Activation of Hepatic STAT3 Maintains Pulmonary Defense during Endotoxemia

Abstract

Pneumonia and infection-induced sepsis are worldwide public health concerns. Both pathologies elicit systemic inflammation and induce a robust acute-phase response (APR). Although APR activation is well regarded as a hallmark of infection, the direct contributions of liver activation to pulmonary defense during sepsis remain unclear. By targeting STAT3-dependent acute-phase changes in the liver, we evaluated the role of liver STAT3 activity in promoting host defense in the context of sepsis and pneumonia. We employed a two-hit endotoxemia/pneumonia model, whereby administration of 18 h of intraperitoneal lipopolysaccharide (LPS; 5 mg/kg of body weight) was followed by intratracheal Escherichia coli (10⁶ CFU) in wild-type mice or those lacking hepatocyte STAT3 (hepSTAT3^−/−). Pneumonia alone (without endotoxemia) was effectively controlled in the absence of liver STAT3. Following endotoxemia and pneumonia, however, hepSTAT3^−/− mice, with significantly reduced levels of circulating and airspace acute-phase proteins, exhibited significantly elevated lung and blood bacterial burdens and mortality. These data suggested that STAT3-dependent liver responses are necessary to promote host defense. While neither recruited airspace neutrophils nor lung injury was altered in endotoxemic hepSTAT3^−/− mice, alveolar macrophage reactive oxygen species generation was significantly decreased. Additionally, bronchoalveolar lavage fluid from this group of hepSTAT3^−/− mice allowed greater bacterial growth ex vivo. These results suggest that hepatic STAT3 activation promotes both cellular and humoral lung defenses. Taken together, induction of liver STAT3-dependent gene expression programs is essential to countering the deleterious consequences of sepsis on pneumonia susceptibility.

INTRODUCTION

Sepsis is a complex immunopathological syndrome defined by the systemic inflammatory response to infection and is a leading contributor to morbidity and mortality in intensive care units as evidenced by approximately 750,000 cases per year (2% of all hospital admissions) (1–3). This multifaceted, systemic inflammatory response can be further complicated by organ dysfunction (severe sepsis) and hypotension (septic shock), all of which lead to a complex, variable syndrome with mortality rates between 30 and 50% (4). While pneumonia is the leading cause of sepsis, with about one-half of all sepsis cases originating as respiratory infections (2), sepsis also greatly increases a patient's subsequent susceptibility to bacterial pneumonia (5). In fact, 10 to 30% of mechanically ventilated, septic shock patients develop ventilator-associated pneumonia (6). This positive association extends beyond ventilator-related circumstances and has been corroborated experimentally by multiple studies demonstrating deleterious effects of sepsis and/or endotoxemia on pneumonia outcomes (7–15). With the rapid increase in prevalence of drug-resistant pathogens and the limited treatment options available, there is a growing need to develop novel pharmaceutical interventions and to improve our understanding of the inflammatory processes involved in both pathologies.

A shared and prominent feature of sepsis, pneumonia, and other inflammatory conditions is the hepatic acute-phase response (APR) (16–19). Induced by the host defense cytokines tumor necrosis factor alpha (TNF-α), interleukin-1 (IL-1), and IL-6 (20), the APR is characterized by significant changes in circulating levels of acute-phase proteins (APPs) (19, 21, 22). While it is well appreciated that sepsis can cause pulmonary immunosuppression and pneumonia susceptibility (5, 7–15), it is unclear whether or how preexisting liver activation (i.e., sepsis-induced APR) modulates subsequent responses to local lung infections.

Signal transducer and activator of transcription-3 (STAT3) is one of two transcription factors (along with NF-κB RelA) required for induction of a strong hepatic APR during pneumonia (23–25). Several lines of evidence implicate beneficial roles for liver-derived APPs during pneumonia (26–29). We have shown that this lung-liver axis, enabled by both transcription factors, is required for maximal protection during pneumonia alone, but the distinct roles of STAT3 (versus RelA) in this process remain unclear. Others have linked hepatic STAT3 activity to the APR in models of sepsis (30, 31). Given the close association between pneumonia, sepsis, STAT3, and the APR, we sought to determine the direct influence of systemic STAT3-dependent liver activity on subsequent pneumonia outcomes. Our results demonstrate the significance of liver activation during sepsis and that the APR drives protective networks of gene expression to maximize local defense responses to pulmonary pathogens encountered during endotoxemia.

MATERIALS AND METHODS

Mice.

Experiments were performed using mice in which hepatocyte STAT3 was functionally deleted using the Cre-LoxP system. Briefly, mice containing homozygous floxed alleles for Stat3 (32) were bred with albumin-driven Cre-recombinase transgenic mice (Alb-Cre^tg/−/Stat3^LoxP/LoxP). Experimental results obtained from hepSTAT3^−/− mice were compared to those from littermate controls lacking the Cre-recombinase transgene (Alb-Cre^−/−/Stat3^LoxP/LoxP). Male and female mice on a mixed genetic background were used between 6 and 12 weeks of age, and each experiment was performed at least twice. All animal protocols were approved by the Boston University Institutional Animal Care and Use Committee.

Experimental endotoxemia and pneumonia.

