Drug Resistance Mutations of HIV Type 1 Non-B Viruses to Integrase Inhibitors in Treatment-Naive Patients from Sub-Saharan Countries and Discordant Interpretations

The human immunodeficiency virus type 1 (HIV-1) integrase is an enzyme of 288 amino acids involved in the integration of the viral DNA into the host cell genome. Several integrase inhibitors (INIs) have recently been developed or are under development, including raltegravir (RAL), elvitegravir (EVG), GSK-1349572, GSK-364735, S-1360, L-870,810, and L-9160. ¹ Currently, RAL is the only FDA approved INI, mainly used in salvage treatment options for antiretroviral-failing patients who have already developed HIV drug resistance to traditional antiretrovirals (ARV), i.e., protease inhibitors (PIs) and reverse transcriptase inhibitors (RTIs). Numerous studies have contributed to identify drug resistance mutations that are selected in the integrase region under treatment or existing as natural viral polymorphisms. ^2,3 However, the majority of these studies are from developed countries where HIV-1 genetic diversity and treatment access are different from what is observed in Africa and other resource-limited countries. Moreover the selection of few mutations in the integrase gene was shown to sufficiently impair the effectiveness of INIs, ^{3, 4} therefore raising concerns about the impact of mutations already present as natural polymorphisms in the IN gene. Additional knowledge on naturally occurring integrase mutations is thus important to identify potential subtype effects and to assess the prevalence of resistance-associated mutations in African and resource-limited countries where the demand for new drug classes is increasing. The aim of this study was to describe, in non-B HIV-1 integrase sequences obtained from treatment-naive African patients, the prevalence of natural polymorphisms potentially influencing the susceptibility to INIs and their significance according to current interpretation algorithms.

We studied 97 sequences obtained from ARV treatment-naive HIV-1-infected patients that were collected between 2006 and 2008 in Cameroon (n=30), Burundi (n=18), Democratic Republic of Congo (DRC) (n=31), and in France (n=18). All patients from France originated from various countries in sub-Saharan Africa and all were infected with non-B HIV-1 viruses. Viral RNA was extracted from plasma samples using the QIAamp Viral RNA extraction kit (Qiagen, Courtaboeuf, France), and RNA was transcribed to cDNA with the Expand Reverse transcriptase system (Roche, Meylan, France) using the reverse primer INT1as (5′-CTTATRGCAGANTCTGHAAAACADTYAAART-3′). The integrase gene was amplified by a nested PCR using the Expand High Fidelity PCR system (Roche, Meylan, France) with outer primers INT1s (5′-AAGTAAAYATAGTAACAGAYTCACARTATGCATT-3′) and INT1as (INT1as- CTTATRGCAGANTCTGHAAAACADTYAAART) and inner primers INT2s (5′-CATGGGTACCAGCACAYAARGGRATTGGAGGAAA-3′) and INT2as (5′-CCTARTGGGATGTGTACTTCTGARCTT-3′). For patients included in France, the integrase gene was amplified with outer primers IN12 (5′-GCAGGATTCGGGATTAGAAG-3′) and IN13 (5′-CTTTCTCCTGTATGCAGACC-3′) and inner primers IN1 (5′-AAGGTCTATCTGGCATGGGTA-3′) and BH4 (5′-TCCCCTAGTGGGATGTGTACTTC-3′). The final PCR fragment of 1070 bp (inner primers INT2s and INT2as) or 1043 bp (inner primers IN1 and BH4) was agarose gel purified and directly sequenced using the BigDye Terminator v3.1 Cycle Sequencing kit (Applied Biosystems, Carlsbad, CA). Subtypes and circulating recombinant forms (CRFs) were determined by phylogenetic tree and recombinant analysis as previously described. ⁵ Differences in amino acid positions of the newly sequenced strains in the integrase region were identified by comparing the new sequences with a consensus subtype B sequence. We considered as relevant only mutations considered by at least one of the three currently used interpretation algorithms: ANRSv2011.05, HIVdbv6.0.11, and RegaV8.0.2 (http://hivdb.stanford.edu/).

The subtype distribution and CRF distribution of the 97 sequences obtained in the IN region were as follow: A (n=3), C (n=26), D (n=5), F1/F2 (n=5), G (n=6), H (n=13), J (n=5) and K (n=1), CRF02 (n=26); CRF01 (n=1), CRF06 (n=1), CRF09 (n=1), CRF11 (n=2), CRF25 (n=1), CRF30 (n=1), and CRF37 (n=1). Comparison of the new IN amino acid sequences with the consensus B sequence identified up to 30 mutated positions from amino acid (aa) 17 to aa 263, some of which were previously or are still reported as only polymorphic or associated with resistance to INIs. ^3,4 Nine of these positions including L74, T97, K156, E157, G163, T206, S230, D232, and R263 were considered to be associated with resistance to INIs (RAL/EVG) by at least one interpretation algorithm (Table 1). Overall, up to 58 (59.8%) out of the 97 non-B sequences analyzed carried at least one of these mutations, 45/97 (46.4%) had one mutation, 12/97(12.4%) had two, and one sequence carried up to three (E157Q, T206S, and R263K). Of these mutations, only one was considered as a major INIs resistance mutation, i.e., E157Q conferring resistance to RAL that we found in 5/97 (5.2%) samples tested, one subtype C, two D, and two CRF02_AG. This mutation was considered as a primary mutation only by the ANRS algorithm and was not considered either by HIVdb or Rega. None of the following primary mutations, T66IAK, E92QV, F121Y, Y143RCH, P147S, Q146P, S145G, Q148HRK, V151AL, and N155HS, known to be reducing RAL or EVG susceptibility by several fold (≥5 fold), ⁴ was found.

