Amino acid residues in HIV-2 reverse transcriptase that restrict the development of nucleoside analogue resistance through the excision pathway

Nucleoside reverse transcriptase (RT) inhibitors (NRTIs) are the backbone of current antiretroviral treatments. However, the emergence of viral resistance against NRTIs is a major threat to their therapeutic effectiveness. In HIV-1, NRTI resistance-associated mutations either reduce RT-mediated incorporation of NRTI triphosphates (discrimination mechanism) or confer an ATP-mediated nucleotide excision activity that removes the inhibitor from the 3′ terminus of DNA primers, enabling further primer elongation (excision mechanism). In HIV-2, resistance to zidovudine (3′-azido-3′-deoxythymidine (AZT)) and other NRTIs is conferred by mutations affecting nucleotide discrimination. Mutations of the excision pathway such as M41L, D67N, K70R, or S215Y (known as thymidine-analogue resistance mutations (TAMs)) are rare in the virus from HIV-2–infected individuals. Here, we demonstrate that mutant M41L/D67N/K70R/S215Y HIV-2 RT lacks ATP-dependent excision activity, and recombinant virus containing this RT remains susceptible to AZT inhibition. Mutant HIV-2 RTs were tested for their ability to unblock and extend DNA primers terminated with AZT and other NRTIs, when complexed with RNA or DNA templates. Our results show that Met73 and, to a lesser extent, Ile75 suppress excision activity when TAMs are present in the HIV-2 RT. Interestingly, recombinant HIV-2 carrying a mutant D67N/K70R/M73K RT showed 10-fold decreased AZT susceptibility and increased rescue efficiency on AZT- or tenofovir-terminated primers, as compared with the double-mutant D67N/K70R. Molecular dynamics simulations reveal that Met73influences β3–β4 hairpin loop conformation, whereas its substitution affects hydrogen bond interactions at position 70, required for NRTI excision. Our work highlights critical HIV-2 RT residues impeding the development of excision-mediated NRTI resistance.

Although the human immunodeficiency virus type 1 (HIV-1) is responsible for most of the global AIDS pandemic, it has been estimated that one to two million people are infected with HIV type 2 (HIV-2) worldwide (1). HIV-2 is endemic in West Africa, but it is also found in regions of Western Europe, India, Brazil, and South Africa. HIV-2 has a longer asymptomatic phase and shows slower progression to AIDS in comparison with HIV-1 (2). However, fewer treatment options are available for those infected with HIV-2. Non-nucleoside reverse transcriptase (RT) 3 inhibitors, the fusion inhibitor enfuvirtide and several protease inhibitors approved for treatment of HIV-1 infection, are ineffective against HIV-2 (3,4). In contrast, HIV-1 and HIV-2 show similar susceptibilities to nucleoside RT inhibitors (NRTIs) (5). Those drugs constitute the backbone of current therapies against infections caused by both types of HIV (6).
HIV RTs are heterodimeric enzymes that convert the singlestranded RNA genome into a double-stranded DNA that can be integrated in the host cell genome (7,8). RTs have two catalytic activities as follows: (i) a DNA polymerase activity that uses RNA or DNA as templates, and (ii) a ribonuclease H (RNase H) activity that degrades RNA from RNA/DNA hybrids. NRTIs are converted to triphosphate derivatives inside the cell and are incorporated into the DNA by the polymerase activity of the RT. Further DNA synthesis is then blocked due to the absence of a 3Ј-OH in the ribose moiety of the NRTI (9). Although HIV-1 and HIV-2 share some NRTI resistance mutations, clinical studies have shown that NRTI resistance emergence is faster in HIV-2.
K65R, Q151M (sometimes accompanied by V111I), and M184V are the major mutations conferring resistance to NRTIs in HIV-2 (18,19). The combination K65R/Q151M/M184V confers class-wide nucleoside analogue resistance (20). Amino acid substitutions in this complex affect residues of the dNTPbinding site in the DNA polymerase domain of the RT, and biochemical studies carried out with the HIV-1 enzyme showed that those changes interfered with the incorporation of nucleotide analogues (9). In contrast, TAMs are difficult to select in cell culture after HIV-2 passage in the presence of AZT (21) and are found with a very low prevalence in HIV-2 obtained from patients treated with AZT and other nucleoside analogues (22)(23)(24). HIV-1 and HIV-2 RTs have similar levels of DNA polymerase activity, but the HIV-2 enzyme incorporates AZT-triphosphate (AZTTP) less efficiently than HIV-1 RT and shows a much lower ability to excise the inhibitor in the presence of ATP (25). Moreover, substituting Tyr for Ser 215 in the HIV-2 RT did not increase its ATP-dependent phosphorolytic activity but had a detrimental effect on polymerase activity (26). In those studies, the authors speculated on the possibility that selection of TAMs would be detrimental for HIV-2 RT function, and this would justify why the excision mechanism has a negligible influence in HIV-2 resistance to NRTIs. Furthermore, the presence of Ile 75 in HIV-2 RT instead of Val as in HIV-1 RT could be responsible in part for the lack of the excision-based resistance mechanism, as suggested by the antagonistic effect of V75I when combined with TAMs in the context of an excision-proficient HIV-1 RT (27). The presence of Ile 75 is expected to facilitate the selection of Q151M, because this substitution together with A62V, V75I, F77L, and F116Y forms the "Q151M complex" that in HIV-1 confers resistance to all approved NRTIs except tenofovir (28,29).
In this work, we show that HIV-2 ROD RT with the four classical TAMs (M41L/D67N/K70R/S215Y) has negligible ATPdependent excision activity. Despite the high level of amino acid sequence similarity between HIV-1 and HIV-2 RTs, there are important differences in their primary structures that affect residues in the ␤3-␤4 hairpin loop (i.e. at positions 63-73), and around Tyr 215 , in addition to residues elsewhere that in HIV-1 have an antagonistic effect on TAMs. The systematic replacement of HIV-2 RT residues by those found in the HIV-1 RT and the subsequent analysis of the ATP-dependent excision activity of the resulting RTs, using different NRTIs and templateprimers, shows that Met 73 in HIV-2 RT (instead of Lys as found in HIV-1 RT) acts as a molecular barrier that curbs the possibility of developing HIV-2 resistance through the accumulation of TAMs.

