Cytokine production by ex vivo (EV)-stimulated leukocytes is commonly used to gauge immune function and frequently proposed to guide immunomodulatory therapy. However, whether EV cytokine production capacity accurately reflects the in vivo (IV) immune status is largely unknown. We investigated relationships between EV monocyte cytokine responses and IV cytokine responses in a large cohort of healthy volunteers using a highly standardized IV model of short-lived LPS-induced systemic inflammation, which captures hallmarks of both hyperinflammation and immunological tolerance. Therefore, 110 healthy volunteers were intravenously challenged with 1 ng/kg LPS twice: on day 0 to determine the extent of the IV (hyper)inflammatory response and on day 7 to determine the degree of IV endotoxin tolerance. Baseline EV monocyte cytokine production capacity was assessed prior to LPS administration. Short-term and long-term EV tolerance was assessed in monocytes isolated 4 h and 7 days after LPS administration, respectively. No robust correlations were observed between baseline EV cytokine production capacity and IV cytokine responses following LPS administration. However, highly robust inverse correlations were observed between IV cytokine responses and EV cytokine responses of monocytes isolated 4 h after IV LPS administration. No correlations between IV and EV tolerance were found. In conclusion, attenuated EV cytokine production capacity reflects ongoing IV inflammation rather than immune suppression. Results of EV assays should be interpreted with caution at the risk of improper use of immuno­stimulatory drugs.

© 2022 The Author(s). Published by S. Karger AG, Basel

Introduction

Inflammatory syndromes such as sepsis, mainly due to bacterial infections but also caused by viral infections such as COVID-19, pose a major health care burden. Even prior to the COVID-19 pandemic, sepsis was considered the leading cause of death worldwide with 11 million deaths annually, accounting for 19.7% of all global deaths [1]. Apart from the lives lost, sepsis also represents an enormous economic burden to society [2-4]. Hence, effective sepsis interventions remain an urgent and unmet medical need.

Unfortunately, dozens of specific adjuvant therapies mainly focusing on inhibition of the immune system in sepsis failed to improve clinical outcomes to date [5]. This is likely explained by its highly complex and heterogeneous pathophysiology, which can entail both hyperinflammation, as well as (concurrent) profound immune suppression, a phenomenon known as sepsis-induced immunoparalysis [6, 7]. Therefore, sepsis research focus is now shifting toward more personalized treatment approaches tailored to the immunological phenotype of the individual patient. As a result, reliable immunoprofiling tools to assess immune function are increasingly warranted, not only to stratify patients into different treatment subgroups but also to monitor treatment responses over time.

A currently commonly used method to gauge immune function in critically ill patients is to determine cytokine production by leukocytes that are stimulated ex vivo (EV) with bacterial endotoxin (lipopolysaccharide [LPS]) or other inflammatory stimuli [8-12]. In this context, a low EV cytokine response is presumed to reflect a suppressed in vivo (IV) “immune status” and has therefore frequently been proposed as a diagnostic tool to identify patients who might benefit from immunostimulatory therapy [13-15]. Furthermore, EV and in vitro stimulations of leukocytes are widely applied techniques in translational studies to investigate immunophysiological mechanisms or to test potential therapeutic interventions [16-18]. However, surprisingly little is known about whether these EV and in vitro models adequately reflect the IV situation.

Comparison of EV and IV immune responses in patients in general but particularly in critically ill patients is challenging because there is no standardized immunological insult due to variance in patient-specific factors such as pathogen, site of infection, bacterial or viral load, time since disease onset, comorbidities, age, sex, and use of medication. In contrast, the experimental human endotoxemia model, in which healthy volunteers are intravenously challenged with LPS, is a highly standardized and reproducible model of systemic inflammation, capturing many immunological hallmarks of early “hyperinflammatory” sepsis [19]. Furthermore, the LPS-induced immune response in this model is followed by a refractory “endotoxin-tolerant” state, which is exemplified by a severely blunted response upon a second challenge with the same dose of LPS [20, 21]. This phenomenon bears many similarities to sepsis-induced immunoparalysis often observed in later stages of sepsis [19-21]. In the present work, we investigated to what extent cytokine responses upon EV and in vitro stimulation of leukocytes with a wide variety of stimuli reflect the IV immune response in a large mixed cohort of healthy volunteers undergoing repeated experimental endotoxemia.

