Mechanical Ventilation Affects Lung Function and Cytokine Production in an Experimental Model of Endotoxemia

Thirty pathogen-free New Zealand rabbits weighing 2.56 ± 0.02 kg (Charles River Laboratories, L’Abresle, France) were studied. The protocol conformed to the guidelines in the Guide for the Care and Use of Laboratory Animals  12and was approved by the animal research committee (Comité d’Ethique Animale de l’Université de la Méditerranée, Université de la Méditerranée, Marseille, France).

The animal preparation has been described elsewhere.13Briefly, animals were anesthetized with ethyl carbamate (initial dose of 1 g/kg intravenous, followed by hourly injections of 0.2 g/kg). Then, they were tracheotomized, and a catheter was placed into a femoral artery for measurements of arterial blood pressure and heart rate, as well as for blood gas sampling and further analyses. Sterile conditions were maintained throughout the experiment. Body temperature was continuously monitored and maintained between 37.5° and 38.5°C by a thermostatically controlled heating pad. All animals were connected to a mixing chamber that delivered a warm and humidified gas mixture containing 25–30% oxygen in nitrogen.

The MV animals that received lipopolysaccharide (MV-LPS) or saline (MV-saline) were paralyzed with 0.2 mg/kg vecuronium bromide followed by hourly injections of 0.2 mg/kg and were connected to a constant-volume ventilator pump that delivered airflow in a sinusoidal wave form (Beaudoin, Paris, France). V^Twas set at 10 ml/kg. Only six respiratory rates were available with this device: 14, 28, 32, 40, 64, and 80 cycles/min. A respiratory rate of 32 breaths/min was the most compatible with a physiologic arterial carbon dioxide tension (Paco²), and that did not produce hypercapnia (hypercapnia was unacceptable because of its possible effect on cytokine production14,15). This setting was also used in our previously published work. The inspiratory:expiratory ratio was 50%, and no inspiratory plateau was applied.

Continuous infusion of saline (initially set at 5 ml · kg⁻¹· h⁻¹) was administered via  a ear vein catheter. When mean arterial pressure decreased below 65 mmHg, the infusion flow rate was increased progressively to a maximum of 20 ml · kg⁻¹· h⁻¹.

Systemic arterial blood pressure was measured continuously with a calibrated pressure transducer and recorded on a polygraph. Arterial oxygen tension (Pao²), Paco², and arterial p  H (p  Ha) were measured hourly from 0.3-ml blood samples (ABL 330; Radiometer, Copenhagen, Denmark). The fraction of inspired oxygen (Fio²) was measured in the inspiratory tube coming from the mixing chamber with a Beckman OM11 analyzer (Fullerton, CA).

Airflow was measured with a Fleisch pneumotachograph attached to the tracheal cannula and connected to a differential electromanometer. V^Twas obtained through electronic integration of the flow signal. Tracheal pressure (Ptr) was measured from a lateral outlet of the tracheal cannula. Maximum tracheal pressure (peak inspiratory pressure) was measured, and V^T, airflow, Ptr, and arterial pressure were continuously recorded on a Gould ES1000 polygraph. Respiratory rate and minute ventilation were calculated from V^Trecordings.

The total resistance of the respiratory system was calculated, according to the method of Amdur and Mead,16from the changes in Ptr and flow rate measured between two successive isovolumetric levels (1/2 V^T).

A pressure–volume curve was performed at reference time and at the end of the experiment after a short inhalation of pure oxygen, using a 100-ml syringe filled with pure oxygen to successively fill the lungs step by step (5–10 ml) until Ptr reached 30 cm H²O and emptying them until Ptr had returned to the atmospheric pressure. For each volume level, Ptr values were determined from the plateau value measured 2 s after the initial peak. To do this in the spontaneously breathing group, transient muscle relaxation was induced using succinylcholine chloride (1 mg/kg intravenous). Static compliance was measured from the static volume–pressure relation.

