The magnitude of the LPS-elicited cytokine response is commonly used to assess immune function in critically ill patients. A suppressed response, known as endotoxin tolerance, is associated with worse outcomes, yet endotoxin tolerance-inducing TLR4 ligands are known to protect animals from infection. Thus, it remains unknown whether the magnitude of the LPS-elicited cytokine response provides an accurate assessment of antimicrobial immunity. To address this, the ability of diverse TLR ligands to modify the LPS-elicited cytokine response and resistance to infection were assessed. Priming of mice with LPS, monophosphoryl lipid A (MPLA), or poly(I:C) significantly reduced plasma LPS–elicited proinflammatory cytokines, reflecting endotoxin tolerance, whereas CpG-ODN–primed mice showed augmented cytokine production. In contrast, LPS, MPLA, and CpG-ODN, but not poly(I:C), improved the host response to a Pseudomonas aeruginosa infection. Mice primed with protective TLR ligands, including CpG-ODN, showed reduced plasma cytokines during P. aeruginosa infection. The protection imparted by TLR ligands persisted for up to 15 d yet was independent of the adaptive immune system. In bone marrow–derived macrophages, protective TLR ligands induced a persistent metabolic phenotype characterized by elevated glycolysis and oxidative metabolism as well as augmented size, granularity, phagocytosis, and respiratory burst. Sustained augmentation of glycolysis in TLR-primed cells was dependent, in part, on hypoxia-inducible factor 1-α and was essential for increased phagocytosis. In conclusion, the magnitude of LPS-elicited cytokine production is not indicative of antimicrobial immunity after exposure to TLR ligands. Additionally, protective TLR ligands induce sustained augmentation of phagocyte metabolism and antimicrobial function.
Endotoxin tolerance is the immunological phenomenon wherein prior exposure to TLR ligands renders innate leukocytes refractory to LPS (also known as endotoxin)-elicited proinflammatory cytokine secretion (1). Endotoxin tolerance is classically induced by exposure to LPS, and is mediated by inducible inhibitors of TLR signaling, such as IRAK-M and SHIP-1, which suppress proinflammatory gene product expression (2, 3). This immunoregulatory process has received widespread attention from the scientific community regarding both the molecular mechanism of induction and clinical significance (4). Yet, whereas the molecular mechanisms of induction are largely understood, the clinical significance remains debated.
There is a body of literature suggesting that endotoxin tolerance represents a state of immunoparalysis and susceptibility to infection (4–6). In support of this hypothesis, an endotoxin tolerance gene signature has been associated with worse patient outcomes (7, 8), and a high incidence of secondary nosocomial infections in critically ill patients (9). Concurrent with this suggestion, an alternative body of literature reports that treatment with TLR4 ligands such as LPS and monophosphoryl lipid A (MPLA), the prototypical ligands used to induce endotoxin tolerance, augments host resistance to infection while facilitating suppressed cytokine production in a variety of infection models (10–12). Although seemingly in contradiction, these hypotheses both assume that the magnitude of the LPS-elicited cytokine response dictates the host response to infection. As such, the LPS-elicited proinflammatory cytokine response is commonly used to assess innate immune function in research settings (13, 14). Despite this, it has not been sufficiently demonstrated that the LPS-elicited cytokine response provides an accurate assessment of the host response to live infection. One novel hypothesis to resolve this contradiction is that the magnitude of the LPS-elicited cytokine response is not relevant to, and thus does not predict, the host response to infection.
Our previous work has consistently demonstrated that TLR ligand-induced immunoprophylaxis requires improved antimicrobial functions. This finding prompts the hypothesis that innate antimicrobial efficiency, rather than the LPS-elicited cytokine response, determines infection outcomes (12, 15). However, the processes that underlie persistent TLR-mediated changes in leukocyte antimicrobial function are unclear. One cellular process that may underlie the ability of TLR ligands to augment cellular antimicrobial functions is metabolic reprogramming, which can rapidly change to support specific innate immune cell needs. Macrophages and dendritic cells are known to shift their metabolic profile to one that favors glycolysis over oxidative metabolism, a state known as aerobic glycolysis, immediately following TLR stimulation to package and secrete cytokines (16). In contrast, anti-inflammatory macrophages exhibit a metabolic program of augmented oxidative phosphorylation (17). How metabolic programming changes as macrophages adapt to priming with TLR ligands has not been explored in depth.
