Nucleotide Excision Repair and Template-independent Addition by HIV-1 Reverse Transcriptase in the Presence of Nucleocapsid Protein*

During HIV replication, reverse transcriptase (RT), assisted by the nucleocapsid protein (NC), converts the genomic RNA into proviral DNA. This process appears to be the major source of genetic variability, as RT can misincorporate nucleotides during minus and plus strand DNA synthesis. To investigate nucleotide addition or substitution by RT, we set up in vitro models containing HIV-1 RNA, cDNA, NC, and various RTs. We used the wild type RT and azidothymidine- and didanosine-resistant RTs, because they represent the major forms of resistant RTs selected in patients undergoing therapies. Results show that all RTs can add nucleotides in a non-template fashion at the cDNA 3′-end, a reaction stimulated by NC. Nucleotide substitutions were examined using in vitro systems where 3′-mutated cDNAs were extended by RT on an HIV-1 RNA template. With NC, RT extension of the mutated cDNAs was efficient, and surprisingly, mutations were frequently corrected. These results suggest for the first time that RT has excision-repair activity that is triggered by NC. Chaperoning of RT by NC might be explained by the fact that NC stabilizes an RT-DNA binary complex. In conclusion, RT-NC interactions appear to play critical roles in HIV-1 variability.

Human immunodeficiency virus type I (HIV-1) 2 replication is characterized by a high level of genetic variability, which leads to the appearance of quasispecies within a single individual (1). The emergence of quasispecies poses serious issues for AIDS treatment and vaccine development. Indeed, during most antiretroviral therapies, drug-resistant variants and viruses escaping the immune surveillance quickly appear (2,3). Understanding the molecular determinants of HIV-1 variability is of critical importance for the design of new highly active antiretroviral drugs.
Several mechanisms contribute to HIV-1 hypermutability. Reverse transcription, which comprises the early stage of the viral replication cycle, is the major source of retrovirus variability. Reverse transcriptase (RT) catalyzes the conversion of the single-stranded genomic RNA into a double-stranded viral DNA that is subsequently imported into the nucleus and integrated into the host cell genome by integrase. RT is error-prone and has the ability to misincorporate nucleotides during polymerization and to extend the mispaired nucleotides (4,5). This is thought to generate one or several mutations per round of virus replication. In addition to the obligatory first and second DNA strand transfers, RT mediates random strand transfers that generate recombinant viruses (6). These transfer events can be triggered by pauses during reverse transcription because of secondary structures in the viral RNA, mispaired nucleotides, or nicks (7)(8)(9). Previous results indicate that 30% of the transfers are coupled with point mutations at the site of transfer (10). Moreover, RT is able to add one or more non-template nucleotides on the cDNA 3Ј-end while copying RNA ends (11)(12)(13). As the genomic RNA is not fully protected against cellular nuclease degradation, RT is thought to encounter multiple ends during reverse transcription (5,14). cDNAs bearing the extra nucleotides are likely transferred onto the second RNA molecule to allow polymerization rescue. Data show that these transfers often result in substitutions (9). During polymerization, RT interacts with and is assisted by the viral nucleocapsid protein NC (5,10,15). NC is a well conserved small basic protein with two zinc finger motifs flanked by regions rich in basic residues. NC molecules coat the dimeric genome, forming the nucleocapsid structure within the virion, and are implicated in the early phases (proviral DNA synthesis) and late phases (genomic RNA dimerization and packaging) of the viral replication cycle. NC promotes reverse transcription by directing the annealing of primer tRNA Lys-3 to the primer binding site (15). In addition, NC stimulates minus and plus strand DNA transfers by promoting hybridization of the complementary repeat (R) sequences from the cDNA and the RNA template and by annealing of the tRNA and primer binding site sequences, respectively (10, 16 -18). NC also promotes specific proviral DNA synthesis by inhibiting self-primed reverse transcription (19 -21). Recently, NC was found to play a role in the maintenance of the newly made HIV-1 DNA by protecting the long terminal repeat sequences from nuclease attack and was also found to assist integrase (22)(23)(24).
