Frontline Science: Endotoxin‐induced immunotolerance is associated with loss of monocyte metabolic plasticity and reduction of oxidative burst

Abstract Secondary infections are a major complication of sepsis and associated with a compromised immune state, called sepsis‐induced immunoparalysis. Molecular mechanisms causing immunoparalysis remain unclear; however, changes in cellular metabolism of leukocytes have been linked to immunoparalysis. We investigated the relation of metabolic changes to antimicrobial monocyte functions in endotoxin‐induced immunotolerance, as a model for sepsis‐induced immunoparalysis. In this study, immunotolerance was induced in healthy males by intravenous endotoxin (2 ng/kg, derived from Escherichia coli O:113) administration. Before and after induction of immunotolerance, circulating CD14+ monocytes were isolated and assessed for antimicrobial functions, including cytokine production, oxidative burst, and microbial (Candida albicans) killing capacity, as well metabolic responses to ex vivo stimulation. Next, the effects of altered cellular metabolism on monocyte functions were validated in vitro. Ex vivo lipopolysaccharide stimulation induced an extensive rewiring of metabolism in naive monocytes. In contrast, endotoxin‐induced immunotolerant monocytes showed no metabolic plasticity, as they were unable to adapt their metabolism or mount cytokine and oxidative responses. Validation experiments showed that modulation of metabolic pathways, affected by immunotolerance, influenced monocyte cytokine production, oxidative burst, and microbial (C. albicans) killing in naive monocytes. Collectively, these data demonstrate that immunotolerant monocytes are characterized by a loss of metabolic plasticity and these metabolic defects impact antimicrobial monocyte immune functions. Further, these findings support that the changed cellular metabolism of immunotolerant monocytes might reveal novel therapeutic targets to reverse sepsis‐induced immunoparalysis.


INTRODUCTION
Secondary infections represent a major complication for patients admitted to the intensive care unit (ICU) with sepsis. 1 In sepsis, hyperinflammatory responses to infection are accompanied and/or followed by counterregulatory anti-inflammatory responses, which can cause a dysfunctional state in which immune cells are unable to adequately respond to pathogens. 2 This phenomenon, called sepsis-induced immunoparalysis, is associated with increased susceptibility to secondary (opportunistic) infections. 3 Microbial infections in the late phase of sepsis including bacteria (e.g., Escherichia coli and Staphylococcus aureus), fungi (e.g., Candida albicans), and viral reactivation all illustrate the clinical significance of sepsis-induced immunoparalysis, which is associated with increased ICU length of stay. [4][5][6] Moreover, the dysregulated host response is considered to be a critical factor that contributes to the persistently high mortality in sepsis, as a result of secondary infections. 7,8 Monocytes are severely affected during immunoparalysis, and they play a pivotal role in the host response during sepsis. A low human leukocyte antigen-antigen D related (HLA-DR) expression on monocytes coincides with an increase in nosocomial infections and higher mortality. 9,10 Another hallmark of sepsis-induced immunoparalysis are impaired ex vivo monocyte cytokine responses, which are likely a result of cellular reprogramming. 11 Therefore, reversing or preventing monocyte immunotolerance represents an attractive strategy to improve immune function and outcome of sepsis. 12 To facilitate the development of effective therapies, the molecular mechanisms underlying the immunotolerant state of monocytes should be unraveled.
In the past decade, it has become clear that during sepsis immune function and cellular metabolism are closely interlinked. 13 A previous study revealed dysfunction of cellular metabolism in immunotolerant monocytes. 14 Recently, in vitro LPS tolerization was found to induce prominent metabolic changes over time in THP-1 monocyte cells. 15 This study demonstrated a consecutive metabolic adaptation of these cells during immune activation, deactivation (tolerance), and resolution of inflammation. These findings support the notion that metabolic rewiring after immune stimulation may be a major mechanism that drives the tolerant state of immune cells. Importantly, this study highlights the need for in vivo studies to further characterize metabolic changes in tolerant immune cells and the impact on immune function in parallel with this metabolic shift. A clear understanding of the essential metabolic routes affected during immunoparalysis and how they interfere with immune function is still lacking. This knowledge gap needs to be bridged before effective therapies can be developed.
In the current study, the primary aim was to better understand the potentially defective adaptation of monocyte metabolic responses during immunoparalysis. To this end, a standardized protocol 16 of endotoxin-induced immunotolerance was employed, where human subjects were intravenously challenged with endotoxin (E. coli LPS) to model sepsis-induced immunoparalysis. Monocytes were studied before, during, and 1 week after endotoxin-induced immunotolerance to characterize metabolic changes associated with immunotolerance and long-lasting metabolic reprogramming following endotoxin challenge. First, the metabolic profile and immunological functions were characterized ex vivo before immunotolerance, during immunotolerance and 1 week after immunotolerance. Second, metabolic pathways found to be affected were validated for their impact on antimicrobial functions in naive monocytes.

