Blood-borne LPS is rapidly eliminated by liver sinusoidal endothelial cells via HDL

Abstract

During Gram-negative bacterial infections, excessive lipopolysaccharide (LPS) induces inflammation and sepsis via action on immune cells. However, the bulk of LPS can be cleared from circulation by the liver. Liver clearance is thought to be a slow process mediated exclusively by phagocytic resident macrophages, Kupffer cells (KC). However, we discovered that LPS disappears rapidly from the circulation, with a half-life of 2–4 minutes in mice and liver eliminates about three quarters of LPS from blood circulation. Using microscopic techniques, we found that ~75% of fluor-tagged LPS in liver became associated with liver sinusoidal endothelial cells (LSEC) and only ~25% with KC. Notably, the ratio of LSEC-KC associated LPS remained unchanged 45 min after infusion, indicating that LSEC independently processes the LPS. Most interestingly, results of kinetic analysis of LPS bioactivity, using modified limulus amebocyte lysate assay, suggest that recombinant factor-C, an LPS binding protein, competitively inhibits HDL-mediated LPS association with LSEC early in the process. Supporting the previous notion 3 min post-infusion, 75% of infused fluorescently-tagged LPS-HDL complex associates with LSEC, suggesting that HDL facilitates LPS clearance. These results lead us to propose a new paradigm of LSEC and HDL in clearing LPS with a potential to avoid inflammation during sepsis.

INTRODUCTION

During Gram-negative bacterial infections, lipopolysaccharide (LPS) activates the innate immune system. It does so by serving as a ligand for toll like receptor-4 (TLR4) and the TLR4 co-receptor MD2 on immune cells. The presumed LPS signaling pathway has been thought to be the following: LPS from bacteria interacts with serum proteins, LBP and CD14, and these complexes bind to and activate TLR4 and MD2. While the phosphates in lipid-A portion of LPS engage MD2, the binding of hexa-acyl-lipid A component to adjacent TLR4 induces TLR4 dimerization(1). This dimerization initiates the intracellular signaling cascades involving several signaling proteins and transcriptional factors(2). Although TLR4 innate immune signaling is critical for eliminating bacterial infection, it can also induce excess cytokine production with concomitant shock and sepsis-mediated death. In the course of sepsis, the host can quickly and proactively remove LPS from circulation, thereby removing the stimulus for TLR4 activation. This rapid and detoxifying LPS clearance mechanism remains poorly understood.

Although the LPS-mediated inflammatory response is systemic, LPS clearance occurs mainly in the liver as a host innate immune mechanism. Accordingly, the liver is the first organ exposed to gut-derived endotoxin. It has been shown that about 80% of intravenously infused rough LPS, a virulent form lacking o-chain is cleared by liver(3,4).

The clearance of LPS by liver is thought to be mediated exclusively by Kupffer cells (KC)(4–8) and hepatocytes(9). Macrophages in general are thought to dephosphorylate and deacylate lipid A after phagocytosis(10,11) inactivating LPS. Although such enzymatic mechanisms appear to be well established, macrophage-mediated uptake and modification of LPS-lipid A happens slowly, typically over hours. By contrast, LPS is eliminated very swiftly from the blood circulation(12), which suggests the involvement of an efficient, non-KC clearing system and a rapid LPS scavenger cell in liver.

Apart from KC, the other major scavenger in liver is the liver sinusoidal endothelial cell (LSEC)(13). We recently discovered the scavenging abilities of LSEC, possessing high endocytic ability, take up particles less than 200nm in diameter, such as blood borne viruses or small immune complexes(14,15). LSEC are well equipped for endocytosis, given that they express major endocytic receptors including those for mannose, collagen, hyaluron, L-SIGN, FcγRIIb and, most importantly, several types of scavenger receptors(16,17). Thus, whereas particles under 200 nm are eliminated from blood by the endocytosis function of LSEC, uptake and clearance of larger particles, such as bacteria or cells, requires the phagocytic action of macrophages.

LPS, which is 10–50nm, can be easily endocytosed(18). This raised the question about how it is actually cleared in the liver – by KC, LSEC, or both? Also not well understood is the actual process by which LPS is delivered to the liver cells where uptake occurs. What had been shown was that LPS incubated with serum could modify the endotoxic property(19). In addition, it had been found that most of the LPS in serum that is known to be captured by LBP gets very rapidly sequestered by lipoproteins(20,21) and majority of the LPS in plasma is bound to high density lipoprotein (HDL)(4). In human plasma, over 90% of LPS has been shown to be bound with lipoproteins; with highest affinity for HDL, medium affinity for low density lipoprotein (LDL), and low affinity for very low density lipoprotein (VLDL)(22). The differential affinity of LPS for lipoprotein is reportedly related to the amount of phospholipid in each of the lipoprotein classes(23). HDL contains the most phospholipid considering its surface area and particle numbers in human plasma and thus has the greatest LPS-binding capacity(24).

Once HDL binds to LPS the complexes show little or no stimulatory effect on cytokine production both in vitro and in vivo(25). Therefore, LPS association with plasma HDL has been assumed to neutralize or dampen the bioactivity of LPS. The LPS-lipoprotein interaction is very stable and is thought to occur through the lipid A portion of LPS interacting with phospholipid of lipoproteins(26). However the eventual fate of LPS-HDL complex is still unknown. Finally, how and to what extent lipoproteins contribute to the clearance of LPS is not clear. Understanding the elimination mechanisms might suggest ways to avoid TLR4 mediated inflammation.

