CROI 2017 Abstract e-Book

Abstract eBook

Poster and Themed Discussion Abstracts

196 BLIND DATING: PHYLOGENETIC DETERMINATION OF LATENT HIV SEQUENCE AGES WITHIN HOST Bradley R. Jones 1 , Joshua Horacsek 1 , Jeffrey Joy 2 , Zabrina Brumme 1 , Art Poon 2 1 Simon Fraser Univ, Burnaby, British Columbia, Canada, 2 Univ of British Columbia, Vancouver, British Columbia, Canada

Background: The ability of HIV to persist within latent cellular reservoirs represents a major barrier to cure. The timing of establishment of individual viral reservoirs over the course of an infection may influence their susceptibility to elimination by immune-mediated or therapeutic approaches. However, methods to accurately estimate the age of reservoir sequences remain scarce. We propose a simple method to date suspected reservoir sequences using phylogenetically informed regression. Methods: Our method is as follows: Taking aligned HIV sequences from a single individual sampled over multiple time points, we reconstruct a maximum-likelihood phylogeny. The location of the root in the phylogeny is determined using root-to-tip regression. Finally, the root-to-tip distances of possibly latent sequences are mapped to the optimal regression line to estimate reservoir establishment dates. To validate our method, we simulated HIV sequence evolution under a standard model and reconstructed phylogenies from these data by maximum likelihood. Latency was simulated by shortening the branch lengths of up to 50% of the tips in the tree. We employed this same approach to validate the method on HIV RNA sequences, which provided a more realistic assessment of sources of error than direct simulation. Results: We were successfully able to extract HIV reservoir ages from simulated data. When this method was applied to HIV DNA sequences in phylogenies containing dated HIV RNA sequences, we observed that the predicted dates significantly preceded dates of sampling, which was consistent with latency. Conclusion: Our results provide evidence that the date that a HIV lineage entered a latent state can be recovered from phylogenetic analysis of HIV RNA sequence variation from the same patient. These dates can provide important insights into the dynamics of HIV reservoirs within hosts. 197 HIV-1 SEQUENCES FROM EARLY INFECTION PREDICT THE AGE OF THE INFECTION Morgane Rolland 1 , Christopher L. Owen 1 , Sodsai Tovanabutra 1 , Eric Sanders-Buell 1 , Lucas Maganga 2 , Hannah Kibuuka 3 , Sorachai Nitayaphan 4 , Fredrick Sawe 5 , Merlin L. Robb 1 , Nelson L. Michael 1 1 US Military HIV Rsr Prog, Silver Spring, MD, USA, 2 MMRP, Mbeya, Tanzania, United Republic of, 3 Makerere Univ Walter Reed Proj, Kampala, Uganda, 4 AFRIMS, Bangkok, Thailand, 5 Kenya Med Reseach Inst/Walter Reed Proj, Kericho, Kenya Background: Being able to infer the age of an HIV infection based on sequences is an invaluable tool, as biological samples often lack information on the date of transmission. Phylogenetic dating methods have been developed and tested with simulated data, but they have not been compared against real data with precise estimates of the date of infection. The RV217 cohort allows such comparisons since it enrolled more than 2,300 seronegative individuals to identify 115 acute HIV infections based on RNA tests performed twice weekly. Methods: HIV genomes were sequenced from plasma samples following PCR amplification by endpoint-dilution. Phylogenies, divergence times, molecular rates of substitution and effective population sizes were estimated in BEASTv1.8.2 using uncorrelated lognormal relaxed-clock rates and the Bayesian Skyride tree prior for 250 million generations. Results: For the 36 subjects in this study, the last negative HIV test occurred a median of 4 days before HIV diagnosis (range: 2-32 days), and a total of 1,190 HIV genomes were sequenced at a median (range) of 5 (1-15), 32 (27-42) and 170 (132-261) days after HIV diagnosis. Insights into the history of an HIV infection can be gained by reconstructing a molecular time-scale of its evolution. HIV in early infection did not conform to a strict molecular clock (average coefficient of variation = 0.783). HIV diversification processes varied greatly between individuals: the median substitution rate ranged between 7.26e-07 and 1.22e-04 (average = 1.55e-05). Ten of 36 individuals were infected with multiple HIV founder variants and they tended to show higher substitution rates (2.77 e-05) than subjects with single founders (1.07 e-05) (p = 0.05). For infections established with multiple founders, estimates for the median age of the infection were not accurate: they were significantly higher than for infections with single founders (366 vs. 204 days, p < 0.001), reflecting that the distinct founder variants in a given individual were sampled from a previous host in chronic infection. Among the 26 subjects with single founders, Bayesian estimates suggested that the infection started at a median of 7 days prior to the last negative HIV test (IQR= 15 day prior to 1 day after). For all but one subject, the 95% HPD encompassed the transmission window. Conclusion: These results show that Bayesian coalescence methods allow to accurately date an HIV infection event, providing estimates that can be used when the date of HIV transmission is unknown. 198 USING PRIMER ID DEEP SEQUENCING TO IDENTIFY RECENT HIV INFECTION IN NORTH CAROLINA Shuntai Zhou 1 , Christopher Sellers 1 , Marc Potempa 1 , Scott J. Zimmerman 2 , Erkia Samoff 3 , William C. Miller 4 , Joseph J. Eron 1 , Myron S. Cohen 1 , Ronald Swanstrom 1 , Ann M. Dennis 1 1 Univ of North Carolina at Chapel Hill, Chapel, NC, USA, 2 North Carolina State Lab of Pub Hlth, Raleigh, NC, USA, 3 North Carolina Div of Pub Hlth, Raleigh, NC, USA, 4 Ohio State Univ, Columbus, OH, USA

