icon-folder.gif   Conference Reports for NATAP  
  XVII International HIV Drug Resistance Workshop
June 10-14, 2008
Sitges, Spain
Back grey_arrow_rt.gif
Gene-Based Tools Getting Better at Predicting Coreceptor Use
  XVII International HIV Drug Resistance Workshop
June 10-14, 2008, Sitges, Spain
Mark Mascolini
Several Resistance Workshop studies yielded evidence that fast new genotyping tools may determine HIV coreceptor use in patients better than earlier gene-based schemes. Researchers have sought a reliable gene-sequencing method for predicting coreceptor use since it became apparent that CCR5 antagonists would join the antiviral armory.
CCR5 antagonists cannot control replication of virus that uses the CXCR4 coreceptor. Current genotyping systems detect as few as 30% to 50% of a person's X4-using virus, according to Richard Harrigan of the British Columbia Centre for Excellence in HIV/AIDS. The phenotypic Trofile test, though reliable, is expensive and slower than genotyping might be.
Harrigan and a German-US team, some of whose members previously cast doubt on the accuracy of genotypic coreceptor-calling tactics [1], took a new, fully automated approach with triplicate independent V3 loop sequence genotyping coupled with two gene-based predictors, PSSM and geno2pheno [2]. Harrigan and colleagues also devised a method for figuring coreceptor use in people with an undetectable plasma load by analyzing viral sequences in cellular HIV DNA.
These investigators amplified each viral sequence in triplicate and used RE-call software to determine HIV V3 loop sequences--the primary determinant of coreceptor use--in 63 viral samples for which the phenotypic Trofile assay made coreceptor calls. Then they plugged the sequencing data into two coreceptor-reckoning tools, PSSM and geno2pheno, to interpret the genotypes as R5 or DM (meaning the virus can use either CCR5 or CXCR4).
With this approach, the investigators calculated a 61.2% sensitivity and 93% specificity with geno2pheno in making the same coreceptor picks as Trofile, and a 75% sensitivity and 83% specificity with PSSM. Combining geno2pheno and PSSM yielded a sensitivity of 75.8% and a specificity of 91.1%. Validating the combined method in 278 samples from the maraviroc MOTIVATE trials, Harrigan figured a sensitivity of 72% and a specificity of 88%.
Scores from the genotyping tools correlated positive with the light-unit readout on the Trofile assay, whose strength determines coreceptor use [r(2) = 0.41]. This result suggested that "V3 sequence variation alone explains much of the variation in the X4 phenotype parameter."
Next these researchers determined V3 sequences in HIV DNA of 26 people whose HIV could use both CCR5 and CXCR4 and in 14 whose virus used only CCR5--all of whom had undetectable plasma HIV RNA as a result of antiretroviral therapy. Such people may want to start a CCR5 antagonist, Harrigan explained, because they find their current combination hard to take, but they could be reluctant to stop their drugs and let their viral load rise into ranges that make genotyping easier. With the DNA-ciphering approach, sensitivity and specificity for reaching the same coreceptor call as Trofile measured 76% and 71% with PSSM and 77% and 93% with geno2pheno.
Harrigan and coworkers concluded that their strategies should screen out most people with CXCR4-using virus, who would be poor candidates for CCR5 antagonist therapy. The cost, they estimated, would be $300 to $400 per assay. But phenotyping with Trofile may still be needed, the researchers cautioned, to confirm R5 status.
Alexander Thielen from the Max-Planck Institute in Saarbrucken and some of these same investigators explored the value of adding genotypic shifts in the HIV V2 loop sequence to V3 loop sequence data in predicting coreceptor leanings [3]. They analyzed 916 Los Alamos database sequences from 312 people for whom a phenotypic assay had determined coreceptor use. Seven V2 mutations correlated significantly with CXCR4 use.
A prediction model incorporating both V2 and V3 mutations proved slightly but significantly more accurate than a model using only V3 mutations (area under the receiver operating characteristic curve (AUC) 0.933 for V2 plus V3 versus 0.914 for V3 alone (P = 0.0019). The investigators then validated their prediction model on an independent set of viral samples from 268 antiretroviral-naive people. At a specificity of 90%, sensitivity improved from 54.2% with V3 genes alone to 62.8% with V2 plus V3. Again the AUC was higher when the model considered both V2 and V3 mutations (0.841) than when relying only on V3 mutations (0.778).
With Max Planck and University of Cologne coworkers, Martin Daumer from the Institute of Immunology and Genetics in Kaiserslautern took a third approach to predicting coreceptor preference with genotypic data [4]. Instead of relying on standard population-based sequencing to determine V3 gene variations, they used an "ultradeep" sequencing method that can detect individual viral clones. The study began with 55 viral samples from antiretroviral-treated people for which the Trofile assay had figured coreceptor use. The investigators chose 7 CCR5-using samples and 7 CXCR4-using samples for further analysis with the ultradeep method and the geno2pheno coreceptor prediction tool.
With the 14 samples selected for ultradeep analysis, standard sequencing plus geno2pheno interpretation resulted in one false R5 prediction and four false X4 predictions. Ultradeep sequencing plus geno2pheno resulted in only one coreceptor prediction differing from the Trofile call. But this type of ultradeep sequencing costs much more than standard sequencing.
1. Low AJ, Dong W, Chan D, et al. Current V3 genotyping algorithms are inadequate for predicting X4 co-receptor usage in clinical isolates. AIDS. 2007;21:F17-F24.
2. Moores A, Thielen A, Dong W, et al. Improved detection of X4 virus by V3 genotyping: application to plasma RNA and proviral DNA. XVII International HIV Drug Resistance Workshop. June 10-14, 2008, Sitges, Spain. Abstract 89.
3. Thielen A, Harrigan PR, Lou AJ, et al. Improved genotypic prediction of HIV-1 coreceptor usage by incorporating V2 loop sequence variation. XVII International HIV Drug Resistance Workshop. June 10-14, 2008, Sitges, Spain. Abstract 90.
4. Daumer MP, Kaiser R, Klein R, et al. Inferring viral tropism from genotype with massively parallel sequencing: qualitative and quantitative analysis. Improved genotypic prediction of HIV-1 coreceptor usage by incorporating V2 loop sequence variation. XVII International HIV Drug Resistance Workshop. June 10-14, 2008, Sitges, Spain. Abstract 91.