We are led by the studies presented above to investigate whether LSEC are responsible for the rapid removal of LPS and also whether HDL facilitates clearance of LPS. In this manuscript, we present evidence that LSEC are the major LPS clearing cells and the HDLs are the plasma component that can transport LPS in plasma to LSEC, thereby facilitating the clearance process. These findings therefore suggest a novel conceptualization of LPS clearance.

MATERIALS AND METHODS

Ethics statement and animals

Male BALB/c and C57BL/6 mice of age 12–15 weeks old were obtained from Taconics and Jackson laboratory, respectively. All protocols were approved by The Ohio State University Institutional Animal Care and Use Committee. Bleedings were performed under Isofluorine anesthesia, and all efforts were made to minimize suffering.

Reagents

Highly pure rough LPS from E. coli K12 strain LCD25 was obtained from List biological labs. The FITC labelled LPS from E. coli 014 (Ra) and ³H/¹⁴C labelled LPS from Salmonella enterica sv. Typhimurium PR122 (Rc) were kind gift from Prof. Robert Munford (NIAID). Pooled normal human plasma was from Innovative Research. Endotoxin free water was from GE Healthcare Life Sciences. Fluorescein isothiocyanate (FITC), Alexa 488 hydrazide and Alexa 594 NHS were obtained from Molecular Probes. Purified apo-A1 from human plasma was obtained from Athens Research & Technology. The rabbit anti-apoA1 antibody was from Abcam. The TNF-α ELISA kit was from R&D. Chromogenic LAL Endotoxin Assay Kit was from Genscript. Recombinant human LBP protein was from R&D.

Cells

Raw 264.7 cells were obtained from American Type Culture Collection and were maintained in Dulbecco’s modified Eagle’s medium supplemented with 5% fetal bovine serum.

Preparation and infusion of LPS

Just before infusion, FITC-LPS and 488-LPS were resuspended into HNEB buffer (20mM HEPES, 150mM NaCl, 0.1mM EDTA, 0.3mg/ml BSA) to the corresponding concentration, and dispersed with Branson Sonifier 450. The mice were then infused intravenously with 10μg of FITC/488-LPS through tail vein.

Clearance kinetics of LPS

To study in vivo LPS clearance kinetics, we infused 7.32 μg ³H/¹⁴C labelled LPS from Salmonella enterica in 50 μl by tail vein infusion. The mice were bled via the retro-orbital plexus of ~10 μl blood at post-infusion times of 30 sec, 2, 5, 15, 30 and 60 min. To adjust for different body weights of individual mice, the blood concentrations of ³H/¹⁴C LPS (dpm/10 μl blood) were normalized to body weight as described before(14). To study in vivo clearance kinetics with non-radioactive LPS, we infused 10 μg 488-LPS in 50 μl HNEB buffer by tail vein infusion. The mice were bled via the retro-orbital plexus of ~10 μl blood at post-infusion times of 30 sec, 2.5, 5, 15 and 30 min. To adjust for different body weights of individual mice, the blood concentrations of 488-LPS (ng/10 μl blood) were normalized to body weight as described before(14). Since a plot of blood concentrations of 488-LPS indicated biphasic decay, we employed a biexponential decay model to fit the LPS concentration-time profile.

Organ distribution of LPS

To study in vivo distribution of LPS in various organs, we infused intravenously (retro-orbitally) Balb/c mice with 5μg ³H/¹⁴C labelled LPS from Salmonella enterica in HNEB buffer. After 10 min, the mice were bled via the retro-orbital plexus of ~20μl blood using heparinized capillaries. To account for different body weights of individual mice, the blood concentrations of ³H/¹⁴C LPS (dpm/10μl blood) were normalized to body weight as described before(14). Organs including liver, heart, lung, spleen and kidney were harvest at 10 min and weighed. Small portion of about 100 mg of each organ was homogenized in PBS containing 0.2% Triton X-100 and 5mM EDTA. Blood was lysed in PBS with 2% SDS and 4mM EDTA. Samples were then subject to scintillation counter to determine ³H/¹⁴C LPS radioactivity (dpm/mg). The dpm/mg tissues were then used to calculate the distribution of ³H/¹⁴C labelled LPS per organ.

Labeling of LPS with FITC

The FITC-LPS was obtained from Dr. Munford and also prepared using a modified protocol of Shao et al and characterized for bioactivity(5). The bioactivity of FITC-LPS was also tested in vivo. When infused intravenously FITC-LPS produced the same amount of endotoxic shock as unlabeled LPS of same concentration (data not shown).

Preparation of Alexa 488 LPS

LPS (2mg/ml) was monomerized by dissolving in 0.1% triethylamine (Sigma). After sonication, LPS was double oxidized with galactose oxidase (57) followed by NaIO4 (58). The oxidized LPS was dialyzed using 2 kDa dialysis cassette against PBS (pH 7.4) buffer overnight with a minimum of 4 volume exchanges. Further, The LPS was labelled by incubating with 200 μg/ml of Alexa-488 hydrazide (Life Sciences) for 1h at 37°C, followed by overnight incubation at 4°C. The Alexa-488 labelled LPS product dialyzed using 2 kDa dialysis cassette against PBS at 4°C to remove free Alexa-488.

Limulus Amebocyte Lysate (LAL) Assay

The samples were quantified for the presence of endotoxin/LPS using the lysate of limulus amebocytes (LAL). Briefly, endotoxin samples and LPS standards were diluted to the concentration between 0.1~1 EU, and incubated for 10 minutes with LAL lysates at 37 °C. Samples with LAL lysates were then treated with chromogenic reagent followed by color stabilizer reagents. Endotoxin levels were determined by absorptions at 545 nm.

