INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The complex events by which hematogenous infection of the lungs with gram-negative bacteria results in alveolar hypoxia and inflammation that progress to the acute respiratory distress syndrome (ARDS) are incompletely understood. An early step in this sequence is thought to be overexpression of inflammatory cytokines within the lower respiratory tract, most notably TNF-, IL-1, and IL-1, following exposure to gram-negative microbial products (1). Subsequent export of these mediators into pulmonary venous blood likely contributes to the multiple organ dysfunction syndrome associated with sepsis-induced ARDS (4, 5). However, neither the cytokine expression profiles in intact lungs after gram-negative bacteremic infection, nor the mechanisms that regulate them have been fully defined. In this context, recent data suggest that regulation of TNF- and IL-1 gene expression in the lungs early after gram-negative bacteremia is considerably more complex than was initially suspected. Coinduction of anti-inflammatory cytokines, among them TGF-1, modulate sepsis-related TNF- and IL-1 biosynthesis (6, 7), and genes regulating immunomodulatory second messengers may further influence postbacteremic cytokine production. Of these, prostaglandin (PG)H synthase-2 (COX-2) is rate-limiting for the synthesis of cyclooxgenase-derived eicosanoids, including the anti-inflammatory agonist PGE₂ (8, 9).

Besides such mediator interactions, the prevailing alveolar PO₂ is a potentially important, though relatively unexplored, physiologic factor that may influence postbacteremic cytokine expression in the lungs. Decreases in alveolar O₂ tensions occur during cytokine-mediated cardiopulmonary dysfunction, pneumonia, and gravity-dependent atelectasis contributing to ventilation/perfusion mismatch. Although commonly viewed as a consequence of septic lung inflammation, alveolar hypoxia may itself modulate inflammatory cytokine production (10), in part by corresponding changes in COX-2 expression. Thus, hypoxic enhancement of LPS-induced TNF- and IL-1 protein secretion in alveolar macrophages coincides with decreases in COX-2 message and PGE₂ (9).

TNF- and IL-1 production by the liver and the lungs are thought to play pivotal roles in ARDS associated with gram-negative sepsis (1, 11). Even so, it is unknown whether postbacteremic reductions in the hepatic and pulmonary O₂ supply result in different organ-specific cytokine responses. We previously showed that intraportal Escherichia coli (EC) infection of perfused rat livers followed by brief secondary hypoxic stress (PO₂ ~ 46 mm Hg for 30 min) and reoxygenation downregulates TNF-, IL-1, and IL-1 gene expression, compared with EC-challenged normoxic livers (12, 13). Moreover, reductions in EC-induced IL-1, but not TNF- expression by such hypoxia/reoxygenation (H/R) were abrogated by the xanthine oxidase (XO) inhibitor allopurinol (12, 13). In this report, we tested the hypothesis that secondary alveolar H/R suppresses EC-induced TNF-, IL-1, and IL-1 expression in perfused rat lungs as in the liver. Hypoxic sensitivity of postbacteremic cytokine production was correlated with expression of TGF-1, COX-2 and its byproduct PGE₂, tissue antioxidant status as reflected by reduced glutathione levels (14), and pulmonary microvascular responses as assessed by the fluid filtration coefficient (K_f) (15, 16). Allopurinol-pretreated lungs were also studied to determine the effects of XO-derived reactive O₂ species (ROS) on cytokine production in this model.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Animals

Male specific pathogen-free Sprague-Dawley rats (275-350 g; Harlan, Indianapolis, IN) were housed in positive-pressure isolation carrels with free access to standard rat chow and water before experiments. Studies were performed according to NIH guidelines and were approved by the Animal Care Committee of Saint Louis University.

Lung Harvesting for ex vivo Perfusion

Anesthesia was induced with sodium pentobarbital (50 mg/kg intraperitoneally), after which a tracheostomy was aseptically performed with a modified 14-g Teflon catheter. Animals were ventilated (VT = 10 ml/kg; f = 25 breaths/min) with a rodent ventilator (Harvard Apparatus, South Natick, MA) using a 15% O₂/6% CO₂/79% N₂ gas mixture. A carotid artery was cannulated with PE-50 tubing through which the animals were exsanguinated following anticoagulation with sodium heparin (500 U intravenously). The lungs were aseptically harvested via a sternotomy and cannulation of the main pulmonary artery by a PE-280 polyethylene catheter (17). Lung perfusion with preoxygenated perfusate (PO₂ > 100 mm Hg, see below) was begun at an initial rate of ~ 5 ml/min. The left ventricle was incised to allow egress of perfusate and passage of a PE-320 catheter into the left atrium. The time required for harvesting was ~ 10 min following pulmonary arterial cannulation. While continuing perfusion, the heart and lungs were excised en bloc, placed within the cabinet for ex vivo perfusion, and suspended from a calibrated force displacement transducer (Model FT 03; Grass Instrument Co., Quincy, IL) to confirm isogravimetric conditions. The lungs were briefly hyperinflated to reverse atelectasis by raising airway pressure (Paw) to 10 cm H₂O for 5 s. Organs with uneven perfusion were discarded.

