CCR5 and CCR5 Screening Assays Comparisons: genotype, 454 Ultra-Deep Sequencing and Trofile
Reported by Jules Levin
18th Intl HIV Drug Resistance Workshop
June 9-12 Ft Myers Florida
Transitions from CCR5 to CXCR4 use in the absence of antiretroviral drug pressure proceed incrementally and may occur through a multitude of genetic pathways
W Huang1, E Coakley1, J Toma1, E Stawiski1,
A Frantzell1, H Schuitemaker2, CJ Petropoulos1 and
AB van't Wout2
1Monogram Biosciences, South San Francisco, CA, USA
2Department of Experimental Immunology, Center for Infectious
Diseases and Immunity Amsterdam (CINIMA), Academic Medical
Center, University of Amsterdam, Amsterdam, the Netherlands
BACKGROUND: The emergence of HIV subpopulations that utilize the CXCR4 coreceptor is clinically important yet poorly understood. CXCR4 use is prognostic for disease progression, negatively predictive for response to CCR5 antagonists and represents a prevalent escape pathway for CCR5 inhibitor treatment. To advance our understanding of the acquisition of CXCR4 use, we characterized coreceptor utilization along with envelope clone sequences derived from patient virus populations that evolved the ability to use CXCR4 in the absence of antiretroviral drug pressure.
METHODS: Longitudinal plasma samples were obtained from eight treatment-naive subjects with previously documented tropism switches (NSI to SI). Envelope clones were isolated from different time points for each subject.
Coreceptor tropism was assigned by Trofile and envelope sequences were determined using conventional methods.
RESULTS: Virus populations at the time of the coreceptor switch comprised mixtures of R5- and dual-tropic clones. The proportion of dual tropic clones in virus populations and their ability to use CCR5 and CXCR4 varied by subject.
Dual-X clones that use CXCR4 efficiently were identified in all eight subjects; dual-R clones that use CXCR4 inefficiently appeared in six patients and presented prior to, or concurrently with, dual-X tropic clones. Phylogenetic analysis of full-length envelope sequences demonstrated that emerging dual-X variants exhibited limited diversity and were markedly divergent from R5 and dual-R clones in some, but not all cases. Generally, efficient CXCR4 use required multiple amino acid changes in the V3 loop, but few discernable patterns were shared among subjects. The presence of positively charged amino acids at positions 11/25 and the absence of potential N-linked glycosylation sites in V3 were associated with, but not unique to, CXCR4 use.
CONCLUSIONS: Detailed characterization of emerging CXCR4-using variants in the virus populations of eight individuals indicates that HIV uses a multitude of mutational pathways to change from CCR5 to CXCR4 use. Furthermore, our observations indicate that CCR5 to CXCR4 transitions are incremental, not quantum, and are consistent with the 'R5, dual-R, dual-X' model that we
have proposed to describe the evolution of CXCR4 use from R5-tropic virus populations.
Estimating evolutionary pathways to CXCR4 usage from cross-sectional data
A Thielen1, A Altmann1, J Bogojeska1, R Kaiser2 and
1Max-Planck Institute for Informatics, Saarbrucken, Germany
2University of Cologne, Cologne, Germany
BACKGROUND: Coreceptor usage of HIV-1 is mainly determined by the V3 loop of gp120. Mutations outside V3, most notably in the bridging sheet, have also been described to be correlated with X4 viruses. So far, however, it is not understand if these mutations prepare for the coreceptor-switch, or if they act as compensatory mutations for fitness losses induced by 11/25 mutations (306R/
K/H and 322R/K/H). In this work, we wanted to study this question with the help of mutagenetic trees, which have previously been used to model HIV resistance pathways in presence of drug pressure.
METHODS: Sequence data containing the V3 loop (n=9,557) as well as one of the gp160 regions V2 (n=2,656), C4 (n=2,098) or GP41 (n=827) were downloaded from the Los Alamos Sequence Database. Only subtype B sequences and at most one sequence per patient were analyzed. Samples with experimentally determined phenotype were used to detect mutations associated with X4 viruses (Fisher's exact test). Mixture models of mutagenetic trees containing the most significant mutations were then generated using the R-package Rtreemix.
