The biomarkers' landscape of post-COVID-19 patients can suggest selective clinical interventions

Abstract

In COVID-19 clinical symptoms can persist even after negativization also in individuals who have had mild or moderate disease. We here investigated the biomarkers that define the post-COVID-19 clinical state analyzing the exhaled breath condensate (EBC) of 38 post COVID-19 patients and 38 sex and age-matched healthy controls via nuclear magnetic resonance (NMR)-based metabolomics. Predicted gene-modulated microRNAs (miRNAs) related to COVID-19 were quantified from EBC of 10 patients and 10 controls. Finally, clinical parameters from all post-COVID-19 patients were correlated with metabolomic data. Post-COVID-19 patients and controls showed different metabolic phenotype ("metabotype"). From the metabolites, by using enrichment analysis we identified miRNAs that resulted up-regulated (hsa-miR146a-5p) and down-regulated (hsa-miR-126-3p and hsa-miR-223-3p) in post-COVID-19. Taken together, our multiomics data indicate that post-COVID-19 patients before rehabilitation are characterized by persistent inflammation, dysregulation of liver, endovascular thrombotic and pulmonary processes, and physical impairment, which should be the primary clinical targets to contrast the post-acute sequelae of COVID-19.

Figure 1

Schematic diagram illustrating the overall study design.

Figure 2

Orthogonal projections to latent structures discriminant analysis (OPLS-DA) of EBC samples from post-COVID patients and controls. Scores plot showing the degree of separation of the model between post-COVID (red circles) and controls (blue circles). The model presents strong regression (95%, CV-ANOVA p < 2.3 × 10⁻¹²) and high-quality parameters (R² = 81% and Q² = 87%). The labels t[1] and t_o[1] along the axes represent the scores (the first 2 partial least-squares components) of the model, which are sufficient to build a satisfactory classification model.

Figure 3

Relative expression of miRNAs in EBC samples obtained from enrolled subjects. (a) hsa-miR-126-3p-3p. (b) hsa-miR-146a-5p-5p. (c) hsa-miR-223-3p-3p. Blue bars refer to control subjects, while red bars refer to post-COVID patients. Analysis was performed with qRT-PCR. p-values are shown.

Figure 4

Heatmap based on Pearson correlation coefficients between EBC metabolites and values obtained from clinical test in negativized COVID-19 patients. Rows and columns are rearranged according to the centroid-based correlation matrix-based hierarchical clustering (CMBHC). Blue tone indicates positive correlations between metabolites and clinical data, whereas light tones indicate negative correlations. Correlation values ρ = |0.7| are marked with a double asterisk, while values ρ = |0.5| are labeled with a single asterisk. EBC metabolites are: 1, acetoin; 2, propionate/isobutyrate; 3, propionate/valine; 4, pyruvate; 5, isobutyrate; 6, methanol; 7, 3-hydroxyisovalerate; 8, isovalerate; 9, fatty acids (FA); 10, lactate; 11, formate; 12, trimethylamine; 13, 2-hydroxyisovalerate; 14, isocaproate; 15, isovalerate; 16, valerate; 17, acetate; 18, acetone; 19, serine; 20, isopropanol; 21, glycine; 22, 3-hydroxybutyrate; 23, ethanol. The arrows on top label significant EBC metabolites, whose trend, with respect to healthy subjects, is symbolized by the arrow direction. Statistically significant clinical data are underlined.

Figure 5

Multilevel PLS-DA scores plot for post-COVID patients. The labels X-variate 1 and X-variate 2 along the axes represent the scores (the first 2 partial least-squares components) of the model, which are sufficient to build a satisfactory classification model. Admission variables (IN) are shown in red, while discharge variables (OUT) are in black.

Figure 6

Contribution plot of the principal component PC1 of the multilevel PLS-DA model including the clinical parameters of the post-COVID patients. Each bar represents the loading value for each variable on PC1. Admission variables (IN) are shown in red, while discharge variables (OUT) are in black.