Perfusion Apparatus and Perfusate Composition

Perfusions were in the recirculating mode in a temperature-controlled (37.0 ± 3° C) humidified plexiglass cabinet by a peristaltic pump (Masterflex; Cole Parmer, Chicago, IL) from a venous reservoir through a bubble trap and into the pulmonary artery (17). Flow rates (18-20 ml/min) were confirmed in each experiment by timed collection of perfusates. Mean pulmonary artery pressure (Ppa) and mean left atrial pressure (Pla) were measured with saline-filled catheters connected to Model 59-DTS-H transducers (Baxter Corp., Irvine, CA) leveled to the midposition of the lungs. The outflow from the left atrium was directed through a waterfall circuit to maintain Pla at 0-3 cm H₂O. Paw was monitored from the tracheostomy tube. Ppa, Pla, Paw, and lung weight were continuously monitored by a polygraph. Perfusates were equilibrated with 15% O₂/6% CO₂/79% N₂ before use during lung harvesting. Respiratory f was adjusted to achieve initial perfusate pH values of 7.35-7.45; PCO₂ < 45 mm Hg; and PO₂ between 100-150 mm Hg (IL-1306 Blood Gas analyzer; Instrumentation Labs, Lexington, MA).

Before each experiment, circuits were sterilized with 3% H₂O₂ and flushes of sterile 0.9% NaCl. To confirm sterile conditions, quantitative streak-plate cultures of perfusate before and after circuit priming were performed on nutrient agar (37° C, 24 h) (12, 13, 18). Organs subsequently found to be perfused under contaminated baseline conditions were excluded from analysis. Residual contamination of circuits with endotoxin was controlled by weekly depyrogenation with an endotoxin-neutralizing eluting buffer (neutralizing potency, 50,000 EU of endotoxin/cm², PyroCLEAN; Alerchek, Inc., Portland, ME) and by changing disposable circuit elements. Measurements of endotoxin in perfusate recirculating in the apparatus before lung placement were  2 ng/ml by a quantitative Limulus assay (QCL-100, M.A. Whittaker Bioproducts, Walkersville, MD). Krebs-Ringer-Henseleit bicarbonate buffer (KHRB) was prepared in endotoxin-free glassware using nonpyrogenic U.S.P. H₂O. Perfusates were freshly prepared by adding 10 mM glucose, canine serum (5% vol/vol; endotoxin content, < 5 pg/ml) for bacterial opsonization (12, 13, 18); penicillin (100 U/L), and streptomycin (0.1 mg/ml). After adding low-endotoxin bovine serum albumin (4% wt/vol, lot #L588404; Intergen, Newark, NJ) to maintain oncotic pressure (17) and pharmacy-grade polymyxin B sulfate (Roerig Pfizer, New York, NY) in a final concentration of 25 ng/ml to neutralize albumin-associated endotoxin, perfusates were filtered (0.22 µm). The endotoxin content of perfusates prior to circuit priming was < 0.2 ng/ml by the Limulus assay.

E. coli Cultures

E. coli serotype O55:B5 (EC) was obtained as strain #12014 from the American Type Culture Collection (ATCC, Bethesda, MD) and resuspended in NS to 10⁹ colony-forming units (cfu)/ml as enumerated by a hemacytometer with serial dilutions (12, 13, 18). Actual inocula were verified by duplicate quantitative cultures (37° C, 24 h). Inocula were kept at 4° C until infusion into the circuit upstream from the lungs.

Experimental Protocol

Lungs (total n = 43) were isogravimetrically equilibrated for 30 min before experiments to ensure steady-state conditions of Ppa, Pla, Paw, lung weight, and gas exchange (Figure 1). Following baseline measurements and perfusate sampling at t = 0, six groups were studied over 180 min after hematogenous infection with 10⁹ viable EC in 1 ml or isovolumetric NS given over 1-2 min after t = 0: (1) Normoxic EC controls (n = 7) and (2) NS controls (n = 7) were ventilated with 15% O₂/6% CO₂/79% N₂ throughout the experiment; (3) EC + H/R (n = 7) and (4) NS + H/R lungs (n = 6) received normoxic ventilation until t = 30 min, after which normocapnic alveolar hypoxia (Pa_O2 ~ 10 mm Hg) was induced for 90 min by ventilating the lungs with 5% CO₂/ 95% N₂ during constant-flow perfusion. Reductions in perfusate PO₂ averaged 66 ± 4% (mean ± SEM) from baseline and as verified in each preparation, occurred within 30 min of induction of alveolar hypoxia. At t = 120 min, hypoxic lungs were reoxygenated with 15% O₂/ 6% CO₂/79% N₂, followed by ventilation for an additional 60 min until t = 180 min. Parallel experiments were performed to determine the effects of lung harvesting and equilibration on cytokine expression in organs removed from the circuit at t = 0 prior to EC or NS challenges (n = 3). To compare the effects of postbacteremic hypoxia with those of combined EC + H/R, additional normoxic or hypoxic EC-challenged lungs (n = 3 each) were frozen in liquid N₂ at t = 120 min, corresponding to peak hypoxia without reoxygenation of EC + H/R lungs. The effects of xanthine oxidase inhibition on cytokine expression after EC + H/R were also studied with pharmacy-grade allopurinol (ALLO, Burroughs Welcome, Research Triangle Park, NC) given as 50 mg/kg by gavage 18 h prior to lung harvest. An additional 3 mg/kg of ALLO was given intravenously before lung harvest and a final conc. of 500 µM was added to perfusates prior to t = 0 in (5) ALLO + EC + H/R lungs (n = 4) and in (6) ALLO + NS + H/R lungs (n = 3) (12, 13, 18).

