Endothelial dysfunction is an early indicator of sepsis and neutrophil degranulation of septic shock in surgical patients.

BACKGROUND
Stratification of the severity of infection is currently based on the Sequential Organ Failure Assessment (SOFA) score, which is difficult to calculate outside the ICU. Biomarkers could help to stratify the severity of infection in surgical patients.


METHODS
Levels of ten biomarkers indicating endothelial dysfunction, 22 indicating emergency granulopoiesis, and six denoting neutrophil degranulation were compared in three groups of patients in the first 12 h after diagnosis at three Spanish hospitals.


RESULTS
There were 100 patients with infection, 95 with sepsis and 57 with septic shock. Seven biomarkers indicating endothelial dysfunction (mid-regional proadrenomedullin (MR-ProADM), syndecan 1, thrombomodulin, angiopoietin 2, endothelial cell-specific molecule 1, vascular cell adhesion molecule 1 and E-selectin) had stronger associations with sepsis than infection alone. MR-ProADM had the highest odds ratio (OR) in multivariable analysis (OR 11·53, 95 per cent c.i. 4·15 to 32·08; P = 0·006) and the best area under the curve (AUC) for detecting sepsis (0·86, 95 per cent c.i. 0·80 to 0·91; P < 0·001). In a comparison of sepsis with septic shock, two biomarkers of neutrophil degranulation, proteinase 3 (OR 8·09, 1·34 to 48·91; P = 0·028) and lipocalin 2 (OR 6·62, 2·47 to 17·77; P = 0·002), had the strongest association with septic shock, but lipocalin 2 exhibited the highest AUC (0·81, 0·73 to 0·90; P < 0·001).


CONCLUSION
MR-ProADM and lipocalin 2 could be alternatives to the SOFA score in the detection of sepsis and septic shock respectively in surgical patients with infection.


Introduction
Sepsis and septic shock are major causes of morbidity and mortality in surgical patients 1 . In a patient with infection, prompt detection of sepsis is key to the initiation of early treatment with appropriate antimicrobials, elimination of the infectious source, administration of fluids and appropriate transfer to the ICU. In patients with sepsis, prompt detection of septic shock could imply a need to modify antibiotic treatment, seek alternative sources of 221 500 (186 250-299 250) 184 000 (105 250-276 000) 123 000 (88 500-258 000)  potentially infectious organisms not already identified, and adjust ICU support. Since publication of the Third International Consensus Definitions for Sepsis and Septic Shock (SEPSIS-3) in 2016 2 , severity stratification in patients with infection has been based on the Sequential Organ Failure Assessment (SOFA) score 3 . The problem with this score is that it is difficult to calculate in non-ICU settings, such as surgical departments or the emergency room. The alternative proposed by the SEPSIS-3 consensus for these settings, the quickSOFA (composed of three simple items: respiratory frequency, BP and the Glasgow Coma Scale score), is very specific but less useful for detecting sepsis 4 .  Benjamini-Hochberg P Mid-regional proadrenomedullin (nmol/l) ED 11⋅53 (4⋅15, 32⋅08) < 0⋅001 0⋅006 Biomarkers could contribute to stratification of the severity of infection. Sepsis is characterized by acute endothelial dysfunction, which increases vascular permeability, promotes activation of the coagulation cascade and tissue oedema, and compromises the perfusion of vital organs 5 . Biomarkers of endothelial responses can be used to categorize patients into homogeneous subgroups with different severity 6 . In turn, sepsis activates emergency granulopoiesis, inducing release of immature neutrophil precursor cells in the peripheral blood, an event related directly to severity 7 -10 . Emergency granulopoiesis can be detected by profiling the mRNA in blood of the genes that are expressed sequentially in the neutrophil precursors 11,12 . Other molecules denoting severity during an infection are proteins released to the plasma during neutrophil degranulation 13,14 . These include matrix metalloproteinase (MMP) 8, neutrophil gelatinase-associated lipocalin and lactotransferrin, which have been shown to be closely related to the development of sepsis 15 , and levels of plasma MMPs 3, 7, 8 and 9 are increased in severe sepsis on admission to the ICU 16 .
In this study, 38 biomarkers of endothelial dysfunction, emergency granulopoiesis or neutrophil degranulation were evaluated to stratify severity in surgical patients with infection. The hypothesis was that these biomarkers might differentiate between three groups of patients: those with infection, those with sepsis, and those with septic shock.

