Improved methods for cross-sectional HIV incidence estimation are needed for surveillance of the epidemic and for evaluating the impact of HIV prevention interventions. We used a massively-multiplexed system, VirScan, to identify novel biomarkers for estimating the duration of HIV infection. This system uses phage display immunoprecipitation sequencing to quantify antibody (Ab) binding to >3,300 peptides spanning the HIV genome.
We analyzed 403 samples from 57 African women with known duration of infection (14 days to 8.7 years). Ab binding to each peptide in the VirScan library was quantified, and peptides with the strongest association of Ab binding with duration of HIV infection were identified. We used generalized estimating equations to analyze the association of Ab binding with duration of HIV infection, accounting for repeat sampling from the same individuals. An independent sample set from the same cohort was used to validate results (72 samples from 32 women infected 84 days to 9.1 years).
We identified 309 peptides that had a significant association between Ab binding and duration of infection (p<0.05 after adjusting for multiple comparisons); 266 peptides had increasing Ab binding over time and 43 peptides had decreasing Ab binding over time. Four peptides were selected for further analysis (two with increased binding: in gp120 and gp41; two with decreased binding: in gag and pol). The binding scores for these four peptides were combined in a simple, unweighted, linear model to estimate the duration of infection. This estimate was highly correlated with the observed (true) duration of infection (p <3 x 10-36 in the independent sample set). The predictive value of the 4-peptide ‘serosignature’ did not appear to be impacted by low viral load, low CD4 cell count, or HIV subtype. We also demonstrated that peptide engineering could be used to improve the association of Ab binding and duration of HIV infection.
Deep profiling of the antibody response to HIV infection identified novel peptide biomarkers for the duration of HIV infection. Peptides identified using this approach could be incorporated into simpler, high-throughput assays for cross-sectional HIV incidence estimation and other applications.