Mice were given an intraperitoneal (i.p.) injection of 5 mg/kg of body weight of ultrapure lipopolysaccharide (LPS; InvivoGen) or saline, followed 18 h later by an intratracheal (i.t.) infection with 1 × 10⁶ CFU of Escherichia coli (Fig. 1A) (serotype 06:K2:H1; ATCC 19138), as previously described by our laboratory (23). For i.t. instillations, mice were anesthetized by i.p. injection of a mixture of ketamine (50 mg/kg) and xylazine (5 mg/kg), the trachea was exposed, and a 24-gauge catheter was inserted into the trachea. A 50-μl bolus of saline containing the E. coli was then instilled into the left bronchus. Mice were euthanized at the indicated time points by isoflurane overdose, and specific tissues were harvested for measurements detailed below.

FIG 1.

The APR is dependent on liver STAT3 activation during endotoxemia followed by pneumonia. (A) Dual challenge of endotoxemia and pneumonia. WT and mutant mice lacking hepatic STAT3 were pretreated with an intraperitoneal injection of 5 mg/kg LPS. After 18 h, mice were intratracheally infected with 1 × 10⁶ CFU of E. coli. Mice were euthanized at 0, 6, or 24 h after E. coli infection. At the indicated time points, serum (B, C) and BALF (D, E) were collected, and SAA and SAP acute-phase protein concentrations were measured using an ELISA. Dashed lines indicate baseline concentrations in vehicle-treated, WT mice without pneumonia. *, P < 0.05 versus WT mice at the indicated time points as determined by a two-way ANOVA followed by a Holm-Sidak test (n = 3 to 9 per group). (F, G) SAA and SAP concentrations from 24-hour serum and BALF were compared, and a correlation with a linear regression was performed. Each individual point represents a single mouse (both WT and mutant). Pearson r values and P values for each correlation are shown (n = 19).

Bronchoalveolar lavage.

Lungs were isolated and bronchoalveolar lavage fluid (BALF) was collected as previously described at the indicated time points (20, 33). Lungs were removed and tethered to a 20-gauge blunted catheter by way of the trachea. Once secured, the lungs were repeatedly lavaged 10 times with 1 ml of phosphate-buffered saline (PBS). The cell-free supernatant from the first lavage was aliquoted and stored at −80°C for protein analysis. Pooled cells from all washes were counted using a hemacytometer, and differential counts were determined after cytocentrifugation and Diff-Quick (Dade-Behring) staining.

Serum collection.

After collection at the indicated time points, blood was incubated in MiniCollect Z Serum Separator tubes (Greiner Bio-One) for 30 min at room temperature and then centrifuged for 15 min at 1,500 × g and 4°C for serum separation. Serum was aliquoted and stored at −80°C for protein analysis.

Protein measurements.

APPs were measured by enzyme-linked immunosorbent assay (ELISA). Serum amyloid A (SAA) and serum amyloid P (SAP) ELISAs were purchased from Immunology Consultants Laboratory, Inc. Cytokine protein concentrations were assessed using a Bio-plex 200 workstation (Bio-Rad) in conjunction with a Bio-plex cytokine bead array (Bio-Rad). Included in the panel were IL-1β, IL-6, IL-10, IL-17, granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), CXCL1, leukemia inhibitory factor (LIF), CXCL2, and TNF-α. Total protein concentrations in BALF were measured using the bicinchoninic acid (BCA) assay (Sigma).

Bacteriology.

At 6 and 24 h following E. coli infection, lungs were isolated and homogenized using a Bullet Blender (Next Advance). Homogenates and heparinized blood were serially diluted in sterile water and plated on 5% sheep blood agar plates (BD Biosciences). After an overnight incubation at 37°C, colonies were counted and expressed as total CFU per lung or per milliliter of blood.

ROS generation.

Mice were administered LPS by i.p. injection followed 18 h later by i.t. E. coli. Six hours after E. coli infection, BALs were performed using ice-cold lavage buffer (Hanks' balanced salt solution [HBSS; Life Technologies, Invitrogen], 2.7 mM EDTA disodium salt solution [Sigma-Aldrich], 20 mM HEPES, 100 U/ml penicillin-streptomycin [Pen-Strep]). BALF cells were stained for reactive oxygen species (ROS) generation using the CellROX Deep Red Reagent as specified by the manufacturer's instructions (Life Technologies). Additionally, cell surface markers were used to identify airspace neutrophils (CD45⁺/7AAD⁻/Ly6G⁺/F4/80⁻) and alveolar macrophages (CD45⁺/7AAD⁻/F4/80⁺/Ly6G⁻/Autofluorescence^hi). The following antibodies were used: CD45-PE/Cy7 clone 30-F11, Ly6G-APC/Cy7 clone 1A8, and F4/80-PE eFluor610 clone CI:A3-1. All antibodies, including 7-amino-actinomycin D (7AAD) viability staining solution, were purchased from Biolegend. Cells were subjected to flow cytometry using the LSRII from BD Biosciences and analyzed for ROS generation in each cell type using FlowJo.

pHrodo phagocytosis assay.

Phagocytosis was measured using red pHrodo E. coli bioparticles (Life Technologies), which fluoresce only in low-pH environments (such as the phagolysosomal compartment). pHrodo bioparticles were prepared by suspension in 250 μl of PBS followed by sonication for 5 min. Mice were given i.p. LPS followed 18 h later by i.t. E. coli. After 6 h, mice were instilled with a 50-μl bolus of the pHrodo bioparticles, and after another hour, the lungs were lavaged with ice-cold lavage buffer (see Fig. 5A). Cells were stained for surface antigens to detect neutrophils and alveolar macrophages as described above. Phagocytosis (phycoerythrin [PE] fluorescence) was examined in each cell type using FlowJo software.

