INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Many features of lung injury in sepsis and severe infection are ascribed to lipopolysaccharides (LPS; endotoxin) of gram-negative bacteria. Harmful effects evoked by LPS in animal model include lung microcirculatory disturbances with pulmonary hypertension, edema formation, and severe impairment of gas exchange (1, 2). These events are ascribed mainly to an activation of inflammatory cells, with subsequent liberation of a large array of vasoactive and inflammatory mediators. In addition, direct cellular toxicity of LPS has been described (3).

In recent years, proteinaceous exotoxins produced by gram-positive or gram-negative bacteria have received increasing attention as potentially important instigators of systemic inflammation and multiorgan dysfunction. Escherichia coli hemolysin (HlyA) is a well-characterized representative of this group of bacterial virulence factors. The pathogenetic relevance of this agent has been established in animal models as well as in human infection, in which a strong association of HlyA production with disease, including pyelonephritis and sepsis, has been described (4). HlyA exerts a particularly potent action on neutrophils, monocytes, and endothelial cells, provoking the generation of inflammatory mediators such as reactive oxygen species, lipid mediators, cytokines, proteases, and nitric oxide (NO) (5). Indeed, HlyA has been identified as the most potent inductor so far described of the preformed phophatidylinositol-hydrolysis-related signal-transduction pathway in human neutrophils (9).

In perfused rabbit lungs, HlyA provoked marked pulmonary vasoconstriction and concomitant ventilation-perfusion (A/) mismatch, largely mediated by exotoxin-elicited thromboxane generation in the lung (10, 11). Vascular leakage, apparently independent of the prostanoid-related vasoconstrictor response, was also noted. Interestingly, prior admixture of the lung perfusate with low doses of LPS, which itself did not evoke lung vascular effects, "primed" the lung vasculature to respond to a subsequent HlyA challenge with a manifold pressor response (12). This vasoconstrictor response was again largely thromboxane mediated, and LPS-evoked induction of cyclooxygenase-2 (COX-2) is currently suspected as the event underlying this response (L. Ermert, F. Grimminger, and W. Seeger; unpublished results). In addition to the pressor response, marked edema formation was noted in these studies. The present study addressed the question of whether this increase in fluid filtration was due only to augmented microvascular pressure accompanied by the vasoconstrictor response, or whether an increase in pulmonary capillary permeability occurs as an independent event in response to the sequential LPS-HlyA challenge. For this purpose, the isolated, perfused rabbit lung model was changed to a pressure-controlled perfusion system, and the experiments were performed in the presence of a thromboxane receptor antagonist. Employing low doses of both LPS and HlyA, we found that prior priming of the rabbit lung vasculature with LPS provokes a manifold increase in the capillary filtration coefficient (Kfc) in response to a subsequent HlyA challenge. This constitutes a further, previously unrecognized example of synergism between bacterial endo- and exotoxins, which may be operative in septic lung failure.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Reagents

HlyA was kindly provided by S. Bhakdi (Institute for Microbiology, Mainz, Germany). The exotoxin was prepared by polyethylene glycol precipitation of culture supernatant, followed by centrifugation in a linear density-gradient preparation of glycerol, as previously described (4). The toxin was recovered in active and monomeric form from the gradient, and the endotoxin content of this preparation was reduced to ~ 3 ng LPS/µg protein. Toxin aliquots were stored at 70° C. The hemolytic titer was assessed directly before use of the toxin, and was expressed in hemolytic units (HU)/ml. The hemolysin protein concentration was determined with an enzyme-linked immunosorbent assay (ELISA) that uses a monoclonal anti-HlyA antibody to capture the antigen, and a second, polyclonal rabbit antibody for development. With this ELISA, it was found that a hemolytic titer of 1 HU/ml corresponded to a toxin protein concentration of ~ 0.1 µg/ml. Salmonella abortus equii endotoxin was kindly provided by C. Galanos (Max Planck Institute of Immunology, Freiburg, Germany), and the thromboxane receptor antagonist BM 13.505 was a generous gift from Boehringer AG (Mannheim, Germany). Hydroxyethylamylopectin (MW 200,000) was obtained from Fresenius (Oberursel, Germany). Dimethylthiazolyl-diphenyltetrazolium-bromide (MTT) was obtained from Sigma (Munich, Germany), and recombinant murine tumor necrosis factor- (TNF-) from Genzyme (Frankfurt, Germany). All other chemicals were obtained from Merck AG (Darmstadt, Germany).

