The role of Plasminogen Activator Inhibitor 1 in predicting sepsis-associated liver dysfunction: an observational study

Ewa Woznica Niesobska (  ewa.anna.woznica@gmail.com ) Uniwersytet Medyczny im Piastów Śląskich we Wrocławiu https://orcid.org/0000-0002-6400-9560 Patrycja Lesnik Uniwersytet Medyczny im Piastow Slaskich we Wroclawiu Ryszard Woznica Wadowice District Hospital Jaroslaw Janc 4 Wojskowy Szpital Kliniczny z Poliklinika SP ZOZ we Wroclawiu Malgorzata Zalewska Uniwersytet Medyczny im Piastow Slaskich we Wroclawiu Lidia Lysenko Uniwersytet Medyczny im Piastow Slaskich we Wroclawiu

In sepsis, infection combined with hyperreactivty of in ammatory response and microcirculatory failure contribute to the organ dysfunction. In available reports, due to the lack of homogenous diagnostic criteria, the incidence of SALD varies from 1,1% up to 34,7% [2]. It may present as hypoxic hepatitis (HH), sepsis-induced cholestasis and/or coagulopathy [3] and the degree of organ injury may range from a mild liver dysfunction to a life-threatening liver failure.
According to Sepsis-3 criteria an acute change in sequential organ failure assessment (SOFA) score of ≥ 2 in response to an infection allows to diagnose a life-threatening organ dysfunction de ned as sepsis [4]. An increase in bilirubin serum concentration can be a signal of developing liver dysfunction. Elevation of bilirubin concentration in human plasma of > 1,9 mg/dl gives SOFA score of 2. Bilirubin itself is late and non-speci c marker of hepatic dysfunction. Hyperbilirubinemia may be caused by hemolysis, cholestasis and hepatic dysfunction of multiple origins: decreased bilirubin transport, uptake and clearance, as well as hepatic ischemia, hepatocellular damage [2].
Acute liver failure (ALF) is de ned as an acute liver injury with an onset of hepatic encephalopathy (HE) and an increase in International Normalised Ratio (INR) of > 1,5, in patient with no pre-existing liver disease [5]. In an ICU setting, due to multiple possible causes of impaired consciousness it is di cult to diagnose HE. Sepsis induced HH is a rare cause of ALF, mainly due to organ hypoperfusion. However, an increased blood ow and cardiac output in the primary stages of septic shock might not be enough to compensate an increased hepatic oxygen demand and also result in the development of HH [2]. Ischemic/hypoxic hepatitis is reported in up to 2,5% of all the ICU patients [6].
We postulated that the de nition of SALD requires rede ning and creating more speci c diagnostic criteria. The aim of the study was to identify plasma biomarkers, which can be used for an early diagnosis of SALD.

Materials And Methods
A single-centre, prospective observational study was conducted in a 14-bed medicosurgical ICU at Wroclaw University Hospital from September 2015 to April 2019. The study was approved by the Bioethics Committee of Wroclaw Medical University, informed consent was obtained from all patients or their families.
The study enrolled adult patients admitted to the ICU due to sepsis/ septic shock and patients who have developed sepsis/septic shock during ICU stay. Sepsis/ septic shock was de ned according to the Sepsis-3 criteria [4]. Exclusion criteria were: age <18years old, pregnancy, pre-existing liver disease (cholestatic disorders; genetic, vascular, metabolic liver diseases; viral hepatitis; liver tumours), decompensated liver cirrhosis (Child-Pugh class B or C), immunosuppression, HIV infection and cancer. Patients with missing data were excluded from the study. All patients have been treated according to Surviving Sepsis Guidelines 2012 (patients admitted to the ICU before publication of new de nition and guidelines) and Surviving Sepsis Guidelines 2016 [4,7].
Once the patient had been quali ed for the study, clinical and demographic data including age, sex, comorbidity, Acute Physiology and Chronic Health Evaluation (APACHE II) score, SOFA score and origin of sepsis were recorded in the study protocol. APACHE II and SOFA score were subsequently recorded on days 1, 3, 5, 7 and 14. On the same days patients were also screened for the development of Disseminated Intravascular Coagulation (DIC) using ISTH Criteria for DIC [8].
Blood samples were collected and centrifuged at 3000 rpm at room temperature in EDTA tubes, consecutively samples were frozen at -28 Celsius degree within 30 minutes of collection.
Levels of human hepcidin, IP-10, PAI-1, ET-1 were measured using quantikine ELISA tests (R&D Systems Inc., Minneapolis, MN, USA). Human TAT complex levels were measured using Assay Max (Assaypro LLC, St. Charles, USA). All assays employ the quantitative sandwich enzyme immunoassay technique. Tests were performed and interpreted according to the manufacturer's instructions.
Taking into consideration SOFA scoring, we de ned SALD as an acute elevation of serum bilirubin level of 2 mg/dl or more, excluding other than sepsis causes of hyperbilirubinemia.
The primary endpoint of the study was the development of SALD, while staying in the ICU. The secondary endpoint was 28-day overall survival.