Endotoxin tolerance can be induced by TLR ligands other than LPS, and some TLR ligands have been reported to induce endotoxin sensitization, or augmented LPS-elicited cytokine production (18). This diversity among TLR ligands offers a novel strategy, which we use in this study, to dissect whether an altered LPS-elicited cytokine response impacts host responses to infection. We hypothesize that the LPS-elicited proinflammatory cytokine response does not predict the host response to infection following exposure to TLR ligands. To address our hypothesis, we examined the ability of various TLR ligands to alter the LPS-elicited proinflammatory cytokine response and to alter antimicrobial immunity against Pseudomonas aeruginosa, one of the most common nosocomial gram-negative pathogens (19). We demonstrate that the magnitude of the LPS-elicited cytokine response does not correlate with host resistance to infection, and provide evidence that TLR immunoprophylaxis can occur without endotoxin tolerance. Additionally, we demonstrate that endotoxin tolerance is associated with a unique metabolic phenotype in macrophages that facilitates persistent improvements in antimicrobial function.
Materials and Methods
Mouse model of infection
All animal procedures were consistent with the National Institutes of Health guidelines for the Care and Use of Experimental Animals, and were approved by the Institutional Animal Care and Use Committee at Vanderbilt University. Male 8- to 10-wk old BALB/c mice were purchased from Envigo Laboratories (Indianapolis, IN). Male wild-type and RAG2−/− C57/Bl6 mice were purchased from the Jackson Laboratory (Bar Harbor, ME). LysMcre-HIF-1αfl/fl mice were received as a gift from Dr. H. Eltzschig (University of Texas at Houston). P. aeruginosa was purchased from American Type Culture and Collection (ATCC 19660, Manassas, VA). The culture was grown in tryptic soy broth and resuspended in sterile saline. For i.p. infection, mice were inoculated with 1 × 108 CFU P. aeruginosa diluted in lactated Ringer’s solution (LR). At 6 h after inoculation, the peritoneal cavity was lavaged with 2 ml of sterile PBS. Peritoneal lavage fluid was cultured on tryptic soy agar overnight and colonies were counted to determine CFU per milliliter. For characterization of peritoneal leukocytes, cells in the peritoneal lavage were counted, centrifuged (300 × g for 10 min at 4°C), and resuspended in PBS at desired cell concentrations for flow cytometry.
TLR ligand treatment
MPLA derived from Salmonella enterica serotype Minnesota Re 595, was purchased from Sigma-Aldrich (St. Louis, MO) or InvivoGen (San Diego, CA). MPLA was solubilized in 0.2% triethylamine solution or DMSO. Ultrapure LPS derived from Escherichia coli 0111:B4, synthetic ODN 1826 (CpG-ODN), synthetic ODN 2138 (GpC-ODN), and high m.w. poly(I:C) were purchased from InvivoGen and solubilized in sterile saline. For treatment, all TLR ligands were diluted in LR (100 μg/ml) and administered by i.p. injection at 20 μg in 0.2 ml, except LPS, which was given at 2 μg in 0.2 ml. Vehicle-treated mice received i.p. injection of 0.2 ml LR.
In vivo plasma cytokine measurements
Whole blood was harvested by carotid artery laceration under isoflurane anesthesia and collected in heparinized syringes. Concentrations of IL-6, IFN-γ, IL-1β, and KC in plasma were measured using a Bio-Plex Multiplex Bead Array and read with the Bio-Plex Magpix Multiplex Reader (Bio-Rad, Hercules, CA).
Ex vivo plasma cytokine measurements
Whole blood was harvested by carotid artery laceration under isoflurane anesthesia and collected in heparinized syringes. Whole blood was mixed 1:1 with RPMI 1640 with glutamine (Life Technologies, Carlsbad, CA) containing 100 ng/ml LPS. After 6 h, cells were centrifuged and plasma was collected. TNF-α concentrations were measured using a Bio-Plex Bead Array and read with the Bio-Plex Magpix Multiplex Reader (Bio-Rad).
Intraperitoneal leukocytes were suspended in cold PBS (1 × 107 cells/ml), incubated with anti-mouse CD 16/32 (1 μl/ml; eBioscience, San Diego, CA) for 5 min to block nonspecific Fc receptor-mediated Ab binding, then with fluorochrome-conjugated Abs or isotype control Abs (0.5 μg/106 cells/0.1 ml) at 4°C for 30 min. Samples were washed and resuspended in 250 μl cold PBS and run immediately on an Accuri C6 flow cytometer (BD Biosciences, San Diego, CA). Data were analyzed using Accuri C6 software. Abs used for these studies included anti-F4/80-FITC (clone BM8; eBioscience), anti-Ly6G-PE (clone 1A8; BD Biosciences), anti-Ly6C, and appropriate isotype controls (eBioscience, BD Biosciences). Neutrophils are identified as Ly6G+F4/80− cells and macrophages are identified as Ly6G−F4/80+ cells.
