Background: The major obstacles to the CCR5-antagonist maraviroc being more widely used in anti-HIV-1 therapy regimens are (i) traditional pre-treatment phenotypic tropism tests to determine virus susceptibility to maraviroc are expensive and time consuming, and (ii) cheaper and more rapid genotypic tropism tests have been developed primarily for HIV-1 subtype B strains, which account for only 10% of infections worldwide. We therefore developed PhenoSeq, a suite of reliable genotypic tropism tests specific for HIV-1 subtypes A, B, C, D and circulating recombinant forms of subtypes AE (CRF01_AE) and AG (CRF02_AG), which together account for 95% of infections worldwide.
Methods: Development of subtype-specific genotypic tropism tests was informed by analysis of all HIV-1 env third variable region (V3) amino acid sequences from the Los Alamos HIV Database that had corresponding phenotypic tropism and HIV-1 subtype data (n=2257; 630 CXCR4-using and 1637 CCR5-using), to elucidate statistically significant sequence alterations that distinguish CXCR4- from CCR5-using viruses. With the aid of a new bioinformatic tool that we developed (bulk2clonal), PhenoSeq algorithms were validated against independent clinical HIV-1 sequence panels from patients previously enrolled in the Pfizer trials A4001064 and MERIT, relative to results of the original Trofile assay (OTA) and the enhanced sensitivity Trofile assay (ESTA).
Results: Our analyses show that the PhenoSeq genotypic algorithms exhibit more favourable sensitivity and specificity profiles for establishing the tropism of HIV-1 subtypes A, B, C, D, CRF01_AE and CRF02_AG than current alternative algorithms, including the clinically validated in-use algorithms geno2pheno (g2p; false positive rate 5.75% and 10%) and WebPSSM (Table 1). Furthermore, PhenoSeq area under the receiver operator characteristic curves (AUROC) were significantly greater than g2p FPR 5.75% for establishing the tropism of HIV-1 subtype C (p=0.01), g2p FPR 10% for CRF01_AE (p=0.03), and WebPSSM for subtypes B (p=0.01) and D (p=0.05) (two-tailed t-test, p≤0.05 considered significant).
Conclusions: As the only platform of algorithms that can reliably infer the tropism of the major global HIV-1 subtypes, PhenoSeq has the potential to inform the appropriate use of maraviroc and future CCR5 blocking drugs in regions of the world where non-B HIV-1 predominates, and which are burdened the most by the HIV-1 pandemic.
| PhenoSeq | g2p FPR 5.75% | g2p FPR 10% | WebPSSMX4R5 | WebPSSMSI/NSI | ||
| B-HIV Test Set 1 ⇥⇥⇥OTA ⇥⇥⇥12 CXCR4-using, 41 R5 |
Sens/Spec | 100/87.8 | 100/97.6 | 100/90.2 | 41.7/95.1 | 16.7/87.8 |
| AUROC | 0.94 | 0.99 (p=0.17) | 0.95 (p=0.44) | 0.68 (p=0.01) | 0.52 (p=0.01) | |
| B-HIV Test Set 2 ⇥⇥⇥LANL ⇥⇥⇥92 CXCR4-using, 269 R5 |
Sens/Spec | 78.4/80.3 | 70.7/92.2 | 72.8/84 | 55.4/96.3 | 62/94.1 |
| AUROC | 0.79 | 0.82 (p=0.32) | 0.78 (p=0.41) | 0.76 (p=0.24) | 0.78 (p=0.41) | |
| C-HIV Test Set 1 ⇥⇥⇥OTA ⇥⇥⇥55 CXCR4-using, 40 R5 |
Sens/Spec | 83.6/92.5 | 78.2/95 | 81.8/95 | – | 85.5/77.5 |
| AUROC | 0.88 | 0.87 (p=0.22) | 0.88 (p=0.30) | – | 0.82 (p=0.22) | |
| C-HIV Test Set 2 ⇥⇥⇥ESTA ⇥⇥⇥18 CXCR4-using, 187 R5 |
Sens/Spec | 77.8/75.4 | 11.1/93 | 50/82.4 | – | 61.1/81.3 |
| AUROC | 0.77 | 0.52 (p=0.01) | 0.66 (p=0.15) | – | 0.71 (p=0.29) | |
| D-HIV Test Set ⇥⇥⇥OTA ⇥⇥⇥43 CXCR4-using, 44 R5 |
Sens/Spec | 80.5/77.3 | 81.4/72.7 | 88.4/61.4 | 81.4/65.9 | 69.8/63.6 |
| AUROC | 0.79 | 0.77 (p=0.40) | 0.75 (p=0.29) | 0.74 (p=0.24) | 0.67 (p=0.05) | |
| CRF01_AE Test Set ⇥⇥⇥OTA ⇥⇥⇥14 CXCR4-using, 25 R5 |
Sens/Spec | 85.7/96 | 78.6/68 | 85.7/56 | 85.7/72 | 85.7/76 |
| AUROC | 0.91 | 0.73 (p=0.046) | 0.71 (p=0.03) | 0.79 (p=0.11) | 0.81 (p=0.15) | |
| CRF02_AG Test Set 1 ⇥⇥⇥OTA ⇥⇥⇥9 CXCR4-using, 69 R5 |
Sens/Spec | 88.9/76.8 | 44.4/97.1 | 55.6/94.2 | 44.4/87 | 66.7/84 |
| AUROC | 0.83 | 0.71 (p=0.18) | 0.75 (p=0.27) | 0.66 (p=0.10) | 0.75 (p=0.28) | |
| CRF02_AG Test Set 2 ⇥⇥⇥OTA ⇥⇥⇥8 CXCR4-using, 36 R5 |
Sens/Spec | 62.5/97.2 | 50/94.4 | 62.5/80.6 | 37.5/88.9 | 37.5/91.7 |
| AUROC | 0.80 | 0.72 (p=0.30) | 0.72 (p=0.29) | 0.63 (p=0.14) | 0.65 (p=0.16) | |
Sens, % sensitivity was calculated by dividing the number of correctly predicted CXCR4-using sequences by the total number CXCR4-using sequences and multiplying by 100. Spec, % specificity was calculated by dividing the number of correctly predicted R5 sequences by the total number of R5 sequences and multiplying by 100. P-values (two-tailed) for comparisons of area under the receiver operator characteristic curve (AUROC) <0.05 were considered significant and are highlighted in bold text. g2p, geno2pheno. OTA, original Trofile assay. ESTA, enhanced sensitivity Trofile assay. LANL, Los Alamos HIV Database. FPR, false positive rate. *The subtype C specific WebPSSMSI/NSI algorithm was used for subtype C HIV-1 predictions.