Cytokine measurement

RAW 264.7 cells were stimulated with different preparations of LPS and LPS HDL complexes for 5 hrs. Cell supernatants were harvested, centrifuged to remove dead cells, and analyzed by ELISA using TNF-α specific kits from R&D Systems.

Immunofluorescence

Liver pieces of about 5 mm were fixed in 4% paraformaldehyde-PBS for 2 hrs at room temperature and were infused in 20% sucrose-PBS overnight at 4°C after washing with PBS. The tissues were then embedded in a tissue freezing medium and stored at −80°C. For IF analysis cryostat sections of 5 μm thickness, were blocked in 5% milk-PBS followed by incubation with primary antibody overnight at 4°C. The primary antibodies were rabbit anti-mannose receptor (Santa Cruz) and mab F4/80 at a concentration of 20 μg/ml. After 3 washes with PBS the sections were incubated with fluorescence tagged secondary antibodies for 1hr at room temperature. Nuclei were stained with DAPI for 10 minutes, and the sections were mounted under cover slips in Prolong gold (Invitrogen). Control incubations included isotype controls along with their respective secondary antibodies and also secondary antibodies alone. The images were acquired in the Olympus FluoView 1000 Laser Scanning Confocal microscope equipped with a spectral detection system for a finer separation of fluorochromes (FV 1000 spectra) using 60× oil immersion lens at room temperature. Image analyses were done using Image J software.

Tissue LAL assay

The LAL assay to detect biologically active LPS in liver tissues, obtained from mice that were previously infused with FITC-LPS, were done using a modified protocol of Uragoh et al(59). Briefly, the liver tissues were blocked with 10% goat serum followed by incubation with recombinant factor C (Prof. Deng, Singapore) for 1 hr at room temperature. After a brief wash with PBS (pH 7.4) the sections were incubated with rabbit anti-factor C antibody (Prof. Kawabata, Kyushu University) at 4°C overnight. The sections were then washed for an hour and then incubated with Alexa 568 goat anti rabbit IgG for 2 hrs. Nuclei were stained with DAPI for 10 min, and the sections were mounted under cover slips in Prolong gold (Invitrogen). Control incubations included isotype controls along with their respective secondary antibodies with or without recombinant factor C.

Quantitative Microscopy

The relative association of FITC-LPS and 488-LPS between LSEC and KC was quantified from merged IF images dual-stained for LSEC using MR and KC using anti-F4/80 for KC in the following manner. The total intensity of FITC-LPS/488-LPS was calculated using Image J software. The intensity of FITC-LPS/488-LPS associated with KC was measured by cropping the individual KC, identified with F4/80. The FITC-LPS/488-LPS associated with LSEC in an image were obtained from subtracted values of total intensity minus the sum of intensities from all cropped KC. Thus, total intensity of FITC-LPS/488-LPS minus FITC-LPS/488-LPS of KC equaled the FITC-LPS/488-LPS intensity of LSEC. Optical sections with a total area of 14 mm² and 8 mm² were analyzed from each of three different mice.

The relative localization of FITC-LPS and 488-LPS between LSEC and KC over the course of time were obtained as was quantified as described for Fig. 2 from an area of 9 mm² and from a total of 3 mice at time 1, 15 and 45.

Fig 2. In liver, FITC-labelled rough LPS is distributed primarily to LSEC.

Four color fluorescence microscopic image of liver from mice infused 1 minute earlier with 7.5 μg of FITC-LPS. a. DAPI showing cell nuclei (blue). b. Green puncta identify FITC-LPS particles. c. Rabbit IgG anti-MR mark LSEC.in red d. Magenta color defines the KC labelled with mab F4/80. e. Merged panels of a,b,c and d. f. Merged panel e plus Differential Interface Contrast (DIC) defining tissue structure including sinusoidal lumens. The bars in the panel d indicate 10 μm. The image presented here is a representative of 30 images of 3 different mice from 3 experiments, which are quantified in Fig 3.

Quantification of modified LAL assay were done using IF signal of factor-C binding and FITC-LPS binding to liver by measuring the area and the mean fluorescence intensity of the green pixels (FITC-LPS) and red pixels (factor-C).

Isolation of HDL from human plasma

The HDL isolation was performed as described previously by us(60). Briefly, the density of the human plasma was adjusted to 1.22 g/ml using potassium bromide (KBr) and ultracentrifuged at 450,000 × g for 3 hrs at 4 °C. Once separated, 200 μl of top layer containing all lipoproteins was collected. The HDL was separated from the total lipoproteins by overlaying 0.15 M NaCl and centrifuging at 625,000 × g for one hour at 4°C. The 3rd layer from top was collected and characterized to be HDL.

Preparation of 594-HDL

The HDL (10 mg/ml) was dialyzed against borate buffer (pH 8.3) using 20 kDa cut-off dialysis cassette for 1 hr. The HDL was labelled by incubation with 25 μg of Alexa 594 NHS (10 mg/ml in anhydrous dimethylformamide) overnight at room temperature with constant mixing. The free Alexa-594 dye was removed by dialysis using 20 kDa cut-off dialysis cassette against 50 mM HEPES buffer for 2 hrs at 4°C. The protein content of the HDL was calculated using the absorbance at 280 nm and confirmed by Bradford protein assay. The purity of HDL fractions was confirmed by SDS-PAGE electrophoresis and immunoblot analysis using anti-apoA1 antibody (Abcam) and then stored at −80°C until use.

