INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The liver plays a central role in regulating multiple host defense, immunologic, biochemical, and metabolic functions during gram-negative sepsis. Critical hepatic responses to gram-negative endotoxemia include synthesis and export of TNF- by Kupffer cells as well as acute-phase proteins by hepatocytes, and hepatobiliary clearance of cytokines and secondary eicosanoid mediators (1). Preexisting liver dysfunction has been linked to enhanced pulmonary inflammation, the acute respiratory distress syndrome (ARDS), and multiple organ failure (3). De novo interactions between the liver and lungs early after endotoxemia are thought to be responsible, particularly those altering the kinetics of circulating endotoxin, TNF-, and IL-6, as well as compartmental leukotriene (LT) metabolism (3). However, neither the extent nor the mechanisms by which different types of liver dysfunction modulate organ injury during endotoxemia are understood (2, 3, 6).

To define threshold conditions and mediator pathways by which liver-lung interactions during endotoxemia lead to pulmonary injury and shock in subjects with compromised hepatic function, models of liver impairment have been developed. Acute, reversible disruption of hepatocellular performance simulating viral hepatitis by the selective toxin D-galactosamine (D-Galn) (8), and hepatic hypoperfusion with preserved cellular function by portacaval shunting (9) strongly suggest the importance of a bidirectional TNF-:LT axis of inflammation in endotoxin-induced lung injury and mortality (3, 9). Endotoxin-initiated influx of neutrophils (PMNs) into the distal air spaces was amplified in both models, reflected by increased recovery of PMNs in bronchoalveolar lavage fluid (BALF), and correlated with elevated BALF LTB₄ and LTC₄, lung microvascular permeability, and mortality (8, 9). Lipoxygenase inhibition with diethylcarbamazine (DEC) or the selective 5-lipoxygenase activating-protein inhibitor MK-886 significantly reduced BALF PMNs and LTs, lung wet/dry weight (W/D) ratios, and serum TNF-, and significantly improved survival (8, 9).

In contrast to reversible hepatocytic dysfunction induced by D-Galn or portacaval shunting with minimally altered parenchymal cells, patients with chronic liver impairment usually manifest hepatic inflammation, deranged sinusoidal perfusion, and cholestasis (6, 10). These changes progress during gram-negative bacterial sepsis or endotoxemia (6, 10). Bile duct ligation (BDL) in rats reproduces these biochemical and histopathologic alterations as well as inducing pulmonary dysfunction in such animals, with the severity of lung injury increasing over time with progression from cholestasis to biliary cirrhosis (7). In this model, end-stage biliary cirrhosis evident 5 to 6 wk after BDL was associated with reduced hypoxic pulmonary vasoconstriction, reinforcing a hepatic origin for the intrapulmonary shunting characterizing the hepatopulmonary syndrome (3). BDL of shorter duration in rats (10 to 14 d) resulted in milder cholestatic injury and fewer pulmonary defects while still resulting in increased sensitivity to lipopolysaccharide (LPS) (6, 7). The mechanism by which BDL- induced liver dysfunction impairs survival during endotoxemia remains speculative. Although changes in TNF- and IL-6 kinetics have been proposed, the relation of such changes to circulating endotoxin, cardiopulmonary function, PMN influx into liver and lungs, and mortality remains unclear since most data are based on in vitro and ex situ studies (6). No investigation has examined the role of LTs on organ inflammation and survival in this model during endotoxemia.

To address this issue directly, we developed a cholestatic rat model in which liver-lung interactions in conscious, catheterized animals were evaluated after an otherwise nonlethal infusion of Escherichia coli LPS administered 14 d after BDL or sham surgery. We analyzed the kinetics of circulating endotoxin, TNF-, and IL-6 over 24 h after endotoxemia with and without prior BDL, and correlated these kinetics with changes in survival, vital signs, and formed elements during continuous hemodynamic monitoring. We further assessed lung injury severity and PMN vascular margination within the liver and the lungs. The importance of LTs in modulating outcome in this cholestatic model compared with previous models of selective liver impairment (8, 9) was evaluated by parallel studies of lipoxygenase inhibition with DEC or MK-886. Our results support the concept that cholestatic liver injury potently enhances hepatic cytokine production by an LT-independent mechanism in response to circulating endotoxin, without altering intravascular LPS clearance. The resulting increases in serum TNF- and IL-6 as critical mediators of the multiple organ dysfunction syndrome (MODS) act in concert with BDL-induced increases in organ PMN influx to promote microvascular fluid flux in nonhepatic organs and to reduce survival.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Animals

As in our past studies (8, 9, 11), male specific pathogen-free Sprague-Dawley rats weighing 275 to 325 g (Harlan, Indianapolis, IN) were housed in positive-pressure isolation carrels with access to chow and water throughout experiments. Surgical manipulations and in vivo studies adhered to NIH guidelines and were approved by the Animal Care Committee of Saint Louis University.

