Background: Severe coronavirus disease 2019 (COVID-19) disease courses are characterized by dysregulated host response, which leads to a broad range of immuno-inflammatory, thrombotic, and parenchymal alterations. 

Purpose: The aim was to characterize a system-wide, molecular serologic fingerprint of severe COVID-19 longitudinal disease courses.

Methods: We performed longitudinal profiling of plasma including metabolome, proteome (Olink^® Cardiovascular and Inflammation panel à 96 proteins) and routine biochemistry from 5 of initially 7 (2 drop outs after data pre-processing) seropositive, well-phenotyped patients with severe COVID-19 referred to the Intensive Care Unit at German Heart Center in Munich during the first phase of the pandemic in 2020. Patient characteristics were; 64±9 years, 40% female, median CRP 300±142 mg/dl, IL-6 279±187 ng/L, D-Dimers 8,5±12 mg/L, NT-proBNP 2425±3934 ng/L, all on mechanical ventilator, 40% in need for mechanical circulatory and pulmonary support via extracorporal membrane oxygenation (ECMO). Our statistical evaluation included a machine learning approach (employing feature selection through LASSO). We compared the non-negative coefficients with a random simulated model incorporating the correlation structure of the underlying dataset revealing potential blood based candidate biomarkers preceding hyperinflammatory immune response (denoted rise in IL-6) and COVID-19 coagulopathy (denoted rise in D-Dimers) by 24-48 h.

Results: Time-series analysis of patient sera revealed a number of candidate proteins involved in biological pathways such as coagulation/thrombosis (e.g. Urokinase plasminogen activator, Carboxypeptidase B2), oxidative stress/inflammation (e.g. IL-1alpha, MMP9, C-C motif chemokine 20), immunoadhesion (e.g. E-selectin), tissue repair (e.g. Epithelial cell adhesion molecule), energy metabolism and growth factor response, as well as regulatory pathways [e.g. Tyrosine-protein kinase receptor UFO which has been described as a host-/co-receptor that promotes the entry of SARS-CoV-2 into cells in the absence of ACE2 and low-density lipoprotein receptor (LDLR)]. Upregulation of LDLR might underly the typical changes in lipoprotein profile observed with severe COVID-19 disease courses. Comparing multivariate with univariate results showed that combination of markers rather than single markers predict the selected markers associated with disease progression of COVID-19.

Conclusions: A better understanding of the proteomic signature preceding clinical deterioration in severe COVID-19 disease courses might shed further light on the mechanisms associated with hyperinflammation/COVID-19 coagulopathy, help guide treatment allocation and help pave the way to development of druggable targets/prognostic tests.