INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Acute lung injury is an important cause of morbidity and mortality in sepsis, but it occurs in a minority of septic patients. The reasons for this variability in clinical response are unknown. Clinical studies of the acute respiratory distress syndrome (ARDS) indicate that the occurrence of acute lung injury (ALI) increases with multiple risk factors (1, 2). Although these studies have not tested for interactions among the wide variety of potential factors predisposing to ALI, the concept that sequential activation of inflammatory processes alters cellular responses to a given stimulus has experimental support from cell-culture and animal studies (3). These cellular responses could translate to alterations in the course and extent of the injury responses of lung and other organs to sepsis. Endotoxins and other microbial products from gram-negative organisms have varying effects on cytokine production and inflammatory-cell trafficking, and may lead to increased or decreased responsiveness to subsequent stimuli. Depending on the type of biologic response, these phenomena have been described as either priming or tolerance. Platelet-activating factor (PAF) and endotoxin, for example, when infused in subinjurious doses, have been shown to produce lung injury on repeat exposure (6). Endotoxin may act as a modulator of cellular responses to inflammatory stimuli, affecting the consequences of a subsequent insult (7, 8).

We hypothesized that the severity of ALI in primates with gram-negative sepsis would be influenced by the effects of recent inflammatory stimuli, which might serve to prime the lung for subsequent injury. We used baboons for testing this hypothesis because the cardiopulmonary responses to gram-negative sepsis in these animals closely resemble those seen in humans (9, 10). A protocol involving live and heat-killed bacteria was developed for use in these animals, rather than one based on endotoxin, to more closely mimic conditions that might be expected to be encountered clinically. We found that prior exposure to a nonlethal dose of heat-killed bacteria increases lung injury from gram-negative sepsis. Our data also suggest that the same exposure attenuates the severity of the hemodynamic compromise accompanying a sudden, lethal infusion of live bacteria.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Animal Preparation

Twenty-three adult male baboons (Papio cynocephalus) weighing 14 to 20 kg were obtained from the Southwest Foundation for Biomedical Research (San Antonio, TX), quarantined for a minimum of 4 wk, and determined to be tuberculosis-free prior to use. Animals were handled in accordance with American Association for the Accreditation of Laboratory Animal Care guidelines, and the protocol for the study was approved by the Duke Institutional Review Board and Institutional Animal Care and Use Committee. After an overnight fast, each animal was sedated with intramuscular ketamine (20 to 25 mg/ kg) and orotracheally intubated. Sedation was maintained with ketamine (3 to 10 mg/kg/h) and diazepam (0.4 to 0.8 mg/kg every 2 h). The animals were mechanically ventilated with a volume-cycled ventilator (Model 7200a; Puritan-Bennett Corp., Kansas City, KS) at an FI_O2 of 0.21, a tidal volume (VT) of 10 to 12 ml/kg, and a positive end-expiratory pressure (PEEP) of 2.5 cm H₂O. The initial rate of 10 to 12 breaths/min was sufficient to maintain a Pa_CO2 of 35 to 45 mm Hg. An indwelling arterial line and a pulmonary artery catheter (size 5 French; Baxter Healthcare Corp., Irvine, CA) were placed via a femoral cutdown for hemodynamic monitoring. Ampicillin (1 g) every 6 h intravenously, and gentamicin (40 mg) and polymyxin (20,000 units) every 4 h intratracheally, were given as prophylaxis against nosocomial pneumonia and secondary bacteremia. Sepsis was induced in 16 animals by infusing 1 to 2 × 10¹⁰ colony forming units (cfu)/kg of Escherichia coli in a volume of 50 ml over a period of 60 min (infusion pump Model 351; Sage Instruments, Cambridge, MA). The time of live E. coli infusion is referred to as time = 0 h. Eight of these septic animals received 10% of this dose (1 × 10⁹ cfu/kg) of bacteria as heat-killed organisms via a 30-min infusion 12 h before receiving the live E. coli (time = 12 h). These animals were considered to be primed. Gentamicin (3 mg/kg intravenously) was administered 60 min after completion of the live E. coli infusion. Fluids were infused at a rate of 100 to 150 ml/kg over the first 24 h to support blood pressure as needed for sepsis-induced hypotension. Typical volume resuscitation during the first 6 to 8 h after infusion of live E. coli was 1,000 ml. Dopamine was used for transient hypotension unresponsive to fluids, but animals were killed by KCl injection when the mean arterial pressure (MAP) could not be maintained above 60 mm Hg or at 36 h after live E. coli infusion for primed or 48 h for unprimed animals. Six normal animals ventilated with air and monitored in identical fashion for 96 h prior to necropsy provided pathology control data. These experiments were interspersed throughout the sepsis experiments. Lung tissue from these animals was used to provide pathology control data for electron microscopic morphometry.

