Variation in presepsin and thrombomodulin levels for predicting COVID-19 mortality

Coronavirus disease (COVID-19) has caused extensive mortality globally; therefore, biomarkers predicting the severity and prognosis of COVID-19 are essential. This study aimed to evaluate the application of presepsin (P-SEP) and thrombomodulin (TM), which are biomarkers of sepsis and endothelial dysfunction, respectively, in the prognosis of COVID-19. Serum P-SEP and TM levels from COVID-19 patients (n = 183) were measured. Disease severity was classified as mild, moderate I, moderate II, or severe based on hemoglobin oxygen saturation and the history of intensive care unit transfer or use of ventilation at admission. Patients in the severe group were further divided into survivors and non-survivors. P-SEP and TM levels were significantly higher in the severe group than those in the mild group, even after adjusting for creatinine values. In addition, TM levels were significantly higher in non-survivors than in survivors. Changes in the P-SEP levels at two time points with an interval of 4.1 ± 2.2 days were significantly different between the survivors and non-survivors. In conclusion, TM and continuous P-SEP measurements may be useful for predicting mortality in patients with COVID-19. Moreover, our data indicate that P-SEP and TM values after creatinine adjustment could be independent predictive markers, apart from renal function.


Basic profiles and laboratory findings
There was a significant difference in age and lengths of hospital stay among the groups, but there was no significant difference in sex (Table 1).During hospitalization, 16 patients, including three patients classified as mild, died 36.6 ± 25.8 days after admission, and 167 patients were discharged after 13.2 ± 14.1 days.The routine test findings were compared among the four severity groups.Various tests showed significant differences among the groups (Table 1).For example, the levels of WBC, CRP, and lactate dehydrogenase (inflammatory markers) and the activated partial thromboplastin time, fibrinogen, and D-dimer (coagulation-fibrinolysis markers) tended to increase with an increase in disease severity.Regarding renal function markers, there were no significant differences in levels of CRE, inorganic phosphorus, or uric acid among the groups, whereas urea nitrogen (UN) levels and estimated glomerular filtration rate (eGFR) were significantly higher and lower, respectively, in the severe group than those in the other groups.As for liver disorder markers, aspartate aminotransferase, alanine aminotransferase, and gamma-glutamyltransferase levels were higher in the severe group than in the mild group.In contrast, levels of red blood cells, hemoglobin, total protein, albumin, and calcium were lower in the severe group than in the mild group.

Differences in P-SEP and TM at administration based on disease severity
Initially, we confirmed whether P-SEP and TM values were associated with levels of the renal dysfunction markers, CRE and UN.P-SEP values were significantly correlated with both CRE and UN values (r = 0.320 and 0.346, respectively, p < 0.001) (Fig. 2A,B), although the bias of scatter blots was observed below 3 mg/dL CRE.Similar tendencies were observed in the comparison between TM, CRE, and UN levels.The levels of TM were positively correlated with those of CRE and UN (r = 0.426 and 0.590, respectively, p < 0.001) (Fig. 2C,D).
Further, we evaluated P-SEP and TM values in each severity group.The averages of P-SEP values in all the groups were above the cut-off value for the diagnosis of sepsis (> 500 pg/mL).The averages of serum TM values in the mild, moderate I, and moderate II groups were within the reference range (12.1-24.9U/mL).Using the P-SEP and TM values, we calculated various parameters to determine whether these values would differentiate the patients classified into groups based on their severity (Table 2).P-SEP levels of patients with COVID-19 at administration were significantly higher in the severe group (1245.6 ± 761.5 pg/mL) than in the mild group (812.9 ± 730.7 pg/mL) (p < 0.001) (Fig. 3A).TM levels at administration were significantly higher in the severe group (31.7 ± 18.4 U/mL) than the mild group (17.0 ± 13.4 U/mL) (p < 0.001), moderate I group (16.9 ± 9.2 U/mL) (p = 0.002), and moderate II group (15.2 ± 10.5 U/mL) (p = 0.009) (Fig. 3B).There were no significant differences in CRE levels among the groups (Table 1); however, P-SEP and TM were excreted by the kidney, as stated above.Therefore, we adjusted both these values for the CRE value and compared the difference in each value among

