Co-receptor usage and prediction of v3 genotyping algorithms in hiv-1 subtype b' from paid blood donors experienced anti-retroviral therapy in chinese central province

Background This study explored co-receptor usage and prediction of V3 genotyping algorithms in HIV-1 subtype B' from paid blood donors experienced anti-retroviral therapy in Chinese central province in order to design effectively therapeutic regimen. Methods HIV-1 strains were isolated in treatment HIV-1 infections and treatment-naïve HIV-1 infections, then co-receptor usage of HIV-1 strains was identified based on Ghost cell lines using flow cytometry. HIV-1 V3 region was amplified and submitted into web-server (WebPSSM and geno2pheno) to predict HIV-1 co-receptor usage. The feasibility of prediction HIV-1 usage with Web-server assay was analyzed by comparing prediction of V3 genotyping algorithms with HIV phenotype assay based on Ghost cell line. Results 45 HIV-1 strains and 114 HIV-1 strains were isolated from HIV-1 infections exposed anti-retroviral therapy and treatment-naïve, respectively. 41% clinical viruses from ART patients and 18% from treatment-naïve patients used CXCR4 as co-receptor. The net charge in the V3 loop was significantly difference in both groups. The sensitivity and specificity for predicting co-receptor capacity is 54.6% and 90.0% on 11/25 rule, 50.0% and 90% on Web-PSSMx4r5, 68.2% and 40.0% on Geno2pheno[co-receptor]. Conclusion Dual/mixed/X4 co-receptor utilization was higher in ART patients than treatment-naïve patients. It is should paid attention to predicting HIV-1 co-receptor usage based on V3 genotyping algorithms in HIV-1 subtype B' from paid blood donors experienced anti-retroviral therapy in Chinese central province.


Background
HIV-1 enters a host cell using the CD4 receptor and coreceptors including the CXCR4 and/or CCR5. In general, R5-tropic strains using CCR5 as co-receptor are responsible for the early stage of infection, while mixed or dualtropic R5/X4 strains using both CXCR4 and CCR5 as co-receptor, and X4 using CXCR4 co-receptor are detected in more advanced disease stages, and are believed to be associated with more rapid CD4 + T cell decline and accelerate disease progression to AIDS [1]. However the X4 viruses usually coexist with R5 viruses in the viral swarm [2]. There are still 50% patients with late stage HIV-1 B infection having only R5 viruses detectable in treatment-naïve HIV-1 patients [3] but not other HIV-1 subtypes [4,5]. The mechanisms that prompt the evolution towards CXCR4 strains from CCR5 strains are not fully understood. Meanwhile, there were different point of views about HIV-1 co-receptor usage after the patients experienced highly active antiretroviral therapy (HAART). After HAART therapy (59 months [6-240 months]), HIV-1 co-receptor usage was fairly stable [6]. But, some drugs are duty to the preferential suppression of CXCR4-special strains of HIV-1 [7].
The third variable loop (V3) sequence of HIV envelope is the major domain associated with HIV co-receptor usage [8]. In general, when the amino acids at codons 11 and/or 25 within the V3 loop is positive charged, the HIV strains usually use CXCR4 as co-receptor. Therefore the 11/25 charge rule is a simple genotypic method to be predicted HIV co-receptor usage. Subsequently, several genotyping algorithms based on V3 loop for predicting HIV co-receptor usage have been published, such as neural networks(NN), decision tree support vector machines(SVM) [9], Position Specific Scoring Matrix approach (PSSM) [10]. However, it reports that current V3 genotyping algorithms are inadequate for predicting X4 co-receptor usage in clinical isolates [11].
Since the first co-receptor antagonist--Maraviroc against HIV-1 was approved in the United States in 2007, which blocking HIV-1gp120 from binding to CCR5, thereby preventing HIV-1 into the host cell. It could effectively inhibit CCR5-tropic strain but not CXCR4-tropic strain, and is a promising agent for treatment-experienced patients infected with multidrug-resistant CCR5 strain [12]. It is necessary to know HIV-1 coreceptor usage before Maraviroc is applied to clinical. Therefore, we collected HIV-1 infections experiencing treatment with reverse transcriptase inhibitors, and isolated HIV-1 strains from HIV-1 infections to evaluate the feasibility that predictes HIV-1 co-receptor usage based on V3 genotyping algorithms.