Preparation of 594-HDL-488-LPS conjugates

The 594-HDL and 488-LPS complex was prepared by incubating 1.5 mg and 200 μg, respectively along with 6μg lipopolysaccharide binding protein (LBP) in 100mM HEPES buffer (pH 7.2) and incubated overnight at room temperature. To separate the unbound LPS from the complex, the density of the reaction mixture was adjusted to 1.22 mg/ml with KBr, and centrifuged at 500,000 g for 75 min. The top layer containing 488-LPS and 594-HDL complex was collected and dialyzed using 20 kDa cut-off Slide-A-Lyzer Dialysis cassettes against PBS pH 7.4.

Size exclusion chromatography analysis

HDL and 594-HDL-488-LPS conjugates were analyzed using high resolution FPLC on Superose 6 Increase 3.2/300 column and preparatively-purified on Superose 6 10/300 GL column (both from GE Healthcare Life Sciences). The FPLC setup consisted of Bio-Rad DuoFlow equipped with C96 autosampler with cooling, and BioFrac fraction collector (all from Bio-Rad). Lipoproteins were eluted at a flow rate of 0.05 mL/min in all analytical FPLC runs and 0.5 mL/min in all preparative runs. All lipoproteins were eluted isocratically using 10 mM phosphate buffer and 154 mM NaCl at pH 7.4. Fractions were collected by automatic fraction collector at 4°C (Bio-Frac) in 1.2 ml sized micro-titer tubes, followed by analysis of fluorescence and absorbance on the plate reader (Molecular Devices SpectraMax i3). FPLC chromatograms were then plotted as fluorescence intensity (or absorbance) against fraction number.

Immunoblot

The VLDL, LDL and HDL preparation isolated from plasma (20 μg) were separated on 8–16% gradient SDS-polyacrylamide gels, and the proteins were transferred to nitrocellulose membranes (0.45 μm). The membranes were blocked with 5% milk for 30 min and then incubated overnight with primary rabbit anti-apoA1 antibody (Abcam) at a concentration of 1:300 at 4°C. The bands were developed using HRP-conjugated goat anti-rabbit IgG antibody at a concentration of 1:2000 and detected for quantification using a laser scanner (Pharos-FX, Bio-Rad). The apoA-1 standard from Athens Research & Technology (GA) was used as a reference.

Competitive Inhibition assay of FITC-LPS with unlabeled LPS

Competitive inhibition studies were carried out using 488-LPS (1 μg) and varying amounts (0, 1, 10, 30, and 50 μg) of unlabeled LPS as the competitive inhibitor in RAW 264.7 cells. Briefly, in 24-well plate, RAW cells (5 × 10⁵/well) were plated on Poly L-Lysine pre-coated coverslips via incubation in serum free RPMI 1640 media for 4 hrs. The cells were then incubated with 488-LPS and unlabeled LPS in serum free RMPI media for 10 min. After incubation and brief wash with PBS, the cells were fixed in 4% paraformaldehyde for 20 min, and then washed with PBS. The nuclei were then stained with DAPI (100 ng/ml) for 10 min, followed by PBS wash for 10 min. The coverslips were then mounted to a slide with prolong gold mounting media and the images were acquired in the Olympus FluoView 1000 Laser Scanning Confocal microscope.

To quantify the relative proportion of FITC-LPS that binds to RAW cells, we employed Image J software developed in JAVA. After thresh-holding the entire image to separate 488 pixels from background pixels we determined the total area and mean fluorescence intensity of the pixels within an image. The pixel intensity associated with one cell is then calculated by dividing the area × MFI by the number of cells counted using DAPI staining.

Statistical analysis

Random-effects linear regression was used to estimate difference in TNF-α production across group in Fig S4 (Resting, 488-LPS, 488-LPS+LBP, and 488-LPS+594-HDL+LBP) where the random term was the experiment. This allows within and between experiment variability when estimating the standard error used to generate p-values. For all other statistical analysis a paired Student’s t-test was used for each statistical comparison and a value of p ≤ 0.05 was considered significant. Random-effects linear regression and Bonferroni’s method was used to adjust the p-values in order to conserve the overall type 1 error rate at 0.05 for the study. All analyses were run using Stata 14.1, StataCorp LP, College Station, TX and GraphPad Prism 6 version.

RESULTS

Rough LPS is cleared efficiently in mice

We measured the LPS clearance rate using Alexa 488 labelled E. coli K12 strain LCD25 LPS and ³H/¹⁴C labelled Salmonella enterica sv. Typhimurium PR122 (Rc) LPS(11), both are rough form of LPS with predicted labeling positions (Fig S1). The clearance kinetics of intravenously infused LPS was plotted based on the disappearance of 488-LPS (Fig 1) and decay of β-irradiation from peripheral blood measured periodically over 60 min (Fig. S2). The LPS concentration at time zero was calculated as the dose divided by the blood volume. The LPS concentrations at experimental time points were determined from interpolating fluorescence intensity from plasma into standard curve from 488-LPS diluted in plasma. We noted a rapid bi-exponential clearance of both types of LPS from the bloodstream, with ~80% cleared in 5 minutes giving a half-life (t_1/2) of 2–4 min (Fig 1A) using assumed blood concentration at 0 min as 100%. About 70% cleared in 5 min using concentration at 30 sec as 100% (Fig 1B). The clearance of LPS was significantly faster during the alpha phase (within 5 minutes) than the beta phase (5–60 minutes). Similar decay kinetics, with a t_1/2 life of 2–4 min, were found between C57BL/6 mice and BALB/c (Fig S2). These results suggest that the known difference in sensitivity to LPS between these mouse strains, is not due to a difference in clearance(27). The rates of clearance of ³H were not different from ¹⁴C counts (data not shown) assuring the reliability of the reagent.