View larger version (19K): [in this window] [in a new window]   Figure 1.   Schematic depiction of the experimental protocol for ex vivo lung perfusion over 180 min following pulmonary intravascular infection with 109 live E. coli (EC, serotype O55:B5) or isovolumetric infusion of sterile 0.9% NaCl (NS) at t = 0. Definition of abbreviations: H/R = hypoxia/reoxygenation; Kf = microvascular fluid filtration coefficient.

At t = 0, 30, 60, 90, 120, 150, and 180 min, perfusate samples (1 ml) were analyzed for pH, PCO₂, and PO₂, and for circulating EC cfu by streak-plating 10 µl on nutrient agar (37° C, 24 h). Additional aliquots of perfusate were immediately placed on ice and frozen at 70° C for subsequent analyses of TNF-, IL-1 isoforms, and lactate dehydrogenase (LDH) (Procedure no. 228-UV; Sigma Chemical Co., St. Louis, MO) as an indicator of endothelial integrity. Fresh perfusate was added back to the system at each time point to replace losses.

Lung K_f Measurements

The microvascular fluid filtration coefficient (K_f) was determined with the lungs in an isogravimetric state at the start of experiments (t = 0) and after 180 min by dividing the slope of lung weight gain 3-5 min after elevation of the Pla during double vascular occlusion as validated for perfused rat lungs (15, 16). Resultant changes in Pla corresponding to vascular filling and filtration phases were confirmed in each preparation by visual inspection of the Pla tracings.

Post-perfusion Studies

After sample collection and double occlusion measurements at t = 180 min, time-dependent changes in each preparation were evaluated by the response of the pulmonary vasculature to the vasoconstrictor PGF₂ (Sigma). A positive response was taken to be a rise in mean Ppa of  5 mm Hg over 1-3 min after a 300 µg bolus infusion (17). In certain experiments, lungs were then airway-fixed with cacodylate-buffered glutaraldehyde for 30 min at a transpulmonary pressure of 20-22 cm H₂O, followed by fixation in glutaraldehyde overnight at 5° C (4). Lung sections (6-µm) were stained with hematoxylin and eosin and clearance of formed elements, and interstitial or alveolar edema were evaluated in a blinded manner. Lungs not undergoing fixation were frozen in liquid N₂ and stored at 70° C, except for a standardized section of the left lower lobe, which was dried to a constant weight at 37° C for determining the wet/dry weight (W/D) ratio (4).

Bioactive TNF- Measurements

TNF- bioactivity in duplicate samples of perfusate was quantitated with mycoplasma-free, actinomycin D-treated murine fibroblasts as previously described (4, 12, 13, 18).

Immunoreactive IL-1 and IL-1

IL- was analyzed in duplicate aliquots of perfusates at t = 0, 60, 120, and 180 min by a RIA for rat IL-1 with a sensitivity of 10 pg/ml (Endogen, Cambridge, MA). Specimens (0.5 ml) were extracted with 1 ml of chloroform (19) and centrifuged at 10,000 × g at 4° C, after which the aqueous phase was again extracted before RIA. IL-1 in perfusates from similar time points was determined in duplicate after chloroform extraction by an ELISA (12) sensitive to rat IL-1 (Genzyme). Cell-associated IL-1 concentrations were determined in lung tissue samples (> 0.5 g) after homogenization in a protease inhibitor solution and lysis by four freeze-thaw cycles (4). Final IL-1 and IL-1 concentrations (pg/mg of protein in tissue lysates) were determined by the bicinchoninic acid protein assay (Pierce, Rockford, IL), with correction for tissue H₂O content.

Northern Hybridization Analyses

Total RNA was isolated from 500-1,000 mg of liquid N₂-frozen tissue and RNA integrity confirmed spectrophotometrically and by ethidium bromide-stained gels (12, 13, 18). Samples were electrophoresed for 4 h on 2% formaldehyde gels followed by transfer to nylon membranes. An 1100 bp cDNA of murine TNF- (11, 18), a 1301-bp cDNA for murine IL-1 and a 1300-bp cDNA for murine IL-1 (12) were used in hybridizations, as was a 1050-bp fragment for human TGF-1 (20). A murine inducible cyclooxygenase PGH synthase-2 cDNA was purchased from Caymen Chemical Co. (Ann Arbor, MI) and a human glyceraldehyde 3-phosphate dehydrogenase (GAPDH) cDNA from Clontech (Palo Alto, CA). The cDNA probes were random primer-labelled with ³²P, and after hybridizations membranes were washed at 65° C prior to autoradiography at 70° C (11, 12, 18). Steady-state mRNA levels were measured in a minimum of three lungs per group by densitometry in the linear range of detection (Molecular Dynamics, Sunnyvale, CA), with signals normalized for differences in RNA loading using the GAPDH values.

