3). DNA sequencing reactions were performed with a Sequenase 2.0 DNA sequencing kit (Amersham Life Sciences, Arlington Heights, Ill.) with the oligonucleotide 5'GCGCAGCTGTTGGGAAGGGCG3' as primer. Sequencing reaction products were resolved on 6% denaturing polyacrylamide gels by using a GENOMYX_LR DNA sequencer.

Calculating error frequencies and error rates. Error frequencies were calculated as the ratio of confirmed mutant plaques to the total plaques screened. Spontaneous background mutation frequency was determined by electroporating uncopied gapped DNA and scoring for mutants as with DNA synthesized by RT. Corrected frequencies were then obtained by subtracting the spontaneous background from the frequency observed for a given RT (3).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Overall mutation frequencies. The fidelity of DNA synthesis by wild-type, E89G, M184V, and E89G/M184V HIV-1 RTs was measured in an M13 phage gap-filling assay. Copying errors resulting in altered function of the phage lacZ reporter gene were phenotypically scored. Gap-filling DNA synthesis reactions (two independent syntheses per enzyme) were performed with purified wild-type, E89G, M184V, and E89G/M184V HIV-1 RTs, and completion of the reaction was confirmed via agarose gel electrophoresis (data not shown). M13 DNA was electroporated into host E. coli, and plaques were generated by plating the transformants on a bacterial indicator lawn. Mutation frequencies for each enzyme were compiled from ratios of total mutant to total wild-type plaques obtained from one to three electroporations per gap-filling reaction per enzyme. The observed mutation frequencies for the wild-type, E89G, M184V, and E89G/M184V HIV-1 RTs were 8.60 × 10³, 6.26 × 10³, 5.53 × 10³, and 12.30 × 10³, respectively (Table 1). These represent background-corrected values, adjusted by using the spontaneous mutation frequency of uncopied gapped DNA of 1.5 × 10³. These results indicated that in terms of overall polymerase fidelity, the wild-type, E89G, M184V, and E89G/M184V HIV-1 RTs exhibit similar levels of fidelity (less than twofold differences). Of these, the E89G and M184V RTs were slightly less error prone than wild-type RT, while the E89G/M184V double mutant was slightly more error prone than wild-type RT.

Analysis of mutation spectra of the variant RTs. DNA sequence analysis of the phage mutants generated by the individual RTs revealed significant differences in the distribution and specificity of the mutations induced (Fig. 1). Mutations induced by wild-type, E89G, and M184V RTs were fairly well distributed throughout the target region, with some clustering at hotspots (sites with at least five mutations), while those produced by E89G/M184V RT were almost exclusively clustered at hotspots (see below). Close inspection of hotspot distribution shows that certain hotspots are shared by all RTs (Table 2). For example, the base substitution hotspot at position 36 is well represented in the spectra of all of the RTs studied here, as is the frameshift hotspot at positions 137 to 139 (Fig. 1; Table 2). In contrast, the base substitution hotspot observed in the mutation spectrum of wild-type RT at position 112 and the deletion hotspot at positions 106 to 108 (Fig. 1) are conspicuously absent from the spectra of the mutant RTs. In fact, with the mutant RTs, these sites are nearly devoid of mutations (Fig. 1). Similarly, the frameshift hotspot found at positions 34 to 36 in the spectra of the E89G and E89G/M184V RTs is completely absent in the spectra of wild-type and M184V RTs (P < 0.0001 for both E89G [versus wild-type and M184V RTs] and E89G/M184V [versus wild-type and M184V RTs]) (Fig. 1). Furthermore, a greater percentage of total mutations induced by E89G/M184V (83.8%) are within hotspots compared to mutations induced by the other RTs (wild type, 53.6% [P < 0.0001]; E89G, 60.2% [P = 0.0002]; M184V, 60.6% [P = 0.0003]). For the E89G/M184V variant, four sites (positions 36, 34 to 36, 70 to 73, and 137 to 139) account for almost 84% (83/99) of all mutations, with the 34 to 36 positions alone accounting for 68% (67/99) of total mutations observed (Fig. 1c; Table 2).

