Lipopolysaccharide Clearance, Bacterial Clearance, and Systemic Inflammatory Responses Are Regulated by Cell Type–Specific Functions of TLR4 during Sepsis

Sepsis affects ∼750,000 patients every year in the United States, resulting in 250,000 deaths (1). Much of the pathobiology of sepsis is the result of the host immune response to the pathogen, which is driven through the detection of microbial-derived molecules by pattern recognition receptors of the innate immune system (2). Pattern recognition receptors, such as TLRs, exhibit selectivity for microbial molecules that can be organism specific and that may be reflected in the subcellular location of the receptor (2). For example, TLRs that recognize bacterial membrane components are expressed on the cell surface of host cells, whereas TLRs that detect microbial nucleic acids are expressed predominantly in the endosomal compartment (3). This spatial organization of TLRs may facilitate the recognition of the microbes based on their unique components that are most likely to be available for host detection within specific subcellular domains. However, specificity of TLR function is likely to be dictated by subcellular location, as well as by cell type, because TLRs are expressed on diverse cell types that have wide-ranging and complementary functions (4–7). However, the integration of host responses to infection through the recognition of the same microbial molecule by different cells using the same receptor is poorly understood. A better understanding of how the host response to microbial challenge is integrated through microbial recognition by different cell types may be essential to successfully embark upon therapeutic strategies to manipulate host–microbe interactions.

The TLR4–MD2 complex is well-characterized as the receptor for Gram-negative bacterial endotoxin, or LPS (8). This receptor complex is expressed on numerous cell types, including cells of the immune system and parenchymal cells (9). Studies using mouse strains deficient in TLR4 signaling (10–12), TLR4 expression (10, 13–16), or using inhibitors of TLR function in wild type (WT) mice (1, 17) confirmed that TLR4 contributes to bacterial clearance and the host inflammatory response in the setting of Gram-negative bacterial infection. Furthermore, the TLR4 Asp²⁹⁹Gly polymorphism, which results in depressed LPS-TLR4 signaling (18), is associated with increased susceptibility to Gram-negative infections in humans (19). Therefore, the TLR4-dependent response of individual cell types in the setting of bacterial sepsis is likely to be determined by the functional role of each cell type in the overall host response.

In vitro studies showed that macrophages respond to LPS through TLR4-MD2 by enhancing phagocytosis and producing inflammatory mediators (20). We showed that hepatocytes use TLR4 to take up LPS (9). To study the relative contribution of various cell types to the host response to polymicrobial sepsis, we generated strains of mice with TLR4 deleted specifically from myeloid cells (macrophages and neutrophils; LysMTLR4KO) (21, 22), or deleted specifically from hepatocytes (HCTLR4KO) (23, 24), and subjected them, along with appropriate control strains, to cecal ligation and puncture (CLP) in the presence or absence of antibiotics. In the absence of antibiotics, the dominant role for LysM-TLR4 is the enhanced phagocytosis and clearance of bacteria. In the presence of antibiotics, which effectively cleared bacteria from the blood and peritoneum, LysM-TLR4 drives cytokine production and organ damage. The efficiency of bacterial clearance and the magnitude of the host inflammatory response are determined by hepatocyte TLR4 expression, which is critical for LPS clearance from the circulation. These observations emphasize the importance of regulating even the low levels of LPS generated during polymicrobial sepsis in coordinating the immune response to bacterial infection. These results also shed light on the cell type–selective roles for TLR4 in coordinating this response. Therefore, strategies directed toward modulating or interfering with microbial molecule recognition should take into consideration the different roles that these receptors play on multiple cell populations.