Results
HIV-1 BH10 and HIV-2 ROD RTs share 61% amino acid sequence identity. The crystal structures of AZT-resistant HIV-1 RT containing TAMs M41L, D67N, K70R, T215Y, and K219Q in its unliganded form, in complex with AZT-terminated DNA/DNA template-primers, and with doublestranded DNA and AZT adenosine dinucleoside tetraphosphate (AZTppppA) have been reported (30). In those studies, Tu et al. (30) showed that TAMs at positions 67, 70, and 219 contribute to the proper orientation of the pyrophosphate moiety during ribonucleotide-dependent nucleotide excision, whereas Tyr 215 interacts with the adenine moiety in AZTp-pppA and defines the ATP-binding site. Based on the available crystal structures and previous work with mutant RTs having insertions and deletions in the ␤3-␤4 hairpin loop (31, 32), we hypothesized that the major determinants of the excision activity in AZT-resistant RTs locate within residues 60 -76 (␤3-␤4 hairpin loop and adjacent residues of ␤-strands 3 and 4), and around position 215. The amino acid sequences of HIV-1 and HIV-2 RTs show differences at those regions (Fig. 1a). In addition, HIV-1 and HIV-2 RTs differ at specific positions that were previously shown to modulate the excision reaction in the presence of TAMs. Thus, previous studies demonstrated that by themselves V75I, W88G, and Y181C decrease the ATP-dependent excision activity of HIV-1 RTs carrying TAMs (6). The wildtype (WT) HIV-2 ROD RT contains Ile 75 , Thr 88 , and Ile 181 instead of Val, Trp, and Tyr, as found in the HIV-1 RT. These differences could be also responsible for the suppression of the excision activity in HIV-2 RTs. Based on those observations, we obtained HIV-2 ROD RT variants whose amino acid sequences are given in Fig. 1b. They included enzymes where the sequence of the ␤3-␤4 hairpin loop of the HIV-1 BH10 RT was introduced in the HIV-2 ROD RT together with TAMs M41L and T215Y (e.g. 5M_STK), others that contained different substitutions in the 210 -215 sequence (e.g. 5M_LRWL), or combinations of mutations, including potential antagonists of TAMs such as I75V, T88W, or I181Y. The enzymes were purified and tested for their ability to remove thymidine analogues and tenofovir, in rescue assays carried out with RNA/DNA and DNA/DNA template-primers.