Bone marrow–derived macrophages
Femurs and tibias were harvested from male C57/Bl6 mice, and flushed with 10 ml cold PBS to obtain bone marrow cells. Bone marrow cells were incubated in RPMI 1640 with glutamine and 25 mM HEPES (Life Technologies) containing 10% certified FBS (Life Technologies), 1% antibiotic-antimycotic (Life Technologies), and 10 ng/ml mouse recombinant M-CSF (R&D Systems, Minneapolis, MN) for 7 d. At the initiation of the growth period, bone marrow cell concentration was 4 × 104/ml, as this has been found to yield a mature, pure population of macrophages (20). After 7 d, macrophages were washed with PBS and incubated with media containing various TLR ligands for 24 h. Macrophages were then washed thoroughly with PBS and rested in media for 3 d before assessment.
In vitro cytokine measurements
Bone marrow–derived macrophages (BMDMs) were plated in 96-well plates at 5 × 105 cells/ml and stimulated with 100 ng/ml LPS or 1 × 106 CFU/ml of heat-killed P. aeruginosa for 24 h. Prior to the assay, P. aeruginosa was heat killed by incubation at 65°C for one hour. Supernatant was harvested 24 h after LPS. IL-6 concentration in the media was determined by IL-6 ELISA (eBioscience).
Assessment of macrophage metabolism
BMDMs were plated in a 96-well Seahorse assay plate at 5 × 105 cells/ml in Seahorse Assay Media and assessed on the Seahorse XFe96 Extracellular Flux Analyzer (Agilent Technologies, Santa Clara, CA). For the glycolysis stress test assay, cells were sequentially treated with 10 mM glucose (RPI, Mount Prospect, IL), 1 μM oligomycin (Agilent Technologies), and 50 mM 2-deoxyglucose (Sigma-Aldrich).
BMDMs plated at 5 × 105/ml were lysed with Luciferase Cell Culture Lysis Reagent (Promega, Madison, WI). ATP concentration in the lysate was measured using the luminescent ATP Determination Kit purchased from Invitrogen (Carlsbad, CA). Luminescence was determined by the Bioek Synergy MX plate reader (Biotek, Winooski, VT).
BMDMs were incubated with pHrodo-tagged E.coli bioparticles purchased from Life Technologies, suspended in RPMI 1640 with glutamine and without phenol red (Life Technologies), for 1.5 h to allow for phagocytosis to occur. Following incubation, BMDMs were washed with PBS and pHrodo mean fluorescence intensity (MFI) was determined via flow cytometry.
Respiratory burst assay
BMDMs were assessed using the Respiratory Burst Assay (Cayman Chemical Company, Ann Arbor, MI). In brief, BMDMs were incubated with dihydrorodamine-123, which fluoresces in the presence of reactive oxygen species. Reactive oxygen species were then elicited by incubation with 200 nM PMA for 45 min. After stimulation, cells were washed with PBS and rhodamine MFI was determined via flow cytometry.
All data were analyzed with GraphPad Prism 7 (La Jolla, CA). Data from multiple group experiments were analyzed using one-way ANOVA followed by Dunnett multicomparison post hoc test. Due to significant heteroscedasticity intrinsic to the Bio-Plex cytokine data as identified by a Levene test of equality of variances with p < 0.0001, all in vivo cytokine data derived from the Bio-Plex were log-transformed prior to statistical comparison. All data values are presented as mean ± SEM, except for bacterial counts and body temperature, for which mean values alone are designated. A p value ≤0.05 was considered statistically significant.