One additional baboon was primed before the onset of live E. coli sepsis as described earlier, and the right middle lobe was biopsied via thoracotomy at 12 h after the priming dose (prior to live E. coli infusion). The biopsy was processed for light microscopy as described subsequently. Physiologic and morphometric data from this experiment were not included in the results or statistical analysis, owing to the possible effects of the thoracotomy on subsequent experimental measurements.

Hemodynamic Monitoring

Heart rate, temperature, arterial blood pressure, pulmonary artery pressures, ventilator pressures and volumes, and fluid balance were recorded every hour. Every 6 h, measurements of cardiac output (by thermodilution; Horizon 2000 monitor; Mennen Medical Inc., Clarence, NY), arterial and mixed venous blood PO₂, PCO₂, and pH (Model 1304 pH/blood gas analyzer; Instrumentation Laboratory, Lexington, MA), and oxygen saturation, oxygen content, and hemoglobin (Model 482 co-oximeter; Instrumentation Laboratory) were obtained.

Preparation of E. coli

E. coli reconstituted from lyophilized specimens (serotype 086a: K61; American Type Culture Collection, Rockville, MD) and stored at 70° C were thawed, streaked onto sheep blood agar plates, and incubated at 37° C for 18 to 24 hours. Plates were checked for purity of culture, and isolated colonies were streaked onto trypticase soy agar slants and incubated for 18 h at 37° C. On the morning of the experiment, organisms were gently washed off the slants with sterile saline, centrifuged, and resuspended in saline. The concentration of bacteria was measured by transmittance at 550 nm and adjusted with saline to give a final dose of 1 to 2 × 10¹⁰ cfu/kg for each baboon, in a volume of 50 ml. This dose of E. coli (1 × 10¹⁰ cfu/kg) is an LD₁₀₀. The number of viable organisms was confirmed by colony counting, using pour plates. For heat-killed E. coli, 1 × 10 ⁹ cfu/kg were heated in a water bath at 65° C for at least 30 min. Efficacy of heat killing was > 99.99%.

Ventilation-Perfusion Measurements

Ventilation-perfusion (A/) distributions were quantified through the multiple inert-gas elimination technique of Wagner and colleagues (11). A dilute solution of six inert gases (sulfur hexafluoride, ethane, cyclopropane, enflurane, diethyl ether, and acetone) in normal saline was infused continuously via a peripheral vein at a rate of 2.5 to 3 ml/ min. Duplicate samples of systemic arterial blood, mixed venous blood, and mixed expired gas were collected at least 40 min after beginning the infusion. Each blood sample was equilibrated with an equal volume of nitrogen in a heated water bath (37° C) for at least 40 min. The equilibrated gas and expired samples were analyzed for sulfur hexafluoride with an electron-capture detector and for the five other gases with a flame-ionization detector (Model 5890 Series II Chromatograph; Hewlett-Packard, Wilmington, DE). The data were stored digitally on a computer for analysis. The blood-gas partition coefficients of the six inert gases were also determined each day. Retention-partition coefficient and excretion-partition coefficient curves were generated, using enforced smoothing and least squares analysis, and the data were transformed into A/ distributions using a 50-compartment model. Shunt was defined as a A/ of less than 0.005 and dead space as a A/ greater than 100. The log (A/) standard deviations for perfusion and ventilation (SDq and SDv) were used to measure dispersion of blood flow and ventilation, respectively (12). A/ measurements were considered satisfactory if the residual sum of squares was 10 or less.

