The recent advent of antiretroviral agents that belong to new drug classes is a major step in the fight against HIV, because a substantial number of patients have already selected for drug resistance to most currently available antiretroviral drugs. CCR5 antagonists block the binding of the virus to the cell surface CCR5 coreceptor, preventing HIV entry into the target cell.1,2 Most HIV-1 variants isolated from asymptomatic drug-naive chronically infected individuals use CCR5 along with CD4 to gain entry into the cells.3,4 HIV-1 variants able to use another coreceptor, CXCR4, emerge over the course of infection, however, reaching up to 40% to 50% in heavily pretreated patients with advanced HIV-1 disease.5 Given the lack of virologic response of patients with mixed/dual-tropic viruses, the determination of HIV-1 tropism is mandatory before the prescription of CCR5 antagonists.6,7
Phenotypic assays have been widely used for determining HIV-1 tropism. These tests are expensive and time-consuming and require specialized facilities and personnel, however. Bioinformatic tools based on genotypic sequences of the HIV-1 envelope gene (V3 sequences) involved in tropism have been examined, and predictive methods have been developed to estimate HIV-1 tropism. Genotypic tools should clearly be more accessible than phenotypic methods, but their reliability and sensitivity toward detecting minority species in the sequencing procedure remain major concerns.
MATERIAL AND METHODS
From a total of 424 recent HIV-1 seroconverters (<12 months since exposure) recorded at the national Spanish registry from January 1997 to June 2007, 61 randomly selected individuals were examined. The main characteristics of this population, including viral load, CD4 cell counts, gender, transmission route, length of infection, and primary drug resistance mutations (interpreted according to the latest International AIDS Society [IAS]-USA mutation list7), have been reported previously,8,9 and no significant differences were noticed between these 61 patients and the entire seroconverter cohort.
HIV Tropism Determination
A population-based experimental phenotypic assay was used to assess HIV-1 coreceptor use. Briefly, the assay consists of 1 reverse transcriptase (RT) polymerase chain reaction (PCR) step in which a large fragment of the HIV envelope glycoprotein gp120 gene (NH2-V4) and amplicons were cloned in a vector and transformed into Escherichia coli cells. Finally, recombinant viruses were cultured in U87-CD4 cells expressing CXCR4 or CCR5.10
Genotypic estimation of viral tropism was carried out in parallel in the same plasma specimens. V3 sequences were amplified by nested PCR using E80/E105 as outer primers and ES7/E125 as inner primers; primers and PCR conditions have been reported elsewhere.8,11 Sequences were analyzed using SeqScape v2.5 (Applied Biosystems, Foster City, CA), and all possible combinations of V3 amino-acid sequences present as a result of nucleotide mixtures were considered.
V3 sequences generated were then interpreted using 8 different bioinformatics genotypic predictors of tropism, accessed in September 2007, which are freely available at 3 Web sites: C4.5, C4.5 with 8 and 12, PART, SVM, and Charge Rule are available at http://genomiac2.ucsd.edu:8080/wetcat/v3.html12; PSSMX4R5 and PSSMsinsi are available at http://ubik.microbiol.washington.edu/computing/pssm/13,14; and geno2pheno is available at http://coreceptor.bioinf.mpi-sb.mpg.de/cgi-bin/coreceptor.pl.15 For Web PSSM, the tropism prediction was based on subtype B.14 For the geno2pheno Web site, the maximum sensitivity value for recognizing X4 variants was chosen for predictions of tropism using the specificity cutoff of 10% and a sensitivity of 80%. In all cases, HIV-1 variants were classified as R5 or X4 tropic, with the latter including pure X4 and dual/mixed X4R5 viruses.
All data were analyzed using the statistical software package SPSS 13.0 (SPSS Inc., Chicago, IL). Results were given as percentage or median values with their interquartile ranges (IQRs). Comparisons were performed using nonparametric or χ2 tests. The level of concordance was estimated using the Cochrane test. The sensitivity and specificity for the detection of X4 variants and their 95% confidence intervals (95% CIs) were calculated considering the phenotypic result as the “gold standard.” Sensitivity values were calculated as the proportion of samples that were considered as harboring X4 or X4R5 viruses by genotype within the whole group of X4 and X4R5 specimens informed by phenotype. Specificity values were reported as the proportion of patients with R5 viruses by genotype within all the R5 specimens informed by phenotype. All reported P values were 2-sided and were considered as significant only when P values were <0.05.