Lung Model

The technique of isolated rabbit lung perfusion has been described previously (13). For the present study, the lung model was extended to allow either constant-flow or constant-pressure perfusion. Briefly, rabbits of either sex, weighing 2.6 to 2.9 kg, were anticoagulated with heparin and deeply anesthetized with a mixture of ketamine and xylazine. Tracheostomy was performed, and the animals were ventilated with room air, using a Harvard respirator (VT = 30 ml, frequency 24 breaths/min; positive endexpiratory pressure [PEEP] of 1 cm H₂O). After midsternal thoracotomy, large-bore catheters were placed into the pulmonary artery and the left atrium, and perfusion with sterile Krebs- Henseleit-hydroxyethylamylopectin buffer (KHHB) was started. The buffer contained 120 mM NaCl, 4.3 mM KCl, 1.1 mM KH₂PO₄, 24 mM NaHCO₃, 2.4 mM CaCl₂, 1.3 mM MgPO₄, and 2.4 g/L glucose, as well as 5% (wt/vol) hydroxyethylamylopectin as an oncotic agent. Sterilized perfusion-circuit tubing was used throughout the perfusion system. In parallel with the onset of artificial perfusion, the rabbits' room air ventilation was supplemented with 4% CO₂. After extensive rinsing of the lung vasculature, the lungs were perfused with recirculating buffer at a constant pulsatile flow of 100 ml/min (total recirculating volume: 350 ml). Left atrial pressure was set at 3 mm Hg (referenced at the hilum), and the whole system was equilibrated at 37° C. Lungs were suspended from a force transducer for continuous registration of organ weight. The perfusion flow rate was measured continuously with an ultrasonic flow probe designed for sterile tubing (Transonic Systems Inc., Ithaca, NY), and pulmonary arterial and venous pressures were monitored with pressure transducers. The system was then switched to pressure-controlled perfusion by inserting an arterial overflow reservoir, which was adjusted to a mean pulmonary arterial pressure (Ppa; referenced at the hilum) of 7 mm Hg. Kfc (normalized for wet lung weight) and the total vascular compliance were determined gravimetrically from the slope of the curve of the lung weight-gain induced by a 7.5-mm Hg-step elevation of both the venous and arterial reservoirs for 8 min (13). Lung weight gain was calculated as the difference in organ weight measured directly before and 5 min after each of these pressure-elevation maneuvers. The microvascular pressure was determined with the arterial and venous double occlusion technique (14).

Lungs included in the study: (1) had a homogeneous white appearance, with no signs of hemostasis, edema or atelectasis; (2) had initial pulmonary artery and ventilation pressures in the normal range; and (3) were isogravimetric during an initial steady-state period of at least 30 min.

Biochemical Measurements

TNF- was determined in a cytolytic cell assay with the mouse fibrosarcoma cell line WEHI 164, clone 13 (kindly donated by Dr. T. Espevik, University of Trondheim, Trondheim, Norway), as previously described (15). The WEHI cells (2 × 10⁴) were incubated with serial dilutions of perfusate in microtiter wells (Nunc, Roskilde, Denmark). After 18 h, MTT (5 mg/ml in phosphate-buffered saline [PBS], 100 µl/ well) was added. The reaction was stopped after 4 h by addition of 5% formic acid in 2-propanol, and the content of reduced MTT was read in a microELISA autoreader (570 nm). Recombinant murine TNF- served as standard in all assays. The sensitivity of the WEHI 164.13 cell assay ranged from 0.8 to 0.02 pg protein/cytolytic unit in the different experiments. In order to establish the source of the cytolytic activity as TNF-, an antiserum directed against human TNF-, having established cross-reactivity with rabbit TNF- (16), was added to the perfusates. It was found that this antiserum neutralized all cytolytic activity measured with the WEHI cells.

Potassium was measured according to standard techniques. Thromboxane B₂ (TxB₂) in samples of pulmonary venous effluent was measured with ELISA techniques, as recently described (17). LPS levels in the lung perfusate were measured according to the protocol of a Limulus-based photometric test (Coatest Endotoxin; Kabi Vitrum, Munich, Germany).