RESULTS: Mutations at 13 positions within V2, 23 in V3, 2 in C4 and 15 in GP41 were significantly (P<0.05) associated with X4 viruses. In almost all generated trees, the mutations 306R/K/H and 322R/K/H appeared in two different arms of the tree suggesting that these are two independent pathways to evolve to X4. 306R/K/H was generally preceded by mutations at positions 306 and 316 while
322R/K/H was selected by I424V and mutations at position 317. Other mutations usually came afterwards. GP41 mutations including the highly predictive insertion A/I/V after position 515 in GP41 were all seen as successors of 306R/K/H. Mutations disrupting the N-glycosylation motif of V3 followed this pathway too. By contrast, V2 mutations, such as the well known X4 mutation S195H, commonly appeared after 322R/K/H. Another mutation in C4 (S440D/E) previously been described as highly correlated with CXCR4 usage evolved independently from
306R/K/H and 322R/K/H.
CONCLUSIONS: Since 11/25 mutations predominantly appeared as one of the first mutations in our trees, it can be assumed that most other mutations associated with X4 viruses are compensatory mutations following after the coreceptor switch and that there is no ongoing evolution towards X4 viruses over time.
Modification of the CCR5 binding site by mutations in gp120 influence drug susceptibility and viral infectivity in one subject with clinical resistance to vicriviroc
From Jules: this was I think a well received presentation that reports CCR5 antagonists have a high barrier to resistance.
RA Ogert1, Y Hou1, L Ba1, C Buontempo1,
N Murgolo2, P Qiu2, J Duca3, J Strizki1, R Ralston1 and JA Howe1
1Virology, Schering-Plough Research Institute, Kenilworth, NJ, USA
2Molecular Design and Informatics, Schering-Plough Research
Institute, Kenilworth, NJ, USA
33D Drug Design, Schering-Plough Research Institute, Kenilworth,
OBJECTIVES: The mechanisms by which HIV type-1 (HIV-1) may develop resistance to vicriviroc, a small-molecule CCR5 antagonist, are not well understood. We sought to characterize the viral adaptations that confer phenotypic resistance in a clinical isolate.
METHODS: Site-directed mutagenesis was used to introduce reciprocal point mutations into resistant or baseline env genes. Relative viral infectivity and VCV susceptibility studies were performed using a single-cycle pseudoparticle assay with U87-CD4/CCR5 cells or 293T cells transfected with CD4 and wild-type or mutant CCR5.
RESULTS: The HIV-1 env gene from the clade D strain isolated at the time of study discontinuation contained six amino acid changes in the V3 loop and one in the C4 region as compared with the baseline clone. Pseudovirus generated with the chimeric env were completely resistant to VCV, but remained exclusively R5-tropic. Back mutations E315Q, G321R and F317L in the tip and the stem of the V3 loop region from the resistant env restored complete and partial susceptibility to VCV, respectively; back mutation of V3 loop residues N320D, K328E and R429G in C4 significantly reduced pseudovirus infectivity without altering the resistant phenotype. Forward mutagenesis showed that none of the six amino acid residues were sufficient to confer the resistant phenotype to the baseline chimeric gp120. A combination of four specific mutations in V3 was required to restore complete resistance. Structural modeling demonstrated that certain mutations identified in the V3 loop map to the binding domain in gp120 for the N terminus of CCR5. Entry of pseudovirus generated with resistant envelopes was more sensitive than baseline counterparts to point mutations Y10A, D11A, Y14A and Y15A that eliminate important binding determinants in the N terminus of CCR5.
CONCLUSIONS: The amino acid changes we characterized primarily conferred VCV resistance or influenced envelope infectivity. A specific combination of four of the seven mutations was required for the complete VCV resistance phenotype, suggesting that CCR5 antagonists have a high barrier to resistance. Functionally, the combination of mutations resulted in increased dependence on the envelope interaction with the N terminus of CCR5 for viral entry.
