INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

During gram-negative bacteremic sepsis, interactions between nonpulmonary organ systems and the lungs are important determinants of pulmonary inflammation, as typified by the acute respiratory distress syndrome (ARDS) (1, 2). However, specific organ dysfunctions that impact postbacteremic lung inflammation and the underlying mechanisms of such interactions are incompletely understood.

The liver plays a key role in multiple regulatory aspects of host defense during gram-negative sepsis (3, 4). Both the production of inflammatory cytokines including tumor necrosis factor-alpha (TNF-) by liver macrophages and the export of these mediators into hepatic venous blood and the systemic circulation are substantial after endotoxemia (5). Moreover, Kupffer cell-mediated uptake of gram-negative bacteria, hepatobiliary clearance of TNF- and phlogistic lipoxygenase-derived leukotrienes (LTs) are important lung protective mechanisms (4, 9, 10). Sepsis-related liver dysfunction from occult changes in hepatic perfusion despite preserved systemic blood pressure may amplify downstream inflammatory events within the pulmonary vasculature and distal air spaces.

Liver impairment in patients with gram-negative sepsis is increasingly recognized as a risk factor for ARDS (2, 3, 7, 8). However, it is not clear whether the enhanced risk of sepsis-related lung inflammation is associated with altered kinetics of circulating microbial products and inflammatory mediators such as TNF-. Accordingly, there is interest in models of liver-lung interactions in which endotoxin-induced TNF- biosynthesis precedes acute hepatic dysfunction as occurs in the critically ill. A common cause of such organ impairment is ischemia/reperfusion (I/R) injury from sepsis-related cardiopulmonary derangements that reduce hepatic O₂ delivery (11, 12). Sepsis-induced changes in portal venous pressure-flow relations (13, 14) and decreased microvascular perfusion due to occlusive aggregates of polymorphonuclear leukocytes (PMNs) within the hepatic sinusoids (15) may also contribute. Based on data from noninfected experimental models, lung inflammation may be augmented by postbacteremic I/R injury of the liver. In rats, hepatic ischemia from 90 min of partial lobar vascular occlusion followed by reperfusion increased circulating TNF-, interleukin-6 (IL-6), epithelial neutrophil-activating protein (16), CXC chemokines (19), and lung microvascular permeability and PMN sequestration. Hepatic and pulmonary TNF- gene expression are independently stimulated by nonischemic endotoxemia (5, 6) or by nonseptic liver I/R (16) through activation of nuclear factor kappa B (NF-B) and other redox-sensitive transcription factors (20). Even so, it is unknown whether lung inflammation and mortality are affected by selective hepatic I/R injury after gram-negative sepsis, particularly in relation to changes in microbial vascular clearance and circulating inflammatory mediators.

We performed these studies to test the hypothesis that postbacteremic I/R injury of the liver alters the kinetics of circulating endotoxin, TNF-, and IL-6, enhancing lung PMN influx and mortality after gram-negative bacteremic sepsis compared with similarly infected, antibiotic-treated animals with normal liver function. As a mechanistic corollary, we determined the effects of LT antagonism in this model using the 5-lipoxygenase-activating protein inhibitor MK-886 (10), based on our previous findings that LTs mediate endotoxin-induced mortality and lung inflammation after galactosamine-induced acute liver injury (9) or portacaval shunting (10). Here we report that secondary reductions in liver blood flow and reperfusion early after normotensive E. coli sepsis augment circulating endotoxin and TNF- levels while inducing a liver-lung interaction by which postbacteremic lung inflammation is enhanced.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Animals and Vascular Catheterizations

Pathogen-free male Sprague-Dawley rats (270 to 300 g; Harlan, Indianapolis, IN) were caged in positive-pressure isolation rooms before surgery and in negative-pressure containment rooms thereafter, with continual access to chow and water. All studies adhered to National Institutes of Health guidelines and were approved by the Animal Care Committee of Saint Louis University. On the day before gram-negative bacteremic infection (Figure 1) and under ketamine:xylazine anesthesia (2:1, 0.9 ml/kg, intramuscularly), the left carotid artery and right jugular vein were aseptically catheterized with heparinized (10 U/ml) normal saline (NS)-filled PE-50 tubing (Becton Dickinson, Sparks, MD) (10, 21). Animals received 300 mg of penicillin and 2.5 mg of amikacin sulfate (intravenously) at surgery and 30 min after bacterial infection.

View larger version (22K): [in this window] [in a new window]   Figure 1.   Schematic depiction of the experimental protocol for studies that began 18 to 24 h after vascular catheterization and ended 195 min or 24 h after bacteremic infection with 109 viable E. coli serotype O55:B5 (EC) or infusion of 0.9% NaCl (NS). Laparotomy at t = 45 min was followed by 90 min of focal no-flow hepatic ischemia in I/R groups; Sham animals had liver manipulation without vaso-occlusive ischemia. At t = 135 min, livers underwent brief (1-h) reperfusion or subacute reperfusion over 24 h before necropsy.