Because NC plays key roles in the early steps of reverse transcription and in the DNA strand transfers, we wanted to investigate its possible function in viral variability. To investigate the influence of NC on nucleotide addition or substitution by RT, we established in vitro model systems containing HIV-1 RNA and cDNA and NC and RT proteins. In addition to the wild type (wt) RT, we used AZT-and ddI-resistant RTs, because they represent the major forms of resistant RTs selected in patients under highly active antiretroviral therapy. Our results show that all RTs can add nucleotides in a non-templated fashion at the cDNA 3Ј-end. Purines were preferentially added, and this was independent of the ultimate residue. NC was found to stimulate this nontemplate addition of nucleotides by RTs. Nucleotide substitutions by RT were examined using the same in vitro systems, where 3Ј-mutated cDNAs were extended. In the presence of NC, RT extension of the mutated cDNAs was efficient, and more importantly in most cases, mutations were corrected.
To further investigate how the presence of NC affects the interaction between RT and its substrate and thus its activity, we examined the influence of NC on the dissociation rate of RT on a duplex DNA. Our data indicate that NC significantly stabilizes the binary complex between HIV-1 RT and duplex DNA and therefore should be viewed as a critical RT co-factor.

EXPERIMENTAL PROCEDURES
RNA and DNA-RNAs were either HIV-1 RNA-generated in vitro by T7 phage RNA polymerase (10) or synthetic oligoribonucleotides. DNA and RNA oligonucleotides were supplied by Invitrogen or Integrated DNA Technologies. All of the other reagents were purchased from Sigma and were of the highest purity available. Radiolabeled oligonucleotides were prepared using phage T4 polynucleotide kinase (Invitrogen) and [␥-32 P]ATP. DNAs were purified by 8% PAGE in 7 M urea. Sequences of nucleic acids used in this study are presented in Table 1.
Template-independent Addition of nt, Nucleotide Excision, and Primer Extension Assays-Once nucleoprotein complexes were formed, the reaction volume was increased to 25 l by adding HIV-1 RT at 0.25 M, one dNTP (Invitrogen) at 0.25 mM, or an equimolar mixture or AZTTP (PerkinElmer Life Sciences) at 0.1 mM, 30 mM NaCl, and 3 mM MgCl 2 . Incubation was for 30 min at 37°C, after which the reaction was stopped by adding 0.5% SDS and 5 mM EDTA. Nucleic acids were purified by phenol-chloroform extraction and then ethanol-precipitated. Pellets were dissolved in formamide, denatured at 95°C for 1 min, and analyzed by 12% PAGE in 7 M urea.
Determination of Dissociation Rate Constants by Single Nucleotide Incorporation-Single nucleotide extension was evaluated on a 27-nucleotide DNA template (5Ј-AGCTACTCGATATGGCA ATAA-GACTCC-3Ј) annealed to a 5Ј-end-labeled 23-nucleotide primer (5Ј-GGAGTCTTATTGCCATATCGAGT-3Ј). The reaction mixture contained 25 nM template-primer and 37 nM HIV-1 RT in 12.5 mM Tris-HCl, pH 7.8, 40 mM NaCl, 9 mM MgCl 2 , 5 mM dithiothreitol, and 0.01% (v/v) Triton X-100. When included in the mixture, the concentration was 1.2 M for NC-(1-71) and up to 2.4 M for NC-(1-55) rec . After preincubation for 3 min at 37°C, heparin was added to sequester dissociated RT and ensure single round nucleotide incorporation. For reactions lacking NC, the minimal heparin concentration (10 g/ml) that inhibits rebinding of RT to the primer-template was used. For reactions containing NC, the heparin concentration was increased to 2 mg/ml to account for a potential decrease in effective heparin concentration because of binding with the basic NC. Incubation was continued for times ranging from 9 to 360 s before dATP was added to a final concentration of 25 M to initiate single nucleotide incorporation. The sample was then placed on ice and the reaction stopped by mixing with an equal volume of 89 mM Tris borate (pH 8.3), 2 mM EDTA, and 95% (v/v) formamide containing 0.1% (w/v) bromphenol blue and xylene cyanol. Reaction products were resolved by high voltage-denaturing 10% PAGE and visualized by phosphorimaging. The fraction of enzyme bound to the template was plotted as a function of time and the dissociation rate constant k off estimated by fitting the data with KaleidaGraph to the single exponential function y ϭ A⅐exp(Ϫk off ⅐t), where A is amplitude, k off is dissociation rate, and t is incubation time with the trap. Note that initiating the primer extension with dATP at varying times after the addition of heparin reveals the fraction of RT still bound to the template at this time point.
