Correlates of HIV-1 control after combination immunotherapy

Abstract

The identification of therapeutic strategies to induce sustained antiretroviral therapy (ART)-free control of HIV infection is a major priority¹. Combination immunotherapy including HIV vaccination, immune stimulation, latency reversal and passive transfer of broadly neutralizing antibodies (bNAbs) has shown promise in non-human primate models^2-6, but few studies have translated such approaches into people. Here we performed a single-arm, proof-of-concept study in ten people living with HIV on ART, combining the following three approaches: (1) therapeutic vaccination with an HIV Gag conserved element-targeted DNA + IL-12 prime/modified vaccinia Ankara (MVA) boost regimen followed by (2) administration of two bNAbs (10-1074, VRC07-523LS) and a toll-like receptor 9 agonist (lefitolimod) during ART suppression, followed by (3) repeat bNAb administration at the time of ART interruption (Clinicaltrials.gov: NCT04357821 ). Seven out of the ten participants exhibited post-intervention control after pausing ART, independent of residual bNAb plasma levels. Robust expansion of activated CD8⁺ T cells early in response to rebounding virus correlated with a lower median viral load after peak viraemia off ART. These data suggest that combination immunotherapy approaches might prove effective in inducing sustained control of HIV by slowing rebound and improving CD8⁺ T cell responses, and that these approaches should continue to be optimized.

Fig. 1. Combination immunotherapy promotes post-intervention control of HIV.

a, Study schema: administration of combination immunotherapy including a DNA vaccine targeting CEs of HIV Gag, two bNAbs with a TLR9 agonist in ten people living with HIV on ART, followed by a second infusion of bNAbs immediately before an ATI. b, The time to HIV rebound after ART discontinuation. c, Plasma viral load rebound kinetics from the first day of rebound (x = 0); the numbers on the graph indicate set point (the median viral load from 2 weeks after peak viral load to the time of ART restart); the colours indicate the rebound phenotype (red, non-control/typical rebound; blue, viraemic post-intervention control; grey, no rebound for over 18 months off ART). The shape indicates the timing of ART after HIV acquisition: circle, chronic infection; triangle, acute infection (<1 month); square, early infection (1–6 months). End points indicate the day of ART restart, except for the participant with no rebound (grey). The empty square represents a participant with an HLA-B*57 allele. d, Comparison of the last available viral load before ART initiation (single value) versus the post-ART set point. Statistical analysis was performed using two-sided Wilcoxon signed-rank tests. e, HIV rebound slope (log₁₀-transformed copies per ml per day) from the time of rebound to the time of peak viral load after pausing ART in people who were known controllers before starting ART (prior controller, n = 7) or not (prior non-controller, n = 13) in the absence of any immunotherapeutic intervention, versus non-controllers (n = 3) and viraemic post-intervention controllers (n = 6) from this trial. LOD, limit of detection; NC, non-control; PIC, post-intervention control; UD, undetectable. Source data

Fig. 2. bNAb PK and susceptibility predict time to HIV rebound but not post-ART set point.

a, Viral load and modelled bNAb concentrations over time after the second dose of bNAbs in each participant; bNAb line colours: black, 10-1074; grey, VRC07-523LS. At the top of each graph, phenotypic susceptibility of autologous HIV to neutralization by bNAbs (IC₉₀) is depicted in pairs of shaded rectangles (left rectangle, 10-1074; right rectangle, VRC07-523LS). Susceptibility was measured from both cell-associated provirus at baseline (depicted in the left-most rectangles in each graph; grey indicates that the value could not be determined (ND)) as well as from post-rebound plasma virus at multiple timepoints, as indicated. b, The correlation between 10-1074 and VRC07-523LS exposure (area under the bNAb concentration–time curve (AUC) after the second dose of bNAbs) and time to rebound. c, The correlation between bNAb susceptibility (IC₉₀) after rebound and the time to rebound after ATI. d, The correlation between bNAb exposure and the post-rebound viral load set point. Statistical analysis was performed using two-sided Spearman’s correlation (b–d). Source data

Fig. 3. The impact of combination immunotherapy on the HIV reservoir and HIV-specific T cell responses.