Identification of suppressors of the ATP-dependent unblocking activity of DNA primers terminated with AZT or tenofovir in the HIV-2 RT
The ability of HIV-2 ROD RTs to rescue AZTMP-terminated primers in the presence of ATP (3.2 mM) was first assessed with the RNA/DNA template-primer 31Trna/21P (Fig. 2). Reactions were carried out in two steps. The RT and the RNA/DNA

NRTI excision suppression in HIV-2 RT
hybrid were incubated in the presence of inhibitor (AZTTP) to generate a blocked 22-nucleotide primer. Then, the addition of dNTPs and ATP (the pyrophosphate donor) facilitated unblocking and further extension of the primer to obtain a product of 31 nucleotides. In these assays, the HIV-1 BH10 RT that contained TAMs M41L and T215Y (LY_HIV-1) showed relatively high levels of excision activity. In contrast, the WT HIV-2 ROD RT and the quadruple mutant M41L/D67N/K70R/ S215Y (4M RT) showed almost negligible activity in the presence of 3.2 mM ATP.

NRTI excision suppression in HIV-2 RT
Although Asp 67 and Lys 70 are conserved in HIV-1 and HIV-2 RTs, there are differences at four residues of the ␤3-␤4 hairpin loop region of both polymerases. By introducing the amino acids found at positions 68, 69, 73, and 75 in HIV-1 BH10 RT in the sequence context of the 4M HIV-2 RT, we obtained enzymes with high ATP-dependent excision activity (Fig. 2). The substitution of Lys for Met 73 in the HIV-2 ROD RT had the largest impact on rescue activity among all tested polymerases. Thus, mutants 5M_STK and 5M_SK showed efficiencies that were almost 2-fold higher than those obtained with the reference HIV-1 BH10 RT mutant M41L/T215Y, and well above the efficiencies of the WT HIV-2 ROD RT. Other RTs showed lower ATP-dependent excision activities. However, substituting Val 75 for Ile in the 4M RT produced a modest but significant increase in rescue efficiencies. The subsequent addition of K68S and/or N69T, several combinations of mutations in the vicinity of Tyr 215 , or substitutions with a potential antagonistic effect on TAMs (i.e. T88W and I181Y) had a minor effect on DNA primer rescue catalyzed by the 5M RT mutant enzyme.
Similar assays carried out with a DNA-DNA complex containing an AZTMP-terminated primer also revealed that the substitution of Lys for Met 73 in the HIV-2 ROD RT had a large impact in the rescue activity of the polymerase (Fig. S1). Mutant derivatives containing M73K (i.e. 5M_SK and 5M_STK) rescued between 30 and 35% of the AZTMP-terminated primers after incubating the template-primer and the RT for 20 min in the presence of ATP and dNTPs. In contrast, 5M, 5M_K68S, and 5M_ST RTs had lower excision activity (Ͻ14%) under the same conditions. Nevertheless, these three enzymes showed a modest increase in rescue activity in comparison with the 4M RT. This could be attributed to the substitution of Val for Ile 75 that could facilitate excision by neutralizing the small but significant antagonistic effect of Ile 75 . As in the case of experiments carried out with the RNA/DNA hybrid, T88W and I181Y had a minor effect on DNA primer rescue when combined with I75V and other TAMs, although we observed a slight increase with the substitution I181Y. In contrast, none of the mutations located in the vicinity of Tyr 215 seemed to affect the rescue activity of the 5M mutant, because all tested HIV-2 RT variants showed rescue activities below 14%, after a 20-min incubation in the presence of ATP and dNTPs.
In general, removal of AZTMP and further extension of the unblocked primer were more efficient with the RNA template than with the DNA template for all tested RTs. Analysis of the efficiencies of the rescue reactions with different RTs and the 31Trna-21P complex confirms the suppressor effect of Ile 75 , because mutant 5M shows higher rescue activity than 4M, although the addition of K68S and N69T produced a slight but non-significant increase in rescue efficiency. In any case, the effects of those substitutions were relatively small in comparison with the addition of M73K.
The ATP-dependent rescue activity was also tested with all mutants using tenofovir-terminated primers and the DNA/ DNA hybrid D38T/25PGA (Fig. S2). The observed ranking of rescue efficiencies was similar to those obtained with AZTterminated primers, supporting the notion that the inhibitor had a minor impact on the relative excision activity of the RTs.
The amounts of rescued primer obtained with the WT HIV-2 RT and its mutant derivative 4M were very small. Again, the presence of the additional substitution I75V produced a significant increase in tenofovir excision and further extension of the unblocked primer. This effect was reduced if other ␤3-␤4 hairpin substitutions such as K68S or N69T were added to the mutational complex. However, in those cases, addition of M73K rendered excision-proficient RTs with rescue efficiencies higher than those observed for the well-characterized HIV-1 BH10 RT mutant M41L/T215Y (LY_HIV-1).

