HIV-1 antibody 3BNC117 suppresses viral rebound in humans during treatment interruption

Abstract

Interruption of combination antiretroviral therapy in HIV-1-infected individuals leads to rapid viral rebound. Here we report the results of a phase IIa open label clinical trial evaluating 3BNC117,a broad and potent neutralizing antibody against the CD4 binding site of the HIV-1 Env protein, during analytical treatment interruption in 13 HIV-1-infected individuals. Participants with 3BNC117-sensitive virus outgrowth cultures were enrolled. Results show that two or four 30 mg kg(-1) 3BNC117 infusions,separated by 3 or 2 weeks, respectively, are generally well tolerated.Infusions are associated with a delay in viral rebound of 5-9 weeks after two infusions, and up to 19 weeks after four infusions, or an average of 6.7 and 9.9 weeks, respectively, compared with 2.6 weeks for historical controls (P < 0.00001). Rebound viruses arise predominantly from a single provirus. In most individuals,emerging viruses show increased resistance, indicating escape.However, 30% of participants remained suppressed until antibody concentrations waned below 20 μg ml(-1), and the viruses emerging in all but one of these individuals showed no apparent resistance to 3BCN117, suggesting failure to escape over a period of 9-19 weeks.We conclude that the administration of 3BNC117 exerts strong selective pressure on HIV-1 emerging from latent reservoirs during analytical treatment interruption in humans.

Extended Data Figure 1. Study participant selection and neutralization of pre-infusion cultures by 3BNC117

a, Flow diagram showing the selection of study participants. b, Bar diagrams showing IC₅₀ values (μg ml⁻¹) in TZM-bl assays for 3BNC117 against bulk virus outgrowth culture supernatants from the indicated time point pre-infusion for each participant (Supplementary Table 3). For some participants both screen and day 0 cultures were obtained and showed less than threefold variation in IC₅₀ values. The red dotted line indicates an IC₅₀ of 2 μg ml⁻¹ which was used as a threshold for inclusion in the study.

Extended Data Figure 2. CD4⁺ and CD8⁺ T cells during study period in participants

a–d, Absolute T cell counts (a, c) and percentage of CD4⁺ and CD8⁺ T cells among CD3⁺T cells (b, d) for group A and B, respectively (Supplementary Table 4). 3BNC117 infusions are indicated with red arrows. e, Comparison of absolute CD4⁺ T cell counts and percentage of CD4⁺ T cells among CD3⁺T cells at screen, day 0, rebound and after re-suppression. Shown is the data from participants 701, 702, 703, 704, 707, 708, 709, 711 and 713 for whom re-suppression CD4 counts were available (Supplementary Table 4). The last available time point was used as re-suppression time point. Red lines indicate the mean value and error bars indicate standard deviation. P values were obtained using a paired t-test comparing the indicated time points. f, Plasma viral loads and CD4 counts in all study participants. 3BNC117 infusions are indicated with red arrows. The left y-axis shows plasma viral loads in RNA copies per ml (black curves), and right y-axis shows absolute CD4 counts in cells per μl (red curves). Grey areas indicate ART therapy.

Extended Data Figure 3. HIV-specific T-cell responses

Total breadth (open squares) and magnitude (bars) of T-cell responses against HIV-1 overlapping peptides (OLPs) at the designated time points following administration of 3BNC117 (yellow arrows indicate infusions of 3BNC117 at 30 mg kg⁻¹). For all study participants, antiretroviral therapy was discontinued on day 2 after the first 3BNC117 administration. Blue arrows indicate the time of viral rebound. For study participants 710, 712 and 715 rebound occurred at week 19, 16 and 11, respectively. Baseline samples for study participant 710 and week 12 samples for study participant 714 were not available for ELISpot analysis. Overall, breadth, magnitude and protein specificity were heterogeneous among the study participants.

Extended Data Figure 4. Viral rebound in ACTG control subjects and trial participants