Escherichia coli Cultures

E. coli (EC, serotype O55:B5) were obtained as number 12014 from the American Type Culture Collection (ATCC, Rockville, MD). Suspension cultures (18 to 24 h) were sedimented (1,000 × g, 10 min, 4° C), washed twice and resuspended in sterile NS, and kept at 4° C until use (6). Inocula of EC enumerated by hemacytometer as 1.0 × 10⁹ colony-forming units (cfu)/ml contained 8.5 ± 0.4 × 10⁸ cfu/ml by quantitative culture. Quantitative whole-blood cultures were performed in duplicate at designated time points in all experimental groups.

Experimental Protocol

Eighty rats were studied herein. Experiments lasting 195 min (n = 41) were performed to obtain liver and lung tissue after 90 min of hepatic ischemia plus 60 min of reperfusion to compare with studies of organ PMN accumulation after nonseptic hepatic ischemia (16). To define the effects of longer reperfusion on circulating endotoxin, TNF-, lung dysfunction, and mortality over 24 h (17), additional studies were performed in 39 rats (Figure 1). Rats were reanesthetized 18 to 24 h after catheterizations with sodium pentobarbital (30 mg/kg, intraperitoneally) and placed on a heating pad to maintain rectal temperatures at 37 to 40° C (10, 21). Arterial pressure and pulse rate were continuously recorded as was respiratory frequency (10, 21). Animals received NS (5 ml/kg intravenously) over 10 min to ensure adequate hydration. Arterial blood (1.5 ml) was obtained 10 min later for hematocrit and leukocyte counts by phase microscopy, differential counts, and blood culture. Serum was stored at 70° C until analyzed for endotoxin, TNF-, IL-6, and alanine aminotransferase (ALT). After NS replacement of this and succeeding blood samples, cardiopulmonary data were obtained for 20 min to ensure stable baseline conditions. Additional NS (approximately 2.5 ml/kg, intravenously) was given as needed to maintain mean arterial pressure > 90 mm Hg during surgery.

Animals were randomly assigned to receive bacteremic infection over 15 min by intravenous infusion of 1.0 × 10⁹ live EC in 1.0 ml NS, or of 1.0 ml NS alone, which began at t = 0. The EC inocula corresponded to a dose that is lethal in 25% of test subjects (LD₂₅) in conscious rats over 24 h (6). After 30 min, another arterial sample (at t = 45 min) was obtained and then an aseptic laparotomy was performed. Rats then randomly underwent either occlusion of the hepatic artery, portal vein, and bile duct branches of the median, left, and caudal lobes of the liver (16- 19) using a microaneurysm clamp for 90 min (focal no-flow ischemic groups), or only laparotomy and manipulation of the liver (sham groups). The laparotomy incision was reapproximated until reopening the abdomen 90 min later (t = 135 min) in ischemic preparations, in which the vascular clamp was removed, reperfusion verified, and the wound closed. Anesthesia was maintained as judged by ablation of reflexes with sodium pentobarbital (10 mg/kg, intravenously).

Brief (1-h) Reperfusion Studies

At the conclusion of hepatic ischemia (t = 135 min), rats underwent 60 min of liver reperfusion until killed with sodium pentobarbital (50 mg/kg, intravenously) and creation of a pneumothorax at t = 195 min. In these studies, blood samples (1.5 ml each) were obtained during peak ischemia (t = 105 min), and just prior to sacrifice (t = 195 min), followed by necropsy. Four groups were studied: (1) EC + Sham (n = 9); (2) EC + I/R (n = 13); (3) NS + Sham (n = 8); (4) NS + I/R (n = 11).

24-h Reperfusion Studies

In experiments lasting up to 24 h after EC infection or NS infusion, animals underwent 90 min of hepatic no-flow ischemia or sham manipulation as previously described, followed by reperfusion starting at t = 135 min and closure of the laparotomy incision. These rats awoke in their cages during continuous monitoring and recording of vital signs for up to 24 h, after which survivors were killed with sodium pentobarbital and necropsied (Figure 1). Animals dying overnight (t = 12 to 20 h) were not necropsied, but examination of blood pressure traces enabled determination of their time of death. To maintain a similar number of four arterial samples/rat as in the 60-min reperfusion studies, blood was obtained at baseline, and at t = 195 min, 6 h, and 24 h. Animals received a second dose of antibiotics at t = 6 h. Groups in these 24-h studies were: (1) EC + Sham (n = 7); (2) EC + I/R (n = 9); (3) NS + Sham (n = 8); and (4) NS + I/R (n = 6).