Histology and Lung Morphometry

Lung tissue from all animals was studied with both light and electron microscopy (EM) to determine the distribution and amount of lung injury. At the end of the exposures, the left lung was removed and sampled for wet/dry lung weight determination, and the right lung was inflation-fixed via the endotracheal tube for 15 min at a fixative pressure of 30 cm with 2% glutaraldehyde in 0.85M Na cacodylate buffer (pH 7.4), prior to immersion in fixative. After 7 to 10 d immersion, the lung was transversely sectioned into 1-cm slices. A stratified sample was obtained by random selection of eight 1-cm ³ cubes (three from the upper, one from the middle, and four from the lower lobe). Large bronchovascular structures were not included. For light microscopy, tissue samples were embedded in paraffin, sectioned, and stained with hematoxylin and eosin (H&E). For EM, five tissue cubes were sectioned into smaller cubes 1 to 2 mm on a side, and 10 to 20 of these small cubes from each site were processed together. The tissue was washed in cacodylate buffer and postfixed with 2% osmium tetroxide. After dehydration in graded ethanol solutions, the tissue cubes were immersed in propylene oxide and embedded in Epon resin. Thin sections were cut with a diamond knife, placed on coated, 200-mesh Cu/ Rh grids, and stained with uranyl acetate and lead citrate. Sections were viewed and photographed on a Zeiss 10C transmission electron microscope (Karl Zeiss, Jena, Germany), and then enlarged to ×8,500 on 11- by 14-inch photographic paper for analysis. Micrographs were analyzed morphometrically, using the methods of Weibel and Bolender (13) and Underwood (14), by a morphometrist blinded to study conditions. The open lung biopsy specimen was fixed with glutaraldehyde by manual inflation and processed for light microscopy as outlined earlier.

Five of the randomly selected sites from each animal were analyzed, two each from the upper and lower lobes and one from the middle lobe. The remaining tissue samples were stored at 4° C. Counts were made on 30 photos per site, using a 112-line overlay and yielding 224 points per micrograph. Morphometry data were normalized relative to the surface density of epithelial basement membrane in a standard volume of alveolar tissue, according to methods used in our earlier morphometric analysis of hyperoxic lung injury in primates (15). This normalization is based on the finding that the alveolar basement membrane does not stretch significantly with physiologic levels of inflation, and remains intact even in badly damaged areas of the lung.

Measurements were made for epithelial volume density, including alveolar type 1 and type 2 cells; endothelial volume density; inflammatory-cell volume densities in intravascular, interstitial, and alveolar compartments; and interstitial volume density. The interstitium was analyzed with both matrix and cell counts. Fibroblast volume density data included other noninflammatory interstitial cells such as myofibroblasts and pericytes, which cannot be differentiated from fibroblasts in fragmentary cytoplasmic profiles present in electron micrographs. These data were normalized as detailed previously, and were expressed as the ratio of the volume of tissue in µm³ to the surface area of alveolar basement membrane in µm². Segments of alveolar basement membrane denuded of epithelial covering as a result of severe injury were recorded as bare basement membrane and added to the calculation of total alveolar-basement-membrane surface area. Similarly surface intercepts of capillary basement membrane where endothelium was fragmented or widely displaced were counted as disrupted capillary basement membrane. These data were expressed as percent coverage of the respective basement membrane.