Ultrapure LPS (Escherichia coli 0111:B4) was from List Biological Laboratories (Vandell Way, CA). This LPS does not contain a significant amount of contaminating proteins that could stimulate TLR2 nonspecifically, and was used in all in vitro experiments. LPS from Sigma (St. Louis, MO) was used for in vivo experiments. Alexa Fluor 488–E. coli LPS was from Invitrogen (Carlsbad, CA). All LPS was tested for purity by separation on silver-stained SDS-PAGE gels, and no detectable TNF was produced in TLR4-null (TLR4^−/−) macrophages in response to any of the LPS used. Williams Medium E was from Life Technologies-BRL (Grand Island, NY); FCS was from HyClone Laboratories (Logan, UT); LysoTracker DND99 was from Invitrogen (Carlsbad, CA); and goat anti-LPS Endotoxin Ab was purchased from Thermo Scientific (Rockford, IL).

Male WT (TLR4^loxP/loxP) mice, cell specific TLR4^−/− mice, and global TLR4^−/− mice were bred at our facility at the University of Pittsburgh and used at the age of 8–12 wk. All mice developed were on a C57BL/6 genetic background. Animal protocols were approved by the Animal Care and Use Committee of the University of Pittsburgh, and the experiments were performed in strict adherence to the National Institutes of Health Guidelines for the Use of Laboratory Animals. TLR4^loxP/loxP and cellular-specific TLR4^−/− mice were generated as previously described (25). In brief, the TLR4^loxP allele was created by inserting loxP sites within intron 1 and intron 2, flanking exon 2 of TLR4. Mice homozygous for TLR4^loxP were generated by Ozgene (Bentley, WA). TLR4^loxP/loxP mice were interbred with stud males (TLR4^{loxP/-; Alb-cre} and TLR4^{loxP/-; Lyz-cre}) to generate the desired genotype. Mice homozygous for Cre recombinase linked to the albumin (alb) and lysozyme (lyz) promoter are commercially available from The Jackson Laboratory. Transgenic male mice used for experiments were confirmed to be the desired genotype via standard genotyping techniques. WT mice used in this study were either C57BL/6 or TLR4^loxP/loxP (flox) mice, as specified in the results. Global TLR4^−/− mice were globally lacking the loxP flanked exon 2 (i.e., they were global homozygotes for the same mutation contained within the conditional knockout [KO] mice).

Sepsis was induced by CLP. Mice weighing 25–30 g were used. Skin was disinfected with 2% iodine tincture. Laparotomy was performed under isoflurane anesthesia (Piramal Critical Care, Bethlehem, PA), and 50% of the cecum was ligated and punctured twice with a 22-gauge needle. Saline (1 ml) was given s.c. for resuscitation immediately after the operation. For analgesia, buprenorphin (0.1 mg/kg, Butler Schein, Dublin, OH) was injected to mice s.c. every 12 h starting 2 h after CLP. In some experiments, mice were also given antibiotics (PRIMAXIN, 25 mg/kg; Merck) s.c. every 12 h. At time points after CLP, mice were anesthetized with isoflurane and euthanized by opening of the chest cavity and withdrawal of blood by cardiac puncture.

Cells were isolated from mice by an in situ collagenase (type VI; Sigma) perfusion technique, modified as described previously (26). Hepatocyte purity exceeded 99% by flow cytometric assay. Cell viability was typically >90% by trypan blue exclusion. Hepatocytes (150,000 cells/ml) were plated on gelatin-coated culture plates or coverslips precoated with Collagen I (BD Pharmingen, San Diego, CA) in Williams medium E with 10% calf serum, 15 mM HEPES, 10⁻⁶ M insulin, 2 mM l-glutamine, 100 U/ml penicillin, and 100 U/ml streptomycin. Hepatocytes were allowed to attach to plates overnight; prior to treatment, the cell culture medium was changed to serum-free medium.

Samples for bacterial culture were collected as mice were euthanized at time points after CLP. The peritoneal cavity was washed with 1 ml PBS, and the peritoneal lavage was collected under sterile conditions. The left lateral liver lobe was removed and mechanically homogenized with 1 ml PBS under sterile conditions. Peritoneal lavage fluid, liver homogenates, and blood were subjected to serial 10-fold dilutions and cultured overnight in 5% sheep blood agar (Teknova, Hollister, CA). CFU were quantified by manual counting.