Immunoreactive PGE₂

PGE₂ levels in circulating perfusates were determined in duplicate (18). Samples were obtained at t = 0, at peak hypoxia in EC + H/R lungs (t = 120 min), and at t = 180 min, with corresponding samples in EC controls. PGE₂ was determined in duplicate 500 µl samples, with recovery assessed by a [³H] PGE₂ tracer. To confirm results, PGE₂ concentrations were also determined by an ¹²⁵I-RIA (Amersham Life Sciences, IL).

Tissue GSH

Lung concentrations of GSH, a sensitive indicator of intrapulmonary oxidative stress (14), were measured in frozen lung samples in duplicate by HPLC separation as previously described (12). Results are expressed as µmol/g tissue weight.

Statistical Analyses

Data are presented as means ± SEM. Repeated-measures analysis of variance was performed to analyze sequential changes in within-group variables. Post hoc comparisons were performed when appropriate by Student-Newman-Keuls tests. Specific between-group variables were compared by analysis of variance, bivariate regression, or by two-tailed t tests for independent groups. Significance was accepted for p values < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sequential E. coli Bacteremia and Alveolar H/R: Effects on Lung GSH Levels, Microbial Clearance, and TNF- Gene Expression

Following hematogenous pulmonary infection or NS infusion at t = 0, PO₂ values in recirculating perfusates remained constant during normoxic perfusion over 180 min. As expected, perfusate PO₂ tensions fell rapidly with induction of alveolar hypoxia at t = 30 min in EC + H/R and NS + H/R lungs, stabilizing at 51 ± 2 mm Hg over the 90-min hypoxic interval (p < 0.01 versus time-matched normoxic EC or NS values). Upon reoxygenation, perfusate PO₂ levels promptly returned to pre-hypoxic values in both EC- and NS-challenged lungs. Despite these changes in alveolar gas and perfusate O₂ tensions during hypoxia and subsequent reoxygenation, lung GSH levels did not change from baseline values (0.75 µmol/g) after 180 min of perfusion in the EC + H/R group (0.80 ± 0.3 µmol/g), or from time-matched normoxic EC values (0.50 ± 0.2 µmol/g). Similarly, no changes in GSH levels were detected at t = 180 min in NS + H/R lungs (0.61 ± 0.1 µmol/g) versus normoxic NS control values (0.70 ± 0.1 µmol/g).

The relationship between these physiologic events and bacterial clearance kinetics in perfusates of EC-infected lungs is shown in Figure 2 (top). No bacterial growth occurred at any time during perfusion of normoxic NS or NS + H/R lungs (not shown). For equivalent bacterial inocula as in normoxic EC controls, alveolar hypoxia and its resolution in EC + H/R lungs affected neither peak cfu's at t = 30 min nor the duration of bacteremia. Despite these findings, circulating bioactive TNF- levels were markedly enhanced by postbacteremic reductions in the alveolar PO₂ in EC + H/R lungs, compared with cytokine concentrations during normoxic pulmonary infection (Figure 2, bottom). Thus, the progressive rise in perfusate TNF- levels in normoxic EC controls which peaked at 247 ± 56 U/ml by 180 min (p < 0.005 versus time-matched NS value) was further accentuated in EC + H/R lungs, in which similarly-timed TNF- values averaged 509 ± 49 U/ml (p < 0.0001 versus peak normoxic EC concentrations). This potent additive effect of secondary H/R on export of TNF- from infected lungs was already evident prior to reoxygenation. Perfusate TNF- levels were 375 ± 45 U/ml at peak hypoxia (t = 120 min) in EC + H/R preparations compared with time-matched values of 172 ± 30 U/ml in normoxic EC lungs (p < 0.005). However, such differences in TNF- concentrations between the EC + H/R and the normoxic EC group became greater during the reoxygenation phase (Figure 2, bottom). Of note, the stimulatory effects of reductions in the alveolar PO₂ on TNF- secretion after bacteremic lung infection contrasted with the lack of effect of alveolar hypoxia on noninfected NS + H/R lungs to increase TNF- above basal levels observed in normoxic NS controls (Figure 2, bottom).

View larger version (27K): [in this window] [in a new window]   Figure 2.   (Top) Bacterial disappearance kinetics in perfusates of E. coli (EC)-infected lungs during normoxia or secondary alveolar hypoxia/reoxygenation (H/R) showing no intergroup differences in peak colony-forming unit (cfu) values or rate of clearance. No bacterial growth occurred at any time point during perfusion of NS or NS + H/R lungs (not shown). (Bottom) Sequential changes in circulating bioactive tumor necrosis factor- (TNF-) levels. Data are means ± SEM. *p < 0.05 versus group-specific baseline value at t = 0; p < 0.05 versus time-matched normoxic NS value; p < 0.005 versus time-matched normoxic NS value; §p < 0.005 versus time-matched normoxic EC control; p < 0.0001 versus time-matched normoxic EC value.

