Background: The development of HIV co-receptor inhibitors demands suitable assays for assessing co-receptor usage from clinically derived isolates.  Two recombinant phenotypic assays for co-receptor usage and a genotypic analysis of the HIV V3 loop were employed on a set of clinically derived HIV-1 samples in order to evaluate concordance between measures.

Methods: Aliquots of previously genotyped HIV-1 samples (or PCR products) derived from antiretroviral naïve individuals were tested for co-receptor usage using two independent phenotype methods, yielding 74 values which could be compared. HIV co-receptor phenotypes were determined by previously validated recombinant assays which incorporate either ~2500 bp (Monogram/Virologic assay) or ~900 bp (Mammano lab/TRT assay) fragments of the HIV envelope. HIV envelope V3 loop sequences (~105 bp) were derived by standard automated sequence analysis. Genotypic predictions were performed using a support vector machine (SVM) analysis of the HIV V3 loop (www.geno2pheno.org) trained on 1110 matched clonal genotype-phenotype pairs from the Los Alamos HIV database.

Results: HIV co-receptor usage was obtained from both phenotypic assays for 74 samples, either as R5 (N=50 or 45 from Virologic or TRT, respectively or {X4 or R5/X4}(N=24 or 29, respectively), with an overall 86.3% concurrence. There was no clear evidence of a difference in sensitivity between the two phenotypic assays, as samples were identified as X4/R5 in one assay but not the other in similar number.  Furthermore, employing different X4 luciferase cut-off values did not increase the degree of agreement between the different phenotype assays. Using only the presence of basic amino acid residues at codons 11 and 25 of the V3 loop resulted in very poor sensitivity for detecting X4 virus, but a bioinformatic algorithm based on a Support Vector Machine using entire HIV V3 genotype data was able to achieve an 82.1% and 71.2% correlation with the Virologic and TRT assays, respectively, approaching the degree of agreement between the two phenotype assays. The presence of minority species of X4 virus in clinical samples likely complicates the interpretation of both phenotypic and genotypic assessments of HIV tropism.

Conclusion: In most cases, the two phenotype assays and the bioinfomatic approach gave similar results.  However, in cases where there are differences in tropism results, it is not clear which of these assays is “correct”.