Cytokine biomarker phenotype for early prediction and triage of sepsis in blunt trauma patients

Sepsis is a life-threatening condition associated with an exacerbated production of cytokines that can promote a hyperactive response to infection or induce immunoparalysis. The trauma victim’s inherent state of hyperinammation frequently camouages septic events, delaying the initiation of targeted therapy. Thus, this study aimed to establish the proles of cytokines in trauma patients to characterize the nature of the immune responses to sepsis, which might enable early prediction and individualized treatments to be developed and targeted. A 15-plex human cytokine magnetic bead assay system was used to measure analytes in citrated plasma samples. Analysis of these cytokine kinetics was performed on 40 patients with severe blunt trauma admitted to our trauma center between March 2016 and February 2017 with Injury Severity Score (ISS) more than 20 with regard to sepsis (Sepsis-3) over a 14-day time course. stay of severe trauma victims is usually prolonged. Therefore, detailed monitoring and documentation of clinical and laboratory parameters are valuable prerequisites to retrospectively identify rapid changes in the conditions of patients. (D) Trauma victims are thought to be more susceptible to infectious complications than any other patient population, allowing for statistically balanced juxtaposition of septic versus nonseptic patients. (E) Trauma represents a rather homogenous entity with directly measurable trauma severity (i.e., ISS and AIS), which can simply be adjusted for statistical analyses.


Introduction
Sepsis is a major public health crisis worldwide. Despite the huge advances in treatment technology, over 5.3 million patients die annually from sepsis, at an estimated overall mortality of 30% [1]. Sepsis, de ned as "life-threatening organ dysfunction caused by a dysregulated host response to infection according to the Third International Consensus De nitions for Sepsis and Septic Shock (Sepsis-3)", is counted among the leading causes of major trauma-related mortality worldwide [2]. But, it has proven di cult to develop clinical and laboratory criteria that accurately predict risk for mortality, particularly in the early period of sepsis. Timely detection and treatment of emerging infections can allow us to take measures for treatment as early as possible, thereby reducing post-trauma mortality and improving patient outcome [3,4]. Hence, in the setting of imperfect diagnostic and prognostic tools for sepsis, novel methods for effective identi cation of severe trauma patients who have the potential to develop life-threatening infectious complications and are at the risk of death is still an urgent problem.
Severe trauma could induce exacerbation of systemic in ammation, which often progresses to sepsis leading to a lethal outcome. In ammatory imbalance represents the most critical basis of sepsis pathogenesis and occurs throughout the whole process of sepsis, and the pathogens eliciting the response include organisms such as bacteria, fungi, parasites, and viruses [5]. The host's initial acute response to invasive pathogens typically causes macrophages to engulf the pathogens and produce a range of pro-and anti-in ammatory cytokines, and this can trigger cytokine storms and activate the immune system [6]. Numerous evidence [7] has been sought to identify effective biomarkers to predict which patients are at high-risk of morbidity and mortality. Many biomarkers [8,9], such as C-reaction protein (CRP), procalcitonin (PCT), tumor necrosis factor (TNF)-α, interleukin-6 (IL-6), and IL-8, which are involved in systemic in ammation caused by trauma and relate to the severity of injury, have been assessed as potential markers to predict poor prognosis in critical ill patients. Extensive data are also available for sepsis-related changes in macrophage in ammatory protein (MIP)-1α/β, heat shock protein (HSP) 70, matrix metalloprotease (MMP)-8, and granzyme B in animal or adult models [7]. Unfortunately, all of these biomarkers lack su cient speci city. In addition, various rule-based disease-severity scoring systems are also widely used in an attempt to identify sepsis patients. These scores, such as the Systemic In ammatory Response Syndrome (SIRS) criteria [10], the Modi ed Early Warning Score (MEWS) [11], and the Sequential Organ Failure Assessment (SOFA) [12], are manually tabulated at the bedside and lack accuracy in sepsis diagnosis.
Therefore, improved biomarkers are required for the prompt diagnosis of sepsis and prediction of outcomes. However, previous reports have often evaluated individual biomarkers. In this study, in ammatory biomarkers, including IL-6, IL-8, IL-10, IL-15, TNF-α, macrophage derived chemokine (MDC), interferon-inducible protein (IP)-10, eotaxin, soluble CD40 ligand (sCD40L), growth regulated oncogene (GRO), monocyte chemotactic protein (MCP)-1, MIP-1β, broblast growth factor (FGF)-2, granulocyte colony-stimulating factor (G-CSF) and granulocyte macrophage colony-stimulating factor (GM-CSF) were analyzed at ve timepoints (1, 3, 5, 7, and 14 days after trauma) in forty severe trauma patients. The objective of the present study was to investigate the time course of these cytokine levels in patients with severe trauma and to determine whether cytokines can serve as a biomarker to predict the development of sepsis in trauma patients.

