Alexander Thielen

Bioinformatics prediction of HIV coreceptor usage.

As sequencing technology and prediction algorithms improve, HIV genotyping and coreceptor usage prediction are likely to play an increasingly important role in guiding patient prognosis and treatment selection.

Deep sequencing to infer HIV-1 co-receptor usage: application to three clinical trials of maraviroc in treatment-experienced patients.

BACKGROUND The Maraviroc versus Optimized Therapy in Viremic Antiretroviral Treatment-Experienced Patients (MOTIVATE) studies compared maraviroc versus placebo in treatment-experienced patients with CCR5-using (R5) human immunodeficiency virus type 1 (HIV-1), screened using the original Trofile assay. A subset with non-R5 HIV infection entered the A4001029 trial. We retrospectively examined the performance of a genotypic tropism assay based on deep sequencing of the HIV env V3 loop in predicting virologic response to maraviroc in these trials. METHODS V3 amplicons were prepared from 1827 screening plasma samples and sequenced on a Roche/454 GS-FLX to a depth of >3000 sequences/sample. Samples were considered non-R5 if ≥2% of their viral population scored greater than or equal to -4.75 or ≤3.5 using the PSSM(x4/R5) or geno2pheno algorithms, respectively. RESULTS Deep sequencing identified more than twice as many maraviroc recipients as having non-R5 HIV, compared with the original Trofile. With use of genotyping, we determined that 49% of maraviroc recipients with R5 HIV at screening had a week 48 viral load <50 copies/mL versus 26% of recipients with non-R5. Corresponding percentages were 46% and 23% with screening by Trofile. In cases in which screening assays differed, median week 8 log₁₀ copies/mL viral load decrease favored 454. Other parameters predicted by genotyping included likelihood of changing to non-R5 tropism. CONCLUSIONS This large study establishes deep V3 sequencing as a promising tool for identifying treatment-experienced individuals who could benefit from CCR5-antagonist-containing regimens.

Population-based V3 genotypic tropism assay: a retrospective analysis using screening samples from the A4001029 and MOTIVATE studies.

Background:The MOTIVATE-1 and 2 studies compared maraviroc (MVC) along with optimized background therapy (OBT) vs. placebo along with OBT in treatment-experienced patients screened as having R5-HIV (original Monogram Trofile). A subset screened with non-R5 HIV were treated with MVC or placebo along with OBT in a sister safety trial, A4001029. This analysis retrospectively examined the performance of population-based sequence analysis of HIV-1 env V3-loop to predict coreceptor tropism. Methods:Triplicate V3-loop sequences were generated using stored screening plasma samples and data was processed using custom software (‘ReCall’), blinded to clinical response. Tropism was inferred using geno2pheno (‘g2p’; 5% false positive rate). Primary outcomes were viral load changes after starting maraviroc; and concordance with prior screening Trofile results. Results:Genotype and Trofile results were available for 1164 individuals with virological outcome data (N = 169 non-R5 by Trofile). Compared with Trofile, V3 genotyping had a specificity of 92.6% and a sensitivity of 67.4% for detecting non-R5 virus. However, when compared with clinical outcome, virological responses were consistently similar between Trofile and V3 genotype at weeks 8 and 24 following the initiation of therapy for patients categorized as R5. Conclusion:Despite differences in sensitivity for predicting non-R5 HIV, week 8 and 24 week virological responses were similar in this treatment-experienced population. These findings suggest the potential utility of V3 genotyping as an accessible assay to select patients who may benefit from maraviroc treatment. Optimization of the predictive tropism algorithm may lead to further improvement in the clinical utility of HIV genotypic tropism assays.

Deep V3 Sequencing for HIV Type 1 Tropism in Treatment-Naive Patients: A Reanalysis of the MERIT Trial of Maraviroc

BACKGROUND Deep sequencing is a highly sensitive technique that can detect and quantify the proportion of non-R5 human immunodeficiency virus (HIV) variants, including small minorities, that may emerge and cause virologic failure in patients who receive maraviroc-containing regimens. We retrospectively tested the ability of deep sequencing to predict response to a maraviroc-containing regimen in the Maraviroc versus Efavirenz in Treatment-Naive Patients (MERIT) trial. Results were compared with those obtained using the Enhanced Sensitivity Trofile Assay (ESTA), which is widely used in clinical practice. METHODS Screening plasma samples from treatment-naive patients who received maraviroc and efavirenz in the MERIT trial were assessed. Samples were extracted, and the V3 region of HIV type 1 glycoprotein 120 was amplified in triplicate and combined in equal quantities before sequencing on a Roche/454 Genome Sequencer-FLX (n = 859). Tropism was inferred from third variable (V3) sequences, with samples classified as non-R5 if ≥2% of the viral population scored ≤3.5 using geno2pheno. RESULTS Deep sequencing distinguished between responders and nonresponders to maraviroc. Among patients identified as having R5-HIV by deep sequencing, 67% of maraviroc recipients and 69% of efavirenz recipients had a plasma viral load <50 copies/mL at week 48, similar to the ESTA results: 68% and 68%, respectively. CONCLUSIONS Reanalysis of the MERIT trial using deep V3 loop sequencing indicates that, had patients originally been screened using this method, the maraviroc arm would have likely been found to be noninferior to the efavirenz arm.