View larger version (105K): [in this window] [in a new window]   View larger version (110K): [in this window] [in a new window]   View larger version (103K): [in this window] [in a new window]   FIG. 1.   (a) Spectrum of mutations induced by wild-type HIV-1 RT and its E89G variant. Comparison of mutations generated in the lacZ target by wild-type HIV-1Hxb2 and E89G HIV-1 RTs is shown. Mutations induced by wild-type HIV-1Hxb2 RT are represented above the template sequence (white), and mutations induced by E89G HIV-1 RT are shown below the template sequence. Substitutions are indicated by the letter corresponding to the new base (in yellow) above or below the base in the template sequence it is replacing. Nucleotide deletions are indicated by an upright (red) triangle over or under the corresponding run of template bases; nucleotide additions are indicated by an inverted (red) triangle over or under the corresponding run of template bases. Triangles with numerals (2 or 3) underneath them indicate deletions involving more than one base (the bases deleted or are thought to be involved in the deletion are underlined in red). A two-base deletion with the asterisk indicates where it was not possible to determine whether bases 140 and 141 or 141 and 142 were deleted. Nucleotide positions are indicated below the template sequence. (b) Spectrum of mutations induced by wild-type and the M184V mutant HIV-1 RTs. Comparison of mutations generated in the lacZ target by wild-type HIV-1Hxb2 and M184V HIV-1 RTs is shown. Mutations induced by wild-type HIV-1Hxb2 RT are represented above the template sequence (white), and mutations induced by M184V HIV-1 RT are shown below the template sequence. (c) Spectrum of mutations induced by wild-type and E89G/M184V mutant RTs. Comparison of mutations generated in the lacZ target by wild-type HIV-1Hxb2 and E89G/M184V HIV-1 RTs is shown. Mutations induced by wild-type HIV-1Hxb2 RT are represented above the template sequence (white), and mutations induced by E89G/M184V HIV-1 RT are shown below the template sequence.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Viral mutation rates are the products of the combined effects of overall polymerase (copying) error rates, replication rates, and biological selection (5). Presumably, increases or decreases in any of these factors could result in corresponding changes in mutation rate. This notion has previously not been tested for any variant of HIV-1 RT. The variants of HIV-1 RT that were recently reported to possess increased fidelities of dNTP insertion and primer extension in in vitro assays (6, 11, 20, 21, 24, 31) appeared suitable to examining the contributions of polymerase fidelity to overall mutation rate in vivo. Thus, we have determined the overall copying fidelity of the E89G, M184V, and E89G/M184V RT variants. We had earlier reported increases of three- to sixfold in average nucleotide insertion fidelity for the M184V and E89G RTs (6, 31). When changes in the fidelity of formation of specific mispairs are considered, the increases of 2- to 45-fold (6, 21, 24, 31) have been seen. With respect to terminal mismatch extension efficiency, increases in fidelity of primer extension, ranging from 3- to 66-fold greater (for extension from specific mispairs) than for wild-type RT, have been reported for E89G and M184V RTs (11, 20, 24). In contrast to these large increases in fidelities of dNTP insertion and primer extension previously observed for specific types of errors, the M184V and E89G RTs in the current study showed only slightly decreased overall polymerase error rates whereas the E89G/M184V variant rate was slightly increased (less than twofold differences when compared to the wild type) (Table 1). This discrepancy could be a result of the influence of certain features of the M13 forward mutation assay, such as host repair and the ability to detect slippage-mediated errors in addition to nucleotide misinsertion and mispair extension. In fact, a substantial proportion of the total mutations observed occurred within the homonucleotide runs in the template sequence which are primary sites for slippage-mediated mutagenesis. Thus, in light of the modest differences observed in overall error rates, it appears that any increases in fidelity of these mutants that may be afforded by the increased dNTP insertion/primer extension fidelity are muted by the contributions of other mutagenic mechanisms.

The lack of increase in fidelity for M184V RT seen in this study would appear to be concordant with the biological data regarding viral evolution of M184V mutant viruses. Drug-resistant variants of such viruses were found to emerge in cell culture at rates equal to those of wild-type viruses (1, 13), suggesting that mutation rates were unaffected by the RT mutation. It is likely that compensatory factors similar to those operating in the M13 forward assay (as described above) were modulating the effect of the viral 184V substitution.