Fig 1. Rough LPS is cleared efficiently in WT BALB/c mice and liver is the major organ involved in clearance.

We infused via the tail vein 488-LPS and then evaluated the clearance from peripheral blood. Panel A: The curve plots the percentage of 488-LPS in 10 μl of blood vs time by keeping time 0 as 100%. The LPS concentration at zero time is calculated as the dose divided by the blood volume. Panel B: The curve plots the percentage of fluorescence from 488-LPS in 10 μl of blood vs time by using the fluorescence from first time point namely 30 sec as 100%. Each data point represents mean percentage ± SD of three mice. Panel C and D: Organ distribution of ³H/¹⁴C LPS showing radioactivity in percentage of infused dose (panel C) and radioactivity in dpm (panel D) in various organs. The data is from 3 different mice representing 3 experiments. Asterisks signify statistically significant differences, p<0.05, using Student’s t-test.

Liver is the major organ involved in LPS clearance

We quantified LPS in various organs to access the distribution kinetics in early time points. After 5 min of ³H/¹⁴C labelled Salmonella enterica LPS infusion, we autopsied 3 mice and quantified radioactivity in major organs. Of the total administered dose we recovered 74±21% in liver, 0% in lung, 4±3% in spleen, 0% in kidney, and 11±5% in blood. Approximately 11% of the infused radioactivity was not accounted for. The amount of LPS in blood is similar to the LPS concentration at 10 min in Fig 1A and Fig 1B, assuring the reproducibility of our experiments. Thus, three quarters of infused radioactive LPS was found in the liver (Fig 1C and Fig 1D).

In liver, LPS is distributed primarily to LSEC

Noting that liver clears LPS(3,4) and wanting to determine the cell type(s) responsible for the rapid uptake, we examined paraformaldehyde fixed liver sections by fluorescence confocal microscopy at the very early time of 1 min post infusion. To visualize LPS, it was fluor-tagged with FITC or Alexa 488 LPS (488-LPS). The predicted labeling site for FITC in LPS is a primary amine connected with phosphates from the heptose of Kdo, and in 488-LPS, the diol groups from galactose, heptose and Kdo (Fig S1). The labeling procedure did not alter the bioactivity (Fig. S4A and B). As seen here, unlabeled LPS competed with 488-LPS, indicates that these molecules bind to same receptor in the cell. Furthermore, in RAW 264.7 cells, both FITC-LPS and 488-LPS induced similar levels of cytokine production as did their respective unlabeled LPS controls (Fig S4A and B), thus validating that the labeling procedure did not alter the functionality of these reagents. We then infused these reagents intravenously and analyzed the liver. In the microscopic images of liver at 1 minute after infusion with FITC-LPS (Fig 2) and 3 minutes after 488-LPS (Data not shown) infusion, liver images showed abundant green punctas lining the sinusoids; which are very obvious when overlaid with DIC. The LSEC were then mapped red with anti-MR, while KC was marked with anti-F4/80 (magenta). The bulk of the LPS punctas perfectly align with the LSEC marker, whereas lesser amounts appear to be localized to KC. No LPS was associated with hepatocytes or larger vein endothelial cells.

Quantification confirmed that LPS localized predominantly to LSEC

Quantification of FITC-LPS/488-LPS from 4-color fluorescence images similar to Fig 2 confirmed LSEC as the predominant site of LPS localization in the liver tissue sections we examined. We quantified the overall percent of liver LPS associated with both LSEC and KC (Fig 3). We found that 74±12 and 80±13% of FITC-LPS and 488-LPS, respectively localized to LSEC while only 26±12% and 21±13% of FITC-LPS and 488-LPS associated with KC, authenticating our visual impressions. The two different fluor-tagged LPS compounds behaved similarly with respect to their distribution between LSEC and KC (Fig 3A and 3C), although the intensity of fluorescence varied (Fig 3B and 3D).

Fig 3. Quantification validates predominant LSEC localization of LPS.

Panel A and B. Quantification of FITC-LPS from 4-color fluorescence images similar to Fig 2 by measuring the total pixel area and the mean fluorescence intensity of green puncta associated with LSEC and KC markers. Panel C and D. Quantification of 488-LPS from 4-color fluorescence images similar to Fig S3, by measuring the total pixel area and the mean fluorescence intensity of green puncta associated with LSEC and KC markers. Panel A and C. The bar graphs show percentage of FITC-LPS/488-LPS ±SD from 3 different experiments. Panel B and D. The bar graph shows area × mean fluorescence intensity ± Standard deviation (SD) of FITC-LPS/488-LPS from 6 mice, 3 for FITC-LPS and 3 for 488-LPS from 3 different experiments. The area of tissue examined microscopically were 14 and 8 mm², respectively for FITC-LPS and 488-LPS. Asterisks signify statistically significant differences, p<0.05, using Student’s t- test.

LSEC-associated LPS is not taken up by KC

We tested whether LPS that associated with LSEC then is taken up by KC. To do so, we analyzed LPS in images similar to the ones in Fig 2 at various time after infusion. Relative location of FITC-LPS in mouse livers at 1, 15 and 45 min after infusion between LSEC and KC suggests that LSEC associated LPS did not move to KC until 45 min post-infusion. The percentage of co-localization relative to LSEC and KC in bar graph shows that the percentage of LPS associated with LSEC and KC is not significantly different over the course of time (Fig 4), suggesting that LSEC processes the LPS independently of KC.

