HIV broadly neutralizing antibody escape dynamics drive the outcome of AAV vectored immunotherapy in humanized mice

Abstract

Broadly neutralizing antibodies (bNAbs) have shown promise for prevention and treatment of HIV. Potency and breadth measured in vitro are often used as predictors of clinical potential; however, human studies demonstrate that clinical efficacy of bNAbs is undermined by both pre-existing and de novo resistance. Here we find that HIV-infected humanized mice receiving bNAbs delivered via AAV as Vectored ImmunoTherapy (VIT) can be used to identify antibody escape paths, which are largely conserved. Path selection, and consequent therapeutic success, is driven by the fitness cost and resistance benefit of emerging mutations. Applying this framework, we independently modulated bNAb resistance or the fitness cost of escape mutants, resulting in enhanced efficacy of VIT. This in vivo escape path analysis successfully explains the therapeutic efficacy of bNAbs, whereas potency and breadth failed to do so, illustrating a tractable means of minimizing viral escape from bNAbs.

Keywords: AAV; Broadly Neutralizing Antibody; Escapability; Evolution; HIV; Humanized Mice; Resistance; Vectored Immunotherapy.

Figure 1.. Vectored delivery of VRC07, but not PGDM1400 or N6, can suppress established HIV_REJO.c infection.

(A) BLT humanized mice were infected with 300 TCID₅₀ of HIV_REJO.c and the viral population was allowed to replicate for 4 weeks. Mice were then IM injected with 5×10¹¹ genome copies (GC) of AAVs encoding for either VRC07, N6, PGDM1400, or Luciferase as control. Mice were followed for 6 months, and blood samples were collected weekly. (B) ELISA-based quantitation of gp120-binding antibodies in the serum of HIV_REJO.c-infected humanized mice following administration of AAV8-Luciferase, AAV8-VRC07, AAV8-N6, or AAV8-PGDM1400 vectors. Black arrow denotes vector administration. Data are plotted as geometric mean ± geometric SD. (C-F) HIV viral load in plasma of HIV_REJO.c-infected mice injected with AAV8-Luciferase as control (C), AAV8-VRC07 (D), AAV8-N6 (E), or AAV8-PGDM1400 (F). Black arrows denote vector administration. Each colored line depicts an individual mouse tracked over time. The sensitivity of qPCR was 1 genome copy per μL of plasma, and 5 μL were used in the reaction, resulting in a 1000 copy per mL limit of detection (solid line). Data are presented as mean ± S.E.M. (G) Kaplan-Meier plot of viral suppression in BLT humanized mice infected with HIV_REJO.c given the indicated bNAb-expressing vector. The total model significance (p<0.0001) and pairwise comparison against the Luciferase control (****: p<0.0001) were assessed independently using Log-rank (Mantel-Cox) tests. The percentage of HIV suppressed was defined as the fraction of mice that were not escaped as described in the methods.

Figure 2.. Fitness cost and resistance benefit accrued along escape paths determine the difficulty of HIV_REJO. escape from each bNAb.

(A) Amino acid divergence from the envelope gene of the HIV_REJO.c parental strain across all AAV8-VRC07 treated mice. Sequences were determined by Illumina Deep Sequencing of the viral envelope isolated from plasma at the final experimental timepoint. The X-axis represents the envelope protein amino acid position relative to HXB2 numbering. The Y-axis represents the percentage of average amino acid divergence from the parental strain, corrected for divergence observed in control mice. Pie charts represent the most common amino acid mutations for sites with the highest divergence. See also Figure S2. (B) In vitro neutralization assays of each HIV_REJO.c mutant identified in (A) against VRC07. Data are plotted as mean ± S.E.M. Each data point was evaluated in quadruplicate. (C) Relative viral growth of each HIV_REJO.c mutant identified in (A). Growth rates were determined in activated CD4⁺ T cells performing QuickFit assays and normalized to the parental strain. Data are plotted as mean ± S.E.M. and statistical differences were assessed by a Kruskal-Wallis non-parametric ANOVA with Dunn’s post hoc test to correct for multiple comparisons (****: p<0.0001). See also Figure S2. (D-F) Escapability maps denoting fitness cost (Y-axis, relative doubling time shown in (C)) and resistance benefits (X-axis, neutralization resistance shown in (B)) for each HIV_REJO.c mutant observed during escape from VRC07 (D), N6 (E), or PGDM1400 (F). Dashed vertical line denotes the geometric mean antibody serum concentration after vectored bNAb administration. Shaded areas represent the neutralization assay limit of detection. Solid arrows represent the likely initial path taken with dashed arrows representing the likely second step paths to escape. In the map legend, each escape path is sorted based on the relative frequency observed in the sequencing data. See also Figures S2 and S3. (G) Survival analysis of HIV_REJO.c-infected mice treated with AAV8-VRC07 by escape path haplotype. The total model significance (p=0.0126) and pairwise comparisons of the non-D279A escape paths to each other escape path (*: p<0.05) were assessed independently using Log-rank (Mantel-Cox) tests. See also Figure S3. (H) Correlation plot of N6 geometric mean-expression vs post-escape HIV_REJO.c viral load. The line represents the semi-log least squares regression, ρ represents the Spearman correlation value determined for the data, with the associated p-value below. See also Figure S3. (I) Correlation plot of the post-N6 escape viral load of HIV_REJO.c vs the frequency of N276D determined by viral envelope sequencing. The line represents the semi-log least squares regression, ρ represents the Spearman correlation value determined for the data, with the associated p-value below. See also Figure S3. (J) Escape Barrier (AUC) Score denoting the aggregate fitness cost for each escape path as determined in the escapability maps. This score was calculated by adding the area under the escapability plot for escape paths A and path B viral escapes from VRC07, N6, and PGDM1400. See also Figure S4.