Statistical Analysis

Data are reported as mean ± SEM, except for survival times, which are shown as median values. Statistical analyses were conducted with commercially available software (Statview 4.01; Apple Computer, Cupertino, CA). Values of p are reported in cases in which tests were performed. A value of p  0.05 was considered significant; a trend was defined as p  0.10. Student's t test was used to compare physiologic measurements in the two groups prior to any bacterial exposure, live or heat-killed. Sequential data from physiologic measurements were analyzed over the first 24 h after live E. coli sepsis, using repeated measures analysis of variance (ANOVA). Statistical analysis was not performed after the 24 h time point, owing to four early deaths in the unprimed sepsis group. In addition, data collected at 12 h (before infusion of heat-killed bacteria) from primed animals were compared with 0-h data (before infusion of live bacteria), using Student's t test to analyze the effect of priming on those endpoints. Histologic data from animals primed prior to live bacterial challenge were compared with data from septic animals and data from six normal animals ventilated with air for 96 h, using ANOVA with Fisher's post hoc test.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Live E. coli sepsis in this model produces a shock syndrome characterized by hypotension, metabolic acidosis, and hypoxemia with decreased pulmonary compliance. The eight animals that received live E. coli alone survived from 24 to 48 h (median: 34 h) and died of hypotension refractory to fluids and dopamine. Four animals died at 24 to 26 h and the remaining four at 42 to 48 h. Each group required vasopressors during the experiment to support the blood pressure, as shown in Table 1. Animals that were primed with heat-killed bacteria prior to the live bacterial challenge survived for a median of 36 h after live E. coli sepsis (range: 30 to 36 h), and their hypotension remained responsive to intravenous fluids in all cases except for one animal that required intermittent low-dose dopamine for the final 4 h of the experiment (see Table 1). This animal died 30 h after live bacteria infusion, and was the only primed animal to die before the scheduled end of the experiment. This animal had a refractory respiratory and metabolic acidosis. In both groups, fluid resuscitation was used to maintain a pulmonary capillary wedge pressure (PCWP) of 8 to 12 mm Hg, and the total fluid administration was the same in both groups. The priming dose of E. coli caused transient hypotension in some animals, but this was easily reversible when intravenous fluid was administered after priming, to maintain a stable PCWP. Total fluid support volume ranged from 900 to 1,700 ml during these first 12 h.

Selected physiologic values from the experiments are shown in Table 2. Initial physiologic parameters, before any bacterial exposure (live or heat-killed), were the same in both primed (n = 8) and unprimed (n = 8) animals. After live E. coli infusion, the two groups had similar decreases over time in MAP. The level of support required to maintain the MAP, however, was greater in the unprimed animals, as noted earlier. Both groups of animals were tachycardic throughout the experiment. Systemic valscular resistance adjusted for weight (SVR* kg) decreased from 12 h to 0 h in primed animals (p = 0.01), and tended to be lower than in unprimed animals after infusion of live E. coli (p = 0.06). Cardiac output/kg was higher in primed animals during the first 24 h after infusion of live E. coli (p = 0.02), due to an increase after priming (p = 0.04) that was sustained after the live bacteria were infused (Figure 1A). Although mean pulmonary artery pressure () increased similarly in both groups over the first 12 h of live E. coli sepsis, this increase reached a higher plateau in the eight animals that had been primed (p = 0.006; Figure 1B). Again, this group tended to show an increase in PA at time 0, due to priming (p = 0.07). Pulmonary vascular resistance adjusted for weight (PVR*kg) was not significantly different in the two groups (Figure 1C).

View larger version (14K): [in this window] [in a new window]   Figure 1.   Hemodynamic responses to sepsis. (A) Cardiac-output responses to sepsis. Weight-adjusted cardiac output (CO/kg in L/min/kg) was higher at the onset of live E. coli infusion (t = 0 h) in primed animals, due to the effect of the infusion of heat killed bacteria 12 h earlier (* = 0.04). This increase was sustained throughout the following 24 h (p = 0.02 by repeated measures ANOVA). (B) Increases in mean pulmonary artery pressure ( in mm Hg) in sepsis. Primed animals showed greater pulmonary hypertension than unprimed animals after infusion of live E. coli (p < 0.01 by repeated measures ANOVA through 24 h); *p < 0.05. (C ) Pulmonary vascular resistance adjusted for weight (PVR*kg in dyn*cm*s5*kg) increased similarly in unprimed and primed sepsis; p = NS by repeated measures ANOVA.

