Targeted plasma proteomics reveals signatures discriminating COVID-19 from sepsis with pneumonia

Plasma samples from SARS-Cov-2 infected patients with COVID-19 admitted at the intensive care or high dependency unit (n = 17, severe COVID-19) or the infectious disease clinic (n = 10, moderate COVID-19) at Karolinska University Hospital, Stockholm were collected for this study. Paired convalescent plasma samples from the COVID-19 groups were collected from 17 patients (8 and 9, from the moderate and severe COVID-19 groups, respectively) approximately 4 months after hospital discharge (median = 136 days, range = 89–153 days). Plasma samples from sepsis patients identified and enrolled in the emergency department were also included. As controls, plasma samples from age- and sex-matched SARS-CoV-2 IgG seronegative healthy volunteers (n = 16) were collected on the same days as the acute COVID-19 patients. Inclusion and exclusion criteria for patient enrollment and collected clinical information of the patients are provided below.

The COVID-19 patients were adult SARS-CoV-2 RNA positive patients who were admitted with acute illness at the Karolinska University Hospital, Stockholm, Sweden in April and May 2020, and were treated at the Infectious Diseases and Intensive Care Unit (ICU) Clinics. Patients with oxygen saturation of 90–94% and/or receiving 0.5–3 L/min of oxygen admitted to the Infectious Diseases Clinic were included to represent moderately ill COVID-19 cases. Patients treated at the ICU or high-dependency unit were included as severely ill COVID-19 cases. Exclusion criteria were age ≥ 80 years, current malignancy, or immunomodulatory treatment prior to hospitalization, to minimize immunosuppression from other causes. Corticosteroid therapy at hospital prior to sampling had been given to 2 moderate- and 12 severe- COVID-19 patients. Plasma samples from the acute phase were collected at the day of study enrollment (i.e., 5–24 days after onset of illness and 1–8 days after hospital admission).

CAP, sepsis, and septic shock patients were identified, enrolled, and sampled for plasma within 2 h after arrival at the Emergency Department of Karolinska University Hospital Huddinge in 2017–2019, as they triggered the department´s sepsis alert [22]. The sepsis alert was triggered in patients with clinical signs of infection combined with either (i) at least one of the following: oxygen saturation < 90% despite oxygen supplementation, respiratory rate > 30 per minute, heart rate > 130 per minute, systolic blood pressure < 90 mmHg, or Glasgow Coma Scale < 8; or (ii) blood lactate > 3.2 mmol/L combined with at least one of the following: oxygen saturation < 95% on room air, respiratory rate > 25 per minute, heart rate > 110 per minute, altered mental status, or temperature > 38.5 °C or < 35 °C.

Patients with pulmonary infiltrates and Influenza virus RNA detected in respiratory tract samples, without any bacterial microorganism detected were selected as Influenza CAP patients (n = 11). Patients with pulmonary infiltrates without any virus detected, but with Streptococcus pneumoniae (n = 11), Haemophilus influenzae (n = 4), or Staphylococcus aureus (n = 2) detected in blood culture or lower respiratory tract culture, or S. pneumoniae detected in nasopharyngeal culture (n = 1), or Mycoplasma pneumoniae (n = 3) detected by specific PCR on respiratory secretions were selected as bacterial CAP patients.

Patients with sepsis with a sequential organ failure assessment (SOFA) score of ≥ 2 within 12 h from arrival at the hospital and an infectious focus not including the lungs were selected as non-pneumonia sepsis patients. In these patients, blood culture was positive for Escherichia coli in 10 patients, Staphylococcus aureus in 3 patients, Group A Streptococcus in 1 patient, Group B Streptococcus in 3 patients, and Group C Streptococcus in 1 patient.

Finally, patients with infection, total SOFA score of ≥ 2, lactate > 2, and who received vasopressors were selected as septic shock patients (n = 12). These patients had the following foci of infection: lungs (n = 3), urinary tract infection (n = 5), skin-joint infection (n = 2), abdomen (n = 1), and other (n = 1). Blood culture was positive for E. coli in 3 patients (one hade multi-bacterial growth), Streptococcus pyogenes in 2 patients, and other gram-negative bacteria in 2 patients.