Improved detection of CXCR4-using HIV by V3 genotyping: application of population-based and "deep" sequencing to plasma RNA and proviral DNA.

Background:Tropism testing should rule out CXCR4-using HIV before treatment with CCR5 antagonists. Currently, the recombinant phenotypic Trofile assay (Monogram) is most widely utilized; however, genotypic tests may represent alternative methods. Methods:Independent triplicate amplifications of the HIV gp120 V3 region were made from either plasma HIV RNA or proviral DNA. These underwent standard, population-based sequencing with an ABI3730 (RNA n = 63; DNA n = 40), or “deep” sequencing with a Roche/454 Genome Sequencer-FLX (RNA n = 12; DNA n = 12). Position-specific scoring matrices (PSSMX4/R5) (−6.96 cutoff) and geno2pheno[coreceptor] (5% false-positive rate) inferred tropism from V3 sequence. These methods were then independently validated with a separate, blinded dataset (n = 278) of screening samples from the maraviroc MOTIVATE trials. Results:Standard sequencing of HIV RNA with PSSM yielded 69% sensitivity and 91% specificity, relative to Trofile. The validation dataset gave 75% sensitivity and 83% specificity. Proviral DNA plus PSSM gave 77% sensitivity and 71% specificity. “Deep” sequencing of HIV RNA detected >2% inferred-CXCR4-using virus in 8/8 samples called non-R5 by Trofile, and <2% in 4/4 samples called R5. Conclusions:Triplicate analyses of V3 standard sequence data detect greater proportions of CXCR4-using samples than previously achieved. Sequencing proviral DNA and “deep” V3 sequencing may also be useful tools for assessing tropism.

Population-based sequencing of the V3-loop can predict the virological response to maraviroc in treatment-naive patients of the MERIT trial.

Background:MERIT was a randomized trial comparing maraviroc (MVC) + Combivir versus efavirenz (EFV) + Combivir in drug-naive patients screened as having R5 HIV-1 by the original Trofile assay (OTA). We retrospectively evaluated treatment response after rescreening for viral tropism using population-based V3-loop sequencing. Methods:HIV env V3-loop was amplified in triplicate using reverse transcriptase–polymerase chain reaction from stored screening plasma and sequenced. Automated base calling was performed using custom software (RECall) and tropism inferred by geno2pheno (5.75% false-positive rate). Tropism results by genotype were compared with those of OTA and Enhanced Sensitivity Trofile assay (ESTA), where all results were available (n = 876). Results:Approximately 8% of patients screened as having R5 virus by OTA were classified as having non-R5 virus by V3-loop genotyping. These patients were less likely to have early or sustained week-48 treatment response to MVC, but not EFV. When restricted to patients with R5 virus by genotype, reanalysis of the primary study endpoint (plasma viral load <50 copies/mL at week 48) showed noninferiority of MVC twice daily to EFV (67% vs. 68%). Rescreening by genotype and ESTA had 84% concordance; patients receiving MVC twice daily rescreened as having R5 virus had greater than 1 log10 copies per milliliter decrease in viral load over those rescreened as having non-R5 virus. Where genotype and ESTA screening results were discordant outcomes were similar. Conclusions:The exclusion of ∼8% of patients with CXCR4-using virus by population-based sequencing would likely have resulted in noninferior responses in the MVC twice-daily and EFV arms. Rescreening by ESTA and population-based sequencing predicted similar virological response.

Added value of deep sequencing relative to population sequencing in heavily pre-treated HIV-1-infected subjects.

Objective To explore the potential of deep HIV-1 sequencing for adding clinically relevant information relative to viral population sequencing in heavily pre-treated HIV-1-infected subjects. Methods In a proof-of-concept study, deep sequencing was compared to population sequencing in HIV-1-infected individuals with previous triple-class virological failure who also developed virologic failure to deep salvage therapy including, at least, darunavir, tipranavir, etravirine or raltegravir. Viral susceptibility was inferred before salvage therapy initiation and at virological failure using deep and population sequencing genotypes interpreted with the HIVdb, Rega and ANRS algorithms. The threshold level for mutant detection with deep sequencing was 1%. Results 7 subjects with previous exposure to a median of 15 antiretrovirals during a median of 13 years were included. Deep salvage therapy included darunavir, tipranavir, etravirine or raltegravir in 4, 2, 2 and 5 subjects, respectively. Self-reported treatment adherence was adequate in 4 and partial in 2; one individual underwent treatment interruption during follow-up. Deep sequencing detected all mutations found by population sequencing and identified additional resistance mutations in all but one individual, predominantly after virological failure to deep salvage therapy. Additional genotypic information led to consistent decreases in predicted susceptibility to etravirine, efavirenz, nucleoside reverse transcriptase inhibitors and indinavir in 2, 1, 2 and 1 subject, respectively. Deep sequencing data did not consistently modify the susceptibility predictions achieved with population sequencing for darunavir, tipranavir or raltegravir. Conclusions In this subset of heavily pre-treated individuals, deep sequencing improved the assessment of genotypic resistance to etravirine, but did not consistently provide additional information on darunavir, tipranavir or raltegravir susceptibility. These data may inform the design of future studies addressing the clinical value of minority drug-resistant variants in treatment-experienced subjects.