Methods
Surgical patients with infection, sepsis or septic shock were recruited prospectively from the surgery departments and surgical ICUs of the three participating hospitals (Hospital Clínico Universitario de Valladolid, Hospital Universitario Río Hortega de Valladolid and Hospital Clínico

Ethical approval
The study was approved by the respective Committees for Ethics in Clinical Research of the three participating hospitals. Methods were carried out in accordance with current Spanish law for Biomedical Research, fulfilling the standards indicated by the Declaration of Helsinki. Written informed consent was obtained from patients' relatives or their legal representative before enrolment.

Microbiology
Standard cultures in biological samples, guided by the presumptive source of the infection, were performed to assess the presence of the causal pathogen. Potentially contaminant microorganisms were not considered.

Biomarker profiling
Thirty-eight biomarkers (10 denoting endothelial dysfunction, 22 indicating emergency granulopoiesis and 6 denoting neutrophil degranulation) were profiled in the three patient groups (infection, sepsis or septic shock) in the first 12 h after diagnosis (Tables S1 and S2, supporting information). The methods used to profile these biomarkers are detailed in Appendix S1 (supporting information). Blood from healthy individuals was collected as part of their blood donation.

Statistical analysis
Statistical analysis was performed with IBM SPSS ® version 20 (IBM, Armonk, New York, USA). Box plots were represented using Minitab ® 19.2 (Minitab, Coventry, UK). For demographic and clinical characteristics of the patients, differences between groups were assessed using the χ 2 test for categorical variables. Differences between groups for continuous variables were assessed with the Kruskal-Wallis test, with post hoc tests adjusting for multiple comparisons.
In the comparison of infection and sepsis, multivariable logistic regression analysis was employed to evaluate the association between biomarker levels and the presence of sepsis. In the comparison of sepsis and septic shock, the same type of analysis was employed to evaluate the association between biomarker levels and the presence of septic shock. Only biomarkers yielding P ≤ 0⋅050 in univariable analysis were tested in multivariable analyses. Potential confounding clinical factors that yielded P ≤ 0⋅100 in univariable analysis were introduced as adjusting variables in multivariable analyses, followed by multiple testing correction by the false discovery rate using the Benjamini-Hochberg procedure. The optimal operating point in the area under the curve (AUC) analysis was identified as described previously 19 .

Results
There was a total of 100 patients with infection, 95 with sepsis and 57 with septic shock. Patients with infection were significantly younger than those in the other groups ( Table 1), and the healthy controls. Proportions of men to women were similar in all patient groups and control subjects. Patients with sepsis and septic shock had more antecedent cardiovascular, respiratory or renal disease. The proportion of patients needing urgent surgery was similar in the three groups. Abdominal surgery was the most frequent type, and the abdomen was the predominant source of infection in all three patient groups. Respiratory infection was more common in patients with sepsis or septic shock than in patients with infection alone.
The prevalence of bacteraemia was highest in patients with septic shock, where Gram-negative bacteria dominated ( Table 1).
Patients with septic shock showed the highest degree of organ failure as assessed by the SOFA score. Duration of hospital stay was directly associated with severity. No patient in the infection group died in hospital, compared with seven of 95 (7 per cent) patients with sepsis and 22 of 57 (39 per cent) with septic shock ( Table 1).
Coagulopathy (as assessed by the international normalized ratio) and decreased lymphocyte and monocyte counts were related to increasing severity. Biomarker levels showed a generalized trend to increase with disease severity (Figs 1 and 2; Table S3, supporting information).
Confounding factors from Table 1 that yielded P ≤ 0⋅100 in univariable analysis, to be introduced as adjusting variables in multivariable analyses, are shown in Table S4 (supporting information).
Multivariable analysis of biomarker levels to evaluate the risk of sepsis versus infection identified seven biomarkers of endothelial dysfunction, two of neutrophil degranulation and 13 of emergency granulopoiesis as independent risk factors for sepsis ( Table 2).
Multivariable analysis to evaluate the risk of septic shock versus sepsis revealed four biomarkers of endothelial dysfunction, six of neutrophil degranulation and 14 of emergency granulopoiesis as independent risk factors for septic shock ( Table 3).
The AUC analysis to assess biomarker sensitivity and specificity indicated that mid-regional proadrenomedullin (MR-ProADM) was the best biomarker for differentiating sepsis from infection, whereas lipocalin 2 in plasma was the best biomarker for distinguishing septic shock from sepsis (Fig. 3).