FIG 5.

The hepatic APR does not modulate phagocytosis in airspace cells during endotoxemia and pneumonia. (A) WT and mutant mice were treated with intraperitoneal LPS for 18 h. Afterwards, E. coli was instilled intratracheally, followed 6 h later by a second instillation of E. coli pHrodo particles. Lungs were lavaged an hour later, and cells were stained as follows for flow cytometry: neutrophils (CD45⁺/7AAD⁻/Ly6G⁺/F4/80⁻), alveolar macrophages (CD45⁺/7AAD⁻/F4/80⁺/Ly6G⁻/Autofluorescence^hi). (B) Representative histograms illustrate the percentages of cells positive for pHrodo particle phagocytosis in WT (black line) and mutant (dashed line) mice. Filled curves (gray) represent cells not exposed to pHrodo particles. Summarized data for all mice studied were calculated to determine the frequency (C) and magnitude (D) of particle ingestion, as determined by the percentage of positive cells and mean fluorescence intensity, respectively. No significant changes between genotypes were detected, as assessed by Student's t test (n = 5 or 6 per group).

Cell-based bacterial killing assay.

HepSTAT3^−/− mutant and wild-type (WT) mice were treated with LPS by i.p. injection, and lungs were lavaged 18 h later with ice-cold lavage buffer. Recovered macrophages were pelleted by centrifugation at 300 × g for 5 min at 4°C and then washed with serum-free, ice-cold RPMI 1640 medium (with 1% penicillin-streptomycin). Cells were plated on opaque, tissue culture-treated 96-well plates (Falcon) at a concentration of 100,000 cells/well. After 1 h at 37°C and 5% CO₂, nonadherent cells were washed away with complete RPMI medium (with 1% penicillin-streptomycin and 10% fetal bovine serum), leaving only the adherent alveolar macrophages. Luminescent, log-phase E. coli cells (strain Xen14; Caliper) were added to the macrophage cultures at 10⁷ CFU/ml for 1 h in antibiotic-free RPMI medium (with 10% fetal bovine serum). Cells were then washed twice with RPMI medium containing 100 μg/ml gentamicin (with 10% fetal bovine serum), which is cell impermeable and thus kills only noninternalized bacteria. Bacterial luminescence was then measured immediately and hourly for the next 4 h using a luminometer. At each time point, baseline cell luminescence was subtracted from each sample, and bacterial killing was assessed by determining decreases in bacterial luminescence over time.

BALF bacterial growth assay.

Luminescent E. coli Xen14 (Caliper) was cultured on blood agar plates overnight at 37°C in 5% CO₂. After 18 h, colonies from those plates were then incubated on new plates until the bacteria reached log-phase growth (for around 4 h) at 37°C and 5% CO₂. Once in log phase, E. coli was suspended in PBS at 1 × 10⁶ CFU/ml. Cell-free BALF from mutant and wild-type mice infected for 0, 6, or 24 h with E. coli after LPS injection was aliquoted into a 96-well plate (90 μl/well) and then incubated with 10 μl of the bacterial suspension (for a starting concentration of 1 × 10⁵ CFU/ml in each well) while rotating at 37°C. Bacterial luminescence (as an indicator of bacterial growth) was measured at the start of and after a 5-hour incubation using a luminometer (Turner BioSystems). Growth was calculated as fold increases based on the starting luminescent values. No viable bacteria were detected in the aliquoted BALF used.

Detection of NETs.

In order to quantify neutrophil extracellular trap (NET) release in the BALF, we performed a myeloperoxidase (MPO)-DNA ELISA as previously described (34). A 96-well plate was coated with 5 μg/ml of an anti-MPO antibody (rabbit polyclonal, catalogue number ab9535; AbCam) overnight at 4°C, washed with PBS, and then blocked for 2 h at room temperature with 5% bovine serum albumin (BSA) in PBS. After washing with PBS, 50 μl cell-free BALF from mutant and WT mice infected with intrapulmonary E. coli for 0, 6, or 24 h after LPS injection was added to the plate and incubated while shaking at room temperature for 2 h. After washing with wash buffer (1% BSA and 0.05% Tween in PBS), a peroxidase-labeled anti-DNA monoclonal antibody diluted 1:100 in 1% BSA PBS (from the Cell Death Detection ELISAPlus kit, catalogue number 11774425001; Roche) was added, and the plate was incubated for another 2 h. After another wash with washing buffer, 100 μl of 2,2′-azinobis(3-ethylbenzthiazolinesulfonic acid) (ABTS) solution (also from the Cell Death Detection ELISAPlus kit) was added for 45 min at room temperature in the dark. The optical density of the plate at a wavelength of 405 nm was recorded.

Statistics.

All statistical analyses were done using GraphPad Prism 6.0 (GraphPad). CFU data are illustrated as individual values with medians, whereas the remaining data are shown as means ± standard errors of the means (SEM). Two groups were compared using Student's t test, while multiple group comparisons (i.e., CFU) were conducted using a nonparametric one-way analysis of variance (ANOVA) (Kruskal-Wallis test), followed by Dunn's test for multiple comparisons. When applicable, multiple group comparisons were also made using a two-way ANOVA followed by a Holm-Sidak test. Data were considered significant at P values of <0.05 and were marked accordingly.