TLR ligands differentially alter the LPS-elicited proinflammatory cytokine response
To determine how different TLR ligands alter the proinflammatory cytokine response to LPS, mice were injected with either the TLR4 ligands LPS or MPLA, the TLR9 ligand CpG-ODN, the TLR3 ligand poly(I:C), or LR (vehicle) via the i.p. route (Fig. 1A). GpC-ODN was used as an additional control. Three days later, mice were challenged with LPS. Six hours after LPS challenge, plasma proinflammatory cytokine concentrations were measured. Mice primed with LPS or MPLA demonstrated significantly reduced plasma IL-6, IFN-γ, IL-1β, and KC concentrations after LPS challenge, and mice primed with poly(I:C) demonstrated reduced IFN-γ, with a trend toward reduced IL-6 and KC (Fig. 1B), as compared with vehicle control. In contrast, CpG-ODN–primed mice demonstrated significantly higher plasma concentrations of IL-6, IFN-γ, and IL-1β compared with vehicle control. Priming mice with GpC-ODN did not significantly alter LPS-induced cytokine production. To use a clinically feasible assessment of LPS responses, as well as to detect TNF-α, which clears from the plasma within 2 h after in vivo LPS, blood was harvested from TLR-ligand primed mice and exposed to LPS for 6 h ex vivo followed by measurement of TNF-α concentrations in the plasma supernatant fraction. The supernatant fraction from MPLA-, LPS-, and poly(I:C)-primed mice had reduced TNF-α concentrations compared with control, whereas CpG-ODN primed blood was found to have elevated TNF-α (Fig. 1C). Thus, various TLR ligands differentially alter the response to endotoxin in vivo and ex vivo. The TLR4 ligands LPS and MPLA potently induce sustained endotoxin tolerance whereas poly(I:C) weakly elicits endotoxin tolerance. In contrast, CpG-ODN potently augments LPS-induced cytokine production.
The magnitude of the LPS-elicited proinflammatory cytokine response does not indicate infection susceptibility
To determine the ability of TLR ligands to change the host susceptibility to live P. aeruginosa infection, mice were treated with LPS, MPLA, GpC-ODN, CpG-ODN, poly(I:C), or LR (vehicle), and 3 d later were challenged with P. aeruginosa (Fig. 2A). Core temperature, bacterial counts, leukocyte numbers at the site of infection, and plasma cytokines were measured 6 h after P. aeruginosa challenge. Core body temperature provides an accurate surrogate for physiologic integrity during infection because mice reliably develop hypothermia as the severity of infection progresses (21). Mice primed with LPS, MPLA, or CpG-ODN were robustly resistant to infection, as indicated by maintenance of normal core body temperature (Fig. 2B), whereas mice primed with LR or GpC-ODN showed dramatic reductions in core body temperature. Core body temperature in mice primed with poly(I:C) was statistically higher than in vehicle-primed mice but significantly lower than in non-infected mice or mice primed with LPS, MPLA, or CpG-ODN. Mice primed with LPS, MPLA, or CpG-ODN demonstrated a significant reduction in i.p. P. aeruginosa (Fig. 2C), whereas mice primed with GpC-ODN, poly(I:C), or vehicle did not. Overall, the magnitude of the LPS-elicited cytokine proinflammatory response did not correlate with susceptibility to infection, as ligands that increase (CpG-ODN) and decrease (MPLA, LPS) LPS-elicited cytokine responses both improved the host response to P. aeruginosa infection.
Protective TLR ligands improve the recruitment of phagocytes
Next, the cellular mechanisms by which TLR ligands generate protection from infection were investigated. Given that mice resistant to infection had reduced bacterial burden, the recruitment of neutrophils and macrophages to the site of infection was characterized. Six hours after i.p. P. aeruginosa infection, the peritoneal cavity of infected mice was lavaged with PBS and leukocytes in the peritoneal lavage fluid were characterized using flow cytometry. Both the total number and percentage of neutrophils were significantly elevated in LPS-, MPLA-, and CpG-ODN–primed mice as compared with LR-primed controls (Fig. 2D, 2E). Poly(I:C)-primed mice demonstrated an increase in percentage, but not total number, of neutrophils recruited to the peritoneal cavity. Additionally, a significant increase in i.p. macrophage numbers and percentage was observed after priming with MPLA and CpG-ODN compared with LR-primed controls. Priming with LPS, GpC-ODN, or poly(I:C) did not elicit a significant increase in the total number or percentage of macrophages compared with LR-primed controls. Resident peritoneal macrophages are terminally differentiated and unable to proliferate rapidly, thus an elevation in total peritoneal macrophage numbers indicates these macrophages are derived from bloodstream monocytes. In support of that contention, the majority of MPLA-recruited F4/80+ cells coexpressed the monocyte-associated surface marker Ly6C (Supplemental Fig. 1). These data indicate that persistent improvements in the recruitment of phagocytes to the site of infection are a likely mechanism by which many TLR ligands exert their protective effect, a finding that is backed by our previous studies with MPLA (10).
