Phenotypic impact of HIV reverse transcriptase M184I/V mutations in combination with single thymidine analog mutations on nucleoside reverse transcriptase inhibitor resistance

Objectives: To analyse the impact of the M184I/V mutation and individual thymidine-associated mutations (TAM) on nucleoside reverse transcriptase inhibitor (NRTI) phenotypic susceptibility and compare these results with those obtained using commercial and public algorithms.

Design: An HIV genotypic/phenotypic database with over 27 000 samples was used to obtain the median fold change (5-95th percentile) in NRTI phenotypic susceptibility for viruses from patients containing individual TAM with or without the M184I or V mutation and for wild-type patient viruses.

Results: The resulting data indicated that in vitro, individual TAM do not have an equivalent impact on NRTI resistance, with some individual TAM having little or no impact on NRTI resistance (e.g. M41L or K219Q/E/H/R). In the presence of the M184I/V mutation, re-sensitization to some drugs, including zidovudine, stavudine and tenofovir was observed despite the presence of a TAM. For didanosine and abacavir, the presence of the M184V mutation and a single TAM did not result in a fold-change increase associated with decreased drug susceptibility. Analysis of public and commercial algorithms revealed a lack of concordance regarding the impact of these mutations, and with the observed phenotypic data.

Conclusion: These analyses should assist in the creation of rules for genotypic drug resistance algorithms for a better reflection of the impact of individual TAM and also the impact of M184I/V on resistance. These data provide additional evidence that retaining lamivudine in those treatment regimens in which TAM can be selected may provide some therapeutic benefit by maintaining the M184V mutation.

Introduction

In recent years, antiretroviral therapy options have increased, and treatment regimens for HIV infection have become more complex. Phenotypic testing allows one to assess the impact of all the HIV protease and reverse transcriptase (RT) mutations in the HIV isolated from a patient's plasma. As HIV genotyping assays are quicker and less costly to perform than phenotypic assays, algorithms have been created to make better predictions of the impact of mutations on phenotypic and clinical response.

The HIV RT M184V mutation can be selected by several drugs, including lamivudine and abacavir, and has been associated with increased fidelity of the polymerase [^1-3], the inhibition of zidovudine- terminated primer unblocking [⁴], decreased viral fitness [⁵], and partial reversal of resistance to zidovudine and stavudine [^6-8], zidovudine hypersusceptibility [^5,9,10], and delays the development of thymidine analogue mutations (TAM) when used in combination with zidovudine [¹¹] or stavudine [¹²].

The rules used to create algorithms for genotypic interpretations are generally based upon the available literature data, when genotypic mutations and phenotypic resistance or clinical responses were determined, but typically do not account for all possible combinations of mutations that may be encountered. Clinically defined cut-offs are available for all antiretroviral agents in the nucleoside reverse transcriptase inhibitor (NRTI) class except zidovudine and zalcitabine.

In this study, we selected patient HIV samples containing a single TAM with or without the M184I/V mutation for which the phenotype was also obtained. The phenotypic results were compared with the results that would be predicted by several public and commercially available predictive genotypic algorithms for HIV drug resistance.

Materials and methods

Plasma samples from HIV-infected patients were analysed for HIV protease and RT mutations and for phenotypic resistance by ViroLogic Inc [¹³]. Phenotypic drug susceptibility was measured by PhenoSense HIV assay (ViroLogic Inc., South San Francisco, CA, USA), and results were expressed as the median of the IC₅₀ fold-change [clinical isolate IC₅₀ divided by IC₅₀ of simultaneously tested drug-sensitive reference virus (NL4-3)] values. Reduced susceptibility for each drug was defined by fold-change values above the cut-off: 4.5-fold for abacavir; 3.5-fold for lamivudine; 1.9-fold for zidovudine; 1.7-fold for didanosine, stavudine, and zalcitabine; and 1.4-fold for tenofovir. Cut-offs were defined on the basis of virological outcome data from clinical trials (abacavir, lamivudine, didanosine, stavudine, and tenofovir) [^14,15], or from the natural variability of wild-type clinical isolates and assay reproducibility (zidovudine and zalcitabine) [¹⁴]. For quantitative analyses, samples with lamivudine IC₅₀ above the highest drug concentration tested in the assay were assigned an arbitrary fold-change value of 200.