To assess the concentration-dependent effect of NC on single nucleotide extension, the reaction was modified as follows: a mixture containing 25 nM template-primer and 25 nM HIV-1 RT was preincubated with varying concentrations of NC (0.1-1. DNA Sequencing-cDNAs synthesized by RT extension of wild type or mutated DNA primers with or without NC-(1-72) were purified by phenol-chloroform extraction and ethanol-precipitated. cDNAs were PCR-amplified using a high fidelity Taq polymerase (Promega) and specific primers bearing restriction sites TARamp(Ϫ) (5Ј-GGGATC-CGCTAGCCAGAGAGCTCCCGG-3Ј) and TARamp(ϩ) (5Ј-GGA-ATTCGTGCTTTAAGTTAGTACC-3Ј). Amplified DNAs were then digested with EcoRI and BamHI and cloned into pSP64 vector digested with the same enzymes. Ligations were performed using Invitrogen T4 DNA ligase transformed and amplified in an E. coli recA Ϫ strain.
Plasmids were sequenced and sequences analyzed using the Edit View software.

Influence of NC on Non-template Nucleotide Addition by RT-Non-
template addition of nucleotides by RT is thought to occur at the blunt end of the newly made duplex template. During HIV-1 reverse transcription, such additions may take place during the two obligatory strand transfers and also at non-obligatory strand transfers caused by nicks in the genomic RNA (11)(12)(13).
To study non-template nucleotide additions at the cDNA 3Ј-end by HIV-1 RT, we derived an in vitro system formed of an RNA template, a primer DNA corresponding to the newly synthesized cDNA, NC, HIV-1 RT, and dNTPs ( Fig. 1). Four types of template-primer were used with either one of the four possible cDNA 3Ј-ends. Sequences used were derived from the HIV-1 genome or from non-viral sequences ( Table 1). The addition of NC to RNA and 5Ј-end 32 P-labeled primer DNA caused the concomitant formation of nucleoprotein complexes (20,27) and hybridization of the primer DNA and template RNA. In the absence of NC, annealing was carried out by heating (see "Experimental Procedures"). Next, either one or a mixture of the four dNTPs and RT were added. Reactions were performed at 37°C for 30 min and [ 32 P]cDNAs resolved on a 12% denaturating polyacrylamide gel and visualized by autoradiography (Fig. 1).
Non-template nucleotide addition activity was assayed with wild type RT, NC-(1-72) and several different classes of template-primer complexes with each cDNA 3Ј-end as indicated in Fig. 2. Results show that RT, in the absence of NC, can efficiently add one nucleotide at the cDNA3Ј-end, preferentially A or G when the 3Ј-end is T, G, or C (Fig. 2, B-D) and not when it is A (Fig. 2A). These results are in good agreement with previous data from several groups (9,(11)(12)(13). In the presence of NC, RT was able to add A, G, or T at the cDNA 3ЈA ( Fig. 2A). Moreover, the addition of 4 -6 consecutive T or G bases was also observed on template-primer complexes 3 and 7 (Fig. 2, B and D). However, NC did not change the pattern of the RT-directed addition, which remained A,GϾTϾ ϾC. (Fig. 2). The non-template addition of nucleotides by RT requires a hybrid, because when DNA was not hybridized to the RNA template, no addition was observed (data not shown).
To further investigate the role of NC in the RT-directed non-template addition of nucleotides, we used NC-(1-55), because it is the ultimately processed form of HIV-1 NC and also the NC-(12-53) mutant corresponding to the central globular domain of the viral protein (see supplemental data). NC-(1-55) was found to be clearly less active than NC-(1-72) (Fig. 3, compare lanes 6 and 7), whereas the NC-(12-53) deletion mutant was very poorly active (Fig. 3, lane 8). These results suggest that the basic rich regions of NC are important for enhancing the non-template addition of nucleotides by HIV-1 RT.