Fig 4. LSEC associated LPS is not taken up by KC.

Quantification of FITC-LPS in mouse livers at 1, 15 and 45 min after infusion using 4-color fluorescence images similar to Fig 2, by measuring the total pixel area and the mean fluorescence intensity of green puncta associated with LSEC and KC markers. The bar graphs show relative percentage of FITC-LPS association ± SD, between LSEC and KC from each of 3 different mice from 3 different experiments for each time point. Lack of asterisk signify lack of any statistically significant differences P<0.05, using Student’s t- test.

LPS associated with LSEC does not bind to factor C until 15 minutes after infusion

Having found that the majority of LPS associated with LSEC (Fig 3) remains associated as late as 45 min (Fig 4), we tested whether the LPS bioactivity decreases over the course of time. The bioactivity of FITC-LPS was assessed based on its ability to bind to the endotoxin binding protein, recombinant factor C (rFc), in tissue sections. We did this using a modified Limulus Amebocyte Lysate (LAL) assay, a highly sensitive primary assay to localize biologically active LPS based on its rFc-binding ability(28). rFC binding to LPS was visualized using anti-rFc Ab followed by fluor-tagged secondary Ab. We then studied LPS bioactivity in vivo in liver tissue previously infused with FITC-LPS and fixed after 1, 15, and 45 min (Fig 5A). Green puncta identified FITC-LPS particles and the red puncta identified the factor C binding to FITC-LPS. Quantification of total pixel area and the mean fluorescence intensity of green puncta and red puncta suggest that at 1 min, factor-C binds significantly less FITC-LPS compared to 15 and 45 min (Fig 5C), although FITC-LPS association to LSEC is not different (Fig 5B). Binding of factor C at 45 min with FITC-LPS suggests that lipid A portion of LPS is not chemically modified and is still biologically active.

Fig 5. LPS associated with LSEC does not bind to factor C at 1 min after infusion.

The biologically activity of FITC-LPS associated with LSEC was detected in vivo in liver tissue previously infused with FITC-LPS and fixed after 1, 15 and 45 min, using modified LAL assay with factor C and anti-factor C antibody. The green puncta identify FITC-LPS particles and the red puncta identifies the factor C binding to FITC-LPS.

Panel A Quantification of FITC-LPS signal from 2-color fluorescence images, similar to Fig 5 is shown here. The total pixel area and the mean fluorescence intensity of green puncta were measured and plotted.

Panel B. Quantification of factor-C binding signal from 2-color fluorescence images, similar to Fig 5 is shown here. The total pixel area and the mean fluorescence intensity of red puncta were measured and plotted.

The bar graphs show the Area × Mean fluorescence intensity ± SD from each of 3 different mouse from 2 experiments. The area of tissue examined microscopically totaled 14 mm². Asterisks signify statistically significant differences, p<0.05, using Student’s t- test.

HDL Isolation and characterization of LPS-HDL

To determine whether LPS that is associated with LSEC is also associated with HDL, we prepared LPS-HDL complexes. We first isolated HDL from lipoprotein fractions of human plasma using density gradient ultra-centrifugation. The fraction that contains HDL was identified using western blotting for the major protein component of HDL, namely apoA-1, using native human apoA1 as a positive control(29). The western blot results (Fig S3A) showed a single apoA-1 band from HDL co-migrating with the apoA-1 standard, confirming the purity in our preparation. Also unlabeled HDL was able to compete with Alexa 594-HDL (594-HDL) in liver suggesting that labeling did not alter the biochemical nature of HDL (Fig S3C). We then prepared 488-LPS and 594-HDL complexes as described in Materials and Methods section and analyzed them by means of high-resolution fast protein liquid chromatography (FPLC) on a Superose 6 Increase 3.2/300 size exclusion column, which separates lipoproteins based on size. The overlay of 488-LPS peak and the 594-HDL with 280 nm absorption peak from apo-A1 protein of HDL (Fig 6A) demonstrates the association of 488-LPS with 594-HDL. The late fluorescent peaks from fractions 30~40 (Fig 6A) represents free 488-LPS and free Alexa 594 dye with smaller size. Given the presence of free starting material (488-LPS and free Alexa 594 dye) in raw preparation, we decided to purify the complex using preparative FPLC. The injection of raw complex into Superose 6 Increase 10/300 GL column and collection of fractions 17 to 26 accomplished this goal. The FPLC profile of purified 594-HDL-488-LPS complex (Fig 6B) represents a single 594-HDL peak and 280 nm protein absorption peak overlaying with 488-LPS, which is a major separated fraction, representing about 95% of the total quantified material. However, a late Alexa 488 peak (Fig 6B) indicated that a small fraction of 488-LPS was present in the preparation, indicating that the complex might have dissociated while moving through the gel of Superose column, or due to other processing (see Materials and Methods). Another possibility is that the removal of excess of LBP during chromatographic separation might have caused slight destabilization of the complex, which resulted in departure of a few 488-LPS molecules from the 594-HDL-488-LPS adduct. The bioactivity or TNF-α inducing abilities of these complexes were measured by ELISA and our data (Fig S4C) confirmed LBP significantly enhance the TNF-α producing ability of LPS(30) and addition of HDL significantly inhibited, as reported earlier(31,32).

Fig 6. FPLC of Superose-6 size exclusion elution profiles of LPS-HDL complex.