Materials and MethodsSubjects

The protocol for this study was approved by the Local Ethics Committee (CMO Arnhem-Nijmegen; reference Nos. NL68166.091.18 and 2018-4983). One hundred and ten healthy volunteers between 18 and 35 years of age were recruited. All subjects provided written informed consent and were included after medical history, physical examination, routine laboratory tests, and a 12-lead electrocardiogram revealed no abnormalities. Smoking, use of any medication (contraceptives precluded), previous participation in experimental human endotoxemia, or signs of acute illness within 2 weeks prior to the start of the study were considered exclusion criteria. All study procedures were performed in compliance with the Declaration of Helsinki and its latest revisions.

Study Design

We performed a prospective experimental cohort study, the design of which is depicted in Figure 1. To evaluate the relationship of EV and in vitro models to the IV setting, both the primary innate immune response as well as the development of LPS-induced monocyte hyporesponsiveness (further referred to as “tolerance”) were studied in these three compartments concomitantly. A detailed description of the experimental setup for each compartment is provided below and in Figure 1. In short, to study the IV immune response and development of IV tolerance, subjects were challenged twice with an intravenous bolus of the same dose of LPS. The first LPS challenge (on day 0) served to quantify the primary cytokine response and to induce endotoxin tolerance. The second LPS challenge (on day 7) served to quantify the degree of tolerance, reflected by a markedly attenuated cytokine response compared to the first challenge. One hour before and 4 h after the first LPS challenge, as well as 1 h before the second LPS challenge, the EV cytokine production capacity was assessed in monocytes that were stimulated with a wide palette of pathogen-associated molecular patterns (PAMPs) and heat-killed pathogens (see section below for details). To evaluate in vitro tolerance, naive monocytes (i.e., isolated 1 h before the first LPS challenge) were isolated and incubated with LPS or culture medium (negative control) for 24 h, washed and rested, and then stimulated again on day 5 with LPS for 24 h, after which cytokine production was quantified. In addition, peripheral blood mononuclear cells (PBMCs) were isolated 1 h before the first LPS challenge and stimulated with LPS to investigate the relationship between monocyte and PBMC cytokine responses.

Fig. 1.

Schematic overview of the study procedures. LPS, lipopolysaccharide; TLR, Toll-like receptor; AUC, area under the cytokine time-concentration curve.

Experimental Human Endotoxemia (IV LPS Administration)

All experimental endotoxemia-related procedures were performed as described previously and were identical on both LPS challenge days (days 0 and 7) [19, 22]. In short, subjects were admitted to the research unit of the Radboud University Medical Center for 8 h. Subjects had to refrain from alcohol and caffeine (24 h) and food and drinks (12 h) prior to LPS administration. A radial artery catheter (BD Infusion Therapy Systems, Sandy, UT, USA) and antebrachial venous cannula were placed to allow serial blood sampling, hemodynamic monitoring, and administration of fluids and LPS, respectively. In the 45 min prior to LPS administration, hydration fluids (2.5% glucose/0.45% sodium chloride) were administered as a 1.5-L prehydration bolus to reduce the risk of vasovagal collapse [23] and thereafter at a rate of 150 mL/h for the remainder of the experiment. Directly after prehydration, a bodyweight-adjusted bolus dose of 1 ng/kg LPS (Escherichia coli-derived, Type O113, lot No. 94332B1; List Biological Laboratories, Campbell, CA, USA) was administered. Blood samples were serially obtained to construct time-concentration curves of circulating cytokines.

EV Stimulation of PBMCs and Monocytes

Cytokine production capacity was assessed 1 h before (T = −1, monocytes and PBMCs, the latter to explore the relationship between monocyte and PBMC-derived cytokine responses) and 4 h after (T = 4, monocytes) the first LPS challenge (on day 0) and 1 h before the second LPS challenge (T = −1, day 7, monocytes). To this end, PBMCs were isolated from ethylenediaminetetraacetic-acid-anticoagulated blood using Ficoll-Paque (GE Healthcare, Chicago, IL, USA) isolation as described in detail elsewhere [24]. Subsequently, CD14+CD16− monocytes were isolated by immunomagnetic negative selection using a monocyte isolation kit (EasySepTM Human Monocyte Isolation Kit, STEMCELL Technologies, Cologne, Germany) as per the manufacturer’s instructions. Purity of the monocyte fraction was determined using a hematology analyzer (Sysmex XN-450, Hamburg, Germany) prior to stimulation, and median (interquartile range) purity was 88 (86–92)% at day 0 T = −1, 90 [86–94]% at day 0 T = 4 and 88 (84–90)% at day 7 T = −1. Next, PBMCs and monocytes were resuspended in culture medium (Dutch-modified Roswell Park Memorial Institute [RPMI] 1640 + gentamycin 50 μg/mL + sodium pyruvate 1 mM + glutamax 2 mM) and seeded in 96-well polystyrene plates (5 × 105 cells/well in round-bottom plates for PBMCs, 105 cells/well in flat-bottom plates for monocytes). Monocytes were stimulated for 24 h at 37°C and 5% CO2 with the PAMPs LPS, β-glucan, resiquimod (R848), Pam3Cys, polyinosinic:polycytidylic acid, and CpG motifs as well as with heat-killed pathogens E. coli, Pseudomonas aeruginosa, Staphylococcus aureus, and Candida albicans (see Table 1 for detailed descriptions and used concentrations) in the presence of 10% human-pooled serum for 24 h at 37°C and 5% CO2. PBMCs were stimulated solely with 10 ng/mL LPS for 24 h at 37°C and 5% CO2. After 24 h, supernatants were collected and stored at −80°C until analysis.