RESULTS

The APR is dependent on liver STAT3 during endotoxemia followed by pneumonia.

In order to determine the effect of STAT3-dependent liver activation in the context of sepsis and pneumonia, we used the Cre-LoxP system to obtain a mouse model of hepatocyte-specific, functional STAT3 deletion (hepSTAT3^−/−). To model the clinical circumstances of sepsis preceding pneumonia, we employed a dual challenge of endotoxemia followed by a bacterial lung challenge (Fig. 1A). Mutant and WT mice were administered an intraperitoneal injection of either 5 mg/kg of LPS or vehicle (saline). After 18 h, 1 × 10⁶ CFU of E. coli was intratracheally instilled into left lung lobes for an additional 0, 6, or 24 h. Gram-negative infections, including E. coli pneumonias, are a major cause of nosocomial pneumonia (35, 36), which are particularly relevant during sepsis, as septic patients have a much greater risk of developing enterobacterial, hospital-acquired pneumonias (5, 6). As such, E. coli pneumonias were utilized in this model of sepsis-induced pneumonia because of its specific relevance to septic patients. Additionally, E. coli pneumonia models induce large amounts of lung injury necessary to induce plasma protein (and thus APP) extravasation into the airspaces, which is key in understanding how the liver response to sepsis can directly influence local lung defense. Because liver STAT3 activation is required for maximal APR induction (24, 25, 30, 31), we measured the concentrations of two representative, circulating APPs: serum amyloid A (SAA) and serum amyloid P (SAP). In WT mice, the concentrations of both SAA and SAP in serum were induced dramatically above baseline with LPS pretreatment alone (Fig. 1B and C, 0 h). Unlike SAA, SAP concentrations in serum were further increased by E. coli infection, as there was a significant effect of infection only for SAP levels in serum (Fig. 1C). Independent of treatment (LPS and/or E. coli pneumonia), APP concentrations remained unchanged in mutant mice but were significantly different from those of WT mice, indicating that hepatic STAT3 function is necessary for a maximal APR.

In order to determine whether an endotoxin-induced (STAT3-dependent) APR could affect the local lung environment, we sampled the protein and cellular content of the airspaces by bronchoalveolar lavage. LPS pretreatment alone was insufficient to alter baseline concentrations of airspace SAA and SAP in either mouse genotype (Fig. 1D and E, 0 h). However, intrapulmonary infection with E. coli markedly increased the concentrations of both acute-phase proteins in the BALF of WT mice. SAP increases in BALF were significantly blunted in mutant mice, with a similar trend observed for SAA (Fig. 1E and D, respectively). These changes resembled those in the blood compartment, suggesting that airspace APP content is a function of plasma extravasation into pneumonic lungs, especially in the case of SAP. This is further evidenced by significant correlations between concentrations of both APPs in serum and BALF after 24 h of E. coli infection (Fig. 1F and G).

Host defense during endotoxemia and pneumonia is compromised by lack of hepatic STAT3.

In order to determine if an endotoxemia-induced hepatic APR affects pulmonary host defense and/or inflammation during pneumonia, we measured 6- and 24-h lung and blood bacterial burdens in both genotypes of mice pretreated with either LPS or vehicle. After 6 h of pneumonia, we observed lung bacterial burdens that were similar to the original inoculum (approximately 10⁶ CFU), with no statistical difference between any of the groups tested (Fig. 2A). Moreover, bacteremia was not detected in any group (Fig. 2B). At 24 h postinfection, however, mutant mice pretreated with LPS had significantly greater lung bacterial burdens than any other group (Fig. 2C), suggesting that STAT3-dependent liver activity is required for local defense in response to preexisting endotoxemia. HepSTAT3^−/− mutant mice pretreated with LPS also had significantly increased bacteremia, possibly due to differences in dissemination and/or systemic clearance (Fig. 2D).

FIG 2.

Host defense during endotoxemia and pneumonia is compromised by lack of hepatic STAT3. WT and mutant mice were treated for 18 h with intraperitoneal LPS or saline followed by intratracheal E. coli. After 6 (A, B) and 24 (C, D) hours of E. coli infection, lung homogenates (A, C) and blood (B, D) were processed for quantification of viable bacteria. †, P < 0.05 between the denoted groups based on a Kruskal-Wallis test followed by Dunn's multiple-comparison test (n = 5 to 16 per group). At the indicated time points, BALF was harvested for determination of recruited neutrophil numbers (E) and total protein concentrations (F). *, P < 0.05 versus WT mice at the indicated time points as determined by a two-way ANOVA followed by a Holm-Sidak test (n = 3 to 9 per group). ND, not detected. (G) Survival was observed through 24 h of E. coli infection. *, P < 0.05 versus all other groups as determined by a Mantel-Cox test (n = 19 to 52 per group).