In the SOFA score calculation, registered values of creatinine, bilirubin, and platelet count values from the period 7–90 days prior to admission were used as baseline values. In the total SOFA score at sampling, baseline SOFA points for pathologic values of creatinine, bilirubin, or platelet count were subtracted.

Plasma was collected in whole-blood tubes containing EDTA (ethylenediaminetetraacetic acid). The COVID-19 patients and healthy controls were sampled using ordinary EDTA tubes that were centrifuged within 2 h and plasma was aspirated and frozen in aliquots at − 80 °C. The sepsis patients were sampled using PPT plasma preparation tubes containing EDTA, that were centrifuged within 2 h and then frozen at − 80 °C. The tubes were thawed once for aspiration and refreezing into aliquots at − 80 °C.

Proximity extension assays (PEA) assays (Olink AB, Uppsala, Sweden) were used for the quantification of selected soluble factors in plasma from all cohorts. All samples were measured using three biomarker panels: organ damage (v.3311), immune response (v.3203), and inflammation (v.3022), each of them targeting 92 protein analytes. Samples that had quality control (QC) warning in all three panels were excluded (1 sepsis and 1 COVID-19 patient), and protein measurements with QC warnings were assigned as missing values. Four proteins were included in two of the panels and measured twice: IL6, CCL11, IL10, and IL5; we assigned the mean of the double measurements as NPX value for the corresponding proteins. For further analyses, only analytes with less than 33% of data under the lower limit of detection (left censored) were considered, yielding a total of 193 proteins analyzed. Individual left-censored values from the analytes included in the study were imputed with the lower limit of detection value.

Protein annotation for GO (gene ontology) biological process terms was performed on the STRING platform [23], all terms related to Innate immune-, adaptive immune-, and Inflammatory- responses were grouped together. The association to specific immune cells was determined with the Human Protein Atlas (HPA) [24] data sets for RNA single cell type, RNA blood cell specific, and RNA blood lineage specific.

Coagulation factors were measured in plasma using three different multiplex panels, including the 6-, 4-, and 3-plex human ProcartaPlex panels (Thermofisher). The samples were measured according to the manufacturer’s instructions and the results are shown in arbitrary units (AU) indicating the log2 of the concentration in the samples in relation to reference plasma provided with the kit. In addition, Thrombomodulin and D-dimer were measured using a Multiplex Luminex^® assay (R&D Systems, UK). Assays were performed according to the manufacturer’s guidelines and samples were acquired on a Luminex MAGPIX instrument using xPonent 4.0 Software (Luminex).

For analysis of PEA data, two-tailed student t test was used for two-group comparisons, the test was paired in the acute-convalescent comparison. Statistical analysis of non-parametric data (clinical parameters and protein concentrations from multiplex) was performed with Mann Whitney U test for two group comparisons and Kruskal Wallis coupled with Dunn’s test. Statistical comparison of categorical variables was performed with Fisher test. All p-values were adjusted for multiple-group comparisons with the false discovery rate (FDR).

To account for the effect of confounders, we built a multivariate limma linear model [25] to adjust protein NPX values of COVID-19, CAP-Sepsis, Other Sepsis patients for age (in years), sex, Charlson comorbidity index, and corticosteroid use prior to sampling. The model included no intercept and contrasted the COVID-19 to CAP-Sepsis and COVID-19 to Other-Sepsis status separately, to derive estimated coefficients and standard errors for a corresponding comparison. Finally, moderated two-sided t statistic and F statistic were calculated for the comparisons based on adjusted log2-Fold change (FC) and empirical Bayes moderation of the standard errors, and the p-values corrected for multiple testing with the FDR.