ATP-dependent excision activities of WT and mutant HIV-2 RTs
Assays described above have consistently shown that M73K and to a lesser extent I75V facilitate the rescue of NRTI-terminated primers by TAM-containing HIV-2 RTs. However, those assays combine excision with DNA primer extension. In Fig. 3, we show that I75V confers some ATP-dependent excision activity on primers terminated with AZT when added to the 4M complex, whereas the subsequent addition of M73K (in combination with K68S) results in a further enhancement of the excision activity. Interestingly, the effects of I75V were not observed in ATP-dependent excision reactions carried out with primers terminated with 2Ј,3Ј-didehydro-2Ј,3Ј-dideoxythymidine monophosphate (d4TMP). Mutants 5M, 5M_K68S, and 4M and the WT HIV-2 enzyme showed negligible activity (k obs Ͻ 1.6 ϫ 10 Ϫ3 min Ϫ1 ). However, addition of the M73K substitution to the 5M_K68S complex rendered an RT with significant excision activity and showed a k obs of 5.8 ϫ 10 Ϫ3 min Ϫ1 . Taken together, the enzymatic assays reveal the strong effect of Met 73 in the context of HIV-2 ROD RTs bearing TAMs by suppressing NRTI excision.

M73K restores ATP-dependent rescue activity of HIV-2 RT bearing TAMs D67N and K70R
Unlike HIV-1 (11,32), the double mutant D67N/K70R HIV-2 RT has very low or negligible rescue activity on primers terminated with AZT either in RNA/DNA or in DNA/DNA template-primer complexes (Fig. 4, a and b). However, adding the M73K substitution produces a large increase in rescue efficiency that was observed both with AZT-terminated (Fig. 4, a  and b) and tenofovir-terminated primers (Fig. 4c). As expected, M73K alone had no effect on the unblocking and extension reactions and showed a phenotype similar to that obtained with the WT HIV-2 RT. The three RTs showed significant DNA polymerization activity in single-nucleotide incorporation assays, carried out with template-primer D38/25PGA, under steady-state conditions. In those reactions, the turnover rates (k cat ) for WT HIV-2 RT and mutants D67N/K70R and D67N/ K70R/M73K were 7.19 Ϯ 0.82, 3.87 Ϯ 0.43, and 3.42 Ϯ 0.19 min Ϫ1 , respectively.
The only available crystal structure of HIV-2 RT (Protein Data Bank (PDB) code 1MU2) has an overall folding similar to that of the HIV-1 RT (33) but has no ligands and is incomplete, lacking several key loops. Therefore, for modeling purposes, we used the crystal structure of a ternary complex of HIV-1 RT, double-stranded DNA and dTTP (PDB code 1RTD). Structural models of catalytically competent WT and mutant HIV-1 RTs were obtained after refinement by molecular dynamics. In

NRTI excision suppression in HIV-2 RT
HIV-1 RT, residue 73 is Lys instead of Met as occurs in the HIV-2 enzyme (Fig. 1). The structural models of mutant HIV-1 RTs D67N/K70R and D67N/K70R/K73M were compared with the one previously obtained for the catalytically competent WT HIV-1 RT (34).
As shown in Fig. S3A, root-mean-square deviations (r.m.s.d.) corresponding to the backbone C␣ atoms of the modeled ternary complexes remained below 2.4 Å. Superimpositions of the simulated structures were consistent with the results of the r.m.s.d. analysis, although significant differences were observed between the conformation of the ␤3-␤4 hairpin loop of D67N/ K70R/K73M and the equivalent structures of the WT HIV-1 RT and mutant D67N/K70R (Fig. S4A). Mutations did not affect critical interactions required to maintain the catalytic attack distance between the 3Ј-OH of the primer and the ␣-phosphorus (Fig. S3B). At the end of the simulation, interactions in the catalytic site of the ternary complex containing two Mg 2ϩ ions were similar for the three studied RTs (Fig. S4).
Molecular dynamics simulations showed that the position of Arg 70 was largely affected by substitutions at the ␤3-␤4 hairpin loop. Interestingly, in the double mutant D67N/K70R, distances between the -carbon of Arg 70 and the ␥-phosphorus of the dNTP remained stable and below 4.5 Å, and they were compatible with the establishment of hydrogen-bonding interactions between the amido groups of Arg 70 and the hydroxyl substituents of the dNTP ␥-phosphate (32). These interactions were not predicted for RTs devoid of excision activity, such as