Bacterial killing in the lungs relies on innate immunity, including that provided by recruited neutrophils and other extravasated plasma constituents during inflammation (37). In order to determine whether local inflammation was compromised by STAT3 deficiency, we measured BALF neutrophils and total protein concentrations (Fig. 2E and F, respectively). We observed an influx of neutrophils at 24 h after infection with E. coli in both WT and mutant mice, consistent with an acute pneumonia (Fig. 2E). Additionally, there were significantly greater numbers of neutrophils recruited to the airspaces in mutant mice than in WT mice at 24 h after infection with E. coli, which was likely secondary to increased bacterial loads. Total protein concentrations in BALF were also increased due to infection, but no differences were observed between genotypes (Fig. 2F), suggesting changes in serum APP concentrations and not changes in protein delivery (from blood to airspaces) as the cause of BALF APP differences between genotypes. To reinforce this concept, we performed an additional correlation analysis comparing BALF APP concentrations with BALF total protein levels. Neither SAA nor SAP levels significantly correlated with pulmonary edema (for SAA, P = 0.4916, r = −0.1733; for SAP, P = 0.3802, r = −0.2201), suggesting that any changes in BALF APP content are linked to expression differences either systemically or in the lungs themselves (particularly in the case of mutant mice). Interestingly, 24 h after administration of E. coli in endotoxemic hepSTAT3^−/− mice, impaired antibacterial defense was associated with increased mortality (Fig. 2G). These data suggest that STAT3-dependent liver responses are protective in the setting of sepsis followed by pneumonia. This response, however, does not appear to be mediated through alveolar neutrophil recruitment.

Pulmonary and systemic cytokine induction is not reliant on STAT3-dependent acute-phase changes.

As another index of lung and systemic inflammation, cytokine protein concentrations in BALF and serum were measured (Fig. 3 and 4). We utilized a multiplex bead array to determine the concentrations of 10 cytokines, all of which are relevant to pneumonia and/or lung injury (37, 38): IL-1β, IL-6, IL-10, IL-17, G-CSF, GM-CSF, CXCL1, TNF-α, LIF, and CXCL2. We observed several patterns of cytokine kinetics in the airspaces, ranging from increases due to E. coli infection to no change at all, but there were no changes in BALF cytokine concentrations due to genotype (Fig. 3). Serum cytokine changes were also variable across targets; however, unlike in the BALF, three serum cytokines were significantly changed due to the absence of liver STAT3: IL-1β, IL-17, and TNF-α. Concentrations of both IL-17 and TNF-α were significantly greater in mutant mice, consistent with increased bacteremia. Interestingly, IL-1β was significantly decreased in mutant mice after LPS pretreatment (0 h after administration of E. coli), but not during the course of infection, potentially indicating a small defect in systemic innate immunity. Whether or how this genotype-dependent decrease in IL-1β contributes to the phenotype of this group remains unclear.

FIG 3.

Pulmonary cytokine induction is unaffected by hepatic STAT3 deletion. WT and mutant mice were treated for 18 h with intraperitoneal LPS followed by intratracheal E. coli. At the indicated time points after E. coli infection, lungs were lavaged, and BALF cytokine protein concentrations were determined using a multiplex bead array. Dashed lines (some of which overlap the x axis) indicate baseline concentrations in vehicle-treated, WT mice without pneumonia. There was no significant overall effect of genotype observed, as determined by a two-way ANOVA (n = 3 to 9 per group).

FIG 4.

Hepatic STAT3 activation has a minimal effect on circulating cytokine concentrations. WT and mutant mice were treated for 18 h with intraperitoneal LPS followed by intratracheal E. coli. At the indicated time points after E. coli infection, serum was collected, and cytokine concentrations were measured with a multiplex bead array. Dashed lines (some of which overlap the x axis) indicate baseline concentrations in vehicle-treated, WT mice without pneumonia. For IL-1β, the asterisk (*) indicates a P value of <0.05 versus WT mice at that time point as determined by a two-way ANOVA followed by a Holm-Sidak test. For IL-17 and TNF-α, the asterisk (*) indicates a P value of <0.05 for overall effect of genotype as determined by a two-way ANOVA. n = 3 to 9 per group.

The hepatic APR does not modulate phagocytosis in airspace cells during endotoxemia and pneumonia.

Hepatic STAT3^−/− mice with a preexistent endotoxemia have increased bacterial burdens both systemically and locally during pneumonia. Neutrophil recruitment and other inflammatory mediators (i.e., cytokines) were either unchanged or increased in hepSTAT3^−/− mice, suggesting that these aspects of host defense are uncompromised in mutant mice. To determine if endotoxin-induced liver STAT3 activation affects cellular defenses during pneumonia, we measured phagocytosis in airspace macrophages and neutrophils using pHrodo E. coli bioparticles. These bioparticles are conjugated to a phycoerythrin (PE) fluorophore that fluoresces only in low-pH environments, characteristic of the phagolysosomal compartment. Multiple laboratories have validated this system as an effective strategy for discriminating between surface-bound and internalized particles (39–41). After 18 h of i.p. LPS administration, mutant and WT mice were i.t. infected with E. coli for 6 h, followed by a second i.t. instillation with pHrodo E. coli bioparticles. After 1 h, the lungs were lavaged and cells were analyzed by flow cytometry (Fig. 5A). Airspace macrophages and neutrophils included cells positive for phagocytosis of pHrodo bioparticles (Fig. 5B). Neither macrophages nor neutrophils, however, exhibited genotype-dependent differences in the frequency (Fig. 5C) or magnitude (Fig. 5D) of phagocytosis. These data suggest that bacterial uptake and phagolysosomal fusion are unlikely to be responsible for impaired bacterial killing in the absence of hepatocyte STAT3 during endotoxemia and pneumonia.

Maximal ROS generation in airspace macrophages is dependent on hepatic STAT3 activation.