To evaluate the role of EC-induced LTs in mediating liver and lung injury in this model, certain animals were pretreated with MK-886 (Merck Frosst, West Point, PA), an inhibitor of 5-lipoxygenase-activating protein. MK-886 was given in doses (3.3 mg/kg in 2% ethanol) that reduce leukotriene B₄ (LTB₄) concentrations in bronchoalveolar lavage fluid (BALF) 24-h after endotoxemia (10), by gavage 2 h before EC infection and again at t = 6 h, giving two additional groups, MK-886 + EC Sham (n = 4) and MK-886 + EC + I/R (n = 5).

Postmortem Studies

At t = 195 min in the brief reperfusion studies and t = 24 h in subacute reperfusion experiments, or at witnessed death if occurring earlier, a tracheostomy and sternotomy were performed. Then the right lungs were excised to determine their wet/dry weight (W/D) ratio after drying to constant weight at 70° C as an index of microvascular fluid flux (10). Left lungs were fixed in situ with cacodylate-buffered 2% glutaraldehyde (transpulmonary pressure 20 to 22 cm H₂O), and then 2- to 3-mm midlobar slices immersed in glutaraldehyde overnight at 5° C (10). Standardized sections of postischemic and nonischemic hepatic lobes, and sections of heart, kidney, and cecum were excised and fixed. Paraffin-embedded sections (6 µm) were stained with hematoxylin/eosin, and with chloroacetate esterase (CAE) for 2 h with or without hematoxylin counterstaining (10).

Slides of CAE-stained liver and lung were coded to eliminate knowledge of animal treatment before scoring for the number of PMNs per high-powered field (HPF) at a magnification of ×400. Tissues were scored in a random pattern across liver and lung lobes from each animal until 25 HPFs had been evaluated, after which the mean number of PMNs/HPF was calculated for each animal. Tissues showing artifactual damage or underinflated alveoli were excluded from analyses.

Serum Endotoxin

Serum samples (50 µl) were analyzed in duplicate for endotoxin content by a chromogenic Limulus amebocyte lysate (LAL) assay (Associates of Cape Cod, Woods Hole, MA). Samples were diluted 1:5,000 in pyrogen-free LAL reagent H₂O, after which endotoxin concentrations were determined at 405 nm with a correction wavelength of 490 nm (EL-311; Bio-Tek, Winooski, VT).

Cytokine Assays

Actinomycin D-treated, mycoplasma-free murine L929 cells (ATCC) were used to quantitate serum TNF- bioactivity in duplicate by cytotoxicity measured at 550 nm (6, 10, 21, 22). Sera underwent a single freeze-thaw cycle before analysis. IL-6 was measured in duplicate in subacute studies with an ELISA specific for rat IL-6 (Biosource, Camarillo, CA).

ALT

Serum ALT levels were determined in duplicate using a colorimetric endpoint assay and appropriate controls as recommended by the manufacturer (505-P; Sigma, St. Louis, MO) (21).

Statistical Analyses

Data are reported as means ± SEM. Serial changes in within-group variables were analyzed by repeated-measures analysis of variance (ANOVA) and a post hoc Newman-Keuls test, and between-group comparisons by Kruskal-Wallis analysis with pairwise comparisons and two-tailed paired t test (6, 10, 21). Mortality data were analyzed using the Fisher exact test. Significance was accepted for p values < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Supplementary data obtained during this investigation are available for review in the online data supplement to this article at www.atsjournals.org (Figures E1 through E7).

Postbacteremic Liver Ischemia and Brief Reperfusion: Effects on Circulating Endotoxin, TNF-, and Organ PMN Influx

Among the two groups of bacteremic animals, 90 min of secondary hepatic ischemia in EC + I/R rats did not alter circulating endotoxin levels at t = 105 min (1,328 ± 239 endotoxin units [EU]/ml) compared with the EC + Sham group (837 ± 296 EU/ml; p = not significant [NS]). Likewise, serum endotoxin levels were not increased by 60 min of hepatic reperfusion in EC + I/R rats, compared with time-matched EC + Sham values (607 ± 115 versus 760 ± 135 EU/ml, respectively).

Circulating bioactive TNF- increased > 20-fold by 45 min after E. coli bacteremic sepsis in both EC + Sham and EC + I/R animals (p < 0.001 versus baseline) compared with time-matched NS + Sham or NS + I/R values. Serum TNF- rose further in both EC-infected groups afterwards, but peak TNF- concentrations were higher in EC + I/R rats during hepatic ischemia (t = 105 min) compared with time-matched values in EC + Sham rats with noninterrupted liver perfusion (p < 0.05) (Figure 2A). Postbacteremic differences in circulating TNF- were associated with greater liver injury in EC + I/R versus EC + Sham rats as assessed by serum ALT (Figure 2B), similar to time-matched differences between NS + I/R versus NS Sham animals.