Experimental human endotoxemia model
Experimental human endotoxemia was used as a model for sepsisinduced immunoparalysis, as extensively described in previously published studies. 16

Peripheral blood mononuclear cells and CD14+ monocyte isolation
Blood was diluted (1:1) in PBS. PBMCs were isolated from blood by differential density gradient centrifugation over

Monocyte Candida killing capacity
To determine the intracellular killing capacity of pathogens after experimental endotoxemia, CD14 + monocytes (1 × 10 5 cells/well) of 5 male volunteers were exposed to live C. albicans (2 × 10 5 yeast/well, strain UC820) for 24 h at 37 • C and 5%. CO 2 at baseline, 4 h after endotoxin challenge, and 7 days following endotoxin administration. After incubation, all cells were lysed in water and serial dilutions of lysates were plated on Sabouraud agar plates. The Candida killing capacity of CD14+ monocytes was quantified by counting remaining colonyforming units (CFUs), after overnight incubation at 37 • C and 5% CO 2 .
The percentage of killed microbes was calculated as 1 − (CFU remained after incubation with microbes/CFU determined before incubation with microbes) × 100.  Cytokines were measured in supernatants to assess the effects of metabolic modulation on cytokine responses.

In vitro oxidative burst assays
Similarly, the effects of metabolic modulation on oxidative burst were determined. Supernatants of seeded PBMCs (5.0 × 10 5 cells/well)

In vitro NADPH assays
For the detection of intracellular NADPH, isolated PBMCs were plated in 12-wells flat round-bottom plates with a density of 4.5 × 10 6 cells/well. PBMCs were preincubated for 24 h with various modulators, as described before (Table 1)

In vitro microbial killing assays
After removal of the metabolic modulators, PBMCs were washed twice with warm PBS in order to prevent interference of the metabolic inhibitors and stimuli with the live C. albicans. PBMCs were stimulated with RPMI, or C. albicans (2 × 10 5 yeast/well, UC820) for 24 h at 37 • C and 5% CO 2 . After incubation, all cells were lysed in water, and serial dilutions of lysates were plated on Sabouraud agar plates (Becton Dickinson, Heidelberg, Germany). The Candida killing capacity of the PBMCs was quantified by counting remaining CFUs overnight incubation at 37 • C and 5% CO 2 . The killing capacity was calculated as described before.

In vitro viability assays
To determine the effect of the inhibitors and stimuli of specific metabolic pathways on cell viability in PBMCs, apoptosis was assessed using Annexin-V and propidium iodide (PI) staining. Cells were stained for 15 min in 300 l RPMI containing 1 l Annexin-V (Biovision) and 5 mM CaCl 2 . Immediately before measurement 1.5 l PI (Invitrogen Molecular Probes) was added. Cells were measured using a CytoFLEX cytometer (Beckman Coulter).

