Introduction
Sepsis is a life-threatening condition resulting from a dysregulated host response to infection [
1]. Sepsis and septic shock continue to represent significant risks for mortality in critically ill patients [
2]. In 2017, there were approximately 49 million new cases of sepsis and 11 million sepsis-related deaths worldwide [
3].
The pathophysiology of sepsis is complex, encompassing a variety of proinflammatory and immunosuppressive responses [
4]. Myeloid-derived suppressor cells (MDSCs) have been implicated in sepsis-induced immune suppression [
5]. Two main subpopulations are usually considered: polymorphonuclear MDSCs (PMN-MDSCs) and monocytic MDSCs (M-MDSCs) [
5]. MDSCs have the capacity to hinder immune responses, encompassing those modulated by T cells, B cells, and natural killer (NK) cells. PMN-MDSCs and M-MDSCs share critical biochemical attributes that enable the suppression of immune responses [
6].
Recent studies applied single-cell RNA sequencing (scRNA-seq) to understand the spectrum of immune cell states in the blood of sepsis patients [
7‐
9]. scRNA-seq has identified an immature bone-marrow-derived CD14
+ monocyte phenotype, denoted as “monocyte state 1” (MS1), which is reminiscent of M-MDSCs [
7,
8]. This monocyte phenotype is characterized by elevated expression levels of
RETN,
ALOX5AP and
IL1R2, and reduced expression of class II major histocompatibility complex (MHC-II). MS1 cells can be induced from bone marrow precursors, and display several immunosuppressive properties, including suppression of T cell proliferation and inhibition of the inflammatory activation of epithelial and endothelial cells [
7,
8]. Furthermore, an independent investigation reported the presence of a neutrophil subset, designated as “IL1R2
+ Neu” in sepsis patients [
9]. These cells exhibit gene expression profiles remarkably similar to those of MS1 cells, suggesting that common myelopoietic processes might underlie the development of both MS1 cells and IL1R2
+ Neu cells [
9].
The proportion of MS1 cells can be estimated by deconvolution of bulk RNA expression data from whole blood [
7,
8,
10]. In this study, we leveraged this validated deconvolution method to evaluate the percentage of MS1 cells using whole blood transcriptome data from a well-characterized cohort of sepsis patients. By doing so, we aimed to determine the association of MS1 cell profiles with disease presentation, outcomes and host response characteristics, using non-infected critically ill patients and healthy individuals as comparators.
Discussion
MDSCs are immature myeloid cells with immunosuppressive features found in increased numbers in the circulation of patients with inflammatory conditions. Expansion of MDSCs is considered to play a key role in sepsis-induced immunosuppression [
5]. Two recent studies reported a newly discovered monocyte state named MS1, reminiscent of M-MDSCs, of which the abundance in blood of patients with sepsis correlated with higher mortality rates [
7,
8]. The present study provides comprehensive information about the association between the proportion of MS1 cells, and disease presentation, complications and host response aberrations in critically ill patients with sepsis or a non-infectious condition.
Previously, expression of the MS1 gene program in blood was reported to be negatively associated with survival in an analysis making use of 11–15 cohorts included in meta-analyses reporting on mortality among sepsis patients [
7,
8]. In our study, MS1 cell abundance did not differ between sepsis survivors and non-survivors, and there were no mortality differences between patient groups stratified according to MS1 cell frequencies. Albeit non-significant, MS1 cell percentages expressed as a continuous variable showed a nonlinear relationship with mortality. While this might indicate that an increase in MS1 cell abundance to a certain extent may improve outcome, this possibility is speculative and requires validation. Notably, in the earlier meta-analyses the association between the MS1 cell abundance and mortality was not consistent in all individual sepsis cohorts [
7,
8].
Thus far, the relation between MS1 cell frequency, and the clinical presentation and disease associated complications in patients with sepsis was not studied in great detail. We found a clear association between the percentage of MS1 cells and disease severity, as evidenced by higher severity scores and more shock in the intermediate and high MS1 groups. Additionally, the high MS1 group more often presented with abdominal infection, which however most likely was linked to the fact that these patients more often had shock on admission to the ICU, more so than that the MS1 expansion was related to the site of infection. In agreement, we and others previously reported a higher incidence of shock in sepsis patients with an abdominal source of infection [
13,
27]. MS1 cells clearly exert immune suppressive effects, and accordingly, the proportion of MS1 cells in sepsis patients displayed a negative correlation with HLA class II gene expression [
7,
8] (considered a classic sign of immunosuppression) [
28]. Nevertheless, the MS1 score did not differ between patients who did and those who did not develop a secondary infection, a complication considered to be linked with immunosuppression [
28]. In contrast, patients with high MS1 cell abundance more often developed ARDS while on the ICU, a complication that is considered to arise from exaggerated inflammation [
29]. These data suggest that, in spite of the immune suppressive properties of MS1 cells [
7,
8], other concurrent host response aberrations may be dominant in the overall immune state of sepsis patients with high MS1 cell proportions. Indeed, our comprehensive host response analyses provide support for this notion.