Study population
This study was conducted with data obtained at the Department of Trauma Surgery in Daping Hospital and the Chongqing Emergency Medical Center from March 2016 to February 2017. The inclusion criteria were the following: (1) the age of patients is between 18 and 65 years, (2) ISS is greater than 20, and (3) patients survive for longer than 48 hours after injury. Patients were excluded if they had penetrating injuries, preexisting organ dysfunction, or immune diseases. This study was approved by the Ethical and Protocol Review Committees of Army Medical University and followed the guidelines of the Declaration of Helsinki. Informed consent was obtained from the patients and their next of kin before enrollment, including explicit permission for plasma analysis and the collection of relevant clinical data.

Clinical evaluation
Demographic characteristics and laboratory and clinical data were collected by expert physicians. The ISS was calculated according to the Abbreviated Injury Scale developed in 2005 [13]. Sepsis was diagnosed based on a rapid change in SOFA score ≥ 2, with evidence of infection during the hospital period according to "Sepsis-3" [2]. Infection [14] was de ned as a clinically obvious source or positive bacterial culture by medical examinations as follows: microbiological tests, including culture of body uids; conventional or real-time polymerase chain reaction; radiological analyses, including ultrasonography, X-ray, and computed tomography; and serology. SOFA scores [12] were calculated daily as the sum of the simultaneously obtained individual organ scores, including respiratory, cardiovascular, hepatic, renal, coagulation, and neurologic systems.

Multiplex cytokine detection
After informed consent, venous blood samples were drawn on days 1, 3, 5, 7, and 14 during the rst 14 days after trauma, encompassing >90% of the outcome events. Plasma was separated, aliquoted, and stored at -80°C until analyzed. The multiplex cytokine kits (IL-6, IL-8, IL-10, IL-15, TNF-α, MDC, IP-10, eotaxin, sCD40L, GRO, MCP-1, MIP-1β, FGF-2, G-CSF, and GM-CSF) were obtained, and the assay was performed on the Luminex200 system in accordance with the manufacturer's instructions (Millipore, Billerica, MA, USA). The detectable range of each cytokine was 0.89 to 3500 pg/mL. All measurements were made by the same person who was blinded to the clinical data.

Statistical analysis
Discrete data are expressed as counts with percentages, whereas continuous data are presented as the median with interquartile range (IQR) or the mean ± standard deviation (SD) as appropriate. Baseline characteristics were compared between groups using the Chi-square test for counts and univariate analysis of variance for continuous data. Biomarker time courses were compared between groups using Wilcoxon's rank sum test. Receiver operating characteristic (ROC) curve analysis was employed at each timepoint to evaluate the predictive performance of the biomarkers tested with regard to traumatic sepsis.
Values of the Area Under Curve (AUC) are reported with the corresponding 95% con dence interval (CI).
An AUC greater than 0.7 was considered good. All tests were 2 tailed; P < 0.05 was considered signi cant.