Dynamic escape of pre-existing raltegravir-resistant HIV-1 from raltegravir selection pressure

Using quantitative deep HIV-1 sequencing in a subject who developed virological failure to deep salvage therapy with raltegravir, we found that most Q148R and N155H mutants detected at the time of virological failure originated from pre-existing minority Q148R and N155H variants through independent evolutionary clusters. Double 148R+N155H mutants were also detected in 1.7% of viruses at virological failure in association with E138K and/or G163R. Our findings illustrate the ability of HIV-1 to escape from suboptimal antiretroviral drug pressure through selection of pre-existing drug-resistant mutants, underscoring the importance of using fully active antiretroviral regimens to treat all HIV-1-infected subjects.

Improved Prediction of HIV-1 Coreceptor Usage with Sequence Information from the Second Hypervariable Loop of gp120

BACKGROUND Human immunodeficiency virus type 1 (HIV‐1) uses the CD4 receptor and a coreceptor to gain cell entry. Coreceptor usage is mainly determined by the V3 loop of gp120. Therefore, coreceptor usage is currently inferred from the genotype on the basis of V3 alone. However, several mutations outside V3 have been repeatedly reported to influence coreceptor usage. In this study, the impact of the V2 loop on coreceptor usage prediction was analyzed. METHODS Sequences were analyzed for differences at specific positions and position‐independent features with the Fisher exact and Student t tests. Prediction models were trained with support vector machines and evaluated in cross‐validation on clonal data. Models trained on the clonal data set were validated on 2 clinical data sets. RESULTS Several mutations and position‐independent features within V2 were statistically significantly different between R5 and X4 viruses. Cross‐validation on the clonal data set revealed a statistically significantly higher area under the receiver operating characteristic curve if features of both loops were used, compared with those using only V2 or V3 alone. Similar results were found with clinically derived data sets. CONCLUSIONS The ability of the V2 loop to improve coreceptor usage prediction has been shown in a large data set. Utilization of this information can lead to considerable improvements in the prediction of coreceptor use both on clonal data sets and on clinically derived data sets.

HIV-GRADE: A Publicly Available, Rules-based Drug Resistance Interpretation Algorithm Integrating Bioinformatic Knowledge

Background: Genotypic drug resistance testing provides essential information for guiding treatment in HIV-infected patients. It may either be used for identifying patients with transmitted drug resistance or to clarify reasons for treatment failure and to check for remaining treatment options. While different approaches for the interpretation of HIV sequence information are already available, no other available rules-based systems specifically have looked into the effects of combinations of drugs. HIV-GRADE (Genotypischer Resistenz Algorithmus Deutschland) was planned as a countrywide approach to establish standardized drug resistance interpretation in Germany and also to introduce rules for estimating the influence of mutations on drug combinations. The rules for HIV-GRADE are taken from the literature, clinical follow-up data and from a bioinformatics-driven interpretation system (geno2pheno[resistance]). HIV-GRADE presents the option of seeing the rules and results of other drug resistance algorithms for a given sequence simultaneously. Methods: The HIV-GRADE rules-based interpretation system was developed by the members of the HIV-GRADE registered society. For continuous updates, this expert committee meets twice a year to analyze data from various sources. Besides data from clinical studies and the centers involved, published correlations for mutations with drug resistance and genotype-phenotype correlation data information from the bioinformatic models of geno2pheno are used to generate the rules for the HIV-GRADE interpretation system. A freely available online tool was developed on the basis of the Stanford HIVdb rules interpretation tool using the algorithm specification interface. Clinical validation of the interpretation system was performed on the data of treatment episodes consisting of sequence information, antiretroviral treatment and viral load, before and 3 months after treatment change. Data were analyzed using multiple linear regression. Results: As the developed online tool allows easy comparison of different drug resistance interpretation systems, coefficients of determination (R2) were compared for the freely available rules-based systems. HIV-GRADE (R2 = 0.40), Stanford HIVdb (R2 = 0.40), REGA algorithm (R2 = 0.36) and ANRS (R2 = 0.35) had a very similar performance using this multiple linear regression model. Conclusion: The performance of HIV-GRADE is comparable to alternative rules-based interpretation systems. While there is still room for improvement, HIV-GRADE has been made publicly available to allow access to our approach regarding the interpretation of resistance against single drugs and drug combinations.