Bile Duct Ligation and Sham Surgeries

After overnight fasting and under isoflurane anesthesia, a midventral laparotomy was performed aseptically on each animal to expose the porta hepatis. In BDL rats, the common bile duct was mobilized, ligated proximally and distally near the choledochoduodenal junction with 4-0 nylon suture, and widely divided from adjoining structures to prevent cicatricial reanastomosis (7). In sham rats, the common bile duct was also mobilized and the viscera gently manipulated as would occur during the BDL procedure. The peritoneum, subcutaneous tissue, and skin were then closed in layers with 2-0 nylon, with the entire operative procedure completed in  20 min. Prior to withdrawal of anesthetic, animals received buprenorphine subcutaneously (0.1 mg/ kg) to alleviate early postoperative discomfort. In preliminary studies we determined that the optimal interval between BDL surgery and later intravenous challenges (below) was 14 d, with most BDL animals remaining in good health and gaining weight despite significant elevations in their baseline serum bilirubin (see RESULTS) as assayed by a direct method (Kit 551-A; Sigma, St. Louis, MO). Pancreatitis secondary to BDL or sham surgery was eliminated as a confounding variable by measuring no increase in baseline serum amylase 14 d after surgery (before induction of endotoxemia) using an enzymatic method (Kit 577-3; Sigma), and by examining abdominal organs at necropsy.

Vascular Catheterizations

Thirteen days after BDL or sham surgery and under ketamine:xylazine anesthesia (2:1, 0.9 ml/kg) given intramuscularly, the left carotid artery and right jugular vein were aseptically catheterized with PE-50 (Clay Adams, Parsippany, NJ) filled with heparinized (10 U/ml) sterile 0.9% NaCl (NS), after which catheters were exteriorized to the dorsal neck (9, 11). Rats received 300 mg penicillin and 2.5 mg amikacin sulfate intravenously during surgery and recovered overnight with their catheters suspended by counterweights to allow free movement and unlimited access to chow and water.

Experimental Protocol

With animals awake and unrestrained 18 to 24 h after catheterization surgery, carotid arterial pressure (mm Hg) and pulse (beats/min) were continuously recorded on a multichannel physiograph (Narco Bio-Systems, Houston, TX) (9, 11). Respiratory frequency (breaths/ min) was assessed by direct observation every 30 min, and rectal temperature (°C) was measured by a miniprobe (Diatek, San Diego, CA) inserted to a depth of 4 cm. Initial measurements were obtained after a 30-min equilibration, followed by a baseline arterial blood sample (1.5 ml) for duplicate analyses of microhematocrit, and leukocyte and platelet counts by phase microscopy. Serum was immediately isolated and stored at 70° C until analyzed in duplicate for immunoreactive and bioactive TNF- and immunoreactive IL-6 (see below), alanine aminotransferase (ALT) by a commercial kit (DG159-UV; Sigma), and endotoxin by a chromogenic Limulus assay (QCL-1000; Whittaker M.A. Bioproducts, Walkersville, MD) (9, 11).

Each animal was then randomly assigned for intravenous infusion of E. coli LPS (serotype O55:B5, 5.0 mg/kg; Sigma) in 1.0 ml NS, or infusion of 1.0 ml NS alone. The major experimental groups, monitored for as long as 24 h after intravenous challenge, comprised 37 rats: BDL + LPS (n = 11), Sham + LPS (n = 14), BDL + NS (n = 6), and Sham + NS (n = 6). Infusions of LPS or NS were made through the jugular venous catheter by calibrated infusion pump (Sage, Cambridge, MA) over 15 min and ended at t = 0. Additional arterial blood samples were taken at t = 1.5, 3.5, and 24 h, when animals were killed (Na-pentobarbital, 50 mg/kg, given intravenously), or at death if occurring earlier. Isovolumetric NS was infused via the jugular catheter after each blood sample. After the 3.5-h sample, catheters were filled with heparinized (10 U/ml) NS for continuous arterial pressure monitoring overnight (9, 11). Two animals that developed increasingly severe respiratory distress were killed as described above in accordance with institutional Animal Care Committee guidelines and were recorded as surviving to the previous time point.

