Performance of a Population-Based HIV-1 Tropism Phenotypic Assay and Correlation With V3 Genotypic Prediction Tools in Recent HIV-1 Seroconverters

Abstract

Pure X4 and X4R5 dual-tropic viruses may be recognized in approximately 15% of drug-naive HIV-1-positive patients. CCR5 antagonists are active against R5 viruses; therefore, HIV tropism should be known before their prescription. A population-based phenotypic assay was performed in 61 recent HIV-1 seroconverters. The results were compared with those obtained using 8 different predictor software programs (C4.5, C4.5 with 8 and 12, PART, SVM, Charge Rule, PSSMsinsi, PSSMx4r5, and geno2pheno), which are freely available at 3 different Web sites and use V3 sequences derived from patient's viruses. Phenotypic testing reported X4R5 dual-tropic viruses in 10 (16.4%) patients. CD4 cell counts and viral loads were significantly lower in X4R5 dual-tropic (450 cells/microL and 3.9 log HIV RNA copies/mL) than in R5 viruses (629 cells/microL, 4.5 log HIV RNA copies/mL) (P<0.05). The overall concordance of genotype and phenotype was relatively good (>80%). Although specificity was >90% using all but 1 genotypic predictor (geno2pheno), however, the sensitivity for the detection of X4 variants was low (<30%), except for SVM and geno2pheno (70%). The prevalence of X4 and X4/R5 dual-tropic viruses in recent HIV seroconverters is 16%. Current genotypic algorithms need to be improved for the estimation of HIV-1 coreceptor use before moving to the clinic. This information is crucial for the selection of candidates to receive CCR5 antagonists in places where phenotypic tropism assays may not be feasible.