Systemic proinflammatory cytokine concentrations correlate with bacterial burden
Plasma proinflammatory cytokine concentrations at 6 h following P. aeruginosa infection were quantified (Fig. 2F). Plasma concentrations of IL-6, IFN-γ, IL-1β, and KC were significantly reduced in the plasma of mice primed with LPS or MPLA compared with LR-primed animals. Even though CpG-ODN increased plasma proinflammatory cytokines after LPS challenge, CpG-ODN–primed mice had a significant reduction in plasma IL-6, IL-1β, and KC, but not IFN-γ, after P. aeruginosa infection. These data indicate that TLR ligands that augment bacterial clearance will facilitate reduced systemic proinflammatory cytokines during infection, even if they sensitize the cytokine response to endotoxin.
TLR4 ligand-induced immunoprophylaxis persists for up to 15 d but is independent of the adaptive immune system
To determine how long TLR immunoprophylaxis can be sustained, mice were challenged with P. aeruginosa 1, 10, and 15 d following MPLA priming. Mice infected 10 d following MPLA had significantly higher core body temperature than vehicle controls and demonstrated significant reductions in bacterial burden (Fig. 3). Mice infected 15 d following MPLA had improvements in core temperature over controls, but did not show statistically significant improvements in bacterial clearance.
There are no established mechanisms for the innate immune system to sustain a phenotype 2 wk after a stimulus. Although we have previously published on the requirement of the innate immune system for TLR ligands to protect animals (10), the importance of the adaptive immune system for protection is unknown. Therefore, we investigated the hypothesis that the adaptive immune system mediates the persistent protection and improved phagocyte recruitment seen after TLR ligand priming. Wild-type and RAG2 knockout (RAG2−/−) mice, which are deficient in functional T and B cells, were primed with MPLA or LR (vehicle) and challenged 3 d later with P. aeruginosa (Fig. 4A). MPLA-primed RAG2−/− mice maintained body temperature and demonstrated reduced bacterial burden in response to infection, as compared with LR-primed RAG2−/− mice (Fig. 4B, 4C). Additionally, MPLA-primed RAG2−/− mice demonstrated a significant elevation in both total and percent of neutrophils in the peritoneal cavity as well as total and percent of macrophages, compared with LR-primed RAG2−/− mice (Fig. 4D, 4E). Similar to wild-type animals, proinflammatory cytokine levels in MPLA-primed RAG2−/− mice were significantly lower than in controls (Fig. 4F). Thus, the adaptive immune system is not required to maintain TLR4-mediated augmentation of host resistance to infection.
Endotoxin tolerant macrophages exhibit persistent alterations to glycolytic and oxidative metabolism
Our in vivo data suggest that the innate immune system can sustain an immunoprotective phenotype despite endotoxin tolerance. To explore this phenomenon on a cellular level, we examined the function of BMDMs following exposure to TLR ligands (Fig. 5A). MPLA-, LPS-, CpG-ODN–, and poly(I:C)-primed BMDMs exhibited reduced IL-6 secretion in response to LPS, reflective of endotoxin tolerance (Fig. 5B). A similar tolerance profile was noted after exposure to heat-killed P. aeruginosa (Fig. 5C). Interestingly, CpG-ODN elevated LPS-elicited cytokine responses in vivo and ex vivo (Fig. 1), but suppressed cytokine responses in macrophage cultures in vitro. This finding has been noted in previous studies, and is likely due to profound potentiation of IFN-γ secretion by CpG-ODN in vivo, which can eliminate endotoxin tolerance (22, 23). To assess broader macrophage activity, macrophage glycolytic metabolism, which is required for a variety of basal myeloid cell inflammatory functions (24), was examined. MPLA-, LPS-, CpG-ODN–, and poly(I:C)-primed BMDMs displayed an elevated glycolytic rate (Fig. 5D, 5E).
Macrophages actively responding to TLR ligands, commonly called M1 macrophages, have been shown to adopt a metabolic phenotype of aerobic glycolysis, which is an increase in glycolysis with a decrease in oxygen consumption. To determine whether this phenotype was also found in TLR-primed macrophages, basal oxygen consumption rate during the glycolysis stress test was determined immediately after glucose addition. Surprisingly, we found that TLR-primed macrophages had a significantly elevated basal oxidative rate that correlated with the observed increases in glycolysis (Fig. 5F). Increases in glycolysis and oxidative metabolism suggest the cells may be producing greater amounts of ATP. Indeed, BMDMs primed with protective TLR ligands produced a significantly greater amount of ATP (Fig. 5G).