a, Kaplan–Meier plot summarizing viral rebound in 52 ACTG trial participants who underwent ATI without antibody treatment (black, Supplementary Table 6) and trial participants (Fig. 2a, b, Supplementary Table 4). Six group A participants are shown in red, seven Group B participants in blue and the combination in green as indicated. The y-axis indicates the percentage of participants with viral levels below 200 RNA copies per ml, x-axis indicates weeks after ATI initiation. The survival curves of all considered partitions of the trial participants (group A, group B and group A + B) differed significantly at significance level α = 0.05 from the survival curve of the ACTG trial participants. For the comparison of group A (group A + B) with the ACTG trial participants, we performed a weighted log-rank test adjusting for the clinical variables ‘years on ART’ and ‘age’ to correct for possible confounding factors (Supplementary Table 7 P <0.00001). We identified those potential confounders by univariate parametric survival regression using a likelihood ratio test (Statistical Methods). Since we did not discover any confounders with the same analysis among all available clinical variables for the comparison between group B participants and the ACTG trial participants, we performed a standard log-rank test in that setting (P < 0.0001). b–d, In order to perform a survival regression, the distribution of the rebound times has to be determined. Therefore, we compared the empirical cumulative distribution function (CDF) of the rebound times (black, solid line) with the CDF of the rebound times to a fitted distribution (Weibull, exponential, normal, logistic, log-normal, and gamma) for each comparison group (combined trial participants, group A or group B with ACTG control patients). Since the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) were smallest for the log-normal distribution (green), we have chosen to model the rebound times with the log-normal distribution. e, Dot plot indicating the relationship between cell associated HIV DNA in pre-infusion PBMCs (y-axis) and the week of rebound (x-axis). Group A and group B participants are coloured red and blue respectively. The P value was derived from calculating the Pearson correlation coefficient.

Extended Data Figure 5. In vitro neutralization of pre-infusion and rebound virus outgrowth cultures by 3BNC117 or 10-1074

a, b, TZM-bl assay neutralization by 3BNC117 (a) and 10-1074 (b) are shown for individual virus outgrowth cultures derived from pre-infusion (black lines/symbols) or rebound (red lines/symbols) time points for each participant. In some cases, multiple independent cultures were grown from a single time point and assayed for neutralization (Supplementary Table 3). ‘Screen’ refers to cultures of PBMC samples taken weeks before infusion during screening, while ‘Day 0’ refers to cultures of PBMCs collected immediately before the first 3BNC117 infusion. Rebound culture time points are denoted by the week (W) at which the samples were collected. Symbols reflect the means of two technical replicates; error bars denote standard deviation.

Extended Data Figure 6. Phylogenetic tree of env nucleotide sequences from trial participants

A maximum likelihood phylogenetic tree was constructed from single-genome-derived viral env sequences from outgrowth culture supernatants as well as plasma from participants 701 (olive), 702 (black), 703 (pink), 704 (yellow), 707 (light blue), 708 (green), 709 (dark blue) and 711 (brown). Hypervariable (as defined in http://www.hiv.lanl.gov/content/sequence/VAR_REG_CHAR/) and other poorly aligned regions were excluded from the analysis. The tree was constructed using PhyML with a GTR+I+G substitution model and midpoint rooted. Asterisks indicate 100% bootstrap support (only values for major nodes are shown). The scale bar indicates 0.01 substitutions per site.

Extended Data Figure 7. Rebound virus clonality and neutralization sensitivity to 3BNC117

Maximum likelihood phylogenetic trees of plasma and culture-derived env sequences are shown for participants 701, 702, 703, 704. Sequences obtained at screening, on Day 0, and consecutive rebound time points (plasma and cultures) are colour coded as indicated. The trees were rooted based on the branch insertion identified in the between-subject tree (Extended Data Fig. 6). Bootstrap values ≥90% are shown. Names of env sequences used to generate pseudoviruses for 3BNC117 neutralization analysis are indicated along with the respective IC₈₀ titres in μg ml⁻¹. Representative rebound viruses selected in Fig. 4b are marked with red stars (Fig. 4b, Supplementary Table 9). Zero branch length viruses in multi-rebounders 702 and 703 are marked with black stars.

Extended Data Figure 8. Rebound virus clonality and neutralization sensitivity to 3BNC117

Maximum likelihood phylogenetic trees of plasma- and culture-derived env sequences are shown for participants 707, 708, 709 and 711. Sequences obtained at screening, on day 0, and consecutive rebound time points (plasma and cultures) are colour coded as indicated. The trees were rooted based on the branch insertion identified in the between-subject tree (Extended Data Fig. 6). Bootstrap values ≥90% are shown. Names of env sequences used to generate pseudoviruses for 3BNC117 neutralization analysis are indicated along with the respective IC₈₀titres in μg ml⁻¹. Representative rebound viruses selected in Fig. 4b are marked with red stars (Fig. 4b, Supplementary Table 9). Zero branch length viruses in multi-rebounder 709 are marked with black stars.