Experimental Protocol

After termination of the steady-state period with performance of the first hydrostatic challenge, the time was set at zero and lungs were perfused for 180 min in the absence or presence of 10 ng/ml S. abortus equi LPS, followed by bolus injection of 0.021 HU/ml HlyA or solvent (see Figure 1). Hydrostatic challenges were again performed at the end of the 180-min LPS priming period, as well as 30 and 60 min after HlyA administration. Double-occlusion maneuvers were performed in duplicate (15 s each) 5 min before each hydrostatic challenge. In all experiments, the thromboxane receptor antagonist BM 13.505 (5 µM) was admixed with the recirculating buffer fluid 10 min prior to HlyA or solvent administration. Perfusate samples for determination of TxB₂, potassium, and TNF- were taken every hour during the period of LPS or sham priming and 5, 15, 30, and 60 min for TxB₂ and potassium, and 60 min for TNF-, after application of HlyA.

View larger version (28K): [in this window] [in a new window]   Figure 1.   Perfusion flow in the different experimental groups. Hydrostatic challenges are indicated by hatched areas, intravascular administration of LPS, BM 13.505 and HlyA by arrows. Means ± SEM are given; error bars are missing when they fall within the symbol.

Data Analysis

For comparison of means, analysis of variance (ANOVA) was performed with Scheffé's post hoc test. If necessary, values were log-transformed to achieve a normal distribution, as assessed by the Kolmogorov-Smirnov test. Values of p < 0.05 were considered to represent a significant difference. All statistical procedures were performed with the SPSS (SPSS, Inc., Chicago, IL) for Microsoft Windows (Microsoft, Inc., Seattle, WA) analysis system.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In sterile control experiments, perfusion flow was virtually constant throughout the entire experimental period (Figure 1). Each hydrostatic challenge induced a transient and reproducible increase in the perfusion flow, owing to vascular recruitment and/or distension. In LPS-primed lungs, flow rates were nearly identical to those in sterile lungs (Figure 1). Application of HlyA induced an immediate vasoconstriction, comparable in magnitude in both sterile and LPS-pretreated lungs, effecting a transient decrease in perfusion flow (Figure 1).

As anticipated from the constant-pressure mode of perfusion, microvascular pressures did not change throughout the experimental period, and displayed no significant variations among the four experimental groups (Table 1). Single values never surpassed 6 mm Hg. Vascular compliance remained unchanged in all experimental groups (Table 1).

Baseline Kfc values were in a range similar to previously reported values (10, 13). In sterile control lungs, Kfc values did not increase for up to 240 min (Figure 2). Administration of LPS alone had no effect on Kfc values within this observation period. At the chosen dosage of 0.021 HU/ml, HlyA induced only a minor, insignificant (p > 0.15) increase in Kfc in sterile lungs (Figure 2). In contrast, after LPS pretreatment, the same HlyA dose provoked a manifold increase in Kfc. This was paralleled by extensive edema formation during the maneuvers of pressure elevation, reflected by the hydrostatic-challenge-induced weight gain (Table 1). In an additional set of experiments, lungs were perfused for 180 min in the absence of LPS, and both endotoxin and HlyA were then administered simultaneously. In these lungs, the Kfc values and hydrostatic-challenge-induced weight gain did not differ significantly from those provoked by HlyA challenge in sterile lungs (data not shown in detail).

View larger version (24K): [in this window] [in a new window]   Figure 2.   Influence of LPS priming and HlyA challenge on the capillary filtration coefficent (Kfc). Means ± SEM for the different experimental groups and the different time points are given; error bars are missing when they fall within the symbol (**p  0.01 compared with all other groups).

In sterile or LPS-perfused lungs, only small amounts of TxB₂ were released into the recirculating buffer fluid (Figure 3). Application of HlyA after 3 h of sterile perfusion provoked only a moderate TxB₂ release, whereas in LPS-primed organs, marked liberation of TxB₂ occurred in response to subsequent challenge with HlyA (Figure 3).

