Background
Intensive care units (ICUs) treat a heterogeneous group of patients with life-threatening injuries requiring close monitoring and support with high-technology medical equipment and multiple medications. In this context, the search for prognostic biomarkers that would help clinicians predict patients’ outcomes has been ongoing for years [
1].
Several aspects of risk stratification biomarkers could be of interest in the ICU, especially in the context of financial constraints faced by the health care system and the strong need for optimization of limited medical resources. Indeed, besides use as benchmarking tools and in inter-ICU comparative studies, accurate outcome prediction based on biomarkers could be extremely useful for improving clinical decision-making. For example, biomarkers could help clinicians to dispense relevant treatments to patients who might benefit from them [
2].
Transcriptomics has recently emerged as a novel tool to identify such biomarkers [
3‐
8]. Most important, the availability of fully automated and user-friendly molecular platforms offers concrete solutions for routine clinical applications, even in ICU and emergency settings.
We previously identified chemokine (C-X3-C motif) receptor 1 (
CX3CR1) as the most differentially expressed gene between survivors and non-survivors after septic shock [
9,
10]. CX3CR1 is the sole receptor for fractalkine (CX3CL1) expressed on leucocyte subpopulations. Interestingly, we observed an association between decreased CX3CR1 expression on monocytes after septic shock and decreased functional responses of monocytes ex vivo [
10]. This provides a link between decreased CX3CR1 and immune dysfunction after sepsis and suggests that the measure of such a host response messenger RNA (mRNA) marker may be useful in evaluation of injury-induced immune alterations. However, whether such a biomarker could be useful in the heterogeneous group of ICU patients is unknown. Therefore, we conducted a prospective, multicentre, non-interventional study with the primary objective of studying the association between CX3CR1 mRNA expression and risk of death in a large cohort of critically ill patients. The secondary objective was to evaluate the association between CX3CR1 mRNA expression and risk of death in the subgroup of septic shock patients.
Methods
Experimental design
This prospective, multicentre, non-interventional study was conducted in six ICUs in Lyon, France. It was approved by our institutional ethical review board (Comité d’Ethique des Centres d’investigation Clinique de l’Inter-Région Rhône-Alpes Auvergne – IRB 5044).
Follow-up and outcomes
The primary outcome was the association of day 1 (D1) and D3 CX3CR1 mRNA expression and risk of death during the hospital stay of ICU patients with systemic inflammatory response syndrome (SIRS). The secondary outcome was the association of D1 and D3 CX3CR1 mRNA expression and risk of death during the hospital stay in the subgroup of septic shock patients. In both cases, mortality was evaluated at 7 days (D7) and at 28 days (D28).
Since half of the deaths in our cohort occurred before D7 and the median ICU stay was 7 days, this time point (i.e., D7) was used to define early mortality among the 725 patients eligible at D1. Conversely, D28 was used to define late mortality after ICU admission among the 515 patients eligible at D3.
Patients
From December 2009 to June 2011, every consecutive patient aged ≥18 years with an expected length of stay in the ICU of more than 2 days was prospectively enrolled if the patient met the criteria for SIRS as described in the American College of Chest Physicians/Society of Critical Care Medicine (ACCP/SCCM) 1992 consensus statement [
11] and if non-opposition to inclusion in the protocol was obtained.
Exclusion criteria comprised immunosuppression as defined by immunosuppressive treatment or corticoid treatment with dosage >10 mg/day or cumulative dose >700 mg equivalent prednisolone, aplasia as defined by number of circulating neutrophils <500 cells/mm3, primary innate immune deficiency, and extracorporeal circulation during the month before ICU admission.
No informed consent was needed, as this study was non-interventional and complementary blood samples in PAXgene® tubes (PreAnalytiX, Hombrechtikon, Switzerland) were obtained during patients’ routine blood sampling after completion of routine follow-up tests. Nevertheless, non-opposition to inclusion in the protocol was recorded from every patient or the patient’s next of kin.
Definitions
SIRS was defined as the presence of at least two of the following clinical criteria: temperature >38 °C or <36 °C, heart rate >90 beats/minute, respiratory rate >20 breaths per minute or PaCO
2 < 32 mmHg (4.3 kPa), and leucocyte count >12,000/mm
3 or <4000/mm
3 [
11].
Sepsis was defined as the presence of a proven (visible either clinically/surgically, radiologically or microbiologically) infection or a highly suspected infection at inclusion. Following the definitions proposed by Vincent et al., which partly rule out the former definitions of sepsis and severe sepsis described in the ACCP/SCCM 1992 consensus conference statement, sepsis was simply defined as an infection requiring ICU admission [
12].