Animals were randomized into three groups: In one group, animals were allowed to breathe spontaneously and received intravenous lipopolysaccharide after an initial 2-h steady state period (SB-LPS; n = 10), one group was mechanically ventilated and treated with 0.9% intravenous NaCl (MV-saline; n = 10), and one group was mechanically ventilated and treated with intravenous lipopolysaccharide (MV-LPS; n = 10). The intravenous injection was of 5 μg/kg lipopolysaccharide (0.1 ml/kg of a solution of 5 μg/0.1 ml lipopolysaccharide, Escherichia coli  serotype O111:B4; Sigma Chemical, St. Louis, MO) dissolved in 0.9% NaCl. This dose was chosen from data from literature17showing that intravenous administration of a single bolus of 5 μg/kg lipopolysaccharide resulted in a systemic cytokine response but not lung injury in the absence of an additional stimulation. In initial experiments, we tested several doses of lipopolysaccharide and found that at substantially higher doses of lipopolysaccharide, the animals were not hemodynamically stable. All animals were monitored continuously for 4 h after the lipopolysaccharide injections and then euthanized.

After a midline sternotomy, a lethal cardiac injection was administered, and the lungs were clamped at the hilum at end inspiration. The left lung was lavaged with 6 ml/kg sterile saline. The right lung was filled with 10% buffered formalin at a transpulmonary pressure of 15 cm H²O via  a 3-French catheter.

Cell analysis was performed on 2 ml fluid recovered from lavage. The remaining fluid was spun to pellet the cells, and the supernatant was stored at −80°C for further cytokine determination. The pelleted cells were homogenized in Trizol reagent (Gibco-BRL; Life Technologies, Eragny, France) and stored at −80°C until tumor necrosis factor α (TNF-α) messenger RNA (mRNA) analysis was performed.

Concentrations of TNF-α, interleukin 8 (IL-8), and growth-related oncogene α (GRO-α)/CXCL1 were measured in plasma and bronchoalveolar lavage (BAL) supernatant using rabbit-specific immunoassay as previously described.18,19For TNF-α, monoclonal anti-rabbit TNF antibodies were used. IL-8 was measured by immunoassay using two different murine monoclonal antibodies raised against recombinant rabbit IL-8. This assay was modified from our previously published work.13Assay lower limits were 0.1, 0.1, and 0.01 ng/ml for GRO-α, TNF-α, and IL-8, respectively.

Real-time polymerase chain reaction was performed using oligonucleotide polymerase chain reaction primers synthesized according to the published complementary DNA sequences of rabbit TNF-α and 18S ribosomal RNA.

In fixed lungs, the cranial, middle, and caudal lobes were sampled from 3-mm horizontal sections including the whole circumference of each lobe. All slides were examined by two pathologists blinded to the group assignment of the lungs. The following histologic criteria were assessed: edema (alveolar and interstitial congestion), hemorrhage, interstitial cell infiltration, and intraalveolar cell accumulation. Their respective intensities were semiquantitatively scored as 0 (absent), + (mild), ++ (moderate), or +++ (severe). Intravascular aggregation of neutrophils on the vessel walls and the existence of thrombi were recorded. When disagreements occurred between the observers, the pathologists were asked to rescore the sections and to give a final consensus result. The lungs of six additional control animals that had not been ventilated and were killed after a lethal injection of anesthetic were also analyzed for comparison. For each criterion, a final score was calculated from the unique final decision of the two pathologists by dividing the sum of individual results by the number of animals (minimum, 0; maximum, 3).

All data were expressed as the mean ± SEM. The Sigmastat 2.03 program was used for statistics (Sigma Company, Erkrath, Germany). Data were tested for normal distribution with the Kolmogorov–Smirnov test. The differences in the values among the three study groups were compared using parametric one-way analysis of variance for normally distributed data. When the data were not normally distributed (cytokine concentrations, neutrophils), the medians were compared using the nonparametric Kruskal–Wallis test. When the difference between groups was significant with the above tests, the group or groups that differed from the others were identified using a multiple comparison procedure such as the Newman–Keuls test for post hoc  analysis of variance and the Dunn method for pairwise multiple comparisons.

For temporally repeated data, changes over time were assessed in each group using one-way repeated-measures analysis of variance for normally distributed data or Friedman repeated measures on ranks when data were not normally distributed (cytokine concentrations). Differences between the times were identified using either the Student–Newman–Keuls comparison test after repeated-measures analysis of variance or the Tukey multiple comparison test after the nonparametric Friedman test.

