Unveiling the proteome-wide autoreactome enables enhanced evaluation of emerging CAR T cell therapies in autoimmunity

Abstract

Given the global surge in autoimmune diseases, it is critical to evaluate emerging therapeutic interventions. Despite numerous new targeted immunomodulatory therapies, comprehensive approaches to apply and evaluate the effects of these treatments longitudinally are lacking. Here, we leveraged advances in programmable-phage immunoprecipitation methodology to explore the modulation, or lack thereof, of autoantibody profiles, proteome-wide, in both health and disease. Using a custom set of over 730,000 human-derived peptides, we demonstrated that each individual, regardless of disease state, possesses a distinct and complex constellation of autoreactive antibodies. For each individual, the set of resulting autoreactivites constituted a unique immunological fingerprint, or "autoreactome," that was remarkably stable over years. Using the autoreactome as a primary output, we evaluated the relative effectiveness of various immunomodulatory therapies in altering autoantibody repertoires. We found that therapies targeting B cell maturation antigen (BCMA) profoundly altered an individual's autoreactome, while anti-CD19 and anti-CD20 therapies had minimal effects. These data both confirm that the autoreactome comprises autoantibodies secreted by plasma cells and strongly suggest that BCMA or other plasma cell-targeting therapies may be highly effective in treating currently refractory autoantibody-mediated diseases.

Keywords: Adaptive immunity; Autoimmune diseases; Autoimmunity; Immunotherapy; Therapeutics.

Figure 1. Individuals harbor a unique set of longitudinally stable autoreactivities.

(A) Graphical representation of the analytical approach to quantitatively compare individual samples’ unique proteome-wide autoantibody signal (“Autoreactome,” containing up to 730,000 unique autoreactivities) to any other sample. (B) Correlation matrices showing Pearson’s correlation coefficients of complete PhIP-Seq signal in healthy individuals. Left: 48 samples representing 24 individuals in technical replicate (same data set as used in A). Right: 79 distinct individuals. (C) Correlation matrix showing Pearson’s r values for complete PhIP-Seq enrichment in 7 distinct individuals, each of whom have serial samples over at least 3 years. (D) Kernel density estimate plot showing distribution of Pearson’s r correlation coefficients among technical replicates, individuals over time, and between different individuals.

Figure 2. Rituximab treatment does not significantly alter the autoreactome.

(A) Correlation matrix showing Pearson’s r values of complete PhIP-Seq signal in 7 distinct individuals with myasthenia gravis, each of whom were either rituximab naive or had not received rituximab for more than 6 months prior to first sample collection. Red arrows represent administration of rituximab. (B) Kernel density estimate plot showing distribution of Pearson’s r value correlation coefficients among longitudinal samples from individuals receiving rituximab relative to longitudinal samples from individuals received no intervention. (C) Line plots showing the autoreactivity (sum of top 10 PhIP-Seq Z scores relative to the 79 individuals acting as healthy controls) for each patient over the first 6 months following the initial rituximab dose. One-way paired-sample Wilcoxon’s test, P = 0.625 at 1 month, 0.125 at 3 months, and 0.3125 at 6 months.

Figure 3. CD19 CAR T cell therapy has minimal effect on the autoreactome 6 months after treatment.

(A) Correlation matrix showing Pearson’s r values of complete PhIP-Seq signal in 14 distinct individuals before and after anti-CD19 CAR T cell therapy. Yellow arrows represent the 6-month posttreatment time point. (B) Kernel density estimate plot showing distribution of correlation coefficients within each individual before and after therapy relative to the distribution among untreated individuals over time and between different individuals. (C) Line plots showing the autoreactivity (sum of top 10 PhIP-Seq Z scores relative to the 79 healthy controls) for each patient before and after treatment. One-sided paired-sample Wilcoxon’s test, P = 0.021.

Figure 4. Anti-BCMA CAR T cell therapy significantly alters the autoreactome.

(A) Correlation matrix showing Pearson’s r values of complete PhIP-Seq signal in 9 distinct individuals before and after anti-BCMA CAR T cell therapy. Yellow arrows represent the 6-month posttreatment time point. (B) Kernel density estimate plot showing distribution of correlation coefficients within each individual before and after BCMA-targeted therapy relative to the distribution among untreated individuals over time and the distribution between different individuals. (C) Heatmap showing top 10 autoreactivities (sum of top 10 PhIP-Seq Z scores relative to the 79 individuals acting as healthy controls) in each individual before and after treatment. (D) Line plots showing the autoreactivity (sum of top 3–12 PhIP-Seq Z scores relative to the 79 healthy controls, therefore accounting for potential paraprotein confounding) for each patient before and after treatment. One-sided paired-sample Wilcoxon’s test, P = 0.0019. (E) Box plots showing the relative distributions of autoreactivity before and after treatment with rituximab (sum of top 10), anti-CD19 CAR T cell therapy (sum of top 10), and anti-BCMA CAR T cell therapy (sum of top 3 through 12 autoreactivities). Anti-BCMA CAR T cell treatment cohort, Mann-Whitney U, P = 0.42, with a median percentage decrease of 52.3%; anti-CD19 CAR T cell treatment cohort, Mann-Whitney U, P = 0.206, with a median percentage decrease of 11.9%; and anti-BCMA CAR T cell treatment cohort, Mann-Whitney U, P = 0.003, with a median percentage decrease of 95.6%. ***P < 0.005.