Antiretroviral Drug Resistance in HIV-2: Three Amino Acid Changes Are Sufficient for Classwide Nucleoside Analogue Resistance

Abstract

Genotypic surveys suggest that human immunodeficiency virus type 1 (HIV-1) and HIV-2 evolve different sets of mutations in response to nucleoside reverse-transcriptase inhibitors (NRTIs). We used site-directed mutagenesis, culture-based phenotyping, and cell-free assays to determine the resistance profiles conferred by specific amino acid replacements in HIV-2 reverse transcriptase. Although thymidine analogue mutations had no effect on zidovudine sensitivity, the addition of Q151M together with K65R or M184V was sufficient for high-level resistance to both lamivudine and zidovudine in HIV-2, and the combination of K65R, Q151M, and M184V conferred classwide NRTI resistance. These data suggest that current NRTI-based regimens are suboptimal for treating HIV-2 infection.

HIV-2 infection is endemic in West Africa and has achieved a limited prevalence in southern Africa, India, Brazil, and parts of western Europe [1]. As with HIV type 1 (HIV-1) infection, drug resistance in HIV-2 is a major barrier to sustained antiretroviral therapy. HIV-2 is intrinsically resistant to nonnucleoside reverse-transcriptase inhibitors (NNRTIs) and the fusion inhibitor T-20 (enfuvirtide) [2], and some HIV-2 isolates also exhibit reduced susceptibility to certain protease inhibitors [2, 3]. In contrast, wild-type (WT) HIV-1 and HIV-2 exhibit comparable sensitivities to nucleoside reverse-transcriptase inhibitors (NRTIs) [4]. Although regimens that include 2 NRTIs and a protease inhibitor can initially suppress viral RNA levels in treatment-naive patients infected with HIV-2 [5], the emergence of drug-resistant variants in response to therapy [5–7] is a major obstacle to clinical treatment because most of the inhibitors that are active against HIV-2 in vitro are not widely available in West Africa and other developing regions.

Currently, efforts to identify the genetic changes responsible for drug resistance in HIV-2 in vivo are limited to a handful of small-scale studies. One potentially important trend observed in HIV-2–infected patients is the frequent emergence of mutations that encode the K65R and Q151M substitutions in reverse transcriptase (RT) [5–8] (figure A1 in appendix A, which appears only in the electronic version of the Journal). These changes are substantially more common in HIV-2 than in HIV-1 and often appear together with M184V in bulk sequences obtained from individuals who have received NRTI therapy. In contrast, the thymidine analogue mutations (TAMs) that provide the principal route to nucleoside analogue resistance in HIV-1 [9] are rarely observed in HIV-2. Although it is appropriate to be cautious in interpreting the limited amount of genotypic data available for HIV-2, existing studies suggest that HIV-1 and HIV-2 evolve NRTI resistance by different mutational pathways. This inference relies on the assumption that known genotype-phenotype relationships for HIV-1 are directly applicable to HIV-2, because the resistance phenotypes conferred by specific mutations in HIV-2 RT have not been quantified in a defined molecular clone and because only limited data from biochemical and culture–based studies are available [10–12]. To obtain a better understanding of the evolution of drug resistance in HIV-2, we directly tested the effects of clinically observed substitutions in HIV-2 RT on NRTI sensitivity using site-directed mutagenesis, cell culture-based phenotypic assays, and cell-free polymerase reactions.

Materials and methods

Site-directed mutants of HIV-2_ROD were constructed in the pROD9 molecular clone, which was kindly provided by Michael Emerman (Fred Hutchinson Cancer Research Center, Seattle, Washington) (see appendix B, which appears only in the electronic version of the Journal). Mutants of HIV-1_NL4-3 were constructed in pNL4-3 or in a modified version of pR9, as described elsewhere [13]. Each WT or mutant molecular clone was transfected into 293T-17 (293tsA1609_neo) cells for the production of virus, and drug susceptibilities were measured using our established HeLa-CD4 (MAGIC-5A) indicator cell assay [4, 13] (see appendix B, which appears only in the electronic version of the Journal). This assay quantifies the dose-dependent inhibition of HIV replication in a single cycle of infection, enabling head-to-head comparisons of HIV-1 and HIV-2 in the same cell type. Zidovudine (AZT), didanosine (ddI), and stavudine (d4T) were obtained commercially (Sigma-Aldrich), as were tenofovir (PMPA) and abacavir (ABC) (Moravek Biochemicals). Lamivudine (3TC) and emtricitabine (FTC) were kindly provided by Raymond Schinazi (Emory University, Atlanta, Georgia) or were obtained commercially (Moravek Biochemicals).