A database of approximately 27 000 clinical samples was queried to identify samples containing specific RT mutation profiles that included a single TAM with or without a mutation at codon 184 (M184V or M184I), or M184V/I alone. TAM were defined as M41L, D67N, K70R, L210W, T215Y/F or K219E/H/N/Q/R. Samples were excluded if they had more than one TAM, mixtures at the TAM in question or at M184I/V, or any other NRTI resistance-associated mutation (E44D, K65R, T69D/N, T69 insertions, L74I/V, V75T/M/S/A, Y115F, V118I, Q151M). Samples were not excluded if non-nucleoside reverse transcription inhibitor mutations or protease inhibitor mutations were present, and could include Y181C or L100I, which were found to have little effect on the median fold-change (MFC) results. Samples were divided into those with M184 or with M184I/V in addition to the specific TAM, and the MFC in phenotypic resistance was calculated for the NRTI. For comparison, 500 samples lacking any TAM or NRTI resistance- associated mutation were selected randomly from the database. The statistical significance of subtle differences in MFC were tested using the Mann-Whitney non-parametric test using Prism 3.0 software (GraphPad Inc., San Diego, CA, USA).

The resistance algorithm comparisons for the specific mutations were run using the Stanford HIValg Resistance Algorithm Comparison version 3.1.1 [¹⁶] (http://hivdb2.stanford.edu/asi) except for the GeneSeq results, which were supplied by ViroLogic Inc (algorithm updated May 2003), and the VGI results, which used rules version 6.0 (downloadable from http://www.trugene.com/Physicians/GuideLines_6.0_Rules.pdf). The algorithm specification interface (ASI) document type definition that the Stanford HIValg adhere to can be viewed on the website and source code downloaded from that site. The proprietary GeneSeq algorithm counts mutations equally without assigning a penalty to specific mutations.

Results

In the absence of TAM or NRTI resistance-associated mutations, M184I/V containing samples in the ViroLogic database (n = 2451) had MFC values indicating susceptibility to all NRTI except for lamivudine (> 200). The fold-changes with M184V or I (M184I/V) were 2.8, 1.4, 0.7, 0.4, and 0.5-fold for abacavir, didanosine, stavudine, zidovudine and tenofovir, respectively (Table 1).

Table 1

The number of samples containing a single TAM, with or without M184I/VI, and the associated NRTI MFC are summarized in Table 1. The presence of a single TAM without M184V or I was associated with MFC below the cut-offs for all NRTI, except for zidovudine in the presence of K70R (2.2-fold), T215Y (7.6-fold), or T215F (5.7-fold). The T215Y and F mutations, without M184V/I, were also associated with increases in MFC for stavudine and tenofovir (1.4-1.5 and 1.2-1.3-fold, respectively; Mann-Whitney P < 0.0001 versus wild-type HIV-1), which approached, but did not exceed, the cut-offs for reduced susceptibility to these drugs (1.7 and 1.4-fold, respectively).

The presence of the M184I/V mutation with any TAM was associated with enhanced susceptibility to zidovudine, stavudine or tenofovir compared with viruses without M184I/V or I (Mann-Whitney P < 0.01). The MFC for all NRTI except lamivudine was below the cut-off for samples with mutations at positions 41, 67, 70, and 219 in combination with M184I/V, as well as for all NRTI except zidovudine and zalcitabine for samples with T215F/Y plus M184I/V. The M184I/V mutation with any TAM was associated with a higher fold-change value for abacavir, didanosine, lamivudine and zalcitabine compared with viruses without the M184I/V mutation. However, M184I/V was only associated with MFC above the assay cut-off for lamivudine (> 300-fold) in the presence of any TAM, and zalcitabine (1.7-1.8-fold) in the presence of T215F/Y. Susceptibility to didanosine and abacavir in the presence of the M184I/V mutation and some TAM (e.g. T215F or Y) approached, but did not exceed, the assay cut-off.

For several groups there was considerable variability in fold-change not accounted for by the NRTI resistance-associated mutations used in the query definition and beyond the variation expected based on assay reproducibility [¹⁴]. The 5th and 95th percentiles are presented in Table 1 to illustrate this. Therefore, whereas the MFC may be above or below the cut-off for a given group, there are often exceptions to the rule.