Poster and Themed Discussion Abstracts

Background: Identifying recent infection is critical for monitoring HIV transmission and incidence but remains challenging. With advances in next generation sequencing (NGS) and Primer ID (PID), intra-host viral diversity biomarkers can be efficiently evaluated over multiple genomic regions and scaled to the population level. We assessed diversity over time from subjects with known infection dates (model) then applied our findings to remnant diagnostic sera from newly diagnosed, antibody positive infections reported in 2014 at the NC State Lab of Public Health (NCSLPH). Methods: Single and longitudinal plasma specimens from ART-naïve CHAVI-001 (n=25; Feibig stage 1-4 at diagnosis) and chronically infected (n=7; sampled >1 year from diagnosis) subjects were obtained for the model. We constructed multiplexed PID libraries and sequenced RT and env V1-V3 using Illumina MiSeq. We used a PID pipeline to construct a consensus sequence (TCS) for each starting template, and calculated % pairwise diversity (π) for each region. Multi-variant or super-infection was excluded by phylogenetic analysis. We evaluated mean π over time and ROC area under the curve (AUC) to distinguish performance for different gene regions. Recent infection (<9 months since infection) was estimated for NCSLPH specimens. Results: In the model, 68 samples were analyzed from 19/25 CHAVI (4 dual infections excluded; 2 had low RNA) and 7 chronically infected subjects. CHAVI samples ranged from 0-1367 days since seroconversion (52%were ≤6 months, 20% 6-12 months, 28%>12 months). Mean π increased significantly over time for all regions including RT, 3rd codon RT only, V1-V3, and C2/V3 only (Fig. 1A). The combined π at RT and V1-V3 (Fig. 1B) showed best discrimination for recent definitions at 9 months (AUC=0.91, 95% CI 0.85-0.97) and 12 months (AUC=0.91, CI 0.84-0.98). Among NCSLPH specimens, 70/91 had sufficient consensus at both RT and V1-V3. Subjects were representative of new HIV diagnoses in NC: 65%MSM, 61% black/AA, 66%≤30 years. Of these, 25/70 (36%) were identified as recent infection at 9 months at combined π threshold of 0.80% (sensitivity 82%, specificity 91% in model).

77

CROI 2017