Immune correlates analysis of the Imbokodo (HVTN 705/HPX2008) efficacy trial of a mosaic HIV-1 vaccine regimen evaluated in Southern African people assigned female sex at birth: a two-phase case-control study

Abstract

Background: The HVTN 705 Imbokodo trial of 2636 people without HIV and assigned female sex at birth, conducted in southern Africa, evaluated a heterologous HIV-1 vaccine regimen: mosaic adenovirus 26-based vaccine (Ad26.Mos4.HIV) at Months 0, 3, 6, 12 and alum-adjuvanted clade C gp140 at Months 6, 12. Per-protocol vaccine efficacy (VE) against HIV-1 diagnosis from seven to 24 months was 14.1% (95% CI: -22.0% to 39.5%). Immune correlates analysis was performed for markers selected based on prior evidence in efficacy trials and/or nonhuman primate models.

Methods: Humoral and cellular immune response markers at Month 7 were evaluated as immune correlates of risk and of protection in a breakthrough case-control cohort (n = 52 cases, 246 non-cases). Primary markers were IgG binding to vaccine-strain gp140, IgG3 binding to diverse Env antigens (IgG3 Env breadth), IgG3 binding to diverse V1V2 antigens (IgG3 V1V2 breadth), antibody-dependent phagocytosis against the vaccine-strain gp140, Env-specific CD4+ and CD8+ T-cell responses, and multi-epitope functions.

Findings: No immune markers were statistically significant correlates of risk. IgG3 V1V2 breadth trended toward an inverse association: hazard ratio 0.70 (95% CI: 0.36 to 1.35; p = 0.29) per 10-fold increase and 0.51 (95% CI: 0.21 to 1.24; p = 0.14) in a Cox model with all primary markers. The VE estimate was 11.8% (95% CI: -17.9% to 34.0%) at all IgG3 V1V2 breadth values below 667 weighted geometric mean net MFI; just above this value, the VE estimate sharply increased to 62.6% (95% CI: -17.9% to 89.6%), and further increased to 80.9% (95% CI: -17.9% to 99.5%) at 1471 MFI, the 95th percentile of the marker distribution. Mediation analysis yielded a VE of 35.7% (95% CI: 15.0% to 51.3%) attributable to the vaccine's impact on this marker.

Interpretation: The trend in association of greater IgG3 V1V2 antibody breadth with lower likelihood of HIV acquisition is consistent with the identification of antibodies against V1V2 as immune correlates in three other HIV vaccine efficacy trials and suggests that a greater emphasis should be placed on studying this region in the HIV-1 envelope as a vaccine immunogen.

Funding: National Institute of Allergy and Infectious Diseases and Janssen Vaccines & Prevention BV.

Keywords: Ad26.Mos4.HIV vaccine regimen; Binding antibodies; Correlates of protection; Correlates of risk; IgG3 V1V2 antibodies; Maximal signal diversity-weighted average.

Fig. 1

Distribution of the seven primary immune markers (measured at Month 7) in per-protocol participants, stratified by randomisation arm (vaccine vs. placebo) and case/non-case outcome status. Markers include (A) IgG gp140 C97ZA, (B) IgG3 V1V2 breadth, (C) IgG3 gp120+gp140 breadth, (D) ADCP gp140 C97ZA, (E) CD4+ IFN-g and/or IL-2 Env, (F) CD8+ IFN-g and/or IL-2 Env, and (G) Multi-epitope functions: the maximum diversity-weighted average of the first six primary markers and of ADCP gp140 Mos1, IgG V1V2 breadth, and ADCC partial area-under-the-curve against each strain CAP8, CH58, WITO. Each violin plot contains a boxplot showing the estimated 25th, 50th, and 75th percentiles of the marker distribution, as well as a (rotated) kernel density estimate of the marker probability density function. Boxplots are based on observed response magnitudes among responders (filled coloured circles) and nonresponders (empty grey triangles). Panels show the actual numbers of participants with available assay data above the response rates (if relevant) and median or geometric mean (GM) values. The response rates and median or GM values accounted for inverse-probability-of-sampling weights and thus are for the population of eligible participants from which the case/control study cohort was randomly sampled. ADCC, antibody-dependent cellular cytotoxicity; ADCP, antibody-dependent cellular phagocytosis; Env, envelope; GM, geometric mean; gp, glycoprotein; HIV, human immunodeficiency virus; IFN, interferon; IgG, immunoglobulin G; IL, interleukin; MFI, mean fluorescence intensity; V1V2, variable regions V1 and V2.

