Background
Sepsis is defined as a life-threatening organ dysfunction due to a dysregulated host response to infection [
1]. Immunological heterogeneity is a well-known phenomenon in sepsis patients, as they may present with both hyperinflammatory and immunosuppressive phenotypes. This might explain the failure of one-fits-all approaches and illustrates that a personalized approach is required for this complex syndrome [
2]. Biomarkers to assess the individual immune status of a patient are currently not used in clinical practice but are essential to facilitate precision medicine in sepsis patients. Decreased monocytic (m)HLA-DR expression is the most studied biomarker of sepsis-induced immune suppression [
3‐
5]. Multiple studies established associations between mHLA-DR expression and outcome parameters, such as secondary infections and mortality [
6‐
10]. Moreover, there are indications from small studies that mHLA-DR expression differs between patients with different causative pathogens and/or sites of infection [
11,
12]. However, little is known about mHLA-DR expression kinetics and their relationship with sepsis characteristics and severity of disease. Detailed insight in mHLA-DR expression kinetics between subgroups of septic patients could pave the way towards increased understanding of sepsis immunopathology and aid the selection of patients that could benefit from immunostimulatory agents. In the current study, we set out to evaluate whether the kinetics of mHLA-DR expression, measured using a standardized assay, vary between patients with different primary sites of infection and different pathogens in a large cohort of septic shock patients. Furthermore, using unsupervised clustering analysis, we aimed to identify specific mHLA-DR trajectories and relate these to outcome parameters such as the occurrence of secondary infections and mortality.
Discussion
In the current study, we evaluated mHLA-expression kinetics in the largest group of septic shock patients so far, using a standardized flow cytometry-based assay. We established that differences in sites of infection or causative pathogens are not associated with differences in mHLA-DR expression kinetics. Most importantly, using unsupervised clustering, we identified three mHLA-DR trajectories (‘early improvers’, ‘delayed or non-improvers’ and ‘decliners’), and these trajectories corresponded with adverse outcome (secondary infection or death), which was significantly higher in both the decliners and delayed or non-improvers vs. the early improvers. Our findings illustrate that trajectories provide a better understanding of the relationship between mHLA-DR expression and adverse outcome and thereby have additional value over single measurements.
Although no statistically significant between-group differences in kinetics were found, patients with abdominal infections showed an increase in mHLA-DR expression over time, whereas this was not the case in the respiratory focus group. This is likely the result of the higher disease severity in patients with a respiratory infection, as we also demonstrate a significant correlation between SOFA score at admission and a decreased mHLA-DR expression at days 3–4 and 6–8. Of note, SOFA score was not associated with mHLA-DR expression at days 1–2, which suggests a delay in mHLA-DR downregulation. The association between disease severity and mHLA-DR expression is further supported by the fact that patients with a declining mHLA-DR expression (trajectory C) display a higher SOFA score, noradrenaline requirement and mortality rate compared to early improvers (trajectory A). In line with these results, others have also reported that disease severity is a main driver of sepsis-induced immunosuppression and sepsis outcome [
14].
Previous work revealed that, in patients with bloodstream infections, those with
S. pneumonia infections showed a swift increase in mHLA-DR expression over time, whereas patients with a
S. aureus infection displayed a delayed recovery [
11]. Patients with an
E. coli infection had expression levels similar to those found in healthy volunteers [
11]. However, the population was heterogeneous and the observed differences might therefore be explained by the large differences in disease severity between patients in the different pathogen categories [
11]. In the current cohort of septic shock patients, we do not find differences in mHLA-DR expression kinetics between patients with Gram-positive, Gram-negative and negative cultures. This finding is strengthened by the fact that disease severity and outcome were similar across pathogen categories. A study using genome-wide gene expression analysis in mice revealed that Gram-positive and Gram-negative sepsis elicit common downstream pathways [
15], which underscores that differences in mHLA-DR between these pathogen categories are not likely.