Based on the protein expression of all 193 proteins, we plotted the 122 samples on a principal component analysis (PCA) plot and further clustered the data with the partitioning around medoids (PAM) algorithm. For PAM, we tested different number of potential clusters k, ranging from k = 2 to the maximum number of samples groups (k = 9), and eventually selected the number of maximum k (k = 4) where the samples had clear separation and the minimum subgroup sample size (n = 8).

All analyses were performed with R (v.4.0.3), in R Studio (v.1.3.959).Correlation analyses were performed with two-tailed non-parametric Spearman test applied on pairwise complete observations using the packages factoextra (v1.0.7), FactoMineR (v2.3), PerformanceAnalytics (v2.0.4), ggplot2 (v3.3.1), gplots (v3.0.4), pheatmap (v1.0.12), vegan (v2.5-6), corrplot (v0.84), lattice (v0.20-41) and latticeExtra (v0.6-29), stats (v4.0.1), and complexheatmap (v2.5.6). The plots were generated with ggplot2(v3.3.6). The package ggpubr (v0.4.0.999) was used for adding statistical significance stars into the plots. The heatmaps with dendrogram integrated were made with pheatmap (v1.0.12).

To build ML models that would accurately classify COVID-19 from CAP-sepsis, we partitioned the dataset (COVID-19 n = 27, CAP-sepsis n = 32) 1,000 times, with random allocation of 75% of the samples to the training & validation dataset (TrnVD) and 25% to the testing dataset (TstD), maintaining the probabilistic distribution of the two conditions in both TrnVD and TstD. On the 1,000 TrnVD, we first ran iterations of random forest (RF) models (R package caret, v6.0-90) for protein selection, using a leave-one-out cross-validation (LOOCV) strategy, with a minimum node size of 3 nodes, 1,000 trees, and Cohen’s kappa as a metric for model training. On the same 1,000 TrnVD, we then ran iterations of logistic regression with lasso regularization (LR-lasso) models (R package glmnet, v4.1-2) for protein selection, using a LOOCV strategy, alpha = 1, selecting minimal lambda for best model in an iteration. We opted for lasso regularization of LR models because it was the best performer of the three modelling approaches, the other two including ridge regression (alpha = 0) and elastic net (EN) regularization (alpha = 0.5); EN had a comparable performance but selected more proteins in the models.

For each iteration of either model on both TrnVD and TstD, we calculated performance metrics: accuracy, F1 score, sensitivity, specificity, positive predictive value, negative predictive value, and Mathew’s correlation coefficient (MCC). MCC is a more robust estimate of accuracy for unbalanced datasets, and it ranges from -1 (extremely low agreement) to 1 (perfect agreement). To be comparable to accuracy estimates, we transformed it to a normalized MCC (nMCC), with a range from 0 to 100%, following the equation: nMCC = (MCC + 1)/2 × 100%. We used the performance metrics on TstD for the final comparison between the RF and LR-lasso models. The performance metrics were plotted on an accuracy radar plot where mean values of accuracy are presented, along with the range and with 95% confidence intervals (CI) (± 1.9685 × SD).

This study included plasma samples from COVID-19 patients (n = 27 acute and 17 paired convalescent) and sepsis patients (n = 62, all acute), as well as healthy controls (n = 16) to obtain baseline readings (Fig. 1A). COVID-19 patients were enrolled as having either moderate or severe disease based on pre-defined inclusion and exclusion clinical criteria, as detailed in materials and methods. The sepsis patients were classified into different clinical endotypes, i.e., CAP caused by Influenza viruses or bacterial causes, non-pneumonia sepsis, and septic shock. The septic shock cohort includes mostly non-pneumonia cases but also three cases with pneumonia.