Antiretroviral drug susceptibility of recombinant HIV-2
Susceptibility to AZT was determined in MT-4 cells for selected HIV-2 mutants to test whether the results obtained in rescue and excision assays correlated with phenotypic drug susceptibility of recombinant HIV-2. As shown in Table 1, the presence of the four classical TAMs (M41L/D67N/K70R/ S215Y) in HIV-2 did not produce an AZT-resistant virus (i.e. compare 4M versus WT HIV-2 ROD ). However, substituting Val 75 for Ile (5M mutant) decreased AZT susceptibility almost four times. Other substitutions in the ␤3-␤4 hairpin loop had no significant effect on thymidine analogue resistance, although we were unable to show the effects of M73K when multiple amino acid substitutions were introduced in the HIV-2 RT. The recombinant HIV-2 with RTs 5M_SK and 5M_STK was not viable and was unable to replicate in MT-4 cells. Nevertheless, a 10-fold increase in the IC 50 for AZT was observed for the triple mutant D67N/K70R/M73K HIV-2 RT

NRTI excision suppression in HIV-2 RT
when compared with the double mutant D67N/K70R. These data are consistent with the biochemical assays showing that amino acid substitutions M73K and to a lesser extent I75V confer resistance through increased primer-unblocking activity on AZT-terminated primers.

Discussion
In HIV-1, high-level AZT resistance develops through the accumulation of TAMs in the viral RT. These amino acid substitutions confer an excision activity that enables the polymerase to unblock DNA primers terminated with AZT and other NRTIs, and resume DNA synthesis. The excision reaction takes place in the presence of a pyrophosphate donor (usually ATP). However, in HIV-2 this pathway is suppressed, and TAMs are rarely observed in treated individuals (23,(35)(36)(37)(38)(39). In HIV-2 RT, AZT resistance is usually conferred by Q151M despite the fact

NRTI excision suppression in HIV-2 RT
that S215F/S215Y, a key substitution in the excision mechanism, requires only one nucleotide change, in contrast to HIV-1, where two mutations are needed to generate the equivalent substitution (i.e. T215F/T215Y). Q151M occurs at the nucleotide-binding site and affects the enzyme's ability to discriminate between NRTI triphosphates and dNTPs (40).
Our results demonstrate that the major determinants that suppress resistance through the excision pathway in HIV-2 are located adjacent to the ␤3-␤4 hairpin loop, at positions 73 and 75 of the RT. HIV-2 RT bearing the four major TAMs (M41L/ D67N/K70R/S215Y) lacks ATP-dependent excision activity and fails to unblock and extend primers terminated with AZT or tenofovir. However, when the HIV-1 RT sequence NSTRWRKLV (residues 67-75) was introduced at the equivalent positions of the HIV-2 RT in combination with TAMs M41L and S215Y, the enzyme was turned into an excisionproficient RT. Interestingly, our mutagenesis analysis shows that major contributors to this phenotype are Lys 73 and to a lesser extent Val 75 . The presence of Ile 75 in HIV-2 may favor the acquisition of resistance through the discrimination pathway because V75I is part of the multi-NRTI resistance Q151M complex of HIV-1 RT (28) and V75I antagonizes the effect of TAMs in HIV-1 RT (27). As in HIV-1, development of Q151M in HIV-2 requires at least two nucleotide changes with potential viable intermediates bearing amino acid substitutions Q151I or Q151L (38,41,42). Residues in the vicinity of Tyr 215 or amino acid substitutions with a potential antagonistic effect on TAMs do not seem to have a major impact on HIV-2 NRTI resistance through the excision mechanism. These results are in agreement with previous findings showing that HIV-2 RT mutants S215Y and F214L/S215Y had negligible excision activity on NRTI-terminated DNA primers (26).
Met 73 in HIV-2 RT plays a critical role in the suppression of the excision pathway. All HIV-2 RT mutants having the substitution M73K were excision-proficient in biochemical assays. However, HIV-2 variants carrying this substitution in the RT showed impaired viral replication, rendering non-viable virus in those cases where HIV-2 contained mutant RTs 5M_SK and 5M_STK. Although HIV-2 RT variants with the substitution M73K were active in biochemical assays, we found that those mutated enzymes were prone to degradation, and therefore additional purification steps had to be introduced to obtain enzymes of highest purity. In HIV-1 RT, random mutagenesis studies have shown that Lys 73 can be replaced by many different amino acids (even non-conservative substitutions) while maintaining catalytic activity, although Met was not identified among the recovered active mutants (43). However, Lys 73 is absolutely conserved in all groups and subtypes of HIV-1, whereas Met 73 is conserved in all HIV-2 groups as well as in major simian immunodeficiency virus (SIV) species (e.g. mac, mne, and smm) (44).
Our results showing that M73K could confer HIV-2 RT-significant ATP-dependent rescue activity in enzymatic assays in the presence of D67N/K70R and decreased AZT susceptibility to the corresponding recombinant virus raise the question of how Lys 73 could affect the conformation of the ␤3-␤4 hairpin loop and the interactions needed for NRTI removal. Previous molecular dynamics studies with insertion-and deletion-containing HIV-1 RTs had revealed that changes at positions 67 and 70 affected critical hydrogen bond interactions between amido groups of Arg 70 and the hydroxyl substituents of the dNTP ␥-phosphate (32). These interactions that could be established in the mutant D67N/K70R RT were consistent with the formation of equivalent hydrogen bonds in the crystal structure of the excision-proficient HIV-1 RT M41L/D67N/ K70R/T215Y/K219Q, complexed with AZT-terminated DNA/ DNA template-primers, and AZTppppA (30).
In contrast, when we replaced Met for Lys 73 in the HIV-1 RT D67N/K70R model and performed molecular dynamics simulations, there were remarkable conformational changes that eventually led to the loss of the potential hydrogen bond between Arg 70 and the incoming dNTP. This proposal is consistent with the results of rescue assays carried out with AZTMP-terminated primers that showed that unlike the double-mutant D67N/K70R, the HIV-1 BH10 RT D67N/K70R/ K73M had negligible ATP-dependent phosphorolytic activity (Fig. 6).
According to the molecular models, the hydrophobic side chain of Met 73 moves away from the dNTP-binding site into the hydrophobic core of the RT to interact with Val 60 . Interestingly, Val or Ile are found at this position in all HIV-1 strains, whereas Thr 60 is highly conserved in all HIV-2 and SIV variants. The