The role of a TNF-:leukotriene axis in modulating LPS-induced shock and mortality (8, 9) during biliary cholestasis was evaluated over 24 h in 16 additional BDL and sham rats. The lipoxygenase inhibitor diethylcarbamazine (DEC), 20 mg/kg in 1.0 ml NS, freshly prepared as the citrate-N,N-diethyl-4-methyl-1-piperazine carboximide (Sigma), was infused intravenously over 15 min in certain rats 30 min after completing the LPS infusion. This DEC regimen suppressed plasma LTB₄ as well as LTB₄ and LTC₄ levels in BALF 24 h after endotoxemia while reducing mortality in rats with D-Galn pretreatment or chronic portacaval shunts (8, 9). The BDL + LPS + DEC and Sham + LPS + DEC groups (n = 5 for each) were evaluated as described above. Additional LPS-infused BDL and Sham rats were pretreated and post-treated with MK-886 (Merck Frosst, Canada), a specific inhibitor of the 5'-lipoxygenase activating protein, using doses previously found effective in decreasing BALF [LTB₄]s indicative of systemic LT synthesis in endotoxemic rats with prior hepatic compromise (9). Three MK-886 doses (each 3.3 mg/kg in 1.5 ml of 2% ethanol) were given by gavage just before catheterization, at 2 h before and at 6 h after LPS. These MK-886 + BDL + LPS and MK-886 + Sham + LPS (n = 3 for each) were monitored as detailed above.

Three additional animals in each of the eight major treatment groups (total n = 24) were killed at t = 1.5 h, immediately after obtaining their hemodynamic data and blood samples. Fresh liver and lungs were excised from each rat, blotted free of blood, freeze-clamped in liquid N₂, and then stored at 70° C until analyzed for cell-associated TNF- as described below.

Postmortem Studies

At death, the trachea was cannulated by tracheostomy, and after sternotomy the right mainstem bronchus was ligated in situ. Right lung lobes were excised for determination of wet/dry weight ratio (W/D) without correction for blood volume by drying to constant weight at 70° C (9, 11). Left lungs were fixed in situ with cacodylate-buffered 2% glutaraldehyde at a transpulmonary inflation pressure of 20 to 22 cm H₂O, followed by fixation of 2 to 3 mm midlobar slices in fresh glutaraldehyde overnight at 5° C before dehydration and paraffin embedding (11). Standardized sections of liver and cecum were excised and similarly isolated for W/D determinations and immersion-fixation for histologic examination. Paraffin-embedded 6-µm serial sections were stained with conventional hematoxylin-eosin, and with chloroacetate esterase (CAE) with hematoxylin counterstaining to visualize neutrophils (PMNs) (11). The PMN counts per high power field (PMNs/hpf) were quantified independently by two investigators and averaged for 25 randomly selected fields per specimen at ×400 (Nikon Labophot, Japan). Animals dying overnight (t = 12 to 20 h) were not necropsied, although pressure traces usually provided the exact time of death.

Serum and Cell-associated Cytokine Analyses

Serum immunoreactive TNF- was determined in duplicate by an ELISA using rabbit antimurine anti-TNF- and goat antirabbit IgG linked to horseradish peroxidase and sensitive to murine TNF- over a range of 100 to 3,200 pg/ml (11). In selected serum samples, TNF- bioactivity was also measured in duplicate using actinomycin D-treated murine L929 cells (American Type Culture Collection, Rockville, MD) over an 8-fold dilution range based on spectrophotometric measurement of cytotoxicity at 550 nm (2, 3, 9, 11, 12). Results were calibrated with standard curves on each plate using murine rTNF- (Genzyme, Cambridge, MA). Internal controls were spiked with rmTNF- to assess recovery. To determine whether differences in peak circulating TNF- at t = 1.5 h reflected altered synthesis or release of functional cytokines by liver and lungs because of cholestasis, frozen tissues (0.25 to 0.50 g) from animals killed at this time point were homogenized in NS containing proteinase inhibitors, with homogenates then analyzed for immunoreactive TNF- and bioactive TNF- as previously described (9).

Serum immunoreactive IL-6 levels were also determined in duplicate by a rat-specific ELISA linked to horseradish peroxidase and with a detection limit of 30 pg/ml (Biosource International, Camarillo, CA). Assay specificity was confirmed and percentage recoveries were determined by spiking experimental serum samples with recombinant rat IL-6 as the standard (13).