When a parameter was measured twice only (i.e. , pulmonary mechanics), comparisons over time were assessed by a paired t  test. A P  value of less than 0.05 was used to determine statistical significance. Agreement between the two pathologists was assessed by the κ coefficient calculation (κ= observed agreement − expected agreement/1 − expected agreement). A κ statistic of less than 0.4 is considered to show a weak correlation, whereas values from 0.4 to 1 are interpreted as moderate, good, and very good.

In the three groups, arterial blood gases and p  Ha remained stable throughout the 2-h period preceding injection of lipopolysaccharide or saline (fig. 1). Table 1shows the mean values of physiologic variables measured at baseline, i.e. , just before injection of lipopolysaccharide or saline. There was no significant difference between the two MV groups (MV-LPS and MV-saline) for any of the baseline parameters. Spontaneously breathing animals had higher Paco², lower V^T, and higher respiratory rate values at baseline than MV animals. However, their minute ventilation, resistance of the respiratory system, and static compliance were similar to those measured in MV animals.

In the three groups, the mean arterial pressure decreased significantly at the end of the experiment, but despite the fact that the MV-saline group tended to have higher mean arterial pressure values, no significant difference between groups was observed at any time during the study. Heart rate did not differ among any of the groups at any time. Therefore, there was no evidence of a significant deleterious effect of our dose of lipopolysaccharide on cardiovascular function. At the end of the experiment, values were 72 ± 8, 73 ± 7, and 66 ± 8 mmHg in the SB-LPS, MV-saline, and MV-LPS groups, respectively.

No modification of arterial blood gases and p  Ha (fig. 1) were observed after saline injection in the MV-saline group. In spontaneously breathing animals (SB-LPS group), lipopolysaccharide injection did not induce a Pao²change, but hypocapnia developed in the first hour after injection, because of hyperventilation. In this group, V^Tincreased (up to 8.9 ± 0.3 ml/kg at the end; P < 0.001 vs.  baseline) with no change in respiratory rate, resulting in increased minute ventilation that reached 533 ± 50 ml · min⁻¹· kg⁻¹at the end of the experiment (P < 0.001 vs.  baseline and vs.  mechanically ventilated groups).

In the MV-LPS group, marked hypoxemia developed 1 h after lipopolysaccharide injection, followed by a modest partial recovery of Pao²:Fio². This Pao²:Fio²decrease in the MV-LPS group was accompanied by an increase in Paco²and a decrease in p  Ha. Despite the decrease in p  Ha in this group, the difference in p  H with the other groups did not reach significance at any time. In contrast, Paco²remained significantly increased above baseline at the end of the experiment (P < 0.05), although there was a trend toward further decline after the early lipopolysaccharide-induced increase. Static compliance was significantly lower at the end of the experiment than at baseline (decreasing to 0.87 ml · cm H²O⁻¹· kg⁻¹; P < 0.01 vs.  baseline), whereas it remained stable in the two other groups. In addition, in MV-LPS rabbits, peak inspiratory pressure increased up to 14.5 ± 1.1 cm H²O (P = 0.08). The resistance of the respiratory system did not change significantly in any of the groups.

All groups had similar baseline concentrations of blood leukocytes. In the two groups challenged with lipopolysaccharide, the blood leukocyte count, particularly the neutrophil count, decreased significantly and with the same magnitude. The peripheral blood leukocyte count did not change in the MV-saline group (fig. 2A).

Lipopolysaccharide injection induced an increase in the BAL total leukocyte count of both ventilated and nonventilated animals. The BAL leukocyte counts were 525 ± 51 and 457 ± 38 × 10³/ml in MV-LPS and SB-LPS groups, respectively (not significant), versus  316 ± 27 × 10³/ml in the MV-saline group (P < 0.01 vs.  MV-LPS and SB-LPS). In endotoxemic animals, this increase was due mainly to an increase in neutrophils and macrophages. Because of the significant increase in neutrophils in the BAL of MV-LPS animals in comparison with the two other groups (fig. 2B), the percentage of macrophages was lower in the MV-LPS group than in the SB-LPS and MV-saline animals (80.8, 88.6, and 97.2%, respectively; P < 0.05).

