HIV-1 genotypic drug-resistance interpretation algorithms need to include hypersusceptibility-associated mutations.

Kai Wang,R. Samudrala,J. Mittler

DOI: https://doi.org/10.1086/423489

2004-12-01

Abstract:To the Editor—Tozzi et al. [1] recently reported a retrospective study investigating the effect that mutations associated with hypersusceptibility to nonnucleoside reverse-transcriptase inhibitors (NNRTIs) have on the virological response to efa-virenz-based therapies. The authors concluded that " the M41L, M184V, L210W, and T215Y mutations were associated with a better, although transient, virolog-ical outcome in patients treated with efa-virenz-based regimens " (p. 1688). Hypersusceptibility is certainly a very interesting concept, and we have checked publicly available HIV-1 genotypic drug-resistance interpretation algorithms for information on hypersusceptibility. The most recent mutations report (October 2003 revision) from the International AIDS Society–USA Drug Resistance Mutations Group [2] includes a footnote that describes M184V as a mutation that is possibly associated with hypersusceptibil-ity to some nucleoside reverse-transcrip-tase inhibitors, but the report does not include this or any other hypersuscepti-bility-associated mutations in its main table. Using the algorithm-specific interface of the Stanford Drug Resistance Database [3], we could not find a " hypersuscepti-ble " category in the rules for either the ANRS algorithm (version 2003.10) or the Rega algorithm (version 6.2). The Stan-ford HIVdb algorithm (version 2004.4) does contain a few hypersusceptibility annotations , but none of the 4 mutations reported by Tozzi et al. is listed as conferring hypersusceptibility to efavirenz or to other NNRTIs. The geno2pheno web server [4] implements 2 machine-learning algorithms (decision tree and support vector machine) for HIV-1 genotypic drug-resistance interpretation, but neither algorithm can provide predictions for hy-persusceptibility. We recently developed a linear statistical model for HIV-1 genotyp-ic drug-resistance interpretation [5]. Our model for efavirenz suggests that the presence of M41L or T215Y can potentially decease IC 50 fold-change values by 38.7% and 33.0%, respectively. Our ge-notypic algorithm actually predicts the existence of at least 1 hypersusceptibility-associated mutation for all 17 protease inhibitors (PIs) and reverse-transcriptase inhibitors used in our study. Included in our models are the predictions that M41L and T215Y cause hypersusceptibility to delavirdine and that T215Y causes hy-persusceptibility to nevirapine. In surveying the literature, we have come across several articles reporting on mutations associated with hypersuscepti-bility to NNRTIs. Shulman et al. [6] observed a positive correlation between phe-notypic hypersusceptibility to efavirenz and virological responses to regimens that include efavirenz; also, using a multivar-iate analysis similar to our linear model, they found 215Y/F to be associated with hypersusceptibility. Whitcomb et al. [7] analyzed genotypic, phenotypic, and clinical data on mutations associated with hy-persusceptibility to NNRTIs and demonstrated that " …