a, Longitudinal peripheral blood sampling timepoints (applies to Figs. 3–5). pVL, plasma viral load. b,c, CD4⁺ T cell-associated potentially intact HIV DNA, as measured by IPDA (b) and HIV RNA (transactivating response (TAR) region, indicating total initiated HIV transcripts) (c). d, The magnitude of IFNγ⁺ CE-specific CD4⁺ and CD8⁺ T cells as measured by ICS; the numbers below the x axis indicate the proportion of participants with detectable CE-specific responses at each timepoint. e, The magnitude of total (Gag + Pol + Nef + Env) HIV-specific CD8⁺ T cell responses by ICS. Statistical analysis was performed using linear mixed-effect analysis with Tukey’s multiple-comparison test. Timepoints (applies to all Figs): BL, baseline before interventions (that is, on ART); post-prime, the day of MVA vaccination (>8 weeks after last DNA vaccination); post-boost, 2 weeks after MVA vaccination; pre-LEF, immediately before lefitolimod dosing; pre-ATI, immediately before ATI; pre-R, the last PBMC sampling timepoint available before rebound (sampled within 1–4 weeks before HIV rebound); post-R1, the first PBMC sampling timepoint after rebound (for participants with a sample available ≤28 days after rebound, at a low viral load; all <3,000 copies per ml). NS, not significant. Source data

Fig. 4. The sequential activation of innate and adaptive immune cells peri-rebound after combination immunotherapy.

a, The subset of manually gated immune cell features of which the abundance changed significantly in the six viraemic post-intervention controllers (PICs) between baseline compared with pre-rebound (top), baseline compared with the first post-rebound timepoint (bottom) or baseline compared with both timepoints (middle), as measured using CyTOF. Statistical analysis was performed using the two-sided Wilcoxon signed-rank test; *P < 0.05. The features include cell types and phenotypes as listed in Supplementary Table 4; median values are z scored by feature. b, Participant-level changes in a subset of immune features from a; the plots also show data from non-controllers (red) and the aviraemic controller (grey). c, Gating (left) and longitudinal assessment (right) of total plasmablast abundance. d, Representative plot (left) and longitudinal assessment (right) of the frequency of T-bet⁺ plasmablasts. Timepoint definitions are provided in Fig. 3. For PID 60610 (aviraemic controller), data shown at the pre-R timepoint were generated from samples collected 44 weeks into the ATI. ABC, activated B cells; ASC, antibody-secreting cells; T_H, T helper cells; T_FH, T follicular helper cells; T_TM, transitional memory T cells; T_ctl, cytotoxic CD4⁺ T cells. Source data

Fig. 5. Post-intervention control of HIV is associated with a robust stem/memory-like, activated/proliferating CD8⁺ T cell response to rebound.

a,b, The gating strategy (a) and the frequency of Ki-67⁺ non-naive CD8⁺ T cells by manual gating at the baseline, pre-R and post-R1 timepoints as measured using CyTOF (b) (timepoint definitions are shown in Fig. 3). c, The distribution of Ki-67⁺ cells across CD8⁺ T cell subsets at the post-R1 timepoint. T_N, naive (CD45RA⁺CCR7⁺CD95⁻); T_SCM, stem cell memory (CD45RA⁺CCR7⁺CD95⁻); T_CM, central memory (CD45RA⁻CCR7⁺), T_TM, transitional memory (CD45RA⁻CCR7⁻CD27⁺), T_EM, effector memory (CD45RA⁻CCR7⁻CD27⁻), T_EMRA, CD45RA-expressing effector memory (CD45RA⁺CCR7⁻). d, The frequency of TCF-1⁺ cells within the Ki-67⁺ non-naive CD8⁺ T cells at the post-R1 timepoint. e, Uniform manifold approximation and projection (UMAP) analysis of non-naive CD8⁺ T cells including cells from all participants at the baseline, pre-ATI, pre-R and post-R1 timepoints (left). Right, expression of individual markers on the UMAP. f, The scaled median expression of markers across all clusters (activation markers CD38, HLA-DR and Ki-67 are boxed for emphasis). g, The frequencies of clusters from f of non-naive activated CD8⁺ T cells across the baseline, pre-R and post-R1 timepoints; c9, c12, c17, c8 and c14 are activated and express CD38 ± HLA-DR ± Ki-67; cells in c3 have a T_EMRA phenotype. h, Two-sided Spearman correlations between each participant’s post-ART set point viral load and the frequency of select non-naive CD8⁺ T cell populations at the post-R1 timepoint. The analysis excludes PID 60610 (no rebound) and 35933 (no post-R1 sample). The dashed boxes in b, c and g encompass datapoints from post-intervention controllers. Source data

Extended Data Fig. 1

CONSORT diagram.

Extended Data Fig. 2. bNAb exposure and susceptibility do not relate to post-intervention control outcomes.