We studied the association between MS1 cell abundance and the host response by analyzing blood gene expression profiles and, in a more targeted way, by evaluating 15 plasma biomarkers reflective of pathophysiological mechanisms implicated in sepsis. We applied different gene set enrichment techniques to show that an increase in MS1 cell frequency is associated with a decrease in lymphocyte-related and interferon response genes. These results support the conclusions by Reyes et al. [
8], highlighting a reduced interferon response in MS1 cells upon stimulation and a negative correlation between the MS1 gene program and expression of interferon response genes, and are consistent with the recognized immune suppressive activity of MDSCs, necessitating the inactivation of the interferon pathway [
30]. On the other hand, an increased percentage of MS1 cells was related to upregulation of TNF signaling pathways via NF-κB, IL-6/JAK/STAT3 signaling, and inflammatory response pathways. These seemingly paradoxical findings fit with the current consensus that hyperinflammation and immune suppression co-exist in sepsis patients upon ICU admission [
4], rather than that they represent successive phases [
31]. Plasma biomarker analysis demonstrated only modest differences in the inflammation domain (lower in patients with low MS1 cell numbers), while endothelial and coagulation responses were similar across MS1 groups.
Recent studies have tried to stratify patients with sepsis into more homogeneous subgroups based on host response characteristics using various unsupervised clustering methods [
17‐
19,
21,
22]. We grouped patients included in our cohort in these previously published subtypes and determined the MS1 cell frequency in each subgroup, thereby seeking to assess potential overlap. Patients with low MS1 cell percentages more often classified in the low mortality risk subtypes Mars3 [
17], adaptive [
18], and hypoinflammatory [
21,
22], which is in agreement with the association between MS1 cell frequency and disease severity. Otherwise, MS1 cell proportions did not clearly align with dominant pathophysiological mechanisms; for example, higher MS1 cell abundances were found in the SRS1 subtype (which mainly reflects an immunocompromised profile) [
19,
20], but also in the hyperinflammatory subtype (reflecting a subtype with dominant inflammatory host response pattern); these seemingly opposing associations are in agreement with our gene set enrichment analyses discussed above. We found positive correlations between MS1 cell abundance and the SRSq [
19,
20] and MDP scores, which indicate the extent of gene expression perturbation relative to a healthy state [
23‐
25]. Concurrently, the proportion of MS1 cells had a negative correlation with HLA class II gene expression, corroborating earlier findings [
7,
8] and pointing at immunosuppression [
28]. Collectively, these results suggest that, while MS1 cells clearly exert immune suppressive effects [
7,
8], their abundance in patients with sepsis upon ICU admission should be considered as one aberrant feature in a broadly disturbed host response in patients who already are critically ill (i.e., have departed from normal immune homeostasis along divergent pathophysiological pathways).
A previous study reported that the MS1 cell fraction is also expanded in ICU patients with a non-infectious condition, although to a lesser extent than in sepsis patients admitted to the ICU [
7]. Contrary to these findings, our study found no difference between sepsis patients and non-infected ICU controls, and we further show that the presence of shock is similarly associated with MS1 cell expansion in both groups. The discrepancy between our study and the one published earlier may be explained by differences in disease severities between sepsis and non-infected patients (not reported in [
7]). Analyses seeking to associate MS1 cell frequencies with other host response deviations in non-infected ICU patients showed strong similarities with results obtained in sepsis patients, including a decrease in lymphocyte-related and interferon response genes, and an upregulation of TNF signaling pathways via NF-κB, IL-6/JAK/STAT3 signaling, and inflammatory response pathways in patients with higher MS1 cell fractions. Together, these data suggest that MS1 cell expansion and its relation to other host response aberrations are primarily determined by the severity of disease and not by the inciting injury, thereby aligning with the recently proposed new concept of critical illness [
32].
Our study has strengths and limitations. We used a large well-annotated cohort of prospectively enrolled patients, allowing studies in ICU patients with or without sepsis on clinically relevant outcomes and associations with diverse host response mechanisms. The study was conducted in two ICUs in the Netherlands; results may not be generalizable to other critical care settings. Our investigation was observational and therefore does not allow conclusions about causal relationships. CIBERSORTx can only estimate cell abundances based on the cells represented in the reference matrix. Therefore, the percentages estimated using whole blood RNA profiles in this and previously reported studies [
7,
8] refer to the subpopulation of cells within the peripheral blood mononuclear cell fraction. Another limitation lies in the inherent nature of deconvolution, which does not provide a definitive picture of cellular composition. Although implementing scRNA-seq analyses and/or mass cytometry could offer more direct information to identify MS1 cells and confirm or refute the hypotheses advanced in this paper, applying these technologies in cohort studies of this magnitude poses financial and logistical challenges.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.