Panel A: The FPLC profile shows the size exclusion chromatographic analysis of 488-LPS with 594-HDL. Panel A shows the elution profile of LPS-HDL complex before purification. Fractions of 17 to 26 from Panel A were collected and concentrated to achieve purified 488-LPS and 594 complexes. Panel B: Shows the elution profile of Alexa 488-LPS-Alexa 594-HDL complex from Superose 6 Increase 3.2/300 analytical column. The chromatogram shows the overlay of absorbance at UV 280 nm from HDL proteins and fluorescence emission spectra at 488 nm excitation from Alexa-488 LPS and 594 nm excitation from Alexa 594-labeled HDL. The data are representative of 3 independent preparations/experiments.

488-LPS-594-HDL complex localize chiefly to LSEC

Having demonstrated that FITC-LPS localized predominantly to LSEC (Fig 2) and LPS associated with LSEC could be co-associated by HDL within 15 minutes (Fig 5), we then wanted to determine whether LPS-HDL complex localizes to LSEC. We did that by examining fixed liver sections previously infused with 488-LPS and 594-HDL complexes at very early time of 3 min post-infusion. Microscopic images of liver from infused mice showed abundant green dots lining the sinusoids which are very evident when overlaid with DIC. The bulk of the LPS puncta and HDL puncta (Fig 7 A) perfectly align with LSEC marker (MR), whereas fewer numbers appear to be localized to KC (F4/80). Interestingly, no LPS-HDL complex was associated with hepatocytes or in the lumen. Around 30 images, similar to those in Fig 7A, were quantified to determine the percentage of 594-HDL-488-LPS complex associated with KC and LSEC (Fig 7B). Our data indicates that ~75% of 488-LPS associated with LSEC while only 25% with KC. The 594-HDL shows similar distribution with 78% in LSEC and 22% in KC. Quantification of co-localization using Pearson’s analysis reveals that around 80% of 488-LPS associates with 594-HDL.

Fig 7. 488-LPS- 594-HDL complex localize chiefly to LSEC.

Panel A: Four color fluorescence microscopic image of liver from mice infused 3 minutes earlier with 488-LPS-594-HDL complex. a. Magenta color delineates the KC. b. Green puncta identify 488-LPS particles. c. Red puncta identifies 594-HDL. d. Blue color shows Rabbit IgG anti-MR marking LSEC. e. Merged panels a, b, c and d. f. panel e plus DIC defining tissue structure including sinusoidal lumens. The bars in the panel d column indicate 10 μm.

Panel B: Quantification of Alexa 488-LPS and Alexa 594-HDL signal from 2-color fluorescence images, similar to panel A is shown here. The total pixel area and the mean fluorescence intensity of green puncta and red puncta were measured and plotted. The bar graphs show percentage and mean ± SD from each of 3 different experiments/mouse. The area of tissue examined microscopically totaled 8 mm². Asterisks signify statistically significant differences, p<0.05, using Student’s t-test.

Animations of 3-color images of liver sections projected in 3 dimensions further documents LPS-HDL complex predominate association with LSEC (Sup video 1) and to a lesser extent with KC (Sup video 2). LPS-HDL associated with LSEC appears as individual yellow puncta signifying co-localization of green (LPS) and red (HDL) in closer vicinity, whereas in KC it appears as clumps of yellow puncta (Sup video 2).

DISCUSSION

Our results lead us to draw three major conclusions. The key finding and first major observation is that circulating LPS is rapidly cleared from the circulation, with a half-life between 2–4 min, and ~80% of infused LPS is cleared within 5 min of infusion. The fast decay kinetics that we observed are similar to findings from other laboratories, although the kinetic calculations are not available(33–35). It is important to note that the remaining ~20% of LPS could be bound to immune cells such as monocytes, tissue macrophages, neutrophils, or platelets and therefore potentially involved in signaling. The amount of LPS cleared in 5 min is remarkably huge, about 650 fold more than what is known to cause inflammation, i.e. 8100 ng LPS cleared per mouse in 5 min relative to 12.5 ng of LPS that can induce inflammation in mice(36,37).

In accordance with the previous literature our data shows that the liver is the major organ involved in clearance of LPS(3,4). Accordingly, the second of our key findings involves the distribution and localization of LPS within liver. These experiments were facilitated by being able to distinguish LSEC from KC microscopically(14). Published reports on LPS clearance have concluded that KC are the major scavenger cell type involved in clearing endotoxin(4,5,7). This conclusion was based on localization of infused LPS using radiolabelled or fluor-tagged LPS in liver several hours or days after infusion, when majority of the LPS is already cleared and only LPS involved in signaling may remain. It is clear from our experiments, which looked at very early post-infusion time points, that LSEC is the major cell type for clearing LPS in liver. We are limited by the techniques used in the study to distinguish whether the cleared LPS is internalized or bound to the membrane of the LSEC. Thus we did not specify but rather used the terms clearance or elimination of LPS from the blood circulation and subsequent uptake by LSEC. In addition, KC also appears to be involved in clearing LPS to a minor extent during the early time points (5). Furthermore, it is possible that the LPS associated with LSEC is inactivated over the course of time, and the structurally and functionally altered LPS(38) could be processed eventually after several hours by hepatocytes (9,39). Moreover, the aggregation status of the LPS seems to be a major determinant of clearance (4), and we have made all necessary efforts to minimize LPS aggregation as described in the materials and methods section. We point out that the clearance phenomenon with monomeric and to some extent aggregated LPS shown in our study is different from previous published studies that used LPS aggregates or bacterial membrane fragments/blebs or bacterial outer membrane vesicles(20,40–42). Additionally, our results show that the LSEC-associated LPS is not taken up by KC over the course of time. To assess that possibility, we used a modified LAL assay (Fig 5A). We predicted that rFc binding to FITC-LPS will decrease over time. Surprisingly, contrary to our prediction, the LAL assay showed substantially less rFc binding to FITC-LPS at 1 min, but significantly more rFc binds to LPS at 15 and 45 min (Fig 5 B and C). These results suggest that a plasma component may mask or compete with LPS from binding to rFc. The plasma inhibitor of LPS binding to rFc was reasoned to be HDL, as shown earlier(43). Moreover, the role of HDL in exerting protection against sepsis possibly through clearance was shown earlier by Guo et.al using ApoA-I-KO mice(44).