To examine the effects of amino acid replacements in HIV-2 RT on nucleoside analogue monophosphate incorporation, we expressed WT and mutant forms of HIV-1 and HIV-2 RT in Escherichia coli and purified the heterodimeric forms of the enzymes by column chromatography. Detailed descriptions of the RT-expressing plasmids, conditions for bacterial growth, purification procedures, and assays used to quantify analogue triphosphate susceptibility are provided in appendix B, which appears only in the electronic version of the Journal.

The drug concentrations required to inhibit focus formation in cell culture by 50% (EC₅₀ values) or to inhibit DNA synthesis by 50% in cell-free reactions (IC₅₀ values) were calculated using a sigmoidal regression equation in Prism (version 4.0; GraphPad Software). Statistical significance was determined by analysis of variance of log₁₀ IC₅₀ or EC₅₀ values in conjunction with Tukey’s posttest.

Results

The frequency with which K65R, Q151M, and M184V appear together in HIV-2 RT suggests that these replacements define a genetic pathway leading to escape from NRTI-based regimens [5–8] (figure A1 in appendix A, which appears only in the electronic version of the Journal). We therefore examined the individual and combined effects of K65R, Q151M, and M184V on NRTI sensitivity in HIV-2 (table 1). We initially introduced point mutations encoding the Q151M replacement into HIV-2_ROD and tested the susceptibility of the resulting variant to AZT. Surprisingly, the dose-response profile of Q151M HIV-2_ROD was comparable to that of the highly resistant Q151M/A62V/V75I/F77L/F116Y (Q151M+4) HIV-1_NL4-3 mutant (figure A2 in appendix A, which appears only in the electronic version of the Journal). Over multiple trials, Q151M HIV-2_ROD exhibited a mean EC₅₀ for AZT that was 43-fold greater than that for WT HIV-2_ROD, whereas the mean EC₅₀ for Q151M HIV-1_NL4-3 was only 4-fold greater than that for WT HIV-1_NLA-3 (table 1). To determine whether this outcome was specific to the ROD isolate of HIV-2, we replaced the RT-encoding region of pROD9 (RT codons 14–542) with an equivalent region from the HIV-2_UC2–encoding plasmid pUC2 and introduced mutations encoding Q151M into the resultant molecular clone. As observed for Q151M HIV-2_ROD, the Q151M variant of HIV-2_ROD/UC2 was highly resistant to AZT (a 30-fold increase in EC₅₀, compared with that of WT HIV-2_ROD-UC2; the mean EC₅₀ values [±SD] were 0.071 ± 0.2 and 2.1 ± 0.2 μmol/L for WT and Q151M HIV-2_ROD/UC2, respectively). Taken together, these data demonstrate that Q151M alone is sufficient to produce high-level AZT resistance in HIV-2.

Table 1.

Susceptibilities of HIV-1_NL4-3 and HIV-2_ROD strains to nucleoside analogue reverse-transcriptase inhibitors.