TABLE 1
Nucleic acids used in this study P/T stands for primer/template. Underlined sequences indicate mutated nucleotide(s) at the primer 3Ј-end used in the assays.

FIGURE 1. Formation of HIV-1 nucleoprotein complexes in vitro to study template-independent addition and excision of nucleotides by RT.
HIV-1 blunt end nucleoprotein complexes were formed by incubating an RNA representing part of the HIV-1 genome and a 5Ј-end 32 P-labeled DNA corresponding to the viral cDNA without or with NC for 10 min at 37°C. HIV-1 RT and each dNTP alone or equimolar amounts of the four dNTPs or AZTTP were then added, and reactions were performed for 30 min at 37°C. cDNAs were fractionated on a 12% denaturing polyacrylamide gel and visualized by autoradiography.
A terminal addition pattern similar to wild type RT was observed for the mutant enzymes, namely A,GϾTϾC (data not shown). Moreover, stimulation of the non-template addition of mutant RTs by NC was dependent on the basic regions of NC (Fig. 4B).
Sequencing results summarized in Table 2 were unexpected, because they show that the original mutations were rarely found when cDNA synthesis was performed in the presence of NC-(1-72). For example, a single mutation was corrected Ͼ80% of the time. Similarly, high correction values were obtained with 3 and 5 consecutive mutations. It was interesting to note that a single mutation was retained ϳ14% of the time, whereas partially corrected sequences were found in 10% of the cDNAs sequenced (Tables 1 and 2, 3Јmut3). These results indicate that RT in the presence of NC-(1-72) can correct mismatches at the cDNA polymerization site in vitro, although not completely. In the absence of NC, RT was still capable of partially correcting a single mutation at the cDNA polymerization site, but this was decreased by at least 3-fold when three consecutive mutations were present ( Table 2, 3Јmut1 and 3Јmut3 without NC). In addition, the absence of NC during reverse transcription was associated with an increased number of mutations elsewhere in the 370-nucleotide-sequenced cDNAs ( Table 2). Taken together, these data show that HIV-1 RT has an excision-repair activity in vitro and that NC stimulates this new activity.
HIV-1 RT-mediated Nucleotide Excision-To understand the repair reaction carried out by RT during elongation of cDNAs originally containing 1-5 mutations at their 3Ј-end (Table 2), we reasoned that the RT enzyme could excise the mismatched residues at the polymerization site, thus allowing cDNA synthesis to occur. To investigate this possibility, we used the template-primer hybrid as before to form a nucleoprotein complex upon NC addition. Thereafter, we added RT and the nucleotide analogue AZT in its triphosphorylated form, AZTTP, to prevent elongation after a putative nucleotide excision (Fig. 1).

Nucleotide Addition and Excision Repair by HIV-1 RT
removal by RT, we again used NC-(1-55) and NC-(12-53) corresponding to the central zinc fingers (see "Experimental Procedures"). The NC-(12-53) deletion mutant, missing the basic regions, was unable to stimulate nucleotide excision (Fig. 5B, lane 8). These data indicate that the NC basic domains are required for an efficient nucleotide excision by RT. The NC-(1-55) variant also did not promote nucleotide removal by RT in the presence of AZTTP (Fig. 5B, lane 7). To examine whether AZTTP has a direct effect on the RT excision activity, we used other chain terminators, such as ddATP, ddTTP, ddGTP, and ddCTP, in the presence or absence of NC. However, no excision was observed under these conditions (data not shown). Because ATP was described as a catalyst of AZT removal by AZT-resistant RT (28), we analyzed its effect on nucleotide removal by HIV-1 RT. In the presence of NC, AZTTP, and ATP, nucleotide excision efficiency by wt RT was lower than in assays performed without ribose ATP (data not shown). This suggests that ATP negatively influences nucleotide excision by RT. In addition, when inorganic pyrophosphatase was added to a reaction with AZTTP, nucleotide excision was partially hampered, indicating that pyrophosphate is involved in nucleotide removal by HIV-1 RT (data not shown).