View larger version (19K): [in this window] [in a new window]   Figure 3.   Influence of LPS priming and HlyA challenge on thromboxane release into perfusate. Means ± SEM for the different experimental groups and the different time points are given; error bars are missing when they fall within the symbol.

In lungs perfused with sterile buffer fluid, only marginal amounts of TNF- were detected. In LPS-primed organs, progressive intravascular liberation of this cytokine was noted (Figure 4), which was, however, not enhanced by subsequent stimulation with HlyA (Figure 4). Application of HlyA provoked an immediate, marked potassium release, which was independent of prior LPS priming (Figure 5) and was followed by some reuptake of potassium in the 60-min perfusion period following HlyA administration.

View larger version (20K): [in this window] [in a new window]   Figure 4.   Release of TNF- into the recirculating buffer fluid in response to LPS priming and subsequent HlyA challenge. Means ± SEM for the different experimental groups and the different time points are given; error bars are missing when they fall within the symbol.

View larger version (21K): [in this window] [in a new window]   Figure 5.   Release of potassium into the recirculating perfusate in response to HlyA challenge. The buffer-fluid potassium values at 180 min were set at zero. Means ± SEM for the different experimental groups and the different time points are given; error bars are missing when they fall within the symbol.

In the lungs perfused in the absence of endotoxin administration, no LPS was detected in any of the perfusate samples (< 10 pg/ml, the detection limit of the assay). HlyA admixture was accompanied by some very minor contamination with LPS, calculated to result in < 5 pg/ml perfusate, and accordingly, no endotoxin was detectable in perfusate samples following HlyA admixture with the assay system. More than 70% of intravascularly applied LPS was recovered upon subsequent analysis of perfusate samples over the 180-min perfusion period (data not shown in detail).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

At the low concentrations used in this study, neither LPS alone nor HlyA alone elicited a significant increase in lung microvascular permeability. When employed in a sequence of endotoxin priming and exotoxin challenge, however, a manifold augmentation of Kfc was noted, with concomitant lung edema formation. Owing to the pressure-controlled mode of perfusion, no variation in lung microvascular pressure occurred, and the leakage response must be ascribed to a markedly increased pulmonary capillary permeability. To our knowledge, this is the first report to describe such synergistic action of endotoxin priming and exotoxin challenge in provoking pulmonary vascular leakage.

As anticipated from the preceding studies of LPS-HlyA cooperativity (11, 12), endotoxin administration in the blood-free perfused lungs in the present study did not evoke any substantial thromboxane release. Limited quantities of this vasoconstrictor were liberated in response to the low HlyA dose employed in the study, but excessive TxA₂ concentrations were measured upon sequential application of LPS and HlyA. Such large amounts of thromboxane would have resulted in severe pulmonary vasoconstriction, but this was largely blocked by the presence of the thromboxane-receptor antagonist BM 13.505. As a result, there was only a transient reduction of perfusate flow in the present model of pressure-controlled perfusion. Interestingly, this decrease in vascular cross-sectional area was comparable in LPS- and non-LPS-primed lungs, despite the large differences in thromboxane liberation, thus suggesting that this residual part of the HlyA-induced vasoconstrictor response was not related to thromboxane formation, but to some yet unidentified mechanism.

A large number of experimental studies with intact animals have demonstrated that intravascular endotoxin administration results in lung vascular leakage and protein-rich edema formation (2). Moreover, simultaneous administration of endotoxin and peptide chemotactic factors was reported to produce a cooperative evocation of neutrophil-dependent pulmonary vascular injury in intact rabbits (18). Under the conditions of blood- and plasma-free lung perfusion used in the present study, the administration of LPS did not provoke any increase in Kfc values or pulmonary edema formation, although the lung vasculature responded to the endotoxin challenge with marked TNF generation. This observation is in accord with findings in previous studies of buffer-perfused goat (19) and rabbit lungs (12). It is most probably explained by the absence of plasma-borne endotoxin binders such as the LPS binding protein LBP (20), the lack of LPS-sensitive humoral mediator systems such as the complement cascade, and the absence of recruitment of circulating leukocytes, since increased pulmonary capillary neutrophil adhesion is a feature commonly observed in LPS-challenged intact animals (18, 21). An increasing dosage of HlyA will by itself increase vascular permeability in rabbit lungs, as previously reported (10, 22), but the current HlyA load was chosen approximate the threshold dose, since some (nonsignificant) increase in Kfc and some minor augmentation of lung weight was noted.