The hypoxic enhancement of postbacteremic TNF- protein secretion in EC + H/R compared with normoxic EC lungs was paralleled by group-specific changes at the mRNA level (Figure 3). Bacteremic infection followed by alveolar H/R more than doubled the number of TNF- transcripts after 180 min of perfusion compared with lung infection under normoxic conditions (p < 0.05). As noted for the enhanced secretion of TNF- protein by EC + H/R lungs, this O₂-sensitive postbacteremic cytokine expression did not depend on reoxygenation-specific events. Significant differences in TNF- mRNA accumulation between EC + H/R and EC groups were already present by t = 120 min at peak hypoxia (p < 0.05, data not shown). Alveolar H/R without preceding infection in NS + H/R lungs also increased TNF- message compared with normoxic NS controls (Figure 3), despite no increase in circulating TNF- protein in NS + H/R perfusates by t = 180 min (Figure 3, bottom). However, such group-specific differences in TNF- transcripts were not evident at t = 120 min (peak hypoxia) in NS-challenged lungs as for EC + H/R organs (not shown).

View larger version (36K): [in this window] [in a new window]   Figure 3.   Northern hybridization analysis of TNF- mRNA in lung homogenates 180 min after pulmonary intravascular infection with 109 E. coli (EC, serotype O55:B5) or infusion of 0.9% NaCl (NS) followed by normoxic perfusion, or by 90 min of secondary alveolar hypoxia/reoxygenation (EC + H/R, NS + H/R). (Top) Means ± SEM of determinations of the ratio of TNF- to glyceraldehyde phosphate dehydrogenase (GAPDH) mRNA densitometric signals (minimum n = 3 lungs per group). (Bottom) Representative TNF- and GAPDH blots from individual lungs in each group. *p < 0.05 versus NS controls and EC + H/R.

Xanthine Oxidase Inhibition: Lack of Effect on Hypoxic Enhancement of E. coli-induced TNF- Secretory Responses

Inhibition of xanthine oxidase-derived ROS by allopurinol pretreatment of rats plus drug addition to perfusates had no effect on the amount or rate of clearance of circulating E. coli by ALLO + EC + H/R lungs (peak cfu at t = 30 min, 1,900 ± 400 cfu/ml), compared with EC + H/R or normoxic EC groups (Figure 2, top). Xanthine oxidase inhibition also failed to attenuate the enhanced postbacteremic export of TNF- from infected lungs undergoing alveolar H/R (Table 1). In this context, interference by allopurinol on TNF--mediated L929 cell cytotoxicity in perfusate samples was excluded by studies in which allopurinol at 50, 500, and 1,000 µM in fresh perfusate had no effect on the lytic effects of low (50 U) or high (250 U) doses of murine rTNF- on L929 cells in vitro (not shown). TNF- values of ALLO + NS + H/R lungs were likewise similar to time-matched NS  H/R cytokine concentrations (Table 1).

Alveolar H/R and Postbacteremic IL-1 Expression

In contrast to the increased TNF- in venous effluents of EC and EC + H/R lungs compared with NS-challenged organs, no circulating IL-1 was found through 180 min after normoxic EC infection or sequential EC + H/R. Similarly, neither NS + H/R nor normoxic NS perfusates contained immunoreactive IL-1. Although detectable, cell-associated IL-1 levels did not differ across groups at t = 180 min, averaging 587 ± 126 pg/mg lung protein in normoxic EC, 732 ± 160 pg/mg in EC + H/R, 444 ± 81 pg/mg in NS, and 623 ± 97 pg/mg in NS + H/R lungs, respectively (p = NS). Even so, postbactermic IL-1 mRNA abundance at t = 120 min representing peak hypoxia was enhanced in EC + H/R lungs compared with EC organs (not shown), and these differences became more striking by t = 180 min, with maximal IL-1 transcript accumulation (p < 0.05, Figure 4). No significant differences in IL-1 message were found between NS + H/R and NS lungs at this time point.

View larger version (38K): [in this window] [in a new window]   Figure 4.   Northern hybridization analysis of IL-1 mRNA in lung homogenates 180 min after pulmonary intravascular infection with 109 E. coli (EC, serotype O55:B5) orinfusion of 0.9% NaCl (NS), followed by normoxic perfusion or by secondary hypoxia/reoxygenation (H/R). (Top) Means ± SEM of determinations of the ratio of IL-1 to glyceraldehyde phosphate dehydrogenase (GAPDH) mRNA densitometric signals (minimum n = 3 lungs per group). (Bottom) Representative IL-1 and GAPDH blots from individual lungs in each group. *p < 0.05 versus normoxic EC control.

Perfusate IL-1 concentrations after bacteremic lung infection also remained low through 180 min irrespective of alveolar H/R, with peak circulating IL-1 levels indistinguishable between EC + H/R and EC controls (9 ± 3 versus 5 ± 3 pg/ ml, respectively, at t = 180 min). As for IL-1, secretion of IL-1 was minimally affected in NS + H/R lungs (15 ± 3 pg/ ml at t = 180 min), which was not different from normoxic NS controls. There were also no intergroup differences in cell- associated IL-1; values were 114 ± 31 pg/mg lung protein in normoxic EC versus 87 ± 10 pg/mg in EC + H/R lungs, compared with 60 ± 10 pg/mg in NS lungs and 44 ± 12 pg/mg in NS + H/R organs. In contrast to elevated TNF- and IL-1 mRNA profiles in EC + H/R lungs, IL-1 transcripts were not increased by either bacteremia or subsequent alveolar hypoxia in EC + H/R organs at t = 120 min (not shown), or by 180 min (Figure 5).