View larger version (24K): [in this window] [in a new window]   Figure 2.   (A) Mean values (± SEM) for serum bioactive TNF- after bacteremic infection with 109 EC or infusion of NS at t = 0, with and without subsequent liver ischemia and brief (1-h) reperfusion. EC + Sham (open squares, n = 9); EC + I/R (filled squares, n = 13); NS + Sham (open circles, n = 8); and NS + I/R (filled circles, n = 11). *p < 0.01 versus within-group baseline values at t = 0; p < 0.05 for EC + I/R versus time-matched EC + Sham group value. (B) Mean values (± SEM) for serum ALT in the same animals. *p < 0.05 versus within-group baseline values; p < 0.01 versus within-group baseline values and time-matched EC + Sham and NS + Sham values.

Notably, 1 h of liver reperfusion did not increase serum TNF- in EC + I/R rats compared with their intra-ischemic peak values at t = 105 min. Even so, TNF- concentrations at t = 195 min remained higher in EC + I/R animals than time-matched values for EC + Sham rats with uninterrupted liver perfusion (Figure 2A). Blood culture positivity and bacterial clearance from the blood of EC + Sham and EC + I/R rats were comparable. Peak circulating cfu (means ± SEM) at t = 45 min were 21 ± 15 × 10³/ml in EC + Sham rats versus 19 ± 12 × 10³ in EC + I/R animals. Subsequent values were 12 ± 8 × 10³/ml in EC + Sham rats at t = 105 min versus 16 ± 9 × 10³/ml in the EC + I/R cohort, and 1.8 ± 0.7 × 10³/ml versus 1.6 ± 1.0 × 10³/ml at t = 195 min, respectively. No positive blood cultures were detected at any time point in NS + Sham or NS + I/R animals. Mean arterial pressure was 90 to 120 mm Hg at baseline in both groups. Respiratory frequencies averaged 55 to 72 breaths/min at baseline and did not vary with treatment or time.

Ischemic (clamped) lobes of livers from EC + I/R rats exhibited more necrosis and influx of PMNs compared with time-matched liver sections from EC + Sham animals (Figures E1a and E1b). Nevertheless, PMN influx after 60 min of reperfusion into ischemic versus nonischemic lobes of EC + I/R livers did not differ from EC + Sham control animals (Figure 3). Increased PMN influx was also noted in postischemic hepatic lobes without infection in NS + I/R versus NS + Sham livers (Figures E1c and E1d). Bacteremic infection and nonbacteremic I/R of the liver in NS + I/R rats both induced leukopenia by t = 105 min (Table 1), with the reductions in EC + I/R rats caused by decreased blood PMNs.

View larger version (20K): [in this window] [in a new window]   Figure 3.   PMN influx into the livers (A) and lungs (B) of the same four experimental groups shown in Figure 2 after bacteremic infection with 109 EC or infusion of NS at t = 0, with and without subsequent liver ischemia and brief (1-h) reperfusion. Note division of liver tissues into ischemic and nonischemic (e.g., nonclamped) segments. Values are means ± SEM of CAE-positive cell counts in at least 25 HPF at a nominal magnification of ×400. (A) *p < 0.05, p < 0.01 for EC + Sham and EC + I/R groups versus organ-specific NS + Sham values; (B) *p < 0.05 for NS + I/R versus NS + Sham animals.

Lung PMNs/HPF increased in EC + Sham rats versus NS + Sham controls (Figure 3B). Neither EC nor ischemia plus reperfusion altered alveolar architecture or caused edema, with PMNs in EC + Sham and EC + I/R rats confined to septal capillaries. No differential changes in PMN influx were found in heart, kidney, and cecum after EC infection, hepatic I/R, or sequential EC + I/R treatment. These results were supported by similar organ W/D ratios (Figure E2), except for cecums from EC + I/R animals having an elevated W/D (6.57 ± 0.67) compared with NS + Sham rats (4.22 ± 0.15, p < 0.05).

Postbacteremic Liver-Lung Interactions During Longer-term (24-h) Hepatic Reperfusion: Endotoxin and Cytokine Responses, Organ Inflammation, and Survival