Blood was collected in heparinized tubes and centrifuged at 10,000 × g for 7 min to obtain plasma. Plasma was assayed for endotoxin using the LAL Chromogenic Assay (Hycult Biotech, Uden, The Netherlands), according to the manufacturer’s instructions, to semiquantitatively determine LPS level in liver and plasma. Cytokines (IL-6, TNF-α) were assayed by ELISA (R&D Systems, Minneapolis, MN), according to the manufacturer’s instructions. Alanine aminotransferase (ALT) was measured using a HESKA DRI-CHEM Veterinary Chemistry Analyzer (Loveland, CO) and slides from Fujifilm (Asaka-shi Saitama, Japan).

Myeloperoxidase (MPO) activity was determined by commercially available MPO assay (Mouse MPO ELISA kit; Hycult Biotech), according to the manufacturer’s instructions, using tissue lysates from lungs, liver, and peritoneal lavage fluid. Briefly, liver and lung tissue harvested from mice after CLP was lysed with 1× cell lysis buffer (Cell Signaling Technologies) containing 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM Na₂EDTA, 1 mM EGTA, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM β-Glycerophosphate, 1 mM Na₃VO₄, 1 μg/ml leupeptin, and 1 μg/ml PMSF on ice for 10 min. Tissue lysate was centrifuged at 10,000 × g for 10 min at 4°C. Protein content in the supernatant was determined by BCA protein assay (Pierce, Rockford, IL.). An equal amount (10 μg) of liver and lung protein and 100 μl peritoneal lavage fluid were assayed.

For in vitro experiments, a phagocytosis assay was performed according to the manufacturer’s instructions (Vybrant Phagocytosis Assay Kit; Life Technologies). Briefly, peritoneal monocytes were isolated from peritoneum by peritoneal lavage using DMEM (Life Technologies). Cells were washed in DMEM and plated on 96-well plates (10⁵/well) in DMEM with 5% FBS. Plates were washed in PBS after 4 h, and adherent cells were treated with heated-killed fluorescent-labeled E. coli (K12 strain) for 2 h and subsequently quenched with trypan blue for 1 min. Fluorescence intensity in cells was determined using a fluorescence microplate reader. For in vivo experiments, heated-killed fluorescent-labeled E. coli (K12 strain) were injected i.p. in mice. After 2 h, peritoneal macrophages were isolated by peritoneal lavage using DMEM. Cells were washed in DMEM and plated in 96-well plates, as above, and after 1 h plates were washed with PBS. Fluorescence of adherent cells was quenched with trypan blue for 1 min. Fluorescence intensity in cells was determined as above.

A total of 5 μM/ml LysoTracker red was added to isolated primary mouse hepatocytes plated on collagen-coated 35-mm plates for 15 min, following by washing twice with serum-free media. A total of 1 μg/ml Alexa Fluor 488–E. coli LPS (Molecular Probes) was added to hepatocytes. LPS uptake was monitored using the Nikon A1 confocal live cell system, and images were captured every 2 min for 2 h. Images were analyzed by MetaMorph (Molecular Devices).

Mice were injected with 5 mg/kg E. coli LPS i.v. via tail vein and sacrificed at 6 or 24 h after injection. Liver and blood were collected for analysis. Liver tissue was fixed immediately in 2% paraformaldehyde for 2 h, followed by overnight rehydration in 30% sucrose. Tissues were then frozen in 2-methyl butane and stored at −80°C until tissue sectioning (4 μm) using a cryostat. For LPS immunofluorescence, liver sections were permeabilized with 0.1% Triton X-100, washed in PBS and 0.5% BSA in PBS, and blocked with 2% BSA in PBS for 1 h. Anti-LPS Ab (Thermo Scientific, Rockford, IL) was added at 1:100 dilution for 1 h at room temperature. Secondary Ab (donkey anti-goat; 1:1000 dilution) was added for 1 h at room temperature. Liver sections were visualized by confocal microscopy and images were analyzed using Metamorph.