Screening for HIV tropism using population-based V3 genotypic analysis: a retrospective virological outcome analysis using stored plasma screening samples from MOTIVATE-1
PR Harrigan1, R McGovern1, W Dong1, A Thielen2,
M Jensen3, T Mo1, D Chapman4, M Lewis5, I James5
and H Valdez4
1BC Centre for Excellence in HIV/AIDS, Vancouver, Canada,
2Max-Planck Institute for Informatics, Saarbrucken, Germany.
3Fortinbras Research, Buford, GA, USA
4Pfizer, Inc., New York, NY, USA
5Pfizer Research and Development, Sandwich, UK
BACKGROUND: MOTIVATE-1 compared maraviroc+ optimised background versus placebo+OB in treatment-experienced patients with R5-HIV (standard Monogram Trofile). A subset screened without known R5-HIV entered a sister trial, A4001029. This study retrospectively examined population V3 loop sequence-based screening.
METHODS: Triplicate V3 loop sequencing performed on stored screening plasma samples was processed without human intervention using custom software (ReCall) blinded to clinical data. Tropism was inferred using either geno2pheno (g2P; 5% false-positive rate) or PSSM (-2.96 cutoff). Primary outcomes were concordance between the screening Trofile and genotype and viral load changes after starting maraviroc. 25% of the dataset was reserved for validation.
RESULTS: Preliminary genotypes and Trofile results were available from 1,230 samples (non-R5 n=553 by Trofile). Compared with Trofile, genotype had sensitivities of 63% and 56% and specificities of 91% and 90% for detecting non-R5 virus using g2P and PSSM, respectively. However, short-term viral load decreases were similar regardless of the assay used. For example, for patients screened as R5 by Trofile, median week 8 decreases were 2.4 logs (IQR
1.5->3) and 2.4 logs (IQR 1.2->3) in the MVC twice-daily and once-daily arms, respectively, compared with 2.5 logs (IQR 1.6->3) and 2.4 logs (IQR 1.2->3) for patients screened as R5 by g2P, and 2.5 logs (IQR 1.5->3) and 2.4 logs (IQR 1.1->3) for patients screened as R5 by PSSM. For patients screened as non-R5 by Trofile, decreases were 1.4 logs (IQR 0.2->3) and 0.6 logs (IQR 0->3) for MVC twice-daily and once-daily, respectively, compared with 1.5 logs (0.2->3) and 0.7 logs (0->3) for patients screened as non-R5 by g2P, and 1.2 logs (IQR 0.2->3) and 1.0 logs (IQR 0->3) for patients screened as non-R5 by PSSM.
R5 groups had n>100 and all non-R5 groups had n>30. Median placebo week 8 viral load decreases were ≤1 log. Additional analyses are in progress to examine outcomes at 24 weeks and account for the activity of the OB regimen.
CONCLUSIONS: Despite apparently poor sensitivity of standard genotyping for predicting non-R5 HIV relative to standard Trofile, early virological reductions in this treatment-experienced population were similar regardless of the assay used.
Evaluation of the genotypic prediction of HIV-1 tropism versus a phenotypic assay and correlation with the virological response to maraviroc: the GenoTropism study.
P Recordon-Pinson1, C Soulie2, D Descamps3, M Lazrek4, C Charpentier5, B Montes6, MA Trabaud7, J Cottalorda8, C Amiel9, L Morand-Joubert10, C Tamalet11, D Desbois12, M Mace13, V Ferre14, A Vabret15, A Ruffault16, J Izopet17, H Fleury1, F Brun-Vezinet3, AG Marcelin2, J Reynes6, P Flandre18, V Calvez2 and B Masquelier1 and the ANRS AC11 Resistance Study Group
1CHU de Bordeaux and EA 2968, Universite Victor Segalen, Bordeaux, France
2CHU Pitie-Salpetriere, Paris, France
3CHU Bichat-Claude Bernard, Paris, France
4CHU Lille. France
5CHU HEGP, Paris, France
6CHU Montpellier, France
7CHU Lyon, France
8CHU Nice, France
9CHU Tenon, Paris, France
10CHU St Antoine, Paris, France
11CHU Marseille, France
12CHU Paul Brousse, Paris, France
13CHU Orleans, France
14CHU Nantes, France
15CHU Caen, France
16CHU Rennes, France
17CHU Toulouse, France
18INSERM U943, Paris, France
OBJECTIVES: Genotypic algorithms for prediction of HIV-1 coreceptor usage need to be evaluated in a clinical setting. We aimed at studying i) the correlation of genotypic prediction of tropism in comparison with a phenotypic assay, and ii) the relation between genotypic prediction of tropism at baseline and the virological response (VR) to a maraviroc (MVC)-including therapy.