Table 1.

Stimuli used for EV and in vitro leukocyte stimulation assays

In vitro Tolerance Induction

To assess the development of in vitro tolerance, monocytes were isolated 1 h before the first IV LPS challenge day (T = −1 on day 0) as described above. Next, monocytes were resuspended in culture medium and seeded in flat-bottom 96-well polystyrene plates (105 cells/well). Cells were incubated with LPS (10 ng/mL) or culture medium alone for 24 h at 37°C and 5% CO2 in the presence of 10% human pooled serum. After this initial incubation, supernatants were discarded, cells were washed and incubated in culture medium with 10% human pooled serum for 4 days. Culture medium was refreshed after 2 days. After 5 days, all monocyte cultures were stimulated for 24 h with 10 ng/mL LPS after which supernatants were collected and stored at −80°C until analysis.

Cytokine Analysis

For plasma cytokine determination, ethylenediaminetetraacetic-acid-anticoagulated blood was centrifuged (10 min, 2,000 g, 4°C) directly after withdrawal, and plasma was stored at −80°C until analysis. Concentrations of tumor necrosis factor (TNF), interleukin (IL)-1 receptor antagonist (Ra), IL-6, IL-8, IL-10, macrophage inflammatory protein (MIP)-1α, monocyte chemoattractant protein (MCP)-1, granulocyte colony-stimulating factor (G-CSF), and interferon-γ-induced protein (IP)-10 were determined batchwise using a simultaneous Luminex assay (Milliplex, Millipore, Billerica, MA, USA) as per the manufacturer’s instructions. Concentrations of TNF, IL-1β, IL-6, IL-1Ra, and IL-10 in cell culture supernatants were measured batchwise using an ELISA kit according to the manufacturer’s instructions (R&D Systems, Minneapolis, MN, USA).

Statistical Analysis

Distribution of data was tested for normality using Shapiro-Wilk’s test, and data with a non-Gaussian distribution were log-transformed prior to statistical testing. Data with a Gaussian distribution are presented as mean ± standard error of the mean and nonparametric data as geometric mean (95% confidence interval) or median (interquartile range). For IV circulating cytokine data, the area under the plasma cytokine concentration-time curve (AUC) was used as an integral measure of the cytokine response over time during each LPS challenge day. IV endotoxin tolerance was then quantified for each cytokine by calculating the log2 fold change between the AUC on day 7 and the AUC on day 0 (Fig. 1). EV tolerance was then quantified for each cytokine by calculating the log2 fold change between the T = 4 and T = −1 EV responses on day 0 (short-term tolerance, Fig. 1) and by calculating the log2 fold change between the day 7 T = −1 and day 0 T = −1 EV responses (long-term tolerance, Fig. 1). In vitro tolerance was quantified for each cytokine by calculating the log2 fold change between the in vitro LPS-LPS and RPMI-LPS exposure conditions (Fig. 1).

Log2 fold change tolerance values were tested for significance using a one-sample t test against a hypothetical mean of 0. For repeated measures, two-way ANOVA (time × day interaction term) was used to analyze differences in the IV cytokine time-concentration curves between the two LPS challenge days. Paired Student’s t tests were used to analyze differences in the AUCs of IV responses and between the different time points and conditions of the EV and in vitro responses, respectively. Pearson’s correlation analysis was used to explore the association between IV, EV, and in vitro responses, and p values were corrected for multiple testing using the false discovery rate method according to Benjamini and Hochberg [25]. A two-tailed p value <0.05 was considered statistically significant. All statistical analyses were performed using GraphPad Prism version 9 (GraphPad Software, San Diego, CA, USA) and R version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria).