All data were analyzed using GraphPad Prism software (GraphPad, San Diego, CA). Data from multiple-group experiments were analyzed using one-way ANOVA, followed by a post hoc Tukey test to compare groups. For measurements of bacterial CFU, groups were compared using a nonparametric Mann–Whitney U test. Survival data were analyzed using the log-rank test. A p value <0.05 was considered statistically significant for all experiments. All values are presented as the mean ± SD, with the exception of bacterial counts, for which median values are designated.

To establish the role of TLR4 on specific liver cell populations during polymicrobial sepsis, CLP without antibiotics was performed on global TLR4^−/− (KO) mice, as well as mice with TLR4 deleted only from macrophages and neutrophils (lyz-cre TLR4KO mice; LysMTLR4KO) or hepatocytes (albumin-cre TLR4KO mice; HCTLR4KO mice). The efficiency and specificity of TLR4 deletion in hepatocytes or liver macrophages were confirmed by assessing LPS-specific responses from cells isolated from the mouse strains used in this study. Liver nonparenchymal cells, which include liver macrophages (Kupffer cells) isolated from WT (C57BL/6) and TLR4^loxp/^loxp (Flox) control mice, as well as from HCTLR4KO mice, responded to LPS (100 ng/ml) by releasing IL-6 and TNF-α into the culture supernatant (Supplemental Fig. 1A). In contrast, liver nonparenchymal cells from either global TLR4KO mice or LysMTLR4KO mice did not release significant amounts of IL-6 or TNF-α in response to LPS (Supplemental Fig. 1B). We showed previously that hepatocytes avidly take up LPS in a TLR4-dependent manner (9), and, as expected, cultured hepatocytes from WT, Flox, and LysMTLR4KO mice took up fluorescent-labeled LPS, whereas cells from global TLR4KO and HCTLR4KO mice did not (Supplemental Videos 1–4). These data show that the conditional KO strains exhibited the expected cell type–selective functional deletion of TLR4.

We first assessed survival over 7 d in the global and cell-specific TLR4KO mice and their WT and Flox (TLR4^loxP/loxP) controls to determine the impact of TLR4 deletion on the host response to polymicrobial infection. WT and Flox mice had similar survival, with ∼40% surviving to day 7 (p = 0.451). However, mortality reached 100% in global TLR4KO and LysMTLR4KO mice (p < 0.05, versus WT/Flox), with improved survival (70% surviving to day 7; p = 0.0453, versus Flox) in HCTLR4KO mice (Fig. 1A).

TLR4 has known roles in activating leukocyte phagocytosis (27), so we examined the efficiency of bacterial clearance to determine whether this would explain the differences in survival between the mouse strains. Bacterial CFU were measured in the blood, liver, and peritoneal washout of mice at 6 and 18 h after CLP. For all results, WT mice served as controls for the global TLR4KO mice, and parent strain Flox mice served as controls for the cell type–selective KO strains. There were no significant differences in bacterial counts between WT and Flox mice after CLP. As shown in Fig.1B–1D, bacterial clearance from the circulation, liver, and peritoneal cavity was significantly impaired in the global TLR4KO and LysMTLR4KO mice. In contrast, bacterial clearance from the blood and peritoneal cavity was more efficient in the HCTLR4KO mice. Circulating LPS levels were below the level of detection at 6 h, but they correlated with bacterial counts at 18 h, with elevated levels in both global TLR4KO and LysMTLR4KO strains (Fig. 1E). There were also significant increases in the plasma ALT level, a marker of hepatocellular injury, in mouse strains that failed to clear bacteria efficiently (Fig. 1F). These observations indicate that TLR4 expressed on macrophages and neutrophils is important for the efficient clearance of bacteria from the circulation, liver, and peritoneal cavity, which correlates with liver inflammation, during bacterial peritonitis.