METHODS: Antiretroviral-experienced patients were included in the MVC-expanded access program if they had a R5 screening result with Trofile™ (Monogram Biosciences). V3 loop sequences were determined at screening,
and tropism was predicted using nine genotypic algorithms. Genotypic CXCR4 predictions were compared with Trofile; dual/mixed (D/M) variants were considered as X4 variants. The relationships between genotypic tropism, V3 variability and VR were tested; the VR was defined as plasma viral load (PVL) <50 c/ml at month 6 (M6). The genotypic evolution was determined in patients with suboptimal response to MVC (PVL >50 c/ ml).
RESULTS: Both genotypic and phenotypic results were obtained for 189 patients at screening, with 54 isolates scored X4 and 135 R5 with Trofile. The concordance with Trofile™ ranged between 74% and 81% for the different algorithms, with Geno2pheno10% and PSSMsinsi having the highest sensitivities (59% and 54%) for X4 detection. In patients receiving MVC, at baseline, the median PVL was 4.1 log10 c/ml (IQR 3.4-4.9) and the median CD4 cell count was 268 cells/µl (130-393). The VR at M6 was completed in 68% of patients. The median CD4 increase was 83 cells/µl at M6. The lowest P-value, although not significant, between VR and genotypic prediction of tropism was for geno2pheno10% combined with PSSMr5x4 (P=0.07), with 71% of VR for R5. No association was found between baseline V3 loop mutations and VR. Genotypic follow-up was determined for 13 patients of whom 3 were failing with an X4 genotype, and 10 with a R5 genotype with changes in V3 loop sequences in 4/10.
CONCLUSION: Genotypic prediction of tropism could be associated with VR even in patients screened by a phenotypic assay. The ongoing replication on MVC with or without genotypic evolution of the V3 loop deserves further investigation.
Comparison of population and 454 genotyping and Trofile phenotyping for tropism determination in patients selected for maraviroc initiation
L Vandekerckhove1, C Verhofstede1, I Vandenbroucke2, C Booth3, H Van Marck2, AM Geretti3 and L Stuyver2 on behalf of the Tropism Study Group
1Ghent University Hospital, Ghent, Belgium
2Virco, Mechelen, Belgium
3Royal Free Hampstead NHS Trust & UCL Medical School,
OBJECTIVES: In order to establish alternative strategies to guide initiation of maraviroc, genotypic tropism prediction obtained by population Sanger sequencing was compared with the results of 454 sequencing. We used plasma samples from HAART-failing patients for which population phenotyping analysis by the standard Trofile™ assay (Monogram Biosciences) had been requested.
METHODS: Plasma samples were retrieved from HIVinfected individuals who underwent standard Trofile testing in routine clinical care. Clinical outcomes data were collected. V3-loop sequences were obtained by Sanger and 454 sequencing. Tropism prediction from V3 sequences was obtained using the WebPSSM algorithm and the SVM of Geno2Pheno.
RESULTS: The study comprised 50 patients, including 45% infected with non-B subtypes. Using Sanger sequencing, an interpretable result was obtained in 48/50 (96%) samples: R5=36 and X4=12. With the same samples, 454 sequencing produced a mean of 10,406 reads (range 1,170-19,436) per sample. The predicted X4 virus was quantified. Of the 36 samples predicted as R5 by Sanger sequencing, 34 had <10% of predicted X4 by 454. Of the 12 samples predicted as X4, 11 had >10% X4. Based on the Trofile result, 24 patients were started on maraviroc. Of these, 21 were in the group of predicted to have a prevalence of X4 virus <10%, all of whom had an initially response. Follow-up of these patients is ongoing.