Overall, we demonstrate that patients with a decreasing mHLA-DR expression over time have the highest occurrence of secondary infections and mortality rate. This is in accordance with literature, as our group and others have previously described the relationship between decreased or non-recovering mHLA-DR expression and unfavourable outcome [
6,
7,
16‐
21]. However, to the best of our knowledge, we are the first to apply unsupervised trajectory modelling. This data-driven ‘bottom-up’ approach might be more sensitive to identify endotypes of septic shock patients associated with unfavourable outcome. Previous results showing that recovery of mHLA-DR expression occurred less in non-survivors [
16] are in line with our data showing that patients with decreasing mHLA-DR expression in trajectory C had a higher mortality. It is remarkable that, although the decliners had the highest disease severity at baseline, their mHLA-DR expression on days 1–2 was the highest. The absence of a correlation between mHLA-DR expression at days 1–2 and the severity of disease or any of the outcome parameters illustrates that a single measurement at this specific time point is of minor value. Furthermore, because the trajectories of the decliners and early improvers intersect at days 3–4, a single mHLA-DR measurement at days 3–4 is also not suitable to predict overall adverse outcome. These observations further indicate that it is the course of mHLA-DR over time (dynamic) rather than its value at a single time point (static), which is of clinical significance. To this end, patient enrichment in studies evaluating immunomodulatory treatments for sepsis-induced immunosuppression should be performed using dynamic mHLA-DR expression.
Strikingly, the overall mHLA-DR expression in our septic shock cohort was very low regarding the fact that the cut-off used for the treatment of sepsis-induced immunosuppression was 8000 AB/cell in a clinical trial [
22]. In our cohort, 77% of the patients at days 3–4 and 68% at days 6–8 were below this threshold. These low values are in agreement with recent results obtained in clinical trials that included mHLA-DR monitoring [
23,
24]. Thus, the identification of the appropriate threshold defining sepsis-induced immunosuppression requires further study, preferentially in a multicentre setting. It is likely that a cut-off for selecting the most severely affected patients will be more in the range of 5000 AB/cell than 8000 AB/cell. Despite the association with adverse outcome, the accuracy of low mHLA-DR expression to predict clinical outcome at the individual patient level was limited in our cohort. That said, considering the amount of literature associating low mHLA-DR and unfavourable outcomes, it remains a valuable marker for enrichment in clinical trials, for example to select a group of patients that may benefit from immunomodulating therapies [
25,
26]. Nevertheless, in light of the fact that septic shock is a highly complex immunological syndrome, involving multiple immune defects that vary between patients and within patients over time [
27], it remains debatable whether a single biomarker will ever be sufficient to gauge a patient’s overall immune status. Therefore, future research should be focused on the identification of combinations of biomarkers to classify immune endotypes in sepsis [
2].
Strengths of the current study include the use of a large cohort comprised of only septic shock patients, which reduces variation compared to other studies in patients with multiple sepsis severities. Furthermore, we used a standardized mHLA-DR assay which facilitates generalizability and comparison with other cohorts using the same methodology. Finally, we used an innovative data-driven approach to identify mHLA-DR expression trajectories. Several limitations also need to be addressed. First, we performed the kinetics and trajectory analyses in subgroups of patients who remained in the ICU for at least 4 or 6 days, respectively. This introduced selection bias, as patients who died or were discharged at earlier time points were not taken into account. Second, disease severity parameters were only available at admission. A longer follow-up of these parameters would have allowed a more extensive analysis of the relationship between the kinetics of disease severity and mHLA-DR expression. Third, the number of patients with specific pathogens (e.g.
S. aureus or
P. aeruginosa), other groups of pathogens (e.g. fungi, viruses) and sites of infection other than abdominal, respiratory and urinary tract were too low to draw conclusions about mHLA-DR expression kinetics. Fourth, despite the large cohort of septic shock patients, the sample size of studied subgroups was relatively low, which might explain the lack of significance for secondary infections across the trajectories. Due to small sample sizes in subgroups, the absence of a statistical difference cannot rule out the presence of an association. Finally, approximately 60% of the patients in our cohort underwent a surgical intervention, which can impact mHLA-DR expression [
28].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.