Clinical characteristics of trauma patients
Baseline characteristics of the study population in total and according to the main outcome (no sepsis vs. sepsis vs. septic shock) are given in Table 1. Forty severely injured patients (9 female) with ISS≥20 were included. The median periods from injury to admission to the hospital were 5.5 hours (4.0-17.3 hours). We pooled the septic and septic shock group for further analyses due to the low number of patients with septic shock (n=7

Dynamic changes in in ammatory biomarkers in trauma patients
We rst characterized the kinetic changes in in ammatory biomarkers in the plasma of trauma patients. As shown in Figure 1 and Table S1, the plasma levels of six cytokines, including IL-6, IL-10, IP-10, sCD40L, GRO, and G-CSF, were signi cantly changed among the different timepoints (days 1, 3, 5, 7, and 14) in all trauma patients. Among them, IP-10, sCD40L, and GRO were signi cantly elevated with the progression of trauma, and the other three cytokines were signi cantly decreased during the disease course. For example, there was a signi cantly higher level of IP-10 on day 14 compared with day 1 (p=0.005) and day 3 (p=0.009). The levels of IL-8, IL-15, TNF-α, MCP-1, MDC, eotaxin, MIP-1β, GM-CSF, and FGF-2 showed no signi cant changes among the different timepoints after trauma.

Time course of cytokine biomarkers as related to sepsis
To delineate sterile in ammation from sepsis, we evaluated the association between the levels of cytokines and traumatic sepsis using Wilcoxon's rank sum test analysis ( Figure 2 and Table S2). Other cytokine levels showed no signi cant difference between the two groups in the rst 14 days.
To further corroborate the predictive ability of the eight cytokines with regard to the status of sepsis, we employed univariate binary logistic regression with ROC curve analysis. Eight cytokines were able to predict septic course at day 3 after trauma with an AUC ranging between 0.69 and 0.85 (P<0.05).
Combination of the eight cytokines demonstrated even higher AUCs of up to 0.90 ( Figure 3 and Table S3). At day 3 after trauma, a remarkable increase (by 5%) was demonstrated when combining the eight cytokines. Giving equal weight to both sensitivity and speci city, we selected -1.19 for the combined eight markers as a cut-off value for being able to predict yet clinically unapparent septic course at day 3 after trauma (sensitivity/speci city: 83%/91%).

Event-related analysis of in ammatory biomarkers
To further elucidate the interaction between time and disease status, we rearranged the individual biomarker courses according to their timepoint of occurrence and used the grand median for patients with an uneventful course as a reference line. This event-related analysis (Figure 4) demonstrated that serum G-CSF levels had a 1.5-fold increase within 72 hours and a 3-fold increase within 48 hours before clinically apparent sepsis with subsequent initiation of therapy. In contrast, IL-10 levels appeared 4-fold higher within 72 hours and 1.5-fold higher before sepsis was clinically overt. IL-6 serum levels showed a circa 1.5-fold increase in patients with septic complications within 72 hours before being clinically overt.
In the same time frame, FGF-2 decreased 2-fold in patients with sepsis. Eventually, serum levels of MDC, sCD40L, GRO, and IL-15 demonstrated unspeci c undulations within 72 hours before diagnosis of the corresponding event.