View larger version (36K): [in this window] [in a new window]   Figure 5.   Northern hybridization analysis of IL-1 mRNA. (Top) Means ± SEM of the ratio of IL-1 to GAPDH mRNA densitometric signals in the same lung homogenates reported in Figures 3 and 4. (Bottom) Representative IL-1 and GAPDH blots from individual lungs in each group.

TGF- and COX-2 Expression

There was no hypoxic enhancement or change in postbacteremic TGF- message levels in EC + H/R lungs, either at peak hypoxia (not shown), or by 180 min (Figure 6). Accumulation of COX-2 mRNA at peak hypoxia of after H/R was also not greater in EC + H/R lungs compared with normoxic EC controls, or in NS + H/R lungs compared with NS controls (not shown). Moreover, perfusate PGE₂ as an index of COX-2 enzyme activity was 177 ± 31 pg/ml in the EC + H/R group at peak hypoxia and 227 ± 51 pg/ml at t = 180 (p = NS), values similar to 169 ± 46 pg/ml and 207 ± 59 pg/ml, respectively, in time-matched normoxic EC. None of these PGE₂ concentrations differed from NS + H/R or normoxic NS values (not shown), even after determination of PGE₂ by RIA (data not shown).

View larger version (38K): [in this window] [in a new window]   Figure 6.   Northern hybridization analysis of TGF- mRNA. (Top) Means ± SEM of the ratio of TGF- to GAPDH densitometric signals (minimum n = 3 lungs/group). (Bottom) Representative TGF- and GAPDH plots from individual lungs in each group.

Pulmonary Hemodynamic and Microvascular Responses

Serial changes in Ppa during normoxic ventilation compared with secondary alveolar H/R are depicted in Figure 7. No differences were found across study groups in baseline (t = 0) values or at subsequent time points, irrespective of reductions in the alveolar PO₂, despite intact pressor responses to PGF₂ (+6.5 ± 3 mm Hg) at the conclusion of experiments. In further support of a general lack of time-dependent or perturbation-specific endothelial dysfunction in the preparations, no increases in release of LDH by the lungs were noted over time in any group from basal values of 11 ± 1 U/ml (not shown).

View larger version (29K): [in this window] [in a new window]   Figure 7.   Serial changes in mean pulmonary artery pressure (Ppa) over 180 min following intravascular infection with 109 E. coli (EC, serotype O55:B5) or infusion of 0.9% NaCl (NS) at t = 0, followed by normoxic perfusion or by secondary constant-flow alveolar H/R starting at t = 30 min. Data are means ± SEM.

Changes in lung K_f values, with and without bacteremia and alveolar H/R, are shown in Figure 8. Baseline K_f did not differ among groups, including values in ALLO-treated lungs (not shown). Postbacteremic K_f values increased in EC lungs after 180 min of perfusion compared with baseline (p < 0.05). However, time-matched K_f values were similar in the EC + H/R group, as were lung W/D ratios which averaged 9.0 ± 0.06 across all groups. K_f differences from the beginning to end of experiments were less pronounced in NS-challenged lungs regardless of H/R (Figure 8), and W/D ratios were similarly not affected by alveolar H/R.

View larger version (17K): [in this window] [in a new window]   Figure 8.   Comparison of group-specific changes in the lung microvascular fluid filtration coefficient (Kf) at baseline (t = 0) and after 180 min of perfusion. Data are means ± SEM. *p < 0.05 versus group-specific t = 0 value.

Lung Histological Findings

All specimens of airway-fixed lungs showed comparably mild peribronchial and perivascular edema formation which rarely impinged on distal airspaces. No group-specific interstitial or alveolar injury was found (not shown). Histological features were likewise similar in all groups of allopurinol-treated lungs irrespective of bacterial infection or alveolar H/R (not shown).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

TNF- and isoforms of IL-1 are canonical mediators of gram-negative bacterial sepsis that have been causally implicated in lung inflammation during sepsis-induced ARDS (1). The main finding of this study is that alveolar H/R early after gram-negative bacteremic lung infection enhances the expression of TNF- and IL-1 in the lower respiratory tract, compared with cytokine profiles during normoxic pulmonary infection. Moreover, such hypoxic enhancement of postbacteremic cytokine expression appears to be a selective rather than a generalized process, at least within the time constraints imposed by ex vivo lung perfusion as performed here. Thus, no corresponding increases in mRNA levels of IL-1 or the anti-inflammatory cytokine TGF-1 were found in EC + H/R lungs compared with EC controls. The novel finding that E. coli-induced expression of TNF- and IL-1 in the lungs is enhanced by secondary hypoxic stress within 2 h of infection identifies the alveolar O₂ tension as a potent immunomodulatory signal in the early postbacteremic phase of sepsis. Of note, the reductions in mean alveolar PO₂ to approximately 10 mm Hg and mean perfusate PO₂ to ~ 50 mm Hg that augmented TNF- and IL-1 expression were insufficient to alter bacterial clearance, tissue GSH levels, Ppa, K_f, or lung histology. Congruent with the lack of change in GSH levels in EC + H/R lungs, pretreatment with the XO inhibitor allopurinol failed to abrogate the hypoxic enhancement of TNF- and IL-1 mRNA accumulation and TNF- protein secretion (Table 1). Collectively, these results suggest that hypoxia-inducible expression of the inflammatory cytokines TNF- and IL-1 is upregulated by a mechanism independent of XO-derived ROS following sequential gram-negative bacteremic lung infection and otherwise well-tolerated reductions in the alveolar PO₂.