Fig. 2

(A–H) Reactivity in per-protocol vaccine recipients of IgG3 antibodies to the eight exploratory individual V1V2 antigens (measured at Month 7) that comprised the IgG3 V1V2 breadth primary marker, stratified by case/non-case outcome status; (I, J) magnitude-breadth curves of Month 7 responses, stratified by case/non-case outcome status, for breadth of antibodies reactive with antigens with preferential reactivity to V2i (I) or V2p (J) antigens; and (K) IgG3 95th percentile levels (net MFI) against each V1V2 antigen. Markers include (A) IgG3 gp70–001428.2.42 V1V2, (B) IgG3 gp70–1012.11.TC21.3257 V1V2 (C), IgG3 gp70-B.CaseA2 V1V2, (D) IgG3 gp70-1394C9G1 V1V2, (E) IgG3 gp70-BF1266 431a V1V2, (F) IgG3 gp70-Ce1086 B2 V1V2, (G) IgG3 C.1086C V1V2 Tags, and (H) IgG3AE.A244 V1V2 Tags 293F and are ordered from highest to lowest mean AUC difference V2p versus V2i mAbs (Figure S20). Each violin plot contains a boxplot showing the estimated 25th, 50th, and 75th percentiles of the marker distribution, as well as a (rotated) kernel density estimate of the marker probability density function. Boxplots are based on observed response magnitudes among responders (filled coloured circles) and nonresponders (empty grey triangles). Panels show the actual numbers of participants with available assay data above the response rates (if relevant) and median or geometric mean (GM) values. The response rates and median or GM values accounted for inverse-probability-of-sampling weights and thus are for the population of eligible participants from which the case/control study cohort was randomly sampled. LLoQ: lower limit of quantitation. Magnitude-breadth curves are shown for (I) the antigens with greater reactivity against V2i monoclonal antibodies [(A), (B), (C)] and (J) the antigens with a relatively lower reactivity for V2i monoclonal antibodies and higher reactivity with V2p monoclonal antibodies [(D)–(H)]. (K) includes the maximal-signal diversity weights assigned to the 8 antigens in the IgG3 V1V2 breadth primary marker. AUC, area under the curve; CI, confidence interval; GM, geometric mean; gp, glycoprotein; HIV, human immunodeficiency virus; IgG, immunoglobulin G; MFI, mean fluorescence intensity; V1V2, variable regions V1 and V2; V2i, β-sheet conformation; V2p, α-helical conformation.

Fig. 3

Analyses of Month 7 IgG3 V1V2 breadth score and Month 7 IgG3 gp70-001428.2.42 V1V2 as correlates of risk and correlates of protection. (A and C) Covariate-adjusted cumulative incidence of HIV-1 by Day 550 (post Month 7) for subgroups of per-protocol vaccine recipients with (A) an IgG3 V1V2 breadth score or (C) reactivity of IgG3 antibodies to the gp70-001428.2.42 V1V2 antigen above a given threshold value, estimated using nonparametric threshold regression. The blue dots represent point estimates at each new HIV-1 acquisition endpoint, and the black lines represent a linear interpolation of these points. The grey shaded area represents pointwise 95% confidence bands, and the upper boundary of the green shaded area represents the estimated reverse cumulative distribution function (CDF) of the marker. The estimates and CIs assume a nonincreasing threshold-response function. The red vertical dashed line is the marker threshold corresponding to the largest observed HIV-1 event time (within the time frame of interest). (B) Controlled vaccine efficacy (CVE) as a function of IgG3 V1V2 breadth score, controlling for covariates. Estimates are computed separately using a Cox proportional hazards model (solid purple line) and a nonparametric method that assumes that the CVE curve is nondecreasing (solid blue line). Corresponding 95% pointwise confidence bands are shown in shaded blue/purple, with dotted lines representing the corresponding lower and upper limits. The distribution of the marker, estimated using KDE, is plotted in orange with the addition of a rectangle at the lower limit representing a point mass at the marker minimum value. The solid grey line represents estimated overall VE, and the dotted grey lines represent the corresponding 95% CI limits. The Cox model estimates are cut off at the 97.5th quantile of the marker distribution and the nonparametric estimates are cut off at the 95th quantile. (C) and (D) are analogous to (A) and (B), respectively, but for the Month 7 IgG3 gp70–001428.2.42 V1V2 marker. CI, confidence interval; CVE, controlled vaccine efficacy; gp, glycoprotein; HIV, human immunodeficiency virus; IgG, immunoglobulin G; KDE, kernel density estimation; LLOQ, lower limit of quantitation; MFI, mean fluorescence intensity; V1V2, variable regions V1 and V2; VE, vaccine efficacy. Baseline covariates adjusted for: region (RSA versus Non-RSA), age, body mass index, baseline risk score.