Results

One hundred and ten subjects were included in this study. Baseline demographic characteristics are displayed in Figure 2. All procedures and symptoms associated with the experimental endotoxemia protocol were well tolerated by all subjects, and no adverse events occurred during the study.

Fig. 2.

Baseline demographic characteristics of the study cohort.

LPS-Induced Cytokine Responses and Tolerance Development in the Different Compartments

As expected, intravenous administration of LPS resulted in a profound, but transient, increase in plasma concentrations of pro-inflammatory cytokines TNF, IL-6, IL-8, IP-10, MIP-1α, MCP-1, and G-CSF as well as anti-inflammatory cytokines IL-1Ra and IL-10 in all subjects on both LPS challenge days (Fig. 3a, b; online suppl. Fig. 1A; see www.karger.com/doi/10.1159/000525572 for all online suppl. material). The response upon the second LPS challenge was severely blunted for all cytokines compared to the first challenge (median log2 fold change in AUC ranging from −2.8 for G-CSF to −0.3 for MCP-1, all p < 0.0001), indicative of IV endotoxin tolerance (Fig. 3a, b; online suppl. Fig. 1A).

Fig. 3.

Cytokine responses and tolerance development in the different compartments. a Time-concentration curves of IV cytokine responses on both LPS challenge days. Data are displayed as geometric mean with 95% confidence intervals. p values were computed using repeated measures two-way ANOVA (time × day interaction term) on log-transformed data. b AUCs of IV cytokine responses on both LPS challenge days (day 0 and day 7). Data are displayed as scatterplots with the horizontal line indicating the median. p values were computed using paired Student’s t tests on log-transformed data. c Cytokine responses of monocytes isolated before the first LPS challenge (day 0, T = −1) and 4 h afterward (day 0, T = 4) that were stimulated EV with various stimuli (see Table 1 for descriptions and concentrations used). Data are displayed as scatterplots with the horizontal line indicating the median. p values were computed using paired Student’s t tests on log-transformed data. d Correlations between cytokine responses in monocytes (marked with “M”) and PBMCs (marked with “P”) upon EV stimulation with 10 ng/mL LPS. Both cell types were obtained before the first LPS challenge (day 0, T = −1). Correlation coefficients are indicated by color and were determined using Pearson’s correlation on log-transformed data, and significant correlations are marked with an asterisk. e In vitro tolerance development in monocytes. Naive monocytes (obtained 1 h before the first LPS challenge [T = −1 on day 0]) were incubated with LPS (10 ng/mL) or culture medium (RPMI) for 24 h. Subsequently, cells were washed and incubated in RPMI for 4 days after which they were stimulated with 10 ng/mL LPS for 24 h. Data displayed as scatterplots with the horizontal line indicating the median. p values were computed using paired Student’s t tests on log-transformed data. p values were corrected for multiple testing using the FDR method according to Benjamini and Hochberg. TNF, tumor necrosis factor; IL, interleukin, Ra, receptor antagonist; G-CSF, granulocyte-colony-stimulating factor; FDR, false discovery rate. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Upon EV stimulation of naive monocytes isolated at T = −1 on day 0 (before the first IV LPS challenge), production of TNF, IL-1β, IL-6, IL-1Ra, and IL-10 was induced by all stimuli except β-glucan and CpG motifs, which only induced IL-1Ra production (Fig. 3c; online suppl. Fig. 1B). LPS-induced EV cytokine production by monocytes and PBMCs obtained at this time point were strongly correlated (Fig. 3d), indicating that monocytes are the primary cytokine producers in these short-term EV stimulation experiments. In monocytes that were isolated 4 h after the first IV LPS challenge (T = 4 on day 0), production of pro-inflammatory and anti-inflammatory cytokines was significantly lower for most stimuli, indicative of short-term EV tolerance (Fig. 3c; online suppl. Fig. 1B). In contrast, 4 h following IV LPS administration, EV monocyte responses to S. aureus were enhanced, reflected by higher production of TNF, IL-6, IL-1Ra (all p < 0.0001), and IL-10 (p < 0.01; Fig. 3c). In contrast to the profoundly suppressed IV cytokine response upon the second LPS challenge on day 7 compared to the first challenge (Fig. 3a, b; online suppl. Fig. 1A), EV cytokine production capacity was largely restored in monocytes that were isolated at T = −1 on day 7 (before the second LPS challenge), although some degree of tolerance persisted at this time point. The latter was particularly apparent for monocytes stimulated with LPS and E. coli, indicating long-term EV tolerance (online suppl. Fig. 2). Of interest, EV cytokine responses to stimulation with S. aureus on day 7 T = −1 were also attenuated compared with day 0 T = −1, indicative of long-term EV tolerance to this type of stimulation, whereas short-term responses (i.e., at day 0 T = 4) were enhanced as described above (online suppl. Fig. 2; Fig. 3c).

