Longitudinal clinical proteomics reveals pneumonia type-specific protein biomarkers and autoantibodies

Abstract

Community-acquired pneumonia is a major cause of morbidity and mortality globally. Specific molecular endotypes are currently not well defined and different viral or bacterial pathogens may trigger specific host responses and pathogenic mechanisms. We performed longitudinal proteomic profiling of bronchoalveolar lavage fluid and plasma from bacterial, influenza and SARS-COV-2 driven pneumonia. Our analysis revealed highly pneumonia type specific proteomic signatures, including COVID-19 specific antibodies locally produced in the lung. These antibodies showed biased immunoglobulin V-domain usage, linked to a CD69/CD83 plasma cell state associated with disease severity and degree of autoimmunity. Using mass spectrometry driven autoantibody profiling in two independent COVID-19 cohorts, we identified 177 putative autoantibodies targeting extracellular matrix, nuclear, and immune-related proteins. Of note, temporal changes in autoantibody profiles correlated with clinical markers of inflammation, organ dysfunction, and duration of hospitalization. These findings highlight the autoimmune aspects of COVID-19 and provide potential biomarkers and therapeutic targets to help improve patient outcomes.

Figure 1.. Pneumonia-specific signatures can be detected in BALF upon intubation in the intensive care unit.

(a) Overview of study design. (b) Lollipop chart displaying patient numbers and the longitudinal resolution of sample collection per pneumonia type. (c) Box plots that compare clinical parameters and phenotypical observations across pneumonia types. Data are represented as mean ± SD, and variance was statistically assessed with the non-parametric Kruskal-Wallis test. (d) Principal component analysis of BALF proteomes at the time of intubation based on the 488 most variable proteins. (e) Hierarchical clustering of 12 BALF pneumonia-specific protein modules at day 0 of intubation as derived from co-expression network analysis. The color scale denotes the mean fold change of the proteins in each module. (f) Pathway analysis of selected BALF module proteins at intubation induction. The top 5 enriched Reactome terms for each module are displayed. Color code depicts the adjusted p-value, and point size refers to the number of proteins detected per term. (g-h) Box plots displaying the longitudinal enrichment of (g) neutrophil degranulation and (h) immunoglobulin Reactome terms in BALF specimens across pneumonia types. Each dot represents the enrichment score for an individual patient. Data are represented as mean ± SD, and the variability between groups was statistically assessed with the non-parametric Kruskal-Wallis test. Proteins involved in the term are displayed on the right side of the plot.

Figure 2.. Integrative analysis of body fluid proteomes reveals lung-resident biomarkers.

(a) Principal component analysis of plasma proteomes at the time of intubation induction based on the 140 most variable proteins. (b) Hierarchical clustering of 7 plasma pneumonia-specific protein modules at day 0 of intubation as derived from co-expression network analysis. The color scale denotes the mean fold change of the proteins in each module. (c) Pathway analysis of selected plasma module proteins at intubation induction. The top 5 enriched Reactome terms for each module are displayed. Color code depicts the adjusted p-value, and point size refers to the number of proteins detected per term. (d) Scatter plot displaying immunoglobulin segment abundance in the BALF (x-axis) and plasma (y-axis) of COVID-19 patients at the time point upon intubation. Dot size corresponds to the number of patients that express a detected protein. (e-f) Upset plot displaying the overlap of (e) BALF and (f) shared BALF/plasma immunoglobulin segments across pneumonia types. (g-h) Box plots displaying the longitudinal enrichment of COVID-19-specific (g) BALF and (f) shared BALF/plasma immunoglobulin signatures across pneumonia types. Each dot represents the enrichment score for an individual patient. Data are represented as mean ± SD and were statistically assessed with the non-parametric Kruskal-Wallis test.

Figure 3.. Peripheral B cell clonal dynamics and transcriptional signatures are associated with COVID-19 severity.

(a) Overview of bioinformatics workflow. (b) UMAP visualization of 38,063 peripheral blood B/plasma cells from 90 COVID-19 patients and 23 healthy controls, colored by cluster identity. (c) UMAP projection of 8,599 plasma cells from 90 COVID-19 patients and 23 healthy controls. Lower panels show clonal expansions of BALF-derived V-fragment lineages mapped to COVID-19 patients vs controls. (d) Heatmap visualizing the relative abundance of B/plasma cell molecular states across stages of COVID-19 disease progression. (e) Bar plot showing clonal percentage of proteomics-identified V-segments per plasma cell state. (f) Heatmap of top differentially expressed genes for each plasma cell state. Color code denotes the scaled average expression across clusters. (g) Heatmap of top differentially expressed proteins for plasma cell state 3. Color code denotes the scaled average expression across clusters. (h) Heatmap of differentially expressed surface protein markers in plasma cell state 3 between healthy controls and COVID-19 donors.

Figure 4.. Comparative analysis of the putative auto-antigen repertoire of two independent COVID-19 cohorts.

(a) Flowchart of the Differential Antibody Capture assay (DAC) and benchmarking processes with anti-IL-6R antibody. (b) Schema of Chicago and Munich cohort demographics. (c) Scatter plot displaying fold changes of individual antigens in COVID-19 patients (n=13) of the Chicago cohort versus bacterial pneumonia patients (n=6) on the x-axis and influenza patients (n=7) on the y-axis at the first time point upon intubation. Dot size and color represent the number of COVID-19 patients with a particular plasma autoantibody. Enriched putative autoantibodies in COVID-19 over bacterial pneumonia and influenza are depicted in the highlighted rectangle. (d) Scatter plot displaying fold changes of individual antigens in severe COVID-19 patients (n=16) of the Munich cohort versus high-inflammatory control patients (n=23) on the x-axis and low-inflammatory control patients (n=24) on the y-axis at the first point upon in hospitalization. Dot size and color and color represent the number of COVID-19 patients with a particular plasma autoantibody. Enriched putative autoantibodies in COVID-19 over both controls are depicted in the highlighted rectangle. (e) Venn diagram demonstrating putative autoantibody repertoires of the Chicago and Munich cohorts grouped by molecular function. Displayed autoantigens are significantly enriched at least at one time point over at least one control group and in at least 3 COVID-19 patients.

Figure 5.. Longitudinal profiling of autoreactivities in mild and severe COVID-19 patients.

(a) T-distributed stochastic neighbor embedding (t-SNE) of 27 COVID-19 patients enrolled in the Ludwig-Maximilians-University Hospital (Munich cohort) based on available clinical parameters at the beginning of intubation. Two clusters (mild and severe) are shown. (b) Heatmap of mixed-effects model for time and COVID-19 severity. Only significant proteins (p<0.05) are displayed, and shared proteins with the Chicago cohort are highlighted. Color code represents average expression in time intervals after hospitalization.

Figure 6.. Blood putative autoantibodies are associated with clinical parameters in two independent severe COVID-19 cohorts.

(a-b) Heatmaps presenting associations between the detection antibodies against shared putative antigens (n=15) in the peripheral blood of (a) n=13 COVID-19 patients of the Chicago and (b) n=16 severe COVID-19 patients of the Munich cohort upon intubation on the x-axis and selected clinical parameters on the y-axis in the Chicago and Munich cohorts. The color denotes the output of the Wilcoxon test (p-value). The dark grey color corresponds to the absence of the test results due to missing values. (c-d) Box plots showing the significant associations between 19 putative shared autoantigens and clinical parameters in two cohorts: Chicago (c) and Munich (d). The x-axis represents COVID-19 patients categorized based on the presence (+) or absence (−) of detected autoantigens at the time of intubation. Statistical significance was assessed using the Wilcoxon test.