We next assessed the importance of TLR4 on the local and systemic inflammatory responses during CLP without antibiotics. Neutrophil MPO levels in the lungs, liver, and peritoneal cavity were measured at 0, 6, and 18 h to estimate PMN influx. As shown in Fig. 2A and 2B, neutrophil influx into the peritoneal cavity and lungs was highest at 6 h and returned to baseline by 18 h. MPO levels were significantly higher in the global TLR4KO and LysMTLR4KO strains, suggesting that neutrophils responded to increased bacterial levels with a greater influx into tissues, consistent with the known and powerful chemotactic effects of bacteria on neutrophils. Liver MPO levels were also elevated at 6 h but were not altered by TLR4 status on either monocytes or hepatocytes (Fig. 2C). Serum IL-6 concentration was not detectable at baseline (time 0), but it was significantly elevated after CLP (Fig. 2D). In the control strains (WT and Flox), circulating IL-6 levels were significantly elevated at 6 h after CLP and decreased slightly by 18 h. However, IL-6 levels in the global TLR4KO and LysMTLR4KO mice were similar to controls at 6 h, but they continued to rise and were significantly elevated compared with controls at 18 h (Fig. 2D). Thus, the magnitude of the inflammatory response, as measured by the circulating level of IL-6, was noted to correlate with the efficiency of bacterial clearance, which, in turn, depends on the expression of TLR4 on macrophages and neutrophils. The failure to clear bacteria correlates with an increase in inflammatory markers, even in the absence of TLR4 on macrophages and neutrophils.

Our results in the CLP model without antibiotic administration indicate that the expression of TLR4 on macrophages is important for bacterial clearance and that the efficiency of bacterial clearance determines the magnitude of the inflammatory response and end-organ injury. To evaluate the impact of TLR4 status, independent of variations in bacterial clearance, CLP was carried out, followed by the administration of the antibiotic imipenem (25 mg/kg, s.c. every 12 h). Antibiotic treatment reduced bacteria in the blood to undetectable levels, as well as levels in the liver and peritoneum by two to four orders of magnitude at 18 h (Fig. 3A, 3B). Furthermore, bacterial counts were independent of TLR4 status, with no statistical differences in bacterial load between mouse strains (Fig. 3A, 3B). Plasma endotoxin (LPS) levels at 18 h were significantly elevated in both global TLR4KO and HCTLR4KO mice (Fig. 3C). However, circulating IL-6 levels were significantly increased only in the HCTLR4KO mice, suggesting that higher endotoxin levels resulted in increased activation of TLR4-sufficient macrophages in these mice (Fig. 3D). This was supported by the significantly lower levels of IL-6 in LysMTLR4KO mice when bacterial load was equalized between strains with antibiotics (Fig. 3D). IL-6 levels also correlated with evidence of cellular injury in the liver (ALT level) and were greatest in HCTLR4KO mice (p < 0.05, versus Flox) (Fig. 3E). Taken together, these data suggest that TLR4 on hepatocytes is important for LPS clearance during polymicrobial sepsis and that circulating LPS stimulates TLR on myeloid cells during sepsis to produce inflammatory cytokines, which increases end-organ damage.

We next confirmed that uptake of Alexa Fluor 488–labeled fluorescent LPS by cultured hepatocytes is dependent on TLR4 (Fig. 4A, 4B, Supplemental Videos 1–4). Furthermore, live cell microscopy revealed that the labeled LPS colocalized with the lysosomal tracker (Fig. 4C) in hepatocytes.