Finally, naive monocytes obtained prior to IV LPS administration that were incubated with culture medium for 5 days and subsequently stimulated with LPS for 24 h (RPMI-LPS condition) displayed high production of TNF, IL-6, IL-1Ra, and IL-10 (Fig. 3e). Conversely, naive monocytes that were initially stimulated with LPS in vitro for 24 h allowed to rest for 4 days and subsequently restimulated with LPS (LPS-LPS condition) showed marked in vitro tolerance, reflected by diminished levels of pro-inflammatory TNF and IL-6 (both p < 0.0001, Fig. 3e). Production of IL-1Ra was also attenuated, albeit to a lesser degree, whereas IL-10 release was enhanced (both p < 0.0001, Fig. 3e).

Correlations among and between IV and EV Cytokine Responses

Correlations between EV and IV cytokine responses (the latter expressed as AUC, reflecting the integral cytokine response over time) on the first IV LPS challenge (day 0) are displayed in Figure 4. No correlations were identified between EV cytokine responses of naive monocytes obtained before the first IV LPS challenge to any of the stimuli and the IV cytokine response to LPS, except for Pam3Cys, for which a single weak but significant correlation was found between IV IL-1Ra production and EV IL-6 production (r = 0.23, p = 0.03) and S. aureus, for which relatively weak, but statistically significant, inverse correlations were found between EV IL-6 production and IV concentrations of TNF (r = −0.30; p = 0.004), IL-6 (r = −0.23; p = 0.03), IL-8 (r = −0.25; p = 0.02), and IP-10 (r = −0.26; p = 0.01). However, in monocytes isolated 4 h after IV LPS administration, widespread and robust inverse correlations between IV and EV cytokine responses were observed for virtually all stimuli, indicating an association between IV cytokine responses and EV tolerance development (Fig. 4). So, plasma levels of predominantly pro-inflammatory cytokines upon the first IV LPS challenge strongly correlated to the extent of short-term EV tolerance (Fig. 5a; online suppl. Fig. 3A).

Fig. 4.

Correlations between IV cytokine responses (expressed as the AUC) upon the first LPS challenge (marked with IV; horizontal axis) and cytokine responses of monocytes stimulated EV with various stimuli (marked with EV; vertical axis; see Table 1 for descriptions and concentrations used). Cytokine responses in monocytes isolated before the first LPS challenge (day 0, T = −1) are displayed in the left columns, whereas cytokine responses in monocytes isolated 4 h after the first LPS challenge (day 0, T = 4) are displayed in the right columns. Correlation coefficients are indicated by color and were determined using Pearson’s correlation on log-transformed data. Significant correlations are marked with an asterisk. Two examples of correlations are displayed in the scatterplots on the right. p values were corrected for multiple testing using the FDR method according to Benjamini and Hochberg [25]. AUC, area under the cytokine time-concentration curves; TNF, tumor necrosis factor; IL, interleukin; Ra, receptor antagonist; IP, interferon gamma-induced protein; MCP, monocyte chemoattractant protein; MIP, macrophage inflammatory protein; G-CSF, granulocyte-colony-stimulating factor; FDR, false discovery rate.

Fig. 5.

Correlations between IV cytokine responses upon the first LPS challenge (expressed as the AUC) and the extent of EV tolerance (a), IV tolerance (b), and in vitro tolerance (c). Tolerance values are expressed as log2 fold change, with lower values indicating more pronounced tolerance and the horizontal lines indicating the mean. Tolerance values were tested for significance using a one-sample t test against a hypothetical mean of 0. Correlation coefficients were determined using Pearson’s correlation, and IV AUCs were log-transformed prior to analysis. p values were corrected for multiple testing using the FDR method according to Benjamini and Hochberg. AUC, area under the cytokine time-concentration curve; TNF, tumor necrosis factor; IL, interleukin; Ra, receptor antagonist; MIP, macrophage inflammatory protein; EC, E. coli; PA, P. aeruginosa; SA, S. aureus; FDR, false discovery rate. * p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

