Michal Juraska, Li Li, Craig A Magaret, James Peng, Elena E Giorgi, James Ludwig, Lindsay N Carpp, Allan C deCamp, Jia J Kee, Cindy Molitor, Marthin Mandig, Paul T Edlefsen, Raabya Rossenkhan, Dylan H Westfall, Wenjie Deng, Lennie Chen, Hong Zhao, Bruna Galvao, Craig Adams, Anna Yssel, David Matten, Talita York, Asanda Gwashu-Nyangiwe, Nonkululeko Ndabambi, Ruwayhida Thebus, Edith M Swann, John A Hural, Janine van Duijn, Susan Buchbinder, Frank Tomaka, Lawrence Corey, Kathryn Mngadi, Glenda E Gray, Peter B Gilbert, Carolyn Williamson, James I Mullins
Despite extensive prespecified and exploratory analyses including Env features supported by prior studies to potentially impact VE, there was only limited, weak evidence that Env sequence features modified VE in HVTN 705. Although previous work suggested a protective role of IgG3 binding to V1V2 in a small subgroup of vaccine recipients, a V1V2 sieve signal was absent.
BACKGROUND: A heterologous Ad26.Mos4.HIV and clade C gp140 vaccine regimen did not show overall significant efficacy against HIV-1 acquisition [point estimate 14.1%; 95% confidence interval (CI), -22.0 to 39.5] in the HVTN 705/HPX2008 trial in southern African women. We examined whether and how vaccine efficacy (VE) against HIV-1 diagnosis over 7-24 months post-first dose varied by HIV-1 Envelope (Env) amino acid sequence features.
METHODS: HIV-1 viral sequences were generated by PacBio SMRT-UMI sequencing from the first RNA-positive sample of participants who acquired HIV-1. Env amino acid sequence features were prespecified for analyses based on 1) being hypothesised to impact VE; and 2) having sufficient variability. Sieve analyses assessed VE by a single representative sequence and by viral population composition.
FINDINGS: The majority of Env features showed no evidence of differential VE, with only two signals having familywise error rate (FWER) P-values <0.10. In single-sequence analyses, VE declined with increasing physicochemical-weighted Hamming distance from the C97ZA vaccine insert in clade C broadly neutralising antibody resistance-associated signature positions (FWER P = 0.08). Sequence-predicted Env structural features showed no significant vaccine vs. placebo differences in structural divergence from the C97ZA vaccine-insert Env sequence. In multi-sequence analyses (median 121 sequences/individual), VE was higher against viral populations with ≥99% vs. <99% L832 prevalence (VE = 91.7%; 95% CI, 67.4-97.9 vs. VE = -7.0%; 95% CI, -55.5 to 26.4) (unadjusted P = 0.0002 for differential VE, FWER P = 0.023).
INTERPRETATION: Despite extensive prespecified and exploratory analyses including Env features supported by prior studies to potentially impact VE, there was only limited, weak evidence that Env sequence features modified VE in HVTN 705. Although previous work suggested a protective role of IgG3 binding to V1V2 in a small subgroup of vaccine recipients, a V1V2 sieve signal was absent.
FUNDING: National Institutes of Health and Johnson & Johnson.