IP-10
IP-10 is induced in different cells, ie. leukocytes (neutrophils, monocytes, macrophages), in response to type 1 and 2 interferons (IFN) and lipopolysaccharide (LPS) stimulation. IP-10 induces apoptosis, cell growth inhibition, chemotaxis and angiostasis [9]. IP-10 activates C-X-C motif chemokine receptor 3 (CXCR3), in response to viral infections, autoimmune diseases, allotransplantation and cancer, which is an important regulator of natural killer (NK), natural killer T (NKT) and T helper (Th)1 lymphocyte tra cking [10]. In liver, IP-10 is secreted by hepatocytes in areas of lobular in ammation and may be responsible for the development of intrahepatic in ammation. IP-10 is involved in the pathogenesis of hepatitis C and hepatitis B and recently it is shown to play a pivotal role in the pathogenesis of experimental steatohepatitis [11,12,13].

ET-1
Elevated ET-1 plasma levels have been reported in patients with septic shock [14]. ET-1 is a vasoconstrictive molecule synthesizsed by endothelial cells in response to its injury. It also stimulates phagocytosis and chemotaxis of monocytes/macrophages and neutrophils therefore augmenting the in ammatory response [15]. In a murine model, hepatic macrophages were shown to be the primary source of elevated plasma levels of ET-1 [16]. Hepatic stellate cells, sinusoidal endothelial cells and Kupffer cells are the hepatic source of ET-1. ET-1 being a local regulator of hepatic sinusoidal microcirculation, acts via ET A / ET B2 receptors on hepatic stellate cells, causing liver sinusoid constriction [17]. Endothelin-mediated microcirculatory failure leads to hepatocellular injury via worsening the oxygen delivery and metabolic dysfunction during sepsis [18].

Hepcidin
Hepcidin is a peptide of mostly hepatic origin involved in iron metabolism. Increased serum hepcidin levels were observed in neoplastic diseases, in ammation and sepsis [19]. Its expression is suppressed by iron de ciency, anaemia, and hypoxia, but induced by iron overload, in ammatory stimuli and LPS [20]. During in ammatory states hepcidin expression is induced via cytokine IL-6 [21]. Antimicrobial function of hepcidin was shown in several studies, but the mechanism of both antibacterial and antifungal action is not yet well understood [19. Iron metabolism may be a useful predictor of outcome in patients with liver disease. In murine N-acetyl-p-amino-phenol (APAP) -induced acute liver failure model, hepcidin was shown to be an independent predictor of mortality [22].
TAT TAT, a marker of thrombin generation, forms following the neutralisation of thrombin by antithrombin III (AT III) [23]. AT III is synthesized in liver and is a natural anticoagulant. Its antiin ammatory function is due to neutralisation of thrombin, which is responsible for leukocyte rolling and adhesion, but also it depends on blocking the effect of protease activated receptor-1 [24]. TAT can be used as a sensitive parameter of latent activation of the clotting pathway. The rise of TAT suggests continuous thrombin generation and antithrombin depletion [25]. In patients with liver cirrhosis, elevated TAT levels are observed also in patients with Child-Pugh A. TAT and AT III are thought to be independently associated with occurrence of liver dysfunction [26].