Although bacteremic lung infection and subsequent alveolar H/R occur commonly in the early phases of ARDS, the consequences of this combinatorial interaction on pulmonary cytokine production are unknown. One reason is that to date, independent lines of investigation have focused on sepsis- related cytokine production by the lungs, or on the effects of nonbacteremic pulmonary oxidative stress. The principal rationale for this study was to model for the first time the sequential nature of these events within intact lungs under dynamic conditions of ventilation and perfusion. In this context, only limited data are available concerning the effects of reductions in the cellular O₂ supply on cytokine-specific gene expression after stimulation by gram-negative endotoxin. Differences in the magnitude and duration of hypoxia in such reports, as well as variable reoxygenation paradigms and diversity of model systems further confound the issue. Hempel and coworkers reported enhanced release of TNF- and IL-1 from human alveolar macrophages after stimulation with E. coli LPS followed by 24 h of exposure to ambient O₂ tensions of < 1 mm Hg, without reoxygenation (9). Anoxia increased LPS-induced TNF- expression compared with normoxic LPS controls while causing no increases in IL-1 message or cell-associated IL-1 protein (9), similar to our findings at the whole-lung level (Figures 3 and 5). In another report (21), concurrent challenge of murine alveolar macrophages with E. coli LPS and 6 h of anoxia or especially, anoxia-hyperoxia enhanced expression of TNF- and macrophage inflammatory protein-1 by a mechanism inhibitable by the ROS scavenger dimethyl sulfoxide (21).

Our results confirm a complex and differential modulation of E. coli-induced TNF- versus IL-1 gene expression in the lungs by reduction-oxidation (redox) events associated with H/R, as shown by our previous findings in the liver (12, 13). In determining the acute effects of alveolar H/R on postbacteremic cytokine expression in ventilated lungs during constant-flow perfusion, we have shown that relatively brief (90 min) and arguably, clinically relevant decreases in the alveolar PO₂ upregulate the early-response cytokines TNF- and IL-1. Our results concerning changes in pulmonary cytokine expression induced by alveolar hypoxia agree with reports of enhanced TNF- message accumulation and protein secretion in LPS-challenged alveolar macrophages subjected to more severe and prolonged hypoxia (9), or to prolonged hypoxia and reoxygenation (21). However, further comparisons with our study are precluded by differences in design, including assessment of bacterial clearance, IL-1 and IL-1 expression, and tissue GSH levels in relation to pulmonary hemodynamics.

Our findings in EC + H/R lungs also confirm and extend data from a rat model of atelectasis produced by unilateral airway occlusion for 60 min (10). In that report, alveolar macrophages harvested from atelectatic but not ventilated lungs secreted more bioactive TNF- and IL-1 following subsequent in vitro challenge with E. coli LPS (serotype O55:B5, 10 µg/ ml) and culture for 24 h. Unlike our protocol (Figure 1), alveolar hypoxia preceded endotoxin exposure in that study (10), and cytokine mRNA levels, lung redox status, and effects of antioxidants were not assessed. A further distinction was our delay in starting hypoxia until 0.5 h after gram-negative bacteremia with normoxic ventilation, in an attempt to model the clinical scenario of early postbacteremic deteriorations in gas exchange. Because pulmonary cytokine responses to evolving hematogenous gram-negative infection may differ from those with only purified LPS, we used an E. coli inoculum corresponding to a defined end point in vivo (LD₂₅ over 24 h in intact rats) (22) as done previously in our studies in perfused rat liver (12, 13, 18). Our data offer further insight into the kinetics of lung bacterial clearance in relation to cytokine induction (Figure 2). These have not been previously defined in perfused lungs during alveolar H/R, particularly in the presence of rigorous control of microbial and endotoxin cross-contamination, as performed here.

Several mechanisms may account for the selective enhancement of TNF- and IL-1 expression in EC + H/R lungs, compared with infected lungs receiving normoxic ventilation. Chief among these is redox regulation of cytokine gene transcription, which during nonseptic severe hypoxic stress is initiated by ROS-mediated shifts in the intracellular thiol pool from GSH to glutathione disulfide (GSSG) and from reduced to oxidized thioredoxin (23). The transcription factor NF-B is subsequently activated via phosphorylation of its cytoplasmic anchoring protein IB (26, 27). Binding of NF-B to DNA sequence motifs follows in the promoter regions of TNF- and IL-1 isoforms, unless high, inhibitory levels of GSSG are present (23). However, it is unknown whether NF-B activation and binding vary over a spectrum of hypoxic stress of the lungs, especially when preceded by exposure to gram-negative endotoxin as performed here. That GSH levels were not reduced in EC + H/R lungs compared with normoxic EC controls suggests that the degree of alveolar hypoxia was mild in these studies. Determining if, and to what extent NF-B activation and cytokine gene transcription in infected lungs is further increased or prolonged by this degree of hypoxic stress will need to be addressed in future investigations.