Increased reperfusion times after hepatic ischemia were associated with significantly higher serum endotoxin and TNF- levels during normotensive E. coli sepsis, compared with EC Sham animals without induced reductions in liver blood flow. Circulating endotoxin at t = 6 h was 2.3-fold higher in EC + I/ R animals versus EC + Sham rats (1,151 ± 265 versus 495 ± 166 EU/ml, respectively; p < 0.05). Similarly, serum TNF- in EC + I/R animals exceeded time-matched values for EC + Sham rats at all times after baseline (Figure 4A). These differences in TNF- kinetics occurred despite similar intravascular clearance of EC as determined by quantitative blood cultures. At t = 195 min circulating cfu were 2.2 ± 1.1 × 10³/ml in EC + I/R animals versus 1.6 ± 0.6 × 10³/ml in the EC + Sham cohort, respectively. Subsequent values at t = 6 h were 4.4 ± 2.2 × 10³/ml versus 4.14 ± 2.9 × 10³/ml in EC + I/R versus EC + Sham animals, respectively. At t = 24 h corresponding values in surviving animals were 1.8 ± 1.6 × 10³/ml versus 1.5 ± 0.6 × 10³/ml. No blood cultures were positive at any time point in NS + Sham or NS + I/R animals. There were no intergroup differences in mean arterial blood pressure at any time point among bacteremic or NS-infused animals. However, the higher serum TNF- levels in EC + I/R rats were associated with elevated serum ALT values versus EC + Sham animals (Figure 4B). Of note, serum ALT in NS + I/R rats was also significantly elevated versus NS + Sham control rats, although no increases in serum bioactive TNF- were found at any time point in the NS + I/R group (Figures 4A and 4B). Peak IL-6 concentrations at t = 195 min and 6 h were similarly increased in both groups of bacteremic animals (Figure 5). However, serum IL-6 was no longer detectable at t = 24 h in surviving EC + Sham animals, whereas in EC + I/R rats serum IL-6 averaged 109 ± 67 pg/ml (p < 0.05). No IL-6 was detected at any time point in the serum of either the NS + I/R or NS + Sham animals.

View larger version (24K): [in this window] [in a new window]   Figure 4.   (A) Mean values (± SEM) for serum bioactive TNF- after bacteremic infection with 109 EC or infusion of NS at t = 0, with and without subsequent liver ischemia and subacute (24-h) reperfusion. EC + Sham (open squares, n = 7); EC + I/R (filled squares, n = 9); NS + Sham (open circles, n = 8); and NS + I/R (filled circles, n = 6). *p < 0.01 versus within-group baseline values at t = 0; p < 0.01 versus EC + I/R baseline and peak EC + I/R versus EC + Sham group values. (B) Mean values (± SEM) for serum ALT in the same animals. *p < 0.05 versus within-group baseline; p < 0.01 versus within-group baseline and time-matched EC + Sham and NS + Sham values.

View larger version (20K): [in this window] [in a new window]   Figure 5.   Mean values (± SEM) for serum immunoreactive IL-6 after E. coli bacteremic infection at t = 0, with and without 90 min of subsequent liver ischemia and subacute (24-h) reperfusion. *p < 0.05 versus time-matched EC + Sham value.

Livers from 24-h survivors in the EC + I/R and NS + I/R groups showed significantly greater PMN influx into their postischemic versus nonischemic lobar regions (Figure 6A) and versus liver after 1 h of reperfusion (Figure 3A). The magnitude of reperfusion-related PMN influx was independent of prior EC infection. In contrast, normotensive E. coli sepsis combined with secondary hepatic I/R injury significantly increased lung inflammation in surviving EC + I/R animals compared with the EC + Sham rats (Figure 6B). Interestingly, the greater PMN influx into the lungs of surviving EC + I/R animals was twice that of NS + I/R rats with similar hepatic injury but no EC infection, which also did not differ from lung PMNs/HPF in NS + Sham rats (Figure 6B). Circulating leukocytes were reduced over 24 h among the survivors of both the EC + I/R and NS + I/R groups, compared with early but transient decreases in EC + Sham rats and no change in NS + Sham animals (Table 2). The greater PMN influx into the lungs of EC + I/R animals also was associated with higher lung W/D ratios in this group (6.34 ± 0.58 versus 5.34 ± 0.26 in EC + Sham rats; p < 0.05). No treatment-related differences were found in the mean W/D ratios for liver, heart, kidney, or cecum in animals surviving to 24 h (Figure E3).

View larger version (18K): [in this window] [in a new window]   Figure 6.   Neutrophil (PMN) influx into the livers (A) and the lungs (B) of the same four experimental groups shown in Figure 4 after bacteremic infection with 109 EC or infusion of NS at t = 0, followed by subsequent liver ischemia and subacute (24-h) reperfusion. Values are means ± SEM of PMNs/HPF. Note increased numbers of PMNs only in ischemic liver tissue of EC + I/R and NS + I/R animals. PMN influx into lungs is significant only in EC + I/R animals. *p < 0.05 versus livers from NS + Sham rats; *p < 0.05 versus EC + Sham, NS + Sham, and NS + I/R lungs.

Although hepatic ischemia with prolonged reperfusion amplified the postbacteremic endotoxin and TNF- concentrations as well as lung inflammatory responses early after E. coli sepsis, the 24-h survival of EC + I/R animals (45%) did not differ from EC + Sham rats (57%); p = NS. Most deaths occurred within 3 to 8 h of EC infection (Figure 7). No deaths occurred through 24 h in either the NS + Sham or NS + I/R groups. Regardless of treatment, all 24-h survivors were normothermic and normotensive with mild tachycardia and tachypnea (Figures E4-E7). However, intragroup respiratory rates at 24 h increased from 80 ± 5 breaths/min at baseline to 153 ± 22 breaths/ min in EC + Sham animals (p < 0.05), and from 71 ± 4 at baseline to 107 ± 5 breaths/min in EC + I/R rats (p < 0.05). Respiratory frequency did not differ at any time in the NS + I/R group compared with NS + Sham animals (Figure E7).

