Cell-free RNA reveals host and microbial correlates of broadly neutralizing antibody development against HIV

Figures

• 

  Fig 1. Study design and development of neutralization breadth. A) Longitudinal trajectories of neutralization breadth in bNAb producers (red) and controls (blue). Time zero corresponds to the estimated HIV infection point, with breadth defined as the ability of plasma antibodies to neutralize a cross-clade panel of 18 env-pseudotyped viruses (S2 Table). B) Heatmap of serum neutralization titers (IC50) against the 18-virus panel, grouped by individual, with time points sorted from first (top) to last (bottom). Colors represent log10-transformed IC50 neutralization titers, with higher values (yellow) indicating stronger neutralization potency and black indicating values below the limit of detection.

• 

  Fig 2. HIV consensus sequences from cfRNA. A) Genomic similarity matrix of consensus HIV sequences obtained from cfRNA, showing clustering within individuals. B) Maximum-likelihood phylogeny of cfRNA-derived HIV consensus sequences (colored labels) with previously published envelope sequences (bold, CAPRISA). Shaded boxes highlight individual-level clustering. With the exception of one sample, cfRNA consensus sequences consistently cluster with the corresponding envelope sequences.

• 

  Fig 3. Correlation between cfHIV levels and host transcriptome. A) Number of genes significantly correlated with cfHIV counts, viral load (VL) and CD4 counts (Spearman’s rank correlation, see Methods). B) Subset of genes from panel A previously described as involved in host immune responses to viral infection. C) Example correlations between cfHIV counts and three HIV-associated genes. Spearman correlation coefficients indicated. No partitioning is observed between bNAb producers (red) and controls (blue).

• 

  Fig 4. Gene level differences in cfRNA between bNAb producers and controls. A) Volcano plot of differentially abundant genes. Genes with a fold change > 2 and FDR < 0.05 are highlighted (red = increased, blue = decreased in bNAb producers). Labelled points are genes involved in host immune responses to viral infection. B) Violin plots of the distribution of elevated genes in bNAb producers, stratified by group and neutralization breadth (number of viral strains neutralized), used here as a proxy for progression along bNAb development. C) Genes in the MHC class I antigen presentation pathway. Most show negative correlations with breadth, and some are also differentially abundant between bNAb producers and controls. Correlations with breadth were assessed using Spearman’s rank correlation (Spearman’s ρ indicated next to each gene, see Methods). * correlated with breadth at FDR < 0.05, † differentially abundant between bNAb producers and controls at FDR < 0.05, ‡ correlated with cfHIV at FDR < 0.05.

• 

  Fig 5. Microbial differences and GBV-C enrichment in bNAb producers. A) Volcano plot of differentially abundant microbial genera between bNAb producers and non-producers, colored by phylum (FDR < 0.05, Benjamini-Hochberg). Positive log2 fold change indicates enrichment in bNAb producers. B) Scatterplots showing cfGBV-C abundance in samples partitioned by time and study group (left, bNAbs producers, right, controls). The dashed line indicates the threshold used to define elevated cfGBV-C levels. Using a logistic mixed-effects model with participant as a random effect, bNAb producers showed higher odds of elevated GBV-C levels (odds ratio [OR]: 6.4, 95% CI: 2.3-17.9). In some participants, GBV-C read counts exceeded those of HIV up to 100-fold.