The above considerations are relevant to the other major finding of this study, namely, that the liver and the lungs in the same species exhibit directionally opposite responses in postbacteremic TNF- and IL-1 expression following secondary hypoxic stress (12, 13, 18, 28). In this context, activation of NF-B in perfused rat livers receiving equivalent E. coli infection and H/R was significantly diminished at peak hypoxia along with TNF-, IL-1, and IL-1 expression at mRNA and protein levels despite preserved hepatic GSH levels (12, 13, 29). Differential activation of NF-B or other trans-acting factors after postbacteremic H/R of the liver versus the lungs may therefore account for the opposing changes in cytokine expression. We cannot resolve this issue from the present data. Still, the thesis that sepsis-induced cytokine expression is sensitive to the prevailing O₂ supply in an organ-specific manner is strongly supported by our finding that postbacteremic TNF- and IL-1 expression is downregulated in the liver by H/R but upregulated in the lungs (28). Our present finding that tissue GSH levels were unchanged by alveolar H/R in both EC + H/R and NS + H/R lungs further indicates that as for the liver, depletion of GSH by ROS is not a prerequisite for hypoxic modification of cytokine expression at the whole-lung level. These results are consistent with the observations that oxidant stress insufficient to reduce cellular GSH nevertheless modulates activation of NF-B and other transcription factors (23). We cannot exclude the additional possibility that an as yet unidentified hypoxia-responsive enhancer element as described in the human erythropoietin gene (30) promoted hypoxic upregulation of E. coli-induced TNF- and IL-1 expression in the lungs.

Xanthine oxidase as a source of ROS that activate NF-B and other transcription factors has been implicated in the upregulation of TNF-, IL-1, and other cytokines in parenchymal lung mononuclear cells of mice after 1 h of hemorrhagic shock (31, 32). These data obtained during reductions in the pulmonary O₂ supply under nonseptic, low-flow conditions cannot be readily extrapolated to postbacteremic reductions in cytokine expression during constant-flow alveolar H/R or hypoxic stress of the liver. Thus, XO inhibition neither diminished the enhanced TNF- production in EC + H/R lungs (Table 1) nor restored the depressed TNF- production in perfused EC + H/R livers (12). We consider it unlikely that incomplete inhibition of XO in this study explains these findings. The 500 µM concentration of allopurinol added to perfusates exceeds the 50 µM dose that reversed shock-induced increases in TNF-, IL-1, and TGF-1 (32). Moreover, equivalent doses of allopurinol abrogated the suppressive effects of H/R on E. coli-induced IL-1 production at the mRNA and protein levels in perfused liver (13). Also, pretreating rats with a similar allopurinol regimen reversed hypoxia-mediated decreases in circulating TNF- after 10⁹ E. coli bacteremia (33). These findings do not exclude the possibility that ROS from other cell sources stimulated TNF- and IL-1 expression in EC + H/R lungs, which should be explored in future studies.

Two additional mechanisms for hypoxic enhancement of postbacteremic cytokine expression were evaluated in this study. Decreased expression of anti-inflammatory cytokines in EC + H/R lungs appears to be an unlikely cause, given the lack of change in TGF-1 transcripts (Figure 6). We have not excluded the possibility that changes in TGF-1 protein or other anti-inflammatory agonists such as IL-4 or heat-shock proteins (7) were increased in EC + H/R lungs. Hypoxia may also increase LPS-induced expression of TNF- and IL-1 by inhibiting COX-2 expression, thereby removing tonic suppression of cytokine production owing to falls in PGE₂-mediated cyclic AMP formation (9). Such hypoxia-associated falls in PGE₂ synthesis or altered PGE₂ isomerase activity (34) are linked to reductions in cellular GSH levels. However, we detected no changes in tissue GSH levels, COX-2 message, or circulating PGE₂, even though ROS reduce inactivation of PGE₂ in perfused rat lungs (35). We cannot exclude changes in COX-2 expression in discrete cell populations within EC + H/R lungs, as we characterized COX-2 and cytokine expression only at the whole-lung level. Nevertheless, it remains unlikely that this mechanism accounted for our findings.

In summary, we have shown that postbacteremic reductions in alveolar oxygenation, as occur with septic lung injury in the critically ill, acutely modify the expression of the pleiotropic cytokines TNF- and IL-1. Recognizing the limitations of this "two-hit" model of lung inflammation, further study of sequential bacteremic lung infection and alveolar H/R will be important to define the boundary conditions of this interaction. This is especially the case with respect to lung infection by taxonomically distinct organisms, during regional alveolar hypoxia, and with differing strategies of ventilatory support.