View larger version (18K): [in this window] [in a new window]   Figure 7.   Mortality in animals after bacteremic infection with 109 EC or infusion of NS at t = 0, with and without subsequent liver ischemia and reperfusion over 24 h.

Effects of Lipoxygenase Inhibition on Inflammation, Postischemic Liver Function, and Survival

Serum endotoxin levels in animals treated with MK-886 before and after EC infection and then sham manipulations or focal no-flow hepatic I/R were not different from their respective EC + Sham and EC + I/R values. Thus, endotoxin values at t = 6 h were 534 ± 127 EU/ml in MK-886 + EC + Sham rats versus 1,072 ± 216 EU/ml in MK-886 + EC + I/R animals (p < 0.05). However, lipoxygenase inhibition significantly reduced circulating TNF- in the MK-886 + EC + I/R group (Table 3). In contrast, lipoxygenase inhibition did not alter peak serum TNF- levels in MK-886 + EC + Sham rats compared with EC + Sham animals.

As Figure 4B shows, most surviving EC +I/R rats showed little attenuation of their elevated serum ALT even by t = 24 h. In contrast, lipoxygenase inhibition significantly diminished serum ALT in both the MK-886 + EC + I/R rats and the MK-886 + EC + Sham animals (Table 3). In particular, MK-886 treatment reduced time-specific peak serum ALT levels by 99% compared with EC + I/R rats. Compared with a baseline serum ALT level in MK-886 + EC + Sham rats of 47 ± 5 U/ml, the ALT values in this group were 46 ± 3, 54 ± 6, and 65 ± 14 U/ml at t = 195 min, 6 h, and 24 h, respectively (p < 0.05 versus time-matched EC + Sham values). Despite these changes, MK-886 neither altered cardiopulmonary indices nor improved 24-h survival (60%), such that organ inflammation could not be statistically assessed.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this report we provide evidence for a novel interaction between the liver and the lungs during normotensive E. coli bacteremic sepsis, in which secondary hepatic ischemia amplified circulating bioactive TNF- in antibiotic-treated rats over 24 h, compared with septic animals without induced impairment in hepatic perfusion. Such postbacteremic ischemia of the liver augmented peak serum TNF- levels (Figure 2A), even though neither the intravascular clearance of E. coli nor serum endotoxin concentrations were altered. Reestablishing in vivo perfusion to ischemic liver regions for 1 h did not increase circulating TNF- above ischemic-phase concentrations, although TNF- concentrations in EC + I/R animals remained greater than concurrent cytokine values in EC + Sham rats. However, extending hepatic reperfusion over 24 h was associated with evolving biochemical injury to the liver and intrahepatic accumulation of neutrophils within previously ischemic (i.e., clamped) tissue, even as serum endotoxin and TNF- levels exceeded time-matched EC + Sham control values (Figures 4A and 5). These events coincided with postbacteremic increases in pulmonary microvascular permeability as well as augmented PMN influx into the lungs, compared with infected rats without secondary liver ischemia/reperfusion injury (Figure 6B). Collectively, these results suggest that localized reductions in hepatic blood flow (despite otherwise normal systemic blood pressure) predispose to lung inflammation by modifying the kinetics of circulating endotoxin and TNF-, a canonical mediator of gram-negative sepsis (23).

Significant, but clinically occult, reductions in hepatic blood flow causing hypoxia of the liver are thought to be common during gram-negative bacteremic sepsis in the critically ill (3, 11, 12). Contributing factors include enhanced intrasinusoidal PMN aggregation that increases de novo resistance to sinusoidal blood flow (3, 15), endotoxin-mediated upward shifts of the portal venous pressure-flow relationship that increase the heterogeneity of sinusoidal perfusion (13, 14), mismatch of hepatic O₂ supply and demand (24), and redirection of blood flow away from the liver by catecholamine therapy (12). Cognizant of the worse pulmonary outcome and survival of severely infected patients with antecedent liver dysfunction (2, 7), we previously modeled liver-lung interactions during gram-negative sepsis over a spectrum of preexisting derangements of hepatic performance and perfusion, including acute hepatitis (9), portacaval shunting (10), and biliary cirrhosis (21). With the present study we have extended the boundary conditions of such organ interactions and their impact on postbacteremic lung dysfunction, by focusing on postbacteremic ischemia of the liver followed by acute and subacute reperfusion phases. No previous studies have investigated the combinatorial interaction of sepsis and hepatic I/R injury despite its presumed frequency.