PAI-1
PAI-1 is a principal inhibitor of brinolysis. Elevated PAI-1 is linked to sepsis-induced coagulopathy and development of disseminated intravascular coagulation (DIC). PAI-1 levels correlate with the severity of MODS in sepsis and DIC [27]. PAI-1 gene is expressed in liver, endothelial cells, macrophages, adipose tissue, heart and kidney [28]. Increase in plasma PAI-1 after LPS-stimulation may be a combined effect of both PAI-1 release from activated platelets and its synthesis associated with PAI-1 gene expression on hepatocytes [29]. Plasma PAI-1 levels are also strongly related to liver steatosis, what supports the thesis of liver being an important source of circulating PAI-1 [30].

STATISTICAL ANALYSIS
Continuous data are presented as median and lower and upper quartiles for non-normally distributed variables or as mean and standard deviation for normally distributed variables. Statistical differences between groups were calculated by the non-parametric Mann-Whitney U test. Statistical signi cance between frequencies was calculated by the chi-square test. Relation between two parameters was assessed using correlation analysis and Spearman correlation coe cients were calculated. Survival curves were obtained with the Kaplan-Meier method and were compared using the log rank test. Receiver operating characteristic (ROC) curve analysis was performed to calculate the area under the receiver operating characteristic curve. The best cut-off values were calculated to maximize the sum of sensitivity and speci city. Positive predictive value (PPV), negative predictive value (NPV) and accuracy were also calculated. P value of less than 0,05 was required to reject the null hypothesis. Statistical analysis was performed using EPIINFO Ver. 7.2.3.1 software package.

Results
During the study period 1660 patients were admitted to the ICU. Sepsis /septic shock was diagnosed in 480 patients (28,9%). According to the study criteria 401 patients were excluded and further 79 were enrolled into the analysis. Sepsis was diagnosed in 32 (40,5%) and septic shock in 47 (59,5%) patients.
Patients were severely ill with mean APACHE II score of 24. The DIC score according to the ISTH Criteria were calculated. None of the patients in the study group developed DIC at any time of observation.
Out of the 79 patients, 24 (30,4%) met the criteria of SALD. None of the patients developing SALD required extracorporeal liver support. Median bilirubin in the SALD group was 2,25(1,1 ÷ 4,16)mg/dl; 3,3(2,1 ÷ 5,67)mg/dl; 5,65(2,25 ÷ 6,95)mg/dl; 5,94(1,9 ÷ 7,6)mg/dl; 4,4(2,2 ÷ 5,6)mg/dl at days 1, 3, 5, 7 and 14 respectively. At the time of enrolment into the study, there was no difference in the severity of the disease between the groups. The characteristics of the patients is shown in table no.1. HTA -hypertension arterial, DM -diabetes mellitus, CCD -chronic cardiac disease, CA -cardiac arrest, OBS -obesity, CKD -chronic kidney disease, CPD -chronic pulmonary disease, ARYarrhythmia, ALC -alcoholic disease, NEUR -chronic neurological disease, OTH -other There was no statistically signi cant difference in comorbidities between the groups. None of the patients in our study group had metabolic syndrome. Neither alcohol abuse nor obesity had statistically signi cant impact on the development of SALD. Apart from bloodstream infections (BSI), there was no difference in the source of sepsis between the groups. There was no case of BSI in bilirubin < 2 mg/dl group. We evaluated the association between plasma biomarkers measured within rst 24 h after enrolment into the study. Analysed biomarkers were also investigated as predictors for 28-day survival. In Table 2. shown are the plasma biomarkers levels at the time of enrolment into the study (day 1). Only AST was signi cantly higher in SALD group, comparing to no-SALD group (p = 0,0234). Taking into consideration no difference in routinely measured biomarkers between the groups, we decided to analyse novel biomarkers at baseline.  At baseline we observed statistically signi cant difference between the groups only for PAI-1. In the group with no SALD the levels of PAI-1 were signi cantly lower than in patients with SALD. Subsequently a ROC curve analysis for PAI-1 was conducted (Fig. 1).
As there was no statistically signi cant difference between the groups in severity of the disease (APACHE II score 24 vs. 24,5, p = 0,979), so we didn't analyze 28-day survival adjusting it to the APACHE II score at the time of enrolment.
The Spearman correlation was also conducted between IP-10, ET-1, PAI-1, hepcidin and TAT on day 1 and APACHE II and SOFA scores. There was statistically signi cant positive correlation between PAI-1 and SOFA (R = 0,26, p = 0,0198), but no correlation between hepcidin, IP-10, ET-1, TAT and SOFA or APACHE II scores, as well as no correlation between PAI-1 and APACHE II. there is a statistically signi cant difference in PAI-1 (p = 0,00432) and TAT (p = 0,0420) values between the groups, but no difference in IP-10 (p = 0,0795), endothelin-1(p = 0,0761) and hepcidin (p = 0,0558). The statistically signi cant difference remains up to day 7. for PAI-1 and TAT with p = 0,00942 and p = 0,0414 respectively. There is no difference between the groups in median values of biomarkers on day 14.