Baseline characteristics and laboratory parameters of the study cohorts are shown in Tables 1 and 2, respectively. The patients with severe or moderate COVID-19 did not differ with respect to age, sex, and comorbidities (Table 1; Fig. 1B). However, patients with severe COVID-19 had higher total SOFA score and significantly impaired lung function (e.g., increased respiratory SOFA, decreased PaO₂/FiO₂), as compared to patients with moderate COVID-19 (Additional file 2: Table S1; Fig. 1C). In addition, both COVID-19 cohorts displayed abnormal levels of common laboratory markers with the most pathologic levels seen among severe cases (Table 2; Fig. 1C). While lymphocyte and leukocyte counts showed no significant differences between the two COVID-19 severity groups (Additional file 2: Table S1), there was a significant increase of neutrophil counts during severe COVID-19 (Additional file 2: Table S1; Fig. 1C). Compared with the COVID-19 patients, the sepsis cohort patients were older and had higher Charlson comorbidity index (Tables 1; Additional file 2: Table S1; Fig. 1B). Severe COVID-19 and septic shock groups had the highest total SOFA score, while the respiratory SOFA was significantly higher in the severe COVID-19 group as compared to the other groups, including those with pneumonia (Fig. 1C; Additional file 2: Table S1). Additionally, higher levels of creatinine were noted in the non-pneumonia sepsis and septic shock cohorts. This last group also displayed higher levels of Procalcitonin in comparison to the other cohorts (Fig. 1C; Additional file 2: Table S1).

Plasma protein profiles from the patient cohorts and from healthy controls were obtained through PEA using the three Olink 96-targets’ panels for inflammation, immune response, and organ damage. Among the 273 analytes measured, proteins with more than 33% missing values (i.e., under the limit of detection) were excluded, yielding 193 proteins for the final comparative analyses (Additional file 2: Table S3). Unsupervised clustering analyses grouped the patients into four groups, separating acute COVID-19 patients (cluster 1) from healthy controls and convalescent COVID-19 patients (cluster 2), the two CAP groups caused by influenza and bacteria (cluster 3), and septic shock and non-pneumonia sepsis patients (cluster 4) (Fig. 2A). The clearest separation was noticeable on the first principal component, where COVID-19 patients clustered closer to healthy individuals than to sepsis patients. Of note, the cluster dominated by CAP patients (cluster 3) also included three acute COVID patients as well as three septic shock patients; the latter of which all had pneumonia. Hierarchical clustering based on the protein abundance revealed that the separation between the groups could be explained by higher levels of most of the measured proteins in the sepsis cohorts as compared to the COVID-19 patients and healthy controls (Fig. 2B). Severe COVID-19 patients displayed higher levels of proteins than in moderate COVID-19. Although COVID-19 has been reported as a cytokine-storm driven disease [26], the levels of the classical sepsis-associated proteins, i.e., IL6, CXCL8 (Interleukin-8), IL10, IL12, TNF, and IFNγ, did not reach the same magnitude as in CAP-sepsis, non-pneumonia sepsis or septic shock (Fig. 2C). The hierarchical clustering also showed that during convalescence, most proteins normalized and displayed a similar profile to healthy controls (Fig. 2B; Additional file 1: Fig. S1).