Critique of the Model

Induction of hepatic ischemia followed by reperfusion over 24 h as performed here is a well-validated tool to characterize hepatic reperfusion injury in rats (15). By this approach, approximately 70% of the liver parenchyma is made reversibly ischemic, even as portal decompression through nonclamped right and caudate lobar vessels averts mesenteric congestion. To date, such hepatic I/R injury has been studied only under nonseptic conditions resembling the NS + I/R cohort in this study, which simulate liver underperfusion during hypovolemic conditions or liver resectional surgery (15). Thus, Colletti and coworkers (16, 18) and McCurry and colleagues (17) reported increased export of TNF-, IL-6, and CXC chemokines into hepatic venous blood after 90 min of ischemia and reperfusion injury of the liver over 24 h, coincident with increases in lung microvascular permeability and intrapulmonary accumulation of PMNs. Yoshidome and coworkers found similar lung inflammation after hepatic I/R in mice which was attributed to intrapulmonary activation of NF-B (19, 20).

In contrast to these reports, we found no increases in circulating bioactive TNF- at any time point in noninfected NS + I/R rats compared with NS + Sham control animals (Figures 2A and 4A). Moreover, lung microvascular permeability was not augmented in these noninfected NS + I/R animals despite ischemic liver injury (Figures E2 and E3), and lung PMN influx was only modestly increased after brief but not prolonged reperfusion (Figures 3B and 6B). The reasons are unclear for these differences in NS + I/R animals from previous reports of hepatic I/R injury in noninfected rats (16). Our rigorously aseptic I/R protocol in chronically catheterized NS- infused rats, including antibiotics, prevented bacterial contamination that may have contributed to postischemic cytokine production in studies by others. Further, hepatic ischemia was induced during the postprandial state in the present study compared with experiments in fasted animals in which hepatic xanthine oxidase activity is increased (18, 25). We would therefore expect reduced oxidative transactivation of NF-B and other redox-sensitive transcription factors (20) and by extension, decreased activation of the TNF- promoter region. Our sampling of arterial versus superior vena caval blood and enumeration of PMNs in lung tissue by direct microscopy rather than by myeloperoxidase activity also differed from previous reports (16). Another potential factor is the apparently enhanced sensitivity of WEHI 164 cells to the cytolytic effects of TNF (16) compared with L929 cells as used herein (26).

To address our study hypothesis, we modified the noninfected hepatic I/R injury model to one of normotensive, antibiotic-treated E. coli bacteremic sepsis with secondary ischemic liver dysfunction (Figure 1). To our knowledge this is the first investigation of the determinants of postbacteremic lung inflammation by selective I/R injury of a nonpulmonary organ system. Even so, there are limitations of this study. Hemodynamic instability during sepsis likely causes recurrent liver ischemia, which was not modeled in our EC + I/R preparations. Emerging data also suggest that endotoxin-induced TNF- gene expression and lung inflammation may be differentially affected by focal versus global liver ischemia (22, 27). Still, our intent was to model a single reduction in hepatic perfusion early after severe but recoverable gram-negative infection. Thus, differences in mortality between EC + I/R and EC + Sham groups (Figure 7) most likely were minimized by our calibration of inoculum, antibiotic therapy, fluid administration, and postoperative analgesia in the 24-h EC + I/R reperfusion group. The effects of hepatic I/R injury on mortality during more severe or compartmentalized infection remain to be determined. Regarding the longer-term (24-h) reperfusion studies, we did not find parallel increases in W/D ratios of the liver or other nonpulmonary organs in EC + I/R rats compared with EC + Sham control rats. Although lung edema formation is enhanced with even modest increases in pulmonary microvascular permeability, we suspect that more aggressive fluid resuscitation in this model would have caused greater edema in nonpulmonary organs as well. Also, W/D ratios as estimates of microvascular permeability were obtained in surviving animals at the 24-h time point, such that animals with more severe pulmonary and nonpulmonary edema may have already succumbed (Figure 7). The full spectrum of effects caused by secondary hepatic ischemia on postbacteremic lung function requires further study, given our finding of increased lung inflammation without histologic evidence of pulmonary injury (e.g., hyaline membranes) in this model. Additional studies of hepatic I/R that quantify other measures of septic lung inflammation and injury are thus necessary to fully characterize this complex interaction. We consider the most important of these to be NF-B activation, cytokine expression profiles, mediator concentrations in BALF, and anatomic indices of pulmonary damage.

