Mapping the Serum Proteome of COVID-19 Patients; Guidance for Severity Assessment

Abstract

Coronavirus disease 2019 (COVID-19) is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), whose outbreak in 2019 led to an ongoing pandemic with devastating consequences for the global economy and human health. According to the World Health Organization, COVID-19 has affected more than 481 million people worldwide, with 6 million confirmed deaths. The joint efforts of the scientific community have undoubtedly increased the pace of production of COVID-19 vaccines, but there is still so much uncharted ground to cover regarding the mechanisms of SARS-CoV-2 infection, replication and host response. These issues can be approached by proteomics with unprecedented capacity paving the way for the development of more efficient strategies for patient care. In this study, we present a deep proteome analysis that has been performed on a cohort of 72 COVID-19 patients aiming to identify serum proteins assessing the dynamics of the disease at different age ranges. A panel of 53 proteins that participate in several functions such as acute-phase response and inflammation, blood coagulation, cell adhesion, complement cascade, endocytosis, immune response, oxidative stress and tissue injury, have been correlated with patient severity, suggesting a molecular basis for their clinical stratification. Eighteen protein candidates were further validated by targeted proteomics in an independent cohort of 84 patients including a group of individuals that had satisfactorily resolved SARS-CoV-2 infection. Remarkably, all protein alterations were normalized 100 days after leaving the hospital, which further supports the reliability of the selected proteins as hallmarks of COVID-19 progression and grading. The optimized protein panel may prove its value for optimal severity assessment as well as in the follow up of COVID-19 patients.

Keywords: COVID-19; SARS-CoV-2; biomarkers; proteomics; severity prognostics.

Figure 1

Experimental workflow Seventy-two serum samples (discovery cohort) from patients at different ages suffering from mild to severe COVID-19 were subjected to digestion, TMT-labeling, peptide fractionation, LC-MS/MS and statistical analysis with SanXoT package. To confirm our previous observations 18 proteins were selected for targeted validation by SRM in an independent cohort of 84 patients that included 21 individuals that successfully recovered from SARS-CoV-2 infection.

Figure 2

Proteins differentially regulated with disease severity. The heatmap shows protein abundance changes (expressed as Zq, or standardized log2-ratios) normalized by the average values of the non-hospitalized patient group. Statistical significance of changes (p-value) is calculated using two-tailed Student’s t-test. Proteins (quantified with more than one peptide and in the 80% in the individuals) whose abundance is significantly (p-value < 0.05) increased (red) or decreased (blue) at least in one comparison are shown (1 = Non-Hospitalized vs. Hospitalized, 2 = Hospitalized vs. ICU, 3 = ICU vs. EXITUS, 4 = Non-Hospitalized vs. EXITUS).

Figure 3

Logistic regression model using selected proteins that correlate with disease progression and severity in the discovery cohort. The panel was constructed with SERPINA3, FGA, PON1, AFM, APOA2 and TTR. The model was trained by logistic regression comparing the NH and E groups. Data are mean ± SEM of the prediction of the model per each individual. Statistical significance (p-value) is calculated using one-way ANOVA, (* p < 0.05, ** p < 0.01) and all the comparisons refer to <60 years-old non-hospitalized patients.

Figure 4

Proteins monitored by SRM. (A) The heatmap shows protein abundance changes (Zq) normalized by the average values of the 100 days after leaving hospital patient group. Up-regulated proteins are shown in red and down-regulated proteins in blue. Statistical significance of changes (p-value) is calculated using two-tailed Student’s t-test (1 = Non-Hospitalized vs. Hospitalized, 2 = Hospitalized vs. ICU, 3 = ICU vs. EXITUS, 4 = Non-Hospitalized vs. EXITUS, 5 = Exitus vs. 100 days after patient discharged). (B) Prediction of disease severity using a logistic regression model. The model was trained using the same proteins as in Figure 3. (C) Protein quantification from TMT and SRM experiments. Protein quantification is represented as the average difference of Zq values between Non-Hospitalized and EXITUS patients. Statistical significance (p-value) is calculated using one-way ANOVA (** p < 0.01, **** p < 0.0001 and all the comparisons refer to <60 years-old non-hospitalized patients.