NRTI excision suppression in HIV-2 RT
interaction between Met 73 and Thr 60 is likely to be stabilized through hydrogen bonding and/or hydrophobic interaction of their side chains. Therefore, the presence of Thr 60 in the HIV-2 RT would be an additional restriction to the development of NRTI resistance through the excision pathway, if wildtype Met 73 is replaced by Lys.
Apart from their influence on nucleoside analogue resistance, differences in the conformation of the ␤3-␤4 hairpin loop may also affect other properties of the HIV-2 RT DNA polymerase activity. Recently, we showed that the tenofovir resistance-associated mutation K65R that confers increased fidelity of DNA synthesis in HIV-1 group M and group O RTs had no effect on the accuracy of the enzyme when introduced in the HIV-2 RT (45). This result was consistent with a previous report showing relative minor differences in the mutation rates determined for WT and K65R RT-containing SIV mac239 infecting pigtailed macaques (46). Interestingly, the amino acid sequence of the ␤3-␤4 hairpin loop region is identical in HIV-2 and SIV mac239 RTs.
Despite having similar NRTI susceptibilities, our work underlines the important differences between HIV-1 and HIV-2 RTs, including the development of resistance through different pathways. However, these differences can be mapped to a couple of residues and suggest that genotypic monitoring for the emergence of mutations affecting the ␤3-␤4 hairpin loop will be needed in the long run to prevent the eventual emergence of HIV-2 NRTI resistance through the excision pathway. In any case, continuing efforts in the development of anti-HIV-2 drugs are needed to overcome limitations imposed by the reduced armamentarium of antiretroviral drugs available to combat HIV-2 infection.