The importance of TLR4 on hepatocytes for LPS clearance was also verified in vivo. Mice were injected i.v. with LPS (5 mg/kg), and liver and plasma were collected at 6 or 24 h after injection. LPS uptake in liver cells was easily detectable by immunofluorescence at 6 h in hepatocytes in the livers of WT, Flox, and LysMTLR4KO mice (Fig. 5A). In contrast, no LPS was visualized in hepatocytes of global TLR4KO or HCTLR4KO mice (Fig. 5A). However, LPS was evident in sinusoidal cells of all strains of mice, which is consistent with uptake into Kupffer cells in a TLR4-independent manner. Circulating LPS levels were elevated in all strains at 6 h after injection, but they remained significantly elevated at 24 h only in global TLR4KO and HCTLR4KO mice, confirming the importance of TLR4 on hepatocytes for endotoxin clearance from plasma (Fig. 5B). Plasma IL-6 levels were also measured as an estimate of the inflammatory response to LPS (Fig. 5C). LPS injection led to similar circulating levels of LPS in WT and Flox mice at 6 and 24 h, with IL-6 levels higher at the 6-h time point, as expected (Fig. 5C). Also, as anticipated, global TLR4KO and LysMTLR4KO mice exhibited no increase in IL-6 in response to LPS in vivo (Fig. 5C), which was due to the lack of initiation of the NF-κB–signaling cascade by LPS in these strains. In contrast, IL-6 levels were significantly higher in the HCTLR4KO mice at both 6 and 24 h compared with the control strains (Fig. 5C). These data confirm that TLR4 on myeloid cells is essential for the systemic inflammatory response during endotoxemia. Taken together, our data show that TLR4 on hepatocytes is critical for the clearance of LPS from the circulation, and the failure of hepatocytes to clear LPS can result in an exaggerated systemic inflammatory response.

Our results show that TLR4 on hepatocytes is important for LPS clearance during endotoxemia. Furthermore, elevated LPS levels observed during CLP with antibiotics in the HCTLR4KO mice suggest that LPS clearance in polymicrobial sepsis is also dependent on TLR4 on hepatocytes. We reasoned that the more efficient clearance of bacteria in HCTLR4KO mice could result from the stimulatory effects of higher LPS levels detected by phagocytic cells early in the course of bacterial infection in this strain of mice. To explore this possibility, we first confirmed that phagocytic activity of macrophages could be increased by stimulation with LPS. To do this, we pretreated isolated WT peritoneal macrophages with LPS at concentrations between 1 and 100 ng/ml for 6 h prior to exposure to fluorescently labeled heat-killed E. coli bacteria for 2 h. As shown in Fig. 6A, macrophage phagocytosis was significantly increased above control levels in a concentration-dependent manner after LPS treatment.

Having shown that macrophage phagocytosis can be stimulated by low concentrations of LPS, we then examined the phagocytic activity of macrophages from WT, Flox, global TLR4KO, HCTLR4KO, and LysMTLR4KO mice using an in vivo phagocytosis assay. Mice were injected i.p. with LPS (5 mg/kg), followed 6 h later by i.p. injection of fluorescently labeled heat-killed E. coli. Peritoneal macrophages were harvested from these mice, and phagocytosis was determined by the amount of fluorescence in these cells. Compared with the control mouse strains, phagocytosis was depressed significantly in global TLR4KO and LysMTLR4KO mice (Fig. 6B). Phagocytosis in macrophages from HCTLR4KO mice was increased significantly compared with controls. Plasma LPS levels in HCTLR4 mice also were elevated, correlating with increased phagocytosis (Fig. 6C), as well as the increased bacterial clearance in these mice shown earlier. The response of macrophages to LPS was further confirmed as being TLR4 dependent, because global TLR4KO mice also had increased plasma LPS levels but failed to activate their TLR4-deficient macrophages and, therefore, exhibited suppressed phagocytosis (Fig. 6C).

Having shown the potential for increased phagocytosis ability in macrophages from HCTLR4KO mice in response to higher circulating LPS, we next sought to determine whether LPS treatment could result in increased bacterial clearance during CLP. WT mice were subjected to CLP without antibiotics and then injected i.v. with a low concentration of LPS (1 μg/kg) or saline (control) after 1 h. Bacterial counts at 18 h after CLP were significantly lower in the blood and peritoneal washout of CLP mice given LPS treatment compared with CLP mice that received saline (Fig. 6D). Decreased bacterial levels also correlated with lower systemic LPS levels (Fig. 6E) and lower levels of liver damage (ALT; Fig. 6F) at 18 h compared with saline-treated control CLP mice. Taken together, these data suggest that higher LPS levels enhance bacterial phagocytosis by macrophages and increase bacterial clearance in the presence of intact TLR4 on phagocytic cells.