Statistical Analyses

Data are reported as means ± SEM. Serial within-group variables were analyzed by repeated-measures ANOVA and a post hoc Newman-Keuls test, and between-group comparisons were made using Kruskal-Wallis analysis with pairwise comparisons and a two-tailed paired t test (11). Mortality was analyzed using Fisher's exact test (11). Significance was accepted for p < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Chronic BDL and Cholestatic Liver Injury

We determined that the optimal interval between BDL and subsequent monitoring during E. coli endotoxemia was 2 wk for consistent biliary cholestasis and significant elevations in serum bilirubin: BDL = 7.34 ± 0.74 mg/dl (n = 17) versus Sham = 0.25 ± 0.05 mg/dl (n = 20) (p < 0.005 between groups). Livers in BDL rats at this time point (before LPS or NS infusion) showed sinusoidal distension, panlobular hepatocytic injury, and mixed leukocytic infiltration (7, 14, 15). Other tissues in BDL animals, notably the lungs, were unremarkable. Extending the interval between surgery and LPS challenge beyond 3 wk resulted in baseline deterioration of many BDL animals to a moribund state characterized by progressive icterus, reduced food and water intake, weight loss exceeding 20%, and, occasionally, acute peritonitis caused by rupture of ductal remnants.

Serum TNF- and IL-6 Levels during Endotoxemia

Serum endotoxin levels were essentially identical between BDL + LPS and Sham + LPS groups at all time points after intravenous LPS infusion (Figure 1, top panel). Despite these results, serum immunoreactive TNF- peaked 1.5 h after completing the LPS infusion among BDL + LPS rats at 17.95 ± 3.97 ng/ml, a value nearly 8-fold higher than time-matched values of Sham + LPS animals (Figure 1, middle panel). Comparably higher values for serum TNF- were noted in BDL + LPS rats versus Sham + LPS animals at t = 3.5 h as well, such that the ratio of the areas under the respective TNF- curves over 24 h for BDL + LPS versus Sham + LPS groups was approximately 8.95. However, chronic BDL and associated cholestasis without experimental endotoxemia did not affect serum TNF-, with circulating levels in BDL + NS rats being similar to Sham + NS animals at all timepoints (Figure 1, middle panel ). Similar results across the four main treatment groups were obtained for serum bioactive TNF-, with mean values at t = 1.5 h equal to 3,291 ± 1,201 U/ml for BDL + LPS rats versus 1,456 ± 383 U/ml for Sham + LPS animals (p < 0.05) while being undetectable in BDL + NS and Sham + NS rats.

View larger version (23K): [in this window] [in a new window]   Figure 1.   (Top panel ) Serum endotoxin did not differ at any time point after intravenous infusion between rats that underwent BDL versus Sham surgery 14 d earlier, using the Limulus assay (see METHODS). (Middle panel ) Serum immunoreactive TNF- was elevated in BDL + LPS and Sham + LPS rats, but BDL without endotoxemia did not increase circulating TNF- (*p < 0.05 versus intragroup baselines). Both peak TNF- at 1.5 h and the area under the cytokine/time curve were greatest in BDL + LPS rats (p < 0.001 versus all other groups). (Bottom panel ) Serum immunoreactive IL-6 increased by t = 1.5 h after LPS and was further elevated by prior BDL, but declined to basal levels by t = 24 h. *p < 0.01 for Sham + LPS versus BDL + NS and Sham + NS groups; p < 0.01 for BDL + LPS versus Sham + LPS group. All data are plotted as means ± SEM for initial n values of: BDL + LPS = 11, Sham + LPS = 14; BDL + NS = 6; and Sham + NS = 6.

Endotoxin-induced increases in circulating immunoreactive IL-6 were significantly augmented by cholestatic liver injury (Figure 1, bottom panel). Serum IL-6 values were nearly 4-fold greater in BDL + LPS rats versus Sham + LPS animals by t = 1.5 h, and this difference persisted through t = 3.5 h (p < 0.01). Cholestasis alone was without effect, however, with serum IL-6 levels undetectable at all time points in BDL and sham animals infused only with NS. By t = 24 h, serum IL-6 was again below detection limits among Sham + LPS rats, whereas trace amounts were still measurable in serum from the few BDL + LPS animals that survived to this time point.

Cell-associated TNF- in Liver and Lung Homogenates

Baseline bioactive TNF- in BDL livers and lungs were 337 ± 45 and 114 ± 23 U/mg fresh weight, respectively, not differing from Shams (not shown). By t = 1.5 h after LPS, BDL livers contained 1938 ± 262 U/mg TNF- versus 821 ± 337 in Shams (p < 0.05), whereas lung TNF- was 437 ± 180 U/mg in BDL versus 670 ± 219 in Sham rats; lung values at 1.5 h differed from baseline (p < 0.05) but not from each other. Bioactive TNF- did not increase after NS. By t = 1.5 h after LPS, liver antigenic TNF- (pg/mg fresh weight) increased to 412 ± 140 in BDL versus 129 ± 67 in Shams (p < 0.05) but did not increase after NS from a value of 28 ± 6 pg/mg at baseline. Lung antigenic TNF- also increased at t = 1.5 h after LPS, to 317 ± 57 pg/mg in BDL versus 167 ± 79 in Shams (p < 0.05 between groups) while not differing in NS-infused rats from a baseline of 8 ± 3 pg/mg.