Discussion
The results of our study show that almost one third of septic patients had an acute elevation of bilirubin up to 2 mg/dl or more. The results are subsequent with available reports [1]. Two third of the enrolled patients ful lled our SALD criteria within 24 hours from diagnosing sepsis/ septic shock. After analysing our study group demographics, we demonstrate that, apart from BSI, neither cause of sepsis, nor comorbidities were contributing to SALD development. Metabolic syndrome, obesity and alcohol intake are known risk factors increasing the probability acute liver damage [31]. None of the patients in our study group had metabolic syndrome, neither alcohol abuse nor obesity in uenced the development of SALD.
Interestingly, there was no difference in severity of the disease (APACHE II score) or the degree of the organ dysfunction (SOFA score) at the time of inclusion in patients who developed SALD and those who did not. This suggests that hyperbilirubinemia may just be a signal of worsening liver function, which contributes to adverse outcomes via independent pathways (not included in APACHE II scoring) or that serum bilirubin cut-off value of 2 mg/dl is too low to identify patients with adverse outcome. This is re ected in the results of our study. Mild elevations in bilirubin levels, consequently the degree of organ dysfunction, wasn't severe enough to cause a difference in mortality between the groups. What is more in sepsis/septic shock there is a wide range of other causes in uencing mortality, than hyperbilirubinemia itself. This is also con rmed by the fact, that there was no difference in APACHE II scoring between our study groups.
Bilirubin as single biomarker is a poor factor in distinguishing newly developing organ dysfunction from preexisting one. Due to its late increase and low speci city, nding a single cause of hyperbilirubinemia in ICU patients remains a challenge. Contrary to Patel et al. [32], in our study, bilirubin was shown to be a poor predictor of survival in septic patients.
Our study was the rst to evaluate the correlation between PAI-1 and liver function in septic patients. Out of ve analysed biomarkers, only PAI-1 at the time of enrolment could be useful to predict the development of SALD. In various studies liver has been proved to be an important source of circulating PAI-1 [28,30]. PAI-1 can be a better marker than bilirubin, predicting organ dysfunction, contrary to bilirubin, as correlation between PAI-1 and SOFA score was shown. The study revealed that cut-off value for PAI-1 of 9 ng/ml was subsequent with an acute increase in bilirubin levels of 2 mg/dl or more. Interestingly this cut-off value holds within normal limits (0,99 − 16,9 ng/ml). In no-SALD group there were also individuals reaching higher PAI-1 levels than the calculated cut-off value of 9 ng/ml. PAI-1 was analysed as a predictive marker, meaning that not all the subjects above cut-off value develop liver dysfunction. There must be other risk factors in uencing the development of SALD, what requires further investigation. What is more our results have shown low speci city of PAI-1 with high negative predictive value. Low speci city of PAI-1 may be a result of many origins and different factors inducing its synthesis. Increase in plasma PAI-1 after LPS-stimulation may be a combined effect of both PAI-1 release from activated platelets and its synthesis associated with PAI-1 gene expression on hepatocytes [29]. The results obtained in our study group may indicate that the activated platelets could play a greater role in an increase in PAI-1 levels, than its hepatic source does.