As described earlier, some degree of EV tolerance was still observed in monocytes isolated prior to the second LPS challenge on day 7 (long-term EV tolerance, online suppl. Fig. 2), but no robust correlations were observed between the IV cytokine response upon the first LPS challenge and the extent of long-term EV tolerance (online suppl. Fig. 3B). The degree of IV endotoxin tolerance following the second IV LPS challenge on day 7 was strongly correlated to plasma levels of predominantly pro-inflammatory cytokines upon the first IV LPS challenge, with high cytokine responders exhibiting more extensive IV tolerance at day 7 (Fig. 5b; online suppl. Fig. 4A). In contrast to IV and EV tolerance, no significant correlations were observed between the IV cytokine response upon the first LPS challenge and the extent of in vitro tolerance (Fig. 5c; online suppl. Fig. 4B). When evaluating relationships between the extent of IV tolerance and the extent of short- and long-term EV and in vitro tolerance, some weak correlations were found for a few stimulus-cytokine pairs, but no robust relationship between tolerance in these different experimental setups was present (Fig. 6).

Fig. 6.

Correlations between the extent of IV endotoxin tolerance (IV; horizontal axis) and short-term EV tolerance (EV; vertical axis (a)), long-term EV tolerance (EV; vertical axis (b)), and in vitro tolerance (vertical axis (c)). Correlation coefficients are indicated by color and were determined using Pearson’s correlation on log2 fold change data. Significant correlations are marked with an asterisk, and p values were corrected for multiple testing using the FDR method according to Benjamini and Hochberg. IV, in vivo; EV, ex vivo; TNF, tumor necrosis factor; IL, interleukin; Ra, receptor antagonist; IP, interferon gamma-induced protein; MCP, monocyte chemoattractant protein; MIP, macrophage inflammatory protein; G-CSF, granulocyte colony-stimulating factor; FDR, false discovery rate.

Discussion

EV stimulation of immune cells is a common and widely used technique to gauge the immune function of critically ill patients [10, 11, 13, 14], and this method is frequently proposed to enrich patient populations in trials, e.g., investigating immunomodulatory therapies. Moreover, EV and in vitro stimulations are also broadly applied in translational studies to explore immunopathophysiological pathways or to test potential new immunomodulatory interventions [16, 18, 26]. However, it is thought that tissue-resident macrophages, not circulating immune cells, are mainly responsible for the IV cytokine response as for example chemotherapy-induced leukopenic mice produce more, not less, cytokines following LPS administration [27]. As a result, cytokines measured in blood mainly reflect spillover from the tissues, and whether EV or in vitro models accurately reflect the IV immune status was therefore hitherto largely unclear. In the present study, we investigated these associations using paired measurements in a large mixed cohort of healthy volunteers undergoing repeated experimental endotoxemia, the only IV sepsis model available in humans. We demonstrate that EV and in vitro cytokine responses to a wide variety of stimuli do not robustly correlate with the IV cytokine response to repeated intravenous LPS administration, suggesting that these measurements have limited or no predictive value for a patient’s immune response evoked by a PAMP or an infection. Furthermore, the degree of endotoxin tolerance observed in the EV and in vitro models also did not correlate to the extent of IV endotoxin tolerance. This finding underscores that the circulating leukocyte compartment behaves completely different than the wide variety of tissue-resident immune cells that are activated in response to intravenous administration of a pathogen-derived stimulus.

We and others have previously reported that EV cytokine production by LPS-stimulated whole blood does not correlate with the IV cytokine response to LPS administration in small cohorts of healthy volunteers (n = 15 and n = 20, respectively) [28, 29]. Moreover, EV endotoxin tolerance was also found to resolve much more swiftly compared to IV endotoxin tolerance in a previous study performed in 16 volunteers undergoing experimental endotoxemia [20]. Similarly, others compared the transcriptomic response of LPS-stimulated whole blood to the transcriptomic whole blood response to intravenous LPS administration in healthy volunteers by analyzing expression patterns of 38 genes in both compartments [30]. Although expression of most genes (68%) changed in the same direction in both compartments, several archetypal pro-inflammatory cytokines such as TNF, IFN-γ, and IL-1β showed attenuated expression EV compared to IV. Furthermore, several markers related to T-cell and neutrophil activation were even expressed in opposite directions, again underscoring that these EV models may not adequately represent the IV innate immune response [30]. However, it must be acknowledged that IV and EV responses were compared in an unmatched fashion (e.g., in different individuals), precluding proper correlation analyses [30]. Our present work confirms these differential IV and EV responses and endotoxin tolerance characteristics in a much larger cohort of volunteers, using paired measurements and using a greater variety of inflammatory stimuli. These data indicate that EV models do not accurately reflect the more complex IV innate immune response. Our data also illustrate once again that the circulating leukocyte compartment does not contribute to the IV cytokine response to intravenous LPS administration.