Protective TLR ligands improve the antimicrobial functions of macrophages
To probe macrophage functions beyond cytokine secretion, we assessed macrophages on a variety of antimicrobial endpoints. Alterations to macrophage morphology indicate gross alterations to function, such as changes to phagocytosis or other antimicrobial functions (25). Using flow cytometry, we found that MPLA-, LPS-, CpG-ODN–, and poly(I:C)-primed macrophages were significantly larger and more granular than unprimed macrophages (Fig. 6A, 6B). We then assessed BMDMs on bacterial phagocytosis and respiratory burst, two key antimicrobial processes. MPLA-, LPS-, and CpG-ODN–primed macrophages showed increased phagocytosis of pHrodo-labeled E. coli bioparticles (Fig. 6C), as well as increased PMA-elicited respiratory burst (Fig. 6D).
Increased glycolytic metabolism facilitates MPLA-mediated improvements in phagocytosis and is dependent, in part, on hypoxia-inducible factor 1-α
To determine whether the observed alterations in metabolism influence antimicrobial function, we assessed the requirement of glycolysis for macrophage phagocytosis. MPLA-primed BMDMs were incubated with 2-deoxyglucose (2-DG) during phagocytosis of pHrodo-tagged E. coli bioparticles. In BMDMs 2-DG rapidly inhibits glycolysis (Fig. 4D) and it induced a significant reduction in phagocytosis in control and MPLA-primed BMDMs (Fig. 7A), although phagocytosis in MPLA-primed BMDMs remained higher than unprimed BMDMs. Hypoxia-inducible factor 1α (HIF-1α) is a transcription factor known to coordinate TLR-induced increases in glycolysis (17). To assess the role of HIF-1α in TLR-specific increases in glycolysis and phagocytosis, HIF-1α knockout BMDMs were derived from LysMcre-HIF-1αfl/fl mice and primed with MPLA. Indeed, HIF-1α−/− BMDMs had a significantly reduced glycolytic rate following MPLA priming compared with wild-type BMDMs (Fig. 7B). Further, MPLA-primed HIF-1α−/− macrophages had significantly reduced phagocytosis when compared with MPLA-primed wild-type BMDMs, although phagocytosis remained significantly higher than unprimed BMDMs (Fig. 7C).
Taken together, these data support the argument that the LPS-elicited cytokine response is not indicative of susceptibility to infection, especially after prior exposure to TLR ligands. We discovered that many TLR ligands persistently alter the LPS-elicited proinflammatory cytokine response, but that both decreased (LPS, MPLA) and increased (CpG-ODN) LPS-elicited cytokine responses occurred concurrent with resistance to infection. To our knowledge, this is the first study to identify an incongruity between the magnitude of the cytokine response to LPS and the response to live infection. This finding, although simple, may have significant implications. It has previously been assumed that the cytokine response to LPS is predictive of the response to infection, because investigators continue to use LPS-induced cytokine shock as a surrogate for bacterial sepsis (26, 27), and the LPS cytokine response assay to determine immune-competency in human patients (13, 28). Moreover, the conclusion that a reduced LPS-elicited cytokine response is immunosuppressive continues to receive wide attention (5, 29). This study suggests that the magnitude of the LPS-elicited cytokine response should not be used as the sole measure of immune function, and its value should be questioned.
Our finding that LPS and MPLA protect against infection reiterates our previous conclusion that endotoxin tolerance is not a state of immunoparalysis. Further supporting this conclusion, we found that MPLA- and LPS-primed macrophages have augmented antimicrobial functions. Although the expression of an endotoxin tolerance gene signature has been associated with worse outcomes in critically ill patients (8), the increased expression of these genes may simply correlate with the severity of the primary inflammatory insult (3, 30). Severe inflammation can lead to a myriad of pathologies, including widespread organ and endothelial cell dysfunction, T cell apoptosis, metabolic instability, and hypothalamic dysregulation, each of which is also associated with worse patient outcomes (31–34). Whether a decreased LPS-elicited cytokine response is an underlying cause of increased infection rates in critically ill patients warrants more careful study. Even if an association is found, other causes of impaired antimicrobial immunity, most notably T cell loss and dysfunction, may be present in critically ill patients. Loss and dysfunction of T cells has been widely reported in critically ill humans and in animal models of critical illness (35–37). Animal models of critical illness demonstrate that T cell dysfunction directly contributes to impaired antimicrobial immunity (38, 39). Thus, although an LPS tolerant phenotype may be present in patients with severe sepsis and other critical illnesses, a cause and effect relationship has not been developed and other mechanisms of immune dysfunction may contribute to impaired antimicrobial immunity.
