Skip Navigation LinksHome > July 11, 2007 - Volume 21 - Issue 11 > Correlation between a phenotypic assay and three bioinformat...
AIDS:
doi: 10.1097/QAD.0b013e32826fb741
Research Letters

Correlation between a phenotypic assay and three bioinformatic tools for determining HIV co-receptor use

Poveda, Evaa; Briz, Verónicaa; Roulet, Vanessab,c; del Mar González, Maríaa; Faudon, Jean-Louisb; Skrabal, Katharinab,c; Soriano, Vincenta

Free Access
Article Outline
Collapse Box

Author Information

aDepartment of Infectious Diseases, Hospital Carlos III, Madrid, Spain

bEurofins Viralliance, Kalamazoo, Michigan, USA

cBioalliance Pharma, Paris, France.

Received 5 March, 2007

Revised 18 April, 2007

Accepted 22 April, 2007

Collapse Box

Abstract

The predictive value of three genotypic methods to determine HIV-1 co-receptor usage was assessed in 83 plasma specimens taking as reference the results obtained using a recombinant phenotypic assay (Phenoscript). The best concordance was found for webPSSM, followed by geno2pheno and wetcat (85.9, 71.8 and 70.5%, respectively). Less than 5.1% of phenotypic X4 viruses were missed by genotypic tools. The genotypic prediction of HIV-1 co-receptor usage can thus assist therapeutic decisions for using CCR5 antagonists.

CCR5 and CXCR4 are the main co-receptors used by HIV-1 to enter into the host cells [1–3]. Several co-receptor antagonists, designed to block the interaction between the HIV-1 gp120 with either CCR5 or CXCR4 receptors, are under clinical development [4,5]. Maraviroc and vicriviroc are CCR5 antagonists currently in the latest steps of clinical development. Given their mechanism of action, viral tropism should be assessed before their prescription and eventually during treatment in order to exclude X4 viruses at baseline or switches on therapy, respectively [6,7]. In this scenario, the availability of rapid and reliable tools for assessing co-receptor tropism in clinical sites is warranted.

There are several methods to test viral tropism, being the most appropriate under debate [8–10]. Phenotypic assays using either primary isolates or recombinant viruses are considered to be the most reliable; however, they are sophisticated, expensive and require special facilities and expertise. Bioinformatic tools based on the interpretation of genotypic sequences derived from the V3 envelope region of the virus have been developed in recent years to predict HIV-1 co-receptor use [11]. At this time their accuracy and concordance with respect to phenotypic results is unclear.

HIV-1 co-receptor usage was examined in plasma samples collected from 83 HIV-infected patients using the Phenoscript Env Assay (Eurofins, Kalamazoo, Michigan, USA) [12], which is based on a recombinant virus technology. Briefly, amplification of viral envelope sequences derived from plasma is initially carried out, followed by homologous recombination between the env amplicons and a deleted NL43 plasmid of the corresponding region during co-transfection of producer cells. Recombinant viruses are used to infect indicator cells expressing CCR5 or CXCR4, in addition to CD4, and carrying the lacZ gene under control of the viral long-terminal repeat. The specificity of infection is assessed in each assay by incubation of indicator cells in the presence and absence of co-receptor inhibitors, and five controls carrying patients' derived env sequences of different tropism and infectivity. The Phenoscript Env assay has been validated for its sensitivity for the detection of minority species, as well as its specificity for samples with different viral loads and distinct HIV-1 non-B subtypes [13].

Three genotypic/bioinformatic tools built on the interpretation of viral envelope sequences were evaluated. All three methods infer HIV-1 co-receptor usage based on the nucleotide or amino acid sequences of the V3 region obtained after amplification from plasma HIV RNA by nested polymerase chain reaction (PCR), using primers E80 (5′–CCA ATT CCC ATA CAT TAT TGT G–3′) and E105 (5′–GCT TTT CCT ACT TCC TGC CAC–3′) as outer primers, and ES7 (5′–CTG TTA AAT GGC AGT CTA GC–3′) and E125 (5′–CAA TTT CTG GGT CCC CTC CTG AGG–3′) as inner primers. Primer sequences were kindly provided by M. Quinones-Mateu (Life Science Division, Diagnostic Hybrids, Case Western Reserve University, Cleveland, Ohio, USA). PCR amplicons were purified using the High Pure PCR Product Purification Kit (Roche, Mannheim, Germany) and were directly sequenced with primers ES7 and E125 using the ABI PRISM dRhodamine Terminator Cycle Sequencing Kit (Applied Biosystems, Foster City, California, USA) following manufacturer's instructions. All sequences generated were edited using the SeqScape v2.5 software (Applied Biosystems). Nucleotide mixtures were considered when the second highest peak in the electropherogram was above 25%. Nucleotide triplets containing mixtures were translated into each of the possible amino acids. For each sample, all different V3 sequences generated as a consequence of all possible amino acid combinations were considered for the tropism analysis. Co-receptor usage was determined for each sample using three genotypic predictor softwares freely available online: wetcat (http://genomiac2.ucsd.edu:8080/wetcat/v3.html) [14], geno2pheno (http://coreceptor.bioinf.mpi-sb.mpg.de/cgi-bin/coreceptor.pl) [15], and webPSSM (position specific scoring matrices) (http://ubik.microbiol.washington.edu/computing/pssm/) [16]. HIV-1 variants were classified as R5 or X4 tropic, including the latter both pure and dual/mixed R5/X4 viruses.