All groups had similar baseline plasma concentrations of cytokines (fig. 3). The changes in plasma concentration of cytokines due to lipopolysaccharide injection were similar at the different times (baseline and 1 and 4 h after lipopolysaccharide injection) in the MV-LPS and SB-LPS groups. The plasma IL-8 concentration markedly increased after lipopolysaccharide injection, and high concentrations were detectable at 4 h. TNF-α increased early (1 h) and then decreased to relatively low concentrations at 4 h. GRO-α increased slightly in the groups challenged with lipopolysaccharide but remained at concentrations less than 1 ng/ml of plasma. Cytokine concentrations remained low in the MV-saline group at all times.

The BAL concentrations of IL-8 and GRO-α were both significantly higher in MV-LPS group than in the two other groups (fig. 4). TNF-α was not detected in the BAL of SB-LPS animals but was detected in some MV-saline and MV-LPS animals; however, the median value was at zero in all groups. Because these negative results differed from what we expected, we performed mRNA analysis in all of the animals. Animals treated with MV only (MV-saline group) had higher expression of TNF-α mRNA than spontaneously breathing endotoxemic animals, and there was nonsignificant further increase with the combination of lipopolysaccharide and MV (9-fold and 11-fold, respectively). The total protein content of the BAL was not different among the three groups.

In the histologic evaluation, disagreement between the two pathologists was never noted for more than two of the histologic criteria for each animal. The disagreements concerned the presence or absence of interstitial edema. For this parameter, the κ coefficient was 0.37 in the SB-LPS group, 0.58 in the MV-LPS group, and 0.8 in the MV-saline group. There was 100% agreement in all cases after reevaluation of the coded samples.

Light microscopic lung examination mainly showed interstitial cell infiltrates in the MV-LPS, SB-LPS, and MV-saline groups. These changes were least noticeable in the MV-saline group and most apparent in the MV-LPS group. In this latter group, an increase in the alveolar wall thickness was noted. Figure 5Ashows four representative sections of control, MV-saline, MV-LPS, and SB-LPS groups. In some sections from MV-LPS and MV-saline animals, leukocytes were present in the alveolar lumen. Neutrophil adherence to extraalveolar vessel walls was observed in 50% of the MV-LPS animals, in one animal of the SB-LPS group, and also in one animal of the MV-saline group but in none of the controls (fig. 5B, top). Congestion was present in each of the animal groups and was mainly located in perivascular exudates (fig. 5B, bottom). No thrombi or hyaline membranes were observed. Perivascular hemorrhages were rarely observed, and only in MV animals. Pathologic scores reflected these changes and are shown in figure 6.

The major goal of this study was to determine whether an MV strategy commonly used in major surgical procedures enhances the effects of transient endotoxemia on the lungs. The results show that injection of 5 μg/kg endotoxemia by itself caused only minor changes in the lungs of spontaneously breathing animals, whereas endotoxemia had a marked effect in ventilated lungs in vivo , reflected by impaired function and increases in lung inflammatory cells and chemotactic cytokines. These results are important because they suggest that conventional MV is a risk factor for inflammatory lung injury when endotoxemia occurs, as during or after major surgery. These observations add to a growing body of experimental work suggesting that PPMV can be harmful even when moderate pressures and volumes are applied to the lungs. The novelty of our approach consists of testing a ventilatory strategy that reflects current practice in healthy humans, i.e. , using moderate volume and moderate pressure levels in previously healthy lungs.