A total of 61 samples from recent HIV-1 seroconverters were examined. Their median age was 30 years old; 84% were men, most of whom had been infected through homosexual relationships. The median estimated time from exposure to the initial diagnosis of HIV infection was 7 months (IQR: 3 to 11 months). The mean viral load was 4.42 log HIV RNA copies/mL, and the mean CD4 count was 600 cells/μL. All but 4 (6.6%) subjects were infected by HIV-1 clade B viruses. The overall rate of primary drug resistance mutations, according to the latest IAS-USA mutation list, was 11.5% (7 of 61). By antiretroviral drug class, drug resistance was as follows: 5 (8.2%) for nucleoside reverse transcriptase inhibitors (NRTIs), 6 (9.8%) for nonnucleoside reverse transcriptase inhibitors (NNRTIs), and 3 (4.9%) for protease inhibitors (PIs). The most frequent substitution in the RT gene was at position K103N, which was found in 5 individuals. Other common changes were M41L (n = 3), D67N (n = 1), L74V (n = 1), V118I (n = 3), Y181C (n = 1), L210W (n = 1), and T215Y and revertants (n = 4). In the protease gene, the most common resistance mutation was V82A, which was found in 3 cases.
Phenotypic tropism testing reported X4R5 viruses in 10 (16.4%) seroconverters. There were significant differences in viral load and CD4 cell counts when comparing patients infected with R5 versus X4R5 viruses. Both parameters were significantly lower in patients with X4R5 dual/mixed-tropic viruses (Table 1). No differences were noticed in the estimated duration of infection, HIV subtype, or presence of drug resistance mutations acquired at the time of infection.
V3 sequences could be obtained from plasma collected from all patients (Genebank accession numbers EU424177 to EU424238). A total of 84 different amino-acid sequences were analyzed considering all nucleotide mixtures and possible amino-acid combinations. For those samples with more than 1 possible sequence that inferred different tropism, an X4R5 dual-tropic prediction was reported. The maximum number of possible combinations was 16, although the median number of amino-acid sequences per sample was 1 (IQR: 1 to 2 sequences). The rate of X4/X4R5 viruses using genotypic tools was as follows: 18% for SVM, 1.6% for C4.5, 1.7% for C4.5 with p8 to p12, 5.1% for the charge rule, 6.8% for PART, 32.7% for geno2pheno, and 5.8% for PSSM.
The overall concordance of genotypic tools with phenotypic results was as follows: 88.5% for SVM, 86.4% for C4.5, 86.4% for C4.5 with p8 to p12, 83% for the charge rule, 81.3% for PART, 71.2% for geno2pheno, and 82.7% for PSSM. No association between the number of possible amino-acid combinations for each sequence and the probability to predict X4 by any tool was found.
The specificity and sensitivity were calculated for each genotypic predictor. Overall, the specificity for the detection of X4 variants was good for all (>90%) except geno2pheno (72.5%). By contrast, the sensitivity for the detection of X4R5 dual-tropic variants was low. The highest sensitivity was obtained using SVM and geno2pheno (70%), whereas the remaining genotypic predictors had sensitivity values <30% (Fig. 1).
In this population of recent HIV-1 seroconverters, in which the phenotypic assay reported a prevalence of X4R5 viruses of 16.4%, most of the genotypic prediction tools tended to underestimate X4 viruses. Discordances were seen in any sense, however. For example, SVM reported 4 samples as X4 when the phenotype identified them as R5. Conversely, 3 samples predicted as R5 by genotype were X4R5 by phenotype. For geno2pheno, the main finding was an overestimation of X4 viruses; a total of 21 samples were assigned as X4R5, whereas 14 of them were R5 by phenotype. No association with low viral load values, low CD4 cell counts, or shorter time since HIV exposure was found, which could explain a preferential clonal detection and the highest degree of discordance.
In a previous study, we reported the prevalence and clinical correlates of X4 viruses in 296 antiretroviral-naive individuals with recent HIV-1 seroconversion in Spain, using the SVM genotypic prediction tool. The proportion of X4 and R5/X4 dual-tropic viruses in this population was 17.2%.8 This rate was higher than expected for persons who had been infected for a short period, because transmission of HIV is thought to be driven by R5 variants in most instances.16-18 This rate is not significantly different from that reported in larger cohorts of antiretroviral-naive chronically infected individuals, however.4,5 A major caveat for our findings is that it remains unclear if genotypic methods for predicting HIV-1 tropism and coreceptor use from V3 sequences can reliably be applied to uncultured and uncloned patients' material, which possibly contains mixtures of X4 and R5 viruses. Using a phenotypic tropism assay, however, we have confirmed our prior findings. In the present study, we tested a random sample of 61 patients from the original population using a phenotypic assay, and 16.4% were found to carry X4R5 viruses.