CONCLUSIONS: In comparison with 454 sequencing, Sanger sequencing showed 100% specificity and approximately 10% sensitivity for X4 detection. As the clinically relevant sensitivity threshold for tropism prediction has not been established, population sequencing may provide a rapid assessment tool on its own or be integrated with more sensitive methodologies. In one possible testing
algorithm, population V3 sequencing could be applied to first-line screening, allowing the rapid identification of samples with a high content of X4 sequences, thereby reducing the number of samples for more extensive analyses as 454. This approach needs further clinical validation and will demand a thorough standardization of the Sanger sequencing and interpretation methodology. Additionally, determination of relevant clinical cutoffs for X4 minorities
(QUEST assay) Performance of an HIV-1 coreceptor tropism assay utilizing replicate V3 loop sequencing and heteroduplex analysis with capillary electrophoresis
(Quest issued a press release during the conference saying they offer their test to the public through Quest labs.
R Baumann1, H Hamdan1, D Schwab1, T Robins2, J Vichayanonda1 and RM Kagan1
1Quest Diagnostics Nichols Institute, San Juan Capistrano, CA, USA
2Quest Diagnostics Clinical Trials, Malibu, CA USA
OBJECTIVES: CCR5 (R5) antagonist therapy may be ineffective in patients with pre-existing CXCR4-tropic (X4) or dual/mixed virus. Genotypic tropism tests afford a more rapid and less complex means of screening for X4-tropic virus than phenotypic assays. In HIV samples from treatment-naive patients, the SensiTrop™ II genotypic assay demonstrated 79% concordance with the standard Trofile™ phenotypic assay. We developed a genotypic tropism assay that couples replicate V3 loop sequencing with heteroduplex analysis by capillary electrophoresis (CE-HDA). Here we assess concordance of the CE-HAD method with the Trofile and SensiTrop II assays in samples
from treatment-experienced patients.
METHODS: Coreceptor tropism was assayed for 40 de-identified HIV-1 subtype B patient samples with genotypically predicted resistance to ≥4 reverse transcriptase or protease inhibitors by the Trofile (Monogram Biosciences), SensiTrop II (Pathway Diagnostics) and CE-HDA assays. A 141 bp HIV-1 envelope V3 loop amplicon was sequenced in duplicate, and tropism was predicted by bioinformatics analysis of amino acid sequences. For heteroduplex analysis, four FAM-labelled V3 loop R5 probes were annealed to the V3 amplicons and heteroduplexes were resolved on an ABI 3130xl CE analyser.
RESULTS: X4-utilizing virus was identified in 48% of samples by Trofile, 50% by SensiTrop II and 56% by CE-HDA, but these differences were not statistically significant (McNemar test). CE-HDA was 72% concordant with Trofile (k=0.44). Sensitivity and specificity for X4 were 79% and 65%, respectively. CE-HDA was 74% concordant with SensiTrop II (k=0.49) and SensiTrop II was 73% concordant with Trofile (k=0.45). Seven Trofile R5's were typed as X4 by CE-HDA, four of which were also typed as X4 by SensiTrop II. Four Trofile X4's were typed as R5 by CE-HDA, two of which were also R5 by SensiTrop II.
CONCLUSIONS: In samples from treatment-experienced patients, concordance between two independently developed genotypic tropism assays was comparable to the concordance of each assay with a phenotypic method. The proportion of X4 viruses detected did not vary significantly by assay type. Further comparisons of CE-HAD to the Trofile-enhanced sensitivity (ES) phenotypic assay are warranted. Clinical outcome studies are needed to assess the predictive values of different tropism assays for patients treated with CCR5 antagonists.
Population and clonal tropism phenotyping of clinical isolates and comparison with next generation sequencing and prediction tools
K Van Baelen1, E Rondelez1, L Van Wesenbeeck1,
W Mostmans1, H Van Marck1, V Van Eygen1,
C Verhofstede2, L Vandekerckhove2, I Vandenbroucke1,
L Stuyver1 and the Tropism Work Group
1Virco, Mechelen, Belgium
2Aids Reference Laboratory, Ghent University Hospital, Belgium
OBJECTIVES: The use of CCR5 antagonists is currently limited to patients with R5-tropic HIV only. Several assays to measure tropism of clinical samples and algorithms to predict tropism have been developed.