Positive-pressure mechanical ventilation using noninjurious V^Tand ZEEP does not usually induce significant lung injury per se  in animals with healthy lungs, in contrast to what is observed when lungs are injured.20Some experimental observations suggest that PPMV may also sensitize the lung to a secondary injury.13PPMV is also associated with qualitative or quantitative changes in surfactant.21However, whether a ventilatory strategy using 10 ml/kg V^Tin ZEEP could favor lung injury in healthy lungs in the presence of a further inflammatory stimulation had not been shown before this study. The MV protocol that we used was based on recent reports in clinical practice.22–24In humans, Wrigge et al.  24found that a potentially injurious ventilatory strategy using 10–15 ml/kg V^Tand ZEEP did not enhance systemic or lung inflammatory responses after 3 h in the operating room. Recently, Altemeier et al.  10found that a moderately high V^T(V^T= 15 ml/kg) had a synergistic effect with high concentrations of lipopolysaccharide (5 mg/kg), significantly higher than we used in this study. In contrast with that study, our experimental approach using a V^Tof 10 ml/kg and lipopolysaccharide at 5 μg/ml more closely simulates the MV strategy commonly used in the operating room and should more closely reflect the concentration range of circulating lipopolysaccharide associated with the perioperative periods of major surgery. At high doses of lipopolysaccharide, plasma endotoxin concentration reaches levels thousands-fold higher than that documented in humans during the operative period or during septic shock.25–28 

Depending on the animal species, experimental conditions, and the dose of lipopolysaccharide, the lung response to endotoxemia ranges from leukocyte accumulation in lung capillaries without changes in vascular permeability to severe lung damage.29Our protocol of intravenous administration of 5 μg/kg was based on the study of Schimke et al. ,17which showed no marked lung injury at this dose of lipopolysaccharide in the absence of an additional stimulus. As expected, we observed only minor changes in spontaneously breathing endotoxemic animals, according to previous reports with doses of lipopolysaccharide varying from 0.1 to 3 mg/kg.17,30–34Because a 2-h period has been sufficient to observe ventilation-induced lung inflammation ex vivo  35and in vivo ,36we chose a 2-h delay before lipopolysaccharide injection to provide a chance for PPMV to prime the lung for an accentuated response to the second stimulus.

We measured IL-8 and TNF-α because of published data suggesting synergy between mechanical stretch and sepsis.9,37,38In addition, the CXC chemokine, GRO-α, appeared interesting to investigate acute inflammation in rabbits.19,39Named CXCL1 in the novel nomenclature, GRO-α is produced by humans and mouse cells and is homologous in rats to the cytokine-induced neutrophil chemoattractant protein. Like other α chemokines, GRO-α is a potent neutrophil attractant and activator that may play an important role in the pathogeneses of ventilator-induced lung injury.19,39 

Neutrophils were moderately increased in the BAL of spontaneously breathing animals, in accord with previous reports showing that in the absence of another lung stimulus, neutrophils remain sequestered in capillaries.40,41The number of sequestered cells is both dose and time dependent33; however, transmigration to the alveolar surface requires chemotactic factors32such as IL-8 and GRO-α.

We report lower concentration of cytokines in the BAL fluid than in plasma on crude measurement. However, performing BAL causes dilution of in situ  molecules of as much as 50- to 100-fold, depending on the method used. Here, we instilled approximately 15–20 ml saline in a left lung of approximately 50 ml total lung capacity in the open thorax. Because the BAL was standardized, we believe the dilution factor was similar in all animals, so comparison should be accurate between groups assessed with the same methodology. It may not be the same when cytokine concentrations from different studies are compared: Lavage procedure and assays for cytokine measurement may differ, generating discrepancies between measured concentration in different studies.13 

In vitro , the potentiating effect of a mechanical stretch and lipopolysaccharide on IL-8 production by lung cells has been reported.37,38,In vivo , potentiation between injurious ventilation and high-dose lipopolysaccharide has been shown for TNF-α, IL-8, and GRO-α.10In accord, we found that both IL-8 and GRO-α were up-regulated in the alveolar compartment of rabbits treated with MV and injected with lipopolysaccharide. These mediators are likely to form the alveolar–capillary chemokine gradient necessary for the transmigration of neutrophils from the vascular bed to the lower airways. In contrast, TNF-α did not increase significantly in the lungs of any of the study groups, and only a few animals receiving MV with or without lipopolysaccharide had detectable concentrations. This differs from what we and others have reported.10The kinetics of release of TNF-α, with elevation at the early stage of exposure to PPMV and then decline in the late stage,42does not necessarily explain these discrepancies. Interestingly, we found that TNF-α mRNA production was markedly increased in the RNA samples of the alveolar cell pellets of the ventilated animals. Because most TNF-α is secreted after transcription, this suggests that TNF-α was present in the lower airways of the ventilated animals. However, the role of TNF-α in ventilation-induced inflammation remains unclear and is debated in the literature.43–45In some animal models of ventilator-induced lung injury, alveolar concentrations of TNF-α are increased,9whereas in other models they are not.46,47Even in acute respiratory distress syndrome patients, TNF-α is present in BAL fluids but is largely inactivated.48,49 