Although both LPS and HlyA were ineffective in terms of increasing Kfc when given as sole agents, a dramatic increase in Kfc with concomitant severe lung edema formation occurred upon sequential coapplication of these bacterial agents. The leakage response was clearly not caused by any change in pulmonary capillary filtration pressure. First, as detailed earlier, the HlyA challenge was undertaken in the presence of a thromboxane receptor antagonist in order to suppress overwhelming vasoconstriction, and second, constancy of pulmonary artery pressure was guaranteed by the use of pressure-controlled perfusion. Even a moderate increase in microvascular pressure, which might ensue from changes in the relative distribution of pre- and postcapillary resistance under the foregoing conditions, was ruled out by demonstrating unchanged double occlusion pressures. Since there was no evidence for a significant increase in the capillary filtration area, the manifold augmentation of Kfc values clearly indicates markedly increased hydraulic conductivity, most probably of the lung microvasculature.

The mechanisms underlying this impressive leakage response remain to be determined. Clearly, the leakage response does not rest solely on the addition of subthreshold effects of two toxins; however, "priming" of the lung vasculature by the preceding application of LPS is required, since simultaneous administration of both LPS and HlyA did not reproduce the leakage response. LPS apparently did not facilitate membrane attack by HlyA, as the rapid and partly reversible HlyA-induced release of potassium, assumed to reflect pore formation by the proteinaceous toxin (4), was identical in LPS-pretreated and sterile perfused lungs. Since the lung microvasculature is known to harbor a large number of neutrophils, which have been shown to respond to LPS priming in in vitro studies, it is tempting to speculate that these cells are involved in the leakage response. As suggested by recent morphometric studies, the pool of marginated neutrophils residing in the capillary bed of rabbit lungs even under conditions of prolonged ex vivo buffer perfusion is as large as 1.6 times the respective circulating leukocyte population in this animal (23, 24). Exposure of neutrophils to LPS in vitro primes these cells to respond to different inflammatory stimuli with increased oxygen radical formation (25- 27), protease liberation (28), and lipid mediator release (29). Because neutrophils are target cells of HlyA, which is a potent inductor of the preformed phosphatidylinositol-hydrolysis- related signaling pathway with concomitant oxidative burst, protease liberation, and leukotriene synthesis (6, 7, 9), it is easily imaginable that enhanced tissue damage with loss of endothelial barrier function may result from a burst of neutrophil activation due to synergistic effects of toxins on these cells. Furthermore, in vitro studies with monocytes have repeatedly shown responsiveness of this cell type to priming effects of LPS, with enhanced oxidant and chemoattractant release upon sequential stimulation with different agents (30, 31). Like neutrophils, monocytes are known to remain marginated in the rabbit lung microvasculature even under conditions of buffer perfusion, with a total pool size surpassing by many fold the number of circulating monocytes (23, 24). Cooperative effects of LPS and HlyA on this endothelium-adherent monocyte population may thus easily contribute to the observed priming of vascular leakage. Moreover, the endothelial cells themselves are target cells of both LPS (32) and HlyA (8), and may thus respond in a cooperative fashion to the sequential action of these agents. Furthermore, the recognized secondary induction of TNF might add to the priming efficacy of LPS, as suggested by in vitro studies (33, 34). Notably, the LPS priming for an enhanced leakage response to HlyA has been found to occur in the absence of plasma-derived endotoxin-binding protein such as LBP, soluble CD14, and septins (20, 35). Further studies with such agents should address the question of whether the LPS priming for the exotoxin-elicited leakage response is further enhanced in the presence of circulating endotoxin-binding proteins.

In conclusion, the present study clearly demonstrates that priming with LPS synergizes with E. coli  HlyA challenge to induce increased lung vascular permeability with concomitant severe pulmonary edema formation. Despite unresolved nature of the mechanisms underlying this synergism, this finding further supports the concept that endotoxin priming induces profound cellular and metabolic alterations in the lung vasculature. The excessive response of such primed vasculature to a second exotoxin challenge may be relevant to the pathogenesis of respiratory failure under conditions of sepsis and severe lung infection.