Discussion
This study investigated the time course of 15 cytokines, including both pro-and anti-in ammatory cytokines and chemokines, in a cohort of 40 patients who experienced severe trauma with regard to the incidence of septic complications within the initial 14 days after trauma. As expected, the levels of ve cytokines, including IL-6, IL-10, sCD40L, GRO, and G-CSF, were signi cantly interactional between time and the presence of sepsis, which was indicated by a signi cantly steeper change in plasma levels in patients with sepsis as opposed to those without sepsis. This interaction suggests that, besides a cut-off value, the time-related kinetics of cytokine biomarkers has a crucial role in the preclinical identi cation of septic patients.
When combining these cytokines, the ROC curve analysis achieved even higher AUCs of up to 0.90 at day 3. On the one hand, the increase in accuracy was favorable; on the other hand, day 3 after trauma demonstrated a remarkable increase (by 5%) when combining these biomarkers. These data suggest that combining multiple cytokines might be advantageous, especially in the early phase after trauma. However, it is to be noted that standard ROC curve analysis represents a theoretical model, which does not account for time dependency of an event. This limitation of considering an event (i.e., sepsis) status for an individual as xed over time is often encountered in biomarker studies and demands cautious data interpretation. Therefore, the event-related analysis as shown in Fig. 4 compensates for that and better re ects the predictive strength of a biomarker. Against this background, IL-6, IL-10, G-CSF, and FGF-2 have excellent predictive ability for septic progression.
The ultimate goal in sepsis biomarker research, however, is the development of a score that is useful for sepsis prediction, diagnosis, and risk strati cation. With the availability of systems biology approaches, such a sepsis score may ideally comprise a combination of several biomarkers, including genomic, transcriptomic, proteomic, and metabolomic candidates, as well as clinical parameters. Examples of preliminary scores in different sepsis cohorts include the combination of clinical and transcriptomic markers, metabolic and protein-mediator pro ling, or the combination of biomarkers with established clinical scores [7,15,16]. In our opinion, some cytokines are also very promising candidates to be included in a sepsis score in general and in particular for discrimination of septic complications in patients with underlying systemic in ammation as indicated by the results of the present study. Nevertheless, based on the experience of previous studies [17], the development of a multibiomarker score for predicting and diagnosing sepsis represents an intricate endeavor. The implementation of novel elds, such as bioinformatics, computational biology, and machine learning, into sepsis research might open up new possibilities to adequately re ect the complexity and dynamics of sepsis for development of a valid and reliable diagnostic tool in the future.
As early decision making with the initiation of therapy is key to survival, especially in trauma, in ammatory biomarkers are urgently needed, aiming at the preclinical detection of deterioration [18]. In contrast, this patient population has some major methodological advantages in regard to testing novel biomarkers compared to other critically ill patients, making up for some key limitations of previous studies as follows: (A) Prophylactic administration of antibiotics, which is generally considered a major incommoding confounder in sepsis-related biomarker research, is strongly avoided in patients with blunt trauma injuries due to resistance breeding. (B) The trauma victim is mostly young to middle-aged with no or few comorbidities, thus greatly diminishing confounding effects. (C) The ICU stay of severe trauma victims is usually prolonged. Therefore, detailed monitoring and documentation of clinical and laboratory parameters are valuable prerequisites to retrospectively identify rapid changes in the conditions of patients. (D) Trauma victims are thought to be more susceptible to infectious complications than any other patient population, allowing for statistically balanced juxtaposition of septic versus nonseptic patients. (E) Trauma represents a rather homogenous entity with directly measurable trauma severity (i.e., ISS and AIS), which can simply be adjusted for statistical analyses.
Elucidating the role of cytokines in severe trauma patients with respect to the occurrence of sepsis, the present study represents a crucial step on the way to testing for potentially helpful biomarkers for early sepsis prediction and detection. In this context, the present study needs to be considered as hypothesis generating and awaiting con rmation in future clinical trials, since there are some limitations. Given the relatively small number of patients included and the single center design of our study, there is no external validation of our data, which has to be addressed in further studies. Likewise, measurement of biomarkers and assessment of clinical parameters were performed only once per 48 hours, neglecting potential alterations between these intervals. Moreover, the in uence of the type of infection and of repetitive surgical tissue damage on cytokine serum levels need to be clari ed in further studies. We believe that these ndings will have implications for the management of patients with trauma.

Conclusions
Cytokine pro les demonstrate high discriminatory ability to timely identify evolving sepsis in patients with severe blunt trauma. These abrupt changes allow sepsis detection up to 72 hours before clinically overt deterioration. Affordable cytokine bedside measurement might enable earlier treatment of sepsis with reduced morbidity and mortality and subsequently reduce costs for trauma patients in the future.

Declarations
Ethics approval and consent to participate The study protocol was approved by the Ethical and Protocol Review Committee of the Army Medical University (No. TMMU2012009). Informed consent was obtained from the patients or their next of kin.

Consent for publication
Not applicable.

Con icts of interest
All authors declare that they have no competing interests. All authors read and approved the nal manuscript.

Availability of data and materials
The datasets used for analysis during the current study are available from the corresponding author on reasonable request.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementTables.docx