To detect alterations of plasma protein levels in each patient cohort, we used as reference the levels in healthy controls. First, we assessed the influence of etiology by comparing the differential expression profiles between patients with lung infections caused by different microbes, i.e., SARS-CoV-2 (COVID-19), Influenza virus (CAP-Infl), or bacteria (CAP-Bac). The results revealed a shared response of 45 proteins, as well as 7, 6, and 5 unique proteins specific for SARS-CoV-2, Influenza, and bacteria, respectively (Fig. 2D; Additional file 2: Table S4). In general, the response profiles, including both shared and unique proteins, showed an equal representation from all three PEA protein panels (inflammation, immune response, and organ damage), except for the set of proteins unique to severe COVID-19, which was dominated by markers of the organ damage response (Fig. 2E). Next, we examined the response profiles in COVID-19 versus other sepsis cases, i.e., non-pneumonia sepsis or septic shock (Fig. 2F). Again, the results revealed a shared core response of 43 proteins with altered abundance, while 5, 9, and 3 were unique to COVID-19, non-pneumonia sepsis and septic shock, respectively. The unique signatures revealed a similar dominance of organ damage proteins in severe COVID-19, while the non-pneumonia sepsis showed a predominance of immune response related proteins (Fig. 2G; Additional file 2: Table S4). The shared core profiles in Fig. 2D, F were essentially identical (42 proteins) and likely depict the basal host response to infection independent of etiology or focus of infection. However, many of the shared factors differed in expression, including several of the pro-inflammatory markers that had substantially higher levels in both CAP-sepsis and other sepsis cohorts, as compared to COVID-19 (Figs. 2H, I; Additional file 2: Table S5). In contrast, only two proteins, pleiotrophin (PTN) and keratin-19 (KRT19), were upregulated in COVID-19. By stratifying the comparison to CAP based on etiology, PTN was consistently higher in COVID-19 compared to both CAP-Infl and CAP-Bac, while KRT19 was only higher when compared to CAP-Bac but not to CAP-Infl (Additional file 1: Fig. S1). Moreover, we observed that the vast majority of these differences remain even after adjustment for the following confounders age, sex, Charlson comorbidity index and the use of corticosteroids prior to sampling (Additional file 1: Fig. S2; Additional file 2: Tables S5, S6).

Taking all patient cohorts into account, a set of five proteins were unique to COVID-19 (Fig. 2J). Among these, four (CLEC4A, DSG4, FAM3B, and RARRES1) displayed significantly lower levels in both moderate and severe COVID-19, as compared to the healthy controls and the sepsis cohorts. ITGB6 had higher levels in severe COVID-19 as compared to all other sepsis cohorts. In contrast, moderate COVID-19 patients had lower ITGB6 levels compared to all other cohorts.

Moderate and severe forms of COVID-19 almost consistently present with pneumonia [27‐29], posing as a challenge to differentiate from CAP-sepsis caused by other agents [30, 31]. Therefore, to identify plasma proteins that can serve as biomarkers for accurate differentiation of COVID-19 and CAP-sepsis patients, we employed two ML algorithms, i.e., RF and LR-lasso (Additional file 1: Figure S3).

Seven proteins, i.e., TRIM21, CASP8, NBN, FOXO1, PIK3AP1, PTN, and BID, had higher average variable importance for the models and were repeatedly selected as biomarkers in the 1000 iterations of the RF models (Fig. 3A). On average, the models had high accuracy in differentiating COVID-19 from CAP on both the training (mean accuracy = 95.01%, range: 91.11–100%) and testing data (mean accuracy = 93.78%, range: 71.43–100%). Four of the five top models with highest accuracy (98% and 100% on training and testing data, respectively), consisted of single proteins TRIM21, PIK3AP1 and NBN, and one model consisted of CASP8 and FLT3LG (Additional file 1: Fig. S4A). In comparison, among the 1000 iterations of the LR-lasso models, four proteins were selected in > 50% of the iterations: PTN, CASP8, CSF1, and TRIM21 (Fig. 3B). Overall, the distribution of performance metrics of the LR-lasso was similar to the RF models, with higher accuracy of LR-lasso models in training data (mean accuracy = 97.24%, range: 93.33–100%, Wilcoxon test: p < 2.2 × 10^–16) and no difference in testing data (mean accuracy: 93.97%, range: 71.43–100%, Wilcoxon test: p = 0.394). There was a trade-off between slightly better positive predictive value (PPV) and specificity versus worse sensitivity in the LR-lasso models compared to RF models (Fig. 3C, Wilcoxon test: p < 0.05). However, the best LR-lasso models outperformed the best RF models; 93 LR-lasso models had 100% accuracy in both training and testing data, where the algorithms selected from 4 to 17 proteins as predictors in the models. As a final biomarker panel, we selected the LR-lasso model with the smallest panel of plasma proteins that had 100% accuracy in all the metrics, including the AUC (Area Under the Curve) in ROC (Receiver Operating Characteristics) curves (Fig. 3D). This model consisted of four proteins; PTN and CSF1, whose higher levels predicted COVID-19, and TRIM21 and CASP8, whose higher levels predicted CAP.