All patients were infected with HIV-1 subtype B strains and were on regular follow-up at our institution. Forty-six patients (55.4%) were heavily antiretroviral experienced and the other 37 were drug naive. As expected, the mean CD4 T-cell count was significantly higher in drug-naive than in antiretroviral-experienced patients (437 ± 361 versus 246.6 ± 204.6 cells/μl; P < 0.01). The mean plasma HIV-RNA level was similar in both groups (3.9 ± 0.8 versus 4.1 ± 0.6 log copies/ml, respectively).

The results of co-receptor usage in this set of clinical samples are depicted in Table 1. Whereas genotypic amplicons could be obtained from all samples, phenotypic testing could only be performed in 78 out of 83 specimens (93.9%), because the lack of enough plasma volume precluded testing of the remaining specimens. The highest degree of concordance with the phenotypic test results was found for webPSSM (85.9%) followed by geno2pheno (71.8%) and wetcat (70.5%). Discordance between phenotypic and genotypic methods were mainly the result of an overestimation of X4 viruses by genotypic tools. This disagreement was seen in 26.9, 24 and 8.9% of tested samples using wetcat, geno2pheno and webPSSM, respectively. Conversely, genotypic prediction of R5 by bioinformatic tools in samples reported to be X4 by phenotype was seen in 2.5, 3.8 and 5.1% of patients using wetcat, geno2pheno and webPSSM softwares, respectively. The sensitivity and specificity for the detection of R5 viruses for wetcat, geno2pheno and webPSSM was as follows: 64.4 and 89.5%, 67.8 and 88.8%, and 88.1 and 78.9%, respectively.

Table 1
Table 1
Image Tools

The discordance found between genotypic and phenotypic methods did not seem to be explained by differences in patient characteristics, such as viral load, CD4 cell counts or treatment status (data not shown). Although the HIV-1 envelope V3 region is the main determinant for co-receptor usage [17], several studies have shown that other env hypervariable regions (V1, V2, V4 and V5) as well as the entire C1–C4 region may also influence HIV tropism [17–21]. Whereas genotypic methods based exclusively on V3 sequences may be a reliable indicator of HIV-1 tropism, it is obvious that phenotypic assays such as the Phenoscript Env Assay, which includes the whole gp120 and the ectodomain of gp41 of the viral envelope [12], may be more accurate to assess viral tropism.

The different results obtained with the three genotypic methods could be a result of technical details such as the different statistical models used for predictions. Whereas wetcat and geno2pheno use support vector machine, webPSSM predicts CXCR4 usage on the basis of the PSSM model. Support vector machine is a statistical tool that offers a binary classification of the sequences (R5 or X4); thus R5/X4 sequences are considered as X4 by wetcat. Likewise, they are included in the two models used by geno2pheno, one for the detection of CCR5 and another to recognize CXCR4 [14,15]. By contrast, PSSM interprets a propensity to use CXCR4 given low, intermediate, and high scores for R5, R5/X4 and X4 viruses, respectively [16]. Accordingly, PSSM offers a higher specificity for the detection of X4 viruses than the other two bioinformatics tools.

As a result of the lower limit of detection of the Phenoscript Env Assay for the identification of R5 or X4 viruses in a mixture (which is approximately 5% in a mixture of primary viruses with comparable infectivity) [13], we can assume that specimens reported to be R5 must harbour less than 5–10% of X4 viruses. As disagreements between genotypes and phenotype were mainly represented by overestimates of X4 viruses by bioinformatic tools, however, it seems unlikely that X4 viruses will be missed with these methods (overall less than 5.1% in our study). Moreover, the overall agreement between geno–pheno methods to predict CCR5 usage was 96.2%. Therefore, current bioinformatics tools, although imperfect, may be useful to assess the opportunity of using CCR5 antagonists on clinical sites when phenotypic assays are not available.

Sponsorship: This work was partly funded by grants from Fundación Investigación y Educación en SIDA (IES) and Red de Investigación en SIDA (RIS, ISCIII–RETIC RD06).

Back to Top | Article Outline

References

1. Feng Y, Broder C, Kennedy P, Berger E. HIV-1 entry cofactor: functional cDNA cloning of a seven-transmembrane G protein-coupled receptor. Science 1996; 272:872–877.