Pulmonary mechanics and gas exchange data are shown in Table 3. Both groups had a 25% decrease in pulmonary-system compliance and developed similar changes in A/ relationships with increases in shunt, dead space, and SDv and SDq. SDv and SDq increased after priming (p = 0.05 and p = 0.01, respectively), but were not different for the two groups after live E. coli were infused. Significant hypoxemia developed in both groups. Although two of the primed septic animals required supplemental oxygen for a Pa_O2 less than 50 mm Hg, the difference in alveolar-arterial (A-a) oxygen ratios between the two groups did not reach statistical significance. Lung edema was increased in both the primed and unprimed sepsis groups (wet/dry lung weight ratios: 6.065 ± 0.558 and 5.778 ± 0.346, respectively, p = NS). The reference value for air-ventilated wet/dry lung weight in control baboons was 4.777 ± 0.188. 

Despite aggressive volume resuscitation in both groups, urine output decreased more over the first 24 h after live E. coli infusion in unprimed than in primed animals (p = 0.03; Figure 2). The arterial blood gas measurements also reflected an ongoing metabolic acidosis that was less pronounced in the primed animals with a lesser decrease in serum [HCO₃] (p = 0.02; Figure 3). Arterial pH values were not different in the two groups of animals, reflecting ventilator management during the experiments.

View larger version (28K): [in this window] [in a new window]   Figure 2.   Urine output in unprimed versus primed animals with E. coli sepsis. Unprimed animals showed decreased urine flow as compared with primed animals; p = 0.03 by repeated measures ANOVA; n = 8 at each time point except 36 h, at which n = 7 in primed sepsis. Data for unprimed sepsis are shown through 24 h only, due to early deaths of four animals.

View larger version (20K): [in this window] [in a new window]   Figure 3.   Change in serum [HCO3] during sepsis. Serum [HCO3] decreased more in unprimed than in primed septic animals after live E. coli infusion; p = 0.02 by ANOVA; n = 8 at each time point shown except 36 h, at which n = 7 in primed sepsis. Data for unprimed sepsis are shown through 24 h only, due to early deaths of four animals.

Light microscopic analysis of lung tissue from animals with E. coli sepsis showed thickened alveolar septae with increased cellularity, primarily due to increased numbers of neutrophils in the intravascular compartment (Figure 4B). Some intraalveolar edema was present. In contrast, animals that were primed with heat-killed bacteria prior to live E. coli sepsis developed thickened alveolar septae with both increased neutrophils and a prominent increase in mononuclear and interstitial cells (Figure 4D). Alveolar edema was also present in this group, and many inflammatory cells were present in the alveolar spaces. Open-lung biopsy, done in one animal at 12 h after priming, showed normal lung architecture by light microscopy. There was no increase in neutrophils in alveolar septae, and no cells or edema in the alveolar spaces (Figure 4C).

View larger version (55K): [in this window] [in a new window]   View larger version (76K): [in this window] [in a new window]   View larger version (60K): [in this window] [in a new window]   View larger version (95K): [in this window] [in a new window]   Figure 4.   Light microscopy of the baboon lung in sepsis. (A) Lung from a normal baboon ventilated with air for 96 h. (B) Lung from an unprimed septic animal, showing inflammatory-cell infiltrate within the vascular space and interstitial edema. (C ) Open lung biopsy obtained 12 h after priming with heat-killed E. coli shows normal lung architecture and absence of inflammation. (D) Animals primed before E. coli sepsis show extensive mixed inflammatory-cell infiltrate in the alveolar spaces and thickened interstitium with increased cellularity. Original magnification: approximately ×30, H&E.

