Dysregulated autoantibodies targeting vaso- and immunoregulatory receptors in Post COVID Syndrome correlate with symptom severity

Figure 1. Study workflow and description of autoantibody targets. (A) After data acquisition, different statistical analyses (written on the top) were carried out in order to characterize the signature of autoantibodies (AAB) against G protein coupled receptors (GPCRs) and COVID-19-associated molecules (e.g. renin-angiotensin system (RAS)) in Post COVID Syndrome (PCS) when compared with healthy controls (HC) and post COVID-19 healthy controls (PCHC). Created with Biorender. (B) The 10 squares on the left represent autonomic nervous system (ANS) related receptors, while the 10 on the right show non-ANS molecules and receptors (e.g. RAS, immune and circulatory systems). Blue edges in the network highlight the interactions among the AAB targets, while gray edges represent other interactions. Node colors map to Gene Ontology (GO) biological processes (BPs) and node size corresponds to number of interacting partners for each target. Circular nodes represent human and SARS-CoV-2 molecules (as well as two Spike (S) proteins with unspecified roles) that are described in the IMEx coronavirus interactome. Circular organization of the proteins on the top middle of the image represent interacting…

Figure 2. Autoantibodies (-Ab) against G protein coupled receptors (GPCR) and COVID-19-associated molecules are dysregulated during Post COVID Syndrome (PCS). (A) Box plots of Ab investigated in PCS patients with and without ME/CFS and healthy controls post or without COVID-19 history (PCHC or HC). Significance determined by Kruskal Wallis test followed by Dunn test as post hoc. Dunn test p values were corrected for FDR. Adjusted p-values are being represented by: *p.adj < 0.05; **p.adj <0.01; ***p.adj < 0.001; ****p.adj < 0.0001. Boxes represent the median and interquartile range (IQR). (B) Forest plot of regression coefficients for the confounding factors age in years, gender (reference being female) and time post COVID-19 in months considering 95% confidence interval (CI). Red dots and CI indicate that variable has a positive influence in the Ab level, blue dots and CI indicate a negative influence and gray ones contain 0 in the confidence interval, therefore are taken as non significant.

Figure 3. Autoantibodies (-Ab) stratify patients by post-acute COVID-19 outcomes. (A) Principal component analysis (PCA) with spectral decomposition based on logarithmic values of 20 Abs show the stratification of the four studied groups. Variables pointing to the same sense of the corresponding principal components are positive correlated. Small ellipses are the concentration around the mean points of each group. (B) Graphs of variables (Abs) obtained by PCA of all individuals in this study. (C) Barplot with the contribution percentages of each variable to each dimension. A black dashed line is plotted on the 5% mark, and blue bars indicate a contribution higher than 5%.

Figure 4. Machine learning classification of study groups based on autoantibodies. (A) Receiver operating characteristic (ROC) curves of 20 antibodies (Abs) with an area under the curve (AUC) of 77% for healthy individuals and 77% for PCS patient group. (B) Stable curve showing number of trees and out-of-bag (OOB) error rate of 20.34%. (C) Variable importance score plot based on Gini decrease and number (no) of nodes, and the mean of minimum depth for each Ab, showing which variable presents a higher score in classifying COVID-19 post-acute infection outcomes. (D) Heatmap of the confusion matrix. Numbers represent the amount of occurrences that happened when training the random forest model in predicted (row) vs actual classification (column), therefore the blueish diagonal identifies the hits, while other cells are mismatches.

Figure 5. Autoantibody correlation signatures associate with post-acute infection outcome. Circular networks based on Spearman’s rank correlation for the level of the 20 autoantibodies (-Ab) in post COVID syndrome (PCS) patients with and without ME/CFS and healthy controls post or without COVID-19 history (PCHC or HC). There is a list with the abbreviations and the Abs names by the right side of the plot. Correlations greater than 0.6 are represented by the blue edges, and thicker edges imply greater correlations.

Figure 6. Correlation between autoantibody (-Ab) levels and clinical scores. Plots represent Spearman correlation coefficient (r) of correlation of Abs with (A) symptom scores and (B) autonomic symptom score assessed by COMPASS-31 questionnaire of PCS/non-ME/CFS (grey) and PCS/ME/CFS (black) patients. p values represented by: *p < 0.05, **p< 0.01 and ***p<0.001.