We further examined the concordance between phenotypic results and those obtained using different genotypic prediction tools of viral tropism, all of which are based on V3 sequences. The overall concordance between geno and pheno was >80%. Although the specificity for recognizing X4 viruses approached 90% using almost all genotypic tools (except geno2pheno), the sensitivity was generally low (<30%), except for SVM and geno2pheno (70% sensitivity). The highest specificity and sensitivity for X4 viruses were observed using geno2pheno and particularly SVM. These data are in agreement with similar studies conducted in chronically infected drug-naive19 and antiretroviral-experienced11,20,21 patients. All these studies concluded that although inference of viral tropism using HIV-1 V3-loop-based coreceptor algorithms performs relatively well in clonal samples,22 genotypic methods are so far inadequate for predicting the presence of X4 variants in clinical samples.
Several reasons have been postulated to explain the limited performance of genotypic tools in clinical samples. In most subjects with progressive disease, multiple HIV-1 tropic viruses coexist in a dynamic fashion23; therefore, the differences between results derived from viral culture and population-based sequencing most likely reflect the presence of viral mixtures and the fact that a different lower limit of detection exists for minority species in these assays. Studies with phenotypic recombinant assays have shown that the limit of detection for minority species is in the range of 5% to 20%,24,25 whereas the limit for conventional bulk sequencing is around 25% to 30%. A further consideration is that although the V3 loop contains the major determinants of coreceptor use, other neighbor regions outside V3 may contribute to the interaction between the virion and the cell surface coreceptor.26,27 It should be pointed out that distinct recombinant phenotypic assays differ in the length of the region cloned; although a few incorporate the complete gp120 or gp160 gene,10,28 others, such as the test used in this study, miss the C4 region and all gp41.10 Recent reports have suggested that the C4 region and gp41 could modulate coreceptor binding.29 Because current sequencing is restricted to a short fragment encompassing the V3 loop, expanding the sequence analysis beyond the V3 loop could increase the sensitivity for identifying X4 viruses using genotypic prediction tools.
The inclusion of CCR5 antagonists into the HIV drug armamentarium requires the implementation of reliable methods to determine HIV tropism. Testing with these tools should be performed in all candidates to be treated with CCR5 antagonists before prescribing them and on treatment failure with these compounds. Our data demonstrate that even in recent HIV-1 infection, the proportion of patients carrying X4 variants is not negligible (∼16%); therefore, even in this particular subset of patients, tropism testing is mandatory. Most of these sites have been designed only for research purposes with no accessible reports of control for accuracy and reliability and with no security methods to ensure confidentiality after submission of patient information. At the moment, most genotypic amplification/sequencing tools lack adequate sensitivity to detect X4 viruses; therefore, they cannot confidently be used to make treatment decisions in clinical practice. Efforts to improve their performance are urgently needed, however, because the sophisticated and expensive phenotypic assays could not be available in many regions where CCR5 antagonists may be needed in a short time.
1. Fatkenheuer G, Pozniak A, Jonson M, et al. Efficacy of short-term monotherapy with maraviroc, a new CCR5 antagonist, in patients infected with HIV-1. Nat Med
2. Gulick R, Su Z, Flexner C, et al. Phase 2 study of the safety and efficacy of vicriviroc, a CCR5 inhibitor, in HIV-1-infected, treatment-experienced patients: AIDS Clinical Trials Group 5211. J Infect Dis
3. Berger E, Murphy P, Farber J. Chemokine receptors as HIV-1 coreceptors: role in viral entry, tropism, and disease. Annu Rev Immunol
4. Brumme Z, Goodrich J, Mayer H, et al. Molecular and clinical epidemiology of CXCR4-using HIV-1 in a large population of antiretroviral naïve individuals. J Infect Dis
5. Moyle G, Wildfire A, Mandalia S, et al. Epidemiology and predictive factors for chemokine receptor use in HIV-1 infection. J Infect Dis
6. Poveda E, Briz V, Quiñónes-Mateu M, et al. HIV tropism: diagnostic tools and implications for disease progression and treatment with entry inhibitors. AIDS
7. Johnson V, Brun-Vezinet F, Clotet B, et al. Update of the drug resistance mutations in HIV-1: 2007. Top HIV Med
8. de Mendoza C, Rodriguez C, Garcia F, et al. Prevalence of X4 tropic viruses in patients recently infected with HIV-1 and lack of association with transmission of drug resistance. J Antimicrob Chemother
9. de Mendoza C, Rodriguez C, Colomina J, et al. Resistance to nonnucleoside reverse transcriptase inhibitors and prevalence of HIV type 1 non-B subtypes are increasing among persons with recent infection in Spain. Clin Infect Dis
10. Van Baelen K, Vandenbroucke I, Rondelez E, et al. HIV-1 coreceptor usage determination in clinical isolates using clonal and population-based genotypic and phenotypic assays. J Virol Methods
11. Poveda E, Briz V, de Mendoza C, et al. Prevalence of X4 tropic HIV-1 variants in patients with differences in disease stage and exposure to antiretroviral therapy. J Med Virol
12. Pillai S, Good B, Richman D, et al. A new perspective on V3 phenotype prediction. AIDS Res Hum Retroviruses
13. Jensen M, Li F, van't Wout A, et al. Improved coreceptor usage prediction and genotypic monitoring of R5-to-X4 transition by motif analysis of HIV type 1 env V3 loop sequences. J Virol
14. Jensen M, Coetzer M, van't Wout A, et al. A reliable phenotype predictor for HIV type 1 subtype C based on envelope V3 sequences. J Virol
15. Sing T, Beerenwinkel N, Lengauer T. Learning mixtures of localized rules by maximizing the area under the ROC curve. In: Hernández-Orallo J, Lachiche N, Flach P, eds. ROCAI- 2004 International Workshop on ROC Analysis in Artificial Intelligence
. London, UK: Premiere Inc. 2004;89-96.