Cytokine measurements
Cytokine measurements of all experiments were performed in col-

Statistical analysis
The data are represented as the mean ± SEM or median, based on their distribution. One-way ANOVA for repeated measurements was used to compare statistical differences for the means of the  and P-values were calculated using a 2-sided paired t-test. A Benjamini-Hochberg false discovery rate procedure was used to correct these P-values for type I errors. 22 Different sample normalization methods were attempted, including total area normalization, and techniques more robust to outliers like the normalization method available in the DESeq2 package. 23 In the end, it was found that normalization was not required, since the same amount of sample was used for each measurement, and the values should be semiquantitative. A P-value < 0.05 was considered statistically significant with *P < 0.05, **P < 0.01, and ***P < 0.001.

Endotoxin-induced immunotolerance affects monocyte cytokine responses, microbial killing capacity, and oxidative burst
Intravenous endotoxin administration induces immunotolerance in circulating monocytes in healthy volunteers, and these otherwise highly responsive cells are rendered irresponsive to ex vivo re-challenge with LPS. 16,18,19 Accordingly, we confirmed that monocytes isolated 4 h after in vivo endotoxin administration were irresponsive to ex vivo LPS stimulation (Fig. 1A). This effect was restored 7 days following endotoxemia. In contrast, monocyte cytokine responses to ex vivo LPS stimulation were unaffected in placebo subjects.
After in vivo endotoxin-induced immunotolerance, monocytes showed a significant attenuated capacity to induce an oxidative burst response to E. coli and S. aureus (Fig. 1B). A similar trend toward C. albicans-induced ROS release was observed. Finally, the microbial killing was investigated during endotoxin-induced immunotolerance. After in vivo endotoxin challenge, monocytes showed a reduced Candida killing capacity in 2/5 subjects (Fig. 1C). The effect of endotoxin tolerance on microbial killing in the other volunteers was inconclusive, with one volunteer showing even enhanced monocytemediated Candida killing during tolerance.

Endotoxin-induced immunotolerance reduces the metabolic plasticity of monocytes
Monocytes isolated from 5 volunteers that were infused with endo-  Fig. 2A and Table S1). In contrast, in immunotolerant monocytes obtained 4 h following in vivo endotoxemia, almost the full metabolic profile was down-regulated compared to baseline in untreated monocytes (Fig. S1A, left panel). Also ex vivo LPS stimulation did not significantly upregulate abundance of metabolites; except for quinolinate, a downstream product of the kynurenine pathway, which was significantly increased (Fig. 2A, Table S1, and Fig. S1). Monocytes (acylcarnitine), whereas this pathway was not significantly affected at other time points (Fig. 2D), which highlights the changed metabolic response 7 days after endotoxin infusion.

Modulation of cellular metabolism in naive immune cells differentially influences cytokine responses
The metabolic profiling of monocytes indicated a significant loss of metabolic plasticity in immunotolerant monocytes. Therefore, the influence of cellular metabolism on immune functions including cytokine production, oxidative burst, microbial (Candida) killing, and cell viability were determined. Metabolic pathways affected by immunotolerance were systematically targeted by inhibitors/ modulators at crucial steps in the metabolic pathways ( Fig. 3A and Table 1). The effect of each metabolic modulation on cell viability was measured after 24 h and did not reveal reduced cell viability compared to their respective vehicle control (Fig. S2). In endotoxin-induced toler-

Glycolysis, PPP, and glutaminolysis are crucial for oxidative burst
The oxidative burst of in vitro LPS-tolerized PBMCs was assessed to confirm that immunotolerance corresponds with reduced oxidative responses. Indeed reduced ROS release upon microbial restimulation was observed in LPS-tolerized PBMCs (Fig. 4A). Inhibition of pathways known to fuel NADPH (glycolysis, PPP, and glutaminolysis) reduced ROS release, while modulation of pathways leading to increased NADPH levels (DCA and NAC) increased ROS release (Fig. 4C). Measurements of intracellular NADPH levels following modulation of these NADPH-dependent pathways validated this.
The blockade of glycolysis using 2DG reduced intracellular NADPH levels if pyruvate was absent in the media, whereas NAC and DCA increased intracellular NADPH levels (Fig. 4D). No changes in intracellular NADPH concentrations were detected by the inhibition of PPP or glutaminolysis.

