(This work was presented at the 14th Conference on Retroviruses and Opportunistic Infection, Los Angeles, CA, 25 to 28 February 2007.)

We selected nine PI-naive HIV-2-infected patients from the French HIV-2 ANRS cohort. Six of these patients subsequently received HAART regimens including a PI (indinavir, nelfinavir, saquinavir, and/or ritonavir) for a median of 13 months (range, 2 to 36) and had plasma viruses harboring mutations in the protease gene. Peripheral blood mononuclear cell (PBMC) coculture isolates were collected from these patients before (T0, n = 6) and during (T1 [time of collection of first plasma specimen during PI exposure], n = 6; T2 [time of collection of second plasma specimen during PI exposure], n = 2) PI exposure. Protease gene sequences of plasma and PBMC isolates were determined as previously described (3). Amino acid changes were compared with those associated with drug resistance in HIV-1 (International AIDS Society-USA [IAS-USA]). We determined the in vitro phenotypic susceptibility of clinical HIV-2 isolates and HIV-2 reference strain ROD and HIV-1 reference strain BRU to amprenavir, atazanavir, darunavir, indinavir, lopinavir, nelfinavir, saquinavir, and tipranavir using the ANRS PBMC assay (4). Phenotypic inhibitory quotients (PIQs) were calculated for each clinical HIV-2 isolate as the ratio between the trough plasma PI concentration and the 50% inhibitory concentration (IC₅₀). The PIQs were not adjusted for protein binding. Table ​Table11 shows the phenotypic results and protease gene sequences of the HIV-2 clinical isolates and the phenotypic results of the HIV-2 and HIV-1 reference isolates.

The protease sequences of the wild-type HIV-2 strains, compared to the HIV-1 clade B consensus sequence, contained several amino acids associated with HIV-1 PI resistance, such as 10I/V, 16E, 20V, 32I, 33V, 35G, 36I, 43T, 46I, 47V, 58E, 62V, 69K, 71V, 73A, 82I, and 93L. Other differences were observed at positions involved in but not associated with HIV-1 resistance, such as 13A, 34A, 60K, 63E, 76M, 77T, 85F, and 89I. Relative to the HIV-1 reference strain, the median IC₅₀ values of the HIV-2 wild-type isolates were 31-fold higher for amprenavir, eightfold higher for atazanavir, sevenfold higher for tipranavir, and threefold higher for indinavir and nelfinavir. Darunavir, lopinavir, and saquinavir median IC₅₀ and IC₉₀ values were similar for HIV-1 and the wild-type HIV-2 isolates (Table ​(Table1).1). Viruses isolated from the six PI-experienced patients at T1 and T2 harbored the I82F, I84V, and L90M substitutions, alone or in combination with minor HIV-1 PI mutations, such as V10I, V33I, I54M, I64V, V71I, and I89V. Compared to the corresponding wild-type isolates, the six mutants showed increases of fourfold to >10-fold in the IC₅₀ and/or IC₉₀ values of all tested PIs, at both T1 and T2. The PIQ values of amprenavir, atazanavir, indinavir, nelfinavir, and tipranavir were, respectively, 33-fold, eightfold, threefold, threefold, and sevenfold lower for HIV-2 wild-type strains than for HIV-1. Darunavir, lopinavir, and saquinavir PIQ values were similar.

PIs are designed to fit the active site of the HIV-1 protease and are sensitive to structural changes in the viral protein. It has been reported that the therapeutic outcome of HIV-2-infected patients might be influenced by the choice of PIs (1, 15). For amprenavir, our data are in keeping with those reported elsewhere, showing significantly less activity against HIV-2 wild-type strains than against HIV-1 (9, 14, 16). These phenotypic results could be explained by the natural presence in HIV-2 protease of amino acids associated with resistance to HIV-1, which might influence the binding affinity of the PIs for HIV-2 protease (2, 3, 9, 10, 11). In our study, all the HIV-2 wild-type strain protease sequences naturally presented the amino acids 32I and 47V, associated with resistance in HIV-1 infection to amprenavir, according to the IAS-USA list and to different genotypic resistance interpretation algorithms (www.hivfrenchresistance.org, http://hivdb.stanford.edu and www.kuleuven.be/rega/cev/links/rega_algorithm). We found that clinical HIV-2 isolates and HIV-1 reference strains had similar phenotypic susceptibility to saquinavir, lopinavir, and darunavir. As amprenavir and darunavir are structurally close, we expected darunavir to be relatively ineffective in HIV-2. Crystallographic studies with HIV-1 showed that darunavir interacts directly with the main chains of aspartic acid residues (Asp-29 and Asp-30), whereas other PIs interact with side chains in the S2 subsite of the HIV-1 enzyme (6, 7). Moreover, it has been reported for HIV-1 that the binding affinity of darunavir for the wild-type protease was >100-fold higher than those of other PIs due to the slower dissociation rate of this molecule from the protease active site (5). In the same way, crystallography structure studies of the HIV-2 protease and binding affinity experiments might help us to understand the difference observed in the natural susceptibilities of HIV-2 strains to these two drugs as well as phenotypic resistance in HIV-2 mutated strains. The IC₅₀ and IC₉₀ values of atazanavir, indinavir, nelfinavir, and tipranavir for the HIV-2 isolates were higher than those observed for HIV-1, suggesting the hypothesis of the lower activities of these PIs against HIV-2. However, these values were lower than their respective trough plasma concentrations. This might be explained by the fact that HIV-2 wild-type isolates harbored several amino acids associated with PI resistance in HIV-1.

PI treatment-associated amino acid changes in the HIV-2 protease gene occurred at positions known to confer PI resistance in HIV-1 and were not associated with the use of a particular PI without any order of accumulation. They altered the phenotypic susceptibility of the isolates to all the PIs tested here. These results are in keeping with data published elsewhere (2, 3, 9, 11, 13, 15). Mutagenesis experiments coupled with phenotypic susceptibility testing might help to determine the impact of each substitution on PI resistance. Saquinavir, lopinavir, and darunavir appear to be the best choices for first-line therapy of HIV-2 infection, while amprenavir should not be used. Atazanavir and tipranavir might be used with care (17). Our results suggest that treatment guidelines for HIV-1-infected patients should not be directly extrapolated to HIV-2-infected patients. Virological efficacy data in vivo might help us to evaluate the place of PIs in HIV-2 antiretroviral strategy.