Association of HIV-1 co-receptor usage with clinical characteristics
Compared with treatment-naive participants, a higher percentage of HIV-1 strains in treated participants were harboring dual/mixed/X4-tropic viruses (51.1% vs. 18%) (See table 2). To further analyze association of HIV-1 co-receptor usage with clinical characteristics, CD4 + T cell count or VL was stratified and the discrepancy was analyzed using the Mantel-Haenszel test. After adjusted by CD4 + T cell count or VL, the HIV-1 co-receptor usage was difference between treatment-naïve and ART group(p < 0.05; see table 2). HIV-1 X4 co-receptor usage utilization has higher percentage in ART group than treatment-naïve group, and increased with CD4 + T cell count decrease and with VL increase (Figure 2).
In treatment groups, there is no association between HIV-1 co-receptor usage and therapeutic regimens (p > 0.05). Also, when treatment time was stratified (treatment time < 18 months as a group, 18 months = < treatment time < 30 months as second group, and treatment > = 30 months as the third group), there was no evidence for association between treatment time and HIV-1 co-receptor usage (P > 0.05)(see table 2).
Association of HIV-1 co-receptor usage with V3 loop sequence 81 sequences of V3 region from the 114 treatment-naïve patients and 42 sequence of V3 region from 45 ART patients were amplified. According to the formula (V3 net charge = (R + K)-(D + E)), net charge of V3 loop was calculated. In the formula, the R and K was short for argentine and lysine, respectively; D and E short for aspartic acid and glutamic acid, respectively. In the ART group, the net charge of V3 loop was distributed from 2 to 7(4.33 ± 1.34), of which 4.86 ± 1.25 for X4/R5 strain, 3.75 ± 1.21 for the R5 strain. In the treatment-naïve    table 3).
In both ART and treatment-naïve group, number of net charge of V3 for R5-tropic viruses distributed mainly below 4, which frequency is more than 70%. However, number of net charge of V3 for X4-tropic viruses distributed mainly above 4 in treatment-naïve group, above 5 in ART group (see table 3).
HIV-1 co-receptor usage was predicted based on genotypic algorithm including 11/25 charge rule, Webserver (Web-PSSM x4r5 and Geno2pheno [coreceptor] ), which is called HIV-1 co-receptor genotype. The consistency between genotype and phenotype of co-receptor usage was evaluated among ART population. The sensitivity and specificity for predicting X4 capacity is 54.6% and 90.0% on 11/25 rule, 50.0% and 90% on Web-PSSM x4r5 , 68.2% and 40.0% on Geno2pheno [coreceptor] (see table 4).   [15]. In September 2003, the "Four Frees and One Care" policy was implemented, which provided free antiretroviral drugs in above areas [16]. In this article, we report that a largescale study of HIV-1 co-receptor usage and their prediction based on V3 genotyping algorithms in population who were infected by paid blood donors in Henan and Anhui province. However, there is limited information to know X4-to-R5 switch of HIV-1 in this population after antiviral therapy. Therefore, the present study was based on the characterization of specimens collected from 45 subjects experienced ART and 114 treatmentnaïve subjects between 2005 and 2008. All the viruses isolated from ART and treatment-naïve population in this study are HIV-1 B' subtype. The combination of B' viral subtype and Chinese host's genetic background has likely provided a unique situation for the understanding of HIV-1 co-receptor usage and their prediction based on V3 genotyping algorithms in a particular population who infected though paid blood donation and then experienced ART. In present study, co-receptor usage of HIV-1 in patients with and without treatment on HAART was detected based on Ghost cell lines (phenotypic assays). The results showed that the HIV-1 CXCR4 utilization among antiretroviral therapy HIV-1 infected patients was higher than in the treatment-naïve population, implying that it should pay attention to the choice of co-receptor antagonists after the treatment failure on HAART. The present study was in agreement with Hunt's results that there is more widely X4-tropism strain in antiretroviral-experienced patients [17]. When CD4 + T cell count or VL was stratified, the HIV-1 coreceptor usage was difference between treatment-naïve and ART group. HIV-1 CXCR4 utilization has higher percentage in ART group than treat-naïve group, and increased with CD4 + T cell count decrease and with VL increase in both group. Usually, the CXCR4 utilization is higher in more advanced disease stages. There is a report that some drugs are duty to the preferential suppression of CXCR4-special strains of HIV-1 [7], However, the frequency of CXCR4 utilization in the two therapeutic regimens (AZT + DDI + NVP or D4T + DDI + NVP) is no difference in our study, This study could not found any association between treatment time with CXCR4 utilization, which agreed with other report [6]. HIV-1 R5 to X4 switch is dynamic processes during the interaction between HIV-1 variation and host immune. Of course, it does not exclude the reason that the criterion that the participants in ART group would initiate antiretroviral therapy is that their CD4 + T counts must be blow 200 cells/μl in China.
Number of net charge of V3 plays an important role in detecting viral R5-to X4 co-receptor switch. 70% R5tropic viral net charge of V3 distributed below 4 whatever exposed to drug or not. However, there is more than 60% for X4-tropic viruses which number of net charge of V3 distributes mainly above 4 in treatmentnaïve group, above 5 in ART group, For exception, there is not any X4-tropic viruses which the number of net charge of V3 is below 4 in treatment-naïve group, whereas there is 18.2% X4-tropic viruses which the number of net charge of V3 is below 4 in ART group, suggesting the number of net charge of V3 is not available for co-receptor prediction of HIV-1 B' subtype exposed to drug.
V3 loop, as the major determinant of viral tropism, is a base of lots of prediction essays of co-receptor usage, Note: a: the number of cases; b:it is the net charge of V3 loop according to the formula (V3 net charge=(R+K)-(D+E)); c:it is the frequency of b in the a. bold: No. of net charge of V3 for R5-tropic viruses distributed mainly below 4, occupied more than 70%; bold and italic: No. of net charge of V3 for X4-tropic viruses distributed mainly above 4 in drug-naïve group, above 5 in ART group, occupied more than 60% Table 4 HIV-1 co-receptor prediction based on genotypic algorithm and its sensitivity and specificity in ART population for example, networks(NN), decision tree [9], support vector machines(SVM) [9], Position Specific Scoring Matrix approach (PSSM) [10]. In this study, PSSM x4/r5 , geno2pheno [coreceptor] and 11/25 charge rule were chosen to assess the concordance with phenotype assay. The specificities and sensitivities in our study is lower than Garrido's study that the specificities for detecting HIV-1 B X4 variants are 92%(PSSM x4/r5 ), 88%(geno2pheno [coreceptor] ), and the sensitivities are 90%(PSSM x4/ r5 ) and 90% (geno2pheno [coreceptor] ) [18], but higher than Whitcomb's study that the specificities for detecting HIV-1 B X4 variants are more than 90%(PSSM x4/r5 , gen-o2pheno [coreceptor] , 11/25rule). And the sensitivities are merely 30.5%(11/25rule), 24.5%(PSSM x4/r5 ) and 44.7% (geno2pheno [coreceptor] ) [11]. The reason for this difference is different method for detecting HIV-1 co-receptor phenotype. Anyway, all the studys reach an agreement that current V3 genotyping algorithms are inadequate for predicting X4 co-receptor usage in clinical isolates.