Mutagenesis
Site-directed mutagenesis was carried out by using the standard QuikChange TM (Stratagene) protocol. The pT5m plasmid encoding the large subunit of the HIV-2 ROD RT was used as template (50) to obtain the HIV-2 RT mutants. Complementary mutagenic primers (Table S1) were used to amplify the entire pT5m(RODRT) plasmid in a thermocycling reaction carried out with high-fidelity Pfu DNA polymerase. Mutant RTs were constructed in an orderly manner. At first, we obtained the double-mutant D67N/K70R that was used as template to generate the M41L/D67N/K70R/S215Y mutant. The I75V substitution was introduced in this construct to obtain M41L/D67N/ K70R/I75V/S215Y, a quintuple-mutant that was then use to generate derivatives containing one additional mutation, such as K68S, T88W, I181Y, N210L, or N210W. The complex M41L/D67N/K68S/K70R/I75V/S215Y was used as template to introduce additional mutations such as N69T, M73K, or N69T/ M73K. The combination G211R/L212W was introduced in the context of M41L/D67N/K70R/I75V/N210L/S215Y, and the obtained plasmid was used as template to generate the mutant M41L/D67N/K70R/I75V/N210L/G211R/L212W/F214L/S215Y that contained the F214L substitution. HIV-1 BH10 mutants D67N/K70R and D67N/K70R/K73M were obtained by following the same protocol, but using plasmid p66RTB as template (48,49) and the corresponding mutagenic primers listed the Table S1. All constructs were verified by DNA sequencing.

NRTI excision suppression in HIV-2 RT
DTT, and 5% (w/v) polyethylene glycol 6000. The labeled template-primers (75 nM) were preincubated with the corresponding RT (60 nM active enzyme concentration) at 37°C for 10 min. Primers were then blocked by adding an equal amount of buffer containing AZTTP or tenofovir-DP, at a final concentration of 25 M, and incubated at 37°C for 30 min. Rescue reactions were initiated by adding a mixture of all dNTPs in the presence of 3.2 mM ATP. In rescue assays carried out with DNA/DNA template-primers, dNTPs were supplied at 100 M (final concentration), except dATP (with D38/25PGA) or dTTP (with D38T/25PGA) whose concentration was kept at 1 M to minimize the inhibitory effect of the next complementary dNTP, caused by the formation of "dead-end" complexes (9,56). Inhibitor and dNTP concentrations were 2-fold higher in assays carried with the RNA/DNA template-primer 31Trna/ 21P. After addition of dNTPs and the pyrophosphate donor, reactions were incubated for up to 20 min at 37°C to facilitate nucleotide excision and primer elongation. Aliquots were removed at appropriate times, and reactions were stopped by adding an equal amount of sample loading buffer (10 mM EDTA in 90% formamide containing 3 mg/ml xylene cyanol FF and 3 mg/ml bromphenol blue). Products were resolved on denaturing 20% (w/v) polyacrylamide, 8 M urea gels, and primer rescue was quantified by phosphorimaging with a BAS 1500 scanner (Fuji), using the program Tina version 2.09 (Raytest Isotopenmessgerate Gmbh, Staubenhardt, Germany).

Kinetics of the ATP-dependent excision reaction
The kinetics of the excision reaction were monitored with template-primers D38/25PGA d4TMP and D38/25PGA AZTMP , obtained after blocking the primer with nucleotide analogues with terminal deoxynucleotidyltransferase (Roche Applied Science) or WT HIV-1 BH10 RT, respectively (13,53). AZTMP and d4TMP excision was determined by using template-primers D38/25PGA AZTMP and D38/25PGA d4TMP , respectively. To ensure single turnover conditions, we used an excess of enzyme over template-primer. Thus, RTs (80 nM) and blocked template-primers (60 nM) were preincubated at 37°C for 10 min in 0.11 M Hepes buffer, pH 7.0, containing 30 mM NaCl, 30 mM magnesium acetate, 130 mM potassium acetate, 1 mM DTT, and 5% (w/v) polyethylene glycol 6000. Reactions were initiated by adding 1 volume of a solution containing 6.4 mM ATP in 130 mM potassium acetate, 1 mM DTT, and 5% (w/v) polyethylene glycol 6000. Excision reactions were carried out for 0 to 80 min at 37°C, and aliquots were removed at appropriate times, stopped with sample loading buffer, and analyzed by denaturing PAGE as described above. The amount of hydrolyzed product generated over time was fitted to a single exponential decay equation: [P] ϭ A ϫ e Ϫ k obs ϫ t, where k obs is apparent kinetic constant of the excision reaction.