Survival, Cardiopulmonary Indices, and Hematology

Our previous work with LPS from E. coli serotype O55:B5 (8, 9) established that 5.0 mg/kg given intravenously was nonlethal over 24 h in conscious rats when unaccompanied by additional stressors (Figure 2). Nevertheless, Sham + LPS rats demonstrated consistent and significant systemic hypotension and tachypnea by t = 1.5 h, and hypothermia by t = 3.5 h, all of which returned to near baseline values by t = 24 h (Table 1). The BDL + LPS rats likewise showed early tachypnea accompanying bradycardic hypotension, as well as severe hypothermia and mortality by t = 3.5 h (Figure 2 and Table 1). By t = 24 h, 64% of the BDL + LPS animals had died of endotoxic shock, with most of the survivors still hypotensive and hypothermic (Table 1). Sham and BDL rats that received only NS infusions showed no such abnormalities in systemic hemodynamics or vital signs.

View larger version (17K): [in this window] [in a new window]   Figure 2.   Effects of cholestatic liver dysfunction on mortality after E. coli endotoxemia or NS infusion. Among the four major study groups, only BDL + LPS rats experienced any mortality within 24 h of completing the LPS infusion. Initial n values are those given in Figure 1; *p < 0.05, **p < 0.01 versus all other groups at each time point.

Rats in all treatment groups developed similarly mild anemia regardless of BDL or sham surgery, with hematocrits reduced about 25% during the 24 h of intensive monitoring and arterial sampling. Baseline total leukocyte counts among all BDL rats were increased to approximately twice those values measured in Sham animals (Table 2). Regardless of cholestatic liver dysfunction, all endotoxemic animals subsequently developed severe leukopenia by t = 1.5 h that persisted through 3.5 h, although leukocyte counts in survivors approximated their intragroup baselines by 24 h (Table 2). Differential counts on peripheral smears indicated that the percentage of PMNs before endotoxemia was about 42% in Sham rats versus 55% in BDL animals (p = NS). These PMN percentages did not vary consistently in the 24 h after LPS or NS infusion, although endotoxemia induced nearly absolute and significant neutropenia at 1.5 h and 3.5 h (Figure 3). Baseline platelet counts did not vary among treatment groups, and all endotoxemic rats showed progressive thrombocytopenia that was significant by t = 1.5 h and persisted through 24 h, regardless of surgical pretreatment (Table 2). The slight reductions in platelet counts by t = 24 h in Sham + NS and BDL + NS rats were consistent with their normovolemic anemia (Table 2).

View larger version (18K): [in this window] [in a new window]   Figure 3.   BDL increased baseline blood neutrophil counts (PMNs/ µl) compared with sham surgery, but all LPS-infused animals were neutropenic from 1.5 through 3.5 h (*p < 0.01 versus within-group baselines). BDL survivors and Sham animals showed reversals in circulating PMN counts by 24 h (p < 0.05 versus intragroup values at 3.5 h).

Serum bilirubin was altered significantly by endotoxemia only in the BDL + LPS rats, increasing to 15.60 ± 2.91 mg/dl by t =1.5 h and thereafter (p < 0.05 versus within-group baseline). Baseline serum ALT values did not vary across the major experimental groups, averaging 81 ± 16 U/ml. However, significant increases in serum ALT caused by cholestasis or endotoxemia alone were noted by t = 24 h, being 168 ± 41 U/ml in Sham + LPS rats and 190 ± 78 in the BDL + NS groups (p < 0.05 for both versus within-group baselines) and 71 ± 20 U/ml in Sham + NS rats. Cholestasis and endotoxemia appeared to act additively, with a mean serum ALT of 284 ± 130 U/ml at t = 24 h among BDL + LPS survivors (p < 0.01 versus all other groups and time points).

Organ Histopathology and Neutrophil Influx

Organs obtained at t = 24 h were evaluated for changes in W/D ratios as an estimate of transcapillary fluid flux (Table 3). Regardless of whether rats were infused with LPS or NS, livers and lungs from BDL rats had increased W/D values versus Sham groups. LPS infusion also increased this estimate of cecal permeability in BDL animals, but it did not alter W/D ratios in Sham rats.