Although both E89G and M184V mutations increase nucleotide insertion/primer extension fidelity, it seems likely that they exert their effects via distinct mechanisms. The residues affected by these mutations lie within the enzyme active site in or near the dNTP-binding pocket (12, 29). Structural data place these residues at opposing sides of the active site, with the E89 residue contacting the phosphate-deoxyribose backbone of the template strand and the M184 residue interacting with the sugar moiety of the primer terminus (12, 29). Consequently, as mutations at these residues alter different surfaces of the dNTP-binding pocket, it might be anticipated that the effects of these modifications on the dNTP-binding pocket conformation would differ. Such differences are reflected in the divergence seen in the distribution and specificity of mutations induced by the variants. For instance, the E89G mutation had an enhancing effect on frameshift mutagenesis (both run and non-run associated). In contrast, the M184V mutation had a minimal effect on this class of errors (Table 3). It is generally thought that most frameshift mutations (insertions and deletions) involving few (one to three) nucleotides arise from transiently misaligned or slipped template-primer intermediates (28). The E89 residue is located within the 5a strand of the palm region of RT, where it is thought to help the polymerase grip the template (12). It is conceivable that the glycine substitution at this residue, by virtue of glycine's smaller side chain, results in a looser grip, which might enable the primed template to more easily form misaligned intermediates while still bound to the enzyme. Alternatively, a looser template grip stemming from the 89G mutation might allow the RT to more readily dissociate from the template, providing greater opportunity for the formation of misaligned template-primers. A determination of dissociation rates using purified RTs and model template-primers may help differentiate between the two possibilities. The M184 residue participates in dNTP binding (21, 32) rather than template-primer grip. Thus, it is possible that substitutions at this position affect utilization of frameshift intermediates rather than the enzyme's ability to promote their formation.

To date, the E89G substitution has been found only in viral isolates bearing the M184V mutation as well. Therefore, the E89G/M184V double mutant was studied in order to understand the cumulative effect of these two mutations on error rate. Curiously, when present independently, the individual mutations increased polymerase fidelity slightly while a combination of the two led to a slight decrease in the overall polymerase fidelity (Table 1). More unexpected was the highly clustered distribution of the mutations induced by the E89G/M184V RT, considering that both E89G and M184V RTs generated mutations that were fairly well distributed. Yet the influence of both RT mutations appears to be seen in the mutational specificity of the double mutant. Hotspots common to both single variants (positions 36 and 137 to 139) are reproduced by the E89G/M184V variant. The strong representation of the deletion hotspot at the 34 to 36 site seems to suggest some dominance of the 89G mutation; however, the lack of non-run-associated frameshifts for the double mutant appears to indicate a significant contribution by the 184V residue to frameshift specificity. Thus, the error specificities of each single mutation are preserved to a degree in the double mutant while the effects of each individual substitution on polymerase fidelity are somewhat modulated by the other. Since these two residues occupy opposite sides of the active site, it may be possible that steric constraints imposed by these substitutions are additive, thus further restricting the range of stable premutational intermediates. If this is so, it may offer an explanation for the limited mutation distribution exhibited by E89G/M184V RT.

Since the influence of these mutations on overall RT fidelity appears minimal, could differences in error specificity have a significant biological effect on viruses harboring these mutations? It is difficult to predict any specific effects of 184V alteration since the change observed in overall fidelity was small. In the case of E89G variant viruses, perhaps the enhancement in frameshift mutagenesis associated with the 89G substitution when expressed in vivo results in a higher proportion of virions with single-base insertions and deletions which are far more likely to be lethal than base substitutions. This may partially account for the failure of this mutation to be seen in a viral isolate independent of the 184V substitution. The ability of the M184V mutation to reduce the frameshift mutagenicity of E89G could permit a variant HIV with both mutations to replicate and to be recovered via in vitro selection protocols (7).

Finally, it is known that sequence context contributes substantially to the mutagenic potential of a given nucleotide site (2, 18, 23), and this likely results in the site bias that produces hotspots. The studies reported here were conducted in the context of the lacZ gene, which may not be representative of viral sequences. Therefore, these RT mutations could impart to replicating viral genomes overall fidelities higher or lower than those observed here. In this context, the double mutant E89G/M184V RT in particular needs to be tested for its effect on viral mutation rate since it appears to display few mutations outside the hotspots which are so characteristic of lacZ gene. It should be possible through single-cycle infection studies with HIVs containing these mutations to establish the in vivo significance of our findings.