Comparison between severe survival and non-survival cases
Patients in the severe group were further divided into survivors (n = 34) and non-survivors (n = 13), and the P-SEP and TM levels were compared between them.There was no significant difference in the levels of P-SEP and inflammatory markers (CRP and WBC) between the survivors and non-survivors (Fig. 4).In contrast, non-survivors had significantly higher TM levels (41.23 ± 23.92 U/mL) at admission than the survivors (27.91 ± 14.53 U/ mL) (p = 0.026), whereas there was no significant difference in the fibrinogen and D-dimer levels between them.When the TM values were adjusted for CRE, the TM/CRE values showed no significant differences between them.As for renal function markers, there was no significant difference in the CRE levels (p = 0.141), whereas UN levels were significantly higher in non-survivors than in survivors (p = 0.013).Moreover, eGFR was lower in non-survivors than in the survivors (p = 0.004).
Furthermore, we focused on the changes in P-SEP and TM levels after administration.Survivors (n = 25 for P-SEP and n = 24 for TM) and non-survivors (n = 13) with more than two P-SEP and TM measurements were analyzed.The elapsed number of days (mean ± standard deviation (SD)) between the first and second tests for both P-SEP and TM was 4.1 ± 2.2 days.As for P-SEP, when the ratio of the second measurement value to the www.nature.com/scientificreports/first value (P-SEP ratio) was calculated, there was a significant difference between survivors and non-survivors (p = 0.004) (Fig. 5A).The ratio in the survivors was less than 1.0 (0.77 ± 0.61), while the ratio in non-survivors was elevated (1.30 ± 0.70).Moreover, when P-SEP was adjusted for CRE, the change in P-SEP/CRE (P-SEP/CRE ratio) was still higher in non-survivors (p = 0.002) (Fig. 5C).In contrast, TM values in both cases increased from the first measurement at admission, and the TM ratio in survivors and non-survivors was 1.46 ± 0.72 and 1.26 ± 0.46, respectively, while there was no significant difference in the ratios between both groups (Fig. 5B).
Similarly, there was no significant difference in ratios of CRP, WBC, D-dimer, and fibrinogen between the survivors and non-survivors (Fig. 5E-H).

Receiver operating characteristic (ROC) curve analysis
We further evaluated the ability of P-SEP and TM values to predict in-hospital mortality using ROC curves (Fig. 5).Since the P-SEP ratio showed stronger statistical power than that of the P-SEP value in terms of differentiation between survivors and non-survivors, the P-SEP ratio was used for ROC analysis.The areas under the ROC curves (AUROC) (95% confidence interval (CI), asymptotic significance) of models using the P-SEP ratio, CRP, and WBC were 0.782 (95% CI 0.637-0.927,p = 0.005), 0.615 (95% CI 0.428-0.803,p = 0.249), and 0.537 www.nature.com/scientificreports/(95% CI 0.327-0.747,p = 0.712), respectively (Fig. 5I).When ROC analysis was performed using TM, D-dimer, and fibrinogen, TM showed an insignificantly higher AUROC than that of the other coagulation-fibrinolysis markers (Fig. 5J).The AUROC of TM, D-dimer, and fibrinogen were 0.667 (95% CI 0.496-0.839,p = 0.083), 0.439 (95% CI 0.240-0.639,p = 0.529), and 0.520 (95% CI 0.329-0.710,p = 0.837), respectively.According to the largest Youden Index, the optimum cut-off values for the P-SEP ratio and TM were 0.67 and 27.3 U/mL, respectively.Since the change in P-SEP concentration can be useful for predicting mortality, we monitored P-SEP and TM levels in patients in the severe group.The representative time courses of P-SEP and TM for survivors and nonsurvivors are shown in Fig. 6 (for P-SEP) and Fig. 7 (for TM).As shown in Fig. 5, there was no significant difference in P-SEP levels on admission between the survivors and non-survivors.However, in survivors, the levels of P-SEP on second testing decreased and continued to decline or remained low until the discharge date (Fig. 6A,B).Similar tendencies were observed for CRP and WBC values.In contrast, in non-survivors, P-SEP values did not decrease on the second testing day and drastically increased before death (Fig. 6C,D).As for TM, as shown in Fig. 5, the values on admission were higher in non-survivors than those in the survivors (Fig. 7).However, unlike P-SEP, a decrease in TM levels was not observed in survivors (Fig. 7A,B), even though fibrinogen levels tended to decrease until discharge.Interestingly, D-dimer levels increased until the middle of hospitalization and then decreased to normal levels.In contrast, TM and D-dimer levels increased drastically, similar to P-SEP among non-survivors, while an increase in fibrinogen levels was not observed (Fig. 7C,D).