As an alternative contributor to cellular host defense, ROS generation was measured in airspace cells from both genotypes following 6 h of pneumonia in endotoxemic mice. Total cells were stained for surface antigens to identify macrophages and neutrophils as described above, and ROS production was measured using the CellROX Deep Red Reagent from Life Technologies (Fig. 6A). Interestingly, airspace macrophages from mutant mice had significantly less ROS production than those from WT mice (Fig. 6B). A similar trend was apparent with neutrophils, but this did not reach statistical significance (Fig. 6B). These data connect compromised macrophage ROS production to impaired pulmonary host defense in endotoxemic hepSTAT3^−/− mice.

FIG 6.

Alveolar macrophage reactive oxygen species (ROS) production and airspace bacterial resistance are dependent on the hepatic STAT3 activation. (A) WT and mutant mice were treated for 18 h with intraperitoneal LPS followed by an intratracheal instillation of E. coli. Six hours later, the lungs were lavaged and recovered cells were stained using the CellROX Deep Red Reagent to determine ROS generation in neutrophils (PMN; CD45⁺/7AAD⁻/Ly6G⁺/F4/80⁻) and alveolar macrophages (Mφ; CD45⁺/7AAD⁻/F4/80⁺/Ly6G⁻/Autofluorescence^hi). Representative histograms illustrate the mean fluorescence intensity (MFI) for the populations positive for ROS generation in WT (black line) and mutant (dashed line) mice. Filled curves (gray) represent cells not exposed to CellROX reagent. (B) ROS generation was quantified in each cell type, and data are shown as the percentage of ROS generation observed in WT mice. §, P < 0.05 versus WT as determined by Student's t test (n = 5 or 6 per group, with each point corresponding to an individual mouse). (C) Mutant and WT mice were treated with either i.p. vehicle or LPS for 18 h. Lungs were lavaged, and alveolar macrophages were isolated and plated in a 96-well plate. After adhering for 1 h, 1 × 10⁷ CFU/ml of luminescent E. coli were incubated with the macrophages for another hour. Additionally, control, cell-free wells containing only E. coli were included. After 1 h, all wells were washed (including the E. coli-only control wells to account for residual bacteria after washing) and then incubated in 100 μg/ml of gentamicin for 4 h. Luminescence, as a metric of bacterial viability, was recorded at the start and every hour thereafter. The WT/saline group was statistically different from the other groups as determined by a two-way ANOVA followed by a Tukey post hoc test for individual comparisons among groups (n = 5 to 9, with each well corresponding to a different mouse). Additionally, the experiment was performed on three separate days. *, P < 0.05 versus all other groups. §, P < 0.05 versus WT/LPS and E. coli + gentamicin. (D) Neutrophil extracellular traps (NETs) were quantified by an MPO-DNA ELISA in the BALF from WT and mutant mice pretreated with LPS for 18 h followed by intrapulmonary infection with E. coli for the indicated time periods. A significant overall effect of infection but not genotype was observed, as determined by a two-way ANOVA (n = 3 to 9 per group, corresponding to BALF used from individual mice infected in at least two independent experiments). (E) The same cell-free BALF as that used in panel D was incubated with log-phase luminescent E. coli, rotating at 37°C for 5 h. Bacterial growth was calculated as fold increases in luminescence compared to the starting values for each sample. *, P < 0.05 versus WT at the specified time point as assessed by a two-way ANOVA followed by a Holm-Sidak test (n = 3 or 4 per group, corresponding to BALF used from 3 or 4 individual mice infected in at least two independent experiments).

Macrophage intracellular bacterial killing is not impaired by liver STAT3 deletion.

As another assessment of cellular host defense, we utilized a gentamicin protection assay in which we incubated primary alveolar macrophages with luminescent E. coli. Macrophages were originally harvested from mutant or WT mice 18 h after treatment with i.p. vehicle or LPS. After an hour to allow for bacterial uptake, gentamicin, which does not penetrate the cell membrane, was added to kill any remaining extracellular bacteria, and luminescence was recorded hourly as a measure of bacterial viability. Interestingly, initial bacterial uptake was reduced in macrophages from either genotype after LPS treatment (Fig. 6C, 0 h), consistent with previous reports detailing alveolar macrophage dysfunction after sepsis. Although the uptake of bacteria was compromised in macrophages isolated from LPS-treated mice, no additional changes were observed due to genotype (Fig. 6C). In fact, luminescence barely exceeded the levels detected in negative-control wells (no macrophages), limiting the opportunity to determine additional effects resulting from liver STAT3 deficiency.

Soluble host defense mediators within the airspaces are dependent on the hepatic APR.

Neutrophils are immediately recruited to the alveolar compartment during early stages of infection to aid in pathogen clearance (42). As an innate defense, in addition to phagocytosis, neutrophils are equipped to release endogenous genomic DNA laced with antimicrobial proteins to effectively trap and lyse invading microbes. These NETs are studded with granulocytic proteins, including MPO (43). As a way to determine whether NET release was affected by the APR, we measured the relative concentrations of NETs in BALF from WT and mutant mice after endotoxemia and pneumonia (Fig. 6D). As anticipated, we observed an overall increase in NET release due to pneumonia, and while there is a trend toward decreased NET release in mutant mice, this difference did not reach statistical significance (Fig. 6D).