Comparison between non-controllers (NC) and post-intervention controllers (PIC) in 10-1074 and VRC07-523LS a, b) exposure (AUC, area under the bNAb concentration-time curve following the second dose of bNAbs), and c, d) phenotypic susceptibility of autologous HIV to neutralization by bNAbs (inhibitory concentration [IC₉₀]). Source data

Extended Data Fig. 3. Longitudinal HIV reservoir measurements and HIV-specific T cell responses.

a) Cell-associated 3′ and 5′ defective HIV DNA in peripheral blood CD4⁺ T cells by IPDA. b) Longitudinal cell-associated HIV RNA transcript measurements. c) IPDA measurements from gut tissue in select participants (ND, not detected). All sampled post-ATI. All participants were aviremic at the time of sampling. On ART: 35933, 10227, 71190. Off ART: 60610. d) Intracellular cytokine staining (ICS) gating scheme. e) Longitudinal phenotypes of HIV Gag/CE-specific (IFNγ⁺) CD8⁺ T cells. Missing points reflect timepoints with no or too few (<50) CE-specific CD8⁺ T cells to phenotype. f) Frequency of Nef-specific (IFNγ⁺) CD4⁺ and CD8⁺ T cells. g) PD-1 expression on Nef-specific CD8⁺ T cells. h) Frequency of total (IFNγ⁺) HIV-specific CD4⁺ T cells by ICS (sum: Gag+Pol+Nef+Env). IPDA, intact proviral DNA assay. Timepoints (applies to all figures): BL, baseline prior to interventions (i.e., on ART); Post-prime, day of MVA vaccination (>6 weeks after last DNA vaccination); Post-boost, 2 weeks after MVA vaccination; pre-LEF, prior to lefitolimod dosing; pre-ATI, prior to ATI; pre-R, last PBMC sampling timepoint available prior to rebound (sampled within 1-4 weeks prior to HIV rebound). Note: for PID 60610 (aviremic controller), the data shown at the “pre-R” timepoint was evaluated 44 weeks into the ATI. IFNγ, interferon gamma. ICS, intracellular cytokine staining. Source data

Extended Data Fig. 4. Minimal differences in baseline immune features between post-intervention controllers and non-controllers.

a) Differentially abundant CyTOF features in the peripheral blood (by two-sided Wilcoxon rank-sum test with Benjamini-Hochberg adjustment for multiple comparisons), and b) differentially expressed genes in total peripheral blood mononuclear cells (PBMCs) between the viremic post-intervention controllers (n = 6) and non-controllers (n = 3) at baseline (on ART, prior to interventions). Differential expression analysis was performed using lmfit through limma using the empirical Bayes method to calculate a t-statistic. Significance cut-offs were set at a fold change > 1.5 and nominal P value < 0.05. c) CyTOF landmark and sub-landmark gating scheme. Populations labelled in blue represent landmark populations and populations labelled in red represent sub-landmark populations. CyTOF, cytometry time of flight. PICs, post-intervention controllers. NCs, non-controllers.

Extended Data Fig. 5. CyTOF landmark and sub-landmark populations.

Frequency of major peripheral blood landmark immune cell populations over time (as a % of total live CD45+ cells) and sub-landmark populations over time (as a % of parent landmark populations). Source data

Extended Data Fig. 6. Additional T cell characterization.

Longitudinal frequency at baseline (BL), before rebound (pre-R), and/or after rebound (post-R1) of: a) Activated (CD38⁺HLA-DR⁺) non-naïve CD8⁺ T cells, b) total CD8⁺ T cells, c) non-naïve CD8⁺ T cells, d) total HIV-specific CD8⁺ T cells (as measured by IFNγ expression after 6-hour in vitro stimulation with HIV Gag, Pol, Nef, or Env peptide pools [responses summed]). e) Fold change (FC) between pre-R and post-R1 in HIV-specific CD8⁺ T cell magnitude. f) Total HIV-specific CD8⁺ T cell proliferative response (as measured by the dilution of the dye, cell trace violet [CTV], after 6-day in vitro stimulation with HIV peptide pools). g) FC between BL and pre-ATI in CD8⁺ T cell proliferation in response to Gag+Pol+Nef+Env peptide pool stimulation. h) CE-specific CD8⁺ T cell proliferative response. Magnitude (i) or proliferative response (j-k) of HIV-specific CD4⁺ T cell responses to Gag+Pol+Nef+Env peptide pools (i-j) or CE peptide pool (k). l) Frequency of Ki67⁺ cells within different CD4⁺ T cell subsets. Within-group differences were not significant in d), f), h), i), j), k) (linear mixed effects models). Source data

Extended Data Fig. 7

Longitudinal non-naïve CD8⁺ T cell cluster abundances by CyTOF. Source data