Influence of NC on RT-DNA Complex Stability-Taken together, the above results show that NC can stimulate non-template additions and excision-repair of nucleotides by RT at the cDNA 3Ј-end. One plausible mechanism underlying this stimulation could be that NC has a stabilizing effect on the RT-substrate complex, prolonging the time RT remains productively bound. As a more quantitative measurement of interactions between RT and its substrate and the influence of NC, dissociation rate constants were evaluated by single nucleotide incorporation. The presence of heparin in the reaction mixture ensured single RT binding events, such that the relative amount of primer extension product at a given time reflects the fraction of enzyme in complex with the substrate. Fig. 6 illustrates the formation of the primer extension product (p ϩ 1) in the absence (A) or presence (B) of NC-(1-71), when single nucleotide extension was initiated at varying time points after the addition of heparin. Fitting the data with the single exponential equation y ϭ A⅐exp(Ϫk off ⅐t), where A is amplitude and t is incubation time, yields the dissociation rate k off for RT on the substrate. For the reaction without NC, the dissociation rate k off was determined as 0.059 Ϯ 0.0014 s Ϫ1 , whereas in the presence of NC-(1-71), k off was 0.0036 Ϯ 0.00028 s Ϫ1 (Fig. 6C). This 10-fold decrease in dissociation rate constant suggests that NC significantly stabilizes the complex between HIV-1 RT and template-primer. In contrast, the ultimate processed form of NC, namely NC-(1-55), had no effect (data not shown).  . The data show that the additional presence of NC-(1-55) did not influence the effect of NC-(1-71) on the complex stability between RT and template-primer. For both sets of experiments, the apparent K d for NC was 270 Ϯ 44 nM. A previous study reported an apparent affinity constant of 600 Ϯ 70 nM for a 1:1 stoichiometry of NC binding to RT (29). This value resulted from an affinity test based on the immobilization of RT heterodimer on Sepharose beads, and this experimental approach might be less accurate than our direct kinetic analysis.

DISCUSSION
In the present study, we have used in vitro systems mimicking the in vivo HIV-1 nucleoprotein complex called the viral nucleocapsid in which reverse transcription occurs. We have revealed new roles for NC in the synthesis of HIV-1 proviral DNA. NC-(1-72) stimulates the template-independent nucleotide addition mediated by wild type and drugresistant RTs. Using four different blunt end template-primer hybrids, we observed variations in the NC stimulation levels, suggesting that NC influence on the non-template addition of nucleotides by RT depends on the local sequence or secondary structure of the blunt end hybrid. In addition, NC-(1-72) was found to stimulate multiple additions of dTMP and dGMP, which might be correlated with RT-NC interactions (30) (see also below) and the higher affinity of NC for T and G residues (31).
In another series of experiments, we observed that HIV-RT chaper- oned by NC-(1-72) was able to efficiently correct one, three, and five consecutive mismatches at the polymerization site in vitro (Table 2). These findings indicate, for the first time, that HIV-1 RT could have an excision-repair activity in vitro. It has been assumed for a long time that retroviral RTs lack a proofreading activity (32,33), because they can efficiently misincorporate nucleotides and subsequently extend mismatches (4, 34 -38). This lack of 3Ј-5Ј-exonuclease activity was proposed to account for the high error rate of retroviral RT. However, in the majority of mispair extension assays, cDNAs extended were not sequenced and thus the presence of the initial mutation was not confirmed. In the present system, data on several hundred sequences confirm extension of one (A:C) and three consecutive mismatches (A:C, G:G, C:U) ( Table 2). More interestingly, we observed that mispair correction occurred more frequently than mispair extension, especially in the presence of NC-(1-72). In agreement with this, HIV-1 RT has the ability to excise a nucleotide from the cDNA 3Ј-end using two different mechanisms, namely by pyrophosphorolysis corresponding to the reverse reaction of polymerization and by ATP-dependent hydrolysis. Also, RT was found to efficiently remove AZT from AZT-blocked DNA (39,40). Excision of AZT by HIV-1 WT and AZT-resistant RT has been proposed to explain resistance to AZT inhibition. Furthermore, RT was recently shown to selectively remove mispairs (41) and to excise an incorporated dUMP in the presence of uracil-DNA glycosylase type 2 (42). Excision of dGMP with the insertion of the correct dAMP opposite a T residue can occur with high efficiency as compared with removal of dCMP and dTMP opposite a T residue. In the in vitro HIV-1 system used here, four types of mispairs were either extended or corrected: A:G, G:G, C:U, and A:C (Table 2). This confirms and extends data showing the efficient extension of A:C mispair by wild type HIV-1 RT (4). Last, recent findings reporting a nucleolytic activity of telomerase (43) bring other evidences supporting a possible nucleolytic activity of HIV-1 RT.