Figure 3.. HIV_REJO.c V5-loop restricts escape from VRC07, by decreasing the resistance benefit of D279 mutations.

(A) ELISA-based quantitation of gp120-binding antibodies in the serum of HIV_JR-CSF-infected humanized mice following administration of 5×10¹¹ genome copies (GC) of AAV8-Luciferase, AAV8-VRC07, AAV8-N6, or AAV8-PGDM1400 vectors. Black arrow denotes vector administration. Data are plotted as geometric mean ± geometric SD. (B) Kaplan-Meier plot of viral suppression in humanized mice infected with HIV_JR-CSF given the indicated bNAb-expressing vector. The total model significance was assessed using a Long-rank (Mantel-Cox) test (p=0.0714). See also Figure S5. (C) Amino acid divergence from the envelope gene of the HIV_JR-CSF parental strain across all AAV8-VRC07 treated mice. Sequences were determined by Illumina Deep Sequencing of the viral envelope isolated from plasma at the final experimental timepoint. The X-axis represents the envelope protein amino acid position relative to HXB2 numbering. The Y-axis represents the percentage of average amino acid divergence from the parental strain, corrected for divergence observed in control mice. Pie chart represents the most common amino acid mutations for the site with the highest divergence. See also Figure S6. (D) Escapability map of HIV_JR-CSF escape from VRC07. Dashed vertical line denotes the geometric mean antibody serum concentration after AAV8-VRC07 administration. Shaded area represents the neutralization assay limit of detection. Arrows represent the likely path taken to escape. See also Figures S6 and S8. (E) Alignment of the D-loop and the V5-loop amino acid sequences for HIV_REJO.c, HIV_JR-CSF, and the HIV_RD, HIV_RV, and HIV_RDV chimeras. (F) In vitro neutralization of HIV_REJO.c, HIV_RD, HIV_RV, HIV_RDV, and HIV_JR-CSF by VRC07 bNAb. Data are plotted as mean ± S.E.M. Each datapoint was evaluated in quadruplicate. (G) Kaplan-Meier plot of viral suppression in humanized mice infected with HIV_REJO.c, HIV_RV, or HIV_JR-CSF following vectored VRC07 administration. The total model significance (p<0.0001) and pairwise comparisons against HIV_REJO.c (***: p<0.001) were assessed independently using Log-rank (Mantel-Cox) tests. See also Figure S7. (H) Escapability map of HIV_RV during escape from VRC07. Lighter square symbols represent the original HIV_REJO.c escape path taken against VRC07. Dashed vertical line denotes geometric mean bNAb concentrations after vectored VRC07 administration. Shaded areas represent the limits of detection. Arrows represent the likely path to escape. See also Figures S7 and S8. (I) Escape Barrier (AUC) Score quantifying the difficulty of escape from VRC07 for HIV_REJO.c, HIV_RV, and HIV_JR-CSF. See also Figure S8.

Figure 4.. Increasing the fitness cost of HIV_REJO.c escape mutations enhances the efficacy of vectored VRC07.

(A) The fitness of HIV_REJO.c-VRC07 escape mutations with or without the Pol M184V mutation were evaluated using QuickFit. Data are presented as mean ± SD. Statistical differences were assessed by a Two-Way ANOVA, with a Šidák post hoc test to correct for multiple comparisons (*: p<0.05; ****: p<0.0001). See also Figure S9. (B) Escapability map of HIV_REJO.c-VRC07 escape mutants with or without the Pol M184V mutation. Solid symbols (diamonds) represent envelope mutations on the Pol M184V mutant background, while lighter symbols (circles) represent the data for envelope mutations on the WT Pol background. Shaded areas represent the limits of detection. Solid arrows represent the likely path taken with dashed arrows representing secondary steps to escape. See also Figure S9. (C) Empirical Escape Barrier score for HIV_REJO.c escape from VRC07 as compared to the theoretical Escape Barrier score for HIV_{REJO.c-PolM184V} escape from VRC07. (D) Experimental setup to determine the impact of Pol M184 mutations on HIV_REJO.c escape from VRC07. Humanized mice were infected with HIV_REJO.c and then treated with a sub-optimal ART starting 2 weeks after infection, to select and maintain Pol M184 mutants. At week 8, 5×10¹¹ GC of AAV8-VRC07 was injected IM, and the ART regimen was stopped at week 10. Mice were followed for 6 months, and blood samples were collected weekly. (E) Percentage of Pol M184V/I mutations determined from viral sequences isolated from the plasma of HIV_REJO.c-infected mice at week 8, prior to AAV8-VRC07 administration. Statistical differences were assessed by an unpaired two-tailed Student’s t-test (****: p<0.0001). Data are presented as mean ± S.E.M. (F) HIV_REJO.c viral load in plasma of AAV8-VRC07 treated mice. Black arrows denote vector administration. Each colored line depicts an individual mouse. The qPCR lower limit of detection was 1 genome copy per μL of plasma and 5μL were used in the reaction (solid line). Data are presented as mean ± S.E.M. (G) Kaplan-Meier plot of viral suppression in humanized mice infected with HIV_REJO.c with or without ART selection and with or without AAV8-VRC07 administration. The total model significance (p<0.0001) and pairwise comparisons against the Luciferase + Selection control (***: p<0.001; ****: p<0.0001) were assessed independently with Log-rank (Mantel-Cox) tests. See also Figure S9.