When evaluated at t = 24 h, PMN accumulation within liver and lung was differentially modulated by cholestatic liver dysfunction and endotoxemia. Among animals infused only with NS, cholestasis caused by BDL increased the PMNs/hpf in liver by approximately 4-fold over sham-surgerized animals without altering PMN influx in lungs (Figure 4). Endotoxemia further amplified intrahepatic PMN accumulation among cholestatic rats, with the BDL + LPS group average nearly 10-fold greater than the mean for Sham + NS animals. In most cases, PMNs adhered to sinusoidal walls and were most abundant in zone 1 of hepatic lobules adjacent to portal triads and coinciding with severe hepatocytic necrosis. Endotoxemia also resulted in enhanced PMN influx into lungs, with mean values being similarly elevated at 24 h in BDL + LPS and Sham + LPS groups, compared with Sham + NS and BDL + NS rats (Figure 4). Intrapulmonary PMNs adhered primarily to arteriolar and capillary endothelia and were rarely noted on the epithelial side of distal air spaces.

View larger version (23K): [in this window] [in a new window]   Figure 4.   By 24 h after intravenous challenge, PMN density was elevated in liver by BDL and/or LPS treatments, but it was increased in lungs only by LPS infusions. Data are the means ± SEM of CAE-stained PMNs in 25 randomly selected high power fields (hpf, ×400) for n = four animals per group; *p < 0.05, p < 0.001 versus the same organ among Sham + NS rats.

Effects of Lipoxygenase Inhibition

Treatment of endotoxemic Sham or BDL animals with DEC (20 mg/kg, given intravenously) approximately doubled their peak circulating TNF- at t = 1.5 h compared with time-matched values for Sham + LPS and BDL + LPS groups of rats not given this lipoxygenase inhibitor (Table 4). However, serum TNF- levels were not consistently elevated by DEC at later time points. Similarly the LPS-induced increases in serum IL-6 levels were not modulated significantly by DEC at either t = 3.5 h or t = 24 h in Sham or BDL animals (Table 4). Of clinical relevance, DEC did not significantly affect other monitored variables among endotoxemic Sham or BDL rats, nor was 24-h survival improved among BDL + LPS animals by this compound.

Pretreatment and post-treatment with MK-886 (3 × 3.3 mg/kg, given orally) had no effect on survival or on peak serum TNF- values at t = 1.5 h among BDL + LPS animals, although this compound appeared to prolong TNF- clearance through t = 3.5 h (Table 4). MK-886 also significantly increased serum TNF- among Sham + LPS rats at both t = 1.5 and 3.5 h. Interestingly MK-886 differentially modulated serum IL-6 by t = 3.5 h after LPS infusion, significantly enhancing circulating levels of this cytokine among BDL + LPS rats while decreasing serum IL-6 in Sham + LPS animals (Table 4). As noted among the DEC-treated groups, MK-886 did not result in any other differences between monitored variables when compared with BDL + LPS or Sham + LPS groups, including survival, hemodynamics, or postmortem data on liver, lung, or cecum.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this study, circulating TNF- and IL-6 were significantly enhanced in rats with chronic biliary cholestasis after gram-negative endotoxemia versus animals with normal liver function. Cholestatic hepatic injury 14 d after BDL was confirmed before LPS infusion by histopathology and by elevations in serum bilirubin but not in TNF- or IL-6. The greater serum TNF- and IL-6 values in BDL + LPS versus Sham + LPS rats occurred despite similar serum endotoxin levels, and they correlated with higher concentrations of cell-associated TNF- caused by cholestasis. These amplifying effects of BDL on postendotoxemic cytokine profiles in the intravascular and hepatic compartments were accompanied by more severe shock, including hypothermia, tachypnea, and reduced survival to otherwise nonlethal endotoxemia. Cholestasis also predisposed BDL rats to increased microvascular fluid flux as estimated by liver and lung W/D ratios, regardless of whether they were challenged with LPS or NS. Intrahepatic PMNs were more numerous in all cholestatic rats, presumably because of ongoing inflammation and fibrogenesis. This hepatic PMN influx into cholestatic livers was augmented by LPS infusion and persisted for at least 24 h thereafter, accompanied by histologic and biochemical evidence of hepatocytic necrosis. Intrapulmonary PMNs increased primarily because of endotoxemia rather than cholestasis alone, although the elevation in lung W/D ratio was exacerbated by the combined effects of BDL and endotoxemia.