The host reaction to polymicrobial sepsis involves an integrated response at both the end-organ and cellular levels. For infections that involve Gram-negative bacteria, TLR4 is central to this integrated response through the sensing of LPS (2). However, TLR4 is expressed on many cell types, including leukocytes and parenchymal cells (5–7). In this study, we used cell type–specific KO mouse strains to characterize the function of TLR on macrophages and hepatocytes, cell types thought to be integral to the host response to sepsis in a clinically relevant model of peritonitis (9, 28). We show that one of the main functions of TLR4 on myeloid cells is to enhance phagocytosis and clearance of bacteria. Importantly, we also show that the systemic inflammatory response to infection correlates with the efficiency of bacterial clearance and is independent of TLR4 in this infection model. Hepatocyte TLR4 is critical for the clearance of LPS in both polymicrobial sepsis and endotoxemia. Interestingly, the absence of TLR4 on hepatocytes actually improved bacterial clearance during CLP in the absence of antibiotics, and our further investigation showed that increased bacterial clearance could be recapitulated through the administration of small amounts of LPS early in the course of CLP sepsis. However, failure to clear LPS in HCTLR4KO mice resulted in greater inflammation and end-organ injury in CLP when antibiotics were also given. Thus, in the absence of antibiotics, TLR4 is essential to host responses through bacterial clearance. However, in the presence of antibiotics (a more clinically relevant situation), the efficiency of LPS clearance (and so TLR4 on hepatocytes) and the presence of even low levels of LPS may be more important.

The essential role of the TLR4/MD2 complex in the recognition of LPS in both in vivo and in vitro settings is well established (8, 20). Inflammation, tissue damage, and lethality are diminished in the absence of TLR4 signaling (29). However, in the setting of infection with live bacteria, the contribution of TLR4 varies depending on the experimental model and type of organism used in the study (13–15, 30–37). Most early studies used mouse strains with mutant forms of TLR4. C3H/HeJ mice express full-length TLR4, but they exhibit defective TLR4 signaling as the result of a single amino acid mutation in the cytoplasmic signaling domain of TLR4 (38). The C57BL/10ScN mouse strain, most commonly referred to as TLR4KO, expresses a truncated form of TLR4 protein that fails to recognize LPS (39). The clearance of live E. coli and neutrophil accumulation were found to vary considerably between these two TLR4-deficient strains of mice (32). Clear differences also exist between C3H/HeJ mice and TLR4KO mice that could impact the response of these mice strains to a bacterial challenge. For example, we showed that LPS uptake by hepatocytes in vitro, as well as the clearance of LPS in vivo, is intact in C3H/HeJ mice but defective in TLR4KO mice (9). Our studies showed that this difference is explained by a unique signaling complex on hepatocytes for the uptake of LPS that involves CD11b/CD18 and TLR4 but does not depend on signaling through the Toll/IL-1R domain of TLR4 (9). In the current study, we mitigate these strain differences by developing several strains of mice on a C57BL/6 background that do not express TLR4 either globally or on specific cell types of interest. TLR4 was deleted from myeloid cells (TLR4^loxP/loxP × LysM-cre) to assess the specific role of these leukocyte populations in the response to polymicrobial infections. TLR4 was also deleted specifically from hepatocytes (TLR4^loxP/loxP × Albumin-cre) to determine the importance of hepatocyte TLR4 and hepatocyte LPS uptake on the host response to sepsis.