Interestingly, in monocytes that were obtained 4 h after LPS administration, very robust inverse correlations were observed between the EV cytokine production capacity to multiple stimuli and the IV cytokine response. Perhaps counterintuitively, these results suggest that an attenuated EV cytokine response is not an indication of systemic immune suppression but might rather reflect ongoing (hyper)inflammation IV. This hypothesis is supported by a very recent study in 40 patients with early septic shock, in whom attenuated EV cytokine responses were associated with more pronounced IV inflammation (e.g., higher blood cytokine levels), whereas high EV cytokine production was associated with more modest IV inflammation [31]. This has important implications for the manner in which these assays are currently used, as low EV cytokine production was even proposed as a gold standard to identify patients who might be eligible for immunostimulatory therapy [14]. Using such an approach, especially during early sepsis, may actually result in the use of immunostimulatory agents in the most inflamed patients, which may present serious risks. In light of the findings mentioned above, EV cytokine production capacity in sepsis patients should be interpreted with care. This may also apply to other leukocytic biomarkers, such as monocytic human leukocyte antigen (mHLA)-DR expression. Reduced mHLA-DR expression is currently the most widely accepted marker to identify immunoparalysis in sepsis. However, it was shown to be correlated to EV LPS-induced TNF production in sepsis patients [32]. Considering the results of the present study, reduced mHLA-DR expression may therefore also reflect ongoing hyperinflammation rather than immune suppression and may also lead to improper use of immunostimulatory drugs.

Several factors might explain the differences between IV and EV cytokine responses. First, although EV and in vitro stimulation models offer the possibility to stimulate different subsets of immune cells (e.g., PBMCs, monocytes, or whole blood), they are by definition solely focused on cytokine production in the blood compartment. However, the systemic IV immune response is much more complex and involves multiple immunological and non-immunological cell types such as tissue-resident macrophages and endothelial cells. Furthermore, IV responses are strongly regulated by feedback loops not only within or between other immunological organs such as the lymphoid tissues, spleen, and the bone marrow [7] but also involving other compartments such as the gut microbiota [33] and the hypothalamic-pituitary-adrenal axis [34]. Moreover, in septic patients, differences in responses to live pathogens IV as opposed to EV stimulation with heat-killed pathogens or Toll-like receptor (TLR) ligands might further limit comparability. Finally, the lack of standardized methodology and differences in (concentrations of) the used stimuli, cell types, and the number of stimulated cells, as well as timing of measurements might also complicate the interpretation of EV cytokine production in septic patients [12, 15].

An unexpected but interesting finding was that EV cytokine responses to the Gram-positive bacteria S. aureus were enhanced in monocytes that were isolated 4 h after IV LPS administration, whereas a mildly tolerated response to S. aureus was observed in monocytes that were isolated 7 days later. Short-term synergistic effects between LPS and (components of) S. aureus have been described previously in primary human leukocytes [35]. This is possibly mediated by an increased recognition of lipoteichoic acid (a cell wall component of Gram-positive bacteria) by TLR2 as transient upregulation of TLR2 expression following exposure to LPS was demonstrated both EV [35] and IV [36]. However, our study also demonstrates that cytokine responses to the TLR2 ligand Pam3Cys were attenuated rather than enhanced in monocytes isolated after IV LPS administration, implying that the enhanced response observed in S. aureus-stimulated cells may not be mediated through the TLR2 pathway. Perhaps other mechanisms, such as increased recognition of lipoteichoic acid by CD14 [37] might contribute to these augmented cytokine responses as CD14 expression on monocytes was also shown to be transiently upregulated after exposure to LPS [35].