To assess how the most frequently selected plasma proteins performed as single biomarkers compared to existing clinical variables, we tested their sensitivity and specificity in differentiating COVID-19 from CAP through ROC curves. Among the existing clinical biomarkers, plasma albumin had the best discriminating power (Fig. 3E). Nonetheless, most plasma protein biomarkers that were identified with the two ML algorithms outperformed all the clinical variables in differentiating COVID-19 from CAP, with TRIM21 having the highest AUC (Fig. 3F, see Additional file 1: Fig. S4B, C for ROC curves of the remaining proteins). Single plasma proteins had higher accuracy in differentiating the two conditions than any of the remaining clinical biomarkers, with only four proteins being sufficient to obtain full differentiation between COVID-19 and CAP.

In comparison to healthy controls, 57 proteins were differentially altered in both COVID-19 groups, regardless of severity, whereas 63 proteins were identified only in severe cases (Additional file 2: Table S4). Most of these markers displayed higher levels than those observed in the healthy controls (Additional file 1: Fig. S5A). Furthermore, a direct comparison of severe to moderate COVID-19 patients showed that among the shared proteins, 44 had higher levels in severe cases, while only 3 proteins had higher levels in moderate cases (Fig. 4A; Additional file 1: Fig. S5B). Most of these proteins correlated with total SOFA, respiratory SOFA score, and PaO₂/FiO₂ ratio; as well as with leukocyte and neutrophil counts in the COVID-19 patients (Fig. 4B). In contrast, such associations were not recapitulated in any of the sepsis cohorts (Additional file 1: Fig. S6), not even in those with pulmonary affection, i.e., the CAP cohort (Fig. 4C), implicating a particular role of these proteins in the pathophysiology of SARS-CoV-2 infections. To validate our findings, we analyzed a publicly available plasma proteomics dataset on acute phase COVID-19 samples classified according to the WHO severity grades [8]. We compared the WHO severity grades IV and II, which approximate our classification of moderate and severe cohorts, respectively. This dataset included 42 out of the 47 severity-associated proteins. Among these, 27 (64%) recapitulated the statistically significant alterations between the patient groups (Additional file 1: Fig. S7).

To seek further functional insight, we annotated the 47 COVID-19 severity-associated proteins with GO terms, which revealed that the most frequent terms related to innate immune responses and furthermore, that the proteins’ main cellular sources were granulocytes and monocytes (Fig. 4D). Developing on our previous studies using high-dimensional flow cytometry for immunophenotyping of subpopulations of granulocytes and monocytes on our COVID-19 patient cohort [16, 17], we were able to correlate severity-implicated plasma proteins to specific cell surface markers defining immune cell subpopulations. Increased levels of almost all soluble proteins correlated to low expression of activation markers in both granulocyte and monocyte populations, indicating a link between the soluble markers and immature or exhausted innate immune cells in severe COVID-19 (Additional file 1: Fig. S8). In particular, we found an association between plasma proteins elevated in severe COVID-19, such as HGF, AREG, CKAP4, S100A12, NCF2, ITGB6, and a subpopulation of monocytes with lower expression of CD86 and HLA-DR, which are characteristics of myeloid-derived suppressor-like cells (Figure S8A). Likewise, the high levels of most soluble factors correlated with decreased expression of activation markers in neutrophils, e.g., CD16 and CD69, indicating an association with increased frequency of immature CD16^dim (Additional file 1: Fig. S8B). Similarly, most soluble factors were inversely correlated to the expression of activation markers in basophils (e.g., CD11b, CD62L and CD177) and eosinophils (e.g., CD66b and CD193), while a positive correlation was observed between the levels of many factors and the activation markers CXCR4 and FceR1 in eosinophils and basophils, respectively.