HIV drug susceptibility assays
AZT and darunavir susceptibility assays were carried out with infectious HIV-2 clones generated in this study and with previously described recombinant HIV-1 carrying either the WT BH10 RT or a double-mutant M41L/T215Y RT (49). To generate infectious HIV-2 carrying the different RT variants, we prepared three DNA fragments by PCR covering the full-length HIV-2 proviral DNA. These three PCR DNA fragments overlap 31-35 bp, and after transfection of susceptible MT-4 cells, they were able to generate infectious HIV-2 particles (57). The template for fragments I and III (HIV-2 ROD strain nucleotides 1-2421 and 4019 -9671, respectively) was proviral DNA obtained from MT-4 cells infected with HIV-2 ROD (Centre for AIDS Reagents, NIBSC, ID 0121) (58). Fragment II, corresponding to HIV-2 RT (HIV-2 ROD nucleotides 2381-4049), was obtained by PCR from the RT expression plasmids described above. PCR oligonucleotides for fragments I, II, and III were as follows: 1HIV-2RODF (5Ј-GGTCGCTCTGCGGA-GAGGCT-3Ј, HIV-2 ROD nts 1-20); 2HIV-RODR (5Ј-GGCTT-TAGCATTATTTTTATTGGGTCTACTTTGGC-3Ј, HIV-2 ROD nts 2384 -2421) and 2HIV-2RODF (5Ј-GCCAAAGTAGACC-CAATAAAAATAATGCTAAAGCC-3Ј, HIV-2 ROD nts 2384 -2421); 3HIV-2RODR (5Ј-GTCTGATACCCTGAGTCAC-TAAATGATCTAC-3Ј, HIV-2 ROD nts 4019 -4049) and 3HIV-2RODF (5Ј-GTAGATCATTTAGTGACTCAGGGTA-TCAGAC-3Ј, HIV-2 ROD nts 4019 -4049); and 5HIV-2RODR (5Ј-TGCTTCTAACTGGCAGCTTTATTAAGG-3Ј, HIV-2 ROD nts 9645-9671), respectively. Proviral DNA was purified using the QuickExtract DNA extraction protocol (Epicenter Biotechnologies). Cycling parameters were 1 cycle of denaturation at 94°C for 2 min, followed by 40 cycles of 45 s at 95°C, 45 s at 55°C, and 2-6 min at 68°C, with a final extension step of 68°C for 7 min. High fidelity Platinum TaqDNA polymerase (Thermo Fisher Scientific) was used for PCR amplifications. After PCR purification (QIAquick PCR purification kit, Qiagen), 100 ng of each of the three PCR fragments were co-transfected into MT-4 cells as we described previously (32). Cell culture supernatants were harvested on days 3, 5, and 7 after transfection when the concentration of HIV-1 p24 antigen (Innotest HIV Antigen mAB, Fujirebio) surpassed 500 ng/ml. If p24 antigen was not detected after 7 days of culture, three blind passages, feeding the cultures with fresh medium and new MT-4 cells, were performed to recover viable virus. The nucleotide sequence of the RT-coding region of the progeny virus was checked for possible reversions or undesired mutations. After three blind passages without p24 antigen detection in the cultured supernatant, the construct was considered non-viable in MT-4 cell culture. Susceptibility to AZT and darunavir was determined in MT-4 cells, as described previously (49), using a multiplicity of infection of 0.003. Viable cells at different drug concentrations were quantified with a tetrazolium-based colorimetric method (59). AZT and darunavir were obtained from the AIDS Research and Reference Reagent Program (National Institutes of Health). HIV-2 shows natural resistance to nonnucleoside RT inhibitors and several protease inhibitors. We used darunavir as a control due to its potent in vitro activity against both types of HIV (4).

Homology modeling and molecular dynamics
A structural model of HIV-1 RT bearing the amino acid substitutions D67N/K70R/K73M bound to double-stranded DNA and dTTP was constructed by standard homology modeling techniques (60), using as reference the one previously obtained for the double-mutant D67N/K70R (32). Molecular dynamics simulations, based on the models, were performed as described

NRTI excision suppression in HIV-2 RT
previously for the WT HIV-1 RT (34). This model for the catalytically-competent WT RT was based on the X-ray structure determined by Huang et al. (61) (PDB file 1RTD) corresponding to a ternary complex of HIV-1 RT, a dideoxyguanosineterminated primer-template and dTTP, and containing two Mg 2ϩ ions at the DNA polymerase active site. Atomic charges for dTTP were obtained with the RESP program (62) to fit potentials calculated at 6 -31G* level using the Gaussian-03 package (63). The van der Waals parameters for Mg 2ϩ were taken from Allnér et al. (64). The total simulation length was around 10 ns, and the analysis of trajectories was performed as reported previously (34). The SHAKE algorithm was applied allowing for an integration time step of 2 fs (34).