Our finding that CpG-ODN elevates in vivo LPS-elicited cytokine responses yet still protects mice from P. aeruginosa infection disputes the hypothesis that endotoxin tolerance is immunoprotective and the mechanism underlying TLR ligand immunoprophylaxis. Further supporting this conclusion, poly(I:C) was able to induce mild endotoxin tolerance, yet was unable to induce infection prophylaxis. Mice protected by a reduced cytokine burden alone do not typically display concurrent increases in bacterial clearance and phagocyte recruitment (40), yet MPLA- and LPS-primed mice exhibited improved bacterial clearance and phagocyte recruitment. Our previous work has found that MPLA induces robust neutrophil and monocyte proliferation and mobilization (41), and that G-CSF, and the accompanying antimicrobial response, is essential for MPLA to reduce cytokine levels in response to P. aeruginosa (42). Thus, our data continue to support the hypothesis that TLR ligands protect animals from infection by improving host antimicrobial efficiency, rather than by altering host LPS-elicited proinflammatory cytokine responses. Live bacteria elicit antimicrobial responses, which LPS does not, that may underlie the reduced cytokines levels in CpG-ODN–primed mice. It is important to consider that TLR ligands can change cytokine responses to a variety of microbial products. Therefore, it is possible that P. aeruginosa contains microbial products beyond LPS that CpG-ODN–primed mice are tolerant to, making it difficult to completely rule out the impact of cross-tolerance on TLR prophylaxis.
Altogether, our in vivo and in vitro work strongly suggests that a reduced LPS-elicited cytokine response does not necessarily entail immunoparalysis or immunoresistance, prompting the conclusion that the LPS-elicited cytokine response does not predict the host response to infection. We have consistently found that TLR ligands protect animals by improving the host antimicrobial response, although it was previously unknown how protective TLR ligands persistently alter phagocyte function to achieve protection. Upon a metabolic assessment of TLR-primed macrophages, we found that TLR ligand priming induces a persistent and unique metabolic phenotype. Namely, we found that macrophages primed with protective TLR ligands are more glycolytic, more oxidative, and produce more ATP. This metabolic programming was surprising as we concurrently observed these cells to be refractory to LPS-elicited cytokine secretion. Our observation of increased glycolysis and oxidative metabolism does not fit into previously defined metabolic classifications of M1 or M2 macrophages. Rather, these metabolic findings demonstrate that MPLA-primed endotoxin tolerant macrophages have a metabolic phenotype reflecting features of both M1 (increased glycolytic rate) and M2 (increased oxidative rate) macrophages. This finding is in concert with data from transcriptional studies of endotoxin-tolerant macrophages that demonstrate that these cells express a unique transcriptional profile with features of both M1 and M2 macrophages, although predominantly express M2 features (43, 44). It is possible that an elevated glycolytic rate is not specific to M1 macrophages, and is also used by endotoxin-tolerant macrophages to facilitate improved antimicrobial function. As such, the identification of this metabolic phenotype raises many questions regarding the purpose of TLR-mediated metabolic reprogramming in endotoxin tolerant macrophages.
Products of glycolysis have been associated with improved macrophage phagocytosis and respiratory burst functions (45, 46). Thus, our metabolic data suggest that functions other than cytokine secretion, particularly antimicrobial functions, may be augmented in TLR ligand-primed macrophages. We found that TLR ligand-primed macrophages have increased size and granularity, as well as improvements in phagocytosis and respiratory burst. Hence, these data show that endotoxin tolerance does not entail complete functional anergy, but rather that endotoxin tolerant macrophages undergo sustained adaptations to improve their antimicrobial capacity while suppressing cytokine responses. Interestingly, although poly(I:C)-primed macrophages showed small increases in glycolysis, size, and granularity, they did not show improvements in phagocytosis or respiratory burst. MPLA, LPS, and CpG-ODN each activate the TLR signaling adaptor MyD88, whereas poly(I:C) only activates the TLR signaling adaptor Toll/IL-1R domain-containing adapter inducing IFN-β (47), thus, activation of MyD88-dependent signaling may be required to completely induce an improved antimicrobial phenotype.