Strain, varianta	EC50 by drug, mean ± SD, μmol/Lb
	AZT	ddI	d4T	PMPA	3TC	FTC	ABC
HIV-1NL4-3
WT	0.16 ± 0.07 (1)	4.7 ± 2.9 (1)	5.5 ± 1.6 (1)	7.2 ± 2.0 (1)	0.87 ± 0.28 (1)	0.25 ± 0.09 (1)	7.3 ± 3.2 (1)
Q151M	0.65 ± 0.25 (4)	10 ± 3.7 (2)	20 ± 7.9 (4)	7.8 ± 3.5 (1)	1.4 ± 0.4 (2)	0.41 ± 0.31 (2)	21 ± 7.1 (3)
Q151M+4	7.0 ± 3.4 (44)	45 ± 17 (10)	79 ± 42 (14)	26 ± 11 (4)	3.8 ± 2.8 (4)	0.79 ± 0.10 (3)	40 ± 24 (5)
HIV-2ROD
WT	0.12 ± 0.06 (1)	8.7 ± 6.3 (1)	7.1 ± 3.7 (1)	8.4 ± 5.3 (1)	1.8 ± 1.2 (1)	0.60 ± 0.31 (1)	6.5 ± 2.1 (1)
Q151M	5.2 ± 2.3 (43)	42 ± 21 (5)	40 ± 14 (7)	2.9 ± 0.8 (0.3)	3.2 ± 0.9 (2)	2.7 ± 1.6 (5)	14 ± 8.3 (2)
M184V	0.073 ± 0.04 (1)	19 ± 14 (2)	2.1 ± 0.9 (0.3)	6.0 ± 2.1 (1)	>400 (>200)	>400 (>200)	6.2 ± 2.3 (1)
Q151M and M184V	3.5 ± 1.4 (29)	72 ± 11 (8)	17 ± 3.8 (2)	6.8 ± 4.6 (1)	>400 (>200)	>400 (>200)	14 ± 5.1 (2)
K65R	0.14 ± 0.03 (1)	37 ± 35 (4)	11 ± 2.1 (2)	8.7 ± 3.1 (1)	68 ± 29 (38)	51 ± 32 (85)	10 ± 4.1 (2)
K65R and Q151M	6.7 ± 1.6 (56)	55 ± 30 (6)	86 ± 46 (12)	18 ± 13 (2)	134 ± 50 (74)	152 ± 40 (250)	35 ± 32 (5)
K65R, Q151M, and M184V	7.9 ± 4.8 (66)	373 ± 71 (43)	29 ± 14 (4)	39 ± 23 (5)	>400 (>200)	>400 (>200)	>100 (>10)

NOTE. Drug concentrations required to inhibit focus formation in cell culture by 50% (EC₅₀ values; data are means ± SDs of data from ≥3 independent experiments) were obtained in cultured MAGIC-5A cells; see Materials for details. Values in bold type differ significantly from values for wild-type (WT) HIV-1_NL4-3 (P < .05, by analysis of variance with Tukey’s multiple-comparison test). Portions of the data for WT HIV-1_NL4-3, WT HIV-2_ROD, and Q151M/A62V/V75I/F77L/F116Y (mutant Q151M+4, which is resistant to multiple nucleoside reverse-transcriptase inhibitors) HIV-1_NL4-3 have been reported elsewhere [4]. ABC, abacavir; AZT, zidovudine; ddI, didanosine; d4T, stavudine; FTC, emtricitabine; PMPA, tenofovir; 3TC, lamivudine.

^aViruses produced by full-length plasmid clones of HIV-1_NL4-3 (pNL4-3 or pR9ΔApa) or HIV-2_ROD (pROD9).

^bNos. in parentheses are the n-fold change in EC₅₀, compared with the corresponding WT value.

In HIV-1 RT, TAMs confer AZT resistance by increasing the primer-unblocking activity of the polymerase, which results in enhanced removal of AZT-5′-monophosphate from the nascent DNA strand [9]. We constructed HIV-1_{NL4 3} and HIV-2_ROD variants encoding 2 pivotal replacements in the TAM series (M41L and T215Y) and compared their sensitivities to AZT. In contrast to HIV-1_NL4-3, HIV-2_ROD mutants that harbored M41L, T215Y, or both substitutions showed no detectable resistance to AZT in the single-cycle assay (<2-fold increase in EC₅₀ relative to WT HIV-2_ROD) (figure A3 in appendix A, which appears only in the electronic version of the Journal). This result is consistent with the infrequent occurrence of TAMs in HIV-2 sequences obtained from patients who had received antiretroviral therapy (figure A1 in appendix A, which appears only in the electronic version of the Journal).