These results extend our previous work on liver-lung interactions during endotoxemia using noninflammatory D-Galn and portacaval anastomotic models of compromised hepatic performance (8, 9). However, the deleterious consequences of endotoxemia with preexisting cholestatic liver impairment were neither reversed nor attenuated by two structurally dissimilar inhibitors of LT biosynthesis, at doses previously shown to be effective at reducing LT production (8, 9). This finding agrees with recent data in a different rat endotoxin model of inflammatory liver impairment caused by focal hepatic ischemia/reperfusion (I/R) injury, in which MK-886 also was ineffective in reducing circulating TNF- or mortality (13, and unpublished data). Collectively, these results confirm the importance of normal hepatic performance in modulating the kinetics of TNF- production and, ultimately, survival during gram-negative infection (1, 4). Our findings further suggest that inflammatory liver impairment importantly modifies the stimulatory effects of LTs on postendotoxemic TNF- synthesis and shock-related phenomena.

This report confirms critical features of cholestatic liver dysfunction early after BDL in rats, notably, hyperbilirubinemia with normal baseline hemodynamics and mild hepatic histopathology when evaluated before endotoxemia (7). By 14 d after BDL, the livers of such cholestatic rats have been reported to contain twice the normal number of resident mononuclear cells and an expanding population of perisinusoidal (Ito) cells with myofibroblast phenotypes (14). The influx and proliferation of leukocytes and reparative cells after BDL in rodents likewise resemble changes seen in human inflammatory liver disorders, ranging in severity from acute hepatitis through fatal obstructive jaundice (6, 15, 16). In addition to enlarging the number of intrahepatic cells capable of producing inflammatory cytokines in response to circulating endotoxin, liver dysfunction in both humans and rodents primes such cells for more rapid and exuberant expression of these mediators in response to LPS (17, 18). Although both elevated and normal values for serum TNF-, IL-1, and IL-6 have been reported in patients with obstructive jaundice, and their peripheral blood mononuclear cells (PBMC) typically exhibit low basal secretion of these cytokines in vitro, their PBMC produce 10-fold more of these cytokines in response to very low levels of LPS (16). Such ex situ hypersensitivity as gauged by TNF- secretion in response to Salmonella enteritidis endotoxin also has been reported using perfused livers obtained from rats within 3 d of BDL (17).

Hypersensitivity to LPS during biliary cholestasis has been attributed in human and rodent studies at least in part to rapid and marked upregulation of monocytic surface antigens, notably CD14 receptors that bind LPS complexed with the LPS-binding protein (15, 19, 20). CD14 receptors are normally expressed in low abundance, or they may even be absent, on PBMC and Kupffer cells derived from subjects without liver disease (15, 19). Kupffer cells isolated from rats 4 d after BDL also secreted exaggerated amounts of TNF- and IL-6 in response to gram-positive bacterial peptidoglycan (20). Those results imply either unexpected interactions between CD14 receptors and peptidoglycan or, more likely, reflect a generalized priming of Kupffer cell phenotypic responses to inflammatory stimuli. In a murine cholestasis model, basal hepatic expression of TNF- was increased within 8 h of BDL and persisted for at least 5 d, measured both by the abundance of mRNA transcripts and by TNF- bioactivity in liver homogenates (21). In cholestatic rats, increased basal hepatic TNF- mRNA within 2 d of BDL localizes to Kupffer cells and is preceded by sustained activation within these cells of the nuclear transcription factor NF-B (20, 22). Not surprisingly, some investigators have reported elevations in basal serum TNF- in rats 3 to 18 d after BDL compared with sham-surgerized animals (23, 24), although this has not been a uniform finding (17), and we detected none at 14 d here. On the basis of electrophoretic mobility shift assays, however, persistent constitutive activation of NF-B in livers of BDL rats has been found even 14 d after interruption of biliary drainage (unpublished data).

Hepatic or peritoneal macrophages from cholestatic rodents in some instances were refractory to E. coli LPS when assessed by their synthesis of TNF- and secondary mediators, including nitric oxide (20, 25). Desensitization to gram-negative LPS after BDL may occur with increased translocation of gut bacteria that colonize other organs (25). Clinically, translocation of gut microflora early in the course of liver diseases is cited frequently as a risk factor predisposing these patients to septic shock, ARDS, and MODS (1). Translocation may be facilitated by reduced intestinal bile salts that normally bind endotoxins, by mucosal injury and epithelial exfoliation caused by cholestatic edema, or by other mechanisms (3, 6, 25). Liver dysfunction promotes mesenteric edema and portal venous endotoxemia (1), reducing responsiveness to specific bacterial endotoxins among resident hepatic mononuclear cells.

In addition to upregulated intrahepatic cytokine production, peripheral blood PMNs from patients with obstructive jaundice also exhibit priming of their respiratory burst activity in response to numerous stimuli, including bacterial peptides and IL-1 (6, 26). Similarly, PMNs isolated from rats at 12 h to 15 d after BDL showed increased superoxide anion production in response to phorbol esters or bacterial peptides, and exhibited greater phagocytosis and chemotaxis compared with PMNs from sham animals (27). Cholestatic increases in PMN function, combined with more numerous blood and tissue PMN, as noted by us (Figures 3 and 4) and by others (21, 27), would amplify any systemic effects of LPS-induced elevations in TNF- and other mediators of endotoxic shock.