Shock was defined as persistent hypotension despite adequate fluid resuscitation requiring the use of epinephrine or norepinephrine at a dose >0.25 μg/kg/minute [
13].
Antibiotic treatment was considered to be appropriate if all the pathogenic microorganisms were sensitive to at least one of the administered antimicrobial drugs as determined by an in vitro sensitivity pattern. In cases of culture-negative infection, treatment was considered appropriate if consistent with the local anti-infectious protocols.
Data collection at inclusion
The following demographic, clinical and biochemical data were collected at patient admission and/or inclusion in the protocol:
1.
Demographic characteristics: age and sex
2.
Clinical scores: Simplified Acute Physiology Score II (SAPS II) (range 0–163 [
14], calculated at inclusion), Sepsis-related Organ Failure Assessment (SOFA) score (range 0–24 [
15], calculated at inclusion and at D3) and Charlson comorbidity index (range 0–32 [
16] for which age was not taken into account).
3.
Lactic acid concentration (at D1 and at D3)
4.
Need for respiratory (mechanical ventilation), haemodynamic (vasopressor treatment) or renal (renal replacement therapy) support
5.
Presence of shock (defined as a cardiovascular SOFA score of 4)
6.
Diagnostic category (planned or urgent, surgery or medical)
8.
Presence of sepsis (either community- or hospital-acquired)
9.
Characteristics of infection if relevant (site, type of detection, identified germs)
10.
Presence of an anti-infectious treatment if relevant and its adequacy either to the antibiogram or to the local anti-infectious protocol
PAXgene® tubes were collected from patients within the first 12 h after inclusion in the protocol (D1) and at D3. Total RNA was extracted using the PAXgene® Blood RNA Kit (PreAnalytiX). Before RNA elution, the residual genomic DNA was digested using the RNase-Free DNase Set (QIAGEN, Hilden, Germany). Total RNA was reverse-transcribed into complementary DNA (cDNA) using the SuperScript VILO cDNA Synthesis Kit (Life Technologies, Carlsbad, CA, USA). The candidate gene was quantified using quantitative real-time polymerase chain reactions. Polymerase chain reactions (PCRs) were performed in a LightCycler instrument (Roche Diagnostics, Risch-Rotkreuz, Switzerland) using the standard TaqMan Fast Advanced Master Mix PCR kit according to the manufacturer’s instructions (Applied Biosystems, Foster City, CA, USA). Thermocycling was performed in a final volume of 20 μl containing 5 μM of required primers and 1 μM of required probe. PCR was performed with an initial denaturation step of 10 minutes at 95 °C, followed by 45 cycles of a touch-down PCR protocol (10 seconds at 95 °C, 29 seconds of annealing at 68 °C and a 1-second extension at 72 °C). The cDNA standards were prepared from purified PCR amplicons obtained with the corresponding primers (Additional file
1: Table S1). The second derivative maximum method was used with the LightCycler software to automatically determine the crossing point for individual samples, as previously described. Standard curves were generated by using quadruplicate cDNA standard. Relative standard curves describing the PCR efficiency of selected genes were created and used to perform efficiency-corrected quantification with the LightCycler Relative Quantification software (Roche Diagnostics). Gene expression normalization was performed using a selected housekeeping gene (hypoxanthine phosphoribosyltransferase 1 [
HPRT1]), and results were expressed as calibrated normalized relative quantity (CNRQ) [
17].
CX3CR1 expression and
CX3CR1 CNRQ are equivalent terms.
Statistical analysis
The descriptive analyses comparing surviving and non-surviving patients at D7 and D28 (all-cause mortality) after ICU admission were completed with the usual appropriate tests, such as the t test or Mann-Whitney U test for quantitative variables and Pearson’s χ2 test for qualitative variables. The Monte Carlo resampling method was used to estimate the p value of the χ2 test of independence in cases of an expected frequency <5. ICU survival curves were created using the Kaplan-Meier method. Every patient was followed for 28 days in the ICU and after ICU discharge. The results of the log-rank tests were associated with the survival representations. ROC curves were built for CX3CR1 expression levels at D1 and D3 according to D7 and D28 mortality, respectively. Best cut-off values (i.e., maximized sensitivity and specificity) for CX3CR1 mRNA expression prediction of mortality were identified according to the calculation of Youden indexes derived from ROC curve analyses. Univariate and multivariate analyses of mortality outcomes at D7 and D28 were studied with logistic regression. Parameters with a p value <0.1 in univariate analyses were kept in multivariate analyses. Parameters with >20 % missing values were not included in multivariate analyses. The 0.632+ bootstrap method was applied to adjust for potential overestimation of diagnostic performance and to estimate internal validity. For each model, 1000 resamplings with replacement were performed. The 95 % CI represents the 2.5th and 97.5th percentiles of estimates obtained in bootstrap resamples. All bootstrap estimations were performed using the ModelGood package in R. The level of significance was set at 5 %, and the results were described with a 95 % CI. The analyses were performed with IBM SPSS Statistics version 20 software (IBM, Armonk, NY, USA).