By what mechanisms were serum TNF- levels increased during hepatic ischemia (Figures 2A) and after reperfusion (Figure 4A)? We postulate different mechanisms during ischemic and reperfusion phases, respectively. Venous drainage from ischemic liver tissue was interrupted by vascular clamping. Therefore, enhanced cytokine production and export from these regions into the systemic circulation cannot account for the increased serum TNF- at t = 105 min in EC + I/R animals (Figure 2A). It is equally problematic to implicate increased TNF- production by the residually perfused right and caudate hepatic lobes during the ischemic phase, given the similar serum endotoxin concentrations in EC + I/R and EC + Sham animals. Considering the hepatobiliary metabolism of TNF- (3, 4, 9), reduced hepatic clearance of TNF- during ischemia may have increased peak circulating TNF- levels. We consider this unlikely, as we previously showed that the disappearance kinetics of infused recombinant TNF- were not altered by reductions in hepatic blood flow of approximately 75% after portacaval shunting in rats (10). We therefore favor the possibility that endotoxin spillover past the ischemic liver led to increased pulmonary endotoxin clearance in EC + I/R rats compared with EC + Sham animals. In this scenario, subsequent pulmonary synthesis and release of TNF- would be enhanced as shown by Halvorsen and coworkers in swine (28). However, our experiments were not designed to discriminate between hepatic versus pulmonary endotoxin clearance and TNF- secretory responses, which should be addressed in future studies.

Besides accentuating peak serum TNF- levels, focal liver ischemia after gram-negative sepsis was associated with an increased area under the cytokine concentration-time curve as reperfusion progressed. Thus, circulating TNF- levels were higher than time-matched EC + Sham values over the entire 24-h protocol (Figure 4A). The mechanisms for such higher reperfusion-phase TNF- concentrations in infected rats during evolving hepatic injury appear multifactorial. Washout of TNF- from ischemic liver regions after 1 h of reperfusion may account for the early reperfusion-phase differences in serum TNF- concentrations between EC + I/R and EC + Sham rats. Thereafter, a dissociation occurred between rising serum endotoxin levels and decreasing bacteremia in EC + I/R and EC + Sham animals, which we attribute to the administration of bactericidal antibiotics. Disproportionately rising serum endotoxin levels in EC + I/R animals suggest that hepatic TNF- clearance may have also been depressed at t = 6 h and beyond. Alternatively, these increased postischemic serum endotoxin levels may have stimulated subsequent TNF- and IL-6 production up to t = 24 h. In this context, the lack of significant earlier increases of circulating immunoreactive IL-6 in EC + I/R rats compared with sham-operated septic rats is noteworthy. Conceivably, differential hypoxic sensitivity of endotoxin-induced TNF- versus IL-6 transcriptional activation, as well as altered stability of cytokine-specific message or impaired posttranslational cytokine processing in the postischemic liver, may have played contributory roles.

Excessive bioavailability of endotoxin-induced leukotrienes contributes to mortality and lung inflammation during D-galactosamine-induced hepatitis and chronic portacaval shunting in rats (4, 9, 10). Lipoxygenase inhibition in this setting attenuates adverse sequelae, including mortality, even as circulating TNF- levels are suppressed (9, 10), suggesting that liver dysfunction primes the host to septic lung injury via a TNF-:leukotriene axis of inflammation. However, we recently reported that lipoxygenase inhibition failed to reduce circulating TNF- or improve the outcome of endotoxemia after cholestatic biliary cirrhosis in conscious rats (21). Thus, the interplay between TNF- and LTs may vary with the type of liver dysfunction or its timing relative to gram-negative infection (22, 27). In the present study, decreases in serum ALT in MK-886 + EC + I/R animals along with suppressed postbacteremic TNF- levels support a hepatoprotective as well as an anti-TNF- effect of MK-886. Nonetheless, these MK-886- induced reductions in circulating TNF- did not improve survival, which prevented assessment of its effects on lung PMN influx which is maximal 24 h after endotoxemia (9, 10). Additional studies are warranted to address this limitation. The apparent beneficial effect of MK-886 on hepatic I/R injury shown here contrasts with the ineffectiveness of the leukotriene synthesis inhibitor L663,536 (Merck Frosst) on liver injury as assessed by plasma ALT levels and liver necrosis (29). Although serum TNF- was not measured in that study, our data agree with reports of protection against postischemic liver injury by anti-TNF- antiserum (16, 18) or compounds such as FK506 (30) that suppress endotoxin-induced TNF- biosynthesis. The attenuation of I/R-mediated hepatic damage and increased serum TNF- concentrations in MK-886 + EC + I/R animals further raises the possibility that leukotriene synthesis inhibition may forestall evolving postbacteremic liver injury and thus, lung inflammation in this setting.

In summary, liver-lung interactions during gram-negative bacteremic sepsis and their impact on lung inflammation and mortality are incompletely understood. The variety of models of liver dysfunction, limited knowledge of organ-specific cytokine expression, and unexpected diversity of potential molecular pathways (31) are all factors. Further analyses of the boundary conditions by which changes in liver function affect the lungs early in sepsis are essential for improved knowledge of these organ interactions during critical illness.