Conclusions
In summary, the study shows that prevalence of dual/ mixed/X4 HIV-1 strain among ART participants is higher than among treatment-naïve participants. V3 genotyping algorithms for predicting HIV-1 co-receptor usage is not enough for HIV-1 B' subtype from patients experienced ART.

Study population
All the subjects were recruited from HIV-1 infected former blood/plasma donors (FBDs) [13] in The blood from all the subjects was collected for viral load, CD4 + T count detection and the peripheral blood mononuclear cells (PBMCs) for isolating primary HIV strains. All subjects signed informed consent forms before blood collection. This study was approved by the Institutional Research Ethics Committee of Chinese Center for Disease Control and Prevention in China. The viral load were tested with COBAS AMPLICOR™ techniques and Analyzer (Roche Diagnostics, Alameda, CA). The count of CD4 + T and CD8 + T lymphocytes was performed by flow cytometry (EPICS-XL, Coulter) with TruCount package from BD Biosciences (San Jose, CA).

HIV-1 isolation from patients' PBMCs
Primary HIV-1 strains were isolated by co-culturing PBMCs from infected individual and those from two or more from healthy individuals after phytohaemagglutinin(PHA)-stimulation. The co-culture was incubated in growth RPMI-1640 medium supplemented with 10% fetal calf serum (FCS), 100 U/ml penicillin, 100 μg/ml streptomycin, 2.9 mg/ml L-glutamine and 100 IU recombinant IL-2 (Roche Diagnostic,Sigma) as previously described [19]. Cultures were maintained by regular addition of uninfected stimulated PBMCs and fresh media. Culture supernatants were collected once a week to measure p24 production levels using a commercial enzyme-linked immunosorbent assay (ELISA) kit according to the instructions from the manufacturer (BioMerieux, Marcy-l'Etoile, France). Virus culture supernatants with p24 consentations higher than 2 ng/ ml were aliquoted and stored in liquid nitrogen until being used.

Detection of HIV-1 co-receptor usage
GHOST cells, expressing CD4 while expressing CXCR4 or CCR5, were seeded in 24-well plates (Corning Inc, Spain) at the density of 1×105 cells/well*0.5 ml. On the following day, the monolayers, about 70% confluent, were infected with virus stocks (200 μl/well) in the presence of 8 μg/ml DEAE-dxtran to enhance the infective efficiency. After 48 hours, cells were harvested and analyzed with flow cytometer (Elite ESP, Beckman Coulter, Germany) and a total of 10,000 to 15,000 events were scored. We expected an approximately 10 fold shift in mean GFP fluorescence of infected cells over uninfected cell [20]. The Ghost-R5 and -X4 cells infected with HIV-1 SF33 , HIV-1 Ba-L and HIV-1 IIIB were positive controls and the cells without HIV-1 infection were negative control.

Bioinformatic prediction
After alignment, sequences with positively charged amino acids at codons 11 and/or 25 within the V3 loop were classified as having an 11/25 genotype. Then the HIV-1 strain with 11/25 genotype was believed as CXCR4 or CXCR4/CCR5 strain.

Statistical analysis
In this study, age, CD4 + T count and treatment time was indication as mean or median and range, and virus load was transformed to log10. The age, CD4 + T count or VL difference between ART and treatment-naïve were performed by using T test, and the distribution of gender between two groups was performed by using chi square test. All the statistical analysis was done on SPSS software (V13.0), and a P value less than 0.05 was considered statistically significant.