The lungs are frequently cited as a major target of the excessive LPS-induced cytokinemia accompanying liver impairment, both clinically and in various animal models of hepatic dysfunction (1, 21, 28). Most relevant to the present study, Chang and Ohara (28) found a greatly expanded population of pulmonary intravascular macrophages (PIMs) in rats within 14 d of BDL. These PIMs avidly phagocytized infused latex particles as well as radiolabeled endotoxins, and the increasing numbers of PIMs with time after ductal ligation correlated with the severity of pulmonary edema and mortality after LPS challenge (28). The lungs of cholestatic mice also contain more PMNs than do the lungs of control animals within 5 d of BDL, with this increase inhibited by anti-TNF- antibodies (21). As for the elevated levels of circulating TNF- after endotoxemia in rats with preexisting liver impairment, intrapulmonary PMNs are similarly increased in a stereotypic manner by diverse forms of hepatic dysfunction. In vivo hepatic ischemia/reperfusion injury (29), acute D-Galn hepatotoxicity (8), and chronic reductions in liver perfusion caused by portacaval shunting (7) are all associated with enhanced PMN influx into the lungs. Thus, an expanded and activated intrapulmonary mononuclear cell population may well account for increased lung production of cytokines and LTs, with the attendant lung injury exacerbated at least in part by the actions of more numerous PMNs marginated within the pulmonary microvasculature and alveoli.

Results herein shed further light on the role of increased TNF- production in modulating in vivo responses to endotoxemia, particularly the complex bidirectional interaction of TNF- with lipoxygenase-derived eicosanoids (8, 9). In rats with acute hepatotoxicity because of D-Galn treatment and reminiscent of clinical viral hepatitis, LPS-induced TNF- production and mortality were amplified (3, 32). Increased mortality and pulmonary inflammatory responses to LPS in D-Galn-treated animals were furthermore associated with elevated serum LTB₄ 90 min after endotoxemia, and LTB₄ and LTC₄ concentrations in plasma and BALF 24 h later (8). DEC, which inhibits generation of lipoxygenase pathway metabolites beyond the 5-HETE step, attenuated these LT responses, even as survival, lung albumin leak, and pulmonary inflammation were improved (8).

Although suggesting that impaired metabolism of endogenously synthesized LTs by the damaged liver plays a causal role in postendotoxemic organ injury, DEC or MK-886 also suppressed LPS-induced serum TNF- by > 90% during reductions in hepatic blood flow caused by portacaval shunting (9). These latter changes were accompanied by significant reductions in mortality, bronchoalveolar PMN counts, and BALF levels of LTB₄ 24 h after endotoxemia (9). Such reductions in serum TNF- by lipoxygenase inhibition in models of noninflammatory liver impairment raise the possibility that shunting of arachidonate precursors to the cyclooxygenase pathway generates prostaglandins (PG) such as PGE₂ that inhibit proinflammatory cytokine synthesis. The resulting hypersensitivity to LPS induced by cholestasis may have preempted the normal negative feedback regulation exerted by PGE₂ and other cyclooxygenase products. However, this mechanism appears unlikely since dual inhibitors of the lipoxygenase and cyclooxygenase pathways similarly depress LPS-induced TNF- production (33).

A different mechanism, based on the known stimulatory effects of LTs on TNF- biosynthesis in vitro (34), would imply that LT synthesis inhibition should block the stimulatory effects of LTs on TNF- production (9). The failure of these agents to reduce serum TNF- or serum IL-6 (Table 4), or to improve survival among animals with BDL or with I/R-mediated hepatic inflammatory impairment (13) suggests two other possibilities. First, augmented hepatic production and export of LTs, associated with the enhanced PMN influx into cholestatic livers, may have exceeded the ability of DEC or MK-886 to control such increased biosynthesis, using doses formerly found to be effective (8, 9). Second, cholestasis and postischemic liver injury may have altered the kinetics and/or bioavailability of these LT synthesis inhibitors, thereby reducing their effectiveness. We cannot distinguish between these or other possibilities at present since BALF LT levels were not quantified before and after LT inhibition. This limitation of the present study occurred primarily because maximal BALF LT levels occur 24 h after endotoxemia (8, 9), by which time all of the BDL + LPS + DEC and BDL + LPS + MK-886 animals had succumbed. Further studies in this model of cholestatic injury will be important to resolve this issue and to address additional mechanisms of inflammation within the liver and the lungs.