To assess the effects of other replacements that are commonly associated with Q151M in HIV-1, we constructed an A62V/F77L/F116Y/Q151M (Q151M+3) variant of HIV-2_ROD. This mutant is genetically equivalent to Q151M+4 HIV-1_NL4-3 because HIV-2_ROD encodes an isoleucine at RT codon 75 that is highly conserved in HIV-2 sequences from treatment-naive patients [14]. Although A62V, F77L, and Y116F increase the level of Q151M-mediated AZT resistance in HIV-1 [15], the Q151M and Q151M+3 variants of HIV-2_ROD exhibited comparable sensitivity to the drug (mean EC₅₀ [±SD], 5.2 ± 2.3 vs. 3.1 ± 0.3 μmol/L, for Q151M and Q151M+3, respectively), indicating that A62V, F77L, and Y116F do not enhance AZT resistance in HIV-2.

We also measured the susceptibilities of Q151M-, K65R-, and M184V-containing variants of HIV-2_ROD to other NRTIs used in antiretroviral therapy. Q151M had no statistically significant effect on viral sensitivity to PMPA, 3TC, or ABC, but it imparted 5–7-fold resistance to d4T, ddI, and FTC, compared with that of the WT (table 1). K65R imparted 38-fold resistance to 3TC and 85-fold resistance to FTC, as well as 4-fold resistance to ddI. The M184V substitution conferred >200-fold resistance to both 3TC and FTC. Importantly, the addition of either K65R or M184V together with Q151M conferred robust resistance to 3TC and FTC without substantially reducing the level of Q151M-mediated AZT resistance; the HIV-2_ROD mutants that contained K65R and Q151M and the mutants that contained Q151M and M184V showed ≥70-fold resistance to 3TC and FTC, as well as 29-fold and 56-fold resistance to AZT, respectively (table 1). This result differs substantially from the outcome observed for HIV-1, in which M184V suppresses TAM-mediated AZT resistance by impairing the primer-unblocking activity of RT [9].

To further examine the potential for dual AZT-3TC resistance in HIV-2, we quantified the effects of K65R and Q151M on the incorporation of analogue-5′-monophosphate by purified HIV-2 RT. In comparison with WT HIV-2 RT, Q151M HIV-2 RT exhibited a 15-fold increase in the IC₅₀ for AZT-5′-triphosphate. The combination of K65R and Q151M conferred high-level resistance to AZT-5′-triphosphate (an 83-fold increase) and substantial resistance to 3TC-5′-triphosphate (a 7-fold increase) (table 2). Taken together with our cell culture data (table 1), these results demonstrate that the combination of Q151M with either K65R or M184V is sufficient to produce dual resistance to AZT and 3TC in HIV-2.

Table 2.

Sensitivity of HIV-1 and HIV-2 reverse transcriptase (RT) to dNTP analogues.

RTa	IC50 by dNTP analogue, mean ± SD, μmol/Lb
	AZTTP	3TCTP
HIV-1NL4-3
WT	0.086 ± 0.019 (1)	1.9 ± 0.1 (1)
Q151M	0.84 ± 0.13 (10)	ND
Q151M+4	4.5 ± 1.7 (52)	ND
HIV-2ROD
WT	0.092 ± 0.014 (1)	6.1 ± 0.4 (1)
Q151M	1.4 ± 0.1 (15)	5.2 ± 2.4 (1)
Q151M and K65R	7.6 ± 2.9 (83)	44 ± 10 (7)

NOTE. IC₅₀ values are from ≥3 independent experiments. Values in bold type are significantly different from the corresponding values for wild-type (WT) RT (P < .05, by analysis of variance of log IC₅₀ values with Tukey’s multiple-comparison test). For lamivudine (3TC)–5′-triphosphate (3TCTP), reactions with M184V HIV-1 RT served as a positive control and yielded an IC₅₀ >100 μmol/L. AZTTP, zidovudine (AZT)–5′-triphosphate; ND, not determined.

^aHeterodimeric p66/p51 HIV-1 and p68/p58 HIV-2 RTs expressed in Escherichia coli and purified by column chromatography.

^bNos. in parentheses indicate the n-fold change in IC₅₀, compared with the corresponding WT value.