Discussion
We demonstrate, for the first time to our knowledge, that decreased CX3CR1 mRNA expression is associated with early and late mortality in critically ill patients independently of initial severity, organ dysfunction, existing comorbidities and metabolic distress. This represents the largest study of such transcriptomic markers’ association with mortality in ICU patients. In addition, this study validates, in a large cohort of patients, our previous results showing that decreased CX3CR1 mRNA is an independent predictor of death after septic shock. This suggests that the monitoring of this host response biomarker in ICU patients may provide information complementary to the usual risk stratification parameters.
Researchers in numerous small observational studies have described the rapid modification of several immune parameters after injury and the association between the extent or duration of these alterations and increased risk of death in these patients [
18]. Therefore, in the context of clinical trials already underway in which researchers are evaluating immunoadjuvant therapies in sepsis [
19], host response biomarkers are being evaluated as risk stratification parameters in ICU patients. Considering the constantly evolving immune response after severe injury, such biomarkers could be used as a stratification tool or for treatment follow-up [
20]. While some markers, such as decreased human leucocyte antigen (HLA)-DR expression on monocytes or lymphopenia, have been proposed as immunomonitoring tools [
20], the monitoring of several different markers included in a panel is probably required to accurately assess patients’ immune status after injury [
21]. In this context, considering the recent developments of standardized molecular biology techniques, the evaluation of transcriptomic markers presents the advantage of now being fully automated and therefore available with very low pre-analytical constraints [
20]. This would most likely facilitate their use on a routine basis, as opposed to markers monitored with less standardized techniques. However, to date, the capacity of such mRNA marker to predict deleterious outcomes in ICU patients has never been evaluated in a large cohort of critically ill patients. In the present study, we evaluated the performance of CX3CR1 mRNA level in the prediction of mortality in a cohort of more than 720 ICU patients. This represents the largest clinical evaluation of a molecular host response marker in ICU patients.
CX3CR1 is a receptor expressed on leucocyte subpopulations. The CX3CR1 ligand (fractalkine) is the only member of the CX3C chemokine subfamily. CX3CR1 has a role in cell chemotaxis to sites of inflammation through interaction with fractalkine fixed on activated endothelial cells. In 2007, Auffray et al. showed that CX3CR1 is overexpressed on “patroller” monocytes and is required for their rapid tissue invasion at the infection site [
22]. Interestingly, several studies investigated the expression of CX3CR1 after severe injuries. First, in a microarray study, we identified this gene as the most downregulated gene between survivors and non-survivors of septic shock [
9]. This result was confirmed in a subsequent cohort of septic shock patients at both the mRNA and protein levels [
10]. Interestingly, decreased CX3CR1 expression on monocytes from septic patients was associated with their decreased activation after ex vivo stimulation. This provides a link between decreased CX3CR1 and immune dysfunction after sepsis. In further studies in murine models of sepsis, investigators proposed a role of this molecule in bacterial killing and host defence against infection as well as in sepsis-induced organ dysfunction [
23‐
26]. This all suggests that measurement of such mRNA markers may help in evaluation of injury-induced immune alterations. Interestingly, besides sepsis, decreased CX3CR1 mRNA expression was also described in sterile inflammation, such as major vascular surgery [
27]. Of note, in a recent study of more than 1100 patients, Hoogendijk et al. reported that plasma fractalkine level was positively correlated with severity in patients with sepsis admitted to the ICU [
28]. Therefore, the tandem CX3CR1-fractalkine may represent a good prognostic marker in ICU patients. However, the performance of this marker as a risk stratification tool in total ICU patients has not been described yet. Therefore, in the present study, we evaluated the association between CX3CR1 mRNA expression and risk of death in a large cohort of critically ill patients. In addition, we confirmed the association between CX3CR1 mRNA expression and risk of death in the subgroup of septic shock patients.