METHODS: Plasma from 44 treatment-experienced patients with standard Trofile (Monogram) assay results was available for viral RNA extraction and gp120 NH2-V4 amplification. All samples were tested in the Virco phenotypic tropism assay, using population and clonal (24 clones/ sample) approaches. The V3-loop was amplified using barcoded primers, amplicons were equimolar pooled and sequenced on the 454 GS-FLX instrument. Tropism was predicted using the WebPSSM algorithm.
RESULTS: Population phenotype results were obtained for 56.8% (n=25) and 84.1% (n=37) of samples in the Trofile and Virco tropism pheno assay, respectively. Using Trofile as a reference, Virco tropism assay showed specificity 100%, sensitivity 88.9%, PPV 100% and NPV 60%. Clonal phenotyping was successful on 39 samples, resulting in 936 plasmids of which 692 (73.9%) yielded infectious virus with the following distribution: R5=627, D=58 and X4=7 clones. 454 V3-loop sequencing generated an average of 13,475 reads/sample. The percentage of predicted X4 virus was calculated and compared with phenotyping RESULTS: (i) 0% predicted X4 (n=18) showed up as R5 in population- and clonal-based phenotyping; (ii) 0.07-10% predicted X4 (n=10) showed up as R5 in population, but an occasional D in clonal phenotyping; (iii) >10% predicted X4 (n=14) showed up as R5/D/X4 in clonal phenotyping, increasing propensity for D/M in population phenotyping; and (iv) >99% predicted X4 (n=6) showed up as phenotypically R5 (n=3) or D/M (n=3).
CONCLUSION: D/M virus could not be detected using the most sensitive methods in 18 plasma samples from patients failing HAART. Phenotypic approaches detected X4 virus present at 10% or more of the virus population. A total of 692 cases of mainly concordant and some discordant datasets (clonal phenotype versus V3 prediction tool) became available, which in turn can be used to improve the prediction algorithms. Most of the samples failing phenotyping approaches could be characterized by deep sequencing and V3-loop prediction. Our data indicate that 454 sequencing combined with improved prediction tools could provide a sensitive alternative to phenotyping.
Amino acid changes in GP41 of HIV-1 associated with coreceptor tropism
E Stawiski, W Huang, J Whitcomb, L Napolitano,
C Petropoulos and E Coakley
Monogram Biosciences, South San Francisco CA, USA
OBJECTIVES: Much work has focused on amino acid changes within the V3 region of the HIV-1 envelope gene that influence coreceptor tropism (CRT) usage in the entry process. Changes outside of V3 are also known to affect CRT including at least one mutation within the GP41 region. Here we describe additional mutations in gp41 associated with CRT.
METHODS: We evaluated available gp41 sequence data from commercial samples submitted for CRT testing by Trofile™ a CLIA-validated cell-based recombinant virus assay. These gp41 sequences being determined for quality
control purposes. We selected subtype B samples (as determined by the Hyphy package) that contained a minimal spanning sequence starting at the beginning of the HR1 region and ending in the intercytoplasmic region of GP41. By the standard Trofile™ assay, these samples had CRT determined to be R5 (n=3,696), DM (n=2,474) or X4 (n=283). Samples that contained a mixture at a position of interest were excluded. Fisher's Exact Test was used to define amino acid changes (relative to HXB2) which were associated with CRT, and P-values were corrected for multiple testing using the Benjamini-Hochberg method.
RESULTS: Forty-eight mutations were found to be significantly associated with CRT (P-value <0.05). Twenty-two mutations were associated with CCR5-use (R5) and 26 were found to be associated with CXCR4-use (DM/X4). The mutations were found throughout GP41, with the majority being on the extracelluar region. Two of these mutations, N42D and V38A, are known enfuvirtide resistance mutations and one, A30T, is associated with enhanced
enfuvirtide sensitivity. The two mutations most strongly associated with CXCR4 use were A30T and L34M (odds ratios of 11.2 and 5.4, respectively).
CONCLUSIONS: These findings support previous observations that that determinants of tropism may lie outside of V3. Further studies will be needed to confirm the degree to which these GP41 mutations contribute directly to CRT determination.