In our test to determine whether LSEC-bound LPS is associated with HDL, we made our third major novel observation, namely that fluor-tagged LPS-HDL complex localizes to LSEC at 3 min after infusion. A number of clinical studies have indicated that low plasma HDL-cholesterol levels are associated with an increased risk for severe sepsis(45). Moreover, humans with low HDL levels have a more robust inflammatory response to LPS administration. Finally, raising the HDL concentration in mice has been shown to have a protective role against endotoxin in vivo (46) and transgenic mice with elevated HDL concentrations are resistant to endotoxin challenge(46). These results support our findings that the level of HDL directly correlates with LPS clearance and, therefore, HDL could be a cargo carrier of LPS, delivering it to the liver for eventual detoxification or inactivation. Although HDL-mediated protection has been attributed to the neutralization of LPS, the dissociation rate of LPS from LPS-lipoprotein complex is significantly slower than the clearance rate of lipoprotein. Thus, HDL promotes the elimination of LPS and thus protects the host from LPS-induced injury. The observed reduction of LPS bioactivity could stem from enhanced uptake and elimination of LPS and lipoprotein functions as a liver-targeted delivery system. This finding supports the idea that lipoproteins clear LPS from the circulation via delivery to the liver where lipoprotein metabolism also occurs. Most interestingly, liver is also the major organ that takes up majority of the aggregated LPS and LPS-HDL complexes(11). Although it is now recognized that HDL can carry several lipids, proteins and micro-RNAs(47); the current study specifically demonstrated that HDL transports LPS, revealing the role of HDL in innate immune functions of the host.

Based on our findings, the predicted overview of LPS uptake mechanism by LSEC is shown in Fig 8. The binding and uptake of LPS by LSEC involves the participation of HDL. HDL sequesters LPS at early post-exposure times, i.e. between 1–15 min, and this interaction facilities the uptake of LPS by LSEC. After 15 min HDL separates from LPS, either inside or outside the plasma membrane of LSEC, when binding partners such as factor C get access to LPS. Apart from LPS presented by HDL, it is possible that aggregated LPS alone can be endocytosed. Early literature demonstrated that the endocytosed LPS delivered to lysosomes is not inflammatory(10,41,48). The binding of 488-LPS and 594-HDL blocked by unlabeled LPS and HDL (Fig S3B and S3C) in our experiments suggest involvement of an endocytic receptor.

Fig 8.

Schematic model for LPS inactivation by LSEC.

The receptor that could facilitate the uptake of LPS in LSEC could be the HDL receptor Scavenger Receptor-B1 (SR-B1). Lipoproteins like HDL are well-known transporters of lipid-rich molecules and are recognized by the cells of the liver by lipoprotein receptors. The liver is also the major organ that expresses SR-B1, and at the same time the major organ involved in clearance of LPS. The SR-B1-null mouse has been shown to have accentuated inflammatory response to LPS compared to wild type strains (49,50). It has been suggested that hepatocyte and macrophage SR-B1 play a role by either clearing or responding to LPS(49). However, we have recently demonstrated that in liver, HDL receptor SR-B1 is expressed abundantly in LSEC but is barely perceptible on hepatocytes(17). These findings offer an irresistible clue that SR-B1 in LSEC plays a role in inflammation by eliminating LPS. In addition, TLR4 was shown not to be involved in uptake of LPS in both monocytes and endothelial cells(51)in vitro. Published work indicates that liver expresses less TLR4 than other organs(52). Further study is warranted to confirm TLR4 expression in LSEC and whether it is involved in clearing LPS, either alone or with scavenger receptors(51,53,54). The other known LPS receptor Caspase 11(55) is likely not involved in LPS clearance because of its intracellular expression, and also there is no sign of apoptosis in LSEC at the time points we have studied.

LPS taken up by the LSEC may undergo one of two known modes of enzymatic inactivation, namely dephosphorylation of the phosphates(10) and deacylation of the primary acyl components of lipid A, mediated by the enzyme acyloxyacyl hydrolase(56). After identifying the major cell type involved in LPS clearance, current studies in our laboratory are attempting to clarify further potential mechanisms such as: receptors, enzymes and the signaling molecules that are involved in LPS clearance by LSEC.

Our research, thus far done only in mice, has potential relevance to human sepsis as the LPS association with lipoproteins also occurs in the decreasing order of HDL>LDL>VLDL in both human and mouse.

ABBREVIATIONS

LPS

Lipopolysaccharide

LSEC

liver sinusoidal endothelial cells

KC

Kupffer cells

DIC

differential interference contrast

IF

immunofluorescence

RFI

relative fluorescence intensity

HDL

High density lipoprotein

rFc

recombinant factor C

LAL

Limulus Amebocyte Lysate

TLR4

Toll like receptor4

SR-B1

Scavenger Receptor-B1