Many authors have investigated PAI-1 as a biomarker of sepsis and DIC. Koyama et al. [33] investigated PAI-1 combined with TAT and protein C as predictors for the development of overt DIC. In our cohort non of the patients developed DIC, so it was impossible to confront these results with ours. The incidence of DIC in septic patients, diagnosed with ISTH criteria, is reported to be approximately 29% [34]. The authors think that the cause of lack of overt DIC in the study group was a result of used inclusion criteria resulting in choosing only 79 out of 480 (28,9%) septic patients treated in our ICU. In terms of survival, Koyama's [33] results were compatible with ours, con rming PAI-1to be a good predictor of survival in septic patients. Higher PAI-1 levels were related to higher SOFA scoring at the onset of sepsis, but there was no re ection in APACHE II scoring.
We also evaluated PAI-1 in combination with AST as a predictor of 28-day survival. AST and ALT are the markers of hepatocellular integrity. However AST is a sensitive marker of liver cell injury, its speci city remains questionable. Apart from liver it can be found in cardiac muscle, skeletal muscle, kidneys, brain, pancreas, lungs, leukocyte and erythrocytes. AST levels usually peak before ALT, especially in ischemic or toxic liver injury. Acinar zone 3 is vulnerable to hypoxic damage and the enzyme is mostly distributed in the peripheral site of the acinus [35]. It was con rmed in our study, as the levels of AST at day 1 were signi cantly higher in SALD group. This may also indicate an ischemic damage as one of the causes of SALD. Low cut-off AST values presented in our study are again a proof that liver dysfunction in our study group wasn't severe, as well as the cut-off bilirubin of 2 mg/dl was too low of a value to diagnose SALD with a potential adverse outcome.
Our analysis revealed that a combination of PAI-1 and AST predicts better the 28-day survival, than PAI-1 alone, but due to low cut-off values of AST it might not be clinically signi cant.
The analysis of further days of the observation showed, that PAI−1 remained higher during the rst week of observation in the group of patients who developed SALD. Higher PAI−1 values observed on day 1 were related to a greater increase in bilirubin levels on days 3, 5 and 7, what con rms a potential usefulness of the marker as a predictor of liver dysfunction.
Lower hepcidin values observed on day 1 were related to a greater increase in bilirubin levels on days 5, 7 and 14. As predictors of SALD, ET−1 and hepcidin on day 3 and TAT on day 5, could be taken into consideration as their difference between SALD and no-SALD group remained signi cant, but our study group was to small to conduct further analysis.
The limitations of the study were: it was a single-centre study and the number of patients enrolled into the study was relatively small; although our results (SALD incidence) matched that demonstrated in previous studies [1], lack of generally accepted SALD de nition remained a challenge.
PAI-1 may be a good marker prognosing SALD, but due to its low speci city and high negative predictive value, it has to be analysed in combination with other markers. AST levels at the time of admission might be a good choice. Further studies are needed nd a unique biomarker of SALD, but rstly to broaden knowledge on its patophysiology and to nd an unanimous de nition and de ne clear diagnostic criteria.

Page 14/19
Sepsis-associated hyperbilirubinemia is frequent, but bilirubin is late and not speci c marker of SALD. Measuring PAI-1 serum levels at the onset of sepsis/septic shock may be useful in predicting the development of SALD. A combination of PAI-1 and AST predicts better the 28-day survival, than PAI-1 alone, but due to low cut-off values of AST it might not be clinically signi cant. Very subtle, but still signi cant differences and low cut off values of the markers prove the authors' hypothesis that a cut-off value of bilirubin > 2 mg/dl is too low, and higher cut-off values of bilirubin or combination of bilirubin with other markers of liver injury, might be needed to distinguish a group of patients with SALD leading to an adverse outcome. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.