On qualitative evaluation by electron microscopy, lung tissue from unprimed septic animals (Figure 5A) showed a predominance of polymorphonuclear neutrophils (PMN) in the capillaries with normal-appearing capillary endothelium. The interstitium was thickened as a result of edema. Type 1 alveolar epithelial cells had areas of swelling and blebbing along the cell membrane, but the epithelial surface remained largely intact. In contrast, lung tissue from animals that were primed prior to live E. coli sepsis (Figure 5B) had damage to the capillary endothelium, with vacuolization and ruffling of the cell membranes. Capillaries were packed with adherent monocytic inflammatory cells as well as PMN. The alveolar interstitium was thickened due to both edema and increased matrix elements, such as collagen fibers. Interstitial fibroblasts and other noninflammatory cells were also increased. Type 1 epithelium was damaged more severely in those animals with diffuse swelling and vacuolization. In some places, disruption of the epithelial layer left denuded basement membrane. The alveolar spaces in primed animals had a mixed inflammatory infiltrate consisting of PMN and alveolar macrophages.

View larger version (103K): [in this window] [in a new window]   View larger version (145K): [in this window] [in a new window]   Figure 5.   Electron micrographs of lung in unprimed and primed septic animals. (A) Polymorphonuclear neutrophils (PMN) are prominent in the vascular space of unprimed septic animals. (B) Primed septic animals have an inflammatory infiltrate in alveolar spaces, consisting of both PMN and macrophages. The interstitium is thickened, with an increase in fibroblasts, and both endothelium and type I epithelium are swollen. Original magnification: approximately ×4,700.

The pathologic changes in the lungs of septic and primed septic animals were compared through EM morphometry with normal baboon lungs, using ANOVA. This analysis showed that lung injury involving endothelium, interstitium, and epithelium was increased in the animals that were primed before infusion of live bacteria. Normalized total epithelial cell volume density was increased in primed animals as compared with unprimed septic animals (p = 0.01). This increase was due to a 50% increase above ventilator control values in the normalized volume density of the type 1 epithelium in primed septic animals (p = 0.003). There was no significant change in the volume density of type 2 cells (Figure 6), and there was no difference in the percent bare basement membrane. Endothelial-cell swelling was not a feature of lung injury in unprimed animals, but was seen in primed animals after E. coli sepsis. Normalized endothelial volume density was increased at 0.365 ± 0.015 µm³/µm² in primed septic animals, as compared with 0.309 ± 0.024 µm³/µm² in air-ventilated pathology control animals, and 0.297 ± 0.016 µm³/µm² in unprimed septic animals (p = 0.03). In the primed animals there was a change in the composition of the interstitial compartment, with an increase in the normalized volume density of fibroblasts and other noninflammatory interstitial cells over that in both air-ventilated controls and unprimed septic animals (p = 0.002; Figure 7).

View larger version (25K): [in this window] [in a new window]   Figure 6.   Normalized pulmonary epithelial volume density in sepsis. Total epithelial volume density increased in animals that were primed prior to live E. coli sepsis, due primarily to an increase in normalized type 1-cell volume density. Data analyzed by ANOVA. *p < 0.05 versus air; p < 0.05 versus sepsis.

View larger version (24K): [in this window] [in a new window]   Figure 7.   Normalized interstitial volume density in sepsis. Fibroblast volume density increased in primed but not unprimed septic animals. Data analyzed by ANOVA; *p < 0.05 versus air; p < 0.05 versus sepsis.

In primed septic animals there was also a prominent increase in the inflammatory response in the alveolar space, with a threefold increase in normalized alvolar macrophage volume density (p = 0.007; Figure 8A). Neutrophils were present in the alveolar spaces in six of the eight primed versus only two of the eight unprimed septic animals although the difference in normalized volume density of alveolar PMN (0.026 ± 0.010 µm³/µm² versus 0.012 ± 0.009 µm³/µm²) did not reach statistical significance (Figure 8B).