We chose to evaluate both early (i.e., before D7) and late (i.e., before D28) mortality, hypothesizing that parameters associated with mortality after ICU admission may evolve over time. Indeed, we observed that the clinical parameters associated with mortality were different when we considered early versus late deaths. This was the case both in the total cohort of ICU patients and in the subgroup of septic shock patients. For example, in the multivariate analyses, Charlson comorbidity index, illustrating patients’ comorbidities, was not associated with early mortality but rather with late deaths. Similarly, SOFA score at D1 or the presence of shock at D1 was not associated with early death, while the SOFA score at D3 was significantly associated with D28 mortality. This suggests that the later course of ICU patients with SIRS may be impacted principally by the patient’s medical background (such as comorbidities) and the persistence of organ dysfunction. Conversely, early deaths may be related mostly to the severity of the initial insult. Interestingly, host immune response assessed by CX3CR1 mRNA expression constantly remained associated with both early and late outcomes. This may illustrate that alterations of the host immune response after severe injury may play a role in mortality in the ICU and therefore supports the use of immunoadjuvant therapies in sepsis.
Different CX3CR1 thresholds were selected for prediction of D7 versus D28 mortality. Indeed, cut-off values for the predictive capacity of CX3CR1 at D1 were very low compared with cut-offs at D3. This may illustrate the evolution of the host immune response over time after severe aggression. Severe injury strongly impacts the host immune response by inducing a “genomic storm” [
4,
6]. However, in patients who survive this initial aggression, their immune status should return to normal. Failure to return to normal may lead to late mortality [
18,
19]. Interestingly, these cut-off values were very similar between ICU patients and the subcohort of septic shock patients. This suggests that underlying pathophysiological mechanisms associated with clinical severity are shared by sterile inflammation and sepsis [
4,
12,
29,
30]. In our results, the association between lower CX3CR1 expression and fatal outcome was independent of the presence of either an infection or a shock. Interestingly, non-infectious SIRS seems to be as deleterious as septic shock [
31,
32]. In the subgroup of non-infected patients with SIRS, we also observed a significant association between decreased CX3CR1 mRNA level at D1 and D7 mortality (AUC 0.75, 95 % CI 0.66–0.84). In univariate and multivariate logistic regression analyses, low CX3CR1 expression (below threshold of 0.12) in non-infected patients remained significantly associated with D7 mortality independently of confounding parameters (aOR 3.63, 95 % CI 1.43–9.20,
p = 0.007). However, in this group of patients, no association between D3 CX3CR1 mRNA level and D28 mortality was observed.
This study has some limitations. In particular, we did not include measurement of other markers of injury-induced immunosuppression, such as monocytic HLA-DR. Therefore, we could not compare the performance of our transcriptomic marker versus such a “gold standard” parameter in the present cohort. In addition, as no specific immune function tests were performed, we cannot definitively prove that decreased CX3CR1 mRNA is associated with immune dysfunction in ICU patients. However, considering the interesting results obtained, these two aspects need to be confirmed in a dedicated clinical study.
Abbreviations
ACCP/SCCM, American College of Chest Physicians/Society of Critical Care Medicine; aOR, adjusted odds ratio; cDNA, complementary DNA; CNRQ, calibrated normalized relative quantity; CX3CR1, chemokine (C-X3-C motif) receptor 1; D, day; HLA-DR, human leucocyte antigen; ICU, intensive care unit; LR, likelihood ratio; mRNA, messenger RNA; NPV, negative predictive value; PCR, polymerase chain reaction; PPV, positive predictive value; qRT-PCR, quantitative real-time polymerase chain reaction; SAPS II, Simplified Acute Physiology Score II; Se, sensitivity; SIRS, systemic inflammatory response syndrome; SOFA, Sepsis-related Organ Failure Assessment; Sp, specificity
Acknowledgements
MIP-REA study group
Dr. Guillaume Marcotte and Dr. Christian Guillaume (Hospices Civils de Lyon, Department of Anaesthesiology and Critical Care Medicine, Groupement Hospitalier Edouard Herriot, University Claude Bernard Lyon 1, Lyon, France); Dr. Romain Hernu, Sylvie De La Salle, Dr. Marie Simon, and Dr. Thomas Baudry (Hospices Civils de Lyon, Medical Intensive Care Unit, Groupement Hospitalier Edouard Herriot, Lyon, France); Elisabeth Cerrato, Emmanuelle Gallet-Gorius, Audrey Larue, and Christine Alberti-Segui (Hospices Civils de Lyon - bioMérieux Joint Research Unit, Groupement Hospitalier Edouard Herriot, Lyon, France); Nathalie Panel and Marion Provent (Hospices Civils de Lyon, Intensive Care Unit, Centre Hospitalier Lyon Sud, Pierre Bénite, France); Prof. Jean-Paul Viale (Hospices Civils de Lyon, Intensive Care Unit, Hôpital de la Croix Rousse, Lyon, France); and Anne Portier (Hospices Civils de Lyon, Immunology Laboratory, Groupement Hospitalier Edouard Herriot, Lyon, France).