2. Dragic T, Litwin V, Allaway G, Martin S, Huang Y, Nagashima K, et al. HIV-1 entry into CD4+ cells is mediated by the chemokine receptor CC-CKR-5. Nature 1996; 381:667–673.

3. Deng H, Liu R, Ellmeier W, Choe S, Unutmaz D, Burkhart M, et al. Identification of a major co-receptor for primary isolates of HIV-1. Nature 1996; 381:661–666.

4. Seibert C, Sakmar T. Small-molecule antagonist of CCR5 and CXCR4: a promising new class of anti-HIV drugs. Curr Opin Pharmacol 2004; 4:447–452.

5. Briz V, Poveda E, Soriano V. HIV entry inhibitors: mechanisms of action and resistance pathways. J Antimicrob Chemother 2006; 57:619–627.

6. Westby M, Lewis M, Whitcomb J, Youle M, Pozniak A, James I, et al. Emergence of CXCR4-using HIV-1 variants in a minority of HIV-1-infected patients following treatment with the CCR5 antagonist Maraviroc is from a pretreatment CXCR4-using virus reservoir. J Virol 2006; 80:4909–4920.

7. Mayer H, Van der Ryst E, Saag M, Clotet B, Fatkenheuer G, Clumeck N, et al. Safety and efficacy of Maraviroc, a novel CCR5 antagonist, when used in combination with optimized background therapy for the treatment of antiretroviral-experienced subjects infected with dual/mixed-tropic HIV-1: 24-week results of a phase 2b exploratory trial. In: XVIth International AIDS Conference. Toronto, Canada 13–18 August 2006; Abstract THLB0215.

8. Poveda E, Briz V, Quinones-Mateu M, Soriano V. HIV tropism: diagnostic tools and implications for disease progression and treatment with entry inhibitors. AIDS 2006; 20:1359–1367.

9. Weber J, Piontkivska H, Quinones-Mateu M. HIV type 1 tropism and inhibitors of viral entry: clinical implications. AIDS Rev 2006; 8:60–77.

10. Skrabal K, Low A, Dong W, Sing T, Cheung P, Mammano F, Harrigan P. Determining HIV co-receptor use in a clinical setting: degree of correlation between two phenotypic assays and a bioinformatic model. J Clin Microbiol 2007; 45:279–284.

11. Jensen M, Van't Wout A. Predicting HIV-1 coreceptor usage with sequence analysis. AIDS Rev 2003; 19:145–149.

12. Labernardiere J, Lebel-Binay S, Faudon JL, Holguin A, Soriano V, Cheret A. Tropism determination and performance of Phenoscript HIV-1 entry inhibitors assay. Antivir Ther 2004; 9:S141.

13. Roulet V, Rochas S, Labernardiere J, Mammano F, Faudon J, Raja N, et al. HIV-1 Phenoscript®ENV: a sensitive assay for the detection of HIV minority species and determination of B and non-B subtypes viral tropism. In: 5th European HIV Drug Resistance Workshop: Cascais, Portugal 28–30 March 2007; Abstract 59.

14. Pillai S, Good B, Richman D, Corbeil J. A new perspective on V3 phenotype prediction. AIDS Res Hum Retroviruses 2003; 19:145–149.

15. Sing T, Beerenwinked N, Lengaver T. Learning mixtures of localized rules by maximizing the area under the ROC curve. In: 16th European Conference on Artificial Intelligence (ECAI) Workshop on ROC Analysis in AI: Valencia, Spain 22–27 August 2004.

16. Jensen M, Li F, van't Wont A, Nickle D, Shiner D, He H, et al. Improved co-receptor usage prediction and genotypic monitoring of R5 to X4 transition by motif analysis of HIV type 1 env V3 loop sequences. J Virol 2003; 77:13376–13388.

17. Hwang S, Boyle T, Lyerly H, Cullen B. Identification of the envelope V3 loop as the primary determinant of cell tropism in HIV-1. Science 1991; 253:71–74.

18. Andeweng A, Leeflang P, Osterhaus A, Bosch M. Both the V2 and V3 regions of the HIV-1 surface glycoprotein functionally interact with other envelope regions in syncytium formation. J Virol 1993; 67:3232–3239.

19. Boyd M, Simpson G, Cann A, Johnson M, Weiss R. A single amino acid substitution in the V1 loop of HIV-1 gp120 alters cellular tropism. J Virol 1993; 67:3649–3652.

20. Groenink M, Fouchier R, Broersen S, et al. Relation of phenotype evolution of HIV-1 to envelope V2 configuration. Science 1993; 260:1513–1516.

21. Hoffman N, Seillier-Moiseiwitsh F, Ahn J, Walker J, Swanstrom R. Variability in the HIV-1 gp120 env protein linked to phenotype-associated changes in the V3 loop. J Virol 2002; 76:3852–3864.

© 2007 Lippincott Williams & Wilkins, Inc.

Login