Discussion
This study found that a panel of seven biomarkers related to endothelial dysfunction (MR-ProADM, syndecan (SDC) 1, thrombomodulin (THBD), angiopoietin (ANGPT) 2, endothelial cell-specific molecule 1, vascular cell adhesion molecule 1 and E-selectin) were associated with the presence of sepsis in patients with infection. This suggests that induction of endothelial injury is an early event as organ dysfunction develops. SDC1 and MR-proADM were the biomarkers showing the highest odds ratios for sepsis. SDC1 is a glycosaminoglycan shed from the endothelial glycocalyx during sepsis, and levels in plasma correlate with the SOFA score 20,21 . In the present study, MR-proADM was the biomarker of endothelial dysfunction showing not only the strongest association but also the best balance between sensitivity and specificity for sepsis, with an AUC of 0⋅86. Adrenomedullin is secreted from various organs and tissues, including vascular endothelial cells. It regulates vascular tone and endothelial permeability 22 . MR-proADM, the mid-regional fragment of proadrenomedullin, is more stable and directly reflects levels of the rapidly degraded active adrenomedullin peptide 23 . There is growing evidence of the value of MR-ProADM as an early marker of severity in patients with infection 24 and as a predictor of organ failure in patients with community-acquired pneumonia 25 .
In the comparison of sepsis and septic shock, the number of biomarkers of endothelial dysfunction independently associated with septic shock dropped to four (SDC1, MR-ProADM, THBD and ANGPT2). In contrast, six biomarkers denoting neutrophil degranulation were associated with septic shock: proteinase 3 (a serine protease), lipocalin 2 (a neutrophil gelatinase-associated protein), interleukin-18 receptor type 1 (an inductor of neutrophil degranulation) 26 , matrix metalloproteinase (MMP) 8 (a neutrophil collagenase), lactoferrin (a major iron-binding protein) and myeloperoxidase (a heme protein). Only two of these biomarkers seemed relevant to differentiate sepsis from plain infection (lipocalin 2 and MMP8), suggesting that neutrophil degranulation may be important in the pathogenesis of septic shock. Proteins released from neutrophil granules could be mediating antibacterial effects 8,27 -30 , and may participate in tissue remodelling 31 , attenuation of inflammation 32 and preventing the deleterious effects of neutrophil extracellular traps 33 . However, increased intravascular levels of degranulated proteins could induce enhanced proteolysis 34 , endothelial injury and organ failure 35 -39 . Proteinase 3 and lipocalin 2 had strongest associations with the presence of septic shock. Neutrophil degranulation can lead to increased endothelial permeability via a mechanism that, in part, involves the actions of proteinase 3 40 , and a multimarker model containing proteinase 3 was able to predict the risk of septic acute kidney injury in patients with septic shock 41 . In the present study, lipocalin 2 was the marker showing the best balance between sensitivity and specificity in detecting septic shock. Lipocalin 2 has been used for risk stratification, early diagnosis and prognostication of sepsis in the emergency department 42,43 . This protein is associated with mortality and multiple organ dysfunction syndrome in severe sepsis and septic shock 44 . Lipocalin 2 has been promoted as a relatively robust predictor of 28-day mortality in severe sepsis 45 .
The present study has shown that emergency granulopoiesis is a preserved signature of both sepsis and septic shock, although to a greater degree in septic shock. The observed parallel between emergency granulopoiesis signatures and severity is in agreement with a previous study 9 demonstrating that, in sepsis, the increased presence of circulating immature granulocytes is linked to clinical deterioration.
Regarding acute-phase biomarkers, procalcitonin showed modest associations with the risk of sepsis and septic shock, while C-reactive protein showed a mild association, exclusively with the risk of sepsis. These results indicate that neither procalcitonin nor C-reactive protein is suitable for severity stratification in patients with infection.
Profiling protein levels in plasma of MR-ProADM and lipocalin 2 could contribute to stratification of the severity of infection, particularly in settings where calculation of the SOFA score is not feasible. Evaluation of protein biomarkers is technically easier than evaluating those of transcriptomic nature. Emerging point-of-care devices could result in evaluation of these biomarkers in clinical practice as results can be obtained in less than 1 h 46 .
This study has an important limitation in that biomarkers were compared only at diagnosis of infection, sepsis or septic shock. Further prospective follow-up studies with serial sampling should validate the potential role of MR-ProADM and lipocalin 2 in predicting clinical worsening of patients with infection or sepsis.