Our study has several strengths. First, more than 100 subjects were included. Given the labor-intensive and costly nature of experimental human endotoxemia studies, sample sizes are usually restricted to approximately 10–20 volunteers, and large cohorts such as ours are unique. Second, we used a wide variety of inflammatory stimuli and performed paired IV and EV measurements in all subjects. Finally, in contrast to studies in sepsis patients, our study population is highly homogeneous and is also not affected by confounding factors due to the highly standardized nature of the experimental endotoxemia protocol. Several limitations of this study also must be acknowledged. First, in contrast to the wide variety of compounds and pathogens used for EV stimulations, only LPS was used to induce an IV immune response. This is because experimental human endotoxemia is the only available model that offers the opportunity to study systemic inflammation and tolerance induction IV in humans. As a consequence, the IV immune response in our study was purely TLR4-driven, and other inflammatory signaling pathways might not be adequately reflected in this model. However, it is well established that immune responses to different pathogens, pathogenic stimuli, and also to other inciting events (trauma, burns) show great analogy [6, 38]. This was underscored again by a study showing that gene expression patterns are highly correlated across heterogeneous groups of patients with fecal peritonitis and community-acquired pneumonia, demonstrating that most (sepsis) response pathways are common and independent of the pathogen or stimulus involved [39]. Also, because no correlations were found between the IV, EV, and in vitro cytokine production upon stimulation with LPS, it appears unlikely that robust correlations would have been found upon IV challenge with another stimulus. Second, it must be acknowledged that the inverse associations in our study between the degree of IV inflammation and EV cytokine production capacity were observed on the first LPS challenge day, which bears many similarities to the early phases of sepsis. Therefore, it cannot be ruled out that in later phases of sepsis, the suppressed IV and EV immune status may be correlated. Nevertheless, this may also represent an epiphenomenon due to ongoing immunological stimulation due to active infection in sepsis as opposed to short-lived bolus stimulation in the endotoxemia model, where cytokine levels are only elevated for several hours. Third, EV stimulations in this study were mainly performed in monocytes, implying that potential interactions with other cell types such as (regulatory) T or B cells or neutrophils are not taken into account. However, since cytokine responses of monocytes and PBMCs were strongly correlated, this likely has little impact on the study findings and interpretation.

In conclusion, we demonstrate that EV and in vitro immune-cell stimulation models that are currently widely used do not accurately reflect the IV immune response nor the development of endotoxin tolerance in a large cohort of healthy volunteers undergoing experimental human endotoxemia. Clinicians and researchers must be careful with the interpretation of these leukocyte responses in the context of determining a patient’s immune status, especially during early sepsis, and these measurements should preferably be combined with other biomarkers and clinical parameters to provide a more complete assessment of immune function. Our results also underscore that results from in vitro and EV experiments should be validated IV for proper interpretation.

Statement of Ethics

This study protocol was reviewed and approved by the Local Ethics Committee (Commissie Mensgebonden Onderzoek [CMO] Arnhem-Nijmegen; reference Nos. NL68166.091.18 and 2018-4983). All subjects provided written informed consent prior to participation in this study.

Conflict of Interest Statement

The authors have no conflict of interest to disclose.

Funding Sources

This work was internally funded by the Department of Intensive Care Medicine of the Radboud University Medical Center in Nijmegen, The Netherlands.

Author Contributions

Matthijs Kox, Peter Pickkers, Niklas Bruse, and Aron Jansen designed the experiments. Aron Jansen and Nicole Waalders conducted all human endotoxemia-related procedures. Niklas Bruse, Jelle Gerretsen, and Daniëlle Rijbroek performed all EV and in vitro stimulation experiments. Aron Jansen analyzed and visualized the data and drafted the primary version of the manuscript. Matthijs Kox and Peter Pickkers critically revised the article and supervised the research. All the authors read and revised the final version of this article.

Data Availability Statement

All data used for the analysis in this study are available upon reasonable request to the corresponding author.

This article is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC). Usage and distribution for commercial purposes requires written permission. Drug Dosage: The authors and the publisher have exerted every effort to ensure that drug selection and dosage set forth in this text are in accord with current recommendations and practice at the time of publication. However, in view of ongoing research, changes in government regulations, and the constant flow of information relating to drug therapy and drug reactions, the reader is urged to check the package insert for each drug for any changes in indications and dosage and for added warnings and precautions. This is particularly important when the recommended agent is a new and/or infrequently employed drug. Disclaimer: The statements, opinions and data contained in this publication are solely those of the individual authors and contributors and not of the publishers and the editor(s). The appearance of advertisements or/and product references in the publication is not a warranty, endorsement, or approval of the products or services advertised or of their effectiveness, quality or safety. The publisher and the editor(s) disclaim responsibility for any injury to persons or property resulting from any ideas, methods, instructions or products referred to in the content or advertisements.