We found that inhibition of glycolysis among TLR-primed macrophages caused impairment of phagocytosis. This finding demonstrates that the unique metabolic programming caused by TLR agonist treatment may directly support improvements in antimicrobial function. Indeed, HIF-1α−/− macrophages, which were partially deficient in TLR-induced glycolysis, were unable to increase phagocytosis as much as wild-type macrophages. There is only limited knowledge regarding the purpose of TLR-mediated metabolic reprogramming in macrophages for functions other than cytokine secretion. In this study, we find a unique role for glycolysis in cells that are refractory to cytokine secretion. It remains to be determined exactly how TLR-mediated glycolytic metabolism influences phagocytosis, and how the observed increase in oxidative metabolism and ATP generation changes antimicrobial function. Glycolysis directly supports the production of cellular building blocks, including fatty acids that are essential for monocytes to differentiate into phagocytic macrophages (48, 49). Hence, TLR-induced alterations in glycolysis may support antimicrobial processes via these anabolic processes. Additionally, it is possible that increased production of ATP in TLR ligand-primed macrophages provides energy for improvements in antimicrobial function, although the specific role for altered TLR-mediated metabolic reprogramming remains to be studied.
In concert with literature reports of increased human monocyte antimicrobial activity during endotoxin tolerance (50–52), these data support the proposal that endotoxin tolerance can be thought of as cellular reprogramming rather than immunoparalysis or immunosuppression (53). Combined with the lack of predictability of the LPS-elicited cytokine response, these findings strongly suggest that future research on macrophage immunomodulation include additional assessments of macrophage antimicrobial function beyond the LPS-elicited cytokine response.
The duration of TLR4 ligand-induced immunoprophylaxis had not been previously determined. The finding that MPLA protects RAG2−/− mice and that MPLA can protect mice for at least 10 d after a single injection suggests the innate immune system has yet undiscovered mechanisms to sustain improved antimicrobial responses. Supporting this hypothesis, our macrophage studies were all conducted 3 d following the clearance of TLR ligands in vitro, suggesting that the metabolic and functional impact of TLR ligands is persistent. Other groups that have found similar glycolytic phenotypes in macrophage-rich tissues during endotoxin tolerance (54). These results are surprising considering that there are no established mechanisms for the innate immune system to sustain any phenotype for a time period long after a stimulus has been cleared. However, such mechanisms may lie in persistent epigenetic alterations, as has recently been observed following LPS administration, and in non-TLR–ligand immunoprophylaxis (55, 56). Further work on TLR ligand-induced persistent epigenetic modifications will likely discover novel ways for the innate immune system to sustain phenotypes long after a stimulus has been cleared.
Finally, many therapeutic trials in sepsis have focused on altering proinflammatory cytokine levels during infection, notably trials with anti-TNF-α, IL-1 receptor antagonist, and a TLR4 antagonist, all of which have been met with failure (57–59). In contrast, the findings presented in this study suggest that improving the host antimicrobial response to infection may be a more effective way to reduce systemic inflammation and resolve sepsis. Hence, these data support clinical trials with immunostimulatory compounds that can improve microbicidal function in immune cells. Ongoing studies of this nature include trials with IL-7, which aim to reverse T cell anergy observed during postsepsis immunosuppression (35), as well as trials with GM-CSF (60). Our findings support the initiation of clinical trials with MPLA or CpG-ODN to improve antimicrobial efficiency in anticipation of sepsis in critically ill or immunocompromised patients. More accurate assessments of immune function, combined with therapies that boost antimicrobial processes, may effectively mitigate the risk of infection in hospitalized patients.
The Agilent Seahorse Extracellular Flux Analyzer is housed and managed within the Vanderbilt High-Throughput Screening Core Facility (an institutionally supported core) and was funded by National Institutes of Health Shared Instrumentation Grant 1S10OD018015.
This work was supported by National Institutes of Health/National Institute of General Medical Sciences Grants R01 GM121711 (to J.K.B., principal investigator), R01 GM104306 (to E.R.S., principal investigator), and T32 GM007347 (to B.A.F.) and by the Vanderbilt University Medical Center Faculty Scholars Award (to J.K.B., principal investigator).
The online version of this article contains supplemental material.
The authors have no financial conflicts of interest.