Finally, in the HIV-2_ROD mutant that contained K65R, Q151M, and M184V we observed greater-than-additive increases in the levels of resistance to ddI and ABC, compared with the resistance levels of the single– or double–amino acid variants (table 1). As a result, the HIV-2_ROD mutant that contained K65R, Q151M, and M184V showed ≥40-fold resistance to AZT, ddI, 3TC, and FTC; >10-fold resistance to ABC; and 4–5-fold resistance to d4T and PMPA. These findings demonstrate that, in HIV-2_ROD, the combination of K65R, Q151M, and M184V confers classwide NRTI resistance, with high-level resistance to AZT, ddI, 3TC, FTC, and ABC.

Discussion

To our knowledge, this is the first study showing the individual contributions of clinically observed amino acid replacements in HIV-2 RT to nucleoside analogue resistance. Our analysis provides 2 important insights that help to explain why the mutations that emerge in HIV-2 during therapy differ from those typically seen in HIV-1. First, in contrast to HIV-1, a single Q151M replacement in HIV-2 RT confers high-level phenotypic resistance to AZT as well as substantial cross-resistance to other nucleoside analogues (table 1). Second, 2 key replacements in the TAM pathway (M41L and T215Y) have no effect on AZT susceptibility in HIV-2 in culture (figure A3 in appendix A, which appears only in the electronic version of the Journal). This result is consistent with the outcome of a recent study showing that, compared with WT HIV-1 RT, WT HIV-2 RT exhibits a substantially lower level of primer-unblocking activity [10]. Taken together, these data demonstrate that equivalent substitutions in HIV-1 and HIV-2 RT can have different effects on NRTI susceptibility and that the genetic algorithms used to predict drug resistance phenotypes in HIV-1 RT are not necessarily applicable to HIV-2.

The dual AZT-3TC resistance phenotypes observed in HIV-2 in cell culture (table 1) and in cell-free RT assays (table 2) are of particular concern because fixed-dose formulations of AZT and 3TC are still widely used in West Africa and other developing regions where HIV-2 infection is endemic. In HIV-1, high-level resistance to both AZT and 3TC typically requires the combination of multiple TAMs, M184V, and additional replacements in RT [9]. In contrast, in HIV-2, only 2 amino acid changes (K65R and Q151M or Q151M and M184V) are required for high-level resistance to both of these inhibitors (table 1). These findings strongly suggest that the mutational barrier to dual AZT-3TC resistance is lower in HIV-2 than it is in HIV-1. Although other NRTIs retained substantial antiviral activity against the HIV-2 variants tested in our study (i.e., PMPA and ABC; see table 1), we cannot presently exclude the possibility that, in HIV-2, high-level resistance to these drugs can also be acquired via a relatively small number of mutational steps. Additional studies are required to address this issue.

In a recent analysis of patient-derived HIV-2 isolates, Damond et al. concluded that Q151M alone only confers resistance to d4T and ABC and that resistance to other NRTIs involved the coselection of a V111I change in RT [11]. Although the strains examined in their study were comprised of mixed populations of variants, further examination of their data indicates that, in fact, Q151M confers broad-spectrum NRTI resistance both in the presence and in the absence of the V111I substitution. In our experiments, we introduced the Q151M change into both an HIV-2_ROD–based and an HIV-2_UC2–based RT back-ground; both of these strains encode a valine at position 111 of RT. Our findings lead us to conclude that Q151M alone is sufficient to produce high-level AZT resistance in HIV-2 and that the V111I substitution is not required for broad-spectrum NRTI resistance.

From a therapeutic perspective, resistance to multiple NRTIs is particularly challenging in HIV-2 because naturally occurring resistance mutations in HIV-2 appear to accelerate the development of high-level resistance to multiple protease inhibitors [2, 3]. Thus, strict adherence to therapy may be particularly important for delaying the outgrowth of drug-resistant HIV-2 variants. Our in vitro analyses of HIV-2 suggest that the emergence of NRTI-resistant mutants during first-line treatment, particularly in patients treated with AZT and 3TC, will likely result in poor responsiveness to subsequent nucleoside-based regimens. Our findings support initiatives to provide a broader array of antiretrovirals to HIV-2–infected patients in resource-limited settings and emphasize the need to identify drug combinations that inhibit HIV-2 strains resistant to multiple NRTIs.