A striking difference in the response to lipopolysaccharide between ventilated and spontaneously breathing animals was the early alteration in gas exchanges at 1 h. This was not related to significant hemodynamic impairment in the MV-LPS group, because the systemic arterial pressure and heart rate remained in normal range. However, the occurrence of a lipopolysaccharide-induced hyperkinesia could not be fully ruled out in the absence of intrapulmonary shunt measurement. The increase in Paco²in MV-LPS animals suggests an increase in physiologic dead space as a marker of lung physiologic alteration.

The exact cause of the significant differences between the spontaneously breathing and the mechanically ventilated animals in the response to lipopolysaccharide is not clear from this study. In mechanically ventilated animals, V^Twas nearly twice that of spontaneously breathing animals (1.7-fold). Compared with negative-pressure spontaneous breathing, PPMV causes differences in the pattern of applied forces across the lungs. As another possible mechanism, there may also have been an increase in capillary transmural pressure, either because of changes in intrathoracic pressure on the pulmonary vasculature or because of the interdependence phenomenon.50Muscle paralysis may also have a role. In a similar rabbit model of MV in ZEEP, avoiding muscle paralysis resulted in fewer lung macroscopic abnormalities and decreased inflammatory response (including cytokines) while the V^Twas unchanged.51Therefore, we asked whether the suppression of diaphragmatic activity could have played a role in the enhancement of lung injury in mechanically ventilated endotoxemic animals. The regional distribution of ventilation, especially in prediaphragmatic locations, may differ between mechanically ventilated paralyzed animals as compared with spontaneously breathing animals. This might favor a harmful effect of MV. In addition, it is not known whether cellular hypoxia develops in microatelectatic areas (that are known to develop during MV in anesthetized and paralyzed subjects) and acts as a trigger for inflammatory responses. The role of the absence of positive end-expiratory pressure in our model should also be questioned because, by decreasing microatelectatic areas, positive end-expiratory pressure may minimize the injury, possibly because of atelectotrauma (biotrauma due to opening and closing unstable alveoli).9,52Major differences in hemodynamic, anesthetic/muscle paralysis management and ventilatory strategy between the current study and the study of Altemeier et al.  10may have influenced the above-cited mechanisms (even hypothetical) and might explain some differences in lung physiologic response between the two studies.

The experimental design comparing ventilated with spontaneously breathing animals resulted in baseline differences between groups in the airway pressure regimen and several physiologic parameters. For example, the spontaneous breathing animals developed marked hypocapnia, which could influence the lung injury, the cytokine response, or both. As demonstrated by some authors, hypocapnia could accentuate both the lung injury and the inflammatory response.13However, our spontaneously breathing animals had fewer lung alterations than the ventilated animals, so hypocapnia cannot explain the differences that we observed between the experimental groups.

In sum, this study shows that MV and circulating lipopolysaccharide have important interactions that could be relevant during and after surgery. MV with routinely used V^Tand ZEEP enhanced responses to circulating lipopolysaccharide. The degree of hypoxemia correlated with the alteration of the pulmonary mechanics and with increased alveolar neutrophils, IL-8 and GRO-α, suggesting a possible role for the inflammatory response in lung functional abnormalities. These results show that further information is needed about more effective ventilatory strategies that could be used in healthy patients undergoing major surgery, particularly those at risk for intraoperative or postoperative endotoxemia or sepsis.

The authors thank Mrs. Sylvie Raveihle (Animal Technician) and Mrs. Nathalie Kipson (Laboratory Technician, Laboratoire de Physiopathologie Respiratoire EA 2201, Institut de Formation et de Recherche Jean Roche, Faculté de Médecine, Université de la Méditerranée Marseille, France). They also thank Marie-Josée Payan, M.D., D.Sc. (Assistant Professor, Laboratory of Histopathology, Centre Hospitalier Universitaire Timone, Marseille, France), for her contribution to lung histologic examination.