Immunotolerance is associated with reduced Candida killing capacity
In vitro experiments showed a significantly reduced killing capacity for C. albicans in LPS-induced tolerant monocytes (Fig. 5A). Inhibition of glycolysis (2DG) did consistently reduce Candida killing capacity in , and C968 (glutaminolysis, n = 7). Data are shown as the median with P-value of statistical comparison by the Wilcoxon Signed Rank Test (*P < 0.05, **P < 0.01, and ***P < 0.001). (C) Simplified schematic overview of the metabolic pathways that were targeted with inhibitor (green labels) or stimulator (red labels) that could lead to increased intracellular NAPDH levels (green arrows) or decreased (red dotted line) NADPH levels. (D) Relative changes in monocyte intracellular NADPH levels measured by colorimetric assay following treatment of the cells for 24 h with various modulators of cellular metabolism (n = 6, red bars). Fold changes are relative to the appropriate vehicle control (black bars; either RPMI, or RPMI without pyruvate [-PYR], or DMSO). Bars represent the mean ± SEM and were compared for significance using the Wilcoxon Signed Rank Test (*P < 0.05) healthy donors, but only when pyruvate was excluded from the culture media during the preincubation period (Fig. 5B). Modulation of the TCA cycle and PPP did influence Candida killing but showed high variability between donors. We found that inhibition of glutaminolysis (BPTES or C968) exhibited a substantial impact on Candida killing in naive monocytes (Fig. 5C)