In the presence of AZTTP, RT (assisted by NC-(1-72)) catalyzed excision of several nucleotides from a cDNA in a blunt end duplex, and this was AZTTP-dependent (Fig. 5). These results could be interpreted as follows. First, this RT activity appears to require NC-(1-72) and occurs in a nucleoprotein complex where multiple molecular interactions are taking place (notably NC-RT interactions in addition to RT and NC with the hybrid nucleic acid (reviewed in Ref. 31 and see below)). Second, the requirement for AZTTP probably indicates that RT initially incorporated AZT in a non-template fashion and subsequently removed nucleotides. This interpretation would be in agreement with previous data showing that WT and AZT-resistant RT were able to excise 1-3 nucleotides from a 3Ј-AZT-blocked primer (44). Nucleotide excision at the 3Ј-end of a blunt end cDNA-RNA complex can be mutagenic for HIV-1, but this will depend on whether the 3Ј-excised cDNA is transferred to the other viral RNA molecule and how polymerization resumes. Because DNA strand transfer is error-prone, this process coupled with nucleotide excision can increase genome variability. The results presented here suggest that antiretroviral treatments probably increase HIV-1 genetic variability as previously suggested (45,46).
The excision and addition of nucleotides, coupled with strand transfers, appear to be major mediators in remodeling viral sequences, thus fueling virus variability. However, these two mechanisms have to be regulated to ensure some level of virus viability. In view of the results presented here, viral sequences seem to act as a regulator of variability, because efficiency of the addition and excision by RT are not the same for all hybrid sequences tested. Hot spots of variability most probably exist along the HIV genome, as already shown for viral recombination processes (47,48). In addition to the nucleotide context, NC appears to be an important regulator of variability. Indeed, we observed differences between the two NC variants -(1-72) and -(1-55) with respect to the levels of nucleotide addition and excision. In fact, the two NC variants act differently on the RT-template complex, because NC-(1-71) stabilizes the binary complex, whereas NC-(1-55) exerts no effect (Fig. 7). These results are consistent with previous reports indicating distinct properties for the two NC variants in their interaction with RT. Cameron et al. (49) have shown that NC-(1-71) efficiently suppresses the defects of strand transfer and RNase H activity exhibited by an RT deletion mutant, whereas NC-(1-55) displays no effect (49). The authors suggested a specific interaction between the carboxyl-terminal amino acids of NC and HIV-1 RT. Similarly, Lener et al. (30) have reported that NC-(1-72) functions more efficiently than NC-(1-55) in forming replicative nucleoprotein complexes in vitro (15). Thus, one can imagine that these two NC variants, which coexist in the viral particle, act differently in the course of cDNA synthesis by RT (30) and in proviral DNA integration (22,50).
Taken together, our findings highlight the dual role played by both NC and RT in viral replication. On one hand, RT and NC ensure virus viability by allowing a specific conversion of the RNA genome into a complete double-stranded DNA flanked by long terminal repeats in a process that necessitates two obligatory strand transfers. In addition and when required, forced interstrand transfers can occur at nicks in the RNA genome to allow continuation of cDNA synthesis by the replicative complex. On the other hand, RT chaperoned by NC appears to induce a balanced variability in the newly synthesized cDNA by virtue of non-template nucleotide addition or nucleotide excision at mispaired sites. This balanced mechanism of virus viability and variability would thus ensure both efficient virus replication and enough diversity to escape drug treatments and immune responses.