The frequency of the other identified resistance mutations varied from 1% to almost 50%: L74M (3/97, 3.1%), T97A (9/97; 9.3%), K156N (2/97; 2.1%), G163K (1/97; 1.0%), T206S (47/97; 48.5%), S230N (1/97; 1.0%), D232N (1/97; 1.0%), and R236K (1/97; 1.0%) and all were considered as accessory resistance mutations by one or several interpretation algorithms, mostly by the HIVdb and Rega algorithms (Table 1). The ANRS rules selected only L74 and T97 as accessory resistance mutations and they are considered significant only if at least two other mutations are present. According to the viral subtype, we did not find any specific mutational pattern and all the subtypes analyzed appeared to randomly harbor no or several mutations. The three subtype A strains and the single subtype K isolate carried no mutation, and the limited sample number can explain this observation. On the other hand, we found up to five different mutations among subtype C and D strains, including L74M, T97A, K156N, E157Q, and T206S; and the remaining subtypes and CRFs carried from one to four different mutations. Few mutations as G163K, S230N, D232N, and R263K were observed in only one subtype, respectively, F1/F2, G, CRF11_cpx, and CRF02_AG, while T206S was identified at a relatively high frequency among almost all subtypes, reaching 100% (25/25) frequency among CRF02 isolates (Table 1).

Naturally occurring resistance to RAL and to other INIs, which are still under clinical development, has not been reported yet and was also not observed in our study population of non-B-untreated patients. Several accessory mutations that contribute to INI resistance only in the presence of primary INI resistance mutations have been, however, reported among B and non-B strains, ^2,6,7 and included some of those we reported here. A recent study analyzing 4470 sequences of INI-untreated patients and including 52% of non-B isolates reported frequencies between 0.5% and 2.5% for L74M (2.5%), T97A (2.2%), E157Q (2.0%), G163K (0.4%), and R263K (0.1%) ⁸ and except for E157Q, the frequencies of these mutations were shown to increase significantly among RAL failing patients ⁴ illustrating their contribution when primary mutations are selected. These mutations were also reported as occurring at higher proportion among non-B compared to B isolates. ⁶

We observed a significantly high prevalence for some mutations as T206S, which is currently considered only by the Rega algorithm, but is likely to be a highly polymorphic mutation as previously reported, especially among non-B subtypes. ^2,3,6 E157Q, found in 5% of INI-naive patients in the present study, is considered as an accessory mutation in many studies. ^4,8 In addition, using the QuickAlign tool from the HIV Sequence Database shows that the E157 is replaced (usually by Q) in 2.56% of sequences in the alignment of 1804 sequences used by that tool (http://www.hiv.lanl.gov/content/sequence/QUICK_ALIGN/QuickAlign.html). However, this mutation was reported as reducing by itself the enzymatic activity of the integrase enzyme and was therefore considered as a primary RAL mutation in the ANRS algorithm. ⁹ Also, this mutation was rapidly selected and archived in HIV-DNA within a short term of 8 weeks low-level replication in a heavily treated patient. ¹⁰ Nonetheless, the fact that these findings relied only on few subtype B strains does not make it possible to fully understand the impact of this mutation in non-B populations, among which other studies also reported a prevalence of up to 4.1% (3/73), in two subtype C and one recombinant. ¹¹ As for almost all resistance mutations identified in this study, additional in vitro and clinical data involving non-B viruses are necessary to better appreciate the clinical relevance of these amino acid changes in the IN region of HIV-1 non-B variants.

Finally, among the nine resistance-associated positions identified in this study by the three currently used interpretation algorithms, only two, L74 and T97, were concordantly interpreted as shown in Table 1. As previously reported for genotypic drug resistance interpretation in the protease and reverse transcriptase regions, especially for non-B viruses, ¹² several mutations were discordantly interpreted by the three algorithms, ANRS considered only three mutations out of nine, four of nine were chosen by HIVdb, and Rega selected up to seven of nine. The major illustrations of this interalgorithms discordance were the selection of E157Q as a primary mutation only by the ANRS algorithm and the highly prevalent T206S mutation selected only by Rega. Additional studies involving non-B subtypes and also the establishment of standardized definition of accessory resistance mutations, as well as in vitro and clinical cutoff harmonization, could help in reducing these interalgorithms discordances.