Comment on: High sensitivity of specific genotypic tools for detection of X4 variants in antiretroviral-experienced patients suitable to be treated with CCR5 antagonists

Sir, We read with interest the article by Seclén et al., on the reliability of genotypic tools for detecting CXCR4-using variants in patients at different stages of HIV-1 infection. Seclén et al. reported that the concordance between the Phenoscript tropism assay and genotypic methods based on plasma virus RNA ranged from 63% to 85%. The best predictors for detecting X4 variants were WebPSSMX4/R5 (sensitivity 77%, specificity 87%) and Geno2phenoFPR1⁄45% (sensitivity 80%, specificity 77%). Seclén et al. also used two simple rules combining the ‘11/25’ and ‘net charge’ rules and termed Garrido’s rule and Delobel’s rule. Delobel’s rule was first derived from clonal analyses of genotype– phenotype correlations. We validated this rule using a large set of clinical data. According to Delobel’s rule, one of the following criteria is required for predicting CXCR4 co-receptor usage: (i) 11 R/K and/or 25K; (ii) 25R and a net charge of ≥+5; or (iii) a net charge of ≥+6. In the article by Seclén et al., Delobel’s rule was ‘a sample is labelled as X4 if a basic amino acid (R or K) is recognized at positions 11/25 within the V3 region AND if the global net charge of the V3 sequence is ≥5, otherwise it is R5’. Thus, Delobel’s rule was not applied adequately. This mistake probably had a great impact on the prediction of co-receptor usage and on the concordance between genotype and phenotype. In fact, Delobel’s rule is closer, although not identical, to Garrido’s rule giving 80% sensitivity and 79% specificity in this study for the global population. Our previous study on antiretroviral-experienced patients showed that Delobel’s rule was 77% sensitive and 96% specific. We also genotyped naive patients, for whom Delobel’s rule was 63% sensitive and 97% specific. This is very different from the performance reported in the Seclén et al. study (sensitivity 20%). These differences in algorithm performance, whatever the stage of the HIV-1 infection, can be explained by the incorrect use of Delobel’s rule. Seclén et al. found that the algorithms were poorly sensitive for detecting CXCR4-using variants in non-B subtypes. Unfortunately, this analysis was global due to the small number of each subtype. In contrast, we used a panel of CRF-02 strains to demonstrate the good performance of Delobel’s rule for predicting CXCR4-using viruses; the sensitivity was 70% and the specificity was 98%. Using the same dataset, the WebPSSMX4/R5 was 80% sensitive, but less specific (76%). In another study, Delobel’s rule predicted the subtype C CXCR4-using viruses with a sensitivity of 93% and a specificity of 96%, and the subtype C WebPSSMSI/NSI was 82% sensitive and 93% specific. 5 Further studies are needed to evaluate the performances of genotypic algorithms for specific subtypes. Finally, we agree that genotypic methods are suitable for determining HIV-1 co-receptor usage in both plasma and peripheral blood mononuclear cells before treatment with CCR5 antagonists.