View larger version (17K): [in this window] [in a new window]   Figure 8.   Increase in pulmonary inflammatory cells in sepsis. (A) Blood monocytes were increased in the pulmonary vasculature in both unprimed and primed septic animals, but alveolar macrophages were increased only in the group primed before sepsis. (B) Polymorphonuclear leukocytes were increased in both sepsis groups. The increase in alveolar PMN in the primed as compared with the unprimed group was not statistically significant. Data analyzed by ANOVA; *p < 0.05 versus air; p < 0.05 versus sepsis.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have shown that nonlethal exposure to heat-killed bacteria modifies the pathologic and hemodynamic consequences of sepsis with live E. coli in baboons. Animals that were primed before the onset of sepsis developed more lung injury as measured by qualitative and quantitative pathology. Prior infusion of heat-killed bacteria also attenuated the hypotensive response to sepsis and resulted in less severe renal dysfunction and metabolic acidosis after a lethal challenge with E. coli. Although survival was not an end point in these experiments, amelioration of the circulatory response to gram-negative sepsis was consistent with the survival benefit of priming reported by others (5, 18, 19). All eight unprimed septic animals in our study required dopamine in addition to volume resuscitation, compared with only one of the primed septic animals. Thus, the increased renal failure in the unprimed animals was probably related to decreased organ perfusion in the setting of refractory hypotension.

Use of endotoxin as an immunomodulating agent has been referred to as tolerance or priming, depending on its effects on the end points of the study (3, 16). We have used the term "priming" to describe the findings in the animals in our study, though we encountered both increased and decreased injury responses in different organ systems. The classical concept of endotoxin tolerance is determined by a decrease in lethality seen with serial exposure to endotoxin. Tolerance can be divided into an early phase, in which changes in macrophage function play a predominant role, and a late phase, which is antibody-mediated and polysaccharide-specific (24). Development of a tolerant state requires exposure to low doses of bacterial products, usually endotoxin or one of its lipid-A derivatives (16). Decreased shock and mortality after early induction of tolerance have been described in small animals (5) and may be due to decreased cytokine release or to decreased responsiveness to mediators (17, 21). At the cellular level, endotoxin tolerance has been associated with a state in which monocytic cells previously exposed to small amounts of endotoxin display decreased production of tumor necrosis factor (TNF) in response to lipopolysaccharide (LPS) (16). Although decreased TNF production is characteristic of tolerant cells, there is evidence that this response is dose specific, and that rather than simply downregulating cytokine production, endotoxin can reprogram macrophage responses to a variety of cytokines and signaling pathways (7, 20).

Despite the hemodynamic tolerance produced by pretreatment of baboons with heat-killed E. coli, we found that this effect did not extend to the lung. Pulmonary artery pressure and cardiac output were increased in the primed as compared with the unprimed septic animals. Similar changes in the pulmonary vascular response to endotoxin and other bacterial products have been described in other models. For example, repeated injections of endotoxin over a 10-wk period in sheep produce pulmonary hypertension (25), and a single previous exposure to endotoxin enhances the vasoconstrictor response to staphylococcal -toxin in the isolated perfused rabbit lung (22). The increase in pulmonary artery pressures during sepsis in the primed baboons in our study is consistent with these observations.

The actions of cytokines and inflammatory cells on pulmonary vasculature and endothelial-cell permeability lead to alterations in gas exchange and lung compliance, to pulmonary edema, and to accumulation of inflammatory cells within the lung (10, 26). Acute sepsis in animal models causes epithelial damage and intravascular neutrophil accumulation, but typically does not cause ultrastructural changes in capillary endothelium in the lung (10, 26). In our unprimed animals this was also the case. Primed septic animals, however, had significant swelling of endothelial cells. Qualitatively, these cells showed vacuolization and swelling of the cytoplasm and ruffling of the cell membrane, suggesting activation of and/or injury to the cells. Similar changes have been described at 24 h in ultrastructural analysis of lung injury in sepsis in humans (27).

Lung injury from sepsis in primed animals differed from that in unprimed animals in morphometric parameters other than endothelial-cell damage. Animals that were primed prior to E. coli sepsis developed an inflammatory response in the alveolar spaces that was not present in unprimed animals. The changes we measured in the lung interstitium, including an increased volume of fibroblasts and other noninflammatory interstitial cells, suggest the presence of early fibroproliferative responses in addition to inflammatory processes. This is an important component of ARDS in humans (27). Since median survival time was similar in the two groups of baboons in our study, these pathologic changes suggest qualitative differences in lung injury.