Table 2 summarizes the results obtained when different genotypic interpretation algorithms were compared for the same mutation patterns. No specific mutation had the same predicted susceptibility pattern for all NRTI in every algorithm; however, for most mutation patterns, there was concordance among a majority of the algorithms. These results are consistent with other studies reporting discordance between genotype algorithms [¹⁷] and between genotype and phenotype [¹⁸].

Table 2

The majority of the algorithms predicted decreased susceptibility to zidovudine for every TAM pattern, although the phenotypic results demonstrated decreased zidovudine susceptibility only in the presence of the K70R, T215Y and T215F mutations. The presence of the M184I/V mutation in combination with any single TAM increased phenotypic susceptibility for zidovudine such that the MFC was below the cut-off for reduced susceptibility, except in combination with T215F or Y. This concurred with predictions by the majority of the algorithms; however, the RCG and the VGI algorithms predicted reduced susceptibility for zidovudine in the presence of any TAM, and the GeneSeq algorithm predicted reduced susceptibility to zidovudine in the presence of K70R, T215Y or T215F, even in the presence of M184I/V.

The majority of the algorithms also predicted decreased susceptibility to abacavir in the presence of the M184I/V mutation and any single TAM. However, only the T215Y plus the M184I/V mutation pattern was associated with a MFC above the clinically defined phenotypic assay cut-off. Similar patterns of over-predictions of reduced susceptibility were also observed for didanosine and zalcitabine, in which phenotypic results showed decreased susceptibility only for zalcitabine in the presence of M184V/I in combination with K70R, T215Y or T215F.

Discussion

The goal of HIV highly active antiretroviral therapy is to reduce morbidity and mortality associated with HIV infection. Suppressing viral load can prevent the selection of drug resistance, preserve therapy options, and maintain quality of life. However, in some studies a residual clinical benefit has been noted for antiretroviral regimens even when resistance-associated viremia is present [^19-21]. This residual benefit may be caused by multiple factors. In-vitro and in-vivo studies on the M184V mutation have documented the reductions in processivity of the RT and loss of viral fitness [^22-27], or replicative capacity [⁵], especially when the M184V mutation is present along with TAM [²⁸], delay the reversal of zidovudine/stavudine resistance [^6-8] and a reduction in RT pyrophosphorolysis [⁴].

In HIV isolates with any single TAM, the presence of the M184I/V mutation can re-sensitize or enhance susceptibility to zidovudine, stavudine and tenofovir. Furthermore, hypersusceptibility (defined as a fold-change < 0.4-fold) was observed for zidovudine in the presence of K219Q/E/N/R plus M184I/V. For zidovudine, in samples with the K70R, T215Y, or T215F mutation and no M184I/V mutation, the MFC were above the assay cut-off for reduced susceptibility, and for K70R the presence of M184I/V reduced the MFC to well below the 1.9-fold cut-off.

In the absence of the M184I/V mutation, the MFC for abacavir, didanosine, lamivudine and zalcitabine in the presence of any of single TAM was below the assay cut-off, and usually very close to the susceptibility of the wild-type reference virus. For abacavir, didanosine, and zalcitabine, the presence of the M184I/V mutation plus a TAM increased the MFC over those observed in isolates with a single TAM alone. However, increases were generally below cut-offs for reduced susceptibility to each drug except for zalcitabine, which was slightly above the cut-off when T215Y or T215F were present. The differing effects of the M184I/V mutation on the two groups of NRTI has previously been reported [²⁹].

Not all TAM had the same impact on drug susceptibility. For stavudine, zidovudine and tenofovir, isolates with the M41L, D67N, L210W or K219Q/E/N/R mutations have MFC below their respective cut-offs. Conversely, T215Y and T215F are clearly associated with slight decreases in susceptibility to these drugs. It should be noted that the specific genetic background and different combination of HIV resistance mutations could affect the phenotypic as well as the clinical outcome for an individual patient. Therefore, clinical studies are necessary to associate virological outcomes with baseline phenotypes.

These results indicate that individual TAM do not have an equivalent impact on NRTI resistance, and that in the presence of the M184I/V mutation, re-sensitization to some drugs is observed despite the presence of a TAM. Therefore, retaining lamivudine in treatment regimens in which TAM can be selected may provide therapeutic benefit by maintaining the M184I/V mutation. Discordance was noted between the phenotypic cut-offs and the currently available algorithms for predicting drug susceptibility for specific mutation patterns. The data from the current study provide additional results for future versions of genotypic algorithms.