Figure 1. 3BNC117 neutralization coverage, trial design and pharmacokinetics of 3BNC117 in HIV-1-infected individuals during ATI

a, b, Sensitivity of virus outgrowth cultures from 63 ART suppressed individuals to 3BNC117 and VRC01 (Supplementary Table 1). The y-axis shows the fraction of viral outgrowth culture supernatants neutralized by a given antibody concentration (x-axis) in Tzm-bl assays. Red line indicates cut-off IC₅₀ (2 μg ml⁻¹) for participation in the trial. c, d, Diagrammatic representation of study groups A and B respectively. 3BNC117 infusions indicated by the red arrows, and sampling for PK and virologic studies indicated below. Numbers indicate study weeks. e, f, 3BNC117 levels as determined by ELISA for group A (n = 6, left panel, red), group B (n = 7, right panel, red (n = 4), black (n = 2) and purple (n = 1)), HIV-1 negative (n = 3, blue) and viraemic individuals (n = 6, green). Curves indicate mean 3BNC117 levels, error bars the standard deviation. Arrows indicate 3BNC117 infusions.

Figure 2. Delay in viral rebound in the presence of 3BNC117

a, b, Plasma viral loads and 3BNC117 levels in group A and group B participants respectively. 3BNC117 infusions are indicated with red arrows. The left y-axis shows plasma viral loads in RNA copies per ml (black curves), and right y-axis shows antibody levels measured by ELISA (red curves). Average rebound time point (2.6 weeks, Supplementary Table 6) in 52 ACTG trial participants who underwent ATI without antibody treatment is shown with dotted lines. Grey areas indicate ART therapy. c, Kaplan–Meier plot summarizing viral rebound in 52 ACTG trial participants who underwent ATI without antibody treatment (black, Supplementary Table 6), and the combination of all 13 participants (red) who underwent ATI with 3BNC117 infusions. The y-axis indicates the percentage of participants with viral levels below 200 RNA copies per ml, x-axis indicates weeks after ATI initiation. The P value is based on a bootstrap version of the weighted log-rank test adjusting for the potential confounders ‘years on ART’, and ‘age’ (Supplementary Table 7, Methods Statistical Analyses). d, Dot plot indicating the relationship between 3BNC117 sensitivity of pre-infusion outgrowth cultures at screening (y-axis, IC₈₀ in μg ml⁻¹) and the week of rebound (x-axis). Group A (n = 6) and group B (n = 7) participants are coloured red and blue respectively. The P value was derived from calculating the Pearson correlation coefficients. e, Dot plot indicating the relationship between 3BNC117 sensitivity of rebound outgrowth cultures (y-axis, IC₈₀ in μg ml⁻¹) and the 3BNC117 serum concentration at rebound (x-axis, in μg ml⁻¹). 704, 708, 709 and 713 did not reach IC₈₀ at the concentrations tested and were assigned a value of 22 μg ml⁻¹. Group A (n = 6) and group B (n = 7) participants are coloured red and blue, respectively. The P value was derived from calculating the Pearson correlation coefficients.

Figure 3. Viral rebound during ATI and 3BNC117 treatment

a, b, Graph of 3BNC117 (a) or 10-1074 (b) IC₈₀ titres against baseline and rebound outgrowth cultures. Blue and green bars represent average IC₈₀ titres against screen and day 0 outgrowth cultures; red and purple bars represent IC₈₀ titres against rebound outgrowth cultures from the indicated time points. Asterisks indicate cultures failing to reach an IC₈₀ up to 20 μg ml⁻¹. c, d, Difference between rebound and pre-infusion culture IC₈₀ titres (from a, b) for 3BNC117 (c) or 10-1074 (d). For cultures failing to reach an IC₈₀ up to 20 μg ml⁻¹, a value of 20 μg ml⁻¹ was assigned, and such cultures are marked with an asterisk. e, f, Clonality of the rebound virus. Maximum likelihood phylogenetic trees comparing pre-ATI single genome derived env sequences (blue and orange) to rebound plasma env sequences (green and pink) are shown for participants whose rebound comprised single (e) versus multiple (f) viruses (Supplementary Table 8). Pre-ATI culture sequences were inferred as described in the Methods section ‘Statistical analyses’.

Figure 4. 3BNC117 resistance in rebound viruses

a, Logogram shows env gp120 regions (amino acid positions; 270–285 and 455–467, according to HXBc2 numbering) indicating sequence changes from pre-infusion culture(s) (first row) to rebound sequences derived from plasma SGS at the indicated time points. The frequency of each amino acid is indicated by its height. Red residues represent mutations predicted to affect neutralization. b, 3BNC117 neutralization sensitivity of pseudoviruses derived from pre-infusion or rebound SGS. Black lines represent pre-infusion virus envs; red lines represent the major env at rebound for each participant (Extended Data Figs 7 and 8, Supplementary Tables 8 and 9 and Methods); grey lines represent minor rebound envs in participants with multiple rebound viruses or variants that evolved after rebound (Extended Data Figs 7 and 8, Supplementary Table 9). Symbols reflect the means of two technical replicates; error bars denote standard deviation.