Reduced oxidative burst in monocytes correlates with defective microbial killing mechanisms
Our data demonstrate that central metabolic pathways like glutaminolysis and glycolysis are both involved in ROS generation (Fig. 4B) and candida killing in naïve monocytes ( Fig. 5B and C). The importance of ROS production for host defense is demonstrated in patients, with genetic defects in the NADPH oxidase complex, who are unable to produce ROS and are highly susceptible to staphylococcal and Aspergillus infections. 24 Hypothetically, metabolic modulation might have influenced monocyte antimicrobial functions in similar ways; by altering NADPH-dependent oxidative burst, thereby influencing phagocytic activity and killing capacity (Fig. 6A). To test this hypothesis, we investigated the correlation between C. albicans-induced oxidative burst and fungal killing (Fig. 6B). In these experiments, C. albicans killing was measured in the presence of diphenyleneiodonium (DPI). DPI is a frequently used and potent ROS inhibitor mediated by flavoenzymes, in particular NADPH oxidase. 25 As expected, DPI-treated monocytes demonstrated drastically reduced capacity to mount an oxidative burst (Fig. 6C). In parallel, we observed a reduced capacity of monocytes to kill C. albicans yeasts, confirming the link between ROS and killing (Fig. 6D). Systemic endotoxin-induced responses in this model are milder than that suffered by septic patients, who by definition suffer from organ injury. This is reflected by the fact that endotoxin-challenged volunteers recover fully without residual deficits and can be discharged 8 h after challenge. 16,18,19 In contrast, sepsis-associated organ dysfunction is often life-threatening and assessed in all patients suspected of sepsis. 27 Moreover, severe sepsis survivors have an increased risk of long-term deficits including physical and cognitive impairment. 28 Given that, in experimental endotoxemia, already major (long-lasting) defects in metabolic plasticity of monocytes were observed, we envision that during sepsis the severe impact on the immune system may trigger even more severe changes in cellular metabolism. This possibly contributes to the sustained immune dysfunction in septic patients, who develop secondary infections days to weeks after diagnosis. 7,29 Of note, systemic LPS stimulation can result in opposite immune responses in vivo. 30 Several studies showed that LPS priming with low concentrations could augment immune responses, resembling in a proinflammatory phenotype. [30][31][32] However, in line with other endotoxemia studies, we showed a proper immunotolerant phenotype in the current study. 16,18,19 Essential metabolic differences were found between homeostasis and endotoxin-induced immunotolerance, which may reveal biomarkers or therapeutic targets for immunoparalysis. The untargeted metabolome analysis identified the metabolites quinolinate and kynurenine to be upregulated, respectively 4 h and 7 days after endotoxin administration. These are metabolites of the tryptophankynurenine pathway, which is known to be activated or dysregulated in several inflammatory conditions including infectious diseases, autoimmune disorders, malignancy, and cardiovascular diseases. 33 Accumulation of kynurenine and downstream metabolites, through increased tryptophan degradation by IDO-1, contributes to hypotension due to vasodilatation, and is therefore considered as a novel target for septic shock treatment. 34 Besides these vascular effects, alterations in this pathway correlate with disease severity and impaired immune function. 35  Glutamine has gained clinical interest as it has regulatory effects on immune cell function and may serve as a beneficial supplement in critically ill patients. 47 We observed variation in monocyte killing capacity in glutamine poor or supplemented conditions.
In line with this, trials in critically ill patients similarly observed divergent outcomes upon glutamine supplementation. 48,49 Some studies demonstrated benefits in critically ill patients, including decreased mortality and infection risk, [50][51][52] but it is now debated that glutamine supplementation can further elucidate hyperinflammation. 53 The benefit of modulating glutamine metabolism, therefore, may depend on the individual inflammatory state. Second, DCA was used in a clinical trial to treat hyperlactatemia, which included patients with septic shock, but did not improve clinical outcome. 56 Besides reducing lactate levels, DCA activates the pyruvate dehydrogenase complex (PDC) and stimulates mitochondrial glucose oxidation through increased mitochondrial pyruvate influx. 57 DCA in a sepsis mouse model promoted restoration of immunometabolism, increased survival, and improved bacterial clearance. 58 New insights and the data presented here suggest that the modulators DCA and NAC might be reconsiderable targets in sepsis treatment. However, the aforementioned trials with DCA and NAC focused on the hyperinflammatory aspects of sepsis.
One can speculate that timing of treatment and the heterogeneity of populations enrolled in these trials have concealed beneficial effects in patients subgroups, in particular for patients with profound immunosuppression. This provides a rationale for future studies on the use of these metabolic modulators in patients with sepsisinduced immunoparalysis.
Our study has some limitations. First, metabolomic profiling was performed in a small number of participants. Therefore, validation experiments were focused on metabolic pathways rather than on single metabolites. We assume that parallel to pathway analyses, individual metabolites may help to identify novel metabolic therapeutic targets. Second, our study population was within an age range of 18-35, whereas 64.9% of cases of sepsis occurs in patients at or over the age of 65. 59 Immune metabolism is a research field that is relatively new, and knowledge on the differences in immunometabolic reprogramming between adults and the elderly is scarce. Our results may, therefore, not be generalizable to populations that commonly develop sepsis. Third, our sepsis model focused on the activation of the innate immune system, although sepsis-induced immunoparalysis is characterized by defects in both innate and adaptive immunity. Fourth, our in vitro model was limited to modulate single metabolic pathways separately. Modulation may also have led to compensatory or alternative mechanisms for metabolic adaptation. Therefore, this study does not provide information about the collateral interactions of various metabolic processes or their connections with molecular, epigenetic, and transcriptional changes, which all contribute to immunotolerance.
Considering that metabolic modulation might provide new therapeutic options for sepsis, a systematic biological approach with validation in larger endotoxemia and patient cohorts is necessary to unravel the complexity of metabolic reprogramming and confirm the therapeutic value of metabolic targets in immune paralysis.
To conclude, this study provides essential insights into metabolic changes in immunotolerant monocytes, and we report a loss of metabolic plasticity in these cells. By modulation of metabolic pathways in naive immune cells, we validated the concept that metabolic pathways changed in immunotolerance can affect antimicrobial functions. Furthermore, several metabolites and metabolic pathways were identified that might represent future therapeutic targets to reverse sepsis-induced immunoparalysis. As such, this study opens up new therapeutic avenues for sepsis-induced immunoparalysis.