Discussion
Sepsis, respiratory failure, and ARDS are among the most common symptoms observed in COVID-19 nonsurvivors 30 ; thus, in this study, we focused on the sepsis biomarker, P-SEP, and endothelial dysfunction observed in ARDS, which are still unclear topics in the literature.For comparison with P-SEP and TM, we selected CRP and WBC as representative inflammatory biomarkers, and fibrinogen and D-dimer as representative coagulationfibrinolysis biomarkers.These are common testing measures upon patient admission, and we thought that they could be compared during inflammatory conditions such as sepsis and endothelial dysfunction.Moreover, to our knowledge, no study has utilized CRE adjustment for evaluating patients with COVID-19 or evaluated the change in these values between two time points.With regard to the classification of severity, P-SEP and TM levels reflected the aggravation of symptoms, while other major markers, such as WBC, CRP, fibrinogen, and D-dimer, were also differentiated based on severity.In contrast, only TM levels showed a significant difference in mortality prediction.TM plays a role in the suppression of the coagulation pathway by activating protein C and inactivating activated coagulation factors V and VIII in cooperation with thrombin 31,32 .However, under various pathological conditions, the expression of TM can be modulated.Cytokines are major factors in the onset of disseminated intravascular coagulation associated with severe infectious diseases such as sepsis.Many tissue factors are produced from monocytes, macrophages, and vascular endothelial cells by inflammatory cytokines such as lipopolysaccharide, tumor necrosis factor, and interleukin-1, resulting in remarkable coagulation activation 33 .The released lipopolysaccharide and cytokines suppress the expression of TM, an anticoagulant protein in vascular endothelial cells, and accelerate coagulation activation.In addition, lipopolysaccharide and cytokines overexpress plasminogen activator inhibitor, which is a fibrinolytic inhibitor, in vascular endothelial cells and suppress fibrinolysis, resulting in multiple organ failure due to residual microthrombus 34,35 .Thus, suppression of TM expression in endothelial cells is associated with the pathological progression of coagulation.At the same time, TM is released from the injured cells into circulation, and the elevated plasma TM levels can be used as a biomarker for endothelial cell injury.With regard to COVID-19, the expression of TM in patient's injured pulmonary endothelial cells is also decreased and is associated with immune cell infiltration in the lungs 36 .Moreover, high circulating platelet aggregates were observed in non-survivor patients with COVID-19 and correlated with TM levels more than with D-dimer and fibrinogen levels 37 , as observed in our current study.Although contradictory results for blood TM levels as a biomarker have been reported, high TM levels may reflect endothelial disorder and coagulation promotion, which could prove fatal, similar to the present study findings.
Furthermore, P-SEP levels did not show a significant difference between survivors and non-survivors, whereas the change in P-SEP values reflected mortality.P-SEP is a relatively novel septic marker, and many studies have demonstrated its usefulness for the early diagnosis of sepsis and prognosis compared with other septic markers.The AUC of P-SEP calculated from the ROC curves in the infection group was significantly greater than that of procalcitonin, CRP, and IL-6 8 .P-SEP is more useful in classifying the stage of sepsis than other septic markers (procalcitonin, CRP, and IL-6) 38 .Considering that most non-survivors with COVID-19 showed symptoms of sepsis 1 , the dynamics of P-SEP values in patients with COVID-19 might be similar to those in patients with sepsis.In this study, 7 patients in the non-survivor group died due to sepsis.However, when serum samples were collected at admission for P-SEP and TM levels assessment, sepsis was deniable, at least, in cases of the 4 patients, while 1 patient had already developed sepsis at admission.Generally, P-SEP values do not increase in patients with viral infections; however, SARS-CoV-2 can directly infect monocytes to reduce CD14 + /CD16 - classical monocytes and increase CD14 + /CD16 + intermediate monocytes 39 , which have an increased phagocytic function 40 , resulting in the release of cytokines including P-SEP in the early stage of the disease.This could explain the high P-SEP values even in the mild group.To our knowledge, this is the first study to evaluate P-SEP values at multiple time points as well as the change in values after admission for patients with COVID-19, except for the case studies.In a previous study that investigated P-SEP levels in patients with sepsis, the P-SEP levels gradually increased from the early stage based on the severity of sepsis 8,41 .Elevation in P-SEP values in nonsurvivors was observed over time; however, these elevations were not observed in survivors 42 .A recent study has reported that P-SEP is one of the valuable prognostic biomarkers in assessing ICU mortality risk in COVID-19 patients independent of dexamethasone administration 43 .As with sepsis, changes in P-SEP levels might provide more important information for the prediction of mortality, even in COVID-19 patients.In addition, some non-survivor patients with COVID-19 showed renal dysfunction, and in this study, some patients in the severe group (particularly the non-survivor group) had high CRE levels.However, significant differences in P-SEP and TM values remained even after CRE adjustment, which indicated that both might be used as predictive markers,

Figure 2 .
Figure 2. Correlation of presepsin (P-SEP) and thrombomodulin (TM) levels with renal function markers.P-SEP values in all patients were compared with (A) creatinine (CRE) and (B) urea nitrogen (UN) values (n = 183).The TM values were also compared with the (C) CRE and (D) UN values (n = 182).One patient who received recombinant human TM treatment was excluded from the TM analysis.Correlations were evaluated using Spearman's rank correlation test.

Table 2 .
Data for P-SEP-and TM-related parameters among the severity groups.P-SEP presepsin, TM thrombomodulin, P-SEP/CRE P-SEP adjusted for CRE, TM/CRE TM adjusted for CRE.*Significant difference; Kruskal-Wallis test.