CROI 2018 Abstract eBook

Abstract eBook

Poster Abstracts

vs. 30%) and suffered more often from psychiatric problems such as depression (32% vs. 20%), mood swings (24% vs. 8%) and tiredness (14% vs. 4%). In the univariable models, several factors were associated with an increased risk for ADR (see figure): i.a. younger age, female gender, black ethnicity, having a language barrier, having any psychiatric problems or side effects, belonging to the heterosexual transmission group. In the multivariable model, unemployment (OR: 4.1, 95% CI: 1.9-8.6, p<0.001), limited education (OR 3.1, 95% CI 1.1-8.1, p=0.022) and psychiatric comorbidities or side effects (OR: 2.1, 95% CI: 1.1, 4.3, p=0.034) remained significantly associated with ADR. Patients with ADR and without any identified risk factor were rare (n=18). Conclusion: Although, ADR has become very rare with cART, patients in socially challenging life situations or patients with psychiatric problems are at higher risk for drug resistance. Prompt identification and adequate support of these patients may prevent further drug resistance acquisition.

Sequencing resolved by NGS clonal sequencing and codon variant calling. PASeq analysis also detected 186 additional amino acid mutations with median frequency of 3.420%(IQR:1.72-10.69%). Of these, none was found in the NL4.3 positive control and all PASeq detected mutations for this sample were also detected in Sanger Sequencing Conclusion: PASeq.org combined with NGS is an effective alternative for HIV NGS data analysis that provides Sanger-like information with improved resolution and increased sensitivity, by overcoming technical and data throughput limitations. 556 SMRT SEQUENCING OF FULL-LENGTH POL AMPLICONS TO INVESTIGATE HIV-1 DRUG RESISTANCE Sergio Ita , Caroline Ignacio, Sanjay Mohan, Venkatesh Kumar, Benjamin Murrell, Douglas D. Richman, Davey M. Smith, Gabriel A. Wagner University of California San Diego, La Jolla, CA, USA Background: Current HIV drug-resistance testing does not identify viral variants present at levels below 20% of the population and fails to simultaneously sequence the principal regions targeted by antiretroviral therapy (ART). Deep sequencing platforms have improved detection of minority drug resistant mutations (DRMs); however, these methods are limited by the relatively short length of the output reads (~150-400 base pairs). We used Pacific Biosciences single molecule real time (SMRT) sequencing to deep sequence en bloc the 3-kilobase pol gene among HIV-infected individuals, and monitor DRMs within full-length (FL) pol amplicons. Methods: Viral RNA was isolated and cDNA generated from a participant of the San Diego Primary Infection Resource Consortium. We optimized amplification of a 3,338-bp pol product using subtype-B-specific primers spanning RT, PR, and IN (HXB2 1736 to 5074). Libraries of each FL pol amplicon were sequenced on the PacBio RS II platform. For sequence analysis, we adapted FLEA, a pipeline for processing full-length HIV env PacBio data. Populations were characterized using interactive browser-based visualizations of sequence alignments and phylogenies, highlighting DRMs from the Stanford HIV Drug Resistance Database. Results: SMRT reads produced from a sample collected 385 days after the estimated date of infection and before initiation of ART (Figure 1) revealed a heterogeneous population with a single outlier variant (estimated frequency: 2.8%), which had APOBEC-induced mutations. Total pol nucleotide diversity (as measured by total tree span under a GTR model) was 2.35%, but reduced to 0.8% after excluding the APOBEC-mutated variant. M184I was observed in this APOBEC-mutated variant, along with 18 other APOBEC mutations and 9 premature stop codons, rendering this variant defective. M184I was not found on any other haplotype variants and no other DRMs were found within the viral population. Conclusion: We showed that SMRT sequencing of plasma HIV-1 RNA: a) generated deep coverage of FL pol amplicons, b) detected DRMs within a defective variant, and c) did not reveal minority DRMs in FL reads, consistent with the local epidemic. Long-read deep sequence platforms, such as PacBio, provide a sensitive method for linkage analysis across the entire length of HIV-1 pol. Further studies could apply this methodology to detect co-evolution of DRMs.

Poster Abstracts

555 PASEQ: ONE-CLICK, CLOUD-BASED WEB SERVICE FOR NGS-BASED HIV GENOTYPING DATA ANALYSIS Jordi Herms 1 , Cristina Rodriguez 2 , Maria Casadella 2 , Teresa Puig 2 , Bonaventura Clotet 2 , Roger Paredes 2 , Marc Noguera-Julian 2 1 Fundació Lluita Contra la Sida, Badalona, Spain, 2 IrsiCaixa Institute for AIDS Research, Badalona, Spain Background: Cost-effective HIV drug resistance testing is needed. Next Generation Sequencing (NGS) technologies produce high quality results at an optimized cost[1]. However, the need for bioinformatics analysis of large data volumes is one of the main bottlenecks for the adoption of NGS in HIV DR testing. Methods: We developed an automatic analysis pipeline, PASeq: Polymorphism Analysis by Sequencing. Briefly, raw NGS data is quality&contamination filtered, aligned to a pre-defined reference (HXB2) and codon-level queried for amino acid variants. Mutations are programmatically used to query Stanford HIVdb for resistance interpretation. All results are interactively provided both through the web and a pdf report. PASeq runs on a computational cloud and can scale up to virtually an unlimited number of concurrent analyses with a <1 hour mean runtime. PASeq provides a one-click friendly user web interface, creates printable reports and streamlines data submission and interpretation. Additionally, PASeq stores results in a structured database that can be exploited for HIV epidemiology and is freely available at www.paseq.org. 26 HIV routinely genotyped plasma and 1 NL4.3 plasmid samples were sequenced using Illumina (MiSeq/Nextera-XT protocol). Illumina raw data was directly uploaded and analyzed through PASeq.org, with a 1% lower threshold. Protease and positions 1-335 in reverse transcriptase were compared Results: High quality Sanger and Illumina NGS data was obtained for all samples. PASeq analysis took approximately 1h time. Median coverage for Illumina data was 7662(IQR:6325-9739). All except 8/11733 (<0.001%) amino acid Sanger calls were also detected by PASeq in Illumina data. Thorough re- inspection of these 8 discrepancies showed that 4 corresponded to low Sanger quality sequences, 3 within the same sample, while the other 4 discrepant readouts corresponded to dual mixes, a known technical limitation of Sanger

CROI 2018 204