CLP models are thought to be the most representative of sepsis commonly encountered by patients with polymicrobial i.p. infection associated with fecal contamination (40). CLP outcomes and responses vary considerably depending on needle puncture size, amount of cecum ligated, and the use of fluids or antibiotics (40), and this variability may be reflected in the conflicting results reported in TLR4-mutant and TLR4KO mice. Using CLP without antibiotics, Ayala et al. (41) showed reduced neutrophil influx in the lungs of C3H/HeJ mice. Dear et al. (42) administered antibiotics following CLP and found no difference between C3H/HeJ mice and WT mice at 12 h when measuring end-organ damage and TNF-α levels. However, Alves-Filho et al. (35) found that C3H/HeJ mice subjected to CLP without antibiotics exhibited improved survival compared with WT controls, with no difference between the strains with regard to neutrophil influx into organs. More recently, Castoldi et al. (43) carried out mild CLP without antibiotics and found that TLR4KO mice exhibited less kidney injury, a reduction in neutrophil influx into the kidney, and a diminished systemic cytokine response at 24 h. Two further studies (42, 44) also investigated the role of MyD88 and showed that MyD88KO mice are protected but TLR4 mutant mice are not protected during CLP. Taken together, these previous studies in models of CLP indicate that both the strain of the mouse and the use of antibiotic impact the consequences of TLR4 mutation or deletion in the setting of polymicrobial intra-abdominal infection. However, it is important to note that none of these previous studies assessed the importance of bacterial clearance in driving the extent of the host response. We are able to show, using the cell-specific TLR4KO mouse strains, that, in the absence of antibiotics, the efficiency of bacterial clearance determines both survival and the magnitude of the inflammatory response; this clearance is determined, in turn, by LPS recognition by TLR4 on macrophages. Neutrophil influx correlated with bacterial counts and did not depend on TLR4 expression. Neutrophil trapping of bacteria was shown to involve the TLR4-dependent release of DNA nets (45). Therefore, it is interesting to speculate that although neutrophil chemotaxis is intact in the absence of TLR4, net release may be impaired.

Our studies reveal the importance of TLR4 on macrophages for the stimulation of bacterial phagocytosis in the setting of polymicrobial sepsis. It was shown that TLR4 stimulation leads to enhanced phagocytosis of bacteria by macrophages (15). Furthermore, other investigators showed that agonists of TLR4 (33, 46) or other TLRs (46), given either as a pretreatment or early postinfection, enhance bacterial clearance and survival in models of sepsis. We confirm in our current work that LPS given 1 h after CLP improves bacterial clearance. Furthermore, we integrate these findings with our previous observations to show that the efficiency of LPS clearance by hepatocytes can impact bacterial clearance and that both LPS clearance and bacterial clearance are controlled through cell type–specific functions of TLR4.

The liver was identified as the primary site of LPS clearance in early studies (47, 48). Both Kupffer cells (49, 50) and hepatocytes (9, 51–53) were shown to participate in LPS clearance. A specific pathway for LPS deacylation in Kupffer cells was identified (54). Other work showed that LPS can interact with low-density lipoproteins, which leads to the uptake of the complex by the low-density lipoprotein receptor (55, 56). Our work establishes hepatocytes as the dominant cell involved in the TLR4-dependent clearance of LPS. It is likely that numerous mechanisms have evolved for the efficient clearance of LPS and that the cells and mechanisms involved may depend on the context. Small intermittent quantities of LPS emanating from the gastrointestinal tract could be removed by Kupffer cells, whereas higher boluses of LPS, as encountered during sepsis, seem to evoke TLR4-dependent clearance by hepatocytes.

Elevated LPS levels are correlated with worse outcome in severe sepsis patients (57). Efforts to neutralize (58) or clear LPS (59) improve hemodynamics, organ dysfunction, and mortality in septic patients. We showed that hepatocytes can become tolerant for LPS uptake by previous exposure to LPS through a suppressor of cytokine signaling–dependent mechanism (52). It is conceivable that a failure to adequately clear LPS in sepsis, resulting from downregulation of hepatocyte LPS uptake, could contribute to adverse outcomes. Synthetic TLR4 antagonists showed no clinical benefit in patients with severe sepsis (2, 60). It is tempting to speculate that this failure to improve outcomes in sepsis may be due to the inhibition of TLR4 functions that are critical to survival, such as enhanced bacterial phagocytosis. Our work showing major differences between sepsis with and without antibiotics highlights the consequences of failed LPS clearance in these two settings, with LPS driving bacterial clearance in the absence of antibiotics and stimulating the inflammatory response in the presence of antibiotics.

We thank Meihua Bo and Hong Liao for excellent technical support.

The authors have no financial conflicts of interest.