In order to determine whether extracellular products other than NETs may contribute to differential bacterial resistance in the alveolar lining fluid, we developed an assay in which we incubated luminescent E. coli (strain Xen14) in cell- and bacteria-free BALF from endotoxemic and pneumonic WT or mutant mice. Bacterial growth was calculated as fold increases in luminescence compared to the starting values for each sample. Interestingly, BALF from mutant mice supported bacterial growth significantly more than did that from WT mice (Fig. 6E), suggesting that the airspace milieu of mutant mice is less resistant to bacterial growth. Whether and how this change in bacterial resistance in the airspaces relies on differences in the antimicrobial proteome or nutrient availability of the alveolar lining fluid remains uncertain.

DISCUSSION

The results of this study demonstrate a novel role for the STAT3-dependent liver acute-phase response in driving innate host defenses during pneumonia in endotoxemic animals. Using a two-hit model of endotoxemia and intrapulmonary E. coli, we observed impaired antibacterial defense and higher mortality in mice that were deficient in hepatic STAT3. While several indices of inflammation (e.g., neutrophilia, edema, and cytokine induction) were largely unaffected by the interruption of hepatic activation, others (e.g., macrophage ROS generation and airway lining fluid content) were dependent on hepatic STAT3.

The physiologic and molecular mechanisms by which hepatic innate responses mediate host defense during sepsis and pneumonia have never been elucidated. Several studies, however, have implicated an important role for hepatic STAT3 activation during either sepsis or pneumonia alone. Alonzi et al. described the necessity of inducible liver STAT3 activation during endotoxemia for induction of the APR (31). Additionally, Sakamori et al. used a hepatocyte-specific STAT3 knockout mouse to show the importance of this signaling pathway in controlling excessive inflammation during polymicrobial sepsis induced by cecal ligation and puncture (CLP) (30). In fact, their results for mutant mice during sepsis alone were consistent with our own, showing decreased survival as well as increases in circulating cytokines; however, they did not detect changes in blood bacterial burdens. Similarly, Sander et al. demonstrated that liver STAT3-dependent signaling was also crucial to attenuate mortality, but not host defense, in response to CLP through a process facilitated by SAA-dependent mobilization of myeloid-derived suppressor cells (44). The last two studies described above, while notable, were not designed to determine the degree to which sepsis-induced liver activation (via STAT3) calibrates subsequent responses to pneumonia, which is a highly distinct and clinically relevant scenario.

It is well established that in both septic patients and animal models, sepsis results in immunosuppression (45), which is thought to promote secondary infections such as those causing pneumonia (8, 46). A multitude of studies have revealed the detrimental consequences of sepsis-induced immunosuppression on critical pneumonia outcomes, including antibacterial defense, alveolar macrophage function, alveolar neutrophil recruitment, and cytokine production (7, 9, 10, 12–15, 47–51). Our own protocol of endotoxemia followed by pneumonia, however, was not sufficient to recapitulate the circumstances of sepsis-induced immunosuppression. We observed no effect of endotoxemia alone on pneumonia outcomes in WT mice, including pulmonary defense, lung cytokine expression, and neutrophil recruitment, but rather found that endotoxemia compromised bacterial clearance only in mice lacking hepatic STAT3 (Fig. 2C). There are many possible explanations for this. First, the dose of LPS (5 mg/kg) and/or its type (an ultrapure, Toll-like receptor 4 [TLR4] agonist) may not be sufficient to induce immunosuppression in the setting of our pneumonia protocol. Additionally, the timing of LPS pretreatment (18 h before E. coli infection) and/or the genetic background of our hepSTAT3^−/− mouse strain could also be factors. The lack of observable LPS-induced immunosuppression in WT mice, however, empowered us to more precisely examine the roles of endotoxin-induced hepatic STAT3 activation on a subsequent lung infection, and this opportunity may have been diminished by overwhelming immunosuppression due to LPS alone.

Independently, our laboratory and others have reported a functional role for the APR in pneumonia alone. We have shown, using an APR-null mouse model (lacking both hepatic STAT3 and RelA), that liver activation is required for survival, hepatoprotection, and maximal pulmonary inflammation during an E. coli pneumonia (23), as well as systemic defense and opsonophagocytosis during pneumococcal pneumonia (24). The common clinical observation that sepsis is frequently followed by pneumonia (5, 6, 8) raises the question of whether or how a preexisting liver response alters pneumonia susceptibility, for better or for worse. Renckens et al. determined that a preexisting APR induced by turpentine impairs the pulmonary inflammatory response to Pseudomonas aeruginosa and Acinetobacter baumannii (52, 53). The model of inducing a preexisting APR via turpentine injection is very different from our method of inducing the APR through endotoxemia. Additionally, turpentine's effects are unlikely to be limited to liver activation. Using our hepatocyte-specific STAT3-null mouse in our model of endotoxemia followed by pneumonia allowed us, for the first time, to interrogate the role of preexisting liver-specific acute-phase changes on pneumonia susceptibility. This is an important distinction from our earlier studies, which examined the global acute-phase changes (driven by both STAT3 and RelA) in the setting of pneumonia alone. Moreover, by examining the effects of preexisting STAT3-dependent liver responses, these studies aim to help clarify an important clinical/immunological scenario in which sepsis modifies subsequent immune responses to lung pathogens.