We further examined the levels of these 47 proteins during convalescent phase four months after acute disease. At this time point, all patients subjected to convalescent sampling had recovered, although some presented with persisting cough. While most proteins associated to severity normalized during convalescence phase (Additional file 1: Fig. S5B), only HGF and KRT19 remained significantly higher as compared to healthy controls (Fig. 4E; Additional file 2: Table S4). Interestingly, a set of six proteins had a similar behavior to KRT19 and HGF during the acute phase of COVID-19, except that they reached healthy levels during convalescence (Fig. 4F). Among the five proteins uniquely upregulated in COVID-19 patients compared with controls (see Fig. 2J), only ITGB6 had higher levels in acute severe patients compared to healthy controls and remained higher during convalescence (Fig. 4G). The three proteins that were lower in COVID-19 cases compared to healthy controls and had the lowest levels in severe cases, i.e., CLEC4C, LTA (Lymphotoxin α), and ITGA11, all normalized during convalescence (Fig. 4H). Although statistically non-significant at 5% FDR, IFNγ showed the largest difference with lower levels in severe compared to moderate cases (log2-FC = − 2.525, p-value = 0.036, adj. p-value = 0.104). Unlike the previous three proteins, IFNγ levels were above the healthy range in both the acute and convalescence phases.

Severe infections that trigger a systemic inflammatory response, like sepsis, commonly present coagulopathies [32]. This effect is also seen in patients with COVID-19 admitted into intensive care, who frequently present with thrombotic complications. Since it has been reported that the coagulation abnormalities presented in COVID-19 patients differ from those in patients with sepsis or trauma [33], we extended the plasma profiling to 14 coagulation factors assessed by Luminex® multiplex.

In our cohorts, there were no significant differences in prothrombin time (INR) and platelet counts as compared to reference levels (Fig. 1C), not even in three out of the four severe COVID-19 cases with reported thromboembolic events. The Luminex^® d-dimer measurements in our patient cohorts were significantly higher compared to healthy controls, but no differences were observed between patients with pulmonary infections of different etiology (COVID-19 vs. CAP). The highest concentrations of D-dimer were measured in septic shock patients and their levels were only significantly higher in comparison to COVID-19 patients, but not to other septic cohorts (Fig. 5A; Additional file 2: Tables S7, S8).

When comparing COVID-19 cases to healthy controls, significant differences were observed in the von Willebrand factor (vWF) and factor XIII levels. Additionally, severe COVID-19 cases also displayed elevated levels of factor VIII and thrombomodulin. All these factors, except thrombomodulin, returned to normal levels in the convalescence phase (Fig. 5A). In contrast, sepsis cases showed more profound changes in the coagulation pathway. Plasma samples from all sepsis cases also displayed an increase in vWF concentration as well as differences in the concentrations of factors VII, V, prothrombin, and thrombomodulin (Fig. 5A). Moreover, differences in factors XII and XI were also noted for CAP and NP patients. Like COVID-19, lower levels of factor XIII were found in CAP-Bac, NP sepsis and septic shock. When comparing the concentrations of antithrombin, protein C and protein S, which regulate the coagulation cascade, we observed that COVID-19 and influenza pneumonia patients did not show significant differences in the concentration of these proteins. In contrast, patients with CAP bacteria and NP sepsis had lower levels of all three of them (Fig. 5A; Additional file 2: Tables S7, S8).

Assessing coagulation factors to clinical markers of severity, we observed that in COVID-19 patients, vWF, XIII, VIII, and thrombomodulin levels correlated with lung function impairment (defined by SOFA respiratory score), whereas such correlation was not seen in the other sepsis cohorts (Fig. 5B). Although our study does not directly conclude on the incidence of coagulopathies in either COVID-19 or sepsis patients, possible tissue damage could have contributed to changes observed in the coagulation cascade, where all proteins altered during COVID-19 (vWF, thrombomodulin, Factor XIII and Factor VIII) are linked to injury and wound repair [34‐36]. Overall, these results indicate that the pathophysiology underlying the coagulation abnormalities in COVID-19 may differ from those in sepsis.