After 12 h, the effects on the lung of priming with heat-killed E. coli appeared to be minimal. At 12 h, the lung appeared normal by light microscopy, and there were no gross changes in number or distribution of inflammatory cells. The lack of evidence of neutrophil sequestration makes it unlikely that priming causes increased lung injury due to degranulation of activated neutrophils sequestered in the lung, as has been implicated in some injuries (e.g., the generalized Shwartzman reaction) (28). The only significant change in pulmonary physiology measured in our study at 12 h after infusion of heat-killed E. coli was an increase in dispersion of ventilation and perfusion comparable to that found with mechanical ventilation alone (29).

The lack of an inflammatory cell infiltrate after priming implicates functional rather than numerical cellular changes induced by the killed bacteria in the development of lung injury in the primed animals. In human volunteers, intravenous administration of endotoxin does not affect cell type or number in bronchoalveolar lavage fluid (BALF); however, recovered macrophages show enhanced secretion of TNF, interleukin-1, and prostaglandin E₂ with subsequent exposure to LPS in vitro (23). Alternatively, an increase in circulating mediators or local production of cytokines from the initial insult could interact with the second, live infusion of E. coli. PAF, for example, which can produce many of the physiologic features of sepsis, acts synergistically with LPS to cause lung injury (6). Alterations in neutrophil function may also play a role in our experiments, via changes in conformation, chemotaxis, adherence, and superoxide production that occur when neutrophils are exposed to small amounts of LPS (30, 31). In naive animals, inflammatory cell influx into the lung is a prominent feature of sepsis-induced ALI and is primarily an intravascular event (26), as we found in unprimed animals. In contrast, the distribution of inflammatory cells into the interstitial and alveolar compartments was more promiment in primed animals. The priming dose of bacteria may have influenced adhesion and migration events through changes in the expression of cell-adhesion molecules on leukocytes and/or pulmonary endothelial cells. The consequence of increased inflammatory cell infiltration into the alveolar spaces may be increased damage to the epithelium.

The responses of noninflammatory cell types such as endothelial cells and fibroblasts may also be modified by the responses to small quantities of bacterial products in the lung. Interleukin-1 and interferon- increase superoxide release by endothelial cells in vitro (32), and cytokines released by activated inflammatory cells alter signaling between the vascular endothelium and its surrounding connective tissue structure (33). Fibroblasts can also interact with endotoxin-activated monocytes, secreting their own immunomodulating substances (34). It is possible that prior exposure to bacteria predisposes the lung to specific injury patterns by such mechanisms.

Although our experiments were designed primarily to measure lung injury responses in sepsis, our data also indicate that priming attenuated the shock associated with lethal E. coli sepsis. Discrepancies in organ responses to intravascular exposure to heat-killed bacteria could be explained in several ways. For instance, differences in presentation of endotoxin and other circulating bacterial products by CD14 or other binding proteins could contribute to variations in injury between tissues. In vitro, LPS pretreatment of macrophages causes differential regulation of TNF and nitric oxide production that varies with the dose of LPS used for priming (8, 20, 35). Differential responsiveness to endotoxin and other immune stimulants by different macrophage populations has also been described (36), and this could contribute to the divergent injury patterns we found in lung and systemic organs.

In summary, the baboon responds to sublethal exposure to killed whole bacteria in a manner that modulates organ damage subsequent to a lethal E. coli infusion. The organ damage may be either attenuated or increased, depending on the tissues involved. Thus the terms "tolerance" and "priming" can be applied to different organ responses to endotoxin in the same primate, in accord with the heterogeneity of the clinical sepsis syndrome in humans. The mechanisms responsible for the enhanced lung injury we found in this study should be further elucidated, because they may provide important clues to the pathophysiology of human ARDS. A variety of immunomodulators, including modified endotoxin preparations, are being considered for use in patients at risk for sepsis (16). Our findings suggest that additional data on tissue-specific responses to such immunomodulators will be important in determining how and when to administer these preparations therapeutically.