In association with impaired APP induction, mutant mice pretreated with LPS had significantly greater bacterial loads in the lungs and blood during pneumonia, implying that local pulmonary defenses are particularly affected during endotoxemia in the absence of an intact liver response. Increased mortality was also observed in this group, suggesting this defect in host defense as a potential cause of mortality. These outcomes were also associated with an increase in serum TNF-α that is likely due to greater amounts of circulating bacteria and could also contribute to death in hepSTAT3^−/− mice, as TNF-α can cause septic shock (54). In trying to determine which aspects of host defense are mediated by the sepsis-induced APR, we measured pulmonary inflammation and injury. We observed no decrease in neutrophil recruitment, pulmonary cytokine concentrations, or proteinaceous edema between genotypes, suggesting that these characteristic measures of inflammation were unlikely to contribute to host defense differences in endotoxemic hepSTAT3^−/− mice. In fact, the only apparent changes in lung cytokine levels (IL-6, G-CSF, and LIF) actually trended toward an increase, which we hypothesize to be secondary to increased bacterial burdens in this experimental group. Overall, the immunosuppression observed in our own study differs from previous findings, which typically involve reduced cytokines and inflammation (9, 10).

Phagocytosis and NET production were also equivalent between groups. Regarding the former, however, we acknowledge the fact that pHrodo E. coli bioparticles (our method of quantifying phagocytosis) may not perfectly replicate interactions between living E. coli and the inflammatory milieu (including opsonins such as extravasated APPs). Yet we observed extremely efficient uptake using this system (around 40 to 60%) in both cell types analyzed, supporting an environment sufficient for comparison of phagocytic functions. Interestingly, ROS generation was significantly attenuated in alveolar macrophages from mutant mice, suggesting that the endotoxemia-induced hepatic APR facilitates at least one fundamental aspect of cell-mediated antimicrobial defense. We also employed a primary alveolar macrophage-based bacterial killing assay to determine if differences in ROS production could manifest as changes in cellular bacterial killing ex vivo. Significantly more bacterial uptake was detected in macrophages recovered from vehicle-treated WT mice than in those from LPS-treated mice of either genotype (Fig. 6C). Yet we observed no differences in killing with or without hepatocyte STAT3, suggesting that either (i) liver STAT3 deficiency (in the setting of endotoxemia) is insufficient to compromise the antibacterial function of an otherwise unchallenged (no pneumonia) alveolar macrophage or (ii) differences in bacterial killing are beyond the detection limit of this experimental system, perhaps due to the small amounts of bacterial uptake in macrophages from LPS-treated mice. The immunosuppressive effect of endotoxemia on macrophage function is consistent with that seen in other studies (11, 49, 55). Future investigations are needed to determine whether or how previously established pathways driving sepsis-induced pulmonary immunosuppression are mechanistically linked to STAT3-dependent liver activity.

We also assessed the capacity of alveolar lining fluid to influence bacterial growth after endotoxemia in mice with and without STAT3-dependent liver responses. Indeed, the composition of soluble mediators in this niche could have large implications on pathogen resistance at the air-liquid interface. In an effort to understand whether the balance of growth-promoting and growth-inhibiting soluble factors is dependent on the APR, we incubated luminescent E. coli with cell- and bacterium-free BALF from mutant and WT mice collected at different time points following endotoxemia and E. coli infection. Interestingly, BALF from mutant mice supported E. coli growth significantly more than that from WT mice, suggesting that products downstream of hepatic STAT3 activation create a less favorable environment for infection in the airspaces. There have been multiple reports of antimicrobial polypeptides, including lactoferrin, lysozyme, lipocalins, and beta-defensins, which are produced in the liver and/or lungs (29, 56–58) and could be modulated by the APR either directly or indirectly. Alternatively, increased growth in the BALF from mutant mice could be attributable to an altered nutrient pool that is more supportive of bacterial replication. While our assay does not allow us to discriminate between the bactericidal and bacteriostatic components of BALF, this finding, combined with the defect in macrophage ROS generation and increased bacterial burdens, strongly supports the concept that the sepsis-induced hepatic APR is a required component for maintaining pulmonary defense.

While our results identify a critical role for STAT3 activation in the setting of endotoxemia, the actual mechanisms of LPS-induced liver STAT3 activity are not entirely clear. We and others have shown IL-6 to be both sufficient and necessary for hepatic STAT3 activation in a variety of contexts (16, 59–61). In response to LPS, specifically, there is evidence suggesting an important role for IL-6, although others have shown evidence of IL-6-independent LPS-induced liver STAT3 as well. Mechanisms whereby the liver promotes lung defense constitute an even greater knowledge gap. Indeed, liver-derived acute-phase proteins are diverse in number and function, promoting a wide variety of immunological processes. Previous studies from our laboratory have implicated the APR in the activation of airspace macrophages during pneumonia (23), and a multitude of APPs have been shown to activate macrophages (61–64). Also, pentraxins such as SAP and C-reactive protein (CRP) engage numerous receptors capable of activating macrophages (and other cells), which can promote ROS generation (65, 66). In addition to this example, which may be linked to our current ROS results, many other liver-derived proteins likely integrate to directly and/or indirectly enhance macrophage activity. Which of these applies to the lung and/or macrophage responses in our current studies is an avenue of future investigation.

This study puts forth evidence of a novel, immunoprotective role for hepatic STAT3 during endotoxemia. Our data suggest that endotoxemia may initiate both immunosuppressive and immunoprotective responses, which were effectively balanced in WT mice. In the absence of liver STAT3, however, the immunosuppressive effects of endotoxemia were revealed, indicating a protective role for STAT3-dependent gene programs. Future insight into the mechanisms by which sepsis-mediated liver activation is protective during subsequent lung infections will provide valuable, alternative avenues for the treatment and prevention of sepsis and pneumonia.