16. Berger E, Murphy P, Farber J. Chemokine receptors as HIV-1 coreceptors: role in viral entry, tropism, and disease. Annu Rev Immunol
17. Kawamura T, Kurtz S, Blauvelt A. The role of Langerhans cells in the sexual transmission of HIV. J Dermatol Sci
18. Bomsel M, David V. Mucosal gatekeepers: selecting HIV viruses for early infection. Nat Med
19. Low A, Dong W, Chan D, et al. Current V3 genotypic algorithms are inadequate for predicting X4 co-receptor usage in clinical isolates. AIDS
20. Wilkin T, Su Z, Kuritzkes D, et al. HIV-1 chemokine coreceptor use among antiretroviral-experienced patients screened for a clinical trial of CCR5 inhibitor: AIDS Clinical Trial Group A5211. Clin Infect Dis
21. Melby T, Despirito M, Demasi R, et al. HIV-1 coreceptor use in triple-class treatment-experienced patients: baseline prevalence, correlates, and relationship to enfuvirtide response. J Infect Dis
22. Jensen M, van't Wout A. Predicting HIV-1 coreceptor usage with sequence analysis. AIDS Rev
23. Karlsson I, Antonsson L, Shi Y, et al. HIV biological variability unveiled: frequent isolations and chimeric receptors reveal unprecedented variation of coreceptor use. AIDS
24. Coakley E, Petropoulos C, Whitcomb J. Assessing chemokine co-receptor usage in HIV. Curr Opin Infect Dis
25. Whitcomb J, Huang W, Fransen S, et al. Development and characterization of a novel single-cycle recombinant-virus assay to determine HIV-1 coreceptor tropism. Antimicrob Agents Chemother
26. Melikyan G, Platt E, Kabat D. The role of the N-terminal segment of CCR5 in HIV-1 Env-mediated membrane fusion and the mechanism of virus adaptation to CCR5 lacking this segment. Retrovirology
27. Walter BL, Wehrly K, Swanstrom R, et al. Role of low CD4 levels in the influence of HIV type 1 envelope V1 and V2 regions on entry and spread in macrophages. J Virol
28. Skrabal K, Low A, Dong W, et al. Determining HIV coreceptor use in a clinical setting: degree of correlation between two phenotypic assays and a bioinformatic model. J Clin Microbiol
29. Suphaphiphat P, Essex M, Lee TH. Mutations in the V3 stem versus the V3 crown and C4 region have different effects on the binding and fusion steps of HIV type 1 gp120 interaction with the CCR5 coreceptor. Virology
Members of the Spanish HIV-1 Seroconverter Study Group
Javier Colomina, Hospital de la Ribera, Valencia; Concepción Tuset, Hospital General, Valencia; Felix Gutiérrez, Enrique Bernal, and Victoria Sánchez-Hellín, Hospital General, Elche; Federico Garcia, Hospital Universitario, Granada; Isabel Viciana, Hospital Virgen de la Victoria, Málaga; Julian Torre-Cisneros, Hospital Reina Sofia, Córdoba; Jose M. Eiros and Raúl Ortíz de Lejarazu, Hospital Clínico, Valladolid; Antonio Aguilera, Hospital Xeral, Santiago de Compostela; Pilar Leiva, Hospital Central de Asturias, Oviedo; Jesús Agüero and Ana Sáez, Hospital Marqués de Valdecilla, Santander; Francesc Vidal, Hospital Joan XXIII, Tarragona; Esteban Ribera and Estrella Caballero, Hospital Vall d'Hebrón, Barcelona; Lidia Ruíz, Fundacio IrsiCaixa, Badalona; José Luis Gómez, Hospital Ntra Sra de la Candelaria, Santa Cruz de Tenerife; Manolo Leal, Hospital Virgen del Rocio, Sevilla; Carmen Rodríguez and Jorge del Romero, Centro Sanitario Sandoval, Madrid; and Carmen de Mendoza, Angélica Corral, Natalia Zahonero, and Vincent Soriano, Hospital Carlos III, Madrid.